Evaluation of Proton Minibeam Radiotherapy on Antitumor Immune Responses in a Rat Model of Glioblastoma

The authors demonstrate how proton minibeam radiotherapy (pMBRT) generates antitumor immunity through CD4⁺ and CD8⁺ T cells tumor infiltration. Single-cell transcriptomics reveals enhanced antibody production, elevated expression of chemotactic cytokines, and activation of interferon signaling pathways after pMBRT.

Abstract

Treating radioresistant tumors like glioblastoma multiforme remains a challenge exacerbated by their immunosuppressive nature. Radiotherapy (RT) plays an immunomodulatory role, exerting both immunosuppressive and immunostimulatory effects. The nature of these effects depends on the total dose, dose per fraction, dose delivery method, and treatment length. Hypofractionation is observed to tip the balance toward immune stimulation. However, the use of hypofractionation is restricted in bulky tumors, such as gliomas, because of the high risk of toxicity. Therefore, finding new strategies leading to more favorable immune responses while reducing normal tissue toxicities could improve cancer treatment. In this study, we examine antitumoral immune responses to proton minibeam RT (pMBRT). Its immunomodulatory effects are not fully understood. To explore this, we conducted an in-depth characterization of the immune response to a curative dose of pMBRT in a preclinical orthotopic rat model of glioblastoma. Our findings revealed a close association between pMBRT and the immune response. pMBRT increased lymphocyte density in tumors more effectively than conventional proton therapy. Single-cell transcriptomics identified several immune cell types and unique transcriptional changes in tumor immune cells following pMBRT, including increased antibody production, chemotactic cytokine expression, and IFN responses. These results underscore the critical role of adaptive immunity, specifically T cells, in pMBRT's mechanism. The potential of pMBRT to trigger an antitumor immune response in a single RT session with minimal damage to healthy tissue makes it a promising candidate for future clinical trials and radioimmunotherapy combinations.

Introduction

Cancer accounts for 25% of all deaths in Europe, and radiotherapy (RT) is a cornerstone of cancer treatment. However, normal tissue tolerances still compromise an effective RT treatment of many late-stage tumors, radioresistant bulky tumors (1), recurring tumors (2), and brain and pediatric tumors (3, 4). Among them, glioblastoma multiforme (GBM) ranks among the most lethal forms of human cancer. The current standard-of-care for GBM, namely, surgery followed by RT and chemotherapy with temozolomide (4), only modestly improves patient survival, and survivors frequently suffer from neurologic sequelae. Novel therapeutic strategies, leading to different biological mechanisms than conventional RT like immune and/or vascular effects, could offer an efficient treatment for those currently hopeless cases aforementioned.

RT has important effects on normal and tumor vasculature, as well as complex and dynamic interactions with the immune system. Indeed, although RT can lead to radiation-induced lymphopenia and immunosuppression, it has revealed an immunomodulatory capacity to potentiate the efficacy of the treatment (5–10). There is a current collective effort in the community to find the best radiation scheme and configuration to induce an optimized antitumor response, and different combination strategies with immunotherapies are being proposed (7, 11–19). Recent clinical trials showed dynamic immune effects of ablative stereotactic RT that influenced the treatment outcome (20).

In particular, partial irradiations and clinical spatial fractionation techniques like lattice RT or GRID RT, in which the tumor is not homogeneously irradiated, have shown significant immune activation and abscopal events and shrinkage of tumors outside of the irradiation field in treated patients (21, 22). Minibeam RT (MBRT), in which the dose is deposited along paths of submillimetric parallel high-dose beams (peaks) and low doses (valleys) in the rest of the tissue, has been shown to generate T-cell immune responses, providing long-term immunization against the tumor (23).

Proton MBRT (pMBRT) combines the lower toxicity of submillimetric spatial fractionation with the therapeutic benefits of proton therapy (PT; refs. 24, 25). PT can better circumvent hematopoietic organs at-risk and blood reservoirs because of its unique dose deposition, diminishing radiation-induced lymphopenia and increasing prognosis in clinical settings (26–28). An increased immunogenicity for PT has been proposed (29, 30). Yet, PT can also induce side effects including brain necrosis (31, 32). In pMBRT, the neurologic toxicities associated with high doses of cranial PT are reduced (33–35), and radiation-induced fibrosis in thoracic PT in mice is also decreased (36). Importantly, pMBRT has been reported to maintain or enhance tumor control in GBM rat models (37–40). The efficacy of pMBRT does not depend on achieving total coverage of lethal dose radiation in the complete tumor volume, as heterogeneous dose distributions in the target demonstrated excellent tumor control (37, 38).

Thus far, minibeams showed an increase in the therapeutic index in glioma-bearing rodents (41–43). A recent canine clinical trial (44) achieved 71% completed pathologic remission in the MBRT-treated group, without acute toxicity. The two first patients treated with MBRT exhibited no acute toxicity and substantial (>50% of the tumor volume) partial response (45). Although the results to date encourage further clinical trials using minibeams, uncovering the mechanism of pMBRT becomes fundamental to reach the full potential of this technique in future treatments. pMBRT has shown potential to achieve curative doses with minimal toxicity (33, 34, 37, 39), which makes brain tumors, such as GBM and brain metastasis (46), as ideal candidates for this technique. MBRT has also shown the ability to establish long-term antitumor immunity, which can be beneficial for tumors with high recurrence rate like GBM (23).

Therefore, this study was designed to tackle this gap in the literature and offer a comprehensive study of the immune response elicited by pMBRT in GBM after one single nontoxic therapeutic dose, with specific attention to T-cell response. This will also give valuable information for the design of future clinical trials involving spatial fractionation, especially if combination strategies with immunotherapies are envisioned.

Materials and Methods

Study design

The purposes of this study were to assess the importance of tumor immunity for spatial dose fractionation and to offer an exhaustive analysis of the immune response in a “cold” preclinical tumor model that could benefit from the reduced toxicity offered by spatial dose fractionation. To achieve this, we used a well-characterized GBM cell line RG2-[D74] orthotopically implanted in the brain of rats and irradiated the cranium with homogeneous [conventionally irradiated PT (conv-PT)] or heterogeneous (pMBRT) proton beam irradiation. Observations of tumor immune response included quantification using flow cytometry and spatial distribution using immunofluorescence (IF) multiplex staining and single-cell transcriptomic analysis of tumor-infiltrating immune cells. Group randomization was performed based on bioluminescence imaging (BLI) signal before irradiation, ensuring that each experimental group exhibited a similar average BLI signal.

The sample size for in vivo studies was estimated on the basis of prior experience with the experimental model. The survival and IHC studies were performed in two independent experimental replicates. The flow cytometry and transcriptomic studies were performed in at least three independent experimental replicates. All experiments had an equivalent number of animals per group per experiment apart from the transcriptomic studies. The experimental group of the animal was annotated in the animal facility, and all the experimental groups were randomly paired in cages and treated. No outliers were removed unless a technical error was noted during sample manipulation.

Ethics statement

All animal experiments were conducted in accordance with the animal welfare and ethical guidelines of our institution. They were approved by the French Ministry of Research (permits #6361-201608101234488, #23434-2019122418442057, and #36372-2022040609163783). Animals were housed at the Institut Curie animal facility accredited by the French Ministry of Agriculture for performing experiments on rodents. Cages were enriched with cardboard tunnels.

Tumor inoculation

The RG2-[D74] (ATCC CRL-2433, RRID: CVCL_3581) and the F98 (ATCC CRL-2397) glioma cell lines were acquired in 2020, and both transduced with a luciferase gene and GFP reporter gene (RG2-Luc-GFP and F98-Luc-GFP) were used. For the lentiviral transduction, the RG2 and F98 glioma cell lines were seeded in 96-well plates at a density sufficient to reach approximately 80% confluency on the day of transduction. The culture medium was then replaced with fresh medium containing polybrene (TR-1003, Sigma-Aldrich) at a final concentration of 10 µg/mL. Packaged lentivirus (CMV-GFP-T2A-Luciferase, BLIV101VA-1, System Biosciences) was added at varying multiplicities of infection (5, 10, and 20) to determine optimal transduction efficiency. The plates were gently swirled to ensure uniform mixing and incubated for 72 hours at 37°C in a humidified atmosphere with 5% CO₂. During this incubation period, the viral genome was integrated into the host cell genome. GFP reporter expression was monitored using fluorescence microscopy. After 72 hours, the medium was aspirated, and the cells were washed with PBS. Transduced cells were subsequently expanded, and FACS was performed to isolate and enrich the GFP-positive population. For intracranial implantation, cells were cultured in DMEM (41965-039, Thermo Fisher Scientific) supplemented with 1 mmol/L sodium pyruvate (11360-039, Thermo Fisher Scientific), 10% FCS (CVFSVF00-01, Eurobio Scientific), 1% amphotericin B (1,25 µg/mL, 15290-026, Thermo Fisher Scientific), and 1% penicillin–streptomycin (100 U/mL and 100 µg/mL, respectively, 15140-122, Thermo Fisher Scientific). RG2 medium was supplemented with 1 mmol/L HEPES (15630-049, Thermo Fisher Scientific). Cells were cultured at 37°C with 5% CO₂ and passaged twice weekly. Cells were used between passages 2 and 12 after thawing or collection. Prior to establishing each new liquid nitrogen stock, the cell lines were tested for Mycoplasma contamination using a PCR-based detection kit (Sigma-Aldrich, MP0035). Authentication of the cell lines was performed via short tandem repeat profiling in 2020. The sex of the cell lines is not specified by the supplier. A total of 50,000 RG2-Luc cells were suspended in 5 µL of DMEM (41965-039, Thermo Fisher Scientific) and then injected intracranially into 6-week-old wild-type (WT) Fischer 344 (F344) rats (Janvier Labs) or Foxn1^rnu mutated rats (Fischer Nude rats, Janvier Labs) using a Hamilton syringe through a burr hole in the right caudate nucleus (coordinates relative to bregma were the following: anterior–posterior: −1 mm; median–lateral: +4 mm; dorsal–ventral: −5.5 mm from the skull). The presence of a tumor was confirmed by BLI before irradiation. A same number of F98 cells were implanted and injected intracranially into 6-week-old WT F344 rats (Janvier Labs). The clinical status of the animals along with the duration of the experiment was checked five times per week. Any rat showing classical adverse neurologic signs related to tumor growth or substantial weight loss was humanely euthanized using CO₂ asphyxia.

BLI was conducted using an IVIS Spectrum system (Lumina LT series III, PerkinElmer). The software employed was Living Image (64-bit) version 4.7.4 advanced acquisition and analysis tools for IVIS imaging systems.

D-Luciferin at a concentration of 150 mg/kg was injected intraperitoneally, and bioluminescence was measured using the IVIS spectrum after 25 minutes (peak of bioluminescence). Confirmation of tumor presence relied on the detection of a bioluminescent signal surpassing the background level. Consequently, only rats displaying a BLI signal significantly exceeding the background signal on the day prior to irradiation were enrolled in the study.

Tumor presence and volume quantification were performed by means of MRI in the F98-bearing rats. A catheter was inserted into the tail vein for contrast agent administration. A 7-T preclinical magnet (Bruker Avance Horizontal 7-T Bruker, Inc.) equipped with a 35 mm-diameter “bird-cage” antenna was employed. The employed T2 and T1 sequences are described in previous works (33). Tumor volume was measured 24 hours before irradiation.

Irradiation and dosimetry

All the irradiations were performed at the pencil beam scanning beamline of the Orsay Proton Therapy Center 14 days after tumor inoculation. The proton energy at the isocenter was 100 MeV, and the animals were irradiated at the plateau of the Bragg curve. In pMBRT irradiations, a 65-mm-thick divergent brass multislit collimator was attached to the nozzle exit to shape the planar minibeams (47). Dosimetry was performed following the methodology previously described (48). Radiochromic films (OC-1) were then placed on the skin for quality assurance of the irradiation. The average dose in pMBRT was defined as the mean dose between the first and last peaks at the tumor depth. The doses received are provided in Fig. 1A.

Figure 1.

Dosimetry for pMBRT and analysis of survival in WT and immunodeficient rats. A, Scheme of irradiation used throughout the study. B, Percentage of survival of rats bearing RG2 GBM. Left, comparison of the nonirradiated control (black, n = 12) and pMBRT-irradiated groups (blue, n = 10) in immunocompetent F344 rats. Right, comparison of the nonirradiated control (black, n = 3) and pMBRT-irradiated groups (blue, n = 6) in Fischer Nude rats. Statistical analysis was done using the log-rank (Mantel–Cox) test.  c-t-c, center-to-center distance between the peaks; FWHM, full width at half maximum; ns, not significant; PVDR, peak-to-valley dose ratio.

The choice to deliver the same average dose of homogeneous PT was based on the team’s experience that 30 Gy of conv-PT constitutes a cytotoxic dose for the tumor with a similar survival rate as 30 Gy in pMBRT (30%; ref. 37). Figure S1 shows a side-to-side comparison of survival curves of RG2-bearing rats irradiated with either conventional proton therapy (CPT) or pMBRT at 30 Gy. Survival is statistically equivalent.

All irradiations were performed under anesthesia to avoid animal movement. The rats received isoflurane at a concentration of 2.5% in a 50:50 mix of air and medical oxygen.

Processing of tumor samples for flow cytometry

Fourteen to 16 days after tumor implantation (control samples) or 7 days after irradiation, tumors were carefully extracted and dissected from the rats and washed in 4°C Dulbecco’s PBS (D-PBS), weighed, and immediately minced into pieces and digested in a digestion mix containing 1 mg/mL collagenase D (Roche), 0.1 mg/mL DNase I (Sigma), and 3% FCS in D-PBS for 30 minutes at 37°C in agitation. Samples were then mechanically disrupted with a 2-mL syringe piston on top of 100-mm filters to obtain a single-cell suspension in FACS buffer containing PBS with 0.5% BSA and 2 mmol/L ethylenediaminetetraacetic acid. Cells were mixed with 30% isotonic Percoll solution, centrifuged to remove dead cells and debris, and subsequently resuspended in FACS buffer with the anti-CD32 (FcγRII) blocking agent.

Cells were incubated in a viability stain at a 1:1,000 dilution (FVS780, BD Biosciences, RRID: AB_2869673) and immunolabeled in FACS buffer with the appropriate antibodies. Counting beads were added to each sample before flow cytometry (CountBright Plus Absolute Counting Beads, Thermo Fisher Scientific). Cell profiles were analyzed using a flow cytometer (LSRFortessa, BD Biosciences) and FlowJo v10.6 software (BD Life Sciences). Gating strategies are provided in the Supplementary Material (Supplementary Fig. S2).

Analysis of intratumoral cytokines

After tissue collection and rinsing in cold D-PBS, the tumors were weighed after removing the excess water. Tumors were placed in M tubes (Miltenyi Biotec) on ice with lysis buffer containing Cell Extraction Buffer (RIPA; Invitrogen, 89900), Pierce Protease Inhibitor Mini Tablets, EDTA-free (Thermo Fisher Scientific), and phenylmethylsulfonyl fluoride (PMSF) Protease Inhibitor (Thermo Fisher Scientific, 36978). Samples were lysed using the protein program in a tissue dissociator (gentleMACS, Miltenyi Biotec) at maximum speed for 1 minute and immediately centrifuged at 4,000 rpm for 5 minutes at 4°C. The supernatant was collected, and aliquots were prepared on ice and stored at −80°C. Cytokines were measured using chemiluminescence-based V-PLEX Proinflammatory Panel 2 (rat) Kit (MSD, K1559D, RRID: AB_2916285).

Histopathology and multiplexed IF

Two or 7 days after irradiation, brains were carefully removed and fixed in zinc formalin fixative (Sigma-Aldrich, Z2902) for histopathologic and multiplex IF analysis. The brains were embedded in paraffin wax, and serial 5-μm-thick coronal sections were cut through the tumoral area. One slide was stained with hematoxylin and eosin for histopathologic evaluation of the lesions.

Two multiplex IF stainings were done using a panel of four antibodies [CD3 (polyclonal, A0452, Dako, RRID: AB_2335677), CD4 (clone D7D2Z, 25229, Cell Signaling Technology, RRID: AB_2798898), CD8 (clone OX-8, ab33786, Abcam, RRID: AB_726709), and FoxP3 (clone EPR22102-37, ab215206, Abcam, RRID: AB_2860568)] and three antibodies [CD68 (clone ED1, MCA341R, Bio-Rad, RRID: AB_2291300), CD8 (clone OX-8, ab33786, Abcam, RRID: AB_726709), and Pax5 (clone EPR3730(2), ab109443, Abcam, RRID: AB_10862070)] and Opal Multiplex IHC (Akoya Biosciences), which was optimized in-house. Successive cycles were done by incubating with each primary antibody diluted in normal goat serum 5% (S2000-100, Dutscher). The slides were then incubated with secondary antibody at a 1:300 dilution (goat anti–rabbit P0448, Dako, RRID: AB_2617138 or goat anti–mouse P0447, Dako, RRID: AB_2617137) followed by incubation with Opal fluorophore diluted at 1:100 in Plus Amplification Diluent (FP1498, Akoya Biosciences). These cycles were repeated for subsequent antibodies. Finally, the slides were stained with 4′,6-diamidino-2-phenylindole (DAPI; D1306, Invitrogen, RRID: AB_2629482) at a 1:1,000 dilution and mounted using an in-house preparation of Mowiol.

The Mantra 1 quantitative pathology workstation (PerkinElmer) was used to acquire images using a fluorescence protocol (Mantra Snap 1.0.4 software, Akoya Biosciences). Three sets of images were acquired per slide: one set with images centered on the edge of the tumor, one set of images further from the tumoral tissue, and one set of images acquired more centrally within the tumor. Cells were counted in three fields for each region, and the average was taken for graph representation. inForm advanced image analysis software (inForm 2.4.4 software, PerkinElmer) was used to process and analyze the multispectral images.

Processing of tumor samples for single-cell RNA sequencing

Fourteen days after tumor implantation (control samples) or 7 days after irradiation, tumors were harvested and immediately placed in a C tube (Miltenyi Biotec) to be processed enzymatically and mechanically by incubation in digestion mix for 40 minutes at 37°C in a tissue dissociator (gentleMACS, Miltenyi Biotec). The number of biological replicates was four nonirradiated controls, three conv-PT–treated tumors, and four pMBRT-treated tumors. One technical replicate per sample was employed. Samples were prepared in six different experiments and sequenced in three separate sequencing rounds.

The resulting single-cell suspension was resuspended in FACS buffer, blocked with purified anti-CD32 (FcγRII), and incubated with anti-CD45 magnetic microbeads (Miltenyi Biotec) for 30 minutes. CD45-positive cells were isolated using MACS separator kit (Miltenyi Biotec) and processed following 5′ VDJ-seq kit (10x Genomics).

Single-cell RNA sequencing analysis

Single-cell RNA sequencing (RNA-seq) fastq reads were processed using the Cell Ranger Single-Cell Software Suite (version Rnor6.0-3.0.2) to perform quality control, barcode and unique molecular identifiers (UMI) processing, and gene expression quantification through single-cell 5′ methodology. Reads were mapped to the Rattus norvegicus transcriptome reference. Single-cell gene expression data for all samples were imported within R (v4.2.1), and downstream analyses were performed using the Seurat R package (v4.3.0.1). For each sample, mitochondrial content was assessed using the PercentageFeatureSet function from Seurat providing as a pattern for the following regular expression: Mrps|Mrpl|Mt-|ND|COX|ATP6|ATP8|CYTB, mapping all mitochondrial genes present in the rat transcriptome gene annotation. Only cells with more than 200 features expressed, more than 750 UMIs but less than 50,000 UMIs, and less than 5% of mitochondrial gene content were kept for downstream analyses. DoubletFinder (v2.0.3) was then used to identify singlet and doublet cells (doublet rate expectation of 7.5%). Only singlet cells were kept for downstream analysis, and Seurat objects for each sample were subsequently merged into one object. After log-normalization, the 5,000 most variable genes were scaled and used to compute principal component analysis. Batch-effect correction was performed using the Harmony R package (v0.1.1) providing the following metadata as batches to correct: sample ID, run ID, and experimental preparation ID. Using harmony embeddings, top 40 components were chosen to compute Uniform Manifold Approximation and Projection for Dimension Reduction and perform clustering steps. Clusters were first identified at resolution 0.7. Class identification per cluster was associated by their unique gene expression using the FindAllMarkers function from Seurat and the expression of well-known sets of genes linked to cell types or biological processes. After filtering out all nonimmune cell clusters (epithelial and stroma cells), principal component analysis, harmony batch correction, Uniform Manifold Approximation and Projection for Dimension Reduction, and cluster identification were computed again, and final clusters were identified using 40 harmony components at resolution 0.7, leading to the 18 final cell clusters.

To minimize effects linked to the number of cells per sample and condition, statistics on cell proportions were computed after performing 1,000 downsampling iterations to 2,200 cells per sample. For each sample and each cluster, the median frequency from these 1,000 downsampling observations was used as the final cluster proportion of each sample. Afterward, t tests were used to compare cluster abundancy for each condition.

Differential expression analysis for all clusters and comparisons of interest was performed using the FindMarkers function from Seurat. The criteria to consider genes as differentially expressed were |log₂FC| ≥ 0.411 and adjusted P value < 0.05. Pathway analysis or enrichment analysis (EA) was explored on every list of differentially expressed genes (DEG) using the Gene Ontology biological process annotation database from the enrichGO function on the clusterProfiler R package (v4.12.0). Dot plots showing the most representative significant term from the EA were generated with the ggplot function from the ggplot2 R package (v3.5.1). The procedure adopted for the differential expression analysis and EA in the T-cell subclustering followed the same procedure as for the object containing the 18 total clusters.

Cell–cell communication analysis was performed using the CellChat R package (v2.1.2) to explore intercellular signaling interactions across the three experimental groups. Following data preprocessing, the createCellChat function was employed to initialize a CellChat object, using the complete “CellChatDB.mouse” as the reference database for cell–cell communication analysis. Gene overexpression and interaction overexpression were identified using identifyOverExpressedGenes and identifyOverExpressedInteractions, respectively. Communication probabilities were then calculated using computeCommunProb. Following calculation, signaling interactions were filtered with a minimum cell threshold (filterCommunication, min.cells = 10), followed by pathway-level probability computation (computeCommunProbPathway). Aggregated networks were finally generated using AggregateNet, with results stored in the netP slot.

Statistical analysis

Survival curves were statistically analyzed using the log-rank (Mantel–Cox) test. Flow cytometry and cytokine statistical analyses were performed using ordinary one-way ANOVA and uncorrected Fisher least significant difference (LSD) test for multiple comparisons. IF quantification statistical analysis was performed with a linear mixed model with repeated measures using the restricted maximum likelihood method for the (co)variance component estimation procedure to account for multiple data points per animal (nonindependent).

All statistical analyses were performed using GraphPad Prism 10 software (GraphPad Software), except for the data from the histopathologic and the transcriptomics analyses, which were analyzed using R version 4.4.0 (The R Foundation for Statistical Computing) and RStudio software version 2024.04.1 + 748 (Posit Software, PBC). The results are presented as sample mean and SD as error bar.

Results

pMBRT is dependent on the action of T cells

To test the immune mechanism necessity of pMBRT, WT immunocompetent F344 rats or T cell–deficient athymic rats of the same genetic background (Fischer Nude) were challenged with the rat orthotopic GBM model RG2 and treated with pMBRT 14 days following implantation to observe the survival curve differences. A single dose of average 30 Gy pMBRT was prescribed allowing for a heterogeneous dose delivery at the tumor depth (Fig. 1A). Whereas F344 pMBRT-treated rats had a significantly higher survival probability than the controls (P value = 0.0001), there was no difference in survival with pMBRT in the Nude groups, indicating that pMBRT relied on T-cell function (Fig. 1B). pMBRT treatment increased the survival at the same probability as conv-PT) with the equivalent average dose (30 Gy; Supplementary Fig. S1; no significant differences in survival between the two irradiated groups). In previous studies, no significant differences in survival were observed when immunocompetent and nude animals were irradiated with conventional RT using an ablative dose in the whole tumor area (22, 23). The surviving animals showed no signs of tumor recurrence 200 days following inoculation.

pMBRT elicits a strong immune response in the tumor microenvironment

As T-cell immunity is involved in the mechanisms of pMBRT, a phenotypic characterization of the ensemble of tumor immune cells was performed. To compare the immune response generated by pMBRT, the same mean dose (30 Gy) was delivered in a homogeneous manner in the tumor as conv-PT. Flow cytometry analysis of tumor was done on the day of irradiation for the control group (nonirradiated) or 7 days after irradiation, as infiltration of T cells into tumors was found to be at its peak (Fig. 2A–L; Supplementary Fig. S2). The results show a significant increase in CD4⁺ T-cell [both conventional and regulatory T cells (Treg)] density in the tumor after both irradiations (Fig. 2A and B) and an increase in CD43⁺ monocytes (Fig. 2K). Interestingly, only pMBRT significantly increased the density of tissue-resident memory CD8⁺ T cells (Fig. 2D), B cells (Fig. 2E), and conventional dendritic cells type 1 (Fig. 2G), suggesting an enhanced antigen presentation and immune memory formation taking place after pMBRT. Similarly, only pMBRT significantly increased the density of CD8⁺ macrophages, a rat-specific type of tumor-associated macrophages (TAM) that express higher RT-1B (antigen presentation molecule, HLA class II). Analysis of a second more immunosuppressive tumor model (F98) yielded similar tendencies, although B cells were practically nonexistent; however, the low number of animals did not allow to study small statistical differences (Supplementary Fig. S3). This model suggests that a minimal number of cells in the environment could be important for pMBRT to have an effect, but more experiments would be needed to reach a definitive conclusion. Analysis of cytokine concentration in the tumor microenvironment revealed a similar increase in proinflammatory IL1β, chemokine (C-X-C motif) ligand (CXCL) 1, and TNFα at 7 days after irradiation (Fig. 2M). No difference in the levels of VEGF or IFNγ was observed in the tissue lysate. However, an early increase in acute inflammatory IL6 was observed only in those treated with conv-PT 24 hours after irradiation, suggesting an early acute inflammation specifically of homogeneous irradiation.

Figure 2.

Analysis of immune cell density and cytokines in the tumor. Density in cells/mg of tumoral tissue of the control group 14 days after implantation (black, n = 12) compared with 7 days after irradiation in the conv-PT (red, n = 12) and pMBRT-treated groups (blue, n = 11): (A) conventional CD25^neg CD4⁺ T cells, (B) CD4⁺ Tregs, (C) CD8⁺ CD103^neg T cells, (D) CD8⁺ tissue-resident memory (TRM) T cells, (E) B cells, (F) NK cells, (G) conventional dendritic cells (cDC) type 1, (H) neutrophils, (I) CD8⁺ macrophages, (J) His48⁺ monocyte/macrophages (mono/mac), (K) CD43⁺ His48^dim monocytes, and (L) CD49d^neg macrophages. M, Cytokines detected in the tumor lysate 14 days after implantation in the control group (black, n = 8) compared with 24 hours after irradiation or 7 days after irradiation in the conv-PT (red, 24 hpi n = 5 and 7 dpi n = 6) and pMBRT-treated groups (blue, 24 hpi n = 5 and 7 dpi n = 6). The data are presented as the mean ± SD. Statistical analysis performed with one-way ANOVA and t test multiple comparisons. dpi, days post implantation; hpi, hours post implantation.

More efficient attraction of lymphocytes to the core of the tumor induced by pMBRT

To obtain a spatial characterization of this infiltration, the tumors were imaged using IF, and the regions were segmented into the center of the tumor, adjacent tissue, and periphery of the tumor. A distance of 320 µm on both sides of the tumor invasion front was considered for the periphery of the tumor and the adjacent tissue. Both CD4⁺ and CD8⁺ T cells were solely significantly increased in the core of the tumor by pMBRT and not the homogeneous irradiation (Fig. 3A and B), whereas both conv-PT and pMBRT increased the T cells in the adjacent and the peripheral regions of the tumor. An increase in Tregs was also observed in the core of the tumor after pMBRT but also after conv-PT in the adjacent and peripheral tissues. Interestingly, pMBRT was also able to increase the concentration of B cells in the border and especially in the periphery of the tumor (Fig. 3C; Supplementary Fig. S4). The increase in macrophages after radiation treatments was also observed in the tissue specifically in the center of the tumor, whereas the increase in CD8⁺ macrophages was observed in all the regions analyzed both after conv-PT and pMBRT (Fig. 3C; Supplementary Fig. S4). These results suggest that pMBRT can enhance the infiltration of antitumoral T and B lymphocytes into the core of GBM.

Figure 3.

Spatial analysis of immune cell density with multiplex immunohistofluorescence. A, CD3, CD4, CD8, and FoxP3 staining segmented into the center of the tumor, adjacent area (border), and periphery to analyze CD4⁺ T cells, CD8⁺ T cells, and Tregs in control (n = 5), conv-PT (n = 7), and pMBRT (n = 7) expressed in cells/mm² of tissue 7 days after implantation. B, Representative images from the IF multiplex with CD3 (magenta), CD4 (yellow), CD8 (green), and FoxP3 (cyan) staining in the tumor center, at the tumor periphery, and in adjacent brain tissue with regard to the irradiation group. Scale bars, 100 µm, (C) Pax5, CD68, and CD8 staining in control (n = 5), conv-PT (n = 7), and pMBRT (n = 7) 7 days after implantation segmented into the center of the tumor, adjacent area (border), and periphery to analyze B cells (Pax5⁺ cells), macrophages (CD68⁺), and CD8⁺ macrophages (CD68⁺ CD8⁺ cells) expressed in cells/mm² of tissue. The data are presented as the mean ± SD.

Single-cell transcriptomics analysis of infiltrating immune cells in rat GBM

To obtain a comprehensive view of the changes induced by pMBRT to the immune cells in the tumor, a single-cell transcriptomics analysis (single-cell RNA-seq) was performed on CD45-expressing cells in the rat RG2 GBM model (Fig. 4A). The dataset contained a total of 14,025 tumor immune cells pertaining to four nonirradiated controls, 13,135 cells pertaining to three conv-PT–treated tumors, and 14,807 cells pertaining to four pMBRT-treated tumors after quality control filtering (Fig. 4B; Supplementary Fig. S5). Cell partitioning into clusters showed well-separated 18 different cell types/states identified by their gene expression (Fig. 4A and C). The most significantly different marker expressed by each cluster showed clear biological separations between the clusters (Fig. 4C) and helped annotate the clusters together with key markers in literature (Fig. 4C; Supplementary Fig. S6). The following 18 clusters were identified: one cluster of neutrophils (C1); four clusters of dendritic cells (DC; C2–C5), among those an activated state (C3) showing upregulation of antigen presentation genes (CD40, CD84, and CD86) and lymph node trafficking chemokine Ccr7; three clusters of monocytes and intermediate monocytes/macrophages (C6–C8); four clusters of TAMs (C9–C12); five clusters of T-cell and NK-cell subtypes (C13–17); and one cluster of B cells (C18). Additionally, two main clusters of proliferation were observed, containing macrophages (C12) and T cells (C13).

Figure 4.

Identification of tumor-infiltrating immune cells in rat glioblastoma by single-cell RNA-seq. A, Uniform Manifold Approximation and Projection for Dimension Reduction (UMAP) representation showing clusterization of the main 18 clusters separated by gene expression. B, Localization of the three experimental conditions analyzed in the dataset. C, Feature plot of each most DEG per cluster (lowest P value). D, Markers used to identify the different populations in the clusters. Arrows show alternative gene nomenclature or the translated protein. Blue notations indicate the family of markers or function of the gene. CTRL, control; Mono/mac, monocytes/macrophages.

Transcriptional changes induced by pMBRT in different immune cell populations

The proportion of cell clusters per condition was analyzed after downsampling and iterating over the lowest cell–containing sample (2,200 cells; Fig. 5A; Supplementary Fig. S7), which revealed an increase in the activation of DCs (C3), only significant in the case of pMBRT, as well as a marked reduction of microglia cells (C11) in both irradiated conditions. Tregs (C15) were also increased in proportion as previously observed in the cytometry and IF studies. Statistical analysis of DEGs between conditions per cluster revealed that irradiation and spatial fractionation influences vary among the different immune populations (Fig. 5B; Supplementary Fig. S8). Although both PT irradiations generated strong gene dysregulation in most clusters compared with the nonirradiated control populations (with the exception of neutrophils), the comparison between conv-PT and pMBRT showed no variation in neutrophils (C1), plasmacytoid dendritic cells (pDC) (C2), and CD4 T cells (C14). However, other cell types showed increased gene dysregulation related to spatial fractionation of the irradiation (Fig. 5B); in particular, activated DCs (C3) and B cells (C18) showed higher upregulated genes in pMBRT than in the other two comparisons, whereas monocytes (C6) and monocytes/macrophages (C7) showed the most differences between the irradiation conditions. This table, in summary, provided qualitative insights into which clusters exhibit more transcriptional changes, guiding the focus of further analyses such as functional enrichment, and it is to be interpreted alongside other metrics, such as fold change and adjusted P value of individual genes or gene expression proportion.

Figure 5.

Statistical comparison of the conditions in single-cell transcriptomics. A, Proportion of clusters by sample compared by condition. Control condition (yellow, n = 4), conv-PT (green, n = 3), and pMBRT (blue, n = 4); statistical analysis using the t test pairwise comparison; only P < 0.05 values are shown. B, Quantification of DEGs per comparison by each cluster in the dataset. Red, upregulated genes; blue, downregulated genes. C, DEGs of pMBRT vs. conv-PT comparison in C3 — aDC1/2, (D) C18 — B cells, (E) C6 — monocytes, and (F) C7 — mono/mac 1. Only genes upregulated or downregulated with more than 0.411 log₂ fold change (FC; 30% change) and adjusted P value of less than 0.05 are considered significant DEGs. Genes with a log₂FC > 1 or log₂FC < −1 (>200% change) are shown in a darker color and annotated in the graph. G, Overrepresentation analysis or pathway analysis using the Gene Ontology (GO) biological process (BP) dataset of C3 — aDC1/2, (H) C18 — B cells, (I) C6 — monocytes, and (J) C7 — mono/mac 1. Pathways are curated to remove duplicates and ranked in descending order by ratio of genes per condition and q value. CTRL, control; Mono/mac, monocytes/macrophages.

Indeed, activated DCs significantly upregulated genes related to antigen presentation and immune activation in pMBRT samples compared with conv-PT, like Cxcl10, RT1-Bb, and Il1b (Fig. 5C), and overall signaling pathways related to T-cell costimulatory signals like Cd80 and Cd86 (Fig. 5C and G). On the other hand, B cells in pMBRT samples upregulated genes related to terminal differentiation into antibody-secreting cells compared with those in conv-PT, such as Jchain, Mzb1, Xbp1, Ssr4, immunoglobulin light chain variable region AABR07034739.1, and transcription of the establishment of endoplasmic reticulum (ER) protein processing like Sec61b, which is associated with antibody-secreting plasma cells (Fig. 5D), predominantly increasing cell signaling pathways related to immunoglobulin production and peptide processing in ER like Creld2 (Fig. 5D and H).

With regard to the clusters containing monocytes/macrophages (C6 and C7), pMBRT-treated samples upregulated members of the HSP family (Hsp70) like Hspa8, Hsph1, and Hspa1b (Fig. 5E and F) and members of the IFN signaling pathway [IFN regulatory factor (Irf) 1, Irf7, and three-prime repair exonuclease 1 (Trex1)] and genes induced by oxidative stress (Hif1a and Hmox1). pMBRT in monocytes (C6) also upregulated pro-inflammatory Clec10a and other genes involved in myeloid cell differentiation like Selenop and Cd163 (Fig. 5E and I), whereas in the Mono/Mac 1 cluster (C7) conv-PT or pMBRT also generated a dysregulation in cytokines, upregulating neutrophil/monocyte chemotactic cytokine Cxcl3 in the case of conv-PT and T-cell and macrophage chemotactic cytokine Cxcl11 in the case of pMBRT (Fig. 5F and J). Overall, pMBRT upregulated genes involved in specific cell signaling pathways across tumor immune cells, including T- and B-cell activation, IFN response, and response to reactive oxygen species (Fig. 5C–J and Supplementary Fig. S7). The comparisons for all the clusters are available in Supplementary Fig. S5, and the Gene Ontology signaling pathways are available in the Supplementary Material (Supplementary Fig. S9).

Analysis of pMBRT-induced T cells

Reanalysis of the original subsets pertaining to T cells (C13–17 in Fig. 5) exposed a contaminant cluster expressing macrophages and other antigen-presenting cell markers, which was removed for the subsequent analysis (Supplementary Fig. S10). A reclustering of the remaining cells unveiled 10 different types of lymphocytes, including a cluster of NK cells that was undetectable in the previous analysis (Fig. 6A). The clusters were annotated based on the expression of canonical T-cell markers (Cd3e, Cd4, and Cd8a), including naïve T-cell markers like Ccr7, Tcf7, Lef1, and Sell and cytotoxicity markers like Prf1, Gzma, Gzmb, and Cx3cr1 and two clusters of terminally differentiated effector/memory cells (TEMRA), as well as one cluster expressing high levels of IFN response genes such as Ifit2, Ifit3, and Irf7 (Fig. 6B; Supplementary Fig. S11). Cluster C10 was annotated as NK cells as it did not express Cd3e and exclusively expressed the NK cell marker Klrg1. Although most T cells of the nonirradiated control condition were located in the naïve, regulatory, and proliferative clusters, T cells of the irradiated conditions were more abundant in the CD8⁺ TEMRA T-cell (C6 and C7) and conventional T-cell (Tconv; C9) clusters (Fig. 6C).

Figure 6.

Reclustering of T-cell clusters. A, Uniform Manifold Approximation and Projection for Dimension Reduction (UMAP) representation showing clusterization of the main T-cell types/states by gene expression. B, Markers used to identify the different populations in the clusters. C, Localization of the three experimental conditions analyzed in the UMAP. D, Chemokine and chemokine receptor expression in the myeloid cells and T cells. Significance is based on adjusted P value < 0.05. E, Quantification of DEGs per comparison by each cluster in the dataset. Red, upregulated genes; blue, downregulated genes. Significance is based on adjusted P value < 0.05. CTRL, control.

CellChat analysis showed an increased number of inferred cell–cell interactions among the tumor immune cells of the irradiated tumors, especially in pMBRT (Supplementary Fig. S12A). The chemokine and chemokine receptor expression in myeloid cells (DCs, monocytes, and macrophages) together with T cells (clusters C1–C9) yielded information of the chemotactic cues present in the tissue after the treatments. Among those, PT significantly reduced the expression of myeloid-derived suppressor cell–attractant Ccl2 and the macrophage-attractant Ccl3, Cxcl1, and Cxcl2, while increasing T-cell attractant and activator Ccl5. Interestingly, only pMBRT significantly increased the expression of IFN-induced Cxcl10, both in myeloid cells and T cells, and Cxcl11 in myeloid cells, related to T-cell chemoattraction in tumors, whereas both PT groups increased the expression of their associated receptor, Cxcr3 (Fig. 6D). Indeed, although both conv-PT and pMBRT similarly increased the level of CXCL chemotactic cell pathway among the cell clusters (Fig. S12B), the specific analysis of CXCL10–CXCR3 and CXCL11–CXCR3 ligand–receptor pair analysis yielded particular activation after pMBRT irradiation (Supplementary Fig. S12C). Cxcl10 was only expressed by C10 (macrophages 2) and C12 (proliferating) macrophages in nonirradiated controls, affecting only certain T-cell clusters, whereas the pair CXCL11–CXCR3 was not even detected in the control group. Conv-PT changed the clusters involved in the CXCL10/CXCL11–CXCR3 interaction. In the case of pMBRT, virtually all myeloid clusters but neutrophils synchronously increased the interaction to all lymphoid clusters via CXCL10–CXCR3 signaling. Similarly, more types of myeloid cells increased the CXCL11–CXCR3 interaction, the main senders remaining monocytes and DCs (Supplementary Fig. S12C).

Statistical analysis of the DEGs in pairwise comparisons showed transcriptomic differences among most of the clusters after both radiations (Fig. 6E), but only clusters C1–C5, C7, and C8 showed compelling transcriptomic differences between pMBRT and conv-PT (Fig. 7), whereas CD8 TEMRA T cells, Tconv, and NK cells showed no significant differences with regard to spatial fractionation. Among the genes dysregulated in the other clusters, HSP chaperones like Hspa8 were differentially upregulated in pMBRT-induced T cells. Type I IFN signaling pathway genes were upregulated in naïve CD8 T cells (Fig. 7C and J), CD4/CD8 TEMRA (Fig. 7F and M), and CD4 Treg (Fig. 7G and N), exemplified by Irf7, IFN-stimulated gene 15, and IFN-induced transmembrane protein 1. Ifng was also more expressed in pMBRT-induced T cells that were proliferating (Fig. 7A and B) and activator protein Jun in all T cells, whereas exhaustion markers Tigit and Tox were more upregulated in all conv-PT–induced T cells (Fig. 7A–G). Although immune checkpoint PD-1 (Pdcd1) was not overly upregulated after proton irradiations, CTLA-4 (Ctla4) was upregulated by both conditions and specifically by conv-PT in proliferating T cells (C1 and C2), CD4/CD8 naïve/effector (C4), IFN response (C5), CD4/CD8 TEMRA (C7), and CD4 Treg (C8; Fig. 7A–G). Thus, both pMBRT and conv-PT generated more CD8 TEMRA and Tconv cells than the control group, but all the other T cells presented differences with regard to toxicity, activation, and exhaustion (Fig. 7H–N), pointing to better preserved T cells after pMBRT.

Figure 7.

Gene analysis between pMBRT and conv-PT in the different T-cell clusters. A, Volcano plot showing DEGs in pMBRT vs. conv-PT comparison (left); selected gene expression levels in the three conditions in clusters C1, (B) C2, (C) C3, (D) C4, (E) C5, (F) C7, and (G) C8 (right). Only genes with more than 0.411 log₂ fold change (log₂FC; 30% change) and adjusted P value of less than 0.05 are considered. Genes with a log₂FC > 1 or log₂FC < −1 (>200% change) are shown in a darker color and annotated in the graph. H, Overrepresentation analysis or pathway analysis using the Gene Ontology (GO) biological process (BP) dataset of clusters C1, (I) C2, (J) C3, (K) C4, (L) C5, (M) C7, and (N) C8. Pathways are curated to remove duplicates and ranked in descending order by ratio of genes per condition and q value. CTRL, control; Ifitm1, IFN-induced transmembrane protein 1; Isg15, IFN-stimulated gene 15.

Discussion

Despite significant advances in cancer therapy in the past few decades, the treatment of bulky radioresistant tumors continues being a challenge, with cases such as brain GBM lacking any efficient treatment. Thus, there is still an urgent medical need to find innovative therapies that provide alternative antitumor mechanisms with limited toxicity in the normal tissue. The lower toxicity together with a similar or even superior tumoral control that has been shown in preclinical studies places pMBRT as a promising technique for difficult-to-treat cancers (37–39). Two first patients have been treated with low-energy X-ray MBRT in superficial tumors (45), and pMBRT will allow for better treatment of deep-seated tumors because of the physical properties of protons. At present, it is fundamental to increase our understanding of the antitumor mechanisms of this technique to achieve its full potential. Thus, this article embarked upon the relevant question of the mechanism of pMBRT for antitumoral effect, with special attention to the immunomodulatory capacities of pMBRT.

The results of this study in terms of the requirement of T cells for minibeam efficacy are in agreement with a previous MBRT study (23), further stressing the importance of T cell–induced cytotoxicity in MBRT. In general, lymphoid cell populations that are related to generation of adaptive immunity and immunologic memory were significantly increased in the tumor after pMBRT treatment albeit not in conv-PT, such as tissue-resident memory CD8⁺ cells, B cells, and DCs. Additionally, pMBRT increased the reach of CD4⁺ and CD8⁺ T cells to the center of the tumor, both of which are linked to the lower recurrence rate (49, 50). This increase in immune cells in the tissue could be related to a better sparing of the vasculature architecture by spatial fractionation, mechanical disintegration of tumoral extracellular matrix, or an increased proportion of immunogenic cell death events, which provokes the release of damage-associated molecular patterns, triggering IFN-I–induced inflammation and leukocyte chemotaxis (51).

Single-cell RNA-seq is a powerful method of analysis that allows to unravel the gene expression profiles inside a heterogeneous population. The analysis of CD45-expressing cells offered a vast view of the immune cells in this tumor model and the transcriptomic and functional differences generated by PT and pMBRT. An increase in the proportion of activated DCs was suggestive of maturation of DCs following both irradiations, although only pMBRT increase was significant. Additionally, pMBRT-induced mature DCs had higher expression of genes related to T-cell activation and response to IFN. Indeed, the activated dendritic cells (aDC) induced by conv-PT showed lower levels of co-stimulatory molecules CD80 and CD86, suggesting a maturation toward lower-immunogenic DCs. Although still not well understood, it has been suggested that the type of cell death such as apoptosis, necrosis, or pyroptosis and the subsequently produced IFN-I signaling cascade could be behind the skewed maturation of DCs into immunogenic or tolerogenic, ultimately determining T-cell clonal expansion and antitumoral immune response (51). Interestingly, the monocytes induced by pMBRT in this model increased the expression of genes of the IFN-I signaling pathway. STING-independent IFN-I activation by monocytes drives focal RT efficacy in high doses and promotes effector CD8⁺ T-cell function (52). Homogeneous doses of 20 Gy and higher can induce TREX1 in the tumor cells, which degrades cytosolic DNA, dampening IFN-I secretion and RT-induced immunogenicity (53). In this study, Trex1 was upregulated in the monocytes of both irradiated conditions. Although this pathway inhibitor was upregulated in pMBRT, other pathway activators like Soc3, Irf1, and Irf7 were significantly more upregulated in pMBRT than the other two conditions.

An increase in B-cell immunity was observed by a quantitative increase in the tissue that was mostly concentrated in the tumor margins. Interestingly, pMBRT created the most differences in the transcriptome compared with conv-PT and control groups in this subset of leukocytes, with regard to differentiation into antibody-producing plasma cells. Plasma cells contain a large ER, necessary to produce large numbers of antibodies. Of note, a recent publication demonstrated that plasma cells are remarkably resistant to radiation (54), which could also be an argument to consider using pMBRT when re-irradiation is programmed.

The single-cell transcriptomics analysis also detected a maturation of T cells into CD8⁺ TEMRA by PT that could not be observed in the flow cytometric or histologic studies, as well as identified a cluster of IFN-response T cells in the dataset and an increase in immune checkpoint CTLA-4 by PT. IFN-I signaling pathway genes were also upregulated in pMBRT naïve CD8⁺ T cells, CD4/CD8 TEMRA, and CD4 Tregs, and an increase in IFNγ was also noticed in proliferating T cells and the IFN-response cluster. Conversely, conv-PT–produced T cells showed an increased exhaustion profile. Altogether, pMBRT-induced T cells seemed more activated than in homogeneous irradiation, which could account for the importance of these T cells for pMBRT antitumor effects.

The immune response generated by pMBRT could be related to both a better sparing of the T cells in the valleys traversing the tissue, leading to T-cell clonal expansion and activation after infiltration to the tumor core, and different nature of cell death events provoked by the heterogeneous dose distribution in the tissue, which could be more immunogenic in the case of pMBRT, more proficiently inducing T-cell activity and chemotaxis. The increase in the IFN-I signaling cascade in the monocytes, main immune producers of IFN-I (52, 55) together with the tumoral and stromal cells, and DC immunogenic activation observed in this study seem to point toward the latter, with a specific upregulation of master IFN-I regulator IRF7 (56) in pMBRT. The exclusive upregulation of chemokine CXCL10 in pMBRT could be determinant in this technique. CXCL10 is a potent T-cell attractant and activator induced by IFN-I and IFNγ, and its expression correlated with antitumor immunity and good prognosis (49, 57, 58), especially by TAMs (57, 58), which could explain the improved T-cell infiltration in pMBRT. However, future studies will be needed to test these hypotheses.

Altogether, the results presented here show that pMBRT boosts the immune response against GBM, enhancing lymphocyte density and transcriptome. This article is also meant to serve as a resource for the community in particle therapy and those using partial and spatially fractionated irradiations by increasing our knowledge on the relationship between radiation and immune response, showing promise for future radioimmunotherapies by minimizing healthy tissue toxicity. Furthermore, this study also provides a dataset of rat tumor immunity that could serve for future immunologic studies involving other preclinical animal models beyond the mouse. Single-cell analysis provided information of the main tumor-associated immune subsets, which are developmentally conserved between rodents (rat and mouse) and human, including main markers of DC activation, macrophage canonical markers, and T-cell subsets, as well as new rat markers like AABR07034739.1.

This study has also some limitations. The difficulty in finding the fairest comparison with a spatial fractionation technique is a contentious topic, due to the multiple parameters that can be modulated. In this study, we decided to deliver a single fraction of radiation in the form of a homogeneous (conv-PT) or heterogeneous (pMBRT) radiation field, as means to compare the effect of spatial fractionation in the same total dose given to the tumor. In this case, the same average dose (30 Gy) was employed in both conv-PT and pMBRT as this condition leads to statistically equivalent survival in both irradiation modes (37).

We chose RG2 as a glioma model as these cells form a highly invasive glioma with low immunogenicity compared with other rat glioma models like C6 and 9L (59, 60). Therefore, the model allowed us to generate a proof-of-concept study to observe relative differences in the immune response generated by a pMBRT irradiation compared with a conventional irradiation. The literature reports contradictory results with regard to the immunogenicity of luciferase expressed by the tumor cells, which was shown in a model of breast cancer (61). Although the contrary was observed in others, it is the case in a model of ovarian cancer (62). Although we have not directly evaluated the role of luciferase in this study, comparing published data at the same concentration of RG2-Luc cells (23) and WT RG2 (63) showed a median survival time of 20 ± 2 days. Hence, these data suggest that luciferase in this specific model plays no role in slowing down the tumor process.

In future studies, the influence of different parameters in pMBRT on tumor-killing capacity and immune modulation should also be explored, as they remain out of the scope in the present article. On the other hand, this study was performed in a murine model, the rat; therefore, the results should be taken with caution and be confirmed in human subjects as the clinical trials start. Indeed, the effect of pMBRT in a bigger volume and different radiosensitivity such as in humans should be validated, yet the results in a recent veterinary clinical trial (44) and the two first patients treated with MBRT yielded optimistic results (45).

Several possibilities arise on the potential applications of pMBRT into current treatments in the clinics. pMBRT could be proposed as a first line of treatment in patients with immunologically cold tumor microenvironment like GBMs and in which the radiosensitivity of the tissue does not allow to use higher doses of conventional radiation. Other clinical perspectives of MBRT include the treatment of tumors local control of which could be improved by means of a dose escalation, which is currently incompatible with the normal tissue toxicities resulting from standard treatments. Finally, the increase in immune memory potential by pMBRT could be valuable as a tumor vaccine, diminishing the high recurrence rate associated with certain cancers. It could also be envisioned as an adjuvant technique to immunotherapy, combined or not with conventional RT or PT. It is currently fundamental to understand the best RT configuration that works in concert with immunotherapy and that avoids toxicities of acute inflammatory conditions exacerbated by ionizing radiation.

Supplementary Material

Table S1

Table S1. List of antibodies used for flow cytometry in Fig. 2.

Figure S1

Figure S1. Survival curves of RG2- glioma bearing rats: non-irradiated controls (black), rats receiving 30 Gy in Conv-PT (blue) and rats receiving 30 Gy average dose in pMBRT (red). The survival of the irradiated groups is significantly longer than the controls (p < 0.0001, log-rank (Mantel-Cox) test). There is no significant differences in survival between the two irradiated groups.

Figure S2

Figure S2. A Gating strategy for tumor immune cell analysis by flow cytometry in Fig. 2.

Figure S3

Figure S3. A, T2 image of a F98-bearing rat. B, Flow cytometry results in a F98-bearing rat.

Figure S4

Figure S4. Representative images from the IF mutliplex

Figure S5

Figure S5. Cells in the dataset after quality control and doublet cell removal.

Figure S6

Figure S6. Heatmap showing the top 5 differentially expressed genes in the clusters

Figure S7

Figure S7. Results of the 1,000 iterations to downsize to 2,200 cells per cluster and per sample.

Figure S8

Figure S8. DEGs of pMBRT vs Conv-PT comparison in all the clusters. , only genes up- or down-regulated with more than 0.411 log2 fold change (FC) (30% change) and adjusted p-value of less than 0.05 are considered as significantly DEGs. Genes with a log2FC > 1 or log2FC < -1 (>200% change) are shown in a darker color and annotated in the graph.

Figure S9

Figure S9. Over-Representation Analysis (ORA), or pathway analysis, using Gene Ontology (GO) - Biological Process (BP) dataset of all the clusters that presented pathways. Pathways are curated to remove duplicates and ranked in descending order by pathway group and qvalue.

Figure S10

Figure S10. Removal of contaminant in T cell subclustering. A) Reclustering of clusters C13-C17 of Fig. 5. B) Heatmap of top 10 most significantly differentially expressed genes per cluster. C) Expression of macrophage markers in the clusters confirms contamination, cluster 3 will be removed for following analysis

Figure S11

Figure S11. Heatmap showing the top 10 differentially expressed genes per cluster in the NK/T cell object

Figure S12

Figure S12. Analysis of cell–cell communication using CellChat.

Data Availability

All data are available upon request to the corresponding author. The data discussed in this publication have been deposited in NCBI’s Gene Expression Omnibus through GEO Series accession number GSE279905 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE279905).

Authors’ Disclosures

P.-E. Bonte reports personal fees from Mnemo Therapeutics outside the submitted work. No disclosures were reported by the other authors.

Authors’ Contributions

L. Iturri: Conceptualization, data curation, formal analysis, investigation, methodology, writing–original draft, writing–review and editing. M. Riquelme-Perez: Formal analysis, methodology, writing–review and editing. P.-E. Bonté: Formal analysis. S. Potiron: Investigation. C. Goudot: Formal analysis. M. Juchaux: Formal analysis, supervision, investigation, writing–review and editing. E. Brisebard: Formal analysis, investigation. C. Gilbert: Investigation. J. Espenon: Investigation. R. Ortiz: Dosimetry, investigation. A. Patriarca: Investigation. L. De Marzi: Irradiation. S. Amigorena: Validation, investigation. Y. Prezado: Conceptualization, resources, supervision, funding acquisition, investigation, project administration, writing–review and editing.