Dorsoventral (DV) patterning of vertebrate embryos requires the concerted action of the Bone Morphogenetic Protein (BMP) and Wnt signaling pathways. In contrast to our understanding of the role of BMP in establishing ventral fates, our understanding of the role of Wnts in ventralizing embryos is less complete. Wnt8 is required for ventral patterning in both Xenopus and zebrafish; however, its mechanism of action remains unclear. We have used the zebrafish to address the requirement for Wnt8 in restricting the size of the dorsal organizer. Epistasis experiments suggest that Wnt8 achieves this restriction by regulating the early expression of the transcriptional repressors Vent and Vox. Our data show that vent and vox are direct transcriptional targets of Wnt8/β-catenin. Additionally, we show that Wnt8 and Bmp2b co-regulate vent and vox in a dynamic fashion. Thus, whereas both Wnt8 and zygotic BMP are ventralizing agents that regulate common target genes, their temporally different modes of action are necessary to pattern the embryo harmoniously along its DV axis.
Formation of the vertebrate embryonic axes requires Wnt signaling at two points: after fertilization, to establish a dorsal signaling center, and during gastrulation, to pattern and specify ventral fates (for reviews, see De Robertis et al., 2000; Schier, 2001). Although canonical Wnt/β-catenin signaling is involved in both processes, it is triggered differently in each case. Specification of the dorsal signaling center appears to be a ligand-independent mechanism involving the accumulation of β-catenin, the nuclear effector of Wnt signaling, in dorsal nuclei(Larabell et al., 1997; Kelly et al., 2000; Schier, 2001). Accumulation of nuclear β-catenin leads to the formation of the Niewkoop center, which induces the dorsal mesodermal structure known as Spemann's Organizer (known as the `shield' in zebrafish or the `node' in the mouse) (for a review, see Moon and Kimelman, 1998). After the establishment of the dorsoventral (DV) axis, Wnt/β-catenin activity stimulated by the ligand Wnt8 is required to antagonize the organizer; thus, zebrafish wnt8 mutants, or Xenopus embryos expressing a dominant-negative Xwnt8, display enlarged organizers and concomitant loss of posterior and ventral tissues(Hoppler et al., 1996; Lekven et al., 2001). Because proteins secreted by the organizer are known to be required for head formation and embryonic patterning (for a review, see De Robertis et al., 2000),understanding the mechanisms that limit organizer expansion is crucial for understanding embryonic patterning.
The organizer influences DV patterning through its secretion of BMP inhibitors such as Chordin (Chd) or Noggin(De Robertis et al., 2000). However, BMP also exerts its own effect on the organizer. The Xvent ventral homeobox genes were identified as transcriptional targets of BMP in Xenopus, and were shown to repress organizer gene expression on the ventral side of the embryo (Gawantka et al., 1995; Onichtchouk et al.,1996; Onichtchouk et al.,1998; Melby et al.,1999; Lee et al.,2002). Indeed, Xvents repress the transcription of targets such as chd and goosecoid (gsc)(Onichtchouk et al., 1996; Melby et al., 1999; Trindade et al., 1999). Analysis of the Xvent1b and Xvent2b promoters revealed the presence of consensus Lef/Tcf binding sites(Friedle and Knöchel,2002). In addition, the Xvent1b promoter is responsive to zygotic Wnt activity, suggesting that the expression of Xvent genes in general may be under the control of Wnt8 (Friedle and Knöchel, 2002). In support of this, Hoppler and Moon found that overexpression of dn-Xwnt8 leads to the reduction of both Xvent1 and Xvent2 expression in Xenopus(Hoppler and Moon, 1998). Thus, these studies suggest that the expression of transcriptional repressors required to restrict organizer gene expression may be under the concerted control of both the BMP and Wnt pathways.
Genetic analysis of zebrafish vent (also known as vega2,similar to Xvent1) and vox (also known as vega1,similar to Xvent2) showed that the proteins encoded by these genes function as redundant transcriptional repressors(Kawahara et al., 2000; Melby et al., 2000; Imai et al., 2001). Zebrafish embryos homozygous for a chromosomal deficiency of the closely linked vent and vox loci show an expansion of organizer gene expression and severe DV patterning defects(Imai et al., 2001). Further epistatic analysis suggested that the primary role of Vent and Vox is to modulate BMP inhibitors secreted by the organizer(Imai et al., 2001). vent and vox are known BMP transcriptional targets in zebrafish as well, but their dependency on BMP signaling starts at around 70-75% epiboly (Kawahara et al.,2000; Melby et al.,2000). As a result, zygotic BMP mutants do not have expanded organizers as vent/vox mutants do at shield stage(Mullins et al., 1996;Miller-Bertoglio, 1997; Imai et al.,2001). To date, only two zebrafish zygotic mutants are known to display significantly expanded organizers: vent/vox mutants and wnt8 mutants. These data suggest that the relationship between BMP, Wnt8 and Vent/Vox is an important one for organizer regulation, the nature of which has been unclear but has been suggested to be complex(Hoppler and Moon, 1998; Marom et al., 1999).
We have used a loss-of-function approach in zebrafish to study the relationship between Wnt8, zygotic BMP and Vent/Vox regulation and activity,in order to understand the mechanism by which Wnt8 antagonizes the organizer. Our results suggest that Wnt8 directly regulates the transcriptional levels of vent and vox, and that the maintenance of high levels of vent or vox is required for the repression of organizer genes on the ventral side of the embryo. Furthermore, we provide evidence that Vent and Vox are absolutely essential to mediate the organizer repression activity of Wnt8. We also show that organizer repression and the maintenance of ventrolateral mesoderm fates appear to be independent events. Finally, we show that the early regulation of both vent and vox is under Wnt8 and BMP control, but that Wnt8 is the primary regulator; that is, at the onset of gastrulation, the requirement for BMP is only revealed in the absence of Wnt8. Zygotic BMP becomes the primary regulator of vent (but not vox) transcription during mid to late gastrulation. Therefore, Wnt8 and BMP contribute to the repression of the organizer, which will, as a consequence, regulate the distribution of Wnt and BMP inhibitors.
Materials and methods
Fish maintenance and genetics
Animals were maintained as described(Westerfield, 2000). Embryos were staged according to Kimmel et al.(Kimmel et al., 1995). Our wild-type strain is AB. Mutants used were Df(LG14)wnt8w8(Lekven et al., 2001), DfST7 (Imai et al.,2001) and swrTC300(Mullins et al., 1996). Results from wnt8 or vent;voxdeficiency mutants were confirmed with morpholinos (MOs).
In situ hybridization
In situ hybridizations were performed as described(Oxtoby and Jowett, 1993). Probes used were gsc (Stachel et al., 1993), chd(Miller-Bertoglio et al.,1997), wnt8 ORF1 and wnt8 ORF1+ORF2(Lekven et al., 2001), eve1 (Joly et al.,1993), vent/vox(Melby et al., 2000), bmp2b (Kishimoto et al.,1997), opl (Grinblat et al., 1998), pax2a(Krauss et al., 1991) and tbx6 (Hug et al.,1997).
Genotyping of embryos
wnt8 mutants were genotyped as described(Lekven et al., 2001). vent;vox mutants were genotyped using vox R1(5′-GATATTGCACACCAGCGTGA-3′) and vox L1(5′-GTTCCAGAACCGAAGGATGA-3′) primers. swr mutants were genotyped as described (Wagner and Mullins, 2002). Embryos were classified according to their phenotype, photographed and genotyped. For wnt8;swr double mutants,at least 85 embryos from an intercross were examined in the same fashion.
Embryo microinjection, morpholinos, constructs
MOs (Genetools, LLC), RNA or DNA were injected into one- to four-cell stage embryos. Approximately 3 nL was injected per embryo. Capped mRNAs were synthesized using mMESSAGE mMACHINE (Ambion) and diluted in water. MOs were diluted in Danieau's buffer as recommended (Genetools). wnt8 MOs(targeting ORF1 and ORF2), and vent and vox MOs, have been described (Lekven et al.,2001; Imai et al.,2001). GR-LEFΔN-βCTA RNA was injected at 300 ng/μL into one-cell stage embryos. Embryos were dechorionated manually in fish water(Westerfield, 2000) prior to treatment. Dexamethasone (DEX; Sigma) treatments were performed for one hour at 1, 2, 3, 4 or 5 hours post-fertilization (HPF). DEX (100 mM stock solution in 100% ethanol) was used at a final concentration of 10 mM in 0.3×Danieau's solution. Treated embryos were fixed at 6 HPF. For the Cycloheximide (CHX; Calbiochem) treatment, embryos were first injected with GR-LEFΔN-βCTA RNA then treated with CHX (10 μg/mL), with or without DEX. For vent induction analysis, n(CHX)=37 and 55, n(DEX)=44, 37 and 11, and n(CHX+DEX)=28, 34 and 28, where n=total number of embryos analyzed in each experiment. For vox induction, n(CHX)=16, 17 and 12, n(DEX)=5, 12 and 20, and n(CHX+DEX)=9, 14 and 19. As a control for CHX treatments,uninjected embryos were treated with CHX from 1.5 HPF to sphere stage, then fixed and stained for gsc (Leung et al., 2003). No treated embryos expressed gsc(n=34). The χ2 test was used to determine statistical significance.
Zebrafish wnt8 and vent;vox mutants have expanded organizers, swr mutants do not
Although BMP and Wnt8 both are described as `ventralizing agents' (i.e. overexpression leads to a shift in mesodermal fates), they play non-equivalent roles in DV patterning. To illustrate this, we compared the expression of DV markers in wnt8 (Dfw8)(Lekven et al., 2001), vent vox (Dfst7)(Imai et al., 2001) and bmp2b (swrtc300)(Mullins et al., 1996)mutants.
In zebrafish, wnt8 contains two open reading frames (ORF1 and ORF2) (Lekven et al., 2001). The two Wnt8 proteins were shown to function redundantly in anteroposterior(AP) and DV patterning, as the Dfw8 phenotype is phenocopied only by co-injection of both ORF1 and ORF2 MOs(Lekven et al., 2001). Similarly, the Dfst7 phenotype is phenocopied by the co-injection of vent and vox MOs(Imai et al., 2001).
Expression analysis of the dorsal markers chd, gsc, floating head(flh) and dharma (bozozok) at shield stage shows that they are expanded ventrally in wnt8 mutants(Fig. 1B,F)(Lekven et al., 2001) (and data not shown) as well as in vent;vox mutants(Fig. 1C, inset, and Fig. 1G)(Imai et al., 2001). swr mutants, however, do not exhibit a similar expansion at shield stage (Fig. 1D,H)(Mullins et al., 1996; Miller-Bertoglio et al.,1997). Importantly, the expansion of dorsal markers is stronger in vent;vox mutants than in wnt8 mutants. For instance, gsc encircles the margin of vent;vox mutants(Fig. 1C, inset) but extends over a ∼90° arc in wnt8– embryos at the same stage (Fig. 1B). This comparative analysis shows that Wnt8 and Vent/Vox, but not BMP, are normally required ventrally during gastrulation to restrict the size of the organizer,which is in agreement with previous reports(Mullins et al., 1996; Miller-Bertoglio et al., 1997; Imai et al., 2001; Lekven et al., 2001).
The expanded organizer phenotype is first observed in wnt8– embryos at 40% epiboly (discussed below), a developmental timepoint when convergence movements have not yet started(Kimmel et al., 1995). Thus,the expansion of dorsal markers in these backgrounds must reflect a change in fate rather than an alteration of cell movements.
Wnt8 is also required to promote ventral fates. eve1, a ventral mesodermal marker, is reduced in wnt8 mutants(Fig. 1B). It is similarly reduced in swr mutants (Fig. 1D) (Mullins et al.,1996). By contrast, eve1 is less reduced in vent;vox mutants (Fig. 1C) than in wnt8 and swr mutants(Fig. 1B,D), despite the fact that the dorsal markers gsc (Fig. 1C, inset) or chd(Fig. 1G) encircle the margin of the same embryos. Hence, Wnt8 and BMP are required in the ventral mesoderm for the maintenance of eve1, a ventral-specific gene, and this function is separable from repression of the organizer.
Wnt8 regulates vent and vox mRNA levels
Because Wnt8 and Vent/Vox share the function of repressing dorsal genes, we analyzed their epistatic relationship. We first examined vent and vox mRNA levels in wild-type versus wnt8–backgrounds (Fig. 2). In zebrafish, vent is expressed at the mesodermal margin during gastrulation, whereas vox displays both ventral mesoderm and ectoderm expression (Melby et al.,2000).
Starting at 30% epiboly (late blastula), the accumulation of ventat the margin is visibly weaker in wnt8 mutants or morphants than in wild type (Fig. 2A-C). We did not detect any differences in vent expression at earlier stages (data not shown). vox expression is not visibly different in wnt8mutants at 30% epiboly (data not shown), but is reduced in the margin of wnt8 mutants/morphants at 40% epiboly(Fig. 2G-I).
To determine the correspondence between vent and voxreduction and the onset of an observable phenotype in wnt8 mutants,we examined chd expression at these early stages. At 30% epiboly, no visible difference in the chd expression domain was observed in wnt8 mutants (data not shown), but we did detect an expansion of chd expression at 40% epiboly, the timepoint at which both vent and vox are reduced in wnt8–embryos (Fig. 2M-O). Hence, our results suggest that a reduction in both vent and vox levels may be required to observe the expanded organizer phenotype at 40% epiboly,which is consistent with Vent and Vox functioning redundantly(Imai et al., 2001).
During the rest of gastrulation, vent and vox mRNA levels stay reduced in wnt8 mutants/morphants compared with in wild type(Fig. 2D-F,J-L; data not shown). By comparison, vent and vox levels are unchanged in swr mutants at shield stage(Kawahara et al., 2000; Melby et al., 2000), which explains the lack of an organizer phenotype(Mullins et al., 1996; Miller-Bertoglio et al.,1997). Indeed, Bmp2b is only required at mid to late gastrulation for the maintenance of vent and ectodermal vox expression(Melby et al., 2000). Therefore, Wnt8 regulation of vent and vox starts at the blastula/gastrula transition (30/40% epiboly), whereas Bmp2b regulation of these genes occurs later (70% epiboly).
To test the reciprocal possibility of wnt8 being regulated by Vent and Vox, we looked at the expression of wnt8 in vent;vox mutants (Fig. 3). As zebrafish wnt8 produces transcripts for both protein coding regions, we used probes to detect either the ORF1/ORF2 bicistronic transcript (ORF1), or both the bicistronic transcript and the ORF2 transcript (ORF1+ORF2) (Lekven et al.,2001). No differences from wild-type expression were observed in 30% or 40% epiboly vent;vox mutants(Fig. 3A,B,G,H). Because vent;vox mutants are affected prior to 30% epiboly(Imai et al., 2001), this suggests that a change in wnt8 expression is not responsible for the vent;vox mutant phenotype. The dorsal domain lacking ORF1 expression is slightly expanded in vent;vox mutants at shield stage(Fig. 3C,D; confirmed with MOs)and is more pronounced at 75% epiboly (Fig. 3F). Although there is an observable difference dorsally, ORF1 levels ventrally seem to be unaffected in vent;vox mutants(Fig. 3C-F), suggesting that the reduction in dorsal wnt8 ORF1 expression is an indirect consequence of an enlarged organizer. Analysis of ORF2 expression at later stages revealed that it is not affected by the loss of Vent and Vox(Fig. 3I-L). This is not unexpected as wnt8 ORF2 accumulates dorsally during gastrulation(Fig. 3K) and is therefore insensitive to molecules present in the organizer. Thus, only wnt8ORF1 expression depends on Vent and Vox, but this dependency is restricted dorsally and may be indirect. By comparison, wnt8 ORF2 expression does not depend on Vent and Vox.
Wnt8 functions through β-catenin to regulate vent and vox transcription
The above data show that Wnt8/β-catenin is necessary to maintain normal vent and vox expression. To test whether Wnt8 is sufficient to induce vent and vox, we injected Wnt8 ORF1 or ORF2 expression plasmids into wild-type embryos and assayed vent and vox expression by in situ hybridization at shield stage. In both cases, ectopic domains were observed in the animal ectoderm region and/or dorsal mesoderm, where vent and vox are normally absent(Table 1, and data not shown). To confirm that canonical Wnt signaling was involved in vent and vox regulation, we modulated β-catenin activity using a hormone inducible β-cat/Lef fusion protein (GR-LEFΔN-βCTA)(Domingos et al., 2001). The GR-LEFΔN-βCTA protein contains the human glucocorticoid receptor domain fused to the DNA-binding domain of murine LEF and the transactivation domain of murine β-catenin. Addition of the hormone dexamethasone (DEX)leads to the nuclear translocation of the fusion protein and toβ-catenin/Lef-induced transcription, thus allowing controlled induction of Wnt signaling (Domingos et al.,2001). Addition of DEX for a one-hour period at 1, 2, 3, 4 or 5 HPF led to ectopic vent and vox expression in a proportion of injected embryos (∼50-70% of embryos; Fig. 4A, panels b,d; data not shown). Consistent with the role of β-catenin in organizer induction,ectopic gsc was observed in a proportion of embryos treated at 1, 2 or 3 HPF, but not at later timepoints (data not shown).
Although our results suggest that Wnt8/β-catenin regulates vent and vox transcription, it is unclear whether this is direct (through β-catenin/Lef-induced transcription) or indirect (through the synthesis of an intermediate transcriptional regulator). Interestingly,the genomic region upstream of zebrafish vox contains consensus Lef/Tcf binding sites consistent with Wnt regulation of voxtranscription (our own observations, and D. Kimelman, personal communication). To address this, we used cycloheximide (CHX) to test whether protein synthesis is required for the induction of ectopic vent or vox by GR-LEFΔN-βCTA. Treatment of GR-LEFΔN-βCTA-injected embryos with DEX at 5 HPF results in ectopic vent or vox RNA expression in 49% and 62.1% of embryos, respectively(Fig. 4B). Addition of CHX simultaneously with DEX did not result in a statistically different number of embryos with ectopic vent and vox domains (72.2% and 59.5%; Fig. 4B), indicating that GR-LEFΔN-βCTA activation of vent and vox does not require de novo protein synthesis. Thus, our results suggest that vent and vox are direct transcriptional targets of Wnt8/β-catenin signaling.
Wnt8 repression of the organizer requires Vent/Vox
As vent and vox transcription is regulated by Wnt8, we hypothesized that Vent and Vox function downstream of Wnt8 to repress dorsal genes, and that the wnt8– organizer phenotype is due to reduced vent and vox levels. If this is correct,injection of vent or vox RNA or DNA into wnt8mutants would suppress the expanded organizer phenotype. We first established amounts of injected Vox or Vent that are sufficient to reduce the expression of dorsal markers (gsc, chd, flh) in wild-type embryos(Fig. 5A, panels a,c; data not shown). When injected into wnt8 mutants, Vox was able to reduce the expression of dorsal genes (Fig. 5A, compare panels b and d; Table 2). Similar results were obtained with either DNA or RNA injection for both vent and vox (Table 2, and data not shown). Thus, Vent and Vox expression can bypass wnt8loss-of-function in repressing organizer genes, thus supporting the placement of vent and vox genetically downstream of wnt8. These results suggest that the difference in severity of the wnt8– and vent–;vox– organizer phenotypes (see Fig. 1) could be explained by residual Vent and Vox activity in wnt8 mutants. In agreement with this, further reduction of Vent and Vox in wnt8mutants by injection of sub-maximal concentrations of vent and vox MOs enhances the severity of the wnt8–phenotype (Fig. 5B).
While Vent and Vox can bypass Wnt8 to repress organizer genes, we wished to assess whether Wnt8 requires Vent and Vox to repress the organizer. If Vent and Vox are essential for this Wnt8 function, then Wnt8/β-catenin activity should be ineffective in their absence. In support of this, vent;vox mutants express nearly normal levels of wnt8 mRNA(see Fig. 3), hence the expansion of the organizer in vent;vox mutants occurs in the presence of wnt8 transcripts.
To confirm that the wnt8 transcripts in vent;vox mutants produce functional proteins, we used two assays of Wnt8 function. First, we examined the expression of the Wnt/β-catenin activity reporter TOPdGFP(Dorsky et al., 2002; Lewis et al., 2004). We analyzed the expression of TOPdGFP mRNA at 100% epiboly in embryos homozygous for the transgene after injection of wnt8 or vent+vox MOs(Fig. 6A-D). As expected, and confirming previous results (Phillips et al., 2004), wnt8 MOs severely reduce TOPdGFP expression in 90% of injected embryos to almost undetectable levels (n=20; Fig. 6B). In vent/voxmorphants, three phenotypic classes were observed: the first class displayed wild-type TOPdGFP expression (50%, n=22; Fig. 6C); the second class showed moderate reduction in TOPdGFP (14%, not shown); and the third class displayed a stonger reduction in staining (36%; Fig. 6D), but this class had significantly more TOPdGFP expression than wnt8 morphants (compare Fig. 6D to Fig. 6B). As a control for the strength of the vent+vox MO injections, a sample of the injected embryos was examined at 24 HPF and all showed a strong vent/vox loss-of-function phenotype (n=23)(Imai et al., 2001). Thus,TOPdGFP is a reporter of Wnt8 activity and is still expressed in vent+vox morphants. Reduced levels of TOPdGFP expression in some vent+vox morphants could reflect the fact that expression of the Wnt antagonists Dickkopf 1 and Frzb is significantly expanded (Imai et al., 2001)(and our own observations).
To confirm that expressed Wnt8 actively patterns vent;vox mutants,we analyzed AP neural patterning, a function known to require Wnt8(Lekven et al., 2001; Erter et al., 2001). To assess the AP phenotype of vent;vox mutants, a combination of three probes was used: opl (anterior neuroectoderm), pax2a(midbrain-hindbrain border) and tbx6 (posterior non-axial mesoderm). In wnt8 mutants or morphants, AP patterning is severely disrupted at 90%-100% epiboly: the opl domain is expanded along the AP axis, pax2a expression is delayed and tbx6 expression is strongly reduced (Fig. 6F). By comparison, vent;vox mutants have only mildly affected AP patterning illustrated by a slight posterior shift of the opl and pax2adomain away from the animal pole, but the distance between opl or pax2a and tbx6 is significantly greater than in wnt8 morphants (Fig. 6G, compare with Fig. 6F). As expected, the expanded organizer of vent;voxmutants results in an enlarged dorsal clearing of tbx6 expression,whereas the levels of tbx6 ventrally are relatively unaffected(Fig. 6G, compare with Fig. 6E). As tbx6expression depends on Wnt8, our results do not support an absence of Wnt8/β-catenin activity in vent;vox mutants. Furthermore,reducing Wnt8 translation in vent;vox mutants results in an additive phenotype. opl extends ventrally, as in vent;vox mutants,whereas pax2a and tbx6 expression is severely reduced, as in wnt8 mutants (Fig. 6H). Taken together, these results show that Wnt8 expression and patterning activity does not depend on Vent and Vox, with the significant exception that Wnt8 is unable to repress organizer genes when Vent and Vox are absent.
To further show that Wnt8 requires Vent and Vox in organizer repression, we tested whether exogenous Wnt8 can repress organizer genes in vent;voxmutants. We injected a wnt8 ORF1 expression plasmid (20 ng/μL)into one-cell stage vent;vox mutants and assayed gscexpression at shield stage. No injected vent;vox mutant embryos(n=25; genotyped by PCR) displayed reduced gsc expression,although this treatment did result in decreased gsc expression in wild-type siblings (n=54). As a control, we checked that the injected wnt8 DNA was sufficient to induce ectopic vent and vox expression in wild-type embryos (64% ectopic expression for vent, n=25; 42.8% ectopic expression for vox, n=35). Thus,repression of the organizer by exogenous Wnt8 requires Vent or Vox.
Our results show that in the absence of Vent and Vox, wnt8 is expressed and is active, as assayed by TOPdGFP reporter expression, tbx6 expression and embryonic AP patterning. Furthermore, ectopic Wnt8 cannot repress gsc in vent;vox mutants. These data strongly support a linear model in which Wnt8 acts directly upstream of Vent and Vox to repress the organizer.
Both Wnt8 and Bmp2b are required at different timepoints for the maintenance of vent and vox
Two pathways are required for the maintenance of vent and vox expression in zebrafish: the zygotic BMP pathway(Melby et al., 2000; Imai et al., 2001) and the Wnt pathway (this work). To understand the combined regulation of ventand vox during gastrulation by the Wnt8 and BMP pathways, we analyzed the phenotype of wnt8;swr double mutants(Fig. 7). Using swr(bmp2b) mutants is sufficient to assess the influence of zygotic BMP signaling, as it was previously shown that loss of Bmp2b produces a zygotic bmp– null phenotype(Schmid et al., 2000). The requirement for both BMP and Wnt8 inputs towards vent and vox expression would be revealed if wnt8;swr double mutants exhibit a phenotype similar to the vent–;vox– phenotype. We found that gsc and chd are expressed in a broader domain around the mesodermal margin in shield stage wnt8;swr double mutants compared with either single mutant (Fig. 7B, compare with Fig. 1; data not shown), and thus they phenocopy vent;voxmutants (Fig. 7A). The same results were obtained when using the wnt8 deficiency or wnt8MO knockdown (Fig. 7G),confirming the specificity of the interaction.
As wnt8;swr double mutants display the same expanded organizer phenotype as vent;vox mutants at shield stage, we expected vent and vox mRNAs to be either absent or strongly reduced. We found both vent and mesodermal vox to be strongly reduced but not completely absent in shield stage wnt8;swr double mutants(Fig. 7E,J). Both ventand vox transcripts are not detectable in the mesoderm of later stage wnt8;swr double mutants (data not shown).
The fact that double mutants appear to be worse than wnt8 or swr single mutants suggests that Wnt8 and BMP function in parallel to regulate vent and vox. Consistent with this, bmp2bexpression in wnt8 mutants/morphants is close to wild type(Fig. 7M-O), and wnt8expression in swr mutants is normal at shield stage(Fig. 7K,L). Hence, both Wnt8 and Bmp2b are early regulators of vent and vox, but Wnt8 has a more prominent role until mid-gastrula stages.
To understand the DV phenotype of wnt8 mutants, we have analyzed the interaction of Wnt8, BMP, Vent and Vox. We found that the levels of both repressors are lower in wnt8– embryos at 40% epiboly when the expanded organizer phenotype initiates(Fig. 8). Consistent with a direct role for Wnt8 in vent/vox regulation, an inducible Lef/β-catenin fusion protein induces ectopic vent and vox transcription in the absence of new protein synthesis. Vent and Vox can repress organizer genes in the absence of Wnt8, suggesting that a simple linear pathway connects Wnt8/β-catenin with Vent/Vox-dependent organizer repression. In support of this, Wnt8 is unable to repress the organizer in the absence of Vent and Vox, although it is able to induce a Wnt reporter gene and to function in AP patterning. In addition, exogenous Wnt8 cannot repress gsc in vent;vox mutants. Finally, vent and vox regulation is under the control of both Wnt8 and zygotic BMP (Fig. 8),although Wnt8 is the primary regulator during early- to mid-gastrula stages.
vent and vox are transcriptional targets of Wnt8/β-catenin signaling
Although it is not known what induces vent and vox, our data show that Wnt8 regulates their early transcriptional maintenance. What is unclear is which Lef or Tcf proteins are involved in Wnt8-mediated transcriptional regulation. Studies in Xenopus suggest that Lef1 and not Tcf3 may mediate Xwnt8 function (Roel et al., 2002), but this has not yet been addressed in zebrafish.
Interestingly, it has recently been observed that overexpression of a conditional dominant repressor form of Tcf (hs-ΔTcf) leads to a more severe phenotype than the loss of Wnt8(Lewis et al., 2004). Lewis et al. found that gsc expression encircles the margin of transgenic hs-ΔTcf embryos heat-shocked at 4 HPF, a phenotype similar to vent;vox or wnt8;swr double mutants. Why would overexpression of a dominant-negative Tcf produce a more severe phenotype than loss of Wnt8 signaling? This could be explained if ΔTcf not only abolishes Wnt8 function but also prevents other factors from positively regulating vent and vox. One such factor could be the Smads that mediate Bmp2b function, as we have shown that zygotic BMP signaling is essential for maintaining vent and vox expression in the absence of Wnt8. In other words, ΔTcf may prevent Smad-dependent regulation of vent and vox.
Regulation of vent and vox by Wnt8: comparison between zebrafish and Xenopus
The transcriptional regulation of Xvent genes has been studied quite extensively in Xenopus, where most were found to be direct targets of Bmp4 signaling (Rastegar et al., 1999; Henningfeld et al.,2000; Henningfeld et al.,2002; Lee et al.,2002). However, the analysis of their regulation by Xwnt8 is less complete. It was found that zygotic Wnt signaling is necessary and sufficient for Xvent1 and Xvent2 expression(Hoppler and Moon, 1998; Marom et al., 1999), in agreement with our findings for zebrafish Wnt8. Analysis of Xenopusembryos overexpressing dominant-negative Xvent1 and Xvent2 revealed that Xwnt8 expression is not affected by the loss of Xvent activity(Onichtchouk et al., 1998). Again, our data agree as wnt8 is expressed in vent;vox mutants. The inability of Xwnt8 to rescue the dominant-negative Xvent phenotype was interpreted to mean that Xwnt8 functions in a different pathway than Bmp4/Xvent(Onichtchouk et al., 1998). However, we propose that, as in zebrafish, Xwnt8 functions upstream of Xvent genes, and that apparent differences between our model and Xenopus models may be due to the different experimental approaches. For example, concomitant reduction of Xwnt8, and Xvent1 and Xvent2, activities using dominant-negative proteins results in a more severe phenotype than reducing Xvent1 and Xvent2 alone(Onichtchouk et al., 1998). This is also what we observed when injecting vent and voxMOs in a wnt8– background. Thus, our results agree with data obtained in Xenopus, although our interpretation of the Wnt8/Vent/Vox relationship is somewhat different.
Wnt8 and zygotic BMP are required during gastrulation to maintain vent and vox expression at different timepoints
Our results show that both Wnt8 and Bmp2b (hence zygotic BMP) are required to maintain vent and vox levels during gastrulation, but that Wnt8 regulation of those genes occurs earlier at the blastula/gastrula transition (Fig. 8). The lack of an expanded organizer in swr mutants can be explained by the late regulation of vent and vox by zygotic BMP after the organizer has been formed. In addition, mesodermal vox levels are unchanged in swr mutants (only ectodermal vox levels are reduced at 70%) (Melby et al.,2000). Hence, mesodermal Vox can repress dorsal genes in swr mutants. Consistent with this, injection of vox MO in swr mutants results in expanded gsc expression at 70%epiboly (M.-C.R. and A.C.L., unpublished).
There are two known BMP signaling pathways in Xenopus and zebrafish (Dale and Jones,1999; Wilm and Solnica-Krezel,2003). In zebrafish, the maternal BMP pathway is thought to establish ventral identity in a manner analogous to the establishment of a dorsal axis by maternal β-catenin activity(Kramer et al., 2002; Sidi et al., 2003). Understanding the regulation of Wnt8 by maternal and zygotic BMP may explain apparently contradictory results from Xenopus and zebrafish. For instance, whereas it was found that regulation of zebrafish vent and vox by zygotic BMP occurs at mid to late gastrulation(Melby et al., 2000), Xenopus Xvent2 regulation by BMP signaling occurs during early gastrulation (stage 10.5) (Ladher et al.,1996). Xvent2 regulation was observed in embryos overexpressing a truncated Bmp2/4 receptor that does not distinguish between Bmp2 or Bmp4 ligands (Suzuki et al.,1994). However, Bmp2 is both maternally provided and zygotically expressed (Dale and Jones,1999). It has therefore been suggested that Xvent2expression may be under the influence of a maternal BMP signal(Ladher et al., 1996). Interestingly, the use of the same BMP-knockdown approach also results in decreased Xwnt8 expression(Schmidt et al., 1995; Hoppler and Moon, 1998). In zebrafish, it has been reported that loss of maternal BMP (Radar) signaling does not interfere with the induction of vent and vox at MBT(Sidi et al., 2003), although embryos homozygous for maternal smad5 display slightly expanded gsc and chd expression(Kramer et al., 2002). Thus,the elucidation of the relationship between Wnt8 and maternal or zygotic BMP in zebrafish using a loss-of-function approach may address whether the regulation of vent and vox is fundamentally different between zebrafish and Xenopus.
We thank Gerri Buckles for invaluable support, William Talbot for the Dfst7 line, Nobue Itasaki for GR-LEFΔN-βCTA,David Kimelman for the vent and vox constructs, Richard Dorsky for the TOPdGFPreporter line, and Bruce Riley and Bryan Phillips for discussion and critical reading of the manuscript. This work was supported in part by a Beginner Grant in Aid from the American Heart Association, Texas Affiliate (no. 0365081Y).