A Regulatory Transcriptional Loop Controls Proliferation and Differentiation in Drosophila Neural Stem Cells

Neurogenesis is initiated by a set of basic Helix-Loop-Helix (bHLH) transcription factors that specify neural progenitors and allow them to generate neurons in multiple rounds of asymmetric cell division. The Drosophila Daughterless (Da) protein and its mammalian counterparts (E12/E47) act as heterodimerization factors for proneural genes and are therefore critically required for neurogenesis. Here, we demonstrate that Da can also be an inhibitor of the neural progenitor fate whose absence leads to stem cell overproliferation and tumor formation. We explain this paradox by demonstrating that Da induces the differentiation factor Prospero (Pros) whose asymmetric segregation is essential for differentiation in one of the two daughter cells. Da co-operates with the bHLH transcription factor Asense, whereas the other proneural genes are dispensible. After mitosis, Pros terminates Asense expression in one of the two daughter cells. In da mutants, pros is not expressed, leading to the formation of lethal transplantable brain tumors. Our results define a transcriptional feedback loop that regulates the balance between self-renewal and differentiation in Drosophila optic lobe neuroblasts. They indicate that initiation of a neural differentiation program in stem cells is essential to prevent tumorigenesis.


Introduction
Stem cells are defined by their ability to self-renew and produce differentiating daughter cells. These two features must be tightly controlled since misregulation can lead to stem cell loss and tissue degeneration or overproduction of stem cells and tumor formation. Drosophila neural stem cells called neuroblasts (NBs) are a well studied model system for investigating molecular and cellular mechanisms of stem cell maintenance and tumorigenesis as their mode of cell division and cell fate determination are well defined [1,2]. In the larval brain several types of NBs are defined by their locations and ways of cell division [3]. NBs in the central brain delaminate from the ventral neuroectoderm during embryogenesis and are subdivided into type I and type II NBs [2]. Type I NBs divide asymmetrically and produce another NB and a ganglion mother cell (GMC), which divides symmetrically into two neurons and/ or glia cells. Type II NBs also divide asymmetrically but produce another NB and an intermediate neural progenitor (INP), which continues to divide asymmetrically producing INPs and GMCs [4][5][6]. While NBs in the central brain are formed during the embryonic stage, NBs in the so-called optic lobes show a different mode of neurogenssis. The optic lobes are located at the lateral side of each brain lobe and NBs in this region produce neurons for visual processing in the adult stage [7]. There are two proliferating centers in the optic lobe, the outer and inner proliferation center. In both areas, the number of NBs increases during larval stages. Neuroepithelial cells (NE cells) in the outer proliferation center produce lamina and medulla neurons, while NE cells in the inner proliferation center mainly give rise to lobula and lobula plate neurons. In early larval stages, NE cells proliferate by repetitive symmetric cell divisions. In late larval stages, the formation of medulla NBs starts on the medial side of the neuroectoderm and a wave of differentiation progresses from the medial to the lateral side ( Figures 1A-1D) [8,9]. Medulla NBs divide asymmetrically and display a lineage similar to the type I NBs in the central brain.
The basic molecular machanisms of asymmetric cell division are common to all NBs. The Par complex proteins Par3/ Bazooka (Baz)-Par6-atypical protein kinase C (aPKC) localize to the apical cortex during mitosis and direct the orientation of the mitotic spindle along the apicobasal axis [10,11]. During mitosis, Baz-Par6-aPKC regulate the asymmetric localization of three cell fate determinants to the basal cortex. These determintants are inherited only by the GMC, where they stop self-renewal, terminate cell cycle progression, and direct the cell towards differentiation. The set of basaly segregating fate determinants includes the Notch repressor Numb, the NHLdomain protein Brain tumor (Brat), and the homeodomein transcription factor Prospero (Pros) [12][13][14][15][16]. In numb, brat, or pros mutants, impaired cell fate determination in larval NBs leads to overproliferation of NBs and transplantation of these mutant brains to the abdomen of adult host flies causes malignant tumors that eventually become metastatic and kill the host [14][15][16][17][18][19][20].
In this study, we characterize the role of Da as a tumor suppressor in the Drosophila larval brain. We show that inhibiting Da function results in overproliferation of medulla optic lobe NBs and leads to the formation of transplantable brain tumors. We explain this phenotype by showing that Da and Ase promote differentiation through regulating Pros expression, suggesting that the differentiation program is set up in neural stem cells and asymmetric segregation of Pros ensures that the differentiation program is implemented only in one of the two daughter cells. Our data indicate that a regulatory loop between Da/Ase and Pros maintains the balance between self-renewal and differentiation in optic lobe NBs.

Fly Genetics
Flies were grown at 25uC unless otherwise noted. w flies were used as wild-type controls.  were raised for 1 day at 25uC and shifted up to 29uC. Conditions for transplantation experiments are descried below.

Histology
Third instar wandaring larvae were dissected in PBS and fixed in 3.7% Formaldehyde in PBS. Samples were washed three times after fixation with PBS containing 0.3% Triton X-100 and transferred to blocking solution (PBS containing 5% normal goat serum and 0.3% Triton X-100). Specimens were incubated with primary antibodies diluted in blocking solution for overnight at 4uC. Primary antibodies were washed four times with PBS containing 0.3% Triton X-100 before the incubation with secondary antibodies for overnight at 4uC. Secondary antibodies were washed four times with PBS containing 0.3% Triton X-100. Specimens were mounted with Vectashield mounting media (Vector) and viewed on a Zeiss LSM710 confocal microscope. Imaris software (Bitplane) was used for preparing three-dimensional images. The following antibodies were provided by

Transplantation of larval brains
Glass needles used in the transplantation experiments were constructed, and a simple micro-injection system was prepared as described previously [19]. 4-6 days old adult w females, kept at 25uC, were used as hosts. The host flies were immobilized on an ice-cold metal plate and stuck on a piece of double-sided sticky tape, with their ventral sides up. Crosses were set up at 29uC between virgin females of UAS-dicer2; insc-Gal4 UAS-mCD8::GFP/ CyO and males of da RNAi or w. GFP-positive wandering third instar larvae were collected and larval brains were dissected in icecold PBS. The dissected brain lobes were transferred into a small drop of cold PBS on a glass microscope slide and cut into two pieces to separate the optic lobes from the central brain. The isolated optic lobes were transplanted into the abdomen of host flies under a GFP microscope to ensure cells were collected by the needle and transplanted into the hosts. After transplantation, host flies were allowed to recover at room temperature for 1-2 hours in fresh standard Drosophila medium before transferred to and maintained at 29uC.

Da acts as a tumor suppressor in optic lobe neuroblasts
To further characterize the overproliferation caused by da RNAi, we induced da RNAi by insc-Gal4 in all larval NBs. The number of Deadpan (Dpn) expressing NBs increased at the expense of Embryonic lethal abnormal vision (Elav) expressing neurons (100%, n = 14) ( Figure S1). Although Da was expressed in all NBs of the central brain and in some progenitor cells ( Figures S2A-S2B''') we did not find any phenotype in these lineages when we induced da 3 amorphic mutant clones using mosaic analysis with a repressible cell marker (MARCM) technique [37] (0%, n = 19 for type I NB lineages, and 0%, n = 16 for type II NB lineages) ( Figures S2C-S2F'').
The visual processing centers of the fly brain arise from the socalled optic lobes. The medial surface of the optic lobes is surrounded by medulla NBs that differentiate from NE cells and generate medulla neurons on the inner side of the brain (Figures 1A-1D) [8,9,36]. In the optic lobe, Da was expressed in NE cells and in medulla NBs ( Figures 1E-1F''). To induce da RNAi in the optic lobe, we used a dpn-Gal4 driver line that showed strong Gal4 expression in NE cells and medulla NBs and weak expression in medulla neurons ( Figure S3) (called dpnOL-Gal4 below, Janelia Gal4 stocks, Bloomington Stock Center #47456) [38]. Expression of da RNAi from dpnOL-Gal4 caused a strong increase of Dpn positive NBs (100%, n = 12) (compare Figures 1G-1G'' and 1I-1I''). We also checked the da RNAi phenotype with the mitotic marker Phospho-Histone H3 (PH3), the NB marker Miranda (Mira) and the neuronal marker Elav. In the wild type, PH3 positive mitotic cells (NBs and GMCs) were restricted to the periphery of the optic lobe ( Figures 1H-1H''). In da RNAi samples, PH3 positive cells were mislocalized and ectopically found in the inner side of the brain (100%, n = 18) ( Figures 1J-1J''). To confirm this phenotype, da 3 mutant clones were induced in the optic lobe. In da 3 clones, Dpn positive NBs were found in the region that was normally occupied by medulla neurons (92%, n = 26) (Figures 1K-1L). Thus, da is required for cell fate determination in medulla NBs.
To test whether the ectopic NBs in da RNAi brains have unlimited growth potential and can induce malignant tumors, optic lobes expressing GFP under the control of insc-Gal4, were dissected and implanted into the abdomen of wild type adult host flies [19]. Transplanted cells from da RNAi brains proliferated and GFP positive cells were observed in the host flies (17%, n = 47), while no substantial growth was observed in control samples (0%, n = 30) (Figures 1M and 1N). PH3 positive mitotically active cells were observed in the tissue from transplanted da RNAi samples, and this tumor tissue consisted of both Dpn-expressing NB-like cells and Elav-expressing neuron-like cells ( Figure 1O). This suggests that the da tumor cells proliferate and some of the cells keep the stem cell state, but these cells also produce differentiating cells. This is consistent with the result from da 3 clones, in which both ectopic NBs and differentiated neurons were observed ( Figures 1K-1L). From these results, we conclude that Da acts as a tumor suppressor in optic lobe NB lineages.

Ase regulates NB differentiation during medulla NB development
Da is an E-box protein that heterodimerizes with other bHLH type transcription factors, such as the proneural proteins of the AS-C [23,24,39]. The AS-C is composed of four transcription factors called Achaete (Ac), Scute (Sc), Lethal of Scute (L(1)sc), and Asense (Ase) [40]. While Ac is not expressed in the optic lobe, three of four AS-C proteins show specific expression [8,9]. Sc is expressed in the NE cells and NBs, L(1)sc is transiently expressed in the transition zone between NE cells and NBs, and Ase is expressed in NBs and GMCs in the developing medulla (Figures 2A-2A''' show expression of Ase) [8,9]. To test which of the AS-C genes might act with Da during cell fate determination in medulla NBs, we induced clones of several deletion lines that uncover the AS-C region ( Figures 2B-2E, clones in the optic lobe were induced by NP7340-Gal4 and UAS-flp [36]). Ectopic NBs were observed in clones of Df(1)260-1 uncovering all AS-C genes or in ase 1 that uncovers the ase coding region (84%, n = 57 for Df(1)260-1 clones, and 88%, n = 73 for ase 1 clones) ( Figures 2B-2C'', 2E). On the other hand, no phenotype was observed in clones of Df(1)sc19, which deletes ac, sc, and l(1)sc (0%, n = 24) ( Figures 2D-2D'', 2E). Since the phenotype of Df(1)260-1 or ase 1 clones was similar to the phenotype of da 3 mutant clones and heterodimerization between Ase and Da has been shown, we conclude that Da acts together with Ase to regulate cell fates in the optic lobe [39]. It has been reported that Da is required for the timly differentiation from NE cells to NBs and L(1)sc is involved in this transition during the optic lobe development [8]. From the expression pattern of AS-C genes and results from the clonal analysis using deficiency lines, we propose a dual function for Da: As a heterodimer with L(1)sc, Da promotes the transition from NE cells to NBs. Later, Da acts with Ase in NBs to promote differentiation and prevent tumor formation.

Pros is a downstream target of Da and Ase
To identify the downstream targets of Da and Ase, we tested the expression of candidate genes. The homeodomain transcription factor Pros acts as a cell fate deteminant in embryonic and larval NBs and is regulated by Da and Ase in embryos [12,15,16,26,41]. In the larval optic lobe, Pros is localized to the basal cortex of dividing NBs and nuclear in GMCs and newly born medulla neurons (Figures 2A-2A''') [9,42]. We tested whether Pros expression is dependent on Da and/or Ase. Pros expression decreased in da 3 or ase 1 mutant clones (87%, n = 38 for da 3 clones, and 74%, n = 38 for ase 1 clones) ( Figures 3A-3B'') suggesting that Pros acts downstream of Da and Ase. To test whether pros is required for cell fate determination in the optic lobe, we induced pros 17 mutant clones. In pros 17 mutant clones, ectopic NBs were observed in the medulla neuron layer, which was similar to the phenotype of da 3 or ase 1 mutant clones (88%, n = 26) ( Figures 3C-3C''). Overexpression of Pros, on the other hand, resulted in a decrease of medulla NBs (100%, n = 8) (Figures 3E-3E'', compare to Figures 1G-1G''). To test whether Pros acts downstream of Da, we overexpressed Pros in a da RNAi background. A reduced number of medulla NBs were observed in optic lobes overexpressing Pros in a da RNAi background, indicating that pros is epistatic to da (100%, n = 10) ( Figures 3F-3F''). Thus, Pros is a key downstream target of Da and Ase in optic lobe NBs.
Next, we asked whether Pros expression is regulated by Da in the central brain where da is not required for NB self-renewal ( Figures S2C-S2F''). Nuclear Pros expression was found in differentiating daughter cells in the wild type. Pros expression remained in da 3 mutant clones (0%, n = 11) ( Figure S4). Thus, unlike in the optic lobe, Da is not essential for Pros expression in the central brain. This explains why the da phenotype is sepecific to the optic lobe NBs, while pros mutations cause overproliferation in all larval NBs ( Figures 3C-3C'') [14][15][16]. We speculate that other factors may act redundantly to regulate Pros expression in the central brain.
If Pros is induced by Da and Ase, then how are their functions turned off after asymmetric division? To test whether Pros can terminate the expression of ase, we examined Ase expression in pros 17 clones. While Ase expression was restricted to the periphery of the optic lobe in wild type, Ase expression continued on the inner side of the optic lobe in pros 17 clones (77%, n = 48) ( Figures 3D-3D''). Thus, Pros turns off Ase expression and this transcriptional negative feedback loop regulates the proliferation and differentiation of NBs.

Discussion
A prevailing view in stem cell biology is that a self-renewal program allows prolonged proliferation in stem cells and is turned off upon differentiation. Our data challenge this view and demonstrate that the ability to differentiate is pre-programmed in neural stem cells. This explains why transcription factors like Da and Ase that are thought to be required for NB specification can be required for proper differentiation and act as tumor suppressors. We propose that a regulatory transcriptional loop assures cell fate determination and inhibits tumor formation ( Figure 3G). In a medulla NB, Da and Ase heterodimers induce Pros expression [39] but Pros is excluded from the nucleus and therefore can not terminate Ase expression. After asymmetric cell division, however, Pros enters the nucleus of the GMC where it initiates differentiation and cell cycle exit [43]. In the GMC, Pros terminates Ase expression and therefore triggers an irreversible decision towards differentiation. The data from embryonic NBs suggest that Pros can directly bind to the ase region and regulates its expression [43]. In the absence of this regulation, GMCs maintain the stem cell fate and continue to grow into malignant tumors.
The role of Da, Ase, and Pros in neural stem cells could be conserved in mammals. Mammalian class I bHLH genes, namely E2A (encoding the E12 and E47 proteins), E2-2, and HEB are expressed in the developing brain. E2A, HEB, or E2A/HEB transheterozygous mutant mice show a brain size defect, suggesting that class I factors also regulate mouse brain development [44,45]. Mash1 and Prox1, the vertebrate orthologs of Ase and Pros, are expressed in proliferating neural precursor cells of the developing forebrain and spinal cord [46]. Like in Drosophila, Mash1 induces Prox1 and Mash1 promotes an early step of differentiation in neural stem cells [46]. Like in vertebrates, NE cells in the Drosophila optic lobe first proliferate by symmetric cell division and then become asymmetrically dividing NBs [8,9,47]. From these molecular and developmental similarities, we speculate that the transcriptional regulatory mechanism we have identified might be well conserved in mammalian brains.
Our data are of particular relevance in light of the recently postulated role of stem cells in the formation of malignant tumors [48,49]. Failure to limit self-renewal capacity in stem cells or defects in progenitor cell differentiation can both lead to the formation of cells that continue to proliferate and ultimately form tumors [50]. While genes acting in stem cells are thought to promote self-renewal, genes required in differentiating cells are thought to promote differentiation and limit proliferation and are therefore candidate tumor suppressors. Our data challenge this view and show that the path to differentiation is initiated in the stem cell and therefore even genes specific to stem cells can act as tumor suppressors. It will be interesting to determine whether a similar mechanism acts in mammalian neural stem cells as well. If it does, the expression pattern of a gene can no longer be used as a main criterium for whether it promotes or inhibits self-renewal in stem cell lineages.