59. Inferring copy number events in Glioblastoma using spatial gene expression data

Abstract

<h3>Background</h3> Glioblastoma (GBM) is the most common malignant brain tumour in adults, with inevitable treatment resistance and poor outcomes. Genomic instability of GBM gives rise to large-scale chromosomal copy number variants (CNVs) that drive resistance. Genetically distinct clones in GBM are conventionally investigated using whole-genome sequencing (WGS), but recent bioinformatic approaches can infer CNVs using transcriptomics data only. <h3>Purpose</h3> To investigate whether patient-derived xenograft (PDX) mouse models of GBM exhibit clonal diversity as a function of space (across tumour regions) and time (early to late time points). <h3>Methods</h3> PDX data from 3 cell lines representing the tumour core, vascular regions, and infiltrating edge of one GBM patient's tumour were profiled at early and late time points in disease progression, using the 10X Genomics Visium platform for spatial transcriptomics. The computational tool inferCNV was used to identify CNVs. <h3>Results</h3> Successfully identified large-scale CNVs were concordant with events detected in matched WGS data. Additionally, we identified a minor subclone not detectable by WGS that expands in multiple PDX mice indicating it is pre-existing in the patient-derived cell lines. Further analysis revealed localization of this subclone to the tumour core region suggesting a microenvironment-specific adaptation. <h3>Conclusions</h3> We show that spatial gene expression data can be leveraged to infer CNVs in PDX models of GBM that correspond to known events from bulk WGS data. Subclones present below the limit of WGS detection can also be inferred. Charting the spatially resolved position of these clones in vivo supports the use of PDX models in GBM.