ABSTRACT
A bone morphogenetic protein (BMP) signaling pathway acts in the establishment of the dorsoventral axis of the vertebrate embryo. Here we demonstrate the genetic requirement for two different Bmp ligand subclass genes for dorsoventral pattern formation of the zebrafish embryo. From the relative efficiencies observed in Bmp ligand rescue experiments, conserved chromosomal synteny, and isolation of the zebrafish bmp7 gene, we determined that the strongly dorsalized snailhouse mutant phenotype is caused by a mutation in the bmp7 gene. We show that the original snailhouse allele is a hypomorphic mutation and we identify a snailhouse/bmp7 null mutant. We demonstrate that the snailhouse/bmp7 null mutant phenotype is identical to the presumptive null mutant phenotype of the strongest dorsalized zebrafish mutant swirl/bmp2b, revealing equivalent genetic roles for these two Bmp ligands. Double mutant snailhouse/bmp7; swirl/bmp2b embryos do not exhibit additional or stronger dorsalized phenotypes, indicating that these Bmp ligands do not function redundantly in early embryonic development. Furthermore, overexpression experiments reveal that Bmp2b and Bmp7 synergize in the ventralization of wild-type embryos through a cell-autonomous mechanism, suggesting that Bmp2b/Bmp7 heterodimers may act in vivo to specify ventral cell fates in the zebrafish embryo.
INTRODUCTION
Dorsoventral axis formation in vertebrates and invertebrates depends on a conserved signal transduction pathway mediated by secreted ligands of the TGFβ superfamily, the BMPs (reviewed in De Robertis and Sasai, 1996; Ferguson, 1996). In dorsoventral axis formation, a gradient of BMP activity is hypothesized to form along the axis and specify different cell fates in a concentration-dependent manner. In vertebrates the gradient of BMP activity is thought to be generated through the diffusion into lateral and ventral regions of dorsally expressed BMP antagonists, which can directly bind and inhibit BMP activity (reviewed in Cho and Blitz, 1998).
BMP ligands are processed from a preproprotein and dimerize to form mature ligands. Within the Bmp family there are several distinct subclasses: bmp2 and bmp4 are most closely related to Drosophila decapentaplegic (dpp), whereas bmp7, also referred to as OP-1, belongs to a bmp5,6,7 subclass and is most closely related to glass bottom boat-60A in Drosophila (reviewed in Hogan, 1996). Bmp dimers bind two distinct classes of transmembrane serine-threonine receptor kinases, type I and type II, which have been suggested to form a tetrameric receptor complex binding two Bmp dimers. This activated receptor complex phosphorylates downstream Smad proteins, which then translocate into the nucleus and regulate transcription of downstream target genes (reviewed in Massagué, 1998). Genetic analyses in the zebrafish demonstrate a requirement for the smad5, chordin and bmp2b genes in dorsoventral axis formation (Hild et al., 1999; Kishimoto et al., 1997; Nguyen et al., 1998b; Schulte-Merker et al., 1997).
The requirement for different Bmp ligands during early vertebrate embryogenesis has been difficult to address. Overexpression studies in Xenopus reveal that Bmp2, Bmp4 and Bmp7 have very similar biological potentials in the early embryo (Hawley et al., 1995; Nishimatsu and Thomsen, 1998). Due to the potential promiscuity in heterodimerization of both Bmp receptors and Bmp ligands (Hazama et al., 1995), overexpression of dominant-negative mutants for these Bmp components may display a lack of specificity and interfere with signaling by several Bmps. In the mouse a small fraction of bmp4 mutants display a lack of ventroposterior mesodermal derivatives, consistent with a role in dorsoventral patterning (Winnier et al., 1995). However, redundancy in Bmp function and/or a maternal Bmp contribution has been suggested to account for the lack of early patterning defects observed in bmp2 and bmp7 mutant mouse embryos (Dudley et al., 1995; Dudley and Robertson, 1997; Zhang and Bradley, 1996). Therefore, the characterization of zebrafish mutants of these Bmp ligands allows one to address the presently unresolved question of the contribution of different Bmp ligands to early dorsoventral patterning of a vertebrate embryo.
Here we demonstrate that the strongly dorsalized zebrafish snailhouse (snh) mutant is caused by a mutation in the bmp7 gene. We show that the null snh/bmp7 mutant phenotype is indistinguishable from the swirl (swr)/bmp2b null phenotype. Moreover, snh/bmp7; swr/bmp2b double mutant embryos do not exhibit a stronger dorsalized phenotype than the single mutants, demonstrating that bmp2b and bmp7 in the zebrafish are equally required to specify ventral cell fates. Furthermore, coinjection of bmp2b and bmp7 mRNAs into the same blastomere elicits a much stronger ventralizing activity than injection of each into adjacent blastomeres, suggesting that Snh/Bmp7 and Swr/Bmp2b might function through intracellular heterodimer formation.
MATERIALS AND METHODS
Zebrafish bmp7 homologue isolation
RT-PCR on gastrula stage RNA was performed as previously described (Connors et al., 1999) with the primers CT(G/T)GGNTGGCA(G/A)GACTGGAT and CC(A/G)CA(G/T)-GC(C/T)(C/T)GNACCACCAT. The amplified fragment was used to screen a blastula/gastrula stage zebrafish cDNA library (gift from Thiery Lepage) to isolate a complete bmp7 cDNA.
Whole-mount in situ hybridization
The bmp7 coding sequence was subcloned into the EcoRI/XbaI sites of pBSKII(+). This plasmid was linearized with KpnI and antisense RNA synthesized using T3 RNA-Polymerase. Whole-mount in situ hybridizations were performed as described by Thisse et al. (http://www-igbmc.u-strasbg.fr/zf_info/zfbook/chapt9/9.82.html).
Chromosomal mapping
Mapping was performed as in Nguyen et al. (1998b).
Cloning of snhty68a
TÜ wild-type embryos and mutant embryos from a cross of homozygous snhty68a fish were collected at bud stage. cDNA was synthesized and PCR performed as in Connors et al. (1999) with the primer pairs 7J-TGTGGACTTCTGCTTAAC and 7G-ATGACCT-TTCTGTTTGCCTC, and 7A-AGCATCAGGAATCCACCTCC and 7D-TCAGAAATCAAGTCAAGTGG to amplify the bmp7 cDNA. Two independent PCR products were subcloned into pGEM-TEasy (Promega) and sequenced.
mRNA injections
The entire ORF of bmp7 and bmp7ty68a was amplified by PCR using the primers 7H-CCGCTTCATATGAAGAAGTC and 7J-TGTGGACTTCTGCTTAACTC, subcloned into pGEM-T and sequenced. We subcloned these inserts into the StuI site of pCS2+. pCS2+bmp7 and pCS2+bmp7ty68a were linearized with ApaI and mRNA synthesized and injected, as described (Nguyen et al., 1998b).
RESULTS
snailhouse mutant phenotypes
The strongest class of dorsalized mutant defects (C5) identified in large-scale mutagenesis screens in the zebrafish is caused by recessive swr/bmp2b mutations and the dominant maternal somitabun (sbn)/smad5 mutation (Hild et al., 1999; Mullins et al., 1996; Nguyen et al., 1998b). In mutant embryos of these genes, dorsolateral derivatives are expanded and ventrally encircle the embryo, while ventral cell fates are strongly reduced or absent. The recessive mutation snh causes the second strongest class of dorsalized phenotypes (C4). snh mutant embryos are morphologically distinguishable as distinctly ovoid shaped embryos just prior to bud stage (Fig. 1A,B). During segmentation stages, the somites extend from their normal dorsal position to ventrolateral regions (Fig. 1D,E). Typically snhty68a mutant embryos exhibit a less severe dorsalization than swr/bmp2b mutants and display a severe posterior truncation at 1 day post-fertilisation (d.p.f.) with the trunk twisting around its axis in a snail shell-like fashion (Fig. 1G). Occasionally snhty68a mutant embryos die around the 14-somite stage, with phenotypes similar to swr/bmp2b and sbn/smad5 mutant embryos.
aubergine (aub) is a dorsalized zebrafish mutant, which arose spontaneously in a wild-type fish stock. This zygotic, recessive mutation gives rise to strongly dorsalized C5 phenotypes, resulting in the extreme elongation (Fig. 1A,C), circumferential somites (Fig. 1F) and ultimate death of all mutants by the 15-somite stage. The dorsalization observed in aub is more severe than the one typically observed in snhty68a (Fig. 1E,F) and appears indistinguishable from swr/bmp2b mutants. Microinjection of synthetic mRNAs encoding bmp2b or bmp4 (Nikaido et al., 1997) rescues the aub mutant phenotype (data not shown), suggesting that this locus encodes a secreted Bmp ligand, rather than a downstream component of the signaling pathway. Sequencing of the bmp2b and bmp4 transcripts from aub mutant embryos revealed, however, no mutational alteration.
Since similar conclusions had been reached for snh (Nguyen et al., 1998b), we addressed whether aub and snh are allelic to each other. In crosses between heterozygous fish of these two mutations, aub failed to complement snhty68a in six independent crosses (22% of the embryos showed a C4 or C5 phenotype, 172/620). We therefore consider aub a stronger allele of snh (as further confirmed by mapping data below) and rename it snhsb1aub.
snailhouse maps to LG11 and is syntenic to bmp7 in the mouse
To identify potential candidate genes for the snh mutations, we mapped snh to a chromosomal location using SSLP markers (Knapik et al., 1998). In pools of mutant and sibling embryos in the F2 generation of a mapping cross, we found that the SSLP marker Z6909 on LG11 was linked to the snh mutation (Fig. 2A). In analyzing 150 snhty68a chromosomes, we identified 11 recombinants, placing snhty68a 7.3 cM from marker Z6909 (Fig. 2A). Examination of three additional SSLP markers shows that snh is located proximal to Z6909 on LG11 (data not shown). In our snhsb1aub line, we found SSLP marker Z13395 on LG11, located 13 cM proximal to marker Z6909 (Fig. 2A), to be 6 cM (5 recombinants in 84 meioses) from the snhsb1aub mutation. Analysis of an additional marker distal to Z6909 places the snhsb1aub mutation at the same position as snhty68a, consistent with these mutations being allelic to each other.
In zebrafish, the snail1 gene also maps to LG11 (Postlethwait et al., 1998) and we found it to be approximately 16 cM telomeric to the snhty68a mutation (Fig. 2A, data not shown). Since we previously observed that snhty68a mutant embryos are most efficiently rescued by Xbmp7 overexpression, as opposed to bmp2b and bmp4 (Nguyen et al., 1998b), we examined the possibility of conserved chromosomal synteny between the snail1 gene and the bmp5,6,7 subclass of Bmps in the mouse and human. Interestingly, the mouse bmp7 gene maps 5 cM from the snail homologue (www.informatics.jax.org/searches/quick_gene_report). Given the previously observed conserved synteny between mouse, human and zebrafish species (Postlethwait et al., 1998) and our previous rescue results, we reasoned that bmp7 was an excellent candidate for the snh gene.
Isolation of a zebrafish bmp7 homologue
By RT-PCR with degenerate primers and cDNA library screening we isolated a zebrafish bmp7 cDNA. The encoded protein belongs to the BMP7/OP1 subgroup of TGFβs and is most closely related to Xenopus XBMP7 (67.8% identical; 75.1% similar amino acids), and mouse BMP7/OP1 (66.2% identity; 73.9% similarity), but is slightly less related to Xenopus XBMP7R (57.2% identity; 64.8% similarity) (Fig. 2B). A predicted 26-amino-acid signal peptide for secretion is followed by a 261-amino-acid prodomain and a conserved BMP maturation cleavage site (RSVR) (Fig. 2B). The 145 C-terminal amino acids of the presumptive mature Bmp7 peptide contain the seven invariant cysteine residues necessary for dimerization and formation of the characteristic TGFβ cysteine-knot structure.
snailhouse is a mutation in bmp7
To investigate whether the bmp7 gene corresponds to snh, we examined linkage of bmp7 to the snh mutation. We identified an RFLP in the 3′UTR of bmp7 in a snhty68a mapping cross line and followed the segregation of the RFLP in individual F2 mutant and wild-type sibling embryos (Fig. 2C). In 162 meioses we always observed segregation of the polymorphism with the mutation (data not shown), placing the snhty68a mutation within 0.6 cM of the bmp7 gene.
To determine if a mutation in the bmp7 gene causes the snh mutant phenotype, we cloned the bmp7 cDNA from snhty68a mutant embryos. The sequence of the snhty68a allele revealed a missense mutation in the prodomain, changing valine 130 to a glycine residue (Fig. 2D). To confirm that this mutation is the cause of the snhty68a phenotype, we assayed the ability of the mutant bmp7 allele (bmp7ty68a) to rescue the snhty68a phenotype. We injected 50 pg of bmp7ty68a mRNA into snhty68a mutant embryos derived from homozygous adult parents (previously rescued to viable adults, as described in Nguyen et al., 1998b) and observed only partial rescue of the mutants to a slightly weaker phenotype (Fig. 2E). In contrast, 50 pg of the wild-type bmp7 allele rescues 99% of the snhty68a mutant embryos to a wild-type or very weakly dorsalized (C1) phenotype (Fig. 2E). At tenfold higher concentrations (500 pg), however, bmp7ty68a mRNA can rescue snhty68a mutants to less severely dorsalized phenotypes and occasionally to a wild-type phenotype (Fig. 2E). This demonstrates that the valine residue at position 130 in the Bmp7 protein is essential for its proper function, although the mutant protein still displays some weak activity. This mutation may affect the stability, secretion or maturation of the ligand.
Absence of bmp7 transcript suggests that snhsb1aub is a bmp7 null mutant
We next pursued the identification of the defect responsible for the stronger snhsb1aub allele. Attempts to amplify the bmp7 cDNA from snhsb1aub mutant embryos by RT-PCR, however, repeatedly failed, despite our ability to isolate both bmp2b and bmp4 from snhsb1aub mutants (data not shown). Moreover, no bmp7 expression is detectable by in situ hybridization in snhsb1aub mutant embryos at any stage of development examined, in contrast to the equally severely dorsalized swr/bmp2b mutant (see below; Fig. 5L,O; data not shown). Thus, no bmp7 transcripts are detectable in snhsb1aub mutants, indicating that snhsb1aub represents a null mutation of the bmp7 gene.
We show by Southern blot analysis probing with the 3′UTR fragment of the bmp7 cDNA that the bmp7 locus is altered in the snhsb1aub mutant. With all restriction enzymes tested, we observed two different size restriction fragments for homozygous mutant and homozygous wild-type DNA, and the presence of both bands from DNA of heterozygous fish (Fig. 3A). Considering the altered genomic structure of this gene, it is possible that the absence of the bmp7 transcript in snhsb1aub mutant embryos is due to an insertion or local genomic rearrangement.
We investigated whether snhsb1aub mutants could be rescued to homozygous adults, as we had previously found with the hypomorph snhty68a (Nguyen et al., 1998b). Indeed, injection of 60 pg of bmp7/snh mRNA rescues a substantial fraction of the mutants to a wild-type or very weakly dorsalized phenotype (60%, Fig. 2E). Furthermore, we could raise the majority of rescued snhsb1aub homozygotes to viable, fertile adults. Like snhty68a, snhsb1aub homozygous females do not exhibit a maternal-effect mutant phenotype, indicating no significant maternal contribution of Bmp7/Snh to embryonic development. These results together with the non-complementation and mapping data confirm that the dorsalized phenotype of snhsb1aub is due to the specific loss of Bmp7 activity.
Generation of bmp7 deficiencies
Since the snhty68a allele is a hypomorph, and the precise molecular lesion responsible for the potential null allele snhsb1aub is not completely characterized, we undertook the isolation of gamma ray-induced deficiencies of snh/bmp7. Gamma rays typically cause deletions and translocations (Fritz et al., 1996), which can be valuable in analyzing the null phenotype of a gene. We performed a non-complementation screen between homozygous snhty68a females and males generated from gamma-ray irradiated sperm. We screened 205 males and identified two males that failed to complement the snh/bmp7 mutation to generate the lines, Df(LG11)snhp11 and Df(LG11)snhp15. Similar to other gamma ray-induced mutations, the Df(LG11)snhp11 and Df(LG11)snhp15 mutations did not segregate in a Mendelian manner. Crosses between homozygous snhty68a females and Df(LG11)snhp11 or Df(LG11)snhp15 males produced 1-25% mutant embryos, instead of the 50% expected. Intercrosses of the Df(LG11)snhp11 and Df(LG11)snhp15 lines generated broods with 1-12% mutant embryos.
To determine if the bmp7 gene was deleted, we performed PCR with bmp7-specific primers on DNA from Df(LG11)snhp11 and Df(LG11)snhp15 mutant embryos. Mutant DNA never yielded a 3′UTR PCR amplification product of bmp7, whereas wild-type sibling DNA always did (Fig. 3B). Primers specific to tolloid served as a positive control for the quality of the mutant DNA (Fig. 3B). Thus, the bmp7 3′UTR is deleted in both lines.
To determine the extent of the deletions we examined SSLP markers surrounding the bmp7 locus. The markers Z8214, Z13411 and Z6909 distal to snh/bmp7 failed to produce a PCR amplification product from mutant DNA, whereas sibling DNA did (Fig. 3C). We amplified the next proximal marker to snh, Z13395, from mutant DNA, indicating that the deficiency does not extend beyond this marker (Fig. 3C). Thus, the deficiency includes the region from snh/bmp7 to at least marker Z8214, encompassing a minimum of 15.4 cM (Fig. 2A).
Expression of bmp7 during embryogenesis
We next studied the embryonic expression of bmp7 to investigate whether its expression domains are compatible with a proposed role in the early specification of ventral cell fates.
Both in situ hybridization and RT-PCR failed to detect any maternally deposited bmp7 RNA (data not shown), consistent with the lack of a maternal-effect phenotype of homozygous snh females. bmp7 expression first becomes detectable uniformly throughout the blastoderm shortly after the activation of the zygotic genome (data not shown). At 30% epiboly, transcripts start to be excluded from the presumptive dorsalmost region of the embryo (Fig. 4A) and this restriction of bmp7 expression to ventrolateral regions becomes more pronounced at the beginning of gastrulation (Fig. 4B,C). At 60% epiboly, bmp7 is expressed in the ventral half of the gastrula (Fig. 4D,E), as well as in the yolk syncytial layer (YSL) (Fig. 4F) and weakly in the prechordal plate (Fig. 4D). The expression of bmp7 in the ventral blastula and gastrula stage embryo supports a role for this factor in the early specification of ventral cell fates. At late gastrulation stages (Fig. 4G), expression delineates the boundary of the anterior neural plate, while weak expression is observed throughout the non-neural ectoderm. Moreover, transcripts are detected in the ventral marginal zone (data not shown).
At early segmentation stages, transcripts have vanished from the border of the anterior neural plate, while very weak expression is still observed outside the neurogenic ectoderm (Fig. 4H). At this time, expression is observed at the border of the hatching gland (Fig. 4H), as well as posteriorly in the tail bud and at the border between neural and non-neural posterior ectoderm (Fig. 4I). At later somitogenesis stages, bmp7 begins to be expressed in the epiphysis, where it persists until late stages of embryogenesis (Fig. 4J). At 24 hours (h.)p.f., bmp7 transcripts are detectable in the posterior aspect of the otic vesicle (Fig. 4K). Further expression domains include the endoderm (Fig. 4J,K) and posterior dorsal neural tissue (Fig. 4L). At 48 hpf, additional expression is observed in a subset of cells in the hyoid arch as well as ectodermal cells surrounding the mouth opening (data not shown).
Regulation of bmp7 expression in dorsalized mutants
We previously showed that expression of the bmp2b and bmp4 genes is initiated in swr/bmp2b, sbn/smad5 and snhty68a/bmp7 mutant embryos, while maintenance of bmp2b and bmp4 expression is affected (Nguyen et al., 1998b). To test if bmp7 gene expression is also altered in these mutants, we examined expression of bmp7 by in situ hybridization. At sphere stage and 50% epiboly, just prior to gastrulation, bmp7 expression in snhty68a, sbn/smad5, and swr/bmp2b mutants is indistinguishable from wild-type embryos (data not shown). After the onset of gastrulation, the expression of bmp7 becomes downregulated in these mutants (Fig. 5C,F,I,L), similar to that observed with bmp2b and bmp4 expression (Fig. 5A,B,D,E,G,H,J,K). In snhty68a and sbn/smad5 mutant embryos, bmp7 expression becomes restricted to a more ventral domain (Fig. 5F,I). However, in midgastrula-stage swr/bmp2b mutant embryos, bmp7 expression is absent from the ventral blastoderm, while it is still detectable in the prechordal plate and the ventral YSL (Fig. 5L). Thus, the initiation of bmp7 expression is not perturbed in swr/bmp2b, sbn/smad5 and snh/bmp7 mutant embryos, while maintenance of bmp7 expression is affected.
In snhsb1aub mutants, a lack of ventral embryonic bmp2b and bmp4 expression at gastrulation stages is observed similar to that in swr/bmp2b mutants (Fig. 5J,M and K,N). In contrast, we did not detect bmp7 expression at any stage examined in snhsb1aub, Df(LG11)snhp11 and Df(LG11)snhp15 mutant embryos (Fig. 5O, data not shown). In these mutant embryos, unlike swr/bmp2b mutants, bmp7 expression is also absent from the YSL and the prechordal plate (Fig. 5, compare L to O, data not shown). Thus, the absence of all bmp7 expression domains in snhsb1aub, Df(LG11)snhp11 and Df(LG11)snhp15 mutant embryos is consistent with these mutations being null alleles for bmp7.
Null alleles of snh/bmp7 are indistinguishable from swr/bmp2b
To characterize and compare the phenotypic strengths of snhty68a, snhsb1aub, and deletions Df(LG11)snhp11 and Df(LG11)snhp15 to the strongest dorsalized mutant, swr/bmp2b, we examined the expression of several markers that distinguish between strong and moderately strong dorsalized phenotypes. We previously showed that all somites of swr/bmp2b mutant embryos encircle the embryo, whereas in snhty68a only somites posterior to somite 2 typically circle the embryo (Fig. 1E; Mullins et al., 1996). We examined the expression of myoD (Weinberg et al., 1996), a somitic mesodermal marker, and found that in both snhsb1aub and Df(LG11)snhp15 mutants all the somites form rings around the embryo, identical to the swr/bmp2b mutant phenotype (Fig. 6A,C-E). We found all aspects of the Df(LG11)snhp15 mutant embryos indistinguishable from snhsb1aub and therefore only show snhsb1aub here. In examining the expression of pax2.1 (Krauss et al., 1991), we observed the midbrain-hindbrain boundary domain to encircle the entire embryo in Df(LG11)snhp15, snhsb1aub and swr/bmp2b fish, whereas in snhty68a embryos it is enlarged laterally, but never encircles the embryo (Fig. 6B-E; Mullins et al., 1996). Thus, a stronger dorsalized phenotype is displayed in the deficiency and snhsb1aub embryos than in snhty68a mutants.
We previously found that specification of the cranial neural crest is very sensitive to variations in Bmp activity. In mutants with presumptive low Bmp signaling activity, cranial neural crest progenitors are expanded, while in mutants with no or very low Bmp signaling activity, cranial neural crest is greatly reduced to absent (Nguyen et al., 1998b). In snhty68a the presumptive cranial neural crest, as revealed by fkd6 expression (Odenthal and Nüsslein-Volhard, 1998), is expanded (Fig. 6F-H), consistent with the hypomorphic nature of the snhty68a allele and the presence of residual Bmp activity. In contrast, swr/bmp2b, snhsb1aub and Df(LG11)snhp15 mutant embryos display little or no fkd6-expressing neural crest progenitors (Fig. 6I,J).
Consistent with snhty68a exhibiting a weaker phenotype morphologically and bmp7ty68a mRNA overexpression providing some weak rescuing activity, our marker analysis also indicates the presence of residual Bmp7 activity in snhty68a mutant embryos. Furthermore, the identical phenotypic strengths of Df(LG11)snhp11 and Df(LG11)snhp15 to snhsb1aub mutant embryos strongly supports our hypothesis that snhsb1aub is a null allele of bmp7. The strength of the dorsalization displayed by snhsb1aub embryos appears equivalent to that of swr/bmp2b mutants, indicating that Bmp activity is reduced to the same degree in complete loss of function alleles of snh/bmp7 and presumptive null alleles of swr/bmp2b.
swr/bmp2b; snh/bmp7 double mutants do not display a stronger phenotype than single mutants
We investigated whether swr/bmp2b and snh/bmp7 display redundant functions in early embryonic development, by examining swr; snh double mutant embryos. Embryos from intercrosses of double heterozygous swrtc300a/+, snhsb1aub/+ fish did not display any detectable morphological alterations distinct from the single mutants (data not shown). We also analyzed msxB expression, a marker of dorsal neuroectodermal tissue during segmentation stages (Ekker et al., 1997) which, similar to neural crest markers (Nguyen et al., 1998b), is very sensitive to variations in the amount of Bmp signaling (V. H. Nguyen and M. C. M., unpublished). In swr/bmp2b and snh/bmp7 mutants, we observe residual msxB expression (Fig. 7A,B), which could reflect redundant Bmp7/Snh signaling still present in swr/bmp2b mutants. In mutant embryos from crosses between double heterozygous swrtc300a, snhsb1aub fish, however, we never observed a strong reduction or absence of msxB expression, which would be expected if the double mutants display reduced Bmp signaling activity compared to the single mutants (Fig. 7B, 855 total embryos examined). The lack of additional phenotypes or a stronger dorsalization in double mutant embryos suggests that bmp2b and bmp7 do not function redundantly during early development in the zebrafish.
bmp2b and bmp7 synergize in overexpression experiments
To investigate the combined versus single activities of bmp2b and bmp7, we injected mRNA for bmp2b or bmp7 either separately or in combination into the yolk of 2-cell stage wild-type zebrafish embryos. The resulting embryos were scored according to their morphology. Class I embryos display a wild-type morphology (Fig. 8A). Weakly ventralized class II embryos show a slight reduction of the head and an expansion of ventroposterior hematopoietic derivatives (Fig. 8B). Class III embryos lack all dorso-axial derivatives, including the head and notochord (Fig. 8C). The most severely affected class IV embryos acquire a spindle shape devoid of morphologically identifiable dorsoventral polarity and contain a large quantity of blood cells at their posterior extremity (Fig. 8D).
Both bmp2b and bmp7 can induce severely ventralized class III/IV phenotypes if overexpressed at sufficiently high doses (20 pg and 400 pg of mRNA, respectively; data not shown). Coinjection of both mRNAs, however, leads to a clear synergistic increase in ventralizing potency. Following the separate injection of 4 pg bmp2b or 80 pg bmp7 RNA, over 80% of the embryos still display a wild-type morphology (Fig. 8E). In contrast, the combined injection of these same doses of bmp2b and bmp7 mRNA leads to a substantial decrease in wild-type embryos (to 19.8%) and a concomitant increase in strongly ventralized class III/IV phenotypes (to 58.2%). The coinjection of 4 pg bmp2b with 80 pg bmp7 mRNA has the same effect as separate injections of either 20 pg bmp2b or 400 pg bmp7 mRNA, suggesting that the combined presence of both factors increases their ventralizing potency.
We next addressed whether the synergy observed in our coinjection experiments was specific to bmp2b and bmp7 or whether similar effects could be obtained by any two of the gastrula-expressed bmps: bmp2b, bmp4 and bmp7. We expressed combinations of bmp2b, bmp4 or bmp7, any one of which had only a very weak ventralizing effect (Fig. 8E). In combinations, however, we observed that the ventralizing potency of either bmp2b or bmp4 was strongly enhanced by coexpression of bmp7, while in contrast no synergistic interaction was detectable between bmp2b and bmp4 (Fig. 8E).
Bmp2b and Bmp7 may activate different receptors, which synergize at a downstream level. Alternatively, the cooperation between Bmp2b and Bmp7 could be mediated through more active heterodimers of these factors. In this case, synergy would depend on the coexpression of bmp2b and bmp7 in the same cell, since Bmp dimers form intracellularly. To test whether the synergistic effect could be mediated through the non-cell-autonomous interaction of Bmp2b and Bmp7, mRNAs encoding bmp2b or bmp7 together with two different fluorescent dyes were injected either separately into two neighboring blastomeres or coinjected into the same cell of 16-cell stage embryos (Fig. 8F). Observation of the clonal derivatives of the injected blastomeres at early gastrula stages revealed extensive intermingling between bmp2b-and bmp7-expressing cells (data not shown). We found that the percentage of ventralized embryos is clearly enhanced, if both mRNAs are injected into the same blastomere as opposed to separate injections into adjacent cells (Fig. 8F). The cooperation between bmp2b and bmp7 is therefore mediated through a cell-autonomous mechanism, suggesting that these two factors cooperate to specify ventral embryonic fates through heterodimer formation.
DISCUSSION
Bmp7 and Bmp2b mediate equivalent, non-redundant genetic functions in dorsoventral patterning
The strongest dorsalized phenotype isolated during zebrafish mutagenesis screens is associated with the loss of swr/bmp2b gene function. We previously showed that embryos homozygous for the snhty68a mutation display a comparatively weaker dorsalization (Mullins et al., 1996; Nguyen et al., 1998b). In this study, we report that snhty68a is a hypomorphic mutation in the zebrafish bmp7 gene. Furthermore, we show that null mutations of snh/bmp7 cause a more severe dorsalization, indistinguishable from the one observed in swr/bmp2b mutant embryos, and swr/bmp2b; snh/bmp7 double mutant embryos display a phenotype identical to that of the single mutants. Therefore, Bmp7 and Bmp2b mediate equivalent, non-redundant, ventralizing activities required for dorsoventral pattern formation of the zebrafish embryo.
Bmp signaling in early vertebrate embryos
In the mouse it has been difficult to address the roles played by various Bmps during gastrulation. Most Bmp4 and all Bmp type I receptor mouse mutants die prior to gastrulation, precluding an analysis of their roles in dorsoventral pattern formation (Mishina et al., 1995; Winnier et al., 1995). A small fraction of bmp4 mutant embryos that survive beyond this stage exhibit defects in ventroposterior tissue consistent with a role in dorsoventral patterning. The lack of gastrulation defects in bmp5, bmp6 and bmp7 mouse mutants or bmp5; bmp7 and bmp5; bmp6 double mutants may reflect possible maternal or redundant expression of these or other bmp genes, as shown for other tissues in the mouse (Solloway et al., 1998; Solloway and Robertson, 1999; Storm and Kingsley, 1996).
In Xenopus bmp2, 4 and 7 behave similarly in many misexpression and explant assays tested. Overexpression of cleavage-defective dominant-negative bmp2, 4 or 7 mutants similarly dorsalize the embryo (Hawley et al., 1995; Suzuki et al., 1997b). Although caveats exist in the interpretation of results generated from dominant-negative mutants, the data are consistent with a requirement for multiple Bmps in dorsoventral patterning, similar to our results in the zebrafish.
Bmp signaling in the fish versus the fly
While arthropods and vertebrates display an inverted body plan with respect to the dorsoventral axis, certain aspects of dorsoventral patterning have been conserved (De Robertis and Sasai, 1996; Ferguson, 1996), while other aspects appear divergent. As in the zebrafish, two TGFβ family members in the fly, dpp and screw, are required for specification of dorsal cell fates. While dpp is the Drosophila orthologue of bmp2/bmp4, screw is a distant relative of the bmp5,6,7 subclass, with no clear vertebrate orthologue (Arora et al., 1994). Null mutations of dpp eliminate all dorsal structures, while screw null mutants lack only a subset of these derivatives, revealing the non-equivalent roles of these genes, unlike the equivalent genetic roles for bmp2b/swr and bmp7/snh in the zebrafish. Moreover, dpp and screw exhibit different rescuing activities in overexpression experiments and can interact non-cell autonomously through the stimulation of different receptors (Arora et al., 1994; Neul and Ferguson, 1998; Nguyen et al., 1998a). We show that the ventralizing synergy between bmp2b and bmp7 depends on their cell-autonomous expression. Therefore, while the fly and the zebrafish both depend on two different Bmps for dorsoventral patterning, the respective roles of these genes differ in the two organisms.
Requirement of bmp2b and bmp7 in early patterning in the zebrafish
The failure of either bmp7 or bmp2b to ensure proper specification of ventral fates alone, together with the similar expression, activities and mutant phenotypes of these genes, suggests that these two factors may cooperate to achieve proper embryonic patterning. This is further supported by our finding that coexpression of bmp7 and bmp2b strongly enhances the ventralizing potential of either bmp alone. Several mechanisms by which such cooperation could be achieved are discussed below.
According to a first model, Bmp2b and Bmp7 could simply have additive, dosage effects. This mechanism may depend on a positive transcriptional autoregulatory loop, in agreement with the observed downregulation of bmp2b, 4 and 7 expression in swr/bmp2b, snh/bmp7 and sbn/smad5 mutant embryos. Insufficient Bmp signaling in the mutants could account for the disruption of a transcriptional autoregulatory loop (Kim et al., 1998; Metz et al., 1998). However, such a model would predict a strong effect following the simultaneous loss of one copy of bmp2b and one copy of bmp7. The lack of consistent severely dorsalized phenotypes in snh/bmp7; swr/bmp2b double heterozygotes (data not shown), therefore, argues against this mechanism as being the sole one for Bmp2b and Bmp7 in dorsoventral patterning.
Other possible mechanisms involve the specific requirement of either Bmp2b and Bmp7 homodimers or Bmp2b/Bmp7 heterodimers. Bmp2b and Bmp7 homodimers may independently stimulate distinct receptors required for the activation of downstream components, or both may be required for the stimulation of the same receptor tetramer, as has been suggested for Dpp and Screw in Drosophila (Neul and Ferguson, 1998; Nguyen et al., 1998a). Alternatively, Bmp2b and Bmp7 may interact by formation of heterodimers that are required for receptor activation. In several experimental systems, Bmp heterodimers display increased potency compared to homodimers (Aono et al., 1995; Israel et al., 1996; Nishimatsu and Thomsen, 1998; Suzuki et al., 1997a). Furthermore, an endogenous role for Bmp heterodimers has been suggested from the analysis of an antimorphic mutation in the human Bmp family member CDMP1 (Thomas et al., 1997). Thus, Bmp2b and Bmp7 may function through the intracellular formation of heterodimers.
In the case of heterodimer formation, cooperativity in our ventralizing assay system (Fig. 8) would depend on the coexpression of bmp2b and bmp7 in the same cell, since heterodimers form intracellularly. However, if the synergy we observe is mediated by Bmp2b and Bmp7 homodimers, then a non-cell-autonomous interaction between these factors is expected. Both swr/bmp2b (Kishimoto et al., 1997; Nguyen et al., 1998b) and snh/bmp7 (S. A. C. and M. C. M., unpublished) act in a non-cell-autonomous manner, although Bmp2b and Bmp4 may act only at short-range (Jones et al., 1996; Nikaido et al., 1999). Our finding that the synergy between bmp2b and bmp7 depends on a cell-autonomous interaction between these factors, therefore strongly suggests that Bmp2b and Bmp7 cooperate by heterodimer formation.
In conclusion, our genetic analysis shows that swr/bmp2b and snh/bmp7 play equivalent, non-redundant roles in dorsoventral pattern formation of the zebrafish embryo. Our experiments further suggest that Bmp2b and Bmp7 cooperate to specify ventral cell fates by forming heterodimers, in agreement with the fact that loss of function of either dimerization partner causes identical, strongly dorsalized mutant phenotypes.
Acknowledgments
We thank M. Hammerschmidt for communicating data prior to publication; V. Heyer and T. Steffan for technical assistance and O. Nkundwa, A. Karim, C. Miller and H. Bey-Robinson for fish care. We are grateful to M. Halpern for gamma-irradiated fish, M. Fishman and E. Knapik for SSLP marker information, J. Postlethwait for chromosomal synteny assistance, M. Moos for the CDMP1 reference, R. Dosch, M. Granato and D. Wagner for critical comments on the manuscript. B. T. and C. T. were supported from the Institut National de la Santé et de la Recherche Médicale, the Centre National de la Recherche Scientifique, the Hôpital Universitaire de Strasbourg, the Association pour la Recherche sur le Cancer and the Ligue Nationale contre le Cancer. M. F. is a stipendiate of the Ministère de l’Enseignement Supérieur et de la Recherche. This work was supported by grant RO1-GM56326 to M. C. M. and training grants T32-HD07305 and HD07516 01 to S. A. C.