The zebrafish colourless gene regulates development of non-ectomesenchymal neural crest derivatives

The process whereby precursor cells become specified to generate progeny with distinct fates is fundamental to development. The neural crest, a vertebrate tissue that generates cells of diverse fates (Le Douarin, 1982), has been widely used to address this issue. Crest-derived fates fall into 4 broad categories: pigment cells, neurons, glia and ectomesenchymal cells. Each category includes several distinct, terminally differentiated cell types; for example, in fish, ectomesenchymal cells include both craniofacial cartilage cells and median fin mesenchyme (Schilling and Kimmel, 1994; Smith et al., 1994).

Neural crest cells delaminate from the neural tube and migrate on well-characterised paths (Weston, 1970; Bronner-Fraser, 1986; Loring and Erickson, 1987; Krotoski et al., 1988; Serbedzija et al., 1989, 1990; Raible et al., 1992; Collazo et al., 1993). Premigratory neural crest is described operationally as a mixture of unrestricted and fate-restricted precursors. An unrestricted precursor generates progeny of two or more types, whereas a fate-restricted precursor generates progeny of only one type. However, these lineage-based definitions do not reveal cell potency, which must be tested independently from fate (see Eisen and Weston, 1993; Raible and Eisen, 1996). Clonal studies in vivo demonstrate that premigratory neural crest cells frequently generate clones containing more than one fate (chick, Bronner-Fraser and Fraser, 1991; Frank and Sanes, 1991; Fraser and Bronner-Fraser, 1991; mouse, Serbedzija et al., 1994; frog, Collazo et al., 1993; zebrafish, Raible and Eisen, 1994). Cultured mammalian and avian neural crest cells,also frequently generate mixed fate clones, as well as fate-restricted clones (Sieber-Blum and Cohen, 1980, Baroffio et al., 1988; Dupin et al., 1990; Baroffio et al., 1991; Stemple and Anderson, 1992; Rao and Anderson, 1997). At its most extreme, in vivo lineage studies suggest that zebrafish premigratory cranial crest consists of only fate-restricted cells (Schilling and Kimmel, 1994).

The mechanism and timing of specification of individual neural crest fates from unrestricted precursors has been a matter of debate and remains largely unknown. Baroffio and colleagues showed that cultured, migratory, avian cranial neural crest includes a very rare class of unrestricted cells, capable of generating progeny that differentiated into all classes of crest derivatives (Baroffio et al., 1988; 1991). However, most clones generated only a subset of fates. Such observations have lead to the suggestion that neural crest development involves a progressive series of restrictions in developmental fate. Thus, unrestricted precursors are postulated to generate progenitors restricted to a subset of fates, which then give rise to fate-restricted daughters (Dupin et al., 1990; Baroffio et al., 1991; Weston, 1991; Anderson, 1993; Le Douarin et al., 1994; Lo et al., 1994). Several progenitors generating specific subsets of fates have been well-characterised, including sympathoadrenal precursors (Anderson and Axel, 1986) and neuronoglial precursors (Dupin et al., 1990; Frank and Sanes, 1991; Pomeranz et al., 1993; Henion and Weston, 1997). The molecular genetic basis of these fate restrictions is poorly understood. Although in vitro some growth factors can direct the fate choice of individual neural crest stem cells, there is no evidence for progressive restriction in this paradigm (Stemple and Anderson, 1992; Shah et al., 1994, 1996; Morrison et al., 1997; Shah and Anderson, 1997).

Models of neural crest specification by progressive fate restriction predict the existence of genes that regulate generation of combinations of neural crest fates. Mutants in such genes would essentially lack cells of those fates, with other derivatives being unaffected. Several mouse mutants have defects in specific subsets of neural crest fates, although the combinations of derivatives affected are perhaps surprising. Thus, mutations in 3 genes, piebald (s), lethal spotting (ls) and Dominant megacolon (Dom) combine defects in the enteric nervous system and melanocytes. Homozygous mutant s and ls mice have a phenotype very similar to Dom heterozygotes, showing reduced melanocytes and posterior colon aganglionosis (Mayer and Maltby, 1964; Webster, 1973; Lane and Liu, 1984; Baynash et al., 1994; Hosoda et al., 1994; Pavan and Tilghman, 1994). Therefore, all three genes are implicated in development of both enteric neurons and melanocytes. In addition, Dom homozygotes have extensive peripheral nervous system (PNS) defects outside the enteric nervous system (Herbarth et al., 1998; Southard-Smith et al., 1998). Molecular characterisation shows that s and ls encode endothelin receptor B (Hosoda et al., 1994) and its ligand, endothelin-3 (Baynash et al., 1994) and Dom encodes the Sox10 transcription factor (Herbarth et al., 1998; Southard-Smith et al., 1998). The precise roles of these genes remains unclear; functions in proliferation, survival and differentiation of crest fates have been proposed (Lahav et al., 1996; Reid et al., 1996; Stone et al., 1997).

In zebrafish, in vivo clonal studies demonstrated that the majority of premigratory crest cells are fate-restricted (Raible and Eisen, 1994; Schilling and Kimmel, 1994). Clonal analysis of trunk crest cells is consistent with production of fate-restricted precursors as daughters of initially unrestricted cells (Raible and Eisen, 1994). Evidence for a completely unrestricted precursor, one forming all major classes of crest derivatives, has not yet been reported.

The advantages of zebrafish for in vivo analyses of cell behaviour are nicely complemented by extensive collections of mutations available for study. Many mutations affecting genes crucial for neural crest development have been isolated (Henion et al., 1995; Driever et al., 1996; Haffter et al., 1996; Kelsh et al., 1996; Neuhauss et al., 1996; Odenthal et al., 1996; Piotrowski et al., 1996; Schilling et al., 1996a,b). To date, characterisation of most of these mutant phenotypes are in early stages. chinless, the best-characterised zebrafish neural crest mutant, has profound defects in craniofacial cartilage consistent with involvement in cartilage specification (Schilling et al., 1996b). Mutants with defects in more than one neural crest category have been described (Henion et al., 1995; Kelsh et al., 1996; Neuhauss et al., 1996; Odenthal et al., 1996; Schilling et al., 1996a). However, it remains to be seen whether any of these mutations affect cell numbers of two or more derivatives, as would be expected for genes involved in progressive fate restriction. Thus, a zebrafish mutation affecting several, but not all, neural crest fates would be particularly interesting. We report here studies of mutations in a zebrafish gene, colourless (cls), in which homozygous mutants show such a phenotype.

cls mutations were isolated independently in two large-scale mutagenesis screens (Kelsh et al., 1996; Malicki et al., 1996). Other than severe defects in pigmentation and ear morphogenesis (Malicki et al., 1996; Whitfield et al., 1996), these mutants show few overt defects prior to death at 10 days post-fertilisation (dpf). Initial phenotypic characterisation demonstrated a dramatic reduction in all three pigment cell types. Here we describe additional, pronounced neural crest defects in neural and pigment cell fates. However, ectomesenchymal fates are remarkably unaffected. Furthermore, we demonstrate that cls functions cell-autonomously within all three crest-derived pigment cell types. We propose that cls functions early in neural crest development and is important for generation of several fates. Such a phenotype is consistent with that of a gene functioning in development of a precursor restricted to non-ectomesenchymal crest fates.

Embryos from the University of Oregon zebrafish colony were raised at 28.5°C and staged according to Kimmel et al. (1995). We examined embryos homozygous for 3 strong cls^– alleles, t3, tw2 and tw11on AB (t3) or mixed AB/Tübingen backgrounds (tw2 and tw11); since no differences between alleles have been described we do not distinguish them here. Homozygous cls^– embryos were obtained by mating known heterozygotes. We see no evidence of defects in heterozygotes and so refer to homozygous mutants as cls^–. We have no evidence whether these alleles are nulls. albino^–(alb^–) embryos were generated by mating alb^b4 homozygotes (Streisinger et al., 1986).

Neural crest cells on the lateral migration pathway (Raible et al., 1992) were counted by mounting anaesthetised embryos between No. 1 coverslips using No. 1 coverslips as spacers. All cells showing the characteristic fibroblastic morphology of migrating neural crest cells, and positioned between the epidermis and lateral surface of the muscles, were counted for somites 5-15 on one side of the embryo. These counts were performed on embryos between 26-somite and prim-24 stages (22-35 hpf). After counting, each embryo was released from between the coverslips and raised overnight in embryo medium so that mutants could be identified.

RNA in situ hybridisation was performed by the method of Thisse et al. (1993), except probes were not hydrolysed. Probes were labeled with digoxigenin or fluorescein using the Genius RNA Labeling Kit (Boehringer Mannheim) and detected with the appropriate antibody. A blue precipitate was formed by incubating with 5-bromo 4-chloro 3-indolyl phosphate/nitro blue tetrazolium. Probes used: dopachrome tautomerase/tyrosinase-related protein 2 (dct) (R. N. K., B. Schmid and J. S. E., unpublished); forkhead-related gene 6 (fkd6; Odenthal and Nuesslein-Volhard, 1998); dlx2 (Akimenko et al., 1994); sonic hedgehog (shh; Krauss et al., 1993); engrailed3 (en3; Ekker et al., 1992); krox20 (Oxtoby and Jowett, 1993); c-ret (Bisgrove et al., 1997). Antibody staining was performed using peroxidase-antiperoxidase (Sternberger Monoclonals, Inc). Double immunolabelings of dissected zebrafish guts were examined with a Zeiss LSM 310 confocal microscope using fluorescent secondary antibodies (Jackson Immunoresearch Laboratories) and assembled using VoxelView (Vital Images). Primary antibodies used were: anti-Hu, mAb 16A11 (Marusich et al., 1994); anti-acetylated tubulin (Sigma); and anti-GFAP (Axell). Embryos were processed as whole mounts. Immunolabeling of larval stages was performed on cryostat sections, as described by Raible et al. (1992). To view the enteric nervous system, larvae were fixed in 4% paraformaldehyde for 2 hours and intact guts (pharynx to anus) dissected out with fine needles and processed for whole-mount antibody staining.

Images were cropped, colour-balanced, contrast-enhanced and labeled using Photoshop (Adobe).

Cartilage was stained with Alcian green according to the method of Kimmel et al. (1998). Larvae were fixed in 4% paraformaldehyde overnight, washed in acid-alcohol (1.85% HCl, 70% ethanol) and incubated overnight in 0.1% Alcian green in acid-alcohol. Larvae were destained by extensive acid-alcohol washes, rehydrated, bleached using peroxide (1% KOH, 3% H₂O₂), cleared and stored in glycerol (0.25% KOH, 50% glycerol).

Donor embryos were labeled with 10× 103 M_r tetramethylrhodamine dextran (Molecular Probes) by intracellular injection into yolk cell cytoplasm during early cleavage stages. Genetic mosaics were generated by transplanting cells from late blastula donors into shield stage hosts. Known wild-type embryos were from AB stocks. To generate wild type→cls^– chimaeras, we transplanted wild-type cells into host embryos from cls heterozygote crosses. Since at these stages wild-type embryos are indistinguishable from cls^– siblings, these experiments generated many wild type→wild type control transplants. Donors and hosts were raised separately at 28.5°C until their phenotypes could be scored. To generate cls^–→wild type chimaeras, donor and host embryos were reversed. In some experiments, wild-type embryos were replaced with alb^– embryos, which completely lack melanin (Streisinger et al., 1986). The alb^– mutation acts cell-autonomously in melanophores (R. N. Kelsh et al., unpublished).

Pigment cells were scored for pigment type (Bagnara and Hadley, 1973) and rhodamine label. Iridophores and xanthophores could be readily scored for fluorescent label between 3-6 dpf. In contrast, melanin strongly quenched the rhodamine fluorescence and only some melanophores could be reliably scored for rhodamine label at 2 dpf. Cells that could not be reliably scored were not considered.

Decreased melanophore number in cls^– embryos (Kelsh et al., 1996) might result either from a decrease in melanoblast precursors or in melanin synthesis. To distinguish these alternatives, we compared melanoblasts and melanophores in wild-type and cls^– embryos using dct riboprobes (R. N. Kelsh et al., unpublished). In mice, dct is expressed in melanoblasts days before melanin becomes visible (Steel et al., 1992). In zebrafish, we detect melanoblasts several hours before melanisation. From 27 hpf, cls^– embryos were readily distinguished from wild types by their lack of body melanin and distinct dct expression patterns (Fig. 1A,B). Wild-type embryos had many strongly labeled melanoblasts migrating throughout the body (Fig. 1A,C). In contrast, cls^– embryos showed only a few, weakly stained melanoblasts, none anterior to the hindbrain, and none appeared to have migrated (Fig. 1B,D). Similar melanoblast defects were detected in 25% (presumably cls^–) of embryos from a heterozygous cls^– cross by 21 hpf, several hours before melanisation begins.

Since there are no published markers for iridoblasts and xanthoblasts, we examined cls^– embryos for neural crest cells migrating on the lateral pathway. These cells are fated to become pigment cells (Raible and Eisen, 1994) and are strongly reduced in numbers in cls^– embryos (Kelsh et al., 1996). Furthermore, they are severely depleted from very early stages (Fig. 2).

The pigment phenotype in cls^– embryos is reminiscent of s, ls and Dom mutant mouse phenotypes. These mutants also show enteric nervous system defects, so we asked whether the PNS was affected in cls^– larvae. A pan-neuronal marker, anti-Hu antibody (Marusich et al., 1994), revealed large numbers of anti-Hu-positive enteric neurons around wild-type larval guts (Fig. 3A). cls^– larvae have only about 13% as many enteric neurons as wild types along the entire length of the gut (Fig. 3B). Similar results were obtained by in situ hybridisation with another marker for enteric ganglion precursors, c-ret (Bisgrove et al., 1997; Marcos-Gutierrez et al., 1997; data not shown). Comparison of the rostrocaudal extent of enteric nervous system precursors using these 2 markers suggested that enteric neuron differentiation shows a rostrocaudal gradient. Thus, anti-Hu labeled neurons are relatively scarce in the caudal gut at 3 dpf, whereas c-ret labeling reveals many enteric neuron precursors in these caudal regions (data not shown).

Bisgrove and colleagues (1997) suggest the identification of a cluster of c-ret-positive cells in 24 hpf embryos as enteric nervous system precursors, based primarily on their location and the later expression of c-ret in enteric neurons. However, it is perhaps surprising that these cells are apparently unaffected in homozygous alyron mutants, which have severe defects in neural crest formation (Cretekos and Grunwald, 1999), so this tentative identification of these cells needs to be confirmed by a lineage-tracing method. In situ hybridisation with c-ret probes of 24 hpf embryos from cls^+/– crosses shows this cell cluster to be present in all embryos. Thus these putative enteric nervous system precursors appear to be properly specified, at least with respect to this marker, in the absence of cls function.

Dorsal root ganglion (DRG) sensory neurons are strongly affected in cls^–embryos. Anti-Hu immunostaining showed that wild types had a reproducible series of DRGs as early as 2 dpf, but this pattern was severely disrupted in cls^– larvae (Fig. 4). Posterior trunk and tail DRGs were almost absent (Fig. 4G), whilst in the anterior trunk the number was reduced, the pattern lost bilateral symmetry and positioning relative to the neural tube varied (Fig. 4E; Table 1). Although more posteriorly cls^– larvae had no DRGs, there were occasional anti-Hu-positive cells dorsal to the neural tube (Fig. 4C). These cells are reminiscent of extramedullary cells (EMCs) described in various amphibians and teleosts, including zebrafish (Hughes, 1957; Myers, 1985). EMCs were more abundant in cls^– larvae than in wild types.

Zebrafish sympathetic neurons are also detected by anti-Hu immunostaining and by an antibody to tyrosine hydroxylase (P. Henion, personal communication). We were unable to detect differentiated sympathetic neurons in sectioned 9 dpf cls^– larvae, although they were readily seen in wild types (data not shown).

To learn whether cls gene function is required for development of all peripheral sensory neurons, we examined cranial ganglia which are reported to contain neurons derived from both neural crest and ectodermal placodes (Le Douarin, 1982). The neural crest contribution to cranial ganglia has not been thoroughly investigated in zebrafish, although a contribution to the trigeminal ganglion has been shown by 24 hpf (Schilling and Kimmel, 1994). Both anti-Hu antibody and c-ret riboprobes label neurons in developing head ganglia at 24 hpf. We saw no obvious change in ganglia size, position or shape in cls^– embryos compared with wild types using these (data not shown) and other markers, including zn-12 (Trevarrow et al., 1990) and zn-12 plus anti-acetylated tubulin (Chitnis and Kuwada, 1990) (Fig. 5A,B).

In avian and mammalian embryos, peripheral glia arise from neural crest (Le Douarin, 1982). The fkd6 gene is expressed initially in premigratory neural crest (Odenthal and Nuesslein-Volhard, 1998) and later in what appear to be cranial ganglion satellite glia and cranial nerve-associated Schwann cells. At 24 hpf, fkd6 was expressed in peripheral cells of cranial ganglia and cells extending along the posterior lateral line (PLL) nerve (Fig. 6A,C). Based on position and absence of anti-Hu labeling, we suggest these are glial cells (R. N. Kelsh et al., unpublished). cls^– embryos had strongly reduced numbers of fkd6-expressing cells associated with cranial ganglia (Fig. 6B,D) and with the PLL and other cranial nerves (data not shown). Thus we propose that cls mutations affect glial cell numbers in the cranial PNS.

Glial fibrillary acidic protein (GFAP) is an intermediate filament protein expressed in glia of the CNS (astrocytes and radial glia) and PNS (enteric glia; Jessen and Mirsky, 1980, 1983; Bjorklund et al., 1984; Jessen and Mirsky, 1985). In zebrafish larvae, an anti-bovine GFAP antibody cross-reacts with radial glia in the CNS, putative Schwann cells in the PNS (P. Henion, personal communication) and putative enteric glia in dissected zebrafish gut preparations. Wild-type embryos showed a network of GFAP-positive fibres extending over the gut (Fig. 7A), around and between Hu-positive enteric neurons. In contrast, most guts from cls^– larvae showed none of these fibres (Fig. 7B), although occasionally short isolated sections of gut were labeled (data not shown). Thus, cls^– larvae have highly reduced or absent enteric glia.

In view of the widespread defects in all other major crest derivatives, we examined ectomesenchymal derivatives to learn whether they were also affected. dlx-2, a putative marker for early branchial arch crest (Akimenko et al., 1994), was expressed normally in cls^– larvae (data not shown). Alcian green staining showed that all wild-type elements of the craniofacial skeleton were present in cls^– embryos at 3-5 dpf (Fig. 8). However, later cls^– embryos showed a slight retardation of arch development.

Neural crest also generates cells with another putatively ectomesenchymal fate in anamniote vertebrates, fin mesenchyme. These cells are prominent in median fins of zebrafish larvae, and at least some median fin mesenchyme cells have been reported to originate from neural crest (Smith et al., 1994; D. W. Raible, personal communication). Fin mesenchyme cells were found in similar positions and numbers in wild-type and cls^– larvae (Fig. 8G,H). Thus both cranial and trunk ectomesenchymal crest derivatives appeared normal in cls^– larvae.

Since so many neural crest fates were strongly affected in cls^– embryos, we investigated whether premigratory neural crest was reduced. Both sna-2 (Thisse et al., 1995) and fkd6 (Odenthal and Nuesslein-Volhard, 1998) have been reported to label premigratory neural crest cells. At both 14-24 hpf (fkd6) and 20-24 hpf (sna-2) the approximate number, position and intensity of labeling of premigratory crest cells was indistinguishable in mutants and wild type (data not shown).

To learn whether cls mutations affect non-ectomesenchymal neural crest cells autonomously, we transplanted cells between wild-type and cls^– embryos.

Cells transplanted from labeled wild-type donors into unlabeled cls^– hosts were able to form neural crest and migrate on both medial and lateral pathways (Fig. 9A). Fluorescently labeled, wild-type pigment cells in cls^– hosts developed wild-type pigment cell phenotypes (Figs 9C,D and 10; Table 2), even in close proximity to unlabelled cls^– melanophores (data not shown). Only wild-type cells developed into normal chromatophores in cls^– hosts and cls^– pigment cells were not rescued by nearby wild-type cells. Therefore, the cls gene product appears to function cell-autonomously within crest-derived pigment cells.

We predicted that cls^– cells transplanted into the neural crest of wild-type hosts would develop the characteristic cls^– melanophore phenotype. However, this prediction remains untested as we cannot recognise these cells amongst the heavily pigmented wild-type melanophores. As an alternative approach, we transplanted cls^– cells into alb^– hosts (Table 3). alb^– embryos have a full complement of melanophores, based on dct in situ hybridisation (R. N. Kelsh et al., unpublished), but do not become pigmented (Chakrabarti et al., 1983; Streisinger et al., 1986). Wild-type cells transplanted into alb^– hosts often formed normal melanophores, but did not rescue the nonpigmented phenotype of alb^– mutant melanophores (data not shown). Since nearly 75% of wild type→alb^– chimaeras had at least one melanophore, we clearly targeted wild-type cells to the neural crest of alb^– hosts at high frequency. In contrast, we never saw any melanophores with wild-type pigmentation when cls^– cells were transplanted into alb^– hosts, consistent with the conclusion that cls mutations act cell-autonomously. However, somewhat surprisingly, we did not see melanophores with the distinctive cls^– phenotype in these alb^– hosts.

Here we demonstrated an unexpectedly broad, but still selective, requirement for the cls gene during neural crest development. We confirmed previous findings (Kelsh et al., 1996) that cls mutations affect pigment cell fates prior to migration, and showed that they severely reduce sensory, sympathetic and enteric neurons and enteric and putative cranial glia. In contrast, craniofacial skeletal derivatives and median fin mesenchyme are initially unaffected in cls^– embryos. In view of the significant contribution of neural crest to the cranial ganglia in mammalian and avian embryos (Le Douarin, 1982), we expected significant reduction in the neuronal populations of at least some cranial ganglia in cls^– embryos. We suggest that the lack of defects at 24 hpf reflects a predominantly ectodermal placode derivation for at least these early neurons, although we cannot rule out later contributions from undifferentiated neural crest cells. However, gross abnormalities of the posterior lateral line ganglion are not visible even in 5 dpf cls^– mutant larvae. Previous studies have shown that some cranial ganglion neurons are crest-derived (Schilling and Kimmel, 1994; Dutton and R. N. K., unpublished), but the relative contributions of ectodermal placode and neural crest to these ganglia is still unknown. Clearly, much work is required to clarify the origins of zebrafish cranial neurons, before this surprising result can be fully explained.

We propose that the cls^– phenotype reflects an early segregation between ectomesenchymal and non-ectomesenchymal crest precursors in which cls gene function is required only for correct development of non-ectomesenchymal precursors (Fig. 11). Non-ectomesenchymal precursors generate neuronal, glial and/or pigment cells, but not ectomesenchymal cells. The cls^– phenotype provides genetic evidence consistent with the suggestion that ectomesenchymal fates segregate early from other crest fates (Weston, 1991; Le Douarin et al., 1994). One prediction of our hypothesis is that trunk crest-derived ectomesenchymal fates segregate early from non-ectomesenchymal fates, probably before neural crest migration. Zebrafish crest lineage studies have not yet examined segregation of the ectomesenchymal lineage (Schilling and Kimmel, 1994; Smith et al., 1994). Similarly, we are not aware of any mutant with defects in both crest-derived ectomesenchymal fates. chinless, the best-studied craniofacial mutation, is not reported to affect median fin mesenchyme (Schilling et al., 1996b). However our model is supported by recent studies showing that altering activity of Wnt signalling in zebrafish non-ectomesenchymal cranial crest cells changes their fate, but altering Wnt signalling in cartilage precursors does not affect their fate (Dorsky et al., 1998).

Although we propose early segregation of ectomesenchymal and non-ectomesenchymal precursors, we cannot rule out two other possibilities. First, that single fate precursors are directly specified from unrestricted neural crest cells. In this model, the cls gene product would be required in all precursors that generate neuronal, glial or pigment cell derivatives. Second, that there is no segregation of ectomesenchymal and non-ectomesenchymal precursors. Instead, crest cells might be specified with a wide variety of partially restricted but overlapping potentials, as proposed by Le Douarin and colleagues in avian crest (Le Douarin, 1986; Baroffio et al., 1988, 1990; Le Douarin, 1990; Baroffio et al., 1991; Le Douarin et al., 1991). In this model, the cls gene product would be necessary for generation of any non-ectomesenchymal fate from all these various progenitors.

The cls mutant phenotype is reminiscent of mouse pigmentation mutants with defects in both melanocytes and enteric nervous system (Lane and Liu, 1984; Baynash et al., 1994; Hosoda et al., 1994; Southard-Smith et al., 1998). The milder defects in s and ls mutants suggest that these genes function only in melanoblasts and enteric nervous system precursors. In contrast, the more extreme phenotype of Dom homozygotes implies an earlier function in a wider range of derivatives (Southard-Smith et al., 1998). cls^– embryos show wide-ranging neural crest defects reminiscent of Dom homozygous mutants and so cls gene function is also likely to act early in crest development. Whilst cls^– mutants lack enteric neurons along the entire extent of the gut, s and ls homozygous mutants show aganglionosis only in the posterior gut. This aspect of the phenotype is most reminiscent of c-ret mutants, which show aganglionosis throughout the gut, but which have no pigment defects (Schuchardt et al., 1994); and Dom homozygous mutants. However, a major difference between cls and Dom homozygous mutant phenotypes is the severe otic vesicle defect in cls^– embryos; inner ear defects have not been described for Dom homozygotes. It will be interesting to know whether cls encodes a zebrafish homologue of any of these mice loci or whether it defines a new gene involved in neural crest development. We are currently testing these ideas. Mice with mutations in s, ls and Dom are all models for human Hirschsprung’s disease, characterised by aganglionosis of the colon, and Waardenburg-Shah syndrome, in which colonic aganglionosis is associated with pigmentary defects (Passarge, 1973; Badner and Chakravarti, 1990; Dow et al., 1994; Attié et al., 1995; Edery et al., 1996; Hofstra et al., 1996; Southard-Smith et al., 1998). We propose that cls mutations are a zebrafish model for these diseases. Given the ease of in vivo lineage analysis (Raible and Eisen, 1994), we believe that this new model will contribute significantly to our understanding of Hirschsprung’s disease and Waardenburg-Shah syndrome.

Importantly, cls gene function is not required for formation of neural crest per se, since premigratory crest markers are expressed normally in cls^– embryos. A priori, cls might function in specification, proliferation, survival, or differentiation of non-ectomesenchymal neural crest precursors. The observations described here are consistent with possible roles in specification, proliferation and differentiation, and we are accumulating evidence for a possible role in survival (K. Dutton and R. N. K., unpublished observation). Further studies should clarify which of these processes is primarily affected in cls^– mutants.

The earliest defects that we see in cls^– embryos are present at 21 hpf or shortly thereafter. Thus, dct-expressing melanoblasts are reduced in number as early as 21 hpf and lateral path neural crest cell migration is defective at 24 hpf. The eventual fate of these pigment cell precursors is unclear. Some differentiate into the residual population of pigment cells present in cls^– mutants, but others presumably die or change fate. Future studies of crest cells, including clonal analysis and investigations of cell death, should distinguish these possibilities.

cls^– embryos show a migration defect for crest-derived pigment cell precursors and possibly DRG neurons (Kelsh et al., 1996; this work). However, a primary defect in migration cannot fully explain the cls^– phenotype since such a role would predict that crest derivatives differentiating near their origin, for instance melanophores in the dorsal pigment stripe, would develop normally. However, this is the location of the few melanophores seen in cls^– mutants, yet even here none differentiate normally. Conversely, our c-ret studies provide evidence that migration of putative enteric nervous system precursors may be normal until at least 24 hpf, although a contribution of migratory difficulties to the later enteric neuron phenotype cannot be ruled out. Thus, we suggest that migratory defects are a secondary effect of the cls mutation.

Finally, whatever its role, we have demonstrated that the cls gene product acts cell-autonomously, at least in pigment cell precursors. The cell-autonomy of cls gene action in other crest fates remains to be tested. Although we expect cls to function cell-autonomously in all non-ectomesenchymal crest derivatives, autonomy in pigment cells and non-autonomy in enteric neurons is proposed for the s gene in mice (Mayer, 1965; Kapur et al., 1995) and may be the case here as well.

We thank R. Bremiller and M. McDowel for expert assistance with sectioning, R. Banally and S. Carneiro for technical assistance with specific experiments, C. Moens, Y.-L. Jan and D. Raible for RNA probes, and staff of the University of Oregon zebrafish facility for fish husbandry. We gratefully acknowledge P. Henion, C. Kimmel and D. Raible for their detailed and helpful criticism of previous drafts. Supported by NIH grant HD22486 and an HFSP postdoctoral fellowship to R. N. K.