Segregating expression domains of two goosecoid genes during the transition from gastrulation to neurulation in chick embryos

Homeobox genes are involved in specifying position, fate, pattern formation and morphogenesis during embryogenesis; together these features define the’identity’ of cells or tissues. A prominent role during axis formation has been attributed to genes with a goosecoid homeobox. Goosecoid genes were found in vertebrates (Blum et al., 1992, 1994; Blumberg et al., 1991; Cho et al., 1991; Izpisúa-Belmonte et al., 1993; Stachel et al., 1993) and in Drosophila melanogaster (Goriely et al., 1996; Hahn and Jäckle, 1996). The earliest expression of the Xenopus goosecoid gene is observed in the blastula, specifically in the dorsal part of the marginal zone. The ventral part of the marginal zone, on the contrary, is characterized by over-lapping domains of the two homeobox genes Xvent1 and Xvent2 (Gawantka et al., 1995; Onichtchouk et al., 1996). While Xgsc responds to TGFβ-like, dorsalizing factors (activin A, Vg1; Cho et al., 1991; Seleiro et al., 1996), the Xvent genes respond to the ventralizing factor BMP-4 (Gawantka et al., 1995; Onichtchouk et al., 1996). Thus, the goosecoid and the two Xvent homeobox genes seem to be involved in specifying dorsal or ventral identities, respectively, in the marginal zone.

In gastrulating vertebrate embryos, goosecoid is expressed in the dorsal/anterior blastopore margin, the embryonic shield in zebrafish, the dorsal blastopore lip in Xenopus or in Hensen’s node in chick and mouse. Goosecoid-expressing cells migrate anteriorly and form the prechordal mesendoderm underlying the prospective forebrain ectoderm. The conspicuous expression patterns of the goosecoid gene has prompted studies addressing its role in the phenomenon of Spemann’s organizer (Spemann and Mangold, 1924; Cho et al., 1991; Blum et al., 1992; Izpisúa-Belmonte et al., 1993; Niehrs et al., 1993). Expression of goosecoid RNA in ventral cells leads to the formation of a second blastopore and consequently to the generation of a secondary embryo.

Gene targeting experiments in mice did not reveal any function of goosecoid in gastrulation and neurulation (Rivera-Peréz et al., 1995; Yamada et al., 1995). The goosecoid null-mutants did not show the early phenotypes, instead the observed phenotypes corresponded rather to late expression domains, such as the first branchial arch. Surprisingly late phenotypes were also observed after inactivation of other homeobox genes expressed early in embryogenesis, and were generally interpreted as indications of redundancy. Thus, Otx-2 gene inactivation resulted in head defects (Ang et al., 1996) and not early epiblast defects, the Nkx2.5 gene inactivation resulted in heart morphogenesis abnormalities and not in defects of the precardiac mesoderm (Lyons et al., 1995), and also several Hox gene inactivations resulted in relatively late abnormalities (Stein et al., 1996a).

In vertebrates, many homeobox genes are present in paralogous pairs, with homeodomains of more than 75% identity. Possibly this is a result of massive genomic duplication, which occurred in the chordate lineage after the cephalochordates (Amphioxus) and thus are found in all vertebrates (Holland et al., 1994). Several examples exist where the expression patterns of two paralogous genes overlap to a large extent, e.g. the CNOT genes (Ranson et al., 1995; Stein and Kessel, 1995; Stein et al., 1996b), the Xvent genes (Gawantka et al., 1995; Onichtchouk et al., 1996) or some Hox genes. A different combination of homeobox gene expression seems to be involved in the specification of cellular identity (Kessel and Gruss, 1991). This concept of a molecular code was derived from studies on the Hox genes, but may well be a more generally applicable principle. The duplication of information resulting from gene duplications allows an increased flexibility for the coding mechanism, leading to complexity, redundancy and safety typical for higher vertebrate development.

Here, we characterize a goosecoid-related gene in the chick. The expression of the GSX gene segregates strikingly from the expression of the previously described chicken goosecoid gene. GSX expression characterizes cells located in the middle and posterior part of the primitive streak, while GSC expression characterizes cells being located in the anterior part of the primitive streak, including Hensen’s node. GSX activity correlates in the mid-streak phase (HH stage 3) with the induction of gastrulation, and from the extended streak phase (HH stage 4) onwards with a neural fate. On the contrary, GSC activity correlates with the induction of neurulation and a mesendodermal fate. We discuss the role of both homeobox genes, GSX and GSC, at the transition from gastrulation to neurulation.

A Xenopus goosecoid cDNA fragment, corresponding to the highly conserved homeobox was labelled with [³²P] CTP by random priming and used to screen a genomic chick library generated in Lambda Fix II (Stratagene). Hybridization under low stringent conditions allowed the isolation of three independent clones. After subcloning into pBlue-script II KS and rescreening a 4853 bp EcoRI restriction fragment, containing the sequence of the entire GSX gene, and a 1618 bp ApaI restriction fragment, containing the 3′-part of the gene, were identified. The sequencing (Sanger et al., 1977) was performed with the Dye Terminator Kit (ABI) on a 377 DNA Sequencer (ABI).

Total RNA from HH stage 3/3⁺ embryos was prepared by the LiCl procedure (Auffray and Rougeon, 1980) and 5 μg were reverse transcribed (Pharmacia). The conserved exon-intron boundaries in the homeobox were confirmed by the isolation of a 163 bp and a 120 bp fragment of the spliced homeobox by RT-PCR using gene-specific primers for the amplification of cDNA derived from embryos of the intermediate primitive streak stages.

Two specific primer pairs were synthesized to amplify GSX sequences from chick RNA. Primer pair LL67/66 amplifies a 163 bp fragment of the spliced GSX homeobox: 5′-primer, CAT CGC ACC ATA TTC ACC GAG, 3′-primer, GGC TCG ACG GTT CTT AAA CCA and primer pair LL68/66 amplifies a 120 bp fragment of the spliced GSX homeobox: 5′-primer, CTG GAA ACA CTT TTC CAC CAG, 3′-primer, GGC TCG ACG GTT CTT AAA CCA. The PCR cycling parameters were 25 cycles of 94°C for 1 minute, 60°C for 2 minutes, 72°C for 1 minute and a final extension at 72°C for 10 minutes. PCR products were ligated into the pCRII vector (Invitrogen) and identified by dideoxy sequencing.

The cDNA sequence of the internal exon and the corresponding splice donor and acceptor sites were deduced from exon trapping experiments (Buckler et al., 1991). The exon trap procedure was applied to the genomic GSX sequence. By subcloning a fragment of the genomic clone into the pSPL1 vector, transfecting COS-7 cells and applying PCR amplification to first strand cDNA prepared from the resultant COS-7 cell mRNA, we obtained the internal exon, which was then subcloned into the pAMP10 vector (Gibco BRL).

Fertilized chick eggs (White Leghorn, obtained from Lohmann Tierzucht, Cuxhaven) were incubated at 37.8°C in a humidified incubator until they had reached the desired stage of development. Embryos were staged according to Eyal-Giladi and Kochav (roman numbers for preprimitive streak stages; Eyal-Giladi and Kochav, 1976) and Hamburger and Hamilton (arabic numerals after appearance of the primitive streak; Hamburger and Hamilton, 1951). Fertilized quail eggs were obtained from Heinrich Linnenschmidt, Wieden brück and quail embryos were staged using the chick tables.

Antisense GSX riboprobes were synthesized from a 1.6 kb genomic ApaI-fragment and a 163 bp fragment of the spliced GSX homeobox isolated by RT-PCR using T3 or SP6 RNA polymerase and incorporating digoxigenin-UTP (Boehringer Mannheim) according to manufacturer’s instructions. Whole-mount in situ hybridization was performed essentially as described by Wilkinson (Wilkinson, 1992), except that the hybridization and the first two washing steps were performed at 55°C in presence of 0.1% CHAPS detergent (Sigma; Stein and Kessel, 1995), and no RNase treatment was done. For paraffin sections (8 μm), stained embryos were dehydrated and embedded in Paraplast plus (Sherwood Medicals). Sense control probes were negative in all experiments.

White Leghorn chicken eggs were incubated in a humidified incubator for 12-15 hours. Embryos, together with large portions of the adjacent vitelline membranes, were removed from the eggs and prepared for a modified New culture (Stern, 1993). The blastoderm, still attached to its vitelline membrane, was oriented ventral side up in a 35 mm Petri dish on a fresh albumen substrate. A glass ring was positioned round the blastoderm and the vitelline membrane was draped over the ring. Such cultured chick blastoderms at HH stage 2-4 were used as hosts.

Chick or quail embryos were used as donors. These were removed from their vitelline membranes and placed dorsal side up in a Petri dish, using Pannett-Compton saline. Primitive streak grafts were obtained from these embryos at HH stage 2-4. Pieces of primitive streak were excised with a tungsten needle. Each excised graft was transferred with a micropipet from the donor blastoderm and placed on top of the host blastoderm. A small hole was made in the hosts hypoblast in the germinal crescent, and the graft was inserted between hypoblast and ectoderm at the level of the anterior margin of the area pellucida. The cultures were incubated for 6-18 hours.

For the generation of chick/quail chimeras, chick embryos grafted with quail tissue were fixed in Zenker’s fixative and embedded in Paraplast plus (Sherwood Medicals). Paraffin sections (8 μm) were processed for quail nucleolar histochemistry according to the Harris’s haematoxylin staining described by Stern (1993).

Embryos processed with the whole-mount procedure were viewed with a Zeiss Stemi SV 11 using a combination of reflected light from two fiber optic sources and transmitted illumination. Sections through the whole mounts were viewed and photographed using Nomarski differential interference contrast optics. Photographs were taken on Fuji 64T tungsten-balanced color film.

By screening a chick genomic library under low stringent conditions with a Xenopus goosecoid cDNA fragment comprising the entire homeobox, we isolated a goosecoid-related gene, GSX, and characterized the locus by RT-PCR, exon trapping and sequence comparison. The GSX gene has a three exon organization, with an intron at a conserved position in the homeobox (Fig. 1, indicated by an arrowhead in B), similar to the known vertebrate and the Drosophila goosecoid genes (Fig. 1A). The initiator methionine for GSX was assigned based on its close vicinity to a sequence motif (FSIENIL) near the N terminus, that aligns with the highly conserved domain in the known goosecoid genes from frog, chick, mouse, human (FSIDNIL), zebrafish (FSIDSIL) and the D-gsc gene of the fruitfly (FTIDSIL).

The coding sequence predicts a 213-residue protein, very similar in length to the vertebrate GSC proteins (chick 246 amino acids, mouse 254 aa, Xenopus 244 aa, zebrafish 240 aa and human 252 aa), but smaller than Drosophila gsc (419 aa). The open reading frame reveals that the gene product has a paired-type homeodomain near the C terminus, which is highly homologous to the other known goosecoid homeodomains: 73.7% identity between chick GSX and Drosophila gsc; zebrafish, chick, mouse and human GSC; and 75.4% identity between chick GSX and Xenopus GSC (Blumberg et al., 1991; Cho et al., 1991; Blum et al., 1992; Izpisúa-Belmonte et al., 1993; Stachel et al., 1993; Goriely et al., 1996; Hahn and Jäckle, 1996).

The GSX homeodomain contains, like all other members of the goosecoid family, a lysine residue at the ninth position of the DNA recognition helix (Fig. 1B, indicated by an asterisk), which is a critical residue for target-specific DNA-binding (Treisman et al., 1989; Blumberg et al., 1991). A lysine in the corresponding position is only found in five other types of homeodomain proteins in different species, as there are: Bicoid (Berleth et al., 1988), Otx/Orthodenticle (Finkelstein et al.,1990; Simeone et al., 1993), Six/Sine Oculis (Cheyette et al., 1994; Oliver et al., 1995) and UNC-30/Ptx (Jin et al., 1994; Lamonerie et al., 1996).

GSX transcripts are already detectable before primitive streak formation and gastrulation, when the hypoblast sheet expands at stage EK XIII (Eyal-Giladi and Kochav, 1976). Expression is seen as a crescent associated with Koller’s sickle at the posterior margin of the blastoderm (Fig. 2A). Sagittal sections reveal that the majority of GSX-positive cells are located in the epiblast (Fig. 3A). The deep layer of the posterior epiblast, the location of presumptive primitive streak cells, shows a weak staining for GSX.

The appearance of the primitive streak is accompanied by an increase in the level of GSX expression. At early to mid-streak stages (HH stage 2-3⁻; Hamburger and Hamilton, 1951), expression is seen in the entire primitive streak including its tip (Fig. 2B). By HH stage 3-3⁺, as the groove appears in the streak, GSX expression extends through the primitive streak, but excludes Hensen’s node (Figs 2C, 3B, 4A).

When the streak has progressed to its maximal length by HH stage 4, GSX transcripts become less abundant in the primitive streak itself, with expression remaining in the primitive ridges (Fig. 2D). The clearly discernible primitive groove and the primitive pit, major sites of avian gastrulation, are negative for GSX expression (Fig. 2D). At this stage, a new domain arises around the anterior third of the primitive streak, with expressing cells in the epiblast surrounding Hensen’s node in a circular and later pear-shaped expansion (Figs 2D and 3C). The limits of GSX expression reveal a sharp boundary between the strongly stained central neuroectodermal region and the unstained more peripheral epiblast (presumptive epidermis). Fate mapping studies have identified this domain as the early neural plate (Rudnick, 1935; Spratt, 1952; Garcia-Martinez et al., 1993).

During regression of the primitive streak, when it shortens by the movements of Hensen’s node towards a more posterior position, GSX expression is maintained in the neural plate, illustrating its convergence and extension (Figs 2E-H, 3D). As the head fold forms (HH stage 6), the elevating neural plate shows a strong expression of GSX (Figs 2F, 3E), which persists during neural tube closure (Figs 2G,H, 3F).

The comparison of GSX with GSC expression reveals segregating domains along the primitive streak, with GSC becoming confined to the anterior streak and the prechordal plate (Izpisúa-Belmonte et al., 1993), and GSX remaining in more posterior parts of the streak and then in the neural plate (Figs 4, 6). It is noteworthy, however, that after segregation GSC expression is induced in the developing forebrain above the prechordal plate, where it overlaps with GSX expression.

We analyzed various parts of the primitive streak for their inductive potential at different times during chick development (Fig. 5). This type of experiments has been described before, mostly addressing the node as the chick organizer (for references, see Dias and Schoenwolf, 1990), but also more caudal levels (Gallera and Nicolet, 1969; Izpisúa-Belmonte et al., 1993; Waddington and Schmidt, 1933; summarized in Gallera, 1971; Waddington, 1952). We characterized the inductions with molecular markers and followed the fate in chick-quail chimeras. Grafts were taken from the tip or more caudal levels, in order to use cells expressing only GSC, only GSX or both genes simultaneously. They were transplanted to naive ectoderm in anterolateral positions of HH stage 3/3+ hosts. A total of 239 primitive streak grafts was made, of which 181 (76%) were used for analysis. Those that were not used include induced structures that merged with the host embryo or did not survive in culture. Two completely different outcomes were observed: either the induction of gastrulation or the induction of neurulation.

Induction of gastrulation was evident by formation of an ectopic primitive streak with a clearly discernible groove, which, in sections, revealed cells leaving the streak exactly as found in primary streaks (92 positive inductions of 128 operations). Gastrulation was induced by grafts from early streaks (HH stage 2/3⁻) whether taken from the tip or more caudally. This result was also obtained with various grafts derived from the posterior half or end of the streak using HH stage 3/3+ donors, but not using HH stage 4 or stage 5 donors. For the majority of the experiments described below, we decided to use grafts from the middle of the primitive streak (HH stage 3/3⁺), which do not express GSC. Middle streak grafts induce nicely elongated streaks and represent the most posterior site being involved in bona fide axis formation, while the second half is mainly concerned with the production of extraembryonal mesoderm (Psychoyos and Stern, 1996). At HH stage 3/3⁺, the capacity for the induction of gastrulation disappears from the tip of the streak (Gallera and Nicolet, 1969), where it is replaced by a novel capacity, inducing neural structures (see next section).

Induced streaks of various stages were analyzed with molecular markers. All were positive for the pan-mesodermal marker Brachyury (Ch-T; Kispert et al., 1995), with strong expression in the streak itself, in the ingressing mesoderm and in axial mesoderm, which is produced after prolonged incubation (Fig. 5A,B). GSC (Izpisúa-Belmonte et al., 1993), HNF3β (Ruiz i Altaba et al., 1995) and CNOT1 (Stein and Kessel, 1995) are markers for the anterior portion of the primitive streak, a region deliberately avoided in the grafts. Expression of these three markers was not typically observed in the induced streaks (Fig. 5D-F,H-J). However, a few ectopic primitive streaks with either GSC (n=1) or HNF3β (n=3) expression were also obtained (data not shown), possibly indicating that anterior streak cells were unexpectedly present in the graft. CNOT1 was only observed in epiblast cells around the tip of induced streaks, a signal corresponding to the described CNOT1 expression in very young, prenodal, paranodal and postnodal neural plate (Fig. 5D,E,G), whereas the expression domain normally found in the anterior streak did not appear (Fig. 5D-F; Stein and Kessel, 1995). Induced gastrulation events showed GSX expression in the streak itself and in the surrounding neuoectoderm (Fig. 5K-M). In order to distinguish between self-differentiation and induction, grafts were taken from HH stage 3/3+ quail embryos (Le Douarin, 1969). Quail cells derived from the graft were found in a horseshoe-like area anterior of the induced streak (Fig. 5N-P). Thus they behaved like the earliest mesoderm in chicken, the precardiac mesoderm, which normally reaches the future heart-forming region very anterior in the early embryo, either by passive displacement or by active migration. The fate of the grafted cells therefore was equivalent to the fate that they would have obtained if left in the primary embryo. Induced primitive streaks, including cells before, during and after gastrulation were completely derived from chick cells, indicating a true induction event (Fig. 5P).

The induction of neurulation proceeded quite differently (89 positive inductions of 111 operations). Neural induction by node transplantations represents the chick equivalent of Spemann’s organizer experiments using the Einsteck procedure and have been characterized in depth elsewhere (Dias and Schoenwolf, 1990; Storey et al., 1992; Beddington, 1994). We corroborated that node grafts (HH stage 3/4) induced in the overlying host epiblast a nervous system, which was clearly rostrocaudally patterned by morphological and molecular criteria (Fig. 5A,C,D). During this process, the epiblast thickened and neural folds elevated, but it appeared that the basal membrane never opened to allow gastrulation of epiblast cells to contribute to notochord or paraxial mesoderm. The grafts themselves grew out, depending on their precise stage, to form either a globular, chordoid mass (HH stage 3 nodes, Fig. 5D), or an elongated notochord (HH stage 4–, Fig. 5A). By using quail donors, we demonstrated the absence of graft-derived cells from the neuroectoderm and their presence in the chordoid material (not shown). Secondary embryos were analyzed with a number of molecular markers: Brachyury, CNOT1, GSC and EN-2. In all cases, they were expressed in their expected patterns: prospective dorsal forebrain was indicated by CNOT1 (Fig. 5D), prospective midbrain by EN-2 (not shown), notochord-like cells by Brachyury (Fig. 5A,C) and CNOT1 (Fig. 5D). Brachyury or CNOT1 expression was never induced in cells of the epiblast overlying the graft. In conclusion, we demonstrated the segregation of inductive potential along the primitive streak, with neural induction becoming confined to the anterior streak and the capacity for induction of gastrulation localized towards more posterior parts (Fig. 5).

Based on the primary structure of the homeodomain, the conservation of a second N-terminal domain and on the genomic organization, we consider GSX to be closely related to the previously described chicken goosecoid gene (GSC). It remains to be seen, however, if, like GSC, the GSX gene also confers an axis-inducing potential after injection of GSX mRNA into Xenopus embryos.

The two chicken genes, GSX and GSC, are expressed simultaneously in early primitive streak stages, when only gastrulation is occurring and neurulation is not yet induced. We assume that GSC expression within the prospective mesoderm, the Brachyury (Ch-T)-positive zone (Kispert et al., 1995), is related to the definition of dorsal values, as has been demonstrated in the amphibian Xenopus laevis (Cho et al., 1991).

With the further extension of the streak, a new inductive capacity segregates and appears in the tip of the streak. The forming node now acquires the potential to induce neurulation, as evident in transplantation experiments and in the appearance of neural commitment around the node (Dias and Schoenwolf, 1990; Gallera and Nicolet, 1969; Storey et al., 1992, Waddington and Schmidt, 1933). In parallel, the initially overlapping expression domains of the related genes segregate, so that GSC expression is correlated with the new inductive potential, while GSX characterizes the remaining part of the streak, i.e. the older, gastrulation-inducing area. Thus, when the primitive streak approaches its maximal extension, GSC and GSX do not form a nested pair of expression domains. The difference concerns an area of major importance for pattern formation processes, the organizer. Therefore, the consequences are drastic. After full extension of the streak at HH stage 4, GSC-expressing cells gastrulate and ingress as axial mesendoderm, while GSX-expressing cells stay strictly ectodermal and demarcate the forming neural plate, for which GSX now represents a unique marker (see Fig. 6 for a schematic summary). In Xenopus, neural induction has been closely linked to the GSC target gene chordin (Sasai et al., 1994, 1995). Thus, the secreted factor chordin may represent a molecular link between GSC and GSX in the neural-inducing and the responding cells of the chick.

With the ingression of the axial mesendoderm, the definition of the anterior values begins. The most anterior aspect of the embryo is specified in the animal hemisphere, where a direct contact between endoderm and mesoderm occurs. The anterior endoderm is characterized by the homeobox gene XANF1 in Xenopus (Zaraisky et al., 1995) and the related gene Rpx (Hesx1) in mouse (Hermesz et al., 1996; Thomas and Beddington, 1996). The overlying, also XANF1/Rpx-positive ectoderm, will form Rathke’s pouch and later the adenohypophysis. Adjacent to the anterior endoderm lies the pre-chordal mesendoderm, consisting of GSC-expressing cells. The overlying, first only GSX-positive, but then also GSC-positive ectoderm will go on to form the neurohypophysis. An inductive interaction between the prechordal plate and the anterior neural plate, each expressing a different member of the goosecoid family, contributes to the patterning of the forebrain anlage. We have argued elsewhere that here again a neuralizing factor like chordin could be the molecular signal, emitted from the prechordal to the neural plate (Pera and Kessel, unpublished data).

During the patterning of the forebrain anlage, GSC expression appears also in the ventral neuroectoderm above the prechordal plate. By the 4-somite stage (HH stage 8), the neuroectodermal expression extends through the mesencephalon into the rhombencephalon. Since GSX labels the complete neuroectoderm at this stage, the two genes have finally generated a true nested expression pattern.

For the induction of gastrulation, we used medial, pure GSX-positive parts of primitive streaks and avoided the anterior, GSC-positive parts. The induced streaks were unusual with regard to genes expressed normally in anterior streaks or the node. Thus neither GSC, nor CNOT1 or HNF3βexpression was found in the tips of these streaks. Nevertheless, they seemed to proceed quite normally through gastrulation and to go into neu-rulation. We demonstrated an early neural plate forming in the ectoderm around the tip, expressing GSX and CNOT1. The tip of ectopically induced streaks produced axial mesoderm as evident by Ch-T expression (Fig. 5A). Taken together, the organizer with its typical genetic set up appears to be dispensable. This finding is further sustained by results from genetic inactivation of the murine GSC or HNF3βgenes, suggesting the presence of redundant gene activities, replacing the inactivated gene in its function (Ang et al., 1994; Weinstein et al., 1994; Rivera-Peréz et al., 1995; Yamada et al., 1995). Both mutants passed quite undisturbed through gastrulation, neurulation and anteroposterior neural patterning, even though the HNF3β mutants did not generate a node or notochord. Mechanic ablation of the node or the complete anterior 40% of the primitive streak in chicken was also consistent with normal embryogenesis, although GSC expression was never regenerated in the healed, anterior streaks (Psychoyos and Stern, 1996). It remains to be seen how far, in these experiments, GSX is taking over the role of GSC.

The common beginning and later segregation of both gene activities during ontogeny could indicate a similar dynamic during phylogeny. This assumption is strengthened by findings concerning the Drosophila goosecoid gene (Goriely et al., 1996; Hahn and Jäckle, 1996). The primary structure of its homeodomain is equally distant (73.7%) from the chicken GSC and the GSX domains. The fly gene is expressed only in ectodermal structures, the brain anlage and the esophagus, and its derivative, the stomatogastric nervous system, and mutants have defects in neural development (Hahn and Jäckle, 1996). We suggest that the basic goosecoid information was split to two genes during the evolution of chordates, allowing the generation of novel fates and inductive potentials. Such a gene duplication most likely occurred together with many other duplications after the cephalochordates separated from the lineage leading to vertebrates (Holland et al., 1994). We speculate that gastrulation of goosecoid cells only became possible when, at the same time, a second goosecoid identity could be maintained in the ectoderm. As a consequence, a simple sequence, going from endoderm to notochord, was converted to a more elaborate system of endoderm-mesendoderm-notochord. The two lower groups of chordates, ascidians (Urochordata) and Amphioxus (Cephalochordata) indeed lack a prechordal mesendoderm, and their notochord is formed in continuity with the endoderm during gastrulation (Conklin, 1932; Satoh, 1994). A direct consequence of a prechordal plate in vertebrates is the more complex organization of the forebrain, as evident by the presence of a unique ventral forebrain or the presence of two eyes, instead of one (Pera and Kessel, unpublished data).

In the chick, the nervous system develops from GSX cells, which are induced and patterned under the influence of GSC cells. This interaction is exerted in a planar fashion, while still in the anterior streak, and vertically after gastrulation from the prechordal plate (Ruiz i Altaba, 1993). Thus, the two related genes are found on both sides of neural development in chicken: GSC on the inducer, GSX on the responder side. The segregation of both gene functions may represent a key feature at the beginning of a chordate nervous system not only during ontogeny, but also during phylogeny.

We thank W. Behrens for excellent technical assistance, B. Herrmann (Freiburg) for the Brachyury and A. Ruiz i Altaba (New York) for the HNF3βprobe, and H. Böger, T. Böttger, P. Gruss, G. Oliver, E. Pera, S. Stein, T. Thomas, D. Treichel and A. Voß for discussions. This work was supported by the Max-Planck-Gesellschaft, the SFB 271, and by grants from the Leyla and Mathers Foundation and NSF.

The EMBL database accession number for the sequence reported in this paper is Y09850.