Molecular analysis of vertebrate nervous system development
Seminar organized by Nicole Le Douarin, Peter Gruss and David G. Wilkinson from 13 – 18 June, 1998.
Yves-Alain Barde (Max-Planck Institute, Martinsried), Gilbert Bernier (Max-Planck Institute, Gôttingen), Heinrich Betz (Max-Planck Institute, Frankfurt), Eduardo Boncinelli (Scientific Institute H. San Raffaele, Milano), Jean-Pierre Changeux (Institut Pasteur, Paris), Patrick Charnay (Ecole normale supérieure, Paris), Jean-Baptiste Charrier (Institut d’embryologie cellulaire et moléculaire, CNRS, Nogent-sur-Marne), Eddy de Robertis (University of California, Los Angeles), Elisabeth Dupin (Institut d’embryologie cellulaire et moléculaire, CNRS, Nogent-sur-Marne), Catherine Dulac (Howard Hughes Institute, Cambridge, USA), Peter Gruss (Max-Planck Institute, Gôttingen), François Guillemot (Université de Strasbourg), Chaya Kalcheim (Hebrew University of Jerusalem), Nicole Le Douarin (Institut d’embryologie cellulaire et moléculaire, CNRS et Collège de France, Nogent-sur-Marne), Salvador Martinez (University of Murcia, Spain), Isabel Palmeirim (Institut d’embryologie cellulaire et moléculaire, CNRS, Nogent-sur-Marne), Fabienne Pituello (Centre de biologie du développement, Toulouse), Martin Schwarz (Max-Planck Institute, Gôttingen), Marie-Aimée Teillet (Institut d’embryologie cellulaire et moléculaire, CNRS, Nogent-sur-Marne), Marion Wassef (Ecole normale supérieure, Paris), David Wilkinson (National Institute for Medical Research, London)
A great deal of progress has been made in understanding the molecular and cellular basis of nervous system development in vertebrates. A central theme that has emerged is the striking conservation of molecular mechanisms between vertebrate and invertebrate species that has enabled the identification of many key genes. In addition, many further genes are being identified through approaches such as the cloning of Xenopus genes expressed in neural-inducing tissue, or gene trap strategies in the mouse. Use of these as markers in cell transplantation experiments has given new insights into signalling interactions between tissues, while overexpression and gene knockout experiments have revealed roles in cascades of transcriptional regulation and cell communication and behaviour.
These themes were reflected in the meeting on the « Molecular analysis of vertebrate nervous system development » organized by Peter Gruss and Nicole Le Douarin, which was generously funded by the Fondation des Treilles and held at their beautiful estate near Tourtour in southern France. This meeting encompassed many aspects of this wide-ranging field, from the first steps of neural induction to the complexifies of neuronal organisation that underlie the functioning of the mature nervous system.
Eddy De Robertis (Los Angeles) discussed the finding that many molecules that can induce the neural plate act by inhibition. These inhibitors are chordin, noggin and follistatin which are expressed in the Spemanns organiser of Xenopus embryos, a group of dorsal cells that has neural-inducing and dorsalising activities, and that can induce a new body axis after transplantation. BMP-4 is expressed ventrally and has ventralising and anti-neuralising activity mediated by binding to BMP receptors. Chordin binds to BMP-4 and prevents the latter from binding to its receptors. In addition, since BMP-4 positively regulates its own expression, the block of this feedback loop by chordin in dorsal regions may underlie the ventral restriction of BMP-4 transcripts. In Drosophila, the homologues of these genes, dpp and sog, have a similar function along the dorsoventral axis, and the tolloid gene was identified as an indirect activator of dpp and found to encode a protease. The evolutionary conservation between systems was extended further by the identification of its Xenopus homologue, Xolloid. Xolloid has ventralising activity and was found to act by cleaving and thus inactivating chordin. A further example of the use of inhibitory mechanisms came from the identification of the cerberus gene in Xenopus. Cerberus is expressed in anterior endoderm that acts as a « head organiser » and, unlike signals from Spemanns (trunk) organiser, is able to induce the most anterior part of the nervous system. Cerberus protein is processed from a long to a short form, and whereas only the former binds to and blocks BMP4, both forms can bind to nodal-related 1, another member of the TGFb superfamily. Thus, induction of the neural plate involves the use of multiple inhibitors, as well as inactivators of some of these inhibitors. Since diffusion of BMPs is normally restricted by their rapid binding to receptor, it is also possible that some « inhibitors » enable BMPs to diffuse as a complex to more distant tissues where the binding partner is proteolysed and the BMP released.
Interactions between the neural plate, surface ectoderm and the notochord lead to the formation of neural crest dorsally, and distinct neuronal and glial cell types along the dorsoventral (D-V)axis. Sonic hedgehog (shh) is a ventralising signal secreted by the notochord that can induce the floor plate and shh expression in the adjacent neural epithelium. However, Marie-Aimée Teillet (Nogent) presented evidence that the progenitors of the notochord and floor plate already exist in Hensens node of the chick (the equivalent to Spemanns organiser of Xenopus). Fate mapping using chick-quail chimaeras reveals that Hensens node plus underlying endoderm (or the chordneural hinge that forms during later, secondary neurulation) will form the floor plate, notochord and dorsal part of the gut, and this correlates with the early expression of the HNF3b gene in these tissues. During the posterior regression of Hensens node, morphogenetic movements lead to the cleavage of the neural plate by the insertion of floor plate progenitors. These data suggest that interactions to form floor plate are already occurring within Hensens node, and indicate that some ablation experiments that had been thought to remove only the source of inducing signais had also removed presumptive floor plate cells.
Further aspects of D-V patterning were discussed by Fabienne Pituello (Toulouse) who showed that Pax–6 expression in the spinal cord is regulated by several mesodermal tissues. Pax–6 is expressed in medial regions along the D-V axis, and had been shown to be repressed dorsally by dorsalin/ BMP-4 and ventrally by shh from the notochord/floor plate. The onset of expression of Pax–6 correlates with the formation of somites in the adjacent paraxial mesoderm, raising the possibility of an inductive relationship. Indeed, somites were found to induce Pax–6 expression, whereas presomitic mesoderm lacks this activity, suggested that the timing of expression in the neural tube may be linked to the sequential segmentation of mesoderm from anterior-to-posterior. Recent progress in elucidating mechanisms underlying this segmentation was presented by Isabel Palmeirim (Nogent). A chick homologue of the Drosophila segmentation gene hairy, named Chairy1, is expressed in a dynamic pattern in presomitic mesoderm, with the expression domain moving in apparent waves that correlate with somite formation. These ‘waves’ are not due to cell movement or inductive interactions, but rather to a cell autonomous dock independent of protein synthesis that switches Chairy1 mRNA expression on and off. Each cell goes through 12 cycles of expression before forming a somite, and this raises the question as to how the cycles trigger somitogenesis. One possible mechanism is that as Chairy1 mRNA cycles, a progressive accumulation of Chairy1 protein occurs to a critical threshold. A further possibility is that segmentation involves a combination of Chairy1 and Chairy2, a related gene that has a more prolonged expression in the posterior mesoderm during each cycle.
Anterior-posterior patterning of the neural plate is already occuring during its induction, and progressively leads to its subdivision into regions that form distinct derivatives. Inductive interactions and genes expressed in spatially-restricted domains that underlie this regionalisation were discussed in a number of talks. Eduardo Boncinelli (Milan) presented evidence that nested domains of Emx and Otx gene expression in the anterior neural tube are involved in specifying forebrain territory. In the mouse, Otx–2 is expressed in anterior mesoendoderm and later in the overlying neural ectoderm, implicating it as a component of interactions that specify anterior tissues. Consistent with this is the anterior expression of an Otx gene during regeneration in planarians, suggesting that it has an evolutionarily ancient role. Indeed, widespread expression of Otx–2 in Xenopus embryos leads to an expansion of anterior tissue, and loss of trunk structures, probably due to a block in the process of posteriorisation that normally induces posterior tissues. Remarkably, this phenotype also occurs alter overexpression of a candidate transcriptional target gene for Otx–2, consistent with this gene being a major mediator of Otx–2 fonction. It is likely that Otx–2 also has later roles, since its expression domain persists during later stages of forebrain development where it correlates with boundaries along which axons migrate. Evidence was presented suggesting that Otx–2 may indeed regulate the expression of cell surface proteins involved in axon guidance.
Later roles of genes that regulate early patterning were also discussed by Marion Wassef (Paris) in the context of midbrain development. Analyses of null mutants have revealed that En–1/2 and Wnt–1 have important roles in the midbrain-hindbrain boundary (isthmus) that organises the formation of the cerebellum. These genes are also expressed later in complex patterns in the developing cerebellum, and the possible role of this was analysed in mice with weaker mutations or in which gene expression has been altered by replacing one gene with another. It was found that in these mice there are disruptions to the development of specific components of the cerebellum, indicating that the fonction and correct regulation of Wnt–1 and En genes are important for later patterning.
Studies of the midbrain by Salvador Martinez (Murcia) continued the theme of patterning and boundaries in discussing the role of FGF–8 in inducing midbrain/hindbrain structures. FGF–8 is expressed in the isthmus, and the implantation of beads carrying FGF–8 protein reveal that it can transform diencephalic tissue into midbrain/anterior hindbrain derivatives (tectum and cerebellum). However, the extent of this induction depends upon where the FGF–8 bead is placed relative to morphological subdivisions of the diencephalon, the prosomeres (p). When placed in p1 (the most anterior prosomere), p1 but not p2 is transformed to midbrain, whereas when placed in p2, Edouardo Boncinelli et Yves-Alain Barde François Guillemot et Catherine Dulac both p 1 and p2 are transformed. These data suggest that the boundary between these prosomeres acts as a directional barrier to induction. Similarly, the p2/p3 boundary (zona limitans) acts as a barrier to the ectopic induction of tectum/cerebellum, and furthermore when transplanted can induce adjacent tissue to form a zona limitans. Thus, the formation of boundaries that limit the action of inductive signals is likely to be an important mechanism for establishing distinct territories in the forebrain and midbrain.
Similar themes also emerged from talks by Martin Schwarz and Peter Gruss (Gôttingen) on the role of Pax genes in midbrain and eye development. Pax–2 and then Pax–5 are upregulated in the developing midbrain, and gene knockouts reveal that they have an overlapping function in specifying midbrain territory. In Pax–2 / Pax–5 double knockouts, Pax–6 expression (which normally occurs in the territories adjacent to the midbrain) spreads caudally into the presumptive midbrain, while in Pax–6 null mutants, there is an expansion of Pax–2 expression rostrally and expansion of the midbrain. Furthermore, this reciprocal relationship exists in the eye, where Pax–6 is expressed in the lens and retina, and Pax–2 in the adjacent optic stalk. In Pax–2 mutants, Pax–6 expression spreads into the presumptive optic stalk, suggesting that the expression of these genes in adjacent territories restricts the effect of inductive signals. In addition, dissection of Pax–6 regulatory elements reveals that a retinal enhancer element is directly repressed by Pax–2 in cooperation with a novel homeobox gene, Vax–1, that is expressed in the optic stalk. Taken together, these studies suggest that barriers to inductive signals at boundaries (perhaps established by the reciprocal expression of Pax genes), together with direct transcriptional repression, establish territories with distinct fates.
Moving further caudally, the hindbrain is another region in which subdivisions of the neural epithelium have a major role in patterning. The hindbrain is segmented to form rhombomeres that underlie the specification of nerves and of branchial neural crest cells that migrate into the branchial arches. An important aspect of this segmentation is that there is a restriction to the intermingling of cells between adjacent rhombomeres, such that each segment can maintain a distinct identity. David G. Wilkinson (London) discussed the roles of Eph receptor tyrosine kinases and their membrane-bound ephrin ligands in this process. Ephrin–B genes encode transmembrane pro-teins, and in vitro experiments suggest that in addition to activating Eph receptors, they may themselves transduce signals.
Three ephrin-B genes are expressed in rhombomeres r2/4/6, while three Eph receptors that they can activate (including EphA4) are expressed in r3/r5. Activation of Eph receptor or of ephrin-B protein each lead to cell sorting, possibly through regulating repulsion or de-adhesion responses similar to those triggered by Eph receptors during neuronal and neural crest cell pathfinding. These data suggest that a bi-directional signalling at rhombomere boundaries is involved in restricting cell intermingling. Studies of EphA4 gene regulation reveal that expression in r3/r5 is controlled by the Krox-20 zinc-finger transcription factor. As discussed by Patrick Charnay (see below), Krox20 is also required for the maintenance of r3/r5 and regulates the expression of Hox genes that confer A-P identity, and thus couples the restriction of cell movement to their identity.
In embryos with a null mutation in Krox-20, presumptive r3/r5 territory is initially present but then disappears, and Patrick Charnay (Paris) described progress in analysing whether this is due to cell death or a switch in the identity of r3/r5 cells. By using a cre-lox system, embryos have been generated in which cells present in presumptive r3/r5 are permanently marked by lacZ expression. Preliminary results show that there are fewer of these r3/r5-derived cells in Krox-20 null mutants, consistent with effects on cell proliferation or death. Krox-20 may also have other roles, as suggested by genetic interactions with the Hoxa-1 gene. In Hoxa-1 null mutants, patches of cells not expressing Krox-20 are present within r3, and this abnormality is increased in combination with a heterozygous Krox-20 mutation. Since Hoxa-1 is expressed transiently in pre-r4 and more caudal regions, these non-autonomous effects may be due to disruption of signalling between adjacent rhombomeres. Indeed, evidence for signais that regulate segmentai identity has also been obtained by transplantation experiments. Another approach to understanding the role of Krox-20 in segmentation is to identify its targets. In addition to regulating Hox genes and EphA4, Krox-20 is upstream of two related genes, nab-1 and nab-2, that are expressed in r3 and r5. The nab proteins had been found to bind the related Egr1 (Krox-24) protein and block its ability to activate target genes. These proteins also block the transactivation domain of Krox-20 and thus are involved in a negative feedback loop initiated by Krox-20.
Following the regional specification of territories in the nervous system, cell differentiation generates a diversity of neuronal and glial cell types. François Guillemot (Strasbourg) discussed studies of the cascade of bHLH transcription factors that progressively restricts the fate of neural progenitors to specific cell types. Mash-1 is expressed early during development of many regions of the CNS and PNS, and analysis of null mutants reveals that fate specification is arrested in multiple tissues. Unexpectedly, this arrest occurs at different steps in diffèrent tissues; in some cases during terminal differentiation of a neuronal precursor, and in other cases prior to selection of neuronal precursors from a stem cell. Analysis of the ventral telencephalon in which there is a loss of the medial part revealed less cell proliferation, white neuronal precursors are still present. In view of the role of the homologous genes in Drosophila, it was possible that Mash-1 regulates Delta expression. Indeed, it was found that Mash-1 regulates Delta-1, which in turn is involved in a negative feedback loop via Notch activation that blocks Mash-1 expression in adjacent cells, maintaining them in an undifferentiated, proliferative state.
The neural crest migrates from the dorsal neural tube to contribute to many tissues, including the peripheral nervous system, melanocytes and, in the head, skeletal tissues. As for the CNS, neural crest development involves A-P specification, restrictions to cell movement, and regulation of cell type differentiation, and several talks discussed these issues. Nicole Le Douarin (Nogent-sur-Marne) reviewed the findings from fate mapping with chick-quail chimaeras that reveal the contributions of neural crest and paraxial mesoderm to bones in the head. Neural crest has the major contribution, with paraxial mesoderm only forming a few bones. Strikingly, a bone that is derived in part from neural crest and in part from mesoderm forms adjacent to the tip of the notochord, at the interface between the chordal and prechordal regions. This finding fits with the idea that the contribution of neural crest to skeletal components evolved together with the formation of a prechordal region in vertebrates. In the branchial arches, neural crest cells differentiate to form a distinct set of skeletal derivatives in each arch. These neural crest cells migrate in streams from specific rhombomeres that correlate with domains of Hox gene expression. Consistent with evidence from analysis of a Hox null mutant that Hox genes control the A-P identity of neural crest, there is a tight correlation between Hox gene expression and the pattern of skeletal structures that form after a variety of experiments in which specific populations of branchial neural crest are transplanted or ablated.
A key aspect of neural crest patterning is the migration of cells along specific pathways. In the trunk, neural crest cells migrate through the anterior but not posterior part of the schlerotomal part of somites, and this underlies the formation of the segmented series of dorsal root ganglia. Chaya Kalcheirn (Jerusalem) presented evidence that the extracellular matrix molecule, F-spondin, has a role in this segmentai migration. F-spondin is expressed in the posterior half of each somite, as well as in dermamyotome which is inhibitory to neural crest cell migration. In explant essays it was found that addition of F-spondin inhibits neural crest migration through anterior somite cells, and that anti-F-spondin antibody allows increased migration through anterior somite cells. Since Eph receptors and ephrins have also been found to restrict neural crest cell migration through the somites, these data suggest that multiple cues are involved in guidance.
Elisabeth Dupin (Nogent-sur-Marne) presented evidence that endothelins and their receptors have important roles in the differentiation of neural crest cells to enteric neurons in the posterior gut, and to melanocytes. These tissues are absent in mice with a null mutation in endothelin-1, raising the question as to which aspects of neural crest cell development is disrupted. To test this, the effects of endothelin-3 (which is expressed in epidermis and gut mesenchyme) on clonal explant cultures of chick neural crest (which express endothelin receptor-B2) were studied. Treatment with endothelin-3 increased the number and size of clones that differentiate into melanocytes, glial cells, or melanocytes plus glial cells. In addition, endothelin-3 stimulated the proliferation of differentiated pigment cells. These data indicate that endothelin-3 regulates the development of melanocyte and glial cell precursors and their common progenitor, as well as differentiated derivatives.
Yves-Alain Barde (Martinsried) pointed out that in contrast to the conservation of many gene families in diverse organisms, the neurotrophins provide an example of an evolutionary innovation, since true homologues are not apparent in Drosophila. In vertebrates, these ligands regulate neuronal cell survival by activating trk receptor tyrosine kinases to generate the correct number of neurons and their connections. However, in contrast to this trophic effect of NGF and other neurotrophins, in tissues such as early retinal ganglion cells, NGF triggers cell death. This distinct response is mediated through the p75 receptor which is expressed in the early retina, whereas trk receptors are not. Cell death in the retina is not random, but occurs along the forming pathways of axons, perhaps forming channels. Intriguingly, NGF is expressed by microglial cells delivered through the bloodstream, and cell death is absent when the retina is cultured in isolation, suggesting that recruitment of microglial cells to specific sites might generate localised cell death.
The ultimate outcome of neural development is the formation of a complex and functional network of connections, and several talks discussed issues ranging from large-scale organisation of these networks to the subcellular organisation of synapses. Catherine Dulac (Cambridge, USA) described studies of the vomeronasal organ (VNO) and its projections which are involved in the perception of pheromones. The use of single cell cDNA libraries for differential screening enabled the identification of 7-pass transmembrane receptors for pheromones that are distinct in structure and site of expression from the previously-identified odorant receptors. There are around 150 distinct pheromone receptors that fall into two classes, and individual cells within the VNO appear to express a single receptor. This raises the question as to how information about pheromones is encoded through the topography of cells expressing distinct receptors in the VNO and their projections to their targets, the glomerulae. It is found that the two subfamilies of pheromone receptors are expressed in distinct layers of the VMO, suggestive of functional units, and that some receptors have sexually dimorphic patterns of expression. Unlike the situation for odorant perception, cens expressing a specific receptor project to multiple glomerulae. This raises the question of whether the pattern of connections of the glomerulae to their central targets leads to a subsequent convergence of information in the brain.
Two talks described studies of how synapses are assembled at specific sites in neurons and their targets. Jean-Pierre Changeux (Paris) discussed the localisation of acetylcholine receptor in muscle cells to the synapse at the neuromuscular junction. After myotube formation, the a1 receptor protein initially has widespread distribution, but then the protein becomes localised through interactions with a complex of pro-teins that form at the neuromuscular junction. Concurrent with this, expression of receptor mRNA becomes localised in the multinuclear muscle cell to the nuclei adjacent to the synapse. This restriction involves electrical activity that through calcium influx suppresses expression, while local signals (not involving acetycholine receptor activation) at the junction up-regulate expression. Insights into the molecular components of this spatial regulation are emerging from mapping of regulatory elements of the a1 receptor gene that are involved in this positive and negative control, and identification of the interacting transcription factors. Heinrich Betz (Frankfurt) described studies of the localisation of inhibitory glycine receptors to synapses in dendrites. This localisation is triggered by activation of the receptor, and a key component is the gephrin protein that is part of a complex linking receptor to the cytoskeleton. Yeast two-hybrid screens have identified proteins that bind and co-localise with gephrin, including components that might mediate localised signal transduction. Further dissection of this system will enable understanding of how signal transduction regulates the localisation of receptor, and also how different receptors (e.g. GABA and glycine receptors) sort differentially.
Since it is clear that for all of these, and other, aspects of neural development, only a fraction of the relevant components have been identified, a major challenge is to identify further genes that have a potential involvement. Peter Gruss summarised progress in a gene trap screen in the mouse, in which the insertion of lacZ into genes visualises their expression pattern, thus suggesting potential sites of their function. Furthermore, the insertion event in many cases will disrupt the function of the gene, which can then be revealed by breeding to homozygosity. More than half of the approx. 200 gene traps thus far characterised are in novel genes, and a number of those that have been bred to homozygocity have given striking developmental phenotypes.
Overall, a message that emerged from this meeting is that much progress has been made in understanding many important aspects of neural development, but there is a long way to go. In particular, a major challenge for many of the systems under study is, in addition to identifying the relevant components, to elucidate how they are linked in pathways of cell communication, signal transduction and gene transcription that underlie the establishment of patterns of cells or subcellular organisation.