Nutrition-Responsive Glia Control Exit of Neural Stem Cells from Quiescence

Summary The systemic regulation of stem cells ensures that they meet the needs of the organism during growth and in response to injury. A key point of regulation is the decision between quiescence and proliferation. During development, Drosophila neural stem cells (neuroblasts) transit through a period of quiescence separating distinct embryonic and postembryonic phases of proliferation. It is known that neuroblasts exit quiescence via a hitherto unknown pathway in response to a nutrition-dependent signal from the fat body. We have identified a population of glial cells that produce insulin/IGF-like peptides in response to nutrition, and we show that the insulin/IGF receptor pathway is necessary for neuroblasts to exit quiescence. The forced expression of insulin/IGF-like peptides in glia, or activation of PI3K/Akt signaling in neuroblasts, can drive neuroblast growth and proliferation in the absence of dietary protein and thus uncouple neuroblasts from systemic control.


INTRODUCTION
The stem cell populations found in tissues as varied as blood, gut, and brain spend much of their time in a mitotically dormant, quiescent state (for reviews, see Ma et al., 2009;Moore and Lemischka, 2006;Woodward et al., 2005;Zammit, 2008). Cellular quiescence, or G 0 , is the reversible arrest of growth and proliferation and is actively maintained by a distinct transcriptional program (Coller et al., 2006). The balance between quiescence and proliferation, as well as the rate and duration of proliferation, can have significant effects on the growth, maintenance, and repair of tissues. When ''choosing'' whether or not to exit the quiescent state and divide, stem cells integrate a variety of local and systemic signals (reviewed in Drummond- Barbosa, 2008; Morrison and Spradling, 2008). In the mammalian brain, the neural stem cells (NSCs) in the subventricular zone (SVZ) and hippocampal subgranular zone (SGZ) transition between quiescence and proliferation, generating new neurons throughout the life of the animal (Ahn and Joyner, 2005;Doetsch et al., 1999;Ma et al., 2009;Morshead et al., 1994). A number of factors have been shown to have mitogenic effects on NSCs; however, it is not clear upon which cells (stem cells or their proliferative progeny) and at what point in the cell cycle these factors act (Zhao et al., 2008).
Drosophila neural stem cells (neuroblasts) in the central brain and thoracic ventral nerve cord (tVNC) are quiescent for 24 hours between their embryonic and larval phases of proliferation (Hartenstein et al., 1987;Ito and Hotta, 1992;Prokop and Technau, 1991;Truman and Bate, 1988). Quiescent neuroblasts are easily identifiable and are amenable to genetic manipulation, making them a potentially powerful model with which to study the transition between quiescence and proliferation. However, the mechanisms regulating the exit from quiescence, either intrinsic or extrinsic, are not well established. Genetic studies found that Drosophila FGF, in concert with Drosophila Perlecan, promotes the neuroblast transition from quiescence to proliferation (Park et al., 2003), but subsequent work revealed that this effect is indirect (Barrett et al., 2008). Britton and Edgar found that the exit from quiescence is physiologically coupled to larval growth and development via a nutritional stimulus (Britton and Edgar, 1998). The Drosophila fat body performs many of the storage and endocrine functions of the vertebrate liver and acts as a sensor, coupling nutritional state to organismal growth (Colombani et al., 2003). In response to dietary amino acids, the fat body secretes a mitogen that acts on the CNS to bring about neuroblast proliferation (Britton and Edgar, 1998). This fat body-derived mitogen (FBDM) initiates cell growth in quiescent neuroblasts and promotes (or at least permits) cell-cycle re-entry (Britton and Edgar, 1998). Yet the identity of the FBDM, the cell type upon which it acts, and the downstream pathway activated in neuroblasts are unknown.
Insulin and insulin-like growth factor (IGF) signaling are powerful regulators of growth and metabolism. In mammals, IGF-I has been shown to drive the proliferation of neural stem cells in both the embryo and adult (reviewed in Anderson et al., 2002;Joseph D'Ercole and Ye, 2008). IGF-I expression is induced in astrocytes (astroglia) in response to a variety of CNS injuries (Yan et al., 2006;Ye et al., 2004) and is thought to be responsible for the increased neural stem cell proliferation seen in the SVZ and SGZ following cortical ischemia (Yan et al., 2006).
In Drosophila, there are seven insulin/IGF-like peptides (dILPs 1-7) and a single insulin/IGF receptor (dInR). dInR activates the highly conserved PI3K/Akt pathway, leading to cellular growth and proliferation (reviewed in Goberdhan and Wilson, 2003). dILPs expressed by the IPC (insulin-producing cell) neurons of the brain are secreted into the circulation, where their endocrine functions include the regulation of growth, carbohydrate metabolism, and germline stem cell division LaFever and Drummond-Barbosa, 2005;Rulifson et al., 2002). dInR is strongly enriched in the developing CNS and its resident neuroblasts (Fernandez et al., 1995;Garofalo and Rosen, 1988), but a role for the insulin/IGF pathway in neuroblast proliferation has not been found.
We show here that the nutritional stimulus (known to be transduced by the fat body [Britton and Edgar, 1998]) induces the expression of dILPs in a subset of glia that neighbors neuroblasts and that the InR/PI3K pathway is required by neuroblasts for the exit from quiescence. Indeed, the forced expression of dILPs in glia, or activation of PI3K/Akt signaling in neuroblasts, can drive neuroblast proliferation in the absence of dietary protein, uncoupling the quiescence and proliferation of neuroblasts from systemic nutritional control. Thus, we identify a paracrine function of dILPs as mediators of the systemic regulation of neuroblast proliferation.

Neuroblast Reactivation and Nutritional Dependence
During embryogenesis, neuroblasts proliferate to generate the neurons that will form the larval CNS. Following the embryonic phase of proliferation, neuroblasts either enter into quiescence or undergo apoptosis. Quiescent neuroblasts reactivate and resume proliferation during larval stages, generating neurons that will contribute to the adult CNS (reviewed in Egger et al., 2008).
At the beginning of the first larval instar, the cell body diameter of quiescent neuroblasts is 3-4 mm, similar to surrounding neurons. Shortly thereafter, neuroblasts begin to enlarge, and by 24 hph, the average diameter is 7 mm (compare Figures  1A and 1B). It is at this time that the first neuroblast divisions are seen (Truman and Bate, 1988 and data not shown). Neuroblasts reactivate asynchronously, but by the end of the second larval instar, all neuroblasts have fully enlarged and begun to proliferate (Truman and Bate, 1988; Figure 1C). Interestingly, the exit from quiescence of neural stem cells from the developing mammalian cortex has also been shown to coincide with an increase in cell size (Alam et al., 2004;Groszer et al., 2006). Quiescent neuroblasts, like quiescent neural stem cells of the mammalian SVZ and SGZ, exhibit a more complex morphology than proliferating cells ( Figure 1B 0 ) (Ma et al., 2009). Quiescent neuroblasts extend a primary cellular process toward the neuropil and also occasionally extend a process toward the ventral surface or toward other neuroblasts (Truman and Bate, 1988;Figures 1A-1B 0 ). These processes are present until neuroblasts begin to divide (Tsuji et al., 2008), but their function has not yet been investigated. In larvae, growth and cell proliferation are triggered by feeding (Britton and Edgar, 1998). In larvae reared on a sucrose-only (amino acid-deprived) diet, neuroblast reactivation never occurs. Neuroblasts display no cellular growth (a prerequisite for neuroblast cell cycle re-entry) and maintain their primary process (Britton and Edgar, 1998;Figures 1D-1F).
Stellate Surface Glia Express dILP6 and dILP2 during Reactivation A transcriptome analysis comparing VNCs from newly hatched larvae and VNCs from larvae at the end of the first instar suggested that the expression of dILP6 and dILP2 increases in the VNC during neuroblast reactivation (J.M.C. and A.H.B., unpublished data). The seven dILPs are expressed in distinct spatiotemporal patterns during development (Brogiolo et al., 2001). dILP6 is reported to be expressed in the larval gut (Brogiolo et al., 2001) and the pupal fat body (Okamoto et al., 2009;Slaidina et al., 2009), whereas dILP2 is known to be expressed in the IPC neurons of the brain (along with dilps 1, 3, and 5) Rulifson et al., 2002).
To determine whether dILP6 is also expressed in the CNS, we generated a dilp6-GAL4 line (see Experimental Procedures). dilp6-GAL4 drives expression in a subset of the surface glia that wraps the CNS (Figures 2A-2B 0 ). Strong expression was evident by mid first instar (11 hph) and was maintained throughout neuroblast reactivation (Figures 2A-2B 0 ). We also assayed the expression of dILP2 by immunohistochemistry and found that it too was expressed in the same surface glial population (Figures 2C and 2C 0 and Figure S1). The glial cells labeled by dilp6-GAL4 are located above the neuroblasts and underneath the surrounding basement membrane (Figures 2D and 2E). They are stellate in appearance, with several processes radiating from the central cell body (Figures 2A-B 0 ). Thus, dILPs, expressed by glial cells, are ideally positioned to activate the dInR pathway in neuroblasts during reactivation.
PI3K Is Active during, and Required for, Neuroblast Reactivation dInR regulates growth and proliferation in other tissues by recruiting PI3K to the cell membrane, where it converts phosphoinositol(4,5)P2 (PIP2) to phosphoinositol(3,4,5)P3 (PIP3) . PIP3 then recruits the protein kinase Akt (among other proteins) to the membrane, leading to Akt activation and signaling Verdu et al., 1999). PI3K activity can be assayed with a pleckstrin homology (PH) domain-green fluorescent protein (GFP) fusion protein (PH-GFP) . PH-GFP is strongly recruited to the membrane when PIP3 levels are high (i.e., when PI3K is active) via the binding of its PH domain to PIP3. We observe a strong increase of membranous PH-GFP in reactivating neuroblasts (compare Figures S2A and S2A 0 with S2B and S2B 0 ), consistent with an increase in PI3K activity. We also see strong expression of S6 kinase (S6K) in reactivating neuroblasts ( Figure S3), a kinase known to promote growth downstream of insulin/PI3K signaling (Lizcano et al., 2003;Miron et al., 2003;Rintelen et al., 2001).
While dInR null mutants are embryonic lethal (Fernandez et al., 1995), PI3K null mutants survive through larval development . Null mutants of the catalytic subunit of PI3K, dp110, display normal growth until the third larval instar. In these mutant larvae, the imaginal discs are not discernible; however, the CNS was reported to appear normal . We examined dp110 mutants and found that the CNS is significantly reduced in size compared to wild-type larvae ( Figure S2). Such a reduction in CNS size is indicative of reduced neuroblast proliferation. The neuroblasts in dp110 mutants are severely reduced in size, with the majority showing no sign of postembryonic growth or division ( Figure S2 and data not shown). These results demonstrate that PI3K signaling is required in order for neuroblasts to reactivate.

Inhibition of dInR/PI3K Signaling Retards the Exit from Quiescence
The neuroblast phenotype seen in dp110 null mutant larvae could result from either an intrinsic requirement for PI3K signaling within neuroblasts or a requirement for PI3K in another cell or tissue type that affects neuroblast reactivation. In order to address whether dInR and PI3K are intrinsically required by (D-F) In larvae deprived of amino acids (sucrose-only diet), neuroblast growth and cell-cycle re-entry never occur (Britton and Edgar, 1998). Neuroblasts maintain their quiescent size and primary process. Compare (D), (E), and (F) with (A 0 ), (B 0 ), and (C 0 ), respectively. Z projections of tVNCs at indicated time points. GFP, green; Deadpan (Dpn; neuroblast nuclei, red); Discs Large (Dlg; cell cortices, blue). Scale bars, 20 mm.
neuroblasts for the exit from quiescence, we used grh-GAL4 to express negative regulators of the pathway within neuroblasts. By the end of the first larval instar (24 hph), the majority of neuroblasts in the tVNC have already enlarged significantly. The average neuroblast diameter increases from 4 mm to 7 mm (Figures 1A and 1B and Figures 3A and 3E). Expression of a dominant-negative form of the PI3K adaptor subunit (Dp60) (Weinkove et al., 1999) within neuroblasts caused a strong reduction in neuroblast growth during the first larval instar, with most neuroblasts maintaining their small quiescent size of 4 mm (Figures 3A, 3B, and 3E). In Drosophila, as in vertebrates, the tumor suppressor PTEN antagonizes PI3K by converting PIP3 to PIP2 (Goberdhan et al., 1999;Maehama and Dixon, 1998). Misexpression of dPTEN  within neuroblasts generated the same phenotype as Dp60 expression, effectively blocking growth and reactivation during the first larval instar (Figures 3A, 3C, and 3E). These two results suggest that the PIP3generating activity of PI3K is required intrinsically by neuroblasts for reactivation to occur. Finally, if dInR is responsible for activating PI3K, then blocking dInR function should phenocopy the expression of Dp60 or dPTEN. Expression of a dominant-negative form of dInR (dInR K1409A ) inhibits neuroblast reactivation in the same manner as Dp60 and dPTEN, with the majority of neuroblasts remaining 4 mm in diameter ( Figures 3A, 3D, and 3E). Neuroblasts that do not express grh-GAL4 act as an internal control, showing that neuroblast reactivation can occur as normal in these cells (see dashed boxes, Figures 3A-3D). These data support a model in which the activation of dInR in neuroblasts and the subsequent upregulation of PI3K are responsible for the exit from quiescence.  (Stork et al., 2008) glia during first-and second-larval instars. dilp6-GAL4-driving UAS-mCD8-GFP, red; glial nuclei, blue (anti-Repo). Scale bars, 15 mm. (C) Anti-dILP2 (green) in the tVNC at 24 hph shows a punctate perinuclear enrichment in surface glial cells (see pink arrows in C 0 ), consistent with secretory vesicle processing. Z projection of ventral surface glial layer. (D and E) dILP6-positive glia (dilp6-GAL4 > UAS-mCD8-GFP [green]) lie just above neuroblasts (Dpn, red) and below the basement membrane (dPerlecan, blue). Sequential sections from ventral surface of VNC (D and D 00 ) and in cross-section (E). Scale bars: D and D 00 , 10 mm; E, 1mm. See also Figure S1.

Activation of PI3K Is Sufficient for Neuroblast Reactivation
If the dInR/PI3K pathway is responsible for neuroblast reactivation in response to nutritional stimuli, then activation of the pathway in the absence of the stimulus might be expected to cause aberrant reactivation. In order to test this hypothesis, we expressed a membrane-targeted, constitutively active, version of the PI3K catalytic subunit (dp110 CAAX )  in neuroblasts of larvae that were reared on a sucrose-only diet. We found that constitutive activation of PI3K can drive neuroblast reactivation during the first larval instar, irrespective of dietary protein ( Figure 4A-4B 0 ). High levels of PI3K activity increased the rate of reactivation beyond those normally seen; at the end of the first larval instar (24 hph), we find neuroblasts that have prematurely reached their full size (10 mm or more) and have already undergone multiple rounds of cell division (as evidenced by the presence of several small GFP-retaining daughter cells; Figure 4B). Thus, PI3K signaling within neuroblasts can drive the cellular growth and proliferation that constitute the exit from quiescence. The divisions proceed with the correct asymmetric partitioning of Miranda and Prospero into the differentiating daughter cell (reviewed in Knoblich (2008) (Figures 4C-4E).  (C and D 0 ) The adaptor protein Miranda (red) is asymmetrically localized and partitioned to daughter cells of dp110 CAAX -reactivated neuroblasts (yellow arrowheads). Scale bar, 5 mm.
The activation of PI3K in this context appeared to cause reactivation in an all or nothing manner. We observed a subset of the grh-GAL4-positive neuroblasts reactivating fully. Of the 141 thoracic neuroblasts (Truman and Bate, 1988), 48 show significant grh-GAL4 expression. Of these 48 neuroblasts, 2-6 (4%-12%) reactivated, with all others remaining completely quiescent ( Figure 4B). We noticed a bias toward the reactivation of lateral neuroblasts ( Figure 4B 0 and data not shown), which may reflect differences in the levels of pathway activation or possibly an intrinsic difference in neuroblast sensitivity to PI3K activity. Normally, the lateral neuroblasts of the thoracic VNC reactivate first (Truman and Bate, 1988), which supports the idea of differential neuroblast sensitivity to dInR/PI3K signaling.

Akt Is Upregulated by PI3K in Neuroblasts and Is Sufficient for Reactivation
Drosophila Akt is a key transducer of increased PIP3 levels, such as those seen in response to dInR/PI3K activation . Following recruitment to the cell membrane, Akt is activated by PDK1-mediated phosphorylation (Cho et al., 2001;Rintelen et al., 2001). We found that, when we increased PI3K activity in neuroblasts by expression of dp110 CAAX , the levels of phosphorylated Akt (pAkt) were concomitantly increased ( Figures 4F-F 00 ). To test whether Akt activation is sufficient for the exit from quiescence, we expressed a membrane-targeted form of Akt (myr-Akt)  in neuroblasts of larvae reared on a sucrose-only diet. myr-Akt expression was sufficient to drive both growth and cell-cycle re-entry (as evidenced by extensive pH3 labeling) in quiescent neuroblasts in the absence of the nutritional stimulus ( Figures 4G-4I and Figure S4). Indeed, expression of myr-Akt was more potent than dp110 CAAX , as all grh-GAL4positive neuroblasts reactivated. The difference in the number of neuroblasts that reactivated in response to dp110 CAAX (4%-12%) and myr-Akt (100%) may reflect a differential sensitivity to negative feedback regulation in the pathway (see, for example, Kockel et al., 2010). Myr-Akt may escape negative control more readily than wild-type Akt that has been activated by dp110 CAAX .
Once neuroblast reactivation has been ectopically triggered by either PI3K or Akt, then neuroblast proliferation occurs at approximately the same rate. When we assayed reactivated neuroblasts at 24 hr, they had generated on average six or seven daughter cells under either condition. For dp110 CAAX , we counted the daughter cells of 29 reactivated neuroblasts from 10 tVNCs; on average, each neuroblast had 6.76 daughter cells. For myr-Akt, we counted the daughter cells of 40 reactivated neuroblasts from four tVNCs; on average, each neuroblast had 6.65 daughter cells. Thus, dInR/PI3K appear to act via their canonical downstream pathway, and when activated in neuroblasts, this pathway is sufficient for reactivation.

dILPs Are Required for Neuroblast Reactivation
There is significant redundancy among the dILP family of InR ligands, with no individual dILP being essential . However, two lethal dILP loss-of-function mutant combinations have recently been generated: Ddilp 2,3,5, and 6, and Ddilp 1,2,3,4,5, and 6 (Grö nke et al., 2010). We assayed neuroblast reactivation in the Ddilp 2,3,5,6 quadruple mutant. We found no sign of neuroblast reactivation in homozygous dilp 2,3,5,6 mutants at 28 hr posthatching (compare Figures 5A  and 5B). These mutants are developmentally delayed, which could explain the smaller neuroblast size. Therefore, we examined neuroblasts from third-instar mutant larvae that had undergone significant organismal growth. We found that neuroblasts were significantly reduced in size, with many neuroblasts showing no sign of reactivation ( Figure 5C). This result is consistent with an acute requirement for dILPs and the insulin/PI3K pathway for neuroblast growth and proliferation.

Glial dILP Expression Is Nutrition Dependent
Are surface glia the source of dILPs that activate dInR/PI3K signaling in neuroblasts in response to nutrition? If so, then we would expect glial dILP expression, or secretion, to be nutrition dependent. It has been demonstrated that nutrition, via the fat body, can control both the expression and secretion of dILPs in the IPC neurons of the brain (Gé minard et al., 2009;. When larvae are reared on a sucrose-only diet, there is a significant decrease in surface-glial dILP2 protein expression (compare Figures 5D and 5E). This suggests that glial dILP2 is nutritionally regulated and that this regulation occurs at the level of expression. No antibody is available for dILP6; therefore, we assayed its response to nutrition at the transcript level. We carried out a Q-PCR analysis on the ventral nerve cords from larvae at different developmental times, reared under different nutritional conditions ( Figure 5F). We found that the levels of dilp6 transcript begin to increase by 12 hph and that, by 24 hph, they have increased 8-fold over the levels seen in VNCs from just-hatched larvae (in which neuroblasts are quiescent). Furthermore, the increase in dilp6 transcription during the first-larval instar is completely abolished when larvae are deprived of amino acids and reared on a sucrose-only diet. Thus, dILP2 and dILP6 expression are both nutrition dependent.

The Glial Expression of dILPs Is Sufficient for Neuroblast Reactivation
If paracrine insulin/IGF signaling from glial cells to neuroblasts is responsible for the nutrition-dependent exit from quiescence, then the forced expression of dILPs within glia should drive neuroblast reactivation in the absence of the systemic nutritional cue. To test this hypothesis, we drove expression of dILP6  with the glial-specific driver repo-GAL4 . When these flies were reared on a sucrose-only diet as (G and G 0 ) Neuroblasts in which Akt signaling is activated by myr-Akt are quiescent at 0-1 hph. (H and H 0 ) These neuroblasts can fully reactivate during the first larval instar despite the absence of a nutritional stimulus. The yellow arrow points to a neuroblast not expressing grh-GAL4 that has failed to reactivate in the absence of the nutritional stimulus. (I and I 0 ) Neuroblasts and their progeny are seen dividing at 48 hph (pH3, white). See also Figure S4. larvae, they initiated neuroblast reactivation despite the absence of organismal growth ( Figures 6A and 6B). The enlargement of neuroblasts proceeded as normal, although the reactivated neuroblasts divided less frequently than in fed larvae, with up to four mitotic neuroblasts per VNC at each time point (Figures 6B-6D; n = 17 tVNCs). It may be that maximal pathway activation requires the simultaneous expression of another nutritionally controlled mitogen or that the glial secretion of dILP6 itself is nutritionally regulated.
It has previously been reported that high-level misexpression of dILP2 causes lethality . We found that misexpression of dILP2 using repo-GAL4 caused lethality early in the first-larval instar. We therefore employed the temperaturesensitive GAL4 inhibitor GAL80 ts  to block expression during embryogenesis. Glial dILP2 expression at larval stages also induced neuroblast reactivation in the absence of amino acids ( Figure S5). Taken together, these data support a model in which the nutritional stimulus, acting via the fat body, induces the expression and/or secretion of dILPs by surface glia. These dILPs then act on neuroblasts in a paracrine manner to bring about the growth and proliferation that constitute reactivation ( Figure 7E).

Disrupting Glial Signaling Blocks Neuroblast Reactivation
The dILPS are able to substitute for one another functionally (Broughton et al., 2008;. Consequently, we see no phenotype when we knock down either dILP2 or dILP6 expression in glia by targeted RNAi (data not shown). Furthermore, it has been reported that knockdown of dILP2 expression results in a compensatory increase in transcription of at least two other dilps (dipl3 and dilp5) (Broughton et al., 2008;. To show that glial-derived dILPs are the specific trigger for neuroblast reactivation would require the directed knockdown of at least four dILPs (2, 3, 5, and 6), and possibly more, within glia. To date, such an experiment has not proven technically feasible.
We reasoned that, if glia are the source of dILPs required for neuroblast reactivation, then blocking the ability of glia to signal should inhibit reactivation. To do this, we expressed a dominantnegative, temperature-sensitive mutant of Drosophila dynamin (shibire ts ; UAS-shi ts ) in glial cells to block vesicular trafficking. When we drove expression of shi ts with the glial-specific driver Repo GAL4, we found that neuroblast reactivation was blocked at the restrictive temperature ( Figures 7A-7D). Neuroblast growth and proliferation were both dramatically reduced. The block in growth was restricted to neuroblasts; overall regulation of growth was unaffected, and larvae exhibited normal organismal growth and progression through larval stages/instars. We conclude that signaling from the overlying glial cells is crucial for neuroblast reactivation as, importantly, neuroblasts were not reactivated by dILPs secreted from another source. This result supports our model that insulinergic glia are the key relay between nutritional state and neural stem cell reactivation and proliferation ( Figure 7E).

Figure 5. dILPs Are Required for Neuroblast Reactivation, and Their Glial Expression is Nutrition Dependent
(A-C) dilp 2,3,5,6 mutants display impaired neuroblast reactivation (compare B and C with heterozygous control A). Dpn, red; Dlg, blue. Scale bars, 20 mm. (D and E) VNCs from Oregon R larvae at 24 hph. (D) dILP2 protein expression in the surface glia of larvae fed a normal diet. (E) In larvae reared on a sucrose-only diet, dILP2 expression is greatly reduced (DILP2, green; repo, red). VNCs were dissected, stained, and imaged together. Identical reagents and microscope settings were employed. Scale bars, 25 mm. (F) Q-PCR analysis of dilp6 in the VNC. dILP6 transcript levels at 12 hr and 24 hr posthatching in VNCs of larvae fed normal or sucrose-only diets, compared to dilp6 transcript levels at 0 hr (just hatched). dilp6 levels normally increase 8-fold during the first instar (0-24 h) but are abolished when larvae are reared on a sucrose-only diet. ***p < 0.02; Student's t test. Error bars represent standard deviations. Larvae fed a normal diet showed a mean fold change in dilp6 mRNA level of 1.7 and 7.9 at 12 and 24 hr, respectively, with SD of 0.01 and 1.55, respectively. Larvae fed a sucrose-only diet showed a mean fold change in dilp6 mRNA level of 1.1 and 1.2 at 12 and 24 hr, respectively, with SD of 0.11 and 0.15, respectively.

Neuroblast Quiescence and Reactivation
Neuroblast entry into quiescence is governed intrinsically by the same transcription factor cascade that controls neuroblast temporal identity (Isshiki et al., 2001;Tsuji et al., 2008). However, the exit from quiescence and the larval reinitiation of the intrinsic temporal cascade (Maurange et al., 2008) is subject to extrinsic, humoral regulation. It has been reported that, in response to dietary amino acids, the fat body secretes a growth factor/ mitogen (FBDM) that acts on the CNS to bring about the cellular growth and cell-cycle re-entry that constitute neuroblast reactivation (Britton and Edgar, 1998). Here, we have identified a population of surface glial cells that respond to the nutrition-dependent stimulus by expressing dILPs and have shown that the dInR/PI3K pathway is required by neuroblasts to exit quiescence in response to nutrition. Forced expression of dILPs in glia or activation of PI3K/Akt signaling in neuroblasts can drive neuroblast growth and proliferation in the absence of dietary protein and thus uncouple neuroblast reactivation from systemic nutritional control.
Cell growth and division are not strictly coupled in neuroblasts. In Drosophila Perlecan (dPerlecan) loss-of-function mutants, the majority of neuroblasts appear to increase in size but then remain G1 arrested (Datta, 1995). This suggested that a dedicated mitogen might exist to promote cell-cycle progression. Drosophila Activin-like peptides (ALPs) are required for normal levels of neuroblast division in the larval brain and appear to be one such dedicated mitogen (Zhu et al., 2008). dPerlecan is expressed by glia and forms part of the basement membrane that enwraps the CNS (Friedrich et al., 2000;Lindner et al., 2007;Voigt et al., 2002). dPerlecan was proposed to modulate Drosophila FGF (Branchless (Bnl)), allowing it to act as a mitogen for neuroblasts (Park et al., 2003). However, it now appears that the action of Bnl is indirect via a still to be identified cell type (Barrett et al., 2008). One possibility is that Bnl acts on glia to modulate the expression of other proteins, such as dILPs or ALPs, which then in turn act on neuroblasts directly. Here, we show that expression of dILPs by glia leads to neuroblast reactivation in the absence of dietary protein; however, the number of mitoses falls short of that seen under normal dietary conditions. This could be explained by the absence of another nutritionally dependent mitogen. It will be of interest to see whether the glial expression of ALPs, like that of dILPs, relies on dietary protein.

Glia and Neural Stem Cell Proliferation
In the larval CNS, neuroblasts and their progeny are completely surrounded by glial cell processes. If the interaction between neuroblasts and surrounding glia is disrupted by expression of a dominant-negative form of DE-cadherin, the mitotic activity of neuroblasts is severely reduced (Dumstrei et al., 2003). In the mammalian brain, glial cells are involved in a wide variety of processes, including axon guidance, synapse formation, and neuronal specification (reviewed in Ma et al., 2005). Glial cells, with the extracellular matrix and vasculature, also make up the adult neural stem cell niche (reviewed in Nern and Momma, 2006). Astrocytes have been shown to promote neural stem cell (C and D) Quantification of neuroblast enlargement and proliferation, respectively. ***p < 0.005; Student's t test. The higher variation seen at 72 hr posthatching is due to a subset of larvae eventually showing neuroblast reactivation after a prolonged delay (40%; n = 10). (C) Box and whisker plot showing neuroblast growth is blocked by glial expression of shi ts . At 29, 48, and 72 hr, control neuroblasts have mean diameters of 8. 13,10.27,and 11.46 mm,respectively,with SD of 1.49,1.71,and 2.06,respectively. At 29,48, and 72 hr, in larvae in which dynamin function has been blocked in glia, neuroblasts have mean diameters of 4. 17,4.87,and 5.58 mm,respectively,with SD of 0.47,0.79,and 1.75,respectively. (D) Bar chart showing neuroblast proliferation is also suppressed by blocking dynamin function in glia. M phase neuroblasts were identified by the presence of pH3. Error bars represent standard deviations. At 29, 48, and 72 hr, control tVNCs have a mean number of M phase neuroblasts of 14.5, 14.6, and 13.14, respectively, with SD of 1.91, 0.71, and 3.02, respectively. At 29, 48, adn 72 hr, in larvae in which dynamin function has been blocked in glia, tVNCs have a mean number of M phase neuroblasts of 0, 0.5, and 4.1, respectively, with SD of 0, 0.58, and 5.55, respectively. (E) A model for the nutritional control of neuroblast reactivation. Previous work (Britton and Edgar, 1998) suggested that dietary amino acids are sensed by the fat body, triggering FBDM secretion into the hemolymph. The FBDM might then stimulate surface glia, which we show express and secrete dILPs in response to amino acids. These dILPs act on neuroblasts in a paracrine manner to activate the dInR/PI3K/Akt pathway, leading to cell growth and cell-cycle re-entry. dILPs, purple; active PI3K/Akt, green; asymmetrically localized cell fate determinants, red. proliferation in culture (Song et al., 2002) and can express proproliferative factors such as FGF-2 and IGF-I (Garcia-Estrada et al., 1992;Shetty et al., 2005). Thus, astrocytes are thought to be a key component of the niches that dynamically regulate neural stem cell proliferation in the adult brain (Ma et al., 2005).
We have shown that Drosophila surface glia can transduce systemic signals and, by expressing dILP2 and dILP6, control neuroblast exit from quiescence. Glial cells also express dPerlecan and ana (Ebens et al., 1993) and are the source of the Activinlike peptides that have been shown to have a direct mitogenic effect on neuroblasts (Brummel et al., 1999;Zhu et al., 2008). Thus, much like mammalian glial cells, Drosophila glial cells perform a number of the functions that define a niche and control the proliferation of neural stem cells (Morrison and Spradling, 2008).

Insulin/IGF Signaling and Neural Stem Cell Proliferation
Recent results suggest a role for IGF-1 in the control of neural stem cell division (Mairet-Coello et al., 2009). IGF-1 injection into rat embryonic brain results in a 28% increase in DNA content postnatally as a consequence of increased DNA synthesis and entry into S phase. Conversely, DNA synthesis and entry into S phase are decreased when the PI3K/Akt pathway is blocked. Furthermore, the loss of PTEN, the tumor suppressor and PI3K antagonist, enhances the exit from G 0 of neural stem cells cultured from mouse embryonic cortex (Groszer et al., 2006). The authors suggest that a concomitant increase in cell size may push the cells to enter G 1 .
Here, we show, in vivo, that glial expression of insulin-like peptides activates the dInR/PI3K/Akt pathway in Drosophila neural stem cells and is responsible for their exit from quiescence. This pathway promotes cell growth and the transition from G 0 to G 1 and is also sufficient to promote G 1 -S and mitosis. Given that IGF-1 and the PI3K/Akt pathway can promote cellcycle progression in vertebrate neural stem cells (Aberg et al., 2003;Yan et al., 2006), this same pathway may regulate vertebrate neural stem cell reactivation in the same way as we have shown here for Drosophila.

Manipulating Glia to Control Neuroblast Behavior
The identity of the proposed FBDM, secreted by the fat body in response to dietary protein, remains unknown. However, explant CNS culture experiments demonstrated that the FBDM can act directly on the CNS to bring about neuroblast reactivation (Britton and Edgar, 1998). We have identified the surface glia as a key relay in the nutritional control of neuroblast proliferation. If we can identify the receptor protein(s) that controls glial dILP expression/secretion, then we may, by extension, identify the FBDM and approach a comprehensive understanding of how neural stem cell proliferation is coupled to nutrition and organism-wide growth.
Finding treatments that stimulate the survival and proliferation of endogenous neural stem cells as potential therapies for neurodegenerative disorders is an area of active research (e.g., Androutsellis-Theotokis et al., 2008). The results reported here highlight the effectiveness of targeting support (or niche) cells in order to manipulate the behavior of stem/progenitor cells as an alternative to the direct targeting of the progenitors themselves.

EXPERIMENTAL PROCEDURES Transgenics
Generation of grh-GAL4: The ''D4'' grainyhead enhancer (4 kb from the second intron of the grainyhead gene) (gift from S. Bray) was excised from pBluescript and ligated into the pPTGAL GAL4 P element vector (Sharma et al., 2002). Generation of dilp6-GAL4: 2 kb, 18 bp ''upstream'' of the first protein-coding exon of the dilp6 gene, was amplified from genomic DNA using the PCR primers: forward, GGAATACGAGATACTCCGAAGAAA; reverse, GTTAGATTGCTTAACAACGCTCTG. The resultant PCR product was initially TOPO cloned (Invitrogen), followed by insertion into the pPTGAL GAL4 P element vector. Standard methods were subsequently used for germline transformation.
Quantitative Real-Time PCR Total RNA was extracted from 60 VNCs (brain dissected away) per sample using TRIZOL reagent (Invitrogen). cDNA was prepared using Superscript II (Invitrogen). Quantitative real-time PCR (Q-PCR) was performed using an ABI 7300 Q-PCR machine and SYBR green (QIAGEN). Results were calculated using the standard curve method and normalized against GAPDH1. Three biological replicates per sample type were generated and each subjected to three technical replicate reactions. dILP6 primers were as in Grö nke et al. (2010). GAPDH1 primers used were: forward, ATTTCGCTGAACGATAAGTTCGT; reverse, CGATGACGCGGTTGGAGTA.

Larval Culture
Embryos were placed on a fresh apple juice plate prior to larval hatching. Larvae that hatched within a 30 min window were then transferred to fresh yeast, and this was called 0 hr posthatching (hph). To deprive larvae of dietary amino acids, larvae were transferred to a solution of 20% sucrose in PBS after hatching instead of fresh yeast.

SUPPLEMENTAL INFORMATION
Supplemental Information includes Extended Experimental Procedures and five figures and can be found with this article online at doi:10.1016/j.cell. 2010.12.007. Figure S1. dILP6 and dILP2 Are Expressed in the Same Cells, Related to Figure 2 24h VNC in which dilp6-GAL4 is driving the nuclear marker histone-mRFP (in red) (under UAS control). Glial cells expressing dILP2 protein (in green) are the same cells that express dilp6 GAL4 (red nuclei). Compare yellow arrows between the separate channels and overlay. Figure S2. PI3K Is Active during, and Required for, Neuroblast Reactivation, Related to Figure 3 (A-B 0 ) PH-GFP distribution during reactivation. During neuroblast reactivation there is a strong accumulation of PH-GFP at the cell membrane (compare arrowheads in A, A' (quiescent neuroblasts) to those in B,B' (reactivating neuroblasts)), indicating increased PIP3 levels and therefore increased PI3K activity. (GFP in green, Deadpan in red, scale bars represent 20mm). (C and E) The CNS is significantly reduced in size in PI3K (dp110) loss of function mutants. CNS from wandering third instar larvae stained with Discs large and false colored (CNS in red, and eye discs in green). Thoracic VNC (tVNC) highlighted with dashed white box. CB marks the central brain. (Scale bars represent 150mm). (D and F) Neuroblasts remain quiescent in PI3K (dp110) mutants. z-projections of the tVNC. In PI3K mutants, neuroblast cell growth is significantly retarded, with many neuroblasts remaining quiescent. Compare yellow arrows in (F), with white arrowheads in (D). (Dpn in red, scale bars represent 20mm). Figure S3. S6K Is Enriched in Reactivating Neuroblasts, Related to Figure 3 Late first instar larval VNCs (genotype exhibits slightly delayed development) that are homozygous for the S6K-GFP protein trap (in which GFP has been inserted into the second intron of endogenous S6K) Kelso et al., 2004).  24h VNC from a larva reared on a sucrose-only diet in which grh-GAL4 is driving UAS myr-Akt and UAS mCD8-GFP expression. myr-Akt expression is sufficient for neuroblast reactivation in the absence of the amino-acid stimulus. Confocal channels have been split to allow better visualization of representative mitotic figures. All grh-Gal4-positive neuroblasts have enlarged, and many have begun to divide as evidenced by smaller, GFP-retaining, daughter cells adjacent to the neuroblasts, and pH3 staining. Mitotic pH3 nuclei can be seen in both neuroblasts (white arrows), and neuroblast progeny (yellow arrow). (Scale bars represent 20mm).
Cell 143, 1161-1173, December 23, 2010 ª2010 Elsevier Inc. S5 Figure S5. dILP2 Can Also Reactivate Neuroblasts, Related to Figure 6 27h VNCs from larvae reared on a sucrose-only diet in which repo-GAL4 is driving the expression of either dILP6, dILP2, or dILP2 and dILP6, in the presence of tubGAL80 ts . Larvae were shifted from 18 C to 29 C at larval hatching to block the repressor activity of GAL80, and allow dilp expression. Both dILP6 and dILP2 were sufficient for neuroblast reactivation under these conditions (white arrowheads point to enlarging neuroblasts). Co-expression of dILP2 and dILP6 had no additive or synergistic effect on neuroblast reactivation under these conditions. (Scale bars represent 20mm).