Helix-loop-helix transcription factors of the Ebf/Olf1 family have previously been implicated in the control of neurogenesis in the central nervous system in both Xenopus laevis and the mouse, but their precise roles have remained unclear. We have characterised two family members in the chick, and have performed a functional analysis by gain- and loss-of-function experiments. This study revealed several specific roles for Ebf genes in the spinal cord and hindbrain regions of higher vertebrates, and enabled their precise positioning along the neurogenic cascade.
During neurogenesis, cell cycle exit appears to be tightly coupled to migration to the mantle layer and to neuronal differentiation. We show that antagonizing Ebf gene activity allows the uncoupling of these processes. Ebf gene function is necessary to initiate neuronal differentiation and migration toward the mantle layer in neuroepithelial progenitors, but it is not required for cell cycle exit. Ebf genes therefore appear to be master controllers of neuronal differentiation and migration, coupling them to cell cycle exit and earlier steps of neurogenesis.
Mutual activation between proneural and Ebf genes suggests that besides their involvement in the engagement of differentiation, Ebf genes may also participate in the stabilisation of the committed state. Finally, gain-of-function data raise the possibility that, in addition to these general roles, Ebf genes may be involved in neuronal subtype specification in particular regions of the CNS.
The construction of the vertebrate central nervous system (CNS) involves the coordinated production and differentiation of specific neuronal and glial cell populations according to developmental stage and location. Neurogliogenesis occurs from the ventricular zone, a proliferative neuroepithelium whose cells have properties of stem cells. Neuroepithelial cells can generate committed progenitors that are more restricted in their fate and proliferation potential, and whose progeny leave the cell cycle, migrate into the mantle layer and differentiate. As development proceeds, the mantle layer thickens, while the ventricular zone becomes thinner. During recent years, important insights into the genetic control of vertebrate neurogenesis have been gained, revealing an increasing complexity. These studies, building on observations initially carried out in Drosophila, first revealed the existence of so-called proneural genes, which encode transcription factors of the basic helix-loop-helix (bHLH) family, are expressed in the proliferative neuroepithelium, and are necessary and sufficient to promote the generation of the committed progenitors (Bertrand et al., 2002). They include genes of the achaete-scute (Mash1 in mouse and Cash1 in chicken) and neurogenin (Ngn1 and Ngn2) families. Proneural genes are subjected to repression by inhibitors (Hes/Her/Esr), which are themselves under the control of Notch signalling. Other bHLH genes are associated with differentiation of the committed progenitors and are expressed in early post-mitotic cells migrating toward the mantle layer (NeuroM, NSCL1), or in cells which have reached the mantle layer (NeuroD) (Begley et al., 1992; Roztocil et al., 1997). However, a specific gene type may not be restricted to a particular function during neurogenesis, as proneural genes have recently been shown to play additional roles, integrating positional information in the process and contributing to the specification of neuronal subtype identity (Fode et al., 2000; Gowan et al., 2001).
Another class of genes, those of the Ebf/Olf1 family (reviewed by Dubois and Vincent, 2001; Liberg et al., 2002), has been involved in the control of neuronal differentiation. Ebf/Olf1 was independently identified as a transcription factor implicated in mouse B-lymphocyte differentiation (Ebf) (Hagman et al., 1993; Travis et al., 1993) and in transcriptional control of the rat olfactory marker protein gene (Olf1) (Kudrycki et al., 1993; Wang and Reed, 1993). Three other rodent genes showing high similarity to Ebf (renamed Ebf1), Ebf2 (also known as Mmot1 or O/E-3), Ebf3 (also known as O/E-2) and O/E-4, were subsequently identified (Garel et al., 1997; Malgaretti et al., 1997; Wang et al., 1997; Wang et al., 2002). In other species, Ebf orthologs have been found in C. elegans, Drosophila, Xenopus and zebrafish (Crozatier et al., 1996; Bally-Cuif et al., 1998; Dubois et al., 1998; Prasad et al., 1998; Pozzoli et al., 2001). All Ebf proteins share an atypical non-basic helix-loop-helix motif (HLH), with an additional type 2 helix and a novel zinc coordination motif. This latter motif is involved in DNA binding, which also requires homodimerisation or heterodimerisation with other family members mediated by the HLH structure (Hagman et al., 1995). A detailed analysis performed in the mouse revealed that Ebf1, Ebf2 and Ebf3 are expressed along the entire rostro-caudal axis of the developing CNS, with overlapping patterns, except in the forebrain where each of them is restricted to specific regions (Garel et al., 1997; Wang et al., 1997). Expression is observed in differentiating neurones, Ebf2 being restricted to early post-mitotic neurones whereas Ebf1 and Ebf3 expression is maintained in more differentiated cells. These data suggest a general role for Ebf genes in neuronal differentiation. In accordance with this, overexpression of a dominant-negative mutant of Xenopus Ebf2, Xcoe2/XEbf2, prevents primary neurone differentiation in Xenopus embryos (Dubois et al., 1998). Both XEbf2 and XEbf3 were shown to have neurogenic activity in this system, but were proposed to be involved in different steps of the neurogenic cascade. Whereas XEbf2 acts at an early stage during the commitment process, downstream to XNgn1 but upstream of XNeuroD, XEbf3, which is expressed later, functions downstream to XNeuroD, in neuronal differentiation (Dubois et al., 1998; Pozzoli et al., 2001). In the mouse, analysis of Ebf1–/– mutants revealed neuronal differentiation defects in CNS regions where Ebf1 is the sole family member to be expressed. Hence, in the striatum, post-mitotic neurones that leave the subventricular zone (SVZ) en route to the mantle layer appear unable to downregulate genes normally restricted to the SVZ, or to activate some mantle-specific genes (Garel et al., 1999). In the hindbrain, facial branchiomotor neurones show an abnormal migratory pathway, presumably as a result of incorrect interpretation of environmental guiding cues (Garel et al., 2000). A migration defect was also observed in gonadotropin-releasing hormone-synthesizing neurones in Ebf2–/– mice (Corradi et al., 2003). These latter animals also show peripheral nerve defects, although it was not established whether this is of central and/or peripheral origin. Altogether these studies confirm that the Ebf genes play important and different roles in neurogenesis, and suggest that significant redundancy exists between them.
Although work performed in Xenopus has provided information on the position of Ebf genes in the neurogenic cascade, it was restricted to primary neurogenesis in the ectoderm, which significantly differs from neurogenesis in higher vertebrates, in which neurones are continuously generated in the proliferative neuroepithelium and migrate to the pial surface when they differentiate. Furthermore, in gene targeting experiments in the mouse, compensation between the different Ebf genes is likely to have prevented the unravelling of important aspects of their function. This prompted us to undertake a study of Ebf gene function in neurogenesis by electroporation of the chick embryo neural tube, where we can use both gain- and loss-of-function approaches in a higher vertebrate. We provide evidence that Ebf genes play general and essential functions in the neurogenic process in higher vertebrates, acting downstream to proneural genes. They are necessary for initiation of both migration toward the mantle layer and neuronal differentiation, but are not required for cell cycle exit, and thus allow the uncoupling of these two aspects of neurogenesis.
Materials and methods
Plasmid constructs and chick cDNAs cloning
All expression constructs were derived from plasmid pAdRSV-Sp (Giudicelli et al., 2003). For mouse Ebf1, a 1.8 kb DNA fragment containing the entire coding sequence was obtained by PCR amplification from plasmid pEBF17 (Hagman et al., 1993) using the following primers: 5′-TTTCATGCTAGCGATCCAGGAAAGCATCC-3′ and 5′-AAATTCCTTAAGGCAATTCTTTCACATGG-3′. The dominant-negative Ebf construct (ΔEbf), lacking 321 bp corresponding to the 107 N-terminal amino acids, was obtained by subcloning a BamHI-AflII fragment from the former construct. For chick Ngn2 and NeuroM, we used 700 bp DraIII-PvuII, and 1.6 kb EcoRI, DNA fragments from plasmids c-ngn-2 (Perez et al., 1999) and pBSK-NeuroM (Roztocil et al., 1997), respectively. For chick NSCL1, we used a 400 bp DNA fragment obtained by PCR amplification from the cDNA (Li et al., 1999), using the following primers: 5′-ATCCGTCTAGACATGCTCAACTCGGAGCA-3′ and 5′-AGACAATCGATGGGCGGCTCAGACATCCA-3′. N-cadherin and CRABPI cDNA probes were obtained by RT-PCR from 5-day-old whole chick embryo total RNA, using the following primers:
N-cadherin, 5′-TGGTAACTGTTGTCAAGCCCA-3′ and 5′-GGGTCTACAGCAGTGATGTTA-3′; and
CRABPI, 5′-ATGCCTAACTTCGCCCGCACCT-3′ and 5′-GCTCATCATTAGCTAATTCTCGAGT-3′.
PCR-amplified DNA fragments and cloning junctions were checked by sequencing. cDNA clones corresponding to chick Ebf1 and Ebf3 DNA binding domains were obtained essentially as described (Garel et al., 1997), using 5-day-old whole chick embryo total RNA. Degenerated oligonucleotide primers corresponding to the conserved amino acid sequences AHFEKQP and DNMFVHNN that flank the Drosophila collier and mouse Ebf1 DNA binding domains were as described previously (Garel et al., 1997). GenBank Accession numbers for chick Ebf1 and Ebf3 clones are AY270034 and AY270035, respectively.
In ovo electroporation
Electroporations were carried out as described (Itasaki et al., 1999; Giudicelli et al., 2001). Briefly, fertilised eggs were incubated at 37°C to the indicated embryonic stages. DNA in 10 mM Tris (pH 8.0) was injected into the neural tube at the following concentrations: ΔEbf, 2 μg/μl; reporter plasmids [pAdRSVβgal (Le Gal La Salle et al., 1993) and pEGFP-N1 (Clontech)], 0.5 μg/μl; other constructs, 1.5 μg/μl. The total DNA concentration in each electroporation experiment was kept constant by adding vector plasmid DNA. Embryos were electroporated at hindbrain or thoracic spinal cord levels. Electroporations were performed with a BTX820 electroporator (Quantum) with the following parameters: hindbrain level, 4 pulses of 25 V and 40 milliseconds at a frequency of 1 Hz; spinal cord level, 6 pulses of 25 V and 50 milliseconds at a frequency of 1 Hz. To evaluate the efficiency of electroporation, a GFP expression vector (pEGFP-N1) was systematically co-transfected in embryos to be processed for in situ hybridisation or peroxidase-revealed immunohistochemistry, and only efficiently electroporated embryos were analysed. Following electroporation, the eggs were incubated for 10 to 40 hours, as indicated. The embryos were harvested in phosphate buffered saline (PBS), fixed in paraformaldehyde (PFA, 4% in PBS) for 3 hours for immunohistochemistry, or for more than 6 hours for in situ hybridisation, and dehydrated in a methanol series. For β-galactosidase detection, the embryos were fixed for 20 minutes and stained with X-gal as previously described (Schneider-Maunoury et al., 1993).
In situ hybridisation
In situ hybridisation with digoxigenin-labelled riboprobes on whole-mount embryos or vibratome sections was performed as described (Wilkinson and Nieto, 1993) and double in situ hybridisation was as described previously (Giudicelli et al., 2001). In this latter case one of the probes was labelled with fluorescein-UTP. Digoxigenin and fluorescein were detected sequentially with alkaline phosphatase-coupled antibodies (Roche). NBT/BCIP (purple) staining was carried out first, typically on digoxigenin-labelled probes. After removal of the antibody, INT/BCIP (orange, red) staining was performed on fluorescein-labelled probes. After whole-mount in situ hybridisation, embryos were flat-mounted or microtome sectioned. For vibratome section (50 μm thick), embryos were embedded in 4% agarose or albumin/gelatine. The chick probes were as follows: cEbf1, cEbf3, N-cadherin and CRABPI (this work); R-cadherin (Inuzuka et al., 1991a); Cash1 (Jasoni et al., 1994); Ngn1 and Ngn2 (Perez et al., 1999); Islet1, Islet2 and Lim1 (Tsuchida et al., 1994).
Cell proliferation and cell apoptosis analyses
Cell proliferation was evaluated by bromodeoxyuridine (BrdU) incorporation. BrdU, 15% (w/v) in PBS, was injected into the lumen of the neural tube and the embryos were harvested 2 hours later. BrdU immunodetection was performed on sections treated with 2 N HCl, 0.5% Triton X-100 in PBS for 30 minutes at 37°C, after blocking and before incubation with primary antibodies. Apoptosis was detected by fluorescein labelling of DNA strand breaks (TUNEL, Roche) on 50 μm-thick vibratome sections.
Immunohistochemistry was normally revealed by fluorescence, except for the detection of neurofilaments in flat-mounted hindbrains. In this latter case, we used a biotinylated hamster antibody directed against mouse IgG (Vector, 1:400) and streptavidin-horse radish peroxidase (Amersham, 1:500), to detect the anti-neurofilament antibody. Peroxidase activity was revealed with diaminobenzidine (Sigma), in the presence of nickel ammonium to enhance the staining. For immunofluorescence, 50 μm-thick vibratome sections were prepared from embryos embedded in 4% agarose, and blocked in PBS containing 0.25% Triton X-100 and 5% donkey serum. Primary antibodies were incubated in the same solution overnight at 4°C. Incubations with the secondary antibodies were at room temperature for 2 hours. All washes were performed with PBS containing 0.25% Triton X-100. Sections were mounted in Vecta Shield (Vector). Immunofluorescence pictures were acquired on a Leica TCS 4D confocal microscope and assembled with Adobe Photoshop. Antibodies and dilutions were as follows: neurofilaments, mouse monoclonal 3A10 (1:20, Developmental Studies Hybridoma Bank (DSHB)); β-tubulin-type III (Tuj1), mouse monoclonal and rabbit polyclonal (1:500, Babco); BrdU, mouse monoclonal (1:100, Becton Dickinson); β-galactosidase, rabbit polyclonal (1:700 Cappel); Islet 1/2, mouse monoclonal 39.4D5 (1:100, DSHB); HA, rat monoclonal (1:400, Roche); Flag, rabbit polyclonal (1:200, Sigma); FITC-, Cy3- and Cy5-conjugated secondary antibodies (1:200-1:800, Jackson Immuno Research).
Identification and expression of chick Ebf genes
Before performing functional analyses of Ebf genes in the chick embryo, it was important to identify chick Ebf family members and to study their expression pattern in the developing neural tube. We followed a strategy that previously led us to the identification of mouse Ebf2 and Ebf3 (Garel et al., 1997). Degenerated oligonucleotide primers derived from highly conserved regions of the DNA binding domains of mouse Ebf1 and the Drosophila Ebf homologue collier (knot – FlyBase) were used for RT-PCR amplification of RNA from five-day-old whole chick embryos (see Materials and methods). Two different amplified sequences were identified as belonging to putative Ebf family members (GenBank Accession Numbers AY270034 and AY270035). Nucleotide and amino acid sequence comparisons strongly suggested that they correspond to the chick orthologs of mouse Ebf1 and Ebf3, respectively. Chick Ebf1 presented 87% and 99% identity with mouse Ebf1 at the nucleotide and amino acid levels, respectively (data not shown). Chick Ebf3 showed 86% and 100% identity with mouse Ebf3 at the nucleotide and amino acid levels, respectively (data not shown).
In situ hybridisation analysis of the expression of chick Ebf1 and Ebf3 in the developing neural tube revealed an activation coincidental with the onset of neurogenesis (Fig. 1). This was particularly obvious in the hindbrain, where neurogenesis is known to occur first in even-numbered rhombomeres (r), and then in odd-numbered ones. A similar dynamic pattern of activation was seen for Ebf1 and Ebf3, which were activated throughout the hindbrain except in r3 and r5 at stages HH13-HH14 (Fig. 1A,B); expression in these latter rhombomeres occurred only from stage HH15 (data not shown). Sections through the spinal cord revealed that Ebf1 and Ebf3 expression was mainly associated with the mantle layer, and was likely to correspond to differentiating post-mitotic neurones (Fig. 1C-F). These expression patterns suggest that Ebf gene activation may constitute a general feature of neuronal differentiation, in this part of the CNS at least. In addition, they are very similar to those observed for the murine orthologs (Garel et al., 1997; Wang et al., 1997). The conservation of gene sequences and expression profiles suggest that Ebf genes perform similar functions in mammals and birds.
Ngn2 and NeuroM promote Ebf gene expression
Since Ebf gene expression is induced during neurogenesis, we investigated whether genes known to promote this latter process might also cause extended Ebf gene expression. The atonal-related genes Ngn1 and Ngn2 are normally activated in proliferating cells in the ventricular zone, and are involved in the initiation of neurogenesis (Ma et al., 1996; Ma et al., 1999; Perez et al., 1999). To test a possible epistatic link between proneural and Ebf genes, we misexpressed Ngn2 in the chick neural tube by in ovo electroporation. The Ngn2 coding sequence was cloned into plasmid pAdRSV-Sp, placing it under the control of the Rous sarcoma virus long terminal repeat promoter, enhanced by the human type 5 adenovirus terminal repeat. This combination of regulatory elements leads to efficient transcription of genes inserted downstream (Giudicelli et al., 2003). Electroporation of the Ngn2 expression plasmid in one half of the neural tube at stage HH9-HH10 led to strong activation of Ebf1 and to weaker, but significant, enhancement of Ebf3 expression, as detected by in situ hybridisation (Fig. 2A,D).
We then investigated whether genes possibly acting more downstream in the neurogenic cascade could also lead to Ebf gene activation. The genes NeuroM and NSCL1 are transiently expressed in early post-mitotic neurones starting their migration to the mantle layer (Begley et al., 1992; Roztocil et al., 1997). As in the case of Ngn2, both genes were ectopically expressed in the neural tube by electroporation using the pAdRSV-Sp expression plasmid, and the expression of Ebf1 and Ebf3 was monitored by in situ hybridisation. The effect of NeuroM misexpression was very similar to that of Ngn2, leading to the activation of both Ebf genes (Fig. 2B,E). By contrast, NSCL-1 ectopic expression did not affect the Ebf gene expression pattern (Fig. 2C,F).
These experiments indicate that Ngn2 and NeuroM can promote Ebf gene expression during neuronal differentiation. They suggest that proneural and early differentiation genes lie upstream of the Ebf genes in the cascade of events controlling neurogenesis.
Ebf1 misexpression promotes neuronal differentiation
To directly assess Ebf gene function in neurogenesis, we first performed gain-of-function experiments by using in ovo electroporation in the chick neural tube with the pAdRSV-Sp expression vector. We chose to misexpress the mouse Ebf1 gene because the complete coding sequence was available and this gene appears to be very similar to the chick ortholog, both in terms of sequence and expression pattern. In the first series of experiments, stage HH15 embryos were co-electroporated with Ebf1 and lacZ expression plasmids to identify the electroporated cells [it has been shown that under these conditions, most electroporated cells receive both plasmids, (Dubreuil et al., 2000)]. As a control, some embryos were electroporated with the lacZ expression plasmid alone. In the control experiments, analysis of the lacZ-positive cells, performed by X-gal staining, indicated that, as expected, 20 hours after electroporation most cells presented a morphology and a localisation of neuroepithelial progenitors (Fig. 3A). This was still the case for the majority of the electroporated cells at 40 hours after electroporation, but at that time some X-gal-positive cells were restricted to the mantle layer and had presumably undergone differentiation (Fig. 3B). Analysis of the distribution of X-gal-positive cells in neural tubes co-electroporated with the Ebf1 expression plasmid revealed a very different situation: whereas at 20 hours after electroporation most of the cells presented a neuroepithelial morphology (Fig. 3C), like in the control, at 40 hours essentially all of the lacZ-positive cells were localised in the mantle layer (Fig. 3F). A time course analysis indicated that the migration of Ebf1-electroporated cells was largely engaged at 24 hours and was almost completed at 30 hours (Fig. 3D,E). Therefore ectopic expression of Ebf1 in neuroepithelial progenitors appears to result in the early onset of cell migration toward the mantle layer.
As neuroepithelial progenitors normally exit the cell cycle before engaging in migration towards the mantle layer, we investigated whether the early migrating Ebf1-expressing cells could still be proliferating. To track the Ebf1-electroporated cells in this case, we modified the expression vector to introduce an HA epitope at the N-terminal end of the protein. We then verified that this modification did not affect the activity of the protein in co-electroporation experiments with lacZ (data not shown). To analyse proliferation, chick neural tubes were electroporated with either the HA-Ebf1 expression construct or a control construct, Krox20m-HA (see below), incubated for 18 or 28 hours, and then subjected to a 2-hour BrdU incorporation pulse, which labels cells in S-phase. The embryos were then processed for immunofluorescence with antibodies directed against HA and BrdU. Analysis of neural tube transverse sections from embryos incubated for 30 hours after electroporation confirmed the previous results, i.e. most HA-Ebf1-positive nuclei were located in the mantle layer (Fig. 3H,J). In addition, it revealed that almost all of these nuclei were BrdU-negative (Fig. 3H), which suggests that the electroporated cells had left the cell cycle. When the embryos were recovered 20 hours after electroporation, more than half of the HA-Ebf1-positive nuclei were still within the ventricular zone (Fig. 3G,J). Nevertheless, the majority of these ventricular nuclei were BrdU-negative (Fig. 3J). The control construct, Krox20m-HA, was designed to produce a mutant derivative of another transcription factor, Krox20. This mutant, R409W, does not bind to DNA and has no transcriptional activity (Warner et al., 1999; Giudicelli et al., 2001); it is therefore expected to have no effect on neurogenesis. Indeed, 30 hours after electroporation, most Krox20m-HA-positive nuclei were still within the ventricular zone (Fig. 3I,J), in agreement with the lacZ electroporation experiments (data not shown). Furthermore, most of these ventricular nuclei were BrdU-positive (Fig. 3J), indicating that the corresponding cells were proliferating. In conclusion, these experiments indicate that ectopic expression of Ebf1 in neuroepithelial progenitors leads to early onset of both exit from the cell cycle and migration toward the mantle layer.
Finally, we investigated whether forced Ebf1 expression would result in modifications in the expression pattern of neuronal and neuroepithelial markers. As indicated above, at around stages HH13-HH14 the hindbrain is an interesting region in which to analyse neuronal differentiation because, in contrast to even-numbered rhombomeres, odd-numbered rhombomeres are still largely devoid of differentiated neurones, as indicated by neurofilament labelling (Fig. 3K). Analysis of neurofilament expression in Ebf1-electroporated hindbrain revealed both an increased number of differentiated neurones in even-numbered rhombomeres and an early onset of neuronal differentiation in odd-numbered ones (Fig. 3L). Furthermore, double-labelling for HA-Ebf1 and neurofilaments or another neurone-specific marker, neuronal class III β-tubulin (Tuj1), indicated that most Ebf1-positive cells also expressed these markers (Fig. 3M-O). An important aspect of the neuronal differentiation programme is the shift in the expression pattern of adhesion molecules that is associated with neuronal migration: for example, N-cadherin is expressed in neuroepithelial progenitors, whereas R-cadherin expression is activated later during neurogenesis and has been suggested to be associated with post-mitotic cells (Inuzuka et al., 1991b; Redies and Takeichi, 1996). We found that Ebf1 ectopic expression led to the repression of N-cadherin expression and the activation of R-cadherin expression (Fig. 3P-R). Therefore, Ebf1 promotes downregulation of a progenitor-associated adhesion molecule and expression of a neurone-associated one. This may play a role in the early onset of the migration of the electroporated cells towards the mantle layer. In conclusion, these data indicate that ectopic expression of Ebf1 in neuroepithelial progenitors promotes neuronal differentiation, as indicated by modifications in the expression of several markers.
Ebf1-mediated neurogenesis involves activation of bHLH genes
As Ebf1 misexpression promotes neurogenesis in a similar way to proneural genes, we wondered whether this outcome might be reached through activation of the proneural genes. We therefore tested the effect of Ebf1 misexpression on Ngn1 and Ngn2, as well as on the achaete-scute-related Cash1 gene. Ten hours after electroporation, expression of both Ngn1 and Ngn2 was induced on the experimental side (Fig. 4A,B). By contrast, Cash1 expression was not affected (Fig. 4C). At later stages, consistent with the neuronal differentiation induced by Ebf1, we observed a general downregulation of the proneural genes. In the case of both Ngn1 and Cash1, this was already visible at 20 hours after electroporation, whereas the decrease in Ngn2 expression was only observed after 30 hours (Fig. 4D-F, and data not shown). These data indicate that Ebf1 can promote the expression of some, but not all, proneural genes, which suggests the existence of specific, positive regulatory loops linking these genes.
Activation of Ngn1 and Ngn2 provides a possible means for induction of neurogenesis by Ebf1. But is proneural, and more generally bHLH, gene function actually mediating aspects of Ebf1 neurogenic activity? To address this issue we used an Id inhibitor of bHLH proteins (Norton, 2000) to antagonise their activity. As expected, electroporation of a chicken Id2 (Martinsen and Bronner-Fraser, 1998; Dubreuil at al., 2002) expression vector prevented migration of neuroepithelial cells towards the mantle layer and neuronal differentiation (Fig. 3S, compare with 3I). Co-electroporation of the Ebf1 and Id2 expression vectors led to only a minority of the electroporated cells migrating into the mantle layer and expressing the neuronal marker Tuj1 (Fig. 3T). This is in sharp contrast with the electroporation of Ebf1 alone, which drives most of the cells into the mantle layer (Fig. 3H,J). Therefore, we can conclude that induction of neurogenesis by Ebf1 relies for a large part on bHLH (including proneural) protein activity. In addition, the fact that no cycling cells were observed among the co-electroporated cells located in the mantle layer suggests that Ebf1 activity is not sufficient to drive cells into the mantel layer if they have not left the cell cycle.
Ebf1 misexpression alters neuronal subtype
As Ebf1 is able to promote neuronal differentiation, we wondered whether it might also interfere with the specification of neuronal identity. To investigate this possibility, we analysed the effect of Ebf1 misexpression on neuronal subtype characteristics. At the stage of collection of the embryos (30 hours after electroporation at stage HH15), thoracic spinal cord neurones are normally distributed between Islet1/Islet2-positive populations, which correspond to motoneurones located in the ventral region, and Lim1- and/or CRABPI-positive populations, which correspond to interneurones, mostly located dorsally to the motoneurone pool (Tsuchida et al., 1994; Vaessen et al., 1990), with no co-expression of these two types of markers (Fig. 5A-E,G). Upon Ebf1 electroporation, we observed an extension of the CRABPI- and Lim1-positive domains into the motor column (Fig. 5A-E). Furthermore, these markers were also detected in cells still located within the ventricular zone; they are normally restricted to the mantle layer (Fig. 5A,B). Double-labelling experiments indicated that CRABPI- or Lim1-positive cells located at the level of the motor column were negative for Islet1 (Fig. 5C,E). Finally, double labelling for electroporated Ebf1 and for Islet1/Islet2 revealed that all electroporated cells were negative for Islet1/Islet2, even though they were located at the level of the motor column (Fig. 5G). Performing the electroporation experiments at different stages of embryo development did not affect this result (Fig. 5F,H). In all cases, the electroporation of Ebf1 resulted in a reduction of the pool of Islet1/Islet2-positive cells (Fig. 5F-I).
Hence, ectopic expression of Ebf1 appears to lead to premature activation of interneurone markers and, surprisingly, to repression of motoneurone markers. Double-labelling experiments suggest that progenitor cells normally fated to become motoneurones according to their dorsoventral (DV) localisation are reprogrammed towards an interneurone differentiation pathway. Therefore, under the conditions of these misexpression experiments, Ebf genes can modulate neuronal fate.
A dominant-negative Ebf protein prevents neuronal differentiation
The previous analyses have indicated that Ebf genes can interfere at various levels of the neuronal differentiation pathway. In order to identify their precise role during normal neurogenesis, we attempted to generate a dominant-negative molecule. We introduced a deletion in the Ebf1 expression vector, resulting in the elimination of the N-terminal 107 amino acids of the wild-type protein. Such a deletion retains the dimerisation domain but has been shown previously to obliterate Ebf1 DNA-binding activity (Hagman et al., 1993). Furthermore, a similar construct derived from the Xenopus Ebf2 gene was shown to have a dominant-negative activity (Dubois et al., 1998).
To investigate whether the deleted protein, termed ΔEbf, possessed a dominant-negative activity, we performed co-electroporation experiments with varying amounts of the wild-type Ebf1 expression construct and examined the effect on neuronal differentiation at the hindbrain level, as revealed by neurofilament staining (Fig. 6A-D). A concentration of 2 μg/μl of the ΔEbf expression plasmid was able to prevent precocious neuronal differentiation induced by wild-type Ebf1 at concentrations varying between 1 and 2μ g/μl (Fig. 6B,C, and data not shown). This suggests that ΔEbf antagonises wild-type Ebf1 and, therefore, can act as a dominant-negative molecule. Furthermore, at lower concentrations of the Ebf1 expression plasmid, or in its absence,Δ Ebf led to inhibition of neuronal differentiation (Fig. 5D, and data not shown). This observation suggests that Ebf family members are involved in normal neuronal differentiation, and that ΔEbf can also antagonise endogenous Ebf1 and, presumably, other members of the family owing to heterodimer formation. To better establish this point, we examined the effect of electroporation of ΔEbf on the expression of the interneurone and motoneurone markers, CRABPI and Islet1, respectively. In both cases, we observed a limited but significant decrease in expression on the experimental side (Fig. 6E,F).
As ΔEbf antagonises endogenous Ebf proteins, which themselves can lead to the activation of proneural genes, we wondered whether ΔEbf inhibition of neuronal differentiation might be mediated by repression of proneural or early differentiation genes. We therefore investigated the effect of ΔEbf electroporation on Ngn2 and NeuroM expression. No modification of the patterns of Ngn2 and NeuroM expression was observed (Fig. 6G,H), which suggests that the effect of ΔEbf on neuronal differentiation does not rely on alterations in the expression of genes involved in early neurogenesis.
In conclusion, our data suggest that Ebf activity is necessary for neuronal differentiation, but that it is not required for earlier stages of neurogenesis.
Antagonizing Ebf blocks neuronal differentiation and migration without interfering with cell cycle exit
As expression of ΔEbf prevents expression of neuronal markers, we then investigated whether it could also interfere with neuronal migration and/or cell cycle exit. For this purpose, we performed co-electroporation of the ΔEbf vector with a construct that expresses green fluorescent protein (GFP), in order to trace electroporated cells. To evaluate cell cycle exit, we carried out pulse-labelling with BrdU as described above. ΔEbf-electroporated cells were mostly found within the ventricular zone, which is in contrast to control cells electroporated with the mutant Krox20 construct, which were distributed between the ventricular zone and the mantle layer, and Ebf1-electroporated cells, which were largely confined to the mantle layer (Fig. 7A-D). These data suggest that expression of ΔEbf prevents migration towards the mantle layer. In addition, ΔEbf-electroporated cells did not express the neuronal marker Tuj1 (Fig. 7I), indicating that ΔEbf prevented neuronal differentiation, which is in agreement with the above data (Fig. 6). Analysis of BrdU-positive cells among electroporated cells indicated that they represented similar proportions after electroporation with ΔEbf or with the control construct, whereas, as expected, this proportion was dramatically reduced after electroporation with Ebf1 (Fig. 7A-D). This indicates that despite blocking neuronal migration and differentiation, ΔEbf did not affect the proportion of cells in S-phase, and presumably of proliferating cells. Accordingly, the proportion of electroporated BrdU-negative cells within the ventricular zone after electroporation with ΔEbf was much higher than in the control, suggesting that cells had exited the cell cycle but stayed within the ventricular zone (Fig. 7D).
As dominant-negative ΔEbf appears to prevent neuronal differentiation/migration without affecting cell cycle exit, we investigated what happens when forcing expression of a proneural gene able to drive both types of processes. For this purpose, we constructed an expression vector for a flagged version of chick Ngn2. As expected, cells electroporated with Ngn2 were found within the mantle layer, were negative for BrdU incorporation and expressed the neuronal marker Tuj1 (Fig. 7E,H,J), indicating that they had exited from the cell cycle and differentiated. Co-electroporation ofΔ Ebf with Ngn2 largely prevented the migration of electroporated cells to the mantle layer (Fig. 7F,H). In addition, electroporated cells staying in the ventricular zone were prevented from differentiating, as they did not expressed the neuronal marker Tuj1 (Fig. 7K). By contrast, co-electroporation of ΔEbf did not impede cell cycle exit, because none of the Ngn2-expressing cells incorporated BrdU (Fig. 7F). Finally, as expected, co-electroporation of Ebf1 with Ngn2 led to both cell cycle exit and migration to the mantle layer (Fig. 7G,H). In conclusion, these data indicate that whereas ΔEbf can largely impede migration towards the mantle layer and neuronal differentiation induced by Ngn2, it does not prevent cell cycle exit.
As expression of ΔEbf appears to block neuronal differentiation of cells engaged in the neurogenic pathway, we wondered about the effects of this contentious situation on cell fate and survival at later stages. For this purpose, we analysed cell apoptosis by TUNEL assay, 40 hours after co-electroporation of the neural tube with either theΔ Ebf expression construct or the mutant Krox20 expression plasmid as a control, together with a lacZ vector as a tracer. We found that in the control case some apoptosis is observed specifically in the electroporated side (Fig. 7L). This phenomenon might be related to non-specific effects of high levels of expression of the electroporated genes. Cells electroporated with ΔEbf were still largely located within the ventricular zone (89±3%). Electroporation ofΔ Ebf led to a significant increase of the number of TUNEL-positive cells (3.5±1-fold) in the electroporated side, as compared with the control case (Fig. 7L,M, note that at least half of the apoptotic cells in the experimental side express lacZ, although this is not visible in the figure because of the very low level of β-galactosidase, possibly due to apoptosis). These data suggest that the ΔEbf-mediated block in neuronal differentiation can lead to cell death in at least a subset of the affected cells.
In this paper, we have identified and characterized two novel chick genes that are the orthologs of mouse Ebf1 and Ebf3. The very high homology of the amino acid sequences, and the similarity of the expression patterns in post-mitotic neurones along the entire CNS, in mouse and chick indicate that Ebf family members have been highly conserved during higher vertebrate evolution, and are likely to have similar roles in different species. Using gain- and loss-of-function experiments in the chick embryo neural tube, we have carried out a detailed analysis of Ebf gene function during neurogenesis. This allowed precise positioning of these genes along the neurogenic cascade and raised the possibility that they may play multiple roles in this pathway in higher vertebrates. In particular, our data establish that Ebf genes are required for the coupling of neuronal differentiation and migration to the mantle layer with cell cycle exit.
The Ebf genes in the neurogenic cascade
Our conclusions on Ebf gene regulation and function are detailed below, and are based on the following observations.
(1) Analysis of Ebf1 and Ebf3 mRNAs in the chick neural tube indicated that their accumulation is coincidental with the onset of neurogenesis and that they are detected within the entire mantle layer (Fig. 1). This is in agreement with the expression pattern of the mouse orthologs (Garel et al., 1997; Wang et al., 1997), and shows that these genes are expressed at a high level in early post-mitotic neurones and that their expression is maintained during neuronal differentiation. Low level, scattered expression has also been observed in the neuroepithelium for mouse Ebf2 and Ebf3, presumably corresponding to cells en route to the mantle layer (Garel et al., 1997; Pattyn et al., 2000).
(2) Forced expression of both Ngn2, a proneural gene, and NeuroM, an early neuronal differentiation regulator, promoted Ebf1 and Ebf3 expression (Fig. 2), indicating that the latter genes are downstream of the former in the neurogenic cascade, consistent with Ebf gene expression pattern.
(3) Expression of a dominant-negative molecule, which presumably antagonizes all Ebf activities, did not affect cell cycle exit, but prevented neuroepithelial precursor migration towards the mantle layer and expression of differentiation markers (Figs 6, 7). Furthermore, the dominant-negative Ebf was also able to prevent neuronal differentiation and migration induced by the forced expression of Ngn2, but it did not affect the endogenous expression of this latter gene. Together, these observations suggest that Ebf genes play an essential role in cell engagement into neuronal differentiation and migration towards the mantle layer, coupling these processes to cell cycle exit.
(4) In agreement with a role of Ebf genes in the control of neuronal differentiation and migration, misexpression of Ebf1 in neuroepithelial progenitors promoted these processes (Fig. 3), which indicates that Ebf genes are both necessary and sufficient. However, surprisingly, forced expression of Ebf1 also led to exit from the cell cycle. This was correlated with a transient reinforcement of Ngn1 and Ngn2 (Fig. 5), and NeuroM (data not shown), expression. We could actually demonstrate that induction of the complete neurogenic programme by Ebf1 was largely dependent on bHLH proteins, presumably including proneural gene products, as shown by its inhibition by the bHLH antagonist Id2 (Fig. 3S,T). At this stage, we cannot exclude that forced high level expression of Ebf1 in neuroepithelial progenitors may lead to non-physiological proneural gene activation, subsequently resulting in the activation of the complete programme. An alternative explanation, involving a second function of Ebf genes, can nevertheless be envisaged and is discussed below.
(5) We have shown that Ebf1 misexpression also leads to changes in the balance of neuronal subtypes (Fig. 5). This suggests the existence of a third level of intervention of Ebf genes in the neurogenic cascade.
On the basis of these different observations, we propose to position the Ebf genes in the neurogenic cascade in the spinal cord and the hindbrain as indicated in Fig. 8. According to this model Ebf genes are downstream to proneural genes and cell cycle exit, but are absolutely required for neuronal differentiation and migration towards the mantle layer. The details of our conclusions are discussed below.
Uncoupling cell cycle exit from migration and differentiation
This work has revealed that expression of a dominant-negative Ebf gene prevents neuronal differentiation and migration towards the mantle layer without interfering with cell cycle exit. Previous studies performed in the Xenopus embryo indicated that blocking cell division by the injection of cell cycle inhibitors did not result in a significant activation of neuronal markers (Hardcastle and Papalopulu, 2000). Together these data indicate that cell cycle exit is not sufficient for induction of neuronal differentiation. In other words, the two types of events, although they are co-induced by proneural genes, are not intrinsically linked, and interfering with Ebf gene function allows them to be uncoupled.
However, uncoupling cell cycle exit from neuronal differentiation and migration may not be without consequences. Our analysis of cell death following expression of dominant-negative Ebf indicates that this treatment increases the proportion of apoptotic cells among electroporated cells (Fig. 7), which suggests that the cells might be able to sense this abnormal uncoupling and, consequently, enter into a cell death programme. A related interpretation is that the cells could simply sense the block in differentiation when they are already too far engaged in the commitment process. In several systems, a block in cell differentiation is accompanied by cell death, and this is also the case in the developing striatum of Ebf1 null mice (Garel et al., 1999).
As Ebf expression is required for both neuronal differentiation and migration towards the mantle layer, a subsequent question is whether Ebf genes independently contribute to each of these two manifestations, or whether differentiation is a consequence of migration or vice-versa. In favour of a direct role of Ebf genes in migration, we have observed that Ebf1 ectopic expression results in a modification of the pattern of expression of adhesion molecules, with repression of N-cadherin and induction of R-cadherin. Such modifications have been proposed to be involved either in the release of undifferentiated progenitors from the ventricular zone and/or in the promotion of their migration towards the mantle layer (Redies and Takeichi, 1996). Furthermore, additional evidence suggests a role of Ebf genes in the control of cellular adhesion, and of cellular and axonal migration (Prasad et al., 1998; Garel et al., 1999; Garel et al., 2000; Corradi et al., 2003). Nevertheless, our work suggests that Ebf genes also control aspects of neuronal differentiation independently from the migration towards the mantle layer. Neuronal markers like Lim1 and CRABPI, which are normally restricted to the mantle layer, are expressed in Ebf1-electroporated cells while they are still within the ventricular zone. In addition, in the striatum primordium of Ebf1 null-mutant embryos, cells in the mantle layer show an aberrant differentiation pattern, being unable to downregulate genes normally restricted to the SVZ, or to activate some mantle-specific genes (Garel et al., 1999). Together these data suggest that the function of Ebf genes is not likely to be restricted either to the promotion of migration towards the mantle layer or to the induction of neuronal differentiation, but rather that these genes directly contribute to both types of processes, and may actually coordinate them.
Additional, putative functions of Ebf genes in neuronal commitment and subtype specification
As indicated above, we do not know whether the transient activation of Ngn2 and NeuroM by Ebf1 (Fig. 4) is physiological, in particular because expression of the dominant-negative Ebf does not seem to affect the expression of these genes (Fig. 6). However the establishment of positive-feedback loops, involving proneural, Ebf and possibly other genes, might be involved in the stabilisation of the committed state among selected progenitors (Fig. 8). Another possibility is that Ebf genes might mediate a signal acting after cell cycle exit and ensuring that differentiating cells cannot resume proliferation, as has been proposed for NeuroD (Mutoh et al., 1998). If there is redundancy in the establishment or maintenance of these loops, the latter might not be revealed by the loss-of-function mediated by the dominant-negative Ebf. A similar feedback mechanism involving Xcoe2/Ebf2 has been proposed during primary neurogenesis in Xenopus (Dubois et al., 1998). This raises the possibility that different aspects of the functions of Ebf genes, and of their relationships with other genes involved in the control of neurogenesis, have been conserved during vertebrate evolution. By contrast, Xenopus Ebf3 is expressed later than Xcoe2, is not involved in maintaining XNgn1 expression and has been proposed to be required only in late neuronal differentiation, downstream of XNeuroD (Pozzoli et al., 2001). Although our experiments did not specifically address the precise level of action of each Ebf gene, the expression data (this work) (Garel et al., 1997; Malgaretti et al., 1997; Wang et al., 1997; Pattyn et al., 2000), together with the absence of a general neuronal phenotype associated with the Ebf2 null mutation (Corradi et al., 2003), are consistent with an early role of Ebf1 and Ebf3, coincidental and redundant with that of Ebf2. Therefore, Ebf2 and Ebf3 have functionally diverged in Xenopus, whereas they may have conserved similar early neurogenic function in higher vertebrates.
Combinatorial expression of homeobox genes in neural progenitors, established according to their DV location and in response, in particular, to sonic hedgehog signalling, has been shown to play an essential role in neuronal subtype specification in the spinal cord (Jessell, 2000). Proneural genes also have restricted patterns of expression along the DV axis and recently they have been implicated in the specification of neuronal subtype as well (Fode et al., 2000; Gowan et al., 2001). In this study, we have shown that Ebf1 misexpression in spinal cord progenitors leads to repression of motoneurone markers (Islet1/Islet2) and activation of interneurone markers (Lim1 and CRABPI), at the level of the motor column (Fig. 5). This suggests that progenitors normally fated to become motoneurones according to their DV location are reprogrammed towards an interneurone fate. This was unexpected as Ebf genes are normally expressed in differentiating motoneurones (Fig. 1) (Garel et al., 1997). We can provide two possible, and non-exclusive, explanations for our observations. Firstly, it is possible that high levels or inappropriate timing (leading to premature cell cycle exit in particular) of Ebf1 expression in neuroepithelial progenitors leads to modifications in the combinatorial expression of genes involved in early DV specification (e.g. homeobox genes, proneural genes) and, consequently, but indirectly, to fate changes. Indeed, we have shown that Ebf1 misexpression has different effects on the expression of three proneural genes, promoting the expression of Ngn1 and Ngn2, but not of Cash1 (Fig. 4). Secondly, because the Ebf genes appear as major regulators of neuronal differentiation, it is possible that they directly control the expression of neuronal subtype-specific genes. Indeed, we have shown that Ebf1 inactivation in the striatum primordium prevents the activation of CRABPI (Garel et al., 1999), and that forced expression of Ebf1 leads to ectopic activation of CRABPI not only in the motor column but also in the ventricular zone (Fig. 5). It is therefore possible that higher than normal levels of Ebf expression in differentiating neurones directly alters the balance between subtype specification genes. Further analyses will be required to precisely delineate the possible function of Ebf genes in this aspect of neuronal differentiation.
We are grateful to D. Anderson, M. Ballivet, C. Goridis, R. Grosschedl, T. Jessell, T. Reh, M. Takeichi and S.-Z. Wang for reagents. We also acknowledge the Developmental Studies Hybridoma Bank (University of Iowa) for monoclonal antibody supply. We thank J. Ghislain, C. Goridis and F. Guillemot for critical reading of the manuscript. This work was supported by grants from INSERM, MENRT, EC, ARC and AFM. M.G.-D. was supported by FEBS and EC Marie Curie fellowships.
- Accepted August 27, 2003.
- © 2003.