Abstract Introduction Replication repair deficiency (RRD) is a pan-cancer mechanism characterized by hypermutagenesis and variable response to immunotherapy. Total tumor mutational burden (TMB) is a potential predictor for cancer response to immunotherapy, however, whether organ-specific immune regulation of mutation accumulation during tumorigenesis is present, remains unclear. Methods To address this question and gain understanding on differential immune surveillance in intracranial and extracranial tumors, we created a novel RRD (MMRD/ POLE mutant) glioma mouse model. We then performed a trans-species analysis comparing RRD mutagenesis of murine and human cancers from intracranial and extracranial origins to study tumor kinetics, clonal evolution, and the effects of immunotherapy on TMB and immunogenic neoantigens during tumorigenesis. Results In humans, CNS tumors exhibited lower immune cell infiltration (p = 0.001) and survival (p = 0.03) than extracranial tumors, similar to our observations in immunocompetent mice, where intracranial tumors equally displayed low immune infiltration and worse survival outcomes (p = 0.002) compared to extracranial tumors. Interestingly, despite obligatory mutagenesis, TMB was reduced during tumor relapse in human cancers and upon re-implantation in immunocompetent mice (p < 0.05). In contrast, TMB dramatically increased in human RRD cell glioma cell lines grown in vitro and in tumors reimplanted in immunocompromised mice, particularly in the CNS (p < 0.05). In immunocompetent mice, the tumor-growth driving ability of RRD glioma cell lines with different TMB exhibited an association between higher TMB, lower penetrance, higher immune infiltration (p = 0.0005), and improved survival outcomes over lower TMB (p = 0.01). Finally, neoantigen analysis in mouse and human gliomas pre- and post-immunotherapy revealed fewer immunogenic neoantigens after therapy (p < 0.05). Longitudinal human biopsies revealed an initial increase in neoantigen burden during immunotherapy response, followed by a marked loss of neoantigens at relapse. Significance This trans-species analysis confirms CNS as an immune “sanctuary” site and highlights continuous organ-specific immunoediting, which determines tumor progression and response to immunotherapy.

Multimodal Contrastive Learning (CL) has shown significant performance in aligning representations across various data modalities and improving downstream tasks, especially in healthcare. It works by minimizing the distance between matched (positive) data modalities, while maximizing the distance between mismatched (negative) samples. Traditional CL frameworks typically assume instance-based correspondence within data batches, treating all non-paired samples as negatives. However, this assumption often fails in medical settings, where samples may share high-level semantic attributes, leading to false negatives that degrade representation quality. In this paper, we propose Multimodal Semantic-Aware Contrastive Learning (MseaCL), a CL framework trained on a pediatric cohort of 3D brain magnetic resonance imaging (MRI) scans and radiology reports. The goal of this framework is to mitigate the impact of semantically similar false negative samples by incorporating semantic similarity between radiology reports, as a guiding signal during the learning process. Our results indicate that applying this framework as a pretraining stage can achieve notable improvements in downstream tasks, e.g., at least a 22.6% increase in the area under the receiver operating characteristic curve (AUC) of pediatric brain tumor molecular classification, demonstrating its potential for more robust and semantically aligned multimodal representations in clinical applications.

Abstract Introduction Targeted therapy is emerging as frontline care for pLGG, yet optimal dose and duration remain unclear. Long-term use risks chronic toxicity, while abrupt cessation can trigger rebound growth, and retreatment response is less favorable. A structured weaning strategy may preserve tumor control while reducing toxicity and financial burden. Methods We analyzed children with pLGG treated with trametinib at our institution to evaluate (1) the minimum dose required for tumor stability, (2) proportion of regrowth after treatment cessation, and (3) differences in durability of tumor control between the weaning and retreatment cohorts. Results Fifty-five patients met inclusion criteria (non-BRAF V600E altered pLGG treated with trametinib). Most pLGG were BRAF fused (69%), followed by NF1-associated (20%). Eight patients (14%) progressed on trametinib, predominantly those with disseminated pLGG, and were excluded from further analysis.When examining the dose of trametinib which enabled stable tumor response, 70% of patients (33/47) maintained stable response at a median of 67% of weight-based dose (range: 26-96%). Of patients who stopped therapy after tumor response/stabilization, 64% (25/39) experienced regrowth at a median 0.63 year (range: 0.3-4 years); of those, 92% (23/25) required further systemic treatment. We therefore initiated a weaning protocol for patients after >1.5 years of stability and compared durability of control against patients who stopped therapy with subsequent retreatment. The tumor size of patients who stopped treatment was on average 48% larger at the end of the second trametinib treatment compared to the best response during first treatment. In contrast, the tumor size of patients who weaned treatment differed only by 4% between best response and the current size during weaning (p = 0.03). Conclusion This study provides evidence for safety and efficacy of weaning rather than abrupt stopping of therapy. These findings guide future strategies which may reduce treatment costs and side effects while ensuring tumor control in this chronic disease.

BACKGROUND:Plexiform neurofibromas (PNs) are observed in up to 50% of patients with neurofibromatosis type 1 (NF1). Trametinib has been used widely to treat PNs but limited data has been reported on its efficacy within a clinical trial. We conducted a phase 2 trial with trametinib for patients with NF1 and inoperable PNs. METHODS:Patients with NF1 and inoperable PNs aged ≥1 month to ≤25 years when starting trametinib were eligible. Patients received trametinib once daily orally continuously for a maximum of 18 cycles of 28 days. Patients completed patient-reported outcome measures. Response evaluations were performed after cycle 3, 6, 12 and at the end of treatment and then every 6 months. The PN volumes were centrally quantified using a semi-automatic 3D segmentation method. RESULTS:Forty-six patients were recruited. The median age was 11.1 years (range 0.7-19.8). The median baseline tumor volume was 104.6 ml (range 16.3 to 650.1 ml). Volumetric assessment demonstrated an overall response rate (volume decrease of ≥20%) in 47.8% (22/46) of patients. Median volume change was -19.3% (range -80.5 to 3.5). Forty-one patients (89.1%) completed the planned treatment. The 3 years progression free survival and the event free survival were respectively 73.1% (95% CI 53.3%-85.5%) and 47.1% (95% CI 28.8-63.5). CONCLUSIONS:We report outcomes and volumetric response of PNs treated with trametinib within a phase 2 trial. Based on these results, trametinib is effective and overall well tolerated.

Abstract Background Pediatric low-grade gliomas (PLGG) are the most frequent brain tumors in children. The majority of PLGG have activation of the MAPK/ERK pathway. Methods This phase II trial includes three recurrent/refractory PLGG groups: NF1 PLGG, KIAA1549-BRAF fusion PLGG and PLGG with other MAPK/ERK pathway activation (excluding BRAF V600E). Treatment phase has been completed by all patients with PLGG. The primary objective was to evaluate the overall response rate based on modified RANO criteria after daily oral trametinib administration for 18 cycles, each cycle lasting 28 days. Results As of November 05, 2024 recruitment was completed with 68 PLGG patients enrolled (NF1: n = 12; KIAA1549-BRAF fusion: n = 42; other: n = 14 including 7 patients with FGFR alterations). Median age at enrollment was 9.2 years (range 1.8-25.4). Median follow-up was 39.5 months (range 0.9-57.2).Sixty-five patients were evaluable to assess response. The overall best responses include: 1 complete response (1.5%), 12 partial response (PR) (18.2%), 17 minor response (MR) (25.8%), 33 stable disease (50%), 3 progressive disease (4.6%). Median time to response was 2.8 months (range 2.4-13.6) with a median duration of response of 23.8 months (range 0-49.9). Treatment was discontinued for 68 patients: including 41 (60.3%) after completing 18 cycles as planned, 8 (11.8%) for progressive disease and 7 (10.3%) for adverse events. Median progression-free (PFS) survival was 33.3 months (95% CI 23.0-48.8), 18 months PFS was 73.2 % (95% CI 23.0-48.8) and the 36 months PFS was 45.5 % (95% CI 38.5-52.5). Fourteen (20.6%) patients progressed after discontinuation of treatment at a median time of 7.4 months (0.1-18 months). Eighteen patients restarted treatment after discontinuation. Median time to restart treatment was 16.3 months (range 2.1-41.7) Conclusion Trametinib can be an effective and well tolerated therapy for patients with recurrent/refractory PLGG

Abstract Background Survivors of pediatric low-grade glioma (pLGG), a chronic disease, experience significant health sequelae. No population-based studies have characterized their health and healthcare service utilization while accounting for tumor location, biology, and treatment. Methods Using a provincial pediatric cancer registry, all 5-year pLGG survivors diagnosed in Ontario, Canada between 1991-2023 at < 18 years of age were matched 1:5 to cancer-free controls based on age, sex, and geography. Individuals were followed from pLGG diagnosis (index) until December 2024/censorship (death, new cancer event). Clinical data were linked to health administrative databases to estimate the cumulative incidences/cause-specific hazard ratios (HR) of mortality, stroke, severe hearing loss requiring aids, and disability support, accounting for matching and competing risks. Rates per 1,000 person-year and HR of emergency department (ED) visits and hospitalizations were also estimated. Results The final N = 1,415 pLGG survivor cohort (N = 7,075 controls) was followed for a median of 13 years [interquartile range (IQR) 7-21] and diagnosed at a median age of 8 years (IQR 4-12); 88 (6.2%) were irradiated and 223 (23.5%) received chemotherapy in childhood. Tumors were predominantly cerebellar (N = 381, 26.9%), hemispheric (N = 373, 26.4%), and optic pathway/hypothalamic (N = 256, 18.1%). Documented molecular alterations (N = 530, 37.5%) included BRAF fusions (N = 237, 44.7%), Neurofibromatosis Type 1 mutations (N = 149, 28.1%), and BRAF-V600E mutations (N = 111, 20.9%). At 20 years post-index, the overall survival was 95.9% in survivors versus 99.7% in controls (HR 17.7, 95%CI 9.4-33.3), yet survivors were at significant risk of stroke (HR 75.1, 95%CI 23.3-242.3), hearing loss (HR 3.0, 95%CI 1.5-5.9), disability (HR 4.5, 95%CI 3.9-5.2), and healthcare utilization (ED visits HR 1.9, 95%CI 1.7-1.9; hospitalizations HR 9.8, 95%CI 8.7-10.9). Summary In this large contemporary pLGG cohort, we demonstrated excellent survival but significant morbidity among survivors, despite radiation omission for most. These disparities must be addressed by minimizing treatment toxicity, preventative measures, and comprehensive survivorship care.

BACKGROUND:Pediatric high-grade gliomas (pHGG) are the leading cause of childhood cancer-related deaths. Those arising in the midline harboring a lysine to methionine substitution at position 27 in histone 3 (H3K27M), termed diffuse midline glioma (or DIPG when occurring in the pons), are particularly deadly and in need of additional therapeutic options. Our group and others have found upregulation of the RAS/MAPK pathway across HGGs, including DIPG; however, RAS is notoriously difficult to target therapeutically, with no approved drugs that can target non-mutant RAS proteins. METHODS:We leveraged a pan-RAS-cleaving biologic (RRSP-DTB) to define RAS dependency in DIPG and employed proteomic profiling of patient-derived tumors to identify enriched cell-surface receptors for tumor-selective targeting. RRSP was re-engineered to engage the lead candidate receptor, tumor endothelial marker 8 (TEM8/ANTXR1), generating a targeted RAS-degrading biologic evaluated in vitro and in orthotopic DIPG xenograft models. RESULTS:Pan-RAS cleavage revealed DIPG to be critically dependent on RAS/MAPK signaling compared to other RAS/MAPK HGGs. Targeting RRSP to TEM8 enabled efficient intracellular delivery, resulting in complete RAS ablation and apoptotic cell death in patient-derived DIPG cells at low picomolar concentrations. In vivo, intracerebroventricular delivery reduced leptomeningeal disease burden (p = 0.04), while convection-enhanced delivery extended survival in orthotopic DIPG models (median survival 68 vs. 57 days; p = 0.016). CONCLUSIONS:We demonstrate RAS-dependency in DIPG targetable by RRSP-DTT-TEM8, a novel, first-in-class pan-RAS biotherapeutic, supporting its potential as a therapeutic strategy for this otherwise untreatable disease.

ALK and ROS1 fusions are key drivers of infant-type hemispheric gliomas (IHG). With diverse gene partners, the impact of ALK and ROS1 oncoprotein heterogeneity on glioma biology remains unknown. We developed an integrative phospho-proteomic and transcriptomic approach to discover biological functions regulated by five IHG-associated fusions: CCDC88A::ALK, PPP1CB::ALK, GOPC::ROS1, CLIP1::ROS1, and KIF21A::ROS1. Here, we report fusion-specific oncogenic functions conferred by the 5' gene partner, including increased cell motility driven by microtubule-interacting fusions CCDC88A::ALK and CLIP1::ROS1. All studied fusions converge on STAT3 activation. Using affinity purification mass spectrometry, we identified SHP2 in direct interaction with all three ROS1 oncoproteins but with none of the ALK oncoproteins, which in turn interact with SHC1/SHC3. ROS1 fusions phosphorylate SHP2 to a greater extent than ALK fusions, and analyses of downstream pathways suggest MAPK-independent, non-canonical SHP2-driven functions. Our findings reveal both common and fusion-specific dependencies, offering opportunities to optimize therapeutic strategies for pediatric gliomas.

Abstract There is an unmet need for effective treatment strategies for diffuse midline glioma (DMG), a devastating pediatric brain tumor. Clinical trials with immune checkpoint inhibitors that are efficacious in other tumors have failed to show a survival benefit for DMG patients. In this study, we analyzed the expression of several known immune checkpoint molecules in human and murine DMG cells by flow cytometry and found that the poliovirus receptor CD155 (an adhesion molecule that interacts with CD96 and TIGIT on CD8+ T cells) is highly expressed in all DMG lines and models. This finding was confirmed in primary pediatric brain tumor samples. To test whether CD155 regulates susceptibility to CD8+ T cell killing, we cultured murine DMG cells with CD8+ T cells. shRNA mediated silencing of CD155 led to a marked increase in T cell-mediated killing in vitro. Strikingly, CD155-deficient DMG cells failed to grow at all in immunocompetent mice, while depletion of CD8+ T cells allowed these tumors to grow. Intriguingly, CD155 also exerted cell-autonomous effects on tumor cells: silencing of CD155 led to induction of apoptosis of DMG cells even in the absence of T cells and led to delayed tumor growth even in immunocompromised mice. Transcriptomic analyses identified the transcription factor FOXM1 as a key target that is silenced in CD155-deficient cells. FOXM1 silencing also led to reduced proliferation of DMG cells in vitro and in vivo, and treatment of DMG-bearing mice with Thiostrepton, a FOXM1-targeting antibiotic, delayed tumor growth and prolonged survival. These studies demonstrate that CD155 functions as a modulator of tumor cell sensitivity to T cells and also regulates tumor cell survival in a T cell-independent manner. Our studies suggest that targeting CD155 or its downstream mediators could be a valuable double-pronged therapeutic strategy for this devastating disease.

Abstract Background Analysis of cerebrospinal fluid (CSF) circulating tumor DNA (ctDNA) has shown promise in multiple brain tumor types. The role of CSF-based liquid biopsy in early detection, surveillance, monitoring response and predicting relapse in patients with cancer predisposition syndromes has not been described. Methods We established a protocol for CSF-ctDNA collection from patients with hereditary replication repair deficiency (RRD) for surveillance, response assessment, and monitoring of gliomas post-therapy. ctDNA was assessed for single-nucleotide variants (SNVs), copy number variants(CNVs) and MMRD signatures (MMRDness score) using a combination of hybrid capture and low-pass whole-genome sequencing. Results A total of 84 CSF samples from 48 patients (CMMRD(32), Lynch (15), POLE(1)) were analyzed. Diagnostic sensitivity for high-grade glioma was 93% with 100% specificity. Three tumors were diagnosed >6 months prior to tumor evidence by diagnostic procedure. CSF-ctDNA status was significantly associated with survival - 58 samples were taken during RRD-glioma immunotherapy, overall survival was 91% and 15% for patients with negative versus positive CSF-ctDNA, respectively (p = 0.005). MMRDness scores were elevated at diagnosis and became negative in association with treatment response. These patients eventually had repeated negative CSF-ctDNA and are being monitored post therapy cessation. Ongoing monitoring of 12 patients post immunotherapy revealed 2 positive samples which were confirmed by biopsy upon positive imaging. One is alive upon changing immunotherapy. Comparative tumor and repeated CSF mutation analysis revealed unique patterns of spatial and temporal clonal evolution shaped by common drivers, new drivers and loss of antigens suggesting immune editing. Conclusions This is the first report of RRD patients benefiting from CSF-ctDNA for early detection, monitoring response and MRD with immunotherapy. In addition to CNV and SNV detection, MMRDness score is a unique quantitative tool to assess tumor burden during therapy. These data can be expanded to other cancer syndromes and malignant glioma management.

BACKGROUND Liquid biopsy has emerged as a minimally invasive method for tumor diagnosis, monitoring, and therapeutic guidance. For CNS tumors, cerebrospinal fluid (CSF) provides a reliable and accessible source of tumor-derived cell-free DNA (ctDNA). METHODS This study evaluates the clinical utility of CSF liquid biopsy in a real-world prospective setting. A total of 148 CSF samples from 120 patients underwent molecular analysis using droplet digital PCR (ddPCR) and/or next-generation sequencing to detect mutations, fusions, copy number alterations, and mismatch-repair deficient signatures (MMRDness). Samples were collected via lumbar puncture ( n = 82; 45% ctDNA positive) or from ventricle sources at the time of surgery or through shunts ( n = 66; 65% ct DNA positive). RESULTS Overall, ctDNA was detected in 54% of samples with higher detection in high-grade gliomas at diagnosis (100%, 1 sample equivocal) compared with low-grade gliomas (50%). Among low-grade gliomas, ctDNA detection was higher in disseminated cases (80% versus 43%) and from ventricular versus lumbar samples (56% versus 38%). CONCLUSION Liquid biopsy distinguished relapse from second malignancy and serial sampling demonstrated the potential for ctDNA levels to track treatment response and disease progression. In patients with MMRD tumors, high MMRDness score from ctDNA supported active disease. These findings demonstrate that combined liquid biopsy assays facilitate diagnosis, monitoring, and personalized treatment decisions, offering a viable alternative to invasive surgical biopsies in pediatric CNS tumors. TRIAL REGISTRATION None. FUNDING Proof of Principle Grant from The Hospital for Sick Children; The Canadian Institutes of Health Research; The Canadian Cancer Society; The We Love You Connie Foundation; Garron Family Cancer Center at SickKids; SickKids Clinician Training Program; Ben Stelter Foundation through the Women and Children’s Health Research Institute; Jeffrey Brock Cancer Genetics Research Fellowship; Garron Family Cancer Center Research Fellowship/Scotiabank Clinician Scientist Fellowship; Atrium/CMCC and Hold’em for Life Oncology Fellowship; Tokyo Children’s Cancer Study Group Scholarship of the Gold Ribbons Network.

Abstract Background Liquid biopsy enables profiling of CNS tumors without surgery and tracking minimal residual disease. Our pilot work indicated high detection rate of circulating tumor DNA (ctDNA) from CSF of patients with CNS germ cell tumors (CNS-GCT) at diagnosis. Herein, we present an expanded, international dataset with longitudinal sampling to evaluate the dynamics of CSF-ctDNA. Methods A multi-institutional cohort of CNS-GCT patients treated with induction chemotherapy and response-adapted radiation was assembled. CSF was collected at diagnosis, during therapy, and after completion of treatment. Cell-free DNA was extracted for studying copy-number variations (CNVs) using low-pass whole-genome sequencing, with additional mutational analysis by panel sequencing when available. ctDNA positivity is inferred by the presence of CNVs +/- mutations. Results One-hundred and fifty-eight samples from 77 patients (median age 13 years; germinoma=48, NGGCT=29) were analyzed. At diagnosis, ctDNA was detected in 95% (37/39) of germinoma, and 88% (15/17) of NGGCT patients, including all patients with negative tumor markers (n = 21). Where mutational analysis was performed (n = 27), oncogenic drivers in the form of amplifications or mutations were identified in 96% (26/27). Longitudinal CSF profiling revealed persistent ctDNA in 27% of samples (6/22) after 1-2 cycles of chemotherapy, and in 25% (6/24) after induction therapy, prior to irradiation. Among patients with persistent ctDNA post-induction, three developed early recurrence despite negative ctDNA following radiation, and one demonstrated refractory disease, representing only events to-date. Evolution of CNV profiles was observed from paired-CSF at diagnosis and relapse, suggesting emergence of resistant clones. Conclusion Liquid biopsy enables the minimally invasive diagnosis and genotyping of CNS-GCTs. Early clearance of ctDNA was observed in the majority, while persistence following induction therapy may serve as a poor prognostic marker. Evidence based on the largest entity-specific CSF cohort supports the prospective incorporation of serial CSF liquid biopsies for personalized, response-adapted, management of CNS-GCTs.

Abstract Background Replication repair maintains genomic fidelity. Germline-inherited replication repair deficiency (RRD) leads to hypermutant, early-onset brain tumours. RRD medulloblastoma (MB) are reported but not well-characterized. Methods We performed multi-omic analyses to elucidate the impact of hypermutation on RRD-MB biology and treatment-outcomes. Results RRD-MB (n = 48) were enriched for anaplastic disease (61%). t-SNE analysis suggested SHH-subgroup affinity despite low confidence of classification on the Heidelberg Classifier (78%). Genomic profile displayed microsatellite instability and hypermutation, with frequent somatic mutations in polymerase proofreading genes (80%). Mutations were also common inTP53 (48%), SHH-pathway genes (56%) and, strikingly, glioma driver genes (ATRX,NF1,RB1; 50%). Copy number changes were infrequent. Comparative analysis of variant allele frequency in tumours with both SHH and glioma drivers indicated that SHH-drivers occur first, suggesting that primary tumor identity is defined by an initial combination of cell of origin and genetic alterations. Deconvolution of single-nuclei RNA-sequencing revealed enrichment for astro-glial expression programs exclusive to RRD-MB compared to non-RRD SHH-MB, suggesting a functional impact of glioma driver mutations. Representative RRD-MB mouse models (Nestin-Cre+/MSH2LoxP/LoxP/POLES459F/+) further substantiated human findings, harbouring both SHH- and glioma-drivers, with ultra-hypermutation and mixed embryonal-glial phenotypes. High levels of immune infiltrates (CD8+ T-cell), in both mouse and human tumors, provided biological insights into the immune mechanisms underlying the efficacy of anti-PD1 monotherapy in human RRD-MB, leading to radiological remission in refractory disease and prolonged survival (p = 0.02). Conclusions RRD-MB form a distinct subtype of SHH-activated MB. Their genetic phenotypes are modified by genomic instability leading to point mutations and indels in both SHH and glioma-pathways. Hypermutation and corresponding immune infiltration confer immune checkpoint inhibitor efficacy at recurrence, allowing the development of prospective combined chemo-immunotherapy treatment protocols for patients with high-risk disease.

Abstract Background Although targeted therapies for paediatric low-grade gliomas (pLGG) have shown efficacy, comparisons of alteration-specific long-term outcomes between targeted therapies and conventional chemotherapy are lacking. Methods We performed a population-based study of BRAF-altered pLGG treated at relapse with trametinib (for BRAF fused pLGG) or dabrafenib +/- trametinib (BRAF V600E), compared with chemotherapy-treated pLGG. Imaging responses (RAPNO) were assessed by neuroradiologists. Results The initial cohort comprised 267 BRAF fused and 161 BRAF V600E pLGG. Across alterations and lines of therapy, imaging responses for chemotherapy-treated pLGG were similar. Clinician-determined progression-free survival (PFS) was poor and, importantly, equivalent between first/second-line chemotherapy. For BRAF fused pLGG treated with chemotherapy or trametinib, there was a higher objective response rate (>25% reduction) with trametinib – 60% (18/30) versus 36% (8/22) [second-line] versus 39% (22/56) [first-line]. Trametinib yielded significantly more partial responses. Moreover, 56% of tumours refractory/non-responsive to first-line chemotherapy responded to trametinib. These responses translated to superior disease control with 18-month PFS of 91% (n = 34) versus 52% for second-line chemotherapy (n = 27) (p < 0.001). At 3.5 years, PFS for patients who remained on therapy was 82% versus 19% for chemotherapy. These differences narrowed for patients who discontinued trametinib, with 49% progressing within 2 years. Interestingly, disseminated tumours were more likely to progress on therapy – 67% probability at 5 years (p = 0.001). BRAF V600E pLGG treated with targeted therapy (n = 19) also showed superior disease control with 5-year PFS of 82% versus 33% for second-line chemotherapy (n = 9) (p = 0.02). With prolonged treatment times (median 5.5 years [1-10.9]), tumour control persisted – only 16% of patients progressed on therapy. Review of V600E imaging responses is ongoing. Conclusion Acknowledging variation in treatment courses and follow-up, targeted therapy for relapsed BRAF-altered pLGG yields superior tumour control relative to chemotherapy. Ongoing work will help elucidate optimal treatment schedules and long-term functional outcomes.

Pediatric brain tumours are highly prevalent and remain one of the leading causes of cancer-related deaths in children. There are numerous different brain tumour types that are now well characterized by magnetic resonance imaging (MRI), patient clinical course, neuropathological and molecular genetic alterations. One of the challenges with treating pediatric brain tumours with systemic chemotherapy is the inability of several chemotherapeutic agents to cross the blood brain barrier (BBB) which serves as a protective mechanism for neuronal homeostasis. The BBB is primarily comprised of microvascular endothelial tight junctions. Controlling BBB permeability to allow for therapeutics to cross and combat brain tumors is now possible using MR-guided Focused Ultrasound (MRgFUS). In this approach, microbubbles are administered intra-venously prior to MRgFUS BBB disruption at the targeted tumour site in the brain. In the presence of MRgFUS, the microbubbles in the brain capillaries oscillate, and temporarily disrupt the BBB enabling systemically administered chemotherapy drugs to cross at the targeted site. In this review, we provide evidence supporting the use of MRgFUS BBB disruption to treat brain tumours in animal models, and in on-going human clinical drug trials. We conclude with efforts to harness the potency of the immune system using MRgFUS against pediatric brain tumours.

Abstract Replication repair deficiency (RRD), caused by germline biallelic mutations in mismatch repair, leads to hypermutant brain tumour development in children and adolescents. RRD medulloblastoma have been reported but how RRD driven mutagenesis contributes to their clinical, genomic, and immune profiles is unknown. Through the International RRD Consortium, we enrolled 43 RRD medulloblastoma, analysed their exomes, methylomes, transcriptomes, and clinical outcomes. To better understand the mechanisms underlying their development and to assess preclinical responses to immunotherapy, we analyzed the histology and exomes of embryonal brain tumours developed in RRD mouse models (Nestin-Cre+/MSH2LoxP/LoxP/POLES459F/+). RRD medulloblastoma were enriched for anaplasia (61%), localised disease (77%), and harboured hypermutation and microsatellite instability, contrasting with “quiet” genomic profiles of non-RRD medulloblastoma (p<0.0001). Point mutations were frequent in POLE/POLD1 (80%), TP53 (48%), SHH pathway genes (56%) and, notably, glioma driver genes (ATRX, NF1, RB1: 50%). TP53 hotspot mutations were distinct from non-RRD medulloblastoma and occurred in trinucleotide sequence contexts which were highly mutated by RRD mutational signatures, suggesting that replication errors shape their genetic evolution. Copy-number changes were infrequent (<20%) and inversely correlated with tumour mutation burden. While most did not classify with high confidence (>0.9) on DNA methylation subtyping tools, RRD medulloblastoma shared methylation programs with SHH medulloblastoma. Non-promoter genomic regions were hypomethylated relative to non-RRD medulloblastoma, partially explaining their failure to perform in DNA methylation classifiers. Transcriptomics reflected shared expression programs between RRD and SHH medulloblastomas. RRD medulloblastoma demonstrated high CD8+ T-cell infiltration and response to anti-PD1 monotherapy, with 60% three-year progression-free survival. Outcome was worse for tumours with TP53-mutation (p=0.04). Recurrent/progressive RRD medulloblastoma treated with anti-PD1 monotherapy had prolonged survival compared those which were not (p = 0.02), including responses in TP53-mutant tumours. RRD mouse models developed tumours that mirrored histological and genetic profiles of human disease and responded to anti-PD1 monotherapy, highlighting their relevance for future preclinical studies. Human RRD medulloblastoma are distinguished by their unique spectrum of driver mutations and hypomethylation profiles, shaped by RRD mutagenesis. Mouse models recapitulate human features and offer a mechanistic model for tumorigenesis and immunotherapy responses. Importantly, immune-hot microenvironments in RRD medulloblastoma contribute to successful salvage therapy for tumours failing chemo-radiation. Citation Format: Nicholas R. Fernandez, Anirban Das, Adrian Levine, Kyle Smith, Evan Wang, Melissa Galati, Zoya Aamir, Jiil Chung, Logine Negm, Hope Friedman, Katharine O'Flaherty, Owen Crump, Quang M. Trinh, Nuno M. Nunes, Vanessa Bianchi, Lucie Stengs, Melissa Edwards, Lincoln Stein, Eric Bouffet, Michael D. Taylor, Paul Northcott, Vijay Ramaswamy, Cynthia Hawkins, Uri Tabori. Trans-species analysis of replication repair deficient medulloblastoma and response to immune checkpoint inhibition: An IRRDC report [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 1159.

Abstract Background Vorasidenib has changed the therapeutic landscape for low-grade gliomas (LGG) with IDH1/IDH2 mutations in adults. IDH mutant (mIDH)-LGG are rare in children and pediatric data for inhibitor use is lacking. We aimed to investigate the prevalence and profile of pediatric mIDH LGG and report our early experience using vorasidenib in children. Methods Using a dual approach, we performed a retrospective population-based analysis of IDH1/IDH2 mutations in all pediatric LGG diagnosed between 1986-2025 and initiated a prospective international registry-based follow-up for children (12-18 years) on vorasidenib. Results IDH1/IDH2 mutations were detected in 2.4% (n = 30/1255) of pediatric LGG. Median age was 15-years (range: 8.8-18), with predominance of IDH1 mutation (93.4%, vs IDH2: 6.7%), and relative frequency of astrocytoma (56.6%) over oligodendroglioma (1p/19q codeletion: 43.4%). ATRX and TP53 mutations were noted in 20% of pediatric mIDH astrocytoma. 13.4% had syndromic associations (Li Fraumeni, Olliers, ANDP, biallelic MBD4-deficiency). While complete resection (23.3%) was the primary goal, diffuse infiltrative/multifocal disease resulted in subtotal resection (30%) or biopsy (46.7%) followed by neuroimaging surveillance in the majority. Ten children received vorasidenib following radiological progression, with a median follow-up of 16.9 (range: 16.4–17.3 months) months on drug. Common adverse effects (≥15%) included fatigue, headache, dizziness, and nausea. Grade 3 transaminitis persisted or recurred, mandating dose-reduction in 40% of patients. Weight gain (n = 2/10) was reported. Importantly, serial reduction in tumor growth-rate and radiological stabilization was observed in n = 9/10 on vorasidenib. None received radiation. Improved seizure-control was reported in 50% with this presenting symptom. Biomarker and CSF ctDNA analyses are ongoing. Conclusions We highlight the impact and challenges of using vorasidenib in mIDH low grade glioma in childhood. Pediatric safety and pharmacokinetic studies are being developed. Combined surveillance and biomarker analyses will continue over the next decade to investigate whether treatment in childhood can impact the natural history and mitigate transformation risk, allowing mIDH-glioma interception.

Pediatric low-grade glioma (pLGG) is the most common pediatric cancer type, accounting for 40% of brain tumours in children, and identifying pLGG molecular subtype is crucial for treatment planning. However, the gold standard to determine the pLGG molecular subtype is biopsy, which can be impractical or dangerous for patients. This research improves the performance of convolutional neural networks (CNNs) in classifying pLGG molecular subtypes through Magnetic Resonance Imaging ( MRI) scans by introducing a loss function that specifically improves the model's Area Under the Receiver Operating Characteristic Curve (AUROC), offering a non-invasive diagnostic alternative. In this study, a retrospective dataset of 339 children with pLGG (143 BRAF fusion, 71 with BRAF V600E mutation, and 125 nonBRAF) was curated. The baseline CNN model was trained using binary cross entropy (BCE), and achieved an AUROC of 86.11% for differentiating BRAF fusion and BRAF V600E mutations, which was improved to 87.71% using our proposed AUROC loss function (p-value 0.045). With multiclass classification, the AUROC improved from 74.42% to 76.59% (p-value 0.002). The proposed AUROC loss function significantly improved the performance in identifying both binary and multiclass scenarios.

Abstract Background Cerebrospinal fluid (CSF) circulating tumor DNA (ctDNA) offers a minimally invasive means into the genomic and epigenomic landscape of pediatric CNS tumors. It outperforms conventional cytology and can often detect recurrence prior to radiographic changes. However, interpretation is highly dependent on sampling context and timing. Disease-specific schedules for CSF collection have yet to be developed. Methods Members of the Canadian National Pediatric Brain Tumor Consortium (CPBTC) developed a consensus for CSF collection schedules across major pediatric CNS tumor entities. This was informed by Canadian multicenter CSF ctDNA experience (>550 specimens), and existing institutional practices for CSF liquid biopsy. Results The panel generated harmonized, disease-specific CSF sampling schedules encompassing diagnosis, post-operative and post-radiation, on-treatment, end-of-therapy MRD evaluation, and structured surveillance timepoints for medulloblastoma, CNS embryonal tumors, ependymoma, CNS germ cell tumors, and low- and high-grade glioma. Recommendations integrate feasibility constraints (age, access, volume), conventional protocol-specific evaluation timepoints, and procedural considerations (lumbar puncture vs Ommaya reservoir, use of stabilizing tubes when sample processing is delayed) to enable safe, reproducible sampling. Conclusion This consensus provides the first harmonized, pragmatic framework for CSF collection to support standardized ctDNA interpretation and MRD-informed decision-making in children with brain tumors. Implementation of these schedules across centers is expected to enhance comparability of results, facilitate multicenter trials, and accelerate evidence generation for liquid biopsy-guided precision therapy in pediatric neuro-oncology.