Research note: Integration of GWAS and RNA-seq identifies mutations linked to within feather patterning in domestic chickens

Abstract

Within feather pattern variation is widespread among chicken breeds globally and is primarily characterized by the distributions of eumelanin on a single feather vane. To analyze the genetic basis underlying different within feather patterns, we conducted the GWAS by pairwise comparing the three common patterns observed in Chinese domestic chickens: stippling, pencilling, and arrow. Based on the coding-region variants, a set of candidate genes were sifted out, notably MC1R, CDH1, MYO5C, FOXI2, and RANBP17, which are functionally associated with melanogenesis and pigment transport. From the perspective of regulatory variants, we integrated GWAS signals with RNA-seq data from growing-phase feather follicles to identify candidates driving transcript-level divergence. With putative regulatory variants, RANBP17, MAPK6, and LOC107051763 were identified as overlapping candidates with differentially expressed genes. Overall, by integrating coding-region and regulatory-region variant analysis, this study identified a set of prioritized candidate genes linked to within feather patterning. This study provides the first systematic genetic dissection of within-feather patterns in domestic chickens by integrating GWAS and RNA-seq, offering new insights into the genetic and developmental mechanisms underlying avian within feather pattern formation.

Introduction

The phenotypic diversity of feather plumage in domestic chickens is characterized by both across the body and within feather patterns. Across the body pattern affects the whole body or major body regions. Within feather patterning refers to the color pattern on an individual feather, such as spangling, autosomal barring, and stippling (Andersson et al., 2020). Moreover, specific patterns are frequently associated with quality chicken meat or egg in various regions, thereby holding substantial economic value.

Extensive crossbreeding experiments conducted throughout the 20th century demonstrated that various within feather patterns are determined by combinations of different alleles at multiple loci. As summarized by Andersson et al., the major loci determining chicken within feather patterns (except mottling) include E (Extension), Pg (Patterning), Ml (Melanotic), Db (Dark brown), and Co (Columbian).

A missense mutation in the Melanocortin 1 Receptor (MC1R) was found to be associated to the E*E allele on E locus, which is responsible for Extended black plumage (Kerje et al., 2003). Besides, Pg has been linked to another missense mutation in MC1R, indicating that this gene can simultaneously influence across body plumage and within feather patterning (Schwochow et al., 2021). The causal variant at the Ml locus was identified as an indel variant upstream of GJA5, which alters its expressions (Li et al., 2021). The Db locus has been mapped to an 8.3 kb deletion upstream of SOX10 (Gunnarsson et al., 2011).

Based on our observations, we documented stippling, pencilling, and arrow as the three representative feather patterns. These specific patterns are among the most prevalent plumage types in Chinese indigenous breeds and exhibit distinct spatial arrangements of melanin within individual feathers. Stippling is characterized by eumelanin being distributed as discrete, evenly spaced dots within the feather, without forming concentric elliptical bands along the feather contour (Fig. 1A). Pencilling consists of periodic elliptical bands of eumelanin and pheomelanin, with two or more concentric eumelanin bands located in the inner portion of the feather rather than at the outermost edge (Fig. 1B). Arrow also features elliptical bands of pheomelanin and eumelanin, but with only a single eumelanin band situated in the inner portion of the feather (Fig. 1C).

Fig. 1.

(A) The phenotype of stippling. (B) The phenotype of pencilling. (C) The phenotype of arrow. (D) The Manhattan plot and Q-Q plot for the comparison of 45 stippling and 21 pencilling samples. The red solid line and blue dashed line indicate the genome-wide suggestive significance threshold (P = 1/number of SNPs) and genome-wide significance threshold (P = 0.05/number of SNPs), respectively. (E) PCA plot of RNA-seq data from six feather follicles per group. Circles represent stippling samples, and triangles represent pencilling samples. (F) Venn diagram for the comparison of stippling and pencilling showing the integration of GWAS-identified genes and DEGs from RNA-seq to identify candidate gene putatively affected by regulatory-region variants. (G) The Manhattan plot and Q-Q plot for the comparison of 45 stippling and 35 arrow samples. (H) PCA plot of RNA-seq data from five feather follicles per group. Circles represent stippling samples, and squares represent arrow samples. (I) Venn diagram for the comparison of stippling and arrow. (J) The Manhattan plot and Q-Q plot for the comparison of 35 arrow and 22 pencilling samples. (K) PCA plot of RNA-seq data from five feather follicles per group. Squares represent arrow samples and triangles represent pencilling samples. (L) Venn diagram for the comparison of arrow and pencilling.

To dissect the genetic basis underlying within feather pattern variation, we employed an integrative approach combining genome-wide association studies (GWAS) with RNA sequencing (RNA-seq), and conducted pairwise comparisons among the three pattern types. This approach identified 11 candidate genes associated with within feather pattern formation. These findings advance our understanding of the genetic and developmental mechanisms regulating pattern differentiation, facilitating the improvements of within-breed phenotypic uniformity and inter-breed distinctiveness for domestic commercial chickens.

Materials and methods

Ethical compliance and phenotypic evaluation

The animal procedures used in this study were approved by the Institutional Animal Care and Use Committee of Huazhong Agricultural University (HZAUCH-2025-0008). The study population consisted of maternal-line hens from the XinHua-2 commercial layer strain, which sharing the genetic makeup and plumage appearance (expressing stippling, pencilling, and arrow) with Chinese domestic chickens. Hens were classified according to the within feather pattern presented on their dorsal feathers. Individuals exhibiting irregular or poorly defined patterns, ambiguous intermediate phenotypes, or inconsistencies between dorsal and pectoral patterning were excluded. Eventually, 45 chickens with stippling patterns, 21 with pencilling, and 35 with arrow were kept for all pairwise analyses.

Genomic sequencing and quality control

Blood samples were collected from the brachial veins. DNA extraction and sequencing procedures followed the protocols described in Guo et al. (2025). After quality control, a total of 14,991,403 high-confidence single nucleotide polymorphisms (SNPs) from 102 individuals were retained for downstream analyses.

Genome-wide association studies

GWAS was conducted using the same analytical framework as in our previous work, employing GEMMA v0.98.1 and a linear mixed model (LMM) for association testing (Guo et al., 2025). The genomic kinship matrix was incorporated as a covariate into the GWAS model to minimize the potential confounding effects of population structure. Genome-wide significance thresholds were assessed using the Bonferroni correction. Specifically, the genome-wide significant threshold was set at 0.05 divided by the total number of SNPs (0.05/N), while the suggestive threshold was defined as 1 divided by the total number of SNPs (1/N). For each significant SNP, a ± 50 kb flanking window was designated as the candidate interval. Variants within the candidate intervals were annotated using SnpEff v5.2a, and genes carrying high-impact variants (e.g. stop-gained, start-lost, splice-donor-variant, stop-lost) or moderate-impact variants (missense-variant) were initially selected. Among these genes containing missense mutations, the ones with known pigmentation related functions were retained. These missense variants were subsequently evaluated using SIFT (http://sift-dna.org) and PolyPhen-2 (http://genetics.bwh.harvard.edu/pph2/) —two online tools that assess sequence conservation and the potential impact of amino acid substitutions—to screen out the finely filtered missense variants. After that, χ² tests were performed to analyze genotype-phenotype associations for these mutations in different comparison groups.

Transcriptome data analysis

For each of the three pattern types, six individuals were randomly selected and their growing-phase feather follicles were obtained for RNA-seq analysis. Transcriptome sequencing was performed by Beijing Biomarker Technologies Co., Ltd. using the Illumina platform, generating paired-end FASTQ reads. Raw reads were subjected to stringent quality control standards. Initial quality assessment was performed using FastQC v0.12.1, followed by adapter removal and low-quality base filtering with Trimmomatic v0.35. Only high-quality clean reads with a Q30 percentage exceeding 90% were retained for subsequent analyses. The resulting clean reads were aligned to the reference genome GRCg6a using HISAT2 v2.1.0, and gene counts were obtained with featureCounts v2.0.3. FPKM values were calculated based on the gene count matrix, and principal component analysis (PCA) was performed using R v4.4.3. Genes with P < 0.05 and | log₂(fold change) | ≥ 1 were considered differentially expressed genes (DEGs).

We utilized the online tool EVenn (http://www.ehbio.com/test/venn/#/) to identify potential non-coding regulatory variants by performing an integrated analysis of genes within GWAS candidate regions and DEGs. To explore the potential regulatory mechanisms, we integrated publicly available Gallus gallus ATAC-seq and chromatin state datasets sourced from Professor Huaijun Zhou's laboratory track hub (https://genome.ucsc.edu/s/zhouhuaijun/chicken_regulatory_element). Functional enrichment analysis of DEGs was performed using KOBAS (http://bioinfo.org/kobas/genelist/).

Results and discussion

GWAS of coding-region variants

Pairwise GWAS comparisons were conducted among the three pattern types: stippling, pencilling, and arrow. For the first comparison, the primary distinction between stippling and pencilling is that stippling features discontinuous, dot-like distributions of eumelanin, whereas pencilling exhibits two or more continuous concentric bands of eumelanin. Following the GWAS analysis, significant SNPs were identified on chromosomes 1, 2, 3, 6, 9, 10, 12, 18, and 33 (Fig. 1D). The corresponding candidate intervals encompassed a total of 78 genes, including RSPO3, MYO5C, and MAPK6. Among them, four carried high-impact variants: LOC101751874 (stop-gained and splice-donor-variant), CCDC138 (start-lost), C9ORF152 (stop-lost), FOXI2 (splice-donor-variant and missense-variant). Additionally, three genes with finely filtered missense variants were identified: FOXI2, MYO5C, and TMOD2. After genotype–phenotype association analyses, a total of 5 coding-region candidate genes were eventually identified: LOC101751874, CCDC138, C9ORF152, FOXI2, and MYO5C.

The second comparison was conducted between stippling and arrow. Arrow features a single continuous band of eumelanin in a feather vane. GWAS analysis identified significant SNPs on chromosomes 1, 2, 3, 4, 6, 13, and 23 (Fig. 1G), give 50 genes in candidate intervals, including GATA6, GJA9, and EXTL. Applying the same criteria as in the first comparison, RANBP17 (splice-donor-variant) and DACT2 (missense-variant) were identified as the coding-region candidate genes for the second comparison.

The third comparison involved the arrow and pencilling, which differ primarily in the number of concentric eumelanin bands. GWAS identified significant SNPs on chromosomes 1, 4, and 11. The candidate interval on chromosome 11 (spanning 18,809,511 to 18,885,097 bp) with concentrated signals indicate strong genetic relevance of this region to the phenotypic divergence (Fig. 1J). Within these candidate intervals, we annotated 35 genes, including the melanogenesis-related MC1R and CDH1 (Table 1). Applying filtering criteria, two coding-region candidate genes (MC1R, CDH1) were identified.

Table 1.

Candidate missense mutations and their functional prediction scores identified from three GWAS comparisons.

Comparison	Chr	Candidate gene	SNP	P-value (Chi-square test)	SIFT score1	PolyPhen-2 score2
Stippling VS Pencilling	6	FOXI2	rs3387873967	0.003	0.05	0.824
	10	MYO5C	rs14946372	<0.001	0.00	0.991
Stippling VS Arrow	3	DACT2	chr3:41979977	0.041	0.07	0.988
Arrow VS Pencilling	11	CDH1	rs3137645	<0.001	0.00	1.000
	11	MC1R	rs313704501	<0.001	0.01	0.043

Note:.

¹The SIFT score predicts the potential functional impact of an amino acid substitution based on sequence homology. Variants with a score≤0.05 are considered to have a relatively strong predicted effect on protein function.

²PolyPhen-2 score evaluates the potential functional impact of an amino acid substitution by integrating evolutionary conservation, protein structural features, and functional annotations. Scores closer to 1.00 indicate a stronger predicted effect.

Integration of GWAS and RNA-seq reveals regulatory variants shaping within feather patterning

We performed pairwise transcriptomic comparisons among three plumage patterns. In the stippling–pencilling comparison (Fig. 1E), 466 DEGs were identified, while MAPK6 is the sole DEG located within the GWAS candidate intervals (Fig. 1F). In the stippling–arrow comparison (Fig. 1H), 828 DEGs were detected, and joint genomic–transcriptomic analysis revealed RANBP17 as the shared candidate gene (Fig. 1I). For the arrow–pencilling comparison (Fig. 1K), 494 DEGs were identified, and integrative integration pinpointed LOC107051763 as the candidate gene (Fig. 1L). For MAPK6, publicly available ATAC-seq data revealed a prominent signal approximately 300 bp upstream of the 5′ end. Regarding RANBP17, two key open chromatin regions were identified: one located roughly 800 bp upstream of the 5′ end and another approximately 20 kb downstream of the 3′ end. For LOC107051763, a significant ATAC signal was detected 30 kb upstream of the gene.

Enrichment analysis indicates that while Wnt signaling and melanogenesis drive pigment synthesis, pathways related to cell adhesion and actin cytoskeleton regulation likely determine feather patterning by modulating melanocyte migration and spatial positioning.

Interpreting the findings

In many domestic chicken breeds, variable within feather patterns frequently co-exist within a single line, leading to inconsistent appearance and reducing the distinctiveness between lines. In this study, we focused on three prevalent within feather pattern types and performed an integrated genomic and transcriptomic analysis. This integrative approach successfully pinpointed a set of candidate genes associated with plumage patterns.

In the stippling-pencilling comparison, we identified MYO5C, a myosin family member involved in organelle transport, as a key candidate. Given that the MYO5A-MLPH complex is essential for melanosome transport (Vaez et al., 2008), we hypothesize that MYO5C similarly regulates intracellular melanosome trafficking. Additionally, FOXI2 is a member of the Forkhead family, and extensive previous research has shown that this family is largely associated with the development of skin and epithelium as well as feather morphology. Among the genes identified based on non-coding region variants, BLASTP analysis revealed that LOC101751874 is highly homologous to the Vinculin and Alpha-catenin families, with Vinculin/Catenin serving as a key bridge connecting cadherin and the cytoskeleton.

In the stippling-arrow comparison, RANBP17 emerged as a DEG. This gene was previously linked to cattle coat color (Chhotaray et al., 2021). BLASTP analysis confirmed high sequence identity (68.73%) between the chicken RANBP17 protein and its bovine (Bos mutus) homolog, underscoring its conserved functional potential. In addition to that, another missense mutation gene, DACT2, regulates the Wnt signaling pathway, which is crucial for melanocyte and feather development. Therefore, it likely plays a key role in the observed phenotypic divergence.

In the third comparison group, between the arrow and pencilling phenotypes, two candidate genes closely related to melanin deposition— MC1R and CDH1—were identified. MC1R is a major-effect gene for plumage color (Kerje et al., 2003). In our genomic data, we detected two missense mutations in MC1R, Thr143Ala and Cys213Arg, and the genotypic distributions of both mutations differed extremely significantly between the two phenotypes (P < 0.001). Notably, while Thr143Ala was reported to be associated with systemic eumelanin suppression in the Wheaten allele (e^y) (Hoque et al., 2013), our findings provide a novel link between this variant and fine-scale within feather patterning. We propose that the intermittent eumelanin stripes observed in pencilling patterns are a consequence of the localized inhibitory effect exerted by Thr143Ala. This discovery shifts the functional paradigm of MC1R from a macro-level "color switch" to a micro-level "pattern regulator". In addition, this comparison also identified CDH1, which mediates the anchoring and attachment of melanocytes to surrounding keratinocytes (Vasioukhin et al., 2000).While genes such as CDH1 exhibit pleiotropic roles across various biological systems, no significant differences in gross morphology, production performance, or general health were observed, based on our study and long-term practices of breeding. These observations suggest that the effects of these variants are likely highly tissue-specific. Consequently, we prioritized our subsequent functional analysis on genes exhibiting significant differential expression specifically within the feather follicles. For instance, we identified LOC107051763 through GWAS and RNA-seq. Although currently unannotated, this lncRNA likely functions as a regulatory element that modulates the expression of genes involved in melanin patterning.

In this study, the minimal overlap between GWAS candidates and DEGs suggests a hierarchy where missense mutations alter protein function in upstream, then trigger downstream transcriptional changes. Although the precise interaction pathways remain to be uncovered, this integrated analysis provides a hypothetical framework for elucidating the genetic mechanisms underlying within-feather patterning.

While this study identified several candidate genes regulating feather patterns, we acknowledge certain limitations. First, the sample sizes for both GWAS and RNA-seq analysis were relatively small. Secondly, we prioritized broad candidate gene inclusion through a ± 50 kb window screening strategy rather than fine-mapping or LD analysis. Furthermore, our current findings are primarily based on statistical associations and lack direct biological validation of causal mutations. Consequently, in-depth mechanistic analysis should be carried out in the future work, including single-cell RNA-seq and to confirm the phenotypic effects of candidate mutations through lentiviral injection or gene-editing technologies.

In summary, our findings provide the definitions and genetic dissections of within feather patterns. A total of 11 candidate genes were identified from both coding-region and regulatory-region variants, including genes such as MC1R, CDH1, and RANBP17. Our findings also provide an important supplement to existing theories of avian feather genetics. Current theories of avian feather patterns formation primarily rely on the reaction-diffusion model. In our study, the adhesion molecules we identified (such as CDH1 and LOC101751874), suggesting that precise cell adhesion is key to the pattern formation. Furthermore, unlike the classic MC1R/Agouti signaling pathway, our results suggest that fine-scale within feather textures may depend more on the intracellular transportation efficiency of melanosomes. This indicates that during avian evolution, complex within feather patterns were likely generated through the fine-tuned regulation of the adhesion and transportation of these cells.

These candidate variants provide promising targets for developing molecular markers for the selection of diverse plumage patterns. These results also suggest that the genetic mechanisms underlying within feather patterns are complex, involving multi-gene interactions and the combined effects of coding and non-coding variants.

Funding

This work was supported by the National Natural Science Foundation of China (NSFC) project (32202663).

Acknowledgments

The authors thank Hubei Xinhua Ecological Livestock and Poultry Development Co., Ltd. for providing the experimental animals.

Data deposition

The raw RNA-seq and whole-genome sequencing data generated in this study have been deposited in the National Genomics Data Center (NGDC) database (https://ngdc.cncb.ac.cn/bioproject/) under accession number PRJCA057422.

CRediT authorship contribution statement

Huanjie Guo: Writing – original draft, Visualization, Validation, Software, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Yuqi Chen: Writing – review & editing, Methodology, Investigation. Yifan Guo: Writing – review & editing, Methodology, Investigation. Binghui Wang: Writing – review & editing, Conceptualization. Jingyi Li: Writing – review & editing, Supervision, Resources, Project administration, Methodology, Funding acquisition, Conceptualization.

Disclosures

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.