Ogon/Secreted Frizzled functions as a negative feedback regulator of Bmp signaling

One of the most important processes in the generation of vertebrate embryos is the formation of the dorsoventral (DV) and anteroposterior (AP) axes, in which cells at the appropriate positions adopt specific fates. During vertebrate embryogenesis, the dorsal organizer (which is also called Spemann's organizer in amphibian embryos, the node in mice and Hensen's node in chicks),plays a pivotal role in establishing these axes. Molecular identification of the dorsal organizer-derived inductive signals has revealed that Chordin and Noggin, generated from the dorsal organizer, bind to and prevent bone morphogenetic proteins (Bmps) from activating their receptors and thereby inhibit Bmp-dependent ventralization in Xenopus embryos(Piccolo et al., 1996; Zimmerman et al., 1996)(reviewed by De Robertis et al.,2000). Interactions between Bmps and the Bmp antagonists at the gastrula stage refine the DV axis in the mesoderm and endoderm, and are also involved in the formation of the neuroectoderm from the dorsal ectoderm. The interaction between Bmps and the Bmp antagonist Chordin is modulated by the metalloproteinase Tolloid (Blader et al.,1997; Piccolo et al.,1997), which cleaves and inactivates Chordin protein and Twisted Gastrulation (TSG), which can make a complex with Bmp and Chordin(Chang et al., 2001; Larrain et al., 2001; Oelgeschlager et al., 2000; Ross et al., 2001; Scott et al., 2001). In addition to Bmp inhibitors, the organizer-derived Wnt inhibitors (such as Dickkopf1, Frzb1 and Cerberus) and Nodal inhibitors (Lefty/Antivin and Cerberus) are thought to prevent the propagation of posteriorizing signals emanating from the mesoderm and endoderm that are involved in formation of the AP axis.

Roles for Bmp signaling in DV axis formation are also supported by analyses of zebrafish mutants that display abnormalities in DV patterning. The dorsalized mutants swirl (swr) and snailhouse(snh) have defective bmp2b and bmp7 genes,respectively (Dick et al.,2000; Kishimoto et al.,1997; Schmid et al.,2000). somitabun, captain hook and piggy tailall have defects in the zebrafish gene for Smad5 (madh5 –Zebrafish Information Network) (Hild et al., 1999; Kramer et al.,2002), which functions as a signal transducer for Bmp signaling. lost-a-fin encodes the gene for the type I Bmp receptor Alk8(Bauer et al., 2001; Mintzer et al., 2001), and mini fin encodes Tolloid (Connors et al., 1999). All of these data indicate that Bmp2 and Bmp7 and their signaling play essential roles in the formation of ventral tissues. In contrast to the dorsalized mutants, there are only two mutants, namely dino and ogon, that have been reported to display clearly ventralized phenotypes without other abnormalities in the early specification of the dorsal organizer (such as those seen in the bozozok mutants)(Hammerschmidt et al., 1996a; Solnica-Krezel et al., 1996). These ventralized mutant embryos display an expansion of ventral tissues, such as the ventral tail fin, posterior somatic mesoderm, blood and pronephron, and a reduction, to various degrees, of the anterior somites and the neuroectoderm. The dino (din) locus encodes the zebrafish ortholog of Chordin (Chordino; Chd – Zebrafish Information Network),whereas the molecular identity of the ogon locus has not yet been elucidated.

Complementation and mapping analyses revealed that ogon^m60, mercedes^tm305 and short tail^b180 are allelic, and thus commonly referred to as ogon (ogo)(Miller-Bertoglio et al.,1999). The ogo^m60 and ogo^b180 mutations are deficiencies in the distal part(close to the telomere) of linkage group 25 (LG25), indicating that the ogo locus is localized to the deleted region. The N-ethyl-N-nitrosourea (ENU)-induced allele ogo^tm305displays viable phenotypes, which are less severe than those of the ogo^m60 and ogo^b180 mutants, suggesting that ogo^tm305 is a hypomorphic allele or that ogo^m60 and ogo^b180 harbor the loss of additional gene(s) in the deletion. The ogo^m60 mutant embryo displays neural degeneration in addition to the ventralized phenotypes. The ventralized phenotypes of ogo are similar to those of chordino mutants, except that the reduction of the anterior neuroectoderm is less severe in the ogo than in the dinmutants. It has also been reported that a maternally derived ogo gene contributes to dorsoventral patterning(Miller-Bertoglio et al.,1999). The ventralized phenotypes of ogo are fully suppressed by the overexpression of the Bmp antagonists Chordin and Noggin, or the expression of a dominant-negative type II Bmp receptor(Miller-Bertoglio et al.,1999). Epistatic analyses revealed that swr/bmp2b and snh/bmp7 are epistatic to ogo in DV patterning(Miller-Bertoglio et al.,1999; Wagner and Mullins,2002). In ventral tail fin formation, lost-a-fin/alk8, is epistatic to ogo (Wagner and Mullins, 2002). All of these data consistently indicate that ogo encodes a dorsalizing factor that inhibits Bmp signaling either directly or indirectly. In contrast to chordino, ogo does not show an epistatic relationship with mini fin/tolloid(Wagner and Mullins, 2002),suggesting that ogo functions differently from chordin. It has been reported that elimination of both the zygotic ogo and chordin genes additively ventralizes the embryo, implying distinct requirements for these genes in DV axis formation(Hammerschmidt et al., 1996a; Miller-Bertoglio et al.,1999). The molecular identification of ogo is required to elucidate the precise relationship between ogo and chordin,and the molecular mechanisms by which ogo regulates the formation of the DV axis.

In this study, we isolated the gene responsible for the ogomutants by a positional cloning strategy and found that the ogo locus encodes a zebrafish homolog of Secreted Frizzled (Sizzled). The sizzled (szl) gene was originally identified in Xenopus based on its ability to dorsalize the Xenopus embryo(Salic et al., 1997). Szl displays sequence similarity with a Wnt receptor Frizzled and is reported to function as an inhibitor of Xenopus Wnt8 (XWnt8)(Salic et al., 1997). However, szl and a szl-related gene sizzled2 reportedly do not inhibit the activity of XWnt8, suggesting they have a different mode of action in DV axis formation (Bradley et al., 2000; Collavin and Kirschner, 2003). We found that Ogo/Szl functions to inhibit Bmp signaling in a manner that does not involve the inhibition of Wnt8-mediated signaling. Ogo/Szl requires Chordin protein for its dorsalizing activity. In contrast to other dorsalizing factors, ogo/szl is expressed on the ventral side and requires Bmp signaling. Our results suggest that Ogo/Szl functions as a negative-feedback regulator of Bmp signaling and provide a novel mechanism by which the DV axis is established during gastrulation.

ENU-treated AB male fish were mated with wild-type AB female fish to generate F1 female progeny. Haploid embryos were generated by in vitro fertilization using oocytes from the F1 female fish and ultra-violet ray-irradiated sperm. The haploid embryos were fixed at the early segmentation stages and analyzed by whole-mount in situ hybridization with the neural markers fez-like, engrailed3, krox20 and a neuronal marker deltaB. Two mutants (rk1 and rk2) displaying ventralized phenotypes (reduced neuroectoderm) were obtained by screening 753 F1 females. Complementation analyses using ogon^tm305 and dino^tt250 indicated that rk1 and rk2were allelic to ogon and dino, respectively. Maternal-zygotic ogon^rk1 embryos were generated by crossing homozygous and fertile male and female ogon^rk1fish. The genotyping of the swirl^tc300 and dino^tt250 mutants was done as described previously(Dick et al., 2000; Hild et al., 1999). To genotype the ogon^rk1 allele, the genomic fragment containing the splicing donor of the first intron of zebrafish sizzled gene was amplified by PCR using rk1-5′(5′-CCTCGATCTGACGACTTGAGGA-3′) and rk1-3′(5′-GCCAGTTCTAAATCATGAGCTACAC-3′), and digested with RsaI, which cleaves the PCR product from the wild-type but not from the ogon^rk1 allele.

ogon^rk1 heterozygous fish were mated with wild-type India fish to generate F1 families. Homozygous ogon^rk1mutant embryos were raised from the F1 cross and selected by morphological criteria (expanded ventral tissue). We used samples of their genomic DNA to carry out segregation analyses. We first examined the SSLP markers z1378,z8380, z23415 and z14408, which are located on the distal region of linkage group (LG) 25. This region is reported to be deleted in ogon^m60 and ogon^b180(Miller-Bertoglio et al.,1999). We found that the SSLP marker z8380 was close to the ogon locus (three recombinations out of 2998 meiotic segregations). Using the PCR primers for z8380 as a probe, we obtained BAC and PAC lines: BAC 185L03, BAC 31 and PAC 203B16. We isolated PAC 35K5, PAC 251M11 and BAC 173L18 using the end sequence of PAC 203B16, and we isolated PAC 259J12 using the end sequence of PAC 35K5. We isolated the fragments from the AB and India genomes,which correspond to the end of the PAC and BAC clones, and found polymorphic markers, SSLPs and STSs (sequence-tagged sites, detected by restriction fragment length polymorphisms and PCR). The precise information on the markers is available on request. ogo/szl cDNA was isolated by hybridizing a lambda Ziplox zebrafish gastrula cDNA library with the inserts of PAC 203B16,PAC 35K3 and a cosmid generated from PAC 203B16. A genomic fragment and cDNA fragment of ogon/sizzled were sequenced by performing shot-gun sequencing and reading the PCR products.

Capped RNAs for Noggin1, Dickkopf1, Chordin, Tlc (a constitutively active type I Bmp receptor), Xenopus Frzb-1, Xenopus Sizzled and Xenopus Crescent were generated as described previously(Furthauer et al., 1999; Hashimoto et al., 2000; Houart et al., 2002; Miller-Bertoglio et al., 1997; Nikaido et al., 1999; Pera and De Robertis, 2000; Salic et al., 1997). The coding region of zebrafish sizzled/ogon cDNA in pZL1 (from the lambda Ziplox clone) was excised and inserted into a modified pCS2+ (pCS2+SN). The sizzled^tm305 cDNA was isolated by PCR from the ogon^tm305 homozygous mutant embryos and inserted into pCS2+SN. The sizzled cDNA containing four mispaired nucleotides without amino acid changes was created by PCR and inserted into pCS2+SN. pCS2+SN sizzled and sizzled^tm305 were digested with AscI, and the capped RNA was transcribed with SP6 RNA polymerase. The antisense morpholino oligonucleotides used in this study were szl MO (5′-ACAGCAGCAGA-CTGAATAGAGACAT-3′) and control MO with four mispaired bases (5′-ACAGgAG-CAcACTGAtTAGAcACAT-3′). The chordin MO has previously been published(Nasevicius and Ekker,2000).

Whole-mount in situ hybridization was performed using BM purple (Roche) as the alkaline phosphatase substrate. The detection of six3.2, bmp2b,chordin, goosecoid, fused somites/tbx24 and eve1 was as described previously (Joly et al.,1993; Kobayashi et al.,1998; Miller-Bertoglio et al.,1997; Nikaido et al.,2002; Oxtoby and Jowett,1993; Stachel et al.,1993). The SalI-BamHI fragment of sizzled was subcloned into pZL1. The sizzled RNA probe was generated by digestion of pZL1-5′ sizzled and transcription with SP6 RNA polymerase. Photographs were taken using an AxioPlan2 microscope and AxioCam(Zeiss).

We isolated a novel ENU-induced allele of ogo, ogo^rk1,using a haploid screening procedure. The ogo locus was reported to be localized to the distal part of LG25(Miller-Bertoglio et al.,1999); therefore, we started the positional cloning of ogo based on this information. By segregation analysis using the ogo^rk1 allele and simple length polymorphism (SSLP)markers in the region of LG25, we mapped the ogo locus close to z8380(0.1 cM, Fig. 1A). Starting with z8380, we performed chromosomal walking with bacterial artificial chromosomes (BACs) and P1-derived artificial chromosomes (PACs)(Fig. 1A). By further segregation analyses using markers on the BACs and PACs, we found that two PAC clones, 203B16 and 35K5, contained the ogo locus. Using inserts from these PAC clones, we screened a gastrula cDNA library and obtained three cDNAs. One of these cDNAs encoded a protein of 282 amino acids, which displayed sequence similarity (52.3% identical at the amino acid level) with Xenopus Sizzled (Salic et al.,1997) (Fig. 1B). Owing to similarities with Xsizzled in its expression profile and function (described below), the isolated gene is most likely a zebrafish ortholog of szl. Like other sFrps, zebrafish Szl has a cysteine-rich domain (CRD) but no transmembrane domain. Sequence comparison revealed that Szl is distantly related to sFrp1-5 and relatively close to Crescent(Fig.1C).

To examine whether mutations in szl are responsible for the ogo phenotypes, we inhibited the function of Szl by injecting an antisense morpholino oligonucleotide (MO)(Fig. 2; Table 1). The embryos that received injections of the szl MO displayed ventralized phenotypes(Fig. 2B), including increases in the blood and ventral tail fin, that were similar to the typical phenotypes of the ogo mutant embryos. Injection of the control MO did not have any significant effects (Fig. 2D), and co-injection of a modified szl RNA containing four mispaired bases in the MO recognition site or Xenopus szl RNA suppressed the ventralized phenotypes caused by the szl MO(Fig. 2C; Table 1). Injection of the wild-type szl RNA rescued the ventralized phenotypes of the maternal-zygotic (MZ) ogo^rk1 in a dose-dependent manner(Fig. 2F,G; Table 2). We sequenced the genomic DNA and cDNA of szl isolated from the ogo^rk1 and ogo^tm305homozygous mutant embryos. We found a mutation in the splicing donor site of the first intron in the ogo^rk1 allele(Fig. 1D) that disrupted the correct splicing and resulted in truncation at the CRD (data not shown). We found several polymorphisms in the coding region of szl in the ogo^tm305 allele, one of which introduced an amino acid change at position 92 from aspartate, which is conserved between the zebrafish and Xenopus Szl, to asparagine(Fig. 1D). The mutation abrogated the function of Szl (described below). All of these data indicate that szl corresponds to ogo and that a mutation in the szl gene leads to ventralized phenotypes in zebrafish.

To examine the function of zebrafish ogo/szl, we misexpressed the ogo/szl RNA in wild-type embryos. The ogo/szl RNA-injected embryos displayed dorsalized phenotypes(Fig. 3B), similar to those of swr and snh (mutants with compromised Bmp signaling)(Dick et al., 2000; Kishimoto et al., 1997; Schmid et al., 2000), and of embryos overexpressing the Bmp antagonists Chordin and Noggin1(Furthauer et al., 1999; Miller-Bertoglio et al.,1997). The ogo/szl RNA-injected embryos did not show any change in chordin and bmp2b expression at the early gastrula stage (data not shown), but exhibited ventrally expanded chordin expression and reducedbmp2b expression at the mid-gastrula stage (Fig. 3D,G). Misexpression of ogo/szl did not affect the expression of goosecoid in the embryonic shield and the prechordal plate (Fig. 3I,K)or no tail expression (data not shown), indicating that Ogo/Szl could dorsalize the embryos without affecting the early specification of the mesoderm and dorsal organizer. Misexpression of ogo/szl RNA derived from ogo^tm305 neither dorsalized the embryo (data not shown) nor induced the ventral expansion of chordin expression(Fig. 3E), indicating that the mutation in the szl gene led to the loss of the dorsalizing activity of Szl.

Because most sFrps function as Wnt inhibitors and XSzl is reported to inhibit Wnt8 (Salic et al.,1997), we compared the activity of Ogo/Szl with that of known Wnt inhibitors and a Bmp inhibitor. Overexpression of ogo/szl as well as Xszl elicited dorsal expansion of the neuroectoderm, similar to the effect of noggin1 (nog)(Fig. 4A). In these embryos,expression of the forebrain-specific gene six3.2 and the mid-hindbrain marker pax2.1/no isthmus was expanded ventrally and formed a circular expression domain. By contrast, the overexpression of the Wnt inhibitor Dkk1 and the sFrp XCrescent reduced or abolished the pax2.1 expression and induced posterior but not ventral expansion of the six3.2 expression (Fig. 4A), indicating that these Wnt inhibitors had an anteriorizing activity on the neuroectoderm (Erter et al., 2001; Hashimoto et al.,2000; Lekven et al.,2001). Similarly, misexpression of XFrzb-1 and Tlc did not appear to dorsalize the zebrafish embryos (data not shown)(Houart et al., 2002). This is consistent with the involvement of Wnt8 and Wnt8 inhibitors in anterior-posterior (AP) patterning rather than in DV patterning(Erter et al., 2001; Lekven et al., 2001). These data indicate that Ogo/Szl functions similarly to the Bmp inhibitor but not the Wnt inhibitors, at least when these molecules are misexpressed. We next examined whether Szl inhibits the functions of wnt8, wnt5/pipetail or wnt11/silberblick, which are known to be expressed at the blastula and gastrula stages (Heisenberg et al.,2000; Kelly et al.,1995; Makita et al.,1998; Rauch et al.,1997). Overexpression of ogo/szl did not inhibit the wnt8-dependent ectopic expression of bozozok/dharma(Fig. 4B), a target of the Wnt canonical pathway (Ryu et al.,2001), but the overexpression of Xcrescent did. Furthermore, the overexpression of ogo/szl did not inhibit the wnt5-and wnt11-dependent inhibition of convergent extension(data not shown). This is consistent with the idea that Wnt5 and Wnt11-mediated signaling is not involved in DV patterning(Heisenberg et al., 2000; Rauch et al., 1997). All of these data indicate that Ogo/Szl does not inhibit the function of Wnt8 in zebrafish and that it functions differently from other sFrps and from the Wnt inhibitor Dkk1. We further examined whether Wnt inhibitors rescue the ventralized phenotypes of the szl MO-injected embryos. The szl MO-injected embryos displayed expansion of ventral ogo/szl expression at the late gastrula stage(Fig. 4H), as ogo/szlis expressed on the ventral side, in a Bmp signal-dependent manner (described below). Overexpression of Xcre or Dkk1 in the szl MO-injected embryos suppressed the expansion of ogo/szl expression (Fig. 4I,J),but concomitantly elicited the expansion of gsc expression (Fig. 4E,F),indicating that Xcre and Dkk1 can suppress the ventralized phenotypes caused by the loss of Ogo/Szl, but possibly through the expansion of the dorsal organizer, which produces Bmp inhibitors. However, it remains possible that Xcre and Dkk1 have a similar dorsalizing activity, which does not depend on Wnt8 inhibition, as Ogo/Szl does.

Maternally deposited ogo/szl transcripts were detected weakly by RT-PC) but not by in situ hybridization (data not shown). This is consistent with a hypothesized role of the maternal ogo gene(Miller-Bertoglio et al.,1999). Zygotic expression of ogo/szl was detected on the ventral side from the late blastula stage(Fig. 5). The ventral expression continued through gastrulation. At the segmentation stages, ogo/szl expression was specifically detected in the ventral tail fin,which is strongly affected in the ogo mutant embryos. The ventral expression of ogo/szl during the gastrula stages was severely diminished or absent in the swr/bmp2b mutant embryos (Fig. 6B,F). By contrast, ogo/szl expression was dorsally expanded in the chordino (din) mutant embryos (Fig. 6C,G)and in embryos that received an injection of RNA for a constitutively active Bmp receptor (Fig. 6D,H). These data indicate that ogo/szl expression is positively regulated by the Bmp signal during gastrulation.

Because Chordin is a dorsalizing factor that is indispensable for dorsalization (Oelgeschlager et al.,2003; Schulte-Merker et al.,1997), we examined whether Chordin is involved in the Ogo/Szl-dependent dorsalization and Bmp inhibition. We injected 200 pg of ogo/szl RNA into embryos obtained from crossing dinheterozygous fish; this treatment induced strong dorsalization in the wild-type and din heterozygous embryos (din/+; Fig. 7A). However,overexpression of ogo/szl in the din homozygous embryos(din/din) did not cause dorsalization, and the ogo/szl-overexpressing din/din embryos displayed strongly ventralized phenotypes, similar to the din mutant embryos, indicating that Chordin is required for the ogo/szl-dependent dorsalization. We next examined whether Ogo/Szl regulates the expression of chordin(Fig. 7B). We overexpressed Ogo/Szl in the din mutant and in chordin MO-injected embryos to exclude the involvement of autoregulated chordin expression(Fig. 7B). The loss of Chordin protein, either after injecting the chordin (din) MO or in the din mutant, led to the reduction or absence of the lateral expression of chordin, as reported previously(Miller-Bertoglio et al.,1997; Schulte-Merker et al.,1997). The misexpression of Noggin1, which binds directly to and inhibits the function of Bmp2/4, rescued and expanded the lateral expression in din mutant and din morphant embryos(Fig. 7B). By contrast, the misexpression of ogo/szl neither rescued the lateral expression nor affected the chordin expression at any developmental stage in the absence of the Chordin protein, indicating that ogo/szl does not regulate the expression of chordin in the absence of Chordin.

These data suggest that Ogo/Szl functions to modulate the activity of Chordin. To address this issue, we examined whether Ogo/Szl functions cooperatively with Chordin in DV patterning. We misexpressed ogo/szland chordin in din mutant embryos and examined the morphology and expression of fused somites (fss)/tbx24,which marks the paraxial mesoderm (Nikaido et al., 2002) (Fig. 7C; Table 3). fss expression was reduced in the din mutant embryos. Injection of a sub-optimal amount (15 pg) of chordin RNA slightly rescued or induced the ectopic expression of fss in the dinmutant embryos, but the total expression of fss in these embryos was less than in the wild-type and din heterozygous embryos. Co-injection of 15 pg chordin RNA with 200 pg ogo/szl RNA elicited a strong ventral expansion of fss in the din mutant embryos. Injection of 15 pg chordin RNA weakly rescued the ventralized phenotype of din at a low penetrance, but 200 pg of ogo/szldid not (Table 3). Co-injection of 15 pg chordin RNA with 200 pg ogo/szl RNA led to strong dorsalization regardless of the genotypes of the injected embryos. Because the misexpression of ogo/szl could not induce the expression of endogenous Chordin in the din mutant embryos, the data simply indicate that Ogo/Szl functions cooperatively with Chordin in DV patterning.

It has previously been reported that disruption of both ogo and chordin leads to more severely ventralized phenotypes than is seen in the single mutants (Hammerschmidt et al.,1996a; Miller-Bertoglio et al., 1999). This is inconsistent with the Chordin-dependency of the dorsalizing activity of Ogo. We generated ogo;din double mutant embryos by crossing fish bearing the ogo^rk1 and din^tt250 alleles, and examined their phenotypes and the expression of the ventral marker eve1 and the neuroectoderm marker krox20 (markers for rhombomeres 3 and 5). We also examined the genotypes of some of these embryos. The din^tt250 mutant embryos obtained from crossing ogo^rk1/+;din^tt250/+ parents displayed reduced anterior neuroectoderm (narrow stripes of krox20 expression),the accumulation of caudal cells, including blood cells, and showed variable phenotypes in the ventral tail fin, from reduction to expansion (Fig. 8C-E,G,L; Table 4), as reported previously (Miller-Bertoglio et al.,1999). We found that neither heterozygous nor homozygous mutations of ogo^rk1 in the din^tt250 mutant enhanced the morphologically ventralized phenotypes, nor did the mutation enhanced the expansion of eve1 expression or the reduction of krox20 expression in the din^tt250 mutant embryos. Taken together with the results of our misexpression studies in the din mutant embryos, we conclude that Chordin is required for the dorsalizing activity of Ogo/Szl.

The interaction between Bmps and Bmp antagonists plays an important role in DV patterning during early vertebrate embryogenesis. Many organizer-specific genes and/or Bmp antagonists have been molecularly identified; however, only chordin (chordino) and ogon have been shown genetically to be required for the early DV patterning in zebrafish(Hammerschmidt et al., 1996a; Schulte-Merker et al., 1997; Solnica-Krezel et al., 1996). This is likely to be due to redundant functions among Bmp inhibitors at the gastrula stages. By contrast, Chordin and Ogon would be expected to have non-redundant functions in the DV patterning. Genetic and phenotypic analyses of ogo mutants predict that the ogo locus encodes a dorsalizing factor that inhibits Bmp signaling. In this report, we demonstrated that a zebrafish ortholog of sizzled corresponds to ogo. First, the szl gene was located in the ogolocus (Fig. 1A). We found mutations in the szl gene for two alleles, ogo^rk1and ogo^tm305, that disrupted the dorsalizing activity of Szl (Figs 1, 3). The loss of function of Szl caused by injecting the szl MO phenocopied the ogo mutant embryos, and the injection of szl RNA suppressed the ventralized phenotype of the ogo mutants (Fig. 2). All these data indicate that the loss of function of szl leads to the ventralized phenotype observed for the ogomutant alleles.

The phenotypes of the ogo^m60 and ogo^b180 mutant embryos are more severe than the phenotype of the ogo^tm305 mutant embryos(Miller-Bertoglio et al.,1999). ogo^m60 and ogo^b180are deficiencies in the chromosome, and ogo^tm305 is suggested to be a hypomorphic allele. However, misexpression of large amounts of the szl gene from ogo^tm305 did not dorsalize the embryo (Fig. 3E),suggesting that ogo^tm305 is a functionally null allele. The loss of Szl protein following the injection of the szl MO led to ventralized phenotypes similar to those of the ogo^m60 and ogo^b180 mutants (Fig. 2), indicating that the loss of function of the single gene szl is solely responsible for the ventralized phenotypes of ogo. The embryonic lethality of ogo^m60 and ogo^b180 is probably due to the deletion of additional gene(s) in LG25.

A contribution from maternally derived ogo has been reported(Miller-Bertoglio et al.,1999). Consistent with this, we detected maternally deposited ogo/szl transcripts by RT-PCR (data not shown), and embryos receiving injections of large amounts of szl MO display phenotypes similar to those of maternal-zygotic (MZ) ogo mutant embryos(Miller-Bertoglio et al.,1999) (Fig. 2). However, as discussed below, the dorsalizing activity of ogo/szlrequires the presence of chordin, which is expressed after the mid-blastula transition (Miller-Bertoglio et al., 1997). Maternally provided Ogo/Szl should support the function of the zygotic Ogo/Szl in dorsalization.

Expression of ogo/szl was detected in the ventral blastoderm from the late blastula through the gastrula stages(Fig. 5). Ventral expression of ogo/szl at the gastrula stages strongly depended on Bmp signaling(Fig. 6). Zygotic Bmp signal-dependent expression has been reported for bmp2b/swr and bmp7/snh (Dick et al.,2000; Kishimoto et al.,1997; Schmid et al.,2000). The ventral expression of bmp2b and bmp7decreases after the mid-gastrula stage in the swr and snhmutant embryos, suggesting that the ventral expression of these genes depends only on Bmp signaling after the mid-gastrula stage. Similarly, the ventral ogo/szl expression was not affected at the late blastula stage in the swr and din mutant embryos (data not shown). These data indicate that ventral ogo/szl expression after the mid-gastrula stage depends on zygotic Bmp signaling. However, the ogo/szl expression was restricted to the ventral side from the time of its initiation at the late blastula stage. It has recently been shown that maternally provided smad5 is involved in the early specification of ventral tissue at the late blastula stage (Kramer et al.,2002). Ventral ogo/szl at the late blastula and early gastrula stages might be regulated by the maternally derived Bmp signal. Alternatively, the early ventral expression of ogo/szl might be regulated by the interaction between the dorsal-specific homeobox gene bozozok and ventrally expressed homeobox genes vox(previously vega1), vent (previously vega2) and ved, which play a role in early DV specification before the zygotic Bmp signaling occurs (Imai et al.,2001; Kawahara et al.,2000a; Kawahara et al.,2000b; Shimizu et al.,2002).

Compared with the expression of bmp2b, bmp4 and bmp7,ogo/szl expression was restricted to the more ventral blastoderm(Fig. 5). Similarly, in Xenopus embryos, the expression of szl and szl2 is confined to the ventral-most part of the ventral blastoderm(Bradley et al., 2000; Salic et al., 1997). The injection of increasing amounts of bmp4 RNA and the injection of chordin MOs in Xenopus embryos revealed that a high level of Bmp signaling is required for the expression of szl(Marom et al., 1999; Oelgeschlager et al., 2003). Thus, the ventral expression of sizzled is regulated by a mechanism that is conserved between zebrafish and Xenopus, and a high level of Bmp signaling activity, which exists in the ventralmost part of the blastoderm, is required for the expression of sizzled on the ventral side. Promoter analyses of ogo/szl will clarify this issue.

Ogo/Szl has sequence similarity with the Wnt receptor Frizzled, suggesting a role for Ogo/Szl in Wnt inhibition. However, the overexpression of Ogo/Szl did not inhibit the Wnt8-dependent ectopic expression of bozozok(Fig. 4). Misexpression of the Wnt8 inhibitor Dkk1 and the sFrp Crescent anteriorized the neuroectoderm, but did not dorsalize the embryo efficiently, whereas misexpression of Ogo/Szl dorsalized the embryo but did not anteriorize the neuroectoderm, unlike the Bmp inhibitor Noggin 1 (Fig. 4). These data suggest that Ogo/Szl functions as a Bmp inhibitor rather than as a Wnt inhibitor.

We found that overexpression of Crescent and Dkk1 suppressed the ventralized phenotypes of szl MO-injected embryos(Fig. 4). However, Crescent and Dkk1 also elicited expansion of the dorsal organizer(Fig. 4)(Hashimoto et al., 2000),which produces the Bmp inhibitors Chordin and Noggin 1. Thus, Crescent and Dkk1 might substitute for the function of Ogo/Szl indirectly by expanding the expression domain of the organizer-derived Bmp inhibitors. It remains possible that Crescent and Dkk1 (less likely) might have a dorsalizing activity other than expansion of the dorsal organizer, just as Ogo/Szl does. However, Ogo/Szl did not inhibit the Wnt8 activity; therefore, the dorsalizing activity of Ogo/Szl should not depend on Wnt8 inhibition.

In zebrafish, the loss of wnt8 or of tcf3/headless, which functions to inhibit Wnt8 signaling, strikingly affects the AP patterning in the neuroectoderm in addition to causing abnormalities in the DV patterning(Erter et al., 2001; Kim et al., 2000; Lekven et al., 2001). The phenotypes of the ogo mutant and the ogo/szl-overexpressing embryos were different from those of embryos with high (e.g. headlessmutant embryos) and low (wnt8 morphant embryos) Wnt8 activities,respectively, further supporting the idea that Ogon does not function to inhibit Wnt8 signaling. In addition to wnt8, there are several Wnt genes reported to be expressed at the blastula and gastrula stages in zebrafish. Among them, wnt5/pipetail and wnt11/silberblick,which activate a non-canonical Wnt signal, are known to be involved in convergent-extension movements during gastrulation(Heisenberg et al., 2000; Rauch et al., 1997), and thus it is unlikely that the dorsalizing activity of Ogo/Szl is due to the inhibition of Wnt5 and Wnt11. Consistent with this, misexpression of ogo/szl did not affect the wnt5- or wnt11-mediated inhibition of convergent extension (data not shown). Similarly, it was reported that Xenopus szl2 does not inhibit the activities of Xenopus Wnt3a, Wnt5a and Wnt8(Bradley et al., 2000). All of these data indicate that Ogon/Szl and Xenopus Szls promote dorsalization through interactions with factors other than the Wnts.

Ogo/Sizzled requires the Chordin protein to dorsalize embryos(Fig. 7), indicating that Ogo/Szl displays a mode of action that is completely different from that of other Bmp antagonists. This finding is consistent with a previous report that misexpression of Xenopus sizzled cannot rescue UV-treated ventralized embryos, which should not express chordin(Salic et al., 1997).

How does Ogo/Szl inhibit Bmp signaling in a Chordin-dependent manner? Our data and the data previously published imply the mode of function of Ogo/Szl:(1) din and ogo mutant embryos have similar ventralized phenotypes (Hammerschmidt et al.,1996b; Miller-Bertoglio et al., 1999; Solnica-Krezel et al., 1996); (2) the ventralized phenotypes of ogo can be suppressed by the expression of Chordin, Noggin and a dominant-negative Bmp receptor (Miller-Bertoglio et al.,1999), but the din phenotypes cannot be suppressed by misexpression of ogo/szl (Fig. 7); (3) overexpression of ogo/szl did not inhibit Wnt8 activity (Fig. 4); (4)overexpression of ogo/szl induced a similar phenotype to that of noggin1 but not that of dkk1 or crescent(Fig. 4); (5) low levels of chordin could act synergistically with ogo/szl in dorsalization (Fig. 7); (6) a mutation in ogo did not enhance the ventralized phenotypes of din embryos (Fig. 8);and (7) loss of tolloid/mini fin can suppress the ogon tail phenotype (Wagner and Mullins,2002). All of these results indicate that Ogo/Szl can augment the activity of Chordin, by inhibiting an inhibitor of Chordin, by directly making Chordin more active, or by modulating the Bmp signal so that it becomes more susceptible to the Chordin-mediated inhibition. The dorsalizing activity of the Chordin protein is regulated by different mechanisms: the chordin protein level is regulated through processing by Tolloid-related metalloproteinases,and Chordin interacts physically and functionally with Bmp and Twisted Gastrulation (Tsg) to modulate Bmp activity(De Robertis et al., 2000). Tolloid-related proteins and Tsg might be involved in the function of Ogo/Szl. Alternatively, Ogo/Szl may function in parallel with Chordin. Both Ogo/Szl and Chordin are required for the formation of posterior dorsal tissues, and the loss of either Ogo/Szl or Chordin might lead to ventralization. In this scenario, the lowering of the Bmp signal by Chordin might work cooperatively with Ogo/Szl to dorsalize the embryo.

A mutation in the cysteine-rich domain (CRD) of the Ogo/Szl in ogo^tm305 implies an essential role for the CRD in the activity of Ogo/Szl. As the CRD of Frizzled is known to interact with Wingless and Wnts (Bhanot et al., 1996),it is still possible that Ogo/Szl functions by inhibiting unidentified Wnt(s). The identification of proteins that interact with Ogo/Szl will shed light on the mechanisms by which Ogo/Szl inhibits Bmp signaling and regulates the specification of the DV axis.

Many feedback inhibitors play roles in the early patterning of vertebrate embryogenesis (Freeman, 2000). The Antivin/Lefties (Lefy1 and Lefy2) function as feedback inhibitors for the Nodal-related molecules in zebrafish, Xenopus and mice(Bisgrove et al., 1999; Cheng et al., 2000; Meno et al., 1999; Thisse and Thisse, 1999). Sprouty4 and Sef function in FGF signaling(Furthauer et al., 2002; Furthauer et al., 2001; Tsang et al., 2002). An inhibitory Smad, Smad7 and Bambi/Nma function as feedback inhibitors of Bmp signaling (Grotewold et al.,2001; Nakayama et al.,1998; Onichtchouk et al.,1999; Souchelnytskyi et al.,1998). ogo/szl is regulated positively by Bmp signaling and in turn Ogo/Szl inhibits Bmp signaling, indicating that Ogon/Szl functions as a negative-feedback regulator of Bmp signaling. In contrast to the other feedback inhibitors described above, Ogo/Szl function requires the Chordin protein. ogo/szl is expressed on the ventral side of the embryo in a Bmp-signal-dependent manner, whereas chordin is expressed on the dorsal side and is negatively regulated by Bmp signaling(Hammerschmidt et al., 1996b). Thus, ogo/szl and chordin are regulated in completely opposite manners, but cooperate in inhibiting the Bmp signal. The Ogo/Szl and Chordin proteins that are diffused from the ventral and dorsal sides might be colocalized at a specific position along the DV axis and function there cooperatively to inhibit the Bmp signal. In support of this idea, Chordin and Szl2 appear to diffuse a long distance from their source(Bradley et al., 2000; Jones and Smith, 1998). Alternatively, molecules that function downstream of the Ogo/Szl-mediated signaling might interact with Chordin or unknown regulator(s) of Chordin to inhibit Bmp signaling. In any case, the functional interaction between Ogo/Szl and Chordin provides a precise positional cue to cells along the DV axis during gastrulation.

We thank L. Solnica-Krezel, R. Moon, E. M. De Robertis, M. Kirschner, N. Ueno, M. Halpern, M. Kobayashi, Y. Imai, T. Tada, K. Araki, C. Houart, S. Wilson, B. Thisse and C. Thisse for providing the DNA samples and fish; S. Schulte-Merker and L. Collavin for sharing unpublished results; R. Ladher for critical reading of the manuscript; C. Fukae and Y. Kuga for fish care; H. Tarui and K. Agata for genomic sequencing; and S. Iwaki for secretarial assistance. Y.-K.B. and T. Hirata are Fellows of the Japan Society for the Promotion of Science. This work was supported by Grant-in-Aids for Scientific Research from the Ministry of Education, Science, Sports and Technology(KAKENHI 13138204), and from JSPS (KAKENHI 13680805); and by a grant from RIKEN.