Zebrafish Mesp family genes, mesp-a and mesp-b are segmentally expressed in the presomitic mesoderm, and Mesp-b confers the anterior identity to the developing somites

During vertebrate embryonic development, the paraxial mesoderm is subdivided into metameric subunits called somites. The somites are the first segmented structures to form during embryogenesis and they govern the metamerism of all somite-derived tissues; axial skeleton, the dermis of the back, and all striated muscle of the adult body (Christ and Ordahl, 1995) and spinal ganglia (Keynes and Stern, 1984). Individual pairs of somites, located symmetrically on either side of the neural tube, are formed in a rostrocaudal progression within the presomitic mesoderm (PSM). It is believed that the process of somitogenesis can be divided into three distinct stages, which may be regulated by different genetic mechanisms (Tam and Trainor, 1994). (1) Specification as paraxial mesoderm: the mesoderm derived from the primitive streak in mice and chick, the marginal zone in amphibians, and germ ring in fish or the tailbud, is arranged on both sides of the neural tube as the paraxial mesoderm; (2) segmentation: the paraxial (somitic) mesoderm generates each segmental border and is divided into the so-called epithelial somites; (3) differentiation: the somites differentiate into the sclerotome and dermomyotome which subsequently segregates into the dermatome and myotome.

Recently, signalling molecules such as Shh, BMP and Noggin have been identified and implicated in patterning and differentiation within the somites (for review see Currie et al., 1998). Molecular and genetic analyses of zebrafish no tail (ntl), floating head (flh) and spadetail (spt) mutants have shed light on the mechanism underlying a cell fate decision between axial and paraxial mesoderm (Halpern et al., 1997; Talbot et al., 1995; Amacher et al., 1998).

The mechanisms of segmentation have been studied most extensively in chick embryos. Transplantation experiments in chick have shown that the anteroposterior subdivision within the somite, which has already been established during segmentation period (Stern and Keynes, 1987; Aoyama and Asamoto, 1988), serves to maintain the segmental arrangement of the somite. The understanding of the mechanisms that establish a segmental prepattern in the PSM has been greatly advanced in the last few years. The remarkably cyclic expression pattern of c-hairy1, a chick bHLH transcription factor, related to the Drosophila pair-rule gene hairy, in the PSM suggests the presence of an intrinsic molecular clock before the appearance of the somite (Palmeirim et al., 1997). Furthermore, the unique expression pattern of zebrafish her1, more distantly related to hairy, suggested a pair-rule prepattern in zebrafish PSM: it is expressed in two or three stripes demarcating the primordia of the odd numbered somites beginning with the 5th somite (Müller et al., 1996). Recent genetic analyses further indicate that segmentation of the PSM involves a cell-cell interaction mediated by the Notch receptors and their ligands. For example, segmental defects are seen in mouse embryos with a targeted inactivation of the genes encoding Dll1 (Hrabe de Angelis et al., 1997), Notch1 (Conlon et al., 1995), lunatic fringe (Zhang and Gridley, 1998; Evrad et al., 1998), and a component of the Notch-Delta signalling pathway called RBP-Jk or Su (H) (Oka et al., 1995). Recently, a dynamic expression pattern has been demonstrated for lunatic fringe in chick and mouse PSM, which resembles c-hairy1 expression, suggesting a link between the segmentation clock and the Notch signalling pathway (McGrew et al., 1998; Forsberg et al., 1998; Aulehla and Johnson, 1999; del Barco Barrantes et al., 1999). Segmental defects are also detected in mice deficient in fibroblast growth factor receptor1 (FGFR1; Deng et al., 1994; Yamaguchi et al., 1994). Furthermore, recent experiments with zebrafish suggest a role of Eph signalling in segmentation (Durbin et al., 1998). All these genes are expressed in the PSM and implicated in early steps of somitogenesis.

Murine MesP1 and MesP2 are bHLH transcription factors (Saga et al., 1996, 1997) and they show a segmental expression pattern in the PSM. Mesp2-deficient mice showed two major defects during somitogenesis: a failure of the initiation of segmentation, and a disturbance of the somitic anteroposterior polarity. Recently, Mesp-related genes were cloned in chick (cMeso-1; Buchberger et al., 1998) and Xenopus (Thylacine; Sparrow et al., 1998; Mespo; Joseph and Cassetta, 1999). Like murine Mesp genes, they are predominantly expressed in the PSM, and both antisense and overexpression experiments suggest they have roles in segmentation. The exact nature of their role, and how their expression is regulated are important issues for understanding how paraxial mesoderm is segmented in different vertebrate species.

In this paper, we describe two zebrafish Mesp-related genes, mesp-a and mesp-b. Their expression in the PSM is confined to the anterior halves of presumptive and/or forming somites and is regulated by fss (van Eeden et al., 1996). Two-colour in situ hybridization with mesp and her1 probes suggests that, like her1, mesp-a is not expressed in the cells that will contribute to the first several somites, supporting an idea of a difference between rostral and caudal somite formation (reviewed in Jiang et al., 1998). Furthermore, we found that the newly established expression domain of her1, near the tailbud, travels (or moves) anteriorly and finally overlaps the mesp expression domain. Thus, like mesp, her1 transcripts are detected in every somite primordium (except for the first several ones) preceding boundary formation. Finally, ectopic expression experiments suggest that Mesp-b is involved in establishing the anterior fate within the presumptive somites.

All studies on wild-type zebrafish (Danio rerio) were carried out in the Oregon AB background. The mutant alleles used in this paper are fused somites (fss^ti1), beamter (bea^tw212b), deadly seven (des^th35b), after eight (aei^tr233) and mindbomb (mib^ta52b). Embryos obtained from natural crosses were maintained in 1/3 Ringer’s solution (39 mM NaCl, 0.97 mM KCl, 1.8 mM CaCl₂, 1.7 mM Hepes at pH 7.2) at 28.5°C and staged according to age (hours postfertilization at 28.5°C) and morphological criteria (Kimmel et al., 1995).

Degenerate oligonucleotide primers were designed for the amino acid sequences conserved between MesP1 and MesP2 proteins. A sense primer (5′ GA(AG)(AC)G(AGCT)GA(AG)AA(AG)(CT)T(AGCT)-(AC)G 3′) encoding the amino acid sequence EREKLR and an antisense primer (5′ (GC)(AT)(AGCT)A(AG)(AG)TG(AGCT)CC-(AGT)AT(AG)TA 3′) encoding the amino acid sequence YIGHLS were used for amplification.

Total RNA from the shield-stage embryos was prepared by the acid guanidium thiocyanate/phenol/chloroform extraction method (Chomczynski and Sacchi, 1987). First-strand cDNA was synthesized by using 10μg of total RNA and oligo-dT primer (RTG cDNA synthesis kit, Pharmacia). For PCR, 1/10 volume of the first-strand cDNA was used for PCR amplification. One fragment gave an unexpected large product which hybridized with the murine Mesp1 probe. The fragment was then cloned into pBluescript II SK+ (Stratagene) and sequenced. The fragment was found to contain the bHLH region similar to that of murine MesP1 as well as the 3′ entire UTR of the gene. This cDNA was designated as zebrafish mesp-a. The 5′ fragment was obtained using a CapFinder_TM PCR cDNA Library Construction Kit (CLONETECH Laboratories, Inc. Palo Alto, CA). During the course of the study, we screened a PAC genomic library and we found a PAC clone that contained both mesp-a and a second Mesp-related gene. From the partial sequence of this clone, we isolated a complete cDNA clone of the second gene by PCR. The second Mesp-related gene was designated as mesp-b. Sequences of the two Mesp-related cDNAs were determined by comparing sequence data from two independently amplified clones to exclude PCR errors.

All capped sense RNAs were synthesized and purified as previously described (Makita et al., 1998; Koshida et al., 1998). The synthesized mRNAs were diluted to the appropriate concentration with distilled water and injected into 1-cell stage embryos. The concentration used for each RNA was as follows: 0.1 μg/μl for mesp genes, 0.1 mg/ml for green fluorescent protein (GFP); 0.1 μg/μl for lacZ (except for the samples shown in Fig. 5E-H in which 0.02 μg/μl of lacZ RNA were coinjected). Approximately 400-500 pl of diluted RNA was injected into each embryo.

The embryos injected with lacZ RNA were fixed with 4% paraformaldehyde in PBS for 30 minutes at room temperature. They were washed 4 times with PBST for 10 minutes each time and washed once with buffer A (17.5 mM K₃[Fe(CN)₆], 17.5 mM K₄[Fe(CN)₆], 1 mM MgCl₂ in PBS) for 10 minutes. Colour staining was performed using 0.4 mg/ml dilution of X-gal in buffer A for 2 hours at 37°C. They were washed three times with PBST for 5 minutes each time and refixed with 4% paraformaldehyde in PBS at 4°C overnight. The embryos were then processed for in situ hybridization.

Single colour in situ hybridization was carried out as described by Koshida et al. (1998). Two-colour in situ hybridization was performed essentially as described by Jowett and Yan (1996). Briefly, one antisense RNA probe was labeled with digoxigenin-UTP (Roche) and the other was labeled with fluorescein-UTP (Roche). Hybridization was performed with both probes to the embryos simultaneously. After reaction with anti-fluorescein alkaline phosphatase Fab fragments at 4°C overnight, colour staining was performed with Fast Red (Roche). Then, alkaline phosphatase was inactivated by incubation with 0.1 M glycine-HCl (pH 2.2), 0.1% Tween 20 twice for 15 minutes each time at room temperature. After washing with MABT (100 mM maleic acid, 150 mM NaCl, 0.1% Tween 20) 4 times for 5 minutes each time, the embryos were fixed with 4% paraformaldehyde in PBS for 20 minutes at room temperature. The samples were washed 5 times with MABT for 5 minutes each time, and reacted with anti-digoxigenin alkaline phosphatase Fab fragments at 4C overnight. Second colour staining was performed with BM purple (Roche) or ELF^TM-97 mRNA In Situ Hybridization Kit (Molecular Probes).

A partial cDNA fragment of a zebrafish Mesp-related gene was obtained from shield stage cDNAs by PCR with degenerate primers corresponding to the bHLH region of murine Mesp1. The cDNA fragment contained the Mesp-related bHLH region and was designated as mesp-a. The second related gene, mesp-b, was obtained from somite-stage cDNA by PCR. The primers used were designed based on a partial sequence of a DNA fragment derived from a PAC clone containing mesp-a. The mesp-a cDNA, which consists of 887 base pairs (bp), encodes a protein of 223 amino acids from the first possible initiation site at the 24th nucleotide (nt) (Fig. 1A). Similarly, translation from the ATG at the 13th nt in the 930 bp mesp-b cDNA will yield a protein of 236 amino acids (Fig. 1A). The sequences surrounding the first in-frame methionine have a good match to the Kozak consensus (Kozak, 1987) and hence this is most likely to be the translation initiation site. The bHLH domains of these two proteins share more than 90% amino acid identity (46/51), while no homology is seen outside the bHLH domain (Fig. 1A).

A database search with the deduced amino acid sequences revealed several closely related bHLH family members (Fig. 1B). The Mesp-a (Mesp-b) bHLH domain showed 73% (72%), 71% (69%), 55% (57%), 67% (65%), 76% (75%) and 75% (73%) identity to those of Thylacine 1, Thylacine 2 (Xenopus; Sparrow et al., 1998), Mespo (Xenopus; Joseph and Cassetta, 1999), cMeso-1 (chick; Buchberger et al., 1998), MesP1 and MesP2 (mouse; Saga et al., 1996 and 1997), respectively. In contrast, it is only 47% (45%), 33% (31%) and 31% (28%) identical to zebrafish NeuroD (Blader et al., 1997; Korzh et al., 1998), MyoD (Weinberg et al., 1996) and Her1 (Müller et al., 1996), respectively (Fig. 1B). Essentially no homology to those or any other proteins was found outside of the bHLH domain at the nucleotide or protein level. Based on this similarity and the phylogenetic tree (Fig. 1C), we concluded that zebrafish mesp-a and mesp-b belong to the Mesp-related subfamily of bHLH transcription factors.

Whole-mount in situ hybridization with digoxigenin-labeled or fluorescein-labeled probes was performed to investigate the embryonic expression patterns of these two genes. To avoid cross hybridization between the similar bHLH transcripts, we used probes which excluded the bHLH regions. The nucleotide sequences outside the bHLH of mesp-a and mesp-b differ significantly and the probes used provided distinct expression.

mesp-a transcripts are first seen at 30% epiboly in the blastoderm margin (data not shown). Expression was detected as an homogenous ring around the margin with a small gap (about one-eighth of the ring). This expression pattern was maintained until early gastrula stages (Fig. 2A). Double staining with mesp-a and goosecoid (Stachel et al., 1993) probes at the shield stage revealed that the most dorsal margin is devoid of mesp-a (Fig. 2B). As gastrulation proceeds, the expression at the dorsolateral margin becomes broad along the animal-vegetal axis, while that at the ventral margin declines (Fig. 2C). Histological sections demonstrated that the transcripts were restricted to the hypoblast cells during gastrulation (data not shown). During mid to late gastrulation, dorsolateral expression above the margin becomes narrowed along the dorsal-ventral axis probably due to dorsal convergence movement. Before the end of gastrulation, at about 95% epiboly, a pair of transverse stripes of mesp-a-expressing cells separates from the marginal domain (Fig. 2D, see also Fig. 4A-C). In contrast to mesp-a, mesp-b is not expressed during early to mid gastrulation (Fig. 2G). Expression of mesp-b is first seen at the end of gastrulation in stripes in the paraxial mesoderm. Throughout the segmentation period, expression of mesp-a and mesp-b is detected in stripes of cells located on either side of the neural tube. As segmentation proceeds, the position of the stripes shifted progressively towards the posterior part of embryos by switching off the most rostral stripe near the forming somites and initiating a new expression domain caudal to the most caudal stripe. The expression domain is initially broad but the most anterior stripes are usually narrower. The expression patterns are dynamic and the number of stripes varies between one to three: one to two stripes are most often observed for mesp-a while two to three are seen for mesp-b (Fig. 2E,F,H,I). Two-colour in situ hybridization with mesp-a and mesp-b probes demonstrated that both genes are activated at approximately the same time and the same level during the segmentation period (Fig. 2J,K). However, the transcripts of mesp-b tend to be more persistent, leaving the most anterior stripe of mesp-b located anterior to that of mesp-a (Fig. 2I, see also Fig. 3B,C)

Histological sections of stained embryos show that the mesp transcripts are in the PSM, posterior to the segmented somites (Fig. 3A). To precisely localize the transcripts within the PSM, we compared the expression patterns of mesp-a and mesp-b with other known genes expressed in this specific region. We first used the myoD expression pattern as a reference. myoD is expressed in the posterior parts of somites adjacent to the furrow, and one or two fainter pairs of bands are seen in the posterior parts of the somite just undergoing furrow formation (S0) and more posteriorly in the paraxial mesoderm (Weinberg et al., 1996). Two-colour in situ hybridization shows that, when three bands of mesp-b expression are present, they are located posteriorly adjacent to the posterior faint bands of myoD (Fig. 3C). It was difficult to determine whether or not any cells expressed both myoD and mesp genes. The staining also shows that, unlike mesp-b, mesp-a expression domain includes the adaxial cells that are located on either side of the notochord and are positive for myoD (asterisk in Fig. 3B). We then used paraxial protocadherin (papc; Yamamoto et al., 1998) to localize mesp transcripts within the somite primordia. Zebrafish papc is expressed in four bilateral pairs of bands in the paraxial mesoderm during the segmentation period. The first band is located in the anterior border of the newest somite formed (SI) and the second in the forming somite (S0). The two posterior, stronger bands seem to be located in successive somite primordia. Two-colour in situ hybridization shows that the three bands of mesp-b are located within the posterior three bands of papc and share the same anterior boundary (Fig. 3D,E). Thus, it is likely that the most anterior stripes of mesp-b correspond to the anterior of the forming somite (S0) and that the posterior two stripes are in two successive somite primordia. Accordingly, mesp-a is expressed in one and/or two primordia posterior to the forming somites. The spatial relationship between myoD, papc and mesp expression is schematically represented in Fig. 3F.

The expression patterns of the mesp genes suggest that they function in the segmentation of the paraxial mesoderm. Given that the fss-type genes (van Eeden et al., 1996), identified by mutation, are involved in segmenting the paraxial mesoderm, we examined the expression patterns of mesp-a and mesp-b in mutants of this group. The mutants of fss-type genes, fss, bea, des, aei and mib, exhibit a defect in somite boundary formation. However, the spatial distribution of the defects is different among mutants: fss controls the formation of all somites while the other four only govern the formation of caudal somites, the first several somites remain intact in these mutants.

In fss mutants, neither mesp-a nor mesp-b expression is detected in the PSM throughout the segmentation period (Fig. 4G-I). During gastrulation, by contrast, all fss mutants show normal mesp-a expression, suggesting that mesp-a is upstream of fss or in a different pathway, at least during gastrulation. However, the expression in the paraxial mesoderm disappears in the mutant at 95% epiboly (arrow in Fig. 4F) when in wild-type embryos the rostral expression domain become restricted to a pair of narrow stripes (Fig. 4A-F). The marginal expression of mesp-a still persists at this stage (arrowhead in Fig. 4F) In the other four mutants, bea, des, aei and mib, the transcripts of both mesp-a and mesp-b are detected but their intensity is highly variable. In all these mutants, the mesp-b expression pattern is no longer striped but shows a mosaic ‘salt and pepper’ pattern, which covers a region two- to three-somites wide in the paraxial mesoderm (Fig. 4M-O). As compared with mesp-b, mesp-a expression in these mutants tend to be very weak and diffuse, and sometimes undetectable except for that in the adaxial cells (Fig. 4J-L).

To study the roles of zebrafish Mesp-a and Mesp-b in somitogenesis, we overexpressed these proteins by injecting capped RNA into one-cell stage zebrafish embryos. Overexpression of Mesp-a caused a severe gastrulation defect, probably due to the inhibition of mesoderm formation (data not shown). In contrast, overexpression of Mesp-b showed specific effects on somitogenesis. Thus, we concentrated on the functional analysis of Mesp-b. RNA encoding β-galactosidase was coinjected with the RNA to be overexpressed to localize the injected RNAs. Control injections using green fluorescent protein (GFP) RNA did not cause any disturbance of somitogenesis or gene expression (Fig. 5A,B).

With the injection of an appropriate amount of mesp-b RNAs (40-50 pg), a disturbance of segmentation was frequently seen (Fig. 5C,D). The severity of this phenotype varied, ranging from the most severe in which many somites failed to form (125/160), to a less severe one where a region of the somites were irregularly formed (11/160). However, we found a close correlation between defects and the presence of the injected RNAs (22/22; Fig. 5C,D,H). Histological sections revealed that, in the affected region, segmentation is severely affected, but no sign of tissue damage such as cell death was observed (Fig. 5G,H).

We then determined whether overexpression of Mesp-b affected the expression of several genes implicated in somitogenesis. Expression patterns of her1, myoD, papc, FGFR1, notch5 and notch6 were examined in injected embryos in which defects in boundary formation were clearly visible at the 8- to 12-somite stage. The severity and number of embryos affected varied between different experiments, depending on the distribution of the injected RNAs. Representative embryos are shown in Fig. 5E,F,I-R.

In control embryos of the 8- to 12-somite stage, her1 is expressed in two or three stripes in the PSM and papc in three to four stripes near and over the point of somite formation. In injected embryos, the length of the PSM (the distance between the tailbud to forming somites) tended to be shorter, but the segmental expression of her1 persisted in the injected region that was positive for β-galactosidase activity (6/8; Fig. 5I,J). In contrast, ectopic expression of Mesp-b caused a uniform expression of papc in the PSM (10/12; Fig. 5K,L). In some cases, the anterior boundary of the papc expression was shifted anteriorly as compared to the uninjected region (arrowhead in Fig. 5L). Furthermore, papc expression was increased in the head mesenchyme (arrow in Fig. 5L). myoD expression was almost lost in regions where Mesp-b was overexpressed (10/10; Fig. 5E,F).

Since papc is expressed in the anterior parts and myoD in the posterior parts of forming and segmented somites, the above results suggest that development of posterior fate within the somites is suppressed in the Mesp-b-injected region. To further verify this idea, we examined the expression pattern of FGFR1. In control embryos, FGFR1 expression is detected uniformly throughout the PSM but after somite formation the expression in a posterior domain of each somite is down-regulated, resulting in a segmented pattern in the somitic region (Fig. 5M; a manuscript on FGFR1 expression pattern is in preparation by K. Y.). This expression pattern is similar to that described for the mouse orthologue (Orr-Urtreger et al., 1991; Yamaguchi et al., 1992). As shown in Fig. 5N (arrow), injected embryos had a disrupted pattern of FGFR1 expression; on the injected side, FGFR1 was not down-regulated in the somitic region, instead it was uniformly expressed throughout the paraxial mesoderm (8/8).

To further examine the effect on the anteroposterior specification within the somites, we looked at the expression of notch5 and notch6 which are segmentally expressed in the anterior PSM and/or formed somites (the probes used were derived from PCR-amplified fragments based on the sequences of notch5 and notch6; Westin and Lardelli, 1997). Transcripts of notch5 are detected in the posterior parts of presumptive and formed somites (Fig. 5O), while notch6 is expressed in the PSM and the anterior parts of somites (Fig. 5Q), in a complementary pattern to myoD expression (Westin and Lardelli, 1997). Although their expression is weak, consistent results were obtained in injected embryos: ectopic expression of Mesp-b significantly down-regulated notch5 (6/6; Fig. 5P), whereas it up-regulated notch6 in the gaps which resulted in uniform expression in the paraxial mesoderm (8/8; Fig. 5R).

All these results support the idea that overexpression of Mesp-b enhances the anterior fate of somites at the expense of posterior one.

mesp-a expression is first detected in the blastoderm margin but stripes of expression become apparent in the paraxial mesoderm during the late gastrulation period (Fig. 2C-D). This early expression pattern is reminiscent of that of her1, a zebrafish bHLH gene related to the Drosophila pair-rule gene hairy (Müller et al., 1996). her1 is expressed in the blastoderm margin and in two distinct transverse bands of hypoblastic cells above the margin. After the completion of epiboly, this pattern remains essentially unchanged throughout the segmentation period. Lineage tracing experiments have demonstrated that the cells in the her1 expression domain I (the first stripe) are incorporated into the 5th somite (Müller et al., 1996).

To determine the relationship between mesp-a and her1, we analyzed their relative expression patterns in embryos of mid-gastrula to early segmentation stage by two-colour in situ hybridization. In these experiments, the embryos hybridized with her1 and mesp-a probes were first stained for her1 in red, then photographed. This was followed by visualization of mesp-a transcripts in blue (Fig. 6A-D,F,G). As shown in Fig. 6A and B, the expression domains of mesp-a and her1 in the marginal region initially overlap at the shield stage. At mid- to late gastrula stages, the anterior expression bands for both genes are separated from the marginal expression domain, and they are segregated such that mesp-a expression is located anteriorly to her1 (domain I) (Fig. 6C,D). Double fluorescent in situ hybridization with mesp-a (red) and her1 (green) show that the expression domains of each gene are completely segregated, although they are juxtaposed to each other (Fig. 6E). These results, together with the lineage tracing experiments of her1-expressing cells (Müller et al., 1996), suggest that the precursors for the first three to four somites do not express mesp-a.

During the segmentation period, the anteriormost stripes of mesp-a and her1 are seen juxtaposed or partially overlapped. There is a tendency for the overlapping region to be observed when two pairs of mesp-a stripes are detected, but not when the only one pair of mesp-a stripes is detected (Fig. 6F-I). These two staining patterns were reproducibly obtained at all segmentation stages examined. However, this result appeared contradictory to that reported by Müller et al (1996) because, in the trunk region, her1 expression has been proposed to follow a pair-rule pattern in alternating somite primordia, whereas mesp expression is detected in all somite primodium. Thus, we re-examined her1 expression using mesp-a expression as a reference. We collected embryos between 5- to 7-somite stages and re-staged them according to the following morphological criteria before fixing. Each stage was further divided into three sub-stages: the n-somite stage when boundary formation is just completed, the n-somite+ stage when the presomitic cells that will contribute to the future next somite become aggregated and the next furrow formation has started, and the n-somite++ stage when furrow formation has proceeded to half way. These morphological criteria were determined by a lateral view of a live embryo. It should be noted that we are not able to judge accurately the lengths of individual stages. Arranging a series of pictures in order, we noticed that her1 expression domains move in a caudal to rostral direction although the posterior tip of the tailbud is always express her1 transcripts (Fig. 7A-G,A′-G′). We examined seven independent series of embryo collections and all results were consistent. The her1 stripes, which appear in the tailbud region, become narrower as they migrate, and finally overlap the most rostral decaying mesp stripes before both transcripts disappear (Fig. 7). A wave leaves the tailbud region every 30 minutes and takes 1.5 hours to complete one cycle and to disappear. Thus, all somite primordia (except for the first few) become positive for her1 and mesp-a prior to segmentation.

In this study, we have isolated two zebrafish cDNAs that encode bHLH transcription factors related to murine MesP1 and MesP2. Recently, Mesp-related genes have been isolated in Xenopus (Thylacine, Sparrow et al., 1998; Mespo; Joseph and Cassetta, 1999) and chick (cMeso-1, Buchberger et al., 1998). Their bHLH domains are highly homologous to each other, forming a novel family of bHLH transcription factors that we refer to as the Mesp family. Outside the bHLH region, however, essentially no homology can be found among Mesp family proteins. Based on the phylogenetic tree (Fig. 1C), it is likely that the independent duplication of the mesp gene occurred in each vertebrate during evolution. In spite of this, the expression pattern and genomic organization of murine Mesp1 and Mesp2 are very similar to those of zebrafish mesp-a and mesp-b. mesp-a and mesp-b have a nearly identical expression pattern to murine Mesp1 and Mesp2, respectively. Only Mesp1 and mesp-a are expressed in the early mesoderm (the primitive streak and the blastoderm margin), while the expression in the PSM is almost the same for all four genes, albeit the number of expression stripes varies. Although the precise genomic organization of mesp-a and mesp-b is not clear yet, our mapping data (see below) and the fact that mesp-b was isolated from a PAC clone containing mesp-a suggest that the two genes are located within a region of less than a few hundred kb. Similarly, Mesp1 and Mesp2 are located in chromosome 7, head to head, and separated by 23 kb (Saga et al., 1997).

All identified Mesp family members are expressed in a subdomain of the PSM, immediately posterior to the forming somites (Saga et al., 1996 and 1997; Sparrow et al., 1998; Buchberger et al., 1998; Joseph and Cassetta, 1999). In addition to the Mesp family, other bHLH transcription factors, such as zebrafish her1, chick c-hairy1, and Xenopus Hairy2A (Jen et al., 1997), are segmentally and/or transiently expressed in the PSM, suggesting that the bHLH family play an important role in establishing a segmental prepattern prior to somitogenesis. Interestingly, some of them show a preferential expression within prospective anterior or posterior parts of somites, suggesting they have roles in anteroposterior specification of the somite primordia. Thylacine expression is restricted to the anterior halves of the somitomeres while Hairy2A is in the posterior halves. Similarly, this study revealed that at least the most rostral expression domain of zebrafish mesp-a and mesp-b resides in the anterior part of the future somite. Although the expression stripes of murine Mesp genes and chick cMeso-1 are approximately one somite wide, the analysis of lacZ expression driven by Mesp2 promoter suggests that the cells that have expressed Mesp2 mainly contribute to the anterior parts of the mature somites (data not shown). These facts strongly support the existence of a segmental prepattern and an anteroposterior polarity in the PSM, in agreement with the results of experimental manipulations performed in several organisms (Keynes and Stern, 1988; Elsdale et al, 1976; Cooke, 1978; Kimmel et al., 1988; Primmett et al., 1988; Roy et al., 1999).

The mesp-a expression domain at the blastoderm margin contains all the cells that will involute, or ingress, to form mesoderm derivatives (Kimmel et al., 1995). During involution and dorsal convergent movement, mesp-a-positive cells are mainly located in the future paraxial mesoderm. However, the expression soon disappears in the paraxial mesoderm except for the most rostral region, leaving a pair of stripes in the PSM. A similar dynamic expression pattern has been reported for zebrafish her1. Two-colour in situ hybridization analysis demonstrates that the expression domains of mesp-a and her1 initially overlap but that they are soon segregated during late gastrulation when their striped expression patterns become apparent. mesp-a stripes are always located just anterior to those of her1 (Fig. 6E). This relationship is maintained until the onset of segmentation.

Since lineage-tracing analysis demonstrated that the first pair of her1 stripes (domain I) resides in a primordium of the future 5th somite (Müller et al., 1996), neither mesp-a nor her1 is expressed in the cells that contribute to the first several somites (maybe four). These results suggest the existence of distinct mechanisms for rostral and caudal segmentation. In addition, as reviewed by Jiang et al. (1998), several other observations further strengthen this idea. First, the earliest few somites in zebrafish seem to form more quickly than later ones: 3 per hour for the first six, and 2 per hour thereafter (Kimmel et al., 1995). Secondly, in fss-type mutants except for fss, segmentation defects are visible only in the region posterior to the 4th to 8th somite level (van Eeden et al., 1996). Thirdly, several early somites in the cervical region are generated in Wnt3a- and Mesp2-deficient mice in which segmentation in the trunk and tail regions is severely impaired (Takada et al., 1994 and Saga et al., 1997).

Holland et al. (1997) found that a homologue of the engrailed gene is expressed only in the first eight somites of amphioxus. In amphioxus embryos the first eight somites are formed by a primitive mechanism by which epithelial outpockets are pinched off the primitive gut, whereas more caudal somites are formed by a common vertebrate mechanisms by which mesenchymal cell sheets are subdivided into segments. In spite of differences in the mode of segment formation in amphioxus and in higher vertebrates, several schemes for aligning the segments in amphioxus and vertebrates have been proposed. For example, the sixth somite of amphioxus is equated to the sixth somite in the lamprey and shark and to the first true somites in the chick (Gilland and Baker, 1993). All these observations suggest the presence of a distinct mechanism for the rostral segmentation. Therefore, it will be of interest to clone Mesp family members from amphioxus and to determine their anterior expression boundary.

Using the mesp-a expression pattern as a reference, we found that zebrafish her1 transcripts oscillate in the PSM with a regular periodicity before decaying near the point of segmentation. In the trunk region, her1 expression overlaps with the anterior mesp-a expression domain and, consequently, every somite primordium is positive for her1 transcripts prior to segmentation. This dynamic expression pattern is highly reminiscent of that reported for c-hairy1 which provided the first molecular evidence for an intrinsic clock linked to somite segmentation (Palmeirim et al., 1997). The expression of c-hairy1 appears as a wave, which sweeps across the PSM once during each somite formation, and stabilizes in the posterior half of the next budding somite. Although zebrafish her1 is not maintained in the somite, it is likely that the cycling behavior of hairy-related genes is conserved among vertebrates. The wave of c-hairy does not result from cell displacement or from signal propagation in the PSM but rather reflects an intrinsically coordinated clock which do not require protein synthesis (Palmeirim et al., 1997). The nature of the her1 wave in zebrafish should also be examined in this context.

During the segmentation period, expression of mesp-a and mesp-b in the PSM is completely abolished by the fss mutation, indicating that they are downstream targets of fss. However, the phenotype of fss mutant is not likely to be due to a mutation in a mesp gene because mapping of mesp genes with a radiation hybrid panel (kindly provided by Dr M. Ekker) showed that zebrafish mesp genes are located in LG 7 to which no fss-type genes have been mapped so far (Dr Scott Holley, personal communication).

In fss mutants, mesp-a expression remains normal up to late gastrulation but its rostral expression becomes weak between 90% and 95% epiboly and finally disappears by 100% epiboly. This indicates that fss is required for mesp expression only in the rostral PSM, suggesting that fss functions near the point of segmentation (Fig. 8A). Consistent with this, van Eeden et al. (1998) have reported that the fss mutation only affects the most rostral stripes of her1: the establishment of a segmental expression pattern is normal in fss mutants, but her1 expression is switched off prematurely in the rostral PSM, resulting in a reduction of number of visible stripes.

In contrast to fss, other fss-type mutants, bea, des, aei and mib, show a weaker effect, and differentially affect mesp-a or mesp-b expression. The expression of mesp-a in the PSM is very weak in these mutants, while the level of mesp-b expression is more or less unaffected but its striped expression pattern is lost, probably due to an effect secondary to the irregular segmentation. Thus, in spite of similar expression patterns during the segmentation period, the expression of mesp-a and mesp-b seems to be differentially regulated. Furthermore, no mutual dependency in their expression has been observed so far. Misexpression of Mesp-b did not affect mesp-a expression (data not shown) and, in Mesp1- or Mesp2-knockout mice, the expression of the intact Mesp gene is unaffected (Saga et al., 1997, Saga, 1998).

The results obtained by misexpression experiments demonstrate that Mesp-b expressed in the anterior parts of the presumptive and forming somites confers anterior properties on the somite cells. Misexpression of Mesp-b in the normal embryos led to a loss or incorrect formation of somite boundaries. In the injected region, papc and FGFR1 expression, which demarcate the anterior domain of presumptive and/or segmented somites, lost their metameric pattern. In contrast, myoD expression in the posterior domain of the somites was greatly reduced. Accordingly, notch5 expression which would normally be detected in the posterior parts was reduced, while notch6 expression domain in the anterior parts of the presumptive and segmented somites was expanded, eliminating the gaps in the striped expression pattern. All these data demonstrate that misexpression of Mesp-b expands the anterior character at the expense of the posterior one within both presumptive and segmented somites. Recently a number of studies have shown that FGF-mediated signalling (Deng et al., 1994; Yamaguchi et al., 1994) and Notch-Delta signalling (Conlon et al., 1995; Jen et al., 1997, 1999) are involved in boundary formation. In this study, the expression of the genes involved in these signalling pathways is altered by misexpression of Mesp-b. Together with the fact that the expression of Notch1, Notch2 and FGFR1 are greatly downregulated in Mesp2-deficient mice (Saga et al., 1997), it is likely that the functions of MesP2 and Mesp-b are mediated via Notch-Delta and FGFR signalling systems.

The defects in boundary formation caused by misexpression of Mesp-b are reminiscent of those reported in Mesp2-deficient mice (Saga et al., 1997). Conversely, in this case, the anterior somite fate is suppressed, and the entire somite is posteriorized in identity. Thus, the conclusion drawn from the present gain- of-function experiment is consistent with that from the loss-of-function in Mesp2-deficient mice. Severe attenuation of somitogenesis has also been reported in Xenopus embryos injected with Thylacine RNA and in chick embryos treated with antisense RNA or oligonucleotides of cMeso-1 (Sparrow et al., 1998; Buchberger et al., 1998), suggesting that the Mesp family genes have a conserved function in vertebrate segmentation.

Taken together, the Mesp family of bHLH transcription factors may act at the same point in vertebrate segmentation, the establishment of anteroposterior polarity within the somite primordia, probably through interacting with the FGFR and the Notch-Delta signalling pathways (Fig. 8B).

We thank Drs Adam Rodaway (King’s College London) and Trevor Jowett (University of Newcastle) for their critical comments on β-galactosidase staining and two-colour in situ hybridization, and Mr Hiroshi Onodera for help in isolating zebrafish FGFR1. We also thank Drs Charles B. Kimmel (University of Oregon), Sharon L. Amacher (University of California, Berkeley) and Julian Lewis (Imperial Cancer Research Fund) for critical reading of the manuscript. This work was supported in part by CREST (Core Research for Evolutionary Science and Technology) of Japan Science and Technology Corporation (JST), by Bio-Design Project (BDV-00-IV-1-1) and Pioneering Research Project in Biotechnology from the Ministry of Agriculture, Forestry and Fisheries of Japan, and by grants-in-aids from the Ministry of Education, Science, and Culture of Japan. Y.-J. J. would like to thank Dr Julian Lewis for support and was funded by an EMBO long-term fellowship (ALTF 225-1996).