The putative Wnt receptor Xenopus frizzled-7 functions upstream of β-catenin in vertebrate dorsoventral mesoderm patterning

The role of wingless (wnt) protein signaling in establishment of the vertebrate dorsal axis is currently unclear. Misexpression of wnt proteins causes ectopic axes in Xenopus embryos (see Moon et al., 1997 for review), suggesting that wnts may be involved in the endogenous dorsal signaling pathway. Furthermore, disruption or depletion of several components of the Xwnt pathway (β-catenin, Heasman et al., 1994; Glycogen-synthase Kinase-3β (GSK-3β), Pierce and Kimelman, 1995; Dominguez et al., 1995; He et al., 1995; axin, Zeng et al., 1997; Xgsk-3-binding protein (GBP), Yost et al., 1998; XTcf3, Molenaar et al., 1996) affect dorsal axis formation. Doubts have been raised recently, however, about the complete conservation of a Drosophila-type wingless/pangolin pathway in Xenopus. Recent experiments designed to test for this requirement have failed to identify any role for wnt signaling in endogenous dorsal axis specification. For example, a dominant-negative Wnt8 protein (dnXwnt8; Hoppler et al., 1996), a secreted Wnt8-binding protein (Frzb; Wang et al., 1997; Leyns et al., 1997) and a dominant negative Dishevelled protein (Xdd1; Sokol, 1996), each capable of reducing the ability of exogenous wnt proteins to induce ectopic dorsal axes in Xenopus embryos, failed to perturb the development of the endogenous axis. Based on these findings, many current research publications (for example, see Glinka et al., 1997; Yost et al., 1998) and reviews (i.e. Wodarz and Nusse, 1998) propose that the dorsal signaling pathway may be activated by a novel intracellular mechanism that does not involve wnts or wnt receptors.

Evidence from Caenorhabditis elegans, Drosophila and vertebrate studies have demonstrated that the frizzled family of seven-pass transmembrane proteins are essential for wnt-mediated signal transduction in a variety of biological processes (Herman and Horvitz, 1994; Bhanot et al., 1996; Yang-Snyder et al., 1996; He et al., 1997; Slusarski et al., 1997; Harris et al., 1996; Sawa et al., 1996; Thorpe et al., 1997; Rocheleau et al., 1997; Nasevicius et al., 1998; Bhat, 1998; Kennerdell and Carthew, 1998; Muller et al., 1999). This genetic evidence, coupled with prior interaction data (Bhanot et al., 1996) and new biochemical work in Drosophila demonstrating direct binding between wnt and frizzled proteins in vivo (Wesley, 1999), argues forcibly for a role as a wnt receptor for this family of serpentine proteins. We have tested whether frizzled gene function is required for axis formation in Xenopus to determine the role of cell-cell signaling in this fundamental vertebrate developmental process.

Here we have studied the role of a novel maternal frizzled gene, Xfz7, in Xenopus axis formation. We report the isolation and characterization of Xfz7. We show that Xfz7 acts synergistically with Xwnt8b, Xwnt11 and Xwnt5a in ectopic axis induction, and that the predicted cytoplasmic terminal residues are essential for this interaction. We study the role of maternal Xfz7 using an antisense depletion approach and provide direct evidence for the requirement of Xfz7 in dorsal axis formation. Depletion of Xfz7 results in loss of dorsoanterior structures and markers of dorsal mesoderm, while partial depletion reduces the ability of ectopic Xwnt8b protein to induce secondary axes. These effects are fully reversible both morphologically and molecularly by the subsequent introduction of exogenously added Xfz7 transcripts, demonstrating that these antisense-mediated effects are specific for the Xfz7 gene. We further establish that Xfz7 lies upstream of β-catenin in the dorsal signaling pathway. Thus this work provides direct evidence for a receptor-mediated Xwnt pathway contribution to Xenopus axis formation.

The degenerate primers YW157 and YW158 (Wang et al., 1996) were used to amplify a 210 bp Xfz7 fragment by PCR from a Xenopus oocyte cDNA library (kindly provided by Ernesto Resnik) using TaqHiFi (Boehringer Mannheim) and subsequently ligated into pBluescript (data not shown). This fragment was used to screen a Xenopus oocyte λgt10 library (kindly provided by Joe Yost) by colony hybridization (Genius system, Boehringer Mannheim) according to manufacturer’s instructions. Three clones were isolated and characterized by direct sequencing. The largest 3 kb Xfz7 cDNA S73 included most of the open reading frame (ORF) and a significant stretch of the 3′UTR. We used library and S73-specific primers to amplify by PCR the 5′ end of Xfz7 ORF. 3′UTR sequences were similarly isolated using standard PCR methods. Both strands of each subclone were subjected to DNA sequence analysis. The composite cDNA sequence has been submitted to GenBank under accession number: AF039215. The complete ORF was isolated by PCR amplification using 5′ and 3′ ORF primers with maternal cDNA as template and confirmed by subsequent sequence analysis. Identification of protein motifs was performed using GeneWorks (2.5).

Full-length protein sequences were aligned using GeneWorks (2.5) evolutionary relationship algorithm. Statistically distinct branchpoints are indicated (data not shown).

Whole-mount in situ hybridization was performed as described (Harland, 1997). DIG-labeled antisense Xfz7 probe was synthesized using T3 RNA polymerase and DIG RNA labeling kit (Boehringer Mannheim) from cDNA subclone plasmid S73 after linearization with BamHI. Electrophoresis and northern blotting were performed as described (Hopwood et al., 1989). RNA was isolated as described (Gurdon et al., 1985). The 2.7 kb EcoRI-AlwNI fragment from S73 was used as probe template for Xfz7 detection and labeled with ³²P by random priming. In all cases, the ribosomal bands on the transferred membrane were used as loading controls for the relative abundance of RNA loaded in any given lane.

Other constructs used for northern hybridization were Xnr3 (Smith et al., 1995) and Xwnt8 (Christian and Moon, 1993; derived from plasmid XE9 kindly provided by R. Moon). goosecoid (gsc) probe was synthesized by PCR amplification of gsc ORF fragment from gsc plasmid (Christian and Moon, 1993, generous gift of R. Moon) and subsequent ³²P labeling of this PCR fragment.

RT-PCR for siamois and histone H4 was performed as described (Zhang et al., 1998) using one embryo for each reaction. The primers and PCR profile for siamois and histone H4 were essentially as described (Darras et al., 1997, and Pillemer et al., 1998, respectively), except that 26 amplification cycles for siamois were used. To detect Xfz7 and Xfz3, we used non-radioactive PCR with the following cycle profile: 94°C denaturation for 30 seconds, 57°C annealing for 30 seconds, 72°C amplification for 15 seconds. The following primer pairs were used (5′ to 3′): Xfz7 (forward – TGCCGGCCAC-CATCGTGCTG, reverse – CCTGCGCCAGGACTGTAGGGT; 26 cycles); Xfz3 (forward – TGGCGTGCCGTGGCCGGAAG, reverse – GTGGCAGAGAGGCAAACTATGG; 30 cycles). In both cases, amplification was found to be in the linear range (data not shown).

Plasmid Xfz7/T3TS was constructed by the PCR amplification of the Xfz7 ORF from a maternal cDNA library and subsequently subcloned into pT3TS (Hyatt and Ekker, 1999). Xfz7ΔT/T3TS was generated by PCR using Xfz7/T3TS as template and subsequently subcloned into pT3TS. Each construct was confirmed by DNA sequence analysis.

All synthetic RNA used in this study was made by in vitro transcription using mMessage mMachine kit (Ambion) according to manufacturer’s instructions. RNA concentration was estimated from the intensity of ethidium bromide staining in the agarose gel.

XbaI-linearized Xfz7/T3TS or Xfz7ΔT/T3TS DNA was used as template for RNA synthesis. We used the following templates for wnt RNA synthesis: Xwnt8b in pCS2+ (Cui et al., 1995), XWnt11 in pSP64T (Ku and Melton, 1993), and XWnt5A in pSP64T (Moon et al., 1993; gift of R. Moon).

For injections of Xfz7 and Xfz7ΔT, dejellied embryos at the 2-cell stage were placed in 3% Ficoll in 1× MMR and injected into both cells. 2 ng total of Xfz7 or Xfz7ΔT RNA were injected in the overexpression experiments. Embryos were placed into 0.1× MMR at the midblastula stage for further development. For injections of Xfz7ΔT RNA into oocytes, 2.5 ng of RNA was injected into marginal zones of defolliculated oocytes 4 hours prior to progesterone stimulation. Oocytes subsequently were fertilized using host-transfer technique as described below.

3-10 pg Wnt8b, 600-900 pg Wnt11 or 300 pg Wnt5A were injected into embryos alone or in conjunction with 0.5-1 ng of Xfz7 (or Xfz7ΔT) into one ventral blastomere at the 4-cell stage. The injected embryos were scored for double axes at tailbud stages. For cell movement assays, embryos were injected into both blastomeres at the 2-cell stage with 200 ng of Wnt5A or 200-600 pg of Wnt11 with or without 0.5 ng of Xfz7. Animal caps from the experimental and control embryos were dissected at the midblastula stage and placed into 1× MMR containing 3 ng/ml bovine activin A (Genzyme) and cultured until sibling stage 15.

For histological analysis, sample embryos were fixed in MEMFA at stages 28-32, embedded in paraplast, sectioned into 20 μm sections and stained with hematoxylin as described (Kelly et al., 1991).

For Xfz7 depletion experiments, ovaries were isolated and oocytes manually defolliculated, as described previously (Heasman et al., 1994). The oocytes were injected into the equatorial region with a combination of 2 or 3 ng of oligo Xfz7-37S (sequence: C*T*G*TTCCAACACGGT*G*C*A, asterisk represents phosphothioate linkages) and 2 or 3 ng of oligo Xfz7-43S (sequence: G*T*T*CATGAGCGCCTC*G*C*A) HPLC-purified antisense oligonucleotides (Genosys). The oocytes were cultured for 24 hours, matured for 10 hours following stimulation with 2 μM progesterone, labeled using vital dyes and fertilized using the host transfer technique, as described previously (Heasman et al., 1994). Sample embryos were frozen at stage 10+ as defined by the appearance of the dorsal lip, for marker analysis. In three independent experiments where oocytes were injected with 5.5-6 ng of the oligo mixture, at tailbud stages 22 embryos total were normal (DAI=5), 22 were categorized as DAI=3-4, 26 corresponded to DAI=1-2, and 6 were evaluated as DAI<1. Average DAI = 2.9±0.4. Corresponding numbers of uninjected control embryos were 74, 7, 4 and 0, respectively (DAI = 4.7±0.3).

Histological analysis was performed on stage 40 embryos as described above.

Xnr3 expression was reduced in 14 out of 17 Xfz7-depleted embryo samples (2 embryos per sample) analyzed in 7 independent experiments. Representative membranes from the Xnr3 analyses were reprobed for goosecoid and wnt8 gene expression (Fig. 5A, lanes 1-4). Reduction in siamois level was observed in 9 out of 13 single Xfz7-depleted embryos analyzed in 3 independent experiments.

Individual analyses of Xfz7-depleted embryos (stage 10+) noted extreme examples of Xnr3 reduction (Fig. 5A, lanes 5 and 6). This level of reduction occurred in 8 out of 30 embryos analyzed in 4 independent experiments.

To compare the depletion level of Xfz7 in oocytes injected with different doses of oligonucleotide, northern blots were exposed to a PhosphorImager screen, and the amount of RNA was quantified using the IPLabGel program.

To analyze specificity of the depletion phenotype, 300 pg of synthetic Xfz7 RNA was injected into marginal zones of Xfz7-depleted oocytes 4 hours prior to progesterone stimulation. For epistasis analysis, 4-10 pg of Wnt8b or 200-450 pg of β-catenin RNA was injected into marginal zones of Xfz7-depleted oocytes 4 hours prior to progesterone stimulation.

To test the specificity of Wnt response in Xfz7-depleted embryos, 10-20 pg of Wnt8b RNA was injected into one of the ventral cells at the 4-cell stage into control and depleted embryos. The injected embryos were scored for double axes at tailbud stages. 300 pg of Wnt5A or 600 pg of Wnt11 RNA was injected at the 2-cell stage into both blastomeres of control and Xfz7-depleted embryos. Animal caps from these embryos were dissected at the midblastula stage, placed into 1× MMR containing 3 ng/ml bovine activin A (Genzyme) and cultured until sibling stage 15.

We isolated and characterized a member of the Xenopus frizzled gene family of wnt receptors. The transcribed RNA is predicted to encode a 549 amino acid (a.a.) residues protein with significant homology to the fz family of serpentine receptors (Fig. 1A). BLAST search of nucleotide and protein databases using the sequence of the isolated Xenopus frizzled homologue identified an exceptionally high homology to mouse frizzled 7 (88% identical a.a. residues), therefore it was called Xenopus frizzled 7 (Xfz7). Xfz7 contains features diagnostic of other frizzled family members, such as an amino-terminal signal sequence (Fig. 1A, purple residues), an amino-terminal cysteine-rich domain (Fig. 1A, conserved cysteines in bold), and seven predicted membrane spanning domains (Fig. 1A, underlined). A relatively short poorly conserved cytoplasmic tail with a single PDZ domain binding motif (S/T-X-V; Fig. 1A, red residues) was also noted. Xfz7 falls into the Dfz1 class of fz proteins (Fig. 1B), with 50% identical a.a. residues to Dfz1.

The expression of Xfz7 was characterized using northern analysis and whole-mount in situ hybridization (Fig. 1C-J). Maternal and zygotic Xfz7 transcripts were identified (Fig. 1C). Early maternal transcripts were detected in both animal and vegetal hemispheres, as analyzed by northern hybridization (data not shown). At the late blastula-gastrula stages, Xfz7 is concentrated dorsally and partially excluded from the yolk plug (Fig. 1C-E). During neurulation Xfz7 expression becomes localized to neural folds in presumptive neural-crest-cell-derived areas (Fig. 1F-H). In the late neurula, Xfz7 is localized to some neural crest cell derivatives such as pharyngeal arches and somites (Fig. 1I). By the tailbud stage, Xfz7 transcripts are detected in the eye, ear, heart regions, pharyngeal arches and pronephros (Fig. 1J).

Up to 2 ng of Xfz7-encoding mRNA was microinjected into both cells of 2-cell-stage Xenopus embryos and assayed at subsequent stages of development. At the neurula stage, Xfz7 RNA-injected embryos display incompletely closed neural folds (Fig. 2A,B), which is similar to the results for Xfz3 (Shi et al., 1998) or zebrafish frizzled genes (S. S. and S. C. E., unpublished data). At tailbud stages, Xfz7-injected embryos display shortened axes and are bent dorsally (Fig. 2C,D). In contrast to Xfz8 (Itoh et al., 1995; Deardorff et al., 1998), ventral blastomere injection of Xfz7 does not result in axis duplication (Fig. 2L, last column). In zebrafish embryos, Xfz7 was capable of dorsalization and axis induction (data not shown) as was observed for zebrafish frizzled A and human frizzled 5 genes (Nasevicius et al., 1998).

To identify candidate ligands for Xfz7, we tested the ability of Xfz7 to synergize with the known maternally expressed Wnt proteins, Wnt5A, Wnt8b and Wnt11, in co-injection experiments. Of these three Wnts, only Wnt8b and Wnt11 are known to induce secondary axes when expressed ectopically. We co-injected Xfz7 with Wnts into ventral blastomeres at doses where Wnts alone produced a very low percentage of axis duplications. Co-injection of Xfz7 with Wnt8b, Wnt5A or Wnt11 greatly increased the number of observed ectopic axes (Fig. 2E-H,L). Secondary notochords were observed in a number of sectioned Wnt11+Xfz7 and Wnt5A+Xfz7-injected embryos (Fig. 2I-K). This synergy provides evidence that Xfz7 is likely to function as a wnt receptor during development.

Wnt proteins fall into two distinct functional classes. Wnt-1 class can cause axis duplication when expressed ectopically, which is thought to occur through β-catenin-dependent mechanism (review by Gradl et al., 1999). Wnt5A and Wnt11 class does not cause a significant percentage of ectopic axes but are known to block the elongation of activin-treated animal caps, which mimics the morphogenetic movements normally associated with gastrulation (Moon et al., 1993; Du et al., 1995), an effect believed to be β-catenin independent (Gradl et al., 1999). If Xfz7 was acting as a receptor in this pathway, we would expect Xfz7 to enhance the ability of cells to respond to limiting amounts of wnt proteins. We co-injected Xfz7 with Wnt5A or Wnt11 into both cells at the 2-cell stages at doses where Wnts produce little or no inhibitory effect on the elongation of animal caps (uninjected animal caps, Fig. 2M,R; animal caps injected with low doses of Wnt5A or Wnt11, Fig. 2N and Fig. 2S, respectively). Co-injection of Xfz7 with either Wnt5A or Wnt11 had no significant effect on the elongation of animal caps (see Fig. 2O,Q for Wnt5A+Xfz7, and Fig. 2T,V for Wnt11+Xfz7). Internal controls for Xfz7 and Wnt activities were included in these experiments, e.g., high doses of Wnt5A or Wnt11 RNA which could inhibit animal cap elongation (Fig. 2P and 2U, respectively). Thus Xfz7 does not synergize with Wnt5A or Wnt 11 in inhibiting morphogenetic movements, and Xfz7 is not likely to be functioning as a receptor in this wnt-mediated bioassay.

To address maternal function of Xfz7, we undertook a depletion by antisense oligonucleotides approach. We identified Xfz7-specific antisense oligonucleotides (see Materials and Methods) for which an injected dose of 4 ng per oocyte resulted in the significant depletion of endogenous Xfz7 transcripts (to 12±6% of endogenous levels), as analyzed by northern hybridization (Fig. 3G). These moderate-level Xfz7-depleted embryos produced no detectable phenotype, however, as analyzed morphologically or by early mesoderm markers (Fig. 3B,E; data not shown). To test if Wnt signaling is affected in Xfz7-depleted embryos and to develop a functional assay for these knock-down studies, we injected maternally expressed Wnt8b into a ventral blastomere of Xfz7-depleted embryos. The number of ectopic axes was significantly lower in Xfz7-depleted embryos compared to control embryos injected with Wnt8b alone (Fig. 3A,C,F). Injection of synthetic Xfz7 RNA into Xfz7-depleted oocytes restored the ability of ectopic Wnt8b to induce secondary axes (Fig. 3D,F), demonstrating the specificity of the antisense depletion to Xfz7. A partial reduction in Xfz7 levels is thus sufficient to compromise the ability of Wnt8b to induce ectopic axes without altering the establishment of endogenous dorsoventral mesoderm.

To test the ability of Wnts to inhibit morphogenetic movements in Xfz7-depleted embryos, we injected Wnt5A and Wnt11 into Xfz7-depleted embryos. Animal caps were explanted, treated with activin and scored for elongation. Depletion of Xfz7 did not significantly affect the ability of Wnt5A or Wnt11 to block elongation (Fig. 3H-M). Thus Wnt signaling leading to the inhibition of morphogenetic movements is not affected in embryos moderately depleted of maternal Xfz7 gene function. We cannot exclude a formal possibility though that the blocking of Wnt inhibition is rescued by the zygotic Xfz7 since its transcription starts soon after the animal caps are explanted.

To test if Xfz7 has any endogenous role, we optimized the depletion of maternal Xfz7 transcripts by increasing the amount of oligonucleotide mixture injected into oocytes (up to 6 ng). This dose of Xfz7 antisense oligonucleotides resulted in the nearly quantitative depletion of endogenous maternal Xfz7 transcripts (to 2.0±1.9% of endogenous levels), as assayed using northern analyses (Fig. 4F). 70% of the resulting experimental embryos had various degrees of deficiencies of dorsoanterior structures (Fig. 4A-D,J). The dorsoanterior index (DAI; Scharf and Gerhart, 1983; Kao and Elinson, 1988) was used to estimate the degree of ventralization more precisely at late tailbud stages. Phenotypes ranged from mild reduction of anterior head structures (Fig. 4B) to almost complete loss of dorsal mesodermal structures (Fig. 4D). Histological analysis of the most extremely ventralized embryos revealed the absence of any dorsal structures (Fig. 4H,I). These embryos were reminiscent of those deprived of dorsal mesoderm resulting from UV-irradiation (Scharf and Gerhart, 1983).

If the observed phenotype is due to the loss of Xfz7 function, injection of synthetic RNA into depleted oocytes would be expected to rescue the phenotype. Indeed, injection of synthetic Xfz7 RNA into Xfz7-depleted oocytes significantly reduced ventralization, as judged by the significantly higher percentage of normal embryos and the change of distribution of any affected embryos to weaker DAI categories upon the re-introduction of Xfz7 message (Fig. 4E,J). Fig. 4F demonstrates that we have effectively replaced endogenous Xfz7 transcripts with exogenous mRNA in these RNA rescue experiments.

Each of the oligonucleotides when injected individually at doses up to 5 ng did not cause any significant phenotype (data not shown). The combination of the two oligonucleotides at 3 ng each resulted in more efficient depletion than each of them separately at 6 ng. It is highly unlikely that both oligonucleotides would by coincidence target the same RNA.

To further demonstrate that the ventralization occurs due to specific loss of Xfz7 RNA and not any other frizzled RNA, we analyzed the expression of the only other known maternally expressed Xfz3 homolog (Shi et al., 1998). Xfz3 expression in oligo-injected oocytes was not affected, as analyzed by RT-PCR (Fig. 4G).

Sample embryos were frozen before gastrulation (stage 9.5) and at the start of gastrulation (stage 10+, as determined by appearance of dorsal lip in each embryo being analyzed) and analyzed by northern hybridization for expression of dorsal and ventral mesoderm markers. Representative samples (2 embryo equivalents per lane) are shown in Fig. 5A, lanes 1-4; the more extreme cases presumably corresponding to DAI<1 embryos are represented in Fig. 5A, lanes 5-6 (1 embryo per lane). Expression of the dorsal marker Xnr3, a known direct target of β-catenin signaling (Smith et al., 1995; McKendry et al., 1997), was significantly reduced in Xfz7-depleted embryos (Fig. 5A; n=7 experiments, see Materials and Methods). Expression of another direct target of Wnt signaling, siamois (Fan et al., 1998; Brannon et al., 1997; Lemaire et al., 1995), was reduced in Xfz7-depleted embryos, as analyzed by RT-PCR (Fig. 5C). Another known dorsal mesoderm marker goosecoid (gsc; Cho et al., 1991) was reduced while expression of a ventral mesodermal marker Wnt8 (Christian et al., 1991) was not affected (Fig. 5A). A later stage analysis demonstrates that Xnr3 levels in Xfz7-depleted embryos never reach control embryo levels (Fig. 5B), indicating that these defects in dorsal marker induction are not due to a simple delay in embryonic development. Injection of synthetic Xfz7 RNA into Xfz7-depleted oocytes restored Xnr3 expression (Fig. 5D). These results show that Xfz7 gene function is necessary for the induction of dorsal mesoderm.

To test whether Xfz7 functions in the known early Wnt pathway, we injected Xwnt8b into Xfz7-depleted oocytes (6 ng of oligonucleotide mixture was used in these experiments). The embryos were frozen just before the onset of gastrulation (stage 9.5) and assayed for induction of the Wnt pathway target gene Xnr3 by northern hybridization. Injection of Wnt8b RNA into control oocytes resulted in strong induction of Xnr3, while injection of Wnt8b into Xfz7-depleted oocytes resulted in the reduced level of Xnr3 expression compared to uninjected controls (Fig. 5E). Indeed, these embryos displayed no ability to respond to ectopic Wnt signaling, as assayed by Xnr3 induction. To test if Xfz7 functions upstream of β-catenin, we injected β-catenin RNA into Xfz7-depleted oocytes and assayed for the induction of Xnr3. In contrast to the Wnt8b results, injection of β-catenin fully restored Xnr3 expression in Xfz7-depleted oocytes (Fig. 5F). These results place endogenous Xfz7 function between an axis-inducing Wnt and β-catenin in the dorsal mesoderm-induction pathway (Fig. 5G).

To test the requirement of the putative Xfz7 cytoplasmic tail for fz-mediated wnt signal transduction, we generated an altered form of Xfz7, which lacked the last 25 amino acid residues in the cytoplasmic tail (Xfz7ΔT). We injected a low dose of ectopic Wnt8b with wild-type Xfz7, Xfz7ΔT or control GFP RNA into a ventral blastomere at the 4-cell stage and scored the injected embryos at tailbud stages for ectopic axes. As expected, adding Xfz7 increased the percentage of embryos with a secondary axis (Fig. 6A-C). Adding Xfz7ΔT, however, did not result in the higher percentage of embryos with secondary axes (Fig. 6A-C). The percentage of the embryos with secondary axes was indeed lower than in the case of injecting Wnt8b with inert GFP RNA. To specifically test for the antagonistic effect of Xfz7ΔT, we injected a higher amount of Wnt8b RNA with or without Xfz7ΔT RNA into a ventral blastomere and scored for secondary axis formation. Addition of Xfz7ΔT greatly reduced but did not abolish the ability of these embryos to form a secondary axis (Fig. 6D-F), demonstrating an antagonistic function of this truncated form of Xfz7. The cytoplasmic tail of Xfz7 is therefore required for Xfz7-mediated wnt signal transduction.

We tested the effects of the addition of this antagonistic form of Xfz7 on embryonic development. Injection of Xfz7ΔT-encoding mRNA resulted in embryos with abnormally elongated blastopores during gastrulation (Fig. 6I,J). Overexpression of Xfz7ΔT caused inhibition of morphogenetic movements as seen by the inhibition of marginal zones elongation explanted at the blastula stage (Fig. 6M,N). Xfz7ΔT-overexpressing embryos developed into shortened embryos with dorsally bent tails (Fig. 6G,H), similar to the ones resulting from overexpressing of the wild-type Xfz7 (see Fig. 2C,D). It may seem surprising that overexpression of wild-type Xfz7 and Xfz7ΔT yields similar phenotypes. However, it is known that Dfz1 overexpression can mimick loss of function in Drosophila (Krasnow and Adler, 1994). The mechanism of this effect is not known but may include the titration of ligand at high doses by wild-type fz protein. At equivalent low doses of Xfz7 and Xfz7ΔT mRNA, only Xfz7ΔT effectively causes the bent back phenotype.

To test if Xfz7ΔT could block endogenous Xfz7 signal transduction when expressed very early in development, we injected Xfz7ΔT into oocytes, which were later fertilized using the host-transfer method. Such overexpression of Xfz7ΔT resulted in a more severe phenotype where embryos were extremely shortened, did not close neural folds properly and had no identifiable heads or dorsal mesoderm structures (Fig. 6K,L). This phenotype correlated with the reduced expression of dorsal organizer gene Xnr3, as analyzed by northern hybridization (Fig. 5O). These results support the idea that Xfz7 functions very early in development in dorsal mesoderm induction.

The data provided in this article strongly argue for the requirement of Wnt signaling in Xenopus dorsoventral axis formation. The axis-induction mechanism, as well as many other developmental events, is quite likely to be evolutionally conserved. A dominant negative frizzled molecule can block dorsal organizer formation in zebrafish (Nasevicius et al., 1998). This study implicated fz signaling in axis formation but, due to the limitations of the methodology, did not identify the endogenous fz protein encoding such a function. Furthermore, Wnt3 knockout mice do not form a primitive streak, mesoderm or node (Liu et al., 1999), demonstrating that wnt pathway is required for murine axis specification. Our results argue that primary embryonic axis induction is highly conserved between frogs, fish and mice.

The question of what is the ligand for Xfz7 still remains open. Each maternally encoded wnt gene that we tested was capable, when expressed at sufficiently high levels, of activating Xfz7 for dorsoventral mesoderm specification. Consequently, we cannot exclude any of these as likely ligands. The only maternal Wnt protein known to be asymmetrically distributed along the dorsoventral axis is Wnt11 (Schroeder et al., 1999), which makes it a likely candidate for the endogenous axis induction. Alternatively, we have found in Xenopus a maternally expressed mouse Wnt3 homolog (S. S. and S. C. E., unpublished data). Future experiments will have to determine the actual ligand for Xfz7.

Our results demonstrate that Xfz7 functions before the mid-blastula transition in axis specification, and we do not really know more precisely when its function is required. Since we show that β-catenin functions downstream of Xfz7, Xfz7 is likely to transduce a signal even earlier than β-catenin. β-catenin has been shown to be differentially localized to nuclei in Xenopus embryos by the 32-cell stage and in the cytoplasm as early as the 2-cell stage of development (Larabell et al., 1997). If β-catenin is solely regulated by Xfz7 activation, then Xfz7 has to function very early in development.

How can Xfz7 be specifically activated in Xenopus this early in embryonic development? An example of wnt pathway functioning at the 4-cell stage in orienting a mitotic spindle comes from C. elegans (Schlesinger et al., 1999). In Xenopus there are no distinct embryonic regions formed yet at this stage of development, yet autocrine or paracrine wnt signaling might participate in polarizing the dividing cells by activating receptors on a particular side of a cell as it is seen in Drosophila wing hair induction (Krasnow et al., 1995).

Alternatively, Xfz7 may be activated ubiquitously in an embryo, possibly even by a number of maternally expressed Wnts. If at least a single downstream component of the pathway is localized asymmetrically, as is known to be the case for Dishevelled (Miller et al., 1999), the ubiquitous activation of Xfz7 would result in a localized signal transduction. This model is actually in agreement with asymmetry established by cortical rotation data. So, according to Miller et al. (1999), cortical rotation would result in the dorsal enrichment of Dsh. An inactivated Dsh may not be able to effectively transduce the signal. Ubiquitous Wnt signaling through Xfz7 would result in ubiquitous Dsh activation. But, since Dsh is already enriched dorsally due to cortical rotation, higher signaling would be transmitted on the dorsal side, resulting in the dorsal enrichment of β-catenin.

We cannot exclude the possibility that β-catenin could also be regulated by another, Xfz7-independent pathway. The strongest phenotypes observed in Xfz7-depleted embryos, however, argue that this putative alternative pathway is not always sufficient for endogenous axis specification. Analysis of a strong loss-of-function phenotype for Xfz7 ligand(s) could potentially distinguish between these single and dual signal models.

The very early requirement for Xfz7 function may explain why previous attempts to block Wnt signaling at the ligand or receptor level have failed. This is supported by our findings that the Fz7 truncation ΔT had to be injected very early (oocytes) to cause ventralization; similar doses injected later into embryos caused other defects, presumably due to interference with zygotic Wnt pathways, but did not cause ventralization.

We thank M. O’Connor, J. Lohr, M. Kofron and anonymous reviewers for critical comments on this manuscript. This work was supported by the Minnesota Medical Foundation and the National Institutes of Health (R01GM55877 to S. C. E. and R01HD33002 to J. H.).