OS10.8.A APPLICATIONS OF NOVEL FFPE BASED TECHNOLOGIES FOR THE DIAGNOSTICS OF GLIOMAS

Abstract

Abstract BACKGROUND Due to the lack of consistent tumour-cell specific markers, the diffuse brain invasion of glioblastoma (GB) presents a significant diagnostic challenge, especially for specimens with a low tumour cell fraction or scarce tissue. The only common alteration found in most GB is the gain of chromosome 7 and loss of chromosome 10. The emergence of new technologies such as spatial and single nucleus transcriptomics that allow for detection of copy number alterations (CNVs) may allow us to push the diagnostic boundaries. MATERIAL AND METHODS We performed 10x Visium spatial transcriptomic analysis of FFPE tissue from bulk resections of 14 FFPE GB and stereotactic biopsies of 6 GB, 7 IDH mutant astrocytomas, 4 oligodendrogliomas and 1 diffuse glioma NOS. Visium tissue spots were then deconvolved using data from single nucleus transcriptomic data from nuclei isolated from FFPE tissue. Fresh frozen and FFPE single nucleus transcriptomic libraries were also generated from 4 GB with matching fresh-frozen and FFPE tissue available. RESULTS The infiltration zone, the tumour core and healthy brain tissue could be differentiated using the inferred CNVs of chromosome 7 and chromosome 10. Additionally, inferred CNVs of tissue fragments smaller than 1 mm2 were in accordance with Infinium MethylationEPIC array analysis (a well-established method for CNV detection) from the same tumour tissue. Using the spatial transcriptomics approach, the loss of chromosome 1p and 19q, which is characteristic for oligodendroglioma was also detected. To overcome the low resolution of Visium, deconvolution approaches using a single cell dataset can be used to determine the tumour content of a given tissue more precisely. To test if the single nuclei data from FFPE samples is suitable for deconvolution of spatial transcriptomics, we compared fresh-frozen and FFPE single nuclei data from matching tumours. As expected, FFPE and fresh frozen nuclei clustered together according to cell types and tumours in the UMAP space. CONCLUSION Spatial transcriptomics can be useful for brain tumour diagnostic workup, especially in cases where a limited amount of tissue prevents analysis by standard molecular diagnostic METHODS . Additionally, CNVs inferred from the spatial data as well as the spatial expression of marker genes help to better distinguish tumour infiltration zones from the core tumour and healthy brain tissue in the case of GB. This analysis can be further supported by deconvolution with a reference set generated from single nucleus transcriptomic analysis of FFPE tissue.