Cross-species multi-omic analysis of intra-tumoral heterogeneity and cell state transition in Glioblastoma

Abstract

Glioblastoma is a malignant brain tumor characterized by intra-tumoral heterogeneity and plasticity, and shows treatment resistance through cell state transition. In this study, I performed a comprehensive analysis of cross-species evolutionarily conserved tumor progression and cell state dynamics using a genetically engineered mouse glioblastoma model to recapitulate the features of human glioblastoma. Due to the characteristics of mouse glioblastoma, which mimics neurodevelopmental hierarchy, it is difficult to distinguish normal cells from cancer cells using transcriptional profiling alone. And unlike humans, criteria for discrimination of cancer cells is not established. Aberrant RNA splicing process is known to be one of the characteristics of cancer cells. Based on the temporal dynamics of the transcriptional process, the integration of unspliced RNA and spliced RNA was used to distinguish cells in a disease state from cells in a normal state. This analysis method increases the resolution of clustering by providing multimodal information to existing single cell transcriptomic data without additional experiments. To compare human and mouse glioblastoma, integrative cross-species analysis is required. Based on the evolutionarily conserved orthologous gene information between human and mouse, an autoencoder based framework was used to achieve appropriate cross-species integration of unpaired transcriptomic data. Through this, I was able to confirm the expression pattern of orthologous genes in cell states shared between different species. Through non-negative matrix decomposition techniques, biological programs commonly expressed in human and mouse glioblastoma are identified. And the 'gliogenesis' program is responsible for tumorigenesis in the early stages of glioblastoma and conserved in both humans and mice. The ‘gliogenesis’ program involves cell signaling receptors and several transcription factors. Through Markov chain-based analysis of cell state dynamics, several transcription factors were identified as common drivers of cell state transitions in humans and mice. Precursor cell-like glioblastoma mainly expresses transcription factors related to ‘gliogenesis’, and it was confirmed that the expression of transcription factors expressed during the development of the nervous system was observed to change the cell state in various forms. Ascl1, together with Olig1 and Olig2, was found to promote early tumor formation and cancer cell proliferation. Through cell signaling analysis of single-cell transcriptome data and cell signaling analysis of spatial transcriptomic data, it was confirmed that signaling pathways involved in ‘gliogenesis’ program is biologically active. Cell signaling analysis predicted Ascl1 as an intracellular target of the corresponding receptors. In tumor-sphere assy, an increase in Ascl1 expression was confirmed along with an increase in the number of colonies upon Pdgf-aa stimulation in vitro. These findings suggest that transcription factors conserved across species are regulators of cell state transition, which leads to intra-tumoral heterogeneity and treatment resistance. The discovery of common drivers for cell state transition in both human and mice highlights a potential for developing therapies that suppress tumorigenesis.