A β-Defensin Mutation Causes Black Coat Color in Domestic Dogs

Linkage and association mapping of the K locus

We showed previously that the K^B mutation mapped to a 12-Mb interval on the distal end of dog chromosome 16 (CFA16), between markers REN292N24 and FH3592 (14). We ascertained additional kindreds segregating K^B, k^br, and k^y to refine the map location [see supporting online material (SOM) text] and defined overlapping critical regions of 3.8 and 7.6 Mb for the K^B and k^br mutations, respectively (Fig. 1Afigs. S1 to S4).

We used an association-based strategy to narrow the critical region. Because most breeds were derived in the past 200 years from small founding populations (15), mutations within a breed are expected to be identical by descent and share extended haplotypes. In Boxers and Great Danes, we identified broad peaks of significant association (Bonferroni-corrected P value < 0.05) that extended over 1.9 Mb and 320 kb, respectively (Fig. 1, B to D). Sixteen genes have been annotated to the region of significant association in Great Danes, including a gene cluster that encodes 12 β-defensins (16): small antimicrobial peptides that are secreted mainly by epithelial cells (17, 18). We sequenced the mature protein-coding regions for nine members of the β-defensin cluster (those known at the time) in dogs carrying K^B and/or k^y and identified several polymorphisms concordant with the K^B allele, including a 3–base pair (bp) deletion in the second exon of CBD103, the ortholog of human DEFB103, that predicts an in-frame glycine deletion (ΔG23).

To evaluate the extent to which the ΔG23 polymorphism distinguishes K^B versus k^y more broadly, we examined dogs from 38 breeds that could be classified into two categories with regard to their putative K locus genotype (SOM text). Among 454 dogs, there were 13 cases where the ΔG23 polymorphism did not correlate with coat-color phenotype. However, sequencing of Agouti and Mc1r revealed that each discordant case could be explained by known epistatic interactions (19, 20) (table S3). These results indicate that K^B alleles in all breeds are probably identical by descent and suggest that the ΔG23 polymorphism or a closely linked variant in complete linkage disequilibrium (LD) is the K^B mutation.

Short-range haplotype and resequencing analysis of K^B-bearing chromosomes

By contrast to the pattern of LD within breeds, which affords a powerful approach for association mapping with megabase resolution, the pattern of LD across breeds is more fine-grained and therefore provides the opportunity for high-resolution haplotype mapping when mutations in different breeds are identical by descent (21). We identified 28 polymorphisms [22 single-nucleotide polymorphisms (SNPs) and six indels including the ΔG23 polymorphism] in a 20-kb interval surrounding CBD103 that were then used to infer short-range haplotypes for 14 K^B-bearing and 16 k^y-bearing chromosomes selected from seven breeds (SOM text). We observed six “parental” k^y-bearing and five “parental” K^B-bearing chromosomes (depicted in yellow and blue, respectively, in Fig. 2A). We also identified eight chromosomes that carried a single ancestral recombination event, which together defined a maximal interval for K^B of 9146 bp (Fig. 2A). Complete resequencing of this interval (except for three homopolymer tracts) in five k^y/k^y animals, one K^B/k^y animal, and four K^B/K^B animals from seven breeds revealed two polymorphisms besides ΔG23 that are perfectly concordant with K locus genotype (S104 and S105) (Fig. 2A and table S2).

The 9146-bp interval contains both exons of CBD103, the first exon of dog expressed sequence tag (EST) CX990240, and dog EST CO665262 (Fig. 2). However, several considerations indicate that CBD103 is, indeed, the K locus and that the ΔG23 deletion in CBD103 is the K^B mutation. First, the other two transcribed elements in the critical interval are represented in the database by single ESTs and are not known to encode proteins or to be expressed in the skin. Second, S104 and S105 lie in a long terminal repeat element that is 3 kb upstream of the first exon of CBD103 and have no effect on mRNA levels of CBD103 (Fig. 3A). Finally, as discussed further below, CBD103 is highly expressed in skin, the ΔG23 deletion affects CBD103 protein function, and pharmacologic studies reveal that CBD103 can modulate melanocortin signaling.

The preceding discussion has referred to K locus variation as though k^y is ancestral, whereas K^B is derived: a hypothesis based on the comparative genetic distribution of coat-color phenotypes and inheritance patterns. Considerations based on sequence alignments confirm this hypothesis: Mammalian CBD103 orthologs that we identified from the available genome sequence are each 67 amino acids in length, and the optimal sequence similarity alignment contains no gaps or insertions (Fig. 2B), indicating that ΔG23 and consequently the K^B mutation occurred specifically within the canid lineage.

Expression of dog defensins in skin and in transfected keratinocytes

We isolated RNA from the skin of a k^y/k^y Doberman Pinscher and a K^B/k^y mixed-breed dog and surveyed the expression of the 19 β-defensin genes that are clustered on chromosomes 16 or 25 by reverse transcription polymerase chain reaction (RT-PCR) (16). Expression was detectable only for two genes: CBD1 and CBD103 (fig. S6). We then used quantitative RT-PCR to measure levels of skin mRNA from four K^B/k^y samples and four k^y/k^y samples, which were all from mixed-breed dogs, and found no effect of K locus genotype on levels of CBD1, CBD103, or Agouti mRNA (Fig. 3A).

Available antisera against human DEFB103 are unable to detect the endogenous dog protein by Western blotting or immunohistochemistry; therefore, we generated epitope-tagged expression constructs for each allele (CBD103V5 and CBD103DG23V5) and studied their patterns and levels of protein expression after transfection of cultured mouse keratinocytes.

In cell extracts analyzed by Western blotting, antisera against the V5 epitope detect a single fragment whose size (about 8 kD) corresponds to the expected molecular mass of the tagged protein after signal peptide cleavage; in media, an additional slightly smaller band is present, which suggests additional processing (Fig. 3B). The relative ratios of the two bands are similar in media from keratinocytes transfected with either construct; however, the total amount of immunodetectable protein in media was significantly greater for CBD103DG23V5 as compared with that for CBD103V5 (P = 0.0021, Cochran-Mantel-Haenszel chi-square test). Thus, loss of the N-terminal glycine from CBD103 does not affect intracellular processing but allows more of the mature protein to accumulate in the media and/or extracellular space.

CBD103 activity in vivo and in vitro

To further explore the function of CBD103 in an experimental genetic system, we generated transgenic mice in which a cDNA encoding either the K^B or the k^y allele was driven by a strong and widely expressed promoter (22). We chose a genetic background that normally has Agouti-banded hairs and observed that two transgenic founders generated with the CBD103ΔG23 cDNA (the K^B allele) displayed a predominantly black coat with small patches of banded hair (Fig. 4). Unexpectedly, the normal CBD103 cDNA (the k^y allele) also produced transgenic mice with a black coat in 20 out of 21 founders (Fig. 4). Furthermore, we observed that transgenic animals were smaller than their nontransgenic littermates. By 2 weeks of age, female transgenic animals were easily recognized by their dark coat and small size; in adult mice, reduced body weight persists in both males and females (fig. S7).

These considerations suggest three possible mechanisms by which CBD103 might act to modulate melanocortin signaling: (i) by binding to and activating the Mc1r, (ii) by binding to the Mc1r and preventing its inhibition by Agouti protein, or (iii) by binding to Agouti protein leading to its sequestration and/or degradation. To distinguish among these ideas, we generated synthetic forms of CBD103 and tested their ability to interact with the Mc1r and agouti signaling protein–YY (ASIP-YY), a synthetic version of the C terminus of Agouti protein that behaves as a competitive antagonist of α-MSH at the Mc1r and Mc4r (23).

The CBD103 k^y allele predicts a mature peptide (after signal sequence cleavage) of 45 amino acids that contains six cysteine residues and begins with the glycine that is deleted in the K^B allele (Fig. 2B). The predicted signal sequence cleavage site is supported by biochemical studies of the orthologous human protein [known as DEFB103 or human β-defensin 3 (HBD3)] purified from human tissues (24). We synthesized the 45-residue k^y form and 44-residue K^B form of the dog protein (hence referred to as CBD103 and CBD103DG23, respectively), allowed oxidative refolding, and used mass spectrometry and high-performance liquid chromatography to confirm recovery of a single congener with three intrachain disulfide bonds.

[Nle⁴, d-Phe⁷]-α–MSH (NDP-MSH), a potent derivative of α-MSH, stimulates robust accumulation of cyclic adenosine monophosphate (cAMP) in melanocytes (25). However, neither CBD103 nor CBD103ΔG23 has any effect on cAMP accumulation (Fig. 5A), indicating that CBD103 is not a conventional Mc1r agonist. To test for receptor binding, we transiently transfected human embryonic kidney (HEK) 293 cells with Mc1r expression constructs (22) and used europium-labeled NDP-MSH (Eu-NDP-MSH) as a fluorescent tracer. We used saturation binding assays to first calculate the affinity of Eu-NDP-MSH for the dog Mc1r (Fig. 5B) and then carried out displacement assays with progressively increasing concentrations of CBD103, CBD103ΔG23, or ASIP-YY. All three peptides exhibited qualitatively similar profiles characteristic for competitive binding to a single high-affinity site (Fig. 5C and Table 1). Quantitatively, inhibition constant (K_i) values estimated from the data depicted in Fig. 5C were 150.6 nM for CBD103 and 34.2 nM for CBD103ΔG23. These estimates varied according to experimental conditions, but CBD103ΔG23 consistently exhibited higher affinity for the dog Mc1r than did CBD103 (mean of fivefold across four paired experiments; Table 1). In these same experiments, ASIP-YY exhibited K_i values of 0.51 to 0.95 nM for the dog Mc1r, in the same range as reported previously for ASIP-YY at the human Mc1r (23). Using quantitative RT-PCR, we found that the levels of CBD103 mRNA in total skin were ~300-fold greater than that of Agouti mRNA; to the extent that this difference reflects protein abundance, the levels of CBD103ΔG23 in dog skin are likely to be much greater than those of Agouti protein, which is consistent with a model in which CBD103ΔG23 competitively inhibits the ability of Agouti protein to antagonize Mc1r signaling.

Finally, we investigated a potential interaction between Agouti protein and CBD103 by preincubating ASIP-YY with either CBD103 or CBD103ΔG23 before the binding assay, and we found no evidence for a functional interaction (fig. S8). We also examined the effects of mixing the two peptides on two-dimensional nuclear magnetic resonance spectra and found no evidence for a structural interaction. Taken together, these observations indicate that CBD103 is a high-affinity ligand for the Mc1r and that the ΔG23 mutation may cause a black coat color in dogs via two different mechanisms: increased affinity for the Mc1r and increased availability of the mature protein in vivo.

Other mammals, other melanocortin receptors, and other defensins

Dogs are the only mammals known for which dominant inheritance of black coat color is not caused by a Mc1r mutation. However, we observed that both CBD103 and CBD103ΔG23 are high-affinity ligands for the mouse (Fig. 5D) and the human (Fig. 5E) Mc1r, with estimated K_i values that were, unexpectedly, lower than that observed at the dog Mc1r (Table 1). Furthermore, unlike at the dog Mc1r, the affinity of CBD103 as compared to that of CBD103ΔG23 was similar at both the mouse and the human Mc1r (Table 1), which may explain why expression of either allele caused a black coat in transgenic mice (Fig. 4).

Given the effects of the CBD103 transgenes on body size, we investigated the potential for cross-talk between β-defensins and other melanocortin receptors. We found that CBD103ΔG23 binds to the human Mc4r with an intermediate affinity (K_i = 104.5 nM) between that of CBD103ΔG23 and CBD103 for the dog Mc1r (Table 1 and Fig. 5E). We also synthesized HBD3, the human ortholog of canine CBD103, and observed high-affinity binding (K_i = 13.8 nM) to the human Mc1r (Table 1 and Fig. 5F). Finally, we tested human β-defensin 1 (HBD1), which lies in the same cluster as HBD3 but is more distantly related; HBD1 also exhibited high-affinity binding to the human Mc1r (K_i = 30 nM) (Table 1 and Fig. 5F).

Defensins and melanocortin signaling

β-defensin genes exhibit strong signatures of diversifying selection (26, 27), which is consistent with their proposed role as endogenous antibiotics. Confirming and studying this role in vivo, however, have been challenging; the results of mouse knockout experiments have been reported for only one gene, Defb1, and revealed no obvious effects on viability or innate defense (28). By contrast, the CBD103 K^B mutation causes a marked change in hair color that is readily apparent among a large number of dog breeds with diverse genetic backgrounds. We propose that β-defensin genes play a broader role than currently envisaged, in which the ability to modulate melanocortin receptor signaling may have been selected during vertebrate evolution to provide camouflage and/or adaptive visual cues. From this perspective, evolutionary lability of the β-defensins with regard to both diversifying selection and copy number variation (26, 27, 29) may represent a response to a shifting spectrum of environmental challenges that includes microbial pathogens, carnivorous predators, and nutrient availability. More generally, our biochemical studies indicate that at least two human β-defensins, HBD1 and HBD3, may also modulate melanocortin receptor signaling in vivo. Both genes are expressed by a broad range of epithelial and other tissues (18) and could therefore act not only on Mc1r but also on other melanocortin receptors, including Mc5r, for which an endogenous melanocortin ligand has not yet been identified. The range of K_i values that we observed suggests that CBD103ΔG23, and potentially HBD1 or HBD3, modulate pigmentation in the low nanomolar range; however, the antimicrobial effects of β-defensins are typically observed in the low micromolar range (30). Finally, several β-defensins promote chemotaxis of immune cells, and although there is controversy regarding which receptors are involved, there is a growing consensus that β-defensins contribute to adaptive as well as to innate immunity (31–33). From this perspective, several studies have pointed to a role for Mc1r signaling in immune cell function (34), which could be mediated by β-defensins acting as melanocortin receptor ligands.

Modulation of melanocortin receptor signaling by β-defensins may also help explain the apparent paradox that mutations of Pomc have a relatively minor effect on pigmentation (11, 35–37). These observations have been often interpreted as evidence that melanocortin receptors have high levels of constitutive activity, but our results suggest that β-defensins may raise “basal” levels of melanocortin receptor signaling in the absence of melanocortin peptides.

Much of the early work on melanocortin signaling was driven by ideas associated with the way in which Pomc had been discovered and named; additional biological functions of melanocortins in behavior and energy homeostasis did not become apparent for decades. An analogous situation may apply in the case of β-defensins, which underscores the utility of phenotype-driven genetics to provide an agnostic view of gene function.

Supplementary Material