Anteroposterior patterning is required within segments for somite boundary formation in developing zebrafish

The mechanisms underlying the establishment of segmental prepattern in vertebrate embryos is not well characterised at a molecular level. This is in contrast to the situation in Drosophila, in which the molecular basis of segmentation has been shown to involve a hierarchy of gene interactions initiated by maternal factors that pattern the egg and early embryo. These maternal factors control the expression of gap genes in domains that constitute several adjacent presumptive segments. Gap genes subsequently regulate the expression of pair-rule genes that are transcribed in overlapping regions corresponding to alternating segments. In turn, pair-rule genes control the expression of segment polarity genes that generate anterior and posterior regions within each segment. Finally segments acquire individual identities through the function of hox genes (reviewed by Lawrence, 1992).

Some of the mechanisms necessary to achieve segmentation of the body axis in Drosophila may be conserved and utilised during vertebrate development. For instance, vertebrate homologues of the Drosophila pair-rule gene hairy are involved in somite segmentation (Muller et al., 1996; Palmeirim et al., 1997). The hox genes are also highly conserved and may regulate segmental identity in both vertebrates and invertebrates (Krumlauf, 1994). However, genes that are involved in the establishment of segment polarity in the Drosophila embryo, such as hedgehog, do not appear to be involved in somite segmentation (Ingham, 1995). Thus, while there are similarities between the mechanisms that establish segments and assign their identities in vertebrates and invertebrates, it is not yet clear how anteroposterior (AP) regionalisation of segments occurs in vertebrates.

The establishment of anterior and posterior differences within segments is essential for the development of somites and the correct patterning of a number of other structures. The vertebrae form from the somites, and the peripheral nervous system is also segmentally patterned, with migrating neural crest cells and outgrowing sensory and motor axons restricted to the anterior half of each somite (Keynes and Stern, 1984). The AP polarity of somites is also necessary for the maintenance of segment boundaries, as when half somites are juxtaposed, boundaries only form when anterior and posterior halves are confronted (Stern and Keynes, 1987).

Several genes expressed in the paraxial mesoderm of the vertebrate embryo are known to regulate AP polarity within somites. These include the signalling proteins Notch (Conlon et al., 1995), Delta (Hrabe de Angelis et al., 1997) and lunatic fringe (Evrard et al., 1998; Zhang and Gridley, 1998), and the bHLH transcription factor MesP2 (Saga et al., 1997). Other molecules may also be involved in the processes of morphological boundary formation. These include members of the Eph family of receptor tyrosine kinases and their ligands, the ephrins (Durbin et al., 1998). The expression patterns of these and other genes also show that a segmental pattern is established within the paraxial mesoderm before morphological boundaries are evident. Although rigorous fate mapping of the presomitic mesoderm has not been carried out, it appears that segments defined by the expression of genes in the presomitic mesoderm subsequently correspond to somites (Muller et al., 1996).

There are a number of zebrafish mutations that lead to defects in somite boundary formation (van Eeden et al., 1996). No somite boundaries form early during development in fused somites (fss⁻); whereas in beamter (bea⁻), deadly seven (des⁻) and after eight (aei⁻) only the first four, seven and eight somites form respectively. Homozygous mutants are viable for all four of these mutant loci. Since vertebrae form with a relatively normal periodic pattern in these mutants (van Eeden et al., 1996), it is presumed that the genes function downstream of the processes that establish a segmental pattern in the paraxial mesoderm. Instead the mutations could affect AP patterning events within each segment (van Eeden et al., 1998).

In this paper, we describe the expression pattern of a zebrafish MesP-related gene, mesp-a, which reveals that each segment is AP regionalised following the establishment of a segmental prepattern in the paraxial mesoderm and before somite boundaries form. We show molecular evidence for the establishment of a segmental prepattern in fss⁻, bea⁻, des⁻ and aei⁻ embryos. However, AP regionalisation within segments is disturbed in these mutants. In fss⁻, segments appear posteriorized whereas in bea⁻, des⁻ and aei⁻ there is a loss of polarity, with markers of anterior and posterior segment identity expressed throughout each segment. Since somite formation is disrupted in these mutants, AP regionalisation within segments would appear to be required for somite boundary formation. In support of this, we rescue boundary formation in fss⁻ embryos through ectopic expression of EphA4, an anterior segment marker, within the paraxial mesoderm, demonstrating that boundary formation requires an interface between cells of anterior and posterior segment identity, and that Eph signalling molecules may underlie this AP character.

Breeding fish were maintained at 28.5°C on a 14 hour light/10 hour dark cycle. Embryos were collected by natural spawning and staged according to Kimmel et al. (1995). The adult viable recessive mutants after eight^tm223 (aei−), beamter^tm98 (bea−), deadly seven^tw239 (des−) and fused somites^te314a (fss−) were used (van Eeden et al., 1996).

mesp-a cDNA was isolated from a zebrafish cDNA library (provided by Dr David Grunwald) in a screen for bHLH genes (Gering et al., 1998). The probe used was a 160 bp DNA fragment amplified from zebrafish genomic DNA using primers 5′ GTCTTCACCAACAGCCG 3′ and 5′ CCAGGAAGTTGATGTACTT 3′, defined according to the 5′ and 3′ ends of the SCL-related bHLH gene SLP1 (Gottgens et al., 1998). Three different cDNAs were isolated, one of which encoded a protein that possessed a bHLH region similar to that of the murine MesP proteins and was named Mesp-a. The DNA sequence is available from GenBank under accession number AF188833 (also assigned AB037939).

Single whole-mount in situ hybridisations were carried out following the protocol of Thisse et al. (1993). Double whole-mount in situ hybridisations were carried out according to the method of Hauptmann and Gerster (1994).

Detection of biotin-dextran in transplanted embryos was carried out as described in Westerfield (1993) using the ABC Vectastain Kit (Vector Laboratories, Inc.).

Embryos from a cross between fss^+/− × fss^+/− fish were labelled at the 1- to 4-cell stage with a mixture of rhodamine-dextran and biotinylated-dextran (Molecular Probes). Transplants were carried out as described in Ho and Kane (1990) using unlabelled host embryos, also from a fss^+/− × fss^+/− cross. Donor and host embryos were allowed to develop until between 8 and 12 somites, when they were genotyped morphologically. Host embryos were in situ hybridised with an fgf8 probe and processed for biotin detection.

In other experiments, wild-type donor embryos were injected at the 1- to 2-cell stage with 200 pl of mRNA encoding an EphA4-GFP fusion construct at 1000 ng/μl or GFP at 200 ng/μl. The fusion construct was made such that the GFP was in-frame at the 3′ end of the receptor. Transplantations were carried out as described above, using wild-type embryos and embryos generated from a cross of fss−/− × fss−/− fish as hosts. Only cells positive for EphA4-GFP or GFP, as determined by fluorescence, were transplanted. Host embryos were allowed to develop to 8-10 somites. Those in which there were cells positive for EphA4-GFP or GFP in the paraxial mesoderm were examined morphologically by DIC optics.

In order to isolate MesP-related genes involved in somite patterning, we performed a low-stringency screen of a zebrafish cDNA library using a PCR-amplified fragment of a zebrafish bHLH gene. One cDNA was isolated, the bHLH region of which showed homology to the murine proteins MesP1 (Saga et al., 1996) and MesP2 (Saga et al., 1997). This gene was named mesp-a.

The bHLH domain of Mesp-a is 76% identical at the amino acid level to equivalent domains of mouse MesP1 and MesP2, and shows 73%, 71% and 67% identity to equivalent domains of the related proteins Thylacine 1, Thylacine 2 (Sparrow et al., 1998) and cMeso-1 (Buchberger et al., 1998), respectively. None of these MesP-related genes show significant homology outside of the bHLH domain, unlike other families of bHLH proteins (Fig. 1), making it difficult to determine orthology within the MesP family.

Expression of mesp-a is initiated at 40% epiboly around the edge of the advancing blastoderm with the exception of cells at the future dorsal region of the embryo (Fig. 2A). Following the start of gastrulation, cells expressing mesp-a involute and expression is seen in the presumptive paraxial mesoderm. At 70% epiboly expression starts to be restricted anteriorly, as mesp-a is lost from the posterior paraxial mesoderm (Fig. 2B,C). By bud stage, narrow stripes of mesp-a expression are present in the anterior presomitic mesoderm (Fig. 2D).

Throughout somitogenesis, expression of mesp-a is highly dynamic, with one or two stripes of expression in the anterior presomitic mesoderm of embryos at morphologically equivalent stages of development (Fig. 2E-H). When two stripes of mesp-a expression are visible, the most anterior stripe is predicted to be in somite −I (S−I), where SI is the newest formed somite, S0 is the somite in the process of forming, and S−I is the next most posterior presumptive somite, as determined by the position of the last formed somite boundary (Christ and Ordahl, 1995). The posterior stripe of mesp-a expression is broader than the anterior one, suggesting that mesp-a expression may be restricted within segments as they mature (Fig. 2F). Indeed, mesp-a is initially expressed in a stripe of 5-6 cell diameters along the AP axis, which may encompass all of S−II (Fig. 2F,G). During the time that it takes one somite to form, this stripe of expression is restricted rostrally within the segment to a stripe of 2-3 cell diameters along the AP axis (Fig. 2E). These narrow stripes of mesp-a are seen in S−I (Fig. 2E-G). During the formation of the next somite, this anterior stripe of expression is lost, with expression persisting for the longest time in the adaxial cells in the anterior of the segment (Fig. 2H). These observations suggest that the anterior stripe of mesp-a constitutes a subset of anterior cells of S−I, whereas the posterior stripe constitutes most or all cells in S−II.

Analysis of double in situ hybridisations between mesp-a, her1 and delta D support the interpretation that mesp-a expression is restricted to the anterior region of S−I. her1 has been described as being expressed in alternating segments (Muller et al., 1996). When two stripes of mesp-a expression are visible, the anterior stripe is a subdomain of the most anterior domain of her1 expression; whereas the posterior stripe of mesp-a is in the segmental domain which is devoid of her1 expression, S−II (Fig. 2J). When one stripe of mesp-a is visible, this constitutes an anterior subset of the cells expressing her1 and delta D in S−I (Fig. 2I,K).

In summary, our data suggest that mesp-a is initially expressed throughout S−II, then is restricted to anterior cells as the domain becomes S−I, and expression is lost by the stage that the cells are in S0. The restriction of mesp-a expression from throughout S−II to an anterior domain within S−I suggests that each segment is regionalised along its AP axis at this time in their development.

The expression pattern of mesp-a during somitogenesis indicates that each segment is patterned along its AP axis following the establishment of a segmental prepattern in the paraxial mesoderm. In zebrafish mutants that show a disruption of somite boundary formation (van Eeden et al., 1996) the defects must lie in the establishment of a segmental prepattern, in the AP regionalisation of each segment, or in the later step of boundary formation. Homozygous fss⁻, bea⁻, des⁻ and aei⁻ mutants form vertebrae with an almost normal periodicity (van Eeden et al., 1996), suggesting that the mutated genes function downstream of the processes in which a segmental prepattern is established in the paraxial mesoderm. To determine whether there is any molecular evidence of the establishment of a segmental prepattern in these mutants during embryogenesis, we examined the expression patterns of mesodermal marker genes in fss⁻, bea⁻, des⁻ and aei⁻.

In agreement with van Eeden et al (1998) we found in fss⁻, as in wild-type embryos, that her1 is expressed in a segmental pattern (Fig. 3B). However, the two stripes of her1 expression present in wild-type embryos, were not seen in fss⁻ embryos. This suggests that a segmental prepattern is initially established in fss⁻, but that patterning in the presomitic mesoderm is disturbed. In contrast to fss⁻, her1 was not obviously segmentally expressed in bea⁻, des⁻ or aei⁻. In bea⁻, her1 was expressed throughout the tail bud and posterior presomitic mesoderm (Fig. 3C), whereas in des⁻ and aei⁻, her1 was expressed strongly in the tailbud, more weakly in the posterior presomitic mesoderm, and in a “salt and pepper” fashion in the anterior presomitic mesoderm (Fig. 3D,E). Although segmental expression of her1 is lost in bea⁻, des⁻ and aei⁻, from the pattern of the vertebrae that form in these mutants, we know that this does not equate to a complete loss of a segmental prepattern.

Analysis of CS131 expression confirmed decisively that a segmental prepattern is established in all of these somite mutants. CS131 is a member of the group of small acidic proteins, which include GADD45 and myd118, that cause cell cycle arrest (reviewed by Fornace, 1992). During somitogenesis CS131 is expressed in a posterior domain of formed somites, and in a stripe in the presomitic mesoderm which, from its position, appears to encompass S−II (Fig. 3F). In fss⁻, bea⁻, des⁻ and aei⁻ a stripe of CS131 expression is present in the paraxial mesoderm in a position that corresponds to the presomitic stripe in wild-type embryos (Fig. 3G-J), showing that a segmental prepattern is established in the presomitic mesoderm of all of these mutants.

Analysis of hox gene expression revealed that AP identities are maintained along the AP axis of fss⁻, bea⁻, des⁻ and aei⁻ embryos. In wild-type embryos, hoxb6 is expressed posterior to the boundary between somites 3 and 4 (Prince et al., 1998; Fig. 3K) and hoxa10 is expressed caudal to the somite 11/12 border (Sordino et al., 1996). In fss⁻, bea⁻, des⁻ and aei⁻, both hoxb6 and hoxa10 are expressed in the paraxial mesoderm in domains that correspond to the wild-type expression domains, despite the lack of somite boundaries (Fig. 3L-O and data not shown). This suggests that patterning along the AP body axis respects segment borders in the paraxial mesoderm of these mutants, even when somite boundaries are not present.

As the establishment of a segmental prepattern in the presomitic mesoderm does not appear to be disturbed in fss⁻, bea⁻, des⁻ and aei⁻ mutants, it is possible that segments are not correctly patterned following their specification. To address this possibility, we examined the expression of mesp-a, EphA4 (Durbin et al., 1998), delta D (Dornseifer et al., 1997) and fgf8/ace (Furthauer et al., 1997; Reifers et al., 1998) markers of anterior segment identity, and ephrin-B2 (Durbin et al., 1998) and myoD (Weinberg et al., 1996) markers of posterior segment character, in the mutants.

A loss of anterior markers and expression of posterior markers throughout the anterior presomitic mesoderm suggests that the somitic mesoderm of fss⁻ is posteriorized. mesp-a and EphA4 expression were absent in anterior segmental domains in fss⁻ (Fig. 4B,F). Weak, diffuse bands of delta D and fgf8 expression were observed in the anterior presomitic mesoderm, and expression of both was lost in the region equivalent to that where somites have formed in wild-type embryos (Fig. 4D,J). As delta D and fgf8 are only anteriorly restricted in wild-type embryos as the segments mature (Fig. 4C,I), the residual expression in fss⁻ may correspond to this unrestricted early expression domain. In contrast, markers of posterior segment identity, ephrin-B2 and myoD, were expressed throughout the somitic mesoderm of fss⁻ (Fig. 4H,L). Thus the AP regionalisation of segments is disrupted in fss⁻ and the somitic mesoderm appears posteriorized.

The regionalisation of segments along the AP axis is also disrupted in bea⁻, des⁻ and aei⁻, but unlike in fss⁻, both anterior and posterior segment markers are expressed throughout the somitic mesoderm. mesp-a expression was dramatically reduced and present in broad diffuse stripes in these mutants (Fig. 5B-D). The segmental restriction of other anterior and posterior segment markers was also lost. EphA4, delta D and ephrin-B2 were expressed throughout the region in which they are normally AP restricted within segments (Fig. 5F-H,J-L, N-P). Double in situ hybridisations with EphA4 and ephrin-B2 showed that cells in this area were expressing markers of both anterior and posterior segment identity (Fig. 5R). Thus AP regionalisation within segments is disturbed in bea⁻, des⁻ and aei⁻, resulting in a loss of polarisation of segments.

To determine whether fss acts in a cell autonomous manner, we transplanted labelled wild-type or fss⁻ cells into wild-type and mutant embryos. fgf8 expression, boundary formation and the location of transplanted cells were subsequently determined. Wild-type cells transplanted into a wild-type host dispersed throughout the somitic mesoderm and contributed normally to anterior and posterior domains within somites, as expected (Fig. 6A). fss⁻ cells transplanted into a wild-type host also dispersed throughout the paraxial mesoderm and contributed to all regions of the somites, including the boundaries as determined by the position and epithelial morphology of cells (Fig. 6B). This suggests that fss acts in a cell non-autonomous manner and the fss⁻ cells were rescued through being in a wild-type environment. This raised the possibility that the presence of wild-type cells in a fss⁻ mutant host may be able to rescue the mutant phenotype. However, no obvious rescue was visible when wild-type cells were transplanted into mutant hosts, as detected by fgf8 expression or morphological boundaries (Fig. 6C). Of course it remains possible that there were not enough wild-type cells or they were positioned in an inappropriate place to bring about rescue. It does suggest, however, that fss acts in a cell non-autonomous manner in the patterning of segments along their AP axis.

The loss of somite boundary formation in fss⁻ correlates with a loss of cells with anterior segment identity. We have shown previously that the juxtaposition of cells expressing EphA4 and ephrin-B2 is required for somite boundary formation (Durbin et al., 1998). This interface is lost in fss⁻ (Fig. 4). Therefore, we attempted to rescue boundary formation in fss⁻ by ectopically expressing EphA4 in the paraxial mesoderm.

Wild-type donor embryos were injected with RNA encoding an EphA4-GFP fusion construct. To test that the EphA4-GFP protein was functional, we showed that it could rescue the effects of overexpression of a co-injected dominant negative receptor (data not shown). Cells expressing ectopic EphA4, as detected by fluorescence, were transplanted into wild-type and fss⁻ hosts. Host embryos were allowed to develop to 8-10 somites, when the morphology of their somitic mesoderm in relation to the presence of transplanted cells was examined.

When cells expressing ectopic EphA4 were scattered through the paraxial mesoderm in fss⁻ embryos, the somitic mesoderm appeared unsegmented as it typically does in the mutants (n=15; Fig. 7A). However, when clusters of transplanted cells were present, morphological boundaries were visible where the cells expressing ectopic EphA4 were in contact with the host cells (n=30; Fig. 7B-D). Only two fss⁻ embryos with small clumps of cells expressing EphA4 in the paraxial mesoderm did not form ectopic boundaries. Boundaries were sometimes only visible on one side of the group of transplanted cells (Fig. 7B), suggesting that boundaries may only form on one side of the polarised cells.

No ectopic boundaries were seen when cells expressing ectopic EphA4 were transplanted into wild-type embryos (n=22; Fig. 7E). Boundaries also did not form in control experiments, when wild-type cells expressing GFP were transplanted into fss⁻ hosts (n=14; Fig. 7F). Thus it is possible to rescue boundary formation in the paraxial mesoderm of fss⁻ embryos by the apposition of adjacent domains of cells expressing EphA4 and ephrin-B2.

The expression of mesp-a shows that segments are internally regionalised along their AP axis by the stage that cells have arrived in the anterior presomitic mesoderm. This precedes somite boundary formation, with borders subsequently forming between cells of anterior and posterior character in adjacent segments. In fss⁻, bea⁻, des⁻ and aei⁻ mutants, we show that a segmental plan is established in the paraxial mesoderm, but that patterning within segments is disturbed and somite boundary formation fails to occur. This suggests that the AP regionalisation of each segment is required for somite boundary formation. In support of this, we were able to rescue boundary formation in fss⁻ embryos by the reconstitution of adjacent domains of cells expressing anterior and posterior segment markers within the paraxial mesoderm, by ectopically expressing EphA4. This suggests that signalling through Eph receptors is sufficient to bring about morphological boundary formation in the paraxial mesoderm.

Several models have been proposed to explain how a segmental pattern is established within the paraxial mesoderm. One of these, the clock and wavefront model, proposes that cells in the mesoderm are continuously oscillating, this being the clock. The interaction of this clock with a wavefront moving along the axis of the embryo results in the definition of regularly sized segments (Cooke and Zeeman, 1976). The identification of genes, c-hairy1 (Palmeirim et al., 1997) and lunatic fringe (Forsberg et al., 1998; McGrew et al., 1998), which show cyclical expression in the presomitic mesoderm in chick and mouse, suggests the existence of a developmental clock linked to segmentation. Until recently no zebrafish genes had been convincingly shown to exhibit cyclical expression, although there is now evidence to suggest that some elements of the expression of her1 are cyclical (Sawada et al., 2000). This raises the possibility that a clock linked to somitogenesis also exists in zebrafish.

A variation of the clock and wavefront model was put forward by Primmet et al. (1989), in which the hypothetical oscillator is proposed to be the cell cycle. In this paper, we have described the expression pattern of CS131, a gene belonging to a family of proteins involved in regulating the cell cycle, in a restricted segmental pattern in the presomitic mesoderm during somitogenesis. This suggests that there could be a link between the cell cycle and segmentation. The expression pattern of CS131 is the first molecular evidence that supports the involvement of the cell cycle in the control of segmentation, although functional experiments will be required to address if the gene truly functions during the initial segmentation process or in events downstream of this. However, the segmental pattern of CS131 is not significantly disturbed in fss⁻, bea⁻, des⁻ and aei⁻ mutants, suggesting that any involvement of the cell cycle in somitogenesis is upstream of the AP regionalisation of segments.

A further model that addresses the question of how and where somite boundaries form, rather than how a segmental pattern is established in the presomitic mesoderm, is Meinhardt’s positional information model. This proposes that gene expression in the presomitic mesoderm is initially oscillating between two states, anterior and posterior, before the fate of cells is fixed such that there are groups of anterior cells adjacent to groups of posterior cells. Boundaries consequently form where cells with different identities are opposed (Meinhardt, 1986). Our results confirm that somite boundary formation does indeed occur as a result of cells with anterior segment identity being juxtaposed to cells of posterior character. However, we have found no evidence to suggest that cells are initially oscillating between these two fates. The expression pattern of mesp-a suggests that this anterior segment marker is initially expressed throughout prospective segments, and the fact that the somitic mesoderm appears posteriorized in fss⁻ shows that all cells are also capable of adopting the posterior fate. This suggests that the AP regionalisation of segments involves a determination event in which cells differentiate from a precursor state to either an anterior or posterior identity, all cells being competent to adopt either fate.

The somitic mesoderm of fss⁻ embryos appears posteriorized and transplantation experiments have suggested that fss functions in a cell non-autonomous manner, suggesting that an intercellular signal is required for the AP regionalisation of segments. The involvement of the Notch signalling system in somitogenesis has been known since mice homozygous for a null mutation in Notch1 were found to have severe somitic defects (Conlon et al., 1995). More recently, analyses of mice mutant in the Notch ligands Dll1 and Dll3, and in lunatic fringe, have shown that the AP polarity of somites is disturbed in all of these (Hrabe de Angelis et al., 1997; Kusumi et al., 1998; Evrard et al., 1998; Zhang and Gridley, 1998). This suggests an involvement of the Notch signalling system in AP regionalisation within segments, and provides a candidate for the intercellular signalling system that is disrupted in fss⁻, and the other mutants that exhibit defects in the AP patterning of segments, bea⁻, des⁻ and aei⁻. Analysis of double mutant combinations between fss⁻, bea⁻, des⁻ and aei⁻ has shown that there is no redundancy between them, and this also suggests that all of these genes function within the same pathway (van Eeden et al., 1998). Recent analysis has suggested that a Notch-mediated signal may be required early in the process of somite formation, possibly in the establishment of the segmental prepattern (Jen et al., 1999; Takke and Campos-Ortega, 1999), whereas Delta and Fringe homologues appear to act downstream of this in the AP regionalisation within segments (del Barco Barrantes et al., 1999; Takke and Campos-Ortega, 1999). This suggests that Delta or Fringe signals could be disrupted in fss⁻, bea⁻, des⁻ and aei⁻ mutants, and indeed the expression of delta D is disturbed in all of these mutants.

The molecular identity of the fss, bea, des and aei genes remains unknown at present. It is possible that one or more of these genes are the zebrafish orthologues of the genes that when disrupted in mouse have resulted in related or comparable somite phenotypes. These mammalian genes include Notch1 (Conlon et al., 1995), Dll1 (Hrabe de Angelis et al., 1997), lunatic fringe (Evrard et al., 1998; Zhang and Gridley, 1998) and MesP2 (Saga et al., 1997). In mice homozygous for null mutations in any of these genes, the anterior somites form but the posterior ones fail to segment from the paraxial mesoderm. Thus, morphologically, these mouse mutants resemble the zebrafish mutants bea⁻, des⁻ and aei⁻. At a molecular level, disruptions in Dll1, lunatic fringe and MesP2 all result in defects in the AP patterning of segments, similar to the zebrafish mutants studied here. In mice homozygous for a null mutations in Dll1 or lunatic fringe, markers of anterior and posterior segment identity are expressed in broad, diffuse regions (Hrabe de Angelis et al., 1997; Evrard et al., 1998; Zhang and Gridley, 1998; del Barco Barrantes et al., 1999), similar to what is seen in bea⁻, des⁻ and aei⁻. In mice in which MesP2 has been disrupted, anterior segment markers are lost (Saga et al., 1997), similar to what has been described here for fss⁻.

We have previously demonstrated that disturbance of Eph signalling results in loss or disruption of somite boundary formation (Durbin et al., 1998). In fss⁻, the somitic mesoderm appears posteriorized, and expression of EphA4 is lost whilst ephrin-B2 is expressed throughout this region. Clusters of cells ectopically expressing EphA4 in the paraxial mesoderm of fss⁻ embryos rescued boundary formation. Thus signalling through the Eph system is sufficient to cause morphological boundary formation in the mesoderm. Ectopic boundaries were sometimes only seen on one side of a group of transplanted cells, suggesting that cells may become polarised during boundary formation, with morphological boundaries only forming on one side of the polarised cells.

Experiments using cultured explants of chick presomitic mesoderm have shown that the establishment of a segmental prepattern and the patterning of segments along their AP axis can occur in the absence of somite boundary formation (Palmeirim et al., 1998). Thus there must be molecules that link these events of pattern formation and morphological segmentation in vivo. The receptor EphA4 and its ligand ephrin-B2, are good candidates to fulfil this role. Their expression in anterior and posterior segment halves, respectively, indicates that these molecules lie downstream of the signals that lead to AP polarisation of segments.

When cells expressing ectopic EphA4 were transplanted into the paraxial mesoderm of wild-type embryos no ectopic boundaries formed, although it might be predicted that boundaries should form between transplanted cells and the host’s posterior segment cells. This is most probably because the ectopic EphA4 could not overcome the endogenous patterning mechanisms that specify boundary formation at every alternate interface of anterior and posterior segment cells. In contrast, in fss⁻ embryos, ectopic boundaries formed wherever the transplanted cells were located, not just at the positions of segment borders. Thus the paraxial mesoderm remains more responsive to somite-boundary-forming signals in fss⁻ than wild-type embryos.

The isolation of mutants such as bea⁻, des⁻ and aei⁻ in which only the most anterior somites form, suggests that the first few somites could be specified by a different mechanism to the more posterior ones. We have shown here that a segmental prepattern is correctly established in the paraxial mesoderm of all of these mutants, and it is patterning events within segments that is disturbed in each case. This suggests that the most anterior somites could be patterned internally by a different mechanism to the more posterior ones. However, as AP regionalisation within every segment is disrupted in fss⁻, there must be some underlying similarities between the patterning mechanisms in all somites.

The expression pattern of mesp-a also supports the idea that anterior and posterior segments are patterned differently. Analysis of the expression pattern of her1 has shown that the first somites are specified during gastrulation. In fact, the first stripe of her1 expression, which corresponds to the fifth somite, is visible at 70% epiboly (Muller et al., 1996). This suggests that mesp-a does not function during the AP patterning of the first somites to form, since it is not segmentally expressed until around 80% epiboly. This again suggests that the first few somites could be patterned differently along their AP axis to the more posterior ones. It will be of great interest to isolate and clone further mutants that help resolve the conserved and divergent mechanisms regulating somite formation along the AP axis.

We thank our colleagues in the Holder and Wilson laboratories for comments and criticisms during the course of this work, Dr Tom Schilling for technical help and comments on the manuscript, Dr Adam Rodaway for initial characterisation of mesp-a, Dr Hiro Takeda and colleagues for sharing their results prior to publication, many colleagues for providing probes and reagents used in this study, V. Heyer and T. Steffan for technical assistance, and B. Gasking, R. Mak, O. Nkundwa and A. Karmim for maintenance of zebrafish. This work was supported by grants from the Wellcome Trust and the BBSRC (to N. H. and S. W.), and funds from the Institut National de la Recherche Médicale, the Centre National de la Recherche Scientifique, the Hôpital Universitaire de Strasbourg, the Association pour la Recherche sur le Cancer and the Ligue Nationale Contre le Cancer (to C. T. and B. T.). P. S. receives fellowship support from EMBO and the EC.