Preliminary mechanistic study of mitochondrial function in intestinal protection mediated by high-energy X-ray FLASH radiotherapy

Abstract

Purpose

Ultra-high dose rate (UHDR) radiation retains tumor-killing efficacy while mitigating toxicity to normal tissues, holding a promising transformative radiotherapy paradigm. This study aimed to explore the potential role of mitochondria in intestinal protection conferred by high-energy X-ray FLASH radiotherapy (FLASH-RT) and the associated signaling pathways.

Method

Healthy female C57BL/6 mice were subjected to whole-abdominal irradiation using three modalities: ultra-high dose rate radiotherapy (FLASH-RT), conventional dose rate radiotherapy (CONV-RT), and sham irradiation (Control). Mouse survival status and body weight changes were monitored within 15 days post-irradiation. At 72 h post-irradiation, whole blood samples were collected for hematological analysis, and intestinal tissues were harvested for pathological detection, transmission electron microscopy (TEM)-based observation of mitochondrial alterations, and two types of mitochondria-targeted metabolomic assays.

Results

A Compact single High-energy X-ray Source FLASH-RT device (CHEX-FLASH) was used, with a dose rate of 200 Gy/s. At 15 days post-irradiation, the survival rates of the Control group (100%, 10/10) and FLASH-RT group (80%, 8/10) were significantly higher than those of the CONV-RT group (30%, 3/10). Body weight decreased in the early post-irradiation period in all groups; however, the decline was milder in the FLASH-RT group with greater late-stage recovery. Hematological results at 72 h showed that CONV-RT induced similar marrow suppression compared to FLASH-RT. Intestinal histopathological analysis revealed that FLASH-RT alleviated intestinal inflammation and promoted enterocyte proliferation, whereas DNA double-strand breaks and apoptosis levels did not differ significantly between the two irradiated groups. FLASH-RT mitigated mitochondrial damage, reduced reactive oxygen species (ROS) levels and slightly activated mitophagy. Mitochondria-related energy metabolomics detection of intestinal tissues showed that the mitochondrial damage marker malonic acid was significantly lower in FLASH-RT than in CONV-RT, and differentially expressed metabolites were primarily enriched in mitochondrial antioxidant pathways. Additionally, upregulated expression of the antioxidant protein nuclear factor erythroid 2-related factor 2 (NRF2) and decreased total superoxide dismutase (SOD) activity were verified.

Conclusion

CHEX-FLASH achieves UHDR irradiation and alleviates radiation-induced intestinal injury. The protective effect of FLASH-RT on intestinal tissues may be mediated by mitigating mitochondrial damage and enhancing antioxidant pathways through the improvement of mitochondrial energy metabolism.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13014-026-02809-w.

Introduction

The global incidence of cancer continues to rise, and it has become one of the main causes of human death, posing a serious threat to human health [1]. Concurrently, radiation-induced toxicity has emerged as a worldwide concern. Radiation enteritis—an inflammatory disorder that arises when ionizing radiation injures the intestinal mucosa and adjacent tissues during treatment of pelvic, abdominal, or retroperitoneal malignancies (e.g., cervical, rectal, or prostate cancer), and it is one of the most frequent gastrointestinal complications of radiotherapy [2–4]. Achieving optimal tumor control while minimizing intestinal toxicity has therefore become a central objective of contemporary radiotherapy.

Ultra-high dose rate radiotherapy (UHDR, ≥ 40 Gy/s), also known as FLASH radiotherapy (FLASH-RT), has attracted extensive attention in recent years due to its potential to retain tumor therapeutic effects while sparing normal tissues [5]. If its efficacy and safety are validated in future clinical trials, FLASH-RT could have a transformative impact on radiotherapy practice. Although several mechanistic hypotheses have been proposed, such as oxygen consumption, DNA damage repair, free radical-free radical interaction, and immune mediation [6–8], there are still conflicting positive and negative research results, and the underlying biological mechanisms remains incompletely understood.

Historically, the biological effects of ionizing radiation were attributed primarily to nuclear DNA damage and its subsequent repair. However, current molecular biology studies have shown that the results of ionizing radiation exposure are highly dependent on the activation and regulation of other molecular components of organelles, which determine the cell’s survival and proliferation ability. For example, mitochondria, as the “energy factories” of eukaryotic cells, play a particularly critical role in regulating cellular metabolism, energy supply, and radiation-induced signal transduction [9].

Surprisingly, mitochondria have recently been implicated in the normal-tissue protection conferred by FLASH-RT. Guo Z et al. [10] exposed normal human lung fibroblasts (IMR90) to 100 Gy/s proton irradiation and observed significant attenuation of mitochondrial injury—manifest as preserved mitochondrial membrane potential (MMP), restored mitochondrial DNA (mtDNA) copy number and reduced reactive oxygen species (ROS)—whereas malignant cells showed no such sparing. Of particular importance, the observed reduction in ROS levels suggests that the endogenous antioxidant defense system may be enhanced. Superoxide dismutase (SOD), as the primary enzyme responsible for scavenging superoxide anion radicals, specifically catalyzes the dismutation of superoxide anions (O₂⁻) into the less toxic hydrogen peroxide (H₂O₂), thereby preventing the initiation of oxidative damage cascades and serving as a key indicator of cellular antioxidant capacity [11–13]. However, whether the unique tissue-sparing effect of FLASH-RT involves a specific upregulation of antioxidant molecules remains to be validated in in vivo models. In addition, mitochondria represent the central hub of cellular metabolism, and radiation-induced mitochondrial dysfunction may lead to alterations in cellular metabolic processes. Therefore, by targeting mitochondria and detecting relevant metabolites, we aim to systematically evaluate whether the protective effects of FLASH-RT are associated with enhanced mitochondrial antioxidant function. These findings, together with the detection approaches employed in the present study, will open new avenues for elucidating the mechanistic basis of the unique protective effects of FLASH-RT.

In recent years, pre-clinical studies have consistently demonstrated FLASH-RT-mediated sparing of normal brain [14], lung [15], intestine [16, 17], skin [18] and bone [19]. The scope of validation has been extended to large mammals [20], and exploratory clinical attempts have also been reported in single-patient case studies [21]. Among the available beam modalities—electrons, protons and high-energy photons—electrons are restricted to superficial targets by their limited penetration; protons incur high facility costs; whereas high-energy X-rays combine deep penetration with minimal beam divergence and favourable economics, positioning them as the most clinically scalable option for FLASH-RT.

However, high-energy X-ray sources capable of achieving UHDR remain scarce worldwide. In 2021, our research group first reported multi-organ protective effects of high-energy X-ray FLASH irradiation using the PARTER platform—a dedicated preclinical facility—installed on the terahertz free-electron laser (CTFEL) at the China Academy of Engineering Physics [15]. Yang Y et al. [22] characterized the beam delivery and dosimetric performance of the PARTER source and demonstrated that it meets the requirements for pre-clinical FLASH studies. Owing to its large footprint and high capital cost, however, PARTER is not suitable for clinical deployment. Thus we developed a compact single high-energy x-ray source FLASH-RT device (CHEX-FLASH, Zhongjiu Flash Medical Technology Co., Ltd.). The device employs a room-temperature radio-frequency linear accelerator: an S-band accelerator to produce 10 MeV electrons that impinge on a rotating tungsten target to generate megavoltage UHDR X rays [23]. The CHEX-FLASH device is highly compact and can be retrofitted into existing medical linac bunkers without structural modification, enabling better adaptation to clinical practice. In 2025, Lin B et al. [24] confirmed that this device elicits protective effects in lung, intestine, and skin while preserving robust anti-tumor efficacy. Based on this equipment, we verified the protective effect of FLASH-RT on normal intestinal tissues in mice, and innovatively conducted an in vivo study to explore the underlying mitochondrial mechanisms, providing mechanistic insights for more comprehensive FLASH-RT research.

Methods and materials

Irradiation device and dosimetry monitoring

All irradiation experiments were performed using the 10 MV CHEX-FLASH device (Zhongjiu Flash Medical Technology Co., Ltd.); detailed parameters are provided in Supplementary Table 1. Two dose-rate modalities were compared on the same platform: UHDR (FLASH-RT, ≥ 40 Gy s⁻¹) and conventional dose-rate (CONV-RT, 0.185 Gy/s). We used a beam current transformer (FCT-082-5.0) to monitor the beam current, and a diamond detector (CIVIDEC B3) installed downstream of the primary collimator to monitor the X-ray beam. A 2 cm thick acrylic plate was integrated into the flattening filter assembly of the experimental equipment to facilitate dose build-up. Gafchromic™ EBT-XD radiographic films (Ashland Inc., Covington, Kentucky, USA) were placed beneath solid water at the central level of the irradiation target area to ensure uniform absolute dose, as previously described [24]. Figure 1a depicts the in vivo irradiation geometry; the experimental set-up is shown in Fig. 1b.

Fig. 1.

The experimental setup diagrams, parameters, and experimental design for FLASH radiotherapy (FLASH-RT) and conventional dose-rate radiotherapy (CONV-RT). (a) Schematic diagram of the experimental setup. (b) Actual in vivo irradiation scenario: the mouse was positioned on a polymethyl methacrylate (PMMA) plate; a 1.0-cm-thick water-equivalent bolus was placed on its dorsum for dose build-up, and an EBT-XD film was interposed between the compensator and the mouse’s dorsal surface. (c) The irradiation field was a full abdominal field of 4 cm × 4 cm, with the upper boundary set below the xiphoid process of the mouse and the lower boundary at the upper end of the anus. (d) EBT-XD film was used to assess the dose distribution. (e) EBT-XD film was applied to evaluate the percentage depth dose (PDD). (f) Horizontal and vertical dose curves, representing the dose distribution. (g) PDD curve for X-ray sources, the dose was evaluated at a water-equivalent depth of 8 mm. (h) Schematic diagram of the research design

After being anesthetized with isoflurane gas (3–4%, Isoflurane, R510-22-10), the mice were placed prone on a 0.5 cm custom polymethyl methacrylate (PMMA) platform; a 1 cm water-equivalent bolus was applied dorsally for dose build-up. The RGB values of the EBT-XD films were scanned 5 min post-irradiation using an Epson Expression 12000XL scanner (Seiko Epson Corporation, Japan), and analyzed using SNC patient software (Sun Nuclear Corporation, USA). Additionally, before the FLASH experiment, the EBT-XD films were also calibrated using a clinically applied 6 MV Elekta Precise linac (Elekta AB, Stockholm, Sweden). After the dose calibration was verified to be acceptable, mice were subjected to irradiation experiments. The films were rescanned and read 24 h after irradiation to confirm the accuracy of the delivered radiation dose and to be used for subsequent dosimetric evaluation and analysis.

Animal model and euthanasia method

Female C57BL/6 mice aged 6–8 weeks were purchased from SPF (Beijing) Biotechnology Co., Ltd. (Beijing, China) and housed at Mianyang Central Hospital (Mianyang, China). All animal experiments were conducted in strict accordance with ethical standards and approved by the Animal Ethics Committee of Mianyang Central Hospital (approval number: S20240203-02).

Prior to euthanasia, the experimental animals were anesthetized with isoflurane (R510-22-10) gas (Shenzhen RWD Life Science Co., Ltd.). This approach effectively prevents animals from experiencing pain, suffocation, or stress responses during euthanasia. Furthermore, isoflurane undergoes rapid metabolism, which minimizes interference of the anesthetic with subsequent experimental test results. During the anesthetic induction phase, the initial concentration of isoflurane was set at 4%–5%. Adequate anesthetic depth was confirmed when the animals exhibited loss of corneal reflex and cessation of limb movements. Following this, procedures including irradiation and blood collection were performed, and euthanasia was ultimately achieved via cervical dislocation. The criteria for confirming death were loss of consciousness, respiratory arrest, and cardiac arrest.

Whole-abdominal irradiation

In this study, mice were subjected to three treatments: FLASH-RT (12 Gy/1F), CONV-RT (12 Gy/1F), or Sham (0 Gy, Control). Irradiations were delivered to a 4 cm × 4 cm abdominal field extending from the xiphoid process to the upper anal margin (Fig. 1c). In addition, the irradiation dose distribution and percent depth dose (PDD) curves are shown in Fig. 1d-g. Fig. 1h presents the study design: at 72 h after irradiation, whole blood and small intestinal tissues were collected (n = 5), while the remaining mice (n = 10) were observed twice weekly for survival and body weight changes during the acute phase (15 days). The endpoints were defined as body weight loss > 20–25%, severe deterioration in general health, or death; mice reaching any of these endpoints, as well as those completing the scheduled observation period or terminal tissue collection, were immediately euthanised.

Whole blood count detection

At 72 h after irradiation, mice were anesthetized with isoflurane gas, and the peripheral blood from the eyeballs was collected into EDTA-coated capillary tubes. A URIT-5160Vet auto hematology analyzer (URIT, Guilin, China) was used to determine the white blood cell count (WBC), hemoglobin (HGB), platelet count (PLT), lymphocyte count (LYM), neutrophil percentage (NEU%), and eosinophil percentage (EOS%).

Hematoxylin-eosin staining (H&E staining)

Following euthanasia, the abdominal cavity was opened and the intestine photographed in situ. The small bowel was then excised, rinsed in ice-cold phosphate-buffered saline (PBS), and fixed overnight in 4% universal fixative (Biosharp BL539A). Paraffin-embedded tissues were sectioned at 4–5 μm and stained with H&E. Subsequently, digital images were acquired with a pathological section scanner (NanoZoomer S360, C13220-01, Hamamatsu Photonics K.K., Japan). Acute radiation-induced intestinal injury was quantified based on the number of intestinal crypts per unit area [14], villus height, and intestinal injury pathological score [14]. The semi-quantitative pathological score was determined according to the proportion of the damaged intestinal segment to the total length: 0 = normal; 1 = < 1/3 segment affected; 2 = 1/3–2/3 affected; 3 = > 2/3 affected).

Immunohistochemical staining (IHC) and image analysis

Following deparaffinization and rehydration, antigen retrieval was performed and sections were incubated with primary antibodies. After washing, species-appropriate secondary antibodies were applied, chromogenic or fluorescent detection was carried out, and nuclei were counterstained. Sections were dehydrated, cleared and coverslipped before digital imaging with either a NanoZoomer S360 bright-field scanner (Hamamatsu Photonics, Japan) or a Pannoramic SCAN II fluorescence scanner (3DHISTECH, Hungary). Among them, Ki-67 (CST, CST9129) was used for proliferation; γH2AX (Sigma-Aldrich, SAB5600038) for double-strand breaks; cleaved caspase-3 (Affinity, AF7022) and Tunel (BrightGreen Apoptosis Detection kit, A112) for apoptosis; Parkin (Abmart, T56641S), PINK1 (Abmart, TD7742S) and LC3 (ABclonal, A15591) for mitophagy; NRF2 (Proteintech, 80593-1-RR) for oxidative stress response.

For each mouse, three non-overlapping fields of view were randomly selected (avoiding edge areas or regions with artificial damage), and quantification of positively stained areas was performed using ImageJ (version 1.54f). The specific analysis workflow was as follows: images were first subjected to color deconvolution (HDAB) to obtain an independent DAB grayscale image. An automatic thresholding algorithm (Otsu method) was uniformly applied to all analyzed images for initial segmentation. Regions confirmed to be devoid of positive staining (the submucosal layer) were defined as background ROIs, and their mean gray values were obtained. The images to be analyzed were then converted to 8-bit grayscale, and a threshold range was set such that the positively stained regions were completely selected while the background regions showed no obvious staining; this threshold was applied consistently to all analyzed images. The software automatically calculated the percentage of the total area occupied by pixels above the threshold relative to the entire region of interest for each section, defined as “% Positive Area.” The final data represent the average value from three sections.

Transmission electron microscopy (TEM)

Fresh small-intestinal samples (~ 1–3 mm³) were excised and immersed within 1 min in ice-cold 2.5% glutaraldehyde fixative (Yuanye Bio-Technology, Shanghai, R20510). After overnight fixation at 4 °C, tissues were post-fixed in 0.1 M phosphate buffer, dehydrated in graded ethanol, infiltrated with resin and polymerised at 60 °C for 48 h. Ultra-thin sections (60–80 nm) were cut, stained with uranyl acetate and lead citrate, and examined using a TEM (Hitachi, HT7800).

Measurement of MMP, ROS and total SOD activity

The MMP, ROS and total SOD activity levels of fresh small-intestinal samples were detected using the Mitochondrial Membrane Potential Assay Kit with JC-1 (Beyotime, C2006), the Tissue ROS Test Kit (DHE) (Bjbalb, HR8821), and the total SOD activity Assay Kit (Solarbio, BC0170). A multimode microplate reader (BioTek, Synergy LX, 23101208) was used.

When the mitochondrial membrane potential is high, JC-1 accumulates in the mitochondrial matrix to form polymers (J-aggregates) that emit red fluorescence; when the mitochondrial membrane potential is low, JC-1 cannot accumulate in the mitochondrial matrix and exists as monomers, which emit green fluorescence. The level of ROS in tissues can be determined by detecting the fluorescence of DHE products. SOD is a metalloenzyme widely present in organisms, serving as an important scavenger of oxygen free radicals. It can catalyze the dismutation of superoxide anions to generate H₂O₂ and O₂. SOD is not only a superoxide anion-scavenging enzyme but also a major H₂O₂-generating enzyme, playing a crucial role in the biological antioxidant system.

Mitochondria-targeted metabolomics (P650)

Small-intestinal samples were rinsed in PBS, snap-frozen in liquid nitrogen, and analyzed by targeted mitochondrial metabolomics (P650 panel, Panomix Biomedical Technology, Suzhou). The panel quantifies > 600 metabolites, including carbohydrates, organic acids, amino acids, bile acids, indoles, purine nucleotides and lipids. Frozen tissues were thawed at 4 °C, mixed with cold methanol/acetonitrile/water (2:2:1, v/v/v), vortexed, sonicated on ice for 30 min, held at − 20 °C for 10 min, and centrifuged (14 000 g, 4 °C, 20 min). Supernatants were vacuum-dried and reconstituted in 100 µL acetonitrile/water (1:1, v/v), vortexed and re-centrifuged (14 000 g, 4 °C, 15 min). Metabolites were separated on an ultra-high-performance liquid chromatograph coupled to a triple-quadrupole mass spectrometer (UHPLC-QTRAP MS). Peak areas were extracted with MultiQuant/Analyst; concentrations were calculated against internal-standard calibration curves.

Mitochondria-targeted energy metabolomics (Kit 100)

The Kit 100 panel quantifies 100 mitochondrial energy metabolites, covering glycolysis (EMP), the pentose phosphate pathway (PPP), the tricarboxylic acid (TCA) cycle, and selected intermediates from fatty-acid oxidation and amino-acid transamination. Intestinal tissue was homogenized in extraction buffer, freeze-ground and centrifuged; the supernatant was diluted four-fold, spiked with internal standards, vortex-mixed and dried. Samples were reconstituted in either 50% acetonitrile/water (alkaline loading) or 0.1% formic acid/40% methanol/water (acidic loading), vortexed and centrifuged (12000 rpm, 4 °C, 10 min). The resulting supernatants were transferred to LC-MS vials. Data were acquired on a SCIEX OS platform and processed for qualitative and quantitative analysis.

Statistical analysis

Statistical analyses were performed using GraphPad Prism version 8.4.0 (GraphPad Software Inc., La Jolla, CA, USA). All data are presented as the mean ± standard error of the mean (SEM). Survival analysis was conducted using Kaplan–Meier curves, and differences between groups were evaluated using the Mantel–Cox (log-rank) test. For metabolite profiling data, after sum normalization, the processed data were imported into SIMCA-P software (version 14.1, Umetrics, Umeå, Sweden) for multivariate data analysis. Metabolites with more than 50% missing values across samples were excluded. Principal component analysis (PCA), partial least squares discriminant analysis (PLS-DA), and orthogonal partial least squares discriminant analysis (OPLS-DA) were performed using the R software package ropls [25] for dimensionality reduction. Seven-fold cross-validation and response permutation testing were applied to assess the robustness of the models. Variable importance in projection (VIP) values derived from the OPLS-DA model were calculated to indicate the contribution of each variable to group discrimination. Pathway annotation was performed using the Kyoto Encyclopedia of Genes and Genomes (KEGG) Mapper tool. For comparisons of metabolite levels between two groups, unpaired two-tailed Student’s t test or the Mann–Whitney–Wilcoxon test was applied, as appropriate. Differential metabolites and pathway enrichment analyses were subjected to Benjamini–Hochberg false discovery rate (FDR) correction. Comparisons among multiple groups were performed using one-way analysis of variance (ANOVA) with Bonferroni multiple-comparison correction. Adjusted p values (adj. p) < 0.05 were considered statistically significant. The significance levels are denoted as follows: “*” denotes p < 0.05; “**” denotes p < 0.01; “ ***” denotes p < 0.001; “****” denotes p < 0.0001; and “ns” denotes not significant.

Results

Dosimetry

Irradiation parameters are listed in Supplementary Tables 1 and 2. The CHEX-FLASH system delivered 200 Gy/s for FLASH-RT and 0.185 Gy/s for CONV-RT. Because both modalities used the same linac, the prescribed dose was identical (FLASH-RT: 12.01 ± 0.12 Gy; CONV-RT: 12.03 ± 0.12 Gy). Lateral and longitudinal dose profiles measured with EBT-XD films were uniform and symmetric (Fig. 1f). The PDD curve is presented in Fig. 1g, where the blue box represents the dose distribution within the range of the abdominal depth of the mice.

FLASH-RT enhances survival rates and mitigates myelosuppression

To compare the toxicity and mortality under a whole abdominal irradiation dose of 12 Gy, mice were subjected to irradiation in the FLASH-RT group, CONV-RT group, and Control group. Peripheral blood cell tests were performed at 3 days post-irradiation (n = 5), and survival rate and body weight changes within 15 days were evaluated (n = 10). The survival rates of the Control group, FLASH-RT group, and CONV-RT group were 100% (10/10), 80% (8/10), and 30% (3/10) respectively, with significant differences among the three groups (Fig. 2a, P = 0.0005); no statistically significant difference was observed between the FLASH-RT group and the control group (P = 0.1672). Standardized body weight of surviving mice rose steadily in Controls, declined transiently in both irradiated groups. Specifically, the FLASH-RT group decreased within 5 days after irradiation and then increased, whereas the body weight of the CONV-RT group decreased within 8 days after irradiation and then increased, with the increasing trend being less pronounced than that of the FLASH-RT group (Fig. 2b, P < 0.0001).

Fig. 2.

Survival status, body weight, and hematological results post-irradiation. (a) Survival curves within 15 days post-irradiation. (b) Standardized body weight change curves of all surviving mice in each group within 15 days after irradiation (left panel), and individual standardized body weight change curves of each mouse in each group (right panel). (c) Results of blood routine tests at 72 h post-irradiation. (d) Representative in vivo photographs of the full abdomen of mice at 72 h post-irradiation, the black arrow indicates the suspected site of intestinal edema or dilation. “*” denotes p < 0.05; “ **” denotes p < 0.01; “ ***” denotes p < 0.001; “****” denotes p < 0.0001; and “ns” denotes not significant

Blood analysis (Fig. 2c) revealed marked decreases in WBC, LYM and HGB in both irradiated groups. NEU % and EOS % were higher in CONV-RT than in FLASH-RT mice (P < 0.05). PLT levels did not differ among groups (P = 0.1001).

FLASH-RT alleviates radiation-induced intestinal injury

At 72 h post-irradiation, we dissected the abdominal cavity of mice in each group (Fig. 2d). Macroscopic observation showed that, compared with the Control group and the FLASH-RT group, the CONV-RT group appeared to exhibit a broader extent of mild intestinal edema and dilatation. Further H&E staining of intestinal sections (Fig. 3a) revealed that the CONV-RT group had more extensive intestinal structural disorder, villous atrophy, and epithelial damage compared with the FLASH-RT group. Quantitative histology (Fig. 3b) confirmed that villus height in FLASH-RT (335.99 ± 45.81 μm) matched that of Controls (343.62 ± 36.20 μm) and exceeded the CONV-RT value (P = 0.0037). Crypt density (per 50 μm) was highest in Controls, intermediate in FLASH-RT and lowest in CONV-RT (P < 0.0001), while the intestinal injury score was correspondingly lowest in FLASH-RT (P < 0.0001). IHC demonstrated that proliferative activity (Ki-67) in FLASH-RT was comparable to Controls and significantly higher than in CONV-RT (P < 0.01; Fig. 3c, d). γH2AX staining indicated irradiation-induced DNA damage in both irradiated groups, with no difference between them (P > 0.9999; Figs. 3e, f). The results of one-way analysis of variance showed that the overall mean levels of cleaved caspase-3 (P < 0.05) and Tunel (P < 0.05) signals differed significantly among the three groups. However, after applying the Bonferroni correction for three pairwise comparisons (adjusted significance level α = 0.05/3 = 0.0167), the mean cleaved caspase-3 signal was statistically significantly between the Control group and the CONV-RT group (P = 0.0162), whereas no significant differences were observed between the Control group and the FLASH-RT group (P = 0.0263) or between the FLASH-RT group and the CONV-RT group (P = 0.2080). In contrast, no statistically significant differences in Tunel signals were observed in any pairwise comparisons among the groups (P > 0.9999, P = 0.0395, and P = 0.0720). Overall, the extent of cellular apoptosis following high-energy X-ray irradiation appeared to be relatively low (Figs. 3g-j).

Fig. 3.

Pathological staining results of intestinal tissues at 72 h post-irradiation. (a) Representative hematoxylin-eosin staining（H&E staining）images (scale bars, 100×, 250 μm; 300×, 50 μm). (b) Comparative analysis of intestinal villus length, number of intestinal crypts, and intestinal tissue damage scores. (c) Representative immunohistochemical (IHC) stained Ki-67 image (scale bars, 200×, 100 μm). (d) Positive Ki-67 area analysis. (e) Representative IHC stained γH2AX image (scale bars, 200×, 100 μm). (f) Positive γH2AX area analysis. (g) Representative IHC stained cleaved-caspase3 (c-caspase3) image (scale bars, 200×, 100 μm).(h) Positive c-caspase3 area analysis. (i) Representative IHC stained Tunel image (scale bars, 200×, 50 μm). (j) Positive Tunel area analysis. “*” denotes p < 0.05; “**” denotes p < 0.01; “***” denotes p < 0.001; “****” denotes p < 0.0001; and “ns” denotes not significant

FLASH-RT mitigates mitochondrial damage and reduces ROS levels in intestinal tissues

We observed the microstructure of intestinal tissues in the three groups at 72 h post-irradiation using TEM (Fig. 4a). The results showed that in the CONV-RT group, mitochondria were extensively abnormally shaped, with mild swelling, a dissolved and pale matrix, dilated cristae or vacuolar changes, and local damage to the membrane structure. In contrast, mitochondria in the FLASH-RT group were similar to those in the Control group, with no obvious abnormal changes. The mitochondrial membrane potential in the CONV-RT group was significantly lower than that in the FLASH-RT group and the Control group (Fig. 4b). Additionally, CONV-RT induced more ROS, while there was no significant difference between the FLASH-RT group and the Control group (P = 0.5806, Fig. 4c). The levels of Parkin, PINK1, and LC3 in the FLASH-RT group were lower than those in the CONV-RT group (P < 0.05). The levels of Parkin and PINK1 in the FLASH-RT group were higher than those in the Control group, while there was no significant difference in LC3 levels (Fig. 4d and e). These findings indicate that both FLASH-RT and CONV-RT can activate mitophagy, but this activation was more pronounced for the latter.

Fig. 4.

Detection of mitochondrial-related indices in intestinal tissues post-irradiation. (a) Representative transmission electron microscopy (TEM) images of intestinal tissue mitochondria (scale bars, 100000×, 1 μm). (b) Quantitative analysis of mitochondrial membrane potential (MMP) levels in intestinal tissues. (c) Quantitative analysis of reactive oxygen species (ROS) levels in intestinal tissues. (d) Representative IHC stained images of Parkin, PINK1, and LC3 (scale bars, 200×, 100 μm). (e) Positive area analysis of Parkin, PINK1, and LC3. “*” denotes p < 0.05; “**” denotes p < 0.01; “***” denotes p < 0.001; “****” denotes p < 0.0001; and “ns” denotes not significant

FLASH-RT may confer intestinal protection by enhancing antioxidant defense against oxidative damage

Given the differences in mitochondrial damage between FLASH-RT and CONV-RT, we performed mitochondria-targeted metabolomics. A total of 538 metabolites were identified in the P650 project; after exclusion of metabolites with missing values, 423 metabolites were included in the subsequent analyses, and their classification based on chemical taxonomy is detailed in Supplementary Fig. 1. PCA was conducted on all samples and quality control (QC) samples using the quantitative results of the identified metabolites. The experimental results (Fig. 5a) showed that the samples were closely clustered, indicating good reproducibility of the experiment. We compared the differential metabolites among the groups (Fig. 5b, P < 0.05) and found that there were 106 differential metabolites between the CONV-RT group and the Control group, and 106 differential metabolites between the FLASH-RT group and the Control group. In the comparison between the FLASH-RT group and the CONV-RT group, a total of 8 differentially expressed metabolites were identified, among which 6 were upregulated and 2 were downregulated. We intuitively displayed the changes of the significant differential metabolites identified in the two groups using a bar chart (Fig. 5c). Further hierarchical cluster analysis was performed on each group of samples to form a cluster tree showing the similarity between samples, and the results are shown in Fig. 5d, where samples clustered in the same group have higher similarity. The 6 upregulated differential metabolites in the FLASH-RT group were adenosine diphosphate ribose (ADP-ribose), aspartylphenylalanine, betaine, gamma-glutamyl-phenylalanine, glycyl-phenylalanine, and hypotaurine; the 2 downregulated differential metabolites were apocholic acid and tauro-α-muricholic acid (α-TMCA), with their quantitative analysis results shown in Fig. 5e. KEGG pathway enrichment analysis of the list of differential metabolites using MetaboAnalyst showed (Fig. 5f) that the FLASH-RT group was enriched in 2 pathways, namely glycine, serine and threonine metabolism, and taurine and hypotaurine metabolism, which are mainly related to the regulation of mitochondrial antioxidant function.

Fig. 5.

Mitochondria-targeted metabolism analysis of intestinal tissues post-irradiation (P650). (a) Principal Component Analysis (PCA) of the overall samples. (b) Volcano plot of differentially expressed energy metabolites between different groups. (c) Linear Discriminant Analysis Effect Size (LEfSe) showing substances with significant differences between the FLASH-RT and CONV-RT groups. (d) Hierarchical clustering heatmap of significantly different metabolites between the FLASH-RT and CONV-RT groups. (e) Quantitative analysis of substances with significant differences between the FLASH-RT and CONV-RT groups. (f) Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment pathway map (bubble chart) for the FLASH-RT vs. CONV-RT groups. “*” denotes p < 0.05; “**” denotes p < 0.01; “***” denotes p < 0.001; “****” denotes p < 0.0001; and “ns” denotes not significant

In addition, we detected 100 substances targeting mitochondrial energy metabolism. The PCA score plot of all samples showed obvious clustering among the groups, indicating good reproducibility of the analysis (Fig. 6a). Further OPLS-DA showed significant separation between the Control group, FLASH-RT group, and CONV-RT group (Fig. 6a). Figure 6b shows the Z-score plot of all metabolites; the closer to the right, the higher the relative content of the current metabolite in the sample, and the closer to the left, the lower the content. Quantitative analysis revealed that malonic acid in the CONV-RT group was significantly higher than that in the FLASH-RT group (Fig. 6c). KEGG pathway enrichment analysis of the metabolites was performed, and the network diagram results (Fig. 6d) showed that the FLASH-RT group was most enriched in 5 pathways, namely central carbon metabolism in cancer, protein digestion and absorption, aminoacyl-tRNA biosynthesis, glucagon signaling pathway, and alanine, aspartate and glutamate metabolism, which are also related to mitochondrial oxidative stress function.

Fig. 6.

Mitochondria-targeted energy metabolism analysis of intestinal tissues post-irradiation (Kit 100). (a) PCA and Orthogonal Partial Least Squares Discriminant Analysis (OPLS-DA) score plots for each group. (b) Z-score plot comparing the CONV-RT and Control groups. The X-axis represents the Z-score of the relative abundance of metabolites, and the Y-axis shows the metabolite names. The further to the right, the higher the relative abundance of the metabolite in the sample. (c) Box plot of malonic acid quantification in the FLASH-RT and CONV-RT groups. (d) Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment pathway map (network diagram) for the FLASH-RT vs. CONV-RT groups. Green dots represent pathways, and other dots represent metabolites. The size of the pathway dots indicates the number of connected metabolites (larger dots signify more connections), and the color of metabolite dots represents the magnitude of the log2FC value. (e) Representative IHC stained NRF2 image (scale bars, 200×, 100 μm). (f) Positive NRF2 area analysis. (g) Quantitative analysis of total superoxide dismutase (SOD) activity. “*” denotes p < 0.05; “**” denotes p < 0.01; “***” denotes p < 0.001; “****” denotes p < 0.0001; and “ns” denotes not significant

Because the differential metabolomics and KEGG enrichment indicated alterations in pathways related to oxidative stress handling, we next sought tissue-level evidence for activation of antioxidant defense. NRF2 is the “master regulator” of the cellular response to oxidative stress, controlling the expression of a series of antioxidant genes. Accordingly, NRF2 expression in intestinal tissues at 72 h post-irradiation was assessed by IHC. We found that the Control group had the highest level, and the FLASH-RT group was also significantly higher than the CONV-RT group (P < 0.01). In addition, the total SOD activity in intestinal tissues of the FLASH-RT group (4.17 ± 1.05 U/mg prot) was significantly lower than that of the CONV-RT group (8.17 ± 1.83 U/mg prot). These results suggest that FLASH-RT may activate mitochondrial antioxidant function to cope with radiation damage.

Discussion

Due to its biological function of protecting normal tissues, FLASH-RT has the potential to become a disruptive radiotherapy modality, but its underlying biological mechanisms remain to be explored. In this study, we utilized the CHEX-FLASH device, focusing on confirming the protective effect of FLASH-RT on normal intestinal tissues in mice. For the first time, we observed the phenomenon of alleviated intestinal mitochondrial damage, and further analyzed and verified the differential mitochondrial metabolites between the FLASH-RT and CONV-RT groups, revealing that FLASH-RT may achieve intestinal protection by enhancing mitochondrial antioxidant function.

We investigated mortality, body weight change curves, acute hematological toxicity, and intestinal damage following 12 Gy whole abdominal irradiation. The results of this study showed that compared with CONV-RT, FLASH-RT significantly improved survival and body weight gain, which is consistent with previous studies on intestinal FLASH irradiation with protons or electrons [16, 26]. We also observed that FLASH-RT could reduce myelosuppression, which has been reported in a study by Zhu H. et al. [27], who further found that CONV-RT significantly reduced WBC and LYM counts, with a more pronounced decrease at 48 h post-irradiation compared to 3 h. We found that NEU% increased significantly after CONV-RT, which may be due to the relative increase in the proportion of neutrophils caused by lymphopenia, or inflammatory responses induced by intestinal tissue damage after radiotherapy. Histopathological examination of intestinal tissues revealed that FLASH-RT alleviated post-radiation intestinal damage, as evidenced by lower intestinal injury scores, and significantly better intestinal villus height, number of intestinal crypts, and intestinal cell proliferation capacity compared to CONV-RT. This confirms that the CHEX-FLASH device exhibits a significant intestinal protective effect, consistent with the abdominal irradiation effects of other electrons, protons, or X-rays [16, 17, 28–30]. Valdés Zayas A. et al. [29] compared the reproducibility of the FLASH intestinal protection effect across different experimental platforms and concluded that FLASH-RT effectively mitigates intestinal damage, particularly by enhancing crypt-cell regeneration, but emphasized that achieving cross-platform reproducibility requires more standardized experimental designs and deeper mechanistic investigation. Although ≥ 40 Gy s⁻¹ is a commonly used threshold, recent negative results indicate that this value may be insufficient under all conditions. Cao W. et al. [31] evaluated heavy-ion UHDR irradiation for intestinal protection and found that when the dose rate exceeded 100 Gy s⁻¹, FLASH-RT significantly improved mouse survival under high linear energy transfer (LET), suggesting that both dose rate and LET are critical determinants of the FLASH effect. Reviews have emphasized that intestinal FLASH protection depends not only on ultra-high mean dose rate but also on multiple beam-delivery parameters, including single-pulse dose, pulse width, pulse frequency, and total irradiation time [24]; these factors may regulate free-radical kinetics and normal-tissue biological responses [32]. FLASH-RT-treated enteroids show enhanced colony-forming potential, possibly linked to up-regulation of WNT-family genes, cell-cell adhesion molecules, and hypoxia-response pathways [33]. Multi-omics analyses of high-energy X-ray FLASH radiotherapy in mouse intestine revealed associations with remodelling of the microbiome-metabolism axis, attenuated inflammatory responses, and increased intestinal stem-cell numbers [34]. It has also been reported that FLASH achieves intestinal protection by reducing cGAS-STING-mediated inflammatory pyroptosis while preserving anti-tumor immunity [30]. However, mechanistic investigation of FLASH-RT intestinal protection is still preliminary.

In addition, we found no statistically significant difference in γH2AX expression levels between FLASH-RT and CONV-RT (P > 0.9999), indicating that the degree of DNA double-strand damage was comparable between the two groups. However, since our detection time point was 72 h post-irradiation, it is possible that DNA damage may have been repaired. Zhu H. et al. [27] found that γH2AX levels were relatively high at 3 h post-irradiation and decreased at 48 h, but γH2AX levels in CONV-RT remained higher than those in FLASH-RT. However, comparative studies on DNA damage are inconsistent: Shi X. et al. [30] irradiated intestinal organoids and human intestinal epithelial cells (HIEC-6) with X-ray FLASH-RT and CONV-RT, and found that the expression levels of γH2AX and 53BP1 were almost identical in both groups. In contrast, other researchers observed that acute irradiation with electrons in human lung cells [35] and intestinal crypt stem cells [16] resulted in lower levels of 53BP1 or γH2AX in FLASH-RT compared to CONV-RT. Overall, discrepancies among studies regarding DNA damage may be related to differences in time points, dose-rate levels, and radiation sources. Moreover, substantial heterogeneity exists between in vitro cell line or organoid models and in vivo animal models, including differences in three-dimensional tissue architecture, in vivo oxygen tension, and the tissue microenvironment, such as the extracellular matrix and cell–cell interactions. In addition, with regard to the minimal apoptotic trend observed after irradiation at the two dose rates, emerging evidence indicates that radiotherapy can induce alternative forms of cell death, including ferroptosis [36, 37] and pyroptosis [30], among other regulated cell death modalities [38]. Therefore, the effects of FLASH-RT on cellular death mechanisms remain to be further elucidated.

Currently, researchers have proposed several hypotheses [6, 39], including oxygen depletion, free radical-free radical interaction, and immune factors, but there are conflicting positive and negative research results reported. Surprisingly, in recent years, a hypothesis has emerged that mitochondria are involved in the protective mechanism of FLASH-RT. A research team irradiated normal human lung fibroblasts (IMR90) with protons and found that FLASH-RT (100 Gy/s) could alleviate mitochondrial damage, including morphological and functional changes such as mitochondrial membrane potential, mtDNA copy number, and ROS [10], while this phenomenon was not observed in cancer cells. A pre-published article also reported that compared with low-dose-rate electron irradiation (0.36 Gy/s), FLASH electron irradiation at 61–610 Gy/s significantly enhanced cytochrome c release from mitochondria in human breast cancer cells MCF-10A, thereby triggering massive caspase activation and inhibiting cytoplasmic mtDNA accumulation and interferon-β (IFN-β) secretion [40]. Another in vivo study on the protective effect of electron FLASH-RT on the esophagus showed via TEM that CONV-RT induced mitochondrial swelling, blurred structure, disappearance or break of cristae, reduced matrix electron density, and vacuolation, while mitochondrial swelling and structural blurring were less obvious in the FLASH-RT group. Proteomic analysis revealed that CONV-RT downregulated the expression of key proteins involved in the TCA cycle and oxidative phosphorylation in mice [41].

In our study, we observed changes in mitochondrial structure following high-energy X-ray irradiation for the first time. TEM also revealed similar obvious characteristics of intestinal mitochondrial damage in the CONV-RT group, while mitochondrial damage in the FLASH-RT group was mild. Meanwhile, compared with FLASH-RT, CONV-RT disrupted mitochondrial membrane potential and increased ROS, confirming functional impairment. In addition, CONV-RT exhibited higher expression of mitophagy-related proteins Parkin and PINK1 than FLASH-RT, indicating that mitophagy is activated to remove dysfunctional mitochondria after mitochondrial damage [42]. Mitochondria are small double-membraned organelles in the cytoplasm of eukaryotic cells, coupled with the electron transport chain (ETC) and oxidative phosphorylation (OXPHOS) through the tricarboxylic acid (TCA) cycle, and are known as the “energy factories” of the cell.

Using targeted metabolomics, we performed qualitative and quantitative analysis of mitochondria-related metabolites and found that FLASH-RT expressed lower levels of malonic acid (an endogenous competitive inhibitor of succinate dehydrogenase), which is a marker of mitochondrial damage [43]. Malonate causes rapid collapse of MMP and ROS production, thereby overwhelming mitochondrial antioxidant capacity and leading to mitochondrial swelling [44].

Differential metabolite analysis revealed that FLASH-RT upregulated ADP-ribose which is an important product of nicotinamide adenine dinucleotide (NAD⁺) metabolism, and its generation is closely associated with the homeostasis of the mitochondrial NAD⁺ pool; NAD⁺ plays a critical role in regulating mitochondrial oxidative metabolism and protein deacetylation modifications [45]. Previous studies have shown that betaine can enhance mitochondrial respiratory chain function and increase cytochrome c oxidase activity, thereby elevating mitochondrial membrane potential and cellular energy levels [46]. It can also reduce mitochondrial superoxide levels, significantly alleviate MSS51-mediated inhibition of mitochondrial respiration, and improve muscle energy metabolism [47]. In addition, the synthesis of betaine depends on mitochondrial choline oxidative metabolism. Betaine can specifically bind to TBK1 and inhibit its kinase activity, thereby blocking activation of the downstream IRF3/NF-κB signaling pathway and attenuating inflammatory responses [48]. Studies have reported that taurine, as an antioxidant, protects mitochondria from oxidative stress–induced functional impairment, potentially through multiple mechanisms such as stabilizing intracellular GSH levels. Taurine derivatives can also activate the nuclear factor NRF2, a transcription factor that suppresses inflammatory responses by regulating the transcription of multiple antioxidant genes [49]. These findings suggest that FLASH-RT may maintain a more favorable balance between mitochondrial energy metabolism and oxidative stress. In the present study, phenylalanine-containing dipeptides and amino acid–related metabolites were upregulated in the FLASH-RT group. Although these metabolites are not classical mitochondrial intermediates, their constituent amino acids, such as aspartate, have been reported to alleviate oxidative stress by mediating RIP-dependent mitochondrial function [50]. Phenylalanine, under hypoxic conditions, can enhance mitochondrial function by improving mitochondrial oxygen consumption rates, reducing ROS levels, and activating signaling axes such as AMPK–PGC-1α [51], suggesting that FLASH-RT may induce amino acid metabolic alterations associated with adaptive changes in mitochondrial oxidative function.

In contrast, apocholic acid and α-TMCA were downregulated in the FLASH-RT group. Apocholic acid is an unsaturated bile acid, and bile acids may induce oxidative stress and mitochondrial damage through mechanisms such as disruption of cellular membranes and induction of ROS generation [52]. α-TMCA is a bile acid derivative that acts on farnesoid X receptor (FXR) metabolism as an antagonist [53, 54]. Previous studies have reported that FXR metabolism is associated with mitochondrial energy metabolism and oxidative stress pathways [54]. FXR agonists can modulate the NRF2 pathway to alleviate hepatic inflammation [55]. Under diabetic conditions, reduced FXR expression can regulate COX6A2, thereby enhancing mitochondria-dependent stress responses and cell death [56].

Through KEGG enrichment analysis, we identified two key pathways: “Glycine/serine/threonine metabolism” and “Taurine and hypotaurine metabolism.” From previous literature, “Glycine/serine/threonine metabolism” can influence carbohydrate patterns and energy utilization efficiency, ultimately delaying intervertebral disc degeneration through antioxidant effects [57], while the “serine/glycine pathway” can maintain cellular function, oxidative stress tolerance, and survival [58]. “Taurine and hypotaurine metabolism” can also protect mitochondria from oxidative and sub-oxidative stress [59, 60].

Based on the above metabolomics analyses, we therefore hypothesize that FLASH-RT may enhance mitochondrial antioxidant defense capacity. We further confirmed this hypothesis by immunohistochemical (IHC) analysis of intestinal tissues, showing that, compared with CONV-RT, FLASH-RT exhibited higher expression levels of NRF2 and reduced total SOD activity, which are antioxidant genes induced in response to ionizing radiation–induced oxidative stress [61, 62]. We speculate that these differential expression patterns may not be contradictory, and the underlying mechanisms may involve the following: FLASH-RT significantly attenuated ROS generation, potentially reducing the demand for radiation stress-induced SOD—the primary enzymatic defense against superoxide anion—to scavenge excessive free radicals, thereby possibly resulting in a passive reduction of total SOD activity. Conversely, the higher oxidative burden in CONV-RT may have driven stress-mediated upregulation of total SOD activity. Given that NRF2 orchestrates an extensive antioxidant gene network, including heme oxygenase-1 (HO-1) and glutathione synthesis-related enzymes [63], FLASH-RT-induced NRF2 activation may preferentially or more efficiently coordinate alternative antioxidant pathways. Furthermore, as an enzymatic activity indicator, total SOD activity may be susceptible to factors such as enzyme consumption, inactivation, and post-translational modifications, or may exhibit temporal and spatial delays in transcriptional regulation. The upregulation of NRF2 observed in this study is consistent with the metabolomics findings related to the “taurine metabolism” pathway and the observed effects of certain metabolites in promoting NRF2. This indicates that FLASH-RT does not merely passively reduce damage, but may actively trigger a cascade antioxidant response ranging from metabolic adaptation to transcriptional activation.

In summary, we propose the following hypothesis (Fig. 7): high-energy X-ray FLASH-RT can reduce ROS generation and attenuate mitochondrial damage, which is closely associated with enhanced mitochondrial antioxidant capacity. The potential mechanisms may involve activation of the NRF2 pathway, modulation of total SOD activity, and positive regulation of mitochondrial-related antioxidant metabolite metabolism. Our study may provide a valuable foundation for subsequent mechanistic investigations into causal relationships.

Fig. 7.

Research hypothesis diagram.

Of course, this study has some limitations. For example, the detection of ROS, myelosuppression, and γH2AX breakage indices lacks dynamic detection at multiple time points, and multi-factor exploration of FLASH-RT dose parameters such as dose rate and fractionation methods is not conducted. And analysis of tissue samples at a single time-point is also insufficient to fully interpret the dynamic processes of mitochondrial dynamics and inflammatory responses. Future studies should adopt a multi-time-point sampling strategy to dissect the temporal dynamics of these key biological processes, thereby enabling a more comprehensive elucidation of the FLASH protective mechanism. However, we propose that FLASH-RT achieves normal tissue protection by enhancing mitochondrial antioxidant function, which provides a favorable research basis for further mechanistic studies by researchers.

Conclusion

CHEX-FLASH achieves UHDR irradiation and alleviates radiation-induced intestinal injury. The protective effect of FLASH-RT on intestinal tissues may be mediated by mitigating mitochondrial damage and enhancing antioxidant pathways through improved mitochondrial energy metabolism.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

ADP-ribose

Adenosine diphosphate ribose

α-TMCA

Tauro-α-muricholic acid

ANOVA

Analysis of variance

CHEX-FLASH

Compact single high-energy X-ray source FLASH radiotherapy device

CONV-RT

Conventional dose-rate radiotherapy

CTFEL

Terahertz free-electron laser

EMP

Glycolysis

EOS%

Eosinophil percentage

ETC

Electron transport chain

FDR

False discovery rate

FLASH-RT

FLASH radiotherapy

FXR

Farnesoid X receptor

H&E

Hematoxylin-eosin

HGB

Hemoglobin

IFN-β

Interferon-β

IHC

Immunohistochemical staining

IMR90

Normal human lung fibroblasts

KEGG

Kyoto Encyclopedia of Genes and Genomes

LET

Linear energy transfer

LYM

Lymphocyte count

MMP

Mitochondrial membrane potential

mtDNA

Mitochondrial DNA

NAD⁺

Nicotinamide adenine dinucleotide

NEU%

Neutrophil percentage

OH-1

Heme oxygenase-1

OPLS-DA

Orthogonal partial least squares discriminant analysis

OXPHOS

Oxidative phosphorylation

PARTER

Preclinical accelerator for radiotherapy research

PBS

Phosphate-buffered saline

PCA

Principal component analysis

PDD

Percent depth dose

PLS-DA

Partial least squares discriminant analysis

PLT

Platelet count

PMMA

Polymethyl methacrylate

PPP

Pentose phosphate pathway

QC

Quality control

ROS

Reactive oxygen species

SEM

Standard error of the mean

SOD

Superoxide dismutase

TCA

Tricarboxylic acid

TEM

Transmission electron microscopy

UHDR

Ultra-high dose rate radiotherapy

UHPLC-QTRAP MS

Ultra-high-performance liquid chromatograph coupled to a triple-quadrupole mass spectrometer

VIP

Variable importance in projection

WBC

White blood cell count

Author contributions

X.H., H.D. and B.L. completed animal experiments and wrote the main manuscript text, H.Z., M.T., W.W., D.W. and Y.Z. participated in radiotherapy experiments and data collection and analysis. X.H., Y.Z. and Y.Y. prepared Figs. 1, 2, 3, 4, 5 and 6. X.H. and X.D. designed this experiment and revised the manuscript. All authors reviewed the manuscript.

Funding

This work was financially supported by the Projects of National Natural Science Foundation of China (Grant NO. U2330122) and the General Program of the Sichuan Natural Science Foundation (Grant No. 2023NSFSC0710).

Data availability

All supporting data are included as supplementary files with this manuscript. Additional raw data can be obtained from the corresponding author upon reasonable request.

Declarations

Ethical approval

All mouse experiments adhered to the approved guidelines of the Ethics Committee of Mianyang Central Hospital (approval number: S20240203-02).

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.