Tumor microenvironment shapes the spatial organization of glioblastoma cell states.

Skip to main content An official website of the United States government Here's how you know Here's how you know Official websites use .gov A .gov website belongs to an official government organization in the United States. Secure .gov websites use HTTPS A lock ( Lock Locked padlock icon ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites. Search Log in Dashboard Publications Account settings Log out Search… Search NCBI Primary site navigation Search Logged in as: Dashboard Publications Account settings Log in Search PMC Full-Text Archive Search in PMC Journal List User Guide PERMALINK Copy As a library, NLM provides access to scientific literature. Inclusion in an NLM database does not imply endorsement of, or agreement with, the contents by NLM or the National Institutes of Health. Learn more: PMC Disclaimer | PMC Copyright Notice Neuro Oncol . 2026 Jan 10;28(3):585–596. doi: 10.1093/neuonc/noag003 Search in PMC Search in PubMed View in NLM Catalog Add to search Tumor microenvironment shapes the spatial organization of glioblastoma cell states Pranav Prakash Pranav Prakash 1 Department of Neurosurgery, University of California, San Francisco (UCSF), San Francisco, CA, USA Find articles by Pranav Prakash 1 , James Trippett James Trippett 2 Department of Neurosurgery, University of California, San Francisco (UCSF), San Francisco, CA, USA Find articles by James Trippett 2 , Cameron Ehsan Cameron Ehsan 3 Department of Neurosurgery, University of California, San Francisco (UCSF), San Francisco, CA, USA Find articles by Cameron Ehsan 3 , Joseph Namkung Joseph Namkung 4 Department of Neurosurgery, University of California, San Francisco (UCSF), San Francisco, CA, USA Find articles by Joseph Namkung 4 , Meeki Lad Meeki Lad 5 Department of Neurosurgery, University of California, San Francisco (UCSF), San Francisco, CA, USA Find articles by Meeki Lad 5 , Manish K Aghi Manish K Aghi 6 Department of Neurosurgery, University of California, San Francisco (UCSF), San Francisco, CA, USA Find articles by Manish K Aghi 6, ✉ Author information Article notes Copyright and License information 1 Department of Neurosurgery, University of California, San Francisco (UCSF), San Francisco, CA, USA 2 Department of Neurosurgery, University of California, San Francisco (UCSF), San Francisco, CA, USA 3 Department of Neurosurgery, University of California, San Francisco (UCSF), San Francisco, CA, USA 4 Department of Neurosurgery, University of California, San Francisco (UCSF), San Francisco, CA, USA 5 Department of Neurosurgery, University of California, San Francisco (UCSF), San Francisco, CA, USA 6 Department of Neurosurgery, University of California, San Francisco (UCSF), San Francisco, CA, USA ✉ Corresponding Author: Manish K. Aghi, MD, PhD, Department of Neurosurgery, University of California, San Francisco (UCSF), Room HD-465, 1450 Third Street, San Francisco, CA 94158 ( [email protected] ). Received 2025 Aug 18; Accepted 2026 Jan 3; Collection date 2026 Mar. © The Author(s) 2026. Published by Oxford University Press on behalf of the Society for Neuro-Oncology. This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial License ( https://creativecommons.org/licenses/by-nc/4.0/ ), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited. For commercial re-use, please contact [email protected] for reprints and translation rights for reprints. All other permissions can be obtained through our RightsLink service via the Permissions link on the article page on our site—for further information please contact [email protected]. PMC Copyright notice PMCID: PMC13070500  PMID: 41520155 Abstract Glioblastoma is characterized by heterogeneous and plastic cellular populations that adopt transcriptional programs shaped by genetic alterations and microenvironmental cues. Recent studies have identified at least 4 partially inconvertible cell states—astrocytic-like, neural progenitor-like, oligodendrocyte progenitor-like, and mesenchymal-like—that represent aberrant developmental programs. Expanded analysis further reveals hybrid and intermediate states that form continuous transcriptional and metabolic gradients. These states exhibit spatial organization, assembling into 3 distinct microanatomical niches: a perivascular niche enriched with mesenchymal-like and oligodendrocyte progenitor-like cells, a hypoxic niche harboring quiescent and stressed cells of all states, and an invasive niche containing astrocyte-like or proneural populations. Niches continuously remodel as cell states transition, migrate, and reestablish new programming in response to angiogenesis, hypoxia, immune infiltration, and neuronal activity. This interplay between states and the microenvironment generates a self-renewing spatial architecture, maintaining expansion at the edge and protection within the core. This review integrates single-cell, single-nucleus, and spatial studies to describe a microenvironmental-driven model of cell state organization. Understanding how these multiscale drives converge to generate a continuum of cell state identities may reveal strategies to disrupt the spatial architecture underlying glioblastoma plasticity and recurrence. Keywords: cell states, glioblastoma, spatial niche Key Points. Glioblastoma is characterized by at least 4 interconvertible cell states that recapitulate neurodevelopment: astrocyte-like, neural progenitor-like, oligodendrocyte progenitor-like, and mesenchymal-like. Glioblastoma cell states exhibit spatial organization, assembling into 3 distinct microanatomical niches that drive tumor progression and resistance: the perivascular niche, hypoxic niche, and invasive niche. Standard therapies promote treatment-resistant populations, particularly mesenchymal-like cells in the invasive niche that predominate in recurrence through enhanced stemness and vascular remodeling. Glioblastoma (GBM) is an aggressive grade 4 astrocytoma marked by profound intratumoral heterogeneity, plasticity, invasive potential, and recurrence. Median survival remains low at 12 to 18 months despite multimodal therapies. 1-3 These poor clinical outcomes reflect the underlying complexity of the disease. GBM is not characterized by a static mass of malignant cells but rather a dynamic tumor microenvironment of interconvertible cell types organized into distinct spatial niches. 4 , 5 Historically, this heterogeneity was attributed to a model of glioblastoma stem cells (GSCs), a subpopulation capable of self-renewal, tumor initiation, and differentiation into multiple or restricted lineages. 6 The recent single-cell era has reframed this concept through transcriptionally defined cellular states. Within this new framework, GSCs may exist in multiple transcriptional states depending on the local microenvironment, where functional plasticity and transcriptional states overlap. These applications of single-cell and single-nucleus analyses of GBM cells have delineated at least 4 principal and partially inconvertible states, including astrocyte-like (AC-like), neural progenitor-like (NPC-like), oligodendrocyte progenitor-like (OPC-like), and mesenchymal-like (MES-like) states. Complementary studies ( Table 1 ) have extended this framework by identifying additional intermediate populations that occupy continuous axes of stemness, metabolic programs, and injury response. 8 Spatial transcriptomics further demonstrates that cell states self-organize into specialized microenvironments within the tumor based on intrinsic and extrinsic factors. 7 , 8 These studies demonstrate a continuum of cell state identities, with each state predominating in specific tumor regions and preferentially associating with malignant cells of other states. 10 This forms regional networks dispersed throughout the tumor in which cell states can localize, facilitating rapid angiogenesis, immunosuppression, growth, and invasion. 10 , 11 Whether such cell state organizational patterns are the result of local microenvironmental triggers or the migration of states toward favorable niches has yet to be elucidated. However, some cell states may evade treatment or differentiate posttreatment given their spatial localization, allowing for recurrence driven by cell states. Table 1. Representative Gene Signatures of Cell State Frameworks Study State Genes Neftel et al. AC-like CST3, GFAP, S100B, HOPX, SLC1A3, MLC1 OPC-like PLP1, ALCAM, OLIG1, OMG, PLLP NPC-like 1 SOX4, DCX, CD24, DLL3, SOX11 NPC-like 2 RND3, SOX11, DCX, CD24, STMN4, STMN2, DLX5, DLX6-AS1 MES-like 1 VIM, ANXA1, ANXA2, CHI3L1, CD44 MES-like 2 DDIT3, ENO2, VIM, ADM, LDHA, HILPDA Garofano et al. Mitochondrial SDHB, ATP5E, COA6, COX6A1, CPT1C, NDUFA2, NDUFAF5, UQCRFS1, TIMM50, OPA3, PARS2, SLC45A1, ENO1, H6PD, SLC16A10, SLC16A3, XBP1, NRAS, PFKP Glycolytic/plurimetabolic PTEN, PIK3CB, HGF, NAMPT, MET, MAFB, TFAM, ATP5S, POLG2, SUCLA2, NARS2, NF1, RB1, CDKN1B, PPARGC1A Neuronal ATRX, TET1, NEUROD6, GABRR2, HTR5A, E2F2, EYA3, HES2, PAX7 Proliferative/progenitor PCNA, SKP2, AURKA, PLK4, MAD2L1, FANCD2, RAD51, CHEK2, BIRC3, SETD7, CBX6, CBX7, BTK, EZH2, RUVBL1, PDGFRA, CCNE1, MAD1L1, TEP1, NEK9 Spitzer et al. OPC-like TNR, LHFPL3, FGF14 AC-like GFAP, AQP4, TNC Hypoxia VEGFA, NDRG1, HILPDA MES-like CD44, ANXA2, CHI3L1 NPC-like SOX4, SOX11, DCX Glial progenitor cell-like MEIS1, MEOX2, EGFR NEU-like RBFOX1, CNTNAP2, SYN3, NRGN, RTN1, SNAP25 Stress HSPA5, HSPD1, UFM1, HSPA1A, UBB, TCP1 Cilia-like DNAH6, DNAH9, CFAP43 Open in a new tab Abbreviations: AC-like, astrocyte-like; GPC-like, glial progenitor-like; MES-like, mesenchymal-like; NEU-like, neuronal-like; NPC-like, neural progenitor-like; OPC-like, oligodendrocyte progenitor-like. The diversity of reported cell states may partly reflect the scope and limitations of different methodologies. Single-cell RNA sequencing first defined 4 cell states by profiling transcriptionally active cells, capturing more proliferative populations. Single-nucleus RNA sequencing extends this framework by recovering hypoxic, quiescent, and treatment-stressed cells lost during dissociation. Integrating these modalities suggests that GBM cell states may align along coordinated axes of aberrant neurodevelopment, metabolic reprogramming, and immune interaction, with the caveat that the definition of populations such as invasive and immune-interacting states varies between studies, which may reflect differences in sample preservation, spatial resolution, sampling time points, or tumor biology. Together, these studies redefine GBM as a continuously evolving system where cell state, spatial organization, and microenvironmental composition are interdependent. Understanding this dynamic is essential to establishing the drivers of heterogeneity, creating a functional and targetable aspect of GBM cell states. This review synthesizes recent developments in single-cell and spatial studies to propose an integrated model of cell state organization within spatial microenvironmental niches that underlie the plasticity of GBM. Drivers of Cell State in GBM Tumor Cells Genetic and Epigenetic Drivers The emergence of GBM cell states begins with genetic or epigenetic alterations in driver genes. 7 IDH-wild-type tumors have been classified as classical, proneural, and mesenchymal subtypes based on the genetic profile of the entire tumor. 10 Classical tumors show EGFR amplification/mutation and are enriched in AC-like cell states. 7 , 11 , 12 Proneural tumors feature PDGFRA mutations and CDKN2A deletions, correlating with OPC-like and NPC-like states. 7 , 11 , 12 Mesenchymal tumors are marked by NF1 and PTEN deletions and are enriched in MES-like states. 7 , 11 , 12 The proportion of specific cell states in GBM tumors correlates with copy number alterations at key loci: PDGFRA (OPC-like), CDK4 (NPC-like), EGFR (AC-like), and NF1 (MES-like). 7 , 11 , 12 The differential expression of additional markers, such as APOE, AQP4, and GFAP for AC-like cells and CHI3L1 and VEGFA for MES-like cells, further delineates this intratumoral heterogeneity. 12 GBM tumors comprise diverse cell states, contributing to both genetic and cellular variability. Clonal populations defined by driver mutations may acquire random passenger mutations, giving rise to subclones. 13 For instance, MES-like cells can be divided into MES-like 1 and 2 subclones, associated with reactivity and hypoxia, respectively. 14 Both driver and passenger mutations may influence quiescent, proliferative, and stem-like traits across cell states. These driver gene alterations influence both cell state identity and the spatial distribution of cells. EGFR- and PDGFRA-amplified cells are highly proliferative and tend to cluster in vascular regions, whereas non-amplified cells are more common in the hypoxic core. 12 , 15 This behavior is not observed in cells with co-amplification of EGFR and CDK4. 16 CDK4-amplified cells are often associated with hypoxic and immune-rich environments. 16 Reports on the localization of NF1-deleted MES-like cells vary, with enrichment in hypoxic and immune-infiltrated regions or enrichment in perivascular regions. These spatial dynamics likely promote distinct cell state distributions. Driver gene mutations, particularly in vulnerable regions such as promoter and enhancer regions, enable the accumulation of epigenetic alterations during malignant transformation. 17 However, rapid proliferation mitigates the impact of DNA hypermethylation. This results in a global hypomethylated cancer genome that permits broad ­transcriptional reactivation and emergence of multiple cell states seen in IDH-wild-type tumors. 12 , 18 In contrast, IDH-mutant tumor cells exhibit hypermethylation at regulatory regions, limiting their cellular plasticity. 19 The enhanced plasticity in IDH-wild-type tumors may underlie their higher recurrence rates and poorer clinical outcomes compared to IDH-mutant tumors. 12 , 19 Driver events such as EGFR amplifications are rare in IDH-mutant tumors, possibly reducing the formation of diverse clonal populations that give rise to spatially segregated lineages. This scarcity of driver alterations may lead to fewer accompanying epigenetic modifications underlying the more uniform cellular architecture and less aggressive behavior of IDH-mutant gliomas. 20 Spatial and clonal tracing studies suggest that these genetic and epigenetic configurations exhibit regional patterning within the tumor mass. Multi-region sampling from spatially distinct tumor fragments demonstrates a variety of copy number alterations and region-specific methylation within the same tumor. 21 Furthermore, many of these genetically defined subclones undergo clonal extinction as the tumor evolves. 22 Thus, spatial segregation may not confer long-term clonal stability as clones compete within their local microenvironments. Although the strength of these associations varies across methodologies, the genetic diversity of GBM is transiently reflected in its spatial organization. Transcriptomic Regulation of Cell States While genetic and epigenetic alterations shape GBM cell states, transcriptional regulation plays a key role in the maintenance of and transition between GBM cell states. Transcription factors like ASCL1, HES1, and OLIG2 guide cell state identity with their oscillation maintaining GSC quiescence. 12 , 23-25 Sustained expression promotes differentiation via neurodevelopmental pathways, while posttranscriptional modifications, such as OLIG2 phosphorylation, support a proliferative stem-like state in healthy neuronal cells. 26 Elevated OLIG2 expression enhances progenitor-like traits and migration, whereas its downregulation halts migration. Transition to a MES-like state has been associated with migration independent of the vasculature and correlates with immune infiltration, microglial/macrophage activation, and increased transcription factor and growth factor activity. 27 This MES-like shift has been shown to remodel the tumor microenvironment, promoting neuronal differentiation or reactive astrogliosis at lesion sites. 12 Aggressive tumor regions show upregulation of transcriptional programs such as PTPRZ1, EGFR, SIVRPD3, TPST1, and GUCD1, which are associated with glycoprotein metabolism, antigen processing, and axon injury responses. 12 Though microRNAs (miRNAs) have not been directly linked to the 4 main GBM cell states, they play key roles in cell cycling, mesenchymal transition, and oncogenic signaling. 28-31 Oncogenic miRNAs regulate mRNA stability and translation in malignant cells. 22 For example, miR-21 promotes proliferation via PTEN suppression, while its knockdown induces apoptosis. 28 , 29 miR-29b supports angiogenesis and stemness, particularly in mesenchymal subtypes, whereas miR-9 can suppress mesenchymal differentiation and targets HES1 to influence neural lineage development. 30 In contrast, miR-7 inhibits EGFR signaling and migration, limiting tumor growth. 31 These miRNAs contribute to tumor progression by promoting proliferation and genetic instability. 28 Imbalances between oncogenic and tumor-suppressive miRNAs may increase susceptibility to genetic and epigenetic changes in driver genes, giving rise to GBM cell states. 12 Despite the established functional roles of these transcriptional regulators in the biology of the 4 main GBM cell states, their spatial context remains poorly defined. Integrating high-resolution spatial profiling and transcriptomics will be necessary to determine whether specific transcriptional regulatory programs predominate within particular spatial niches or fluctuate across them. Influences of Microenvironmental Factors on Cellular Landscapes Microenvironmental factors are major determinants of GBM cell fate across spatial niches. 32 While genetic, epigenetic, and transcriptomic changes provide the capacity for genomic plasticity, the surrounding microenvironment may shape which cell states are expressed. Triggers like hypoxia, cytokines, and cell–cell interactions can guide gene expression and impose selective pressures that favor specific phenotypes. EGFR amplification is frequently associated with enrichment of AC-like transcriptional programs, while expression of the constitutively active EGFRvIII variant promotes partial astrocytic differentiation without full maturation, 33 potentially giving rise to progenitor-like astrocytic phenotypes rather than fully differentiated cells. Similarly, PDGFR- and NF1-altered tumors show immature OPC-like cells, with PDFRA overexpression mirroring EGFRvIII’s effect of blocking complete differentiation. 33 NPC-like cells remain dormant in the tumor core, driven by interferon signals, particularly IFN-γ from T cells, which suppress motility and induce quiescence. 33 In contrast, margin cells tend to differentiate toward AC-like or proneural populations, likely influenced by the neuron- and astrocyte-rich microenvironment. Regions within the tumor core marked by hypoxia show increased immune infiltration, including microglia, macrophages, NK cells, and T cells. 34 Immune infiltration, especially by myeloid cells, is closely associated with regions enriched in MES-like cells, which are often linked to a hypoxic microenvironment. 34 , 35 Under hypoxic conditions, inflammatory macrophages increase the production of pro-inflammatory cytokines while impairing T-cell function, supporting the concept of a hypoxia-driven immunosuppressive microenvironment. 36 In addition, myeloid cells actively promote the transition toward an MES-like phenotype. 35 Thus, diverse microenvironmental landscapes across GBM tumors help drive lineage divergence and ­cellular heterogeneity. Redefining Traditional Glioblastoma Niches with Cell State Organization Historically, GBM architecture has been described through 3 microanatomical compartments, or spatial niches: perivascular, hypoxic, and invasive. These niches are both functionally and morphologically distinct, creating specialized microenvironments that contribute to the tumor’s cellular heterogeneity. The perivascular niche is characterized by abnormal vasculature and supports tumor growth, perivascular migration, and the maintenance of stem-like populations. 37 High-grade gliomas are marked by hypoxic necrotic cores surrounded by hypercellular pseudopalisades and thrombosed vessels, a hallmark histopathologic feature. 37 At the tumor margins, infiltrating tumor cells and glioma-neuronal synapses form the invasive niche. 38 Importantly, these niches are dynamic and may transition into one another over time. 37 Human spatial transcriptomics and single-cell studies have preserved this tripartite topographic framework while adding a continuum of molecularly defined domains hidden from histopathology. Across multiple studies, these 3 spatial regions may be broadly mapped to traditional tumor histologic architecture: necrosis and hypoxia adjacent regions localize to the hypoxic niche, angiogenic and immune-interacting hubs localize to the perivascular niche, and neurodevelopmental and infiltrated brain zones localize to the invasive niche. Specific GBM cell states are shown to be preferentially enriched within each niche, potentially driving proliferation, invasion, and resistance to treatment. Although the spatial localization of the Neftel subtypes has been well described, the architecture of other cell state frameworks remains understudied. Earlier studies defined GSCs as a principal cell type within niches, whereas newer analyses suggest overlap between transcriptional cell states and functionally defined stem-like cells. How GSCs exist within the cell state hierarchy remains uncertain, underscoring a conceptual gap between earlier models and contemporary single-cell interpretations of tumor heterogeneity. Perivascular Niche Serving as the proliferative hub of GBM, the perivascular niche is characterized by rapid angiogenesis, driven by enhanced vascular endothelial growth factor (VEGF), fibroblast growth factor (FGF), and platelet-derived growth factor (PDGF). 37 , 39 These pro-angiogenic factors are secreted by CD133 + GSCs located along the vasculature. 40 VEGF and related factors induce vascular expansion through pericyte proliferation and degradation of the vascular basement membrane, resulting in disorganized endothelial and pericyte architecture and the formation of leaky, dysfunctional, or thrombosed vessels. 37 In parallel, recruitment of immature endothelial cells and pericytes further supports neovascularization. 41 Beyond endothelial and pericyte populations, the perivascular niche is enriched with nonmalignant stromal cells including cancer-associated fibroblasts (CAFs) and various immune cells such as neutrophils, macrophages, and myeloid-derived suppressor cells. 42 Tumor-associated macrophages (TAMs), which exist along a continuum of phenotypes beyond the classical M1/M2 dichotomy in GBM, are often found in close proximity to GSCs. 43 Analysis of this spatial association has suggested that stem-like cells may actively recruit TAMs via their secretion of extracellular matrix (ECM) proteins. 44 Once recruited, macrophages and monocyte progenitors promote local immunosuppression through factors such as TGF-β, which also enhances GSC invasion. 45 Notably, TAM density positively correlates with tumor grade, underscoring their potential role in sustaining a supportive microenvironment for GSCs and other plastic tumor cell states. This niche, formed by the vasculature and infiltrating nonmalignant cells, is further suggested to be enriched with MES-like and OPC-like populations. In human recurrent GBM, MES-like and OPC-like GBM cells have been shown to cluster along blood vessels within the perivascular niche, mimicking the vasculature-dependent behavior of normal neural stem cells. 16 Among these, MES-like cells promote rapid angiogenesis through VEGF signaling, potentially expanding the perivascular microenvironment and reinforcing the MES-like state. VEGF production has been attributed to MES-like cells associated with CAFs in human spatial datasets and mouse models. 16 , 46 Functional GSCs also reside within this niche, often situated adjacent to endothelial cells and possibly integrated with MES-like and OPC-like populations. 47 At the interface between the perivascular niche and hypoxic regions, transcriptionally hybrid cells emerge, exhibiting features of both MES1-like and AC-like states in spatially resolved multi-omics of human tissue. 48 These hybrid populations may represent either true intermediate states or spatially co-localized mixtures of MES1-like and AC-like cells. Typically found near vasculature bordering necrotic regions, these cells may coexist with MES-like cells displaying fibroblast-like properties and active VEGF signaling. The MES-like program itself is subclassified into MES1 and MES2 states: MES1 cells are implicated in hybridization with AC-like cells and may be more invasive, while MES2 cells are more strongly associated with a fibroblast-like transcriptional signature, VEGF production, and hypoxia. Analysis of human primary and recurrent tumors indicates that tissues containing MES1/AC-like hybrid populations show elevated fatty acid biosynthesis, suggesting a highly proliferative and invasive phenotype, which correlates with worse patient prognosis. 49 A similar hybrid phenotype was described by Pichol-Thievend et al., who termed it “vessel co-opting and resistant” state (VC-resist) in treatment-refractory GBM. 50 In addition, OPC-like cells are frequently found near pseudopalisading regions. Interestingly, while tumor-associated OPC-like cells are localized to these areas, canonical OPCs typically avoid such environments as hypoxia induces premature differentiation. Onubogu et al.identified a collagen-rich perivascular rim in human recurrent GBM formed by DCN-positive fibroblast-like cells. 16 This structure inhibits VEGF signaling and angiogenesis, exerting a localized antitumoral effect. However, spatial transcriptomic analysis revealed that MES-like tumor cells cluster just outside this decorin-expressing rim, beyond its antiproliferative reach. 16 This spatial exclusion could allow MES-like cell expansion, suggesting that the rim’s limited diffusion radius allows adjacent pro-tumoral growth. Hypoxic Niche The dysfunctional, clot-prone vessels formed in the perivascular niche eventually fail to meet the metabolic demands of the growing tumor mass, leading to the hypoxic niche in GBM. 37 This loss of perfusion is driven by the collapse of enlarged vessels, endothelial cell apoptosis, and vascular regression. 37 Thrombosed vessels commonly found in hypoxic niches suggest that the structurally abnormal vasculature generated in the perivascular niche is ultimately unsustainable. As oxygen supply diminishes, GBM cells undergo apoptosis and reorganize their elongated nuclei into circular, layered patterns known as palisades around necrotic foci. 37 This spatial arrangement likely reflects a migratory response of cells attempting to escape the necrotic core. Hypoxic regions are not simply byproducts of vascular failure but rather actively contribute to tumor progression by fostering invasion, resistance to therapy, and maintenance of quiescent populations. These areas are characterized by upregulation of hypoxia-inducible factors, which in turn drive the expression of VEGF and other survival and invasion-associated factors. 51 Functional GSCs often localize to hypoxic regions, either embedded within or adjacent to necrotic cores, where they may enter a dormant state protected from both hypoxia-induced stress and therapeutic insult. 37 , 47 , 51 HIF activation further reinforces their stem-like phenotype and proliferation potential. The hypoxic niche could serve as a reservoir for therapy-resistant GSCs, which may later migrate to repopulate the tumor and drive recurrence. Some evidence suggests that GSCs in the perivascular niche may supply those in hypoxic areas via chemotactic signals driven by hypoxia. Meanwhile, dying cells within necrotic cores release inflammatory factors that recruit immune cells and skew them toward an immunosuppressive state. 52 Dormant macrophages and CD8 + T cells have been observed to accumulate in these hypoxic cores in syngeneic mouse models, where the local environment dampens their cytotoxic functions. 52 Notably, human single-cell data suggest that macrophages in hypoxia may undergo differentiation, possibly to adapt to oxygen deprivation and support neovascularization. 34 The cellular composition of the necrotic core remains poorly understood, though one study suggests that this region may be enriched with injured or dormant NPC-like cells in mouse models. 33 IFN-γ secreted by infiltrating T cells influences the dormancy of these cell states within the hypoxic niche. 33 Surrounding the necrotic core, peri-necrotic palisades have the highest levels of VEGF expression, indicating that MES-like or AC-MES hybrid cells may localize here in response to hypoxia. 37 These palisading regions, which overlap spatially with elements of the perivascular niche, may contain similar cellular populations, including OPC-like cells bordered by MES-like cells from both niches. This spatial proximity between OPC-like and MES-like cells in patient samples has been associated with a poorer prognosis. 53 The hypoxic gradient from the necrotic core through the palisades and adjacent hypoxic zones and toward the tumor margins has been identified as a key driver of layered cellular organization in human GBM. 11 In contrast, IDH-mutant tumors, which often lack this hypoxic gradient, exhibit a more diffuse distribution of cell states and tend to show less resistance to therapy compared to IDH-wild-type GBMs. 11 Ultimately, the hypoxic niche not only fosters a protective environment for accumulating stem-like cells but also promotes the adoption of more invasive phenotypes across multiple tumor cell states. Invasive Niche GBM cells representing at least 3 developmental states (AC-like, OPC-like, and NPC-like) may escape the hypoxic niche, which predominates at the tumor core, by migrating through the perivascular niche toward the invasive niche at the tumor margins. Within the perivascular niche, astrocytes and endothelial cells support the migration of GSCs while helping maintain their stem-like properties in biomimetic microfluidic tumor-microenvironment systems. 54 AC-like migration tracts provide a pathway for these stem-like cells to infiltrate marginal tumor regions and spread into adjacent healthy brain tissue. 54 In response to hypoxia, GBM cells often undergo a metabolic shift in vitro from the pentose phosphate pathway to glycolysis, which may promote migration toward more favorable microenvironments. 55 MES-like cells, in contrast to other cell states, may migrate independently of the vasculature, making them particularly more effective at driving invasion. 11 This vasculature-independent mobility suggests that transition into an MES-like phenotype may be a key step in establishing the invasive niche. In addition, microglia, similar to TAMs in the perivascular and hypoxic niches, also migrate to tumor margins in human tissue. 34 Once there, they may contribute to shaping the invasive niche by recruiting additional immune cells through cytokine secretion into the surrounding healthy brain tissue. AC-like cells at the tumor margins within the invasive niche can be subdivided into 2 distinct states: AC1 and AC2. The AC2 subtype closely resembles outer radial glial cells, a developmental cell type that may arise from migrating stem-like cells originating in the perivascular niche, as these cells also express stemness markers in patient tissue and organoid models. 56 , 57 As these cells travel along AC-like migration tracts, they may adopt radial glial-like properties while retaining cellular plasticity. These outer radial glial-like cells are capable of further differentiation into both neuronal and astrocytic lineages. 56 , 57 Recent advances in cancer neuroscience have identified glioma populations capable of forming synaptic connections with neurons, a phenomenon linked to enhanced tumor growth. 58 , 59 These synaptogenic tumor cells are thought to arise from AC-like radial glial subtypes, using their plasticity to integrate into neural circuits in mouse xenografts. 60 Neuronal synaptic and paracrine signaling in turn promotes the proliferation of these malignant cells. 60 , 61 At the tumor margins, AC-like cells in somatic mouse models demonstrate resistance to temozolomide (TMZ) through elevated expression of glutathione S-transferases (GSTs), enzymes involved in detoxification. 33 These cells appear to form 2 interrelated populations: AC1 cells, which exhibit a highly infiltrative phenotype, and AC2 cells, which are more integrated into neuronal networks that support tumor growth. In response to treatment, AC-like cells may transition between these 2 states, shifting toward AC1 to evade surgical resection via infiltration into healthy brain tissue and reverting to AC2 to establish synaptic networks that drive tumor recurrence and expansion. While many studies associate invasion with AC-like populations, the Winkler group and others propose an alternative model of invasion. In this alternative model, motile OPC-like and NPC-like cells at the infiltrative front can migrate as single cells through brain parenchyma with neurite-like microtubule extensions. 62 Furthermore, tumor cells migrating along white matter tracts have been shown to adopt a quiescent OPC-like phenotype, supporting a proneural model of invasion. 63 This promotion of quiescence may act to slow rapid tumor expansion, consistent with observations that inhibiting white matter degeneration can slow growth. 64 These findings suggest that GBM invasion is not uniform but is dynamically guided by distinct signaling and structural contexts. Consequently, regional variance between gray and white matter further stratifies the invasive edge, giving rise to parallel invasion modes along gray versus white matter within the same tumor, where different cell states may exhibit different behaviors. Dynamic Niches Comprise a Layered Global Structure As previously discussed, expansile tumor growth in GBM, driven by both proliferation and invasion, fosters dynamic interactions between the tumor’s spatial niches, where each niche can give rise to another in a continuous cycle. 37 For example, unchecked malignant growth within the perivascular niche can lead to regions of poor perfusion, resulting in hypoxia and subsequent necrosis. In response to increasing hypoxic stress, some tumor cells undergo apoptosis and contribute to the formation of necrotic cores, while others adapt through MES-like transitions, allowing them to migrate toward healthier brain regions. These more resilient, migratory cells establish the invasive niche at the tumor margins. This niche cycling not only drives tumor expansion but also promotes the survival and proliferation of highly plastic cells that resist treatment. As the tumor grows outward, these niches may layer concentrically, creating a layered global structure. Greenwald et al. describe this organization as a multilayered system of cell states shaped by a hypoxia gradient based on integrative spatial analysis of human samples. 11 Their model divides the tumor into 5 spatial layers: a hypoxic/necrotic core, a hypoxic-adjacent zone, an angiogenesis/immune hub, neurodevelopmental layers, and infiltrated brain tissue. 11 The stratification resembles the traditional niche organization of GBM in which the hypoxic and hypoxic-adjacent layers correspond to the hypoxic niche. The angiogenesis and immune hub align with the perivascular niche. Lastly, the neurodevelopmental and infiltrated brain layers represent the invasive niche. Together, these layers illustrate how local microenvironments may guide tumor organization, proliferation, and therapeutic resistance. The spatial organization of distinct GBM cell states could drive the formation of layered tumor architecture through their specialized functions within particular niches. For example, MES-like cells in the perivascular niche promote aberrant angiogenesis, resulting in dysfunctional vasculature. These vessels fail to adequately perfuse adjacent highly proliferative OPC-like cells, ultimately leading to hypoxia and the formation of necrotic cores. Within this emerging hypoxic niche, malignant cells may adopt a dormant GSC, NPC-like, or migratory MES-like phenotype to survive or escape hypoxic conditions. Migratory MES-like cells can then move toward the tumor margins, where they interact with AC-like populations in the invasive niche. This interaction may induce reactive astrogliosis in AC-like cells, prompting their conversion into MES-like cells as suggested by in vitro coculture studies. 12 , 65 These newly transformed MES-like cells at the invasive edge may then promote angiogenesis, transforming the invasive edge into a secondary perivascular niche with expanded vascularization. This continuous reorganization of niche environments by migrating and transforming cell states may explain the layered structure described by Greenwald et al. 1 where niches are stacked concentrically within the tumor. Through this niche cycling process ( Figure 1 ), GBM may maintain both its expansive growth and a protective architecture, one that shields stem-like cells in the hypoxic core while enabling outward progression through plastic and invasive cell states. Figure 1. Open in a new tab Dynamic niche cycling drives a layered glioblastoma architecture. A dysfunctional perivascular niche supports VEGF-expressing AC-like cells that promote aberrant angiogenesis. Resulting hypoxia leads to necrotic core formation and induces MES-like transformation and migration toward oxygen-rich regions. At the invasive edge, MES-like cells trigger reactive astrogliosis, converting AC-like cells into pro-angiogenic MES-like states that seed secondary perivascular niches. This niche cycling establishes a layered tumor structure, preserving GSCs in hypoxic cores while enabling outward progression through plastic, treatment-resistant states. Created in BioRender. Trippett, J. (2026) https://BioRender.com/i7apum0 . Cell State Populations Associate to Form Networks During tumor expansion, GBM proliferation can occur in 2 distinct spatial patterns. Computational reconstructions of human tumor evolution demonstrate that growth may occur as discrete masses of a single dominant cell state or as intermixed masses containing multiple cell states. 13 These growth patterns may give rise to what are termed “state-specific” and “state-state” clusters in the spatial transcriptome of human GBM. In state-specific clusters, tumor cells predominantly associate with others of the same transcriptional state ( Figure 2 ). 11 For example, the invasive niche is often dominated by AC-like cells that form tight clusters with phenotypically similar cells. This specific clustering of AC-like populations in patient histology has been linked to worse patient prognosis, likely due to their enhanced capacity for marginal invasion and treatment resistance. 53 In contrast, when AC-like cells are more dispersed or intermixed with other cell states, outcomes tend to improve. 53 One proposed mechanism for the poor prognosis associated with AC-like clustering involves their ability to form GFAP+ tumor microtubules (TMs), which establish multicellular anatomical networks that protect against cell death. 53 These networks are reinforced by connexin-43 (Cx43) gap junctions, allowing for the exchange of growth factors and signaling molecules that support cell survival and proliferation. 58 These networks have been shown to facilitate calcium wave propagation through TMs, protecting them from cell death in mouse xenografts. 66 Gap junctions between astrocytes and tumor cells enable mitochondrial exchange that fuels oxidative metabolism. 67 Connexin further exhibits state-specific expression with connexin-46 in stem-like cells, while more differentiated populations expressed Cx43 in similar models. 68 Interestingly, bipolar GBM cells with 2 TMs demonstrate greater invasiveness compared to those with 4 or more, suggesting that cells with fewer connections may be more migratory, while more connected cells remain stationary to maintain supportive networks. 12 Figure 2. Open in a new tab State-specific and state-state clustering networks in glioblastoma. State-specific clusters of AC-like cells form GFAP+ microtubule networks linked by connexin gap junctions, with bipolar cells driving invasion and multipolar cells anchoring stable hubs. OPC-like cells align along axonal tracts, forming invasive tendrils at the tumor periphery. TAMs aggregate in necrotic zones, establishing immunosuppressive microenvironments that protect GSCs. Activated microglia integrate into AC-like networks, potentially extending immune suppression toward the invasive edge. These coordinated state-state and state-specific clusters enable structural support, immune modulation, and directional invasion. Created in BioRender. Trippett, J. (2026) https://BioRender.com/48aft2g . This pattern of state-specific clustering also applies to nonmalignant cell populations. For example, TAMs and microglia, particularly TAMs, tend to cluster within their own lineages, which form distinct immunosuppressive niches, particularly within hypoxic and invasive regions. 11 , 34 , 69 Such clustering aids in sustaining the pro-tumoral tumor microenvironment, especially in protecting GSCs within the hypoxic core of both mouse and human tissue. 11 , 34 , 52 , 69 In contrast, state-state clusters represent regions where different cell states co-localize and interact closely. 11 Human spatial analyses have shown proximity between MES-like cells and AC-like cells as well as between OPC-like and MES-hypoxia cells, both of which are associated with worse clinical outcomes. 48 , 53 The biological implications of these mixed clusters remain poorly understood, but they may reflect microenvironments that favor cooperative behavior among distinct cell types. Certain microenvironmental factors may allow for the synergistic favoring of multiple cell states, while others favor only one. Importantly, “state-specific” and “state-state” clusters may interchange over time as migrating malignant cells establish themselves in new niches, adapt to local environmental cues, and compete. Treatment-Protected Cell States Remodel Tumor Architecture The spatial organization of cell states within distinct tumor niches may confer protection against both chemoradiotherapy and surgical resection. For example, the hypoxic niche may support a quiescent GSC population that resists chemoradiotherapy as these treatments are less effective against non-proliferative cells in the G 0 phase. These dormant GSCs can survive initial therapy and later reactivate, leading to recurrence. In the oxygen-rich perivascular niche, GSCs and MES-like cells organized along vasculature may evade treatment through migration and resistance to TMZ, which is hydrolyzed less effectively at the neutral pH (7 to 7.4) typical of this microenvironment. In matched human primary-recurrent tumors, MES-like cells have been observed to express more pronounced stem-like state properties than in primary tumors. 70 These recurrent human tumors also show enrichment of MES-like cells within microvascular regions, where they may promote angiogenesis and further tumor spread. 71 AC-like and MES-like cells within the invasive niche are uniquely protected against the 3 treatment modalities used against GBM. First, cells in the invasive niche avoid surgical resection, which removes grossly abnormal tissue but unfortunately is unable to safely access invasive niches harbored microscopically in otherwise surrounding healthy brain tissue. Second, the invasive niche often extends beyond areas of blood-brain barrier disruption into regions with poor chemotherapy penetration. Third, the invasive niche is also typically outside the typical 2 cm margin of MRI gadolinium enhancement targeted by radiotherapy. In this manner, spatially organized niches may enable cell populations that are resistant to treatment to persist, supporting recurrence. Transcriptomic and proteogenomic analyses of human recurrent GBMs reveal increased expression of neuronal, oligodendrocyte, and ECM genes, many of which are associated with MES-like cells. 72 , 73 Chemoradiotherapy may inadvertently promote this MES-like phenotype in surviving tumor cells, leading to angiogenesis and tumor growth. In recurrent tumors, MES-like populations have been shown to employ vessel hijacking processes such as vessel co-option and vasectasia to expand and remodel the vasculature. Spinelli et al. report that MES-like tumors are characterized by dilated vessels formed through vasectasia rather than a dense network of smaller angiogenic-type blood vessels in mouse xenografts. 74 In addition, vessel co-option causes the involution of co-opted vessels, inducing local hypoxia and further promotion of an angiogenic phenotype ( Figure 3 ). 75 Consistent with these mechanisms, Stadlbauer et al. observed that human recurrent tumors demonstrate a 170% increase in vessel density and a 120% increase in oxygen perfusion compared to primary tumors. 76 The predominance of MES-like cells in recurrent tumors allows for rapid vascular remodeling, regrowth, and faster relapse of paired human samples. 77 In addition to vascular changes, recurrent tumors show elevated markers of neural and synaptic activity. 73 The upregulation of neuronal and oligodendrocyte markers suggests increased glioma-neuronal integration at tumor margins. These findings raise the possibility that recurrent tumors may also harbor elevated populations of NPC-like and OPC-like cells, potentially driven by or contributing to glioma-neuron network formation. In contrast to findings proposing MES-like dominance at recurrence, Spitzer et al. observe that matched patient recurrent tumors retain AC-like, OPC-like, NPC-like, or hybrid transcriptional states. 9 This suggests that recurrence represents a heterogeneous adaptive process where diverse states may emerge depending on each tumor’s spatial and cellular identities. Divergent sampling time points may also explain these differences, as early recurrence may capture more neurodevelopmental-like states that precede mesenchymal programs associated with tumor expansion. Further research is needed to elucidate the role of these neurodevelopmental states in GBM recurrence. Figure 3. Open in a new tab Treatment-induced vascular remodeling in recurrent glioblastoma. Recurrent glioblastoma displays co-option of preexisting vessels, followed by involution and localized hypoxia. Sprouting angiogenesis and vasectasia driven by VEGF-secreting MES-like cells generate structurally abnormal, dilated vasculature. TAMs remodel the matrix to support vessel formation, while perivascular fibroblasts potentiate endothelial expansion. A fibrotic scar rich in CAFs reinforces angiogenic signaling. Hypoxic niches persist beneath these structures, preserving dormant, treatment-resistant GSCs that fuel relapse. Created in BioRender. Trippett, J. (2026) https://BioRender.com/mivm5dp . Conclusions The transcriptomic heterogeneity revealed by single-cell and spatial studies reflects the continuous impact of the tumor microenvironment on cellular identity. The spatial organization of niches with their respective cues establishes gradients of stress and support that influence the transcriptional programs observed across the tumor. In this framework, we propose that cellular states represent functional equilibria imposed by their local microenvironment rather than arising from defined sets of lineages. Within this model, GSCs exist as a functional phenotype of the same regulatory network. Defining aspects of GSCs, including self-renewal, quiescence, and therapeutic resistance, may emerge from various transcriptional states when niche conditions favor the necessary context for stem-like properties. Prior models of GSCs could exist within the continuum of cell states where the microenvironment is a fundamental driver of both cellular diversity and plasticity. Although single-cell technologies have mapped this diversity, they capture a snapshot of transcriptomic states, offering limited insight into the functional mechanisms that sustain them. Future studies must incorporate functional models of cell states to assess how the microenvironment stabilizes or redirects specific transcriptional programs over tumor evolution. Therapeutically, these advances suggest that interventions directed against cell states may fail to yield durable responses, as cell states are transient outcomes of their environmental context. While strategies such as ZNF117 downregulation to promote OPC-like states or AP1 inhibition to reverse MES-like phenotypes show promise, the durability and extent of lineage commitment achieved with these approaches remain unclear. 78 , 79 Under stress, GBM cell states often revert to a stem-like phenotype, a protective mechanism also observed in normal neural biology such as in reactive astrocytes. 80 Given this plasticity, targeting the tumor microenvironment rather than individual cell states may offer a more promising approach. Transforming heterogeneous niches into a more homogenous state susceptible to treatment could deplete source populations responsible for recurrence. Disrupting the spatial organization that supports lethal cell states may render them vulnerable to lineage commitment and reduce their ability to revert to stem-like phenotypes. Thus, targeting the niche itself may encourage more stable lineage commitment without triggering reactive plasticity. Experimental systems further differ in spatial complexity and often produce varying hierarchies of cell states. While human tissue studies capture the full extent of clinical heterogeneity and evolutionary complexity, mouse models provide mechanistic validation yet lack the diversity inherent to the human disease. Additional methodological differences such as those between single-cell and single-nucleus sequencing, can also lead to divergent interpretations. These discrepancies highlight the need for standardized and integrative approaches in this context-dependent field. As analytical definitions evolve, conceptual frameworks must remain flexible to describe the relationship between transcriptional plasticity and microenvironmental regulation that drives GBM plasticity. Contributor Information Pranav Prakash, Department of Neurosurgery, University of California, San Francisco (UCSF), San Francisco, CA, USA. James Trippett, Department of Neurosurgery, University of California, San Francisco (UCSF), San Francisco, CA, USA. Cameron Ehsan, Department of Neurosurgery, University of California, San Francisco (UCSF), San Francisco, CA, USA. Joseph Namkung, Department of Neurosurgery, University of California, San Francisco (UCSF), San Francisco, CA, USA. Meeki Lad, Department of Neurosurgery, University of California, San Francisco (UCSF), San Francisco, CA, USA. Manish K Aghi, Department of Neurosurgery, University of California, San Francisco (UCSF), San Francisco, CA, USA. Author Contributions P.P. conceived of the topic, drafted the manuscript, and edited the manuscript. J.T. designed figures and edited the manuscript. C.E. and J.N. edited the manuscript. M.L. conceived of the topic and edited the manuscript. M.K.A. procured funding for the research team and edited the manuscript. Conflict of Interest Statement P.P., J.T., C.E., J.N., M.L., and M.K.A. report no conflicts of interest. Funding M.K.A. was supported by the National Institutes of Health grants R01CA227136, R01NS079697, and R01CA260443. Data Availability No original data were produced or examined in this research. References 1. Wirsching HG, Galanis E, Weller M.  Glioblastoma. Handb Clin Neurol. 2016;134:381–397. 10.1016/B978-0-12-802997-8.00023-2 [ DOI ] [ PubMed ] [ Google Scholar ] 2. Davis ME.  Glioblastoma: overview of disease and treatment. Clin J Oncol Nurs. 2016;20:S2–S8. 10.1188/16.CJON.S1.2-8 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 3. Nørøxe DS, Poulsen HS, Lassen U.  Hallmarks of glioblastoma: a systematic review. ESMO Open. 2016;1:e000144. 10.1136/esmoopen-2016-000144 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 4. Torrisi F, Alberghina C, D’Aprile S, et al.  The hallmarks of glioblastoma. Biomedicines. 2022;10:806. 10.3390/biomedicines10040806 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 5. Eisenbarth D, Wang YA.  Glioblastoma heterogeneity at single-cell resolution. Oncogene. 2023;42:2155–2165. 10.1038/s41388-023-02738-y [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 6. Prager BC, Xie Q, Bao S, Rich JN.  Cancer stem cells: the architects of the tumor ecosystem. Cell Stem Cell. 2019;24:41–53. 10.1016/j.stem.2018.12.009 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 7. Neftel C, Laffy J, Filbin MG, et al.  An integrative model of cellular states. Cell. 2019;178:835–849.e21. 10.1016/j.cell.2019.06.024 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 8. Garofano L, Migliozzi S, Oh YT, et al.  Pathway-based classification of glioblastoma uncovers a mitochondrial subtype with therapeutic vulnerabilities. Nat Cancer. 2021;2:141–156. 10.1038/s43018-020-00159-4 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 9. Spitzer A, Johnson KC, Nomura M, et al.  Deciphering the longitudinal trajectories of glioblastoma ecosystems by integrative single-cell genomics. Nat Genet. 2025;57:1168–1178. 10.1038/s41588-025-02168-4 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 10. Johnson AL, Laterra J, Lopez-Bertoni H.  Exploring glioblastoma stem-cell heterogeneity. Front Oncol. 2022;12:995498. 10.3389/fonc.2022.995498 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 11. Greenwald AC, Darnell NG, Hoefflin R, et al.  Integrative spatial analysis reveals a multi-layered organization of glioblastoma. Cell. 2024;187:2485–2501.e26. 10.1016/j.cell.2024.03.029 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 12. De Silva MI, Stringer BW, Bardy C.  Neuronal and tumourigenic boundaries of glioblastoma plasticity. Trends Cancer. 2023;9:223–236. 10.1016/j.trecan.2022.10.010 [ DOI ] [ PubMed ] [ Google Scholar ] 13. Mathur R, Wang Q, Schupp PG, et al.  Glioblastoma evolution and heterogeneity. Cell. 2024;187:446–463.e16. 10.1016/j.cell.2023.12.013 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 14. Chanoch-Myers R, Wider A, Suva ML, Tirosh I.  Elucidating the diversity of malignant mesenchymal states. Genome Med. 2022;14:106. 10.1186/s13073-022-01109-8 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 15. Varn FS, Johnson KC, Martinek J, et al. ; GLASS Consortium. Glioma progression is shaped by genetic evolution. Cell. 2022;185:2184–2199.e16. 10.1016/j.cell.2022.04.038 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 16. Onubogu U, Gatenbee CD, Prabhakaran S, et al.  Spatial analysis of recurrent glioblastoma reveals perivascular niche organization. JCI Insight. 2024;9:e179853. 10.1172/jci.insight.179853 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 17. Carén H, Pollard SM, Beck S.  Epigenetic mechanisms in glioblastoma. Mol Aspects Med. 2013;34:849–862. 10.1016/j.mam.2012.06.007 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 18. Cadieux B, Ching TT, VandenBerg SR, Costello JF.  Genome-wide hypomethylation in human glioblastomas. Cancer Res. 2006;66:8469–8476. [ DOI ] [ PubMed ] [ Google Scholar ] 19. Chaligne R, Gaiti F, Silverbush D, et al.  Epigenetic encoding, heritability and plasticity of glioma transcriptional cell states. Nat Genet. 2021;53:1469–1479. 10.1038/s41588-021-00927-7 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 20. Moffet JJD, Fatunla OE, Freytag L, et al.  Spatial architecture of high-grade glioma reveals tumor heterogeneity within distinct domains. Neurooncol Adv. 2023;5:vdad142. 10.1093/noajnl/vdad142 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 21. Sottoriva A, Spiteri I, Piccirillo SG, et al.  Intratumor heterogeneity in human glioblastoma reflects cancer evolutionary dynamics. Proc Natl Acad Sci U S A. 2013;110:4009–4014. 10.1073/pnas.1219747110 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 22. Ceresa D, Alessandrini F, Lucchini S, et al.  Early clonal extinction in glioblastoma progression revealed by genetic barcoding. Cancer Cell. 2023;41:1466–1479.e9. 10.1016/j.ccell.2023.07.001 [ DOI ] [ PubMed ] [ Google Scholar ] 23. Vue TY, Kollipara RK, Borromeo MD, et al.  ASCL1 regulates neurodevelopmental transcription factors. Glia. 2020;68:2613–2630. 10.1002/glia.23873 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 24. Liu ZH, Dai XM, Du B.  Hes1: a key role in stemness, metastasis and multidrug resistance. Cancer Biol Ther. 2015;16:353–359. 10.1080/15384047.2015.1016662 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 25. Szu JI, Tsigelny IF, Wojcinski A, Kesari S.  Biological functions of the Olig gene family. Front Neurosci. 2023;17:1129434. 10.3389/fnins.2023.1129434 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 26. Sun Y, Meijer DH, Alberta JA, et al.  Phosphorylation state of OLIG2 regulates proliferation. Neuron. 2011;69:906–917. 10.1016/j.neuron.2011.02.005 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 27. Singh SK, Fiorelli R, Kupp R, et al.  Post-translational modifications of OLIG2 regulate glioma invasion. Cell Rep. 2016;16:950–966. 10.1016/j.celrep.2016.06.045 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 28. Buruiană A, Florian ȘI, Florian AI, et al.  Roles of miRNA in glioblastoma tumor-cell communication. Int J Mol Sci. 2020;21–1950. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 29. Bassot A, Dragic H, Haddad SA, et al.  Identification of a miRNA multi-targeting therapeutic strategy. Cell Death Dis. 2023;14:630. 10.1038/s41419-023-06117-z [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 30. Sana J, Busek P, Fadrus P, et al.  MicroRNAs differentially expressed in glioblastoma stem-like cells. Sci Rep. 2018;8:2836. 10.1038/s41598-018-20929-6 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 31. Xu B, Mei J, Ji W, et al.  MicroRNAs involved in the EGFR pathway in glioblastoma. Biomed Pharmacother. 2021;134:111115. 10.1016/j.biopha.2020.111115 [ DOI ] [ PubMed ] [ Google Scholar ] 32. Sharma P, Aaroe A, Liang J, Puduvalli VK.  Tumor microenvironment in glioblastoma. Neurooncol Adv. 2023;5:vdad009. 10.1093/noajnl/vdad009 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 33. Garcia-Diaz C, Pöysti A, Mereu E, et al.  Glioblastoma cell fate is differentially regulated. Cell Rep. 2023;42:112472. 10.1016/j.celrep.2023.112472 [ DOI ] [ PubMed ] [ Google Scholar ] 34. Haley MJ, Bere L, Minshull J, et al.  Hypoxia coordinates the spatial landscape of myeloid cells. Sci Adv. 2024;10:eadj3301. 10.1126/sciadv.adj3301 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 35. Hara T, Chanoch-Myers R, Mathewson ND, et al.  Cancer-cell/immune-cell interactions drive transitions to mesenchymal-like states in glioblastoma. Cancer Cell. 2021;39:779–792.e11. 10.1016/j.ccell.2021.05.002 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 36. Ravi VM, Neidert N, Will P, et al.  T-cell dysfunction in the glioblastoma microenvironment. Nat Commun. 2022;13:925. 10.1038/s41467-022-28523-1 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 37. Hambardzumyan D, Bergers G.  Glioblastoma: defining tumor niches. Trends Cancer. 2015;1:252–265. 10.1016/j.trecan.2015.10.009 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 38. Cornelison RC, Yuan JX, Tate KM, et al.  A patient-designed tissue-engineered model of the infiltrative glioblastoma microenvironment. NPJ Precis Oncol. 2022;6:54. 10.1038/s41698-022-00290-8 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 39. Mosteiro A, Pedrosa L, Ferrés A, et al.  The vascular microenvironment in glioblastoma. Biomedicines. 2022;10:1285. 10.3390/biomedicines10061285 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 40. Calabrese C, Poppleton H, Kocak M, et al.  A perivascular niche for brain-tumor stem cells. Cancer Cell. 2007;11:69–82. 10.1016/j.ccr.2006.11.020 [ DOI ] [ PubMed ] [ Google Scholar ] 41. Sattiraju A, Mintz A.  Pericytes in glioblastomas: multifaceted role. Adv Exp Med Biol. 2019;1147:65–91. 10.1007/978-3-030-16908-4_2 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 42. Charles N, Holland EC.  The perivascular niche microenvironment in brain-tumor progression. Cell Cycle. 2010;9:3012–3021. 10.4161/cc.9.15.12710 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 43. Qian Y, Yin Y, Zheng X, et al.  Metabolic regulation of tumor-associated macrophage heterogeneity. Biomark Res. 2024;12:1. 10.1186/s40364-023-00549-7 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 44. Aramini B, Masciale V, Grisendi G, et al.  Cancer stem cells and macrophages: molecular connections. Oncotarget. 2021;12:230–250. 10.18632/oncotarget.27870 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 45. Caverzán MD, Beaugé L, Oliveda PM, et al.  Exploring monocytes-macrophages in the immune microenvironment of glioblastoma. Brain Sci. 2023;13:542. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 46. Zarodniuk M, Steele A, Lu X, et al.  Perivascular fibroblasts predict immunotherapy resistance in glioblastoma. NPJ Genom Med. 2023;8:35. 10.1038/s41525-023-00381-w [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 47. Sahoo O, Mitra R, Nagaiah N.  The hidden architects of glioblastoma multiforme: glioma stem cells. MedComm Oncol. 2024;3. 10.1002/mog2.66 [ DOI ] [ Google Scholar ] 48. Ravi VM, Will P, Kueckelhaus J, et al.  Spatially resolved multi-omics deciphers bidirectional tumor–host interdependence in glioblastoma. Cancer Cell. 2022;40:639–655.e13. 10.1016/j.ccell.2022.05.009 [ DOI ] [ PubMed ] [ Google Scholar ] 49. Al-Dalahmah O, Argenziano MG, Kannan A, et al.  Re-convolving the compositional landscape of primary and recurrent glioblastoma. Nat Commun. 2023;14:2586. 10.1038/s41467-023-38186-1 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 50. Pichol-Thievend C, Anezo O, Pettiwala AM, et al.  VC-resist glioblastoma cell state: vessel co-option as a key driver of chemoradiation resistance. Nat Commun. 2024;15:3602. 10.1038/s41467-024-47985-z [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 51. Choudhry H, Harris AL.  Advances in hypoxia-inducible factor biology. Cell Metab. 2018;27:281–298. 10.1016/j.cmet.2017.10.005 [ DOI ] [ PubMed ] [ Google Scholar ] 52. Sattiraju A, Kang S, Giotti B, et al.  Hypoxic niches attract and sequester tumor-associated macrophages and cytotoxic T cells. Immunity. 2023;56:1825–1843.e6. 10.1016/j.immuni.2023.06.017 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 53. Zheng Y, Carrillo-Perez F, Pizurica M, et al.  Spatial cellular architecture predicts prognosis in glioblastoma. Nat Commun. 2023;14:4122. 10.1038/s41467-023-39933-0 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 54. Adjei-Sowah EA, O’Connor SA, Veldhuizen J, et al.  Microfluidic tumor-microenvironment model of glioma stem-cell interactions. Adv Sci. 2022;9:e2201436. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 55. Kathagen-Buhmann A, Schulte A, Weller J, et al.  Glycolysis and the pentose-phosphate pathway in glioblastoma migration vs proliferation. Neuro Oncol. 2016;18:1219–1229. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 56. Ren Y, Huang Z, Zhou L, et al.  Spatial transcriptomics reveals niche-specific enrichment of radial glial stem-like cells. Nat Commun. 2023;14:1028. 10.1038/s41467-023-36707-6 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 57. Bhaduri A, Di Lullo E, Jung D, et al.  Outer radial glia-like cancer stem cells contribute to heterogeneity of glioblastoma. Cell Stem Cell. 2020;26:48–63.e6. 10.1016/j.stem.2019.11.015 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 58. Mancusi R, Monje M.  The neuroscience of cancer. Nature. 2023;618:467–479. 10.1038/s41586-023-05968-y [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 59. Winkler F, Venkatesh HS, Amit M, et al.  Cancer neuroscience: state of the field. Cell. 2023;186:1689–1707. 10.1016/j.cell.2023.02.002 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 60. Krishna S, Choudhury A, Keough MB, et al.  Glioblastoma remodelling of human neural circuits decreases survival. Nature. 2023;617:599–607. 10.1038/s41586-023-06036-1 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 61. Taylor KR, Barron T, Hui A, et al.  Glioma synapses recruit mechanisms of adaptive plasticity. Nature. 2023;623:366–374. 10.1038/s41586-023-06678-1 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 62. Venkataramani V, Yang Y, Schubert MC, et al.  Glioblastoma hijacks neuronal mechanisms for brain invasion. Cell. 2022;185:2899–2917.e31. 10.1016/j.cell.2022.06.054 [ DOI ] [ PubMed ] [ Google Scholar ] 63. Brooks LJ, Clements MP, Burden JJ, et al.  The white matter is a pro-differentiative niche for glioblastoma. Nat Commun. 2021;12:2184. 10.1038/s41467-021-22225-w [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 64. Clements M, Tang W, Florjanic Baronik Z, et al.  Axonal injury is a targetable driver of glioblastoma progression. Nature. 2025;646:452–461. 10.1038/s41586-025-09411-2 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 65. Iser IC, Lenz G, Wink MR.  EMT-like process in glioblastomas and reactive astrocytes. Neurochem Int. 2019;122:139–143. 10.1016/j.neuint.2018.11.016 [ DOI ] [ PubMed ] [ Google Scholar ] 66. Osswald M, Jung E, Sahm F, et al.  Brain tumour cells interconnect to a functional and resistant network. Nature. 2015;528:93–98. 10.1038/nature16071 [ DOI ] [ PubMed ] [ Google Scholar ] 67. Watson DC, Bayik D, Storevik S, et al.  GAP43-dependent mitochondria transfer from astrocytes enhances glioblastoma tumorigenicity. Nat Cancer. 2023;4:648–664. 10.1038/s43018-023-00556-5 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 68. Mulkearns-Hubert EE, Hajdari N, Hong ES, et al.  Connexin 43 drives glioblastoma cancer stem cell phenotypes through a WNK lysine-deficient protein kinase 1-c-MYC signaling axis. Cell Rep. 2025;44:116303. 10.1016/j.celrep.2025.116303 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 69. Solomou G, Young AMH, Bulstrode HJCJ.  Microglia and macrophages in glioblastoma: landscapes and treatment directions. Mol Oncol. 2024;18:2906–2926. 10.1002/1878-0261.13657 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 70. Drexler R, Khatri R, Schüller U, et al.  Temporal change of DNA-methylation subclasses between matched newly diagnosed and recurrent glioblastoma. Acta Neuropathol. 2024;147:21. 10.1007/s00401-023-02677-8 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 71. Wang X, Sun Q, Wang W, et al.  Decoding key cell sub-populations in glioblastoma at recurrence by single-cell analysis. Acta Neuropathol Commun. 2023;11:125. 10.1186/s40478-023-01613-x [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 72. Li J, Shih LK, Brat DJ.  Integrated proteogenomics uncover mechanisms of glioblastoma evolution. Cancer Res. 2024;84:1379–1381. 10.1158/0008-5472.CAN-24-0452 [ DOI ] [ PubMed ] [ Google Scholar ] 73. Hoogstrate Y, Draaisma K, Ghisai SA, et al.  Transcriptome analysis reveals tumor-microenvironment changes in glioblastoma. Cancer Cell. 2023;41:678–692.e7. 10.1016/j.ccell.2023.02.019 [ DOI ] [ PubMed ] [ Google Scholar ] 74. Spinelli C, Adnani L, Meehan B, et al.  Mesenchymal glioma stem cells trigger vasectasia via extracellular vesicles carrying EGFR. Nat Commun. 2024;15:2865. 10.1038/s41467-024-46597-x [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 75. Seano G, Jain RK.  Vessel co-option in glioblastoma: emerging insights and opportunities. Angiogenesis. 2020;23:9–16. 10.1007/s10456-019-09691-z [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 76. Stadlbauer A, Kinfe TM, Eyüpoglu I, et al.  Tissue hypoxia and microvascular architecture predict glioblastoma recurrence. Clin Cancer Res. 2021;27:1641–1649. 10.1158/1078-0432.CCR-20-3580 [ DOI ] [ PubMed ] [ Google Scholar ] 77. Ho WM, Chen CY, Chiang TW, et al.  Longer time to relapse is associated with greater divergence between paired primary and recurrent IDH-wild-type glioblastomas. Acta Neuropathol Commun. 2024;12:77. 10.1186/s40478-024-01829-5 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 78. Liu J, Wang X, Chen AT, et al.  ZNF117 regulates glioblastoma stem-cell differentiation toward the oligodendroglial lineage. Nat Commun. 2022;13:2196. 10.1038/s41467-022-29884-3 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 79. Wang L, Jung J, Babikir H, et al.  A single-cell atlas of glioblastoma evolution under therapy. Nat Cancer. 2022;3:1534–1552. 10.1038/s43018-022-00475-x [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 80. Simpson Ragdale H, Clements M, Tang W, et al.  Injury primes mutation-bearing astrocytes for dedifferentiation in later life. Curr Biol. 2023;33:1082–1098.e8. 10.1016/j.cub.2023.02.013 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Associated Data This section collects any data citations, data availability statements, or supplementary materials included in this article. Data Availability Statement No original data were produced or examined in this research. Articles from Neuro-Oncology are provided here courtesy of Society for Neuro-Oncology and Oxford University Press ACTIONS View on publisher site PDF (5.5 MB) Cite Collections Permalink PERMALINK Copy RESOURCES Similar articles Cited by other articles Links to NCBI Databases Cite Copy Download .nbib .nbib Format: AMA APA MLA NLM Add to Collections Create a new collection Add to an existing collection Name your collection * Choose a collection Unable to load your collection due to an error Please try again Add Cancel Follow NCBI NCBI on X (formerly known as Twitter) NCBI on Facebook NCBI on LinkedIn NCBI on GitHub NCBI RSS feed Connect with NLM NLM on X (formerly known as Twitter) NLM on Facebook NLM on YouTube National Library of Medicine 8600 Rockville Pike Bethesda, MD 20894 Web Policies FOIA HHS Vulnerability Disclosure Help Accessibility Careers NLM NIH HHS USA.gov Back to Top