FLASH radiotherapy preserves hepatic function and maintains metabolic homeostasis in a murine breast cancer model: an experimental preclinical study Authors: Xingyu Lu, Dehuan Xie, Liang Cui, Hang Shang, Linghong Zhou Published: 2026 ABSTRACT AbstractBackgroundRadiotherapy (RT) for treating breast cancer can result in incidental dose deposition to the liver, leading to functional impairment. FLASH radiotherapy (FLASH-RT), delivered at ultra-high dose rates (UHDR), has demonstrated remarkable normal tissue sparing. We investigated the hepatoprotective effects and metabolic alterations associated with FLASH-RT and conventional radiotherapy (CONV-RT) in a preclinical breast cancer model.MethodsFemale BALB/c mice bearing syngeneic 4T1 breast tumors were randomized to receive a single 20 Gy fraction of either FLASH-RT (625 Gy/s) or CONV-RT (0.54 Gy/s) using a 6 MeV electron beam. Tumor kinetics and systemic toxicity were monitored for 14 days. Hepatic integrity was assessed via histology, immunohistochemistry, oxidative-stress markers, serum biochemical assays, and non-targeted UPLC-MS/MS metabolomics of liver tissue.ResultsBoth modalities achieved isoeffective tumor growth delay. However, FLASH-RT induced only transient body-weight reduction followed by rapid recovery, and was associated with reduced hepatic injury markers compared with CONV-RT. Histopathological analysis revealed that FLASH-RT preserved hepatic architecture and attenuated early pro-fibrotic signaling, as reflected by reduced α-SMA and Nestin expression compared with CONV-RT. FLASH-treated mice exhibited significantly lower serum ALT/ALP levels and reduced oxidative stress (lower MDA levels and higher GSH-PX levels). Furthermore, FLASH-RT attenuated immune-mediated inflammation, as evidenced by diminished infiltration of CD8 + T cells and F4/80 + macrophages. Metabolomic profiling demonstrated that FLASH-RT preserved hepatic metabolic homeostasis, preventing the profound lipid and choline metabolism disruptions observed following CONV-RT.ConclusionsFLASH-RT was associated with substantial hepatic sparing by preserving structural integrity, attenuating oxidative–inflammatory signaling, and maintaining metabolic stability. These findings suggest that FLASH-RT may help expand the therapeutic window by uncoupling antitumor efficacy from collateral hepatic toxicity.Supplementary InformationThe online version contains supplementary material available at 10.1186/s13014-026-02878-x.Keywords: FLASH radiotherapy (FLASH-RT), Ultra-high dose rate (UHDR), Metabolomics, Radiation-induced liver injury (RILI), Breast cancer Background Breast cancer (BC) remains the most prevalent malignancy among women worldwide [1]. Standard therapeutic strategies typically involve surgery, chemotherapy, and radiotherapy (RT), the latter of which is essential for local tumor management [2–5]. Despite its clinical efficacy, RT can induce significant collateral damage to normal tissues, thereby diminishing patients’ quality of life and limiting treatment tolerability. Cumulative evidence suggests that severe radiation-induced toxicity affects approximately 59% to 73% of patients [6], imposing a ceiling on the maximum radiation dose that can be safely administered to the target volume. This clinical challenge necessitates innovative strategies that preserve therapeutic efficacy while minimizing adverse side effects. In this regard, FLASH radiotherapy (FLASH-RT) has emerged as a promising approach in radiation oncology, employing ultra-high dose rates exceeding 40 Gy/s to improve the therapeutic index [7–9]. Compared with conventional radiotherapy (CONV-RT), FLASH-RT has a distinctive biological advantage known as the “FLASH effect”, characterized by normal-tissue sparing and preserved antitumor efficacy [7, 10–13]. These benefits are hypothesized to involve rapid oxygen depletion, altered DNA-damage responses, and distinct reactive oxygen species (ROS) kinetics [14–18]. Extensive preclinical evidence suggests that FLASH-RT can decrease systemic toxicity and improve post-irradiation functional outcomes [19–21]. Further clinical translation of FLASH-RT will require careful evaluation of dose rate, beam delivery, and tumor-specific treatment contexts [22]. Specifically for breast cancer, FLASH-RT has the potential to mitigate incidental injury to adjacent organs while preserving antitumor efficacy. However, how FLASH-RT influences hepatic responses to non-primary-field exposure during thoracic or axillary irradiation remains poorly understood. Given the vital role of the liver in drug metabolism and systemic detoxification, thorough investigations into how FLASH-RT preserves hepatic integrity are essential. This study aimed to compare FLASH-RT and CONV-RT in terms of tumor control and hepatic normal-tissue responses in a 4T1 syngeneic breast cancer model. Specifically, we evaluated inflammation, early pro-fibrotic signaling, and metabolic alterations associated with acute liver injury after irradiation. These findings may help refine radiotherapy protocols and provide insight into strategies for maintaining tumor control while reducing normal-tissue injury. Methods Syngeneic 4T1 breast tumor model The Animal Management Committee of Southern Medical University approved all the animal procedures (Approval No. L2023039), which were conducted in compliance with institutional ethical guidelines. A total of 36 female BALB/c mice (4 weeks old, average weight 14.5 ± 1 g) were purchased from Beijing Sinogenetic Biotechnology Co., Ltd. (production license No. SCXK (Beijing, China) 2022-0006). The mice were maintained under controlled conditions (temperature, 22 °C ± 3 °C; humidity, 55% ± 5%; 12-hour light/dark cycle). Following one week of acclimatization, a syngeneic 4T1 breast tumor model was established by subcutaneous injection of 5 × 106 murine 4T1 cells (Guangzhou Xinyuan Biotech Co. Ltd., Guangzhou, China) suspended in 0.2 mL of a 1:1 (v/v) mixture of phosphate-buffered saline (PBS) and Matrigel into the right axillary region. The tumor volume (V) was monitored daily using digital calipers and calculated using the formula: V = 0.52×(L×W2), where L and W represent the length and width of the tumor, respectively. Body weights were also recorded daily throughout the experimental period. A total of 36 mice were randomized into three groups, with 12 mice per group: Sham-RT, CONV-RT, and FLASH-RT. Survival, tumor-volume, and body-weight analyses were based on the initial randomized cohorts. Because animals in each group reached humane endpoints or died during follow-up, the final number of animals available for terminal serum biochemical assays, histological analysis, immunostaining, TUNEL assay, and metabolomic profiling was n = 8–10 per group. Irradiation setup and dosimetry Mice received a single fraction prescription dose of 20 Gy using a 6 MeV generated electron beam delivered via either conventional radiotherapy (CONV-RT) or ultra-high dose rate radiotherapy (UHDR-RT, also known as FLASH-RT). The irradiation field was precisely aligned with the tumors to ensure accurate targeting. Beam characteristics were verified dosimetrically to confirm effective FLASH irradiation at the prescribed dose rates. On day 14 post-inoculation, when the tumors reached an average volume of 500 mm³, the mice were randomly assigned to the Sham-RT (Control), CONV-RT, or FLASH-RT treatment groups, which were matched for body weight and tumor size. Anesthesia was induced using tribromoethanol (0.2 mL/10 g body weight). For irradiation, mice were placed in the prone position on a custom-designed platform with a 0.5-cm-thick polymethyl methacrylate (PMMA) build-up holder. The electron beam was delivered in a dorsal-to-ventral direction (Fig. 1h). A 5-cm-thick lead shield with a 1 × 1 cm aperture was positioned so that the aperture was aligned only with the right axillary tumor. Because the tumor was located in the unilateral axillary region close to the lateral thoracic wall rather than the abdomen, the liver was outside the primary irradiation aperture. EBT-XD films were placed at the dorsal entrance plane and ventral exit plane to verify dose delivery and beam reproducibility. These film measurements were used because a dedicated small-animal treatment-planning system or direct tumor-center dose measurement was not available in this experimental setup. Tumor growth was evaluated daily for two weeks post-irradiation. Effects of FLASH and conventional radiotherapy on tumor growth in a breast cancer model. (a) Dose distribution heat maps and corresponding profile curves at the dorsal entrance plane. (b) Dose distribution heat maps and corresponding profile curves at the ventral exit plane. (c) Relative tumor volume changes over 14 days for the treatment and sham groups. (d) Comparison of relative tumor volumes among the treatment and sham groups on day 14 (sampling day). (e) Longitudinal changes in relative body weight after irradiation. (f) Survival curves representing the outcomes of various treatment modalities at 14 days post-treatment. (g) Schematic depiction of the experimental timeline, illustrating the progression of treatments and assessments. (h) Irradiation geometry showing dorsal-to-ventral beam delivery through a tumor-aligned 1 × 1 cm aperture, with the liver outside the primary field Dosimetry calibration for Gafchromic EBT-XD films was performed independently under both CONV-RT and FLASH-RT conditions. To account for potential ion recombination effects inherent in ion chambers at UHDR, film dosimetry served as the primary standard for dose verification, with a focus on cross-calibrating the pulse-to-pulse consistency of the 6 MeV electron beam. The minor dosimetric variance between cohorts and its potential biological impact are thoroughly addressed in the Discussion section. Sample collection At the end of the experiment (2 weeks post-irradiation), following a 12-hour fasting period, mice were euthanized via an intraperitoneal overdose of 1% pentobarbital sodium (150 mg/kg). Deep anesthesia was verified by the absence of the pedal withdrawal reflex prior to sacrifice. Following confirmed cessation of respiratory and cardiac functions, blood was harvested via cardiac puncture and centrifuged at 3500 rpm for 15 min at 4 °C to isolate the serum. Liver tissues were fixed in 4% paraformaldehyde, embedded in paraffin, and sectioned into 3–5 μm slices for subsequent histopathological analyses, including hematoxylin and eosin (H&E) staining and immunohistochemistry (IHC). The remaining tissue samples were rapidly frozen and stored at -80 °C for future analyses. Histological and immunohistochemical (IHC) analyses After dewaxing and rehydration, the liver sections were stained with H&E. For IHC analysis, the sections were treated with 3% hydrogen peroxide, blocked with goat serum, and incubated overnight at 4 °C with primary antibodies. Secondary antibodies were applied, followed by staining with diaminobenzidine (DAB), counterstaining with hematoxylin, and mounting with neutral balsam. The scanned slides were documented via an SQS-40R slide scanner (Shenzhen Shengqiang Technology Co., Ltd., China), and image analysis was performed via ImageJ (v1.52a) software. Immunofluorescence (IF) assay For IF staining, tissue Sects. (3–5 μm) were dewaxed, rehydrated, and blocked with 10% goat serum containing 0.3% Triton X-100 for one hour at room temperature. The sections were then incubated overnight at 4 °C with primary antibodies. After the samples were washed, fluorophore-conjugated secondary antibodies were applied for one hour at room temperature, followed by DAPI staining for nuclear visualization. Images were acquired via a KFBIO Digital Pathology Scanner (Konfoong Bioinformation Technology Co., Ltd., China). TUNEL assay Apoptotic cells in tissue sections were identified via a TUNEL apoptosis kit (Red, AF596; Elabscience Biotechnology Co., Ltd., Wuhan, China) per the manufacturer’s instructions. Biochemical analysis Liver tissues were homogenized, sonicated, and centrifuged to remove debris, with the supernatants stored at -80 °C for subsequent analyses. Commercial kits from Jiancheng Biology (Nanjing, China) were used to quantify oxidative stress markers (MDA, SOD, and GSH-PX) in liver samples. Serum biochemical parameters (ALT, AST, ALP, TBIL, BUN, UA, CR, TG, and CHO) were measured via kits from Nanjing Jiancheng Bioengineering Institute according to the provided protocols. Metabolite analysis For metabolomic analysis, 50 mg samples of liver tissue were extracted via a 4:1 methanol: water solution, followed by sonication and centrifugation at 13,000 rpm for 15 min at 4 °C. The cleared supernatants were analyzed via ultra-performance liquid chromatography coupled with tandem mass spectrometry (UPLC‒MS/MS). Data processing, including peak extraction, alignment, and identification, was performed via Progenesis QI software (Waters Corporation, Milford, USA). Principal component analysis (PCA) and enrichment analyses were conducted via Omic Studio tools (v3.6, https://www.omicstudio.cn/tool). Differentially abundant metabolites were primarily identified using VIP > 1 and p < 0.05 as the primary exploratory feature-selection criteria. To evaluate the robustness of metabolite selection under a more conservative threshold, an additional sensitivity analysis was performed using VIP > 1.5 while retaining p < 0.05. KEGG pathway annotation and enrichment analyses were conducted based on the Kyoto Encyclopedia of Genes and Genomes database (http://www.genome.jp/kegg/). Statistical analyses Statistical analyses were performed using GraphPad Prism 9.6 (GraphPad Software, La Jolla, CA, USA). Data are presented as mean ± SEM. For comparisons between two groups, unpaired two-tailed Student’s t-tests or Mann–Whitney U tests were used according to data distribution. For comparisons among multiple groups, one-way or two-way ANOVA with Bonferroni post hoc correction was used when appropriate. Longitudinal tumor-volume and body-weight data were plotted to illustrate temporal trends during follow-up; statistical comparisons were performed only for specified endpoints or comparisons indicated in the corresponding figure legends. Survival curves were compared using the log-rank test. A p value < 0.05 was considered statistically significant. Results Both FLASH-RT and CONV-RT demonstrated comparable efficacy in suppressing tumor progression and improving survival in the 4T1 breast cancer model. However, FLASH-RT (625 Gy/s) showed a more favorable normal-tissue response profile, with acceptable acute tolerability and reduced treatment-related hepatic injury markers. Notably, FLASH-RT attenuated acute hepatic inflammation, early pro-fibrotic signaling, and oxidative stress, thereby preserving hepatic functional integrity and metabolic homeostasis. Dosimetric characterization and UHDR execution Both cohorts were irradiated using a 6 MeV generated electron beam. While both groups received a nominal single-fraction dose of 20 Gy, the delivery kinetics differed markedly: CONV-RT was delivered at a standard rate of 0.54 Gy/s, whereas FLASH-RT achieved an ultra-high dose rate (UHDR) of 625 Gy/s. Dose deposition and coverage were verified using Gafchromic EBT-XD films positioned at the dorsal entrance plane and ventral exit plane. The dose distribution heatmaps and axial profiles (Fig. 1a and b) confirmed robust beam homogeneity and precise target alignment across the tumor volume. Based on the film dosimetry results, the mean delivered doses were recorded as 23.64 ± 0.79 Gy for the FLASH-RT group and 21.33 ± 0.67 Gy for the CONV-RT group (Table 1). Although the FLASH-RT cohort received a slightly higher physical dose (approximately 10.8% variance), both modalities achieved comparable tumor growth inhibition (Fig. 1c and d), providing an appropriate experimental setting for evaluating normal-tissue responses under comparable tumor-control conditions. Table 1Experimental grouping and irradiation parametersModeMeasured Dose(Gy)AverageDR (Gy/s)FR(Hz)PW(µs)NOPGun HV(%)Field size(cm2)FLASH-RT23.64 ± 0.79~ 62525048751 × 1CONV-RT21.33 ± 0.67~ 0.54101.137089.51 × 1DR: Dose Rate; FR: Frequency; PW: Pulsed Width; NOP: Number Of Pulses; Gun HV: Gun high voltage Experimental grouping and irradiation parameters DR: Dose Rate; FR: Frequency; PW: Pulsed Width; NOP: Number Of Pulses; Gun HV: Gun high voltage Isoeffective tumor suppression and enhanced systemic tolerability Following a single 20 Gy fraction, 4T1 tumor-bearing mice were monitored until humane endpoints were reached (Fig. 1g). Although the measured physical dose was modestly higher in the FLASH-RT group than in the CONV-RT group, both modalities yielded comparable tumor growth inhibition (Fig. 1c and d). These findings confirm that for the malignant targets, the two irradiation protocols were biologically isoeffective. Body-weight monitoring showed transient post-irradiation weight fluctuations across the treatment groups. In the FLASH-RT cohort, the initial reduction in body weight was followed by rapid recovery within approximately 2 days, and body-weight changes tended to converge among groups by the end of the observation period (Fig. 1e). Survival rates remained consistent across all cohorts, suggesting that the radiation doses employed were within the acute tolerability threshold (Fig. 1f). Attenuation of early pro-fibrotic signaling and preservation of hepatic architecture Histological evaluation at 2 weeks post-irradiation demonstrated that FLASH-RT was associated with better preservation of the hepatic parenchyma. H&E staining (Fig. 2a) revealed that CONV-RT induced classic signs of radiation-induced liver injury (RILI), including ballooning degeneration, focal necrosis, and eosinophilic body formation. In contrast, the hepatic architecture in the FLASH-RT group remained largely intact with negligible inflammatory infiltration. Histological and immunofluorescent analysis of liver tissues across different treatment groups. BALB/c mice underwent FLASH, conventional (CONV), or sham therapy. (a) Hematoxylin and eosin (HE) staining of liver sections from the various treatment groups. (b) Double immunofluorescence staining of liver sections, highlighting alpha-smooth muscle actin (α-SMA, red) and Nestin (green), with DAPI (blue) used for nuclear labeling. (c) Violin plots depicting the percentage area of immunofluorescence for α-SMA and Nestin in liver tissues among the treatment groups. Statistically significant differences are indicated as follows: ns, not significant; *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001 Immunofluorescence analyses of α-SMA and Nestin further revealed the pro-fibrotic activation status (Fig. 2b). CONV-RT triggered a significant up-regulation of α-SMA expression, indicating the rapid transition of hepatic stellate cells (HSCs) into myofibroblasts. Conversely, FLASH-RT maintained significantly lower levels of α-SMA and fewer Nestin-positive repair cells (Fig. 2c). This attenuation of early pro-fibrotic signaling, despite comparable tumor delay, supports a more favorable therapeutic window for FLASH-RT. Multi-parametric evaluation of hepatoprotection and immune homeostasis Preservation of hepatic functional and renal markers Biochemical profiling (Fig. 3a) revealed that CONV-RT increased hepatocellular injury and biliary dysfunction markers, as evidenced by significantly elevated ALT, ALP, and TBIL levels (p < 0.05). Remarkably, FLASH-RT maintained these markers at levels approaching those of the sham controls. Interestingly, AST levels tended to decrease (CONV> Sham> FLASH) but were less sensitive to acute radiation injury in this model. With respect to renal function, the CONV-RT group exhibited increased renal stress (elevated UREA and BUN, p < 0.05), whereas FLASH-RT significantly mitigated these elevations, consistent with better preservation of renal biochemical markers. Furthermore, FLASH-RT stabilized lipid metabolism, preventing the radiation-induced increase in TG and CHO levels observed in the CONV-RT group. Evaluation of biochemical and oxidative stress parameters in liver tissues from different treatment groups. BALB/c mice were treated with FLASH, conventional (CONV), or sham therapies. (a) Serum biochemical parameters, including alanine aminotransferase (ALT), alkaline phosphatase (ALP), aspartate aminotransferase (AST), creatinine (CREA), urea (UREA), blood urea nitrogen (BUN), triglycerides (TG), cholesterol (CHO), and total bilirubin (TBIL) for each treatment group. (b) Oxidative stress indicators in liver tissues, including malondialdehyde (MDA), superoxide dismutase (SOD), and glutathione peroxidase (GSH-PX) levels. (c) Violin plots representing the percentage area of TUNEL-positive fluorescence in liver tissues across the treatment groups. (d) TUNEL staining of liver sections, with TUNEL marked in red and DAPI in blue for nuclear staining. (e) Immunohistochemical analysis of liver sections for myeloperoxidase (MPO), F4/80, CD8, NLRP3, and Caspase-1, alongside violin plots displaying the percentage of positive area. (f) Violin plots illustrating the positive area of immune markers in liver sections. Data are expressed as mean ± SEM (n = 8–10 terminal samples per group, depending on sample availability after humane endpoints or death during follow-up). Statistically significant differences are indicated as follows: ns, not significant; *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; analyzed using appropriate statistical tests FLASH-RT attenuates oxidative stress and apoptotic signaling FLASH-RT was associated with reduced oxidative damage (Fig. 3b). CONV-RT induced extensive lipid peroxidation (elevated MDA, p < 0.05), whereas FLASH-RT maintained superior endogenous antioxidant capacity, specifically preserving GSH-PX activity (p < 0.005 vs. CONV). Correspondingly, TUNEL assays (Fig. 3b and c) revealed that FLASH-RT markedly reduced the radiation-induced increase in apoptosis observed in the CONV-RT cohort, consistent with improved preservation of cellular viability and hepatic functional markers. Suppression of the NLRP3 inflammasome and immune dysregulation IHC profiling revealed that CONV-RT provoked robust recruitment of MPO+ neutrophils and F4/80 + macrophages (Fig. 3d). Crucially, the CONV-RT group exhibited a marked activation of the NLRP3/Caspase-1 inflammasome axis (p < 0.00005), a key driver of sterile inflammation in RILI. FLASH-RT significantly downregulated NLRP3 and Caspase-1 expression, suggesting that UHDR delivery was associated with reduced DAMPs-related inflammatory signaling and cytotoxic CD8 + T-cell infiltration to levels akin to those of the sham controls. Metabolomic profiling of hepatic metabolic homeostasis Exploratory metabolomic pattern recognition To characterize the global metabolic shifts induced by different irradiation modalities, we employed a non-targeted metabolomics approach across both positive and negative ion modes. A partial least squares-discriminant analysis (PLS-DA) model was constructed to evaluate the variance among the Sham-RT, CONV-RT, and FLASH-RT cohorts. The score plot showed partial group separation, with PC1 and PC2 accounting for 21.33% and 22.90% of the total variance, respectively (R² = 0.854, Q² = 0.324; Fig. 4a). The permutation test did not suggest obvious overfitting (intercept Q² = -0.501, intercept R² = 0.535; Fig. 4b). However, the Q² value of 0.324 indicated modest predictive performance; therefore, the PLS-DA results were interpreted as exploratory pattern-recognition findings rather than as a stand-alone predictive classification model. Analysis of metabolomic profiles among different treatment groups. BALB/c mice were subjected to various treatments: sham, FLASH, and conventional (CONV). (a) Partial least squares-discriminant analysis (PLS-DA) score plot showing group separation among Sham-RT, FLASH-RT, CONV-RT, and quality-control (QC) samples. (b) Permutation test results for the PLS-DA model across the three groups. (c) Heatmap depicting the abundance of differential metabolites (z-score normalized) among the treatment groups. (d) Circular plot categorizing the classes of differential metabolites identified across the groups. (e) Correlation heatmap illustrating the relationships among differentially expressed metabolites. Statistically significant differences are indicated as follows: ns, not significant; *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; analyzed using appropriate statistical methods Differential metabolite signatures and the protective effect of FLASH Among the detected metabolites, 81 were identified as differentially abundant using the primary exploratory criteria of VIP > 1 and p < 0.05. To further evaluate feature-selection robustness, we performed a conservative sensitivity analysis using VIP > 1.5 while retaining p < 0.05. This stricter threshold retained 31 of the 81 metabolites, corresponding to 38.3% of the primary metabolite set (Fig. S1A; Table S1; see Additional file 1). KEGG pathway enrichment and functional impact To evaluate the functional impact of radiation-induced metabolic alterations, KEGG pathway enrichment was performed (Fig. 5a). Integrated mapping using Sankey and String diagrams revealed the top 10 enriched pathways, with a predominant clustering in metabolism and organismal systems. Notably, a bubble plot analysis (Fig. 5b) revealed that Glycerophospholipid, Choline, and Ether lipid metabolism—key components of cell membrane integrity and inflammatory signaling—were the biological domains most susceptible to ionizing radiation. While CONV-RT induced a broad perturbation of these metabolic networks, FLASH-RT significantly mitigated these disruptions. These exploratory pathway-level findings suggest that preservation of lipid- and membrane-associated metabolic profiles may be associated with the hepatic-sparing effect of FLASH-RT. However, targeted metabolite validation and functional mechanistic studies are required to establish causality. Exploratory metabolite-pathway analysis of hepatic metabolic alterations across treatment groups. BALB/c mice were subjected to sham, FLASH, or conventional radiotherapy (CONV-RT). (a) KEGG enrichment circular plot showing pathway entries associated with differentially abundant metabolites identified using the primary exploratory criteria of VIP > 1 and p < 0.05. The outer circle represents pathway IDs, and color coding indicates pathway classifications. (b) Metabolite-pathway chord diagram illustrating the relationships between differentially abundant metabolites and enriched pathways. (c) Sankey diagram showing the connections between differential metabolites and enriched pathways, accompanied by a bubble plot of enriched pathways KEGG top 10 pathway comparison between the primary VIP > 1 analysis and the conservative VIP > 1.5 sensitivity analysis showed that 6 of the top 10 pathways overlapped between the two analyses (Table S2; see Additional file 1). Importantly, glycerophospholipid metabolism, ether lipid metabolism, and glycerolipid metabolism remained represented under the stricter threshold. The VIP > 1.5 analysis also highlighted lipid- and membrane-associated pathways, including sphingolipid metabolism/signaling and phosphatidylinositol-related pathways (Fig. S1B; see Additional file 1). These results support the robustness of lipid- and membrane-associated metabolic alterations while remaining exploratory. Exploratory association between metabolic perturbation, hepatic inflammation, and early pro-fibrotic signaling Our findings suggest that radiation-associated metabolic perturbation may be linked to hepatic inflammatory injury rather than establishing a direct causal pathway. In the CONV-RT group, disruption of glycerophospholipid and choline metabolism was accompanied by increased F4/80 + macrophage and CD8 + T-cell infiltration. Because Kupffer-cell activation and cytokine release, including TNF-α and IL-6, are recognized contributors to hepatic inflammation and fibrogenesis [23], these observations support a potential association between altered lipid-associated metabolism and immune-mediated hepatic injury. Specifically, changes in palmitic-acid-related metabolism may reflect mitochondrial and membrane stress and may contribute to ROS generation. An oxidative microenvironment can promote the release of damage-associated molecular patterns (DAMPs), such as HMGB1, and activate inflammatory cell-death pathways, including the NLRP3/Caspase-1 axis [24, 25]. Consistent with this interpretation, CONV-RT increased oxidative stress markers, TUNEL-positive cells, and NLRP3/Caspase-1 expression. The concurrent increase in α-SMA and Nestin should be interpreted as early pro-fibrotic signaling rather than established fibrosis within the 2-week observation window. In contrast, FLASH-RT was associated with preservation of lipid- and membrane-associated metabolic profiles and attenuation of oxidative stress, inflammatory-cell infiltration, apoptotic signaling, and early pro-fibrotic activation. Together, these findings provide a hypothesis-generating metabolic perspective on FLASH-mediated hepatic sparing, suggesting a possible link between preserved lipid homeostasis and reduced metabolic–oxidative–inflammatory injury. Targeted metabolite validation and functional pathway experiments are required to establish causality. Discussion The clinical delivery of radiotherapy for breast cancer, particularly when it involves right-sided tumors or internal mammary chain (IMC) irradiation, frequently involves unintended dose deposition in the hepatic dome. In this study, we provide experimental evidence that, compared with conventional-rate irradiation (CONV-RT), FLASH-RT (625 Gy/s) was associated with better preservation of hepatic structural and functional integrity in a 4T1 syngeneic breast cancer model. Our multi-omics approach suggested that FLASH-RT was associated with preservation of metabolic homeostasis and attenuation of ROS-NLRP3-associated inflammatory and early pro-fibrotic signaling. Uncoupling antitumor efficacy from normal tissue toxicity A critical prerequisite for validating the biological advantages of ultra-high dose rate (UHDR) delivery is the demonstration of isoeffective tumor control. A notable observation in our study is that the actual delivered dose for FLASH-RT (23.64 Gy) was numerically higher than that for CONV-RT (21.33 Gy). Under standard radiobiological principles, this ~ 10.8% increase in physical dose would typically be expected to exacerbate hepatic injury and systemic toxicity. However, our findings showed a different pattern: the FLASH-RT group exhibited significantly preserved liver function and metabolic stability compared to the CONV-RT group. This phenomenon is consistent with a normal-tissue sparing effect despite the modestly higher measured physical dose in the FLASH-RT group [7, 14, 26]. Given that both modalities exhibited comparable tumor growth kinetics (Fig. 1c, d), the better preservation of hepatic integrity in the FLASH-RT cohort supports the potential of UHDR delivery to uncouple malignant target irradiation from collateral organ toxicity. Consistent with the findings of Wang et al. and the paradigm shift proposed by Loo et al. [27], our results support the potential of FLASH-RT to expand the therapeutic window by improving normal-tissue tolerance. UHDR-associated attenuation of the ROS-NLRP3 inflammatory axis Radiation-induced liver injury (RILI) is orchestrated by a cascade of sterile inflammation triggered by DAMPs. Our results suggest that FLASH-RT may alter the early radiochemical and oxidative response associated with this process. By delivering doses at ultra-high rates, FLASH-RT may promote rapid radical–radical recombination and transient oxygen depletion, thereby minimizing the generation of persistent reactive oxygen species (ROS). Specifically, FLASH-RT attenuated oxidative stress, as evidenced by reduced MDA levels and increased antioxidant activity (GSH-PX) [28]. This preservation of the oxidative environment may contribute to reduced activation of the NLRP3 inflammasome and Caspase-1 axis [29]. We observed that while CONV-RT induced a marked inflammatory response and significant hepatocellular apoptosis (TUNEL+), FLASH-RT markedly attenuated these responses. This suppression may attenuate early pro-fibrotic activation during the acute observation window, as reflected by reduced α-SMA and Nestin expression [30]. Immune modulation and metabolism‒immune crosstalk Our immune profiling revealed significant reductions in CD8 + T-cell infiltration and macrophage numbers in the FLASH-RT group, indicating an anti-inflammatory response distinct from that in the CONV-RT group. This finding aligns with that of Loo et al. [27], who reported that UHDR radiotherapy modulates immune response dynamics—likely through altered cytokine profiles [31], while maintaining tumor control [32]. In addition to traditional histopathology, non-targeted metabolomic profiling provided systemic insights into the protective effect of FLASH. CONV-RT was associated with broad metabolic perturbation, particularly in the context of cholesterol, choline, and ether lipid metabolism [12]. In contrast, the metabolic profile of the FLASH-RT group was markedly congruent with that of the sham control group. Notably, metabolites associated with cellular defense, such as Palmitic acid and Palmitoyl sphingomyelin, remained stable in the FLASH-RT cohort. Based on these exploratory metabolomic and immunohistochemical findings, we hypothesize that preservation of lipid-associated metabolic profiles may contribute to reduced oxidative and inflammatory responses after FLASH-RT. This hypothesis requires targeted metabolite validation and functional pathway studies. This preservation may reduce mitochondrial stress and DAMP-related inflammatory signaling, thereby contributing to lower NLRP3/Caspase-1 activation and hepatic inflammation. These findings correlate with those of Martínez-Rovira et al. [33], who utilized infrared microspectroscopy to underscore the importance of maintaining biomolecular integrity for optimal metabolic and structural outcomes following radiotherapy. Clinical implications and limitations Integrating FLASH-RT into the current treatment paradigm could represent a significant shift in radiotherapy [12]. For breast cancer patients, the ability of FLASH-RT to preserve liver function while attenuating acute hepatic injury and early pro-fibrotic responses may be relevant for reducing radiation-associated hepatic toxicity. By maintaining metabolic homeostasis, FLASH-RT could reduce systemic hepatotoxicity and radiation-associated fatigue, which are often associated with incidental hepatic exposure [27, 32]. However, several limitations must be acknowledged. First, this was a preclinical study using a 4T1 murine breast cancer model, and the findings may not fully recapitulate hepatic responses in patients receiving breast radiotherapy. Second, no formal a priori sample-size calculation was performed. Although 36 mice were initially randomized, animals in each group reached humane endpoints or died during follow-up, resulting in final terminal sample sizes of n = 8–10 per group. In the Sham-RT group, early humane euthanasia was mainly due to excessive tumor burden. Third, the 14-day observation window allowed assessment of acute hepatic injury, oxidative stress, inflammation, and early pro-fibrotic signaling, but was insufficient to evaluate established fibrosis or long-term collagen deposition. Masson’s trichrome staining, Sirius Red staining, hydroxyproline quantification, and long-term functional follow-up were not performed. Therefore, the α-SMA and Nestin findings should be interpreted as early pro-fibrotic responses rather than mature hepatic fibrosis. Fourth, the metabolomic findings should be interpreted as exploratory and hypothesis-generating. Although the primary metabolite set was defined using VIP > 1 and p < 0.05, an additional VIP > 1.5 sensitivity analysis retained 31 high-weight metabolites and preserved several lipid- and membrane-associated pathways. Nevertheless, targeted LC-MS/MS validation and functional mechanistic experiments were not performed. Therefore, these data support an association between FLASH-RT and preservation of hepatic metabolic homeostasis but do not establish causality. Fifth, the measured physical dose was modestly higher in the FLASH-RT group than in the CONV-RT group; although FLASH-RT still showed lower hepatic injury markers despite this dose difference, the dosimetric difference should be considered when interpreting the magnitude of normal-tissue sparing. Finally, although the 1 × 1 cm aperture was aligned only with the right axillary tumor and the liver was outside the primary irradiation field, this animal irradiation geometry cannot fully reproduce the complex hepatic dose distribution encountered in clinical breast radiotherapy. Future studies should include longer follow-up, direct assessment of collagen deposition, targeted metabolite validation, functional pathway experiments, and more clinically representative irradiation geometries. Conclusions In summary, FLASH-RT (625 Gy/s) provided hepatic sparing while maintaining comparable tumor control in a preclinical 4T1 breast cancer model. FLASH-RT attenuated radiation-associated oxidative stress, inflammatory activation, hepatocellular apoptosis, and early pro-fibrotic signaling. Non-targeted liver metabolomics suggested that FLASH-RT was associated with preservation of lipid- and membrane-associated metabolic profiles. These exploratory findings provide a hypothesis-generating metabolic perspective on FLASH-mediated normal-tissue sparing. Long-term functional studies and targeted mechanistic validation are required to support clinical translation. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1: Additional file 1. File format: .pdf. Title: Supplementary Figure S1 and Supplementary Tables S1–S2. Description: This file contains the VIP-threshold metabolomic sensitivity analysis, including the VIP > 1.5 retained metabolite list and the KEGG top 10 pathway comparison between the primary and conservative analyses. Acknowledgements The authors would like to express their sincere gratitude to the Department of Engineering Physics at Tsinghua University for providing the irradiation equipment and technical support essential for UHDR execution, as well as to Nanfang Hospital for providing the necessary experimental facilities and resources. Abbreviations Alkaline phosphatase Alanine aminotransferase Aspartate aminotransferase Breast cancer Blood urea nitrogen Cholesterol Conventional radiotherapy Creatinine Diaminobenzidine Damage-associated molecular patterns Dose-modifying factor FLASH radiotherapy Glutathione peroxidase Hematoxylin and eosin High mobility group box 1 Hepatic stellate cells Immunofluorescence Immunohistochemistry Internal mammary chain Kyoto Encyclopedia of Genes and Genomes Malondialdehyde Principal component analysis Phosphate-buffered saline Partial least squares-discriminant analysis Polymethyl methacrylate Radiation-induced liver disease Radiation-induced hepatic injury Radiation-induced liver injury Reactive oxygen species Radiotherapy Standard errors of the means Superoxide dismutase Total bilirubin Triglycerides Terminal deoxynucleotidyl transferase dUTP nick end labeling Uric acid Ultra-high dose rates Ultra-performance liquid chromatography-tandem mass spectrometry Variable importance in projection Author contributions XL and DX contributed equally to this work and are considered co-first authors. XL was responsible for conceptualization, methodology, investigation, data curation, visualization, and writing the original draft. DX contributed to methodology, data curation, resources, funding acquisition, and writing – review & editing. LC performed methodology, formal analysis, validation, and data curation. HS was involved in methodology, resources, and data curation. LZ provided conceptualization, supervision, project administration, funding acquisition, and writing – review & editing. All authors read and approved the final manuscript and agreed to be personally accountable for their own contributions. Funding This work was supported by grants from the National Key Research and Development Program of China (No. K120303106 to LHZ) and the National Natural Science Foundation of China (No. 82404187 to DHX), the Provincial Natural Science Foundation of Guangdong, China (No. 2025A1515012682 to DHX). Data availability The datasets generated and analyzed during the current study, including the dosimetric measurements (Table 1) and metabolomic profiles, are available from the corresponding author on reasonable request. All key findings supporting the conclusions of this article are included within the manuscript and its supplementary information files. Declarations Ethics approval and consent to participate All animal procedures and experimental protocols in this study were reviewed and approved by the Animal Management Committee of Southern Medical University (Approval No. L2023039). All experiments were conducted in strict compliance with the institutional ethical guidelines and the relevant regulations for the care and use of laboratory animals. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Xingyu Lu and Dehuan Xie contributed equally to this work. References Associated Data Supplementary Materials Supplementary Material 1: Additional file 1. File format: .pdf. Title: Supplementary Figure S1 and Supplementary Tables S1–S2. Description: This file contains the VIP-threshold metabolomic sensitivity analysis, including the VIP > 1.5 retained metabolite list and the KEGG top 10 pathway comparison between the primary and conservative analyses. Data Availability Statement The datasets generated and analyzed during the current study, including the dosimetric measurements (Table 1) and metabolomic profiles, are available from the corresponding author on reasonable request. All key findings supporting the conclusions of this article are included within the manuscript and its supplementary information files.