Formation of the gastrula organizer requires suppression of ventralizing signals and, in fish and frog, the need to counteract the effect of ubiquitously present maternal factors that activate the expression of Bmps. How the balance between dorsalizing and ventralizing factors is shifted towards organizer establishment at late blastula stages is not well understood. Mutations in zebrafish bozozok (boz) cause severe defects in axial mesoderm and anterior neurectoderm and affect organizer formation. The boz gene encodes the homeodomain protein Bozozok/Dharma and its expression in the region of the organizer is activated through β-catenin signaling. Here, we investigate the molecular mechanism by which boz contributes to the establishment of the organizer. We demonstrate that the homeodomain protein Boz acts as a transcriptional repressor in zebrafish: overexpression of an En-Boz fusion protein can rescue the boz phenotype, whereas a VP16-Boz fusion protein acts as an antimorph. Expression analysis of bmp2b indicates that Boz negatively regulates bmp2b in the prospective organizer. We demonstrate that this Boz activity is independent of that of other zygotic genes, because it also occurs when translation of zygotic genes is suppressed by cycloheximide(CHX). We identify two high-affinity binding sites for Boz within the first intron of the bmp2b gene. Deletion of these control elements abolishes Boz-dependent repression of bmp2b in the early blastula. Thus, Boz directly represses bmp2b by binding to control elements in the bmp2b locus. We propose that early transcriptional repression of bmp2b by Boz is one of the first steps toward formation of a stable organizer, whereas the later-acting Bmp antagonists (e.g. Chordin, Noggin)modulate Bmp activity in the gastrula to induce patterning along the dorsoventral axis. Thus, similar to Drosophila Dpp, asymmetry of Bmp expression in zebrafish is initiated at the transcriptional level, and the shape of the gradient and its function as a morphogen are later modulated by post-transcriptional mechanisms.

The inhibition of Bmp/Dpp signals by Chordin/Sog and other antagonists is central to dorsoventral patterning mechanisms in animal development(Holley et al., 1995; Sasai et al., 1995; Marqués et al., 1997). In zebrafish, mutations in bmp2b (Swirl)(Kishimoto et al., 1997; Nguyen et al., 1998), bmp7 (Snailhouse) (Dick et al., 2000; Schmid et al.,2000) and smad5 (Somitabun)(Hild et al., 1999; Kramer et al., 2002) result in strong dorsalization, whereas mutations in chordin (chd)(Schulte-Merker et al., 1997)produce ventralized phenotypes. Thus, in zebrafish, Bmp2b and Bmp7 cooperate and ventralize the embryo by activating transforming growth factor β(TGFβ) family receptors and signaling via downstream Smads. Whereas Dpp expression in Drosophila is initiated in a spatially restricted dorsal domain, Bmp2/4 family proteins in vertebrates are initially expressed in a fairly ubiquitous manner throughout the blastoderm, as seen in Xenopus (bmp4)(Hemmati-Brivanlou and Thomsen,1995) and zebrafish (bmp2b and bmp7)(Martinez-Barbera et al.,1997; Nikaido et al.,1997; Dick et al.,2000; Kramer et al.,2002). In Drosophila, the dorsally restricted expression of Dpp is established by transcriptional repression of Dppby Dorsal on the ventral side of the embryo. In vertebrates, the mechanisms for the initial transcriptional repression of Bmp genes on the dorsal side of the embryo are not well understood. In late blastulae and early gastrulae, Bmp gene expression is limited dorsally by organizer-derived secreted factors such as Chordin and Noggin, which inhibit Bmp signaling and block the feedback loop by which Bmp maintains its own expression(Onichtchouk et al., 1996; Kishimoto et al., 1997). In Xenopus, chd expression is directly repressed by the homeodomain-protein Vox, which is produced under the control of Bmp4(Melby et al., 1999). Because Bmp signaling efficiently represses chd expression, it has remained unclear how a dorsal organizer is initiated at blastula stages in the presence of high levels of Bmp activity.

Mutations in bozozok (boz) affect proper establishment of the zebrafish gastrula organizer. boz mutant embryos lack axial mesoderm and have severe patterning defects in the anterior neuroectoderm(Solnica-Krezel et al., 1996; Solnica-Krezel and Driever,2001). boz encodes a homeodomain protein also known as Dharma and Nieuwkoid (Koos and Ho,1998; Yamanaka et al.,1998; Fekany et al.,1999). It is transcribed in dorsal blastomeres and in the dorsal yolk syncytial layer from the earliest zygotic gene expression until early gastrula (Yamanaka et al.,1998). Hyperdorsalization of zebrafish embryos by incubation in lithium chloride, which activates β-catenin signaling, results in expression of boz around the margin(Yamanaka et al., 1998). Active components of β-catenin signaling are required for bozexpression (Shimizu et al.,2000), which is mediated through TCF/LEF binding sites in the boz promoter (Ryu et al.,2001). The function of boz is required for expression of gsc and many other organizer-specific genes(Fekany et al., 1999). Thus, boz might be directly activated by Wnt/β-catenin signaling and might directly mediate some activities of the Nieuwkoop center to establish the organizer. boz encodes a homeodomain protein and has been suggested to be involved in regulating expression of zygotic patterning genes acting both in the Nieuwkoop center and in the gastrula organizer. Recent studies indicate that boz is required for downregulation of bmp2b expression on the dorsal side(Koos and Ho, 1999) and it has also been suggested to interfere with Wnt signaling(Fekany-Lee et al., 2000) and Nodal signaling (Shimizu et al.,2000). However, the mechanism by which boz exerts its function at the crossroads of anterioposterior and dorsoventral patterning is not yet understood.

In this study, we investigate the role of boz in establishing dorsoventral polarity at the molecular level. We show that Boz acts as a transcriptional repressor, because fusion proteins containing the Boz homeodomain and the Drosophila Engrailed repressor domain can rescue the boz mutant phenotype. By contrast, a fusion of Boz with the VP16 transcriptional activator domain acts as a Boz antimorph, phenocopying the boz mutant phenotype and ventralizing the embryo. Repression of bmp2b does not require translation of zygotic gene products and thus must be a direct effect of Boz. We identify two high-affinity binding sites for Boz in the first intron of bmp2b and demonstrate by deletion analysis that the Boz-binding sites mediate bmp2b repression. We suggest a molecular pathway for initiation of dorsoventral asymmetry of bmp2b expression in which β-catenin signaling activates boz, which, in turn, represses bmp2b transcription in the prospective organizer.

Genetic strains and phenotypic analysis

The zebrafish bozm168 allele(Solnica-Krezel et al., 1996)was crossed from its AB/Tü strain background into an India strain wild-type background, in which the boz phenotype shows higher penetrance and expressivity. Staging was performed according to Kimmel and colleagues (Kimmel et al.,1995). In situ hybridization was performed as described by Hauptmann (Hauptmann, 1999). Phenotypes of boz mutant embryos were classified according to Fekany et al. (Fekany et al., 1999)(Table 1).

Constructs and overexpression of mRNA in fish embryos

boz full-length cDNA was amplified by RT-PCR from 30% epiboly stage embryos, using primers 5′-ATACTCACGCAGCTTTTGGG and 5′-CAAAATGTTGGCATTATTCTGTC (Yamanaka et al., 1998) and subcloned into the pCS2+ expression vector. The En-boz construct was generated by recombinant PCR to fuse the Engrailed repressor domain (amino acids 1-296) to the homeodomain of Boz(amino acids 105-192). The VP16-boz construct was generated by recombinant PCR to fuse the VP16 activation domain (C-terminal 72 amino acids)to the homeodomain of Boz (amino acids 105-192). Boz(δ-eh1) was generated by recombinant PCR to delete the eh1 motif (amino acids 9-15). Boz(HD) was generated by subcloning the C-terminal amino acids 105-192 with the homeodomain into pCS2+. Capped sense RNA was synthesized using SP6 RNA polymerase and the mMESSAGE mMACHINE system (Ambion) after NotI digestion in pCS2+.

For CHX treatment (modified from Gard et al., 1990), embryos were injected with mRNA at the early one-cell stage, subsequently dechorionated using pronaseE (Sigma, 1 mg ml-1) and incubated in 0.3× Danieau's medium. At 1.5 hours post-fertilization (hpf), the medium was replaced with 0.3× Danieau's medium containing 10 μg ml-1 CHX (Calbiochem, diluted from 100 mg ml-1 stock in ethanol). Embryos were maintained in CHX until control embryos had reached sphere stage and then fixed for whole-mount in situ hybridization.

bmp2b regulatory region

Genomic PAC clones containing bmp2b were isolated by filter hybridization of a zebrafish genomic library (RZPD, Berlin) with a bmp2b cDNA probe. The 3.5 kb upstream of the bmp2b coding region, including the first intron, were sequenced by primer walking. The transcription start site was determined by primer extension analysis (MMLV reverse transcriptase, Ambion. bmp2b primer:5′-TTCCCGTCGTCTCCTAAGTTC-3′).

Electrophoretic mobility shift assay

The Boz bacterial expression construct was generated by subcloning the boz full-length cDNA into the 6×His-tagged bacterial expression vector pET15b (Novagen), followed by transformation into BL21(DE3pLysS) strain bacteria. Bacterial recombinant Boz protein was induced by IPTG at 25°C for 4 hours and then isolated and purified over a Talon (Clontech) affinity column according to the manufacturer's instructions.

For each of the double-stranded oligonucleotides used, one of the two strands was 32P end-labeled by the T4 kinase reaction or by Klenow fill-in reaction. Probe (10,000 cpm per 10 fmoles) was incubated with 0.1μg of purified Boz protein in the presence of 0.76 μg of poly d(I-C) in binding buffer (20 mM HEPES, 50 mM EDTA, 5 mM MgCl2, 5% glycerol, 1 mM DTT, pH 8) in a total volume of 10 μl as described(Brannon et al., 1997). Samples were incubated on ice for 20 minutes, followed by a further 20-minute incubation with the radiolabeled probe at room temperature. Electrophoresis was performed in 4% polyacrylamide gel with 0.25× TBE buffer at room temperature.

For the analysis of the bmp2b promoter, 3.5 kb of upstream region was sequenced and seven 32P end-labeled DNA fragments (F1, -807 to-445; F2, -466 to -124; F3, -151 to +214; F4, +280 to +901; F5, +880 to +1494;F6, +1470 to +1741; F7, +1723 to +2343) were generated by PCR to cover almost the entire region. The fragments F7 and F4 were each digested by two separate sets of enzymes to generate overlapping sets of smaller fragments: (1) AluI (for F7) or MboI and Sau3AI (for F4), followed by radiolabeling with Klenow; (2) MseI digestion, ligation of MseI adapters to the fragments and amplification of the ligated fragments by PCR using end-labeled MseI adapter oligonucleotides. The radiolabeled MseI PCR fragments were confirmed by electrophoretic mobility shift assay (data not shown). Double-stranded oligonucleotides were synthesized to verify binding sites: B1 (+1980 to +2033:5′-GGTAAGTAAAATAATCTTATTTCAAACTAAAAGCAAGATTATTTTACTCACCAA),B2 (+1470 to +1494: 5′-AAAGCAAGATTAGTTTACTGGCTTG), B3 (+537 to +565: 5′-GCGTGCCTGCATGTAATGTGTGAGGTCAG), B4 (+430 to +467:5′-GCATTCAATTACGTGCTTGATATTACGTATTAGCAAAC),B5 (+343 to +369: 5′-TCGCTTGTGGATTAAAACACGAATTCA), B6 (-151 to -124: 5′-GTATTGCGTATACATTACATTCTCGTTC) and B7 (-786 to-759: 5′-GATGTAAAGTCAGAATTATTAGCCCCCT). Bicoid binding site double-stranded oligonucleotide:(5′-GTCACCTCTGCCCATCTAATCCCTTGACGC) from the hunchback gene (Driever et al.,1989). Underlined nucleotides indicate potential Bicoid-type homeodomain binding sites.

Luciferase assay of promoter activity

The Bmp2b-pLUC promoter fusion (bmp2b -806 to +2538) was generated by subcloning the bmp2b fragment at the BamHI and HindIII sites into the pLUC vector (a luciferase derivative of pBLCAT6 provided by J. Altschmied, Würzburg)(Boshart et al., 1992). Promoter fusions that have the binding sites B1 and /or B4 deleted were constructed by PCR-based subcloning. Bmpδ1pLUC has a deletion (basepairs+1983 to +2030) spanning site B1, Bmpδ4pLUC has a deletion (basepairs+432 to +490) spanning site B4 and Bmpδ1,4pLUC has both of these sites deleted but still contains the sequence +491 to +1982 of the first intron.

Embryos were harvested within 5 minutes of fertilization and the chorions were removed with pronase E (Sigma, 1 mg/ml). The embryos were co-injected with linearized pLUC reporter DNA derivatives and mRNA during early one-cell stage. The embryos were then incubated at 28.5°C for 5 hours. Ten microfuge tubes of ten embryos were collected for each experimental condition,the supernatant was aspirated and the embryos were frozen at -80°C immediately after addition of 80 μl lysis buffer (0.1 M sodium phosphate buffer pH 7.3, 1 mM DTT, 0.2% Triton-X100, 0.2 mM AEBSF[4-(2-aminoethyl)benzenesulfonylfluoride hydrochloride; Sigma]. For the luciferase assays, embryos were thawed and dissolved by repeated pipetting,and the lysate cleared by centrifuging 10 minutes at 16,000 g(4°C). 60 μl of supernatant were transferred to 96-well microtiter plates on ice. 350 μl of reaction buffer (22.5 mM glycylglycin pH 7.8, 2 mM ATP, 10 mM MgSO4) were mixed (vortex) with 50 μl embryo extract in luminometer tubes and the reaction was started by injection of 0.2 mM D-luciferin in 20 mM glyclglycin pH 7.8. Measurements were recorded using a Lumat LB9501 (Berthold). Each experiment (ten measurements of ten embryos each) was performed at least in triplicate and the standard deviation was calculated.

Boz acts as a transcriptional repressor during zebrafish organizer formation

The severe dorsoanterior patterning defects in boz mutant embryos(Solnica-Krezel et al., 1996; Fekany et al., 1999) could be caused by lack of activation of dorsoanterior genes, by lack of repression of ventralizing genes or by a combination of both. We assayed Boz for its potential effects on transcription in two heterologous systems, yeast and Drosophila (see supplemental data S1 and S2). In both assay systems, Boz or Boz fusion proteins did not appear to be activators of target genes but rather affected target gene expression as repressors or as competitors of endogenous activators.

To determine whether Boz functions as a transcriptional repressor in zebrafish, we analysed the effects of repressor and activator fusion constructs on embryonic axis formation. We created a fusion of the Engrailed(En) repressor domain (amino acids 1-296)(Han and Manley, 1993) and the homeodomain of Boz (amino acids 105-192)(Fig. 1). Overexpression by microinjection of 2 pg En-boz synthetic mRNA at the one-cell stage induced dorsalization (Fig. 1A-D), which, at 24 hpf, is similar to that resulting from overexpression of wild-type boz(Yamanaka et al., 1998). Injection of 0.05-0.25 pg of En-boz mRNA into progeny of bozheterozygous parents provides some phenotypic rescue, shifting the distribution of boz phenotypic classes from severe class I and II mutant phenotypes to the less severe classes IV and V(Table 1). We observed a reduction in the penetrance of the mutant phenotype, indicating complete rescue in some cases. In addition, at the higher concentration, a large proportion of embryos become dorsalized, resembling snailhouse or piggytail phenotypes (Mullins et al., 1996). This dorsalizing activity occurs during early patterning. Overexpression of either En-boz(Fig. 2E,F,M,N) or bozmRNA (Fig. 2C,D,K,L) induced ectopic chd and gsc expression in 50% epiboly stage wild-type embryos. Thus, the En-Boz fusion protein has activities similar to wild-type Boz: overexpression of either Enboz or boz mRNA can rescue boz mutant phenotypes and hyperdorsalize wild-type embryos.

The N-terminus of Boz contains an eh1 motif (FSIDYIL; see supplemental data S1)that is conserved among genes homologous to Engrailed(Koos and Ho, 1999). The eh1 motif is also present in other classes of homeodomain proteins that have been shown to act as repressors, such as Gsc(Ferreiro et al., 1998), Hex(Ho et al., 1999), Xvent1(Friedle et al., 1998; Rastegar et al., 1999), Vox(Melby et al., 1999), Xanf1(Ermakova et al., 1999) and Xvent2 (Trindade et al.,1999). To investigate the function of this motif we created a Boz variant with the eh1 motif deleted, Boz(δ-eh1). When injected into one-cell stage embryos, boz(δ-eh1) mRNA lacks dorsalizing activity, even when as much as 25 pg mRNA were injected (data not shown). Thus, the eh1 motif is essential for the dorsalizing activity of Boz in zebrafish embryos. A Boz derivative lacking the homeodomain,Boz(δ-HD), has been reported to show no biological activity in zebrafish embryos (Koos and Ho, 1999). We overexpressed the Boz homeodomain region alone and could not induce abnormal development (amino acids 105-192; up to 25 pg mRNA injected; data not shown). Thus, the endogenous activity of Boz might be exclusively mediated through the eh-1 domain and the DNA binding homeobox domain.

VP16-Boz fusion protein functions as a Boz antimorph

We created the transcriptional activator fusion construct VP16-bozby fusing the VP16 activator domain (C-terminal 72 amino acids; Sadowski et al., 1988) to the Boz homeodomain (amino acids 105-192; Fig. 1). Injection of 5 pg VP16-boz mRNA into one-cell-stage embryos produced the full range of boz mutant phenotypes(Fig. 1G,H). Increasing amounts of injected VP16-boz mRNA resulted in a gradual increase in the incidence of severe boz mutant phenotypes(Table 2). In addition, some embryos were ventralized, showing an expansion of ventral fates such as the blood-forming region and enlarged ventral tail fins. Consistent with the morphological defects, expression of the organizer specific genes gscand chd was reduced in VP16-boz injected wild-type embryos at 50% epiboly (Fig. 2G,H,O,P). Taken together, these data suggest that the VP16-Boz fusion protein acts like a boz antimorph, induces ventralization and phenocopies the boz mutant phenotype. As VP16 fusion proteins cause increased levels of transcription of target genes, our finding of diminished expression of chd and gsc indicates that VP16-Boz only indirectly affects these genes. We postulate that VP16-Boz activates expression of ventral factors, which in turn repress chd and gsctranscription.

Regulation of bmp2b expression by boz

boz RNA can dorsalize zebrafish embryos, and so we determined the effects of boz, En-boz and VP16-boz on bmp2bexpression. Wild-type gastrula stage embryos express a gradient of bmp2b mRNA, with the highest concentration at the ventral side(Fig. 3A,B). Injections of 2 pg of synthetic boz or En-boz mRNA into one-cell-stage wild-type embryos almost completely abolished bmp2b expression at 50%epiboly (Fig. 3C-F). To determine whether boz influences initiation of bmp2bexpression, we analysed embryos at sphere stage, when zygotic bmp2bexpression begins. Overexpression of boz caused a dramatic reduction of bmp2b expression at sphere stage(Fig. 3J,K) but did not change the expression level of chd (Fig. 3I). These findings indicate that Boz might directly repress bmp2b but does not directly affect the expression of chd. Following overexpression of boz, we could not detect any changes in the expression of bmp7 at sphere stage (data not shown). If boz directly represses bmp2b then VP16-boz might activate it. Indeed, we found that overexpression of VP16-boz clearly enhances bmp2b expression by sphere stage(Fig. 3G,H).

To further analyse the function of boz in bmp2bregulation during early development, we compared bmp2b expression at sphere stage in wild-type and boz mutant embryos. In wild-type embryos, bmp2b expression is ubiquitously initiated in the blastoderm, except for a small domain at the dorsal blastoderm margin(Fig. 3L,N). The early dorsal repression does not depend on chd function(Hild et al., 1999). In boz mutants, the repression of bmp2b expression in this dorsal domain is abolished (Fig. 3M,O) (see also Koos and Ho,1999). This early bmp2b expression phenotype is found in all boz mutant embryos, although they later exhibit highly variable morphological phenotypes (classes I-V; Table 1)(Fekany et al., 1999). In crosses between heterozygous boz parents, one-quarter of the progeny lack repression of bmp2b at the dorsal margin between sphere stage and 30% epiboly. Taken together, these findings suggest that bozfunctions to repress bmp2b during the initial phase of its expression from sphere stage to 30% epiboly.

The complexity of dorsoventral patterning interactions could still make the interaction between boz and bmp2b an indirect one, by Boz repressing an activator of zygotic bmp2b transcription. To investigate whether other early zygotic genes contribute to the effect of Boz on bmp2b transcription, we inhibited translation of zygotically expressed mRNAs with CHX by incubating embryos from 1.5 hpf onwards in 10μg ml-1 CHX. In this experiment, maternal and injected mRNAs are translated during the first 90 minutes before CHX is added but zygotic transcripts generated after the mid-blastula transition are not translated in the presence of CHX. At this CHX concentration, we find that gscexpression is not detectable and thus depends on the presence of other zygotic gene products (Fig. 4B,H). bmp2b is not affected by CHX and is thus activated by maternal factors (Fig. 4A,G). Repression of bmp2b transcription by overexpression of boz is not affected by CHX (Fig. 4C,I). Similarly, activation of bmp2b expression by overexpression of boz-VP16 also occurs in the presence of CHX(Fig. 4E,K). This demonstrates that zygotic gene products are not required to mediate the repressive effect of Boz on bmp2b transcription. Further, our data on gscexpression (Fig. 4D,J,F,L)demonstrate that the effect of Boz on gsc requires the expression of zygotic proteins.

Boz binds to regulatory elements in bmp2b

To determine whether boz affects bmp2b expression by directly binding to the bmp2b regulatory region, we characterized the genomic organization of bmp2b(Fig. 5A). Primer extension analysis revealed the putative transcriptional start site 2551 bp upstream of the ATG (data not shown) and we identified a 2.1 kb intron 15 bp upstream of the start codon. Seven PCR fragments were generated to cover the 3.5 kb region upstream of the ATG (Fig. 5A)and these were subjected to DNA electrophoretic mobility shift assays (EMSA). Two DNA fragments (F4, +280 to +901; and F7, +1723 to +2343) showed changes in electrophoretic mobility in the presence of Boz, indicating specific binding(data not shown). These large fragments were further digested into sets of smaller fragments to identify Boz binding sites by EMSA (data not shown) and two smaller fragments (B1, +1980 to +2033; B4, +430 to +467) were identified that were strongly bound by Boz. Five other sites containing a potential Bicoid-class homeodomain binding motif (C)TAAT(C) were identified in the promoter and analysed by EMSA (double-stranded oligonucleotides B2, B3, B5,B6, B7; Fig. 5A) but they were bound only very weakly or not at all by Boz. Only the B1 and B4 double-stranded oligonucleotides showed specific binding to recombinant Boz protein (Fig. 5B,C). The binding of Boz to the radiolabeled double-stranded oligonucleotides B1 and B4 could be specifically competed by either B4 or B1 competitor oligonucleotides,but not efficiently by the Bicoid-site, B2, B3 or B5. The specific binding of Boz to B1 and B4 sites in bmp2b supports the hypothesis that bmp2b is a direct target of the transcriptional repressor Boz.

Boz binding sites in bmp2b confer Boz-dependent repression of bmp2b

To investigate whether the B1 and B4 sites mediate Boz-dependent bmp2b repression in vivo, we performed luciferase reporter assays in blastula stage embryos. A fusion of the luciferase reporter (BmppLUC) with 3.5 kb of sequences 5′ of the Bmp2b ATG, including the first intron and 929 bp upstream of the transcription start site, was injected into one-cell-stage embryos and found to drive expression of luciferase after mid-blastula transition. Co-injection of BmppLUC with boz mRNA or En-bozmRNA resulted each in about 6.5-times repression, whereas co-injection with VP16boz mRNA resulted in 8.8-times increase in luciferase activity(Fig. 6A). Neither bozmRNA nor En-boz mRNA has a significant effect on the expression of LUC from the pCMVtkLUC control plasmid(Fig. 6A). In a separate set of experiments, deletions of Boz binding sites B1 (Bmpδ1pLUC; 4.6-times repression) or B4 (Bmpδ4pLUC; 4.5-times repression) each reduced the repression observed with BmppLUC (10.2-times repression) by more than half(Fig. 6B). Deletion of both B1 and B4 (Bmpδ1,4pLUC; 1.6-times repression) reduced the repression six times to levels within the error margin of control injections. Thus, we show that Boz can repress bmp2b transcription directly by binding to sites B1 and B4 within the first intron of bmp2b.

Boz acts as a transcriptional repressor in organizer development

The N-terminus of Boz contains an eh1 motif, which is conserved among Engrailed genes and contributes substantially to the ability of Engrailed to repress transcription in Drosophila(Smith and Jaynes, 1996; Koos and Ho, 1998). We tested the functional significance of the eh1 motif in Boz and found that its deletion completely abolished the dorsalizing activity of Boz. We compared the activity of Boz with the activity of fusion proteins in which the Boz homeodomain is fused to the Engrailed repressor domain or the VP16 activator domain. The En-Boz repressor fusion mimicked the endogenous activity of Boz. En-Boz caused dorsalization with activation of the organizer genes gsc and chd, and, significantly, could rescue the boz mutant phenotype. By contrast, overexpression of the VP16-Boz activator fusion phenocopied boz mutant defects including downregulation of the dorsal organizer genes and resulted in ventralization of the embryo (Shimizu et al.,2002). These data support the hypothesis that Boz acts as a transcriptional repressor required for proper organizer formation. How does the repressor activity of Boz promote organizer formation and expression of organizer-specific genes? Because the removal of Bmp activity from the dorsal side of the embryo is crucial for dorsal development, we hypothesize that bmp2b is a direct target of Boz repression.

Transcriptional repression of bmp2b in the nascent organizer by Boz

Bmp2 and Bmp4 function as instructive signals that determine positional identities along the dorsoventral axis(Dosch et al., 1997). Zebrafish swirl/bmp2b mutants(Kishimoto et al., 1997; Nguyen et al., 1998) share common phenotypic features with bmp4 mutant mice(Winnier et al., 1995) and the Swirl/bmp2b expression pattern is similar to that of Xenopus bmp4 (Nikaido et al.,1997). Thus, zebrafish bmp2b seems to be similar in function to mouse or Xenopus bmp4, playing essential roles in the establishment of the dorsoventral axis in these organisms. In zebrafish, bmp2b is initially ubiquitously expressed at sphere stage. Only a small domain at the dorsal margin is devoid of bmp2b expression. bmp2b expression becomes progressively restricted to the ventral region during late blastula and early gastrula stages to form the Bmp gradient(Nikaido et al., 1997). Expression in the yolk syncytial layer (YSL) is maintained during gastrulation. Although the initiation of bmp2b expression does not depend on its function, the maintenance of bmp2b expression from shield stage on requires functional Bmp2b protein and its downstream effector Somitabun/Smad5 (Kishimoto et al., 1997; Hild et al.,1999; Kramer et al.,2002). Although the Bmp antagonist Chordin is required to form and maintain the dorsoventral gradient of bmp2b expression during gastrulation, bmp2b expression at sphere stage is normal in chd mutant embryos(Miller-Bertoglio et al.,1997). Likewise, chd expression is normal through early shield stage in the dorsalizing mutants swr, sbn and snh. In these mutants, chd expression expands in lateral and ventral directions only as epiboly proceeds(Miller-Bertoglio et al.,1997). The initiation of the dorsoventral asymmetry of bmp2b expression at sphere stage is thought to depend on dorsally localized factor(s), but the exact mechanism by which bmp2bexpression is initially blocked on the dorsal side was previously unclear.

Our data provide five lines of evidence that the earliest repression of bmp2b in the presumptive organizer is a direct activity of Boz at sphere stage.

  1. Although overexpression of boz or En-boz represses bmp2b, it does not affect chd expression at sphere stage. Thus, the repressor activity of Boz on bmp2b during sphere stage does not depend on Chordin.

  2. Overexpression of VP16-boz causes increased levels of bmp2b transcription from mid-blastula transition onward. Together with the reciprocal effect (repression) of boz or En-bozoverexpression on bmp2b, this suggests a direct interaction.

  3. Boz exerts its repression of bmp2b even when translation of zygotic gene products is inhibited by CHX. Thus, the effect of Boz is not mediated through a relay mechanism, in which Boz would repress a zygotic activator of bmp2b.

  4. We have identified binding sites for Boz in the bmp2b control region. These Boz binding sites are located close to splice donor and splice acceptor sites of the first intron, a location in which repressor sites for eukaryotic gene regulation are frequently found.

  5. Deletion analysis and reporter assays in zebrafish embryos reveal that the two Boz binding sites in the bmp2b control region are required for normal levels of bmp2b repression.

Model for the role of boz during initiation of the organizer

Our data show that dorsoventral asymmetry of bmp2b expression is initiated at the level of transcription(Fig. 7). In the absence of boz, maternal factors like Smad5/Sbn(Kramer et al., 2002) lead to ubiquitous zygotic expression of bmp2b. The presence of bmp2b in the organizer region of boz mutants interferes with the establishment of the organizer because Bmp2b mediates repression of chd and other organizer genes. In the presence of Boz, a bmp2b-free dorsal marginal zone is generated, which promotes proper organizer formation through other targets of β-catenin/Tcf/Lef signaling. The variable expressivity of the boz mutant phenotype can be understood as the result of the competition between Nieuwkoop-center signaling activity to establish the organizer and Bmp2b-mediated repression of organizer genes. This competition might have a variable outcome, as reflected by the variable penetrance and expressivity of the boz mutant phenotype(Fekany et al., 1999):occasionally, β-catenin signaling is strong enough to induce a near-normal organizer even in the absence of functional Boz. More often, bmp2b repression by Boz is required to permit the formation of a potent organizer. Thus, the Nieuwkoop center itself is involved in two activities: (1) induction of organizer genes (e.g. chd, noggin and gsc); and (2) direct repression of ventralizing signals at the dorsal margin.

Boz thus is a component of a complex network of negative regulatory interactions between dorsalizing and ventralizing activities. It was previously shown that mutual repression exists between Boz and the ventralizing transcription factors Vega1 (Vox) and Vega2 (Vent) (Kawahara et al., 2000a,b; Melby et al., 2000; Imai et al., 2001) as well as ved (Shimizu et al.,2002). Vega1 (Vox), Vega2 (Vent) and Ved also repress other dorsal genes, such as gsc and chd. Such multiple inhibitory interactions might be essential to control the extent of dorsal and ventral territories, which otherwise would predominantly depend on extracellular diffusion of the Bmp2b morphogen – a mechanism that might not be sufficient to generate stable borders of organizer gene expression.

Several questions remain. How can transcriptional repression of bmp2b by overexpression of boz explain the concurrent expansion of gsc and chd expression? Interestingly,following boz overexpression, gsc and chdexpression expand to marginal and dorso-animal positions, but not to ventro-animal positions. We suggest that this ventrolateral expansion of gsc and chd is not a consequence of repressed bmp2bexpression, because gsc and chd expression are normal in bmp2b/swr mutants during late blastula and early gastrula stages(Miller-Bertoglio et al.,1997; Mullins et al.,1996). Instead, we postulate that the expansion of gscand chd results from the repression of additional boz target genes like vega1 (vox) and vega2 (vent),which are ventrolaterally expressed and can repress gsc and chd (Imai et al.,2001). In the zebrafish Dfst7 mutant,which lacks both vox and vent loci, chd and gsc expression are ventrolaterally expanded(Imai et al., 2001). Following boz-mediated repression of vox and vent, and the elimination of their repressive effect on chd and gsc,β-catenin or its effectors might induce chd and gscexpression around the margin. The accumulation of β-catenin even at ventrolateral positions has been reported for late blastula and early gastrula stage zebrafish (R. Warga, Origin and specification of the endoderm in the zebrafish, Danio rerio. PhD thesis, Eberhardt-Karls-Universität Tübingen, Germany, 1996). Similarly, enhanced levels of nuclear translocation of β-catenin have been observed in Xenopus early gastrulae in a region extending to more lateral positions than those of gsc or chd expression(Schneider et al., 1996). Homologs of Siamois (Lemaire et al., 1995) and Twin(Laurent et al., 1997), which mediate the effect of β-catenin signaling to cause activation of organizer genes like gsc (Cho et al., 1991; Laurent et al.,1997) in Xenopus, have not yet been described for zebrafish.

Our findings correlate with data obtained from experiments in Xenopus, which indicate that inhibition of Bmp and Nodal, Bmp and Wnt signaling, or of Bmp, Nodal and Wnt pathways is sufficient for head induction(Glinka et al., 1997; Piccolo et al., 1999). Induction of neurectoderm in Xenopus can be promoted by Wnt/β-catenin signaling (Baker et al.,1999). This suggests that, in zebrafish, boz might serve as a downstream effector of Wnt signaling in this process. Early expression ofβ -catenin induces the expression of neural-specific markers and inhibits the expression of bmp4 in Xenopus ectoderm. Further, Wnt, but not the Bmp antagonist Noggin, can inhibit bmp4expression at early gastrula stages. Boz could be the functional homolog of an as-yet-unidentified gene in Xenopus that mediates Wnt-signaling-dependent repression of bmp4 expression.

The alteration in bmp2b expression in boz mutants from that of the wild type is not sufficient to explain the observed degree of anterioposterior and dorsoventral patterning defects. The boz mutants have less severe expansion of bmp2b expression than do chdmutants, but more severe dorsoventral patterning defects. Further,overexpression of zebrafish dkk1, a Wnt antagonist, can rescue anterior neural plate and axial mesoderm defects in boz mutant embryos (Hashimoto et al.,2000). Our finding that Boz acts as a repressor suggests that Boz might also directly interact with some component of the zygotic Wnt signaling pathway, that functions in the dorsal blastula and early gastrula.

From an evolutionary viewpoint, our findings point at potential similarities in the initiation of dorsoventral Bmp/Dpp asymmetry in both vertebrates and invertebrates. In Drosophila, transcriptional repression by Dorsal initiates dorsally restricted expression of the potent Dpp morphogen and our data show that an analogous strategy is applied to restrict the early source of Bmp2b locally in zebrafish. The later Dpp/Bmp2b– Sog/Chd antagonism then fine tunes and maintains the activity of the Dpp/Bmp2b morphogen on the appropriate side of the embryo. The requirement for a combination of initial restriction by transcriptional control and later inhibition of protein function by antagonists to mediate the initiation,establishment and maintenance of the Dpp/Bmp2b morphogen gradient is thus a feature found in both arthropods and vertebrates.

We are grateful to L. Solnica-Krezel, S. Schulte-Merker, M. Hammerschmidt,F. Argenton, J. Altschmied and F. Schioch for in situ probes and DNA constructs, and to S. Götter for maintenance of the fish. We thank D. Meyer and K. Lunde for helpful comments on the manuscript. This work was supported by grant DR 362/1 from the DFG (W.D.) and fellowships from the Alexander von Humboldt Foundation (T.C.L.), the Boehringer Ingelheim Fonds(D.N.) and Landesgraduiertenstipendium Baden-Württemberg (J.B.).

Baker, J. C., Beddington, R. S. and Harland, R. M.(
1999
). Wnt signaling in Xenopus embryos inhibits bmp4 expression and activates neural development.
Genes Dev.
13
,
3149
-3159.
Brannon, M., Gomperts, M., Sumoy, L., Moon, R. T. and Kimelman,D. (
1997
). A β-catenin/XTcf-3 complex binds to the Siamois promotor to regulate dorsal axis specification in Xenopus.
Genes Dev.
11
,
2359
-2370.
Boshart, M., Kluppel, M., Schmidt, A., Schütz, G. and Luckow, B. (
1992
). Reporter constructs with low background activity utilizing the cat gene.
Gene
110
,
129
-130.
Cho, K. W. Y., Blumberg, B., Steinbeisser, H. and De Robertis,E. M. (
1991
). Molecular nature of Spemann's organizer: the role of the Xenopus homeobox gene Goosecoid.
Cell
67
,
1111
-1120.
Dick, A., Hild, M., Bauer, H., Imai, Y., Maifeld, H., Schier, A. F., Talbot, W. S., Bouwmeester, T. and Hammerschmidt, M.(
2000
). Essential role of Bmp7 (snailhouse) and its prodomain in dorsoventral patterning of the zebrafish embryo.
Development
127
,
343
-354.
Dosch, R., Gawantka, V., Delius, H., Blumenstock, C. and Niehrs,C. (
1997
). Bmp-4 acts as a morphogen in dorsoventral mesoderm patterning in Xenopus.
Development
124
,
2325
-2334.
Driever, W., Thoma, G. and Nüsslein-Volhard, C.(
1989
). Determination of spatial domains of zygotic gene expression in the Drosophila embryo by the affinity of binding sites for the bicoid morphogen.
Nature
340
,
363
-367.
Ermakova, G. V., Alexandrova, E. M., Kazanskaya, O. V.,Vasiliev, O. L., Smith, M. W. and Zaraisky, A. G. (
1999
). The homeobox gene, Xanf-1, can control both neural differentiation and patterning in the presumptive anterior neurectoderm of the Xenopus laevis embryo.
Development
126
,
4513
-4523.
Fekany, K., Yamanaka, Y., Leung, T., Sirotkin, H. I.,Topczewski, J., Gates, M. A., Hibi, M., Renucci, A., Stemple, D., Radbill, A. et al. (
1999
). The zebrafish bozozok locus encodes Dharma, a homeodomain protein essential for induction of gastrula organizer and dorsoanterior embryonic structures.
Development
126
,
1427
-1438.
Fekany-Lee, K., Gonzalez, E., Miller-Bertoglio, V. and Solnica-Krezel, L. (
2000
). The homeobox gene bozozokpromotes anterior neuroectoderm formation in zebrafish through negative regulation of BMP2/4 and Wnt pathways.
Development
127
,
2333
-2345.
Ferreiro, B., Artinger, M., Cho, K. and Niehrs, C.(
1998
). Antimorphic goosecoids.
Development
125
,
1347
-1359.
Friedle, H., Rastegar, S., Paul, H., Kaufmann, E. and Knöchel, W. (
1998
). Xvent-1 mediates BMP-4-induced suppression of the dorsal-lip-specific early response gene XFD-1′ in Xenopus embryos.
EMBO J.
17
,
2298
-2307.
Gard, D. L., Hafezi, S, Zhang, T. and Doxsey, S. J.(
1990
). Centrosome duplication continues in cycloheximide-treated Xenopus blastulae in the absence of a detectable cell cycle.
J. Cell Biol.
110
,
2033
-2042.
Glinka, A., Wu, W., Onichtchouk, D., Blumenstock, C. and Niehrs,C. (
1997
). head induction by simultaneous repression of Bmp and Wnt signalling in Xenopus.
Nature
389
,
517
-519.
Han, K. and Manley, J. L. (
1993
). Functional domains of the Drosophila Engrailed protein.
EMBO J.
12
,
2723
-2733.
Hashimoto, H., Itoh, M., Yamanaka, Y., Yamashita, S., Shimizu,T., Solnica-Krezel, L., Hibi, M. and Hirano, T. (
2000
). Zebrafish Dkk1 functions in forebrain specification and axial mesendoderm formation.
Dev. Biol
.
217
,
138
-152.
Hauptmann, G. (
1999
). Two-color detection of mRNA transcript localizations in fish and fly embryos using alkaline phosphatase and β-galactosidase conjugated antibodies.
Dev. Genes Evol.
209
,
317
-321.
Hemmati-Brivanlou, A. and Thomsen, G. H.(
1995
). Ventral mesodermal patterning in Xenopusembryos: expression patterns and activities of BMP-2 and BMP-4.
Dev. Genet
.
17
,
78
-89.
Hild, M., Dick, A., Rauch, G.-J., Meier, A., Bouwmeester, T.,Haffter, P. and Hammerschmidt, M. (
1999
). The smad5mutation somitabun blocks Bmp2b signaling during early dorsoventral patterning of the zebrafish embryo.
Development
126
,
2149
-2159.
Ho, C. Y., Houart, C., Wilson, S. W. and Stainier, D. Y.(
1999
). A role for the extraembryonic yolk syncytial layer in patterning the zebrafish embryo suggested by properties of the hexgene.
Curr. Biol
.
9
,
1131
-1134.
Holley, S. A., Jackson, P. D., Sasai, Y., De Robertis, E. M.,Hoffmann, F. M. and Ferguson, E. L. (
1995
). A conserved system for dorsal-ventral patterning in insects and vertebrates involving Sog and Chordin.
Nature
376
,
249
-253.
Imai, Y., Gates, M. A., Melby, A. E., Kimelman, D., Schier, A. F. and Talbot, W. S. (
2001
). The homeobox genes voxand vent are redundant repressors of dorsal fates in zebrafish.
Development
128
,
2407
-2420.
Kawahara, A., Wilm, T., Solnica-Krezel, L. and Dawid, I. B.(
2000a
). Antagonistic role of Vega1 and bozozok/dharma homeobox genes in organizer formation.
Proc. Natl. Acad. Sci. USA
97
,
12121
-12126.
Kawahara, A., Wilm, T., Solnica-Krezel, L. and Dawid, I. B.(
2000b
). Functional interaction of Vega2 and Goosecoid homeobox genes in zebrafish.
Genesis
28
,
58
-67.
Kimmel, C. B., Ballard, W. W., Kimmel, S. R., Ullmann, B. and Schilling, T. F. (
1995
). Stages of embryonic development of the zebrafish.
Dev. Dyn.
203
,
253
-310.
Kishimoto, Y., Lee, K. H., Zon, L., Hammerschmidt, M. and Schulte-Merker, S. (
1997
). The molecular nature of zebrafish swirl: BMP2 function is essential during early dorsoventral patterning.
Development
124
,
4457
-4466.
Koos, D. S. and Ho, R. K. (
1998
). The Nieuwkoid gene characterizes and mediates a Nieuwkoop-center-like activity in the zebrafish.
Curr. Biol.
8
,
1199
-1206.
Koos, D. S. and Ho, R. K. (
1999
). The Nieuwkoid/dharma homeobox gene is essential for bmp2brepression in the zebrafish pregastrula.
Dev. Biol
.
215
,
190
-207.
Kramer, C., Mayr, T., Nowak, M., Schuhmacher, J., Runke, G.,Bauer, H., Wagner, D. S., Schmid, B., Imai, Y., Talbot, W. S., Mullins, M. C. and Hammerschmidt, M. (
2002
). Maternally supplied Smad5 is required for ventral specification in zebrafish embryos prior to zygotic Bmp signaling.
Dev. Biol
.
250
,
263
-279.
Laurent, M. N., Blitz, I., Hashimoto, C., Rothbächer, U. and Cho, K. W. Y. (
1997
). The Xenopus homeobox gene Twin mediates Wnt induction of Goosecoid in establishment of Spemann's organizer.
Development
124
,
4905
-4516.
Lemaire, P., Garrett, N. and Gurdon, J. B.(
1995
). Expression cloning of Siamois, a Xenopus homeobox gene expressed in dorsal-vegetal cells of blastulae and able to induce a complete secondary axis.
Cell
81
,
85
-94.
Marqués, G., Musacchio, M., Shimell, M. J.,Wünnenberg-Stapleton, K., Cho, K. W. Y. and O'Connor, M. B.(
1997
). Production of a DPP activity gradient in the early Drosophila embryo through the opposing actions of the SOG and TLD proteins.
Cell
91
,
417
-426.
Martinez-Barbera, J. P., Toresson, H., Da Rocha, S. and Krauss,S. (
1997
). Cloning and Expression of three members of the zebrafish Bmp family – Bmp2a, Bmp2b and Bmp4.
Gene
198
,
53
-59.
Melby, A. E., Clements, W. K. and Kimelman, D.(
1999
). Regulation of dorsal gene expression in Xenopusby the ventralizing homeodomain gene Vox.
Dev. Biol.
211
,
293
-305.
Melby, A. E., Beach, C., Mullins, M. and Kimelman, D.(
2000
). Patterning the early zebrafish by the opposing actions of bozozok and vox/vent.
Dev. Biol
.
224
,
275
-285.
Miller-Bertoglio, V. E., Fisher, S., Sanchez, A., Mullins, M. C. and Halpern, M. E. (
1997
). Differential regulation of Chordin expression domains in mutant zebrafish.
Developmental Biology
192
,
537
-550.
Mullins, M. C., Hammerschmidt, M., Kane, D. A., Odenthal, J.,Brand, M., van Eeden, F. J., Furutani-Seiki, M., Granato, M., Haffter, P.,Heisenberg, C. P. et al. (
1996
). Genes establishing dorsoventral pattern formation in the zebrafish embryo: the ventral specifying genes.
Development
123
,
81
-93.
Nguyen, V. H., Schmid, B., Trout, J., Connors, S. A., Ekker, M. and Mullins, M. C. (
1998
). Ventral and lateral regions of the zebrafish gastrula, including the neural crest progenitors, are established by a bmp2b/Swirl pathway of genes.
Dev. Biol
.
199
,
93
-110.
Nikaido, M., Tada, M., Saji, T. and Ueno, N.(
1997
). Conservation of BMP signaling in zebrafish mesoderm patterning.
Mech. Dev.
61
,
75
-88.
Onichtchouk, D., Gawantka, V., Dosch, R., Delius, H.,Hirschfeld, K., Blumenstock, C. and Niehrs, C. (
1996
). The Xvent-2 homeobox gene is part of the BMP-4 signalling pathway controlling dorsoventral patterning of Xenopus mesoderm.
Development
122
,
3045
-3063.
Piccolo, S., Agius, E., Leyns, L., Bhattacharyya, S., Grunz, H.,Bouwmeester, T. and De Robertis, E. M. (
1999
). The head inducer Cerberus is a multifunctional antagonist of Nodal, MBP and Wnt signals.
Nature
397
,
707
-710.
Pogoda, H.-M., Solnica-Krezel, L., Driever, W. and Meyer, D.(
2000
). The zebrafish forkhead transcription factor FoxH1/Fast1 is a modulator of Nodal signaling required for organizer formation.
Curr. Biol
10
,
1041
-1049.
Rastegar, S., Friedle, H., Frommer, G. and Knöchel, W.(
1999
). Transcriptional regulation of Xvent homeobox genes.
Mech. Dev
.
81
,
139
-149.
Ryu, S. L., Fujii, R., Yamanaka, Y., Shimizu, T., Yabe, T.,Hirata, T., Hibi, M. and Hirano, T. (
2001
). Regulation of dharma/bozozok by the Wnt pathway.
Dev. Biol
.
231
,
397
-409.
Sadowski, I., Ma, J., Triezenberg, S. and Ptashne, M.(
1988
). GAL4-VP16 is an unusually potent transcriptional activator.
Nature
335
,
563
-564.
Sasai, Y., Lu, B., Steinbeisser, H. and De, R. E.(
1995
). Regulation of neural induction by the Chd and Bmp-4 antagonistic patterning signals in Xenopus.
Nature
376
,
333
-336.
Schmid, B., Furthauer, M., Connors, S. A., Trout, J., Thisse,B., Thisse, C. and Mullins, M. C. (
2000
). Equivalent genetic roles for bmp7/Snailhouse and bmp2b/Swirl in dorsoventral pattern formation.
Development
127
,
957
-967.
Schneider, S., Steinbeisser, H., Warga, R. M. and Hausen, P.(
1996
). β-Catenin translocation into nuclei demarcates the dorsalizing centers in frog and fish embryos.
Mech. Dev.
57
,
191
-198.
Schulte-Merker, S., Lee, K. J., McMahon, A. P. and Hammerschmidt, M. (
1997
). The zebrafish organizer requires chordino.
Nature
387
,
862
-863.
Shimizu, T., Yamanaka, Y., Ryu, S.-L., Hashimoto, H., Yabe, T.,Hirata, T., Bae, Y.-k., Hibi, M. and Hirano, T. (
2000
). Cooperative roles of Bozozok/Dharma and Nodal-related proteins in the formation of the dorsal organizer in zebrafish.
Mech. Dev
.
91
,
293
-303.
Shimizu, T., Yamanaka, Y., Nojima, H., Yabe, T., Hibi, M. and Hirano, T. (
2002
). A novel repressor-type homeobox gene, Ved, is involved in dharma/bozozok mediated dorsal organizer formation in zebrafish.
Mech. Dev.
118
,
125
-138.
Smith, S. T. and Jaynes, J. B. (
1996
). A conserved region of engrailed, shared among all En-, Gsc-, Nk1-, Nk2- and Msh-class homeoproteins, mediates active transcriptional repression in vivo.
Development
122
,
3141
-3150.
Solnica-Krezel, L. and Driever, W. (
2001
). The role of the homeodomain protein Bozozok in zebrafish axis formation.
Int. J. Dev. Biol
.
45
,
299
-310.
Solnica-Krezel, L., Stemple, D. L., Mountcastle-Shah, E.,Rangini, Z., Neuhauss, S. C., Malicki, J., Schier, A. F., Stainier, D. Y.,Zwartkruis, F., Abdelilah, S. et al. (
1996
). Mutations affecting cell fates and cellular rearrangements during gastrulation in zebrafish.
Development
123
,
67
-80.
Trindade, M., Tada, M. and Smith, J. C. (
1999
). DNA-binding specificity and embryological function of Xom (Xvent-2).
Dev. Biol
.
216
,
442
-456.
Winnier, G., Blessing, M., Labosky, P. A. and Hogan, B. L.(
1995
). Bone morphogenetic protein-4 is required for mesoderm formation and patterning in the mouse.
Genes Dev.
9
,
2105
-2116.
Yamanaka, Y., Mizuno, T., Sasai, Y., Kishi, M., Takeda, H., Kim,C. H., Hibi, M. and Hirano, T. (
1998
). A novel homeobox gene, dharma, can induce the organizer in a non-cell-autonomous manner.
Genes Dev.
12
,
2345
-2353.

Supplementary information