Segment-specific requirements for dorsoventral patterning genes during early brain development in Drosophila

The Drosophila central nervous system, which can be subdivided into ventral nerve cord (VNC) and brain, provides a useful model with which to study the generation of neural cell diversity and patterning. It develops from bilaterally symmetrical neuroectoderm, which gives rise to a fixed number of neural stem cells, called neuroblasts (NBs). NBs, which form the VNC, descend from the truncal region of the neuroectoderm, whereas those that form the brain derive from the procephalic neuroectoderm (pNE)(Hartenstein and Campos-Ortega,1984; Poulson,1950). The developmental processes that underlie NB formation,cell fate specification and pattern formation have been extensively studied in the embryonic VNC. In early stages of embryogenesis, each segment of the truncal neuroectoderm is subdivided by segment-polarity genes into transverse rows along the anteroposterior axis, and by `columnar genes' (DV patterning genes) into longitudinal columns along the dorsoventral axis. The superimposition of their gene activities establishes a Cartesian coordinate system of positional cues conferring specific identities to neural equivalence groups, of which each gives rise to a single NB (reviewed by Bhat, 1999; Skeath, 1999; Skeath and Thor, 2003). Accordingly, each NB acquires a unique identity, which is reflected by the time and position of its delamination from the neuroectoderm, by the combination of genes it expresses, and by the production of a specific cell lineage (e.g. Bossing et al.,1996; Broadus et al.,1995; Doe, 1992; Schmidt et al., 1997). Through the expression of proneural genes, each cell of an equivalence group (or proneural cluster) acquires the potential to become a NB (e.g. Skeath and Carroll, 1992). Notch signaling ensures that only a single cell adopts a NB fate and the others assume an epidermal fate (reviewed by Campos-Ortega, 1995).

The truncal neuroectoderm is subdivided along the dorsoventral axis into adjacent longitudinal columns by the expression of three genes: ventral nervous system defective (vnd) in the ventral, intermediate neuroblasts defective (ind) in the intermediate and muscle segment homeobox [msh; Drop (Dr) - FlyBase] in the dorsal neuroectodermal column(Buescher and Chia, 1997; Chu et al., 1998; D'Alessio and Frasch, 1996; Isshiki et al., 1997; Jimenez et al., 1995; Mellerick and Nirenberg, 1995; Weiss et al., 1998). Experimental data provide evidence that these homeobox genes interact in a hierarchical cascade of transcriptional repression, which is necessary to establish and maintain the sharp boundaries between their domains of expression. According to that hierarchy, vnd represses ind(and msh) in the ventral column, and ind represses msh in the intermediate column (reviewed by Cornell and Von Ohlen, 2000; Skeath, 1999).

vnd, the prototype of the Nkx2-type gene family, is expressed not only in the ventral column neuroectoderm, but also in descending NBs, and is maintained in subsets of their daughter cells. Coincident with its pattern of expression, vnd is required for specification of ventral column neuroectoderm, as well as for formation and specification of ventral NBs and their progeny (Chu et al.,1998; Jimenez and Campos-Ortega, 1990; McDonald et al., 1998; Mellerick and Nierenberg, 1995; Mellerick and Modica, 2002; Shao et al.,2002; Skeath et al.,1994). Similar roles are fulfilled by ind in intermediate, and msh in dorsal, column neuroectoderm and NBs,although msh is dispensable for the formation of dorsal NBs(Isshiki et al., 1997; Weiss et al., 1998). Owing to much higher complexity and hidden segmental organization, the mechanisms that control pattern formation in the brain are largely unknown. The brain can be subdivided (from posterior to anterior) into the tritocerebrum, deutocerebrum and protocerebrum, which derive from the intercalary, antennal and ocular(plus labral) segments, respectively (Schmitt-Ott and Technau, 1992) (reviewed by Urbach and Technau, 2003a). Recently, we have described the pattern of the entire population of brain NBs(Urbach et al., 2003) and could show that, based on the combination of genes expressed, each brain NB acquires a unique identity (Urbach and Technau, 2003b). Furthermore, we have presented first data on the expression of the columnar genes in the early embryonic brain(Urbach and Technau,2003c).

In this study, we have investigated the function of columnar genes during early brain development with the main focus on vnd. We show that vnd is required for formation of ventral brain NBs, and that cell death contributes to the loss of brain NBs in embryos lacking vndfunction. Detailed analysis of the expression of an array of molecular markers in loss- and gain-of-function backgrounds indicates that vnd is also necessary, and to some extent sufficient, for the proper specification of ventral and intermediate pNE and brain NBs. We further demonstrate that control of ind and msh expression by vnd differs among brain neuromeres and between brain and VNC. Finally, we show that the generation of normal numbers of early glial cells in the protocerebrum is independent of msh and vnd, which is in contrast to the situation in the trito-/deutocerebrum and the VNC.

The following fly strains were used: Oregon R (wild type), engrailed-lacZ (ryXho25) (Hama et al., 1990), svp-lacZ (H162)(Mlodzik et al., 1990), vnd⁶ (Jimenez and Campos-Ortega, 1990) and msh^d68(Isshiki et al., 1997). The UAS-Gal4 system (Brand and Perrimon,1993) was used to overexpress vnd in the neuroectoderm,crossing UAS-vnd (Chu et al.,1998) with the sca- Gal4 driver line(Klaes et al., 1994).

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

Embryos were dechorionated, fixed and immunostained according to previously published protocols (Patel,1994). The following primary antibodies were used:rabbit-anti-Atonal [1:5000 (Jarman et al.,1993); kindly provided by A. Jarman], rabbit anti-cleaved human caspase 3 [1:50 (Brennecke et al.,2003); Cell Signaling Technology], mouse-anti-Dachshund [1:250(Mardon et al., 1994); DSHB],rabbit-anti-Deadpan [1:300 (Bier et al.,1992); kindly provided by H. Vaessin], anti-DIG-AP (1:1000,Roche), rat-anti-Empty spiracles [1:1000(Walldorf and Gehring, 1992);kindly provided by U. Walldorf], mouse-anti-Engrailed [4D9,1:7(Patel et al., 1989); DSHB],rabbit-anti-Eyeless (1:1000, kindly provided by U. Walldorf),mouse-anti-β-Galactosidase (1:500, Promega),rabbit-anti-β-Galactosidase (1:2500, Cappel), rat-anti-Gooseberry-distal[16F12 and 10E10, 1:2 (Zhang et al.,1994); kindly provided by B. Holmgren], mouse-anti-Ladybird early[1:2 (Jagla et al., 1997);kindly provided by K. Jagla], rabbit-anti-Muscle segment homeobox (1:500;kindly provided by M. P. Scott), rabbit-anti-Repo (1:500)(Halter et al., 1995),rat-anti-Sloppy paired [1:300 (Cadigan et al., 1994); kindly provided by W. Gehring], rabbit-anti-Ventral nervous system defective [1:2000 (McDonald et al., 1998); kindly provided by C. Doe] and mouse-anti-Wingless(1:10, DSHB). The secondary antibodies (donkey anti-mouse, goat anti-rabbit;donkey anti-rat; Dianova) were either biotinylated or alkaline phosphatase-conjugated and diluted 1:500.

DIG-labeled intermediate neuroblast defective RNA probe (kindly provided by M. P. Scott) was synthesized using HindIII linearized pNB40-ind as a template with T7 polymerase, and DIG-labeled lethal of scute RNA probe was synthesized using the EST clone lsc RE59335 (BDGP)cloned in PstI linearized pFLC1 as a template with T3 polymerase according to the manufacturers protocol (Roche). The hybridization on embryos was performed as described previously(Plickert et al., 1997; Tautz and Pfeifle, 1989).

Embryos were viewed under a Zeiss Axioplan equipped with Nomarski optics using 40×, 63× and 100× oil immersion objectives. Pictures were digitized with a CCD camera (Contron progress 3012) and different focal planes were combined using Adobe Photoshop 7.0. Semi-schematic presentations are based on camera lucida drawings.

Expression of Vnd protein is first detectable in the blastoderm in bilateral longitudinal columns along the ventral neuroectoderm of the trunk and the ventral procephalic neuroectoderm (pNE), covering the prospective ventral parts of the trito-(TC), deuto-(DC) and protocerebrum (PC; Fig. 1A). Although in the trunk Vnd expression is maintained within a continuous ventral neuroectodermal column during subsequent stages, it becomes more complex and diverse among head segments. By stage 8 [before first brain neuroblasts (NBs) have developed], Vnd becomes downregulated in parts of the procephalic domain(Fig. 1B). Until stage 10, Vnd has largely vanished in the anterior pNE and NBs of the DC, but its level remains high in the ventral pNE and delaminating NBs of the TC, posterior DC and PC (Fig. 1C-E). During stages 10/11, Vnd becomes downregulated at the ventral border between TC and DC (Fig. 1F-K). Accordingly, by late stage 11, Vnd expression is restricted to separate domains at the posterior border of the TC, DC and PC, respectively(Fig. 1I-K). Whereas the number of Vnd-positive NBs in the domains of the TC and DC is rather small, a large population of about 13 NBs is found in the PC(Fig. 1J,K) (see also Urbach and Technau,2003c).

Thus, in contrast to the situation in the trunk, Vnd expression in the early brain is highly dynamic and becomes progressively confined to three separate ventral domains, encompassing different numbers of NBs and their progeny in the posterior compartments of the TC, DC and PC [for Vnd expression at later stages, see Sprecher et al.(Sprecher et al., 2006)].

To assay the presence of individual brain NBs in embryos lacking vnd function (referred to as `vnd embryos'), we used morphological criteria (size and position), the NB marker Deadpan (Dpn) and Seven up-lacZ (Svp-lacZ). Between stages 9 and 11 Svp-lacZ-positive brain NBs represent stable reference points, also allowing identification of Svp-lacZ-negative NBs surrounding them(Urbach et al., 2003). We focused mainly on the pattern of NBs at stage 9 (corresponding to the S2 NB pattern in the ventral nerve cord, VNC), and at stage 11 (corresponding to the S5 NB pattern in the VNC) (Doe,1992) when the full complement of brain NBs is established(Urbach et al., 2003). The expression of all molecular markers used in this study for the identification of individual NBs in wild-type and mutant embryos is summarized in Fig. 2 and Table 1. The analysis of large samples of embryos allowed us to quantify the data and thus to determine the variability of the phenotypes. The frequency at which each NB is found in vnd embryos is indicated in percent in Fig. 2B′,D′.

In contrast to the PC and DC and the neuromeres of the VNC, NB formation in the TC initiates after stage 9. In the PC and DC of stage 9 vndembryos almost all ventral brain NBs are improperly specified (see below) or missing. Penetrance of the effects significantly differs among the individual NBs (Fig. 2B′). For example, Dv2 (27%) fails to form about three times more often than the neighboring Dv3 [90%; for nomenclature of brain NBs, see Urbach et al.(Urbach et al., 2003)].

By stage 11, a lack of NBs becomes more apparent in the ventral domains of the TC, DC and most prominently the PC of vnd embryos stained with Dpn antibody (Fig. 3A-D). Svp-lacZ in wild type is expressed in about two thirds of the 26 ventral brain NBs (Fig. 3E). In vnd mutants, all ventral Svp-lacZ-positive, as well as the surrounding immunonegative, NBs are affected, but to different degrees(Fig. 2D,D′, Fig. 3F). By contrast, all dorsal brain NBs can be identified in vnd mutants, including Dd5, the only dorsal brain NBs that, in wild type, transiently co-express mshand vnd.

Taken together, a considerable proportion of ventral NBs is not formed in the TC, DC and PC of vnd embryos. The probability of being part of this fraction significantly differs among the individual NBs.

In order to investigate whether cell death accounts for loss of ventral brain NBs, we performed antibody staining against activated caspase 3 (Casp3). Casp3 activation is one of the early hallmarks of cell death. In the wild-type pNE, cell death first becomes detectable by late stage 10, and increases rapidly during stages 11/12 (see also Nassif et al., 1998). A comparable time course is observed in vnd embryos. However, apoptosis is significantly more pronounced in the ventral pNE of the developing DC and PC, but not TC (Fig. 4A,B). During early/middle stage 11, Casp3 staining was found in cells of the pNE and in brain NBs (Fig. 4C), at relative positions corresponding to the loss of deuto- and protocerebral NBs at later stages (compare Fig. 4C with 4D,E). We did not detect apoptotic NBs in the TC. These data suggest that apoptosis of neuroectodermal progenitor cells, and of already developed NBs in the DC and PC, partially accounts for the defects in the pattern of brain NBs in vnd embryos [for apoptosis in vndmutant brain at later stages, see Sprecher et al.(Sprecher et al., 2006)].

As shown above, the ventral pNE of vnd embryos produces a reduced set of brain NBs. In order to uncover the identity of residual NBs at stage 11, we analyzed the expression of marker genes, which are normally exclusively(Muscle segment homeobox, Msh) or preferentially (Empty spiracles, Ems;Dachshund, Dac; Ladybird early, Lbe; Eyeless, Ey) expressed in dorsal pNE and specific subsets of dorsal NBs (for wild-type marker gene expression, see Fig. 2A,C) (see also Urbach and Technau, 2003b). The combinations and frequencies in which the various markers are expressed in individual brain NBs of vnd embryos are summarized in Fig. 2B,D and in Table 1. In wild type, Msh is expressed in the dorsal pNE of the TC and DC, but not of the PC(Fig. 6A-D). In vndembryos, Msh is derepressed in the ventral pNE and residual ventral NBs in the TC and DC, but again not detected in the PC(Fig. 6E-H). In vndmutants, Ems expression becomes ectopically expanded into large parts of the ventral antennal and intercalary pNE, and into part of the residual ventral NBs (three to four in the DC and one in the TC; Fig. 5A-D), resembling the pattern of ectopic Msh. Lbe expression in the pNE of the TC (and slightly in the DC) is expanded ventrally, although we did not detect descending ectopic Lbe-positive NBs (Fig. 5E-H). Dac is derepressed in part of the ocular Vnd mutant domain(Fig. 5I-L), and we identified Dac in two to four additional ventral NBs in the PC descending from this ectopic Dac domain (Fig. 5L). Furthermore, Ey is lacking in intermediate pNE (and respective NBs) of the TC,but is ectopically expressed in the ventral pNE(Fig. 5M-P). These data indicate that the expression of several dorsal marker genes is expanded in vnd mutants into the ventral pNE and deriving NBs, suggesting that loss of Vnd at least in part leads to a ventral-to-dorsal transformation of ventral pNE and residual ventral NBs.

To find out whether `intermediate NBs', which normally derive from the insulated domains of ind expression (e.g. Td1, Td2, Td3 and Dd1), are also affected in the mutant, we investigated the expression of intermediate neuroblasts defective (ind), Ey,Gooseberry-distal (Gsb-d), Wingless (Wg) and Svp-lacZ, markers which are co-expressed in part of these precursors in wild type (see Fig. 2A,C, Table 1)(Urbach and Technau, 2003b). In vnd embryos, Td3 in most cases develops properly(Fig. 2D, Fig. 3D, Fig. 6T). By contrast, lack of ind is found in positions of Td1, Td2 and Dd1(Fig. 2D, compare Fig. 6P with 6T). Expression of the other marker genes suggests that Td2 and Dd1 are mostly formed but become improperly specified, whereas Td1 is not formed(Fig. 2D, Fig. 3D).

Together, these data indicate that Vnd is required for proper specification of ventral pNE and residual ventral, as well as for intermediate NBs.

To test whether vnd is able to alter the fate of dorsal pNE and NBs, we ectopically expressed Vnd (using sca-Gal4; embryos in the following termed sca-vnd) and examined the expression of Dac, Ems,En, Ey, Lbe, Msh and Wg, which are specifically expressed in dorsal and/or ventral parts of the pNE and brain NBs. In sca-vnd embryos, these markers are largely suppressed or completely absent in the dorsal pNE and NBs of the TC (Ems, Ey, Lbe, Msh and Wg), DC (Ems, En, Ey, Lbe, Msh, and Wg) and PC (Dac, Ems, En, Ey, and Wg) (Fig. 5Q-X, Fig. 6I-L and data not shown). The effects of ectopic Vnd on the level of individual NBs are summarized in Table 1. Susceptibility to ectopic Vnd significantly differs among dorsal NBs in a cell-specific manner. For example, Lbe is frequently missing in the dorsal Dd7(40%), and always in the dorsal Td4 (0%). The two ventral Lbe-positive NBs in the PC, Ppv3 and Pcv8, are maintained (100%), and, in addition, we usually identified ectopic Lbe in one to six adjacent NBs in the dorsal PC (compare Fig. 5F with 5T). This indicates that ectopic Vnd suppresses Lbe in the dorsal TC and DC, but confers ventral fates to a subset of dorsal NBs in the PC.

Taken together, in the vnd gain-of-function situation, we recognized a wide-ranging loss of specific marker gene expression and a partial dorsal-to-ventral transformation of dorsal pNE and corresponding brain NBs.

In wild type, the expression pattern of ind and mshdiffers between head and trunk and among head segments. ind is expressed in three separate pNE patches, of which the intercalary and antennal ind spot are located at intermediate position between the dorsal Msh and ventral Vnd domain. The domains of Msh and Vnd share a common border at sites lacking an intervening ind domain. In the ocular segment, the ind domain is spatially separated from the Vnd domain, and Msh is not expressed (Fig. 6A-D,M-P) (see also Urbach and Technau,2003c). To assay whether vnd regulates these DV patterning genes in the pNE and brain NBs, we examined the expression of Msh and ind in vnd loss-of-function and gain-of-function embryos.

In vnd mutants, ind is derepressed within the ventral part of the truncal neuroectoderm (Weiss et al., 1998). Surprisingly, we did not observe such a derepression of ind in the ventral intercalary and antennal segment(compare Fig. 6M,O with 6Q,S). Instead, expression of ind in the intercalary segment is delayed(starting after stage 9) and found in a smaller area. Even more unexpectedly, ind in the antennal pNE and Dd1 was never detected [0%, n=30; `cluster1' in McDonald et al. (McDonald et al., 1998)]. This indicates that in the antennal segment, Vnd is necessary for activating ind, rather than for repressing it. ind transcript in the ocular pNE often appears slightly expanded ventrally (70%, n=30; Fig. 6O,S).

Msh expression in vnd embryos is ectopically expanded into the ventralmost pNE of the intercalary and antennal segment and corresponding residual NBs, apparently owing to the lack of repression by vnd and ind (compare Fig. 6A-D with 6E-H). We observed derepression of Msh already before the early phase of NB formation (stage 8/9; Fig. 6E,F). This is in contrast to observations made in the trunk,where ind (instead of msh) is derepressed in the ventral neuroectoderm at these stages (McDonald et al., 1998) [although Msh is found in the ventral neuroectoderm at later stages (Chu et al.,1998)]. Early ectopic Msh is likely to `dorsalize' the ventral pNE and developing residual NBs. We found, that ectopic Msh also encompasses the region of the missing antennal ind spot, suggesting that cell fate of intermediate Dd1 (which is mostly formed) may also be affected by ectopic Msh. By contrast, ectopic Msh is largely absent at positions of the intercalary ind spot (compare Fig. 6H with 6T), indicating that, at least in the TC, the genetic interaction of ind and msh (repression of msh by ind)may be comparable with the situation in the trunk.

Upon ectopic expression of Vnd (sca-vnd) in the entire neuroectoderm, we found that Msh, before and during the phase of NB formation,is largely (15%, n=28), and sometimes completely, suppressed in the dorsal pNE and descending NBs (compare Fig. 6A-D with 6I-L). In those cases in which residual Msh is found, it is usually confined to dorsalmost ectodermal patches in the head, which, based on their position, are unlikely to give rise to brain NBs.

In sca-vnd embryos, expression of ind is lacking in the intercalary domain in 85% of cases, and diminished in the ocular domain in 37%cases (n=50), indicating that, as in the trunk neuroectoderm, ectopic Vnd is able to suppress ind expression (compare Fig. 6M-P with 6U-X). However,the antennal ind spot is almost unaffected under these conditions(observed in 92% cases by stage 10/11; n=50). Thus, indexpression in the antennal (and residual ocular) domain is less sensitive to ectopic Vnd.

Taken together, these data indicate that interactions between vnd,ind and msh differ among brain segments and compared with the VNC (Fig. 6Y). In the brain, ind is expressed in three separate segmental domains, and Msh is not expressed in the PC. Whereas in the trunk Vnd negatively regulates ind, Vnd does not repress ind (but msh) in the ventral pNE and NBs of the DC and TC. Instead, in the deutocerebral region, Vnd is necessary for activating and/or maintaining ind expression.

In the VNC, most glial cells derive from dorsal NBs(Schmidt et al., 1997), and msh is implicated in proper specification of these NBs and their progeny (Buescher and Chia,1997; Isshiki et al.,1997). Likewise, Msh is expressed in dorsal pNE and NBs of the TC and DC (Fig. 6A-D). First,glial cells in the TC and DC are closely associated with dorsal NBs, and are thus likely to derive from these progenitors(Urbach et al., 2003) (data not shown). In order to find out whether msh influences glial cell development in the brain, we counted Repo-positive cells in mshmutant embryos at early stage 12; until this stage it is possible to relate glial cells to their neuromere of origin. We observed that the number of glial cells in the DC and TC is strongly reduced, but is unchanged in the PC(Fig. 7A-D; Table 2).

Conversely, in vnd embryos, we observed an increase in the number of glial cells in the DC (>125%, n=22) and TC (>30%, n=22). In the PC, the number of glial cells seems to be unchanged, as in msh mutants (Fig. 7E,F; Table 2). The supernumerary glial cells in the TC and DC were found at more ventral positions, suggesting that they stem from ventral rather than dorsal brain NBs. This is in accordance with the derepression of Msh in the ventral pNE(described above), which may `dorsalize' residual ventral NBs. Such misspecified NBs may produce glial cells, as their dorsal counterparts normally do.

In sca-vnd embryos, the glial cell numbers in the TC and DC are significantly reduced compared with wild type (>90% of glial cells are missing in the TC and >50% in the DC, n=22), and match exactly with the numbers counted in msh mutants(Table 2). Interestingly, in sca-vnd embryos, the number of glial cells is also diminished in the PC (Fig. 7G,H; Table 2). This effect cannot be due to Msh repression, as Msh is normally not expressed in the early PC (this study) (Urbach and Technau,2003c). Instead, specification of dorsal NBs and their glial progeny in the PC must depend on other factors. Although in a vndmutant background these factors (unlike msh) do not expand the territory of their glial promoting function, they (like msh) remain repressable by ectopic Vnd expression.

Taken together, these data indicate that, in the posterior brain (TC and DC), production of normal numbers of glial cells depends on mshexpression (as in the VNC), which is ventrally delimited by Vnd (but by Ind in the VNC). This is in clear contrast to the anterior brain (PC), where early glial development is independent from msh and vnd.

The NK-2 type homeodomain protein Vnd is dynamically expressed in the pNE and the descending brain NBs. In vnd mutant embryos, we found at embryonic stages 9 and 11 that ventral NBs in the TC, DC and PC are largely absent, although at different frequencies. Analogously, in the primordium of the VNC of vnd embryos a significant loss of ventral NBs has been reported (Jimenez and Campos-Ortega,1990; Skeath et al.,1994; Chu et al.,1998; McDonald et al.,1998).

In the absence of Vnd, we found an increase in cell death, which contributes to the loss of ventral brain NBs. Apoptosis acts at the level of both pNE progenitor cells and NBs [for cell death occurring at later stages,see Sprecher et al. (Sprecher et al.,2006)]. It is not yet clear whether the reduction of ventral NBs is solely due to cell death, or whether it also involves activity of proneural genes. In the truncal neuroectoderm, proneural genes of the AS-C complex promote NB formation (reviewed by Campos-Ortega, 1995). There is evidence that vnd interacts with proneural genes, but also that it has additional function in promoting NB formation apart from activating proneural genes. The latter assumption is supported by the finding that, in vnd embryos, lethal of scute (l'sc) can still be expressed in the ventral proneural clusters of, for example, NB5-2, although the respective NB is missing (Skeath et al., 1994; Chu et al.,1998). In the pNE, genes of the AS-C complex are expressed in large proneural domains, of which those of achaete, but especially of l'sc (Urbach et al.,2003; Younossi-Hartenstein et al., 1996), seem to overlap with the vnd expression domain, suggesting a possible genetic interaction. However, in vndembryos, we observe no substantial differences in the expression pattern of l'sc transcript compared with the wild type (R.U., unpublished). Thus, similar to the situation in the trunk, Vnd does not appear to exert proneural function through activation of l'sc. However, our data propose a possible interaction between vnd and the proneural gene atonal. In vnd mutants, expression of atonal is often missing in proneural clusters of the sensory organ precursors of the hypopharyngeal-latero-hypopharyngeal organ (see Fig. S1 in the supplementary material). Clearly, further investigations are required to clarify in how far interactions between vnd and proneural genes play a role in the formation of ventral brain NBs.

We noticed that, in vnd mutants, not only ventral, but also intermediate brain NBs in the TC and DC show defects in their formation or specification, comparable with the situation in the trunk(Jimenez and Campos-Ortega,1990; McDonald et al.,1998; Mellerick and Modica,2002). As intermediate brain NBs do not express vnd (but ind), these defects appear to be non-cell-autonomous. Another, more likely explanation is that determination occurs at the blastodermal stage,when Vnd is transiently expressed in a much larger population of cells in the pNE, which presumably include progenitors of intermediate NBs. A similar proposal was made for intermediate NBs in the trunk(McDonald et al., 1998). Furthermore, early commitment of ventral neuroectodermal cells and cell-autonomous expression of ventral and intermediate NB fates has been demonstrated by heterotopic transplantations of neuroectodermal cells from ventral to dorsal sites at the early gastrula stage(Udolph et al., 1995).

In the trunk, a segmentally reiterated combinatorial code of genes expressed within each particular proneural cluster specifies the individual identity of the NB it gives rise to. These include DV patterning genes and segment polarity genes, which provide positional information in the neuroectoderm, as well as a number of other factors (reviewed by Bhat, 1999; Skeath, 1999; Skeath and Thor, 2003). Most of these genes are also expressed in specific domains of the pNE before NBs delaminate, although in a segment-specific manner(Urbach and Technau, 2003b). Our present data show that Vnd influences the expression of such site-specific marker genes ('NB identity genes') already in the pNE, before NBs are formed. In vnd embryos, we noticed a derepression of dorsal-specific genes in the ventral pNE (e.g. of msh and ems in the intercalary and antennal segment, and dac in the ocular segment) and in the descending NBs, and conversely, a loss of ventral-specific gene expression(e.g. lbe in the PC). The altered expression of `NB identity genes'in vnd mutants reflects a ventral-to-dorsal transformation of ventral pNE and residual NBs. Further evidence for such a transformation is the production of (ectopic) glial cells by these ventral NBs, which normally is a trait specific to dorsal NBs (see below). By contrast, in the trunk, absence of Vnd results in a ventral-to-intermediate transformation, owing to the derepression of ind (instead of msh in pNE), which induces specification of intermediate NB fates(Chu et al., 1998; Weiss et al., 1998).

Together, our data in the vnd loss- and gain-of-function backgrounds indicate that vnd is required, and is at least partially sufficient, for the induction of ventral fate in brain NBs through the activation of genes specific for the ventral pNE, and through the repression of genes specific for dorsal pNE.

Our analysis revealed differences in the regulation of DV patterning genes among the intercalary (IC), antennal (AN) and ocular (OC) head segments,giving rise to the TC, DC and PC, respectively. Overexpression of vndleads to repression of ind within the IC, but loss of vnd-function does not seem to cause ventral expansion of the ind intercalary spot. Unexpectedly, ind is completely absent in the AN of vnd mutants, suggesting that in this segment vnd is necessary for activation and/or maintenance of ind(rather than repression). This is supported by our previous finding that the ind antennal spot transiently co-expresses Vnd(Urbach and Technau, 2003c),which is unique in the neuroectoderm, and by our present finding that in vnd gain-of-function background the ind antennal spot is almost unaffected. In the OC, however, ind expression is partially repressed upon Vnd overexpression, and ventrally expanded in the absence of Vnd, similar to the situation in truncal segments. However, because, in wild type the ind ocular spot does not adjoin the ocular vnddomain (Urbach and Technau,2003c), its expansion in vnd embryos cannot be due to a cell-autonomous effect.

Overexpression of vnd abolishes Msh almost completely in the neuroectoderm of all body segments. Yet, absence of Vnd reveals segment-specific differences in the regulation of msh. Owing to insulated ind expression in the IC and lack of ind in the AN of vnd mutants, Msh (instead of ind) is found in the ventral pNE of these segments, which is unique in the CNS anlagen, except for the mandibular segment, which exhibits equivalent expression (compare Fig. 6A,M with 6E,Q).

Among the pregnathal segments, the degree of conservation with regard to the expression and interactions of DV patterning genes seems to be highest in the posterior IC (TC) [ind and msh being repressed by(ectopic) vnd, and msh by ind]. In the anterior head, endogenous Msh expression in the dorsal pNE reaches the segmental border between AN (DC) and OC (PC), but does not cross it(Urbach and Technau, 2003c). Ectopic Msh in vnd mutants does also not cross this border, which suggests interference with regulatory factors acting in AP axis.

Significant differences between the anterior head segments and the trunk have also been reported for the initial mode of activation and cross-regulatory interactions of segment-polarity genes(Gallitano-Mendel and Finkelstein,1997).

As discussed above, in the pNE, vnd is necessary for the formation and specification of brain NB. It remains to be shown whether ind and msh exert analogous functions. However, more than 50% of the identified brain NBs do not express any of the three DV patterning genes(Urbach and Technau, 2003c). Most of these NBs derive from pNE of the preantennal head, which implies that further factors are involved in DV patterning of the anterior pNE and brain. Several other genes have been reported to be crucial for DV patterning in the truncal neuroectoderm, such as the EGF-receptor homolog Egfr(Skeath, 1998; Udolph et al., 1998; Yagi et al., 1998), the Sox genes SoxNeuro and Dicheate(Buescher et al., 2002; Sanchez-Soriano and Russell,2000; Zhao and Skeath,2002), and Nk6 (Uhler et al., 2002). For most of them it has been shown that they are involved in formation and/or specification of truncal NBs. Egfr, both Sox genes (J.S. and R.U., unpublished) and Nk6(Uhler et al., 2002) are also expressed in the pNE, before and during the phase of NB formation. However, in Egfr mutant embryos the number and pattern of brain NBs is unaffected(Dumstrei et al., 1998). How far the Sox genes and Nk6 contribute to the formation and/or specification of brain NBs awaits further investigation.

Most of the glial cells in the VNC derive from dorsal NBs (neuroglioblasts or glioblasts) (Schmidt et al.,1997), which depend on msh for proper specification. Accordingly, glial cells deriving from these progenitors are missing or improperly differentiated in msh mutants(Isshiki et al., 1997), as well as in sca-vnd embryos(Mellerick and Modica, 2002). Likewise, in the TC and DC, first glial cells are closely associated with dorsal NBs that descend from Msh-expressing pNE(Urbach and Technau, 2003c). In the TC, some dorsal NBs have been identified as glial progenitors, e.g. the neuroglioblast Td4 and the glioblast Td7, which are putative serial homologs of the truncal neuroglioblast NB5-6 and the glioblast LGB, respectively(Urbach et al., 2003). In absence of Vnd, we found the number of glial cells in the TC, and especially in the DC, to be increased. This is most probably due to the segment-specific early derepression of Msh in the ventral pNE and NBs of the TC and DC. In the truncal segments, however, ind instead of Msh is derepressed in the ventral NE (Chu et al., 1998; McDonald et al., 1998), and the number of glial cells is not significantly affected in the VNC of vnd mutants (data not shown). Furthermore, in msh mutants,glial development in the TC and DC is almost completely abolished, which parallels the phenotype observed upon vnd overexpression (leading to repression of msh in the dorsal pNE and NBs). Thus, comparable with the situation in the VNC, Msh promotes glial fate in the TC and DC. However,in the PC, glial development must be regulated differently (at least in its early phase). Until stage 12 we did not detect Msh in this part of the brain,and in msh mutants the number of glial cells in the PC is normal. Glial cell fate in the PC is also not affected by loss of vnd,although it remains repressable by ectopic Vnd. Therefore, as opposed to the TC and DC, and to the VNC, normal production of early glial cells in the PC does not depend on msh, nor indirectly on vnd or ind.

There are striking similarities in the spatial order of expression of vnd, ind and msh in the Drosophila neuroectoderm and homologous genes in the neural plate and neural tube of vertebrates: vnd homologs of the Nkx2 family are expressed in ventral regions; the ind homologs Gsh1 and Gsh2 are expressed in the intermediate regions; the msh homologs Msx1,Msx2 and Msx3 are expressed in the dorsal region of the neural tube (reviewed by Arendt and Nübler-Jung, 1999; Cornell and Ohlen, 2000). This dorsoventral order of expression is conserved not only in the anlagen of the truncal CNS but also in those that form the posterior part of the brain (in Drosophila, TC and DC; in vertebrates, hindbrain)(Urbach and Technau, 2003c). Moreover, the anterior borders of the expression domains of these columnar genes correspond in the early brains of Drosophila and mouse:expression of vnd/Nkx2 extends most rostrally (mouse ventral forebrain), followed by ind/Gsh1 and, finally, msh/Msx3expression (reviewed by Urbach and Technau, 2004). Thus, the expression of columnar genes in the brain is, to some extent, evolutionarily conserved both along the DV axis and along the AP axis.

In this study, we have presented evidence that in Drosophila vndmutant embryos a large fraction of ventral brain NBs is missing, and that ventral pNE and residual ventral NBs show significant traits of a ventral-to-dorsal transformation owing to derepression of msh (as opposed to ind in the VNC). Again, this displays obvious similarities to findings made in mice carrying a deletion of Nkx2.1. Consistent with the pattern of expression in wild type, in the mutant embryonic brain a substantial loss of ventral (especially forebrain) structures has been observed (Kimura et al., 1996; Sussel et al., 1999). Moreover, ind/Gsh2 expression is not expanded in Nkx2.1 mutants(Corbin et al., 2003)(reviewed by Rallu et al.,2002), and residual basal (ventral) pallidal structures become transformed into dorsal striatal structures(Sussel et al., 1999). Thus,in both Drosophila and mouse, loss of vnd/Nkx2 in the brain leads to a transformation of ventral into dorsal structures, rather than into intermediate structures, which has been shown to be the case in the truncal CNS of both species (Briscoe et al.,1999; Chu et al.,1998; McDonald et al.,1998). Therefore, in the developing brains of Drosophilaand vertebrates, vnd/Nkx2 is crucial for the formation and specification of ventral brain structures, and interacts with other dorsoventral patterning genes in a region-specific manner.

We are grateful to Frank Hirth, Heinrich Reichert, Ana Rogulja-Ortmann and Simon Sprecher for critical comments on the manuscript; and to Marita Buescher, Bill Chia, Thomas Czerny, Chris Doe, Walter Gehring, Frank Hirth,Bob Holmgren, Krzysztof Jagla, Andrew Jarman, Akinao Nose, Ward Odenwald,Heinrich Reichert, Matthew Scott, Simon Sprecher, Harald Vässin, Uwe Walldorf and the Bloomington stock center for providing antibodies, cDNA and fly stocks. This work was supported by a grant from the Deutsche Forschungsgemeinschaft to R.U. and G.M.T.