CNS midline cells in Drosophila induce the differentiation of lateral neural cells

The principal organisation of the CNS is similar in insects and vertebrates. In both cases, the nervous system can be subdivided into two lateral halves connected by a special set of cells found at the CNS midline. These midline cells are the first cells to be specified in Drosophila as well as in vertebrates; in the fly, they are called the midline or mesectodermal cells whereas, in vertebrates, they are known as floor plate cells (Crews et al., 1988; Thomas et al., 1988; Schoenwolf and Smith, 1990). The development of the floor plate cells depends on inductive signals originating from the mesoderm (van Straaten et al., 1988; Placzek et al., 1990). Inductive events might control the development of the CNS midline cells in Drosophila as well, since in mutants, which lack the mesoderm, no midline cells are formed (Kosman et al., 1991; Leptin, 1991).

The midline cells are not only special regarding the early specification in the CNS but also exert a number of important and evolutionary conserved functions. In the developing neural tube, the floor plate cells contribute to the dorsoventral patterning in the CNS by inducing the development of lateral neurons (Yamada et al., 1993; Hynes et al., 1995b; Roelink et al., 1995; Ericson et al., 1996). Later in development the floor plate cells attract commissural axonal growth cones and guide them across the midline to the contralateral side (Tessier-Lavigne and Goodman, 1996).

The Drosophila CNS midline comprises a small set of about 20 individually identifiable cells per abdominal neuromere. In each segment, four midline glial cells, two MP1 neurones, six ventral unpaired median (VUM) neurones, two unpaired median interneurons (UMI) as well as the progeny of the median neuroblast are found (Klambt et al., 1991; Bossing and Technau, 1994). The development of all CNS midline cells depends on the function of the master regulatory gene single-minded (sim). sim encodes a transcription factor of the bHLH family, which is expressed almost exclusively in the CNS midline cells (Crews et al., 1988; Thomas et al., 1988; Nambu et al., 1991; Lewis & Crews, 1994). In mutant sim embryos, all CNS midline cells degenerate and, conversely when sim is expressed ectopically, neuroectodermal cells are routed towards a midline cell fate (Nambu et al., 1991). In addition to acting as a transcriptional activator, sim can mediate transcriptional repression (Wharton et al., 1994; Xiao et al., 1996).

Similar to vertebrate floor plate cells, the Drosophila midline cells exert inductive functions during embryonic development. On the one hand, midline cells control pattern-ing of the ventral ectoderm possibly through the production of an activating ligand of the EGF receptor homologue (Kim and Crews, 1993; Golembo et al., 1996). On the other hand, CNS midline cells also induce the formation of specific mesodermal cells (Lüer et al., 1997). During the early development of the ventral nerve cord, however, midline cells have apparently no function, as neuroblasts form in the correct number and pattern independent of the presence of a CNS midline (Rao et al., 1991). Nevertheless, heterotopic cell transplantations of neuroectodermal cells in mutant single-minded (sim) embryos, which lack differentiated CNS midline cells (Sonnenfeld and Jacobs, 1994) suggest that CNS midline cells might be required for the correct positioning or the maintenance of correct positions of lateral cells along the dorsoven-tral axis (Udolph et al., 1995).

Here we report evidence suggesting that the CNS midline cells are required during later CNS development. In embryos lacking the CNS midline cells due to mutation or specific cell ablation about 15% fewer cells develop in the lateral CNS cortex and the axonal scaffold collapses at the ventral midline. This phenotype can be partially rescued by homotopic transplantation of wild-type midline cells or directed expression of single-minded. In embryos lacking all CNS midline cells, the number of CNS cells expressing specific markers like rR226 is reduced, whereas the number of these cells increases following homotopic expression of single-minded cDNA in mutant single-minded embryos. In addition, we ectopically expressed single-minded in a strip of cells perpendicular to the normal CNS midline, which subsequently results in the formation of a perpendicular strip of midline cells. Such ectopically located midline cells are able to induce additional rR226-positive cells in the lateral CNS. We thus conclude that the Drosophila CNS midline cells are required for the development of neurons in the lateral nerve cord.

The following Drosophila stocks were kindly supplied by the Bloomington Stock Centre: sim^H9; Notch^ts1; otd^H1; and pnt8^B74. pnt^3D2-026, a pointedP2-specific allele was isolated in a saturating EMS mutagenesis (T. Hummel and C. Klambt, unpublished). To drive expression in the CNS midline, we used a sim-GAL4 driver line (Scholz et al., 1997). A UAS-sim effector strain was generated following germline transformation of a construct containing a fulllength single-minded cDNA clone (Nambu et al., 1991) in pUAST (Brand and Perrimon, 1993). Insertions of this construct on the second and third chromosome were used. UAS-hh, UAS-secreted-spitz were kindly provided by M. Frasch and B.-Z. Shilo. The following activator lines were used: scabrous-GAL4, hairy-GAL4 andpatched-GAL4 (Hinz et al., 1994); paired-GAL4 and engrailed-GAL4 (Brand and Perrimon, 1993).

To label subsets of embryonic CNS cells, we used the following antibodies: anti-EVE (Patel et al., 1994); 22C10 (Fujita et al., 1982); BP102 (A. Bieber, N. Patel and C. S. Goodman, personal communication), anti-ENGRAiLED (Patel et al., 1989); anti-REPO (Halter et al., 1995); anti-PROSPERO (Spana and Doe, 1995); anti-DEADPAN (Bier et al., 1992). The following enhancer trap lines were used: apterous-tau-lacZ (apC2.1) and 3.883 (Callahan and Thomas, 1994); M84 and P101 (Klambt and Goodman, 1991); AA142 X55 (Klambt et al., 1991); rK77 and rR226 (C. Klambt and C. S. Goodman unpublished); rI50 (Klaes et al., 1994); 1912 (Doe, 1992); sim-lacZ (Nambu et al., 1991). To generate single-minded mutant phenocopies using the Notch^ts1 allele, we employed a heat-shock regime as described previously (Menne and Klambt, 1994).

Immunohistochemistry and X-Gal staining were performed as described (Menne and Klambt, 1994; Lüer et al., 1997). To determine the cell number per neuromere in early stage 16, embryos were fixed and embedded as described (Stollewerk et al., 1996). Serial 1 pm sections covering several segments were stained with toluidine blue and photographed. Only segments represented by complete series of sections were used for subsequent analysis. Individual nuclei were traced and counted.

Ablation and transplantation of midline cells were performed as described (Lüer et al., 1997).

In each abdominal hemi-neuromere about 30 neuroblasts generate specific cell lineages consisting of neuronal and/or glial progeny cells (Bossing et al., 1996; Schmidt et al., 1997). The CNS midline comprises about 20 cells in each abdominal neuromere (Klambt et al., 1991; Bossing and Technau, 1994). In single-minded mutant embryos most of the midline cells die (Sonnenfeld and Jacobs, 1994) but the CNS cortex of mutant single-minded embryos appears much thinner then anticipated from the loss of the only 20 midline cells (Fig. 1A,B and C,D). However, the expression of singleminded is not restricted to the CNS midline but is also found in the mesoderm (Lewis and Crews, 1994) and CNS midline cells die relatively late in single-minded mutant embryos (after germband extension, stage 12). We therefore analysed the phenotypes following mechanical ablation of the midline cells from 1 to 2 consecutive segments at stage 7, as well as following genetic ablation of the CNS midline anlage using the Notch^ts1 allele at stage 5. The mechanical and the genetic ablation of the CNS midline anlage lead to a similar thinning of the ventral cord cortex (Fig. 2B). This suggests that the thinning of the cortex indeed results from the loss of midline cells.

To determine whether the observed reduction in size of a single-minded mutant ventral cord might include a loss of cells in the lateral CNS, we counted cells in neuromeres of wild type (Fig. 1A), single-minded (Fig. 1C) and Notch^ts1 embryos lacking the CNS midline due to specific heat-shock treatment (Menne and Klambt, 1994). All embryos were fixed at stage 16 and segmental units were defined by the entry points of the segmental nerves into the CNS (arrow in Fig. 1A). We counted about 490 cells in an abdominal wild-type neuromere (cortex plus midline in three neuromeres of three embryos with 496, 492 and 489 cells). In mutant singleminded embryos, we counted an average number of 411 cells (four neuromeres of four embryos with 419, 417, 401 and 408 cells, each). Similar CNS cell numbers were counted in heat-shocked Notch^ts1 embryos (two neuromeres of two embryos with 412 and 403 cells each). These data show that, in addition to lacking about 20 midline cells per neuromere, mutant single-minded as well as heat-shocked Notch^tsI embryos also lack about 15% of the normal complement of lateral ventral cord cells.

Two questions emerge from the above analysis. First, do the midline cells influence neuroblast development, formation of ganglion mother cells (GMCs) or terminal neural differentiation? And second, what is the nature of the missing cells? In a previous study, it was shown that the segmental neuroblasts form properly in mutants that lack the CNS midline cells (Rao et al., 1991). Similarly, neuroblasts develop normally following ablation of the midline cells during gastrulation (Fig. 2A). The formation of GMCs, as assayed by expression of prospero, also appeared to be unaffected in single-minded mutant embryos (data not shown).

In order to determine which CNS cells are affected in embryos that lack the midline, we assayed the expression of a number of enhancer trap lines, which were selected based on their expression of 0-galactosidase in only a few cells per hemi-neuromere.

To analyse CNS glial cells, we studied the expression of 0 -galactosidase directed by the enhancer trap lines rI50, PI0I and M84 in mutant single-minded embryos, as well as the pattern of anti-PROSPERO staining in stage 16 single-minded mutant embryos. In each case, we observed a reduction in the number of cells expressing the different marker genes relative to wild type (Table 1). It should be noted, however, that we cannot determine whether cells that normally express the different lacZ markers are absent or whether the lacZ expression is reduced below the level of detection. Given the reduction in the overall number of CNS cells in single-minded mutant embryos, we prefer to attribute the observed reduction in the number of cells expressing the different markers to a reduction in the number of lacZ-expressing cells.

In addition to the glia markers, we used several marker lines that allow the labelling of subsets of neuronal cells. The 0 -galactosidase expression pattern directed by some of these marker lines is unchanged in a single-minded mutant background (e.g. 3.883 (Callahan and Thomas, 1994)). Similarly the expression of eve, highlighting the aCC/pCC neurons, appeared unaffected in the single-minded mutant embryos (Table 1).

However, lacZ expression associated with several other marker lines appeared to be dependent on the presence of the CNS midline cells (Table 1, Figs 3, 4). The enhancer trap line rK77 is associated with β-galactosidase expression in all midline cells as well as in a pair of neuronal cells in each hem-ineuromere. Single–minded mutant embryos lack rK77 expression in most segments; however, in some neuromeres, we observed β-galactosidase expression in two lateral CNS cells (Fig. 3A,D).

β-galactosidase expression directed by the rR226 enhancer trap insertion is found in only three cells per hemi-neuromere, which are located laterally to the longitudinal tracts at the position of the posterior commissure (Fig. 3B). In mutant single-minded embryos, the number of rR226-positive cells found in each hemi-neuromere ranges from 1 to 3 (Fig. 3E). A similar reduction in the number of rR226-positive cells can be seen in Notch^ts1 embryos that have been shifted to the restrictive temperature (31°C) for 30 minutes just before gastrulation and thus lack most of the CNS midline anlage (Fig. 3F; Menne and Klambt, 1994).

To test further the hypothesis that the midline cells are responsible for the observed reduction in the number of cells expressing the rR226 marker and the alteration in the axonal pattern, we attempted to rescue the single-minded mutant phenotype by expression of a single-minded transgene using the GAL4 technique, as well as by homotopic transplantation of wild-type midline cells.

Upon directed expression of single-minded in all midline cells using a sim-GAL4 driver line (Scholz et al., 1997), we observed a partial rescue of the mutant single-minded axon pattern phenotype (Fig. 4B,C). The number of lateral CNS cells expressing the rR226 marker or the marker apC2.1 (Fig. 4D-F) appears to be rescued as well. Interestingly, persistent overexpression of single-minded in all midline cells leads to additional rR226-positive cells (Fig. 6C) but does not alter the wild-type number of apC2.1 -positive cells. Similarly, transplantation of midline cells also led to a partial rescue of the mutant single-minded phenotype and commissures develop (Fig. 2C).

Loss of all midline cells resulted in a decrease in the total number of lateral CNS cells and in a reduction in the number of cells that express specific molecular markers. In order to determine whether midline neurons or midline glial cells are required for proper differentiation of lateral CNS cells, we analysed the expression of the rR226 marker in the CNS of two additional mutant strains.

To address the role of the midline neurones, we studied mutant orthodenticle embryos. orthodenticle encodes a homeodomain protein that is expressed specifically in the midline neurones. Loss of orthodenticle function results in the degeneration of many midline neurones (Finkelstein et al., 1990). Compared to wild type, the number of rR226-positive cells is reduced in orthodenticle mutant embryos. We counted the number of rR226-positive cells in 96 hemineuromeres. 27% showed one rR226-positive cell; 43% showed two rR226-positive cells and 30% showed three rR226-positive cells. We never observed hemineuromeres with more than three or with no rR226-positive cells (Fig. 3C).

In mutant pointed embryos, the presumptive midline glial cells fail to differentiate and eventually die. Within the embryonic CNS, pointed is expressed specifically in glial cells (Klambt, 1993). pointed encodes two transcripts, pointedPI, which is expressed in the lateral glial cells and pointedP2, which is expressed only in the midline glial cells. To study the effect of midline glial cells on the development of lateral CNS cells, we used the allele pointed^3D2–026, which affects only the pointedP2 function (T Hummel and C. Klambt, unpublished data). In embryos lacking all pointed function or the pointedP2 function only, the number of rR226-positive cells decreases (Fig. 3B,G). Of 144 hemineuromeres counted, 23% showed one rR226-positive cell, 49% showed two rR226-positive cells and 28% showed three rR226-positive cells (Fig. 3G). We never observed more than three or no rR226-positive cells in a hemineuromere.

These data suggest that both glial and neuronal midline cell lineages are required for the differentiation or maintenance of the normal complement of lateral CNS cells.

Cell or tissue transplantation experiments are useful to demonstrate inductive properties. Unfortunately, in Drosophila, such experiments are not possible. For still unknown reasons midline cells transplanted laterally in the neuroectoderm have the tendency to migrate back to the midline and do not remain at the point of implantation (Udolph et al., 1995). We therefore tried to induce midline cell fates in the lateral CNS by genetic means, using the GAL4 system. Ubiquitous expression of sim during embryogenesis causes all neuroectodermal cells to adopt a midline cell fate (Nambu et al., 1991). Since the midline comprises neuronal as well as glial cell types, we first asked whether overexpression of single-minded in the midline would interfere with correct midline cell specification. Embryos of the genotype sim–GAL4, UAS–sim express elevated levels of single–minded in all CNS midline cells. This increased single-minded expression, however, does not lead to any change in midline cell fate and the normal number of midline glial cells and midline neurones is formed (data not shown). We thus concluded that the level of single-minded is not crucial for the acquisition of the correct cell fate in the midline.

Next we assayed the effects of single–minded expression in the entire neuroectoderm. Here cells appear to be routed towards the different midline cell fates (glial or neuronal) strictly according to their position in the segment. Thus, upon single-minded expression in the neuroectoderm, midline glial cells form in the anterior part of the segment, whereas neuronal cells types form in the posterior part of the segment (Fig. 5A,D and B,E). Whereas the ectopic ‘lateral’ midline neurones remained where they were induced by the ectopic expression of single–minded, we noted that the ectopic midline glial cells migrate towards axonal tracts, which might be comparable to the normal migration of the midline glial cells towards commissural axons (Fig. 5C,F).

To test for inductive properties of the CNS midline cells, we generated embryos that expressed single–minded in cell strips perpendicular to the normal ventral midline in single–minded mutant embryos. To generate such a pattern, we used flies carrying a UAS-sim construct as well as flies carrying different pair rule or segment polarity GAL4 driver constructs. engrailed-GAL4 directs expression of UAS coupled transgenes in a strip of cells just posterior to rR226-positive cells in each segment (Fig. 6A). Upon single–minded e xpression, neuronal midline cells form (see Fig. 5). In addition, up to five rR226-positive cells can be detected in many hemineuromeres (Fig. 6B). We could not use a paired-GAL4 activator line, since the rR226-positive cells are within the paired expression domain (not shown).

The number of rR226-expressing cells is similarly increased when we express single-minded in all midline cells. This effect is also observed in a single-minded mutant background (Fig. 6C). In all cases, we observed the appearance of additional rR226-positive cells only in close association with the ‘normal’ equivalent of rR226-positive cells, which suggests that not all CNS cells are equally competent to respond to signalling cues emanating from the midline. The single-minded GAL4 activator line used in these experiments does not drive detectable expression in cortical CNS cells. The expression of the apC2.1 marker was not affected by expression of single–minded, indicating that different signals are required for the differentiation or maintenance of different cortical CNS cells.

To address whether EGF-receptor-mediated signalling might be involved in controlling rR226 cell number, we expressed secreted SPITZ in all CNS midline cells. This expression did not result in an increase in numbers of rR226-positive cells, suggesting that EGF-receptor signalling is not needed in controlling rR226 cell number. When we transplanted single midline cells that secrete SPITZ into the CNS midline of wild-type hosts, we also observed no increase in the number of rR226-positive cells. However, in the same embryos, the number of aCC/pCC neurons was increased (Fig. 7). It should be noted that the number of aCC/pCC neurons is not changed in mutant single-minded embryos. This might again indicate the existence of several independent signals that regulate the differentiation or maintenance of different lateral CNS cells.

In summary, these data suggest that all CNS midline cells, midline glia as well as midline neurons, are capable of influencing the development of lateral CNS cells. However, the nature of the underlying signals remains elusive.

Here we have investigated the function of the midline cells of the CNS during the formation of the lateral CNS. Embryos lacking functional midline cells show a reduction in the number of CNS cells and a reduction in the number of cells expressing a variety of different markers. In addition, cell transplantation experiments as well as expression of singleminded in different regions of the embryonic CNS support the notion that the CNS midline is required for the differentiation or the maintenance of a subset of lateral CNS cells.

Despite the obvious differences between the invertebrate ventral nerve cord and the vertebrate neural tube, the development of both systems appears to follow rather similar principles. The definition of the neurogenic region is controlled by homologous dorsoventral patterning genes leading to neural tube formation on the dorsal side or nerve cord formation on the ventral side, respectively (Arendt and Nübler-Jung, 1996; Bier, 1997). In Drosophila, as well as in vertebrates, the subsequent selection of individual neural precursor cells from an uncommitted field of cells depends on the interplay of the neurogenic genes Notch and Delta (Campos-Ortega, 1993; Chitnis et al., 1995; Chitnis and Kintner, 1995; Dornseifer et al., 1997). Similarly, the bHLH proteins encoded by proneural genes appear to be required for neural development in flies as well as in vertebrates (Chitnis and Kintner, 1996; Ma et al., 1996; Hinz, 1997).

In both systems, the cells located at the midline of the CNS, the floor plate cells or the midline cells, appear to be required for the subsequent differentiation of cells in the lateral CNS cortex. The development of motoneurons in the neural tube as well as dopaminergic neurons in the midbrain, is dependent on the floor plate (Yamada et al., 1993; Hynes et al., 1995a). Here we have shown that, in Drosophila, the differentiation of a number of lateral cortex cells also requires the presence of the CNS midline. Beside specifying CNS cell number, the CNS midline performs further conserved functions during the later development of the CNS, and commissural growth cones are guided towards the midline. Molecules mediating this guidance have been identified as the NETRINS. Not surprisingly, in both vertebrates and invertebrates, NETRINS are expressed in the CNS midline cells, where they perform homologous functions (Hamelin et al., 1993; Harris et al., 1996; Mitchell et al., 1996; Serafini et al., 1996; Tessier-Lavigne and Goodman, 1996).

One simple way to assay possible inductive effects of a given tissue is by tissue or cell transplantation (van Straaten et al., 1988; Placzek et al., 1990; 1991; Hatta et al., 1991; Yamada et al., 1991). This approach, however, is not feasible for midline cells in Drosophila since, upon transplantation into the lateral neuroectoderm, midline cells have the tendency to migrate to the normal CNS midline (Udolph et al., 1995). We thus took advantage of the fact that single–minded acts as a master regulatory gene of midline development (Nambu et al., 1991). The CNS midline comprises different lineages with different functions during the formation of the CNS (Klambt et al., 1991; Bossing and Technau, 1994). Following expression of single–minded in the entire neuroectoderm, midline glial cells and midline neurons are formed in specific segmental positions, indicating that the segment polarity genes are required for individual cell fate specification. Preliminary analyses of the different midline cell lineages in several segment polarity mutants confirmed this observation.

Interestingly, overexpression of single-minded in the CNS midline of wild-type embryos using the sim-GAL4 driver line leads to the appearance of additional rR226-positive cells. This could imply that single-minded participates in the control of the inductive signal postulated to emanate from the CNS midline.

The data presented in this paper suggest that the midline provides signals required to establish the normal complement of lateral CNS cells. However, the molecular nature of these signals remains elusive. So far we have analysed the effects of two candidate genes, spitz and hedgehog, both of which encode secreted proteins that are involved in other inductive signalling processes (Roelink et al., 1994; Marti et al., 1995; Burke and Basler, 1996; Freeman, 1996; Pringle et al., 1996; Lüer et al., 1997). spitz acts as an activating ligand for the Drosophila EGF receptor homologue (DER) controlling eye and ventral ectoderm development (Freeman, 1996; Schweitzer et al., 1995; Gabay et al., 1996; Golembo et al., 1996). hedgehog is a segment polarity gene involved in patterning the ectoderm as well as the imaginal discs (Burke and Basler, 1997). Whereas the expression of single-minded in all midline cells leads to the formation of additional rR226-positive cells in the lateral CNS, we could not detect an increase in numbers of rR226-positive cells following the expression of secreted spitz in all CNS midline cells. However, expression of secreted SPITZ resulted in the appearance of additional aCC/pCC neurons pointing to the existence of several independent signals that regulate the development of different lateral CNS cells. When we expressed HEDGEHOG in all midline cells, we did not observe any increase in the number of rR226-positive or in the number of eve-positive CNS cells (data not shown).

Cells that perform important organizing functions are generally specified earlier than the surrounding tissue, which eventually becomes competent to respond to the inductive signals. This has been described for the vertebrate floor plate within the neural tube (Schoenwolf and Smith, 1990) and the ectodermal ridge of the limb bud (Laufer et al., 1997; Rodriguez-Esteban et al., 1997). Similarly the Drosophila CNS midline cells are the first cells in the nervous system to be specified (Crews et al., 1988; Thomas et al., 1988; Menne and Klambt, 1994; Martin-Bermudo et al., 1995). An additional characteristic feature of cells with organising properties is their location in the center of symmetry of a developing organ. Gradients of diffusible signals can then be generated, which serve as positional information to further subdivide the organ. Like the floor plate cells of the vertebrate neural tube (Munsterberg and Lassar, 1995; Pourquie et al., 1996), the Drosophila CNS midline exerts a bi-directional inductive influence. on the one hand, the midline controls the patterning of ventral somatic muscles and induces the formation of specific mesodermal cells (Lüer et al., 1997). on the other hand, the midline controls the differentiation of the ventral ectoderm (Kim and Crews, 1993) as well as the development of the lateral CNS (this paper). From our experiments, we cannot decide whether the latter is due to diffusible signals emanating from the midline or whether signals are conveyed via direct cell-cell contact. Further genetic experiments will help to identify the relevant genes and the nature of the signals.

We thank J. A. Campos-Ortega for continuous support. J. A. Campos-ortega, P. Hardy and two reviewers for comments on the manuscript. C. Q. Doe, C. S. Goodman, G. M. Rubin, J. Thomas, A. Travers and H. Vaessin for generously providing antibodies and flies. This work was funded by grants of the German-Israeli Foundation and the DFG to C. K. and by the DFG and the Biotech Programme of the EC to G. M. T.