To identify tumor “cells of origin,” we created a single-cell RNA sequencing (scRNA-seq) atlas of the human cerebellum, spanning neurogenesis to adulthood. This atlas captures all major neuronal and glial types—Purkinje neurons, granule cells, Bergman glia—and maps their developmental trajectories. We compared cerebellar tumors—medulloblastoma, posterior fossa ependymoma, pilocytic astrocytoma, and radiation-induced glioma (RIG)—to normal populations. Each tumor aligned with a distinct cerebellar lineage, revealing tumor-specific biology (PMIDs: 38029793, 37534924).Using this reference, we identified transcriptomic features shared between tumors and their matched normal cells, as well as tumor-specific differences. MHCII gene upregulation in pilocytic astrocytomas aligned with oligodendrocyte lineage. RIGs displayed unique profiles, supporting their classification as distinct tumor types (PMID: 37534924). We initiated functional studies on selected targets, including testing whether MHCII-expressing gliomas present tumor-specific peptides for CAR-T therapy. While key pathways have been identified, follow-up validation is ongoing and beyond the proposal scope. To study tumor heterogeneity, we generated a single-cell multiomic atlas of Group 3/4 medulloblastoma—the most comprehensive to date. It revealed transcription factor–gene regulatory networks (TF-GRNs) driving intra- and inter-tumor variation. We validated TFs that can reprogram tumor identity in vitro and in vivo, enabling models of underrepresented subtypes like non-MYC–driven tumors. For tumor evolution, we analyzed oncogene-driven subsets within this atlas. Single-cell copy number analysis showed MYC and MYCN as late events, while early tumorigenesis likely begins with large chromosomal alterations. Whole-genome sequencing and modeling placed tumor origin in the first trimester, with clonal expansion near birth—matching predicted cells of origin. Spatial single-cell profiling linked subclonal tumor regions to microenvironmental features, highlighting heterogeneity’s impact on tumor structure. These insights, published in Nature (PMID pending), reshape our understanding of medulloblastoma development.
We also built a single-nucleus RNA-seq atlas of the human lower brainstem (pons and medulla), profiling 400,000 cells across development. Like the cerebellar atlas, it maps glial and neuronal trajectories and represents the most complete dataset of human hindbrain development to date. Additionally, we produced single-nucleus ATAC-seq atlases of mouse (90,000 cells) and human (110,000 cells) cerebellar development. Integrated with RNA-seq data, these reveal gene regulatory elements and programs guiding hindbrain differentiation. Together, these efforts provide a high-resolution framework for understanding brain development and pediatric brain tumors—from origins and regulatory control to evolution and heterogeneity—informing future diagnostics and therapies.
