Modeling the unmet need in rare pediatric brain tumors: Preclinical models for ependymoma and ETMR

This thesis investigates preclinical models for pediatric brain tumors, focusing on ependymoma (EPN) and embryonal tumor with multilayered rosettes (ETMR), both thought to originate from radial glial cells. We developed and characterized organoids, tumoroids, and patient-derived xenograft (PDX) models to study tumor biology, tumor–microenvironment interactions, and to evaluate potential therapies. The goal was to provide robust models for preclinical research and to identify compounds with therapeutic promise.

Chapter 2 explores oncogenic drivers, particularly fusion proteins, in pediatric central nervous system (CNS) tumors. We compiled a comprehensive overview of reported fusions, identifying two main classes: (1) kinase fusions, often involving MAPK pathway proteins (e.g., NTRK, RAF, FGFR), which typically drive pediatric low-grade gliomas; and (2) fusions with transcriptional regulators, which have helped refine tumor classification beyond histological definitions. Some fusions occur across multiple tumor types or with varied partners. Better understanding the mechanisms of these fusions can provide opportunities for better suited therapies.

In Chapter 3, we modeled supratentorial ependymoma (ST-EPN) subtype-specific tumors using cerebral organoids. Introducing oncogenic fusions into these organoids yielded tumors closely resembling ST-EPN patient tumors, confirmed by histology, transcriptome and single-cell analyses. ZFTA-fusion tumors retained radial glia-like properties with neuronal differentiation tendencies, while YAP1-fusion tumors showed stronger ependymal lineage differentiation. These lineage trajectories mirrored patient tumor data. Fusion-specific effects on the microenvironment were also observed: YAP1-fusion tumors promoted neuronal proliferation, whereas ZFTA-fusion tumors induced a more tumor-supportive gene expression. Given these signatures Dasatinib was deemed a suitable candidate therapy for ZFTA driven tumors.

Chapter 4 describes patient-derived tumoroid models of EPN. We generated four PF-EPN-A and three ST-EPN-ZFTA tumoroids, which faithfully recapitulated patient and subtype specific features. A drug screen of over 200 compounds identified histone deacetylase inhibitors, HSP90 inhibitors, and topoisomerase II inhibitors as effective against ST-EPN-ZFTA models. These findings support the use of patient-derived tumoroids as preclinical avatars for therapeutic discovery.

In Chapter 5, we applied similar approaches to ETMR, a rare and aggressive pediatric tumor. Tumoroids were derived from both patient material and PDXs. These closely resembled patient tumors, though they primarily captured the embryonal component. Comparisons confirmed their distinction from other embryonal tumors, such as ATRT and medulloblastoma. Compound screening showed resistance to standard chemotherapies, consistent with clinical experience. However, targeted agents including XPO1 and MDM2 inhibitors displayed some efficacy. Notably, SN-38 and its pegylated form PLX-038 produced strong responses in vitro and in vivo, suggesting novel therapeutic avenues for ETMR.

Finally, Chapter 6 summarizes these findings, emphasizing the value of organoid, tumoroid, and PDX models in advancing understanding of pediatric brain tumor biology. These models not only illuminate tumor biology and tumor–microenvironment dynamics but also serve as powerful platforms for identifying and validating new treatment strategies.