Skip to main content
Browse Subject Areas

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

The origin and evolution of fibromelanosis in domesticated chickens: Genomic comparison of Indonesian Cemani and Chinese Silkie breeds

  • Anik Budhi Dharmayanthi,

    Affiliations Department of Evolutionary Studies of Biosystems, SOKENDAI (The Graduate University for Advanced Studies), Kanagawa, Japan, Museum Zoologicum Bogoriense, Research Center for Biology, Indonesian Institute of Science (LIPI), Cibinong, Indonesia

  • Yohei Terai,

    Affiliation Department of Evolutionary Studies of Biosystems, SOKENDAI (The Graduate University for Advanced Studies), Kanagawa, Japan

  • Sri Sulandari †,

    † Deceased.

    Affiliation Museum Zoologicum Bogoriense, Research Center for Biology, Indonesian Institute of Science (LIPI), Cibinong, Indonesia

  • M. Syamsul Arifin Zein,

    Affiliation Museum Zoologicum Bogoriense, Research Center for Biology, Indonesian Institute of Science (LIPI), Cibinong, Indonesia

  • Toyoko Akiyama,

    Affiliation Department of Biology, Keio University, Yokohama, Japan

  • Yoko Satta

    Affiliation Department of Evolutionary Studies of Biosystems, SOKENDAI (The Graduate University for Advanced Studies), Kanagawa, Japan


Like Chinese Silkie, Indonesian Ayam Cemani exhibits fibromelanosis or dermal hyperpigmentation and possesses complex segmental duplications on chromosome 20 that involve the endothelin 3 gene, EDN3. A genomic region, DR1 of 127 kb, together with another region, DR2 of 171 kb, was duplicated by unequal crossing over, accompanied by inversion of one DR2. Quantitative PCR and copy number variation analyses on the Cemani genome sequence confirmed the duplication of EDN3. These genetic arrangements are identical in Cemani and Silkie, indicating a single origin of the genetic cause of Fm. The two DR1s harbor two distinct EDN3 haplotypes in a form of permanent heterozygosity, although they remain allelic in the ancestral Red Jungle Fowl population and some domesticated chicken breeds, with their allelic divergence time being as recent as 0.3 million years ago. In Cemani and Silkie breeds, artificial selection favoring the Fm phenotype has left an unambiguous record for selective sweep that extends in both directions from tandemly duplicated EDN3 loci. This highly homozygous tract is different in length between Cemani and Silkie, reflecting their distinct breeding histories. It is estimated that the Fm phenotype came into existence at least 6600–9100 years ago, prior to domestication of Cemani and Silkie, and that throughout domestication there has been intense artificial selection with strength s > 50% in each breed.


With conspicuously diversified phenotypes with respect to morphological, physiological, and behavioral traits, domesticated animals are excellent model organisms for investigating underlying genetic changes as well as for elucidating the underlying evolutionary mechanisms. It is widely accepted that phenotypes currently observed in domesticated organisms are usually selected from variations that arose spontaneously in wild, ancestral populations. There have been several attempts to identify the genetic bases of such phenotypes and compare them with those of wild ancestors [15]. One of the common underlying ideas is that artificial selection reduces the level of genetic variability at linked neutral sites when a selected allele rapidly increases in frequency toward fixation (selective sweep) and/or is kept fixed in a breeding population for a relatively short period. On the basis of 2.8 million SNPs, the International Chicken Polymorphism Map Consortium [6] scanned the genomes of three chicken breeds (Broiler, Layer, and Silkie) and Red Jungle Fowl (RJF), the wild ancestor of domestic poultry. It was found that relatively high levels of genetic variation with nucleotide diversity π = 0.5% are maintained within chicken breeds; however, little evidence is provided for selective sweeps by adaptive (favored) alleles on length scales greater than 100 kb. One reason for the lack of such long-stretch signals could be the rather high recombination rate in chickens [7]. In a small-scale study of 32 introns randomly selected in two chicken breeds (Silkie and Koshamo or fighting cock), RJF and Green Jungle Fowl (GJF), Sawai et al. [8] also showed that domesticated chickens usually maintain nearly the same level of nucleotide diversity as their ancestral RJF population. The authors further argued that genomic regions that respond to domestication might be rather limited. However, re-sequencing of genomic DNA pools representing eight different populations of domesticated chickens and RJF demonstrated a number of regions under selective sweeps [9]. Another selective sweep analysis of feral Kauai chickens derived from domesticated populations identified genomic regions that are associated with comb mass, maternal brooding behavior and fecundity [10]. Unfortunately, however, these studies did not cover all interesting phenotypes of the various chicken breeds, including Silkie (S1 Fig).

In contrast to these genome-wide scan approaches, we took a candidate gene approach and focused on a particular phenotype known as fibromelanosis (Fm) or dermal hyperpigmentation [11]. Mutations of the Fm gene result in excessive accumulation of black pigment in the skin and several other tissues or organs such as blood vessels, muscles, gonads and tracheas. Chinese Silkie is one of the domesticated chicken breeds with the Fm. The phenotype is inherited in a Mendelian fashion with semidominance [12]. Recombinant analysis using Silkie and Black Minorca (a homozygote for the wild-type chromosome regarding Fm) located the genomic region of Fm between 10.2 Mb and 11.7 Mb on chromosome 20 [12, see also 13,14].

It has been established that the Fm mutation is positively correlated with the duplication of a segment that contains the EDN3 gene encoding endothelin 3 [12,14,15]. EDN3 is a major controller of neural crest cell movement and proliferation. Neural crest cells are pluripotent and thus can develop into several cell types, such as melanoblasts [1619]. Melanocytes, which differentiated from melanoblasts, produce eumelanin (black and dark pigment melanin) and phaeomelanin (colored melanin) in the skin, comb and other organs [20]. The amount of EDN3 mRNA in whole Silkie embryos at 18 days is approximately twice as high as that in wild-type chicken embryos [12,14]. Thus, EDN3 is the most likely candidate gene for such coloring phenotypes in Silkie as well as other domesticated animals, including cats [21] and cattle [22]. Indeed, PCR and next-generation sequencing (NGS) analyses of the Silkie genome unveiled segmental duplication in the Fm region [14,23]. Previously, Dorshorst et al. [14] showed that two regions (DR1 and DR2), separated by a 417 kb spacer, underwent inverted duplication. In the reference genome (Gallus_gallus_4.0,, DR1 is located at nt positions 11,111,559 to 11,238,796 and DR2 at positions 11,651,876 to 11,822,527 on chromosome 20. Each of the duplicated DR1s is 127 kb long, and contains not only EDN3 but also HIVEP3, SLMO2, ATP5E, and TUBB1, whereas each of the duplicated DR2s is 171 kb long and is devoid of genes.

Dermal hyperpigmentation is found in other domesticated chicken breeds, such as Ayam Cemani in Indonesia (S1 Fig), Kadakhnath in India, Black H’Mong in Vietnam, Argentinean Tuzo type in Argentina, and Svarthöna in Sweden. While they all show excessive melanin accumulation, the overall phenotypes of Cemani and other black chickens differ greatly from those of Silkie [13,14]. For instance, unlike Silkie, which shows fluffy feathering, Cemani shows black plumage and non-fluffy feathering. Moreover, comb shape in Cemani males is very different from that in Silkie males with rose combs. In light of these similarities and differences between Cemani and Silkie, Shinomiya et al. [12] and Dorshorst et al. [14] examined whether the Fm region in Cemani is functionally and structurally the same as that in Silkie. Shinomiya et al. [12] analyzed the progenies of sib-crosses of F1 hybrids between Cemani and Ayam Arab (a wild-type domesticated breed in Indonesia). Based on the copy number variation (CNV) observed in EDN3 by quantitative PCR (qPCR), they suggested EDN3 duplication in Cemani, but not in Ayam Arab. Similarly, using a PCR-based diagnostic test, Dorshorst et al. [14] found that the complex arrangement of DR1 and DR2 is shared among Silkie, Cemani, Black H’Mong and Svarthöna. However, because the two copies of DR1 and DR2 cannot be easily distinguished from each other by PCR or NGS, the precise genomic arrangement of these four regions has not fully been elucidated even in Silkie.

In this study, we compared the genomic structure, the pattern and level of DNA variation, and the evolutionary history of the Fm region between Cemani and Silkie. We paid particular attention to the genomic signature of artificial selection on a target gene, EDN3, and used it to estimate the duration and strength of artificial phenotypic selection.

Materials and methods

Ethics statement

This study was conducted in accordance with the animal research guidelines of SOKENDAI (The Graduate University for Advanced Studies), Hayama, Japan. The Research Ethics Committee of SOKENDAI approved the research protocol No.46 on June 16, 2016. Chicken blood and DNA samples were provided by the Indonesian Institute of Science (LIPI). Blood was sampled according to the procedure of animal welfare of the Museum Zoologicum Bogoriense (MZB), Division of Zoology, Research Center for Biology, LIPI, Indonesia. DNA samples were also obtained from Keio University via the Avian Bioscience Research Center (ABRC) of Nagoya University in Japan. Bird maintenance and blood collection were performed in accordance with the University-institutional guidelines for animal experiments.

In addition, Silkie DNA samples from the USA were obtained from the Japanese Society for the Study of H. I. H. Prince Akishinonomiya Collections (JSAC).

Chicken breeds and DNA samples

Most domestic chickens of Cemani, Silkie and other breeds as well as the two jungle fowl species (RJF and GJF) were collected at various locations in Indonesia (Table 1). Chickens for DNA isolation were collected at farms in rural areas of Java, Sumatra, Sulawesi and Nusa Tenggara, Indonesia, mainly in 2005–2010, although some were obtained in more recent years. During sample collections, we always carried a letter of assignment from LIPI; however, no explicit permission from the farmers was required as chickens are not protected animals in Indonesia.

Table 1. IDs of 75 sampled domesticated chickens and jungle fowl together with their collection sites and sample sizes.

Construction of Cemani genomic sequence library

Total DNA of one Cemani chicken (sample “Cemani 41”) was sheared into ~500-bp fragments using an M220 focused ultrasonicator™ (Covaris) and a genomic library NGS was prepared in accordance with the True Seq DNA PCR-free sample preparation protocol. Another genomic DNA library was prepared in accordance with protocols provided with the Illumina Nextera X, for nine regions, each of about 3 kb in length with ~200-kb intervals. Each 3-kb region was amplified by primers designed in-house (S1 Table) in a 30 μl reaction mixture (see S2 Table for the reaction conditions of PCR1). PCR products of each sample (5 μl) were pooled and purified using Agencourt AMpure XP (Beckman Coulter). The libraries were qualitatively and quantitatively verified using an Agilent Bioanalyzer and sequenced on the Illumina HiSeq2000 platform (Illumina).

Public data used

The chicken reference genome was downloaded in September 2015 from the UCSC Genome Browser ( and has the same sequence as that deposited in the GenBank database (Gallus_gallus_4.0). Additionally, full data sets of Silkie (accession numbers: SRX286765, SRX286766, SRX286773, SRX286776, and SRX286777, see ref. 23) and Taiwanese L2 (accession numbers: SRX286779-SRX286781, SRX286798, and SRX286799, see ref. 23) were retrieved from GenBank (

Amplification of duplication boundaries

The presence or absence of duplication boundaries was examined by PCR with two previously published primer pairs, A2 and B2 [14], on a 96-Well GeneAmp® PCR System 9700 from Applied Biosystems (see S2 Table for the detailed reaction conditions of PCR2).

Quantification of gene copy numbers

PCR products for EDN3 (113 bp, primer set qAS044) and uridine-cytidine kinase 1-like 1 (UCKL1) (124 bp, primer set q46) were ligated to the pMD20 vector (TaKaRa Japan). Plasmid DNA was extracted using alkaline lysis [24] and the concentration was determined using NanoVue spectrophotometer (GE Healthcare). Plasmid DNAs were diluted to 10−1 to 10−6 ng/μl in distilled water and were used to draw a standard curve for quantification. qPCR for absolute quantitative analysis was carried out with the SYBR® Premix Ex Taq™ II (Tli RNaseH Plus; TaKaRa). All reactions were run in triplicate on a Thermal Cycler Dice (Applied Biosystems), and the thermal cycling conditions were as indicated under “PCR3” in S2 Table.

Amplification and sequencing of EDN3

A ~1-kb genomic fragment encompassing exons 4 to 5 of EDN3 was amplified and sequenced with the previously reported primer pair AS044F and AS044R [12]. PCR was conducted under reaction conditions listed under “PCR4” in S2 Table. Amplified PCR products were purified by isopropanol precipitation and directly sequenced. For heterozygous sequences, the PCR products were cloned into pMD20, and eight clones for each product were sequenced using the BigDye Terminator Cycle Sequencing Kit (Applied Biosystems) with M4 and Rv universal primers on an ABI 3130xl sequencer (Applied Biosystems).

Reads of data from the chicken WGS and nine 3 kb regions, and CNV analysis of the Fm region

The CLC Genomic Workbench 8.0.3 ( was used to map the 3-kb region reads to the reference genome with 90% length and similarity fractions.

To analyze the WGS of Cemani, Silkie, and Taiwanese L2, low-quality bases were removed with the Trimmomatric software [25], using the following parameter settings; leading = 10, trailing = 10, sliding windows = 4:15, and minlen = 40. The Samtools workflow [2630] ( was used for mapping of the WGS data with 30X coverage.

To examine CNV in the Fm region, the reads from each of three pairs among Cemani, Silkie, and L2 WGS data were mapped onto nt positions 10,700,000–12,000,000 on chromosome 20 using the CNV-Seq software [31]. Default parameters (log2-threshold = 0.6, p-value = 0.01, and minimum windows = 4) were used to produce the CNV list.

Statistical and population genetics analyses

The DNA sequences were aligned with the ATGC software (GENETIX). The number of nucleotide differences per site (p-distance) was calculated with Molecular Evolutionary Genetics Analysis (MEGA6) [32]. Neighbor-joining (NJ) trees [33] were constructed with 1000 bootstrap resampling with an option of complete deletion of gaps/missing nucleotides. The ratio of the extent of divergence to that of polymorphism between any of the nine 3-kb regions was tested using the HKA test [34] implemented in DnaSP [35]. Genetic components in Cemani, Silkie, other domesticated breeds, RJF, and GJF were examined using STRUCTURE (version 2.2) analysis [36] on the website. Heterozygosity at individual SNP sites was computed based on allele frequencies in the samples (S3 Fig).

Calculations of the allele age or the time span of artificial selection operation were based on the same idea used for adaptively introgressed tracts [37,38]. It was assumed that the probability of observing a tract length ≥ x follows the exponential distribution: (1) where L is the mean tract length. This L is given approximately by , in which t is the number of generations elapsed during artificial selection, r is the recombination rate between two adjacent sites per generation, and r* = r(1 − f) is the recombination rate in the presence of inbreeding, with inbreeding coefficient f. If L is equated to an observed mean tract length , we have an estimate of .

To estimate the selection coefficient, we used the following formula for the expected nucleotide diversity (π) at linked neutral sites under recent selective sweep. The ratio of π to the diversity before the sweep (π0) is given by (2) where s is the selection intensity for mutant homozygotes and Ne is the effective population size [1,39, 4043]. It is clear from the formula that the substantial reduction is expected only if 2r*x/s in the exponent is as small as 0.01 [1] or roughly s = 200r*x. We note that this estimate is almost independent of 2Nes, unless Ne is unlikely large.

Data deposition and availability

The nucleotide sequence data was deposited into the DDBJ databank. The sequences of the nine 3-kb regions by NGS are accessiblee from URL of (accession number: DRA004942) and 1-kb sequences for EDN3 gene from URL of (accession numbers: LC194635-LC194778). Aligned sequences for both 1-kb and 3-kb regions are available upon request.


Single origin of the Fm phenotype

As Cemani and Silkie exhibit the same Fm phenotype (S1 Fig), the chromosomal rearrangement including duplicated EDN3 was suspected to be the common genetic cause of Fm in both breeds [12,14]. Therefore, we first confirmed that the genomic rearrangement is indeed of single origin and common to Cemani and Silkie.

First, we studied genetic variation at EDN3 in nine chicken breeds—Ayam Cemani (n = 5), Silkie (n = 3), Ayam Arab (n = 2), Ayam Kedu (n = 5), White Kedu (n = 2), Ayam Kalosi (n = 2), Ayam Kate (n = 1), Ayam Sentul (n = 1), and Kampung Chicken from Lombok (n = 1), and two jungle fowl populations, RJF (n = 6) and GJF (n = 6) (Table 1). To obtain unambiguously phased genomic sequence data for possibly four different EDN3 genes in certain individuals, we analyzed DNA sequences of about 1 kb in length spanning exons 4 to 5. We selected this rather short fragment to avoid any complication due to intragenic recombination in inferring ancestral relationships among the DNA sequences. Indeed when we used 3-kb sequences for determining the EDN3 phylogeny, we found strong evidence for intragenic recombination in multiple haplotypes, though not in those of Cemani and Silkie for an obvious reason (see NJ tree of region 4 in S4 Fig). The 1-kb sequences obtained from the 36 individuals in our sample can be classified into 12 haplotypes. Of these, six (Hap2’, 5, 6, 8, 10, and 11) are restricted to the jungle fowls, three (Hap1, 3, and 9) are restricted to domesticated chickens, and the remaining haplotypes (Hap2, 4, and 7) occur in both domesticated chickens and jungle fowls (Table 2). RJF and GJF individuals exhibit a relatively large number of distinct haplotypes and maintain higher haplotype diversity than domesticated chickens. Importantly, however, no individual possesses more than two distinct haplotypes, indicating that individuals with EDN3 duplication are highly inbred and homozygous. All eight Cemani and Silkie individuals possess only the Hap2/Hap4 haplotype combination (Table 2). This is in sharp contrast to the presence of the Hap4 homozygous BKL2 and Hap2/Hap10 heterozygous RJF9527 in RJF. The absence of segregation of Hap2 and Hap4 in Cemani and Silkie indicates that they are homozygous with respect to the single Hap2-Hap4 haplotype. In other words, Hap2 and Hap4 are no longer allelic in these breeds. This observation strongly suggests that EDN3 was duplicated by unequal crossing over, and the two resulting loci produced permanent heterozygosity for these alleles.

Curiously, four Kedu (KD3, KD16, KDH3, and KDH8) and some other (STC13 and LOM39) individuals also show the same Hap2/Hap4 haplotype combination as Cemani and Silkie. As the phenotypes of KDH3 and KDH8 are quite similar to that of Cemani (S1 Fig), we speculate that these Kedu individuals are heterozygotes, each possessing one Fm chromosome with duplicated EDN3 and one wild-type chromosome with a single EDN3. This suggests interbreeding between Cemani and Kedu, and is based on the likelihood of the allele on the wild-type chromosome being either Hap2 or Hap4, in light of their high frequencies in Indonesian chicken breeds. On the other hand, KD3 and KD16 show apparent wild-type phenotypes for comb and face color (S1 Fig), suggesting that they possess two wild-type chromosomes with distinct Hap2 and Hap4 alleles. In any case, as other individuals show different haplotypes (Table 2), the Kedu population appears to be much more heterogeneous than Cemani and Silkie with respect to haplotypes and copy number of EDN3 genes.

We tested whether Cemani and the other chicken breeds have the same duplicated regions of DR1 and DR2 as Silkie does. We amplified the regions from DNA of 56 individuals using two sets of specific primers [14] (A2 and B2 in Fig 1) (Table 3). The A2 primer set is designed to amplify a region that may contain either the boundary between inverted DR1 and DR2 (1RD-DR2) or that between inverted DR2 and DR1 (2RD-DR1) in a head-to-head configuration, whereas the B2 primer set is designed to amplify a region that contains either the boundary between DR1 and inverted DR2 (DR1-2RD) or that between DR2 and inverted DR1 (DR2-1RD) in a tail-to-tail configuration. The control primer sets A1 and B1 successfully amplified target sequences in all the samples, as reported previously [14]. Amplification with A2 and B2 was consistently successful for 12 Cemani and 12 Silkie individuals and for seven Kedu samples. These findings indicate that the nucleotide sequences of 1RD-DR2/2RD-DR1 and DR1-2RD/DR2-1RD amplification products from Cemani are identical to those from Silkie (S2 Fig).

Fig 1. Three possible arrangements of duplicated DR1s and DR2s in the Fm region.

Duplication of DR1 and DR2 is absent in the wild-type chromosome (top bar). A2 and B2 primer sets are designed for detecting the duplication boundaries between DR1 and DR2; A1 and B1 are for amplification control. This figure is modified from Dorshorst et al. [14]

Table 3. PCR amplification of the duplication boundaries between DR1 and DR2 in the Fm region.

We confirmed that the Hap2/Hap4 combination in STC13 and LOM39 does not result from duplicated DR1, but stems from the segregation of the Hap2 and Hap4 alleles. However, the results of the A2 and B2 amplifications for the 24 Kedu individuals were somewhat puzzling. Amplification was successful for KDH3, KDH8, and KD16, but not for KD3, despite the fact that all four carry the Hap2/Hap4 combination. These observations for KDH3, KDH8, and KD3 agree with the aforementioned speculation that KDH3 and KDH8 have at least one Fm chromosome, while KD3 has two wild-type chromosomes. However, the result for KD16 is unexpected and suggests that, despite its wild-type phenotype, KD16 might possess at least one Fm chromosome. This speculation is supported by the presence of noticeable black pigment on the comb of KD16 (S1 Fig). This may also be true for KD17, KD19, KD21, and KD22 samples which exhibited successful A2 and B2 amplification (Table 3 and S1 Fig). This observation corroborates the high heterogeneity of Fm in the Kedu population. Although it is conceivable that the heterogeneity is related to Cemani breeding in the same geographic area of Central Java, further investigation of the genotype-phenotype relationship is required to draw any definitive conclusion. The high heterogeneity in the Kedu Fm phenotype also suggests that the Dermal Melanin inhibitor (ID) locus, on chromosome Z [44], is worth further investigation.

Second, we studied CNV in EDN3 using qPCR. We measured the absolute concentrations of amplified EDN3 and UCKL1 amplicons in reaction mixtures of each sample and normalized the copy number of EDN3 based on the single-copy gene of UCKL1. Cemani and Silkie have twice to four times larger copy numbers than non-Fm chickens (Fig 2). Although the exact number of EDN3 copies in Cemani and Silkie genome is difficult to estimate by the qPCR alone, the Fm phenotype surely shows excessive EDN3 copies [12]. In addition, we carried out WGS for a single Cemani individual (Cemani 41). Using CNV-Seq [31], we confirmed that approximately twice as many reads were mapped onto the DR1 and DR2 in Cemani as compared to Taiwanese native chicken L2 with non-Fm phenotype (Fig 3A) and a similar result was obtained in Silkie with respect to L2 (Fig 3B) [23]. However, when the Cemani genome was compared with the Silkie genome, neither DR1 nor DR2 showed any excess of reads (Fig 3C). Together, these results consistently indicate that the DR1 and DR2 arrangement in Cemani is identical to that in Silkie and strongly support a single common origin of the Fm phenotype in Cemani and Silkie.

Fig 2. EDN3 CNV.

Copy numbers of EDN3 were normalized to those of UCKL1 in qPCR. Red bars represent the copy numbers of three Silkie and five Cemani individuals, and blue bars represent those of eight wild-type chickens. The average copy number in Fm-phenotype chickens is 3.39 ± 0.44 and that in wild types is 1.15 ± 0.09 (P = 1.50 × 10−6).

Fig 3. Log2 ratio for CNV in chicken chromosome 20 between nt positions 10,700,000 and 12,000,000.

(a) Comparison of read mapping between Cemani and Taiwanese L2. A blow-up of the DR1 region containing EDN3 is shown below the map. Comparison of read mapping between combinations of (b) Silkie and Taiwanese L2 and (c) Cemani and Silkie.

Haplotype diversity and duplication of EDN3

To study the origin of Hap2 and Hap4 at duplicated EDN3 loci, we examined the sequence differences among the 12 haplotypes or alleles in more detail. The haplotype sequences of the 36 individuals (Table 2) contain 35 variable sites consisting of one 1-bp deletion, two 3-bp deletions, and 28 point mutations (S3 Table). Of these haplotypes, five are singletons in the sample, whereas Hap2 and Hap4 are represented in 17 and 23 individuals, respectively. Hap2’ is one of the singletons and differs from Hap2 by a single point mutation at position 784. As it occurs in RJF, it has likely been derived from Hap2 in the RJF population. Likewise, Hap9 differs from Hap4 by a single 3-bp deletion and descends in indigenous Ayam Arab ARG19. More importantly, Hap10 was found only in RJF and differs from Hap2 and Hap4 by one and two point mutations, respectively. Therefore, Hap10 likely is a common ancestor of Hap2 and Hap4. Thus, the allelic divergence among Hap2, Hap4, and Hap10 must have occurred in the RJF population, which still harbors all these alleles.

To examine the phylogenetic relationship among the 12 haplotypes, we constructed an NJ tree [33] rooted by the orthologous quail sequence and statistically evaluated with 1000 bootstraps based on all nucleotide substitutions in 34 1-kb fragments derived from 29 individuals (Fig 4). Although the tree showed several intermingling patterns of ancestral allelic lineages leading to RJF and domesticated chickens owing to incomplete lineage sorting, it did support that Hap10 is a recent common ancestor of Hap2 and Hap4. Next, we estimated the divergence time between Hap2 and Hap4 based on two calibrated substitution rates. One is based on the published substitution rate in introns [8,45]. When the rate of (1.7–1.8) × 10−9 per site per year is applied to the average per-site nucleotide differences between Hap2 and Hap4 (0.0020 ± 0.0014), the divergence time of 0.6 ± 0.4 million years is obtained. Alternatively, we can directly calibrate the substitution rate in the present EDN3 sequences using the divergence time of RJF/domesticated chickens from GJF. This divergence time can be inferred from such a geological event as the emergence of Java island 3–4 million years ago [46]. Further evidence from fossil records regarding the 4–5 million year-old ancestor of Gallus (Gallus bravardi) consistently suggests that GJF originated around 4 million years ago [47,48]. In this calibration, however, it has to be noted that four distinct haplotypes (Hap5, 7, 8, and 11) exist in GJF, of which Hap11 clustered together with Hap6 in RJFs (Aceh1 and Aceh4), and Hap7 is the same haplotype as that in the domesticated KDP7. Provided that GJF is indigenous to Java and the Lesser Sunda Islands, these Hap7 and Hap11 in GJF raise the possibility of recent introgression from RJF and/or domesticated chickens [8]. Therefore, we excluded Hap7 and Hap11 when calculating the average nucleotide difference per site (0.0107 ± 0.0025) between GJF and other chickens, which resulted in the substitution rate of 1.3 × 10−9 per site per year. This rate is a little slower than the previous one and yields a somewhat earlier estimate of the divergence time (0.8 ± 0.5 million years) between Hap2 and Hap4. In either case, a rough estimate of divergence time (0.6–0.8 myr) implies that these EDN3 alleles in fact originated in the ancestral RJF population. At some point after this allelic divergence, the EDN3 locus was duplicated, and the Fm phenotype appeared. We will discuss a lower limit of this allelic divergence that can be set by the divergence time between the Cemani and Silkie breeds, together with a re-evaluation of the above estimates with large standard errors.

Fig 4. NJ tree of EDN3 haplotypes rooted by the quail sequence (accession number NC_029535).

The tree was constructed with 1000 bootstrap resampling with an option of complete deletion of gaps/missing nucleotides [33]. The nucleotide divergence was measured by using the number of nucleotide differences per site, without multiple-hit correction.

High- homozygosity tracts (HHTs) in Cemani and Silkie

To detect any genomic signature of artificial selection on the Fm phenotype, we investigated the pattern and degree of DNA polymorphism in DR1 and its surrounding genomic regions. Using 9 Cemani, 10 Silkie, 11 other domesticated chickens including a single RJF, and two GJFs, we first examined nine regions of about 3 kb long and separated by ~200-kb intervals. As a whole, they span a 1.4-Mb genomic region that includes the 254-kb duplicated DR1s, 342-kb duplicated DR2s, and 413-kb spacer (Fig 5 and S3 Fig). Table 4 shows summary statistics of the genetic variability in these nine regions (see S4 Fig for NJ trees). First, the number of haplotypes (Hk) in a sample of k chromosomes is generally much smaller than the number of segregating sites (Sk) within the same region. Each region is thus in fairly strong linkage disequilibrium and is consistent with relatively large values of |D′| (not shown) or the squared correlation coefficient (r2): the absolute value of r is greater than 0.56 in all regions of all four populations. Second, the pattern and level of DNA polymorphism in Cemani and Silkie greatly differ from those in “Others” and GJF. We note that region 3 is located upstream of DR1, region 4 is within DR1, and regions 5–7 are within the spacer, while regions 1, 2, 8, and 9 are further away from the Fm region. Compared with regions 1, 2, and 6–9, regions 3–5 in Cemani and Silkie show a remarkable reduction in Hk, Sk, Watterson’s θw and nucleotide diversity π [4951]. For instance, in regions 1, 2, and 6–9, the average number of segregating sites per kb is about 12 in Cemani and Silkie. The expected number in each 3-kb region is thus about 30; however the actual number observed in regions 3 and 5 is 0 in Cemani and at most 2 in Silkie. We regarded this extremely low level of genetic variability as evidence for selective sweep via artificial selection for Fm. To verify this, we carried out the HKA test for Cemani and Silkie [34] using the divergence data in comparison with GJFs. The test indicated a significantly lower level of polymorphism in regions 3–5 than in any other region (S5 Fig). Additionally, we applied the STRUCTURE analysis region by region [35]. Although Cemani and Silkie individuals are generally assigned to different multiple genetic components in regions 1, 2, and 6–9, they are assigned to a single common component in regions 3–5 (DDX27, EDN3, and TH1L) (S6 Fig).

Fig 5. The nucleotide diversity (π) in nine regions surrounding END3 in chicken breeds.

The π values in Cemani (green), Silkie (purple), other domesticated breeds (red), and GJF (blue) are shown under a schematic diagram of their locations together with duplicated DR1 and DR2 regions on chromosome 20. Each π is measured in a 1 kb window with an overlapping sliding size of 100 bp. All regions except 7 and 8 are located within genes whose exon and intron structures are indicated below diversity plots.

Table 4. DNA polymorphism in nine regions in three populations of chickens and GJF (see also Fig 5).

Among regions 3–5, region 4 within DR1 shows an exceptionally high level of genetic variability; however, this is deceptive and results from the inevitable mixture of duplicated sequences. The homologous sequences within duplicated DR1 cannot be amplified separately by the present PCR method; genetic variability in region 4 within DR1 is simply owing to a mixture of the two paralogous sequences. Despite this caveat, the level of genetic variability in region 4 is somewhat higher in Silkie than in Cemani. This is largely due to the inclusion of three sequences of Indonesian Silkie (KPS16, 17, and 30) and is also visible in the STRUCTURE analysis (S6 Fig). The nucleotide sites at positions, 1828, 2015, 2049, 2230, 2279, 2630, 2637, 2664, 2837, and 3005 in region 4 (ranging from 11,157,992 to 11,158,169 in the reference genome) are variable with respect to two distinct haplotypes. One haplotype is identical to that in Silkie and the other is identical to that in indigenous Indonesian chicken breed KAL28 (see NJ tree of region 4 in S4 Fig), suggesting frequent occurrence of interbreeding between Indonesian Silkie and other indigenous domesticated chickens.

We aimed to determine whether the regions with reduced genetic variability or the HHTs are identical for Cemani and Silkie. For practical purposes, we operationally defined an HHT as a consecutive genomic region over 10 kb with π < 10−4 or < 2% of the normal level π0 = 0.005 (Table 4). We determined the WGS of one Cemani individual and compared it with those of Silkie and Taiwanese L2. Cemani and Silkie exhibit a similar extent of reduction in DR1 and its surrounding regions, but the HHT length differed greatly between these two breeds (Fig 6). The Cemani HHT is long and extends toward regions 2 and 7, whereas the Silkie HHT is relatively short and limited from a little beyond regions 3 to 5. The left and right HHT lengths are, respectively, 118 and 387 kb in Cemani and 52 and 101 kb in Silkie, respectively. However, it is highly probable that such a tract can differ from individual to individual even within a breed. Therefore, using the same samples of 19 Cemani and Silkie in Table 4, we further examined the genetic variability surrounding the boundaries in nine windows, each about 1 kb each in length (Fig 6). The windows are localized into three clusters; the left HHT in Cemani still extends beyond the central cluster, while that of Silkie already ends there. Likewise, the right HHT in Cemani terminates just within the right cluster, whereas that in Silkie ends well within in this cluster. We measured the left and right HHT from the 5’ and 3’ ends, respectively, of EDN3 (11,144,657–11,161,475) and estimated the tract lengths. The minimum left and right HHT lengths are 118 and 224 kb, respectively, in Cemani and 52 and 101 kb, respectively, in Silkie (Fig 6).

Fig 6. Nucleotide diversity (π) based on NGS genotype data for one Cemani (red) and one Silkie (green) individual.

Diversity is measured in 10-kb non-overlapping windows. The left-shaded region represents DR1, in which nearly the same patterns and degrees of polymorphism are exhibited by Cemani and Silkie. This supports that DR1 was duplicated prior to their diversification and has since been frozen from recombination in both breeds. The same trend is observed in the right-shaded DR2; however, the ancestral haplotype is obscured by recombination. The three upper panels show the proportion of per-SNP heterozygotes at the population level for Cemani (red) and Silkie (green). Observations are made in 9 windows, each 1 kb long. The 9 windows are grouped into sets of 3, 2, and 4 windows. The three windows at the left are consecutive and contain 16 SNPs in total. The two windows in the middle and four at the right are separated by 16–40 kb and contain 22 and 19 SNPs, respectively.


Genomic configuration of the Fm region

Dorshorst et al. [14] proposed three possible rearrangements (FM1–FM3) of duplicated DR1s and DR2s in the Fm region (Fig 1). Although all these rearrangements possess the same boundaries of 1RD-DR2/2RD-DR1 and DR1-2RD/DR2-1RD, one major difference exists with respect to the relative position of the 413-kb spacer. In models FM1 and FM3, duplicated DR1s sandwich the spacer. If either FM1 or FM3 is valid, the HHT is expected to cover the entire spacer as the two EDN3 loci in DR1s are simultaneous targets of artificial selection. In contrast, in model FM2, the spacer is located outside the duplicated DR1s, and can therefore recombine with wild-type chromosomes without disrupting the Fm phenotype. In this case, the spacer region is expected to be polymorphic because of recombination. Our data (Figs 5 and 6 and Table 4) clearly showed that the patterns and degrees of polymorphism exhibited by Cemani and Silkie are consistent with FM2, but inconsistent with FM1 and FM3.

DR1 duplication and emergence of the Fm phenotype

We estimated an upper limit of EDN3 duplication time of 0.6 ± 0.4 ∼ 0.8 ± 0.5 million years based on the allelic divergence between Hap2 and Hap4. Although the standard error is large because of the usage of the short sequences, Hap2 and Hap4 seem to diverge from each other much earlier than is documented in any known archeological record of domesticated chickens. As mentioned earlier, EDN3 duplication and the Fm phenotype emerged in the ancestral RJF population of chickens; therefore, this phenotypic variation was highly likely to be selected once domestication began in Asia. Xiang et al. [52] dated ancient mtDNA sequences from the earliest archaeological chicken bones in China back to 10,000 years ago.

The analysis of the Cemani and Silkie genome sequences revealed that the 71.4-kb region spanning nt 11,183,600 to 11,255,000 is located within the joint set of the right HHTs in both breeds (Fig 6 and S7 Fig). In this region, recombination has been apparently prohibited by artificial selection on EDN3 and therefore, the two breeds are most closely related to each other in terms of nucleotide substitutions (S7 Fig). Because of the paralogy between DR1s, 50 variable sites in Cemani and 51 in Silkie are observed within a stretch of an approximately 55 kb of 71.4-kb region. It is important to note that a great majority (49) of these variable sites are shared between the two breeds, implying that they accumulated in their common ancestor (S4 Table). As the per-site differences amount to approximately (9.2 ± 1.3) × 10−4, we can estimate the sequence divergence time between the duplicated DR1s as 0.26 ± 0.04 ∼ 0.35 ± 0.05 million years ago. These are more recent, but more reliable than the previous estimates for the upper limit of DR1/EDN3 duplication time. In either case, we conclude that the Fm phenotype caused by duplicated DR1/EDN3 originated in RJF long before the domestication process began in China.

Additionally, we are interested in the divergence time between Cemani and Silkie, to use it as a lower limit for the DR1 duplication time. For this purpose, we used breed-specific substitutions. Provided that recombination is rare or absent within the 71.4-kb region, we treated such substitutions as derived variants and proportional to the divergence time between Cemani and Silkie breeds. In the entire region of 2 × 55 + 16 = 126 kb, there are one Cemani-specific and two Silkie-specific nucleotide substitutions (S4 Table). The mean per-site sequence differences are therefore given by for a pair of Cemani and Silkie genomes. Using the calibrated nucleotide substitution rate of (1.3 ∼ 1.8) × 10−9, we obtain the divergence time of 6600 ∼ 9100 years and regard it as an upper limit of the divergence time between the Cemani and Silkie breeds. Because this also gives a lower limit of the DR1/EDN3 duplication time, the estimate suggests that the Fm phenotype emerged by this time (Fig 7).

Fig 7. Evolutionary history of EDN3 genes in Cemani and Silkie.

The green lines represent the allelic divergence of Hap2 and Hap4 at EDN3 in the ancestral RJF population. The blue lines represent the ancestral lineages in the 71.4-kb region (shadow), from which the divergence time between Cemani and Silkie was estimated. The DR1 duplication is placed somewhere between 10,000 and 300,000 years ago. The horizontal red line corresponds to the beginning of the domestication process in China [52].

Ayam Cemani and Ayam Kedu in Indonesia

The extent of nucleotide diversity in Cemani is almost the same as that in other domesticated chickens and jungle fowl, except near the EDN3 locus (regions 3, 4, and 5), despite the fact that Cemani is a local breed of Kedu, in Indonesia. This is in sharp contrast to Silkie which we could be sampled in Indonesia, Japan, and the USA in the present study. There are two possible explanations for this high genetic variability in Cemani: a relatively large founding population, and frequent genetic exchanges with other domesticated chickens or jungle fowls. The presence of KDH3 and KDH8, which are heterozygous for the Fm and wild-type chromosomes, supports the latter hypothesis of interbreeding of Cemani with other domesticated chickens. In this case, there must have been intense selection to maintain the Fm phenotype. However, the effect of intense selection can be limited in genomic regions closely linked to a target site. Further study of Ayam Kedu with abundant Fm variation will provide useful information on their breeding schemes and the history of the Fm phenotype in Indonesia.

History and strength of artificial selection

We used information on HHTs in two simple ways [53] without any sophistication for inference [5456]. One is to infer the allele age or the history of artificial selection based on the idea underlying the inference of adaptively introgressed tracts [38,39], and the other is to infer the strength of artificial selection, as has been done for maize domestication [1]. Both estimates depend heavily on the recombination rate r per bp per generation. The recombination rate is known to vary considerably among as well as within chromosomes [7]. The rate is 3.7 cM/Mb if averaged over the entire chicken genome, whereas it is approximately 3.0–5.0 cM/Mb for chromosome 20. We assume here r = 3.0 cM/Mb = 3.0 × 10−8 per bp per generation. When considering effect of inbreeding in the domestication process, r is replaced with the effective recombination rate r* = r(1 − f) in which f is the inbreeding coefficient [39].

As mentioned previously, we measured the minimum lengths of the left and right HHTs in sequences of Cemani or Silkie individuals. Based on both the population and NGS data, the left and right HHT lengths are 118 kb and 224 kb, respectively, in Cemani, and 52 kb and 101 kb, respectively, in Silkie. Using formula (1) [37, 38], with an observed mean tract length , we calculated the number of generations elapsed under artificial selection (see Materials and Methods). In the case of Cemani, kb so that t ≈ 200/(1 − f), whereas in the case of Silkie, kb so that t ≈ 440/(1 − f). It thus appears that the history of Cemani is approximately half of that of Silkie. In the absence of inbreeding, the tract erodes quickly, but even intense inbreeding such as full-sib mating, with f = 1/4, can increase the time only by 30%. Furthermore, if we define the generation time of chickens and fowls based on the mean age (m) of hens at a given time [39], we can convert the above t generations into m × t years. If chickens can lay eggs for 7 years (with the age at first reproduction being 1 year and the mean longevity being approximately 15 years), it might be reasonable to assume m = 3–4. Therefore, it appears that Cemani and Silkie have been bred for roughly (600 ∼ 800)/(1 − f) years and (1300 ∼ 1700)/(1 − f) years, respectively. Thus, the history of Indonesian Ayam Cemani appears to be rather short, whereas the relatively long history of Silkie is consistent with the relatively short HHT length in its Fm region.

Second, we used formula (2) for the expected nucleotide diversity π at linked neutral sites under recent selective sweep with selection coefficient s (see Materials and Methods). In the virtual absence of variation, we can have such a rough relationship as s = 200r*x. With r = 3 × 10−8 [7,57] and x ≥ 100 kb, we have s ≥ 0.6(1 − f). This is inevitably a crude estimate but it indeed suggests intense artificial selection in both the Cemani and Silkie breeds.

As a final caveat, it may be asked why the tract boundaries in the NGS data separate HHT very sharply from the neutral level (Fig 5). The two Fm chromosomes in an individual must be flanked by wild-type chromosomes and likely have different recombination breakpoints. However, the two Fm chromosomes have virtually identical DNA sequences in the focal site and nearby linked sites, but are different from wild-type chromosomes, which might also differ from each other. Therefore, we can identify only sharp HHT boundaries in a single diploid Fm individual. However, as these abrupt boundaries can differ among individuals, the tract boundaries might become more gradual for large samples or at the population level. This would explain intermediate values of π observed in regions 1, 2 and 6 as well as in the right HHT in a sample of nine Cemani individuals (Table 4, Fig 6).


We showed that fibromelanosis (Fm) in Indonesian Ayam Cemani and Chinese Silkie resulted from the same genetic change involving EDN3 duplication on chromosome 20. This genomic change of a single origin arose spontaneously in the ancestral population of RJF in Asia, probably well before the first domestication of chickens. Strong artificial selection for the Fm phenotype is evident in the genetic variability near the target site of duplicated EDN3, although the pattern and level of variability differ sensitively between these two breeds, which have undergone different domestication processes.

Supporting information

S1 Fig. Characteristic morphological traits of several Indonesian chickens and Chinese Silkie.

(a) Female Cemani, (b)—(d) female Kedu, (e)—(i) male Kedu, (j) male white Silkie and (k) male black Silkie.


S2 Fig. Sequence information for duplication boundaries generated by the A2 and B2 primer sets.

The A2 and B2 sequences of Cemani (CM6_A2 and CM6_B2) are identical to those of Silkie (SIB17_A2 and SIB17_B2). The boundary was determined by comparison between A1 (CM31_A1) and A2 (upper panel), and between B1 (CM6_B1) and B2 (lower panel).


S3 Fig. Expected heterozygosity at individual SNP sites in the nine regions in chicken breeds.

(a) Domesticated chickens, RJF, and GJF, (b) Ayam Cemani, (c) Silkie chicken.


S4 Fig. NJ trees for the nine regions in domesticated chicken breeds, and RJF.

The phylogenetic relationship differs greatly from region to region. Two GJF haplotype sequences were used as outgroups.


S5 Fig.

Results of the HKA test in each of the nine regions of Cemani (a), Silkie (b), and other domesticated chickens (c). The significant reduction in DNA polymorphism is found in Cemani and Silkie only for DDX27 in region 3, EDN3 in region 4, and TH1L in region 5 are compared.


S6 Fig. STRUCTURE analysis of each of the nine regions of GJF, RJF, Cemani, Silkie and other domesticated chickens.

For regions 3–5, Cemani and Silkie exhibit nearly identical genetic components, whereas in other regions, there are no noticeable structural differences among chicken breeds and RJF.


S7 Fig. Nucleotide diversity between Cemani and Silkie based on NGS data.

Bars with R1–R9 indicate the positions of the nine regions. Green square parentheses indicate the position of EDN3, and a red bar indicates the 71.4-kb region with low divergence between the two breeds.


S1 Table. Sequences of primers used in this study.


S2 Table. Reaction mixtures and PCR conditions used in this study.


S3 Table. Segregating sites in EDN3 haplotypes.


S4 Table. Variable sites in the 71.4-kb region in Cemani, Silkie and Taiwanese L2 as compared to the reference genome.

The region ranges from nt 11,183,600 to 11,255,000 and includes part of DR1. Insertions and deletions are excluded. The colored columns indicate the Silkie (green)- or Cemani (red)-specific mutations.



This paper is dedicated to the late professor Dr. Sri Sulandari.

We thank Dr. Naoyuki Takahata for helpful discussion and critical reading of early versions of this manuscript and Drs. Monty Slatkin and Wen-Hsiung Li for comments and suggestions. We are deeply grateful to the ABRC in Nagoya University, and to the JSAC for providing us with chickens as well as blood and DNA samples. We are grateful for the support of SOKENDAI and LIPI throughout this project. ABD expresses her sincere thanks to the MEXT for a scholarship. SS passed away before the completion of this study. YS is responsible for the integrity and validity of the data collected by the deceased author.

Author Contributions

  1. Conceptualization: ABD YS.
  2. Data curation: ABD YS.
  3. Formal analysis: ABD YS.
  4. Investigation: ABD.
  5. Methodology: ABD YS.
  6. Project administration: YS.
  7. Resources: YT SS TA MSAZ.
  8. Supervision: YS.
  9. Validation: YS.
  10. Visualization: ABD.
  11. Writing – original draft: ABD YS.
  12. Writing – review & editing: YS.


  1. 1. Wang RL, Stec A, Hey J, Lukens L, Doebley J. The limits of selection during maize domestication. Nature. 1999; 398: 236–239. pmid:10094045
  2. 2. Cong B, Barrero LS, Tanksley SD. Regulatory change in YABBY-like transcription factor led to evolution of extreme fruit size during tomato domestication. Nat Genet. 2008; 40: 800–804. pmid:18469814
  3. 3. Bovine HapMap Consortium. Genome-wide survey of SNP variation uncovers the genetic structure of cattle breeds. Science. 2009; 324: 528–532. pmid:19390050
  4. 4. vonHoldt BM, Pollinger JP, Lohmueller KE, Han E, Parker HG, Quignon P, et al. Genome-wide SNP haplotype analyses reveal a rich history underlying dog domestication. Nature. 2010; 464: 898–902. pmid:20237475
  5. 5. Amaral AJ, Ferretti L, Megens HJ, Crooijmans RPMA, Nie H, Ramos-Onsins SE, et al. Genome-wide footprints of pig domestication and selection revealed through massive parallel sequencing of pooled DNA. PLoS ONE. 2011; 4: e14782.
  6. 6. International Chicken Polymorphism Map Consortium. A genetic variation map for chicken with 2.8 million single-nucleotide polymorphisms. Nature. 2004; 432: 717–722. pmid:15592405
  7. 7. Groenen MAM, Wahlberg P, Foglio M, Cheng HH, Megens HJ, Crooijmans RPMA, et al. A high-density SNP-based linkage map of the chicken genome reveals sequence features correlated with recombination rate. Genome Res. 2009; 19: 510–519. pmid:19088305
  8. 8. Sawai H, Kim HL, Kuno K, Suzuki S, Gotoh H, Takada M, et al. The origin and genetic variation of domestic chickens with special reference to junglefowls Gallus g. gallus and G. varius. PLoS ONE. 2010; 5: e10639. pmid:20502703
  9. 9. Rubin CJ, Zody MC, Erikson J, Meadows JRS, Sherwood E, Webster MT, et al. Whole-genome resequencing reveals loci under selection during chicken domestication. Nature. 2010; 464: 587–591. pmid:20220755
  10. 10. Johnsson M, Gering E, Willis P, Lopez S, Van Dorp L, Hellenthal G, Henriksen R, et al. Feralisation targets different genomic loci to domestication in the chicken. Nature Comm. 2016; 7: 12950.
  11. 11. Hutt FB. Genetics of the Fowl. New York: McGraw-Hill; 1949.
  12. 12. Shinomiya A, Kayashima Y, Kinoshita K, Mizutani M, Namikawa T, Matsuda Y, et al. Gene duplication of the endothelin 3 is closely correlated with the hyperpigmentation of the internal organs (Fibromelanosis) in Silkie chickens. Genetics. 2011; 111.136705. E-pub 2011.Nov. 30.
  13. 13. Dorshorst B, Okimoto R, Ashwell C. Genomic regions associated with dermal hyperpigmentation, polydactyly and other morphological traits in the Silkie chicken. J Hered. 2010; 101: 339–350. pmid:20064842
  14. 14. Dorshorst B, Molin AM, Rubin CJ, Johansson AM, Strömstedt L, Pham M-H, et al. A complex genomic rearrangement involving the Endothelin 3 locus causes dermal hyperpigmentation in the chicken. PLoS Genet. 2011; 7: e1002412. pmid:22216010
  15. 15. Saldana-Caboverde A, Kos L. Roles of endothelin signaling in melanocyte development and melanoma. Pigment Cell Melanoma Res. 2010; 23: 160–170. pmid:20128875
  16. 16. Lahav R, Ziller C, Dupin E, Le Douarin NM. Endothelin 3 promotes neural crest cell proliferation and mediates a vast increase in melanocyte number in culture. Proc Natl Acad Sci U S A. 1996; 93: 3892–3897. pmid:8632985
  17. 17. Lahav R, Lecoin L, Ziller C, Nataf V, Camahan JF, Martin FH, et al. Effect of the Steel gene product on melanogenesis in avian neural cell cultures. Differentiation. 1998; 58: 133–139.
  18. 18. Dupin E, Douarin NML. Development of melanocyte precursors from the vertebrate neural crest. Oncogene. 2003 22: 3016–3023. pmid:12789276
  19. 19. Akiyama T, Shinomiya A. Overview on the melanocyte precursor migration from the neural crest. In: Smith J, Haworth M, editors. Skin Pigmentation. New York: Nova Science Publishers, Inc.; 2013, pp. 175–196.
  20. 20. Smyth JR Jr. Melanin pigmentation: Its Biological roles, Inheritance and Expression in the Chicken. University of Mass Amherst; 1994.
  21. 21. Kaelin CB, Xu X, Hong LZ, David VA, McGowan KA, Schmidt-Küntzel A, et al. Specifying and sustaining pigmentation patterns in domestic and wild cats. Science. 2012; 337: 1536–1541. pmid:22997338
  22. 22. Qanbari S, Pausch H, Jansen S, Somel M, Strom TM, Fries R, et al. Classic selective sweeps revealed by massive sequencing in cattle. PLoS Genet. 2014; 10: e1004148. pmid:24586189
  23. 23. Fan W-L, Ng CS, Chen C-F, Lu M-YJ, Chen Y-H, Liu C-J, et al. Genome-wide patterns of genetic variation in two domestic chickens. Genome Biol Evol. 2013; 5: 1376–1392. pmid:23814129
  24. 24. Birnboim HC. A rapid alkaline extraction method for the isolation of plasmid DNA. Methods Enzymol. 1983; 100: 243–255. pmid:6353143
  25. 25. Bolger AM, Lohse M, Usadel B. Trimmomatic: a flexible trimmer for Illumina sequence data. Bioinformatics. 2014;30: 2114–2120. pmid:24695404
  26. 26. Li H. Aligning sequence reads, clone sequences and assembly contigs with BWA-MEM; 2013. Avilable: arXiv:1303.3997v2, [q-bio.GN]
  27. 27. Li H, Handsaker B, Wysoker A, Fennell T, Ruan J, Homer N, et al. The Sequence Alignment/Map format and SAMtools. Bioinformatics. 2009; 25: 2078–2079. pmid:19505943
  28. 28. McKenna A, Hanna M, Banks E, Sivachenko A, Cibulskis K, Kernytsky A, et al. The Genome Analysis Toolkit: a MapReduce framework for analyzing next-generation DNA sequencing data. Genome Res. 2010; 20: 1297–1303 pmid:20644199
  29. 29. DePristo M, Banks E, Poplin R, Garimella K, Maguire J, Hartl C, et al. A framework for variation discovery and genotyping using next-generation DNA sequencing data. Nat Genet. 2011; 43: 491–498 pmid:21478889
  30. 30. Van der Auwera GA, Carneiro MO, Hartl C, Poplin R, del Angel G, Levy-Moonshine A, et al. From FastQ Data to High-Confidence Variant Calls: The Genome Analysis Toolkit Best Practices Pipeline. Curr Protoc Bioinformatics 2013; 43: 11.10.1–11.10.33.
  31. 31. Xie C, Tammi MT. CNV-seq, a new method to detect copy number variation using high-throughput sequencing. BMC Bioinformatics. 2009; 10: 80. pmid:19267900
  32. 32. Tamura K, Stecher G, Peterson D, Filipski A, Kumar S. MEGA6: Molecular Evolutionary Genetics Analysis version 6.0. Mol Biol Evol. 2013; 30: 2725–2729. pmid:24132122
  33. 33. Saitou N, Nei M. The neighbor-joining method: a new method for reconstructing phylogenetic trees. Mol Biol Evol. 1987; 4: 406–425. pmid:3447015
  34. 34. Hudson RR, Kreitman M, Aguadé M. A test of neutral molecular evolution based on nucleotide data. Genetics. 1987; 116: 153–159. pmid:3110004
  35. 35. Librado P, Rozas J. DnaSP v5: A software for comprehensive analysis of DNA polymorphism data. Bioinformatics. 2009; 25: 1451–1452. pmid:19346325
  36. 36. Prichard JK, Stephens M, Donnelly P. Inference of population structure using multilocus genotype data. Genetics. 2000; 155: 945–959. pmid:10835412
  37. 37. Liang M, Nielsen R. The length of admixture tracts. Genetics. 2014; 197: 953–967. pmid:24770332
  38. 38. Racimo F, Sankararaman S, Nielsen R, Huerta-Sanchez E. Evidence for archaic adaptive introgression in humans. Nat Rev Genet. 2015; 16: 359–371. pmid:25963373
  39. 39. Charlesworth B, Charlesworth D. Elements of Evolutionary Genetics. Colorado: Roberts and Company Publishers; 2012.
  40. 40. Maynard Smith J, Haigh J. The hitch-hiking effect of a favorable gene. Genet Res. 1974; 23: 23–35. pmid:4407212
  41. 41. Kaplan NL, Hudson RR, Langley CH. The “hitch-hiking” effect revisited. Genetics. 1989; 123: 887–899. pmid:2612899
  42. 42. Barton NH. The effect of hitch-hiking on neutral genealogies. Genet Res. 1998; 72: 123–133.
  43. 43. Nielsen R, Williamson S, Kim Y, Hubisz MJ, Clark AG, Bustamante C. Genomic scans for selective sweeps using SNP data. Genome Res. 2005; 15: 1566–1575. pmid:16251466
  44. 44. Bateson W, Punnett RC. The inheritance of the peculiar pigmentation of the silky fowl. J Genet. 1911; 1: 185–203.
  45. 45. Axelsson E, Smith NG, Sundstrom H, Berlin S, Ellengren H. Male-biased mutation rate and divergence in autosomal, Z-linked and W-linked introns of chicken and turkey. Mol Biol Evol. 2004; 21: 1538–1547. pmid:15140948
  46. 46. Bellwood PS. Man’s conquest of the Pacific: The Prehistory of Southeast Asia and Oceania. Oxford University Press; 1979.
  47. 47. Lambrecht K. Handbuch der Paläornithologie. Berlin: Verlag Grebrüder, Borntaeger; 1933.
  48. 48. Brodkorb P. Catalogue of fossil birds. Bull Florida State Mus. 1964; 8: 195–335.
  49. 49. Ewens WJ. The sampling theory of selectively neutral alleles. Theor Popul Biol. 1972; 3: 87–112. pmid:4667078
  50. 50. Watterson GA. On the number of segregating sites in genetical models without recombination. Theor Popul Biol. 1975; 7: 256–276. pmid:1145509
  51. 51. Nei M. Molecular Evolutionary Genetics. New York: Columbia University Press; 1987.
  52. 52. Xiang H, Gao J, Yu B, Zhou H, Cai D, Zhang Y, et al. Early Holocene chicken domestication in northern China. Proc Natl Acad Sci U S A. 2014; 111: 17564–17569. pmid:25422439
  53. 53. Stephens JC, Reich DE, Goldstein DB, Shin HD, Smith MW, Carrington M, et al. Dating the origin of the CCR5-Delta32 AIDS-resistance allele by the coalescence of haplotypes. Am J Hum Genet. 1998; 62: 1507–1515. pmid:9585595
  54. 54. Slatkin M. A Bayesian method for jointly estimating allele age and selection intensity. Genet Res. 2008; 90: 119–128.
  55. 55. Chen H, Slatkin M. Inferring selection intensity and allele age from multilocus haplotype structure. Genes Genomes Genet. 2013; 3: 1429–1442.
  56. 56. Chen H, Hey J, Slatkin M. A hidden Markov model for investigating recent positive selection through haplotype structure. Theor Popul Biol. 2015; 99: 18–30. pmid:25446961
  57. 57. Mugal CF, Nabholz B, Ellegren H. Genome-wide analysis in chicken reveals that local levels of genetic diversity are mainly governed by the rate of recombination. BMC Genomics. 2013; 14: 86. pmid:23394684