Applications of spatial transcriptomics for the diagnostics of gliomas

Abstract Due to their infiltrative growth, diffuse gliomas can pose a major challenge for neuropathologists. Lack of specific immunohistochemistry markers for some gliomas like IDH-wildtype glioblastoma can make it hard to identify the tumour cell content or distinguish between tumour and reactive gliosis. Other hurdles are high fractions of DNA from healthy cells in molecular diagnostic analysis, or scarce tissue, which can result in a failed DNA extraction. To overcome these limitations, we evaluated the application of spatial transcriptomics on 14 glioblastoma bulk resections and 18 stereotactic biopsies of 7 glioblastoma, IDH wildtype, 4 oligodendroglioma IDH-mutant and 7 astrocytoma IDH-mutant samples with the aim to generate a diagnostic report containing all information needed by the neuropathologist and clinician. Using the SPATA2 R package, we were able to infer the copy number variation (CNV) profiles for each spot. When comparing the inferred CNVs to the CNVs calculated by the 850K methylome analysis, a correlation between CNVs identified from the 850K data and the inferred CNVs could be seen. For diagnostic relevant CNVs of chr. 7, 10, 1p and 19q, the inferred CNVs were concordant with the 850K data in 100% of the cases with confirmed tumour tissue. CNVs from spatial transcriptomics were able to reveal sub-clones, which could not be detected by the 850K array data. The CNVs could then be used in combination with a deconvolution approach to determine the tumour burden of the tissue. Using RNA expression, the immunohistochemistry staining of important diagnostic markers like NeuN, GFAP and Ki67 was predicted. In addition, the spatial gene expression could be used to identify the glioblastoma subtype based on the gene expression of different methylation subtypes. All the above-mentioned analysis also generated meaningful results for tissue fragments smaller than 1mm in diameter, which are too small for molecular diagnostic analysis.