Search

Melanoma-derived brain metastases (MBM) represent an unmet clinical need because central nervous system progression is frequently an end stage of the disease. Immune checkpoint inhibitors (ICI) provide a clinical opportunity against MBM; however, the MBM tumor microenvironment (TME) has not been fully elucidated in the context of ICI. To dissect unique elements of the MBM TME and correlates of MBM response to ICI, we collected 32 fresh MBM and performed single-cell RNA sequencing of the MBM TME and T-cell receptor clonotyping on T cells from MBM and matched blood and extracranial lesions. We observed myeloid phenotypic heterogeneity in the MBM TME, most notably multiple distinct neutrophil states, including an IL8-expressing population that correlated with malignant cell epithelial-to-mesenchymal transition. In addition, we observed significant relationships between intracranial T-cell phenotypes and the distribution of T-cell clonotypes intracranially and peripherally. We found that the phenotype, clonotype, and overall number of MBM-infiltrating T cells were associated with response to ICI, suggesting that ICI-responsive MBMs interact with peripheral blood in a manner similar to extracranial lesions. These data identify unique features of the MBM TME that may represent potential targets to improve clinical outcomes for patients with MBM.

Diffuse midline gliomas driven by lysine27-to-methionine mutations in histone 3 (H3-K27M DMGs) are among the most fatal brain tumors. Molecular studies including single cell RNA-sequencing (scRNA-seq) of pediatric and predominantly pontine H3-K27M DMGs have shown that the H3-K27M oncohistone keeps glioma cells locked in a stem-like oligodendrocyte precursor cell (OPC) state that is capable of self-renewal and tumor-initiation. However, a comprehensive dissection of the cellular architecture of H3-K27M DMGs across different midline regions and age groups is required to better understand the cell-intrinsic and contextual regulation of H3-K27M DMG cell identities. In particular, the more recently described group of adult H3-K27M DMGs remains understudied. Here, we have collected and characterized 45 H3-K27M mutant patient tumors, spanning pontine (n=26), thalamic (n=17), and spinal (n=2) locations. Median age at surgery was 12 (2-68) years, encompassing 21 early childhood (0-10 years), 12 adolescent (11-20 years), and 12 adult (≥ 21 years) tumors. The majority of samples were obtained pre-treatment (n=28), as opposed to post-treatment or at autopsy (n=17). We profiled all 45 tumors by single cell/single nucleus RNA-seq and selected tumors were further characterized by the single cell assay for transposase-accessible chromatin (scATAC-seq). Our integrated analyses highlight the predominance of transcriptionally and epigenetically defined OPC-like tumor cells as the main cell population of H3-K27M DMGs across all age groups and locations. We further identify distinct age- and location-specific OPC-like cell subpopulations. Comparison of pediatric and adult tumors further demonstrates a significant increase of mesenchymal cell states in adult H3-K27M DMGs, which we link to differences in glioma-associated immune cell compartments between age groups. Together, this study sheds light on the effects of age- and region-dependent microenvironments in shaping cellular identities in H3-K27M DMGs.

Histone 3 lysine27-to-methionine mutant diffuse midline gliomas (H3-K27M DMGs) are among the most lethal brain tumors. Their putative cellular hierarchy has been shown to be driven by self-renewing stem-like cells arrested in an oligodendrocyte precursor-like (OPC-like) state, of which few cells are able to differentiate towards more mature astrocyte (AC)-like and oligodendrocyte (OC)-like cells. However, the spatial organization underlying this tumor cell architecture and its microenvironmental interactions in intact H3-K27M DMG tissues remain unknown. Here, we profiled the single cell transcriptomes of 45 patient H3-K27M DMGs and derived cell population-specific marker gene combinations to characterize the single cell spatial organization of 16 tumors using targeted in situ sequencing. We thereby resolved different malignant and non-malignant cell populations including cycling, OPC-like, AC-like, OC-like, mesenchymal tumor cells, and non-malignant oligodendrocytes, astrocytes, neurons, myeloid cells, T cells, and vascular cells directly in situ. Global neighborhood analyses indicate a higher tendency of cycling OPC-like cells, vascular cells, and neurons to localize within a more restricted homogeneous compartment, whereas AC-like cells, non-malignant astrocytes and myeloid cells tend to intermingle with different cell populations in a more diffuse manner. Among malignant cells, we observed cycling OPC-like and OC-like cells to co-localize within a niche-like structure that is surrounded by more differentiated AC-like cells. We further validated this stem-like niche at the protein level using multiplexed immunofluorescence via the CODEX system. Finally, we characterized relationships between malignant and non-malignant cells, consistently identifying preferred neighborhoods of mesenchymal tumor cells with vascular and myeloid cells. Together, this study resolves the spatial architecture of H3-K27M DMG malignant and non-malignant cells at single cell resolution and identifies a local niche of the oligodendroglial lineage containing the OPC-like cancer stem-like cells, thus providing novel insights into the cancer stem-like compartment in H3-K27M DMGs and suggesting potential avenues for its perturbation.

Atypical Teratoid/Rhabdoid Tumors (ATRTs) are known for exhibiting high inter-tumor heterogeneity, even though they are almost all characterized by a common loss of SMARCB1 (or rarely SMARCA4). Three subgroups have been identified at bulk methylome and transcriptome level: ATRT-TYR, ATRT-SHH, and ATRT-MYC. To better understand the biology underlying each subgroup and potentially unveil their (different) cell(s) of origin, we performed single-cell transcriptomic analyses in 22 ATRTs using fresh frozen samples and both 10X and Smartseq technology. All data, grouped by technology, underwent quality control and normalization, regressing out the biases introduced by each sample. Tumor microenvironment (TME) and tumor bulk (TB) clusters were characterized by a combination of copy number variant analyses, enrichment in literature lists of marker genes for specific cell populations, and in-depth analysis of differentially enriched (DE) genes. Non-negative Matrix Factorization (NMF) was applied to TB to reveal major transcriptional profiles, which were grouped into meta-signatures. A total of 71 gene lists were retrieved from NMF (TB) and DE analyses (TME + TB), that gathered into 11 signature groups by Jaccard similarity, with one extra group accounting for unique signatures. Three groups targeted TME, accounting for either microglia, fibroblasts and endothelial cells, or OPCs, oligodendrocytes, astrocytes and neurons. These signatures are enriched in specific clusters across technologies. The remaining eight groups divide into two types, either enriched in clusters predominantly formed by cells of one or two ATRT subgroups or signatures enriched for a particular phenotype, such as cilial, cycling, axonogenesis or EM transition. While the first type is enriched across clusters in a gradient fashion, the second shows enrichment for selected clusters across technologies. Further analyses on the integrated dataset and additional samples are ongoing to validate and refine these 11 signature groups in ATRTs to see how this may lead to new treatment approaches.

Melanoma-derived brain metastases (MBM) represent an unmet clinical need due to central nervous system (CNS) progression as a frequent, end-stage site of disease. Immune checkpoint inhibition (ICI) represents a clinical opportunity against MBM; however, the MBM tumor microenvironment (TME) has not been fully elucidated in the context of ICI. To dissect unique MBM-TME elements and correlates of MBM-ICI response, we collected 32 fresh MBM and performed single cell RNA sequencing of the MBM-TME and T cell receptor clonotyping on T cells from MBM and matched blood and extracranial lesions. We observed myeloid phenotypic heterogeneity, most notably multiple distinct neutrophil states including an IL-8 expressing population that correlated with malignant cell epithelial-to-mesenchymal transition. Additionally, we observe significant relationships between intracranial T cell phenotypes and the distribution of T cell clonotypes intracranially and peripherally. We found that the phenotype, clonotype, and overall number of MBM-infiltrating T cells were associated with response to ICI, suggesting that ICI-responsive MBMs interact with peripheral blood in a manner similar to extracranial lesions. These data demonstrate unique features of the MBM-TME, which may represent potential targets to improve clinical outcomes for patients with MBM.

Embryonal tumor with multilayered rosettes (ETMR) is a malignant brain tumor that typically occurs in children under the age of three. Most patients die within two years of diagnosis, and more effective, targeted therapies are urgently needed. To better characterize the oncogenic mechanisms of key driver alterations and identify novel therapeutic targets, we studied the cellular heterogeneity of ETMR using single-cell RNA sequencing. Analyses conducted on >3,000 high-quality cells collected from ten primary and relapse specimens revealed a common cellular hierarchy across all tumors: A highly proliferative neural stem cell-like population (SOX2+) gives rise to intermediate progenitors (ASCL1+) and more differentiated neuron-like cells (STMN2/4+). These malignant populations closely match histological patterns of ETMR (i.e. rosettes, neuropil), as observed by immunofluorescence microscopy. Comparison to single-cell datasets from human embryos indicates resemblance to cell populations of the developing brain, but also reveals key ETMR-specific differences, including expression of the chromosome 19 miRNA cluster (C19MC, the presumed genetic driver of most ETMRs), which is restricted to the stem cell-like population. We next investigated if targeting C19MC is a viable strategy to disrupt the cellular hierarchy of ETMR. Silencing with antisense oligonucleotides shows pronounced reduction of cell line growth for a specific subset of the 46 members of C19MC. These miRNAs share seed sequences with evolutionary conserved miRNAs that have been shown to regulate pluripotency and self-renewal of embryonic stem cells. We hypothesize that select C19MC members play similar roles in ETMR and represent bona fide targets for therapeutic targeting using antisense technology.

Ependymoma represents a heterogeneous disease affecting the entire neuraxis. Extensive molecular profiling efforts have identified molecular ependymoma subgroups based on DNA methylation. However, the intratumoral heterogeneity and developmental origins of these groups are only partially understood, and effective treatments are still lacking for about 50% of patients with high-risk tumors. We interrogated the cellular architecture of ependymoma using single cell/single nucleus RNA-sequencing to analyze 24 tumor specimens across major molecular subgroups and anatomic locations. We additionally analyzed ten patient-derived ependymoma cell models and two patient-derived xenografts (PDXs). Interestingly, we identified an analogous cellular hierarchy across all ependymoma groups, originating from undifferentiated neural stem cell-like populations towards different degrees of impaired differentiation states comprising neuronal precursor-like, astro-glial-like, and ependymal-like tumor cells. While prognostically favorable ependymoma groups predominantly harbored differentiated cell populations, aggressive groups were enriched for undifferentiated subpopulations. Projection of transcriptomic signatures onto an independent bulk RNA-seq cohort stratified patient survival even within known molecular groups, thus refining the prognostic power of DNA methylation-based profiling. Furthermore, we identified novel potentially druggable targets such as IGF- and FGF-signaling within poorly prognostic transcriptional programs. Ependymoma-derived cell models/PDXs widely recapitulated the transcriptional programs identified within fresh tumors and are leveraged to validate identified target genes in functional follow-up analyses. Taken together, our analyses reveal a developmental hierarchy and transcriptomic context underlying the biologically and clinically distinct behavior of ependymoma groups. The newly characterized cellular states and underlying regulatory networks could serve as basis for future therapeutic target identification and reveal biomarkers for clinical trials.

Ependymoma represents a heterogeneous disease affecting the entire neuraxis. Extensive molecular profiling efforts have identified molecular ependymoma subgroups based on DNA methylation. However, the intratumoral heterogeneity and developmental origins of these groups are only partially understood, and effective treatments are still lacking for about 50% of patients with high-risk tumors. We interrogated the cellular architecture of ependymoma using single cell/nucleus RNA-sequencing to analyze 24 tumor specimens across major molecular subgroups and anatomic locations. We additionally analyzed ten patient-derived ependymoma cell models and two patient-derived xenografts (PDXs). Interestingly, we identified an analogous cellular hierarchy across all ependymoma groups, originating from undifferentiated neural stem cell-like populations towards different degrees of impaired differentiation states comprising neuronal precursor-like, astro-glial-like, and ependymal-like tumor cells. While prognostically favorable ependymoma groups predominantly harbored differentiated cell populations, aggressive groups were enriched for undifferentiated subpopulations. Projection of transcriptomic signatures onto an independent bulk RNA-seq cohort stratified patient survival even within known molecular groups, thus refining the prognostic power of DNA methylation-based profiling. Furthermore, we identified novel potentially druggable targets including IGF- and FGF-signaling within poorly prognostic transcriptional programs. Ependymoma-derived cell models/PDXs widely recapitulated the transcriptional programs identified within fresh tumors and are leveraged to validate identified target genes in functional follow-up analyses. Taken together, our analyses reveal a developmental hierarchy and transcriptomic context underlying the biologically and clinically distinct behavior of ependymoma groups. The newly characterized cellular states and underlying regulatory networks could serve as basis for future therapeutic target identification and reveal biomarkers for clinical trials.

Medulloblastoma is a malignant childhood cerebellar tumor comprised of distinct molecular subgroups. Whereas genomic characteristics of these subgroups are well defined, the extent to which cellular diversity underlies their divergent biology and clinical behaviour remains largely unexplored. We used single-cell transcriptomics to investigate intra- and inter-tumoral heterogeneity in twenty-five medulloblastomas spanning all molecular subgroups. WNT, SHH, and Group 3 tumors comprised subgroup-specific undifferentiated and differentiated neuronal-like malignant populations, whereas Group 4 tumors were exclusively comprised of differentiated neuronal-like neoplastic cells. SHH tumors closely resembled granule neurons of varying differentiation states that correlated with patient age. Group 3 and Group 4 tumors exhibited a developmental trajectory from primitive progenitor-like to more mature neuronal-like cells, whose relative proportions distinguished these subgroups. Cross-species transcriptomics defined distinct glutamatergic populations as putative cells-of-origin for SHH and Group 4 subtypes. Collectively, these data provide novel insights into the cellular and developmental states underlying subtype-specific medulloblastoma biology.

High-grade gliomas harboring H3 G34R/V mutations exclusively occur in the cerebral hemispheres of adolescents and young adults, suggesting a distinct neurodevelopmental origin. Combining multimodal bulk and single-cell genomics with unbiased genome-scale CRISPR/Cas9 approaches, we here describe a GABAergic interneuron progenitor lineage as the most likely context from which these H3 G34R/V mutations drive gliomagenesis, conferring unique and tumor-selective gene targets essential for glioma cell survival, as validated genetically and pharmacologically. Phenotypically, we demonstrate that while H3 G34R/V glioma cells harbor the neurotransmitter GABA, they are developmentally stalled, and do not induce the neuronal hyperexcitability described in other glioma subtypes. These findings offer a striking counter-example to the prevailing view of glioma origins in glial precursor cells, resulting in distinct cellular, microenvironmental, and therapeutic consequences.

High-grade gliomas (HGG) are among the most prevalent and fatal cancers in pediatric, adolescent, and young adult (AYA) patients. Especially understudied are older children and young adults, aged 16–39 years. Previously, we profiled primary pediatric HGGs through single-cell transcriptomics and identified the genetic, epigenetic and developmental programs that drive their malignant progression. However, the questions of how these programs compare to those in older HGG patients, what the mechanisms are by which these tumors ultimately evolve to drive recurrence and treatment resistance, and how distinct tumor cell subpopulations bidirectionally communicate with their microenvironment remain to be elucidated. Here, we use single-nucleus RNA sequencing to compare 11 paired, matched high-grade gliomas at diagnosis and recurrence and 15 additional H3K27M primary and recurrent DMG samples in pediatric and AYA patients. In all tumors, we find both undifferentiated and differentiated tumor cells recapitulating distinct glial lineages, as well as diverse microenvironmental cell populations. When longitudinally comparing this tumor architecture within matched pairs, we find substantial differences in transcriptional program expressions. Diagnostic samples include more differentiated, astrocyte-like tumor cells, while cells from recurrent samples more highly express ribosomal and heat-shock protein genes, suggesting tumor progression- and treatment-related shifts. Ongoing sequencing and analysis will allow for unprecedented insight into the evolutionary dynamics of pediatric and AYA high-grade gliomas as well as delineate differences in the biology of DMGs occurring in different age groups. This multi-institutional project was funded by the National Institute of Health.

Brain tumors are comprised of cells with heterogeneous genetic and transcriptional states, resulting in substantial phenotypic diversity. This diversity is particularly evident in embryonal tumor with multilayered rosettes (ETMR), which shows a striking bi-phasic pattern for which it is named. A better understanding of its underlying molecular makeup is urgently needed to develop more effective therapeutic strategies that eliminate all malignant cell types underlying ETMR initiation, maintenance, progression, and relapse. Furthermore, the cellular origin of ETMR is currently poorly understood. We used plate-based single-cell RNA sequencing to assess the intratumoral heterogeneity in 6 fresh and 4 snap-frozen surgical biopsies, following a workflow that we have previously established to study pediatric high grade gliomas, medulloblastomas, and ependymomas. Computational analyses conducted on >4,000 single cells identified cellular hierarchies ranging from a proliferative, undifferentiated cell population to more differentiated, predominantly neural-like progeny in all samples. Patient-derived cell line and xenograft models partially recapitulated this hierarchy. We further integrated transcriptional programs identified in single cells with available datasets of the developing normal brain, as well as with programs identified in other pediatric brain tumor entities, to inform both putative cellular origins and ETMR-specific oncogenic pathways. These timely results provide unparalleled insights into the molecular underpinnings of the phenotypic heterogeneity observed in ETMR. Analyses aimed at further integrating malignant cell type abundances with genetic alterations and clinical annotations, and therapeutical targeting of malignant cell populations using in-vitro models are currently ongoing.

Children diagnosed with diffuse intrinsic pontine glioma (DIPG), a type of high grade glioma in the brainstem, currently have a dismal 5-year overall survival of only 2%. The majority of DIPG patients harbor a K27M mutation in histone 3.3 encoding genes (H3.3K27M). To understand if the aberrant epigenetic landscape induced by H3.3K27M provides an opportunity for novel targeted therapies, we conducted the first CRISPR/Cas9 screen using a focused library of 1,350 epigenetic regulatory and cancer related genes in six H3.3K27M DIPG patient-derived primary neurosphere cell lines. We identified gene dependencies in chromatin regulators, polycomb repressive complexes 1 and 2 (PRC1 and PRC2), histone demethylases, acetyltransferases and deacetylators as novel tumor cell dependencies in DIPG. We hypothesized that targeting dysregulated functions of chromatin regulators by genetically deleting and chemically targeting these epigenetically induced vulnerabilities, we could ameliorate, or even reverse the downstream oncogenic effects of the aberrant epigenetic landscape of DIPG. In our secondary CRISPR nanoscreen, we first used six single guide RNAs (sgRNA) to knockout each gene using CRISPR/Cas9 ribonucleoprotein nucleofections, followed by use of three best sgRNAs combined with homology directed repair templates. Compared to lentiviral delivery, nucleofection is a rapid method, with reduced off-target toxicity, suitable for single gene knockouts in DIPG neurospheres. Secondary CRISPR validations confirmed dependencies in BMI1, CBX4, KDM1A, EZH2, EED, SUZ12, HDAC2, and EP300. Next, we conducted a chemical screen using 20 inhibitors and degraders to target the aberrant activity of HDAC, KDM1A, P300/CBP, PRC1 and PRC2. We identified eight chemical compounds that were effective in H3.3K27M DIPG neurosphere cell lines at low drug concentrations. Among these, an inhibitor and degrader targeting P300/CBP activity indicates a novel strategy of epigenetic therapy in DIPG. Through our combinatorial testing, we will identify a synergistic combination of epigenetic therapy for treating children diagnosed with H3.3K27M DIPG.

Summary Medulloblastoma is a malignant childhood cerebellar tumour comprised of distinct molecular subgroups. Whereas genomic characteristics of these subgroups are well defined, the extent to which cellular diversity underlies their divergent biology and clinical behaviour remains largely unexplored. We used single-cell transcriptomics to investigate intra- and inter-tumoural heterogeneity in twenty-five medulloblastomas spanning all molecular subgroups. WNT, SHH, and Group 3 tumours comprised subgroup-specific undifferentiated and differentiated neuronal-like malignant populations, whereas Group 4 tumours were exclusively comprised of differentiated neuronal-like neoplastic cells. SHH tumours closely resembled granule neurons of varying differentiation states that correlated with patient age. Group 3 and Group 4 tumours exhibited a developmental trajectory from primitive progenitor-like to more mature neuronal-like cells, whose relative proportions distinguished these subgroups. Cross-species transcriptomics defined distinct glutamatergic populations as putative cells-of-origin for SHH and Group 4 subtypes. Collectively, these data provide novel insights into the cellular and developmental states underlying subtype-specific medulloblastoma biology.

The cellular composition of H3K27M gliomas Diffuse midline gliomas with histone H3 lysine27-to-methionine mutations (H3K27M-glioma) are an aggressive type of childhood cancer with few options for treatment. Filbin et al. used a single-cell sequencing approach to study the oncogenic programs, genetics, and cellular hierarchies of H3K27M-glioma. Tumors were mainly composed of cells resembling oligodendrocyte precursor cells, whereas differentiated malignant cells were a smaller fraction. In comparison with other gliomas, these cancers had distinct oncogenic programs and stem cell–like profiles that contributed to their stable tumor-propagating potential. The analysis also identified a lineage-specific marker that may be useful in developing therapies. Science , this issue p. 331

Human cancers are composed of cells with heterogeneous genetic and epigenetic states, resulting in substantial phenotypic diversity. Medulloblastoma (MB), a clinically challenging, malignant childhood brain tumor, is no exception. A better understanding of its cellular heterogeneity is urgently needed in order to develop more efficacious therapeutic strategies that eliminate all tumor subclones underlying MB initiation, maintenance, progression, and relapse. We used single-cell RNA sequencing to assess intratumoral heterogeneity in 20 MB samples representing all four molecular subgroups. Fresh surgical biopsies were dissociated and flow sorted into single cells prior to full-length transcriptome sequencing using the Smart-seq2 protocol. Computational analyses conducted on >6,000 single cells identified cellular hierarchies ranging from proliferative, undifferentiated cell types to more differentiated neural- and astrocyte-like progeny. Transcriptional differences in undifferentiated cell populations between MB subgroups suggested distinct cellular origins, and informed oncogenic pathways believed to promote MB pathogenesis. Identification of transcriptional programs active in single cells aided the deconvolution of published bulk RNA-seq profiles, providing a deeper understanding of the heterogeneity characteristic of Group 3/4 subtypes. Lastly, inference of copy-number variations and probable driver-gene mutations from the single-cell data provided evidence for genetic subclones in a subset of patients. Overlaying transcriptional profiles on MB genetic subclones unveiled how oncogenic programs adapted during tumor evolution. These interim results provide unparalleled insights into the cellular heterogeneity of MB subgroups. Analyses aimed at further resolving molecular substructure in Group 3/4, verifying the developmental origins of MB subgroups, and implicating defined cell populations driving tumorigenesis are ongoing.

Schwannomas are found either sporadically or as part of a syndrome. Genetic alterations found in schwannomas include those in NF2 and SMARCB1 (implicated in schwannomatosis), and BRAF V600E. Reduced copy number of chromosome 22q can be the mechanism for loss of the unaltered allele ofNF2 and/orSMARCB1in some schwannomas. We analyzed a cohort of 57 vestibular and non-vestibular schwannomas from three clinical subtypes (Neurofibromatosis 2, schwannomatosis, and sporadic) for reduced copy number (NF2 and SMARCB1), and mutations (NF2, SMARCB1, LZTR1, NF1 and BRAF). Tumors were histologically classified as conventional/classic, or hybrid. Genomic DNA was extracted from FFPE biopsies. Full coding regions of genes NF1, NF2, LZTR1, SMARCB1 and BRAF were enriched by anchored multiplex-PCR (AMP). Resulting libraries were quantitated and sequenced on the Illumina MiSeq System. Bioinformatics analyses were performed using custom scripts. Copy number alteration was detected using a read depth approach after consolidation of single molecule derived reads. INI1 immunohistochemistry was performed. RNA ISH for SMARCB1 was performed. Five tumors were of hybrid histology, and all five were syndromic. Reduced copy number for NF2 was identified in 33% of vestibular schwannomas and 18% of non-vestibular schwannomas. There was no significant difference in reduced copy number of NF2 between sporadic and syndromic forms. BRAF V600E and NF1 mutations were not identified in any case. For most INI1 mosaics, there was no molecular event explaining the partial loss of INI1 IHC. This study confirms the association of hybrid histology with syndromic schwannomas. NF1 mutations did not explain hybrid histology. We did not find BRAF V600E mutations in this series. INI1 mosaic pattern was not explained by reduced copy number or mutation of SMARCB1 as this was found in only a small number of the cases. These results suggest that loss of INI1 expression could be the result of epigenetic alterations.

While human tumors are shaped by the genetic evolution of cancer cells, evidence also suggests that they display superimposed hierarchies reminiscent of normal development. Yet, relating the genetic and hierarchical models of cancer organization has been limited by technical challenges. Here, studying H3 K27M-mutant diffuse midline gliomas we profiled thousands of single cells from fresh tumors by RNA-seq and reconstructed the genetic and epigenetic cellular architecture of tumors in patients. We identified a subset of cancer cells that are differentiated along specialized glial programs while a distinct subset expresses neural stem/progenitor expression programs. By identifying distinct genetic clones, we show that these cellular architectures evolve during the progression of the tumors, suggesting that developmental programs and genetic influences coordinately determine tumor architecture. These results provide unprecedented insight into the cellular composition of midline brain tumors at single-cell resolution and may help harmonize the cancer stem cell and the genetic models of cancer, with critical implications for disease management.

Summary Ependymoma is a heterogeneous entity of central nervous system tumors with well-established molecular groups. Here, we apply single-cell RNA sequencing to analyze ependymomas across molecular groups and anatomic locations to investigate their intratumoral heterogeneity and developmental origins. Ependymomas are composed of a cellular hierarchy initiating from undifferentiated populations, which undergo impaired differentiation toward three lineages of neuronal-glial fate specification. While prognostically favorable groups of ependymoma predominantly harbor differentiated cells, aggressive groups are enriched for undifferentiated cell populations. The delineated transcriptomic signatures correlate with patient survival and define molecular dependencies for targeted treatment approaches. Taken together, our analyses reveal a developmental hierarchy underlying ependymomas relevant to biological and clinical behavior., Graphical Abstract, Highlights • Cellular hierarchies in ependymoma reflect impaired neurodevelopment • Ependymoma cells follow neuronal-, astrocytic-, and ependymal-like trajectories • Undifferentiated programs confer inferior prognosis and are enriched at recurrence • Preclinical data suggest feasibility of inhibiting undifferentiated subpopulations, Gojo et al. use single-cell RNA sequencing to investigate intratumoral heterogeneity and cellular hierarchy in pediatric ependymoma, identifying impaired neurodevelopmental trajectories. The malignant trajectories reveal therapeutic targets and prognostic signatures in ependymoma.