Project Summary/Abstract In preliminary studies, my colleagues and I used a single cell RNA sequencing to profile gene expression within individual cells of pediatric brain tumors. My single cell profiles provided the first direct evidence of a concept long suspected but never directly tested – namely that pediatric brain cancers arise from relatively uncommitted neural progenitors that become developmentally stalled in a replication-competent state. That being said, my preliminary observations have a central limitation – namely that scRNAseq destroys information about spatial relationships: who is next to whom within the tumor and within the brain? Why should we care about nearest neighbor relationships between tumor cells and the cellular components of normal brain? The answer is embedded within an emerging body of data highlighting functional interactions between tumor and tumor microenvironment. Within brain tumors, cell division, fate choice decisions, development and survival are regulated by short range interactions between presenter and receiver cell types. All of these informative cell:cell interactions are erased by conventional genomic profiling methods including scRNAseq. I propose here that combinatorial transcriptional codes identified by scRNAseq can localize cancer stem cells and cancer cell subtypes in 3D-space within the brain where conventional immunochemical markers fail. To test this proposal, I will (i) generate combinatorial transcriptional codes unique to the various cancer cells and tumor-associated normal cells from my existing single cell dataset; (ii) design specific primers for these combinatorial identifiers and; (iii) visualize cell:cell contacts by performing in situ sequencing on patient-derived tumor samples and on normal brain tissue samples. I will focus on pediatric gliomas, the most common tumor type of childhood. My study plan has two question-oriented, hypothesis-driven aims and a third goal-oriented aim (with a useful deliverable). Aim one asks why some cells within high grade gliomas exit the replication competent state to differentiate while others remain replication competent. Is the developmental cue intrinsic (i.e. a developmental “clock”) or extrinsic (e.g. a short range ligand:receptor interaction)? Aim two asks why some gliomas invade and others do not. I will use methodology summarized above to test the hypothesis that malignant or benign phenotypes are an outcome of differential “addictions” to local mitogenic signals from normal neural cell types. Aim three is to expand my data sets on glioma into a tumor-wide single cell atlas of normal cell contacts with most major forms of pediatric brain cancer.

Project Narrative Brain tumors have surpassed the leukemias as the leading cause of cancer-related death in children. Therapeutic options for pediatric brain cancers have shown little expansion over the past few decades and the prognosis for many tumor types (especially high-grade gliomas) remains bleak. The research proposed here will identify tumor-associated contacts with the brain microenvironment that guide cellular state transitions, account for tumor grade and can be weaponized to therapeutic ends.