Integorated multi-omics reveals the NOS2-metabolite-microbiota regulatory axis underlying feed efficiency in laying hens

Abstract

Feed efficiency (Feed conversion ratio, FCR), a critical economic indicator for poultry breeding, is closely associated with the sustainable advancement of the poultry production. However, the molecular mechanisms governing differences in FCR among chickens have not yet been fully elucidated. Transcriptome analysis identified 189 differentially expressed genes, among which NOS2 was identified as a key gene significantly involved in the arginine biosynthesis pathway. Following NOS2 interference for 48 h in primary hepatocytes, CAT, Nrf2, FASN were downregulated, while overexpressed NOS2 showed an opposite tendency. Metabolomics analysis identified 116 and 87 differential metabolites in the liver and cecum contents, respectively. Metagenomics analysis identified s__Methanocorpusculum_vombati as the dominant taxa for LFCR. Integrated analysis revealed that NOS2 expression was positively correlated with 23 liver metabolites. Combined analysis found that genus Oscillibacter as a central hub microbiota, significantly connecting to 2-aminobicyclo[3.1.0]hexane-2,6-dicarboxylic acid in the cecum content. This study elucidates the regulatory role of NOS2 to antioxidant, lipogenesis and inflammatory of chicken hepatocyte, and predicted that NOS2 gene as a central regulator to improve feed efficiency in laying hens.

Introduction

Laying hens provide a cost-effective source of high-quality animal protein for human consumption, including eggs and meat. However, feed accounts for more than 60% of the total cost of poultry production (Arango, 2009; Wilkinson, 2011), making feed conversion ratio (FCR) as a key determinant of the sustainability and profitability of the laying hen industry (Zhang et al., 2024b; Zhou et al., 2023). The heritability of FCR falls within the range of 0.07 to 0.41 (Sell-Kubiak et al., 2017), reflecting the synergistic effects of host genetics, metabolic homeostasis, and intestinal microbial community (Zhang et al., 2024b; Yuan et al., 2015b; Yuan et al., 2015a; Zhou et al., 2023; Clark et al., 2019). Therefore, a comprehensive multi-omics investigation of FCR in laying hens will facilitate precision nutritional strategies, genetic improvement, and eventually benefit economy (Gao, Zheng and Xu, 2025c)

The liver is the central metabolic organ responsible for converting dietary nutrients into energy, protein, and lipid required for egg production (Van Eck et al., 2023; Yin et al., 2023; Guo et al., 2025), and is closely associated with FCR in chickens (Gao et al., 2025a; Gao et al., 2025c; Zhang et al., 2026). Chickens with different FCR values exhibit substantial disparate in hepatic gene expression, lipid metabolism, and hormone secretion (Zhang et al., 2026; Alexandre et al., 2015; Jin et al., 2019; Reyer et al., 2017). Specifically, chickens with high feed efficiency displayed lower hepatic lipid deposition and abdominal fat, but higher breast muscle yield (Gao et al., 2025a; Zou et al., 2023). In addition, hepatic lipid accumulation, lipogenesis, and metabolism have been identified as factors significantly associated with feed efficiency in chicken (Gao et al., 2025a; Gao et al., 2025c). More recently, an integrated multi-omics analysis combined with gene interference and overexpression experiments in hepatocytes, identified a hepatic lipid deposition gene of propionyl-CoA carboxylase beta subunit (PCCB), as an important regulator to feed efficiency (Zhang et al., 2026). Existing evidence from dietary nutritional additive trials reveals that optimized antioxidant status contributes to higher feed efficiency (Pirgozliev et al., 2019; Su et al., 2016; Russell et al., 2016). Nevertheless, the specific molecular regulatory network responsible for this antioxidant-mediated feed efficiency improvement has not yet been characterized at the genetic level. As the genetic determinants of FCR vary considerably among chicken breeds, breed-specific investigations are essential for genetically improving feed efficiency in native breeds, such as Wenshang Barred hens.

The gastrointestinal tract (GIT) plays an essential role in digestion and nutrient absorption, with the cecum harboring the most abundant microbial community in mature birds and exerting profound effects on host physiological status as well as production performance (Huang et al., 2021; Stanley, Hughes and Moore, 2014). Cecal microbiota not only ferment dietary crude fiber and carbohydrates into short-chain fatty acids (SCFAs) to facilitate the host metabolic performance (Gabriel et al., 2006), but also interact with host genetic factors, leading to efficient feed utilization in chickens (Huang et al., 2021; Zhang et al., 2024b; Xiong et al., 2024; Wen et al., 2021). The liver-cecum axis, which connects the central metabolic organ with the primary site of microbial fermentation, plays a crucial role in regulating feed efficiency through liver metabolism, host-microbiota interaction, and the maintenance of host immune homeostasis (Bai et al., 2025; Zhang et al., 2024b).

Therefore, our study will systematically elucidate the molecular mechanisms and regulatory networks underlying FCR in Wenshang Barred hens by integrating liver transcriptomics and metabolomics with cecal content metabolomics and metagenomics. Our findings provide new insights into the molecular basis of feed efficiency and establish a theoretical foundation for enhancing the feed utilization through molecular breeding and nutritional interventions.

Materials and methods

Experimental design

In this study, 273 Wenshang Barred hens with comparable body weights were housed at the Breeder Chicken Farm of Dongying Base, Shandong Academy of Agricultural Sciences. All experimental chickens were raised under identical environmental conditions with free access to feed and water. Data including egg production, egg weight, and daily feed intake were monitored over a 32-day period (52-57 weeks of age). A corn-soybean meal basal diet formulated according to the Chinese Feed Ingredients and Nutritional Values Table (32nd Edition, 2021) and the Chicken Feeding Standards (NY/T 33-2004) was used (Table S1).

Serum biochemical, and hormones assays

Blood samples were obtained from the wing vein of experimental chickens to determine 8 serum biochemical indices by automated biochemical analyzer (mindray, Shenzhen, China), including total cholesterol (TC, 105-000483-00), total protein (TP, 105-000486-00), aspartate transaminase (AST, 105-000478-00), alanine aminotransferase (ALT, 105-000477-00), triglyceride (TG, 105-000484-00), high-density lipoprotein cholesterol (HDL-C, 105-000498-00), Low-Density Lipoprotein (LDL-C, 105-000499-00) and glucose (GLU, 105-000495-00). All the serum hormone indices, including estradiol (E2, JYM0049Ch), follicle-stimulating (FSH, JYM0055Ch), luteinizing hormone (LH, JYM0032Ch), growth hormone (GH, JYM0053Ch), and insulin (JYM0002Ch) were bought from Colorful-Gene Biotech Co., Ltd. (Wuhan, China).

Transcriptome sequencing

RNA was extracted from 12 liver samples in a high FCR (HFCR, n = 6) and a low FCR (LFCR, n = 6). After confirming the RNA concentration and purity (NanoDrop2000, ThermoFisher, MA, USA), quality (Agilent5300, Agilent, CA, USA) as well as integrity (agarose gel electrophoresis) of liver tissue (Table S3), Illumina Novaseq™ 6000 (Illumina, CA, USA) were used for sequencing. Low-quality sequences were filtered by using fastp (v0.23.4) (Chen et al., 2018). The cleaned reads were aligned to bGalGal1.mat.broiler.GRCg7b by using Hisat2 (v2.2.1) (Kim et al., 2019) and assembled by StringTie (v2.2.1) (Pertea et al., 2016). Gene expression levels were measured using transcripts per million reads (TPM) values. Differential expression analysis was performed using DESeq2 (v1.42.0) (Love, Huber and Anders, 2014). Differentially expressed genes (DEGs) were identified with |log₂FoldChange| > 1 and p < 0.05. The soft threshold at 14 when using WGCNA (Langfelder and Horvath, 2008) to analysis the DEGs (Fig. S2).

Real-time fluorescence quantitative PCR

Seven DEGs (ASL1, CYP26B1, GREM2, WNT11, NOS2, IL4I1, and TGM4) were randomly chosen for qRT-PCR (Table S5). To assess the functional consequences of NOS2 manipulation in chicken primary hepatocytes, the mRNA expression of genes involved in antioxidant defense (CAT, SOD, Nrf2, GPX1), lipogenesis (SREBF1, SCD1, ACCα, FASN), and inflammatory response (IL-6, TNF-α, IL-10, HO-1) was determined following NOS2 knockdown or overexpression. Primers (Table S5) for the genes were synthesized by Sangon Biotech Co., Ltd. (Shanghai, China).

RealStar Fast SYBR qPCR mix (GenStar, #A301 Beijing, China) on a Roche fluorescence quantitative PCR platform was used for qRT-PCR. Data were processed using the 2^-△△Ct method with β-actin as the endogenous control.

Metabolites extraction and sequencing

Liver and cecal content samples from HFCR (n = 6) and LFCR (n = 6) group were used for untargeted metabolomics by using LC-MS/MS at Majorbio Bio-Pharm Technology Co., Ltd. (Shanghai, China) with same analysis process and threshold (p < 0.05, VIP >1) of our previous study (Ren et al., 2022; Hao et al., 2026).

Metagenomic sequencing

DNAs were extracted from the cecal content from the HFCR and LFCR group (n = 6 in each group) by using FastPure Stool DNA Isolation Kit (MJYH, shanghai, China). After DNA quality evaluation (Table S7, Fig. S1), paired-end sequencing was carried out on the Illumina NovaSeq™ X Plus (Illumina Inc., San Diego, CA, USA) through Majorbio Bio-Pharm Technology Co., Ltd. (Shanghai, China).

Raw reads were trimmed of adapters and low-quality reads by using fastp (v0.23.4). Contigs (≥300 bp) were assembled from raw reads by MEGAHIT (v1.2.9) (Li et al., 2016), followed by Prodigal (v2.6.3) (Hyatt et al., 2010) to predicate ORF, and CD-HIT (v4.6.1) to generated a non-redundant gene set clustering at 90% identity and coverage (Fu et al., 2012).

Primary hepatocytes isolation

Liver tissue was isolated from 14-day-old chicken embryos (E14), washed 3 to 5 times with d-Hanks solution at room temperature, then minced the liver tissues into small pieces by using sterilized scissor. After digested with 0.25% collagenase IV at 37°C for 30 min, and then filtered through 200-mesh screens (Solarbio, Beijing, China) and 70-μm cell strainer (nylon mesh, Cat. No. BS-70-CS, Biosharp), respectively. A 2-ml aliquote of cell suspension was carefully layered onto 5 mL of a percoll gradient solution prepared with percoll, 10 × PBS, and 1 × PBS at a volumetric ratio of 27: 3:20, then centrifugation at 3000 rpm for 15 min. Finally, resuspended in Williams’E complete medium (5% FBS, 4 nM dexamethasone, 1% penicillin-streptomycin, 2 nM l-glutamine, and cultured at 37 °C with 5% CO₂ (Fig. 1).

Fig. 1.

Primary hepatocytes isolation from 14-day-old chicken embryos (E14).

Cell transfection

Three siRNAs of NOS2 gene were purchased from GenePharma (Shanghai, China) to inhibit the expression of NOS2 in primary hepatocytes, including siRNA-NOS2-1, siRNA-NOS2-2 and siRNA-NOS2-3 (Table S5). A NOS2 cDNA fragment of 3477 bp were constructed into a plasmid of pcDNA3.1(+)eGFP and digested with BamHI and NotI to confirm the fragment of NOS2. The plasmid was named pcDNA3.1(+)eGFP-NOS2 which provided by GENCEFE Biotech (Wuxi, China). Cells were transfected siRNAs or plasmids by using TransIT-X2 Dynamic Delivery System (TransIT-X2).

Statistical analyses

Pearson’s correlation analysis was used to assess mRNA–metabolite and metabolite–microbiome correlations with false discovery rate (FDR) controlled by the Benjamini-Hochberg procedure. Student’s t-test were employed for group comparisons. Data are presented as means ± SD, p < 0.05 considered statistically significant.

Results

Serum biochemical parameters between LFCR and HFCR

Egg-laying rate, average daily feed intake, and average egg weight were measured from 52 to 57 weeks of age. Based on the calculated FCR of Wenshang Barred hens, the experimental birds were divided into HFCR and LFCR group (n = 6, each) (Table 1). No statistically significant differences of serum biochemical parameters and hormone concentrations were observed between the two groups (p > 0.05) (Table S2). Notably, the GH exhibited a slight upward trend in the low FCR group (p = 0.074). Collectively, these results suggest that homeostasis in FCR between the two groups was maintained through internal regulatory mechanisms.

Table 1.

Grouping information of a high feed conversion ratio (HFCR, n = 6) and a low FCR group (LFCR, n = 6).

Items	Average Daily Feed Intake (g/d)	Average Egg Weight (g/d)	Feed Conversion Rate (%)	Laying Rate (%)
Low FCR (LFCR)	105.77	38.71	2.75±0.22	80.2
High FCR (HFCR)	118.50	31.80	3.77±0.32	69.0
P-value	0.052	0.130	<0.001	0.104

Transcriptomics analysis in liver tissue

On average, a total of 48,273,038 and 43,066,602 raw reads were obtained from the liver tissues from HFCR and LFCR group, respectively (Table S4). Approximate 95.31% and 95.29% of clean reads were mapped to the chicken reference in a manner of uniquely mapped reads, respectively (Table S4). Unmapped reads or reads that mapped multiple times were removed. In total, 81.47 Gb of clean data was generated. After filtering low-abundance genes, 189 DEGs were identified, including 72 up-regulated and 117 down-regulated genes (p < 0.05, |Log₂FC| >1) (Fig. 2A). RT-qPCR analysis of seven randomly selected genes confirmed consistent expression trends with the transcriptome profiles (Fig. 2B). In order to detect the highly correlated genes with FCR, the 189 DE genes were used for WGCNA. Genes in the turquoise module were significantly positively correlated with FCR (R²=0.67) (Fig. 2C), among which NOS2 gene and ASL1 gene were notably enriched in the arginine biosynthesis pathway (gga00220) (Fig. 2D). In addition, seven hub genes were commonly identified by Degree, Closeness, Bottleneck and ECP algorithms (Fig. 2E and Table 2), which also have connections (Fig. 2F).

Fig. 2.

(A) Differentially expressed genes numbers in high feed conversion ratio (HFCR, n = 6) and low FCR (LFCR, n = 6) groups (B) qRT-PCR validation of transcriptome sequencing results (C) Module trait relationship with FCR (D) KEGG enrichment analysis in turquoise module (E) Top ten hub genes calculated by five algorithms including Degree, Closeness, Bottleneck, ECP and DMNC (F) Hub-gene-centered PPI network generated via the Degree-based algorithm.

Table 2.

Top 10 genes calculated by five algorithms.

BottleNeck	Degree	Closeness	EPC	DMNC
IL4I1	IFI6	IFI6	IFI6	MX1
IFI6	ACOD1	DDX60	DDX60	DDX60
GINS1	DDX60	ACOD1	ACOD1	IFI6
ACOD1	IL4I1	IL4I1	MX1	IL4I1
DDX60	MX1	MX1	IL4I1	MCM10
MX1	GINS1	PLAC8	PLAC8	CNTF
TGM4	MCM10	TGM4	TGM4	FGF13
NCAPG2	CNTF	GINS1	GINS1	ACOD1
MCM10	FGF13	MCM10	MCM10	PLAC8
CNTF	PLAC8	NCAPG2	NCAPG2	TGM4

Note: IL4I1: interleukin 4 induced 1 gene; IFI6: interferon alpha inducible protein 6 gene, GINS1: GINS complex subunit 1 gene; ACOD1: aconitate decarboxylase 1 gene; DDX60: DExD/H-box helicase 60 gene; MX1: MX dynamin like GTPase 1 gene; MCM10: minichromosome maintenance 10 replication initiation factor gene; PLAC8: PLAC8 like 1 gene; TGM4: transglutaminase 4; NCAPG2: non-SMC condensin II complex subunit G2; CNTF: ciliary neurotrophic factor; FGF13: fibroblast growth factor 13.

Interference and overexpression efficiency of the NOS2 gene in primary hepatocytes

As a member of the nitric oxide synthase (NOS) family, NOS2 was highly expressed in the LFCR group and notably enriched in the arginine biosynthesis pathway, promoting its selection for functional validation in chicken primary hepatocytes. It showed that siRNA-NOS2-1 downregulated the expression of NOS2 efficiently than other two siRNAs at mRNA levels after 48 h transfection interference fragments of NOS2 in the primary hepatocytes (Fig. 3A). Next, the NOS2 cDNA fragment of 3477 bp were constructed into a plasmid of pcDNA3.1(+)eGFP and digested with BamHI and NotI to confirm the fragment of NOS2 (Fig. 3B). The transfection efficiency was confirmed by green fluorescent signal under microscope (Fig. 3C) and qRT-PCR which showed that the mRNA expression of NOS2 gene was extremely higher after transfected with pcDNA3.1(+)eGFP-NOS2 for 48 h (Fig. 3D).

Fig. 3.

(A) Nitric oxide synthase 2 gene (NOS2) expression at mRNA level after transfected with three siRNAs including siRNA-NOS2-1, siRNA-NOS2-2 and siRNA-NOS2-3 (n = 3) for 24 h and 48 h (B) The NOS2 cDNA fragment of 3477 bp and digested with BamHI and NotI to construct the plasmid of pcDNA3.1(+)eGFP-NOS2 (C) The green fluorescent signal under microscope after transfected with pcDNA3.1(+)eGFP and pcDNA3.1(+)eGFP-NOS2 (D)NOS2 mRNA expression level after transfected with pcDNA3.1(+)eGFP and pcDNA3.1(+)eGFP-NOS2 for 24 and 48 h, the relative expression level of pcDNA3.1(+)eGFP-NOS2 treated for 48 h was showed as log(2^-△△Ct). Error bars indicate the SD among three repeats. Values are mean ± SD for three biological replicates. SD: standard deviation. *p < 0.05, ** p < 0.01, *** p < 0.001.

Gene expression after interference and overexpressed NOS2 in primary hepatocytes

Following transfection of chicken primary hepatocytes with siRNA-NOS2-1 for 24 h and 48 h, the mRNA expression levels of the antioxidant-related gene CAT (p = 0.047; Fig. 4A) and Nrf2 (p = 0.031; Fig. 4C) was significantly downregulated after 48 h transfection with siRNA-1, whereas that of GPx1 (p = 0.056; Fig. 4D) and SOD (p = 0.118; Fig. 4B) remained unchanged both in 24 h and 48 h. Regarding lipogenesis-related genes, transcript levels of SCD1 (p = 0.368; Fig. 4E), ACCa (p = 0.361; Fig. 4F), and SREBF1 (p = 0.295; Fig. 4H) were unaffected, but FASN (p = 0.007; Fig. 4G) expression was significantly decreased after 48 h transfection with siRNA-1. Inflammatory factor genes of TNF-α (p = 0.047; Fig. 4J) and HO-1 (p = 0.017, Fig. 4L) was downregulated only after 48 h transfection with siRNA-1, while IL-6 (p = 0.341; Fig. 4I) and IL-10 (p = 0.105; Fig. 4K) showed no significant alterations both in 24 h and 48 h.

Fig. 4.

Twelve genes expression at mRNA level after transfected with siRNA-NOS2-1. (A) Catalase gene (CAT) (B) Superoxide dismutase gene (SOD) (C) Nuclear factor, erythroid 2 like 2 gene (Nrf2) (D)Glutathione peroxidase 1 gene (GPX1) (E) Stearoyl-CoA desaturase 1 gene (SCD1) (F) Acetyl-CoA carboxylase alpha gene (ACCa) (G) Fatty acid synthase gene (FASN) (H) Sterol regulatory element binding transcription factor 1 gene (SREBF1) (I) Interleukin 6 gene (IL-6) (J) Tumor necrosis factor-alpha gene (TNF-α) (K) Interleukin 10 gene (IL-10) (L) Heme oxygenase 1 gene (HO-1). Error bars indicate the SD among three repeats. Values are mean ± SD for three biological replicates. SD: standard deviation. *p < 0.05, ** p < 0.01.

Following transfection of chicken primary hepatocytes with pcDNA3.1(+)eGFP-NOS2 for 24 h and 48 h, the mRNA expression levels of the antioxidant-related gene CAT (Fig. 5A) and Nrf2 (Fig. 5C) were extremely upregulated (p < 0.001), while SOD (Fig. 5B) and GPx1 gene (Fig. 5D) were downregulated (p < 0.001) after transfection with pcDNA3.1(+)eGFP-NOS2 for 48 h. Lipogenesis-related genes of SCD1 (p = 0.811; Fig. 5E), ACCa (p = 0.525; Fig. 5F), and SREBF1 (p = 0.134; Fig. 5H) were also unaffected, but FASN (p < 0.001; Fig. 5G) expression was increased after transfected with pcDNA3.1(+)eGFP-NOS2 for 48 h. Inflammatory factor genes of TNF-α (p < 0.01; Fig. 5J) and IL-6 (p < 0.001) were also downregulated at 48 h, while IL-10 (p = 0.524; Fig. 5K) showed no significantly alterations both in 24 h and 48 h.

Fig. 5.

Twelve genes expression at mRNA level after transfected with pcDNA3.1(+)eGFP and pcDNA3.1(+)eGFP-NOS2. (A) Catalase gene (CAT) (B) Superoxide dismutase gene (SOD) (C) Nuclear factor, erythroid 2 like 2 gene (Nrf2) (D)Glutathione peroxidase 1 gene (GPX1) (E) Stearoyl-CoA desaturase 1 gene (SCD1) (F) Acetyl-CoA carboxylase alpha gene (ACCa) (G) Fatty acid synthase gene (FASN) (H) Sterol regulatory element binding transcription factor 1 gene (SREBF1) (I) Interleukin 6 gene (IL-6) (J) Tumor necrosis factor-alpha gene (TNF-α) (K) Interleukin 10 gene (IL-10) (L) Heme oxygenase 1 gene (HO-1). Error bars indicate the SD among three repeats. Values are mean ± SD for three biological replicates. SD: standard deviation. *p < 0.05, ** p < 0.01, *** p < 0.001.

Metabolites profiles in liver and cecum content

Untargeted metabolomics profiling in liver tissues and cecum contents was performed to systematically elucidate the regulatory mechanism to FCR. The cumulative proportion of peaks were 96.73% and 82.44% in liver and cecum content, respectively (Fig. S3) when RSD value less than 30% (Fig. S3). Totally, 17,830 and 20,519 mix ion mode were identified, including 2623 and 4046 identified metabolites in liver and cecum contents, respectively (Table S6). The PLS-DA derived PC1 and PC2 collectively explained 21.3% and 16.2% in liver (Fig. 6A), 11.3% and 22.3% in cecum contents (Fig. 7A). The model was suitable for the subsequent data analysis with R² = 0.93 and Q² = 0.25 in liver tissue (Fig. 6B), and R² = 0.95 and Q² = 0.15 in cecum contents (Fig. 7B).

Fig. 6.

Comparison analysis of metabolites between high feed conversion ratio (HFCR, n = 6) and low FCR (LFCR, n = 6) groups in liver. (A) Partial least squares-discriminant analysis (PLS-DA) plot (B) Permutation test (C) Volcano plot of HFCR and LFCR metabolites. The horizontal coordinate is the difference multiple of metabolic ions in HFCR and LFCR; the ordinate is the p-value. (D) KEGG topology analysis for the DE metabolites.

Fig. 7.

Comparison analysis of metabolites between high feed conversion ratio (HFCR, n = 6) and low FCR (LFCR, n = 6) groups in cecum contents. (A) Partial least squares-discriminant analysis (PLS-DA) plot (B) Permutation test (C) Volcano plot of HFCR and LFCR metabolites. The horizontal coordinate is the difference multiple of metabolic ions in HFCR and LFCR; The ordinate is the p-value. (D) Heatmap for the differential metabolites (E) KEGG topology analysis for the differentially identified metabolites.

Totally, 116 differentially metabolites (VIP >1 and p < 0.05) were identified in liver, including 25 up-regulated and 91 down-regulated metabolites (Fig. 6C, Supplementary file 1), such as the down-regulated positive metabolites of serotonin and deoxycholylcysteine (Fig. 6D and Table 3). Metabolites enrichment analysis found that top four significantly KEGG pathways were significantly enriched, including tryptophan metabolism (map00380), folate biosynthesis (map00790), gap junction (map04540) and pyrimidine metabolism (map00240) (Fig. 6E, Supplementary file 2).

Table 3.

Top 12 differential identified metabolites in liver tissue.

Metabolites	p-value	VIP	FC	Regulate	Ion Mode
Docosahexaenoyl Ethanolamide	0.016	1.44	1.03	Up	pos
N-Stearoyl Lysine	0.027	1.08	1.03	Up	pos
Oleoyl Ethanolamide	0.035	1.42	1.03	Up	pos
Serotonin	0.041	3.32	0.87	Down	pos
Deoxycholylcysteine	0.038	2.98	0.70	Down	pos
7,8-Dihydroneopterin	0.012	3.62	0.86	Down	pos
Proscillaridin	0.048	2.46	1.11	Up	neg
N-Desmethyl-O-Hydroxy Rosiglitazone	0.043	1.68	1.06	Up	neg
Pg (36:4)	0.040	1.13	1.02	Up	neg
Paramethadione	0.004	2.25	0.94	Down	neg
3-Methyldioxyindole	0.037	1.50	0.97	Down	neg
1-Methyluric Acid	0.037	1.88	0.93	Down	neg

Note: FC: fold change; VIP: Variable importance in the projection.

Eighty-seven differentially metabolites were identified in cecum content, including 23 up-regulated and 64 down-regulated metabolites (Fig. 7C, Supplementary file 3). Such as the down-regulated positive metabolites of aspartame, and gamma-linolenic acid, down-regulated negative metabolites of N-acetylcystathionine and protocatechuic acid (Fig. 7D and Table 4). Metabolites enrichment analysis found that top three significantly KEGG pathways were biosynthesis of unsaturated fatty acids (map01040), phenylalanine, tyrosine and tryptophan biosynthesis (map00400) and polycyclic aromatic hydrocarbon degradation (map00624) (Fig. 7E, Supplementary file 4).

Table 4.

Top 12 differential identified metabolites in cecum contents.

Metabolites	p-value	VIP	FC	Regulate	Ion Mode
Piperidine	0.028	3.59	1.13	Up	pos
Methyl Dioxindole-3-Acetate	0.010	2.33	1.05	Up	pos
4-Ethynylaniline	0.016	1.08	1.01	Up	pos
Aspartame	0.028	1.34	0.97	Down	pos
Acetolein	0.016	1.41	0.98	Down	pos
Gamma-Linolenic Acid	0.023	1.78	0.97	Down	pos
2,8-Quinolinediol	0.013	2.25	1.05	Up	neg
8-Hydroxy-9,10-Epoxystearic Acid	0.016	1.51	1.03	Up	neg
Melleolide K	0.018	3.64	1.17	Up	neg
N-Acetylcystathionine	0.014	2.21	0.96	Down	neg
Octadecanedioic Acid	0.034	1.88	0.97	Down	neg
Protocatechuic Acid	0.010	2.29	0.95	Down	neg

Note: FC: fold change; VIP: Variable importance in the projection.

Correlation between differentially expressed genes and metabolites in liver

To elucidate the association between DE genes and differential metabolites, correlation analysis was performed using 189 DE genes, 87 positive differential metabolites and 29 negative differential metabolites. A total of 485 mRNA–positive metabolite pairs and 287 mRNA–negative metabolite pairs demonstrated statistically significant correlations (p < 0.05 and FDR < 0.05). Notably, 14 metabolites (e.g., 2-picoline, N-methylaniline and N-stearoyl cysteine) exhibited strong positive correlations (FDR < 0.001) with multiple DEGs. In contrast, a distinct cluster of metabolites (e.g., ergothioneine, calcium pantothenate, N-stearoyl lysine and 3-ketocholanic acid) displayed predominantly negative correlations with the overlapping DEG sets, suggesting potential antagonistic regulatory mechanisms (Supplementary file 5 and Fig. 8A). To further characterize the topological architecture of these metabolite-gene interactions, a co-expression network was constructed incorporating all statistically significant correlations (FDR < 0.05; Fig. 8B and 8D). Such as, the NOS2 gene was significantly positively correlated with 14 positive metabolites (e.g., 2-picoline, N-methylaniline and N-stearoyl lysine), and 9 negative metabolites (e.g., paramethadione and 14-methylpentadecanoic acid) (Figure abscript).

Fig. 8.

Heatmap for the Pearson correlation coefficients of positive ions (A) or negative ions (B) metabolites with differentially expressed genes (DEGs) Network for the correlation analysis between positive ions metabolites (C) or negative ions metabolites (D) and differentially expressed genes. Note: diamond-shaped: metabolite nodes; circular: DEG nodes; positive correlations colored in red; negative correlation colored in blue.

Metagenomic sequencing and function predication in cecal contents

To elucidate the microbial community structures and functional gene repertoires driving the differential cecal metabolite profiles observed between the HFCR and LFCR groups, metagenomic sequencing analysis of cecum contents was performed to compare the microbial community variation in the HFCR and LFCR group. An average of 4,470,047 and 4,360,951 clean reads were obtained in the HFCR and LFCR group, respectively. After assembly, an average of 414,655 and 403,398 contigs with an average N50 length of 1,224 bp and 1,212 bp were produced in HFCR and LFCR group, respectively (Table S8). An average of 618,495 and 604,352 ORFs were predicted with gene lengths between 200 and 400 bp accounted for the largest proportion (Table S9 and Fig. S4). Alpha diversity analysis revealed no significant differences in Simpson, Shannon, observed species, and Chao1 indices of cecal microbiota between the two groups (p > 0.05) (Fig. S5).

At the genus level, Bacteroides was the dominant genera in the cecal community, accounting for 16.84% and 15.74% in the HFCR and LFCR group, respectively (Fig. S6). At genus level, the relative abundances of Oscillibacter (p = 0.04) and Escherichia (p = 0.04) were significantly lower in the LFCR than that of HFCR group. However, the relative abundance of Candidatus_Gallacutalibacter (p = 0.04) and Methanocorpusculum (p = 0.04) were significantly higher in the LFCR than HFCR group (Fig. 9A). At species level, the relative abundances of Candidalus Paranacloroides intestinipulorum (p = 0.005) and Bacteripphage_sp (p < 0.05) were significantly higher in the LFCR than HFCR group (Fig. 9B). Linear discriminant analysis (LDA) effect size found that g__unclassified_f__Methanobacteriaceae and s__Methanobacteriaceae_archaeon were the major differential taxa in HFCR group (Supplemental file 6), while s__Methanocorpusculum_vombati was the major taxa in LFCR group (Fig. 9C). Functional annotation found that bacterial chemotaxis (p = 0.01) was the major KEGG in LFCR group, while RNA degradation, thiamine metabolism and tuberculosis were the major KEGGs in HFCR group (p < 0.05) (Fig. 9D).

Fig. 9.

Wilcoxon rank-sum test bar plot on (A) genus and (B) species level (C) Linear discriminant analysis (LDA) effect size (LEfSe) identification of cecum content taxa, and (D) KEGG at level 3 with significant differences (LDA > 1, P < 0.05), between high feed conversion ratio (HFCR, n = 6) and low FCR (LFCR, n = 6) group.

Correlation between cecal microbiomes and metabolites in cecum content

To investigate whether the observed compositional shifts in cecal microbiota were functionally associated with metabolic alterations, we performed an integrated correlation analysis between differentially abundant microbial genera and significantly altered metabolites detected in cecum contents (Supplementary file 7). In the positive (Fig. 10A) and negative (Fig. 10C) ion mode, the correlation heatmap revealed distinct co-clustering patterns between cecal microbial general and metabolites, which highlighted g__Oscillibacter as the central hub microbiota, significantly connecting to 2-aminobicyclo[3.1.0]hexane-2,6-dicarboxylic acid (Fig. 10) and 1-methoxyficifolinol (Fig. 10). Taken together, the convergent hub roles across both ionization modes suggest its broad regulatory influence over the cecal metabolome.

Fig. 10.

Heatmap for the Pearson correlation coefficients of positive ions (A) or negative ions (B) metabolites with microbiome. Network for the correlation analysis between positive ions metabolites (C) or negative ions metabolites (D) and cecum microbiome correlation. Note: diamond-shaped: metabolite nodes; circular: microbiome nodes; positive correlations colored in red; negative correlation colored in blue.

Discussion

The present study established an integrated multi-omics framework combining liver transcriptomics, liver and cecal metabolomics, and cecal metagenomics to systematically elucidate the molecular mechanisms underlying feed efficiency in Wenshang Barred hens. We identified NOS2 as a key candidate regulator of FCR, with potential roles in modulating antioxidant capacity, lipogenesis, and inflammatory responses. Furthermore, we identified an NOS2-metabolite-cecal microbiota regulatory axis, revealing the regulatory network governing the interactions among host genes, metabolites, and cecal microbiota that collectively modulate FCR in Wenshang Barred hens.

Transcriptional co-expression network analysis in liver identified the turquoise module as the principal regulatory module associated with feed efficiency, with NOS2 exhibiting high expression in the LFCR group. This finding is consistent with previous studies in pigs, where NOS2 exerted a beneficial effect on improved feed efficiency (Tan et al., 2017a). NOS2, a member of the nitric oxide synthase (NOS) family, catalyzes the conversion of l-arginine to nitric oxide (NO), an important signaling molecule involved in antimicrobial defense and immune responses (Coleman, 2001; Bogdan, Röllinghoff and Diefenbach, 2000). In addition, NOS2 interference in primary hepatocytes significantly downregulated the expression of Nrf2, but when overexpressed NOS2 gene in hepatocytes, the Nrf2 and CAT were upregulated, which indicated that NOS2-Nrf2-CAT regulatory axis may contribute to improve the host antioxidant capacity and nutrient absorption to regulate FCR (Deng et al., 2026). However, as the Nrf2-dependent antioxidant genes, GPx1 and SOD were not upregulated following NOS2 overexpression. Such a seemingly paradoxical response is attributable to NO, the direct enzymatic product of NOS2, which acts as a double-edged sword in redox regulation (Thomas et al., 2015; Um et al., 2011). NO can S-nitrosylate Keap1, thereby releasing Nrf2 and promoting its nuclear translocation and subsequent transcriptional activation of select ARE-driven genes such as CAT (Ma, 2013). Excessive NO production can generate peroxynitrite (ONOO⁻) through reaction with superoxide, which inactivates both SOD and GPx1 (Sies et al., 1997). NOS2-derived NO may act as a molecular rheostat to fine-tune the antioxidant defense system, instead of uniformly upregulating the entire set of Nrf2-driven target genes. This selective activation of Nrf2-driven target genes could adequate antioxidant protection via CAT induction, thereby permitting greater allocation of metabolic resources to productive processes including egg synthesis and energy conversion in LFCR hens. Furthermore, the simultaneous downregulation of pro-inflammatory cytokines (TNF-α, IL-6) upon NOS2 overexpression suggests that NOS2 may exert a protective, anti-inflammatory role in the liver, consistent with the emerging concept that efficient nutrient utilization is associated with a balanced hepatic immune-metabolic state (Amevor et al., 2025). Notably, NOS2 suppression significantly reduced FASN expression, whereas NOS2 overexpression markedly elevated FASN levels, indicating that NO signaling participates in regulating hepatic lipogenesis. This observation agrees with previous reports that highly feed-efficient hens exhibit reduced hepatic lipid deposition (Guo et al., 2025). Together, these results indicate that NOS2 favors efficient nutrient utilization in LFCR group by simultaneously enhancing antioxidant capacity, modulating inflammatory factors, and promoting liver lipid metabolism.

Serotonin was identified as a differentially abundant metabolite in LFCR, enriched in the tryptophan metabolic pathway. Serotonin has been reported to regulate appetite (Blundell, 1984), contribute to hepatic steatosis (Choi et al., 2018), and promote hepatocellular lipogenesis (Hwang et al., 2025; Osawa et al., 2011). Accumulating evidences indicate the existence of functional crosstalk between arginine and tryptophan metabolic to immunoregulatory (Mondanelli et al., 2019; Mondanelli et al., 2017; Zhang, Mishra and Jagannath, 2024a). Transcriptome analysis found that NOS2 enrichment in the arginine biosynthesis pathway, raises an intriguing possibility of metabolic cross-talk between the tryptophan and arginine pathways (Mondanelli et al., 2017; Grohmann and Bronte, 2010). Integrated gene-metabolite correlation analysis revealed that NOS2 was extensively correlated with a broad metabolic connectivity, reinforcing its role as a key molecular hub linking transcriptional regulation to metabolic output. These findings suggest that NOS2 functions not merely as an individual effector gene but as a network hub regulator that coordinates multiple metabolic pathways relevant to feed efficiency.

The cecum serves as a critical interface between linking gut microbiota-derived signals to hepatic metabolic and immune responses (Choi et al., 2018). Cecum and cecal microflora play an indispensable role in absorbing additional nutrients and maintaining normal physiological functions (Tan et al., 2019; SVIHUS, Choct and Classen, 2013). Genus Oscillibacter is significantly associated with FCR (Tan et al., 2017b; McCormack et al., 2017; Bergamaschi et al., 2020; Liu et al., 2021). Consistently, cecal microbiome and metabolome combined analysis also identified Oscillibacter as a central microbial hub that was strongly associated with differential metabolites, suggesting it as one of the key microbial signatures to distinguish LFCR from HFCR group.

PPI network also identified IFI6 as another hub gene. IFI6 is also a well-characterized interferon-stimulated gene that has played a definitive role in mitochondrial antioxidant defense (Liu et al., 2020; Qi et al., 2015; El-Merahbi et al., 2015). Its central role within the PPI networks suggests that innate immune-metabolic cross-regulation (Gao et al., 2025a; Gao, Zheng and Xu, 2025; Amevor et al., 2026) may constitute a previously underappreciated regulatory axis contributing to feed efficiency in laying hens. Nevertheless, the specific molecular regulatory mechanisms underlying FCR in Wenshang Barred hens require further validation through in vivo functional experiments. In addition, the relatively limited number of biological replicates may lead to potential bias and reduce the reliability and statistical power of the transcriptomic, metabolomic, and microbiome integration analysis and should therefore be considered a limitation of the present study.

Conclusion

This study systematically elucidated the molecular mechanisms underlying feed efficiency in Wenshang Barred hens through an integrated multi-omics approach encompassing liver transcriptomics, liver and cecal metabolomics, and cecal metagenomics. NOS2 was identified as a central hub gene that was highly expressed in LFCR group and significantly enriched in the arginine biosynthesis pathway. Functional validation in chicken primary hepatocytes demonstrated that NOS2 regulated the mRNA expression of genes related to antioxidant (CAT, Nrf2, SOD and GPx1), lipogenic (FASN), and inflammatory factors (TNF-a, IL-6 and HO-1). Furthermore, integrated gene-metabolite correlation analysis recommended NOS2 as the central regulator of the hepatic metabolic network, showing significant associations with 23 differential metabolites. Hepatic metabolomics identified tryptophan metabolism as the most significantly enriched pathway, with serotonin downregulation in LFCR hens potentially reflecting reduced lipogenic pressure and more efficient energy partitioning. Cecal metagenomics identified Oscillibacter as a key microbial hub linking gut microbial metabolism to host metabolic regulation through the gut-liver axis. Collectively, these findings establish a NOS2-metabolite-microbiota regulatory axis underlying feed efficiency, providing promising molecular targets for practical molecular breeding programs and nutritional intervention strategies in Wenshang Barred hens.

Ethics approval

All animal procedures were approved by the Institutional Review Board of the Institutional Animal Care and Use Committee of Shandong Academy of Agricultural Science (protocol code: SAAS-2025-037).

Abbreviation

ACC: Acetyl-coA carboxylase gene; ASL1: Argininosuccinate lyase 1 gene; AST: Aspartate transaminase; ALT: Alanine aminotransferase; ACOD1: Aconitate decarboxylase 1 gene; AMPKα1: AMP-activated protein kinase catalytic subunit alpha 1 gene; CNTF: Ciliary neurotrophic factor; CYP26B1: Cytochrome P450 family 26 subfamily B member 1; cDNA: Complementary DNA; DEGs: Differentially expressed genes; DDX60: DExD/H-box helicase 60 gene; E2: Estradio; FASN: Fatty acid synthase gene; FCR: Feed conversion ratio; FSH: Follicle-stimulating; FGF13: Fibroblast growth factor 13 gene; FDR: False discovery rate; GLU: Glucose; GH: Growth hormone; GREM2: Gremlin 2, DAN family BMP antagonist gene; GINS1: GINS complex subunit 1 gene; GIT: Gastrointestinal tract; HDL-C: High-density lipoprotein cholesterol; HFCR: High FCR group; HMDB: Human Metabolome Database; HMGR: 3‑hydroxy-3-methylglutaryl-CoA reductase gene; IL4I1: Interleukin 4 induced 1 gene; IFI6: Interferon alpha inducible protein 6 gene; LDL-C: Low-Density Lipoprotein; LH: Luteinizing hormone; LFCR: Low FCR group; LXRα: Liver X receptor α gene; MX1: MX dynamin like GTPase 1 gene; MCM10: Minichromosome maintenance 10 replication initiation factor gene; NCAPG2: Non-SMC condensin II complex subunit G2 gene; NOS2:Nitric oxide synthase 2 gene; ORFs: Open reading frames; OPLS-DA: Orthogonal partial least squares discriminant analysis; PCCB: Propionyl-CoA carboxylase beta subunit gene; PCA: Principal component analysis; PLAC8: PLAC8 like 1 gene; PPARα: Peroxisome proliferator activated receptor alpha gene; QC: Quality control; RQN: RNA Quality Number; SCFAs: Short-chain fatty acids; SCD1: Stearoyl-CoA desaturase 1 gene; SOD: Superoxide dismutase gene; SREBF1: Sterol regulatory element binding transcription factor 1 gene; TP: Total protein; TC: Total cholesterol; TG: Triglyceride; TPM: Transcripts per million reads; TGM4: Transglutaminase 4; TNF-α: Tumor necrosis factor-alpha; VIP: Variable importance in projection; WNT11:Wnt family member 11.

Author contribution

Dan Hao: Conceived and designed the study, interpreted the data, drafted the first version of the manuscript, and critically revised the manuscript.

Huifang Li: Conducted the formal analysis, data interpretation and visualization.

Wei Liu: Performed animal experiments and interpretation the data.

Guangqiang Shang: Conducted formal analysis and data interpretation.

Yan Sun: Participated in experimental design, and revised the manuscript.

Yan Zhou: Conducted data analysis and critically revised the manuscript.

Jie Wang:Participated in experimental phenotypic data collection and part of the data interpretation.

Dingguo Cao: Methodology support and critically revised the manuscript.

Chuanyan Che: Participated in experimental phenotypic data collection, and for the methods improvement for the revised the manuscript.

Fuwei Li: Conceived and designed the study, and critically revised the manuscript.

Disclosures

We declare that we 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 file 1. Differentially identified metabolites in liver tissue between high feed conversion ratio (HFCR, n = 6) and a low FCR (LFCR, n = 6) groups.

Supplementary file 2. Enriched KEGGs for the differentially identified metabolites in liver tissue between high feed conversion ratio (HFCR, n = 6) and a low FCR (LFCR, n = 6) groups.

Supplementary file 3. Differentially identified metabolites in cecum content between high feed conversion ratio (HFCR, n = 6) and a low FCR (LFCR, n = 6) groups.

Supplementary file 4. Enriched KEGGs for the differentially identified metabolites in cecum contents between high feed conversion ratio (HFCR, n = 6) and low FCR (LFCR, n = 6).

Supplementary file 5. Pearson correlation coefficients of differentially expressed genes and metabolites in liver tissue.

Supplementary file 6. Identification of differentially abundant taxa in cecal contents using the linear discriminant analysis (LDA)effect size (LEfSe) algorithm.

Supplementary file 7. Pearson correlation coefficients of differentially metabolites and microbial in cecum content.

Data Availability

The liver transcriptome is available in the National Genomics Data Center under the accession number CRA042604. The metagenomic sequencing data is available in the National Genomics Data Center under the accession number CRA042657.