Mutations affecting the formation of the notochord in the zebrafish, Danio rerio

The notochord is an early embryonic structure which is the source of signals important for axis formation and neural differentiation. Grafting and ablation experiments mainly performed in the chick embryo have shown that the notochord can induce ventral cell fates, such as the floor plate and motor neurons, in the neural tube (reviewed by Smith, 1993). The floor plate is a morphologically visible structure in the ventralmost part of the neural tube. In the chick, transplantation of an ectopic notochord into a region lateral to the prospective neural tube leads to the formation of an additional floor plate in the region in contact with the transplanted notochord, while removal of the notochord prevents floor plate formation (Yamada et al., 1991; Placzek et al., 1991 and references therein).

The notochord also plays a crucial role in somite patterning (reviewed by Bumcrot and McMahon, 1995). Grafting of an ectopic notochord to a position adjacent to presomitic mesoderm causes the development of extra sclerotome adjacent to the graft and prevents development of dermomyotome in the vicinity of the graft. Conversely, removal of the notochord prevents the formation of sclerotome and results in the fusion of somites, of exclusively dermomyotomal origin, beneath the neural tube in place of the notochord (Münsterberg and Lassar, 1994, and references therein). In the zebrafish, the presence of a differentiated notochord appears to be required for the induction of muscle pioneer cells (Halpern et al., 1993), a set of early differentiating muscle cells which are thought to be descendants of the adaxial mesoderm (Felsenfeld et al., 1991).

In the zebrafish, two genes that are required for notochord development, no tail (ntl) and floating head (flh), have previously been identified. Loss-of-function mutations in the flh gene cause a lack of notochord along the entire length of the embryo, with somites in the trunk fused below the neural tube (Talbot et al., 1995). flh was recently shown to be the zebrafish homologue of the Xenopus gene Xnot (Talbot et al., 1995 and references therein). This gene is expressed in the notochord and floor plate of early zebrafish embryos (Talbot et al., 1995). Expression studies performed in flh embryos suggest that cells lacking flh function differentiate into muscle rather than notochord (Halpern et al., 1995). A fate map analysis of the dorsal marginal region in gastrula stage flh mutant embryos supports this finding (Melby et al., 1996).

ntl is required for notochord and tail formation (Halpern et al., 1993; Schulte-Merker et al., 1994). Molecular analysis has shown that no tail (ntl) is the zebrafish homologue of the mouse T (Brachyury) gene, which encodes a transcription factor (Schulte-Merker et al., 1994; Kispert and Herrmann, 1993). The ntl gene is expressed in the germ ring of gastrulating embryos, in the developing notochord during somitogenesis (axial mesoderm) and in the prospective mesodermal cells of the tailbud (Schulte-Merker et al., 1992). In contrast to flh, mutant ntl embryos have undifferentiated axial mesodermal cells separating flanking somitic mesoderm (Halpern et al., 1993). The floor plate is present in ntl embryos, whereas muscle pioneer cells in the somites are not formed and the embryos do not develop a horizontal myoseptum which normally separates the dorsal and ventral parts of the somites (Halpern et al., 1993).

We have performed a systematic mutational analysis of early development in the zebrafish (Haffter et al., 1996). In this screen we have isolated alleles of ntl and flh, as well as of two new genes, mom and doc, which are also required for early notochord formation. This paper presents a preliminary characterisation of the four genes by comparing the expression pattern of various genes specifically expressed in the notochord, the floor plate and the adaxial mesoderm in mutant embryos. The mom phenotype is similar to flh, while doc resembles ntl. In both mom and flh, the formation of the floor plate is reduced, whereas in ntl it is enlarged, and in doc a normal floor plate is formed. mom and doc appear to be required predominantly in the trunk as mutant embryos exhibit a normal notochord in the posterior part of the tail. Mutations in all four genes disrupt the formation of the adaxially derived muscle pioneer cells and the horizontal myoseptum. Our analysis suggests that flh and mom have a common function in the early specification of the axial versus paraxial mesoderm, while ntl and doc are required for the development of the axial mesoderm into notochord.

Fish stocks were maintained as previously described by Mullins et al. (1994). The screen was performed as described by Haffter et al. (1996).

Antibody staining was done as described by Bernhardt et al. (1990) except that the embryos were fixed in 4% PFA in PBS and permeabilised with acetone (described by Schulte-Merker et al., 1992). The detection was performed using a peroxidase kit (Vektor) according to the manufacturer’s instructions.

In situ hybridization was performed as described by Hammerschmidt and Nüsslein-Volhard (1993).

The protocol of Ho and Kane (1990) was used for transplantation, with modifications described by van Eeden et al. (1996). About 10-15 cells from sphere stage wild-type donor embryos, previously marked with the lineage tracer dye rhodamine dextran, were transplanted to the margin of unlabeled host embryos at the same stage. The fate of the transplanted wild-type cells was scored at 24 hours of development under a compound microscope with fluorescence illumination. By transplanting labeled wild-type cells into doc recipients, offspring of donor cells were found in the region of the notochord in 30 doc recipient embryos. In 10 of these, big vacuolated notochord cells were obtained in trunk regions. 12 other transplants were not informative, because the wild-type cells were in the posterior part of the embryo where a normal appearing notochord forms in mutant embryos. The number of labeled wild-type cells in the remaining 7 mutant embryos was too large for a meaningful interpretation.

Mutations in any of the four genes floating head (flh), momo (mom), no tail (ntl) and doc lead to the failure of proper notochord formation. flh and ntl have been described before (Talbot et al., 1995; Halpern et al., 1993), whereas mom and doc are newly identified genes. Although the absence of a differentiated notochord in the trunk region is common to strong alleles of all four genes, mutations in each of the four genes display a distinct phenotype. The characteristic features of the four genes are summarized in Table 1.

We have isolated two flh alleles whose sequence alterations suggest that they are most likely to be functional null alleles (Talbot et al., 1995). The resulting phenotype is indistinguishable from that of the previously described allele flhⁿ¹, that is the notochord is completely absent in both trunk and tail. In the trunk region, the left and right somites are fused underneath the neural tube by early somite stages, whereas tail somites are separated by a large blood sinus (Fig. 1B). Discontinuous groups of floor plate cells can be seen in the ventral spinal cord (arrowhead in Fig. 1B).

Only one allele of mom was isolated in our screen. In mom mutant embryos, the trunk notochord is missing while the notochord in the tail region appears to be normal (Fig. 1C). As in flh embryos, somites are fused underneath the neural tube in the trunk of mom embryos. This phenotype is not fully penetrant and a few notochord cells can be detected in the trunk region at early somite stages of some mom embryos (Fig. 1H). As revealed by the counts of mutant embryos in individual egg lays, a variable fraction, averaging 20%, of genotypically mutant embryos even display a normal notochord and are indistinguishable from wild type. In mom embryos displaying a strong notochord phenotype, disruption of the floor plate appears more severe compared to flh, whereas in the tail tip both a floor plate and a notochord is formed (Fig. 1C). In addition to the defects in the trunk, about 15% of mom embryos also display variable head defects such as small eyes (Fig. 1J,K).

Two alleles of the ntl gene of equal strength, ntl^b160 and ntl^b195, have previously been identified and characterized (Halpern et al., 1993). In ntl mutant embryos notochord precursor cells are present in the trunk but fail to differentiate (Halpern et al., 1993; Schulte-Merker et al., 1994). In addition, the tail does not form and the notochord is completely missing from the posterior part of the embryo, resulting in fusion of the somites (Halpern et al., 1993, Fig. 1D). We have isolated three new alleles of ntl that differ in phenotype strength. In embryos with the weakest allele, ntl^ts260, small vacuolated cells in the axial region are detectable over the entire length of the embryo, and a tail is formed whose length is reduced by only one somite (Fig. 1E). ntl^tb244e is of intermediate strength; it results in failure to form vacuolated notochord cells, and mutant embryos are reduced in length by four somites (data not shown). The phenotype of ntl^tc41 mutants is stronger than that of the previously described ntl^b160 and ntl^b195 embryos. The post anal region in ntl^tc41 is shorter than in ntl^b160, even though the same number of somites are formed (Fig. 2A, and see below).

To define the molecular lesion of the strongest allele ntl^tc41, we PCR amplified and cloned genomic DNA from the ntl region of homozygous mutant embryos. The sequence analysis revealed a single nucleotide exchange (T to A transversion) changing a leucine (TTG) at position 156 into a premature stop codon (TAG). The truncated protein is shorter than the predicted product of ntl^b160 (245 amino acids) but slightly longer than that of ntl^b195 (103 amino acids plus 35 amino acids encoded by an insertion; Schulte-Merker et al., 1994). From the sequence analysis, it is not obvious why the three alleles differ in strength. The molecular nature of the other two alleles, ntl^tb244e and ntl^ts260, is so far not clear.

Three alleles of doc have been isolated and all of them cause the axial mesoderm to remain undifferentiated in most parts of the embryo (Fig. 1F). In embryos with the strongest allele, doc^tt258, vacuolated notochord cells are only formed in a few posterior segments (Fig. 1L). In embryos with the weaker allele, doc^tt202, a thin notochord forms in both the tail and the trunk (data not shown). The amount of differentiated notochord formed in the tail region is variable and depends on the strength of the allele. The phenotype is slightly variable in all three alleles, even between embryos of the same clutch.

At high temperature (29°C), doc heterozygous individuals may display a dominant phenotype, that is at day six of development they are shorter, and the notochord has an irregular shape, with smaller cells (Fig. 2C). Most of these heterozygous fish survive to adulthood and are fertile.

To gain a better understanding of the individual function of the four genes at early stages of development, we have carried out in situ hybridization and antibody staining with a number of molecular probes expressed early in the axial mesoderm of wild-type embryos (ntl, twist (twi), sonic hedgehog (shh), and axial (axl)).

Antibodies to the Ntl protein label nuclei in cells of the presumptive mesoderm of the germ ring and, later, the tailbud of wild-type embryos, as well as the axial mesoderm, which later forms the notochord (Schulte-Merker et al., 1992; Fig. 3A). During early gastrulation stages (shield stage to 60% epiboly) flh mutant embryos labeled with an antiserum against the Ntl protein are indistinguishable from wild-type embryos (data not shown). At later gastrulation stages mutant embryos exhibit persistent ntl expression in the tailbud, whereas in the midline mesoderm expression is restricted to a few cells in the posterior axis near the tailbud (Fig. 3B).

In mom embryos with a strong phenotype, Ntl protein distribution is similar to that of flh embryos. Expression is normal in cells of the presumptive tailbud mesoderm, but in the axial mesoderm Ntl-expressing cells are largely confined to posterior regions, with only a few anterior cells exhibiting reduced levels of Ntl protein (Fig. 3C). In mom embryos with a weak phenotype, Ntl expression in the notochord may be detectable over the entire length in a thin stripe, or even appears normal (data not shown). These data indicate that flh and mom are both required for ntl expression in the more anterior axial mesodermal region, but not for the early expression in the presumptive mesoderm of germ ring and tailbud, of the embryo.

We analyzed ntl expression of all ntl alleles at the RNA and protein levels in tailbud stage embryos. Ntl protein is strongly reduced in embryos mutant for the two strong alleles ntl^b160 and ntl^b195, whereas transient weak RNA expression can be observed in the tailbud and in the axial mesoderm (Schulte-Merker et al., 1994). Ntl protein is present at low levels in the notochord of mutants with the intermediate and weak alleles, ntl^tb244e and ntl^ts260. The strongest allele, ntl^tc41, results in neither Ntl protein nor RNA being detectable in the axial mesoderm (data not shown). These data confirm the morphological observations on a molecular level and suggest that ntl^tc41 is a null allele.

In doc mutant embryos Ntl expression is normal until the tailbud stage. The expression in the axial mesoderm decreases during early somite stages and it is no longer detectable in the anterior axial mesoderm by the 10-somite stage (Fig. 3H). In heterozygous doc embryos displaying a dominant phenotype (see above), Ntl staining reveals that at the 10-somite stage the nuclei in the anterior notochord cells are not arranged as regularly as in wild-type embryos (Fig. 3G). Some nuclei have lost Ntl protein completely (enlargement in Fig. 3G). These findings suggest that doc is required for the maintenance of ntl expression in the axial mesoderm in a dose dependent manner after gastrulation.

We also analyzed twist (twi), which is expressed in the axial mesoderm during gastrulation (Halpern et al 1995). twi is not expressed in ntl embryos (Fig. 3D,E). In flh and mom embryos expression of twi is only detectable in axial cells near the tailbud that also express ntl (data not shown) suggesting a requirement of ntl function for the expression of twi in the axial mesoderm. twi expression in doc embryos is indistinguishable from wild-type expression until the tailbud stage, it then decreases in the anterior notochord during early somite stages in parallel with the decreasing expression of Ntl. These data suggest that the initial twi expression is not dependent on doc function, while for its maintenance it requires doc, probably through its function in the maintenance of ntl expression.

The two genes shh (Krauss et al., 1993) and axl (Strähle et al., 1993) are expressed in both the notochord and the overlying floor plate. Similarly to the pattern of expression observed for twist, in the midline mesoderm of mutant embryos shh and axl transcripts are only detected in cells that also express Ntl. This is illustrated for shh in doc in Fig. 3J, and mom in Fig. 4C.

We conclude that flh and mom are required for the expression of ntl in the axial mesodermal region while the function of doc is to maintain this expression. The expression of other notochord markers, such as twi, shh, and axl depends on high levels of Ntl in the axial mesoderm.

We investigated a possible correlation between notochord and floor plate formation in mutant embryos. The floor plate is reduced to small patches (arrowhead in Fig. 1B) in embryos mutant for flh and mom, whereas a floor plate is formed in ntl and doc embryos (Fig. 1D-F). We also investigated the expression of the floor plate marker shh. In wild-type embryos, shh, in addition to its expression in the developing notochord, is expressed in a one-cell-wide stripe in the ventral neural tube (the midline floor plate cell; Krauss et al., 1993; Fig. 4A). At the 15-somite stage short interrupted stretches of shh expression are found in the ventral neural tube of flh embryos, although there are no notochord cells in the underlying mesoderm (Fig. 4B). In mom embryos at the 15-somite stage, shh expression in the ventral neural tube is often detectable in single dispersed cells (Fig. 4C). Double labeling with the Ntl antiserum reveals that these shh-expressing cells often, but not always, overlie residual notochord cells, expressing both shh and Ntl (arrows in Fig. 4C and data not shown). Overall, we observe that in mom the floor plate defects are stronger than in flh, despite its weaker effect on the notochord, suggesting overlapping but distinct functions for the two genes.

In ntl embryos, in contrast to flh and mom embryos, the shh-expressing domain in the ventral neural tube is enlarged. In embryos mutant for the strongest allele, ntl^tc41, shh is expressed in a two to three cell-wide stripe in the ventral neural tube (Fig. 4D, note that shh expression in the axial mesoderm is completely abolished). In contrast, only a single row of cells expresses shh in the ventral spinal cord in embryos with the weak ntl allele, ntl^ts260 (Fig. 4E), and in doc embryos (Fig. 4F). Together with the observation of a broadened domain of axial mesoderm in strong, but not weak ntl alleles (see below), this effect of ntl on the floor plate supports an involvement of the axial mesoderm in floor plate induction. This induction does not require ntl or doc (see Discussion).

Somites in mutant embryos of flh, mom, ntl and doc appear u-shaped instead of showing the normal v-shape (shown for doc in Fig. 5B). The horizontal myoseptum, which divides the somites into dorsal and ventral parts, is missing in mutants of all four genes. We analyzed the somite phenotype with an antibody against Engrailed-like proteins (4D9, or Eng), which labels the nuclei of a subset of the paraxial cells, called the muscle pioneer cells, from early somite stages on (Patel et al., 1989; Hatta et al., 1991 and Fig. 5C). Eng expression is not detectable in the trunk somites of mutant embryos of flh, mom, ntl and doc (shown for doc in Fig. 5D). However, Eng-expressing cells can be detected in regions where the notochord is present in mom and doc embryos (data not shown). These observations are in agreement with findings of Halpern et al. (1993) that a differentiated notochord is required for the development of muscle pioneer cells in the adjacent somites.

Muscle pioneer cells are thought to be derived from a single cell-wide band of cells bounding the notochord, called the adaxial cells (Felsenfeld et al., 1991). At the tailbud stage, myoD is expressed in these adaxial cells (Weinberg et al., 1996 and Fig. 6A). This adaxial myoD expression appears unaffected in flh and mom mutants. In addition, in these mutants, axial cells between the two adaxial stripes also express myoD (Halpern et al., 1995 and Fig. 6B,C). These results are consistent with findings of Melby and coworkers (1996), who have shown that the notochord fate in axial cells of flh mutants is altered into muscle fate. The similarity of myoD staining in flh and mom mutants suggests that these two genes have similar functions in the specification of the axial mesoderm.

In ntl embryos, expression of myoD is strongly reduced in the adaxial cells at the tailbud stage (Weinberg et al., 1996). In embryos with the strongest allele, ntl^tc41, expression is barely detectable in two faint stripes flanking a region of axial mesoderm that is broader than normal (arrows in Fig. 6D). Levels of myoD expression increase with decreasing strength of the ntl allele (Fig. 6E,F). These data suggest that ntl is required to induce adaxial cell development and myoD expression in the paraxial mesoderm (see Discussion). Early expression of myoD in adaxial cells of doc embryos is only slightly reduced (Fig. 6G), suggesting that doc has no, or only a weak role in the initial formation of the adaxial cells.

At later somite stages myoD continues to be expressed in adaxial cells, and it is also expressed in the posterior part of each somite (Weinberg et al., 1996 and Fig. 6H). No adaxial expression of myoD is detectable at the 15-somite stage in embryos with the strong ntl allele, ntl^tc41, while myoD expression appears normal in the somitic mesoderm (Fig. 6K). In flh and mom embryos, adaxial expression of myoD is disrupted in regions anterior to the tailbud (Fig. 6I,J), while it appears normal close to the tailbud where myoD-expressing cells of the right and left side are separated by ntl-expressing cells. This suggests that axial mesodermal cells initially present in flh and mom embryos are able to initiate adaxial myoD expression, but that adaxial development in somitic regions is not maintained in the absence of axial mesoderm. In embryos with the weak allele, ntl^ts260, and in doc, myoD expression is reduced in the anterior adaxial mesoderm during somitic stages. These observations reveal a correlation between ntl expression in the axial mesoderm and the development of adaxial cells.

In summary, we have found for all four genes that the failure to form a horizontal myoseptum correlates with the absence of the muscle pioneer cells. It can be traced back to the specification of the adaxial cells of the paraxial mesoderm. ntl expressing cells in the axial mesoderm appear to be required for the development of adaxial cells. Furthermore, development of the notochord, or a signal downstream of doc, is necessary for the development of muscle pioneer cells from a subset of the adaxial cells (see Discussion).

Genetically mosaic embryos produced by cell transplantation can give important information about the requirement and function of gene expression in particular tissues. Such chimeric embryos have been produced for ntl and flh (Halpern et al., 1993, 1995). These experiments showed that ntl and flh function are autonomously required in the notochord. In addition, it was shown that ntl function is required in the notochord for the formation of muscle pioneer cells in the somites, indicating an inductive signal from the notochord (Halpern et al., 1993).

We have carried out similar experiments for doc. Labeled wild-type cells were transplanted into doc embryos. The recipient embryos were analyzed during the pharyngula period (Kimmel et al., 1995). Descendants of labeled wild-type cells differentiated into large vacuolated notochord-like cells in an anterior notochord region (Fig. 7A-D). However, the wild-type cells were not able to induce neighboring doc mutant cells to form notochord. Mutant doc cells transplanted into wild-type recipients never formed vacuolated cells in the notochord. These transplantation experiments therefore show that the doc gene product acts cell autonomously in the notochord for its formation.

Some of the transplanted wild-type cells that formed a stretch of notochord in doc embryos were found to be associated with a horizontal myoseptum in the neighboring somites, which consisted entirely of doc mutant cells (arrow in Fig. 7E,F). This illustrates that doc gene activity is required in the notochord for the formation of the horizontal myoseptum. Transplanting cells from doc embryos into wild-type hosts support this finding. Large areas of doc mutant cells in the somites of wild-type hosts were able to form a morphologically normal horizontal myoseptum, indicating that doc mutant cells in the somites are able to respond to the signal from the notochord (arrows in Fig. 7G,H). This experiment indicates that the formation of the horizontal myoseptum depends on the function of doc in the notochord, resulting in the production of the inducing signal, and not in the somites where the signal from the notochord is received.

We have isolated multiple alleles of two genes previously known to be required for notochord formation: 3 alleles for ntl (Halpern et al., 1993) and 2 alleles for flh (Talbot at al., 1995; Halpern et al., 1995). Only one allele of mom was isolated. The variability observed in the mom phenotype suggests that the allele we isolated is not a null allele. For doc three alleles of different strength were isolated. doc has a temperature dependent dominant phenotype that may lead to a reduced viability of heterozygous individuals. Dominant lethality, even with low penetrance, reduces the number of heterozygous carriers and biases against the recovery of strong alleles. No dominant effects have been observed for flh, mom and ntl.

The flh alleles are all of the same strength and probably reflect the amorphic state of the gene. During the time of this study, flh has been shown to encode the zebrafish homologue of Xnot, a homeobox-containing transcription factor from Xenopus laevis (Talbot et al., 1995 and references therein). Two alleles isolated in our screen, flh^tm229 and flh^tk241, carry point mutations causing premature stop codons in the coding region upstream of the homeodomain. It is likely that these two alleles represent a complete loss of function (Talbot et al., 1995). For no tail, we have isolated three alleles, all of different strength. One of them, ntl^tc41, results in a phenotype that is stronger than that of ntl^b160 and ntl^b195. This is surprising, as all three alleles introduce stop codons in the N-terminal DNA binding domain and are expected to be amorphic.

In flh embryos, axial mesodermal cells appear to form somitic muscle cells instead of notochord. myoD, which normally marks the paraxial cells, is ectopically expressed in the axial mesodermal cells of flh (Halpern et al., 1995). Fate map studies support this finding (Melby et al., 1996). mom resembles flh with respect to the fate of the axial mesodermal cells. In contrast, axial mesodermal cells are present in ntl and doc embryos but remain undifferentiated. This suggests a role for flh and mom in the early specification of the axial mesoderm, preventing it becoming paraxial mesoderm instead. In ntl and doc mutant embryos, however, axial mesodermal development is initiated and notochord precursor cells are formed that fulfill some of the functions of the notochord, such as signalling to the ventral neuroectoderm, while signals required for the formation of a horizontal myoseptum by the paraxial mesoderm is impaired in embryos mutant of all four genes.

A phenotype reminiscent of the flh and mom phenotype has been described in mice. Mutant HNF-3β embryos do not form an organized node and notochord (Ang and Rossant, 1994; Weinstein et al., 1994). In addition, mutant embryos show marked defects in the organization of the somites and the neural tube. axial is the zebrafish homologue of HNF-3β?(Strähle et al., 1993). A linkage analysis between mom and?axial is currently being performed.

In order to investigate the function of the individual genes in notochord formation, we analyzed the expression of several notochord markers in flh, mom, ntl and doc mutants (summarized in Table 2). We found that ntl is expressed at the beginning of gastrulation in the axial mesoderm of flh and mom embryos. During later stages, however, ntl expression remains restricted to a small region close to the tailbud. Only in this ntl expressing region are other notochord specific markers, such as twi, shh and axial detectable. In more anterior regions, cells of the axial mesoderm now express adaxial markers (myoD, snail; Halpern et al. 1995). These findings suggest that flh and mom suppress adaxial development in these cells. In both mutants, disruption of axial mesoderm development is not absolute and faint expression of axial mesodermal markers such as ntl can be detected anterior to the tailbud in flh and mom embryos (Talbot et al., 1995; this study). It is possible that flh and mom have partially redundant functions in the specification of the axial mesoderm. Double mutant analysis will clarify this point.

In doc embryos, Ntl expression is initiated in the axial mesoderm but it is not maintained beyond early somitic stages. The expression of other notochord markers parallels the presence of Ntl. These notochord markers are not expressed in the axial mesoderm of strong ntl alleles. We propose that the similarity in phenotype of doc and ntl is explained by a function of doc in maintaining Ntl at high levels in the axial mesoderm, which in turn is required for the further development of the notochord, and the expression of twi, shh and axial.

The notochord in mutant flh embryos is missing, whereas the floor plate is disrupted in most regions of the embryo (Talbot et al., 1995). flh is expressed in the notochord and in the floor plate (Talbot et al., 1995) and therefore could have a function in both structures. However, descendants of transplanted mutant cells into wild-type embryos contribute to a morphologically normal floor plate, indicating that flh function in the floor plate is not necessary for its proper formation (Halpern et al., 1995). Although the axial mesoderm is less affected in mutant mom than in flh embryos, the floor plate is affected more strongly. In mom embryos, floor plate cells often overly residual notochord cells, which is consistent with a direct induction of floor plate by the notochord. This is not the case for flh embryos in which the notochord is almost completely absent (as seen by the failure of the midline mesoderm to express notochord markers) while floor plate is formed to a substantial degree. It has been suggested that the midline mesoderm in flh retains some of its capacity to induce floor plate development, despite the absence of further notochord differentiation (Halpern et al., 1995). Alternatively, floor plate development in flh embryos could be induced early on while the presumptive ventral neural tube cells are still in contact with the axial mesoderm close to the shield. In flh and mom embryos these cells do express axial mesodermal markers. We suggest that floor plate development in normal embryos depends on both the initial specification of ventral neural tube during shield stage, and later, on the adjacent axial mesodermal cells, as has been shown for other vertebrate systems (Yamada et al., 1991; Placzek et al., 1991).

Evidence for a direct role of the axial mesoderm in floor plate induction is provided by the examination of the floor plate in ntl embryos. In embryos with the strong ntl alleles the floor plate is broadened as indicated by the enlarged expression domains of shh in the ventral spinal cord (Fig. 4D,H). The enlargement of the shh domain correlates with the region occupied by undifferentiated axial mesoderm cells in ntl embryos (dependent on the phenotypic strength of the ntl allele). This region is also enlarged in ntl embryos as seen by flh expression (Talbot et al., 1995) and by the increased distance between the two bilateral stripes of adaxial cells expressing myoD (Fig. 6D,E) and snail (Hammerschmidt and Nüsslein-Volhard, 1993). We propose that cells of the ventral neural tube are induced to develop floor plate in ntl mutants by their contact with undifferentiated axial mesodermal cells. This induction does not require ntl (Halpern et al., 1993) or doc.

The broadened axial mesoderm in ntl embryos suggests that convergent extension in the axial mesoderm during gastrulation is affected. Our results are consistent with the finding that the brachyury gene, the mouse homologue of ntl, is required for convergent extension movements in the axial mesoderm (Yamada, 1994 and references therein).

The notochord in vertebrates plays a crucial role in somite patterning by inducing sclerotome in the presomitic mesoderm (reviewed by Bumcrot and McMahon, 1995). In the zebrafish the embryonic sclerotome is a rather small structure and specific markers expressed in this structure are not yet available. However, signals from the notochord seem to be required for the formation of muscle pioneer cells and the horizontal myoseptum (Halpern et al., 1993).

Somites are abnormally shaped in mutant embryos of flh, mom, ntl and doc. A morphologically distinct horizontal myoseptum does not form and Eng expression in the nuclei of muscle pioneer cells is not detectable in regions where a differentiated notochord is missing. Normal levels of adaxial myoD expression in flh and mom embryos early on indicate that a differentiated notochord is not necessary for the initiation of myoD expression. Strongly reduced levels of myoD expression in the adaxial cells were found in ntl mutant embryos (Weinberg et al., 1996). We find that the level of myoD expression in ntl embryos is dependent on the allele strength. Therefore ntl itself appears to have a function in the initiation of adaxial myoD expression to full levels. ntl is transiently expressed in the presumptive mesoderm (Schulte-Merker et al., 1992), and therefore could in principle function early in the paraxial precursor cells to activate myoD expression. However, ntl has been shown to be required cell autonomously in the notochord for the patterning of the adjacent paraxial mesoderm (Halpern et al. 1993). Furthermore, shh expressed ubiquitously in the embryo induces myoD in the entire paraxial mesoderm, even in ntl embryos (Weinberg et al., 1996). This suggests that the requirement for ntl in the induction of myoD is via its role in the expression of shh.

Reduced levels of myoD expression in the adaxial cells at later stages of development in flh, mom and doc embryos correlates with the loss of ntl expression in the axial mesoderm. In doc embryos Eng-expressing muscle pioneers are missing, even though the adaxial expression of myoD is maintained until later somite stages. As doc is also required autonomously in the notochord, maintained notochord development or a signal downstream of doc appears to induce muscle pioneers in a subset of the myoD-expressing adaxial cells. Mutations affecting this signal or its reception are expected to display a somite phenotype similar to that observed in flh, mom, ntl and doc mutants, while retaining a normal notochord. During the course of this screen, six other genes were identified with defects in the horizontal myoseptum and the muscle pioneers (chameleon (con), choker (cho), sonic-you (syu), u-boot (ubo), you and you-too (yot); van Eeden et al., 1996). In contrast to the notochord mutants described in this study, these you-type mutants do not show any morphological notochord defects. However, myoD expression is severely reduced or absent in adaxial cells of embryos mutant for you, yot, syu, and con (van Eeden et al., 1996). While ntl and doc are shown by mosaic analysis to be required in the notochord, similar experiments performed in yot and ubo indicate a requirement for these two genes in the somites, suggesting a function in the reception of the signal from the notochord. The similarity between the somite phenotypes of you-type mutants and that of flh, mom, ntl and doc indicate that our screen has identified genes involved in different steps of a signaling pathway from the notochord to the somites.

In addition to the mutants with defects in early notochord formation we have identified a number of mutants with later notochord defects (Tables 3 and 4, Fig. 8B). Mutations in six genes, sleepy (sly), grumpy (gup), bashful (bal), happy (hap), sneezy (sny) and dopey (dop), produce embryos without fully differentiated notochords. The defects in notochord differentiation result in short embryos with thin notochords that are poorly vacuolated. Although the somites of mutant embryos appear disorganized, Eng-expressing muscle pioneer cells are present. In addition, three of these mutants (sly, gup and bal) also show a severe disorganization of the overall brain morphology (Fig. 8D) and a number of axons that normally cross the midline grow abnormally in sly, gup and bal. In particular retinotectal axons frequently turn rostrally after crossing the midline or project to the tectum on the ipsilateral side instead of crossing the midline (Karlstrom et al., 1996). Similarly, we found that in sly the Mauthner cell axons frequently fail to cross the midline and grow along the ipsilateral side (Fig. 8F). We have isolated mutations in five genes, which lead to undulations in the normally straight and rod-like notochord. In crash test dummy (ctd), the notochord is bent variably in many positions, causing severe distortions of the body axis (Fig. 8H). In zickzack (ziz), the notochord specifically undulates sideways at regular intervals along the body axis. In quasimodo (qam), there are irregular bulges and thickenings of the notochord that increase in size over time. In addition to the notochord phenotype, qam also displays reduced melanophore pigmentation in the body and in the eyes. Mutations of kinks and wavy tail lead to undulations and kinks in the notochord predominantly in the tail.

The vacuolated notochord cells are surrounded by an epithelial monolayer of cells termed the notochord sheath. The notochord sheath is thought to keep the notochord cells in a linear array, resulting in the rod-like stiff structure of the notochord. Embryos mutant for korken (kon) do not form the notochord sheath. Instead, round cells that look similar to undifferentiated notochord cells are present all around the periphery of the notochord at 40 hours of development whereas the vacuolated cells of the notochord appear normal (Fig. 8J). Embryos homozygous for kon also have reduced motility and only jump once after touch.

26 mutants display a variety of other notochord defects, mostly notochord degeneration (Table 5). 10 of these mutants are homozygous viable.

We thank Bob Riggleman for providing the twist probe prior to publication, William Talbot, Anne Melby, Marnie Halpern and Chuck Kimmel for communicating results prior to publication, Bill Trevarrow for the flhⁿ¹ allele, Stefan Schulte-Merker for providing the ntl antibody, and N. H. Patel for providing the Eng antibody (4D9). We thank Klaus Trummler, Frank Uhlmann and Mathias Metz for assistance in the analysis of the ntl alleles, Silke Rudolph for technical assistance, Heike Schauerte for helping with the in situ hybridization, and Joel Wilson and Cornelia Fricke for their help with the fish work, and finally Tanya Whitfield, Francisco Pelegri, Darren Gilmour and Stefan Schulte-Merker for discussion and help with the manuscript.