Coordinated sequential action of EGFR and Notch signaling pathways regulates proneural wave progression in the Drosophila optic lobe

The central nervous system (CNS) is composed of many types of cells placed in the right place at the right time. The process begins by inducing neural stem cells/neural progenitors from neuroepithelial cells. In the embryonic CNS or in sensory organ precursor (SOP) development in Drosophila, neural stem cells called neuroblasts produce a variety of neurons and glia. Neuroblasts arise from neuroepithelial cells that express proneural proteins such as Achaete (Ac), Scute (Sc), Lethal of Scute [L(1)sc] and Atonal (Ato) (Cabrera et al., 1987; Jarman et al., 1993; Jarman et al., 1994; Skeath and Carroll, 1992). In these systems, one or only a few cells are selected as neuroblasts from proneural cluster cells and other cells are fated into non-neuronal cells. This phenomenon is explained by a mechanism called lateral inhibition, which is governed by the Notch signaling pathway and has served as a model system for neurogenesis in both vertebrates and invertebrates (Artavanis-Tsakonas et al., 1991; Beatus and Lendahl, 1998; Hassan and Vaessin, 1996). The precise mechanisms underlying the selection of neuroblasts from neuroepithelial cells have not been fully documented, because the process of activating Notch signaling in lateral inhibition is thought to be stochastic. However, lateral inhibition is not the sole mechanism for inducing neuroblasts; distinct transitions from neuroepithelial cells to neuroblasts have also been reported, such as the medulla neurogenesis described in this report.

To analyze the molecular mechanisms that regulate the transition from neuroepithelial cells to neuroblasts, we used the Drosophila optic lobe as a model system (Southall et al., 2008; Yasugi et al., 2008). The Drosophila visual system is composed of the compound eye and the optic lobe in the brain. The latter contains three neural ganglia: the lamina, the medulla and the lobula complex. These optic ganglia are derived from two neuroepithelia: the outer optic anlage (OOA) and the inner optic anlage (IOA) (Meinertzhagen and Hanson, 1993). The OOA generates the outer medulla and the lamina neurons, whereas the IOA generates the inner medulla, the lobula and the lobula plate neurons. The number of neuroepithelial cells of the OOA increases by repetitive symmetric cell divisions during early larval stages. At the early third-instar larval stage (L3 stage), neuroepithelial cells located at the medial edge start to differentiate into medulla neuroblasts; these neuroblasts undergo asymmetric division and produce ganglion mother cells (GMCs), which divide again and become medulla neurons (Egger et al., 2007; Nassif et al., 2003; Toriya et al., 2006). Differentiation of neuroepithelial cells to medulla neuroblasts progresses from the medial to the lateral direction. Expression of L(1)sc always precedes medulla neuroblast formation and this dynamic expression pattern of L(1)sc was named as the proneural wave (Yasugi et al., 2008). The Janus kinase/Signal transducer and activator of transcription (JAK/STAT) pathway is activated in lateral neuroepithelial cells and negatively regulates proneural wave progression (Yasugi et al., 2008). We also studied the role of the EGFR and Notch signaling pathways in proneural wave progression because the impairment of the pathways has previously been suggested to cause abnormal optic lobe development (Sawamoto et al., 1996; Toriya et al., 2006).

The EGFR signaling pathway is a primary signaling pathway that has multiple functions in Drosophila development (Doroquez and Rebay, 2006; Shilo, 2003; Shilo, 2005). Signals from the binding of the ligand to the receptor propagate through Ras-Mitogen activated protein kinase (MAPK) signaling inside the cell. Phosphorylated (activated) MAPK translocates to the nucleus and phosphorylates the ETS-domain transcription factors Pointed P2 (PntP2) and Yan (Brunner et al., 1994; O'Neill et al., 1994). The pnt gene encodes two isoforms, PntP1 and PntP2, both of which are ETS-domain transcription factors (Klambt, 1993). The PntP2 isoform is phosphorylated by activated MAPK and activates target genes, whereas the PntP1 isoform is thought to be a primary target of EGFR signaling (Brunner et al., 1994; Gabay et al., 1996; O'Neill et al., 1994). Yan competes with Pnt for access to the regulatory regions of target genes and represses their expression in the absence of an EGFR signal. Spitz, the primary EGF ligand, is first transcribed in a membrane-tethered precursor form (mSpi). Rhomboid (Rho) is a seven-pass transmembrane protein that localizes in the Golgi apparatus and acts as a protease that cleaves mSpi to make the secreted form (sSpi) (Bier et al., 1990; Golembo et al., 1996; Sturtevant et al., 1993). Argos (Aos) is a diffusible protein that acts as a negative regulator of EGFR (Freeman et al., 1992b).

In the Notch (N) signaling pathway, after binding of the ligand Delta (Dl) with its receptor N, N is processed by proteolysis to release its intracellular domain (NICD). NICD translocates to the nucleus and regulates expression of target genes in association with Suppressor of Hairless [Su(H)] (Lecourtois and Schweisguth, 1995; Schweisguth and Posakony, 1992). The seven clustered Enhancer of Split complex [E(spl)-C] [E(spl)mβ, mγ, mδ, m3, m7 and m8] encodes basic helix-loop-helix (bHLH) transcription factors, and their expression is regulated by Notch signaling (Jennings et al., 1994). The EGFR and Notch signaling pathways act cooperatively in some cases and antagonistically in others (Doroquez and Rebay, 2006; Sundaram, 2005). In this study, we focus on the mechanisms underlying proneural wave progression and report that the EGFR and Notch signaling pathways play pivotal roles.

Flies were grown at 25°C unless otherwise noted. Canton-S flies were used as wild-type controls. The following mutant strains were used: aos^w11 (Freeman et al., 1992b), EGFR^CO (Schupbach and Wieschaus, 1989), EGFR^tsla (Kumar et al., 1998), pnt^Δ88 (Scholz et al., 1993), spi^SC1 (Tio et al., 1994), rho^PΔ5 (Freeman et al., 1992a), Df(1)260-1 (Hinz et al., 1994), N^55e11 (Kidd et al., 1983), N^ts1 (Shellenbarger and Mohler, 1978), Su(H)^Δ47 (Morel and Schweisguth, 2000), Dl^RevF10 (Heitzler and Simpson, 1991), E(spl)^b32.2 p[gro] (Heitzler et al., 1996), and Stat92E^85C9 (Silver and Montell, 2001). EGFR^CO/EGFR^tsla and N^55e11/N^ts1 flies were raised for 5 days at 17°C and then shifted up to 29°C. To investigate the phenotype of pnt^Δ88, Dl^RevF10, Dl^RevF10 pnt^Δ88, E(spl)^b32.2 p[gro] and Stat92E^85C9, we generated clones in a Minute background with hsflp122; FRT82 ubi-GFP M(3)^w124/Tm6B, Tb¹ (Ferrus, 1975) to overcome the defect in proliferation of these mutant cells (data not shown). Su(H)^Δ47 FRT40/T(2;3)SM6a-TM6B, Tb¹ flies were crossed with hsflp122; ubi-GFP FRT40 flies. hsflp122; spi^SC1 FRT40/T(2;3)SM6a-TM6B, Tb¹ flies were crossed with M(2)z arm-lacZ FRT40/CyO Act-GFP flies. rho^PΔ5 clones were induced by crossing with hsflp122; RpS17⁴ arm-lacZ FRT80/Tm6B, Tb¹ flies. spi^SC1 and rho^PΔ5 clones were induced by Minute background to induce large clones. Df(1)260-1 clones were induced by NP7340-Gal4, UAS-flp (Hayashi et al., 2002; Sato et al., 2006). To generate the N^act-expressing pnt^Δ88 clones, hsflp122; act-Gal4, UAS-GFP; FRT82 tub-Gal80 flies were used (Lee and Luo, 2001). Overexpression clones of UAS-EGFR^λtop (Queenan et al., 1997), UAS-Ras^V12 (Karim and Rubin, 1998), UAS-sspi (Schweitzer et al., 1995), UAS-rho (Guichard et al., 1999), UAS-N^act (Go et al., 1998) and UAS-hop^Tum-l (Harrison et al., 1995) were induced by hs-flp; AyGal4 flies (Ito et al., 1997). Overexpression of UAS-EGFR^DN (Freeman, 1996) or UAS-pntP2 (Klaes et al., 1994) was induced by NP3605-Gal4, UAS-GFP.nls flies. RNAi experiments of spi, rho or N were performed by crossing with NP3605-Gal4, UAS-GFP.nls; UAS-dicer2 flies with UAS-IR flies (Dietzl et al., 2007). F1 progenies were raised for 3 days at 25°C and shifted up to 29°C. E(spl)mg-GFP is a reporter that reflects the activity of the Notch signal (Almeida and Bray, 2005).

Immunohistochemistry was performed as described (Huang and Kunes, 1996; Takei et al., 2004). The following antibodies were provided by the Developmental Studies Hybridoma Bank (DSHB): mouse anti-Dac (mAbdac2-3, 1:1000), mouse anti-Arm (N2 7A1, 1:40), mouse anti-Repo (8D12, 1:10) and rat anti-Elav (7E8A10, 1:50). We also used rat anti-L(1)sc (1:800, A. Carmena), guinea pig anti-Dpn (1:1000, J. B. Skeath), rabbit anti-Mira (1:2500, F. Matsuzaki), rabbit anti-Rho (1:1000, E. Bier), rabbit anti-PntP1 (1:1000, J. B. Skeath), rabbit anti-PatJ (1:1000, H. J. Bellen), mouse anti-Yan (1:500, G. M. Rubin), mouse anti-β-gal (1:250, Promega) and rabbit anti-β-gal (1:2000, Cappel). The following secondary antibodies (Jackson) were used at 1:200 dilutions: anti-mouse Cy3, anti-mouse Cy5, anti-mouse FITC, anti-guinea pig Cy3, anti-rat Cy5, anti-rat FITC and anti-rabbit Cy5. Specimens were mounted with vectashield mounting media (Vector) and viewed on a Zeiss LSM510 confocal microscope. Zen software (Zeiss) was used for 2.5D analysis in Fig. S1 in the supplementary material. L3 larvae were staged according to the eye development. Early, no Elav-expressing photoreceptors (72-96 hours AEL); Mid, one to 8 rows of ommatidia are formed (72-96 hours AEL); Late, more than eight rows of ommatidia are formed (96-120 hours AEL). One or two rows of neuroblasts, three to five rows of neuroblasts and more than five rows of neuroblasts are formed in early, mid and late 3L stage, respectively.

In situ hybridization protocols were performed as described previously (Nagaso et al., 2001). For fluorescent in situ hybridization analysis, hybridized probes were detected with anti-digoxigenin-POD (Roche) and fluorescein-labeled tyramid (TSA Fluorescence System, Perkin Elmer) was used as a substrate. A DNA template for spi was amplified from fly genomic DNA with ATGCATTCCACAATGAGTGTAC and GTAGGGTAGCTTGCGCTCCAGA primers and for E(spl)mγ with ATGTCGTCGCTACAAATGTCCG and GGCTGGTCGAACTGGTATTGGA primers.

During development of the optic lobe of Drosophila, both lamina neurons and medulla neuroblasts differentiate from OOA neuroepithelial cells located between the two types of neuronal cells (Fig. 1A,B,D) (Egger et al., 2007). The proneural protein L(1)sc is expressed in only one or two rows of cells in the medial neuroepithelial region (Fig. 1C,E-H). Careful observation revealed a low level of Dpn expression in two to three rows of cells lateral to those expressing L(1)sc (Fig. 1G″, arrow; see also Fig. S1 in the supplementary material). For convenience, we classified cells into four developmental stages: neuroepithelial cells expressing PatJ, neuronal progenitor I expressing a low level of Dpn, neuronal progenitor II expressing L(1)sc and neuroblasts (Fig. 1C). We searched for genes that might promote the proneural wave and focused on the EGFR signal, because it was reported that the projections of photoreceptor axons (R axons) were disorganized in optic lobes of aos mutants, a gene encoding an antagonist of the EGFR signaling pathway (Sawamoto et al., 1996). Relative to wild-type flies, the number of Dpn-expressing medulla neuroblasts and Elav-expressing neurons was smaller and Dac-expressing lamina neurons were absent in aos^w11 mutants (Freeman et al., 1992b) (compare Fig. 1I with 1J; 100%; n=16). We next examined the loss-of-function of EGFR using a combination of null (EGFR^CO) and temperature-sensitive (EGFR^tsla) alleles. No lamina neurons, medulla neuroblasts or medulla neurons were observed in the optic lobes of EGFR^CO/EGFR^tsla mutants (Fig. 1K; 100%; n=10). Overexpression of the dominant-negative form of EGFR (EGFR^DN) using the NP3605-Gal4 line, which expresses Gal4 in optic lobe cells but not in photoreceptor cells (see Fig. S2A-F in the supplementary material), showed phenotypes similar to those of EGFR^CO/EGFR^tsla mutants (see Fig. S2G-O in the supplementary material). Activation of EGFR signaling through the overexpression of pntP2 resulted in premature neuroblast differentiation, the loss of the lamina and fewer neurons (see Fig. S2P-R in the supplementary material). These results suggest that the EGFR signal in the optic lobe is required for medulla neurogenesis. We determined that the impairment of lamina development was not due to the lack of retinal axon innervations because eye development was normal (see Fig. S2L,O,R in the supplementary material), suggesting that the EGFR signaling pathway is required for the development and proliferation of the neuroepithelial cells from which both lamina and medulla neurons develop.

Expression of PntP1 was examined because the EGFR signal induces pntP1 (Gabay et al., 1996). We observed PntP1 expression in three or four rows of neuroepithelial cells, some of which also expressed L(1)sc throughout the L3 stage (Fig. 2A-D), indicating that the EGFR signal is activated in neuroepithelial cells at and near the proneural wave. We also observed PntP1 expression in the glial cells of the lamina at the late L3 stage (Fig. 2C; see Fig. S3A in the supplementary material). Yan is an ETS-domain transcription factor and the phosphorylation of Yan by activated MAPK leads to its degradation (O'Neill et al., 1994; Rebay and Rubin, 1995). Yan is expressed in lateral neuroepithelial cells and the medial part of the lamina, and PntP1 and Yan are expressed in a complementary manner (see Fig. S3B in the supplementary material).

To examine the functions of EGFR signaling, we carried out clonal analyses using a pnt^Δ88 mutant (Scholz et al., 1993). Cells in the pnt^Δ88 mutant clones express neither L(1)sc nor Dpn (Fig. 2E,J; n=42). These pnt^Δ88 mutant cells express the neuroepithelial cell marker PatJ, but not the neuronal marker Elav (Fig. 2F,G,J; n=16 for F and n=25 for G), suggesting that neuroepithelial cells mutant for pnt remain in the neuroepithelial cell state. By contrast, when EGFR^λtop, which encodes a constitutively active form of EGFR (Queenan et al., 1997), was overexpressed, expression of L(1)sc moved laterally and neuroblast formation occurred earlier than in the surrounding wild-type tissue (Fig. 2H,K; n=20). This result suggests that activation of the EGFR signaling speeds up the progression of the proneural wave. We also observed a relatively rare case in which a few ectopic neuroblasts formed separately from the original medulla neuroblasts (see Fig. S4 in the supplementary material). Earlier expression of L(1)sc and Dpn was also observed when Ras^V12, an active form of Ras (Karim and Rubin, 1998), was expressed, suggesting Ras mediates the EGFR signal (Fig. 2I,K; n=18). These results suggest that the EGFR signal is required and sufficient for the progression of the proneural wave.

Among the EGFR ligands, spi is broadly expressed in the optic lobe (Fig. 3A,B). Rho, required for Spi maturation, is expressed at and in front of the proneural wave throughout the L3 stage (Fig. 3C-E). RNAi against spi or rho resulted in disruption of the medulla structure, although Dac-expressing lamina cells were generated (see Fig. S5 in the supplementary material). When large spi^SC1/SC1 clones were generated, disruption or delay of neuroblast formation was observed and L(1)sc expression disappeared in spi^SC1/SC1 cells located away from spi^SC1/+ cells (Fig. 3F; n=20). Similarly, rho^PΔ5/PΔ5 cells distant from rho^PΔ5/+ cells failed to differentiate into neuroblasts, and L(1)sc expression was not observed in rho^PΔ5/PΔ5 cells that were far away from rho^PΔ5/+ cells (Fig. 3H; n=22). In both cases, L(1)sc and Dpn expression was observed in mutant cells adjacent to the wild-type cells, probably because Spi emanating from the wild-type cells induced the development of nearby mutant cells. When sspi or rho was overexpressed, expression of L(1)sc and Dpn moved laterally, as in the case of overexpressing EGFR^λtop (Fig. 3G,I; n=17 for G and n=31 for I: see also Fig. 2H).

It has previously been reported that Rho itself is one of the targets of the EGFR signal (Sapir et al., 1998; Wasserman and Freeman, 1998), therefore we tested whether Rho expression is dependent on EGFR signaling activity. In pnt^Δ88 mutant clones, expression of Rho decreased autonomously (Fig. 4A; n=25). When cells expressing Rho were clonally induced, Rho was found both in cells of the clones and in cells adjacent to earlier differentiating medulla neuroblasts (Fig. 4B; n=23). These results suggest that Rho itself is one of the targets of the EGFR signal, and that Rho stimulates Spi secretion and activates EGFR signaling in more lateral neuroepithelial cells. This sequential induction of EGFR signaling progresses the proneural wave from the medial to the lateral direction.

Clones of cells mutant for Df(1)260-1, which is deficient for all AS-C genes (Hinz et al., 1994), result in a delay of neuroblast formation (Yasugi et al., 2008). Rho expression was only slightly downregulated in Df(1)260-1 cells, suggesting that Rho expression is not dependent on AS-C function and that the EGFR signal regulates expression of L(1)sc and Rho in a parallel way (Fig. 4C; n=26: see also Fig. S6 in the supplementary material for the expression of NP7340-Gal4 line used for generating mutant clones). Conversely, PntP1 expression remained medially in Df(1)260-1 cells, suggesting that AS-C shuts off EGFR signaling (Fig. 4D,E; n=28).

We next examined the role of Notch signaling because its ligand Delta (Dl) is highly expressed in a domain two to three cells wide that includes neuronal progenitor I and II (Fig. 5A,B: see also Fig. S7A-C in the supplementary material). In addition, the expression of E(spl)mγ, a Notch target gene, was examined with the reporter E(spl)mγ-GFP (Almeida and Bray, 2005). This GFP signal is detected strongly in neuronal progenitor I and II and weakly in medulla neuroblasts (Fig. 5C,D). Expression of E(spl)mγ in medial neuroepithelial cells was also confirmed by in situ hybridization (see Fig. S7D in the supplementary material). Compared with wild type, late L3 optic lobes of N^55e11/N^ts1a mutants were smaller, had a much smaller number of medulla neuroblasts and neurons, and had no lamina neurons (Fig. 1I,5E; 100%; n=16). Downregulation of N function in the optic lobe by RNAi showed similar disruption of the optic lobe (see Fig. S8 in the supplementary material).

To further examine the function of Notch signal, clones of cells mutant for Su(H)^Δ47, a null allele of Su(H) (Morel and Schweisguth, 2000), were generated. In and around the Su(H)^Δ47 clones, L(1)sc expression and the onset of Dpn expression moved laterally (Fig. 5F,K; n=22). Advanced progression of the proneural wave was also observed when Dl^RevF10 clones were induced (Fig. 5G,K; n=21). When the signal was activated by inducing clones of cells expressing a constitutively active form of N (N^act), it resulted in the prolonged expression of L(1)sc and PatJ, together with a delay in the onset of Dpn expression (Fig. 5H,I,L; n=20 for 5H and n=14 for 5I). We also observed a loss of weak Dpn expression in the Su(H)^Δ47 mutant clones, whereas ectopic expression of N^act resulted in a level of Dpn expression with an intensity equivalent to or slightly more than that of Dpn expression in neuronal progenitor I (see Fig. S9 in the supplementary material). These results suggest that the Notch signal maintains cells in states equivalent to neuronal progenitor I and II, and that the Notch signal must be downregulated in order to allow the neuronal progenitors to develop into neuroblasts. In E(spl)^b32.2 mutant clones deficient for all E(spl)-C genes (Heitzler et al., 1996), Dl expression was elevated and Dl-positive punctate signals were observed (Fig. 5J; n=28), suggesting that E(spl)-C downregulates Dl expression and thus the Notch signal.

Neuronal progenitor I and II express both PntP1 and Dl (Fig. 6A), indicating that both the EGFR and Notch signaling pathways are activated in these cells. In pnt^Δ88 clones, Dl expression decreased (Fig. 6B; n=18), suggesting that the EGFR signal stimulates activation of the Notch signal through regulation of Dl expression. In Su(H)^Δ47 clones, PntP1 expression decreased (Fig. 6C; n=20). When clones of cells expressing N^act were induced, PntP1 expression remained medial at the expense of Dpn expression (Fig. 6D; n=23). These results suggest that the EGFR and Notch signals are mutually dependent. Neither L(1)sc nor Dpn expression was observed in Dl^RevF10 pnt^Δ88 double mutant clones (Fig. 6E; n=36), indicating that the EGFR signal, irrespective of the Notch signal, is a prerequisite for cells to develop into neuronal progenitor II and neuroblasts. No L(1)sc was detected in clones of cells mutant for pnt^Δ88 and expressing N^act (Fig. 6F; n=23), but a low level of Dpn was found in the cells more lateral to L(1)sc-expressing cells, suggesting that neuroepithelial cells, which receive the Notch signal but not the EGFR signal, respond by expressing a low level of Dpn. Co-activation of EGFR and Notch signals by co-expression of Ras^V12 and N^act resulted in a persistent medial expression of L(1)sc and a delay of Dpn expression, indicating again that the Notch signal can maintain L(1)sc expression in conjunction with the EGFR signal (Fig. 6G; n=19).

We have previously reported that the JAK/STAT signal is activated in neuroepithelial cells and negatively regulates the progression of the proneural wave (Yasugi et al., 2008). In and around the stat92E^85C9 mutant clones, expression of PntP1 and Rho moved laterally in a cell non-autonomous manner (Fig. 7A,C; n=22 for 7A and n=13 for 7C). By contrast, activation of the JAK/STAT signal by expression of hop^Tum-l, a constitutively active form of hop, resulted in medial expression of PntP1 and Rho (Fig. 7B,D; n=26 for 7B and n=14 for 7D). These results suggest that JAK/STAT regulates the timing of Rho expression and EGFR activation. Expression of Dl was similarly regulated by the JAK/STAT signal. Dl expression moved laterally near the Stat92E^85C9 mutant clones and medially near the hop^Tum-l-overexpressing clones (Fig. 7E,F; n=24 for E, n=10 for F). These results suggest that the JAK/STAT signal acts upstream of the EGFR and Notch signals and regulates the activation timing of both signals.

Loss of EGFR function in progenitor cells caused failure of L(1)sc expression and differentiation into neuroblasts. In addition, elevated EGFR signaling resulted in faster proneural wave progression and induced earlier neuroblast differentiation. The activation of the EGFR signal is regulated by a transient expression of Rho, which cleaves membrane-associated Spi to generate secreted active Spi. We also demonstrated that Rho expression itself depends on EGFR function, and thus the sequential induction of the EGFR signal progresses the proneural wave (Fig. 8). Clones of cells mutant for pnt were not recovered unless Minute was employed (data not shown), suggesting that the EGFR pathway is required for the proliferation of neuroepithelial cells. However, the progression of the proneural wave is not regulated by the proliferation rate per se (Yasugi et al., 2008).

The function of the Notch signaling pathway in neurogenesis is known as the lateral inhibition (Artavanis-Tsakonas et al., 1991; Hassan and Vaessin, 1996). A revision of this notion has recently been proposed for mouse neurogenesis, in which levels of the Notch signal oscillate in neural progenitor cells during early stages of embryogenesis, and thus no cell maintains a constant level of the signal (Shimojo et al., 2008). The oscillation depends mainly on a short lifetime and negative-feedback regulation of the Notch effecter protein Hes1, a homolog of Drosophila E(spl). This prevents precocious neuronal fate determination. The biggest difficulty in analysis of Notch signaling is the random distribution of different stages of cells in the developing ventricular zone, which is thus called a salt-and-pepper pattern. In medulla neurogenesis, however, cell differentiation is well organized spatiotemporally and the developmental process of medulla neurons can be viewed as a medial-lateral array of progressively aged cells across the optic lobe (Fig. 8A). Such features allowed us to precisely analyze the functions of Notch. We classified cells into four types according to their developmental stages: neuroepithelial cells expressing PatJ, neuronal progenitor I expressing a low level of Dpn, neuronal progenitor II expressing L(1)sc and neuroblasts expressing high levels of Dpn (Fig. 8B). The Notch signal is activated in neuronal progenitor I and II (Fig. 8B). The EGFR signal turns on in the neuronal progenitor II stage and progresses the stage by activating L(1)sc expression (Fig. 8B). Cells become neuroblasts when the Notch and EGFR signals are shut off (Fig. 8B). Cells stay as neuronal progenitor I when Notch signal alone is activated, whereas cells stay as neuronal progenitor II when the Notch signal is activated in conjunction with the EGFR signal (Fig. 5J). Although the Notch signal is once activated, it must be turned off to let cells differentiate into neuroblasts. In neuronal progenitor II, E(spl)-C expression is induced by Notch signaling, and the increased E(spl)-C next downregulates Dl expression and subsequent activation of the Notch signal (Fig. 5J).

What does Notch do in medulla neurogenesis? We infer that the Notch signal sustains cell fates, whereas the EGFR signal progresses the transitions of cell fate. This was well documented when a constitutively active form of each signal component was induced. EGFR, or its downstream Ras, induces expression of L(1)sc but does not fix its state, even though the constitutively active form is employed. At the same time, a constitutively active Notch sustains cell fates in a cell-autonomous manner (Fig. 5H, Fig. 6F). Constitutively active N receptors, by contrast, autonomously determine cell fates depending on the context: cells become neuronal progenitor I in the absence of EGFR and neuronal progenitor II in the presence of EGFR. The precocious neurogenesis caused by the impairment of Notch signaling suggests that Notch keeps cells in the progenitor state for a certain length of time in order to allow neuroepithelial cells to grow into a sufficient population. In the prospective spinal cord of chick embryo, the development from neural stem cells to neurons progresses rostrocaudally, during which the transition from proliferating progenitors to neurogenic progenitors is regulated by Notch signaling (Hammerle and Tejedor, 2007).

Although Notch plays a pivotal role in determining cell fate between neural and non-neural cells, the function may be context dependent and can be classified into three categories. (1) Classical lateral inhibition is seen in CNS formation in embryogenesis and SOP formation in Drosophila (Artavanis-Tsakonas et al., 1991; Hassan and Vaessin, 1996). Cells that once expressed a higher level of the Notch ligand maintain their cell states and become neuroblasts. (2) Oscillatory activations are found in early development of the mouse brain (Shimojo et al., 2008). Progenitor cells are not destined to either cell types. (3) An association with the proneural wave found in Drosophila medulla neurogenesis as is described here. The Notch signal is transiently activated only once and then shuts off in a synchronized manner. The notable difference in the outcome is the ratio of neural to non-neural cells; a small number of cells from the entire population become neuroblasts or neural stem cells in the former cases (1 and 2), whereas most of the cells become neuroblasts in the latter case (3). The differences between (1) and (2) can be ascribed at least in part to the duration of development. Hes1 expression has been shown to oscillate within a period of 2 hours in the mouse (Hirata et al., 2002; Masamizu et al., 2006), whereas in Drosophila embryogenesis, selection of neuroblasts from neuroectodermal cells takes place within a few hours (Hartenstein and Campos-Ortega, 1984). Thus, even if Drosophila E(spl) has a half-life time equivalent to Hes1, the selection process during embryogenesis probably finishes within a cycle of the oscillation. The process of medulla neuroblast formation continues for more than 1 day, but Notch signaling is activated for a much shorter period in any given cell. This raises the possibility that E(spl)/Hes1 may have a similarly short half-life but outcome would depend on the developmental context.

The functions of EGFR and Notch described here resemble their roles in SOP formation of adult chordotonal organ development (zur Lage and Jarman, 1999); the EGFR signal provides an inductive cue, whereas the Notch signal prevents premature SOP formation. In addition, restricted expression of rho and activation of the EGFR signal assure reiterative SOP commitment. Several neuroblasts are also sequentially differentiated from epidermal cells in adult chordotonal organs.

Unpaired, a ligand of the JAK/STAT pathway is expressed in lateral neuroepithelial cells and shapes an activity gradient that is higher in lateral and lower in the medial neuroepithelium. The JAK/STAT signal acts as a negative regulator of the progression of the proneural wave (Yasugi et al., 2008). In this report, we showed that activation of both EGFR and Notch signaling pathways depends on the activity of the JAK/STAT signal. The JAK/STAT signal probably acts upstream of EGFR and Notch signals in a non-autonomous fashion. These three signals coordinate and precisely regulate the formation of neuroblasts.

We thank F. J. Tejedor for acquainting us with chick neurogenesis, S. J. Newfeld for critical comments on the manuscript, and members of the Tabata lab for helpful comments and discussion. We also thank S. Kato for the use of a confocal microscope and Y. Maeyama for her excellent technical help. We are grateful to A. Carmena, G. H. Baeg, H. J. Bellen, E. Bier, S. J. Bray, M. Freeman, S. Hayashi, K. Matsuno, F. Matsuzaki, N. Perrimon, J. B. Skeath, G. Struhl, the Drosophila Genetic Resource Center Kyoto, the Bloomington Stock Center, VDRC, and the Developmental Studies Hybridoma Bank for flies and antibodies. This work was supported by Grants-in-Aid from MEXT of Japan; Global COE Program (Integrative Life Science Based on the Study of Biosignaling Mechanisms), MEXT, Japan; the Toray Science and Technology Grant; and Yamada Science foundation. Support for T.Y. was provided by Grant-in-Aid for Young Scientists (B) from JSPS and by Sasakawa Scientific Research Grant from The Japan Science Society.