Abstract General Information

Title

STEMNESS SIGNATURE STRATIFIES GLIOMAS BASED ON AGGRESSIVENESS, HISTOPATHOLOGIC AND MOLECULAR FEATURES

Introduction, Objectives, Methods, Results, and Conclusion.

Gliomas are the most common form of central nervous system (CNS) neoplasm, accounting for nearly 80% of all malignant primary brain tumors and the cellular origins of these tumors are still unknown. For a long time, the classification, diagnosis and treatment of gliomas were based on histological characteristics, following the classification of the World Health Organization (WHO). In this work we propose a Stemness prediction model based on gene expression signatures using publicly available gene expression data from neural progenitor cells (fetal astrocyte-AST) that can be used to measure the dedifferentiation state (or Stemness) of glioma samples. To build the prediction model, publicly available single-cell RNA sequencing data was used to access and identify gene expression data from fetal astrocyte population. All data analysis was performed using the R software, where the samples obtained were normalized through the expression of control genes and later the subpopulations of interest were identified through the expression of marker genes. Then the one-class logistic regression (OCLR) machine learning algorithm was applied to build the prediction model using the cell populations previously identified. The model was applied to TCGA glioma data and the stemness index based on fetal astrocytes (ASTsi) was able to separate gliomas by their degree of malignancy (grade), by histology and by molecular subtypes. Grade 4 IDHwt tumors had the highest rates, where oligodendroglioma IDHmut had the lowest. When applied to gliomas longitudinal samples from the GLASS consortium, we observed ASTsi increase in IDHmut second recurrence tumors while it decreases in IDHwt recurrences, compared to primary gliomas. In addition, we observed in both the TCGA and GLASS primary tumors that the TCGA classical-like subtype were the ones with the highest ASTsi. When we analyzed the recently described functional subtypes of gliomas in primary tumors from the GLASS cohort, we observed highest ASTsi in the proliferative IDHmut and in the mithocondrial IDHwt tumors. The prediction model built using fetal astrocyte signatures was able to generate stemness indices that, when applied to glioma samples, stratified glioma samples by histopathology and molecular subtypes and showed correlation with cancer aggressiveness, thus proving to be a good metric to be added in the identification and diagnosis of patients.

Keywords (separated by comma on a single line)

Gliomas, Stemness, Gene expression, Single-cell RNAseq, Bioinformatics