Anterior identity is established in chick epiblast by hypoblast and anterior definitive endoderm

Early patterning of the vertebrate central nervous system involves a complex and interwoven set of spatiotemporal inductions, tissue movements and patterning mechanisms. Two major tasks are achieved early on: the definition of anterior positional identity and the segregation of neural tissue identity. In mouse (Beddington and Robertson,1998; Beddington and Robertson,1999), zebrafish (Houart et al., 1998; Koshida et al.,1998) and Xenopus(Jones et al., 1999),signalling centres have been identified that are distinct from the classical organiser and capable of establishing anterior positional identity separately from neural specification. In chick, therefore, positional and tissue identity may also be separable. Three mechanisms could exist for establishing initial anterior positional and tissue identity in the epiblast/prospective neural plate: first, neural specification leads to neuralised tissue, which by default is anterior in character (Mangold,1933; Nieuwkoop et al.,1952; Nieuwkoop and Nigtevecht, 1954; Spemann,1931; Spemann,1938); second, positional identity is conferred by a separate mechanism upon neuralised tissue, which is initially positionally neutral(Waddington and Needham,1936); and third, initial anterior positional identity is established in the epiblast independent of and before neural specification occurs. The latter two possibilities require that anterior positional identity be established separately from neural specification.

In chick, the process of neural induction begins before the onset of gastrulation, with competence being conferred by FGF signals emanating from the posterior of the embryo (Muhr et al.,1999; Streit et al.,2000; Wilson et al.,2001; Wilson and Edlund,2001; Wilson et al.,2000). The cellular interactions leading to the specification of competent tissue as neural, and the timing over which they occur, remain unclear. Although transplants of posterior epiblast can induce transient expression of pre-neural markers such as Sox3 in epiblast, stable expression of Sox2 in specified neuroectoderm requires both central and posterior epiblast cells to come together at mid-streak stages (3+ or 3c/d), forming a functional organiser(Streit et al., 2000). Further support for the timing of neural specification at mid-streak stages comes from explant studies in which competent tissue at stage 3d, but not 3c, cultured in isolation, was able to self differentiate, developing the columnar neuroepithelial morphology of specified neuroectoderm, as well as having stable expression of Sox2 (Darnell et al., 1999).

The relative spatiotemporal positions of early embryonic tissues(Fig. 1) suggest that several tissues could function potentially in anteroposterior patterning. Candidate tissues able to produce `organising' signals include a population of central epiblast (CE) cells (Darnell et al.,1999; Garcia-Martinez et al.,1993; Hatada and Stern,1994; Healy et al.,2001; Lawson and Schoenwolf,2001a; Lawson and Schoenwolf,2001b; Schoenwolf et al.,1989b; Streit et al.,2000), and the underlying lower layer, the hypoblast and ingressing anterior definitive endoderm (ADE). The CE population is a group of epiblast cells rostral to the tip of the primitive streak between stages 2 and 4. They are in a position equivalent to that of the mouse early gastrula organiser (EGO), which has been shown to have a role in head patterning when combined with epiblast and anterior visceral endoderm (AVE)(Tam and Steiner, 1999). As the primitive streak extends forward, the CE population becomes incorporated into the streak (Garcia-Martinez et al.,1993; Garcia-Martinez and Schoenwolf, 1993; Joubin and Stern, 1999; Lawson and Schoenwolf, 2001a; Lawson and Schoenwolf, 2001b; Schoenwolf and Alvarez, 1989; Schoenwolf et al., 1989a; Schoenwolf et al., 1989b; Schoenwolf et al.,1992; Smith and Schoenwolf,1991). Early fate-mapping studies used quail/chick chimaeras and fluorescent dye injections to determine the fate of cells in the rostral streak (Garcia-Martinez et al.,1993; Garcia-Martinez and Schoenwolf, 1993; Schoenwolf et al., 1992; Selleck and Stern, 1991). Homotopic and isochronic cell grafts from stage 3a/b rostral streak contributed extensively to head mesenchyme and foregut endoderm, whereas a small proportion was also detected in notochord and the median hinge-point cells (i.e. the future floor plate of the neural tube). Stages 3c-4 rostral streak cells contributed mainly to notochord and median hinge-point cells, although a small number were traced to the head mesenchyme and foregut endoderm.

Other tissues in chick with putative `organising' ability are those comprising the lower layer - the hypoblast and ADE. Based on similar gene expression patterns, fate and a dual role in `protecting' the prospective forebrain from caudalising influences of the organiser, the hypoblast at stage XII/XIII has been proposed to be the homologue of the mouse anterior visceral endoderm (Foley et al., 2000). The hypoblast forms the primitive endoderm underlying the epiblast from stage X/XI until the ADE begins to ingress at stage 3a, displacing the hypoblast rostrally. Transplant experiments of the hypoblast only, produced transient induction of Sox3 and Otx2(Foley et al., 2000). These authors argue that the hypoblast, therefore, protects the epiblast against caudalising influences rather than influencing cell fate. Neither hypoblasts from stage XIV, nor the ADE present under the epiblast from stage 3a until the beginning of axial mesoderm ingression from stage 4+, have previously been implicated in anterior patterning. Potentially both planar and vertical signalling mechanisms operate in patterning positional identity through central epiblast, and lower layer hypoblast and ADE, respectively.

Owing to contradictory results and gaps in our understanding regarding the role of CE, the hypoblast and ADE, we have tested these tissues for a role in establishing anterior positional identity. Previous chick transplant studies have failed overall to show a role for lower layer tissues in determining cell fate (Foley et al., 2000). Removing the lower layer has also met with little previous success, mostly because of the embryos' ability to recover and replace ablated tissues, such as the hypoblast at early stages(Vanroelen et al., 1982), or to tissue being removed later than the time in question(Withington et al., 2001). We used the transection assay (Fig. 2) (Darnell et al.,1999; Healy et al.,2001; Schoenwolf et al.,1989b; Yuan et al.,1995a) to address what the role for these lower layer tissues is,and determined that anterior positional identity seems to be separable from neural specification. Isolating rostral epiblast (prospective anterior neural plate) from the influences of Hensen's node and ingressing axial mesoderm was crucial, because ingressing axial mesoderm from stage 4+ has a role in the induction of Ganf, the earliest specific maker of anterior neural plate (Knoetgen et al., 1999). The expression patterns of Sox2, the most definitive early neural specification marker identified to date(Rex et al., 1997; Streit et al., 2000; Streit et al., 1997; Uchikawa et al., 2003), and Ganf were re-examined and used to define anterior and neural identity. We show that anterior positional identity is established and maintained in the epiblast by the hypoblast at stage 3a/b and ADE at stages 3d and 4, apparently separately from neural specification, and we propose a revised model for establishing anteroposterior polarity, neural specification and head patterning, based on this new evidence.

Hens' eggs (White Leghorn) were incubated at 38°C for desired stages. Prestreak embryos were staged according to Eyal-Giladi and Kochav(Eyal-Giladi and Kochav, 1976)(EGK; Roman numerals), and Hamburger and Hamilton(Hamburger and Hamilton, 1951)for primitive streak and later stages (HH; Arabic numbers), with HH stage 3(gastrula) embryos refined according to Schoenwolf and co-workers(Fig. 1)(Chapman et al., 2002).

In situ hybridisation was performed as described previously(Chapman et al., 2002). Embryos were then cleared in 80% glycerol/PBS, embedded in 20% gelatin, fixed with 4%PFA and sectioned using a Leica vibratome at 40-50 μm. Embryos were imaged with a SPOT, Coolsnap or Zeiss Axiocam digital camera. The following markers were used: Sox2, specified neuroectoderm (R. Lovell-Badge): Wnt8c, ingressing mesodermal cells (J. Dodd): Ganf, earliest marker of anterior neuroectoderm (A. Zaraisky); Fgf8, primitive streak (G. Martin); Chordin, primitive streak, Hensen's node and ingressing axial mesoderm (A. Graham); Crescent, hypoblast and ADE(P. Pfeffer).

Transection of embryos was performed as described by Darnell et al.(Darnell et al., 1999). In experiment 1, embryos were transected to determine the effect of separating rostral tissues from the primitive streak, prospective node and ingressing mesoderm. Embryos were transected at the rostralmost level of the streak (Type B), or 125 μm rostral to the streak (Type C), at stages 3a-4+ and cultured on an agar/albumen substrate with no added culture media for 24 hours(Fig. 2). The blastoderm isolates were then processed for Ganf and Sox2 transcripts(Table 1). In experiment 2, the lower layer of rostral isolates was removed to determine whether this layer has a role in patterning the rostral epiblast. Isolates were cultured in collagen: 3.3 mg/ml rat tail collagen (Roche) was prepared in 0.2% acetic acid. 480 μl collagen, 36 μl DEPC-H₂O, 60 μl 10×DMEM and 20 μl 0.75% bicarbonate solution were added together on ice. Rostral and caudal isolates of each transected embryo were embedded, and after 30 minutes at 37°C in a 5% CO₂ incubator, carbonated Neurobasal medium supplemented with Glutamax was added. Embryos were transected (Type B)in saline (123 mM), followed by removal of the lower layer using tungsten needles (0.125 mm tungsten wire, WPI). No enzymatic treatments were used. RBIs with an intact lower layer served as controls(Table 2). To test for mesoderm in the RBIs, in experiment 3, transected embryos were fixed immediately and then processed for Wnt8c expression. Experiments 4 and 5 were designed to test sufficiency of the lower layer to induce Ganf:either rostral (experiment 4) or caudal (experiment 5) lower layer was recombined with rostral epiblast in collagen culture for 24 hours and then processed for Ganf transcripts.

Before discussing our results, it is important for the purposes of clarity to define the terminology that we will be using. This is especially true because early development of the avian embryo is complex, and different investigators often use different terms, or even the same terms but in different ways.

The lower (ventral) layer is formed by polyingression of cells from the overlying epiblast at stage X/XI, forming islands of cells in the subgerminal cavity (Fig. 1)(Harrison et al., 1991; Lawson and Schoenwolf, 2001a). Together with cells moving rostrally from Koller's sickle and the posterior marginal zone (PMZ), a complete lower layer is formed, called the primary hypoblast (Callebaut et al.,1999; Stern and Canning,1990; Vakaet,1970). The endoblast (secondary hypoblast) forms at stage XIII/XIV, with cells moving rostrally from Koller's sickle and the PMZ. Primary and secondary hypoblast form a continuous sheet of cells under the epiblast by stage XIV/2 - the primitive endoderm. The primitive streak forms at stage 2, as a triangularly shaped structure that elongates rostrally. Definitive endoderm begins ingression through the rostral end of the primitive streak from stage 3a to stage 4/4+, by which time the lower layer has fully displaced the hypoblast sheet rostral to the embryo, forming the germ cell crescent (Lawson and Schoenwolf,2003). At stages 3a/b the rostral streak gives rise to the ADE,including the midline prechordal plate endoderm (PCPE) that lies beneath the forebrain. During subsequent development, the prechordal plate endoderm buds off proliferative mesoderm and together with ingressing mesoderm, forms a middle layer, the prechordal plate mesoderm, which contributes to the head mesenchyme (Seifert et al.,1993). Axial mesoderm ingresses through the rostral streak from stage 4+ (the head process), and consists of a mixed cell population of prechordal plate mesoderm and rostral notochord, which intercalates between the neuroectoderm and ADE (Foley et al.,1997; Vesque et al.,2000). The molecular basis for the spatial separation of these two populations is unclear, although SEM studies of the morphological movements have been described (England,1984; England and Wakely,1977; England et al.,1978; Wakely and England,1979). The laying down of more caudal notochord occurs as Hensen's node and the definitive streak regress caudally. Intercalation of the fan-shaped prechordal plate mesoderm results in the prechordal plate mesoderm coming to partially overlie the PCPE.

Sox2 is the earliest pan-neural marker stably expressed in the specified neuroectoderm (Rex et al.,1997; Streit et al.,2000; Streit et al.,1997). Embryos were tested for expression of Sox2 from stage XI/XII (not shown), with expression first detected at stage 3d, as expected (Fig. 3A). Expression began just rostral and lateral to Hensen's node and later expanded rostrally and laterally toward the outer boundary of the neural plate, away from the streak (stages 4/4+) (Fig. 3B). This pattern suggests that neural specification occurs in a spatiotemporal manner across the prospective neuroectoderm. The caudal boundary of expression remained constant, suggesting that the first cells to express Sox2are not the anteriormost neuroectoderm, but rather are the more caudal neural plate. By stage 5, Sox2 was expressed throughout the neural plate,which is still flat prior to formation of the head fold and neural tube(Fig. 3C). Concomitant with node regression, the caudal boundary of Sox2 expression extended caudally through convergent extension (Fig. 3D). The ventral neural plate was lighter in colour as the neural tube formed, while the neural plate narrowed as the neural folds rose up and moved medially, with varying levels of expression within the rostrocaudal length of the neural plate, with stronger expression of Sox2 in the rostralmost neural plate. In summary, Sox2 is detected from stage 3d onwards and is the earliest available stable marker of specified neuroectoderm.

The time course of Ganf expression was examined by in situ hybridisation from stage X/XI to stage 17. Ganf transcripts were detected from stage 4 (Fig. 4A), earlier than previously reported(Knoetgen et al., 1999). This is important because Knoetgen and colleagues reported that Ganf is expressed only once ingressing axial mesoderm underlies the anterior neural plate. Our wholemounts and sagittal sections(Fig. 4A,B) reveal a gap between Ganf expression (blue) and Chordin (red), which is a marker of Hensen's node and ingressing axial mesoderm. In addition, previous studies using scanning electron microscopy of dissected blastoderms show that ingressed mesoderm is absent rostral to the node at this stage(Lawson and Schoenwolf,2001a). Furthermore, unlike in mouse, where transcripts of the Anf homologue Hesx1 are found in both neuroectoderm and the underlying anterior visceral endoderm(Hermesz et al., 1996),expression in chick was detected only in neuroectoderm fated to become the forebrain rostral to the ZLI (Kazanskaya et al., 1997). At later stages, Ganf-positive mesodermal cells may be present, as in mice and frogs, but this is not the case at stage 4 (Fig. 4A,B)(Kazanskaya et al., 1997). At stage 4/4+ the definitive streak has extended maximally(Fig. 4C) and ingression of axial mesoderm begins. Although these axial cells migrate rostrally, Hensen's node regresses caudally with the streak, laying down the notochord. The domain of expression of the axial mesodermal marker Chordin lengthens during this process (Fig. 4C-F). Ganf expression narrows mediolaterally from stage 5 as the neural plate extends rostrally and begins to fold(Fig. 4E), leading to formation of the neural tube. Ganf expression becomes progressively restricted,until at stage 17 when expression is detected only in the floor of Rathke's pouch (not shown), a region fated to form part of the anterior pituitary. Double in situ hybridisation confirms that Ganf expression is co-localised with the rostralmost Sox2 domain, beginning in stage 4 embryos (Fig. 5). These data indicate that Ganf is expressed before ingression of axial mesoderm at stage 4/4+ and may, therefore, be induced by tissues other than Hensen's node and the axial mesoderm.

Transected rostral blastoderm isolates (RBIs) were processed for Sox2 (Fig. 6) and Ganf (Fig. 7)transcripts to establish the percentage of type B and C RBIs with expression at each stage (experiment 1). Results are summarised in Table 1, with stage 3a/b, type B, RBIs (directly rostral to the node) expressing Sox2 in 6/11 cases and Ganf in 7/11 cases. This number drops at stage 3c when the central epiblast (CE) cells, which have the ability to act as an organiser or inducer of an organiser (Darnell et al.,1999), become incorporated into the extending streak and are thus excluded from the RBIs, with only 3/11 isolates expressing Sox2 and 2/13 expressing Ganf. At stage 3d, with neural specification, 7/11 Sox2 and 6/9 Ganf-expressing RBIs are detected. Likewise at stage 4, 8/8 and 6/10 RBIs expressed Sox2 and Ganftranscripts, respectively, with later stages all expressing Ganf. Control embryos (stages 3a-4) were processed immediately after transection to ensure the accuracy of the transection. Fgf8 (primitive streak) and Chordin (primitive streak, Hensen's node and ingressing axial mesoderm) transcripts were detected only in the caudal isolate and not the RBI(6/6, each marker, not shown). Sox2 and Ganf were not expressed in the RBIs from type C transections (125 μm rostral to streak)until stage 3d when 1/8 RBIs was positive for Ganf transcripts,followed by stage 4 when 3/8 expressed Ganf(Table 1). This result is consistent with the position of CE cells and the expression pattern of Sox2, the leading edge of which does not extend more than 125 μm rostral to the node until stage 4 (Fig. 5). As Sox2 is expressed in a spatiotemporal manner spreading from Hensen's node toward the outer edge of the neural plate, RBIs resulting from Type C transections would have had to receive the signals resulting in neural specification before being transected, and CE cells would be excluded from all the isolates. RBIs that had not received signals to undergo neural specification would be unable to express Ganf, as is the case here. This and other studies show that Ganf expression only occurs in neural specified cells after receiving signals from the organizer,CE and/or ingressing axial mesoderm(Knoetgen et al., 1999). As transections exclude the organiser and ingressing axial mesoderm, this raises the question of whether neural specification alone is sufficient for expression of Ganf or whether an additional signal from the CE or lower layer might be required. To test the hypothesis that CE or lower layer is required, we removed the lower layer from RBIs at each stage and determined whether Ganf was expressed. As type B transections were regarded as the most informative, no further type C transections were performed.

To determine a role for the lower layer in induction of Ganf,embryos were probed for transcripts after removal of the lower layer from RBIs(experiment 2) (Fig. 7 and Table 2). To determine the presence or absence of the lower layer, control RBIs, with the lower layer intact or removed, were processed immediately after transection for Crescent. Intact RBIs had strong expression in the lower layer (8/8),as expected, while in lower-layer deficient RBIs Crescent was absent(4/4). All caudal isolates expressed Crescent (12/12, data not shown). For longer-term culture, RBIs were embedded in collagen, with either the lower layer removed, or in control isolates, left intact to ensure that the collagen itself or culture medium did not affect Ganf expression. First, in the control RBIs transected at stage 3a/b, 11/19 RBIs (57.9%)expressed Ganf. By stage 3c, the percentage reduces as expected when CE cells are excluded, with only 5/19 (26.3%) of the RBIs with detectable expression. Once neural specification occurs at stage 3d the percentage rises again to 58.3% in 7/12 RBIs. This result is the same as for stage 4 isolates,7/12 RBIs expressing Ganf. By contrast, removal of the lower layer results in a decreased number of RBIs expressing Ganf. This indicates that in addition to neural specification, the lower layer [i.e. hypoblast and anterior definitive endoderm (ADE)], are required for Ganf expression(see Table 2). Each stage tested suggests a progressive role for the lower layer. At stage 3a/b the lower layer, rostral to the primitive streak, consists almost exclusively of hypoblast, with ADE only beginning to ingress through the primitive streak. When this layer is removed, only 2/17 isolates (11.8%) continue to express Ganf, compared with 11/19 in controls. Sox2 expression was ascertained for RBIs from stage 3a/b embryos in which the lower layer was removed. 6/9 (66.7%) isolates were positive for Sox2 transcripts. This suggests that removal of the lower layer has no effect on the number of RBIs that are neuralised. This result further suggests that CE is not the inducer of Ganf in the epiblast, although it is crucial for neural specification and the induction of Sox2.

At stage 3c, the number of RBIs with Ganf transcripts is similar to that of intact embryos, 5/19, compared with 4/15 isolates lacking the lower layer. The lack of neural specification in these RBIs is due to the exclusion of CE cells and results in low numbers of RBIs expressing Ganf. Lower layer removal, now composed of hypoblast and ADE, has no effect on the numbers of RBIs expressing Ganf, indicating that the inducing activity of the lower layer may be no longer required. Interestingly, the ADE expresses only Crescent, Cerberus, Hex and Otx2 at stage 3c, whereas at stage 3d, Lim1 and Hnf3β are also induced(Chapman et al., 2002). This may indicate that at stage 3c the ADE cannot perform a maintenance role, but by stage 3d has developed sufficiently to do so. At stage 3d, removal of hypoblast and ADE again causes a reduction in the numbers of embryos expressing Ganf, 3/15 (20%) compared with 7/12 intact RBIs. This is indicative of a maintenance function being lost. At stage 4 this effect is even more pronounced with 0/8 isolates (7/12 intact RBIs) positive for Ganf transcripts, suggesting that although the lower layer is responsible for initial induction and maintenance of anterior identity in epiblast, factors from other tissues may be required to maintain and even stabilise the expression. Tissue candidates for this role include Hensen's node and ingressing axial mesoderm, which have been identified previously as important in Ganf expression(Knoetgen et al., 1999). In summary, removal of the lower layer in RBIs demonstrates that vertical signalling by the lower layer is required for the expression of the anterior neuroectoderm marker Ganf, apparently separately from and before neural specification, and that following neural specification the lower layer has a maintenance role, without the involvement of mesoderm.

Patterning of the rostral ectoderm in intact isolates could be due to the presence of mesodermal cells that are inadvertently included in the RBIs when transecting (experiment 3). When the lower layer is removed in these isolates,any mesodermal cells present could conceivably also be stripped away,resulting in loss of Ganf expression. To test this we, transected embryos from stages 3a/b through to stage 4 and tested for Wnt8cexpression, which marks ingressing mesoderm(Fig. 8). Isolates were fixed immediately after transection, and in no case were Wnt8c transcripts detected by in situ hybridisation in the RBIs (n=29; stage 3a/b, n=12; stage 3c, n=7; stage 3d, n=7; stage 4, n=3). The caudal blastoderm isolates from these transections acted as controls for the presence of Wnt8c and were positive for mesodermal cells in all cases.

Having established a requirement for the lower layer in patterning regional identity in the overlying ectoderm we wanted to know whether the hypoblast tissue at stage 3a/b was sufficient (Fig. 9). Loss of Ganf expression could be due simply to damage to the epiblast while removing the lower layer. To test this, RBIs at stage 3a/b were stripped of the lower layer and then rostral ectoderm and hypoblast were recombined in collagen culture for 24 hours, followed by processing for Ganf expression (experiment 4). Ganf was detected in 7/11 cases (63.6%), the same as in intact RBIs, indicating that the lower layer is sufficient to induce Ganf in the rostral ectoderm(Fig. 9A). Caudal endoderm(lateral to the primitive streak) at stage 3a/b was used to test whether another regional population of endoderm cells could substitute for the rostral lower layer (experiment 5). Embryos were transected as normal, the rostral endoderm removed and then rostral ectoderm was recombined in collagen culture with the caudal endoderm. In none of the cases was Ganf induced (0/6)(Fig. 9B). This was interesting because even if the caudal endoderm is not sufficient to induce Ganf,a small number of cases might be expected to express Ganf, raising the possibility that the caudal endoderm was not only insufficient to induce Ganf but actually inhibited Ganf expression, although the later possibility remains to be tested more vigorously.

The transection assay provides evidence that suggests that the establishment of anterior positional identity in the epiblast is a separate event from neural specification. This evidence was obtained by using the earliest available specific anterior neural identity marker Ganf, as only tissue that is both neural and anterior in character expresses Ganf (Knoetgen et al.,1999). Transection separates prospective anterior neural plate from the influence of the node (the classical organiser), preventing ingression of axial mesoderm, whereas removing the lower layer from these RBIs provides insight into the inductive ability of component tissues at various stages. Our results demonstrate a novel role for the lower layer hypoblast and ADE in patterning overlying epiblast.

After transection at stage 3a/b, the epiblast still neuralises, as indicated by the expression of the definitive pan-neural marker Sox2(Rex et al., 1997). Neural specification in RBIs depends on the presence of a population of central epiblast (CE) cells with `organising' ability, acting either as an organiser or inducer of an organiser (Darnell et al.,1999). After removal of Hensen's node in whole embryos and tissue isolates, the organiser reconstitutes(Joubin and Stern, 2001; Psychoyos and Stern, 1996; Yuan et al., 1995a; Yuan et al., 1995b; Yuan and Schoenwolf, 1998; Yuan and Schoenwolf, 1999). We have not determined whether the same mechanism operates in the rostral blastoderm isolates, although markers of notochord (Not1), node(Shh) and primitive streak (Brachyury/T) were detected in RBIs, suggesting that the organiser is reconstituted(Darnell et al., 1999). Therefore, tissue identified as able to specify neural identity in RBIs was present at stage 3a/b and reduced at stage 3c as the primitive streak extended rostrally, incorporating the CE cells(Darnell et al., 1999; Lawson and Schoenwolf, 2001a; Lawson and Schoenwolf, 2001b),although long-range neural specification signalling, prior to transection,cannot be ruled out entirely. When lower layer, composed only of hypoblast,was included in the RBI, Ganf was expressed at the same frequency as Sox2. By contrast, removing lower layer from these RBIs resulted in the loss of Ganf expression, but did not affect neural specification. Ganf was not transiently induced, suggesting that lower layer signals are required to establish positional identity in the overlying epiblast before neural specification at stage 3d.

Hypoblast seems to be required for only a brief period, because in stage 3c transections, removal of the hypoblast does not abolish Ganfexpression. Only a small proportion of RBIs undergo neural specification at this stage, as CE cells are excluded from transected RBIs. By contrast,concomitant with neural specification at stage 3d, transection does not affect the neural character of RBIs, whereas removal of lower layer still leads to a reduction in the percentage of RBIs with Ganf expression, similar to that for stage 3a/b. Hypoblast has been displaced rostrally by the ADE(including midline prechordal plate endoderm) now underling the region where Ganf is induced. The ADE may perform a maintenance role from stage 3d when expression of Lim1 and Hnf3β is induced, in addition to Crescent, Cerberus, Hex and Otx2(Chapman et al., 2002). Further work will be needed to determine whether the ADE is directly involved in the induction of Ganf, or whether ADE maintains anterior character specified earlier by the inductive interaction with the hypoblast.

There is an ongoing debate as to which lower layer tissue in the chick is equivalent to the mammalian AVE, and whether the hypoblast and ADE have any patterning role. In mouse, adjoining the rostral boundary of the primitive streak the early gastrula organiser (EGO), together with epiblast and AVE, is required for head formation(Martinez-Barbera and Beddington,2001). Chick CE cells are in a position equivalent to the EGO and have `head organiser' properties (i.e. the ability to induce neural identity)(Darnell et al., 1999; Garcia-Martinez et al., 1993; Healy et al., 2001; Schoenwolf et al., 1989b). Determining whether CE cells can be considered a true head organiser still requires that roles in neuralising naïve epiblast and re-patterning more caudal areas of the neural plate be demonstrated. At stage 2 and 3a/b, the CE population is rostral to the extending streak, but by stage 3c it becomes incorporated into the rostrally extending streak forming Hensen's node(Schoenwolf et al., 1989b). The node acts like a head organiser, establishing and refining neural identity, and maintaining and embellishing patterning in overlying neuroectoderm. The properties of axial mesoderm as it ingresses through Hensen's node at stage 4+ are reported to be the result of its origin in the node and vertical signals from the definitive endoderm as it intercalates between the upper and lower germ layers(Vesque et al., 2000). An anteriorising signalling centre in the lower layer could act as the source of signals that operate to further pattern the extending axial mesoderm,indicating a relay mechanism operates, where the anterior endoderm patterns,directly or indirectly, the prechordal plate mesoderm, which in turn patterns the overlying neuroectoderm (Dale et al.,1997; Foley et al.,1997; Pera and Kessel,1997). Our results suggest that hypoblast and ADE also have an earlier role in directly patterning the overlying epiblast.

With ingression of axial mesoderm through Hensen's node, head organiser ability is lost, perhaps allowing remaining cells to perform the role of trunk/tail organiser, refining the patterning of more caudal parts of the neural plate. An important related issue is whether neural identity is a neutral fate, with lower layer providing positional identity. In our experiments, removal of the lower layer does not affect neural specification,because RBIs still express the definitive pan-neural marker Sox2. The significance of the Sox2 expression pattern, which expands progressively from medial tissue adjacent to Hensen's node and then laterally across the neuroectoderm, suggests that neural specification does not occur first in the most anteriorly positioned cells of the prospective neural plate. However, Ganf expression is lost when the lower layer is removed. Therefore, these data together support a model in which loss of anterior identity does not affect the neural character of tissue, suggesting that neural identity itself is neutral with respect to position.

Transplanted chick lower layer was unable to induce anterior neuroectodermal (ANE) markers in epiblast, whereas rabbit AVE and chick axial mesoderm induced Ganf (Knoetgen et al., 1999). A heterochronic shift in patterning of the ANE was proposed, with chick prechordal mesoderm taking over the role played by mouse AVE. Why did the transplanted chick tissue not induce expression of Ganf? The signals needed to induce positional identity were either no longer present (transplanted hypoblast was older than stage 3a/b), i.e. necessary signals could have been reduced by enzymatic treatments used to facilitate isolation of the lower layer, or the responding tissue was not neuralised and, therefore, not competent to express Ganf. Our data demonstrate that intact hypoblast at stage 3a/b is required for the neuroectoderm to express Ganf. Extirpation of hypoblast at stage 3 did not lead to loss of Ganf expression at later stages; however,stage 3 is highly dynamic and prospective ANE was not separated from the influence of the node or ingressing axial mesoderm, both of which are sufficient to induce the expression of Ganf(Knoetgen et al., 1999). Transections demonstrate that in RBIs the lack of a node does not affect induction of Ganf, whilst axial mesoderm begins ingressing only after Ganf expression has begun and, therefore, is unlikely to be the initial endogenous inducer. Thus, chick axial mesoderm is probably not the homologue of the mouse AVE and a heterochronic shift is unlikely. The later patterning role of axial mesoderm is important in refining regional neuroectoderm identity (Dale et al.,1997; Foley et al.,1997; Pera and Kessel,1997), but initial anterior positional identity must be assigned to earlier endodermal tissues.

An alternative hypothesis, a modified Nieuwkoop model, proposes that AVE and hypoblast at stage XII/XIII are equivalent tissues. However, the lower layer in this model is responsible for cell movements, rather than cell fate,directing cells away from the caudalising influence of Hensen's node(Foley et al., 2000). Transplanted hypoblast induced transient expression of Sox3 and Otx2 in epiblast, indicating signalling capability, but as expression was not maintained the authors suggested that anteriorising the epiblast is not the hypoblast's main role. Our data do not support the interpretation of the proposed early pre-forebrain state, defined by the expression of the pre-neural markers Sox3 and chick ERNI(Streit et al., 2000). ERNI has been shown to be a retrotransposon only present in the Galliform genome,requiring clarification of its biological significance(Acloque et al., 2001), and Sox3 is expressed in a mosaic of epiblast cells from prestreak stages across the entire area pellucida, only becoming restricted to the neuroectoderm after stage 3d (Rex et al.,1997). Rather, in our model successive inductive interactions anteriorises epiblast, with molecular signals establishing and maintaining anterior identity separately from neural specification. An explanation for the failure to maintain the initial Sox3 and Otx2 expression is that hypoblast is unable to either stabilise or maintain this expression,requiring later ADE to provide the necessary signals. More importantly, this early induction suggests that, as in the mouse, the ability to pattern the early embryo is not confined to Hensen's node and its derivatives.

We suggest a revised hypothesis, where successive inductive interactions between hypoblast/ADE and epiblast act to promote anterior character(Fig. 10). As hypoblast is displaced rostrally by the ADE, signals from ADE stabilise/maintain this rostral identity in the overlying epiblast/neuroectoderm. Rostrally located hypoblast (stages XII-XIV) is remarkably similar to the mouse AVE, expressing Lim1, Hnf3b, Otx2, Gsc, Cerberus, Hex and Crescent(Chapman et al., 2002). Genes expressed in ADE include Crescent, Cerberus, Hex and Otx2,whereas Lim1 and Hnf3b are detected only after stage 3d(Chapman et al., 2002). Ganf is the earliest marker detected in the rostral epiblast in response to anteriorising signals from the lower layer and neural specification by the head organiser. Head organiser cells leave Hensen's node,as ingressing axial mesoderm, permitting the remaining population to perform the role of trunk/tail organiser. Changing gene expression reflecting this include Otx2, Nodal and Dkk1, which are lost from the streak at stages 5+/6, while Bmp7 is now expressed in rostral streak from which it was previously excluded (Chapman et al., 2002). This novel hypothesis allows for separate signalling pathways to pattern anterior and neural identity, and for the hypoblast to direct cells with a rostral fate away from the caudalising influence of the trunk/tail organiser(Foley et al., 2000). It further takes into account results suggesting that definitive endoderm is required for patterning neural plate, as stage 4+ removal results in loss of the forebrain because of lack of vertical signals to ingressing head process,and also direct maintenance signals to the overlying neuroectoderm(Withington et al., 2001). Loss of these stabilising and maintenance signals results in the loss of forebrain identity. This hypothesis is further supported by the Foxa2conditional mouse mutant, where loss of Foxa2 results in axial mesoderm losing its identity; anterior neuroectoderm in turn is not stabilised, resulting in forebrain truncation(Hallonet et al., 2002).

Hensen's node is, therefore, able to act as head and trunk/tail organiser,with spatiotemporally separated signals. The early head organiser producing neural identity in the plane of the ectoderm and the late organiser posteriorising more caudal neural plate. Anterior identity results from successive inductive interactions between the hypoblast, ADE and overlying epiblast, stabilised and refined later by ingressing axial mesoderm.

This work was supported by the MRC, a Wellcome Trust Prize Fellowship(S.C.C.), and NIH grants NS18112 and DC04185 (G.C.S.).