Asymmetric cell division generates daughter cells with different developmental fates. In Drosophila neuroblasts, asymmetric divisions are characterized by (1) a difference in size between the two daughter cells and (2) an asymmetric distribution of cell fate determinants, including Prospero and Numb, between the two daughter cells. In embryonic neuroblasts, the asymmetric localization of cell fate determinants is under the control of the protein Inscuteable (Insc), which is itself localized asymmetrically as an apical crescent. Rapsynoid (Raps), which interacts in a two-hybrid assay with the signal transduction protein Galphai, is localized asymmetrically in dividing larval neuroblasts and colocalizes with Insc. Moreover, in raps mutants, the asymmetric divisions of neuroblasts are altered: (1) Insc is no longer asymmetrically localized in the dividing neuroblast; and (2) the neuroblast division produces two daughter cells of similar sizes. However, the morphologically symmetrical divisions of raps neuroblasts still lead to daughter cells with different fates, as shown by differences in gene expression. The data show that Raps is a novel protein involved in the control of asymmetric divisions of neuroblasts (Parmentier, 2000).

Because Insc is involved in the control of asymmetric division of neuroblasts, the colocalization of Raps with Insc suggests a possible role of Raps in the same process. To analyze Raps function in neuroblasts, flies mutant for rapsynoid were obtained by imprecise excision of a P-element, l(3)S031807, adjacent to the rapsynoid gene. The P-element is situated in an intron of another gene proximal to rapsynoid, oriented in the opposite direction. Flies homozygous for any of the rapsynoid deletions die as young pupae and, based on their phenotype over Df(3R)IR16 that deletes the whole raps gene, behave genetically as strong hypomorphs for the phenotypes studied (Parmentier, 2000).

Because colocalization of Raps with Insc was observed in wild type, disruption of Insc localization was examined in raps mutants. The Insc crescent fails to form in the mutant metaphase neuroblasts, and only a punctate staining similar to that seen in interphase is visible. This phenotype is rescued in all raps mutants when a rapsynoid transgene is added under the control of the hsp70 promoter. The basal expression, at 25°C, of two copies of the transgene is sufficient for a complete rescue of Insc asymmetrical localization. It is thus concluded that Raps is necessary for the asymmetrical localization of Insc in neuroblasts (Parmentier, 2000).

The effect of a raps mutation was studied on the localization of Miranda, whose localization on the GMC side of the neuroblast during division is dependent on Insc function. As expected, Miranda is asymmetrically localized in wild-type mitotic neuroblasts, but is no longer asymmetrically localized in raps mutant neuroblasts (Parmentier, 2000).

An important aspect of neuroblast asymmetric divisions is their morphology. In wild-type third-instar larvae, at day 5 after egglaying, all neuroblast divisions are morphologically asymmetric, producing a GMC that is one-eighth the size of the neuroblast. This is not the case in raps mutants, where 28% of neuroblasts divisions are morphologically symmetrical. Correlatively, the neuroblasts of third-instar larvae are smaller than in wild type, and their size approaches that of GMCs. These phenotypes are rescued in the presence of a rapsynoid transgene (Parmentier, 2000).

Molecular markers have been used to determine whether the loss of morphological asymmetry in dividing raps neuroblasts reflects a defect in cell fate determination during these divisions. A neuroblast marker (deadpan-LacZ) was used that is expressed at a high level in neuroblasts and at a low level in GMCs, and a GMC marker, Prospero, was used which is expressed in the nuclei of GMCs but not in the nuclei of neuroblasts. In raps mutants, a neuroblast-like cell expresses the neuroblast marker and no Prospero (or expresses Prospero very weakly), whereas the other daughter cell expresses Prospero strongly. Although the division is morphologically symmetrical, and although Insc is not apically localized, there is still some asymmetry in the division to lead to the differentiation of two dissimilar daughter cells (Parmentier, 2000).

Drosophila neuroblast asymmetric divisions generate two daughters of unequal size and fate. A complex of apically localized molecules mediates basal localization of cell fate determinants and apicobasal orientation of the mitotic spindle, but how daughter cell size is controlled has remained unclear. Mitotic spindle geometry and unequal daughter cell size were shown to be controlled by two parallel pathways (Bazooka/DaPKC and Pins/Galphai) within the apical complex. While the localized activity of either pathway alone is sufficient to mediate the generation of an asymmetric mitotic spindle and unequally sized neuroblast daughters, loss of both pathways results in symmetric divisions. In sensory organ precursors, Bazooka/DaPKC and Pins/Galphai localize to opposite sides of the cortex and function in opposition to generate a symmetric spindle (Cai, 2003).

Thus members of the NB apical protein complex control the generation of daughter cells of unequal size. There are two redundant pathways: (1) Baz/DaPKC/ (and presumably DmPar6) as well as (2) Pins/Gαi, either of which, when asymmetrically localized to the NB cortex, can lead to the formation of an asymmetric mitotic spindle through the preferential elongation of the proximal spindle arm and the displacement of the spindle toward the distal cell cortex, resulting in the production of unequal-sized daughter cells. In addition, in NBs, Insc is required for the function of the Baz/DaPKC/(DmPar6) pathway. When both pathways are inactivated/attenuated, spindle asymmetry and displacement fail to occur and equal-sized daughter cells are produced at high frequency. In the PNS progenitor, pI, where Baz/DaPKC are localized to the posterior cortex and Pins/Gαi are localized to the anterior cortex, the mitotic spindle is symmetric. Consistent with this hypothesis that both pathways can act to cause the preferential elongation of the proximal spindle arm relative to the distal spindle arm, removing posterior baz function without abolishing the localization and function of the anterior components results in the production of an asymmetric spindle with an anterior bias; removing anterior pins function without affecting the function of the posterior components results in a posteriorly biased asymmetric spindle; if components of both pathways are localized to the anterior cortex through the ectopic expression of Insc, an anteriorly biased asymmetric spindle results. These findings suggest that DaPKC and hetrotrimeric G protein signaling work in conjunction in the NB to produce an asymmetric spindle and in opposition in pI to produce a symmetric spindle (Cai, 2003).

Several lines of evidence suggest that localized signaling is essential to generate an asymmetric spindle and daughter cells of unequal size. (1) When both signaling pathways are abolished/attenuated (e.g., in insc/pins double mutant) or when signaling is uniform, which is assumed to be the case when Baz/DaPKC/Pins/Gαi are all uniformly localized throughout the cell cortex (e.g., in the case of Gαi overexpression in wt NBs), equal-sized daughters are generated. (2) When pins function is removed and DaPKC/Baz is asymmetrically localized (e.g., in pins mutant NB) or when Pins/Gαi are uniformly cortical but DaPKC/Baz are asymmetrically localized (e.g., in 69% [n = 51] of wt NBs overexpressing C-Pins), the site of the DaPKC/Baz localization coincides with the position where the larger daughter forms. (3) When Pins/Gαi is asymmetrically localized but baz/DaPKC function has been compromised (e.g., in insc mutant) or when Pins/Gαi is asymmetrically localized but Baz/DaPKC is uniformly cortical (in the case of NBs with basal Pins-C-Pon crescents), the site of localization coincides with the larger daughter and the extended spindle arm. These observations indicate that just one localized signal source, mediated presumably by either heterotrimeric G protein or DaPKC, is sufficient to cause proximal spindle arm elongation and the generation of unequal-sized daughters (Cai, 2003).

The situation is different in pI where Baz/DaPKC/(DmPar6) and Pins/Gαi act in opposition and where Insc is not required for the function of the Baz/DaPKC/(DmPar6) with respect to spindle elongation. Here, a distinction can be made between two possible models for explaining how spindle asymmetry/geometry is mediated. The first model is that the presence of either asymmetrically localized Baz/DaPKC/(DmPar6) or Pins/Gαi on one side of the cell is sufficient to cause elongation of the proximal spindle arm, regardless of what occurs on the other side of the cell. A second model would be that the signals from the opposite sides of the cortex are integrated and the bias in the spindle geometry depends on the relative magnitude of the two signals. The simplest prediction of the first model would be that the distance from the cleavage furrow to the spindle pole of wt telophase pI should be equivalent to the longer of the two spindle arms in telophase pI mutant for either baz or pins. This appears not to be the case. The average length of the longer spindle arm in telophase pI mutant for pins or baz is greater than that of a wt spindle arm and the length of the shorter of the spindle arms in mutant pI is less than that of a wt spindle arm . An equivalent analysis is difficult to do with NBs, since the size of the 30 or so NBs found in each hemisegment is more variable. Nevertheless, based on these observations the second type of model is favored (Cai, 2003).

Previous work has shown that Pins binds to the GDP bound form of Gαi and can cause Gαi to dissociate from Gβ13F; moreover, some phenotypes seen when Gαi is overexpressed in wt NBs (e.g., equal size divisions) are not seen when GαiQ205L, an activated form of Gαi lacking GTPase activity that should be in the GTP bound form, is overexpressed, or when Gβ13F function is abolished. These phenotypes therefore are unlikely to be induced by GTP bound Gαi or by depletion of Gβγ, suggesting that the GDP bound form of Gαi may be responsible for the equal size NB divisions seen when wt Gαi is overexpressed. These findings clearly support the view that the Pins/GDP-Gαi complex has a role for generating the signal associated with spindle asymmetry. (1) Equal size divisions seen when Gαi is overexpressed in wt NBs is drastically reduced when overexpression is performed in the absence of Pins. (2) Whenever unequal size division occurs when Baz/DaPKC function is compromised, Pins and Gαi are always colocalized to the side of the cell where the future larger daughter is formed (Cai, 2003).

Although in the nematode embryo generation of unequal-sized daughters involves only the posterior displacement of a symmetric spindle, there appears to be some parallels between the two model systems. In the wt nematode P₀ division, the magnitude of the forces acting on the two spindle poles apparently depend on the character of the anterior and posterior cortex. In wt P₀, PAR-3 and PAR-2 localize to the anterior and posterior cortex, respectively, and the mitotic spindle is displaced toward the posterior pole, correlating with a greater net posterior force acting on the posterior spindle pole relative to the net anterior force acting on the anterior spindle pole. In par-2 mutants, PAR-3 expands to occupy the whole of the cortex, imparting anterior character throughout, and the net force acting on both spindle poles has a magnitude equivalent to that of the wt force acting on the anterior spindle pole. Conversely in par-3 P₀, PAR-2 becomes cortical, imparting posterior character to the entire cortex, and the magnitude of both forces acting on the spindle poles is equivalent to that of the wt posterior acting force. In both par-2 and par-3 mutants, the forces acting on the spindle poles are equalized, mitotic spindle is no longer displaced, and equal-sized daughters result (Cai, 2003).

In Drosophila NBs, although spindle displacement occurs, the generation of an apically biased mitotic spindle mediated by either asymmetrically localized Baz/DaPKC or Pins/Gαi makes the major contribution to the difference in daughter cell size. It is proposed that the asymmetric localization of components of either of these pathways can make the region of the cell cortex they occupy different from the cortical regions that they don't occupy through localized DaPKC or heterotrimeric G protein signaling mediated through Pins/Gαi. In wt NBs, the components of either pathway would impart apical character to the cell cortex where they are localized. One effect of the asymmetric signaling is to generate the preferential elongation of the spindle arm closest to the site of the localized signal. If signaling is symmetric, for example either when Baz/DaPKC and Pins/Gαi are all uniformly cortical, or when Baz/DaPKC and Pins/Gαi are localized to opposite sides of a dividing progenitor, as in pI, a symmetric spindle results. Hence, in both the nematode P₀ and in Drosophila NBs the generation of unequal-sized daughters is regulated by asymmetrically localized cortical components. In the nematode there is compelling evidence that differential forces acting on the two spindle poles mediate spindle displacement and the generation of unequal daughters. However, NBs of Drosophila asterless mutants are apparently devoid of functional centrosomes and astral microtubules, yet they form functional asymmetric anastral mitotic spindles and undergo unequal cytokinesis to generate unequal size daughters. It remains to be seen how the localized properties of the NB cell cortex influences its spindle geometry (Cai, 2003).

How a cell chooses to proliferate or to differentiate is an important issue in stem cell and cancer biology. Drosophila neuroblasts undergo self-renewal with every cell division, producing another neuroblast and a differentiating daughter cell, but the mechanisms controlling the self-renewal/differentiation decision are poorly understood. This study tested whether cell polarity genes, known to regulate embryonic neuroblast asymmetric cell division, also regulate neuroblast self-renewal. Clonal analysis in larval brains shows that pins mutant neuroblasts rapidly fail to self-renew, whereas lethal giant larvae (lgl) mutant neuroblasts generate multiple neuroblasts. Notably, lgl pins double mutant neuroblasts all divide symmetrically to self-renew, filling the brain with neuroblasts at the expense of neurons. The lgl pins neuroblasts show ectopic cortical localization of atypical protein kinase C (aPKC), and a decrease in aPKC expression reduces neuroblast numbers, suggesting that aPKC promotes neuroblast self-renewal. In support of this hypothesis, neuroblast-specific overexpression of membrane-targeted aPKC, but not a kinase-dead version, induces ectopic neuroblast self-renewal. It is concluded that cortical aPKC kinase activity is a potent inducer of neuroblast self-renewal (Lee, 2005).

Drosophila neuroblasts are an excellent model system in which to investigate the molecular control of self-renewal versus differentiation. Larval neuroblasts repeatedly divide asymmetrically to self-renew a neuroblast and to produce a smaller daughter cell, called a ganglion mother cell (GMC), that typically makes two postmitotic neurons; this process enables a single neuroblast to generate many hundreds of neurons. Self-renewal is defined as the capacity of a neuroblast to maintain all attributes of its cell type (molecular markers and proliferation potential). In this regard, a neuroblast is very similar to a germline stem cell: both maintain their stem cell identity while generating differentiating progeny. About 100 neuroblasts per brain lobe are formed during embryogenesis, where they proliferate briefly before entering quiescence. Brain neuroblasts re-enter the cell cycle between 10 and 72 h after larval hatching (ALH), and then a stable population of ~100 mitotic, self-renewing neuroblasts is maintained. This invariant neuroblast number was used to screen for mutants altering self-renewal versus differentiation: mutants in which a neuroblast makes two neuroblast progeny (ectopic self-renewal) will have >100 neuroblasts, whereas mutants in which a neuroblast makes two GMC progeny (failure in self-renewal) will have <100 neuroblasts. This assay was used to test known cell polarity mutants for a role in neuroblast self-renewal (Lee, 2005).

Two classes of cell polarity regulators were assayed for an effect on larval neuroblast self-renewal. lgl and discs large (dlg) zygotic mutants were examined, because these mutants form brain tumours and promote basal protein targeting in embryonic and larval neuroblasts. Lgl and Dlg have several protein interaction motifs and are localized around the neuroblast cortex. In addition, pins and Galphai zygotic mutants were examined; these genes regulate cell polarity in embryonic neuroblasts, but have not been well characterized in larval neuroblasts. Pins and Galphai are colocalized with Inscuteable and the evolutionarily conserved Bazooka-Par6- aPKC proteins at the apical cortex of mitotic neuroblasts, and all of these proteins are partitioned into the neuroblast during cytokinesis (Lee, 2005).

In wild-type larvae, a population of ~100 neuroblasts could be identified by the markers Worniu, Deadpan and Miranda, and by labelling with a pulse of 5-bromodeoxyuridine (BrdU); by contrast, the thousands of differentiating GMCs and neurons rapidly downregulate neuroblast markers and express nuclear Prospero and/or Elav. A clear increase in neuroblast number is observed in lgl and dlg mutants; there are supernumerary neuroblasts at all stages examined; all extra neuroblasts expressed Deadpan and Miranda and are proliferative on the basis of their ability to incorporate BrdU. Galphai zygotic mutants have a complex phenotype that will be described in a later publication; however, pins zygotic mutants show a marked decrease in neuroblast number. Notably, this phenotype is not due to a subset of neuroblasts remaining quiescent, because neuroblast numbers peak and then decline over time, and it is not due to neuroblast cell death. The relatively late onset of the pins phenotype is probably due to the gradual depletion of maternal pins gene product in these larvae (Lee, 2005).

To determine whether the pins and lgl larval brain phenotypes are due to defects in neuroblast self-renewal, positively marked genetic clones were induced in single neuroblasts to trace their progeny. Clone induction parameters were adjusted to ensure that each clone was derived from a single neuroblast (1.2 clones per lobe). In wild-type brains, neuroblast clones always contained a single Worniu⁺ Miranda⁺ nuclear-Prospero^- neuroblast and numerous smaller Worniu^- Miranda^- nuclear-Prospero⁺ progeny, confirming that wild-type neuroblasts always divide to self-renew and to generate a smaller differentiating GMC. By contrast, lgl mutant brains had an average of 2.3 neuroblasts per clone, with up to six neuroblasts per clone, showing that lgl mutant neuroblasts can divide symmetrically to yield two neuroblasts. The opposite phenotype was seen in pins mutant brains: 72.8% of the clones had no neuroblast and the remainder had a single neuroblast. The neuroblasts did not die in the pins mutants as evidenced by the following: the cell death marker caspase-3 was not upregulated, neuroblast-specific expression of the p35 cell death inhibitor did not rescue the missing neuroblasts, and one clone was observed in which the largest cell coexpressed neuroblast and GMC markers, consistent with an intermediate stage in neuroblast-to-GMC differentiation. It is concluded that wild-type neuroblasts exclusively generate neuroblast/GMC siblings; lgl mutant neuroblasts occasionally undergo ectopic self-renewal to generate neuroblast/neuroblast siblings; and pins mutant neuroblasts occasionally fail to self-renew, resulting in GMC/GMC siblings and termination of the lineage (Lee, 2005).

Next to be examined was whether lgl pins double mutants had fewer neuroblasts (like pins mutants) or extra neuroblasts (like lgl mutants). Unexpectedly, a phenotype was detected in which the larval brain was full of cells expressing the neuroblast markers Worniu, Miranda and Deadpan and lacking expression of the neuronal marker Elav. Additional markers that distinguish neuroblasts and GMCs were examined to determine whether these cells were neuroblasts or a hybrid neuroblast/GMC identity. Both wild-type neuroblasts and lgl pins cells actively transcribed the worniu, deadpan, miranda and prospero genes, maintained proliferation, did not express the Elav neuronal differentiation marker, and did not extend axons. The only potential GMC attribute found in lgl pins neuroblasts was nuclear Prospero protein but, because wild-type neuroblasts and GMCs both contain Prospero protein, which can accumulate in neuroblast nuclei if not properly localized, this protein is not a definitive marker for the GMC cell type. Thus, lgl pins brains contain large numbers of ectopic, proliferating, self-renewing neuroblasts. Combining these lgl, pins and lgl pins mutant data leads to the conclusion that Lgl inhibits self-renewal, whereas Pins has dual functions in promoting and inhibiting self-renewal (Lee, 2005).

To understand how Lgl and Pins regulate neuroblast self-renewal at the cellular level, cortical polarity marker localization was examined in mitotic larval neuroblasts. In wild-type larval neuroblasts, the Par complex (Bazooka-Par6-aPKC) and Pins-Galphai proteins forms an apical crescent at metaphase and are partitioned into the self-renewing neuroblast at telophase, whereas the Miranda and Prospero proteins form a basal crescent at metaphase and are partitioned into the differentiating GMC at telophase. In lgl pins double mutants, in which all neuroblasts divide symmetrically to generate self-renewing neuroblast/neuroblast siblings, most mitotic neuroblasts show uniform cortical aPKC, cytoplasmic Bazooka and Par6, and uniform cortical Miranda at metaphase and telophase. Thus, only aPKC maintained its correct subcellular localization and correlated with neuroblast self-renewal (Lee, 2005).

aPKC localization was examined in lgl and pins single mutants, in which symmetric divisions occurred at lower frequency. In lgl mutants, aPKC showed weak ectopic cortical localization in about half the metaphase neuroblasts, whereas Miranda was delocalized from the cortex; by telophase, however, both proteins appeared to be localized normally. Ectopic cortical aPKC was also observed in dlg mutant larval neuroblasts. A role for Lgl in restricting aPKC to the apical cortex of neuroblasts has not been reported but would be consistent with the observation that basolateral Lgl restricts aPKC to the apical surface of Drosophila and vertebrate epithelia and Xenopus blastomeres. In pins mutants, aPKC and cytoplasmic Miranda showed weak uniform cortical distribution in metaphase neuroblasts, but were properly localized in most telophase neuroblasts Thus, both Lgl and Pins are required to restrict aPKC to the apical cortex in metaphase neuroblasts (Lee, 2005).

Whether aPKC is required for neuroblast self-renewal was examined. aPKC mutant clones in larval mushroom body neuroblasts showed premature lineage termination, consistent with aPKC being required for neuroblast self-renewal. In addition, aPKC null mutants died as second instar larvae with reduced neuroblast numbers. Because this was a relatively mild phenotype and there was no detectable aPKC protein at this stage, it is likely that there are additional pathways for stimulating neuroblast self-renewal. Next, whether aPKC is required for ectopic neuroblast self-renewal in the lgl mutants was tested. lgl aPKC double mutants had normal numbers of neuroblasts, showing that aPKC is required for the ectopic neuroblast self-renewal seen in lgl mutants. aPKC mutants also suppressed ectopic neuroblast self-renewal in several independently isolated lgl mutations, further supporting a role for aPKC in self-renewal. In addition, it was found that aPKC is fully epistatic to lgl in regulating Miranda localization. Thus, aPKC is required for the ectopic neuroblast self-renewal and Miranda delocalization phenotypes seen in lgl mutants (Lee, 2005).

These data are most consistent with a model in which Lgl negatively regulates aPKC, and aPKC directly promotes self-renewal. This model is based on the observations that Lgl restricts aPKC localization to the apical cortex of neuroblasts and that a reduction in aPKC blocks the lgl self-renewal phenotype. To test this model, worniu-Gal4 line was used to drive neuroblast-specific expression of constitutively active aPKC or Lgl proteins, and an increase or decrease in neuroblast numbers was assayed. Neuroblast-specific expression of aPKC targeted to the plasma membrane with a CAAX prenylation motif (UAS-aPKC^CAAXWT) resulted in ectopic cortical aPKC localization, loss of cortical Miranda, and a large increase in the number of neuroblasts. These effects were not observed after overexpression of wild-type aPKC or a membrane-targeted kinase-dead aPKC (UAS-aPKC^CAAXKD). Expression of a constitutively active aPKC (UAS-aPKC^DeltaN) that was predominantly cytoplasmic gave only a slight increase in neuroblast number, showing that cortical localization of aPKC is essential to generate ectopic neuroblasts. By contrast, neuroblast-specific expression of a constitutively active Lgl protein (Lgl3A) resulted in the expected uniform cortical localization of Miranda, but no change in neuroblast numbers. Combined overexpression of both Lgl3A and aPKC^CAAXWT, however, resulted in strong suppression of the aPKC^CAAXWT ectopic neuroblast phenotype, even though Lgl3A alone had no effect on neuroblast numbers, consistent with Lgl inhibiting aPKC function either directly or through its downstream effectors. Thus, neuroblast-specific overexpression of aPKC can expand the neuroblast population (most probably by promoting symmetric neuroblast/neuroblast cell divisions) without eliminating the ability of these neuroblasts to undergo asymmetric neuroblast/GMC divisions to generate differentiating progeny. It is concluded that aPKC is sufficient to promote neuroblast self-renewal, Lgl can inhibit aPKC function, and membrane-targeting and kinase activity are essential for aPKC function (Lee, 2005).

This study has established Drosophila larval neuroblasts as a model system for studying self-renewal versus differentiation. A simple model is proposed in which Pins anchors aPKC apically and Lgl inhibits aPKC localization basally, thereby restricting aPKC to the apical cortex where it promotes neuroblast self-renewal. In addition, aPKC can phosphorylate and directly inhibit Lgl function, which together with the current data provides evidence for mutual inhibition between Lgl and aPKC in neuroblasts, similar to the mutual inhibition seen between these two proteins in epithelia. Mutual inhibition between aPKC and Lgl would result in stabilization of apical aPKC localization and more reliable partitioning of aPKC into the neuroblast during mitosis. In pins mutants, aPKC is delocalized and nonfunctional owing to Lgl activity, thereby reducing self-renewal; in lgl mutants, aPKC shows weak ectopic cortical localization that increases self-renewal, and in lgl pins double mutants, aPKC is both delocalized and fully active: thus all neuroblasts undergo symmetric self-renewal. Although the targets of aPKC involved in self-renewal are unknown, aPKC may directly phosphorylate and inactivate GMC determinants, and/or phosphorylate and activate neuroblast-specific proteins. Notably, lgl1 mutant mice have neural progenitor hypertrophy and knockdown of a pins mammalian homologue (AGS3) leads to depletion of neural progenitors: phenotypes that are very similar to those described in this study. In the future, it will be important to determine the role of aPKC in mammalian neural progenitor self-renewal and to identify the aPKC-regulated phosphoproteins that regulate neuroblast self-renewal in Drosophila (Lee, 2005).