A golden fish reveals pigmentation loss in Europeans

Pigment color and pattern are important for camouflage and the communication of visual cues. In vertebrates, body coloration is a function of specialized pigment cells derived from the neural crest (1). The melanocytes of birds and mammals (homologous to melanophores in other vertebrates) produce the insoluble polymeric pigment melanin. Melanin plays an important role in the protection of DNA from ultraviolet radiation (2) and the enhancement of visual acuity by controlling light scatter (3). Melanin pigmentation abnormalities have been associated with inflammation and cancer, as well as visual, endocrine, auditory, and platelet defects (4).

Despite the cloning of many human albinism genes and the knowledge of over 100 genes that affect coat color in mice, the genetic origin of the striking variations in human skin color is one of the remaining puzzles in biology (5). Because the primary ultrastructural differences between melanocytes of dark-skinned Africans and lighter-skinned Europeans include changes in melanosome number, size, and density (6, 7), we reasoned that animal models with similar differences may contribute to our understanding of human skin color. Here we present evidence that the human ortholog of a gene associated with a pigment mutation in zebrafish, SLC24A5, plays a role in human skin pigmentation.

The zebrafish golden phenotype. The study of pigmentation variants (5, 8) has led to the identification of most of the known genes that affect pigmentation and has contributed to our understanding of basic genetic principles in peas, fruit flies, corn, mice, and other classical model systems. The first recessive mutation studied in zebrafish (Danio rerio), golden (gol^b1), causes hypopigmentation of skin melanophores (Fig. 1) and retinal pigment epithelium (Fig. 2) (9). Despite its common use for the calibration of germ-line mutagenesis (10), the golden gene remained unidentified.

The golden phenotype is characterized by delayed and reduced development of melanin pigmentation. At approximately 48 hours postfertilization (hpf), melanin pigmentation is evident in the melanophores and retinal pigment epithelium (RPE) of wild-type embryos (Fig. 2A) but is not apparent in golden embryos (Fig. 2B). By 72 hpf, golden melanophores and RPE begin to develop pigmentation (Fig. 2, F and G) that is lighter than that of wild type (Fig. 2, D and E). In adult zebrafish, the melanophore-rich dark stripes are considerably lighter in golden compared with wild-type animals (Fig. 1, A and B). In regions of the ventral stripes where melanophore density is low enough to distinguish individual cells, it is apparent that the melanophores of golden adults are less melanin-rich than those in wild-type fish (Fig. 1, A and B).

Transmission electron microscopy was used to determine the cellular basis of golden hypopigmentation in skin melanophores and RPE of ∼55-hpf wild-type and golden zebrafish Wild-type melanophores contained numerous, uniformly dense, round-to-oval melanosomes (Fig. 1, C and E). The melanophores of golden fish were thinner and contained fewer melanosomes (Fig. 1D). In addition, golden melanosomes were smaller, less electron-dense, and irregularly shaped (Fig. 1F). Comparable differences between wild-type and golden melanosomes were present in the RPE (fig. S1, A and B).

Dysmorphic melanosomes have also been reported in mouse models of Hermansky-Pudlak syndrome (HPS) (11, 12). Because HPS is characterized by defects in platelet-dense granules and lysosomes as well as melanosomes, we examined whether the golden mutation also affects thrombocyte function in the zebrafish. A comparison of golden and wild-type larvae in a laser-induced arterial thrombosis assay (13) revealed no significant difference in clotting time (35 versus 30 s). The golden phenotype thus appears to be restricted to melanin pigment cells in zebrafish.

The zebrafish golden gene is slc24a5/nckx5. Similarities between zebrafish golden and light-skinned human melanosomes suggested that the positional cloning of golden might lead to the identification of a phylogenetically conserved class of genes that regulate melanosome morphogenesis. Positional cloning, morpholino knockdown, DNA and RNA rescue, and expression analysis were used to identify the gene underlying the golden phenotype. Linkage analysis of 1126 homozygous gol^b1 embryos (representing 2252 meioses) revealed a single crossover between golden and microsatellite marker z13836 on chromosome 18. This map distance of 0.044 centimorgans (cM) [95% confidence interval (CI), 0.01 to 0.16 cM] corresponds to a physical distance of about 33 kilobases (kb) (using 1 cM = 740 kb) (14). Marker z9484 was also tightly linked to golden but informative in fewer individuals; no recombinants between z9484 and golden were identified in 468 embryos (95% CI, distance <0.32 cM). Polymerase chain reaction (PCR) analysis of a γ-radiation–induced deletion allele, gol^b13 (15), showed a loss of markers z10264, z9404, z928, and z13836, but not z9484 (fig. S2A). Screening of a zebrafish genomic library (16) led to the identification of a clone (PAC215f11) containing both z13836 and z9484 within an ∼85-kb insert. Microinjection of PAC215f11 into golden embryos produced mosaic rescue of wild-type pigmentation in embryonic melanophores and RPE (Fig. 2, H and I), indicating the presence of a functional golden gene within this clone.

Shotgun sequencing, contig assembly, and gene prediction revealed two partial and three complete genes within PAC215f11 (fig. S2B): the 3′ end of a thrombospondin-repeat–containing gene (flj13710), a putative potassium-dependent sodium/calcium exchanger (slc24a5), myelin expression factor 2 (myef2), a cortexin homolog (ctxn2), and the 5′ end of a sodium/potassium/chloride cotransporter gene (slc12a1). We screened each candidate gene using morpholino antisense oligonucleotides directed against either the initiation codon (17) or splice donor junctions (18). Only embryos injected with a morpholino targeted to slc24a5 (either of two splice-junction morpholinos or one start codon morpholino) successfully phenocopied golden (Fig. 2C). In rescue experiments, injection of full-length, wild-type slc24a5 transcript into homozygous gol^b1 embryos led to the partial restoration of wild-type pigmentation in both melanophores and RPE (Fig. 2, J and K). Taken together, these results confirm the identity of golden as slc24a5.

To identify the mutation in the gol^b1 allele, we compared complementary DNA (cDNA) and genomic sequence from wild-type and gol^b1 embryos. A C→A nucleotide transversion that converts Tyr²⁰⁸ to a stop codon was found in gol^b1 cDNA clones (GenBank accession number AY682554) and verified by sequencing gol^b1 genomic DNA (fig. S3C). Conceptual translation of the mutant sequence predicts the truncation of the golden polypeptide to about 40% of its normal size, with loss of the central hydrophilic loop and the C-terminal cluster of potential transmembrane domains.

In wild-type embryos, the RNA expression pattern of slc24a5 (Fig. 3A) resembled that of the melanin biosynthesis marker dct (Fig. 3B), consistent with expression of slc24a5 in melanophores and RPE. In contrast, slc24a5 expression was nearly undetectable in golden embryos (Fig. 3C), the expected result of nonsense-mediated mRNA decay (19). The extent of protein deletion associated with the gol^b1 mutation, together with its low expression, suggests that gol^b1 is a null mutation. The persistence of melanosome morphogenesis, despite likely absence of function, suggests that golden plays a modulatory rather than essential role in the formation of the melanosome. The pattern of dct expression seen in golden embryos (Fig. 3D) resembles that of wild-type embryos, indicating that the golden mutation does not affect the generation or migration of melanophores.

Conservation of golden gene structure and function in vertebrate evolution. Comparison of golden cDNA (accession number AY538713) to genomic (accession number AY581204) sequences shows that the wild-type gene contains nine exons (fig. S2C) encoding 513 amino acids (fig. S3A). BLAST searches revealed that the protein is most similar to potassium-dependent sodium/calcium exchangers (encoded by the NCKX gene family), with highest similarity (68 to 69% amino acid identity) to murine Slc24a5 (accession number BAC40800) and human SLC24A5 (accession number NP_995322) (fig. S3B). The zebrafish golden gene shares less similarity with other human NCKX genes (35 to 41% identity to SLC24A1 to SLC24A4) or sodium/calcium exchanger (NCX) genes (26 to 29% identity to SLC8A1 to SLC8A3). Shared intron/exon structure and gene order (slc24a5, myef2, ctxn2, and slc12a5) between fish and mammals further supports the conclusion that the zebrafish golden gene and SLC24A5 are orthologs. The high sequence similarity among the orthologous sequences from fish and mammals (fig. S3A) suggested that function may also be conserved. The ability of human SLC24A5 mRNA to rescue melanin pigmentation when injected into golden zebrafish embryos (Fig. 2, L and M, and fig. S4) demonstrated functional conservation of the mammalian and fish polypeptides across vertebrate evolution.

Tissue-specific expression of Slc24a5. Quantitative reverse transcriptase PCR (RT-PCR) was used to examine Slc24a5 expression in normal mouse tissues and in the B16 melanoma cell line (Fig. 3E). Slc24a5 expression varied 1000-fold between tissues, with concentrations in skin and eye at least 10-fold higher than in other tissues. The mouse melanoma showed ∼100-fold greater expression of Slc24a5 compared with normal skin and eye. These results suggest that mammalian Slc24a5, like zebrafish golden, appears to be highly expressed in melanin-producing cells.

Model for the role of SLC24A5 in pigmentation. SLC24A5 shares with other members of the protein family a potential hydrophobic signal sequence near the amino terminus and 11 hydrophobic segments, forming two groups of potential transmembrane segments separated by a central cytoplasmic domain. This structure is consistent with membrane localization, although the specific topology of these proteins remains controversial (20). Elucidation of the specific role of this exchanger in melanosome morphogenesis requires knowledge of its subcellular localization and transport properties. Although previously characterized members of the NCKX and NCX families have been shown to be plasma membrane proteins (21), the melanosomal phenotype of golden suggested the possibility that the slc24a5 protein resides in the melanosome membrane. To distinguish between these alternatives, confocal microscopy was used to localize green fluorescent protein (GFP)– and hemagglutinin (HA)-tagged derivatives of zebrafish slc24a5 in MNT1, a constitutively pigmented human melanoma cell line (22). Both slc24a5 fusion proteins displayed an intracellular pattern of localization (Fig. 4, A and B), which is distinct from that of a known plasma membrane control (Fig. 4C). The HA-tagged protein showed phenotypic rescue of the golden phenotype (Fig. 4D), indicating that tag addition did not abrogate its function. Taken together, these results indicate that the slc24a5 protein functions in intracellular, membrane-bound structures, consistent with melanosomes and/or their precursors.

Several observations suggest a model for the involvement of slc24a5 in organellar calcium uptake (Fig. 4E). First, the intracellular localization of the slc24a5 protein suggests that it affects organellar, rather than cytoplasmic, calcium concentrations, in contrast with other members of the NCX and NCKX families. Second, the accumulation of calcium in mammalian melanosomes appears to occur in a transmembrane pH gradient–dependent manner (23). Third, several subunits of the vacuolar proton adenosine triphosphatase (V-ATPase) and at least two intracellular sodium/proton exchangers have also been localized to melanosomes (24, 25). In the model, active transport of protons by the V-ATPase is coupled to slc24a5-mediated calcium transport via a sodium/proton exchanger. The melanosomal phenotype of the zebrafish golden mutant suggests that the calcium accumulation predicted by the model plays a role in melanosome morphogenesis and melanogenesis. The observations that processing of the melanosomal scaffolding protein pmel17 is mediated by a furin-like protease (26) and that furin activity is calcium-dependent (27) are consistent with this view. The role of pH in melanogenesis has been studied far more extensively than that of calcium, with alterations in pH affecting both the maturation of tyrosinase and its catalytic activity (25, 28). The interdependence of proton and calcium gradients in the model may thus provide a second mechanism, in addition to calcium-dependent melanosome morphogenesis, by which the activity of slc24a5 might affect melanin pigmentation.

Role of SLC24A5 in human pigmentation. To evaluate the potential impact of SLC24A5 on the evolution of human skin pigmentation, we looked for polymorphisms within the gene. We noted that the G and A alleles of the single nucleotide polymorphism (SNP) rs1426654 encoded alanine or threonine, respectively, at amino acid 111 in the third exon of SLC24A5. This was the only coding SNP within SLC24A5 in the International Haplotype Map (HapMap) release 16c.1 (29). Sequence comparisons indicate the presence of alanine at the corresponding position in all other known members of the SLC24 (NCKX) gene family (fig. S5). The SNP rs1426654 had been previously shown to rank second (after the FY null allele at the Duffy antigen locus) in a tabulation of 3011 ancestry-informative markers (30). The allele frequency for the Thr¹¹¹ variant ranged from 98.7 to 100% among several European-American population samples, whereas the ancestral alanine allele (Ala¹¹¹) had a frequency of 93 to 100% in African, Indigenous American, and East Asian population samples (fig. S6) (29, 30). The difference in allele frequencies between the European and African populations at rs1426654 ranks within the top 0.01% of SNP markers in the HapMap database (29), consistent with the possibility that this SNP has been a target of natural or sexual selection.

A striking reduction in heterozygosity near SLC24A5 in the European HapMap sample (Fig. 5A) constitutes additional evidence for selection. The 150-kb region on chromosome 15 that includes SLC24A5, MYEF2, CTNX2, and part of SLC12A1 has an average heterozygosity of only 0.0072 in the European sample, which is considerably lower than that of the non-European HapMap samples (0.175 to 0.226). This region, which contains several additional SNPs with high-frequency differences between populations, was the largest contiguous autosomal region of low heterozygosity in the European (CEU) population sample (Fig. 5B). This pattern of variation is consistent with the occurrence of a selective sweep in this genomic region in a population ancestral to Europeans. For comparison, diminished heterozygosity is seen in a 22-kb region encompassing the 3′ half of MATP (SLC45A2) in European samples, and more detailed analysis of this genomic region shows evidence for a selective sweep (31). However, the gene for agouti signaling protein (ASIP), which is known to be involved in pigmentation differences (32), shows no such evidence.

The availability of samples from two recently admixed populations, an African-American and an African-Caribbean population, allowed us to determine whether the rs1426654 polymorphism in SLC24A5 correlates with skin pigmentation levels, as measured by reflectometry (33). Regression analysis using ancestry and SLC24A5 genotype as independent variables revealed an impact of SLC24A5 on skin pigmentation (Fig. 6). Despite considerable overlap in skin pigmentation between genotypic groups, regression lines for individuals with GG versus AG and GG versus AA genotypes were separated by about 7 and 9.5 melanin units, respectively (Fig. 6A). These differences are more evident in plots of skin pigmentation separated by genotype (Fig. 6B). SLC24A5 genotype contributed an estimated 7.5, 9.5, or 11.2 melanin units to the differences in melanin pigmentation among African-Americans and African-Caribbeans in the dominant, unconstrained (additive effect plus dominance deviation), or additive models, respectively.

The computer program ADMIXMAP provides a test of gene effect that corrects for potential biases caused by uncertainty in the estimation of admixture from marker data (34). Score tests for association of melanin index with the SLC24A5 polymorphism were significant in both African-American (P = 3 × 10–6) and African-Caribbean population subsamples (P = 2 × 10–4). The effect of SLC24A5 on melanin index is between 7.6 and 11.4 melanin units (95% confidence limits). The data suggest that the skinlightening effect of the A (Thr) allele is partially dominant to the G (Ala) allele. Based on the average pigmentation difference between European-Americans and African-Americans of about 30 melanin units (33), our results suggest that SLC24A5 explains between 25 and 38% of the European-African difference in skin melanin index.

Relative contributions of SLC24A5 and other genes to human pigment variation. Our estimates of the effect of SLC24A5 on pigmentation are consistent with previous work indicating that multiple genes must be invoked to explain the skin pigmentation differences between Europeans and Africans (5, 35). Significant effects of several previously known pigmentation genes have been demonstrated, including those of MATP (36), ASIP (32), TYR(33), and OCA2 (33), but the magnitude of the contribution has been determined only forASIP, which accounts for ≤4 melanin units (32). MATP may have a larger effect (37), but it can be concluded that much of the remaining difference in skin pigmentation remains to be explained.

Variation of skin, eye, and hair color in Europeans, in whom a haplotype containing the derived Thr¹¹¹ allele predominates, indicates that other genes contribute to pigmentation within this population. For example, variants in MC1R have been linked to red hair and very light skin [reviewed in (37)], whereas OCA2 or a gene closely linked to it is involved in eye color (7, 38). The lightening caused by the derived allele of SLC24A5 may be permissive for the effect of other genes on eye or hair color in Europeans.

Because Africans and East Asians share the ancestral Ala¹¹¹ allele of rs1426654, this polymorphism cannot be responsible for the marked difference in skin pigmentation between these groups. Although we cannot rule out a contribution from other polymorphisms within this gene, the high heterozygosity in this region argues against a selective sweep in a population ancestral to East Asians. It will be interesting to determine whether the polymorphisms responsible for determining the lighter skin color of East Asians are unique to these populations or shared with Europeans.

The importance of model systems in human gene discovery. Our identification of the role of SLC24A5 in human pigmentation began with the positional cloning of a mutation in zebrafish. Typically, the search for genes associated with specific phenotypes in humans results in multiple potential candidates. Our results suggest that distinguishing the functional genes from multiple candidates may require a combination of phylogenetic analysis, nonmammalian functional genomics, and human genetics. Such cross-disciplinary approaches thus appear to be an effective way to mine societal benefit from our investment in the human genome.

Supporting Online Material

www.sciencemag.org/cgi/content/full/310/5755/1782/DC1

Materials and Methods
Figs. S1 to S7
References

References and Notes

1. R. N. Kelsh, Pigment Cell Res. 17, 26 (2004).

2. N. P. M. Smit et al., Photochem. Photobiol. 74, 424 (2001).

3. A. T. Hewitt, R. Adler, in Retina, S. J. Ryan, Ed. (The C. V. Mosby Company, St. Louis, MO, 1989), p. 57.

4. R. A. King, V. J. Hearing, D. J. Creel, W. S. Oetting, in The Metabolic and Molecular Bases of Inherited Disease, C. R. Scriver, A. L. Beaudet, W. S. Sly, D. Valle, Eds. (McGraw-Hill, St. Louis, MO, 2001), p. 5587.

5. G. S. Barsh, PLoS Biol. 1, 19 (2003).

6. J. L. Bolognia, S. J. Orlow, in Dermatology, J. L. Bolognia, J. L. Jorizzo, R. P. Rapini, Eds. (Mosby, London, 2003), p. 935.

7. R. A. Sturm, T. N. Frudakis, Trends Genet. 20, 327 (2004).

8. S. Fukamachi, A. Shimada, A. Shima, Nat. Genet. 28, 381 (2001).

9. G. Streisinger, C. Walker, N. Dower, D. Knauber, F. Singer, Nature 291, 293 (1981).

10. M. C. Mullins, M. Hammerschmidt, P. Haffter, C. Nusslein-Volhard, Curr. Biol. 4, 189 (1994).

11. T. Nguyen et al., J. Invest. Dermatol. 119, 1156 (2002) .

12. T. Nguyen, M. Wei, J. Invest. Dermatol. 122, 452 (2004).

13. M. Gregory, P. Jagadeeswaran, Blood Cells Mol. Dis. 28, 418 (2002).

14. N. Shimoda et al., Genomics 58, 219 (1999).

15. C. Walker, G. Streisinger, Genetics 103, 125 (1983).

16. C. T. Amemiya, L. I. Zon, Genomics 58, 211 (1999).

17. A. Nasevicius, S. C. Ekker, Nat. Genet. 26, 216 (2000).

18. B. W. Draper, P. A. Morcos, C. B. Kimmel, Genesis 30, 154 (2001).

19. L. E. Maquat, Curr. Biol. 12, R196 (2002).

20. K. D. Philipson, D. A. Nicoll, Annu. Rev. Physiol. 62, 111 (2000).

21. X. Cai, J. Lytton, Mol. Biol. Evol. 21, 1692 (2004).

22. M. Cuomo et al., J. Invest. Dermatol. 96, 446 (1991).

23. R. Salceda, G. Sanchez-Chavez, Cell Calcium 27, 223 (2000).

24. V. Basrur et al., Proteome Res. 2, 69 (2003).

25. D. R. Smith, D. T. Spaulding, H. M. Glenn, B. B. Fuller, Exp. Cell Res. 298, 521 (2004).

26. J. F. Berson et al., J. Cell Biol. 161, 521 (2003).

27. G. Thomas, Nat. Rev. Mol. Cell Biol. 3, 53 (2002).

28. H. Watabe et al., J. Biol. Chem. 279, 7971 (2004).

29. The International HapMap Consortium, Nature 426, 789 (2003); available at www.hapmap.org, release 16b, 31 May 2005.

30. M. W. Smith et al., Am. J. Hum. Genet. 74, 1001 (2004).

31. M. Soejima et al., Mol. Biol. Evol., doi: 10.1093/molbev/ msj018 (2005).

32. C. Bonilla et al., Hum. Genet. 116, 402 (2005).

33. M. D. Shriver et al., Hum. Genet. 112, 386 (2003).

34. C. J. Hoggart et al., Am. J. Hum. Genet. 72, 1492 (2003).

35. G. A. Harrison, G. T. Owen, Ann. Hum. Genet. 28, 27 (1964).

36. J. Graf, R. Hodgson, A. van Daal, Hum. Mutat. 25, 278 (2005).

37. K. Makova, H. Norton, Peptides 26, 1901 (2005).

38. T. Frudakis et al., Genetics 165, 2071 (2003).

39. We thank P. Hubley for excellent management of our zebrafish facility; B. Blasiole, W. Boehmler, A. Sidor, and J. Gershenson for experimental assistance; R. Levenson, B. Kennedy, G. Chase, J. Carlson, and E. Puffenberger for helpful discussions; L. Rush for help on the cover design; and V. Hearing and M. Marks for MNT1 cells. Supported by funding from the Jake Gittlen Memorial Golf Tournament (K.C.), NSF (grant MCB9604923 to K.C.), NIH (grants CA73935, HD40179, and RR017441 to K.C.; HD37572 to D.G.; HG002154 to M.S.; EY11308 to N.M.; and HL077910 to P.J.), the Pennsylvania Tobacco Settlement Fund (K.C.), and the Natural Sciences and Engineering Research Council of Canada (E.P.). This work is dedicated to the open, trusting, and generous atmosphere fostered by the late George Streisinger.