All experiments in our study were approved by the Ethics Committee for Animal Experiments of the General Hospital of the Northern Theater of Operations of the Chinese People’s Liberation Army (2023M734296). A total of 33 healthy male rats (weighing 160–230 g, Liaoning Changsheng Biologicals) were used and randomly divided into the control (3 rats), 20 Gy (15 rats), and 30 Gy (15 rats) groups. Prior to subsequent processing, all rats were anesthetized using inhaled isoflurane after fasting for 2 h. Rats in the experimental groups were irradiated with a single 20 Gy and a single 30 Gy using the Infinity linear gas pedal of Medtronic, and a source-skin distance of 100 cm, respectively. During irradiation, rats were well aligned with the localization outline (Supplementary Figures S1A–E). Rats in the control group were anesthetized only without irradiation. One hour after irradiation, all rats resumed normal feeding and drinking. Three rats in each group were executed 10 days later to harvest the livers (Supplementary Figures S1F, G), half of which were immediately immersed in saline, cut into small pieces, placed in a freezing tube and stored on ice to prevent protein denaturation, and ultimately stored in a −80 °C refrigerator. The other half of the liver tissue was fixed by 4% neutral paraformaldehyde and then paraffin tissue embedded to produce 5 μm tissue sections. To verify the successful establishment of the rat RILI model, liver sections were subjected to hematoxylin-eosin (HE) staining for pathological examination. The proportion of inflammatory cell infiltration area and the proportion of fibrosis (blue staining) area were analyzed using ImageJ software.

A portion of rat liver tissue preserved at −80 was ground into powder, added to PASP protein lysis solution (100 mM NH4HCO3, 8M Urea, pH 8), and lysed by ultrasonication in an ice water bath for 5 min. Then, the obtained samples were centrifuged at 4 °C and 12,000 g for 15 min to collect the supernatant. After that, another 1 h of reaction was performed at 56 °C after the addition of 10 mM DTT, followed by another reaction for 1 h at room temperature in the dark when IAM was added. Subsequently, 4x volume of pre-cooled acetone were added to precipitate at −20 °C for 2 h. The generated precipitate was collected by centrifugation, followed by repeated precipitation with acetone. After air-drying, the proteins were solubilized using a proteolytic solution for quantification of proteins using a Bradford kit.

After the lysis of proteins, trypsin and 100 mM TEAB buffer were added for subsequent 4 h of digestion at 37 °C. Another overnight digestion was performed with the addition of trypsin and CaCl₂. The pH was adjusted with formic acid to less than 3. The next steps were centrifugation after mixing, de-salting using a C18 column, as well as washing and elution with 70% acetonitrile. The samples were lyophilized and dissolved in TEAB, followed by reaction at room temperature for 2 h after the supplementation of TMT labeling reagent. The reaction was terminated with ammonia, with the reaction mixture desalted and lyophilized finally.

LC-MS was performed with the preparation of mobile phase A (2% acetonitrile, and 98% water) and mobile phase B (98% acetonitrile, and 2% water). The lyophilized powder was dissolved with liquid A, centrifuged and passed through a Waters BEHC18 column (4.6 × 250 mm, 5 μm) for gradient separation at a column temperature of 45 °C. With the collection of one tube per minute, it was subjected to lyophilization and dissolution with formic acid. Mass spectrometry was performed using a TMT tag with a mass spectral range of 350–1,500 m/z, primary resolution of 60,000, secondary resolution of 45,000, and fragmentation collision energy of 32% (Supplementary Figures S2A, B).

The identified proteins were analyzed for Gene Ontology (GO), InterPro (IPR), Cluster of Orthologous Groups of Proteins (COG) as well as Kyoto Encyclopedia of Genes and Genomes (KEGG) functions and pathways using InterProScan software. DEPs were analyzed for volcano map, GO, IPR, KEGG and subcellular localization. Furthermore, to identify key modular genes, clustering analysis was performed by WGCNA package in R software (version 1.73). Protein-protein interaction (PPI) networks were constructed using the STRING database (https://string-db.org) and Cytoscape 3.9, associated with functional annotation by the Reactome database. PPD was performed using the HDOCK 1.1 software, and the number of docking samples was set to 100,000. Finally, the top-10 results with the best binding energy were exported, visualized and analyzed using PyMOL 2.6.

Paraffin-embedded sections of rat liver tissues were deparaffinized in xylene for 10 min (×2), passed stepwise through 100%, 95%, 80%, and 70% ethanol for 5 min of hydration each time, and finally immersed in distilled water for 5 min. Then, the processed sections were immersed in hematoxylin staining solution for 5 min, rinsed under running water for 5 min, and differentiated with 1% hydrochloric acid in alcohol for 10–20 s. After rinsing under running water for another 5 min, these sections were finally blued with ammonia or lithium carbonate for 30 s. After that, sections were stained in 1% eosin solution for 1–2 min, rinsed under running water, and then dehydrated in 70%, 80%, 95%, and 100% ethanol for 2 min, respectively. Finally, the sections were transparent in xylene for 10 min (×2) and sealed with neutral resin.

Rat hepatocyte lines (BRL and BRL-3A) were obtained from iCell (Shanghai, China) and cultured in Dulbecco’s Modified Eagle’s Medium (DMEM, Gibco, China) supplemented with 10% fetal bovine serum (FBS, Gibco, United States) and 1% penicillin/streptomycin. The human hepatocyte line (THLE-2) was purchased from ATCC (Manassas, VA, United States; catalog numbers CRL-2706 and CRL-11233) and maintained in BEGM SingleQuots medium (iCell, Shanghai, China) containing 10% FBS.

For gene silencing, small interfering RNA (siRNA) oligonucleotides specifically targeting THBS1 were obtained from GenePharma (Suzhou, China). Cells were seeded at an appropriate density and transfected with siRNA using Lipofectamine™ 2000 (Life Technologies, Thermo Fisher Scientific, MA, United States) following the manufacturer’s instructions. The final siRNA concentration was 37.5 nM. After 48 h of transfection, cells were exposed to 10 Gy of irradiation, with a non-targeting siRNA used as the negative control. Transfection efficiency was verified by Wb. For in vitro inhibition experiments, Ruxolitinib (S1378, Selleck Chemicals, United States) was dissolved in DMSO and diluted to a final concentration of 5 μM. Cells were treated with Ruxolitinib for 48 h, followed by 10 Gy irradiation, and protein samples were collected 48 h post-irradiation for analysis.

Total protein concentration was determined by BCA protein quantification kit (Coolaber, China) as instructed. For protein separation, the SDS-PAGE was performed on 10% polyacrylamide gel (with 12% separation gel). Then, 20 μg of quantitative protein sample was taken and mixed with loading buffer, boiled at 95 °C for 5 min, and stored at 4 °C for use. The gel was electrophoresed at 80 V for separation until the dye front reached the bottom of the gel. The proteins were transferred to PVDF or nitrocellulose membranes at 300 mA for 1.5 h and cooled at 4 °C. Membranes were sealed with 5% skim milk powder for 2 h of incubation at room temperature. Then, these membranes were incubated with specific primary antibody overnight at 4 °C and secondary antibody for 2 h at room temperature. After 3 times of washing (7 min each) using TBST, membranes were treated with ECL luminescent substrate (Thermo Scientific Pierce ECL), and signals were captured using a chemiluminescent imaging system (Bio-Rad ChemiDoc or GE Amersham Imager). Quantitative analysis of the protein bands was performed by ImageJ to calculate the relative expression of the target proteins, with normalization to the internal reference GAPDH. Specific information of the antibodies (i.e., company, stock number, and dilution ratio) used for WB are listed in the Supplementary Material.

ELISA was performed (THBS1; Animaluni, Shanghai; 0.156–10 ng/mL) to determine the concentration of THBS1 in plasma. The plasma samples of 0 Gy, 20 Gy, and 30 Gy groups were diluted 5-fold for the assay, and the absorbance was measured at 450 nm by a microplate spectrophotometer (Supplementary Figures S2E).

Paraffin-embedded rat liver sections were dewaxed in xylene, rehydrated through graded ethanol, and treated with 3% hydrogen peroxide to block endogenous peroxidase. After blocking with 5% BSA, sections were incubated overnight at 4 °C with the THBS1 primary antibody (1:2000). A HRP-conjugated secondary antibody was then applied, followed by DAB visualization and hematoxylin counterstaining. Slides were mounted with neutral resin and examined under a Carl Zeiss optical microscope.

Data were searched and obtained using ProteomeDiscoverer 2.4 with the rattus_norvegicus_uniprot database, with a precursor mass tolerance of 10 ppm and a fragment mass tolerance of 0.02 Da, as well as modifications including cysteine alkylation, methionine oxidation, and TMT tagging. Results were filtered using a False-Detection-Rate (FDR) to remove peptides and proteins with FDR >1%. Meanwhile, DEPs were analyzed using T-test (P < 0.05, Log2FC > 0.55 or < -0.55). GraphPad Prism 10 (San Diego, CA, United States) was employed for data analysis and result visualization. Normally distributed measurements were expressed as mean ± standard deviation (SD). Inter-group comparisons were made using independent samples t-tests. A value of P < 0.05 was considered statistically significant (*P < 0.05, **P < 0.01, and ***P < 0.001).

After radiation exposure, rats in the 30Gy group showed significantly decreased body weight loss compared to that of rats in the 20Gy group (average weight loss: 5.20 g vs. 1.98 g). All three rats in the control group showed a gradual increase in body weight. All mice survived from the beginning of irradiation to the 10th day after irradiation. Rats in the 30Gy group had dull fur from the third day onwards, a large asymmetric hair loss in the irradiated field on day 4, and localized erythema on the skin, accompanied by pain reactions such as licking, scratching of the irradiation field. Rats in the 20Gy group showed hair thinning and localized hair loss from day 7 onwards, with smooth and slightly erythematous skin, none significant pain reaction of licking and scratching. Furthermore, 5 rats in the 30Gy group showed obvious diarrhea (Bristol 6 type) on day 4, and the remaining 10 rats showed obvious diarrhea (Bristol 6 type) on day 5, with feces adhering to the anus in all rats. In addition, rats in the 30Gy group had obviously decreased food intake compared with that of the 20Gy group; and only two rats in the 20Gy group showed mild diarrhea on day 5 (Bristol type 5).

Based on the results of HE staining, on the 10th days after radiation exposure, rats were exhibited pathologic changes of liver specimens under light microscope. One of the three rats in the 20 Gy group showed hydropic degeneration of hepatocytes, and all of the three rats in the 30 Gy group showed varying degrees of hydropic degeneration of hepatocytes and inflammatory infiltration. The control group showed no obvious pathological changes in the liver sample. It indicated successful establishment of an early stage RILI model in rat (Figure 1). Given that all three samples in the 30Gy group showed different degrees of pathological changes, the rat liver samples from this group were selected for subsequent proteomic analysis and screening for DEPs.

PCA showed significant separation in the gene expression profiles of the 30Gy group and the control group in three-dimensional space, especially in the PC1 direction. Therefore, there might be obvious differences between the two groups in gene expression or proteomic features. The relative intra-group clustering of the three samples indicated relative spatial consistency in terms of the main characteristics of samples within group, suggesting low variability and good reproducibility in intra-group samples (Figure 2A). The CV cumulative curve can facilitate the assessment of the difference in inter-group variability, where the X-axis with lower CV indicates a higher concentration of the sample value, and higher CVs indicate greater inter-sample variation. Meanwhile, the Y-axis indicates the cumulative fraction, i.e., the proportion of samples with different CV values and below. As the CV increases, the cumulative fraction elevates and eventually approaches 1, indicating 100% of the sample data. As shown in Figure 2B, the steeper CV cumulative curve for the RILI and the control groups revealed less variability in the sample at 100%, i.e., better repeatability of the overall samples.

Weighted Gene Co-expression Network Analysis (WGCNA) was performed to identify gene modules closely associated with the radiation-induced liver injury (RILI) phenotype. As shown in Figures 3A–C, genes were clustered based on their expression profiles, with those exhibiting similar patterns grouped into the same module. Hierarchical clustering identified 15 distinct gene modules, among which the Cyan module showed the strongest correlation with RILI (R = 0.95, Figures 3D,E). Liver tissues from the modeled and control rats were taken for proteomics, resulting in 176 DEPs, of which 106 were upregulated and 70 were downregulated. These DEPs were visualized using volcano and heat maps (Figures 2C,D). By Venn diagram analysis, THBS1 gene was enriched in both WGCNA key modules and DEGs list, with both MM and GS of >0.85, suggesting its potentially significant role in the development of RILI with high biological significance (Supplementary Figures S2D). The key DEPs intersected with the WGCNA (Cyan) are shown in Table 1, mainly including THBS1, STAT3, COL6A2, etc.

To elucidate the biological functions of the DEPs following radiation exposure, multiple enrichment analyses were performed, including GO, KEGG, domain annotation, and subcellular localization. Collectively, the results indicated that radiation-induced changes in the liver proteome were closely associated with ECM remodeling, stress response, and metabolic adaptation.In the GO enrichment analysis (Figures 4A,B), biological process (BP) terms were predominantly enriched in wound healing, response to stress, and extracellular matrix organization, suggesting enhanced liver tissue remodeling and repair activities after irradiation. Cellular component (CC) terms were highly concentrated in the extracellular region and plasma membrane, aligning with the molecular function (MF) enrichment in endopeptidase inhibitor activity, receptor binding, and metal ion binding. These results indicate that radiation primarily activated extracellular and secreted proteins involved in ECM structure and intercellular signaling. Such enrichment patterns are consistent with the biological behavior of structural matrix proteins like collagens (COL family) and matricellular regulators such as THBS1, both of which contribute to ECM integrity and remodeling. KEGG pathway analysis (Figures 4C,D) further demonstrated that DEPs were mainly involved in ECM–receptor interaction, focal adhesion, and amino acid metabolism (particularly alanine, aspartate, glutamate, and tyrosine pathways). These metabolic pathways are essential for collagen synthesis and oxidative stress regulation, implying that radiation-induced metabolic reprogramming may support ECM reconstruction and fibrotic progression.

Subcellular localization analysis (Figure 4E) revealed that over one-quarter of DEPs were distributed in the extracellular space and plasma membrane. This finding corroborates the idea that major proteomic alterations occurred in the extracellular microenvironment. Structural domain enrichment analysis (Figure 4F) revealed significant enrichment of the thrombospondin type I repeat domain, a THBS1 hallmark, as well as of protease inhibitor I25. These results emphasize the roles of anti-proteolytic and matrix-binding motifs in ECM stabilization and tissue remodeling. These results imply that RILI triggers a coordinated response centered on ECM reorganization. Upregulation of ECM-associated domains and localization patterns implies that key structural proteins, including THBS1 and collagens, play central roles in mediating matrix remodeling and stress adaptation during RILI repair.

PPI analysis revealed that radiation-induced DEPs were mainly enriched in metabolic pathways, platelet degranulation, ECM organization, and PDGF signaling (Figure 5A). THBS1 occupied a central hub position within this network, bridging the ECM structural module (COL6A1, COL6A2, and COL5A1) and the signaling module (the PDGF–STAT3 axis) (Figures 5B,C). This topological configuration suggests that THBS1 acts as a key mediator that links extracellular matrix remodeling to the activation of intracellular profibrotic signaling.

Molecular docking analysis further supported the physical and functional relevance of this interaction network. The THBS1–PDGFA complex exhibited a stable binding conformation (binding energy = −39.5 kcal/mol), involving key residues such as SER108, ARG28, and ASP55 (Figure 5D). This high-affinity binding implies that THBS1 can facilitate PDGFA engagement and amplify downstream signaling, reinforcing its regulatory role in ECM remodeling and profibrotic signaling cascades.

The key protein THBS1, identified through screening, was further validated by WB. As shown in Figures 6A–D, THBS1 expression in the 30 Gy group was significantly higher than that in the control (0 Gy) group and exhibited a clear dose-dependent increase with radiation exposure. The WB results were consistent with the IHC findings (Figures 7A,B), indicating that the radiation-induced upregulation of THBS1 occurred not only at the overall expression level but also within specific tissue regions. IHC analysis further revealed that THBS1 was predominantly localized in ECM of hepatic tissue, particularly within the injured areas, supporting its role as an essential ECM component. Given its potential alteration under pathological conditions, we speculate that THBS1 may serve as a biomarker reflecting disease status. To further assess this hypothesis, THBS1 levels in the plasma of irradiated rats were quantified using an ELISA kit (Animaluni, Shanghai, China). THBS1 concentrations were significantly higher in the 20 Gy and 30 Gy groups than in the control group (P < 0.05; Figure 6E). Moreover, Pearson’s correlation analysis revealed a strong positive association between plasma THBS1 levels and tissue expression (P = 0.0010, R = 0.8973; Figure 6F), indicating that the content of THBS1 assayed in plasma could reflect the degree of RILI in liver tissues to a certain extent.

WB analysis showed that radiation (30 Gy) significantly increased the expression of PDG-FA, p-JAK2, and p-STAT3 in rat liver tissues, indicating activation of the JAK2/STAT3 pathway (Figures 6A,B). To determine whether this activation was mediated by THBS1, siRNA knockdown of THBS1 in BRL, BRL-3A, and THLE-2 cells markedly reduced p-JAK2 and p-STAT3 levels after irradiation, along with a decline in downstream collagens (COL5A1, COL6A1, COL6A2) (Figure 7). Furthermore, inhibition of JAK2/STAT3 signaling with Ruxolitinib produced similar effects, confi-rming that THBS1 promotes collagen expression through activation of the JAK2/STAT3 pathway.