P12.12.A Resolving the cellular architecture of infiltration zones in glioblastoma

Abstract

Abstract Background Therapy resistance and infiltration still pose major challenges in the treatment of glioma. In the tumour border niche, an interaction between different healthy cell types and malignant cells leads to therapy resistance, acquisition of stem-cell like features, and recurrence. However, studying the tumour border is quite challenging due to the lack of specific glioma markers. Although single cell datasets contain information about the abundance of different cell types, they still lack the spatial arrangement at the border niche. Spatial transcriptomic approaches might overcome these problems, but are still limited by a low resolution or a limited number of detectable transcripts. Material and Methods We applied the Visium spatial transcriptomics platform on 18 FFPE and 11 fresh frozen gliomas to image the cellular architecture of the tumour border niche. In addition, the transcriptome of single nuclei isolated from the fresh frozen sections was also analysed. Using Seurat and a single cell reference set from Darmanis et al. (2017) the cell types were mapped onto the section and the copy number variation profile (CNV) was determined. Results Using a single cell reference set we were able to map different cell types onto different areas of the tumour border niche of gliomas. To control the mapping the expression of cellular markers was used to detect healthy brain cells and the CNV profile was used to identify the tumour. The mapping matched the expression of markers and the CNV profile and only showed minor differences. First analysis confirmed distinct distribution of oligodendrocytes and neurons at the tumour border. We could not detect an enrichment of any tumour subtypes at the tumour border so far. However, one sample showed an enrichment of the mesenchymal subtype at the perinecrotic region. CNVs also showed low intratumoural heterogeneity for most sections. Conclusion Although the first results of the mapping of different cell types look promising, a larger reference set including a larger variety of cell types and tumour subtypes could improve the mapping even further. Nevertheless, we were able to show the cell type distribution at the tumour border and generate robust CNVs for most sections using spatial transcriptomics.