Ems and Nkx6 are central regulators in dorsoventral patterning of the Drosophila brain

The development of the central nervous system in vertebrates and invertebrates involves the transformation of a two-dimensional sheet of neuroectodermal cells into a complex three-dimensional structure comprising a variety of different neural cell types. The specification of neural cell types is a multi-step process, which at early stages of embryogenesis critically depends on conveying positional information in the neuroectoderm (NE) to neural stem cells. The underlying molecular genetic mechanisms have been extensively studied in the embryonic ventral nerve cord (VNC) of the trunk (reviewed by Dessaud et al., 2008; Skeath and Thor, 2003).

In Drosophila, the VNC is generated by segmental arrays of neural stem cells, called neuroblasts (NBs), which delaminate from the truncal NE. Each NB acquires a unique identity that is finally reflected in the production of a specific cell lineage (Bossing et al., 1996; Schmidt et al., 1997). NB identity is specified by the combinatorial code of positional cues in the NE provided by the products of early patterning genes, which act in the anteroposterior (AP) and dorsoventral (DV) axes (reviewed by Skeath and Thor, 2003). The graded activities of the nuclear factor Dorsal, BMP and Epidermal growth factor receptor (Egfr) signaling pathways determine the DV boundaries of the NE, and further regulate the expression of a set of evolutionary conserved DV patterning genes (Hong et al., 2008; Mizutani et al., 2006; Skeath, 1998; von Ohlen and Doe, 2000). The expression of these ‘DV genes’ subdivides the trunk NE into three longitudinal columns along the DV axis: ventral nervous system defective (vnd/Nkx2) in the ventral, intermediate neuroblasts defective (ind/Gsx) in the intermediate, and muscle segment homeobox (msh/Msx; Drop - FlyBase) in the dorsal NE column (Chu et al., 1998; Isshiki et al., 1997; McDonald et al., 1998; Weiss et al., 1998). In Drosophila, these homeobox genes interact in a hierarchical cascade of transcriptional repression, according to which vnd represses ind (and msh) in the ventral column, and ind represses msh in the intermediate column. The DV genes encode key regulators of NB identity and each NE column thereby gives rise to a population of distinctly specified NBs. However, whereas vnd, ind and Egfr have also been shown to be central for the formation of NBs in their respective NE column, this role appears dispensable for msh (reviewed by Cornell and von Ohlen, 2000; Skeath, 1999).

Much less is known about the genetic mechanisms of DV regionalization in the developing brain. The procephalic NE and the descending population of ~100 brain NBs in each hemisphere can be subdivided (from anterior to posterior) into the presumptive proto- (PC), deuto- (DC) and tritocerebrum (TC) (Urbach et al., 2003). We previously showed that DV genes are expressed in a segment-specific manner in the procephalic NE and brain NBs (Urbach and Technau, 2003a; Urbach and Technau, 2003b). It was further shown that the way in which vnd controls expression of ind and msh differs between the brain and the VNC, and that vnd is necessary for proper development of the ventral procephalic NE and of the brain NBs that descend from these NE domains. These experiments suggested that the role of vnd/Nkx2 in brain development exhibits striking parallels between vertebrates and Drosophila (Urbach et al., 2006).

In this work, we uncover a novel regulatory network of homeodomain transcription factors in Drosophila that is specifically necessary to pattern the TC and DC in the DV axis, whereas the underlying genetic interactions in the PC diverge. We provide evidence that empty spiracles (ems) and Nk6 homeobox (Nkx6; HGTX - FlyBase) encode key regulators in the DV genetic network. Intriguingly, ems, a cephalic gap gene, is crucial for the brain-specific regulation of DV gene expression (Nkx6, vnd, ind, msh) and, conversely, DV genes control the expression of ems. Moreover, we demonstrate that cross-repressive interactions between pairs of homeodomain proteins establish DV gene expression domains in the fly brain, which bears similarity to DV patterning in the vertebrate neural tube.

The following fly strains were used: Oregon R (wild type), Df(3L)XG3 (Bloomington Stock Center), ems^9H83 (Jürgens et al., 1984), ind^16.2 (Weiss et al., 1998), msh⁶⁸ (Isshiki et al., 1997), Nkx6^D25 (Broihier et al., 2004), vnd⁶ (Jiménez and Campos-Ortega, 1990), UAS-vnd (Chu et al., 1998), UAS-ind (von Ohlen et al., 2007), UAS-Nkx6 (Broihier et al., 2004), UAS-msh-m25-m6 (Isshiki et al., 1997) and UAS-ems (Lovegrove et al., 2006). The vnd⁶; msh⁶⁸ double-mutant flies were created using standard genetic methods. The UAS-Gal4 system (Brand and Perrimon, 1993) was used to overexpress the UAS constructs in the neuroectoderm, crossing the UAS lines with the sca-Gal4 (Klaes et al., 1994) or Matα-Gal4-VP16 (Häcker and Perrimon, 1998) driver line. In the deficiency Df(3L)XG3, the chromosome region 70E3 to71D4 is deleted, including the closely adjacent loci of the ind and Nkx6 genes; the deletion, notably, does not include any other gene for which a role in DV patterning is known and that might otherwise confound our observed expression phenotype. To confirm whether the deficiency Df(3L)XG3 includes Nkx6 and ind, complementation tests were performed: the deficiency was crossed with either single mutant (ind^16.2 and Nkx6^D25) and each was found to be non-complementary. All mutants have been blue balanced and distinguished from heterozygotes via antibody staining against β-galactosidase. The deficiency Df(3L)XG3 was additionally identified by lack of staining by in situ hybridization against Nkx6 or ind mRNA.

Staging of the embryos was carried out as previously described (Campos-Ortega and Hartenstein, 1997). Flat preparations of the head ectoderm of stained embryos and mounting were carried out as previously described (Urbach et al., 2003).

Embryos were viewed under a Zeiss Axioplan microscope equipped with Nomarski optics using 40×, 63× and 100× oil-immersion objectives. Pictures were digitized with a CCD camera (Contron Progress 3012) and processed with Adobe Photoshop CS2.

Embryos were dechorionated, fixed and immunostained according to previously published protocols (Urbach et al., 2003). For the anti-En and anti-Ems antibodies, biotinyl tyramide (TSA Biotin System, PerkinElmer) was used to amplify the signal following the manufacturer's protocol. The following primary antibodies were used at the indicated dilutions: rabbit anti-β-gal (1:2000; Promega), mouse anti-En (1:7; 4D9, DSHB), rat anti-Ems (1:1000) (Walldorf and Gehring, 1992), sheep anti-DIG alkaline phosphatase conjugated (1:1000; Roche Diagnostics), sheep anti-FITC alkaline phosphatase conjugated (1:1000; Roche Diagnostics) and rat anti-Nkx6 (1:100) (Brohier et al., 2004). The secondary antibodies (donkey anti-mouse, donkey anti-rabbit, donkey anti-rat; Dianova) were either biotinylated or alkaline phosphatase conjugated and used at 1:500.

ind, vnd and msh RNA probes were synthesized with T7 RNA polymerase using pBS SKII(+) linearized with XhoI(ind), SacI(vnd) or HindIII (msh) as a template. The Nkx6 RNA probe was synthesized with T3 RNA polymerase, using the EST clone RE18506 [Berkeley Drosophila Genome Project (BDGP)] cloned into SphI-linearized pFLC1 as a template, according to the manufacturer's protocol (Roche). All riboprobes were DIG- as well as FITC-labeled. In situ hybridization was performed as described previously (Urbach et al., 2006) and the probes processed with NBT/BCIP solution (Roche Diagnostics), which, after a wash in methanol, results in blue staining. Then, the embryos were immunolabeled with a second primary antibody, followed by incubation with a biotinylated secondary antibody and processing with DAB. For double in situ hybridizations, the FITC-labeled probe was processed with NBT/BCIP, followed by several washes with glycine buffer. Then, embryos were incubated with anti-DIG-AP antibody and the second probe processed with Vector Red (Vector Laboratories).

The Drosophila brain, developing from the procephalic NE, can be divided into the TC, DC and PC. These AP subregions can be subdivided further into distinct DV domains based on the domain-specific expression of different DV genes (vnd, ind and msh), which is most clearly displayed by stage 9 (Fig. 1M′,M″) (Urbach and Technau, 2003a). In relationship to these genes, we investigated the expression of Nkx6 (Fig. 1A-M′), another important DV gene (as we will show). Expression of Nkx6, which initiates in the blastodermal procephalic NE by stage 6 (Fig. 1C,D) (see Uhler et al., 2002), clearly exhibited segment-specific differences by stages 8/9: it was expressed in the ventral (which coexpresses vnd, Fig. 1K) and intermediate NE (which coexpresses ind, Fig. 1L) of the TC, as well as in the intermediate DC (which coexpresses ind). Thus, the expression of Nkx6 is complementary to that of dorsal msh (Fig. 1C,M′). Nkx6 was also detected in distinct subsets of early- and later-born trito- and deutocerebral NBs (Fig. 1M; data not shown), which develop from Nkx6-expressing NE (Fig. 1E,I,J,M′). Nkx6 was not expressed in the NE of the PC, but by stage 11 was detected in a single protocerebral NB (Ppd5, Fig. 1J). Hence, expression of Nkx6 at high levels and from early embryonic stages onwards is observed only in the NE and NBs of the TC and DC.

We previously described that, in contrast to the trunk NE, the expression domains of msh and vnd partially share a common border in the TC and DC (Urbach and Technau, 2003a) (see Fig. 1B,D,M′), and that Vnd represses msh (Urbach et al., 2006). vnd is downregulated early in development in the intermediate NE and in large parts of the ventral NE of both neuromeres (Urbach et al., 2006), raising the question of how suppression of msh is sustained in this area after vnd has been lost. We found that Nkx6 is expressed in the intermediate TC and DC just before the expression of vnd is downregulated and expression of ind is activated (Fig. 1A,C,D; Fig. 2B,C). Since the expression of Nkx6 is complementary to that of msh (Fig. 1C; Fig. 2A), we speculated that Nkx6 could act as a repressor of msh. In support of this, we observed ectopic msh in the intermediate TC (70%, n=20 brain hemispheres) and in the intermediate/ventral DC (95%, n=20) in Nkx6-null mutant embryos (Nkx6^D25) by stages 10/11 (Fig. 2D,D′,E,E′). In a reciprocal experiment, we ectopically expressed Nkx6 in the entire NE using sca-Gal4 (termed sca>Nkx6), which drives expression in the NE from stage 8 onwards, and found that ectopic Nkx6 represses msh in the dorsal NE of the TC and DC (10% complete loss, 90% strong reduction of msh, n=35) (Fig. 2F,F′). Thus, Nkx6 acts as a repressor of msh and compensates for the early loss of vnd, keeping dorsalizing signals out of the ventral/intermediate TC and DC. To investigate whether Msh in turn represses Nkx6 in the dorsal TC and DC, we explored Nkx6 expression in msh-null mutants (msh⁶⁸). In these embryos, the dorsal limit of the Nkx6 expression domain was shifted into the entire dorsal NE of the TC and DC (100%, n=28) (Fig. 3A,E,E′). Vice versa, after ectopic expression of msh using the maternal Matα-Gal4-VP16 (termed Matα-Gal4), which drives expression ubiquitously (Bossing et al., 2002), we observed that Nkx6 expression is largely reduced (83%, n=18) and sometimes absent (17%, n=18) (Fig. 3F,F′). Hence, these data demonstrate that Nkx6 and Msh cross-repress each other to stabilize the boundary between intermediate and dorsal NE in the TC and DC.

Our observation that Nkx6 is expressed in specific DV neuroectodermal domains in the TC and DC prompted us to look in more detail into how Nkx6 is regulated there. We detected Nkx6 coexpression with vnd in the ventral TC, and with ind and vnd (transiently) in the intermediate TC and DC (Fig. 1K,L). It has been reported previously that Vnd is a positive regulator of Nkx6 in the head NE (Uhler et al., 2002). In agreement, we found a lack of Nkx6 expression in the TC and DC in vnd-null mutant embryos (vnd⁶; 100%, n=20; Fig. 3A,C,C′). Since vnd is necessary to activate ind (Urbach et al., 2006) in the intermediate DC, this raises the possibility that vnd could regulate Nkx6 expression indirectly via ind. However, expression of Nkx6 was unaffected in the DC in ind-null mutants (ind^16.2), as was also the case in the TC (100%, n=16; Fig. 3B), indicating that Nkx6 depends on vnd but not on ind. Since ind expression is unaffected in Nkx6 mutant embryos (data not shown), this suggests that Nkx6 and ind are positively regulated in parallel in the DC by Vnd.

Although we provide evidence that Vnd is a positive regulator of Nkx6 expression in the TC and DC, it is unclear whether Vnd regulates Nkx6 directly or indirectly (see Uhler et al., 2002). In both brain neuromeres, Vnd is a repressor of msh (Urbach et al., 2006), and here we show that Msh acts as a repressor of Nkx6. This raises the possibility that Vnd regulates Nkx6 indirectly by repressing msh. In accordance with this, Nkx6 expression is abolished in vnd mutant embryos in which msh is derepressed in the ventral TC and DC (Fig. 3C,C′) (Urbach et al., 2006), analogous to the situation after ectopic msh expression (Fig. 3F,F′). Conversely, upon overexpressing vnd, which leads to repression of msh (Urbach et al., 2006), we observed a significant expansion of Nkx6 expression into the dorsal NE of TC and DC (100%, n=25; Fig. 3D,D′), similar to the results obtained in msh mutants (Fig. 3E,E′). Even though all of these observations suggest that Vnd regulates Nkx6 indirectly by repressing the Nkx6-repressor Msh, they do not exclude the possibility of a more direct interaction of vnd with Nkx6. We therefore tested Nkx6 expression in vnd⁶; msh⁶⁸ double-mutant brains. We hypothesized that if Vnd does not directly activate Nkx6 but is only needed to repress msh, then Nkx6 should be detected in its endogenous expression domain (despite the absence of Vnd) and should additionally be expanded into the dorsal NE because Msh repressor function is absent in this genetic background. Exactly this pattern of Nkx6 expression was observed in the double-mutant brain (100%, n=10; Fig. 3G,G′), indicating that Vnd facilitates the expression of Nkx6 by repressing the Nkx6-repressor msh.

The sharp restriction of the domain of Nkx6 expression at the anterior DC and posterior TC suggests an interference with regulatory factors acting along the AP axis. Unexpectedly, we identified the cephalic gap gene ems to be involved. ems expression initiates in a blastodermal circumferential stripe in the procephalon (Walldorf and Gehring, 1992) that subsequently resolves into three smaller domains that later encompass part of the TC, DC and PC (e.g. Fig. 4C) (Hirth et al., 1995; Urbach and Technau, 2003a). In double labelings, we found that Nkx6 expression initiates (by stage 6) within the Ems domain (in the prospective intermediate DC; Fig. 4A). Slightly later, the anterior borders of the Nkx6 and Ems expression domains coincided exactly. Nkx6 was subsequently expressed in NE cells of the TC, which retain high levels of Ems expression before Nkx6 is activated (Fig. 4B). By stage 8, when the Nkx6 expression domain is entirely established, the posterior borders of the Nkx6 (at the posterior TC) and Ems expression domains corresponded precisely (Fig. 4B). Thus, ems appears to define the AP boundaries of Nkx6 expression in the TC and DC. This was also observed upon vnd overexpression or in msh mutants, in which the Nkx6 domain expands into the dorsal NE (Fig. 3D,E); in both cases, ectopic Nkx6 is expressed within the AP borders of the early ems domain. Since these expression data imply that Ems plays a role in the activation of Nkx6, we tested the expression of Nkx6 in ems-null mutant embryos (ems^9H83). Remarkably, in these embryos Nkx6 expression was completely abolished in the TC and DC (100%, n=25) (Fig. 4C,J), indicating that the AP patterning gene ems is required for the activation of Nkx6. The procephalic ems domain in addition covered the areas of ind and msh expression in the TC and DC (Fig. 4D,E,G,H). Expression of ind (similar to Nkx6) initiated in the DC where the anterior borders of the ind and ems domains corresponded precisely (Fig. 4D,E), suggesting that ems also regulates the anterior expansion of ind expression in the DC. In ems mutants, the expression of ind and msh was entirely absent in the NE of the TC and DC (Fig. 4F,K,I,L), indicating that ems is needed for the activation of ind and msh as well.

These results led us to ask whether Ems alone is able to activate Nkx6, ind or msh. Since endogenous ems is expressed throughout the entire NE of the early TC and DC (Fig. 4A,B), we investigated expression of Nkx6, ind and msh in the PC and in the NE of the VNC upon mis-expression of Ems. In Matα>ems or sca>ems embryos, msh expression was found in a nested ectopic domain in the central PC (Fig. 4M,N), whereas expression of Nkx6 and ind appeared unaltered (data not shown); also, in the trunk NE, ectopic expression of these DV genes could not be detected (data not shown). This suggests that in the PC, Ems is to a certain extent sufficient to activate msh but not Nkx6 or ind; however, we assume that this is due to interference with factors specific for the PC [e.g. the ems-repressor Tailless (Hartmann et al., 2001)] or for the trunk [where ems function is repressed by the activity of homeotic genes (Schöck et al., 2000)] NE. Moreover, high levels of Ems seem to be necessary early on in order to initiate expression of Nkx6, so that it is additionally possible that the onset and intensity of ectopic Ems expression are inadequate to induce activation of the DV genes. Nevertheless, an indication that Ems activates Nkx6 was observed in msh mutant embryos. As mentioned above, in msh mutants, ectopic Nkx6 expanded into the entire dorsal NE of TC and DC (Fig. 3E,E′). Since Ems is also expressed in this NE region (Fig. 4A,B), this suggests that Ems might be able to activate Nkx6 in the absence of Msh.

From stage 9 onwards, Ems and Nkx6 or ind are expressed in mutually exclusive NE domains (and in corresponding populations of brain NBs) in the TC and DC (Fig. 4C,F; see Fig. 6A; data not shown), raising the possibility that one or both of these genes might become a repressor of ems. Therefore, we examined Ems expression in Nkx6 or ind loss- and gain-of-function embryos. Since Ems expression appeared unaltered under all these conditions (data not shown), this indicates that neither Nkx6 nor Ind is alone sufficient to repress ems. However, in the absence of both ind and Nkx6 in the small deficiency Df(3L)XG3, Ems was specifically derepressed in the intermediate DC (where ind and Nkx6 are normally coexpressed). Derepression of ems in Df(3L)XG3 embryos was not observed before stage 10 (Fig. 4O,P; complete derepression in 44%, strong derepression in 56%, n=25), when Vnd can be detected in the respective NE (Urbach et al., 2006). Since Vnd is an early repressor of ems (see below), but later disappears from the intermediate DC, we suggest that repression of ems (as initially carried out by Vnd) is maintained by the combined activity of Nkx6 and Ind. This further argues for a regulation of ems expression via a negative-feedback loop in which Ems is needed to activate its own late repressors Nkx6 and Ind.

vnd is important for the formation and specification of brain NBs. In contrast to the trunk, in the early brain vnd is dynamically expressed and becomes progressively confined to three ventral domains in the posterior TC, DC and PC (Sprecher et al., 2006; Urbach et al., 2003). We found that vnd is transiently expressed also in the intermediate NE of the TC and DC (Fig. 1A,D; data not shown), which confirms our previous assumption that in vnd mutant embryos the observed defects in the formation and specification of intermediate brain NBs are also cell-autonomously regulated (Urbach et al., 2006). However, it has remained unclear how vnd expression becomes depleted in the intermediate NE and in large parts of the ventral NE in the TC and DC. Interestingly, we again identified involvement of the cephalic gap gene ems. In double labelings for Ems protein and vnd mRNA, we observed Ems coexpression with vnd by stage 5 in the prospective NE of the intermediate/ventral TC and DC (Fig. 5A,A′). During stages 6/7, expression of vnd began to disappear within the Ems domain in the intermediate and distinct parts of the ventral NE in the DC (Fig. 5B,C,C′). This early reduction of vnd expression suggests that Ems represses vnd. To test this, we examined vnd expression in ems mutant embryos. Indeed, we found that until stages 10/11, vnd expression in these embryos is kept at high levels in those NE domains in the TC and DC (100%, n=26) (Fig. 5I,J), where it is normally repressed already during stages 6-9 (Fig. 5A-E′). Thus, Ems is necessary to repress vnd in the intermediate NE and in large parts of the ventral NE in the TC and DC.

However, Ems expression in the prospective NE of the TC and DC does not completely repress vnd in the ventral NE. Surprisingly, during stages 7/8, cells began to lose Ems expression in domains where vnd expression remained high (Fig. 5C,C′). During stage 8, the residual Ems/vnd-coexpressing NE became partitioned into Ems- or vnd-expressing subdomains (Fig. 5D,D′), so that by stage 9, when neurogenesis initiates, the Ems and vnd expression domains were largely complementary (Fig. 5E,E′). This suggests that Vnd might also repress ems. In vnd mutant embryos, ectopic Ems expression was found in the intermediate/ventral DC and in the ventral TC (100%, n=30; Fig. 5F,G). Conversely, mis-expression of vnd (Matα>vnd) abolished Ems expression almost completely in the TC/DC, except for a small domain in the ventral DC where expression levels of Ems were significantly reduced (100%, n=18; Fig. 5F,H) (see Urbach et al., 2006). These data suggest that Vnd is necessary and sufficient to repress ems. Taken together, we conclude that after the blastodermal phase of coexpression, Ems and Vnd act as mutual inhibitors.

By stage 11, Nkx6 was additionally detected in one NB (Ppd5) in the PC, but not in the overlaying NE (Fig. 1J; Fig. 6A). Interestingly, Ems, but not vnd, was coexpressed with Nkx6 in Ppd5, and Ems was also detected in its NE of origin (Fig. 6A,B). Accordingly, Nkx6 expression was not affected in Ppd5 in vnd mutants (Fig. 6C). That Nkx6 is regulated independently of vnd in the PC was further corroborated by our observations made upon vnd overexpression (Mata>vnd or sca>vnd), showing that Nkx6 cannot be ectopically induced in the PC (Fig. 3D; Fig. 6D) (see Uhler et al., 2002). By contrast, in ems mutants Nkx6 expression was absent at the position of Ppd5 (100%, n=22) (Fig. 6E), indicating the involvement of Ems in activating Nkx6. In ems mutants, the development of brain NBs has been reported to be affected (Younossi-Hartenstein et al., 1997), and even though we observe NBs in the position of Ppd5, we cannot exclude the possibility that loss of Nkx6 is a secondary effect of the non-formation of Ppd5. To see whether Msh (which is not normally expressed in the PC in the investigated time window) is able to interact with Nkx6 in the PC, we overexpressed msh (sca>msh) in this region. We found that Nkx6 expression is abolished in Ppd5 (100%, n=24; Fig. 6F), indicating that Msh, similar to the situation in the TC and DC, is capable of repressing Nkx6 also in the PC. Taken together, these results suggest that Ems is necessary to induce the expression of Nkx6 in the protocerebral NB Ppd5. In contrast to in the TC and DC, Vnd does not play a role in the regulation of Nkx6 expression in the PC, most likely because the Nkx6-repressor Msh is absent from the early PC.

We show for the first time that the evolutionarily conserved homeodomain protein Ems is an integral component of the gene regulatory network that governs DV patterning in the posterior brain neuromeres, the TC and DC (Fig. 7A,B). This novel function is surprising because ems has hitherto been exclusively connected with patterning functions along the AP axis. It has been proposed that the combined activities of the gap genes ems, buttonhead and orthodenticle (ocelliless - FlyBase) generate head segments (Cohen and Jürgens, 1990; Grossniklaus et al., 1994) and that ems mutants exhibit defects in the formation of the intercalary and antennal segment (Cohen and Jürgens, 1990; Schmidt-Ott et al., 1994) as well as in the corresponding TC and DC (Hirth et al., 1995; Younossi-Hartenstein et al., 1997) in accordance with the early pattern of ems expression (Dalton et al., 1989; Walldorf and Gehring, 1992; Urbach and Technau, 2003b). ems probably also has a homeotic function in specifying aspects of intercalary segment identity (Schöck et al., 2000). We provide evidence that another crucial function of Ems is its cross-repressive interaction with Vnd (Fig. 7B). Previously, we showed that vnd expression is dynamic and exhibits specific differences in the TC and DC (Urbach et al., 2006). Here, we demonstrate that Ems is involved in the regulation of brain-specific differences in vnd expression, and that Vnd acts to repress ems in complementary parts of the TC and DC (Fig. 7B). These interactions help to refine the pattern into mutually exclusive domains at the onset of neurogenesis, which is important as both genes provide positional information that subsequently specifies the identity of individual brain NBs (Urbach and Technau, 2003b). Depending on the context, Vnd/Nkx2 can act as a transcriptional activator or repressor, as determined by physical interaction with the co-repressor Groucho, which enhances repression (Chu et al., 1998; Cowden and Levine, 2003; Muhr et al., 2001; McDonald et al., 1998; Stepchenko and Nierenberg. 2004; Uhler et al., 2007; Yu et al., 2005). Interestingly, we observe that Ems also regulates the expression of two Nkx genes in an opposing manner: it represses vnd/Nkx2 but is necessary to activate Nkx6. The repressor function of Ems most likely also depends on Groucho, as Ems has been reported to bind Groucho in vitro (Goldstein et al., 2005).

In ems mutants, defects in proneural gene expression (lethal of scute and achaete) are restricted to NE regions where ems is normally expressed during early neurogenesis, leading to the loss of a subset of NBs in the TC and DC (Younossi-Hartenstein et al., 1997; Hartmann et al., 2000). This contrasts with the phenotype of the late embryonic ems mutant brain, which exhibits a severe reduction, or entire elimination, of the TC and DC (Hirth et al., 1995), suggesting that the proper development of a larger NE domain and/or fraction of NBs in the TC and DC must be affected. However, in ems mutants the organization of the early procephalic NE appears normal until stages 9/10 and apoptosis is not detected (Hartmann et al., 2000). A possible explanation for the subsequent complete loss of TC and DC is that in ems mutants, vnd becomes derepressed in the ventral/intermediate NE of both neuromeres, and expression of msh, ind and Nkx6 is not activated. As we have shown previously, ectopic vnd prevents the expression of many NB identity genes (Urbach et al., 2006). Indeed, the expression of a number of molecular markers has been reported to be absent in the ems mutant brain (Hartmann et al., 2000). It is therefore conceivable that in the TC and DC of ems mutants, as a consequence of lacking ems and ectopic vnd (and the absence of proneural gene activation), some NBs do not form. Additionally, owing to mis-specification of the NE (where neural identity gene expression is absent or altered), the other NBs and their progeny might still form but degenerate at later stages.

It has been largely unclear how expression of Nkx6 is regulated in the brain NE, although Vnd has been suggested to act as a positive regulator (Uhler et al., 2002). At the blastodermal stage, coexpression of ems and vnd is only observed in the intermediate and ventral NE of the TC and DC, which might account for early Nkx6 expression being limited to the respective NE in the brain and absent from the trunk. Our data indicate that Ems and Vnd together facilitate the activation of Nkx6. Ems expression closely prefigures the domain of Nkx6 expression in the TC and DC (Fig. 7A), and together with the fact that Nkx6 is completely abolished in ems mutants, this suggests that Ems might act as a direct activator to regulate the extension of the Nkx6 domain along the AP axis. Vnd indirectly regulates the enlargement of the Nkx6 domain along the DV axis by repressing the Nkx6-repressor Msh. That DV patterning in the brain NE integrates AP signals is additionally supported by the fact that Ems is also necessary for activation of ind and msh, indicating that ems is a key regulator in DV patterning of the TC and DC. We also provide evidence for a negative-feedback control in the DV regulatory network, in which Ems is needed to activate its own later-stage repressors, Nkx6 and Ind (Fig. 7B). Together, our data suggest not only that Ems regulates the expression of all DV genes (activating Nkx6, ind, msh and repressing vnd), but also that DV factors (Nkx6, Ind and Vnd) control expression of ems, indicating that integration of DV and AP patterning signals takes place at different levels in the DV genetic network.

We identified Nkx6 as specifically involved in DV patterning of the TC and DC (Fig. 7B). In addition to later suppression of ems (in concert with Ind), a further pivotal function of Nkx6 is to maintain the suppression of msh in the intermediate/ventral TC and DC that was initiated by Vnd. Since in both neuromeres the expression of Nkx6 starts before and persists longer than that of ind (Fig. 7A), and because msh is ventrally derepressed in Nkx6 but not in ind mutants (J.S. and R.U., unpublished observations), this implies that Nkx6 (but not Ind) is the major msh suppressor necessary to prevent intermediate/ventral NE and the descending NBs from adopting dorsal fates (Urbach et al., 2006). Consequently, Nkx6 indirectly regulates the proper specification of brain NB identity by suppressing msh (and ems). Further experiments are required to show whether Nkx6 is also more directly involved in the fate specification of NBs and progeny cells in the brain, as has been shown in the VNC, where Nkx6 promotes the fate of ventrally projecting, and represses the fate of dorsally projecting, motoneurons (Broihier et al., 2004).

Additionally, we observed cross-inhibitory interactions between Nkx6 and Msh (Fig. 7A,B). We assume that this mutually repressive regulation in the TC and DC is necessary to stabilize the boundary between dorsal and intermediate NE, and to ensure the regionalized expression of msh and Nkx6 over time. It is likely that Nkx6 and Msh/Msx interact with the co-repressor Groucho (Andersson et al., 2006; Broihier and Skeath, 2002; Muhr et al., 2001; Uhler et al., 2002; Syu et al., 2009) to repress each other at the transcriptional level. Interestingly, aspects of the genetic interactions between Nkx6 and Msh/Msx seem to be evolutionarily conserved, as Msx1, which is expressed in the vertebrate midbrain and functions as a crucial determinant in the specification of dopamine neurons, represses Nkx6.1 in ventral midbrain dopaminergic progenitors of mice (Andersson et al., 2006).

It had not been shown until now that domains of DV gene expression in the Drosophila brain become established through cross-repressive regulation, and it is possible that such genetic interactions are more common than previously thought (e.g. we have further evidence that Ind and Msh act as mutual inhibitors; J.S. and R.U., unpublished). This suggests that in the fly brain, cross-inhibition between pairs of homeodomain transcription factors is fundamental for establishing and maintaining DV neuroectodermal and corresponding stem cell domains. By contrast, in the NE of the VNC, where DV patterning is much better understood (reviewed by Cornell and Von Ohlen, 2000; Hong et al., 2008; Skeath, 1999), cross-repressive interactions of homeobox genes are largely omitted. There, DV patterning is proposed to be conducted by a strict ventral-dominant hierarchy according to which ventral genes repress more-dorsal genes (Cowden and Levine, 2003). However, one exception to the rule seems to be the cross-inhibitory interaction between Vnd and Ind (Zhao et al., 2007). Interestingly, in the developing vertebrate neural tube, cross-repressive interactions of homeodomain proteins are common and indeed crucial for the establishment of discrete DV progenitor domains (reviewed by Dessaud et al., 2008). This bears a marked resemblance to the mutually antagonistic relationship between pairs of homeodomain proteins that dorsoventrally pattern the fly brain.

A predominant feature of the brain-specific DV genetic network we describe here, and a general design feature of gene regulatory networks (Levine and Davidson, 2005), is the extensive use of transcriptional repression to regulate target gene expression in spatial and temporal dimensions. All factors involved in the network operate as repressors (except Ems, which may also serve as an activator), via mutual repression (between Ems and Vmd, and between Nkx6 and Msh), a double-negative mechanism (Vnd represses Msh, which represses Nkx6), and a negative-feedback loop (Ems is needed to activate Nkx6 and Ind, which in turn repress Ems) (Fig. 7B). The spatial and temporal complexity of the regulatory interactions we have deciphered implies similar complexity in the underlying cis-regulatory control of these factors. For example, the domain of msh expression is regulated by the input of at least two transcriptional repressors acting in subsequent time windows (Vnd early and Nkx6 late), and the input of at least three repressors regulates the dynamics of ems expression (Vnd early, Ind and Nkx6 late). The brain-specific DV patterning network probably comprises further genes in addition to those that we have identified, and it is likely that interactions with other putative regulators (e.g. Dorsal, Egfr, Dpp) will complement our present model. Altogether, our data provide the basis for a systematic comparison of the genetic processes underlying DV patterning of the brain between different animal taxa at the level of gene regulatory networks.

The genetic factors considered in this study in the developing fly brain are expressed in similar NE domains from early embryonic stages onwards in the anterior neural plate in vertebrates. Emx2, for example, is expressed in the laterodorsal region, and Nkx2 genes in the ventral region, of the early vertebrate forebrain (Fig. 7C). At the four-somite stage (~E8), these two domains exhibit a common border (reviewed by Rubenstein et al., 1998; Shimamura et al., 1995), similar to that observed in Drosophila after Ems and Vnd have, through cross-repression, regulated their mutually exclusive expression domains. Moreover, whereas Msx genes are mainly expressed in dorsal regions of the posterior forebrain, midbrain and hindbrain (reviewed by Ramos and Robert, 2005), expression of Nkx6 genes is reported in more lateroventral regions, overlapping ventrally with the expression of Nkx2 genes (Prakash et al., 2009; Rubenstein et al., 1998) (Fig. 7C). However, even though these patterns of gene expression exhibit certain similarities between insects and vertebrates, it remains to be shown whether their genetic interactions are also conserved.

We thank Ethan Bier, James Castelli-Gair Hombría, Takako Isshiki, Akinao Nose, Tonja von Ohlen, Jim Skeath, Olaf Vef, Uwe Walldorf, BDGP and the Bloomington Stock Center for providing antibodies, cDNA and fly stocks. We are grateful to Gerd Technau for general support, and to Gert Pflugfelder, Ana Rogulja-Ortmann, Gerd Technau and Joachim Urban for critically reading the manuscript. This work was supported by a grant from the Deutsche Forschungsgemeinschaft to R.U. and G.M.T. (UR163/1-2, 1-4).