Integrated multi-tissue transcriptomic analysis identifies hub genes associated with egg production recovery after fasting-induced molting Authors: Xiaomeng Miao, Jia Liu, Fugui Li, Xing Chen, Yimin Wei, Li Zhang, Yulong Feng, Zhonghua Ning Published: 2026 ABSTRACT AbstractFasting-induced molting is widely employed in the poultry industry to rejuvenate the reproductive system of aging laying hens, thereby improving post-molt laying performance and extending productive lifespan. However, substantial individual variation in post-molt reproductive recovery may compromise the overall effectiveness of molting programs. Elucidating the molecular basis underlying this variation is therefore important for poultry production and genetic improvement. In this study, laying hens were classified into high-recovery (HR) and low-recovery (LR) groups according to their egg production recovery following fasting-induced molting. Transcriptomic profiling of the hypothalamus, pituitary, ovary, and liver was performed using RNA sequencing (RNA-seq). A total of 271, 435, 3,425, and 289 differentially expressed genes (DEGs) were identified between the HR and LR groups in the hypothalamus, pituitary, ovary, and liver, respectively. Functional enrichment analysis showed that these DEGs were primarily associated with protein homeostasis, hormone response, ribosome function, and immune regulation, indicating distinct tissue-specific biological functions during post-molt recovery. Integration of weighted gene co-expression network analysis (WGCNA) and protein-protein interaction (PPI) network analysis further identified HSP90AA1, GDF7, CTSS, and STS as candidate hub genes associated with post-molt reproductive recovery. Collectively, this study provides a comprehensive multi-tissue transcriptomic landscape of post-molt egg production recovery and identifies key biological processes and candidate genes potentially involved in this process. These findings improve our understanding of the molecular mechanisms underlying post-molt reproductive recovery and provide valuable genetic resources for the genetic improvement of laying hens with superior post-molt performance.Keywords: Laying hen, Fasting-induced molting, Egg production recovery, Multi-tissue transcriptomics, Hub gene Introduction China is the world’s largest producer and consumer of eggs (Chen et al., 2024). In 2025, national egg production exceeded 30 million tons, and the laying hen population surpassed 1.3 billion (Xu et al., 2026). The scale of the industry places considerable pressure on breeder stock management, feed resource supply, and environmental sustainability. Increasing total egg output per hen or extending the productive lifespan of laying hens is therefore considered an effective strategy to reduce flock size, improve feed utilization, and alleviate environmental burdens. In line with this goal, the poultry industry proposed a breeding target of “500 eggs in 100 weeks” in 2011 (Bain et al., 2016). However, under conventional production conditions, laying hens typically undergo natural molting before culling, which is accompanied by a marked decline in egg production and egg quality (Zhang et al., 2025). In addition, hens show considerable individual variation in the timing of laying cessation and recovery during natural molting, and some individuals require more than 10 weeks to resume egg production (Ma et al., 2025). To address this problem, induced molting has been widely adopted in commercial production. Management practices such as feed withdrawal, water restriction, and photoperiod manipulation are used to induce synchronized molting within a flock, leading to temporary regression of the reproductive system followed by redevelopment during recovery (Zhu et al., 2025). As a result, laying cessation occurs rapidly and synchronously, and most hens resume egg production within 7–9 weeks. This practice improves post-molt laying performance and egg quality while extending the productive lifespan of laying hens (Wang et al., 2024a). Nevertheless, substantial individual variation in post-molt reproductive recovery is still observed under practical production conditions. Such variation delays flock-level recovery and reduces overall production efficiency. Previous studies have mainly focused on comparisons of molting methods (Saeed et al., 2023; Torki and Sedgh-Gooya, 2026), changes in laying performance, physiological parameters, gut microbiota, and gene expression during molting (van Eck et al., 2025; Xue et al., 2025), and nutritional or additive interventions aimed at promoting reproductive recovery (Luo et al., 2025; Sun et al., 2025). However, the molecular mechanisms underlying individual variation in post-molt egg production recovery remain largely unclear. From a systems biology perspective, transcriptomic sequencing combined with bioinformatics analysis provides a powerful approach for investigating the molecular regulation of economically important traits in poultry (Han et al., 2025). In laying hens, transcriptomic profiling of ovarian follicles from high- and low-yielding hens revealed activation of signaling pathways including PI3K-AKT-FOXO3, TGF-β, and Wnt/β-catenin in high-producing birds, providing potential targets for improving egg production performance (Wadood et al., 2025). Similarly, comparative transcriptomic analyses of hypothalamic, pituitary, and ovarian tissues from high- and low-yielding indigenous chickens identified key candidate genes, including POMC, NPY, CRH, and WNT3A, involved in processes related to neuroendocrine regulation, neurotransmitter transport, steroid hormone synthesis, and signal transduction (Liu et al., 2026). Egg production is a complex quantitative trait regulated by multiple tissues and numerous genes with small individual effects (Fu et al., 2023). The hypothalamic–pituitary–ovarian (HPO) axis controls follicular development and ovulation through endocrine signaling (Leng et al., 2024; Wang et al., 2024c), whereas the liver is the primary organ responsible for lipid and protein synthesis required for egg formation and plays a critical role in laying performance (Wang et al., 2024b). Therefore, integrated analysis of multi-tissue transcriptomic data is essential for elucidating the molecular mechanisms underlying post-molt egg production recovery in laying hens subjected to fasting-induced molting. Based on this framework, the present study aimed to elucidate the molecular basis underlying individual variation in post-molt egg production recovery in laying hens subjected to fasting-induced molting. By comparing hens with different post-molt egg production recovery performance and integrating transcriptomic data from the hypothalamus, pituitary, ovary, and liver, we sought to identify tissue-specific and shared molecular pathways associated with post-molt egg production recovery and to characterize the potential regulatory features across these tissues. These findings are expected to improve our understanding of the molecular regulation of post-molt egg production recovery and provide a theoretical basis for breeding laying hens with extended laying cycles and improved post-molt recovery performance. Materials and methods Ethics statement All animal procedures were approved by the Ethics Committee of the Guizhou Institute of Animal Husbandry and Veterinary Medicine (Approval No. AWE-GZSXMSY-2024-10) and were conducted in accordance with the institutional guidelines for the care and use of experimental animals. Animals and experimental design A total of 1,084 healthy Hy-Line Gray laying hens (61 weeks of age) from a single production batch were obtained from Ceheng Lunda Livestock and Poultry Co., Ltd. (Guizhou, China). Birds were housed individually in cages with ad libitum access to water and were provided 120 g of feed per bird daily. The photoperiod was maintained at 16 h light and 8 h dark (16L:8D). After a 1-week acclimation period, individual egg production was recorded daily beginning at 62 weeks of age. Egg number during weeks 63–69 (EN63–69 w) was calculated for each hen as the pre-molt egg production trait. Vaccinations against avian influenza (AI) and Newcastle disease (ND) were administered at the end of week 65. At 69 weeks of age, all hens were weighed individually after an overnight fast, and body weight at 69 weeks of age (BW69w) was recorded as the pre-molt body weight trait before the hens were subjected to a two-phase fasting-induced molting program. During the fasting phase (weeks 70–71), feed was completely withdrawn, water was withheld for the first 2.5 days, and the photoperiod was reduced to 8L:16D. The fasting phase ended on day 12, when the average body weight loss reached 26.98%, which was within the target range (25%–30%) for successful molting induction. During the recovery phase (beginning in week 72), feed was reintroduced at 20 g/day and increased by 20 g/day until the standard feeding level of 120 g/day was reached. During the first 11 days of recovery, birds received a pullet-grower diet containing 2% limestone. The diet was then changed to a layer ration containing 3% limestone for 3 days. Thereafter, the limestone level was increased by 1 percentage point for every 10% increase in laying rate until a final level of 8% was reached and maintained. Individual egg production was continuously recorded during the recovery period. Egg numbers during weeks 72–75 (EN72–75 w), weeks 76–80 (EN76–80 w), and weeks 72–80 (EN72–80 w) were calculated for each hen. Age at first egg (AFE) after molting was determined from the daily egg production records as the age of each hen at the first egg laid during the recovery period. The photoperiod was gradually restored to 16L:8D once the laying rate reached 5% and body weight recovered to 90% of the pre-molt level. Light duration was increased by 2 h/day when the weekly increase in laying rate exceeded 20% and by 1 h/day when the increase was less than 10%. Representative photographs illustrating feather changes during fasting-induced molting are shown in Fig. 1. Representative feather appearance of laying hens at different stages of fasting-induced molting. (A) Before fasting-induced molting. (B) At the end of the fasting period. (C) When the flock laying rate reached 50% during the recovery period. Birds with comparable EN63–69 w and BW69w were pre-selected. Based on cumulative egg production during the recovery period (weeks 72–80), hens exhibiting rapid recovery and high egg production were classified as the high-recovery (HR) group, whereas those showing slow recovery and low egg production were classified as the low-recovery (LR) group. Four clinically healthy hens were selected from each group and euthanized by CO₂ inhalation. The hypothalamus, pituitary, ovary, and liver were rapidly collected, snap-frozen in liquid nitrogen, and stored at −80°C until RNA extraction and sequencing. RNA extraction, library construction, and sequencing Total RNA was extracted from hypothalamus, pituitary, ovary, and liver tissues using TRIzol reagent (Invitrogen, Carlsbad, CA, USA) according to the manufacturer's instructions. RNA concentration and purity were measured using a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific, Wilmington, DE, USA). RNA integrity was assessed by 1% agarose gel electrophoresis (200 V for 10 min) using approximately 400 ng of total RNA per sample. All 32 RNA samples met the quality requirements for library construction, with concentrations >50 ng/μL, A260/A280 ratios of 1.8–2.1, A260/A230 ratios >2.0, and distinct 28S and 18S rRNA bands with an approximate intensity ratio of 2:1. For library preparation, mRNA was enriched from 5 μg of total RNA by two rounds of purification using Dynabeads Oligo (dT) (Thermo Fisher Scientific). The purified mRNA was fragmented at 94 °C for 6 min. First-strand cDNA was synthesized, followed by strand-specific second-strand cDNA synthesis in which dTTP was replaced with dUTP. The resulting cDNA fragments were subjected to end repair, A-tailing, and ligation with dual-index adapters. Fragments of the desired size were selected using AMPure XP beads (Beckman Coulter, Brea, CA, USA), followed by uracil-DNA glycosylase (UDG) treatment. The libraries were amplified by eight PCR cycles, resulting in an average insert size of 300 ± 50 bp. Paired-end sequencing (2 × 150 bp) was subsequently performed on the DNBSEQ-T7 platform (MGI Tech Co., Ltd., Shenzhen, China) according to the manufacturer's standard protocol. RNA-seq data processing and quantification Raw sequencing reads were processed using fastp (v0.23.4) to generate high-quality clean reads. Adapter sequences were removed, and reads containing more than five ambiguous nucleotides (N), with >40% of bases having a Phred quality score ≤20, or shorter than 20 bp were discarded. The resulting clean reads were aligned to the Gallus gallus reference genome (GRCg7b.115), downloaded from the Ensembl database, using HISAT2 (v2.2.1). Gene-level read counts were generated using featureCounts (v2.0.6). The resulting count matrix was used for all downstream analyses. Differential expression analysis Differential expression analysis was performed using the DESeq2 package (v1.42.1) in R based on raw gene count data. Gene expression counts were normalized, followed by variance-stabilizing transformation for downstream analyses. Principal component analysis (PCA) was performed using the transformed expression matrix of all 32 samples to evaluate global transcriptional variation, sample relationships, and clustering patterns among tissues and experimental groups. DEGs were identified using a threshold of |log₂ fold change| ≥ log₂(1.5) and a Benjamini–Hochberg adjusted P-value < 0.05. Multi-group DEG scatter plots, UpSet plots, and hierarchical clustering heatmaps were generated using the ggplot2 (v4.0.2), UpSetR (v1.4.1), and pheatmap (v1.0.13) packages in R, respectively. Functional annotation and enrichment analysis To investigate the biological functions of the identified DEGs, Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses were performed using the clusterProfiler package (v4.10.1) in R. Gene identifiers were converted from Ensembl IDs to Entrez IDs using the chicken annotation database org.Gg.eg.db (v3.18.0). GO enrichment analysis was conducted for the three ontology categories: biological process (BP), cellular component (CC), and molecular function (MF). Significantly enriched GO terms and KEGG pathways were identified using a threshold of P < 0.05. Enrichment results were visualized using the ggplot2 package, and bar plots were generated to present the functional enrichment profiles of DEGs in the four tissues. Weighted wene co-expression network analysis (WGCNA) To characterize major gene co-expression patterns across the four tissues, weighted gene co-expression network analysis (WGCNA) was performed using the WGCNA package (v1.74) in R. Prior to network construction, the 10,000 genes with the highest median absolute deviation (MAD) across all 32 samples were selected to reduce noise from low-variance genes and improve computational efficiency. An unsigned scale-free co-expression network was constructed using the blockwiseModules function. The adjacency matrix was calculated as Aij = |cor(xi, xj)|β, where cor(xi, xj) represents the Pearson correlation coefficient between genes i and j. The soft-thresholding power (β) was determined using the pickSoftThreshold function to achieve an approximate scale-free topology (R² ≥ 0.80) under an unsigned network topology criterion. The adjacency matrix was subsequently transformed into an unsigned topological overlap matrix (TOM) to measure network connectivity. Gene modules were identified by average-linkage hierarchical clustering based on TOM dissimilarity (1 − TOM), with a deepSplit value of 2 and a minimum module size of 100 genes. Modules with high similarity were merged using a module eigengene (ME) dissimilarity threshold of 0.25, corresponding to a correlation coefficient ≥ 0.75. Module–trait relationships were first evaluated to characterize tissue-associated co-expression patterns. Sample traits were defined as eight tissue–group combinations generated by combining four tissues [hypothalamus (H), pituitary (P), ovary (O), and liver (L)] with two recovery groups [high-recovery (HR) and low-recovery (LR)]. Correlations between module eigengenes and these sample traits were calculated, and statistical significance was assessed using Student asymptotic P-values implemented in the corPvalueStudent function. Modules showing the strongest associations with the corresponding tissues and consistent directions of association in the high-recovery (HR) and low-recovery (LR) groups were defined as tissue-associated modules. To further assess the association between co-expression modules and post-molt egg production recovery, module eigengene (ME) values for each WGCNA module were compared between the HR and LR groups within each tissue using Welch’s two-sample t-test. P-values were adjusted for multiple testing using the Benjamini–Hochberg procedure to control the false discovery rate (FDR), with FDR < 0.05 considered statistically significant. Differences in ME values between the HR and LR groups within each tissue were visualized using the ggplot2 package (v4.0.2) in R. Integration of transcriptomic data and Protein-Protein Interaction (PPI) network analysis To identify candidate genes associated with post-molt egg production recovery, DEGs between the HR and LR groups within each tissue were intersected with genes from the corresponding tissue-associated WGCNA modules. Protein–protein interaction (PPI) networks were constructed for the overlapping genes using the Multiple Proteins module in the STRING database (v12.0) for Gallus gallus, with the network type set to full STRING network and a minimum required interaction score of 0.400 (medium confidence). For tissues containing fewer than 20 input genes, up to 20 shell proteins were added to improve network connectivity. The resulting PPI networks were visualized using Cytoscape (v3.10.4). Network topology was analyzed using the cytoHubba plugin. Four topological algorithms, including Maximal Clique Centrality (MCC), Degree, Closeness, and Betweenness, were applied independently to rank genes within each tissue-specific network. Genes consistently ranked among the top 10 by all four algorithms were defined as hub genes. Statistical analysis The pre-molt traits (EN63–69 w and BW69w) and post-molt recovery traits (AFE, EN72–75 w, EN76–80 w, and EN72–80 w) were organized using Microsoft Excel 2021 and analyzed with IBM SPSS Statistics (v19.0). Data normality and homogeneity of variance were evaluated using the Shapiro–Wilk and Levene's tests, respectively. Differences between the HR and LR groups were assessed using independent-samples Student's t-tests for normally distributed data with homogeneous variances, whereas the Mann–Whitney U test was used for non-normally distributed data. Results are presented as mean ± SEM. Statistical significance was defined as ns (P > 0.05) or ** (P < 0.01). Fig.s and statistical annotations were generated using GraphPad Prism (v10.1.2). Results Divergence in phenotypic traits between HR and LR groups To evaluate differences in reproductive recovery following fasting-induced molting, egg production performance and body weight were compared between the high-recovery (HR) and low-recovery (LR) groups (Table 1; Fig. 2). Before molting, the two groups showed comparable baseline characteristics. No significant differences were observed in cumulative egg production during weeks 63–69 (Fig. 2A, P > 0.05) or body weight at 69 weeks of age (Fig. 2B, P > 0.05), indicating similar pre-molt production performance and physiological status. Table 1Production performance of the HR and LR groups before induced molting and during the post-molt recovery period.ItemsHRLRP-valueEN63-69w47.50±0.5348.63±0.380.11BW69w1786.30±14.841790.70±19.720.86AFE14.13±0.35B44.38±1.50A< 0.01EN72-75w14.75±0.37A0.00±0.00B< 0.01EN76-80w33.88±0.40A18.50±1.30B< 0.01EN72-80w48.63±0.26A18.50±1.30B< 0.01Note:Pre-molt traits include EN63-69 w (egg number during weeks 63-69) and BW69w (body weight at 69 weeks of age). Post-molt recovery traits include AFE (age at first egg), EN72-75 w (egg number during weeks 72-75), EN76-80 w (egg number during weeks 76-80), and EN72-80 w (egg number during weeks 72-80). Values are presented as mean ± SEM. Within each row, means with different uppercase superscript letters differ significantly (P < 0.01), whereas means without superscripts do not differ significantly (P > 0.05). Production performance of the HR and LR groups before induced molting and during the post-molt recovery period. Note:Pre-molt traits include EN63-69 w (egg number during weeks 63-69) and BW69w (body weight at 69 weeks of age). Post-molt recovery traits include AFE (age at first egg), EN72-75 w (egg number during weeks 72-75), EN76-80 w (egg number during weeks 76-80), and EN72-80 w (egg number during weeks 72-80). Values are presented as mean ± SEM. Within each row, means with different uppercase superscript letters differ significantly (P < 0.01), whereas means without superscripts do not differ significantly (P > 0.05). Comparison of production performance between HR and LR groups during pre-molt and post-molt recovery periods. (A) Egg number at 63-69 weeks of age. (B) Body weight at 69 weeks of age. (C) Age at first egg. (D) Egg number at 72-75 weeks of age. (E) Egg number at 76-80 weeks of age. (F) Egg number at 72-80 weeks of age. Data are presented as mean ± SEM. ns indicates no significant difference, and ** indicates significant difference (P < 0.01). Following molting, the two groups exhibited marked differences in reproductive recovery. The HR group resumed laying significantly earlier than the LR group, with the interval to first egg being 14.13 ± 0.35 d and 44.38 ± 1.50 d, respectively (Fig. 2C, P < 0.01). During weeks 72–75, hens in the HR group produced 14.75 ± 0.37 eggs on average, whereas no egg production was recorded in the LR group (Fig. 2D, P < 0.01). The HR group also maintained significantly higher egg production during weeks 76–80 (Fig. 2E, P < 0.01). Consequently, cumulative egg production during weeks 72–80 was significantly greater in the HR group than in the LR group (Fig. 2F, P < 0.01). Transcriptomic landscapes and identification of DEGs RNA-seq analysis of 32 hypothalamic, pituitary, ovarian, and liver samples generated a total of 1.55 billion raw reads. After quality filtering, each library yielded 5.87–9.02 Gb of clean data (Supplementary Table 1). The Q30 values ranged from 95.42% to 97.75%, and GC content ranged from 45.06% to 46.28%. The proportion of clean reads mapped to the Gallus gallus reference genome (GRCg7b.115) ranged from 95.51% to 97.66%, with an average mapping rate of 96.23%. These results indicated that the sequencing data were of high quality and suitable for subsequent analyses. Principal component analysis (PCA) showed that samples clustered primarily according to tissue type, with PC1 and PC2 explaining 28.74% and 21.46% of the total variance, respectively (Fig. 3A). Differential expression analysis identified 3,425 DEGs in the ovary, including 2,554 upregulated and 871 downregulated genes in the HR group (Supplementary Table 2). In comparison, 435 DEGs were identified in the pituitary (245 upregulated and 190 downregulated; Supplementary Table 3), 289 in the liver (142 upregulated and 147 downregulated; Supplementary Table 4), and 271 in the hypothalamus (84 upregulated and 187 downregulated; Supplementary Table 5) (Fig. 3B). Transcriptomic comparison between HR and LR groups across four tissues. (A) PCA of 32 samples from hypothalamus (H), pituitary (P), ovary (O), and liver (L). Circles and triangles indicate HR and LR samples, respectively. (B) DEGs identified in pairwise comparisons (HR-H vs LR-H, HR-P vs LR-P, HR-O vs LR-O, and HR-L vs LR-L). Orange and green dots represent up- and downregulated genes, respectively. (C) UpSet plot showing shared and tissue-specific DEGs across the four tissues; pink bars indicate set sizes and blue bars indicate intersection sizes. (D-G) Heatmaps of DEGs in hypothalamus (D), pituitary (E), ovary (F), and liver (G). Expression values are normalized and scaled from low (blue) to high (red). UpSet analysis showed that most DEGs were tissue-specific (Fig. 3C). The ovary contained the largest number of tissue-specific DEGs (3,181), followed by the pituitary (276), liver (216), and hypothalamus (186). Only 10 DEGs were shared across the hypothalamus, pituitary, and ovary, including HSPB8, KIF20A, COL27A1, ERCC6L, PIK3R6, KCNG2, ABCA9, A2ML3, LOC101751097, and ENSGALG00010002671. Hierarchical clustering based on tissue-specific DEGs clearly separated the HR and LR samples in all four tissues (Fig. 3D–G). Functional enrichment analysis of the identified DEGs DEGs identified in the hypothalamus, pituitary, ovary, and liver were subjected to GO and KEGG enrichment analyses using a significance threshold of P < 0.05. Across the four tissues, a total of 539 significantly enriched GO terms (Supplementary Table 6) and 55 significantly enriched KEGG pathways (Supplementary Table 7) were identified. The ovary showed the largest number of enriched GO terms (265), whereas the liver had the fewest (59). Similarly, the ovary contained the largest number of enriched KEGG pathways (25), whereas only 4 pathways were identified in the hypothalamus. The top 10 enriched GO terms (BP, CC, and MF) and KEGG pathways for each tissue are shown in Fig. 3. In the hypothalamus, DEGs were enriched in 101 GO terms, including 74 BP, 12 CC, and 26 MF terms (Fig. 4A). KEGG analysis identified four enriched pathways: protein processing in endoplasmic reticulum, nitrogen metabolism, apelin signaling pathway, and glutathione metabolism (Fig. 4E). In the pituitary, DEGs were enriched in 103 GO terms, including 77 BP, 10 CC, and 16 MF terms (Fig. 4B). KEGG analysis identified 12 enriched pathways, with the top 10 including neuroactive ligand-receptor interaction, hormone signaling, cytoskeleton in muscle cells, intestinal immune network for IgA production, ECM-receptor interaction, integrin signaling, calcium signaling pathway, arachidonic acid metabolism, pyrimidine metabolism, and metabolism of xenobiotics by cytochrome P450 (Fig. 4E). In the ovary, DEGs were enriched in 265 GO terms, including 204 BP, 29 CC, and 32 MF terms (Fig. 4C). KEGG analysis identified 25 enriched pathways, with the top 10 including cell adhesion molecule (CAM) interaction, efferocytosis, integrin signaling, cytokine-cytokine receptor interaction, ECM-receptor interaction, intestinal immune network for IgA production, calcium signaling pathway, apoptosis, cytoskeleton in muscle cells, and herpes simplex virus 1 infection (Fig. 4E). In the liver, DEGs were enriched in 59 GO terms, including 51 BP, 3 CC, and 5 MF terms (Fig. 4D). KEGG analysis identified 14 enriched pathways, with the top 10 including metabolic pathways, glycosphingolipid biosynthesis-ganglio series, insulin signaling pathway, biosynthesis of cofactors, phosphatidylinositol signaling system, glycerolipid metabolism, FoxO signaling pathway, intestinal immune network for IgA production, adipocytokine signaling pathway, and PPAR signaling pathway (Fig. 4E). The numbers and composition of enriched GO terms and KEGG pathways varied among the four tissues. Functional enrichment analysis of DEGs between HR and LR groups across tissues. (A-D) GO enrichment analysis of DEGs in hypothalamus (A), pituitary (B), ovary (C), and liver (D). The x-axis represents −log₁₀(P-value) of enriched GO terms. Bars represent Biological Process (BP), Cellular Component (CC), and Molecular Function (MF). (E) KEGG pathway enrichment analysis of DEGs in the four tissue comparisons. The y-axis represents −log₁₀(P-value) of enriched pathways. Colors represent different comparisons: hypothalamus (HR-H vs LR-H, red), pituitary (HR-P vs LR-P, blue), ovary (HR-O vs LR-O, orange), and liver (HR-L vs LR-L, green). Multi-tissue WGCNA identifies tissue-associated co-expression modules To characterize gene co-expression patterns across tissues, WGCNA was performed using the 10,000 genes with the highest median absolute deviation (MAD) across 32 hypothalamic, pituitary, ovarian, and liver samples from the HR and LR groups. The scale-free topology fit index (R²) exceeded 0.80 when the soft-thresholding power (β) reached 6 and approached 0.90 at β= 9 (Fig. 5A). At β= 9, the mean connectivity decreased to below 1,000 (Fig. 5B). Therefore, a soft-thresholding power of 9 was used for subsequent network construction. Based on the dynamic tree-cutting algorithm, 9,890 of the 10,000 genes were assigned to six co-expression modules, whereas the remaining 110 genes were assigned to the grey module (Fig. 5C). The turquoise, blue, brown, yellow, green, and red modules contained 2,737, 2,648, 1,971, 1,473, 543, and 518 genes, respectively (Supplementary Table 8). WGCNA of genes in HR and LR groups across tissues. (A) Scale-free topology fit index for different soft-thresholding powers. (B) Mean connectivity under different soft-thresholding powers. (C) Gene dendrogram with module color assignments. (D) Heatmap of correlations between module eigengenes and tissue-group combinations (HR-H, LR-H, HR-P, LR-P, HR-O, LR-O, HR-L, LR-L). The color scale indicates Pearson correlation coefficients. (E) Within-tissue comparisons of module eigengene values between the HR and LR groups for MEgreen in the hypothalamus, MEred in the pituitary, MEyellow in the ovary, and MEbrown in the liver. *FDR < 0.05; ns, not significant. (F) Gene network heatmap based on the TOM, where darker colors indicate higher connectivity. Correlation analysis between module eigengenes and tissue-group combinations identified significant associations between specific modules and tissues (Fig. 5D). The MEgreen module showed the strongest positive correlation with hypothalamic samples from both the HR and LR groups (HR_H: r = 0.57, P = 7 × 10⁻⁴; LR_H: r = 0.56, P = 8 × 10⁻⁴). The MEred module was positively correlated with pituitary samples from both groups (HR_P: r = 0.52, P = 0.002; LR_P: r = 0.51, P = 0.003). Similarly, the MEyellow module showed significant positive correlations with ovarian samples (HR_O: r = 0.52, P = 0.003; LR_O: r = 0.53, P = 0.002). In contrast, the MEbrown module was negatively correlated with liver samples (HR_L: r = −0.45, P = 0.010; LR_L: r = −0.46, P = 0.008). The similar correlations observed for the HR and LR groups within the same tissue indicated that these module–trait relationships mainly reflected tissue-associated co-expression patterns. To further assess the association between co-expression modules and post-molt egg production recovery, module eigengene (ME) values were compared between the HR and LR groups within each tissue (Fig. 5E; Supplementary Table 9). After Benjamini–Hochberg correction, nine tissue–module combinations showed significant differences between the HR and LR groups (FDR < 0.05). Among the tissue-associated modules identified above, MEgreen values in the hypothalamus were significantly higher in the HR group than in the LR group (P = 0.0108, FDR = 0.0348), whereas MEyellow values in the ovary were significantly lower in the HR group than in the LR group (P = 0.0013, FDR = 0.0150). In contrast, MEred in the pituitary (P = 0.2188, FDR = 0.2764) and MEbrown in the liver (P = 0.1817, FDR = 0.2565) did not differ significantly between the two groups. The topological overlap matrix (TOM) heatmap showed clear separation among the identified modules, with high topological overlap within individual modules (Fig. 5F). Based on the tissue associations described above, the MEgreen, MEred, MEyellow, and MEbrown modules were selected as the corresponding hypothalamus-, pituitary-, ovary-, and liver-associated modules, respectively, for subsequent integration with within-tissue DEGs. Identification of candidate hub genes through integrated DEG, WGCNA, and PPI network analyses Candidate genes associated with post-molt recovery were identified by intersecting DEGs between the HR and LR groups within each tissue with genes from the corresponding tissue-associated WGCNA modules (MEgreen for hypothalamus, MEred for pituitary, MEyellow for ovary, and MEbrown for liver). Venn diagram analysis identified 15, 12, 197, and 22 overlapping genes in the hypothalamus (Fig. 6A), pituitary (Fig. 6B), ovary (Fig. 6C), and liver (Fig. 6D), respectively (Supplementary Table 10). The overlapping genes were used to construct protein-protein interaction (PPI) networks using the STRING database and were subsequently visualized in Cytoscape. Hub genes were identified using the cytoHubba plugin based on four topological algorithms: Maximal Clique Centrality (MCC), Degree, Closeness, and Betweenness. The top 10 ranked genes from each algorithm are provided in Supplementary Table 11. Based on the integrated rankings across the four algorithms, HSP90AA1, TULP1, and C3orf14 were identified as hub genes in the hypothalamus (Fig. 6E); GDF7, FZD10, and TRPA1 in the pituitary (Fig. 6F); CTSS, TLR7, and RPL12 in the ovary (Fig. 6G); and STS, PNPLA2, and PPARGC1A in the liver (Fig. 6H). Integration of DEGs, WGCNA, and PPI networks to identify key genes associated with differences in egg production performance between HR and LR groups. (A-D) Venn diagrams showing overlapping genes between DEGs and WGCNA-identified genes in hypothalamus (A), pituitary (B), ovary (C), and liver (D). (E-H) PPI networks of key genes in hypothalamus (E), pituitary (F), ovary (G), and liver (H). Nodes represent genes (protein-coding genes). Blue nodes indicate hub genes, and edges represent known or predicted protein-protein interactions. Discussion Fasting-induced molting is widely used in commercial laying hen production because it promotes reproductive tract remodeling, extends the laying cycle, and improves post-molt egg production performance (Gickel et al., 2026). However, considerable individual variation in post-molt reproductive recovery remains, even under the same genetic background and management conditions. In the present study, the high-recovery (HR) and low-recovery (LR) groups of Hy-Line Gray laying hens showed comparable body weight and egg production performance before molting, indicating similar physiological and production status at baseline. During the recovery phase, however, the two groups exhibited markedly different reproductive performance. The HR group resumed laying significantly earlier than the LR group and maintained a higher cumulative egg production throughout the recovery period. These findings suggest that differences in post-molt reproductive recovery are unlikely to result from pre-molt production performance, but are more likely associated with molecular regulatory mechanisms involved in the re-establishment of reproductive function. Elucidating these mechanisms will improve our understanding of post-molt reproductive recovery and may facilitate future genetic improvement of laying hens. Differential expression analysis revealed a progressive increase in the number of DEGs along the hypothalamic-pituitary-ovarian (HPO) axis, with the ovary exhibiting the largest transcriptional response. Specifically, ovarian DEGs accounted for 82.91% (3,425/4,132) of all DEGs identified across the HPO axis, indicating substantially greater transcriptional reprogramming in the ovary than in the hypothalamus or pituitary during post-molt recovery. Similar findings have been reported in previous poultry transcriptomic studies, in which the ovary consistently exhibited more extensive transcriptional changes than the hypothalamus and pituitary (Fan et al., 2025; Wang et al., 2025b). Together, these observations suggest that the ovary represents a major site of transcriptional regulation during post-molt reproductive recovery. In addition, 10 DEGs were shared across the HPO axis, including HSPB8, KIF20A, COL27A1, ERCC6L, PIK3R6, KCNG2, ABCA9, A2ML3, LOC101751097, and ENSGALG00010002671. Among these genes, HSPB8 encodes a small heat shock protein involved in protein homeostasis and cellular stress protection, which may facilitate adaptation to the metabolic and endocrine changes associated with molting (Basaki et al., 2022). KIF20A plays an important role in mitosis, cytokinesis, and microtubule organization, suggesting a potential role in ovarian tissue remodeling through the regulation of cell proliferation (Zhou et al., 2026). PIK3R6 encodes a regulatory subunit of phosphoinositide 3-kinase and participates in the PI3K/AKT signaling pathway, which regulates cell proliferation, survival, and metabolism. Given its involvement in energy metabolism, PIK3R6 may contribute to follicular development during post-molt recovery (Shao and Xu, 2025). ABCA9 is an endoplasmic reticulum-associated cholesterol transporter involved in lipid homeostasis and cholesterol transport (Wang et al., 2025a). Its role in lipid metabolism suggests that it may contribute to follicular development by regulating cholesterol utilization in ovarian cells (Hwang et al., 2023). To further elucidate the biological functions of the identified DEGs, GO and KEGG enrichment analyses were performed. In the hypothalamus, DEGs were significantly enriched in GO terms related to protein homeostasis, including protein folding, chaperone-mediated protein folding, and heat shock protein-dependent protein folding. These biological processes are essential for maintaining protein quality control and cellular homeostasis under physiological and stress conditions, thereby supporting normal hypothalamic neuroendocrine activity (Kim et al., 2013). Consistently, the protein processing in endoplasmic reticulum pathway was significantly enriched in the KEGG analysis, further suggesting that maintenance of protein homeostasis may be associated with recovery of hypothalamic function following induced molting. In the pituitary, DEGs were mainly enriched in GO terms associated with hormone responsiveness and circadian regulation, including response to hormone, cellular response to hormone stimulus, and circadian rhythm. Circadian regulation has been reported to influence reproductive activity and laying performance in poultry through modulation of endocrine function (Xu et al., 2023). Therefore, the observed enrichment suggests that differences in post-molt recovery between the HR and LR groups may be associated with hormone responsiveness and circadian regulation in the pituitary. Consistently, KEGG analysis identified enrichment of multiple hormone-related signaling pathways. In the ovary, DEGs were predominantly enriched in GO terms associated with ribosomal structure and function, including cytosolic ribosome, ribosome, ribosomal subunit, and structural constituent of ribosome. Ribosomes constitute the core machinery for protein synthesis and are closely associated with cellular biosynthetic activity (Nakanishi et al., 2023). Because follicular growth and steroid hormone synthesis require continuous protein production, these findings suggest that translational activity may contribute to ovarian functional recovery after induced molting. Consistent with the GO results, the Ribosome pathway was also significantly enriched in the KEGG analysis. In the liver, DEGs were significantly enriched in GO terms associated with immune regulation, including regulation of immune system process, positive regulation of leukocyte activation, positive regulation of leukocyte differentiation, regulation of leukocyte activation, positive regulation of cell activation, and regulation of hemopoiesis. As the central organ coordinating metabolism and immunity (Jenne and Kubes, 2013), the liver may contribute to post-molt recovery by maintaining systemic immune homeostasis. KEGG analysis further identified enrichment of the intestinal immune network for IgA production pathway, which was also enriched in the pituitary and ovary. Given the established roles of IgA-mediated mucosal immunity and the gut-liver axis in maintaining immune homeostasis (Lin et al., 2014; Pan et al., 2026), this pathway may participate in systemic immune regulation during post-molt recovery. WGCNA further revealed that the major co-expression modules were closely associated with tissue identity, consistent with the clear tissue-specific separation observed in the PCA. This pattern is expected in multi-tissue transcriptomic datasets, as differences among tissues represent a major source of transcriptomic variation. Notably, within-tissue comparisons of module eigengene (ME) values revealed recovery-associated differences in a subset of modules. After FDR correction, MEgreen showed significantly higher ME values in the HR group in the hypothalamus, whereas MEyellow showed significantly lower ME values in the HR group in the ovary. These findings suggest that, beyond the dominant tissue-associated co-expression patterns, specific modules may exhibit coordinated transcriptional changes associated with post-molt egg production recovery. In contrast, MEred in the pituitary and MEbrown in the liver did not differ significantly between the HR and LR groups, indicating that these tissue-associated modules did not show detectable overall expression differences between recovery groups. To further identify candidate regulators associated with post-molt egg production recovery, within-tissue DEGs between the HR and LR groups were integrated with genes from the corresponding tissue-associated WGCNA modules, and the overlapping genes were subsequently subjected to protein–protein interaction (PPI) network analysis. Network topology analysis identified several tissue-specific hub genes. In the hypothalamus, HSP90AA1 was identified as a hub gene and was enriched in multiple GO terms and KEGG pathways related to protein homeostasis. HSP90AA1 encodes a molecular chaperone that plays a central role in protein folding and stabilization under both normal and stress conditions (Yang et al., 2025). Considering its known biological function and enrichment profile, HSP90AA1 may contribute to post-molt recovery by maintaining hypothalamic protein homeostasis and neuroendocrine function. In the pituitary, GDF7 was identified as a hub gene. As a member of the transforming growth factor-β (TGF-β) superfamily, GDF7 is involved in cell differentiation and tissue development (Mott et al., 2024). Previous studies have shown that increased GDF7 expression suppresses neck feather development during chicken embryogenesis (Mou et al., 2011). Because induced molting involves extensive feather regeneration and tissue remodeling, GDF7 may participate in developmental processes during the recovery phase and thereby influence pituitary function associated with reproductive recovery. In the ovary, CTSS was identified as a hub gene. CTSS encodes cathepsin S, a lysosomal protease involved in extracellular matrix remodeling, apoptosis, and tissue remodeling (Bae et al., 2014). These biological functions suggest that CTSS may participate in ovarian tissue reconstruction during post-molt recovery. In the liver, STS was identified as a hub gene. STS encodes steroid sulfatase, an enzyme that catalyzes the hydrolysis of sulfated steroid precursors into biologically active steroid hormones (Purohit et al., 2011). Steroid hormones play essential roles in hepatic synthesis of yolk precursors, including very low-density lipoprotein (VLDL) and vitellogenin (VTG) (Wang et al., 2013). Therefore, STS may contribute to post-molt recovery by regulating steroid hormone metabolism and supporting follicular development. From a systems-level perspective, the hypothalamus, pituitary, ovary, and liver exhibited distinct tissue-specific transcriptional profiles associated with different biological processes related to post-molt egg production recovery. Specifically, the hypothalamus was primarily associated with protein homeostasis, the pituitary with hormone responsiveness and circadian regulation, the ovary with protein synthesis and tissue remodeling, and the liver with metabolic and immune regulation. Despite these tissue-specific functional features, the biological processes identified across tissues collectively involved multiple physiological systems, including neuroendocrine regulation, reproductive function, and metabolic homeostasis. Taken together, these findings suggest that post-molt egg production recovery may not be regulated by a single tissue alone but rather be associated with coordinated changes across the neuroendocrine, reproductive, and metabolic systems. Conclusion In summary, using multi-tissue transcriptomic data from the hypothalamus, pituitary, ovary, and liver, this study systematically characterized the molecular regulatory differences associated with post-molt recovery of laying performance between high- and low-recovery laying hens following fasting-induced molting. Tissue-specific transcriptional responses were observed across the four tissues, with the ovary exhibiting the greatest extent of transcriptional reprogramming during the recovery phase. Functional enrichment analysis revealed distinct biological characteristics among tissues, with the hypothalamus mainly associated with protein homeostasis, the pituitary with hormone response and circadian rhythm regulation, the ovary with ribosome-mediated protein synthesis, and the liver with immune homeostasis and metabolic regulation. Furthermore, integration of differential expression analysis, WGCNA, and protein-protein interaction network analysis identified several tissue-specific hub genes, including HSP90AA1, GDF7, CTSS, and STS. Collectively, the multi-tissue transcriptomic analysis indicates that post-molt egg production recovery is associated with complementary molecular responses across neuroendocrine, reproductive, and metabolic systems. These findings improve our understanding of the molecular mechanisms underlying post-molt reproductive recovery in laying hens and provide candidate genes for future studies on laying recovery capacity and genetic improvement of laying performance. Authorship contribution statement Xiaomeng Miao: Conceived and designed the study, developed the experimental methodology, conducted the experiment, collected and organized the data, performed the statistical and bioinformatic analyses, prepared the figures, and drafted the manuscript. Jia Liu: Contributed to the experimental methodology, conducted the animal experiment, and collected and organized samples and data. Fugui Li: Conducted the animal experiment and collected and organized the data. Xing Chen: Conducted the animal experiment and contributed to the validation experiments. Yimin Wei: Conducted the animal experiment and collected and organized the data. Li Zhang: Contributed to the experimental methodology and validation and supervised the study. Yulong Feng: Conducted the animal experiment, contributed to the validation experiments, and collected and organized the data. Zhonghua Ning: Conceived and supervised the study, managed the research project, obtained funding, and reviewed and revised the manuscript. All authors critically reviewed the manuscript, approved the final version for publication, and agreed to be accountable for their contributions and for the accuracy and integrity of the work. Data availability The RNA-seq datasets generated during this study have been deposited in the NCBI Sequence Read Archive (SRA) database under BioProject accession number PRJNA1482247. 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. Acknowledgements This work was supported by the Guizhou Provincial Key Technology Research and Development Program (Grant No. [2023]004), the National Key Research and Development Program of China (Grant No. 2024YFD200030403), and the China Agricultural University–Jiajiang County People's Government Professor Workstation for Poultry Breeding and Genetic Resources (Grant No. 202105510410408). Footnotes Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.psj.2026.107779. Supplementary materials References Associated Data Supplementary Materials Data Availability Statement The RNA-seq datasets generated during this study have been deposited in the NCBI Sequence Read Archive (SRA) database under BioProject accession number PRJNA1482247.