Abstract
Distinct classes of neural cells differentiate at specific locations within the embryonic vertebrate nervous system. To define the cellular mechanisms that control the identity and pattern of neural cells we have used a combination of functional assays and antigenic markers to examine the differentiation of cells in the developing spinal cord and hindbrain in vivo and in vitro. Our results suggest that a critical step in the dorsoventral patterning of the embryonic CNS is the differentiation of a specialized group of midline neural cells, termed the floor plate, in response to local inductive signals from the underlying notochord. The floor plate and notochord appear to control the pattern of cell types that appear along the dorsoventral axis of the neural tube. The fate of neuroepithelial cells in the ventral neural tube may be defined by cell position with respect to the ventral midline and controlled by polarizing signals that originate from the floor plate and notochord.
Introduction
A basic problem in vertebrate development is to define the mechanisms that control the identity and patterning of cells in the nervous system. The initial event in neural cell patterning is the induction and regionalization of the neural plate. The establishment of polarity along the anteroposterior axis of the neural plate is evident soon after neural induction; however the patterning of cells along the dorsoventral axis is delayed, occurring as the neural plate folds to form the neural tube (Roach, 1945; Jacobson, 1964) (Fig. 1). The position of neuroepithelial cells along these two primary axes is likely to contribute significantly to their fate. In anterior regions of the neural plate, small differences in cell position along the anteroposterior axis result in markedly different cell fates (Puelles et al. 1987). However, in more posterior regions such as the spinal cord, the overall pattern of cell types is essentially identical at different segmental levels (Hollyday and Hamburger, 1977; Altman and Bayer, 1984; Eisen, 1991). Thus, a major determinant of neural cell identity in the spinal cord is likely to be cell position along the dorsoventral axis of the neural tube.
Each of the neuronal cell groups that differentiates along the dorsoventral axis of the nervous system exhibits bilateral symmetry (Fig. 1). The dorsoventral polarity of the neural tube and the bilateral symmetry of the spinal cord could be established at earlier stages by cells at the midline of the neural plate which provide signals that spread laterally to control the fate of adjacent neuroepithelial cells. The cells located at the midline are one of the first cell groups to undergo differentiation within the neural plate (Bancroft and Bellairs, 1975; Schoenwolf, 1982; Altman and Bayer, 1984). During neurulation these midline cells give rise to the floor plate (Baker, 1927; Kingsbury, 1930), an epithelial cell group that later occupies the ventral midline of the neural tube (Fig. 1) and has been implicated in the patterning of axonal projections in the developing central nervous system (Jessell et al. 1989). The floor plate releases a diffusible chemoattractant that may orient the growth of the axons of commissural neurons towards the ventral midline of the embryonic spinal cord (Tessier-Lavigne et al. 1988; Placzek et al. 1990a, 19906). The floor plate also appears to act as an intermediate target involved in changes in axonal trajectory at the midline (Bovolenta and Dodd, 1990, 1991; Kuwada et al. 1990a, 19906; Kuwada and Hatta, 1990; Hatta et al. 1990; Yaginuma et al. 1991). In addition, floor plate cells are the source of a polarizing signal that can respecify the pattern of cell differentiation along the anteroposterior axis of the developing chick limb (Wagner et al. 1990), mimicking the effect of the limb polarizing region and of retinoic acid (Tickle et al. 1975, 1982).
The characterization of these floor plate properties has led us to investigate the cellular interactions that control floor plate differentiation and to determine whether the floor plate has a role in regulating the identity of other neural cells. Here we discuss experiments that provide evidence firstly, that local inductive signals from the notochord control the differentiated properties of floor plate cells and secondly, that the floor plate and notochord together regulate the differentiation of neural cell types along the dorsoventral axis of the neural tube.
Results
Induction of the floor plate
The first sign of floor plate differentiation is a change in the morphology of cells at the midline of the neural plate (Schoenwolf and Smith, 1990). Floor plate cells exhibit basally located nuclei and apical constrictions, leading to the characteristic wedge-shaped appearance of the floor plate in the neural tube (Schroeder, 1970; Schoenwolf and Franks, 1984). Early studies on the control of floor plate differentiation examined the morphology of the spinal cord in chick embryos with duplicated notochords (Watterson et al. 1955). In these embryos, spinal cord cells with characteristic floor plate morphology were observed adjacent to both notochords, suggesting that the differentiation of the floor plate depends on the notochord. More recent experimental studies have provided additional support for this idea. Grafting of a notochord next to the lateral part of the neural tube results in the appearance of cells with floor plate-like morphology in the adjacent neural epithelium (van Straaten et al. 1985a,b, 1988; Smith and Schoenwolf, 1989). In addition, in embryos in which the notochord has been deleted, either experimentally or as a result of mutations, wedge-shaped cells are absent from the midline of the spinal cord (Wolff, 1936; Grabowski, 1956; Youn and Malacinski, 1981; van Straaten and Drukker, 1987; Kuwada and Hatta, 1990; Hatta et al. 1990; Bovolenta and Dodd, 1991).
These results are consistent with the idea that the notochord can influence the morphology of the neural tube, generating a region of cells with the appearance of a floor plate. However, comparable changes in neuroepithelial cell shape occur during the development of many regions of the central nervous system (Lewis, 1947; Martins-Green, 1988; Schoenwolf et al. 1988; Moury and Jacobson, 1989), making it difficult to identify the floor plate solely by morphology. The characterization of floor plate-specific antigens (Table 1) and functions (Dodd and Jessell, 1988; Tessier-Lavigne et al. 1988; Jessell et al. 1989) has permitted a more critical test of the dependence of floor plate differentiation on signals that derive from the notochord. To examine the origins of the floor plate, we first determined whether surface antigens expressed by the floor plate are induced in other regions of the neural tube in response to a segment of notochord grafted adjacent to the neural tube of stage 9-11 (Hamburger and Hamilton, 1951) chick embryos. The floor plate antigens, FP1, FP2 and SCI were induced in neuroepithelial cells adjacent to notochord grafts placed dorsally or laterally to the neural tube (Fig. 2 and Yamada et al. 1991). These results show that the expression of specific floor plate antigens can be induced by the notochord.
Previous studies have shown that the embryonic rat floor plate releases a diffusible chemoattractant which promotes the outgrowth and orientation of spinal commissural axons in vitro (Tessier-Lavigne et al. 1988; Placzek et al. 1990a, 1990ft). The chick floor plate also releases this chemoattractant (Placzek et al. 1990c). In both rat and chick embryos, the chemoattractant is restricted to the floor plate and thus provides a functional marker with which to monitor the differentiation of the floor plate. To assay for induction of the chemoattractant in vivo, a segment of notochord was grafted adjacent to the neural tube of stage 9-11 chick embryos. Explants of spinal cord from the region adjacent to the grafted notochord consistently evoked axon outgrowth from rat dorsal spinal cord explants (Placzek et al. 1990c). In contrast, neither the neuroepithelium from the contralateral, unoperated side of the spinal cord nor the grafted notochord itself exhibited chemotropic activity (Placzek et al. 1990c). Thus, the notochord can induce expression of this chemoattractant in regions of the chick neural tube that normally lack this activity.
There appears to be a marked rostrocaudal gradient in the onset of floor plate differentiation within the neural tube. Midline neural plate cells isolated from the anterior region of stage 6 chick embryos are already capable of expressing differentiated floor plate characteristics as assessed by production of the chemoattractant (Fig. 3C, D and Placzek et al. 1990c). In contrast, floor plate differentiation in more caudal regions is delayed, and does not occur until about stage 12 (Placzek et al. 1990c). Removal of the notochord before this stage should therefore reveal whether the notochord is required for the differentiation of the floor plate during normal development. To test this, a piece of notochord from the caudal region of stage 10–12 embryos was removed and the expression of antigenic and functional properties of the floor plate examined at stage 19-23. At segmental levels lacking the notochord, cells at the ventral midline of the spinal cord did not express either the SCI or FP1 antigens (Yamada et al. 1991). Similarly, the chemotropic activity of regions of the ventral midline of the neural tube isolated from the notochordless region and from regions just rostral and caudal was assayed (Placzek et al. 1990c). The ventral midline, at levels lacking a notochord, did not evoke axon outgrowth, whereas the ventral midline rostral and caudal to the notochord region evoked extensive outgrowth. These results provide evidence that the notochord is required for the differentiation of the floor plate during normal development.
These in vivo studies do not exclude the possibility that other tissues, for example somites, act in combination with the notochord to induce differentiation of the floor plate. To test whether the notochord alone is sufficient for induction, stage 7 chick notochord was placed in vitro together with lateral regions of stages 6 and 7 neural plate (Fig. 3A,B). Explants of the lateral region of the neural plate alone did not exhibit chemotropic activity (Fig. 3C,D). However, significant chemotropic activity was observed when explants of lateral neural plate and notochord were placed in direct contact (Figs 3C,D and 4A) (Placzek et al. 1990c). In contrast, chemotropic activity was not detected when the notochord and lateral neural plate were placed 50 – 100 μ m apart (Figs 3C,D and 4B). This argues against the possibility that the uninduced lateral neural plate and notochord release different factors that act in combination to evoke commissural axon outgrowth and suggests that the notochord is sufficient to induce a floor plate.
These results also suggest that proximity or direct contact between notochord cells and the neural plate is necessary for induction of the floor plate. In support of this, during the early development of the neural tube only those cells contacted by the notochord exhibit the shape changes characteristic of early floor plate differentiation (Schoenwolf and Smith, 1990). Because there is little cell death within the chick neural tube before stage 15 (Homma et al. 1990) the ectopic expression of floor plate-specific properties in neural tube cells is likely to result from induction rather than from the rescue of committed precursor cells that normally die in the absence of the notochord. The induction of chemotropic activity in isolated explants of lateral neural plate also makes it unlikely that in vivo the ectopic notochord simply directs the dorsolateral migration of committed floor plate progenitors originally located at the ventral midline.
More generally, these results suggest that mesodermally derived inductive signals have sequential and progressively more refined roles in regulating neural differentiation. During gastrulation, these signals appear to control the initial induction of the neural plate and its early regionalization along the anteroposterior axis (Mangold, 1933; Spemann, 1938; Dixon and Kintner, 1989; Ruiz i Altaba, 1990). Later, during neural tube development, the mesoderm also determines the identity of a specific subset of cells within the neural plate, inducing them to differentiate into floor plate cells. As discussed below, induction of the floor plate by the notochord appears to be an important step in subsequent aspects of cell patterning in the developing nervous system.
Signals from the floor plate and notochord control cell pattern along the dorsoventral axis of the neural tube
The differentiation of floor plate cells precedes that of other cell types within the neural tube (Fig. 1 and Schoenwolf and Smith, 1990). In the spinal cord, floor plate differentiation is followed by the appearance of motor neurons, commissural neurons and neural crest cells at different dorsoventral positions (Fig. 1). The characterization of the floor plate as a source of polarizing activity in the developing limb (Wagner et al. 1990), together with its midline location within the neural tube, raises the possibility that the floor plate controls the pattern of cell differentiation along the dorsoventral axis of the neural tube (Jessell et al. 1989).
To determine whether the fate of neuroepithelial precursors within the chick neural tube depends on position with respect to the floor plate, we first identified antigens that are selectively expressed by distinct subsets of neural cells distributed at different positions along the dorsoventral axis of the spinal cord and hindbrain (Table 1, and Yamada et al. 1991). The expression of these antigens was monitored to determine how the pattern of cells in the embryonic chick nervous system changes in the presence of an additional floor plate or notochord at ectopic positions in the neural tube, and after elimination of the floor plate and notochord.
Changes in the pattern of neural cell differentiation after notochord grafts
Induction of motor neurons
To determine whether the floor plate controls the differentiation of other neural cells, the location of motor neurons was examined in embryos in which an additional floor plate had been induced at the dorsal midline by a notochord graft. Motor neurons normally differentiate in the ventral spinal cord near, but not immediately adjacent to the floor plate (Fig. 5A). In embryos with dorsal notochord grafts, labelling with MAb SCI revealed the presence of two columns of motor neurons in a normal ventral position. In addition, the dorsal half of the spinal cord contained two ectopic motor columns that differentiated in a bilaterally symmetric manner around the newly induced floor plate (Fig. 5B) (Yamada et al. 1991). When notochord grafts were placed next to the neural tube, midway between the roof plate and the floor plate, an ectopic floor plate was detected adjacent to the graft. In addition, an ectopic column of motor neurons developed in the dorsal spinal cord on the side of the graft, separated from the induced floor plate by a region of unlabelled cells. There was about a 1.5-fold increase in the number of motor neurons on the side of the spinal cord adjacent to the graft but no significant change in the total number of neural cells (Yamada et al. 1991).
Induction of cells adjacent to the floor plate
The expression of the SCI antigen by floor plate cells and motor neurons revealed a region of unlabelled neuroepithelium that was invariably interposed between the floor plate and motor columns in the spinal cord of normal and operated embryos (Yamada et al. 1991). In the chick spinal cord, primitive longitudinal (PL) interneurons differentiate in this region (Yaginuma et al. 1990) (termed region X) although antigenic markers that identify, selectively, PL cells or other cells in region X in the spinal cord have not been defined. However, in the hindbrain, a subset of cells in region X differentiate into serotonergic neurons (Wallace, 1985). We therefore examined whether the differentiation of cells in region X is dependent on the floor plate and notochord by monitoring the expression of serotonergic neurons after grafting a piece of notochord adjacent to the neural tube at the level of the hindbrain. Notochord grafts at the level of the hindbrain induced an ectopic floor plate, which could be detected with MAbs SCI and FPL Serotonergic neurons were present immediately adjacent to the ectopic floor plate whereas SC1-labelled cranial motor neurons were induced at a distance from the ectopic floor plate (Yamada et al. 1991). These findings show that cells characteristic of region X can be induced by midline-derived signals and occupy a position between ectopic floor plate cells and motor neurons.
Decrease in dorsal cell types
We next examined whether the appearance of ventral cell types in dorsal regions after notochord grafts is accompanied by changes in the pattern of differentiation of dorsal cell types. The AC4 antigen is expressed by essentially all neuroepithelial cells at the neural plate stage. During neurulation, expression of the AC4 antigen disappears from neuroepithelial cells in a ventral to dorsal progression and by stage 17 the antigen is restricted to the intermediate and dorsal regions of the spinal cord with the exception of a small region near to the roof plate (Yamada et al. 1991). After this stage expression remains constant and is restricted to the dorsal half of the spinal cord. The progressive dorsal restriction of the AC4 antigen raises the possibility that its expression is repressed by signals that derive from the ventral midline.
To test this possibility, notochord grafts were placed next to the lateral part of the neural tube. A marked decrease in expression of the AC4 antigen was observed in the dorsal spinal cord adjacent to the graft (Yamada et al. 1991). Changes in the distribution of identified neurons in the dorsal and intermediate region of the spinal cord were assessed by monitoring the expression of cellular retinoic acid-binding protein (CRABP), a marker of spinal relay neurons (Maden et al. 1989; Kitamoto et al. 1989). The number of CRABP-labelled neurons decreased after lateral notochord grafts (not shown). Thus, the appearance of ventral cell types in dorsal regions after notochord grafts is accompanied by a change in the pattern of differentiation of cell types characteristic of the dorsal spinal cord. As discussed later, the signals that control the differentiation of cells in the dorsal neural tube remain unclear.
Changes in neural cell pattern mediated by the floor plate
The change in neural cell pattern observed after notochord grafts raises the possibility that a cascade of local inductive signals may control the pattern of cell differentiation in the developing nervous system. Induction of the floor plate by the notochord would be the first step in this process. Signals from the floor plate may then control the differentiation of other cells along the dorsoventral axis of the neural tube. One prediction of this model is that the floor plate should be able to influence directly the pattern of neural cell differentiation. To test this, segments of floor plate were grafted adjacent to the neural tube of host embryos.
Floor plate grafts were found to produce changes in the pattern of cell differentiation along the dorsoventral axis of the spinal cord that were indistinguishable from the changes evoked by notochord grafts (Yamada et al. 1991). When a floor plate was grafted above the dorsal midline of a host neural tube, labelling with the MAb SCI revealed the presence of two additional motor columns that differentiated in a bilaterally symmetric position with respect to the grafted floor plate (Fig. 5C). Similarly, lateral floor plate grafts induced an adjacent ectopic column of motor neurons. In addition, floor plate grafts were able to suppress the expression of markers such as the AC4 antigen that are characteristic of dorsal cell types.
To determine whether the change in pattern of cell differentiation was dependent on signals from the floor plate itself and not from flanking neural tissue, the region of neuroepithelium immediately adjacent to the floor plate and, in addition, the ventrolateral and dorsal regions of the neural tube were tested for their ability to induce motor neuron differentiation. No changes in the pattern of motor neuron differentiation were observed with such grafts (Yamada et al. 1991). Thus, the pattern of cell differentiation in the neural tube appears to be modified by signals that originate from the floor plate.
Homeogenetic induction by the floor plate
Fate mapping studies in avian embryos have shown that the floor plate and notochord derive from overlapping regions of the epiblast (Rosenquist, 1966,1983; Nicolet, 1970; Hara, 1978; Schoen wolf et al. 1989; Selleck and Stern, 1991). Moreover, these two cell groups exhibit many common biochemical properties (Jessell et al. 1989) and signalling functions (Wagner et al. 1990). These observations raise the possibility that the floor plate mimics the notochord in its ability to induce floor plate differentiation. To test this, a segment of chick floor plate was grafted adjacent to the neural tube, and floor plate differentiation within the neural tube assayed by expression of antigenic markers. Grafts of quail and rat floor plate in combination with antibodies specific for chick floor plate antigens excluded the possibility that grafted floor plate cells invaded the host neural tube. Both the FP1 and SCI antigens were induced in adjacent regions of the spinal cord after floor plate grafts (Fig. 6A) (Yamada et al. 1991). These results show that the floor plate can mimic the notochord in inducing floor plate differentiation in adjacent regions of the neuroepithelium, providing evidence that the floor plate has homeogenetic inductive properties.
The notochord may induce motor neurons independently of the floor plate
The results described in the preceding sections suggest that a sequential cascade of inductive interactions initiated by the notochord and perpetuated by the floor plate can operate to control the dorsoventral pattern of cell differentiation in the neural tube. However, it also possible that signals from the notochord c regulate motor neuron differentiation directly without inducing an intermediary floor plate.
To test this possibility a stage 10 notochord w grafted next to a stage 15-18 neural tube, which is in longer competent to respond by forming a recognization floor plate (van Straaten et al. 19856; Placzek et < 1990c). In contrast to the results of notochord grafts younger hosts, expression of the FP1 antigen was n detected. Despite this, expression of the SCI and 3A antigens revealed an increase in the size of the mot column adjacent to the notochord graft (Yamada et < 1991). Further evidence for a direct effect of t notochord on motor neuron differentiation is provide by dorsal notochord grafts into younger embryos which, in some cases, were able to induce motor neurons in the absence of expression of floor plate antigens (Yamada et al. 1991). These results provide some evidence that the notochord can induce motor neurons without prior induction of a floor plate. It remains possible, however, that in both experiments, notochord grafts induced partial floor plate properties which include the ability to induce motor neurons.
Neural cell identity is defined by position with respect to the floor plate
One feature to emerge from these results is the consistency of the position at which motor neurons appear in relation to the floor plate. This is apparent in reconstructions of embryos in which notochord grafts were placed diagonally with respect to the dorsoventral axis and thus occupied a series of different dorsoventral positions adjacent to the spinal cord (Fig. 7A). In dorsal regions, the notochord graft induced an additional floor plate in the adjacent neuroectoderm and an ectopic dorsal motor neuron column at a distance from the induced floor plate (Fig. 7B). As the graft approached more ventral positions the distance between the host and induced floor plates decreased and there was a progressive decrease in the size of the intervening ventral column of motor neurons (Fig. 7B). The decrease in the area occupied by the ventral motor neuron column was greater, proportionally, than the decrease in area of the region of neuroepithelium between the two floor plates (Fig. 8). Thus, at distances between the two floor plates of less than 80 μ m, motor neurons were absent from the intervening region. When the grafted notochord was located immediately adjacent to the host notochord the ventral midline had a single expanded floor plate (Fig. 2D). The simplest interpretation of these findings is that motor neuron differentiation is determined primarily by position with respect to the floor plate.
Changes in neural cell pattern in the absence of the floor plate and notochord
To examine whether the notochord and floor plate are required for regulating the pattern of cell differentiation in the neural tube during normal development, a segment of notochord underlying the caudal ventral midline of stage 10 – 12 embryos was removed, thus preventing the development of the floor plate at these segmental levels (Placzek et al. 1990c; Yamada et al. 1991). The floor plate extended for about 60 μ m into the region of neural tube from which the underlying notochord had been removed. Further into the notochord-free region, the expression of floor plate antigens disappeared although distinct lateral motor columns were still present for a further 60 pm. Beyond this point, the bilateral motor columns merged to form a single expanded column at the ventral midline of the neural tube. Thus an arc of motor neurons differentiates around the end of the floor plate. At distances greater than 100 um from the end of the floor plate, motor neurons were absent. These results support the idea that motor neurons differentiate at a constant position with respect to the floor plate.
Analysis of the cell types remaining in the spinal cord after elimination of the notochord and floor plate revealed a marked difference in the dependency of distinct neuronal classes on ventral midline-derived signals. Although there was a complete absence of motor neurons (Fig. 9A,B), other cell types normally found in the dorsal spinal cord were present over an expanded dorsoventral domain. First, after removal of the notochord, the AC4 antigen which is normally absent from ventral regions, was expressed along the entire dorsoventral extent of the spinal cord, except for the small region near the roof plate. Second, sensory relay neurons that normally differentiate in the dorsal and intermediate regions of the spinal cord and express a cellular retinoic acid binding protein (CRABP) (Maden et al. 1989; Kitamoto et al. 1989) (Fig. 9C) were located at all dorsoventral positions including the ventral midline after notochord removal (Fig. 9D).
Moreover, in normal embryos, sensory relay axons remain within the spinal cord with commissural axons projecting to and across the floor plate (Fig. 9C). However, after notochord removal, the axons of CRABP-labelled sensory relay neurons projected out of the spinal cord (Fig. 9D). These results show that in the absence of a floor plate and motor neurons, sensory relay neurons in the ventral spinal cord exhibit an abnormal axonal projection pattern reminiscent of that of motor neurons in normal embryos. The outgrowth of commissural axons into the mesenchyme surrounding the spinal cord has also been observed after ectopic floor plate grafts (Placzek et al. 19906) and in mouse embryos homozygous for mutations at the Danforth’s short tail locus in which the floor plate is absent (Bovolenta and Dodd, 1991). These results emphasize the importance of using cell-type specific markers rather than cell position and axonal trajectory to analyze the control of cell identity in the developing nervous system.
More generally, the changes in neural cell pattern observed in the absence of the notochord and floor plate support the idea that the differentiation and patterning of many distinct cell types in the ventral region of the nervous system is controlled by midline-derived signals.
Discussion
The experiments described here provide evidence that cell patterning within the neural tube is controlled by signals that derive from two axial midline cell groups, the notochord and the floor plate. An early step in the establishment of cell pattern within the neural tube is the induction of the floor plate by the notochord. The floor plate may then regulate the pattern of cell differentiation along the dorsoventral axis of the neural tube. Although the mechanisms underlying these patterning events have not been resolved, the identity of neuroepithelial cells in the ventral region of the neural tube appears to be defined by position with respect to the ventral midline (Fig. 10). These observations support the idea that cell fate within the developing nervous system, as in other embryonic tissues (Chasan and Anderson, 1991; Tickle et al. 1975), depends largely on the interpretation of positional signals that derive from local sources (Wolpert, 1969). In vertebrates such polarizing signals appear to derive from discrete cell groups with specialized organizing properties (Spemann, 1938; Gerhart and Keller, 1986).
Induction of the floor plate
Studies of amphibian and avian embryos have provided evidence that the cells that will eventually populate the midline of the neural plate undergo extensive cell rearrangements and possess specialized migratory properties prior to overt signs of neural differentiation (Jacobson, 1981; Keller et al. 1985; Alvarez and Schoenwolf, 1991). However, the acquisition of the specialized signalling properties of the floor plate that control neural cell pattern and axonal trajectory require local inductive signals from the notochord. The demonstration of floor plate induction in vitro (Figs 3 and 4) indicates that notochord-derived signals are sufficient to induce floor plate properties in the absence of factors supplied by other embryonic cell types. In addition, these in vitro studies suggest that contact between the notochord and neural plate is required for induction of the floor plate. The contact-dependence of floor plate induction could reflect the membrane anchoring of an inductive ligand, the requirement for a high concentration of a diffusible factor, or the sequestration of a diffusible signal in the extracellular matrix that surrounds the notochord.
Studies in other vertebrate species have provided evidence consistent with the induction of the floor plate by the notochord. UV-treatment of fertilized Xenopus eggs can lead to the development of tadpoles lacking a notochord. The neural tube of these embryos lacks ventral midline cells with characteristics of the floor plate (Youn and Malacinski, 1981; Clarke et al. 1991). In addition, analysis of neural differentiation in the embryonic mouse mutants Danforth’s short tail and truncate in which the notochord fails to differentiate or undergoes early degeneration has shown that the caudal region of the spinal cord lacks a morphogenetically- and antigenically-defined floor plate (Theiler, 1959; Bovolenta and Dodd, 1991). However, at more rostral regions of the spinal cord of Danforth’s short tail embryos the floor plate is present even though the notochord is absent (Bovolenta and Dodd, 1991). This may result from the persistence of primitive notochord cells for a time sufficient to induce a floor plate in anterior regions.
Relative contributions of the notochord and floor plate to the control of neural cell pattern
The dependence of floor plate differentiation on inductive signals from the notochord makes it difficult to distinguish whether the dorsoventral patterning of the neural tube is controlled by signals from the notochord, the floor plate or from both cell groups.
Evidence discussed below suggests that signals from either the floor plate or the notochord may be sufficient to control the differentiation of distinct cell types in the ventral neural tube. Nevertheless, it is difficult at present to rule out the possibility that polarizing signals are synthesized only by the notochord and merely sequestered by the floor plate. The slow release of such signals could then account for the apparent polarizing activity of the floor plate in grafting experiments.
The advantage of two midline cell groups with similar or identical signalling properties is not clear. This which the pattern of cell differentiation within the neural tube is established. At the neural plate stage, the notochord is in direct contact with the midline of the neural plate and may initiate the mediolateral patterning of the nervous system by inducing the floor plate. However, during subsequent stages of neurulation, the notochord is displaced ventrally, becomes surrounded by an extracellular matrix and is separated from the neural tube by mesenchymal cells (Jurand, 1962). Thus, if the notochord were the sole source of signals that control the pattern of cell differentiation along the dorsoventral axis of the neural tube, the entire pattern would have to be established at a time when the notochord is still in proximity to the neural ectoderm. The acquisition of polarizing properties by the floor plate provides a mechanism by which midline-derived signals could continue to impose pattern on the adjacent neuroepithelium at stages after the displacement of the notochord.
Homeogenetic induction and heterogeneity of the floor plate
One of the apparent signalling properties of the floor plate is the ability to induce additional floor plate cells (Fig. 10). Evidence for this derives from grafting experiments in chick embryos, in which rat, quail or chick floor plate can induce floor plate-specific antigens in the adjacent neuroepithelium (Fig. 6A and Yamada et al. 1991). Similar conclusions have been reached from an analysis of zebrafish cyclops mutant embryos which lack a recognizable floor plate even though the notochord is present (Hatta et al. 1991). Wild-type ectodermal cells transplanted into a cyclops host embryo can populate the midline of the neural tube and acquire antigenic and morphological characteristics of the floor plate (Hatta et al. 1991). This suggests that the cyclops mutation affects the response properties of the neural ectoderm to notochord-derived signals rather than the signalling properties of the notochord. In addition, mutant host midline cells that are adjacent to wild-type floor plate cells also acquire floor plate morphology. These observations suggest that wild-type cells can be induced by the notochord to differentiate into floor plate cells, and then are able to induce some floor plate properties in adjacent cyclops neuroepithelial cells. Both the notochord and floor plate, therefore, appear to have homeogenetic inductive properties. The notochord induces notochord-like signalling properties in overlying neuroepithelial cells and newly differentiated floor plate cells induce floor plate properties in adjacent neuroepithelial cells.
The ability of wild-type floor plate cells to induce floor plate properties in cyclops neuroepithelial cells that are refractory to notochord-derived signals (Hatta et al. 1991) also raises the possibility that the notochord and floor plate are sources of distinct signalling molecules, each with the capacity to induce the floor plate. However, the cyclops result could also be obtained if the mutation attenuates the response of neuroepithelial cells to a common signal that is provided at greater concentration by the floor plate than by the notochord.
The ability of the floor plate to induce additional floor plate cells together with the apparent contact dependence of floor plate induction provides a potential mechanism for maintaining the lateral dimensions of the floor plate during early neural tube development. In chick and amphibian embryos, cells that eventually occupy the midline of the neural plate undergo convergent extension movements in which cell intercalation results in the longitudinal expansion of the set of cells that initially occupy the midline (Jacobson et al. 1986; Schoenwolf and Alvarez, 1989; Schoenwolf and Smith, 1990; Fraser et al. 1990). If the number of floor plate cells remained constant, these movements would result in a progressive decrease in the lateral dimensions of the floor plate. However, analysis of floor plate antigen expression in chick embryos indicates that the width of the floor plate does not decrease during elongation of the neural tube and, if anything, increases (Yamada et al. unpublished observations). Maintenance of the lateral dimensions of the floor plate could be achieved if the initial population of midline floor plate cells were to recruit lateral neuroepithelial cells into the floor plate by homeogenetic induction. The notochord may therefore induce primary floor plate properties in those cells at the midline of the neural plate that are in direct contact with it. A second phase of floor plate development appears to involve the induction of secondary floor plate properties in cells of the adjacent neuroepithelium (Fig. 10).
The homeogenetic inducing activity of the floor plate requires a mechanism for preventing the lateral propagation of this signal throughout the neuroepithelium, converting the entire neural tube into a floor plate. One possibility is that secondary floor plate cells are not able to recruit additional floor plate cells. A second possibility is that the notochord and/or floor plate provide long range signals that induce more lateral neural plate cells to differentiate into ventral neural cell types such as cells in region X and motor neurons (Fig. 10). The specification of these ventral cell types may occur over the same time period, or shortly after the time that the midline cells acquire primary floor plate properties. Thus, by the time that secondary floor plate cells have acquired homeogenetic inducing properties only a small number of cells remain uncommitted and thus available for additional recruitment into the floor plate.
A distinction between primary and secondary floor plate cells is further suggested by the heterogeneity observed in the antigenic properties of floor plate cells in vertebrate embryos. In the chick, the SCI antigen is restricted to the medial subset of floor plate cells that express the FP1 antigen (Fig. 6B, C and Yamada et al. 1991). Similarly in rat embryos the KI antigen (Dodd and Jessell, 1988) defines a medial subset of floor plate cells that express the P84/6G3 antigen (Chang and Lagenauer, 1990; Jessell and Morton, unpublished observations) and the P35 protein (McKanna and Cohen, 1989). There may also be cellular heterogeneity in the floor plate of zebrafish embryos. The morphological and antigenic features of the single cell that occupies the ventral midline of the spinal cord differ from those of the adjacent lateral cells (Hatta et al. 1991). These two lateral cells in the zebrafish spinal cord exhibit properties characteristic of the rat floor plate (Bovolenta and Dodd, 1990) in that commissural axons appear to alter their trajectory from transverse to longitudinal growth in contact with these lateral cells rather than with the midline cell (Kuwada et al. 1990a, b).
Control of motor neuron differentiation
The cellular mechanisms involved in the differentiation of discrete neuronal cell types, for example motor neurons, in response to notochord and floor plate-derived signals remain unclear. Motor neuroblasts derive from the ventricular zone in the ventral region of the neural tube and migrate laterally to form a discrete cluster of motor neurons beginning at stage 15 (Langman and Haden, 1970; Hollyday and Hamburger, 1977). The increase in the total number of motor neurons after notochord and floor plate grafts (Yamada et al. 1991) excludes the possibility that the appearance of ectopic motor neurons results from the redistribution of a fixed number of motor neuron precursors. Moreover, there is very little cell death in the lateral neural tube before stage 15 (Homma et al. 1990) which makes it unlikely that the appearance of ectopic motor neurons results from the rescue of a population of committed motor neuron progenitors. It seems more likely that the differentiation of motor neurons in ectopic regions of the spinal cord, in response to notochord or floor plate grafts, results from a change in the fate of pluripotent neuroepithelial progenitors. In support of this, lineage analysis of retrovirally marked cells in the chick neural tube indicates that motor neurons derive from progenitors that give rise to other classes of neurons and also to glial cells (Leber et al. 1990). Such multipotential progenitors may be distributed throughout the neural tube with signals from the floor plate and notochord required for their differentiation into motor neurons (Fig. 10). Progenitors that are not exposed to floor plate signals appear to acquire dorsal cell fates (see following section).
This view of the interactions controlling motor neuron differentiation differs from that of Hirano et al. (1991), who surgically deprived one side of the neural tube of stage 12 chick embryos of a floor plate. Histological examination of such embryos led to the proposal that neurons located in dorsal regions of the neural tube can give rise to motor neurons but normally are inhibited from doing so by signals from the floor plate. Our experiments suggest that by stage 12 the floor plate has already differentiated, and that the specification of motor neurons by the floor plate and/or notochord has already occurred (Yamada et al. 1991).
Thus, elimination of the notochord or separation of the floor plate from one side of the neural tube at stage 12 or later may not result in the conversion of dorsal neural tube cells into motor neurons as proposed by Hirano et al. (1991). Instead, the presence of motor neurons in dorsal regions may reflect the dorsal displacement of the normal column of motor neurons as a result of distortion of the neuroepithelium caused by slitting the ventral midline of the neural tube.
The influence of the notochord and other mesodermal tissues on the differentiation of ventral neuroblasts in the chick neural tube has been examined previously. Inversion of the neural tube along its dorsoventral axis at stage 11 – 12 has been reported to result in the appearance of ventral characteristics in the formerly dorsal spinal cord, although specific cell types were not identified (Steding, 1962). Similarly, notochord grafts placed next to the neural tube have been shown to lead to the proliferation of ventral neuroblasts (van Straaten et al. 19856, 1988). In addition, grafts of notochord and muscle next to the spinal cord of older embryos have been shown to increase the proliferation of ventral neurons (van Straaten et al. 19856; van Straaten and Drukker, 1987; van Straaten et al. 1989; Fontaine-Perus et al. 1989). However, it is unclear if the proliferative effects of grafts into older embryos are related to the changes in cell pattern observed in young embryos.
Control of cell differentiation in the dorsal neural tube
Distinct neuronal cells within the spinal cord differ markedly in their dependency on ventral midline-derived signals (Yamada et al. 1991). Elimination of the notochord and floor plate results in the absence of motor neurons but other classes of cells including neural crest cells and commissural neurons, are present. The differentiation of the dorsal neural tube therefore does not appear to depend on the notochord and floor plate. Thus, the acquisition of dorsal cell properties may represent the constitutive fate of neuroepithelial cells. An alternative possibility is that neuroepithelial cells in the ventral region of the neural tube require floor plate and notochord-derived signals for their proliferation. In the absence of these signals, these ventral cells fail to divide, resulting primarily in a deletion rather than a change in fate of cells in the ventral neural tube.
The differentiation of neuroepithelial cells into dorsal cell types may require signals from other cell groups, perhaps the roof plate. Indirect support for the existence of roof plate-derived signals that affect neural tube patterning comes from analysis of chick embryos which have received dorsal grafts of notochord or floor plate (Yamada et al. 1991). The induction of a floor plate in response to a dorsally placed notochord or floor plate occurs only when fusion of the dorsal midline of the neural tube fails, presumably preventing roof plate differentiation. Thus, signals from the roof plate may render the adjacent dorsal neural tube refractory to notochord and floor plate-derived signals. The nature of such roof plate-derived signals is not known but several members of the wnt family of molecules are localized in and around the roof plate (Wilkinson et al. 1987; Roelink and Nusse, 1991).
Control of neural cell pattern by contact-dependent or diffusible signals
The floor plate and notochord could influence the pattern of cell differentiation within the neural tube by at least three different strategies.
First, a cascade of locally-acting, possibly contact dependent signals might originate from the floor plate, with each cell type responsible for inducing its dorsolateral neighbors (Fig. 11A). The inability of grafts of the region of neuroepithelium immediately adjacent to the floor plate to induce motor neurons provides evidence against such a model. However, this does not exclude a second scheme (Fig. 11B), in which the specification of different cell types depends both on contact-dependent signals and a long range diffusible signal that derives solely from the notochord and/or floor plate. Thirdly, cell differentiation in the ventral neural tube could be controlled solely by signals from the floor plate and notochord (Fig. 11C). For example, the notochord and floor plate could act as local sources of a factor that diffuses through the adjacent neuroectoderm, establishing a concentration gradient with its high point at the ventral midline (Fig. 11C). In this scheme, cell identity and the overall pattern of cell types would be defined by the concentration of diffusible signal to which precursor neuroepithelial cells are exposed (Yamada et al. 1991). Neural plate cells exposed to highest concentrations of this factor would differentiate into floor plate; cells exposed to a lower concentration would differentiate into cells of region X and cells exposed to a still lower concentration would differentiate into motor neurons. A variation on this model is that multiple signals from the floor plate or notochord have different ranges of action, thus specifying distinct cell types at different distances from the ventral midline.
Graded signals have been invoked in the generation of cell pattern along the anteroposterior axis of the developing chick wing bud (Tickle et al. 1975). There, anteroposterior pattern appears to be under the control of a specialized region of posterior mesoderm known as the zone of polarizing activity (ZPA), which can evoke mirror-image digit duplications when grafted to ectopic sites (Tickle et al. 1975). Retinoic acid mimics the effects of the ZPA (Tickle et al. 1982) and appears to be distributed unevenly along the anteroposterior axis of the limb bud with its highest concentration in the posterior mesenchyme (Thaller and Eichele, 1987). On this basis, it has been suggested that retinoic acid functions as an endogenous morphogen involved in establishing axial polarity in the developing chick limb (see Brockes, 1989; Eichele, 1989).
The notochord and floor plate, but not other regions of the neural tube, mimic the action of the ZPA and retinoic acid in respecifying digit pattern in the chick limb (Wagner et al. 1990). Biochemical studies show that the floor plate can synthesize and release active retinoids in vitro (Wagner et al. 1990). Moreover, the γ and β retinoic acid receptors are expressed at high levels in the neural plate and neural tube (Smith and Eichele, 1991; Ruberte et al. 1991). The ability of the notochord and floor plate to control the pattern of cell differentiation in both the developing limb bud and neural tube may therefore have a common molecular basis, possibly involving retinoids. However, grafts of the ZPA or of retinoic acid-impregnated beads do not mimic the floor plate or notochord in changing the pattern of cell differentiation in the neural tube (Yamada, unpublished observations). Thus, retinoids released by the floor plate may have functions independent of dorsoventral patterning or act only in combination with additional factors.
Candidate genes that may be involved in the response of undifferentiated neuroepithelial cells to floor plate and notochord-derived signals have been identified. Several of the Hox genes exhibit dorsoventral restrictions in their expression pattern in the early spinal cord (Graham et al. 1991). Similarly, Pax genes are expressed in restricted regions along the dorsoventral axis of the neural tube and spinal cord (Goulding et al. 1991a). In particular, Pax-3 exhibits a striking dorsal restriction within the neural tube (Goulding et al. 19916). Finally, Mash-1, a mammalian helix-loop-helix protein is expressed in cells in region X, immediately adjacent to the floor plate (Lo et al. 1991).
Control of neural cell pattern in different vertebrates
The cellular signals involved in the control of neural cell pattern are likely to be conserved in different vertebrate species. However, recent studies on Xenopus embryos suggest that the time of action and cellular origin of these signals may differ. For example, in UV-treated Xenopus embryos which lack a notochord and floor plate the spinal cord contains primary motor neurons, but appears to lack secondary motor neurons (Clarke et al. 1991). Other neural cell classes, in particular Rohon-Beard neurons (Lamborghini, 1980) and Kolmer-Agdhur cells (Dale et al. 1987) also appear in UV-treated embryos (Clarke et al. 1991). A likely reason for this difference is that in Xenopus the specification of neuronal cell types such as primary motor neurons (Hughes, 1959; van Mier et al. 1985; Nordiander, 1986) and Rohon-Beard neurons (Lamborghini, 1980) occurs during early gastrulation, before overt signs of neural plate formation. These results indicate that in Xenopus the determination of primary neurons must be controlled by signals that derive from cells other than the floor plate.
The absence of secondary motor neurons in the spinal cord of notochordless Xenopus embryos (Clarke et al. 1991) suggests that the differentiation of the majority of motor neurons and perhaps other neurons that are specified at later stages is dependent on the notochord and floor plate. Recent studies of the development of neural cell types in the ectoderm of Xenopus exogastrulae have provided support for this idea (Ruiz i Altaba, personal communication). Neural cells characteristic of secondary motor neurons are virtually absent from the exogastrula ectoderm even though many other classes of neural cells are present. Secondary motor neurons can, however, be induced in exogastrula ectoderm by implanting a notochord, consistent with the idea that signals from the notochord and floor plate are required. In birds and mammals, the determination of most classes of neurons including motor neurons does not appear to occur until after the onset of neurulation (Hollyday and Hamburger, 1977). Thus, most neurons in the avian and mammalian spinal cord may correspond to the secondary neurons in Xenopus with respect to the source of the signals involved in their differentiation.
In apparent contrast to observations in Xenopus (Clarke et al. 1991) the analysis of motor neuron differentiation in the zebrafish cyclops embryos indicates that both primary and secondary motor neurons are present despite the apparent absence of a floor plate (Kimmel et al. 1991; Eisen, 1991). The presence of primary motor neurons in cyclops embryos is likely to reflect the existence of inductive signals that operate during gastrulation, as described above in Xenopus. However, the presence of secondary motor neurons in cyclops differs from expectations based on analysis of UV-treated Xenopus embryos and notochordless chick embryos. There are two possible explanations for this difference. First, the absence of morphological and antigenic features of the midline floor plate cell in cyclops embryos does not rule out that partial floor plate properties remain and that these are sufficient to induce other ventral cell types including secondary motor neurons. Alternatively, as discussed above, the notochord may be able to control motor neuron differentiation directly, in the absence of a floor plate. If this is the case the cyclops mutation would appear to block the signalling pathway involved in floor plate differentiation without affecting the pathway involved in motor neuron specification. This raises the possibility that the notochord activates distinct signalling pathways that control floor plate and motor neuron differentiation.
Control of neural cell identity along the anteroposterior axis
The notochord and floor plate may also contribute to the pattern of neural cell differentiation along the anteroposterior axis of the neural tube. Many classes of neurons are absent from the forebrain and restricted to more posterior regions of the CNS. For example, the most anterior group of motor neurons, the oculomotor nucleus, is located in the midbrain near the anterior end of the floor plate. Thus, motor neurons represent one of several classes of neurons found at more posterior regions of the CNS which appear to be dependent on the floor plate for their differentiation. These observations reinforce the suggestions of Ahlborn (1883) and Kingsbury (1930) that the anteroposterior domain of the neuraxis over which motor neurons differentiate is defined by the notochord and floor plate.
Studies of amphibian neural development have provided evidence that regional differentiation along the anteroposterior axis is established during the early development of the neural plate, before that along the dorsoventral axis (Mangold, 1933; Roach,1945; Jacobson, 1964; Ruiz i Altaba and Melton, 1989). These findings, together with the present results suggest that the fate of a cell within the nervous system is restricted initially by its anteroposterior position within the neural plate. The location of cells along the dorsoventral axis of the neural tube with respect to the floor plate appears later to determine which of a limited number of possible fates, appropriate for a given anteroposterior position, is selected.
ACKNOWLEDGEMENTS
We thank David Anderson, Chuck Kimmel and Ariel Ruiz i Altaba for permission to cite unpublished results, Ariel Ruiz i Altaba for helpful comments and Vicki Leon, Ira Schieren and Eric Hubei for help in preparing the manuscript. T. M. Jessell is an Investigator and T. Yamada is a Research Associate of the Howard Hughes Medical Institute. J. Dodd was supported by grants from NIH and NSF, a Klingenstein Fellowship and the Irma T. Hirschi Foundation. M. Tessier-Lavigne was supported by a scholarship from the Lucille P. Markey Foundation.