Maternal xNorrin, a Canonical Wnt Signaling Agonist and TGF-β Antagonist, Controls Early Neuroectoderm Specification in Xenopus

Xenopus maternal Norrin, which activates Wnt signaling but inhibits TGF-β family molecules, is essential for neuroectoderm formation. Loss of TGF-β inhibition in Norrin may contribute to the development of Norrie disease.


Introduction
Dorsal-ventral axis specification is one of the earliest patterning events in the embryo. In vertebrates, early dorsal ectoderm gives rise to the neural plate, which in turn develops into the central nervous system (CNS). Previous studies have found that dorsal axis formation in amphibians is initiated during cortical rotation after fertilization. Current evidence strongly suggests that the canonical Wnt signaling pathway, operating at blastula stages, plays a critical role in dorsal specification [1]. For example, Wnt signaling was discovered to induce secondary axes when ectopically activated in the ventral cells of early embryos. Loss-of-function studies indicate that the Wnt/b-catenin signaling pathway is also essential for dorsal specification [2][3][4]. More recently, Heasman and colleagues provided strong evidence that vegetally localized maternal Wnt11 cooperates with Wnt5A to activate the canonical Wnt pathway and is required for dorsal axis formation [5][6][7]. However, despite extensive studies on dorsal specification, some observations cannot be fully explained. For example, although the cortex is rotated only 30u toward the dorsal side, activated nuclear b-catenin is observed in all dorsal cells, including dorsal cells near the animal pole [8]. Previous studies suggested that Wnt pathway components may be transferred beyond 30u to the dorsal animal region [8,9]. However, it remains unknown whether such movements can fully account for Wnt activation in dorsal animal cells, and it is also unclear how these movements precisely regulate the earliest steps of neuroectoderm formation in the blastula.
In addition to canonical Wnt signaling, the BMP pathway has also been implicated in neuroectoderm specification and patterning. During early gastrulation, Noggin, Chordin, and Follistatin expressed in the Spemann organizer bind to BMPs in the extracellular space and antagonize their epidermal-promoting effects [10][11][12]. These results support a ''default model'' for neural induction in which ectoderm cells are predisposed to become neurons if they receive no BMP signals [13,14]. Genetic screens in Drosophila and zebrafish have yielded mutants that affect dorsalventral patterning. Interestingly, most of these mutants show defects in the BMP signaling pathway, indicating that BMP signaling has a conserved role in dorsal-ventral patterning [1].
On the other hand, dorsal animal cells in the Xenopus blastula can develop into dorsal and neural tissues cell-autonomously when cultured in a saline solution [15,16]. De Robertis and colleagues found that a subset of the dorsal ectoderm cells in the late blastula expressed Chordin, Noggin, Siamois, and Xnr3 prior to Spemann organizer functioning and referred to these cells as the blastula Chordinand Noggin-expressing center (BCNE center) [16]. Early Chordin and Noggin transcription is activated by maternal b-catenin, but the precise mechanism underlying this activation remains to be uncovered [16].
We report here that maternal Xenopus Norrin (xNorrin) is required for b-catenin activation in dorsal animal cells in the Xenopus blastula and in early neuroectoderm development. Norrin is a non-Wnt ligand that was previously shown to activate bcatenin through LRP5 and Frizzled4 or TSPAN12 during retina vascular development [17][18][19]. In humans, mutations in Norrin cause Norrie disease [20]. We further show that xNorrin can directly antagonize TGF-b/BMP signaling. Our results not only identify an endogenous maternal factor required for early neuroectoderm specification, but may also add TGF-b inhibition to the increasingly complex regulatory activities of Norrin in retinal vascular development [17,19].

xNorrin Promotes Dorsal and Anterior Neural Formation
We sought to identify additional secreted molecules that are involved in neuroectoderm formation. Neuroectoderm is derived from dorsal animal regions in early Xenopus embryos. Therefore, we used early neural markers to search for molecules that may be responsible for early neural specification. In Xenopus, ultraviolet (UV) irradiation of the vegetal pole in embryos causes severe dorsal axis development defects [21] ( Figure 1A-1F) in otherwise normal embryos ( Figure 1A). We selected a set of candidate genes

Author Summary
A key step during early embryogenesis is the generation of neural precursors, which later form the central nervous system. In vertebrates, this process requires proper dorsalventral axis specification, and we know that the canonical Wnt and BMP signaling pathways help pattern the dorsal ectoderm. In this study, we examine other factors that are involved in neuroectoderm development in the frog species Xenopus laevis. We find that maternal Xenopus Norrin (xNorrin) is required for canonical Wnt signaling in the dorsal ectoderm, functions upstream of neural inducers, and is required for neural formation. We also find that xNorrin not only activates Wnt signaling, but also inhibits BMP/Nodal-related signaling. In humans, mutations in Norrin cause Norrie disease. Using Norrin mutants identified in patients with Norrie disease, we find that some Norrin mutants fail to inhibit BMP/Nodal-related signaling (specifically, TGF-b) but retain the ability to activate the Wnt pathway, suggesting that loss of TGF-b inhibition may contribute to Norrie disease development. that were previously shown to activate Wnt/b-catenin pathways and tested their ability to reorganize the dorsal axis or anterior neural tissues by injecting them individually into UV-irradiated embryos. Among the maternally expressed Wnt genes (Wnt5a, Wnt8b, and Wnt11) tested, Wnt11 and Wnt8b were able to induce some dorsal axis structures ( Figure 1C and data not shown) [22,23]. However, none of these molecules triggered the formation of anterior neural tissues (data not shown).
We also cloned X. laevis xNorrin (GenBank accession number: EU528658) from unfertilized eggs. This gene encodes a homolog of human Norrin that can activate b-catenin [18]. The injection of xNorrin mRNA into UV-ventralized embryos produced a welldefined head-like structure (Figure 1), including cement gland, eye, and brain-like tissues (85%, n = 97). In contrast, only 5% of UVventralized embryos (n = 76) developed any dorsal axial structures (such as notochord and neural tube), and 49% of Wnt11-injected, UV-irradiated embryos (n = 81) developed dorsal ridges without notochords and neural tubes ( Figure 1C and 1F). Gene expression analysis showed that xNorrin induced not only pan-neural markers, such as Sox3, Sox2, and NCAM, but also anterior neural markers, such as otx2, Xpax6, and En2, in stage 20 embryos ( Figure 1J and 1K). In contrast, Wnt11 induced the expression of the rhombomere marker Krox20 ( Figure 1K) and only weakly induced the expression of the pan-neural marker Sox3 ( Figure 1I). These results indicate that xNorrin can promote anterior neural tissue formation in an otherwise non-neural background. The neural formation triggered by xNorrin expression in UV-ventralized embryos may perhaps be attributable to early neuroectoderm induction by the injected xNorrin.
We reasoned that for maternal xNorrin to act in specifying the neuroectoderm in a cell-autonomous fashion, it should meet two criteria. First, it should be expressed in the dorsal ectoderm of the blastula [16,24]. Second, it should be able to activate canonical Wnt signaling [16,25]. Indeed, we confirmed that xNorrin mRNA is expressed in the animal pole of stage 6 oocytes and early cleavage embryos (Figures 2A, 2C, and S2A). In addition, much more xNorrin mRNA was detected in dorsal blastomeres than in ventral blastomeres in 16-cell-stage embryos ( Figure 2B).
Norrin proteins are highly conserved among vertebrates ( Figure  S1A and S1B). xNorrin, like its mouse ortholog, can activate Wntresponsive reporters (data not shown) and induce LRP6 phosphorylation in HEK293T cells ( Figure S2C). Next, we examined whether xNorrin could activate early Wnt target gene expression in vivo. The injection of xNorrin into UV-irradiated embryos robustly induced the expression of the known Wnt targets Chordin, Noggin, Xnr3, and Siamois ( Figure 2D). Further, animal caps injected with xNorrin plus Xenopus Frizzled4 plus human Lrp5 mRNA (NFL) also expressed Xnr3 and Siamois, but not Xbra ( Figure  S2B). We noted that xNorrin injection alone did not induce Xnr3 or Siamois expression in animal caps ( Figure S2B), suggesting that some components of the xNorrin pathway may not be expressed in the caps (see Discussion). However, the injection of xNorrin into dorsal animal cells enhanced Chordin expression during the late blastula and early gastrula stages ( Figure 2E). These results suggest that maternal xNorrin may promote neuroectoderm specification by activating canonical Wnt signaling.

xNorrin Is Required for Neuroectoderm Specification
To address whether maternal xNorrin is required for neuroectoderm specification and hence anterior CNS formation at a later stage, we used an xNorrin antisense morpholino (MO) oligonucleotide (xNor-MO) to inhibit xNorrin translation ( Figure 3A). The inhibition of xNorrin mRNA translation by xNor-MO was both specific and dose-dependent ( Figure 3B). We injected xNor-MO into the animal region of the two dorsal blastomeres in the four-to eight-cell embryo stage to suppress endogenous xNorrin translation. The majority of xNor-MO-injected embryos (61%, n = 64) displayed anterior head truncations, and another 15% of the embryos lacked morphological eye structures and other anterior neural structures at tadpole stages ( Figure 3D). The injection of a mismatched MO (misMO), xNor-misMO, that failed to block xNorrin-Myc translation (data not shown) produced no discernible phenotype compared to uninjected controls ( Figure 3C and 3E). The specificity of xNor-MO was further tested by the co-injection of a wild-type xNorrin mRNA lacking the xNor-MO target sequence. The injection of 25 pg of xNorrin mRNA significantly rescued the anterior neural development defects in xNor-MOinjected embryos ( Figure 3F) (n = 81, 77% rescued). Furthermore, the injection of xNor-MO into one dorsal animal cell in eight-cellstage embryos, while leaving the other side intact, resulted in severe defects in eye development at later stages (compare Figure 3G and 3H). Because xNorrin is also expressed zygotically at later stages, we designed a splicing MO (spMO) to specifically block its splicing ( Figure S3A and S3B). While xNor-MO inhibited anterior development, xNor-spMO had almost no effect on axis development ( Figure S3C-S3G). We further confirmed that xNor- MO preferentially inhibited XBF-1 (an anterior neural marker [26]) expression in the injected side, while xNor-spMO had a much weaker effect ( Figure S3H). Neither MO had a significant role in regulating the expression of HoxB9 (a posterior marker [27]) ( Figure S3H). These results suggest that maternal xNorrin signaling is required for anterior CNS formation.
The loss of anterior head development may be an indirect effect due to a lack of early neuroectoderm specification. Thus, we tried to address whether b-catenin activation in the ectoderm, which is indispensable for full dorsal axis formation [3], depends on xNorrin activity. First, we used a SuperTopFlash Wnt reporter, which can be activated by injection into the dorsal animal blastomeres of eight-cell-stage embryos [5] ( Figure 3I). The coinjection of xNor-MO with the reporter plasmid largely blocked reporter activity compared to co-injection with xNor-misMO ( Figure 3I). In a separate assay, we examined whether maternal xNorrin was required for the expression of Chordin, Noggin, Xnr3, and Siamois in dorsal animal cells, which is one of the earliest indications of b-catenin activation [16]. We found that xNor-MO reduced the expression of these genes in late blastula embryos ( Figures 3J, S4A, and S4B) but did not interfere with the expression of gsc or Xnr1 ( Figure S4A and S4B) at the early reporter plasmids were co-injected into the dorsal animal cells of eight-cell embryos. F/R luciferase: ratio of firefly luciferase reading to renilla luciferase reading. (J) Whole-mount in situ hybridization shows that Chordin expression is reduced at stage 9 (53% of injected embryos, n = 80) and stage 10 (61% of injected embryos, n = 79) in xNor-MO-injected embryos, compared to xNor-misMO-injected embryos or uninjected embryos. (K-N) Whole-mount in situ hybridization for Chordin in bisected xNor-MO-injected embryos (stage 9.5) showing that xNor-MO inhibits Chordin expression in neuroectoderm precursors (arrow) (reduction in 66% of injected embryos, n = 104) (L) compared to wild-type embryos (K) and embryos with xNor-misMO injected into dorsal animal cells at the eight-to 16-cell stage (reduction in 13% of injected embryos, n = 78) (M). Note that xNorrin mRNA (100 pg) rescues Chordin expression in the dorsal ectoderm (80% of co-injected embryos showed expression comparable to wild-type embryos, n = 50) (N). Embryos are oriented such that their dorsal side is on the right. Dotted lines indicate the boundaries between the deep mesoderm and the superficial ectoderm. doi:10.1371/journal.pbio.1001286.g003 xNorrin Controls Early Neuroectoderm Specification PLoS Biology | www.plosbiology.org gastrula stage. The reductions in the expression of these genes in xNor-MO embryos can be rescued by the co-expression of xNorrin ( Figure S4B). In late blastula xNor-MO embryos (stage 9.5), the reduction of Chordin expression was mostly restricted to the ectoderm, while deep dorsal mesoderm cells retained weak expression ( Figure 3L). The ectoderm expression of Chordin in the later blastula was fully restored by the co-injection of wild-type xNorrin mRNA lacking the MO target sequence ( Figure 3N). In the blastula ectoderm, Chordin expression is controlled by maternal bcatenin [16,28]. Thus, the control of early Chordin expression by xNorrin should partially reflect how xNorrin functions in neuroectoderm precursors. Together, these results indicate that, besides vegetally localized Wnt11 activity, maternal xNorrin is required to activate the canonical Wnt pathway in the dorsal ectoderm and is essential for the proper expression of early zygotic neural inducers before gastrulation.

xNorrin-Activated Wnt Signaling Fails to Dorsalize Ventral Mesoderm
Mouse Norrin is a secreted protein that is tightly associated with the extracellular matrix [29]. However, we found that xNorrin was secreted into culture medium when expressed in HEK293 cells and Xenopus embryo explants (data not shown). The secretion of xNorrin in the culture cells and its potent activity in early embryos prompted us to speculate that other mechanisms may be required to restrict xNorrin activity in early embryos.
We first tested whether xNorrin was active when expressed ectopically in embryos. Previous studies indicated that the ectopic activation of the canonical Wnt pathway in the ventral side of early embryos is sufficient for secondary dorsal axis formation [30][31][32]. In addition, the co-expression of NFL was shown to activate canonical Wnt signaling in tissue culture cells and in animal cap explants ( Figure S2B and S2C). Thus, we examined whether NFL could mimic canonical Wnt proteins and induce secondary axes in early embryos. Surprisingly, when injected into the ventral vegetal cells of early embryos, NFL failed to generate any complete secondary axes ( Figure 4C), while Wnt8 was able to generate secondary axes, as shown previously [31,32] ( Figure 4B). However, NFL was able to weakly induce partial secondary axes in which the neural marker Sox3 was detected ( Figure 4F). NFLinjected embryos had neural tubes but not notochords in their secondary axes ( Figure 4I, 4L, and 4L'), while Wnt8-injected embryos had complete secondary axes containing both neural tubes and notochords ( Figure 4E, H, K, and K') [31,32].
The failure of NFL-injected embryos to form complete secondary axes was not due to a lack of activation of Wnt signaling by NFL, because Chordin, Siamois, and Xnr3 expression could be detected in the ventral side of the early gastrula (stage 10) ( Figure S4C and S4D). However, Chordin expression was mostly induced in the superficial layer and not in the deep ventral mesoderm ( Figure 4O). The much lower expression of Chordin in the deep layer was considered unlikely to be a staining artifact because strong Chordin signal was readily detected in the dorsal mesoderm ( Figure 4O). Embryos injected ventrally with Wnt8 strongly induced Chordin expression in both germ layers, as expected ( Figure 4N and 4Q). Similarly, the injection of NFL into ventral animal cells induced Chordin transcription only in the ventral ectoderm and not in the mesoderm ( Figure 4R).
These results suggest that an intrinsic mechanism may exist to restrict endogenous xNorrin activity to the prospective neuroectoderm. Alternatively, injected NFL may alter the cell fate of endomesoderm, making it incompetent to form dorsal endomesoderm, even in the presence of canonical Wnt signaling.
xNorrin Inhibits Activin/Nodal-Related Induced Mesoderm Formation Because NFL injection failed to activate Wnt target genes in the endomesoderm ( Figure 4O and 4R), we initially proposed that an xNorrin-specific inhibitor might exist in the endomesoderm. However, after extensive investigation, we were not able to identify any molecule that could fulfill the proposed criteria for the inhibitor, i.e., that it should be expressed specifically in the endomesoderm and exert its antagonizing activity on xNorrin but not Wnt8. We thus turned to an alternative possibility, that NFL may influence the fate of endomesoderm precursor cells, making the germ layer incapable of conversion into dorsal endomesoderm. Previously, TGF-b family members, such as Xnr1, -2, -4, -5, and -6 and derriere were shown to be essential for mesoderm induction in Xenopus embryos [33]. Zygotic transcription of Xnr genes is activated by maternal transcription factor VegT and b-catenin. The Nodal-related molecules form a dorsal-ventral gradient that induces dose-dependent endomesoderm formation. Higher concentration of Nodal-related molecules results in dorsal specification [34,35]. In a mesoderm induction assay, we used Activin, in lieu of Nodal-related molecules, to induce strong axial mesoderm and convergent extension in animal cap cells ( Figure 5A-5C) [33,36]. When co-expressed in animal cap cells, xNorrin completely blocked the Activin-induced elongation of animal cap explants ( Figure 5D; compare to Figure 5B and 5C). The inhibition of mesoderm formation was confirmed by the lack of expression of the mesoderm markers Xbra, Xwnt8, MyoD, and mactin in the co-expressing explants ( Figure 5E). In whole embryos, xNorrin injection into the vegetal pole also blocked Xbra expression ( Figure S5A-S5D). These results suggest that xNorrin may negatively regulate mesoderm induction in vivo. Next, we tested whether Xnr1 and xNorrin could be directly associated extracellularly. We combined and incubated conditioned medium from Xnr1-transfected HEK293 cells and from xNorrin-transfected cells and used the medium for immunoprecipitation. Indeed, we detected an association between Xnr1 and xNorrin ( Figure S6B), suggesting that maternal xNorrin may restrict Nodal-related activity from extending into the animal pole.

Reciprocal Inhibition between xNorrin and BMP4
Because xNorrin can inhibit Activin/Nodal-related activity, we hypothesized that it may also antagonize other members of the TGF-b superfamily. Indeed, we found that xNorrin also strongly inhibited the activity of a BMP4 reporter (BRE-Luc) ( Figure 5F). As expected, xNorrin also inhibited Smad1 phosphorylation induced by BMP4 ( Figure 5G). One possible mechanism for inhibition between proteins is through direct binding. We examined this possibility between BMP4 and xNorrin. To this end, we injected differently tagged BMP4 and xNorrin mRNAs into adjacent blastomeres in advanced four-cell-stage embryos to allow secretion of the respective proteins into the extracellular space. At late gastrula, protein extract was immunoprecipitated with one tag antibody and blotted with the other tag antibody. Results showed that BMP4 was indeed associated with xNorrin extracellularly. Thus, the inhibition by xNorrin is likely through direct binding to BMP4 ( Figure 5I).
The direct interactions between xNorrin and BMP4 led us to investigate whether xNorrin activity was regulated by BMP4. We showed that xNorrin induced neural marker expression in animal caps ( Figure 5H). In an animal cap assay, BMP4 significantly inhibited the otx2, Xpax6, and NCAM expression induced by xNorrin ( Figure 5H). Thus, reciprocal inhibition between xNorrin and BMP4 may also be implicated in dorsal-ventral ectoderm development.
Previous studies indicated that the dorsally expressed BMP4 inhibitors Chordin, Noggin, and Follistatin could induce neural formation through direct binding [10][11][12]. Because xNorrin can also inhibit BMP4, we investigated xNorrin neural induction activity. Indeed, we found that xNorrin alone can induce the expression of neural-specific genes in animal cap cells in a dosedependent manner (Figures 6A and S5A). The neural promoting activity of xNorrin in ectodermal cells was confirmed by the Sox3 (a neural marker) expression in xNorrin-injected animal caps ( Figure 6B). Further, xNorrin, like the truncated BMP receptor DBMPR, can induce ectopic Sox3 and XAG1 (an anterior marker) expression when injected into one ventral blastomere of 32-cell embryos ( Figure 6C). More importantly, neural induction was observed when a b-catenin-specific MO was co-injected, indicating that canonical Wnt pathway activation is not required for neural formation in this setting ( Figure 6A, compare the two bcatenin-MO-injected lanes). Furthermore, we did not observe activation of Xnr3 or Siamois in xNorrin-injected animal caps, confirming the lack of canonical Wnt activation ( Figure S2B). We conclude that xNorrin and BMP4 are reciprocally inhibited and that xNorrin may promote neural development independent of Wnt signaling activation.

Loss of TGF-b Inhibition in a Subset of Norrin Mutants
Norrin mutations are responsible for both X-linked familial exudative vitreoretinopathy (FEVR) and Norrie disease (Online Mendelian Inheritance of Man MIM#310620) in humans [37,38]. The finding that Norrin can inhibit BMP/TGF-b activity prompted us to test whether this regulation is involved in the disease development. We noticed that some previously identified Norrin mutants isolated from human patients did not significantly affect Wnt pathway activation [18,39]. We hypothesized that these human Norrin mutants might instead be compromised in their ability to antagonize BMP/TGF-b activities in vivo.
The ectopic expression of xNorrin or mouse Norrin in the vegetal cells of whole embryos potently inhibited the expression of the mesoderm-specific marker Xbra, which is dependent on Nodalrelated activity in vivo ( Figure 7D compared to 7B and 7C; Figure  S5B-S5D). We thus used this assay to examine the activity of various xNorrin mutants on Xbra expression. We constructed three xNorrin point mutants (R40K, L60P, and K57N) based on mutations identified from human patients. Compared to wild-type xNorrin, the xNorrin R40K and L60P mutants showed decreased Xnr3, Siamois, and Chordin expression when co-expressed with Lrp5 and Frizzled4 in animal caps, while xNorrin K57N strongly activated these Wnt target genes ( Figure 7A). This is consistent with previous findings using cell culture assays [18,39]. In a wholeembryo assay, the xNorrin R40K mutant largely inhibited Xbra expression, while the xNorrin L60P mutant showed only slight inhibitory activity ( Figure 7E and 7G compared to Figure 7B and 7D). In an extreme case, the xNorrin K57N mutant completely failed to suppress Xbra expression ( Figure 7F). A lack of BMP4 binding ability might explain this loss of TGF-b inhibition. However, only a minor reduction in BMP4 binding was observed for the xNorrin K57N mutant compared to wild-type xNorrin. The xNorrin R40K mutant also did not show significantly reduced binding to BMP4 ( Figure S6).
Next, we examined whether a lack of TGF-b inhibition by xNorrin compromised its neural induction function in a loss-of-function background. Because we could not directly study the K57N mutation through a knock-in experiment in Xenopus, we tested K57N mutant function in xNor-MO-injected embryos. In contrast to wild-type xNorrin, which was able to significantly rescue the anterior defects of the morphants (including eyes in 23% of the embryos) ( Figure S7A-S7D), the K57N mutant was far less efficient, often producing phenotypes similar to those of xNor-MO-injected embryos ( Figure  S7E). We did not observe normal eye formation in any K57Nmutant-injected embryos ( Figure S7F), suggesting that TGF-b inhibition is crucial for the full activity of xNorrin.
Together, these results indicate that Wnt activation and TGF-b inhibition activities are encoded by distinct domains in Norrin proteins and that the loss of TGF-b inhibition in Norrin mutants may be a novel mechanism implicated in the development of Norrie disease in humans (see Discussion).

Discussion
The present work addresses the molecular nature and mechanism of a maternal signal that specifies the early neuroectoderm. Our findings reveal an essential coordination of canonical Wnt signaling activation and extracellular BMP/TGF-b inhibition by maternal xNorrin and further highlight the integration of the two major signaling pathways during early neuroectoderm specification (Figure 8). Our results also point to the de-repression of BMP/TGF-b as a new molecular mechanism in Norrie disease.

Wnt Signaling Induction by xNorrin and Early Neuroectoderm Specification
Canonical Wnt signaling activation in early embryos is essential for the initial dorsal specification [2,25]. Heasman and colleagues previously provided strong evidence that Wnt11 and Wnt5A are endogenous ligands required for b-catenin signaling in all dorsal cells, including dorsal animal cells [5][6][7]. These important findings seem to indicate that any additional Wnt agonists specifically required for b-catenin activation in dorsal animal cells would be redundant. However, previous studies suggested that in Xenopus, an animal-to-vegetal signal was implicated in promoting neural fate before gastrulation, and dorsal animal cells from the blastula are able to develop into neural tissues cell-autonomously in culture [18,28,40]. Noggin and Chordin were discovered to act as neural inducers prior to gastrulation. We demonstrated a lack of bcatenin activation in xNor-MO-injected embryos, which strongly indicated that Wnt11 activity was not sufficient to compensate for the loss of xNorrin activity in vivo ( Figure 3I). The severe neural tissue formation defect in xNor-MO embryos is likely due to a failure in the specification of the early neuroectoderm. The significant down-regulation of the dorsal marker Chordin supports this hypothesis ( Figure 3J-3N).
If both Wnt11 and xNorrin are involved in dorsal specification, then why does maternal xNorrin, which is likely retained in Wnt11-depleted embryos, fail to compensate for the loss of Wnt11 RNA in generating anterior dorsal formation [5]? It is possible that additional molecules are required for xNorrin function in dorsal animal cells. For example, cortical rotation may play a role in the activation of xNorrin signaling. In fact, we found that the dorsal enrichment of xNorrin was lost in UV-irradiated embryos ( Figure 2B and 2C and data not shown). One possibility is that a vegetal signal, such as Wnt11, may be required to fully activate xNorrin signaling in the dorsal ectoderm during cortical rotation. Candidate targets of this vegetal signal may include Xenopus Frizzled4 and Xenopus LRP5, two known receptors for xNorrin [18]. Similarly, the absence of Xnr3 and Siamois expression in xNorrin-injected animal caps can be attributed to the lack of functional xNorrin receptors, which are required for xNorrin signaling ( Figure S2B).

Reciprocal Inhibition between xNorrin and TGF-b
In early embryos, balanced signaling activities from opposite domains are critical for patterning the dorsal-ventral, anteriorposterior, and animal-vegetal axes. For example, in the Xenopus gastrula, ventral BMP molecules antagonize Chordin and Noggin from dorsal cells through direct binding in the extracellular space [1]. Similarly, mesoderm-promoting Nodal activity in the vegetal pole is negatively regulated by maternal TGF-b signaling inhibitors, such as Coco and Ectodermin, from the animal half [41,42]. In addition, the competence of blastomeres to form neural and retinal progeny is repressed by endomesoderm-promoting factors in the vegetal pole [43].
Previously, Coco expressed at the animal pole was proposed as a competence factor to block Nodal signaling and ensure the correct patterning of the ectoderm [41]. Our results indicate that xNorrin also directly inhibits BMP/TGF-b signaling, likely through direct extracellular binding without the activation of Wnt signaling ( Figures 5 and 6). It is possible that this BMP antagonizing activity is required to predispose the dorsal ectoderm toward neural fates before zygotic BMP inhibitors are expressed. Both maternal Coco and xNorrin are expressed in overlapping domains in the animal pole of Xenopus oocytes [41]. It would be interesting to investigate how distinct TGF-b antagonists are coordinated to modulate multiple TGF-b signaling pathways in vivo. Although both Coco and xNorrin are TGF-b antagonists, there is a clear difference in that Coco also functions as a Wnt inhibitor by some unknown mechanism, while xNorrin is a Wnt agonist.
We also showed that the ectopic expression of xNorrin in the vegetal-marginal regions inhibited mesoderm formation and blocked gastrulation ( Figure 7D and data not shown), underscoring the importance of restricting xNorrin activity to the dorsal animal pole. This result seems to suggest that the xNorrinmediated inhibition of mesoderm formation may account for the unexpected failure of NFL to induce complete secondary axes on the ventral side ( Figure 4). However, we observed that the combination of the xNorrin K57N mutant, Frizzled4, and Lrp5 also failed to induce secondary axes (data not shown), suggesting that an alternative mechanism must prevent NFL from inducing secondary axes (Figure 4) [18]. A putative Norrin-specific inhibitor in the endomesoderm other than TGF-b cannot be excluded.
The BMP/TGF-b inhibition function of xNorrin may be attributed to a predicted cysteine-knot domain in the carboxyl terminal ( Figure S1) [44]. A previous bioinformatics study classified the putative Norrin cysteine-knot domain as a mucin protein, along with secretory mucin and von Willebrand factor [45]. Other members of this subgroup may be tested for their potential ability to negatively regulate TGF-b family members.
Conversely, BMP4 was shown to repress Norrin-induced neural formation ( Figure 5H). In addition to xNorrin RNA localization in the dorsal animal region, ventrally expressed BMP4 and vegetally expressed mesoderm inducers, such as Nodal, may further restrict xNorrin activity to the prospective neuroectoderm. Thus, reciprocal inhibition between BMP/TGF-b and xNorrin are equally important for appropriate embryonic patterning.

A Link between TGF-b Signaling and Norrie Disease
Norrin has been identified as an activator of the canonical Wnt signaling pathway through two separate receptor complexes, Frizzled4/Lrp5 and Frizzled4/TSPAN12 [17][18][19]. Given the direct link between Norrin mutations and Norrie disease, and the roles of TGF-b signaling in multiple human diseases, it is important to recognize that Norrin also functions as a potent inhibitor of TGF-b family members. Two lines of evidence indicate that canonical Wnt signaling and TGF-b inhibition are induced separately by Norrins. First, xNorrin can induce neural formation in the absence of Wnt target activation ( Figure 6A). Second, selected xNorrin point mutants (e.g., K57N) potently suppress endogenous TGF-b target gene expression but maintain robust Wnt activation capability (Figure 7).
One of the major defects caused by Norrie disease is abnormal vascular development in the retina and inner ear [18]. The development of the elaborate vascular structure in the retina is strongly influenced by VEGF, which in turn can be positively regulated by TGF-b/ALK5 signaling [46]. Although Norrin mutations cause retinal hypovascularization, a previous study showed that the numbers of blood vessels in the ganglion cell layer and the nerve fiber layer are actually increased in the Norrin knock-out mice [47], suggesting a pro-angiogenesis activity that may be enhanced in these cell layers. Because Norrin hemizygous mutant mice also have severe defects in ear and brain development [18,48], further investigation of local TGF-b regulation in these organs is also warranted.
Finally, the finding that two independent activities are encoded in the small Norrin protein (mature human Norrin has only 109 amino acid residues) raises the question of how Wnt pathway activation and TGF-b signal tuning are coordinated by Norrin in vivo. Solving the three-dimensional structures of the wild-type Norrin protein and selected point mutants may help answer this question and may even help elucidate the molecular mechanisms of Wnt agonist signaling through its receptors.

Embryo Manipulation and Injection
All animal studies in this report were approved by the Institutional Review Board of the Institute of Genetics and Developmental Biology, Chinese Academy of Sciences. X. laevis eggs were isolated in 16 MBS plus high salt solution and fertilized using sperm suspensions in 16 MMR. Embryos were cultured in 0.16 MMR. Embryo dissection was performed as previously described [52]. Briefly, mid-blastula embryos were transferred into 16Steinberg's solution, the vitelline membrane was removed, and 363 mm 2 animal caps were cut. Explants were cultured in 16 Steinberg's solution until they reached the indicated stages [53].
For MO oligonucleotides and mRNA injections, embryos were transferred into 16 MMR containing 2% Ficoll (Amersham Biosciences). Pigment intensity was used to differentiate the dorsal and ventral sides. After injections, embryos were washed thoroughly and returned to 0.16 MMR during the blastula stage.
For UV treatment, embryos were irradiated by placing them in a quartz colorimetric cup oriented with the animal pole upwards and UV-irradiated at 50 mJ using the Stratagene Crosslinker 1800. Immediately after irradiation, the embryos were transferred into 16 MMR containing 2% Ficoll. For rescue experiments, four-to eight-cell-stage embryos were injected with 500 pg of xNorrin mRNA or 500 pg of Wnt11 mRNA [23].

Reverse Transcription PCR
Total RNA was prepared using the Proteinase K method and treated with 10 mg of yeast tRNA and RNase-free DNase (Promega) before cDNA synthesis [52]. cDNA was synthesized by reverse transcription, and the reactions were performed in a volume of 20 ml using 200 ng of random primer (Promega), 56 first-strand buffer, 0.01 M DTT, 40 U RNase inhibitor (TaKaRa), 1 mM each dNTP, and 200 U M-MLV RT (Invitrogen) at 37uC for 50 min. Reactions were then heat-inactivated at 70uC for 10 min and stored at 220uC. One-tenth of the mixture was used as a template for PCR. PCR was carried out in a volume of 25 ml containing 100 mM dNTPs, 0.2 mM each primer, and 1 U of rTaq DNA polymerase (TaKaRa). The PCR parameters and DNA primers are described in Table S1. PCR cycles were determined such that no amplification saturation was reached in semi-quantitative assays.

Luciferase Assays
SuperTopFlash DNA (20 pg), containing eight copies of the TCF-binding site upstream of a minimal TK promoter and the luciferase open reading frame, and pRL-TK DNA (10 pg) (Renilla luciferase was used as an internal control) [5] were co-injected into two dorsal animal cells at the eight-cell stage of wild-type, xNor-MO (20 ng)-injected, or xNor-misMO (20 ng)-injected embryos Three replicate samples for each of the three embryo types were frozen at the late blastula stage, and luciferase assays were carried out using a Promega Luciferase Assay system.

Western Blot
To test for xNor-MO activity, one-cell-stage embryos were injected with 5 ng, 10 ng, or 20 ng of xNor-MO at the marginal zone and then injected four times with a total of 1.5 ng of xNorrin-Myc mRNA into the marginal zone at the four-cell stage. A total of five blastula embryos were homogenized in 100 ml of ice-cold lysis buffer [54]. Protein lysates were spun for 15 min at high speed at 4uC. Protein detection by Western blot was performed using antic-Myc (9E10) primary (Santa Cruz Biotechnology) and HRPconjugated secondary (Pierce) antibodies with Pierce Western blot detection reagents.
For co-immunoprecipitation assays, 500 pg of xNorrin-Myc and 500 pg of BMP4 mRNA were injected into different cells of fourcell-stage embryos. The injected embryos were frozen at stage 10 in batches of five and lysed with 500 ml of ice-cold lysis buffer. The cleared lysate was mixed with anti-FLAG-M2 agarose beads and incubated overnight at 4uC. The beads were pelleted, washed four times with lysis buffer, mixed with SDS-PAGE sample buffer, and processed in a standard electrophoresis and Western blot protocol.

Whole-Mount In Situ Hybridization
Whole-mount in situ hybridization was performed according to a standard protocol as described previously [55], with minor modifications for dissected embryos. For dissected embryos, whole pigmented embryos were fixed for 1 h in MEMFA, bisected along the dorsal-ventral axis with a scalpel blade, fixed for two additional hours in MEMFA, and washed and stored in 100% methanol. The embryos were hybridized at 65uC overnight. BM Purple was used as a substrate (Roche). Pigment was then bleached. The RNA probes were labeled with digoxigenin-UTP (Roche) with the appropriate RNA polymerase using linearized plasmids.

Histological Analysis
For histological analysis, embryos were fixed overnight in Bouin's solution and then dehydrated and embedded in paraffin. Sections of 10-mm thickness were prepared and stained with hematoxylin and eosin as previously described [56]. ODC (ornithine decarboxylase) served as a loading control. RT-: no reverse transcription. Embryos were staged according to Nieuwkoop and Faber [53]. (B) Expression of Siamois and Xnr3 (Wnt target genes) and Xbra (mesoderm marker) in isolated animal caps from embryos co-injected with NFL (200 pg each), xNorrin (200 pg), or Wnt8 (10 pg) mRNAs. Nor, Norrin; WE, whole wild-type embryo; WT, wild type. ODC served as a loading control. (C) Norrin can lead to phosphorylation of its receptor, LRP6. LRP6 phosphorylation at three specific threonine (T) and serine (S) sites (T1479, S1490, and T1493) was analyzed in HEK293 cells transfected with Lrp6/Axin/mFz4, with or without mouse Norrin or xNorrin, using site-specific antibodies. Total LRP was detected using a general LRP antibody.  Figure S6 xNorrin interacts with TGF-b family members. (A) xNorrin binds to BMP4. Epitope-tagged wild-type xNorrin or xNorrin point mutants (R40K or K57N) and BMP4 were separately expressed in HEK293 cells. The conditioned medium from cells expressing individual xNorrin and BMP4 were mixed and incubated. The FLAG-tagged protein complexes were immunoprecipitated using an anti-FLAG antibody and separated in SDS-PAGE and blotted. An anti-c-Myc antibody was used to detect Myc-tagged BMP4. The expression of FLAG-tagged proteins was detected using an anti-FLAG antibody. xNorrin was shown to bind to BMP4. The R40K mutant retained this binding activity, while the K57N mutant showed slightly reduced BMP4 binding. Conditioned medium of the parent FLAGplasmid-transfected cells was used as a control (ctrl). Arrowhead, BMP4-Myc; arrow, immunoglobulin light chains (LC). a-FLAG, anti-FLAG tag monoclonal antibody; a-Myc, anti-c-Myc-tag monoclonal antibody. (B) xNorrin binds to Xnr1 but not DKK-1. The assay performed was similar to that described in (A). Conditioned medium (ctrl) was the same as in (