Genome-wide association study reveals loci associated with cytokine levels in chicken serum

Abstract

Cytokine-mediated immune responses constitute a critical line of defense that protects chickens from invasion of viruses, bacteria, parasites and other pathogens. Accordingly, the present study aimed to identify genomic variants and candidate genes linked to serum cytokine concentrations in chickens, with the long-term objective of uncovering potential molecular markers to support selective breeding for improved immune traits in chickens. For this purpose, a total of 206 ten-week-old Qiandongnan Xiaoxiang chickens were selected as experimental animals. Serum concentrations of six key cytokines—interferon gamma (IFN‑γ), interleukin‑1β (IL‑1β), interleukin‑6 (IL‑6), interleukin‑17 (IL‑17), interleukin‑22 (IL‑22), and tumor necrosis factor alpha (TNF‑α)—were quantified and regarded as immune-related phenotypic traits. Subsequently, genome-wide resequencing data were analyzed using the FarmCPU model implemented in the rMVP package to conduct genome-wide association studies (GWAS), enabling the identification of critical loci significantly associated with the measured immune phenotypes. Importantly, the GWAS identified 58 loci significantly associated with serum cytokine traits; these loci reside in or near 20 candidate genes such as SCRN1, GALT-CNTFR, and MESDC1-MESDC2. Overall, this research delineates the population variation patterns of major serum cytokines in Qiandongnan Xiaoxiang chickens. These loci and candidate genes can be used to screen molecular markers for immune traits.

Introduction

Diseases are a major constraint on production performance in the poultry industry and often lead to economic losses accounting for 10%–20% of total industry output (Berghof et al., 2019). In recent years, China has strictly enforced regulatory restrictions on the prophylactic use of veterinary antibiotics in livestock and poultry breeding, (Ministry of Agriculture and Rural Affairs of the People’s Republic of China, 2019). Against this regulatory backdrop, the domestic chicken industry faces an urgent need to develop alternative strategies to address mounting challenges in disease prevention and control. Genetic selective breeding, which excavates the natural genetic variation of chicken populations, represents a sustainable route to improve chicken disease resistance. In this context, screening and verifying molecular markers serve as the essential prerequisite for implementing targeted genetic breeding (Wang et al., 2014). Acting as the frontline defense against exogenous pathogens, innate immunity lays the critical physiological foundation for broad-spectrum disease resistance in chickens, and cytokines serve as its core mediators. Cytokines include interleukins (ILs), interferons (IFNs), tumor necrosis factor (TNF) superfamily proteins, colony-stimulating factors (CSFs), and chemokines. Upon activation of chicken innate immunity, diverse cytokines and chemokines are secreted, which subsequently activate adaptive immune cells to eliminate or limit pathogen invasion. A wealth of previous studies has demonstrated that signature cytokines including interferon‑γ (IFN‑γ), interleukin‑6 (IL‑6), and interleukin‑18 (IL‑18) are closely involved in the early host immune response following pathogen infection. For example, in chickens infected with Salmonella (Swaggerty et al., 2014) or Marek's disease virus (Kaiser et al., 2003), the expression level of IFN-γ in resistant lines is significantly higher than that in susceptible lines, suggesting that IFN-γ expression could serve as a reliable indicator reflecting the disease resistance of chickens against pathogen infection. Additionally, prior studies have revealed that the secretion capacity of IL-6 and IL-18 may be correlated with resistance to Marek’s disease in chickens (Kaiser et al., 2003). Notably, when circulating IL-6, CXCLi2 and CCLi2 were used as biomarkers for selective breeding targeting disease resistance, the generated chicken lines displayed substantially enhanced resistance to a wide range of pathogens, including Salmonella, Clostridium perfringens, Eimeria and Campylobacter (Swaggerty et al., 2017). Taken together, these findings demonstrate that the level of cytokine responses can reliably reflect chicken susceptibility to various pathogens. A clinical study on human influenza vaccines proposed the concept of the "naturally adjuvanted immune setpoint". This idea offers mechanistic evidence indicating that immune disparities among healthy individuals stem from differences in these baseline setpoints (Mulè et al., 2024). A moderate increase of baseline serum cytokine concentrations suggests the immune system maintains a "naturally pre-activated" or alert immune status with an elevated setpoint. Such a physiological state enables more rapid and robust immune reactions during pathogen infection or vaccination. Research on geese revealed that favorable raising environments (e.g., cornfield free-range rearing) correlated with elevated serum IFN-α and IFN-γ concentrations. The authors speculated that this phenotype strengthened the geese’s capacity to resist viral invasion while lowering the incidence of autoimmune disorders. (Wang et al., 2015). Hence, exploring cytokine responses helps elucidate the molecular mechanisms regulating disease resistance in chickens, and also provides valuable immune biomarker traits to support molecular breeding for disease-resistant chicken lines.

Genetic regulation of cytokine production has been thoroughly characterized in human functional genomic studies. For instance, the Human Functional Genomics Project (HFGP) verified that host genetic factors explain 25% to 50% of the variation in cytokine production capacity (Li et al., 2016a). Numerous studies have utilized genome-wide association studies (GWAS) and other approaches to systematically map genetic loci associated with cytokine-related phenotypes (Sliz et al., 2019). Multiple genes such as PDGFRB and ABO have been confirmed to participate in cytokine response networks (Nath et al., 2019). Moreover, the NAA35-GOLM1 locus has been proven to modulate IL-6 production upon diverse pathogen challenges and is correlated with human susceptibility to candidemia (Li et al., 2016a). In another research cohort from the HFGP, the TLR1-TLR6-TLR10 cluster located on chromosome 4 was shown to potentially modulate the secretion levels of IL-6, IL-1β, and TNF-α (Li et al., 2016b). In contrast, analogous GWAS analyses targeting chicken cytokine traits are still relatively scarce. Even so, the available published studies have identified abundant genetic variants located in genomic regions associated with chicken cytokine regulation. For example, quantitative trait loci (QTLs) correlated with IFN‑γ show strong associations with chicken cellular immune responses, and multiple key loci controlling IFN‑γ expression have been preliminarily mapped (Kianpoor et al., 2025). Furthermore, Zhang et al. detected 39 single-nucleotide polymorphisms (SNPs) that were significantly associated with immune phenotypes, including serum IgY levels in Beijing You chickens (Zhang et al., 2015). Overall, the identification of cytokine quantitative trait loci (cQTLs) has accelerated studies exploring immune regulatory mechanisms and human host susceptibility to infectious diseases. Nevertheless, due to considerable differences in the genetic backgrounds of chickens and humans, findings from human research cannot be directly applied to disease-resistance selective breeding in chickens.

Accordingly, cytokines can serve as stable and quantifiable indicator traits to assess disease resistance. Compared with traditional phenotypic detection methods for disease resistance, cytokine measurement is convenient to operate and high-throughput, making it highly suitable for deciphering the genetic mechanisms regulating immune modulation and identifying molecular markers in chickens. Qiandongnan Xiaoxiang chicken is an indigenous chicken breed native to the Qiandongnan Miao and Dong Autonomous Prefecture, Guizhou Province, China. As an excellent miniature local poultry germplasm resource, this breed exhibits strong disease resistance and favorable environmental adaptability, which makes it an optimal animal model for poultry disease resistance research. In this study, 10-week-old individuals of this breed were used as experimental animals to characterize population-wide variation in cytokine abundance, screen genomic variants significantly associated with cytokine immune traits, and identify key candidate genes modulating immune reactions. We measured the concentrations of six critical cytokines and performed whole-genome resequencing combined with GWAS to locate trait-linked SNPs and functional candidate genes. This research intends to resolve the genetic architecture governing cytokine-associated immune traits across the whole genome, laying theoretical and empirical support for marker-assisted selection (MAS) of immune traits in indigenous chickens and the exploration of superior genes from local poultry germplasm resources.

Materials and methods

Experimental animals

The experimental animals used in this study were Qiandongnan Xiaoxiang chickens, raised at the animal housing facility of the College of Animal Science, Guizhou University between April and September 2024. A total of 206 chickens (114 males and 92 females) were hatched and reared under consistent incubation conditions, feeding schedules and standardized husbandry protocols. Specifically, all chickens were kept in a unified breeding room from 1 to 4 weeks of age, then transferred to another identical rearing room until reaching 10 weeks of age. The stocking density was 16 birds/m² from 0 to 4 weeks of age and 8 birds/m² from 4 to 10 weeks of age. Males and females were reared in separate single-sex groups throughout the experiment. A 16 h daily light photoperiod was maintained throughout the trial, with ambient temperature controlled within a range of 15°C to 35°C. All chickens were fed commercial feed at 9:00 and 17:00 daily. The feed, including broiler starter feed (Formula 510) and grower feed (Formula 511), was purchased from Guiyang Hengchen Feed Co., Ltd. Feeding procedures were implemented as follows: Formula 510 starter feed was supplied to all chickens from 1 to 5 weeks of age. A transitional mixed diet consisting of Formula 510 and Formula 511 was offered at week 6, followed by exclusive Formula 511 grower feed from 7 to 10 weeks of age. All 206 chickens were kept under uniform feeding and management conditions. Feed and drinking water were supplied ad libitum, and the detailed feed composition is presented in Table 1.

Table 1.

Composition and nutrient levels of experimental diets (air-dried basis) %.

Item	Age period
	0–42 d	43–70 d
Ingredients, %
Corn	56.30	58.62
Soybean meal	18.52	25.00
Rapeseed meal	10.45	0.00
Corn gluten meal	6.33	3.05
Wheat bran	2.94	5.63
Soybean oil	1.63	3.34
Limestone	1.18	1.18
Phytase	0.04	0.00
Choline chloride	0.15	0.00
Methionine	0.15	0.10
Lysine	0.22	0.32
NaCl	0.15	0.15
CaHPO4	1.89	1.61
Premix1)	0.05	1.00
total	100.00	100.00
Nutrients,%
CP	21.18	19.05
ME (MJ/kg)	12.12	12.56
Ca	1.00	0.90
AP	0.45	0.40
Met +Cys	0.90	0.72
Lys	1.06	0.90

The premix provided the following per kg of diets: 0–42 d: VA 6000 IU, thiamin 2.0 mg, riboflavin 4.0 mg, niacin 42.0 mg, pyridoxine 4.0 mg, cobalamin 0.01 mg, VD₃ 2000 IU, VE 30 IU, VK₃ 1.8 mg, calcium pantothenate 10.0 mg, biotin 0.15 mg, folic acid 0.85 mg, Fe (as ferrous sulfate) 80.0 mg, Cu (as copper sulfate) 8.0 mg, Mn (as manganese sulfate) 80.0 mg, Zn (as zinc sulfate) 65.0 mg, I (as potassium iodide) 0.50 mg, Se (as sodium selenite) 0.25 mg; 43–70 d: VA 10000 IU, thiamin 2.50 mg, riboflavin 7.5 mg, pyridoxine 3.0 mg, VD₃ 2000 IU, VE 25.0 mg, VK 2.8 mg, nicotinamide 40.0 mg, calcium pantothenate 25.0 mg, biotin 0.20 mg, folic acid 1.5 mg, cobalamin 0.015 mg, Fe (as ferrous sulfate) 80.0 mg, Cu (as copper sulfate) 8.0 mg, Mn (as manganese sulfate) 100.0 mg, Zn (as zinc sulfate) 60.0 mg, I (as potassium iodide) 0.35 mg, Se (as sodium selenite) 0.3 mg.

Sample collection

At 10 weeks of age, all chickens were subjected to an 8-hour fasting period prior to body weight measurement. Following this procedure, 2–3 mL of blood was collected from the brachial vein. Two types of vacuum blood collection tubes were used for sampling: anticoagulant-free tubes for serum separation, and EDTA-containing tubes for whole blood preservation. Blood samples were centrifuged at 3,000 rpm for 3–5 min, and the supernatant was collected to harvest serum samples. Whole blood samples were repeatedly and gently inverted to ensure adequate homogenization. All samples were temporarily stored at −80 °C and subsequently shipped to Shanghai Majorbio Bio-Pharm Technology Co., Ltd. for whole-genome resequencing.

Cytokine level measurement

Serum concentrations of six key cytokines, including IFN-γ, IL-1β, IL-6, IL-17, IL-22, and TNF-α, were quantified using commercial ELISA kits specific for chicken immune cytokines (Zike Bio Co., Ltd., Shenzhen, China). Microplate reader model:Infinite F50 (Tecan). Kit catalog numbers: ZK-C6300 Chicken Interferon γ (IFN-γ) ELISA Kit, ZK-C6332 Chicken Interleukin 1β (IL-1β) ELISA Kit, ZK-C6329 Chicken Interleukin 6 (IL-6) ELISA Kit, ZK-C6337 Chicken Interleukin 17 (IL-17) ELISA Kit, ZK-C7108 Chicken Interleukin 22 (IL-22) ELISA Kit, ZK-C6305 Chicken Tumor Necrosis Factor α (TNF-α) ELISA Kit. All 206 chicken serum samples were tested in five independent batches. Before cytokine quantification, serum aliquots were taken from the −80 °C freezer and thawed at room temperature. Meanwhile, the ELISA kits were moved from 2–8 °C refrigeration to room temperature and equilibrated for 30 min prior to use. The detailed experimental procedures were performed as follows: standard wells and sample wells were arranged in accordance with the experimental design. Gradient concentrations of standard solutions (50 μL per well) were sequentially loaded into the standard wells. For each test sample, 10 μL serum sample was pipetted to the corresponding sample well, followed immediately by 40 μL sample diluent. Each sample was assayed in technical triplicate. Subsequently, 100 μL of horseradish peroxidase (HRP)-conjugated detection antibody was dispensed into all reaction wells. The plate was sealed with a adhesive sealing film and incubated at 37 °C for 60 min in the dark. After incubation, the liquid within each well was aspirated, and residual liquid on the microplate was blotted dry with absorbent filter paper. Each well was filled with 330 μL wash buffer, incubated for 1 min, then emptied and blotted dry. This step was repeated five times in total. Following the final blot, 50 μL of substrate A and 50 μL of substrate B were sequentially added to each well, gently mixed, and the plate was covered with a new sealing film and incubated at 37 °C in the dark for 15 minutes. Following color development, 50 μL of stop solution was added to each well to halt the enzymatic reaction. The optical density (OD) of each well was measured at a wavelength of 450 nm using a microplate reader within 15 minutes of reaction termination. A standard curve was generated for each plate using the provided standards, and sample concentrations were interpolated from the standard curve.

Genome sequencing and quality control

Genomic DNA was isolated from chicken anticoagulated whole blood using a magnetic bead-based universal genomic DNA extraction kit (Shanghai Majorbio Bio-Pharm Biotechnology Co., Ltd., China). Briefly, 200 μL whole blood was mixed with proteinase K, lysis buffer and RNase A, then purified on a fully automatic nucleic acid extractor following the standard program. DNA integrity and purity were jointly evaluated via 1% agarose gel electrophoresis and NanoDrop (Thermo Fisher Scientific, USA) ND-2000 spectrophotometer. Only DNA samples with OD₂₆₀/₂₈₀ ranging from 1.8 to 2.0 and OD₂₆₀/₂₃₀ ≥ 2.0 without obvious degradation were retained for library construction. Approximately 0.5 μg of qualified genomic DNA was fragmented to ∼350 bp by sonication, followed by end repair, A-tailing, adapter ligation and PCR amplification. Purified libraries were quantified and quality-check, then sequenced on Illumina NovaSeq X Plus platform (Illumina, San Diego, USA) with PE150 mode. A total of 2,566.41 Gb raw data were generated for 206 chickens, with average raw Q30 of 95.35%. The actual effective sequencing depth per sample ranged from 10.19× to 15.93×, with a mean depth of ∼12× across the entire population. The average genome coverage of ≥1× and ≥4× reached 98.12% and 94.36%, respectively. Raw FASTQ reads were initially processed for quality filtering with Fastp software (v0.23.2) (Chen, 2023). After quality filtering, the resulting high-quality reads were aligned against the chicken (Gallus gallus) reference genome using the BWA-MEM algorithm (v1.0.6) (Gao et al., 2025). The resulting BAM files were sorted via SAMtools (v1.15.1) (Vargas-Salgado et al., 2024), and duplicate sequences arising from PCR amplification were marked using the MarkDuplicates tool (v3.3.0) (Zhang et al., 2024). Subsequently, single nucleotide polymorphisms (SNPs) were genotyped for all samples utilizing the HaplotypeCaller and GenotypeGVCFs modules of the GATK software (v4.3.0.0) (Zhang et al., 2025), and functional annotation of variant loci was completed via SnpEff software (v5.1d) (Riccio et al., 2024). Ultimately, a total of 13,560,419 high-quality SNPs and Indels were obtained for downstream analyses. These SNPs satisfied the following filtering criteria: mean site depth > 4×, missing rate < 30%, minor allele frequency (MAF) > 5% and linkage disequilibrium (LD) r² threshold < 0.95.

Genome-wide association study

A genome-wide association study (GWAS) was performed using the rMVP software (v1.0.6) (Yin et al., 2021) based on the Fixed and random model Circulating Probability Unification (FarmCPU) method. In this study, SNP effect sizes were estimated under an additive genetic model. Genotypes were coded as follows: homozygous for the reference allele (REF/REF) = 0, heterozygous (REF/ALT) = 1, and homozygous for the alternative allele (ALT/ALT) = 2. The effect allele was defined as the alternative allele (ALT). Accordingly, a positive effect estimate indicates that each copy of the ALT allele is associated with an increased trait value, while a negative effect estimate indicates a decreased trait value. This genotype coding scheme matches the default additive model implemented in the rMVP software. We performed principal component analysis (PCA) using PLINK software (v1.90b6.20) based on the genome-wide SNP data. To eliminate confounding effects from population structure and non-genetic factors on phenotypic variation, the first three principal components (PC1–PC3) and experimental batch were incorporated as fixed effects in the FarmCPU model. Batch information was recoded into four dummy variables and integrated into the fixed-effect component of the model. Sex was not included as a fixed effect because it showed no statistically significant effect on the phenotypes in univariate or bivariate analyses. Population structure and ancestry were inferred using ADMIXTURE (v1.3.0)(Alexander et al., 2009). To determine the optimal number of ancestral clusters, we tested K values from 1 to 20. The optimal number of clusters was selected based on the lowest CV error. Pairwise kinship was estimated with PLINK (Purcell et al., 2007) using the method reported by Manichaikul et al. (Manichaikul et al., 2010). This was used to assess relatedness among individuals. Separately, a genomic relationship matrix (G matrix) was built from all genome‑wide SNPs. This matrix was included as a random effect in the FarmCPU model to correct for cryptic relatedness. Fixed effect model:

$$\text{yi}=\text{Batch}+P{C}_1{a}_1+P{C}_2{a}_2+P{C}_3{a}_3+M_{i1}{\mathrm{b}}_1+M_{i2}{\mathrm{b}}_2\dots +M_{it}{b}_t+S_{ij}{\mathrm{d}}_j+e_i$$

Where: y_i represents the phenotypic value (cytokine concentration) of the i-th individual after adjustment for fixed effects. Batch stands for the categorical experimental batch covariate. Batch is a categorical variable with five levels (B, C, D, E, F). This variable contains five levels (B, C, D, E, F). Batch B was designated as the reference group, and batch effects were recoded into four dummy variables that were incorporated into the fixed-effect component of the model. PC₁, PC₂, and PC₃ stand for the first three principal components obtained from principal component analysis (PCA); a₁, a₂ and a₃ represent their corresponding effect estimates. M_i1, M_i2, …, M_it refer to the genotypes of candidate associated loci added to the model at the t-th iteration (this term is empty at the first iteration); b₁, b₂, …, b_t correspond to their effect values. S_ij represents the genotype of the i-th individual at the j-th marker, and d_j corresponds to its effect value. e_i indicates the residual error, which follows a normal distribution with a mean of 0 and a variance of σ_e². Functional annotation of loci passing the genome-wide significance cutoff was performed using SnpEff (version 5.2b). In this GWAS analysis, we first calculated a stringent Bonferroni-corrected threshold as a reference. The genome-wide significance threshold was computed to correspond to -log10(P) ≈ 8.43, which was found to be overly conservative in follow-up analyses. Given the moderate sample size of the present study, we adopted a more lenient threshold to lower the probability of false-negative results. The threshold of `-log10(P) = 6` was set with reference to commonly used GWAS significance criteria in poultry and other species such as pigs (Cai et al., 2024; Han et al., 2025; Reyer et al., 2017). Additionally, general linear model (GLM) and mixed linear model (MLM) were also implemented for complementary verification, and their outcomes are presented for reference.

Quantitative real-time PCR (qPCR) analysis

An exploratory transcriptional profiling experiment was conducted with an independent chicken population. Eighteen 35-day-old male broilers were randomly chosen for this assay. Animals exhibiting extreme phenotypes were categorized into high and low subgroups according to serum cytokine concentrations. Intergroup differences between the two subgroups were compared via the Wilcoxon rank-sum test, with six chickens allocated to each group. Notably, this cross-age trial merely functions as exploratory correlation verification; it cannot provide conclusive causal validation for cytokine-associated candidate genes discovered in the GWAS performed on 10-week-old chickens.

Total RNA was extracted from whole blood using TRIzol reagent, followed by reverse transcription using the NotionScript cDNA First-Strand Synthesis Kit (with dsDNase). Primers for the three target genes were designed via NCBI Primer-BLAST and synthesized by Sangon Biotech (Shanghai). Relative gene expression was determined using the 2^(-ΔΔCt) method, and detailed primer sequences are listed in Supplementary Table 1. Differences in expression data were analyzed using unpaired two-tailed Student’s t-test; Welch’s t-test was adopted when variance heterogeneity existed between groups. Statistical significance was determined by unpaired two-tailed Student's t-test comparing the raw 2^−ΔCt values between the high- and low-expression groups. For visualization, all expression data were normalized against the low-expression group, whose value was calibrated to 1.0. Normality was assessed by Shapiro-Wilk test.

Other statistical methods

The Wilcoxon rank-sum test was adopted to assess sex-associated disparities in serum cytokine concentrations. Spearman’s rank correlation analysis for the six cytokines were completed using the corr.test. The Kruskal–Wallis H test was additionally utilized to explore the impacts of experimental batches on target cytokines to quantify the statistical significance of batch effects. To jointly assess the independent contributions of sex and batch to serum cytokine levels and to rule out potential confounding interactions between these two variables, we performed Scheirer-Ray-Hare tests, the nonparametric alternative to two-way analysis of variance, for each of the six cytokines with sex, batch, and their interaction as fixed factors. All statistical computations and data visualization were accomplished in R software (v4.5.2). Furthermore, functional enrichment analysis of candidate genes was performed using Gene Ontology (GO) annotation (https://www.geneontology.org/) and the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway database (https://www.genome.jp/kegg/) (Bu et al., 2021).

Results

Effects of experimental batch and sex on serum cytokine levels in chickens

We quantified the serum concentrations of six cytokines in experimental chickens (Fig. 1 and Table 2). The coefficient of variation (CV) for the six cytokines ranged from 7.26% to 34.74%, indicating substantial inter-individual variabilityin serum cytokine levels among the tested chickens. Normality testing revealed that all cytokine data deviated from a normal distribution, thereby justifying the adoption of non-parametric statistical analyses in subsequent analyses.

Fig. 1.

Density distribution of six serum cytokines in chicken (n = 206). The x‑axis shows concentration (pg/mL). The y‑axis indicates density. (A) IFN‑γ; (B) IL‑1β; (C) IL‑6; (D) IL‑17; (E) IL‑22; (F) TNF‑α.

Table 2.

Quantitative profiles of six serum cytokines in chickens.

Trait	Min (pg/ml)	Max (pg/ml)	Mean (pg/ml)	SD (pg/ml)	CV (%)
IFN-γ	176.05	579.03	375.95	109.25	29.06
IL-1β	55.52	226.52	141.49	45.43	32.11
IL-6	13.85	56.75	34.05	11.83	7.26
IL-17	20.79	87.75	53.11	18.19	34.74
IL-22	21.39	91.74	55.93	18.66	33.36
TNF-α	34.17	131.15	81.15	26.17	32.25

To explore the potential effects of sex and experimental batch on cytokine concentrations, we performed the Wilcoxon test to assess sex-dependent differences. The results showed that sex exerted no significant influence on the concentrations of all six cytokines (P > 0.05; Fig. 2). In comparison, the Kruskal-Wallis H test revealed statistically significant batch-dependent differencesin the levels of all six cytokines (P < 0.0001; Fig. 4). We further conducted Scheirer-Ray-Hare tests to evaluate the main effects of sex, experimental batch, and their interaction on serum cytokine levels. The results verified that experimental batch exhibited a highly significant main effect on all six cytokines (P < 0.001). By contrast, sex showed no significant main effect on any of the detected cytokines (P > 0.05), and no significant sex × batch interaction was observed for all cytokine indices (P > 0.05). These findings confirmed that sex had a negligible effect on cytokine levels and did not confound the batch effect (Fig. 5 and Table 3). Accordingly, batch effect correction was applied to all subsequent statistical analysesto eliminate experimental bias. In addition, Spearman’s rank correlation analysis identified strong positive correlations among all six cytokines, with correlation coefficients varying from 0.89 to 0.91 (Fig. 3).

Fig. 2.

Effects of sex on serum cytokine levels in experimental chickens assessed via the Wilcoxon test (males, n = 114; females, n = 92). The x-axis shows the two sex groups (male and female), and the y-axis represents cytokine concentration (pg/mL). (A) IFN-γ; (B) IL-1β; (C) IL-6; (D) IL-17; (E) IL-22; (F) TNF-α.

Fig. 4.

Violin plots with boxplots and individual data points showing the distribution of six serum cytokines across five experimental batches (B–F). The violin contours show the probability density of cytokine data; internal boxplots present medians and interquartile ranges, and data points denote individual sample values. The x-axis denotes experimental batch, and the y-axis indicates cytokine concentration (pg/mL). Kruskal-Wallis tests revealed highly significant batch effects on the levels of all six cytokines (P < 0.001). (A) IFN-γ; (B) IL-1β; (C) IL-6; (D) IL-17; (E) IL-22; (F) TNF-α.

Fig. 5.

Distribution of six serum cytokine concentrations grouped by sex and batch. Boxplots show medians, interquartile ranges, and whiskers (1.5 × IQR). Males (blue, n = 114) and females (orange, n = 92). The x-axis denotes batch (B–F), y-axis shows cytokine concentration (pg/mL). Scheirer-Ray-Hare tests were used for statistics (see Table 3 for P values). (A) IFN-γ; (B) IL-1β; (C) IL-6; (D) IL-17; (E) IL-22; (F) TNF-α.

Table 3.

Scheirer-Ray-Hare test P-values for the effects of sex, batch, and their interaction on six serum cytokines.

Trait	Sex (P value)	Batch (P value)	Sex × Batch (P value)
IFN-γ	0.899	<0.001	0.998
IL-1β	0.764	<0.001	0.999
IL-6	0.987	<0.001	0.941
IL-17	0.619	<0.001	0.998
IL-22	0.796	<0.001	0.979
TNF-α	0.845	<0.001	0.990

Fig. 3.

Spearman’s correlation analysis of six serum cytokines. Diagonal panels present frequency histograms with fitted density curves for individual cytokine levels. The upper triangular matrix shows pairwise Spearman’s correlation coefficients (* P < 0.05, ** P < 0.01, *** P < 0.001). The lower triangular matrix displays scatter plots with linear regression fitting, with the coefficient of determination (R²) annotated to evaluate the linear fit goodness between paired cytokines.

GWAS for six serum cytokines in chickens

A total of 206 chickens underwent whole-genome resequencing in this study. After quality control filtering, 13,560,419 SNPs and Indels were retained for subsequent GWAS analyses targeting serum cytokine levels. PCA results revealed no evident population stratification (Fig. 6A). Population structure analysis determined the optimal grouping as K = 1 (Fig. 6B). The population structure plot displayed solid single color at K = 1 (Fig. 6C), demonstrating a common ancestral source. The kinship heatmap contained only scattered closely related individual pairs rather than large dense colored zones (Fig. 6D), confirming overall homogeneous ancestry in this population. We further evaluated the performance of the association model in eliminating population stratification bias. The Q-Q plots illustrated that the λ values corresponding to all six serum cytokines were within the acceptable threshold range, verifying reliable correction of population stratification (Fig. 7).

Fig. 6.

Population genetic structure and kinship of the studied population. (A) Three-dimensional PCA plot colored according to PC3 scores. X-axis: PC1 (explaining 0.88% variance); Y-axis: PC2 (0.82% variance); Z-axis: PC3 (0.77% variance). (B) Cross-validation errors obtained with different K values for population structure analysis. The X-axis represents the putative number of ancestral populations (K), and the Y-axis indicates cross-validation errors. (C) Admixture population structure bar plots for K = 1, 2 and 3. Colors represent different ancestral components; the Y-axis shows the proportion of each component. (D) Kinship heatmap of all 206 experimental individuals generated using genome-wide SNPs.

Fig. 7.

GWAS results identifying genomic variants (SNPs and Indels) associated with concentrations of six serum cytokines. Manhattan plots (left) show −log10(P) for each SNP and Indel across chromosomes. The gray horizontal line indicates the significance threshold (−log10(P) = 6). Variants above this line were considered significantly associated. Q‑Q plots (right) show observed versus expected −log10(P) values. The genomic inflation factor (λ) is shown in each panel. All λ values were within the normal range (0.97–1.03), indicating no systematic inflation of test statistics. (A‑B) IFN‑γ; (C‑D) IL‑1β; (E‑F) IL‑6; (G‑H) IL‑17; (I‑J) IL‑22; (K‑L) TNF‑α.

A total of 58 significant SNPs and Indels were identified to be associated with serum cytokine traits. Twelve significant loci were related for IFN‑γ on chromosomes 2, 6, 7, and Z. Chromosome 7 harbored the maximum number of significant loci, yet no candidate genes were annotated in these genomic intervals (Fig. 7A–B). Fourteen significant loci linked to IL‑1β were detected on chromosomes 2, 3, 6, and 18; chromosomes 3 and 6 contained the most significant loci, with no candidate genes annotated within these regions (Fig. 7C–D). Eleven loci associated with IL‑6 were distributed on chromosomes 1, 5, 6, 7, and 10 (Fig. 7E–F). Eleven significant loci associated with IL-17 were associated on chromosomes 1, 3, 5 and 14 (Fig. 7G–H). Two genome-wide significant loci were found for IL‑22 (Fig. 7I–J). Eight significant loci were associated with TNF-α. A genomic region with clustered loci was identified on chromosome 4, and no genes were annotated in this interval (Fig. 7K–L). Notably, no shared loci or genes were associated among these six cytokines. Detailed information of loci annotated with candidate genes is summarized in Table 4, while loci without gene annotation are listed in Supplementary Table 2.

Table 4.

Genomic loci (SNPs and Indels) associated with serum concentrations of six chicken cytokines identified via GWAS.

Trait	CHROM	POS	Allele	SNP	Effect	P value	Gene	Functional region
IFN-γ	chrZ	2509490	G/A	rs15002007	15.120	4.87e-08	EPG5	upstream gene variant
	chrZ	8575766	C/G	-	-23.768	8.78e-07	GALT-CNTFR	intergenic_region
IL-1β	chr2	33785080	C/T	rs313465756	8.064	3.73 e-07	SCRN1	intron variant
	chr6	28952820	A/G	rs16561035	-6.432	6.22 e-07	ATRNL1	intron variant
	chr6	28955425	A/C	rs735758099	-6.251	6.39 e-07	ATRNL1	intron variant
IL-6	chr1	12849130	T/C	rs1060363658	1.764	3.54 e-07	PTPN12	intron variant
	chr1	12856384	C/T	rs731646092	1.988	3.48 e-07	PTPN12	intron variant
	chr1	12856411	C/A	rs3384382188	1.939	6.23 e-07	PTPN12	intron variant
	chr1	15495882	G/A	rs15193762	-1.897	2.69 e-07	ABCD2	intron variant
	chr7	20872911	C/T	rs734144895	1.837	6.85 e-07	SLC4A10	intron variant
	chr7	20875207	C/CT	-	1.875	7.23 e-07	SLC4A10	intron variant
	chr7	32749724	TA/T	-	1.672	4.48 e-07	GTDC1	intron variant
	chr10	12242169	G/A	rs316438708	-1.653	1.66 e-07	MESDC1-MESDC2	intergenic_region
IL-17	chr1	9894265	A/G	rs740100375	4.538	9.15 e-08	SEMA3E-PCLO	intergenic_region
	chr5	15503126	C/A	rs737805936	2.361	7.12 e-07	RASSF7	intron variant
	chr5	15503131	G/T	rs741350209	2.403	3.98 e-07	RASSF7	intron variant
	chr5	15503134	C/T	rs731095050	2.368	7.10 e-07	RASSF7	intron variant
	chr5	46125894	C/T	rs15720728	-2.722	4.81 e-07	VRK1	upstream_gene_variant
	chr5	55488502	G/A	rs315555895	-2.441	7.00 e-07	TMEM260	intron variant
	chr5	55488542	G/A	rs15741542	-2.402	3.30 e-07	TMEM260	intron variant
	chr5	55488543	C/T	rs15741544	-2.402	3.30 e-07	TMEM260	intron variant
	chr5	55488546	A/T	rs15741546	-2.402	3.30 e-07	TMEM260	intron variant
	chr14	8095372	C/G	rs738875761	4.972	8.55 e-09	XYLT1	intron variant
IL-22	chr1	45727074	T/C	rs731069417	3.909	9.99 e-08	AMDHD1	intron variant
	chr4	2939863	T/C	rs13641014	2.880	4.03 e-07	PLS3	intron variant
TNF-α	chr2	48133746	T/C	rs736359839	7.272	2.58 e-07	PDE1C	Intron variant
	chr8	2552204	G/A	rs731120491	-6.828	6.34 e-07	CRB1	intron variant
	chr8	16102782	G/A	rs16633398	-7.062	9.60 e-07	MCOLN3	intron variant

Alleles are presented as reference allele (REF)/alternative allele (ALT). Effect sizes were calculated based on an additive genetic model. The alternative allele (ALT) was defined as the effect allele. A positive effect means the ALT allele correlates with higher phenotypic values of the cytokine trait, while a negative effect corresponds to lower trait values.

Analysis of favorable genotypes for serum cytokines at GWAS-significant genomic loci in chickens

To screen advantageous genotypes at key cytokine-associated loci, we performed Kruskal-Wallis H tests to compare serum cytokine levels among chickens carrying distinct genotypes at the 58 GWAS-significant associated loci. The results revealed that 55 of these loci showed no statistically significant differences in cytokine levels across different genotypes. Only three loci displayed obvious genotype-specific differences in cytokine concentrations, suggesting the existence of elite advantageous genotypes (Fig. 8). Specifically, at the IFN-γ-associated locus chrZ:8575766, chickens with the CC genotype exhibited significantly higher IFN-γ concentrations compared with GG genotype carriers (P=0.0092). At the IL-1β-associated locus chr2:33785080, chickens carrying the CT genotype had higher IL-1β concentrations than those with the CC genotype (P=0.0214). At the IL-6-associated locus chr10:12242169, chickens with the GG genotype exhibited higher IL-6 concentrations than AA homozygotes (P=0.0072). Chickens carrying advantageous genotypes tend to have elevated baseline serum cytokine levels.

Fig. 8.

Box plots display serum cytokine concentrations across distinct genotypes. The x-axis presents genotype categories, and the y-axis indicates cytokine concentration. (A) IFN‑γ levels among genotypes at chrZ:8575766. The CC genotype (n = 176) had higher IFN‑γ levels than the CG genotype (n = 20) (P = 0.0092). (B) IL‑1β levels among genotypes at chr2:33785080. The CT genotype (n = 83) had higher IL‑1β levels than the CC genotype (n = 113) (P = 0.0214). (C) IL‑6 levels among genotypes at chr10:12242169. The GG genotype (n = 63) had higher IL‑6 levels than the AA genotype (n = 39) (P = 0.0072). Only loci with significant phenotypic differences among genotypes are displayed. Statistical comparisons were conducted via Kruskal‑Wallis tests (* P < 0.05, ** P < 0.01).

GO and KEGG enrichment analyses of GWAS-derived candidate genes linked to chicken serum cytokines

GO functional annotation and KEGG pathway enrichment analyses were performed on candidate genes significantly correlated with the six serum cytokine phenotypes in chickens (Figs. 9 and Supplementary Figure 1). The enrichment findings revealed that these candidate genes were mainly enriched in immune and inflammatory response pathways, which is consistent with the essential biological roles of cytokines in host immune modulation. Specifically, the candidate genes associated to IFN‑γ were significantly enriched in terms related to cytoskeletal structures (keratin filaments, intermediate filaments), innate immune signaling pathways (e.g., Toll‑like receptor 9 signaling and double‑stranded DNA response), as well as immune defense pathways including Staphylococcus aureus infection and autophagy. Candidate genes associated with IL‑1β were predominantly enriched in signal transduction and inflammatory response biological processes, including Notch binding activity, signal transduction, and response to stimuli, with prominent enrichment in membrane-associated cellular components. Candidate genes linked to IL-6 were functionally enriched in transmembrane transport and ion homeostasis, such as solute inorganic anion antiporter activity and transmembrane transporter activity, and participated in pathways including ABC transporters, peroxisome function, and ribosome biogenesis. Candidate genes associated with IL‑17 exhibited highly specific enrichment in glycosaminoglycan and proteoglycan metabolic pathways, including heparan sulfate and chondroitin sulfate biosynthesis, as well as xylosyltransferase activity. Candidate genes associated with IL‑22 were exclusive significantly enriched in the histidine metabolism pathway, which is closely associated to histidine catabolism and formate metabolism. Candidate genes associated with TNF‑α were enriched in calcium signaling pathways, cyclic nucleotide metabolism and cytoskeletal remodeling, along with physiological pathways such as taste transduction and renin secretion.

Fig. 9.

GO and KEGG pathway enrichment analyses of candidate genes linked to serum cytokines. The x-axis represents the enrichment factor or number of enriched genes. The y-axis lists enriched biological pathways and functional terms. The color gradient corresponds to statistical significance quantified as (−log10(P‑value)). Only pathways with P < 0.05 are visualized. The enriched pathways were mainly involved in immune defense, inflammatory response, and metabolism. (A) GO enrichment for IFN‑γ‑related genes; (B) KEGG enrichment for IFN‑γ‑related genes; (C) GO enrichment for IL‑1β‑related genes; (D) GO enrichment for IL‑6‑related genes; (E) KEGG enrichment for IL‑6‑related genes. Enrichment results for IL‑17, IL‑22, and TNF‑α are provided in Supplementary Figure 1.

Analysis of candidate gene expression via qPCR for verifying chicken serum cytokine GWAS results

To experimentally validate the GWAS analytical results, we quantified the mRNA expression of candidate genes (CNTFR, SCRN1, and MESDC2) in cytokine high-concentration and low-concentration subgroups via quantitative real-time PCR (qPCR). Grouped serum cytokine concentrations were listed as follows: 84.41 ± 2.6 pg/mL, low- IFN‑γ group: 62.95 ± 3.62 pg/mL; high- IL‑1β group: 127.03 ± 15.11 pg/mL; low- IL‑1β group: 82.76 ± 6.52 pg/mL; high- IL‑6 group: 37.87 ± 7.34 pg/mL; low- IL‑6 group: 24.35 ± 1.74 pg/mL. Six biological replicates were included per group. The results indicated that the transcript abundance of all three candidate genes was markedly elevated in the high-cytokine groups relative to their corresponding low-cytokine counterparts (Fig. 10). The consistent trends between phenotypic cytokine levels and gene expression offer preliminary validation for the GWAS outcomes, confirming the genetic association of these three genes with circulating serum cytokine concentrations in chickens. Moreover, these three trait-related genomic loci could be repeatedly identified under multiple statistical models (FarmCPU, GLM and MLM models, Supplementary Table 3). Collectively, qPCR experimental verification combined with cross-model reproducibility supplies multiple independent lines of evidence to verify the reliability of candidate genes screened in the current research.

Fig. 10.

High and low groups were selected based on extreme phenotypic values (Wilcoxon test, **P < 0.01, ***P < 0.001). The x-axis shows the high and low groups, and the y-axis displays phenotypic values. Phenotypic levels were significantly different between groups. (A) IFN‑γ; (B) IL‑1β; (C) IL‑6. Relative mRNA expression of (D) CNTFR, (E) SCRN1, and (F) MESDC2 in high and low cytokine groups. The x-axis represents grouping information, and the y-axis refers to relative gene expression levels. Relative mRNA expression differed significantly between groups. Data were normalized against the low group (value set as 1.0) with the 2^−ΔΔCt method; β-actin served as the internal reference. Error bars denote SD. **P < 0.01, ***P < 0.001 (unpaired two-tailed Student's t-test or Welch's t-test).

Discussion

Coordinated regulation of serum cytokines under physiological homeostasis and batch effect correction

Cytokines are mainly synthesized and secreted by immune cells (macrophages, T lymphocytes, NK cells, etc.) in response to exogenous or endogenous stimuli (Kaminska et al., 2024). In the present study, strong positive correlations were detected among IFN‑γ, IL‑1β, IL‑6, IL‑17, IL‑22 and TNF‑α, with correlation coefficients ranging from 0.89 to 0.91. Under healthy steady-state conditions, pro-inflammatory and anti-inflammatory cytokines usually exhibit positive co-regulation. This is fundamentally different from the "cytokine storm" or the "Th1/Th2 seesaw" pattern seen in disease states. In disease conditions, immune responses often undergo a Th1-to-Th2 polarization or a pro-inflammatory to anti-inflammatory shift. This is typically reflected by a relative decrease in IFN-γ (a Th1 marker) and a relative increase in IL-4/IL-10 (Th2 markers) (Hong et al., 2013). But IL-10 and IFN-γ may have a complex dynamic regulatory relationship (Magombedze et al., 2015), rather than a simple linear inverse relationship. In healthy individuals, cytokines act as key mediators of immune homeostasis, maintaining immune surveillance without triggering overt inflammation (Nirschl et al., 2017). For example, IFN-γ is not merely a pro-inflammatory effector; it also plays an essential instructive role in tissue immune homeostasis under physiological conditions (Nirschl et al., 2017). Similarly, pro-inflammatory cytokines such as IL-1β, IL-6, and TNF-α act as sentinels that alert the immune system to potential threats, while operating within a tightly regulated homeostatic network(2019). In poultry, studies on broilers have shown that multiple immune indices, including IFN-γ and the NF-κB pathway (which regulates IL-1β and IL-6), increase coordinately during normal development (Song et al., 2021). Therefore, the positive correlations observed among these cytokines in our healthy cohort likely reflect coordinated regulation under steady-state conditions. We hypothesize the strong correlations (0.89–0.91) among these six cytokines arise from shared upstream signaling regulation or overlapping cellular sources of immune cells responsible for their production. Furthermore, serum cytokines were quantified at a unified time point in our study, yet batch exerted a significant effect on cytokine concentrations. This phenomenon may be explained by divergent microbial exposure environments across different rearing batches. Accordingly, all subsequent analyses in the present study were adjusted to account for multi-batch environmental effects. Each ELISA plate was equipped with standard curve and quality control samples. The coefficient of variation (CV) of all samples was less than 15%, indicating favorable technical repeatability.

Cytokine‑associated genes differ across species and breeds

A total of 20 putative candidate genes, including SCRN1, GALT and CNTFR, and MESDC1 and MESDC2, were annotated based on the GWAS results. The candidate genes detected in this research differ from those reported in previous relevant studies. For instance, prior GWAS of Jinghai yellow chickens confirmed MYOM1 (myomesin 1) as a key candidate gene significantly linked to IFN‑γ concentrations (Wang et al., 2022). In humans, loci/genes such as PDGFRB, ABO (Nath et al., 2019; Sliz et al., 2019), NAA35‑GOLM1 (Li et al., 2016a), and the TLR1‑TLR6‑TLR10 locus (Li et al., 2016b) have been reported to be associated with cytokine production. Such inconsistencies are mainly attributable to the breed-specific genetic features of this complex quantitative trait. Notably, no overlapping candidate genes were uncovered across the six cytokines in our study, a phenomenon explained by immune cell lineage specificity. Different cytokines are synthesized by distinct immune cell subsets, and their genetic regulation is predominantly controlled by lineage-specific cis-regulatory elements (Kalim et al., 2024; Qi et al., 2024). At the level of expression regulation, coordinated covariation of cytokines derived from different immune cells may be closely related to their shared key transcription factors and signaling pathways. NF-κB, an ubiquitously expressed transcription factor, can directly induce the transcription of multiple pro-inflammatory cytokines, including IL-1β, IL-6, and TNF-α (Ma et al., 2024; Moneva-Sakelarieva et al., 2025). When the NF-κB pathway is activated, these cytokines show coordinated expression (Sun et al., 2022). Meanwhile, the JAK-STAT pathway serves as a common hub for signaling of cytokines such as IL-6, IL-10, and IFN-γ (Oliveira et al., 2025). Its activation can simultaneously regulate the expression of multiple cytokine genes (Mogensen, 2013). Critically, there is direct crosstalk between NF-κB and JAK-STAT pathways: STAT3 can activate NF-κB, and NF-κB can induce IL-6 expression, which in turn feeds back through the JAK-STAT pathway to form a positive regulatory loop (Jarnicki et al., 2010; Zhao and Sanyal, 2022). Therefore, the lineage-specific origins of different cytokines do not prevent them from exhibiting proportional co-variation across individuals. This co-variation reflects systematic differences in the overall activation state of upstream regulatory networks among individuals. Meanwhile, the crosstalk between NF-κB and JAK-STAT pathways operates at low levels under healthy steady-state conditions and is tightly constrained by negative feedback (Harhaj and Shembade, 2020; Herrera-Uribe et al., 2025). This maintains baseline cytokine expression and immune surveillance, rather than triggering uncontrolled inflammatory amplification.

Immune and metabolic pathway enrichment of cytokine candidate genes

Most enriched pathways identified in this study were associated with immune defense against pathogens and inflammatory responses, highlighting their essential roles in cytokine synthesis and homeostasis maintenance. KEGG pathway enrichment analysis revealed that IFN‑γ-related candidate genes were significantly enriched in immune defense pathways, including Staphylococcus aureus infection and autophagy. Studies have shown that in a mouse model of pulmonary S. aureus infection, genes such as Eif2ak2, Ikbke, and Nfkbiz participate in the infection process by promoting autophagy and are key immune regulatory genes (Zha et al., 2025). This functional pattern matches the biological property of IFN‑γ as a macrophage-activating factor. Macrophage studies have verified that S. aureus infection triggers activation of the JAK-STAT signaling pathway (Zhu et al., 2015). IFN-γ is a key upstream signal of this pathway (Liu et al., 2025). CNTFR encodes the ciliary neurotrophic factor receptor, which belongs to the type I cytokine receptor family and is a known component of the JAK-STAT pathway (KEGG: hsa04630). As a candidate gene for the IFN-γ phenotype, CNTFR may mediate pathway crosstalk between the JAK-STAT cascade and the S. aureus infection pathway. IFN-γ induces autophagy, a vital host mechanism for eliminating intracellular S. aureus (Brauweiler et al., 2016). Upon binding of its ligand CNTF (ciliary neurotrophic factor) to the candidate gene CNTFR, intracellular STAT3 protein is activated and participates in the regulation of autophagy (Peterson et al., 2000; Yong et al., 2022). When STAT3 activity is pharmacologically inhibited prior to treatment, the autophagy-inducing effect of CNTF is markedly attenuated (Yong et al., 2022). These findings provide strong evidence that CNTF/CNTFR regulates autophagy through STAT3. Unlike IFN-γ-related genes enriched in immune cascades, candidate genes correlated with IL-6 were primarily enriched in metabolic pathways including ABC transporters, peroxisome, and ribosome biogenesis. ABC transporters mediate the transmembrane transport of diverse substrates such as lipids and metabolites (Albrecht and Viturro, 2007; Neumann et al., 2017). The peroxisome is the primary site of fatty acid β‑oxidation (Wanders et al., 2001; Wanders et al., 2020), and ribosome biogenesis is directly linked to the rate of protein synthesis (Kaczanowska and Rydén-Aulin, 2007; Leiva et al., 2023; Ni and Buszczak, 2023). This enrichment pattern suggests that the genetic regulation of IL‑6 may be achieved more through influencing cellular metabolic status than through directly regulating immune effector molecules. MESDC2, along with MESDC1, is a well-documented modulator of the Wnt signaling pathway (Tran et al., 2021). The Wnt pathway has clear functions in regulating cellular metabolism. For example, Wnt/β-catenin signaling can promote glycolysis by activating mTORC2 (Esen et al., 2013). It is also involved in the regulation of glutamine oxidation and fatty acid oxidation (Karner and Long, 2017; Moorer and Riddle, 2018). In addition, β-catenin, an effector molecule of Wnt signaling, can directly interact with transcription factors that regulate lipid metabolism, such as PPARγ and FOXO (Almeida et al., 2009; Manolagas, 2010). These interactions enable direct crosstalk between signaling pathways and metabolism. The Wnt pathway regulates cellular metabolism. Genetic variants within the MESDC1 and MESDC2 locus may affect Wnt pathway activity. This could change cellular metabolic states. Through the metabolism–immune crosstalk network, these changes may ultimately influence IL-6 expression levels. At the molecular mechanism level, complex cross-regulation exists between between IFN‑γ and IL‑6. IFN‑γ can directly induce neutrophils to produce IL‑6 via the JAK‑STAT signaling pathway (Yoshida et al., 2021). Moreover, in mouse S. aureus infection, the protective effect of IFN‑γ depends on the sequential activation of the TLR2‑IL‑6‑Mincle signaling axis (Matsumura et al., 2019).

Candidate genes and molecular pathways regulating major cytokines

For IFN‑γ level phenotype, the candidate genes annotated in this study included GALT and CNTFR. According to the UniProt/GO database annotations for chickens, CNTFR exhibits cytokine receptor activity and localizes to the cell membrane. Experiments on mouse microglial cells confirm functional synergy between CNTFR signaling and IFN‑γ signaling (Lin et al., 2009). Studies have shown that CNTFRα expression is significantly induced by IFN‑γ, and the complex of CNTF with soluble CNTFRα can synergize with IFN‑γ to enhance CD40 expression and Cox‑2 protein expression. These collective findings support CNTFR as a credible candidate gene underlying the IFN‑γ phenotype (Lin et al., 2009). CNTFR belongs to the IL‑6 family cytokine receptor network. This network has known functional synergy with IFN‑γ. Both pathways communicate through JAK‑STAT crosstalk (Costa-Pereira, et al., 2002). IFN‑γ mainly signals through the JAK‑STAT1 pathway (Owen et al., 2019). CNTF‑CNTFR signals mainly through the JAK‑STAT3 pathway (Jiao et al., 2003; Kaur et al., 2002). Studies have shown that STAT1 and STAT3 pathways regulate each other (Avalle et al., 2012). This crosstalk means that differences in CNTFR expression may affect the cellular response threshold to IFN‑γ signaling, possibly through changes in STAT3 activity. For example, in the absence of STAT3, IL‑6 can induce a response similar to IFN‑γ. This is characterized by persistent STAT1 activation and upregulation of IFN‑γ‑induced genes (Costa-Pereira et al., 2002). In contrast to IFN‑γ, the candidate gene annotated for IL‑1β was SCRN1 (secernin 1). SCRN1 encodes a member of the secernin family, and its rat homolog participates in the modulation of mast cell exocytosis (Way et al., 2002). SCRN1 is widely expressed in humans, rats, and mice (Yu et al., 2014). Exocytosis constitutes a core process through which inflammatory cells release cytokines and chemokines. SCRN1 may modulate the secretion and biological function of IL‑1β by regulating vesicular transport and exocytosis efficiency. For IL‑6, the candidate genes identified were MESDC1 and MESDC2. In mammals, MESDC2 participates in the regulation of Wnt signaling, which undergoes extensive crosstalk with IL‑6 during inflammation and tissue homeostasis (Edara et al., 2020; Kim et al., 2024; Spanjer, et al., 2022). A study published in Biomedicines in 2020 reported that β‑catenin affects IL‑6 expression. β‑catenin is a key effector of Wnt signaling. This finding was shown in activated human astrocytes (Edara et al., 2020). The study also revealed crosstalk between β‑catenin and the NF‑κB pathway. NF‑κB is a classical upstream activator of IL‑6 (Edara et al., 2020). In human synoviocytes, IL‑6 alone or in combination with TNF‑α significantly inhibited Wnt3a‑induced Wnt signaling (Malysheva et al., 2016). This creates a bidirectional regulatory relationship between the two pathways. Inflammatory cytokines, such as IL‑6 and TNF‑α, can also upregulate Wnt signaling through the NF‑κB pathway (Chen et al., 2022). This further indicates that the crosstalk between the two pathways is multi‑layered and bidirectional. As pivotal regulators of Wnt signaling, genetic variants within MESDC1 and MESDC2 may modulate IL-6 expression by altering the magnitude of Wnt pathway activity (Stafford and Nachbur, 2016; Thoudam et al., 2016).

Limitations of the study

Although we annotated core genes tightly correlated with serum cytokine concentrations in chickens in the present study, several limitations of this work need to be addressed. First, the results were obtained from a study population consisting of 206 indigenous chickens, which is a limited sample size. Large-cohort verification experiments using expanded chicken populations are required for follow-up research. All samples in this study were collected from healthy birds at a single time point (10 weeks of age). Therefore, our findings are restricted to baseline cytokine levels at this specific age and health status. Immune parameters may show different dynamics at other production stages (e.g., brooding or laying periods) or under actual pathogen challenges. Thus, extrapolating these results directly to long‑term immune potential or clinical disease resistance should be done with caution. Second, the key genes associated in this study have not been experimentally validated. Subsequent studies should employ approaches such as gene editing to conduct in‑depth investigations of gene function. Lastly, the regulatory mechanisms governing serum cytokine phenotypes in chickens are intricate and demand more comprehensive research. Multi-omics approaches including genomics, proteomics and metabolomics can be applied to further decipher the interactive mechanisms between host genes, proteins and functional metabolites.

Conclusion

We detected prominent inter-individual differences in serum IFN‑γ, IL‑1β, IL‑6, IL‑17, IL‑22 and TNF‑α levels among chickens. Via GWAS analysis, 58 genome-wide significant loci were uncovered for these six cytokine traits. GALT and CNTFR, SCRN1, and MESDC1 and MESDC2 were annotated as candidate genes responsible for the serum phenotypic variation of IFN‑γ, IL‑1β, and IL‑6, respectively. These findings characterize the genetic architecture of circulating cytokines in Qiandongnan Xiaoxiang chickens and support genomic breeding programs focused on the improvement of immune performance.

Author Contribution

Hongfa Zhang: Performed data analysis and visualization, drafted the original manuscript, revised the manuscript, reviewed and approved the final version of the manuscript, and agreed to be accountable for all aspects of this work. Yongxian Yang: Constructed the experimental animal population, and supervised animal feeding, sample collection and phenotypic measurement, revised the manuscript for important intellectual content, reviewed and approved the final version of the manuscript, and agreed to be accountable for all aspects of this work. Liqi Wang: Completed phenotypic measurement and sample collection, conducted genomic DNA extraction and data preprocessing, revised the manuscript for important intellectual content, reviewed and approved the final version of the manuscript, and agreed to be accountable for all aspects of this work. Zhong Wang: Conceived the study, designed the research scheme, supervised the whole experimental work, revised and edited the final version of the manuscript, reviewed and approved the final version of the manuscript, and agreed to be accountable for all aspects of this work.

Funding

This research was funded by the National Natural Science Foundation of China (32260829 and 32160853), the Guizhou Provincial Science and Technology Project (QKH-ZK2022-113 and QKH-ZC2022-key34).

Ethics statement

The Animal Experiment Management and Use Committee of Guizhou University (Guiyang, China) approved this study’s animal protocol (Project No.: EAE-GZU-2022-T050). All procedures were carried out in compliance with the Regulations on the Management and Use of Laboratory Animals issued by the Ministry of Agriculture and Rural Affairs of the People’s Republic of China.

Disclosures

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

Data availability

• The whole-genome resequencing data generated for this study have been deposited in the CNGB Sequence Database (CNSA) of the China National Gene Bank (CNGBdb) and are accessible under accession number CNP0008107 (https://db.cngb.org/).