Early prognostication of overall survival for pediatric diffuse midline gliomas using MRI radiomics and machine learning: a two-center study.

Background: Diffuse midline gliomas (DMG) are aggressive pediatric brain tumors that are diagnosed and monitored through MRI. We developed an automatic pipeline to segment subregions of DMG and select radiomic features that predict patient overall survival (OS). Methods: We acquired diagnostic and post-radiation therapy (RT) multisequence MRI (T1, T1ce, T2, T2 FLAIR) and manual segmentations from two centers of 53 (internal cohort) and 16 (external cohort) DMG patients. We pretrained a deep learning model on a public adult brain tumor dataset, and finetuned it to automatically segment tumor core (TC) and whole tumor (WT) volumes. PyRadiomics and sequential feature selection were used for feature extraction and selection based on the segmented volumes. Two machine learning models were trained on our internal cohort to predict patient 1-year survival from diagnosis. One model used only diagnostic tumor features and the other used both diagnostic and post-RT features. Results: For segmentation, Dice score (mean [median]±SD) was 0.91 (0.94)±0.12 and 0.74 (0.83)±0.32 for TC, and 0.88 (0.91)±0.07 and 0.86 (0.89)±0.06 for WT for internal and external cohorts, respectively. For OS prediction, accuracy was 77% and 81% at time of diagnosis, and 85% and 78% post-RT for internal and external cohorts, respectively. Homogeneous WT intensity in baseline T2 FLAIR and larger post-RT TC/WT volume ratio indicate shorter OS. Conclusions: Machine learning analysis of MRI radiomics has potential to accurately and non-invasively predict which pediatric patients with DMG will survive less than one year from the time of diagnosis to provide patient stratification and guide therapy.

Early prognostication of overall survival for pediatric diffuse midline gliomas using MRI radiomics and machine learning: A two-center study.

Background: Diffuse midline gliomas (DMG) are aggressive pediatric brain tumors that are diagnosed and monitored through MRI. We developed an automatic pipeline to segment subregions of DMG and select radiomic features that predict patient overall survival (OS). Methods: We acquired diagnostic and post-radiation therapy (RT) multisequence MRI (T1, T1ce, T2, and T2 FLAIR) and manual segmentations from 2 centers: 53 from 1 center formed the internal cohort and 16 from the other center formed the external cohort. We pretrained a deep learning model on a public adult brain tumor data set (BraTS 2021), and finetuned it to automatically segment tumor core (TC) and whole tumor (WT) volumes. PyRadiomics and sequential feature selection were used for feature extraction and selection based on the segmented volumes. Two machine learning models were trained on our internal cohort to predict patient 12-month survival from diagnosis. One model used only data obtained at diagnosis prior to any therapy (baseline study) and the other used data at both diagnosis and post-RT (post-RT study). Results: Overall survival prediction accuracy was 77% and 81% for the baseline study, and 85% and 78% for the post-RT study, for internal and external cohorts, respectively. Homogeneous WT intensity in baseline T2 FLAIR and larger post-RT TC/WT volume ratio indicate shorter OS. Conclusions: Machine learning analysis of MRI radiomics has potential to accurately and noninvasively predict which pediatric patients with DMG will survive less than 12 months from the time of diagnosis to provide patient stratification and guide therapy.

Tumor volume features predict survival outcomes for patients diagnosed with diffuse intrinsic pontine glioma.

Background: Diffuse intrinsic pontine glioma (DIPG) is a fatal childhood central nervous system tumor. Diagnosis and monitoring of tumor response to therapy is based on magnetic resonance imaging (MRI). MRI-based analyses of tumor volume and appearance may aid in the prediction of patient overall survival (OS). Methods: Contrast-enhanced T1- and FLAIR/T2-weighted MR images were retrospectively collected from children with classical DIPG diagnosed by imaging (n = 43 patients). MRI features were evaluated at diagnosis (n = 43 patients) and post-radiation (n = 40 patients) to determine OS outcome predictors. Features included 3D tumor volume (Twv), contrast-enhancing tumor core volume (Tc), Tc relative to Twv (TC/Twv), and Twv relative to whole brain volume. Support vector machine (SVM) learning was used to identify feature combinations that predicted OS outcome (defined as OS shorter or longer than 12 months from diagnosis). Results: Features associated with poor OS outcome included the presence of contrast-enhancing tumor at diagnosis, >15% Tc/Twv post-radiation therapy (RT), and >20% ∆Tc/Twv post-RT. Consistently, SVM learning identified Tc/Twv at diagnosis (prediction accuracy of 74%) and ∆Tc/Twv at <2 months post-RT (accuracy = 75%) as primary features of poor survival. Conclusions: This study demonstrates that tumor imaging features at diagnosis and within 4 months of RT can predict differential OS outcomes in DIPG. These findings provide a framework for incorporating tumor volume-based predictive analyses into the clinical setting, with the potential for treatment customization based on tumor risk characteristics and future applications of machine-learning-based analysis.

Pan-cancer atlas of somatic core and linker histone mutations.

Recent genomic data points to a growing role for somatic mutations altering core histone and linker histone-encoding genes in cancer. However, the prevalence and the clinical and biological implications of histone gene mutations in malignant tumors remain incompletely defined. To address these knowledge gaps, we analyzed somatic mutations in 88 linker and core histone genes across 12,743 tumors from pediatric, adolescent and young adult (AYA), and adult cancer patients. We established a pan-cancer histone mutation atlas contextualized by patient age, survival outcome, and tumor location. Overall, 11% of tumors harbored somatic histone mutations, with the highest rates observed among chondrosarcoma (67%), pediatric high-grade glioma (pHGG, >60%), and lymphoma (>30%). Previously unreported histone mutations were discovered in pHGG and other pediatric brain tumors, extending the spectrum of histone gene alterations associated with these cancers. Histone mutation status predicted patient survival outcome in tumor entities including adrenocortical carcinoma. Recurrent pan-cancer histone mutation hotspots were defined and shown to converge on evolutionarily conserved and functional residues. Moreover, we studied histone gene mutations in 1700 pan-cancer cell lines to validate the prevalence and spectrum of histone mutations seen in primary tumors and derived histone-associated drug response profiles, revealing candidate drugs targeting histone mutant cancer cells. This study presents the first-of-its-kind atlas of both core and linker histone mutations across pediatric, AYA, and adult cancers, providing a framework by which specific cancers may be redefined in the context of histone and chromatin alterations.

PNOC015: Repeated convection-enhanced delivery of MTX110 (aqueous panobinostat) in children with newly diagnosed diffuse intrinsic pontine glioma.

BACKGROUND: The objective of this study was to determine the safety, tolerability, and distribution of MTX110 (aqueous panobinostat) delivered by convection-enhanced delivery (CED) in patients with newly diagnosed diffuse intrinsic pontine glioma (DIPG) who completed focal radiation therapy (RT). METHODS: Patients with DIPG (2-21 years) were enrolled after RT. CED of MTX110 combined with gadoteridol was completed across 7 dose levels (DL) (30-90 µM; volumes ranging from 3 mL to 2 consecutive doses of 6 mL). An accelerated dose escalation design was used. Distribution of infusate was monitored with real-time MR imaging. Repeat CED was performed every 4-8 weeks. Quality-of-life (QoL) assessments were obtained at baseline, every 3 months on therapy, and end of therapy. RESULTS: Between May 2018 and March 2020, 7 patients who received a total of 48 CED infusions, were enrolled (median age 8 years, range 5-21). Three patients experienced dose-limited toxicities. Four grade 3 treatment-related adverse events were observed. Most toxicities were transient new or worsening neurologic function. Median overall survival (OS) was 26.1 months (95% confidence interval: 14.8-not reached). Progression-free survival was 4-14 months (median, 7). Cumulative percentage of tumor coverage for combined CED infusions per patient ranged from 35.6% to 81.0%. Increased CED infusions were negatively associated with self-reported QoL assessments. CONCLUSION: Repeat CED of MTX110 with real-time imaging with gadoteridol is tolerable for patients with DIPG. Median OS of 26.1 months compares favorably with historical data for children with DIPG. The results support further investigation of this strategy in a larger cohort.

Circulating tumor DNA sequencing provides comprehensive mutation profiling for pediatric central nervous system tumors.

Molecular profiling of childhood CNS tumors is critical for diagnosis and clinical management, yet tissue access is restricted due to the sensitive tumor location. We developed a targeted deep sequencing platform to detect tumor driver mutations, copy number variations, and heterogeneity in the liquid biome. Here, we present the sensitivity, specificity, and clinical relevance of our minimally invasive platform for tumor mutation profiling in children diagnosed with CNS cancer.

Upfront Biology-Guided Therapy in Diffuse Intrinsic Pontine Glioma: Therapeutic, Molecular, and Biomarker Outcomes from PNOC003.

PURPOSE: PNOC003 is a multicenter precision medicine trial for children and young adults with newly diagnosed diffuse intrinsic pontine glioma (DIPG). PATIENTS AND METHODS: Patients (3-25 years) were enrolled on the basis of imaging consistent with DIPG. Biopsy tissue was collected for whole-exome and mRNA sequencing. After radiotherapy (RT), patients were assigned up to four FDA-approved drugs based on molecular tumor board recommendations. H3K27M-mutant circulating tumor DNA (ctDNA) was longitudinally measured. Tumor tissue and matched primary cell lines were characterized using whole-genome sequencing and DNA methylation profiling. When applicable, results were verified in an independent cohort from the Children's Brain Tumor Network (CBTN). RESULTS: Of 38 patients enrolled, 28 patients (median 6 years, 10 females) were reviewed by the molecular tumor board. Of those, 19 followed treatment recommendations. Median overall survival (OS) was 13.1 months [95% confidence interval (CI), 11.2-18.4] with no difference between patients who followed recommendations and those who did not. H3K27M-mutant ctDNA was detected at baseline in 60% of cases tested and associated with response to RT and survival. Eleven cell lines were established, showing 100% fidelity of key somatic driver gene alterations in the primary tumor. In H3K27-altered DIPGs, TP53 mutations were associated with worse OS (TP53mut 11.1 mo; 95% CI, 8.7-14; TP53wt 13.3 mo; 95% CI, 11.8-NA; P = 3.4e-2), genome instability (P = 3.1e-3), and RT resistance (P = 6.4e-4). The CBTN cohort confirmed an association between TP53 mutation status, genome instability, and clinical outcome. CONCLUSIONS: Upfront treatment-naïve biopsy provides insight into clinically relevant molecular alterations and prognostic biomarkers for H3K27-altered DIPGs.

Imipridones affect tumor bioenergetics and promote cell lineage differentiation in diffuse midline gliomas.

BACKGROUND: Pediatric diffuse midline gliomas (DMGs) are incurable childhood cancers. The imipridone ONC201 has shown early clinical efficacy in a subset of DMGs. However, the anticancer mechanisms of ONC201 and its derivative ONC206 have not been fully described in DMGs. METHODS: DMG models including primary human in vitro (n = 18) and in vivo (murine and zebrafish) models, and patient (n = 20) frozen and FFPE specimens were used. Drug-target engagement was evaluated using in silico ChemPLP and in vitro thermal shift assay. Drug toxicity and neurotoxicity were assessed in zebrafish models. Seahorse XF Cell Mito Stress Test, MitoSOX and TMRM assays, and electron microscopy imaging were used to assess metabolic signatures. Cell lineage differentiation and drug-altered pathways were defined using bulk and single-cell RNA-seq. RESULTS: ONC201 and ONC206 reduce viability of DMG cells in nM concentrations and extend survival of DMG PDX models (ONC201: 117 days, P = .01; ONC206: 113 days, P = .001). ONC206 is 10X more potent than ONC201 in vitro and combination treatment was the most efficacious at prolonging survival in vivo (125 days, P = .02). Thermal shift assay confirmed that both drugs bind to ClpP, with ONC206 exhibiting a higher binding affinity as assessed by in silico ChemPLP. ClpP activation by both drugs results in impaired tumor cell metabolism, mitochondrial damage, ROS production, activation of integrative stress response (ISR), and apoptosis in vitro and in vivo. Strikingly, imipridone treatment triggered a lineage shift from a proliferative, oligodendrocyte precursor-like state to a mature, astrocyte-like state. CONCLUSION: Targeting mitochondrial metabolism and ISR activation effectively impairs DMG tumorigenicity. These results supported the initiation of two pediatric clinical trials (NCT05009992, NCT04732065).

Mechanisms of imipridones in targeting mitochondrial metabolism in cancer cells.

ONC201 is the first member of the imipridone family of anticancer drugs to enter the clinic for the treatment of diverse solid and hematologic cancers. A subset of pediatric and adult patients with highly aggressive brain tumors has shown remarkable clinical responses to ONC201, and recently, the more potent derivative ONC206 entered clinical trials as a single agent for the treatment of central nervous system (CNS) cancers. Despite the emerging clinical interest in the utility of imipridones, their exact molecular mechanisms are not fully described. In fact, the existing literature points to multiple pathways (e.g. tumor necrosis factor (TNF)-related apoptosis-inducing ligand (TRAIL) signaling, dopamine receptor antagonism, and mitochondrial metabolism) as putative drug targets. We have performed a comprehensive literature review and highlighted mitochondrial metabolism as the major target of imipridones. In support of this, we performed a meta-analysis of an ONC201 screen across 539 human cancer cell lines and showed that the mitochondrial caseinolytic protease proteolytic subunit (ClpP) is the most significant predictive biomarker of response to treatment. Herein, we summarize the main findings on the anticancer mechanisms of this potent class of drugs, provide clarity on their role, and identify clinically relevant predictive biomarkers of response.

Standardization of the liquid biopsy for pediatric diffuse midline glioma using ddPCR.

Diffuse midline glioma (DMG) is a highly morbid pediatric brain tumor. Up to 80% of DMGs harbor mutations in histone H3-encoding genes, associated with poor prognosis. We previously showed the feasibility of detecting H3 mutations in circulating tumor DNA (ctDNA) in the liquid biome of children diagnosed with DMG. However, detection of low levels of ctDNA is highly dependent on platform sensitivity and sample type. To address this, we optimized ctDNA detection sensitivity and specificity across two commonly used digital droplet PCR (ddPCR) platforms (RainDance and BioRad), and validated methods for detecting H3F3A c.83A > T (H3.3K27M) mutations in DMG CSF, plasma, and primary tumor specimens across three different institutions. DNA was extracted from H3.3K27M mutant and H3 wildtype (H3WT) specimens, including H3.3K27M tumor tissue (n = 4), CSF (n = 6), plasma (n = 4), and human primary pediatric glioma cells (H3.3K27M, n = 2; H3WT, n = 1). ctDNA detection was enhanced via PCR pre-amplification and use of distinct custom primers and fluorescent LNA probes for c.83 A > T H3F3A mutation detection. Mutation allelic frequency (MAF) was determined and validated through parallel analysis of matched H3.3K27M tissue specimens (n = 3). We determined technical nuances between ddPCR instruments, and optimized sample preparation and sequencing protocols for H3.3K27M mutation detection and quantification. We observed 100% sensitivity and specificity for mutation detection in matched DMG tissue and CSF across assays, platforms and institutions. ctDNA is reliably and reproducibly detected in the liquid biome using ddPCR, representing a clinically feasible, reproducible, and minimally invasive approach for DMG diagnosis, molecular subtyping and therapeutic monitoring.

Mesenchymal Stem Cells Successfully Deliver Oncolytic Virotherapy to Diffuse Intrinsic Pontine Glioma.

PURPOSE: Diffuse intrinsic pontine glioma (DIPG) is among the deadliest of pediatric brain tumors. Radiotherapy is the standard-of-care treatment for DIPG, but offers only transient relief of symptoms for patients with DIPG without providing significant survival benefit. Oncolytic virotherapy is an anticancer treatment that has been investigated for treating various types of brain tumors. EXPERIMENTAL DESIGN: Here, we have explored the use of mesenchymal stem cells (MSC) for oncolytic virus (OV) delivery and evaluated treatment efficacy using preclinical models of DIPG. The survivin promoter drives the conditional replication of OV used in our studies. The efficiency of OV entry into the cells is mediated by fiber modification with seven lysine residues (CRAd.S.pK7). Patients' samples and cell lines were analyzed for the expression of viral entry proteins and survivin. The ability of MSCs to deliver OV to DIPG was studied in the context of a low dose of irradiation. RESULTS: Our results show that DIPG cells and tumors exhibit robust expression of cell surface proteins and survivin that enable efficient OV entry and replication in DIPG cells. MSCs loaded with OV disseminate within a tumor and release OV throughout the DIPG brainstem xenografts in mice. Administration of OV-loaded MSCs with radiotherapy to mice bearing brainstem DIPG xenografts results in more prolonged survival relative to that conferred by either therapy alone (P < 0.01). CONCLUSIONS: Our study supports OV, CRAd.S.pK7, encapsulated within MSCs as a therapeutic strategy that merits further investigation and potential translation for DIPG treatment.

Addition of Multimodal Immunotherapy to Combination Treatment Strategies for Children with DIPG: A Single Institution Experience.

Background: The prognosis of children with diffuse intrinsic pontine glioma (DIPG) remains dismal despite radio- and chemotherapy or molecular-targeted therapy. Immunotherapy is a powerful and promising approach for improving the overall survival (OS) of children with DIPG. Methods: A retrospective analysis for feasibility, immune responsiveness, and OS was performed on 41 children treated in compassionate use with multimodal therapy consisting of Newcastle disease virus, hyperthermia, and autologous dendritic cell vaccines as part of an individualized combinatorial treatment approach for DIPG patients. Results: Patients were treated at diagnosis (n = 28) or at the time of progression (n = 13). In the case of 16 patients, histone H3K27M mutation was confirmed by analysis of biopsy (n = 9) or liquid biopsy (n = 9) specimens. PDL1 mRNA expression was detected in circulating tumor cells of ten patients at diagnosis. Multimodal immunotherapy was feasible as scheduled, until progression, in all patients without major toxicity. When immunotherapy was part of primary treatment, median PFS and OS were 8.4 m and 14.4 m from the time of diagnosis, respectively, with a 2-year OS of 10.7%. When immunotherapy was given at the time of progression, median PFS and OS were 6.5 m and 9.1 m, respectively. A longer OS was associated with a Th1 shift and rise in PanTum Detect test scores. Conclusions: Multimodal immunotherapy is feasible without major toxicity, and warrants further investigation as part of a combinatorial treatment approach for children diagnosed with DIPG.

Mass cytometry detects H3.3K27M-specific vaccine responses in diffuse midline glioma.

BACKGROUNDPatients with diffuse midline gliomas (DMGs), including diffuse intrinsic pontine glioma (DIPG), have dismal outcomes. We previously described the H3.3K27M mutation as a shared neoantigen in HLA-A*02.01+, H3.3K27M+ DMGs. Within the Pacific Pediatric Neuro-Oncology Consortium, we assessed the safety and efficacy of an H3.3K27M-targeted peptide vaccine.METHODSNewly diagnosed patients, aged 3-21 years, with HLA-A*02.01+ and H3.3K27M+ status were enrolled in stratum A (DIPG) or stratum B (nonpontine DMG). Vaccine was administered in combination with polyinosinic-polycytidylic acid-poly-I-lysine carboxymethylcellulose (poly-ICLC) every 3 weeks for 8 cycles, followed by once every 6 weeks. Immunomonitoring and imaging were performed every 3 months. Imaging was centrally reviewed. Immunological responses were assessed in PBMCs using mass cytometry.RESULTSA total of 19 patients were enrolled in stratum A (median age,11 years) and 10 in stratum B (median age, 13 years). There were no grade-4 treatment-related adverse events (TRAEs). Injection site reaction was the most commonly reported TRAE. Overall survival (OS) at 12 months was 40% (95% CI, 22%-73%) for patients in stratum A and 39% (95% CI, 16%-93%) for patients in stratum B. The median OS was 16.1 months for patients who had an expansion of H3.3K27M-reactive CD8+ T cells compared with 9.8 months for their counterparts (P = 0.05). Patients with DIPG with below-median baseline levels of myeloid-derived suppressor cells had prolonged OS compared with their counterparts (P < 0.01). Immediate pretreatment dexamethasone administration was inversely associated with H3.3K27M-reactive CD8+ T cell responses.CONCLUSIONAdministration of the H3.3K27M-specific vaccine was well tolerated. Patients with H3.3K27M-specific CD8+ immunological responses demonstrated prolonged OS compared with nonresponders.TRIAL REGISTRATIONClinicalTrials.gov NCT02960230.FUNDINGThe V Foundation, the Pacific Pediatric Neuro-Oncology Consortium Foundation, the Pediatric Brain Tumor Foundation, the Mithil Prasad Foundation, the MCJ Amelior Foundation, the Anne and Jason Farber Foundation, Will Power Research Fund Inc., the Isabella Kerr Molina Foundation, the Parker Institute for Cancer Immunotherapy, and the National Institute of Neurological Disorders and Stroke (NINDS), NIH (R35NS105068).

Harmonization of postmortem donations for pediatric brain tumors and molecular characterization of diffuse midline gliomas.

Children diagnosed with brain tumors have the lowest overall survival of all pediatric cancers. Recent molecular studies have resulted in the discovery of recurrent driver mutations in many pediatric brain tumors. However, despite these molecular advances, the clinical outcomes of high grade tumors, including H3K27M diffuse midline glioma (H3K27M DMG), remain poor. To address the paucity of tissue for biological studies, we have established a comprehensive protocol for the coordination and processing of donated specimens at postmortem. Since 2010, 60 postmortem pediatric brain tumor donations from 26 institutions were coordinated and collected. Patient derived xenograft models and cell cultures were successfully created (76% and 44% of attempts respectively), irrespective of postmortem processing time. Histological analysis of mid-sagittal whole brain sections revealed evidence of treatment response, immune cell infiltration and the migratory path of infiltrating H3K27M DMG cells into other midline structures and cerebral lobes. Sequencing of primary and disseminated tumors confirmed the presence of oncogenic driver mutations and their obligate partners. Our findings highlight the importance of postmortem tissue donations as an invaluable resource to accelerate research, potentially leading to improved outcomes for children with aggressive brain tumors.

Detection and Monitoring of Tumor Associated Circulating DNA in Patient Biofluids.

Complications associated with upfront and repeat surgical tissue sampling present the need for minimally invasive platforms capable of molecular sub-classification and temporal monitoring of tumor response to therapy. Here, we describe our dPCR-based method for the detection of tumor somatic mutations in cell free DNA (cfDNA), readily available in patient biofluids. Although limited in the number of mutations that can be tested for in each assay, this method provides a high level of sensitivity and specificity. Monitoring of mutation abundance, as calculated by MAF, allows for the evaluation of tumor response to therapy, thereby providing a much-needed supplement to radiographic imaging.

Mitochondrial Dysfunction Leads to Cortical Under-Connectivity and Cognitive Impairment.

Under-connectivity between cerebral cortical association areas may underlie cognitive deficits in neurodevelopmental disorders, including the 22q11.2 deletion syndrome (22q11DS). Using the LgDel 22q11DS mouse model, we assessed cellular, molecular, and developmental origins of under-connectivity and its consequences for cognitive function. Diminished 22q11 gene dosage reduces long-distance projections, limits axon and dendrite growth, and disrupts mitochondrial and synaptic integrity in layer 2/3 but not 5/6 projection neurons (PNs). Diminished dosage of Txnrd2, a 22q11 gene essential for reactive oxygen species catabolism in brain mitochondria, recapitulates these deficits in WT layer 2/3 PNs; Txnrd2 re-expression in LgDel layer 2/3 PNs rescues them. Anti-oxidants reverse LgDel- or Txnrd2-related layer 2/3 mitochondrial, circuit, and cognitive deficits. Accordingly, Txnrd2-mediated oxidative stress reduces layer 2/3 connectivity and impairs cognition in the context of 22q11 deletion. Anti-oxidant restoration of mitochondrial integrity, cortical connectivity, and cognitive behavior defines oxidative stress as a therapeutic target in neurodevelopmental disorders.

Liquid biopsy for pediatric central nervous system tumors.

Central nervous system (CNS) tumors are the most common solid tumors in children, and the leading cause of cancer-related death. Over the past decade, molecular profiling has been incorporated into treatment for pediatric CNS tumors, allowing for a more personalized approach to therapy. Through the identification of tumor-specific changes, it is now possible to diagnose, assign a prognostic subgroup, and develop targeted chemotherapeutic treatment plans for many cancer types. The successful incorporation of informative liquid biopsies, where the liquid biome is interrogated for tumor-associated molecular clues, has the potential to greatly complement the precision-based approach to treatment, and ultimately, to improve clinical outcomes for children with CNS tumors. In this article, the current application of liquid biopsy in cancer therapy will be reviewed, as will its potential for the diagnosis and therapeutic monitoring of pediatric CNS tumors.