Weighted single-step GWAS identified candidate genes associated with semen traits in Rhode Island Red chickens

Abstract

Semen quality traits in chickens are critical indicators reflecting reproductive efficiency, breeding progress, and economic profitability in the poultry industry. Understanding the genetic architecture and enabling targeted genetic improvement of these traits are essential for enhancing poultry production efficiency. In this study, we estimated genetic parameters and performed weighted single-step GWAS for five semen traits in RIR chickens to reveal their genetic architecture and identify key candidate genes and QTLs associated with semen quality. The results showed that heritability estimates for SEVOL, SECON, SPMOT, SPABR, and SPCOUNT ranged from 0.097 to 0.322, indicating low to moderate heritability. 29 significant QTL regions and 12 candidate genes (NFKB1, UBE2D3, PPP3CA, EIF4E, H2AFZ, DOCK2, MTNR1A, TACC3, ADCYAP1, GFRA1, ABLIM1, and CASP7) associated with semen traits were identified by WssGWAS. These QTL regions were located on chromosomes 4 and 13 for SEVOL, chromosome 4 for SPMOT, and chromosomes 2 and 6 for SPABR. Interestingly, four prominent consecutive QTL regions were shared between SEVOL and SPMOT. These QTL were located at 59.83–61.43 Mb, explaining 11.55% and 16.33% of the genetic variance for SEVOL and SPMOT, respectively. The largest-effect SNPs within these four QTL intervals exhibited significant effects on both semen volume and sperm motility, and these markers could be used for marker-assisted selection of semen traits. This study provides further insights into the genetic architecture of chicken semen traits, improves our understanding of their molecular regulation, and identifies valuable QTL and candidate genes. These findings offer a scientific basis for genetic improvement and marker-assisted selection of semen quality.

Introduction

The rapid adoption of artificial insemination in chicken breeding has made the accurate selection of roosters with high‑quality semen increasingly important for improving overall reproductive performance. A rooster’s reproductive potential is mainly determined by semen volume, sperm motility, sperm concentration, and other key indicators (Thananurak, et al., 2020). Breeding programs have mainly focused on selecting chickens with uniform plumage color (Peng, et al., 2025; Yang, et al., 2025; Yu, et al., 2026), excellent growth performance (Ding, et al., 2025; Jiang, et al., 2025; Wen, et al., 2025), superior carcass traits (Pan, et al., 2026; Zhu, et al., 2025c), disease resistance traits (Li, et al., 2025b; Liu, et al., 2025; Yevshin, et al., 2024), and outstanding egg production (Volkova, et al., 2025; Zhang, et al., 2026; Zhu, et al., 2025a) in hens, while often neglecting the potential contribution of semen traits to reproductive performance (Li, et al., 2025a). Laying hen breeding programs usually conduct intensive selection based on a male-to-female ratio of 1:6 to 1:10. In recent years, researchers have shown growing interest in the molecular processes and genetic mechanisms underlying sperm traits (Li, et al., 2026). Identifying genetic markers and candidate genes associated with chicken semen traits can accelerate the genetic progress of these traits.

With the rapid development of high-throughput genotyping and molecular technologies, as well as the continuous decline in sequencing costs, genome-wide association studies (GWAS) have been widely applied to identify genetic markers for important economic traits in animal and plant breeding. However, only a few studies have reported candidate genes and markers associated with chicken semen traits. (Romanov, et al., 2025) used 60K SNP chip data to identify 26 candidate genes associated with semen traits. (Zhang, et al., 2023a) performed imputation using 55K SNP chip data and identified four candidate genes related to semen volume.

Weighted single-step GWAS (WssGWAS), proposed by (Wang, et al., 2012), estimates single nucleotide polymorphism (SNP) effects using genomic estimated breeding values (GEBV) from single-step genomic best linear unbiased prediction (ssGBLUP) (Aguilar, et al., 2010) based on all animals with phenotypic, genotypic, and pedigree information. Furthermore, this approach accommodates heterogeneous variances among SNPs, thereby improving the precision of SNP effect estimation. Therefore, when the number of animals with both phenotypic and genotypic records is limited and the traits are controlled by quantitative trait loci (QTL) with large effects, WssGWAS may perform better than traditional GWAS methods. This method has been widely applied in growth, carcass, reproductive, and semen traits in pigs and cattle (Buaban, et al., 2022; de Araujo Neto, et al., 2019; Vahedi, et al., 2022; Zhang, et al., 2023b), as well as growth traits in chickens (Zhu, et al., 2025b), but has not yet been reported for chicken semen traits.

In this study, we used WssGWAS to identify genetic regions and candidate genes associated with semen traits in Rhode Island Red chickens. Rhode Island Red chickens are renowned worldwide for their excellent laying performance and strong adaptability, and are commonly used as parental lines in chicken breeding programs (Xiao-Ke, et al., 2025). This study aims to provide an in-depth understanding of the genetic architecture of semen traits and offer a theoretical basis for the genetic improvement of semen traits.

Materials and methods

Animal population and phenotypic data

A total of 262 semen records were collected from 138 roosters (60 half-sib families) of the same hatching batch at the experimental farm of IAS-CAAS (Changping, Beijing, PR China). All individuals were raised under uniform feeding standards and provided ad libitum access to water. They were housed in individual cages and maintained under a 16L:8D photoperiod (light intensity: 20 lx), with the dark period from 22:00 to 06:00 h. The nutrient level of the diet was designed according to China Agricultural Industry Standards (NY/T 33-2004). Semen traits were measured at 46 and 48 weeks of age. Before the formal semen quality evaluation, roosters were trained for semen collection every other day for one week using the abdominal massage technique (Burrows and Quinn, 1937).

Five semen traits were measured, including semen volume (SEVOL), semen concentration (SECON), sperm motility (SPMOT), sperm abnormality (SPABR), and total sperm count (SPCOUNT). The ejaculate was collected manually in a weighted micropipette, maintained at 37 °C, and taken for semen quality analysis within 10 min of collection. SEVOL was measured by a weighing method following World Health Organization (WHO) laboratory manuals (W.H.O., 2021). SECON, SPMOT, and SPABR were evaluated using a CASA system (ML-608JZII; Nanning Songjingtianlun Biotechnology Co., Ltd., Nanning, China). Briefly, 10 µL of semen was diluted in 990 µL pre-warmed (37°C) Dulbecco’s Modified Eagle’s Medium (DMEM, Gibco, C11995500BT, Beijing, China). After mixing, 10 µL of the diluted semen was loaded into a 20‑µm‑deep sperm counting chamber placed on a 37°C pre-warmed microscope stage. For each sample, five microscopic fields were captured using the CASA system equipped with a 10 × negative phase‑contrast lens (ML-608JZII). Image brightness and contrast were adjusted to improve sperm detection, and sperm parameters were analyzed automatically by the system. The SPCOUNT was calculated by multiplying the SEVOL by the SECON.

Genotyping and quality control

Genomic DNA was extracted from 121 individuals using the Magnetic Universal Genomic DNA Kit (TIANGEN BIOTECH (BEIJING) CO., LTD.). Library preparation was performed using the BGI Optimal DNA Library Prep Kit (BGI, Shenzhen, China), and 150-bp paired-end (PE150) sequencing reads were generated on the DNBSEQ-T7 platform (BGI, Shenzhen, China). We then called SNPs for these samples using GRCg6a (v112) reference genome with GATK (v4.6.0.0) (McKenna, et al., 2010) default parameters. Subsequently, we performed haplotype phasing and genotype imputation using Beagle (v5.1) (Browning, et al., 2018). After genotype imputation, we removed SNPs with MAF < 0.01, resulting in 22,857,844 autosomal biallelic SNPs. We use PLINK (v1.9) and conduct quality control with the following parameters: –mind 0.05, –geno 0.05, –maf 0.05, and –hwe 1e-5. Finally, a total of 7,728,936 single nucleotide polymorphisms (SNPs) were retained. To avoid the influence of linkage disequilibrium (LD), SNP pruning was performed using PLINK (v1.9) with the parameter "–indep-pairwise 50 5 0.2". A total of 393,565 SNPs were retained for subsequent genetic evaluation and WssGWAS analysis. Genome-wide SNP density plots were constructed using the CMplot (v4.5.1) package to visualize the distribution of SNPs across autosomes.

Statistical models

Variance components for five semen traits were estimated using the average information restricted maximum likelihood (AIREML) method implemented in the BLUPF90 software (Misztal, et al., 2002). The following repeatability model was used to estimate variance components:

$$\mathbf{y}=\mathbf{X}\beta +\mathbf{Za}+\mathbf{Wp}+\mathbf{Dt}+\mathbf{Intv}+\mathbf{e}$$

where y is the vector of semen traits phenotypic value, β is the fixed effects (tier‑specific cage effects), a∼N(0, H${\sigma }_a^2$) is a vector of additive genetic effects, H is the matrix that combines pedigree and genomic information (Aguilar, et al., 2010; Ole F Christensen, 2010), p∼N(0, I${\sigma }_p^2$) is a vector of random permanent environmental effects, covariates Dt and Intv denote the measured date of semen traits and the interval between two subsequent semen collections in days, respectively, e∼N(0, I${\sigma }_e^2$) is the vector of random residuals, and X, Z, and W are the incidence matrices of β, a, and p, respectively. I is the identity matrix. The ${\sigma }_a^2$, ${\sigma }_p^2$and ${\sigma }_e^2$ components are the additive genetic, permanent environmental and residual variances, respectively. H matrix was calculated as:

$${\mathbf{H}}^{-1}={\mathbf{A}}^{-1}+\left[\begin{array}{cc}0& 0\\ 0& {\mathbf{G}}^{-1}-{\mathbf{A}}_{22}^{-1}\end{array}\right]$$

where ${\mathbf{A}}_{22}$ is a submatrix of A for the genotyped individuals; and $\mathbf{G}$ is the genomic relationship matrix, calculated as $\mathbf{G}=\frac{{\mathbf{ZDZ}}^{\prime }}{{\sum }_{i-1}^{m}2p_i\left(1-p_i\right)}$ (VanRaden, 2008), where Z is a genotype matrix adjusted for allele frequencies (with 0, 1, and 2 representing genotypes AA, Aa, and aa, respectively). D is a diagonal matrix of weights for SNP variance (initially D = I, I is the identity matrix), m is the number of SNPs, and $p_i$ is the minor allele frequency of the $i^{th}$ SNP.

Weighted single-step GWAS

Estimates of SNP effects and weights for WssGWAS were obtained according to (Wang, et al., 2012) with the following steps:

• 1.First iteration (t = 1), ${\mathbf{D}}_{\left(t\right)}$= I, ${\mathbf{G}}_{\left(t\right)}$= λ${\mathbf{D}}_{\left(t\right)}{\mathbf{Z}}^{\prime }$ and λ =$\frac{1}{{\sum }_{i=1}^{m}2p_i\left(1-p_i\right)}$ (VanRaden, 2008);
• 2.Calculated GEBV of the entire dataset via ssGBLUP;
• 3.Estimated of SNP effects via ${\widehat{\mathbf{g}}}_t=\lambda {\mathbf{D}}_{\left(t\right)}{\mathbf{Z}}^{\prime }{\mathbf{G}}_{\left(t\right)}^{-1}\widehat{\mathbf{a}}$, where $\widehat{\mathbf{a}}$ is the GEBV of individuals that were also genotyped;
• 4.Calculated SNP weights for the next iteration via $d_{i\left(t+1\right)}$=${\widehat{g}}_{i\left(t\right)}^{2}2p_i\left(1-p_i\right)$, where $i$ is the $i^{th}$ SNP;
• 5.Normalized SNP weights to keep the SNP weights the total genetic variance via ${\mathbf{D}}_{\left(t+1\right)}=\frac{tr\left({\mathbf{D}}_{\left(1\right)}\right)}{tr\left({\mathbf{D}}_{\left(t+1\right)}\right)}{\mathbf{D}}_{\left(t+1\right)}$;
• 6.${\mathbf{G}}_{\left(t+1\right)}=\lambda \mathbf{Z}{\mathbf{D}}_{\left(t+1\right)}{\mathbf{Z}}^{\prime }$ was calculated;
• 7.Let t = t + 1 and iterate from step 2.

This procedure was run for two iterations based on the predicted accuracies of GEBV according to and (Wang, et al., 2014). The SNP window consisted of a region of consecutive SNPs located within 0.4 Mb, which was on the basis of the linkage disequilibrium (LD) decay range for chicken pure breeds (Fu, et al., 2015) (Fig. S1). The LD decay analysis was conducted by PopLDdecay (v3.40) software. The genetic variance explained by the $i^{th}$ set of consecutive SNPs ($i^{th}$ SNP window) was calculated via:

$$\frac{Var\left(a_i\right)}{{\sigma }_a^2}\times 100\%=\frac{Var\left({\sum }_{j=1}^m{\mathbf{Z}}_j{\mathrm{g}}_{\mathrm{j}}\right)}{{\sigma }_a^2}\times 100\%$$

where $a_i$ is the genetic variance of the $i^{th}$ SNP window, ${\sigma }_a^2$ is the total additive genetic variance, ${\mathbf{Z}}_j$ is the vector of the $j^{th}$ SNP for all individuals and ${\mathrm{g}}_j$ is the effect of the $j^{th}$ SNP within the $i^{th}$ window. Manhattan plots of these windows were shown using the R software.

The WssGWAS was conducted using the BLUPF90 software family in an iterative way adapted for genomic analyses (Wang, et al., 2014). Briefly, for phenotype, pedigree and genomic file preprocessing, we made use of RE-NUMF90 and variance components were estimated using AIREMLF90, which were then used in BLUPF90 to predict GEBV. SNP effects were then calculated using the postGSF90 procedure.

Relevant SNP windows, candidate genes, and gene enrichment analysis

Based on the WssGWAS results, SNP windows explaining 1% or more of the total genetic variance (Sollero, et al., 2017; Zhu, et al., 2025b) were selected as candidate QTL regions. Assuming equal contribution across all windows, the expected proportion of genetic variance explained by each window was 0.043% (100/2345). The threshold of 1% applied in the present study was deemed appropriate for our analysis. Candidate QTL regions on the same chromosome that contained three or more overlapping or adjacent QTLs were defined as significant QTL regions for semen traits.

Genomic annotations were derived from the GRCg6a (v112) reference gene database (Ensembl, http://www.ensembl.org). For each candidate gene, a manual search was conducted in the National Center for Biotechnology Information (NCBI, http://www.ncbi.nlm.nih.gov) database to identify previously established correlations with the target traits. Enrichment analyses based on Gene Ontology (GO) (Ashburner, et al., 2000) and the Kyoto Encyclopedia of Genes and Genomes (KEGG) (H, et al., 1999) were subsequently performed using the Database for Annotation, Visualization, and Integrated Discovery (DAVID) (Sherman, et al., 2022) to characterize the functional attributes of the identified candidate genes.

Results

Descriptive statistics and genetic parameters for the semen traits

Descriptive statistics for the phenotypes and heritability estimates for semen traits in the RIR chickens are shown in Table 1. The mean values for SEVOL, SECON, SPMOT, SPABR, and SPCOUNT traits were 0.6 mL, 5.31 × 10⁹/mL, 77.39%, 14.55%, and 3.23 × 10⁹, respectively. Among the semen traits, the coefficients of variation ranged from 12.05% to 47.74%, with sperm motility showing the smallest value and total sperm count the largest. Density histograms indicated no outliers for any of the semen traits (Fig. S2). The heritability estimates for SEVOL, SECON, SPMOT, SPABR, and SPCOUNT were 0.322, 0.160, 0.097, 0.186, and 0.260, respectively. These traits exhibited low-to-moderate heritability, which supports the feasibility of subsequent molecular marker screening and marker-assisted selection (MAS).

Table 1.

Descriptive statistics and heritability estimation of the semen traits in the RIR chickens.

Traits	Records	Mean±SD	Median	CV/%	Min	Max	$h^2$±se
SEVOL/mL	262	0.60±0.18	0.60	30.66	0.13	1.19	0.322±0.081
SECON/(109/mL)	262	5.31±1.79	5.36	33.65	0.79	10.39	0.160±0.078
SPMOT/%	262	77.39±9.32	78.06	12.05	44.17	98.22	0.097±0.083
SPABR/%	262	14.55±3.23	14.15	22.21	8.00	25.74	0.186±0.082
SPCOUNT/109	262	3.23±1.54	3.09	47.74	0.32	8.72	0.260±0.082

SEVOL: semen volume; SECON: semen concentration; SPMOT: sperm motility; SPABR: sperm abnormality; SPCOUNT: total sperm count; CV: Coefficient of variation; Min: minimum; Max: maximum; $h^2$: heritability.

Phenotypic and genetic correlations among the five semen quality traits are shown in Table S1. For phenotypic correlations, SPCOUNT showed strong positive correlations with SEVOL (0.722 ± 0.093) and SECON (0.705 ± 0.093); SPMOT had a moderate positive correlation with SECON (0.331 ± 0.093) and a weak negative correlation with SPABR (-0.216 ± 0.093); weak correlations were observed among the remaining phenotypic traits. For genetic correlations, SPCOUNT exhibited extremely strong positive genetic correlations with SEVOL (0.815 ± 0.011) and SECON (0.743 ± 0.039); SPMOT showed an extremely strong negative genetic correlation with SPABR (-0.526 ± 0.056) and a moderate positive genetic correlation with SPCOUNT (0.486 ± 0.052); SEVOL had a moderate negative genetic correlation with SPMOT (-0.206 ± 0.057), and the genetic correlations among the remaining traits were weak.

Genome-wide SNPs density

The distribution of SNP density across autosomes before and after linkage disequilibrium (LD) pruning was visualized using the number of SNPs within a 1 Mb sliding window (Fig. 1). SNPs exhibited a uniform distribution across the autosomes. LD pruning effectively mitigated the effects of linkage disequilibrium while preserving this overall distribution pattern, thus providing a more independent and robust marker set for subsequent genomic analyses. Principal component analysis (PCA) based on the genomic relationship matrix revealed no outlier individuals (Fig. S3).

Fig. 1.

Genome-wide SNP density plots across autosomes. A and B represent the SNP density distribution within a 1 Mb sliding window before and after LD-based pruning, respectively.

Weighted single-step GWAS of semen traits

This study employed a WssGWAS to analyze the genetic variation of semen traits in RIR chickens explained by 0.4 Mb windows. The window number for the proportion of genetic variance in semen traits of Rhode Island Red chickens explained by each 0.4 Mb window are shown in Table S2. A total of 18, 15, 15, 18, and 11 candidate QTL regions (each explaining more than 1% of the genetic variance) were identified for the SEVOL, SECON, SPMOT, SPABR, and SPCOUNT traits, respectively (Fig. 2 and Fig. S4). Significant QTLs associated with the SEVOL trait were located on chromosomes 4 and 13, respectively. These significant QTL regions explained 17.69% (7 significant QTLs) and 4.46% (3 significant QTLs) of the genetic variance, respectively. The candidate genes identified included NFKB1, UBE2D3, PPP3CA, EIF4E, H2AFZ, and DOCK2 (Fig. 2A and Table 2). Significant QTLs associated with the SPMOT trait were located on chromosome 4. These significant QTL regions explained 21.12% (7 significant QTLs) of the genetic variance. The candidate genes identified included PPP3CA, NFKB1, UBE2D3, MTNR1A, TACC3, EIF4E, and H2AFZ (Fig. 2C and Table 2). Significant QTLs associated with the SPABR trait were located on chromosomes 2 and 6, respectively. These significant QTL regions explained 11.65% (6 significant QTLs) and 12.10% (5 significant QTLs) of the genetic variance, respectively. The candidate genes identified included ADCYAP1, GFRA1, ABLIM1, and CASP7 (Fig. 2D and Table 2). No significant QTL regions were identified for the SECON and SPCOUNT traits (Fig. 2B and E). Interestingly, four prominent consecutive QTL regions were shared for the SEVOL and SPMOT traits. These QTL regions were located within the interval of 59.83–61.43 Mb, explaining 11.55% and 16.33% of the genetic variance for SEVOL and SPMOT, respectively.

Fig. 2.

Genetic variances of five semen traits in RIR chickens explained by 0.04 Mb genomic windows. The A-E are the SEVOL, SECON, SPMOT, SPABR, and SPCOUNT traits, respectively. The red horizontal lines indicate the thresholds for the 1% significance level of genetic variance. SEVOL: semen volume; SECON: semen concentration; SPMOT: sperm motility; SPABR: sperm abnormality; SPCOUNT: total sperm count.

Table 2.

The significant genetic variance regions and candidate genes for semen traits in RIR chickens.

Traits	Chr	Regions(Mb)	gVar(%)	Candidate genes
SEVOL	4	60.63-61.03	5.60	NFKB1, UBE2D3, CISD2
	4	60.23-60.63	3.14	PPP3CA
	4	65.86-66.26	1.89	USP46, SGCB, DCUN1D4
	4	65.46-65.86	1.66	CHIC2, LNX1, FIP1L1
	4	59.83-60.23	1.45	TSPAN17, EIF4E, METAP1, ADH4, C4orf17, TRMT10A, MTTP, DAPP1, LAMTOR3, DNAJB14, H2AFZ, ADH1C
	4	59.01-59.41	1.42	–
	4	61.03-61.43	1.36	KLKB1, PDLIM3, SORBS2, TLR3, CYP4V2
	4	58.61-59.01	1.17	UNC5C
	13	4.76-5.16	1.96	DOCK2, SLIT3
	13	4.36-4.76	1.45	DOCK2, KCNIP1, KCNMB1, LCP2
	13	8.41-8.81	1.06	PTTG2, SLU7, C1QTNF2, FABP6, PWWP2A, TTC1
SPMOT	4	60.23-60.63	6.35	PPP3CA
	4	60.63-61.03	5.49	NFKB1, UBE2D3, CISD2
	4	61.03-61.43	3.42	KLKB1, PDLIM3, SORBS2, TLR3, CYP4V2
	4	61.43-61.83	2.02	MTNR1A
	4	83.64-84.04	1.73	TACC3, TMEM129, SLBP, FAM53A
	4	59.83-60.23	1.06	TSPAN17, EIF4E, METAP1, ADH4, C4orf17, TRMT10A, MTTP, DAPP1, LAMTOR3, DNAJB14, H2AFZ, ADH1C
	4	59.42-59.82	1.03	RAP1GDS1, TSPAN17
SPABR	2	100.88-101.28	3.32	MYL12A, MYOM1, LPIN2, SMCHD1, NDC80
	2	100.48-100.88	2.25	DLGAP1, TGIF1
	2	102.50-102.90	2.02	MIB1, GATA6, CABLES1
	2	102.09-102.49	1.95	THOC1, USP14, COLEC12, ROCK1, GREB1L, MIB1
	2	101.69-102.09	1.06	CLUL1, COLEC12, YES1, ADCYAP1, TYMS
	2	101.28-101.68	1.05	–
	6	29.39-29.79	4.39	ATRNL1, GFRA1
	6	28.58-28.98	2.67	ABLIM1, FHIP2A
	6	28.98-29.38	1.89	FHIP2A, ATRNL1
	6	28.18-28.58	1.62	HABP2, NRAP, CASP7, DCLRE1A, NHLRC2
	6	30.20-30.60	1.54	–

SEVOL: semen volume; SECON: semen concentration; SPMOT: sperm motility; SPABR: sperm abnormality; SPCOUNT: total sperm count; Chr: chromosome; gVar: the proportion of genetic variance explained by a 0.4 Mb window.

To reveal potential biological pathways associated with the target traits, GO and KEGG enrichment analyses were performed on 94 genes located within all significant QTL regions. A total of 14 GO terms related to these key candidate genes were identified (Table 3 and Fig. 3). These GO terms were mainly involved in cytosol, protein binding, positive regulation of connective tissue replacement, lamellipodium, mRNA transport, toll-like receptor 3 signaling pathway, stress fiber, ubiquitin-dependent protein catabolic process, pattern recognition receptor activity, cellular response to exogenous dsRNA, cardiac muscle hypertrophy in response to stress, positive regulation of endocytosis, muscle alpha-actinin binding, and cardiac muscle cell differentiation.

Table 3.

GO terms of the significant genes associated with semen traits of RIR chickens (P < 0.05).

Term	Count	P-value	Genes
GO:0005829∼cytosol	22	4.829E-03	USP46, TRMT10A, YES1, ADH1C, RAP1GDS1, MTTP, CABLES1, TTC1, TYMS, NFKB1, NDC80, PPP3CA, ADH4, CASP7, PDLIM3, FABP6, TACC3, LCP2, NHLRC2, LPIN2, SLU7, METAP1
GO:0005515∼protein binding	17	7.107E-03	USP14, MYOM1, ATRNL1, RAP1GDS1, THOC1, TTC1, SORBS2, UNC5C, NFKB1, MIB1, SMCHD1, C1QTNF2, LNX1, LCP2, NHLRC2, KLKB1, TLR3
GO:1905205∼positive regulation of connective tissue replacement	2	8.343E-03	PPP3CA, ROCK1
GO:0030027∼lamellipodium	4	1.701E-02	ABLIM1, PDLIM3, ROCK1, UNC5C
GO:0051028∼mRNA transport	3	1.994E-02	SLBP, THOC1, EIF4E
GO:0034138∼toll-like receptor 3 signaling pathway	2	2.073E-02	COLEC12, TLR3
GO:0001725∼stress fiber	3	3.008E-02	ABLIM1, PDLIM3, MYL12A
GO:0006511∼ubiquitin-dependent protein catabolic process	4	3.027E-02	TMEM129, UBE2D3, LNX1, MIB1
GO:0038187∼pattern recognition receptor activity	2	3.275E-02	COLEC12, TLR3
GO:0071360∼cellular response to exogenous dsRNA	2	3.296E-02	COLEC12, TLR3
GO:0014898∼cardiac muscle hypertrophy in response to stress	2	3.701E-02	PPP3CA, GATA6
GO:0045807∼positive regulation of endocytosis	2	4.505E-02	PPP3CA, MIB1
GO:0051371∼muscle alpha-actinin binding	2	4.873E-02	PDLIM3, NRAP
GO:0055007∼cardiac muscle cell differentiation	2	4.904E-02	GATA6, GREB1L

Fig. 3.

GO terms plot of the significant genes associated with semen traits of RIR chickens.

Within the four shared QTL regions for the SEVOL and SPMOT traits, the SNPs with the largest effects were extracted. These four SNPs were chr4:59901214, chr4:60230489, chr4:60980330, and chr4:6117366, and they exhibited high minor allele frequencies (MAF) and were widely distributed in the population (Table 4). The four SNPs exhibited significant effects on both the SEVOL and SPMOT traits (P < 0.05). The favorable alleles of chr4:59901214, chr4:60230489, chr4:60980330, and chr4:6117366 for the SEVOL trait were T, C, A, and G, respectively, whereas the favorable alleles for the SPMOT trait were C, A, G, and T, respectively. The favorable alleles were exactly opposite between the two traits (Fig. 4).

Table 4.

The SNP with the largest effect within the shared QTL region for semen traits in RIR chickens.

QTL regions(Mb)	Chr	Position	SNP name	SNP effect	Allele	MAF
Chr4:59.83-60.23	4	59901214	rs733115920	1.79E-05	C/T	0.4959
Chr4:60.23-60.63	4	60230489	rs3387813394	1.95E-05	C/A	0.4669
Chr4:60.63-61.03	4	60980330	rs313052484	2.17E-05	A/G	0.4876
Chr4:61.03-61.43	4	61117366	rs3387738824	2.03E-05	G/T	0.4793

Chr: chromosome

Fig. 4.

Phenotypic values of the SNP with the largest effect size within the shared QTL region for SEVOL and SPMOT traits in RIR chickens.

Discussion

Research object of this study

Rooster semen quality traits are core indicators determining reproductive efficiency, breeding progress, and economic benefits in the poultry industry. Accurate evaluation, genetic dissection, and improvement of these traits represent a critical strategy for enhancing poultry productivity and driving innovation in poultry breeding. In this study, we estimated genetic parameters and conducted a Weighted single-step GWAS for five semen traits in RIR chickens, with the aim of revealing the genetic architecture underlying these traits and identifying key candidate genes and QTLs associated with semen quality.

Unlike pure commercial broiler breeds, Rhode Island Red chickens are widely used as parental lines in commercial crossbreeding systems and therefore have considerable research value. This population was introduced from the University of Guelph, Canada in 2018 (Zhang, et al., 2024), and has been maintained as a conservation population consisting of 60 families for a long time, accumulating abundant genetic and phenotypic variations. Therefore, conducting genome-wide association studies (GWAS) on this population under strictly controlled experimental conditions is of great theoretical and practical significance for the precise identification of genetic loci and candidate genes associated with semen traits.

Phenotypes and heritability of semen traits

The results of this study showed that the coefficients of variation for semen traits in roosters were relatively high, ranging from 12.05% to 47.74%. These findings were consistent with the phenotypic variation patterns of semen traits reported in bulls (Abril-Parreño, et al., 2023) and boars (Zhang, et al., 2025), suggesting that such variation may be closely related to the maturation process of sperm cells. Meanwhile, the results also indicated that these traits have considerable potential for genetic improvement and selection. Among these traits, semen volume was similar to that of white-feathered broilers (Zhang, et al., 2023a) and higher than the values reported for Thai native roosters (Radwan, et al., 2025).

Compared with studies on semen traits in pigs and cattle (Aguilar, et al., 2010; Vahedi, et al., 2022), the lack of large-scale and standardized artificial insemination centers in chickens results in greater difficulty in collecting semen phenotypic data. Compared with previous chip‑based studies (Huang, et al., 2025; Vahedi, et al., 2022; Zhu, et al., 2025b), the present study used single nucleotide polymorphism data derived from whole‑genome resequencing, which enabled more uniform genome-wide coverage and improved the accuracy of the analysis. Heritability estimates for semen traits ranged from 0.097 to 0.322, indicating a moderate to low heritability overall. These results were similar to those reported by (Zhang, et al., 2025) for semen traits in pigs, and slightly higher than the estimates reported by (Radwan, et al., 2025) in chickens. Traits with low to moderate heritability indicate that significant genetic progress can still be achieved through genetic selection.

Weighted single-step GWAS

WssGWAS can simultaneously integrate phenotypic, genotypic, and pedigree data, assign different weights to SNPs according to their importance, and utilize consecutive SNP windows to improve the efficiency of QTL region mapping. This study represents the first application of this method for GWAS of semen traits in chickens. In this study, 29 significant QTL regions and 12 candidate genes associated with semen traits were identified by WssGWAS. These QTL regions were located on chromosomes 4 and 13 for semen volume, chromosome 4 for sperm motility, and chromosomes 2 and 6 for sperm abnormality rate. Four prominent consecutive QTL regions were shared for the SEVOL and SPMOT traits. These QTL regions were located within the interval of 59.83–61.43 Mb, explaining 11.55% and 16.33% of the genetic variance for SEVOL and SPMOT, respectively. These results confirm the hypothesis of gene pleiotropy. In contrast to boar semen collected from large‑scale commercial artificial insemination stations, chicken semen samples are difficult to obtain due to the absence of standardized and large‑scale collection systems. Therefore, future studies with enlarged sample sizes and independent populations are needed to further validate the QTLs identified in this study.

Candidate gene associated with semen traits

NFKB1, UBE2D3, PPP3CA, EIF4E, H2AFZ, and DOCK2 were identified as key candidate genes for semen volume. Polymorphisms in the NFKB1 gene can affect protein expression levels, thereby correlating with the risk of reproductive disorders such as oligospermia (Tunçdemir, et al., 2018). Bull sperm transcriptome analysis has shown that UBE2D3 is significantly and negatively correlated with the proportion of progressively motile sperm (Parthipan, et al., 2017). PPP3CA has been identified as an important candidate gene involved in sperm development and sperm motility in pigs (Wang, et al., 2025), chickens (Guo, et al., 2024; Guo, et al., 2023), and sheep (Bai, et al., 2022). The EIF4E protein is localized to the distal region of the elongated spermatid cyst, and its mutants exhibit complete penetration defects during spermiogenesis, resulting in the failure to produce mature sperm (Shao, et al., 2023). H2AFZ mainly functions in regulating chromatin structure and gene transcription and is involved in embryonic development and reproductive processes. It acts as an important epigenetic regulator of semen traits in cattle (Elango, et al., 2023) and pigs (Zhao, et al., 2018). DOCK2 mainly functions to activate Rac GTPase and regulate actin cytoskeleton remodeling, and it has been identified to be associated with semen quality in sheep (Serrano, et al., 2021).

PPP3CA, NFKB1, UBE2D3, MTNR1A, TACC3, EIF4E, and H2AFZ were identified as key candidate genes for semen motility. MTNR1A is a high-affinity G protein-coupled receptor that primarily mediates melatonin signaling. It regulates spermatogenesis, sperm maturation, and capacitation, thereby directly affecting male fertility. MTNR1A has been identified as a key candidate gene for semen traits (sperm motility, concentration, and morphology) and reproductive performance in sheep and goats (Abecia, et al., 2020; Peña-Delgado, et al., 2025). TACC3 is a core regulator of the microtubule cytoskeleton and maintains the stability of meiotic spindle poles during male meiosis through interaction with Aurora‑A kinase (Simerly, et al., 2024).

ADCYAP1, GFRA1, ABLIM1, and CASP7 were identified as key candidate genes for semen abnormality. ADCYAP1 is a neuropeptide with diverse functions. Studies in humans (Brubel, et al., 2012) and mice (Yan, et al., 2020) have shown that its expression affects sperm development and function. GFRA1 is a GPI-anchored cell surface co-receptor involved in the regulation of testicular development and spermatogenesis. ABLIM1 is a key protein with dual functions in cytoskeleton regulation and E3 ubiquitin ligase activity. It is expressed in the testis and during spermatogenesis, where it modulates spermatogenic cell migration and sperm morphogenesis. Polymorphisms in this gene have significant effects on human semen quality (Cerván‐Martín, et al., 2021). The core function of CASP7 is to mediate apoptosis, and its dysfunction can lead to impaired spermatogenesis and male infertility (Espinoza, et al., 2009). For neighboring genes within other significant QTL regions, currently no functional evidence associated with male reproduction has been identified from public databases.

As purebred Rhode Island Red lines serve as the universal paternal stock for all commercial brown-shelled and pink-shelled layer breeds worldwide, the candidate genes and QTLs associated with semen traits identified in this study provide critical insights for the genetic improvement of major commercial layer lines, such as Lohmann, Hy-Line, and Jingfen series. Furthermore, we identified shared QTL regions for semen volume and sperm motility, and genetic correlation analysis revealed a significant negative genetic correlation between these two traits, which reflects an intrinsic biological trade-off in male reproductive resource allocation in chickens. From an evolutionary and physiological perspective, this negative correlation likely arises from limited energy allocation during spermatogenesis: increased semen production diverts energy away from sperm development and motility maintenance, thereby resulting in reduced sperm motility. Such antagonistic pleiotropy creates a critical bottleneck for practical poultry breeding. Single‑trait directional selection for higher semen volume may inadvertently reduce sperm motility, potentially lowering overall fertilization efficiency and resulting in sub‑fertile roosters. A comparable breeding example exists in pigs, where intensive selection for elevated lean meat percentage increases the incidence of leg disorders(Susanne Stern, et al., 1995). Accordingly, independent selection for individual chicken semen traits is not recommended. Instead, a comprehensive multi‑trait selection index integrating semen volume, sperm motility, sperm concentration, and other semen quality indicators should be constructed to balance antagonistic genetic effects and maximize overall male reproductive performance. This limitation may be mitigated via gene‑editing technology in future studies, which can alter linkage disequilibrium between different traits to facilitate simultaneous multi‑trait selection.

Conclusion

In this study, we estimated genetic parameters and conducted weighted single-step GWAS for five semen traits in RIR chickens to explore their genetic architecture and identify key QTLs and candidate genes. Heritability estimates for SEVOL, SECON, SPMOT, SPABR, and SPCOUNT ranged from 0.097 to 0.322, indicating low to moderate heritability. 29 significant QTL regions and 12 candidate genes associated with semen traits were identified by weighted single-step GWAS. Notably, four consecutive QTLs at 59.83–61.43 Mb were shared by SEVOL and SPMOT, explaining 11.55% and 16.33% of the genetic variance, respectively. The largest effect SNPs in these regions significantly affected both semen volume and sperm motility and can be used for marker-assisted selection. This study provides new insights into the genetic basis of chicken semen traits and supports their genetic improvement.

Ethics approval

Not applicable.

CRediT authorship contribution statement

Hailai Hagos Tesfay performed Investigation, Conceptualisation, Data curation, Formal analysis, and Writing – review & editing. Xiaoke Zhang was responsible for Methodology, Formal analysis, Validation, Visualisation, Validation, Writing – original draft, and Writing – review & editing. Yunlei Li provided Resources and contributed to Conceptualisation, Writing – review & editing, and Validation. Yunhe Zong, Zhong Ma, Xintong Han, Yi Zhao, and Adamu Isa Mani undertook Investigation and Data curation. Ziting Pan contributed to Validation. Fujian Yang, Zongyao Zhang, Jiming Chen, and Yanyan Sun acquired Funding acquisition. Jilan Chen was responsible for Conceptualisation, Writing – review & editing, Supervision, and Funding acquisition.The determination of semen quality traits requires substantial manpower input to ensure all quality parameter measurements are completed within 15 minutes after semen collection from the poultry house. Yunhe Zong, Zhong Ma, Xintong Han, Yi Zhao, and Adamu Isa Mani played a pivotal role in this assessment procedure; Fujian Yang, Zongyao Zhang, Jiming Chen, and Yanyan Sun provided essential financial support for this study, which ensured the smooth implementation of the project. Hailai Hagos Tesfay, Xiaoke Zhang, Yunlei Li, Ziting Pan, and Jilan Chen provided constructive guidance and suggestions for the conduct of this study and the preparation of the manuscript. All co-authors of the present study have played indispensable roles in this research. Hailai Hagos Tesfay: Investigation, Conceptualisation, Data curation, Formal analysis, Writing – review & editing. Xiaoke Zhang: Methodology, Formal analysis, Validation, Visualisation, Validation, Writing – original draft, Writing – review & editing. Yunlei Li: Resources, Conceptualisation, Writing – review & editing, Validation. Yunhe Zong: Investigation, Data curation. Zhong Ma: Investigation, Data curation. Xintong Han: Investigation, Data curation. Yi Zhao: Investigation, Data curation. Ziting Pan: Validation. Adamu Isa Mani: Data curation. Fujian Yang: Funding acquisition. Zongyao Zhang: Funding acquisition. Jiming Chen: Funding acquisition. Yanyan Sun: Funding acquisition. Jilan Chen: Conceptualisation, Writing – review & editing, Supervision, Funding acquisition.

Disclosures

The authors have declared no conflict of interest.