Identification of central regulators related to residual feed intake in Huainan chickens based on weighted gene co-expression network analysis

Abstract

The residual feed intake (RFI) is a crucial economic trait in chickens. However, the genetic network and regulatory mechanisms that underpin RFI traits and the effects of RFI on meat quality and slaughter performance in chickens remain unclear. In this study, a total of 315 male Huainan chickens were reared from 7 to 13 weeks of age, and feed intake and weight gain were recorded for each individual bird. Based on the calculated RFI values, the 30 chickens with the highest RFI values are classified into the high residual feed intake (HRFI) group, while the 30 chickens with the lowest RFI values are classified into low residual feed intake (LRFI) group. The results revealed a significantly lower abdominal fat percentage in the LRFI group; however, no significant differences were detected in other meat quality traits or slaughter performance parameters. This phenotypic difference may be associated with the high expression of PCK1 in the HRFI group, which is likely to enhance glucose metabolism and thereby promote abdominal fat deposition. Two groups randomly selected 9 samples each for RNA seq analysis. The obtained transcriptome data were subjected to differential gene expression analysis, which revealed that 170 genes exhibited down-regulation while 109 genes displayed up-regulation in HRFI group relative to LRFI group. A total of 23097 genes were used to construct the weighted gene co-expression network analysis (WGCNA), and 27 co-expression gene modules were identified. Among these modules, the magenta module (R = 0.66, P = 0.003) has a significant positive correlation with RFI, while the pink module(R=﹣0.6,P = 0.009) has a significant negative correlation. The hub genes within the above modules were identified based on MM > 0.8 and GS > 0.4. Combining differential and hub genes, 56 key genes were identified as being significantly correlated with RFI traits. Several genes were identified as central regulator genes due to their involvement in the regulation of mitochondrial function (e.g., ACE2, ACMSD), glucose metabolism (e.g., FABP2, FETUB, PCK1) and lipid metabolism (e.g., APOA1). The findings will contribute to a more profound comprehension of the genetic expression and regulation of RFI traits, thereby providing a foundation for genetic breeding.

Introduction

Feed costs account for approximately 60-70 % of the total cost of producing poultry (Willems et al., 2019) and reducing feed costs is essential for the development of the poultry industry. Improving feed efficiency presents a sustainable way to reduce production costs (Godfray, et al., 2010).

Residual feed intake (RFI), also termed net feed efficiency, is defined as the difference between actual feed intake and the intake predicted based on requirements for production and maintenance of body weight (Kennedy., et al., 1993). As RFI is phenotypically independent of growth rate and body size (Gilbert, et al., 2017), selecting for low residual feed intake (LRFI) enables lower feed costs without compromising growth performance. Given its moderate heritability (ranging from 0.2 to 0.5) (Xu, et al., 2016), RFI is widely regarded as the most suitable index for selecting feed-efficient poultry (Aggrey, et al., 2010). However, research findings on the effects of RFI on slaughter performance and meat quality are inconsistent. The eviscerated yield and PH value were been found to be significantly higher in the high RFI (HRFI) group (Yang, et al., 2020; Bai, et al., 2024). Others found no significant differences in slaughter performance or meat quality between different RFI groups (Wu, et al., 2025). Therefore, further investigations are warranted to clarify the effect of RFI on slaughter performance and meat quality.

Recently, many studies have been conducted to elucidate the molecular mechanisms related to RFI trait. Transcriptome analysis revealed that the SLC2A5 gene, which encodes glucose transporter 2, and the CTRL gene, which is involved in proteolysis, were upregulated in the LRFI group (Liu, et al., 2019). This suggests that nutrient absorption is closely associated with RFI. Whole-genome sequencing (WGS) had identified several key genes influencing RFI in Cobb chickens, including PLCB4 and CAMK4, which are associated with cAMP-responsive element-binding protein signaling, as well as CDC42, CSK, and PIK3R3, involved in Rac signaling and actin cytoskeleton regulation (Liu, et al., 2018). Similarly, seven significant SNPs associated with RFI were identified in yellow-feather dwarf broiler using genome-wide association analysis (GWAS) (Ye, et al., 2020). However, the RFI is a complex trait, that arises from complex interactions among multiple genes and pathways and the regulatory network affect this trait remain unclear.

Weighted gene co-expression network analysis (WGCNA) is a technique employed to construct gene co-expression networks through the clustering of genes into modules. This allows for the identification of groups of genes that are co-expressed under specific biological conditions, as well as the mining of potential hub genes (Zhang and Horvath, 2005; Langfelder and Horvath, 2008).It is a widely employed technique in the annotation of gene function and the study of the molecular mechanisms underlying complex traits.

This study compared slaughter performance and meat quality between different RFI groups in Huainan chickens. Furthermore, leveraging small intestinal transcriptome data from 18 individuals, differential gene expression analysis and WGCNA technique were combined to identify central regulator genes that may influence RFI. These findings provide a foundation for further inquiry into the regulatory network of RFI.

Materials and methods

Ethics statement

The study was conducted in accordance with the guidelines set forth by the Animal Care and Use Committee of Anhui Agricultural University (Hefei, China) (permit number: SYXK (WAN) 2021-009).

Animals

315 one-day-old male Huainan chickens were obtained from the Jianlang Poultry Breeding Co., Ltd., Hefei, Anhui Province, China. All cocks were housed individually in cages. Throughout the experimental period, all cocks were granted unrestricted access to feed and water. The body weight gain (BWG) and total feed intake (TFI) of the cocks were meticulously recorded to calculate the RFI from days 49 to 91. The following equation:

$$Metabolicmidweight\left(\text{MMW}\right)={\left(\frac{B{W}_{49}\left(kg\right)+B{W}_{91}\left(kg\right)}{2}\right)}^{0.75}$$

$$RFI=ADFI-\left(b_0+b_{1}MMW+b_{2}ADG\right)$$

According to the RFI values, the 30 cocks with the highest RFI values were classified into the HRFI group, while the 30 cocks with the lowest RFI values were classified into the LRFI group (Supplementary table S1).

Sample collection and slaughter performance

At 91 days of age, a total of 30 cocks from two groups were selected, fasted for 12 hours, weighted, electrocuted, and then euthanized by bloodletting. Following euthanasia, samples measuring approximately 2-3 cm were excised from the small intestine of each bird and placed in liquid nitrogen. After bleeding and plucking, the carcass weight was recorded. After evisceration, the eviscerated carcass, breast muscle, leg muscle, and abdominal fat were weighed. Formulas for calculating carcass traits are as Zou, et al. (2023).

Meat quality

The meat color, pH, cooking loss, and shear force were determined using the left side of the breast muscles. The meat color and pH value were measured at 45 min postmortem with a chroma meter (Konica Minolta, CR-400, Osaka, Japan) and pH meter (pH-STAR, Matthaus, Berlin, Germany). According to the method of Xiao, et al. (2021), the cooking loss and shear force were measured using a thermostatic water bath (DK-S24, Shanghai Jinghong Experimental Equipment Co. Ltd., Shanghai, China) and a digital tenderness meter (C-LM3B, Tenovo, Beijing, China). Each index of each sample was measured repeatedly 3 times, and its average value was taken for statistical analysis.

RNA isolation and sequencing

The 9 samples of small intestine tissue from each group were subjected to RNA extraction in accordance with the provided instructions using TRIzol reagent (Invitrogen, Carlsbad, CA) to ensure consistent extraction conditions. To control technical variability arising from the extraction step, all samples were processed in parallel by the same operator, and the extraction reagents were from the same batch. The NanoDrop 2000 Spectrophotometer (Thermo Scientific, Wilmington, DE) was employed for the assessment of RNA integrity and concentration. Only those RNA samples that met the strict quality standards (with an OD₂₆₀/₂₈₀ ratio ranging from 1.8 to 2.0 and no obvious degradation) were subjected to library preparation. the 18 RNA libraries (9 samples per group) were constructed in a randomized manner, with samples from different groups evenly distributed across multiple construction batches to avoid batch-specific bias. Meanwhile, a reference RNA sample (from a pooled small intestine tissue sample) was included in each batch as an internal quality control to monitor batch-to-batch variation. The generation of the 18 RNA libraries was conducted in accordance with the instructions provided by the manufacturer with the mRNA-Seq Sample Preparation Kit (Illumina, San Diego, CA), and all reagents used for library construction (including enzymes, adapters, and reaction buffers) were from the same production lot to ensure standardized reaction conditions. Subsequently, the qualified RNA libraries were sequenced using the Illumina NovaSeq 6000 platform with 150 bp paired-end reads, and all libraries were sequenced in a single run to minimize sequencing batch effects.

Transcriptome data analysis

The quality control process was executed using Trim_Galore (https://github.com/FelixKrueger/Tri-mGalore/) with parameters “-q 20 –phred33 –stringency 3 –paired”. After quality control, the clean reads were aligned to the chicken reference genome (GRCg7b) using the HISAT2 (Kim, et al., 2019) with default parameters. The BAM files were subjected to sorting and indexing using Samtools (Li, et al., 2009). Count numbers for each gene were tallied using the htseq-count script in Python (van Dam, et al., 2017). Differential expression gene (DEG) analysis was conducted between the two groups using the DESeq2 (v4.2.2) tool (Love, et al., 2014). The genes that demonstrated differential expression were identified on the basis of |log2FoldChange| ≥1 and Q-value < 0.05.

Weighted gene co-expression network analysis

The WGCNA was conducted using the WGCNA package (v1.7.1)(Langfelder and Horvath, 2008) in the R software environment. Initially, the "hclust" function is utilized to test all samples for outliers. Next, to distinguish modules with different expression patterns, the best soft threshold power β (β = 1 to 20), which is obtained by the "pickSoftThreshold" algorithm, is used to generate the weighted adjacency matrix. Subsequently, the adjacency matrix was converted into a topological overlap matrix. Then, the cutreeDynamic function with the parameters "minModuleSize = 30, deepSplit = 2, pamRespectsDendro = F" was used for module detection, and cutHeight = 0.4 was used to merge modules. Finally, the quantitative analysis of the phenotype data associated with RFI and gene modules was conducted using Pearson's correlation coefficient. Significant consensus modules were identified based on the criteria of a correlation coefficient |R| ≥ 0.51 and a p-value < 0.05. The hub genes were selected from the significantly related modules in accordance with the following criteria: module membership (MM) > 0.8 and gene significance (GS) > 0.4 (Wei, et al., 2024), and then overlapped with DEG. Protein-protein interactions were then performed using the STRING tool (Szklarczyk, et al., 2023), and visualized using Cytoscape (v3.10.0).

Functional enrichment analysis

The online tool G: Profiler (Reimand, et al., 2016) was employed for the purpose of conducting a functional enrichment analysis for GO (Reimand, et al., 2016). The Benjamini-Hochberg false discovery rate (FDR) procedure was applied with an adjusted p-value<0.05.

Quantitative RT-PCR analysis

A total of 18 samples of the small intestine were employed for qPCR analysis. Specific primers for 9 selected genes were designed using NCBI (Supplementary table S12). The quantitative real-time polymerase chain reaction (qPCR) was conducted on an ABI Prism 7500 instrument (Applied Biosystems, Carlsbad, CA). GAPDH was used as an internal reference, and 3 replicates were performed for each sample and relative gene expression levels were calculated using the $2^{-\Delta \Delta Ct}$ method (Livak and Schmittgen, 2001).

Statistical analysis

The independent sample t-test in SPSS 26.0 software (SPSS Inc., Chicago, IL, USA) was utilized to compare the differences between LRFI and HRFI means. The results of qPCR and RNA-seq were analyzed by Pearson's correlation coefficient. A statistically significant difference was considered at P < 0.05.

Results and discussion

Compare the growth performance in different RFI groups

The growth performance of the two groups is presented in Table 1. Compared with the HRFI group, the LRFI group showed significantly lower ADFI (72.58 g/d vs. 88.46 g/d, P < 0.05) and RFI values (–8.07 g/d vs. 3.07 g/d, P < 0.05). In addition, the FCR was significantly higher in the HRFI group, exceeding that of the LRFI group by 0.69. These findings align with those reported in Wannan Yellow chickens (Yang, et al., 2020), suggesting that LRFI cocks may consume less feed while achieving greater weight gain. Furthermore, consistent with studies on Korat chickens (Kongthungmon, et al., 2025), no significant differences were observed between the two groups in BW49, BW91, or MMW (P > 0.05), indicating that RFI is independent of body weight. Overall, the results of this study provide a basis for RFI-based breeding strategies.

Table 1.

The production performance in different RFI groups in HuaiNan chickens from 49 to 91 days of age.

Item	HRFI	LRFI	p
BW49(g)	685.33 ± 27.74	686.67 ± 47.46	0.926
BW91(g)	1627.60 ± 78.76	1635.13 ± 86.38	0.805
ADG(g/d)	23.56 ± 1.68	23.71 ± 2.21	0.83
MMW(g)	198.28 ± 6.24	198.85 ± 6.90	0.816
ADFI(g/d)	88.46 ± 4.11	72.58 ± 4.83	<0.001
FCR(g/g)	3.76 ± 0.18	3.07 ± 0.16	<0.001
RFI(g/d)	8.14 ± 2.89	−8.07 ± 2.19	<0.001

BW₄₉: body weight at 49 days, BW₉₁: body weight at 91 days, ADG: average daily gain, ADFI: average daily feed intake, MMW: metabolic midweight.

Compare the slaughter performance in different RFI groups

Slaughtering performance is an important indicator for evaluating poultry productivity and economic benefits. The slaughter performance of two groups is shown in Table 2. The LRFI group has a significantly lower abdominal fat deposition (P<0.05), which suggests enhanced metabolic efficiency and a superior ability to convert feed into meat growth. This phenotypic difference may be associated with the high expression of PCK1 in the HRFI group, which is likely to enhance glucose metabolism and thereby promote abdominal fat deposition. This reduced fat accumulation indicates that genetic or management strategies optimizing RFI could improve poultry production efficiency by minimizing wasteful lipid deposition. Although no significant differences were observed between the groups in carcass yield, eviscerated yield, breast meat yield, or leg muscle yield (P > 0.05), all of these traits showed numerically higher mean values in the LRFI group. These findings are consistent with reports in other chicken populations, such as high-quality chicken a line (HQLA) and HB hybrid offspring from Guangdong Weizi Agricultural Technology Co., Ltd.(Zou, et al., 2023) and slower-growing broilers (Wen, et al., 2018). Collectively, the results support the breeding selection of LRFI chickens can effectively reduce abdominal fat deposition in roosters.

Table 2.

The slaughter performance in different RFI groups in HuaiNan chickens from 49 to 91 days of age.

Item	HRFI	LRFI	p
LW, g	1524.73 ± 74.11	1533.07 ± 79.55	0.769
Carcass Yield, %	89.29 ± 1.63	89.30 ± 1.49	0.99
Eviscerated Yield, %	67.27 ± 1.50	67.39 ± 1.39	0.819
Breast Meat Yield , %	13.51 ± 2.25	14.52 ± 2.23	0.152
Leg Meat Yield, %	20.55 ± 1.62	21.64 ± 1.40	0.106
Abdominal Fat Yield, %	2.95 ± 1.61	1.64 ± 1.47	0.027

LW: live weight.

Compare the meat quality in different RFI groups

Meat quality, a critical determinant of consumer acceptance, is commonly evaluated based on attributes such as color, pH, cooking loss, and shear force. PH serves as an important indicator of meat freshness. In agreement with Yang, et al. (2020), our study found that pH was not affected by RFI. However, Chen, et al. (2025) reported that the LRFI group exhibited significantly lower pH values in Huaibei Partridge chickens, suggesting that the effect of RFI on pH may vary across different breeds (Bai, et al., 2022). Meat color, characterized by the L* (lightness), a* (redness), and b* (yellowness) values, showed no significant differences (P>0.05) between the LRFI and HRFI groups in this study（Table 3), indicating that RFI did not affect flesh color. Cooking loss, which reflects water-holding capacity and is associated with nutrient retention and flavor integrity, was numerically lower in the LRFI group, though not statistically significant, potentially indicating a higher water-holding capacity. Shear force is a pivotal indicator in the evaluation of muscle tenderness. The LRFI group demonstrated a lower shear force compared to the HRFI group, suggesting that the meat from the LRFI group is more tender.

Table 3.

The meat quality in different RFI groups in HuaiNan chickens from 49 to 91 days of age.

Item	HRFI	LRFI	p
pH45	6.34 ± 0.20	6.32 ± 0.31	0.807
L*	44.06 ± 2.12	42.48 ± 2.52	0.074
a*	6.10 ± 0.82	6.60 ± 0.87	0.122
b*	12.79 ± 1.56	13.57 ± 1.58	0.185
Cooking loss, %	21.35 ± 5.94	18.48 ± 3.31	0.117
Shear force, N	26.33 ± 8.63	24.22 ± 8.83	0.514

pH₄₅: value measured 45 min after slaughter.

Summary of transcriptome data

The transcriptome sequencing of the small intestines from 18 Huainan chickens identified 23,097 genes, which were used to identify genes involved in the regulation of RFI. After filtration, the number of clean reads per sample ranged from 20,029,883 to 33,149,224. A minimum of 97.05 % of Q20 and a minimum of 91.87 % of Q30 for each sample. The guanine and cytosine (GC) content of each sample was found to fall within the range of 47.74 % to 50.68 % and the percentage of mapped reads ranged from 85.10 % to 93.46 % (Supplementary table S2).

Differentially gene expression and functional analysis

The obtained transcriptome data were subjected to differential gene expression analysis (Supplementary table S3). The volcano diagram illustrates the DEGs between the HRFI group and LRFI group (Fig. 1A). In comparison to LRFI group, 170 genes demonstrated a downregulation, while 109 genes exhibited an upregulation in HRFI group. Among the examined genes, several genes, including cluster of differentiation 36 (CD36) and ATP-binding cassette transporter G8 (ABCG8), were significantly enriched in the processes of “intestinal absorption” and “digestive system process” (Supplementary table S4). Upregulation of CD36 is known to enhance intestinal lipid absorption (Li, et al., 2022), whereas ABCG8 forms obligate heterodimers that reduce intestinal cholesterol and phytosterol uptake while promoting their biliary excretion (Yu, et al., 2014). In the present study, elevated expression of ABCG8 in the HRFI group suggests a reduction in intestinal absorption in these individuals. In addition, Glutamate-ammonia ligase (GLUL) and phosphoenolpyruvate carboxykinase 1 (PCK1) were significantly enriched in “Metabolic pathways”. GLUL has been closely associated with lipid deposition (Peng S, et al., 2023). Upregulation of GLUL in the HRFI group indicates a greater capacity for lipid deposition, which is consistent with our experimental findings of increased abdominal fat in this group. Collectively, these results imply that the LRFI group exhibits more efficient feed conversion and absorption.

Fig. 1.

The results of differential gene expression analysis of 18 small intestine samples. (A) Volcano plot of DEG between HRFI and LRFI. (B) Partial differential genes enrichment pathway.

Key genes identification and PPI analysis

To identify the hub genes that regulate RFI traits, a WGCNA analysis was performed. In this study, data from 18 small intestinal transcriptomes were used to construct expression matrices. A total of 23097 genes were obtained to build the weighted gene co-expression network. Based on the Optimal Soft Threshold (β=7) we selected (Fig. 2A), the differential genes with similar expression patterns were grouped using Average Link Hierarchical Clustering (Fig. 2B). The cluster tree of module eigenvalues was constructed through the calculation of the correlation coefficient among modules. Following the implementation of dynamic tree trimming, a total of 27 co-expression gene modules were identified. Among these modules, the grey modules contain the fewest number of genes, 20, and the black modules contain the most, 4,196 (Supplementary table S5).

Fig. 2.

WGCNA analysis of 18 intestine samples. (A) Selection of the optimal soft threshold (B) The cluster dendrogram. (C) A correlation analysis of the modules and traits.

In order to identify co-expression modules associated with RFI traits, the relationship between RFI and module eigengene (ME) was evaluated (Fig. 2C). The magenta module (R = 0.66, p = 0.003) exhibited a significantly positive correlation with RFI traits, whereas the pink module (R=﹣0.6,p = 0.009) displayed a significantly negative correlation. A total of 3,557 genes and 646 genes were identified in the magenta module and the pink module, respectively (Supplementary table S5). Based on MM > 0.8 and GS > 0.4, the 559 hub genes were identified in the magenta module (Supplementary table S6) and 139 hub genes were identified in the pink module (Supplementary table S7). In order to guarantee the veracity of the results, the correlation between MM and GS was calculated for each module individually. (Fig. 3A and Fig. 3C). A total of 80 hub genes were found to be up-regulated in the magenta module, while 3 hub genes were down-regulated. In contrast, the pink module exhibited a different expression pattern, with 3 hub genes being down-regulated and no hub genes showing up-regulation. In order to enhance comprehension of the mechanism by which these genes regulate RFI, a Gene Ontology (GO) enrichment analysis of hub genes was conducted. The results of the GO analysis indicate that the hub genes in the magenta module were significantly enriched in 177 GO terms (Supplementary table S9), including “oxidoreductase activity”, “organic substance transport”, “organic substance catabolic process” and “endomembrane system” (Fig. 3B). And the hub genes in the pink module were significantly enriched in 47 GO terms (Supplementary table S8), including “system development”, “multicellular organism development” and “regulation of cell motility” (Fig. 3D).

Fig. 3.

Selection and enrichment of key genes in magenta module and pink module. (A) Scatter plot of gene significance in magenta module. (B) Partially enriched pathways in magenta module. (C) Scatter plot of gene significance in pink module. (D) Partially enriched pathways in pink module.

To identify key regulatory genes associated with RFI, we intersected hub genes with differentially expressed genes, which led to the identification of 83 key genes in the magenta module and three in the pink module (Supplementary table S10). The present study concentrated on the magenta module. A total of 65 annotated genes from the magenta module were included in a PPI network analysis performed using the STRING database. The minimum required interaction score for the STRING database was set to 0.600 for high confidence. After excluding genes deemed unrelated, the PPI network was visualized using Cytoscape software (version 3.10.2) (Fig. 4A). RFI is a complex quantitative trait influenced by multiple factors, including digestive, metabolic, biosynthetic, and energy homeostasis processes (Yi, et al., 2015). Therefore, the genes regulating RFI are likely involved in analogous biological processes. Among the key genes, several, including Angiotensin Converting Enzyme-2 (ACE2) and Aminocarboxymuconate Semialdehyde Decarboxylase (ACMSD), were significantly enriched in “Small molecule metabolic process” and “Metabolic pathways” (Fig. 4B and Supplementary table S11). ACE2 plays a critical role in thermogenesis and energy expenditure, activating mitochondrial function through the Akt/FoxO1 and PKA pathways (Cao, et al., 2022). Additionally, ACE2 modulates cellular proliferation by regulating angiotensin II (Ang II) activity (Gong, et al., 2023). ACMSD regulates nicotinamide adenine dinucleotide (NAD) synthesis, and its inhibition promotes NAD production and enhances mitochondrial function (Katsyuba, et al., 2018). In this study, ACE2 and ACMSD were significantly upregulated in the HRFI group, which is indicative of enhanced mitochondrial function. This elevation in mitochondrial activity likely increases overall energy expenditure—whether allocated for thermogenesis or other metabolic processes—resulting in a relative reduction in the energy available for somatic growth and thereby impairing feed efficiency in roosters. Furthermore, fatty acid binding protein 2 (FABP2), fetuin B (FETUB) and PCK1 were significantly enriched in “Triglyceride metabolic process”, “PPAR signaling pathway” and “Glycolysis / Gluconeogenesis” (Fig. 4B and Supplementary table S11). FABP2 influences fatty acid β-oxidation in mitochondria by enhancing enzyme activity and plays a role in fatty acid uptake, thereby affecting glucose metabolism (Rubin, et al., 2012). It also facilitates the uptake, transport, and utilization of dietary fatty acids, promoting fat oxidation by binding saturated and unsaturated long-chain fatty acids with high affinity (Baier, et al., 1995). FABP2 may affect RFI by influencing fatty acid uptake. FETUB reduces insulin-stimulated glucose uptake in myotubes in a time- and dose-dependent manner, and systemic glucose metabolism improves with reduced circulating fetuin B levels in obese mice (Meex, et al., 2015). PCK1 inhibits adipogenesis and lipid synthesis-related gene expression (Ye, et al., 2023). PCK1 is the rate-limiting enzyme for gluconeogenesis, and the protein kinase activity of PCK1 is of significant importance with regard to the process of adipogenesis (Xu et al., 2020). Furthermore, some studies have found a positive correlation between PCK1 expression and abdominal fat accumulation in roosters (Chowen, et al., 2013), which is consistent with our experimental results: the upregulation of PCK1 expression in the HRFI group led to significantly more abdominal fat in male roosters compared to the LRFI group. The upregulation of FABP2, FETUB, and PCK1 in the HRFI group affected the glucose metabolism of roosters, thereby affecting their feed efficiency and abdominal fat. Moreover, apolipoprotein AI (APOA1) is a regulatory factor in lipoprotein metabolism and lipid homeostasis. (Xiao, et al., 2024). The study on the transcriptome of broiler chicken livers found a significant negative correlation between APAO1 expression and abdominal fat deposition (Pirany, et al., 2019), which contradicts the results of this study. This discrepancy may be due to differences in the tissues selected and the grouping methods used. Among these six candidate genes, PCK1 (Chowen, et al., 2013) and APOA1 (Pirany, et al., 2019) have been confirmed to be associated with feed efficiency, while FABP2 (Vigors, et al., 2016) may affect feed efficiency by influencing intestinal absorption. No relevant research has been found for the remaining three genes. Collectively, these findings underscore the multifunctional roles of core genes within metabolic pathways and their potential regulatory contributions to RFI in poultry. Furthermore, our results demonstrate that selection for LRFI not only confers economic benefits by reducing feed costs but also exerts a pronounced effect on mitigating fat deposition in roosters.

Fig. 4.

protein-protein interaction networks of the hub genes and GO functional enrichment analysis of hub genes. (A) Protein-protein interaction networks of key genes with connecting nodes. (B) GO functional enrichment analysis of key genes.

Quantitative real‑time PCR validation

To further confirm the accuracy of our screening results, we conducted qPCR analysis. The expression trends of these genes, measured by log2fold-change in both RNA-Seq and qPCR, were found to be consistent (Fig. 5A). The relative expression levels (log2fold-change) of the 9 genes from RNA-Seq and qPCR were further analyzed using Pearson correlation in SPSS software, revealing a high correlation coefficient of 0.9744 (Fig. 5B). These results confirm the accuracy and reliability of our transcriptome data.

Fig. 5.

(A) Comparisons between qPCR and RNA-seq measurements of the expression abundance of 9 genes. (B) Correlations of the mRNA expression levels of 9 genes based on RNA-seq and qPCR analyses.

Conclusion

The present study compared the production performance, slaughter performance and meat quality in different RFI groups. The results showed that a significantly lower abdominal fat yield was observed in the LRFI group; however, no significant differences were observed in other meat quality traits or slaughter performance.

A number of genes were identified that exert a significant influence on RFI. These genes are involved in a variety of physiological processes, including mitochondrial function (e.g., ACE2, ACMSD), glucose metabolism (e.g., FABP2, FETUB, PCK1), and lipid metabolism (e.g., APOA1). The identification of these key regulatory genes provides a foundation for further research into the molecular networks underlying RFI and supports the development of molecular breeding strategies to improve feed efficiency in chickens.

Data availability

The raw sequence data reported in this paper have been deposited in the Genome Sequence Archive (Genomics, Proteomics & Bioinformatics 2025) in National Genomics Data Center (Nucleic Acids Res 2025), China National Center for Bioinformation / Beijing Institute of Genomics, Chinese Academy of Sciences (GSA: CRA053043) that are publicly accessible at https://ngdc.cncb.ac.cn/gsa.

CRediT authorship contribution statement

Hao Wang: Writing – review & editing, Writing – original draft, Visualization, Methodology, Funding acquisition, Formal analysis, Conceptualization. Zihan Chen: Writing – original draft, Visualization, Validation, Methodology, Formal analysis, Data curation. Chuchu Zhang: Validation, Methodology, Formal analysis, Data curation. Wei Wei: Data curation. Yanghao Liu: Validation. Chaohui Xing: Visualization. Aofan Zou: Visualization, Validation. Jiansheng Cheng: Validation. Runshen Jiang: Writing – review & editing, Writing – original draft, Funding acquisition, Conceptualization.

Disclosures

The authors confirm that there are no conflicts of interest.