Preview

PDF, English Download (12MB) | Terms of use

Abstract

Tumors are complex tissues with substantial intra-tumor heterogeneity, intricately linked to tumor progression and therapeutic resistance. Emerging single-cell RNA sequencing (scRNA-Seq) technologies empower researchers to elucidate these diverse tumor subpopulations. This thesis presents the characterization of a distinct glioblastoma (GB) cell population and introduces a novel bioinformatics tool designed to quantify similarity among cell populations.

Cell-to-cell connectivity through tumor microtubes (TMs) has been discovered among glioma tumor cells, conferring self-repair capabilities, augmenting therapy resistance, and driving tumor progression. Yet, a comprehensive molecular understanding and precise quantification of this connectivity have remained elusive. This study delves into the transcriptomic landscape of the highly connected glioma cell population using scRNA-Seq and RNA-Seq. I found that these highly connected cells exhibited a notable predominance of astrocyte-like (AC) and mesenchymal-like (MES) cell states, while lowly connected cells were characterized by a prevalence of neuronal progenitor-like (NPC) cell states. I established a 71-gene connectivity signature by comparing highly and lowly connected cells. A connectivity signature score (CSS) was developed based on the relative average expression levels of the connectivity signature. This CSS was then applied to several GB patient tumor scRNA-Seq and RNA-Seq datasets, consistently revealing higher CSS values for AC and MES cell states compared to NPC cell states. Furthermore, correlations were observed between CSS values and mesenchymal expression subtypes as well as between CSS values and the mutation status of NF1, PTEN, and TP53. One key finding is that higher CSS values were linked to poorer patient survival. Additionally, CHI3L1 — one of the connectivity signature genes — was identified as a robust marker for cell connectivity and a potential prognostic marker for GB patients. Investigating CHI3L1 overexpression RNA-Seq and proteomics datasets revealed that CHI3L1 upregulated multiple cell state markers and elevated CSS values. Notably, CHI3L1 overexpression also led to increased phosphorylation of the TM-connectivity marker GAP43.

In this thesis, I present a new bioinformatic tool named Interactive Explorer of Single-Cell Cluster Similarity (ieCS). This tool serves to link similar cell populations that share the same biological cell types/states across various donors or experimental conditions. ieCS utilizes an innovative metric to quantify similarity between cell populations. ieCS offers three distinct methods for identifying superclusters comprising similar cell populations. Featuring a user-friendly graphical interface, ieCS enables interactive and intuitive visualization of these superclusters. In a demonstration dataset, ieCS accurately, robustly, and quickly identified superclusters across various experimental conditions.

In conclusion, this thesis characterizes the highly connected GB cell population and introduces a bioinformatics tool for mapping similar cell populations.