Distinct modes of floor plate induction in the chick embryo

The floor plate of the vertebrate embryo develops at the ventral midline of the neural tube. It acts as an embryonic organiser and plays an essential role both in the generation of specific neuronal subtypes along the dorsoventral axis of the brain and spinal cord and in the guidance of axons(Giger and Kolodkin, 2001; Jessell, 2000).

Although the role of the floor plate in patterning the neural tube is well accepted, its ontogeny has been a controversial subject(Le Douarin and Halpern, 2000; Placzek et al., 2000). The prevailing view, for which there is extensive evidence in the chick, is that axial mesodermal notochord cells are the source of an instructive inducing signal that mediates floor plate differentiation in medial cells of the overlying neural plate (Jessell,2000; Placzek et al.,2000). A large body of evidence, moreover, suggests that the secreted signalling molecule Sonic hedgehog (Shh) mediates the ability of notochord to induce floor plate differentiation. Shh is expressed in the node and the notochord prior to floor-plate differentiation. Gain-of-function experiments show that Shh can induce the ectopic differentiation of floor-plate cells in the neural plate in vitro(Marti et al., 1995; Roelink et al., 1994), while blockade of Shh in the notochord eliminates its ability to induce floor plate cells (Ericson et al., 1996). In support of studies in the chick, mutations in the Shh gene, and in components of the Shh signalling pathway, in mouse, block ventral midline differentiation (Chiang et al.,1996; Ding et al.,1998; Matise et al.,1998; Wijgerde et al.,2002).

Several recent studies, however, have questioned this model of floor plate induction. In particular, analysis of chick-quail chimaeras in which quail cells from the chordoneural hinge (CNH), a derivative of the Node, are grafted into chick embryos, have led to the proposal that medial floor plate cells are derived from a population of precursors that are initially situated in the Node, can segregate into either notochord or floor plate, and are already pre-specified within this region (Le Douarin and Halpern, 2000; Teillet et al., 1998). In this model, floor-plate cells thus derive from pre-specified cells that intercalate from the node into the neural midline.

A further challenge to the paradigm of notochord/Shh-mediated floor plate induction arises through observations of zebrafish embryos. Floor-plate cells persist in embryos in which notochord precursors are surgically ablated,demonstrating that a normally developed notochord is not a pre-requisite for floor plate differentiation in this species(Shih and Fraser, 1995). Analyses of zebrafish mutant embryos further supports this contention. Mutations in both no tail (ntl) and floating head(flh) affect notochord formation(Amacher and Kimmel, 1998; Halpern et al., 1993; Schulte Merker et al., 1994; Talbot et al., 1995). Despite this, in ntl mutant embryos the most medial set of floor plate cells are present, and even expanded (Halpern et al., 1997; Odenthal et al.,1996; Odenthal et al.,2000). flh mutants likewise contain patches of cells at the ventral midline that express medial floor-plate markers(Halpern et al., 1995; Halpern et al., 1997; Schier et al., 1997). Moreover, while there is compelling evidence for a requirement for Shh signalling in the induction of floor plate character in amniotes, in zebrafish embryos Hh signalling appears crucial to the induction of lateral floor plate cells, but is not required for the differentiation of medial floor-plate cells(Chen et al., 2001; Etheridge et al., 2001; Odenthal et al., 2000; Schauerte et al., 1998; Varga et al., 2001). Instead,the TGFβ superfamily member Nodal appears essential for medial floor plate induction. Mutations in the zebrafish nodal-related gene cyclops (ndr2) and one-eyed pinhead (oep),an obligate co-factor for Nodal signal transduction(Gritsman et al., 1999) both cause loss of the medial floor plate throughout the length of the neural tube(Hatta, 1992; Hatta et al., 1991; Krauss et al., 1993; Rebagliati et al., 1998; Sampath et al., 1998; Schier et al., 1997; Shinya et al., 1999; Strahle et al., 1997; Zhang et al., 1998). Intriguingly, while analysis of the requirement for Nodal signalling in zebrafish has suggested that medial floor plate specification occurs early in development and within the organiser region, evidence from the analysis of oep mutants suggests that it is nevertheless the result of an inductive interaction (Gritsman et al.,1999; Strahle et al.,1997).

These extensive analyses of floor plate development appear to point towards very disparate models of floor plate formation in distinct species. However, a caveat, and possible explanation for the varied conclusions of these studies is their failure to analyse floor plate development at equivalent stages of embryogenesis. To perform a more direct comparative analysis, we have examined the differentiation of floor-plate cells in the early chick embryo, over the period of gastrulation/early neural plate formation. Our studies show that many medial floor plate cells that form at this time do not derive from Hensen's node itself. Instead, they derive from a region of the prenodal epiblast that lies anterior to Hensen's node, previously shown to contribute to the floor plate, and designated `area a'(Garcia-Martinez et al., 1993; Schoenwolf et al., 1989; Schoenwolf and Sheard, 1990). Real-time lineage analyses reveal that `area a'-derived cells exist as a separate population of floor-plate precursors and do not contribute progeny to Hensen's node, arguing that the shared origins with notochord cells are not a requirement for floor-plate development along the length of the neuraxis. Our studies show that `area a' floor-plate precursors are induced by rapid signalling events mediated by the early forming prechordal mesendoderm. Together, our evidence shows that floor-plate cells along the neuraxis are induced to differentiate and argues against a requirement for pre-specification within the organiser.

In addition, we observe that Shh and Nr1 are co-expressed in the nascent prechordal mesoderm at the time of `area a' differentiation,suggesting that Nodal signalling may play a role in amniote floor plate induction. In support of this, we find that Nodal and Shh can co-operate to induce floor plate character in `area a' cells in vitro. The data presented in this study thus indicate that different signalling events mediate early and late floor plate induction in the chick and support the development of an integrated model of floor plate differentiation in both amniote and anamniote embryos.

Focal injections of fluorescent lipophilic dyes designed to label fewer than 50 cells were made by controlled pressure injection into live embryos either in ovo or in New culture. During New culture, embryos were explanted into L15 medium at HH (Hamburger and Hamilton, 1951) stage 4 and injected with a solution of DiI and/or DiD, 5 mg/ml in 100% ethanol (Molecular Probes). Focal injections were made using a picospritzer II microinjection system (General Valve). Following dye injection, embryos were replaced onto their vitelline membrane and prepared for New culture according to established techniques(New, 1955; Stern and Ireland, 1981).

For time-lapse analysis of cell movement, embryos were then cultured in plastic culture dishes over thin albumen in a culture dish in which the central plastic area had been replaced with a thin glass coverslip to facilitate visualisation. Embryos were visualised using an inverted confocal microscope as previously described (Kulesa and Fraser, 1998). The microscope was surrounded with an insulating chamber maintained at 38°C for the duration of the time-lapse experiment. Single confocal images were taken at 5 or 10 minute intervals for the duration of the analysis.

All embryos were staged and dissected in cold L15 medium (Gibco-BRL). `Area a' explants were prepared from HH stage 4 embryos by making two parallel cuts either side of and anterior to Hensen's node, followed by two cuts at right angles to remove a square of tissue from the region anterior to Hensen's node. The epiblast layer was then isolated from underlying tissue with dispase (1 mg/ml). Explants of `area a'-derived tissue at HH stages 4+, 5 and 6 were isolated by taking an equivalent area of tissue just anterior to Hensen's node. In all cases, explant culture was performed in collagen gels according to published techniques (Placzek and Dale,1999).

Nodal protein, produced by transient transfection of 293T cells with pcDNA3-mNodal (containing the coding sequence for mouse Nodal) was concentrated tenfold using Centri-plus columns (Amicon) and then diluted 1:10 in explant culture medium. Human Shh-N protein (Biogen) was added to the tissue culture medium at the concentrations indicated.

For tissue recombination experiments, HH stage 4+ prechordal mesendoderm was identified by morphology. Explants were prepared by making cuts either side of Hensen's node and at the anterior and posterior limits of the prechordal mesendoderm using sharpened tungsten needles prior to separation of the tissues using 1 mg/ml dispase. Intermediate neural plate tissue from E9.5 rat embryos was isolated as previously described(Placzek et al., 1993). Prechordal mesendoderm was placed in contact with either `area a' or rat neural plate explants in collagen gels and cultured as previously described(Placzek and Dale, 1999).

In vivo grafting experiments were performed as previously described(Placzek et al., 1990). Briefly, a small incision was made between the open neural groove and adjacent presomitic mesoderm in the caudal region of HH stage 10 chick embryos in ovo. Explants of notochord taken from the caudal region of HH stage 10 embryos or nascent prechordal mesendoderm from HH stage 4+ embryos were inserted into the incision adjacent to the neural plate at an intermediate position, in between basal and alar plates. After operations were performed eggs were resealed and incubated until HH stage 19-21 prior to fixation and analysis by immunohistochemistry.

HH stage 4, 4+ or 5-embryos were prepared for New culture, leaving the ventral surface of the embryo exposed. Removal of the prechordal mesendoderm was performed by making a shallow cut through the endodermal and mesodermal layers just anterior to Hensen's node and then scraping away the mesendoderm anterior to it. Following operation, embryos were prepared for New culture as described (New, 1955; Stern and Ireland, 1981) and allowed to develop prior to fixation and further analysis.

Embryos and explants were analysed by immunohistochemistry according to standard techniques (Placzek et al.,1993). The following antibodies were used (dilutions in parentheses): 68.5E1, anti-Shh mAb (1:50)(Ericson et al., 1996); 4C7,anti-HNF3β mAb (1:40) (Ruiz i Altaba et al., 1995); anti-Sox2 pAb (1:500)(Pevny et al., 1998);anti-Lim1/2 (1:50); and anti-Not 1 (1:50). Appropriate secondary antibodies(Jackson Immunoresearch) were conjugated to Cy3.

Embryos and explants were processed for in situ hybridisation as described previously (Vesque et al.,2000). The following template DNAs were to generate a digoxigenin labelled antisense RNA probes: plasmid pCM21 containing a cDNA encoding chick Netrin 1 was linearised with EcoRI and transcribed with T7 polymerase; plasmid pcvhh containing a cDNA encoding chick sonic hedgehog was linearised with SalI and transcribed with SP6 polymerase; plasmid pcp7 containing a cDNA encoding HNF3β was linearised with HindIII and transcribed with SP6 polymerase; plasmid pcGsc containing a cDNA encoding chick Goosecoid was linearised with EcoRI and transcribed with SP6 polymerase (Vesque et al., 2000).

In vivo time-lapse confocal microscopy was performed to analyse the migration of `area a' epiblast cells during gastrulation over the period HH stages 4 to 8. Focal injections of DiI were made into HH stage 4 embryos. Two sites were labelled; `area a', just anterior to Hensen's node and as a reference point, a region of midline epiblast ∼ 200μm more anteriorly(Fig. 1A,C; see Movie 1).

At the beginning of analysis, a few cells from `area a' had already moved away from their original position (Fig. 1C, white arrowhead). Over the next 9 hours, many `area a' cells moved posteriorly, along the axis of the embryo(Fig. 1D-J), and a smaller number of cells moved anteriorly (Fig. 1B-G, arrow in Fig. 1B). Very occasionally, cells moved laterally but with time moved back towards the midline to join the main, axial stream of cells (arrows, Fig. 1C-F). During their migration, `area a'-derived cells extended long filamentous processes polarised in the direction of their movement (arrowhead in Fig. 1B).

To confirm that `area a' cells colonise the midline, we examined whether`area a'-derived cells express the ventral midline cell marker Shh. DiI was injected into `area a' cells in vivo at HH stage 4, and embryos developed in ovo until HH stage 8. Examination of sections confirmed that DiI-labelled cells were confined to the ventral midline floor plate and revealed that they populated medial-most floor plate cells(Fig. 1K-M). `Area a'-labelled cells were never detected within axial mesoderm. These analyses also indicated that `area a' cells contributed largely to anterior ventral midline regions extending from the diencephalon, through the midbrain and hindbrain, with only sporadic labelling detected more posteriorly. Labelled cells were never detected anteriorly within the telencephalon (n=10).

These results demonstrate that cells from the midline, prenodal region of the chick epiblast rapidly populate the midline of the developing neural tube during gastrulation, extending long cellular processes as they migrate. In addition, they demonstrate that `area a'-derived cells largely populate medial-most ventral midline cells that form in the anterior neural tube.

We next addressed whether, in addition to populating the ventral midline,`area a' cells contribute to Hensen's node to form a population of floor plate precursors within this structure.

Focal injections of DiI and DiD were made into the epiblast layer at HH stage 4 (n=5), and the embryos were followed for a 4 hour period,until they reached the equivalent of HH stage 6(Fig. 2; see Movie 2). DiI was used to label cells in `area a' (red labelling in Fig. 2). DiD was used to label cells in the superficial, epiblast layer of Hensen's node (blue labelling in Fig. 2); this also served to mark the position of Hensen's node. During time-lapse confocal imaging analyses, DiI-labelled cells from `area a' were again observed to rapidly move both anteriorly and posteriorly to populate the midline of the embryo(Fig. 2B-H). However, at no point in this analysis were cells from `area a' observed to colonise Hensen's node itself.

During the 4 hour period, Hensen's node regressed posteriorly (blue labelling in Fig. 2B-H). Simultaneously, a stream of Hensen's node-derived cells moved first laterally and then anteriorly to populate the midline. The rate at which this occurred was visibly greater than the rate of node regression, suggesting not only a depositing of cells by Hensen's node, but an active anterior movement of some of these node-derived cells. Focussing through the embryos during time-lapse analysis indicated that such DiD labelled cells populated only the mesodermal layer. To confirm this, single focal injections of DiI were made into the epiblast layer of Hensen's node at HH stage 4, and the embryos analysed at HH stage 6. Sectioning revealed that epiblast cells that leave Hensen's node over the period HH stage 4-6 populate the forming axial mesendoderm, and not the superficial neural layer (Fig. 2I; n=5). By contrast, when identically labelled embryos were allowed to develop beyond HH stage 6, labelled cells were detected in both notochord and floor plate (Fig. 2I, inset; n=3).

Together, these analyses indicate firstly that `area a'-derived cells remain separate from Hensen's node as neurulation proceeds and do not contribute at this time to any population of floor plate precursors residing in Hensen's node. Second, it suggests that the principal contributor to the earliest forming, and most anterior ventral midline is `area a' and not Hensen's node, with Node-derived cells only contributing to the later forming floor plate.

The majority of studies in amniotes showing the induction of floor plate by notochord have examined the differentiation of floor-plate cells that form in thoracic regions of the neuraxis (Artinger and Bronner-Fraser, 1993; Placzek et al., 2000; van Straaten and Hekking,1991; Yamada et al.,1991) and have not examined the induction of the anterior, `area a'-derived population. We therefore next addressed the mechanisms by which`area a'-derived cells differentiate into floor plate and examined whether these differ from those which generate floor plate in the posterior neural tube.

We first addressed the state of specification of `area a'-derived cells as development proceeds. Explants of `area a'(Fig. 3A), or its derivatives,were cultured and examined for expression of markers of floor plate and neural character. Despite being adjacent to Shh-expressing cells in Hensen's node(Fig. 3A, arrowhead), cells explanted from `area a' at HH stage 4 do not express the floor-plate markers Shh, Hnf3b/Foxa2 or Netrin1(Fig. 3B-D), indicating that at this stage they are not specified to become floor plate. Expression of floor plate markers was not observed at any time point analysed (20, 24, 36, 40 hours; n>50). Analysis of Sox2, a marker of undifferentiated neuroepithelial cells (Streit et al.,1997) indicates that `area a' cells are specified as neural at the time of isolation (Fig. 3E). Thus, prior to the overt formation of axial mesoderm, floor plate precursors in `area a' exist as committed neural precursors.

After HH stage 4, however, increasing numbers of `area a'-derived cells express ventral midline characteristics. Explants isolated at HH stage 4+contain a small number of cells which appear to co-express Hnf3b/Foxa2 and Shh (20-40% of cells in all explants analysed, n=10; Fig. 3F,G). By HH stage 5 the proportion of cells in each explant co-expressing Hnf3b/Foxa2 and Shh has increased to between 50 and 70% (n=10; Fig. 3H,I), and by HH stage 6 to between 60 and 90% (n=10; Fig. 3J,K). These cells do not express Lim1/2, brachyury or 3B9 (not shown), markers that label axial mesodermal cells, ruling out the possibility that `area a' cells have differentiated to an axial mesoderm fate, and indicating instead that they are specified to ventral midline floor plate cells. Thus, over a timespan of less than 4 hours, the majority of `area a'-derived cells become specified as floor plate.

The specification of `area a' cells to a floor-plate identity coincides with the onset of axial mesoderm formation, raising the possibility that the first emerging axial mesoderm cells are responsible for inducing `area a'cells to a floor-plate fate. To test this, we removed the earliest forming axial mesendoderm, together with the deep layers of Hensen's node at HH stage 4 (Fig. 4A,B; n=8). Previous studies have indicated that these layers contribute to axial mesoderm and not to neural tissue (Selleck and Stern, 1991).

After 18 hours in culture, operated embryos appeared to be morphologically normal, although the anterior neuropore was rather pronounced and development was retarded, embryos only reaching HH stage 6 to 7(Fig. 4D). In situ hybridisation revealed that expression of the early floor-plate markers Hnf3b/Foxa2 and Shh could not be detected(Fig. 4D and not shown). Analysis of sectioned embryos revealed that the two halves of the neuroepithelium were conjoined (Fig. 4G, arrow). Sectioning confirmed that both prechordal mesoderm and notochord were absent in operated embryos(Fig. 4G arrowhead) and revealed that Shh expression was almost completely absent throughout the neuroepithelium (Fig. 4G,arrow), weak expression being detected on only individual cells on 2% of sections (not shown). Together, this analysis shows that ablation of the emerging mesendoderm leads to the lack of formation of axial mesoderm and the concomitant loss of floor-plate differentiation in `area a'-derived cells.

We next determined whether `area a'-derived floor plate cells require a prolonged period of contact with early emerging mesendoderm for their differentiation, by performing ablations at HH stage 4+. Embryos could be staged precisely, as when endoderm was removed at HH stage 4+, a fan of axial mesendoderm could be seen extending from Hensen's node(Fig. 4C, white arrowhead). After removal of this mesendoderm, and culture for 18 hours, embryos again appeared morphologically normal, although development was retarded to HH stage 7-8 (Fig. 4F; n=5). In situ analysis revealed that Shh was expressed within anterior ventral midline cells (Fig. 4F,I),although far fewer Shh-expressing cells were detected in the neural midline than were present in control embryo, with expression of Shhon these cells much weaker than on control floor-plate cells (compare Fig. 4F,I with Fig. 4E,H).

Together these analyses suggest that the early emerging mesendoderm is crucial for the normal differentiation of the anterior floor plate. In addition, they suggest that a short exposure to this mesendodermal population is sufficient to begin to induce the differentiation of `area a' cells to a floor-plate identity.

The ablation of emerging mesendoderm at HH stage 4 and 4+ removes both prechordal mesendoderm and notochord progenitor cells. To distinguish which of these two cell types might be responsible for the induction of `area a' cells to a floor plate fate, we performed ablations on slightly older, HH stage 5– embryos, and ablated only the early notochord, leaving anterior-most prechordal mesendoderm cells intact (Fig. 5A-D).

Analysis of operated embryos after 6 hours in culture revealed that notochord cells were not re-established; a clear gap was observed at the midline of the axis (Fig. 5D,arrowhead; n=3). In operated embryos analysed after 20 hours in culture, development appeared largely normal(Fig. 5E; n=6). Analysis of Shh expression in whole-mount preparations showed apparently normal expression in the midline along the length of the embryo. Analysis of sectioned embryos revealed that some prechordal mesendoderm remained intact (Fig. 5F,arrowhead). By contrast, the notochord was missing between diencephalic and spinal cord levels of the axis (Fig. 5G). Despite the lack of notochord, Shh-expressing cells differentiated throughout the neuraxis, including cells located between the diencephalon and hindbrain that derived from `area a'(Fig. 5F,G). The levels of Shh expression, moreover, appeared to be similar to those in control embryos.

These analyses show that early exposure to the prechordal mesoderm can rescue anterior floor-plate cells and reveal that the notochord is not required for their differentiation.

Our studies indicate that prechordal mesoderm is required to rapidly induce`area a' cells to a floor-plate fate. We next determined whether it is sufficient to induce their differentiation, by performing in vitro recombinations of HH stage 4+ prechordal mesendoderm with intermediate neural tissue from E9.5 rat embryos or with `area a' from HH stage 4 chick embryos(Fig. 6A). Prechordal mesoderm induced mature floor-plate cells within rat neural plate, as assessed by the expression of the floor plate marker, FP3(Fig. 6B)(Placzek et al., 1993). Similarly, when HH stage 4+ quail prechordal mesoderm was recombined with`area a', expression of HNF3β and Shh, but not markers of axial mesoderm,were induced (Fig. 6C,D and not shown). Thus, HH stage 4+ prechordal mesendoderm is able to induce floor-plate cells in vitro.

To compare the inductive ability of the early prechordal mesoderm with that of the notochord in vivo, a standard assay of floor-plate induction was performed. Here, either notochord or prechordal mesendoderm were grafted adjacent to the forming neural tube in ovo(Fig. 6E). As previously described, notochord induced a discrete region of floor plate directly adjacent to the grafted tissue (Fig. 6G). By contrast, prechordal mesendoderm induced floor-plate character not only adjacent to the graft itself, but also in more dorsal regions of the neural tube on both the ipsilateral and contralateral sides(Fig. 6H). This result suggests that the nascent prechordal mesoderm is a more potent inducer of floor plate character than is notochord, and supports the idea that in vivo, early floor plate cells are induced by a rapid vertical induction mediated by underlying prechordal mesoderm.

Given the requirement for prechordal mesoderm in anterior floor plate induction, we assessed the early axial mesodermal expression of Shhand Nodal, the factors most strongly implicated in floor plate induction in amniote and anamniote embryos respectively. In situ hybridisation at HH stage 4 reveals that neither Shh or Nr1 is expressed in tissues underlying `area a' at this stage of development(Fig. 7A,B). Coincident with the appearance of the nascent prechordal mesoderm at HH stage 4+, however,both Shh and Nr1 are expressed in prechordal mesoderm cells as they pass beneath `area a' (Fig. 7C,D). In notochord cells that follow immediately behind,expression of Nr1 is completely absent while Shh is expressed only very weakly in a subset of cells(Fig. 7E,F). Thus, when `area a' cells are being specified to a floor-plate fate, co-expression of Shh and Nodal is detected in the prechordal mesoderm cells lying directly underneath them. Subsequent to their transient exposure to Shh/Nr1-expressing prechordal mesoderm, `area a'-derived cells themselves begin to express Shh, while underlying notochord cells express Shh at barely detectable levels. Given our observation that prechordal mesoderm can rapidly specify `area a' cells to a floor-plate fate we therefore tested the ability of both Shh and Nodal to specify `area a'cells to a floor plate fate in vitro.

HH stage 4 `area a' explants that do not express floor-plate markers if cultured alone (Fig. 3B-D; Fig. 7K,P,U,Z) were exposed to Shh, Nodal or a combination of the two signalling molecules. Expression of the ventral midline markers Shh (mRNA and protein), HNF3β and Netrin1, and of the axial mesoderm markers brachyury and 3B9 was assessed at different time points (12, 20, 40 hours). Neither brachyury nor 3B9 was induced (not shown). However, addition of Shh protein at high concentration to `area a' explants was sufficient to induce all three ventral midline markers after 20 hours in culture(Fig. 7G,L,Q,V; 100%, n>40), while tenfold lower concentrations were insufficient to elicit this response (Fig. 7H,M,R,W; 0%, n>40). However, when explants were exposed to low concentrations of Shh together with Nodal protein, a strong induction of ventral midline markers was observed, again, after 20 hours in culture (Fig. 7J,O,T,Y; 100%, n>40). Although a weak induction of ventral midline markers was observed when Nodal protein was provided alone(Fig. 7I,N,S,X; 50%, n>40) the response to a combination of Shh and Nodal was notably robust, in many cases more so than the response to Shh alone. Induction in response to Nodal alone, or Nodal and Shh was first detected at the 20 hour time-point. To examine whether we could distinguish a differential induction of HNF3β and Shh in response to either Nodal or Nodal/Shh, a subset of explants were examined by double-labelling. HNF3β and Shh were induced to an identical extent (not shown). Taken together, these data are suggestive of a cooperation between Nodal and Shh signalling during the rapid induction of floor-plate character in `area a' cells by nascent prechordal mesoderm at HH stage 4+.

The floor plate is common to all vertebrate embryos, possibly with an evolutionary origin earlier in the chordate lineage(Corbo et al., 1997). The importance of this structure in the regulation of CNS patterning has made it the focus of a large number of studies. In recent years, studies into floor plate formation in different model organisms have revealed apparent differences in the mechanism of floor-plate formation. The studies suggest that floor-plate induction requires the presence of the notochord and Shh in amniote embryos, but that neither is required for the differentiation of medial floor plate cells in anamniote embryos, which instead are dependent on Nodal signalling. The studies we describe here go some way towards reconciling these different observations, suggesting that in the early chick embryo, the floor plate arises from at least two principal sources. Our experiments show that an epiblast-derived population of floor plate precursors, `area a',primarily populate medial cells in the anterior ventral midline and suggest that this population is rapidly induced to a floor-plate fate early in gastrulation. Our studies suggest that the nascent prechordal mesoderm, and not the notochord, is responsible for the rapid vertical induction of floor-plate character in `area a'-derived cells, and suggest that `area a'-derived cells are left behind in a specified state as they extend along the ventral midline of the neural plate and are not reliant upon further inductive signals from the earliest forming notochord. Strikingly, in vitro experiments also suggest that, as in anamniotes, Nodal signalling may indeed play a role in floor plate induction in the chick embryo.

Previous fate-mapping studies have shown that in the HH stage 4 chick embryo, epiblast cells in `area a' contribute to the floor plate(Schoenwolf and Sheard, 1990). However, these studies did not analyse whether, prior to populating the midline, `area a' cells might transiently populate Hensen's node. Our in vivo time lapse analyses reveal no evidence for this possibility: we do not observe that `area a' cells enter the node. Previous lineage analyses have shown that,prior to HH stage 4, Hensen's node cells do not give rise to floor plate,suggesting in turn that `area a'-derived cells do not themselves migrate out of Hensen's node (Selleck and Stern,1991; Lopez-Sanchez et al.,2001). Together, these results suggest that `area a' and Hensen's node cells are distinct populations. Importantly, this separation demonstrates that a shared lineage between notochord and floor plate cells is not a prerequisite for floor plate differentiation.

Our real-time analysis of cell movement also reveals that, although floor plate precursors are actually present in the node at HH stage 4, they do not migrate out until HH stage 6 (see also Selleck and Stern, 1991; Lopez-Sanchez et al., 2001). Together these studies suggest the existence of two early populations of floor-plate precursors in the chick, one in the prenodal epiblast (`area a'),which gives rise exclusively to cells of the neural midline, principally in anterior regions, and one in the epiblast layer of Hensen's node, the descendants of which leave the Node only after HH stage 6 and are later found in the more posterior ventral midline. Our observation that cells in `area a'form an earlier floor plate population than cells in the Node is supported by previous studies in the chick(Lopez-Sanchez et al., 2001)and raises the possibility that `area a' cells, or their progenitors, exist as a specialised population prior to formation of the organiser. In support of this, even before primitive streak formation, the midline of the epiblast exhibits specialised properties, and itself undergoes powerful anterior extension movements (Kelly et al.,2002; Lawson and Schoenwolf,2001).

Our analyses show that `area a' cells are induced to a floor-plate identity between HH stage 4 and 4+, and suggest that the nascent prechordal mesoderm mediates this induction. Ablation of the mesendoderm as it forms in the deep layers of Hensen's node leads to the loss of the entire floor plate at early stages of development, including floor-plate cells that normally arise from`area a'. This ablation removes both prechordal mesoderm and notochord precursor cells; thus, in principle, either of these could contribute to the induction of `area a' cells to a floor plate identity. However, a number of lines of evidence suggest that prechordal mesoderm, and not notochord cells,are responsible for `area a' induction. First, anterior floor plate cells form normally in embryos in which prechordal mesoderm is present, but notochord is absent. The differentiation of these cells is dependent upon only a very short exposure to the prechordal mesoderm: in embryos in which prechordal mesoderm is eliminated after only a short exposure to `area a' cells, early floor-plate cells still form, albeit fewer in number and expressing lower levels of Shh than normal. It is likely that homeogenetic lateral induction mediated by these early specified cells accounts for the complete rescue of the anterior floor plate in the absence of notochord signals(Placzek et al., 1993). Second, in an ectopic situation the prechordal mesendoderm can induce floor plate in neural tissue with marked potency, supporting the assertion that in vivo, prechordal mesoderm mediates the rapid induction of `area a' cells. Finally, we find that the chick nodal homologue Nr1 is co-expressed with Shh in the nascent prechordal mesoderm at the time at which this tissue is required for the rapid induction of floor plate character in `area a' cells. By contrast, nascent notochord cells do not express Nr1 and barely express Shh.

Many lines of evidence have suggested that in zebrafish, Nodal signalling is required for medial floor plate formation early in development. Both cyc and oep mutant phenotypes include a loss of medial floor plate cells (Hatta, 1992; Hatta et al., 1991; Krauss et al., 1993; Rebagliati et al., 1998; Sampath et al., 1998; Schier et al., 1997; Shinya et al., 1999; Strahle et al., 1997; Zhang et al., 1998). Importantly, the cell-autonomous requirement for oep indicates that the formation of the medial floor plate occurs as the result of an inductive interaction (Gritsman et al.,1999; Strahle et al.,1997). In addition to loss of the medial floor plate, both cyc and oep have defects in prechordal plate formation, in the case of oep a complete loss of this tissue(Schier et al., 1997). This correlation may suggest that the prechordal plate is in fact the source of a floor plate-inducing Nodal signal during gastrulation, a possibility supported by the fact that rescue of the floor-plate phenotype in cyc mutants requires the presence of wild-type cells within the prechordal plate(Sampath et al., 1998).

Our analyses provide a first indication that Nodal may play a role also in chick floor-plate induction: Nodal can cooperate with low levels of Shh to induce floor plate character in `area a' cells. Studies of the zebrafish have suggested that Nodal can induce Shh expression within the neural tube, providing a potential mechanism of cooperation(Muller et al., 2000). Whether such a cooperation does in fact operate in vivo in the chick remains unclear,but is indicated through studies of mouse embryos: both Shh-null mice and mice that are conditionally mutant for Smad2, a downstream effector of Nodal signalling, lose Shh-expressing cells in the anterior neuraxis (Chiang et al.,1996; Heyer et al.,1999). Taken together, these data are suggestive of a cooperative role for Shh and Nodal signalling during floor plate formation in amniote embryos, potentially via Shh activation.

Our studies, together with earlier work(Artinger and Bronner, 1993; Placzek et al., 2000; van Straaten and Hekking,1991; Yamada et al.,1991) suggest that floor-plate cells are induced by two different mechanisms along the anteroposterior axis of the chick embryo. Floor-plate cells that derive from `area a' arise during gastrulation and largely populate anterior regions of the neuraxis. Our studies show that `area a'-derived floor plate cells are induced through a rapid interaction mediated by prechordal mesoderm that may involve a previously unrecognised role for Nodal signalling in an amniote embryo. By contrast, floor-plate cells that differentiate in the neurula stage embryo, and occupy posterior regions of the neuraxis, require a prolonged period of contact with underlying notochord(Fig. 8). Studies in both mouse and zebrafish embryos support the idea that distinct mechanisms operate to specify the floor plate in anterior and posterior regions. Distinct cis-acting regulatory sequences have been identified within the mouse Shh promoter that direct Shh expression to specific regions of the neural tube, supporting the view that multiple genes are involved in activating Shh transcription along the length of the CNS (Epstein et al.,1999). In zebrafish flh and ntl/spt double mutants, the anterior floor plate develops normally but the posterior floor plate is severely affected (Amacher et al.,2002; Halpern, 1995; Schier et al., 1997). Similarly, while ntl acts a partial suppressor of the oep or cyc phenotypes, rescue is observed only posteriorly (Halpern et al.,1997; Schier et al.,1997; Strahle et al.,1997).

An unresolved issue is whether the distinction between these two apparently different schemes is absolute. Intriguingly, studies have shown that posterior floor-plate cells eventually form even after notochord ablation or neural plate isolation in culture (Artinger and Bronner Fraser, 1993), raising the possibility that the induction mechanism leading to anterior floor-plate specification may in some way contribute to floor plate differentiation in the posterior neuraxis. Studies in zebrafish have likewise indicated an additive effect of floor-plate phenotypes: in oep mutants, a few floor-plate cells are present,while in flh mutants the floor plate seems normal anteriorly but scattered posteriorly. By contrast, double mutants show a complete absence, or very severe reduction in number, of floor plate cells(Schier et al., 1997; Strahle et al., 1997).

Our observations raise the question of why the floor plate should arise in this dual manner. A likely explanation is that it occurs because of the different modes of cellular movements in the gastrula and neurula embryo. Early in development, the rapid morphogenetic movements associated with gastrulation and formation of the early neural tube mean that the registration of floor plate and underlying axial mesoderm is not stable(Dale et al., 1999; Woo and Fraser, 1995). By contrast, during neurulation, the ventral midline of the caudalmost neural tube is formed in register with the notochord, so that floor-plate cells arise through the interaction of two stably apposed tissues. It is likely, then,that the functional significance of rapid specification of `area a' cells by the early prechordal mesoderm is to circumvent the requirement for prolonged exposure to a Shh-expressing notochord until such time as stable tissue interactions and Shh expression are re-established in more posterior regions of the embryo following gastrulation.

Finally, the early specification of a population of floor plate cells by signalling from the prechordal mesoderm suggests parallels with anamniote embryos. Our observations that `area a'-derived cells occupy a medial position in the developing anterior floor plate, and that a floor plate is able to develop in the absence of notochord signalling, contingent upon early specification of `area a' cells by the nascent prechordal mesoderm suggest similarities between early floor-plate specification in amniotes and the generation of the medial floor plate in anamniote embryos. In addition, our observation that Nodal signalling may be responsible for mediating this rapid induction of a population of floor-plate cells indicates further parallels with the situation in anamniote embryos. Thus, our studies may go some way towards reconciling models of floor-plate formation in different vertebrate systems.

We thank Simon Ellis for help in early experiments, Liz Illet for technical help and Stephen Szabo for illustrations. We are grateful to Andrew Furley,Sandrine Soubes and the reviewers for helpful comments on the manuscript. This work was supported by the Medical Research Council of Great Britain, the Wellcome Trust and the Yorkshire Cancer Research Trust (to M.P.), by the NIH(to S.E.F. and M.M.S.) and Computational Molecular Biology at the Burroughs-Wellcome Foundation (to P.K.).