Transposable element insertion polymorphisms identify LOC101798742 as a candidate gene for breast muscle yield in ducks

Abstract

Breast muscle yield is an economically important trait in ducks, but the contribution of population-level transposable element insertion polymorphisms (TIPs) to its genetic regulation remains poorly understood. Whole-genome resequencing data from 200 ducks representing Cherry Valley ducks, Liancheng white ducks, and five Jinling White Duck lines were used to identify population-level TIPs. After quality filtering, TIP-based genome-wide association study (TIP-GWAS) and TIP-based expression genome-wide association study (TIP-eGWAS) were performed to identify loci associated with breast muscle traits and gene expression. Candidate genes were prioritized by integrating TIP-associated expression genes, weighted gene coexpression network analysis, differential expression analysis, and gene–trait correlation analysis, followed by functional evaluation in primary duck myoblasts. A total of 3,628 TIPs were identified, and 543 high-confidence TIPs were retained for population structure and association analyses. TIP-GWAS detected 14 significant trait–locus associations, mainly involving breast muscle yield traits, which were merged into 6 nonredundant TIP loci. Targeted TIP-eGWAS followed by conservative filtering retained 1,410 priority eGenes, of which 147 were associated with at least one target trait. These genes were enriched in pathways related to the tricarboxylic acid cycle, oxidative phosphorylation, carbon metabolism, fatty acid metabolism, and fibroblast growth factor receptor signaling. Integrative analysis prioritized LOC101798742 as a candidate gene corresponding to the chr6:36133712 TIP locus. In silico annotation indicated that LOC101798742 encodes the duck ortholog of alpha-aspartyl dipeptidase (DNPEP). In primary duck myoblasts, LOC101798742 knockdown reduced proliferation-related gene expression, altered cell-cycle distribution, and enhanced myoblast fusion, whereas overexpression promoted cell-cycle progression and increased apoptosis-related changes. These findings suggest that population-level TIP variation may contribute to duck breast muscle yield through gene regulatory networks and myoblast proliferation–differentiation dynamics. This study expands the range of regulatory variants considered in duck breeding and provides candidate loci and genes for future functional validation and genetic improvement.

Introduction

Duck meat is an important source of high-quality animal protein and has substantial economic value in the poultry industry (Xie et al., 2025; Zhou et al., 2022). Modern breeding programs have achieved remarkable progress in terms of growth rate, feed efficiency, and carcass yield (Cai et al., 2025a). However, the molecular mechanisms underlying breast muscle yield, a key economic trait, remain incompletely understood, limiting the efficiency of genetic improvement. Phenotypic divergence in carcass and meat quality traits is often accompanied by variation in underlying molecular regulation, suggesting that genetic and regulatory differences contribute to these phenotypes (Zhao et al., 2024). Integrative molecular approaches have been widely applied in poultry to investigate muscle growth and meat quality (Zhou et al., 2025). In ducks, several studies have explored the genetic basis of growth, carcass traits, and meat quality via genomic prediction, GWAS, and multiomics approaches (Cai et al., 2025b). However, these studies have focused mainly on SNPs, protein-coding genes, or expression differences, largely overlooking the potential contribution of noncoding sequences.

Transposable elements (TEs) are mobile DNA sequences that can influence gene expression by acting as regulatory elements (Lanciano and Cristofari, 2020), as sources of lncRNAs (Kapusta et al., 2013), or by giving rise to transcription factor genes through transposase capture (Cosby et al., 2021). The evidence indicates that TE insertions can have substantial phenotypic effects. For example, the blue eggshell phenotype in chickens is caused by an EAV-HP insertion, which promotes the expression of SLCO1B3 in the shell gland of the oviduct (Wang et al., 2013). Wang et al. (2024) reported that a 6945-bp TE insertion causes a functional mutation in IGF2BP1, which significantly affects avian body weight and explains 27.61 % of the phenotypic variance. In addition, a 6634-bp TE insertion within an intron of MITF can induce a novel transcript and thereby affect white plumage formation (Wang et al., 2024). Together, these findings suggest that TEs provide a useful perspective for understanding phenotypic variation and genetic improvement in animals, particularly poultry.

Although previous studies have provided preliminary characterization of the composition and distribution of TEs in duck reference genomes, systematic research on population-level transposable element insertion polymorphisms (TIPs) and their effects on duck economic traits and gene expression is still lacking (Bai et al., 2024). In this study, population-level TIP variation and its relationship with breast muscle traits were investigated. Specifically, phenotypic variation in breast muscle traits and the population distribution of TIPs were compared among Cherry Valley ducks (CV), Liancheng white ducks (LCW), and 5 Jinling White Duck lines (collectively referred to as JLW). A TIP-GWAS was then performed to identify TIP loci associated with breast muscle yield and meat quality traits, and a TIP-eGWAS was conducted to identify TIP-related eGenes. Candidate genes were further prioritized by integrating weighted gene coexpression network analysis (WGCNA), differential expression analysis, and gene–trait correlation analysis. Because expansion of the myogenic cell population and subsequent differentiation and fusion are fundamental processes in skeletal muscle formation, proliferation and differentiation were prioritized for functional evaluation in primary duck myoblasts (Bentzinger et al., 2012). Cell-cycle progression and apoptosis were also examined as complementary indicators of cell proliferation and survival. This study provides new evidence for the genetic regulation of duck breast muscle yield from the perspective of TIP variation and offers candidate markers and genes for future functional and breeding studies.

Materials and methods

Animal sample collection

A total of 200 ducks from seven populations comprising CV, LCW, and five JLW lines were included in this study. Detailed sample information, including population assignment, sex, and data availability for whole-genome sequencing and RNA-seq, is provided in Table S1. All the ducklings were supplied by the Nanjing Institute of Livestock and Poultry Science (Baisheng Family Farm, Nanjing, China). All the ducklings hatched within the same period and were reared under standardized management and uniform environmental conditions.

At 49 d of age, which was the predetermined slaughter endpoint for this experiment and is commonly used in meat-type duck studies (Quaresma et al., 2025; Zaremba et al., 2024), the ducks were electrically stunned in a water bath and then slaughtered by severing the major cervical blood vessels, followed by complete exsanguination. The left breast muscle was collected immediately after slaughter. The breast muscle weight (BMW), breast muscle percentage (BMP), breast muscle thickness (BMT), and meat color parameters (L, a and b) were measured immediately after slaughter. The remaining muscle tissue was rapidly frozen on dry ice, transported to the laboratory of Sichuan Agricultural University, and stored until further analysis. All animal procedures were conducted in accordance with institutional and national guidelines for the care and use of animals and were approved by the Institutional Animal Care and Use Committee of Sichuan Agricultural University (Permit No. DKY20170913).

Determination of slaughter performance

The evaluated yield traits included BMW, BMP, BMT, and breast muscle area (BMA). BMW and BMP were determined according to the Chinese National Standard GB/T 29389-2024. BMT was measured at 3 locations (upper, middle, and lower) via a digital calliper, and the mean value was recorded. The BMA was determined from photographs taken on grid paper at a fixed camera height and quantified via Image-Pro Plus software.

Meat quality assessment

The measured meat quality indices included shear force (SF), meat color, water content (W), protein content (P), and intramuscular fat content (IMF) of the breast muscle. L, a and b were measured with a colorimeter (LS171, Linshang Technology, Shenzhen, China) at 3 locations on the left breast muscle, and the mean L, a, and b values were used for analysis. P, W, and IMF were determined with an NIR food analyser (F3020, Spectrastar Instruments, Guangzhou, China). For SF analysis, thawed breast muscle samples were trimmed along the muscle fibre direction into strips, and the tendons and epimysium were removed. The samples were heated in a constant-temperature water bath at 75°C for 30 to 40 min, cooled to room temperature, and analysed via a texture analyser (TA. XTC-18, Bosin Tech, Shanghai, China).

Whole-genome resequencing and TIP identification

At 49 d of age, after a 12-h fast, blood samples (5 mL per duck) were collected from the subaxillary vein. Genomic DNA was extracted from blood via the phenol/chloroform method. The DNA concentration, purity, and integrity were evaluated via a Nanodrop 2000 (ND-2000, Thermo Fisher Scientific, Waltham, MA, USA) and agarose gel electrophoresis. DNA samples that passed quality control were used for library construction and whole-genome resequencing, which was performed by Shanghai Personal Biotechnology Co., Ltd. Raw sequencing data were subjected to quality control to obtain clean reads, which were then aligned to the duck reference genome ZJU1.0 (GCF_015476345.1) via BWA-MEM. On the basis of the genome-wide variant data, loci with a missing rate ≤10 % and minor allele frequency (MAF) ≥5 % were retained via PLINK software (v1.9). Principal component analysis (PCA) based on genome-wide variant data was then performed via GCTA software (v1.94) to evaluate population genetic structure.

TEs in the reference genome were annotated via RepeatMasker to generate genomic position files for TE regions. On this basis, TIPs were identified via TEPID. Briefly, TEPID uses the tepid-map and tepid-discover modules to detect TE insertions and deletions relative to the reference genome on the basis of alignment information. A total of 3,628 population-level TIPs were initially identified and used for descriptive analyses, including chromosomal distribution, TE category and family composition, genomic feature annotation, and insertion frequency characterization. For downstream PCA and TIP-GWAS, the TIP genotype matrix was further filtered via PLINK v1.9, with a missing rate ≤10 % and MAF ≥5 %. After filtering, 543 TIPs were retained for population structure analysis and TIP-GWAS. PCA based on the filtered TIP genotype matrix was also performed to compare the population structure inferred from TIP variation with that inferred from genome-wide variants.

Transcriptome data analysis

Total RNA was isolated from the same location of the left breast muscle via the TRIzol method. Following reverse transcription, cDNA fragments of approximately 400–500 bp were size-selected for PCR amplification. The resulting PCR products were subsequently purified to construct the sequencing libraries. These libraries were subjected to paired-end sequencing on an Illumina platform via next-generation sequencing technology. The transcriptome sequencing service was provided and completed by Shanghai Personal Biotechnology Co., Ltd.

Clean reads were aligned to the duck reference genome (ZJU1.0, GCF_015476345.1) via HISAT2 (v2.1.0) (Kim et al., 2015), and an overall alignment rate of >95 % was achieved. The read counts mapped to each gene were quantified via featureCounts (Liao et al., 2014) and served as the raw gene expression values. Differential expression analysis was performed via the DESeq2 package. Genes with a false discovery rate (FDR) < 0.05 and a |log2(fold change)| > 1 were identified as significantly differentially expressed genes (DEGs). Gene expression levels were quantified as transcripts per million (TPM) values.

TIP-GWAS and TIP-eGWAS

Before association analysis, phenotypic traits and gene expression levels were normalized. On the basis of the population-level TIP genotype matrix, genome-wide association analyses for target traits and gene expression levels were performed via a mixed linear model implemented in EMMAX to control for confounding effects caused by population structure and individual relatedness. After filtering for missing data rate and MAF, 543 TIPs were tested in TIP-GWAS. The loci identified by phenotype-based TIP-GWAS were subsequently tested for associations with the expression levels of 10,619 genes. The model was specified as follows:

$$y=X\alpha +Z\beta +W\mu +e$$

In this model, y represents the vector of observed phenotypic values; Xα denotes the vector of fixed effects used to correct for population stratification, including breed, sex, and the first 5 principal components (PCA eigenvectors) derived from genome-wide TIP genotypes; Zβ represents the allelic substitution effect; Wμ denotes the random animal effect, with a variance–covariance matrix estimated from the kinship matrix on the basis of genome-wide TIP genotypes; and e represents the random residual vector.

For TIP-GWAS, although only 543 filtered TIPs were tested, the significance threshold was conservatively calculated on the basis of all 3,628 initially identified TIPs and was defined as P < 1/3628 (-log10(P) > 3.5597). For targeted TIP-eGWAS, Benjamini–Hochberg correction was applied across all 63,714 TIP–gene tests (6 loci × 10,619 genes), and FDR < 0.05 was used as the significance criterion. Associations that additionally met P < 1/3628 were retained as a conservative priority subset for downstream candidate-gene analyses.

WGCNA and identification of key modules and genes

WGCNA was used to construct a gene coexpression network on the basis of the transcriptome data. For percentage traits, values were first converted to proportions and then logit-transformed after a small offset was applied to avoid boundary values of 0 and 1. The other traits were log2-transformed. All transformed trait values were subsequently adjusted for breed by fitting linear models, and the residuals were used for downstream WGCNA trait association analysis (Fig. S1A–D). During quality control of the gene expression data, genes and samples with excessive missing values were excluded, and low-variance genes were filtered out via an interquartile range (IQR) filter with a cut-off of 0.5. Finally, 10,992 genes were retained for downstream analysis. Given that sex explained less than 5 % of the unique phenotypic variance across traits, whereas breed explained a much larger proportion, only breed was included in the WGCNA-related trait adjustment procedure (Fig. S1E).

The WGCNA package was used to construct the gene coexpression network. By systematically evaluating a soft-thresholding power range of 1–20, the optimal power was determined to be 11 on the basis of the scale-free topology fit index (R² > 0.85). An unsigned scale-free topology network was constructed using Pearson correlation coefficients to create a weighted adjacency matrix, which was then transformed into a topological overlap matrix (TOM). Gene modules were identified via hierarchical clustering with the following parameters: minModuleSize = 30, deepSplit = 2, and mergeCutHeight = 0.25. This process yielded 12 distinct gene modules. To identify key modules associated with the target traits, Pearson correlation analysis was performed between module eigengenes and phenotypic traits. Modules with an absolute correlation coefficient ≥ 0.2 and a P value < 0.01 were selected for further investigation.

To identify genes directly associated with trait variation, we selected significantly correlated genes for downstream analysis on the basis of the strength and statistical significance of gene–trait correlations. Specifically, within the WGCNA modules of interest, gene significance (GS) was calculated for each gene as the absolute value of the Pearson correlation coefficient between gene expression and the target trait. Gene–trait correlations were calculated using Pearson correlation, and Benjamini–Hochberg correction was applied separately for each trait. Priority eGenes associated with at least one of BMW, BMP, BMT, or BMA at FDR < 0.05 were used for downstream functional analysis. Genes associated with BMW, BMP, and BMT at FDR < 0.05 were used for candidate-gene prioritization.

Functional enrichment analysis

To explore the biological functions of genes associated with duck breast muscle traits, Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses were performed for 3 gene sets: genes within the selected WGCNA modules, DEGs, and eGenes significantly associated with the target traits. All enrichment analyses were conducted via the clusterProfiler package in R. Because a dedicated annotation package for ducks is not available in the R environment, gene symbols were converted to chicken (Gallus gallus) Entrez IDs via the org.Gg.eg.db database, and only successfully mapped genes were retained for enrichment analysis.

In silico functional annotation of LOC101798742

To characterize the candidate gene LOC101798742 (NCBI accession: XM_038182506.2), a series of in silico analyses were performed. The open reading frame (ORF) was predicted via NCBI ORFfinder (v3.10.1). The protein-coding potential of the transcripts was evaluated via CPC2 (Coding Potential Calculator 2). Conserved functional domains and motifs were identified via InterProScan (v5.66) against the Pfam and PANTHER databases. To assess evolutionary conservation, protein homology searches were conducted via NCBI BLASTp against the nonredundant (nr) and clustered NR databases.

Culture and treatment of primary duck myoblasts

Primary myogenic cells were isolated from duck embryos at E15–E17. All embryos were obtained from the Waterfowl Breeding Base of the Ya’an Campus, Sichuan Agricultural University. Briefly, pectoralis major muscles were excised and rinsed with PBS containing 1 % penicillin–streptomycin (P1400, Solarbio, Beijing, China) and then minced into approximately 1 mm³ fragments. The tissue was sequentially digested with 0.1 % type I collagenase for 60 min and 0.25 % trypsin for 20 min. Digestion was stopped by adding complete culture medium consisting of DMEM/F12 (1:1) (C11330500BT, Gibco, CA, USA) supplemented with 15 % fetal bovine serum (BL205A, Biosharp, Hefei, China) and 1 % penicillin–streptomycin. The resulting cell suspension was filtered through 200- and 400-mesh cell strainers, followed by centrifugation at 2,000 rpm for 10 min. After the supernatant was removed, the pellet was resuspended in PBS containing 1 % penicillin–streptomycin and washed 2–3 times. To reduce fibroblast contamination, a differential adhesion step was performed: rapidly adherent fibroblasts were allowed to attach, and the nonadherent cells (myogenic cells) were collected and transferred to new culture plates/dishes. The cells were maintained at 37°C in a humidified incubator with 5 % CO₂, and the medium was replaced every 48 h.

Transfection of siRNA and plasmids

The siRNAs targeting LOC101798742, the pEGFP plasmid, and the corresponding negative controls (NCs) were synthesized by Tsingke Biotechnology Co., Ltd. (Beijing, China). Myogenic cells at 60–70 % confluence were transfected with siRNA or plasmid DNA via Lipofectamine 3000 (Invitrogen, CA, USA) following the manufacturer’s protocol. The final concentrations were 0.15 μM for the siRNAs and 2.5 μg/mL for the plasmid DNA. All nucleotide sequences are listed in Table 1.

Table 1.

Nucleotide sequences of siRNAs.

Gene name	siRNA sequences (5′–3′)
si-NC	Sequence: UUCUCCGAACGUGUCACGUTT Antisense: ACGUGACACGUUCGGAGAATT
si-LOC101798742	Sequence: GCUAGUAGAAGGAGACAAA(dT)(dT) Antisense: UUUGUCUCCUUCUACUAGC(dT)(dT)

Cell cycle and apoptosis analysis

After treatment, the cells were detached with 0.25 % trypsin and collected in 1.5-mL microcentrifuge tubes. The cells were washed with PBS 1–2 times and fixed in 1 mL of 70 % ethanol for 24 h. After fixation, the cells were washed once with PBS and stained with propidium iodide (PI) via a cell cycle and apoptosis detection kit (C1052, Beyotime, Shanghai, China) according to the manufacturer’s instructions, followed by incubation at 37°C for 30 min in the dark. Flow cytometric analysis was performed using a CytoFLEX flow cytometer (Beckman Coulter, Brea, CA, USA). The main cell population was first gated on FSC—H versus SSC—H to exclude debris. Singlet cells were subsequently selected on the basis of PI pulse area and height (Y585-A versus Y585-H) to minimize interference from doublets and cell aggregates. DNA-content histograms were generated from the Y585-A signal for cell-cycle analysis, and the sub-G1 fraction was used as an indirect indicator of apoptosis. Data were analyzed using Kaluza Analysis software (v2.4.0.20305; Beckman Coulter).

Cellular immunofluorescence

After 48 h of treatment, immunofluorescence staining was performed. The culture medium was removed, and the cells were washed three times with PBS. The cells were fixed with 4 % paraformaldehyde for 15 min, permeabilized with 0.2 % Triton X-100 for 10 min, and blocked with immunostaining blocking solution (P0260, Beyotime, Shanghai, China) for 1 h. Primary antibody working solution was added, and the cells were incubated overnight at 4°C. The next day, the cells were incubated with the secondary antibody working solution and DAPI solution in the dark for 40 min and 5 min, respectively. Between each step, the cells were washed 3–5 times with PBS on a shaker. Finally, fluorescence images were captured via an inverted fluorescence microscope. All the antibodies were purchased from ABclonal Technology Co., Ltd. (Wuhan, China). The dilutions of the primary and secondary antibodies used were as follows: anti-MYH4 and anti-MYH7 (1:100) and anti-rabbit IgG (H+L) (1:500).

RNA extraction and real-time quantitative RT‒PCR (RT‒qPCR)

Total RNA was extracted from myogenic cells via an RNA extraction reagent (BL1620A, Biosharp, Hefei, China). The RNA concentration was measured via a nucleic acid concentration detector (Eppendorf, Hamburg, Germany), and RNA integrity was verified via agarose gel electrophoresis. The RNA was subsequently reverse transcribed into cDNA via a reverse transcription kit (A505-02; Exongen, Chengdu, China) according to the manufacturer’s protocol. RT‒qPCR was performed via a fluorescence quantitative PCR kit (A403-01, Exongen, Chengdu, China) on a fluorescence quantitative PCR machine (Bio-Rad, CA, USA). RT‒qPCR amplification was performed under the following conditions: initial denaturation at 95°C for 3 min, followed by 40 cycles of denaturation at 95°C for 15 s, annealing and extension at 60°C for 30 s. GAPDH was used as the internal reference gene, and relative gene expression was calculated via the 2^−ΔΔCt method. The primer sequences are listed in Table 2.

Table 2.

Sequences of Primers used for RT‒qPCR.

Gene name	Primer sequences (5′−3′)	Tm (°C)	GenBank ID
LOC101798742	F:GGCAACACATTCCGTCTCCT R:CCCATGTACGGAACCCCATC	60	XM_038182506.2
Pax7	F:GAGTTCAGGTGTGGTTCAGCA R:GAAATGGTGGTGGTTGGGTAG	60	XM_072026709.1
MyoD1	F:CGACGGCATGATGGAGTACA R:TGAGAGGCAATCAAGGCTGG	60	NM_001310358.1
MyoG	F:CGACGGGGAAAACTTCTTGG R:GTCGATGGACACCGTCTTGC	60	NM_001310376.1
Myf5	F:AGGAGGAGGCTGAAGAAAGTGA R:GCTCTGTCTCGGCAGGTGATA	60	NM_001310776.1
THY1	F:ACTGCCGCTACGAGAACAAA R:GTAGTCGTTGGTGGCCTTGA	60	XM_038167558.2
Col1A1	F:AAGAACGGCGATGATGGTGA R:ACCATCCAGACCACTGAAACC	60	XM_038168724.2
Cyclin D1	F:CTTGGATGCTGGAGGTTTGC R:GCGGCTTTTCTTGAGGGGTT	60	XM_013102661.5
PCNA	F:GAGCCAGTCCAACTGACCTT R:TCTGCAGACATGCTGAGTGTT	60	XM_038166796.2
CDK2	F:CCCGACTACAAACCCACCTT R:GTCGTAGTGCAGCATTTGCG	60	XM_072031804.1
Caspase3	F:GGGGTGACAAGTGCAGAAGT R:CTGTCTGCCTCAACACCAGA	60	XM_027456288.3
Bax	F:CGAAATGTTCGCGGACGGAG R:AGGAGACGATGGTGCGGAAA	60	XM_038172585.2
Bcl-2	F:TGAGGCCTTTGTTCGATTTC R:ATAAGCGCCAAGAGTGATGC	60	XM_027451677.3
GAPDH	F:GGTTGTCTCCTGCGACTTCA R:TCCTTGGATGCCATGTGGAC	60	XM_038180584.2

General statistical analysis

Statistical analyses were performed via R (v4.5.1) and SPSS (v27). Figures were generated via R (v4.5.1) and Origin (v2025b). Unless otherwise indicated, the data are presented as the means ± SDs from at least 3 independent biological replicates. Comparisons between 2 groups were performed via the Mann–Whitney U test. Comparisons among multiple groups were performed via the Kruskal–Wallis test followed by Bonferroni-corrected pairwise comparisons. Different lowercase letters in the figure indicate significant differences between groups (P < 0.05), whereas the same letters indicate no significant differences (P > 0.05). Additionally, * indicates P < 0.05, ** indicates P < 0.01, and *** indicates P < 0.001.

Results

Breast muscle yield and meat quality traits in different duck populations

The breast muscle yield traits differed significantly among the populations (Fig. 1A–D, Table S2). CV consistently presented the highest values for BMW, BMP, BMT, and BMA, whereas LCW presented the lowest values, with significant differences between the 2 populations for all 4 traits (P < 0.05). The five JLW lines presented intermediate phenotypes between those of CV and LCW. Among them, JLW(DAC) tended to be closer to CV for several yield traits, whereas JLW(CC) generally presented lower values than did the other JLW lines.

Fig. 1.

Breast muscle yield and meat quality traits in different duck populations. (A) Breast muscle weight (BMW), (B) breast muscle percentage (BMP), (C) breast muscle thickness (BMT), (D) breast muscle area (BMA), (E) shear force (SF), (F–H) meat color parameters (L, a, and b), (I) water content (W), (J) intramuscular fat content (IMF), and (K) protein content (P) in CV, LCW, and five JLW line ducks. The data are presented as the means ± SDs. Different lowercase letters indicate significant differences among groups (P < 0.05), whereas the same letter indicates no significant difference (P > 0.05).

Meat quality traits also varied among populations, although the overall differences were less pronounced than those observed for yield traits (Fig. 1E–K). CV presented a significantly greater SF than several other populations did, whereas LCW presented relatively greater W and P. IMF also differed among populations. In contrast, no significant differences were detected among populations for L, a, or b (P > 0.05). Overall, population differences were more evident for the breast muscle yield traits than for the meat quality traits. Therefore, subsequent analyses primarily focused on breast muscle yield traits, while meat quality traits were retained for comparison.

Genomic distribution of TIPs and population structure analysis

A total of 3,628 TIPs were initially identified from 200 whole-genome resequencing samples and were subsequently used for descriptive characterization (Table S3). These unfiltered TIPs were distributed across nearly all chromosomes, although their distribution was uneven, with more TIPs observed on larger chromosomes (Fig. 2A). On the basis of insertion frequency characterization, many TIPs presented low population frequencies, with a substantial proportion occurring in only one or a few individual ducks, suggesting that at least some of these insertions may be relatively recent or population specific (Fig. 2A). In terms of TE classification, LINEs represented the largest category, followed by DNA transposons and LTR elements (Fig. 2B). According to the results of the genomic feature annotation, the TIPs were located mainly in intergenic and intronic regions, with only a small proportion distributed in exonic, promoter, and downstream regions (Fig. 2C).

Fig. 2.

Genomic distribution of TIPs and population structure analysis. (A) Chromosomal distribution and insertion frequency distribution of the initially identified, unfiltered TIPs across the duck genome. Different colors represent class I retrotransposons and class II DNA transposons, and the inner histogram shows the insertion frequency distribution of TIPs in the population. (B) Abundances of different TE categories. (C) Distribution of TIPs across genomic functional regions, including promoter, downstream, exon, intron, and intergenic regions. (D) PCA based on the filtered TIP genotype matrix and genome-wide variants. (E) Distribution of TIP counts carried by individuals in different duck populations.

After filtering for the missing data rate and MAF, 543 TIPs were retained for population structure and association analyses. To compare patterns of genetic differentiation among duck populations, PCA was performed on the basis of both the filtered TIP genotype matrix and genome-wide variants. The TIP-based PCA was largely consistent with the genome-wide variant-based PCA: the AA, BB, and CC JLW lines clustered together, DAB and DAC grouped closely with each other, and LCW and CV were clearly separated (Fig. 2D). These findings indicate that filtered TIP variation can generally reflect the genetic structure of duck populations. In addition, the number of TIPs carried by individuals varied among populations, with the LCW generally carrying more TIPs and the CV tending to carry fewer TIPs (Fig. 2E).

Identification of TIP loci and eGenes associated with duck breast muscle traits

To determine whether TIP variation was associated with breast muscle traits, TIP-GWAS was subsequently performed. A total of 14 significant trait‒locus associations were identified via TIP-GWAS, including 4 TIP loci associated with BMW, 3 associated with BMP, 4 associated with BMT, 2 associated with BMA, and 1 associated with P (Fig. 3A, Fig. S2, and Table S4). These significant associations were mainly concentrated in breast muscle yield-related traits, suggesting that TIP variation may contribute more to the genetic regulation of breast muscle yield than to most meat quality traits. After redundant loci were merged, 6 nonredundant significant TIP loci were retained. To further investigate their potential effects on gene expression, TIP-eGWAS was performed.

Fig. 3.

Association analysis of TIP loci with phenotypic traits and functional enrichment of eGenes associated with target traits. (A) Heatmap of TIP-GWAS results. Each cell represents the -log₁₀(P) value for the association between a TIP locus and a phenotypic trait. (B) Chromosomal distribution of all eGenes associated with the representative locus chr6:36133712. (C) Correlation analysis between the eGenes associated with chr6:36133712 and the target traits. (D, E) GO and KEGG enrichment analyses of the 147 eGenes significantly associated with the target traits. * indicates P < 0.05, ** indicates P < 0.01, and *** indicates P < 0.001.

Using FDR < 0.05 as the significance criterion, targeted TIP-eGWAS identified 8,229 significant TIP–gene pairs involving 4,796 unique eGenes. Among these, 1,553 pairs involving 1,410 eGenes additionally met P < 1/3628 and were retained as a conservative priority subset for downstream analyses. Of the priority eGenes represented in the gene–trait correlation dataset, 147 were associated with at least one of BMW, BMP, BMT, or BMA (FDR < 0.05; Fig. 3B and 3C; Tables S5 and S6). For the representative locus chr6:36133712, 55 significant eGenes were identified, and these genes were distributed across multiple chromosomes, suggesting potential trans-regulatory effects (Fig. 3B). Further correlation analysis revealed that some eGenes were significantly associated with BMW, BMP, BMT, and BMA. Among them, LOC101798742, LOC119717899, MB, and LOC119716356 were positively correlated with multiple breast muscle yield traits (P < 0.05), whereas LOC101800226 and ACVR1C were negatively correlated (P < 0.05; Fig. 3C).

GO and KEGG enrichment analyses were further performed for all 147 eGenes significantly associated with the target traits. The GO results revealed that these genes were enriched mainly in terms related to fibroblast growth factor receptor binding, fibroblast growth factor receptor signalling, the cellular response to fibroblast growth factor stimulus, RNA splicing, and oxidoreductase activity (Fig. 3D). KEGG analysis further revealed that these genes were enriched mainly in energy- and metabolism-related pathways, including the tricarboxylic acid cycle, carbon metabolism, fatty acid metabolism, oxidative phosphorylation, and fatty acid degradation, as well as the MAPK and calcium signalling pathways (Fig. 3E). Overall, the enrichment results strongly pointed to metabolism-related networks, suggesting that TIP-associated expression variation may contribute to the regulation of duck breast muscle yield by affecting energy metabolism and acting in concert with growth factor signalling pathways.

WGCNA and candidate gene prioritization

To further narrow the candidate gene set, transcriptome coexpression patterns were subsequently examined by WGCNA. A total of 21,985 genes were detected in the breast muscle transcriptomes of CV, LCW, and JLW ducks. After removing genes and samples with excessive missing values and filtering out low-variance genes, 10,992 genes were retained for WGCNA, and these genes were assigned to 12 coexpression modules (Fig. 4A). Using the criteria of ∣r∣ ≥ 0.2 and P < 0.01, 2 modules, MEpurple and MEmagenta, were found to be significantly associated with the target traits (Fig. 4B). Among them, MEpurple was associated with breast muscle yield traits, including BMW and BMT, as well as meat color traits (L and a), whereas MEmagenta was associated mainly with b. Given the primary focus of this study on breast muscle yield, subsequent analyses focused on MEpurple, which showed significant correlations with BMW and BMT.

Fig. 4.

Transcriptome coexpression analysis and candidate gene prioritization. (A) Hierarchical clustering tree and module assignment of WGCNA coexpression modules. (B) Heatmap showing correlations between module eigengenes and breast muscle yield and meat quality traits. (C, D) GO and KEGG enrichment analyses of genes in the MEpurple module. (E) Venn analysis of the overlap among genes in the MEpurple module, genes differentially expressed between CV and LCW, genes significantly associated with BMW, BMP, and BMT, and eGenes identified for significant TIP loci. (F) Correlation analysis between the candidate gene LOC101798742 and breast muscle yield and meat quality traits in different duck populations and in the overall dataset. * indicates P < 0.05, ** indicates P < 0.01, and *** indicates P < 0.001.

GO and KEGG enrichment analyses of genes in the MEpurple module revealed that these genes were involved mainly in processes such as electron transport, cellular respiration, fatty acid β-oxidation, lipid oxidation, and the tricarboxylic acid cycle and were enriched in pathways such as oxidative phosphorylation, carbon metabolism, the citrate cycle, fatty acid degradation, and fatty acid metabolism (Fig. 4C and 4D). These results suggest that differences in breast muscle yield traits may be closely related to mitochondrial energy metabolism. Considering the significant differences in breast muscle yield traits between CV and LCW, differential expression analysis was further performed between these 2 populations. The GO and KEGG enrichment patterns of the DEGs were generally consistent with those of the MEpurple module (Fig. S3, Table S7).

To further prioritize TE-related candidate genes at the transcriptomic level, Venn analysis integrated the 1,410 eGenes, 55 genes in the MEpurple module, 3,574 genes differentially expressed between CV and LCW, and 204 genes significantly associated with BMW, BMP, and BMT. LOC101798742 remained the sole gene shared among all four sets and was therefore prioritized as the candidate gene corresponding to chr6:36133712 (Fig. 4E). Further analysis revealed that LOC101798742 was positively correlated with BMW, BMP, BMT, and BMA in the overall dataset. Population-specific analysis further revealed that this gene was generally positively correlated with breast muscle yield traits in most populations (Fig. 4F). In addition, genotype-based comparisons revealed that individuals carrying the chr6:36133712 insertion tended to have lower phenotypic values and lower LOC101798742 expression levels (Tables S8 and S9). In silico annotation further indicated that LOC101798742 is predicted to encode a 246-amino-acid protein containing a conserved peptidase S51 domain, supporting its annotation as the duck orthologue of DNPEP (Fig. S4). The predicted protein was highly conserved among avian species.

Effect of LOC101798742 on myogenic cell proliferation

Compared with that of the control, LOC101798742 expression was reduced by 55.19 % after siRNA transfection (P < 0.01; Fig. 5A). At 48 h after transfection, Cyclin D1, PCNA, and CDK2 expression decreased by 52.34 %, 53.06 %, and 43.18 %, respectively (P < 0.05; Fig. 5B–D). Cell cycle analysis revealed an increased proportion of G1-phase cells and a decreased proportion of S-phase cells in the knockdown group (P < 0.01; Fig. 5I). In contrast, LOC101798742 expression increased to 32.75-fold greater than that of the control after overexpression (P < 0.001; Fig. 5E). Overexpression significantly increased PCNA and CDK2 expression (P < 0.05; Fig. 5G and H), did not significantly affect Cyclin D1 (P > 0.05; Fig. 5F), reduced the proportion of G1-phase cells, and increased the proportion of G2-phase cells (P < 0.01; Fig. 5J). Together, these results indicate that LOC101798742 affects duck myoblast proliferation and cell cycle progression.

Fig. 5.

Effect of LOC101798742 on myoblast proliferation-related traits. (A) LOC101798742 expression after siRNA transfection. (B–D) Expression of proliferation-related genes (Cyclin D1, PCNA, and CDK2) after knockdown. (E) LOC101798742 expression after pEGFP-LOC101798742 transfection. (F–H) Expression of proliferation-related genes after overexpression. (I–J) Cell cycle distribution after knockdown or overexpression. 742 denotes LOC101798742; si-742 indicates siRNA targeting LOC101798742; pEGFP-742 indicates overexpression of LOC101798742. The data are presented as the means ± SDs. * indicates P < 0.05, ** indicates P < 0.01, and *** indicates P < 0.001.

Effect of LOC101798742 on myogenic cell differentiation

To evaluate the role of LOC101798742 in myogenic differentiation, myoblasts were transfected with siRNA or an overexpression plasmid and cultured in differentiation medium for 48 h. Knockdown significantly decreased Pax7 and MyoD1 expression and increased MyoG and Myf5 expression (P < 0.05; Fig. 6A–D). Overexpression significantly increased the expression of Pax7, MyoD1, MyoG, and Myf5 (P < 0.05; Fig. 6E–H). Immunofluorescence and fusion index analysis revealed that knockdown significantly increased myoblast fusion (P < 0.05), whereas overexpression had no significant effect (P > 0.05; Fig. 6I–J). These results suggest that LOC101798742 modulates the expression of myogenic regulatory factors and may influence the balance between proliferation and differentiation in duck myoblasts.

Fig. 6.

Effect of LOC101798742 on myogenic differentiation. (A–D) Expression of differentiation-related genes (Pax7, MyoD1, MyoG, and Myf5) after knockdown. (E–H) Expression of differentiation-related genes after overexpression. (I) Immunofluorescence staining of myotube formation (200 ×). (J) Fusion index after knockdown or overexpression. 742 denotes LOC101798742; si-742 indicates siRNA targeting LOC101798742; pEGFP-742 indicates overexpression of LOC101798742. The data are presented as the means ± SDs. * indicates P < 0.05, ** indicates P < 0.01, and *** indicates P < 0.001.

Effect of LOC101798742 on myogenic apoptosis

Finally, apoptosis-related changes were assessed. Knockdown decreased Caspase3 expression (P < 0.05), whereas overexpression had the opposite effect (Fig. 7A and C). Overexpression also increased the Bax/Bcl-2 ratio (P < 0.05; Fig. 7D). In flow cytometry, the sub-G1 fraction was used as an indirect indicator of apoptosis-related DNA content changes. Moreover, overexpression significantly increased the proportion of cells in the sub-G1 phase (P < 0.01), whereas knockdown had no significant effect (P > 0.05; Fig. 7E). These results suggest that the overexpression of LOC101798742 increases apoptosis-related changes in duck myoblasts.

Fig. 7.

Effect of LOC101798742 on apoptosis-related changes in duck myoblasts. (A, B) Apoptosis-related gene expression (Caspase3 and Bax/Bcl-2) after knockdown. (C, D) Apoptosis-related gene expression after overexpression. (E) Sub-G1 fraction measured by flow cytometry as an indirect indicator of apoptosis-related DNA content changes. 742 denotes LOC101798742; si-742 indicates siRNA targeting LOC101798742; pEGFP-742 indicates overexpression of LOC101798742. The data are presented as the means ± SDs. * indicates P < 0.05, ** indicates P < 0.01, and *** indicates P < 0.001.

Discussion

Recent studies have shown that TE insertions can alter local chromatin states and transcriptional regulatory elements, thereby reshaping gene regulatory activity (Lawson et al., 2023). In addition, TE insertions can influence splicing and expression networks and are increasingly recognized as important sources of intraspecific phenotypic variation (Ilık et al., 2025). In ducks, recent work has demonstrated that transposon insertions can directly contribute to increased body weight and white plumage formation, indicating that TEs have practical importance in duck domestication and improvement (Wang et al., 2024). Therefore, the present study extended the analysis to include breast muscle traits at the population level and showed that TIPs may represent an important source of genetic variation affecting duck breast muscle yield. In this study, significant TIP association signals were mainly concentrated in breast muscle yield-related traits, including breast muscle weight, breast muscle percentage, breast muscle thickness, and breast muscle area. These findings suggest that TIP variation in duck populations may preferentially affect muscle deposition and growth processes rather than act broadly on all quality traits. This pattern may reflect differences in trait architecture. Breast muscle yield is a more direct outcome of muscle growth and tissue accretion, whereas many meat quality traits are more strongly influenced by physiological, environmental, and postmortem factors, which may dilute the detectable effects of TIP variation (Baéza et al., 2022; Velleman, 2023). All populations were evaluated at the same chronological age under uniform rearing conditions, which allowed direct comparison of breast muscle traits among populations while minimizing age-related variation. However, given the differences in growth rate among these populations, the same chronological age may not necessarily represent an identical physiological stage.

Furthermore, eGWAS results revealed that significant TIP loci corresponded to multiple eGenes distributed across different chromosomes, indicating that the effects of these loci may not be limited to local cis-regulation of nearby genes but may instead involve broader expression regulatory networks (Bhati et al., 2023; Fueyo et al., 2022). These findings suggest that the relationship between TIPs and breast muscle yield may not follow a simple one locus–one gene–one trait model but may instead be mediated through a set of coordinately changing gene networks.

From a functional perspective, the eGenes significantly associated with the target traits identified in this study, together with the results of the subsequent coexpression module analysis, were more consistently enriched in pathways related to the tricarboxylic acid cycle, oxidative phosphorylation, carbon metabolism, and fatty acid metabolism. Skeletal muscle growth is highly energy-demanding and requires a sustained ATP supply to support protein synthesis and myofiber hypertrophy (Bentzinger et al., 2012; Sartori et al., 2021). Consistent with this, Hubert and Athrey (2022) compared the breast muscle transcriptomes of fast- and slow-growing chickens and reported signals indicative of mitochondrial dysfunction and reduced oxidative phosphorylation capacity in fast-growing breast muscle, suggesting that energy metabolism may represent a key limiting factor under rapid muscle accretion. Therefore, the convergence of functional themes between TIP-based eGWAS and WGCNA in the present study supports the idea that TIP-associated regulatory effects may influence duck breast muscle yield primarily through metabolic networks.

On the basis of the integrated screening of multiple lines of evidence, including eGenes corresponding to significant TIP loci, WGCNA module genes, differentially expressed genes, and gene–trait correlations, LOC101798742 was prioritized as the candidate gene. In addition, in silico annotation revealed that the protein encoded by LOC101798742 contains a conserved peptidase S51 domain and is a DNPEP homologue. Previous studies have shown that DNPEP is involved in maintaining intracellular amino acid homeostasis and that amino acid availability and aspartate metabolism are closely associated with mTOR signalling, nucleotide synthesis, and cell cycle progression (Cantor et al., 2009; Roberts et al., 2023; Sullivan et al., 2018). Therefore, LOC101798742 may participate in regulating the proliferation–differentiation balance of duck myoblasts by affecting local amino acid metabolism.

Notably, although the genotype-based expression differences of LOC101798742 did not reach statistical significance, the direction of change was consistent with the overall results of the present study. Breast muscle yield is a typical complex quantitative trait that is usually influenced by many loci with small effects (Deng et al., 2019). Moreover, distal expression regulatory signals often have weaker marginal effects and are subject to a greater multiple-testing burden, making them less likely to reach statistical significance (Zhou and Cai, 2021). Nevertheless, trans-acting effects may still connect broader regulatory networks and contribute to the formation of complex traits (Dutta et al., 2022). Therefore, in this study, we placed greater emphasis on the consistency of multilayer evidence rather than relying solely on statistical significance from any single analysis to prioritize candidate genes.

In vitro experiments showed that LOC101798742 knockdown inhibited duck myoblast proliferation, altered cell-cycle distribution, and promoted cell fusion. In contrast, LOC101798742 overexpression affected cell-cycle progression and apoptosis-related signaling but did not produce a fully reciprocal phenotype. This asymmetric response may reflect dose-dependent and network-level regulation within the myogenic regulatory system, such that increased and decreased gene expression do not necessarily lead to opposite biological effects (Melendez et al., 2021; Reid et al., 2020; Yamamoto et al., 2018). Therefore, the increased expression of MyoG and Myf5 observed under both knockdown and overexpression conditions should not be interpreted as evidence that LOC101798742 exerts the same directional effect on myogenic differentiation. Importantly, muscle yield formation depends not only on differentiation strength but also on the expansion of the myogenic progenitor pool at early stages and the coordination of differentiation timing thereafter (Bentzinger et al., 2012). In mouse satellite cells, BMP signalling has been shown to support population expansion by preventing premature myogenic differentiation (Ono et al., 2011). In addition, recent evidence indicates that granulin restrains muscle stem cell differentiation, and its deficiency leads to impaired hypertrophy due to premature differentiation of these cells (Zhang et al., 2024). Accordingly, the phenotype observed after LOC101798742 knockdown, characterized by suppressed proliferation together with enhanced fusion, is more likely to reflect a tendency toward premature differentiation rather than a simple muscle growth-promoting effect.

Taken together, the results of TIP-GWAS, TIP-eGWAS, transcriptome analyses, and in vitro functional assays support an association among chr6:36133712, LOC101798742, and duck breast muscle yield, but do not establish a direct causal chain. The current evidence is derived mainly from population-level association analyses and in vitro myoblast assays; therefore, whether chr6:36133712 directly regulates LOC101798742 and whether LOC101798742 causally affects breast muscle yield in vivo remain to be determined. In addition, TIP identification on the basis of short-read resequencing data may still be limited by mapping bias in complex genomic regions and by the resolution of insertion breakpoint detection. The cellular assays used here mainly addressed myoblast proliferation, differentiation, and survival and did not evaluate other determinants of breast muscle yield, such as myofiber hypertrophy, fiber-type composition, fibrosis, or protein turnover. Future studies integrating long-read sequencing, allele-specific expression analysis, developmental-stage expression profiling, and targeted editing of the insertion locus will be needed to further clarify the roles of chr6:36133712 and LOC101798742 in the formation of duck breast muscle yield traits.

Conclusion

This study systematically characterized the distribution patterns of TIPs in duck populations and demonstrated that TIP variation is closely associated with breast muscle yield traits. Through the combined analysis of TIP-GWAS and TIP-based eGWAS, we identified multiple significant TIP loci associated with breast muscle weight, breast muscle percentage, breast muscle thickness, and breast muscle area, together with their corresponding eGenes, suggesting that transposable element insertion variation may participate in the formation of duck breast muscle yield traits by affecting gene expression regulatory networks. On the basis of multiple layers of bioinformatic evidence, LOC101798742 was prioritized as a candidate gene. Further in vitro functional experiments indicated that this gene may be involved in the regulation of duck myoblast proliferation, cell cycle progression, and differentiation. In summary, this study provides new evidence for the genetic regulation of duck breast muscle yield traits from the perspective of transposable element insertion variation and provides an important foundation for future molecular marker mining and functional gene studies.

Author Contributions

Weikang Ling performed the major bioinformatic analyses in this study, including transposable element insertion polymorphism analysis and transcriptomic analysis. He also conducted the validation experiments and related data analysis, participated in the interpretation of the results, and drafted the original manuscript. Cui He participated in the validation experiments, analyzed the experimental data, prepared and organized the relevant figures, and contributed to the interpretation and verification of the experimental results. Hong Tang participated in the validation experiments, analyzed the experimental data, and contributed to the interpretation and verification of the experimental results. Yuan Bai developed the methods and analytical workflow for the identification of transposable elements and transposable element insertion polymorphisms, contributed to the establishment of the initial analytical strategy and framework of the study, and participated in the related data processing and analysis. Yang Xi contributed to the development of the analytical framework and methods, including evaluation of data characteristics, data-processing strategies, measurement approaches, and the design and optimization of analytical methods. He also participated in data analysis and interpretation of the results. Jingjing Qi provided guidance on the analytical methods and participated in data analysis, interpretation of the results, and evaluation of the analytical findings. Shihao Guo participated in the implementation of the analytical methods and was mainly responsible for data organization, preprocessing, and statistical analysis. He also contributed to the summarization and interpretation of the analytical results. Zhao Yang participated in the implementation and optimization of the analytical workflow and was mainly responsible for data processing, quality checking, and statistical analysis. He also contributed to verification of the analytical results and to resolving issues encountered during implementation of the analytical methods. Xu Han participated in data analysis and result validation and was mainly responsible for consistency checks of analytical outputs, organization of the results, and verification of figures and other analytical outputs. He also contributed to the interpretation of the study findings. Anqi Huang contributed to the overall conception and methodological design of the study, with particular involvement in the discussion and refinement of the early animal-management procedures, experimental implementation plan, and sampling strategy. She also participated in biological sample collection and optimization of the study design. Shengqiang Hu contributed to the study design and experimental implementation, with particular involvement in animal management, biological sample collection, and the organization and implementation of the relevant experimental procedures. He also contributed to refinement of the sampling procedures and study design. Chunchun Han contributed to the overall conception and methodological design of the study, with particular involvement in the phenotypic measurement strategy, data recording and organization, and statistical analysis and interpretation of the relevant results. She also participated in the discussion and refinement of the study design and analytical framework. Liang Li contributed to the overall conception and methodological design of the study, provided guidance on data-analysis strategies and specific analytical approaches, and participated in the verification, interpretation, and discussion of key analytical results. He also provided academic guidance on methodological issues and interpretation of the findings and contributed to the revision and improvement of the study design and important intellectual content of the manuscript. Hehe Liu was responsible for the overall conception of the study and development of the overall research framework. He provided guidance on the study design, analytical strategy, and experimental validation plan and supervised the overall implementation and progress of the research. He participated in the analysis, interpretation, and discussion of key data and findings, provided academic oversight of the study conclusions, and critically reviewed and revised the overall structure, major research content, and important intellectual aspects of the manuscript. As the corresponding author, he was responsible for the overall academic supervision of the study and final review of the manuscript. All authors critically reviewed and revised the manuscript for important intellectual content, reviewed and explicitly approved the exact final version of the manuscript for publication, and agreed to be personally and publicly accountable for all aspects of the work. All authors also agreed to ensure that any questions concerning the accuracy or integrity of any part of the work are fully investigated and appropriately resolved.

Declaration of competing interest

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.

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.

Data availability

The whole-genome resequencing and RNA-seq data supporting this study are currently being deposited in the Genome Sequence Archive under BioProject PRJCA062477. All other data supporting the findings of this study are available from the corresponding author upon reasonable request.