ID #266 Matched primary-relapse Diffuse Midline Glioma (DMG) patient-derived xenograft models to identify therapy resistance profiles

Abstract

A major challenge in cancer care is predicting a patient’s risk of relapse after primary treatment. Diffuse Midline Gliomas (DMGs) are intrinsically resistant to standard chemoradiotherapies, and high-throughput genomic profiling of primary and relapse tumors integrated with functional precision medicine are used here to target relapse-initiating cells early. We have established patient-derived xenografts (PDXs) from matched primary–relapse pairs of pediatric diffuse high-grade glioma to study tumor evolution within individual patients using integrated analyses of patient tissue, PDXs, three-dimensional assembloids, and xenograft-derived cell models (XDCLs).

Tumoral single cell suspensions from matched primary-relapse pairs were injected orthotopically in immunodeficient NOG mice. We established in vitro XDCLs in stem cell conditions from the first generation of animals transplanted. Co-culture of these cells in 3D assembloids were performed using 180-250 days old cortical organoids. Single-cell RNA sequencing and bulk RNA sequencing (RNA-seq) on patient biopsies, their corresponding PDXs from first generation of mice, and XDCLs, was performed to track potential changes in vivo and in vitro and to further validate primary-relapse specific evolution.

Multi-omics profiling across models revealed conserved yet distinct transcriptional programs between primary and relapse, with identification of persistent ACVR1 mutation in both primary and relapse. Phenotypic characterizations across different models through density mapping and invasion analyses confirmed proliferative-driven growth mode in relapse tumor models. Relapse assembloid cultures displayed reduced sensitivity to radiation treatment which highlighted the emergence of resistance in the relapse induced through increased mTOR signaling and reduced TP53 pathway.

DMG primary-relapse in vitro XDCLs / 3D assembloid models together with PDXs is a robust approach to study different aspects of relapse-specific tumor biology. Our findings reveal a phenotypic transition from an infiltrative, astrocytic-like, low-proliferative state in primary tumor to a bulky, high proliferative, OPC-like, TP53 mutated, mTOR-associated growth state in relapse. While certain PI3K/mTOR inhibitors showcased attenuated growth in relapse derived models, incomplete responses suggested compensatory signaling pathways, motivating our ongoing evaluation on combination strategies targeting both mTOR and ACVR1 pathways. Our work underlines the importance of modelling disease evolution, showcasing distinct therapeutic vulnerabilities in relapse compared to primary counterparts.