A Serrate–Notch–Canoe complex mediates essential interactions between glia and neuroepithelial cells during Drosophila optic lobe development

Glial cells are not a mere structural filler within the brain but they perform multiple and vital tasks for the proper development and functioning of the nervous system. The roles that glial cells display during the development of the nervous system are as diverse as the multiple glial types specified, including axon ensheathment, axon guidance, phagocytosis and the establishment of the blood–brain barrier (Banerjee and Bhat, 2007; Bundgaard and Abbott, 2008; Edenfeld et al., 2005; Lemke, 2001; Nave and Trapp, 2008; Parker and Auld, 2006). In Drosophila, three main different types of glial cells have been very well characterized in the larval brain, namely surface glia (subdivided into the outermost perineurial glia and the underlying subperineurial glia), cortex glia and neuropile glia (Hartenstein, 2011; Pereanu et al., 2005; Stork et al., 2012). Embryonic neuroblasts (NBs), specifically neuro-glioblasts, give rise to the precursors of the larval glia that will increase in number throughout the larval life, mainly at late larval stages and fundamentally from neuroglioblast division, although the mitosis of differentiated glia also contributes (Pereanu et al., 2005). These three types of glia perform crucial functions during the development of the Drosophila brain. For example, surface glial cells provide signals at early stages of the larval period to induce embryonic quiescent NBs to resume proliferation (Ebens et al., 1993). Cortex glial cells have important trophic functions for neurons, and neuropile and surface glia act as key intermediate targets during axon pathfinding in the brain (Hidalgo, 2003; Hoyle et al., 1986; Pielage and Klämbt, 2001; Poeck et al., 2001; Sepp et al., 2001; Tayler and Garrity, 2003). Glial processes engulf NBs and neurons in the Drosophila larval brain, which is formed by the central brain and the optic lobes (Hartenstein et al., 2008).

The optic lobes, which are located at the lateral side of both brain hemispheres, form part of the Drosophila visual system. They derive from neuroectodermal placodes in the embryonic head that invaginate, lose contact with the epidermis and attach to the brain (Green et al., 1993). At the beginning of the larval life, just after larval hatching, cells of the optic lobe start to proliferate and they separate into an outer proliferation center (OPC), which will give rise to the outer medulla and lamina neurons, and an inner proliferation center (IPC), which generates the inner medulla, the lobula and the lobula plate neurons (Hofbauer and Campos-Ortega, 1990; Meinertzhagen and Hanson, 1993). The OPC anlage is formed by neuroepithelial (NE) cells, which proliferate by symmetric cell division until the OPC reaches a proper size, at third larval instar. At this point, NE cells switch to asymmetric differentiative divisions generating medulla NBs at the medial edge of the OPC anlage (Egger et al., 2007). Differentiation of NE cells to medulla NBs progresses from the medial to the lateral edge of the OPC in a ‘proneural wave’, which was recently identified by the transient and local expression at the wave front of the proneural gene lethal of scute (l′sc) (Yasugi et al., 2008). Over the past few years multiple signaling pathways have proved to be essential for regulating the proneural wave progression within the neuroepithelium in a tissue-autonomous way, including Notch, JAK–STAT, EGFR–Ras–PointedP1 (PntP1) and Fat–Hippo signaling pathways (Egger et al., 2010; Ngo et al., 2010; Reddy et al., 2010; Yasugi et al., 2010; Yasugi et al., 2008). However, not much is known about non-autonomous regulatory mechanisms. Given that glial cells proliferate and differentiate in close contact with the OPC and IPC neuroepithelia, the glia might influence the development of the optic lobe, including the progression of the proneural wave.

In this work, we have analyzed the relationships between NE cells of the optic lobe and the surrounding glia. We found that the PDZ (PSD-95, Discs Large, ZO-1)-domain-containing protein Canoe (Cno) (Miyamoto et al., 1995) is expressed in the NE cells of the OPC and IPC. Cno and its vertebrate homologues AF-6/Afadin are scaffolding proteins that are predominant at adherens junctions (AJs) where they regulate the linkage of AJs to the actin cytoskeleton (Lorger and Moelling, 2006; Mandai et al., 1997; Matsuo et al., 1999; Miyamoto et al., 1995; Sawyer et al., 2009). Cno also performs AJ-independent functions regulating neuron–glia interactions and asymmetric cell division (Slováková and Carmena, 2011; Speicher et al., 2008). In addition, Cno acts as an integration hub of different signaling pathways, including Ras and Notch during muscle and heart progenitor specification (Carmena et al., 2006).

The Ras pathway triggered by the EGFR ultimately activates the ETS transcription factor Pnt that exists in two isoforms, PntP1 and PntP2, the former acting in the optic lobe (Klämbt, 1993; Yasugi et al., 2010). PntP1 is very locally activated at the transition zone, where, in turn, it induces the local expression of L′sc (Yasugi et al., 2010). The Notch receptor is activated by two different ligands, Delta (Dl) and Serrate (Ser), which trigger the proteolytic cleavage of the intracellular domain of Notch. This domain translocates into the nucleus where it associates with Suppressor of Hairless [Su(H)] and activates target genes. The best characterized Notch targets are the genes of the Enhancer of Split-Complex [E(spl)-C], which comprises seven genes that encode basic-helix-loop-helix (b-HLH) transcription factors, namely mδ, mγ, mβ, m3, m5, m7 and m8 (Bailey and Posakony, 1995; Delidakis and Artavanis-Tsakonas, 1992; Jennings et al., 1994; Knust et al., 1992; Lecourtois and Schweisguth, 1995; Rebay et al., 1991).

The data presented in this work strongly suggest that a complex between Cno, Notch (present in NE cells) and Ser (present in the glia) is key for interactions of NE cells and glia during the development of the optic lobe. We show that the Notch ligand Ser displays tissue-autonomous and non-autonomous effects in the glia and in the neuroepithelium, activating different Notch E(spl) target genes in each tissue. By activating Notch in the glia, Ser regulates its proliferation and, by triggering Notch signaling in NE cells, Ser restricts the activation of Ras–PntP1 signaling, and hence the activation of L′sc, to the transition zone.

Cno is expressed in Drosophila embryonic neuroectoderm and in the delaminated NBs where Cno displays an essential role in asymmetric NB division (Speicher et al., 2008). In an attempt to characterize a potential function of Cno in the differentiation of NE cells to medulla NBs during the development of the larval optic lobe, we first analyzed Cno expression in this tissue. Cno was highly enriched at the apical most region of NE cells in the OPC and IPC (Fig. 1B–E′). Cno and its vertebrate orthologs are present at the AJs of different epithelial tissues (Mandai et al., 1997; Matsuo et al., 1999; Sawyer et al., 2009). Indeed, Cno colocalized at the apical region of NE cells with Bazooka (Baz), another well-known component of the zonula adherens, where Baz colocalizes with the DE-cadherin Shotgun (Shg) (Krahn et al., 2010) (Fig. 1F–G″).

Next, we wanted to analyze the effect in the optic lobe of knocking down cno. With that purpose, we overexpressed a cnoRNAi plus the gene encoding the Dicer2 enzyme under the neuroepithelia-specific Gal4 driver c855a (Manseau et al., 1997). The loss of Cno, as tested by immunofluorescence, was complete and specific (Fig. 2E–F″). We analyzed anterior views (see Fig. 1A), in which changes in the proneural wave progression respect to well-defined landmarks, such as the lamina furrow, are easily detected (Fig. 2A–A″). In UAS-cnoRNAi optic lobes, the proneural wave was advanced compared with control optic lobes (68%; n = 41 brain hemispheres), even reaching the lamina region at some points (Fig. 2A–C″). The overexpression of cno caused the opposite phenotype (100%; n = 17), with a marked delay in the progression of the proneural wave (Fig. 2D–D″) and the concomitant decrease in the differentiation of medulla NBs (compare Fig. 2A and D). To further prove a specific role of cno in this process, we carried out mosaic analysis with repressible cell marker (MARCM) clones in the neuroepithelium using armadillo (arm)-Gal4 as a driver. Clones for the null allele cno^R2, labeled with CD8-GFP, revealed L′sc-expressing cells within the clone in an advanced position relative to the cno⁺ cells outside the clone (40%; n = 20) (Fig. 2G,G′).

Notch signaling is required to regulate the progression of the proneural wave (Egger et al., 2010; Ngo et al., 2010; Reddy et al., 2010; Wang et al., 2011; Weng et al., 2012; Yasugi et al., 2010). Given that Cno and Notch functionally interact in other systems (Carmena et al., 2006; Miyamoto et al., 1995) and that cno mutant phenotypes were very reminiscent of those previously described for Notch mutants in the neuroepithelium (Yasugi et al., 2010), we decided to analyze the relative localization of Cno and Notch in the optic lobe. Confocal analysis of double immunofluorescence for Cno and Notch showed colocalization of these proteins at the AJs of NE cells (Fig. 3A–B″).

Next, we wondered whether Cno had some effect on the Notch activity, which is required in NE cells to maintain their fate (Egger et al., 2010; Wang et al., 2011; Weng et al., 2012). First, we wanted to analyze the Notch activity in wild-type brains. For this analysis we used different Notch activity reporters, because it has been previously shown that different E(spl) Notch target genes display distinct expression patterns in other tissues (Cooper et al., 2000; de Celis et al., 1996; Wech et al., 1999). Specifically, we looked at Gbe+Su(H)lacZ (see the Materials and Methods) (Furriols and Bray, 2001), E(spl)mβ-CD2 (mβ-CD2 hereafter) (de Celis et al., 1998), E(spl)mδ-lacZ (mδ-lacZ) (Cooper et al., 2000) and E(spl)m7-nuclacZ (m7-nuclacZ) (Pines et al., 2010). At third larval instar, Gbe+Su(H)lacZ was barely detected in NE cells (Fig. 3C–F′); it started to be highly expressed at the transition zone where the Notch ligand Dl is enriched in a characteristic punctuated pattern (Fig. 3C–D″) (Weng et al., 2012; Yasugi et al., 2010). L′sc is also very locally activated at the transition zone, in progenitor I (PI) and progenitor II (PII) cells; the latter juxtaposed to the emerging medulla NBs (Yasugi et al., 2010) (Fig. 3F,K). Gbe+Su(H)lacZ was highly detected in PI but its expression dropped markedly in PII (Fig. 3F,F′,K), and it was again highly activated in NBs along with the Notch-responding gene deadpan (dpn) (Krejcí et al., 2009) (Fig. 3E,E′,K). A similar expression pattern was shown for the Notch activity reporter mγ–GFP (Weng et al., 2012) (Fig. 3K). A weak expression of Gbe+Su(H)lacZ was detected in the surface glia in close contact with NE cells (Fig. 3D,D″,K). Interestingly, the mβ-CD2 reporter was highly expressed in these surface glia, the same location in which the mδ-lacZ reporter was detected, although at much lower levels (Fig. 3G,G′,I,I′,K); none of them was clearly present in NE cells (Fig. 3G,G′,I,I′,K), but the mβ-CD2 reporter was also detected in medulla NBs (Fig. 3H,H′,K). The only reporter that was expressed in the neuroepithelium was m7-nuclacZ, which also showed a uniform expression in the transition zone and in the emerging NBs (Fig. 3J–K). Specific effects of Notch loss and gain of function on the expression of these reporters were observed in the optic lobe and in the surrounding glia, confirming them as bona fide Notch target genes in the brain (supplementary material Fig. S1).

Given this pattern of Notch activity and the fact that Dl is enriched at the transition zone with almost undetectable levels in NE cells or in the surrounding glia (Fig. 3C,C′,D,D′) (Weng et al., 2012; Yasugi et al., 2010), we wondered whether Ser, another ligand of Notch that activates it in different tissues and cell contexts (Fleming et al., 1990; Rebay et al., 1991; Thomas et al., 1991), would be expressed in the optic lobe. A UAS-CD8::GFP construct under a Ser-Gal4 driver revealed that, whereas Dl was highly restricted to the transition zone, Ser was active in what seemed to be the brain glia, with no detectable expression in NE cells (Fig. 4A–B′). Double labeling with the glial marker Nervana (Nrv) confirmed the expression of Ser (Ser-Gal4>>UAS-CD8::GFP) in the glia (Fig. 4C–C″). Given that CD8-GFP is a membrane protein that reflects where Ser-Gal4 is expressed but not the real localization of the Ser protein within the glia, we used a specific Ser antibody to analyze Ser distribution. Unlike CD8-GFP, Ser was not uniformly present around all glial cell membranes but it was particularly concentrated at the glial membranes in contact with the apical side of NE cells, where Ser colocalized with Notch (Fig. 4D–E″).

Cno and Notch, as well as Notch and glial Ser, strongly colocalized at the most apical side of NE cells, at the interface with the surrounding glia. Hence, we speculated that these proteins might be in a complex. To test this hypothesis, we performed co-immunoprecipitations (co-IPs) from larval head extracts and found that Cno co-immunoprecipitated with both Notch and Ser (Fig. 4F). Hence, Ser, Notch and Cno exist in a complex in vivo.

Colocalization and co-IP experiments suggested that Cno could contribute to stabilize Notch at the AJs of NE cells to facilitate Notch binding to its ligand Ser in the adjacent glia. Indeed, we observed that not only Notch but also Ser were highly reduced after knockdown of cno (UAS-cnoRNAi) in NE cells (63%; n = 19), compared with controls (Fig. 5A–B″). Then, we reasoned that the Notch activity should be also affected in this cno mutant condition. In fact, the analysis of the Notch reporter m7-nuclacZ showed a slight but consistent reduction of its expression in the neuroepithelium (67%; n = 9) (Fig. 5C–D′).

To analyze in more detail the effect of glial Ser in the development of the neuroepithelium, we expressed in the glia a dominant-negative form of Ser (DNSer) called Bd E24 (recently renamed Ser^sec), which is a strong dominant-negative form (Fleming et al., 2013). The expression of DNSer in the glia caused a marked advance of the proneural wave (100%; n = 15) (Fig. 5E,E′,H,H′), a phenotype reminiscent of that displayed when cno is knocked down in the neuroepithelium. The expression of DNSer in the glia also caused ectopic expression of L′sc within the neuroepithelium (100%; n = 21) (Fig. 5F,F′,I,I′). Moreover, PntP1 was also ectopically activated in DNSer (100%; n = 11), colocalizing with L′sc in several cells, indicating that the Ras pathway was misregulated (Fig. 5G,G′,J,J′). Hence, Ser in the glia is crucial for determining the correct spatial activation of the Ras–PntP1 pathway in the neuroepithelium. The expression of DNSer in the glia caused a significant depletion in the glial levels of the Notch reporter mβ–CD2 (100%; n = 12) (Fig. 5K,L), expression that was rescued after the simultaneous expression of DNSer and an activated form of Notch, Nintra (Ni), in the glia (Fig. 5M,M′). However, the expression of DNSer plus Ni in the glia did not rescue the phenotype of ectopic L′sc⁺ and PntP1⁺ cells in the neuroepithelium (Fig. 5N,N′). Intriguingly, the overexpression of DNSer in the glia led to a striking reduction in the expression of the Notch activity reporter m7-nuclacZ in NE cells (100%; n = 7) (Fig. 5O-P′). To further support the role of glial Ser in the underlying neuroepithelium, we performed mosaic analysis with MARCM clones in the glia using the nrv2-Gal4 driver. Clear ectopic L′sc was observed underlying glial Ser^RX82 mutant clones in 69% of the brain hemispheres analyzed (n = 26) (Fig. 5Q). Thus, altogether, these results strongly suggest that, in normal conditions, Ser non-autonomously triggers Notch signaling in the neuroepithelium and this activity avoids the premature activation of Ras–PntP1 signaling in NE cells.

By using Repo, as a specific glial marker, and phospho-histone3 (PH3), as an indicator of mitotic cells, we found that the expression of DNSer in the glia led to a statistically significant decrease in the number of Repo⁺ and PH3⁺ cells, compared with control optic lobes (Fig. 6D–D″,G,H). This phenotype was opposite to the overexpression of the constitutive active form Ni in the glia (Fig. 6E–E″,G,H). Interestingly, the expression of UAS-cnoRNAi in NE cells showed a similar phenotype to the expression of DNSer in the glia (Fig. 6B–B″,H): a significant decrease in the number of glial PH3⁺ cells. In addition, the overexpression of cno in NE cells showed, similar to the overexpression of Ni in the glia, a significant increase in Repo⁺ and PH3⁺ cells, compared with control optic lobes (Fig. 6C–C″,G,H). Indeed, in UAS-cnoRNAi, the Notch reporter mβ-CD2 was downregulated in the surface glia, specifically in the subperineurial glia, in close contact with the neuroepithelium (53%; n = 19) (Fig. 6K–L′) and the overexpression of cno in the neuroepithelium led to an upregulation of the Notch activity reporter Gbe+Su(H)lacZ in the glia (100%; n = 12), whereas, in normal conditions, it is barely detected (Fig. 6I–J′; see also Discussion). The simultaneous expression of DNSer and Ni in the glia rescued the DNSer phenotype in the glia, but not in the neuroepithelium (Fig. 6F–H). Indeed, the expression of mβ-CD2, which was downregulated in the glia after expressing DNSer in this tissue, was also rescued when DNSer and Ni were simultaneously expressed in the glia (Fig. 5M,M′). Hence, Ser is required autonomously in the glia to regulate their proliferation.

Taking all our data together, we propose a working model in which the brain glia would display an important role during the progression of the proneural wave in the optic lobe (Fig. 7H). Ser, present in the glia in close contact with the neuroepithelium, can activate Notch in the glia, as detected by the Notch activity reporters mβ-CD2, mδ-lacZ and, to a lesser degree, Gbe+Su(H)lacZ contributing to regulate glia proliferation. Ser can also activate Notch in the neuroepithelium, as revealed by the Notch activity reporter m7-nuclacZ, activity that restricts the EGFR–Ras–PntP1 signaling and hence L′sc expression to the transition zone. Cno in a complex with Notch at the AJs of NE cells somehow stabilizes Notch at the membrane, favoring the binding of its ligand Ser present in the adjacent glia. Dl, which is enriched at the NE cells of the transition zone (Fig. 7A,A′), could preferentially activate the Notch target gene Gbe+Su(H)lacZ at this location (and in the emerging NBs). Indeed, the ectopic activation of Dl in all NE cells (Fig. 7B,B′) led to a concomitant ectopic activation of the Gbe+Su(H)lacZ reporter along with Dpn (100%; n = 12) (Fig. 7D,E), as well as to a repression of the m7-nuclaz reporter in NE cells (100%; n = 7) (Fig. 7F,G). Intriguingly, when Dl was ectopically activated in the whole surface glia, it was exclusively detected in the perineurial glia (Fig. 7C–C′″), being completely absent in the subperineurial glia where Ser accumulates strongly in contact with neuroepithelial Notch (Fig. 7A; see the Discussion).

Glial cells are key players during the development of the nervous system, and their number is indicative of nervous system complexity (Nave and Trapp, 2008). In Drosophila, larval glia are essential during the development of the brain, where they display multiple functions (Hartenstein, 2011). In this work, we provide evidence that glia play a key role during optic lobe development and proneural wave progression.

Cno and its vertebrate homologues AF-6/Afadin localize at epithelial AJs where they regulate the linkage of AJs to the actin cytoskeleton by binding both actin and nectin family proteins (Lorger and Moelling, 2006; Mandai et al., 1997; Matsuo et al., 1999; Sawyer et al., 2009; Takahashi et al., 1998). Here, we have found that Cno colocalizes with Notch at the AJs of NE cells in the optic lobe proliferation centers. Notch also colocalized with its ligand Ser; they accumulated strongly at the interface between NE cells and the surrounding glia. Our co-immunoprecipitation experiments indicate the formation of a Ser–Notch–Cno complex in vivo, and the mutant analysis shows the functional relevance of such a complex at the glia–neuroepithelium interface. The data presented in this study support the hypothesis that Cno stabilizes Notch at the AJs of NE cells, favoring the binding of Ser present in the adjacent glial cells. Indeed, in cno loss-of-function mutants, both Notch and Ser distribution is affected; this alteration is accompanied by an abnormally advanced proneural wave, a phenotype reminiscent of that shown by optic lobes with Notch loss of function (Yasugi et al., 2010), and also with Ser loss of function, as we show here. Activation of the Notch pathway is essential to maintain the integrity of the neuroepithelium and to allow the correct progression of the proneural wave (Egger et al., 2010; Wang et al., 2011; Weng et al., 2012). Our results show that glial Ser is responsible for such activation, promoting the expression of the m7-nuclacZ reporter in NE cells. In fact, the reduction of glial Ser either by knocking down epithelial cno or by expressing DNSer in the glia led to a decrease in the expression of the m7-nuclacZ reporter in NE cells and to an ectopic activation of the Ras–PntP1 pathway and of L′sc. We propose that this is ultimately the cause of the proneural wave advance observed in those genotypes. Thus, the activation of Notch in the neuroepithelium by glial Ser, in normal conditions, would be essential to avoid a premature activation of the EGFR–Ras–PntP1 pathway and hence of L′sc. Indeed, Notch has been shown to downregulate different EGFR–Ras signaling pathway components such as Rhomboid (Rho), required for the processing of the EGFR ligand Spitz, in other developmental contexts where both pathways are actively cross-talking (Carmena et al., 2002). Therefore, Notch activity in NE cells could be contributing to inhibit Rho, restricting its presence to the transition zone where Rho is very locally expressed (Yasugi et al., 2010).

We observed that, in wild-type conditions, Ser is present in all surface glia (perineurial and subperineurial), as shown by the expression of CD8::GFP (SerGal4>>UAS-CD8::GFP). Notch, as tested by different reporters, is active in this tissue and highly reduced with Ser loss of function in the glia. This suggests the existence of Ser–Notch-mediated intercellular communication between the glial cells that comprise both the perineurial and subperineurial glia. Intriguingly, the knockdown and overexpression of cno in NE cells also had a clear effect on Notch activity in the glia: a reduction and an increase, respectively. This is more challenging to explain. Because the cno loss of function in the NE led to a high reduction of both neuroepithelial Notch and glial Ser, the easiest explanation is that an ‘excess’ of unbound glial Ser is degraded and this impinges on the general thresholds of glial Ser, therefore causing a general reduction in the Notch activity in this tissue. This will be an interesting area to explore in detail and we leave the question open for future investigation.

The activity of Notch in the neuroepithelium and in medulla NBs seems controversial. For example, Notch has been shown to be active in the neuroepithelium at low or null levels (Wang et al., 2011; Weng et al., 2012; Yasugi et al., 2010) or in a ‘salt and pepper’ pattern (Egger et al., 2011). Weak or no activity of Notch in NBs has also been reported (Egger et al., 2010; Yasugi et al., 2010), as well as a high activation (Ngo et al., 2010; Weng et al., 2012). One possibility to conciliate all these results and apparently contradictory data is that different Notch target genes used as Notch activity reporters are, in fact, differentially activated in particular regions or tissues. Our results support this proposal. Here, we have used four different Notch reporters, Gbe+Su(H)lacZ, E(spl)mβ-CD2, E(spl)mδ-lacZ and E(spl)m7-nuclacZ. Whereas m7-nuclacZ was expressed throughout the neuroepithelium, Gbe+Su(H)lacZ was restricted to the transition zone, although both were expressed in medulla NBs along with mβ-CD2. In addition, mβ-CD2 was strongly activated in the glia, whereas the Gbe+Su(H)lacZ and the mδ-lacZ reporters were expressed at much lower levels at this location. Differential activation of Notch target genes has been previously reported and tissue-specific factors could contribute to this differential expression (Cooper et al., 2000; de Celis et al., 1996; Wech et al., 1999). This is an intriguing scenario to analyze in the future. The in-depth analysis of other Notch reporter genes in the developing optic lobe would contribute to clarify this issue.

At third larval instar during optic lobe development, Dl is highly restricted to two to three cells at the transition zone in the neuroepithelium, where Dl activates Notch (Ngo et al., 2010; Weng et al., 2012; Yasugi et al., 2010). In this work, we have found that the other ligand of Notch, Ser, is expressed in the surrounding glia at this larval stage and it is strongly accumulated at the interface with NE cells. Ser activates Notch in the neuroepithelium and this, in turn, would contribute to restrict the activation of the Ras–PntP1 pathway and L′sc to the transition zone. Intriguingly, we have observed that Ser preferentially activates the Notch target gene m7-nuclacZ in the neuroepithelium, whereas Dl activates other Notch target genes, including Gbe+Su(H)lacZ, in the transition zone. For example, the overexpression of Dl in NE cells caused an ectopic activation of Gbe+Su(H)lacZ throughout the neuroepithelium, along with dpn, which also behaves as a Notch target in other systems (Krejcí et al., 2009), and a repression of m7-nuclacZ (Fig. 7D–G). In addition, the loss of function of Ser in the glia caused a striking decrease in the expression of m7-nuclacZ in the neuroepithelium (Fig. 5O–P′). One possibility to explain these observations is that the pool of Notch associated with the AJs and activated by glial Ser is subject to particular post-translational modifications or/and is associated with other AJ proteins (including Cno) that somehow make Notch more receptive to Ser and able to activate specific target genes (i.e. m7). In this regard, it is interesting to note that Dl ectopically expressed in the glia (i.e. repoGal4>>UAS-Dl) was not detected at the interface with NE cells, where glial Ser is strongly localized in contact with Notch (Fig. 4E–E″), but Dl was restricted to the outermost surface glia (perineurial glia) (Fig. 7C–C′″). This result strongly indicates that Dl cannot bind or has very low affinity for this pool of Notch at the AJs, hence it is actively degraded in the subperineurial glia. The low affinity of Dl for Notch at this location further suggests that this pool of Notch at the AJs must be endowed with particular characteristics, as mentioned above, that ultimately could alter the activity properties of Su(H), explaining in turn the distinct expression pattern of Notch target genes. Another possibility, which is not necessarily exclusive, to explain the differential activation of the Notch reporters is that they respond to different Notch thresholds. For example, m7-nuclacZ could require very low levels of Notch activation, whereas Gbe+Su(H)lacZ might require high amounts of Notch signaling in NE cells. All these questions remain open for further investigation.

All stocks used in this study are from the Bloomington Stock Center and the Vienna Drosophila RNAi Center, unless otherwise noted: Gbe+Su(H)lacZ [41 bp of the E(spl)m8 promoter that contains Grainyhead binding sites plus Su(H) binding sites] (Furriols and Bray, 2001), E(spl)mβ-CD2 (de Celis et al., 1998), E(spl)mδ-lacZ (Cooper et al., 2000), E(spl)m7-nuclacZ (a gift from S. Bray) (Pines et al., 2010), FRT82B cno^R2 (Sawyer et al., 2009), hsFLP, FRT82B-tubGal80, arm-Gal4, c855a-Gal4 (Manseau et al., 1997), FRT82B Ser^RX82 (Thomas et al., 1991), repo-Gal4, nrv2-Gal4, Ser-Gal4, UAS-mCD8::GFP, UAS-cnoRNAi, UAS-cno::GFP (Slováková and Carmena, 2011), UAS-Dicer2, UAS-BdE24 or Ser^sec (DNSer, strong expressor that produces the first 1020 amino acids, up to the first BamHI site in Serrate; a gift from Robert Fleming) (Fleming et al., 2013). UAS-Ni, UAS-NDN (Rebay et al., 1993), UAS-Notch::GFP (Kawahashi and Hayashi, 2010), UAS-Dl, UAS-GFP. The crosses Gal4xUAS were carried out at 29°C with the following exception: the cross Gal4xUAS-DNSer was also carried out at less-stringent conditions, specifically at 25°C followed by an incubation overnight at 29°C, for analyzing the expression of L′sc and PntP1 in the neuroepithelium.

To generate clones of cells homozygote for the null allele cno^R2, hsFLP; UAS-CD8::GFP; FRT82B tubGal80 flies were crossed with armGal4; FRT82B cno^R2, identifying the clones by the presence of CD8::GFP. hsFLP was activated for 2 hours at 37°C in first- and second-instar larvae. Clones of the null allele Ser^RX82 were performed in a similar way using in this case the nrv2Gal4 driver to identify glial mutant clones.

Brains were dissected from third-instar larvae and fixed and stained with antibodies using standard protocols unless specified below. The following primary antibodies were used: rabbit anti-Cno 1:400 (Speicher et al., 2008); rat anti-L′sc 1:100 (Martín-Bermudo et al., 1991); guinea pig anti-Dpn 1:1000 (a gift from J. Skeath); rat anti-Baz 1:500 (Wodarz et al., 1999); mouse anti-Notch intracellular domain (C17.9C6) 1:50–1:100 [Developmental Studies Hybridoma Bank (DSHB)]; mouse anti-Notch extracellular domain (C458.2H) 1:50–1:100 (DSHB); mouse anti-Dl 1:40 (DSHB); rat anti-Shg 1:100 (DSHB); goat anti-Ser 1:20–1:100 (Santa Cruz); mouse anti-Dlg 1:100 (DSHB); mouse anti-Repo 1:100 (DSHB); mouse anti-Nrv5F7 1:600 (DSHB); rabbit anti-PH3 1:400 (Upstate); rabbit anti-PntP1 1:300 (a gift from J. Skeath); rabbit anti-βgal 1:3000 (Cappel); mouse anti-βgal 1:800 (Promega); mouse anti-CD2 1:600 (AbD Serotec). Secondary antibodies coupled to biotin (Vector Labs), Alexa Fluor 488, 546 or 633 (Molecular Probes) were used. For immunostaining with the anti-Cno antibody, brains were fixed using the heat and methanol method (Tepass, 1996). Fluorescent images were recorded using a Leica upright DM-SL microscope and assembled using Adobe Photoshop. Most of the micrographs shown in figures represent single sections from confocal Z-stacks, with the exception of some average projections in the following figures: Fig. 2A-D,D′ and Fig. 5E,E′,H,H′.

For in vivo Co-IPs, lysates were prepared from third-instar larval heads obtained from the following crosses: c855a-Gal4>>UAS-Notch::GFP, c855a-Gal4>>UAS-cno::GFP and c855a-Gal4>>UAS::GFP as a negative control. Sectioned heads were homogenized in lysis buffer [50 mM Tris-HCl, pH 8, 150 mM NaCl, 0.1% SDS, 1 mM EDTA, 1% Triton X-100, 1 mM NaF, 100 µM Na₃VO₄, 2 mM PMSF and complete protease inhibitors (Roche)]. Extracts were centrifuged for 2 minutes at 14,000 rpm (18,700 g) at 4°C. The supernatant was incubated with Rb polyclonal antibody to GFP Sepharose beads (Abcam) for 2 hours at 4°C. The beads then were washed three times with lysis buffer without inhibitors, resuspended in protein-set buffer (Fluka) and heated at 95°C for 3 minutes. Precipitates were resolved by SDS-PAGE and immunoblotted with mouse anti-GFP (Clontech), rabbit anti-Cno (affinity purified) or goat anti-Serrate (Santa Cruz). Each experiment was performed at least twice.

Measurement of Repo⁺ and PH3⁺ cells was made by imaging brain frontal sections every 3 μm to ensure that the same cell was not counted twice. The area analyzed was defined according to anatomical references, starting when the lobula plate neuropile appears (Ngo et al., 2010) and following in an antero-posterior direction until the disappearance of the outer neuroepithelium. 8–12 sections were sampled for each brain and only surface glia were counted for statistics. Parametric and non-parametric analyses were carried out with similar results to test differences in the proliferation of glial cells with respect to control brains; slight differences in statistical significance were found in the genotype UAS-Ser+UAS-Ni between the parametric (**) and the non-parametric (*) tests for the percentage of PH3⁺ cell number (Fig. 6H); GLM and N-PAR1WAY procedures were performed to compare variances between treatments and Duncan test was used for comparison of averages using SAS v9.2.

We thank Sarah Bray, Jim Skeath, Bob Fleming, Ben Ohlstein, Ibo Galindo, Mark Peifer, Andreas Wodarz, Yuh N. Jan, Utpal Banerjee, the Bloomington Drosophila Stock Center at the University of Indiana, the Vienna Drosophila RNAi Center and the Developmental Studies Hybridoma Bank at the University of Iowa for kindly providing fly strains and antibodies.