The Impacts of FLASH Radiation Therapy and Conventional Radiation Therapy on the Cognitive Abilities of Mice

Abstract

Purpose

Preclinical studies have demonstrated that FLASH radiation therapy (RT) delivered at ultrahigh-dose rates exerts a preferential normal tissue-sparing effect. This study aimed to delineate the disparities in delayed cognitive function and biological outcomes of whole-brain irradiation in healthy mice, comparing FLASH-RT with conventional RT (CONV-RT).

Methods and Materials

Eighty adult male C57BL/6J mice were divided into 5 groups: Sham, CONV-RT10Gy, CONV-RT20Gy, FLASH-RT10Gy, and FLASH-RT20Gy. Whole-brain irradiation was conducted on mice using a miniaturized x-ray FLASH platform at field-average dose rates of 2 Gy/min for CONV-RT and 213 Gy/s for FLASH-RT. Two months after irradiation, we assessed the mice’s cognitive function and the number of astrocytes and neurons in the hippocampus, then conducted proteomic analyses of their hippocampus.

Results

Following the administration of a 20 Gy FLASH-RT dose, the incidence of radiodermatitis was markedly lower than that observed with CONV-RT, accompanied by an improvement in survival rates. Compared with the Sham and FLASH-RT groups, the CONV-RT group exhibited reduced exploration of the open arms in the elevated plus maze, diminished preference for the novel arm in the Y-maze, a lower discrimination index in the novel object recognition test, and prolonged latency to reach the platform in the water maze test. Compared with the FLASH-RT group, the CONV-RT group exhibited an increase in astrocytes and a decrease in neurons in the hippocampus. Proteomic analysis revealed that FLASH-RT may improve the oxidative stress damage caused by CONV-RT.

Conclusions

This study demonstrated that FLASH-RT conferred significant advantages over CONV-RT in preserving delayed cognitive function and reducing radiation-induced toxicity in healthy mice. Compared with CONV-RT, FLASH-RT mitigated behavioral deficits across multiple cognitive domains and attenuated hippocampal oxidative stress, highlighting its neuroprotective potential. These findings provided compelling preclinical evidence supporting the therapeutic promise of FLASH-RT as a safer alternative to conventional RT for protecting normal brain function.

Introduction

The primary aim of radiation therapy (RT) is to administer tumoricidal doses to malignant tissues¹ while safeguarding nearby healthy structures and essential organs from excessive toxicity.² Nevertheless, conventional RT (CONV-RT) is frequently accompanied by either acute to subacute or chronic side effects that considerably diminish patient quality of life and hinder the ability to escalate doses for improved local tumor control.³

Data from the US Central Brain Tumor Registry reveal that during the period 2017-2021, the age-standardized incidence rate for all brain and central nervous system tumors—both malignant and benign—averaged 25.34 cases per 100,000 individuals.⁴ Furthermore, clinical data indicate that nearly 30% of individuals who received diagnosis for cancer experience brain metastasis,⁵ and over half of these patients are treated with whole-brain RT, which frequently leads to neurologic deficits such as memory loss,⁶ reduced attention span,⁷ and emotional disturbances.⁸ These adverse outcomes are primarily attributed to heightened neuroinflammation,⁹ diminished cerebral microvascular density,¹⁰ decreased dendritic spine density in neuron,¹¹ impaired myelin, and structural disruptions in glial cells,¹² culminating in both anatomic and functional impairments.¹³ Whole-brain RT undermines the integrity of the blood-brain barrier by altering the function of supportive cells,¹⁴ and impairing tight junction proteins.¹⁵ Therefore, there is a pressing need to develop innovative RT strategies to alleviate the cognitive impairments faced by patients with brain tumor.

In 2014, Favaudon et al¹⁶ reintroduced FLASH-RT. Research indicated that this approach dramatically curtailed radiation-induced harm in healthy tissues while still ensuring robust tumor suppression¹⁷—a benefit known as the “FLASH effect.”¹⁸ In contrast to CONV-RT, which typically operated at dose rates of 0.01 to 0.02 Gy/s, FLASH-RT administered doses exceeding 40 Gy/s and concluded the entire irradiation process in <500 ms.^19,20 Moreover, recent investigations focusing on the central nervous system suggested that x-ray-based FLASH-RT can substantially reduce acute hippocampal injury and the associated inflammatory response usually observed with CONV-RT.²¹ However, the exact biological mechanisms underpinning this protective effect—as well as the delayed negative impacts of radiation—remained insufficiently understood, warranting further research into the processes driving the “FLASH effect.”

Until recently, research on low linear energy transfer FLASH-RT primarily relied on electron beams.²² This reliance poses a considerable barrier for clinical translation because electrons exhibit limited tissue penetration.²³ Therefore, our work concentrated on using 10 MV x-rays delivered at 213 Gy/s, which offer distinct clinical advantages over FLASH electron beams. In our study, we performed whole-brain irradiation via a high-energy x-ray FLASH platform and compared the delayed cognitive outcomes in mice subjected to CONV-RT vs FLASH-RT.

Methods and Materials

Irradiation device

For CONV-RT, a 6 MV beam generated by a clinical Elekta Precise Linac (Elekta AB, 151535) from Zhongjiu Flash Medical Technology Co, Ltd, was employed at a mean dose rate of 2 Gy/min.

FLASH-RT was carried out using a miniaturized x-ray FLASH platform based on a room temperature RF Linac at the Zhongjiu Flash Medical Technology Co, Ltd, in China (Table 1). The electron beam, generated by a laser-driven cathode, was accelerated to 10 MV and then used to bombard a tungsten target to produce bremsstrahlung x-rays. The electron beam was monitored in real time using a fast current transformer (Bergoz, ict-082-012-05) installed upstream of the target. X-rays were monitored using a diamond detector (PTW, Ref. TW60025) installed at the entrance of the collimator. The signal was sent through a coaxial cable to an oscilloscope (Rigol, DS7034) for digitalization. This oscilloscope was capable of automatically capturing, storing, and integrating the transient pulses. The integrated value provided by the oscilloscope served as a monitor unit for measuring the accelerator output, and the beam current was precisely regulated by adjusting the driving laser. In this study, we attained an average dose rate of 213 Gy/s, and the total delivered dose was regulated by adjusting the macropulse duration.

Table 1.

Irradiation parameters

Delivery mode	Prescribed dose (Gy)	Radiation source	Energy (MV)	Field size (cm)	Mean dose rate (Gy/s)	Total delivery time (ms)	SSD (cm)
CONV-RT	10	x-rays	6	1.2 × 30	0.033	300,000	100
	20	x-rays	6	1.2 × 30	0.033	600,000	100
FLASH-RT	10	x-rays	10	1.2 × 2	213	47	43
	20	x-rays	10	1.2 × 2	213	91	43

Abbreviations: CONV-RT = conventional radiation therapy; FLASH-RT =FLASH radiation therapy; RT = radiation therapy; SSD = source-skin distance.

Brain irradiation setup

Irradiation was conducted under sodium pentobarbital anesthesia. To perform FLASH-RT, we positioned the mouse head directly behind and in contact with the aperture of a graphite applicator measuring 1.2 × 2 cm². Because the clinical Elekta Precise Linac for CONV-RT allows for the precise configuration of any radiation field and delivers a stable and uniform output, we set the radiation field to 1.2 mm × 30 cm to simultaneously irradiate 5 mice, thereby reducing the time required for the experiment. In addition, a scintillation detector is used to ensure the accuracy of the irradiated area. This arrangement allowed us to irradiate the entire encephalon region while limiting the dose to the eyes, mouth, and rest of the body. Mice received a single dose of 10 or 20 Gy in a dorsoventral beam direction.

Dosimetry

Dosimetry was performed as previously described²¹. The irradiation dosimetry system comprised both passive and active detectors. Absolute dose measurements were obtained using Gafchromic EBT3 radiochromic films (Ashland Inc, 10052201). Before experimental irradiation, the EBT3 films were calibrated with an Elekta Precise linear accelerator in conjunction with a diamond detector (PTW, Ref. TW60025). Twenty-four hours postexposure, the films were digitized with an Epson Expression 12000XL scanner (Seiko Epson Corporation), and optical density values were quantified through in-house algorithms developed in Matlab 2020 (MathWorks). Active dosimetry was performed using a pinpoint ionization chamber (PTW, Ref. TW30013). Comparative analyses were undertaken between the passive and active measurements under identical FLASH irradiation settings. For individual animal exposures, an EBT3 film was positioned between the subject and the polymethyl methacrylate holder to determine the delivered dose precisely.

Animals and experimental design

All procedures were performed using 8-week-old male C57BL/6J mice obtained from SiPeiFu Biotechnology in China. They were maintained under a regular 12-hour light/dark cycle at a controlled temperature of 23 to 25°C. In the whole-brain irradiation study, the animals were randomly allocated into 5 groups (n = 16): Sham, CONV-RT10Gy, CONV-RT20Gy, FLASH-RT10Gy, and FLASH-RT20Gy. If the body weight of the mice decreases by >20% or if the skin score reaches 5, a humane endpoint will be applied. A flowchart of the study protocol is shown in Figure 1A. Behavioral experiments commenced on the 50th day after irradiation, with all surviving animals undergoing assessments in the following sequence: elevated plus maze, Y-maze, novel object recognition, and water maze, 1 test per day to ensure adequate recovery time. All experimental procedures were conducted in accordance with a protocol approved by the ethics committee of the Academy of Beijing Institute of Basic Medical Sciences (approval number: IACUC-JSRZ-24-0801).

Figure 1.

Parameters and dosimetry for FLASH-RT and CONV-RT. (A) The detailed timeline of the experimental procedures. (B) The schematic diagram of the FLASH-RT experiment. (C) The schematic diagram of the CONV-RT experiment. (D) An EBT3 film was placed between the PMMA plate and the anterior surface of the irradiated mouse for dose evaluation. (E) Laser lines were applied for positioning in CONV-RT. (F) Percentage depth dose of FLASH-RT. (G) Pulse structure of the electron beam from a superconductor accelerator for FLASH-RT. Representative scanned images of monitor films 24 hours after FLASH-RT (H) and CONV-RT (I) irradiation are displayed. (J) The FLASH dose distribution of an EBT3 film mounted at a 5-mm depth in solid water. A 1.2 × 2 cm FLASH irradiation field is indicated by the black rectangle. (K) Horizontal dose distribution of FLASH-RT. (L) Vertical dose distribution of FLASH-RT. (M) Dose distribution of FLASH-RT. (N) The CONV dose distribution of an EBT3 film mounted at a 5-mm depth in solid water. A 12 × 300 mm CONV irradiation field is indicated by the black rectangle. (O) Horizontal dose distribution of CONV-RT. (P) Vertical dose distribution of CONV-RT. (Q) Dose distribution of CONV-RT.

Abbreviations: CONV-RT = conventional radiation therapy; FCT = fast current transformer; FLASH-RT =FLASH radiation therapy; PMMA = polymethyl methacrylate; RT = radiation therapy; SSD = source-skin distance.

Skin toxicity assessment

Skin toxicity was assessed in irradiated mice using a previously published scoring system.²⁴ A region spanning the entire brain was evaluated daily from days 30 to 60 after irradiation. Grade 0: the skin appears normal and healthy. Grade 1: only minimal erythema is observed. Grade 2: noticeable redness accompanied by early inflammatory signs and hair loss is present. Grade 3: diffuse microlesions and the onset of ulcerations can be detected. Grade 4: there is confluent moist desquamation with ulcerations measuring <0.5 cm. Grade 5: open wounds with tissue necrosis are evident, marking the terminal stage.

Immunofluorescence

Briefly, following saline perfusion, mouse brains were fixed using a 4% paraformaldehyde solution. The fixed brains were then dehydrated in a 30% sucrose solution (in phosphate-buffered saline [PBS]). Coronal sections, each 20 μm thick, were cut with a Leica CM3050S microtome and subsequently prepared for immunohistochemical analysis. After 3 washes in PBS, the sections were incubated for 2 hours at room temperature with a blocking buffer composed of 0.3% Triton X-100 and 5% bovine serum albumin in PBS, followed by incubation with a guinea pig anti-Glial fibrillary acidic protein antibody (1:500; OB-PGP055-01, Oasis Biofarm) and a rabbit anti-NeuN antibody (1:500; OB-PRB039-01, Oasis Biofarm). The secondary antibody consisted of Alexa Fluor 647-conjugated goat anti-guinea pig IgG (1:500; A-21450, Invitrogen) and Alexa Fluor 488-conjugated donkey anti-rabbit IgG (1:500; A-21206, Invitrogen). Nuclear morphology was visualized using 40,6-diamidine-20-phenylindole dihydrochloride (D8200, Solarbio). Finally, the slides were imaged using a Nikon A1 confocal microscope (Nikon).

Elevated plus maze test

Animals were habituated to the test room for 60 min before experimentation. At the start of tests, individual mice were placed in the center facing a closed arm and allowed 5 minutes of free exploration, recorded by a video camera above the apparatus. ANY-maze software was employed to record the mouse's trajectory over a 5-minute period, quantifying both the frequency of entries and the duration of time spent in the open arms.

Morris water maze

The water maze measured 90 cm in diameter and 50 cm in height, with a water depth of 20 cm. The water temperature was maintained at 23°C. Preliminary training to locate the platform precedes the formal procedure, which consists of 7 consecutive days of place navigation testing and a single probe trial. A stainless-steel platform raised 1 cm above the water was placed at the center of the pool. Each mouse was given 3 trials—randomly released from different quadrants—to find the platform within 60 seconds. If a mouse failed to locate it within 60 seconds, the experimenter guided it to the platform, where it remained for 15 seconds. During the place navigation test, a transparent platform was submerged 1 cm below the water’s surface in the first quadrant. For each trial, the mouse faced the pool wall and was randomly released from one of the 4 quadrants, with a 30-second interval between tests. One hour after completing the spatial navigation assessment on day 6, the platform was withdrawn for the ensuing probe trial. The original platform location becomes the “virtual platform.” Each mouse was released once from a drop point in the quadrant opposite the target quadrant. The ANY-maze software was used to measure the time to first reach the virtual platform, the duration spent in the target quadrant within 60 seconds, the number of crossings over the virtual platform, and the distance traveled in the target quadrant.

Y-maze test

Each arm of the Y-maze measured 35 cm in length, 5 cm in width, and 20 cm in height. The experiment proceeds in 2 stages: first, the novel arm was blocked with a partition. The mouse was gently placed in the central area, oriented away from the novel arm, and allowed to freely explore the other 2 arms for 5 minutes. It was then removed. Next, the partition was opened, and the mouse was placed again in the central area, oriented away from the novel arm, and allowed to explore all 3 arms for another 5 minutes. The exploration zone was defined as the distal half of each chosen arm: a mouse must enter the farthest half of an arm for it to be considered exploration. The ANY-maze software was used to measure the time in novel arm and entries in novel arm.

New object recognition test

The test was performed in a 40 × 40 × 35 cm box. The familiar objects were a cylinder measuring 2 cm in diameter and 3 cm in height, whereas the novel objects were cubes with sides of 2 cm. Mice were acclimated to the box for 5 minutes. The procedure consisted of 2 phases: a 10-minute familiarization period, during which 2 identical objects were presented. 1 hour later, followed by a 5-minute test period, in which one of the familiar objects was replaced by a novel object. The time spent by the mice exploring each object was recorded with a video camera mounted overhead. On completing video collection, data were analyzed with ANY-maze software.

Oxidative stress detection

Sixty days after irradiation, 4 mice per group were euthanized and their hippocampal tissues extracted. Each sample was weighed and homogenized in lysis buffer, then centrifuged at 3000 × g for 10 minutes at 4°C. Protein concentrations were determined using a BCA protein assay kit (Beyotime). Furthermore, levels of malondialdehyde (MDA), glutathione peroxidase (GSH-PX), superoxide dismutase (SOD), and total antioxidant capacity (T-AOC) were measured following the guidelines provided by the assay kits (Beijing Leagene Biotechnology Co, Ltd).

Proteomic analysis

Hippocampus from each group of mice (n = 4) was collected for protein extraction and proteomic analysis and then stored at -80°C. Tissues were lysed in a solution containing protease inhibitors and 8 M urea, subjected to ultrasonic disruption on ice for 45 minutes, and subsequently centrifuged at 12,000 rpm for 20 minutes at 4°C. The protein concentration in the supernatant was quantified using the BCA assay. Proteins were reduced and alkylated with iodoacetamide at room temperature in the dark for 40 minutes, followed by repeated washing with NH₄HCO₃. Subsequently, trypsin was introduced at an enzyme-to-protein mass ratio of 1:50, and the mixture was incubated overnight at 37°C. Following this, peptides were retrieved via vacuum extraction, then desalinated and measured. Mass spectrometric analysis was conducted on an Orbitrap Exploris 480 (Thermo Fisher Scientific), employing a full scan range from 300 to 1800 m/z, with primary and secondary mass resolutions set at 20,000 and 15,000, respectively. After data collection, differential expression analysis was executed using the DESeq2 package in R, with significance defined by P < .05 and an absolute log₂ fold change exceeding 1. Thereafter, the Cluster Profiler tool was used to perform Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes enrichment analyses on the genes exhibiting differential expression, which led to the identification of significantly enriched signaling pathways (P < .05). Finally, gene set enrichment analysis was carried out using the software developed by the Broad Institute.

Statistics

All data and statistical analyses were carried out using the statistical software SPSS (version 27.0, Inc). Behavioral outcomes were assessed with ANY-maze (Stoelting Co). One-way analysis of variance (ANOVA) was employed to evaluate the data. When this analysis indicated statistically significant differences, Tukey-Kramer hoc multiple-comparisons tests were performed to pinpoint disparities among specific groups. P values <.05 were regarded as significant and are depicted in the figures as *P < .05, **P < .01, and ***P < .001. All results are presented as the mean ± SEM.

Results

Parameters and dosimetry of x-ray-based FLASH-RT and CONV-RT

For each irradiated mouse, the output of the electron beam was continuously monitored using a fast current transformer situated upstream of the target. Simultaneously, x-ray exposure was recorded with a diamond detector positioned at the collimator’s entry (Fig. 1B). The detailed records of the delivered dose and corresponding dose rate for each subject were maintained (Table E1).

CONV-RT was delivered through radiation using the clinical Elekta Precise Linac (Fig. 1C). Additionally, an EBT3 film was attached to the head of the irradiated mouse to record the delivered dose (Fig. 1D, E). We positioned the EBT3 film horizontally along the beam direction at the beam's central point to derive the percent depth dose and the beam profiles of FLASH-RT (Fig. 1F). The pulse characteristics of the x-ray-based FLASH-RT were depicted in Figure 1G, which showed a sinusoidal modulation of pulse width accompanied by continuous electron micropulses lasting roughly 16.5 µs; each micropulse persists for about 500 µs. Dose distributions for both FLASH-RT (Fig. 1H, J) and CONV-RT (Fig. 1I, N) were determined by positioning an EBT3 film at a 5-mm depth in a solid water phantom simulating the mouse brain’s central region. This setup demonstrated the horizontal (Fig. 1K, O) and vertical (Fig. 1L, P) dose profiles of whole-brain RT, confirming that the dose distribution within the irradiated regions was uniform and consistent for both FLASH-RT and CONV-RT. Additional details on the dose distribution histograms were provided in Figure 1M, Q.

Comparative analysis of fundamental physiological conditions in mice following FLASH-RT and CONV-RT

Both the CONV-RT20Gy group and the FLASH-RT20Gy group exhibited pronounced weight loss starting on day 6 postirradiation and continuing until day 10, after which recovery ensued. Remarkably, mice in the FLASH-RT20Gy cohort experienced a milder degree of weight loss that rebounded more rapidly than those in the CONV-RT20Gy cohort (Fig. 2A). According to the humane endpoint criteria: if the body weight of the mice decreases by >20%, a humane endpoint will be applied, we monitored the survival of mice following FLASH-RT and CONV-RT. Thirty days after treatment, the survival rate of CONV-RT20Gy mice was a mere 25%, in stark contrast to 87.5% in the FLASH-RT20Gy group (Fig. 2B). No deaths were observed in either the CONV-RT10Gy or FLASH-RT10Gy cohorts. Given that radiodermatitis is a frequent adverse effect of RT, we assessed skin changes in the cranial irradiation field under both modalities. At 30 days postirradiation, CONV-RT20Gy mice exhibited hair loss, prominent ulcerations, erythema, swelling, and crust formation. By day 60, follicular damages were observed (Fig. 2C, D). In contrast, FLASH-RT 20 Gy mice showed no discernible lesions on day 30, with only mild hair loss on day 60. Also, we did not observe any clinical symptoms induced by radiation damage in the laryngo-tracheal or bucco-pharyngeal tracts.

Figure 2.

Comparative analysis of physiological conditions in mice following FLASH-RT and CONV-RT. (A) Postradiation weight changes in mice. (B) Survival curves of healthy mice in the FLASH-RT and CONV-RT that received whole brain irradiation. (C) Photographs of postradiation skin injuries in mice. (D) Postradiation skin damage scoring in mice. (E) Astrocyte and neuron were observed using immunofluorescence staining. (F) The GFAP-positive cells in hippocampal DG regions. (G) The Neun positive cells in hippocampal DG regions. Scale bar, 20 µm. n = 16 mice per group. All data are presented as the mean ± SEM, and error bars represent the SEM. Significance was assessed using one-way ANOVA with Tukey-Kramer post hoc analysis, *P < .05, **P < .01, and ***P < .001.

Abbreviations: ANOVA = analysis of variance; CONV-RT = conventional radiation therapy; DG = dentate gyrus; FLASH-RT = FLASH radiation therapy; GFAP = glial fibrillary acidic protein; RT = radiation therapy.

FLASH-RT mitigates the neuroinflammation and anxiety-like behaviors

Glial fibrillary acidic protein is an astrocyte marker, and we examined the activation of astrocytes in the Dentate Gyrus (DG) region of the mouse hippocampus using immunofluorescence after FLASH-RT and CONV-RT (Fig. 2E). Two months after whole-brain irradiation, we found that the CONV-RT group significantly increased the number of astrocytes in the DG, whereas the FLASH-RT group dramatically reduced their number compared to the CONV-RT group (2F). In addition, we measured the number of viable neurons in the DG 2 months after whole-brain irradiation. The results revealed that CONV-RT caused significant neuronal loss in the DG, whereas the number of neurons in the DG of FLASH-RT groups was significantly increased than that of the CONV-RT group at the same dosage (Fig. 2G).

RT for patients with brain tumor often precipitates neuropsychiatric complications, with anxiety being particularly pervasive. Thus, we examined anxiety-like behaviors in mice 2 months following FLASH-RT or CONV-RT. In the elevated plus maze, CONV-RT10Gy and CONV-RT20Gy groups (n = 4) exhibited fewer entries into the open arms, spent less time there, and covered a shorter distance than Sham controls, signifying pronounced anxiety-like symptoms (Fig. 2E-H). In contrast, FLASH-RT10Gy and FLASH-RT20Gy (n = 12) improved these behaviors. Collectively, 2 months after irradiation, CONV-RT induced marked anxiety-like behaviors, whereas FLASH-RT reduced these effects. However, we acknowledged that the small sample size in the CONV-RT20Gy group (n = 4) constitutes a limitation of this study. This limitation may impact the statistical robustness of the results and should be carefully considered when interpreting the observed "FLASH effect."

FLASH-RT mitigates the cognitive impairment

To evaluate the effects of FLASH-RT and CONV-RT on cognition, we performed Morris water maze, Y-maze, and novel object recognition tests 2 months after irradiation. In the Morris water maze acquisition phase (Fig. 3E), CONV-RT20Gy (n = 4) mice exhibited a marked increase in escape latency compared with the Sham group, requiring significantly more time to locate the platform (Fig. 3F). During the probe trial, CONV-RT10Gy and CONV-RT20Gy (n = 4) mice showed prolonged latencies to first reach the target platform, reduced swimming distances in the target quadrant, fewer platform crossings, and less time spent in that quadrant (Fig. 3G-K). In contrast, FLASH-RT10Gy and FLASH-RT20Gy (n = 12) animals demonstrated shorter times to first platform contact, increased distances in the target quadrant, more crossings, and extended durations in that region compared with CONV-RT counterparts. Furthermore, in the Y-maze test (Fig. 4A), CONV-RT10Gy and CONV-RT20Gy (n = 4) mice spent less time in the novel arm and entered it significantly fewer times than Sham, whereas FLASH-RT10Gy and FLASH-RT20Gy (n = 12) reduced these trends (Fig. 4B-D). Similarly, in the novel object recognition test (Fig. 4E), CONV-RT10Gy and CONV-RT20Gy groups (n = 4) exhibited lower recognition indices relative to Sham, while the FLASH-RT groups showed no significant difference (Fig. 4F-H). Collectively, 2 months following CONV-RT, mice displayed pronounced cognitive dysfunction and diminished learning and memory, whereas FLASH-RT mitigated these radiation-induced impairments. However, we acknowledged that the small sample size in the CONV-RT20Gy group (n = 4) represents a limitation in this study. This limitation might influence the statistical robustness of the cognitive assessment results and should be considered when interpreting the observed differences between the FLASH-RT and CONV-RT groups.

Figure 3.

Comparison of mouse behavior after FLASH-RT and CONV-RT. (A) Open arms entry number. (B) Time in open arms. (C) Distance in open arms. (D) Representative trajectory diagram of the elevated plus maze test. (E) Experimental paradigm for Morris’ water maze. (F) Escape latency of Morris’ water maze during the place navigation test. (G) Latency. (H) Distance in the target quadrant. (I) Target entries. (J) Time spent in the target quadrants. (K) Representative trajectory diagram of the Morris’ water maze. Animals (n = 16, 16, 4, 16, and 12 per group) were assigned to Sham, CONV-RT 10 Gy, CONV-RT 20 Gy, FLASH-RT 10 Gy, and r FLASH-RT 20 Gy, respectively. All data are presented as the mean ± SEM, and error bars represent the SEM. Significance was assessed using one-way ANOVA with Tukey-Kramer post hoc analysis, *P < .05, **P < .01, and ***P < .001.

Abbreviations: ANOVA = analysis of variance; CONV-RT = conventional radiation therapy; FLASH-RT =FLASH radiation therapy; RT = radiation therapy.

Figure 4.

Comparison of mouse cognition function after FLASH-RT and CONV-RT. (A) Experimental paradigm for Y maze. (B) Time in novel arm. (C) Entries in novel arm. (D) Representative heatmaps of the Y maze. (E) Experimental paradigm for new object recognition test. (F) Discrimination index. (G) Distance. (H) Representative trajectory diagram of the new object recognition test. Animals (n = 16, 16, 4, 16, and 12 per group) were assigned to Sham, CONV-RT 10 Gy, CONV-RT 20 Gy, FLASH-RT 10 Gy, and r FLASH-RT 20 Gy, respectively. All data are presented as the mean ± SEM, and error bars represent the SEM. Significance was assessed using one-way ANOVA with Tukey-Kramer post hoc analysis, *P < .05, **P < .01, and ***P < .001.

Abbreviations: ANOVA = analysis of variance; CONV-RT = conventional radiation therapy; FLASH-RT = FLASH radiation therapy; RT = radiation therapy.

FLASH-RT-reduced oxidative stress

To investigate the differences in radiation-induced damage between CONV-RT20Gy and FLASH-RT20Gy, we conducted proteomic analyses on the hippocampal tissues of both mouse groups. We performed a partial least squares discrimination analysis on the 5 groups (Fig. 5A). As shown in Figure 5B, C, the CONV-RT20Gy group exhibited 386 significantly upregulated and 376 downregulated genes compared to the FLASH-RT20Gy group. GO analysis revealed that the differentially expressed genes were mainly enriched in positive regulation of DNA repair, response to nitric oxide, and negative regulation of vascular endothelial growth factor receptor signaling pathway (Fig. 5D). Kyoto Encyclopedia of Genes and Genomes pathway analysis suggested that exposure to CONV-RT20Gy had multiple biological impacts, including effects on biosynthesis, ferroptosis and autophagy (Fig. 5E). Consistently, gene set enrichment analysis demonstrated upregulation in signaling pathways related to the cell death in response to oxidative stress, and regulation of DNA replication in the CONV-RT20Gy group. Conversely, pathways associated with the reactive oxygen species biosynthetic process, negative regulation of wound healing, and the reactive oxygen species metabolic process were downregulated (Fig. 5F). To corroborate our findings and assess oxidative stress in the mouse hippocampus following irradiation, we employed assay kits to quantify MDA levels and overall antioxidant capacity 60 days after irradiation. The FLASH-RT20Gy cohort demonstrated a substantial decrease in MDA levels accompanied by increased values for SOD, T-AOC, and GSH-PX relative to the CONV-RT20Gy group (Fig. 5G-J). Overall, these results indicated that, 60 days postirradiation, the CONV-RT20Gy group experienced pronounced oxidative stress and diminished antioxidant capacity, whereas the FLASH-RT20Gy group reduced oxidative damage by lowering MDA levels and enhancing the activity of key antioxidant enzymes such as SOD, T-AOC, and GSH-PX, thus improving the overall antioxidant capacity and mitigating the delayed detrimental effects of irradiation.

Figure 5.

FLASH-RT reduced oxidative stress. (A) Partial least squares discrimination analysis of 5 groups. (B) Heatmap showing the expression levels of the genes with the largest expression differences between CONV20Gy and FLASH20Gy groups. (C) Volcano plot showing the number and distribution of differentially expressed genes between CONV20Gy and FLASH20Gy groups. (D) Gene Ontology (GO) enrichment plot showing the biological processes enrichment of differentially expressed genes between CONV20Gy and FLASH20Gy groups. (E) Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment plot showing the metabolic pathways enrichment of differentially expressed genes between CONV20Gy and FLASH20Gy groups. (F) Gene set enrichment analysis (GSEA) enrichment plot showing the signaling pathways enrichment of differentially expressed genes between CONV20Gy and FLASH20Gy groups. (G-J) Levels of malondialdehyde (MDA), superoxide dismutase (SOD), total antioxidant capacity, and glutathione peroxidase (GSH-PX) in mouse hippocampal tissue. n = 4 mice per group. All data are presented as the mean ± SEM, and error bars represent the SEM. Significance was assessed using one-way ANOVA with Tukey-Kramer post hoc analysis, *P < .05, **P < .01, and ***P < .001.

Abbreviations: ANOVA = analysis of variance; CONV-RT = conventional radiation therapy; EMC = endoplasmic reticulum membrane complex; FLASH-RT = FLASH radiation therapy; GABA = gama-aminobutyric acid; NS = no significance; RT = radiation therapy; SNARE = soluble N-ethylmaleimide-sensitive factor attachment protein receptors; T-AOC, total antioxidant capacity.

Discussion

Both preclinical research and clinical trials have indicated that FLASH-RT can not only eradicate the tumors effectively but also significantly diminish the harmful effects of ionizing radiation on healthy tissues.^25,26 Research indicated that CONV-RT induced severe oxidative stress in tissues; surprisingly, FLASH-RT mitigated the oxidative stress response in exposed normal tissues.^21,27 Moreover, FLASH-RT mitigated both inflammatory immune responses and the radiation-induced cytotoxicity affecting circulating immune cells,²⁸ thereby maintaining its antitumor efficacy while protecting normal tissues.²⁹ However, this study was conducted on healthy, tumor-free mice, highlighting the need for additional research involving tumor-bearing models to evaluate more realistic delayed outcomes.

A primary focus of this investigation was the meticulous tracking of radiation dose and its rate in both x-ray-based FLASH-RT and CONV-RT. To achieve precise dose measurement, we employed EBT3 films, fast current transformer beam detectors, and ionization chambers. Detailed film analysis definitively verified that our apparatus consistently administered the designated dose at an ultrahigh rate of 213 Gy/s, thus confirming the accurate delivery of both treatment modalities. Indeed, in our study, CONV-RT was delivered at 6 MV whereas FLASH-RT was performed at 10 MV, mainly because of technical limitations of the available accelerator system, where ultrahigh-dose rates could only be achieved with 10 MV. We acknowledged that beam energy differences may potentially influence dose distribution and biological outcomes. However, in small animal cranial irradiation models, the depth-dose differences between 6 MV and 10 MV were relatively minor, and the brain tissues of mice were fully encompassed within the high-dose region. Therefore, we believed that the substantial differences observed in cognitive performance and proteomic signatures are more likely attributable to the dose-rate effect rather than the modest variation in beam energy.

Existing studies have demonstrated that FLASH-RT offers protective effects on brain tissue,³⁰ alleviating damage to the synaptic structure and function of hippocampal neurons under ultrahigh-dose rates.³¹ This was achieved through the reduction of neuroinflammation and oxidative stress,²⁷ thus safeguarding cognitive function. Similarly, our study was the first to demonstrate the advantages of x-ray-based FLASH-RT delivered at an ultrahigh-dose rate of 213 Gy/s in the brains of healthy mice. Even 2 months postirradiation, FLASH-RT continued to mitigate the activation of hippocampal astrocytes and preserve neuronal survival. Moreover, we observed oxidative stress responses in the brain tissue, which align with previous findings that FLASH-RT effectively reduced radiation-induced oxidative stress. This finding was consistent with previous studies and further substantiated the potential of FLASH-RT in mitigating radiation-induced brain damage. However, despite the compelling evidence from existing research, further validation is necessary under conditions involving larger sample sizes and more diverse experimental parameters. Moreover, caution is warranted when extrapolating these findings to human patients, because human tumors require substantially larger treatment volumes than murine models, which may attenuate the protective effects on normal tissues. Human tumors require substantially larger treatment volumes than those in murine models, which may attenuate the degree of normal tissue sparing observed here. Moreover, because the present work was performed in healthy, tumor-free mice, further investigations in tumor-bearing models are essential to assess therapeutic efficacy and delayed outcomes under clinically relevant conditions.

At a dose of 20 Gy, CONV-RT induced severe acute weight loss, high mortality (survival rate 25%), and progressive skin and follicular damage, accompanied by anxiety-like behavior and cognitive deficits. It should be emphasized, however, that only a limited number of animals were included in the behavioral assays, which restricts the strength of conclusions regarding cognitive impairment. In contrast, FLASH-RT caused milder weight loss with rapid recovery, markedly improved survival (75%), and minimal late skin or hair changes, while alleviating radiation-induced behavioral and cognitive dysfunction. Our cognitive findings are broadly consistent with prior work showing that FLASH-RT mitigates radiation-induced neurocognitive decline compared with CONV-RT. For example, Montay-Gruel et al³² demonstrated that FLASH-RT preserved memory performance in irradiated mice. The present study corroborated these observations and further suggested that antioxidant defense mechanisms might have played a role in mediating these protective effects. Proteomic analysis of hippocampal tissues revealed that CONV-RT predominantly affected pathways related to DNA repair, oxidative stress, ferroptosis, and autophagy, whereas FLASH-RT reduced oxidative damage, as indicated by lower MDA and higher SOD, T-AOC, and GSH-PX levels. Because they imply that FLASH-RT better preserved redox homeostasis and antioxidant capacity, this may underlie its ability to spare normal tissues and protect cognitive function.³³ This interpretation is supported by earlier reports showing that FLASH-RT reduced oxidative stress-driven cellular damage and preserves normal tissue function.²¹ Although our analysis centered on the hippocampus, other regions such as the amygdala are also crucial for anxiety-related behaviors.³⁴ Future studies should therefore extend beyond the hippocampus to provide a more comprehensive understanding of the neural basis of FLASH-RT mediated behavioral protection. Although these encouraging outcomes indicated that FLASH-RT may offer clinical advantages, caution must be exercised when applying these findings to human patients with cancer. Notably, human tumors have a substantially larger treatment volume than those in mice, which could constrain the degree of normal tissue protection observed in preclinical models. Moreover, most devices available today are tailored for animal studies with restricted RT fields, necessitating substantial modifications to broaden their irradiation range for cancer treatment. Finally, it is crucial to consider the inherent heterogeneity of tumors, because responses to FLASH-RT may differ depending on tumor type and microenvironment.

One limitation of this study was the small sample size in the CONV-RT20Gy group (n = 4), which affected the robustness and generalizability of the findings. The smaller sample size in this group than the other groups introduced a potential source of bias and may limit the statistical power of the results, particularly when comparing the effects of CONV-RT20Gy with FLASH-RT. Future studies with larger sample sizes are necessary to validate these findings and further explore the long-term effects of conventional RT on cognition and behavior. Expanding the sample size will help strengthen the reliability of the observed outcomes and provide more definitive conclusions regarding the comparative efficacy of FLASH-RT and CONV-RT.

Conclusions

Initially, we confirmed the feasibility and physical characteristics of the miniaturized x-ray FLASH platform. Second, we observed that x-ray-based FLASH-RT conferred a sustained reduction in cognitive dysfunction compared with CONV-RT. Moreover, proteomic analysis of hippocampal tissues revealed that CONV-RT predominantly affected pathways related to DNA repair, oxidative stress, ferroptosis, and autophagy, whereas FLASH-RT was associated with reduced oxidative damage. Collectively, these results demonstrated that FLASH-RT not only mitigated radiation-induced oxidative stress but also preserved redox homeostasis and cognitive function, thereby underscoring its translational potential for clinical application.

Disclosures

All contributing authors report no conflict of interest in this work.