The specification of distinct neural cell types in central nervous system development crucially depends on positional cues conferred to neural stem cells in the neuroectoderm. Here, we investigate the regulation and function of the epidermal growth factor receptor (EGFR) signalling pathway in early development of the Drosophila brain. We find that localized EGFR signalling in the brain neuroectoderm relies on a neuromere-specific deployment of activating (Spitz, Vein) and inhibiting (Argos) ligands. Activated EGFR controls the spatially restricted expression of all dorsoventral (DV) patterning genes in a gene- and neuromere-specific manner. Further, we reveal a novel role of DV genes—ventral nervous system defective (vnd), intermediate neuroblast defective (ind), Nkx6—in regulating the expression of vein and argos, which feed back on EGFR, indicating that EGFR signalling stands not strictly atop the DV patterning genes. Within this network of genetic interactions, Vnd acts as a positive EGFR feedback regulator. Further, we show that EGFR signalling becomes dependent on single-minded-expressing midline cells in the posterior brain (tritocerebrum), but remains midline-independent in the anterior brain (deuto- and protocerebrum). Finally, we demonstrate that activated EGFR controls the proper formation of brain neuroblasts by regulating the number, survival and proneural gene expression of neuroectodermal progenitor cells. These data demonstrate that EGFR signalling is crucially important for patterning and early neurogenesis of the brain.
The central nervous system in insects and mammals arises from multipotent neural stem cells that give rise to a vast array of distinct cell types. The underlying molecular genetic mechanisms have been extensively studied in the developing embryonic truncal nervous system (reviewed in [1,2]). In Drosophila, the ventral nerve cord (VNC) is generated by segmental arrays of neural stem cells, called neuroblasts, which delaminate from the truncal neuroectoderm (NE). Each neuroblast acquires a unique identity that is reflected by the typical developmental time point and position, and the combination of developmental control genes it expresses , which finally determines the number and types of progeny it generates [4,5]. The identity of each neuroblast is initially determined by the combinatorial code of positional cues in the NE, provided by the products of early patterning genes (reviewed in ).
A group of genes essentially involved in patterning of the VNC is the evolutionary conserved dorsoventral (DV) genes. Their activity subdivides the NE along the DV axis into longitudinal columns: ventral nervous system defective (vnd) is expressed in the ventral, intermediate neuroblasts defective (ind) in the intermediate and muscle segment homeobox (msh; Drop [Dr], FlyBase) in the dorsal neuroectodermal column, where they control the formation of neuroblasts (except msh) and specify aspects of their fate [6–16]. The expression domains of DV genes and, accordingly, the DV boundaries of the NE are regulated by the graded activity of the nuclear factor Dorsal and Bone morphogenetic protein [17–19]. Another signalling pathway that controls the regionalized expression of DV genes is the epidermal growth factor receptor (EGFR) pathway [17,20,21], which is highly conserved from fly to human (reviewed in [22,23]). The EGFR pathway is required to induce ind expression and to maintain vnd expression [17,24,25], and thus to control the formation of intermediate and the identity of ventral neuroblasts [20,24]. Localized EGFR activation depends on the neuregulin-like ligand Vein (Vn) and the TGF-α homologue Spitz (Spi) [26–28]. Spi is processed and secreted by the combined activity of the transmembrane protease Rhomboid (Rho) and the chaperone Star (S), and serves as the cardinal activating ligand [28–31]. Expression of rho is tightly controlled and represents the key to the dynamic activation of EGFR, whereas the inactive Spi precursor and S are rather broadly expressed in the NE [32–35]. The spatio-temporal pattern of EGFR activity also depends on the inhibiting ligand Argos (Aos), which is induced in response to high levels of EGFR activity , and antagonizes EGFR activation by sequestering Spi [27,37–39]. Although EGFR signalling is initially induced by ligands (Spi, Vn) expressed and secreted from the ventral NE [20,26,27,40], by gastrulation EGFR signalling becomes dependent on Spi, which is secreted from the ventral midline [20,35,41,42].
Similar to the situation in the trunk, it has been shown that EGFR signalling is required in midline cells of the embryonic head, which follow a particular mode of neurogenesis to give rise to the larval visual system, stomatogastric nervous system and most medial parts of the brain [43–46]. However, the regulation of EGFR signalling and its role in neuroectodermal patterning and specification of cell fate along the DV axis are not well understood in the early embryonic brain. The NE of the embryonic brain gives rise to an array of about 100 neuroblasts in each hemisphere, which can be subdivided (from anterior to posterior) into the presumptive proto- (PC), deuto- (DC) and tritocerebrum (TC) [47,48]. Based on a distinct combination of regulatory genes expressed, each brain neuroblast acquires a unique identity , which suggests that the expression of patterning genes in the overlying NE has to be precisely controlled during neuroblast formation. In previous reports, we showed that the regionalized expression of DV genes exhibits neuromere-specific differences in the NE and neuroblasts of the embryonic brain [50,51]. We then uncovered a genetic network in which evolutionarily conserved factors encoded by DV genes (vnd, ind, msh, Nkx6) and AP patterning genes (empty spiracles, engrailed) closely interact to properly pattern the NE and specify neuroblast identity along the DV axis of the brain [52,53].
In this study, we shed light on the regulation and function of EGFR signalling in early embryonic brain development. We show that a neuromere-specific deployment of activating (Spi, Vn) and inhibiting (Aos) ligands controls the localized activation of EGFR in the brain NE, which, in turn, is necessary for the spatially restricted expression of all DV genes. We also show that DV genes (vnd, Nkx6, ind) regulate the expression of vn and aos, which indicates that within the genetic network EGFR stands not strictly atop the DV genes. Moreover, Single-minded (Sim), a master regulator of CNS midline cells , is needed for EGFR signalling specifically in TC, but not in the anterior brain (DC, PC). Finally, we show that activated EGFR promotes the formation of brain neuroblasts, as it controls the number, survival and proneural gene expression of neuroectodermal progenitor cells. Thus, EGFR signalling plays a central role in DV patterning and early neurogenesis of the fly brain.
2.1. The pattern of EGFR activity compared with DV gene expression in the early brain
To test whether EGFR controls DV patterning in the early brain, we first compared the activation pattern of EGFR with the known expression patterns of the DV genes vnd, ind, Nkx6 and msh [52,53]. We visualized EGFR activity in the brain NE by using an antibody against double-phosphorylated (activated) MAPK . Additionally, the segmental marker Engrailed (En) was used to delineate the borders between brain neuromeres (according to [48,50]). MAPK is initiated at stage 5 in a broad longitudinal stripe (data not shown), which by stage 6 precisely overlaps with expression of vnd, and includes expression of ind and Nkx6 (figure 1a–d,j). Thus, EGFR signalling in the presumptive TC and DC is confined to the ventral and intermediate NE (and ventrally adjacent mesectoderm). In the presumptive PC, MAPK overlaps with Vnd in the ventral NE, but is additionally detected in two large stripes in the intermediate/dorsal NE (the ‘anterior stripe’ and ‘posterior stripe’; according to ). By stage 9, MAPK is kept at strong levels in PC, DC and ventral TC, but is largely vanished from the intermediate TC (figure 1e–h,k). Interestingly, ind expression, which depends on EGFR in the developing VNC , is not initiated in the intermediate TC before MAPK has vanished (figure 1g), unlike in PC and DC (figure 1c). From early stage 10 onwards, MAPK becomes confined to smaller cell clusters in the brain NE (figure 1i). During stages 5–11, MAPK remains complementarily expressed to msh (figure 1j,k; electronic supplementary material, figure S1a,b). EGFR is therefore not active in the dorsal NE of TC and DC. During stages 8–11, MAPK is also transiently detected in subsets of neuroblasts that mostly develop from the MAPK-positive brain NE (electronic supplementary material, figure S2a,b).
2.2. EGFR is required for neuromere-specific expression of ind, vnd and Nkx6
Because EGFR activity overlaps with the expression domains of vnd, Nkx6 and ind in the brain NE, we tested whether EGFR controls the expression of these genes. Analysing EGFRf2 mutants, we found that ind expression is delayed and strongly reduced in DC and PC (87% and 100% of brain hemispheres, respectively; n = 35; figure 2a,b), or entirely missing in DC (13% of hemispheres; n = 35). ind levels in TC of EGFRf2 mutants seemed to be unaffected, being consistent with ind expression complementary to MAPK in wild-type embryos. Ectopic activation of EGFR by overexpression of the secreted (i.e. active) EGFR ligand Spitz (sSpi) using the maternal Matα-Gal4 line (which drives expression ubiquitously; in the following termed Matα > sSpi)  led to ectopic activation of ind in intermediate/dorsal PC and dorsal TC (100% and 25% of hemispheres, respectively; n = 32), while ind was not affected in DC (figure 2c,d). Thus, EGFR signal is necessary for ind expression in DC and PC, and sufficient to activate ind expression in TC and PC.
Expression of vnd expression was activated normally in EGFRf2 mutants, but strongly reduced at early stage 10 in the NE of TC, DC (100% and 83% of hemispheres, respectively; n = 25) and, to a lower extent, PC (83% of hemispheres; n = 25; figure 2e,f). To ectopically activate EGFR, we misexpressed sSpi using the scabrous (sca)-Gal4 line  which drives expression within the NE stronger than the maternal driver used in this study. Because sca-Gal4 induces misexpression not before stage 7, thus later than the maternal driver, we usually took this driver to investigate effects at later stages (stages 11–13). In these sca > sSpi embryos, vnd was ectopically expressed in TC and DC, but not in PC (88% and 79% of hemispheres, respectively; n = 48; figure 2g,h). We conclude that EGFR is necessary for the maintenance and sufficient for induction of vnd expression especially in TC and DC. As with vnd, we observed similar effects of EGFR on the expression of Nkx6 in TC and DC (figure 2i–l). However, we did not observe effects of EGFR on the expression of scarecrow, another Nkx gene with close homology to vnd  (data not shown).
In TC and DC, we furthermore observed a slight derepression of msh in EGFRf2 mutants, and conversely, a reduction of msh expression in Matα > sSpi embryos (electronic supplementary material, figure S3a–d). To test whether EGFR signal directly represses msh, we analysed patterns of msh and MAPK (which normally exclude each other) in vnd6 mutants; in these mutants, msh was derepressed in the intermediate/ventral NE, and MAPK remained unaffected, both factors now being largely coexpressed in this NE (electronic supplementary material, figure S3e,f). These results suggest that activated EGFR regulates msh indirectly, through the activity of other DV genes vnd, ind and Nkx6 (as shown above).
2.3. EGFR induces ind expression in DC by phosphorylation of the co-repressor Groucho
vnd and ind are exceptionally coexpressed in the intermediate DC  (figure 2m), where EGFR is also activated. EGFR can regulate gene expression by MAPK-dependent phosphorylation and thus inactivation of the co-repressor Groucho (Gro) [58–60]. As Vnd is a Gro-dependent repressor , we hypothesized that EGFR signal phosphorylates Gro and thus inactivates Vnd/Gro-mediated repression of ind. To test this, we analysed ind expression after ectopically expressing unphosphorylatable (i.e. constitutively active; GroAA) and pseudo-phosphorylated (inactive; GroDD) Gro-constructs  in the brain NE using the sca-Gal4 and maternal NGT40-Gal4  driver lines, which both drive Gal4 expression in the NE (although NGT40-Gal4 induces misexpression earlier and at weaker levels than sca-Gal4; data not shown). ind was significantly reduced in DC of sca > GroAA and NGT40 > GroAA embryos (figure 2n,p; data not shown), with effects being stronger with the later-initiating but stronger driver sca-Gal4, but unaltered in DC of sca > GroDD control embryos (figure 2o,p). Altogether, our results strongly suggest, that EGFR activity induces ind expression in DC specifically by phosphorylation of Gro, thus inactivating Vnd/Gro.
2.4. rho, vn and S are differentially expressed in the brain neuroectoderm and flanking mesectoderm
The complex spatio-temporal pattern of EGFR activation in the brain NE led us to determine the sources of activating Spi and Vn ligands. Spi is expressed as an inactive precursor and requires Rho and S to be secreted [28–31]. spi itself and S are rather broadly expressed, thus playing a minor role in controlling the spatio-temporal pattern of Spi secretion [32–35]. By contrast, rho expression is tightly controlled and represents the key regulatory step controlling Spi secretion in the developing VNC [35,40]. As we cannot exclude that the spatio-temporal Spi secretion is also regulated by spi and S expression in the early brain, we analysed rho, spi and S expression to determine Spi ligand sources in the brain NE. At stage 5, rho expression in the NE corresponded with the MAPK pattern, but became largely restricted to the sim-expressing mesectoderm by gastrulation (at stage 6), although rudimentary rho expression was still detectable in the NE of DC and PC (electronic supplementary material, figure S4a,b). By stage 9, when EGFR signalling and DV regionalization of the brain NE are most pronounced (figure 1) , rho expression was weakly detected in DC and ventral PC, and the anterior MAPK stripe of PC; stronger rho expression was observed in the posterior MAPK stripe and mesectoderm (ventral to TC and DC) (figure 3a,b). Thus, rho expression closely correlates with EGFR activity, even though rho expression levels vary within the brain NE and flanking mesectoderm. S and spi were broadly expressed in the brain NE (figure 3c; and data not shown). We noted that S expression was particularly strong in the anterior MAPK stripe of PC, where rho expression was weak, despite strong EGFR signalling (figure 3a–c). In contrast with Spi, Vn is expressed as a secreted protein which requires no further processing to be activated . vn expression was codetected with MAPK in DC, ventral PC and mesectoderm (ventral to TC and DC) at stage 9 (figure 3d; electronic supplementary material, figure S4c). Because vn expression was never observed in both MAPK stripes of PC, this suggests that Vn only partially acts as a positive EGFR feedback regulator in the brain, in contrast with the VNC .
In sum, rho, vn and S are differentially expressed during patterning of the brain (summarized in figure 3e). Primarily vn and little rho is expressed in DC and abutting ventral PC. By contrast, rho and S are expressed in the two protocerebral MAPK stripes, with weak rho and strong S expression levels in the anterior stripe, opposite to the posterior stripe. Only S, but not vn or rho, is expressed in the TC, although the adjacent mesectoderm reveals strong vn/rho expression levels. These data suggest that EGFR activity in the brain is regulated by a region-specific deployment of distinct ligands.
2.5. rho and vn are differentially required to activate EGFR in the brain
To test for requirements of Rho and Vn to activate EGFR, we analysed EGFR activity in the brain NE of rhoPΔ5 and vnRG436 mutant embryos. At stage 9, vnRG436 mutants exhibited a specific loss of MAPK only in DC and ventral PC (strong reduction in 40%, moderate reduction in 49%, normal in 11% of hemispheres; n = 35), whereas rhoPΔ5 mutants showed a near total loss of MAPK in the entire brain NE (100% of hemispheres; n = 22) (figure 3f,g). We conclude that Rho is required for EGFR activation in the entire brain NE, whereas Vn is required in addition to Rho for proper EGFR activation in DC and ventral PC. As MAPK was lost in DC/ventral PC in both, rho and vn mutants, we asked if vn expression depends on Rho. Indeed, vn expression was reduced in these brain regions in rhoPΔ5 mutants (electronic supplementary material, figure S4d), indicating that Rho-dependent factors normally induce vn expression, which in turn activates EGFR. However, we cannot rule out that low levels of Rho act in parallel to Vn to achieve proper EGFR activation in DC and ventral PC.
2.6. vn expression is controlled by Vnd in the brain
We found that expression of vn closely correlates with expression of vnd in the early brain NE (figure 4a–c). Therefore, we tested if Vnd controls vn expression. At stage 9, vnd6-mutant embryos showed strong reduction of vn expression in the brain NE (figure 4d). Correspondingly, we found ectopic vn expression in DC and PC upon vnd-overexpression (NGT40 > vnd) (figure 4e). These data suggest that Vnd is necessary, and to some extent sufficient, to activate vn expression. Next, we tested if vn expression is induced by activated EGFR, as it has been proposed in the VNC . At stage 9, we observed a slight reduction of vn expression in the brain NE of EGFRf2 mutants (figure 4f), compared with the strong reduction in vnd6 mutants. Further, in EGFRf2 mutants, the reduction of vn expression was closely correlated with the reduction of vnd expression (as shown above); residual vn expression was always co-detected with residual vnd expression (electronic supplementary material, figure S4e). This suggests that in the brain, EGFR induces vn expression indirectly, via Vnd. Our data thus provide the first evidence that Vnd, a DV gene which has been considered as an EGFR target so far, regulates the expression of an EGFR ligand.
2.7. aos is expressed in dorsal TC independently of EGFR but under control of Vnd, Ind and Nkx6, and inhibits EGFR signalling
Argos (Aos) is a secreted EGFR regulator that antagonizes Spi activity . We detected aos expression within the neuroectodermal MAPK domains in all brain neuromeres (figure 4g), which was expected because aos is known as a negative EGFR feedback regulator induced downstream of EGFR/MAPK . Consistently, aosΔ7 mutants showed a massive overactivation of MAPK in the brain NE at stage 11, when endogenous EGFR activity is largely downregulated (electronic supplementary material, figure S4f,g). Interestingly, we recognized a prominent aos expression domain outside the MAPK pattern, in dorsal TC (figure 4g; electronic supplementary material, figure S4 h,i). Accordingly, this aos domain was not affected in EGFRf2 embryos, whereas the remaining aos expression was entirely missing in the brain (electronic supplementary material, figure S4j,k). We conclude that aos is regulated independently of EGFR in dorsal TC.
Expression of aos restricted to dorsal TC suggests a regulation by DV genes. Therefore, we tested whether vnd, ind or Nkx6, expressed in ventral/intermediate TC (see §2.1.), act to restrict aos to the dorsal TC. In vnd6 mutants, which are characterized by an additional loss of ind and Nkx6 in the TC  (electronic supplementary material, figure S4l,m), aos is derepressed in the ventral/intermediate TC (figure 4h). Conversely, aos is efficiently repressed in NGT40 > vnd embryos (lost in 87%, strongly reduced in 13% of hemispheres; n = 30) (figure 4i). We did not find substantial effects on aos expression in ind or Nkx6 mutants (data not shown), most probably because both factors act redundantly . However, overexpression of either ind or Nkx6 with NGT40-Gal4 led to a strong reduction of aos expression in dorsal TC (figure 4j,k). We conclude that vnd, ind and Nkx6 act in concert to restrict aos to dorsal TC. This provides further evidence of DV patterning genes regulating the regionalized production of EGFR ligands.
Being controlled by the DV gene network, we asked whether Aos is involved in regulating EGFR activity along the DV axis in TC. At stage 9, EGFR is activated only in ventral TC. However, in aosΔ7 mutants, EGFR was additionally activated in intermediate TC (figure 4l–l″), indicating that Aos normally inhibits EGFR signalling in this NE. Thus, Aos is crucial for regulation of EGFR activity along the DV axis in TC.
2.8. Vnd is part of a positive feedback loop controlling EGFR activity in the brain
As Vnd controls the activity of both activating EGFR ligands, Vn and Spi (via aos, as shown above), we analysed the effect of Vnd on EGFR activity in the brain NE. At stage 9, vnd6 mutants exhibited a reduction of MAPK in TC, DC and adjacent ventral PC (figure 4h). In TC, this reduction is probably due to the ventrally expanded domain of aos expression which inhibits Spi-induced EGFR activation. In the vnd6-mutant DC/ventral PC, elevated Aos levels, secreted from the enlarged aos domain in TC, come together with the loss of Vn (figure 4d,h), which we propose to mutually account for the strong reduction of EGFR activation. To test whether Vnd is sufficient to induce EGFR, we analysed MAPK after vnd-overexpression. In NGT40 > vnd embryos, we found ectopic MAPK in PC and dorsal DC (figure 4i). We assume that EGFR is overactivated in PC primarily due to ectopic Vn (figure 4e), and in DC due to elevated levels of Vn and Spi, the latter owing to the loss of Aos in dorsal TC (figure 4e,i).
Taken together, these data demonstrate that Vnd plays a central role for the spatial activity of EGFR as it controls the ligand activity of Vn and Spi (figure 4m). As the maintenance of vnd expression in turn depends on activated EGFR (as shown in §2.2.), Vnd acts as a positive EGFR feedback regulator in the brain.
2.9. EGFR signalling crucially depends on Tll in PC, and on mesectodermal Sim in TC
As EGFR activity in PC emerges largely in the tailless (tll) expression domain (figure 5a–c), we analysed if Tll regulates EGFR activity. tlll49 mutants exhibited a loss of rho in the intermediate/dorsal PC while tll-overexpression (NGT40 > tll) induced ectopic rho expression predominantly in ventral/intermediate TC and DC, but not in PC where endogenous Tll is expressed (figure 5d–f). Consistent with these results, we detected in corresponding brain regions a loss of MAPK in tlll49 mutants and ectopic MAPK in NGT40 > tll embryos (figure 5g–i; electronic supplementary material, figure S4n). Further, we observed a loss of vn expression in ventral PC of tlll49 mutants (figure 5j). As Vnd, necessary for vn expression (as shown in §2.6.), is also lost in these mutants (figure 5k,l), we conclude that in ventral PC, Tll normally induces vn expression via Vnd. Taken together, in PC, Tll induces production of both ligands, Vn (in ventral PC) and Spi (in intermediate/dorsal PC), by regulating the expression of vnd and rho, respectively (figure 5m).
Next, we analysed the role of single-minded (Sim), the main regulator of midline-dependent Spi secretion in VNC patterning [35,64,65], in controlling EGFR activity in the early brain. Onset of EGFR signalling occurs independently of Sim, because expression of rho initiates in the brain NE (at stage 5) before sim in the ventral mesectoderm (at stage 6) (electronic supplementary material, figure S4a,b). By stage 9, sim is expressed in the mesectoderm ventral to TC and DC (figure 5n), where it overlaps with rho and vn expression (figure 5o,p,q). Analysis of simH9 mutants revealed a loss of rho, vn and MAPK signal in this mesectoderm (figure 5r–u), accompanied by a loss of MAPK specifically in TC (figure 5t,u). This indicates that Sim is necessary for the production of Spi and Vn secreted from the mesectoderm to induce EGFR signalling in ventral TC, but is dispensable in DC and PC. Furthermore, sim overexpression (NGT40 > sim) was sufficient to induce ectopic MAPK in the brain NE by stage 11 (figure 5v), probably due to action of rho and vn, which both can be ectopically induced by Sim in the trunk . Altogether these data suggest that mesectodermal Sim is normally important for ongoing EGFR signalling exclusively in TC.
2.10. EGFR controls formation of brain neuroblasts by regulating number, survival and proneural gene expression of neuroectodermal progenitor cells
Having established the regulation of EGFR signalling and its interaction with DV patterning genes in the procephalic NE, we asked if EGFR signal functions in the formation of brain neuroblasts. In stainings against Death caspase-1 (Dcp-1), a hallmark of cell death, we recognized extensive cell death in the brain NE (figure 6a,b; electronic supplementary material, figure S5a–d), which is accompanied by a significant loss of Deadpan (Dpn)-positive brain neuroblasts in all three neuromeres of EGFRf2-mutant brains at stage 11, when compared with wild-type (figure 6a′,a″,b′,b″,c). However, we did not detect apoptotic brain neuroblasts, although EGFR is transiently active in a subset of them (electronic supplementary material, figure S2a,b). To test if the loss of neuroblasts is caused by cell death of NE progenitor cells, we estimated the neuroblast number in EGFRf2;Df(3L)H99 double mutants which are cell death deficient . In total, 20–25% of brain neuroblasts were lost in these double mutants, instead of 40% in EGFRf2 mutants (figure 6c–e′; electronic supplementary material, figure S5e,f), whereas neuroblast numbers were unaffected in Df(3L)H99 mutant control brains (electronic supplementary material, figure S5g,h). These data demonstrate that only a subfraction of 15–20% of brain neuroblasts is missing due to cell death of NE progenitor cells in EGFRf2 embryos.
Because during the period of neuroblast formation EGFR is active in the NE of ventral/intermediate DC and ventral TC, we analysed the number of ventral and intermediate neuroblasts in both neuromeres of EGFRf2;Df(3L)H99 mutants. Ventral neuroblasts were reduced in TC and DC, whereas intermediate (Ind-positive) neuroblasts were all formed (figure 6d,d′,e, e′,f; electronic supplementary material, figure S5e′,f′i). Comparing the small number of neuroblasts missing in TC and DC with the total amount of brain neuroblasts missing in EGFRf2 mutants (figure 6c,f; electronic supplementary material, figure S5i), we conclude that neuroblast formation is primarily affected in PC.
To ascertain if the failure in neuroblast formation in EGFRf2 mutants is due to a deregulation of proneural genes, we investigated expression of lethal of scute (l'sc), the key proneural factor for the development of brain neuroblasts [47,48]. In EGFRf2;Df(3 L)H99 mutants, l'sc expression is strongly reduced in regions where EGFR is normally active (figure 6g–j). Correspondingly, l'sc expression is significantly prolonged upon overexpression of activated Spi (sca > sSpi) (figure 6k,l). To test whether ectopic EGFR activation in those embryos is also sufficient to generate additional brain neuroblasts, we counted neuroblast numbers at stage 13 (until when sca-Gal4 is broadly active; data not shown). The number of Dpn-positive brain neuroblasts was significantly increased (figure 6m,n), but not the mitotic index of those neuroblasts: stainings against the M-phase marker Phospho-Histone3 (PH3) showed that 23% of neuroblasts were mitotic in wild-type and 25% in sca > sSpi (n = 10 hemispheres each). Thus, EGFR activation is sufficient to induce proneural gene expression, and the formation of ectopic brain neuroblasts, but does not enhance their proliferative activity.
As EGFR is activated in the brain NE from stage 5 onwards, we finally asked if early EGFR signalling already impacts the extent of the NE by controlling the early mitotic activity (see ), and thus the final number of NE progenitor cells from which brain neuroblasts develop. Therefore, we analysed PH3-labellings with focus on the ventral/central head NE where EGFR is largely activated, and observed that the amount of mitotic NE cells is reduced by approximately 20% in EGFRf2 mutants (n = 14 hemispheres) (electronic supplementary material, figure S5j–m). Nevertheless, this reduction in NE progenitor cell number alone does not account for the observed reduction in l'sc expression domains.
In sum, our data provide evidence that EGFR signalling affects formation of brain neuroblasts at multiple steps of development: first, by controlling mitosis in the NE anlagen to establish the proper number of neuroectodermal progenitor cells; second, by positively regulating expression of proneural gene l'sc in those cells; and third, by ensuring their survival (summarized in figure 6o).
3.1. Localized EGFR signalling in the embryonic brain is controlled by neuromere-specific deployment of distinct ligands
In the VNC, EGFR is activated in two phases. In the early, midline-independent phase, EGFR is induced by Rho (via processing of Spi) and Vn, which are both expressed in the ventral/intermediate NE; rho expression becomes restricted to the midline during gastrulation, while vn expression restricts towards the ventral NE [20,26,40,41]. In the following midline-dependent phase, EGFR activity essentially depends on Sim, a master regulator of midline development which induces rho expression (and Spi secretion) in the ventral midline and is required for vn expression in the ventral NE [35,64,65,68,69]. As summarized in figure 7, in the brain we observe a similar early period of EGFR activity before Sim expression is initiated in ventral mesectodermal cells (corresponding to the ventral midline; see also ). EGFR is initially induced by Rho and Vn expressed in the ventral/intermediate NE of TC, DC and ventral PC. However, only in the posterior brain, the TC, do both factors become confined towards the ‘ventral midline’, followed by a midline-dependent phase of EGFR signalling (figure 7). These findings support that patterning in the TC closely resembles the situation in the trunk, but is more derived in DC and PC where EGFR activity remains midline-independent, even though weakly Sim-positive midline cells ventral to DC are likely to secrete limited amounts of Spi and Vn. Our data show that in DC/PC, EGFR activity relies on neuroectodermal sources of Vn and Spi (processed by Rho) that are controlled by the DV gene vnd (as discussed below) and the terminal gap gene tll. Surprisingly, Vn, which plays only a minor role in VNC patterning [20,26,42,70], proved to be a major activating EGFR ligand in DC and ventral PC. The reason is probably the exceptionally low level of rho expression in this area, leading to low Spi-levels that alone cannot sufficiently activate EGFR. By contrast, Rho-dependent Spi is the only ligand in the largest part of PC (i.e. intermediate/dorsal PC). Despite a close spatial correlation, however, rho and MAPK signal levels correlate poorly in PC: a strong MAPK signal is detected in the anterior stripe despite very low rho-expression levels, whereas MAPK is weaker in the posterior stripe despite stronger rho expression levels. We noted that S expression levels correlate with MAPK activity, being strong in the anterior MAPK stripe, and weaker in the posterior MAPK stripe. S is responsible for transport of the Spi precursor to the Golgi, where Rho-dependent secretion of Spi occurs [30,31]. Rho also cleaves and inactivates S, thus compromising the levels of secreted Spi . Therefore it seems likely that S together with Rho modulate localized levels of secreted Spi, and hence the activity of EGFR within distinct regions of the brain NE.
The nuclear Dorsal gradient is active in the early embryo to activate expression of rho, vnd and presumably vn in the truncal NE (each being expressed in a uniform longitudinal domain in the ventral NE) [9,72–74]. As the early expression domains of rho, vnd and vn cover in addition the ventral NE of the presumptive TC, DC and PC (figure 4a,b; electronic supplementary material, figure S4a), we assume that the early expression of these genes is likewise activated by the nuclear Dorsal gradient. In dorsal/intermediate PC, however, the induction of the slightly later emerging two stripes of rho expression might rather depend only on Tll (figure 7). The positive regulation of EGFR ligands by Tll in PC seems to be opposed to its function in the development of the larval visual system, where Tll has been proposed to block transcriptional programmes induced by EGFR (via Spi) .
Thus, during early brain development diverse ligands produced in distinct tissues (mesectoderm, NE) control localized activity of EGFR in a neuromere-specific manner (figure 7): in TC depending on midline-specific production of Vn and Spi (and Aos secreted from the dorsal TC, as discussed below), and in DC and PC primarily on Vn and Spi (via Rho), respectively, secreted from defined neuroectodermal domains. Other factors are known to control EGFR signalling, such as the feedback regulators Sprouty and Kekkon [75–78], MAPK phosphatase MKP-3 [79,80], or extracellular regulators of ligand travelling/activity such as Sulf1 and CG4096 . Except Sulf1 and CG4096, which we did not detect in the embryonic brain NE (data not shown), the potential role of the other regulators in patterning of the embryonic brain has to be clarified in further investigations.
3.2. DV genes control activity of EGFR through regulation of vn and aos
In this study, we demonstrate a novel role for Vnd, Nkx6 and Ind in regulating the ligands of EGFR, which indicates that within the hierarchical gene network EGFR stands not strictly atop the DV genes (figure 7). The function of Vnd in this context is far-reaching, because Vnd induces vn expression in all brain neuromeres; this is of importance particularly in DC where Vn is the main EGFR ligand. In the VNC, vn expression is induced by the transcriptional activator Pointed P1 in response to EGFR signal  (figure 7). We cannot exclude that during posterior brain development Vnd induces vn expression by positively regulating Pointed P1.
Notably, Vnd also regulates the expression of the inhibitory ligand Aos in TC. Vnd suppresses aos expression in the ventral (and early intermediate) TC, and keeps Aos secretion limited to the dorsal TC (figure 7). Accordingly, Spi elicited from the ventral midline is able to activate EGFR only in ventral TC, where it maintains vnd expression. Vnd is also required for inducing expression of Nkx6 and Ind , which both keep aos suppressed in the intermediate TC after Vnd is downregulated.
Thus, via the deployment of EGFR ligands (Vn, Aos), Vnd acts positively on itself, and thus stabilizes at later stages the cross-repressive interactions between Msh (dorsal)/Nkx6 (intermediate) and Ind (intermediate)/Vnd (ventral), essential for establishing the boundaries of DV neuroectodermal and corresponding stem cell domains [52,53,82].
3.3. EGFR regulates expression of DV genes in the embryonic brain in a neuromere-specific manner
Subdivision of the NE into discrete gene expression domains is essential for the correct specification of neural stem cells. During DV patterning of the truncal NE, EGFR is necessary for regionalized expression of vnd and ind (summarized in ). In previous reports, we uncovered a network of genetic interactions underlying DV patterning in the brain (including vnd, ind, msh, Nkx6, Ems, En) [51–53]. Here, we expand on this knowledge and show that EGFR strongly participates in the control of DV gene expression in the early brain. EGFR signal is necessary for the maintenance of the expression of two Nkx genes, Nkx6 and vnd, similar to its role for vnd expression in the VNC . It is likely that EGFR signal regulates the regionalized expression of other patterning genes. For example, ectopic En is detected in a few NE cells in EGFRf2-mutant DC (figure 2a,b), making it likely that EGFR is involved in the control of en in this neuromere. As we previously showed that En negatively regulates the expression of ind , possibly the reduction of ind in the EGFRf2-mutant DC is partly due to ectopic En.
EGFR signal is necessary for ind activation in the trunk , whereas its effect on ind expression in the brain strongly differs between neuromeres (figure 7). In TC, ind expression is activated independently of EGFR, which is unique in the entire embryo. However, in aos mutants in which EGFR is ectopically activated in the intermediate TC (figure 4l′′), we observed a significant reduction of ind expression (data not shown). Thus, we propose that in TC, EGFR controls ind expression indirectly via the maintenance of the ind-repressor Vnd which in wild-type fades early in the intermediate TC to allow for ind expression.
In DC, ind activation requires both, EGFR signal and phosphorylation of the co-repressor Gro. Gro has been shown to be directly phosphorylated (and thereby inactivated) by EGFR/MAPK activity (and other RTK/MAPK pathways) , thereby regulating ind expression in the trunk . Given that Ind is co-expressed with its Gro-dependent repressor Vnd , our results strongly suggest that EGFR allows ind expression in DC by phosphorylation of Gro, thus inactivating Vnd/Gro repressor complexes. We noted that in the ventral NE of the trunk and TC, unlike in DC, Vnd manages to repress ind expression despite EGFR being active and Gro being phosphorylated. As Vnd has been shown to form multiple complexes in the embryo , a possible explanation is that Vnd associates with RTK-insensitive co-repressors in trunk segments, and exclusively with Gro in DC (thus being sensitive to Gro inactivation). In PC, EGFR is also necessary for ind activation, but not sensitive to Gro-phosphorylation (data not shown), indicating an EGFR-dependent regulatory mechanism different from VNC and other brain neuromeres.
3.4. EGFR controls the formation of brain neuroblasts at different developmental steps
About 40% of brain neuroblasts are missing in EGFRf2-mutant embryos at embryonic stage 11. Our data suggest that EGFR function affects the formation of brain neuroblasts at multiple steps of development. First, activated EGFR positively controls the early mitotic activity within the neuroectodermal anlagen, in accordance with EGFR function in other developmental contexts (e.g. [85–88]). Second, as many neuroectodermal progenitor cells undergo premature cell death in EGFRf2 mutants, EGFR signalling is critical for their survival (see also [43,45]). EGFR-dependent survival has been reported also for midline glial cells in which Spi-activated EGFR suppresses the proapoptotic protein Hid [89,90]. Also in the large dorsal/intermediate PC, where apoptosis of EGFRf2-mutant neuroectodermal cells is substantial, Spi seems to be the only EGFR ligand, raising the possibility that a similar mechanism regulates neuroectodermal cell survival. We identified a small number of brain neuroblasts with activated EGFR, but never apoptotic neuroblasts in EGFRf2 mutants, suggesting that EGFR signal is rather dispensable for their survival. Third, in addition to the decrease of neuroectodermal progenitor cells, impairment of neuroblast formation in all brain neuromeres of EGFRf2 mutants is due to the loss of proneural gene expression; l'sc needs EGFR signal to be properly activated, in compliance with findings in the larval optic lobe . In accordance with the distribution of activated EGFR in wild-type, in cell-death-deficient EGFRf2;Df(3L)H99 mutants, we found a large population of neuroblasts to be missing at all DV positions in PC, and specifically ventral neuroblasts in the posterior brain (TC/DC). By contrast, in the EGFR-mutant VNC intermediate neuroblasts do not develop, whereas ventral neuroblasts usually form but are often misspecified . In VNC, EGFR promotes the formation of intermediate neuroblasts by activating ind expression [17,20]. At least in PC, it is likely that activity of Ind, in addition to L'sc, both of which are strongly reduced in EGFRf2 mutants, control the development of a small subset of protocerebral neuroblasts. Even though EGFR is active, in DC it does not impact the formation of intermediate neuroblasts; as these neuroblasts, opposed to the NE, express ind in EGFRf2 mutants, this further suggests that EGFR activity is dispensable for ind expression in these progenitors. In TC, onset of ind expression is delayed and regulated independently of EGFR (in NE and neuroblasts), explaining that intermediate neuroblasts are unaffected in the EGFRf2-mutant TC. However, in the mutant TC, we recognized an almost entire loss of ventral neuroblasts, which develop late. Moreover, L'sc was largely lacking there, and Vnd dissipated in the NE before these neuroblasts normally develop. Thus, it is likely that Vnd (see also ), together with L'sc, promote formation of these late-developing ventral neuroblasts. As Vnd is still expressed in the remaining ventral brain neuroblasts in EGFRf2 mutants (electronic supplementary material, figure S5f′), this suggests that vnd (similar to ind) expression is differently regulated in NE and neuroblasts (see also ), and further, that these ventral brain neuroblasts do not undergo a fate shift towards intermediate identity, as has been observed for approximately 50% of ventral neuroblasts in the EGFR-mutant VNC .
3.5. Phylogenetic considerations of EGFR-regulated patterning in the brain
The key components of the EGFR signalling pathways are evolutionarily highly conserved from fly to human. In vertebrates, 4 EGFR family members (ErbB1-4, with ErbB1 homologous to EGFR) and 11 EGF-like ligands are known (reviewed in [93,94]). In the forebrain, ErbB ligands secreted from a narrow region between the dorsal and ventral telencephalon (called ‘antihem’) have been proposed to assist in maintaining DV fates, which suggests a possible involvement of EGFR signalling in regional patterning of the cerebral cortex . ErbB signalling might also be involved in patterning and differentiation of structures at the midbrain–hindbrain boundary (reviewed in ). However, ErbB signalling has not been connected with regulation of DV patterning genes (i.e. vnd/Nkx2, Nkx6, ind/Gsh). Instead, several other extrinsic signalling molecules are involved in their regulation, including the key player Shh, which is secreted from the floorplate (reviewed in [97–99]). This suggests that different upstream signalling pathways are used to control the expression of DV patterning genes in insect and vertebrate brains, even though the regionalized expression of these genes exhibits certain similarities in the embryonic brain of both animal phyla [52,100].
4. Material and methods
4.1. Drosophila genotypes
The following fly strains were used: Oregon R (wild-type); aosΔ7 , Df(3L)H99 , EGFRf2 , UAS-sim ; UAS-sSpi ; Matα-Gal4 , NGT40-Gal4 , simH9 , tlll49  (all provided by Bloomington Drosophila Stock Center); rhoPΔ5  (provided by Marta Llimargas Casanova); sca-Gal4  (provided by Uwe Hinz); UAS-GroAA, UAS-GroDD  (provided by Ze'ev Paroush); UAS-tll  (provided by Mitsuhiko Kurusu); vnRG436  (provided by Amanda Simcox); UAS-vnd , vnd6  (provided by James Skeath).
4.2. Staging, flat preparation and mounting of embryos
Flat preparations of the head/truncal ectoderm of stained embryos and mounting were carried out as described previously .
Embryos were dechorionated, fixed and immunostained according to previously published protocols . The following primary antibodies were used: mouse-anti-Dachshund 2–3 (1 : 250) , mouse-anti-Invected 4D9 (1 : 7) , mouse-anti-Prospero (1 : 10) (all provided by DSHB); rabbit-anti-Death caspase-1 (#9578) (1 : 50), rabbit-anti-p44/42-MAPK (1 : 500) (both provided by Cell Signalling Technology); guinea pig-anti-Deadpan (1 : 5000)  (provided by Jürgen Knoblich); rabbit-anti-Engrailed (1 : 800) (Santa Cruz Biotechnology); rabbit-anti-Ind (1 : 1000)  (provided by Tonia von Ohlen); mouse-anti-p44/42-MAPK (1 : 2000) (provided by Sigma Aldrich); rabbit-anti-PH3 (1 : 500) (provided by Merck Millipore); guinea pig-anti-Runt (1 : 300)  (provided by Ralf Pflanz); guinea pig-anti-Sim (1 : 1500)  (provided by Stephen Crews); rabbit-anti-Vnd (1 : 2000)  (provided by Marshall Nirenberg); sheep-anti-DIG alkaline-phosphatase conjugated (1 : 1000) (provided by Roche Diagnostics). The secondary antibodies were either biotinylated, conjugated with alkaline-phosphatase, or DyLight, Cyanine (all Jackson Immunoresearch) and Alexa (Life technologies) fluorescent dyes (all diluted 1 : 500). Tyramide signal amplification (TSA biotin system; PerkinElmer) was used in DAB stainings according to the manufacturer's protocol.
4.4. Whole mount in situ hybridization
Probes were synthesized using either linearized cDNA/EST-clones, cloned PCR products or PCR products containing a RNA polymerase adapter (both from genomic DNA) (electronic supplementary material, table S1) as a template. T3, T7 or SP6 Polymerase and DIG-RNA Labelling Mix (all Roche Diagnostics) were used for probe synthesis according to the manufacturers protocol. In situ hybridization was performed as described previously  and the probes processed with NBT/BCIP or VectorRed (Vector Labs) solution. Afterwards, the embryos were immunolabelled with a second primary antibody followed by incubation with biotinylated secondary antibodies, and processed with DAB.
The non-fluorescent stainings were documented on a Zeiss Axioplan. Pictures were digitized with a CCD camera (Contron progress 3012). Fluorescent confocal images were acquired on a Leica TCS SP5 II. Pictures were processed with ImageJ, Adobe Photoshop CS4 and Adobe Illustrator CS4. Data (shown in figures 2p and 6c,f; electronic supplementary material, figure S5i,m) were analysed with a two-tailed unpaired Student's t-test.
R.U. conceived the study. D.J. and R.U. designed experiments. D.J. performed the majority of experiments and analysed the majority of data. J.v.H. participated in the analysis of the wild-typic MAPK pattern and EGFR-dependent regulation of DV genes. D.J. prepared all figures. D.J. and R.U. wrote the manuscript.
We declare we have no competing interests.
This work was supported by grants from the Deutsche Forschungsgemeinschaft (UR163/2-2, UR163/3-3 to R.U.).
We thank Dagmar Volland for excellent technical assistance and Gerd Technau for critical reading of the manuscript. We are indebted to James Skeath, Ralf Pflanz, Tonia von Ohlen, Joachim Urban, Jürgen Knoblich, Chris Doe, Stephen Crews, Marshall Nirenberg, Marta Llimargas Casanova, Uwe Hinz, Amanda Simcox, Mitsuhiko Kurusu, Ze'ev Paroush, Ethan Bier, the Bloomington Drosophila Stock Center, the Developmental Studies Hybridoma Bank and the Drosophila Genomic Resource Center for providing flies, antibodies and cDNA.
Electronic supplementary material is available online at https://dx.doi.org/10.6084/m9.figshare.c.3584039.
- Received July 7, 2016.
- Accepted November 14, 2016.
- © 2016 The Authors.
Published by the Royal Society under the terms of the Creative Commons Attribution License http://creativecommons.org/licenses/by/4.0/, which permits unrestricted use, provided the original author and source are credited.