henhaobio
Sign in

bioRxivCancer BiologyPreprint

Differential and regionally distinct molecular responses distinguish FLASH and conventional RT in a medulloblastoma model while preserving therapeutic outcomes

Michèle KnoliD, Javier Franco‐PéreziD, Aymeric AlmeidaiD, Louis V. KunziD, Benoît PetitiD, Antoine Job, Jonathan Ollivier, Jackeline F. Romero, Jeannette JanseniD, Veljko GriljiD, Charles L. LimoliiD, Marie‐Catherine VozeniniD, Paola Ballesteros‐ZebadúaiD

PDFDOICodeData
In-depth full-text digest available · Jump to digest ↓

Abstract

Background. FLASH radiotherapy might minimize treatment-related complications in medulloblastoma patients. To support its clinical translation, we investigated the therapeutic index of hypofractionated FLASH- and CONV-RT and their respective molecular determinants using an orthotopic human medulloblastoma mouse model. Methods. Five cohorts of juvenile UW228-MB-bearing nude mice were irradiated or not using 3x10 Gy (BED10 = 60 Gy), delivered 48 h apart at 0.1 Gy/s (CONV) or 5.5x10 Gy/s (FLASH) using an electron beam (eRT6). Digital spatial profiling (DSP) and immunofluorescence were performed 24h after radiotherapy in one cohort, while four cohorts were followed for long-term tumor response, cognition and neuroinflammation. Results. Both CONV- and FLASH-RT induced a long-lasting anti-tumor response in all irradiated animals. Long-term NOR performance was preserved in irradiated animals, with no significant difference compared with age-matched, non-tumor-bearing controls. A higher but non-significant proportion of very good discriminators was observed in the FLASH cohort while neuroinflammatory markers were statistically reduced. In the FLASH-irradiated mice, DSP revealed a sustained microglial activation in the cerebellar tumor micro-environment, with enhanced expression of genes with phagocytic and proteolytic activity. In the hippocampus, FLASH-induced genes involved a preferential astrocyte/neuron transcriptional crosstalk including metabolic support and synapse-glia interaction. The relevance of this metabolic support provided by astrocytes to neurons in the FLASH cohort was further supported by the enhanced expression of the astrocytic monocarboxylate transporter MCT4, a key transporter of lactate. Conclusion: The study shows that despite similar therapeutic index, hypofractionated FLASH and CONV-RT induce a distinct molecular signature in a human medulloblastoma mouse model.

In-depth summary

Claude Opus 5.5 · xhigh effort · 8 Nov 2026 · from the full text of version 2 (1 Nov 2026) · CC BY 4.0

Hypo-fractionated FLASH and conventional radiotherapy (3x10 Gy) each ablated orthotopic human medulloblastoma in nude mice; FLASH evoked distinct microglial and hippocampal transcriptional programs and less neuroinflammation.

Radiotherapy cures most children with medulloblastoma, but standard fractionation leaves lasting neurocognitive deficits and requires repeated sedation, so regimens that spare cognition with fewer sessions are needed. This work tests a clinically relevant hypo-fractionated FLASH schedule in an orthotopic human tumor model and maps transcriptional responses by cell type and region that may explain the sparing effect of FLASH.

Key results

  1. ShownHypo-fractionated FLASH and conventional (CONV) irradiation (BED 60 Gy) were equally ablative: every irradiated mouse had a complete, lasting tumor response and survived long term, while all untreated tumor-bearing mice progressed.

    Complete long-lasting response in 32/32 irradiated mice vs 16/16 non-irradiated mice (2 cohorts) with micro-CT-visible MB at 8 weeks post-engraftment and neurological signs requiring euthanasia within 9-12 weeks; tumors were detectable by micro-CT in 65% of mice at the time of RT. · Both FLASH- and CONV-RT induce a complete and…; Hypo-fractionated radiotherapy; Tumor response

  2. SuggestedBoth modalities lowered novel object discrimination at 6 months post-RT. FLASH-treated mice were more often very good discriminators, and less often bad discriminators, than CONV-treated mice.

    Very good discriminators at 4 months: 18% after FLASH vs 7% after CONV (the abstract gives 38% vs 7%); bad discriminators were significantly more frequent after CONV at all time points (two-proportion Z-test); 10 age-matched tumor-free controls. · FLASH-RT preserves cognition better than CONV-RT; Novel object recognition; Abstract

  3. SuggestedIn the tumor-bearing cerebellum, microglia dominated the acute response, and the two modalities induced largely non-overlapping programs: FLASH favored phagocytic, proteolytic, mitochondrial and remodeling genes; CONV favored iron, lipid biosynthesis, proliferation and CXCL10/IL5 signaling.

    FLASH-only genes: 87 up, 59 down; CONV-only: 24 up, 22 down; shared: 2 up (Fam187a, Klf5) and 3 down; FLASH-specific genes include Megf10, Mmp16, Adam19 and Spon2; 3 mice per group, an average of 30-60 microglia per segment, |log2FC|>1 and p<0.05. · FLASH and CONV induce a distinct microglial…; Bioinformatics; GeoMx Digital Spatial Profiling

  4. SuggestedFLASH altered neurons in DG, CA3 and CA1, and astrocytes in CA3 and CA1, far more than CONV, which acted mainly on DG astrocytes. The programs were region-specific: DNA repair genes in DG neurons, Arc/Camk2g plasticity genes in CA3 neurons, respiration and synaptic genes in CA3 astrocytes.

    Genes changed at 24 h (up/down). Neurons, FLASH: DG 28/144; CA3 97/218 (CONV 16/23); CA1 64/182 (CONV 3/9). Astrocytes, FLASH: CA3 330/383 (CONV 68/33); CA1 213/201 (CONV 19/6); DG 140/106 vs CONV 140/300. Clec2j was upregulated in several regions under both modalities. · FLASH modulates the neuronal transcriptome in…; FLASH and CONV induce a distinct transcriptomic…; FLASH-induced transcriptomic response is region-…

  5. ShownSix months after RT, pro-inflammatory transcripts and microglial activation markers rose only after CONV, while FLASH-treated brains resembled tumor-free controls. Neither modality produced astrogliosis detectable by GFAP staining.

    Il-1β, TNF-α, IL4 and Nlrp3 mRNA rose only after CONV vs control; IL-6 and Gasdermin-d were unchanged; Iba1 and CD68 staining was high after CONV in cerebellum and hippocampus but low and control-like after FLASH; 6 mice per group; the text reports no test statistics. · FLASH reduces neuroinflammation and glial…; Evaluation of neuroinflammation at endpoint

Methods, data and reproducibility

Orthotopic UW228 human medulloblastoma xenografts in juvenile nude mice received whole-brain 6 MeV electron RT, 3x10 Gy given 48 h apart at 0.1 Gy/s (CONV) or 5.5x10^6 Gy/s (FLASH). Readouts: micro-CT tumor volume, novel object recognition, GeoMx digital spatial profiling at 24 h with mixed-model differential expression and GSEA by region and cell type, and inflammation assays at 6 months.

  • Orthotopic UW228 human medulloblastoma xenografts in female Swiss nude mice (NU(Ico)-Foxn1^nu), produced in five sequential cohorts · mouse (host) carrying human tumor cells · n = 57 mice
  • Tumor response and survival cohorts followed by weekly contrast-enhanced micro-CT · mouse · n = 48 mice (32 irradiated, 16 non-irradiated)
  • GeoMx spatial profiling cohort sampled 24 h post-RT; ROIs in the cerebellum and in hippocampal DG, CA1 and CA3 (microglia, neurons, astrocytes) · mouse · n = 3 mice per group
  • Age-matched tumor-free controls for novel object recognition testing · mouse · n = 10
  • Endpoint neuroinflammation cohort at 6 months post-RT (cytokine qRT-PCR; Iba1, CD68 and GFAP staining) · mouse · n = 6 mice per group
  • UW228 human medulloblastoma cell line, derived from a 9-year-old female patient · human
Methods & tools
Oriatron eRT6 6 MeV electron linear accelerator (PMB, Alcen), MILabs micro-CT with contrast enhancement, GeoMx Digital Spatial Profiling, Novel object recognition (NOR) test, R v4.4.0, Linear mixed model (slide as random intercept), Gene set enrichment analysis (GSEA), GraphPad Prism 9.1 (Shapiro-Wilk, one-way ANOVA, Tukey post hoc), Two-proportion Z-test (socscistatistics web tool), qRT-PCR pro-inflammatory cytokine panel, Iba1, CD68 and GFAP immunostaining
Software & environment
R v4.4.0 for spatial transcriptomic analysis; GraphPad Prism version 9.1 for the behavioral statistics.
Reproducibility gaps
The GEO accession for the raw spatial profiling data is stated as pending. Quality control, pre-processing and neuroinflammation methods are deferred to supplementary material that is not in the provided text.
Limitations (6)
  • The intermediate-energy electron FLASH source has limited penetration and cannot reach deep-seated lesions such as most primary brain tumors, which restricts immediate clinical use. · stated by the authors
  • The dose rate used (>10^6 Gy/s) exceeds what most clinical proton or photon FLASH platforms achieve (~100 Gy/s). Whether sparing holds at lower dose rates or with other radiation qualities has not been validated. · stated by the authors
  • The whole brain was irradiated because conformal delivery was not available. Conformal, image-guided FLASH systems still need to be developed and tested before clinical transfer. · stated by the authors
  • Hosts were nude mice lacking adaptive immunity. Cognition after FLASH stayed below levels previously reported in tumor-free immunocompetent juvenile models, possibly because of this immune status. · stated by the authors
  • The transcriptomic conclusions rest on 3 mice per group at a single 24 h timepoint, with a nominal p<0.05 and |log2FC|>1 and no stated multiple-testing correction. Functional programs are inferred from gene lists without protein-level or functional validation. · noted in this summary
  • Some figures conflict between sections. The abstract reports 38% vs 7% very good discriminators, but Results give 18% vs 7% at 4 months; mouse ages at enrollment also differ between the abstract and Methods. · noted in this summary

Read this if you work on FLASH radiotherapy, pediatric brain tumor radiobiology or radiation-induced cognitive injury and want spatially resolved microglia, neuron and astrocyte responses to FLASH vs conventional dose rates.

Citations
0

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Related papers

Similarity Map
SelectedSame fieldCross-field
THISPadilla '22Jansen '25Drayson '24Melemenidis '25Furini '25Kay '25
73% semantic matchImmunology
O. Padilla, H. E. Minns, H.-J. Wei+22 more·bioRxiv·Nov 9, 2022