Neurogenic role of Gcm transcription factors is conserved in chicken spinal cord

Glial cells missing transcription factors (Gcm and Gcm2) are considered as the glial determinants in the fly embryonic central nervous system (CNS) (for a review, see Van De Bor and Giangrande,2002). Their absence leads to glia-to-neuron transformation, and ectopic expression of gcm or gcm2 leads to differentiation of supernumerary glia at the expense of neurons, indicating a role as a molecular switch between neuronal and glial fates(Alfonso and Jones, 2002; Hosoya et al., 1995; Jones et al., 1995; Kammerer and Giangrande, 2001; Vincent et al., 1996).

Although structurally conserved throughout evolution, the function of these two genes in vertebrate gliogenesis has remained elusive(Altshuller et al., 1996; Anson-Cartwright et al., 2000; Basyuk et al., 1999; Kanemura et al., 1999; Kim et al., 1998; Schreiber et al., 2000). Previous studies have shown that Gcm1- deficient mice die during embryogenesis owing to developmental defects of the placenta(Anson-Cartwright et al., 2000; Schreiber et al., 2000),whereas Gcm2-deficient mice are viable but lack a parathyroid gland(Gunther et al., 2000). No defects were found in the CNS; however, Gcm1-knockout mice die by E10, thereby preventing functional analyses in the nervous system. Moreover,RT-PCR experiments revealed the presence of Gcm1 transcripts in mouse brain, starting from E12 (Iwasaki et al.,2003).

We and others have recently shown that fly gcm genes are required in a specific neuronal population of the larval visual system(Chotard et al., 2005; Yoshida et al., 2005). During optic lobe development, neuroblasts produce lamina precursor cells (LPCs),which divide once to produce lamina neurons(Selleck and Steller, 1991). gcm and gcm2 are coexpressed in LPCs and, in their absence,lamina neurons are not produced, indicating that these two genes are required in neuronal precursors to induce neuronal fate(Chotard et al., 2005; Yoshida et al., 2005). This puzzling observation raises the possibility that vertebrate gcm genes might also be required for neuronal differentiation, and calls for the role of gcm genes in evolution to be re-evaluated.

We here show that the chicken ortholog of fly gcm (herein referred to as c-Gcm1), but not the chicken ortholog of fly gcm2(herein referred to as c-Gcm2), is expressed in early neuronal lineages of the developing chick spinal cord. Moreover, full neuronal differentiation is prematurely induced upon c-Gcm1 overexpression and inhibited upon blocking its pathway, thus revealing a neurogenic, rather than a gliogenic, role. We show that c-Gcm1 is required for precursors to enter the post-mitotic state, acting downstream of the neurogenin and Sox1-3 genes and upstream of NeuroM. Thus, we demonstrate for the first time that a vertebrate gcm gene is expressed and required in the CNS. We also show that, in flies, the neurogenic role of the gcm genes extends to the newly identified interneurons of the central brain. Moreover, gcm or gcm2 overexpression can induce ectopic neuronal differentiation after embryonic development. Finally, we show that fly, chick and mouse gcm genes induce expression of neuronal and glial markers in transfected HeLa cells. Thus, gcm genes have a double potential that is conserved during evolution.

The wild-type strain used was Sevelen. Flies were raised at 25°C, except for experiments with tub-gal80^ts lines. gcm-gal4 carries a gal4 P-element inserted into the gcm promoter. UAS-gcm^DN was used to block gcm-gcm2 function with UAS-gcm^N7-4DN as a control(Soustelle et al., 2004). UAS-ncGFP targets GFP to the nucleus and cytoplasm, whereas UAS-mCD8GFP targets GFP to the membrane. actin-gal4 was used for overexpression experiments with UAS-gcm(Bernardoni et al., 1998) or UAS-gcm2 (Kammerer and Giangrande, 2001).

The c-Gcm1 RNA probe was synthesized from chicken EST pgr1n.pk002.g21 (Chicken EST project, Delaware Biotechnology Institute,Newark, USA). NeuroM and Sox2 probes were generated from specific PCR fragments to avoid cross-reactivity with transgenes. The c-Gcm2 probe was synthesized from the partial cDNA template described by Okabe and Graham (Okabe and Graham,2004). Chicken embryos were staged according to Hamburger and Hamilton (Hamburger and Hamilton,1992). In situ hybridization was performed according to Braquart-Varnier (Braquart-Varnier et al.,2004). Sections were digitalized and analyzed using Zeiss software, and images were manipulated using Adobe Photoshop. In situ hybridizations were performed according to Kammerer and Giangrande(Kammerer and Giangrande,2001), except that Drosophila brains were fixed overnight at 4°C in 4% paraformaldehyde. Double in situ hybridization was performed using probes labeled with either digoxigenin or fluorescein. Mouse anti-digoxigenin (1:100, Boehringer Mannheim) or rabbit anti-FITC (1:10,000,Molecular probes) and fluorescent secondary antibodies were used for immunolabeling.

The Drosophila CNS was immunolabeled as described previously(Ceron et al., 2001). Primary antibodies used were: mouse anti-Repo (1:100, DSHB), rat anti-Elav (1:200,supernatant from DSHB), rabbit anti-phospho-Histone H3 (1:500, Upstate Biotechnology), rabbit anti-GFP (1:500, Molecular probes), mouse anti-Acj6(1:50, DSHB). Secondary antibodies conjugated with FITC, Cy3 or Cy5 (Jackson)were used at 1:500. DAPI was used at 100 ng/ml for nuclei counterstaining. Brains were mounted in Vectashield (Vector) mounting medium.

Chicken tissue vibratome sections fixed in 4% formaldehyde were processed using standard protocols. Primary antibodies used were: rabbit anti-Pax6(1:150, Covance), mouse (1:2000) and rabbit (1:1000) anti-βIII-tubulin(Covance), mouse anti-Mnr2 (1:4, DSHB), mouse anti-Pax7 (1:2, DSHB), mouse anti-Lim1/2 (1:2, DSHB), mouse anti-BrdU (1:2000, DSHB), mouse anti-HuC/D(1:500, Molecular Probes), rabbit anti-GFP (1:500, Torrey). Rabbit anti-Sox1(1:800), anti-Sox2 (1:4000) and anti-Sox3 (1:500) were used, but as similar results were obtained from each only data relative to Sox3 are shown. Secondary antibodies used were: anti-mouse Ig Alexa-546, anti-rabbit Ig Alexa 647 (1:1000, Molecular Probes). Sections or brains were analyzed using confocal microscopes (Leica).

actin-gal4, tub-gal80^ts, UAS-gcm or UAS-gcm2LII larvae were incubated at 30°C and brains dissected at LIII. The c-Gcm1 coding region(Hashemolhosseini et al.,2004) was cloned into the pCIG vector. The c-Gcm1BD-ER vector was constructed by cloning a PCR-amplified fragment (equivalent to amino acids 1-193 of c-Gcm1) into the CMV-based vector pCS2, which contains the repressor domain (amino acids 1-298) of the D. melanogaster Engrailed protein(Smith and Jaynes, 1996). The coding region of Ngn2 was cloned into expression plasmid pAdRSV-Sp. In ovo chicken electroporation was performed as previously described(Danesin et al., 2006; Itasaki et al., 1999). Expression vectors (1 μl) were used at 2 μg/μl, except for the c-Gcm1BD-ER construct which was used at 3 μg/μl, and were coinjected with control vectors pEGFP-N1 (Clontech, 0.6 μg/μl) or pCIG to visualize the transfected area. Flat-mounted spinal cord explants were cultivated in an organotypic culture system as previously described(Agius et al., 2004). Embryos or spinal cord explants were fixed in 4% formaldehyde in PBS for 2 hours for immunohistochemistry, or overnight for in situ hybridization, and dehydrated in an ethanol series. Cell proliferation was evaluated by bromodeoxyuridine(BrdU, Roche) incorporation. BrdU (10 mM) was injected into the lumen of the chicken neural tube and embryos were harvested 1 hour later. BrdU immunolabeling was performed on sections treated with 2N HCl in 0.2% Triton X-100/PBS for 30 minutes, after blocking and before incubation with primary antibodies.

Quantifications were performed by calculating the percentage of electroporated (GFP-positive) cells labeled with a particular marker, divided by the total number of GFP-positive cells. Counts were performed on at least ten sections from three embryos in each experiment. Data were obtained from at least three independent experiments, in each of which at least 150 cells were counted. Results are expressed as the mean percentage±s.e.m. of labeled electroporated cells. Significance was analyzed using the Student's ttest and P values are indicated in the figure legends.

HeLa cells were grown in DMEM supplemented with fetal bovine serum and antibiotics and transfected with Effecten Transfection Reagent (Qiagen). Each cDNA was cloned into the pCIG vector. After 48 hours, transfected cells were analyzed by immunolabeling, using rabbit anti-Tuj1 (1:1000, Covance), mouse anti-GFAP (1:100, ICN Biomedicals), fluorescent secondary antibodies (1:600)and DAPI as above. Preparations were mounted in Polymount medium(Polysciences) and analyzed by confocal microscopy. Labeled cells were counted from three independent transfection assays. Results are expressed as the mean of counted cells.

To investigate the role of vertebrate gcm genes during neural development, we determined the expression profile of the two chicken orthologs c-Gcm1 (Hashemolhosseini et al.,2004) and c-Gcm2(Okabe and Graham, 2004). c-Gcm1 RNA was observed as early as the primitive streak stage, in the presumptive neural plate (Fig. 1A). c-Gcm1-positive cells were found in the ectoderm(Fig. 1A,A′) and were included in the domain expressing Sox2(Fig. 1B), a marker of early neural progenitors (Bylund et al.,2003; Graham et al.,2003; Uwanogho et al.,1995). At the onset of rostral neural tube folding, high levels of transcripts were detected at all rostrocaudal levels of the developing CNS along the entire dorsoventral axis (Fig. 1C,D), except for the dorsal-most region(Fig. 1E). At E2.5, when neuronal differentiation has started(McConnell and Sechrist, 1980; Sechrist and Bronner-Fraser,1991), c-Gcm1 expression was seen to be markedly reinforced in cells located in the lateral wall of the forming mantle layer,which are likely to correspond to newly generated neurons(Fig. 1F). From E3, c-Gcm1 RNA was not detected in the ventricular zone nor in the forming mantle layer (Fig. 1Gand data not shown), indicating that c-Gcm1 expression is extinguished in late neuronal progenitors, before the appearance of glial progenitors (Rowitch, 2004). By contrast, c-Gcm2 RNA was never detected in the developing spinal cord at any of the stages analyzed, from E1 to E8 (see Fig. S1 in the supplementary material). However, in E1.5 embryos, c-Gcm2 was expressed in a restricted rostrocaudal domain of the hindbrain (see Fig. S1B,B′ in the supplementary material).

These results reveal for the first time the expression profile of gcm genes in the embryonic vertebrate CNS.

To directly assess the function of c-Gcm1, we in ovo electroporated a c-Gcm1-expressing vector at E1.5 and analyzed neural tubes using markers specific to progenitor or post-mitotic cells. Electroporation of the c-Gcm1 vector cell-autonomously suppresses the expression of Sox1-3, which code for HMG-box transcription factors specific to CNS proliferating progenitors(Fig. 2A,A′,G)(Bylund et al., 2003; Graham et al., 2003; Uwanogho et al., 1995). Only 15% of c-Gcm1-overexpressing cells coexpressed Sox1-3, as compared with 51% when control vector was electroporated(Fig. 2G). The effects of c-Gcm1 overexpression on Sox1-3, which were observed as early as 6 hours after electroporation, suggest that they may be direct(Fig. 2A,A′) and concomitant with upregulation of NeuroM(Fig. 2B,B′), a marker for early post-mitotic neurons (Roztocil et al., 1997).

To further demonstrate that c-Gcm1 is involved in directing neural progenitors towards a neuronal fate, we analyzed the proliferation profile of c-Gcm1-overexpressing neural progenitors that were subjected to a 1-hour BrdU incorporation pulse 30 hours after electroporation. Only 9% of c-Gcm1 electroporated cells were BrdU-positive(Fig. 2F,H), as compared with 24% for the control (Fig. 2E,H). Thus, c-Gcm1 overexpression causes cell cycle exit. To determine whether this phenotype is associated with loss of proliferative neural progenitors, we analyzed the expression of Pax6 and Pax7,homeodomain transcription factors expressed in intermediate and dorsal progenitors of the ventricular zone, respectively(Briscoe and Ericson, 2001; Ericson et al., 1997). Electroporation with c-Gcm1 cell-autonomously suppressed Pax6(Fig. 2I,K,K′,M) and Pax7(Fig. 2N,N′,P)expression. These data show that NeuroM induction upon c-Gcm1 overexpression is linked to a reduced rate of proliferation in the neural tube and to cell-autonomous loss of progenitor status.

Vertebrate neurogenesis is driven by proneural bHLH transcription factors such as the neurogenins (Ngns), which direct cell cycle exit of neural progenitors by repressing the expression of Sox1-3, and which promote neuronal differentiation by inducing the expression of NeuroM and NeuroD(Bertrand et al., 2002). Interestingly, Ngn2 overexpression upregulated c-Gcm1(Fig. 2D,D′), whereas the reverse was not true (Fig. 2C,C′), placing c-Gcm1 downstream of proneural and upstream of neuronal differentiation factors.

To determine whether premature engagement of neural progenitors towards a neuronal fate results in a full neuronal phenotype, we analyzed the expression of the pan-neuronal markers class III β-tubulin (Tuj1, Fig. 3A-F) and HuC/D(Fig. 3G-H′), as well as that of interneuron marker Lim1/2 (Fig. 3F,I-J′). Twenty-four hours after electroporation, 55% of c-Gcm1-overexpressing cells expressed βIII-tubulin and 40% were Lim1/2-positive, whereas only 9% of cells electroporated with a control vector differentiated into neurons in the same time window(Fig. 3F). Overexpression of c-Gcm1 also triggered neuronal differentiation in the embryonic brain, as assessed by premature expression of βIII-tubulin in the forebrain (Fig. 3D). Neuronal progenitors normally exit the ventricular zone when they leave the cell cycle and start expressing neuronal differentiation markers. This behavior was retained in neural cells electroporated with a control vector(Fig. 3A,A′,G,G′,I,I′), whereas c-Gcm1-induced neurons were also found in the ventricular zone(Fig. 3B,B′,H,H′,J,J′) and no longer expressed Pax7(Fig. 3E), a feature indicative of premature differentiation.

These data demonstrate that c-Gcm1 overexpression prompts neural progenitors to differentiate prematurely by downregulating neural progenitor genes and upregulating a repertoire of neuronal characteristics.

We then asked whether c-Gcm1 has the ability to induce glial cells in the embryonic spinal cord. We used Glast and O4 to identify astrocyte and oligodendrocyte precursors, respectively. Both markers start to be expressed in precursor cells of the neuroepithelium from E5/E6, corresponding to initiation of gliogenesis in chicken (Agius et al., 2004; Soula et al.,2001).

First, c-Gcm1 was electroporated at E1.5 and glial marker expression was analyzed 1.5 to 4 days later. Irrespective of the stage of phenotypic analysis (E3 or E5.5), c-Gcm1 overexpression did not trigger premature and/or ectopic expression of glial markers(Fig. 4A,B and data not shown). Strikingly, O4-positive cells were not detected in the ventral domain of the neuroepithelium (Fig. 4B).

Second, we overexpressed c-Gcm1 at late stages in a gliogenic context. E4.5/E5 spinal cords were electroporated ex ovo, plated in an organotypic culture system and expression of glial markers assessed three days later on transverse sections of spinal cord explants(Danesin et al., 2006). As with early electroporation, late c-Gcm1 overexpression did not induce ectopic expression of glial markers (Fig. 4C-G), and a reduction of the O4-expressing domain was observed(Fig. 4E). Accordingly, we found that c-Gcm1-overexpressing cells invariably expressed Lim1/2(Fig. 4H,I) and most had left the neuroepithelium, indicating that they had adopted a neuronal fate. Altogether, our data show that c-Gcm1 overexpression induces neurogenesis at both early and late stages of spinal cord development.

To further investigate the function of c-Gcm1 during neurogenesis,we generated a construct containing the repressor domain (ER) of the D. melanogaster Engrailed protein fused to the DNA-binding domain of c-Gcm1(c-Gcm1BD-ER). ER fusion constructs have been successfully used to produce dominant-negative mutations that reproduce loss-of-function phenotypes in vertebrates and in flies (Feig,1999).

Electroporation of c-Gcm1BD-ER invariably lead to a marked reduction in the number of cells expressing βIII-tubulin(Fig. 5A,A′), Lim1/2(Fig. 5B,B′,E,F,I) and the motoneuron marker Mnr2 (Fig. 5J) (Tanabe et al.,1998). Similarly, NeuroM expression was reduced as compared with electroporation with a control vector(Fig. 5C-D′). Thus,c-Gcm1 activity is required to induce expression of neuronal differentiation genes both in ventral and dorsal regions of the neural tube. A similar proportion of GFP-positive cells incorporated BrdU after electroporation of c-Gcm1BD-ER or control vectors (Fig. 5G,H,M), indicating that antagonizing c-Gcm1 activity does not affect the proportion of cells in S phase. Similarly, the expression patterns of Sox1-3, Pax6 and Pax7 were not modified in neural tubes electroporated with c-Gcm1BD-ER as compared with the control(Fig. 5K,L,N-O′ and data not shown). In addition, no apoptosis was observed by TUNEL assay in cells electroporated with control or c-Gcm1BD-ER expression vectors (data not shown). Thus, repressing c-Gcm1 target genes prevents expression of neuronal differentiation genes but does not impede cell cycle exit nor does it affect progenitor identity. Moreover, c-Gcm1 does not control spatial patterning in the chick spinal cord. The finding that c-Gcm1BD-ER does not affect general developmental pathways excludes the possibility that this construct induces non-specific effects, as also confirmed by the fact that coexpressing wild-type c-Gcm1 and c-Gcm1BD-ER is sufficient to rescue the loss of neurons induced by c-Gcm1BD-ER alone (see Table S1 in the supplementary material).

The puzzling finding that c-Gcm1 induces neuronal differentiation prompted us to revisit fly gain-of-function phenotypes, as gcm and gcm2 are known to induce gliogenesis when overexpressed in the embryo(Hosoya et al., 1995; Jones et al., 1995; Vincent et al., 1996). Ubiquitous gcm or gcm2 expression [shift at restrictive temperature during second larval instar (LII)] induced a discrete increase in glial cell number in the CNS but did not induce ectopic expression of neuronal markers (data not shown). By contrast, numerous cells positive for the Elav neuronal marker were induced outside the CNS (for example, see the wing disc in Fig. 6), indicating that gcm as well as gcm2 (data not shown) can trigger the neuronal fate during post-embryonic development. Moreover, a subset of Elav-positive cells expressed Acj6, an interneuron marker(Certel et al., 2000) that is not expressed in wild-type imaginal discs(Fig. 6D-F″). This indicates that at least two different neuronal subpopulations are induced following gcm or gcm2 ectopic expression. Discs also displayed ectopic glial marker labeling, which did not colocalize with neuronal labeling (Fig. 6A-C″). Elav-positive cells were observed adjacent to Repo-positive cells or in clusters that only expressed the neuronal marker(Fig. 6B″,C″),indicating that the two phenotypes can be induced independently. Similar data were obtained in all the other imaginal discs (data not shown). Altogether,these results show that gcm and gcm2 are able to trigger glial as well as neuronal differentiation during post-embryonic development.

In situ hybridization and immunolabeling using independent gcm-specific enhancer trap lines (rA87, gcm-gal4) showed that gcm and gcm2 are coexpressed in two clusters(dorsolateral and medial) per central brain hemisphere(Fig. 7A-C′, Fig. 8A, and see Figs S2-S4 in the supplementary material). Each of these clusters, which we will refer to as central brain clusters (or cbcs), contains some forty neurons that persist in the adult, where they represent the only labeled cells of gcm-gal4,UAS-GFP animals (Fig. 7C,C′ and see Figs S3, S4 in the supplementary material). Cbcs appeared and coexpressed gcm-gcm2 at early LII, reaching their final size by the end of the third instar larval stage (LIII) (data not shown). At no stage did cbc cells express the Repo glial marker (see Figs S2,S4 in the supplementary material). Interestingly, phospho-Histone H3 labeling revealed the presence of one mitotically active cell(Fig. 7D-D‴). Altogether, these data imply that cbcs originate from pure neuronal progenitors and that progenitors as well as neurons express gcm-gcm2,suggesting that the neurogenic role of gcm-gcm2 is not restricted to the visual system.

Expression of a dominant-negative construct (gcm^DN)containing the Gcm DNA-binding domain and the repressor domain of Engrailed blocked both Gcm and Gcm2 activities, producing the same phenotypes as those observed in the double gcm-gcm2 deficiency(Chotard et al., 2005; Soustelle et al., 2004; Yoshida et al., 2005). By contrast, gcm^DN expression in neurons that do not express gcm-gcm2 did not induce any defect(Fig. 8C,D), further confirming that the gcm^DN construct acts by blocking the Gcm pathway rather than by non-specifically repressing gene expression.

This construct provided us with a unique opportunity to explore the role of gcm-gcm2 in cells that can be specifically visualized by the gcm-gal4 driver. As gcm-gal4;UAS-gcm^DN animals die as embryos, we crossed gcm-gal4;UAS-gcm^DN flies with tub-Gal80^ts transgenic flies, which ubiquitously express a temperature-sensitive Gal80 protein that represses Gal4 at 18°C(McGuire et al., 2003). gcm-gal4, tub-gal80^ts;UAS-gcm^DN animals are viable and fertile when grown at 18°C but do not hatch when grown at 30°C. Flies expressing gcm^DN (shift at early LII)showed the same visual system neuronal and glial loss (see Fig. S5 in the supplementary material) as observed in the gcm-gcm2 double deficiency(Chotard et al., 2005),providing a final validation of the approach. Importantly, these phenotypes were not observed in animals containing a gcm^N7-4DNtransgene, which carries a point mutation abolishing DNA binding(Soustelle et al., 2004; Vincent et al., 1996). All cbc neurons were missing in gcm-gal4,tub-gal80^ts;UAS-gcm^DN, UAS-ncGFP animals(Fig. 8B), whereas they were still present in gcm-gal4,tub-gal80^ts;UAS-gcm^N7-4DN, UAS-ncGFP control animals (Fig. 8A). Interestingly, the cbc phenotype was observed when using one gcm^DN transgene, whereas defects in the visual system were only induced when using two transgenes, suggesting that different cells display distinct requirements for the Gcm pathway and further confirming the specificity of gcm^DN phenotypes.

Altogether, these data indicate that the Gcm pathway is necessary for the differentiation of specific neuronal populations of the larval brain.

The finding that fly gcm genes induce different cell fates,whereas c-Gcm1 only induces neuronal differentiation, prompted us to determine the potential of fly and vertebrate gcm genes in a cellular context. We transfected HeLa cells, a human cell line of non-neural origin,with an expression vector carrying fly gcm or c-Gcm1. Because neurogenic potential had never been reported in mammals, we also performed the same type of experiments using a mouse Gcm1 expression vector.

Strikingly, the three gcm genes induced the expression ofβIII-tubulin in almost half of the transfected HeLa cells, indicating that all three genes share neurogenic potential (data not shown). Because previous studies in mouse fibroblasts showed that mouse Gcm1 induces expression of an astrocytic marker(Iwasaki et al., 2003), we also assayed the gliogenic role of the three gcm genes in a similar transfection assay and indeed observed GFAP expression in almost two-thirds of the transfected HeLa cells (data not shown). We then asked whether gcm genes induce the expression of neuronal and glial markers in the same cells, and indeed found colabeling upon transfection of fly, chicken and mouse gcm/Gcm1 genes(Fig. 9A-D‴). Interestingly, upon quantification of labeling, we found that most cells expressed both GFAP and βIII-tubulin (60% to 80%, depending on the species), the remainder expressing either marker(Fig. 9E). Thus, expression in a naive context reveals the double neurogenic and gliogenic potentials of gcm genes throughout evolution. These findings allow us to reconcile the apparent discrepancy between the gliogenic potential previously described for fly (Hosoya et al., 1995; Jones et al., 1995; Vincent et al., 1996) and mouse (Iwasaki et al., 2003)genes and their neurogenic potential in fly and chicken(Chotard et al., 2005; Yoshida et al., 2005) (see also present study). Interestingly, transfection of gcm genes does not induce expression of the O4 antigen, which is specific to oligodendrocyte precursors (data not shown), indicating that the observed increase in GFAP andβIII-tubulin labeling is not due to a general, non-specific enhancement of gene expression. Altogether, in vitro and in vivo data support the hypothesis that the Gcm class of transcription factors plays versatile roles in the fly and vertebrate CNS.

The present study provides the first evidence for a vertebrate gcmgene, c-Gcm1, being expressed and required for neuronal differentiation in the CNS. Fly gcm and gcm2 are also coexpressed and required in central brain neurons and are able to induce neuronal differentiation. In addition, gcm genes from fly, chicken and mouse are all able to induce the expression of neuronal markers in vitro. These data point to an unpredicted and conserved neurogenic role of gcm genes and further our understanding of the multistep process that builds the nervous system.

Data concerning the neurogenic role of fly gcm genes and expression of c-Gcm1 at the onset of neurogenesis formed the basis for a functional analysis in the chicken spinal cord. We here show that c-Gcm1 overexpression leads to the downregulation of genes specifically expressed in proliferative neural progenitors of the ventricular zone and concomitantly induces the full neuronal differentiation program. Conversely, blocking the c-Gcm1 pathway does not modify the proliferation rate and expression of progenitor genes, but prevents neuronal differentiation gene expression. c-Gcm1 acts as a neuronal differentiation gene,downstream of proneural genes and upstream of NeuroM. Neuronal differentiation relies on the repression of the Sox genes, which has been proposed to be mediated by neurogenin(Bertrand et al., 2002; Bylund et al., 2003; Graham et al., 2003; Uwanogho et al., 1995). Blocking Sox gene expression, however, leads cells to acquire an early post-mitotic phenotype, but does not allow these cells to fully differentiate into neurons (Bylund et al.,2003), indicating that late events require other Ngn-dependent pathways. c-Gcm1 is likely to constitute a relay of the neurogenin pathway that is indispensable to trigger complete neuronal differentiation.

The widespread expression of c-Gcm1 and its overexpression phenotype strongly argue in favor of a general neurogenic role. Indeed,altering the c-Gcm1 protein does not modify expression of ventral,intermediate or dorsal markers (Fig. 5 and data not shown), although further experiments are required to definitively exclude a role for c-Gcm1 in specifying neuronal subpopulations. Absence of c-Gcm2 expression in the spinal cord excludes the possibility that the two c-Gcm genes play a redundant role in this tissue, although we cannot exclude a c-Gcm2neurogenic/gliogenic role in other regions/stages.

Our study provides the first evidence for c-Gcm1 playing an essential role in the neurogenic process and points to an unexpected conservation of gcm gene function in neuronal differentiation in chicken and Drosophila. The question now is whether gcmneurogenic expression and function are also conserved in mammals. Our HeLa cell transfection data showing that neuronal characteristics are induced by fly gcm, c-Gcm1 and mouse Gcm1 support the hypothesis of conservation. These results may explain the stimulation of secondary neurulation observed in transgenic mice overexpressing mouse Gcm1 in the developing tail bud (Nait-Oumesmar et al., 2002). In the light of present data, and based on the fact that the expression of mouse and human orthologs is detected by RT-PCR in developing brains (Altshuller et al.,1996; Iwasaki et al.,2003; Kim et al.,1998), it will be crucial to determine the expression profile of mammalian gcm genes within the CNS and score for neuronal defects in Gcm-knockout mice.

Fly gcm genes are expressed and required in the neurons of the central brain. Owing to the fact that the only available tool to identify and target cbcs is the gcm-gal4 line, it is not possible to assess the specific roles of gcm and gcm2. Their similar levels of expression, however, suggest that both genes are required, as demonstrated in the visual system (Chotard et al.,2005). gcm genes are necessary for neuron generation rather than maintenance, in line with the early expression of gcm-gcm2 and with the lack of any phenotype in experiments removing gcm-gcm2 function in post-mitotic neurons (data not shown). Generating independent markers for the two cbcs will be crucial for characterizing the gcm pathway and the role of these neurons, which integrate information from different centers of the fly brain (see Fig. S3 in the supplementary material).

Our data show that fly Gcm/Gcm2 transcription factors induce glia or neuron differentiation depending on the cellular context. First, Gcm or Gcm2-induced neuron to glia transformation is more effective in the embryonic than in the larval CNS (data not shown). Second, Gcm or Gcm2 overexpression induces both Repo and Elav expression outside the CNS, but only Repo expression within the CNS. This implies that most neural cells express factors that do not allow Gcm to induce a neuronal pathway, or contain a positive factor that induces Gcm to trigger the glial pathway.

Although fly gcm genes are also necessary and sufficient for glial differentiation, c-Gcm1 overexpression does not induce gliogenesis in vivo and indeed c-Gcm1 activity inversely correlates with the production of glia in the spinal cord. Similarly, no glial defect has been reported for the Gcm1 and Gcm2 knockout mice(Anson-Cartwright et al., 2000; Gunther et al., 2000; Kim et al., 1998; Schreiber et al., 2000) and the only reported in situ expression (mouse Gcm1) concerns cells that do not correspond to known glial lineages(Iwasaki et al., 2003). On the other hand, the gliogenic potential is conserved in vitro because gcmgenes induce glial markers in HeLa cells. We therefore speculate that gcm genes play a gliogenic role in discrete cell populations, in line with mouse Gcm1 overexpression inducing the expression of glial markers in brain cultures (Iwasaki et al.,2003), but not in retina (Hojo et al., 2000). The observation that most cells coexpress neuronal and glial markers in vitro, a situation that is not observed in vivo, strongly suggests that HeLa cells provide a naive context in which both potentials of gcm genes are revealed.

Altogether, the present data merit revisiting the role and potential of the classically defined fly glial master genes. We also identify a conserved neurogenic potential in this class of transcription factors and define a novel step in the pathways leading to neuronal differentiation. Finally, based on in vivo and in vitro data, we propose that fly as well as vertebrate gcmgenes induce neurons and/or glial cells depending on the cellular context. The reiterated use of Gcm developmental pathways within and outside the CNS further emphasizes the need for cell-specific factors. In the future, one of the most challenging issues will be to perform screens to identify cues that are necessary for the neuronal program and to determine common versus specific features of the different Gcm pathways. This will enable us to understand how spatio-temporal patterning regulates and integrates these pathways to control cell specificity.

We thank the Bloomington Center and U. Tepass and C. Desplan for flies; M. Boeglin, D. Hentsch and J. L. Vonesch for assistance with imaging; J. Colonques and F. Tejedor for sharing unpublished data; all group members for advice and C. Diebold for technical assistance and N. Arbogast for keeping stocks; colleagues at CBD for advice, especially S. Bel-Vialar and F. Pituello, and N. Escalas for technical assistance. We thank P. Charnay for Ngn2 vector; T. Edlund for Sox1-3 antibodies; A. Graham for c-gcm2 cDNA; and D. Henrique for Sox2 probe. The confocal microscopy facility was financed by MESR (95.V.0015). This work was supported by INSERM, CNRS, Hôpital Universitaire de Strasbourg, ARC (A.G.), LIGUE(A.G.) and EEC (A.G.). L.S. was supported by ARC, AFM; F.T. by ARSEP and ARC;C.J. by MRT; J.C. by EMBO (short-term fellowship).