Wise, a context-dependent activator and inhibitor of Wnt signalling

Signalling molecules play key roles in developmental events and their actions are highly regulated by endogenous modulators and antagonists in order to obtain precisely balanced outputs. The process of neural anteroposterior(AP) patterning involves integration of various signals such as retinoic acid(RA), fibroblast growth factors (FGFs) and members of the Wnt family. However,the relative roles of these cascades, the degree to which they are used at any particular axial level, and how they are integrated in organising normal AP patterning is poorly understood. Identification of factors that can modulate existing pathways or that represent novel signalling inputs will be beneficial to our understanding of how AP patterning is coordinated.

Analysis of neural patterning is complicated by the tissue interactions and dynamic morphogenetic movements that occur during gastrulation. Xenopus animal caps provide a simplified system for studying patterning events separate from morphogenetic movements. Animal caps alone form epidermis in culture, but when treated with antagonists of BMP signalling they can be induced to adopt an anterior neural fate(Hemmati-Brivanlou and Melton,1997). This anterior neural tissue is capable of altering its positional identity to a more posterior character under the influence of signals from tissues surrounding the neural tube or by ectopic application of posteriorising factors, such as RA, FGF and Wnt family members(Baker et al., 1999; Blumberg et al., 1997; Domingos et al., 2001; Itoh and Sokol, 1997; Kolm et al., 1997; Lamb and Harland, 1995; McGrew et al., 1997; McGrew et al., 1995; Pownall et al., 1996).

Experiments in Xenopus have shown that planar signals within the neuroectoderm and vertical signals from the underlying mesoderm work in concert to control regional identity of the nervous system(Doniach, 1993). Although early AP specification of the nervous system occurs during gastrulation, the neural cells are not irreversibly committed to a particular identity. Grafting experiments in several species have revealed that plasticity in regional character is retained after gastrulation(Cox and Hemmati-Brivanlou,1995; Gould et al., 1998; Grapin-Botton et al., 1997; Itasaki et al., 1996; Muhr et al., 1997; Trainor and Krumlauf, 2000; Woo and Fraser, 1997),suggesting that neural cells are actively receiving signals and communicating with surrounding tissues at later stages.

In this study, we performed a functional screen to search for novel factors derived from tissues surrounding the neural tube with the potential to alter the AP character of neuralised Xenopus animal caps. We have identified a novel gene, Wise, expressed in the surface ectoderm. Wise encodes a secreted protein that is capable of inducing posterior neural markers, and modulates the Wnt signalling pathway in a context-dependent manner. Our results provide a novel mechanism for modulating the Wnt pathway and support a role for Wnt signalling in the neural patterning process.

A cDNA library was made from stage 8-13(Hamburger and Hamilton, 1951)chick embryos using tissues surrounding the neural tube(Fig. 1A) from axial levels capable of inducing Hoxb4 expression in grafting experiments(Itasaki et al., 1996). Size-selected (>1 kb) cDNAs were directionally inserted into a modified 64T vector (Tada et al., 1998). The library contained 250,000 unamplified clones, and 50,000 of these were divided into 100 pools (500 clones per pool). For initial screening, 10 pools were mixed to prepare a single large DNA pool (5000 clones) used to synthesise capped RNA. Xenopus eggs were fertilised, cultured and injected as previously described (Jones and Smith,1999). Animal caps were cut at stage 8, and incubated until siblings reached stage 25. Explants were collected and analysed for specific markers using RT-PCR (Hemmati-Brivanlou and Melton, 1994). For explant recombination assays, separate sets of embryos were injected with either noggin and FIDx (Molecular Probes) or Wise and lacZ RNA. FIDx and lacZ were used as lineage tracers. Caps were cut at stage 8, combined and cultured for assay at stage 25. To detect subcellular localisation of endogenousβ-catenin, relevant RNAs were injected into the animal half of Xenopus embryos at the eight-cell stage. Animal caps were cut at stage 7 and fixed 10-15 minutes later when the cap edge was healed. Cryosections (16 μm) were then stained with an anti-β-catenin antibody(Sigma) and a secondary antibody conjugated with HRP.

Xenopus Wise was isolated from stage 25 embryos by RT-PCR. The degenerate primers used were: upstream,5′-GCTTT(C/T)AA(A/G)AA(C/T)GATGCCAC-3′; downstream,5′-GTGAC(T/C)AC(T/G/A)GT(T/G)ATTTTGTA-3′. Two different clones were identified, presumably resulting from the pseudotetraploid Xenopus genome. For each clone, a longer version covering the start codon was isolated from a Xenopus stage 35 cDNA library.

The dominant-negative Dishevelled construct DIX, specific to the canonical Wnt pathway (Axelrod et al.,1998) was made by creating a stop codon after amino acid 170(Glutamine) by PCR and sub-cloned into pCS2+. Tagged Wise constructs were generated by PCR, and their activity was confirmed by injection into Xenopus embryos with noggin and assayed for induction of en2 in animal caps. ΔE1-2 IgG (lacking EGF repeat 1 and 2) andΔE3-4 IgG (lacking EGF repeat 3 and 4) of human LRP6 were generated by fusing the extracellular domains E3-4 and E1-2 of LRP6(Mao et al., 2001) to the IgG Fc domain (Tamai et al.,2000). A FLAG tag was attached to chick Frizzled 1 extracellular domain (amino acids 1-199) by PCR. Other constructs were as previously published.

The relevant amounts of RNA injected per embryo were as follows. Initial screening: noggin (500 pg) and RNA from pools (12 ng). Fig. 1B: noggin (500 pg) and Wise (150, 300, 600, 1200 pg). Fig. 1C: Wise (30 ng). Fig. 1D,E: noggin (500 pg), Wise (600 pg) and lacZ (100 pg). Fig. 3: Wise (300-500 pg) and lacZ (50 pg); Wise morpholino (30 ng). Fig. 4A: Wise (chick, Xenopus) (500 pg). Fig. 4B-K: control or Wise morpholino (33 ng). Fig. 5A: noggin (500 pg), Wise (600 pg); ΔWnt8 (200 pg),Δ LRP6 (1 ng); ΔDsh(dd1) (1.2 ng), GSK3(500 pg) and LEFΔN (200 pg). Fig. 5B: noggin (500 pg), Wnt8 (50 pg), Wise (600 pg) and ΔFz8 (2 ng). Fig. 5C: noggin(500 pg), Wnt8 (600 pg), Wise (800 pg), ΔWnt8(800 pg). Fig. 5E-G: Tcf3 (300 pg), Wnt8 (25 pg), Wise (300 pg). Fig. 6A-C: Wnt8 (5 pg), Wise (200 pg) and ΔDsh (DIX) (1 ng). Fig. 6D: Wise (1 ng), Wnt8 (100 pg), Dsh (1 ng) and β-catenin (200 pg). Fig. 6E,F: tBR(900 pg) and Wise (50 pg). Fig. 7C: Wise (1 ng). Fig. 7D: Wise (0.5 ng, 1 ng).

Morpholino antisense oligonucleotides were designed against the beginning of the coding region of the two Xenopus Wise genes. The sequences were: 5′-AGCACTGGAGCCTTGAGACAACCAT-3′;5′-AGCAGTGAAGCCTTGAGACAACCAT-3′. A 1:1 mixture of these oligonucleotides were diluted in PIPES (5 mM) buffered water and used for injection. The control morpholino was;5′-CCTCTTACCTCAGTTACAATTTATA-3′, which is a generic control from the supplier (Gene Tools), designed against human β-globin. 30-60 ng of morpholino were injected into whole embryos, and 13-60 ng were injected into a dorsal-animal or ventral-animal blastomere to target the surface ectoderm.

To test the secretion of Wise protein, 15 oocytes were injected with RNA encoding HA-tagged Wise and incubated in a 96-well dish with 150 μl of OR2 medium (Wallace et al., 1973)+ 0.01% BSA for 2 days. Oocytes and the conditioned medium were collected separately and used for western blotting with an anti-HA antibody (Roche).

For immunoprecipitation of conditioned medium, 293 cells were transfected with DNA and the conditioned medium (opti-MEM, Gibco) was concentrated 20- to 40-fold by ultrafiltration. Wnt8-myc medium was collected from S2 cells as described (Hsieh et al.,1999). The medium was mixed and incubated overnight at 4°C. Immunoprecipitation was performed using protein A beads (Amersham) and wash-buffer 150 mM NaCl, 50 mM Tris-HCl (pH 7.5) and 0.1% Triton X-100 as previously described (Hsieh et al.,1999; Tamai et al.,2000).

The functional screen is based on the ability of noggin to induce neural tissue in Xenopus animal caps with an anterior character that in the presence of other factors can be converted to a more posterior character (Lamb and Harland,1995; Smith and Harland,1992). As a source of putative factors, we prepared a cDNA library from dissected tissue (somites, surface ectoderm and endoderm) from stages 8 to 13 (Hamburger and Hamilton,1951) chick embryos (Fig. 1A), which is known to alter AP character when grafted to an ectopic location (Itasaki et al.,1996). After co-injection of noggin RNA with pools of RNA from the library, the induction of posterior character was monitored by RT-PCR on animal caps (Fig. 1A) by assaying for expression of en2, Krox20 and Hoxb9, which mark the midbrain, hindbrain and spinal cord, respectively. Myosin was used as a marker for mesoderm induction to allow us to focus on pools that influence neural patterning in the absence of mesoderm. Successive rounds of subdivision and sib selection identified two clones with this inducing activity. One encoded a 5′-truncated version of β-catenin, consistent with data that the canonical Wnt pathway induces posterior neural markers(Domingos et al., 2001; McGrew et al., 1995). The second clone encoded a novel protein, which we have designated Wise(Wnt modulator in surface ectoderm) based on this study. Injection of noggin with increasing amounts of Wise induced progressively more posterior markers (en2 and Krox20) in the absence of mesoderm (myosin) (Fig. 1B). Wise alone exhibited no neural- or mesoderm-inducing activity, as confirmed using NCAM, myosin (Fig. 1B), brachyury, wnt8 and Xhox3 (data not shown)as markers.

The predicted Wise protein consists of 206 amino acids and contains a cysteine knot-like domain found in a number of growth factors, as well as in Slit, mucin and CCN (Cef10/Cyr61, CTGF and Nov) family members(Bork, 1993)(Fig. 2A). Among these, the C-terminal domain of the CCN family members showed the highest homology to Wise, but other motifs conserved within the CCN family were absent in Wise(Fig. 2B). Hence, Wise is related to but not a member of the CCN family. A homology search revealed that Wise showed the highest amino acid identity (38%) to Sclerostin (SOST),identified by positional cloning of the gene mutated in sclerosteosis(Brunkow et al., 2001). There are a number of EST sequences homologous to Wise in zebrafish, mouse and human databases (Fig. 2A), but none was found in the Drosophila or C. elegans genomes.

A signal sequence motif is present at the N terminus of Wise, and its secretion was confirmed by western blotting after expression of an HA-tagged version of the protein in Xenopus oocytes(Fig. 1C) and COS cells (data not shown). We tested the ability of Wise to posteriorise neural tissue in a cell non-autonomous manner by using a tissue recombination assay in which a Wise-expressing animal cap was combined with a noggin-expressing animal cap. We found that both en2 and Krox20 were induced in discrete domains in the noggin caps (Fig. 1D,E). Hence, Wise has the ability to induce posterior markers at a distance.

In situ hybridisation analyses revealed that Wise is highly expressed in the surface ectoderm of the embryo in a dynamic pattern. In chick, expression was first detectable broadly at stage 9, and then localised in the posterior surface ectoderm overlying the presomitic mesoderm by stage 10-11 (Fig. 1F,G). This expression resolved into a small posterior domain in the tail bud by stage 13(data not shown). At later stages, Wise was expressed in other tissues such as branchial arches, limbs and feather buds (data not shown). By cloning the Xenopus homologue, we found that transcripts were first detected at the gastrula stage (Fig. 1H), and in comparison to chick more extensively along the AP axis in the surface ectoderm at the neurula stage(Fig. 1I). Other localised expression at neurula and tailbud stages was seen in the stomodeal-hypophyseal anlage (Fig. 1J-L), the ventral diencephalon (Fig. 1N) and presumably cranial and lateral line placodes(Fig. 1J,M,N). Some of the corresponding patterns were seen in chick embryos as well (data not shown).

The function of Wise was first studied in Xenopus embryos using overexpression approaches. Injection of Wise RNA in amounts >200 pg into the whole embryo lead to gastrulation defects and loss of eyes (data not shown). At lower amounts (<100 pg), gastrulation proceeded normally,but neural tube closure was abnormal and the neural plate appeared thicker and shorter than controls (data not shown). To evaluate further the effects of Wise on development of the neural tube, RNA or DNA was injected into specific blastomeres at 4-16 cell stages. When Wise RNA injections were targeted to presumptive neural regions, lateral expansion of the neural plate on the injected side was observed(Fig. 3A,B). AP specific markers (en2 and Krox20) were generally displaced laterally and posteriorly (Fig. 3D,E,G,H). When Wise injection was targeted to the forebrain region, ectopic expression of Krox20 and slug was observed (Fig. 3H,K). This indicates that forebrain cells acquired a more posterior character in response to Wise. Other defects included a failure in eye and cement gland formation(Fig. 3M,N, and data not shown). Conversely, when Wise RNA was injected ventrally, ectopic cement glands were induced (Fig. 3P,Q). Identical results were obtained using DNA constructs for injection, where Wise expression commenced at mid-blastula stages under the control of a cytoskeletal actin promoter (data not shown). Thus,ectopic expression of Wise altered aspects of AP patterning in embryos, as well as in animal caps.

To analyse the role of Wise in normal Xenopusdevelopment, we injected antisense morpholino oligonucleotides that specifically interfere with translation of Wise(Fig. 4A). By injection of morpholino oligonucleotides into both blastomeres at the two-cell stage,embryos became cyclopic (Fig. 4B-E). Localised injection of the morpholino into one of the dorsal animal blastomeres at four to eight cell stages resulted in a reduction of neural tissue (Fig. 3C),decreased amount of slug-positive neural crest cells(Fig. 3L), and the formation of a smaller eye (Fig. 3O) on the injected side. AP neural markers such as en2 and Krox20 and a cement gland marker XCG were not obviously affected(Fig. 3F,I,R). In the forebrain, Wise morpholino oligonucleotides caused loss of the olfactory placode revealed by Emx2 staining(Fig. 4F-I). Histological sections at the trunk level of neurula stage embryos showed that Wise morpholino injection caused thicker surface ectoderm(Fig. 4J,K). These results indicated that Wise has endogenous roles in controlling eye,olfactory placode and surface ectoderm formation. The strong expression of Wise in cranial placodes at the tailbud stage(Fig. 1M,N) further suggests a key role for Wise in placode formation. Wise is likely to play a permissive role in the eye development, as the eye tissue shows only a small domain of expression at the tailbud stage(Fig. 1M), and depletion of Wise by morpholino injection affects only growth of the eye and not the initial formation of the eye primordium, as marked by Xrx1(Fig. 3O).

The fact that ectopic expression of Wise RNA and depletion of endogenous translation by the morpholino oligonucleotides both result in similar defects in eye formation (Fig. 3N,O; Fig. 4L)suggests that this process requires a precise level of signalling mediated by Wise. Defects in eye formation are also observed following injection of molecules that activate Wnt signalling (Wnt8 DNA, Frizzled3)(Christian and Moon, 1993; Rasmussen et al., 2001) as well as molecules that inhibit the pathway (Frzb1, Crescent)(Pera and De Robertis, 2000). Other aspects of the phenotypes observed by altering Wise expression are also reminiscent of those seen when the Wnt signalling pathway is perturbed. For example, ectopic induction of cement gland is observed following injection of GSK3β (Itoh et al., 1995), and proper gastrulation and convergent extension movements require Wnt11 and Dishevelled(Heisenberg et al., 2000; Tada and Smith, 2000; Wallingford et al., 2000). Wnts/β-catenin function as posteriorising factors in animal cap assays(Domingos et al., 2001; McGrew et al., 1995), and also regulate induction of neural crest cells(Saint-Jeannet et al., 1997; LaBonne and Bronner-Fraser,1998; Garcia-Castro et al.,2002). These similarities prompted us to test whether Wise acts through or modulates the Wnt signalling pathway.

As Wnts and Wise both induce en2 expression in noggin-injected animal caps, we investigated whether the ability of Wise to induce en2 requires Wnt signalling. To test this, Wise RNA was co-injected with blocking reagents of the Wnt pathway such as wild type GSK3β (Dominguez et al., 1995) and dominant-negative (dn) versions of Wnt8(Hoppler et al., 1996),Frizzled8 (Itoh and Sokol,1999), LRP6 (Tamai et al.,2000), Dishevelled (Sokol,1996) or Lef1 (Vleminckx et al., 1999). These reagents either eliminated or attenuated the ability of Wise to induce en2 in neuralised animal caps(Fig. 5A,B). With respect to the intracellular component Dishevelled, only dominant-negative constructs affecting the canonical Wnt pathway abolished en2 induction (data not shown). Wnt8 and Wise showed an additive effect in induction of en2(Fig. 5C). To confirm the activation of the canonical Wnt pathway by Wise, we examined its effects in other assays. First, in animal cap explants in the absence of noggin,Wise showed a weak dorsalising activity by inducing siamois and Xnr3, two direct targets of the Wnt signalling pathway(Brannon et al., 1997)(Fig. 5D). Second, Injection of Wise RNA increased nuclear accumulation of β-catenin in animal caps, in a manner similar to that of Wnt8. This phenomenon was enhanced by co-injection of Tcf3, a co-factor of β-catenin for transcriptional activation(Fig. 5E-G). As animal caps were assayed before mid-blastula transition, we believe the observed effect is not due to secondarily induced transcription. These data suggest that Wise activates Wnt signalling and requires components of the canonical pathway to induce the signal in animal caps.

It is important to note that, although Wise and Wnts both activate the canonical pathway, there are distinct differences in their outputs. Wnt8 RNA (50 pg) is sufficient to induce both en2 and Krox20 (Domingos et al.,2001), and a higher amount (600 pg) induces only Krox20(Fig. 5C). By comparison, it takes much higher amounts of Wise RNA (300-600 pg) just to induce en2, and 1.2 ng of Wise RNA is only sufficient to weakly induce Krox20 (Fig. 1B). Similarly, Wnt8 robustly induces siamois and Xnr3 at a low amount of RNA (100 pg), although Wise only induces these genes in a relatively weak manner, even at the highest levels of RNA (100-1000 pg)(Fig. 5D). These results show that Wise has weaker posteriorising and dorsalising activities in comparison with Wnt8, raising the possibility that there are both quantitative and qualitative differences in the outputs of the Wnt pathway when activated by these two proteins. The fact that Wnt8 induces dorsal mesoderm in animal caps in the presence of noggin(Domingos et al., 2001)although Wise does not (Fig. 1B), further supports their qualitative difference.

Although induction of en2 can be explained by activation of Wnt signalling, the effects of injected Wise RNA on cement gland induction (Fig. 3Q) resemble those observed when the Wnt pathway is inhibited(Itoh et al., 1995). Thus, it is possible that in some contexts Wise blocks Wnt signalling. When Wnt8 RNA is injected into a ventral vegetal blastomere at the four to eight cell stage, it induces an ectopic secondary axis(Smith and Harland, 1991; Sokol et al., 1991)(Fig. 6A). Based on the ability of Wise to induce siamois and Xnr3, we expected that Wise on its own might induce a secondary axis or work in synergy with Wnt8 in this process. However, injection of Wise RNA did not exhibit secondary axis formation (data not shown). Rather, co-injection of Wise RNA completely blocked Wnt8-induced secondary axis formation(Fig. 6B), as did a dominant negative form of Dishevelled (Fig. 6C). This inhibitory activity was confirmed at the molecular level in ventral marginal zone explants by demonstrating that the Wnt-dependent induction of siamois and Xnr3 is greatly reduced by co-injection of Wise (Fig. 6D). These results suggest that in the presence of both Wnt8 and Wise, Wise interferes with the level of activity of Wnt8. Wise had no effect on the ability of injected intracellular components such as Dishevelled and β-catenin to induce Xnr3 and siamois(Fig.6D), suggesting that Wise functions extracellularly to interfere with the canonical Wnt pathway. The inhibitory effect of Wise on Wnt signalling was further examined by assaying secondary head induction, which can be induced by simultaneous inhibition of both BMP and Wnt signalling (Glinka et al., 1997) (Fig. 6E). Co-injection of Wise and a dominant-negative BMP receptor(Suzuki et al., 1994) induced a complete secondary axis with eyes and cement gland(Fig. 6F), demonstrating that Wise functions as a Wnt inhibitor in this context. This contrasts with our analysis in animal caps, which reveals that Wise does not interfere with the action of Wnt in the induction of Krox20(Fig. 5C). Therefore,modulation of the Wnt pathway by Wise (activation or inhibition) varies with respect to both target genes and cellular contexts.

Although the activating and inhibiting properties of Wise in animal caps and embryos described above are dependent upon the canonical Wnt pathway, it is possible that Wise also influences the planar cell polarity (PCP) pathway that branches at Dishevelled. Wnt11 is required for proper convergent extension movements of mesoderm during gastrulation in Xenopus and Zebrafish, and this has been shown to be dependent upon the PCP pathway of Wnt signalling (Heisenberg et al.,2000; Tada and Smith,2000; Wallingford et al.,2000). Animal caps cultured in the presence of activin form mesoderm, and undergo convergent extension movements that can be blocked by reagents that either elevate or decrease Wnt signalling(Tada and Smith, 2000). This implies that precise levels of Wnt signalling through the PCP pathway are essential for coordinated cell movements.

In this animal cap assay, injection of Wise RNA blocked cell movements preventing elongation of animal caps, but had no affect on activin-induced mesoderm formation (Fig. 7). This suggests that Wise might influence the Wnt-dependent PCP pathway, but whether this is mediated by its ability to activate or inhibit the pathway cannot be distinguished. This effect on cell behaviour in animal caps is consistent with and may explain the phenotypic effects observed in Wise-injected whole embryos. Wise perturbed the morphogenesis of the neural tube, which failed to close. It was thickened and shortened, and there was a lateral expansion, broadening or posterior-shift of AP markers. Many of these defects appear to relate to abnormal convergent extension movements during gastrulation. However, the fact that morpholino antisense oligonucleotides do not interfere with gastrulation (Fig. 4D) and neural AP patterns(Fig. 3F,I), and that Wise is not predominantly expressed at gastrula stage(Fig. 1H), both suggest that endogenous Wise is unlikely to be involved in normal gastrulation movements. This assay suggests that Wise has a potential to interfere with Wnt-mediated PCP as well as the canonical pathway.

Wise encodes a secreted protein and interacts with the Wnt pathway extracellularly (Fig. 5A,B; Fig. 6D). Therefore, to begin to approach the mechanisms of action, we investigated potential physical interactions of Wise with Wnt family members or their putative co-receptors Frizzled8 (Hsieh et al., 1999)and LRP6 (Tamai et al., 2000)(Fig. 8). We mixed conditioned medium of 293 or S2 cells containing a secreted form of LRP6(Tamai et al., 2000) or Frizzled8 (Hsieh et al., 1999)with Wise conditioned medium, and assayed for interactions by immunoprecipitation (IP). In this assay, Wise bound to LRP6 but not to Frizzled8 (Fig. 8A) or Wnt8(Fig. 8B). Recent studies have shown that individual members of the Dickkopf (Dkk) family of secreted proteins can either antagonise or stimulate Wnt signalling through interaction with LRP6 (Brott and Sokol,2002; Mao et al.,2001; Wu et al.,2000). Therefore we performed IP experiments to determine if Wise shares common binding sites with Dkk1 or Wnt on LRP6. The extracellular domain of LRP6 contains four EGF repeats and Dkk1 interacts with repeats 3-4, while Wnt interactions seem to involve mainly repeats 1-2(Mao et al., 2001). We found that Wise binds to LRP6 and a variant where EGF repeats 3 and 4 are deleted(ΔE3-4), but not to one in which EGF repeats 1 and 2 are removed(ΔE1-2) (Fig. 8A). Conversely, Dkk1 binds to LRP6 and ΔE1-2, but not to ΔE3-4(Fig. 8A). These results show that Wise shares the domain on LRP6 essential for interaction with Wnts and that Wise and Dkk1 modulate LRP6 activity by interacting through different domains. We also tested whether Wise and Wnt8 can bind to LRP6 at the same time or whether they compete for binding. As shown in Fig. 8C, Wise interferes with the binding of Wnt8 to LRP6. This suggests a mechanism whereby Wise inhibits Wnt signalling by competing with Wnt8 for binding to LRP6(Fig. 8D).

In this study, we have identified and characterised a novel secreted protein, Wise, that modulates the Wnt signalling pathway. Wise is conserved in vertebrates and contains a cysteine knot-like domain, present in members of the CCN family and other growth factors. Wise is expressed in the surface ectoderm, in which Wnt signalling plays multiple roles. The ability of Wise to affect gastrulation, neural tube morphogenesis and AP patterning in Xenopus explant and embryonic assays are consistent with its functioning to alter Wnt signalling. The novel aspect of Wise compared with other extracellular Wnt modulators is that it both activates and inhibits Wnt signalling in different contexts. In some contexts, Wise stimulates the canonical Wnt pathway, whereby it acts through Dishevelled andβ-catenin. The ability of Wise to interact physically with LRP6 by sharing the same binding domain with Wnt8 suggests that it might function as an alternative ligand for the receptor. In other contexts, Wise antagonises Wnt signalling through the canonical pathways presumably by blocking accessibility of certain Wnts to their receptors. The ability of Wise to both activate and inhibit Wnt signalling provides a new mechanism for modulating Wnt signalling and adds a new level of complexity to how the array of ligands and inhibitors are integrated to control the balance of Wnt signalling in different developmental contexts.

The ability of Wise to interact with the Wnt pathway and the fact that it is normally expressed in a transient manner in the non-neural surface ectoderm, suggest that it might have a role in modulating Wnt signalling in this tissue. Morpholino antisense oligonucleotide against Wise caused thick surface ectoderm (Fig. 4K), and overexpression of Wise by RNA injection caused expanded neural plate(Fig. 3B). These results suggest that endogenous expression of Wise at the edge of the neural plate might regulate the balance of neural and non-neural ectoderm transition. It is known that a balance between Wnt and BMP signalling in the surface ectoderm and dorsal neural tube is important in modulating dorsal fates and the generation of neural crest cells(Dickinson et al., 1995; Garcia-Castro et al., 2002; Liem et al., 1995; Trainor and Krumlauf, 2002). Furthermore, Wnts in the surface ectoderm influence patterning of the underlying somites and their derivatives(Capdevila et al., 1998; Munsterberg et al., 1995). The distribution and timing of Wise expression in the surface ectoderm together with the result of morpholino experiments suggest that it promotes precise levels of Wnt signalling to control some of these interactions.

Wise was isolated on the basis of its ability to alter AP neural patterning, an activity consistent with its interaction with the Wnt cascade. Recent studies provide us with evidence on involvement of Wnt pathway in AP patterning (Davidson et al.,2002; Domingos et al.,2001; Kiecker and Niehrs,2001). Furthermore, an increasing number of extracellular and intracellular inhibitors of Wnt signalling have been found, highlighting the considerable complexity in the nature of regulating this cascade. The roles for Wise in Wnt signalling raises the possibility that other members of this class of cysteine-knot proteins may also exert some of their functions by modulating Wnt activity. Indeed, the CCN family member Cyr61 is also capable of regulating Wnt signalling, although its mode of action is unknown(Latinkic et al., 2003). Wise is distinguished from other Wnt modulators as it seems to have multiple roles in modulating and integrating the readout of Wnt signalling depending upon the local context.

Even though Wise requires the canonical Wnt pathway to posteriorise noggin-treated animal caps, there are differences in the patterns of induction compared with stimulation by Wnt ligands such as Wnt8. Wise induces en2 at low levels of injected RNA, and en2 plus Krox20 only at high levels (Fig. 1B). By contrast, en2 and Krox20 are simultaneously induced by Wnt8, even with very low amounts of RNA, and Wnt8 can induce more posterior Hox genes such as Hoxb9(Domingos et al., 2001). Another difference is that although Wnts or β-catenin downregulate forebrain markers (Otx2, BF1) at the same time as inducing posterior genes (McGrew et al., 1997; McGrew et al., 1995) Wise does not (Fig. 1B). The basis of these differences is not clear. It has been suggested that en2 is under the direct regulation of Tcf3(McGrew et al., 1999), which seems to reflect processes in the isthmic region, where Wnt1 is required for expression of en2 (Danielian and McMahon, 1996; McMahon et al.,1992). However, the activation of other downstream targets, such as Krox20 and Hoxb9, could be indirect and involve multiple steps downstream of the direct action of β-catenin and Lef/Tcf. For example, Wise does not induce mesoderm in the presence or absence of noggin, whereas Wnts or β-catenin do(Sokol, 1993). As mesoderm can influence AP patterning, the differences associated with induction by Wnt might in part be mediated indirectly through mesoderm. Hence, even though Wise and Wnts stimulate the same pathway, there are differences in the nature of their outputs and Wise appears to be a much weaker inducer of posterior genes.

Recent studies on secreted proteins that affect Wnt signalling suggest complex mechanisms modulating the canonical Wnt pathway. Different Frizzled-related protein and Dkk family members exhibit opposite effects in a variety of in vivo and in vitro assays(Bradley et al., 2000; Brott and Sokol, 2002; Li et al., 2002; Mao and Niehrs, 2003; Wu et al., 2000). The activation of the Wnt pathway by Wise is either weaker or different than that seen using Wnts because it takes higher concentrations of Wise to induce en2 and Krox20 and the relative levels of induction of siamois and Xnr3 are much lower. With respect to inhibition,in the presence of both Wnt and Wise, Wise competes with Wnts for the binding to LRP6. This could result in either a less efficient activation of the receptors, which masks Wnt dependent activity, or a complete block of receptor activity (Fig. 8D). It is also possible that Wise could affect Wnt signalling through additional mechanisms. The interaction of Wise with LRP6 may also interfere with the function of Dkks, which could result in either activation or inhibition of the Wnt pathway depending upon which Dkk family member is present. There remains the possibility that Wise interacts with other receptors or modulators that work through intracellular Wnt signalling components. We observed that Wise interferes with cell movements in activin-treated animal caps(Fig. 7), consistent with the gastrulation defects observed in Wise-injected whole embryos. As the pathway involved in cell movements does not appear to require LRP6(Semenov et al., 2001), this result implies that Wise could interact with other proteins for its function. The studies presented here reveal new mechanisms through which a fine balance in Wnt signalling is regulated in various developmental processes.

We thank the following people for probes and constructs: R. M. Harland(noggin), M. Tada (modified p64T), C. Tabin (Fz1), R.T. Moon (Wnt8, dnWnt8,Wnt8-myc), S.Y. Sokol (dsh, dd1, Fz8, Fz8ECD, GSK3β), N. Papalopulu(Sox3), E. Boncinelli (Emx1,2), G Barsacchi (Xrx1), A. Hecht (LefΔN),J.-C. Hsieh (Fz8-IgG, Wnt8-myc), X. He (LRP6 IgG), C. Niehrs (Dkk1-FLAG, LRP6ΔE1-2, ΔE3-4) and T.J. Mohun (Xenopus cDNA library). We specifically thank M. Tada, J.-C. Hsieh, C. Niehrs, X. He and R. M. Harland for technical suggestion and discussion. N.I. was supported by HFSP, and PMD by Gulbenkian Foundation. This work was funded by Core MRC Programme support to R.K. and J.C.S. and an EEC Biotechnology Network grant (#BIO4 CT-960378) to R.K. GenBank Accession Numbers: chick Wise, AY319926; Xenopus Wise, AY319927, AY319928.