Abstract Supratentorial ependymomas (ST-EPN) are aggressive childhood brain cancers that retain features of neurodevelopmental cell types and segregate into molecularly distinct subgroups. The developmental signatures of malignant cell states and microenvironmental factors underlying aberrant cellular behavior across all subgroups are unknown. In this study, we integrated single cell- and spatial transcriptomics as well as live-cell imaging to define ST-EPN cell states, tumor microenvironment and dynamic behavior of malignant cells within the neural microenvironment. We profiled 44 ST-EPN patient tumors encompassing ZFTA-RELA (n=20), ZFTA-Clusters 1 to 4 (n=20), and ST-YAP1 (n=4) subgroups by single nuclei/cell RNA sequencing (sn/scRNA-seq) and identified cell states that are both cancer generic process related (cycling and mesenchymal/hypoxia) as well as human early brain developmental signatures. ST-EPN subgroups displayed distinct signatures, mapping to two temporally restricted progenitors of neuroepithelial-like and embryonic-like cells. Additionally, tumor subgroups showed diverging patterns of neuronal or ependymal differentiation. We further uncovered distinct modes of spatial organization in these tumors by single cell spatial transcriptomics (10X Xenium), including both small local environments that usually were enriched in one major cell state, and global patterns of tissue architecture influenced by mesenchymal and hypoxia signatures. Notably, recurrent tumors exhibited higher proportion of mesenchymal/hypoxia cells with greater global spatial structure than primary tumors. Finally, by superimposing molecular and functional assays in both in vitro and in vivo models, we found that cell states exhibit distinct morphology and cell behavior. Neuronal-like cells especially were the most invasive cell state, with distinct migratory patterns reminiscent of immature neurons during development. Moreover, we demonstrated the crucial role of neural microenvironment in promoting plasticity of malignant cells towards this neuronal lineage. Taken together, we provide a multidimensional framework to integrate transcriptional and phenotypic characterization of tumor heterogeneity in ST-EPN tumors and its potential clinical implications.
Jeong et al. (Fri,) studied this question.