dino and mercedes, two genes regulating dorsal development in the zebrafish embryo

Studies in amphibian embryos have revealed that induction and patterning of the mesoderm and neuroectoderm is a multi-step process that starts before gastrulation and requires distinct maternal and zygotic inductive activities (for review see Kimelman et al., 1992; Sive, 1993). Four different signaling activities have been described: two mesoderm inducing signals emanate from vegetal cells which themselves do not contribute to the mesoderm (Nieuwkoop, 1973). One signal from the ventrolateral vegetal hemisphere induces ventral mesoderm in most of the marginal zone, and the second signal originating from dorsal vegetal cells induces the dorsal most mesoderm. The dorsal mesoderm itself generates a third, zygotic signal that converts the initially ventrally specified mesoderm to a whole spectrum of intermediate fates. According to more recent studies, this dorsalizing activity is counteracted by a fourth, ventralizing signal from the ventral mesoderm (Köster et al., 1991; Dale et al., 1992; Jones et al., 1992; Graff et al., 1994; Maeno et al., 1994; Suzuki et al., 1994; Schmidt et al., 1995; Hawley et al., 1995; Sasai et al., 1995; Wilson and Hemmati-Brivanlou, 1995).

Both the dorsal vegetal cells of blastula embryos (the Nieuwkoop center; Gimlich and Gerhard, 1984), and the dorsal mesodermal cells of early gastrula embryos (the Spemann organizer; Spemann and Mangold, 1924), induce the formation of a secondary axis when transplanted to the ventral side of an embryo. From these experiments, two activities can be assigned to the Spemann organizer during normal development: dorsalization of the mesoderm of the ventrolateral marginal zone and neural induction in the ectoderm of the dorsal animal zone.

Several aspects of early dorsoventral patterning appear to be conserved among vertebrates. An activity similar to that of the Spemann organizer has been found in the dorsal mesoderm of early fish gastrula embryos, the shield (Luther, 1935; Ho, 1992), and in the chick and the mouse node (Kintner and Dodd, 1991; Beddington, 1994). In addition, the embryos of the different vertebrate species have similar fate maps (Kimmel et al., 1990; Hogan et al., 1994), and homologous mesodermal marker genes are expressed in similar patterns (Beddington and Smith, 1993; Kane and Warga, 1994). Thus, early patterning processes might be similarly regulated in the different vertebrates.

In order to identify genes that are required for dorsoventral patterning, we have used a genetic approach, by screening for randomly induced zygotic mutations that lead to visible embryonic phenotypes in the zebrafish (Haffter et al., 1996). Dorsalized mutants are described in the accompanying paper of Mullins et al. (1996). Here, we describe four ventralized zebrafish mutants defining two genes, dino and mercedes. The analysis of the altered expression patterns of various marker genes before, during and after gastrulation suggests that in dino and mercedes mutants, patterning processes reminiscent of those initiated by Spemann’s organizer in amphibia are perturbed.

Mutations were induced and zebrafish were maintained, mated and raised as previously described (Mullins et al., 1994). Embryos were kept at 28°C and staged according to Kimmel et al. (1995). For photography, embryos were dechorionated and mounted in 5% methyl-cellulose in E3 medium (Westerfield, 1994). Older embryos were anesthetized with 1/10 volume of 0.2% MESAB (Sigma).

Immunostainings were carried out as described elsewhere (Schulte-Merker et al., 1992) with slight modifications: the blocking, but not the incubation solution was supplemented with 100 μM biotin, the incubation solution of the secondary antibody was supplemented with 1% normal goat serum (Vector) and the secondary antibody was preincubated for 2 hours with zebrafish embryos.

In situ hybridization was carried out as described elsewhere (Schulte-Merker et al., 1992) with the following modifications: hybridization and all washes were carried out at 65°C. Washes were as follows: 1× 20 minutes in 5× SSCT, 50% formamide; 2× 20 minutes in 2× SSCT, 50% formamide; 1x 10 minutes in 2× SSCT, 25% formamide; 1× 10 minutes in 2× SSCT; 2× 30 minutes in 0.2× SSCT; 1× 10 minutes in PBST. Stained embryos were fixed for 2-6 hours at room temperature in 4% paraformaldehyde/PBS, transferred to methanol and stored at 4°C in benzylbenzoate/benzylalcohol 2:1. For a better visualization of the morphological structures, in some cases embryos were transferred to xylene and photographed in Permount (Fisher Scientific).

Acridine orange is a membrane permeable aromatic derivative that becomes fluorescent inside acidic lysosomal vesicles, thereby preferentially staining dying cells (Abrams et al., 1993).

Live embryos were dechorionated and incubated in 2 mg/ml acridine orange in E3 medium for 1 hour at room temperature. After 2 washes, staining was analyzed under fluorescent light and images were captured using a cooled CCD video camera (Photometrics LTD). As a positive control, embryos homozygous for a mutation causing general necrosis were stained in parallel. These embryos always showed a strong and ubiquitous staining, whereas in wild-type embryos only very few isolated cells were positive (data not shown).

The plasmid pSP64T-xBMP-4 (Dale et al., 1992; Jones et al., 1992) was linearized with XbaI, and capped mRNA was synthesized and injected into wild-type embryos at the 1-4 cell stage (approximately 25 pg per embryo) as described by Hammerschmidt et al. (1996).

Four mutants with a ventralized phenotype were isolated that define two genes, dino and mercedes. dino (din) mutant embryos have a small head and a large tail (Fig. 1B,M,Z). mercedes (mes) embryos display a characteristic duplication of the ventral tail fin (Fig. 1Q, see also accompanying paper of van Eeden et al., 1996). For both genes, two independent alleles were isolated. The dino alleles din^tm84 and din^tt250 are both recessive lethal and of similar strength. The mercedes alleles are both recessive viable. The embryonic phenotype of mes^tm305 is somewhat stronger than that of mes^tz209. mes^tm305 maintains its tail fin phenotype to adulthood, while homozygous mes^tz209 adults appear normal.

Intercrosses of din and mes heterozygotes gave only a partial complementation: a fraction of the embryos exhibited a partial ventral tail fin duplication which was weaker than the phenotype of homozygous mes mutants (see below, Fig. 1T). The fraction of these embryos was higher when the mother was din/+ (23%; 47/207; 3 crosses) than when it was mes/+ (6%; 25/391; 3 crosses). This phenotype was not observed when din/+ females were crossed to wild-type males (0/139; 2 crosses). All of the mutant embryos with the partial ventral tail fin duplication that were grown to adulthood were shown to be din/+, mes/+ double heterozygotes: intercrosses lead to a segregation of din and mes which is typical for 2 factor crosses, with a new double mutant category in the expected proportion (7.7%; 54/699; 4 crosses (expected 6.25%)). The double mutants could be clearly distinguished from din and mes single mutants because of their stronger phenotype (see below). The segregation pattern of din and mes in these crosses suggests that they are not partially complementing alleles of one gene. This conclusion is further supported by seg-regation analysis of outcrosses from din/+, mes/+ transheterozygotes, which shows that din and mes mutations reside at unlinked loci (data not shown).

The partial non-complementation and the observation that the same embryonic structures are affected in both mutants suggests that din and mes are involved in the same patterning processes. The observation that the penetrance of the partial ventral tail fin duplication in din/+, mes/+ trans-heterozygotes is stronger when the mother is heterozygous for din suggests that din may also have a maternal function. Nevertheless, conclusive evidence will require the generation of embryos from a homozygous din mutant female germ line. In the case of mes, homozygous mes females can be grown to adulthood and tested for maternal effects. When homozygous mes females were crossed with homozygous or heterozygous mes males, the homozygous mutant offspring showed, on average, a stronger phenotype than in crosses between heterozygous parents (see below, Fig. 1V,W), while the mes/+ embryos were normal. Thus, both mes and din may have a maternal function in addition to their zygotic function.

Morphological differences between din mutant embryos and their wild-type siblings become obvious at the tailbud stage (Fig. 1A,B). The head region of din mutant embryos appears much thinner, while the posterior region is broader and thicker than in wild-type embryos (Fig. 1A-D). The tailbud emerges in a more anterior position than in wild-type embryos, and many more cells are found ventral to the tailbud. Although the head region is much smaller, the prechordal plate or pillow, where the hatching gland cell precursors are located (Kimmel et al., 1990), appears normal (Fig. 1A,B).

There is an alteration in the size of somites along the axis of din mutant embryos. The first somites in the anterior trunk region are smaller than those in wild-type embryos. In an optical cross section through the 3rd somite of a 3-somite stage din embryo, the somites are about 20% of the normal size (Fig. 1E,F). However, the size of somites progressively increases along the anteroposterior axis. Somites 6-10 are of similar size to those of wild-type embryos, while more posterior somites are larger (Fig. 1I,J). Posterior somites are often fused in the dorsal midline (Fig. 1G,H).

Disruptions in notochord development are also regularly seen in din mutants. In most cases, the notochord is of normal size in the anterior part of the body axis and absent in the posterior part of the trunk, beginning at somites 10-20 (Fig. 1C,D). In some mutant embryos, the notochord is continuous and goes all the way to the tailbud, however, it is thinner in the posterior region.

A second quite variable feature of din mutants is the size of the head in older embryos and larvae. All mutant embryos have a much smaller head with smaller eyes at 24 to 36 hours post-fertilization (Fig. 1K,M). In some mutants, however, the head phenotype becomes progressively weaker at later stages, and eventually, an almost normally sized head is formed (Fig. 1N). The spinal cord is of about normal size in the trunk (Fig. 1E,F) but is clearly smaller in the tail of all mutant embryos (Fig. 1I,J).

More consistent are the later phenotypes of ventral and posterior structures. Mutant embryos contain many more cells on the ventral side of the tail than wild-type embryos (Fig. 1K,M). In the region posterior to the anus, two or three parallel ventral tail fins are formed instead of one (Fig. 1P,R). Often, the fins form further distal branches leading to a tree-like appearance (not shown). A very similar ventral tail fin phenotype (Fig. 1S) is often (41%, 59/142) found in wild-type embryos injected with approximately 25 pg of synthetic mRNA encoding Xenopus Bone Morphogenetic Protein-4 (Bmp4), a signaling molecule of the TGF-β family which has been shown to have ventralizing activities in Xenopus embryos (Köster et al., 1991; Dale et al., 1992; Jones et al., 1992; Schmidt et al., 1995). Other ventral and posterior tissues are also enlarged in din mutant embryos, e.g. the blood islands, which contain many more cells (Fig. 1N), and a population of cells anterior to the anus on the ventral side of the yolk extension. These cells look brownish and form a rough surface rather than a normal epithelium (Fig. 1K,M). Acridine orange staining revealed a high degree of cell death in this region of din mutants at the 15-somite stage (Fig. 1Z), while no significant cell death was detected in the same location in 20-somite stage wild-type (Fig. 1X) and 2-somite stage din mutant embryos (not shown). No increased cell death has been found in dorsal or anterior regions of mutant embryos.

The phenotype of mercedes mutant embryos is similar to that of din mutants, but less severe. As with din mutants, mes mutant embryos can be identified morphologically at the tailbud stage, when the posterior part of the embryo is broader and thicker than in wild-type embryos (see Fig. 2C). In 1-day old mes mutant embryos, the head and the notochord appear more or less normal, and the sizes of the somites are only slightly altered, while significantly more cells are found in the blood islands and on the ventral side of the yolk extension than in wild-type embryos (Fig. 1K,L). As in din mutants, apoptotic cells are found on the ventral side of the yolk extension (Fig. 1Y). At later stages of development, two parallel ventral tail fins are formed (Fig. 1Q). The duplication of the ventral tail fin begins in the region just posterior to the anus and can extend to the tip of the tail (Fig. 1V,W). The fraction of mes homozygous embryos with a complete ventral fin duplication (extending to the tail tip) is greater when the mothers are mes/mes homozygous than when the mothers are mes/+ het-erozygotes (from mes^tm305/mes^tz209 mothers: 38% of mutant embryos, 255/673, 10 crosses; from mes^tm305/+ mothers: 24% of mutant embryos, 143/603, 16 crosses; note that mes^tm305 is the stronger mes allele). In the din/+; mes/+ double heterozy-gous embryos mentioned above, the ventral tail phenotype is weaker. The two ventral branches are narrower than in homozygous mes embryos, and the tail fin duplication often is just visible at its posterior boundary. Fig. 1T shows the embryo with the strongest phenotype that was observed among the analyzed 47 din/+,mes/+ double heterozygotes.

The din/din, mes/mes double mutants show a significantly stronger phenotype than din single mutants over the entire length of the body axis, with even more blood cells and multiple ventral tail fin branches (up to 8) in the posterior region of the embryo, and a strong and persistent loss of head structures (Fig. 1O). As in din single mutants, however, the hatching glands and the anterior part of the notochord are normal.

In order to examine whether the altered organization of the body axis of dino and mercedes mutant embryos is preceded by a shift of positional information, we performed in situ hybridizations using marker genes which are expressed in a dorsoventral-specific manner before and during gastrulation.

eve1, a zebrafish homeobox gene related to the Drosophila pair rule gene even skipped (Joly et al., 1993), is expressed in ventrolateral marginal cells. In wild-type embryos, eve1 expression becomes progressively restricted to the ventral half of the embryo, while in din and mes mutants, it maintains a broader expression, which is first apparent at 40% epiboly (Fig. 2A) and more obvious at 70% epiboly, when the eve1 expression domain spans about 180° of the margin in wild type, 240° in mes mutant, and 320° in din mutant embryos (Fig. 2B). At the end of gastrulation, the eve1-positive cells are located in the posterior region of the embryo, which in the mutants is strongly enlarged (Fig. 2C for mes; compare with Fig. 1D for din). Similarly, the hematopoietic transcription factor gene, gata2 (Detrich et al., 1995), which is normally transiently expressed in the ventral half of the animal region is expressed in a much broader and stronger fashion in the mutants (Fig. 2D,E). Together, the eve1 and gata2 expression patterns indicate that in both the marginal and the animal zone of din mutant embryos, ventral fates expand into more dorsal regions before and during gastrulation. Taken together, analysis of paraxial and axial markers (sna1, sonic hedgehog and goosecoid) indicates that dorsal cell fates are reduced.

sna1, a zebrafish homologue of Drosophila snail, is expressed in involuting cells in ventrolateral regions of the germ ring and in the presomitic mesoderm (Hammerschmidt and Nüsslein-Volhard, 1993). In din mutants at 80% epiboly (Fig. 2F), the paraxial expression domains of sna1 are slightly smaller and shifted to more dorsal regions.

sonic hedgehog (shh) is normally expressed in the form of a triangle in the presumptive notochordal cells of the axial mesoderm at 70% epiboly (Krauss et al., 1993). In din mutants, the shh expression domain is of normal width in the anterior part of the notochord anlage, but becomes progressively narrower than in wild-type embryos in more posterior regions (Fig. 2G), while in mes mutants, no significant alteration in the shh expression pattern has been observed (not shown).

goosecoid (gsc), which is expressed on the dorsal side in early involuting cells of the presumptive prechordal plate (Stachel et al., 1993; Schulte-Merker et al., 1994) is expressed normally in most din mutants (13/15) before and during gas-trulation. Only very few cases (Fig. 2H; also stained for eve1 mRNA to distinguish wild-type and mutant embryos) show a slight reduction of gsc expression. Taken together, the expression patterns of gsc and shh in din mutants show that the loss of dorsal fates is very mild in the early formed anterior mesoderm, but more pronounced in the later formed posterior mesoderm.

The reduction of dorsal anlagen in the presumptive ectoderm of the animal zone is already apparent at the onset of gastrulation, coincident with the expansion of the ventral marginal fates. The forkhead domain gene fkd3, which marks presumptive neuroectodermal cells (J. O., unpublished data) is expressed in a much narrower and weaker fashion on the dorsal side in din mutants at the shield stage (Fig. 2J) and at 70% epiboly (Fig. 2K). mes mutants also display a significant, but less severe reduction in size of the fkd3 expression domain (Fig. 2K). For both mutants, the reduction is strongest in anterior and lateral regions of the fkd3 expression domain, the regions where gata2 is ectopically expressed. At later stages of development, the reduction of the neural tube is anticipated and reflected by the expression of many marker genes, such as pax2, which shows that the number of a specific class of interneurons (Mikkola et al., 1992) is reduced to about 50% of the wild-type level in the spinal cord of din mutants (Fig. 3E).

Cells from the different marginal positions of the late blastula zebrafish embryo give rise to different mesodermal tissues (Kimmel et al., 1990). Thus, the described shift in the expression of the early marker genes like eve1 in dino mutants might lead to a shift in the relative sizes of the different meso-dermal tissues at later stages of development. To examine this notion, we studied the expression pattern of various tissue specific genes after gastrulation.

As markers for the various dorsally and ventrolaterally derived mesodermal tissues (see Fig. 4), we used fkd2 for the hatching gland cells (J. O., unpublished data), no tail (ntl) for the notochord (Schulte-Merker et al., 1992), myoD for the somitic muscle (Weinberg et al., 1996), pax2 for the pronephros (formerly called paxb; Krauss et al., 1991) and gata1 (Detrich et al., 1995) for the blood.

Immunostaining of the protein encoded by the forkhead domain gene fkd2 (J. O. and R. M. W., unpublished data) revealed that the number of presumptive hatching gland cells, derivatives of the prechordal plate, is normal in most din mutants (8/10) and only slightly reduced (150 versus 180) in the strongest cases (2/10; Fig. 3A), consistent with the normal expression pattern of gsc during gastrulation (Fig. 2H).

Using anti-Ntl immunostaining, we have also quantified the number of notochord cells. A normal number of notochord cells is found in the anterior part of din mutant embryos (Fig. 3B). However, the number of notochord cells in the posterior of the body axis, beginning between somites 5 and 10, is reduced. Strong din mutants do not have any ntl-positive notochord cells in the posterior part of the midline (Fig. 3B), and even in din mutants with very weak notochord phenotypes, the number of posterior ntl-positive cells is reduced to about 50% of the number in wild-type embryos (not shown).

The anterior somites of dino mutants contain much fewer myotomal cells, as revealed by the expression pattern of myoD (Fig. 3C). In contrast, posterior somites are larger than in wild-type embryos and are fused in the dorsal midline (Fig. 3C).

A similar enlargement has been found for the expression domain of genes marking ventrally derived mesodermal tissues such as the blood and the pronephros: gata1 which is expressed in presumptive hematopoietic cells in two lateral stripes in the posterior part of the trunk, shows a twofold broader expression domain in din mutants at the 10-somite stage (Fig. 3D). At 22h of development, just before the presumptive blood cells start to migrate anteriorly (Detrich et al., 1995), many more gata1-positive cells can be detected in the region of the anus in din mutants, indicating the presence of more blood cells. Similarly, the posterior part of the pronephros is enlarged in din mutants, as revealed by the expression pattern of pax2 (Fig. 3E). mercedes embryos show a gata1 expression pattern similar to that of din mutant embryos. However, consistent with the weaker phenotype, the expansion of the gata1 expression domain is not as strong (Fig. 3D).

We have described two genes that are required for the dorsalization of early zebrafish embryos, dino and mercedes. Mutants in din and mes display an expansion of ventrolateral fates at the expense of dorsolateral fates in both the ectoderm from the animal zone and the mesendoderm from the marginal zone of the late blastula embryo (see fate map in Fig. 4A). In the ectoderm of both mes and din mutants, the population of ventral tail fin cells (ventral derivatives) is strongly enlarged, while the dorsally derived neuroectoderm is reduced. Consis-tent alterations are also found in the mesoderm: mutants show an increased amount of blood, and an enlargement of pronephros and posterior somites (ventrolateral derivatives), while the anterior somites and the posterior notochord (dorso-lateral derivatives) are reduced. The notion that both din and mes mutant embryos are ventralized is supported by the finding that wild-type embryos overexpressing the ventralizing signal Bmp4 display a similar enlargement of ventral and posterior structures, including the characteristic duplication of the ventral tail fin.

The derivatives of the dorsal most mesendoderm, the anterior notochord and the hatching glands, are unaffected in the mutants. These structures are derived from the shield, the fish equivalent of the Spemann organizer in amphibia (see intro-duction and Fig. 4B). As illustrated in Fig. 4C, the altered organization of both mesodermal and ectodermal structures in din and mes mutant embryos could be explained by a reduction in dorsalizing signaling from cells of the Spemann organizer at late blastula and early gastrula stages. Alternatively, it is possible that the phenotype is caused by gain-of-function mutations in counteracting, ventralizing components. These alternative possibilities will be addressed by segregation analysis of ventralizing and dorsalizing mutations.

Dorsoventral pattern is set up before the onset of gastrulation, as evident from a comparison of the eve1 and fkd3 expression. In late blastula stages, eve1 is expressed in most of the marginal zone, excluding only the dorsal most cells, the region of the presumptive organizer. During further development, still before the onset of gastrulation, eve1 expression normally becomes progressively restricted to more ventral regions of the margin. This suggests that most of the marginal zone of pregastrula embryos is initially ventrally specified, and that the establishment of intermediate mesodermal fates is accomplished in a later process, probably driven by dorsalizing signals spreading from the shield. At the same time as the ventral retraction of the eve1 expression in the marginal zone, the expression of fkd3 is initiated in presumptive neuroectodermal cells in an animal region directly adjacent to the shield, suggesting that neural induction in the ectoderm is initiated simultaneously with the dorsalization of the marginal zone. In din and mes mutants, the repression of the eve1 expression and the activation of the fkd3 expression are impaired to the same extent, leading to a perfectly complementary image of their expression domains in an animal view in both wild-type and mutant embryos. This indicates that din and mes are required for both the dorsalization of the mesoderm and the induction and patterning of the neuroectoderm. The phenotype of the mutants suggests that these two inductions are achieved by one and the same process.

mes mutant embryos display a generally weaker phenotype than din mutants. Even in the stronger din mutants, the dorsalization of the ventrolateral mesoderm and the neural induction is not completely abolished, as illustrated in Fig. 4C. This could have several reasons. The remaining induction could result from maternally supplied gene products. For both din and mes, a maternal function is suggested by the observation that the mutant phenotypes are stronger in a genetic situation in which the maternal contribution is reduced. Alternatively, the isolated alleles might be hypomorphic and retain some zygotic activity. Finally, the remaining induction could be accomplished by other signals of the dorsal mesoderm that are partially redundant to din or mes. din and mes themselves may be partially redundant, and other genes might have been identified but not yet recognized to be involved in the dorsalizing process because their corresponding mutant pheno-types are different from those of din and mes mutants.

During the course of development, the morphogenetic processes are affected. Maternal Drosophila mutants with an altered Bicoid gradient, for example, display severe fate shifts along the anteroposterior axis during gastru-lation but may show less pronounced segmentation defects in the larval stages or even be normal (Driever and Nüsslein-Volhard, 1988). Similarly, mutations in the ventral-specific Drosophila genes twist and snail display a completely penetrant dominant reduction of the mesodermal anlage, but are fully viable (C. Nüsslein-Volhard, unpublished observations).

In addition to the ventralized mutants dino and mercedes described in this paper, several mutations defining at least 6 genes, snailhouse, swirl, somitabun, piggy tail, lost-a-fin and mini fin have been identified that cause a dorsalized phenotype complementary to that of din and mes mutants (Mullins et al., 1996). The expression of eve1 is strongly reduced at late blastula stages, and the ventrolateral mesoderm is dorsalized, while more neuroectoderm is formed. These mutants might be defective in ventralizing activities which normally counteract dorsalizing activitives like those encoded by the genes dino and mercedes.

A similar view has emerged for the mechanisms regulating dorsoventral pattern formation in Xenopus and Drosophila embryos. Although not confirmed in a loss-of-function situation, dorsalizing activities like those assigned to dino have morphological consequences of the early pattern defects of din and mes mutants become progressively weaker. Many din mutant embryos develop a relatively normal head at later stages of development, and mes mutants are even viable. Such a progressive weakening of the phenotypic strength has been observed in other mutants in which early been found for three Xenopus genes: noggin (Smith and Harland, 1992; Knecht et al., 1995), chordin (Sasai et al., 1994) and follistatin (Hemmati-Brivanlou et al., 1994; Kessler and Melton, 1995, D. A. Melton, personal communication). All three genes encode signaling molecules produced by cells of the Spemann organizer, and all dorsalize the ventrolateral mesoderm and induce neuroectoderm in ectopic expression experiments. In addition, Chordin has been reported to counteract and repress Bmp4 (Sasai et al., 1995), a ventralizing signal generated by the ventral mesoderm (Köster et al., 1991; Dale et al., 1992; Jones et al., 1992). A similar interaction has been described for the structural and functional homologues of chordin and Bmp4, short gastrulation (sog) and decapentaplegic (dpp) in dorsoventral pattern formation of Drosophila embryos (François and Bier, 1995; Holley et al., 1995).

It is likely that similar mechanisms are at play in the zebrafish embryo. The observation that the overexpression of the ventralizing signal Bmp4 in wild-type embryos results in defects characteristic of the dino mutant phenotype suggests that Dino, similarly to Chordin, might counteract Bmp4-like activities during normal zebrafish development.

We are very grateful to Dr Andrew McMahon in whose laboratory much of the mutant analysis has been carried out. Additionally, we would like to thank Ed Sullivan for his help and advice during the setting up of a fish facility in the McMahon laboratory. Drs Eric Weinberg and Leonard Zon generously supplied us with reagents prior to publication. Published reagents were obtained from Drs Jon Graff, Jean-Stéphane Joly, Stefan Krauss and Stefan Schulte-Merker. Drs Mary Dickinson, Andrew McMahon, Siegfried Roth and Stefan Schulte-Merker read earlier versions of the manuscript.