Combined minibeam and conventional radiotherapy reduces severe skin toxicity and improves tumor control

Abstract

Radiotherapy (RT) is an important tool for cancer treatment, enabling the delivery of curative doses to tumors while minimizing exposure to surrounding healthy tissue. Advances in beam delivery techniques and precise image-guidance systems have significantly reduced radiation exposure to normal tissues; however, treatment-related side effects remain a major limitation. To address these challenges, novel preclinical strategies are being explored. This study investigated minibeam radiotherapy (MBRT), which employs narrow (>100 microns) parallel beams of X-rays that produce a distinctive peak-and-valley dose distribution. We aimed to assess the efficacy and skin toxicity of MBRT compared to conventional radiotherapy (CONV), and to examine the role of macrophages in treatment response. We used a syngeneic glioblastoma mouse model, created by subcutaneously injecting GL261 cells into the flank, and applied single-fraction treatments of MBRT, CONV, or a combined minibeam-conventional regimen (MBRT-CONV). Irradiation was performed using a dedicated, image-guided small animal radiotherapy system, combined with a custom minibeam collimator. We evaluated tumor-specific survival and acute radiation-induced skin damage at macroscopic and microscopic levels through histopathological analysis to compare treatment outcomes. Our results showed that MBRT led to significantly reduced incidence and severity of skin ulceration compared to CONV at high single doses (25-30 Gy), although MBRT alone provided less effective tumor control. Additionally, we evaluated the effects of these treatments on healthy skin to assess the tissue-sparing potential of MBRT. Specifically, we investigated the impact of RT on macrophage recruitment in healthy skin. Based on the rapid inflammatory resolution observed in healthy skin, we implemented a delayed combined treatment strategy to enhance tumor suppression, by administering a CONV dose fraction seven days after the initial MBRT. Among the tested regimens, the 30 Gy MBRT + 5 Gy CONV combination achieved the most pronounced tumor growth delay and survival benefit, while maintaining a low incidence of toxicity.

Introduction

Radiotherapy (RT) aims to accurately deliver a curative dose to the tumor while minimizing exposure to surrounding healthy tissues. Despite significant reductions in radiation doses to healthy tissues, such as the skin, achieved through advanced beam delivery techniques and precise image-guidance systems, treatment-related side effects remain a major limitation to the overall effectiveness of RT. The skin represents one of the dose-limiting tissues in clinical practice, as the delivery of the high radiation doses required to eradicate aggressive brain tumors, while remaining within the tolerance limits of critical structures such as the spinal cord, brainstem, and the skin itself, constitutes a major challenge, particularly in pediatric patients. Therefore, the treatment of these tumors generally remains palliative¹.

Novel preclinical approaches have been developed to face problems like these, such as microbeam radiotherapy (MRT) or minibeam radiotherapy (MBRT)². MRT and MBRT are referred to as spatially fractionated radiation therapy techniques (SFRT). The fundamental principle is based on the reduction of the irradiated normal tissue volume to increase radiation tolerance^3,4. In general, for MRT, kilovoltage X-ray beams delivered by a synchrotron source of 25-100 μm size separated by 100-400 μm intervals are used to generate a spatially fractionated dose distribution⁵. Unlike standard RT, this beam configuration results in non-uniform dose profiles composed of high and low-dose regions defined as peak and valley, respectively. The peak dose is the dose in the center of the minibeam. The valley dose, caused by the scattering of the primary radiation, is the minimum dose between two consecutive minibeams.

The exact radiobiological mechanisms behind the sparing effect of MBRT are far from being understood; however, it is typically described in terms of the bystander effect⁶ which is the ability of irradiated cells to convey manifestations of damage to neighboring cells not directly irradiated. It has been assumed that the peak-to-valley dose ratio (PVDR) is a key dosimetric parameter related to the therapeutic effect of any SFRT: the higher the PVDR, the lower the toxicity^2,7. Therefore, major efforts have been made to maximize PVDR. However, a recent comprehensive review has highlighted a strong correlation between valley dose and increased life span in preclinical SFRT experiments⁸. In preclinical MRT animal experiments performed at synchrotron facilities, radiation doses (peak dose) of hundreds of Gy in a single fraction were well tolerated by a range of normal tissues^9–11. It has been emphasized that the valley dose has to be less than the tissue tolerance level to broad beams to maintain the sparing effect of MRT⁷. Importantly, MRT seems to enable remarkable tumor control in radioresistant tumors although a significant part of the tumor volume is exposed to a lower dose¹². Such very high doses are delivered in a short time to avoid the beam smearing due to tissue motion. At the moment, only synchrotron sources are capable of producing ultra-high dose rates MRT, making it also impossible to decouple spatial from ultra high dose rate (FLASH) effects¹³.

Regarding MBRT, Dilmanian et al.¹⁴ investigated the healthy tissue tolerance in a rat model. No side effects were observed up to 170 Gy peak dose delivered to the brain using 0.68 mm sized minibeams. Prezado et al.¹⁵ performed experiments with minibeams of 0.6 mm and found that the rat brain tolerated a peak dose of 100 Gy in one fraction. Additionally, animal studies demonstrated a significant delay in tumor growth of radioresistant tumors treated with MBRT^15,16. These results suggest that MBRT could serve as an alternative to MRT, with the key advantage of being more easily delivered, potentially using a preclinical X-ray irradiator. Additionally, MBRT dose distributions are less susceptible to beam blurring caused by organ movement, allowing MBRT experiments to be conducted at more conventional dose rates. The limited availability of synchrotron sources, combined with the promising outcomes of MBRT studies, has motivated a few researchers to implement MBRT in preclinical settings equipped with small animal irradiators.

In this study, we employed a syngeneic subcutaneous glioblastoma (GBM) model in immunocompetent mice. GBM was selected due to its clinical relevance, marked radioresistance, and high unmet therapeutic need, which pose significant challenges for current radiotherapy strategies. The syngeneic model allows preservation of host immune competence, enabling the assessment of tumor-immune interactions in response to radiation. This approach is particularly relevant for MBRT, which has shown a normal tissue-sparing effect while maintaining anti-tumor efficacy in preclinical studies. Using this model, we were able to simultaneously evaluate tumor growth control and radiation-induced skin toxicity, providing a comprehensive framework to assess the potential of MBRT alone or in combination with CONV. Overall, this preclinical model provides a biologically relevant and translationally informative system to investigate both efficacy and toxicity endpoints, supporting the optimization of novel radiotherapy regimens. Specifically, we report on skin toxicity and tumor response from a dose-response experiment in which cohorts of mice received single-fraction MBRT, single-fraction CONV or combined therapy (MBRT-CONV) using a small animal irradiator (X-RAD225Cx SmART, PXI, North Branford, CT, USA). The severity of skin toxicity and overall survival were evaluated as key endpoints. Additionally, we assessed the impact of these treatments on healthy skin to explore the tissue-sparing potential of MBRT and to understand the role of macrophages in resolving inflammation.

Methods

Study design

This preclinical study was designed to investigate the radiobiological and therapeutic differences between MBRT, CONV, and their sequential combination (MBRT-CONV) in a murine syngeneic GBM model. The primary objectives were to assess tumor control efficacy, normal tissue toxicity, and the underlying inflammatory response dynamics.

To achieve these goals, the experimental workflow was structured in sequential phases. First, tumor-bearing mice were treated with single-fraction monotherapies (CONV or MBRT) at varying dose levels to establish baseline efficacy and macroscopic skin toxicity profiles. Second, to understand the tissue-sparing effect of MBRT, we evaluated macrophage infiltration kinetics (ex vivo culture and histopathological analyses) at specific time points post-irradiation. Finally, capitalizing on the rapid inflammatory resolution observed after MBRT, we designed a delayed MBRT-CONV combination protocol. This combined regimen was formulated to test whether the initial spatially fractionated dose could prime the tumor microenvironment, thereby allowing a subsequent conventional irradiation to achieve superior tumor suppression with reduced normal tissue toxicity.

Detailed methodologies regarding tumor implantation, dosimetric characterization, clinical scoring, and histopathological analyses are provided in the respective subsections below.

Irradiation setup and dosimetric characterization

All irradiations were performed using an XRAD-SmART small-animal image-guided irradiation system (Precision X-Ray, North Branford, CT, USA), operating with a 225 kV X-ray source that produces a divergent beam. Because of this intrinsic beam divergence, MBRT treatments had to be delivered with the collimator in direct contact with the skin surface; otherwise, beam smearing would occur, leading to the loss of the spatial dose fractionation (see Fig. 1). The primary collimator aperture was set to 14.4 x 14.4 mm, and a dedicated minibeam collimator was placed below it. This setup generated the spatially fractionated pattern used for MBRT, producing 500-μm wide minibeams spaced 1 mm center-to-center. The resulting irradiation geometry consisted of alternating high dose peak and low-dose valley regions.

Fig. 1.

Experimental setup and minibeam collimator assembly for MBRT delivery. (a) The custom minibeam collimator (MBC) and the auxiliary mounting tool. (b) The MBC properly inserted into the auxiliary tool. (c) The assembled MBC mounted onto the standard reference collimator of the small-animal irradiation system. (d) Representative irradiation geometry and alignment setup, illustrated here using a solid phantom to demonstrate the necessary contact between the MBC and the target surface to prevent beam smearing.

Peak, valley, and mean doses, as well as the PVDR and full-width at half-maximum (FWHM), were experimentally determined by placing Gafchromic radiochromic films on the surface of a PMMA phantom to capture the entrance dose. This procedure replicated the exact setup geometry used for the in vivo treatments. For all experimental groups, the radiation prescription was explicitly defined by the mean dose at the tissue entrance (skin surface). Specifically, for MBRT treatments, the 25 Gy mean dose corresponded approximately to an entrance peak dose of 56 Gy and a valley dose of 4.2 Gy, while the 30 Gy mean dose corresponded approximately to a peak of 67 Gy and a valley of 5 Gy. Each RT treatment was administered using a single anterior-posterior beam to maintain consistent radiation geometry. We acknowledge that in this preclinical setup, it is not possible to determine the exact dose delivered at the tumor location for each individual animal. Tumor depth can vary slightly due to biological variability and positioning, and direct in vivo dosimetric measurement at the tumor level is not feasible in this model. Although the spatial dose distribution (e.g., PVDR) and absolute dose naturally vary with depth, the superficial subcutaneous location of the tumors limits beam degradation. For this reason, and consistently with previous studies using the same irradiation platform, the dose was defined at the entrance (surface) level, which provides a reproducible and standardized reference. By applying this approach consistently to all irradiation conditions (MBRT and CONV) and maintaining constant irradiation times, we ensured strict dosimetric consistency, allowing relative comparisons between groups to remain valid. These experimental measurements were comprehensively characterized and corroborated using Monte Carlo simulations^17–19. Experimental uncertainties for the delivered doses were within ± 5%, as reported in Table 1. The source-to-surface distance (SSD) was kept constant during all irradiations to ensure dosimetric consistency.

Table 1.

Summary of irradiation protocols, delivered doses, and number of animals assigned to each treatment group. These regimens (CONV, MBRT, and MBRT-CONV) were applied to tumor-bearing female C57BL/6 mice once tumor volumes reached 150-200 mm. Experimental uncertainties for the delivered doses were within ± 5%.

N° of mice	Mean dose (± 5%)	Treatment
17	0 Gy	CTRL
7	10 Gy	CONV
17	15 Gy	CONV
7	25 Gy	CONV
8	25 Gy	MBRT
13	30 Gy	MBRT
6	25+5 Gy	MBRT-CONV
6	30+5 Gy	MBRT-CONV

In vivo CONV and MBRT of a GBM preclinical model

The in vivo effects of MBRT compared to CONV were assessed using a syngeneic GBM preclinical model. This was established by subcutaneously injecting GL261 cells suspended in 100 μL of PBS into the left flank of female C57BL/6 mice (weighing 17-23 g).

Treatment was initiated when the tumor volume reached approximately 150-200 mm. Mice were anesthetized and treated in a supine position. The animals were randomly assigned to different treatment groups as detailed in Table 1, which included untreated controls, single-fraction CONV, single-fraction MBRT, and the sequential MBRT-CONV combination. Specifically, the combination MBRT-CONV therapy consisted of animals receiving either 25 or 30 Gy mean dose MBRT, followed 7 days later by a 5 Gy CONV fraction. This 7-day interval was explicitly chosen to coincide with the post-MBRT tissue recovery phase, allowing the safe delivery of an additional uniform dose to enhance tumor control without compromising the normal tissue-sparing benefits of MBRT (see “Ex vivo culture of skin treated with CONV and MBRT in a static system to investigate the tissue-sparing effect” section).

Due to the requirement of direct contact between the MBRT multislit collimator and the skin surface (see “Irradiation setup and dosimetric characterization” section), standard image-guided CBCT positioning could not be used. Instead, tumor alignment was performed using the positioning lasers of the irradiation platform combined with visual and palpable identification of the subcutaneous tumor to accurately center the irradiation field.

Sample sizes varied among groups (e.g., n=17 for CTRL and CONV 15 Gy, versus n=7-8 for other single-dose groups). No formal a priori sample size calculation was performed, as this was an exploratory study evaluating a novel irradiation platform; instead, starting group sizes of n=7-8 per condition were established heuristically based on historical variance data from our laboratory for the GL261 model. The discrepancy in final group sizes is the direct result of our sequential experimental design and adherence to the 3Rs (Reduction) principle. In the initial cohort, a broad range of single-modality CONV and MBRT doses was evaluated. Based on these early results, which identified CONV 15 Gy and MBRT 30 Gy as the optimal comparative baselines for iso-efficacy, we designed subsequent cohorts to test the MBRT-CONV combination. To minimize animal use, we selectively replicated only the essential reference arms (e.g., CTRL, CONV 15 Gy and MBRT 30 Gy) in the subsequent cohorts, deliberately avoiding the unethical replication of highly toxic or less clinically relevant doses (e.g., CONV 25 Gy). Consequently, pooling the data across these sequential cohorts naturally resulted in larger sample sizes for the shared reference groups, while preserving absolute data transparency for all enrolled animals.

Skin toxicity scoring

To systematically evaluate radiation-induced skin reactions, an ordinal scoring scale (0-3), adapted from the literature²⁰ in a simplified form, was established and applied throughout the study. Skin toxicity was assessed visually and categorized using a four-tier grading system: Grade 0 (normal), indicating no visible alteration in skin appearance; Grade 1 (mild), characterized by focal depigmentation or alopecia confined to the irradiated field; Grade 2 (moderate), defined by evident desquamation within the irradiated field with partial loss of the superficial epidermal layer; and Grade 3 (severe), marked by ulceration, with or without associated alopecia and depigmentation, extending across the radiation field. Each mouse was inspected at regular intervals following irradiation (daily during the first 10-14 days and 2-3 times per week thereafter) to monitor the onset, progression, and resolution of radiation-induced damage. Observations were conducted under consistent lighting conditions, and both skin appearance and functional impairment (e.g., leg movement restriction, wound reopening) were recorded. Two independent observers, blinded to the treatment allocation, performed the scoring. In cases of discrepancy between their evaluations, the average of the two scores was assigned as the final value for each time point. Acute radiation damage was further rated on a parallel 0-3 scale considering the extent of tissue involvement and functional impact, ensuring comprehensive assessment of both macroscopic and functional endpoints. This semi-quantitative scoring approach enabled consistent comparison of skin toxicity severity across treatment groups (CONV, MBRT, and combined MBRT-CONV) and facilitated longitudinal analysis of recovery kinetics following different radiation modalities. Importantly, to ensure that the reported toxicity profiles strictly reflect radiation-induced damage, any animal that developed spontaneous mechanical ulceration driven purely by the physical expansion of the subcutaneous tumor mass was explicitly excluded from the final skin toxicity quantitative analysis.

Mouse monitoring and humane endpoints

All animal procedures were approved by the Institutional Animal Care and Use Committee (IACUC) of Ospedale San Raffaele (protocol number 6EEAF.276.EXT.245) and conducted in accordance with national and institutional guidelines for the care and use of laboratory animals. C57BL/6 mice were obtained from Charles River, Italy. All mice were carefully monitored for clinical signs of radiation-induced toxicity, tumor progression, and overall health status throughout the study. Animals were examined daily during the early phase and 2-3 times per week thereafter, with additional daily checks during periods of acute skin reaction. Clinical parameters recorded included body weight, posture, activity, grooming behavior, and food and water intake, together with detailed inspection of the irradiated skin area. Tumor growth was measured three times per week using digital calipers, and tumor volume was calculated according to the formula: . Tumor volume served as the primary quantitative endpoint, and individual length and width measurements for each animal are provided in the Supplementary Table S1. Body weight was monitored strictly for welfare purposes and was not recorded as a quantitative efficacy endpoint.

Humane endpoints were established prior to study initiation in accordance with institutional animal care guidelines to minimize animal suffering. Animals were euthanized if radiation-induced skin toxicity reached Grade 3 (ulceration), if tumor volume exceeded 1,500 mm, if body weight loss was greater than 15%, or if signs of systemic distress (e.g., ruffled fur, impaired mobility, reduced activity) were observed. In addition to the predefined volume threshold, animals were euthanized if spontaneous mechanical tumor ulceration occurred. This complication, driven by the physical expansion of the subcutaneous mass stretching the overlying skin, is clinically distinct from radiation-induced toxicity, as evidenced by its occurrence even in untreated controls, and represents a recognized humane endpoint even when the tumor volume had not yet reached 1500 mm. Furthermore, while the strict tumor volume endpoint was set at 1500 mm, animals exhibiting rapid exponential growth that were euthanized at slightly lower volumes (e.g., 1000-1200 mm) were appropriately scored as tumor-related events. This proactive euthanasia was implemented because their steep growth trajectory would have inevitably caused them to exceed the ethical limits during impending non-monitoring periods (e.g., weekends). Despite these proactive measures, the notoriously aggressive and unpredictable late-stage exponential kinetics of the GL261 model resulted in rare, isolated cases where tumor volumes explosively exceeded the threshold between consecutive measurements. In such instances, animals were immediately euthanized upon detection at the subsequent inspection, in strict adherence to welfare protocols. Euthanasia was performed by gradual-fill CO inhalation using a certified delivery system, followed by cervical dislocation as a secondary physical method. This monitoring and endpoint protocol ensured standardized welfare assessment, compliance with international ethical standards, and reliable comparison of treatment effects across all experimental groups.

Survival analysis

Overall survival was monitored for up to 90 days post-irradiation or until animals reached predefined humane or tumor volume endpoints. To strictly evaluate loco-regional tumor control, the survival event for the Kaplan-Meier analysis was defined exclusively as tumor progression reaching the predefined volume threshold (1,500 mm). This definition included animals proactively euthanized at slightly lower volumes (e.g., 1000-1200 mm) due to impending exponential growth during non-monitoring periods. Conversely, animals that were euthanized prior to reaching the tumor volume endpoint due to independent ethical criteria (specifically, severe radiation-induced skin toxicity or spontaneous mechanical tumor ulceration) were strictly treated as right-censored observations on the day of euthanasia. Furthermore, the actual follow-up duration for animals that did not reach a terminal event was determined by the specific termination date of each sequential experimental cohort; consequently, the maximum observation period ranged from 40 to 90 days post-treatment. Animals that remained alive and below the tumor volume threshold at the end of their respective cohort’s follow-up period were administratively right-censored on their final day of assessment.

Survival probabilities over time were estimated and plotted using the Kaplan-Meier method, allowing direct comparison among treatment groups. Statistical differences in survival distributions were evaluated using the log-rank (Mantel-Cox) test. This rigorous censoring strategy, which accounts for both competing risks and varying follow-up durations through administrative censoring, ensures that all animals contribute accurately to the survival estimates without introducing systematic downward bias.

Statistical analysis

Differences in survival distributions were assessed using the log-rank (Mantel-Cox) test. All statistical analyses were conducted using GraphPad Prism software (version 10.6.1), and a p-value of < 0.05 was considered statistically significant.

Ex vivo culture of skin treated with CONV and MBRT in a static system

Eleven healthy 8-week-old female C57BL/6 mice were used for this exploratory pilot ex vivo assay. To strictly minimize animal use in accordance with the 3Rs principle, a limited number of mice was employed (N=1 mouse per treated condition per time point, plus three untreated overall controls). Mice were irradiated bilaterally on both flanks with one of the following regimens: conventional irradiation (CONV, 15 Gy), minibeam radiation therapy (MBRT, 30 Gy), or combined treatments (MBRT 25 Gy + CONV 5 Gy, or MBRT 30 Gy + CONV 5 Gy). For the combined protocols, the MBRT fraction was delivered first, followed by the CONV dose 7 days later. Animals were euthanized either 3 hours or 7 days after their final irradiation, with the three control mice allocated to the 3-hour (N=1) and 7-day (N=2) evaluation points, respectively.

Skin samples were excised from both flanks immediately after euthanasia induced by CO₂ overdose, in accordance with the regulations of the Italian Ministry of Health (DL March 4, 2014, n. 26, directive implementation 2010/63/EU) and the European Communities Council Directive 2010/63/EU. All procedures involving live vertebrates were approved by the Italian Ministry of Health (authorization no. 336/2023-PR) and were conducted in compliance with ARRIVE guidelines. Freshly excised skin samples were rapidly washed in physiological solution and subsequently in pre-warmed culture medium at 37°C. The culture medium consisted of DMEM (Dulbecco’s Modified Eagle’s Medium) supplemented with 4.5 g/L D-glucose, 10% fetal bovine serum (FBS), 2% penicillin-streptomycin, 200 mM L-glutamine, and 0.3 μg/mL amphotericin B (Gibco, Waltham, MA, USA). Skin explants were maintained under static culture conditions at 37°C in a humidified atmosphere with 5% CO₂ for 6 hours in Petri dishes containing 4 mL of culture medium prior to fixation and subsequent histological and immunohistochemical analyses.

Histopathological analysis

Skin samples were dissected, washed with PBS 0.1 M, and maintained in culture for six hours. This short-term ex vivo incubation was performed to preserve tissue viability and allow controlled assessment of early inflammatory responses following irradiation. Then, samples were fixed in 4% paraformaldehyde for 24 h at 4°C. Tissues were then embedded in paraffin, cut into 5 μm thick sections with a microtome, and dried at 37°C for 24 h. Slides were warmed to 60°C for a minimum of 20 minutes and subjected to a series of deparaffinization and rehydration steps. Sections were stained with Hematoxylin and Eosin, and with anti-CD68 antibody (EPR23917-164, Abcam), 1:200, according to the manufacturer’s instructions to evaluate macrophage infiltration. Immunohistochemical analysis of skin sections was performed by incubating with antigen retrieval buffer (Citrate Buffer pH 6.0, ab64236, Abcam) and then labelled with anti-CD68. The Leica BOND RX autostainer was used for the immunohistochemistry with Kit-HRP MOM KIT (Abcam).

Entire histological sections were digitally acquired using Aperio GT 450 DX. Image analysis was performed using ImageJ software (NIH, USA). Color deconvolution was applied to isolate the DAB signal, and uniform threshold parameters were applied across all samples. The total number of CD68-positive cells was automatically quantified and normalized to the total analyzed tissue area. Results are expressed as CD68-positive cells per mm. The entire viable tissue area was included in the analysis. To strictly avoid pseudoreplication, cell counts from multiple histological sections (technical replicates) derived from the same irradiated flank were averaged. Consequently, each final data point represents a single, independent irradiated lesion (n=2 independent lesions per treated group; n=4 lesions for the 7-day control group). Given the exploratory pilot nature of this assay, data are presented to illustrate qualitative temporal trends, and no inferential statistical tests were applied.

Results

Tumor growth

As shown in Fig. 2, tumor volume measurements illustrate distinct treatment responses across the experimental groups. In particular, the CONV 25 Gy group exhibited complete tumor control throughout the observation period, with no detectable progression. The MBRT 25 Gy group showed good tumor suppression up to 21 days post-treatment, although a gradual increase in tumor volume became evident thereafter in individual trajectories, suggesting a partial but not sustained response. Conversely, both the MBRT 30 Gy and CONV 15 Gy groups demonstrated effective early tumor control.

Fig. 2.

Tumor volume measurements over time for single-dose treatment groups (CTRL, CONV, and MBRT). Tumor volumes were measured using digital calipers. Animals were proactively euthanized upon reaching predefined humane endpoints (tumor volume threshold, severe skin toxicity, or spontaneous mechanical ulceration) or administratively censored at the end of the experimental cohort. (top) Mean tumor volume over time. To strictly prevent mathematical drop-out bias caused by animal reduction, each mean growth curve is rigorously truncated on the exact day the first animal in that respective group was euthanized or censored. Error bars represent the standard error of the mean (SEM). (bottom) Individual tumor growth trajectories for all mice within each treatment group. All individual group panels are formatted with identical scales for both the X-axis and Y-axis to allow direct, unbiased visual comparisons. A Y-axis break (2500–3500 mm) was implemented to accommodate a single fast-growing outlier in the untreated CTRL group without compressing the visual resolution of the therapeutic response ranges.

To rigorously prevent mathematical drop-out bias, mean tumor volume curves (Fig. 2a) were strictly truncated on the exact day the first animal in each respective cohort was euthanized (e.g., day 7 for the MBRT 30 Gy group). While this ethical truncation limits the evaluation of long-term trends in the averaged format, the individual growth trajectories (Fig. 2b) clearly demonstrate that both MBRT 30 Gy and CONV 15 Gy maintained reduced tumor volumes comparable to one another over an extended period. In contrast, individual plots reveal that the MBRT 25 Gy group exhibited a more pronounced volume increase compared to CONV 15 Gy, highlighting a distinct dose-dependent response in the MBRT treatment arm. Moreover, the CONV 10 Gy group did not exhibit appreciable tumor growth inhibition, reinforcing the notion that this lower dose is insufficient to induce meaningful tumor control under the experimental conditions.

MBRT results in reduced severe skin toxicity compared to CONV irradiation at high doses

We tested different radiation doses for both modalities to directly compare the biological effects of MBRT and CONV irradiation. Skin toxicity was evaluated using a semi-quantitative scoring system (0-3; 0 = no lesion, 3 = severe ulceration). Mice were monitored daily during the first 10-14 days post-irradiation and 2-3 times per week thereafter. As predefined in the methodology, animals euthanized due to spontaneous mechanical tumor ulceration rather than radiation damage were excluded from this specific scoring to prevent confounding data. Histologic examination further confirmed the macroscopic observations by revealing tissue-level alterations associated with radiation injury. These included depigmentation, alopecia, desquamation, and ulceration, each feature reflecting varying degrees of damage to the epidermal and dermal layers. These microscopic findings provided structural validation for the clinical scoring system, strengthening the reliability of our toxicity assessment.

Figure 3 presents the results of acute radiation toxicity. In the CONV 25 Gy group, 71.4% (n=5 of 7) of the mice experienced severe damage, 14.3% (n=1 of 7) showed moderate damage, and 14.3% (n=1 of 7) showed mild damage. In the CONV 15 Gy group, 56.2% (n=9 of 16) of mice showed no damage, 25.0% (n=4 of 16) had mild damage, and 18.8% (n=3 of 16) experienced severe damage. Mice treated with CONV 10 Gy exhibited no skin lesions (n=0 of 7). The MBRT 25 Gy group showed no signs of skin or leg damage (n=0 of 6), while 60.0% (n=6 of 10) of mice in the MBRT 30 Gy group had no skin damage, 20.0% (n=2 of 10) had mild lesions, and 20.0% (n=2 of 10) had severe lesions.

Fig. 3.

Assessment of acute skin toxicity following CONV and MBRT irradiation. (a) Representative images of skin lesions observed in treated mice. (b) Schematic representation of the skin toxicity scoring system based on macroscopic features (depigmentation, alopecia, desquamation, ulceration). (c) Semi-quantitative analysis of skin lesion severity across treatment groups. The y-axis indicates the percentage of animals per severity category (0-100%). (d) Summary table detailing the incidence of severe skin damage (Grade 3) and the final evaluated sample sizes for each treatment cohort. Animals were evaluated daily during the first 10-14 days post-irradiation and 2-3 times per week thereafter. To ensure accuracy, animals euthanized due to spontaneous mechanical tumor ulceration were explicitly excluded from this quantitative assessment.

Survival

To evaluate the therapeutic impact of MBRT compared with CONV on true loco-regional tumor control, tumor-specific survival curves were generated (Fig. 4). In this analysis, events were strictly defined as reaching the tumor volume endpoint, while animals euthanized due to radiation-induced toxicity, mechanical ulceration, or administrative end-of-study were right-censored. A comprehensive breakdown of the specific causes of euthanasia and the exact number of censored events for each individual animal and treatment group is provided in Supplementary Table S1.

Fig. 4.

Tumor-specific survival analysis following CONV and MBRT treatments. Kaplan-Meier curves were generated to evaluate true loco-regional tumor control. Events were strictly defined as reaching the predefined tumor volume endpoint (1,500 mm). Animals euthanized prior to this endpoint due to radiation-induced skin toxicity, mechanical ulceration, or administrative end-of-study were treated as right-censored observations (indicated by tick marks on the curves). MBRT 30 Gy and CONV 15 Gy demonstrated comparable efficacy in delaying tumor progression, with no significant difference between their survival distributions (p = 0.2804, Log-rank test). Both treatments provided a statistically significant survival benefit compared to the untreated control group (p = 0.0014 for MBRT 30 Gy; p < 0.0001 for CONV 15 Gy). Notably, the apparent 100% tumor-specific survival probability observed in the CONV 25 Gy group is entirely attributable to the premature right-censoring of all animals in this cohort prior to tumor progression, predominantly due to dose-limiting skin toxicity. Detailed event and censoring counts for each group are provided in Supplementary Table S1.

As shown in Fig. 4, the survival profiles for the MBRT 30 Gy and CONV 15 Gy groups exhibited a similar trend, with overlapping curves that suggest comparable overall efficacy in delaying tumor progression. Statistical comparison using the Log-rank test confirmed this observation, yielding no significant difference between these two treatment modalities at the doses tested (p = 0.2804). When the survival curves of the MBRT 30 Gy and CONV 15 Gy groups were compared individually with the untreated control group, marked differences emerged. The Log-rank test demonstrated that both treatments provided a substantial and statistically significant survival benefit relative to untreated animals (p = 0.0014 for MBRT 30 Gy; p < 0.0001 for CONV 15 Gy).

Notably, the CONV 25 Gy group displayed a 100% tumor-specific survival probability (flat curve), as no animals reached the tumor volume endpoint. However, this apparent loco-regional control was accompanied by prohibitive normal tissue damage. As detailed in Supplementary Table S1, 5 out of 7 mice (71%) had to be prematurely right-censored due to the rapid onset of severe, dose-limiting skin toxicity. The remaining 2 mice (29%) were administratively right-censored without tumor progression upon the scheduled termination of their respective experimental cohorts.

Ex vivo culture of skin treated with CONV and MBRT in a static system to investigate the tissue-sparing effect

To evaluate the tissue-sparing effect of MBRT compared with CONV, preliminary control experiments were performed using single-dose CONV (15 Gy) and MBRT (30 Gy) exposures (see Fig. 5a–c). These studies aimed to characterize radiation-induced inflammatory dynamics and establish dose-dependent differences in normal tissue response. Skin samples were analyzed histologically and immunohistochemically for macrophage infiltration using CD68 staining. Seven days after irradiation, CONV-treated skin showed persistent inflammatory damage, characterized by dense infiltration of CD68-positive macrophages within the dermis and subcutaneous tissue. In contrast, MBRT-treated samples exhibited only transient macrophage recruitment, with CD68-positive cell levels returning to baseline values comparable to untreated controls by day 7 (see Fig. 5b,c). Based on these findings, combined treatment regimens were designed to exploit the favorable inflammatory responses induced by MBRT. Two combination protocols were tested: MBRT 30 Gy + CONV 5 Gy and MBRT 25 Gy + CONV 5 Gy. In both cases, the MBRT fraction was delivered first, followed seven days later by the CONV exposure. This timing was selected to coincide with the recovery phase of macrophage activation observed after MBRT.

Fig. 5.

Ex vivo skin response and combined MBRT-CONV treatment effects. (a) Experimental design for ex vivo skin culture and combined MBRT-CONV treatment. (b) Quantification of CD68-positive macrophages by immunohistochemistry at 3 hours and 7 days post-irradiation. (c) Representative immunohistochemical images of CD68 expression in healthy skin sections. Ex vivo analysis revealed a marked increase in CD68-positive macrophages 7 days after CONV, whereas MBRT-treated samples showed CD68 levels comparable to controls. Three hours post-irradiation, both CONV and MBRT groups exhibited elevated macrophage counts, while the combined MBRT-CONV group maintained low levels at this time point. By day 7, macrophage numbers remained elevated in CONV-treated tissues but returned to baseline in MBRT and combined treatment groups. Data points represent independent irradiated flanks (lesions). Technical replicates (multiple histological sections per flank) were averaged to yield a single value per lesion to avoid pseudoreplication. Sample sizes: N=1 mouse per treated group (n=2 irradiated flanks); N=2 mice for the CTRL 7-days group (n=4 flanks). Given the exploratory pilot nature of this assay, no inferential statistical tests were applied.

Histopathological analysis

Three hours after radiotherapy, a clear increase in CD68-positive macrophages was detected in both the CONV and MBRT treatment groups (see Fig. 5b,c), indicating an acute inflammatory response triggered shortly after irradiation. In contrast, the combined treatment group showed minimal macrophage activation at this early time point, with CD68 levels remaining close to baseline. This suggests that the interaction of the two modalities in the combined approach may attenuate the immediate inflammatory response typically observed after radiation exposure. By day 7 post-RT, distinct differences in macrophage dynamics emerged between treatment groups. Elevated CD68-positive cell counts persisted in tissues collected from CONV-treated mice, reflecting a sustained inflammatory state and potentially prolonged tissue damage or slower healing. In MBRT-treated samples, however, CD68 levels had returned to values comparable to those of the untreated control group, indicating a more rapid decline in inflammatory activity. This pattern was even more evident in the combined MBRT-CONV treatment group, where macrophage levels remained low, further supporting the hypothesis of accelerated inflammatory resolution. These preliminary observations collectively suggest that MBRT, even when administered at higher doses, may favor a faster normalization of inflammatory markers compared to CONV. The combined treatment appears to enhance this effect even further. These qualitative findings reinforce the potential of MBRT to reduce prolonged inflammatory responses in healthy tissues.

Delayed MBRT-CONV therapy enhances tumor control with reduced toxicity

In this model, CONV 15 Gy and MBRT 30 Gy produced comparable tumor control while maintaining similar levels of skin toxicity, and were therefore considered biologically equivalent dose conditions for the two irradiation modalities. For this reason, these dose levels were used as the reference for subsequent comparisons and for the design of the combined MBRT-CONV treatment protocols. To further improve tumor suppression while maintaining a favorable toxicity profile, we implemented a delayed combined treatment strategy. Mice treated with the combined regimen received either 25 Gy MBRT + 5 Gy CONV or 30 Gy MBRT + 5 Gy CONV. In both cases, the initial MBRT fraction delivered a spatially fractionated dose distribution that minimized normal tissue damage, while the subsequent CONV fraction was intended to reinforce tumoricidal efficacy during the post-MBRT recovery phase. Animals tolerated the combined treatments well, exhibiting minimal acute skin reactions and no severe or persistent lesions. Among the tested regimens, the 30 Gy MBRT + 5 Gy CONV combination achieved the most pronounced tumor growth delay and survival benefit, while maintaining low incidence of toxicity (see Fig. 6). These findings suggest that a temporally separated MBRT-CONV sequence can exploit the transient window of enhanced tissue tolerance following MBRT, providing a promising strategy to expand the therapeutic window of radiotherapy for aggressive tumors such as GBM.

Fig. 6.

Combined MBRT-CONV treatment protocol and therapeutic response. (a) Schematic representation of the sequential irradiation protocol. A CONV fraction of 5 Gy was delivered 7 days after the initial MBRT fraction. (b) Quantitative assessment of acute skin toxicity across the combined treatment groups compared to untreated controls. (c) Mean tumor volume over time. To rigorously prevent mathematical drop-out bias, mean curves are strictly truncated on the exact day the first animal in each respective cohort was euthanized or censored. Day 1 is defined as the day the initial MBRT treatment was administered. Error bars represent the standard error of the mean (SEM). (d) Tumor-specific survival analysis (Kaplan-Meier curves). Events were strictly defined as reaching the tumor volume endpoint, with right-censoring applied for non-tumor-related mortality or administrative end-of-study. (e) Individual tumor growth trajectories for mice treated with the combined regimens, formatted with identical scales for unbiased visual comparison. Initial sample sizes: n=6 for MBRT 25 Gy + CONV 5 Gy; n=6 for MBRT 30 Gy + CONV 5 Gy.

Discussion

This study demonstrates that MBRT provides a significant tissue-sparing effect compared with CONV radiotherapy, while maintaining relevant anti-tumor activity. In our preclinical GBM model, single high dose MBRT (25-30 Gy) induced substantially less severe skin toxicity than CONV, despite comparable or even higher delivered doses. This observation is consistent with previous preclinical studies showing that SFRT modalities, including MBRT and MRT, can markedly reduce normal tissue complications while preserving tumor control potential^17,21,22.

The pronounced normal tissue protection observed with MBRT is thought to arise from its unique dose distribution, characterized by alternating peak and valley regions. Cells located within the low-dose valleys remain viable and contribute to tissue repair, vascular preservation, and DNA damage recovery in surrounding areas²³. Several independent studies have shown that the preservation of stromal and endothelial cells in valley regions facilitates accelerated wound healing and reduces late radiation toxicity²⁴. For instance, Fazzari and colleagues demonstrated that spatially fractionated minibeam irradiation induces only transient vascular permeability, with normalization occurring within 24 hours, supporting rapid microvascular recovery following MBRT²³. Similarly, in epidermal 3D models, valley regions exhibit minimal γ-H2AX and 53BP1 foci formation, indicating limited and rapidly repairable DNA damage.

Consistent with these mechanisms, our preliminary ex vivo findings indicated a trend toward a faster resolution of inflammatory responses following MBRT compared with CONV. This was qualitatively reflected by the transient recruitment of CD68-positive macrophages following MBRT, in contrast to the more persistent inflammatory infiltration observed after conventional irradiation. The rapid resolution of macrophage-driven inflammation suggests that MBRT better preserves microvascular and stromal integrity, creating a tissue environment less prone to chronic inflammatory signaling. Similar immunomodulatory effects of spatial fractionation have been described in other preclinical studies, where preservation of viable stromal and immune niches supports balanced cytokine release and improved tissue repair processes^22,23,25.

Despite its favorable toxicity profile, MBRT alone was less effective than CONV in achieving long-term tumor control, particularly at 25 Gy. This limitation is likely related to the intrinsic dose inhomogeneity within the tumor volume, where cells located in valley regions receive substantially lower doses. Similar observations have been reported in other tumor models, where MBRT induces tumor growth delay but may require combination strategies to achieve optimal tumor control²⁶. It should be noted that our survival analysis was rigorously designed to evaluate true loco-regional tumor control (tumor-specific survival) by treating non-tumor-related mortality (e.g., severe toxicity or mechanical ulceration) as competing risks through right-censoring. This approach mathematically isolates the anti-tumor efficacy of the treatments, demonstrating that MBRT provides solid tumor control while simultaneously preventing the early toxicity-induced drop-outs strictly observed with high dose CONV.

In this context, the sequential combination of MBRT followed by a conventional irradiation fraction improved tumor suppression while maintaining reduced normal tissue toxicity. Notably, the 30 Gy MBRT + 5 Gy CONV regimen produced a substantial delay in tumor growth with minimal incidence of severe skin lesions. These findings support the hypothesis that MBRT may act as a priming modality, reshaping the tumor microenvironment in a way that enhances the efficacy of subsequent treatments.

Importantly, the conventional dose used in the combined regimen (5 Gy) was intentionally kept relatively low. The aim was to evaluate whether a modest secondary irradiation delivered during the post-MBRT recovery phase could enhance tumor control while preserving the favorable normal tissue tolerance associated with MBRT. Different dose levels or fractionation schedules may further optimize this strategy and should be explored in future studies.

Emerging evidence indicates that spatially fractionated radiation can modulate both local and systemic immune responses. For example, Rivera et al. reported that MBRT combined with PD-L1 checkpoint inhibition induced stronger abscopal responses than uniform irradiation, suggesting that spatial fractionation may promote beneficial immunological reprogramming²⁶. Our findings support a related concept: the initial MBRT exposure may induce favorable immune and vascular adaptations that increase tissue tolerance and improve responsiveness to subsequent irradiation.

Macrophages may play an important role in mediating these effects. In glioblastoma, tumor-associated macrophages and microglia are major contributors to tumor progression, angiogenesis, and treatment resistance²⁷. While our preliminary macrophage analysis was specifically conducted on healthy skin to evaluate normal-tissue sparing, the observation of a reduced and transient macrophage recruitment after MBRT raises an intriguing hypothesis. If a similar rapid resolution of inflammation occurs within the tumor stroma, MBRT could theoretically help attenuate the chronic pro-tumoral inflammatory environment typically associated with GBM. Further mechanistic studies, including characterization of macrophage polarization states, will be required to determine whether MBRT promotes a shift toward anti-tumoral (M1-like) immune phenotypes.

Several limitations of this study should be acknowledged. First, the experiments were performed using a subcutaneous flank GBM model rather than an orthotopic intracranial model. While the flank model allows precise tumor volume measurements and facilitates irradiation procedures, it does not fully replicate the complex brain microenvironment and anatomical constraints of intracranial glioblastoma. Future studies using orthotopic models will therefore be important to validate the translational relevance of these findings. Second, macrophage infiltration was evaluated using an ex vivo short-term culture approach prior to fixation. Although this method allowed standardized assessment of early inflammatory responses, it does not fully reproduce the in vivo microenvironment over longer time frames. Complementary analyses on directly fixed tissues at multiple time points could provide a more detailed characterization of macrophage dynamics following MBRT.

Finally, the experimental cohorts were conducted sequentially rather than in parallel. While treatment protocols and assessment criteria were standardized, the MBRT-CONV combination groups were evaluated after the single-modality experiments. This design may introduce potential bias when comparing results across cohorts and should therefore be interpreted with caution. Future studies including fully parallel experimental groups would strengthen the robustness of these observations. In addition to sequential MBRT-CONV strategies, other MBRT delivery approaches have been proposed, including fractionated MBRT regimens or cross-fire irradiation geometries, which may improve dose homogeneity within the tumor while preserving the normal tissue-sparing properties of spatial fractionation.

Overall, our findings highlight the potential of MBRT as a complementary radiotherapy strategy capable of widening the therapeutic window. While MBRT alone may not provide optimal tumor control, its strong normal tissue sparing effect and its apparent ability to modulate inflammatory and immune responses support its integration into combined or sequential treatment approaches.

Conclusion

In summary, our findings show that MBRT provides a pronounced normal-tissue sparing benefit compared with conventional radiotherapy while retaining meaningful anti-tumor efficacy in a preclinical GBM model. Although high single-dose MBRT did not achieve the same long-term tumor control as CONV, its markedly reduced toxicity and its capacity to facilitate rapid inflammatory resolution underline its therapeutic relevance. Notably, the specific combination of a 30 Gy MBRT fraction followed by a 5 Gy CONV fraction further enhanced tumor suppression without diminishing normal-tissue protection, suggesting that spatially fractionated irradiation may prime both tumor and stromal compartments for improved responsiveness to additional therapies. These results strengthen the rationale for MBRT as a complementary approach capable of broadening the radiotherapy therapeutic window. Future work should focus on optimizing dose-scheduling strategies, exploring combinations with immunotherapies and radiosensitizers, and clarifying the vascular and immune mechanisms driving MBRT’s selective tissue protection. Overall, this study supports continued translational advancement of MBRT for radioresistant tumors such as GBM.

Supplementary Information

Supplementary Information. ^{(90.4KB, pdf)}

Author contributions

Data curation, formal analysis, data interpretation, investigation, methodology, visualization, and writing - original draft, F.V.; formal analysis, data interpretation, review and editing, S.P.; review and editing, M.M., M.L., L.A.; conceptualization, funding acquisition, project administration, supervision, and writing - review and editing, A.E.S., C.F. All authors reviewed the manuscript.

Funding

Funded by the European Union - Next Generation EU - NRRP M6C2 - Investment 2.1 Enhancement and strengthening of biomedical research in the NHS, Codice progetto: PNRR-POC-2022-12376062, CUP master: C43C22001370007. The research leading to these results has also received funding from AIRC under IG 2022, ID. 27336, project PI: Antonello E. Spinelli.

Data availability

Data utilized and/or analysed during this study are available from the corresponding author upon reasonable request.

Declarations

Competing interests

The authors declare no competing interests.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1038/s41598-026-56825-7.