Small molecule screening identifies targetable zebrafish pigmentation pathways

Authors

  • Sarah Colanesi,

    1. Developmental Biology Programme, Centre for Regenerative Medicine, Department of Biology and Biochemistry, University of Bath, Bath, UK
    Search for more papers by this author
  • Kerrie L. Taylor,

    1. Institute for Genetics and Molecular Medicine, MRC Human Genetics Unit and the Edinburgh Cancer Research UK Centre, The University of Edinburgh, Edinburgh, UK
    Search for more papers by this author
  • Nicholas D. Temperley,

    1. Institute for Genetics and Molecular Medicine, MRC Human Genetics Unit and the Edinburgh Cancer Research UK Centre, The University of Edinburgh, Edinburgh, UK
    Search for more papers by this author
  • Pia R. Lundegaard,

    1. Institute for Genetics and Molecular Medicine, MRC Human Genetics Unit and the Edinburgh Cancer Research UK Centre, The University of Edinburgh, Edinburgh, UK
    2. NeuroSearch A/S, Ballerup, Denmark
    Search for more papers by this author
  • Dong Liu,

    1. Institute for Genetics and Molecular Medicine, MRC Human Genetics Unit and the Edinburgh Cancer Research UK Centre, The University of Edinburgh, Edinburgh, UK
    Search for more papers by this author
  • Trista E. North,

    1. Department of Pathology, Beth Israel Deaconess Medical Center, Boston, MA, USA
    Search for more papers by this author
  • Hironori Ishizaki,

    1. Institute for Genetics and Molecular Medicine, MRC Human Genetics Unit and the Edinburgh Cancer Research UK Centre, The University of Edinburgh, Edinburgh, UK
    Search for more papers by this author
  • Robert N. Kelsh,

    1. Developmental Biology Programme, Centre for Regenerative Medicine, Department of Biology and Biochemistry, University of Bath, Bath, UK
    Search for more papers by this author
  • E. Elizabeth Patton

    1. Institute for Genetics and Molecular Medicine, MRC Human Genetics Unit and the Edinburgh Cancer Research UK Centre, The University of Edinburgh, Edinburgh, UK
    Search for more papers by this author

Robert N. Kelsh, e-mail: bssrnk@bath.ac.uk
E. E. Patton, e-mail: e.patton@hgu.mrc.ac.uk

Summary

Small molecules complement genetic mutants and can be used to probe pigment cell biology by inhibiting specific proteins or pathways. Here, we present the results of a screen of active compounds for those that affect the processes of melanocyte and iridophore development in zebrafish and investigate the effects of a few of these compounds in further detail. We identified and confirmed 57 compounds that altered pigment cell patterning, number, survival, or differentiation. Additional tissue targets and toxicity of small molecules are also discussed. Given that the majority of cell types, including pigment cells, are conserved between zebrafish and other vertebrates, we present these chemicals as molecular tools to study developmental processes of pigment cells in living animals and emphasize the value of zebrafish as an in vivo system for testing the on- and off-target activities of clinically active drugs.

Significance

Small molecules are bioactive tools to probe pigment cell biology within specific developmental time windows and at different dose concentrations. While some of the compounds directly link the cellular phenotype to a known molecular mechanism, others reveal new pathways for the control of pigment cell biology. Thus, the zebrafish system is a valuable, rapid, and inexpensive system for screening small molecules and gaining new insight into pigment cell processes in vivo.

Introduction

Zebrafish have three pigment cell types: melanocytes, xanthophores, and iridophores, which produce black melanin, yellow pteridine, and reflective crystalline guanine, respectively (Kelsh et al., 2009). Combinations of these cells create a stereotypical pigmentation pattern in the embryo that is maintained until metamorphosis, when the fish develops the bluish-black and yellowish-silver striped adult pattern (Rawls et al., 2001). Melanocytes (often called melanophores in fish) are dendritic cells that become visible from prim-12 stage (28 h post-fertilization; hpf); iridophores become visible as shiny, oval cells under incident light from high-pec stage (42 hpf), and are often associated with melanocytes; and xanthophores, also dendritic, become most visible by 5 days post-fertilization (dpf). Within each of the pigment cell types, pigment is contained within membrane-bound organelles (melanosomes, reflecting platelets, and pteridine granules, respectively). During zebrafish embryonic development, pigment cells become specified from multipotent progenitors through combinations of intrinsic factors such as Sox10 (Dutton et al., 2001) with extrinsic factors, including Wnt (melanocytes) and Ltk signaling (iridophores) (Dorsky et al., 1998; Lopes et al., 2008) that together promote expression of cell-type-specific transcription factors such as Mitfa (melanocytes) (Lister et al., 1999). These cell-type-specific transcription factors then coordinate expression of the whole suite of genes required for full expression of the differentiated characteristics of the pigment cells.

As in the zebrafish and the mouse, human skin melanocytes are derived from the neural crest lineage, and many of the melanocyte molecular factors such as Sox10, Mitf, tyrosinase, Dct, and SLC25A4 are highly conserved between species (Levy et al., 2006; Lin and Fisher, 2007). Human skin is pigmented by melanocytes that produce and distribute melanin to surrounding keratinocytes, thereby coloring human skin and playing an important role in protection from the damaging effects of UV-light (Tran et al., 2008). In contrast, fish change color by altering pigment location and reflectivity during background adaptation or stress (Logan et al., 2006; Richardson et al., 2008). Many of the human melanocyte pigmentation and disease states can be modeled using zebrafish genetics, including skin color, albinism, Waardenburg Syndrome, vitiligo, nevi, and melanoma (Ceol et al., 2008; Ishizaki et al., 2010; Lamason et al., 2005; Navarro et al., 2008; Patton et al., 2005, 2010; Richardson et al., 2011; Schonthaler et al., 2008; Taylor et al., 2010; White and Zon, 2008). Thus, a detailed understanding of pigment cell biology in zebrafish is relevant to mammalian melanocyte biology.

Genetic screens in zebrafish and medaka have identified genes that control pigment cell processes, many of which are conserved and relevant to other animal species (Kelsh et al., 1996, 2004; Odenthal et al., 1996; Rawls and Johnson, 2003). These genetic tools have been complemented by efforts to identify small molecules affecting melanocyte pigmentation, distribution, survival, and stem cells (Hultman et al., 2009; Ishizaki et al., 2010; Jung et al., 2005; Mendelsohn et al., 2006; O’reilly-Pol and Johnson, 2008, 2009; Sheets et al., 2007; White et al., 2011; Yang and Johnson, 2006). The small size of the zebrafish embryo, coupled with the optical transparency and homologous target pathways in human cells, has made zebrafish an advantageous model organism for screening small molecules in vivo (Taylor et al., 2010; Zon and Peterson, 2005, 2010). An important aspect of chemical biology in zebrafish is the ability to alter time and dose of the compound, to identify discrete time of action of a specific pathway, and the ability to simultaneously target multiple members of a given protein family (Taylor et al., 2010; Zon and Peterson, 2005, 2010). Chemical biology in zebrafish can also contribute to the identification of the mode of action as well as unknown target pathways of small molecules in vivo (Ishizaki et al., 2010; Kokel et al., 2010; Rihel et al., 2010).

To expand the available molecular tools to probe pigment cell biology, we have performed a screen for small molecules that alter iridophore and melanocyte development in zebrafish embryos. Here, we present the pigment cell-specific compounds identified from the Sigma LOPAC library, the Enzo Life Sciences Screen-WellTM Kinase Inhibitor Library, and Enzo Life Sciences Screen-WellTM Phosphatase Inhibitor Library. Some have been identified in previous studies (Ishizaki et al., 2010; Mendelsohn et al., 2006; Sheets et al., 2007), but others provide new insight into developmental processes and should prove to be a useful resource to study both fundamental aspects of pigment cell development and models of pigment cell disease.

Results

We screened for compounds that caused specific pigment cell phenotypes or altered the pigment cells in addition to targeting other tissue structures (Figure 1A, B). Zebrafish embryos were collected, and five embryos were arrayed at sphere stage (4 hpf) in each well of a 24-well plate. This developmental stage is well before neural crest development, which arises during the segmentation period at about 10 hpf. Screening in 24-well plates provided ample space for manipulation of the live embryos to visualize as much of the pigmentation as possible, while keeping only five embryos per well helped prevent a decrease in water quality from embryo crowding during the screen. Each well contained 10 μM of a compound from the Sigma LOPAC collection (1280 bioactive compounds), or the Enzo Life Sciences Screen-WellTM Kinase and Phosphatase libraries (80 and 33 compounds, respectively). Many of the compounds are active in other systems at 10 μM (e.g. cell cultures) and are expected to be stable for the duration of the screen, although the activity of the compounds may diminish over time (for example, see roscovitine below). The developing fish were screened under the dissecting microscope for changes in iridophore and melanocyte number, location, and pigmentation/iridescence at long-pec stage (48 hpf), protruding mouth stage (72 hpf), and larval day 4 stage (96 hpf). We initiated screening at long-pec stage (48 hpf) because both melanocytes and iridophores are clearly visible in a stereotypic pattern at this time. To facilitate screening of iridophores, we screened independently both AB wild-type and mitfa mutant embryos with each compound. General toxicity phenotypes (e.g. necrotic tissue and general delay) were observed with 81 compounds and were eliminated from the screen. To confirm the effects of those compounds that had a more specific effect, we re-tested compounds from the screening plate at three different concentrations and confirmed their phenotype (Table 1 and Table S1, see Supporting Information; shaded text). For some compounds, we obtained fresh supplies and re-tested to confirm the phenotype (Table 1 and Table S1, see Supporting Information; black text). We also performed targeted retesting of compounds identified from the literature that were not confirmed in the second round of rescreening, most likely due to lower than expected concentrations in the screening plates (Table 1 and Table S1, see Supporting Information; asterisk).

Figure 1.

 Screen for small molecules that modulate pigment cell biology in zebrafish. (A) Adult zebrafish were bred, embryos collected and arrayed in 24-well plates, each well with a compound at 10 μM dissolved in DMSO and E3 embryo medium. Zebrafish embryos were screened at long-pec stage (48 hpf), protruding mouth stage (72 hpf), and larval day 4 (96 hpf) for melanocyte and iridophore phenotypes and classified, where possible, based on the phenotypic analysis used to characterize genetic mutations altering pigment cell biology (Kelsh et al., 1996). (B) Pie-chart showing the range and distribution of phenotypes observed in the screen. Numbers of chemicals identified for each phenotypic category are indicated.

Table 1.   Results of screening small molecules for pigment cell phenotypes Thumbnail image of

Phenotypes were classified using the scheme for the large-scale genetic screen in zebrafish (Kelsh et al., 1996). This phenotypic classification is based on four broad categories: reduced chromatophore numbers (Class I–III), abnormal chromatophore distribution (Class IV, V), reduced chromatophore pigmentation (Class VI.A–J), and abnormal chromatophore morphology (Class VII.A–D). Fifty-seven compounds were confirmed in our screen that produced reproducible melanocyte phenotype (Figure 2) and iridophore phenotype (Figure 3; Table 1 and Table S1, see Supporting Information).

Figure 2.

 Small molecules that modulate melanocyte biology in zebrafish embryos. Images of zebrafish embryos treated with small molecules identified in the screen or DMSO as a control. Embryos (prim-25 stage; 36 hpf) treated with DMSO (A, A′), ellipticine (B, B′, C, C′), or sanguinarine (D, D′) reveal examples of wild type, spindly morphology, and pale and small melanocytes, respectively. A′, B′, and D′ are enlarged images to show phenotypic detail. Melanocyte fragmentation is indicated (arrow). Dorsal view of DMSO (C) and ellipticine (C′)-treated embryos show the spindly morphology of the melanocytes (arrow). (E, E′, F, F′) Pec-fin stage embryos (60 hpf) treated with DMSO, AG1478 (20 μM) or hydroquinone (40 μM) reveal examples of wild type, an abnormal melanocyte pattern, and reduced pigmentation. Arrow indicates reduced numbers of melanocytes at the ventral stripe; asterisks indicate accumulation of melanocytes at the dorsal stripe.

Figure 3.

 Small molecules that modulate iridophore biology in zebrafish. Images of zebrafish embryos treated with small molecules identified in the screen (A′–D′) or DMSO as a control (A–D, respectively). Phenotype examples include dull iridophores (fiduxosin), reduced numbers of iridophores (ellipticine), and ectopic iridophores (CyPPA). CyPPA treatment most likely causes an indirect effect on iridophore location because of a notochord defect (arrowhead). All panels except C, C′ are imaged in incident light.

Class I–III: Reduced chromatophore numbers

We screened for changes in melanocyte and iridophore development and did not screen for changes in xanthophores; thus, unlike in the genetic mutant screen (Kelsh et al., 1996), we did not identify compounds that affected all three cell types (Class I). However, we did identify 13 compounds that affected iridophore number (Class II) and five compounds that affected melanocyte number (Class III). Small molecules that affected melanocyte number included leflunomide that has recently been described as affecting transcription of neural crest genes and is effective against melanoma xenografts (White et al., 2011).

Of particular interest, we found that zebrafish embryos treated with roscovitine had fewer visible melanocytes than DMSO control-treated embryos (Figure 4, Table 1 and Table S1, see Supporting Information) Zebrafish melanocytes first become visible at about 25 hpf, and by 60 hpf, there are about 460 pigmented melanocytes that make up the pattern of the embryo (Yang and Johnson, 2006). Fewer observed melanocytes could be due to roscovitine promoting melanocyte cell death, modulation of proliferation of melanocyte precursors, or clustering of melanocytes so that they cannot be distinguished as separate units. We treated 3 and 4 dpf embryos (with fully pigmented melanocytes) with roscovitine and found no evidence of cell death (e.g. fragmentation of melanin or extrusion of pigmented cell fragments; data not shown; Yang and Johnson, 2006). To define the phenotype more precisely, we next examined the concentration and time of action window for roscovitine in melanocyte development. We found roscovitine to be a relatively unstable compound, which needed to be refreshed daily in the water. A 20 μM roscovitine continuous treatment from prim-5 stage (24 hpf) allowed the embryo to develop without obvious general developmental defects. However, although melanocytes of the roscovitine-treated fish were darkly pigmented and appeared in the correct spatial location, they were reduced in total number and some of the melanocytes appeared to have a short dendrites compared with control-treated zebrafish (Figure 4A, B). Higher concentrations of roscovitine caused non-specific toxicity in the zebrafish embryo, including heart edema and brain necrosis (data not shown). To quantitate the observed melanocyte effect at 20 μM, we defined a region of the head of an immobilized embryo and counted the number of melanocytes in the dorsal stripe (Richardson et al., 2008; Figure S1). Roscovitine caused a significantly reduced number of melanocytes in the developing zebrafish, compared with control-treated fish (Figure 4C, D). Melanocyte loss was not restricted to the head region and caused fewer melanocytes over the entire body of the 4 dpf zebrafish embryo [28% reduction in melanocyte number: control mean: 300.8 (SD = 53.53, n = 5); roscovitine 20 μM mean: 217.2 (SD = 15.90, n = 6) two-sample unpaired t-test P = 0.0051]. Thus, roscovitine appears to affect total melanocyte cell number in the developing embryo.

Figure 4.

 Roscovitine alters melanocyte cell number in zebrafish embryos. (A, B) Images of 5-day old zebrafish embryos treated with DMSO or 20 μM roscovitine. (C, D) Box and whisker plots of the number of melanocytes in the head region (Figure S2) in DMSO-treated embryos compared with embryos treated with roscovitine from prim-5 stage (24 hpf) until (C) 4 dpf and (D) 5 dpf. Standard mean is indicated, outliers are represented by an asterisk.

Class IV: Melanocyte pattern abnormal

Neural crest-derived melanocytes in zebrafish embryos undergo extensive migration from their neural crest origin to generate a highly stereotypical pattern of four stripes (dorsal, lateral, ventral, and yolk sac) (Kelsh et al., 2009). We identified five compounds that affected pigment cell pattern, defined by an absence of part of the normal pigmentation pattern (Class IV). These compounds all affected the normal melanocyte pattern and included inhibitors of PI3-kinase signaling (LY294002), p56LCK signaling (Emodin), ERB signaling (Tyrphostin AG1478), and retinoic acid signaling. One kinase inhibitor, Tyrphostin AG1296, caused a strong melanocyte migration delay in the zebrafish embryo (Figure 5A, B). Tyrphostin AG1296 treatment from bud stage (10 hpf) permitted melanocyte specification and pigmentation, but their migration remains incomplete, so that cells remain clustered around the ear and in dorsal regions of the embryo and fail to cover the yolk sac and yolk sac extension (Figure 5A). Treatment with Tyrphostin AG1296 was dosage dependent, with the strongest defect seen at 20 μM (Figure 5C). The time of addition was also important: later addition of the compound could produce intermediate phenotypes in which melanocytes at the anterior of the fish had migrated correctly while the posterior had not (data not shown). Tyrphostin AG1296 is a PDGFR and c-kit receptor inhibitor, and we find inhibitory effects on other kinases such as Aurora B by in vitro kinase profiling (Table S2, see Supporting Information). The receptor tyrosine kinase c-kit is critical for melanocyte migration, and mutations in this gene lead to white midline spots in mice and piebaldism in humans (Lamoreux et al., 2010). In zebrafish, there are two kit genes (kita/sparse and kitb), and kita is required for melanocyte migration and survival (Parichy et al., 1999; Rawls and Johnson, 2003). Based on the similar phenotypes between the zebrafish kit mutant (Parichy et al., 1999) and Tyrphostin AG1296 chemical treatment, we suggest that the primary target of Tyrphostin AG1296 in melanocytes is kit signaling.

Figure 5.

 Tyrphostin AG1296 alters melanocyte movement in zebrafish embryos. (A, B) A prim-25 stage (36 hpf) zebrafish embryo treated with DMSO as a control or 20 μM of Tyrphostin AG1296 from the 2 to 4 somite stage (10.7–11.3 hpf). Treated embryos develop melanocytes, but the melanocytes are clustered behind the ear (asterisk; A′, B′) and have been retarded in their migration along the medial migration pathway (red arrow; A″, B″). Melanocytes also fail to populate the yolk sac and yolk sac extension (black arrows). (C) Embryos treated with 12 μM Tyrphostin AG1296 (C′) show reduced numbers of melanocytes at yolk extension and ventral stripe compared with embryos treated with 20 μM Tyrphostin AG1296 (C″; arrows).

Class V: Ectopic chromatophores

Five compounds were identified that caused ectopic chromatophores, defined as ectopic chromatophores in addition to the wild-type zebrafish pattern. One compound, CyPPA, caused clusters of iridophores associated with a buckling and wavy notochord (Figure 3). This phenotype is similar to the phenotype caused by copper deficiency (Ishizaki et al., 2010). We found this phenotype could be prevented with the addition of copper chloride to the water (data not shown), suggesting that the appearance of ectopic iridophores in CyPPA-treated embryos is most likely due to the structural changes within the embryo affecting iridoblast migration, rather than a direct change in iridophore development.

Class VI: Reduced chromatophore pigmentation

Pale melanin (VI.A) was the most commonly altered characteristic identified in the screen, with 23 compounds affecting pigment synthesis and/or intensity. For some compounds, we understand the mode of interference with pigmentation, for example by disrupting copper metabolism and/or tyrosinase activity (e.g. PTU, U0126, and hydroquinone). While we attempted to remove compounds that simply caused general toxicity, for other compounds, we note that pale pigmentation is often associated with general developmental delay and/or reduced fitness, and therefore, some of the compounds identified may not reflect direct interference with pigment synthesis.

We identified four compounds that affected iridophore pigmentation, such that iridophores were present, but were dull and lacked reflective properties (Class VI.I). While melanin synthesis is well understood, much less is known about iridophore pigmentation. Iridophores in some fish species can respond to transmitter release by associated adrenergic nerves (Fujii, 2000; Maeno and Iga, 1992; Mathger et al., 2003; Nagaishi and Oshima, 1989). For example, motile iridophores have been described in adult zebrafish blue stripes that can change from a blue to yellow color via a norepinephrine-alpha-adrenoreceptor response (Oshima, 2002), and earlier studies have shown that adrenoreceptor signaling may promote changes to reflecting platelet dispersion in some species of fish (Matsuno and Iga, 1989). In our study of zebrafish embryonic iridophores, we find no evidence of motility in the light-reflecting platelets of the iridophores (S. Colanesi and R. N. Kelsh, unpublished observations). Thus, it was unexpected that we identified fiduxosin as a potent inhibitor of iridophore reflection in zebrafish development (Figure 6). Fiduxosin is an alpha-adrenoreceptor antagonist that blocks G-protein-coupled signaling through the catecholamine adrenergic receptors (Hancock et al., 2002). Previous work has associated alpha-adrenoceptor signaling with pigment aggregation (Fujii, 2000), yet in our screen, this effect was not seen with fiduxosin. In contrast, fiduxosin interfered with iridophore reflectiveness in a dose-dependent manner, with complete loss of iridophore reflection at 15 μM (Figure 6A). Iridophores were still present, however, as indicated by expression of the iridophore gene, leukocyte tyrosine kinase (ltk, Lopes et al., 2008) by RNA in situ hybridization (Figure 6A). This suggests that fiduxosin may interfere with iridophore light-reflective properties or directly with an iridophore differentiation program.

Figure 6.

 Fiduxosin treatment affects reflective properties of iridophores. (A) Fiduxosin causes a dose-dependent loss of iridophore reflectivity (left panel, close up of dorsal stripe at 72 hpf, incident light), while still expressing the iridophore-specific marker ltk (right panel, close up of corresponding area of dorsal stripe at 72 hpf). (B) Fiduxosin disrupts pigmentation of iridophores. Fiduxosin-treated embryos maintain expression of early markers of the iridophore lineage, ltk and ednrb1, but show reduced expression of a late iridophore differentiation marker, cb632 (also called gmps).

If fiduxosin disrupts the iridophore reflective capacity by altering the reflective plates, we would expect the effects to be rapidly reversible upon drug ‘wash-out’ (transfer to fresh water). For example, in the tropical paradise whiptail fish, the rapid changes to iridophore reflectivity from blue to red occur within seconds to minutes and are reversible (Mathger et al., 2003). Likewise, in zebrafish melanocytes, melanin distribution in response to background color or epinephrine is rapid and readily reversible (Logan et al., 2006; Richardson et al., 2008; Sheets et al., 2007). In contrast to our experience with melanin distribution, we found very slow recovery (requiring at least 24 h) of iridophore reflectivity after ‘wash-out’ of fiduxosin (data not shown), suggesting Fiduxosin alters the formation of the guanine plates.

To examine the effects of fiduxosin in more detail, we examined expression at prim-25 (36 hpf), long-pec stage (48 hpf), and protruding mouth stage (72 hpf) of two iridoblast markers, ltk and endothelin receptor B (ednrb1), and the iridophore marker guanine monophosphate synthetase (gmps; also known as cb632,http://zfin.org/) in embryos treated with 10 μM fiduxosin from 20 hpf. Expression of ednrb1 and ltk at prim-25 stage (36 hpf) and long-pec stage (48 hpf) in the distinctive pattern of iridophores demonstrates that specification and early differentiation of the iridophore lineage has occurred normally (Figure 6B). In contrast, these cells did not express the gmps gene (Figure 6B), which is required for guanine platelet formation in the iridophore (Ng et al., 2009). Thus, we suggest fiduxosin directly interferes with the iridophore differentiation program resulting in a failure to express gmps at the levels required for guanine platelet formation.

Class VII: Abnormal chromatophore morphology

We identified 11 compounds that resulted in pale, spindly melanocytes (VII.A) and nine compounds that caused small, rounded melanocytes (VII.B). Strikingly, three cyclooxygenase (COX) inhibitors – ibuprophen, indomethacin, and nimesulide – caused melanocytes to have a spindly morphology at lower-dose treatments and to be small and spot-like at higher concentration treatments (Figure 7). Progesterone and the GABAA receptor activator were also identified in both categories. We did not identify any compounds that affected iridophore shape, or compounds that caused stellate chromatophores (VII.C). Finally, while we did not actively screen for alterations in melanocyte background adaptation (VII.D), we identified three compounds that had fully expanded, darkly pigmented melanocytes. In zebrafish, pigment in melanocytes is motile, and the melanin area can expand or contract within the cell in response to hormones and catecholamines via a cyclic AMP (cAMP) response signaling pathway (Fujii, 2000; Logan et al., 2006; Richardson et al., 2008; Sheets et al., 2007). Two of the identified compounds, forskolin and nocodozole, are known to affect melanin distribution in background adaptation (Logan et al., 2006; Richardson et al., 2008; Sheets et al., 2007) and were also confirmed in our screen (VII.D; Table 1, data not shown).

Figure 7.

 COX inhibitors cause spot-like melanocytes. Images of long-pec stage (48 hpf) embryos treated with (A, D) DMSO, (B, B′) nimesulide, (C, C′) indomethacin, or (D′) ibuprophen. Melanocytes with a spindly morphology (black arrow) or spot-like (red arrow) are indicated.

Discussion

Small molecules provide a means for controlling pigment cell development that complements the wealth of genetic mutations identified in nature and in the laboratory. The zebrafish is an advantageous system for pigment cell chemical biology because the embryos are (i) transparent, (ii) develop ex vivo, and (iii) small enough for multiple embryos to be immersed in relatively little total compound. By screening a small series of libraries, we have identified more than 50 compounds that affect a range of pigment cell processes including specification, migration, pigmentation, and differentiation. A number of the compounds identified are consistent with known processes both in zebrafish and in other melanocyte systems. For example, forskolin and nocodozole were identified as causing no background adaptation (VII.D), consistent with the known role of cAMP signaling and microtubules in melanin movement within zebrafish melanocytes (Logan et al., 2006; Richardson et al., 2008; Sheets et al., 2007). Also, the melanocyte migration phenotype caused by the c-kit inhibitor AG1296 corresponded well to that expected based upon the known pharmacology of the compound.

Many compounds affected more than one biological process. This is likely to reflect the repeated use of biochemically similar activities, such as intracellular signaling pathways, in different biological processes. As demonstrated in the literature and confirmed by our in vitro kinase profiling analysis of AG1296 and roscovitine (Table S2, see Supporting Information), many kinase inhibitors target more than one kinase. Thus, the phenotypes we observe may, unlike genetic mutations, reflect inhibition of multiple kinases. Indeed, we hypothesize that the iridophore phenotype of AG1296 indicates inhibition of leukocyte tyrosine kinase (Ltk; Lopes et al., 2008). Furthermore, some compounds are designed to target a class of kinases, such as MEK inhibitors that target both MEK1 and MEK2 (Barrett et al., 2008).

In many cases, the link between the characterized pharmacological target of the compound and the pigment phenotype observed is unknown. For example, how fiduxosin inhibits iridophore pigmentation is unknown, indicating a possible new pathway in pigment cell biology. Alternatively, there may be additional unexpected targets of the compounds that affect known pigment cell biology regulators. This is seen with both U0126 and CyPPA that prevent melanin synthesis and alter iridophore location by an indirect effect on copper metabolism required for tyrosinase activity and notochord development (Figure 3C; Ishizaki et al., 2010). We speculate that an unexpected target may also be responsible for the small and pale melanocytes caused by the VEGF inhibitor SU1498, a phenotype not detected in treatments with other VEGF inhibitors such as SU4312 (Figure S2).

Some phenotypes were difficult to assign unambiguously to specific categories. For example, it was not always easy to distinguish between reduced number of iridophores (II) and dull iridophores (VI.I) phenotypes based on visualization by incident light because dull iridophores may be present but not clearly visible and/or the intrinsic variability in appearance of these cells dependent upon the angle of incidence of illumination. Likewise, depending on the phenotype, it was sometimes difficult always to accurately assign phenotypes to the melanocyte pattern abnormal (IV) or reduced number of melanocytes (III) classes. Nonetheless, we have made a first attempt to characterize the phenotypes, but appreciate that future detailed studies of the compounds in zebrafish development may indicate additional or different classifications.

Roscovitine, fiduxosin, and COX inhibitors were of particular interest from our screen, not least because they are all currently used in humans to treat a range of conditions. We identified roscovitine as a compound that specifically interfered with the numbers of melanocytes in the developing embryo (Figure 4). Roscovitine (Seliciclib) is a cyclin-dependent kinase (CDK) inhibitor that has activity against CDK2/7/9 and is being developed as an anti-cancer, anti-viral, and anti-inflammatory drug. Roscovitine also effectively inhibits ERK8 and DYRK1A in our kinase profiling assays (Table S2, see Supporting Information). CDK2 appears to play a specific role in melanoma as a transcriptional target of Mitf, and CDK2 expression levels in melanoma lines correlate with sensitivity to roscovitine treatment (Du et al., 2004). A specific role for CDKs in melanocyte development has not been previously shown, in part, because mice deficient for CDK2 have normal coat color (Berthet et al., 2003; Ortega et al., 2003). Roscovitine is also a potent activator of p53, possibly by inhibition of a CDK-dependent mechanism (Lu et al., 2001). While the molecular mechanism underlying the roscovitine phenotype is unknown, roscovitine treatment reveals a novel zebrafish pigmentation phenotype, distinct from those defined genetically to date.

Fiduxosin is used as a muscle relaxant for symptomatic treatment of benign prostatic hyperplasia. We identified fiduxosin as a compound that specifically interferes with an iridophore differentiation transcriptional program (Figure 6). Currently, we know almost nothing about the transcriptional basis for iridophore differentiation, so identifying the target inhibited by fiduxosin might be very informative. While fiduxosin is an alpha-adrenoreceptor antagonist, however, it is not clear whether this is the target in iridophore differentiation. As a practical application, fiduxosin may be a useful tool for removing iridophore pigmentation and autofluorescence, in just the way that PTU is useful for preventing melanin synthesis.

Finally, COX inhibitors are non-steroidal anti-inflammatory drugs commonly used in people to treat pain, reduce heart disease, and more recently shown to affect cord blood stem cells (North et al., 2007). We found COX inhibitors to cause spindly melanocytes at lower concentrations, and distinctive small, spot-like melanocytes at higher concentrations (Figure 7). COX enzymes produce prostaglandins that can increase differentiation of epidermal melanocytes. In human skin, keratinocytes are thought to produce the prostaglandin PGE2 as a pro-inflammatory mediator after erythema caused by excessive UV exposure. More recently, prostaglandin synthesis enzymes have been identified in melanocytes, and prostaglandins have been shown to be a UV irradiation-inducible autocrine factor for epidermal melanocytes that can stimulate tyrosinase activity (Gledhill et al., 2010; Masoodi et al., 2010; Scott et al., 2005; Starner et al., 2010). In our screen, the identification of three small molecules with the same target and same phenotype is compelling evidence that COX inhibition (and not an alternative target) is directly responsible for the melanocyte phenotypes in zebrafish. This may provide a useful in vivo system to address how prostaglandins interact with the pigmentation response, cAMP signaling, and alpha-MSH signaling (Gledhill et al., 2010; Masoodi et al., 2010; Scott et al., 2005; Starner et al., 2010). Notably, COX inhibitors, such as aspirin, can decrease the incidence of colorectal cancer. Thus, a target for COX inhibitors in zebrafish melanocyte homeostasis may provide insight when applied to the study of prostaglandins in melanoma (Fricke et al., 2010).

In conclusion, we have identified small molecules that promote phenotypes that mirror some of the zebrafish genetic pigment cell mutants as well as identify new pigment cell pathways and phenotypes. Small molecules often have unintended targets in vivo, and additional efforts are required to determine the direct target within the pigment cell. However, testing multiple inhibitors that target the same pathway will help to identify both the target pathway of interest within the pigment cell as well as unintended targets specific to a particular compound. Many of the compounds identified here have been effective in other melanocyte contexts (e.g. melanoma and melanocytes in culture), underscoring the value of the approach. Importantly, we have identified compounds and phenotypes not previously reported, providing new resources for exploring melanocyte and iridophore development in detail.

Methods

Zebrafish husbandry

Zebrafish care and procedures were approved by the University of Edinburgh veterinary staff and Ethics committee and the University of Bath Ethical Review Committee, and in compliance with the Animals Scientific Procedures Act 1986 of the UK. Fish were housed in fish facilities at the Universities of Edinburgh (MRC funded) and Bath (Wellcome Trust funded). Embryos were acquired by breeding wild-type strains AB, AB × TL or mitfa mutant zebrafish lines.

Chemical screening

The chemical libraries screened were the Sigma LOPAC collection (1280 bioactive compounds), the Enzo Life Sciences Screen-WellTM Kinase (80 compounds), and the Enzo Life Sciences Screen-WellTM Phosphatase libraries (33 compounds). Five sphere stage (4 hpf) embryos were arrayed in 24-well plates (Corning, Amsterdam, Netherlands) containing 10 μM of compound in 1% DMSO in 1 ml of E3 embryo medium. Embryos were assessed for phenotypic changes under standard and incident light conditions at long-pec stage (48 hpf), protruding mouth stage (72 hpf), and larval day 4 (96 hpf).

In situ hybridization

In situ hybridization with hydrolized RNA probes was performed as described Thisse & Thisse (2008). Probes were labeled with digoxigenin using the DIG RNA Labeling Kit (Roche, Mannheim, Germany) and detected with the appropriate antibody. The signal was detected by incubating the whole-mount samples in BMPurple AP Substrate (Roche).

Kinase profiling

Kinase profiling was performed at the International Centre for Kinase Profiling, MRC Protein Phosphorylation Unit, Dundee (http://www.kinase-screen.mrc.ac.uk/). Compounds were supplied in DMSO and screened in duplicate against a panel of 131 protein kinases, using a radioactive (33P-ATP) filter-binding assay, as described in detail in Bain et al. (2007).

Melanocyte counting assay

Embryos were assessed for melanocyte cell number by first exposing the zebrafish to light to contract the melanin within the melanocyte and then fixed in 4% paraformaldehyde and imaged. Melanocytes were counted within a defined head region (Richardson et al., 2008) in photographs by two observers, or over the entire body.

Acknowledgements

We are grateful to Professor Ian Jackson for critical reading of the manuscript and helpful discussions, Dr. Karthika Paranthaman for excellent zebrafish husbandry, and Dr. Corina Anastasaki and Craig Nicol for help with Figure 1. This work was funded by an EC Framework Programme 7 ZF-CANCER (N.T., E.E.P), Medical Research Scotland (H.I., E.E.P), the Medical Research Council (K.T., E.E.P., R.N.K.), the BBSRC (R.N.K.) and the University of Bath (S.C.).

Ancillary