Embryonic requirements for ErbB signaling in neural crest development and adult pigment pattern formation

The generation of adult form remains an enduring problem. An interesting system for studying the salient genetic and cellular mechanisms is the larva-to-adult transformation, or metamorphosis, of amphibians and fishes(Moran, 1994; Parichy, 1998; Webb, 1999; Brown and Cai, 2007). In zebrafish, Danio rerio, metamorphosis includes dramatic changes in the digestive, excretory and sensory systems; the peripheral nervous system;the skeleton, fins and integument; as well as physiology and behavior(Cubbage and Mabee, 1996; Brown, 1997; Elizondo et al., 2005; Tingaud-Sequeira et al., 2006; Engeszer et al., 2007).

The pigment pattern is another, particularly accessible, trait altered at metamorphosis (Kelsh, 2004; Parichy, 2006). The early larval pigment pattern develops in the embryo from neural crest-derived pigment cells, or chromatophores; this pattern is completed by 5 days post-fertilization (dpf) and comprises stripes of black melanophores with intervening yellow xanthophores. This pattern persists until metamorphosis(∼14 dpf) when melanophores begin to differentiate outside of the early larval stripes. During the next 2 weeks, new adult stripes begin to form as some metamorphic melanophores migrate to sites of adult stripe formation and other melanophores differentiate already at these sites. The result is a juvenile pigment pattern with two `primary' stripes of melanophores, bordering an `interstripe' of xanthophores.

Embryonic/early larval chromatophores and metamorphic chromatophores might have commonalities as well as differences. For example, several mutants lack chromatophore types or pigments both before and after metamorphosis(Lister et al., 1999; Parichy et al., 2000b; Lamason et al., 2005). Others exhibit defects in the adult but not in the embryo(Parichy et al., 2000a; Iwashita et al., 2006; Watanabe et al., 2006). Mutants in this latter class are interesting because they identify genes uniquely required to establish, maintain or recruit latent precursors that contribute to adult form. Included among these are two mutants, pumaand picasso, each having a normal early larval pigment pattern but fewer metamorphic melanophores (Fig. 1A,B) (Parichy and Turner,2003b; Parichy et al.,2003; Quigley et al.,2004). Whereas puma is required autonomously to metamorphic melanophore precursors during pigment pattern metamorphosis, the cellular and temporal requirements for picasso are not known.

Here, we show that picasso is allelic to erbb3b, which encodes an epidermal growth factor receptor (EGFR)-like tyrosine kinase. erbb3b is one of two zebrafish orthologues of human epidermal growth factor receptor 3 (HER3, ERBB3) and part of a larger family that includes EGFR (ERBB1), ERBB2 and ERBB4(Citri and Yarden, 2006; Stein and Staros, 2006). Ligands for ErbB receptors include Egf and neuregulins 1, 2 and 3. The receptors form dimers with individual monomers exhibiting different activities and ligand specificities: for example, ErbB3 lacks endogenous kinase activity,while ErbB2 lacks its own high affinity ligand. Whereas several heterodimers are possible, only a subset seems to have biological significance, with ErbB3 acting with ErbB2 (Graus-Porta et al.,1997; Jones et al.,1999; Oda et al.,2005) and potentially with Egfr(Soltoff et al., 1994; Frolov et al., 2007; Poumay, 2007). ErbB receptors function in glial morphogenesis (Lyons et al., 2005; Britsch,2007), and their misregulation is associated with a variety of cancers (Breuleux, 2007; Sergina and Moasser,2007).

In this study, we find that metamorphic melanophores express erbb3b, suggesting an autonomous activity that occurs late, during the larva-to-adult transformation. Nonetheless, we show that erbb3bfunctions both autonomously and non-autonomously to the metamorphic melanophore lineage. We also identify a major critical period for ErbB signals during embryonic neural crest cell migration, 2 weeks before metamorphosis,indicating a novel role for ErbB signals in establishing precursors to adult chromatophores. Finally, we demonstrate cryptic requirements for ErbB signals during metamorphosis itself, suggesting redundant functions with other pathways at this later stage. Our study provides new insights into the development of adult form and the genetic requirements of a trait expressed before and after metamorphosis.

Fish were maintained at 26-28°C, 14L:10D(Westerfield, 2000). picasso mutants were recovered in screens for N-ethyl-N-nitrosourea-induced mutations and mapped using the partially inbred strains AB^wp and wik^wp.

Chimeric embryos were generated by transplanting cells at blastula stages(3.3-3.8 hours post-fertilization) and then were reared through metamorphosis(Parichy and Turner,2003a).

Stock solutions of AG1478 [4-(3-chloroanilino)-6,7-dimethoxyquinazoline;Calbiochem] or PD158780(4-[(3-bromophenyl)amino]-6-(methylamino)-pyrido[3,4-d]pyridimine; Calbiochem)were diluted in DMSO. Fish were treated with 3 μM of either drug in 10%Hanks solution. To facilitate penetration, 0.5% DMSO was added to all media and embryos were dechorionated prior to treatment. Fish were reared in agar-lined Petri dishes or glass beakers and solutions were changed daily. Fish reared in either drug throughout development invariably died prior to formation of the adult pigment pattern, so could not be analyzed.

A splice-blocking morpholino against erbb3b[TGGGCTCGCAACTGGGTGGAAACAA (Lyons et al.,2005)] was obtained from GeneTools (Eugene, OR). One- or two-cell embryos were injected with 300-500 pg and reared through formation of the adult pigment pattern.

For RT-PCR, metamorphosing larvae were euthanized and rinsed in 10% Hanks solution, after which tissues were dissected and placed in dissociation medium(1 mg/ml collagenase type IV; 0.1 mM epinephrine; 2 mg/ml bovine serum albumin; 0.1 mg/ml trypsin inhibitor)(Clark et al., 1987). Cells were picked and transferred to 5% fetal bovine serum in PBS for 10 minutes,then picked and extracted for RNA. cDNAs were synthesized with the Superscript III CellsDirect cDNA Synthesis System (Invitrogen) and RT-PCRs were performed using the following primers (forward, reverse): erbb3b,ACTCCCTAAAAATCCCTGTGG, GGCGAAGGTGTTGAAGTAAT; erbb2,CACCGGAAGTTTACTCACCAA, GATCTCCAACATTTGACCAT; erbb3a,TGACTCCATCCACTACTGCTG, TTCTTCACCAGCACCTCTGTT; egfr,CCGTTGGTGTGTGTTTTGAG, GCTTTTCAGGAGGGAGACTTTC; dct,ACCTGTGACCAATGAGGAGATT, TACAACACCAACACGATCAACA; β-actin,GTTTTCCCCTCCATTGTT, GGTGTTGAAGGTCTCGAACA; erbb4,CTGCTGCTCAACTGGTGTGT, CCAGTGCCATCACAGCTTCT.

For genotyping picasso^wp.r2e2, we amplified genomic DNA(pcs-wpr2e2^*: TTGGTTACCATTGTGGTTGTTT, TCTTCATGGTAGCTCAGA AACATC)from individual embryos and digested PCR products with RsaI restriction enzyme. The wild-type amplicon cuts with RsaI at position 219, whereas the mutant allele does not cut.

Analyses of mRNA distributions in embryos followed standard protocols(Parichy et al., 2000b). In situ hybridization on larvae followed(Elizondo et al., 2005), but overnight incubation was used for probes and antibodies (a detailed protocol is available at http://protist.biology.washington.edu/dparichy/). For analyses of gene expression in families segregating picassomutant alleles, individual embryos or larvae were imaged after staining then transferred to DNA extraction buffer and processed to determine genotypes retrospectively.

Trunks of 12 dpf larvae were fixed in 4% PFA in PBS for 6 hours at room temperature then permeabilized by washing overnight in deionized water. Specimens were equilibrated in PDTX (PBS containing 1% DMSO and 0.3% Triton X-100) three times for 30 minutes each, then blocked with 5% goat serum in PDTX for 4 hours. Larvae were incubated overnight at 4°C with primary antibody mAB 16A11 (Marusich et al.,1994; Henion et al.,1996) against HuC/D (1:200 in blocking solution), washed in PDTX,incubated with secondary antibody (Alexa Fluor 568; Molecular Probes), then washed and visualized.

Embryos or larvae were viewed with Olympus SZX-12 or Zeiss Lumar stereomicroscopes, or with Zeiss Axioplan 2 or Zeiss Observer compound microscopes. Digital images were collected with Zeiss Axiocam cameras using Zeiss Axiovision and corrected for contrast and color balance when necessary.

Statistical analyses were performed with JMP 7.0 (SAS Institute, Cary, NC). For counts of melanophores, individual cells were distinguished from one another by treating fish with epinephrine to contract melanosomes towards cell bodies. Densities of melanophores were determined by counting melanophores within a rectangular region delimited by: anteriorly, the anterior margin of the dorsal fin insertion; posteriorly, the posterior margin of the anal fin insertion; dorsally, the posterior margin of the dorsal fin insertion;ventrally, the posterior margin of the anal fin insertion. To control for variation in larval development stage, we tested for effects of larval size(measured as flank height at the posterior margin of the anal fin, hpa) as a covariate in analyses (Parichy and Turner,2003b), and retained this factor if P<0.05, though analyses without the co-factor yielded qualitatively equivalent results. Analyses of melanophore densities were treated as multifactorial analyses of variance or covariance with replicates as blocks. Residuals in all analyses were confirmed to be normally distributed and homoscedastic. Least squares means (correcting for size, replicate variation, or both) are presented in figures below, with significant differences assessed by Tukey-Kramer comparisons to preserve an experiment-wide α=0.05.

For analyses of embryonic critical periods for Erbb signals in kitmutant larvae (see below), adult pigment patterns were scored qualitatively for stripe disruption. Breaks in stripes were considered present when three or fewer melanophores were present over a defined anterior-posterior length, as scaled by hpa (above): stripes exhibiting breaks ≤0.5 hpa were scored `0';breaks between 0.5 and 1 hpa were scored `1'; breaks >1 hpa were scored`2'. Dorsal and ventral stripes were scored individually, then summed to generate a `stripe break score' of 0-4. To test for differences among treatment groups, we compared ordinal scores using non-parametric Wilcoxon tests and contingency table analyses. Both methods yielded equivalent results;for simplicity, we present only the former (complete analyses available on request).

To learn when picasso mutants first exhibit pigment pattern defects, we examined embryos and early larvae, and we imaged individual fish daily from early larval stages through formation of the adult pigment pattern. Pigment cell complements of embryos and early larvae were normal(Fig. 1C,D). Subsequently, picasso mutants largely failed to develop metamorphic melanophores,particularly in the mid-trunk, and instead retained early larval melanophores even as adults (Fig. 1E-L). In the posterior trunk of picasso mutants, seemingly more complete melanophore stripes form (Fig. 1B). Posterior stripes in picasso mutants appear more complete because more embryonic/early larval melanophores become situated in the position of adult stripes in this region, and because more metamorphic melanophores differentiate near these cells when compared with the mid-trunk(Fig. 2).

We mapped picasso to chromosome 23 in the vicinity of erbb3b and found that picasso failed to complement an excess neuromast phenotype of an erbb3b-null allele (all of these alleles are recessive and homozygous viable, though weaker than wild type)(Lyons et al., 2005). erbb3b cDNAs had premature stop codons in each of two picasso alleles (Fig. 3), demonstrating that the picasso phenotype arises from mutations in erbb3b.

In embryos, erbb3b is expressed in neural crest cells and glia(Lyons et al., 2005). In metamorphosing larvae, we similarly found erbb3b expression in glia(Fig. 4A). To determine whether erbb3b might be expressed in other tissues below the threshold of detection by in situ hybridization, we used RT-PCR. We detected erbb3b transcripts in both isolated melanophores and in juvenile fin(comprising melanophores, melanophore precursors, bone, skin, vasculature and other cell types; Fig. 4C). We also detected the erbb3b paralogue, erbb3a, in glia and fin,though not in metamorphic melanophores(Fig. 4B,C). As ErbB receptors act as heterodimers, we tested where other ErbB genes are expressed(Fig. 4C): erbb2 was expressed in metamorphic melanophores and in fin; egfr was not expressed in metamorphic melanophores, although it was expressed in fin; and we could not detect erbb4 in melanophores or in fin (data not shown). erbb2 and egfr are also widely expressed in zebrafish embryos (Goishi et al., 2003; Lyons et al., 2005).

To determine what steps in metamorphic melanophore development require erbb3b, we examined molecular markers(Fig. 5A-F). picassomutants were deficient during metamorphosis for cells expressing early neural crest markers (crestin, sox10), as well as early and late markers of the melanophore lineage (mitfa, dct). picasso mutants also had transiently fewer cells expressing xanthophore lineage markers (xdh,csf1r) and fewer myelin basic protein⁺(mbp⁺) glia (Fig. 5G,H and data not shown).

erbb3b might promote adult pigment pattern formation by acting autonomously to the metamorphic melanophore lineage, but also could have non-autonomous effects if, for example, erbb3b-dependent cells provide signals required by metamorphic melanophores or their precursors. To test these possibilities, we constructed genetic mosaics by transplanting cells between blastula stage embryos.

If erbb3b acts autonomously to the metamorphic melanophore lineage, then wild-type melanophores should develop in picassomutants and these cells should form wild-type stripes. If erbb3b acts non-autonomously, then wild-type melanophores should develop where picasso mutant melanophores develop (anteriorly and posteriorly), but not where picasso mutant melanophores are absent (mid-trunk)(Fig. 1B). Wild-type(β-actin::EGFP⁺) → picasso chimeras developed wild-type metamorphic melanophores at high density anteriorly and posteriorly(Fig. 6A,B), but often developed few if any metamorphic melanophores in the mid-trunk(Fig. 6A,C), like picasso mutants. In reciprocal picasso(β-actin::EGFP⁺) → wild-type chimeras, we never found donor picasso mutant metamorphic melanophores in the adult pigment pattern. These findings suggest both non-autonomous and autonomous roles for erbb3b.

Given that metamorphic melanophores express erbb3b and erbb2, differences between wild-type and picasso mutant melanophores could be further revealed as differences in their abilities to populate a flank lacking melanophores. We therefore transplanted wild-type or picasso mutant cells to nacre^w2 mutant hosts,which lack their own melanophores because of a mutation in mitfa,which functions cell-autonomously in melanophore specification(Lister et al., 1999; Parichy and Turner, 2003a). In wild-type → nacre chimeras, embryonic/early larval melanophores often developed, and metamorphic melanophores differentiated to form patches of stripes (Fig. 6D). In picasso → nacre chimeras, embryonic/early larval melanophores developed about as often, but metamorphic melanophores did not appear and, instead, embryonic/early larval melanophores persisted into the adult (Fig. 6E). Interestingly,metamorphic melanophores failed to develop in picasso → nacre chimeras, even anteriorly and posteriorly; this difference from picasso mutants may arise because the melanophore-free nacrebackground would preclude community effects from contributing to pattern regulation in these regions (Fig. 2) (Parichy et al.,2000b; Parichy and Turner,2003b). Together, genetic mosaic analyses indicate that ErbB signals are required autonomously and non-autonomously during metamorphic melanophore development.

The adult pigment pattern of picasso mutants could reflect erbb3b activities early or late. For example, erbb3b could function in the embryo to establish a population of precursors that differentiates at metamorphosis. Or erbb3b could act later in maintaining or expanding such a population, or still later, during their differentiation into metamorphic melanophores. To distinguish among these possibilities, we blocked ErbB signaling using pharmacological inhibitors AG1478 (Levitzki and Gazit,1995; Lyons et al.,2005; Levitzki and Mishani,2006) and PD158780 (Fry et al., 1997; Rewcastle et al.,1998; Frohnert et al.,2003). Preliminary analyses showed that treating wild-type embryos with either AG1478 or PD158780 resulted in excess neuromasts that phenocopy erbb3b mutants (data not shown)(Lyons et al., 2005). As both drugs inhibit kinase activity by interfering with ATP-binding sites, and wild-type Erbb3 already has impaired or absent kinase activity(Guy et al., 1994), inhibitors presumably suppress signals originating with erbb3b:erbb2, erbb3:egfr or other heterodimers. Functions of these receptors that are independent of kinase activity should not be affected.

Wild-type embryos treated with AG1478 developed normal early larval pigment patterns. When these same fish metamorphosed, however, their pigment patterns and melanophore densities were indistinguishable from picasso mutants(Fig. 7B,E). By contrast, fish treated with AG1478 during the pre-metamorphic (early larval) period or during metamorphosis, developed adult pigment patterns and melanophore densities indistinguishable from controls (Fig. 7A,C-E; but see below). Treatment of wild-type fish with PD158780 yielded identical results (Fig. 8 and data not shown).

To test whether the embryonic requirement for ErbB signaling is unique to zebrafish, we examined two more species. We chose D. albolineatusbecause its more uniform pigment pattern(Fig. 7F) might depend on mechanisms different than zebrafish(Quigley et al., 2005; Mills et al., 2007). Danio albolineatus embryos developed defects in metamorphic melanophores similar to D. rerio when treated with AG1478(Fig. 7G) or PD158780 (data not shown). We also examined D. nigrofasciatus(Fig. 7H), in which few metamorphic melanophores develop and, instead, most embryonic/early larval melanophores persist and reorganize to form adult stripes(Quigley et al., 2004). If AG1478 effects are limited to metamorphic melanophores, then the D. nigrofasciatus pigment pattern should be refractory to perturbation. Consistent with this prediction, D. nigrofasciatus embryos treated with AG1478 developed adult pigment pattern defects(Fig. 7I) less severe than those of zebrafish or D. albolineatus.

In mammalian systems, AG1478 is highly selective for EGFR-dependent signals and is less effective against other ErbB receptors(Levitzki and Gazit, 1995),whereas PD158780 is highly effective against all ErbB family members(Fry et al., 1997; Frohnert et al., 2003; Stonecypher et al., 2005). Specificities in zebrafish are not known. Although the similarity of adult pigment patterns between drug-treated fish and the picasso mutant is consistent with the suppression of erbb3b-dependent signals, we would expect these inhibitors to affect signaling through other ErbB receptors as well,particularly as protein tyrosine kinase domains are highly conserved between zebrafish and human orthologues (e.g. domain-specific identities,similarities: Erbb2, 84%, 92%; Egfr, 87%, 95%; Erbb3, 75%, 87%). We therefore repeated these experiments on picasso mutants: if signals independent of erbb3b are inhibited, then the severity of the picassophenotype should be enhanced. When we treated sibling picasso^wp.r2e2 and picasso^wp.r2e2/+embryos with AG1478 for the first 4 days of development, homozygotes unexpectedly developed edema and died by 7 dpf. Thus, a single copy of erbb3b protects against AG1478-dependent lethality, implying that erbb3b has functions that are independent of kinase activity (that,itself, presumably originates with erbb3b:erbb2 or erbb3b:egfr heterodimers). Consistent with this idea are several studies that have revealed kinase-independent activities of receptor tyrosine kinases, including Erbb3(Offterdinger et al., 2002; Rawls and Johnson, 2003; Massie and Mills, 2006; Hsu and Hung, 2007).

Given the preceding results, we treated embryos for shorter periods:wild-type embryos treated for only 2 dpf developed pigment patterns similar to wild-type embryos treated for 4 dpf (Fig. 9A,B); moreover, both picasso^wp.r2e2 and picasso^wp.r2e2/+ embryos treated for 2 dpf survived and developed pigment patterns indistinguishable from untreated picasso^wp.r2e2 controls(Fig. 9C,D; log-transformed melanophore densities: F_1,25=2.46, P=0.13). We observed identical outcomes with PD158780 (data not shown). By comparison with the picasso^wp.r2e2 null phenotype, these data suggest that inhibitors affect adult pigment patterns largely or exclusively by suppressing erbb3b-dependent signals.

These data indicate that ErbB signals are required in embryos for adult pigment pattern formation. To further test this conclusion, we sought an independent means of blocking erbb3b activity. We reasoned that the limited perdurance of morpholino oligonucleotides (3-5 days) should allow us to knock-down erbb3b in the embryo, while permitting later activity at metamorphosis (Nasevicius and Ekker,2000; Mellgren and Johnson,2004). We therefore injected embryos with a morpholino oligonucleotide against erbb3b(Lyons et al., 2005) and raised them into adults. Morpholino-injected fish showed defects qualitatively similar to picasso mutants (Fig. 7K).

Overall then, two independent lines of evidence show that erbb3bis required early for much later adult pigment pattern development. Specifically, the erbb3b mutant adult pigment pattern phenotype can be phenocopied in wild-type fish by: (1) embryonic knockdown of erbb3b via morpholino injection; and (2) treating embryos with either of two pharmacological inhibitors that, in this context, are specific to erbb3b-dependent signals.

The critical period for ErbB signals includes neural crest migration. To further test this coincidence, we treated embryos with ErbB inhibitors for shorter intervals. Preliminary analyses with wild-type yielded extensive variability in defect severity, perhaps owing to stochastic differences in pattern regulation (Parichy and Turner,2003b; Yamaguchi et al.,2007). We therefore used a sensitized background, kit^b5, to delineate the critical period more precisely. kit mutants lose embryonic/early larval melanophores, subsequently develop metamorphic melanophores already in stripes, and also have defects in pattern regeneration (Johnson et al.,1995; Parichy et al.,1999; Rawls and Johnson,2000; Yang and Johnson,2006).

We treated kit mutant embryos beginning between 8 hpf and 70 hpf for periods of 2-26 hours. Such analyses across multiple independent experiments revealed peak sensitivities for adult pigment pattern formation between ∼14 and 22 hpf, with affected individuals developing stripe defects reminiscent of picasso mutants (AG1478: Fig. 10A-C,E,F; PD158780: data not shown). As adult pigment patterns were comparatively refractory to treatments after ∼22 hpf, we asked whether longer treatments at later stages would enhance adult pattern defects. Treating embryos between 26 and 48 hpf did not alter later phenotypes (Fig. 10F), consistent with an earlier critical period. Finally, because erbb3b, erbb2 and egfr are expressed as early as 8-11 hpf[(Goishi et al., 2003; Thisse and Thisse, 2004; Lyons et al., 2005); data not shown], we tested whether ErbB signals have reiterated activities by treating embryos twice. When early treatments (8-11 hpf) were combined with later treatments (beginning at 22 hpf and later), we observed more severe melanophore deficiencies in the adult. Remarkably, treatments at least 8 hours apart often resulted in spatially separated melanophore-deficient patches(e.g. Fig. 10D). The increased severity of these defects suggests early and late functions even in the embryo: defects arising from early ErbB kinase inhibition can presumably be regulated so long as ErbB function is allowed later.

The major critical period for ErbB signals (∼14-22 hpf) corresponds approximately to the time when neural crest cells are migrating at the axial levels affected in the picasso mutant(Raible et al., 1992; Vaglia and Hall, 2000). Therefore, we explored the role of erbb3b in the early patterning of neural crest-derived cells. In comparison with wild-type siblings, picasso mutants at 26 hpf had similar numbers of cells expressing the pan-neural crest marker crestin, but these cells did not localize at sites of ganglion formation in the medial migratory pathway and were instead found further ventrally (Fig. 11A). We observed a similar defect for cells expressing mitfa (Fig. 11B). By contrast, we did not find clear defects in the distributions of dct⁺ melanoblasts(Fig. 11C) or cells in the lateral migratory pathway, consistent with the normal patterning of picasso mutant embryonic/early larval melanophores. Finally, given the defects in ventromedial migrating cells, and defects in the peripheral nervous system of mammalian ErbB3 and ErbB2 mutants(Britsch et al., 1998; Britsch, 2007), we examined ganglion development. We observed gross reductions in the numbers of dorsal root and sympathetic ganglia in picasso mutant larvae at 12 dpf(Fig. 11D). These data reveal an erbb3b-dependence of neural crest morphogenesis that correlates with the early erbb3b-dependence of adult pigment pattern formation.

The preceding experiments demonstrated a critical period for ErbB signaling in embryos. During later development, metamorphic melanophores express both erbb3b and erbb2 (Fig. 4C), but inhibitor treatments during metamorphosis had no effect on the adult pigment pattern (Fig. 7C-E). This suggests redundancies with other pathways, regulation in cell behaviors, poor penetration into tissues or higher thresholds for inhibition. Given these possibilities, we sought to further test roles for ErbB signals during metamorphosis. As higher doses were lethal, we re-tested the inhibitors on sensitized mutant backgrounds: kit, csf1r ^j4e1 and kit/+ csf1r/+. We chose these because they reveal distinct populations of metamorphic melanophores(Johnson et al., 1995; Parichy et al., 1999; Parichy et al., 2000b): kit mutants lack early metamorphic melanophores, but retain late metamorphic melanophores; csf1r mutants retain early metamorphic melanophores, but are missing late metamorphic melanophores. Comparing further deficits should therefore indicate whether one or the other population has a greater requirement for ErbB signaling. Finally, we also examined D. albolineatus because of the differences in melanophore development in this species compared with zebrafish.

In each background, we observed moderate to severe reductions in metamorphic melanophore numbers upon treatment with AG1478 or PD158780 during metamorphosis (Fig. 12; 19-58%fewer melanophores than corresponding controls), changes that were considerably more severe than observed for wild-type larvae(Fig. 7D,E; 3% fewer melanophores than controls). Furthermore, kit mutant and csf1r mutant zebrafish exhibited similar reductions in residual melanophore numbers, suggesting that ErbB signals are required by precursors to both early and late metamorphic melanophores. Consistent with the specificity of these effects, the picasso^wp.r2e2 mutant pigment pattern defect was not enhanced under these conditions and kit mutants treated with AG1478 exhibited pigment patterns that fell within the range observed for fish doubly mutant for kit and picasso^wp.r2e2 (data not shown). These data support a model in which ErbB signals are essential in the embryo but also function redundantly with other pathways during metamorphosis.

We have identified a crucial role for ErbB signaling in Danioadult pigment pattern development. While picasso mutants homozygous for erbb3b null alleles exhibit normal early larval pigment patterns,they are grossly deficient for metamorphic melanophores of the adult pigment pattern. Unexpectedly, metamorphic melanophore precursors require ErbB signals during neural crest development, ∼2 weeks before they begin to differentiate, and also exhibit a cryptic requirement for ErbB signals during metamorphosis. Genetic mosaic analyses further suggest complex modes by which ErbB signals promote adult pigment pattern formation.

A glial requirement for ErbB signals is well documented(Riethmacher et al., 1997; Britsch et al., 1998; Lyons et al., 2005; Pogoda et al., 2006; Britsch, 2007), but roles in pigment cell development have remained obscure. Normal human melanocytes express EGFR, ERBB2, ERBB3 and ERBB4, and stimulation with ligand promotes migration in vitro (Gordon-Thomson et al.,2001; Stove et al.,2003; Gordon-Thomson et al.,2005; Mirmohammadsadegh et al., 2005). ErbB receptors also are expressed in melanoma cells,and are associated with melanoma progression in a teleost model(Wellbrock et al., 2002; Gomez et al., 2004) and with human melanoma proliferation in vitro(Stove et al., 2003; Gordon-Thomson et al., 2005; Funes et al., 2006). Our finding that the picasso mutant phenotype results from lesions in erbb3b provides the first evidence that ErbB signals are required to promote normal pigment cell and pigment pattern development in vivo [although Fitch et al. (Fitch et al.,2003) describe melanocytosis resulting from EGFR overexpression in skin].

This study indicates that adult pigment pattern formation requires ErbB signaling in the embryo. Our analyses used pharmacological inhibitors that several lines evidence suggest are specific to ErbB-dependent signals. First,we observed the same phenotypes with two inhibitors. Second, pigment pattern defects phenocopied erbb3b-null alleles. Third, inhibitors failed to enhance defects of these null alleles. Fourth, similar defects resulted from morpholino-knockdown of erbb3b. Our results thus point to a model in which ErbB signals - depending in part on erbb3b - play an essential embryonic role in promoting much later adult pigment pattern formation.

This early critical period contrasts with other genes. For example, csf1r and puma are required during pigment pattern metamorphosis (Parichy and Turner,2003a; Parichy et al.,2003) and kit is required during pattern formation in the fin (Rawls and Johnson, 2001). The critical period for ErbB signals in zebrafish also may be earlier than for Ednrb of mouse (Shin et al.,1999). Nevertheless, our analyses can suggest only a range of times: both drugs act rapidly and are quickly reversible(Fry et al., 1997; Lenferink et al., 2001; Levitzki and Mishani, 2006),but we do not know how long it takes for concentration changes in solution to reach beneath the epidermis. The peak sensitivity observed for embryos treated between 14-22 hpf may therefore indicate somewhat later critical periods,presumably during neural crest migration.

We can envisage at least two complementary models in which embryonic ErbB signals contribute to later metamorphic melanophore development. In the first model, these signals act autonomously to establish precursors of metamorphic melanophores. This activity could be specific to metamorphic melanophores or could apply to a broader range of neural crest derivatives. For example, both pigment cells and glia share a common precursor(Dutton et al., 2001; Dupin and Le Douarin, 2003; Dupin et al., 2003), both can be generated by adult neural crest-derived stem cells(Sieber-Blum et al., 2004; Amoh et al., 2005; Wong et al., 2006), and both are affected by the erbb3b mutation and by ErbB inhibitor treatments. ErbB signals also may expand a precursor population. If so, then incomplete regulation could explain why the early larval pigment pattern is normal in picasso mutants: if precursors are allocated to fill a defined number of `embryonic/early larval niches' before filling `metamorphic niches', a depleted total number of cells could leave metamorphic niches vacant.

In a second model, ErbB signals promote adult pigment pattern formation non-autonomously to the metamorphic melanophore lineage. This could occur if ErbB-expressing cells provide trophic support to metamorphic melanophore precursors or contribute otherwise to a micro-environment where these precursors reside. Such interactions would be analogous to the non-autonomous mechanisms by which ErbB signals in glia promote neuronal survival and nerve integrity (Riethmacher et al.,1997; Chen et al.,2003; Sharghi-Namini et al.,2006). These observations also raise the possibility that peripheral nerves or ganglia serve as niches for metamorphic melanophore precursors. Consistent with this idea are the defects in ganglion development seen here and in the accompanying study(Honjo et al., 2008), and the presence of cells in peripheral nerves or ganglia of other organisms that are able to produce melanocytes and other cell neural crest derivatives(Nichols et al., 1977; Nichols and Weston, 1977; Ciment et al., 1986; Nataf and Le Douarin, 2000; Rizvi et al., 2002; Joseph et al., 2004).

Beyond the embryo, our data also indicate a role for ErbB signals during metamorphosis. At this stage, ErbB signals are likely to act autonomously to metamorphic melanophores, given their expression of erbb3b and erbb2, but also could have non-autonomous effects if interactions among melanophores promote the survival, proliferation or differentiation of these cells (Parichy et al.,2000b; Parichy and Turner,2003b).

In conclusion, this study supports a model in which ErbB signals, which are mediated in part through erbb3b, are required in the embryo to establish latent precursors that will subsequently generate metamorphic melanophores. Later, during metamorphosis, ErbB signals contribute to melanophore development but are partly or entirely redundant with other pathways. We speculate that erbb3b promotes the development of latent precursors intrinsically, and also is required extrinsically to form a niche where these cells reside until recruited at metamorphosis.

Thanks to members of the Parichy laboratory for helpful discussions and for assistance with fish rearing, to Will Talbot's laboratory for complementation testing of picasso against erbb3b, to Melissa Harris for comments on the manuscript, to Dave Raible's laboratory and especially Hillary McGraw for discussions and advice, and to Judith Eisen and Yasuko Honjo for sharing data prior to publication. Supported by NIH R01 GM62182 and NSF IOB 0541733 to D.M.P.