The role of the msh homeobox gene during Drosophila neurogenesis: implication for the dorsoventral specification of the neuroectoderm

Both in vertebrates and in invertebrates such as Drosophila, the central nervous system (CNS) arises from a two-dimensional structure, a homogeneous sheet of neuroectodermal cells. An early and important step in the generation of diverse sets of neurons and glia is the specification of neural precursor cells according to their position within the neuroectoderm along the anteroposterior (AP) and dorsoventral (DV) axes (reviewed by Lumsden and Krumlauf, 1996; Tanabe and Jessell, 1996; Doe and Skeath, 1996).

The Drosophila ventral nerve cord is an excellent model system for understanding how cellular diversity is generated in the nervous system (reviewed by Goodman and Doe, 1993; Doe and Skeath, 1996). Its development begins with the delamination of about 30 neural and glial precursors, called neuroblasts (NBs), from the ventral neuroectoderm per hemisegment. Each of these NBs has a unique identity that is defined by its position, the timing of delamination, the characteristics of the neurons and/or glia it generates, and the expression of specific molecular markers (Hartenstein and Campos-Ortega, 1984; Hartenstein et al., 1987; Doe, 1992; Broadus et al., 1995; Bossing et al., 1996). NBs undergo several asymmetric cell divisions to bud off ganglion mother cells (GMCs) that divide once more to generate two neurons or glia. Consequently, the NBs give rise to approx. 300 neurons and approx. 30 glia in the mature embryonic CNS per hemisegment.

The acquisition of NB identity is the first and a crucial step in the generation of cellular diversity in the Drosophila CNS. This process is believed to be initiated before or at the time of NB delamination from the neuroectoderm. The stereotypy in the spatial pattern of NB formation and previous cellular analysis in the grasshopper have suggested that a major factor in the determination of NB identity is the position at which a NB forms (reviewed by Goodman and Doe, 1993; Doe and Skeath, 1996). Recent studies showed indeed that several segment polarity genes, including gooseberry-distal (gsb-d), wingless, hedgehog and patched, are regionally expressed in the neuroectoderm and specify the NB identities along the AP axis (Chu-LaGraff and Doe, 1993; Skeath et al., 1995; Bhat, 1996; Matsuzaki and Saigo, 1996). In contrast, the molecular mechanisms that specify NBs along the DV axis have been poorly understood.

The msh/Msx genes are one of the most highly conserved families of homeobox genes (reviewed by Davidson, 1995). During vertebrate spinal cord development, Msx genes (Msx1-3) are regionally expressed in the dorsal portion of the developing neuroectoderm (e.g., Liem et al., 1995; Shimeld et al., 1996; Wang et al., 1996). Similarly in Drosophila, msh is initially expressed in two longitudinal bands that correspond to the dorsal half of the neuroectoderm, and subsequently in many dorsal NBs and their progeny (Lord et al., 1995; D’Alessio and Frasch, 1996; this study). These observations led to the suggestion that msh/Msx genes may play phylogenetically conserved roles in the DV patterning of the CNS (e.g., D’Alessio and Frasch, 1996; Wang et al., 1996). However, the function of this gene family during CNS development has remained largely unknown due to the lack of information on the mutant phenotype (see Discussion).

In this study, we report on the genetic analysis of msh in Drosophila neurogenesis. We show that msh loss-of-function mutations lead to cell fate alterations of a number of NBs formed in the dorsal aspects of the neuroectoderm, including a possible dorsal-to-ventral fate change. Conversely, ectopic expression of msh in the entire neuroectoderm severely disrupts the proper development of the midline and ventral NBs. These results strongly implicate this gene in the DV patterning of the neuroectoderm.

Genomic DNA flanking the P-element insertion in the enhancer trap line rH96 was recovered by plasmid rescue. A 1.5 kb recovered sequence was then used to isolate overlapping genomic clones from a λFIXII genomic library (Stratagene). A DNA fragment adjacent to the P-element insertion point detected a transcript in a Northern blot analysis and was subsequently used to isolate several cDNAs from a 13-to 17-hour embryonic cDNA library (A.N., unpublished). Additional cDNA clones were also cloned by using the isolated cDNA as a probe. cDNA clones were sequenced by the Taq polymerase dideoxy chain method and autosequencer (ABI). The precise location of the P-element insertion in rH96 was determined by sequencing the genomic clone and rescued DNA.

The enhancer trap line rH96, which contained a P[ry⁺, lacZ] insertion, was viable and expressed the normal level of msh transcripts. msh mutant alleles were generated by imprecise excision of the P-element in rH96. The transposon was mobilized by hybrid dysgenesis, and approx. 150 independent ry^- excision chromosomes were balanced. Four embryonic lethal lines, msh^Δ60, msh^Δ68, msh^Δ138 and msh^lacZ-Δ89, were isolated and further analyzed by Southern blotting of the genomic DNA.

RNA in situ hybridization of whole-mount embryos was performed as described by Lehmann and Tautz (1994). Digoxigenin-labeled RNA probes were generated from a 0.6-kb fragment of msh cDNA cloned in pBluescript SK(+).

The GAL4-UAS system (Brand and Perrimon, 1993) was used for the ectopic expression of msh. The EcoRI-XhoI fragment of a msh cDNA containing the entire open reading frame (ORF) plus 236 bp of 5′ and 124 bp of 3′ untranslated sequence was cloned into the EcoRI-XhoI sites of a pUAST vector to generate the UAS-msh reporter construct. This construct was introduced into y, w embryos by germline transformation according to a standard protocol, and five UAS-msh reporter lines were obtained. Two of them, UAS-msh-m25-m6 and UAS-msh-m25-m1, each containing one UAS-msh transgene on the second and third chromosome respectively, were found to express the highest level of ectopic msh expression when crossed to the GAL4 activator lines and were used in this study. Ectopic expression of msh in all NBs was achieved by crossing the UAS-msh reporter lines with the sca-GAL4 effector line (Klaes et al., 1994). The same CNS phenotype was observed when either of the two UAS-msh reporter lines was used.

Antibody staining and dissection of embryos were carried out as previously described (Nose et al., 1992; Patel, 1994). The following primary antibodies were used: rabbit sAb against β-galactosidase (β-gal, Capell), mouse mAb against β-gal (Promega), rabbit sAb against EVE (Frasch et al., 1987), mouse mAb BP102 and 22C10 (Patel, 1994), mouse mAb against EN (Patel, 1994), rat sAb against RK2/REPO (Campbell et al., 1994), rat sAb against GCM (Jones et al., 1995), and rabbit sAb against TOLL (Hashimoto et al., 1991).

An eagle-kinesin-lacZ transgene in the K42 transformant line (Higashijima et al., 1996) and the hkb-lacZ in the 5953 enhancer trap line (Doe, 1992; Chu-LaGraff et al., 1995), both on the third chromosome, were utilized to visualize specific NBs and their putative progenies in msh mutants or in sca-msh. msh^Δ68, eagle-kinesin-lacZ /TM3 and msh^Δ68, hkb-lacZ /TM3 strains were generated by meiotic recombination between the transgenes and the msh allele. sca-GAL4 /+; hkb-lacZ /UAS-msh-m25-m1 embryos were obtained by crossing sca-GAL4 /+; hkb-lacZ /+ and UAS-msh-m25-m1 homozygous flies.

The enhancer-trap line rH96 was isolated and characterized for its expression in specific subsets of muscles and neurons (Klämbt et al., 1991; Nose et al., 1992). By standard methods (see Materials and Methods), approx. 10 kb of genomic DNA, spanning the P-element insertion site, was isolated (Fig. 1A). In this region, a single transcriptional unit of approximately 2.6 kb was detected by northern blot analysis of embryonic poly(A)⁺ RNA (data not shown), and the corresponding cDNA clones were subsequently isolated and sequenced. An open reading frame (ORF) that can encode a protein of 515 amino acid was found, a portion of which was identical to the partial sequence of the msh gene (Roberts et al., 1989). Thus, rH96 was a P-element insertion in the msh gene. While this work was in progress, two groups independently reported on the entire ORF of msh (Lord et al., 1995; D’Alessio and Frasch, 1996). The ORF deduced from our sequence data (submitted to GenBank accession no. AF009038) completely matches that reported by D’Alessio and Frasch (1996).

NB formation occurs between embryonic stage 8 to 11 in five phases (called S1-S5; Hartenstein and Campos-Ortega, 1984; Hartenstein et al., 1987; Doe, 1992). Early forming (S1-S3) NBs are mostly arranged in three longitudinal columns along the DV axis of the embryo. Later forming (S4 and S5) NBs are interspersed between the existing NBs. Along the AP axis of a segment, NBs are numbered from anterior to posterior as rows 1-7 (see Fig. 3). Recent in vivo tracing of DiI-labelled NBs showed that the positions of the NBs in the NB layer correlate with their site of origin in the neuroectoderm (Bossing et al., 1996). Accordingly, ventrally (medially) positioned NBs are derived from the ventral half of the neuroectoderm (thus referred to as ventral NBs); and dorsally (laterally) positioned NBs, from the dorsal half of the neuroectoderm (referred to as dorsal NBs). Due to the cell rearrangement within the neuroectoderm, the boundary between the ventral and dorsal NBs becomes a zigzag line at later stages (see Fig. 3F). NBs, with a few exceptions, behave like stem cells, undergoing asymmetric cell divisions to bud off ganglion mother cells (GMCs), which divide once to generate two neurons or glia. Some NBs (e.g., longitudinal glioblast [LGB]) exclusively give rise to glia, and thus are actually glioblasts. However, for simplicity, they are also referred to as NBs throughout this paper.

Expression of msh during neurogenesis was studied by in situ hybridization for MSH RNA and immunohistochemistry for β-galactosidase (β-gal) in rH96. As described below, β-gal was expressed fundamentally in the same pattern as MSH RNA in rH96 (thus referred to as msh-LacZ). The neuroectodermal expression of msh was first detected at stage 5 as discontinuous patches in several segments, which later extended and merged to form bilateral stripes that ran along the length of the embryo (Fig. 2A; Lord et al., 1995; D’Alessio and Frasch, 1996). The msh-expressing domain corresponds approximately to the dorsal half of the neuroectoderm (Fig. 2B,C; D’Alessio and Frasch, 1996). From this region, four S1 NBs of the lateral column delaminate. Strong MSH RNA expression was only detected in one (NB7-4) out of the four lateral NBs, although the other three NBs also appeared to express the transcript at a low level (Fig. 2B). In contrast, msh-LacZ was detected in all of the four lateral NBs (Fig. 2C). The initial msh expression in the dorsal neuroectoderm was transient and largely disappeared by late stage 9 (Fig. 2D).

Beginning at stage 10, msh expression was re-initiated in many dorsal S3-S5 NBs and their putative neuroectodermal proneural clusters (Fig. 2E-I, summarized in Fig. 3). These include the S3 NBs, NB6-4, LGB (longitudinal glioblast), and a putative glioblast (probably identical to the peripheral glioblast [PGB] described by Jones et al., 1995); S4 NBs, NB2-4, 4-3, and 5-4; and a S5 NB, NB3-4. MSH expression was also seen in some of their immediate progeny (e.g., GMC; Fig. 2H). It is notable that all the msh-positive NBs are those derived from the dorsal half of the neurogenic ectoderm (Bossing et al., 1996), a region that had earlier expressed MSH (see Fig. 3F).In contrast, no MSH expression was seen in NBs derived from the ventral portion of the neuroectoderm. Some dorsal NBs (e.g., NBs 3-3, 3-5, 4-4, 5-5, 5-6), although they were derived from the msh-positive dorsal neuroectoderm, did not themselves express MSH.

During S3-S5 NB formation, msh-LacZ was expressed in the same NBs that expressed the RNA (Fig. 2F,I). Whereas MSH RNA expression in NBs was mostly transient and highest when they were initially formed, msh-LacZ persisted much longer and was expressed in the putative progeny of the msh-positive NBs. At stage 16, msh-LacZ was detected in many neurons and glia mostly located in the lateral portion of the nerve cord (see Fig. 4B). Some LacZ-positive cells were seen to be born in the lateral neuroectoderm and later migrate either medially towards the midline or laterally towards the periphery. From their characteristic migration pattern, morphology and their final locations, these cells were identified as glial cells, including six longitudinal glia, (LG1-6, derived from the LGB, see Fig. 4A,B), two cell body glia (MM-CBG and M-CBG, derived from NB6-4, see Fig. 2I), some channel glia (CGs, derived from NB7-4; G.M. Technau, personal communication), and some exit and peripheral glia (EGs and, PG2 and 3, probably derived from the PGB; nomenclature of glia according to Klämbt and Goodman, 1991; Ito et al., 1995).

The msh gene had previously been mapped to the 99B region of the left arm of chromosome 3 (Roberts et al., 1989; Lord et al., 1995; D’Alessio and Frasch, 1996). No mutants that lacked msh function had been previously reported. The original P-element insertion in rH96, which was located at a position 171 bp upstream of the presumed transcription start site of msh (Fig. 1A), did not abolish the function of msh. We therefore generated four msh alleles, msh^Δ60, msh^Δ68, msh^Δ138 and msh^lacZ-Δ89, each of which contained a deletion in the msh gene by imprecise excision of the P-element (Fig. 1B, see Materials and Methods for details). All four alleles were embryonic lethal and failed to complement each other for lethality. Three of them, msh^Δ60, msh^Δ68 and msh^Δ138, did not express detectable levels of MSH RNA, whereas msh^lacZ-Δ89 expressed the transcript. msh^Δ68 contained a deletion of approx. 4 kb that removed most of the first exon in which the translational initiation site and part of the ORF was localized. Thus this allele probably represents a null allele. msh^Δ68 and msh^lacZ-Δ89, exhibited indistinguishable CNS phenotypes (described below) either as homozygotes or as transheterozygotes. We thus assume that msh^lacZ-Δ89 is also a null mutation and that the transcribed mRNA in this allele lacks msh function. In msh^lacZ-Δ89, the lacZ gene in the P element was left intact and was normally expressed in the msh^lacZ-Δ89 /+ embryos. Thus the LacZ expression could be utilized to trace the cells that normally express msh in the mutant. We mainly used these two alleles, msh^Δ68 and msh^lacZ-Δ89, to analyze the mutant phenotype. The other two alleles, msh^Δ60 and msh^Δ138, displayed similar but slightly milder phenotypes than msh^Δ68 and msh^lacZ-Δ89 (data not shown).

Since msh is expressed in the neuroectoderm before and during NB delamination, we first examined if NB formation took place normally in the msh mutants. Analysis of msh^lacZ-Δ89 homozygous embryos showed that all the msh-LacZ-positive NBs formed normally (data not shown). Thus, msh is not required for the formation of NBs. We then examined the fate of two msh-expressing NBs, LGB and NB6-4. These NBs were chosen because all (for LGB) or part (for NB6-4) of their lineage had been previously characterized by specific markers (Jacobs et al., 1989; Ito et al., 1995; Higashijima et al., 1996). Furthermore, since many of their progeny normally migrate towards the midline away from other msh-LacZ-positive cells, their fate could be unequivocally followed by anti-β-gal staining in normal (rH96) and msh mutant (msh^lacZ-Δ89) embryos.

The LGB is born in the lateral-most portion of the neuro-genic ectoderm, where its first division is symmetrical. Then, the two progeny migrate medio-anteriorly, and interiorly towards the inner surface of the neuroepithelium, and divide further to generate six longitudinal glia, which later align along the longitudinal axon tracts. In normal (rH96) embryos, msh-LacZ was detected in the LGB and its progeny (Figs 2F, 4A,B). The development of the LGB lineage in msh embryos was analyzed by staining for msh-LacZ (in msh^lacZ-Δ89 homozygous embryos), or for REPO, a glial-specific homeobox-containing protein (Xiong et al., 1994; Campbell et al., 1994; Halter et al., 1995). In msh embryos, LGB formed and conducted its first cell division, to produce two progeny, normally. However, their further cell division and migration were found to be abnormal. During early-to-mid stage 12, four LGB progeny that expressed msh-LacZ and REPO were seen in the medial region of the interior surface of the neuroepithelium in normal embryos (Fig. 4A,C). In msh embryos, 92% of the hemisegments (n=74) showed abnormality in the number and/or the position of the LGB progeny (Fig. 4D,F). Only two LGB progeny, which were larger than the normal LGB progeny at this stage, were present in many hemisegments, suggesting that they failed to conduct their second cell division to produce four cells. Many of them also failed to migrate properly and were found in the lateral and outer neuroepithelium close to the position where LGB had initially formed. Later, at stage 16, the msh-LacZ- or REPO-positive longitudinal glia were missing along the longitudinal axon tract (Fig. 4E).

The development of NB6-4 was studied by staining for msh-LacZ, REPO, and eagle-Kinesin-LacZ. The eagle-kinesin-lacZ fusion gene serves as a marker to visualize the morphology of eagle-positive NBs and their derivatives (Higashijima et al., 1996). NB6-4 is born in the lateral edge of the neuroectoderm, migrates medially, and divides in a characteristic quasi-symmetrical manner to generate two glial cells, MM-CBG and M-CBG (Fig. 5A). MM-CBG migrates further towards the midline and surrounds the VUM neurons by stage 13 (Fig. 5B). In msh embryos, NB6-4 formed normally and expressed specific markers such as REPO and eagle-Kinesin-LacZ. However, the timing and pattern of its subsequent cell division to generate MM-CBG and M-CBG appeared abnormal. Furthermore their medial migration often failed to occur or was retarded (Fig. 5E). Their morphology, as visualized by eagle- Kinesin-LacZ, was also dramatically altered (Fig. 5F). The abnormality in the migration and/or the morphology of MM-CBG was observed in 71% of the hemisegments (n=59).

It should be noted that although patterns of cell division and migration were severely affected, both LGB and NB6-4 nonetheless expressed their specific markers (REPO for both LGB and NB6-4, and eagle-Kinesin-LacZ for NB6-4). GCM, a transcription factor that controls glia versus neuronal fate (Jones et al., 1995; Hosoya et al., 1995), was also normally expressed in the LGB and NB6-4 progeny in msh mutants (data not shown). These results suggest that msh is required for some aspects of LGB and NB6-4 development (e.g. cell division and migration) but not for others (e.g., specification as a glial cell).

Although lineage analysis was possible for only LGB and NB6-4, due to the lack of specific lineage markers for other NBs, we obtained evidence suggesting that the development of other dorsal NBs is also abnormal. For example, we observed abnormalities in the division and/or migration of several other glia, including some channel glia (progeny of NB7-4) and peripheral glia (putative progeny of PGB; Jones et al., 1995; data not shown). Abnormality was also seen in the eagle- Kinesin-LacZ-positive PM and PQ neurons (putative progeny of NB2-4 and/or 3-3; Higashijima et al., 1996; Fig. 5C,G). Presumably due to the improper migration and differentiation of many neurons and glia, the axon tracts of msh embryos were severely disrupted (Fig. 5D,H). The longitudinal connectives were often reduced in width or completely missing between segments. The commissures were also distorted and somewhat fuzzy. However, development of ventral NBs appeared largely normal in the msh mutant. EVE-positive GMC1-1a and aCC and pCC neurons (progeny of NB1-1), GMC4-2a and RP2 neuron (progeny of NB4-2), and CQ neurons (progeny of NB7-1) formed normally (see below). hkb-LacZ-positive medial cell clusters (putative progeny of NBs 1-1, 2-1, and 2-2; ChuLaGraff et al., 1995) and identified neurons RP1 and RP3 (progeny of NB3-1) also formed normally (NB lineage according to Bossing et al., 1996).

A different type of msh mutant phenotype was revealed by staining for EVE. EVE is normally expressed in a small subset of neurons including aCC, pCC, RP2, CQ, fpCC and EL neurons. In msh mutants, an additional EVE-positive RP2-like cell was often detected adjacent to the normal RP2 neuron (56%, n=148; Fig. 6A,B). At a lower frequency, approx. 2 additional EVE-positive cells were also seen adjacent to aCC/pCC neurons (see below). Except for the presence of these additional EVE-positive cells, no abnormality was seen in the EVE expression pattern. The RP2 neuron is normally produced by a ventral NB, 4-2 (Doe, 1992). Since neither NB4-2 nor its progeny express msh, it is unlikely that the formation of the additional RP2 neurons was caused by some cell fate change within the NB4-2 lineage. An interesting possibility thus is that in the msh mutants, the fate of a msh-positive dorsal NB was transformed to a more ventral NB4-2-like fate, so as to produce the additional RP2-like cells. To test this hypothesis, we determined whether the ectopic RP2 arose from a msh-positive NB by double staining msh^lacZ-Δ89 embryos for msh-LacZ and EVE.

During normal development of NB4-2, EVE is initially expressed in its first GMC, GMC4-2a, and then in its two progeny, a larger RP2 and smaller RP2sib (Patel et al., 1989; Doe, 1992). In msh^lacZ-Δ89 embryos, at a stage when NB4-2 had already produced the normal EVE-positive RP2 and RP2sib, an additional EVE-positive GMC4-2a-like cell emerged in a position immediately internal to NB4-3 (Fig. 6C). This ectopic ‘GMC4-2a’ expressed msh-LacZ suggesting that it was derived from a msh-positive NB, most likely NB4-3. The later onset of EVE expression in the ‘GMC4-2’ compared with that in the normal GMC4-2 is also consistent with its having been derived from the later forming NB4-3. The ectopic ‘GMC4-2a’ produced two EVE-positive progeny, one larger and another smaller, just like the normal RP2 and RP2sib (Fig. 6D). The ‘RP2’ then migrated medially, changing their nuclear shape, in a characteristic manner similar to that of the normal RP2, and took a position adjacent to the normal RP2 by stage 13 (Fig. 6E). These results strongly suggest that in the absence of msh, the fate of NB4-3 was at least partially transformed to a NB42-like fate, so as to produce a NB4-2 lineage. Since the normal lineage of NB4-3 is unknown and no markers exist that can distinguish between NB4-2 and 4-3, it remains to be determined if the entire lineage of NB4-3 was converted to NB4-2 or the transformation occurred at the level of GMC. The analysis of the origin of the other ectopic EVE-positive cells found near aCC/pCC showed that they were also derived from a msh-LacZ positive NB(s) in the lateral CNS, suggesting that a similar dorsal-to-ventral cell fate transformation produced a duplicate ventral NB lineage (possibly NB1-1, the progenitor cell for aCC/pCC neurons) at the expense of dorsal NB(s) in msh mutants.

To further analyze the function of msh during CNS development, we then studied the effect of ectopic expression of msh by using the GAL4-UAS system (Brand and Perrimon, 1993). We generated transgenic lines carrying UAS-msh transgenes and crossed them to the scabrous(sca)-GAL4 line that drives pan-neuronal expression (Klaes et al., 1994, see Materials and Methods for details). In the resultant sca-msh embryos, msh was expressed in the entire neuroectoderm and subsequently in all NBs.

Ectopic expression of msh resulted in a severe disruption of the axon tracts; the commissures were almost completely absent, and the longitudinal connectives were broken and interrupted (Fig. 7A). Furthermore, gaps were often seen along the midline, suggesting the malformation of the medial CNS. To analyze the effect of ectopic msh in more detail, we stained the sca-msh embryos for various specific markers. The results showed that development of the midline and ventral NBs was seriously perturbed. For example, TOLL-positive midline cells including the midline glia (Hashimoto et al., 1991; Nose et al., 1992) were often missing or showed abnormal morphology, although initial expression of TOLL in the midline precursor cells occurred normally (data not shown). Defects in the midline were also shown by the absence of En-positive median NB progeny (Patel et al., 1989, Doe, 1992; Fig. 7B,C). Many identified neurons located in the medial CNS also failed to differentiate in sca-msh embryos, including the EVE-positive aCC, pCC, CQ, fpCC, and RP2 neurons, all known to be derived from ventral NBs (Patel et al., 1989, Doe, 1992, Bossing et al., 1996; Fig. 7F-I), and hkb-LacZ-positive medial cell clusters (putative progeny of ventral NBs 1-1, 2-1, and 2-2; Doe, 1992; Chu LaGraff et al., 1995; Fig. 7D,E). EVE expression in GMCs 1-1a, 4-2a and 7-1a (Doe, 1992; Broadus et al., 1995) was also nearly absent, suggesting that ectopic msh affects early aspects of NB development (data not shown). The initial formation of ventral NBs, however, occurred normally (as revealed by SNAIL expression, data not shown). In contrast to the severe disruption of the midline and ventral NB lineage, development of neurons and glia derived from the dorsal NBs appeared to be largely normal. These include the EN-positive lateral neurons (Patel et al., 1989; Fig. 7C), EVE-positive EL neurons (Patel et al., 1989; Fig. 7I), and REPO-positive longitudinal glia (Fig. 7J,K).

A prominent feature of msh expression is its restriction to the dorsal region of the developing CNS. It is expressed prior to NB delamination in longitudinal bands that correspond roughly to the dorsal half of the neuroectoderm. The initial msh expression in the dorsal neuroectoderm is transient and largely disappears by the time S2 NBs form, with the exception of the persistent expression in NB7-4. During later development, msh is re-expressed in a subset of S3-S5 NBs. Again, the expression is confined to those that delaminate from the dorsal neuroectoderm that had earlier expressed msh (see Fig. 3F).

The analysis of the msh loss-of-function mutants showed that msh is required for the proper development of dorsal NBs. During the lineages of LGB and NB6-4, the NBs form but fail to conduct their correct developmental program such as cell division and migration. Nonetheless, they express cell-type specific markers such as REPO, GCM, and eagle-Kinesin-LacZ, suggesting that part of their specification was executed normally. The defects thus appear not to reflect a simple cell fate transformation (e.g., to a more medial NB identity), but rather an inability to realize specific aspects of their normal fate. However, a phenotype suggestive of a dorsal-to-ventral fate change was observed for at least one dorsal NB. A msh-positive NB (most likely NB4-3) gives rise to EVE-positive progeny normally produced by a ventral NB, 4-2. These results suggest that msh can act as a switch between alternate fates along the DV axis for some NBs but not others.

In addition to the above examples, defects were seen in the differentiation of a number of neurons and glia formed in the dorsal aspect of the CNS, suggesting that msh function is required for many dorsal NB lineages. Since msh is expressed both in the neuroectoderm and in the NBs, it is currently unknown if the msh function is required in the neuroectoderm, in the NBs, or in both. Initial longitudinal msh expression may confer positional identities (e.g. dorsal positional identities) to the neuroectodermal cells, and this may in turn influence the fate of NBs that will later delaminate from them. msh may also function autonomously in the NBs to specify their fate. It should also be noted that although the phenotypes seen in early aspects of NB development most likely reflect the intrinsic requirement of msh in the NBs and/or in the neuroectoderm, some of the defects seen during later development may also be partially due to the secondary effects caused by altered environment in the lateral neuroectoderm. For example, changed neighbor-relationship between GMCs and neurons due to the abnormalities in cell proliferation and migration may in turn induce changes in the fate of cells that will form later.

Another candidate gene for the DV specification of the neuroectoderm, ventral nerve cord defective (vnd), encodes an NK2-like homeodomain protein (Jiménez et al., 1995; Mellerick and Nirenberg, 1995). While msh is expressed in the dorsal neuroectoderm, vnd is transcribed in longitudinal bands that correspond to the ventral half of the neuroectoderm. Thus, expression of these two homeobox genes subdivides the neuroectoderm into two distinct domains along the DV axis. In vnd mutants, the expression of proneural genes in the medial neuroectoderm fails to occur; and, consequently, the formation of some medial NBs is disrupted (Skeath et al., 1994). The cell fate change of NBs in vnd mutants has not been studied. In contrast, in the msh mutant, NBs form normally but their subsequent development to generate specific lineage is affected. These observations suggest that these two genes may function in somewhat different manners during CNS development, although both are regionally expressed in the neuroectoderm.

Ectopic expression of msh in the entire neuroectoderm severely disrupted the differentiation of cells derived from the ventral neuroectoderm, indicating that restricted expression of msh is critical for proper CNS development. Although midline cells and ventral NBs initially formed, their subsequent development to generate specific GMCs and neurons was perturbed. The lack of EVE expression in GMCs1-1a, 4-2a and 7-1a, the first progeny of their parental NBs, indicates that early aspects of NB specification were affected. No obvious ventral-to-dorsal transformation was seen, however, suggesting that msh expression is not sufficient to convert ventral NBs to a dorsal fate. We observed that ventral markers such as vnd were still expressed largely normally in sca-msh embryos (T. I. and A. N., unpublished observations). Thus the defects seen in the ventral NBs appear to be caused by the concomitant execution of ventral and dorsal developmental programs. (Note, however, that we obtained preliminary results that the fate of some ventral NBs was partially transformed to a more dorsal one.) Due to the improper specification of ventral NBs (and the midline cells), and presumably also due to the changed environment that follows, most cells in the ventral neuroectoderm failed to differentiate or died, leading to a drastic disruption of the ventral CNS.

D’Alessio and Frasch (1996) recently reported that the msh expression domain in the neuroectoderm is expanded ventrally in the flb mutant. flb encodes a Drosophila EGF-receptor homolog (DER), and is required for the establishment of ventral fate in the neuroectoderm (Raz and Shilo, 1993; Golembo et al., 1996). This result suggests that msh is normally repressed in the medial neuroectoderm by the ventralizing signaling mediated by DER. Since ectopic expression of msh inhibits the proper differentiation of the ventral neuroectoderm, suppression of msh in this region during normal development must be an important component of DER function. Similarly, during vertebrate CNS development, the dorsally restricted expression of Msx1-3, and genes belonging to another family of homeobox genes, Pax3/7, is established in part by repressive action of a ventralizing signal, Sonic Hedgehog (SHH; reviewed by Tanabe and Jessell, 1996). Again, inhibitory events appear to play crucial roles in regulating vertebrate CNS patterning. Medial neural plate cells that are never exposed to SHH maintain PAX7 expression and lose their capacity to realize ventral fates (e.g., the floor plate and motor neurons; Ericson et al., 1996). Furthermore, ectopic expression of Pax3 inhibits floor plate differentiation (Tremblay et al., 1996). Thus, suppression of dorsally restricted transcription factors by ventralizing signals appears to be an essential element in DV patterning of the CNS both in Drosophila and in vertebrates.

Dorsally restricted expression of msh, its regulation by flb and other DV patterning genes (D’Alessio and Frasch, 1996), and the loss-of-function and gain-of-function mutant phenotypes described in this study, are all consistent with the notion that msh plays a role in specifying regional identity. msh may act as a transcriptional activator to confer dorsal positional identities to NBs. Alternatively, msh may function in a repressive manner to suppress ventral cell fate in the dorsal neuroectoderm. A role in transcriptional repression has been shown for murine Msx-1 (Zhang et al., 1996).

Since msh is expressed in the neuroectodermal region that corresponds to two to three longitudinal columns of dorsal NBs (see Fig. 3F), msh expression alone would not differentiate the DV identity of NBs within this domain. The lack of apparent dorsal-to-ventral cell fate transformation in LGB and NB6-4 in the loss-of-function mutants is also consistent with the idea that msh is only part of the mechanisms that generate the final NB specificity. We suggest that msh functions as a component of the genetic system that regulates NB identity along the DV axis.

The msh/Msx genes are an ancient family, whose members have been found in a variety of animal species (reviewed by Davidson, 1995). The high degree of sequence conservation in and around the homeobox, and the similar spatial expression pattern in vertebrate and Drosophila CNS suggest an interesting possibility that this gene family performs a conserved function in the DV patterning of the CNS. Two Msx mutations reported so far, a knock-out of the mouse Msx1 and a dominant mutation in the human MSX2, both display craniofacial abnormalities, with defects in bones and teeth (Jabs et al., 1993; Satokata et al., 1994). However, no gross abnormalities have been found in the neural tube, presumably due to redundancy among the Msx genes. The genetic analyses of msh presented in this study provide the first in vivo evidence for the essential role of this gene family in neural development. Further studies on the function of msh in Drosophila should greatly help in elucidate the roles played by this gene family in CNS patterning.

We would like to thank C.S. Goodman for his continuous encouragement. We also thank C. Doe, S. Higashijima, S. Hayashi and colleagues in our lab for helpful discussions; C.S. Goodman, M. Frasch, S. Hayashi, C. Doe, B. Jones, S. Higashijima, T. Hosoya, T. Kojima, K. Saigo, H. Okano, K.V. Anderson, and the Bloomington stock center for providing antibodies and fly strains; and C. Doe and G.M. Technau for comments on an earlier version of the manuscript. T.I. is a postdoctoral research fellow of the Japan Society for the Promotion of Science. This work was supported by research grants to A.N. and M.T. from the Ministry of Education, Science, and Culture of Japan.