Histology and matched multi-omics profiling resolve developmental transitions in Wugang Copper Goose leg muscle

Abstract

Wugang Copper Goose is an indigenous meat-type breed, but the molecular events accompanying leg muscle development around hatch remain unclear. We examined breast and leg muscle histology at embryonic days 15 (E15), 19 (E19), 23 (E23), and 28 (E28) and posthatch day 1 (P1). Transcriptomic, proteomic, and metabolomic profiling of leg muscle at E15, E23, and P1 used matched samples from the same five individuals per stage. Leg muscle fibers became progressively organized before hatch and enlarged markedly at P1, and multi-omics analysis distinguished two developmental transitions. From E15 to E23, increasing cytoskeletal and calcium-signaling programs accompanied lower DNA-replication- and nucleotide-related signals, consistent with structural assembly and declining replication-related activity. From E23 to P1, fiber enlargement coincided with higher oxidative phosphorylation, citrate-cycle, peroxisome proliferator-activated receptor signaling, and fatty-acid-associated programs and lower proliferative programs. RNA and protein fold changes were moderately concordant across both transitions (Spearman ρ = 0.55 and 0.58), whereas metabolite changes were largely layer specific. Reverse-transcription quantitative PCR (RT-qPCR) supported the directional expression patterns of eight representative genes. Together, these data distinguish an early transition from replication-related activity to structural assembly and a posthatch transition dominated by oxidative and lipid-associated maturation.

Introduction

Skeletal muscle is the principal edible tissue in poultry and directly contributes to carcass yield and meat-related quality traits. Its developmental foundation is established largely before hatch, when myofibers are formed, whereas posthatch growth depends mainly on hypertrophy and maturation of existing fibers. This transition involves a shift from proliferative and differentiation programs toward sarcomere assembly, mitochondrial function, and metabolic support. Accordingly, the embryonic-to-posthatch interval can shape the structural foundation and subsequent growth potential of poultry muscle (Oviedo-Rondón et al., 2020; Halevy, 2020; Wang et al., 2022; Velleman, 2023).

Goose muscle development has a distinctive temporal context because incubation is longer than in chickens and ducks, and breast and leg muscles develop asynchronously. In Yangzhou geese, leg muscle differentiation begins around embryonic day 21 and is largely completed between embryonic days 23 and 25, whereas breast muscle differentiation continues after hatch (Weng et al., 2025). Transcriptomic studies have identified stage-dependent changes in myogenic regulation and cell-cycle activity in Dingan, Shitou, and Wuzong geese (Xu et al., 2021; Shao et al., 2025). Related studies have further described epigenetic regulation and cellular heterogeneity during goose muscle development (Zhang et al., 2023; Xu et al., 2025). These findings establish that goose muscle development is both stage- and muscle-dependent, but they do not resolve how leg-muscle structure relates to molecular remodeling from embryonic development to hatch.

Because transcript abundance does not map directly onto protein abundance or biochemical state, resolving this transition requires complementary measurements across molecular layers. RNA profiles capture regulatory responses, protein abundance reflects much of the cellular machinery, and metabolites provide a complementary view of biochemical state. Integrated studies in poultry have implicated focal adhesion, actin-cytoskeleton regulation, extracellular-matrix interactions, cell-cycle control, and energy metabolism in muscle growth (Hu and Liu, 2023; Zhao et al., 2024; Xiong et al., 2024). Many existing studies, however, have focused on a single muscle, one or two molecular layers, or isolated developmental comparisons. It therefore remains unclear which changes are concordant across RNA, protein, and metabolite layers, which are layer-specific, and how these patterns correspond to muscle structure.

Wugang Copper Goose is an indigenous meat-type breed for which leg muscle development around hatch has not been characterized across these biological levels. We therefore examined breast and leg muscle histology at embryonic days 15 (E15), 19 (E19), 23 (E23), and 28 (E28) and posthatch day 1 (P1). Leg muscle collected at E15, E23, and P1 underwent matched transcriptomic, proteomic, and metabolomic profiling. The five histological stages span early embryogenesis, the period around reported leg-muscle differentiation, late incubation, and the immediate posthatch state. The molecular stages bracket an early embryonic state, E23 within the reported differentiation interval, and P1 immediately after hatch. Leg muscle was selected for molecular profiling because it develops earlier than breast muscle over this interval, with breast muscle retained as a histological comparator. We asked how muscle structure and molecular programs changed across two successive intervals and whether the three omics layers provided concordant or complementary evidence. Using all three assays on samples from the same five individuals per stage enabled direct cross-omics comparison. Selected transcript profiles were further examined by reverse-transcription quantitative PCR (RT-qPCR).

Materials and methods

Ethics statement

All experimental procedures were approved by the Laboratory Animal Ethics Review Committee of Hunan Institute of Animal Husbandry and Veterinary Science (Hunan Provincial Livestock Breeding Station) (Approval No. HNSYDW20240008). All procedures were conducted in accordance with the approved guidelines for animal welfare.

Animals, incubation conditions, and sample collection

A total of 200 fertilized eggs of Wugang Copper Goose were obtained from a commercial breeder farm in Wugang City, Hunan Province, China. All eggs were collected from the same batch and were laid by breeder geese aged 300 to 400 d under standardized feeding and management conditions. Before incubation, eggs were disinfected by formaldehyde fumigation and incubated in a fully automatic tunnel-type incubator (GD-1056; Dezhou Youqin Mechanical Equipment Co., Ltd., Dezhou, China). Incubation temperature was set at 37.8°C from embryonic day 1 (E1) to E14, decreased to 37.5°C from E15 to E27, and further reduced to 37.0°C from E28 until hatching. Relative humidity was held at 60–65% from E1 to E27, then raised to 70–75% following transfer to hatching baskets on E28. Eggs were turned 90° every 2 hours throughout incubation until E28, and twice-daily cooling via spraying with lukewarm water (≈35°C) was implemented from E15 to E28.

At each designated sampling time point, embryos and P1 goslings were randomly selected and humanely euthanized by cervical dislocation. Because sex could not be reliably determined based on external morphology during embryonic development, sex was not included as a factor in this study. Breast muscle and leg muscle samples were collected at E15, E19, E23, E28, and P1 for histological analysis. Leg muscle samples from E15, E23, and P1 were used for transcriptomic, proteomic, and metabolomic profiling. For histology, 3 independent embryos or goslings were used at each stage. For each omics analysis, 5 independent biological replicates were collected at each of E15, E23, and P1, with each replicate derived from a single embryo or gosling.

For histological analysis, the pectoralis major muscle and gastrocnemius muscle were dissected and fixed in 4% paraformaldehyde for paraffin embedding. For multi-omics analysis, leg muscle samples were rapidly dissected, rinsed with sterile normal saline, snap-frozen in liquid nitrogen, and stored at −80°C until RNA, protein, and metabolite extraction.

Histological analyses

Paraffin-embedded breast and leg muscle samples were sectioned at 4 to 5 μm and stained with hematoxylin and eosin using a standard protocol. Histological images were captured under a light microscope (Nikon Eclipse E100, Nikon, Tokyo, Japan) equipped with an imaging system (Nikon DS-U3, Nikon, Tokyo, Japan). Representative sections were used to evaluate developmental changes in muscle fiber organization and bundle formation across E15, E19, E23, E28, and P1. For quantitative morphometric analysis, three non-overlapping fields with intact tissue morphology and no obvious tissue damage were randomly selected from each sample, and all clearly identifiable muscle fiber profiles within the selected fields were evaluated per sample. Muscle fiber diameter and cross-sectional area were measured using ImageJ software. Damaged, incomplete, or indistinct fibers were excluded from the analysis. For each developmental stage and muscle type, three biological replicates were analyzed, and the mean value of each biological replicate was used for statistical analysis.

RNA sequencing (RNA-seq) and bioinformatic analysis

Total RNA was extracted from leg muscle samples collected at E15, E23, and P1. RNA quality was assessed before library construction. Poly(A)+ RNA was enriched, fragmented, reverse-transcribed, and used for cDNA library preparation. Qualified libraries were sequenced on an Illumina NovaSeq 6000 platform to generate 150-bp paired-end reads.

Raw reads were filtered using fastp (v0.23.2), aligned to the Anser cygnoides reference genome (NCBI GCF_002166845.1_GooseV1.0) using HISAT2 (v2.2.1), and summarized at gene level using featureCounts (v2.0.3). Genes with at least 10 raw counts in at least 3 of the 15 samples were retained. Analyses were conducted in DESeq2 (v1.52.0) using a design containing developmental stage. A likelihood-ratio test comparing the full model, ∼ stage, with the reduced model, ∼ 1, tested the global stage effect. E23-versus-E15 and P1-versus-E23 effects were estimated by Wald contrasts. Pairwise differential expression required a Benjamini-Hochberg false discovery rate (BH FDR) below 0.05 and an absolute log₂ fold change of at least 1. DESeq2-normalized counts were used to display candidate abundance, and variance-stabilized values were used for sample correlations, principal component analysis (PCA), and three-stage pattern visualization.

Data-independent acquisition (DIA) proteomic analysis

Proteins were extracted from frozen leg muscle samples collected at E15, E23, and P1. Protein concentration was determined using the Bradford method, and protein quality was evaluated by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE). After tryptic digestion, peptides were desalted, lyophilized, and analyzed using a Vanquish Neo ultra-high-performance liquid chromatography (UHPLC) system coupled to an Orbitrap Astral-series mass spectrometer in DIA mode.

Raw DIA data were searched against the Anser cygnoides protein database (UniProt UP000694521) using DIA-NN (v1.8.1). Carbamidomethylation of cysteine was fixed, N-terminal methionine excision was variable, and up to 2 missed cleavages were allowed. Peptides with Lib.Q.Value ≤ 0.01 and proteins with Lib.PG.Q.Value ≤ 0.01 were retained. Protein-matrix zeros were treated as unquantified rather than as biological zero values. For inference, proteins quantified in at least 3 samples at every stage were retained and their nonzero abundances were log₂-transformed. Stage effects were analyzed with limma (v3.68.5) using robust empirical-Bayes moderation with an intensity trend. A moderated F test assessed the global stage effect, and moderated contrasts assessed E23 versus E15 and P1 versus E23. Pairwise differential abundance required BH FDR below 0.05 and an absolute log₂ fold change of at least 1. Missing values remained missing during inference. For quality assessment only, proteins quantified in at least 12 of 15 samples were log₂-transformed and missing values were replaced by the within-protein median; sample correlations used all retained proteins and PCA used the 2,000 most variable retained proteins after protein-wise centering.

Untargeted metabolomic analysis

Untargeted metabolomic profiling was performed using leg muscle samples collected at E15, E23, and P1. Metabolites were extracted from frozen tissue powder using methanol-based extraction. Quality control samples were prepared by pooling equal volumes from all biological samples, and blank samples were included to remove background ions. Liquid chromatography–tandem mass spectrometry (LC-MS/MS) analysis was performed on a Vanquish UHPLC system coupled to a high-resolution Orbitrap mass spectrometer in both positive and negative electrospray ionization modes.

Raw mass spectrometry (MS) files were converted to mzXML format using ProteoWizard and processed using XCMS (v3.20.0) for peak detection, retention-time correction, alignment, and quantification. Metabolites were annotated using the NovoMetDB spectral database and cross-referenced to the Kyoto Encyclopedia of Genes and Genomes (KEGG), Human Metabolome Database (HMDB), and LIPID MAPS. Processed positive- and negative-ion intensity matrices were log₂-transformed and analyzed separately with limma using robust empirical-Bayes moderation with an intensity trend. A moderated F test assessed the global stage effect, and moderated contrasts assessed E23 versus E15 and P1 versus E23. Ion features were assigned to a common chemical identity using the InChIKey when available, followed by the KEGG compound identifier and standardized compound name. P values for ion features assigned to the same identity were combined by the Simes method, and BH correction was applied at the chemical-identity level. The ion feature with the smallest P value supplied the representative abundance profile and fold change. Pairwise differential abundance required chemical-level BH FDR below 0.05 and an absolute log₂ fold change of at least log₂(1.5). For quality assessment, both ion modes were used for sample correlations; each feature was standardized before the two modes were combined for PCA.

Integrated multi-omics analysis

Features with a global stage FDR below 0.05 were classified from their E15, E23, and P1 means as monotonically increasing, monotonically decreasing, E23-peaking, or E23-valley. KEGG over-representation analysis used the measured, KEGG-mapped features in each omics layer as the background. Global enrichments were tested within each three-stage pattern, and pairwise enrichments were tested separately for features higher or lower at the later stage. Hypergeometric P values were BH-adjusted within each omics, pattern, or contrast-direction family. Pathways required at least 3 mapped background features and at least 3 selected overlapping features and contained no more than 500 background features. RNA and protein effects were matched by gene symbol. For duplicated symbols, the transcript with the highest mean normalized count and the protein with the highest mean log₂ abundance were retained. Spearman correlations were calculated between RNA and protein log₂ fold changes. Matched features were classified as significant and directionally concordant in both layers, significant and discordant in both, significant in only one layer, or nonsignificant in both using the layer-specific BH FDR and effect-size criteria. Exact intersections of KEGG pathways enriched at BH FDR below 0.05 were summarized with UpSet plots. An equal-block-weighted consensus PCA was constructed from the 15 matched samples. Each omics block contributed its 500 most variable features after variance stabilization for RNA, log₂ transformation and within-protein median imputation for proteins, and log₂ transformation for metabolites. Features were standardized within each block, and each block was divided by the square root of its feature number before concatenation. Pairwise pathway-module associations were evaluated for carbon metabolism, the citrate cycle, biosynthesis of amino acids, arginine and proline metabolism, glycerophospholipid metabolism, and nucleotide metabolism. Each pathway contained at least 3 mapped features in every omics layer. Module scores were the mean standardized abundance of all mapped features; metabolite ion features assigned to the same chemical identity were averaged first. Scores were centered within developmental stage before RNA-protein, RNA-metabolite, and protein-metabolite Spearman correlations were calculated across the matched samples. BH correction was applied across the 18 displayed correlations.

RT-qPCR validation

RT-qPCR was used to validate the expression patterns of 8 representative genes: PAX7, MYOG, NEB, CKMT2, GATM, ACSL1, GPAM, and FABP3. Total RNA was extracted from leg muscle samples collected at E15, E23, and P1, with 3 independent biological replicates per stage. First-strand cDNA was synthesized using the HiScript RT SuperMix for qPCR kit (Vazyme, China). RT-qPCR was performed using AceQ qPCR SYBR Green Master Mix (Vazyme, China) in a 20 μL reaction system. TBP and YWHAZ were used as dual reference genes. Primer sequences are listed in Table 1. Relative gene expression was calculated using the 2^−ΔΔCt method.

Table 1.

Primers used for gene expression analysis.

Gene	Primer Sequence (5′–3′)	Annealing Temperature (°C)	Accession No.
TBP	Forward: TGGGCAGCTGTGATGTGAAA	60	XM_066994068.1
	Reverse: GCTCGAACTTTAGCGCCAGT
YWHAZ	Forward: GTTCCCTTGCAAAAACGGCTT	60	XM_066990626.1
	Reverse: AGACGGAAGTTGGAAGGCTG
PAX7	Forward: CCGTGCCCTCAGGTTTAGTG	60	XM_066982025.1
	Reverse: GTTACCTTTGTCGCCCAGGA
MYOG	Forward: CCAACCCTGCGGATCACCT	60	XM_048070946.2
	Reverse: TCAGTTTTGAACCCGCTCCT
NEB	Forward: AAAGCCACAGGAGTGAGTGG	60	XM_066999882.1
	Reverse: CACCTCATGTGGTTCACCCT
CKMT2	Forward: GCATGGTGGCAGGTGATGAA	60	XM_013191808.3
	Reverse: CTGACCATGGGTGATCTTGGA
GATM	Forward: GCGTGTCCCGAGCAGTAAA	60	XM_067003987.1
	Reverse: GTAAAGACTCTTCCAAGCCGC
ACSL1	Forward: TCCGCTTAAATAGGGCGGC	60	XM_048078356.2
	Reverse: GTCCCAGAGCTGGAATCTCCC
GPAM	Forward: CCAAGACTGCCAAAGCCAAG	60	XM_067000606.1
	Reverse: CTGAAGATCAGAAGCTGGCCT
FABP3	Forward: GAGGAGTTCGACGAGACCAC	60	XM_066982390.1
	Reverse: GCCCATGGTGAGAGTCAGAAT

Statistical analysis

Analyses were performed in R (v4.6.1). The individual goose was the independent experimental unit. Histological measurements were averaged across fields within each individual, and developmental stages were compared separately within each muscle using one-way analysis of variance (ANOVA) followed by Tukey's honestly significant difference test. RT-qPCR stage effects were tested by one-way ANOVA on ΔC_t values, with BH adjustment across the 8 genes; Tukey-adjusted comparisons were used within each gene. Comparisons of RNA-seq and RT-qPCR stage effects were summarized descriptively using Spearman correlations. Transcriptomic, proteomic, metabolomic, and pathway multiplicity adjustments followed the analysis families specified in the respective sections above. Quantitative plots show individual biological samples together with the mean and the variability measure stated in each figure legend.

Results

Leg muscle showed an apparent structural transition before hatch

At E15, breast and leg muscles contained loosely arranged fibers separated by relatively large intercellular spaces, and muscle-bundle boundaries were poorly defined (Fig. 1A). Fiber organization increased from E19 to E23, with more distinct bundle formation in leg than in breast muscle. By E28 and P1, leg muscle contained densely packed fibers and clearer bundle architecture, whereas breast muscle remained less compact.

Fig. 1.

Histological development and quantitative morphology of breast and leg muscles in Wugang Copper Goose from E15 to P1.

(A) Representative hematoxylin-eosin-stained sections of breast muscle and leg muscle collected at E15, E19, E23, E28, and P1. The upper row shows breast muscle, and the lower row shows leg muscle. Scale bars = 50 μm. (B) Quantitative analysis of muscle fiber diameter and muscle fiber cross-sectional area. Data are presented as means ± standard error of the mean (SEM). Different lowercase letters indicate significant differences among developmental stages within the same muscle type (P < 0.05).

Morphometric analysis showed no difference in leg-muscle fiber diameter between E23 and E28 (13.51 ± 0.13 versus 13.72 ± 0.38 μm; P = 0.910), but diameter was greater at P1 (17.36 ± 0.45 μm) than at either stage (both P < 0.001; Fig. 1B). Cross-sectional area followed the same pattern, with similar values at E23 and E28 (66.88 ± 2.26 versus 73.40 ± 4.42 μm²; P = 0.659) and a higher value at P1 (127.74 ± 7.33 μm²; both P < 0.001). Quantifiable breast-muscle fiber profiles were obtained only at P1, when mean diameter and cross-sectional area were 10.44 ± 0.64 μm and 42.08 ± 5.60 μm², respectively.

Matched multi-omics profiles distinguished the three developmental stages

Transcriptomic, proteomic, and metabolomic profiling was completed for all 15 matched leg-muscle samples. Within-stage sample correlations ranged from 0.88 to 0.98 for RNA, 0.91 to 0.98 for proteins, and 0.68 to 0.97 for metabolites (Fig. 2A to C). The first two principal components explained 82.9%, 77.3%, and 60.5% of the variance in the RNA, protein, and metabolite datasets, respectively. The three stages separated in each omics layer, although E23 metabolite profiles were more dispersed than the E15 and P1 profiles (Fig. 2D to F).

Fig. 2.

Multi-omics data quality and developmental-stage separation.

(A to C) Pearson correlation matrices for the RNA, protein, and metabolite samples, respectively. RNA correlations were calculated from the 2,000 most variable variance-stabilized genes. Protein correlations used proteins quantified in at least 12 of 15 samples after log₂ transformation and within-protein median imputation of unquantified values. Metabolite correlations used log₂-transformed positive- and negative-ion intensities. (D to F) PCA of the corresponding RNA, protein, and metabolite datasets. RNA PCA used the 2,000 most variable variance-stabilized genes. Protein PCA used the 2,000 most variable retained proteins after protein-wise centering. Metabolite PCA used standardized features from the combined positive- and negative-ion datasets. Each point represents one individual goose, with 5 biological replicates per stage. Labels identify all samples.

Global stage analysis resolved monotonic and E23-centered molecular patterns

Global stage tests identified 11,357 transcripts, 4,443 proteins, and 1,742 metabolite identities at FDR below 0.05 (Fig. 3). Of these, 11,355 transcripts, all 4,443 proteins, and all 1,742 metabolite identities were assigned to one of four patterns based on their stage means. Two significant transcripts had identical E23 and P1 stage means and were therefore not assigned to one of the four directional patterns. Monotonic increases or decreases accounted for 70.4% of the classified transcripts, 83.9% of the proteins, and 69.0% of the metabolite identities. The remaining features showed an E23 peak or valley.

Fig. 3.

Global three-stage analysis resolves monotonic and E23-centered molecular patterns.

(A to C) Standardized stage means for transcripts, proteins, and metabolite identities with a significant global developmental-stage effect. Features were classified as monotonically increasing, E23-peaking, E23-valley, or monotonically decreasing according to their E15, E23, and P1 means. Thick lines and ribbons show the median and interquartile range. For visual clarity, 60 features per pattern are shown as faint background lines; all classified features were included in the pattern counts and enrichment analyses. (D) Distribution of the four patterns among globally significant features. Numbers indicate feature counts. (E) Representative pattern-specific KEGG enrichment. Position shows −log₁₀ of the BH FDR, point size shows the number of overlapping features, and color indicates the three-stage pattern. Up to 2 pathways per omics-pattern combination with BH FDR below 0.10 are displayed; the dotted line denotes BH FDR = 0.05. RNA stage effects were tested with a DESeq2 likelihood-ratio test comparing ∼ stage with ∼ 1. Protein stage effects were tested with a limma moderated F test using log₂ abundance values and requiring at least 3 quantified samples per stage. Metabolite-ion effects were tested with limma, combined across matching chemical identities using Simes P values, and corrected at the chemical-identity level. BH correction was applied within each declared analysis family.

Monotonically increasing transcripts were enriched for oxidative phosphorylation and branched-chain amino-acid degradation, whereas increasing proteins were enriched for oxidative phosphorylation and cytoskeleton-related functions (Fig. 3E). Monotonically decreasing transcripts were enriched for cell-cycle and DNA-replication pathways, and decreasing proteins were enriched for spliceosomal and nucleocytoplasmic-transport processes. E23-peaking transcripts were enriched for ribosomal and extracellular matrix (ECM) pathways, whereas E23-peaking proteins were enriched for ECM and muscle-cytoskeleton pathways. E23-valley transcripts included autophagy and mechanistic target of rapamycin (mTOR)-related pathways, whereas E23-valley proteins included oxidative-phosphorylation and peroxisome proliferator-activated receptor (PPAR) signaling pathways.

The E15-to-E23 transition combined structural assembly with declining replication-related programs

The E23-versus-E15 contrast identified 3,262 differentially expressed genes, including 2,028 with higher and 1,234 with lower expression at E23 (Fig. 4A). The same contrast identified 887 differentially abundant proteins, of which 378 were higher and 509 lower at E23, and 608 differentially abundant metabolite identities, of which 322 were higher and 286 lower at E23.

Fig. 4.

The E15-to-E23 and E23-to-P1 transitions engage distinct structural and metabolic programs.

(A) Numbers of differential RNA, protein, and metabolite features in the E23-versus-E15 and P1-versus-E23 contrasts. Orange and blue indicate higher and lower abundance at the later stage, respectively. RNA and protein features required a BH FDR below 0.05 and an absolute log₂ fold change of at least 1. Metabolite identities required chemical-level BH FDR below 0.05 and an absolute log₂ fold change of at least log₂(1.5). (B and C) Direction-specific KEGG enrichment for selected muscle-development and metabolic processes during E15-to-E23 and E23-to-P1. Filled circles indicate BH FDR below 0.05; open circles indicate that the pathway did not meet this criterion in the corresponding omics layer. Fill color indicates the direction of the contributing differential features. Point size represents −log₁₀(FDR). The reference circles at 2, 5, 10, and ≥20 correspond to FDR values of 0.01, 10⁻⁵, 10⁻¹⁰, and ≤10⁻²⁰, respectively; values above 20 were capped for display.

Cytoskeleton-in-muscle-cell and calcium-signaling pathways were enriched among RNA and protein features that increased at E23 (Fig. 4B). ECM interaction and oxidative phosphorylation were additionally enriched among increasing transcripts. DNA-replication and spliceosomal pathways were enriched among proteins that decreased at E23. Nucleotide metabolism and alanine, aspartate, and glutamate metabolism were enriched among metabolites that decreased at E23, whereas the latter pathway was enriched among transcripts that increased. The early transition therefore contained both cross-layer agreement in structural programs and layer-specific changes in intermediary metabolism.

The E23-to-P1 transition emphasized energy- and lipid-associated maturation

The P1-versus-E23 contrast identified 3,503 differentially expressed genes, including 1,562 with higher and 1,941 with lower expression at P1 (Fig. 4A). It also identified 1,654 differentially abundant proteins, including 273 higher and 1,381 lower at P1, and 1,099 differentially abundant metabolite identities, including 531 higher and 568 lower at P1.

Oxidative phosphorylation, the citrate cycle, PPAR signaling, fatty-acid metabolism, and pyruvate metabolism were enriched among RNA and protein features that increased at P1 (Fig. 4C). Cell-cycle and DNA-replication pathways were enriched among RNA and proteins that decreased at P1. Spliceosomal proteins and ECM-related RNA and proteins also decreased, whereas nucleotide metabolism was enriched among lower metabolite identities. These changes distinguished the late transition from the earlier interval and coincided with the increase in leg-muscle fiber size observed at P1.

RNA and protein effects were moderately concordant across both transitions

RNA and protein fold changes were compared for 2,107 matched gene symbols. The effects were positively correlated during both E15-to-E23 (Spearman ρ = 0.55, P = 3.67 × 10⁻¹⁷⁰) and E23-to-P1 (ρ = 0.58, P = 2.42 × 10⁻¹⁹²; Fig. 5A and B). During E15-to-E23, 115 matched features met the significance and effect-size criteria in both layers with the same direction, whereas 9 met the criteria in opposite directions; a further 372 were significant in only one layer. During E23-to-P1, 133 were concordant and significant in both layers, 8 were discordant and significant in both, and 665 were significant in only one layer.

Fig. 5.

Cross-omics integration identifies RNA-protein concordance and pathway-level complementarity.

(A and B) RNA and protein log₂ fold changes for 2,107 matched gene symbols during E15-to-E23 and E23-to-P1. For duplicated symbols, the transcript with the highest mean normalized count and the protein with the highest mean log₂ abundance were retained. Colors distinguish significant concordant effects in both layers, significant discordant effects in both layers, significance in only one layer, and absence of significance in both layers. Significance required a BH FDR below 0.05 and an absolute log₂ fold change of at least 1. Vertical and horizontal dotted lines mark absolute log₂ fold change = 1, and the dashed diagonal is the identity line. The ACSL1 transcript–ACSL1 protein pair, shown in purple, represents a concordant significant RNA–protein change. The GPAM transcript–GPAM protein pair, shown in ochre, met the transcript but not the protein significance criterion in both transitions. These were the only genes from the 8-gene RT-qPCR panel with unambiguous one-to-one RNA–protein matches in both transitions. Spearman correlations use all 2,107 matched symbols. (C and D) Standard UpSet plots show exact intersections of KEGG pathways enriched at BH FDR below 0.05. All 7 possible intersections are shown, including zero-count intersections. Enrichment direction was ignored because direction-specific results are shown in Fig. 4. (E) Equal-block-weighted consensus PCA of the 15 matched samples. Each omics block contributed its 500 most variable features and was weighted by the inverse square root of its feature number before concatenation. Labels identify all samples; ellipses show descriptive 68% data regions. (F) Pairwise Spearman correlations among RNA, protein, and metabolite KEGG pathway-module scores across the same 15 samples. Modules contained at least 3 mapped features in every omics layer and were calculated from all mapped features rather than differential features alone. Scores were centered within developmental stage before correlation. Asterisks denote BH FDR below 0.05 across the 18 displayed correlations. In panels C and D, R, P, and M denote RNA, protein, and metabolite, respectively.

Within the 8-gene RT-qPCR panel, ACSL1 transcript and ACSL1 protein, and GPAM transcript and GPAM protein, had unambiguous one-to-one matches in both transitions. ACSL1 transcript and ACSL1 protein increased significantly during both intervals. GPAM transcript increased significantly, whereas GPAM protein did not meet the differential-abundance criterion. These pairs therefore represented distinct forms of cross-layer agreement within the selected candidate panel.

Pathway integration revealed pairwise overlap and substantial omics specificity

During E15-to-E23, 5 enriched KEGG pathways were shared by RNA and protein, 1 was shared by RNA and metabolites, and none was shared by proteins and metabolites or by all 3 layers (Fig. 5C). Twenty-six pathways were unique to RNA, 6 to protein, and 3 to metabolites. During E23-to-P1, 20 pathways were shared by RNA and protein, whereas no enriched pathway was shared by either metabolite-containing pair or by all 3 layers (Fig. 5D). Nineteen pathways were unique to RNA, 7 to protein, and 3 to metabolites.

The equal-block-weighted consensus PCA separated E15, E23, and P1 samples along axes explaining 63.3% and 10.4% of the integrated variance (Fig. 5E). After pathway-module scores were centered within stage, the RNA and protein scores for arginine and proline metabolism remained positively correlated across the matched individuals (ρ = 0.76, FDR = 0.016; Fig. 5F). None of the displayed RNA-metabolite or protein-metabolite module correlations met FDR below 0.05.

Stage-associated transcript, protein, and histological profiles supported a working model of leg muscle development

RNA-seq profiles showed increasing abundance of NEB, CKMT2, GATM, ACSL1, GPAM, and FABP3 relative to E15, an E23-centered MYOG profile, and decreasing PAX7 abundance (Fig. 6A). RT-qPCR showed the same direction as RNA-seq for all 8 genes in both the E23-versus-E15 and P1-versus-E15 comparisons. Each RT-qPCR assay showed a global stage effect after BH correction across the 8 genes, with all FDR values at or below 2.01 × 10⁻⁷. Across the selected genes, the descriptive Spearman correlations between RNA-seq and RT-qPCR stage effects were 0.476 for E23 versus E15 and 0.619 for P1 versus E15. The corresponding non-centered normalized RNA abundances are shown in Fig. S1.

Fig. 6.

RNA-seq and RT-qPCR profiles of selected transcripts with corresponding protein abundance.

(A) Developmental profiles of PAX7, MYOG, NEB, CKMT2, GATM, ACSL1, GPAM, and FABP3 measured by RNA-seq and RT-qPCR. RNA-seq values are log2(DESeq2-normalized count + 1) centered on the mean E15 value for each gene. RT-qPCR values are log2 relative expression values calculated independently from ΔΔCt and centered on E15 by definition. RNA-seq included 5 individual geese per stage. Non-centered normalized RNA abundances are provided in Fig. S1. RT-qPCR included 3 biological samples per stage, with 3 technical wells averaged within each sample. ΔCt was calculated as target-gene mean Ct minus the arithmetic mean Ct of TBP and YWHAZ, and the mean E15 ΔCt was used as the calibrator. Small symbols represent individual samples; large symbols and error bars show mean ± standard deviation (SD). Strip labels report the FDR from the global three-stage DESeq2 test for RNA-seq and from one-way ANOVA on ΔCt values for RT-qPCR, with BH correction across the 8 genes. The methods were analyzed separately. (B) Log2 normalized abundance of 6 representative proteins identified by the UniProt accessions in the strip labels. Five biological samples were analyzed per stage when quantified. Protein-matrix zeros were treated as unquantified values. PAX7 was not quantified in any P1 sample and therefore did not meet the requirement of at least 3 quantified samples per stage for the global protein test. In the strip labels, q denotes the corresponding BH FDR.

At the protein level, individual-sample data showed increasing abundance of ACSL1 and ACACB and decreasing abundance of MCM2, FN1, and FASN across development (Fig. 6B). PAX7 protein was quantified in all E15 and E23 samples but not in any P1 sample. Because it did not meet the requirement for quantification in at least 3 samples at every stage, PAX7 was not included in the global protein test. These profiles show the observed proteomic abundance of the selected candidates across the 3 stages.

Together with the histological findings, these transcript and protein profiles informed a working model that summarizes the developmental stages without assigning direct regulatory relationships among the measured features (Fig. 7). E15 was characterized by immature fiber organization, cell-cycle and DNA-replication programs, relatively high PAX7 transcript and MCM2 protein abundance, and nucleotide-related metabolite signals. E23 combined improved fiber organization with cytoskeletal, ECM, and calcium-signaling programs, an E23-centered MYOG transcript profile, and stage-specific structural and metabolic features. P1 combined compact, enlarged fibers with oxidative-phosphorylation, citrate-cycle, PPAR-signaling, and fatty-acid-related programs; higher NEB, CKMT2, ACSL1, and FABP3 transcript abundance; and higher ACSL1 and ACACB protein abundance.

Fig. 7.

Evidence-supported working model of Wugang Copper Goose leg muscle development.

The 3 columns summarize histology, biological programs, and representative omics evidence at E15, E23, and P1. Histological fields from Fig. 1 were cropped and resized; scale bars were retained. Solid boxes show observed stage-associated findings from Fig. 1, Fig. 3 to 6. Dashed arrows summarize structural organization with lower replication-related programs (E15 to E23) and energy- and lipid-related maturation with lower proliferative programs (E23 to P1). Candidates and pathways are representative. AA, amino acid; Metab., metabolite; nuc., nucleotide; TCA, tricarboxylic acid.

Discussion

Histological remodeling defines two developmental transitions

Serial histology and matched multi-omics profiling revealed two successive transitions in Wugang Copper Goose leg muscle around hatch. By E23, leg muscle had acquired clearer bundle organization before comparable organization appeared in breast muscle. This timing agrees with the earlier development reported for goose leg muscle (Weng et al., 2025). Fiber diameter and cross-sectional area then increased markedly at P1, indicating that late embryogenesis established tissue organization before pronounced posthatch enlargement. This sequence is consistent with the broader avian pattern in which most myofibers form during embryogenesis and later growth depends mainly on hypertrophy (Halevy, 2020; Velleman, 2023). The aligned molecular changes further separated an early program of structural assembly from a later program of oxidative and lipid-associated maturation. E23 therefore emerges as a biologically informative transition point rather than a simple intermediate stage.

Structural assembly precedes posthatch metabolic maturation

The E15-to-E23 transition combined declining progenitor and replication-related signals with increasing cytoskeletal and calcium-signaling programs. Relatively high PAX7 transcript and MCM2 protein abundance at E15 were consistent with an earlier myogenic and proliferative state (Seale et al., 2000; Buckingham and Rigby, 2014; Simon and Schwacha, 2014). At E23, RNA and protein enrichment for cytoskeleton-in-muscle-cell and calcium-signaling pathways supported advancing contractile organization and excitation-related maturation. Higher NEB transcript abundance was compatible with sarcomeric assembly, whereas CKMT2 and GATM profiles suggested increasing creatine-linked energy support (Ottenheijm et al., 2012; Bonilla et al., 2021; Rizo-Roca et al., 2024). Decreasing DNA-replication and nucleotide-related signals further indicated reduced proliferative activity. Together, the histological and molecular evidence places structural assembly at the center of the first developmental transition.

From E23 to P1, fiber enlargement coincided with stronger oxidative-phosphorylation, citrate-cycle, PPAR-signaling, and fatty-acid-related programs. These changes fit the energetic, substrate-handling, and membrane-remodeling demands of enlarging posthatch fibers. Reorganization of the mitochondrial reticulum during postnatal muscle development provides a physiological context for the oxidative program observed here (Kim et al., 2019). ACSL1 transcript abundance and ACSL1 protein abundance increased across both transitions, while ACACB protein was highest at P1. Higher FABP3 and GPAM transcripts further linked fatty-acid transport and activation with glycerolipid-related processes (Storch and Thumser, 2010; Gimeno and Cao, 2008). Because membrane lipid composition influences mitochondrial and cellular function in skeletal muscle (Hishikawa et al., 2014; Heden et al., 2016; Ferrara et al., 2021), the lipid-associated program may support the functional maturation of expanding fibers.

Matched multi-omics integration distinguishes convergence from layer-specific responses

The metabolome complemented this two-step model without simply reproducing transcript or protein patterns. Metabolite profiles separated the three stages, although E23 samples were more dispersed than the corresponding RNA and protein profiles. Early reductions in nucleotide-related and alanine, aspartate, and glutamate metabolism were consistent with declining replication-associated activity. Nucleotide-related signals also decreased during the later interval. No enriched pathway was shared by all three molecular layers, and no displayed metabolite-containing module correlation remained significant after stage centering and multiple-testing correction. This pattern suggests that metabolite pools capture biochemical states shaped by substrate supply, enzyme activity, transport, compartmentation, and utilization.

RNA-protein comparisons showed the clearest cross-layer agreement. Fold changes were moderately correlated in both developmental intervals, and concordant significant features greatly outnumbered discordant features. RNA-protein pathway overlap was also greater from E23 to P1 than during the earlier transition. However, many features and pathways remained specific to one layer, indicating that transcription and protein abundance captured related but nonidentical aspects of remodeling. The positive stage-centered association for arginine and proline metabolism further identified coordinated RNA-protein variation among individual geese. Because all three omics assays were performed on the same five individuals per stage, the analysis could distinguish cross-layer convergence from layer-specific responses without the added variation of unmatched sampling.

Candidate profiles support a stage-resolved developmental framework

RT-qPCR reproduced the direction of the RNA-seq changes for all eight selected transcripts in both comparisons with E15. These genes represented progenitor status, myogenic differentiation, sarcomeric structure, creatine metabolism, and lipid handling. The protein profiles provided complementary support for selected components: ACSL1 and ACACB increased, whereas MCM2, FN1, and FASN decreased across development. PAX7 protein was quantified at E15 and E23 but not at P1, consistent with its declining transcript profile. Together, these candidate profiles connect the broader omics programs to molecular features of the transition from early myogenesis to structural and metabolic maturation.

Biological significance and future directions

By aligning serial histology with multi-omics measurements from matched individuals, this study provides a stage-resolved framework for goose leg muscle development around hatch. The framework identifies E23 as a useful window for examining the shift from tissue organization to posthatch hypertrophy and metabolic maturation. It also separates observed RNA-protein convergence from metabolite-layer specificity, reducing reliance on pathways shared across all omics layers. These findings can guide developmental-stage selection and candidate prioritization in future studies of avian muscle growth. Targeted protein validation and flux-based measurements would help determine which abundance changes have direct regulatory or metabolic roles.

Conclusions

Histology and matched multi-omics profiling distinguished two major intervals of Wugang Copper Goose leg muscle remodeling. The E15-to-E23 interval combined increasing structural and calcium-signaling programs with declining replication- and nucleotide-related signals. The E23-to-P1 interval combined greater fiber size with higher oxidative, citrate-cycle, PPAR-signaling, and fatty-acid-related programs and lower proliferative activity. RNA and protein changes were moderately concordant, whereas metabolite associations were predominantly layer-specific. RT-qPCR supported the developmental profiles of 8 selected transcripts. Together, these data provide an evidence-based framework for stage-associated structural and metabolic maturation of goose leg muscle around hatch.

Author contributions

Xu He conceived and designed the study, developed the methodology, conducted the investigation, curated and analyzed the data, prepared the figures, and drafted the manuscript. Weican Wan conducted the investigation, collected samples, and performed validation. Chuze Gao conducted the investigation, provided experimental support, and curated the data. Xianxin Zhang provided resources and research support. Bocheng Liu provided experimental support. Yu Tan managed the animals, collected samples, and provided resources. Guitao Jiang conceived the study, supervised the work, administered the project, acquired funding, and reviewed and edited the manuscript. All authors critically reviewed and approved the final version of the manuscript and agree to be accountable for all aspects of the work.

Disclosures

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

Data availability

The RNA-seq data are available in the Gene Expression Omnibus under accession GSE335189. The transcriptomic and proteomic datasets are also deposited in the China National Center for Bioinformation (CNCB) BioProject database under accession PRJCA066375.