Epistatic and combinatorial effects of pigmentary gene mutations in the domestic pigeon

Abstract

Understanding the molecular basis of phenotypic diversity is a critical challenge in biology, yet we know little about the mechanistic effects of different mutations and epistatic relationships among loci that contribute to complex traits. Pigmentation genetics offers a powerful model for identifying mutations underlying diversity, and for determining how additional complexity emerges from interactions among loci. Centuries of artificial selection in domestic rock pigeons have cultivated tremendous variation in plumage pigmentation through the combined effects of dozens of loci. The dominance and epistatic hierarchies of key loci governing this diversity are known through classical genetic studies [1-6], but their molecular identities and the mechanisms of their genetic interactions remain unknown. Here we identify protein-coding and cis-regulatory mutations in Tyrp1, Sox10, and Slc45a2 that underlie classical color phenotypes of pigeons, and present a mechanistic explanation of their dominance and epistatic relationships. We also find unanticipated allelic heterogeneity at Tyrp1 and Sox10, indicating that color variants evolved repeatedly though mutations in the same genes. These results demonstrate how a spectrum of coding and regulatory mutations in a small number of genes can interact to generate substantial phenotypic diversity in a classic Darwinian model of evolution [7].

RESULTS AND DISCUSSION

In the domestic rock pigeon (Columba livia), hundreds of years of accumulated experience by amateur and professional geneticists provide strong evidence that many complex color traits can be partitioned into combined effects of multiple loci, and that the same loci control similar traits across breeds [6]. The classical major color locus (B) is a sex-linked gene that confers one of three “base” colors [1-5]: wild-type blue/black (B⁺), ash-red (B^A), and brown (b) (Fig. 1A-C). The B^A allele is dominant to B⁺ and b, while b is recessive to the others.

Figure 1. Common color phenotypes of domestic rock pigeons.

(A-C) Allelic variation at sex-linked major color locus (B) in order of dominance. (D) Recessive autosomal mutation, recessive red (e). (E-H) Recessive allele at another sex-linked locus, dilute (d) reduces color intensity to generate ash-yellow (E), dun (F), khaki (G), or recessive yellow (H) phenotypes. Black chevrons indicate order of dominance among alleles, red chevron indicates epistasis between loci. Breeds shown: (A,B,E-G) showtype racing homer, (C) mookee, (D) parlor roller, (H) Birmingham roller.

Blue/black and brown phenotypes result from high amounts of eumelanin and low amounts of pheomelanin, while melanin ratios are reversed in ash-red birds [8]. In addition, the autosomal recessive mutation recessive red (e) acts epistatically to the B locus to elevate pheomelanin production, generating red plumage color irrespective of B locus genotype [2, 8] (Fig. 1D). Mutant alleles of a third locus, the sex-linked recessive dilute (d), interact additively with B and e to lighten plumage color and further enrich pigmentation diversity [1, 2, 8] (Fig. 1E-H). This detailed Mendelian understanding of key phenotypes provides a robust foundation to investigate how genes and alleles interact to generate color variation. However, the molecular basis of this diversity – including the identities of genes underlying major pigmentation variants, and a mechanistic explanation for their intra- and inter-locus interactions – remains unknown [9, 10].

Multiple mutations in Tyrp1 underlie base color variation in pigeons

Previously, we reported whole-genome sequences for 41 rock pigeons [11] with diverse color phenotypes. To investigate the molecular identity of the B color locus, we compared the genomes of 6 ash-red to 26 blue/black pigeons for coding changes associated with pigmentation phenotypes using the Variant Annotation, Analysis, and Search Tool (VAAST) [12]. A single gene achieved genome-wide significance: Tyrosinase-related protein 1 (Tyrp1) (P = 1.3×10⁻⁶; Fig. S1A), which encodes a key enzyme in the melanin synthesis pathway. All blue/black pigeons were homozygous G on the Tyrp1 sense strand at scaffold6:214991 (B⁺ allele), while ash-red pigeons were hetero- or homozygous for C (B^A allele), consistent with the dominant mode of inheritance of ash-red. The B^A mutation causes an alanine-to-proline substitution at codon 23 (A23P), corresponding to the cleavage site of the signal peptide (Fig. 2A). In addition to finding a single haplotype containing the B^A allele in our whole-genome panel (Fig. S1B), we found a perfect association between the dominant B^A mutation and the ash-red phenotype in an additional 49 ash-red birds from 20 breeds, and 105 blue/black or brown birds from 36 breeds (Fig. 2B). These results suggest that the ash-red mutation occurred only once and spread species-wide through selective breeding, similar to our previous finding that the same mutation in EphB2 underlies the head crest phenotype in multiple pigeon breeds [11].

Figure 2. Tyrp1 is the major color locus B in domestic pigeons.

(A) Schematic of the genomic Tyrp1 locus with putative B^A and b mutations. (B) Histogram of genotypes of pigeons displaying wild-type or ash-red phenotypes. (C) Schematic and Western blot analysis of cleavage of B⁺ and B^A TYRP1 proteins demonstrate reduced cleavage efficiency of the B^A allele. HA, N-terminal hemagglutinin epitope tag; V5, C-terminal V5 epitope tag. (D) Ultrastructural analysis of melanocytes from B⁺, B^A, and B⁺B^A feathers. Asterisk, premelanosome. Arrow, untreated melanosome. Arrowhead, DOPA-treated melanosomes. Scale bar = 500 nm. (E) Tyrp1 mRNA abundance in B⁺ and b³ feathers by qRT-PCR. Boxes in (C,E) span first to third quartiles; black line, median. *, P = 0.05; **, P < 0.002.

Quantitative RT-PCR analysis revealed that Tyrp1 mRNA levels from developing feathers of B⁺ and B^A pigeons were indistinguishable (Fig. S1C); however, the location of the B^A mutation at the highly conserved cleavage site of the signal peptide (Fig. S1E) suggested that cleavage efficiency might be affected. We therefore expressed N- and C-terminal tagged B⁺ and B^A TYRP1 proteins in cell culture, and found that cleavage efficiency was dramatically reduced by the B^A mutation (relative efficiency: B⁺ = 1 ± 0.18, B^A = 0.14 ± 0.04, n = 4 independent transfections each; P < 0.002) (Fig. 2C). Furthermore, spatial organization of pigment synthesis differed between B genotypes: premelanosomes in regenerating B⁺ feathers had a well-organized, lamellar matrix and melanosomes were darkly pigmented, while ash-red feathers (B^A and B⁺B^A) had a disorganized matrix and only lightly pigmented melanosomes (Fig. 2D). After incubation with the melanin precursor L-DOPA, melanosomes from both wild-type and ash-red birds became darkly pigmented, indicating normal catalytic activity of the melanogenic enzyme Tyrosinase (TYR) in ash-red birds. However, pigment synthesis in B⁺ feathers showed strongest staining localized to the limiting membrane of the melanosome (Fig. 2D), while staining was diffuse in melanosomes from B^A and B⁺B^A feathers. Thus, the striking reduction in TYRP1 cleavage efficiency may disrupt the spatial organization of pigment synthesis activity, providing insight into the molecular basis of dominance of the B^A allele. The dominant Light (B^lt) Tyrp1 allele of mice, a missense mutation near the same cleavage site, causes melanocyte death, probably through the accumulation of cytotoxic pigment intermediates [13]. Whether a similar accumulation of cytotoxins contributes to the pheomelanic phenotype of ash-red pigeons is unknown. However, unlike the mouse B^lt allele, the pigeon B^A allele results in a different kind and localization of melanin production rather than abrogation of melanogenesis.

In contrast to the single ash-red mutation, Tyrp1 sequences from 51 brown pigeons from 30 breeds revealed three different nonsense and frameshift mutations (R72X, b¹; 411-418del, b²; 893delA, b³) (Figs. 2A, S1D), predicted to be null alleles. Indeed, Tyrp1 mRNA abundance in b³ pigeons – the most common b allele in our sample – is greatly reduced or absent (relative expression: B⁺ = 1 ± 0.53, n = 4; b³ = 0.009 ± 0.005, n = 3; P = 0.05) (Fig. 2E), consistent with the activation of nonsense-mediated decay. This indicates that, in contrast to the single origin of the ash-red phenotype, brown color has evolved multiple times in pigeons. Several brown pigeons did not have any of the identified b alleles, raising the possibility that additional mutations might also cause brown feather color (Fig. S1D).

Together, our results demonstrate distinct effects of different mutations in the same gene, and also confirm the predicted orthology of the classical mouse and pigeon B loci [5, 14]. Our analyses suggest a model in which B^A is a neomorphic allele that alters processing of the mutant TYRP1 protein within the cell. Since TYRP1 can modulate TYR activity [15-18], we postulate that the B^A TYRP1 protein alters normal TYR functionality, resulting in an increased ratio of pheomelanin to eumelanin production. In contrast, Tyrp1 loss-of-function alleles b^1-3 cause brown pigment production, consistent with findings in other vertebrates [19].

Recurrent deletions of a Sox10 enhancer underlie recessive red

In addition to the dominant, sex-linked B^A allele, the autosomal locus recessive red (e) acts epistatically to B to increase pheomelanogenesis and decrease eumelanogenesis (Fig. 1D). VAAST scans for the e locus did not identify a strong candidate, suggesting that an unannotated structural variant, such as a large insertion or deletion, might underlie this phenotype. To identify candidates for e, we compared expression of several genes involved in melanin biosynthesis and found that the transcription factor Sox10 and one of its target genes, Tyrp1 (the B locus), were significantly downregulated in feathers of recessive red birds (Fig. 3A) (Sox10 relative expression: blue/black = 1 ± 0.62; recessive red = 0.14 ± 0.07, P = 0.001; Tyrp1 relative expression: blue/black = 1 ± 0.556, recessive red = 0.0001 ± 0.00006, P = 0.002; n = 6 for each). Other melanin biosynthesis genes did not show altered transcript levels (Fig. S2A), suggesting that a mutation directly or indirectly affecting Sox10 expression might underlie the recessive red phenotype.

Figure 3. Sox10 and Slc45a2 are the recessive red (e) and dilute (d) loci of domestic pigeons.

(A) qRT-PCR analysis of Sox10 and Tyrp1 in wild-type vs. recessive red feathers (see Fig. S2 for additional genes). (B) Schematic of deletions upstream of Sox10 in recessive red pigeons (e¹, e²), dark-brown (db) chicken [20], and Hry mutant mouse [21]. Red asterisk denotes conserved element deleted in all three species. Conservation track based on Multiz alignment to chicken, human, mouse, rat, opossum, Xenopus tropicalis, and zebrafish in UCSC Genome Browser (genome.ucsc.edu; chicken assembly v2.1 used as framework). (C) Histogram of genotypes of pigeons displaying wild-type or recessive red phenotypes. (D) Expression of SNPs in e² allele relative to E⁺ allele of Sox10 in feathers of E⁺e² heterozygous pigeons. Blue dashed line, normalized expression level of E⁺ allele. (E) Schematic of SLC45A2 protein with putative d mutation in red. Mutations in chicken and quail associated with lightened feather color in grey. Adapted from [28]. (F) Histogram of genotypes of pigeons displaying wild-type or dilute phenotypes. Boxes in (A,D) span first to third quartiles; black line, median. **, P ≤ 0.002; ***, P < 0.0001.

Deletions of a conserved Sox10 enhancer result in pigmentation defects in other vertebrates, including a lack of pigmentation in mouse and increased pheomelanin production in chicken [20, 21]. Alignment of the pigeon reference genome assembly (a recessive red Danish tumbler [11]) upstream of Sox10 to the orthologous regions of the chicken and zebra finch genomes identified a 7.5-kb deletion in the pigeon genome (Figs. 3B,S2B). Furthermore, four recessive red birds in our genome resequencing panel – but no wild-type birds – were homozygous for this deletion. Importantly, the deletions in pigeon, chicken, and mouse all span a conserved enhancer element that drives Sox10 expression in melanocytes [20, 22].

To test for broader association between the pigeon Sox10 enhancer deletion and recessive red, we genotyped 41 recessive red pigeons from 19 breeds and 103 wild-type pigeons from 45 breeds, and found that 21 recessive red birds (but no wild-type birds) were homozygous for the deletion harbored by the reference genome (e¹ allele, Figs. 3B,C). An additional 17 recessive red birds (but no wild-type birds) were homozygous for a second, 2.5-kb deletion (e²) that partially overlaps e¹, and the remaining 3 birds were heterozygous e¹e² (Fig. 3B,C). Since both pigeon deletions span the Sox10 melanocyte enhancer, we predicted that the reduction in Sox10 expression in recessive red birds was due to a cis-regulatory change. We therefore assayed allele-specific expression of E⁺ and e² alleles of Sox10 in E⁺e² heterozygous birds and found that the e² allele was expressed at only ~10% of E⁺ levels (SNP1 = 0.126 ± 0.055, SNP2 = 0.056 ± 0.043, SNP3 = 0.127 ± 0.059; P < 0.0001 for each, n = 10 E⁺e² birds) (Fig. 3D). Since, in heterozygotes, both the E⁺ and e² alleles are in the same cellular environment, this experiment confirmed that the reduction in Sox10 expression from the e² allele is due to a cis-acting mutation. Together, these genetic and expression results implicate the deletion of a Sox10 melanocyte enhancer as the molecular basis of recessive red in domestic pigeons (Fig. 3B). These results also demonstrate that the E (extension) loci of mammals (Mc1r) and pigeons (Sox10) are not orthologous [5, 9, 23-25]. Moreover, similar to the brown phenotype, recessive red has evolved more than once in pigeons. While we do not observe obvious phenotypic distinctions between e¹ and e² homozygotes, it is possible that the different deletions generate subtly different effects on color by altering other unidentified regulatory elements [26].

The epistatic relationship of e to B is now easily reconciled in light of their molecular identities and mutations: Sox10 directly regulates Tyrp1 expression in melanocytes [27] (Fig. 3A), explaining how loss of Sox10 expression abrogates phenotypic effects of Tyrp1 genotypes. Interestingly, the recessive red phenotype caused by Sox10 downregulation is distinct from the brown phenotype of Tyrp1 loss-of-function mutants, possibly owing to contributions of additional Sox10 regulatory targets or residual b allele activity.

Missense mutation in Slc45a2 is associated with color dilution

While the B and e loci affect pigment color, the sex-linked recessive dilute (d) reduces pigment quantity, further enriching pigmentation diversity [8] (Fig. 1E-H). To identify candidates for d, we compared the genomes of 5 birds with diluted feather color, and 31 birds with non-diluted pigment intensity using VAAST. A single gene achieved genome-wide significance: Solute carrier family 45 member 2 (Slc45a2, P = 2.65×10⁻⁶; Fig. S3A), which is associated with pigmentation phenotypes in diverse vertebrates, including other birds [28-32], but is not orthologous to the dilute locus in mouse (Myo5a) [33]. In pigeons, the d mutation causes a histidine-to-arginine substitution (H341R) at a highly conserved intramembrane residue of SLC45A2 (Figs. 3E, S3C). We genotyped an additional 59 diluted birds from 26 breeds and 67 non-diluted birds from 41 breeds, and found a strong (but not perfect) association between d genotypes and color intensity under a recessive model (Fisher’s exact test, P < 2.2×10⁻¹⁶) (Fig. 3F). Fourteen birds not homozygous for d had diluted feather color and one homozygote was reported to have non-diluted color. However, several other loci can cause either lightened (e.g. milky, reduced, and faded) or darkened (e.g. dirty, sooty, and smoky) pigmentation in pigeons [6], and it is expected that a broad hobbyist-identified sample should include birds with varied genetic bases for color intensity.

Mutations and color traits co-segregate in a controlled cross

As an independent test of our association analyses, we examined co-segregation of pigmentation phenotypes and our three candidate loci in progeny of a male pigeon doubly heterozygous for ash-red and dilute (B^AB⁺,D⁺d,E⁺E⁺) mated to two recessive yellow females (B⁺,d,e²e²; ZW females are hemizygous at the sex-linked B and D loci). As anticipated, segregation of the ash-red (dominant) and dilute phenotypes was observed in the F₁ and F₂ generations, while recessive red was only observed in the F₂. We genotyped all three generations for the candidate mutations in Tyrp1, Sox10, and Slc45a2, and found complete co-segregation between alleles at these loci and their respective pigmentation phenotypes (Fig. 4A). Coupled with our genetic association results, these transmission genetics data strongly support the molecular identities of the classical B^A (and b), e, and d alleles as mutations in Tyrp1, Sox10, and Slc45a2, respectively.

Figure 4. Segregation analysis and mechanistic model of common color phenotypes.

(A) Representative feathers from F₁ and F₂ offspring in a cross segregating B^A, e², and d mutations; numbers of birds with a given genotype are listed below each phenotype. (B) Schematic illustrating common feather colors in pigeons, and mutations responsible for their production. Other pigmentation genes (PGs) are probably also affected by a decrease in Sox10 expression, thereby causing differences between b (Tyrp1 loss-of-function) and e (Sox10 loss-of-function) phenotypes.

Few loci generate many phenotypes

Similar to the genetic architecture of dog coat morphology [34], a relatively small number of loci generate a wide range of plumage color phenotypes in pigeons (Fig. 4B). We found that coding changes at the base color locus B result in a neomorphic dominant allele (B^A, ash-red) that interferes with melanosome formation and localization of melanogenesis to cause one derived phenotype, and multiple recessive, putative null alleles (b^1-3, brown) that underlie another. The e mutation is epistatic to B genotypes due to a regulatory mutation in Sox10, which is a transcriptional regulator of Tyrp1. A mutation at the d locus influences the color of birds of all genotypes at B and e by reducing the quantity of pigment produced, generating an additional layer of genetic complexity and phenotypic diversity. We find evidence that some color phenotypes, such as ash-red, appear to have a single origin while others, such as brown and recessive red, originated multiple times.

Many color phenotypes in pigeons and other domestic animals result from artificial selection [35]. Nevertheless, the combination of coding and regulatory changes, combined effects of multiple loci on a common phenotypic output, as well as single and multiple origins of derived alleles, is reminiscent of the genetic architecture of a variety of adaptive traits in the wild [e.g., 26, 36, 37, 38]. Additionally, the specific genes we implicated in plumage color phenotypes in pigeons also contribute to both natural pigmentation diversity and skin disease in humans, including melanoma risk [39, 40]. Thus, by elucidating the complex interactions among these loci, we enrich our mechanistic understanding of adaptive and non-adaptive variation across species.

HIGHLIGHTS.

• Interactions among three genes control multiple color phenotypes in pigeons

• Pigmentation genes incur a spectrum of coding and regulatory mutations

• Some color variants evolved repeatedly though mutations in the same genes