High-grade gliomas (HGGs) are the most aggressive gliomas of the central nervous system and remain associated with an extremely poor prognosis despite advances in diagnosis and therapy1,2. Their infiltrative growth, molecular heterogeneity, and plasticity render HGGs among the most challenging solid tumors to treat. Surgical resection remains the cornerstone of treatment, but standard Stupp-protocol therapy provides only modest survival benefit, with median overall survival around 15 months and <5% five-year survival3. In the recurrent setting, available options extend survival by only 6–9 months on average4.
To better capture glioma heterogeneity, adult diffuse glioma classification has shifted from histopathological to integrated histo-molecular frameworks5. In the current WHO Classification of Tumors of the Central Nervous System, gliomas are stratified by combinations of molecular alterations, most notably IDH mutation status and 1p/19q codeletion, with additional alterations refining grading and diagnosis (Table 1)5,6. Although diagnostically essential, this classification does not fully capture the heterogeneity introduced by adaptive cellular states and plasticity. Although stem-like tumor-propagating populations occur across diffuse gliomas, the strongest mechanistic evidence linking GSCs to therapy resistance and DDR plasticity derives from IDH-wildtype glioblastoma, which therefore represents the primary focus of this Review.
Table 1 WHO 2021 histo-molecular classification of adult diffuse gliomas
Similarly, molecular markers, such as MGMT promoter methylation, TERT promoter mutations, and EGFR amplification contribute to prognostic and therapeutic stratification, but do not explain the adaptive mechanisms sustaining therapy resistance, particularly in IDH-wildtype HGGs (Table 2)7,8.
Table 2 Molecular markers with prognostic and predictive value and relevance to treatment response in HGGs
These limitations have intensified interest in alternative therapeutic strategies, including immunotherapy and molecularly targeted approaches9, alongside large-scale molecular profiling efforts that have uncovered recurrent genetic alterations, deregulated signaling pathways, and distinct cellular states driving tumor progression and treatment failure10. Among these, glioma stem cells (GSCs) have emerged as a key tumor-maintaining cellular state characterized by self-renewal capacity, lineage plasticity, and exceptional resistance to genotoxic stress induced by radiotherapy and chemotherapy. Importantly, GSCs represent a dynamic and inducible cellular state that can be regenerated from more differentiated tumor cells under microenvironmental and therapeutic pressure. A defining hallmark of GSCs is extensive rewiring of the DNA damage response (DDR). Rather than simply enhancing DNA repair efficiency, DDR plasticity involves dynamic regulation of checkpoint control, replication stress tolerance, repair pathway usage, chromatin and RNA-based regulation, metabolism, translation, and niche signaling to sustain survival under genotoxic stress. Accordingly, adaptive DDR states support not only lesion processing but also damage tolerance and survival-oriented cell-fate decisions.
In this review, we synthesize current evidence illustrating how GSCs exploit multilayered DDR networks to sustain survival, self-renewal, and therapy resistance in HGGs. By integrating checkpoint signaling, replication stress management, DNA repair pathway choice, RNA processing, epigenetic regulation, translational control, metabolism, and microenvironmental interactions, we outline the molecular logic of DDR plasticity in GSCs and highlight vulnerabilities that may inform mechanism-driven therapies. We also highlight transcriptional recovery after damage as a potential additional dimension of DDR fitness in highly transcriptional GSC states.
Current DNA repair-targeting therapies
Current standard-of-care treatment for HGGs is largely based on DNA-damaging strategies. Radiotherapy primarily induces DNA double-strand breaks and oxidative base damage, whereas TMZ introduces methyl adducts at several DNA positions, including N7-guanine, O3-adenine, and the clinically most relevant O6-guanine11. TMZ cytotoxicity depends on unsuccessful lesion processing, leading to replication stress, strand break accumulation, and apoptosis12. However, resistance to TMZ frequently arises through enhanced DNA repair, damage tolerance, or apoptosis suppression13,14. In the recurrent setting, additional alkylating agents such as lomustine and procarbazine are commonly used, either alone or in combination regimens such as PCV (Procarbazine, Lomustine, and Vincristine). Lomustine, a highly lipophilic chloroethylating agent, readily crosses the blood-brain barrier and induces DNA and RNA alkylation as well as interstrand cross-links at the O6 position of guanine, while procarbazine similarly converges on O6-guanine methylation. Their efficacy is likewise limited by the ability of glioma cells to tolerate and repair alkylation-induced lesions15.
Together, these observations indicate that clinical failure is driven not by insufficient DNA damage induction, but by adaptive DNA damage response and repair programs shaped by clinically relevant molecular markers (Table 2). Importantly, this resilience is enriched within specific tumor cells subpopulations displaying heightened survival plasticity. In this context, glioma stem cells have emerged as central drivers of therapy resistance and tumor recurrence, explaining why DNA-damaging treatments fail to achieve durable disease control.
Glioma stem cells
GSCs are recognized as key drivers of glioma initiation, progression, therapy resistance, and recurrence16. Early studies identified tumor-initiating populations by enriching for CD133⁺ cells; however, CD133⁻ fractions can also retain tumor-initiating capacity, indicating that stemness cannot be defined by a single marker17. Accordingly, multiple surface markers have been associated with GSC populations, including CD4418, CD1519, A2B520, CD9021, integrin α622, CD171/L1CAM23, and EpCAM, whose overexpression identifies highly tumorigenic and therapy-resistant GBM subsets24. Beyond surface phenotypes, stem-like states are defined by transcriptional and epigenetic programs linked to neural progenitor identity and self-renewal, including BMI1, SOX2, MSI1/225, NANOG26, and Nestin27, as well as YAP/TAZ signaling28. Stemness is also highly plastic: OCT4 reactivation in therapy-adapted GBM cells promotes stem-like programs, metabolic rewiring, and stress survival, highlighting the reversibility of the GSC state29. Consistently, a core transcriptional network composed of OCT4, SOX2, SALL2, and OLIG2 is sufficient to reprogram differentiated GBM cells into induced GSCs that retain self-renewal, tumor-initiating capacity, and therapy resistance30. In vivo, GSCs are enriched in anatomical niches such as the subventricular zone, where niche-derived signals support stemness, therapy resistance, and recurrence31.
Together, these findings indicate that GSCs represent a dynamic and plastic state whose adaptability underlies their capacity to survive genotoxic stress, setting the stage for DDR-centered mechanisms of therapy resistance.
Clinical and biological basis of therapy resistance in GSCs
Long before GSCs were molecularly defined, clinical observations suggested that therapeutic response in malignant gliomas could not be explained solely by histology, tumor size, anatomical location, or treatment intensity. In a seminal study, Rosenblum et al. showed that younger patients had significantly longer post-operative survival than older patients despite comparable clinical characteristics, attributing this difference to increased intrinsic chemosensitivity of clonogenic tumor cells to the nitrosourea BCNU rather than to reduced tumor aggressiveness32. Importantly, survival correlated inversely with in vitro clonogenic survival following chemotherapy.
These findings provided early evidence that therapy resistance is encoded within specific tumor cells subpopulations, anticipating the later identification of stem-like, therapy-resistant compartments in HGGs. GSCs can therefore be viewed as the cellular basis of clinically observed heterogeneity in treatment response and disease progression. Subsequent studies showed that GSC resilience reflects not simply enhanced DNA repair, but a highly plastic DDR state in which checkpoint control, replication stress tolerance, and cell fate decisions are rewired to sustain survival under genotoxic stress.
Together, these interconnected mechanisms define a multilayered and adaptive DNA damage response framework underlying therapy resistance in GSCs (Fig. 1). Accordingly, DDR plasticity emerges as a defining hallmark of GSCs and a central driver of therapy resistance in HGG. Recent single-cell and spatial transcriptomic studies further revealed that glioblastoma is composed of dynamic malignant cellular states and spatially organized ecosystems rather than a single fixed cellular hierarchy33,34. Although not specific to GSCs, these findings support a model in which stem-like tumor-maintaining states are heterogeneous, inducible, and context-dependent. Consistently, GSC-focused studies show that stem-like cells occupy distinct transcriptional states and niches, suggesting that DDR plasticity is shaped by transcriptional identity, microenvironmental cues, and therapeutic pressure35,36. Accordingly, the mechanisms discussed below are organized into two complementary levels: core checkpoint and DNA repair pathways, and broader regulatory layers sustaining adaptive DDR competence in GSCs. Some signaling nodes, therefore, recur across sections as they are redeployed in distinct biological contexts. The following sections dissect these mechanisms, with supporting experimental models summarized in Supplementary Table 1. Details of the literature search strategy and the PRISMA flow diagram (Supplementary Fig. 1) are provided in the Supplementary Information.
Fig. 1: Multilayered DDR plasticity in glioma stem cells and therapeutic vulnerabilities.The alternative text for this image may have been generated using AI.
Genotoxic stress induced by radiotherapy and temozolomide generates diverse DNA lesions, including double-strand breaks, alkylated bases, oxidative damage, and replication stress. These insults activate core DNA damage response (DDR) processes, including checkpoint signaling, replication stress management, and the selection of DNA repair pathways. In glioma stem cells (GSCs), these processes are embedded within a broader network of regulatory layers comprising chromatin and epigenetic regulation, RNA processing and epitranscriptomic control, translational adaptation, metabolic rewiring, and microenvironmental interactions. Dynamic, state-dependent DDR plasticity enables adaptive switching across these layers, promoting survival under genotoxic stress. Transcriptional recovery, defined as restoration of RNA polymerase II activity following damage, is highlighted as a functional DDR endpoint and an emerging determinant of therapy resistance. Color-coded modules indicate major therapeutically targetable axes, while solid arrows denote experimentally established mechanisms and dashed arrows represent emerging or proposed connections. This integrated network ultimately sustains GSC persistence and drives glioblastoma recurrence. Blue modules indicate DDR-targeting strategies (e.g., ATR, PARP, RAD51, DNA-PK inhibitors), pink modules indicate epigenetic targeting approaches (e.g., LSD1 inhibition), cyan modules represent RNA/epitranscriptomic targeting (e.g., METTL3, ALKBH5), green modules indicate metabolic targeting strategies (e.g., SCD1, PHGDH), and yellow modules represent transcriptional recovery as an emerging therapeutic dimension without an established drug class. Solid arrows denote experimentally established mechanisms, whereas dashed arrows indicate emerging or proposed functional connections. Created in BioRender. Cerutti, E. (2026) https://BioRender.com/7fjatyc.
Checkpoint signaling and replication stress
GSCs are increasingly recognized as a tumor-propagating state sustained by reinforced DDR signaling and hyperactive cell-cycle checkpoints, forming a central axis of radio- and chemoresistance. Seminal studies showed that, following irradiation, GSCs rapidly and persistently activate the ATM/ATR-Chk1/Chk2 cascades, enforcing a robust G2/M checkpoint and resolving DNA double-strand breaks (DSBs) more effectively than differentiated glioma cells. This phenotype reflects checkpoint hyperactivation and replication slowing rather than intrinsically superior repair capacity, enabling prolonged damage tolerance and delayed cell-cycle transitions37,38,39,40,41,42. Pharmacologic ATM inhibition restores sensitivity to DSB-inducing agents, supporting checkpoint reinforcement as central to GSC survival43.
A key driver of this state is chronic intrinsic replication stress. GSCs exhibit reduced fork velocity, γH2AX/53BP1 accumulation at replication sites, and elevated RNA-DNA hybrids driven by transcription of long neural genes. Rather than inducing fork collapse, this stress sustains ATR-Chk1 signaling and checkpoint addiction, conferring radioresistance but exposing a vulnerability, as combined ATR/PARP inhibition abolishes self-renewal in vitro and in vivo41,44,45. CDK12/CDK13 inhibition similarly disrupts transcriptional elongation and replication fork progression in glioblastoma, linking transcription-replication conflict to checkpoint-dependent survival46, while PARG inhibition in glioma stem cells induces replication arrest, intra-S-phase checkpoint activation, and apoptosis in an NAD + -dependent manner, identifying replication stress buffering as an actionable dependency47.
Multiple buffering systems further support tolerance to replication-associated damage. Telomere integrity represents a major DDR node: the G-quadruplex ligand telomestatin induces telomeric dysfunction in GSCs, triggering ATR-Chk1 activation and replication-dependent death while sparing non-stem tumor cells48. Resolution of replication-associated topological stress relies on topoisomerase IIβ (TOP2β); its depletion sensitizes GSCs to TMZ and alkylating stress, shifting their response toward differentiated cells49. Replication protein A (RPA) overexpression likewise supports GSC survival, and its inhibition causes DSB accumulation and radiosensitization, highlighting dependence on fork protection rather than lesion removal50. Additional work indicates that checkpoint protection is modulated by upstream signaling, with IGF1R promoting radioresistance51 and CDC20 perturbation increasing therapy sensitivity52, further linking checkpoint function to broader survival and cell-cycle networks.
Within this checkpoint-dominated landscape, canonical repair pathways are temporally reshaped to sustain damage tolerance. Delayed Fanconi anemia pathway activation and differential reliance on homologous recombination (HR), including RAD51 and FA components, permit prolonged G2 arrest and survival with unresolved lesions44,53,54. Telomere-associated DDR further reinforces this state: ALT-positive GSCs maintain chronic telomeric damage signaling while preserving chromosomal stability, supporting long-term persistence55,56. Preclinical studies in 2D and 3D GSC models support the therapeutic relevance of this checkpoint-addicted state57, while altered radiation delivery, including extra-high-dose-rate irradiation58, further supports context-dependent of checkpoint robustness.
Collectively, these studies define GSC resistance as a replication-stress-driven, checkpoint-addicted DDR state, critically dependent on sustained ATR/Chk1 signaling, fork protection, telomere integrity, and delayed engagement of repair pathways. Checkpoint reinforcement should therefore be viewed less as a marker of superior repair than as a mechanism of damage tolerance preserving GSC survival under sustained genotoxic stress.
DDR plasticity and adaptive survival programs
Beyond checkpoint addiction and replication-stress dependence, GSCs exhibit profound plasticity in DDR outputs. DNA damage sensing remains intact, whereas checkpoint enforcement, apoptosis, and differentiation are selectively attenuated, converting DDR into a permissive survival program shaped by oncogenic, inflammatory, niche-derived, and therapy-induced signals59,60.
Autocrine niche-like signaling is a major mechanism for modulating DDR. The EDN3-EDNRB axis maintains undifferentiated, anti-apoptotic and clonogenic GSC states under stress; its disruption abolishes self-renewal and tumorigenicity, indicating that survival-oriented DDR tuning sustains recurrence61. Inflammatory cues further decouple DDR activation from tumor suppression: prolonged IL-1β induces oxidative DNA damage and DDR signaling, while COX-2 suppresses p53, disabling apoptosis and checkpoint arrest under chronic genotoxic stress62. Periarteriolar and engineered perivascular niches similarly reinforce GSC enrichment and radioresistance, indicating niche-dependent stabilization of adaptive DDR states63,64. Laminin α2-dependent adhesion likewise supports stem-cell maintenance and stress tolerance65.
Oncogenic and adhesion-dependent pathways further tune DDR output. L1CAM links invasive traits to nuclear genome surveillance by regulating the MRN-ATM-Chk2 axis via NBS166. MET-high GSCs exhibit elevated ATM, Chk2, and RAD51 activation together with cytoplasmic anti-apoptotic p21; MET inhibition impairs DSB repair and restores radiosensitivity67. Integrin α6 likewise becomes essential after fractionated radiotherapy, supporting DSB repair and survival68. Distinct TMZ exposure conditions also drive different resistant states, indicating that DDR plasticity is dosage-dependent59.
Mitotic control, metabolic state, and post-translational regulation further reinforce checkpoint permissiveness. MELK sustains chromosomal integrity through FOXM1-Aurora B and ATM/ATR signaling, and its inhibition induces mitotic catastrophe and radiosensitization69. PDK1 cooperates with Chk1 to maintain G2/M control, and dual targeting, phenocopied by UCN-01, triggers catastrophic checkpoint collapse70. Chloroquine likewise underscores the context dependence of this balance, as it concurrently activates pro-survival and death-inducing signaling in GSCs71.
Metabolic support of DDR is also critical: elevated glycolysis sustains checkpoint and HR signaling, whereas GAPDH inhibition disrupts ATM/Chk activation and induces genomic fragmentation72. The deubiquitinase USP1 stabilizes ID1 and CHEK1, sustaining stem-like identity and checkpoint signaling; USP1 inhibition enhances radiation-induced DNA damage and improves survival in vivo53. Likewise, PAF/KIAA0101 promotes translesion synthesis (TLS) and radioresistant self-renewal; its depletion or TLS inhibition reduces sphere formation and sensitizes cells to irradiation73.
These adaptive DDR states are increasingly exploited therapeutically. ATR inhibition synergizes with temozolomide in MGMT-methylated cells, increasing γH2AX, apoptosis, and clonogenic loss74. The cancer-testis lncRNA PITAR promotes adaptive rewiring by stabilizing TRIM28 and promoting p53 degradation, whereas PITAR knockdown restores p53 signaling and sensitizes cells to TMZ75. Multi-target approaches reinforce this concept: the brain-penetrant Hsp90 inhibitor NXD30001 suppresses self-renewal while attenuating DDR and ER stress and synergizes with radiotherapy76, whereas arsenic trioxide combined with (-)-gossypol downregulates DDR genes and selectively kills GSCs ex vivo77. The natural product taccaoside A similarly targets DDR-supporting oncogenic signaling by inhibiting HRAS/KRAS-PI3K-AKT and MAPK-ERK pathways, inducing apoptosis and loss of stemness in vivo78. Additional approaches, including nitric oxide, PBI-05204, and oHSV-P10, similarly destabilize adaptive survival states without directly targeting repair enzymes79,80,81.
Radiobiological context further shapes DDR plasticity. Total radiation dose, rather than dose rate, governs response58, while proton beams outperform photons by inducing overwhelming ROS and DDR collapse82. Targeted radio-pharmacologic strategies can override quiescence: Sonic Hedgehog activation forces dormant cells into S phase, enabling [I-125]ITdU incorporation and selective elimination of GSC populations83. Biological therapies directly exploit DDR rewiring: oHSV-G47Δ with TMZ sequesters ATM and suppresses ATR signaling84, oHSV-TRAIL induces potent apoptosis in recurrent tumors85, and parvovirus MVM preferentially replicates in p53-altered GSCs, triggering DDR-dependent bystander damage in vivo86. Inhibition of Src signaling with AZD0530 further radiosensitizes glioma cells by blocking radiation-induced Src activation87. Advanced tumor models further underscore DDR heterogeneity and adaptability. Profiling of patient-derived lines reveals recurrent driver mutations and DDR-enriched transcriptional clusters with prognostic relevance88. Combined Nbs1/p53 loss generates highly unstable gliomas resembling pediatric and post-radiotherapy tumors89. Three-dimensional platforms further show that Tumor Treating Fields synergize with TMZ and PARP inhibitors independently of p53 or MGMT status, reducing BRCA1 levels and impairing DNA repair90,91. In parallel, ALT glioma models further support tolerance of persistent, non-lethal genome maintenance states92.
Together, these studies establish DDR plasticity as a central, multi-layered survival program in GSCs, in which signaling, microenvironmental cues, and stress-adaptive responses dynamically rewire DDR outputs to favor persistence under genotoxic stress.
Direct reversal and base excision repair in chemoradioresistance
Downstream of checkpoint activation and replication-stress tolerance, GSCs rely on direct reversal and base excision repair (BER) against alkylation-induced DNA damage. MGMT-mediated reversal of O6-methylguanine lesions and PARP-dependent BER critically buffer the cytotoxic effects of TMZ and radiotherapy, sustaining clonogenic survival under genotoxic stress.
PARP1 inhibition disrupts BER and converts repairable lesions into replication-associated DSBs in an MGMT-dependent manner93. MGMT itself remains a central determinant of therapy response: type I interferons suppress MGMT transcription via NF-κB inhibition94, PRIMA-1MET reduces MGMT expression and stemness independently of p5395, while hypoxia enhances both MGMT levels and stem-like traits96. Consistently, MEK-ERK signaling likewise sustains MGMT expression and temozolomide resistance through the MDM2-p53 axis, linking MGMT regulation to oncogenic signaling97. Conversely, delphinidin glycosides inhibit NF-κB-driven MGMT transcription and synergize with TMZ, particularly in mesenchymal GSC populations98. Therapeutic MGMT blockade can also be achieved through targeted delivery strategies, as Biomimetic co-delivery of TMZ with an MGMT inhibitor suppresses orthotopic glioblastoma growth, supporting direct reversal as a tractable vulnerability in vivo99. MGMT targeting also radiosensitizes GSCs by prolonging γH2AX persistence and promoting mitotic catastrophe after irradiation100. Nanoparticle-mediated silencing of drug-resistance programs likewise prolongs survival in orthotopic glioblastoma models101.
Replication-associated buffering mechanisms functionally intersect with direct reversal and BER. Telomestatin selectively induces telomeric dysfunction and ATR-Chk1 hyperactivation in GSCs, triggering replication-dependent arrest and death48. Resolution of replication-associated topological stress depends on topoisomerase IIβ, whose depletion sensitizes GSCs to TMZ and alkylating agents49, whereas RPA inhibition causes extensive DSB accumulation and radiosensitization50. Consistently, PARP inhibition with talazoparib markedly enhances radiosensitivity, particularly with high-LET carbon ions, by preventing DSB resolution and enforcing durable G2/M arrest102. In this context, ATR-Chk1 signaling becomes a critical compensatory dependency, and ATR inhibition synergizes with TMZ specifically in MGMT-methylated GSCs74. BER-centered defenses are further modulated by hypoxia, as the lncRNA LUCAT1 regulates DDR output in hypoxic GSCs103.
Cell state and microenvironment further modulate these repair-centric defenses. Dedifferentiation driven by RB loss, KRAS activation, and PTEN inactivation generates a stem-like, MGMT-positive TMZ-resistant phenotype104, while tumors arising near the subventricular zone exhibit enriched stemness programs, elevated MGMT signaling, and inferior outcome105. Therapeutic strategies increasingly exploit these vulnerabilities: dual ALKBH2/ALKBH5 inhibition reverses TMZ resistance and reduces MGMT expression106. Tumor Treating Fields sensitize GSCs to PARP inhibition and TMZ independently of p53 and MGMT status91, PDCD10 loss promotes MGMT upregulation and TMZ tolerance107, and differentiation therapy with all-trans retinoic acid downregulates MGMT and stemness, restoring TMZ sensitivity108. Oltipraz also shows antitumor activity in glioblastoma, although its connection to MGMT or BER appears indirect109.
Together, these studies indicate that direct reversal and BER operate within a broader replication-stress-adaptive network in which MGMT regulation, PARP activity, and fork protection sustain therapy resistance. Targeting these interconnected layers may therefore dismantle repair-centered defenses in HGG, positioning BER and direct reversal as proximal buffering layers within a broader adaptive DDR network.
Homologous recombination dynamics
Following checkpoint addiction and replication-stress tolerance, HR emerges as a key determinant of therapy response heterogeneity. In GSCs, HR spans a continuum from defective, genomically unstable states to RAD51-dependent HR-addicted phenotypes.
Early evidence indicates that not all GSCs are radioresistant: CD133⁺ populations can exhibit impaired HR and incomplete checkpoint activation, leading to radiosensitivity and chromosomal instability110. However, subsequent studies predominantly identified reinforced HR states. RAD51 is consistently upregulated in radioresistant GSCs and further induced by irradiation; its pharmacologic inhibition (RI-1, B02) suppresses RAD51 foci, disrupts HR, induces persistent DSBs and apoptosis, and selectively collapses clonogenic survival while sparing normal neural stem cells111,112. Clinically, high RAD51 expression correlates with poor response and shorter progression-free survival.
This HR dependency is sustained by convergent transcriptional and regulatory circuits. STAT3-dependent FOXM1 signaling transcriptionally controls core HR genes, and its inhibition increases γH2AX, impairs HR, and induces mitotic catastrophe113. The lncRNA DARS1-AS1 stabilizes FOXM1, RAD51, and BRCA1 transcripts via YBX1, sustaining HR and tumor growth; its depletion radiosensitizes GSCs in vivo114. BRCA1 itself modulates TMZ response in p53 wild-type GSCs, where its knockdown suppresses HR activation and increases apoptosis115. PRMT5 inhibition likewise impairs Fanconi anemia pathway-mediated HR and enhances temozolomide efficacy, supporting FA/HR-associated repair dependencies as actionable vulnerabilities116. Likewise, MEOX2 promotes DNA repair and therapy resistance in glioblastoma stem-like cells through PARP1 interaction, indicating that HR competence depends on broader regulatory networks beyond RAD51 abundance alone117.
HR addiction is therapeutically exploitable. Artesunate synergizes with TMZ by suppressing RAD51 and HR without increasing primary DSB burden118. Oncolytic HSV combined with PARP inhibition induces proteasomal degradation of RAD51 and CHK1, causing massive DSB accumulation and apoptosis selectively in GSCs; RAD51 knockdown phenocopies PARPi sensitivity, confirming HR disruption as the synthetic-lethal driver119. Conversely, resistance to PARP inhibition can arise through a Myc-CDK18-ATR axis: Myc-amplified cells exhibit an HR-defective, PARPi-sensitive state, while CDK18-driven ATR activation restores HR in non-amplified cells; inhibition of CDK18 or ATR re-sensitizes these populations120. In line with this, lipid metabolic regulation also feeds into HR-associated repair, as SCD1 and SCD5 modulate PARP-dependent DNA repair through fatty acid desaturation121.
HR regulation further integrates chromatin and stress-adaptive pathways. The elongation factor SPT6 maintains BRCA1 and RAD51 expression; its inhibition suppresses HR, induces G2 arrest, polyploidy, and loss of tumorigenicity122. RECQL4 depletion disrupts HR, activates checkpoint signaling, and sensitizes GSCs to TMZ123. Telomere-directed stress also intersects with this circuitry, as G-quadruplex stabilization imposes replication-associated stress incompatible with GSC survival124.
Metabolic and ER-stress signaling converge on HR through SCD1, which sustains RAD51 expression downstream of IRE1-SREBP1; its inhibition induces persistent DNA damage and near-complete GSCs depletion125. In 3D cultures, VEGF-Akt signaling promotes HR and DNA-PKcs activation, whereas its inhibition causes durable γH2AX accumulation and radiosensitization126. The chaperone HSP90 stabilizes RAD51 and CHK1; onalespib-mediated degradation of both proteins abrogates HR and enhances radio- and chemosensitivity127.
Finally, post-translational and epitranscriptomic regulators reinforce HR proficiency. The UCHL3-POLD4 axis stabilizes replication polymerase POLD4, sustaining HR/NHEJ and radioresistance128, whereas the MST4-USP14-ALKBH5 module preserves RAD51 expression via m6A-dependent stabilization; its disruption impairs HR and radiosensitizes GSCs129. In vivo genetic modeling using RCAS-TVA-CRISPR systems further enables functional dissection of HR dependencies and synthetic-lethal vulnerabilities during glioma evolution130.
Together, these studies define HR as a dynamically regulated vulnerability ranging from HR-deficient to HR-addicted states. Across this spectrum, RAD51-centered HR networks, integrated with transcriptional, metabolic, and stress-response pathways, represent a central and targetable axis of therapy resistance in HGG.
DNA-PK-dependent end joining
Whereas homologous recombination defines a major resistance axis in proliferating GSCs, non-homologous end joining (NHEJ) provides a complementary and often dominant survival strategy under replication stress, quiescence, and acute irradiation. In GSCs, rapid, cell-cycle-independent DSB repair mediated by DNA-PKcs and XRCC4 enables survival after genotoxic stress at the cost of increased mutagenic tolerance.
Multiple studies demonstrate a selective reliance of GSCs on DNA-PK-driven NHEJ. Depletion of DNA-PKcs impairs DSB repair and redirects damaged cells toward LC3-Beclin-1-dependent autophagic cell death through mTOR inhibition, mechanistically linking defective NHEJ to autophagic vulnerability131. Consistently, temporal DNA-PK activation has been shown to drive genomic instability and therapy resistance in GSCs, supporting the view that persistent NHEJ signaling promotes survival at the cost of genome integrity132. Cathepsin L depletion likewise enhances radiosensitivity in vitro and in vivo, further supporting NHEJ-linked stress tolerance as a therapeutically actionable dependency133. Consistently, NHEJ dominance over apoptotic signaling and persistence in quiescence are defining features of GSCs exposed to genotoxic stress134.
This dependency is tightly modulated by oncogenic signaling. Akt inhibition robustly radiosensitizes GBM cells, particularly those harboring mutant or lost TP53, by reducing DNA-PKcs levels and crippling NHEJ-mediated DSB repair. Silencing TP53 in wild-type cells phenocopies this vulnerability, indicating that TP53 loss enforces compensatory reliance on PI3K-Akt-NHEJ signaling. Combined Akt and DNA-PKcs inhibition radiosensitizes even p53-proficient cells, supporting cooperative targeting of this axis135. Direct pharmacologic exploitation of this circuitry is exemplified by the EGFR-directed combi-molecule ZR2002, which suppresses EGFR/Erk/Akt signaling while inducing DNA damage and selectively eliminates TMZ-resistant, EGFRvIII-positive GSCs in vivo in a p53-dependent manner136. This concept is reinforced by ZYH005, which similarly couples EGFR targeting to DNA damage induction and mitotic catastrophe in glioblastoma137.
Spatial context further shapes NHEJ reliance. In 3D GSC models, VEGF-Akt signaling activates DNA-PKcs and sustains both NHEJ and HR; VEGF deprivation abrogates these pathways, induces persistent γH2AX and pDNA-PKcs foci, and triggers mitotic catastrophe with profound radiosensitization. Akt inhibition reproduces these effects across both 2D and 3D systems, improving survival in vivo126.
Post-translational regulation integrates NHEJ with stemness and plasticity. In mesenchymal GSCs, the UCHL3-POLD4 axis stabilizes polymerase δ, supporting efficient NHEJ and HR, self-renewal, and tumorigenicity; its disruption induces persistent DSBs, loss of stemness, and potent radiosensitization128. Transcription-associated repair scaffolds further reinforce NHEJ function. AATF, upregulated in GSCs, binds and stabilizes XRCC4 at replication forks; following DNA damage, ATM-dependent phosphorylation releases XRCC4 to enable DSB repair. Disrupting this interaction impairs NHEJ, increases unrepaired damage and apoptosis, and sensitizes tumors to radio- and chemotherapy138.
Together, these studies establish DNA-PKcs- and XRCC4-dependent NHEJ as a dominant, druggable survival pathway in GSCs that cooperates with oncogenic signaling and niche cues to sustain DSB repair and stress tolerance.
Beyond canonical DNA repair pathways, GSC resistance is further sustained by regulatory layers that do not directly remove DNA lesions, but instead maintain stemness, stress tolerance, and the broader cellular conditions required for adaptive DDR states. The following sections focus on signaling, chromatin and RNA-based regulation, translation, and metabolism as modules shaping DDR adaptation in GSCs.
PI3K/Akt signaling as an upstream regulator of stemness and DDR adaptation
The PI3K/Akt pathway is one of the most consistently activated survival programs in GSCs and acts primarily as an upstream regulator of stemness-associated survival rather than as a direct DNA repair pathway. By integrating mitogenic, metabolic, and stress-adaptive signals, Akt establishes a permissive context in which GSCs suppress apoptosis and sustain recovery after genotoxic stress. PI3K/Akt signaling, therefore, acts as a coupling layer between stemness and DDR adaptation. This view is further supported by studies showing that pharmacologic inhibition of VEGF/PI3K/Akt signaling or dual PI3K/mTOR blockade is sufficient to induce cell-cycle arrest, apoptosis, and radiosensitization in GSCs, consistent with a role in maintaining adaptive survival states rather than directly executing DNA repair139,140.
Notch signaling provides a prototypical example of Akt-dependent radioresistance. Following irradiation, Notch activation promotes GSC survival by engaging PI3K/Akt and modulating anti- versus pro-apoptotic Mcl-1 isoforms, without increasing intrinsic DNA repair capacity. Pharmacologic inhibition of Notch using γ-secretase inhibitors selectively radiosensitizes the CD133⁺ stem-like fraction while sparing normal neural cells, highlighting a GSC-specific dependence on this axis141. GSCs also sustain Akt activation through a distinct autocrine sphingosine-1-phosphate (S1P) loop. Unlike normal neural cells, GSCs constitutively produce and secrete S1P, activating S1P1-Akt signaling to maintain survival under genotoxic stress. This signaling limits DSB accumulation and supports viability even when MGMT-mediated repair is compromised, identifying S1P as a critical upstream regulator of Akt-driven resistance142.
Disruption of PI3K/Akt signaling unmasks pronounced vulnerabilities. Hyperthermia prevents radiation-induced Akt activation, abolishes self-renewal, induces persistent DSBs and apoptosis, and suppresses tumor growth. In vivo, combined hyperthermia-radiotherapy significantly prolongs survival, with Akt inhibition driving radiosensitization143. Recent work further extends this concept upstream and downstream of Akt signaling: PTEN reactivation impairs GSCs by disrupting cytosolic iron-sulfur assembly, whereas brain-targeted co-delivery of osimertinib and bortezomib suppresses radioresistant glioblastoma in a differentiation-informed therapeutic setting, supporting the tractability of PI3K/Akt-regulated resistant states144,145.
Together, these studies indicate that PI3K/Akt signaling does not primarily enhance DNA repair, but preserves stemness-associated survival states in which adaptive DDR programs can be deployed. Targeting this pathway may therefore be required to dismantle the adaptive survival state of tumor-maintaining GSCs.
Epigenetic and epitranscriptomic control of stemness and DDR competence
Epigenetic and epitranscriptomic regulation are central determinants of DDR competence and therapy resistance in GSCs. By coordinating the control of chromatin structure, histone modifications, DNA/RNA methylation, and protein stability, GSCs establish a plastic regulatory landscape that reinforces DNA repair and adaptive survival under genotoxic stress. Chromatin- and RNA-based mechanisms, therefore, represent complementary layers of DDR regulation. Disrupting these circuits may therefore destabilize adaptive DDR states.
RNA methylation has emerged as a major regulatory layer. METTL3-dependent m6A methylation is highly enriched in GSCs and stabilizes SOX2 mRNA via 3′UTR modification, sustaining stemness and radioresistance; METTL3 depletion reduces neurosphere formation, increases γH2AX persistence, and radiosensitizes GSCs, effects rescued by SOX2 lacking the regulated 3′UTR146. METTL3 also promotes chemoresistance by stabilizing MGMT and APNG transcripts, lowering TMZ sensitivity in vitro and in vivo when depleted147.
Global DNA and RNA methylation patterns further distinguish GSCs from normal neural stem cells. GSCs exhibit reduced 5mC/5hmC, increased 5fC/5caC, aberrant TET activity, and enhancer remodeling linked to proliferation and therapy tolerance. TET2 overexpression supports stemness maintenance and DNA repair efficiency, implicating TET-dependent oxidation programs in survival under therapy148.
m6A demethylases provide an additional layer of DDR control. The m6A demethylase FTO promotes radioresistance and stemness maintenance in GSCs, supporting reversible RNA methylation as a regulator of DDR adaptation149. Likewise, ALKBH5, highly expressed in GSCs, promotes radioresistance and invasion by sustaining HR proficiency through regulation of CHEK1, RAD51, and other HR genes; its inhibition radiosensitizes GSCs and suppresses growth in 3D models150. At the post-translational level, the MST4-USP14-ALKBH5 axis prevents ubiquitin-mediated ALKBH5 degradation, preserving RAD51 expression and HR capacity; disruption of this pathway induces persistent DNA damage, apoptosis, and radiosensitization129.
Histone-modifying enzymes also critically shape DDR outcomes. The lysine demethylase KDM1A/LSD1, overexpressed in GSCs, transcriptionally activates HR and NHEJ genes, including BRCA1, RAD51, and FOXM1; its inhibition suppresses DSB repair, increases γH2AX accumulation, reduces stemness, and sensitizes GSCs to TMZ in vivo151. Similarly, KDM2B supports glioma stem-like cell survival and chemoresistance, extending the role of histone demethylases beyond LSD1 to a broader epigenetic framework that preserves resistant stem-like states152. Similarly, G9a/GLP inhibition sensitizes both glioma cells and GSC-enriched neurospheres to TMZ by inducing apoptosis and autophagy, independently of MGMT promoter methylation or stemness gene expression, identifying G9a as a broad epigenetic vulnerability across GBM cell states153. HDAC6 inhibition promotes GSC differentiation and radiosensitivity through SHH/Gli1 suppression, further linking stemness maintenance to DDR competence154.
Clinically relevant signals further intersect with epigenetic DDR regulation. Dexamethasone induces a CEBPB-driven transcriptional program that enhances proliferation, survival, and TMZ resistance, with upregulation of DNA repair genes such as XRCC2 and BRCA2. This mesenchymal repair-competent state can be antagonized by camptothecin, suggesting that steroid exposure may reinforce DDR-proficient states155. Ionizing radiation itself can also reshape this layer by regulating MYC and NBN expression, suggesting that therapy-induced transcriptional responses feed back into epigenetically controlled DNA repair programs156.
Epigenetic reactivation of tumor suppressors also modulates therapy response. The DNA methylation inhibitor 5-azacitidine restores the expression of TUSC3, reducing stemness and re-sensitizing GSCs to TMZ independently of MGMT status; in MGMT-unmethylated lines, combination with lomeguatrib enables robust tumor suppression157.
Finally, higher-order chromatin organization directly couples inflammatory signaling to DDR. NUP98, acting as a chromatin scaffold through interaction with NF-κB (p65/RelA), sustains transcription of core HR factors including BRCA1/2, RAD51, RAD54L, BLM, and XRCC2. NUP98 depletion collapses HR capacity, increases post-irradiation apoptosis, reduces stemness, and sensitizes tumors to radio-chemotherapy, identifying NUP98-NF-κB as a chromatin-based DDR hub158. At the same time, miR-128-mediated targeting of Polycomb repressor complexes further indicates that non-coding RNA networks epigenetically restrain stem-like programs159.
Together, these studies establish epigenetic and epitranscriptomic regulation as a major layer coordinating DDR competence in GSCs. Targeting RNA methylation, histone modifiers, chromatin scaffolds, and protein-stability circuits may therefore destabilize treatment-refractory GSC states.
RNA splicing as a post-transcriptional link between stemness and DDR competence
Building on epigenetic regulation, RNA splicing emerges as a post-transcriptional layer through which GSCs fine-tune stemness and DDR competence. Alterations in spliceosome integrity and alternative splicing directly shape DNA repair and oncogenic signaling networks.
The spliceosomal component SNRPB is overexpressed in GBM and patient-derived GSCs, with high levels correlating with poor prognosis. SNRPB knockdown induces apoptosis and widespread splicing disruption, including intron retention and downregulation of genes involved in RNA processing, chromatin remodeling, HR, and DNA repair. Affected pathways include RTK, PI3K-AKT, RAS/MAPK, RB, and p53, indicating that spliceosome integrity is required to maintain DDR-proficient stem-like programs160. Splicing regulation also underlies functional heterogeneity across GSC subtypes. Comparative analyses of proneural and mesenchymal GSCs identify ~4900 differential splicing events, with strong concordance between in vitro models and patient tumors. Mesenchymal GSCs preferentially exhibit alternative splicing of DNA repair and cell-cycle genes (ERCC1, FANCD2, RAD17) and increased expression of prognostic lncRNAs (CRNDE, MYOSLID, SOX21-AS1), correlating with enhanced radioresistance161. More broadly, post-transcriptional RNA regulation beyond core spliceosome function also contributes to resistant stem-like states, as miR-146b-5p and proneurogenic miRNA programs modulate stemness and therapy response in GSCs162,163.
Together, these findings indicate that aberrant RNA splicing contributes directly to the acquisition and maintenance of DDR competence in stem-like glioma states. This supports spliceosome-associated regulators as therapeutic vulnerabilities in HGG.
Translational control of DDR competence
Beyond checkpoint activation and repair pathway selection, GSCs depend heavily on translational control to sustain DDR signaling after genotoxic stress. Following irradiation, rapid replenishment of short-lived repair and checkpoint proteins becomes rate-limiting, rendering protein synthesis a key determinant of survival. Recent translatome-level analyses further support a major role for mRNA translation in adaptive glioblastoma biology164.
The mTOR pathway is central to this dependency. Dual mTORC1/2 inhibition with AZD2014 (Vistusertib) impairs post-irradiation recovery, causing persistent γH2AX and delayed DSB resolution by suppressing eIF4E/4E-BP1-dependent cap-dependent translation of DDR proteins165. Polysome profiling confirms that radiation responses in GSCs are predominantly regulated at the translational level: INK128 disrupts eIF4F assembly, prolongs DNA damage signaling, alters organelle remodeling, and markedly radiosensitizes GSCs166.
Metabolic inputs converge on this axis. G0S2, upregulated in radioresistant GSCs and recurrent GBM, activates mTOR-S6K signaling and stabilizes 53BP1 through RNF168 suppression, promoting efficient DSB repair. Targeting G0S2 disrupts translational signaling, increases DNA damage, and restores radiosensitivity167.
Translational dependence is further reinforced by control of ribosome biogenesis. Inhibition of XPO1 with selinexor blocks nuclear export of 5S and 18S rRNA, suppresses polysome formation, and prevents synthesis of essential DDR proteins. Combined with irradiation, selinexor induces persistent γH2AX, unrepaired DSBs, and significantly prolongs survival in orthotopic models, while sparing normal neural tissue168. In parallel, extrinsic ribosome stimuli can drive glioma cells toward stem-like states, supporting a role for ribosome-linked translational control in maintaining DDR-competent stem-like states169.
Together, these studies place translational control downstream of stemness-associated survival circuitry and upstream of repair execution, identifying it as a systems-level determinant of adaptive DDR competence. This exposes a broader vulnerability that can be exploited to overcome radioresistance in GSCs.
Metabolic reprogramming and DDR support
Effective DDR execution in GSCs is tightly coupled to metabolic fitness. Metabolic rewiring provides the energetic, redox, and biosynthetic conditions required for adaptive DDR states under therapy.
Early studies showed that resistant GSCs adopt a caloric restriction-like metabolic state characterized by reduced glycolysis, enhanced β-oxidation, and constitutive AMPK/SIRT1-PGC-1α activation promoting autophagy and DDR reinforcement170. Disrupting this state sensitizes GSCs: resveratrol plus irradiation suppresses clonogenicity and impairs DSB repair in vivo171, whereas ATP depletion by D609 triggers GSC death independently of direct DNA damage172. Consistently, compromising mitochondrial fitness with 3-acetonyltabersonine selectively kills GSCs by inducing mitochondrial depolarization, suppressing ATM signaling, and impairing post-damage recovery173. CRISPRi screens further link metabolic stress responses to chemoresistance programs in glioblastoma174.
Metabolic-DDR coupling also emerges through oxidative stress and lipid signaling. The compound NEO100 induces ER stress and apoptosis while suppressing invasion and prolonging survival175. EGFR-amplified GSCs, characterized by chronic ROS and elevated baseline DDR, show a selective dependence on PARP1 and exquisite sensitivity to talazoparib176. At the population level, NRF2 maintains antioxidant defenses and stemness; its depletion increases oxidative stress and radiosensitizes GSCs to photons and carbon ions177. Chaperone-mediated autophagy further intersects with metabolic regulation by promoting degradation of enzymes such as IDH1, linking autophagy to metabolic-DDR coupling178.
Direct metabolic control of DNA repair has recently been uncovered. ALDH1A3 drives glycolytic flux and lactate-dependent XRCC1 lactylation, promoting BER/NHEJ under TMZ and radiation; disrupting the ALDH1A3-PKM2 interaction restores chemosensitivity179. Similarly, PHGDH-dependent serine synthesis sustains one-carbon metabolism, nucleotide supply, and HR proficiency; PHGDH inhibition increases ROS and DSBs and synergizes with radiation in PDX models180. The PGK1/PHGDH axis likewise drives radioresistance in GSCs, reinforcing metabolic support of DDR competence181. Lipid metabolism also directly modulates repair capacity: de novo lipid synthesis supports DNA repair, whereas its inhibition impairs DDR, and lipid droplet formation counteracts PARP inhibition182,183.
Microenvironmental and treatment-related factors further shape metabolic DDR states. Periventricular GSC niches exhibit enhanced stemness and DNA repair signaling184, while MRSI-defined GSC-rich regions correlate with aggressive metabolic and DDR programs185. Ex vivo, neurosphere cultures show greater radioresistance and slower γH2AX resolution than tumor bulk186. Macrophage-derived lactate suppresses cGAS-STING signaling and antitumor immunity, whereas lactate transport inhibition restores immune activation, linking lactate metabolism to DDR signaling and immune evasion187. Nuclear cholesterol also contributes to DDR regulation by controlling nuclear architecture and DNA damage responses in cancer stem cells188. High-LET carbon ions induce complex, slowly repaired damage in GSCs189, whereas fractionated radiotherapy can enrich CD133⁺ GSCs and promote mitochondrial biogenesis and invasiveness190. Therapeutically, IKCa/BKCa channel inhibition radiosensitizes patient-derived GSCs191, and an oncolytic HSV-1 armed with a BiTE exploits DNA damage-induced NKG2DL expression to amplify immune-mediated killing192. Iron metabolism also contributes to radiation-response heterogeneity, as ferritin heavy chain modulation alters sensitivity in glioblastoma-initiating cells193.
Together, these studies indicate that metabolism functions as a permissive layer of adaptive DDR competence, enabling stem-like glioma cells to buffer genotoxic stress and preserve long-term tumor-propagating capacity. Targeting these adaptive metabolic layers, therefore, offers a powerful route to destabilize therapy-resistant GSC states and enhance chemoradiotherapy efficacy.
Therapeutic implications of targeting DDR plasticity in GSCs
The multilayered DDR architecture described above reveals several tractable therapeutic vulnerabilities in GSCs.
Clinically, MGMT promoter methylation remains the strongest predictive biomarker for temozolomide response in IDH-wildtype glioblastoma, where it predicts benefit from alkylating chemotherapy in randomized trials194,195. In contrast, its predictive utility is limited in IDH-mutant gliomas, where MGMT methylation is nearly universal and does not reliably predict temozolomide benefit196,197. Beyond MGMT, replication stress and ATR–Chk1 dependency provide a rationale for combining ATR inhibitors, such as ceralasertib or gartisertib, with radiotherapy or temozolomide, particularly in MGMT-methylated tumors, although brain penetrance remains limiting for some agents such as berzosertib198,199,200. PARP inhibition, including olaparib and talazoparib, shows genotype-selective activity in EGFR-amplified GSCs, where EGFR-induced oxidative stress creates dependence on PARP-mediated BER, and in tumors with MYC/MYCN amplification or IDH mutations that confer a “BRCAness” phenotype amenable to synthetic lethality120,176,201,202. RAD51- and DNA-PKcs-targeting approaches have also entered clinical evaluation; for example, peposertib, a DNA-PKcs inhibitor, is being evaluated clinically in newly diagnosed MGMT-unmethylated glioblastoma (NCT04555577), with favorable safety and median OS of 22.9 months203. However, radiosensitization by DDR inhibitors may affect tumor and normal tissues in parallel, making therapeutic window optimization a central issue204,205. Additional opportunities emerge from epigenetic and epitranscriptomic targeting. METTL3 inhibition is particularly attractive because m6A-dependent regulation supports HR repair, stabilizes SOX2, and enhances MGMT/APNG expression, thereby promoting TMZ resistance146,206,207,208,209. LSD1 inhibition primarily disrupts GSC maintenance through deregulation of the ATF4-dependent integrated stress response, inducing differentiation and senescence rather than directly suppressing DDR genes210, whereas G9a inhibition can impair both HR and NHEJ repair and radiosensitizes glioma cells, although effects on stemness appear context-dependent211,212,213,214,215. Metabolic dependencies including PHGDH, SCD1, and ALDH1A3 are also being evaluated in orthotopic models, although GBM-specific clinical trials remain limited179,180,216,217,218.
Finally, translational control through mTOR and XPO1 inhibition is clinically advanced: selinexor has completed phase II trials in recurrent GBM (PFS6 17%) and is being evaluated in combination with standard therapy (NCT04421378), although it may paradoxically induce MGMT expression in MGMT-unmethylated tumors and antagonize TMZ efficacy168,219,220,221,222.
Effective therapy will likely require combination regimens simultaneously targeting checkpoint addiction (ATR/Chk1), repair execution (RAD51/DNA-PK), and adaptive survival programs (PI3K/Akt, epigenetic plasticity), guided by molecular DDR and metabolic biomarkers rather than histology alone. The main targetable axes, representative compounds, and candidate selection markers are summarized in Table 3.
Table 3 Therapeutic playbook for targeting DDR plasticity in glioma stem cells
