Met and Hgf signaling controls hypaxial muscle and lateral line development in the zebrafish

In amniote embryos, the developmental mechanisms employed to generate the appendicular muscles of the limbs differ markedly to those used to make the muscles of the axis. Limb muscles exhibit discontinuous development, being specified initially within the hypaxial or lateral dermomyotome, adjacent to the forming limb bud, through the action of yet to be identified inductive signals. Specified myoblasts then undergo an epithelial-to-mesenchymal transition (EMT) before delaminating from the somites and migrating into the limb environment where differentiation is initiated. Chimaeras of chick and quail embryos have revealed that avian limb musculature originates from the lateral half of limb level somites (Christ et al., 1977; Chevallier et al., 1977; Ordahl and Le Douarin, 1992), specifically within the lateral lip of the dermomyotome (Christ and Ordahl,1995). These cells are marked by the expression of a number of different genes whose activities are required for aspects of their consequent morphogenesis (Mennerich et al.,1998; Schäfer and Braun,1999; Dietrich et al.,1998; Dietrich et al.,1999; Mankoo et al.,1999).

Until recently it was believed that the specification of limb muscle precursors was entirely dependent on limb-derived environmental cues. Flank somites grafted into the limb region result in the grafted tissue contributing to the forming limb musculature, despite their non-limb level origin(Chevallier et al., 1977; Christ et al., 1977; Hayashi and Ozawa, 1995). Furthermore, when ectopic limbs are induced adjacent to flank somites, these somites are re-specified to generate migratory muscle precursors that develop into normal limb muscles. However, Alvares et al.(Alvares et al., 2003) have recently shown that limb muscle precursors are not naïve, and the ability of somites to produce limb muscle precursors depends on the axial level from which individual somites derive. Somites transplanted from limb level to the flank retain the ability to initiate limb myoblast specific gene expression. Furthermore, altering the positional identity of somites, through the manipulation of Hox gene expression, again changes the ability of cells to express limb myoblast-specific marker genes. However, despite an underlying reliance on positional information, the positional identity of somites can be completely over-ridden by signals emanating from the limb environment, as flank somites can contribute to limb muscle formation when grafted adjacent to the limb environment. Thus, the specification of limb muscle precursors is influenced by two different layers of regulatory control: one reliant on the position of an individual somite; the second dependent on signals produced by the limb environment, which are sufficient, in themselves, for limb myoblast induction.

Several genes have been identified that are required for the formation of the limb musculature. Perhaps the best understood of the cohort of limb muscle control genes, from a mechanistic point of view, is the receptor tyrosine kinase Met and its ligand hepatocyte growth factor (Hgf) (reviewed by Birchmeier et al., 2003). The EMT of limb myoblasts is mediated by Met and Hgf, a process shown to be necessary for limb myoblast migration(Yang et al., 1996; Heymann et al., 1996; Brand-Saberi et al., 1996). During limb formation, met is expressed in the medial and lateral lips of the dermomyotome at all axial levels(Yang et al., 1996). Localized activation of Met occurs through the restricted expression of hgf,which is present within the mesenchyme of the limb bud, during limb muscle migration. Targeted inactivation of either met or hgf within the mouse embryo results in a lack of appendicular muscle of the limbs, as well as other hypaxially derived migratory muscles such as the diaphragm and tongue, despite these cells being initially specified normally(Bladt et al., 1995; Schmidt et al., 1995). Limb muscle precursors, however, are unable to migrate from the somite into the limb bud in met and hgf mutants, as indicated by the continued expression of lbx1 in the dorsolateral lip of somites at limb levels (Dietrich et al.,1999). In addition, implanting Hgf-soaked beads at inter-limb level, where Hgf is not normally expressed, can result in cells undergoing an EMT before delaminating from the somites in this region(Heymann et al., 1996; Brand-Saberi et al., 1996). Thus, the formation of appendicular muscles of the amniote limb is characterized by a Met/Hgf dependent EMT and consequent long-range migration of limb muscle precursors to their site of differentiation within the dorsal and ventral muscle masses of the limbs.

Analyses in fish species, however, have revealed the existence of two phylogenetically distinct processes operating to generate the appendicular muscles of the fin. The first mechanism, which is thought to be analogous to that present in amniote embryos, has been shown to operate in zebrafish(Neyt et al., 2000). However,a second primitive mode of appendicular muscle formation was found to occur within chondricthyan species such as the spotted dogfish shark(Scyliorhinus canicula). An examination of fin muscle formation at a number of different developmental stages revealed the existence of direct myotomal extensions headed by epithelial buds within the developing fin bud,which lay down differentiating myocytes in a proximal-to-distal manner as the epithelial bud extends within the fin(Dohrn, 1884; Braus, 1899; Goodrich, 1958; Neyt et al., 2000) (reviewed by Galis, 2001). Furthermore, S. canicula somite extensions do not express genes that mark migrating limb/fin myoblasts in other species, reinforcing the primitive nature of the control Selachian fin muscle formation.

These studies raise a number of important issues about the molecular nature of fin muscle formation in zebrafish. Are fin muscle precursor cells influenced by the local fin adjacent environment, or do they possess an intrinsic ability to form fin musculature based on anterior posterior positioning within the embryo? Is the Met-dependent EMT evident within amniote species a true component of the derived mode of fin and limb muscle formation,or is it a synapomorphy that was generated in response to the differing architecture of the tetrapod body plan? To answer these questions, we have investigated the competence of zebrafish somites positioned at different anteroposterior levels to contribute to pectoral fin muscle formation and examined the role that Met-mediated signaling plays in the formation of hypaxial and specifically appendicular muscle formation in zebrafish.

Donor somites were dissected from 13-17 somite stage α-actin GFP transgenic embryos (Higashijima et al., 1996). We chose the 13-17 somite stage of development to perform transplantation because it is the earliest point at which successful transplantation could be achieved. It also encompasses a developmental period prior to both formation of the fin primordia and to the onset of fin muscle precursor specific gene expression within fin adjacent somites. Consequently, any restriction in the ability of somites to contribute to the fin musculature would reflect an intrinsic fate restriction of somitic cells rather than previous exposure to inductive cues from the fin, which may influence the fate of naïve somites. An initial dissection was made in which the trunk was separated from the head and yolk using two 30-gauge needles. The trunk section of the dissected embryo was then incubated in 25 mg/ml pancreatin (Sigma) until the somites visibly started to dissociate. Pancreatin was then inactivated by serially moving the somites into small droplets of L15+10% FCS media. Somitic tissue was selected under a fluorescent dissection microscope (Leica MZ FLIII), and individual somites dissociated using a fire-polished small-bore Pasteur pipette and stored on ice in L15 media during host preparation. Dechorionated wild-type host embryos (13-17 somites), derived from spawnings of AB strain adults were placed on a glass coverslip and the majority of surrounding medium removed. Two drops of 0.7%agarose containing 0.17 mg/ml tricaine were placed directly on the embryo and the embryo oriented such that the lateral surfaces of anterior somites are directly positioned against the surface of the coverslip. Once the agarose has set, the agarose block containing the embryo was trimmed, inverted and remounted again in a drop of agarose/triacane placed on a deep well depression slide. Prior to transplantation, the coverslip was removed and the host embryo, together with the depression slide in which it is mounted, is immersed into a petri dish containing L15 media. Under a dissecting microscope, the somitic region into which the donor somitic tissue is to be transplanted is extirpated using a microforged, hooked glass capillary and the region for transplant enlarged using a second ball-headed, microforged glass capillary needle. The dissected donor somites were directly pipetted, again using a fire-polished glass pipette, adjacent to the extirpated region of the host and positioned into the transplant region using a ball-headed glass capillary needle. Transplanted embryos were left undisturbed, immersed within the L15 media for at least 30 minutes before removing the embryo from the agarose by manual dissection. Individual transplanted embryos are transferred to 12-well tissue culture plates containing systems water and antibiotics and incubated to the desired age at 28.5°C.

Embryos transgenic for the α actin GFP transgene were anaesthetized(0.17 mg/ml tricane, Sigma) at 36 hpf, embedded in 0.7% agarose/0.17 mg/ml tricaine, and orientated obliquely on a large coverslip such that the anterior somites and yolk were positioned directly against the glass. The embedded embryo was inverted into a deep well depression glass slide into which embryo media (Westerfield, 1994) plus 0.17 mg/ml tricaine had been placed. The edge of the depression was coated with small amounts of petroleum jelly so that when the coverslip was inverted onto the depression slide, an airtight seal was produced, and the sample did not dry out during time-lapse. Cell movements were visualized on an Axioskop FS (Zeiss, Germany) by taking near simultaneous DIC and fluorescent images every 5 minutes using a Hamamatsu ORCA digital camera driven by IPLab software(Scanalytic, Fairfax, VA, USA).

A full description of the cloning of met and hgfsequences can be found in the legend for Fig. S1 in the supplementary material.

Two antisense morpholino oligos were designed against the metsequence (Genetools, OR). The initial morpholino (CM1 ATAGTGAATTGTCATCTTTGTTCCT) was designed to anneal against ATG-containing sequences, and the second was targeted against the 5′ untranslated region of met (CM2 CTGTAAAATAAAGACACCTGTCGGA). A control morpholino was also injected that contained a 4 bp mismatch to CM1 (CM1mm,ATAATGGATTGTCATCCTTGCTCCT). Morpholinos were injected into α actin GFP transgenic embryos at the two-cell stage, at concentrations of 0.5 and 0.75 mM. In situ hybridization was carried out as described previously(Jowett and Lettice, 1994)with some minor modifications. Cryosectioning of somite transplants were performed after overnight fixation in 4% PFA, using standard procedures(Westerfield, 1994). Wax sections of embryos processed for in situ hybridization or immunohistochemistry were performed using standard techniques(Nusslein-Volhard and Dahm,2002).

Anti-human HGF antibody (R&D Systems, UK) and a control antibody,anti-human slow MyHC (Chemicon, USA), were both diluted to 500 μg/ml in PBS and injected into α-actin GFP embryos at either 16-20 somites or 28 hpf stages of development. Embryos were incubated at 28.5°C in system water with Methylene Blue, until they reached 48 hpf and analyzed for defects in hypaxial muscle formation. Recombinant Hgf protein (R&D Systems, UK) was diluted in PBS to 50 μg/ml and incubated with shards of Affi Blue beads(BioRad, USA) that had been trimmed with 30 gauge needles to an appropriate diameter, for 1-2 hours at room temperature in a moist chamber. Beads were then back loaded into an injection pipette, under control of a hydraulic microdrive (Narashige, Japan), and injected sub-epidermally at the appropriate positions.

In order to determine if a rostrocaudal restriction in the ability of somites to contribute to fin muscle formation existed, techniques were developed for homo and heterotopic somite transplantation in zebrafish. Somites were obtained from 13- to 17-somite stage zebrafish donor embryos,genetically marked with the expression of a GFP transgene driven by the skeletal muscle-specific α actin promoter(Higashijima et al., 1997) and transplanted into similar stage wild-type host embryos at the approximate level of somite 4. Fate mapping studies had previously shown that somites 2-4 contribute myoblasts to the fin, with somite 4 being the major source(Neyt et al., 2000).

Three different types of transplantation experiment were performed. Initially, we randomly chose somites from any rostral/caudal level and transplanted into the region of somite 4 within the host. Contribution to the fin musculature by donor tissue could be detected by the presence of GFP-positive cells, derived from the transgenically marked donor tissue transplanted into somite 4, within the developing fin. Surprisingly, our analyses revealed that only 5% of successful transplants (n=22) were able to contribute to the fin musculature. This suggested that the somites already possessed, at the 13- to 17-somite stage, some restriction in competence for provision of fin muscles.

To determine how somites are restricted in their ability to contribute to fin musculature, they were isolated from α actin GFP donors and dissected into two groups. The first series of transplants used tissue derived from the six rostral-most somites and the second group of dissections used somites caudal to somite 9 at the 13-17 somite stage. This combination of dissections was aimed at minimizing errors in dissection level leading to an overlap in the rostrocaudal level from which donor somites were derived. Individual somites from each group were transplanted ventrally into the host somite at the approximate level of somite 4, and the contribution of donor tissue to the fin musculature monitored by the presence or absence of GFP-positive myoblasts within the host fin. Rostral transplants gave rise to fin myoblasts 45% of the time (Fig. 1, n=11), while transplants caudal to somite 9 gave rise to fin myoblasts in only one instance (n=18), which we attribute to an experimental error in correctly assigning dissected somites to either the rostral or caudal donor class. We could also show that the failure to give rise to fin myoblasts did not result from a difference in developmental age between the transplanted tissue of posterior and anterior somites at 13-17 somites. Transplants of donor somites performed at 24 hpf, a time at which somites posterior to somite 9 would have been of roughly similar developmental age as the anterior somites within 13- to 17-somite stage embryo, also failed to give rise to fin myoblasts (n=16). Collectively, we have interpreted these results to mean that somites transplanted in this manner possess an anteroposterior fate restriction at a very early stage of their development that cannot be over-ridden by local inductive signals.

In the course of this and previous analyses, we noticed that a second hypaxial muscle was also derived from anterior somites(Fig. 1C-H). This muscle, which we have termed the posterior hypaxial muscle (PHM), was observed in initial fate-mapping experiments to be derived from somites 5 and 6(Neyt et al., 2000) (data not shown). We have extended our initial observations, revealing that this muscle was only ever labeled in transplants of anterior somites that included either somite 5 or 6 (n=5, Fig. 1C-H), which occurred as a bi-product of our attempts to target transplantation to somite 4. Furthermore, the morphogenesis of the PHM appeared to have some unique features and in order to better understand the processes that control the formation of this muscle we undertook a series of time-lapse analyses using the α actin transgene strain to trace the migration of the PHM (Figs 2, 3). The expression of the GFP transgene during the migration of the PHM, as well as markers of muscle differentiation such as MyHC and MyoD(Neyt et al., 2000) (data not shown), is in contrast to fin muscle formation in zebrafish and hypaxial muscle formation in amniotes, which do not express markers of muscle differentiation until migration of these cells is complete. However,expression of the GFP transgene during PHM migration allows us to time lapse the movement of these cells as they emerge from the somites. Cells appeared to exit somites 5 and 6 in a stereotypical pattern, maintaining a strict order in the anteroposterior relationship as they exit the somite, such that the cells that first exit the somite always remain at the anterior or distal tip of the migrating primordium and cells are progressively added from behind, as additional cells exit the somite. Migration of the PHM primordia always follows a strict path (Figs 2, 3; see Movie 1 in the supplementary material). Cells exit the somite at an oblique angle, migrate ventrolaterally around the fin primordia and consequently turn anteriorly,where they eventually attach to the cleithrum, the first forming bone in the zebrafish embryo.

It is also clear that the cues that guide this muscle to its insertion point on the cleithrum are distinct from those that are responsible for the migration of the muscles of the fin. Indeed, it can be seen that cells that occasionally leave this stereotypical PHM migration pathway (and this only appears to happen anterior to the migrating stream, see Movie 1 in the supplementary material), adjacent to the fin primordium and become trapped in the space between the fin and the migrating PHM. This suggests that the pectoral fin primordium may be acting as a negative guidance cue for PHM migration. However, the potential role of other unknown guidance cues is suggested by the behaviors that individual muscle cells undergo during migration. Filopodial protrusion and retractions are clearly evident in cells at the leading edge of the migrating PHM primordia(Fig. 3; see Movies 1 and 2 in the supplementary material), behaviors associated with the sensing of guidance cues on route to articulation with the cleithrum. Filopodial extensions have been previously associated with migrating fin myoblasts in chick explant cultures, suggesting that a conserved mechanism may underlie hypaxial myoblast migration in vertebrate embryos (Knight et al., 2000).

The results of our transplantation studies revealed a surprisingly early competence restriction to the anterior somites for the ability to produce hypaxial musculature. We wished to determine if molecular correlates of this competence were similarly restricted. In order to address this issue, we identified zebrafish orthologs of met and hgf, and examined their expression during fin myoblast migration(Fig. 4). A single open reading frame was identified within the zebrafish genome that encoded the Met receptor(Fig. 4A, see Fig. S1 in the supplementary material). We find that, in contrast to amniote embryos, met somitic expression is specifically restricted to the ventral lateral margin of anterior somites, which we have shown generates hypaxial muscle in zebrafish and cannot be detected within other somites positioned at more posterior levels (Fig. 4C-E). During fin myoblast and PHM migration, expression can be detected in migrating myoblast cells, but in the case of the PHM, expression is restricted only to cells exiting the somite(Fig. 4H-K). By contrast, at 48 hpf, a stage at which the majority of fin myoblast migration is thought to have been completed (Neyt et al.,2000), met expression remains high in the post migratory dorsal and ventral muscle masses of the fin with expression within the PHM no longer detectable at this stage, despite its continued migration towards its attachment point at the cleithrum (Fig. 4L, Fig. 2).

A second migratory cell type in which met is highly expressed is the cells of the posterior lateral line primordia (PLLP). The PLLP gives rise to the lateral line, which is a mechanosensory system deployed in fish and amphibia to perceive and localize movement(Schulze, 1870; Stone, 1933). Detection of water displacement is made through neuromasts, which caudally are derived from the PLLP that originates adjacent to the otic vesicle and undergoes long-range posterior migration along the length of the body(Metcalfe et al., 1985; Metcalfe, 1989). During migration, pro-neuromast cells are deposited approximately every 3.5 hours from the primordium and differentiate to produce seven or eight neuromast clusters spaced evenly along both sides of the body(Metcalfe et al., 1985; Gompel et al., 2001). Expression of met is first detected in the PLLP at ∼20 hpf, at the onset of migration (data not shown). By 24 hpf, met is highly expressed within the PLLP, which has begun its migration caudal to the otic vesicle (Fig. 4C-E). Expression of met in the PLLP is evident throughout the entire period of its migration (Fig. 4F,G) towards the tail tip but is not expressed at any stage within deposited proneuromast or neuromast clusters.

Given that it is the localized expression of the Met ligand, Hgf, which restricts the activation of the Met receptor in amniote embryos, it was of interest to determine how hgf was expressed during fin myoblast migration. We also isolated fragments of two genes encoding Hgf, which we have termed hgf1 and hgf2. The two different genes gave identical expression profiles during development and hence will collectively be called hgf for the purposes of this study. hgf initiates expression broadly and at low level within somites at all axial levels at 22 hpf, with stronger expression evident at somite boundaries(Fig. 4M,N). Expression is also evident within the caudal aspect of the notochord at this stage, with widespread expression also evident in the neural tube by 24 hpf(Fig. 4N). By 30 hpf, hgf transcripts can be detected throughout the fin bud mesenchyme with expression in the fin increasing at 36 and 48 hpf(Fig. 4O-Q). At 48 hpf, hgf transcripts can no longer be detected within the trunk of the embryo.

To investigate the function of Met in the tissues in which it is expressed,we `knocked-down' Met activity, using two different antisense morpholino oligonucleotides. One morpholino was targeted to overlie the sequences encoding the ATG (CM1) and the other to sequences 5′ of the ATG, within the 5′ untranslated region of the met gene (CM2). Microinjection of either of these morpholinos (0.5 mM) into embryos derived from the α actin GFP transgenic strain resulted in a similar, severe,reduction in the amount of GFP-positive cells in the dorsal and ventral muscle masses of the pectoral fin as well as a reduction in cells of the PHM (89%, n=269, Fig. 5A-J). By contrast, no defect in fin muscle and PHM migration was evident in embryos injected with a control morpholino oligonucleotide (CM1mm, n=118, 0.5 mM), containing a 4 bp mismatch to the met ATG spanning morpholino oligonucleotide CM1.

In order to determine whether in met morphant embryos, hypaxial muscle cells were specified normally in anterior somites, we used whole-mount in situ hybridization analysis to study the expression pattern of genes involved in pectoral fin muscle development(Neyt et al., 2000). At 22 hpf, lbx1 is upregulated in the lateral edge of the anterior somites,in wild-type and morphant embryos (Fig. 5K,L) (Neyt et al.,2000), indicating that lbx expression is initiated correctly in embryos with reduced levels of Met. At 36 hpf, a stage when myoblasts are delaminating from the somites and undergoing long-range migration to the fin buds and head, lbx1 expression in morphant embryos is greatly reduced in these migratory cells and absent from the pectoral fins (Fig. 5M,N). Expression of myod, which is expressed in differentiating myoblasts of the pectoral fin bud, is significantly reduced in morphant embryos(Fig. 5O,P), indicating reduced numbers of myoblasts in morphant pectoral fins and the PHM. Thus, we can conclude that Met is required for the exit of myoblasts from the anterior somites but is not necessary for the initial specification of these cells.

As a second prominent region of expression of met was within another migratory cell type, the PLLP, we were interested to determine if there was a similar requirement for Met in directing either the migration of,or the exit of, neuromast clusters from, the PLLP. The vital dye DASPEI(Whitfield et al., 1996) marks the regularly spaced neuromast-derived hair cells present on both sides of the trunk of the embryo (Fig. 6A),and identification of fluorescent hair cell clusters therefore can be used to monitor neuromast deposition. Microinjection of either morpholino CM1 or CM2(0.5 mM) resulted in a similar, severe, reduction in the number of deposited neuromast clusters as monitored by DASPEI staining(Fig. 6A-C). By contrast, no defect in neuromast deposition was evident in embryos injected with a control morpholino oligonucleotide (CM1mm, n=148, 0.5 mM), containing a 4 bp mismatch to the met ATG-spanning morpholino oligonucleotide, CM1. On average, CM1/CM2 injected embryos possessed 2.87±1.31 (n=114)clusters per side, as defined by the presence or absence of fluorescent hair cells at 48 hpf. Uninjected embryos possess 6.0±0.58 (n=12)per side at this stage of development. Furthermore, when neuromast-derived hair cell clusters are deposited in met morpholino-injected embryos,the number of hair cells within individual clusters is drastically reduced(Fig. 6D,E), possessing fewer than 50% (n=11) of the number of differentiated hair cells present in wild-type clusters. Spacing of hair cell clusters, when deposited, was also altered in met morphants, with larger than usual gaps evident between clusters, spanning two to three times the number of somites present between neuromasts in wild-type embryos (Fig. 6C). Furthermore, differentiated clusters appear to be randomly deposited within a somite, instead of possessing the stereotypical location at somite boundaries. The non-neural component of the lateral line, the supporting cells, are believed to be marked by the expression of the follistatin gene (Fig. 6F,G) (Mowbray et al.,2001). An analysis of follistatin expression within met morpholino injected embryos, reveals a similar lack of support cells to that evident for the hair cells(Fig. 6H,I), suggesting all PLLP-derived fates are affected by the injection of metmorpholinos.

The lack of PLLP-derived cells fates could arise as a consequence of failure of a number of developmental mechanisms. Met could be required for the correct migration of the PLLP along the body axis, or alternatively it could act to control the process of neuromast deposition. To distinguish between these alternatives hypotheses, we hybridized met morphant embryos with an antisense probe to prox1, which marks all cells of the PLLP throughout its migration (Glasgow and Tomarev, 1998). An analysis of prox1 expression in met morpholino-injected (n=62) and uninjected embryos(n=30) revealed no difference in the extent or timing of PLLP migration at comparable stages of development(Fig. 6J-L), suggesting that met function is not required for migration of the PLLP. However, an examination of the PLLP at 36 hpf, towards the end of the migratory process,revealed that it was grossly enlarged in met morpholino-injected embryos (Fig. 6L,M). The lack of neuromast formation coupled with an increased size of a normally positioned PLLP strongly suggests that met is required to control the process of pro-neuromast deposition from the migrating PLLP.

In order to determine if ectopic activation of the Met/Hgf signaling pathway could perturb formation of migratory myoblasts, we developed techniques for implanting beads soaked in Hgf protein adjacent to anterior somites. Using such techniques, we hoped to determine if application of ectopic Hgf could produce delamination of ectopic muscle cells from adjacent somites, as had been shown in the chick embryo, and whether there was any restriction in the ability of somites at different anteroposterior levels to do this, as suggested by the restricted nature of met expression. These experiments demonstrated that if a bead was implanted lateral to the fin at 20 hpf, by 36 hpf ectopic muscle fibers could be detected in two stereotypical locations that were never present in control embryos. We observed a spur of muscle from somite 4 and elongating muscle fibers on the yolk adjacent to this somite (Fig. 7A,B,D, n=12). On the control side, where no bead or a control bead soaked in PBS (n=14) was implanted, no ectopic fibers were detected (Fig. 7C, data not shown). In addition, implantation of Hgf-soaked beads at other positions within the embryo did not result in ectopic muscle fiber development from other somites or induction of met expression within adjacent somites(n=5, data not shown). However, implanting a bead in the neural tube of the embryo did result in the ectopic expression of met around the bead (n=2, data not shown). These results reveal that the ability of Hgf to stimulate exit of somitic cells is restricted to anterior, and specifically, fin adjacent somites. It also reveals that the initiation of hypaxial myoblast differentiation is triggered when precociously delaminating cells emerge from the somite environment, rather than by the deployment of some intrinsic cell-autonomous timing of the myoblasts themselves.

As the failure to define the entire coding regions of both hgfgenes (see Fig. S1 in the supplementary material) precluded the use of morpholinos to knock down Hgf function, we injected an antibody against the human HGF protein which was shown to block Hgf function in a number of other contexts (Sheehan et al.,2000; Kolatsi-Joannou et al.,1995). This experimental approach has the additional benefit being able to add antibody at any time during fin myoblast migration and thereby ascertain the requirement for Hgf-mediated signaling during the migration process. We injected the antibody, immediately adjacent to the somites that will give rise to hypaxial muscles, sub-epidermally into a cavity that naturally forms against the axis as the dermis extends from the body axis to the yolk. Injections were carried out in α actin GFP embryos at two different stages of development: the 16-20 somite stage, before myoblasts had begun delaminating from the somites; and at 28 hpf, during delamination and migration of myoblasts from the somites. Both sets of embryos were allowed to develop and analyzed at 48 hpf.

Injection of the antibody at 16- to 20-somite stage resulted in a reduction to complete absence of GFP-positive cells in the dorsoventral muscle masses of the fin and PHM in 49% of injected embryos(Fig. 7H, n=47). Fin bud formation was not affected in these embryos, as evidenced by its normal development, viewed by DIC optics (Fig. 7I). However, if the anti-HGF antibody was injected at 28 hpf, a reduction, but never a complete absence, in the number of GFP-positive cells in the dorsoventral muscle masses was seen within the fin, suggesting that cells that had already exited the somites could migrate normally to contribute to the fin musculature, but those that had yet to exit could not(Fig. 7J, 50% n=26). Strikingly, injection of the anti-HGF antibody at 28 hpf also resulted in a gap forming within the migrating PHM (Fig. 7J). The anterior most cells of the PHM on the injected side migrate to the same anterior position as the PHM cells on the contralateral uninjected side, revealing that Hgf function was not required for the onward migration of these cells. However, the presence of a posterior gap within the PHM suggests that Hgf is required for the initial delamination of PHM precursors from the somite, and once the antibody is removed or degraded from the extracellular space, PHM precursor cells are able to exit from anterior somites. This observation also reinforces the unique nature of PHM morphogenesis, which results from the ordered addition of myoblasts distal to the migrating primordia from specific somites, and the retention of this addition order throughout the migration of the PHM to its attachment to the cleithrum. Control injections with an identically prepared antibody raised against an unrelated epitope resulted in no phenotype at any stage at which it was injected (n=20).

The inability of somites from posterior levels to give rise to hypaxial and specifically appendicular muscle when transplanted into anterior somites indicates a precocious commitment of somitic cells to a given myotomal fate. Reinforcing this notion is the observation that the expression of genes that mark fin myoblasts such as lbx1 and mox2 initiates within anterior somites prior to formation of the fin primordia(Neyt et al., 2000). This suggests that the fin may not be the source of fin myoblast specifying signals, despite expressing many of the same genes involved in limb formation in amniotes (Grandel et al.,2000; Ng et al.,2002). The conclusion that must be drawn is that at the stages we have examined, positional information imparted upon nascent somites is fixed at specific somite levels in fish embryos and that this positional information cannot be over-ridden by local inductive signals present within the fin environment.

Why this should be the case remains to be fully resolved, but one clue could come from the stereotypical position that the pectoral fin occupies in all fish species, both extinct and extant. The pectoral fin is always tightly physically linked to the bones of the skull through a series of fish-specific bones such as the cleithrum, and consequently its relative rostrocaudal positioning may not be easily altered. Within tetrapods, the pectoral girdle is structurally and functionally detached from the skull and this lack of physical association is believed to have allowed the fore-limb repositioning or `posteriorization' evident within early and extant tetrapods to have occurred (Goodrich, 1958). This would have required the evolution of a mechanism to uncouple the positioning of paired appendages from a particular somite level and must have necessarily included within it the ability to generate appendicular muscle from any somite level at which the limb became juxtaposed. In early tetrapods,for example, limbs may have needed to be placed at more posterior positions to support the extra body weight that would be evident in evolving modes of aquatic habitation or terrestrial environs. Once such a mechanism evolved, it would have allowed greater freedom in the formation of different tetrapod body plans, resulting in the marked alteration in the number of pre- and post-forelimb vertebrae (a proxy for somite positioning in this context)evident in different vertebrate species(Burke et al., 1995). Fish,with their highly evolutionarily successful design, never required any mechanism other than the specification of fin myoblasts by rostrocaudal identity (presumably imparted by the Hox code) as the positioning of the pectoral fin bud always occurred at the same somitic level regardless of the species involved. Although a detailed phylogenetic survey of the relationship of pectoral fin position and somite number has yet to be performed to definitively support such an argument, an analysis of the existing literature(Killeen et al., 1999; Okamoto and Kuwada, 1991; Ballard and Needham, 1964; Detlaff et al., 1993),together with our direct observations (embryos examined, zebrafish, salmon,trout, carp, stickleback, paddlefish, sturgeon and lungfish; N.J.C. and P.D.C., unpublished) reveal that in every bony fish species for which we could obtain data, the pectoral fin always arises adjacent to the first six somites. Hence, the two layers of regulation of limb myoblast formation elegantly demonstrated by Alvares et al. (Alvares et al., 2003) may in fact represent the primitive `Hox-imparted' mode of specification, which is present in all vertebrates with paired appendages,and the derived `local induction' mode that may have arisen to accommodate the designs of the tetrapod body plan.

Fate mapping and somite transplantation have defined a hypaxial muscle with a distinct origin and morphogenesis to that of the fin musculature. Although the continuous nature of the extension of the PHM from its origin is reminiscent of inter-limb level hypaxial muscle formation in amniotes and hypaxial muscle formation in general within cartilaginous fish (the primitive mode of hypaxial muscle morphogenesis), we believe that the two processes are not related. First, the PHM highly expresses markers such as lbx(Neyt et al., 2000) and met (this study), which specifically mark limb/fin and anterior migratory myoblasts in this and other contexts. Furthermore, inter-limb level somites exhibit the primitive epithelial morphogenesis of hypaxial muscle formation and the PHM does not migrate as an epithelium. However, we believe the existence of the PHM and the fin associated migration that it undergoes may underlie a number of previous reports, based on comparative morphology,that suggest the pectoral fin muscle of a number of teleost species are produced via the primitive mode of epithelial extension (reviewed by Galis, 2001). We also believe that the PHM is not likely to have a directly analogous muscle in amniote species, given its attachment to the cleithrum, which is a bone of dermal origin found predominantly in fish and primitive tetrapods.

We have demonstrated that the restricted competence of anterior zebrafish somites to produce hypaxial and appendicular muscle is mirrored by a restricted expression of met. In contrast to the postulated function of Met in stimulating active migration of amniote limb myoblasts(Birchmeier et al., 2003), we can only implicate Met-mediated signaling in the control of the initial delamination of hypaxial muscle precursors from zebrafish somites. However, we note that little genetic evidence exists to definitively show that Met signaling is required during the migratory process in amniotes, as conditional knockout mice have yet to be generated that remove Met or Hgf function solely during the migratory period. Furthermore it should be noted that our results cannot rule out a role for Hgf in mediating cell proliferation and differentiation of post-migratory myoblasts within in the fin environment as has been postulated previously to occur for post-migratory limb myoblasts in the chick (Scaal et al.,l999).

Further models of Met and Hgf function developed from results of in vitro studies suggest a chemotactic role for Hgf in guiding migration of stimulated cells (Lee et al., 1999). Again, our results do not support such a role, as beads implanted with Hgf protein do not alter the path of migration of either fin myoblasts or cells of the PHM, instead resulting in a precocious delamination of cells from Met-expressing somites. These results are in line with previous studies in chick and mice that similarly discount a chemotactic role for Hgf in limb myoblast migration (Heymann et al.,1996; Dietrich et al.,1999).

How might Met signal activation control proneuromast deposition from the PLLP? Our results suggest that whatever this mechanism may be, it is independent from the process that directs PLLP migration, which has been shown to be controlled by the chemokine SDF1 and its receptor CXCR4(David et al., 2002), as PLLP migration proceeds normally in Met-deficient embryos.

One clue may come from the fact that the PLLP, despite its migratory morphogenesis, expresses a number of markers of epithelial identity at high levels. We have shown that a number of these markers, such as prox1and met itself are specifically downregulated in the trailing edge of the PLLP before proneuromast deposition occurs. One possible model, which is consistent with the available data, is that migration of the PLLP beyond discrete regions of Hgf expression, present at somite boundaries, could generate pulses of Met-mediated de-epithelization within a subset of predetermined cells of the PLLP (Itoh and Chitnis, 2001). Once sufficient somite boundaries have been traversed by the PLLP and trailing edge cells have downregulated epithelial markers sufficiently, they loose contact with the remainder of the primordium and are released. This model explains a number of puzzling features of neuromast deposition, including its attenuated periodicity of approximately every five somites and the association of differentiated neuromasts with somite boundaries. Furthermore, the model suggests a novel mechanism of`de-epithelialization thresholding', which is controlled by the Met receptor,may act as a developmental timekeeper for cellular morphogenesis.

We thank Kate Hammond and Tanya Whitfield for probes and reagents. We are indebted to Tom Schilling, Susanne Dietrich and Tom Hall for comments and suggestions on the manuscript. This work was funded by the Scottish Hospital Research Trust, Medical Research Council UK and the Wellcome Trust UK.