Research note: An Adaptive Lasso-based genome-wide association study of blood and meat spots in chicken eggs

Abstract

Blood and meat spots (BMS) are common internal defects in chicken eggs that compromise egg quality. To reveal genes contribute to BMS formation, we performed an Adaptive Lasso-based genome-wide association study (ALGWAS) of Rhode Island Red (RIR) hens. The results showed that the heritability of BMS was 0.13. Based on ALGWAS, we identified one significant and 144 suggestive SNPs associated with BMS. Within the significant locus, eight candidate genes were annotated (TNFRSF1A, CECR1, TRIM45, CALD1, SIDT1, VTCN1, USF3, ATP6V1A). Among these, VTCN1 emerged as the strongest candidate, supported by eQTL annotation and its established function as an immune checkpoint regulator that modulates inflammation, vascular integrity, and T-cell responses.

Introduction

Eggs are a staple of the modern diet. However, as a common internal defect, blood and meat spots (BMS) often provokes negative consumer responses. Consumers often mistake these spots for embryos, which causes discomfort that markedly reduces purchase intent. Beyond sensory and market concerns, BMS are linked to greater susceptibility to Salmonella contamination (Smith and Musgrove, 2008), presenting significant challenges to food safety.

BMS include two traits during egg formation in hens. Blood spot may come from occasional ovarian blood vessel rupture, and meat spots may from tissue fragments, with both now collectively termed BMS in recent research(Wu et al., 2025). The incidence of BMS varies by breed, brown-shelled hens are higher than white-shelled (Campo and Garcia Gil, 1998), indicating that genetic factors exert a substantial influence on it.

Rhode Island Red (RIR) hens are globally recognized as a premier brown-shelled egg layer. However, the genetic architecture of BMS remain unclear, exerts an adverse impact on the production. To address this problem, we recorded BMS phenotypes in 1,000 RIR hens to estimate heritability and conducted an Adaptive Lasso-based genome-wide association study (ALGWAS) in a subset of 80 individuals to identify candidate genes.

Materials and methods

Ethics statement

This study was conducted in accordance with the guidelines for the care and use of experimental animals established by the Animal Care and Use Committee of China Agricultural University.

Animals and data collection

A total of 1000 RIR hens (26 weeks old) provided by Hebei Zhongnongmufeng Layer Breeding Co., Ltd. (Hebei, China) were used in this study. All birds were reared in individual cages under identical environmental and management conditions with a 16 h:8 h light:dark cycle and a temperature of 20–22 °C. Birds had ad libitum access to feed and water. Phenotypic scores of BMS were recorded by raters for 5 consecutive days (Fig. 1A, B). All observers were trained using the following scoring criteria: 0 Normal egg content, no blood or meat spots. 1 Minor abnormality, small or faint spots. 2 Moderate abnormality, larger or more extended spots. 3 Severe abnormality, irregular and widespread spots. And inter-rater agreement was assessed using quadratic weighted kappa, calculated with the VassarStats online tool (http://www.vassarstats.net/). Based on consistent scoring over this period, 40 hens laying normal eggs (grade 0) and 40 hens laying BMS eggs (grade ≥ 1) were selected as the control and case groups, respectively.

Fig. 1.

Overview of samples characteristics. (A) Blood spot recording. Blood spot scores are 0, 1, 2, 3 from left to right. (B) Meat spot recording. Meat spot scores are 0, 1, 2, 3 from left to right. (C) IBD distribution. (D) Alignment quality metrics by group.

Estimation of heritability

Heritability was estimated using the DMU software. The model was specified as:

$$y=Xb+Za+e$$

Where y is the vector of BMS, b is the vector of fixed effects, a is the vector of random additive genetic effects, e is the vector of random residuals, and X and Z are the corresponding incidence matrices for fixed and genetic effects, respectively.

Genotyping and quality control

Genomic DNA was extracted with the Tiangen Blood DNA Kit (Tiangen Biotech, Beijing, China). Concentration and purity were quantified using a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific, USA). Qualified samples were submitted to Beijing Berry Hekang Biotechnology Co., Ltd. for whole-genome resequencing on the MGI DNBSEQ-T7 platform, generating 150 bp paired-end reads at an average depth of 10 × across 80 individuals.

Raw reads were quality-filtered with FASTP (v0.20.1) and aligned to the chicken reference genome (GRCg7b) using BWA-MEM (v0.7.17). Alignment quality was assessed using samtools (v1.20), mosdepth (v0.3.3), and Picard (v2.25.5). SNP calling was performed with GATK HaplotypeCaller (v4.1.9.0), followed by joint genotyping across all samples. Variants were filtered (QD < 2.0, MQ < 40.0, FS > 60.0, SOR > 3.0, MQRankSum < −12.5, ReadPosRankSum < −8.0) and further subjected to population-level quality control: individuals with >4 % missing genotypes, SNPs with >2 % missingness, MAF < 0.05 were excluded. Genotype quality (GQ) distribution was assessed using bcftools (v1.20). After these steps, independent SNPs (n = 193,364) were identified using PLINK (v1.90). Genome-wide significance thresholds were set via Bonferroni correction: significant level at 2.59 E-7 (0.05 / 193,364) and suggestive level at 5.17E-6 (1 / 193,364).

ALGWAS and post statistical power calculation

Adaptive Lasso-based genome-wide association study (ALGWAS) is divided into two stages (Yang et al., 2023). In stage 1, trait-associated SNPs are selected using the Adaptive Lasso variable selection method. The base model is:

$$y=X\beta +\epsilon$$

Where y is the phenotypic vector, X is the SNP genotype design matrix, β is the vector of SNP effect coefficients, and ε is the residual vector.

Adaptive Lasso applies weighted penalties: higher for SNPs with negligible effects, lower for those with non-zero effects. The selected SNPs constitute set B, obtained by:

$$\widehat{\beta }=\text{arg}\text{mi}{\mathrm{n}}_{\beta }\left(\parallel y-X\beta {\parallel }^2+\lambda \sum _{j=1}^p{w}_j\left|{\beta }_j\right|\right)$$

In Stage 2, we perform a genome-wide single-SNP scan. For each SNP tested, all SNPs located outside a ± 71.8 kb distance-window around it are included as covariates to account for background polygenic effects while minimizing linkage disequilibrium interference. The model for SNP $k$ is:

$$y=Xb+z_k{\gamma }_k+\epsilon$$

where $z_k$ is the genotype vector of SNP $\mathrm{k}$, ${\gamma }_{\mathit{k}}$ is its effect, and $\mathrm{X}$ is the design matrix of covariates including the intercept and selected SNPs from the set B. Based on this approach, we can obtain that the effect size of SNP $\mathrm{k}$ is estimated by adjusting for these covariates:

$${\widehat{\gamma }}_{\mathit{k}}=\frac{{\mathit{z}}_{\mathit{k}}^{\mathit{T}}\mathit{y}}{{\mathit{z}}_{\mathit{k}}^{\mathit{T}}{\mathit{z}}_{\mathit{k}}}-\sum _{j=1}^{\mathit{q}}\frac{{\mathit{z}}_{\mathit{k}}^{\mathit{T}}{\mathit{X}}_{\mathit{j}}}{{\mathit{z}}_{\mathit{k}}^{\mathit{T}}{\mathit{z}}_{\mathit{k}}}$$

Following this formula, we checked the effect sizes of selected SNPs in stage 1, and then calculated the significant loci using R (v 4.5.1). After this step, the statistical power was estimated for a given sample size at significant level (α = 2.59E-7) by Quanto (v1.2.2).

Linkage disequilibrium analysis and candidate genes identification

Linkage disequilibrium (LD) decay was estimated with PopLDdecay (v3.40). The LD decay distance was defined as the distance at which the mean $r^2$ first fell below 0.2 (Bordbar et al., 2024). Candidate genes within associated genomic regions were annotated using BEDTools (v2.30) against the chicken reference genome (GRCg7b). Functional enrichment analysis was conducted with DAVID to identify Gene Ontology terms. In addition, eQTL annotations were retrieved from the ChickenGTEx Atlas to prioritize candidates.

Data analysis and visualization

Phenotypic data were analyzed using SPSS (v26.0) and are presented as mean ± SD. BMS scores were recorded and preprocessed in Microsoft Excel. Representative egg illustrations corresponding to different BMS scores were created using CorelDRAW 2017. Linkage disequilibrium (LD) decay curves were constructed using PopLDdecay (v3.40). Identity-by-descent (IBD) segments were estimated using PLINK (v1.90). Population structure was assessed via principal component analysis (PCA), and chi-square test were performed in R (v4.5.1). The lead SNP was visually inspected to verify the consistency with the known mutations using the IGV (v 2.19.7).

Results and discussion

Phenotypic data and heritability analysis

In a total of 1,000 RIR hens, the phenotypic score of BMS was 0.67 ± 0.77, while it was 2.15 ± 0.76 in our case group (P<0.05). This reveals that familial patterns further support genetic influence, as certain lineages consistently show higher BMS frequencies than others (Jeffrey and Pino, 1943). Notably, approximately 80 % of BMS cases were meat spots, while only 20 % were blood spots, suggesting that oviduct inflammation may play a key role in BMS formation. Since phenotypic measurement noise can substantially bias association results and affect locus localization, we evaluated rating consistency using the kappa statistic. The obtained value of 0.643 indicates high inter-rater agreement, thereby reducing noise in the phenotypic data for subsequent ALGWAS. Based on the pedigree of population, we calculated the heritability of BMS at 26 weeks of age, and it was estimated to be 0.13, indicating a low component.

Adaptive Lasso-based genome-wide association study of BMS

A Model selection is crucial in GWAS. In this study, we treated BMS as an ordinal trait and implemented the Adaptive Lasso approach to enhance detection precision. To mitigate confounding due to population structure, we selected 80 hens from 47 distinct families and the IBD distribution indicated low relatedness within the population (Fig. 1C). The cohort consisted of 40 hens that consistently produced eggs with a BMS score of 0 and 40 hens with an average BMS score of ≥ 2.15. Phenotypic records showed that approximately 70 % of BMS eggs contained meat spots, while the remaining 30 % exhibited blood spots. Quality control metrics revealed no systematic differences between the two groups, confirming the absence of batch effects (Fig. 1D).

Population stratification was assessed using PCA plot, which showed no evident substructure (Fig. 2A). The λGC was 1.037 in the quantile–quantile plot (Fig. 2B), further supporting adequate control of population stratification. ALGWAS identified one SNP on chromosome 1 that exceeded the genome-wide significance threshold (P < 2.59 E-7) and 144 additional SNPs on chromosomes 1, 2, 3, 5, 10, 17, and 25 that reached suggestive association (P < 5.17E-6). Post power analysis indicated that the lead SNP had a minor allele frequency of 0.37 and an effect size of 0.67 (95 % CI: 0.58–0.76). Assuming a statistical power of 0.8, the estimated required sample size was 82, which closely aligns with our actual sample size.

Fig. 2.

Genetic structure of BMS. (A) PCA. (B) Q-Q plot of BMS. (C) LD decay analysis. (D) GO enrichment analysis of identified genes. (E) Manhattan plot of BMS. Horizontal lines mark the genome-wide (2.59E-7) and suggestive (5.17E-6) significance thresholds.

Visual inspection using IGV confirmed an A > G variant at position 79,394,981 on chromosome 1, corresponding to the significant association signal and annotated in the Ensembl database. We used chi-square test to compare the allele frequency of two groups, and the result showed a significant difference (χ² = 5.047, P < 0.05). LD decay was estimated at 71.8 kb in this population (Fig. 2C). Within this region, 8 candidate genes were identified (Fig. 2E): tumor necrosis factor receptor superfamily member 1A (TNFRSF1A), cat eye syndrome chromosome region candidate 1 (CECR1), tripartite motif-containing 45 (TRIM45), caldesmon 1 (CALD1), SID1 transmembrane family member 1 (SIDT1), V-set domain-containing T-cell activation inhibitor 1 (VTCN1), upstream transcription factor 3 (USF3), and ATPase H⁺-transporting V1 subunit A (ATP6V1A).

Candidates identification and function analysis

GO enrichment analysis linked the 8 candidate genes to inflammation, membrane dynamics, and lipid metabolism (Fig. 2D). Enriched terms included nervous system development, cytokine-mediated signaling, and glucosylceramide catabolism, suggesting possible roles for neuro-immune crosstalk and sphingolipid metabolism in BMS. Molecular functions highlighted cytokine receptor and glucosylceramidase activity, consistent with immune signaling and lysosomal function. Cellular components were enriched for the external plasma membrane and lysosomal membrane, supporting involvement of membrane receptors and degradation machinery in BMS pathogenesis. Further, we looked up the eQTL annotation of these genes from the ChickenGTEx Atlas, and the result provided a regulatory support for VTCN1, one of the eight candidate genes. A variant near VTCN1 (rs731689404) was significantly associated with its expression in ovary, indicating potential regulatory activity in reproductive tissues.

VTCN1, also known as B7-H4, B7S1, and B7x, is involved in negatively regulating T-cell activation (Vaishnav et al., 2022). VTCN1 is an immune checkpoint molecule expressed in antigen-presenting cells and non-lymphoid tissues. Research showed it could suppress T-cell activation, proliferation, cytokine production, and cytolytic activity via a receptor distinct from the CD28 family, likely BTLA (Prasad et al., 2003; Zang et al., 2003; Sica et al., 2003). In the context of BMS, identification of VTCN1 as a candidate gene—along with enriched cytokine-related pathways—points to localized immune dysregulation in the oviduct. Aberrant VTCN1 expression may disrupt the balance between pro‑inflammatory and immunosuppressive signals during ovulation, impairing inflammation resolution or immune surveillance. This imbalance could lead to microhemorrhages or cellular debris retention, ultimately manifesting as BMS. As these deductions are drawn from bioinformatics analyses, further experimental validation in independent cohorts is needed.

CRediT authorship contribution statement

Xiaolong Ma: Writing – review & editing, Writing – original draft, Methodology, Investigation, Formal analysis, Conceptualization. Xing Chen: Writing – review & editing, Validation, Project administration, Investigation. Xue Cheng: Investigation, Data curation. Wei Liu: Resources. Zhonghua Ning: Writing – review & editing, Supervision, Project administration, Funding acquisition, Conceptualization.

Disclosures

No conflict of interest exists in the submission of this manuscript, and manuscript is approved by all authors for publication. On behalf of all the authors, I declare that this paper is original and none of the material in the paper has been published or is under consideration for publication elsewhere. All the authors listed have read the manuscript and approved the submission of the paper to your journal.