The zebrafish bozozok locus encodes Dharma, a homeodomain protein essential for induction of gastrula organizer and dorsoanterior embryonic structures

A series of inductive events establishes the vertebrate embryonic axis. The dorsal gastrula organizer, or Spemann organizer, is a key component identified as a region on the dorsal side of the amphibian gastrula that induces a secondary axis when transplanted to the ventral side of a host embryo. The donor organizer differentiates into axial tissues including notochord, prechordal plate and ventral neural fates, induces and patterns neural tissue and coordinates gastrulation movements (Spemann, 1938). Structures homologous to the Spemann organizer have been identified in other vertebrates based on similar inductive properties and gene expression patterns, and include the node in mouse and chick, and the embryonic shield in fish (Harland and Gerhart, 1997; Lemaire and Kodjabachian, 1996). Secreted factors like Chordin and Noggin perform some of the inductive functions of the dorsal gastrula organizer by antagonizing ventrally expressed bone morphogenetic proteins-2/4 (BMPs). A gradient of BMP-2/4 activity is instructive in dorsoventral patterning of the germ layers during gastrulation (Neave et al., 1997; Nguyen et al., 1998; Thomsen, 1997).

The genetic hierarchy underlying the establishment of the gastrula organizer is only beginning to be elucidated (Moon and Kimelman, 1998). In frog and fish embryos, microtubule-mediated transport of dorsal determinants from the vegetal pole towards the prospective dorsal blastomeres (Jesuthasan and Strahle, 1997; Rowning et al., 1997) is essential for the establishment of the dorsal blastula organizer, or Nieuwkoop center (Dale and Slack, 1987; Nieuwkoop, 1973). The Nieuwkoop center of the frog embryo acts predominantly in the dorsal vegetal endodermal precursors and induces the organizer via non-autonomous signals in the overlying blastoderm, but might also contribute to the organizer through self-induction (reviewed in Moon and Kimelman, 1998). A key step in the formation of the Nieuwkoop center is nuclear localization of β-catenin, an effector of the Wnt signaling pathway (Gumbiner, 1995; Heasman et al., 1994; Larabell et al., 1997; Schneider et al., 1996). Dorsally localized β-catenin, in complex with Tcf-like transcription factors, is thought to trigger axis formation by transcriptional activation of zygotic target genes. Xenopus Siamois and Twin, related homeodomain transcription factors, may be direct targets of the Tcf/β-catenin complex and could act as intermediaries between the maternal and zygotic programs executing axis formation (Brannon and Kimelmann, 1996; Fan and Sokol, 1997; Kessler, 1997; Laurent et al., 1997; Lemaire et al., 1995).

Formation of the gastrula organizer may require the coordination of the β-catenin, TGF-β and possibly other signaling pathways (Heasman, 1997; Kimelman et al., 1992). The TGF-β superfamily members Activin, Vg1 and Nodal-related proteins, which are often classified as mesoderm-inducing factors (MIFs), have axis-inducing activities (Kessler and Melton, 1995; Smith et al., 1990). The integration of the TGF-β and β-catenin pathways is best seen in the regulation of the organizer-specific expression of goosecoid (gsc), whose promoter features both β-catenin-binding sites and TGF-β response elements (Crease et al., 1998; Watabe et al., 1995).

In the zebrafish blastula, the functional equivalent of the Nieuwkoop center is likely to be located in the dorsal yolk syncytial layer (YSL) underlying the blastoderm (reviewed in Schier and Talbot, 1998; Solnica-Krezel, 1999). Accordingly, the YSL has mesoderm-inducing and patterning activities (Mizuno et al., 1996) and nuclear accumulation of β-catenin is detected in the dorsal YSL (Schneider et al., 1996). The homeobox gene dharma is a likely target of β-catenin. In the zebrafish blastula, dharma is expressed predominantly in the dorsal YSL and it can induce gastrula organizer formation when overexpressed in the yolk syncytium (Yamanaka et al., 1998). Furthermore, two nodal-related genes, cyclops and squint, have been implicated in organizer development and function, with squint being expressed in the YSL (Erter et al., 1998; Feldman et al., 1998; Rebagliati et al., 1998a,b; Sampath et al., 1998).

Extraembryonic structures have also been implicated in axis formation during murine development. Recent important discoveries indicated that, in the mouse embryo, notwithstanding the central role of the organizer/node in axis formation, the extraembryonic anterior visceral endoderm (AVE) may be responsible for development of anterior embryonic structures. Surgical removal of extraembryonic AVE leads to reduced expression of anterior neural markers (Thomas and Beddington, 1996). Furthermore, chimera experiments have shown the wild-type functions of TGF-β pathway components, Nodal and Smad2, are necessary in the AVE for proper anterior-posterior (AP) patterning (Varlet et al., 1997; Waldrip et al., 1998). It has been proposed that the AVE is an equivalent of dorsoanterior yolky endoderm, a component of the amphibian gastrula organizer. AVE might fulfill a function of a head organizer that is physically distinct from the trunk organizer activity residing in the node (Beddington and Robertson, 1998; Bouwmeester and Leyns, 1997).

To investigate the genetic basis of axis formation, we have analyzed the zebrafish bozozok (boz) locus. At 1 day postfertilization (dpf), boz^m168/m168 mutants exhibit a reduction or loss of the main organizer derivatives, notochord and prechordal plate, as well as deficiencies in anterior and ventral CNS structures, suggesting a key role of this gene in axis formation (Driever et al., 1997; Solnica-Krezel et al., 1996). Here we report that boz is required at the blastula stages for development of the dorsal gastrula organizer and specification of organizer-derived axial mesodermal fates. Moreover, we show that boz encodes the homeodomain protein, Dharma, recently isolated based on its ability to induce axis formation in the zebrafish embryo (Yamanaka et al., 1998). We demonstrate that boz function in the YSL is sufficient to induce the organizer and development of anterior structures in the overlying mutant blastoderm. These studies establish boz as an essential component of the axis induction pathway in zebrafish, acting presumably at the top of a hierarchy of zygotic genes activated by maternal β-catenin.

Fish and embryos were maintained essentially as described in Solnica-Krezel et al. (1994). boz mutants were maintained in an AB/India background (Knapik et al., 1996).

F₁ progeny of γ-irradiation mutagenized males and wild-type females (obtained from M. Halpern, Carnegie Institution of Washington) were crossed with boz^m168/+ females. One out of 250 males repeatedly produced 4% boz mutant progeny in such crosses.

RNA in situ hybridizations were performed essentially as described in Oxtoby and Jowett (1993) and double in situ hybridizations as in Marlow et al. (1998). β-catenin was detected with diluted raw antisera (gift from P. Schneider and R. T. Moon) as in Solnica-Krezel and Driever (1994).

mRNA was synthesized using Ambion’s mMessage mMachine kit. mRNA injections into 1-to 8-cell embryos were performed as described in Marlow et al. (1998). For YSL injections, embryos were dechorionated with Pronase. After YSL formation (1000-cell stage), embryos were coinjected with boz/dharma and fluorescent dextran into the yolk just below the blastoderm. After injections, the localization of the RNA was inferred from visualization of the fluorescent dextran using epifluorescence.

5-day-old embryos from boz^m168/+; AB/Tu, were digested in 90 μg/ml protease (Qiagen), 10 mM Tris-HCl pH 8.0, 1 mM EDTA, 0.05% SDS at 37°C for 48 hours. A dharma DNA fragment was amplified with PCR primers Dha39 5′-ATA CTC ACG CAG CTT TTG GG-3′ and Dha 350R 5′-GGT TGC CCT GCA TAG TAA GTC-3′. The PCR conditions were 94°C 2 minutes (1 cycle), then 94°C 30 seconds, 56°C 30 seconds and 72°C 40 seconds (30 cycles), followed by 72°C 7 minutes (1 cycle). Amplified products were digested by HaeIII and resolved on 1.8% agarose gels.

Oligonucleotides corresponding to dharma were used in PCR reactions with genomic DNA from boz^m168/m168 mutant embryos and from wild-type siblings. Amplified fragments, purified by agarase treatment after electrophoresis in low-melting-point agarose, were sequenced with an ABI 377 automated sequencer. The stop codon in boz^m168 was identified in several independently amplified fragments and confirmed by RFLP analysis.

Following in situ hybridization, single embryos were digested in 20 μl of water with 20 μg/ml proteinase K and 5 μl PBT at 55°C overnight. After boiling for 5 minutes, 4 μl of digested solution served as a template in 20 μl PCR reactions. PCR conditions were as above, but with 40 cycles.

bozozok^m168(boz^m168) was identified as a recessive, zygotic effect, embryonic lethal mutation that disrupts development of midline embryonic structures with variable penetrance and expressivity (Driever et al., 1997; Solnica-Krezel et al., 1996). We observed an increased severity of the boz^m168 phenotype in an India/AB hybrid genetic background, in which mutant embryos frequently exhibited severe anterior truncations. Five boz phenotypic classes (I-V) were defined based on morphological criteria (Fig. 1). Class I mutants are the most severe, with complete loss of notochord, eyes and forebrain, reduction of midbrain and an abnormal neural keel. In these mutants, expression of the forebrain marker emx1 is reduced or absent. However, krox20 expression domains in rhombomeres 3 and 5 of the hindbrain are enlarged relative to wild type (Fig. 1E). A similarly variable phenotype is exhibited by a putative allele, bozⁱ², described recently (Blagden et al., 1997). Furthermore, in crosses between boz^m168/+ and fish heterozygous for a new γ-irradiation-induced allele, boz^v9, 6% of the progeny (n=1,675) displayed a variable boz^m168-like phenotype. boz^v9 segregates like a translocation, presumably leading to a deficiency of the boz locus (see below). Notably, the boz^v9/m168 phenotype was not stronger than the boz^m168/m168 phenotype, suggesting that the boz^m168 allele strongly reduces or eliminates boz⁺ function (see below).

Due to the decreased penetrance in progeny of older females (K. D. F. and L. S. K., unpublished data), boz^m168/m168 homozygotes are sometimes viable. In crosses of boz^m168/m168 females with wild-type males, 100% of the progeny are phenotypically wild-type boz^m168/+ heterozygotes, indicating that there is not a strict maternal requirement for boz function. All experiments described in this manuscript utilized the boz^m168 mutant allele (boz) unless indicated otherwise.

Notochord and prechordal plate are reduced or missing in boz mutants by the tailbud stage (Solnica-Krezel et al., 1996). Here, we asked when the loss of these axial structures can first be detected in strongly affected boz mutants. At 60% epiboly (7 hours postfertilization, hpf), ntl expression in prospective notochord cells above the dorsal blastoderm margin was reduced or absent in boz embryos (Fig. 2A,B; Schulte-Merker et al., 1992). Furthermore, expression of ntl mRNA in dorsal forerunner cells, located just below the blastoderm margin in wild-type embryos, was not observed in boz mutants, consistent with the absence of these cells in boz gastrulae (Fig. 2A,B and data not shown). Dorsal forerunner cells give rise to the epithelial lining of Kupffer’s vesicle, a transient structure located midventrally in the extending tailbud during somitogenesis (Fig. 2G; Cooper and D’Amico, 1996; Kimmel et al., 1995). Examination of boz embryos revealed a lack of a morphologically distinct Kupffer’s vesicle (Fig. 2H).

Further analysis at 1 dpf revealed that type II collagen (col2a1) expression, normally marking the hypochord and the floor plate of the spinal cord, as well as the nascent notochord in the posterior part of the tail (Yan et al., 1995), was reduced or missing in boz mutant embryos (Fig. 2I,J). Therefore, the boz^m168 mutation interferes with formation of multiple axial cell types including notochord, prechordal plate, hypochord, floor plate and dorsal forerunner cells, already during gastrulation, when these cell types are being specified.

Since boz mutant embryos displayed deficiencies in organizer-derived axial tissues during gastrulation, we investigated effects of the boz^m168 mutation on organizer development. The zebrafish embryonic shield, an equivalent of the Spemann gastrula organizer, is first observed as a thickening of the germ ring on the dorsal side of wild-type gastrulae at 6 hpf (Kimmel et al., 1995; Fig. 2E). In contrast, strongly affected boz gastrulae exhibited a circumferentially uniform germ ring (Fig. 2F). Additionally, a dorsal thickening of the ntl expression domain due to accumulation of ntl-expressing cells was not observed in boz embryos (Fig. 2C,D). Expression of other genes in the embryonic shield was decreased or absent in boz mutants, including sonic hedgehog, lim1, sek1 and TARAM-A (data not shown; Krauss et al., 1993; Renucci et al., 1996; Toyama et al., 1995; Xu et al., 1994). Analysis of axial (HNF-3β homolog) expression demonstrated that boz^m168 affected only the dorsal mesodermal expression domain of this gene, corresponding to prechordal plate and notochord precursors, while its expression was normal in the presumptive endodermal cells localized close to the YSL (Fig. 2K,L; Strahle et al., 1996).

Notably, at the onset of gastrulation, expression of the notochord-specific floating head (flh) gene in the dorsal marginal cells of the forming germ ring was very reduced in boz embryos (Fig. 2M,N; Talbot et al., 1995). Furthermore, at 40% epiboly (5 hpf), the normal dorsal expression of gsc was strongly decreased (Fig. 3A,B; Schulte-Merker et al., 1994; Stachel et al., 1993). Since the embryonic shield does not form in boz mutants, and expression of dorsal-specific genes is affected before the shield stage, the boz⁺ function must be required at the blastula stages for formation of the complete organizer.

The dorsal gastrula organizer is thought to be induced by signals regulated by maternal β-catenin, emanating from the Nieuwkoop center at the blastula stage (reviewed in Moon and Kimelman, 1998; Heasman, 1997). Considering the effects of boz^m168 on organizer formation, we carried out experiments designed to determine where boz acts with respect to components of this pathway.

Treatment of early zebrafish blastulae with 0.3 M lithium chloride (LiCl) dorsalizes embryos by activating the β-catenin pathway (Hedgepeth et al., 1997; Klein and Melton, 1996). Accordingly, treatment of wild-type embryos with LiCl led to expansion of the dorsal expression domain of gsc around the circumference of the gastrula (Stachel et al., 1993; compare Fig. 3A and C). In contrast, when clutches containing boz embryos were treated with LiCl a distinct class of embryos (<25%) was observed, expressing very low levels of gsc around the margin (Fig. 3D). Therefore, LiCl treatment can only induce ectopic gsc in boz mutants at low levels, similar to those found in untreated mutant embryos (compare Fig. 3B and D). In another experiment, embryos treated with LiCl were dorsalized, displaying, during early segmentation, laterally expanded somites and ectopic notochords. LiCl-treated boz embryos also had slightly expanded somites, but they still lacked notochord (data not shown). Among 1,068 progeny of boz heterozygotes treated with LiCl, 84% exhibited ectopic notochords, 12% displayed the boz phenotype, and 4% were unaffected wild type. Similarly, from 571 untreated control embryos, 11% were phenotypically boz. Since the frequency of boz^m168 mutants in the controls corresponds to those not responding to LiCl, we conclude that the boz^m168 mutation blocks the ability of LiCl to induce axial mesoderm.

The relationship between boz and β-catenin was investigated further by overexpression experiments. Injections of β-catenin mRNA (80-240 pg) dorsalized wild-type embryos and induced complete secondary axes, consistent with previous reports (Fig. 3E,G; Table 1; Kelly et al., 1995). Among the progeny of boz^m168/+ fish injected with β-catenin mRNA, boz mutants were still easily identifiable, with a penetrance comparable to uninjected controls (Table 1). Notably, a small fraction of embryos exhibited two boz-like axes, lacking notochord (Fig. 3F,H; Solnica-Krezel et al., 1996). In separate experiments, injection of β-catenin into wild-type embryos induced multiple thickenings of ntl expression, corresponding to ectopic shields. Conversely, 21% of the progeny of boz^m168/+ fish, injected with β-catenin exhibited uniform ntl expression in the margin (data not shown). Since β-catenin mRNA failed to induce an embryonic shield or notochord in boz mutant embryos, we concluded that β-catenin is unable to suppress the boz phenotype. Hence, boz acts downstream or in a parallel pathway to β-catenin.

To determine if the lack of suppression of the boz phenotype in β-catenin overexpressing embryos was due to interference of the boz^m168 mutation with nuclear accumulation of β-catenin, the subcellular localization of β-catenin was analyzed in such embryos at the late blastula stages (4-4.7 hpf) using a specific polyclonal antibody (Larabell et al., 1997; Schneider et al., 1996). Among wild-type embryos, 89% exhibited nuclear staining (n=88). Among the progeny produced by boz^m168/+ and/or boz^m168/m168 fish, 34% (n=167) were phenotypically boz mutant embryos and 93% (n=140) exhibited nuclear staining (Fig. 3J). These results indicated that boz^m168 does not interfere with the nuclear accumulation of β-catenin, further placing boz downstream or parallel to the maternal β-catenin pathway.

The dorsalizing signals of the TGF-β superfamily, Nodal and zDVR1/Vg1, which are thought to act either downstream and/or in parallel to β-catenin, can induce incomplete secondary axes when overexpressed in zebrafish embryos (Dohrmann et al., 1996; Toyama et al., 1995). Similarly, overexpression of a constitutively active form of an activin/TGF-β-related type I receptor, TARAM-A-D, but not the wild-type receptor TARAM-A, can induce complete secondary axes (Renucci et al., 1996). Hence we tested the relationship between boz and dorsalizing TGF-β signals.

Overexpression of TARAM-A mRNAs did not produce any obvious morphological effects in wild-type embryos and failed to suppress the boz phenotype (Table 2). In contrast, injection of mRNA encoding the activated receptor TARAM-A-D was able to suppress the axial mesoderm deficiency in boz mutant embryos. TARAM-A-D mRNA-injected embryos obtained from boz^m168/+ parents could not be distinguished as boz or wild type; all embryos ectopically expressed the chordamesoderm marker ntl and a marker of the prechordal plate-derived hatching gland, hgg1 (Fig. 4E; Table 2; Schulte-Merker et al., 1992; Thisse et al., 1994), whereas boz embryos have reduced or absent ntl and hgg1 expression (Fig. 4A-D).

To test the possibility of boz acting upstream of the receptor, we analyzed the response of boz embryos to two possible ligands, Nodal and the zebrafish homolog of Vg1, zDVR1. Overexpression of murine nodal mRNA in wild-type zebrafish dorsalized the embryos, inducing ectopic ntl and hgg1 consistent with previous reports (Toyama et al., 1995). Similarly, when nodal mRNA was injected into progeny of boz^m168/+ fish the axial mesoderm deficiency in boz mutant embryos was suppressed (Fig. 4F; Table 2). A similar rescue of hgg1 and ntl notochord expression in boz mutants was achieved by ectopic expression of BzDVR-1, a hybrid form containing the BMP pro region and Vg1 mature region (Table 2; Fig. 4G; Dohrmann et al., 1996). These studies indicate that both TGF-β ligands and a constitutively active type I receptor can suppress the deficiency of prechordal and chordamesoderm in boz mutants, suggesting that boz acts either upstream or in parallel to TGF-β signaling in the process of axis formation.

The early action of boz and the epistatic analyses presented here indicated that boz could be an early zygotic target of β-catenin/Tcf action. Recently, a homeobox gene called dharma with axis-inducing activity, was identified in zebrafish (Yamanaka et al., 1998). Expression of dharma is observed predominantly in the dorsal YSL from the early blastula stage and can be induced by LiCl treatment, suggesting that it is an early target of β-catenin (Yamanaka et al., 1998). The spatiotemporal expression pattern as well as the inductive potential of dharma made it a good candidate for corresponding to the boz locus.

The boz^m168 mutation was mapped to LG15 based on its linkage with the chordino locus (Schulte-Merker et al., 1997; E. Gonzales and L. S. K., unpublished observations) and subsequently by the use of simple sequence-length polymorphism (SSLP) markers (Knapik et al., 1998). This analysis positioned Z3358 approximately 1.8 cM (29 recombinants among 1620 meioses) and Z3760 about 6.9 cM (112/1620) to either side of the boz locus (Fig. 5A). Linkage analysis in wild-type and boz mutant mapping crosses placed dharma within the same interval of LG15 (Fig. 5A), prompting us to sequence the genomic DNA from boz^m168/m168 mutants to identify possible mutations in dharma. Sequence analysis revealed a guanine-to-adenine transition (TGG→TGA) that introduced a premature stop codon and truncated the dharma open reading frame (ORF) at codon 70, 45 amino acids upstream of the homeodomain (Fig. 5B).

To test linkage between the lesion in the dharma gene and the boz^m168 mutation, we analyzed a HaeIII restriction site polymorphism introduced by the mutation in boz^m168 crosses. This dharma polymorphism segregated with boz in a DNA mapping panel of 810 diploid boz mutants. This indicated that boz and dharma are less than 0.06 cM apart (Fig. 5B), corresponding on average to a physical distance of 30 to 50 kb (Postlethwait et al., 1994). Furthermore, we found that the dharma sequences are deleted by boz^v9, a γ-ray-induced allele (Fig. 5D).

To test directly if Dharma could rescue the boz phenotype, dharma mRNA (20-500 pg) along with a rhodamine dextran lineage tracer was injected into the yolk of 1-to 4-cell-stage embryos generated by boz^m168/m168 and boz^m168/+ fish. These injections delivered mRNA to most cells of the developing embryos (Fig. 6E,F; Hammerschmidt et al., 1996). All doses dorsalized the embryos and induced ectopic axial mesoderm tissues (Fig. 6C,D). Furthermore, all doses suppressed the absence of notochord and prechordal plate in boz embryos (Table 3).

The close linkage of dharma and boz, presence of a nonsense mutation in the boz^m168dharma ORF, deletion of dharma sequences by the boz^v9 allele and ability of dharma RNA to suppress the boz mutant phenotype provide conclusive evidence for Dharma being encoded by the boz locus. Therefore, we refer to this gene hereafter as bozozok (boz).

Among the progeny of boz^m168/+ and boz^m168/m168 fish at high stage (3.3 hpf), most embryos (97%) exhibited a normal boz/dharma expression pattern in the dorsal blastomeres (Yamanaka et al., 1998). However, already by sphere stage (4 hpf) and continuing during gastrulation, boz^m168/m168 mutants (retrospectively genotyped as described in Materials and Methods) did not express boz/dharma (Fig. 6A,B; Table 4). Therefore, the maintenance but not the establishment of boz expression requires its own function.

Expression of boz/dharma in the dorsal YSL and in dorsal blastomeres at the blastula stages could indicate that boz function is required in both compartments (Yamanaka et al., 1998). To determine whether expression of boz in the YSL is sufficient for its function, boz/dharma mRNA was coinjected with rhodamine-conjugated dextran into the YSL at 1000-cell stage, after the YSL became a separate compartment (Fig. 6G,H). Although injections of boz/dharma mRNA into the YSL were not as effective at dorsalizing wild-type embryos as injections into 1-to 8-cell-stage embryos (Yamanaka et al., 1998), they significantly reduced the penetrance of the boz^m168/m168 mutant phenotype (Table 3), as judged by analyzing injected embryos at day 1 of development for the presence of notochord, floor plate and anterior structures (data not shown). We conclude that overexpression of boz in the YSL of the boz mutant embryos can fully rescue the mutant phenotype. Hence, boz function in the extraembryonic YSL is sufficient for normal development of the overlying blastoderm.

Many genes have been identified in recent years that can induce axis formation when ectopically expressed in a vertebrate embryo, although only a few have been shown to be essential in this process (Moon and Kimelman, 1998). This work provides genetic evidence for the zebrafish boz locus being an early zygotic component required for development of a complete dorsal gastrula organizer, dorsal mesodermal tissues and anterior neural structures. We have found that the boz locus corresponds to the dharma gene encoding a paired-class homeodomain protein (Yamanaka et al., 1998). The presented molecular and genetic evidence argues that the boz^m168 mutation is a null allele.

The absence of the embryonic shield and the loss or reduction of organizer-specific gene expression in boz^m168/m168 mutants demonstrates that boz function is necessary for the formation of a complete gastrula organizer. Furthermore, boz affects specification of numerous organizer-derived dorsal midline structures, including notochord, prechordal mesendoderm, hypochord, floor plate and dorsal forerunner cells. This phenotype strikingly resembles defects obtained after surgical removal of the embryonic shield, consistent with the idea that the genetic and mechanical ablations of this important signaling center have similar developmental consequences (Shih and Fraser, 1996).

However, while removal of the embryonic shield does not lead to anterior-posterior patterning defects (Driever et al., 1997; Shih and Fraser, 1996), strongly affected boz mutants lack the most anterior neural structures. This indicates that boz mutations perturb development more extensively than extirpation of the embryonic shield. One possibility is that some organizer and boz-dependent patterning events occur before morphological manifestation of the embryonic shield. Accordingly, the dorsal expression of a number of genes like gsc and lim1 precedes shield formation (Schulte-Merker et al., 1994; Stachel et al., 1993; Toyama et al., 1995) and is reduced or absent in boz mutants. Another intriguing possibility is that boz might specify anterior fates in an organizer/shield-independent fashion (Fig. 7).

Where and when in the embryo is boz function required for normal development? Since boz/dharma is expressed predominantly in the YSL and its ectopic expression in the YSL led to ectopic expression of the organizer-specific gene gsc in the overlying blastoderm, it was suggested that boz acts in the YSL to induce the gastrula organizer in a nonautonomous manner (Yamanaka et al., 1998). These studies could not exclude, however, a possible requirement for boz function in the blastoderm cells in this process. Here, we demonstrated that injections of wild-type boz mRNA exclusively in the YSL of boz mutant embryos fully rescued all aspects of the boz mutant phenotype, which strongly argues that boz function in the YSL is sufficient for normal development.

The nuclear localization of β-catenin in the dorsal YSL, together with the inductive properties of the yolk syncytium, led to the proposal that the dorsal YSL in fish embryos corresponds to the Nieuwkoop center of frog embryos (Mizuno et al., 1996; Schneider et al., 1996). We hypothesize that boz acts in the YSL as a component of the Nieuwkoop center and that it is essential for induction of the gastrula organizer in the overlying blastoderm. In Xenopus, the Nieuwkoop center is thought to induce the gastrula organizer in the overlying blastomeres in a non-autonomous fashion (Wylie et al., 1996). But, there is also evidence that the gastrula organizer is established directly by inheritance of dorsal determinants (Laurent et al., 1997). Our work provides evidence for a nonautonomous induction of the organizer by the dorsal YSL in zebrafish. However, Bozozok may also act in the blastoderm, e.g. in a dorsal group of non-involuting endocytic marginal (NEM) cells that remain in cytoplasmic confluence with the YSL even after formation of this extraembryonic layer and subsequently give rise to dorsal forerunner cells in the superficial layers of the shield (Cooper and D’Amico, 1996). Since forerunner cells are missing in strongly affected boz mutants, it is plausible that boz acts both in these cells and in the YSL during normal development as well as when ectopically expressed in the YSL. It remains to be determined if signals from NEM cells contribute and are necessary for organizer formation.

Several lines of evidence indicate that boz is a downstream target of dorsal maternal β-catenin and is essential for β-catenin-mediated organizer induction. Activation of the β-catenin pathway failed to suppress the boz phenotype, placing boz downstream or in parallel to β-catenin. The observation that ectopically injected β-catenin exhibits nuclear localization in boz mutants, that in LiCl-treated embryos expression of boz mRNA is increased (Yamanaka et al., 1998) and that there are several consensus Tcf/Lef binding sites in the boz promoter region (Y. Y., M. H. and T. H., unpublished data) strongly argue for boz being a downstream target of β-catenin.

Frog embryos in which maternal β-catenin mRNA was depleted fail to develop any dorsoventral asymmetry (Heasman et al., 1994). In contrast, boz mutant embryos exhibit an incomplete axis with somites and neural tube. Therefore, the loss of boz function most likely only partially blocks β-catenin signaling. We hypothesize that boz is required for specification of dorsoanterior fates but not for the ability of β-catenin to induce the remaining dorsal tissues, including somitic mesoderm and neuroectoderm (Fig. 7).

The functional similarities between the frog homeobox genes siamois and twin and the zebrafish boz gene are striking. All three exhibit similar spatiotemporal expression patterns encompassing, but not limited to, the position of the Nieuwkoop center in frog and fish embryos, their expression is induced by β-catenin before organizer formation and they possess similar activities in gain-of-function experiments (Laurent et al., 1997; Lemaire et al., 1995; Yamanaka et al., 1998). Notably, inhibition of Siamois function via overexpression of a dominant repressor construct (SE) in which the Siamois homeodomain was fused to an active repression domain of D. melanogaster Engrailed inhibits Spemann organizer and axis formation in the frog embryo (Fan and Sokol, 1997; Kessler, 1997). This phenotype is similar to, but stronger than that described here for boz^m168. Considering that the homeodomains of Siamois and Twin are highly related, it is likely that the SE repressor interferes with the function of both proteins (Laurent et al., 1997). Similar to its failure to rescue the boz phenotype, β-catenin cannot suppress the loss of axial structures caused by the SE repressor. Therefore, like Bozozok, Siamois and Twin could be essential for gastrula organizer formation, acting downstream of β-catenin. However, definitive evidence for and identification of specific functions of Siamois and Twin in axis formation has to await the loss-of-function experiments for each of these two genes.

The key to understanding the mechanisms by which boz functions in the YSL in axis specification will be the identification of downstream targets encoding signaling molecules that can affect fates of overlying blastoderm cells. We provided evidence that ectopic expression of both ligands and an activated receptor of the TGF-β family, Vg1, mouse Nodal and TARAM-A-D can suppress the notochord and prechordal plate deficiency in boz mutants. One interpretation of these results is that Bozozok acts, directly or indirectly, upstream of a TGF-β-like signal that induces dorsal fates. Alternatively, boz could act in a parallel pathway, or downstream of TGF-β only partially inhibiting the signaling (Fig. 7). Consistent with boz acting upstream of TGF-β signaling, dorsal expression of two zebrafish nodal-related genes, cyclops and squint, is greatly reduced in boz mutants (Sampath et al., 1998; C. E. Erter, L. S. K. and C. V. E. Wright, unpublished observations). Notably, cyclops and squint fulfill partially redundant functions in establishment and function of the organizer (Feldman et al., 1998; Rebagliati et al., 1998a,b; Sampath et al., 1998). Furthermore, squint, like boz, is expressed in and can induce expression of organizer genes from within the YSL, making it an attractive candidate for a downstream boz target (Feldman et al., 1998; Erter et al., 1998).

While boz is the first gene shown to be required for development of both axial mesoderm and anterior-posterior neural patterning in the fish, mutations in several murine genes, such as nodal, HNF-3β, lim1, otx2 and smad2 lead to deficiencies in both dorsal and anterior fates (Acampora et al., 1995; Ang et al., 1996; Matsuo et al., 1995; Shawlot and Behringer, 1995; Waldrip et al., 1998). During murine development, formation of anterior structures requires function of the components of the Nodal signaling pathway, Nodal and Smad2, in the extraembryonic AVE (Varlet et al., 1997; Waldrip et al., 1998). While boz homologs have not been identified in mammalian embryos, it is worth noting that boz is required for expression of axial (HNF3-β homolog), lim1 and otx2, and is also likely to act upstream of TGF-β signals like Nodal-related proteins. Therefore, it is tempting to speculate an existence of a mammalian boz homolog that would regulate a conserved signaling pathway inducing anterior fates by extraembryonic structures.

We thank C. Wright, B. Hogan, B. Appel, P. Kolodziej and members of our laboratories for critical comments on the manuscript and discussions and F. Marlow for excellent technical assistance and support. β-catenin antibodies were kind gifts from R. T. Moon and P. Schneider and γ-irradiation mutagenized fish from M. Halpern. We thank D. Meyer for assistance and support in mapping as well as E. Knapik and M. Fishman for providing information on the zebrafish genomic map. cDNA clones were kindly provided by our colleagues P. Ingham, C. and B. Thisse, S. Schulte-Merker, D. Grunwald, M. Mishina, T. Jovett, N. Holder, R. Moon, H. Clevers, C. Wright, I. Dawid, E. Weinberg and C. Dhorman. The initial analysis of boz in W. D. laboratory was supported by NICHD HD29761, and subsequently by Alexander von Humboldt postdoctoral fellowship for T.-C. L. Support was obtained from an NIH postdoctoral fellowship to H. I. S. and NIH grant R01 RR12349 to W. S. T.; T. H. was supported by a Grant-in-Aid for COE Research of the Ministry of Education, Science, Sports and Culture in Japan. The work in L. S.-K. laboratory was supported by a Basil O’Connor Starter Scholar Award from March of Dimes Birth Defects Foundation and a Natural Sciences grant from Vanderbilt University and K. D. F. was supported by a NIH training grant.