Integrative transcriptomic analysis reveals tissue associated molecular responses to heat stress in chickens Authors: Ali Hassan Nawaz, Munwar Ali, Shaokat Ali, Jiqiang Ding, Ying Duan, Muhammad Sayyam Akram, Nasir Mukhtar, Ting Xu, Xiaoli Shi, Thobela Louis Tyasi, Diyan Li, Yiqiang Zhao, Chungang Feng Published: 2026 ABSTRACT AbstractClimate change is intensifying extreme heat events that compromise poultry health, welfare, and productivity. Although RNA-seq has been widely used to investigate heat stress responses, individual studies are constrained by small sample sizes, experimental heterogeneity, and discordant differential expression signatures, limiting the identification of conserved molecular responses across tissues. We conducted an integrative analysis of 11 independent RNA-seq datasets comprising 212 samples (110 liver, 62 breast muscle, and 40 heart) from broiler chickens exposed to heat stress. Differential expression analysis, weighted gene co-expression network analysis (WGCNA), and multi-criteria consensus filtering were combined to identify cross-validated hub genes. We identified conserved hepatic, and breast muscle transcriptional responses that showed distinct tissue associated division of labor model. Liver exhibited metabolic reprogramming involving altered gluconeogenic and fatty acid metabolic pathways, including PCK1 and ACACB, together with changes in insulin-PPAR signaling. Breast muscle showed activation of p53 mediated cell cycle arrest and proteostatic responses involving CDKN1A and DNAJA4. Heart exhibited transcriptional changes in adrenergic-calcium signaling, including CACNA1H and PPP3CA, together with suppression of extracellular matrix remodeling. Transcription factor (TF) activity inference using decoupleR and orthology-mapped DoRothEA regulons suggested tissue associated regulatory patterns involving HSF2, NCOA3, and MEF2A/MEF2C in liver; FOXO4, SOX9, and SP-family TFs in breast muscle; and reduced MEF2A/MEF2C, TEAD1, and JUN activity in heart. This integrative framework identifies high confidence candidate genes and regulatory networks validated by multiple analytical approaches, providing molecular candidates for subsequent genomic and functional studies of climate resilience in poultry.Keywords: Chicken, RNA-seq, Heat stress, WGCNA, Climate change Introduction Anthropogenic climate change is intensifying extreme thermal events beyond historical baselines, with projections indicating further escalation through the 21st century (Murali et al., 2023). As the global population approaches 9.7 billion by 2050, poultry, which constitutes the largest share of global meat production and provides the highest feed conversion efficiency among livestock faces acute thermoregulatory limitations when ambient temperatures exceed the upper critical limit of the thermoneutral zone (Babinszky et al., 2011; Oke et al., 2025). Heat stress consequently induces coordinated physiological responses across multiple systems, compromising poultry health, welfare, and productivity. These responses include oxidative stress, acid-base imbalance, and impaired immune competence, which can lead to reduced feed efficiency, decreased body weight gain, and increased mortality (Wasti et al., 2020; Eletu et al., 2026; Idowu et al., 2026). With poultry producing regions projected to face escalating extreme heat exposure by mid-century (Murali et al., 2023), deciphering the molecular architecture of heat resilience has become an urgent priority (Hosseinzadeh and Hasanpur, 2023). Conventional differential expression analysis identifies genes exhibiting large expression changes between heat-stressed and control conditions but does not directly capture coordinated gene networks or relationships among co-expressed genes. Combining differential expression analysis and weighted gene co-expression network analysis (WGCNA) makes it possible to identify statistically significant and network supported genes, increasing confidence in their biological relevance (Liu et al., 2024a; Ding et al., 2025). Cross-study integration can identify transcriptional changes recurring across heterogeneous experimental conditions without directly pooling biologically distinct datasets (Keel and Lindholm-Perry, 2022). However, differential expression and co-expression networks alone cannot resolve the upstream transcriptional regulators that orchestrate tissue associated responses. Transcription factor (TF) activity inference from transcriptomic data using regulon based approaches such as decoupleR enables the reconstruction of regulatory programs driving observed expression shifts, providing a mechanistic layer beyond gene level changes (Garcia-Alonso et al., 2019; Badia-i-Mompel et al., 2022). Integrating these insights, we asked two questions. First, which conserved tissue-associated transcriptional programs characterize the response to heat stress in chickens? Second, which inferred upstream transcription factors are associated with these regulatory programs? This study presents an integrative analysis of 11 independent RNA-seq datasets comprising 212 samples (110 liver, 62 breast muscle, and 40 heart) to construct a consensus molecular map of heat stress response in broiler chickens. By combining differential expression analysis with WGCNA, we identified consensus transcriptional signatures that persist across heterogeneous experimental conditions, genetic backgrounds, and heat stress protocols. We further infer upstream transcription factor activities using decoupleR with the DoRothEA regulon database and evaluate these inferred activities through gene set enrichment analysis, revealing tissue associated regulatory divergence. Our analysis reveals a division of labor model in which liver suppresses gluconeogenesis (PCK1, ACACB), muscle activates p53-mediated proteostasis (CDKN1A, DNAJA4), and heart shows transcriptional changes in adrenergic-calcium signaling (CACNA1H, PPP3CA). Integrating differential expression, WGCNA, and directional consistency, we identify consensus hub genes that provide robust candidates for subsequent genomic and functional studies of climate resilience in chickens. Materials and methods Data acquisition RNA-sequencing data of heat stress studies on Gallus gallus (chicken) were retrieved systematically over NCBI Sequence Read Archive (SRA) database. The selection of studies was done according to the following criteria: (1) both the heat stress and the control condition were performed in a biological replicate, (2) the tissue samples were liver, heart, or breast muscle, (3) the experimental metadata was clearly documented. Eleven independent studies were found in the three tissues, which included 110 liver samples, 62 breast muscle samples, and 40 heart samples (Table S1). Validation and curation of experimental metadata through SRA accession numbers, condition of treatment, depth of sequencing and groupings of samples manually verified and corrected errors in assigning samples to heat stress (HS) and thermoneutral control (CN). Studies whose metadata was not clear or whose tissues were of mixed type was eliminated to ensure analytical rigor. Quality control and read processing Raw FASTQ files were assessed using FastQC v0.11.9. Quality control metrics included per base sequence quality, per-sequence quality scores, GC content distribution, sequence duplication, overrepresented sequences and adapter content. Samples were retained for downstream analysis after confirming acceptable overall read quality. Genome alignment and gene quantification Quality assessed reads were aligned to the Gallus gallus GRCg7b reference genome using STAR v2.7.10a in two pass mode to improve splice junction detection. The corresponding GRCg7b GTF annotation file was used during genome index construction and read alignment. STAR was run with the principal parameters –runThreadN 8, –outSAMtype BAM SortedByCoordinate, and –quantMode GeneCounts. Coordinate sorted BAM files were processed using Picard Tools v2.27.4, including duplicate marking with the MarkDuplicates function, and were indexed using SAMtools v1.15. Gene level raw counts were obtained from the STAR ReadsPerGene.out.tab output and assembled into separate study specific count matrices using Ensembl gene identifiers. These raw count matrices were subsequently used as input for DESeq2. Picard processed BAM files were used for alignment level processing and quality assessment; duplicate marking did not modify the STAR derived count matrices used for differential expression analysis. The complete analytical workflow is presented in Supplementary Fig. 1. DEGs analysis Each study was analyzed independently to compare gene expression between heat-stressed and control samples within the respective tissue using DESeq2 v1.34.0 on R v4.2.1. Raw count matrices were screened to keep the genes that were expressed at a minimum of 10 counts in at least 20% of the samples in each study to eliminate lowly expressed transcripts. DESeq2 models were fitted separately within each study. For simple two group experiments in which the analyzed samples differed only by heat exposure, the design formula was specified as condition. When an additional experimental factor varied within a study and was estimable from the available metadata, it was included in the design. For example, PRJNA1103706 included two genetic backgrounds, and breed was therefore included together with condition when estimating the heat stress effect. Variables that were constant within a study, such as sex or age where all analyzed samples shared the same value, were not included as covariates because their effects could not be independently estimated. DEGs were defined as genes with FDR < 0.05 and |log₂FC| ≥ 1 for liver, or > 0.5 for breast muscle and heart. Only genes showing significant and reproducible differential expression between heat stressed and control samples across independent studies within the same tissue were included in downstream analyses to identify consensus DEGs. As a sensitivity analysis, liver differential expression was also evaluated using the common effect size threshold applied to breast muscle and heart (FDR < 0.05 and |log₂FC| ≥ 0.5). Cross-study recurrence was recalculated using the same requirement of concordant differential expression in at least two independent studies without an opposing significant effect. Because the included studies differed in breed, age, heat exposure protocol and sample size, differential expression analysis was performed independently for each study rather than by directly pooling samples across datasets. Results were subsequently integrated using a concordance based vote-counting approach, in which a consensus DEG was required to show significant differential expression in the same direction in at least two independent studies within a tissue. Weighted gene co-expression network analysis (WGCNA) WGCNA v1.72-1 was performed independently for each study within each tissue. Genes first passed the same low count filter used for the study-wise differential expression analysis, requiring at least 10 raw counts in at least 20% of samples, and DESeq2 variance-stabilized expression values were used for network construction. The pickSoftThreshold function was used separately for each study to evaluate candidate powers based on scale-free topology fit and mean connectivity. Soft thresholding powers were selected independently for each study using pickSoftThreshold, considering scale-free topology fit and mean connectivity, with R2 = 0.80 used as a reference criterion (Supplementary Fig. 3–5). The selected powers were β = 12, 16, 12, and 12 for liver S1, S3, S4, and S5; β = 16, 16, and 10 for breast muscle S1, S2, and S3; and β = 16 for cardiac S1. Corresponding network sample sizes were n = 34, 24, 12, and 32 for liver; n = 24, 20, and 18 for breast muscle; and n = 24 for cardiac S1. The selected study specific power was then supplied to blockwiseModules, together with TOMType = "signed", minModuleSize = 30, reassignThreshold = 0, mergeCutHeight = 0.25, and deepSplit = 2. The module eigengene, defined as the first principal component of each module expression profile, was correlated with heat exposure status (HS = 1, CN = 0) using Pearson correlation. Module–trait associations were evaluated using Student asymptotic P values, and modules with nominal P < 0.05 were retained as exploratory heat stress associated modules. No multiple testing correction was applied across module trait tests; therefore, these associations are described as nominal rather than significant after multiplicity correction. Positive and negative correlations indicate the direction of association between the module eigengene and heat exposure and do not imply uniform upregulation or downregulation of all genes within a module. WGCNA was performed independently for each study within each tissue. Hub genes identification Module membership (kME) was calculated as the correlation between each gene’s expression profile and its corresponding module eigengene. Genes with high absolute module membership were classified as module hub genes. Gene significance was calculated as the correlation between gene expression and heat-stress status. Additionally, the gene significance (GS) module was employed to analyze the quantitative relationship between genes and traits. Hub genes were then further filtered to only include those with directional consistency where the sign of log2FC of DEG analysis was equal to the sign of the module-trait correlation, which provides biological consistency between differential expression and co-expression network positioning. The top hub genes with the highest connectivity in heat stress associated modules were selected to construct interaction networks and were further intersected with DEGs to identify overlapping candidates. Consensus gene identification For liver and breast muscle, network-supported candidates were identified by intersecting cross-study consensus DEGs with hub genes from heat-stress-associated modules. Candidates were required to show significant differential expression in the same direction in at least two independent studies, and directional consistency between differential expression and module–trait association. Only four cardiac genes met the cross-study consensus criterion. Because this set was insufficient for broad functional interpretation, the 419 gene union of the three cardiac DEG sets was analyzed separately for exploratory pathway and network integration. Cardiac candidates derived from this union were therefore not considered directly equivalent to the liver and breast-muscle consensus candidates. DEG hub overlap provided complementary evidence from the same transcriptomic datasets and was not treated as independent functional validation. Functional enrichment analysis Functional enrichment analysis was performed at three analytical stages: (1) consensus DEGs, (2) WGCNA hub genes from heat stress associated modules, and (3) integrated consensus genes (DEG and hub gene overlap). Gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses of DEGs and hub genes of heat stress associated modules were performed. For each gene set, Gene Ontology (GO) term enrichment and KEGG pathway enrichment were conducted using the KOBAS. Afterwards, protein-protein interaction (PPI) network for the final high confidence candidate genes were constructed using STRING v12.0. Transcription factor activity inference using decoupleR Transcription factor (TF) activities were inferred from the normalized gene expression matrix using decoupleR with the DoRothEA human regulon database. Human TF and target gene identifiers were mapped to chicken orthologues. Only interactions for which both the TF and target could be mapped to chicken genes were retained. Human DoRothEA was used because a comparably curated, confidence-graded chicken TF–target database was unavailable. DoRothEA integrates literature-curated TF target interactions from multiple sources, assigning each interaction a mode of regulation score (+1 for activation, –1 for repression). We employed the Univariate Linear Model (ULM) method, which models the expression of each target gene as a function of TF activity, yielding a signed t-statistic (activity score) and associated p-value for each TF (Huang et al., 2025). TFs with false discovery rate (FDR) < 0.05 were considered significantly differentially active. Regulon enrichment To validate that inferred TF activities reflected biologically coherent regulatory programs, we performed two-tailed Gene Set Enrichment Analysis (GSEA) on the ranked differential expression profile (Campbell et al., 2016). For each significantly differentially active TF, target genes were stratified into "activated" (+1) and "repressed" (-1) sets based on curated DoRothEA annotations (Huang et al., 2025). GSEA was executed independently on each signed target set using fgsea with 10,000 permutations. For each validated TF, we extracted four metrics: (i) DecoupleR activity score (ULM t-statistic), (ii) GSEA NES, (iii) –log10 (FDR), and (iv) regulon size (number of detectable targets in the expression matrix). Metrics were Z-scaled within each category to enable cross-TF comparison. GSEA enrichment plots were generated using the fgsea native plotEnrichment function, with line color indicating enrichment direction (red: positive NES; blue: negative NES). GSEA was used as computational corroboration of regulon-level consistency rather than independent experimental validation. Results Conserved heat stress responsive genes across independent transcriptomic studies Differential expression analysis across the independent transcriptomic studies revealed substantial variation in the magnitude and direction of transcriptional responses, reflecting differences in sample size, heat stress protocols, and breed-specific effects (Fig. 1). To identify recurrent responses across heterogeneous experiments without directly pooling samples, we applied a conservative cross-study vote counting approach (Fischer and Hoffmann, 2022), retaining only genes that were significantly differentially expressed in the same direction in at least two independent studies per tissue. Consensus DEGs analysis across independent RNA-seq studies. Volcano plots (A, C, E) show the DEGs identified in studies, with bar plots below indicating the total DEG count per study, for liver (A), breast muscle (C), and heart (E). Barplots (B, Liver; D, Breast muscle; F, Heart) display the number of upregulated (red) and downregulated (blue) DEGs within each individual study. In liver, this stringent filtering identified 67 conserved hepatic DEGs that were consistently responsive to heat stress across datasets, representing strong candidate biomarkers of the hepatic stress response (Table S2). Under the common sensitivity threshold of FDR < 0.05 and log2FC ≥ 0.5, the conserved liver set increased from 67 to 283 genes, while 64 of the original 67 genes (95.5%) retained strict cross study directional concordance. In breast muscle, 48 genes met the consensus criteria (29 were consistently upregulated and 19 downregulated) (Table S3). In contrast, cardiac tissue yielded the lowest overlap, with only four genes showing concordant differential expression across studies. Because only four cardiac genes met the cross-study consensus criteria, the union of the three cardiac DEG sets, comprising 419 unique genes, was analyzed separately for exploratory functional characterization (Table S4). Results derived from this union were not considered directly comparable in evidential strength with the liver and breast muscle consensus analyses. Functional enrichment analysis of DEGs identified across three tissues To elucidate the biological functions of the tissue-specific heat stress responses, we performed Gene Ontology (GO) and KEGG pathway enrichment analyses on the consensus DEG sets from each tissue. Functional enrichment analysis of the hepatic downregulated genes revealed a coordinated suppression of energy-demanding metabolic pathways, prominently involving pyruvate metabolism, adipocytokine signaling, and the insulin-PPAR signaling axis (Supplementary Fig. 2A). Gene Ontology analysis further confirmed significant enrichment of pyruvate metabolism, gluconeogenesis, and cellular response to heat (Supplementary Fig. 2B). On the other hand, upregulated genes were enriched in insulin signaling pathway and protein processing in endoplasmic reticulum and MAPK signaling (Supplementary Fig. 2A). Gene Ontology analysis identified that the most enriched molecular functions were glycogen binding and protein phosphatase type 1 complex, which are unique to PPP1R3B and PPP1R3C (Supplementary Fig. 2B). Additional enriched terms included cholesterol catabolic process, cellular response to glucose stimulus and calcium ion transmembrane transport, which indicates the increased biosynthetic and ion homeostatic capacity. DEGs downregulated in the breast muscle were primarily involved in mitochondrial oxidative phosphorylation and maintenance of the mitochondrial genome, consistent with reduced mitochondrial oxidative metabolism and mitochondrial maintenance. (Supplementary Fig. 2C). Gene ontology analysis further identified enrichment of response to reactive oxygen species, positive regulation of ROS metabolic process, and cellular oxidant detoxification (Supplementary Fig. 2D). In contrast, upregulated genes were enriched in stress response activation and cell cycle regulation. Chaperone-mediated proteostasis was indicated by heat shock protein binding protein refolding activities. Also, Cell cycle arrest emerged as an enriched term driven by cyclin-dependent kinase inhibitor CDKN1A. KEGG pathway analysis revealed an enrichment of the p53 signaling pathway with CDKN1A and growth arrest gene GADD45B as key nodes, alongside enrichment of cell cycle, FoxO signaling, mTOR signaling and Wnt signaling pathways (Supplementary Fig. 2C, D). Collectively, these findings suggest a coordinated activation of growth arrest, autophagy, and stress-responsive transcriptional programs, consistent with a strategy that prioritizes cellular integrity and proteostasis over proliferation under thermal stress. In heart, downregulated genes exhibited enrichment in extracellular matrix organization, with collagen fibril organization represented by multiple collagens (COL4A1, COL4A2) and associated collagen binding activity. This structural pattern was further supported by KEGG enrichment of ECM-receptor interaction, focal adhesion, and actin cytoskeleton regulation, collectively indicating coordinated suppression of contractile apparatus remodeling (Supplementary Fig. 2E, F). Upregulated genes were enriched in pathways linked to contractile performance and metabolic adaptation. Calcium signaling emerged as a prominent pathway, with six genes, including CACNA1H, NOS1, and CAMK2B, contributing to the enrichment of calcium-signaling pathways associated with excitation-contraction coupling (Supplementary Fig. 2E, F). Additionally, adrenergic responsiveness was represented by four genes (PPP1R1A, CACNA2D2, CACNB1, GNAQ) involved in cardiomyocyte stress response. Weighted gene co-expression network analysis (WGCNA) identified key hub genes in liver, breast muscle and heart To identify highly connected transcriptional hubs underlying the heat stress response, we performed weighted gene co-expression network analysis (WGCNA) independently for each study within each tissue. In liver, ten modules were nominally associated with the heat stress phenotype (Fig. 2A; Supplementary Fig. 3). Merging hub genes from these modules yielded a combined set of 220 liver associated hub genes. In breast muscle, eight nominally heat stress associated modules were identified across the three independent studies, with a merged hub gene set of 312 genes (Fig. 2B; Supplementary Fig. 4). Cardiac tissue revealed five nominally heat associated modules, yielding 233 hub genes (Fig. 2C; Supplementary Fig. 5). Functional enrichment of these merged hub gene sets showed convergence with the tissue specific pathways identified through DEG analysis, providing complementary network level support for the identified biological response programs. Heat stress associated modules correlated to heat stress identified through weighted gene co-expression network analysis and then functional enrichment of hub genes. Left panels display module–trait correlations for liver (A), breast muscle (B), and heart (C), with module identifiers and correlation coefficients shown. Red and blue indicate positive and negative module eigengene correlations with heat-stress status, respectively. Middle and right panels show KEGG pathway enrichment analysis of hub genes from all heat stress associated modules. Overlapping genes among DEGs and module hub genes To prioritize genes supported by both differential expression and co-expression network analyses, DEG sets were intersected with genes belonging to heat stress associated WGCNA modules (Fig. 3). Module membership (kME) and gene significance (GS) were retained as quantitative measures of network connectivity and association with heat stress status, respectively. These overlaps provided complementary evidence from the same expression datasets and were not considered independent validation. Expression changes pattern and module membership of high confidence genes identified in both DEGs and module hub genes in liver breast muscle and heart. Scatter plots of module membership (kME) versus gene significance (left) and corresponding expression fold-changes (right) identify candidate hub genes driving tissue associated heat stress responses in liver (A), breast muscle (B), and heart (C). In liver, the DEG–module overlap included metabolic regulators such as PCK1, ACACB, and PPP1R3B, together with xenobiotic and oxidative stress associated genes (Fig. 3A; Table S5). The recurrence of PCK1 and ACACB in both the differential-expression and co-expression analyses provides additional network-level support for their involvement in hepatic metabolic responses to heat stress. In breast muscle, the overlap set included genes associated with proteostasis and cell cycle regulation, including CDKN1A and DNAJA4, as well as lipid-metabolism and developmental candidates such as LIPG, SEMA6D, and FOXD2 (Fig. 3B; Table S6). These genes were therefore prioritized as network supported candidates rather than being interpreted as independently validated hub genes. In heart, DEG–module overlap genes were identified across several heat stress associated modules (Table S7). Candidate genes included GNAQ, GNAI1, PPP3CA, TRPC6, and other genes associated with calcium, G-protein, and cardiovascular signaling. Their kME and GS values were retained to describe network connectivity and association with heat-stress status (Fig. 3C). Collectively, the DEG–WGCNA overlap sets represent prioritized candidates for subsequent functional and genomic investigation. Functional enrichment of network-supported candidates identifies tissue associated response programmes To characterize the functional associations of the DEG–hub intersections, we performed enrichment and network analyses (Fig. 4). Tissue associated functional enrichment of consensus hub genes. (A) Circos plot linking hepatic hub genes to enriched pathways. (B) PPI network of hepatic hub genes. (C) Circos plot of breast muscle hub genes. (D) PPI network of breast muscle hub genes. (E) Circos plot of cardiac hub genes. (F) PPI network of cardiac hub genes. In liver, hub gene enrichment converged on the insulin-PPAR signaling axis, with PCK1 and ACACB as central nodes linking glucose and lipid metabolism, alongside stress-responsive cytochromes (CYP1A1, CYP4B7) and glycogen storage regulators (PPP1R3B/C) (Fig. 4A, B). Downregulated hub genes (PCK1, ACACB) were prominently associated with adipocytokine, insulin, and PPAR signaling, whereas upregulated genes included xenobiotic-processing cytochromes (Fig. 4A, B). In breast muscle, hub gene enrichment converged on p53-mediated cell-cycle arrest (CDKN1A), chaperone-assisted protein refolding (DNAJA4), and lipid catabolic processes (LIPG) (Fig. 4C, D). Additional enrichment of cytoskeletal and microtubule regulators (DIXDC1, RHPN1, KLHL42) further supports a coordinated cellular response that preserves structural integrity. This network-level convergence on growth arrest, proteostasis, and metabolic remodeling corroborates the muscle-specific transcriptional program identified in the consensus DEG analysis and supports a strategy that prioritizes cellular integrity and protein quality control over proliferation under thermal stress. In cardiac tissue, hub gene enrichment converged on calcium signaling and adrenergic responsiveness, with key nodes including the calcineurin modulator PPP3CA, G-protein subunits (GNAQ, GNAI1), and calcium signaling associated genes (TRPC6, SNTA1) (Fig. 4E, F). Apelin signaling emerged as a secondary regulatory axis, consistent with its established role in cardiovascular protection. This network-level convergence on excitation-contraction coupling pathways corroborates the cardiac specific transcriptional program and supports a strategy that prioritizes the preservation of contractile function over tissue reorganization under heat stress. Transcription factor activity inference supports hepatic regulatory patterns To identify upstream transcriptional regulators driving the hepatic heat stress response, we inferred transcription factor (TF) activities using DecoupleR (Badia-i-Mompel et al., 2022; Huang et al., 2025), with the DoRothEA regulon database (Garcia-Alonso et al., 2019). Among the top differentially active TFs, heat stress induced strong activation of stress-responsive regulators (Fig. 5A; Table S8). HSF2 showed the strongest positive activity (ULM = 9.03), followed by NCOA3 (ULM = 6.42), ARID2 (ULM = 4.39), MEF2A (ULM = 4.33) and MEF2C (ULM = 3.01). In contrast, growth related TFs MYC (ULM = −4.48) and FOXP1 (ULM = −6.32) were strongly suppressed, indicating reduced proliferative signaling under heat stress. Transcription factor activity inference and GSEA validation in liver under heat stress. (A) Top 20 differentially active TFs ranked by ULM t-statistic (HS vs CN). (B) Spearman correlation heatmap between TF activity scores and expression of high-confidence hepatic hub genes. (C) Enrichment plots for NCOA3, MYC, MEF2A, and MEF2C regulons. Here NES is normalized enrichment score while, n is regulon size). (D) Multi-metric validation heatmap integrating DecoupleR activity scores, GSEA NES, −log₁₀ (FDR), and regulon size for the four validated TFs. Correlation analysis between TF activity scores and hepatic hub gene expression revealed that NCOA3 and MEF2A/C activities correlated positively with PCK1 and other metabolic regulators, whereas MYC showed negative correlations with the same genes (Fig. 5B). These associations suggest that the observed hub gene expression changes are driven by specific upstream transcriptional programs rather than random variation. We further assessed these inferred activities using gene set enrichment analysis (GSEA) of the ranked regulon targets (Fig. 5C). NCOA3 showed significant positive enrichment (NES = 2.39, padj = 1.3 × 10⁻³, regulon size n = 23), confirming that its target genes are coordinately upregulated under heat stress. MEF2A (NES = 1.85, padj = 1.3 × 10⁻³, n = 93) and MEF2C (NES = 1.54, padj = 1.4 × 10⁻², n = 63) also showed significant positive enrichment. While MYC showed significant negative enrichment (NES = −1.68, padj = 1.3 × 10⁻³, n = 268), indicating negative enrichment of MYC target genes in the heat stress expression profile. The concordance analysis showed consistency between DecoupleR scores, GSEA normalized enrichment scores, and statistical significance across these four TFs (Fig. 5D). Transcription factor activity inference in breast muscle Among the top differentially active TFs, heat stress induced strong activation of SP1 (ULM = 4.68), SP3 (4.40), TEAD1 (4.32), FOXO4 (3.97) and SOX9 (3.60), alongside other stress-responsive regulators (Fig. 6A, Table S9). In contrast, lipogenic regulator SREBF2 (ULM = −2.82) and proliferative TFs E2F4 (ULM = −3.65) and MYCN (ULM = −2.90) were strongly suppressed. Transcription factor activity inference and GSEA validation in breast muscle under heat stress. (A) Top 20 differentially active TFs ranked by ULM t-statistic (HS vs CN); red indicates activation and blue suppression. (B) Spearman correlation heatmap between TF activity scores and breast muscle hub gene expression. (C) Enrichment plots for FOXO4, FOXO1, and SOX9 (activated, positive NES) and SREBF2. (D) Multi-metric validation heatmap integrating DecoupleR scores, GSEA NES, −log₁₀ (FDR), and regulon size. Correlation analysis revealed that FOXO4 activity correlated positively with CDKN1A (ρ = 0.38) (Fig. 6B), consistent with the established role of FOXO transcription factors in inducing CDKN1A-mediated cell-cycle arrest and autophagy (Calnan and Brunet, 2008; Webb and Brunet, 2014). The suppression of SREBF2, a positive regulator of lipogenic gene expression, aligns with the observed downregulation of lipid biosynthetic genes and the shift from anabolic growth to stress-responsive proteostasis. We further assessed the inferred TF activities using gene set enrichment analysis (GSEA) of their regulon targets (Fig. 6C). FOXO4 showed positive enrichment (NES = 1.94, FDR = 0.079, regulon size n = 13). FOXO1 (NES = 1.62, FDR = 0.079, n = 28) and SOX9 (NES = 1.92, FDR = 0.062, n = 9) showed similar positive enrichment trends. SREBF2 showed significant enrichment of its regulon targets (NES = 1.76, FDR = 0.041, n = 3) despite its suppressed activity. The multi-metric heatmap summarizes DecoupleR scores, GSEA NES, FDR, and regulon size for these TFs (Fig. 6D). Collectively, these findings support the involvement of FOXO4 associated cell-cycle regulation, SOX9 associated structural responses, and SP/TEAD associated transcriptional regulation in the breast muscle response to heat stress. Transcription factor activity inference suggests cardiac-associated regulatory network Among the top differentially active TFs, cardiac tissue showed dominant suppression under heat stress, with only four TFs activated (including RARA and FOXP1) while sixteen were suppressed (including HSF2, TEAD1, and MEF2A) (Fig. 7A).To validate whether these inferred TF activities were reflected in the expression patterns of their downstream targets, we computed Spearman correlations between TF activity scores and hub gene expression. The resulting correlation matrix revealed strong negative associations between ATF4 activity and stress-responsive hub genes, as well as coordinated suppression of STAT family members and FOXO3 with contractile and extracellular matrix-related targets (Fig. 7B; Table S10). Transcription factor activity inference and validation in heart under heat stress. (A) Top 20 differentially active TFs ranked by ULM t-statistic; red denotes activation and blue suppression. (B) Spearman correlation heatmap between TF activity scores and cardiac hub gene expression. (C) Enrichment plots for MEF2C, MEF2A, JUN, and TEAD1 showing negative NES values, indicating enrichment of regulon targets among downregulated genes consistent with suppressed activity. (D) Multi-metric validation heatmap integrating DecoupleR scores, GSEA NES, −log₁₀ (FDR), and regulon size. We further assessed the top TFs using gene set enrichment analysis (GSEA) of their regulon targets (Fig. 7C). MEF2C showed significant negative enrichment (NES = -1.4, regulon size n = 63), confirming its suppressed status. MEF2A (NES = −1.32, n = 94), JUN (NES = −1.49, n = 84), and TEAD1 (NES = −1.51, n = 120) also showed significant negative enrichment. Final validation confirmed concordance between DecoupleR scores, GSEA normalized enrichment scores, and statistical significance across all four validated TFs (Fig. 7D). The dominant suppression of MEF2A and MEF2C in heart differs with their strong activation in liver, revealing a tissue-specific divergence in MEF2 family regulatory deployment. The suppression of TEAD1, a key effector of Hippo signaling and cardiac developmental programs, indicates attenuated proliferative and structural remodeling under acute heat stress (Liu et al., 2017, 2019). JUN suppression reflects reduced AP-1-associated transcriptional activity, consistent with reduced transcriptional remodeling alongside the maintenance of calcium and adrenergic signaling programs (Windak et al., 2013). Discussion The analyses identified tissue associated patterns involving hepatic metabolic regulation, breast muscle cell cycle and proteostatic responses, and cardiac calcium related signaling. Because the tissues originated from different experiments, these patterns should not be interpreted as formal tissue by treatment differences. Previous studies have demonstrated that liver is more susceptible to oxidative stress than other organs under heat stress in broilers, functioning simultaneously as the primary site of energy metabolism, bile production, detoxification, and immune regulation (Oluwagbenga and Fraley, 2023). The strong metabolic enrichment of hepatic DEGs in insulin, PPAR, and pyruvate pathways is consistent with previous studies that heat-tolerant chickens exhibit transcriptional responses enriched in metabolic pathways, with liver metabolism being a primary target during both acute and chronic heat stress (Wang et al., 2020; Barreto Sánchez et al., 2022; Li et al., 2026). The downregulation of key metabolic genes, including the gluconeogenic gene (PCK1) and the lipid-metabolic gene (ACACB) in this study may represent a metabolic adjustment to heat stress as numerous studies have reported that suppressed hepatic glucose production during heat stress is an adaptive mechanism to reduce metabolic heat production (Jastrebski et al., 2017; Guo et al., 2021; Siddiqui et al., 2021). The coordinated downregulation of PCK1 and ACACB is consistent with a metabolic adjustment that may reduce energy expenditure during heat exposure. The gluconeogenic enzyme PCK1 has also been implicated in heat-stress responses beyond its canonical metabolic function (Jastrebski et al., 2017). Likewise, the downregulation of ACACB, which encodes an enzyme involved in fatty-acid metabolism is consistent with the recent studies that heat stress in skeletal muscle reduced the abundance of phosphorylated ATP-citrate lyase and fatty acid synthase, which causes suppression of the de novo lipogenesis as a way of minimizing metabolic heat generation (Ganesan et al., 2018). Upregulation of the components of the insulin signaling pathway such as glycogen-targeting protein phosphatase regulatory subunits PPP1R3B and PPP1R3C indicates the switch to a metabolic state that favors glycogen storage over gluconeogenesis (Barreto Sánchez et al., 2022). Transcription factor activity analysis further supported that these hepatic gene expression shifts are driven by coherent upstream regulatory programs. The strong activation of HSF2 is consistent with its established role in proteotoxic and oxidative stress responses (Alasady and Mendillo, 2024). In poultry, HSF-family activation constitutes the first tier of thermal defense, driving molecular chaperone induction to maintain protein homeostasis under heat stress (Balakrishnan et al., 2023; Hossain et al., 2026). Previous studies reported that HSF2 is active in circumstances where the stress is moderate and prolonged supporting our hypothesis that HSF2 might involve in heat adaptation where high environmental temperature sustains for longer period of time (Joutsen et al., 2020; Hästbacka et al., 2025). NCOA3 (ULM = 6.42) functions as a nuclear receptor coactivator that integrates metabolic and inflammatory signaling; its strong activation suggests coordinated upregulation of hormone-responsive metabolic switches in the liver (Horn and Tacke, 2024; Koo et al., 2026). The simultaneous activation of MEF2A and MEF2C aligns with their established roles as calcium-regulated transcription factors that integrate intracellular Ca²⁺/calmodulin-dependent and MAPK signaling with metabolic gene regulation (Kim et al., 2008; Dewenter et al., 2017). In contrast, MYC and FOXP1 exhibited suppressed activity, consistent with the observed downregulation of proliferative gene programs under thermal stress (Xu et al., 2023; Zhu et al., 2024). These inferred activities suggest that liver heat adaptation is transcriptionally orchestrated through HSF2-mediated thermal sensing, NCOA3 associated metabolic coactivation, and MEF2 calcium-regulated metabolic switching, directly complementing the insulin-PPAR metabolic reprogramming identified at the gene level. Breast muscle analysis revealed activation of proteostatic and cell cycle control machinery with upregulated DEGs enriching in p53 signaling, proteasome activity, and heat shock protein-mediated protein refolding. The presence of CDKN1A (cyclin-dependent kinase inhibitor 1A) and GADD45B in our high confidence key hub genes is indicative of the previously identified role of p53 transcriptional programme in protecting cells under thermal stress so that they can pause proliferation and instead use resources to mitigate stress (Furukawa et al., 2016; Wu et al., 2022). In parallel with this cell cycle arrest, previous studies reported that heat stress triggers muscle protein degradation by activation of the ubiquitin-proteasome system, mainly via atrogin-1 and its transcription factor FoxO3, which involves both the production of mitochondrial reactive oxygen species and secretion of corticosterone (Furukawa et al., 2016). These two mechanisms (growth arrest and proteolytic clearance) operate in parallel: while CDKN1A and GADD45B enforce proliferative quiescence, the simultaneous enrichment of ubiquitin-proteasome pathway genes in our upregulated hub gene set reflects activation of protein quality control machinery to clear heat-damaged polypeptides (Malila et al., 2021). The identification of proteasome-associated genes in our analysis is consistent with increased proteostasis and protein turnover as components of the heat-stress response as a recent study found that 26S proteasomes are stably activated during heat shock regardless of new protein synthesis and persist in their activity after restoring normal temperatures (Lee and Goldberg, 2022). Moreover, the strong upregulation of the heat shock protein binding (DNAJA4) and protein refolding functions that we see in our analysis of the breast muscle is consistent with literature that the expression of HSP70 in muscle tissue offers cytoprotection (Grune et al., 2011; Shehata et al., 2020). At the regulatory level, these proteostatic and cell-cycle programs were mirrored by coordinated transcription factor activation. DecoupleR identified significant activation of SP1, SP3, and TEAD1, broad transcriptional activators that govern cell proliferation and tissue integrity programs, alongside FOXO4 and FOXO1, canonical regulators of metabolic stress adaptation that transcriptionally induce CDKN1A-mediated cell-cycle arrest and autophagy (Calnan and Brunet, 2008; Webb and Brunet, 2014; Lasick et al., 2023). SOX9, a regulator of extracellular matrix homeostasis and tissue integrity (Faleeva et al., 2024), was significantly activated, consistent with involvement of SOX9-associated structural and developmental programs under stress. Conversely, SREBF2 and E2F4 exhibited suppressed activity, SREBF2 suppression is consistent with the downregulation of lipogenic gene programs under heat stress (Mitra et al., 2025). Together, these transcriptional responses suggest that breast muscle prioritizes tissue maintenance, FOXO and AP-1 associated proteostasis, and SREBF2 dependent lipogenic suppression over proliferation during heat stress. Functional enrichment analysis of the cardiac upregulated genes revealed a strong activation of calcium signaling pathways and adrenergic signaling pathways that are involved in excitation-contraction coupling and cardiovascular regulation during heat stress (Liu et al., 2024). The up-regulation of voltage-gated calcium channel genes such as CACNA1H, CACNB1, CACNA2D2 and calcium-modulating genes NOS1 and CAMK2B are in line with studies that heat stress triggers calcium overload in cardiomyocytes (Shaheen et al., 2011; McKee et al., 2021). On the other hand, the downregulation of extracellular matrix elements such as various collagen genes (COL4A1, COL4A2) and enrichment of focal adhesion and ECM-receptor interaction networks among downregulated genes is consistent with recent results that chronic cyclic heat stress represses structural gene networks that regulate muscle tissue integrity and cellular interactions (Liu et al., 2024). G-protein subunits (GNAQ and GNAI1), calcineurin component PPP3CA, and apelin signaling mediators emerged as high-confidence hub genes in our cardiac co-expression network. This further supports the involvement of GPCR, adrenergic, and calcium-associated signaling in the cardiac heat-stress response (Parra and Rothermel, 2017; Roy and Tamuli, 2022; Murali and Aradhyam, 2023). The integration of adrenergic signaling genes (PPP1R1A, GNAQ) and calcium handling machinery in our hub gene analysis highlights the heart's reliance on neurotransmitter-mediated contractile regulation to maintain cardiac output during heat stress. In contrast to the transcriptionally active responses observed in liver and breast muscle, cardiac tissue exhibited a strategy of regulatory suppression. The suppression of MEF2A and MEF2C in heart contrasts with their strong activation in liver, revealing a tissue-specific divergence in MEF2 family regulatory deployment. The suppression of TEAD1, a key effector of Hippo signaling and regulates cardiac growth, indicates attenuated proliferative and structural remodeling under heat stress. A recent study establishes that TEAD1 is essential for NRF2 expression and activity under both basal and stress conditions and plays a crucial role in the oxidative stress response in cardiomyocytes (Jagannathan et al., 2025). This dominant suppression pattern suggests that cardiac tissue preserves contractile function through constitutive calcium-adrenergic architecture rather than de novo transcriptional remodeling. These patterns of energy reallocation observed in this study are consistent with conserved strategies of organismal survival that extend beyond avian physiology. In human and experimental models of thermal stress, p53 has been shown to protect cells at temperatures near the heat-stroke threshold by restraining an excessive heat-shock response and thereby maintaining proteostasis (Gong et al., 2019; Janovska et al., 2023; Karakostis et al., 2024), illustrating how stress-responsive pathways can prioritize survival over growth in warm environments. At the systems level, central and peripheral thermoregulatory circuits coordinate sympathetic, cardiovascular, and metabolic outputs. This ensures that thermogenesis and circulation are maintained even when overall energy balance is disturbed, highlighting a functional hierarchy among organs during heat stress (Morrison et al., 2008; Qu and Ajuwon, 2018; Zhang et al., 2025). Our transcriptomic analysis supports this systems-level prioritization at the molecular level. Hepatic suppression of gluconeogenic and lipogenic axes, skeletal muscle activation of p53-mediated cell-cycle arrest and proteostatic maintenance, and cardiac preservation of contractile calcium signaling collectively suggest an energy reallocation strategy that may prioritize cellular maintenance and stress adaptation over growth under heat stress (Supplementary Fig. 6). Liver, breast muscle, and heart were obtained from separate studies that differed in breed, age, sample size, and heat exposure protocol, and a formal tissue × heat stress interaction could not be estimated. The observed patterns are therefore described as tissue associated rather than tissue specific. Concordance based vote counting identified conserved directional responses but did not estimate pooled effect sizes or statistically model between study heterogeneity. In addition, the cardiac analysis relied on an exploratory DEG union. Finally, TF activities were inferred using human regulons transferred to chicken through orthology and require species specific experimental validation. Conclusions This integrative analysis of RNA-seq datasets across three tissues provides a systems level map of the heat stress response in broiler chickens. The liver exhibits metabolic reprogramming involving gluconeogenic and fatty acid metabolic pathways, including downregulation of PCK1 and ACACB. Breast muscle activates p53-mediated cell cycle arrest and proteostasis through CDKN1A and DNAJA4. The heart preserves adrenergic-calcium signaling with CACNA1H and PPP3CA. These transcriptional programs are coordinated by distinct upstream transcription factors including HSF2, NCOA3, and MEF2A/MEF2C in the liver and FOXO4, SOX9, and SP family members in muscle. Collectively, these findings reveal a tissue associated division of labor in which each organ deploys a distinct transcriptional strategy to cope with heat stress. The liver curtails energy intensive biosynthesis; the muscle arrests proliferation and activates protein quality control; and the heart preserves contractile function while suppressing structural remodeling. This coordinated response represents an energy reallocation strategy that prioritizes short term survival over growth. The prioritized genes and regulatory networks provide candidates for subsequent validation in independent populations and controlled multi tissue experiments. Functional studies will be required before these candidates can be considered for genomic selection programmes or interpreted as determinants of heat tolerance. Funding This work was supported by National Key R&D Program of China (Grant No. 2021YFD1300100), and Key R&D Program of Shandong Province, China (No. 2026LZGC004). Author Contributions Ali Hassan Nawaz conceived and designed the study, curated the RNA-seq datasets, performed the bioinformatic and statistical analyses, interpreted the results, prepared the figures, and drafted the manuscript. Munwar Ali contributed to study design, data interpretation, and critical manuscript revision. Shaokat Ali contributed to formal data analysis and critical manuscript revision. Jiqiang Ding contributed to data analysis, result interpretation, and manuscript revision. Ying Duan contributed to data collection, review and visualization of findings. Muhammad Sayyam Akram contributed to data analysis, interpretation, and manuscript revision. Nasir Mukhtar contributed to data collection and critical manuscript review. Ting Xu contributed to data analysis and manuscript revision. Xiaoli Shi contributed to result interpretation and critical manuscript review. Thobela Louis Tyasi contributed to interpretation of the findings and manuscript revision. Diyan Li contributed to result interpretation and critical manuscript review. Yiqiang Zhao contributed to study design, interpretation, and critical manuscript revision. Chungang Feng conceived and supervised the study, contributed to study design and interpretation, and critically revised the manuscript. All authors read and approved the final manuscript. Disclosures All authors declare that they have no known competing interests. Footnotes Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.psj.2026.107884. Supplementary materials References Associated Data Supplementary Materials