The columnar gene vnd is required for tritocerebral neuromere formation during embryonic brain development of Drosophila

In Drosophila, molecular genetic studies have identified genetic cascades that pattern and specify the embryonic ventral nerve cord (VNC) along the anteroposterior (AP) and dorsoventral (DV) axes. These studies show that segment polarity gene activity divides each segment along the AP axis into four parallel transverse rows, whereas columnar gene activity results in a tripartite dorsoventral subdivision of the neuroectoderm into longitudinal domains (Bhat, 1999; von Ohlen and Doe, 2000). Thus, ventral nervous system defective (vnd) is expressed in ventral neuroectodermal cells (Jimenez et al., 1995; Mellerick and Nirenberg, 1995), while intermediate neuroblasts defective (ind) expression is restricted to intermediate neuroectodermal cells (Weiss et al.,1998) and muscle specific homeobox (msh) is expressed in lateral neuroectodermal cells(D'Alessio and Frasch, 1996; Isshiki et al., 1997). These neuroectodermal cells in turn give rise to the three columns of ventral,intermediate and lateral neuroblasts. The different fates of the individual neuroblasts that form in different dorsoventral columns are influenced by expression of the corresponding columnar genes (reviewed by Skeath, 1999; Skeath and Thor, 2003).

The functional role of the columnar genes is exemplified by vnd,an Nkx2-type homeobox gene, which is continuously expressed within the developing VNC from cellularization until the completion of embryonic development. In vnd loss-of-function mutants, ventral neuroblasts are absent or mis-specified. These precursor cell defects correlate with loss or mis-specification of neuronal progeny, errors in axonal pathfinding and an overall reduced number of cells in the developing VNC of vnd mutant embryos. Thus, in vnd loss-of-function mutants, commissures are fused, ventral unpaired medial neurons show pathfinding defects and midline glia are reduced in number. Conversely, overexpression of vnd can lead to transformations in the identity of intermediate and lateral neuroblasts (Chu et al., 1998; McDonald et al., 1998; Mellerick and Modica,2002).

In contrast to the situation in the embryonic VNC, much less is known about the expression and function of the columnar gene vnd in the developing brain of Drosophila. A recent study on early brain neurogenesis shows that vnd is expressed in the procephalic neuroectoderm as well as in subsets of identified brain neuroblasts(Urbach and Technau, 2003a; Urbach and Technau, 2003b). However, information on the expression and function of vnd during later embryonic brain development is lacking.

Here, we analyse the role of the columnar patterning gene vndduring embryonic brain development of Drosophila. Using immunocytochemistry, we first map the expression of vnd and show that it is confined to subsets of neural cells in the developing protocerebrum,deuterocerebrum and tritocerebrum. We then carry out a functional analysis of vnd focussed on the intercalary segment and tritocerebral neuromeres,and show that vnd is essential for the formation of the tritocerebrum in the developing brain. Thus, loss of vnd function leads to a gap-like brain phenotype that is due to the defective formation of a subset of tritocerebral neuroblast and the subsequent loss of neural tissue in the mutant domain. Moreover, we show that this loss of neural tissue is associated with increased apoptotic activity, resulting in the loss of the tritocerebral commissure and the longitudinal connectives that normally run through this neuromere, whereas blocking apoptosis in vnd-null mutant embryos results in partial restoration of tritocerebral cells and axon tracts. These findings suggest that the dorsoventral patterning gene vnd is essential for development and identity specification of tritocerebral neural lineages in embryonic brain development of Drosophila.

The wild type was Oregon-R. For vnd mutant analysis, the null allele vnd⁶ (Jimenez and Campos-Ortega, 1990) was used, balanced over FM7,ftz-lacZ. Homozygous null mutants were identified by the absence of ftz-lacZ. To analyse the development of the tritocerebral lab expression territory in the vnd mutant background, we used the line 7.31 lab-lacZ/7.31 lab-lacZ(Tremml and Bienz, 1992)crossed into vnd⁶. To identify vnd expression in former lab-expressing tritocerebral cells in the labialmutant background, we used line 7.31 lab-lacZ/7.31 lab-lacZ;lab^vd1/TM3, hb-lacZ(Tremml and Bienz, 1992). Homozygous null mutants were identified by the absence of hb-lacZ. For comparison with the wild-type situation, 7.31 lab-lacZ was crossed back to wild type. 7.31 lab-lacZ shows cytoplasmic distribution of βgal and reflects endogenous labexpression with additional ectopic expression patterns in the deutocerebral anlage (Hirth et al., 1998; Hirth et al., 2001). The UAS/Gal4 transcriptional activation system(Brand and Perrimon, 1993) was used in order to perform overexpression experiments. As a Gal4-driver line, we used the sca::Gal4 (Klaes et al.,1994; Sprecher et al.,2004) which in the embryonic CNS is active from neurectodermal stage to early neural lineage formation. For the ubiquitous block of apoptosis within neural lineages of vnd⁶ mutant embryos, we used the sca::Gal4 driver line in order to activate UAS::p35transcription (Mergliano and Minden,2003). Embryos were staged according to Campos-Ortega and Hartenstein (Campos-Ortega and Hartenstein,1997).

Embryos were dechorionated, fixed, immunostained, flattened and staged according to previously published protocols(Patel, 1994; Therianos et al., 1995; Urbach et al., 2003). Primary antibodies were rabbit anti-Deadpan [1:300(Bier et al., 1992); kindly provided by H. Vaessin], rabbit anti-HRP [FITC-conjugated; 1:100(Jan and Jan, 1982); Jackson Immunoresearch], rabbit anti-LAB at 1:100 (F.H., unpublished), rat anti-LAB at 1:500 (F.H., unpublished), mouse anti-NRT at 1:20 (BP106 antibody, DSHB),rabbit anti-VND at 1:200 (McDonald et al.,1998) (kindly provided by C. Q. Doe), rabbit anti-βGAL 1:200-1:400 (Milan Analytika), mouse anti-βGAL 1:50 (DSHB), mouse anti-Fasciclin II 1:5 (Lin and Goodman,1994), rat anti-ELAV 1:30 (DSHB), mouse anti-PROS 1:4(Spana and Doe, 1995), mouse anti-REPO 1:20 (DSHB), mouse anti-Engrailed [4D9,1:6(Patel et al., 1989); DSHB]. Secondary antibodies used for confocal microscopic analysis were Alexa-488,Alexa-568 and Alexa-647 antibodies generated in goat (Molecular Probes), all at 1:150 dilution. Secondary antibodies used for flat-mount preparations analyzed using Nomarski optics were either biotinylated or alkaline phosphatase-conjugated antibodies generated in goat all at 1:500 (Dianova). Apoptotic activity was assayed by TUNEL analysis using a commercial TUNEL kit(ApoTag, Oncor) as previously described(Richter et al., 1998) with the following modifications. After fixation, embryos were washed in PBT for 2×5 minutes, then washed in Equilibration Buffer (from the ApoTaq kit)for 2 minutes. Embryos were incubated in the working strength TdT mixture(from the ApoTaq kit) for 1 hour at 37°C. After incubation, supernatant was removed and embryos were washed for 2×2 minutes with Stop/Wash solution (from the ApoTaq kit), and subsequently washed for 3×2 minutes,2×30 minutes in PBT before starting immunolabelling. Embryos were mounted in Vectashield H-1000 (Vector).

For laser confocal microscopy, a Leica TCS SP was used. Optical sections ranged from 0.2-1.5 μm recorded in line average mode with picture size of 512×512 pixels, or 1024×1024 pixels. Captured images from optical sections were arranged and processed using IMARIS (Bitplane). Complete series of optical sections were imported and processed using ImageJ. For the generation of 3D digital models, raw tiff stacks (stacks of optical sections)were imported into AMIRA (Mercury Computer Systems). User-defined materials were drawn manually around the labelled structures (immunoreactivity of a given antibody) in each layer of a given tiff stack, which were to be included in the model. Subsequently, the program synthesizes a surface by triangulation around the defined materials, which was further processed by increasing the number of triangles per material and smoothening of the surface. For 2D representations in figures, screen shots were generated showing the model in the appropriate angle, virtual lighting and transparency of individual material surfaces (for details, see Pereanu and Hartenstein,2004). Figures were arranged and labelled using Adobe Photoshop.

During an initial phase of embryonic neurogenesis, vnd expression is seen in the neurectoderm and delaminating neuroblasts in the ventral domains of the protocerebral, deutcerebral and tritocerebral brain neuromeres[for a detailed description of vnd expression during the early phase of neurogenesis, Urbach et al. (Urbach et al., 2006)]. During later stages of embryogenesis, vndexpression is also seen in specific cell clusters within these brain neuromeres. Thus, at late stage 12, a large expression domain is seen in the protocerebral neuromere (Fig. 1A,B) and two smaller expression domains are observed in the deutocerebrum and in the tritocerebrum. Although the tritocerebral and deuterocerebral vnd expression clusters are in close proximity to each other, they do not overlap. Towards the end of embryogenesis, at embryonic stage 15, expression of vnd is still visible in these three neuromeric domains (Fig. 1C,D). Throughout embryonic neurogenesis, we find vnd expression in neuroblasts, ganglion mother cells and neurons as judged by immunolabelling with anti-PROS and neuron-specific anti-ELAV antibodies. Immunolabelling with glia-specific anti-REPO antibody indicates that none of the glia cells of the embryonic brain express vnd (data not shown). vnd-expressing cells are also seen in the neuromeres of the suboesophageal ganglion and the VNC, as well as in peripheral sense organs; these vnd-expressing cells will not be considered further in this report.

Mutational inactivation of vnd results in a pronounced brain phenotype in the late stage embryonic brain. Immunolabelling with neuron-specific anti-HRP and anti-ELAV antibodies identifies a large gap separating the anterior deutocerebral brain region from the neuromeres of the posterior suboesophageal ganglion (Fig. 2A,B). Evaluation of the penetrance of the vnd-null mutant phenotype reveals that in 36.5% of the mutant embryos (27/74) this gap is completely devoid of Elav- and HRP-immunoreactive cells (not shown), while in the majority of vnd mutant embryos (63%; 46/74) a thin strand of ELAV- and HRP-immunoreactive cells remains and interconnects the protocerebrum and the suboesophageal ganglion (Fig. 2B). This cell loss is associated with axonal patterning defects in the embryonic brain. The longitudinal connectives that normally run from the protocerebrum to the suboesophageal ganglion are missing or strongly reduced (Fig. 2, arrows) and the tritocerebral commissure is completely absent. Glia-specific anti-REPO immunoreactivity reveals that glial cells are present in the mutant but fail to be correctly localized in the affected region, most probably owing to the absence of neuronal tissue (Fig. 2C,D).

To delineate the region affected in vnd mutants in more detail, we studied the expression of engrailed (en), which in the wild-type embryonic brain is located in several small clusters of cells that demarcate the posterior boundary of the brain neuromeres(Schmidt-Ott and Technau,1992; Hirth et al.,1995). The b1 en-stripe (or en head spot)delimits the posterior protocerebrum (several en cells are also seen more anteriorly in the protocerebrum as the secondary head spot), the b2 en-stripe (or en antennal stripe) delimits the posterior deutocerebrum, and the b3 en-stripe (or en intercalary stripe) delimits the posterior tritocerebrum(Fig. 2E). In late vndmutant brain (embryonic stage 13 onwards), only the b1 en-stripe and the secondary head spot are visible; neither the b2 en-stripe nor the b3 en-stripe can be identified(Fig. 2F). This supports the observation that major parts of the embryonic tritocerebrum and parts of the deutocerebrum are lacking in the vnd mutant. In addition to the cell loss defect in the tritocerebral/deutocerebral brain region, a less marked reduction in overall size of the protocerebrum is also seen in vndmutant embryos. Moreover the organization of the suboesophageal ganglion and the VNC is affected in the vnd mutant (see also Mellerick and Modica,2001). These latter two phenomena were not studied further.

At the gross histological level, the vnd mutant brain phenotype described above is, in part, reminiscent of the mutant brain phenotype observed for the Hox gene labial (lab). In lab-null mutants, tritocerebral cells are generated and positioned correctly; however,these cells fail to differentiate into neurons and marked axogenesis defects occur, including the disruption of longitudinal connectives and lack of the tritocerebral commissure (Hirth et al.,1998; Page, 2000; Hirth et al., 2001). As lab and vnd also show overlapping expression in a subset of tritocerebral neuroblasts (Urbach and Technau, 2003a), these findings suggest that lab-expressing tritocerebral neuroblasts are affected in vndmutant embryos. To investigate this, we focussed on the developing tritocerebrum, and specifically on the lab expression domain of this neuromere, and first determined whether loss of vnd function affects formation of lab-expressing neuroblasts.

During the early phase of brain neurogenesis, the lab-expressing neuroectodermal domain gives rise to 15 neuroblasts, which include all of the tritocerebral neuroblasts and two deutocerebral neuroblasts(Urbach and Technau, 2003a). By stage 11, all of these neuroblasts are present and express lab;they include a ventral group of tritocerebral neuroblasts, Tv1-Tv5, a more dorsal group of tritocerebral neuroblasts, Td1-Td8, and two deutocerebral neuroblasts, Dv2 and Dv4 (Fig. 3A,B). In the wild type, the most ventral part of the neuroectodermal domain, from which the tritocerebral neuroblasts Tv1-Tv5 and the two deutocerebral neuroblasts originate, dynamically co-expresses lab and vnd between stages 8 and 11(Urbach and Technau,2003a).

In vnd mutants this ventral-most part of the lab-expressing domain appears to be reduced in size and accordingly,the number of lab-expressing neuroblasts that derive from this brain region is diminished (Fig. 3C,D). Generally only four to six large rounded cells are observed that co-express lab and the neuroblast-specific marker Deadpan (this may be a slight underestimate as a few enlarged rounded cells in sub-ectodermal position lacking Deadpan expression are sometimes observed in this region). Based on the expression of molecular markers indicative of dorsal neuroblasts [e.g. ladybird early, empty spiracles, wingless(Urbach and Technau, 2003a; Urbach et al., 2006)], this reduction in lab-expressing neuroblasts appears to affect preferentially ventral neuroblasts of the tritocerebrum and adjacent part of the deutocerebrum. These data imply that vnd is required for the formation of a ventral subset of lab-expressing neuroblasts in the developing tritocerebrum.

Although the reduction in tritocerebral neuroblast number seen in vnd mutants can account for some of the tritocerebral defects, this mechanism alone is unlikely to be the exclusive cause for the massive cell loss phenotype observed in the late embryonic vnd mutant brain. This is because a dorsal subset of the lab-expressing tritocerebral neuroblasts, as well as large number of lab-expressing neural progeny are generated in the tritocerebrum of stage 11 vnd mutant brains(compare Fig. 3B with 3D). Hence, in addition to defective neuroblast formation, other phenomena must be responsible for the gap-like phenotype observed in vnd mutant brains,implying that vnd is required also later in embryogenesis - either by acting directly on lab expression in tritocerebral cells or through a lab-independent requirement.

To investigate this, we first determined whether vnd and lab show overlapping expression during later stages of tritocerebral neuromere formation. Immunocytochemical analysis indicates that a partial overlap of vnd and lab expression persists in the differentiating tritocerebrum throughout embryogenesis and is prominent in the ventral region (according to neuraxis) of this neuromere(Fig. 4A,B). Next, we analysed lab expression in late vnd loss-of-function mutant brains. Owing to extensive cell loss in the vnd mutant tritocerebrum, this analysis was limited to the remaining strand of cells that interconnects the protocerebrum and the remaining part of the deutocerebrum with the suboesophageal ganglion. Despite the extensive cell loss seen in vndmutant brains, remaining cells of the interconnecting strand do show lab expression (Fig. 4C,D).

We next investigated whether expression of vnd occurs in the lab mutant tritocerebrum by studying lab loss-of-function mutants. For this, we took advantage of the fact that in lab-null mutants, cells in the tritocerebral mutant domain are generated and can be visualized by a 7.31 lab-lacZ reporter construct(Tremml and Bienz, 1992; Hirth et al., 1998). Surprisingly, despite the lack of expression of neuronal differentiation markers in cells of the lab mutant domain (see also Hirth et al., 1998), vnd is expressed normally and shows partial overlap with tritocerebral lab mutant cells, as visualized by the lab-specific reporter construct(Fig. 4F). This indicates that expression of vnd is not affected by the absence of labduring late stages of embryonic brain development.

What is the nature of the vnd requirement for proper tritocerebral neuromere development during later stages of embryogenesis? One possibility is that vnd might be involved in maintaining differentiating cells during tritocerebral neuromere formation. To investigate this, we focussed on the tritocerebral cells that are labelled by the 7.31 lab-lacZreporter construct and compared wild-type with vnd null mutant embryonic brains. At stage 12, expression of this reporter construct in a wild-type background reflects endogenous lab expression in the tritocerebrum (Fig. 5A,B,D,E). A comparison of reporter gene expression in wild type(Fig. 5B) with reporter gene expression in vnd-null mutants(Fig. 5C) at early stage 12 indicates that the lab-expressing tritocerebral domain is somewhat reduced in size but not deleted in vnd mutants (see also Fig. 3B,D). By contrast, by late stage 12, the lab-expressing tritocerebral domain is significantly reduced in vnd mutants(Fig. 5F). Thus, an average of 27 cells expressing the lab-lacZ reporter gene are detectable in the vnd mutant tritocerebrum when compared with an average of 43 cells detectable in wild type (Fig. 5K). This suggests that in addition to defective neuroblast formation, loss of vnd function affects proper development of neural cells in the tritocerebral lab-expression domain.

Analysis of programmed cell death reveals that this loss of neural cells in the lab-expressing domain is associated with increased apoptotic activity in the developing tritocerebrum. In early stage 12 wild type, low level apoptotic activity is associated with the tritocerebral lab-expressing domain as assayed by TUNEL staining(Fig. 5G). [A low level of apoptosis is seen throughout the developing brain at this stage(Abrams et al., 1993; Nassif et al., 1998).] In early stage 12 vnd mutants, a significant increase in the number of TUNEL-positive apoptotic cells is seen associated with the tritocerebral lab-expressing domain when compared with wild type and this correlates with a decrease in the number of lab-expressing cells in this region (Fig. 5H). Accordingly, by late stage 12, a significant reduction of the number of anti-LAB positive cells has occurred in vnd mutants when compared with wild type (Fig. 5I,J). Thus, on average, 20 TUNEL-positive apoptotic cells are identifiable within the vnd mutant tritocerebral lab-expression domain when compared with 10 apoptotic cells in wild type(Fig. 5K). These findings suggest that marked reduction of tritocerebral lab-expressing cells occurs as early as embryonic stage 12 and that this reduction is caused by increased apoptosis of neural tissue.

In order to further substantiate these observations, we next determined whether blocking apoptosis can prevent cell loss in the vnd mutant brain. For this, we used the UAS/Gal4 transcriptional activation system(Brand and Perrimon, 1993) in order to perform misexpression experiments in vnd mutant embryos. For the ubiquitous block of apoptosis from neurectodermal stage to early neural lineage formation, we used the sca::Gal4 driver line(Sprecher et al., 2004) in order to activate UAS::p35 transcription, an inhibitor of cell-death effector caspases (Mergliano and Minden,2003), in a vnd-null mutant background. When compared with wild type (Fig. 6A) and vnd mutant brain (Fig. 6B), ubiquitous block of apoptosis leads to a remarkable restoration of HRP-immunoreactive tissue in the vnd mutant tritocerebrum (Fig. 6C). Notably, and in contrast to the vnd mutant situation, descending longitudinal connectives that transverse the tritocerebrum are detectable(Fig. 6C, arrow). Moreover, a pronounced lab-expression domain is observed in the cell death prevented vnd mutant tritocerebrum (compare Fig. 6D with Fig. 4B). These data suggest that lab and vnd act in a genetically independent manner in tritocerebral neuromere development: the expression of lab appears to be largely unaffected by the absence of vnd as long as apoptosis is prevented.

To characterize in more detail the extent of neural tissue restoration of the vnd mutant brain phenotype resulting from block of apoptosis, we carried out immunocytochemistry using antibodies against Fasciclin II (FAS2)(Lin and Goodman, 1994) and ELAV. In the wild-type embryonic brain, anti-FAS2 immunostaining labels a number of early differentiating neurons, as well as axon tracts(Fig. 7A), whereas anti-ELAV labels differentiating postmitotic neurons(Fig. 7B). In the vndmutant tritocerebrum, both FAS2 and ELAV-positive cells, as well as FAS2-immunoreactive longitudinal axon tracts, are severely perturbed or lacking (Fig. 7C,D; arrow). When compared with wild-type and vnd mutant brain, ubiquitous block of apoptosis significantly reduces the gap-like defects observed in vnd loss-of-function mutants. Thus, FAS2-immunoreactive longitudinal connectives are detectable and form a continuous band along the anteroposterior neuraxis. However, neural fibres contributing to longitudinal axon tracts display fasciculation defects and appear only loosely bundled when compared with wild type (Fig. 7E, arrow). Moreover, a large number of ELAV-positive cells are detectable in the cell death prevented vnd mutant tritocerebrum(Fig. 7F, arrow). These data provide further evidence that the gap-like brain defect observed in vnd mutants is largely due to increased apoptotic activity and the subsequent loss of neural cells in the tritocerebral domain.

Previous analyses have demonstrated that the genes vnd, ind and msh are required for the columnar subdivision of the VNC neuroectoderm and subsequent formation and determination of neuroblasts along the dorsoventral axis during Drosophila embryogenesis (reviewed by Skeath and Thor, 2003). In the case of vnd, detailed studies have shown that vnd is required for specification of the ventral neuroectodermal column and specific neuroblasts, and that absence or misspecification of ventral neuroblasts correlates with the loss or mis-specification of neural progeny. For example,the aCC/pCC and dMP2/vMP2 neurons are lost, and the RP2 neuron is frequently absent in vnd mutant embryos, resulting in axonal pathfinding defects and defective commissure formation in the developing VNC(Jimenez et al., 1995; Chu et al., 1998; McDonald et al., 1998; Mellerick and Modica,2002).

In this study, we show that the vnd gene is also required during embryonic brain development. Starting at early stages of neuroectoderm and brain neuroblast formation, vnd expression is seen in the developing neuromeres of the embryonic brain (Urbach and Technau, 2003b; Urbach et al., 2006) and in the subsequent course of embryogenesis, vnd expression is seen in specific clusters of differentiating neural cells in each brain neuromere. For an investigation of the functional role of vnd in brain development, we focused on the tritocerebral neuromeres,where our mutant analysis suggests that vnd acts at least during two important steps in its development: precursor cell development and neural progeny maintenance. Thus, early in neurogenesis, a ventral subset of tritocerebral neuroblasts are lacking in vnd mutants, suggesting that vnd is required for the formation of neuroblasts in the developing tritocerebrum. Later in embryogenesis, vnd mutants display a severe loss of neural tissue, together with axonal patterning defects in the tritocerebrum. This gap-like phenotype is associated with increased apoptotic activity, and blocking apoptosis in vnd-null mutant embryos results in partial restoration of tritocerebral cells and axon tracts.

Together, these data suggest that, similar to its role in development of the embryonic VNC (Jimenez et al.,1995; Chu et al.,1998; McDonald et al.,1998; Mellerick and Modica,2002), vnd activity is required for proper formation and identity specification of tritocerebral neural lineages. Accordingly, our observation of increased apoptotic activity in the vnd mutant tritocerebrum might be a consequence of identity changes imposed on vnd deficient neural lineages. Indeed, vnd loss-of-function embryos are characterized by the ectopic expression of dorsal marker genes such as Ems and Msh in residual ventral neuroblasts of the vnd mutant tritocerebrum (Urbach et al.,2006). It is therefore conceivable that ectopic activity of genes such as Ems and Msh, together with the absence of vnd activity ultimately impose identity changes on residual tritocerebral vndmutant lineages that are incompatible with their survival and eventually lead to cell death. In this sense, the gap-like brain phenotype observed in late vnd mutant embryos can be regarded as a consequence both of defects in neuroblast formation and of improper identity specification of tritocerebral neural lineages during embryonic brain development.

Our results indicate that the homeotic gene lab, which is part of the anteroposterior patterning system, and the columnar gene vnd,which is involved in dorsoventral patterning, act in an integrated manner but independently in the formation and specification of the tritocerebral neuromere. Although vnd and lab show overlapping expression in tritocerebral neuroblasts (Urbach and Technau, 2003b; Urbach and Technau, 2004) and subsequently in neural cells of the posterior tritocerebrum (this study), expression of vnd appears unaffected in lab mutant cells. Conversely, vnd does not act on lab expression as the complete absence of lab expression in vnd mutants (with the exception of a rare thin strand of neuronal cells) reflects a secondary defect because of the absence of cells that normally express lab. This independent genetic activity of vnd and lab is further supported by the fact that blocking apoptosis restores tritocerebral lab expression in vnd-null mutant embryos.

Thus, although the lab and vnd mutant brain phenotypes result in comparable axonal patterning defects (loss of the tritocerebral commissure and perturbation of the longitudinal connectives that normally run through this neuromere), their mode of action within the developing tritocerebrum is discriminable. Our results suggest that vnd is required for the specification of neural lineages within the developing tritocerebral neuromere, whereas the Hox gene lab appears to be independently required for the specification of neuronal identity within the same territory during later stages (Hirth et al., 1998; Page,2000; Hirth et al.,2001; Sprecher et al.,2004). This indicates that the activity of the columnar gene vnd is integrated into pattern formation along the anteroposterior neuraxis by ensuring proper formation and development of tritocerebral neural lineages that subsequently become further specified by the activity of the Hox gene lab.

The Drosophila columnar gene vnd belongs to the highly conserved Nkx2 class of transcription factors that have been found in various animals, including mammals (Harvey,1996; Cornell and von Ohlen,2000). Notably, the vnd/Nkx2 family of genes is exceptionally well conserved, both in terms of expression and function. Thus,the vertebrate homologues of vnd are expressed in the neural plate,or tube, in topologically similar positions as is vnd in the Drosophila ventral neuroectoderm and in the absence of vnd/Nkx2 genes, ventral-most cells in the spinal cord and the Drosophila VNC are missing or transformed(Cornell and von Ohlen, 2000; Rallu et al., 2002). Moreover,this evolutionary conservation in expression and function of vnd/Nkx2genes appears to apply to some extent to brain development. A comparison of the anteroposterior order of vnd/Nkx2 gene expression in the early embryonic brains of Drosophila and mouse reveals remarkable similarities (Urbach and Technau,2003b; Urbach and Technau,2004).

In terms of function, genetic knockouts in mice have shown that Nkx2 genes appear to play a crucial role in patterning and neuronal specification during embryonic development of the telencephalon and hindbrain. Nkx2.1 mutant mice display numerous brain patterning defects: the entire pituitary is missing (Kimura et al., 1996); the number of cortical interneurons is halved; there is a complete absence of TrkA-expressing cells in the developing telencephalon; and the ventral-most aspect of the telencephalon (the medial ganglionic eminence) becomes trans-fated to that of the adjacent more dorsolateral ganglionic eminence (Sussel et al., 1999). Thus, comparable with the role of vndduring Drosophila brain development (this study)(Urbach et al., 2006), Nkx2.1 is involved in pattern formation and in cell fate determination during embryonic brain development in mice(Rallu et al., 2002).

In addition, recent studies have shown that Nkx2.2 is involved in neural lineage specification in the developing hindbrain. In particular, the sequential generation of visceral motoneurons and serotonergic neurons from a common pool of neural progenitors located in the ventral hindbrain crucially depend on the integrated activities of Nkx2.2- and Hox1/2-class homeodomain proteins (Pattyn et al.,2003a; Pattyn et al.,2003b). An important function of these proteins is to coordinate the spatial and temporal activation of the homeodomain protein Phox2b, which in turn acts as a binary switch in the selection of motor neuron or serotonergic neuronal fate (Pattyn et al.,2003a; Samad et al.,2004). De-repressive activity of Nkx2.2 at or in vicinity of Pbx/Hox-binding sites proximal to the Phox2b enhancer enhances transcriptional activation of Phox2b by Hox1 and Pbx factors(Samad et al., 2004). These data suggest that comparable with the integrated activity of vnd and lab in Drosophila brain neuromere specification, integrated activity of the Nkx2.2 and Hox1/2 proteins is involved in the specification of segmental neural lineages. Thus, integration of anteroposterior and dorsoventral patterning systems by homeodomain transcription factors of the Hox and vnd/Nkx2 genes might represent an ancestral feature of insect and mammalian brain development(Hirth et al., 2003).

We thank A. H. Brand, the Developmental Studies Hybridoma Bank, C. Q. Doe,P. Fichelson, R. Finkelstein, K. Matthews, D. M. Mellerick, J. B. Skeath, H. Vaessin and K. White for flies and antibodies. This work was supported by the Swiss NSF (to H.R.), the DFG (to R.U. and G.M.T.) and ELTEM-NEUREX (to F.M.R. and F.H.).