The adult zebrafish (Danio rerio) pigment pattern includes yellow xanthophores, black melanophores, at least three types of iridescent iridophores, and two types of white cells (melanoleucophores and xantholeucophores) (Hirata et al., 2003; Lewis et al., 2019; Patterson and Parichy, 2019; Gur et al., 2020). Zebrafish does not have red erythrophores. Because erythrophores occur in many other species of teleosts, we surveyed the distribution of these cells across the Danio genus more broadly. Of 17 Danio species assessed, 14 had erythrophores indicating this cell type is common and most likely was present in the common ancestor of all Danio species (Figure 1B–D).

We focused on D. albolineatus since erythrophores are abundant in this species and are separated spatially from other pigment cells, an arrangement likely to facilitate analysis (Goodrich and Greene, 1959). In the anal fin of adults, red erythrophores were located proximally and were separated from the more distal yellow xanthophores by a narrow stripe of melanophores (Figure 2A). Although erythrophores were present in both sexes, the cells were more deeply and consistently red in males than females and we therefore focused on males at stages when sexes were distinguishable (Figure 2—figure supplement 1A). Male fish older than 1 year often lacked fin stripe melanophores, indicating that some pattern remodeling continues even after sexual maturation (Figure 2—figure supplement 1B). As compared to xanthophores, erythrophores occur at lower densities and were more likely to be binucleated (Figure 2B, middle and right panels; Figure 2—figure supplement 1C and D), a characteristic associated with a mature state of differentiation in stripe melanophores of zebrafish (Saunders et al., 2019).

Because subtle differences in color can be difficult to discern, we sought metrics to describe mature and developing cells, under brightfield and fluorescent illumination in which these cells were distinguishable as well: xanthophores displayed green autofluorescence upon excitation with blue-green light (488 nm) owing to the presence of yellow carotenoids (Granneman et al., 2017; Saunders et al., 2019), whereas erythrophores showed only weak autofluorescence at this wavelength but much stronger red autofluorescence upon exposure to green-yellow light (561 nm)(Figure 2C). We then compared hue in brightfield illumination with relative red:green autofluorescence, confirming the differences between erythrophores and xanthophores (Figure 2D and E).

To understand the anatomical context of erythrophore development, we imaged fish during the larva-to-adult transformation. The first pigmented cells in the anal fin were lightly melanized melanophores. Subsequently, orange xanthophore-like cells appear that were pale and had smaller areas of visible pigment than mature erythrophores and xanthophores (Figure 2B, left panel; Figure 2—figure supplement 2). These early orange cells autofluoresced in both red and green channels, consistent with their appearance in brightfield (Figure 2F). During later development, however, cells in proximal regions were increasingly red whereas cells in distal regions were increasingly yellow. Cell densities gradually diverged between proximal and distal regions as well (Figure 2F; Figure 2—figure supplement 1D).

As a first step in dissecting lineage relationships of erythrophores and xanthophores, we sought to determine whether these cells arise from a common early progenitor. Since red and yellow colors were likely carotenoid-based, we reasoned that lineage relationships should be revealed by clones of cells in fish mosaic for scavenger receptor b1 (scarb1), which is essential for carotenoid accumulation in avian integument and zebrafish xanthophores (Toews et al., 2017; Toomey et al., 2017; Saunders et al., 2019). If erythrophores and xanthophores share a lineage, then rare wild-type clones should contain both red and yellow cells in an otherwise colorless background. If erythrophores and xanthophores have distinct lineage origins, however, wild-type clones should often contain only red cells or only yellow cells. In D. albolineatus injected with high efficiency AltR CRISPR/Cas9 reagents targeting scarb1, wild-type clones most often contained both red cells and yellow cells (Figure 3A).

We further assessed relationships by labeling individual clonal lineages by tol2 transgenesis. We found that orange cells of larvae and both erythrophores and xanthophores of adults expressed transgenes driven by regulatory elements of aldehyde oxidase 5 (aox5) isolated from zebrafish. aox5 functions in the synthesis of pteridines present in xanthophores and erythrophores (see below) and is expressed by xanthophores and their specified precursors in zebrafish (Parichy et al., 2000; McMenamin et al., 2014). When we injected an aox5 reporter transgene at limiting dilutions to express membrane-targeted EGFP, labeled cells were restricted to narrow regions along the anterior–posterior axis, consistent with derivation from single clones observed in other contexts (Tu and Johnson, 2010; Singh et al., 2014; Spiewak et al., 2018). Such cells occurred on the body and fin, and in these presumptive clones, erythrophores and xanthophores were almost always co-labeled (Figure 3B; Figure 3—figure supplement 1), consistent with a common progenitor for erythrophores and xanthophores.

A common progenitor could be specified for erythrophore or xanthophore fates either before or after colonizing the fin. To distinguish between these possibilities we used aox5 reporter expression as an indicator of specification and a nuclear localizing photoconvertible (green→red) fluorophore, nucEosFP, to determine whether cells already expressing this marker transit from body to fin. We generated a transgenic line, Tg(aox5:nucEosFP)^vp37albTg, which allowed us to photoconvert all aox5:nucEosFP+ cells on the body prior to anal fin development (6.5 mm SL). We then assessed the distribution of converted and unconverted nucEosFP 4 days later, when the anal fin had started to form (7.5 mm SL). Because aox5 expression persists once initiated, cells photoconverted at one stage will later have converted fluorophore (displayed in magenta), as well as new, unconverted fluorophore (green), so nuclei will appear white; cells that initiate aox5 expression only after photoconversion will have only unconverted fluorophore and nuclei that are green.

We found that cells on the body had converted and unconverted fluorophore, whereas cells in the fin had only unconverted fluorophore, consistent with initiation of aox5 expression only after progenitors had colonized the fin (Figure 3C). Because it remained possible that some cells had migrated from body to fin and proliferated so extensively that signal of converted nucEosFP was lost by dilution, we repeated these analyses but assessed distributions of cells 1 d after photoconversion; we then photoconverted (or reconverted) all cells on body and fin and repeated this process on successive days. Such labeling failed to reveal cells that translocated from body to fin, although it did reveal numerous cells that acquired aox5 expression when already in the fin (Figure 3D). Together these observations suggest that progenitors migrate to the fin, become specified for erythrophore or xanthophore lineages within the fin, and then contribute to both populations as they proliferate to populate the proximal–distal axis during fin outgrowth.

Clones identified by scarb1 activity or aox5 transgene expression (Figure 3A and B) likely represented progenitors segregated from other lineages during early development when injected Cas9 is active and transgene integration occurs (e.g., Tryon et al., 2011); these clones presumably also included melanophores or other cell types not revealed by these markers (Tu and Johnson, 2010; Tu and Johnson, 2011; Lewis et al., 2019). We therefore asked whether progeny of such clones that had already colonized the fin were restricted to either erythrophore or xanthophore fates by photoconverting individual nucEosFP+ cells at early stages of fin development (7.0–8.5 mm SL) and then assessing phenotypes of resulting clones 30–36 days later (15.0 mm SL). At 7.0 mm SL, only unpigmented nucEosFP+ cells were present (Figure 4A, top). Preliminary observations indicated that proximally located cells tended to remain in the proximal region where erythrophores develop, so we photoconverted cells in distal regions that might become more broadly distributed. Resulting clones consisted of erythrophores if daughter cells remained relatively proximal as the fin grew out, or both erythrophores and xanthophores if daughter cells became distributed across the proximodistal axis (Figure 4B and C; Figure 4—figure supplement 1A). At 7.5 mm, many nucEosFP+ cells had acquired a pale orange color (Figure 4A, bottom) and so we asked whether these cells had become fate-restricted with the onset of pigmentation. Similar to unpigmented cells, however, initially proximal orange cells generated only erythrophores, whereas initially distal orange cells could generate clones of only erythrophores, both erythrophores and xanthophores, or only xanthophores, depending on where daughter cells were distributed (Figure 4C; Figure 4—figure supplement 1B). Finally, at 8.5 mm we found that still-unpigmented nucEosFP+ cells near the distal fin tip generated distal clones restricted to a xanthophore fate (Figure 4—figure supplement 1C). These results show that individual unpigmented cells and early-developing orange cells in the fin can generate both erythrophores and xanthophores, depending on initial location and where progeny localize.

We further asked whether phenotypes of erythrophores and xanthophores might be plastic even after they differentiate by challenging cells in a regenerative context. To test for erythrophore → xanthophore conversion, we amputated fins through the region containing erythrophores, expecting that regeneration distally might allow for repositioning of erythrophores into regions where regenerative xanthophores would be expected, with conditions favorable to fate conversion, should cells retain such potential. We first assessed the possibility that transfating occurs by repeatedly imaging individual fish in brightfield, to learn whether cells near the amputation plane might lose their red color during regenerate outgrowth. Individual erythrophores could often be reidentified using other cells as well as distinctive features of fin ray bones and joints as landmarks (Figure 5A; Figure 5—figure supplement 1). As regeneration proceeded, small groups of cells having paler red or orange coloration, were sometimes observable where individual cells of deep red coloration had been found, suggestive of proliferation and dilution of pre-existing pigments. Later, only yellow cells were found in these same locations. These observations were consistent with the possibility of erythrophore → xanthophore conversion. To test this idea directly we marked nucEosFP+ erythrophores by photoconversion prior to amputation and followed labeled cells through regeneration (Figure 5B; Figure 5—figure supplement 2A). Many erythrophores divided to replenish their complement in proximally regenerating tissue, and a few erythrophores differentiated from unpigmented precursors, as indicated by the presence, or absence of photoconverted nucEosFP, respectively (Figure 5—figure supplement 2B). Additionally, some initially marked erythrophores came to occupy more distal regions and were indistinguishable from regenerative xanthophores that had developed from unpigmented progenitors even 36–51 days post-amputation (Figure 5C). These findings suggest a reduction in pre-existing red pigment as cells divide, and a failure to accumulate new red pigments once proliferation has ceased. We also sought to determine whether xanthophores can transition to an erythrophore fate by ablating central regions of fin and then assessing whether distal xanthophores can move into the regenerating proximal region. However, these experiments were not informative, as regenerative tissue was colonized by erythrophores or xanthophores differentiated from progenitors rather than pre-existing xanthophores (Figure 5—figure supplement 3).

These observations indicate that erythrophores and xanthophores of the adult anal fin share a lineage, that individual progenitor cells within the fin can contribute to both cell types, and that some plasticity in fate persists even after differentiation, with erythrophores able to transition to a yellow-pigmented phenotype when challenged to do so.

To better understand molecular mechanisms of red coloration we compared gene expression between fin regions containing only erythrophores or only xanthophores. Erythrophores occurred at only ~one-third the density of xanthophores in two dimensional images (Figure 2—figure supplement 1D and see below) and proximal and distal fin regions presumably differ in ways other than chromatophore content. Nevertheless, we reasoned that comparisons of bulk tissue preparations might still identify genes having marked differences in expression between erythrophores and xanthophores, as would be expected for loci functioning in pigment synthesis (Saunders et al., 2019). Mapping D. albolineatus sequencing reads to the zebrafish genome identified 18,050 expressed genes. Transcripts of 162 genes were more abundant in proximal erythrophore-containing tissue, whereas transcripts of 200 genes were more abundant in distal xanthophore-containing tissue (q < 0.05; fold-changes = 0.4–6.8) (Figure 6A and B; Supplementary file 1—Table 1).

To identify genes required for red or yellow coloration, we used CRISPR/Cas9 mutagenesis to knock out selected candidates that might have roles in processing of carotenoids, synthesis of other pigments, or fate specification (Supplementary file 1—Tables 2 and 3). We screened mosaic (F0) fish and isolated stable lines of mutant alleles for target genes with pigmentary phenotypes. Of 25 targets derived from RNA-seq, three yielded mutants with defects in pigmentation. To determine which pigments contributed to colors present in wild-type, and which were affected in mutants, we further assayed the carotenoid content of fin regions by HPLC.

In the wild type, fin tissue containing erythrophores was markedly enriched for the red ketocarotenoid astaxanthin; additional peaks had profiles consistent with other ketocarotenoids (Figure 7A, peak 3; Figure 7—figure supplement 1; Supplementary file 1—Table 4). Fin tissue containing xanthophores lacked astaxanthin and instead contained yellow zeaxanthin (peak 10), similar to zebrafish xanthophores (Saunders et al., 2019), as well as additional peaks characteristic of other yellow xanthophyll carotenoids.

To confirm that carotenoids rather than other pigments are principally responsible for pigmentation, we recovered mutant alleles of scarb1, required for carotenoid uptake and localization (Toomey et al., 2017; Saunders et al., 2019), as residual color in such mutants would suggest a non-carotenoid contribution. We isolated two alleles, scarb1^vp38ac1 (V84Δ16X) and scarb1^vp38ac2 (V84X), and found that scarb1^{vp38ac1/vp38ac2} individuals had a phenotype concordant with that of F0 mosaics (Figure 3A): they lacked color in the visible range and lacked carotenoids detectable by HPLC (Figure 7B; Figure 7—figure supplement 2A; Supplementary file 1—Table 4). The absence of residual red or yellow coloration suggested that pteridine pigments do not contribute to visible color in these cells, as they do in some other species (Goodrich et al., 1941; Matsumoto and Obika, 1968; Grether et al., 2001; Weiss et al., 2012; Olsson et al., 2013). Moreover, targeting of differentially expressed genes known to function in pteridine synthesis did not yield visible pigmentation defects in F0 mosaics (erythrophore region: spra, xdh; xanthophore region: aox5; Supplementary file 1—Tables 1 and 4). Pteridine pigments were detectable histologically in erythrophores and xanthophores, however, and could be visible to fish in the UV range (Figure 7—figure supplement 2B).

Mutants for two genes, cyp2ae2 (FO904880.1) and bdh1a (zgc:113142), lacked overt red coloration (Figure 7C; Figure 7—figure supplement 2A), although densities of erythrophores and xanthophores did not differ significantly from wild-type (Figure 7F). Both genes had transcripts that were more abundant in fin tissue containing erythrophores than xanthophores (log₂ fold-changes = 4.9, 2.1; q = 9.6E-23, 8.9E-8; Figure 6). We confirmed by RT-PCR that both genes were expressed in erythrophores picked manually by micropipette from dissociated fin tissue (Figure 7—figure supplement 3A).

cyp2ae2 encodes an enzyme within the large family of cytochrome P450 monooxygenases (Kirischian et al., 2011). A related gene encoding a different P450 family member, CYP2J19, is essential for red coloration in ‘red factor’ canary (Lopes et al., 2016) and zebra finch (Mundy et al., 2016), and likely has similar roles in other birds and turtles (Twyman et al., 2016; Twyman et al., 2018). CYP2J19 expression is testosterone-dependent (Khalil et al., 2020) and its product is believed to play an essential role in the conversion of yellow carotenoids like zeaxanthin into red ketocarotenoids like astaxanthin (Figure 1A). Orthologs of CYP2J19 appear to be restricted to birds and turtles (Twyman et al., 2016). Reciprocally, cyp2ae2 (FO904880.1) is clearly a member of the cyp2 family, and is likely orthologous to cyp2ae1 loci of other teleosts by sequence similarity and chromosomal position, yet no clear orthologues of this gene are present in amniotes (Ensembl Release 103) (Kirischian et al., 2011; Yates et al., 2020). The loss of red color in D. albolineatus erythrophores thus raises the possibility that CYP2J19 and Cyp2ae2 may have acquired carotenoid ketolase activity convergently. Supporting the idea that cyp2ae2 might encode a carotenoid ketolase, the cyp2ae2 mutant had markedly reduced amounts of astaxanthin in fin regions containing erythrophores but relatively greater amounts of zeaxanthin (peak 10) in both erythrophore and xanthophore containing tissue as compared to wild-type, consistent with cyp2ae2 expression in both tissues, albeit at different levels (Figure 7—figure supplement 1, Supplementary file 1—Table 5). Consistent with these findings, erythrophores of cyp2ae2 mutants had markedly reduced red/green fluorescence ratios (Figure 7D) and a reduced diameter of visible pigment (Figure 7E). Trace residual astaxanthin and other ketocarotenoids are unlikely to reflect residual activity of cyp2ae2, as the mutant allele, cyp2ae2^vp39ac1, harbors a five-nucleotide frameshift within the first coding exon leading to 42 novel amino acids followed by a premature stop codon (L46Δ43X). A paralogous locus, cyp2ae1, lies adjacent to cyp2ae2 and was expressed at very low levels in both fin regions (Figure 7—figure supplement 3B; Supplementary file 1—Table 1). Compensatory activity of cyp2ae1 might account for trace levels of ketocarotenoids in cyp2ae2^vp39ac1.

The second red-deficient mutant, bdh1a, encodes 3-hydroxybutyrate dehydrogenase type 1a, a short-chain dehydrogenase/reductase. Homologues of this gene in mammals are known to interconvert hydroxyl and ketone groups, and in particular acetoacetate and 3-hydroxybutyrate, two major ketone bodies (Green et al., 1996; Langston et al., 1996; Persson et al., 2009; Otsuka et al., 2020). Although not implicated previously in carotenoid processing or red coloration, transcripts of a homologous gene were enriched in orange skin of clownfish Amphiprion ocellaris (Salis et al., 2019). The bdh1a^vp40ac1 mutant (D141Δ12X) completely lacked astaxanthin and other ketocarotenoids in erythrophore-containing tissue, and did not exhibit increased levels of zeaxanthin, as observed in cyp2ae2^vpa39c1 (Figure 7C). Red/green fluorescence ratios of erythrophores are similar to those of xanthophores (Figure 7D) and the diameter of visible pigment is reduced from wild-type levels in erythrophores though not xanthophores (Figure 7E), confirming the visible phenotype.

Erythrophores of cyp2ae2 and bdh1a mutant fish appeared normal in size and shape in young adults yet became morphologically heterogeneous as fish age, with pigment-containing cell fragments and fewer cells evident, as well as an onset of whole-fish kyphosis by ~12 months post-fertilization (Figure 7—figure supplement 3C). These phenotypes suggest requirements for both loci in the accumulation of red carotenoids and subsequent homeostasis of erythrophores and other tissues, perhaps associated with a systemic dysregulation of carotenoid–Vitamin A—retinoid metabolism (von Lintig et al., 2005; Ghyselinck and Duester, 2019).

β-carotene oxygenase 1 (Bco1) symmetrically cleaves β-carotene to produce vitamin A, a precursor of retinoic acid; whereas β-carotene oxygenase 2 (Bco2) cleaves a variety of carotenoids in an asymmetric fashion, often leading to their degradation (Widjaja-Adhi et al., 2015; Li et al., 2017; Harrison and Kopec, 2020; Poliakov et al., 2020). bco1 and bco2b were more abundant in tissue containing xanthophores than erythrophores (log₂FC = 4.9, 2.1; q = 9.6E-23, 8.9E-8); a third locus, bco2l was similarly abundant at both sites (Supplementary file 1—Table 1). Given heterogeneities in transcript abundance, we asked whether β-carotene oxygenase genes might also contribute to differences in carotenoid accumulation between cell types. Only bco1-targeted fish exhibited an overt pigmentary phenotype in F0 mosaic animals, with pigment-free patches alongside patches of cells having apparently normal pigmentation consistent with a pigment-cell autonomous function (Figure 7—figure supplement 2C). Mutants stably carrying bco1 alleles (bco1^vp41ac1, 52DΔ5X; bco1^vp41ac2, 51FΔ9X) had reduced carotenoid levels in both xanthophores and erythrophores as well as smaller diameters of contracted pigment granules (Figure 7C–E; Figure 7—figure supplement 2A and B). The mechanism of this effect on chromatophore carotenoid content remains unclear.

Danio albolineatus were derived from individuals collected in Thailand by M. McClure in 1995 (McClure et al., 2006), provided to the laboratory of S. Johnson, and then maintained in our laboratory from 2000 until the present. Additional species of Danio used for assessing erythrophore complements were obtained directly from the field or through the pet trade [D. aesculapii; D. quagga, D. kyathit (McCluskey et al., 2021) D. nigrofasciatus, D. tinwini (Spiewak et al., 2018); D. kerri; D. choprae; D. margaritatus, D. eythromicron] and maintained subsequently in the lab or were observed in the field [D. meghalayensis, D. dangila (Engeszer et al., 2007)]. Fish were reared under standard conditions to maintain D. rerio (~28 °C; 14 L:10D) with larvae fed initially marine rotifers, derived from high-density cultures and enriched with Rotimac and Algamac (Reed Mariculture), with older larvae and adults subsequently fed live brine shrimp and a blend of flake foods enriched with dried spirulina. Stocks of mutant or transgenic D. albolineatus were: scarb1^vp38ac1, scarb1^vp38ac2, cyp2ae2^vp39ac1, bdh1^vp40ac1, bdh1^vp40ac2, csf1ra^vp4ac1, Tg(aox5:nucEosFP)^vp43aTg.

CRISPR/Cas9 mutagenesis bdh1a^vp40ac1 and cyp2ae2^vp39ac1 were generated by injecting one-cell stage embryos with 200 pg sgRNAs and 500 pg Cas9 protein (PNA Bio) using standard procedures (Shah et al., 2015). bco1^vp41ac1, bco1^vp41ac2, scarb1^vp38ac1 and scarb1^vp38ac2 mutagenesis as well as scarb1 targeted clonal labeling were conducted by injecting one-cell stage embryos with approximately 1 nanoliter of 5 μM gRNA:Cas9 RNP complex (IDT). These AltR CRISPR/Cas9 reagents allowed for highly efficient mutagenesis (Hoshijima et al., 2019) even in F0 fish, enabling clonal analyses of rare wild-type cells. For production of mutant lines, individual fish were sorted for anal fin phenotypes at juvenile stages and alleles recovered by intercrossing and outcrossing.

Transgenesis aox5:nucEosFP and mitfa:nucEosFP plasmids were made by assembling 8 kb aox5 and 5 kb mitfa promoters (Budi et al., 2011; McMenamin et al., 2014) with nuclear-localizing photoconvertible fluorophore EosFP using the tol2 Gateway Kit (Kwan et al., 2007) and were injected at the one-cell stage with tol2 mRNA at 25 pg per embryo (Suster et al., 2009) or 6 pg per embryo for clonal analyses.

Whole fish were euthanized then embedded in 1 % low-melt agarose, and captured on a Nikon D-810 digital single lens reflex camera with MicroNikkor 105 mm macro lens. Anal fin details were imaged using a Zeiss Axio Observer inverted microscope or Zeiss AxioZoom stereomicroscope equipped with Zeiss Axiocam cameras.

Carotenoid autofluorescence was imaged using a Zeiss LSM880 inverted laser confocal microscope in Airyscan SR mode. Laser intensity for red (excitation wavelength 561 nm) and green (excitation wavelength 488 nm) channels were set to be identical. For comparison with brightfield illumination, anal fins were stabilized with a few drops of 1 % low-melt agarose then specimens transferred to a Zeiss AxioObserver inverted microscope with Axiocam camera. Cells along the mid line of 10th inter-fin ray were imaged. Background of bright-field images was corrected to white using software Fiji imageJ (Schindelin et al., 2012): Duplicate> Gaussian Blur = 50 > Image Calculator to subtract background> Invert. For larva to juvenile comparisons (Figure 1F), 10 cells in 3rd and 4th inter-fin ray in proximal and distal were imaged. For wild-type and mutant comparisons (Figure 7D), 20 cells in 10th inter-fin ray in proximal and distal were imaged. Other fluorescent images (e.g. photoconverted images, aox5+ clonal labeling) were acquired using a Zeiss AxioObserver inverted microscope equipped with Yokogawa CSU-X1M5000 laser spinning disk and Hamatsu camera.

Images were captured either as single frames or as tiled sets of larger areas that were then stitched computationally using ZEN Blue software, the Autoblend feature of Adobe Photoshop, or manually in Adobe Photoshop. Color balance and display levels were adjusted manually for entire images as needed, with corresponding transformations applied across matched sets of images (e.g. across genotypes). In some instances, gradients of brightness across fields of view (e.g. proximal to distal) were adjusted by applying inverse density gradients in Adobe Photoshop.

Photoconversion was performed on Tg(aox5:nucEosFP) or plasmid-injected F0 mitfa:nucEosFP, using a Zeiss LSM 800 scanning laser confocal with a 405  nm laser and ZEN blue software. Fish were subsequently reared in tanks shaded from ambient light to prevent spontaneous photoconversion (McMenamin et al., 2014; Gur et al., 2020). Brightfield images were taken before photoconversion and fish inspected to ensure that no photoconversion had occurred as a result. Subsequent imaging used fluorescence channels only, except for end-point imaging in fluorescence followed by brightfield. Although pigments autofluoresce in the same channels as nucEosFP, treatment with epinephrine allowed contracted pigment granules to be distinguished unambiguously from nuclei.

For amputation experiments, fins were transected through the middle of the erythrophore-containing region, and imaged subsequently in brightfield (Figure 5A) following treatment with epinephrine. Alternatively, Tg(aox5:nucEosFP) adult males were exposed under an external Zeiss HXP 120  V compact light source for 15 min until all nucEosFP+ cells in fins had been converted (Figure 5C). Fish were anesthetized and anal fins amputated through erythrophore regions, then reared in a shaded tank as above. Unconverted controls reared in the same tank did not show any converted nucEosFP signal when examined concurrently at subsequent time points. Sham control (photoconverted without amputation) of regeneration experiment in Figure 5C. Converted nucEosFP signal remained very strong after 41 days. Images were taken 1 day after amputation and at the end point of the experiment. For excisions of middle fin regions (Figure 5—figure supplement 2), internal fin ray and inter-fin ray regions were removed and xanthophores close to the wound photoconverted.

Adult male anal fins were dissected and dissociated enzymatically with Liberase (Sigma-Aldrich cat. 5401119001, 0.25 mg/mL in dPBS) at 25°C for 15 min followed by gently pipetting for 5 min. Cell suspensions were then filtered through a 70 µm Nylon cell strainer to obtain a single-cell suspension. Individual cells were then picked manually under Zeiss Axio Observer inverted microscope. Cells were identified by their morphology: red erythrophores, yellow xanthophores, black melanophores and transparent small skin cells. Total RNAs were isolated by RNeasy Protect Mini Kit (Qiagen) and cDNAs synthesized with oligo-dT priming using SuperScript III Cells Direct cDNA Synthesis System (Thermo). Primers pairs were designed to span exon-intron junctions or long introns for assessing genomic contamination, targeting (Supplementary file 1—Table 5). Amplifications were performed using Taq polymerase with 35 cycles of 95°C for 30 s, 56°C for 30 s, 72°C for 15 s.

To assess pteridine content, amputated fins were imaged after exposure to dilute ammonia (pH 10.0), which liberates pteridines from protein carriers resulting in autofluorescence under DAPI illumination.

Adult male Danio albolineatus were euthanized, anal fins were dissected and tissue collected from proximal erythrophore or distal xanthophore regions in PBS. RNA was extracted using TRIzol and Direct-zol RNA MiniPrep Kit. mRNA was enriched using NEBNext Poly(A) mRNA Magnetic Isolation Module and sequencing libraries were constructed using NEBNext Ultra RNA Library Prep Kit for Illumina and sequenced on an Illumina Nextseq-500. Reads were aligned to Danio rerio reference genome GRCz11 using Kallisto (Bray et al., 2016) and analyzed using DESeq2 (Love et al., 2014). RNA-seq data are available through GEO (accession ID GSE174713).

Proximal (erythrophore containing) and distal (xanthophore containing) portions of the anal fin were dissected from nine individuals of each genotype and like samples were combined in pools of three for pigment extraction. The pooled fin tissue was homogenized with zirconia beads in 1.2 ml of 0.9% sodium chloride and protein content was quantified a bicinchoninic acid (BCA) assay (23250, Thermo). Carotenoids were extracted from the homogenates by combining 1 ml methanol, 2 ml distilled water, and 2 ml of hexane:tert-methyl butyl ether (1:1 vol:vol), collecting and drying the resulting solvent fraction under nitrogen. Each sample was then split and saponified with 0.02 M NaOH or 0.2 M NaOH in methanol at room temperature to maximize the recovery of ketocarotenoids or other xanthophylls, respectively (Toomey and McGraw, 2007). The saponified extracts were then injected into an Agilent 1,100 series HPLC fitted with a YMC carotenoid 5.0 µm column (4.6 mm × 250 mm, YMC). Carotenoids were separated with a gradient mobile phase of acetonitrile:methanol:dichloromethane (44:44:12) (vol:vol:vol) through 11 minutes, a ramp up to solvent ratios of 35:35:30 for 11–21 min and isocratic conditions through 35 minutes. The column was maintained at 30°C with a mobile phase flow rate of 1.2 ml min^–1 throughout. The samples were monitored with a photodiode array detector at 400, 445, and 480 nm, and carotenoids were identified and quantified by comparison to authentic standards (a gift of DSM Nutritional Products, Heerlen, The Netherlands).

Analyses of quantitative data were performed in JMP Pro 16 (SAS Institute, Cary NC). Numerical data presented in figures are provided in Supplementary file 1.