Epen-11. characterization of the spatial architecture of intratumoral cell states in pediatric high-risk ependymoma

Abstract By single-cell analyses, we previously uncovered distinct intratumoral transcriptomic ependymoma (EPN) cell states and developmental trajectories correlating with molecular subtype and biological behavior. However, the relation of these cell states to characteristic EPN tissue architecture and tumor microenvironment remains widely unknown. We aimed to characterize the spatial architecture and elucidate regulatory circuits in EPN via spatial transcriptomics (10X Visium) of tumor specimens matching our previously generated single-cell transcriptome data spanning all EPN subtypes (ZFTA, YAP1, PFA, PFB). We investigated whether spatially restricted cell-cell interactions correlate with our previously identified cell states. Analyses revealed distinct spatially resolved cell populations in supratentorial and posterior fossa EPN. All subgroups showed several (5-9) spatially discrete cell clusters. Via non-negative matrix factorization (NMF), cell-specific spatial clusters emerged correlating with previous single-cell programs. ZFTA EPN highly correlates with ZFTA-specific metaprograms whereas all other subgroups did not show ZFTA-specific gene expression. Explicit genes such as DLK1 are exclusively upregulated in ZFTA EPN indicating spatially distinct aberrant Notch pathway signaling. Interestingly, this subgroup revealed spatial zones with upregulation of metabolic genes and endothelial markers anti-correlating with a ZFTA-specific program. Ependymal and neuronal-precursor-like signatures were enriched in our unique YAP1 EPN sample in a spatially restricted pattern, whereas GFAP was ubiquitously expressed. PFA was characterized by a neural stem cell (NSC)-like niche accompanied by metabolic and immune-reactive gene signatures. PFB showed a similar cell niche associated with NSC-like and glial-progenitor-like genes enriched simultaneously, further highlighting the NSC characteristic of posterior fossa tumors. Ependymal signatures were absent in PFA and present in PFB indicating more differentiated cell populations in PFB. Summarizing, we confirm presence of distinct intratumoral cell states within ependymoma tissue and reveal their spatial organization. In depth analysis and functional validation in matched ependymoma models will enable interrogation of underlying tumor promoting signaling events driving intratumoral developmental trajectories.