Endothelin receptor Aa regulates proliferation and differentiation of Erb-dependent pigment progenitors in zebrafish

Skin pigment patterns are important, being under strong selection for multiple roles including camouflage and UV protection. Pigment cells underlying these patterns form from adult pigment stem cells (APSCs). In zebrafish, APSCs derive from embryonic neural crest cells, but sit dormant until activated to produce pigment cells during metamorphosis. The APSCs are set-aside in an ErbB signaling dependent manner, but the mechanism maintaining quiescence until metamorphosis remains unknown. Mutants for a pigment pattern gene, parade, exhibit ectopic pigment cells localised to the ventral trunk, but also supernumerary cells restricted to the Ventral Stripe. Contrary to expectations, these melanocytes and iridophores are discrete cells, but closely apposed. We show that parade encodes Endothelin receptor Aa, expressed in the blood vessels, most prominently in the medial blood vessels, consistent with the ventral trunk phenotype. We provide evidence that neuronal fates are not affected in parade mutants, arguing against transdifferentiation of sympathetic neurons to pigment cells. We show that inhibition of BMP signaling prevents specification of sympathetic neurons, indicating conservation of this molecular mechanism with chick and mouse. However, inhibition of sympathetic neuron differentiation does not enhance the parade phenotype. Instead, we pinpoint ventral trunk-restricted proliferation of neural crest cells as an early feature of the parade phenotype. Importantly, using a chemical genetic screen for rescue of the ectopic pigment cell phenotype of parade mutants (whilst leaving the embryonic pattern untouched), we identify ErbB inhibitors as a key hit. The time-window of sensitivity to these inhibitors mirrors precisely the window defined previously as crucial for the setting aside of APSCs in the embryo, strongly implicating adult pigment stem cells as the source of the ectopic pigment cells. We propose that a novel population of APSCs exists in association with medial blood vessels, and that their quiescence is dependent upon Endothelin-dependent factors expressed by the blood vessels.


Introduction
Pattern formation is a crucial aspect of development since it creates the functional arrangements of cell-types that allow an organism to thrive. Pigment pattern formation-the generation of correctly distributed pigments or pigmented cells within the skin or elsewhere in the body-is a case in point, with pigmentation crucial for diverse aspects of an animal's ecology, including avoidance of predators, kin recognition, mate selection, thermal regulation and UV protection [1].
In vertebrates, all pigment cells except those of the pigmented retinal epithelium, are derived from a transient embryonic tissue called the neural crest. Neural crest cells are multipotent, generating numerous types of neurons, glia, pigment cells and other derivatives. They are also highly migratory, moving from their origin in the dorsal neutral tube to occupy diverse sites throughout the embryo. Thus, correct positioning of the different cell-types is a crucial aspect of their development.
Pigment cells in mammals consist only of melanocytes, making (and secreting) black eumelanin or yellow pheomelanin granules. In fish, amphibians and reptiles, pigment cells are much more diverse [2], allowing the generation of the varied and often beautiful pigment patterns these groups display. The zebrafish Danio rerio has rapidly become a paradigmatic example for the genetic and cellular study of pigment pattern formation [1,[3][4][5][6]. Zebrafish pigment patterns consist of three pigment cells, black melanocytes making melanin, yellow xanthophores making pteridines and carotenoids, and iridescent iridophores containing reflecting platelets [2].
Zebrafish, in common with most fish, develop two distinct pigment patterns, an early larval pigment pattern generated in the embryo by direct development of pigment cells from neural crest cells, and an adult pattern formed during metamorphosis, mostly through the de novo differentiation of pigment cells from adult pigment stem cells (APSCs, also formerly known as melanocyte stem cells; [7][8][9]). The adult pigment pattern consists of prominent stripes consisting of melanocytes and associated blue iridophores, alternating with pale stripes (interstripes) consisting of dense silver iridophores and xanthophores. Pigment pattern formation in adults is partially well characterised, with many genes identified that regulate the production of different pigment cell-types from the APSCs or which control their cellular interactions to create the bold horizontal stripe pattern [1,4,5,10]. A key aspect of adult pigment pattern formation that is less well-understood is the generation of the APSCs from neural crest cells. Remarkably, elegant experimental studies from multiple laboratories have established that these are setaside from the neural crest in a narrow time-window (9-48 hours post-fertilisation (hpf); [11]), with at least some occupying a niche within the dorsal root ganglia (DRGs; [9,12]) of the peripheral nervous system. They remain quiescent until metamorphosis begins around 20 days post-fertilisation (dpf), when they become activated. The mechanisms controlling their quiescence and their activation are largely unknown.
Pigment pattern formation in embryos is also poorly understood [3]. The embryonic pigment pattern consists principally of four longitudinal stripes of melanocytes, with iridophores arranged in a characteristic association with the melanocytes in three of these (Dorsal, Ventral and Yolk Sac Stripes), whereas the Lateral Stripe consists only of melanocytes; xanthophores then occupy the space under the epidermis between these stripes. Whereas in the adult the stripes are in the dermis, in the embryo they are associated with other structures, including the CNS, horizontal myoseptum of the body muscle blocks, the internal organs and the ventral most yolk sac. One study has investigated the detailed mechanism driving the association of melanocytes with the horizontal myoseptum [13]. In general, stripes form through migration of pigment cell precursors down migration pathways used by neural crest. These neural crest migration pathways are known as the dorsal (or dorsolateral) migration pathway, consisting of cells migrating under the epidermis and over the outer face of the somites/developing muscle blocks, and the medial migration pathway, running between the neural tube and notochord and the medial face of the somites/developing muscle blocks [14]. Pigment cell precursors of different fates use distinct migration pathways. Thus, xanthoblasts only use the lateral migration pathway, iridoblasts use only the medial migration pathway, whereas melanoblasts use both [15][16][17][18]. Note that migrating pigment cell precursors often show early signs of pigmentation i.e. they are differentiating as they migrate. During the migration phase, early differentiating melanocytes and iridophores can be found in the ventral trunk on the medial migration pathway, but these have disappeared by 72 hpf as those cells migrate into the Ventral and Yolk Sac Stripes [17].
As in adult pigment pattern formation, mutants affecting embryonic pigment pattern offer an exciting entry-point to the study of the mechanisms controlling pigment pattern formation. In a large-scale ENU mutagenesis screen performed in 1996, we identified two zebrafish mutant alleles, pde tj262 and pde tv212 , that defined the parade (pde) gene [19]. These mutants showed ectopic melanophores and iridophores in a well-defined region of the ventral side of the posterior trunk (Fig 1B and 1E), in addition to a stripe pattern similar to the wild type (WT) pigment phenotype (Fig 1A and 1C). The striking coincidence of melanin and reflecting platelet distribution lead us to initially propose that the pde mutant phenotype results from differentiation of pigment cells of mixed fate i.e. with both melanin granules and reflecting platelets within the same cell [19]. Here we perform a comprehensive analysis of the pde mutant phenotype. We show that the pde locus encodes a zebrafish Endothelin receptor A, Ednraa, which in the trunk is expressed only in the developing blood vessels, although in the head it is also expressed in cranial neural crest cells and is associated with craniofacial patterning [20]. Using chemical genetics, coupled with analysis of cell fate studies, we propose that APSCs occupy a niche associated with the medial blood vessels of the trunk, likely to be a second PNS niche within the sympathetic ganglion chain, and that these become precociously activated in pde/ednraa mutants. Thus we hypothesise that key components of that blood-vessel niche are Ednraa-dependent factors that promote APSC quiescence. Endothelin receptor Aa affects pigment progenitor differentiation

Early larval pde mutants have supernumerary iridophores and ectopic melanocytes and iridophores
To test our initial working model, that the ectopic cells are cells of mixed melanocyte/iridophore fate, we began by asking exactly where the ectopic cells were located, and which pigment cell-types were involved. Whilst WT siblings at 5 dpf ( Fig 1D) show no pigment cells in the ventral trunk between the notochord and the Ventral Stripe, in pde tj262 mutants (from now on referred to simply as pde unless specified otherwise) ectopic chromatophores are located in a medial position, directly beneath the dorsal aorta (DA) and above the posterior cardinal vein (PCV; Fig 1F). Thus, their location corresponds precisely to the ventral region of the medial neural crest migration pathway.
In zebrafish, pigment cell migration (both as unpigmented progenitors and as pigmenting cells) is pathway specific, with cells of the melanocyte and iridophore, but not xanthophore, lineages using the medial pathway [3]. Direct observation clearly shows cells with the pigmentation characteristics of both melanocytes and iridophores in this region of pde mutants, but detection of xanthophores by their pigmentation is difficult at these stages. Thus, to assess the presence of ectopic xanthophores, we used whole-mount in situ hybridisation (WISH) using fate-specific markers. Detection of the melanocyte marker dopachrome tautomerase (dct; Fig  1J), the iridoblast marker endothelin receptor ba (ednrba; Fig 1K) and the xanthophore marker GTP cyclohydrolase I (gch; Fig 1L) at 72 hpf, provided clear confirmation that pde mutants display ectopic melanocytes and iridophores, but importantly revealed the complete absence of ectopic xanthophores in the ventral trunk ( Fig 1L).
Initial observations of pde mutants indicated the intriguing possibility that some of the ectopic chromatophores displayed mixed characteristics of both iridophores and melanophores, i.e. both reflecting platelets and melanin-containing melanosomes (Fig 1E and 1F; [19]). This suggested that the phenotype might result, at least in part, from a failure of the normal mutual repression of alternative pigment cell fates. To test this, we used transmission electron microscopy to assess the structure of the ectopic cells. Unexpectedly, we consistently saw that melanosomes, the melanin synthesising organelles of melanocytes, and the reflecting platelets that contain the reflective crystals in iridophores, formed separate clusters, and were never intermingled (Fig 2A-2C). Furthermore, these clusters were consistently separated from each other by double membranes (Fig 2A-2C), similar to the appearance of melanocytes and iridophores in the WT Yolk Sac Stripe (S1 Fig), and consistent with them being separate cells. We conclude that the ectopic pigment cells are not of mixed fate, but instead are tightly associated individual melanocytes and iridophores; we note that this arrangement is characteristic of iridophores in their normal locations, where in each of the Dorsal, Ventral and Yolk Sac Stripes iridophores are tightly associated with melanocytes.
We hypothesised that the ectopic cells might reflect a defect in pigment cell migration through the ventral pathway, with cells that would normally contribute to the Ventral Stripe becoming stuck during migration. To test this, we quantitated the ectopic pigment cells and the cells of both the Dorsal and Ventral Stripes (Fig 2D-2I). Our model predicted that the counts in the Dorsal Stripe would be unaffected, but that pigment cells in the Ventral Stripe might be reduced, perhaps in proportion to the number of ectopic cells. Quantitation of the ectopic pigment cells (melanocytes + iridophores) in the ventral medial pathway revealed a variable phenotype, with a consistently larger number of iridophores than melanocytes in the ectopic position ( Fig 2D). The number of melanocytes in the Dorsal ( Fig 2F) and Ventral ( Fig  2G) Stripes in pde mutants was not statistically different to those of WT siblings. Similarly, there was no significant difference in the number of iridophores in the DS of pde mutants and Endothelin receptor Aa affects pigment progenitor differentiation WT siblings ( Fig 2H). Interestingly, and in contradiction to the disrupted migration model, the number of iridophores in the ventral stripe of pde mutants shows a 58% increase compared to WT siblings ( Fig 2I). Thus, we reject the disrupted migration hypothesis, and instead note that pde mutants display a regionally localised increase in melanocytes and iridophores in the ventral trunk, with some as supernumerary cells in the Ventral Stripe, but many in an ectopic position nearby.

pde encodes the endothelin receptor Aa gene
To identify the mutation that causes the pde phenotype we crossed pde heterozygotes onto a WIK wild-type background, and mapped the locus of the mutation using two sets of microsatellite markers, G4 and H2 [21]. The pde mutation showed strong linkage to markers Z15424 and Z23059 on chromosome 1,~1.5 cM and~3.8 cM away from the mutation, respectively ( Fig 3A). Using the 8 th version of the zebrafish genome assembly, further analysis showed that the marker P249 in clone BX51149 was only 0.2 cM (4 recombinants in 17959 embryos) away from the mutation, placing the pde mutation about 132 kb away in clone CU462997, in which three genes were annotated: 1) mineralocorticoid receptor (now renamed nuclear receptor subfamily 3, group C, member 2, nr3c2), 2) Rho GTPase activating protein 10 (arhgap10), and 3) endothelin receptor Aa (ednraa). Of these three candidate genes, ednraa has been previously reported to be required in the development and patterning of the neural crest-derived ventral cranial cartilages. Furthermore, ednraa is expressed in the developing blood vessels of the posterior trunk [20]. This striking correlation between ednraa expression and the region with ectopic pigment cells in pde mutants made ednraa a strong candidate. Previous studies had not reported a pigment phenotype during morpholino-mediated knockdown, but the focus in that work had been on craniofacial development, reflecting another region of ednraa expression [20].
To test the possible role of ednraa in pigment development, we used morpholino-mediated knockdown by injection of 1-2-cell stage wild-type embryos using the previously validated splice-blocking ednraa-morpholino (ednraa-MO1; S1 Table and Fig 3B; [20]). As controls, we used injection of a random-sequence morpholino (cMO; standard control provided by Genetools) and the ednrab morpholino (ednrab-MO; S1 Table; [20]), neither of which resulted in ectopic pigment cells, nor any other disruption to the early larval pigment pattern ( Fig 3C). In contrast, after injection of ednraa-MO1 we saw ectopic melanocytes and iridophores in the ventral medial pathway in a pde-like phenotype ( Fig 3D). As a further test, we designed a new translation blocking morpholino against ednraa (ednraa-MO2, S1 Table); injection of this morpholino phenocopied the pde mutant pigment phenotype, but also was prone to non-specific deformations, likely due to off-target effects [22]. Together, these data strongly support the hypothesis that the pde mutants disrupt ednraa function.
To assess directly the link between pde mutations and disruption of ednraa function we amplified ednraa cDNAs from pde mutant alleles to examine mRNA sequence and structure from mutants. In addition to the two original alleles, pde tj262 and pde tv212 , we also analysed pde hu4140 , a third mutation identified in an independent screen at the Hubrecht Institute which showed a phenotype indistinguishable from the original alleles, and which failed to complement those original alleles (S2 Fig). This cDNA sequencing showed that pde hu4140 has a single base transition mutation in the 3' region of exon 5 (bp 847; AGA > TGA; Fig 3B), which is predicted to result in a premature translation stop in Ednraa. The pde tj262 and pde tv212 alleles each showed deletions, of 103 bp and 135 bp respectively. cDNA sequence alignment indicates a deletion of exon 7 and a couple of extra bases, predicted to cause a frame shift of translation in exon 8 in pde tj262 , while pde tv212 has a deletion of exon 6 and a couple of extra bases which  Fig 3B). Given that the pde tj262 and pde tv212 alleles were isolated from an ENU-mutagenesis screen and hence are likely to result from induced point mutations, we propose that the mutations in pde tj262 and pde tv212 are likely to affect key bases involved in ednraa splicing.
The EdnrAa protein is a member of the rhodopsin-like G-protein coupled receptor family (GPCR, class A), and as such is characterised by seven transmembrane domains. In silico translation and structural predictions for the ednraa alleles ( Fig 3B; S5A Fig) indicate that the N-terminus and the early transmembrane domains are likely to be intact, but that the other transmembrane domains and the C-terminus are absent. Our three ednraa mutant alleles show indistinguishable phenotypes, and are fully recessive, consistent with the similar predicted molecular effects of the mutations, and strongly indicating that the receptor is not functional. Consequently, we propose that these alleles are all likely null mutants, although formal proof of this will require generation of an N-terminal truncation allele.
We used an in situ hybridization time-course between 6 and 72 hpf to assess the domain of ednraa expression, and in particular, whether it was detectable in neural crest or pigment cells. Our data was fully consistent with the earlier demonstration of ednraa expression in the developing blood vessels, but we saw no evidence of neural crest expression in the trunk (Fig 3E-3H). We then tested whether blood vessels morphology was affected in pde mutants, which we will refer to as ednraa mutants from now on. One possible explanation for the pigment cells might be that blood vessel morphology might be disrupted in ednraa mutants, resulting in misplaced pigment cells. However, neither in situ hybridisation for ednraa (Fig 3E-3H), nor examination of trunk blood vessel morphology using the transgenic line Tg(flia:GFP) (Fig 3I  and 3J) showed differences between WT and ednraa mutant embryos. We conclude that gross morphology of blood vessels is not affected in the ednraa mutants, but that the supernumerary and ectopic pigment cells in the ednraa mutants result from a non-cell autonomous effect of endothelin signalling in the blood vessels.

The ednraa phenotype does not result from neural crest cell transdifferentiation in the ventral trunk
The location of the ectopic pigment cells on the ventral medial pathway corresponds with the location of the nascent sympathetic ganglia, which form on the medial neural crest migration pathway in the vicinity of the dorsal aorta. We considered a transdifferentiation model in which neural crest cells fated to form sympathetic neurons switch to generating pigment cells, predicting that sympathetic neuron numbers would be reduced in ednraa mutants. However, immuno-detection of the early neuronal marker Elav1 (Hu neuronal RNA-binding protein) showed no differences in the number of sympathetic neurons between phenotypically WT embryos and their ednraa mutant siblings (Fig 4A, 4B and 4K). Furthermore, we also tested whether other neural crest-derived neurons are affected in the trunk, but neither DRG sensory neuron nor enteric neuron numbers differed between ednraa mutants and their WT siblings. Thus, at 5 dpf, trunk DRGs contained around 3 neurons per ganglion (Fig 4C, 4D and 4L), while the posterior gut contained around 125 enteric neurons (Fig 4E, 4F and 4M). Endothelin receptor Aa affects pigment progenitor differentiation As a further test of the sympathetic neuron transfating hypothesis, we reasoned that if a key signal driving sympathetic neuron specification was reduced, this might result in enhanced numbers of ectopic pigment cells. In mammals and birds, secreted BMP signals from the dorsal aorta have been shown to induce sympathetic neurons [23][24][25], but it is not known if this mechanism is conserved in zebrafish. To test this, we treated zebrafish embryos with Dorsomorphin (2.5 μM), a well characterised BMP signalling inhibitor [26]. Given the well-known roles for BMP signaling in early patterning in the embryo, we chose a treatment window from 1-4 days post fertilisation (dpf). Although this treatment left the larvae looking morphologically normal, immunofluorescent detection of Elav1 showed that treated larvae had a strong reduction the number of sympathetic neurons compared to DMSO carrier-treated controls ( Fig 4G, 4H and 4N). Having shown that zebrafish sympathetic neurons were BMP-dependent, we then asked whether ectopic pigment cells in ednraa mutants were increased if sympathetic neuron specification was inhibited; using the same treatment conditions, we saw no enhancement of ectopic pigment cells in ednraa mutants treated with dorsomorphin compared to DMSO-treated controls (Fig 4I, 4J and 4O). Thus, although we provide the first evidence to our knowledge that specification of zebrafish sympathetic neurons is BMPdependent, we discount the hypothesis that the ectopic pigment cells in ednraa mutants result from transfating of sympathetic neurons.

The ednraa phenotype results from localised increased neural crest cell proliferation in the ventral trunk, in the vicinity of the medial blood vessels
In order to identify the earliest stage at which ectopic pigment cells appear in the ventral trunk of ednraa mutants, we performed in situ hybridization with the melanocyte marker dct and the iridoblast marker ltk. Comparison of gene expression between WT and ednraa mutant embryos at 24, 30 and 35 hpf showed that ectopic/supernumerary expression of both dct and ltk is detected in the ventral trunk from 35 hpf (Fig 5A-5D).
We tested whether this production of extra pigment cells correlated with an increased proliferation of NC-derived cells in ednraa mutants. Given that ectopic and supernumerary pigment cells appear in the ventral trunk of ednraa mutants between 30 and 35 hpf, this indicates that enhanced proliferation might be expected to be detectable during this time-window. We used our Tg(sox10:cre) driver [27]combined with a Tg(hsp70:loxP-dsRed-loxp-lyn-egfp) redgreen switch reporter to label all neural crest cells with membrane tagged GFP (GFP fused with the membrane LYN Tyrosine-kinase), in conjunction with expression of the proliferation marker phosphohistone H3 (pH3) by immunofluorescent labelling. GFP expression in the neural crest-derived cells was induced by brief heat shock at 30 hpf. Subsequent double labelling of GFP and pH3 at 32 hpf showed a 55% overall increase in NC-derived proliferating cells in ednraa mutants compared with WT siblings, albeit over a very low basal rate (Fig 5E, 5F and 5G). Furthermore, over-proliferation of neural crest cells was detectable only in the ventral region of the medial pathway (Fig 5G). Our results show that ectopic pigment cells in ednraa Endothelin receptor Aa affects pigment progenitor differentiation During the proliferation assays, we noted that the ventral proliferation of neural crestderived cells was often clustered at the ventral end of the migrating streams (arrowheads, Fig  5F). Furthermore, this association was true also of the ectopic pigment cells themselves, as shown by imaging of all NC-derived cells using the same transgenic line (Tg(-4725sox10:cre) ba74; Tg(hsp:loxp-dsRed-loxp-EGFP))(S3 Fig). This imaging confirmed the widespread ventral migration of neural crest-derived cells in ednraa mutants, comparable to that in WT siblings, at 35 hpf (S3A and S3B Fig), reinforcing the conclusion from our WISH studies (Figs 1 and 5). At 5 dpf, this imaging readily showed the ectopic pigment cells, but clearly revealed that their location is ventral to the notochord, in close proximity to the dorsal aorta and associated with the sympathetic ganglial chain, and also near the ventral projections of the spinal nerves (S3C- S3F Fig).

The formation of ectopic pigment cells in ednraa mutants requires ErbB signalling
In order to identify possible mechanisms involved in the formation of the ednraa phenotype, we performed a small molecule screen of 1396 compounds using three different libraries: 1) the Sigma LOPAC library, which contains 1280 small organic ligands including marketed drugs and pharmaceutically relevant structures; 2) the Screen-WellTM Kinase Inhibitor Library that comprises 80 known kinase inhibitors; and 3) the Screen-WellTM Phosphatase Inhibitor Library, containing 33 phosphatase inhibitors of well-characterised activity. Screening was performed using the methodology of our previous screen for pigmentation modifiers [28]; we took advantage of the adult viability of the ednraa mutants, to perform the screen on ednraa mutant embryos. Thus ednraa mutant embryos were treated from 4 hpf to 96 hpf at a standard concentration of 10 μM and rescue or enhancement of ednraa mutant phenotype was systematically assessed at 4 dpf. After rescreening, we identified 23 compounds able to rescue the ednraa mutant phenotype, as well as 3 that enhance the phenotype (S2 Table). Of these hits, four (Tyrphostin AG 1478, U0126, PD 98059 and PD325901) target the MAPK/ERK pathway: Tyrphostin AG-1478 is an inhibitor of the Epidermal Growth Factor Receptor (EGFR) [29] while U0126, PD98059 and PD325901 are highly selective inhibitors of MEK1/2 signalling [30][31][32]. Our previous work has shown that MEK inhibitors interfere with production of regenerative melanocytes, whilst not affecting direct developing pigment cells forming the early larval pattern in zebrafish [33,34]. Treatment of ednraa mutants with each of two of the latter compounds shows a clear dose-response in the degree of rescue of the ednraa phenotype (S4 Fig). The identification of Tyrphostin AG-1478 in our screen is especially notable, because it and the more specific EGFR inhibitor, PD158780, have been shown to affect APSC biology, but not embryonic pigment cells (except a small population contributing to the Lateral Stripe), in zebrafish [11]. Similarly, mutants for the epidermal growth factor receptor (EGFR)-like tyrosine kinase erbb3b (picasso), despite developing a normal embryonic/larval pigment pattern, fail to develop a normal adult pigment pattern and are unable to regenerate melanocytes after embryonic melanocyte ablation due to a lack of APSCs [11]. Thus inhibition of Erb signalling by either AG-1478 or PD158780 selectively affects the biology of NC- Endothelin receptor Aa affects pigment progenitor differentiation derived APSCs that give rise to adult and regenerative melanocytes. The rescue of the ednraa mutant by treatment with the Erb inhibitors suggested the exciting hypothesis that the ectopic cells in the ednraa mutants might result, not from embryonic pigment cells, but by precocious differentiation of APSCs. Interestingly, this role for ErbB signaling shows a tightly constrained time window, since inhibition of ErbB signalling with AG-1478 and PD158780 during window 9-48 hpf of embryonic development is sufficient for these effects [11]. Moreover, after melanocyte ablation throughout 24-72 hpf, melanocyte regeneration normally occurs by 5 dpf, but this is prevented when embryos are treated with AG-1478 during a 9-30 hpf window [7].
Thus, our hypothesis makes the testable prediction that rescue of the ectopic pigment cells in ednraa mutants would share the very specific temporal window known to regulate APSCs development. We treated ednraa mutant embryos with a range of concentrations (0.1-2.0 μM) of PD158780 from 12-48 hpf (Fig 6A-6D). Quantification of the number of ectopic pigment cells showed that ednraa mutant embryos treated with PD158780 have a significant dose dependant-reduction in the number of ectopic cells compared to DMSO treated embryos ( Fig  6M). Moreover, shorter treatment with PD158780 revealed that the ednraa phenotype is effectively rescued when treatment is restricted to a 19-30 hpf window (Figs 6E-6H and 7M), showing a striking match to the critical period for establishment of APSCs, while treatment after this window (24-30 hpf) does not rescue the ednraa phenotype (Fig 6I-6M). This data strongly supports the hypothesis that the source of ectopic pigment cells in ednraa mutants is likely to be APSCs and not embryonic pigment cells.
To date there is only one marker that has been proposed to label APSCs: Dooley and colleagues identified cells expressing Tg(mitfa:GFP) in the DRGs as APSCs as early as 12 dpf

Discussion
In this study we show that ednraa encodes one of two zebrafish Ednra orthologues, Ednraa. In mammals the EdnrA receptor binds selectively to Edn1 and Edn2, mediates vasoconstriction, and is overexpressed in many cancers [35]). In contrast, loss of function in mouse results in homeotic transformation of the lower jaw towards an upper jaw morphology, and in humans underlies Auriculocondylar syndrome (ACS [OMIM # 602483 and # 614669]) [36][37][38][39]. These studies did not identify pigmentation phenotypes, although an Ednra lacZ knockin mouse strain shows prominent expression in the hair follicles [36]. In contrast, Ednrb and Edn3 mutants, as well as mutations in the Edn-processing enzyme Ece1, lack neural crest-derived melanocytes [40][41][42].
Analysis of the zebrafish genome identifies eleven components of endothelin signalling system: Four ligands, Edn1, Edn2, Edn3a and Edn3b; three Endothelin Converting Enzymes, Ece1, Ece2a, Ece2b that activate the ligands; and four receptors, Ednraa, Ednrab, Ednrba and Ednrbb [43]. In adult zebrafish, ednrba and ece2b loss-of-function mutants have all been shown to display reduced iridophores and broken stripes, indicating their coordinated role in iridophore development and pigment patterning, but no effect on embryonic pigment pattern [44,45]. In contrast, edn1 mutants and ednraa and ednrab morphants revealed disruption of the lower jaw, similar to the mammalian role in dorsoventral patterning, but did not examine pigmentation [20]. Here we identify a novel function for Ednraa signaling in pigment cell development. Our experimental studies assess in turn a series of hypotheses regarding the embryonic basis of the ectopic pigment cell phenotype, initially exploring disrupted biology of embryonic (direct developing) pigment cells, before coming to the conclusion that the phenotype must result from disruption of APSC biology. The ednraa mutant phenotype is restricted to supernumerary and ectopic pigment cells in the ventral trunk and anterior tail; this spatial localization had been perplexing, but identification of the gene as ednraa, which is consistently expressed strongly in developing blood vessels (but not in the trunk NC) from well before the onset of detectable ectopic cells, helps to explain the restricted phenotype of ednraa mutants. We find no evidence that neural crest migration is disrupted, since not only is blood vessel morphology normal, but neural crest cell migration and patterning is normal except for the ectopic pigment cells themselves, and pigment cells are abundant in the Ventral and Yolk Sac Stripes. Similarly, we disprove the 'mixed-fate' hypothesis by showing using TEM that the ectopic cells are indistinguishable from those melanocytes and iridophores in the Yolk Sac Stripe where these two cell-types are consistently found in tight apposition. Instead, we conclude that the close association of the two cell-types reflects the natural tendency for iridophores to adhere to melanocytes (as seen in all locations in the early larval pattern). Finally, we find no evidence that the ectopic cells derive from transfating of sympathetic neurons, since neurons are unaffected in ednraa mutants and even when sympathetic neuron specification is inhibited using a BMP inhibitor, ectopic pigment cell number is unchanged.
However, these studies were unable to address the possible role of glial or progenitor cells in the ventral trunk. We reasoned that the supernumerary and ectopic cells were likely to be associated with increased proliferation of a subset of neural crest cells. Strikingly, we found that enhanced proliferation of neural crest did characterise the ednraa mutants, but that this proliferation was specifically associated with neural crest cells in ventral regions i.e. near the dorsal aorta, at around the time (35 hpf) that ectopic pigment cells begin to be identifiable. This is before the sympathetic ganglia show detectable differentiation and suggested that a Endothelin receptor Aa affects pigment progenitor differentiation subset of NC-derived cells undergoes ectopic or precocious proliferation. A key insight into the identity of these cells came from the observation that both ErbB inhibitors rescued the homozygous phenotype. Inhibition of Erb signaling within a defined embryonic time window by either AG-1478 or PD158780 selectively affects the biology of NC-derived APSCs that give rise to adult and regenerative melanocytes [7,11]. We show that ednraa mutants are rescued by ErbB inhibitors in a dose-and time-dependent manner, and with a time-window overlapping that known to regulate APSC development. We conclude that the ectopic pigment cells in ednraa mutants result from precocious differentiation of pigment cells from APSCs, cells that would normally be quiescent until metamorphosis. An alternative model, that the ectopic cells result from simple overproliferation of embryonic melanoblasts and iridoblasts in the ventral region would be consistent with the localised enhanced proliferation, but in our pH3 labeling studies, labelled melanocytes are rare (1 in 20 WT embryos, and 0 in 20 pde mutants). Furthermore, this model would not explain the observations that ErbB inhibition rescues the pde mutant phenotype.
APSCs give rise to the numerous melanocyte and iridophores of the adult skin [12]. Our model would be consistent with the lack of ectopic xanthophores, since the evidence to date is that the majority of adult xanthophores derive from embryonic xanthophores and not from APSCs, although a small contribution from the stem cells is indicated by clonal analyses [12,46]. Finally, the localised increase in neural crest cell proliferation that we document is also consistent with the activation of otherwise quiescent stem cells. APSCs in zebrafish have been closely-linked to the developing peripheral nervous system. One intriguing aspect of our data is that this proliferation is exclusively localised to the ventral trunk, consistent with the idea that the blood vessels form a key aspect of the stem cell niche, but surprising in that to date APSCs have been exclusively associated with the DRGs [9,47]. We note that homozygous ednraa mutants are adult viable, but have no visible skin pigment pattern defect. Our data are consistent with a second source of APSCs, associated with the peripheral nervous system in the ventral trunk, likely the sympathetic ganglia. We propose a model in which these APSCs reside in a novel niche in close proximity to the medial blood vessels, which provide signals holding them in a quiescent state (Fig 7). In the ednraa mutants, these signals are reduced or absent, such that the APSCs become precociously activated. They then undergo proliferation and differentiate as melanocytes and iridophores. Many of these move into the Ventral Stripe location, but the excess cannot be accommodated in this stripe and remain ectopically located in the ventral trunk near the sympathetic ganglia. It is somewhat surprising that homozygous ednraa mutants show no disruption of the adult pigment pattern, but we propose several hypotheses, none mutually exclusive, to explain this observation: 1) although as we have shown APSCs are precociously activated in ednraa mutants, APSC renewal may be unaffected so that the stem cells are not exhausted; 2) ventral APSCs may primarily populate internal, rather than skin, pigment cell populations; 3) APSCs localised in DRGs may compensate for the effects on the more ventral APSCs. A key future test of this model will require direct evidence for the presence of a quiescent set of APSCs in the vicinity of the medial blood vessels.
Although novel in the context of APSCs, blood vessels form an important part of adult stem cell niches in other contexts, especially that of adult Neural Stem Cells (NSCs) [48]. Indeed in the case of NSCs in the subependymal zone of mouse, blood vessel-mediated signals play a role in maintaining quiescence [48], although other sources for these quiescence signals have also been identified in the neurepithelial component of this niche [49][50][51][52]. In these studies of NSCs, notch signaling has been shown to be important in maintenance of quiescence, so it will be fascinating to compare the molecular signals derived from the vasculature that regulate APSC quiescence in the zebrafish. Our discovery of a second niche for APSCs and identification of the key role for blood vessels in controlling their behaviour provides an entry point for uncovering an important but currently understudied aspect of zebrafish pigment pattern formation.
Our work may also have implications for human disease. For example, Phakomatosis pigmentovascularis (PPV) is a rare mosaic disorder defined as the simultaneous occurrence of a widespread vascular nevus and an extensive pigmentary nevus, and associated with activating mutations of Gα subunits of heterotrimeric G proteins [53]. It is currently unknown if there is a common progenitor that gains an oncogenic mutation and leads to both large nevi and vascular proliferations, or if changes in one cell type impact upon another. Our study provides evidence that melanocytes can expand/differentiate in association with a modified blood vessel niche, making it conceivable that, for instance, oncogenic activation in blood vessels might result in non-cell autonomous activation of melanocyte stem cells in the niche to generate a nevus. Reference zebrafish line. The standard mapping wild-type line WIK was crossed with mutant parade tj262 carriers (in AB wild-type background). This founder generation F0 was incrossed to produce heterozygous F1. Eight pairs of F1 fish were incrossed to produce the F2 generation. In total 1796 F2 embryos at 5 dpf were sorted into two groups, those that display the mutant phenotype and those with a normal wild-type pigment phenotype. These were transferred into 96-well plates where the extraction of genomic DNA was performed and stored. Additionally, genomic extracts of 10 homozygous parade tj262 and 10 wild-type sibling embryos of each of the 8 parent pairs were pooled for identification of new SSLP markers.

Fish husbandry
Mapping procedure. The pde locus was mapped to linkage group 1 (LN 1) by bulked segregant analysis using pooled genomic DNA of 48 wild-type siblings and 48 parade tj262 mutant embryos of the F2 generation obtained from F1 Pair 2. For fine mapping we designed new mapping primers using the Primer3 online software tool (http://frodo.wi.mit.edu/primer3/, default settings plus 1 primer pair per 300-400 bp [54]). Primer pairs were designed to generate PCR products of 300-400 bp, which if generating polymorphic PCR amplicons, were then tested on the 1796 single F2 embryos to determine the frequency of recombination.  (30 sec 94C , 30 sec 60˚C, 30 sec 72˚C); then 5 min 72˚C; then stored at 10˚C). The PCR products were then analysed by electrophoresis.

Morpholino injection
Custom morpholinos were purchased from Gene Tools LLC (Philomath, USA) and resuspended in autoclaved MilliQ water to a stock dilution of 20 μg/μl. Stock solutions were stored at -20˚C to prevent evaporation and heated at 70˚C for 5 min prior to dilution to eliminate precipitates. Wild-type embryos were injected into the yolk at the 1-cell stage with 5 ng/nl dilutions. Phenotypes were observed under the microscope at 3 dpf and 4 dpf. Volumes varied between 5 and 10 nl per embryo. Sequences of morpholino oligos can be found in S1 Table.

Whole mount in situ hybridization
Our protocol for whole-mount in situ hybridization is based on (Thisse et al., 1993). All solutions were prepared with DEPC-treated autoclaved MiliQ water or PBS. All probes for in situ hybridization were synthesized using the Dig RNA Labeling Kit (Boehringer). Depending on the orientation of the gene and choice of plasmid, we have chosen either T3, T4 or SP6 RNA polymerases for creating the antisense RNA probe.
Briefly, embryos between 5 and 120 hpf were euthanized with a lethal overdose of tricaine and fixed overnight (4% PFA in PBS). Embryos older than 18 hpf were dechorionated with forceps before fixation, younger embryos were dechorionated with flamed forceps after fixation. All following steps were carried out in 1.5 ml microfuge tubes with 1 ml liquid volumes; up to 40 embryos were processed per tube. The fixative was rinsed out with PBTween (DEPCtreated PBS plus 0.1% Tween). Embryos were then gradually dehydrated on ice with 25, 50, 75 and 100% methanol. For the ISH procedure, embryos were rehydrated on ice using 5 min washes with 75, 50, 25% methanol followed by 3 x 5 min washes with PBTween at room temperature. To improve permeability, embryos between 30-48 hpf were treated for 10 min with 0.01 mg/ml proteinase K/PBTween, embryos; older than 50 hpf were treated for up to 20 min. The proteinase K was removed by washing samples for 3 x 5 min with PBTween followed by a brief re-fixation step with 4% PFA for 20 min at room temperature. The fixative was then washed out with 3 x 5 min PBTween. To prepare the samples for optimal binding conditions, embryos were pre-hybridized with hybridization mix (formamide, 20 x SSC, heparine, tRNA, Tween20, citric acid; stored at -20˚C) in a water bath at 65˚C for 4 h. The hybridization of the RNA probe (diluted 1:400) was done over night at 65˚C in 200 μl hybridization mix. The samples were washed with decreasing concentrations of hybridization mix (100% hyb mix; 75% hyb mix/25% 2x SSCT; 50% hyb mix/50% 2x SSCT; 25% hyb mix/75% 2x SSCT; 100% 2x SSCT), each for 10 min, followed by 2 x 30 min washing with 0.2x SSCT. All steps were carried out in a water bath at 65˚C. Samples were then infiltrated for 3 x 5 min with MABT (from 5 x MAB stock solution plus 0.1% Tween) at room temperature. Samples were blocked for 3 h in MABTween with 5% sheep serum at room temperature under gentle shaking. Antibody binding reaction was performed over night at 4˚C using 200 μl of 1:5000 diluted anti-dioxigenin alkaline phosphatase conjugated antiserum in blocking solution. Samples were then washed for 6 x 15 min in MABT, then prepared for signal detection reaction by infiltrating for 3 x 5 min with NBT/BCIP buffer (100 mM Tris HCl pH 9.5, 50 mM MgCl2, 100 mM NaCl, 0.1% Tween). The embryos were transferred into 9-well glass dishes to observe the staining reaction under the dissecting microscope. 300 μl of BMPurple substrate solution were added per well and the development of the reaction regularly monitored. Signal development was stopped by washing with PBTween.

Chemical screening
Compounds used. The compounds used in this investigation originated from the 1280 compounds from the LOPAC library (Sigma LO1280), 80 compounds from a kinase inhibitor library (Biomol 2832) and 33 compounds from a phosphatase inhibitor library (Biomol 2834A). The compound libraries were stored at -80C and plates thawed at room temperature prior to use. For screening potential effects on developing zebrafish embryos compounds were diluted to 10uM in E3 embryo medium.
For drug treatments, six embryos were placed in each well of 24 well plates. Chemical treatments were prepared in 10uM doses in 1ml E3 medium. To ensure that compounds did not precipitate, the plates containing small molecule treatments were placed on The Belly Dancer (Sorval) for a minimum of 20 minutes. When embryos reached 8hpf E3 medium was removed and 1ml of E3 containing 10uM of compound was added to each well. Wells containing E3 only and 10uM DMSO were used in each chemical treatment as controls. Plates containing treated embryos were incubated at 28C until the embryos reached 5dpf.
Detailed observations were recorded daily and any dead embryos removed to avoid contamination of the medium. At day 5 embryos were anaesthetised with a small dose of tricaine to enable easy manipulation for accurate detailing of pigment phenotype. Pigment cell phenotypes and rescue of the parade mutant phenotype were recorded as well as additional phenotypes such as developmental abnormalities (e.g. shortening of the tail or cardiac edema).
Re-screening of important chemical 'hits'. Once molecules of specific interest to pigment cell development were identified the treatment was repeated on embryos at different developmental time points and across a concentration gradient in order to identify the developmental time point of greatest phenotypic significance and the optimal small molecule concentration.

Whole mount immunostaining
Antibody staining was largely performed as described in [55]. Prior to primary antibody incubation, embryos were permeabilized with 1 ml of 5 μg/ml of proteinase K/distilled water during 30 minutes at 37˚C, then rinsed with 5% of goat serum/distilled water at RT for 5 minutes and washed 3x1 hour with distilled water at RT. The antibodies used were Hu (1:500,) and Alexa Fluor 488 goat-mouse IgG (1:750, Molecular Probes, Cat. # A11001).

Image acquisition and processing
For in vivo microscopy, embryos were mounted at the Leica Fluo III or Zeiss Axiovision dissection microscope in 0.6% agarose or 3% methyl cellulose. For anaesthesia, 0.02% Tricaine was added freshly to the mounting medium. The embryos were transferred to a microscope slide or to a glass-bottomed Petri dish. For immunofluorescence, embryos were visualized with a Nikon Eclipse E800 or Zeise M2 Imager compound microscope under bright field, incident light or epifluorescence illumination, mainly using a 10x, 20x and 63x objective for magnification. Images were taken with a DS-U1 (Nikon) or Orca and color 546 camera (Zeiss). Confocal fluorescence imaging was performed on an inverted Zeiss LSM 510 Meta or LSM 880 confocal microscope with 20x and 40x objectives. Incident light was provided by installing an additional antler lamp (Leica) around the microscope to allow visualisation of iridophores. NIS-Elements D2.30, Zen blue, Fiji and/or Adobe Photoshop 7 were used to adjust white balance and exposure and to assemble the images to figures. Supporting information S1 Table. Morpholino sequences. A random sequence morpholino (cMO) provided by Gen Tools and a splice-blocking morpholino against ednrab (ednrab-MO) were used as control. ednraa was targeted with either a splice-blocking morpholino (ednraa -MO1) or a translationblocking ednraa morpholino (ednraa-MO2). Scale bar = 100 μm (A-C). (TIFF) S2 Table. Screening of small molecules in pde mutants. Name and known targets of small molecules that enhance (E) or rescue (R) the pde phenotype.