The dead ringer/retained transcriptional regulatory gene is required for positioning of the longitudinal glia in the Drosophilaembryonic CNS

In animals, the initial development of the central nervous system (CNS)occurs during embryogenesis when neurons send out their processes, axons, to establish a stereotypical arrangement of axon bundles. Glial cells constitute the second major component of the nervous systems of all multicellular organisms. The function of the nervous system as a whole depends on the correct specification and function of both neurons and glia. Glia are essential for controlling proliferation rates of vicinal neuroblasts(Ebens et al., 1993) and for the correct establishment and survival of the axon scaffold(Booth et al., 2000;Hidalgo and Booth, 2000;Hidalgo and Brand, 1997;Hidalgo et al., 1995;Xiong and Montell, 1995). Furthermore, glia ensheath axon bundles, providing electro-insulation and spatial separation from neighbouring haemocytes, thereby creating the blood-brain barrier (Granderath et al.,1999b).

The two characterized classes of glial cells studied so far in the embryonic CNS of Drosophila are the midline glia (proposed to correspond to the floor plate of vertebrates) and longitudinal glia(reminiscent of vertebrate oligodendrocytes). These cells are distinguished on the basis of their different positions and patterns of gene expression. Midline glia, which originate from the mesectoderm, depend on expression of the single-minded (sim) gene and activation of the EGF signalling pathway for their development(Hummel et al., 1999;Klambt et al., 1991;Stemerdink and Jacobs, 1997). The ultimate phenotype of embryos mutant for genes essential for midline development is the absence of axon commissures and collapse of the longitudinal axon connectives.

Longitudinal glia are a heterogenous population of cells, the specification of which depends on the activity of the gene glial cells missing/glial cell deficient (gcm/glide). gcm encodes a transcription factor which binds to a conserved eight nucleotide DNA-binding motif called the GCM motif (Akiyama et al.,1996; Bernardoni et al.,1998; Hosoya et al.,1995; Jones et al.,1995; Schreiber et al.,1997; Vincent et al.,1996). GCM is a master regulator of glial development, acting as a switch between neural and glial cell fates. GCM performs its function through activation of the gene tramtrack (ttk), which encodes a zinc-finger protein that acts as a repressor of neural cell fate(Badenhorst, 2001;Giesen et al., 1997;Van De Bor et al., 2000). Concomitantly with repression of the neural cell fate, GCM triggers expression of proteins that mediate glial cell differentiation. These include a GCM homolog GLIDE2, the ETS domain protein Pointed (PNT), the homeodomain protein Reversed polarity (REPO) and Locomotion defects (LOCO), a regulator of G-protein signalling (Campbell et al.,1994; Granderath et al.,1999b; Halter et al.,1995; Kammerer and Giangrande,2001; Klaes et al.,1994; Klambt,1993; Klambt and Goodman,1991; Xiong et al.,1994). Activation of the expression of another glial-specific homeodomain protein, Prospero (PROS), in a subset of LG (six out of ten LG per hemisegment), depends on the activity of the DROP/MSH/Lottchen protein (DR– FlyBase) (Buescher and Chia,1997; Doe et al.,1991). In gcm and glide2 mutants, all glial cells acquire a default neural cell fate(Akiyama et al., 1996;Hosoya et al., 1995;Jones et al., 1995;Kammerer and Giangrande,2001). Surprisingly, when PNT, REPO, LOCO or PROS functions are abrogated, longitudinal glia are still formed, albeit in a spatially disorganized fashion, indicating a failure to undergo terminal differentiation(Buescher and Chia, 1997).

Longitudinal glial cells are involved in the formation of the longitudinal axon fascicles by aiding navigation of the pioneer axon growth cones and by directing the fasciculation and defasciculation of axons(Hidalgo and Booth, 2000). Moreover, longitudinal glia are essential for follower axon survival duringDrosophila embryogenesis (Booth et al., 2000).

The loss of correctly specified glia in either gcm, glide2 orDrop/MSH/ltt mutant embryos accounts for the severe defects in the longitudinal connectives observed in these embryos(Akiyama et al., 1996;Buescher and Chia, 1997;Hosoya et al., 1995;Jones et al., 1995;Kammerer and Giangrande,2001). By contrast, mutations in any of the pnt, repo,loco and pros genes result in fasciculation defects that are much weaker than those observed in the gcm mutant embryos(Buescher and Chia, 1997;Granderath and Klambt, 1999a). Embryos lacking any of these gene products develop with minor defasciculation of the longitudinal connectives.

Despite some progress in understanding the specification of the longitudinal glia, many genes involved in this process remain unidentified and the role of some of the known genes is not fully elucidated(Egger et al., 2002). One candidate for a role in embryonic CNS development is the Drosophila dead ringer/retained gene (dri/retn, here referred to asdri) (retn – FlyBase), because of its expression in the embryonic CNS (Gregory et al.,1996). dri encodes a nuclear protein with a conserved DNA binding domain termed the ARID [AT-rich interaction domain(Gregory et al., 1996)(reviewed by Kortschak et al.,2000)]. Analysis of the early stages of Drosophilaembryogenesis showed that DRI is involved in aspects of dorsal/ventral and anterior/posterior axis formation acting either as a repressor or an activator depending on a particular developmental context(Valentine et al., 1998;Hader et al., 2000;Shandala et al., 1999). Later in embryogenesis, dri is required for myogenesis and hindgut development (Shandala et al.,1999).

We describe a role for dri in the formation of a functional CNS during Drosophila embryogenesis. We show that dri is expressed in a subset of glial cells as part of the cascade of transcriptional regulation that occurs during glial cell differentiation. dri mutant embryos are shown to exhibit mild defasciculation defects, similar to those caused by LG differentiation defects in pnt, repo, loco andpros mutant embryos. Finally, the dri mutant longitudinal glia are shown to be defective in their ability to migrate along the longitudinal axonal tracts to their proper positions, providing a cellular basis for the observed axon phenotype.

Two protein null alleles, dri¹ anddri², described previously(Shandala et al., 1999), were used during the course of this work. Enhancer trap and reporter lines used were: AA142, which labels midline glia; X55, which labels midline neurons(Klambt et al., 1991);gcm enhancer trap lines rA87(Jones et al., 1995;Vincent et al., 1996) andgcm^P (Hosoya et al.,1995); rc56, a loco enhancer trap line(Klaes et al., 1994); and P2333, a Star enhancer trap(Spradling et al., 1999). For targeted expression of dri in the embryonic CNS, the yeast Gal4-UAS system was used (Brand and Perrimon,1993). A previously reported UAS-dri construct(Shandala et al., 1999) was placed under the control of several Gal4-expressing constructs:KrGal4 (which expresses GAL4 in T2-A4), simGal4 (which expresses GAL4 in the embryonic midline, provided by C. Klambt, University of Muenster, Germany) and MZ1580 (a PGAL4 insertion in theactin locus, which directs GAL4 expression in the longitudinal glia and its progeny as well as in MP2 neurons and macrophages, provided by G. Technau, Institute of Genetics, University of Mainz, Germany). Thegcm^P1 amorphic allele was provided by B. W. Jones(Skirball Institute, NYU School of Medicine, NY). Unless specified, fly stocks were obtained from the Bloomington Drosophila Stock Center (Indiana University, Bloomington, IN).

A rat polyclonal antibody raised against a bacterially expressed pGEX-DRI fusion protein (Gregory et al.,1996) was used to detect the distribution of DRI in embryos. Polyclonal anti-β-gal was obtained from Rockland Immunochemicals(Gilbertsville, PA), anti-REPO was provided by A. Travers (Laboratory of Molecular Biology, Medical Research Council, Cambridge, UK), anti-Eve was provided by J. Reinitz (The University at Stony Brook, NY), the monoclonal antibodies 1D4 anti-FasII and 1B7 were provided by C. Goodman (University of California, Berkeley, CA), and anti-GFP was provided by P. Silver (Dana Faber Cancer Inst, Boston, MA) (Seedorf et al.,1999). The monoclonal antibodies 9F8A9 anti-ELAV, 2B10 anti-CUT,MR1A anti-PROS, 22C10, BP 102 and 4D9 anti-EN were all obtained from the Developmental Studies Hybridoma Bank (The University of Iowa, Iowa City, USA). For immunohistochemistry, anti-IgG secondary antibodies conjugated with AP,HRP, Cy5 and Lissamine-Rhodamine (Jackson ImmunoResearch Laboratories, PA),and Alexa488 (Molecular Probes, OR) were used. Biotinylated anti-IgG secondary antibodies and AP-, Lissamine-Rhodamine-conjugated streptavidin (Jackson ImmunoResearch Laboratories, PA) were used where signal enhancement was required. Immunohistochemical staining was carried out according to the method of Foe (Foe, 1989). Embryos were mounted in phosphate-buffered saline (PBS) containing 80% glycerol and viewed with epifluorescence using a Zeiss Axiophot or by confocal microscopy using a BioRad MRC1000 scanhead equipped with a krypton/argon laser confocal microscope.

Gregory et al. have reported expression of dri in the CNS ofDrosophila embryos (Gregory et al., 1996). To better characterize the pattern of expression, we co-stained embryos with the axon marker, monoclonal antibody 22C10 and anti-DRI antibodies. The staining revealed that some of thedri-expressing cells lie close to the midline(Fig. 1A, arrow). To further examine the nature of these cells, we used enhancer trap lines with previously described patterns of β-galactosidase expression in mesectoderm-derived cells: X55 (P element insertion at 56E, expresses in the VUM neurones, the MNB and its support cells, and the posterior pair of the midline glia,Fig. 1B), P 2333 [aStar enhancer trap insertion(Spradling et al., 1999)(Fig. 1C)] and AA142 (a P element insertion at 66D that expresses β-galactosidase in the midline glia, MGA, MGM and MGP; Fig. 1D). No correspondence was observed between these markers anddri-expressing cells. However, using the midline-derived cell markers as positional landmarks, we could infer that the dri-positive cells correspond to dorsally located glial cells.

To confirm this, we employed specific molecular markers: the enhancer trap lines rA87 (gcm, Fig. 2A,B), pnt^rm254(Fig. 2C) and rC56(loco, Fig. 2D), and the antibodies anti-REPO (Fig. 2E), anti-PROS (Fig. 2F) and anti-CUT (Fig. 2G). Co-expression of DRI with these glial markers confirmed that,with the exception of one cell per hemisegment at the very lateral edge of the ventral nerve cord (VNC, notched arrowheads inFig. 1A-D), thedri-positive cells are longitudinal glia.

Based on their position within the VNC, the dorsal glia can be subdivided into the longitudinal glial (LG), the A and B subperineural glia (A/B SPG),which are positioned at the medial edge of the longitudinal glia, and the medialmost cell body glia (MM-CBG), which are located in the very medial position within the cortex (Halter et al.,1995; Ito et al.,1995; Jacobs et al.,1989; Klambt and Goodman,1991; Udolph et al.,1993) (summarized in Fig. 2H). All LG, A/B SPG and MM-CBG express gcm, repo andcut (Fig. 2A,B,E,G).dri appears to be expressed in eight out of ten of the LG(Fig. 2, short arrows). It is known that the LG comprise eight cells that are the progeny of lateral glioblasts, plus approx. two cells per hemineuromere that are of unknown origin. The latter intermingle with other longitudinal glial cells and cannot be distinguished by their position or any molecular marker from stage 13 of embryogenesis (Halter et al.,1995; Ito et al.,1995; Jacobs et al.,1989; Klambt and Goodman,1991; Udolph et al.,1993). Our data would be explained if dri is expressed in the eight progeny of the lateral glioblasts, but not in the two cells of unknown origin, but this is yet to be confirmed.

The A/B SPG and MM CBG express dri(Fig. 2). pnt andloco are expressed in the A/B SPG, but not the MM-CBG(Fig. 2C, data forloco are not shown). pros colocalizes with dri in only five LG per hemineuromere, but is absent from the A/B SPG and the MM-CBG(Fig. 2F).

In addition to the dri-positive glia, each hemineuromere has one lateral cell located posterior to the intersegmental nerve root (as revealed by the mAb 22C10 axon marker; Fig. 1A, arrowhead). The origin of these cells is unknown, but they do not express any glial specific marker (Fig. 2A-F) and they project axons along the longitudinal tracts (K.-L. Harris and P. Whitington, personal communication), confirming their identity as interneurons. The nature of these cells and the role of dri in their development is not considered further in this report.

Expression of dri in the embryonic CNS is first clearly detected at stage 11 in a single cell per hemineuromere, probably corresponding to the lateral glioblast (data not shown). DRI continues to be expressed while the daughter glial cells divide and migrate anteromedially(Fig. 3A).

The early expression of dri in a subset of longitudinal glia suggested that dri might be required for the initial neurone/glia fate decision. To test this, we first examined the distribution ofrepo-positive cells in dri mutant embryos from stage 11 up until stage14. The correct position and number of these repo-positive cells (Fig. 3B) suggested that initiation of gliogenesis is not affected in the absence of DRI.

We then tested whether ectopic expression of dri could induce glial fates in non-glial cells, as has been shown for gcm andpnt (Bernardoni et al.,1998; Klaes et al.,1994). However, sim-GAL4-induced expression ofdri in midline cells was found not to increase the number of glial cells, as measured by the expression of repo(Fig. 3C, compare with 3D). Furthermore, when dri is driven by the MZ1580-GAL4 enhancer trap line that is specific for longitudinal glia and their progeny, as well as for MP2 neurones and macrophages, there were no ectopic pros-positive cells(data not shown). Ectopic expression of dri is therefore not sufficient to induce a glial cell fate.

Glial differentiation requires repression of the neuronal fate by the 69 kDa Tramtrack isoform (TTK69) (Badenhorst,2001; Giesen et al.,1997). Overexpression of TTK69 in neuronal cells blocks neural differentiation. To test whether dri is capable of repressing the neural fate, we drove expression of UAS-dri in the T2-A4 segments using Kr-GAL4. In contrast to the effect of expressing TTK69, ectopic expression of DRI did not prevent normal neural differentiation in the CNS or PNS, as assayed by expression of the neuronal marker, ELAV(Fig. 3E, between arrowheads). The normal differentiation of neural cells also suggests that glial cell fate is not induced by DRI in the Kr-specific domain. Loss of TTK69 results in general derepression of the 22C10 neural specific antigen in somatic muscles (Giesen et al.,1997). In dri mutants, however, this antigen was observed in its normal pattern, i.e. in a single somatic muscle fibre(Fig. 3F). These results show that dri does not induce the glial fate, nor does it repress the neural cell fate.

As the next step in exploring possible roles for dri in the LG, we investigated the organization of the longitudinal connectives. Axon projections along the VNC and the subsequent formation of longitudinal connectives depend on the proper function of longitudinal glia(Hidalgo and Booth, 2000;Hidalgo and Brand, 1997). During formation of the first longitudinal tract, LG navigate ipsilateral pioneer axons. Upon establishing contacts with axons, glia continue migrating along the longitudinal bundles, occupying choice points of axon fasciculation or defasciculation. The final fasciculation pattern can be visualized as three longitudinal Fasciclin 2 (Fas2)-positive (mAb1D4-staining) bundles on either side of the midline of stage 15-16 embryos(Grenningloh et al., 1991)(Fig. 4C).

In dri mutant embryos, the first processes projected by neurons were found to be normal (data not shown). This is expected, given thatdri is not required for the specification of the longitudinal glia. Furthermore, at stages 11-14, no gross abnormalities were observed in the initial formation of the longitudinal connectives(Fig. 4B, compare with 4A). However, in stage 15-16 embryos, moderate but consistent fasciculation defects were observed in dri mutant embryos. Some axons were not bundled correctly within the fascicles, which appeared thinner and were occasionally found at the dorsal surface of the VNC(Fig. 4C-E). Axons were observed to occasionally join the wrong bundle or cross inappropriately between bundles, although they never crossed the midline. Moreover, the spacing between the first and second, and between the second and third parallel bundles was slightly increased. The severity of this phenotype was much less than that observed in gcm mutant embryos(Hosoya et al., 1995;Jones et al., 1995;Vincent et al., 1996) (seeFig. 4G), but was similar to the level of disruption observed in pnt mutant embryos(Klaes et al., 1994;Scholz et al., 1993)(Fig. 4F).

This phenotype could arise as a consequence of neural or glial defects. With the aim of showing the glial-specificity of the dri phenotype,we attempted to rescue the axon defects by directing expression of UAS-dri in longitudinal glia using the MZ1580-GAL4 transgene(Hidalgo et al., 1995). Unfortunately, overexpression of dri interfered with normal glial development, causing misplacement of pros-positive cells and severe defasciculation of axons (data not shown). However, the possibility that the phenotype arose as a result of an indirect effect of DRI on neuronal development was ruled out by staining dri mutant embryos with anti-EVE, anti-EN, mAbBP102 and mAb22C10, and finding no defects in neural differentiation (data not shown).

The fasciculation defects showed that the dri mutant longitudinal glia were defective in their axon guidance role. To understand the nature of this defect further, we examined the behaviour of these cells in drimutant embryos. In wild-type stage16 embryos, the longitudinal glia form two flat sheets of cells either side of the midline underlying the dorsal surface of the axon tracts (Halter et al.,1995; Jacobs et al.,1989; Schmidt et al.,1997) (Fig. 5A,C,E). In dri mutant embryos, these cells were found to occupy a normal position relative to the midline, indicating that the LG had undergone their appropriate initial anterior-medial migration from their place of birth at the very lateral edge of the VNC(Fig. 5B,D). However, a pronounced defect in spatial organisation of the glia was observed. The mutant cells were ventrally shifted and not localized to the characteristic flat cell sheet on the dorsal side of the longitudinal connectives with only one or two out of ten glial cells per hemineuromere being in any one focal plane(Fig. 5F, compare with 5E).

Significantly, there is a high degree of similarity in the spatial disorganisation of the two rows of longitudinal glia in dri(Fig. 5F), loco andpnt mutant embryos (Granderath et al., 1999b). The pnt phenotype is associated with partial misregulation of loco expression, suggesting possible regulatory relationships between pnt, loco and dri, a possibility that is explored further below.

To characterize the cellular dri phenotype further, we used thegcmP1 enhancer trap, which expresses β-galactosidase, and monoclonal antibody 1B7, which recognizes both the neuronal and non-neuronal form of the Neuroglian protein. Comparison between the wild-type(Fig. 6A-C) anddri-mutant (Fig. 6D-F)phenotypes of stage 14 CNS preparations revealed that some glia lose expression of the cell adhesion protein, Neuroglian(Fig. 6E,F, arrows). Significantly, Neuroglian-deficient glial cells show somewhat round shapes compared with the more elongated wild-type glia(Fig. 6, compare B with E). Misregulation of the cell adhesion protein Neuroglian might impede cell-cell contacts leading to cell shape changes and failure of glia to migrate normally along the axon tracts.

As noted above, the similarity between the dri phenotypes and those of repo, loco and pnt suggested that gene regulatory relationships might exist between these genes. A considerable amount of information already exists about the nature of the transcriptional cascade required to establish longitudinal glial cells. The glial fate is induced by expression of gcm, while later expression of transcription factors encoded by repo and pnt direct glial differentiation(Akiyama et al., 1996;Bernardoni et al., 1998;Campbell et al., 1994;Hosoya et al., 1995;Jones et al., 1995;Klambt, 1993;Klambt and Goodman, 1991;Schreiber et al., 1997;Vincent et al., 1996;Xiong et al., 1994). We examined dri expression in embryos mutant for genes required for glial formation and differentiation. As expected, dri expression in all of the dorsal glia (but not in the dri-expressing lateral neural cells) depends on gcm (Fig. 7B). Moreover, with the probable exception of the A/B SPG, the normal levels of dri glial expression requires repo(Fig. 7C), as repomutant embryos show a significant reduction in the levels of driexpression and in the numbers of LG that contain DRI. However, driexpression in all dorsal glia does not depend on pnt(Fig. 7D). Analysis of thedri promoter region did not reveal any consensus binding sites for GCM (A/GCCCGCAT) or REPO (NNATTA, data not shown), suggesting thatdri might be not a direct transcriptional target of these genes. The finding that dri expression in all glia is not affected in embryos mutant for faint little ball (flb), a null allele of theDrosophila EGFR gene (Fig. 7E,F), is in line with previous observations that only the midline glia require EGFR signalling.

In a complementary set of experiments, the expression of glial differentiation markers in a dri mutant background was examined. Indri loss-of-function mutants, repo, pnt and cutcontinue to be expressed in the appropriate glia(Fig. 5; data not shown). However, we observed a reduction in the number of pros- andloco-positive glial cells (Fig. 8). One pros-positive glial cell was found to be consistently missing in dri mutant embryos (depicted with arrows inFig. 8A-D), while there was frequent, if irregular, reduction or loss of pros expression in three or four other LG (data not shown). Similarly, loco expression was reduced or lost in some LG, although this phenotype also exhibited variability in different segments (Fig. 8E,F). We scored the number of loco-positive cells in two neuromeres of abdominal segments in ten stage 15 embryos. The average number of dorsal glia per hemineuromere in dri¹/CyOwglacZheterozygotes was 9.8 (392 totally), not significantly different from wild-type numbers (Campbell et al.,1994; Halter et al.,1995; Klaes et al.,1994; Klambt,1993; Klambt and Goodman,1991; Xiong et al.,1994). By contrast, there was an average of 4.8 (193 scored)loco-positive cells per hemineuromere in dri¹homozygotes, confirming the significance of the apparent loss ofloco-expressing cells.

To understand the mechanisms that control formation of the central nervous system, it is important to identify the genetic and molecular components involved. We provide evidence that the transcriptional regulatory gene,dri, is one such component, required for normal embryonic nervous system development. The primary evidence for this comes from the axon bundling and tracking defects observed in dri mutant embryos. The axon phenotype includes incorrectly bundled and routed connectives, and axons that sometimes join the wrong bundle or cross from one tract to another. However,the axon defects do not reflect a role for dri in neurons.dri is not expressed in the neurons that give rise to the longitudinal axon bundles, nor do these neurons exhibit a mutant phenotype indri mutant embryos. Rather, dri was found to be expressed in a subset of glial cells, the longitudinal glia, A and B subperineural glia and the medialmost cell body glia, where it is required for normal differentiation and, as a consequence, normal axon fascicle formation.

Although dri is expressed early in glial development, two lines of evidence allow us to conclude that dri does not play a role in the neural/glial cell fate decision. First, ectopic expression studies showed that DRI cannot induce a glial cell fate in non-glial cells, nor was it involved in repression of the neural fate during glial cell formation. Second, loss ofdri function was found not to impair the initiation of gliogenesis. The first axon tracts were established correctly, indicating that thedri mutation does not cause any dramatic glial cell fate changes. In addition, dri mutant longitudinal glia were still able to undergo their initial migration to the correct position relative to the midline and to navigate the growth cones of pioneer axons. This process depends on the ability of the Robo receptor, on the glial cell, to recognize the repulsive Slit signal emanating from the midline(Halter et al., 1995;Jacobs et al., 1989;Kinrade et al., 2001;Schmidt et al., 1997). Thus,we conclude that Robo function is not affected in dri mutants.

After the initial migration and pioneer axon navigation, however, the behaviour of dri-expressing glial cells becomes aberrant. The normal final positions of these cells are never adopted and the cells exhibit cell shape defects. The mild misplacement of LG in dri mutants is probably caused by defects in glia-glia and axon-glia contacts, resulting at least in part from downregulation of the glial cell surface marker Neuroglian. These defects may interfere with correct migration of glia along the axon bundles which, in turn, causes the axon tract defects.

What is the molecular basis of the mutant phenotype found in drimutants? DRI is a transcription factor, so the link between loss ofdri function and the failure to differentiate properly is likely to be indirect, mediated through misregulation of dri targets required for normal longitudinal glial development. The most informative data came from our analysis of the position of dri in the glial transcriptional regulatory cascade. In general terms, dri activity was found to be downstream of gcm and repo, and independent of pntand cut. It was also found to be upstream of two genes, locoand pros, which are essential for normal development of some glial cells. In this developmental context dri acts as an activator of downstream targets. A consensus DRI binding-sequence has been defined(Gregory et al., 1996), but neither the loco nor pros promoter regions contain such sequences. This raises the possibility that loco and prosmight be indirect targets of DRI activity in the hierarchy of glial cell differentiation. Confirmation of this will require further analysis ofloco and pros transcriptional regulation and perhaps the discovery of additional transcriptional regulatory factors involved in glial development.

The requirement for DRI in the activation of loco is unexpected.loco has been found to be a transcriptional target of PNT but not of REPO (Granderath et al., 2000;Granderath et al., 1999b),while we found that dri expression depends on REPO and not on PNT. It is possible that expression of loco is co-dependent on PNT and DRI in some cells and that the reduced level of dri expression observed inrepo mutants is enough to permit loco expression.

The genetic analysis presented here strengthens the hypothesis that there are different genetic controls for different subsets of dorsal glia. For example, dri expression in all glial cells requires GCM activation,but only some of them requires REPO. The REPO-independentdri-positive cells, two per hemineuromere, appeared to correspond to the A and B subperineural glia (A/B SPG). These derive from neuroglioblast NB1.1 (Halter et al., 1995;Klambt and Goodman, 1991;Udolph et al., 1993),suggesting that REPO is required only for the expression of dri in cells derived from the lateral glioblasts. Unlike dri, pnt and its downstream target loco are not expressed in the MM CBG, which do not have a lateral glioblast origin. This suggests that there are different pathways of pnt and dri induction downstream ofgcm.

At least some of these hierarchical transcriptional interactions may explain the phenotypes observed. The axon and mild positional defects of glia in dri mutants resemble phenotypes of other known late gliogenesis factors, such as those observed in pnt, repo, loco or prosembryos. It is known that early distribution of the glycoprotein Neuroglian is perturbed in pros mutant embryos(Jacobs, 1993). locoencodes a regulator of G-protein signalling (RGS) that has been shown to bind to a Gαi-subunit (Granderath et al.,1999b) and could regulate a G-protein signalling pathway involved in LG migratory behaviour. In addition, expression of Heartless, theDrosophila FGF receptor, in LG, and similarities between theloco and heartless mutant phenotypes(Granderath et al., 1999b;Shishido et al., 1997) leaves open the possibility that FGF could trigger final migration of glia along the longitudinal connectives. This hypothesis is strengthen by the recent finding that subcellular redistribution of Neuroglian from the plasma membrane to cytoplasm, which normally happens during final glial migration to enwrap axon bundles, is disrupted in heartless mutants(Takizawa and Hotta, 2001). Alternatively, it remains possible that additional targets of drimediate the role of this gene in longitudinal glial differentiation.

Our studies add dri to the list of genes, including pnt, repo,loco and pros, that exhibit phenotypes that are much milder than those of the gcm, glide2 and Drop/Ltt genes at the head of the dorsal glia hierarchy (Buescher and Chia, 1997; Campbell et al.,1994; Giesen et al.,1997; Granderath et al.,1999b; Halter et al.,1995; Kammerer and Giangrande,2001; Klaes et al.,1994; Klambt,1993; Klambt and Goodman,1991; Xiong et al.,1994). It appears that diversification of these downstream regulators produces different types of glial cells. Nonetheless, each plays an essential role in driving the required behaviour of glial cells during CNS development. In the case of the DRI transcription factor, this role includes fine tuning the cell shape and migration characteristics of longitudinal glia that enable them to establish a normal axon scaffold.

This work was supported by the Australian Research Council. We are very grateful to Paul Whitington, Barry Dickson, Simon Koblar and William Chia for helpful discussions. We thank C. Goodman, P. Silver and A. Travers for providing antibodies. The monoclonal antibodies 9F8A9 anti-ELAV, 2B10 anti-CUT(developed by G.M.Rubin), MR1A anti-Prospero (developed by C.Q Doe), 22C10(developed by S.Benzer), BP 102 anti-CNS axons (developed by C. Goodman) were obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the NICHD and maintained by The University of Iowa, Department of Biological Sciences, Iowa City, IA 52242.