Altered expression of metabolic pathways and immune-related genes revealed by RNA sequencing and metabolomic analysis in meat-type chickens with different growth rates

Abstract

Differences in growth rates among broiler chickens within a single commercial genetic line have important economic implications; however, their molecular basis remains incompletely understood. This study analyzed male Ross 308 broilers classified into early- and late-growth lines based on weight gain from 1 to 5 days of age. We integrated RNA sequencing, metabolomics based on gas chromatography–mass spectrometry, and exploratory SNP analysis of pectoralis major muscle tissue collected at 35 days of age. Our integrative analysis revealed contrasting energy utilization programs. Slow early-growth phenotype chickens showed a Warburg-like metabolic profile characterized by glycolytic reliance, lactate fermentation, ketone metabolism, and enhanced proteolysis, accompanied by a bottleneck in mitochondrial oxidative phosphorylation. In contrast, fast early-growth phenotype chickens displayed enhanced oxidative phosphorylation, elevated glycerol-3-phosphate levels, and coordinated activation of pathways related to mitochondrial function and immune responses. Notably, reduced CARNS1 expression in the fast early-growth group suggested a potential trade-off with muscle quality, consistent with the role of carnosine in pH buffering and maintaining redox balance. Multi-omics integrated analysis revealed coordinated changes in metabolites and gene expression within glycolysis, lipid metabolism, and mitochondrial pathways. These findings indicate that the weight gain phenotype during early growth is associated with specific transcriptional and metabolic states during later development.

Introduction

Selective breeding produces broilers with rapid growth performance and high meat yield, making them a major global source of animal protein. Modern broilers attain market weight in less than 6 weeks (Tallentire et al., 2016) and now account for nearly 40% of global meat production (FAO, 2024). However, considerable inter-individual variations in body weight and muscle development remain (Galosi et al., 2023; Vasdal et al., 2019), even within the same genetic line. These variations often prolong the rearing period beyond the standard market age, thereby increasing feed and overall production expenses (de Verdal et al., 2011).

The determinants of growth differences can be broadly divided into environmental factors, including external stressors and feed intake, and genetic factors that contribute to phenotypic variations. Skeletal muscle mass, the primary determinant of edible yield, is determined by protein turnover involving the coordinated regulation of protein synthesis and degradation (Russell, 2010). In growing animals, muscle accretion reflects the predominance of protein synthesis, with protein degradation largely associated with normal tissue turnover (Goldspink, 1976; Goldspink and Goldspink, 1977). However, the molecular basis of growth variation remains poorly defined due to complex interactions among genetic, transcriptional, and metabolic processes.

Previous nutritional and metabolomic studies in broilers have indicated that growth rate differences are closely associated with muscle protein turnover and branched-chain amino acid (BCAA) metabolism (Kang et al., 1985; Kim et al., 2022; Musigwa et al., 2025). In particular, enhanced BCAA catabolism and sustained protein degradation have been suggested to limit protein accretion, thereby constraining growth. However, most of these studies have focused on metabolite profiles or systemic nutritional indicators and have not fully addressed the contributions of tissue-specific gene expression, alternative metabolic pathways, immune responses, or underlying genetic variations (Selle et al., 2023).

To understand growth variations, it is necessary to examine not only individual gene expression levels but also their coordinated functions within metabolic networks. Pathways related to glycolysis, carbohydrate metabolism, and immune regulation have been implicated in distinct energy-allocation strategies in chickens exhibiting fast and slow early-growth phenotypes. Given the interdependence of metabolism and immunity, differences in these pathways may critically affect long-term growth efficiency. In this study, we integrated RNA sequencing, metabolomic, and whole-genome analyses to identify key biological processes associated with growth differences. By clarifying the regulatory roles of specific genes and metabolites, our study provides new insights into the biological interactions underlying growth variation and offers potential strategies for improving poultry production efficiency and sustainability.

Materials and methods

Animal samples, total RNA purification, and metabolome sample preparation

All experimental protocols and procedures were reviewed and approved by the Experimental Animal Committee of Kagoshima University (Approval No. A19002). One hundred male Ross 308 broilers were reared in groups under identical environmental conditions, with feed and water provided ad libitum. Birds were housed in wooden cages with wire floors (1,200 × 3,000 × 400 mm) in a temperature-controlled room maintained at 25°C. The cage was equipped with two feeders (2-kg capacity each) and 12 nipple drinkers. A commercial starter diet (Nichiwa Sangyou Co., Hyogo, Japan), formulated according to National Research Council (1994) specifications (≥23% crude protein, 3.1 Mcal/kg metabolizable energy), was provided until 21 days of age. From 21 to 35 days of age, birds were fed a grower diet formulated according to the Japanese Feeding Standard for Poultry (MAFF, 2011). Body weight was recorded at 1 and 5 days of age to calculate early body weight gain. Differences in body weight and body weight gain between the fast and slow early-growth phenotype groups were evaluated using Student’s t-test. At 5 days of age, chicks were categorized into two groups based on their weight gain: the fast early-growth phenotype group (top 20%) and the slow early-growth phenotype group (bottom 20%). Classification at 5 days of age was performed to capture early-growth phenotypes prior to the establishment of social hierarchy, rather than to define final growth performance. Chicks were group-reared in a single cage (40 birds per cage, consisting of both fast and slow early-growth phenotype chicks) until 21 days of age to maintain uniform social conditions, after which seven birds per group were randomly selected and continuously reared under the same conditions. Stocking density was controlled to maintain approximately 10 kg/m² throughout the rearing period. Seven birds per group were used for transcriptomic and metabolomic analyses, and a subset of four birds per group was used for exploratory SNP analysis. Each bird was treated as the experimental unit in all statistical analyses. At 35 days, birds were euthanized by decapitation following carbon dioxide inhalation. Pectoralis major muscles (breast muscles) were collected, snap frozen in liquid nitrogen, and stored at −80°C until use for total RNA extraction and untargeted metabolomic analyses. The breast muscle was selected because it represents the primary site of skeletal muscle growth and edible yield in broiler chickens and therefore provides a relevant tissue for investigating muscle-specific metabolic and transcriptional features associated with growth variation.

Gene expression analysis using RNA-Seq

Total RNA was extracted using an RNeasy Fibrous Tissue Mini Kit (Qiagen, Hilden, Germany). For RNA-seq analysis, TruSeq cDNA libraries were constructed using the NEBNext® Ultra™ Directional RNA Library Prep Kit (Illumina, USA) after confirming the quantity and quality of the extracted RNA. Raw RNA-seq data were acquired using the Illumina NovaSeq 6000 platform. For some samples, publicly available data were retrieved from the database and combined with newly generated datasets (PRJDB17938; Ishihara et al., 2025). Sample information is provided in Table S1. Raw sequencing reads were processed using Trimmomatic program version 0.36 (Bolger et al., 2014) to remove adapter sequences and low-quality reads. STAR version 2.7.10a (Dobin et al., 2013) was used to map reads to the chicken reference genome (bGalGal1.mat.broiler.GRCg7b). For expression analysis, the RSEM v1.3.1 program (Li and Dewey, 2011) was used to estimate read counts. The expression counts were normalized to DESeq2, an R package for RNA-seq data analysis (Love et al., 2014), implemented in TCC-GUI, a graphical user interface for the Tag Count Comparison (TCC) R package (Su et al., 2019). Genes with fewer than 30 total counts across all samples were excluded (Sultan et al., 2008). Differentially expressed genes (DEGs) were identified based on a false discovery rate (FDR) < 0.1.

Sample preparation for gas chromatography/mass spectrometry (GC–MS/MS) analysis

Metabolomic analysis was performed using GC/MS, as previously described (Katafuchi et al., 2023; Shigematsu et al., 2018), with some modifications. Approximately 20 mg of freeze-dried breast muscle sample was suspended in 250 μL of methanol–chloroform–water (5:2:2), with 5 μL of 1 mg/mL 2-isopropylmalic acid added as an internal standard. The hydrophilic fraction was extracted and derivatized using methoxyamine hydrochloride and N-methyl-N-(trimethylsilyl)trifluoroacetamide. To prepare trimethylsilyl derivatives, the tubes were shaken at 1200 × g at 37°C for 45 min in the dark. Metabolome analysis was performed using a GCMS-QP2010SE and the Smart Metabolites Databases (both from Shimadzu Corporation, Kyoto, Japan) to identify characteristic major metabolites. A 30 m × 0.25 mm i.d. DB-5 with a film thickness of 1.00 μm (GL-Science, Tokyo, Japan) was used as the GC column. Selected ion monitoring conditions from the Smart Metabolites Database were used for GC-MS measurements.

Data processing for untargeted metabolomics

Samples were processed using a previously described method (Shimamoto et al., 2020). Chromatogram acquisition, mass spectral peak detection, and waveform and data processing were performed using Shimadzu GC/MS Solution software (Shimadzu, Kyoto, Japan) and the MRMPROBS program version 2.60 (Tsugawa et al., 2014). Data quality was assessed automatically using the following thresholds implemented in the metabolomics data processing pipeline: −10 < Retention Index < 10, dot production > 0.8, and presence > 0.6. The remaining data were manually checked. In total, 92 metabolites were identified in the breast muscle samples. Metabolite levels were semi-quantified using the peak area of each metabolite relative to that of the internal standard (2-isopropylmalic acid). Between-group differences in metabolite levels were evaluated using Student’s t-test, and FDR correction was applied to account for multiple testing.

Multi-omics integration using DIABLO with block sparse partial least squares discriminant analysis (Block sPLS-DA)

To integrate transcriptomic and metabolomic datasets, RNA-seq count data were normalized separately from the DEG analysis to ensure comparability across data types. Specifically, raw read counts were converted to counts per million (CPM) by dividing each gene count by the total number of mapped reads in the sample (library size) and multiplying by one million. To stabilize the variance and approximate a normal distribution, log₂(CPM + 1) values were calculated. Correlation analysis between transcriptomic and metabolomic features was performed using the mixOmics R package (version 6.30.0) (Rohart et al., 2017). To focus on biologically relevant gene–metabolite relationships, we selected DEGs involved in metabolic pathways as defined by Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG), and Reactome categories related to carbohydrate, energy, and amino acid metabolism (Table S2). The metabolomic data were preprocessed to remove missing values, and the input data were structured as a list comprising two blocks: mRNA expression data and metabolite concentrations. A multivariate integration model was constructed using block sPLS-DA implemented with the block.splsda function. The model design matrix was set with a uniformly low correlation between blocks (0.1 off-diagonal). Model tuning was performed using 5-fold cross-validation (10 repeats) to select the optimal number of components and features per block based on centroid distance.

Pathway enrichment analysis using over-representation analysis (ORA) and gene set enrichment analysis (GSEA)

DEGs identified after filtering were analyzed using a combined approach of ORA and GSEA across the GO, KEGG, and Reactome databases. For GO, enrichment analysis was performed for all DEGs, as well as separately for upregulated and downregulated genes, focusing on the Biological Process category to highlight functionally distinct pathways. To determine pathway directionality, GSEA was applied to the entire ranked gene list, with normalized enrichment scores (NESs) used to indicate enrichment in fast or slow early-growth phenotype chickens. The ORA results for upregulated and downregulated DEGs were included as complementary evidence. To summarize functionally relevant pathways, enrichment results from GO, KEGG, and Reactome analyses were further annotated into metabolic or immune categories. For visualization, the top 10 pathways per database were selected based on adjusted P-values, and unrelated terms were excluded. All analyses were performed using the ClusterProfiler package version 4.10.1 (Wu et al., 2021; Xu et al., 2024; Yu et al., 2012; Yu, 2024) in R, with gene annotation from org.Gg.eg.db version 3.20. KEGG and Reactome analyses were conducted using ClusterProfiler and the corresponding databases. P-values were adjusted using the Benjamini–Hochberg method, and pathways with an adjusted P < 0.05 were considered significant. In GSEA, a positive NES indicates that gene sets ranked higher in slow early-growth phenotype chickens, whereas a negative NES indicates enrichment in fast early-growth phenotype chickens.

SNP detection between fast and slow early-growth phenotype chickens

Whole-genome sequencing was performed using an Illumina NovaSeq 6000 platform, generating paired-end reads (150 bp) at an average depth of approximately 45 × . For the present study, we reanalyzed these previously generated data, which were deposited in the DDBJ under accession number PRJDB17938 (Table S3, Ishihara et al., 2025). These eight individuals (slow early-growth phenotype chickens, n = 4; fast early-growth phenotype chickens, n = 4) were the same as those used for transcriptomic and metabolomic analyses in this study, enabling a direct comparison among the three types of data. The resulting reads were aligned to the chicken reference genome GRCg7b using the Burrows–Wheeler Aligner MEM algorithm version 0.7.15 (Li, 2013). Variant calling was conducted using the Genome Analysis Toolkit (GATK) Best Practices pipeline (DePristo et al., 2011; McKenna et al., 2010; Poplin et al., 2017; Van der Auwera et al., 2013). Specifically, we used Samtools v1.3.1 (Li, 2009), Picard v2.13.3 (Broad Institute, 2019, available at https://broadinstitute.github.io/picard/), and GATK HaplotypeCaller (v4.2.2) in Genomic Variant Call Format (GVCF) mode, followed by joint genotyping using GATK GenotypeGVCFs. Variants were filtered using GATK VariantFiltration with the following parameters: “QD < 2.0 || FS > 60.0 || MQRankSum < −12.5 || MQ < 40.0 || ReadPosRankSum < −8.0″. Functional annotations for each variant were assigned using SnpEff v5.0 (Cingolani et al., 2012) with gene models corresponding to the GRCg7b reference genome. To identify variants associated with growth rate, a case–control association analysis was performed between fast and slow early-growth phenotype chickens using PLINK v1.9 (Purcell et al., 2007). A permutation-based association test with 10,000 permutations was performed using default settings. We focused on genes associated with significantly enriched pathways and selected those harboring variants with significant differences in allele frequency between the fast and slow early-growth phenotype chickens. The SNP-based pathway analysis was conducted in an exploratory manner. Pathway information was used to assist interpretation, and the primary conclusions of this study relied on transcriptomic and metabolomic evidence.

Results

Final weights of slow early-growth phenotype and fast early-growth phenotype chickens

Chickens were classified into slow and fast early-growth phenotype groups based on body weight gain at 5 days of age. Body weight remained significantly higher in the fast early-growth phenotype group than in the slow early-growth phenotype group at all subsequent measurement points (P < 0.05), including at 35 days of age (Table 1).

Table 1.

The body weight of chickens from introduction to final measurement.

Phenotype	Day										Weghit gain
	1	5*	6	7	10	15	18	21	28	35
Slow-growth	41.6 ± 1.6	76.8 ± 3	93.1 ± 3.9	109.7 ± 5.3	194.2 ± 12.1	408.4 ± 22.1	623.1 ± 28.6	880.3 ± 40.1	1531.8 ± 61.7	1973.7 ± 82.6	1932.1 ± 81.2
Fast-growth	45.4 ± 1.3	109.8 ± 1.4	130.8 ± 2.1	151.5 ± 2.1	275 ± 6.2	525.4 ± 14.1	775.6 ± 27.8	1058.8 ± 27.6	1762.3 ± 69.4	2249.9 ± 80.1	2204.5 ± 79.9
*Selected into the Slow-growth group or Fast-growth group based on the weight at this point.
Values are expressed as mean ± standard error of the mean (SEM) (n = 7).

DEGs detected using RNA-seq analysis

In total, 25,466 genes were examined using TCC-GUI. After filtering for low-count genes (Sultan et al., 2008), 15,007 genes remained for analysis. Among these, 1,648 genes were identified as differentially expressed (FDR < 0.1; Table S4). Volcano plot analysis revealed that the expression levels of 102 genes were significantly upregulated (P < 0.05, log₂(fold-change) > 1) in slow early-growth phenotype chickens, whereas those of 480 genes were significantly downregulated (P < 0.05, log₂(fold-change) < −1) in fast early-growth phenotype chickens (Fig. 1A, top left). Notably, the expression of CARNS1, a gene encoding carnosine synthase, was higher in slow early-growth phenotype chickens than in fast early-growth phenotype chickens, which is consistent with the potential differences in muscle buffering and antioxidant capacities between the growth phenotypes.

Fig. 1.

Transcriptomic, metabolomic, and integrative analyses of breast muscle from fast- and slow-growing chickens. (A) Volcano plot of differentially expressed genes (DEGs) identified by RNA-seq (FDR < 0.1). The x-axis shows log₂(fold-change) (slow/fast), and the y-axis shows the −log₁₀(FDR). Blue and red dots indicate significantly downregulated and upregulated genes, respectively, in slow early-growth phenotype chickens compared to those in the fast early-growth phenotype group. Gray dots represent genes not reaching significance. Dashed lines denote the thresholds for statistical significance. (B) Volcano plot of metabolites detected by GC–MS/MS analysis. The x-axis shows the log₂(fold-change) (slow/fast), and the y-axis shows the −log₁₀ (nominal P-value). Red and blue dots indicate metabolites with higher levels in slow and fast early-growth phenotype chickens, respectively, based on a nominal P < 0.05 (unadjusted). Gray dots represent metabolites that do not meet this threshold. Dashed lines denote a |log₂(fold-change)| = 1 (vertical) and P = 0.05 (horizontal). (C) Block sparse partial least squares discriminant analysis (block sPLS-DA) of metabolomic data (left) and transcriptomic data (right). Each point represents an individual sample, with orange triangles indicating fast early-growth phenotype and blue circles indicating slow early-growth phenotype chickens. Ellipses represent the 95% confidence intervals for each group. The clear separation of groups is visible along the first two principal components. (D) Integrative network analysis linking differentially expressed genes (blue) and metabolites (green). Edges represent significant correlations, with red lines indicating positive and blue lines indicating negative associations.

Metabolome analysis

Ninety-two metabolites were profiled in skeletal muscle. At the nominal (unadjusted) level, a t-test analysis identified significant between-group differences, notably for glycerol-3-phosphate (higher in fast early-growth phenotype chickens) and acetoacetic acid (higher in slow early-growth phenotype chickens) (Fig. 1B, top right; Table S5), which are metabolites linked to glycolysis/lipid metabolism and ketone metabolism, respectively. Other metabolites, such as cysteine, also showed trends toward between-group differences at the nominal level (fig. 1B). However, none of the metabolites exceeded the FDR correction threshold (Table S5). Therefore, metabolite trends were interpreted in conjunction with RNA-seq and multi-omics (DIABLO; block sPLS-DA) integration rather than as standalone significant effects.

Multi-omics integration

To relate transcriptomic and metabolomic changes, block sPLS-DA analysis was performed. Both datasets clearly distinguished fast and slow early-growth phenotype chickens (Fig. 1C, bottom left). An integrative network analysis further revealed coordinated changes between metabolites and DEGs, such as associations between glycerol-3-phosphate and genes linked to mitochondrial electron transport (e.g., DHODH) and between acetoacetic acid and genes involved in energy metabolism (Fig. 1D, bottom right). These results indicate that altered coordination among lipid utilization, ketone metabolism, and mitochondrial function may underlie the observed growth rate differences.

Functional enrichment analysis of DEGs

Pathway enrichment analysis of DEGs revealed clear differences between fast and slow early-growth phenotype chickens, notably in metabolism, immune processes, and protein degradation (Fig. 2; Table S6). Central carbon metabolism (e.g., “generation of precursor metabolites and energy”) was among the GO Biological Process terms enriched in slow early-growth phenotype chickens. Consistently, the KEGG pathways carbon metabolism (gga01200), citrate cycle (tricarboxylic acid [TCA] cycle; gga00020), and pyruvate metabolism (gga00620) were enriched in slow early-growth phenotype chickens. In rank-based GSEA (slow vs. fast), a positive NES indicated enrichment in slow early-growth phenotype chickens; accordingly, these pathways exhibited positive NES values, indicating the transcriptional upregulation of their constituent genes. In contrast, transcripts encoding distal electron transfer steps (COX/complex IV subunits) and DHODH were less abundant in slow early-growth phenotype chickens (Fig. 2; Table S6). In fast early-growth phenotype chickens, immune-related pathways, including Phagosome (gga04145) and cytokine–cytokine receptor interaction (gga04060), were enriched, in addition to GO terms such as “regulation of immune system process” and “leukocyte activation” (Fig. 2; Table S6).

Slow early-growth phenotype chickens also exhibited higher expression of genes involved in protein catabolism, consistent with the enrichment of the KEGG ubiquitin-mediated proteolysis (gga04120) pathway (Fig. 2, Table S6). These genes overlapped with the Reactome enrichment of “Antigen processing: Ubiquitination & Proteasome degradation,” further supporting enhanced protein degradation in slow early-growth phenotype chickens. In addition to ubiquitin-mediated proteolysis, the expression levels of several autophagy-related genes were upregulated in slow early-growth phenotype chickens, further supporting enhanced protein degradation activity. In contrast, Reactome analysis of fast early-growth phenotype chickens showed strong enrichment of immune-related pathways, including the innate immune system, GPCR ligand binding, and chemokine receptor signaling.

Fig. 2.

Enriched pathways identified by GO, KEGG, and Reactome analyses of differentially expressed genes.

The top 10 enriched pathways per database are shown. The x-axis represents the normalized enrichment score (NES), where values ≥ 0 indicate enrichment in slow early-growth phenotype chickens, and values ≤ 0 indicate enrichment in the fast early-growth phenotype group. The y-axis lists significantly enriched terms from Gene Ontology (GO, top), KEGG pathway (middle), and Reactome pathway (bottom) analyses. Each dot represents an enriched pathway, with dot size corresponding to −log₁₀(FDR) and the dot color indicating the system category (red = immune-related pathways; blue = metabolism-related pathways).

To further characterize muscle energy metabolism, we examined gene expression patterns in three major modules using heatmap analysis (Fig.s S1). First, electron transport chain (ETC)-related genes showed a mixed pattern. Specifically, the expression levels of NDUFS1 (complex I) and CoQ biosynthesis genes (e.g., COQ3 and COQ10B) were upregulated in slow early-growth phenotype chickens, whereas those of COX subunits (complex IV) and DHODH were downregulated, suggesting a potential imbalance between electron input and output. Second, genes associated with lactate fermentation were differentially expressed. The KEGG pyruvate metabolism pathway (gga00620) was significantly enriched in slow early-growth phenotype chickens. LDHA, which encodes a protein that catalyzes the conversion of pyruvate to lactate, was expressed at high levels, indicating enhanced lactate fermentation (Fig. S1). Third, lipid metabolism-related genes showed group-specific patterns (Fig. S1). The slow early-growth phenotype chickens exhibited upregulated expression of glycerol-3-phosphate pathway genes (GPD1 and GPD2), whereas fast early-growth phenotype chickens showed upregulated expression of LPL, ACAA2, and PPARG.

SNP analysis

In total, 42 genes harboring variants with significantly different allele frequencies were identified between the fast and slow early-growth phenotype chickens (Table S7). Most annotated variants were classified as the MODIFIER type, which included non-coding transcript variants, upstream gene variants, and downstream gene variants as primary subcategories (Fig. S2). The number of transcript isoforms per gene ranged from 1 to 17. Although variant profiles were generally consistent across transcripts within the same gene, minor differences were observed. Therefore, the average number of variants per gene was calculated and used for visualization. Genes, such as ROCK2 and NCOA1, had relatively high average counts of MODIFIER variants, with consistent values across multiple isoforms.

Discussion

In this study, we integrated transcriptomic and metabolomic datasets, supplemented by a genome-wide SNP screen, to investigate the molecular basis of early growth divergence in broiler chickens, with a specific focus on skeletal muscle metabolism. This combined multi-omics framework revealed consistent shifts in energy-utilization strategies that distinguish fast and slow early-growth phenotype chickens. Previous studies have examined growth-related metabolic regulation in organs such as the liver and intestine, highlighting their roles in nutrient absorption, energy partitioning, and systemic metabolic homeostasis. For example, dynamic metabolic remodeling of the chicken liver during the post-hatch period has been reported as chicks transition from yolk-derived to feed-derived nutrient utilization (Van Every and Schmidt, 2025). In addition, coordinated changes in gut microbiota and hepatic gene expression under energy deprivation and refeeding conditions demonstrate functional intestine–liver interactions in regulating energy metabolism in poultry (Wang et al., 2025). In contrast, the present study focused on skeletal muscle to directly assess tissue-level processes related to muscle growth and protein accretion. Integrating muscle-specific findings with multi-organ analyses will be an important direction for future studies.

SNP analysis was performed in a subset of four fast and four slow early-growth phenotype chickens and revealed several variants in genes with known regulatory functions, including ROCK2 and NCOA1. Due to the limited sample size, no formal association analysis was performed, and these variants are therefore considered exploratory. They are presented to complement the RNA-seq results rather than to serve as independent evidence for the main conclusions. It should also be noted that classification at 5 days of age was used to define early-growth phenotypes and was not intended to reflect long-term growth potential. Despite the small number of individuals, the two groups showed a clear phenotypic separation under uniform rearing conditions. Moreover, transcriptomic and metabolomic analyses independently identified consistent pathway-level differences between the groups. Together, these concordant multi-omics patterns support the biological relevance of the observed molecular differences.

RNA-seq and enrichment analyses indicated that slow early-growth phenotype chickens preferentially utilized glycolysis, with oxidative metabolism being constrained. The upregulation of LDHA and PDK3 expression would limit the entry of pyruvate into the TCA cycle and promote its conversion to lactate (Gray et al., 2014; Pathria et al., 2018). Elevated expression of the gene encoding the lactate exporter SLC16A3 (MCT4) (Bonen, 2001) further reflects dependence on glycolysis under conditions of restricted oxidative capacity. In the metabolome, acetoacetate levels were nominally higher in slow-growing chickens than in fast early-growth phenotype chickens, which is consistent with the diversion of excess acetyl-CoA toward ketone processing rather than complete oxidation via the TCA cycle (Atherton et al., 2011). In addition, upregulation of GPD1 and GPD2 expression suggested activation of the glycerol-3-phosphate shuttle; however, the glycerol-3-phosphate pool was higher in fast early-growth phenotype chickens than in slow early-growth phenotype chickens. This transcript–metabolite discordance is consistent with a flux effect (greater shuttle turnover lowering the steady-state glycerol-3-phosphate pool), rather than a contradiction. Simultaneously, the reduced expression of COX subunits and DHODH in slow early-growth phenotype chickens indicates a distal electron transport bottleneck, which would increase cytosolic NADH pressure and favors lactate export despite upstream transcriptional compensation. Taken together, our data support a model in which slow early-growth phenotype chickens exhibit a “Warburg-like” glycolytic preference with compensatory activation of upstream oxidative pathways but limited electron outflow at Complex IV/COX and DHODH. Consequently, they rely more on fermentation under conditions of constrained respiratory throughput, with concomitant activation of stress-adaptive catabolic programs, such as the ubiquitin–proteasome system and autophagy. Concordantly, CARNS1 expression was higher in slow-growing chickens than in fast early-growth phenotype chickens, consistent with a compensatory increase in histidine-containing dipeptide biosynthesis to buffer lactate-associated acid load and reactive species under glycolytic, hypoxia-like conditions. As CARNS1 is ATP-dependent, this enhanced flux would incur a modest ATP cost and could slightly compete with protein synthesis; however, we did not quantify carnosine/anserine levels and therefore consider this a mechanistic hypothesis.

In contrast, fast early-growth phenotype chickens displayed a metabolic profile based on oxidative phosphorylation and lipid utilization. Elevated glycerol-3-phosphate levels and increased DHODH expression support activation of the electron transport chain via lipid-derived energy substrates. Although upregulation of genes related to lipid uptake and regulation—such as PPARG and LPL—was observed, direct evidence for enhanced fatty acid oxidation remains limited. This finding is consistent with reports that mitochondrial fatty acid oxidation capacity in broiler breast muscle is restricted compared with that in layer chickens (Hakamata et al., 2020). Moreover, the fast early-growth phenotype chickens showed significant enrichment of immune pathways, including Phagosome and Cytokine–cytokine receptor interaction, whereas Reactome analysis highlighted the “Innate immune system.” These patterns likely reflect both the sufficient energy supply to meet the metabolic demands of immune activation (Hotamisligil, 2006; Straub, 2007) and local tissue stress or remodeling during rapid hypertrophy. Reduced CARNS1 expression in fast early-growth phenotype chickens may also contribute to poor meat quality, as carnosine functions as a pH buffer and antioxidant that protects muscle integrity (Boldyrev et al., 2013).

Notably, oxidative phosphorylation and fatty acid β-oxidation have been reported to be impaired in wooden breast-affected fast-growing broilers (Wang et al., 2023). However, in our study, mitochondrial constraints in slow early-growth phenotype chickens were associated with an ETC bottleneck, indicating different mechanisms that converge on similar mitochondrial inefficiencies. Thus, both “too fast” and “too slow” growth processes may ultimately lead to suboptimal mitochondrial function, albeit through distinct metabolic mechanisms.

These distinct metabolic strategies were evident at 35 days of age and may reflect differences established during early post-hatch development. In the present study, chickens classified as fast or slow early-growth phenotypes based on body weight gain from days 1 to 5 exhibited coherent and contrasting transcriptional and metabolic profiles in skeletal muscle at 35 days of age. This temporal consistency suggests that early phenotypic divergence in growth is associated with sustained differences in muscle energy metabolism rather than transient variation. From a practical perspective, recognizing such early phenotypic differences may provide opportunities to better understand individual growth patterns and to inform future strategies for growth prediction and management in poultry production.

Conclusion

This study demonstrated that broiler chickens with different growth phenotypes employ distinct metabolic strategies in breast muscle at 35 days of age. Slow early-growth phenotype chickens showed glycolytic reliance, lactate fermentation, ketone metabolism, and enhanced protein degradation, accompanied by an electron transport chain bottleneck that limits oxidative phosphorylation. In contrast, fast early-growth phenotype chickens relied more on oxidative phosphorylation supported by lipid-derived substrates. In addition, fast early-growth phenotype chickens exhibited enrichment of immune-related pathways in skeletal muscle, likely reflecting metabolic–immune crosstalk associated with enhanced oxidative metabolism rather than classical inflammatory immune responses. Multi-omics integration confirmed coordinated changes between metabolites and gene expression, whereas SNP analysis suggested minor regulatory variation. These findings provide robust evidence that divergent growth performance is associated with consistent and measurable differences in muscle energy metabolism and related pathways. Integrating these insights may help balance the goals of productivity and meat quality in modern poultry production.

Data availability

All RNA-seq data generated in this study were deposited in the DNA Data Bank of Japan under accession number PRJDB20787.

CRediT authorship contribution statement

Shinya Ishihara: Writing – review & editing, Writing – original draft, Visualization, Validation, Supervision, Software, Resources, Project administration, Methodology, Investigation, Funding acquisition, Formal analysis, Data curation, Conceptualization. Saki Shimamoto: Writing – review & editing, Writing – original draft, Resources, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Shinobu Fujimura: Writing – review & editing, Resources, Methodology, Conceptualization. Miyu Kamimura: Writing – review & editing, Validation, Methodology. Daichi Ijiri: Writing – review & editing, Writing – original draft, Validation, Supervision, Resources, Project administration, Methodology, Conceptualization.

Disclosures

The authors declare no competing financial interests or personal relationships that could have influenced the work reported in this paper. This research was supported by a grant from the Ito Foundation, which had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript.