A morphogen gradient of Wnt/β-catenin signalling regulates anteroposterior neural patterning in Xenopus

The vertebrate central nervous system (CNS) is divided anteroposteriorly into forebrain, midbrain, hindbrain and spinal cord during early embryogenesis. Neural inducers and modifiers establish a crude anteroposterior (AP) pattern before and during gastrulation that becomes refined during later stages (Saha and Grainger, 1992; Lumsden and Krumlauf, 1996; Sasai and De Robertis, 1997; Chang and Hemmati-Brivanlou, 1998a; Beddington and Robertson, 1999; Gamse and Sive, 2000; Stern, 2001). In amphibia, classical experiments have established a central role for Spemann’s organizer in this process. This organizer region does not only induce ectoderm to acquire neural fate, it also emits regionally specific inducers, with anterior mesendoderm inducing forebrain and posterior chordamesoderm inducing spinal cord (Gilbert and Saxén, 1993; Ruiz i Altaba, 1993; Gould and Grainger, 1997; Harland and Gerhart, 1997; Nieuwkoop, 1997). Similarly, classical experiments by Nieuwkoop, Saxén, Toivonen and colleagues have suggested a two-step model where a gradient of a posteriorizing factor (‘transformer’) confers progressive posterior identity to tissue which has initially been induced as anterior neuroectoderm by early involuting mesendoderm (Gilbert and Saxén, 1993; Doniach and Musci, 1995; Nieuwkoop, 1997; Sasai and De Robertis, 1997). The molecular identity of the transforming gradient has remained elusive, although a number of candidate posteriorizing signals have been proposed, notably fibroblast growth factors (FGFs), retinoic acid (RA) and Wnts (Durston et al., 1989; Ruiz i Altaba, 1993; Slack, 1994; Doniach, 1995; McGrew et al., 1995; Pownall et al., 1996; Kolm and Sive, 1997; McGrew et al., 1997; Moon et al., 1997; Isaacs et al., 1998; Pownall et al., 1998; Wodarz and Nusse, 1998; Holowacz and Sokol, 1999; Gavalas and Krumlauf, 2000; Ribisi et al., 2000; Gamse and Sive, 2001). Yet, for none of them has it been shown that (1) it is required for AP patterning, (2) it acts directly on neuroectoderm and (3) it functions endogenously in a graded fashion.

Among the candidate transformers, Wnts have recently been implicated as endogenous antagonists of the head organizer (Niehrs, 1999). Consistent with the distinct inducing activities of anterior mesendoderm and chordamesoderm noted by Spemann and Mangold, inhibitors of transforming growth factors β (TGFβs) and Wnts, which are thought to mediate inductions by the organizer, are differentially expressed along the AP axis of the organizer and its derivatives (Harland and Gerhart, 1997; De Robertis et al., 2000). Notably, Wnt inhibitors are expressed predominantly in the anterior mesendoderm (Niehrs, 1999; De Robertis et al., 2000; Kiecker and Niehrs, 2001). In line with these inhibitors regulating AP neural patterning, co-injection of Wnt and bone morphogenetic protein (BMP) inhibitors on the ventral side of Xenopus embryos induces ectopic heads including forebrain structures while BMP inhibitors only induce ectopic trunk including spinal cord (Glinka et al., 1997). Thus, a distinguishing feature of the head organizer is the repression of Wnt signalling (Niehrs, 1999).

Other data support the idea that Wnts antagonize the head organizer and have posteriorizing properties. Overexpression of various Wnts (Christian and Moon, 1993; McGrew et al., 1995; Fredieu et al., 1997; McGrew et al., 1997; Saint-Jeannet et al., 1997; Chang and Hemmati-Brivanlou, 1998b; McGrew et al., 1999; Gamse and Sive, 2001), the Wnt pathway components β-catenin and XTcf3 (McGrew et al., 1995; Darken and Wilson, 2001; Hamilton et al., 2001) or treatment with the artificial activator of the Wnt pathway lithium chloride (Fredieu et al., 1997) all lead to repression of anterior and concomittant induction of posterior neural markers in Xenopus. Conversely, overexpression of secreted Wnt antagonists anteriorizes Xenopus and zebrafish embryos (Itoh et al., 1995; Hoppler et al., 1996; Leyns et al., 1997; Deardorff et al., 1998; Glinka et al., 1998; Hsieh et al., 1999a; Fekany-Lee et al., 2000; Hashimoto et al., 2000; Shinya et al., 2000). A posteriorizing role for Wnts has also been suggested from studies on transgenic mice expressing Wnt8 (Pöpperl et al., 1997). Furthermore, mice mutant for Wnt3A (Takada et al., 1994), the nuclear transducers of Wnt signalling Lef1 and Tcf1 (Galceran et al., 1999) and the Wnt co-receptor LRP6 (Pinson et al., 2000) show posterior truncations, indicating a requirement for Wnt signalling in posterior structures. Finally, inhibition of Wnt signalling is required for formation of anterior neural structures as evidenced by interference with inhibitors of Wnt signalling: both inhibition of the secreted Wnt antagonist Dkk1 in Xenopus (Glinka et al., 1998; Kazanskaya et al., 2000), as well as inactivation of the Wnt repressors tcf3 and axin1 in the zebrafish headless and masterblind mutants, respectively, lead to microcephalic embryos (Kim et al., 2000; Heisenberg et al., 2001).

These properties make Wnts likely candidates for Nieuwkoop’s posteriorizing factor. Thus, we investigated whether an endogenous Wnt signalling gradient could act as transformer in AP neural patterning during gastrulation, i.e. before the well-established function of Wnts during the pattern-refinement phase of neural development (Patapoutian and Reichardt, 2000; Wilson and Rubenstein, 2000). We show that dose-dependent Wnt signalling is both necessary and sufficient for AP patterning of the neuraxis, and that it occurs in a direct and long-range fashion. Furthermore, we reveal an AP activity gradient of Wnt/β-catenin signalling across the presumptive neural plate. Our results implicate a morphogen gradient of Wnt/β-catenin signalling as a component of Nieuwkoop’s transformer.

XWnt8 and Fc-tagged MFz8-CRD proteins were produced as described (Hsieh et al., 1999b). XWnt8 protein concentration of conditioned media was determined by comparing its signal intensity produced in Western blot with signal intensities of known amounts of ³⁵S-labelled, Myc-tagged XWnt8 protein synthesized by in vitro translation.

Microinjections, conjugation and grafting of animal caps were performed in agarose-coated dishes containing 1× modified Barth’s solution (MBS). Microsurgical operations were performed using eyebrow knives. For ectodermal grafts (Fig. 5, Fig. 6), a small piece of ectoderm was cut from the pigmented donor embryo, and a hole of the same shape and size was cut into the presumptive neural plate of the albino host. The graft was positioned in the recipient’s hole with gentle pressure and transferred to culture dishes after ∼20 minutes of healing. Explants and operated embryos were cultured at 15°C in plastic dishes coated with HEMA (poly(2-hydroxyethylmethacrylate)) in 0.5× MBS. Explantation of early (stage 13) neuroectoderm for luciferase assays (Fig. 7C) was performed in 1× Ca²⁺- and Mg²⁺-free modified Barth’s solution (CMFB). The presumptive neural plate was excised and cut into four AP slices together with the underlying tissue. After few minutes, the ectoderm disadhered and could easily be separated from the underlying material. Ectodermal explants were collected in 1× MBS and processed for luciferase assay.

Concentrated XWnt8-conditioned and control media were dialyzed overnight at 4°C against 0.5× CMFB. Protein dilutions were prepared on ice using control medium, adjusted to 0.1 mg/ml heparin and kept at 15°C. For cell dissociation, 15 to 20 animal caps were collected per sample and transferred to a HEMA-coated tube containing 0.5 ml 1× CMFB + 2mM EDTA and incubated at room temperature for 5 minutes. Subsequently, the CMFB + EDTA was replaced by 0.2 ml of protein sample and the caps were resuspended after further incubation for 20 minutes by gentle pipetting using a HEMA-treated plastic tip. The tubes were incubated for 3 hours at room temperature in a horizontal position to prevent aggregation of the dissociated cells. Following treatment, the tubes were briefly centrifuged and the supernatant was replaced by 0.5 ml 0.5× MBS. The tubes were incubated in an upright position overnight at 18°C to allow reaggregation of the animal cap cells. When control siblings reached the desired stage, RNA was extracted from aggregates using Trizol (Gibco BRL).

RT-PCR analysis was carried out as described previously (Gawantka et al., 1995) using gene-specific primer pairs for Bf1 (Bourguignon et al., 1998), En2 (Hemmati-Brivanlou et al., 1991), H4, muscle actin, Otx2, sia, Xbra (Glinka et al., 1998), Krox20, NCAM2 (Hemmati-Brivanlou et al., 1994) and Xnr3 (Smith et al., 1995). Reporter gene assays on embryonic explants were performed using the Dual-Luciferase Reporter Assay System (Promega).

The expression construct for the activated form of Xenopus β-catenin lacking the N-terminal 90 amino acids was cloned by oligonucleotide-mediated ‘loop-out’ mutagenesis (Sambrook et al., 1989), using pCS2+Xβcat-GFP (Miller and Moon, 1997) to yield the plasmid pCS2+Xβcat*-GFP.

In situ hybridization was carried out using standard procedures (Gawantka et al., 1995; Hollemann et al., 1998b). All embryos and explants were photographed following immersion in 80% glycerol/”phosphate-buffered saline (PBS). The embryos in Fig. 5D were bleached by illumination in 30% H₂O₂/methanol/water (1:1:1). β-catenin staining on mid-gastrula embryo halves was performed as reported previously (Schneider et al., 1996) using a rabbit β-catenin antibody (generous gift from H. Steinbeisser). Stained halves were treated with Murray’s clear, mounted on glass slides and pictures were taken using a Zeiss Axioplan microscope, a Sony Power HAD 3CCD color video camera and the Adobe Photoshop software. Eight consecutive pictures were recorded along the AP axis within the presumptive neural plate, each window typically containing 25 to 40 cells within the analysed ectodermal layer. Grey values were determined for all nuclei using the NIH Image software, the grey value of the cytoplasmic background was subtracted and the resulting densities were averaged for individual pictures.

An important experimental approach used to reveal the function of TGFβs as morphogens during dorsoventral (DV) mesoderm induction was the use of dissociated Xenopus ectodermal (animal cap) cells which were treated with different doses of Activin protein (Green and Smith, 1990). Previously, similar experiments using Wnts were difficult because they could not be obtained in soluble form. Recently though, various sources of soluble, bioactive vertebrate Wnts have been described (Shibamoto et al., 1998; Hsieh et al., 1999b; Piccolo et al., 1999). We used a soluble XWnt8 produced as conditioned medium from Drosophila S2 cells (Hsieh et al., 1999b) to test its dose-dependent effects on neuralized animal cap cells. Animal caps were explanted at late blastula stage and dissociated in Ca²⁺- and Mg²⁺-free medium. This treatment characteristically results in anterior neural induction presumably owing to dilution of anti-neuralizing BMPs (Chang and Hemmati-Brivanlou, 1998a). Cells were treated with XWnt8-conditioned medium until control embryos developed to late gastrula stage. After reaggregation, the cells were further incubated without XWnt8 and harvested at neural tube stage for analysis of marker gene expression by RT-PCR (Fig. 1A).

Dissociation of animal cap cells lead to induction of the pan-neural marker NCAM2 (Kintner and Melton, 1987), the forebrain marker Bf1 (Bourguignon et al., 1998) and Otx2, a marker of fore- and midbrain (Blitz and Cho, 1995; Pannese et al., 1995), indicative of anterior neural induction (Fig. 1B). Treatment with increasing doses of XWnt8 lead to a progressive posteriorization of the neural character of these cells. At 2 nM XWnt8, Bf1 expression was repressed and Otx2 was maintained, while the mid-hindbrain boundary marker En2 (Hemmati-Brivanlou and Harland, 1989) became induced. At 20 nM XWnt8, these anterior neural markers were repressed and the hindbrain marker Krox20 (Bradley et al., 1993) became induced.

To confirm that this posteriorization was due to XWnt8 in the conditioned medium, we also performed XWnt8 treatment in the presence of MFz8-CRD, a secreted form of the ligand-binding domain of mouse Frizzled8, which binds to XWnt8 and acts as Wnt antagonist (Deardorff et al., 1998; Hsieh et al., 1999b). MFz8-CRD abolished the ability of XWnt8 to induce Krox20, indicating that posteriorization was due to Wnt signalling (Fig. 1C). Furthermore, posteriorization by XWnt8 was not accompanied by mesoderm induction, as indicated by the absence of muscle actin expression, and hence occurred in a direct fashion (Fig. 1B,C).

These results show that Wnt protein can posteriorize anterior neuroectoderm in a dose-dependent manner, specifying with increasing dose forebrain (Bf1), midbrain (Otx2, not Bf1), mid-hindbrain (En2) and hindbrain (Krox20), respectively. The modulation of marker gene expression occurs over a 10-fold concentration range.

We tested if Wnts also elicit dose-dependent responses in solid tissue and whether the responses occur as a function of the distance from a local Wnt source. We performed animal cap sandwich experiments similar to those used by Gurdon and colleagues (Gurdon et al., 1994). Animal caps were taken from albino embryos injected with noggin mRNA (Smith and Harland, 1992) and were used as anterior neuralized responding tissue. These caps were conjugated with animal caps from pigmented embryos injected with Xwnt3A mRNA (Christian et al., 1991) or Xwnt8 DNA as sources of posteriorizing Wnt protein (Fig. 2A). The conjugates were allowed to develop until control embryos reached neural plate stage and expression of neural markers was analysed by in situ hybridization. The use of albino and pigmented animal caps allowed easy identification of the Wnt-responding tissue. Cell mixing of pigmented with albino cells was negligible under these conditions, as shown previously (Gurdon et al., 1994), and this was also confirmed by lineage tracing with colloidal gold labelled caps (not shown).

Without Noggin, no neural induction occurred, even when Xwnt3A was expressed (Fig. 2B, a-a′′). Noggin injection resulted in induction of Bf1 throughout the responding albino cap but no expression of the posterior markers En2 or Krox20 was detectable (Fig. 2B, b-b′′). When pigmented caps injected with a low XWnt3A dose were used as Wnt source, Bf1 expression was typically restricted to the distal portion of the responding cap and was completely repressed at high XWnt3A dose (Fig. 2B, c,d). This was paralleled by induction of En2 and Krox20. Importantly, double in situ hybridization revealed that the posterior marker Krox20 was always expressed closer to the XWnt3A source than the more anterior marker En2, as would be expected if their expression requires different Wnt concentrations (Fig. 2B, c′′,d′′). Comparable results were obtained with injected Xwnt8 DNA (Fig. 2C) and mRNA (not shown). The results indicate that a local Wnt source can induce different AP neural fates at long range as a function of the distance in solid tissue.

Next we tested the requirement for Wnt signalling during AP neural patterning in vivo. To modulate Wnt signalling, embryos were injected either with low or high doses of Xdkk1 mRNA, encoding a secreted Wnt antagonist (Glinka et al., 1998), or with the plasmid pCSKA-Xwnt8 (Christian and Moon, 1993) to drive Wnt expression after the midblastula transition (MBT; Fig. 3A). Early neural marker gene expression, including Xanf1 (presumptive forebrain) (Zaraisky et al., 1992), Otx2 (presumptive fore- and midbrain), Gbx2 (presumptive hindbrain and anterior spinal cord) (von Bubnoff et al., 1996) and HoxD1 (presumptive posterior hindbrain and spinal cord) (Kolm and Sive, 1997), was analysed at late gastrula stage when AP neural patterning is already overt. Expression of a low XDkk1 dose shifted the expression of HoxD1 and Gbx2 posteriorly, and expanded Otx2 and Xanf1 ventroposteriorly (Fig. 3A, compare b-b′′′ with c-c′′′). At high XDkk1 doses, HoxD1 and Gbx2 became repressed, Otx2 remained expanded while Xanf1 showed further ventral expansion (Fig. 3A, d-d′′′). Comparable results were obtained with the secreted Wnt antagonist NXFz8 (Deardorff et al., 1998) (not shown). Conversely, injection of pCSKA-Xwnt8 left HoxD1 unchanged, expanded Gbx2, restricted Otx2 anteriorly and repressed Xanf1. These results (summarized in Fig. 3B) show (1) that Wnt signalling is required for expression of spinal cord and hindbrain markers already at gastrula stage, (2) that presumptive forebrain specification is inhibited by endogenous Wnt signalling in the ventroposterior, and (3) that endogenous Wnt signalling is not uniformly distributed but graded, with peak levels in the posterior (Fig. 3A, c) and lowest levels in the anterior (Fig. 3A, a′′). Thus, the presumptive neural plate is under control of posteriorizing Wnt signalling at late gastrula stage.

To investigate how this patterning manifests itself at later stages of neurulation we analysed marker gene expression in embryos injected with Xdkk1, NXfz8 or Xfrzb1 (Leyns et al., 1997) mRNAs or pCSKA-Xwnt8 at neural plate stage (stage 15; Fig. 4A). Overexpression of Xwnt8 repressed Bf1 and reduced Otx2 expression, presumably to midbrain fate, indicating posteriorization of the neural plate (Fig. 4A, a,b). With injected Wnt antagonists, two transformation thresholds were observed with increasing doses. At low doses, expression of Krox20 in rhombomere 5 was repressed but maintained in the more anterior rhombomere 3 (Fig. 4A, c-e). At high dose, both XDkk1 and NXFz8 repressed Krox20 (Fig. 4A, c′,d′), while residual Krox20 expression was observed with XFrzb1 (Fig. 4A, e′, arrowhead). The expression of Bf1 and Otx2 was expanded by all Wnt inhibitors. Interestingly, each Wnt antagonist altered the pattern of neural markers in a characteristic and reproducible manner. XDkk1 lead to a horseshoe-shaped expression of Bf1 and Otx2, indicating a posterior expansion (Fig. 4A, c′). By contrast, XFrzb1 resulted in a lateral, semicircular expansion of expression (Fig. 4A, e′). These differences may relate to the inhibitory specificities of these Wnt antagonists. The results (summarized in Fig. 4B) confirm a dose-dependent requirement for Wnt signalling with at least two thresholds for AP neural patterning.

In the previous experiments, whole embryos were microinjected with Wnts or Wnt antagonists which may affect Wnt signalling in all germ layers. This raises the question of whether the changes in AP neural patterning observed after Wnt modulation in vivo are due to a direct effect on neuroectoderm or if they occur indirectly via an effect on dorsal mesoderm. To specifically modulate Wnt signalling in neuroectoderm, we implanted small groups of animal cap cells from pigmented embryos injected with Xwnt3A or Xdkk1 mRNAs into the presumptive neural plate of albino embryos. The host embryos were allowed to develop until neural plate stage (stage 15) and analysed for marker gene expression (Fig. 5A). The pigmentation of the transplanted animal cap cells allowed again easy identification of the grafted donor tissue. Implantation of Xwnt3A-expressing cells into the presumptive forebrain induced ectopic Krox20 and En2 at some distance from the graft and with inverted polarity compared to the primary AP neuraxis (Fig. 5B, a,b). The transplant also initiated the outgrowth of host tissue, presumably because convergent-extension movements characteristic of posterior neural plate morphogenesis were induced. Thus, a local Wnt source can induce an AP pattern in neuroectoderm at long range. When Xdkk1-expressing cells were implanted into the neural plate, Bf1 expanded posteriorly in the vicinity of the grafted cells (Fig. 5C, a). Conversely, Krox20 expression became downregulated and shifted posteriorly (Fig. 5C, b). It is noteworthy that expression in rhombomere 5 was more sensitive to XDkk1 than in rhombomere 3 systematically, a phenomenon also observed in zebrafish (Shinya et al., 2000) and with other Wnt antagonists in Xenopus (Pera and De Robertis, 2000). This is consistent with the two thresholds of Wnt signalling defined for Krox20 in whole embryo injections (Fig. 4). The results indicate a requirement for different concentrations of Wnt signalling within neuroectoderm.

In order to test if also more posterior marker gene expression requires Wnt signalling and to rule out the possibility that the effect of Xdkk1-expressing implants is due to an indirect effect via underlying mesoderm, we grafted cells expressing the intracellular Wnt inhibitor Xgsk3β (Itoh et al., 1995) into the presumptive posterior neural plate. Uninjected control grafts expressed HoxD1 indicating that transplantation does not interfere with posterior neural gene expression per se (Fig. 5D, a,a′). By contrast, GSK3β grafts located in the presumptive spinal cord failed to express HoxD1, an effect best seen after bleaching of the pigment of the grafted cells (Fig. 5D, b′). This suggests that Wnt signalling is autonomously required for the specification of posterior (trunk) nervous system.

An important question raised by the preceding experiments is if the observed long-range effects (Fig. 2B,C, Fig. 5B) are due to a direct action of diffusing Wnt protein acting as morphogen or if they occur via a relay mechanism. To distinguish between a direct and an indirect mechanism, we asked whether an implant expressing an intracellular component of the Wnt pathway is also able to affect neural patterning. We chose a constitutively active form of β-catenin (Xβcat*) (Barth et al., 1997) to activate the Wnt pathway in a cell-autonomous fashion. First, we titrated Xβcat* and Xwnt3A mRNAs to activate Wnt/β-catenin signalling. In an axis duplication assay similar frequencies of induced secondary embryonic axes were achieved at an Xβcat*/Xwnt3A mRNA ratio of 10:1 (Fig. 6A). When this mRNA ratio was used in animal cap assays, both mRNAs induced the immediate early Wnt target genes siamois (sia) (Brannon and Kimelman, 1996; Carnac et al., 1996) and Xenopus nodal-related 3 (Xnr3) (McKendry et al., 1997) to similar levels by RT-PCR (Fig. 6B).

Implanted cells injected with XWnt3A repressed the forebrain marker Bf1 (Fig. 6C, a,b) and induced the posterior markers En2 and Krox20 ectopically at some distance from the implant in the host neural plate (Fig. 6C, a′,b′,a′′,b′′). By contrast, cells expressing Xβcat* induced no changes in marker gene expression in the host embryo (Fig. 6C, c-c′′) even at highest doses of Xβcat* (20 ng/animal cap; not shown). We conclude that the long-range effect of XWnt3A on AP neural patterning is direct.

Gbx2 and the posterior-dorsolateral neural marker Pax3 (Bang et al., 1997) were expressed in both, Xwnt3A- and Xβcat*-expressing grafts themselves, but not in control grafts (not shown). For Xβcat*, Gbx2 and Pax3 expression were not detectable in adjacent albino tissue, indicating cell-autonomous posteriorization. Thus, both types of grafted cells can acquire comparable AP positional values.

We have shown that dose-dependent Wnt signalling is both necessary and sufficient for AP patterning of neural fates and that Wnts affect neuroectoderm in a direct, long-range fashion. This raises the crucial question of whether there is indeed an endogenous Wnt/β-catenin signalling gradient in the presumptive neural plate that can be directly detected. To analyse endogenous Wnt/β-catenin signalling, we made use of the Wnt-responsive luciferase reporter construct TOP-Flash, carrying Tcf-binding sites, which are directly activated by the Tcf/β-catenin complex (Korinek et al., 1997). The reporter was microinjected into the four animal blastomeres of eight-cell embryos, whose descendants populate predominantly ectoderm. Injected embryos were divided into four AP slices at the end of gastrulation. A gradient of reporter activation was detected for TOP-Flash, increasing 2.4-fold from the anterior towards the posterior of the late gastrula (Fig. 7A). We did not observe this gradient of reporter activation with FOP-Flash, an altered TOP-Flash construct in which the Tcf-binding sites are inactivated by point mutations (Korinek et al., 1997), suggesting that the graded activation of TOP-Flash was due to proper activation of the reporter via Lefs/Tcfs. Because the animal blastomeres of eight-cell stage Xenopus embryos do not give rise to ectoderm exclusively, the graded TOP-Flash activation could still reflect a gradient of Wnt signalling within the mesendoderm rather than in the ectoderm. Thus, we isolated AP slices of presumptive neural plates and performed the reporter experiment with these neuroectodermal explants. In line with the results using whole embryos, a graded activation of TOP-Flash could be observed with the explants, increasing approx. twofold from anterior to posterior (Fig. 7B). Some representative neuroectodermal explants were assayed by in situ hybridization for expression of the paraxial mesodermal marker XmyoD (Hopwood et al., 1989). No XmyoD expression was detected in the explants confirming the absence of mesodermal contamination (not shown).

However, measuring a Wnt signalling gradient by a reporter assay may not reflect acute levels of pathway activation because of the accumulation of luciferase protein between MBT and late gastrula stage. Thus, we analysed Wnt signalling in a more direct way. Upon Wnt pathway activation, β-catenin is known to translocate to the nucleus (Wodarz and Nusse, 1998). To visualize Wnt pathway activation in situ, we therefore detected nuclear β-catenin immunochemically in the presumptive neural plate of mid-gastrula embryos. Because there are few AP morphological landmarks at this stage, we cut gastrulae sagittally into left and right halves, and analysed them in parallel by in situ hybridization for the early posterior marker Gbx2 and by β-catenin staining (Schneider et al., 1996; Rowning et al., 1997). Pictures of β-catenin staining were taken at eight consecutive AP levels within the presumptive neural plate (Fig. 8A). At the anterior-most levels (Fig. 8A,B, numbers 7 and 8), corresponding to future cement gland and forebrain territories, we detected no nuclear but only plasma membrane localization of β-catenin (Fig. 8B). Towards the posterior, β-catenin was seen increasingly localized to the nucleus. Densitometric analysis revealed a sixfold gradient of nuclear localization of β-catenin between the anterior- and posterior-most levels, with the region expressing Gbx2 showing intermediate to peak levels which are maintained up to the blastopore (Fig. 8C). We conclude that there is an AP gradient of endogenous Wnt/β-catenin pathway activation in the presumptive neural plate of the Xenopus gastrula.

Classical experiments have predicted gradients of posteriorizing factors involved in AP patterning of the early vertebrate neural plate. We provide evidence that the Wnt/β-catenin pathway constitutes one such signalling gradient. Our data show (1) that dose-dependent Wnt signalling is both necessary and sufficient for AP patterning of the neuraxis, (2) that Wnt/β-catenin signalling occurs in a direct and long-range fashion within the ectoderm, and (3) that there is an endogenous AP gradient of Wnt/β-catenin signalling in the presumptive neural plate of the Xenopus gastrula.

We present three lines of evidence that different doses of Wnt signalling are sufficient to provide AP positional information to neural cells. First, experiments employing soluble protein show that different doses of XWnt8 can specify at least three distinct neural fates at increasing dose. Second, in animal cap conjugates an organized AP neural pattern is induced by a local Wnt source. Third, Wnt expressing grafts induce an ectopic AP neural pattern in the anterior neuroectoderm of host embryos in a direct fashion. Evidence that Wnts may have graded effects on neural patterning also comes from overexpression of the Wnt effector Dishevelled in Xenopus animal caps, which leads to induction of increasingly posterior neural markers with increasing dose (Itoh and Sokol, 1997). Furthermore, we show by titrations with Wnt antagonists that different doses of Wnt signalling are required for AP neural patterning during gastrulation.

Soluble XWnt8 protein specifies forebrain, midbrain and hindbrain fates with increasing concentrations in animal cap cells neuralized by dissociation (Fig. 1). These inductions occur at 2-20 nM concentrations, which is consistent with the binding constant (∼9 nM) of Wnt8 for Frizzled receptors (Bhanot et al., 1996; Hsieh et al., 1999b). To our knowledge, this is the first demonstration of dose-dependent Wnt signalling effects with any Wnt protein. It should be noted, however, that neural induction of animal cap cells by dissociation may not be comparable with neural induction in the embryo, which may involve instructive FGF and early Wnt/β-catenin signalling (Baker et al., 1999; Harland, 2000; Streit et al., 2000; Wilson et al., 2000; Wilson et al., 2001). Yet, similar to mammalian stem cells, dissociated animal cap cells may provide a powerful system to reconstitute neural differentiation in vitro by applying recombinant Wnt protein in combination with other growth factors.

Thresholds to different doses of Wnt/β-catenin signalling are also suggested by animal cap conjugate experiments (Fig. 2), where an organized AP pattern is induced as a function of the Xwnt3A mRNA dose and of the distance from the Wnt source. XWnt3A and XWnt8 are both able to posteriorize neural fates in these conjugates at long range, suggesting that a variety of type I Wnts signalling via β-catenin (Wodarz and Nusse, 1998) are able to elicit these effects.

In vivo, we show that the presumptive neural plate of the late gastrula embryo is under control of a Wnt signalling gradient between future forebrain and spinal cord. This gradient is detected (1) by the pattern with which marker gene expression reacts after injection of Wnt antagonists and XWnt8 (Fig. 3), (2) by the graded activation of a Wnt-responsive reporter (Fig. 7), and (3) by the graded nuclear accumulation of β-catenin (Fig. 8). The orientation of the gradient, with peak levels in the posterior and lowest levels in the anterior, is consistent with the predicted properties of a transformer gradient proposed by Nieuwkoop and colleagues. In the late gastrula, the gradient appears to specify rather crude AP levels, because Gbx2 and HoxD1 are co-repressed at high XDkk1 doses. This crude pattern is refined at neurula stage when two Wnt thresholds for Krox20 expression in rhombomeres 3 and 5 can be detected. A similar refinement of an initially crude response to a morphogen, which gives rise to distinct thresholds at later stages, has been described for Activin (Green et al., 1994; Wilson and Melton, 1994) and may depend on secondary cell-cell communication.

Although the titrations with Wnt antagonists (Fig. 4, Fig. 5) reveal a dose-dependent requirement for Wnt signalling, the gradual expansion of more anterior markers such as Bf1 and Otx2 is in contrast to the thresholds observed in dissociated animal cap cells responding to different doses of XWnt8 protein (Fig. 1). For example, in the in vitro experiments, populations of cells expressing only Bf1 and Otx2 or only Krox20 are observed at different Wnt doses. Extrapolating from the in vitro results one might expect that in vivo the entire neural plate expresses Krox20 when Wnts, or Bf1 when Wnt antagonists are overexpressed. However, such a complete respecification of the neural plate is never observed in vivo, rather the changes in marker gene expression respect specific boundaries. This is most obvious for Bf1 in Fig. 4, whose anterior stripe of expression only becomes broader but not significantly thicker with increasing doses of Wnt inhibitors. Similarly, after implantation of Xdkk1-expressing cells as in Fig. 5C (panel a), ectopic Bf1 expression is never seen to extend into the midline but straddles the dorsal neural plate posteriorly. This suggests that factors in addition to Wnts restrict the competence of cells to express anterior markers. Obvious candidates for such modifiers are RA, Sonic Hedgehog, Nodals, BMPs and FGFs, which have been implicated in neural patterning (Lumsden and Krumlauf, 1996; Pownall et al., 1996; Sasai and De Robertis, 1997; Chang and Hemmati-Brivanlou, 1998a; Isaacs et al., 1998; Pownall et al., 1998; Gamse and Sive, 2000; Ribisi et al., 2000; Wilson and Rubenstein, 2000). To test if the inhibition of these factors abolishes this restriction of competence we implanted cells co-expressing Xdkk1 and either mouse Hip1, cerberus-short, noggin or FGF receptor-1 and FGF receptor-4a ectodomains into the neural plate, which encode inhibitors of Hedgehog, Nodal, BMP and FGF signalling, respectively (Zimmerman et al., 1996; Ye et al., 1998; Chuang and McMahon, 1999; Piccolo et al., 1999; Streit et al., 2000). None of these growth factor antagonists in conjunction with XDkk1 was able to expand Bf1 expression into the entire neural plate (not shown). Similarly, inhibition of RA signalling by injecting the host embryos with XCYP26 mRNA, encoding a RA degrading enzyme (Hollemann et al., 1998a), did not allow Xdkk1-expressing grafts to expand Bf1 expression further, despite a posterior shift of hindbrain markers induced by XCYP26 (not shown). The restriction of competence for anterior markers to become expressed posteriorly may occur via factors other than those tested. Alternatively, this/these factor(s) may be active before early gastrula, the stage of grafting. An ectodermal prepattern regulating the expression of Otx2 has also been identified in the zebrafish central nervous system. This initial AP prepattern depends on a differential competence of the epiblast and is not imposed by organizer-derived signals (Koshida et al., 1998). We observed the same phenomenon in our experiments, where Otx2 expression never expanded into the ventroposterior ectoderm even at high XDkk1 dose (Fig. 3A, d′′).

Clearly then, a morphogen gradient of Wnt/β-catenin signalling is not sufficient to account for all AP patterning of the neural plate and indeed other signalling molecules, such as RA (Ruiz i Altaba, 1993; Kolm and Sive, 1997) and FGFs (Slack and Isaacs, 1994; Cox and Hemmati-Brivanlou, 1995; Doniach, 1995; Kengaku and Okamoto, 1995; Lamb and Harland, 1995; Holowacz and Sokol, 1999; Ribisi et al., 2000) have been implicated in posteriorization. FGFs are potent posteriorizing agents and FGF8 mediates effects of the isthmic organizer (Pownall et al., 1996; Isaacs et al., 1998; Pownall et al., 1998; Rhinn and Brand, 2001). However, the posteriorizing effects of FGFs can be reversed with Wnt antagonists, suggesting that they may be mediated through Wnts at least partially (McGrew et al., 1997; Kazanskaya et al., 2000). A direct requirement for RA signalling during early neural patterning has been demonstrated in the hindbrain but not in anterior head structures (Blumberg et al., 1997; Kolm et al., 1997; Hollemann et al., 1998a; Gavalas and Krumlauf, 2000; Niederreither et al., 2000; Dupé and Lumsden, 2001). Thus, hindbrain and spinal cord patterning are regulated by FGF, RA and Wnt signalling.

A major finding of our study is the discovery of a Wnt/β-catenin signalling gradient across the presumptive gastrula neural plate. How is this gradient established? Several Wnts are expressed in the Xenopus gastrula, including Xwnt3A, Xwnt5A, Xwnt7B, Xwnt8, Xwnt8B and Xwnt11 (Fig. 9B, a). Xwnt3A and Xwnt8 are expressed in chordamesoderm and lateroventral mesoderm, respectively (Fig. 9A) (Christian and Moon, 1993; McGrew et al., 1997; Bang et al., 1999), both of which harbour posteriorizing activity in Xenopus, zebrafish and chick (Hemmati-Brivanlou et al., 1990; Doniach and Musci, 1995; Bang et al., 1997; Muhr et al., 1997; Woo and Fraser, 1997). As we show that Wnts are able to signal at long range, XWnt3A and XWnt8 may diffuse from these sources to set up a concentration gradient. Xwnt5A and Xwnt11, although expressed in the gastrula, are unlikely to be posteriorizing because they do not appear to signal via β-catenin (Ku and Melton, 1993; Moon et al., 1993; Sokol, 2000). Xwnt7B and Xwnt8B are both expressed maternally and may contribute to posteriorization (Cui et al., 1995; Chang and Hemmati-Brivanlou, 1998b). Thus, the signalling gradient probably involves multiple Wnts. A Wnt signalling-free zone is maintained by Wnt antagonists secreted within the anterior neuroectoderm and from adjacent anterior mesendoderm, including Cerberus, Dkk1, Frzb1, Crescent/Frzb2 and Sfrp2 (Bouwmeester et al., 1996; Leyns et al., 1997; Glinka et al., 1998; Piccolo et al., 1999; Bradley et al., 2000; Pera and De Robertis, 2000) (Fig. 9B, b). As these factors are potentially diffusible, they may also contribute to the establishment of a Wnt activity gradient. In conclusion, we suggest that the Xenopus AP neuraxis is patterned by a Wnt activity gradient formed in the presumptive neuroectoderm of the Xenopus gastrula by the interplay between posteriorizing Wnts and Wnt antagonists expressed in the anterior (Fig. 9C).

That multiple Wnts are involved in setting up the gradient is also suggested by the response of marker genes to different Wnt antagonists. Unlike XDkk1 and NXFz8, XFrzb1 failed to completely repress Krox20 expression. In addition, the expansion of Bf1 expression elicited by XDkk1 and XFrzb1 was reproducibly of different shapes, horseshoe versus semicircular, respectively (Fig. 4A). These differences probably reflect different specificities of the Wnt antagonists. For example, XFrzb1 is unable to inhibit signalling by XWnt3A, unlike XDkk1 (Wang et al., 1997; Krupnik et al., 1999; Kazanskaya et al., 2000).

The formation of the Wnt/β-catenin signalling gradient resembles formation of the BMP morphogen gradient involved in DV patterning. BMP2, BMP4, BMP7 and BMP7R are expressed in overlapping domains in the gastrula and an activity gradient is formed by the diffusion of multiple BMP antagonists secreted from Spemann’s organizer (Harland and Gerhart, 1997; Barth et al., 1999; Dale and Wardle, 1999). Thus, the BMP and Wnt morphogen gradients regulate patterning in the Xenopus gastrula, providing DV and AP positional information, respectively (Fig. 9D, a). Interestingly, a system of orthogonal Decapentaplegic and Wingless morphogen gradients regulate AP and DV patterning of the Drosophila wing imaginal disc, respectively (Fig. 9D, b) (Strigini and Cohen, 1999). This raises the possibility that orthogonal BMP-Wnt gradients represent an evolutionarily conserved module that specifies bilateral symmetry.

We are indebted to Drs J. Nathans and H. Steinbeisser for generous gifts of XWnt8-expressing insect cells and β-catenin antibody, respectively, and for advice. Drs H. Clevers, M. Eilers, J. Gurdon, R. Harland, D. Kimelmann, P. Klein, J. Nathans and R. Moon kindly provided reagents. We thank C. Rumig and P. Steinmetz for support with in situ hybridization and RT-PCR, and C. Walter for skillful assistance in insect cell culture. We are grateful to Drs G. Davidson, O. Kazanskaya and J. Wittbrodt for critical reading of the manuscript. C. N. was supported by the DFG.