Interactions with Iridophores and the Tissue Environment Required for Patterning Melanophores and Xanthophores during Zebrafish Adult Pigment Stripe Formation

Skin pigment patterns of vertebrates are a classic system for understanding fundamental mechanisms of morphogenesis, differentiation, and pattern formation, and recent studies of zebrafish have started to elucidate the cellular interactions and molecular mechanisms underlying these processes. In this species, horizontal dark stripes of melanophores alternate with light interstripes of yellow or orange xanthophores and iridescent iridophores. We showed previously that the highly conserved zinc finger protein Basonuclin-2 (Bnc2) is required in the environment in which pigment cells reside to promote the development and maintenance of all three classes of pigment cells; bnc2 mutants lack body stripes and interstripes. Previous studies also revealed that interactions between melanophores and xanthophores are necessary for organizing stripes and interstripes. Here we show that bnc2 promotes melanophore and xanthophore development by regulating expression of the growth factors Kit ligand a (Kitlga) and Colony stimulating factor-1 (Csf1), respectively. Yet, we found that rescue of melanophores and xanthophores was insufficient for the recovery of stripes in the bnc2 mutant. We therefore asked whether bnc2-dependent iridophores might contribute to stripe and interstripe patterning as well. We found that iridophores themselves express Csf1, and by ablating iridophores in wild-type and mutant backgrounds, we showed that iridophores contribute to organizing both melanophores and xanthophores during the development of stripes and interstripes. Our results reveal an important role for the cellular environment in promoting adult pigment pattern formation and identify new components of a pigment-cell autonomous pattern-generating system likely to have broad implications for understanding how pigment patterns develop and evolve.


Introduction
The pigment patterns of teleost fishes are extraordinarily diverse and have important functions in mate choice, shoaling and predation avoidance [1][2][3][4]. These patterns result from the spatial arrangements of several classes of pigment cells including black melanophores that contain melanin, yellow or orange xanthophores with pteridines and carotenoids, and iridescent iridophores having purine-rich reflecting platelets [5][6][7]. In recent years, mechanisms underlying pigment pattern development, as well as pattern diversification among species, have started to be elucidated. Much of this work has used the zebrafish Danio rerio or its relatives [5,8].
In zebrafish, two distinct patterns develop over the life cycle. The first of these arises in embryos and persists through early larval stages [9][10][11][12][13][14]. Pigment cells of this early larval pattern develop directly from neural crest cells and generate stripes of melanophores at the edges of the myotomes and at the horizontal myoseptum; a few iridophores occur within these stripes whereas xanthophores are scattered widely over the body. The second, adult pigment pattern begins to develop during the larval-to-adult transformation and largely replaces the early larval pigment pattern [15]. Most cells comprising the adult pigment pattern differentiate from post-embryonic latent precursors, with the best studied of these cells, the melanophores, differentiating primarily between ,2-4 weeks post-fertilization [16][17][18][19]. By the end of this period a juvenile pigment pattern has developed consisting of two dark stripes of melanophores bordering a light interstripe of xanthophores and iridophores. As the fish grows, stripes and interstripes are added dorsally and ventrally. In the adult, some iridophores are also found within the melanophore stripes, including an ultrastructurally distinct class of these cells having large, rather than small, reflecting platelets [20]. Cells comprising the body stripes and interstripes are found within the hypodermis [20,21], between the epidermis and the myotome; pigment cells are also found in the scales, fins, and epidermis.
Previous studies showed that development of adult stripes and interstripes requires interactions between different pigment cell classes. For example, colony stimulating factor 1 receptor (csf1r) encodes a receptor tyrosine kinase required for xanthophore survival and migration [22]; csf1r mutants are deficient in xanthophores and also have disorganized melanophores. Yet stripes and interstripes could be restored in these fish by reintroducing xanthophores, either through cell transplantation or in the context of temperature-shift experiments using a temperature-sensitive csf1r allele [23,24]. These experiments suggested that xanthophores are required to organize melanophores into stripes. Subsequent studies identified additional short-range and long-range interactions between these cell types [25][26][27], the dynamics of which are consistent with a process of local self-activation and lateral inhibition, sometimes referred to as a ''Turing mechanism'' [28][29][30]. Such models often assume single, diffusible activators and inhibitors, though other cellular mechanisms can be accommodated as well. Indeed, theoretical and empirical analyses of melanophore and xanthophore behavior can recapitulate a wide range of pattern variants [31,32].
Despite the importance of interactions among pigment cells, the environment in which these cells reside also influences their development and patterning. Such effects are illustrated dramatically by mutants for basonuclin-2 (bnc2) [33], which encodes a highly conserved zinc finger protein that may function as a transcription factor or in RNA processing [34][35][36][37][38]. In contrast to the wild-type, bnc2 mutants exhibit far fewer hypodermal melanophores, xanthophores and iridophores and, consequently, lack body stripes and interstripes, though an apparently normal pigment pattern persists in the fins and in the scales ( Figure 1A, 1B). During the larval-to-adult transformation of bnc2 mutants, differentiated pigment cells of all three classes die at high frequency. Nevertheless, precursors of melanophores and xanthophores are abundant and widespread, suggesting late defects in their survival, terminal differentiation, or both. By contrast, iridophore precursors are markedly fewer, raising the possibility of additional defects in the earlier specification of this lineage. Genetic mosaic analyses showed that bnc2 acts non-autonomously to the melanophore lineage and likely the other pigment cell classes as well. Consistent with this interpretation, bnc2+ cells are initially found along horizontal and vertical myosepta but are later widely dispersed, both in the hypodermis and epidermis, a distribution resembling that of fibromodulin-expressing fibroblasts (LP and DP, unpublished data) but distinct from that of pigment cells and their precursors.
Here, we investigated the mechanisms by which bnc2 supports pigment cell development and the subsequent interactions between pigment cells during pigment pattern formation. We found that bnc2 mutants have reduced expression of Csf1r ligands and the ligand of the Kit receptor tyrosine kinase, Kitlga, which is required for the migration, survival and differentiation of teleost melanophores as well as mammalian melanocytes [9,[39][40][41][42][43][44]. Although restoring Csf1 and Kitlga in bnc2 mutants was sufficient to restore xanthophores and melanophores, these cells failed to organize into a normal striped pattern, indicating a requirement for additional factors or cell types. Because iridophores are deficient in bnc2 mutants, we asked whether these cells might normally contribute to the formation of stripes and interstripes. We found that iridophores are the first adult pigment cells to develop, that they express Csf1, and that xanthophores localize in association with them. To test if interstripe iridophores contribute to pattern development, we ablated these cells in wild-type and mutant larvae, resulting in perturbations to stripes and interstripes and confirming roles for iridophores in stripe and interstripe

Author Summary
Pigment patterns are some of the most distinctive, diverse and aesthetically pleasing traits of vertebrates. In turn, these patterns offer an outstanding opportunity to understand the mechanisms underlying the development of adult form and how such mechanisms change evolutionarily. Among the especially wide-ranging pigment patterns of teleost fishes, the most thoroughly studied example is the horizontal striping of zebrafish. In this species, stripes result from the precise arrangements of three classes of pigment cells: black melanophores, yellow or orange xanthophores and silvery iridophores. Previous studies showed that stripe formation requires interactions between melanophores and xanthophores. Nevertheless, roles for factors in the tissue environment experienced by pigment cells, as well as roles for iridophores in the pattern-forming process, have remained largely unexplored. Here we identify molecular mechanisms through which pigment cells are supported as the pattern develops. We further show that stripe development requires not only interactions between melanophores and xanthophores but iridophores as well, identifying a complex, pattern-generating system that may be applicable to understanding patterns and diversity across species. Our findings thus highlight the critical role of the ''canvas'' on which the pattern is painted, as well as the developmental artistry through which the ''paints'' are applied.
development. Together, our analyses suggest a model in which bnc2 supports the development and survival of melanophores, xanthophores and iridophores, and allows for subsequent interactions involving all three cell types. These results extend our understanding of environmental influences on pattern formation as well as pigment-cell autonomous patterning mechanisms.

Results
bnc2-dependent expression of Kitlga and Csf1 promotes melanophore and xanthophore development yet is insufficient for normal stripe patterning The death of melanophores and xanthophores in bnc2 mutants resembles the death of melanophores in mutants for kita, encoding a zebrafish Kit orthologue [41], and the death of xanthophores in csf1r mutants [24]. As kita and csf1r act autonomously to melanophore and xanthophore lineages [24,41], respectively, whereas bnc2 acts non-autonomously [33], we speculated that bnc2 might contribute to the development and maintenance of melanophores and xanthophores by promoting expression of the receptor ligands, Kitlga and Csf1. Consistent with this idea, quantitative RT-PCR of isolated body skins (with attached pigment cells) revealed significantly reduced expression of kitlga, as well as the two Csf1encoding loci, csf1a and csf1b, in bnc2 mutants compared to the wildtype ( Figure 1C). Quantitative RT-PCR comparisons of fins, in which melanophores and xanthophores persist in bnc2 mutants, failed to reveal differences in kitlga, csf1a or csf1b expression compared to the wild type (all P.0.5; data not shown).
If bnc2 acts through Kitlga and Csf1 to promote the development and survival of melanophores and xanthophores on the body, then restoring the expression of these ligands in the bnc2 mutant should restore melanophores and xanthophores and possibly a striped pattern. To test this idea, we generated transgenic lines using the ubiquitous, heat-shock inducible promoter of hsp70l to express Kitlga, Csf1a, or Csf1b individually, as well as Kitlga simultaneously with either Csf1a or Csf1b.
Restoration of Kitlga expression partially rescued melanophores in bnc2 mutants but did not restore stripes ( Figure 2A); this outcome was not unexpected given requirements for interactions between melanophores and xanthophores and the continued deficiency of the latter [23,24,26]. Restoration of Csf1a rescued xanthophores, and also increased melanophore numbers ( Figure 2B). Despite the abundance of both cell types, normal stripe patterns again failed to develop, with melanophores and xanthophores ranging widely over the flank ( Figure 2B). Similar outcomes were observed upon expressing Kitlga simultaneously with either Csf1a or Csf1b ( Figure 2C), for Csf1b alone, and in genetic mosaics combining cells from Kitlga and Csf1a transgenic embryos (data not shown). Together, these findings support the idea that bnc2-dependent expression of Kitlga, Csf1a and Csf1b promotes the development and survival of hypodermal body melanophores and xanthophores, yet the presence of these cell types alone is insufficient for organizing a normal pattern of body stripes and interstripes.
bnc2-dependent iridophores differentiate before melanophores and xanthophores and mark the prospective interstripe The failure to recover a normal pigment pattern in bnc2 mutants suggested that bnc2 might contribute to interstripe and stripe development through another factor or cell type. We reasoned that such a role could be fulfilled by iridophores, which are dramatically fewer in bnc2 mutants [33]. Consistent with this idea, residual xanthophores in the weak interstripe of bnc2 mutants were found almost exclusively within patches of residual iridophores (compare images of xanthophores and iridophores between wildtype and bnc2 mutant controls in Figure 2B).
If iridophores contribute to patterning interstripe and stripe development, these cells should develop prior to xanthophores and melanophores. We confirmed this by repeated imaging of wild-type and bnc2 mutant larvae, which showed that iridophores are the first adult pigment cell type to develop during the larval-to-adult transformation ( Figure 3). Iridophores developed as early as 4.5 mm standardized standard length (SSL) [45] and were restricted initially to the prospective interstripe region anteriorly, then developed in progressively more posterior regions. In contrast, the first melanophores and xanthophores differentiated later at ,6.0 SSL and ,6.5 SSL, respectively. In bnc2 mutants, xanthophore development was significantly delayed (F 1,5 = 383.8, P,0.001), typically occurring at ,7.5 SSL. The time and place of iridophore development relative to xanthophores and melanophores make iridophores a good candidate for contributing to interstripe location and orientation, and potentially later stripe patterning and maintenance.

Iridophores influence the localization of xanthophores and melanophores during interstripe and stripe development
To test whether iridophores contribute to specifying the location of interstripe xanthophores, we sought to ablate iridophores specifically and autonomously. To this end, we isolated a 3.2 kb fragment upstream from the transcriptional start site of the iridophore marker gene purine nucleoside phosphorylase 4a (pnp4a) [11,33] that drives iridophore-specific transgene expression ( Figure 4A). We used this element to express bacterial nitroreductase (NTR), which converts metronidazole (Mtz) into toxic metabolites that kill cells without bystander effects, even amongst cells that are coupled gap-junctionally [46][47][48][49][50]. We injected embryos with this pnp4a:NTR construct at the one-cell stage and then treated these genetically mosaic larvae with Mtz at stages when adult iridophores first develop in the prospective interstripe. Iridophores were lost over several days and reflecting-platelet containing fragments were identified in typical ''extrusion bodies'' [33,41,42] at the surface of the epidermis ( Figure 4B, 4C, 4D). In contrast to transient, F0injected transgenic larvae, it was not possible to ablate iridophores in stable pnp4a:NTR lines, presumably because of reduced transgene copy numbers. Thus, all subsequent analyses used genetically mosaic F0 larvae with repeated Mtz treatments.
Ablation of interstripe iridophores prior to xanthophore development resulted in fewer xanthophores in regions from which iridophores were lost ( Figure 5A, 5B, 5C), although both iridophores and xanthophores were recovered gradually during later development. Ablations of interstripe iridophores after xanthophores had developed typically did not affect xanthophore survival or patterning (data not shown).
Because interactions between xanthophores and melanophores contribute to organizing melanophore stripes, we anticipated that iridophore ablation and delayed xanthophore development could perturb melanophore patterning as well. Consistent with this prediction, we observed more melanophores in interstripe regions where iridophores (and xanthophores) had been depleted; nevertheless, melanophores occupying these regions were frequently found adjacent to residual or regenerated iridophores ( Figure 5D, 5E, 5F).

Iridophores express Csf1
Given the dependence of Csf1 expression ( Figure 1) and iridophore development on bnc2 ( Figure 3) [33], the requirement of xanthophores for signaling through Csf1r [22,24], and the dependence of xanthophores on iridophores (above), we hypothesized that iridophores supply a localized source of Csf1 to promote xanthophore development in the interstripe. We confirmed that csf1r is expressed by xanthophores during the larval-to-adult transformation using a transgenic reporter line derived from a bacterial artificial chromosome containing the csf1r Figure 2. Re-expression of Kitlga, Csf1a, and Csf1b in bnc2 mutants promoted melanophore and xanthophore development but was insufficient for stripe patterning. (A) Melanophore recovery following heat-shock induction of Tg(hsp70l:kitlga). Although Kitlga expression increased melanophore numbers in bnc2 mutant larvae, the restored melanophores failed to develop into stripes. Plots show means6SE with different letters above bars denoting means that differed significantly from one another in Tukey Kramer post hoc comparisons. All wild-type larvae are bnc2/+ siblings to bnc2 mutants. Sample sizes: bnc2/+, n = 10; bnc2, n = 10, bnc2/+ hsp70l:kitlga, n = 14; bnc2 hsp70l:kitlga, n = 14. (B) Xanthophore and melanophore recovery following heat-shock induction of Tg(hsp70l:csf1a). Upper plot, xanthophores were classed as either associated with iridophores (larger, lower segment of each bar), or not associated with iridophores (smaller, upper segment of each bar): total xanthophore numbers, including xanthophores not associated with iridophores were increased in bnc2 mutants by Csf1a expression. Lower plot indicates that melanophore numbers were increased as well. Images show xanthophores (yellow-orange cells) over iridophores (patches of grey cells in this illumination, denoted by blue arrowheads in the bnc2 mutant). Red dashed circle in bnc2 mutant +Csf1 panel shows a xanthophore that has developed at a distance from iridophores. Lower magnification images (bottom) show typical patterns and the absence of organized stripes in the bnc2 mutant after Csf1 expression, despite increased numbers of melanophores and xanthophores (compare to controls in A). Sample sizes: bnc2/+, n = 15; bnc2, n = 19, bnc2/+ hsp70l:csf1a, n = 19; bnc2 hsp70l:csf1a, n = 22. Results for Tg(hsp70l:csf1b) were equivalent (not shown; total sample size, N = 17). (C) Xanthophore and melanophore numbers were restored by heat shock induction of Tg(hsp70l:kitlga-csf1a) and Tg(hsp70l:kitlga-csf1b) yet stripes failed to form (total sample sizes, N = 7, 12, respectively). Scale bars: in (A) 500 mm for (A); in (B, upper) 80 mm for (B upper for images); in (B, lower) 500 mm for (B bottom 2 images); in (C) 500 mm for (C). doi:10.1371/journal.pgen.1003561.g002 locus ( Figure S1) [51]. To test if interstripe iridophores express csf1a and csf1b, we first used RT-PCR, which detected transcripts for both loci in iridophores isolated individually ( Figure 6A). By in situ hybridization, we found csf1a transcripts in hypodermal cells including cells likely to be iridophores according to their positions before and after in situ hybridization, and their locations at the base of the caudal fin and along the horizontal myoseptum, where iridophores develop ( Figure 6B, 6C, 6D). In cross-sections, csf1a transcript was detectable in the hypodermis where iridophores are found, as revealed by expression of pnp4a [11,33] ( Figure 6E, 6F). In contrast to wild-type larvae, far fewer cells stained for pnp4a and csf1a in the prospective interstripe region of bnc2 mutants. To further test the correspondence of csf1a expression and iridophores we examined the iridophore-free mutant of leucocyte tyrosine kinase (ltk), which is expressed by iridophores and required for their development [52]. ltk mutants lacked csf1a expression where iridophores are Figure 3. Interstripe xanthophores developed after iridophores in wild-type larvae and were further delayed in bnc2 mutants. Shown are a representative wild-type (bnc2/+) larva (A) and a sibling bnc2 mutant (B) imaged repeatedly over 27 d beginning at 6.0 SSL, just prior to the appearance of iridophores at the anteroposterior region imaged, dorsal to the anus. In both the wild-type and the bnc2 mutant iridophores started to appear by day 2 of imaging (blue arrowheads). Xanthophores started to differentiate by day 9 of imaging in wild-type; newly arising xanthophores are indicated by red dashed circles. In contrast, xanthophores did not appear until day 25 of imaging in the bnc2 mutant. As iridophores (and xanthophores) in the interstripe became more abundant, some early larval melanophores along the horizontal myoseptum disappeared from view (e.g., green arrows in A, d12 and d15). For easier visualization of melanophores and other cell type, fish were treated briefly with epinephrine immediately prior to imaging, which contracts melanosomes towards the cell body; the distribution of melanin thus indicates the centers of melanophores whereas processes extending out from the cell body are not visible. Bottom panels schematize the distribution of iridophores (light blue) and xanthophores (red) on the final day shown. Samples sizes for which complete image series were obtained were: bnc2, n = 4; bnc2/+, n = 6.  found normally in wild-type larvae ( Figure 6G, 6G9). We also observed strong, iridophore-independent expression of csf1a in fins of wild-type and ltk mutants ( Figure 6H, 6H9). csf1b was expressed similarly to csf1a by in situ hybridization and was also detectable in a population of dorsal hypodermal cells in both wild-type and bnc2 mutants. Together, these analyses indicate that iridophores express Csf1, and do so at a time and place that marks the prospective interstripe, though additional cell types express these ligands as well.
Localized expression of Csf1 promotes regionally specific xanthophore development If Csf1 expressed by early interstripe iridophores provides a spatial cue for xanthophores, we reasoned that ectopic expression of Csf1 should result in ectopic xanthophore development. To test this possibility we transplanted cells at the blastula stage from bnc2 mutant embryos transgenic for hsp70l:csf1a to bnc2/+ or bnc2 hosts and then induced mosaic expression of Csf1a by heat shock. We additionally expressed Csf1a in a temporally controlled manner within the myotome adjacent to the hypodermis: we identified a 2.2 kb region upstream of slow myosin heavy chain 1 (smyhc1) that drives expression in superficial slow muscle fibers and used this in a TetA-GBD [53] transgene to express Csf1a in these cells specifically during the larval-to-adult transformation. Using both paradigms to induce Csf1a outside of the developing interstripe, we observed corresponding patches of ectopic xanthophores in both bnc2/+ and bnc2 mutant siblings (Figure 7). These findings, and analyses of csf1a and csf1b expression, support a model in which interstripe iridophores provide a localized source of these ligands that contributes to specifying the position of interstripe xanthophores.

Iridophores influence melanophore patterning independently of xanthophores
Because xanthophores contribute to melanophore stripe organization [23,24,26], the mis-patterning of melanophores following iridophore ablation could simply reflect perturbations to the distribution of xanthophores. Yet, iridophores also might influence melanophores independently of xanthophores. To test this possibility, we ablated iridophores in csf1r mutant larvae. These mutants exhibit a few very lightly pigmented xanthophores limited to the immediate vicinity of the horizontal myoseptum but lack xanthophores in the more ventral interstripe region and elsewhere ( Figure S2A, S2B) [22,23,54]. Although stripes in csf1r mutants are disorganized and melanophores initially differentiate more widely over the flank than in wild-type larvae [22], quantitative analyses of final melanophore distributions in unmanipulated csf1r mutants revealed a residual stripe pattern in which melanophores tended to be dorsal or ventral to where the interstripe would form normally ( Figure 8A, 8C, 8E). At later stages, melanophores tended to be situated close to, but not directly over, iridophores, and iridophores were more widely distributed than in the wildtype ( Figure S2C). In csf1r mutants in which iridophores had been ablated, however, melanophores were more likely to occur in the middle of the flank where iridophores had been lost ( Figure 8B, 8D, 8E). Repeated imaging of individual larvae showed that melanophores both migrated to, and differentiated in, regions where iridophores had been ablated; once in these regions, melanophores often settled adjacent to residual iridophores ( Figure 8F, 8G, 8H). Together, these observations suggest that iridophores can influence melanophore patterning independently of interactions between xanthophores and melanophores. Although kitlga is a good candidate for contributing to an interaction between iridophores and melanophores, kitlga expression by that were either uninjected or injected with pnp4a:NTR plasmid. Xanthophore numbers did not differ between groups at the onset of the experiment but iridophore-ablated individuals showed an increasingly severe xanthophore deficiency compared to uninjected larvae as the experiment proceeded (genotype6day interaction, F 1,10 = 2.7, P,0.005; initial sample sizes: uninjected, n = 13; pnp4:NTR, n = 13). During later development, new xanthophores ultimately developed more broadly over the flank and in association with regenerating iridophores; iridophore ablations after xanthophores had differentiated typically did not affect these cells (not shown). (D-F) Examples of larvae (9.5 SSL) exhibiting melanophore patterning defects following earlier iridophore ablations (started at 6.0 SSL). Melanophores have colonized regions from which iridophores were ablated, though a few regenerative or persisting iridophore remained. In the lighting used here, iridophores are blue or gold iridescent. (D) Melanophores occupy a region from which iridophores were ablated (residual or regenerated iridophores outlined by dashed yellow lines). Green arrowhead, one of several melanophores localized adjacent to remaining iridophores. Fish shown in A, C and D were treated with epinephrine prior to imaging. (E) Melanophore stripes are broken at site of iridophore ablation and melanophores appear to ''wrap around'' residual interstripe iridophores on either side of the ablation. (F) In another individual, melanophores stripes are constricted where iridophores have been ablated (arrow). Close-up in F9. Fish in E and F were not treated with epinephrine, so that melanin reveals peripheral processes of melanophores. Most small melanophores in dorsal regions are associated with developing scales and will not contribute to the stripe pattern [45].  Melanophore and xanthophore patterning are defective in additional iridophore-deficient mutant backgrounds To further test inferences from cell ablation studies, we examined melanophore and xanthophore patterning in additional mutant backgrounds, ltk, described above, and endothelin receptor b1a (ednrb1a). ltk mutants lack iridophores and repeated imaging of individual larvae revealed increased frequencies of melanophore death, as well as delays in xanthophore differentiation by an average of 661 d (paired t = 6, P,0.05) as compared to stagematched wild-type siblings ( Figure 9A). When xanthophores did develop they did so widely over the flank, rather than being restricted to the interstripe region ( Figure 9B). ednrb1a is expressed in precursors to all three pigment cell classes and is maintained at high levels in iridophores [55]. ednrb1a mutants exhibit severely reduced numbers of iridophores ( Figure 9D). Although adults exhibit a dorsal melanophore stripe and ventral melanophore spots, examination of pattern development in daily image series showed that ventral spots arise further ventrally than the normal location of the ventral stripe, being localized instead to the site of the second ventral interstripe ( Figure 9C). Together these observations indicate that melano-phore and xanthophore patterning are disrupted in two additional iridophore-deficient mutants, consistent with roles for iridophores in promoting normal stripe and interstripe development.

Discussion
Our analyses together with previous studies suggest a model for adult body stripe and interstripe development in zebrafish (Figure 10). At the onset of adult pigment pattern formation, iridophores begin to differentiate in the prospective interstripe region and the expansion of this population depends on bnc2. Melanophores and xanthophores then start to differentiate, supported by bnc2-dependent Kitlga and Csf1, respectively. Melanophores avoid settling in the interstripe region in part owing to short-range inhibitory interactions with iridophores, whereas xanthophores differentiate specifically in the interstripe, receiving Csf1 both from the skin and from iridophores already there. Subsequently, interactions among all three classes of pigment cells contribute to organizing the definitive pattern of stripes and interstripes.
Previous analyses of adult pigment pattern formation in zebrafish highlighted the importance of interactions between melanophores and xanthophores [23,24] and a combination of short-range and long-range interactions between these cell types is consistent with a Turing mechanism of pattern formation or maintenance [26,27]. Nevertheless, one might anticipate roles for additional cues in specifying stripe position or orientation. For example in studies using a temperature-sensitive allele of csf1r, the orientation of stripes in the fin was randomized when xanthophores developed only at late stages [24], suggesting that cues required for orienting stripes during development either were not present, or not recognized, at later stages. Similarly in this study, the recovery of widespread melanophores and xanthophores in bnc2 mutants was insufficient for stripe formation on the body. This observation suggested that additional factors specify the location and orientation of stripes and interstripes, and support melanophores and xanthophores during pattern formation.
This study indicates that iridophores contribute to adult pigment pattern formation, with several lines of evidence implicating interstripe iridophores in the development of interstripe xanthophores. First, image analyses showed that iridophores are the first adult pigment cells to develop, and do so at the interstripe. Second, Csf1r signaling is necessary for xanthophore development [22,24] and we found that interstripe iridophores express csf1a and csf1b whereas xanthophores express csf1r. Third, misexpressing Csf1 resulted in the development of ectopic xanthophores, indicating this pathway can promote xanthophore localization. Fourth, xanthophore development was delayed when iridophores were ablated transgenically and in the bnc2 mutant, which has a severe iridophore deficiency. Fifth, the few xanthophores that do develop in bnc2 mutants were associated exclusively with the few residual iridophores. From these observations we suggest that iridophores promote the timely appearance of xanthophores within the interstripe ( Figure 10C, interaction #1), thereby positioning xanthophores to interact with melanophores during the subsequent patterning of dorsal and ventral stripes.
Our finding that xanthophore development is delayed in iridophore-deficient ltk mutants is consistent with these inferences. That xanthophores ultimately differentiated in these mutants presumably reflects the persistence of iridophore-independent sources of Csf1 that are not present or not sufficient for xanthophore development in bnc2 mutants. Interestingly, when xanthophores did develop in ltk mutants, they did so more widely over the flank than in the wild-type, in which xanthophores were restricted to the interstripe. A similar restriction of xanthophores to the vicinity of interstripe iridophores has been reported for mitfa mutants, which retain iridophores yet lack melanophores [23]. These observations raise the possibility that iridophores both promote xanthophore development at short-range and repress xanthophore development at long-range ( Figure 10C, interaction #2), though we cannot yet exclude other explanations for this phenomenon.
Our analyses also suggest roles for iridophores in melanophore development and patterning. Our finding that melanophores localized to regions from which iridophores had been ablated could reflect a delay in the development of xanthophores and the inhibitory effects that xanthophores have on melanophore localization [26]. Although this may have contributed to the mis-patterning of melanophores, our finding that iridophore ablation perturbs melanophore patterning even in xanthophoredeficient csf1r mutants suggests that iridophores also influence melanophores independently of xanthophores. Melanophores frequently migrated to, or differentiated within, iridophore-free sites; melanophore centers (as indicated by melanosomes contracted by epinephrine) rarely overlapped with iridophores, yet melanophores often settled adjacent to iridophores. These observations are consistent with a very short-range inhibitory effect of iridophores on melanophore localization ( Figure 10C, interaction #3), as might occur if the two cell types compete for a common substrate, as well as a longer-range attractive or stimulatory effect of iridophores on melanophores ( Figure 10C, interaction #4). Our findings of increased melanophore death in ltk mutants, and the increased death of mitfa:GFP+ cells [16] as well as mis-patterning of melanophores in ednrb1a mutants, are likewise consistent with a model in which iridophores influence melanophores. Finally, we note that our examination of csf1r mutants revealed iridophores to be more widespread in this xanthophore-deficient background than in the wild-type, raising the possibility that xanthophores interact reciprocally with iridophores as well as melanophores. A definitive test of the interactions hypothesized in Figure 10C will await the elucidation of molecular mechanisms underlying these various patternforming events.
In addition to interactions among pigment cells, our study provides new insights into roles for bnc2 in pigment pattern development. Expression analyses and rescue experiments suggested that bnc2 promotes the development and survival of melanophores and xanthophores by ensuring adequate expression of kitlga, csf1a, and csf1b ( Figure 10B). These observations are consistent with previously known roles for Kit ligand [16,[40][41][42][43][44][56][57][58] and Csf1 [22][23][24]59], and identify a novel role for Iridophore ablation at the mid-trunk region (dashed yellow lines in D) likewise resulted in increased numbers of melanophores compared to stagematched control (C)(shown here at ,10.2 SSL). Note that some iridophores have regenerated within previously ablated regions and that melanophores are present at the left edge of the ablated region, adjacent to remaining interstripe iridophores. Larvae in these images were not treated with epinephrine. (E) Quantification of melanophore distributions within dorsal-ventral regions of the flank for larvae that were uninjected but treated with Mtz (left) and for regions of injected, Mtz-treated larvae from which iridophores were unablated (middle) or ablated (right). Plots show means6SE within each region. Asterisk denotes the residual interstripe in csf1r mutants, where melanophore numbers differed significantly between unablated and ablated regions (paired t = 5.6, d.f. = 2, P,0.05). (F,G) Details showing melanophore behaviors in an uninjected control larva (F) and an injected larva (G) in the region of iridophore ablation. Day 0 panels show initial distribution of iridophores and melanophores, prior to Mtz treatment (7.0 SSL). Following iridophore ablation (G), some melanophores moved short distances ventrally (red arrows at d0 and d2 show starting and stopping positions of two melanophores). Melanophores also differentiated within the ablated region (dashed red circles in G, d2); dashed orange circle in G, d7 shows a lightly melanized cell just beneath the surface of the myotome that emerges within the skin by d8. In unablated individuals (F), melanophores typically differentiated further ventrally at sites lacking iridophores (an exception is the left-most melanophore that appeared at d7). Also see Figure S3C. (H) Detail from another individual showing a lightly melanized cell initially near an iridophore that was ablated (blue arrowhead); the melanophore subsequently translocated to settle adjacent to another iridophore. All larvae in F-H were treated with epinephrine. (  . Melanophore and xanthophore development is disrupted in additional iridophore-deficient mutants. (A) Comparison of xanthophore and melanophore development in wild-type and ltk mutants. Shown are details at the horizontal myoseptum from larger images of representative wild-type (ltk/+) and ltk mutant, stage-matched siblings imaged daily (beginning at 6 SSL). In the wild-type, nearly all melanophores persisted through the image series. A xanthophore had already developed at the onset of imaging (day 0, red dashed circle), and additional xanthophores differentiated shortly thereafter. In the ltk mutant, however, melanophores were frequently lost between days (green arrowheads) and melanin-containing debris and extrusion bodies were often apparent (green arrows). Unlike the wild-type, no xanthophores differentiated until day 5 of imaging. (B) During later development (9.6 SSL), xanthophores were confined principally to the interstripe region of the wild-type whereas xanthophore developed widely over the flank in the ltk mutant. The horizontal myoseptum lies at the lower edge of both images. Lower panels show positions of xanthophores in red. (C) Comparison of wild-type and ednrb1a mutant. Shown are ventral flanks of representative stage-matched, sibling wild-type (ednrb1a/+) and ednrb1a mutant larvae imaged daily (8.8-10 SSL). At the onset of imaging, wild-type melanophores are largely absent from a region where the second interstripe will form by day 7 of imaging (blue bars). In ednrb1a mutants, however, melanophores are relatively uniformly distributed in this region at the onset of imaging, and, by day 7 of imaging, formed clusters where the second interstripe would normally form (green Bnc2 in regulating the expression of these genes. It will be interesting to learn if Bnc2 has similar functions in providing trophic support to other stem-cell derived lineages as this locus is also expressed in the ovary, central nervous system, and skeleton [33]. Indeed, zebrafish bnc2 mutant females are infertile and human BNC2 variants are associated with ovarian cancer predisposition [60]; potential defects in other systems have yet to be ascertained. bars). Images shown were rescaled to control for growth. (D) Closeups showing reduced iridophores in ednrb1a mutant compared to wild-type (9.0 SSL) as well as wider distribution of xanthophores. Fish in A, B and D were treated briefly with epinephrine prior to imaging. Sample sizes for which complete image series were obtained were: ltk, n = 6; ltk/+, n = 5; ednrb1a, n = 4; ednrb1a/+, n = 5. Scale bars: in (A, d0) 60 mm for (A); in (B) 100 mm for (B); in (C, d0) 200 mm for (C, d0); in (D) 100 mm for (D). doi:10.1371/journal.pgen.1003561.g009 Figure 10. Summary of results and model for stripe and interstripe patterning in zebrafish. (A) Development of pigment pattern phenotypes in wild-type, bnc2 mutants, iridophore-ablated larvae (pnp4a:NTR), and ltk mutants. Blue circles, iridophores; orange circles, xanthophores; grey and black circles, melanophores. In bnc2 mutants, there are fewer iridophores and increased rate of cell death (open circles) amongst all three pigment cell classes. Xanthophores are restricted to the vicinity of iridophores. In iridophore-ablated larvae, melanophores localize where iridophores have been lost but also organize adjacent to residual iridophore patches. In ltk mutants, iridophores are missing, melanophores tend to die, and xanthophores develop both later and over a wider area than in wild-type larvae. (B) An unknown, bnc2-dependent factor expands an initial population of iridophores, whereas bnc2-dependent Kitlga and Csf1 support the expansion of melanophore and xanthophore populations. (C) Hypothesized interactions amongst pigment cell classes. Black lines, suggested by this study; grey lines, suggested previously [23,24,26]. Solid lines, short-range interactions; dotted lines; longer-range interactions. Iridophores promote xanthophore localization to the prospective interstripe at shortrange through Csf1 (interaction #1), and are hypothesized to repress xanthophore development at a distance (#2). Iridophores also affect melanophores, which are inhibited from localizing at sites already occupied by iridophores (#3), and instead differentiate or localize nearby (#4). Once melanophores and xanthophores have developed, these cells exhibit mutual, short-range inhibitory interactions that affect localization, survival or both (#5, #6); xanthophores also promote melanophore survival at a distance (#7) and melanophores repress the development of other melanophores at a distance (#8) [26]. See main text for additional details. doi:10.1371/journal.pgen.1003561.g010 At least two aspects of bnc2 function remain ambiguous. First, although it is clear that bnc2-dependent iridophores provide one source of Csf1 to developing xanthophores, csf1a and csf1b are also expressed more broadly, whereas kitlga is expressed in skin, and it has not yet been possible to establish whether bnc2+ cells express these factors themselves, or induce other cells to do so. The development of transgenic reporters for all of these loci will address this issue definitively. Second, iridophores are the most severely affected cell type in bnc2 mutants [33] yet the mechanism by which bnc2 promotes iridophore development remains unknown. The distribution of bnc2+ cells [33] does not perfectly mark the prospective interstripe so it seems likely that other factors specify where iridophores will develop, with bnc2+ cells promoting the expansion of the interstripe iridophore population once it has been established. It will be interesting to learn which bnc2dependent and bnc2-independent factors are required for iridophore development and whether manipulation of these factors is sufficient to alter the location or orientation of stripes and interstripes. Finally, the continued high expression of Csf1 and Kitlga in the fins of bnc2 mutants seems likely to explain the persistence of stripes and interstripes at this location; why fin melanophores and xanthophores can organize into stripes in the absence of bnc2 activity, whereas body melanophores cannot awaits further investigation.
Several studies have highlighted the pigment-cell autonomous nature of pattern-generating mechanisms in zebrafish. Our study suggests two extensions to this paradigm. First, environmental factors are required to support pigment cells during pattern formation and are likely to provide cues that bias the initial development of pigment cells (e.g., the first interstripe iridophores) to specific regions, thereby influencing the subsequent locations and orientations of stripes and interstripes. Second, interactions among pigment cells appear to involve all three major classes. Analyses presented here support a model in which iridophores exert positive and negative effects on both xanthophores and melanophores and we can imagine that additional interactions will be identified as well. In this regard, it will be interesting to learn whether pigment pattern formation occurs through additional dimensions of Turing-like interactions. Because interactions amongst more than two cell types are not readily accommodated by existing frameworks for describing local self-activation with lateral inhibition mathematically, additional theoretical effort will be needed to capture biological complexity involving multiple cell types and multiple, reciprocal interactions. Finally, we envisage that evolutionary changes in factors both non-autonomous and autonomous to pigment cell lineages are likely to have contributed to the extraordinary diversification of pigment patterns among ectothermic vertebrates; it will be exciting to discover what general themes emerge as mechanisms of pigment pattern formation are elucidated in other species.

Ethics statement
All animal studies were conducted in accordance with regulations of the University of Washington and the United States Department of Health and Human Services, and received the approval of the Institutional Animal Care and Use Committee of the University of Washington.
Fish stocks were reared in standard conditions at 28.5uC 14L:10D. For transgene inductions using hsp70l promoters, fish were heat-shocked at 38uC twice daily for 1 h beginning when fish had reached 8.5 SSL and extending for period 2-4 weeks. For fish injected with plasmid smyhc1:TetGBD-TREtightBactinTRX:nls-Venus-V2a-csf1a, induction with dexamethasone and doxycycline followed [53]. For ablating iridophores in larvae mosaic for plasmid pnp4a:nlsVenus-V2a-NTR, larvae were incubated overnight in 10 mM Mtz. For time series of individual ablations in wild-type and csf1r mutants, larvae were allowed to recover one night prior to imaging. Fish were then imaged a second day and treated with Mtz again that evening. Repeated treatments were required to repress iridophore regeneration, though in many cases, iridophores eventually recovered. Treatments alternating every third night were also administered to batches of wild-type or csf1r mutant larvae mosaic for pnp4a:nlsVenus-V2a-NTR that were later assessed for iridophore ablations.

In situ hybridization
Characterization of mRNA transcript distributions in whole mount and transverse vibratome sections followed [64]. For comparing distributions of csf1a transcripts and iridophores, larvae were imaged prior to fixation, processed individually and then the corresponding regions re-imaged after color development.

RT-PCR and quantitative RT-PCR
For quantitative RT-PCR, single skins were collected from ,9.0 SSL bnc2 or bnc2/+ larvae and placed in either Trizol Reagent (Invitrogen) or RNAlater (Ambion). RNA was isolated using either Trizol or RNaqueous Microkit (Ambion), followed by LiCl precipitation. cDNA was synthesized with either Superscript III First-Strand Synthesis (Invitrogen) or iScript cDNA Synthesis Kit (Bio-Rad). Quantitative RT-PCRs were performed and analyzed with a StepOnePlus System (Life Technologies) using a Custom Taqman Gene Expression Assay for kitlga (Life Technologies) and the following Taqman Gene Expression Assays (Life Technologies): csf1a, Dr03432536_m1; csf1b, Dr03110811_m1; gapdh, Dr03436842_m1.
For RT-PCR of isolated iridophores, 10-14 SSL larvae were euthanized and 3 skins placed in PBS. Tissue was briefly vortexed to remove scales, then centrifuged and washed again in PBS. Skins were incubated 10 min at 37uC in 0.25% trypsin-EDTA (Invitrogen). Trypsin was removed and tissue incubated 10 min at 37uC in trypsin-inhibitor (Sigma T6414) with 3 mg/ml collagenase, and 2 ml RNase-free DNase I (Thermo Scientific), followed by 3 h at 28 C in a Benchmark Multi-Therm Shaker set to 800 rpm. Cells were washed in PBS and filtered through a 40 mm cell strainer (BD Falcon). Cell mixtures were placed on a glass bottom dish and examined on a Zeiss Observer inverted compound microscope. Individual iridophores were picked using a pulled capillary and Narishige 1M 9B microinjector then expelled directly into resuspension buffer from the Superscript III Cells Direct cDNA Synthesis Kit (Invitrogen). cDNA was synthesized from approximately 50 cells per sample and RT-PCR performed with the following primers designed to span introns (forward, reverse): actb1, ACTGGGATGACATGGAGAAGAT, GTGTT GAAGGTCTCGAACATGA; pnp4a, GAAAAGTTTGGTC-CACGATTTC, TACTCATTCCAACTGCATCCAC; csf1a, TA CACCTTCACAGAGCGTCAGA, CTTCGTTGGACTGTC CTCAATC; csf1b, AACACCCCTGTTAACTGGACCT, GAG GCAGTAGGCAGTGAGAAGA.

Imaging and quantitative analyses
For time-course imaging of interstripe development, fish from bnc2/+, ltk/+, or ednrb1a/+ backcrosses were reared individually and imaged daily on a Zeiss Observer inverted compound microscope or an Olympus SZX-12 stereomicroscope, using Zeiss Axiocam HR cameras and Axiovision software. Individuals from bnc2/+ backcrosses were genotyped retrospectively for the bnc2 utr16e1 lesion [33].
For transgenic rescue experiments of bnc2 mutant melanophores and xanthophores, larvae were viewed and imaged as described above. For assessing melanophore numbers, all melanophores were counted ventral to the horizontal myoseptum in a region bounded by the anterior margin of the dorsal fin and the posterior margin of the anal fin. For assessing xanthophore numbers and localization, xanthophore were counted at three separate locations along the anterior to posterior axis (posterior swim bladder, anus, center of anal fin) within the interstripe region (as marked by iridophores).
To quantify melanophore dorsal-ventral location in csf1r mutants mosaic for pnp4a:nlsVenus-V2a-NTR and uninjected controls, we measured the distance of each melanophore from the dorsal and ventral margins of the myotomes, then divided dorsal length by total distance. Positions were determined for all melanophores between the anterior of the dorsal fin and posterior of the anal fin. Regions were considered ablated when they lacked most iridophores.
For presentation, images were color-balanced and in some cases adjusted for color saturation to assist in visualizing xanthophores.
All statistical analyses were performed using JMP 8.0.2 (SAS Institute, Cary, NC). For analyses of xanthophore numbers, counts were square-root transformed prior to analysis to correct for unequal variances across groups. Figure S1 Expression of csf1r reporter by xanthophores during the larval-to-adult transformation. Interstripe xanthophores autofluoresce in the same channel as GFP and coexpress an mCherry csf1r reporter (arrowhead). Individual shown is Tg(csf1r: Gal4.VP16) i186 injected with a plasmid containing 4xUAS-mCherry, resulting in mosaic mCherry expression. In an individual in which iridophores have been ablated and partially recovered, residual xanthophores remained confined to the horizontal myoseptum and did not enter the region from which iridophores had been lost (likely owing to csf1r requirements for xanthophore migration [22,54]). Shown here is the same region of the larva shown in main text Figure 8G