Somitogenesis is the basis of segmentation of the mesoderm in the trunk and tail of vertebrate embryos. Two groups of mutants with defects in this patterning process have been isolated in our screen for zygotic mutations affecting the embryonic development of the zebrafish (Danio rerio). In mutants of the first group, boundaries between individual somites are invisible early on, although the paraxial mesoderm is present. Later, irregular boundaries between somites are present. Mutations in fused somites (fss) and beamter (bea) affect all somites, whereas mutations in deadly seven (des), after eight (aei) and white tail (wit) only affect the more posterior somites. Mutants of all genes but wit are homozygous viable and fertile. Skeletal stainings and the expression pattern of myoD and snail1 suggest that anteroposterior patterning within individual somites is abnormal. In the second group of mutants, formation of the horizontal myoseptum, which separates the dorsal and ventral part of the myotome, is reduced. Six genes have been defined in this group (you-type genes). you-too mutants show the most severe phenotype; in these the adaxial cells, muscle pioneers and the primary motoneurons are affected, in addition to the horizontal myoseptum.
The horizontal myoseptum is also missing in mutants that lack a notochord. The similarity of the somite phenotype in mutants lacking the notochord and in the you- type mutants suggests that the genes mutated in these two groups are involved in a signaling pathway from the notochord, important for patterning of the somites.
Segmentation is an important strategy for patterning the anteroposterior (A/P) axis in both vertebrate and invertebrate phyla. In Drosophila many of the genes involved in segmentation were identified by genetic screens for mutations affecting embryogenesis (Nüsslein-Volhard and Wieschaus, 1980; Jürgens et al., 1984; Nüsslein-Volhard et al., 1984; Wieschaus et al., 1984). Investigation of these mutants and the subsequent cloning of the corresponding genes has allowed a detailed molecular analysis of this process (Martinez Arias, 1993; Pankratz and Jäckle, 1993). In vertebrates there is still little known about the segmentation process. Rotation experiments in chick indicate that metamerism in the tail depends on the somites, which are derived from mesoderm (Keynes and Stern, 1984). This differs from Drosophila, where the ectoderm is likely to impose segmentation onto the mesoderm (Bate, 1990).
In zebrafish (Danio rerio), as in most other vertebrates, somites form as epithelial spheres from the presomitic mesoderm in an anterior to posterior direction. Starting at 10 hours postfertilization (hpf), one pair of somites is formed every 20-30 minutes by formation of a new somitic furrow. Roughly 30 somite pairs form in a normal embryo; 7 above the yolk cell, 10 above the yolk extension and 13 posterior to the anus. Perpendicular to the A/P axis of the embryo the somites are subdivided into sclerotome and dermomyotome, which later gives rise to myotome and dermatome. In zebrafish, myotome is the major part of the somites, in contrast to mouse and chick, where the major part of the somites becomes sclerotome. The zebrafish sclerotome is composed of a relatively small group of cells in the ventral part of the somite (B. MorinKensicki, personal communication). Vertebrae are derived from the sclerotome, and are formed comparatively late in zebrafish development.
In the course of development three distinct structures are recognizable in the zebrafish myotome: adaxial cells, muscle pioneers and the horizontal myoseptum. In presomitic mesoderm, adaxial cells are distinguishable as large cuboidal cells, adjacent to the notochord. These differ from the rest of the paraxial mesoderm by morphology, behavior and gene expression (Felsenfeld et al., 1991; Hammerschmidt and Nusslein-Volhard, 1993; Thisse et al., 1993; Weinberg et al., 1996). Until the 7-somite stage, the somites have the shape of epithelial spheres. After this timepoint, the adaxial cells in the first five somites start to elongate and intercalate, until they span an entire somite. Adaxial cells from somites that form later in development elongate and intercalate shortly after somite boundary formation. The muscle pioneers, probably a subset of the adaxial cells, are the first cells to show muscle striation in the myotome (Felsenfeld et al., 1991). They lie adjacent to the notochord in the region of the horizontal myoseptum, and are labeled by anti-Engrailed antibody (4D9, Patel et al., 1989; Hatta et al., 1991) starting from the 8-to 10-somite stage and later by Zn5 monoclonal antibody (Trevarrow et al., 1990). The horizontal myoseptum is a fibrous sheet dividing the myotome into a dorsal and a ventral part, and is first visible around 28 hpf. Adaxial cells, muscle pioneers and the horizontal myoseptum are affected in mutants lacking the notochord. It has been shown by cell transplantation experiments that wild-type notochord can induce muscle pioneers and horizontal myoseptum in such mutants (Halpern et al., 1993; Odenthal et al., 1996).
Analysis of somitogenesis in other vertebrates has defined a number of properties of this process. In frogs, a short heatshock during somite formation disturbs somitogenesis. After the heatshock, 4-5 normal somite pairs form, followed by 1-2 abnormal ones (Elsdale et al., 1976). This suggests that the process of segmentation is acting in the presomitic mesoderm, before visible boundaries are formed. Similar effects have been observed in zebrafish (Kimmel et al., 1991). Somites form relatively normally from isolated presomitic mesoderm in culture (Packard and Jacobson, 1976). If newly formed somites are rotated along their A/P axis they form vertebrae with the corresponding inverse polarity (Aoyama and Asamoto, 1988). These experiments indicate that patterning of the paraxial mesoderm along its A/P axis is an intrinsic property of this tissue.
Patterning in the mediolateral (M/L) plane is dependent on induction by the notochord, neural tube and the overlying ectoderm (Brand-Saberi et al., 1993; Münsterberg and Lassar, 1995, and references therein). Transplantation and rotation experiments in chick suggest that polarity in this plane is determined shortly after somite formation (Aoyama and Asamoto, 1988).
A small number of genes are known to be involved in the process of somitogenesis. In mouse, targeted disruption of the fibronectin (George et al., 1993), integrin alpha5 (Yang et al., 1993), FGFr1 (Deng et al., 1994; Yamaguchi et al., 1994), Wnt-3a (Takada et al., 1994) or Notch1 genes (Conlon et al., 1995) results in impaired somite formation.
In our screen (Haffter et al., 1996) two groups of genes were identified that will be valuable for understanding patterning of the somites in both the A/P and M/L axes. Mutations in five genes affect the formation of early somite boundaries. Two of these, fused somites (fss) and beamter (bea), affect the formation of all somite boundaries, while mutations in deadly seven (des), after eight (aei) and white tail (wit) have an effect on more posterior somite boundaries only. 11 genes are involved in the formation of the horizontal myoseptum. Six of these, you, you-too, sonic-you, chameleon, u-boot and choker are required for formation of the horizontal myoseptum but not for notochord formation. The five others, no tail, floating head, doc, momo and dino, also have an effect on notochord formation (Hammerschmidt et al., 1996b; Odenthal et al., 1996). This paper presents an initial characterization of mutants with defects in somite boundary or horizontal myoseptum formation, and other mutants that affect the paraxial mesoderm as a secondary effect (Table 1).
MATERIALS AND METHODS
Maintenance of fish stocks
Maintenance of fish stocks and matings were done as described by Mullins et al. (1994).
All analyses were done with the strongest alleles of the respective genes, if there were detectable differences. The alleles used were fsste314a, beato202, destx201, aeitr233, yotty119, syutq252, contm15a, youty97, ubotp39 and chotm26 . The viability of fss was tested with fssti1 .
Antibody staining and whole mount in situ hybridization
In situ hybridization and antibody staining was done as described by Hammerschmidt et al. (1996b). The probes that were used were myoD (Weinberg et al., 1996), snail1 (Hammerschmidt and NussleinVolhard, 1993; Thisse et al., 1993), and sonic hedgehog (Krauss et al., 1993). The 4D9 monoclonal antibody (Patel et al., 1989), anti-Ntl rabbit polyclonal antiserum (Schulte-Merker et al., 1992), anti-Snail1 rabbit polyclonal antiserum (Hammerschmidt and Nusslein-Volhard, 1993), znp-1 monoclonal antibody (Trevarrow et al., 1990) and monoclonal anti-acetylated tubulin antibody (Sigma) were used for stainings.
Pictures of live embryos were taken on a Zeiss Axiophot microscope using Kodak ektachrome 64T or 160T slide films; rhodamine-labeled cells were photographed using Kodak ektachrome 400T slide film. Pictures were scanned on a Nikon Coolscan slidescanner; composite pictures were made using the Adobe Photoshop software package on a Macintosh computer.
Cell transplantation was basically done as described by Ho and Kane (1990), with the following modifications. For mounting donor and host embryos, small wells were made in agar (2% in E2; Westerfield, 1989), holding single embryos. Transplantations were done using a Zeiss Stemi 2000 stereomicroscope at 20-25×, around the sphere stage (Kimmel et al., 1995).
To determine the effect of the fss mutation on the muscle fiber length, small numbers (5-10) of fss mutant cells were transplanted into fss mutant hosts. Embryos were allowed to develop for 48 hours. Only the length of clearly isolated muscle fibers was measured on a video screen. In addition, the A/P level of these fibers was recorded (all fibers were parallel to the A/P axis). The values obtained were divided by the length that would be expected for muscle fibers parallel to the A/P axis at that level. As a control, exactly the same experiment was done with a set of transplants from wild-type to wild-type embryos.
For the rescue of the myoseptum defect of yot and ubo mutants, slightly larger numbers of cells were transplanted (10-30 cells). Mutant chimaeric embryos were scored for rescue of the myoseptum phenotype after 2-3 days, using a Zeiss Axiophot.
Fish were fixed in 3.7% formaldehyde/PBS for 24-72 hours. After fixation specimens were rinsed and stained with 0.1 mg/ml Alcian blue 8 GX (Sigma) in ethanol:acetic acid 4:1 (v:v). Specimens were incubated in 90%, 50% and 30% v/v ethanol in water. After a final 2 hour wash in water, specimens were digested overnight in a 50 mg/ml solution of trypsin (crude, Sigma) in a 30% saturated NaB 4 O 7 solution in water. Bones were stained with 0.4 ml solution of Alizarine red S (saturated in ethanol; Sigma) in 10 ml 0.5% KOH. Destaining was done in a 1% KOH/glycerin series (3:1, 1:1, 1:3). Scales and sometimes muscles were removed manually. Specimens were stored at 4°C in glycerol. Sometimes grease was removed by another incu-bation in acetone and sometimes specimens were bleached in a 1% H 2 O 2, 1% NH 3 solution, after Alcian blue staining.
Mutations affecting somitogenesis
20 mutants with defects in the formation of epithelial somites were isolated. Complementation analysis defined five complementation groups: fused somites (fss), beamter (bea), deadly seven (des), after eight (aei) and white tail (wit) (Table 1, fss- type genes).
The two alleles of the fss gene are indistinguishable in phenotypic strength. Embryos mutant for fss do not show any morphological evidence of boundary formation in the paraxial mesoderm at early somite stages (Fig. 1A,B). Eventually, however, irregular borders are formed in the paraxial tissue (Fig. 2A,B). The horizontal myoseptum and muscle differentiation are not obviously affected by fss mutations (Fig. 2C,D). The phenotype of bea mutant embryos differs slightly from that of fss mutants. At early somite stages, before differentiation of the most anterior somites, segment boundaries can be seen, in contrast to fss mutants. These boundaries are irregularly placed and indistinct in comparison to those of wild-type siblings (Fig. 1C,D). Later in development, somitogenesis seems to be more severely disturbed, and in the tail the somite boundary defect is as severe as that of fss mutants (Fig. 3). As a result, at 24 hpf, bea mutants seem to have 1-4 relatively normal-looking anterior somite boundaries, followed by irregular boundaries more posteriorly.
In des and aei mutants the first 7±2 somites form as in wildtype siblings, but more posterior somite boundaries change gradually from a wild-type arrangement to an irregular arrangement similar to that of fss. The transition from normal to defective somite boundaries occurs in a similar region in all cases analyzed and no difference in allele strength has been detected among different alleles of aei and des.
The single wit allele exhibits a somite defect that is very similar to that of des and aei. In addition wit mutant embryos have an increased number of primary neurons. A detailed account of this phenotype is given elsewhere (Jiang et al., 1996).
A morphological comparison of the three different phenotypes is given in Fig. 3. For all genes except wit, homozygous mutants survive to adulthood. There are no obvious motility defects, but sometimes these fish are slightly kinked. For fss mutants, survival rates were 75% of the level of the siblings in an India/Tübingen hybrid background (33 versus 46 survivors out of 50).
We have raised homozygous mutants for fss, bea, aei and des. Embryos from crosses between homozygous females and heterozygous males did not have a stronger phenotype than embryos from the reciprocal cross, indicating that these genes do not have a maternal component that might be sufficient to allow somitogenesis in the anteriormost somites, or for the irregular boundaries seen in older mutant embryos.
To investigate further why the anteriormost somites are formed normally or relatively normal in mutants for des, bea and aei, we have made all three double mutant combinations between these three genes. No clear enhancement of the phenotype was found, indicating that these three genes do not have overlapping functions in the anterior part of the paraxial mesoderm.
Segmentation is irregular in fss mutants
At 72 hpf irregular somite boundaries are present in fss mutants (Fig. 2). Observation of live embryos around the 7-somite stage revealed that in fss mutants the elongation and intercalation of the adaxial cells proceeds normally, the only abnormality being the lack of somite boundaries. Soon after these movements have taken place, somite boundaries, although irregular, are formed (Fig. 4).
The average length of muscle fibers in homozygous fss embryos was estimated by transplanting fluorescently labeled mutant fss cells into mutant fss embryos before gastrulation. After 2 days of development labeled muscle fibers were measured and their lengths compared to those of muscle fibers at that particular A/P level in wild-type embryos. The length of individual fibers was variable, but on average only slightly longer in fss than in wild type (Table 2).
In the zebrafish, primary motoneurons are segmentally organized with each myotome having three primary motoneurons (Eisen et al., 1986), namely the Caudal, Middle and Rostral primary motoneurons (CaP, MiP and RoP, respectively). The CaP axon projects to the ventral myotome and the cell body of this neuron lies most posterior within each segment. Stainings with the znp-1 monoclonal antibody, which labels the axons of primary motoneurons (Trevarrow et al., 1990), reveals that in fss mutants CaP axons are present, but are more irregularly spaced, frequently truncate prematurely (10-20%, Fig. 5A,B) and sometimes bifurcate or join with a neighboring axon. MiP axons are present, and affected in a similar way. The effect on the RoP has not been analyzed so far. The motoneuron pattern seems to reflect the irregular segments in fss mutants at 28 hpf. Similar effects were found in des, aei and bea mutants.
Anteroposterior patterning within the somite might be affected
Whole-mount skeletal stainings of fss mutants were performed to see the effect of this mutation on the vertebral column. Tail vertebrae in the zebrafish consist of a centrum, the hemal arch ventrally, and the neural arch dorsally (Fig. 6C). In anterior vertebrae the hemal arches are replaced by ribs, the ribs and both arches being attached to the anterior of the centrum. In fss mutants the centra are relatively normal in number and shape but the hemal and neural arches seem to grow in an irregular fashion and from ectopic positions (Fig. 6B,D). Ribs are present, but are irregularly placed, frequently not attached to the centra, and sometimes bifurcate (data not shown).
A centrum can be divided into eight parts according to anterior/posterior, left/right and dorsal/ventral position. Thus in a wild-type fish there are four anterior positions with, and four posterior positions without a half hemal or neural arch. In the tail of two fss mutant fish, 15 vertebrae were analyzed for the position of these half arches. A position was scored as anterior if an outgrowth was present and posterior if not. In 35% of the positions a change in polarity was scored (82/232, 8 uncertain). If the polarity was completely randomized, a frequency of 50% would be expected, showing that some polarity is still present in the A/P axis of the centra.
Other markers for the A/P polarity within single somites are myoD (Fig. 7) and snail1 (not shown). These markers are expressed in the most posterior somitic cells just after their formation, generating a clear striping pattern, and expression subsequently spreads to more anterior cells. In fss mutants this expression is affected such that, although there still seems to be a wave of expression that passes from anterior to posterior over the paraxial mesoderm as a whole, there are no detectable stripes of posterior somite expression (Fig. 7E,F). The expression still appears slightly patchy, suggesting that there might be some residual polarity within the somites. The adaxial cells, however, express myoD and snail1 normally.
The vertebrae of the other three viable mutants, bea, des and aei, were analyzed in the same way as fss. In all three cases the frequency of positions with abnormal polarity was lower (±12%). In the tail region of bea, des and aei, myoD expression appears homogeneous, as in fss mutants, but more rostrally where somite formation is affected less severely (bea), or not at all (des and aei), a striping pattern is still visible. The transition from apparently normal to abnormal somite formation is shown by in situ hybridization with myoD on des and bea mutant embryos in Fig. 7A-D. We conclude that in fss, bea, des and aei mutants, A/P polarity within the individual somite is affected, although the A/P polarity in the paraxial mesoderm as a whole is still present.
Mutations affecting the horizontal myoseptum
The horizontal myoseptum is a fibrous sheet separating dorsal and ventral myotome (Fig. 8). In jawed fishes this septum is very prominent morphologically and in zebrafish it can be seen at 28 hpf. Such a separation between the dorsal and ventral axial muscles is not restricted to fish but is also found in the axial musculature of terrestrial vertebrates.
Mutants in six genes (you-type genes), you (you), you-too (yot), sonic-you (syu), chameleon (con), u-boot (ubo) and choker(cho) have no or a reduced horizontal myoseptum, but a morphologically normal notochord. Many of these genes were given names with ‘you’ to refer to the U-shape of the somites in these mutants. Both formation of vertical somite boundaries and the notochord are unaffected in you-type mutants. A short phenotypic comparison between you-type mutants is given in Table 3.
In mutants for yot, syu, you and con the floorplate is indistinct at 20 hpf, but at 72 hpf it is morphologically distinguishable. sonic hedgehog (shh) expression in the floorplate is normal in yot, you and con mutants, but in syu shh expression is reduced at 24 hpf (data not shown; Krauss et al., 1993; Brand et al., 1996). In you-type mutants motility in response to touch is variably reduced; yot and con mutants are severely affected, whereas ubo, youtm146c and youtz310b mutants have a very mild phenotype. Mutations in three genes also show fin defects; in con and syu the pectoral fins are variably reduced, and in ubo all fin edges are indented irregularly (van Eeden et al., 1996). Mutants for you, yot, syu and con have a circulation defect, probably because formation of a functional dorsal aorta is delayed or does not occur at all (Chen et al., 1996). In ubo and cho formation of the dorsal aorta is unaffected. Finally, in yot and con mutants the spacing of the eyes is variably reduced, suggesting a variable reduction in ventral brain structures. In addition, these mutants have defects in the pathfinding or outgrowth of retinal axons. Details of the neural phenotypes of yot and con are given elsewhere (Brand et al., 1996; Karlstrom et al., 1996).
All you-type mutations are lethal, except for youtm146c and youtz310b. Lethality is probably due to the circulation defect or to lack of an air-filled swimbladder. Homozygous youtm146c or youtz310b fish do not have a detectable adult phenotype, and do not show maternal effects. Skeletal staining of homozygous fish did not reveal obvious defects.
Mutations in another five genes, dino (din), doc (doc), momo (mom), no tail (ntl) and floating head (flh), affect notochord formation and these mutants also lack the horizontal myoseptum (Halpern et al., 1993; Talbot et al., 1995; Hammerschmidt et al., 1996a; Odenthal et al., 1996). no tail and doc mutants have an undeveloped notochord in the trunk, and form somites lacking the myoseptum. flh mutants lack a notochord, and notochord precursor cells are absent. Somites from the left and the right side fuse under the neural tube and these also lack the horizontal myoseptum. momo lacks the notochord in the trunk and dino lacks the notochord in the tail. In these two mutants the somite phenotype is restricted to the region where the notochord is missing, but otherwise it is similar to the flh phenotype. There is evidence that the lack of the horizontal myoseptum in these mutants is caused by the lack of the notochord (Halpern et al., 1993; Odenthal et al., 1996). In addition to these mutants, six more genes have been defined that are necessary for later notochord development. Somites are condensed in mutants of these genes but the horizontal myoseptum and the muscle pioneers are present (Table 1). Detailed descriptions of mutants with defects in the notochord are given elsewhere (Hammerschmidt et al., 1996a; Odenthal et al., 1996).
In you-type mutants adaxial cells and muscle pioneers are affected
Muscle pioneers are a subset of muscle cells in the region of the horizontal myoseptum. They can be recognized morphologically at 24 hpf, by their early formation of myofibrils. In wild-type embryos 2-6 nuclei of these cells per segment express the Engrailed antigen, which is recognized by the 4D9 monoclonal antibody (Hatta et al., 1991). Stainings with 4D9 monoclonal antibody on yot, the mutation with the strongest phenotype, shows that Engrailed-expressing cells are absent from the somites (Fig. 9A,B). Engrailed expression in the midbrain-hindbrain boundary is normal (Fig. 9A,B).
During somite stages, myoD is expressed at high levels in the adaxial cells. In situ hybridization with myoD on yot mutant embryos shows that expression is reduced at the tailbud stage (data not shown). At 15 somites no expression is detectable in the adaxial cells (Fig. 9E,G), but the expression in the posterior part of the somites, which comes up immediately after their formation, is initially normal. More anteriorly, however, myoD expression is reduced prematurely, and the anterior part of the somite never expresses myoD at normal levels (Fig. 9E,G). yot has a dominant effect on myoD expression, since an unsorted mixture of yot mutants and siblings revealed an approximate 1:2:1 (7:19:7) ratio in embryos with high, low or no detectable myoD expression in the adaxial cells, respectively (Fig. 9E,F,G). In situ hybridization with snail1 (Hammerschmidt and Nusslein-Volhard, 1993; Thisse et al., 1993) on yot mutants indicates that the adaxial expression of this gene is affected in a similar way (data not shown). The morphology of adaxial cells is also affected in yot mutants. At the 2-to 3-somite stage, before the abnormal somite shape appears, yot mutants can be recognized by the morphology of the adaxial cells.
The reduction in myoD expression is less pronounced in syu, con and you mutants, especially in the tailbud where some adaxial expression is present. No difference was detected between ubo mutants and wild-type siblings, suggesting that this gene is not required to establish adaxial expression of myoD. In syu, con and mutants with the strongest you allele (ty97), segments rarely have a few Engrailed-positive cells (data not shown). Engrailed staining of a mutant with the weakest phenotype, youtm146c, showed approximately one weakly staining nucleus per segment (Fig. 9C,D). ubo mutants show reduced Engrailed staining in all segments, and in cho mutants muscle pioneers are present, as judged by their early striation.
The somite phenotype of the five mutants that also affect notochord formation is very similar to the phenotype of yot mutants. Engrailed expression is not detectable in the somites and adaxial expression of myoD is severely reduced (Halpern et al., 1993; Weinberg et al., 1996; Odenthal et al., 1996).
In you-type mutants primary motoneuron patterning is abnormal
The motility defect of you-type mutants suggested that there might be an effect on motoneurons as well. Primary motoneurons were stained by the znp-1 monoclonal antibody in most you-type mutants. In yot mutants, CaP and MiP axons, which normally project to the ventral and dorsal myotome, respectively, are absent in many segments; instead, axons can be seen running along the neural tube (Fig. 5C,D,E). Staining with a monoclonal antibody against acetylated tubulin shows that early axonal tracts in the spinal cord are normal (data not shown). This indicates that the yot mutation does not have a general effect on the early axonal tracts, and suggests that the effect on the motoneurons is specific. At present it is not clear, however, if the primary motoneuron phenotype is due to abnormal somite patterning or to a defect in the motoneurons themselves.
The primary motoneuron phenotype is also detectable in you, syu and con mutants, but the frequency of abnormally projecting axons is lower. Mutations in the flh gene show a similar primary motoneuron defect: axons do not enter the paraxial mesoderm, but instead truncate or project along the neural tube (J.O., unpublished; Talbot et al., 1995).
you-too and u-boot functions are required in the paraxial mesoderm
Because there is evidence that the myoseptum is induced by the notochord (see Discussion), it was interesting to determine whether the gene products for the you-type genes are required in the notochord or in the somites. Cell transplantations from wild-type embryos to yot and ubo mutant embryos were carried out. A small number of fluorescently labeled wild-type cells were transplanted into mutant embryos before the onset of gastrulation (Ho and Kane, 1990). The resulting genetic mosaics were analyzed morphologically. In both mutants a myoseptum is only formed if wild-type donor cells are present at the level of the horizontal myoseptum (Fig. 10). A large number of wild-type cells in the notochord, in contrast, never leads to rescue of a myoseptum (Table 4). In syu, you and con mutants the incomplete penetrance of the myoseptum phenotype did not allow such an analysis.
Other mutations affecting the paraxial mesoderm
A number of mutations were found to affect formation of the entire paraxial mesoderm and thus also to influence somitogenesis. Mutations in five genes result in a variable dorsalization of the embryo and show a ventral expansion of the paraxial mesoderm (Mullins et al., 1996). We have also identified two new alleles of spadetail, which causes a severe reduction of cells in the paraxial mesoderm (Ho and Kane, 1990; Hammerschmidt et al., 1996a). Mutations in dino cause partial ventralization of the embryo; in addition to the effect on posterior somite patterning as described, this mutation also leads to a reduced amount of paraxial mesoderm in the anterior region (Hammerschmidt et al., 1996b). 15 genes are required for formation of striated muscle in the myotomes and elsewhere in the embryo. A detailed description of these genes is given elsewhere (Granato et al., 1996).
Anteroposterior polarity within somites is affected in fss-type mutants
We have isolated mutations in five genes in our zygotic screen that are involved specifically in somite formation. Embryos mutant for these genes fail to form epithelial somites properly. Eventually, however, irregular segment boundaries become visible in the myotomes. Mutations in fss show this defect over the entire length of the paraxial mesoderm, while bea mutants have a weaker defect in the anterior paraxial mesoderm but look similar to fss posteriorly. Mutations in the three other genes (aei, des and wit) only cause this defect in the posterior region of the paraxial mesoderm.
Skeletal stainings and myoD expression indicate that antero-posterior polarity within individual somites is affected. The effect of fss is stronger than that of bea, des or aei, as judged by the skeletal phenotype. Although antero-posterior patterning within single somites is affected, the wave of differentiation that passes from anterior to posterior over the paraxial mesoderm as a whole seems to be relatively normal in fss-type mutants. The observed phenotype can be explained quite well by the ‘clock and wavefront’ model that has been proposed (Cooke and Zeeman, 1976). This model assumes that somite formation is the result of two processes, the first a differentiation ‘wavefront’ that moves in a smooth way over the paraxial tissue from anterior to posterior, and the second a ‘clock’ that periodically promotes and blocks the progression of this wavefront. The genes described above might be interpreted as essential for the clock process only.
Irregular segmentation in fss-type mutants
At first glance it is surprising that a mutant like fss, which has abnormal somite formation, survives to adulthood. The following observations might account for this. Differentiation in the mediolateral plane is only weakly affected; adaxial cells, muscle, muscle pioneers, horizontal myoseptum and vertebrae are present. Primary motoneurons have a segmental arrangement. The irregular segmentation seen later in development results in only slightly more variable muscle fiber length and an apparently normal motility.
In fss mutants the adaxial cells flatten and intercalate as in wild type, resulting in irregular stacks of these cells. These stacks are quite similar in length to the length of a somite. (Fig. 7F, arrowheads). Boundaries form where two stacks of adaxial cells meet (Fig. 4). Thus adaxial cells may play a role in forming the irregular boundaries between the myotomes seen at later stages. In fact, in double mutants between fss and yot, where adaxial cells are absent, most of the irregular boundaries are lost, resulting in embryos with almost unsegmented paraxial mesoderm (F.v.E, unpublished).
Somite formation appears in all vertebrates as a continuous repetition of the same process, and hence a gene required for somite formation might be expected to be required for the formation of all somites. It was therefore surprising that only fss shows such a requirement, mutations in des, aei and wit having defects only in the posterior paraxial mesoderm. The phenotype of bea is stronger in posterior than anterior somites. It is possible that there is a stronger requirement for these genes in the posterior versus the anterior paraxial mesoderm, and that all alleles that we isolated are weak alleles. An argument against this possibility is that all alleles we isolated for a particular gene (10 alleles for des) have the same strength. Furthermore, there are no further mutants in our collection that could be candidates for a null-phenotype of these genes. However, we cannot rule out that these candidates were missed in our screen.
There may be other reasons why the first few somites form normally in bea, aei, des and wit. At present we can say that there is no indication of a maternal contribution of bea, aei, and des, which might suffice for somitogenesis in the anteriormost somites. Furthermore, we have made all possible double mutant combinations between aei, bea and des, but did not detect an enhancement of the phenotype, suggesting that these three genes do not have overlapping functions; overlap with wit could still be possible, however. We speculate that other, maternally expressed genes may replace the function of aei, des and possibly bea in the early embryo. The products of such genes might become exhausted early in somitogenesis, their function taken over by the three zygotic genes described here.
At present there are no cloned genes that could be candidates for fss, bea, aei and des. Mice mutant for fibronectin, integrin alpha5, FGFr1, Wnt-3a and the Notch1 gene show additional morphological defects, and always result in lethality. The neural phenotype of wit mutants suggests that a more general factor, involved in neurogenesis, is affected (Jiang et al., 1996).
Mutations affecting the horizontal myoseptum may define a signaling pathway
We have identified 11 genes that affect the formation of the horizontal myoseptum. Six of these, you, you-too, chameleon, sonic-you, u-boot and choker, have a normal notochord. The other five also show defects in the notochord, and these are discussed elsewhere (Hammerschmidt et al., 1996b; Odenthal et al., 1996). Transplantation experiments in ntl and doc mutant embryos have shown that the notochord is required for induction of the horizontal myoseptum and the muscle pioneers (Halpern et al., 1993; Odenthal et al., 1996). If the muscle pioneers are required for myoseptum formation, mutants that lack the horizontal myoseptum might lack the muscle pioneers, and thus define genes that are involved in a signaling pathway from the notochord to the somites. The you-type mutants might define such genes since they exhibit defects that are very similar to those observed in mutants lacking the notochord. In both groups, adaxial cells, muscle pioneers, horizontal myoseptum and motoneuron axon projection are affected.
Although you-type mutants have many aspects of their phenotypes in common with mutants of genes required for notochord specification, there are also differences. For example the yot and con mutations have effects on the spacing of the eyes and on retinal axon projection and outgrowth (Brand et al., 1996; Karlstrom et al., 1996), indicating that the function of these genes is also required in regions anterior to the notochord.
There are indications that Ntl expression has to be maintained in the notochord in order to induce Engrailed expression in the somites (Odenthal et al., 1996). We found that Ntl expression is normal in the notochord of you-type mutants, suggesting that the defect in you-type mutants lies downstream of Ntl. We have not yet been able to identify a you-type gene that might encode an inducing molecule required in the notochord instead of the paraxial mesoderm. We have shown by transplanting wild-type cells into yot and ubo mutants that these genes are required in the paraxial mesoderm for the formation of the horizontal myoseptum. This leaves three good candidate genes, con, you and syu, which might encode the inducer. A preliminary model of where the different you-type genes might act is shown in Fig. 11.
It is not clear how the identified signaling pathway can be compared to the patterning signals from the notochord, which have been defined in the chick. Only syu has an effect on floorplate formation, as seen by sonic hedgehog (shh, Krauss et al., 1993) expression. So far yot and con mutants have been shown to have a reduced number of secondary motoneurons (Brand et al, 1996), which can be induced by both the notochord and the floorplate. Other you-type mutants have not been tested yet. We have not yet analyzed the effect of the you-type mutants on the induction of the sclerotome. In zebrafish, the sclerotome is a rather small group of cells and the vertebral column develops long after day 6 of development. Our screening procedure did not allow isolation of mutants with defects restricted to the vertebrae.
sonic hedgehog is expressed in both floorplate and notochord and is thought to induce structures in neighboring tissues. Injection of shh mRNA into zebrafish embryos causes an increase of adaxial myoD expression (Weinberg et al., 1996), the opposite effect to the reduction of adaxial myoD, seen in most you-type mutants. shh is therefore a good candidate for either con, you or syu. We are currently testing these mutations for linkage to this gene.
We would like to thank P. Ingham and T. Whitfield for valuable comments on the manuscript and cDNA probes, S. Schulte-Merker for the Ntl antibody and J. Eisen and R. BreMiller for the znp-1 antibody.