PO-390 Image-guided stereotactic molecular profiling of samples taken in- and outside conventional tumour boundaries highlight extensive infiltration of tumour cells in diffuse gliomas

Abstract

IntroductionDiffuse gliomas are highly infiltrative tumours of the central nervous system that lack effective treatment options and suffer from poor survival. We and others have used DNA methylation and gene expression profiling to stratify these tumours into clinically relevant subtypes. Nevertheless, diffuse gliomas are notoriously heterogeneous and a single biopsy is unlikely to reflect all tumour cell populations. Advanced sampling schemes guided by imaging are needed for accurate delineation of heterogeneity.Material and methodsWe obtained 74 stereotactic image-guided biopsies in eight patients with glioma prior to craniotomy in addition to 102 multi-sector glioma samples from 25 patients from outside sources. Patients underwent extensive imaging including standard magnetic resonance imaging and various advanced modalities. For each patient, multiple biopsies were taken from the tumour core, tumour margin and outside of visible imaging abnormalities and each sample was histologically assessed by two board-certified pathologists. We performed methylation profiling on all samples. We used supervised machine learning to determine methylation subtype composition and predict transcriptional subtypes in each sample. Genome wide DNA copy number was inferred.Results and discussionsPrincipal component analysis of DNA methylation separated samples based on IDH status and tumour cell content. Tumour signal could be detected in samples far outside tumour margins on standard MRI, despite demonstrating lower tumour purity and inconclusive tumour subtypes. Samples taken from inside imaging abnormalities demonstrated concordant methylation subtypes, suggesting that this classifier is reliable and not subject to heterogeneity due to microenvironment cell populations as often the case with gene expression subtyping. Despite uniformity in classification, IDH mutant tumours demonstrated a higher degree of epigenetic heterogeneity compared to wild-type tumours, reflecting hypermethylation established by IDH oncometabolites. Although still imperfect, advanced imaging substantially improved tumour delineation. While samples from inside imaging abnormalities are optimal for diagnostic use, tumour cells infiltrate well beyond these boundaries and treatment of these regions should be considered when safe and feasible.ConclusionThese data show methylation profiling is stable across multiple biopsies and could serve as a powerful diagnostic biomarker. Moreover, advanced imaging can more accurately delineate tumour margins and could help direct treatment.