Dual origin of the floor plate in the avian embryo

The early neural tube of the vertebrate embryo is formed of two lateral regions in which neurogenesis proceeds, which are separated by medial structures located in the midsagittal plane of the body, the dorsal roof plate and the ventral floor plate. In the avian embryo, where the morphogenetic movements through which neurulation takes place have been analysed using cell marking techniques, it is established that the lateral regions of the neural tube (i.e. the alar plates dorsally and the basal plates ventrally), together with the roof plate and the neural crest, have an embryological origin that is different from that of the floor plate (Catala et al., 1995; Catala et al., 1996; Teillet et al., 1998a; Le Douarin, 2001). Selective cell labelling by the quail-chick chimera system revealed that the ventral midline cells of the neural tube, and the notochord, are derived from Hensen’s node: the avian organiser. During the posterior elongation of the embryo, Hensen’s node moves caudally and the active proliferation of its constitutive cells generates a medial cord of cells that are left in its wake. This cell cord promptly splits into two components: a dorsal one, which is progressively inserted within the prospective neural plate and thus becomes part of the neurectoderm as the floor plate; and a ventral one, which generates the notochord and becomes part of the mesodermal germ layer. Underneath, firmly adhering to the notochord during early neurulation, lies a strand of endodermal cells that accompanies Hensen’s node in its rostrocaudal movement.

In quail-chick chimeras in which chick Hensen’s node has been replaced by its quail counterpart at the five- to six-somite stage (5-6 ss), the quail node-derived floor plate exhibits a characteristic polarised epithelial organisation with basal nuclei. This structure is distinct from the pseudo-columnar neuroepithelium derived from the neural plate itself (Catala et al., 1996) (see Fig. 2).

The floor plate of the vertebrate neural tube plays a major role during neurogenesis as it is a source of the morphogen Sonic hedgehog (SHH), a secreted glycoprotein essential for motoneurone and interneurone specification (Ericson et al., 1996; Briscoe and Ericson, 1999; Briscoe and Ericson, 2001; Lewis and Eisen, 2001) and for oligodendrocyte differentiation (Orentas and Miller, 1996; Poncet et al., 1996; Pringle et al., 1996). In addition, the floor plate provides guidance cues for outgrowing axons of many neurones (for reviews, see Colamarino and Tessier-Lavigne, 1995; Stoeckli and Landmesser, 1998) (Matise et al., 1999) and participates in directing neurone migration (de Diego et al., 2002).

In the avian embryo, the floor plate and notochord cells, as well as Hensen’s node (from which they are derived), express a gene of the Forkhead family of transcription factors, HNF3β, together with the gene encoding SHH (Charrier et al., 1999). Although Shh is transiently downregulated in the floor-plate cells as they become segregated from the notochord [over the length of a few prospective somites immediately rostral to Hensen’s node at the trunk level (see Marti et al., 1995b)], it is considered that these two gene activities characterise the ventral midline cells of the developing embryo (i.e. the floor plate, the notochord and the cells that constitute Hensen’s node) (Echelard et al., 1993; Ruiz i Altaba et al., 1993).

Grafting experiments carried out in the chick embryo have indicated that floor-plate characteristics can be induced in the lateral neural tube by signals arising from the notochord or the floor plate itself (van Straaten et al., 1985; van Straaten et al., 1988; Placzek et al., 1990; Placzek et al., 1991; Yamada et al., 1991; Pourquié et al., 1993). In vitro cultures have indicated that this process, as well as the development of motoneurones, is mediated by SHH, which is produced during neurulation by both these structures. Indeed, HNF3β and a motoneurone-specific marker Islet1 are expressed in explants of the 10 ss chick posterior neural plate subjected to a culture medium containing the SHH protein (Marti et al., 1995a; Roelink et al., 1995; Ericson et al., 1996). Moreover, in vivo ectopic Shh expression can induce the expression of various floor-plate markers in zebrafish (Krauss et al., 1993), mouse (Echelard et al., 1993) and Xenopus embryos (Ruiz i Altaba et al., 1995a).

The description of spatially restricted expression of various immunocytochemical and molecular markers in the ventral neural tube of several vertebrate species has led to the distinction of two cell populations in the floor plate. Thus, in the chick ventral neural tube at incubation day 3 (E3), a medial region where the cells express both SC1 and FP1 antigens can be distinguished from lateral areas where the cells express FP1 but not SC1 (Placzek et al., 1991; Yamada et al., 1991). In rat embryos, the antigen FP3 is expressed in all floor-plate cells, while FP4 is restricted to medial cells (Placzek et al., 1993; Roelink et al., 1994). Moreover, in mouse and rat embryos, Shh is expressed only in the medial cells (Roelink et al., 1994), whereas HNF3β transcripts are present in a larger region of the ventral neural tube (Monaghan et al., 1993; Sasaki and Hogan, 1993; Ang and Rossant, 1994).

In the zebrafish, the medial floor plate (MFP) consists of a single row of cells flanked on each side by one or two additional rows of lateral floor-plate (LFP) cells (Odenthal and Nusslein-Volhard, 1998; Odenthal et al., 2000). MFP and LFP, although belonging to the same cuboido-epithelial type, differ in their gene expression patterns. Cells of the MFP express Netrin1 (Strähle et al., 1997) and members of the Hedgehog (Hh) family: Shh (Krauss et al., 1993) and tiggy-winkle hedgehog (twhh) (Ekker et al., 1995). They also express several forkhead family members: axial and fkd7 (Strähle et al., 1993; Strähle et al., 1996; Odenthal and Nusslein-Volhard, 1998), and the Xenopus Pintallavis homologue fkd4 (Odenthal and Nusslein-Volhard, 1998). Netrin1, axial and fkd4 transcripts are in addition present in the LFP (Odenthal et al., 2000). Interestingly, in the zebrafish, the Hh paralogues exhibit a dynamic expression in the developing midline. In the early gastrula, both Shh and twhh are expressed in the organiser region (Ekker et al., 1995), but later on, deep midline cells of the embryonic axis fated to become notochord express only Shh, whereas the overlying cells (i.e. the future floor plate) retain only twhh expression (Etheridge et al., 2001). Although Shh is re-expressed later in the zebrafish floor plate, as it is in other vertebrates, the phenotypic analysis of mutations of this gene (Schauerte et al., 1998; Odenthal et al., 2000) or downregulation experiments (Etheridge et al., 2001) indicate that Shh is not required for development of the MFP. Furthermore, twhh is not required for floor-plate development either (Etheridge et al., 2001). However, the Hh pathway seems to be necessary for the formation of the LFP. Thus, in sonic-you (syu) zebrafish mutants lacking the shh gene, LFP cells are absent (Schauerte et al., 1998). In addition, the gene smoothened, part of the Hh receptor machinery, is required for induction of LFP (Varga et al., 2001).

We aimed to explore the possible existence of a lateral expansion of the floor plate in the neural epithelium of the avian embryo. We show that the medial floor plate, as defined by its origin from Hensen’s node, induces a lateral floor plate in which not only floor-plate markers, such as HNF3β and Shh genes, but also genes that are normally activated in the neural ectoderm, such as Sox1, are expressed. We also observed that, during a short time window, the newly induced neural ectoderm can acquire MFP characteristics under the influence of an exogenous notochord or MFP. By contrast, SHH alone can only induce a LFP, not a MFP.

Chick and quail embryos were used as hosts or donors in various in ovo grafting experiments. They were staged according to the number of somites (ss) or embryonic day (E). Operated embryos were sacrificed between E2.5 and E8.

In a first series of experiments, Hensen’s node cells located at the level of the median pit [zone b (see Charrier et al., 1999)] were excised micro-surgically in 5-6 ss chick embryos and replaced by their quail counterpart as previously described (Catala et al., 1996). In another series, the presumptive territory of the neural plate, immediately posterior to the median pit, was replaced by its quail counterpart, while Hensen’s node cells remained untouched. These experiments (see Fig. 2E) resulted in the differential quail labelling of the floor plate and of the basal plate neural epithelium, respectively, as previously shown by Catala et al. (Catala et al., 1996).

Notochords and neural tubes were enzymatically dissociated from E2 (10-25 ss) quail embryos using pancreatin. Floor-plate fragments were dissected from the dissociated neural tubes at the thoracic and cervical levels. Notochord and floor-plate fragments (the latter comprising MFP and LFP components) were grafted between the neural epithelium and the segmental plate in 7-25 ss chick embryos, as described by Pourquié et al. (Pourquié et al., 1993). Clumps of QT6 quail fibroblasts, stably transfected with a construct carrying the chick SHH-coding region (Duprez et al., 1998), were implanted into chick hosts in the same situation.

The caudalward movement of Hensen’s node can be prevented by removing its posterior-most region together with the rostral tip of the primitive streak forming the axial-paraxial hinge (APH). Under these conditions, the posterior neural tube develops without midline cells (notochord and floor plate) and abundant apoptosis was observed in the neural epithelium as well as in the paraxial mesoderm of the affected region because of the absence of the sources of SHH, i.e. floor plate and notochord (Charrier et al., 1999; Charrier et al., 2001). APH excisions were performed in 5-6 ss quail embryos (Fig. 1A). One day later, (Fig. 1B-D), caudal neural tubes deprived of node-derived cells were isolated by enzymatic dissociation and transplanted into stage-matched chick embryos, after microsurgical excision of a fragment of their own truncal neural tube and notochord (Charrier et al., 2001). Quail neural tubes were transplanted either alone or together with a fragment of E2 chick notochord or floor plate, or with clumps of SHH-producing cells (Fig. 1D, parts a-c).

In situ hybridisation on whole mount or on alternate 7 μm serial sections was performed as described previously (Charrier et al., 1999) using probes for the following chicken genes: HNF3β (Ruiz i Altaba et al., 1995b), Shh (Riddle et al., 1993), Netrin1 (Kennedy et al., 1994), CSox1 (Rex et al., 1997), Pax6 (Goulding et al., 1993), Sim1 (Yamada et al., 1991), Nkx2.2 (Ericson et al., 1997). In each case, adjacent sections were treated overnight with the monoclonal antibody (mAb) QCPN (Developmental Studies Hybridoma Bank), which specifically recognises quail cells. Mounted sections were photographed using Nomarski optics (Leica).

When Hensen’s node is replaced in 5-6 ss chick embryo by its quail counterpart (Fig. 2E, red), floor-plate and notochord cells are of quail origin (Fig. 2H,L) from the level of the graft (thoracic level) down to the posterior end of the spinal cord. By contrast, when a region of the prospective neural plate posterior to the node region is replaced by its quail equivalent (Fig. 2E, blue area), cells of the basal plate, but not of the floor plate, carry the quail marker (Fig. 2P). These experiments demonstrated that floor-plate and basal plate cells, in the avian neural tube, have distinct embryological origins (Catala et al., 1996). We have now recorded the dynamic expression patterns of several genes expressed in the ventral neural tube in both types of quail-chick chimeras.

At E3.5, one day after Hensen’s node replacement, Shh and HNF3β were strongly co-expressed both in the node-derived cells and in a fringe of cells deriving from the presumptive neural plate (Fig. 2F-H). From E4 to E5, the HNF3β+ territory became progressively restricted to the node-derived cells, whereas Shh gene remained activated both in the node-derived cells, where it was strongly expressed, and in adjacent neural plate-derived cells, where it was only weakly expressed (not shown). This composite character of the floor plate, as defined by Shh expression, was even more obvious at later stages. At E7, the Shh-expressing floor plate was clearly formed by (1) a medial component originating from Hensen’s node that expressed HNF3β and (2) two lateral areas derived from the original neural plate that are now devoid of HNF3β transcripts (Fig. 2J-L). It was noticeable that, in the Hensen’s node-derived part, the floor plate was made up of polarised cylindrical epithelial cells, whose nuclei were situated in a basal position (Fig. 2L). This cellular arrangement remains typical of the medial floor plate (MFP) throughout development. By contrast, the neural plate-derived floor plate was made up of a pseudostratified epithelium like the rest of the ventricular epithelium (except for the roof plate). The same observations were made whether the MFP was derived from the grafted or from the host node (Fig. 2J-L,N-P).

CSox1, the expression of which correlates with the formation of the early neural plate (Rex et al., 1997) and which has been shown in vitro to be implicated in neural determination and differentiation (Pevny et al., 1998), was not detected in the medial node-derived cells, but was found in lateral neural plate-derived cells of the floor plate, as well as in the rest of the ventricular epithelium derived from the neural plate (Fig. 2I). Interestingly, in situ hybridisation at earlier stages showed that CSox1 was never expressed in the node and node-derived cells, even at later stages when the node becomes the chordo-neural hinge (CNH) (see Fig. 2B-D). By contrast, CSox1 was highly expressed in the neural epithelium (Fig. 2A-C) and to a lesser extent in the neural plate caudal to the node (not shown). Moreover, from E5 to E7, transcripts of the homeobox transcription factor Nkx2.2 (Ericson et al., 1997) were found in a region corresponding to the Shh+, HNF3β– neural epithelium-derived area, that we consider as a lateral floor plate (LFP) (Fig. 2M). At the same time, Sim1 transcripts (Yamada et al., 1991) were confined to two groups of cells situated in the external area of the Nkx2.2 territory in the mantle (not shown). Moreover, Netrin1 (Kennedy et al., 1994) transcripts were abundant in the MFP and decreased laterally through the LFP areas and beyond in the ventral neural tube (see Fig. 4K, control right side of the neural tube and Fig. 8).

Thus, between E3.5 and E7, molecular analysis of the Shh+ territory of the ventral neural tube of quail-chick chimeras allows us to distinguish a medial, node-derived region, the MFP, which is HNF3β+, Netrin1+, CSox1–, Nkx2.2– and Sim1– from lateral, neural plate-derived areas, the LFP, which are CSox1+, HNF3β transiently+, Nkx2.2+, Sim1+ and Netrin1+. Given this distinction, we examined the molecular characteristics of supernumerary floor plates induced by grafts of notochord, floor-plate or SHH-producing cells in the avian embryo in ovo.

Many authors, over the past decade, have shown that a fragment of notochord grafted in contact to the neural tube in E2 chick embryos can induce floor-plate-like characteristics in the neural epithelium (van Straaten et al., 1985; van Straaten et al., 1988; Placzek et al., 1990; Placzek et al., 1991; Placzek et al., 1993; Yamada et al., 1991; Pourquié et al., 1993). However, such an induction was not observed in 100% of the cases (our own observation) (van Straaten et al., 1988), and the variability of the restuls has not been completely explained. We decided to explore the response of the neural epithelium to contact with an inducing tissue in terms of gene activity.

Notochords were dissected enzymatically from 10-21 ss quail embryos. Notochord fragments two to three somites long were implanted in 7-23 ss chick embryos (E2), between the neural epithelium and the segmental plate in the region immediately rostral to the endogenous node. During the first day after the operation (up to E3), only discreet molecular alterations could be detected in the host neural tube but obvious dorsoventral overgrowth of the tube wall was noticed on the side of the graft (not shown). Two days after the graft (E4), the asymmetry of the neural tube had increased. Moreover, the region of the neural epithelium in contact with the grafted notochord was thinner than the rest of the neural tube wall (Fig. 4A-D). Transcripts of Pax6, a lateroventral marker of the neural tube (Goulding et al., 1993), were not found in this region, while Shh and HNF3β began to be expressed (Fig. 4A-C).

From E5 to E7 (3 to 5 days post operation), MFP and LFP characteristics appeared progressively in the induced territory (n=4/7, see Fig. 3) (Fig. 4E-L): HNF3β transcripts were restricted to a medial region (Fig. 4F,J) strongly expressing Shh (Fig. 4E,I); Shh was weakly expressed in a lateral domain where Nkx2.2, Sim1 and CSox1 transcripts were present (Fig. 4G,H,L); CSox1 expression progressively disappeared in the HNF3β+ domain between E6 and E7 (Fig. 4L). Moreover, the chemotropic factor Netrin1, which is known to be expressed in a ventral domain broader than the floor-plate territory (Kennedy et al., 1994) was also present in the region of the notochord-induced floor plate (Fig. 4K).

Thus, a region of the neural epithelium placed in contact with a notochord, can progressively acquire the molecular MFP characteristics (i.e. be HNF3β +, Shh+ and CSox1–) in addition to the LFP traits (i.e. be Shh+, HNF3β transiently+, CSox1+, Nkx2.2+ and Sim1+). However, induction of the MFP-like structure was generally observed only over a short length of the grafted notochord. Moreover, floor-plate induction, as already observed by van Straaten et al. (van Straaten et al., 1988), occurred only when the graft was performed before 15 ss in the most caudal region of the embryo (see Fig. 3). These observations suggest that there is a narrow time window when this full transformation of a neural epithelium into a MFP can take place. By contrast, induction of LFP characteristics is widespread over the full length of the caudal neural tube subjected to induction from the grafted notochord, at any stage of the chick host up to 25 ss. HNF3β and Shh induction by exogenous notochord requires at least 2 days of exposure. Moreover, MFP complete differentiation necessitates 4-5 days of exposure and occurs on a shorter length than LFP phenotype.

Exposure of E2 (9-25 ss) chick posterior neural tube to E2 (12-25 ss) quail floor plates produced various results, generally depending on the developmental stage of both donor and host tissues (see Fig. 3). For 9-14 ss hosts that received a fragment of floor plate from the cervical or thoracic region of 20-25 ss donors, a dorsoventral overgrowth of the lateral neural tube was seen on the side of the graft one day after the graft (i.e. at E3; not shown). Two days after the graft (E4), transcripts of Shh and HNF3β were present in the neural epithelium immediately opposite to the floor-plate graft (n=3/3) (Fig. 5A,B). At this stage, Shh and HNF3β expression co-existed with that of CSox1 and Nkx2.2, the latter of which was induced by the graft (Fig. 5C,D). Pax6 was not expressed in this region (not shown). Five to 6 days after the operation (E7-8), a typical MFP area (HNF3β +, Shh+, Netrin1+, CSox1–, Nkx2.2–) could be observed in four out of 12 cases. The MFP-like structure differentiated generally over a short length (about 50 μm) in embryos operated before 15 ss (see Fig. 3). Only grafts in which donor MFP was in close contact to the host neural epithelium were efficient in MFP induction (see Fig. 5E-H).

In conclusion, notochord and medial floor plate, both of which are derived from Hensen’s node, are able to induce a supernumerary floor-plate-like structure in the neural epithelium, in which the morphological and molecular characteristics of the natural MFP and LFP can be detected. MFP induction can take place only in the posterior-most region of the neural plate of embryos younger than 15 ss.

Grafts of clumps of cells engineered to produce SHH (Duprez et al., 1998) yielded similar results to the grafts of notochord or floor-plate fragments after the first day of exposure (Fig. 6A-C). Later, the grafted cells dispersed and the field resulting from the induction was larger than with notochord and floor-plate grafts. As a result, after 3 days or more of SHH-exposure (E5-E7), the lateral floor plate of the host embryo was widely enlarged on the side of the graft, as seen in Fig. 6D-K showing the expression pattern of several LFP markers including not only Shh and Netrin1 but also Nkx2.2, Sim1 and the neural marker CSox. A characteristic MFP devoid of neural markers was never observed.

The caudalward movement of Hensen’s node was prevented by excising the APH in 5-6 ss quail embryos (see Fig. 1). One day after the excision, the midline structures (floor plate and notochord), and consequently the source of SHH in the medial ectodermal and mesodermal layers, were absent. Under these circumstances, neither the neural tube nor the paraxial mesoderm can survive (Charrier et al., 1999). Many of their constitutive cells undergo apoptosis within the 24 hours that follow the operation. This cell death can, however, be prevented by the experimental addition of cells producing SHH in close proximity to the axial tissues, thus showing that one of the major roles of SHH emanating from the floor plate and/or notochord is to antagonise the built in programmed cell death of the embryonic tissues that form the neural tube and the somites (Charrier et al., 2001; Teillet et al., 1998b).

Such midline cell-deprived neural tubes taken at E2.5 from operated quail embryos were transplanted into 20-25 ss chick embryos in a region from which endogenous neural tube and notochord had been previously removed (Fig. 1D). Transplantation was made in three different situations: (1) together with a chick notochord; (2) with a chick floor plate; (3) with clumps of SHH-producing cells, as already described by Charrier et al. (Charrier et al., 2001). Observation of the transplanted neural tube the first day after the operation showed a decrease in apoptosis in the three situations when compared with a graft in the absence of these tissues (Charrier et al., 2001). No transcripts of the floor-plate markers Shh and HNF3β were then observed, although Pax3 and Pax6 transcripts, previously present in the entire neural tube, were downregulated ventrally in contact with the notochord, floor-plate or SHH-producing cells [see Fig. 2 by Charrier et al. (Charrier et al., 2001)]. However, two days after the graft (E4), Shh and HNF3β transcripts were present in the region facing the notochord, the floor-plate or the SHH-producing cells, while CSox1 was still uniformly expressed in the transplanted neural tube (not shown).

At E7 (5 days after the graft), when the inducer was a notochord, a complete floor plate with its medial and lateral components (MFP and LFP) could be observed in the region facing the graft, although over a short length (Fig. 7A-D). The induced MFP exhibited the typical structure of a columnar highly polarised epithelium (Fig. 6A), like a node-derived floor plate in normal development. It expressed Shh, HNF3β and Netrin1, and failed to exhibit transcripts of the neural marker CSox1 (not shown). Moreover, on each side of this MFP-like structure, lateral areas expressing Shh, CSox1, Netrin1 (not shown) and Nkx2.2 (Fig. 7D) were present.

When the inducer was either the floor plate (Fig. 7E-H) or SHH-producing cells (Fig. 7I-L), the ventral neural tube was incompletely patterned: CSox1 (not shown) continued to be expressed ventrally in a region where Shh and Nkx2.2 were co-expressed and where HNF3β became downregulated, thus showing the molecular characteristics of a LFP.

We demonstrate in this work that, during normal development of the avian embryo, the floor plate, an epithelial structure located ventrally in the neural tube, is heterogeneous and composed of regions that can be distinguished on the basis of their embryological origin and molecular characteristics. The medial region of the floor plate (MFP) is formed by cells originating from the avian organiser, Hensen’s node. During its rostrocaudal movement, which is concomitant with the posterior elongation of the embryo, the node leaves in its wake two closely associated axial structures: the floor plate that becomes inserted into the ectoderm of the neural plate; and the notochord that becomes part of the mesoderm (Catala et al., 1996). These two structures are located midsagittally in the embryo. Data obtained previously (Charrier et al., 1999; Le Douarin and Halpern, 2001; Le Douarin, 2001) support the hypothesis that the notochord and floor plate arise from a population of pluripotent cells present in Hensen’s node, where they might function as stem cells. At midneurulation (i.e. 5-6 ss) these cells occupy the posterior end of Hensen’s node (zone c) [see Fig. 11 by Charrier et al (Charrier et al., 1999)]. In the zebrafish embryo, a common pool of midline precursor cells has also been hypothesised in the early organiser region designated as the shield (Halpern et al., 1997; Appel et al., 1999).

Lateral to the node-derived MFP, cells of the neural ectoderm acquire molecular and functional characteristics that allow them to be considered as lateral floor-plate cells (LFP), analogous to the zebrafish LFP cells (Odenthal et al., 2000).

The mechanisms underlying the neuralisation of the dorsal ectoderm of the vertebrate embryo and leading to the formation of the neural plate have been under investigation for almost a century (see Streit and Stern, 1999; Harland, 2000; Wilson and Edlund, 2001). According to the presently accepted view, neuralisation results from the diffusion by the dorsal organiser (Spemann’s organiser or its equivalent in various vertebrate species) of dorsalising factors that are able to antagonise ventralising influences exerted mainly by members of the TGFβ and Wnt families of secreted proteins. FGF signalling could be necessary for neural induction in avian embryo at earlier stages, before gastrulation (Streit et al., 2000).

In the avian embryo, determination of the neural plate occurs very early in development as the expression pattern of a pan-neural marker, CSox1, at the primitive streak stage (Rex et al., 1997) coincides with the fate map of the neural plate as defined by the quail-chick cell marking technique (Garcia-Martinez et al., 1993). The same exact correspondence between the CSox2 expression domain and the neural tube fate map was found at later stages of chick and quail development [see Fig. 3 by Le Douarin (Le Douarin, 2001)]. Our preceding results showed that the early neural plate, formed by a planar induction in the dorsal ectoderm, lacks a floor-plate territory. The latter is intercalated during Hensen’s node regression (Catala et al., 1996). Only after the floor plate has been incorporated into the neural ectoderm is the definitive neural plate formed. The gene activities of these two domains of the definitive neural plate are different: the neural ectoderm expresses several genes, such as CSox1, CSox2, Msx1, Msx2 and FrzB, that are not expressed in the node derived midline structures (floor plate and notochord) during neurogenesis (J.-B. C., F. L., N. M. L. D. and M.-A. T., unpublished) (Duprez et al., 1999).

The floor plate and the notochord express several genes that were activated in the organiser, such as HNF3β, Shh and Chordin [see Fig. 2 by Charrier et al. (Charrier et al., 1999)]. None of these genes are expressed in the neural ectoderm prior to the time it becomes associated with the node-derived midline structures. Moreover, when insertion of the floor plate is inhibited by caudal Hensen’s node excision, the neural tube never expresses Shh or HNF3β (Charrier et al., 1999). Contact between the node-derived medial cells fated to become the MFP and the adjacent neural ectoderm appears to result in the induction of these genes in cells flanking the MFP that we designate here as lateral floor plate (LFP).

Thus, soon after the incorporation of the node-derived cells into the neural ectoderm, HNF3β and Shh transcripts are found laterally in the neuroepithelial LFP. Similarly, Netrin1 transcripts, which encode a chemotropic secreted protein (Serafini et al., 1994) are, like those of Shh, abundant in the MFP and also expressed in the LFP. However, HNF3β, after expanding laterally in LFP territories, becomes progressively restricted to the MFP (see Figs 2, 8). The presence of CSox1 transcripts distinguishes the LFP from the MFP which in fact never displays neural traits. Moreover, from E4 onwards, the LFP starts to express Nkx2.2, a transcription factor induced by Shh and involved in ventral nervous system patterning (Briscoe et al., 1999; Pabst et al., 2000; Soula et al., 2001).

One can therefore distinguish in the avian neural tube a MFP that is CSox1–, HNF3β+, Shh+, Netrin1+ and Nkx2.2–, and a LFP that is CSox1+, HNF3β transiently+, Shh+, Netrin1+ and Nkx2.2+. In addition, LFP cells maintain the pseudostratified structure of the neuroepithelium and do not acquire the polarised morphology of the MFP.

Genetic analysis has revealed that the zebrafish floor plate is also composed of medial and lateral components whose patterning is differently regulated. The presence of MFP in sonic-you (syu) mutants showed that the shh zebrafish homologue is not involved in the formation of the MFP (Schauerte et al., 1998). By contrast, it is required for the induction of LFP cells (Odenthal et al., 2000). Experimental analysis in the zebrafish embryo was pushed one step further by the injection of morpholinos directed against tiggy-winkle hedgehog (twhh), another member of the zebrafish family of HH-producing genes, into a syu mutant (which is devoid of SHH). In these embryos, the MFP develops normally (Etheridge et al., 2001; Lewis and Eisen, 2001), thus showing that its formation does not depend on either SHH or TWHH. These results are consistent with our observations in the avian embryo, according to which a floor plate can develop independently from the notochord (Teillet et al., 1998a).

Induction of a floor-plate-like structure in the lateral neural tube of the chick embryo by notochord or floor-plate graft has been described several times in the past (van Straaten et al., 1985; van Straaten et al., 1988; Placzek et al., 1990; Placzek et al., 1991; Yamada et al., 1991; Pourquié et al., 1993). In this work we show that fragments of notochord or MFP can induce both MFP and LFP if applied to embryos at stages ranging from 7-15 ss, but not later (see Fig. 3). The full molecular characters of an induced MFP are present only after 5 days of exposure to the notochord or floor-plate graft. Moreover, the MFP is induced only over a short length of the neural tube of the host located in close vicinity to Hensen’s node at operation time. More rostrally, the graft merely induced LFP- but not MFP-type gene activities.

These results demonstrate that, although the MFP and the neural epithelium have different embryological origins in normal development, MFP can be induced in the neural ectoderm by Hensen’s node-derived tissues. However, the neural ectoderm, as it stands in 7-15 ss avian embryos, is able to respond to this induction only over a short period of time and in a region where it is itself in close proximity to the endogenous node of the recipient embryo. Induction of a MFP by an exogenous floor plate can occur only if the donor MFP is in close contact with the neural ectoderm.

In cultures of the 10 ss chick neural plate, SHH has been shown to induce floor-plate and motoneurone markers (for a review, see Roelink et al., 1995). In vivo experiments performed in Xenopus embryos produced similar results (Roelink et al., 1994). We decided to document this phenomenon further in the chick embryo in ovo. We have found that SHH-producing cells grafted in close contact to the neural tube are able, after 3 days of exposure, to induce a transient and weak expression of HNF3β and a robust and durable expression of Shh that co-exist with the presence of neural markers such as CSox1, Nkx2.2 and Sim1. In summary, the pattern of gene expression in the chick neural territory subjected to the SHH protein is characteristic of LFP but not of MFP. The most spectacular effect of SHH on the neural tube is to induce its enlargement on the side of the graft (see Fig. 6D,E). The same gene inductions were obtained in a neural tube that had developed in the absence of midline structures after APH excision.

Thus, SHH is not sufficient to fully transform neuroepithelial cells into medial floor-plate cells. BMP antagonists like chordin might be implicated in this process as suggested by recent results of Patten and Placzek (Patten and Placzek, 2002). BMP antagonists secreted by Hensen’s node derivatives (notochord and MFP) are probably not produced by SHH-producing cells.

This work suggests that, in chick, like in zebrafish, SHH is not involved in specifying the MFP itself but is essential for inducing the LFP. Experiments using Shh^–/– mutant mouse embryos (Chiang et al., 1996) and downregulated Shh chick embryos (Ahlgren and Bronner-Fraser, 1999; Britto et al., 2002) must be carried out to verify this hypothesis.

In conclusion, we have shown that the floor plate, a structure playing an important role in patterning the neural tube of the vertebrate embryo, is a composite a structure in an amniote embryo as it is in lower teleost vertebrates. It is formed by a medial component derived from the organiser, which induces the adjacent neural ectoderm to develop floor-plate markers that co-exist with neural epithelial markers (Fig. 8). The SHH protein plays a key role in inducing the lateral floor plate but cannot, by itself, induce the characteristics of a medial floor plate in the neural ectoderm. This can be achieved during a brief window of time by the MFP itself or by the notochord, meaning that floor-plate induction in experimental conditions requires factor(s) specific to the organiser or its derivatives.

We thank P. Brickell, A. Grapin-Botton, P. Gruss, R. Riddle, J. Rubenstein, A. Ruiz i Altaba and C. Tabin for the chick nucleic acid probes that they provided. QCPN monoclonal antibody was obtained from the Developmental Studies Hybridoma Bank. The authors thank Françoise Dieterlen and Heather Etchevers for critical reading of the manuscript. They are grateful to Bernadette Schuler for technical assistance, and to Sophie Gournet, Francis Beaujean and Michel Fromaget for the illustrations. The helpful comments and constructive suggestions of the referees are greatly appreciated. This work was supported by the Centre National de la Recherche Scientifique and the Association pour la Recherche contre le Cancer (grant number 5578 to M.-A. T.).