Decoding Tumor Heterogeneity: A New Framework for Supratentorial Ependymoma

Supratentorial ependymomas (ST-EPNs) represent one of the most challenging childhood brain cancers, accounting for significant cancer-related mortality in pediatric populations. These aggressive tumors occur in the central nervous system and have historically shown varied outcomes with limited therapeutic options. Recent advances in molecular classification have revealed distinct subgroups with different fusion genes and clinical trajectories, yet the underlying cellular heterogeneity and developmental origins have remained poorly understood. A comprehensive study published in Nature now provides a multidimensional framework for profiling tumor heterogeneity in ST-EPNs, integrating cutting-edge technologies to reveal previously unrecognized cellular states, spatial organization patterns, and dynamic behaviors that drive tumor progression.

Molecular Subgroups and Developmental Signatures

ST-EPNs have been classified into multiple molecular subgroups based on genome-wide DNA methylation profiling, including the canonical ST-ZFTA subgroup characterized by ZFTA-RELA fusions and the ST-YAP1 subgroup with YAP1 fusions. The recent research reveals that these molecularly distinct subgroups are transcriptionally distinct and align with different developmental timepoints across human cortical development. According to the study, ZFTA cluster 3 and 4 tumors project to the earliest developmental windows, while ST-YAP1 tumors align with the latest timepoints. This correlation suggests divergent developmental patterns across subgroups, with each following distinct cellular trajectories that may explain their varying clinical behaviors and therapeutic responses.

Cellular States Mirror Early Brain Development

The study identified eight recurrent cellular states, or metaprograms, across ST-EPN tumors that reflect specific developmental programs. These include two progenitor-like states—neuroepithelial-like and embryonic-like cells—that are reminiscent of early human brain development. The neuroepithelial-like cells strongly project to neuroepithelial stem cells of the developing human cortex and are among the most proliferative populations. Embryonic-like cells express genes essential for embryo formation and patterning but don't match specific brain cell types. The remaining states include radial glial-like, embryonic-neuronal-like, neuronal-like, and ependymal-like populations, each with distinct gene expression profiles and functional characteristics.

Distinct Cellular Composition Across Subgroups

The proportion of these cellular states varies significantly across ST-EPN subgroups. ZFTA-RELA tumors are predominantly composed of neuroepithelial-like, neuronal-like, and ependymal-like cells, while ZFTA cluster 3 tumors display a distinct composition with mostly embryonic-like and embryonic-neuronal-like subpopulations. This is consistent with the undifferentiated histology of ZFTA cluster 3 tumors, which have often been diagnosed as sarcoma or other primitive tumors. ST-YAP1 tumors mainly exhibit ependymal-like signatures with very few progenitor cells present. These findings suggest that ST-EPN tumors from different molecular subgroups have distinct developmental signatures with variability in neuronal and ependymal lineages.

Spatial Organization Patterns in Tumors

Using 10x Genomics Xenium spatial transcriptomics on ZFTA-RELA tumors, researchers discovered distinct global organization patterns. Tumors segregated into structured tumors with high compartmentalization of cell states and disorganized tumors with cell states scattered throughout the tissue. The degree of spatial organization, quantified as the spatial coherence score, revealed that higher organization positively correlated with the presence of mesenchymal/hypoxia cells, while disorganization correlated with embryonic-neuronal-like cells. This suggests that hypoxia may drive tissue structural organization, a phenomenon also observed in adult gliomas.

Local Spatial Architecture

Beyond global patterns, the study identified local spatial structures within tumors. Using computational frameworks like CellCharter, researchers discovered 26 stable local spatial patterns across tumor sections. These included myeloid-enhanced regions, mesenchymal/hypoxia-enhanced clusters, and neuronal-like-enhanced areas. Some spatial clusters were shared across multiple samples, while others were sample-restricted, highlighting the heterogeneity of spatial arrangement within tumors. These local structures often corresponded to morphologically distinct regions, such as areas with high proportions of endothelial cells or ependymal rosettes.

Cellular Morphology and Migratory Behaviors

The research extended beyond transcriptomics to examine how cellular states manifest in morphology and behavior. In coculture models with brain-resident cells, ZFTA-RELA tumor cells exhibited distinct morphological subtypes: cells without processes, cells with 1-3 short primary processes, cells with more than 3 short processes, and cells with 1-3 long processes (longer than 100 μm). These morphologies correlated with specific molecular states, with neuronal-like cells displaying long tumor microtubes and adopting highly migratory behaviors reminiscent of immature neurons during development.

Migration Patterns and Proliferation

Live-cell imaging revealed that neuronal-like and neuroepithelial-like-2 cells were the most migratory populations, while neuroepithelial-like-1 and immature ependymal-like cells were more stationary. Neuronal-like cells frequently adopted saltatory migration patterns—intermittent bursts of rapid motion—similar to migrating immature neurons during development. Neuroepithelial-like-2 cells emerged as particularly significant, exhibiting both high migratory capacity and high proliferation rates. This dual functionality suggests these cells may drive both tumor expansion and migration to other brain regions, representing a critical population for therapeutic targeting.

Clinical Implications and Therapeutic Opportunities

The multidimensional profiling framework provides several avenues for advancing ST-EPN treatment. The distinct developmental signatures across subgroups lay groundwork for personalized therapeutic strategies tailored to specific tumor subtypes. The identification of neuroepithelial-like-2 cells as both migratory and proliferative highlights a potential therapeutic target that could simultaneously address tumor expansion and invasion. Additionally, the role of hypoxia in spatial organization suggests that targeting hypoxic microenvironments could disrupt tumor structure and potentially enhance treatment efficacy.

The integration of extensive transcriptomic, spatial, morphological, and cellular behavioral characterization at the single-cell level has identified signatures of ST-EPN tumors related to migration, proliferation, and plasticity mechanisms.

Future Research Directions

The study opens several important avenues for future investigation. There is an urgent need for comprehensive clinical studies comparing patient outcomes and therapeutic resistance across all defined molecular subgroups in ST-EPN tumors. The role of neuronal-like cells in tumor invasion warrants examination of tumor material from resection borders in patients. Additionally, direct investigation of hypoxia-driven tissue structural changes and their functional implications in ST-EPN tumor models could reveal new therapeutic targets. The methodological framework established in this study provides a blueprint for similar multidimensional profiling across other tumor types.

Conclusion

The multidimensional profiling of heterogeneity in supratentorial ependymomas represents a significant advancement in understanding these challenging pediatric brain cancers. By integrating single-cell and spatial transcriptomics with live-cell imaging, researchers have uncovered the complex cellular states, spatial organization, and dynamic behaviors that drive tumor progression. The identification of distinct developmental signatures across molecular subgroups, the discovery of neuroepithelial-like-2 cells with dual migratory and proliferative capacities, and the characterization of spatial organization patterns all contribute to a more comprehensive understanding of ST-EPN biology. This framework not only advances fundamental knowledge but also provides concrete targets for developing much-needed personalized therapeutic strategies for children affected by these aggressive brain tumors.