New insights into the transcriptomic profile during the late stages of chicken embryonic development

Abstract

While comprehensive transcriptomic characterization of muscle development during the hatching stage remains limited, we conducted an integrated analysis of coding and non-coding RNA profiles in the pectoral muscles of Arbor Acres (AA) broilers and TaoYuan (TY) chickens at embryonic days 17 (E17), 19 (E19), and 21 (E21). Our findings revealed notable phenotypic differences: AA broilers exhibited greater embryo weight and larger muscle fiber cross-sectional areas compared to TY chickens. Across three developmental stages, we identified 4,577 differentially expressed genes (DEGs), 143 differentially expressed microRNAs (DEMs), 90 differentially expressed circRNAs (DECs), and 3,159 differentially expressed lncRNAs (DELs). By integrating weighted gene co-expression network analysis (WGCNA) analysis with differential expression profiling, we prioritized five coding genes (FOSL2, PDE4B, TRIB1, THBS1, and FBXO32) as key regulators of muscle development. Pathway enrichment analysis further revealed significant activation of the glycolysis/gluconeogenesis pathway in TY chickens, likely supporting the energy demands of shell pipping. To elucidate transcriptional regulatory mechanisms, we constructed competing endogenous RNA (ceRNA) networks based on miRanda predictions. This analysis identified four microRNAs (gga-miR-206, gga-miR-383-3p, gga-miR-449a, and gga-miR-449c-5p) that may modulate developmental genes through lncRNA and circRNA mediated sponge interactions. These ceRNA networks provide novel insights and a valuable framework for investigating the molecular regulation of embryonic muscle development in AA broilers and TY chickens.

Introduction

Embryonic development represents a critical phase in the poultry life cycle, with normal progression significantly influencing both hatchability and meat production. The late stage of embryonic development is the preparation stage for energy storage, mainly through muscle mobilization, especially starting to convert into aerobic respiration before hatching, and glycolysis once again becomes the main energy source for this process (Moran, 2007). Concurrently, the embryo relies on the yolk sac for nutrient uptake from the egg yolk, driving increased fat metabolism and enhanced respiratory capacity (Liu et al., 2021). Beyond the liver, the pectoral muscle plays a key role in glycogen storage and mobilization during pre-hatching, supplying energy critical for successful hatching. Therefore, we hypothesize that the muscle development in fast-growing and slow-growing chickens is regulated by transcriptional differences and distinct regulatory networks by the functional analyses.

Transcription factors are involved in the proliferation and differentiation of myoblasts and the formation of myotubes during embryonic development, including myogenic regulatory factors (MRFs) (Hernández-Hernández et al., 2017), paired box (PAX) (Dahl et al., 1997), and myocyte enhancer factor 2 (MEF2) family (Brand, 1997). In addition to coding RNA, the role of non-coding RNA (ncRNAs) has also been reported by many studies in myogenesis (Buckingham and Rigby, 2014; Luo et al., 2013). Despite the limited characterization to date of functional small RNAs, including microRNAs (miRNAs), lncRNAs, and circRNAs, they appear to regulate key biological processes affecting skeletal muscle development and muscle type transformation (Ma et al., 2015; Shen et al., 2023; Wang et al., 2018). The ceRNA regulation network (circRNA225-gga-miR-1306-5p-heat shock protein alpha 8 (HSPA8)) was revealed in chicken muscle development from the embryonic to post-hatching periods (Lei et al., 2022). And the miRNA-34a-5p inhibited chicken myoblast proliferation and differentiation by inhibiting NOTCH1 expression (Ling et al., 2025). The molecular regulatory mechanism of embryonic muscle development in Chengkou mountain chicken was identified by weighted gene co-expression network analysis (WGCNA), including gga-miR-130b-5p for embryonic days 12 (E12), gga-miR-1643-5p for E16, gga-miR-12218-5p for E19, and gga-miR-132b-5p for E21 (Shi et al., 2022). However, there are still few studies on the interactions of several small RNAs, making it difficult to fully reveal the regulatory network mechanisms at the transcriptional level.

The TaoYuan (TY) chicken is a local slow-growing, yellow-feathered chicken with originating from Taoyuan County, Hunan Province, China. In contrast, the Arbor Acres (AA) chicken is a popular commercial broiler known for its fast growth. Our aim is to explore the muscle fiber hypertrophy of these two breeds in the late stage of embryonic development, identify the differentially expressed genes (DEGs) between local chickens and specialized broilers, and comprehensively analyze small RNAs (miRNA, lncRNA, and circRNA), with a focus on investigating the competitive endogenous RNA (ceRNA) regulatory network. To reveal whether the muscle development of fast-growing AA broilers and slow-growing TY broilers in the later stage of embryonic development is regulated by different transcriptomes and ceRNA mechanisms, and thereby provide new insights and references for understanding the characteristics of muscle development in the later stage of embryonic development and identifying specific genes and significantly enriched pathways between local breeds and commercial broilers.

Materials and methods

Ethical statement

The animal experiments detailed in this paper were carried out in strict compliance with relevant national regulations regarding animal ethics and welfare, as well as the ethical guidelines and the animal experimental safety review system of Hunan Agricultural University (approval number: HAU ACC 2024089).

Sample collection

The fertilized eggs we used were respectively sourced from Hunan Shuncheng Industrial Co., Ltd Company (AA broilers) and Hunan Xiangjia Husbandry Limited by Share Ltd Company (TY chickens). And the age of the laying hens was between 38 and 40 weeks. A total of 120 fertilized eggs were incubated at 37.8°C and 55 % humidity in the incubator (YD-1000, Shandong Yuda Incubation Equipment Co., Ltd, China) (Huo et al., 2021; Ren et al., 2021). On the eighth day of incubation, eggs were candled with a flashlight to remove unfertilized eggs and those containing dead embryos. We collected fresh pectoral muscle tissues at three time points (E17, E19, and E21) respectively, with at least six embryos collected at each time point to ensure that there were three duplicate female samples at each time point. Eventually, we determined 18 pectoral muscle samples, and the right pectoral muscle samples were obtained and preserved in 4 % paraformaldehyde for subsequent histological analysis. At the same time, left pectoral muscle samples were collected and rapidly frozen in liquid nitrogen for transcriptome sequencing. Gender identification was carried out by collecting leg muscle for DNA isolation, using chromodomain helicase DNA binding protein 1 (CHD1) primers (F:5′-GTTACTGATTCGTCTACGAGA-3′, R:5′-ATTGAAATGATCCAGTGCTTG-3′).

Morphological analysis of chicken breast muscles

For each sample, two sections cut at thicknesses of 5 microns were prepared and then subjected to Hematoxylin and Eosin (HE) staining. The simplified procedure for HE staining were as follows: the sections were stained with haematoxylin for 2 minutes, rinsed twice with water, successively immersed in 60 % ethanol for 2 minutes, 70 % ethanol for 2 minutes, 80 % ethanol for 2 minutes, 90 % ethanol for 2 minutes, stained with eosin for 2 seconds, placed in 95 % ethanol for 1 minute, dehydrated in absolute ethanol for 6 minutes, cleared in xylene for 15 minutes, and finally mounted using neutral resin. The 18 prepared sections were scanned using a light microscope at a magnification of × 100 to obtain images for subsequent analysis. The cross-sectional area (CSA) (µm²) was calculated by randomly selecting three views through Image Pro Plus software (version 6.0, Media Cybernetics Corporation, USA). The number of muscle fibers in each field of vision was at least 100, muscle fiber size (µm²) was defined as the mean CSA of counted fibers (Huo et al., 2021).

RNA extraction and library preparation

Total RNA was isolated from breast muscle using TRIzol reagent (Invitrogen, Carlsbad, CA, USA). The integrity of the RNA was evaluated using agarose gel electrophoresis and an Agilent 2100 Bioanalyzer (Agilent Technologies, Waldbronn, Germany). The concentration of high quality RNA samples (i.e., A260/280 was between 1.8-2.2, A260/230 ≥ 2.0, and RNA integrity numbers ≥ 7) was normalized to 500 ng/µL, and then used in subsequent experiments. For mRNA, lncRNA, and circRNA, each RNA sample underwent ribosomal RNA (rRNA) depletion using the Ribominus Eukaryotic kit (Invitrogen, Carlsbad, CA, USA). Briefly, reverse transcription was carried out, followed by cDNA library construction (Gao et al., 2019). Each qualified library was paired-end sequenced on the Illumina Novaseq 6000 platform. At the same time, the miRNA libraries were constructed using the TruSeq Small RNA Sample prep Kit (Invitrogen, California, USA) and enriched via PCR amplification. All steps were performed according to the manufacturer's protocols. The miRNA libraries were analyzed for quality control and the average size of inserts was approximately 140-150 bp (Wang et al., 2022). Subsequently, the libraries were purified by gel electrophoresis, examined with the Agilent High Sensitivity DNA Kit (5067-4626, Agilent Technologies, California, USA), and quantified using the Quan-iTTM PicoGreenTM dsDNA Assay Kit (Q33120, Thermo Fisher Scientific, California, USA). Amplification was executed for the PCR products. The cDNA library and miRNA library were then sequenced on a Hiseq platform (Illumina) by Shanghai Personal Biotechnology Cp. Ltd (Shanghai, China) (Fu et al., 2022).

High-throughput sequencing

Raw reads were processed using Trimmomatic (Bolger et al., 2014) with default settings to obtain clean reads. The clean reads were subsequently aligned to the chicken reference genome GRCg7b (http://asia.ensembl.org/Gallus_gallus/Info/Index) by Hierarchical Indexing for Spliced Alignment of Transcripts 2 (HISAT2) v2.2.1 (Kim et al., 2019) and assembled by StringTie v2.1.6 (Kovaka et al., 2019). The transcript sequences were reconstructed through splicing of the clean reads. By comparing these reconstructed sequences with known mRNA and lncRNA transcripts, novel lncRNAs were identified. Fragments per kilobase of transcript per million mapped reads (FPKM) was applied to represent lncRNA expression levels. After the clean reads were aligned to the chicken reference genome GRCg7b, the junctions of unmapped reads were detected using a back-splice algorithm. Subsequently, circRNAs were predicted by the Findcirc software (Memczak et al., 2013). Mapped back-splicing junction reads per million mapped reads (RPM) was used to indicate the expression levels of circRNAs.

Construction and sequencing of small RNA library

Initially, the raw data underwent de-jointing and mass-filtering processes. Subsequently, it was re-processed using the filtered sequence. Through comparison with the chicken reference genome GRCg7b, repetitive small RNA sequences were eliminated, and the expression abundance of miRNA was calculated. To precisely quantify miRNA expression levels, we computed and normalized them to transcripts per million (TPM). For the prediction of miRNA target genes, we established stringent parameters for the miRanda software. Specifically, we required binding free energy values to be less than -10 kcal/mol and binding scores to be greater than 50.

Quantitative real-time PCR assays

The cDNA was synthesized by reverse transcription using miRNA 1st Strand cDNA Synthesis Kit (by tailing A) (Vazyme, Nanjing, China). Quantitative analysis was performed on a the LightCycler 480 System (Roche, Switzerland) using miRNA Unimodal SYBR qPCR Master Mix (Vazyme, Nanjing, China). Primers were synthesized by Sangon Biotech Co., Ltd. (Shanghai, China) and listed in Table S1. U6 and 18s were selected as the internal controls (Table S1). Calculation of relative expression was same with previously described 2^−ΔΔCt method (Livak and Schmittgen, 2001).

Weighted gene co-expression network analysis (WGCNA)

WGCNA analysis was conducted in the R environment following a standard pipeline for 16,142 genes. Briefly, the soft threshold (β = 7) was determined based on the scale-free distribution (R² > 0.75). In WGCNA, to assess whether two genes exhibit similar expression patterns, it is customary to set a soft threshold for screening, genes with values above this threshold are regarded as having similar patterns. Moreover, WGCNA employs correlation coefficient weighting, wherein the correlation coefficient of genes is raised to the Nth power, ensuring that the connections between genes in the network adhere to a scale-free distribution. Subsequently, we adopted stepwise and dynamic cutting methods to construct the gene network and identify modules, with parameters set as minModuleSize = 50 and mergeCutHeight = 0.25. For each embryonic age, the module having the highest correlation coefficient was selected as the target module. Finally, hub genes within the target module were defined according to the criteria of gene significance (GS) > 0.2 and module membership (MM) > 0.8.

Differential expression analysis and pathway enrichment analysis

In the R environment, gene expression levels were normalized using the DESeq2 package (Love et al., 2014). Differentially expressed RNAs (DERs) were identified, including differentially expressed mRNAs (DEGs), differentially expressed miRNAs (DEMs), differentially expressed lncRNAs (DELs), and differentially expressed circRNAs (DECs). The criteria for identification were a fold change > 2 and P value < 0.05. And the volcano maps, venn maps, and upset maps of DERs were all visualized by Hiplot (ORG) (http://hiplot.cn).

Subsequently, all these DERs along with the hub genes were extracted to perform Kyoto Encyclopedia of Genes and Genomes (KEGG) and Gene Ontology (GO) enrichment analyses using KOBAS 3.0 (http://bioinfo.org/kobas). The enrichment pathways were then visualized using Hiplot (ORG).

Prediction of the ceRNA regulatory network

To identify potential ceRNAs among the differentially expressed RNAs, a comprehensive two-part filtering strategy was employed. The first part involved examining the targeting relationship and inversely correlated expression levels between miRNAs and candidate ceRNAs. Predictions from the miRanda software were required to have binding free energy values lower than -15 kcal/mol and binding scores greater than 70. Additionally, the positively correlated expression patterns of candidate ceRNAs and the enrichment level of shared miRNA binding sites were taken into consideration. In the second part, the focus was on DERs (mRNAs, lncRNAs, and circRNAs) that had regulatory relationships with DEMs. A stringent filtering criterion based on the Spearman correlation coefficient was applied. Specifically, target gene pairs with coefficients of -0.7 or less were selected. Given the ceRNA hypothesis stating that ceRNAs show positive correlations in their expression levels, the analysis was further refined by identifying ceRNA pairs with a Spearman correlation coefficient of 0.7 (P < 0.05). Finally, the ceRNA regulatory network was visualized using Cytoscape (version 3.10.1) (Shannon et al., 2003). This tool effectively portrays the complex interactions among ceRNAs.

Statistical analysis

Principal component analysis (PCA) of four RNAs was carried out using R software (version 4.3.3). The data are reported as the mean ± standard deviation for each set of independent experiments. When applicable, statistical significance was evaluated via two-way analysis of variance using SPSS software (version 18.0). Significance levels of P < 0.05, 0.01, or 0.001 were utilized. All results were presented as means and standard deviations and analyzed with GraphPad Prism 8 (GraphPad Software, Inc, San Diego, CA, US).

Results

Phenotypic differences between AA broilers and TY chickens

Based on the set time into incubation, we ultimately selected three time points (E17, E19, and E21) for histomorphological assessments and transcriptome sequencing, the samples at each time point were all female and met three biological replicates (Fig. 1A). There was a significant interaction between the breeds and time for the embryo weight (P < 0.01). Phenotypically, AA broiler embryos exhibited significantly higher embryo weights compared to TY chickens at all late embryonic stages, with embryonic weight showing a progressive increase with advancing age in both groups (Fig. 1B). Histomorphological analysis revealed well-defined muscle fiber structures and a temporal expansion of CSA during development (Fig. 1C). Notably, while AA broilers consistently demonstrated larger CSAs than TY chickens at each examined time point, neither breed showed statistically significant CSA increases across the studied period (Fig. 1D).

Fig. 1.

Embryo weight and cross-sectional area of muscle fibers between AA broilers and TY chickens. (A) Research technology roadmap. HE: Hematoxylin and Eosin. (B) The embryo weight of AA broilers and TY chickens at three time points. * P < 0.05, ** P < 0.01, *** P < 0.001. (C) Histological characterizations of AA broilers and TY chickens at three time points. (D) The cross-sectional area (CSA) of muscle fibers in the AA broilers and TY chickens at three time points. n.s. P > 0.05, * P < 0.05, ** P < 0.01, *** P < 0.001. AA: Arbor Acres, TY: TaoYuan. E17: embryonic days 17, E19: embryonic days 19, E21: embryonic days 21.

Transcriptome profiling of mRNA in embryonic muscle

A total of 18 libraries were generated, with three biological replicates. We acquired 13.92-19.80 billion raw reads and 90.86-128.19 million high-quality clean reads, with an average mapping rate of 91.62 %. And Q20 and Q30 were both above 96.50 % and 93.84 % respectively (Table S2). In the two chicken breeds, a total of 16,142 genes exhibited non-zero expression across the three time points (E17, E19, and E21). PCA analysis revealed a separation trend among the six sample groups, with the E19 samples showing the most distinct separation (Fig. 2A). We conducted WGCNA analysis on 18 samples (Fig. S1A). Using an R² value greater than 0.75 as the standard, the optimal soft threshold was determined to be β = 7 (Fig. S1B). Modules with a similarity greater than 0.75 were clustered based on module eigengenes (Fig. S1C), resulting in the identification of 22 modules (Fig. S1D). As shown in Fig. 2B, different breeds displayed specific gene-expression modules at the three time points. The coral1, honeydew1, lightpink4, mediumpurple3, green, and thistle1 modules were identified as the target modules, which were considered to be the most relevant ones. A total of 1,635 hub genes including GABRA4, FOSL2, PDE4B, TRIB1, THBS1, and FBXO32, within these target modules were determined using the criteria (GS > 0.2 and MM > 0.8) (Fig. S2). Based on KEGG annotation, although the functional enrichment of hub genes in time-specific modules varied, the pathways in which these different hub genes were significantly enriched were nearly consistent. These pathways included PPAR signaling pathway, MAPK signaling pathway, Wnt signaling pathway, mTOR signaling pathway, and FoxO signaling pathway, all of which were significantly enriched in both groups (P < 0.05) (Fig. S3 and Table S3).

Fig. 2.

Screening of key mRNA expression genes in embryonic muscle. (A) Principal component analysis (PCA) of mRNA expression across all samples. (B) Correlation analysis between the module and embryo weight. Red represents a positive correlation, blue represents a negative correlation, and a darker color indicates a stronger the correlation. (C-E) Volcano plot of differentially expressed mRNAs (DEGs) between AA broilers and TY chickens at three time points. Red dots: Represent the DEGs significantly up regulated in the TY group. Blue dots: Represent the DEGs significantly down regulated in the AA group. Gray dots: Represent genes that did not meet the significance threshold. (F- G) Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis in DEGs (P < 0.05). (H) The overlap genes from hub genes and DEGs. AA: Arbor Acres, TY: TaoYuan. E17: embryonic days 17, E19: embryonic days 19, E21: embryonic days 21.

Despite significant differences in coding-gene expression during late embryonic development, our focus was primarily on the gene expression differences between the two breeds. Through differential expression analysis, 4,577 DEGs were identified, with E19 having the largest number of DEGs (Fig. 2C–E and Table S4). Up regulated genes (FOSL2, PDE4B, TRIB1, THBS1, and FBXO32) were predominantly more highly expressed in TY chickens, while down regulated genes were more characteristic of AA broilers (Fig. 2C–E). Down regulated genes (GABRA4) were significantly enriched mainly in the regulation of actin cytoskeleton and cardiac muscle contraction pathways (Fig. 2F). In contrast, the significantly up regulated differential genes in TY chickens were mainly enriched in the glycolysis/gluconeogenesis pathway at the three time points (Fig. 2G). This indicates that the two breeds have different mechanisms of muscle energy supply. In addition, by integrating data from both breeds at the three time points based on WGCNA and differential expression analyses, we identified 129 overlapping genes, such as GABRA4, in AA broilers and 437 overlapping genes, including FOSL2, PDE4B, TRIB1, THBS1, and FBXO32, in TY chickens (Fig. 2H).

Comprehensive miRNA expression in embryonic muscle

We randomly selected 6 miRNAs (gga-miR-206, gga-miR-449a, gga-miR-1786, gga-miR-383-3p, gga-miR-1454, and gga-miR-1744-3p) and 9 samples to verify the reliability of the miRNA library. The results showed that the sequencing quality met the standards (Fig. S4). Analysis of miRNAs in the 18 samples revealed that known miRNAs constituted the largest proportion among all detected miRNA types (Fig. 3A). In total, 1,066 miRNAs were successfully identified, with 884 known miRNAs and 182 novel ones. Each of the 18 samples contained over 300 miRNAs (Fig. 3B). PCA analysis of miRNAs from each time period clearly distinguished between the two chicken breeds (Fig. 3C). Employing differential expression analysis with the screening criteria of P < 0.05 and |log2FC| ≥ 1, we identified 9 significantly down regulated miRNAs and 3 significantly up regulated miRNAs at E17 in the comparison between AA broilers and TY chickens (Fig. 3D). At E19, 33 miRNAs were significantly up regulated and 89 were significantly down regulated (Fig. 3E). At E21, 6 miRNAs were significantly up regulated and 7 were significantly down regulated (Fig. 3F). A total of 143 DEMs were screened at the three time points. Similar to the pattern observed with mRNAs, E19 had the largest number of DEMs (Table S5). Specifically, gga-miR-383-3p, gga-miR-449a, and gga-miR-449c-5p were significantly down regulated among the DEMs between E19 and E21, while gga-miR-206 was the only miRNA significantly up regulated between E17 and E19 (Fig. 3G).

Fig. 3.

Descriptive statistics and differential expression analysis of miRNA in embryonic muscle. (A) The percentage of all kinds of miRNA detected across all samples. (B) The known miRNA number of all samples. (C) Principal component analysis (PCA) of miRNA expression across all samples. (D-F) Volcano plot of the differentially expressed miRNAs (DEMs) between AA broilers and TY chickens at three time points. Red dots: Represent the DEMs significantly up regulated in the TY group. Blue dots: Represent the DEGs significantly down regulated in the AA group. Gray dots: Represent genes that did not meet the significance threshold. (G) Venn map of the differential expressed miRNAs. AA: Arbor Acres, TY: TaoYuan. miRNA: microRNA. E17: embryonic days 17, E19: embryonic days 19, E21: embryonic days 21.

Characteristics and expression of lncRNAs in embryonic muscle

In total, 29,130 lncRNAs were identified across 18 samples. PCA analysis revealed a trend of separation among the six groups (Fig. 4A). To assess the expression levels of each lncRNA in every sample, the quantity of lncRNA expression in each sample was visualized. Similar distribution patterns were observed (Fig. 4B). A total of 3,159 DELs were screened at the three time points. In the volcano plots, 436 DELs were found to be significantly down regulated, while 561 DELs were significantly up regulated at E17 (Fig. 4C and Table S6). And 734 DELs were found to be significantly down regulated, while 824 DELs were significantly up regulated, and DELs were most abundant at E19 (Fig. 4D and Table S6). 435 DELs were found to be significantly down regulated, while 480 DELs were significantly up regulated (Fig. 4E and Table S6). Additionally, there was a minor difference in the number of differential genes between E17 and E19 (Fig. 4F). Subsequently, it was discovered that the expression of three DELs (ENSGALT00010044323, ENSGALT00010072195, and MSTRG.29244.1) was remarkably high in AA broilers at the three time points (Fig. 4G). Moreover, 7 DELs, including MSTRG.19399.1, MSTRG.1986.1, MSTRG.20244.6, and so on, were significantly expressed in the TY group (Fig. 4H).

Fig. 4.

Descriptive statistics and differential expression analysis of lncRNAs in embryonic muscle. (A) Principal component analysis (PCA) of lncRNA expression across all samples. (B) Bar graph showcasing the expression levels of all discovered lncRNAs across 18 individual samples. (C-E) Volcano plot of the differentially expressed lncRNAs (DELs) between AA broilers and TY chickens at three time points. Red dots: Represent the DELs significantly up regulated in the TY group. Blue dots: Represent the DELs significantly down regulated in the AA group. Gray dots: Represent genes that did not meet the significance threshold. (F) Upset map of the DELs between AA broilers and TY chickens at three time points. (G-H) Venn map of the DELs between AA broilers and TY chickens at three time points. AA: Arbor Acres, TY: TaoYuan. E17: embryonic days 17, E19: embryonic days 19, E21: embryonic days 21.

Expression patterns of circRNAs in embryonic muscle

We explored the distinctive characteristics and expression patterns of circRNAs during the late stages of embryonic development (E17, E19, and E21). The classification of 19,231 circRNAs indicated that exonic circRNAs were the most prevalent in pectoral muscle, with a total of 12,922 being identified (Fig. 5A). Intriguingly, 1,455 circRNAs were formed from a single exon. Additionally, we discovered 1,975 circRNAs derived from exon-intron regions. In contrast, intronic circRNAs, intergenic circRNAs, and antisense strand circRNAs were less abundant, numbering 692, 547, and 1,640 respectively (Fig. 5A). Meanwhile, exonic circRNAs constituted over 60 % of the total in the 18 samples (Fig. 5B). All six types of circRNAs were identified in chicken breast muscle, with exonic derived circRNAs showing remarkable abundance (Fig. 5B). The results of PCA analysis on circRNAs from each period demonstrated a clear grouping between the two breeds. However, circRNAs from the six groups were nearly all clustered together (Fig. 5C). Subsequently, we conducted a differential expression analysis of circRNAs in each period. DECs were screened using a fold change greater than 2 and a P value less than 0.05 as the criteria. Ultimately, we obtained 33, 63, and 11 DECs respectively (Table S7). Among these, 15 genes were up regulated and 18 were down regulated at E17, 31 genes were up regulated and 32 were down regulated at E19, and 5 genes were up regulated and 6 were down regulated at E21 (Fig. 5D-F). Furthermore, we found that only one DEC (ggacirc-134224) had significantly high expression in AA broilers at the three time points. In contrast, no significant expression was observed in the TY group (Fig. 5G).

Fig. 5.

Descriptive statistics and differential expression analysis of circRNAs in embryonic muscle. (A) The number of all circular RNA types detected in the breast muscle. (B) The percentage of all kinds of circRNA detected across all samples. (C) Principal component analysis (PCA) of circRNA expression with all samples. (D-F) Volcano plot of the differentially expressed circRNA (DECs) between AA broilers and TY chickens at three time points. Red dots: Represent the DECs significantly up regulated in the TY group. Blue dots: Represent the DECs significantly down regulated in the AA group. Gray dots: Represent genes that did not meet the significance threshold. (G) Venn map of the differentially expressed circRNAs between AA broilers and TY chickens at three time points. AA: Arbor Acres, TY: TaoYuan. E17: embryonic days 17, E19: embryonic days 19, E21: embryonic days 21.

Construction of the ceRNA regulatory network in breast muscle

To construct the ceRNA regulatory network, we first predicted miRNA binding to mRNAs by miRanda software to form miRNA-mRNA pairs (Fig. S5A). First, we screened 4 DEMs, including gga-miR-206 (between E17 and E19), gga-miR-383-3p, gga-miR-449a and gga-miR-449c-5p (between E19 and E21). Meanwhile, we filtered RNAs that were differentially expressed in our data, and established 79 pairs of mRNA-miRNA interactions, 27 pairs of lncRNA-miRNA interactions and 27 pairs of circRNA-miRNA interactions, respectively (Fig. S5B). Then, we calculated the spearman correlation of mRNA and lncRNA/circRNA respectively. The lncRNA-miRNAs/circRNA-miRNAs pairs with significant correlation coefficient greater than 0.7 were filtered out. Finally, we visualized the ceRNA regulatory network using Cytoscape software (Fig. S5A). Regarding the circRNA-miRNA-mRNA relationships (Table S6), the circRNAs did not solely originate from the intersection of two time points. There were 2 ceRNA regulatory networks constructed from lncRNA (lncRNA (8)-miRNA (1)-mRNA (27)) and circRNA (circRNA (3)-miRNA (1)-mRNA (6)) respectively (Fig. 6A-B and Table S8-9), and the gga-miR-206 was highly expressed in the TY chickens. Meanwhile, the 2 ceRNA regulatory networks were constructed from lncRNA (lncRNA (10)-miRNA (3)-mRNA (25)) and circRNA (circRNA (6)-miRNA (3)-mRNA (19)), respectively, and the 3 DEMs were highly expressed in the AA broilers, including gga-miR-383-3p, gga-miR-449a and gga-miR-449c-5p (Fig. 6C-D and Table S8-9).

Fig. 6.

The competing endogenous RNA regulatory network from four microRNAs. (A) CircRNA-miRNA-mRNA network at E17 and E19 (B) LncRNA-miRNA-mRNA network at E17 and E19. (C) CircRNA -miRNA-mRNA network at E19 and E21. (D) LncRNA-miRNA-mRNA network analysis at E19 and E21. E17: embryonic days 17, E19: embryonic days 19, E21: embryonic days 21.

Discussion

AA broilers are fast-growing white-feathered broiler breeds, specifically selected for high meat production performance. In contrast, TY broilers are a local breed in Hunan, representing the slow-growing yellow-feathered broiler type to meet the demand of Chinese consumers for high quality meat. This study found that the embryo weight and muscle fiber CSA of TY chickens in the late stage of embryonic development were significantly lower than those of AA broilers, indicating that yellow-feathered chickens have greater potential for genetic improvement.

To comprehensively understand the developmental patterns of the two lines during the late embryonic stage, we carried out comprehensive transcriptome sequencing. Regarding coding RNA, we separately conducted differential expression analysis and WGCNA analysis. Based on the module-traits association results of WGCNA, the significantly enriched gene sets of each strain at the three time points (E17, E19, and E21) were not consistent, demonstrating breed-specific characteristics. Pathway enrichment analysis results, which included pathways such as the adipocytokine signaling pathway, PPAR signaling pathway, regulation of actin cytoskeleton, and glycerolipid metabolism, indicated that the main function of genes in the late-stage development was to mobilize myocytes and adipocytes to supply energy for hatching. This finding is in line with previous research (De Lima et al., 2020; Grabner et al., 2021; Kistner et al., 2022; Pi et al., 2023). Meanwhile, the results of differential expression analysis also showed that the above-mentioned pathways were significantly enriched. Notably, the glycolysis/gluconeogenesis pathway was particularly prominent in TY chickens at three time points. This observation suggests that embryonic muscle development during this phase may prioritize glycolytic energy production to support hatching rather than substantial fiber hypertrophy.

Generally speaking, miRNAs are a class of widely distributed single-stranded RNA molecules, approximately 18 to 26 nucleotides in length. They play a pivotal role in transcriptional gene regulation (Du and Zamore, 2007). In the ceRNA hypothesis, miRNAs play a key negative regulatory role, combining with mRNAs, lncRNAs, and circRNAs to regulate gene expression and participate in biological processes such as cell aging and apoptosis. Significantly, mRNAs, lncRNAs, circRNAs, and pseudogenes can competitively bind to miRNAs with similar binding sites, thereby exerting mutual regulatory effects on one another (Tay et al., 2014). They are associated with crucial regulatory functions in muscle development, including myoblast proliferation, differentiation, myotube fusion, and fiber hypertrophy or atrophy (Bartel, 2004). For example, miR-29 was shown to influence muscle development, and it can target Akt3 to reduce mouse myoblast proliferation and promote myotube formation (Wei et al., 2013). In our study, we screened three miRNAs (gga-miR-29a-3p, gga-miR-29b-3p, and gga-miR-29c-3p), all belonging to the gga-miR-29 family at E19. It is possible that these three miRNAs also contribute to muscle fiber transformation (Wang et al., 2012). Moreover, miR-499 is a muscle-specific miRNA that plays a vital role in muscle development (Liu et al., 2016). It is involved in regulating the proliferation and differentiation of myocytes, as well as muscle fiber type conversion by targeting different genes. At E21, gga-miR-499-3p and gga-miR-499-5p were significantly differentially expressed (Table S5). The functions of these differentially expressed miRNAs, particularly gga-miR-384-3p and gga-miR-206, require further verification.

LncRNAs are a class of RNA molecules with a length exceeding 200 nucleotides. They play crucial regulatory roles in diverse biological processes, including muscle development and regeneration (Chen et al., 2024; Niu et al., 2024). LncRNAs are involved in gene expression regulation through multiple mechanisms, among which the regulation of miRNAs is notable. By means of the ceRNA mechanism, lncRNAs can sequester miRNAs by harboring miRNA-response elements. This action alleviates the inhibitory effect of miRNAs on their target genes. In our study, we successfully identified 29,130 lncRNAs and characterized their unique features in chicken breast muscles during the late embryonic development stage. Notably, when comparing AA broilers and TY chickens, 1,558 DELs were detected at E19. This number was the highest among the three developmental stages examined, indicating their potential significance in the developmental processes of chicken breast muscles. For example, SYISL, a lncRNA highly expressed in muscle, promotes myoblast proliferation and fusion while inhibiting myogenic differentiation. In mouse models, knockout of SYISL leads to a significant increase in muscle fiber density and muscle mass (Jin et al., 2018). Additionally, another lncRNA, lncMREF, interacts with SMARCA5 upon the activation and differentiation of muscle satellite cells. This interaction enhances chromatin accessibility, facilitating the genomic binding of p300/CBP/H3K27ac and subsequently upregulating the expression of myogenic regulators such as MyoD, thereby promoting cell differentiation (Lv et al., 2022). Therefore, lncRNAs play pivotal roles in the spatio-temporal regulation of muscle development and regeneration, particularly in embryonic muscle tissue.

CircRNAs are a unique class of ncRNAs characterized by covalently linked 5′ and 3′ ends, which impart increased stability (Patop et al., 2019). These circular RNA molecules can be derived from diverse sources, including exons, introns, a combination of exons and introns, intergenic regions, and the antisense strand of genes (Ouyang et al., 2017). Our results revealed that the proportion and quantity of annot-exons were the highest, suggesting their potential involvement in regulating gene expression. This might occur by influencing the splicing process of mRNA, thereby generating different protein isomers. CircRNAs are ubiquitously present in the cells of eukaryotic organisms and play substantial roles in various biological processes. Current research indicates that circRNAs are implicated in muscle development and form a ceRNA regulatory network with mRNA and miRNA, which is involved in the molecular mechanisms underlying muscle development (Yan et al., 2023; Yuan et al., 2022; Zhang et al., 2019). For instance, through dual-luciferase reporter assays and functional studies, circCSDE1 was shown to regulate the proliferation and differentiation of C2C12 myoblasts by sponging miR-21-3p (Sun et al., 2022). Interestingly, circFGFR2 functions as a sponge for miR-133, regulating the MAP3K20 and JNK/MAPK pathway. This regulation inhibits myoblast proliferation while promoting differentiation and skeletal muscle regeneration. It inhibits chicken skeletal muscle satellite cell differentiation by sponging miR-128-3p (Shen et al., 2019). Similarly, the largest number of differentially expressed DECs between the two chicken breeds (AA and TY) was detected at E19. This finding implies that the muscle development process may be particularly active during this period. In future experiments, we will further incorporate the interactions identified in this study into ceRNA regulatory network.

We constructed lncRNA-miRNA-mRNA and circRNA-miRNA-mRNA ceRNA networks. The gga-miR-206, which was derived from the E17-E19 intersection, exhibited significantly higher expression in TY chickens. Muscle-specific miR-206 is known to promote muscle differentiation (Kim et al., 2006) and has been screened during the late post-natal development in Gushi chickens (Li et al., 2021). Furthermore, three miRNAs, namely gga-miR-383-3p, gga-miR-449a, and gga-miR-449c-5p, were identified from the E19-E21 intersection. To date, only gga-miR-449a has been recognized as a sponge for circEML1 and IGF2BP3, facilitating steroid synthesis in the follicular granulosa cells of chickens (Li et al., 2023). We believed that this network (ggacirc-007702/ggacirc-0988536-gga-miR-206) was a high expression regulatory network on TY chickens at E17, and ggacirc-116006-gga-miR-206 was a high-expression regulatory network on TY chickens at E19.

Notably, five genes, FOSL2, PDE4B, TRIB1, THBS1 and FBXO32, were found to bind to all three of these miRNAs. In terms of gene functions, FOSL2 has been reported to regulate glycogen content in chicken muscle (Liu et al., 2021) and THBS1 has been identified as a potential candidate gene influencing inosine monophosphate deposition in muscle (Yu et al., 2023). In addition, THBS1 inhibited myoblast proliferation and differentiation in vitro and delayed muscle regeneration in vivo. The lncMPD2 counteracted the inhibitory effects of miR-34a-5p on THBS1 and myogenesis-related gene mRNA and protein expression (Niu et al., 2024). Meanwhile, lncRNA MIR217HG acts as a potent inducer of cardiac remodeling that may contribute to heart failure by activating the miR-138/THBS1 pathway (Nie et al., 2024). Moreover, miR-431 promotes cardiomyocyte proliferation by targeting FBXO32, providing a potential molecular target for preventing myocardial injury (Li et al., 2024). We propose that FOSL2 and PDE4B interact with gga-miR-383-3p, gga-miR-449a, and gga-miR-449c-5p to form a core regulatory network. Integration with circRNAs and lncRNAs may generate a complex regulatory network that promotes muscle fiber hypertrophy in AA broilers. When the five core genes combine with three miRNAs, they interact with lncRNAs to regulate the muscle fiber hypertrophy of AA broilers. However, within the networks we constructed, the specific functions of the majority of genes and ncRNAs in late embryonic muscle development remain incompletely explored. Thus, these ceRNA networks have the potential to function as candidate regulatory networks for embryonic development, particularly when considering different breed.

Conclusion

This study establishes the first comparative transcriptomic atlas of late embryonic pectoral muscle development between fast-growing AA broilers and slow-growing TY chickens. Our findings reveal a distinct metabolic strategy in TY chickens, where muscle tissue preferentially activates glycolysis/gluconeogenesis pathway to meet the energy demands of shell pipping. Temporal analysis identified E19 as a critical developmental checkpoint, exhibiting peak differential expression of coding and non-coding RNAs between breeds. Four hub miRNAs (gga-miR-206, gga-miR-383-3p, gga-miR-449a, and gga-miR-449c-5p) emerged as central regulators, potentially orchestrating muscle development via sponge interactions with lncRNAs/circRNAs and subsequent modulation of five key genes (FOSL2, PDE4B, TRIB1, THBS1, and FBXO32). These results provide mechanistic insights into the transcriptional regulation of embryonic development between commercial and indigenous breeds. The identified ceRNA networks and metabolic signatures serve as valuable resources for future studies targeting muscle development optimization. Table 1

Table 1.

All differentially expressed RNA numbers.

RNAs	Item	E171	E192	E213
mRNA	Down4	506	2,003	175
	Up5	478	1,774	390
microRNA	Down	9	89	7
	Up	3	33	6
circRNA	Down	436	734	435
	Up	561	824	480
lncRNA	Down	18	32	6
	Up	15	31	5

¹E17: embryonic days 17.

²E19: embryonic days 19.

³E17: embryonic days 17.

⁴Down: represent the genes significantly down regulated in the AA group.

⁵Up: represent the genes significantly up regulated in the TY group.

Ethics approval

All animal experiments were performed in accordance with the protocols approved by the Committee on the Ethics of Animal Experiments of Hunan Agricultural University (HAU ACC 2024089).

CRediT authorship contribution statement

Di Zhao: Writing – review & editing, Writing – original draft, Visualization, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Yifan Shi: Validation. Jiatai Deng: Software, Methodology. Bolin Zhong: Data curation. Yuanyuan Zeng: Data curation. Qingyuan Ouyang: Writing – review & editing, Supervision. Haihan Zhang: Writing – review & editing, Supervision. Zehe Song: Writing – review & editing, Supervision, Resources, Conceptualization. Xi He: Writing – review & editing, Supervision, Resources.

Disclosures

All authors disclosed no relevant relationships.