Anterior neural development requires Del1, a matrix-associated protein that attenuates canonical Wnt signaling via the Ror2 pathway

During early embryogenesis, vertebrate embryos are patterned along the dorsal-ventral (DV) and anterior-posterior (AP) axes. The embryonic regions are subdivided along the AP axis into three major domains: the head, trunk and tail. The classic ‘Einsteck’ studies in amphibian embryos demonstrated that these three domains are distinctly induced in a temporally controlled fashion by the actions of the early, intermediate and late Spemann organizers (reviewed by Gerhart, 2001; Mangold, 1933; Spemann, 1931).

In the two-signal model proposed by Nieuwkoop (Nieuwkoop, 1952a; Nieuwkoop, 1952b) (reviewed by Sasai and De Robertis, 1997), the forebrain is the default fate induced by the neural inducer, whereas more caudal neural tissues are secondarily transformed by caudalizing factors. Signaling factors such as Wnt, FGF, Nodal and retinoic acid cause the posteriorization of neural tissues and suppress head formation (reviewed by Rallu et al., 2002). In particular, canonical Wnt signaling has a crucial and conserved role in suppressing anterior embryonic development in vertebrate embryos (reviewed by Niehrs, 2004; Lewis et al., 2008). Although the microinjection of Wnt or β-catenin mRNA, which mimics maternal Wnt activation before the mid-blastula transition (MBT), induces the formation of a secondary axis (Molenaar et al., 1996; Sokol et al., 1991; Steinbeisser et al., 1993), the activation of Wnt signaling after the MBT (e.g. via plasmid injection) causes a strong posteriorization in the embryo and also in the neuralized animal cap (Christian and Moon, 1993; Kiecker and Niehrs, 2001). Moreover, the inhibition of endogenous Wnt signals by extracellular Wnt antagonists, such as Dkk1, Frzb and WIF1, induces substantial anteriorization, resulting in the formation of a large head (Glinka et al., 1998; Hsieh et al., 1999; Leyns et al., 1997). Wnt signaling in the head is also attenuated by non-secreted factors. Shisa, an anteriorly produced protein located in the endoplasmic reticulum (ER), inhibits the transport of the Wnt receptor Frizzled from the ER to the plasma membrane (Yamamoto et al., 2005). The headless mutant in zebrafish is caused by a mutation in the tcf3 gene, which is expressed in the head and inhibits Wnt signals (Kim et al., 2000). In Xenopus, Tcf3 expression in the head is regulated by the forebrain-specific zinc-finger protein XsalF (Onai et al., 2004).

In this study, we reveal an essential role for the extracellular matrix (ECM) protein Developmental endothelial locus-1 [Del1 (Hidai et al., 1998); also known as EGF-like repeats and discoidin I-like domains 3 (Edil3)] in the negative control of zygotic Wnt signaling to permit proper anterior development of the early Xenopus embryo. Del1 is a secreted protein that belongs to the Discoidin domain family (Kiedzierska et al., 2007); it is localized to the ECM and the basement membrane (Aoka et al., 2002). Previous studies have shown that mammalian Del1 is important in vascular development (Zhong et al., 2003), hair regrowth (Hsu et al., 2008), leukocyte-endothelial adhesion (Choi et al., 2008) and phagocytosis (Hanayama et al., 2004). Most of these activities of Del1 require its integrin-binding RGD motif, which is located in the second EGF domain.

We previously found that Del1 interferes with BMP signaling in the DV patterning of the gastrula marginal zone, albeit moderately (Arakawa et al., 2007). Interestingly, during our subsequent study of the effects of Del1 on ectodermal development, we noticed that some of the gain- and loss-of-function phenotypes of Del1 in Xenopus could not be explained simply by BMP modulation. In this study, we demonstrate that Del1 strongly regulates intracellular Wnt signaling during the anterior specification of the neuroectoderm. Del1 is required for forebrain development and interferes with the canonical Wnt pathway downstream of the nuclear accumulation of β-catenin. In addition, we provide evidence that the Ror2 (Receptor tyrosine kinase-like orphan receptor 2) pathway, which is implicated in the non-canonical Wnt pathway, plays an essential role in the inhibitory effects of Del1 on canonical Wnt signaling. Thus, Del1 is an ECM protein that induces a unique intracellular signaling pathway that is antagonistic to both the Wnt and BMP pathways.

Xenopus laevis embryos were obtained, cultured, microinjected and subjected to whole-mount in situ hybridization as described previously (Arakawa et al., 2007). As needed, Venus mRNA (100 pg) was co-injected as a tracer. For all animal cap assays, embryos were injected animally at the 4- to 8-cell stage. Unless indicated otherwise, animal caps were excised at stage 8 and cultured in 1× LCMR (low calcium/magnesium Ringer's) containing 0.2% BSA until siblings reached stage 11. For dissociated animal cap assays, excised animal caps were dissociated in CMFM (calcium/magnesium-free medium) containing 0.2% BSA. After removal of the outer layer, the dissociated inner layer cells were cultured in CMFM. For the Topflash assays, embryos were co-injected with 25 pg of Topflash reporter (Upstate) and 1 pg of phRL-null (Promega) plasmids, and subjected to luciferase assays using the Dual Luciferase Reporter Assay system (Promega), according to the manufacturer's protocol. All explant experiments were repeated at least three times. All recombinant proteins were obtained from R&D Systems.

Total RNA was isolated from embryos or explants using the RNeasy Mini Kit (Qiagen). RT-PCR and quantitative RT-PCR (qPCR) were performed as described previously (Inomata et al., 2008; Onai et al., 2004; Schambony and Wedlich, 2007; Sasai et al., 1995). Additional primers used in this study were (5′ to 3′): Six3 fwd CTCTCTCCCTTTTACTTTCTCACAC and rev GATAGAGGGTTA AACAAAGTTGCAT; Krox20 fwd GATTCAGATGAGCGGAGTGA and rev CAAGGGGTAGTTGGACGAGT; En2 fwd GTGTCAGCAAAGAGGACAAGAG and rev CCTCTGCTCAGTCAAATACCTG; NCAM fwd ATTCCCCAACTGGTGAGAAG and rev CCTTCAGATTCTCCCTCTGC; Rx2a fwd CAAACTTGAAACTAGGTCTCTGTGA and rev TGTGTTGATCTCCAGCTTTATTTAG; Otx2 fwd TTTCACCAGAGCTCAACTGG and rev CTGGACTCGGGTAGGTTGAT; HoxD1 fwd TTCTGCGTAAAACCTCCACA and rev TCTCTAGGGTGAAGCGTCCT; Ror2 fwd TGCAACTGGAGTCCTCTTTG and rev GGCTGACAAAACCCATCTTT; MyoD fwd CCTGAAGCGATACACCTCAA and rev GTTCCAGAAC CGGGTAGAAA; Xnot fwd GCCCAGACCCTACCTGTAAA and rev TGCGGATTCTCTTCATCTTG; alpha-actin fwd GCTGACAGAATGCAGAAGGA and rev CCGATCCAGACGGAGTATTT. Gene expression levels were normalized to that of ODC (ornithine decarboxylase). Unless indicated otherwise, the expression level of each gene in the control embryo was defined as 100.

Del1, Del1-delC, Del1-delN and Del1(RGE) were subcloned into the expression vector pCS2 (Arakawa et al., 2007). To generate V5-tagged constructs, the V5 tag sequence was added to the C-terminus of each coding sequence. For microinjection, capped mRNAs were synthesized using the SP6 mMessage Machine Kit (Ambion). Del1-MO (5′-GACCCCCTTTAGGATCATGCTTGA-3′), Del1 five-base-mismatched 5mis-MO (5′-cACCCgTTTAcGATCtTGCTTGt-3′), Ror2-MO (5′-GTCAGGCGAGGTAAGGGGCAACACT-3′) and standard control-MO were obtained from Gene Tools. The total amount of injected mRNAs or MOs was equalized with lacZ mRNA or control-MO, respectively.

Cryosections of animal caps were prepared as described previously (Onai et al., 2007) and immunostained with an anti-β-catenin antibody (Sigma, 1/500). For adhesion culture, dissociated animal cap cells were prepared with BSA-free medium, cultured on chamber slides (Nunc) coated with 100 μg/ml human plasma fibronectin (Invitrogen) until the siblings reached stage 13, then fixed with 1× MEMFA [0.1 M MOPS (pH 7.4), 2 mM EGTA, 1 mM MgSO_4, 3.7% formaldehyde] at 4°C for 15 minutes and subjected to immunostaining with an anti-V5 antibody (Invitrogen, 1/1000). The membrane components were visualized by co-injecting mRNA for membrane monomeric Cherry (Lyn-mCherry) protein. For permeabilization, the fixed cells were incubated with PBS containing 0.3% Triton X-100 at room temperature for 15 minutes, then washed five times with PBS before immunostaining. z-stack images were taken using a Zeiss LSM710 confocal microscope.

Animal cap cells were lysed with TNE buffer (50 mM Tris-HCl pH 7.8, 1% NP40, 20 mM EDTA, 150 mM NaCl) containing Protease Inhibitor Cocktail (Nacalai). For the detection of phospho-Erk, Phosphatase Inhibitor Cocktail II (Sigma) was added to the buffer. The primary antibodies were anti-phospho-Erk, anti-Erk (Cell Signaling, 1/500), anti-β-catenin (Sigma, 1/500) and anti-Hsc-70 (Santa Cruz, 1/500), and the secondary antibodies were anti-mouse IgG and anti-rabbit IgG (GE Healthcare, 1/50,000). Signals were detected with ECL Plus reagents (GE Healthcare) and captured with a LAS-3000UVmini (Fuji Film).

For statistical analysis, we used the Prism4 program (GraphPad). Two-group and multiple-group analyses were performed with Student's t-test and one-way ANOVA (Dunnett or Tukey post-hoc test), respectively.

In our previous report (Arakawa et al., 2007), the equator zone injection of Del1 mRNA into all four blastomeres of the 4-cell Xenopus embryo caused a moderate dorsalization in marginal zone (mesodermal) tissues. These phenotypes were explained, at least in part, by the anti-BMP effect of Del1 on mesodermal patterning.

In the early phase of the present study, we sought to analyze the effects of Del1 on ectodermal development by microinjection of Del1 mRNA into the four animal blastomeres of the 8-cell embryo, which contribute mainly to the ectodermal derivatives. Interestingly, this resulted in a phenotype distinct from that observed with the equator injection at the 4-cell stage: the formation of an enlarged head (52%, n=27; Fig. 1A-D), including large eyes (Fig. 1D, arrow) and an expanded cement gland (Fig. 1D, arrowhead), without other obvious signs of general dorsalization of the body. These phenotypes were considered to reflect the anteriorizing effect of Del1 on the pattern of the ectoderm rather than mesoderm.

Overexpression of Del1 on the left side induced an obvious expansion of the anterior neural markers Six3 (forebrain; 55%, n=40), Bf1 (telencephalon; 66%, n=56), Otx2 (forebrain/midbrain; 67%, n=46) and Pax6 (forebrain; 57%, n=42) on the ipsilateral side (arrowheads in Fig. 1E-I, anterior views). Del1 overexpression also caused a posterior shift or expansion of the posterior neural markers En2 (midbrain/hindbrain boundary; 67%, n=30), Krox20 (hindbrain; 54%, n=28) and MafB (posterior hindbrain; 57%, n=30) (arrows in Fig. 1J-L, dorsal views). Under these conditions, in contrast to the 4-cell stage injection in our previous report, we did not observe strong expansion in the expression of mesodermal markers such as Chordin (Chd; axial mesoderm; 100%, n=26) and Goosecoid (Gsc; anterior mesendoderm; no change in 71% and a minimal expansion in 29%, n=28) (Fig. 1M-P, dorsal views).

We next tested the anteriorizing effect of Del1 in a secondary-axis assay. The ventral injection of BMP inhibitors, such as Chd, induces the formation of a secondary axis, which rarely contains the anterior-most head structures, such as the eyes (Glinka et al., 1997; Pera et al., 2001) (Fig. 1Q). Our previous study showed that the ventral injection of Del1 alone does not induce a secondary axis (Arakawa et al., 2007). Interestingly, the co-injection of Chd and Del1 caused the formation of eye tissues in the secondary axis (33%, n=36; Fig. 1R, arrow). The induction of the eye-forming field by Chd and Del1 was confirmed by in situ hybridization with a probe for Rx (a forebrain and eye-field marker; 61%, n=22; Fig. 1S, anterior view).

These findings show that overexpression of Del1 in the animal region promotes anteriorization of the neuroectoderm. This anteriorizing effect was presumably overlooked in our previous analysis because the 4-cell stage injection induced dorsalization of the mesoderm, which secondarily caused some degree of anteriorization of the embryo, whereas in the present study the 8-cell stage injection could uncouple the two phenotypes.

Del1 has two EGF-like domains and two Discoidin domains (Fig. 2A). An integrin-binding RGD motif is present in the first EGF domain. We investigated the relationship between these domains and the activity of Del1 (Fig. 2). Using a V5-tagged construct, we first analyzed the effects of deleting certain domains on the accumulation of Del1 protein in the ECM. RNA-injected animal cap cells were dissociated, cultured on a plastic culture slide, and subjected to immunostaining with an anti-V5 antibody. When the cells were fixed without permeabilization of the plasma membrane (Fig. 2B-F′; Lyn-mCherry, red), a gradient of Del1-V5 and Del1-delN-V5 (lacking the EGF domains) labeling was detected in the pericellular areas of the overexpressing cells (Fig. 2C,E, green), suggesting that these molecules were associated with the pericellular matrix (see also Fig. S1A in the supplementary material). By contrast, almost no Del1-delC-V5 protein was immunodetected around the non-permeabilized cells (Fig. 2D). With permeabilization of the plasma membrane (Fig. 2G-K′), Del1-delC-V5, Del1-V5, and Del1-delN-V5 were detected inside the cells (Fig. 2H-J), indicating that all of these proteins were translated in the cells. Immunoblotting analysis (see Fig. S1B in the supplementary material) showed that Del1-delC-V5 was not detectable in the conditioned medium but was instead detected in the cell lysate, whereas Del1-V5 and Del1-delN-V5 were present in both, suggesting that Del1-delC-V5 was produced by the cells but not secreted. By contrast, the mutation in the RGD motif [Del1(RGE)-V5, in which the integrin-binding RGD motif was mutated to non-integrin-binding RGE; Fig. 2A] did not significantly affect the localization of Del1. The Del1(RGE)-V5 protein was detected in the pericellular areas of the non-permeabilized cells (Fig. 2F,F′) and was secreted into the conditioned medium (see Fig. S1B in the supplementary material).

We next examined which domains were involved in the anteriorizing activity of Del1 (Fig. 2L-S). Del1-delN-V5 expanded the area of Bf1 expression (71%, n=24; Fig. 2N). It also induced Rx expression in the secondary axis generated by Chd (83%, n=23; Fig. 2R), indicating that the C-terminal Discoidin domains are sufficient for the anteriorizing activity. Consistent with this idea, Del1(RGE)-V5 retained its anteriorizing activity (Bf1 expansion in 71%, n=38, Fig. 2O; Rx expression in the Chd-induced secondary axis, 53%, n=45, Fig. 2S). These findings demonstrate that the anteriorization activity of Del1 is mediated by the Discoidin domains and is independent of its RGD motif.

Sagittal sections of neurula stage embryos showed a strong accumulation of Del1 transcripts in the anterior neural plate (Fig. 3A, bracket), in addition to their diffuse distribution throughout the neural plate and dorsal mesoderm.

We next performed a loss-of-function study of Del1 in the embryo. The injection of Del1-MO (see Fig. S2A,B in the supplementary material) into all the animal blastomeres of the 8-cell embryo caused a head defect (microcephaly) without obvious malformation of the trunk or tail (88%, n=24; Fig. 3B,C). In situ hybridization analysis (Fig. 3D-O) showed that Del1-MO injection suppressed expression of the forebrain marker Six3 (83%, n=36; Fig. 3E). This suppression was reversed by co-injection of an MO-insensitive Del1 mRNA (no suppression in 64%, n=28; Fig. 3F; see also Fig. S2C,D in the supplementary material). Del1-MO also attenuated the expression of other forebrain markers, including Pax6 (63%, n=27) and Rx (81%, n=42), whereas the expression of these marker genes was restored by MO-insensitive Del1 mRNA injection (75%, n=28 and 74%, n=52, respectively) (Fig. 3G-L). By contrast, posterior neural markers, such as Krox20 and HoxB9 (spinal cord), were largely unaffected, and the expression of the midbrain/hindbrain marker En2 was modestly decreased or was shifted anteriorly (81%, n=42; Fig. 3N). This effect on En2 was also reversed by co-injection of MO-insensitive Del1 mRNA (no effect in 53%, n=32; Fig. 3Q). Under these conditions, Del1-MO did not substantially change the expression of the mesodermal markers Xnot (axial mesoderm) and Goosecoid (see Fig. S2E-H in the supplementary material). Taken together with the gain-of-function data (Fig. 1), these findings show that Del1 plays a crucial role in anterior neuroectodermal development in the early Xenopus embryo.

Next, we examined whether Del1 is directly involved in the AP patterning of neural tissues. Consistent with previous studies (Onai et al., 2004; Sasai et al., 1995), qPCR analysis showed that Chd-injected neuralized animal cap explants strongly expressed the anterior neural marker Six3, whereas they showed little expression of the posterior neural markers En2 and Krox20 (see Fig. S3A in the supplementary material, lanes 2, 5 and 8). The co-injection of Del1-MO induced the expression of En2 and Krox20 and moderately attenuated that of Six3 (lanes 3, 6 and 9), whereas the expression of the general neural marker NCAM was largely unaffected (lanes 11 and 12). These findings indicate that Del1 is directly required for protecting the neuroectoderm from caudalization.

We next performed overexpression analysis of Del1 in neuralized animal caps that were caudalized by Wnt signaling. The co-injection of Wnt1, even at a low dose, suppressed anterior neural markers and induced posterior genes (see Fig. S3B in the supplementary material, lanes 2, 5 and 8). This posteriorizing effect of Wnt1 was reversed by Del1 co-injection, which increased anterior marker and suppressed posterior marker expression (lanes 3, 6 and 9). The co-injection of Wnt1 at a moderate dose preferentially induced Krox20 but not En2 expression (see Fig. S3C in the supplementary material), presumably reflecting an even stronger caudalizing effect of Wnt. In this case, Del1 co-injection induced both Six3 and En2 and suppressed Krox20 expression, showing that Del1 shifts the regional identity of neural tissues in an anterior direction.

These findings suggest that Del1 plays a crucial role in the anterior specification of the neuroectoderm.

Since Del1 reversed the Wnt-induced posteriorization, we next examined the effect of Del1 on Wnt signaling in vivo (Fig. 4A-D). As previously reported (Christian and Moon, 1993), Wnt plasmid injection suppressed the expression of the anterior neuroectodermal markers in the embryo (Fig. 4B; Six3 suppression in 95%, n=40). This Wnt-induced posteriorization was reversed by the co-injection of Del1 (Fig. 4C; no Six3 suppression in 87%, n=38). Importantly, such a reversal did not occur upon co-injection of the BMP inhibitor Chd (Fig, 4D; Six3 suppression in 95%, n=43), showing that attenuation of the BMP signal by Del1 cannot by itself explain the reversal. In the Topflash-luciferase reporter assay using animal cap cells, Wnt1 injection substantially increased the luciferase activity (Fig. 4E, lane 2), and this increase was strongly suppressed by co-injecting Del1 (Fig. 4E, lane 3; in lane 4, ΔN-Tcf3 was used as a positive control) (Molenaar et al., 1996). Again, such a suppression did not occur upon co-injection with Chd (Fig. 4F, lane 4). These data show that Del1 causes attenuation of canonical Wnt signaling and that this is unlikely to occur via the suppression of BMP signaling, suggesting a direct anti-Wnt activity for Del1.

Previous studies have shown that, in addition to Wnt signals, FGF and Nodal play an important role in early AP patterning in Xenopus (Kengaku and Okamoto, 1995; Piccolo et al., 1999). FGF signals also posteriorized the AP identity of Chd-induced neural tissues (see Fig. S4A, lane 4, in the supplementary material), and Del1 co-injection reversed the posteriorized phenotypes caused by bFGF in the neuralized animal cap (see Fig. S4A, lane 5, in the supplementary material). However, Del1 injection did not inhibit the phosphorylation of Erk (Fig. 4G, lane 3; in lane 4, dnFGFR was used as a positive control) (Amaya et al., 1991) or substantially suppress the induction of Xbra expression in the animal cap treated with bFGF (Fig. 4H, lane 3). This discrepancy may be explained, at least in part, by a previous report that the posteriorizing activity of FGF can be reversed by the Wnt inhibitor Dkk1 (Kazanskaya et al., 2000), indicating the requirement for Wnt signaling downstream of FGF.

Del1 injection did not substantially affect Nodal/Activin signaling either. Del1 injection neither affected the activin response element (ARE)-luciferase activity (Sasai et al., 2008) nor strongly interfered with the Activin-induced induction of Mix.2 expression in the animal cap (see Fig. S4B,C in the supplementary material).

These observations indicate that the anteriorizing activity of Del1 is relevant to the attenuation of Wnt signaling. Consistent with this idea, injection of Del1-MO elevated the Topflash-luciferase activity in the anterior region of the embryo as compared with the level in the control injected with 5mis-MO (Fig. 4I). In addition, the decrease in Six3 expression upon Del1-MO injection was reversed by co-injecting the Wnt inhibitor Dkk1 in a dose-dependent manner (Fig. 4J-M).

In Topflash assays using animal cap cells, Del1 suppressed the luciferase activity induced by the injection of Dishevelled (Dsh) (Fig. 5A), dominant-negative GSK3β (Fig. 5B) (Pierce and Kimelman, 1995) and β-catenin (Fig. 5C), suggesting that Del1 functions antagonistically to the canonical Wnt pathway downstream of these intracellular mediators.

We next analyzed the nuclear localization of β-catenin (Fig. 5D-H), a key regulatory step for the transduction of Wnt signaling (Moon et al., 2004). Immunostaining showed that the Wnt1-induced nuclear level of β-catenin (Fig. 5E) was not substantially reduced by co-injection of Del1 (Fig. 5F), in contrast to the clear reduction caused by co-injection of Dkk1, an upstream antagonist of Wnt signaling (Fig. 5G). These observations support the idea that Del1 acts downstream of the nuclear localization of β-catenin in the canonical Wnt pathway.

In dissociated animal cap cells, the addition of recombinant Wnt3a protein increased Topflash-luciferase activity (Fig. 5I, lane 2) and the level of β-catenin protein (Fig. 5J, lane 2), which is indicative of a reduction in GSK3β-mediated degradation. GSK3β injection inhibited the increases in β-catenin and luciferase levels (Fig. 5I,J, lane 3). By contrast, although Del1 injection also suppressed the Wnt3a-induced luciferase activity, it did not reduce the level of β-catenin (Fig. 5I,J, lane 4). These findings support the idea that Del1 function is not mediated by the conventional GSK3β-regulated β-catenin degradation.

The evidence above suggests that Del1 interferes with the canonical Wnt pathway downstream of the nuclear accumulation of β-catenin. Previous studies have shown that, like Del1, the non-canonical Wnt signal can inhibit the pathway downstream of β-catenin via the Ror2 (Mikels and Nusse, 2006) or NLK (Ishitani et al., 2003) pathway. Therefore, we examined the involvement of Ror2 and NLK in the anteriorizing activity of Del1 (Fig. 6A-F).

The expansion of Bf1 expression by overexpression of Del1 mRNA (Fig. 6B; 71%, n=21) was strongly suppressed by co-injection of Ror2-MO (Fig. 6D; no expansion in 100%, n=23), although Ror2-MO injection alone had only a marginal phenotypic effect at this dose (Fig. 6C). By contrast, the injection of NLK-MO (Satoh et al., 2007) did not effectively suppress the Del1-induced Bf1 upregulation (Fig. 6F; Bf1 expansion in 44%, n=25), even at the dose at which the NLK-MO injection alone showed an inhibitory effect on Bf1 expression (Fig. 6E). These results raised the possibility that the anteriorizing activity of Del1 preferentially depends on the Ror2 pathway.

Ror2 expression was widely observed in the embryonic ectoderm during the early and mid-gastrula stages (later, its expression becomes stronger in the caudal neural tissue) (Hikasa et al., 2002) and also in neuralized animal cap explants (see Fig. S5A-E in the supplementary material). We performed an animal cap assay to examine whether the Ror2 pathway was directly involved in the Del1-induced anteriorization of the neuroectodermal tissue. In Chd/Wnt1-injected animal caps, co-injection of Del1 induced the expression of Six3 and En2 (Fig. 6G, lanes 2 and 6) at the cost of expression of Krox20 (lane 10). Interestingly, this Del1-induced anteriorization in the explants was strongly suppressed by Ror2-MO co-injection (Fig. 6G, lanes 4 and 8), and the reduction of Krox20 expression by Del1 became marginal in the Ror2-depleted animal caps (compare lanes 11 and 12 with 9 and 10). In the Topflash-luciferase assay using animal caps, suppression of the Wnt-induced luciferase activity by Del1 (Fig. 6H, black bars in lanes 2 and 3) was diminished when Ror2-MO was co-injected (white bars in lanes 2 and 3), indicating that Del1 inhibited the canonical Wnt signaling in a Ror2-dependent manner.

Previous studies using cultured cells have shown that non-canonical Wnt ligands, such as Wnt5a and Wnt11, bind directly to the Ror2 receptor and inhibit the canonical Wnt pathway downstream of β-catenin (Hikasa et al., 2002; Maye et al., 2004; Mikels and Nusse, 2006). Consistent with this idea, we observed in the animal cap assay that overexpression of the non-canonical ligands Wnt11 or Wnt5a almost completely suppressed the luciferase activity induced by Wnt3a or Δβ-catenin [which encodes a degradation-resistant form of β-catenin (Yost et al., 1996)] (see Fig. S6A and Fig. S7A in the supplementary material). By contrast, this suppression was only partial (~50%) when Ror2 was knocked down by MO (see Fig. S6B in the supplementary material, compare lanes 5 and 6 with 2 and 3), suggesting that Ror2 also mediates a substantial portion of the Wnt11- or Wnt5a-induced suppression of the canonical signal in the embryonic Xenopus cells. The injection of Wnt11 or Wnt5a induced expression of anterior neural markers, such as Six3, at the cost of posterior gene expression in Chd/Wnt1-injected animal caps (see Fig. S6C and Fig. S7B in the supplementary material). This induction of Six3 by Wnt11 or Wnt5a was reversed by Ror2-MO co-injection (Fig. 6I; see Fig. S7C in the supplementary material, lanes 2 and 4). These results suggested that activation of the Ror2-mediated non-canonical pathway substantially inhibited canonical Wnt signaling and promoted anteriorization in the neuralized animal cap.

Finally, given that the anteriorizing activities of both Del1 and Wnt11 or Wnt5a depend on the Ror2 pathway, we examined whether the Del1 signal and the Wnt11-induced non-canonical signal exhibited any functional interaction. We injected small doses of Del1 and Wnt11 (Fig. 6J) or Wnt5a (see Fig. S7D in the supplementary material) mRNAs into the Chd/Wnt1-injected animal cap. When injected individually, Del1 or Wnt11 showed only a moderate effect on Six3 expression (Fig. 6J, lanes 2 and 3). By contrast, when injected together, they synergistically elevated Six3 expression (lane 4). Importantly, this strong upregulation of the anterior marker was suppressed by co-injection of Ror2-MO (Fig. 6K, compare lanes 2 and 4), showing that this phenomenon is Ror2 dependent. Similarly, in the Topflash assay (see Fig. S6D in the supplementary material), a sub-effective dose of Wnt11 (lane 3) enhanced the Del1-induced attenuation of luciferase activity (lanes 4 and 5). Then, we tested whether the functional interaction of Del1 and Ror2 was also observed in vivo (Fig. 6L-O; see also Fig. S8 in the supplementary material). The single injection of a sub-optimal dose of Ror2-MO or Del1-MO only marginally suppressed Bf1 expression (Fig. 6M,N). By contrast, the injection of both Ror2-MO and Del1-MO strongly inhibited Bf1 expression (Fig. 6O), suggesting that Ror2 and Del1 work together in neural anteriorization in the embryo.

Collectively, these findings indicate that Del1 promotes anterior neuroectoderm development via inhibition of the canonical Wnt signaling pathway in a Ror2-dependent manner.

Several soluble inhibitors of Wnt signals have been identified, including Cerberus, Frzb, Dkk1, WIF1 and Crescent. These factors inhibit the cellular reception of Wnt signals by binding to the Wnt ligands or their receptor complexes (reviewed by Kawano and Kypta, 2003). Compared with these factors, Del1 is unique in two aspects. First, Del1 interferes with the Wnt intracellular pathway, presumably via a non-integrin receptor (as discussed below). Second, Del1 is a matrix protein that accumulates pericellularly. In this sense, Del1 is not a typical diffusible inducer that emanates from the head organizer (e.g. Cerberus from the prechordal plate), but rather acts as a local modulator of the signaling microenvironment within the anterior neuroectoderm. A reasonable interpretation is that the anterior neuroectoderm protects itself from the posteriorizing Wnt signal by accumulating Del1 pericellularly and shifting its microenvironment in an ‘anti-Wnt’ direction. Previous studies have also identified several Wnt-modulating matrix proteins, such as Syndecan-4 (Muñoz et al., 2006), Glypican-4 (Ohkawara et al., 2003) and R-spondin (Wei et al., 2007; Kim et al., 2008). Thus, it appears that the fine-tuning of local Wnt responsiveness is orchestrated by an intricate network of positive and negative Wnt-modulating matrices.

As our previous report indicated that the anti-BMP activity of Del1 is also independent of the RGD motif, it was important to rule out the possibility that the anti-Wnt activity of Del1 is secondary to its anti-BMP function, as this direction of interaction was previously suggested in another context (Nishita et al., 2000). The following observations in the present study indicate that such a secondary effect is unlikely. First, Del1 has a qualitatively distinct activity, i.e. eye induction, in the secondary axis that is induced by BMP inhibition. Second, similarly, Del1 promotes an obvious anteriorization in the neural tissue induced from animal caps by Chd, suggesting that Del1 plays more roles than the BMP antagonist Chd does. Third, Wnt-induced posteriorization of the embryo is not reversed by co-injecting Chd. Furthermore, Del1 strongly inhibits Wnt signaling in the animal cap, whereas Chd affects the Topflash activity only marginally. In addition, a BRE-luciferase assay (see Fig. S9 in the supplementary material) showed that inhibition of the BMP signal by Del1, unlike the inhibition of the canonical Wnt signal (as discussed below), is largely independent of the Ror2 pathway.

From a functional viewpoint, whatever the molecular mechanism might be, the double inhibition of Wnt and BMP signals by the same protein (i.e. the multi-functional Del1) is an efficient way to promote head development (Niehrs, 2004), as has been shown for the secreted head organizer factor Cerberus (Piccolo et al., 1999). It is likely that Del1, a matrix protein, acts more locally to consolidate the forebrain fate.

The present study showed that Del1 inhibits canonical Wnt signaling downstream of the nuclear accumulation of β-catenin. In addition, we demonstrated that this anti-canonical Wnt activity of Del1 is dependent on the Ror2 pathway, which is activated by the non-canonical Wnt ligands. This synergistic interaction between Del1 and the Ror2 pathway is essential for anterior neural development in vivo (Fig. 6L-O). Fig. 6P illustrates our working hypothesis deduced from the results of the present study. Del1 attenuates canonical Wnt signaling by evoking a Ror2-sensitive intracellular cascade that interferes with β-catenin/Tcf-dependent activation of target genes (Mikels and Nusse, 2006). It is an important future challenge to elucidate the detailed molecular cascade involved in Del1 signaling. It has been reported that Wnt5a/Ror2 signaling induces XPAPC expression by activating the Cdc42-JNK-ATF2 pathway (Schambony and Wedlich, 2007). However, in our preliminary study, unlike Wnt5a-induced XPAPC expression, which is suppressed by dominant-negative Cdc42 (dnCdc42) (Choi and Han, 2002), the anti-canonical Wnt activity of Del1 or Wnt5a was not affected by dnCdc42 (see Fig. S10 in the supplementary material), suggesting that the anti-Wnt activity of Del1/Wnt5a is mediated by other signaling pathways.

In addition, it remains to be determined whether Del1 signaling merges with the Ror2 signaling pathway at the Ror2 receptor level or acts in the intracellular pathway downstream of Ror2 (Fig. 6P, dotted arrows). We have so far failed to detect any significant binding of Del1 to Ror2-Fc protein (the extracellular domain of Ror2 fused with the human IgG Fc domain in the C-terminal region) or any facilitated binding of Wnt11 to Ror2-Fc protein in the presence of Del1. The finding that the anti-BMP activity of Del1 is independent of Ror2 suggests the presence of a non-Ror2/non-integrin receptor (R in Fig. 6P) system for Del1, at least for its anti-BMP activity. One technical hindrance for Ror2 signaling analysis at present is the lack of a method to activate Ror2 in a ligand-independent manner. For instance, Ror2 overexpression does not cause the anteriorization of the Chd/Wnt1-injected animal cap (see Fig. S5F in the supplementary material). It is inferred that endogenous Ror2 expression might already be close to the threshold level, as indicated by the fact that the optimal amount of Ror2 required to reverse the Ror2-MO phenotype is fairly low (see Fig. S5G in the supplementary material).

Another intriguing question is whether Del1 also affects the Ror2-mediated planar cell polarity (PCP) pathway (Hikasa et al., 2002), which is a typical non-canonical Wnt cascade, in the opposite direction. Activin-treated animal caps differentiate mainly into dorsal mesodermal tissues (Osada and Wright, 1999) and exhibit a massive convergent-extension movement that depends on the PCP pathway. In our preliminary study using Activin-treated animal caps, Del1 injection inhibited the extension of the mesodermalized animal cap. However, it was not easy to judge whether this is a direct consequence of the effects of Del1 on the PCP pathway or a secondary influence of its anteriorization effect on the dorsal mesoderm (see Fig. S11 in the supplementary material). Del1 injection increased the expression of the anterior dorsal marker Goosecoid, but suppressed the trunk mesodermal markers MyoD, alpha-actin and Xbra. By contrast, Del1 did not substantially affect the expression of the dorsal mesodermal marker Chd, which is expressed in both the anterior and posterior axial regions (Sasai et al., 1994), suggesting that Del1 promotes anteriorization not only of the ectodermal tissue but also of the dorsal mesodermal tissue. Because the convergent-extension movement is observed predominantly in the trunk mesoderm but not strongly in the head mesoderm (Wallingford et al., 2002), the relationship between the suppression of the convergent-extension movement and the promotion of anteriorization is not readily separable in this experimental system and will require careful future investigation.

An additional remaining question is why the overexpression of the anti-Wnt factor Del1 (even when its mRNA is injected radially at the 4-cell stage) does not inhibit the formation of the dorsal axis (Arakawa et al., 2007), which is known to depend on intracellular β-catenin signaling (Heasman, 2006; Weaver and Kimelman, 2004). Consistent with this finding, Del1 did not suppress the induction of a secondary axis by Wnt8 mRNA injection, which mimics the maternal effect (see Fig. S12 in the supplementary material). The axis-inducing β-catenin signal is known to reflect maternal Wnt activity. In contrast to early/maternal Wnt signaling, late/zygotic Wnt signaling that occurs after the MBT (particularly after the late blastula stage) plays a major role in the posterior and ventral specification of embryonic tissues (Niehrs, 2004). In the neural tissue, late/zygotic Wnt signaling has a strong patterning activity that promotes posterior differentiation at the expense of the development of anterior tissues (Kiecker and Niehrs, 2001) (Fig. 3A,B). The gain- and loss-of-function Del1 phenotypes shown in this report are consistent with the idea that Del1 antagonizes Wnt signaling preferentially in its late/zygotic posteriorization function. The dependence of Del1 activity on Ror2 explains, at least in part, its late Wnt-specific phenotype, as Ror2 expression is zygote specific (Hikasa et al., 2002) and no obvious maternal effects, such as axial phenotypes, have been reported for Ror2-MO injection (Schambony and Wedlich, 2007) (our unpublished observations).

In conclusion, the matrix-associated protein Del1 acts as an extracellular modulator that controls the zygotic canonical Wnt pathway in a close relationship with Ror2-dependent non-canonical Wnt regulation.

We thank members of the Y.S. laboratory for comments on this work, Hiroko Takai for constant encouragement, and Ms Masako Suzuki for excellent assistance in maintenance of the frog facility. This work was supported in part by grants-in-aid from MEXT.