ABSTRACT
Developmental mechanisms underlying traits expressed in larval and adult vertebrates remain largely unknown. Pigment patterns of fishes provide an opportunity to identify genes and cell behaviors required for postembryonic morphogenesis and differentiation. In the zebrafish, Danio rerio, pigment patterns reflect the spatial arrangements of three classes of neural crest-derived pigment cells: black melanocytes, yellow xanthophores and silver iridophores. We show that the D. rerio pigment pattern mutant panther ablates xanthophores in embryos and adults and has defects in the development of the adult pattern of melanocyte stripes. We find that panther corresponds to an orthologue of the c-fms gene, which encodes a type III receptor tyrosine kinase and is the closest known homologue of the previously identified pigment pattern gene, kit. In mouse, fms is essential for the development of macrophage and osteoclast lineages and has not been implicated in neural crest or pigment cell development. In contrast, our analyses demonstrate that fms is expressed and required by D. rerio xanthophore precursors and that fms promotes the normal patterning of melanocyte death and migration during adult stripe formation. Finally, we show that fms is required for the appearance of a late developing, kit-independent subpopulation of adult melanocytes. These findings reveal an unexpected role for fms in pigment pattern development and demonstrate that parallel neural crest-derived pigment cell populations depend on the activities of two essentially paralogous genes, kit and fms.
INTRODUCTION
Recent years have seen dramatic advances in our understanding of the developmental genetic bases for the patterning of embryonic axes, tissues and organ rudiments. In contrast, we know relatively little about mechanisms underlying the expression of traits during later stages of development and in adults (Tata, 1993). Nevertheless, identifying the genes and cell behaviors underlying trait expression is an essential step in understanding the origins of naturally occurring trait variation and the evolution of form (Atchley and Hall, 1991; Phillips, 1999). One ecologically important trait that is particularly amenable to analysis is the externally visible pigment pattern of fishes in the genus Danio, which includes the zebrafish D. rerio.
Pigment cells in Danio and other vertebrates are derived from neural crest cells that arise along the dorsal neural tube then disperse along stereotypical pathways throughout the embryo (Hörstadius, 1950; Erickson and Perris, 1993; Hall, 1999; Groves and Bronner-Fraser, 1999). In addition to pigment patterns, neural crest cells contribute to a host of other tissues and organ systems in vertebrates, including bones of the craniofacial skeleton, teeth, neurons and glia of the peripheral nervous system, endocardial cushion cells and endocrine glands. Indeed, Gans and Northcutt (1983) have argued that much of vertebrate morphology and its evolution can be understood in terms of the patterning of neural crest cells and their derivatives, and how these patterning mechanisms have changed phylogenetically. Elucidating the mechanisms by which the final form of larval and adult pigment patterns arise may thus shed light on more general mechanisms of trait variation and evolution in vertebrates.
In danios and many other ectothermic vertebrates, pigment patterns result from the spatial arrangements and coloration of three major classes of neural crest-derived pigment cells: black melanocytes (or melanophores), yellow xanthophores and silver iridophores (Bagnara, 1998; Reedy et al., 1998). In D. rerio, these different classes of pigment cell combine to generate different pigment patterns during different phases of the life cycle (Kirschbaum, 1975; Johnson et al., 1995b). In larvae, a relatively simple pattern is evident by hatching (approx. 2.5 days; Fig. 1A). This ‘early larval’ pigment pattern consists of several stripes of melanocytes and iridophores, as well as xanthophores that are widely distributed over the flank, giving an overall yellow cast to the body. This pattern persists until approx. 14 days, at which time a metamorphosis begins that ultimately results in the formation of the striped pigment pattern of the adult (Fig. 1C). Between 14 and 21 days, melanocytes increase in number and become visible dispersed throughout the skin in regions not previously occupied by these cells. Subsequently, between 21 and 28 days, melanocyte numbers increase more sharply and an adult stripe pattern begins to emerge. This pattern comprises two melanocyte stripes by 28 days, but additional stripes form as the fish grow. Dark stripes consist of melanocytes and iridophores, whereas light ‘interstripe’ regions consist of xanthophores and iridophores (Fig. 2A). Finally, in addition to stripes that are found deep within the dermis adjacent to the myotomes (Hawkes, 1974), more superficial melanocytes and xanthophores cover the dorsal scales, and together give a dark cast to the dorsum of the fish.
The mechanisms of adult stripe development remain largely unknown in D. rerio. Nevertheless, several mutants have started to provide insights into the genes and cell populations that are involved in pigment pattern metamorphosis in this species. A previous analysis (Johnson et al., 1995b) identified roles for three genes that are required for the development of adult stripes: sparse, rose and leopard. sparse mutant larvae have fewer melanocytes than wild type at 3 days. These cells then die, and the fish completely lack melanocytes until approx. 21 days, when a new population differentiates and contributes to an adult pigment pattern with one-half the wild-type complement of stripe melanocytes. In contrast, rose and leopard mutants each exhibit normal pigment patterns through 21 days, but melanocyte numbers increase less rapidly than wild type between 21 and 28 days, again resulting in one-half the wild-type complement of stripe melanocytes. These findings suggested that sparse on one hand, and rose and leopard on the other, identify genes required for the development of distinct classes of melanocytes in the adult pigment pattern: an early developing adult population dependent on sparse; and a later developing adult population dependent on rose and leopard. In support of this model, virtually all body stripe melanocytes are ablated in fish homozygous for both sparse and rose mutations, or sparse and leopard mutations, whereas no additional melanocyte deficit found in fish homozygous for rose and leopard mutations, as compared to either single mutant alone. Thus, the pigment pattern of adult D. rerio depends on contributions from temporally and genetically distinct populations of melanocytes, dependent on the activities of either sparse (early appearing melanocytes), or rose and leopard (later appearing melanocytes). More recently, we showed that sparse is a zebrafish orthologue of kit, which encodes a type III receptor tyrosine kinase that has long been studied for its role in pigment pattern development in amniotes (Parichy et al., 1999). What additional genes contribute to the morphogenesis and differentiation of these melanocyte populations, and other pigment cell classes, remain largely unknown.
Here, we identify another gene required for stripe development in D. rerio. We show that the pigment pattern mutant panther corresponds to an orthologue of c-fms, which is the closest known homologue of the previously identified pigment pattern gene kit. We find that fms is required for the morphogenesis of embryonic xanthophore precursors and the development of adult xanthophores. We also demonstrate roles for fms during adult stripe development. fms promotes the normal patterning of cell death and migration among early larval melanocytes and kit-dependent (rose-independent) stripe melanocytes. fms is also essential for the differentiation of late appearing, rose-dependent (kit-independent) stripe melanocytes. These findings indicate that two essentially paralogous genes (fms, kit) promote the morphogenesis and differentiation of parallel pigment cell populations in D. rerio.
MATERIALS AND METHODS
Fish rearing, mutant alleles, genetic mapping, cloning and molecular analysis
Rearing of D. rerio followed Westerfield (1993). All mutant panther alleles analyzed are recessive and exhibit pigment patterns that are indistinguishable from one another. pantherj4e1 was induced by the point mutagen ethyl nitrosourea (ENU) in the AB genetic background (Johnson and Zon, 1999). pantherj4blue was identified as an allele of panther by complementation testing of ‘blue’ variety D. rerio obtained from the pet trade. Finally, pantherj4e2 was induced by ENU in the inbred C32 genetic background.
To map panther, we crossed pantherj4e1 (maintained in the C32 background) to the inbred wild-type strain SJD to produce highly polymorphic progeny segregating alleles of C32/AB and SJD strains. To rapidly focus our mapping efforts, we used half-tetrad centromere linkage analysis (Johnson et al., 1995a) to exclude chromosomes from consideration. We produced gynogenetic offspring from polymorphic pantherj4e1/+ females, reared these to maturity, then scored them for is phenotype and isolated genomic DNA by standard procedures. We then assessed linkage to centromere markers for each of the 25 D. rerio chromosomes: a complete absence of SJD alleles among 8 gynogenetic panther adults suggested the presence of panther on Linkage Group (LG) 14. Subsequently, we mapped panther more precisely with backcross families segregating pantherj4e1. We cloned a partial D. rerio fms cDNA using degenerate reverse transcriptase PCR from a kidney cDNA library, then cloned a full-length cDNA by rapid amplification of cDNA ends (primer sequences available on request). To map pantherj4e1 relative to fms (GenBank accession #AF240639; see Results), we designed primers (forward: AGGTGAGCGAAGGAGACGATG; reverse: GTATTGGCCGATCTCCTCGAAG) to amplify a polymorphism within predicted intron 6 of the fms gene. Finally, to test fms as a candidate for panther, we designed primers to amplify overlapping fragments of fms cDNAs derived by RT-PCR, and sequenced the PCR products directly to avoid cloning artefacts. We used BigDye dye terminator chemistry (Perkin Elmer) on an ABI-377 automated sequencer and sequenced forward and reverse directions for 2-5 individuals of each genetic background. For pantherj4e1 and pantherj4e2, we compared sequences to those for ancestral chromosomes in AB and C32 backgrounds, respectively. Phylogenetic analyses were performed with PAUP 4.0.0d55 (Swofford, 1999) after aligning sequences with the Pileup program of the GCG software package, and used parsimony and heuristic searching with 1000 bootstrap replications. Nucleotide and protein sequences also were compared against existing databases by BLAST, and predicted protein domains were assessed by comparison with the Pfam database (http://pfam.wustl.edu).
In situ hybridization
Methods for riboprobe preparation and in situ hybridization have been described (Jowett and Yan, 1996), and employed digoxigenin- and fluorescein-labeled riboprobes, alkaline phosphatase-conjugated anti-digoxigenin or anti-fluorescein Fab fragments, and NBT/BCIP (Roche, Indianapolis, IN) or FastRed (Sigma, St. Louis, MO) substrates. Hybridizations were carried out for 12-24 hours at 68°C or 70°C, followed by stringency washes at the same temperatures. Larvae >5 days were cut transversely after fixation to facilitate penetration of reagents. For kit (Parichy et al., 1999) and fms, we used full-length cDNAs (approx. 3 kb) as templates for riboprobe generation. We cloned a 540 bp fragment of D. rerio xdh for riboprobe synthesis (this gene is also identified as D. rerio EST fc18g02; accession numbers: AI657925, AI641077), and used D. rerio ESTs to generate riboprobes for gch (fa05h05; AA494989, AA494990) and cathepsin-K (fa95f03; AI331947, AI330487). cDNAs for mitf (Lister et al., 1999) and dct (Kelsh and Eisen, 1999) were provided by D. Raible and R. Kelsh, respectively.
Analyses of melanocyte behaviors and distributions
To identify melanocyte behaviors that underlie the adult panther mutant phenotype, we followed individual melanocytes in image series of wild-type and panther mutant larvae (n=4 per genotype). Larvae were imaged at 32× using a Spot II digital imaging system (set to capture a high-speed monochrome image) interfaced to an Olympus SZX12 zoom stereomicroscope and Macintosh computer. Images were taken every 24 hours for 12 days, beginning on day 18 of development. Larvae were anesthetized long enough to record an image, then were revived and maintained individually (14 hours L:10 hours D light cycle, 2 feedings per day). All larvae developed normal wild-type or panther mutant adult pigment patterns.
We followed individual melanocytes by comparing melanocyte positions at each of 4-5 body regions along the anteroposterior axis. To facilitate tracking cells and analyzing different morphogenetic behaviors, we labeled individual melanocytes in each body region on day 18 (cells A-M), day 24 (cells N-R), and day 30 (cells S-Z) using Adobe Photoshop (see below). We then followed these cells forwards and backwards in the image series. Identification of different populations of cells at different times allowed us to test whether cells at the beginning of the series died later in the series, and whether cells at the end of the series were present at the beginning, or had differentiated during the time of imaging. We were able to continuously re-identify >90% of labeled melanocytes across all time points, and death of cells could be inferred by the unambiguous disappearance of a cell from one image to the next (see below). To test for movements of melanocytes, we first determined the relative dorsal-to-ventral position of each cell on the flank; a position of 0.0<0.5 is dorsal, 0.5<1.0 is ventral, and 0.5 represents the dorsoventral midpoint of the flank and the approximate position of the developing interstripe region. To assess whether cells moved dorsally or ventrally during the image series, we subtracted the dorsoventral position of each melanocyte when first identified from its position when last visible. This provided an estimate of the net change in dorsoventral position for each cell, such that values >0.0 represent a net ventral movement, and values <0.0 represent a net dorsal movement. These estimates do not, however, reflect anteroposterior changes or total distances migrated. We haphazardly picked cells to label and follow, but attempted to choose melanocytes distributed widely over the flank; dorsoventral distributions of melanocytes chosen for analysis did not differ between wild-type and panther mutants (t878=0.69; P=0.5). In total, we were able to follow 880 melanocytes in image series of wild-type and panther mutant larvae.
To assess changes in melanocyte numbers over time, we counted the total numbers of these cells within each of four myotomal segments distributed evenly across the anteroposterior axis of the same larvae used for imaging. Additionally, to assess differences among genotypes, we counted melanocytes in juvenile fish that had already developed an adult pigment pattern. For these latter counts, we limited our analysis to a region bordered anteriorly and posteriorly by the base of the anal fin; we further limited this analysis to melanocytes that were not present on scales, but were present deeper in the dermis where melanocytes normally are found that contribute to adult stripes. To control for differences in size and area among individuals, we normalized melanocyte counts according to the length of the region examined, resulting in counts with units of melanocytes/mm. Other methods of normalizing melanocyte numbers (e.g., relative to total area or height of the flank) yielded identical results.
Statistical methods
All statistical analyses were performed with JMP Statistical Software (SAS Institute, 1998). For image series, we analysed data as univariate models (considering only effects due to genotype), as well as full, multifactorial models that controlled for effects due to interindividual variation independent of genotype. Both simple and full analyses yielded equivalent results that are presented below. Complete statistical analyses for full models and results of testing all effects are available from DMP on request. We analyzed melanocyte death using multiple logistic regression (Shanubhogue and Gore, 1987): a full model tested for effects due to genotype, individual (nested within genotype), deviation of melanocytes from the midpoint of the flank, final height of the flank, and an interaction between genotype and deviation from the midpoint (testing whether the relationship between probability of death and position depended on genotype). We analyzed melanocyte movements using multiple linear regression (Sokal and Rohlf, 1981): a full model tested for effects due to genotype, individual (nested within genotype), starting melanocyte position, total growth of the flank (controlling for passive movements due to isometric growth), segment identity (controlling for effects due to anteroposterior position), and an interaction between genotype and starting position (testing whether the relationship between starting position and movement depended on genotype). We analyzed melanocyte differentiation using multiple logistic regression: a full model consisted of effects due to genotype and starting height of the flank. All models treated individuals and interactions involving individuals as random effects (Sokal and Rohlf, 1981).
RESULTS
panther corresponds to a D. rerio orthologue of fms
panther mutant adult D. rerio have poorly formed stripes anteriorly and lack stripes posteriorly (Fig. 1D). To test roles for panther in promoting pigment pattern development, we first identified this locus at the molecular level. We used half-tetrad centromere linkage analysis (Johnson et al., 1995a) to map pantherj4e1 to LG 14, and subsequently mapped pantherj4e1 in a 161 individual mapping panel between microsatellites z5435 and z1536 (Fig. 3A). Interestingly, we had previously mapped to this same interval an orthologue of the c-fms gene (fms; also known as colony stimulating factor 1 receptor, Csf1r). fms encodes a type III receptor tyrosine kinase and is the closest known homologue of kit in D. rerio (Fig. 3B); fms and kit are believed to have arisen by duplication of a common ancestral gene during the evolution of vertebrates (Yarden and Ullrich, 1988; Rousset et al., 1995). fms in amniotes has roles in the development of macrophages and bone-resorbing osteoclasts (Motoyoshi, 1998; Flanagan and Lader, 1998), and has not been implicated in neural crest or pigment pattern development. Nevertheless, the structural similarity of fms to kit, and the linkage between fms and pantherj4e1, led us to test fms as a candidate for panther.
To test the correspondence of fms and panther, we identified a length polymorphism within predicted intron 6 of fms [repeating GAAT and TGG motifs; wild-type (SJD) allele: 1000 bp; pantherj4e1 (AB) allele: 820 bp] and used this to assess segregation relative to the panther mutant phenotype. We identified 160 non-recombinant individuals, placing the pantherj4e1 mutant lesion approx. 0.6 cM from fms intron 6. Given the relatively large size of the fms gene in humans (32 kb coding region; Roberts et al., 1988), we hypothesized that pantherj4e1 might correspond to a lesion located within fms but relatively distant from intron 6. To test this possibility, we sequenced fms cDNAs from wild-type and pantherj4e1 mutant backgrounds. We identified a single, non-conservative (val-met) substitution in predicted exon 12, within the first kinase domain at a site that is otherwise invariant in all receptor tyrosine kinases used for our phylogenetic analysis (Fig. 3B,C). This lesion segregated with the panther mutant phenotype in all 161 individuals of the mapping cross, supporting the allelism of panther and fms.
To further test the correspondence of panther and fms, we sequenced fms cDNAs from additional panther backgrounds. In pantherj4blue, we identified a 5 nucleotide deletion near the middle of the second kinase domain, leading to 32 novel amino acids and a premature stop codon. In pantherj4e2, induced in a defined genetic background, we identified a non-conservative asp-arg substitution within the third immunoglobulin loop of the extracellular domain (Yarden and Ullrich, 1988; Fig. 3C).
These independent fms lesions in three panther mutant alleles (two of defined genetic background) confirm the correspondence of fms and panther. Thus, we here designate the mutant and its corresponding gene, fms. Because all three alleles are recessive and have indistinguishable pigment patterns, all are likely to be loss of function; this inference is supported by studies associating lesions in fms or kit kinase domains with reductions in kinase activity (Reith et al., 1990, 1991, 1993), and our observation that fms mRNA levels are dramatically reduced in fms mutant embryos and larvae (data not shown). These data show that Fms, a Kit-like type III receptor tyrosine kinase, is required for pigment pattern development in D. rerio.
Neural crest-derived cells and osteoclasts express fms
In light of the pigment pattern defect in fms mutants, we asked whether this gene is expressed by neural crest cells or their derivatives. In situ hybridization reveals fms+ cells along the mid-trunk dorsal neural tube by 18 hours (not shown), and subsequently in neural crest migratory pathways (Fig. 4A-D). To verify that these cells are neural crest-derived, we tested for coexpression of fms and four markers of neural crest–pigment cell lineages: the heptahelical receptor endothelin receptor B (ednrb), which is expressed by precursors of all three classes of pigment cells (D. M. P., R. N. Kelsh, S. L. J., unpublished data; Nataf et al., 1996; Shin et al., 1999); the transcription factor microphthalmia (mitf), which is expressed by melanocyte precursors (melanoblasts) but is downregulated in late differentiating melanocytes (Opdecamp et al., 1997; Lister et al., 1999); and the melanin synthesis enzyme dopachrome tautomerase (dct) and kit, which are expressed by melanoblasts and melanocytes (Wehrle-Haller and Weston, 1995; MacKenzie et al., 1997; Parichy et al., 1999; Kelsh and Eisen, 1999). These analyses reveal fms expression by ednrb+ cells and some mitf+ cells (Fig. 4E-H). In contrast, we could not detect fms expression by the majority of kit+ cells, or cells either expressing dct or containing melanin (Fig. 4I,J). These data demonstrate that fms is expressed by a subset of neural crest-derived cells, but not late stage, differentiating embryonic melanocytes.
To test whether previously described roles for fms in hematopoiesis and osteoclastogenesis are conserved between amniotes and teleosts, we examined fms expression in non-neural-crest-derived lineages. In embryos, fms+ cells are observed in the position of macrophage progenitors, and these cells coexpress the macrophage lineage marker pu.1 (data not shown). Osteoclasts can be identified by the expression of the protease cathepsin-K (Roodman, 1996). In the regenerating fin, cathepsin-K+ presumptive osteoclasts coexpress fms, and fms mutant regenerates have fewer of these cells than wild type (mean±s.d. cells per fin in wild type: 44.1±14.98, n=10; fms mutants: 1.7±2.73, n=6; t14=6.78, P<0.0001; Fig. 4K-M).
Thus, fms is expressed in non-melanogenic neural crest-derived cells and has phylogenetically conserved expression in macrophage and osteoclast lineages.
fms is essential for development of embryonic xanthophores
fms mutant larvae lack the yellow color of wild-type larvae at 3 days (Fig. 1A,B), indicating a defect in the xanthophore lineage. This could reflect a failure of xanthophore precursors (xanthoblasts) to differentiate. Alternatively, xanthoblasts could require fms for their dispersal from the neural crest, much as melanoblasts or melanocytes require kit for their dispersal (Bernex et al., 1996; Parichy et al., 1999). To test these hypotheses, we isolated two markers for the xanthophore lineage: xanthine dehydrogenase (xdh) and GTP-cyclohydrolase I (gch). Xdh is required for synthesizing the xanthophore pteridine pigment xanthopterin, and thus should be upregulated in xanthoblasts (Epperlein and Löfberg, 1990; Reaume et al., 1991). Gch converts guanosine triphosphate to tetrahydrobiopterin (BH4), a precursor to pteridine pigments. BH4 also is a co-factor for phenylalanine hydroxylase, which converts phenylalanine to tyrosine (a precursor to melanin), and regulates the activity of tyrosinase, which converts tyrosine to the melanin intermediate, dopa (O’Donnell et al., 1989; Wood et al., 1995; Nagatsu and Ichinose, 1999). Thus, gch should be expressed by xanthoblasts, and possibly melanoblasts. In situ hybridization reveals xdh+ cells in neural crest migratory pathways and, at later stages, in regions occupied by xanthophores (Fig. 5A,B), suggesting these cells are xanthoblasts. Additionally, some cells close to the neural crest express both xdh and mitf, and may represent uncommitted precursors of melanocytes or xanthophores (Fig. 5E). gch expression is similar to xdh, and many cells coexpress both gch and xdh (not shown), though gch is also detectable in lightly melanized melanocytes (Fig. 5C,F). Thus, gch is likely to be expressed both by xanthoblasts and melanoblasts. Many non-melanized xdh+ and gch+ cells coexpress fms (Fig. 5D,F), consistent with fms expression by the xanthophore lineage. To test whether fms is required for the migration of xanthoblasts, we examined the distribution of xdh+ and gch+ cells in fms mutant embryos. In contrast to wild type, fms mutant xdh+ cells were confined to the vicinity of the neural crest, and few unmelanized gch+ cells were observed during and after neural crest migration (Fig. 5G-I). These data indicate that fms is expressed by the xanthophore lineage, and promotes the dispersal of these cells from the neural crest. Thus, the xanthophore requirement for fms parallels the melanocyte requirement for kit.
fms promotes patterning of melanocyte death and migration during adult stripe development
fms mutant adults lack normal stripes (Fig. 1D). To determine how fms promotes stripe development, we compared melanocyte numbers and behaviors (n=880 melanocytes total) in image series of wild-type and fms mutant larvae during pigment pattern metamorphosis. Because we imaged larvae between 18 and 30 days, these analyses include early larval melanocytes that may have persisted until metamorphosis, adult stripe melanocytes comprising the early (kit-dependent) and late (rose-dependent) populations, as well as melanocytes that will contribute to adult scales. Xanthophores and iridophores were not visible in these images. At the beginning of the series, wild-type larvae exhibited melanocytes still in the position of early larval stripes, as well as some melanocytes dispersed over the flank (presumptive kit-dependent melanocytes). By the end of the image series, an adult pigment pattern had been established, in wild type consisting of two melanocyte stripes (a dorsal stripe at dorsoventral position approx. 0.35, and a ventral stripe at position approx. 0.65; see Materials and Methods), a single, light interstripe region (position approx. 0.5; i.e., the approximate dorsoventral midpoint of the flank), and scale-associated melanocytes in the dorsal region of the flank (position 0-0.5).
Our analyses reveal that melanocyte death in the prospective interstripe region contributes to stripe development in wild-type larvae, whereas a defect in the patterning of melanocyte death is associated with the failure of normal stripe development in fms mutant larvae. Previous studies of stripe development in the fin (Goodrich and Nichols, 1931) and early larval melanocytes on the body (Milos et al., 1983) suggested that some melanocytes die if present in prospective interstripe regions (i.e., regions occupied by xanthophores; see Discussion). To test whether melanocyte death has a role in adult body stripe development, we determined the probability that melanocytes followed from day 18 of development (day 1 of imaging) died by day 30 (end of imaging). These melanocytes principally represent cells persisting from the early larval pigment pattern, as well as melanocytes comprising the early appearing kit-dependent population of adult stripe melanocytes (and exclude late appearing rose-dependent melanocytes that differentiate after approx. 21 days). We could see melanocyte death as an unambiguous disappearance of a melanocyte from one image to the next (Fig. 6). In wild-type larvae, 17±6.8% of these melanocytes (n=234) died by day 30. To assess whether melanocytes were more likely to die if located in the prospective interstripe region than stripe regions, we tested for a dependence of melanocyte death on dorsoventral distance from the midpoint of the flank. This analysis revealed that wild-type melanocytes were more likely to die if close to the midpoint (i.e., in the prospective interstripe region). This effect can be visualized as differences in the proportions of melanocytes dying when distance from the midpoint of the flank is divided into discrete regions (Fig. 7A). To assess whether fms influences melanocyte death during stripe development, we determined the fate of melanocytes followed from day 18 in fms mutant larvae. In fms mutants, 26±16.2% of these melanocytes died by day 30 (n=207), a significantly greater percentage than in wild type (χ2=5.88, d.f.=1, P<0.05). Also in contrast to wild type, the likelihood of fms mutant melanocytes dying did not depend significantly on distance from the midpoint of the flank (Fig. 7A). Taken together, these data indicate that melanocytes are more likely to die in the prospective interstripe region in wild-type larvae, and a perturbation of Fms signaling results in both an increased total likelihood of melanocyte death and a defect in the patterning of this death.
Image analyses also show that directed melanocyte migration contributes to stripe development in wild-type larvae, and a failure of directed migration is associated with the failure of stripe development in fms mutants. In other species, stripes form because initially dispersed melanocytes migrate into prospective stripe regions (Epperlein and Lofberg, 1990; Parichy, 1996a,b). To test whether melanocyte migration is involved in stripe formation in D. rerio, and whether migration is perturbed in fms mutants, we calculated the relative dorsoventral position of each melanocyte when first identified, and when last observed. For each melanocyte followed more than one day, we then estimated the net change in dorsoventral position by subtracting the starting position from the ending position. These analyses thus include early larval melanocytes, as well as early and late appearing adult melanocytes (analyses limited to melanocytes present at the beginning of imaging yielded qualitatively similar results). In wild-type larvae, dorsal melanocytes tended to move ventrally, whereas ventral melanocytes tended to move dorsally (Fig. 7B). This reflects migration of melanocytes in dorsal and ventral regions of the flank towards the middle of the flank, to form dorsal and ventral stripes, respectively. Although some melanocytes initially in the prospective interstripe region migrated in the reverse direction (e.g., dorsal cells moving further dorsally; Fig. 6), these cells contributed little to the overall pattern of movement (presumably because many melanocytes in the prospective interstripe region died, rather than migrated into stripes). Examination of melanocyte movements in fms mutant larvae (n=390 melanocytes) revealed that absolute changes in melanocyte position (i.e., independent of direction) did not differ from wild type (least squares means ± s.e. in wild type, fms mutants, respectively=0.044±0.004, 0.036±0.005; F1,835=0.06, P=0.8), indicating that overall melanocyte motility is not impaired in fms mutants. Nevertheless, the relationship between direction of movement and position was significantly weaker in fms mutant larvae as compared to wild type (Fig. 7B). This difference can be understood as a reduction in organized, directional melanocyte movements when Fms signaling is perturbed.
These data identify roles for fms in patterning melanocyte death and migration during stripe development. This could indicate a direct effect on melanocyte behavior, if these cells express fms. Or, effects on melanocyte patterning could be indirect, and mediated via interactions between melanocytes and other fms-expressing cell types. To distinguish between these possibilities, we examined fms expression during adult stripe development. In wild-type larvae, fms is expressed by unmelanized cells between the epidermis and myotome (Fig. 8A,C). Although some melanocytes express kit during these stages, differentiated melanocytes do not express fms (Fig. 8B,D,E). These data suggest that fms does not directly promote melanocyte death or migration (although we cannot formally exclude the possibility that differentiated melanocytes express functionally significant levels of fms mRNA at a level below the threshold of detection by in situ hybridization). Since xanthophores influence melanophore morphogenesis in other species (Epperlein and Löfberg, 1990; Parichy, 1996b) and xanthophores depend on fms for their morphogenesis in D. rerio embryos (above), we hypothesized that effects of fms on melanocyte death and migration in D. rerio might be mediated by interactions between melanocytes and fms-dependent xanthophores. To test whether xanthoblasts are present during adult stripe development, we compared the distribution of gch+ cells and differentiated xanthophores between wild-type and fms mutant larvae and adults (xdh staining resulted in excessive background and was not useful during these stages). In wild type, numerous, non-melanized gch+ cells are present between the epidermis and the myotome (Fig. 8F), and xanthophores occur outside of prospective stripe regions during metamorphosis, and in interstripe regions in adults (Fig. 2A). In contrast, fms mutants lack unmelanized gch+ cells (Fig. 8F,′), exhibit greatly reduced numbers of xanthophores on the body and in the fins of metamorphosing larvae, and completely lack xanthophores in adults (Fig. 2B). These findings are consistent with a model in which interactions between melanocytes and fms-dependent xanthophores contribute to patterning melanocyte death and migration during adult stripe development.
Late appearing, kit-independent stripe melanocytes require fms for development
In addition to effects on melanocyte death and migration, the pigment pattern defect in fms mutant adults could reflect a difference in melanocyte differentiation. To test this possibility, we first counted melanocytes in imaged larvae between 18 and 30 days. Since many dorsal melanocytes localize on scales rather than in stripes, we limited our analysis to the ventral half of larvae, which lacks prospective scale melanocytes. Fig. 7C shows that ventral melanocytes nearly doubled in number in wild-type larvae, but did not increase in fms mutants. To determine whether this difference reflected, in part, a difference in melanocyte differentiation, we assessed the probability that melanocytes identified at 30 days were derived from melanocytes that differentiated between 19 and 30 days (i.e., were not present at the beginning of imaging at 18 days). Newly differentiated melanocytes were readily distinguishable as small, lightly melanized cells in regions not previously occupied by melanocytes. This analysis revealed that over the entire flank in wild-type larvae, 49±6.5% of melanocytes differentiated de novo, whereas in fms mutant larvae only 23±12.2% of melanocytes differentiated de novo, resulting in a significant difference between genotypes (χ2=7.8, d.f.=1, P<0.01). Wild-type melanocytes that differentiated during this time were limited principally to dorsal regions (prospective scale and dorsal stripe melanocytes) and ventral regions (prospective ventral stripe melanocytes), and few were found in the middle of the flank (prospective interstripe region; Fig. 7D). In contrast, newly differentiated melanocytes in fms mutants were limited principally to dorsal regions (presumably scale melanocytes) and very few differentiated in the region of the ventral stripe. Thus, fms is required for the differentiation of normal numbers of melanocytes during adult stripe development. Consistent with this inference, 6 week fms mutant juveniles have only approx. 70% as many stripe (i.e., non-scale) melanocytes as wild-type juveniles in dorsal and ventral regions of the flank (Fig. 9E).
Adult melanocyte stripes in D. rerio depend on two populations of melanocytes: a kit-dependent population arises dispersed over the flank beginning at approx. 14 days, and a rose-dependent population arises already in stripes beginning at approx. 21 days (Johnson et al., 1995b). Given the defect in melanocyte differentiation in fms mutants, we asked whether fms is required by kit-dependent melanocytes, rose-dependent melanocytes, or both. The increasingly severe deficit in fms mutant melanocytes after 22 days (Fig. 7C), and the failure of these cells to differentiate in stripes (Fig. 7D), are consistent with a requirement for fms in promoting the appearance of the later, rose-dependent melanocytes. To further test this model, we constructed fish doubly mutant with kit or rose. If fms is required by late appearing rose-dependent melanocytes, but not early appearing kit-dependent melanocytes, two predictions can be made. First, double mutants for fms and rose should not have a different number of melanocytes than the more severely melanocyte-deficient rose single mutant (since the two genes would be acting to affect only one melanocyte population). Second, double mutants for fms and kit should have an additive deficit in melanocyte number, equal to the reduction attributable to the fms mutant alone, plus the reduction due to the kit mutant alone (since the two genes would be acting independently to affect two different melanocyte populations). Fig. 9 presents tests of these predictions. We find that fms; rose double mutants do not have significantly fewer dermal stripe melanocytes (i.e., non-scale melanocytes) than rose single mutants (means=123, 132 melanocytes/mm, respectively; t6=1.36, P=0.2; Fig. 9A,B,E), although these cells remain in a dispersed pattern, presumably reflecting the requirement of fms for melanocyte migration into stripes (Fig. 7B). In contrast, fms; kit double mutants exhibit a dramatic reduction in melanocyte number (mean=36 melanocytes/mm) as compared to either single mutant (fms and kit, means=181, 96 melanocytes/mm, respectively; Fig. 9C-E). Moreover, this reduction in fms; kit double mutants is not significantly different from that predicted (21.7 melanocytes/mm) by additive effects of the two mutations (t6=2.19, P=0.1). Together, these data suggest that fms is required by late appearing, rose-dependent melanocytes, but is not required by early appearing kit-dependent melanocytes. Thus, two essentially paralogous genes, fms and kit, are required by parallel populations of pigment cells: fms promotes the development of embryonic and adult xanthophores, as well as late appearing adult melanocytes; whereas kit promotes the development of embryonic melanocytes and early appearing adult melanocytes.
DISCUSSION
We have shown that the D. rerio pigment pattern mutant panther corresponds to an orthologue of the kit-related gene fms, and that fms is required by embryonic and adult xanthophores, and a subpopulation of adult melanocytes. These results have implications for our understanding of the roles played by receptor tyrosine kinases during development, the cell populations constituting the D. rerio adult pigment pattern, and the morphogenetic mechanisms underlying pigment pattern formation.
Novel roles and phylogenetically conserved roles for fms during development
Studies of mouse mutants have identified several genes required for neural crest and pigment cell development, including kit, mitf and ednrb (Bennett, 1993; Shibahara et al., 1998; Reedy et al., 1998). Mutants corresponding to each of these genes now have been identified in D. rerio, and all exhibit pigment pattern defects (Parichy et al., 1999; Lister et al., 1999; D. M. P., R. N. Kelsh, S. L. J., unpublished data), revealing a conservation of molecular mechanisms for pigment pattern formation across amniotes and teleosts. In contrast, our demonstration that panther is allelic to fms is unexpected, as this gene has not previously been implicated in neural crest or pigment cell development. fms is the closest known homologue of kit and the ligand for the Fms receptor is Colony Stimulating Factor 1 (CSF1), which has structural similarities to the Kit ligand, Steel Factor (Bazan, 1991). Studies of amniotes, including a CSF1-deficient mouse mutant reveals that Fms is expressed and required by cells of the mononuclear phagocyte system, including macrophages in the reproductive tract and elsewhere, microglia in the central nervous system, and bone-resorbing osteoclasts (Ducy and Karsenty, 1998; Motoyoshi, 1998; Flanagan and Lader, 1998). Although a role for Fms in the development of the amniote neural crest–melanocyte lineage has not been excluded, Csf1 mutant mice do not exhibit pigmentation defects (Marks and Lane, 1976).
Several studies have indicated that an additional round of genome duplication has occurred in teleosts as compared to amniotes (Postlethwait et al., 1998; Amores et al., 1998). This inference raises the possibility that the gene that we have identified might correspond to a teleost-specific D. rerio paralogue of kit, rather than a true orthologue of amniote fms. Two lines of evidence argue against this notion. First, phylogenetic reconstructions place this gene as more closely related to mammalian fms genes than other type III receptor tyrosine kinases, including D. rerio kit. Second, our analyses reveal phylogenetically conserved expression in macrophage precursors and osteoclasts. These data strongly support the interpretation that we have isolated a D. rerio fms orthologue with a role in pigment pattern development. Nevertheless, we do not exclude the possibility that a second fms gene may be present in D. rerio, perhaps with more restricted functions than the locus reported here.
Xanthophore requirement for fms parallels melanocyte requirement for kit
Pigment patterns in amniotes typically reflect the physiology of epidermal, neural crest-derived melanocytes that contribute melanin to growing hair, feathers or scales. In contrast, pigment patterns in ectothermic vertebrates typically reflect the spatial organization of dermal neural crest-derived melanocytes, xanthophores and iridophores (Quevedo and Holstein, 1998; Bagnara, 1998). Although pigment synthesis genes are presumably expressed differentially across cell types, the extent to which different pigment cell classes share mechanisms of specification and morphogenesis remains unknown. Here, we have shown that gch+ and xdh+ presumptive xanthoblasts express fms and require this gene for their development, though they do not express or require kit. In contrast, dct+ late-stage melanoblasts express and require kit, but not fms (Parichy et al., 1999). These findings indicate a common requirement of xanthophores and melanocytes for signaling through type III receptor tyrosine kinases. Indeed, the failure of xdh+ cells to disperse from the neural crest in fms mutant embryos is reminiscent of the failure of melanoblasts to disperse in kit mutant embryos (Parichy et al., 1999) and is consistent with shared signal transduction pathways downstream of fms and kit (Dubreuil et al., 1991). Thus, migration from the neural crest depends, at least in part, on parallel signaling mechanisms in melanocytes and xanthophores.
fms promotes differentiation of a subpopulation of adult melanocytes
A previous study suggested that adult stripes in D. rerio arise from temporally and genetically distinct populations of melanocytes (Johnson et al., 1995b). An early developing population initially appears dispersed over the flank and depends on kit; a later developing population arises already in the position of adult stripes and depends on rose. Support for this model came from the finding that kit and rose mutations have additive effects, consistent with these genes acting on two different melanocyte populations: whereas kit and rose single mutants each lack half the normal complement of stripe melanocytes, kit; rose double mutants lack virtually all body stripe melanocytes. Two independent lines of analysis support a model in which fms promotes the differentiation of rose-dependent, but not kit-dependent melanocyte precursors. First, image series revealed that fms mutants have a severe deficiency in melanocyte differentiation, particularly in prospective stripe regions during late stages of pigment pattern metamorphosis. Second, genetic analysis demonstrated that fms; rose double mutants do not have significantly fewer body stripe melanocytes than rose single mutants. In contrast, fms; kit double mutants have a severe deficiency of stripe melanocytes, approximately that expected from additive effects of the two mutations. Taken together, these results strongly support the inference that early appearing melanocytes depend on kit, but not fms, whereas later appearing melanocytes depend on fms and rose, but not kit.
At least two models could explain the fms-dependence of late appearing melanocytes: signaling through Fms could promote the development of these cells indirectly, via interactions between melanocyte precursors and other fms-expressing cell lineages (e.g., see below); or, fms could promote the development of melanocyte precursors directly. For example, if fms is required for establishing or maintaining a population of precursor cells that is recruited to differentiate at metamorphosis, a disruption of Fms signaling could be manifested late in development as a deficit in the number of melanocytes that differentiate in stripes. Consistent with this idea, we observe fms expression in embryonic cells that express both the transcription factor mitf, which is expressed by melanoblasts (Opdecamp et al., 1997; Lister et al., 1999), and gch, which is expressed both by xanthophore and melanocyte lineages. Existence of pigment stem cells also has been inferred in studies of amniotes (e.g., Grichnik et al., 1996; Kunisada et al., 1998). Thus, we hypothesize that Fms signaling allows the development of a subpopulation of pigment cell precursors – that may or may not be pluripotent – and some of these cells differentiate as melanocytes in the position of stripes during pigment pattern metamorphosis.
fms mutant reveals roles for melanocyte death and migration during adult stripe development
In addition to contributions from localized differentiation and proliferation, adult melanocyte stripes could reflect roles for cell migration and differential cell death. Although several studies have described changes in melanocyte distributions during pigment pattern metamorphosis (Goodrich and Nichols, 1931; Kirschbaum, 1975; Johnson et al., 1995b; McClure, 1999), previous workers have not followed individual cells and thus could not distinguish between roles for different morphogenetic behaviors (e.g., a decrease in cell density could reflect either emigration or death of melanocytes). By observing the behavior of melanocytes in wild-type and fms mutants during the larval-to-adult transition, we identified roles both for melanocyte migration and for melanocyte death in pigment pattern metamorphosis. From our analyses of image series and gene expression, we suggest the following model, in which stripe morphogenesis employs both mechanisms (Fig. 10). Initially, melanocytes arise dispersed over the flank. Subsequently, xanthophores arise outside of prospective stripe regions and contribute to organizing melanocytes into stripes in two ways. First, xanthophores stimulate the directed migration of melanocytes into stripes. Second, by organizing melanocytes, xanthophores also increase the local densities of these cells, thereby increasing melanocyte survival (a community effect); melanocytes outside of stripes, in low density regions, lack this community effect on survival and tend to die. In contrast, mutants for fms lack xanthophores. Consequently, melanocytes fail to migrate into stripes, resulting in lower melanocyte densities, and thereby lower melanocyte survival throughout the flank.
Roles for fms in melanocyte migration and death appear to be indirect, as we did not observe fms expression in differentiated melanocytes. A direct role of fms in xanthophore development is suggested, however, by the absence in fms mutants of both unmelanized gch+ presumptive xanthoblasts during metamorphosis and xanthophores in adults. Therefore, the failure of directional melanocyte migration in fms mutants, concomitant with the failure of xanthophore development, suggests a role for fms-dependent xanthophores in organizing melanocytes into stripes. This inference is consistent with pattern-forming mechanisms in salamander larvae, in which interactions between melanophores and xanthophores stimulate the directed migration of melanophores into horizontal stripes and vertical bars (Epperlein and Löfberg, 1990; Parichy, 1996a,b). Thus, initially dispersed melanocytes in D. rerio may be stimulated to coalesce into stripes by ‘population pressure’ (Tucker and Erickson, 1986a,b; Thomas and Yamada, 1992) exerted by fms-dependent xanthophores. Indeed, fms+ and gch+ cells (presumptive xanthoblasts), as well as fully differentiated xanthophores, are abundant both within the prospective interstripe region, and further dorsally and ventrally over the flank, bounding the prospective stripe regions.
Our study also supports a requirement for community effects among melanocytes during normal stripe development. In wild-type larvae (this study) and in the fin (Goodrich et al., 1954; Goodrich and Greene, 1959), xanthophores are abundant in prospective interstripe regions and melanocytes die if present within these regions. Although this might suggest the hypothesis that xanthophores directly promote the death of melanocytes, our findings argue against this possibility as approx. 10% more melanocytes died in xanthophore-deficient fms mutants as compared to wild type. Instead, we suggest that increased melanocyte death and the different pattern of this death in fms mutants may result from decreased local melanocyte densities owing to a failure of these cells to form stripes. More specifically, analyses of patchwork mutant mice imply community effects during melanocyte development, such that a minimum density of these cells is required for their survival (Aubin-Houzelstein et al., 1998). In fms mutants, a failure of melanocytes to coalesce and differentiate in stripes would tend to increase the likelihood of melanocyte densities falling below a critical threshold for survival at any given point on the flank. Such a mechanism would explain both the disorganized pattern of melanocyte death in fms mutants, and the death of melanocytes isolated within the prospective interstripe region during normal development. Thus, our findings are consistent with a model in which stripe development in D. rerio depends both on melanocyte-xanthophore interactions, and interactions among melanocytes. We emphasize, however, that additional factors are likely to be required for stripe development (e.g., Tucker and Erickson, 1986a; Parichy, 1996a). Indeed, the persistence of poorly formed stripes in the anterior of fms mutant adults (Fig. 1D) indicates that stripe development is likely to depend in part on different underlying mechanisms at different axial positions. Finally, the results of this study set the stage for future investigations into the developmental bases for naturally occurring variation in adult pigment patterns and, in this regard, fms has recently been implicated as a candidate gene for the evolutionary loss of stripes in the pearl danio, D. albolineatus (D. M. P. and S. L. J., unpublished data).
Acknowledgements
For helpful comments and suggestions on the manuscript, we thank Carol Erickson, Kathy Iovine, Joan Manuel, Mark Reedy and Jim Skeath. Additionally, we are grateful to Matt Clark, Raymond Lee and others of the Washington University-MPIMG Zebrafish EST project (RO1-DK55379) for identification of ESTs used in this study. We thank David Beier for genotyping fms on our standard mapping panel. Supported by NIH RO1-GM56988 to S. L. J., NSF-Sloan Postdoctoral Fellowship in Molecular Evolution DBI-9750006 to D. M. P., and NIH 1 P50 DK49216-07 to L. I. Z.; S. L. J. is a PEW Scholar in Biomedical Sciences and L. I. Z. is an Associate Investigator of the Howard Hughes Medical Institute. We thank J. Sheppard and J. Manuel for fish care.