Integrative multi-omics analysis deciphers the regulatory mechanisms of egg weight traits in chickens

Abstract

Egg weight is an economically important trait in the chicken, and affects the hatchability and chicks’ performance in broiler breeding programs. Our comprehensive analysis of 22,375 chickens revealed that the hens’ egg weight was linked to their body weight, egg production and hatchability, with higher egg weight potentially increasing the body weight and delaying the female sexual maturity. Egg weight is a dynamic trait, however, previous studies usually focused on single time point and overlooked the dynamic changes during egg-laying period. We performed both single and longitudinal genome-wide association studies in 2,350 hens, combining selective sweep analysis, to identify genetic variants. Then, we integrated multi-omics data of 40 chickens to determine key genes and metabolites. A multi-omics analysis identified 22 key candidate genes, such as ATF6, CSPG4, SH3GL3, C4, LMX1B, CDC34, and CCDC171, of which four (BSG, CFD, MAP2K2, and POLRMT) were associated with egg weights in the ChickenGTEx database. In particular, the SNP rs315726522 may regulate MAP2K2 and FSHB gene expression by modulating the binding of transcription factors, and the SNP rs738839430 caused amino acid change that affected function of CFD protein. This, in turn, affected gonadotropin expression within the GnRH signaling pathway, which ultimately influenced egg weights. Metabolomic analysis revealed 13 metabolites associated with oxidative stress and metabolism of fatty acids, which potentially influenced reproductive performance through stress reduction and hormonal regulation. This study comprehensively analyzed the effects of egg weight in broiler breeders and enhanced our understanding of the genetic mechanisms underlying egg weight in chickens.

Introduction

Chicken meat serves as a major source of high-quality protein for humans. In recent years, extensive research on the growth characteristics of meat pure lines has provided valuable references for improving production efficiency. However, selection for reproductive performance in meat pure lines is still inadequate, although it is closely related to broiler productivity.

Egg weight trait is an essential economic indicator in broiler breeding programs. It affects the important economic performance of two generations simultaneously. For example, it influences the reproductive performance of hens, and it is affected by the body weight of hens, and in turn, it affects the subsequent chick growth (Jiang and Yang, 2007). Egg weight is significantly correlated with the health, hatch weight, and uniformity of chicks after hatching (Ma et al., 2024). A study reported that medium-weight eggs exhibited better hatchability and produced chicks with superior performance regarding body weights, carcass yields and stress tolerance (Duman and Şekeroğlu, 2017; Javid, et al., 2017). Consistent with these findings, eggs laid by 32-week-old breeder flocks exhibited markedly better quality parameters than those from 60-week-old hens, consequently yielding improved embryonic development outcomes and higher hatchability rates (Alo, et al., 2024). Egg weight increases along with the age of the hen during the laying period, and both smaller egg weight at onset of laying period, and excessive egg weight in later stages are undesirable (Minvielle, et al., 1994). Genetic control methods are effective in addressing these issues, and exploring the genetic mechanisms is a practical approach for managing egg weight in chicken production.

Multiple studies have demonstrated that egg weight exhibits consistently high heritability estimates. Comprehensive genetic evaluations indicate that the heritability of egg weight ranges from 0.40 to 0.44 (Liu, et al., 2018b; Ma, et al., 2024). This robust heritability confirms that egg weight is primarily determined by genetic factors. Given these genetic characteristics, investigating the regulatory mechanisms of egg weight through genetic approaches is highly viable and can provide clear direction for genetic improvement in poultry breeding programs. Previously, researchers explored single nucleotide polymorphisms (SNPs) in potential candidate genes that were associated with egg weight at the single gene level. For example, a study involving three hundred and nine Erlang Mountainous chicken found that the PRLR gene polymorphisms are associated with egg production traits such as the age at first egg and egg weight (Zhang, et al., 2012). The advancement of sequencing technologies has resulted in a significant surge in research activities in the domain of egg weight. As of now, 393 quantitative trait loci (QTLs) have been identified, which are distributed across 22 autosomes and Z chromosome (Lyu, et al., 2017; Wang, et al., 2023). However, these studies were all based on single time point and did not account for variations in egg weight across different laying stages. Consequently, these findings are insufficient to completely explain the genetic mechanisms underlying egg weight throughout the entire laying period in hens.

In this study, we specifically focused on the dynamic nature of egg weight as a longitudinal trait that varies with the age of hens. Along with conventional genome-wide association study (GWAS) tools such as GEMMA that analyze specific time points along the trajectory (Ning, et al., 2019; Zhang, et al., 2021), we also applied longitudinal trait GWAS methodology (Teng, et al., 2023). This approach integrates data from multiple points to effectively identify SNPs associated with trait developmental trajectories, enabling the simultaneous detection of both age-dependent loci and consistent effect loci (Teng, et al., 2023). This methodology thoroughly explores the dynamic characteristics of longitudinal traits.

In addition, we conducted a comprehensive analysis of egg weight effects using a well-documented population of 22,375 chickens, systematically evaluating the impact of egg weight on traits including body weight, egg production performance, and fertility. Afterwards, we applied GWAS and selective sweep analysis to 2,350 hens, identified key regulatory genes and explored the genetic mechanisms underlying egg weight traits. Furthermore, we biologically validated the identified genetic regulatory mechanisms through tissue gene expression profiling and serum metabolomics data from 40 hens. In summary, through large-scale data collection, this study revealed the influence of egg weight on other important production traits. By integrating multi-omics analyses, we elucidated the molecular mechanisms affecting egg weight, providing a theoretical foundation for improving egg quality and uniformity in breeding chickens.

Materials and methods

Ethics approval

All procedures were conducted in accordance with the Chinese laws on animal experimentation and were approved by the Institutional Animal Welfare and Ethics Committee of Nanjing Agricultural University, Nanjing, China [certification no: SYXK (Su) 2022-0031], and conducted under the authority of the Project License.

Resource population and phenotypic collection

A total of 22,375 chickens of the White Rock pure line were used as the experimental population. Among them, 2,350 hens were the previous generation of the remaining chickens. The egg weight at four periods were recorded, namely the weight of the first egg (FEW), at 31 wk of age (EW31), at 37 wk of age (EW37), and at 52 wk of age (EW52). The egg number was recorded until the hens were 440 days old. Moreover, phenotypes with important economic values such as the age at first egg (AFE), the fertilization rate of eggs at 37-39 wk of age, and the hatching rate of fertilized eggs were also recorded.

Among the other 20,025 chickens, there are 9,713 descendant roosters and 10,312 hens. They are the offspring of 986 hens among the above mentioned 2,350 chickens. The body weight of the roosters at 35 days of age was recorded. For the hens, the body weight at 35 days of age (BWD35), the body weight at 49 days of age (BWD49), the weight gain from 35 to 49 days of age (BWD35-49), and the age at first egg were recorded.

The experimental subjects were fed and managed in accordance with the breeding company's standard breeding and management protocols for broilers. At 140 d of age, they were transferred to individual cages for egg-laying. In addition to the FEW, five eggs were collected and the average egg weight at four specific time points during egg laying was calculated. A total of 2,350 blood samples were collected from the brachial veins of each hen for second-generation DNA sequencing. Forty hens at 66 weeks of age, including 20 high egg weight hens (HEW) and 20 low egg weight hens (LEW), were selected. Serum samples were collected from blood samples of brachial veins and separated for metabolomics analysis. Following euthanasia of these hens, hypothalamus, pituitary, ovary, and liver samples were collected, totaling 160 samples, for RNA sequencing analysis.

Genotyping and quality control

Genomic DNA was extracted from whole blood samples using the E.Z.N.A.® NRBC Blood DNA Kit (Omega Bio-tek, Norcross, GA). A Tn5-based protocol was used to prepare sequencing libraries. The libraries were sequenced on the Illumina platform (PE 150 model, two libraries). Each bird was sequenced at a mean depth of 1.0 ± 0.14 × . We used BaseVar(v1.01) (Liu, et al., 2018a) for identifying polymorphic sites and to infer allelic frequencies, and STITCH(v 2.0) for the imputation of SNPs (Yang, et al., 2021), as our primary algorithms. A total of 7,916,553 SNPs were identified from the autosomes and from chromosome Z. The SNP calling and imputation processes were both based on the chicken genome (bGalGal1.mat.broiler.GRCg7b). Furthermore, we used PLINK(v1.9) (Purcell, et al., 2007) to perform quality control. We retained SNPs with a minor allelic frequency (MAF) of > 5 %, removed SNPs with call rate of < 90 %, and excluded SNPs with a genotype-missing rate that was > 10 %. After applying these specified filters, a total of 6,920,502 SNPs were retained for further analysis across chromosomes 1-39 and Z.

Genetic parameters estimate

Variance and heritability components were estimated using ASREML (v4.1). The GBLUP model was utilized in this study, as follows:

$$y=Xb+Z\alpha +e$$

where y is a vector of phenotypic values, X is the design matrix that associates observations with the appropriate combination of fixed effects, Z is the design matrix that associates observations with the appropriate combination of random effects, with generation and batch added to the model as fixed effects in this study, b is a vector of fixed effects, α is a vector of random effects, and e is the vector of residual errors.

Single-trait genome‑wide association study

Univariate linear mixed models (LMM) were used to perform a GWAS analysis using GMMMA (v0.98.1) (Ding, et al., 2022; Zhou and Stephens, 2012), and the model was as follows:

$$y=W\alpha +X\beta +u+ϵ;u\sim MV{N}_{\mathrm{n}}\left(0,\lambda {\tau }^{-1}K\right),ϵ\sim MV{N}_{\mathrm{n}}\left(0,{\tau }^{-1}{\mathrm{I}}_n\right)$$

where y represents the phenotype vector, W is the indicator matrix of the fixed effects, with the first column being 1 and subsequent columns being the values of fixed effects or covariates, and α is the coefficient vector of the fixed effects. x represents the genotype vector, β is the SNP effect, u is the random effect vector, and ϵ is the residual vector. MVNn denotes the n-dimensional multivariate normal distribution, λ is the ratio of two variance components, τ⁻¹ is the residual variance, K is the kinship matrix based on SNPs, and In is the unit matrix.

We acknowledge that the Bonferroni correction can be too conservative for the large number of linkage disequilibrium regions among genetic markers in the genome (Johnson, et al., 2010; Tan, et al., 2023), so we employed the PLINK’s -indep-pairwise command to calculate the number of independent SNPs on all chromosomes. To be specific, we set the following parameters: 100 SNPs were used as a window, 20 SNPs were used as each move step, and 0.2 was used as the r² threshold. We adjusted genome-wide significance thresholds and genome-wide potential significance thresholds according to the number of independent SNPs (Liu, et al., 2019). A total of 74,362 SNPs passed the independence test, with a genome-wide significance threshold of 0.05/74,362 and a potential significance threshold of 1/74,362.

Longitudinal trait genome-wide association study

We performed a longitudinal genomic multivariate analysis of egg weight traits at four age points using the Genomic Multivariate Analysis Tool (GMAT) (Ning, et al., 2019).

The matrix form of the model was:

$$y=Xf+Wb+M\alpha +Qa+Zp+\mathrm{e}$$

$$aN\left(0,G\otimes Ka\right),pN\left(0,I\otimes Kp\right),eN\left(0,I\sigma 2e\right)$$

where y is the vector of test-day records, f is the vector of herd-test-day effects, X is the design matrix for f, b and α are vectors of fixed regression coefficients for individuals and the analyzed SNP, respectively. a and p are vectors of random regression coefficients for additive genetic effects and permanent environmental effects, respectively. W, M, Q, and Z are the corresponding co-variable matrices for b, α, a, and p, respectively W, Q, and Z consist of Legendre polynomial values, M consists of products of genotype codes and Legendre polynomial values, e is the vector of random residuals, G is the genomic relationship matrix, I is an identity matrix, ⊗ denotes the Kronecker product, Ka and Kp are variance-covariance matrices for random regression coefficients of additive polygenic effects and permanent environmental effects, respectively, and σ2e is the residual variance.

According to the results of Al-Mamun et al. (Al-Mamun, et al., 2015), the calculation formula of SNP that explains the genetic variation of traits is as follows:

$$\%V_{gi}=\frac{2p_i{q}_i{\beta }_i^2}{{\sigma }_g^2}\times 100$$

In the formula, pi and qi represent the allelic frequencies of the i^th SNP, which were calculated by GEMMA software. βi represents the effect size of the i^th SNP, which was estimated by GEMMA software. The genetic variance was estimated using the genomic relatedness matrix, calculated by ASReml (v4.1).

Gene-based association analysis

Gene-based association analysis was conducted using MAGMA (v1.10) (De Leeuw, et al., 2015). SNPs were annotated according to each gene's genomic position, which included 3 kb upstream and 1.5 kb downstream of these genes. We annotated 26,217 genes with SNPs. To improve the statistical power and sensitivity of the traits being studied, we used two models for the association analysis: snp-wise = mean and snp-wise = top. The Bonferroni correction was applied to adjust for multiple comparisons, which resulted in an adjusted significance threshold of 0.05/26,217.

Selective sweep analysis

Using egg weight phenotype, top 50 high egg weight individuals (HEW50) and bottom 50 low egg weight individuals (LEW50) in each group were selected for further examination at different time points. We used Vcftools version 0.10.1 to generate a comprehensive list of candidate genes within the target regions by calculating the nucleotide polymorphism (π) and population differentiation index (Fst) values of these individuals, using a window size of 40 kb and a step size of 10 kb (Danecek, et al., 2011).

Functional annotation

The evaluation of genetic characteristics was conducted using PopldDecay (Zhang, et al., 2019) and Haploview software version 4.2 (Barrett, et al., 2005). To elucidate the biological function of significant SNPs, we conducted functional annotation on the significant SNPs and sought to identify candidate genes by querying the reference genome GalGal1.mat.broiler.GRCg7b assembly with the Variant Effect Predictor (VEP) and Biomart tools, both of which were supported by the Ensembl database. Concurrently, enrichment analysis using Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) was performed to identify the pathways associated with the candidate genes.

RNA sequencing and data analysis

RNA sequencing was performed on four tissues (hypothalamus, pituitary, ovary, and liver) in 40 hens that were categorized into two groups based on EW52, with 20 in the high egg weight group (HEW) and 20 in the low egg weight group (LEW). A total of 160 samples were prepared, and all samples were sequenced on the DNBSEQ-T7 platform, which resulted in paired-end reads of 100 bp. The raw reads underwent processing through fastp software, which ensured that low-quality reads were removed to obtain clean reads for subsequent data analysis. HISAT2 software facilitated the alignment of data to the reference genome, and the gene expression levels were computed as Fragments Per Kilobase of transcript per Million mapped reads (FPKM). The read counts for each gene were obtained using HTSeq-count.

DESeq2 software was used to analyze the differentially expressed genes (DEGs) in HEW and LEW. DEGs were identified based on the following criteria: FPKM ≥ 1, |log₂(fold change) | ≥ 1, and P < 0.05.

Serum metabolite mass spectrometry

Serum samples were collected from the 40 individuals for subsequent metabolite analysis, which were grouped in the same way as for RNA sequencing. Ultra-high-performance liquid chromatography-tandem mass spectrometry (UHPLC-MS/MS) analysis was conducted by Novogene Co., Ltd. (Beijing, China) using a Vanquish UHPLC system (Thermo Fisher) coupled with an Orbitrap Q Exactive HF or Orbitrap Q Exactive HF-X mass spectrometer (Thermo Fisher). The raw data files generated by UHPLC-MS/MS were analyzed using Compound Discoverer 3.3 (CD3.3, ThermoFisher) for peak alignment, peak picking, and quantification for each metabolite. To enable comparison among samples with varying distributions, the data were standardized using the following formula: sample raw quantitation value / (the sum of sample metabolite quantitation value / the sum of quality control sample metabolite quantitation value). Compounds with a coefficient of variation in relative peak areas for quality control samples that exceeded 30 % were excluded. Subsequently, the metabolites were identified and quantified(Wang, et al., 2024; Zheng, et al., 2024).

DESeq2 software was used to determine the differences in serum metabolite composition between HEW and LEW groups. The identification of differentially expressed metabolites (DEMs) was conducted following the established criteria of |log₂(fold change) | ≥ 1 and P < 0.05.

Weighted gene co-expression network analysis

A gene co-expression network was constructed using the WGCNA package in R to identify genes and metabolites that influenced egg weight across four tissue transcriptomes and serum metabolome. The network was constructed using the topological overlap matrix, followed by module detection and a module emergence function. The minimum requirement for a module was 30 genes or metabolites. The relationship between the identified modules and the phenotypic traits was analyzed using FEW, EW31, EW37, and EW52. A significance threshold of P < 0.05 was applied to determine the significant correlation between modules and traits. The genes within these modules are identified as module hub genes (MHGs).

Prediction of transcription factor binding sites

We used the online website for transcription factor binding sites (https://alggen.lsi.upc.es/cgi-bin/promo_v3/promo/promoinit.cgi?dirDB=TF_8.3) to predict the impact of candidate loci mutations on transcription factor binding.

Analysis of sequence conservation

Sequence conservation was assessed using PhastCons scores based on 77 vertebrate species. The analysis was performed using the SnpSift (v5.0) software. PhastCons scores were retrieved from the UCSC database (http://hgdownload.soe.ucsc.edu/goldenPath/galGal6/phastCons77way/galGal6.77way.phastCons/).

Combined analysis of transcriptomes and metabolomes

A joint analysis was conducted utilizing the Orthogonal Projections to Latent Structures (O2PLS) method to investigate the interrelationship between transcriptome and metabolome data. Both the transcriptome and metabolome data were preprocessed using UV scaling, followed by the calculation of the load value for each gene and metabolite. The genes and metabolites with high load values were crucial for establishing similarity between the two datasets. We focused on the 25 genes or metabolites that exhibited the highest load value.

Results

The influence of egg weight on important economic traits

The traits including growth, egg production, egg weight and hatching rate from a total of 22,375 chickens were recorded. The egg weight of 2,350 hens at four periods (FEW, EW31, EW37, and EW52) were recorded (Fig. 1A, Table 1). The average age of hens when they laid their first egg was 165.3 days (Fig. 1B). The distribution of egg weight data conforms to a normal distribution, and the egg weight gradually increases along with hens' age (Fig. 1D). The average FEW was 47.76 g, while the average EW52 is 61.78 g. Throughout all periods, the coefficients of variation were below 6.5 %. The coefficient of variation of FEW was the highest, and it showed a slight downward trend with the growth of the hens' age.

Fig. 1.

Analysis of the egg weight phenotypic data of hens at four stages. (A) Weigh 5 eggs per hen at each period. (B) Distribution of the age at first egg. (C) PCA analysis of resequencing data of 2,350 hens. (D) The data distribution of egg weight in four periods. FEW, EW31, EW37, and EW52 were the egg weight of the first egg, 31 wk of age, 37 wk of age, and 52 wk of age, respectively. (E) Analysis of the relationship between egg weight and body weight. BW24, BW37, and BW52 refer to the body weight at 24, 37, and 52 wk of age, respectively. (F) The correlation between egg weight, chick body weight and the age at first egg of chicks. BWD35, and BWD49 the body weight at 35, and 49 days of age, respectively. BWD35-49 represents the weight gain from 35 to 49 days of age. AFE stands for the age at first egg.

Table 1.

Descriptive statistics of the mean egg weight by 2,350 hens at four ages from the first egg laid until 52 weeks of age.

Traits	Mean (g)	Min (g)	Max (g)	N	SD	CV
FEW	47.76	34.20	62.30	2312	3.07	6.4 %
EW31	52.16	39.92	62.50	2324	2.98	5.7 %
EW37	57.67	44.37	69.80	2311	3.24	5.6 %
EW52	61.78	49.85	73.96	1245	3.34	5.4 %

Note: FEW, the weight of first egg laid; EW31, EW37, and EW52 represent the egg weight at 31, 37, and 52 wk of age, respectively; N, Number of hens; SD, standard deviation; CV, coefficient of variation.

Egg weight positively correlated with hens’ body weight (Fig. 1E). Egg weight EW37 positively correlated with hens’ AFE, and negatively correlated with egg number (Figure S1A). Egg weight negatively correlated with the fertilization and hatching rates (Figure S1B).

We evaluated the influence of the breeder egg weight with 20,025 subsequent chicks including 9,713 males and 10,312 females. Chicks hatched from heavier eggs tended to have higher body weights at both 35 and 49 days of age (Fig. 1F). Moreover, the heavier eggs positively correlated the weight gain from 35 to 49 days of age. Chicks hatched from heavier eggs had a later age at first egg and will also lay heavier eggs.

Estimation of genetic parameters

A total of 2,350 hens were sequenced to identify genetic variation linked to egg weight traits. After quality control, a total of 6,920,502 SNPs were retained that were distributed across chromosomes 1 - 39 and Z. Genomic PCA analysis showed that the population was not stratified (Fig. 1C), so PCA was not added as a covariate in the GWAS.

The heritability of egg weights was estimated to be 0.458 - 0.571 using SNP data. FEW, which is more affected by the environment, had the highest coefficient of variation and the lowest heritability. The genetic correlation of egg weight was consistently high throughout all stages (0.829 - 0.984) (Table 2).

Table 2.

Heritability, genetic and phenotypic correlations of egg weight at four stages.

Traits	FEW	EW31	EW37	EW52
FEW	0.458 (0.031)	0.944 (0.015)	0.829 (0.027)	0.869 (0.032)
EW31	0.729 (0.012)	0.491 (0.031)	0.942 (0.012)	0.904 (0.023)
EW37	0.669 (0.015)	0.819 (0.009)	0.571 (0.034)	0.984 (0.011)
EW52	0.671 (0.019)	0.773 (0.014)	0.844 (0.010)	0.499 (0.056)

Note: Bold type is heritability. The upper and lower triangles are genetic and phenotypic correlations, respectively. Standard errors (SE) are reported in parentheses.

Potential genetic variants and genes associated with egg weight

A total of 609 significant loci and 2,020 potentially significant loci were identified through single-trait GWAS (Fig. 2A-D, Tables S1-4). The genome inflation factor varied between 0.934 and 0.992, which indicated that the results were reliable and that the level of statistical significance was not amplified by systematic errors.

Fig. 2.

Manhattan and quantile-quantile (Q-Q) plots for genome-wide association analysis and selective sweep analysis. (A-D) Manhattan plot and quantile-quantile (Q-Q) plots were used to analyze FEW, EW31, EW37, and EW52. (E) Longitudinal GWAS of egg weight in four periods. The linear mixed model was used for correlation analysis. The significance was corrected by a relaxed correction for multiple comparisons and was self-defined as the genome-wide significance threshold (P = 0.05/74,362; −log₁₀ (P) = 6.17; the horizontal solid line) or the potential significance threshold (−log₁₀ (P) = 4.87; the horizontal dashed line). (F) Manhattan plot of selective sweep analysis. The horizontal solid line is the threshold line 0.1, and the horizontal dashed line is the threshold line 0.2.

For FEW, a total of 12 significant SNPs were identified on chromosomes 10 and 28, and 264 potentially significant SNPs were detected on chromosomes 10, 17, and 28. For EW31, a total of 565 significant SNPs were identified on chromosome 28, along with 742 potentially significant SNPs on chromosomes 8, 9, 16, 17, 23, 28, and Z. For EW37, 32 significant SNPs were identified on chromosomes 8, 16, and 28, in addition to 809 potentially significant SNPs. For EW52, 167 potentially significant SNPs were found on chromosomes 10 and 33. We used longitudinal GWAS to identify egg weights related loci in multiple periods. A total of 425 significant loci, 2,115 potentially significant loci, and 111 candidate genes were identified (Fig. 2E, Table S5).

Significant regions on chromosome 8 were found for EW31, EW37, and for long-trait, and the most significant SNP rs15899483 was located in intron of the ATF6 gene. Two significant regions were found on chromosome 10 for FEW, EW52, and long-trait. The most significant SNPs were rs312408820 that was upstream of the CSPG4 gene, and rs740320811 that was located in intron of the SH3GL3 gene. There were two significant regions on chromosomes 16 and 17. The most significant SNPs were rs732098243 that was upstream of the C4 gene, and rs739173234 that was located in intron of the LMX1B gene; both SNPs were associated significantly with EW31, EW37, and long-trait. One significant region was found on chromosome 28, which was significant for all traits except EW52. The most significant SNP rs738361772 was located in intron of the CDC34 gene. A region was found on chromosome Z that was associated significantly with both EW37 and long-trait, with the most significant SNP rs740565056 located in intron of the CCDC171 gene.

The selective sweep analysis (Fst) was conducted using HEW50 and LEW50, with the top 1 % SNPs identified as significant selection areas. Of these, 956 genes were identified as potentially selected genes (Fig. 2F, Table S6). A total of 82 genes were annotated jointly in the context of longitudinal GWAS and selective sweep analysis, and regarded as candidate genes that affected egg weight.

To identify genes related to egg weight further, two complementary models of gene association analysis were used to screen 31 genes that were associated significantly with egg weight (P < 1.91E-6) (Fig. 3A, Tables S7-8). Among these, 22 genes (e.g., C2CD4C, FSTL3, HCN2, CDC34, MAP2K2, CFD, BSG, and POLRMT) overlapping with the 82 genes, were selected as key candidate genes (KCGs) (Fig. 3B). Their impacts on egg weight traits were analyzed using the ChickenGTEx database. Four genes, BSG, CFD, MAP2K2, and POLRMT, were associated with egg weight or related traits in the TWAS results (Fig. 3C). Regional association analysis indicated that genomic regions that contained these four genes exhibited significant signals in both GWAS and selective sweep analysis (Fig. 3D).

Fig. 3.

Screening of genes related to egg weight. (A) Venn diagram of significant gene distribution screened by two models of gene association analysis. (B) The distribution among gene association analysis, longitudinal GWAS, and the annotated genes in the top 1 % SNP of the selective sweep analysis. (C) The key candidate genes (KCGs) on TWAS results in the ChickenGTEx database. EW: egg weight, ESW: eggshell weight, NSYF: number of small yellow follicles, SL: Shank length, SW: Shank weight. (D) The location of four KCGs related to egg weight on the genome, and the selective sweep analysis. (E) KEGG pathway of annotated genes at significant SNPs in longitudinal GWAS of egg weight in four periods. (F) KEGG pathway of annotated genes of the top 1 % SNPs of selective sweep analysis. (G) KEGG pathways of 31 genes screened by gene association analysis. (H) KEGG pathways of 22 KCGs.

KEGG analysis was conducted on four gene sets: longitudinal GWAS candidate genes, the annotated genes from selective sweep analysis, significant genes for gene association analysis, and 22 KCGs (Fig. 3E-H, Tables S9-12). Notably, the GnRH signaling pathway emerged as significant with MAP2K2 playing a role in regulating gonadotropin expression.

Gene expression influenced egg weight

To elucidate the multi-tissue regulatory molecular network that affected egg weight, transcriptomic analysis was conducted on the hypothalamus, pituitary, ovary, and liver tissues of high egg weight group (HEW) and low egg weight group (LEW). A total of 160 mRNA libraries were constructed, and 113.92 Gb of clean data were obtained after sequencing. The effective data volume of each sample ranged from 6.86 Gb to 7.06 Gb, with Q30 bases that ranged from 93.69 % to 96.47 %. Additionally, the average GC content was 48.72 %. In total, 17,057 genes were identified as expressed across the four tissues. Principal component analysis distinguished the four tissues effectively, although it did not distinguish fully between HEW and LEW (Fig. 4A-B).

Fig. 4.

Multi-tissue transcriptome analysis to identify candidate genes that regulate egg weight. (A) PCA of transcriptomes of four tissues. (B) PCA of hypothalamus, pituitary, ovary, and liver in HEW and LEW, respectively. (C) WGCNA analysis of transcriptomes of hypothalamus, pituitary, ovary, and liver, respectively; The numbers in each small rectangle represent the correlation between the module and the trait. An asterisk (*) indicates P < 0.05, and two asterisks (**) indicate P < 0.01. (D) The differentially expressed genes (DEGs) of hypothalamus, pituitary, ovary, and liver in HEW and LEW groups. (E-F) Venn diagrams of DEGs, key candidate genes (KCGs), and module hub genes (MHGs), respectively. pHMHGs, pPMHGs, and pOMHGs are positively correlated hub genes in pituitary, ovary, and liver transcriptome modules, respectively, and n is negatively correlated.

The hub genes that were associated significantly with egg weight were identified through WGCNA (Fig. 4C, Table S13-16). Eight modules showed significant correlation with egg weights (P < 0.05), which included 1,849 module hub genes (MHGs). Among these, three modules showed negative correlation, and five modules showed positive correlation. 1,186 genes showed significant positive correlation (pMHGs), and 663 genes showed significant negative correlation (nMHGs).

Transcriptome analysis of the hypothalamus did not reveal modules associated with egg weight. In the pituitary, 297 pMHGs and 326 nMHGs were identified, and the ovary contained 636 pMHGs and 337 nMHGs. In the liver, 252 pMHGs were identified. Furthermore, 26, 25, 50, and 31 DEGs were identified in the hypothalamus, pituitary, ovary, and liver, respectively (Fig. 4D, Table S17-20). Five genes (i.e., ENSGALG00010024111, ENSGALG00010003257, ENSGALG00010024056, MAP2K2, and CFD) in the pituitary were identified as both DEGs and MHGs (Fig. 4E). Two KCGs (i.e., MAP2K2 and CFD) were identified in the pituitary, and referred as MHGs (Fig. 4F).

These results were consistent with GWAS analysis, selective sweep analysis, and ChickenGTEx data analysis, and together demonstrated that MAP2K2 and CFD were important candidate genes (Fig. 3C-D). Six genes (i.e., ENSGALG00010021470, ENSGALG00010007182, ENSGALG00010010295, AQP10, CDO1, and TAGLN) in the ovary were identified as both DEGs and MHGs. Three KCGs (i.e., RNF126, CDC34, and RBM18) were identified in the ovary, and referred as MHGs. One gene (i.e., WDR86) in the liver was identified as both DEG and MHG (Fig. 4E).

Serum metabolites influenced egg weight

The serum metabolome identified 13 key metabolites, of which seven were down-regulated and six were up-regulated (Fig. 5A, Table S21). The WGCNA analysis revealed one module that were associated significantly with egg weight (P < 0.05), which included 112 module hub metabolites (Fig. 5B, Table S22).

Fig. 5.

Analysis of candidate metabolites and candidate SNPs that regulate egg weight. (A) Volcano map for differentially expressed metabolites (DEMs). (B) WGCNA analysis of serum metabolome and egg weight. (C) The correlation analysis of the pituitary, ovary, and liver with serum metabolome, which show the 25 genes or metabolites with the highest contribution. (D) Network analysis between differentially expressed genes (DEGs) and key candidate genes (KCGs) overlapped with module hub genes (MHGs) in the pituitary. The color of the node represents the number of genes/metabolites associated with it, blue represents a small number and red represents a large number; the node size represents the abundance of the gene/metabolite; the line color represents the positive or negative correlation, blue represents a negative correlation and red represents a positive correlation; the line thickness represents the strength of the correlation, the thicker the line, the stronger the correlation. (E) A network analysis was conducted between 13 DEMs and five genes in the pituitary. The full name of the metabolite that corresponds to each abbreviation can be found in Table S21. (F) Analysis of the effects of key candidate variants on transcription factor binding regions in MAP2K2 gene. (G) MAP2K2 gene expression in the pituitary of individuals with different genotypes of key candidate SNP. (H) PhastCons scores of candidate SNPs. (I) CFD gene expression in the pituitary of individuals with different genotypes of key candidate SNP.

Association analysis was performed between four tissue transcriptomes and serum metabolites using O2PLS analysis, and showed 25 genes and 25 metabolites with the highest priority across each tissue (Figs. 5C and S2A, Tables S23-30). Subsequently, KEGG analysis was conducted on the 25 genes of each tissue. 33 KEGG pathways, most of which were related to metabolism, were identified as enriched and might affect serum metabolite levels (Figure S2B-E, Tables S31-34).

Regulatory network analysis

Gene regulatory network and gene-metabolite regulatory network were constructed to analyze the correlation between specific genes and metabolites. The regulatory network analysis involved five genes where pituitary DEGs overlapped with hub genes (Fig. 5D). There was a strong positive correlation between MAP2K2 and CFD, and were positive correlations between the other three novel genes. Additionally, regulatory networks of the five genes and 13 differential metabolites were analyzed further (Fig. 5E). Among these metabolites, there was a weak negative correlation between two metabolites (HexCer (d19:1;2O/9:0) and QA), which were not correlated with any other metabolite and gene. In addition, five genes had varying degrees of correlation with 11 important metabolites. These identified genes and metabolites interacted with each other to act as co-regulators of egg weight.

In the ovary, the regulatory network analysis involved nine genes: six with overlapping DEGs and hub genes, and three with overlapping KCGs and hub genes (Figure S3A). There were correlations between CDO1 and five genes, and strong negative correlation between CDC34 and ENSGALG00010021470. In addition, the regulatory networks of nine genes and 13 differential metabolites were analyzed further, and five important metabolites were connected with theses candidate genes (Figure S3B). In the liver, regulatory network analysis showed there was no connection between the only candidate gene and metabolites (Figure S3C).

Analysis of key candidate SNPs

We evaluated the effects of 22 SNPs for MAP2K2 and CFD genes. We looked at the expression of the two genes in four tissues. MAP2K2 was highly expressed in the hypothalamus, pituitary, and ovary, and CFD was highly expressed in the pituitary (Figure S4A-B). For the identified SNP rs315726522 in the intron 8 of MAP2K2, the C allele causes increased binding of NF-1 (Fig. 5F), thereby influences MAP2K2 expression (Fig. 5G). The expression of gonadotropin-related genes in the pituitary and ovary was analyzed in individuals with different genotypes of SNP rs315726522 (Figure S4C-D). The expression of FSHB in the pituitary of individuals with TT genotype was higher than that with CC genotype, although the expression of FSHR in the ovary in TT individuals was lower than CC individuals. A missense mutation (rs738839430) with a high PhastCons scores was identified in the CFD gene, which led to the conversion of threonine to methionine, and didn’t affect CFD gene expression (Fig. 5H-I). These two SNPs (rs315726522 and rs738839430) were significantly associated with egg weights at all four stages (Figure S5A-B). These results indicated that these SNPs could regulate the gene expression and influence the egg weights.

Discussion

All the egg weight, body weight, and reproductive traits from 22,375 chickens were utilized to comprehensively evaluate the impacts of egg weight. Hens’ egg weight is not only related to their own body weight but also affects egg number. More importantly, it influences the hatching rate of fertilized eggs. This correlation is based partly on a polygenic effect (Liu, et al., 2013; Ma, et al., 2024). Chicks hatched from heavier eggs exhibit greater hatching weights and faster growth rates. Typically, chick weight constitutes 62 % to 78 % of egg weight, which suggests that heavier eggs produce heavier chicks (Akil and Zakaria, 2015). In addition, egg weight influences the growth rate of chicks after hatching, and chicks hatched from eggs with heavier egg tended to exhibit faster growth rate. Specifically, each 1 g increase in egg weight corresponded to a 2 - 13 g increase in broiler weight at 6 - 8 wk of age (Wilson, 1991). The influence appears to be more pronounced in the eggs of younger breeders. The offspring of hens with heavier eggs have a later age of first egg, which also explains the problem that has always been challenging for poultry industry, that is why hens with heavier body weights tend to lay fewer eggs. A comprehensive evaluation of egg weight as an indicator for the transmission of economic traits between generations is rarely reported before, and this study can provide important reference value for poultry breeders.

In this study, heritability estimates for egg weight ranged from 0.45 to 0.63, and were consistent with previous studies, where a heritability of 0.63 was observed in brown-egg dwarf layers (Zhang, et al., 2005), Zhakao layers demonstrated a heritability of 0.50 (Alipanah, et al., 2013), and Rhode Island Red hens showed a heritability estimate of 0.45, with White Leghorn hens exhibiting a similar value of 0.46 (Bécot, et al., 2023). Increased heritability indicates that genetic factors account for a higher proportion of the phenotypic variance associated with this trait. Furthermore, correlation analysis across different time points revealed strong positive phenotypic and genetic correlations, which implied that egg weight was affected by similar genetic factors at different ages. This allows for greater integration of egg weights across multiple time periods for spatial and temporal analyses (Yi, et al., 2015).

We integrated single-trait and longitudinal GWAS to analyze genetic loci for egg weight across four stages. Five regions were identified on chromosomes 8, 10, 16, 28, and Z. The 2.97 - 3.32 Mb region on chromosome 28 was significantly associated with three time points, which indicated its ongoing influence on egg weight during the laying period. The integration of multi-point data with longitudinal GWAS enhanced the efficiency of QTL detection and minimized the incidence of false positive errors (Zhang, et al., 2021). Compared with single-trait GWAS, several novel intervals were identified on chromosomes 1, 2, 3, 4, 5, 15, 20, 21, and 27.

Through GWAS analysis, we obtained 111 candidate genes located on multiple chromosomes, such as ATF6, CSPG4, SH3GL3, C4, LMX1B, CDC34, and CCDC171. ATF6 is a transcription factor that is a target gene for activating the endoplasmic reticulum stress response, and plays a key role in the development of ovary architecture and is required for follicular physiology (Xiong, et al., 2016). CSPG4 and SH3GL3 are associated with the occurrence of cancer and relatively broad cancer suppressors (Lin, et al., 2021; Zhang, et al., 2024). C4 is part of the complement system and involved in immune responses (Wang and Liu, 2021).

We used longitudinal GWAS, gene association analysis, and selective sweep analysis to identify 22 KCGs (e.g., MAP2K2, CFD, CDC34, FSTL3, BSG, and LMX1B). Previous studies have highlighted that several genes associated with body weight and development have pleiotropic effects on egg weight (Fan, et al., 2017; Wolc, et al., 2014). For example, FSTL3 gene, which is associated with bone development, may contribute to calcium deposition and calcification during the formation of the eggshell (Nam, et al., 2015). CDC34, known as the ubiquitin-conjugating enzyme E2R1 (UBE2R1), plays a critical role in regulating various cellular processes, such as cell cycle progression, DNA repair, signal transduction, and protein turnover. This protein-coding gene is implicated in the ubiquitination pathway. Although no studies have directly shown an association between the CDC34 gene and egg weight, CDC34 is likely to be a critical factor in the aging process of human oocytes (Yang, et al., 2021). The LMX1B gene appears to have a significant association with egg weight (Kim, et al., 2022; Yeboah, et al., 2023). LMX1B is crucial for the regulation of limb regeneration and embryonic bone development, and plays a role in calcium deposition and mineralization during eggshell formation (Hanlon, et al., 2022; Kim, et al., 2022). Calcium deposition is related closely to the increase in egg weight, and an adequate supply of calcium helps to ensure that hens deposit calcium efficiently, thus improving egg quality and quantity (Jiang, et al., 2013).

The integration of TWAS and KCGs data identified four key candidate genes, which were located on chromosome 28. In the ChickenGTEx database, POLRMT gene is expressed in brain and trachea tissues and mainly affects crown weight, shell weight, egg weight at 44 wk of age, tibia length, and the number of small yellow follicles (Shen, et al., 2017). POLRMT is associated with the replication and transcription of mtDNA, which is linked to cellular and energy metabolism (Kühl, et al., 2016). Mitochondrial function affects energy production directly and may influence reproductive physiology and egg production in chickens. BSG is a transmembrane glycoprotein, which promotes angiogenesis and is linked to retinal development. The knockout of BSG in mice resulted in retinopathy and infertility (Muramatsu, 2016; Muramatsu and Miyauchi, 2003). We hypothesized that the expression of BSG genes may affect light perception, digestion and absorption, and calcium deposition in chickens, thereby regulating egg weight.

Five genes, which included MAP2K2 and CFD, were screened simultaneously based on WGCNAs, KCGs, and DEGs of the pituitary. In ChickenGTEx, both genes were expressed in multiple tissues and were associated directly with egg weight (Guan, et al., 2025). Further analysis conducted on MAP2K2 and CFD revealed that the variant rs315726522 might exert an influence on egg weight by modulating the expression levels of MAP2K2 in the pituitary. MAP2K2, which is a member of the MAP kinase family, is involved in the estrogen signaling pathway and the GnRH signaling pathway, both of which are essential for reproductive processes (Aoidi, et al., 2016; Dennis, et al., 2024; Nadeau, et al., 2009). The SNP rs315726522 may influence the expression levels of FSHB in the pituitary and FSHR in the ovary. The FSHB and FSHR genes are essential for regulating reproductive traits because they encode the key hormones and receptors involved in this process. CFD is a factor that influences egg production (Santoso, et al., 2019). A missense mutation (rs738839430) in the CFD gene did not affect its expression level, but altered the amino acid composition.

In total, we identified 13 DEMs in HEW and LEW and observed a significant association between egg weight and hub metabolism. 11-Deoxy prostaglandin F1α (11-deoxy PGF1α) is a synthetic analog of prostaglandin F1α. In animal experiments, 11-deoxy PGF1α stimulated uterine contractions in pregnant rats with an efficacy of about three times that of PGF1α (Lippmann, 1969; Zheng, et al., 2024). Prostaglandins play a critical role in poultry reproduction because they regulate the secretion of pituitary gonadotropins and prolactin, thereby affecting ovulation, fertilization, and embryonic development (Caldwell and Behrman, 1981; Chen, et al., 2024). Diethyl maleate, enrofloxacin, genistein 4′-O-glucuronide, and fenpropimorph are associated with oxidative stress in vivo and can affect animal performance through mechanisms that regulate oxidative stress. Among these, enrofloxacin affected the fatty acid content in animals (Kaur, et al., 2006; Oke, et al., 2024; Yang, et al., 2010). Quinolinic acid is an endogenous N-methyl-d-aspartate receptor agonist synthesized from l-tryptophan through the kynurenine pathway, which affects the metabolism of tryptophan in the body, thereby affecting the serotonin pathway to produce N-acetyl serotonin and melatonin, both of which are necessary for regulating circadian rhythms (Hardeland, 2010; Lee, et al., 2021). SM (8:1;2O/28:1) and SM (8:1;2O/28:2) are two isomers of sphingomyelin SM, which affected reproductive performance by affecting fat deposition in poultry (Fan, et al., 2022). Although these metabolites are not related directly to chicken reproductive performance, they may affect the egg-laying process indirectly through a variety of pathways that include regulation of energy metabolism, oxidative stress response, and hormone secretion.

Conclusion

The study has demonstrated that egg weight is not only related to the egg production of hens but also has a significant correlation with the hatching rate of fertilized eggs. In addition, egg weight is closely related to the growth performance and reproductive performance of chicks.

This study elucidated the genetic mechanisms that affect egg weight using multi-omics analyses. Using longitudinal GWAS, we identified 425 significant, 2,115 potentially significant SNPs, and 111 candidate genes related to egg weight. Combining selective sweep, transcriptome and metabolome analyses, we found 22 key candidate genes and 13 key metabolites that might affect egg weight. This study identified that the causative mutation rs315726522 affected egg weight by increasing the expression of the MAP2K2 gene in the ovary. An exon mutation rs738839430 was identified to affect egg weight by influencing the structure and function of the CFD gene. This study revealed the molecular mechanism of egg weight changes in chickens from a new perspective.

Data and model availability statement

The raw data reported in this paper have been deposited in the Genome Sequence Archive in the National Genomics Data Center, the China National Center for Bioinformation / Beijing Institute of Genomics, and the Chinese Academy of Sciences (CRA018570, CRA018550 and OMIX010193). The data are publicly accessible at https://ngdc.cncb.ac.cn/gsa.

Declaration of generative ai and ai-assisted technologies in the writing process

During the preparation of this work the author(s) did not use any AI and AI-assisted technologies.

CRediT authorship contribution statement

Jiqiang Ding: Writing – review & editing, Writing – original draft, Visualization, Validation, Formal analysis, Data curation, Conceptualization. Xiangping Liu: Visualization, Software, Formal analysis, Conceptualization. Ali Hassan Nawaz: Writing – review & editing, Supervision. Dong Leng: Visualization, Software, Formal analysis. Ni Li: Data curation. Doudou Ge: Data curation. Dongfeng Li: Writing – review & editing, Conceptualization. Chungang Feng: Writing – review & editing, Supervision, Resources, Funding acquisition, Conceptualization.

Disclosures

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

Appendix. Supplementary materials

Supplementary Figures 1-5

Supplementary Tables 1-34.