The chinless mutation and neural crest cell interactions in zebrafish jaw development

In vertebrates, much of the head skeleton and peripheral nervous system develops from migratory cells of the neural crest. Neural crest cells arise from dorsal regions of the neural tube and form the anterior neurocranium and segmented pharyngeal skeleton, as well as sensory neurons, glia and pigment cells throughout the body (LeDouarin, 1982). The cranial neural crest of the pharyngeal arches, in particular, is thought to produce signals that organize the patterns of surrounding tissues such as mesodermally derived myogenic cells (Noden, 1983). Thus, an important issue in vertebrate development is how these neural crest cells are specified to form the appropriate cell types and relate this spatial information to their neighbors.

Neural crest cells form the pharyngeal skeleton while mesoderm forms the muscles, and these two populations interact extensively with each other and with surrounding epithelia (Platt, 1893; Hall, 1980; Schilling and Kimmel, 1994; reviewed in Hall, 1987; Lumsden, 1988; Noden, 1988; LeDouarin et al., 1994). These eventually form the reiterated bones and muscles of the jaw and branchial arches that function in feeding and breathing. Skeletomuscular patterning has been best characterized in the avian embryo where neural crest cells migrate segmentally and their fates in the skeleton are determined prior to their emigration from the neural tube (Lumsden et al., 1991; Noden, 1988). When transplanted heterotopically, neural crest cells organize ectopic cartilages and muscles specific for their original anterior-posterior positions (Noden, 1983). The transplanted cells themselves form cartilage and connective tissues and these cells reorganize surrounding mesoderm to form muscles with ectopic orientations and attachment sites. These results have led to the proposal that cranial neural crest cells or their skeletal and connective tissue derivatives pattern the ventral head region during normal embryonic development. Thus the regional patterning of the neural crest can now be seen as providing a basis for the establishment of organizing centers that, after migration, generate positional information. The organizing properties of neural crest may be a general feature of vertebrate embryos, since similar contributions of neural crest to cranial cartilage have been described in every vertebrate species that has been examined (reviewed in Smith and Hall, 1990).

Some of the positional cues that specify neural crest cells and their interactions in the arches have been identified but there have been few insights into their genetic or molecular basis. Notable exceptions are relatives of Antennapedia and Bithorax genes of flies, the Hox genes, and their roles in particular head segments and cell types (reviewed in Krumlauf, 1994). Mouse mutants for some Hox genes show homeotic transformations of pharyngeal arch segments or lack subsets of neural-crest-derived cell types. For example, the loss of Hoxa-2 function results in some skeletal elements of the second (hyoid) arch developing characteristics of first (mandibular) arch structures, suggesting that this gene normally specifies the precursors of hyoid cartilages and bones (Rijli, 1993). This has supported the idea that segment identity in the head periphery is transmitted from the neuroectoderm through a Hox code in the neural crest (Hunt et al., 1991a,b).

To learn developmental genetic mechanisms that establish cell fates in the arches, we are using random mutagenesis in the zebrafish to identify genes whose products are necessary for jaw and gill development. In the zebrafish, a mutational approach has led to the identification of genes involved in early embryogenesis (Streisinger et al., 1981; Kimmel, 1989; Hatta et al., 1991; Halpern et al., 1993 ; Mullins et al., 1994; Solnika-Krezel et al., 1994). Here we describe mutations at the locus chinless (chn^b146 and chn^pa20) that produce embryos lacking pharyngeal cartilages and muscles. We use cell transplantation to show that cells derived from wild-type embryos can autonomously form cartilage in chn mutant hosts and rescue aspects of skeletal patterning. In contrast, mesodermal cells transplanted from mutant donors form muscles in a wild-type environment, where proper influences from their neural crest neighbors are restored. We suggest that chn is required for neural crest signals involved in cartilage and muscle development.

Zebrafish (Danio rerio) from the laboratory colony at the University of Oregon were used in this study. Embryos were obtained either through spontaneous or induced spawnings, raised at 28.5°C and staged according to Kimmel et al. (1995).

The original allele, chn^b146, was induced by exposure of midblastula embryos to gamma-rays, as described previously (Walker and Streisinger, 1983). Recessive mutations were subsequently found in the haploid, parthenogenetic progeny of the mutagenized females by fertilizing their eggs with sperm rendered genetically impotent by exposure to ultraviolet light (Streisinger et al., 1986). Mutants were identified by visual inspection of the jaw and gills at low magnification (∼80×) during the late hatching period (approx. 72 hours after fertilization; Kimmel et al., 1995). A second allele, chn^pa20, was found by screening diploid, F₂ intercross progeny.

An antibody originally produced against the Drosophila Invected protein (4D9) was used to reveal the expression of Engrailed (Eng) proteins in mandibular muscle precursors (Hatta et al., 1990). Briefly, embryos were fixed overnight in 4% buffered paraformaldehyde, frozen in acetone to increase permeability, incubated in primary and secondary antibodies, and reacted using the peroxidase anti-peroxidase method and diaminobenzidine. They were then dehydrated, cleared in methylsalicylate and whole mounted on their sides for photography.

For cartilage staining, embryos were fixed for 2 hours to overnight in buffered 4% formaldehyde, rinsed briefly in distilled water and transferred directly into a 0.1% solution of Alcian blue to stain overnight (Dingerkus and Uhler, 1977). They were then cleared, first by incubating for 1-4 hours in 0.05% trypsin followed by bleaching pigmentation in a 3% hydrogen peroxide solution, cleared in 70% glycerol and whole mounted. For a more general histological overview, we used a ‘trichrome’ stain in which embryos at late hatching stages were immersed in Bouin’s fixative for 24-48 hours, embedded in Paraplast and sectioned in the horizontal plane. Serial sections were cut at 10 μm, treated with HgCl₂ and stained with Mallory’s triple stain (Pantin, 1960). Terminology follows the conventions for zebrafish and other ostariophysines (Cubbage and Mabee, 1996).

Embryos were fixed in 4% paraformaldehyde dissolved in phosphatebuffered saline. In situ hybridizations were performed with antisense riboprobes as previously described (Thisse et al., 1993). The templates used for synthesizing the probes for dlx2 have been described previously (Akimenko et al., 1994).

For clonal analysis of neural crest cell fates, single cells were injected in the premigratory neural crest in a series of embryos at early segmentation stages (12-13 hours) as described in Schilling and Kimmel (1994). Embryos were mounted in 1.5% agar and a 5% solution of tetramethylrhodamine-dextran (Molecular Probes) was iontophoresed intracellularly into single neural crest cells. The animals were then removed and allowed to develop for observations at pharyngula (24 hours) and late hatching (72 hours) stages. The clones were visualized using a SIT video camera to amplify fluorescence.

Donor embryos were labeled at the 1-cell stage with yolk injections of tetramethylrhodamine dextran, a non-fixable fluorescent tracer, or a combination of this and a fixable biotin-dextran (Dingerkus and Uhler, 1977). A suction micropipette was used to transplant neural crest cells from future anterior hindbrain and midbrain levels orthotopically to one side of the head of an unlabeled sibling host at early segmentation stages (12-13 hours). At this stage, some of the neural crest cells that will contribute to mandibular and hyoid arches have segregated at the interface between the lateral ectoderm and the neural keel to form a large lateral mass but have not yet begun to migrate (Schilling and Kimmel, 1994). These segregated cells or their unsegregated neighbors in the keel were moved, typically in groups of six to ten cells. Mutants were indistinguishable from their wild-type siblings at the time of transplantation. Thus all experiments were done blindly and individual donors and hosts were kept separate until they could be identified phenotypically 18 hours later. The cases where transplants were performed between wild-type and wild-type embryos, or mutant and mutant, served as controls.

Mesoderm was transplanted earlier, in the blastula or early gastrula periods, but with similar techniques. Blastomeres from donor embryos, labeled as described above, were transplanted into the involuting, marginal zones of unlabeled hosts. Most or all of these cells involuted to join the hypoblast as it converged towards the body midline and only the most dorsal of these formed axial structures (notochord, hatching gland), while most were paraxial, as shown previously in fate mapping studies (Kimmel et al., 1990) of labeled blastomeres. We did not study embryos in which we supposed, by the presence of donor cells in ectodermal derivatives, mesodermal development may have been influenced by the presence of genotypically different ectoderm.

A gamma-ray-induced mutation chn^b146 (Fig. 1) was identified among the parthenogenetic haploid offspring of a female that had been irradiated at the midblastula stage, as part of a systematic F₁ screen for mutants that disrupt embryonic zebrafish development (Walker and Streisinger, 1983). The phenotype segregated as a single recessive trait. In crosses between chn/+ heterozygotes, 25% of the progeny showed the phenotype (n=2013), 50% of the haploid progeny of heterozygous females were mutant (n=1256). A second mutation, chn^pa20, chemically induced with ethylnitrosourea, was then recovered in a subsequent screen that had a phenotype indistinguishable from chn^b146. The two mutations fail to complement and we infer they are allelic. Transheterozygotes between the two mutations are phenotypically identical to homozygotes. Neither allele shows visible dominant effects in heterozygotes. The mutations are lethal and homozygous mutants die within 5 days of development. Nearly all of the analyses described here were of the original allele, chn^b146.

The chn mutant phenotype can first be recognized early in the pharyngula period, around 30 hours, by the small eye size and elongated noses of mutants as compared to wild-type siblings (Fig. 1A,B). By the late hatching period, 72 hours, mutants have reduced heads and lack arch-derived structures of the jaws and gills (Fig. 1C,D). chn mutant embryos lack the entire pharyngeal skeleton, including Meckels cartilage and palatoquadrate of the mandibular arch, the hyosymplectic, ceratohyal, interhyal and basihyal of the hyoid arch, and basibranchials, hypobranchials and ceratobranchials in arches 3-7. All of these cartilages originate, at least in part, from the neural crest (Schilling and Kimmel, 1994). Some cartilage does form in the chn neurocranium, although overall the elements are strongly reduced in size, including the more anterior trabeculae which are thought to originate from the neural crest. Finding neurocranial cartilage development reveals that lack of chn function does not simply block biosythesis of a key cartilage component and suggests that it is more likely a patterning or specification defect.

As determined further in sectioned material, chn embryos have no cartilage or muscle in the arches and, as a consequence, no protruding jaw (Fig. 1E,F). The mouth opens further ventral and posterior in mutants than in wild type. The phenotype probably is the result of failure of the extension of the jaw and the anterodorsal protrusion of the mouth that characterizes embryos late during the hatching period. Somite-derived trunk and tail muscles are generally well developed in mutants.

Defects in the pharyngeal arches of chn could be the consequence of impaired migration of neural crest. However, mesenchyme that forms the arch primordia appears normal in mutants as determined by inspection, Nomarski optics, utilizing live embryos (not shown) and also by the expression patterns of several genetic markers including dlx2 (Fig. 2A,B). Three major streams of neural crest that express dlx2 appear normal in size and position in embryos with 10 somites derived from incrosses between heterozygous chn/+ parents. By the pharyngula period, when the phenotype becomes visible morphologically, pharyngeal arch primordia and early differentiating, neural-crest-derived sensory neurons in the cranial ganglia appear normal in chn (Fig. 2B). Hindbrain rhombomeres also appear normal in number and size. We do not know if there are additional defects in other neural crest derivatives such as enteric neurons or in the valves of the heart. However, chn mutant embryos have a strong heartbeat and apparently normal gut motility.

To determine if neural crest cells contribute to the mutant mesenchyme, clonal analyses of wild-type and chn embryos were compared (Fig. 3; Table 1). In wild-type embryos, single neural crest cells generate clones of cells that all express the same fate and migration occurs in predictable patterns. Hence we can predict the fates of cells by their premigratory positions; cartilage and pigment precursors lie more dorsomedially than neurogenic crest cells (Schilling and Kimmel, 1994). In chn, neural crest cells labeled with rhodaminedextran from the medial, largely chondrogenic, region migrated ventrally into the arches as in wild-type embryos, but remained undifferentiated in positions where cartilage normally forms. By late hatching (72 hours) stages, labeled crest cells in the mutant pharynx had no recognizable differentiated morphology, whereas in wild types many cartilage cells have already differentiated. Other neural crest derivatives normally derived from the same mass of premigratory crest, including glial cells of cranial ganglia and pigment cells, were as commonly observed in mutants as in wild types. Thus we propose that at least some undifferentiated cells in the pharyngeal arches of chn mutants are arrested cartilage precursors.

chn mutants also lack the segmental array of striated muscles that normally are associated with the arch cartilages, but that stem from mesoderm (Schilling and Kimmel, 1994). To determine if precursors for two jaw muscles were present, we used an antibody that recognizes zebrafish Eng homeoproteins. Precursors of two mandibular muscles, the levator arcus palatini and dilator operculi, in wild types express Eng in their nuclei beginning during the pharyngula period (30 hours), as they migrate, condense, elongate and form striated muscle fibers by 72 hours (Fig. 4; Hatta et al., 1990) In chn mutants, a group of nuclei were labeled in a similar location in the mesenchyme at 30 hours, but did not differentiate as late as 5 days of development, several days after hatching, when the young fry is normally feeding actively. Thus, along with cartilage precursors, at least some muscle precursors derived from the paraxial mesoderm are present in chn but do not differentiate.

To learn if chn mutations block neural crest lineages directly and to define its time of action, we analyzed mosaics, made by transplanting fluorescently labeled wild-type or mutant cells into unlabeled hosts at several different stages (Fig. 5; Table 2). Cells were moved orthotopically into the premigratory neural crest at anterior hindbrain and midbrain levels on only one side of an unlabeled host, of either genotype, and followed until their differentiated phenotypes could be determined.

Wild-type neural crest cells form cartilage in chn hosts (Figs 5, 6, 7). When wild-type neural crest cells were transplanted into the chondrogenic region of the premigratory neural crest in embryos derived from crosses between chn/+ heterozygotes at 12 hours, they formed cartilage in mutant hosts in four of five successful transplants (Table 2), partially rescuing the mutant phenotype. At the time of transplantation the genotype of the host was unknown and was ascertained later by determining if cartilage was present on the unoperated side. By ‘rescue’ we usually mean that some cartilage cells were visible in the mutant host, not that a fully patterned skeletal structure has formed. Typically, cells formed only small clusters of cartilage (Fig. 6) on the transplanted side. However, in one case the cells formed a well developed cartilage bar that was unilateral and stopped abruptly at the ventral midline (Fig. 7). The experiments show that groups of wild-type neural crest cells, transplanted at this stage, differentiate autonomously into cartilage in mutant hosts. The results suggest that cues appropriate to signal cartilage development are present in the mutant environment and that the primary action of the chn mutation is directly in the neural-crest-derived cartilage precursor cells..

However, the same analysis showed that adjacent mutant cells also formed cartilage in these mosaics, chn cells in this case acting according to the genotypes of their wild-type neighbors (Fig. 5, 6; Table 2). Interestingly, in 100% of rescued embryos, the cartilage contained unlabeled mutant cells in addition to the labeled wild-type cells. Mutant chondrocytes were found as far as ten cell diameters away from the wild-type cells but were always incorporated into a contiguous mass of cartilage. Wildtype cells were always located at the distal ends of rescued masses of cartilage, suggesting that they had migrated through and interacted at close range with differentiating mutant cells. Recruitment of mutant cells into cartilage only occurred in the presence of wild-type cartilage in the mosaic, not other neural-crest-derived cell types (e.g. pigment cells, neurons).

In reciprocal transplants, between chn donors and wild-type hosts, mutant cells readily form cartilage (Table 2). In these cases, every mutant cell had a wild-type neighbor. Thus, regardless of the donor or host genotype in mosaics, mutant crest cells could be recruited into cartilage differentiation.

An interaction that seems to maintain the chondrogenic ability of neural crest cells was revealed by mosaics made only one hour later in development (13 hours) the time when neural crest migration begins. Neither wild-type nor mutant cells form cartilage in mosaics made at that time, while wild type to wild type controls at this same stage often do (Table 2). Whereas transplanted mutant cells migrated successfully into the arch primordium in a wild-type host, they were never incorporated into the developing cartilage (0/10). Crest cells that migrated into other regions in these mosaic embryos formed pigment cells and several neural derivatives. Thus mutant cells, as well as their environment, quickly and specifically lose chondrogenic potential, possibly due to the loss of necessary interactions during migration that were succesfully restored in mosaics made earlier at 12 hours.

Surprisingly, when chn embryos were used as hosts in reciprocal transplants at this stage, wild-type cells never formed cartilage. Rather they remained apparently undifferentiated after migrating successfully, for as long as they could be followed (4-5 days). The mesenchymal morphologies of these labeled cells resembled the undifferentiated clones observed in the single cell labeling experiments (see Fig. 3) in chn. Thus the mutant environment had lost the ability to support chondrogenesis.

The absence of head muscles in chn may be a direct effect of the mutation on paraxial mesoderm, or a secondary effect due to environmental factors, such as the lack of interactions with adjacent neural crest cells. To address this issue, paraxial mesodermal precursors were transplanted between mutant and wild-type embryos (Fig. 8, Table 3). Labeled cells were placed along the involuting, marginal zone of the gastrula so that they later entered the hypoblast that gives rise exclusively to mesodermal and endodermal derivatives (Warga and Kimmel, 1990). Transplants spread extensively along the body axis, contributing to both head and trunk mesodermal tissues or endodermal tissues. We observed that wild-type precursors of the head mesoderm, transplanted at the late blastula stage, do not differentiate into striated pharyngeal muscles in chn hosts. In contrast, mutant mesoderm can contribute to head muscles in a wild-type host (Fig. 8). Mutant muscle differentiation did not appear to require adjacent wild-type muscle cells, since, in some cases, mutant cells appeared to constitute entire muscles, such as the adductor mandibulae (am) and intermandibularis posterior (imp) muscles, with no wild-type muscle cells nearby. However, muscles formed from mutant donor cells always neighbored wild-type cartilage and connective tissues.

We have described recessive, noncomplementing lethal mutations in the zebrafish, chn^b146 and chn^pa20, that disrupt the development of pharyngeal arches. Induced by different means both mutations produce similar phenotypes. We infer that both involve a single gene, chinless, and that the mutant alleles are severe, perhaps null. chn homozygous embryos do not develop neural-crest-derived cartilages or mesodermderived muscles in the head. From the early expression of the mutant phenotypes (∼30 hours), we infer that chn must normally function within the arch precursor populations since differentiation of both of these cell types does not normally occur until mid to late hatching (60-72 hours; Schilling and Kimmel, 1994; Kimmel et al., 1995). Transplanting groups of neural crest cells from wild type into mutants to make mosaic embryos can partially restore the missing cartilage phenotype. Transplanting wild-type mesodermal cells cannot restore muscles. Based on these results, we suggest that chn normally functions in the specification of chondrogenic neural crest cells, not in mesoderm, and that this specification occurs at an early, premigratory stage.

In addition, at least two types of cell signaling events that involve neural crest appear to be revealed by the mutation. Mutant cells can make pharyngeal cartilage in mosaics, whereas they never do when they develop in mutants, perhaps reflecting ‘homeogenetic’ induction (Hamburger, 1988) between cartilage precursors, i.e. a developing chondrocyte influences its neighbors to also form cartilage (as discussed further below). Since mutant cells participate productively in this interaction, it is clear that chn does not block the ability of cells to respond, nor does it seem to block the signal. Homeogenetic induction between cartilage cells may be blocked in chn because specified cartilage precursors are not present to initiate it. A second signal, from neural crest to surrounding mesodermally derived myogenic cells, probably occurs later when the two lie together in the pharyngeal primordium (Noden, 1988).

Interestingly, there is an additional temporal aspect of chn action revealed by transplanting neural crest at different stages. The ability for neural crest cells to respond to and probably produce a signal is lost during a narrow window of time as cartilage no longer develops in mosaics, regardless of the host genotype. Thus, at this later stage both the mutant cells and their environment may have lost some aspect of chondrogenic ability. The time period of loss roughly corresponds to the time when neural crest migration begins (Schilling and Kimmel, 1994; Kimmel et al., 1995).

Cranial neural crest cells participate in the formation of the skeletal components and connective tissue of the face and neck as well as many pigment and neural cell types (D’Amico-Martel and Noden, 1983; Hammond and Yntema, 1964; LeDouarin, 1982; Noden, 1983). The smaller eye size and slightly reduced pigmentation in mutant embryos, along with the cartilage defect, suggests that chn may play some role in the development of the neural crest cells that contribute to these structures. Similarly, eye reduction in the small eye mutation in mice results from defects in midbrain neural crest migraton (Matsuo et al., 1993). It is of particular interest that sensory neurons appear to be specifically unaffected in chn. This specificity is reminiscent of some neural crest mutations, such as Patch (Ph), in the mouse embryo (Morrison-Graham et al., 1992) which similarly spares neuronal derivatives but severely affects crest-derived melanocytes. Ph/Ph mutants have a deletion of the beta-subunit of the receptor for plateletderived growth factor (Stephenson et al., 1991).This suggests that the function of chn, like Ph, is needed for the normal development of a non-neuronal subset of neural crest derivatives. The absence of the pharyngeal cartilage derivatives of neural crest and not other derivatives in chn, reveals a subtle genetic control over crest cell specification that we propose occurs before migration, as based on our mosaic analyses. By selectively disrupting the development of particular cell types, mutations like chn may also allow us to begin to understand the roles of each cell type in head patterning.

The many patterning roles of neural crest in the pharyngeal arches clearly involve a complex network of interactions. However, we can explain the results of our mosaic analyses in a unified way by supposing that the chn phenotype reflects a defect in cartilage specification that results not only in an intrinsic defect in the cells’ specification, but also results in mutant cells unable to support subsequent pharyngeal chondrogenesis. This idea is discussed further elsewhere (Schilling and Kimmel, 1994). In the future, by performing heterochronic transplantations, we can distinguish whether this is because mutant cells stop modifying their environment or lose their ability to respond to it.

Taken together, the results suggest that, while the wild-type function of chn may be to autonomously specify pharyngeal chondroblasts, there are several non-autonomous aspects to the mutation, involving both early and late cell signalling. An early specification of a subset of the cranial neural crest might be expected in zebrafish, since cartilage arises from a distinct region of the fate map of premigratory crest and since single cells are clonally restricted to cartilage (Schilling and Kimmel, 1994). Growth factor/receptor systems or extracellular matrix components are obvious candidates for the signals, since they are required for the normal development of subsets of neural crest derivatives in mammalian embryos (Morrison-Graham and Takahashi, 1993). For example, interactions between Steel-factor and c-kit, its receptor, have been demonstrated to influence survival, proliferation and differentiation of melanocytes as well as cells of other tissue types. Members of the TGF-β family of signalling molecules, particularly the bone morphogenetic proteins, are also attractive candidates for mediating such interactions, with known roles in initiation and differentiation of the skeleton in mammals (reviewed in Kingsley, 1994).

In mosaics made by cell transplantation, chn neural crest cells will only form pharyngeal cartilage in the presence of other wild-type cartilage cells. This suggests that the wild-type precartilage cells homeogenetically induce neighbors, bypassing any deficiency in chn cells. Cartilage precursors may normally interact to provide the necessary specification signals, since wild-type neural crest cells induce differentiation in their mutant neighbors. In the cartilage that forms, mutant cells are not always adjacent to wild-type cells but always have cartilage neighbors. Neural crest cells are known to produce extracellular matrix material that might mediate such interactions (Weston et al., 1978; Hay, 1981). Similar ‘homeogenetic’ inductive interactions have been observed between floor plate precursors in the CNS (Hatta et al., 1990) and may represent a common mechanism for organizing discrete, coherent groups of cells with similar fates (DeRobertis et al., 1989).

chn mesodermal cells can make pharyngeal muscles in a wild-type environment. Because chn is required autonomously for cartilage development, we suggest that muscle differentiation may be blocked nonautonomously in chn mutants, because necessary cues from surrounding neural-crest-derived cells are absent. In mutants, some mesodermal cells begin to develop along a muscle pathway, as inferred from Eng expression, but then they are blocked. Wild-type neighbors, possibly in the cartilage, may overcome this block, although we have not observed this directly for the Eng-expressing muscles. Since muscles containing mutant cells are always located near wild-type cartilage, we presume that the interactions occur in the pharyngeal arch primordium after neural crest migration.

The results are not unexpected, based on heterotopic transplantations of crest cells in avian embryos. In such transplants, ectopic skeletal structures reorganize muscles around them (Noden, 1983). Furthermore, the patterning role of these skeletal and connective tissues may reflect a common type of mesenchymal signal, regardless of their germ layer origins. Experiments on both the limb (Chevallier, 1979; Chevallier and Kieny, 1982; Gumpel-Pinot, 1984) and the head (Noden, 1983) have clearly demonstrated that patterns of myocyte condensation and subsequent muscle growth are dependent upon prepatterns within nonmyogenic connective and skeletogenic tissues. Our results in chn are consistent with such a patterning process in the mandibular arch and may add information about its control. Simply the presence of the appropriate neural-crest-derived cells is not sufficient for muscle patterning, but may depend on some crest function that depends on the early function of chn⁺.