Human cancers are complex ecosystems composed of cells with distinct phenotypes, genotypes and epigenetic states, but current models do not adequately reflect tumor composition in patients. We used single cell RNA-seq to profile 430 cells from five primary glioblastomas, which we found to be inherently variable in their expression of diverse transcriptional programs related to oncogenic signaling, proliferation, complement/immune response and hypoxia. We also observed a continuum of stemness-related expression states that enabled us to identify putative regulators of stemness in vivo. Finally, we show that established glioblastoma subtype classifiers are variably expressed across individual cells within a tumor and demonstrate the potential prognostic implications of such intratumoral heterogeneity. Thus, we reveal previously unappreciated heterogeneity in diverse regulatory programs central to glioblastoma biology, prognosis, and therapy.Tumor heterogeneity poses a major challenge to cancer diagnosis and treatment. It can manifest as variability between tumors, wherein different stages, genetic lesions or * Corresponding authors: Bernstein.bradley@mgh.harvard.edu (BEB), aregev@broadinstitute.org (AR), Suva.Mario@mgh.harvard.edu (MLS). † These authors contributed equally to this work. ‡ These authors contributed equally to this work. Glioblastoma is an archetypal example of a heterogeneous cancer and one of the most lethal human malignancies (9, 10). Intratumoral heterogeneity and redundant signaling routes likely underlie the inability of conventional and targeted therapies to achieve long-term remissions (11-13). These tumors contain cellular niches enriched for distinct phenotypic properties, including transient quiescence and self-renewal (14-16), adaptation to hypoxia (17), and resistance to radiation induced DNA damage (18,19). DNA and RNA profiles of bulk tumors have enabled genetic and transcriptional classification of glioblastomas (20,21). However, the relationships between different sources of intratumoral heterogeneitygenetic, transcriptional and functional -remain obscure.Single cell transcriptome analysis by 23) should in principle enable functional characterization from landmark genes and annotated gene sets, relate in vivo states to in vitro models, inform transcriptional classifications based on bulk tumors, and even capture genetic information for expressed transcripts. To interrogate intratumoral heterogeneity systematically, we isolated individual cells from five freshly resected and dissociated human glioblastomas and generated single cell full-length transcriptomes using SMART-seq (96-192 cells/tumor, total 672 cells; Fig. 1A). Prior to sorting, the suspension was depleted for CD45 + cells to remove inflammatory infiltrate. As a control, we also generated population (bulk) RNA-seq profiles from the CD45-depleted tumor samples. All tumors were IDH1/2 wild type primary glioblastomas ( Fig. S1) and three were EGFR amplified as determined by routine clinical tests (Table S1). We exclude...

SUMMARY The diverse malignant, stromal, and immune cells in tumors affect growth, metastasis and response to therapy. We profiled transcriptomes of ~6,000 single cells from 18 head and neck squamous cell carcinoma (HNSCC) patients, including five matched pairs of primary tumors and lymph node metastases. Stromal and immune cells had consistent expression programs across patients. Conversely, malignant cells varied within and between tumors in their expression of signatures related to cell cycle, stress, hypoxia, epithelial differentiation, and partial epithelial-to-mesenchymal transition (p-EMT). Cells expressing the p-EMT program spatially localized to the leading edge of primary tumors. By integrating single-cell transcriptomes with bulk expression profiles for hundreds of tumors, we refined HNSCC subtypes by their malignant and stromal composition, and established p-EMT as an independent predictor of nodal metastasis, grade, and adverse pathologic features. Our results provide insight into the HNSCC ecosystem and define stromal interactions and a p-EMT program associated with metastasis.

Gain-of-function IDH mutations are initiating events that define major clinical and prognostic classes of gliomas1,2. Mutant IDH protein produces a novel onco-metabolite, 2-hydroxyglutarate (2-HG), that interferes with iron-dependent hydroxylases, including the TET family of 5′-methylcytosine hydroxylases3–7. TET enzymes catalyze a key step in the removal of DNA methylation8,9. IDH mutant gliomas thus manifest a CpG island methylator phenotype (G-CIMP)10,11, though the functional significance of this altered epigenetic state remains unclear. Here we show that IDH mutant gliomas exhibit hyper-methylation at CTCF binding sites, compromising binding of this methylation-sensitive insulator protein. Reduced CTCF binding is associated with loss of insulation between topological domains and aberrant gene activation. We specifically demonstrate that loss of CTCF at a domain boundary permits a constitutive enhancer to aberrantly interact with the receptor tyrosine kinase gene PDGFRA, a prominent glioma oncogene. Treatment of IDH mutant gliomaspheres with demethylating agent partially restores insulator function and down-regulates PDGFRA. Conversely, CRISPR-mediated disruption of the CTCF motif in IDH wildtype gliomaspheres up-regulates PDGFRA and increases proliferation. Our study suggests that IDH mutations promote gliomagenesis by disrupting chromosomal topology and allowing aberrant regulatory interactions that induce oncogene expression.

High-grade gliomas are lethal brain cancers whose progression is robustly regulated by neuronal activity. Activity-regulated growth factor release promotes glioma growth, but this alone is insufficient to explain the effect that activity exerts on glioma progression. Here, we use singlecell transcriptomics, electron microscopy, whole-cell patch-clamp electrophysiology and calcium imaging to demonstrate that neuron-glioma interactions include electrochemical communication through bona fide AMPA receptor-dependent neuron-glioma synapses. Neuronal activity also evokes non-synaptic activity-dependent potassium currents that are amplified through gap junction-mediated tumor interconnections forming an electrically-coupled network. Glioma membrane depolarization assessed with in vivo optogenetics promotes proliferation, while pharmacologically or genetically blocking electrochemical signaling inhibits glioma xenograft growth and extends mouse survival. Emphasizing positive feedback mechanisms by which gliomas increase neuronal excitability and thus activity-regulated glioma growth, human intraoperative electrocorticography demonstrates increased cortical excitability in gliomainfiltrated brain. Together, these findings indicate that synaptic and electrical integration in neural circuits promotes glioma progression.

Summary Developmental fate decisions are dictated by master transcription factors (TFs) that interact with cis-regulatory elements to direct transcriptional programs. Certain malignant tumors may also depend on cellular hierarchies reminiscent of normal development but superimposed on underlying genetic aberrations. In glioblastoma (GBM), a subset of stem-like tumor-propagating cells (TPCs) appears to drive tumor progression and underlie therapeutic resistance, yet remain poorly understood. Here, we identify a core set of neurodevelopmental TFs (POU3F2, SOX2, SALL2, OLIG2) essential for GBM propagation. These TFs coordinately bind and activate TPC-specific regulatory elements, and are sufficient to fully reprogram differentiated GBM cells to ‘induced’ TPCs, recapitulating the epigenetic landscape and phenotype of native TPCs. We reconstruct a network model that highlights critical interactions and identifies novel therapeutic targets for eliminating TPCs. Our study establishes the epigenetic basis of a developmental hierarchy in GBM, provides detailed insight into underlying gene regulatory programs, and suggests attendant therapeutic strategies.

INTRODUCTION Tumor fitness, evolution, and resistance to therapy are governed by selection of malignant cells with specific genotypes, by expression programs related to cellular phenotypes, and by influences of the tumor microenvironment (TME). Although bulk tumor analysis can interrogate the genetic state of tumor cells with high precision, bulk expression profiles average the diverse cells within each tumor, thereby masking critical differences and providing limited insight into cancer cell programs and TME influences. Single-cell RNA sequencing (scRNA-seq) can help to address those challenges but incurs financial and logistic considerations, including the time required to accrue large cohorts of fresh tumor specimen for single-cell analysis. RATIONALE We reasoned that scRNA-seq of a limited number of representative tumors could be combined with bulk data from large cohorts to decipher differences between tumor subclasses. In this approach, bulk samples collected for large cohorts, such as from The Cancer Genome Atlas (TCGA), are first used to define the combined effects of differences in cancer cell genotypes, phenotypes, and the composition of the TME. Single-cell analysis of a limited set of representative tumors is then used to distinguish those effects. We applied this approach to understand the differences between two types of isocitrate dehydrogenase (IDH)-mutant gliomas: astrocytoma (IDH-A) and oligodendroglioma (IDH-O). IDH-A and IDH-O are distinguished by co-occurring signature genetic events and by histopathology and are thought to recapitulate distinct glial lineages. By combining 9879 scRNA-seq profiles from 10 IDH-A tumors, 4347 scRNA-seq profiles from six IDH-O tumors, and 165 TCGA bulk RNA profiles, we could decipher differences between these two tumor types at single-cell resolution. RESULTS We find that differences in bulk expression profiles between IDH-A and IDH-O are primarily explained by the impact of signature genetic events and TME composition, but not by distinct expression programs of glial lineages in the malignant cells. We infer that both IDH-A and IDH-O share the same developmental hierarchy, consisting in each case of three subpopulations of malignant cells: nonproliferating cells differentiated along the astrocytic and oligodendrocytic lineages, and proliferative undifferentiated cells that resemble neural stem/progenitor cells. By analyzing tumors of different clinical grades, we observe that higher-grade tumors present enhanced proliferation, larger pools of undifferentiated glioma cells, and an increase in macrophage over microglia programs in the TME. CONCLUSION Our approach provides a general framework to decipher differences between classes of human tumors by decoupling cancer cell genotypes, phenotypes, and the composition of the TME. The shared glial lineages and developmental hierarchies observed in IDH-A and IDH-O suggest a common progenitor for all IDH-mutant gliomas, shedding light on a longstanding debate in gliomagenesis. In contrast to the similarity in gl...

Summary The aberrant transcription factor EWS-FLI1 drives Ewing sarcoma yet its molecular function is incompletely understood. We find that EWS-FLI1 reprograms gene regulatory circuits in Ewing sarcoma by directly inducing or repressing enhancers. At GGAA repeat elements, which lack evolutionary conservation and regulatory potential in other cell types, EWS-FLI1 multimers induce chromatin opening and create de novo enhancers that physically interact with target promoters. Conversely, EWS-FLI1 inactivates conserved enhancers containing canonical ETS motifs by displacing wild type ETS transcription factors. These divergent chromatin-remodeling patterns repress tumor suppressors and mesenchymal lineage regulators, while activating oncogenes and new potential therapeutic targets, such as the kinase VRK1. Our findings demonstrate how EWS-FLI1 establishes an oncogenic regulatory program governing both tumor survival and differentiation.

Triple-negative breast cancer (TNBC) is an aggressive subtype characterized by extensive intratumoral heterogeneity. To investigate the underlying biology, we conducted single-cell RNA-sequencing (scRNA-seq) of >1500 cells from six primary TNBC. Here, we show that intercellular heterogeneity of gene expression programs within each tumor is variable and largely correlates with clonality of inferred genomic copy number changes, suggesting that genotype drives the gene expression phenotype of individual subpopulations. Clustering of gene expression profiles identified distinct subgroups of malignant cells shared by multiple tumors, including a single subpopulation associated with multiple signatures of treatment resistance and metastasis, and characterized functionally by activation of glycosphingolipid metabolism and associated innate immunity pathways. A novel signature defining this subpopulation predicts long-term outcomes for TNBC patients in a large cohort. Collectively, this analysis reveals the functional heterogeneity and its association with genomic evolution in TNBC, and uncovers unanticipated biological principles dictating poor outcomes in this disease.