Targeting LDHA Palmitoylation Sensitizes GBM to Chemoradiotherapy via Suppressing Palmitic Acid Induced Glycolytic Reprogramming Zijie Gao, Rongrong Zhao, Zheyuan Chen, Qingtong Wang, Huanyu Yang, Boyan Li, Yanhua Qi, Shulin Zhao, Jixing Zhao, Jiawei Qiu, Ziwen Pan, Zhe Han, Ruochuan Li, Huize Xia, Weijie Tang, Xing Guo, Ping Zhang, Hongyu Zhao, Xu Qian, Gang Li, Hao Xue Abstract Glioblastoma (GBM) is a lethal malignancy characterized by therapeutic resistance and recurrence, which are driven by dynamic cellular metabolic adaptation within the tumor microenvironment. However, the precise molecular metabolic mechanisms remain poorly understood. Here, we found that palmitic acid (PA) accumulates in hypoxic pseudopalisading regions of GBM and drives the formation of mesenchymal (MES) GBM and the hypoxia myeloid‐derived macrophages (MDM) subpopulation, establishing an interactive tumor‐promoting niche that fuels GBM recurrence. Maechanistically, PA induces ZDHHC12‐mediated palmitoylation of lactate dehydrogenase A (LDHA) at cysteine 163. This enhances LDHA enzymatic activity, promoting glycolytic reprogramming that reinforces the MES phenotype and confers resistance to chemoradiotherapy. Concurrently, palmitoylation facilitates the exosomal sorting of LDHA, which is internalized by MDMs, inducing their polarization into an immunosuppressive, glycolytic hypoxia‐MDM subset, which cooperatively shapes an immunosuppressive pseudopalisading niche with MES glioma stem cells to promote GBM recurrence. Critically, we identified the small molecule Rutin as a potent inhibitor of LDHA‐Cys163 palmitoylation, thereby re‐sensitizing GBM to chemoradiotherapy. This study defines a metabolic circuit in recurrent GBM driven by crosstalk between PA‐activated MES‐GBM cells and hypoxia‐MDMs. The small molecule Rutin disrupts this circuit, re‐sensitizing GBM to chemoradiotherapy and offering a promising strategy to overcome post‐treatment recurrence. Mesenchymal GSCs preferentially uptake palmitic acid, thereby promoting ZDHHC12‐mediated LDHA palmitoylation. This modification enhances LDHA enzymatic activity, drives glycolytic reprogramming, and facilitates exosomal LDHA‐mediated TAM regulation. Pharmacological targeting of the LDHA Cys163 palmitoylation site by Rutin markedly suppresses LDHA palmitoylation and sensitizes MES GSCs to chemoradiotherapy. [ADVS-9999-e77536-g006.webp] Introduction Glioblastoma multiforme (GBM) represents the most aggressive primary intracranial malignancy, distinguished by near‐universal recurrence following multimodal therapy [, , ]. This therapeutic recalcitrance arises from profound intratumoral heterogeneity, wherein molecularly distinct cellular subpopulations exhibit differential susceptibility to conventional chemoradiotherapy [, , ]. Among them, the mesenchymal (MES) subtype, defined by high malignancy and aggressiveness, exhibits particularly pronounced resistance, thereby undermining therapeutic efficacy and compromising patient survival [, ]. Critically, a continuously sustained metabolic symbiosis between MES tumor cells and the immunosuppressive microenvironment collectively fosters the development of adaptive resistance mechanisms [, , ]. Thus, elucidating the niche‐specific metabolic drivers that coordinate cross‐compartment crosstalk to sustain therapeutic resistance is critical for developing targeted interventions. Within this ecological landscape, hypoxic pseudopalisading necrosis emerges as a pathognomonic feature and independent predictor of poor prognosis [, , ]. These pathophysiological niches impose extreme metabolic stress on both malignant glioma stem cells (GSCs) and infiltrating stromal populations, activating parallel adaptation pathways including glycolytic rewiring, lipid metabolic reprogramming, and immunosuppressive signaling [, , ]. The precise mechanisms by which spatially localized metabolites within these microenvironments coordinate functional remodeling of malignant cells and immune cell populations to promote therapeutic tolerance remain incompletely defined. Elucidating this metabolic crosstalk is critical for developing effective microenvironment‐targeted therapeutic interventions. Tumor‐associated myeloid cells (TAMs), the predominant immune infiltrates in GBM, encompass both resident microglia and infiltrating monocyte‐derived macrophages (MDMs) []. With tumor progression or recurrence, the proportion of MDMs progressively surpasses that of proinflammatory microglia, shifting the myeloid landscape toward a more immunosuppressive phenotype [, , ]. Critically, the GBM‐characteristic hypoxic microenvironment recruits MDMs and induces phenotypic reprogramming mediating therapeutic resistance [, , ]. Despite conserved hypoxic‐responsive MDM subsets in primary solid tumors, whether hypoxia‐driven MDM phenotypes dynamically evolve hierarchy during therapy, and how tumor metabolic fitness orchestrates the formation of these subsets require further investigation. In this study, we found that palmitic acid (PA) accumulates in hypoxic pseudopalisading regions of GBM and drives the formation of MES GSCs and the hypoxia‐induced MDM subpopulation, establishing an interactive tumor‐promoting niche that fuels GBM recurrence. Mechanistically, PA drives ZDHHC12‐mediated palmitoylation of lactate dehydrogenase A (LDHA) at cysteine residue 163. This modification enhances LDHA enzymatic activity, thereby promoting glycolytic reprogramming, reinforcing the MES phenotype of GSCs, and conferring resistance to chemoradiotherapy. Concurrently, palmitoylation facilitates exosomal sorting of LDHA. The exosomal LDHA is subsequently internalized by MDMs, inducing an immunosuppressive glycolytic hypoxia‐MDM subset. This subpopulation cooperatively shapes an immunosuppressive pseudopalisading niche with MES‐GSCs to promote GBM recurrence. Critically, structure‐based virtual screening identified Rutin as a potent inhibitor of LDHA‐Cys163 palmitoylation. By disrupting this signaling axis, Rutin re‐sensitized GBM to chemoradiotherapy, and suppressed tumor progression. This work elucidated the potential tissue adaptation mechanism of intratumoral dynamic interactions between MES‐GBM cells and hypoxia‐MDM in recurrent GBM and identified the therapeutic potential of LDHA‐Cys163 palmitoylation to reduce therapy resistance in GBM. Results Hypoxic Pseudopalisading Necrosis Region Acquires Palmitic Acid‐Driven Pro‐tumor Recurrence Traits Pseudopalisading necrosis regions in glioblastoma (GBM) are critically associated with patient prognosis and tumor recurrence [, ]. To identify metabolic pathways specifically upregulated in hypoxic necrotic zones, we performed integrated single‐cell RNA sequencing (scRNA‐seq), spatial transcriptomics (ST), and spatial metabolomics (SM) analyses on surgically resected tumor core and infiltrative margin specimens from six IDH‐wildtype GBM patients, comprising three primary and three recurrent cases (Figure ). Through unbiased single‐cell clustering coupled with canonical lineage marker expression profiling, we delineated transcriptionally distinct cell populations from 107 723 high‐quality cells derived from clinical tumor samples (Figure and Table ). Crucially, malignant cells and tumor associated macrophages (TAMs) emerged as the two most dominant cell populations, suggesting their core roles in tumor progression and recurrence. Malignant cells were further clustered into neural‐progenitor‐like (NPC), oligodendrocyte‐progenitor‐like (OPC), astrocyte‐like (AC), and mesenchymal‐like subpopulations (MES1 and MES2) based on canonical cell type markers (Figure and Table ). Within these specimens, MES1 subpopulations demonstrated the most complex functional profiles and the strongest epithelial‐mesenchymal transition (EMT) and fatty acid metabolism signature, being most pronounced in recurrent GBM samples (Figure ). We observed significant enrichment of the MES subpopulation, in recurrent disease (RD) GBM specimens (Figure ), where these cells predominantly accumulate within the hypoxic core regions of tumors (Figure and Figure ). These results demonstrated that the perinecrotic hypoxic MES1 subpopulation serves as a critical driver of tumor recurrence in GBM. To further validate this phenomenon, we established orthotopic patient‐derived xenograft models using human GBM patient‐derived xenograft (hGBM PDX) in immunodeficient mice. Mice treated with extended standard chemoradiotherapy (SCRT) developed therapy‐resistant tumors, showing reduced apoptosis and attenuated DNA damage post‐chemoradiotherapy (Figure ) versus control (Ctrl) group. This regimen selectively cleared therapy‐sensitive cells while enriching resistant populations. Subsequently, RNA transcriptome sequencing and proteomic profiling were performed on Ctrl and SCRT tumor cells to identify the key features associated with chemoradiotherapy resistance in GBM (Figure ). We observe that the upregulated transcriptome signature (transcriptome‐sig) and proteome signature (proteome‐sig) in chemoradioresistant tumor subpopulations predominantly map to the MES‐like subpopulations, with heightened prominence in the functionally more complex MES1 subpopulation (Figure ), indicating its central role in tumor recurrence. Further metabolic activity analysis revealed a marked elevation in fatty acid transport activity within SCRT group (Figure and Table ). In addition, scRNA functional enrichment analysis revealed significant enrichment of both MES1 program activity and fatty acid metabolic pathways (particularly the uptake and catabolism of long‐chain fatty acids) in recurrent GBM cells (Figure ). We further validated using an independent scRNA‐seq dataset that the MES subtype of GBM is characterized by enhanced fatty acid transport activity (Figure ). To further identify key regulatory fatty acids in GBM, we conducted metabolomic profiling and found that palmitic acid (PA), oleic acid (OA), stearic acid (SA), and linoleic acid (LA) were significantly enriched in GBM tissues (Figure ). We then established a fatty acid library and performed high‐throughput screening, which revealed that PA markedly enhanced the therapy resistance of glioma stem cells (GSCs) compared to other fatty acids (Figure ). Spatial transcriptomics and metabolomics confirm co‐enrichment of the MES1 tumor subpopulation, proteomics_sig and palmitate within hypoxia‐characterized pseudopalisading niches (Figure ), indicating that palmitate activities serve as a critical driver in GBM chemoradiotherapy resistance and recurrence. SLC27A3‐Mediated PA Uptake Facilitates the MES Subtype and Therapy Resistance in GSCs We next examined the relationship between palmitate availability and GBM MES transition. Validation studies using GSCs via BODIPY FL C16 pulse‐chase assays demonstrate time‐dependent potentiation of PA in MES GSCs compared to proneural (PN) GSCs (Figure ). Furthermore, spatial transcriptomic data demonstrated significant enrichment of long‐chain fatty acid uptake activity in regions densely populated with MES cells (Figure ). The solute carrier family 27 (SLC27) plays a crucial role in mediating palmitic acid transport (PAT), thereby facilitating its uptake and intracellular trafficking. To investigate the underlying mechanism driving elevated PAT in MES GSCs, we systematically knocked down individual SLC27 family using RNA interference. The results indicated that silencing SLC27A3 significantly reduced PAT activity in MES GSCs (Figure ). Consistently, SLC27A3 was upregulated in MES GBM tissues and MES GSC, correlating with poorer GBM patient survival, and its RNA and protein levels increase in SCRT tumor cells (Figure ). Collectively, these data suggested SLC27A3 as the principal molecular conduit for PA influx in MES GSCs. Furthermore, we found that PA promoted the maintenance of the MES subtype, as indicated by increased expression of MES markers including CD44, LYN, and Vimentin (Figure ), as well as promotion of invasive capacity, the latter being a fundamental trait of MES phenotypes, in a co‐culture model of tumor spheroids and normal rat brain organoids (Figure ). Critically, SLC27A3 knockdown reversed PA‐induced MES transition (Figure ). Additionally, MES GSCs cultured in PA‐deficient dialyzed medium (DM) exhibited loss of MES‐subtype markers and therapy resistance compared to their counterparts maintained in complete medium (CM). This phenotypic shift and associated therapy resistance were partially alleviated by exogenous PA supplementation (Figure and Figure ). Therapy‐induced DNA double‐strand breaks (DSBs) triggered γH2AX foci accumulation, a hallmark of treatment resistance when rapidly resolved by enhanced DNA repair machinery in tumor cells []. We observed that PA treatment significantly mitigated chemoradiotherapy‐induced DNA damage in MES GSCs, as evidenced by reduced comet tail moments (Figure and Figure ) and γH2AX foci formation (Figure and Figure ), concomitant with reduced apoptosis (Figure and Figure ). Given the critical role of homologous recombination (HR) in the repair of DNA double‐strand breaks, we employed a DR‐GFP reporter assay and found that PA significantly enhanced HR repair efficiency (Figure ). We further investigated its impact on glioma therapeutic resistance in vivo. Mice fed a palm oil‐rich diet exhibited accelerated tumor growth and reduced chemoradiotherapy efficacy compared to controls (Figure , and Figure ). Collectively, these data demonstrated that SLC27A3‐mediated PA uptake in GSCs potentiated MES and chemoradiotherapy resistance phenotypes, ultimately driving GBM recurrence. Palmitoyl‐dependent Glycolysis Fuels MES Subtype and Therapy Resistance in GBM PA is converted to palmitoyl‐CoA by long‐chain acyl‐CoA synthases, after which it can either enter mitochondria for β‐oxidation or serve as a substrate for protein palmitoylation [, ] (Figure ). To determine the dominant pathway, GSCs were treated with the β‐oxidation inhibitor Etomoxir or the palmitoyl‐transferase inhibitor 2‐bromopalmitate (2‐BP). Immunoblot analysis revealed that 2‐BP, but not Etomoxir, effectively abrogated PA‐induced upregulation of MES markers (Figure ), implicating palmitoylation as the primary mediator of the PA‐induced MES subtype. Moreover, 2‐BP substantially restored GSC apoptosis following therapy in PA‐treated cells, while Etomoxir exerted only marginal effects (Figure and Figure ). Similarly, both the alkaline comet assay and γH2AX IF staining further confirmed that, while PA attenuated therapy‐induced DNA damage, this protective effect was significantly abrogated by 2‐BP treatment (Figure , and Figure ). Collectively, these findings indicate that the pro‐tumorigenic and resistance‐enhancing effects of PA in GSCs are largely dependent on protein palmitoylation. To investigate the downstream targets and functional consequences of palmitoylation, we carried out acyl‐biotin exchange (ABE) experiments combined with mass spectrometry using both MES GSCs and GBM tumor tissues (Figure and Table ). This analysis identified palmitoylated proteins consistently modified in both GSCs and GBM tumor tissues. Enrichment analysis revealed that palmitoylated proteins were significantly associated with the glycolytic pathway (Figure ). Hypoxia, a hallmark of pseudopalisading necrotic regions, compels tumor cells to rely heavily on glycolytic energy production to adapt to this hostile microenvironment [, ]. Additionally, both our scRNA and ST analysis consistently demonstrated enhanced glycolytic activity in hypoxic pseudopalisading necrotic regions and areas enriched with the MES1 cell subpopulation (Figure ). GSEA further revealed significantly enhanced hypoxia‐associated glycolytic activity in both necrotic inner regions and recurrent tumor populations (Figure ). And seahorse analysis of extracellular acidification rate (ECAR) revealed that MES GSCs exhibited markedly higher glycolytic capacity and reserve compared to PN GSCs (Figure ). We next assessed whether PA regulates glycolytic processes. Glycolytic stress tests revealed that PA treatment significantly elevated the ECAR, indicating enhanced glycolytic flux in GSC 20 and GSC 267, whereas 2‐BP effectively reversed this effect (Figure and Figure ). Correspondingly, PA treatment also increased intracellular lactate levels, which were attenuated by 2‐BP co‐treatment (Figure and Figure ). Additionally, SCRT tumor cells exhibited higher glycolytic activity compared to Ctrl tumor cells, whereas treatment with 2‐BP suppressed glycolysis in both groups (Figure ). To further evaluate the impact of PA‐induced palmitoylation on glycolysis in vivo, we employed 18F‐FDG PET/CT imaging to assess glycolysis in xenograft models. The results demonstrated that PA‐enriched diet markedly increased glucose uptake in tumors, whereas co‐administration of 2‐BP significantly attenuated this effect (Figure ). Collectively, these findings demonstrate that MES GSCs possess greater metabolic flexibility and an enhanced ability to upregulate glycolytic flux under metabolic stress, which may contribute to their increased survival, adaptability, and resistance to therapeutic interventions. ZDHHC12 Contributes to MES Subtype and Chemoradiotherapy Resistance in GBM Protein palmitoylation is catalyzed by zinc finger DHHC domain‐containing palmitoyl‐transferases (ZDHHCs), a family comprising 23 members in mammals [, ]. In terms of the important role of palmitoylation on glycolysis and MES subtype, we initially performed correlation and survival prognosis analysis based on the TCGA GBM cohort. We noted that ZDHHC12 was highly positively correlated with glycolysis, MES subtype, and poor prognosis in the GBM dataset (Figure ). Further analysis of sequencing results from the SCRT and Ctrl tumor cells revealed that the mRNA expression levels of ZDHHC1, 3, 5, 12, 14, 16, and 24, as well as the protein levels of ZDHHC12 and ZDHHC13, were significantly upregulated in the SCRT tumor cells (Figure ). Knockdown of ZDHHC12 in SCRT tumor cells enhanced their sensitivity to therapy, whereas overexpression of ZDHHC12 in Ctrl tumor cells promoted therapy resistance. In contrast, manipulation of ZDHHC13 expression did not result in a significant change in therapy sensitivity (Figure ). Furthermore, analysis of public GBM datasets indicated that high ZDHHC12 expression is associated with reduced overall survival (Figure ). ScRNA‐seq analysis further revealed elevated ZDHHC12 expression specifically in the MES1‐GBM subpopulation (Figure ). Pseudotime trajectory analysis using Monocle2 indicated progressive upregulation of ZDHHC12 expression during tumor progression, with notably stronger expression in recurrent GBM cells (Figure ). And the IHC staining suggested that ZDHHC12 protein expression increased with elevated glioma grade (Figure ). Moreover, knockdown of ZDHHC12 significantly attenuated therapy resistance in MES GSCs (Figure and Figure ) and suppressed the expression of MES signature markers (Figure ). Simultaneously, rescue experiments demonstrated that ZDHHC12‐knockdown cells exhibited substantially diminished responses to PA administration compared to control cells (Figure , and Figure ). Collectively, these data indicated that the palmitoyl‐transferase ZDHHC12 mediates MES transition and therapy resistance in GSCs through palmitoylation‐dependent pathways. To elucidate the downstream effects of ZDHHC12, transcriptomic analysis revealed that genes significantly downregulated upon ZDHHC12 knockdown were prominently enriched in the MES1 subpopulation. (Figure ). GSEA further demonstrated that knockdown of ZDHHC12 markedly suppressed MES transition and glycolytic signatures (Figure ). Consistently, correlation analysis of scRNA data identified genes positively correlated with ZDHHC12 to be functionally implicated in both MES transition and glycolytic activity (Figure ). Seahorse glycolytic stress test further demonstrated that PA‐induced enhancement of glycolysis, glycolytic capacity, and glycolytic reserve was abolished by ZDHHC12 knockdown in MES GSCs (Figure and Figure ). Consistent results were obtained from measurements of lactic acid content (Figure and Figure ). Metabolic rewiring toward glycolysis, accompanied by a reduction in OXPHOS‐derived mitochondrial ROS, may promote the survival of GSCs following chemoradiotherapy []. Therefore, we assess the regulation of OXPHOS by ZDHHC12 by mitochondrial stress assays. Knockdown of ZDHHC12 significantly increased the OCR of GSC 20 and GSC 267, accompanied by a significant increase in ROS induced by chemoradiotherapy (Figure ). These findings suggested that inhibition of ZDHHC12‐mediated palmitoylation leads to elevated mitochondrial respiration and heightened oxidative stress following chemoradiotherapy (Figure ). To evaluate the effect of ZDHHC12 on the therapy sensitivity of GSCs in vivo, we established xenograft models using control and ZDHHC12‐knockdown GSC 267 followed by chemoradiotherapy. The xenograft models established with ZDHHC12‐KD GSCs showed significantly reduced tumor growth under both control and therapy conditions (Figure and Figure ). Survival analysis revealed prolonged survival in mice bearing ZDHHC12‐knockdown tumors, particularly following therapy (Figure ). Hematoxylin and eosin (H&E) staining, along with IHC staining of ZDHHC12 and γH2AX, further confirmed diminished tumor burden and enhanced DNA damage in the ZDHHC12 knockdown group under chemoradiotherapy (Figure ). Collectively, these findings demonstrate that ZDHHC12 drives glycolytic metabolic reprogramming and contributes to the therapy resistance of GSCs in vitro and in vivo. ZDHHC12‐Mediated LDHA Palmitoylation at Cys163 Drives MES Subtype and Therapy Resistance in GBM To identify potential palmitoylated substrates regulated by ZDHHC12, we performed co‐immunoprecipitation (Co‐IP) assays to screen for ZDHHC12‐interacting proteins. By intersecting these candidates (Table ) with palmitoylated proteins previously identified in GSCs and GBM tissues (Figure ), we identified six overlapping targets (Figure ). Among them, VDAC1, PKM2, SLC25A5 and LDHA are closely related to metabolic reprogramming and mitochondrial/glycolytic regulation, which are key biological features associated with the MES subtype of GBM [, , , ]. We therefore prioritized these four candidates for ABE‐based validation of ZDHHC12‐dependent palmitoylation. The results showed that ZDHHC12 knockdown reduced the palmitoylation levels of all four candidates to varying degrees, with the most pronounced decrease observed for LDHA, followed by a moderate reduction in PKM2, whereas VDAC1 and SLC25A5 were only minimally affected (Figure ). To further evaluate the functional contribution of LDHA and PKM2, we performed rescue experiments. Knockdown of LDHA significantly suppressed PA‐induced upregulation of MES markers and therapy‐resistance of GSC, whereas PKM2 knockdown had a negligible effect (Figure ), indicating LDHA as a critical effector of palmitoylated targeted protein. Co‐IP assays further confirmed the interaction between ZDHHC12 and LDHA (Figure ). We next detected the palmitoylation of LDHA using ABE assay and observed that LDHA could be palmitoylated, whereas 2‐BP treatment or knockdown of ZDHHC12 reduced the levels of palmitoylation on LDHA (Figure and Figure ). We further assessed the palmitoylation levels of endogenous LDHA in clinical specimens, including normal brain tissue and glioma samples of varying grades. LDHA palmitoylation levels progressively increased with glioma grade, correlating with elevated ZDHHC12 expression (Figure ). Collectively, these data established LDHA as a functionally significant palmitoylation substrate of ZDHHC12. As a post‐translational modification, palmitoylation covalently anchors palmitoyl groups to cysteine residues via thioester bonds, which directly modulate substrate stability, subcellular trafficking, and enzymatic activity [, , ]. Through screening, we identified that LDHA contains five highly conserved cysteine residues (Figure ). To identify the critical palmitoylation site on LDHA, we generated a series of LDHA point mutants in which specific cysteine residues were substituted with alanine (C‐A) (Figure ). We found that mutating all potential sites completely abolished LDHA palmitoylation, whereas reintroducing cysteine at position 163 restored it, thereby identifying Cys163 as the major palmitoylation site of LDHA (Figure ). Additionally, single‐point mutant validation revealed that the palmitoylation level of the LDHAC163A mutant was significantly lower than that of wild‐type (WT) LDHA, further confirming the critical role of Cys163 in LDHA palmitoylation (Figure ). To further confirm the palmitoylation site of LDHA, we constructed a CPP‐fused competing peptide encompassing the LDHA amino acid sequence 158–168 containing the palmitoylation motif (CPP‐PWT), along with its site‐specific mutant variant (CPP‐PMUT). Notably, treatment with CPP‐PWT markedly reduced LDHA palmitoylation levels compared to CPP‐PMUT, supporting the critical role of Cys163 in mediating LDHA palmitoylation (Figure ). Rescue assays in LDHA‐knockdown GSCs showed that while LDHAWT restored MES marker expression, the C163A mutant failed to do so (Figure ), highlighting the functional role Cys163 in mediating LDHA palmitoylation. To determine whether C163 mutation affects ZDHHC12‐LDHA binding, we further performed ZDHHC12 Co‐IP and PLA assays. LDHAC163A showed comparable association and proximity with ZDHHC12 relative to LDHAWT (Figure ), indicating that Cys163 mutation impairs LDHA palmitoylation without disrupting ZDHHC12‐LDHA interaction. LDHA Cys163 Palmitoylation Modulates Its Structure, Enzymatic Activity and Glycolytic Reprogramming It has been shown that palmitoylation play a key regulatory role by influencing protein expression, enzyme activity, and extracellular secretion [, , , ]. We first explored the regulatory mechanisms of endogenous palmitoylated LDHA. The LDHA protein expression and subcellular localization were not significantly altered by PA treatment or 2‐BP (Figure ). We found that PA promoted LDHA enzymatic activity. Notably, rescue experiments demonstrated that inhibition of palmitoylation by 2‐BP or knockdown of ZDHHC12 significantly reversed this stimulatory effect (Figure ). We further performed an in vitro LDHA enzymatic assay to evaluate the direct effect of Cys163 palmitoylation on LDHA activity. LDHAWT showed a faster decline in NADH absorbance at 340 nm compared with LDHAC163A (Figure ). In addition, Palmitoyl‐CoA increased LDHAWT activity, elevated Vmax and reduced Km, but failed to affect LDHAC163A (Figure ), indicating that Cys163 palmitoylation enhances LDHA catalytic efficiency and pyruvate binding. To probe the structural and functional consequences of LDHA palmitoylation at site Cys163, we performed molecular dynamics (MD) simulations comparing wild‐type and palmitoylated LDHA‐pyruvate complexes (Figure ). A 200 ns simulation revealed distinct differences in structural stability and dynamic behavior between the two complexes. The root mean square deviation (RMSD) analysis showed that the WT complex remained relatively stable throughout the simulation, with RMSD values fluctuating between 0.4 and 0.6 nm. In contrast, the LDHAC163‐Palmitoylation complex exhibited a progressive increase in RMSD, reaching up to 0.8 nm after 120 ns, indicating reduced structural stability and increased conformational fluctuations (Figure ). Consistent with this, root mean square fluctuation (RMSF) and B‐factor analyses revealed markedly enhanced flexibility in the region surrounding Cys163, as well as increased dynamic activity in the substrate‐binding domain (Figure ). These results suggest that palmitoylation introduces local structural perturbations that propagate to broader regions of the LDHA, potentially impacting its catalytic behavior. Despite decreased structural stability, MM/PBSA analysis revealed enhanced binding affinity of LDHA for pyruvate upon palmitoylation (the total binding free energy (ΔG_total) decreased from −14.04 kcal/mol in the WT complex to −15.16 kcal/mol in the palmitoylated complex, driven primarily by strengthened van der Waals interactions facilitated by the hydrophobic palmitoyl moiety (Figure ). These data indicate that palmitoylation may stabilize the enzyme–substrate interaction through favorable nonpolar contacts, despite introducing greater conformational flexibility. To validate the functional role of LDHA C163 palmitoylation, we generated LDHA‐knockdown GSCs and reconstituted them with either wild‐type or C163A‐mutant LDHA plasmids, enabling the evaluation of palmitoylation‐dependent effects on tumor cell behavior. LDHA silencing reduced LDH enzyme activity, and the recovery of LDHAWT expression exhibited restored effects, while LDHAC163A failed to rescue LDHA enzyme activity in response to PA (Figure ). Seahorse analysis was performed to assess the role of LDHA Cys163 palmitoylation in glycolysis. In LDHA‐knockdown GSC 267, re‐expression of LDHAWT restored glycolytic activity, glycolytic capacity, and glycolytic reserve upon PA treatment, whereas the LDHAC163A failed to do so (Figure ). These results indicate that Cys163 palmitoylation is essential for PA‐induced glycolytic reprogramming in GSCs. ZDHHC12‐Mediated LDHA Palmitoylation Promotes Its Exosomal Sorting in Hypoxic MES‐GSCs MES‐GBM cells exhibit intrinsic hypoxic properties (Figure and Figure ), further corroborated by spatial transcriptomics (ST) data revealing their extensive localization within tumor hypoxic niches (Figure ). Our proteomic analysis detected pronounced enrichment of LDHA in exosomes derived from hypoxia‐exposed GSCs (Figure ), validated by immunoblot analysis (Figure ). Building on established evidence that palmitoylation directly governs exosomal protein sorting [], we propose that PA facilitates exosomal packaging of LDHA through this post‐translational modification. Crucially, Duolink proximity ligation immunoblot assays demonstrated that hypoxia enhanced ZDHHC12‐LDHA binding capacity (Figure and Figure ). And PA promoted LDHA enrichment in exosomes without significantly altering exosomal morphology or size (Figure ), while ZDHHC12 knockdown attenuated PA‐induced LDHA accumulation in exosomal fractions (Figure ). And mutation of LDHA palmitoylation sites (Cys163) profoundly impaired its exosomal sorting (Figure ). In summary, PA promoted exosomal LDHA sorting via ZDHHC12‐mediated palmitoylation. Emerging evidence identifies tumor cell‐TAM interactions as critical mediators of GBM therapy resistance, with MES‐GBM orchestrating both TAM recruitment and immunosuppressive polarization [, , ]. The specific TAM subcluster engaged by MES‐GBM and molecular determinants of this interplay require systematic elucidation. Based on canonical cell‐type markers, TAMs were further clustered into myeloid‐derived macrophages (MDMs) and tissue‐resident microglias (MGs) (Figure ). Intriguingly, our data revealed significant enrichment of MDMs within hypoxic core regions and recurrent GBM specimens (Figure ). Spatial mapping detected pronounced MDM enrichment in hypoxic niches, coupled with hypoxia‐glycolysis pathway activation (Figure ). Functional enrichment analyses demonstrated that these MDMs exhibit characteristic hypoxia‐associated glycolytic signatures (Figure ). We proposed the MES1‐MDM crosstalk axis as a key orchestrator of recurrence dynamics. Further subclustering of MDMs resolved seven transcriptionally distinct subtypes (Figure and Table ), thereby specifying the intratumoral heterogeneity within these tumor‐infiltrating myeloid populations. Specifically, the hypoxia_MDM subcluster manifested a pronounced hypoxia‐responsive transcriptional program. This signature was characterized by upregulation of hypoxia activity, concomitant enhancement of glycolytic and pro‐angiogenic activities, and acquisition of epithelial‐to‐mesenchymal transition (EMT) traits (Figure ). Notably, these molecular signatures demonstrated elevated enrichment in recurrent tumors and tumor core regions (Figure ). Critically, MDM‐hypoxia subpopulation was significantly enriched in recurrent tumors and tumor core regions (Figure and Figure ). Building upon pseudotime trajectory analysis, MDM‐hypoxia was identified at the terminal differentiation state of monocytic lineages (Figure ), with progressively enhanced glycolytic activity correlating with advancing tumor malignancy, particularly prominent in recurrent tumors and hypoxic core regions (Figure ). Critically, ST profiling reveals robust co‐localization of MDM‐hypoxia with MES1‐GBM subtypes and palmitate accumulation (Figure ). This spatial concordance implicates palmitate as a pivotal driver of intercellular crosstalk within the MES1‐MDM‐hypoxia niches. Exosomal Palmitoylated LDHA Drives Macrophage Glycolytic Reprogramming and Reciprocal Macrophage–GBM Crosstalk Next, we investigated whether PA induces MDM‐hypoxia subset formation by promoting packaging of the glycolytic rate‐limiting enzyme LDHA into exosomes for targeted delivery to MDMs. Single‐cell analysis revealed elevated transcriptional levels of immunosuppressive markers CD163/CD206 in recurrent tumor specimens, with heightened expression specifically observed in the low oxygen‐MDM subclass (Figure ). To dissect GSC‐macrophage crosstalk, we established a co‐culture system (Figure ). Functional assays using PBMC‐derived macrophages revealed that GSC‐derived exosomes drove macrophage polarization toward immunosuppressive CD163+/CD206+ phenotypes, and this effect was also potentiated by PA preconditioning but abrogated upon ZDHHC12 knockdown in GSCs (Figure and Figure ). Furthermore, exosomes derived from LDHAWT‐reconstituted GSCs significantly promoted immunosuppressive macrophage polarization, whereas exosomes from LDHAC163A‐expressing cells abrogated this effect (Figure ). Notably, Seahorse metabolic analysis demonstrated that LDHAC163A exosomes exhibited markedly impaired capacity to enhance macrophage glycolysis compared to LDHAWT counterparts. Crucially, ZDHHC12 knockdown significantly attenuated the enhancement of macrophage glycolysis mediated by exosomes derived from LDHAWT reconstituted GSCs. (Figure ), establishing Cys163 palmitoylation as essential for exosome‐mediated macrophage reprogramming. In turn, exosome‐primed macrophages reciprocally promoted MES‐GSC remodeling (Figure and Figure ), suggesting a bidirectional crosstalk between GSCs and macrophages. To further dissect the macrophage‐derived signals underlying this feedback regulation, we examined inflammatory cytokine expression in exosome‐treated macrophages. GSC‐derived exosomes markedly increased the expression of IL‐1β, IL‐6, TNF‐α and TGF‐β, among which IL‐1β showed the most prominent induction; this effect was further enhanced by PA‐preconditioned GSC exosomes but attenuated by ZDHHC12 knockdown (Figure ). Consistently, ELISA confirmed increased secretion of IL‐1β from PBMC‐derived macrophages after treatment with PA‐GSC exosomes, whereas ZDHHC12 knockdown reduced IL‐1β release (Figure ). We next investigated whether macrophage‐derived IL‐1β could feed back to regulate PA uptake in GSCs. IL‐1β treatment increased p65 phosphorylation and SLC27A3 expression in GSC20 and GSC267 cells in a dose‐dependent manner at 0, 10 and 20 ng/mL (Figure ). Pharmacological inhibition of NF‐κB signaling with BMS‐345541 markedly suppressed IL‐1β‐induced p65 phosphorylation and SLC27A3 upregulation (Figure ). Functionally, IL‐1β enhanced palmitic acid uptake in GSCs, whereas BMS‐345541 largely abolished this effect (Figure ). These data suggest that exosome‐reprogrammed macrophages promote GSC palmitic acid uptake through an IL‐1β–NF‐κB–SLC27A3 feedback axis. Clinically, patients with high proportions of both MES1‐GBM cell and hypoxia‐MDM had shorter OS than did those with lower proportions of both cell subpopulations (Figure ). Moreover, intracranial co‐implantation of ZDHHC12‐depleted MES‐GSCs with human PBMC‐derived macrophages revealed that palmitoylation deficiency attenuates tumor growth (Figure ) and reduces CD163+/CD206+ immunosuppressive macrophage infiltration (Figure ), concurrently improving survival outcomes (Figure ). Notably, we obtained similar results in immunocompetent mice (Figure ). Collectively, these results indicated that PA potentiated immunosuppressive MDM‐hypoxia subset formation via ZDHHC12‐mediated palmitoylation of LDHA at Cys163, enabling its exosomal packaging for targeted metabolic reprogramming of macrophages toward a protumorigenic glycolytic state. Rutin Blocks Palmitoylation Site of LDHA, Enhancing Therapeutic Sensitivity of GBM Given the pivotal role of palmitoylated LDHA in sustaining therapeutic resistance of MES GSCs and promoting an immunosuppressive TME, we performed a structure‐based virtual screening centered on the palmitoylation site at LDHAC163 to molecularly dock a library of approximately 120 000 small molecules (Figure and Table ). We identified a series of small molecules that are likely to interact with LDHAC163 site. Among the top‐ranked candidates, surface plasmon resonance (SPR) validation revealed that Rutin exhibited the highest binding affinity to LDHA (Figure and Figure ). Cellular thermal shift assay (CETSA) demonstrated that Rutin increased the thermal stability of LDHA, providing additional evidence for a direct molecular interaction (Figure ). Rutin, also known as vitamin P, is a flavonoid compound abundantly present in herbs, fruits, and vegetables []. It possesses favorable blood–brain barrier permeability and has demonstrated antitumor effects in various malignancies, including glioma, breast cancer, liver cancer, and colorectal cancer [, , , ]. However, the underlying mechanisms of its action have not been fully elucidated. To clarify how Rutin inhibits LDHA palmitoylation, we examined whether it affects the interaction between LDHA and its palmitoyl‐transferase ZDHHC12. Co‐IP and PLA assays showed that Rutin did not substantially disrupt ZDHHC12‐LDHA binding or proximity (Figure ). However, ABE assays revealed that Rutin markedly reduced ZDHHC12‐mediated palmitoylation of LDHAWT, while LDHAC163A displayed minimal palmitoylation and was largely insensitive to Rutin (Figure ). These findings indicate that Rutin suppresses LDHA palmitoylation mainly by binding to the Cys163‐centered palmitoylation site of LDHA, thereby blocking its modification, rather than by interfering with ZDHHC12‐LDHA interaction. Treatment with Rutin significantly inhibited LDHA palmitoylation, suppressed the MES subtype, and enhanced the therapy sensitivity of GSCs (Figure and Figure ). Functionally, Rutin reversed the PA‐induced upregulation of MES markers in GSCs (Figure ). Extracellular flux analysis revealed that Rutin abrogated PA‐induced increases in glycolysis, glycolytic capacity, and glycolytic reserve, which was accompanied by a reduction in lactate production (Figure and Figure ). Additionally, in vivo 18F‐FDG PET imaging revealed that Rutin‐treated mice exhibited significantly reduced glucose uptake in intracranial tumors, indicating suppression of tumor glycolytic activity (Figure ). Moreover, Rutin significantly elevated chemoradiotherapy‐induced apoptosis, DNA damage, as evidenced by increased comet tail moments and γH2AX foci formation (Figure and Figure ), suggesting that Rutin enhances therapy sensitivity in GSCs by disrupting LDHA palmitoylation and glycolytic reprogramming. Using immunocompetent mouse models, we investigated the role of Rutin in modulating the immunosuppressive TME in GBM. Rutin administration led to a marked reduction in tumor burden and a significant extension in survival (Figure and Figure ). IF staining and Fc analysis further revealed decreased infiltration of Immunosuppressive tumor‐associated macrophages (Figure ). To further dissect the cellular landscape of the TME, we performed single‐cell transcriptomic profiling. tSNE visualization revealed a marked decrease in M2‐like immunosuppressive macrophages and proliferating tumor cells, along with an altered immune cell composition in Rutin‐treated tumors (Figure ). Collectively, these findings demonstrate that Rutin potently inhibits LDHA palmitoylation, thereby attenuating GSC chemoradiotherapy resistance and mitigating the immunosuppressive tumor microenvironment in GBM, highlighting its potential as a therapeutic agent. Rutin Enhances the Therapeutic Efficacy of Chemoradiotherapy in GBM To further assess the in vivo therapeutic potential of Rutin in combination with standard treatments, we established orthotopic glioma models and evaluated responses to radiotherapy and TMZ, with or without Rutin co‐administration (Figure ). Bioluminescence imaging revealed that the combination of IR, TMZ, and Rutin led to a significant reduction in tumor burden compared to IR/TMZ or Rutin alone (Figure ). Histological analysis further confirmed this observation, with H&E staining showing reduced tumor size in the combination group (Figure ), and γ‐H2AX IHC demonstrating enhanced DNA damage in the IR/TMZ/Rutin group (Figure ). Survival analysis demonstrated a substantial extension of overall survival in the triple‐treatment group compared to IR/TMZ or control groups, in both immunodeficient and immunocompetent mouse models (Figure ), supporting the synergistic effect of Rutin in enhancing therapeutic response. Tumor organoids serve as valuable models for drug discovery and therapeutic efficacy evaluation, providing a complementary platform to conventional animal models [, ]. Organoids treated with the triple combination therapy exhibited the highest levels of DNA damage, further validating the enhanced therapeutic efficacy of this regimen (Figure ). In addition to the conventional orthotopic model, we further established a postoperative recurrent GBM model by surgically resecting intracranial GSC‐luciferase tumors, followed by IR/TMZ and/or Rutin treatment (Figure ). IVIS imaging revealed that Rutin combined with IR/TMZ markedly delayed postoperative tumor recurrence and reduced tumor burden compared with either treatment alone (Figure ). Survival analysis further showed that the triple‐treatment group had the longest overall survival (Figure ). In conclusion, these findings systematically illustrated the proposed mechanism: MES GSCs, characterized by high fatty acid uptake and LDHA palmitoylation, exhibit elevated glycolysis and resistance to IR and TMZ. Rutin disrupts this axis by inhibiting LDHA palmitoylation, restoring therapy sensitivity and moderating the immunosuppressive TME. Collectively, these findings demonstrate that Rutin synergistically enhances the efficacy of radiotherapy and TMZ by targeting metabolic reprogramming and overcoming therapy resistance in MES‐type GBM (Figure ). Discussion Surgical resection remains the cornerstone of GBM treatment. However, the highly infiltrative nature of GBM, combined with the need to preserve functional brain regions, often precludes complete tumor removal []. As a result, postoperative adjuvant therapies are essential for eliminating residual tumor cells, and understanding resistance to such therapies is particularly critical in incompletely resectable malignancies like GBM. If surgical resection represents a coordinated assault, postoperative chemoradiotherapy serves to eliminate residual stragglers. However, a subset of tumor cells either intrinsically resists therapy or acquires resistance through treatment‐induced selective pressure or mutagenesis. These resistant populations ultimately drive GBM recurrence and treatment failure. To better model this process, we established orthotopic intracranial GBM patient‐derived xenograft (PDX) models using tumor tissues from GBM patients and administered a standard chemoradiotherapy regimen. Multi‐omics analysis of GBM cells that had acquired therapy resistance revealed a selective enrichment of the MES subtype, consistent with previous reports implicating MES GSCs in driving treatment resistance and postoperative recurrence of GBM []. Through integrated data analysis and experimental validation, we found that both therapy resistant cells and MES GSCs exhibited markedly elevated fatty acid transport activity. This metabolic feature suggests a dependency on exogenous lipid uptake to support survival under therapeutic stress, prompting us to investigate the downstream consequences of fatty acid–driven metabolic reprogramming in MES GSCs. Building on this link between fatty acid metabolism and therapy resistance, we further investigated the downstream molecular mechanisms by which elevated lipid uptake promotes chemoradiotherapy resistance in GBM. A metabolite screen of fatty acid derivatives identified PA as a key molecule capable of attenuating the cytotoxic effects of therapy on GSCs. Additionally, multi‐omics analysis, through mapping metabolites to therapy‐resistant regions, further validated that PA drives MES GBM progression and therapeutic resistance. PA is the most abundant dietary saturated fatty acid, commonly derived from animal fat and palm oil, and serves not only as a critical component of membrane phospholipids and an energy substrate via β‐oxidation, but also as a signaling molecule implicated in inflammation, immune modulation, and cancer progression. Recent study has shown that mice subjected to a short‐term palm oil–rich diet, as well as tumor cells briefly exposed to PA in vitro, exhibited sustained high metastatic potential []. Another study revealed that intra‐tumoral accumulation of PA drives extracellular matrix remodeling by promoting the transition of normal fibroblasts into matrix‐associated cancer‐associated fibroblasts via tumor cell–derived cytokine signaling, thereby enhancing tumor progression and tissue stiffening []. While PA is widely utilized by tumor cells as a key metabolic substrate, it has also been associated with lipid peroxidation and ferroptosis under certain conditions. In our study, however, exogenous PA pre‐administration promoted the survival and therapy resistance of GSCs, suggesting that its intracellular fate may be context‐dependent. Rather than contributing to oxidative lipid stress, PA may be selectively directed toward alternative pathways, such as protein palmitoylation, that support tumor adaptation during therapeutic challenge. Consistent with this hypothesis, our experiments demonstrated that PA enhances therapy resistance in GSCs primarily through protein S‐palmitoylation. The study of protein palmitoylation has long been limited by the absence of a consensus substrate motif and the reliance on low‐sensitivity, labor‐intensive methods such as radioactive [3H]‐palmitate labeling. The development of the acyl‐biotin exchange (ABE) assay has markedly improved the efficiency and sensitivity of palmitoylation detection [, , ]. Using the ABE assay on GBM tissue samples and GSCs, we identified a significant enrichment of palmitoylated proteins within the glycolytic pathway, suggesting a strong association between PA‐mediated palmitoylation and metabolic reprogramming in GBM. Further analysis revealed ZDHHC12 as a key palmitoyl‐transferase responsible for regulating GSC glycolysis and resistance to chemoradiotherapy. ZDHHC12 has been reported to palmitoylate NLRP3, promoting its degradation via the chaperone‐mediated autophagy pathway [, , ]. Additionally, ZDHHC12‐mediated palmitoylation of MAVS has been shown to facilitate its activation in antiviral innate immune responses []. However, the functional role of ZDHHC12 in cancer, particularly in glioma, remains largely unexplored. In this study, we demonstrate that ZDHHC12 catalyzes the palmitoylation of LDHA, thereby promoting glycolytic reprogramming and contributing to therapy resistance in GBM. Palmitoylation enhances the hydrophobicity of substrate proteins, thereby influencing their subcellular localization, stability, enzyme activity, membrane association, and protein–protein interactions. We systematically investigated each of the above functional roles and ultimately found that ZDHHC12‐mediated palmitoylation primarily regulates the enzymatic activity of LDHA and facilitates its translocation into exosomes. The former promotes maintenance of the MES subtype of GSCs and attenuates the cytotoxic effects of chemoradiotherapy by enhancing glycolytic metabolism, while the latter contributes to the establishment of an immunosuppressive TME by modulating TAM in GBM. Recent studies have revealed that TAMs secrete extracellular vesicles enriched in LDHA, which enhance glycolysis, proliferation, and survival of GBM cells []. In contrast, we identified that palmitoylated LDHA is also secreted by GSCs via exosomes, where it contributes to the induction of an immunosuppressive phenotype in infiltrating macrophages. Notably, targeting LDHA palmitoylation significantly disrupted the crosstalk between TAMs and GBM cells. Therefore, our study provides important insights into the regulatory mechanisms and functional roles of LDHA in GBM, while also highlighting the therapeutic potential of targeting its palmitoylation. Importantly, our study reveals LDHA palmitoylation as a previously unrecognized mechanism linking fatty acid uptake to metabolic reprogramming and therapeutic resistance in GBM. Rather than serving merely as an energetic substrate or membrane component, PA promotes ZDHHC12‐mediated LDHA palmitoylation, thereby enhancing LDHA enzymatic activity, sustaining the MES phenotype, and facilitating exosomal LDHA release to remodel the immunosuppressive TME. These findings extend the functional significance of LDHA beyond canonical glycolysis and identify LDHA palmitoylation as a metabolic–immune regulatory node in GBM recurrence. Therapeutically, targeting this modification may selectively disrupt the therapy‐resistant LDHA pool and provide a promising strategy to enhance the efficacy of standard chemoradiotherapy. Given the pathological significance of LDHA palmitoylation in GBM, we sought to identify pharmacological agents capable of disrupting this modification. Previous studies have shown that competitive inhibitors, such as peptides containing palmitoylation motifs, can block the palmitoylation of target proteins [, , ]. However, the clinical translation of such short peptides remains challenging. In our study, we identified cysteine 163 (C163) as the critical site for LDHA palmitoylation. Based on this finding, we performed structure‐based screening and identified Rutin as a small molecule capable of occupying the C163 site, thereby inhibiting LDHA palmitoylation and enzymatic activity. Rutin is a flavonoid compound with favorable blood–brain barrier permeability, and has been reported to exert antioxidant, anti‐inflammatory, anti‐tumor, and anti‐aging effects [, , , , ]. Its multifunctional properties and bioavailability make it a promising candidate for therapeutic applications in central nervous system disorders. Recent studies developed a hybrid layered composite nanofiber mesh co‐loaded with TMZ and Rutin, which exhibited a pronounced antitumor effect [, ]. However, the precise molecular mechanism underlying Rutin's activity remains unclear. Our findings demonstrate that Rutin blocks LDHA palmitoylation and inhibits its enzymatic function, thereby enhancing tumor cell sensitivity to chemoradiotherapy and exerting potent antitumor effects in combination therapy in vitro and in vivo. In conclusion, this study demonstrates that PA accumulation is a defining feature of the hypoxic pseudopalisading necrosis regions in GBM, where it drives MES subtype maintenance and therapeutic resistance. Mechanistically, MES GSCs preferentially uptake PA, which promotes LDHA palmitoylation via the ZDHHC12. Palmitoylated LDHA operates via two distinct mechanisms: by enhancing enzymatic activity to promote glycolytic reprogramming, and by being secreted in exosomes to regulate TAM. Furthermore, targeting the palmitoylation site of LDHA (Cys163) with the small molecule Rutin significantly reduced LDHA palmitoylation and increased the chemoradiotherapy sensitivity of MES GSCs. These findings uncover a previously unrecognized regulatory mechanism of LDHA palmitoylation in GBM and highlight a potential therapeutic strategy for overcoming treatment resistance. Experimental Section Human Samples Human glioma tissue samples were collected from patients undergoing surgical resection at Qilu Hospital of Shandong University. The use of human samples and associated clinical data was approved by the Medical Ethics Committee of Qilu Hospital, Shandong University. All participants provided written informed consent, and the study was approved by the Ethics Committee on Scientific Research of Shandong University Qilu Hospital (approval number: KYLL‐2023(ZM)‐657). Tumor tissues were used for histological evaluation, IHC, and molecular analyses, including RNA sequencing and other downstream assays. Cell Lines and Cell Culture Patient‐derived PN‐GSCs (GSC 8‐11, GSC 11), MES‐GSCs (GSC 20, GSC 267, GSC 28), and neural progenitor cells (NPCs) were generously provided by Dr. Frederick F. Lang and Dr. Krishna P.L. Bhat (MD Anderson Cancer Center, University of Texas). GSC subtypes were classified based on the genetic signatures described by Phillips and Verhaak, which are widely accepted and used for GBM subtype study. GSCs and NPCs were dissociated into single cells using Accutase (A6964, Sigma‐Aldrich) and cultured in DMEM/F12 (A4192001, Gibco) supplemented with B‐27 (17504044, Gibco), 20 ng/mL recombinant human EGF (236‐EG, R&D Systems), and 20 ng/mL bFGF (3718‐FB, R&D Systems). The human GBM cell lines U87MG (RRID: CVCL_0022) and U251 (RRID: CVCL_0021) (Chinese Academy of Sciences Cell Bank) and mouse GBM cell line GL261 (RRID: CVCL_Y003, ATCC) were cultured in DMEM (D6429, Sigma) supplemented with 10% FBS (A5670701, Thermo Fisher Scientific). The THP1 cell lines (RRID: CVCL_0006, Chinese Academy of Sciences Cell Bank) were cultured in RPMI1640 (R8758, Gibco) with 10% FBS. All cells were maintained at 37°C in a humidified incubator with 5% CO2 and 5% O2. Single Cell Data Preprocess Raw FASTQ reads were aligned to the GRCh38/hg38 reference genome with Cell Ranger v3.0 (10× Genomics). A Cell Ranger–compatible reference was constructed following the 10× Genomics guidelines. The Cell Ranger pipeline then carried out read alignment and quality filtering, demultiplexed cellular barcodes, and quantified UMIs. After mapping, digital gene expression (DGE) matrices were produced for each individual sample as well as for the combined dataset. Sample‐level DGE outputs were integrated using Cell Ranger's aggr function, with libraries depth‐normalized by balancing effective read counts across batches. This yielded a unified expression matrix containing all cells and their transcript counts across genes. TSNE Visualization and Determination of the Major Cell Types After quality control, downstream expression analysis and cell‐type assignment were performed in R using Seurat. Because the input counts had already been processed/normalized upstream, the data were imported into Seurat without additional normalization, scaling, or centering steps. Per‐cell metadata (including spatial/location information) were incorporated alongside the expression matrix. Highly variable genes were then selected and used for principal component analysis (PCA). The PC loadings and associated genes were evaluated to choose the number of components retained for subsequent analyses. Cells were visualized in two dimensions using t‐SNE based on the top 20 principal components. Cluster‐specific differentially expressed genes were identified with Seurat's FindAllMarkers, requiring detection in at least 10% of cells, a log fold‐change > 0.25, and statistical significance (adjusted p < 0.05; Wilcoxon rank‐sum test). Finally, clusters were assigned to known cell types by comparing their marker profiles with established canonical markers. Pseudotime Analysis Single‐cell trajectories were inferred using Monocle 2 on the cell‐by‐gene expression matrix. Genes showing substantial variation across cells (or differential expression across states) were selected as ordering features to reconstruct the dynamic trajectory and arrange cells along pseudotime. Briefly, gene‐level transcript abundance was quantified for each cell, and genes were prioritized by their relationship between coefficient of variation and mean expression. For each cell type, the top 20 ranked genes were retained as trajectory features. In parallel, RNA velocity estimates were computed and overlaid onto the UMAP embedding generated in Seurat to visualize directional transitions among cell states. Functional Analysis of Differential Genes Differential gene expression across cell types was assessed in Seurat using the Wilcoxon rank‐sum test. To estimate pathway activity at single‐cell resolution, gene set variation analysis (GSVA) was performed with default parameters using the GSVA package. Gene set enrichment analysis (GSEA) was conducted with GSEA software based on curated gene sets from the Molecular Signatures Database (MSigDB). Spatial Transcriptomics Data Processing Briefly, data were normalized with SCTransform, followed by dimensionality reduction with PCA and UMAP. Clustering was performed using the first 20 principal components at the default resolution. Spatial expression patterns were visualized using SpatialFeaturePlot, and signature scores for MES‐like malignant cells and monocyte‐derived macrophages (MDMs) were computed using AddModuleScore. Spatial Metabolomics Sample preparation: Tissues were stored at −80°C prior to analysis. Tissues were placed at −20°C 30 min before use. Before sectioning, fix the frozen tissue samples onto the tray with water. The tissues were sectioned at 10 μm thickness using a Leica CM1950 cryostat (Leica Microsystems GmbH, Wetzlar, Germany) at −20°C. Subsequently, the cut tissue sections were attached to the adhesive microscope slides, and the slides with tissue sections were dried in a vacuum desiccator for 30 min. AFADESI Parameter Settings: Place the adhesive microscope slides with tissue sections on the AFADESI mobile platform. Determine the scanning area according to the size of the tissue, and the scanning is carried out in a line‐by‐line scanning manner. The composition of the spraying solvent is acetonitrile: water = 90:10 (V/V), and the flow rate of the spraying solvent is 3 μL/min. The pressure of the spraying gas is 0.65 Mpa. The spraying voltage is 4500 V in the positive mode and −4000 V in the negative mode. Mass spectrometry imaging: The metabolites in the samples were imaged using the AFADESI imaging platform. The mass spectrometry scanning range was 100–1300 Da. The Massimager software was used to process the data and extract the peaks. The substances of the mass spectrometry peaks are identified by matching the primary molecular weights with those in the database. Antibodies and Reagents The following antibodies were generated by Cell Signaling: anti‐CD44 (37259, RRID: AB_2750879); anti‐LYN (2796, RRID: AB_2138391); anti‐γH2AX (Ser139) (9718, RRID: AB_2118009); anti‐PKM2 (4053, RRID: AB_1904096); anti‐Flag (D6W5B) (14793, RRID: AB_2572291); anti‐β‐Tubulin (2128, RRID: AB_823664); anti‐CD63 (52090, RRID: AB_2924771); anti‐Cd163 (93498, RRID: AB_2800204); anti‐Cd206 (24595, RRID: AB_2892682) and anti‐Caspase3 (9662, RRID: AB_331439). The following antibodies were generated by Proteintech: anti‐LDHA (19987‐1‐AP, RRID: AB_10646429); anti‐Vimentin (10366‐1‐AP, RRID: AB_2273020); anti‐β‐Actin (66009‐1‐Ig, RRID: AB_2687938) and anti‐SLC27A3 (12943‐1‐AP, RRID: AB_2286200). The anti‐ZDHHC12 (PA5‐113144) was obtained from Invitrogen. For western blots, antibodies were diluted 1:1000. For immunofluorescence, antibodies were diluted 1:200. For IHC, antibodies were diluted 1:100. Rutin (HY‐N0148), Etomoxir (HY‐50202) and TMZ (HY‐17364) were purchased from MedChemExpress. Palmitic acid (P0500) and 2‐BP (21604) were purchased from Sigma‐Aldrich. BODIPY FL C16 (D3821) was purchased from Invitrogen. Fatty Acid Uptake Test For fluorescence imaging of fatty acid uptake, 2 × 104 indicated GSC cells were seeded into pre‐coated 8‐well chamber slides (ibidi) and incubated with 1 μM BODIPY FL C16 in culture medium for 12 h. Cells were then washed three times with PBS, and fluorescent fatty acid signals were visualized using a Leica TCS SP8 confocal microscope. For the flow cytometric analysis of fatty acid uptake, 2 × 105 GSC cells of a defined subtype were seeded into 12‐well plates and treated with BODIPY FL C16 (final concentration: 25 μM) for 30, or 60 min. After three washes with PBS, fatty acid uptake was quantified by flow cytometry (Agilent Technologies). Fatty acid uptake was also assessed using the Fatty Acid Uptake Kit (MAK156, Millipore Sigma) following the manufacturer's instructions. GSC cells were seeded into 96‐well plates at a density of 1 × 105 cells per well. The culture medium was replaced with 100 μL of Fatty Acid Dye Loading Solution, and fluorescence intensity (λex = 485 nm, λem = 515 nm) was measured after 1, 10, and 30 min of incubation using a BioTek Synergy plate reader. Xenograft Models All animal studies were performed in accordance with Institutional Animal Care and Use Committee (IACUC) guidelines and were approved by the IACUC of the Qilu Hospital of Shandong University (Approval No. DWLL‐202400062). Male C57BL/6 (RRID: MGI: 2159769) or BALB/c nude (RRID: MGI: 2161072) mice (aged 4–6 weeks), were obtained from GemPharmatech Co., Ltd. (Nanjing, Jiangsu, China) and housed at the Neurosurgery Laboratory of Qilu Hospital of Shandong University (Jinan, Shandong, China). To evaluate the efficacy of standard in vivo chemoradiotherapy in eliminating therapy‐sensitized GBM progenitor cells, PDX models were established in BALB/c nude mice. 7 days after tumor implantation, mice were randomly assigned to either a control group or an IR+TMZ group. Mice in the IR+TMZ group received irradiation at 2 Gy per fraction, once daily for five consecutive days, with synchronous TMZ administration at 25 mg/kg daily via intraperitoneal injection. At 28 days post‐tumor implantation, GBM tissues were harvested for in vitro culture or subjected to deep sequencing. To investigate palmitate diet–induced chemoradiotherapy resistance in vivo, BALB/c nude mice were randomized into a high‐palmitic acid diet group (supplemented with palm oil, TD150067, Envigo) and a normal diet group prior to intracranial tumor implantation. Luciferase‐expressing GSC 267 cells (5 × 105 per mouse) were implanted intracranially. 7 days after tumor seeding, mice were further randomized into control and IR+TMZ groups. Mice in the IR+TMZ group received irradiation at 2.5 Gy per fraction every other day for a total dose of 7.5 Gy, combined with daily TMZ treatment (25 mg/kg, i.p.) from day 10 to day 14. Tumor progression was monitored using the IVIS Spectrum in vivo imaging system (PerkinElmer Inc., Waltham, MA, USA) on days 7 and 28 post‐implantation. In animal experiments involving Rutin combined with standardized treatment, mice were acclimated and then orthotopically implanted with GSC 267 on Day 0. Rutin treatment (100 mg/kg, oral gavage, once daily) started on Day 7 post‐implantation and continued until Day 28. Mice in the IR&TMZ and IR&TMZ&Rutin groups received TMZ (25 mg/kg, i.p., daily) from Day 10 to 14 and focal cranial irradiation (2.5 Gy per session) on Days 10, 12, and 14. Bioluminescent imaging (IVIS) was performed on Days 7 and 28 to monitor tumor burden. All mice were housed in a specific pathogen‐free (SPF) barrier facility with individually ventilated cages. Anesthesia was administered using isoflurane (#R510‐22‐10, RWD Life Science), with induction in a 2.0%–2.5% isoflurane/oxygen mixture and maintenance in 1.5%–1.8% isoflurane/oxygen. Moribund animals were identified based on the following criteria: (1) difficulty in breathing, feeding, or drinking; or (2) ≥15% body weight loss within 4 days. To minimize suffering, animals were exposed to high‐concentration CO2 (30%) for at least 3 min until complete cessation of respiration was observed. Isolation and Culture of Primary GBM Cells Surgically resected GBM specimens from patients at Qilu Hospital, Shandong University, China, as well as tumor samples extracted from the brains of mice, were processed similarly. Briefly, after washing with precooled DMEM containing streptomycin and penicillin, the tumor tissues were gently minced and digested using the tissue enzyme‐lysis kit (DHTEH‐2505. RWD Life Science) following the instructions. Tumor cells were then filtered through a 200‐mesh filter to obtain a single‐cell suspension, and then erythrocytes were removed using erythrocyte lysate (WL1000, R&D Systems). Finally, the primary GBM cells were cultured in serum‐free medium (DMEM/F12, 2% B‐27, 20 ng/mL EGF, 20 ng/mL bFGF). Chemoradiotherapy Treatment Radiation treatment for both in vitro and in vivo experiments was performed using an X‐ray irradiator (X‐Rad225 OptiMAX, PXi). And TMZ (HY‐17364) were purchased from MedChemExpress. For in vitro assays, tumor cells were exposed to ionizing radiation (6 Gy) and TMZ (200 μM), and subsequent functional analyses were carried out 48 h after chemoradiotherapy. For in vivo studies, mice received irradiation at the indicated dose and were treated with TMZ (25 mg/kg) between days 10 and 14 following tumor cell implantation. A detailed schematic of the treatment regimen is shown in the figure. RNA‐seq and Data Analysis 1. Sample Preparation and Library Construction. Total RNA was extracted and assessed for integrity and purity according to standard protocols. Messenger RNA (mRNA) was enriched using poly‐T oligo‐attached magnetic beads. For non‐strand‐specific libraries, cDNA was synthesized using random hexamer primers for both strands. For strand‐specific libraries, the second strand was synthesized using dUTP to maintain directionality. Libraries were prepared through end repair, A‐tailing, adaptor ligation, size selection, and PCR amplification. Library quality and concentration were evaluated using Qubit, real‐time PCR, and Bioanalyzer. 2. Sequencing. Qualified libraries were pooled based on concentration and expected data yield, and subjected to Illumina sequencing. Sequencing was performed using the Sequencing by Synthesis (SBS) strategy, where fluorescently labeled nucleotides were incorporated and imaged in real‐time to generate raw reads. 3. Bioinformatics Analysis. Quality Control: Raw reads were processed using fastp to remove adaptors, poly‐N sequences, and low‐quality reads. Quality metrics including Q20, Q30, and GC content were calculated. Read Mapping: Clean reads were aligned to the reference genome using HISAT2 (v2.2.1, RRID: SCR_015530) with the corresponding gene annotation file to ensure splice‐aware alignment. Expression Quantification: Gene‐level counts were obtained using featureCounts (v2.0.6). Gene expression was normalized as FPKM (Fragments Per Kilobase of transcript per Million mapped reads). Differential Expression Analysis: For datasets with biological replicates, differential expression analysis was performed using DESeq2 (RRID: SCR_000154, v1.42.0) based on a negative binomial distribution. For datasets without replicates, edgeR (v4.0.16) was used with scaling normalization. Differentially expressed genes (DEGs) were identified using a threshold of padj ≤ 0.05 and | log2(fold change) | ≥ 1. Public Data Collection and Analysis Transcriptomic data and corresponding clinical information for TCGA GBM were obtained from the GDC Data Portal (https://portal.gdc.cancer.gov/). RNA‐seq data and clinical annotations for CGGA GBM samples were downloaded from the CGGA database (http://www.cgga.org.cn/). Transcriptome profiles for the Rembrandt (GSE108476) and Gravendeel (GSE16011) GBM cohorts were retrieved from the GEO database (https://www.ncbi.nlm.nih.gov/geo/). Single‐cell RNA sequencing data were obtained from GSE131928. The fatty acid transport activity gene signature (GO:0015908) was obtained from the GSEA website (https://www.gsea‐msigdb.org/gsea/index.jsp, RRID: SCR_016863). Single‐cell RNA‐seq data were analyzed using the R package Seurat (v4.1.0), and t‐SNE was applied for visualization of distinct cell clusters. The R package irGSEA was used to calculate and visualize enrichment scores for the FAT signature using the “UCell” method. For GSEA analysis of TCGA bulk RNA‐seq data, gene signatures were also downloaded from the GSEA website, and enrichment analysis was performed using GSEA software (v4.1.0). Plasmid, siRNA and Cell Transfection All shRNA and overexpression plasmids were purchased from GenePharma (Shanghai, China). Selected shRNA sequences were cloned into the pLKO.1‐puro lentiviral backbone. Recombinant lentiviral particles were generated by co‐transfecting 293T cells (RRID: CVCL_0063) with 5 μg of the target plasmid, 3.25 μg of psPAX3, and 1.75 μg of pMD2.G packaging plasmids using linear polyethylenimine (PEI; 23966, Polysciences). Lentiviral supernatants were harvested 48 h after transfection, filtered through a 0.45 μm membrane (Corning), and used to infect target cells in the presence of 8 μg/mL polybrene. After 24 h of infection, cells were selected with 2 μg/mL puromycin (P8833, Sigma‐Aldrich) to establish stable lines. For transient gene silencing or overexpression, cells were transfected with siRNAs or plasmids using the Lipofectamine 3000 reagent (L3000015, Invitrogen), following the manufacturer's protocol. The RNA‐interfering sequences used in this study are listed in Table . HR Reporter Assays To evaluate the effect of PA on homologous recombination (HR) repair of DNA double‐strand breaks (DSBs), GSC 267 cells stably integrated with DR‐GFP plasmid (RRID: Addgene_26475). Following 48 h of PA treatment, approximately 5 × 105 GSC 267 DR‐GFP cells were transfected with 1 μg of the pCMV‐I‐SceI using Lipofectamine 3000 Transfection Kit (Invitrogen). GFP fluorescence was measured 3 days post‐transfection using a flow cytometer (Agilent Technologies). HR repair efficiency was determined by calculating the percentage of GFP‐positive cells among the total number of analyzed cells. Western Blotting (WB), Immunohistochemistry, and Immunofluorescence (IF) For Western blotting, GSCs or GBM tissues were lysed in RIPA buffer (9806, Cell Signaling Technology) supplemented with 1% protease and phosphatase inhibitors (IKM1010, Solarbio). Protein lysates were separated via SDS‐PAGE and transferred to PVDF membranes. Membranes were incubated with primary antibodies overnight at 4°C, followed by incubation with appropriate secondary antibodies at room temperature for 1 h. Protein signals were detected using an Odyssey fluorescence scanner (ChemiDoc XRS+, Bio‐Rad, USA). For IHC, tissue sections from GSC xenograft‐bearing mice were stained with anti‐γH2AX or anti‐ZDHHC12 primary antibodies. Images were captured using a light microscope and quantitatively evaluated based on the German IHC scoring system. For IF, GSCs were seeded and fixed in μ‐slide 8‐well plates (ibidi). Cells were permeabilized with 0.3% Triton X‐100 and blocked with 5% goat serum. After overnight incubation with primary antibodies at 4°C, cells were stained with fluorescent secondary antibodies and DAPI (D9542, Sigma‐Aldrich) for 1 h and 15 min, respectively. Cells were washed three times with PBS after each step. Representative images were acquired using a Leica SP8 confocal microscope. Apoptosis Assay Tumor cells from different experimental groups were harvested and enzymatically dissociated into single‐cell suspensions using Accutase, followed by gentle washing with PBS to avoid mechanical damage caused by excessive pipetting. Apoptosis was assessed using the Annexin V‐FITC Apoptosis Detection Kit (556547, BD Biosciences), following the manufacturer's protocol. Briefly, cells were resuspended in 100 μL of binding buffer, and 2.5 μL of Annexin V‐FITC and 5 μL of PI were added. The mixture was incubated for 15 min at room temperature in the dark. Following incubation, cells were washed once with PBS to terminate the staining reaction and resuspended in 300 μL of flow cytometry buffer. Samples were analyzed using a flow cytometer (Agilent Technologies), and the percentages of apoptotic cells were quantified. Screening for Fatty Acid Metabolites Associated With Therapy Resistance GSC 267 cells were treated with a human endogenous fatty acid compound library (MedChemExpress) containing about 200 metabolic molecules at concentrations of 100 μM. Following treatment, cells were divided into two groups: a control group and chemoradiotherapy group (6 Gy + 200 μM TMZ). 48 h post‐treatment, cell viability was assessed using a standard viability assay. The relative cell survival was determined by calculating the ratio of cell viability in treatment group versus control cells for each fatty acid treatment: GSC (FA+T) / GSC (FA). Alkaline Comet Assay The comet assay was conducted using the Single Cell Gel Electrophoresis Assay Kit (Catalog #4250‐050‐K, R&D Systems,) in accordance with the manufacturer's instructions. GSCs from different groups were resuspended in PBS at a concentration of 1 × 105 cells/mL. A 10 μL aliquot of the cell suspension was mixed with 100 μL of 0.5% low melting point agarose (LMA) dissolved in calcium‐ and magnesium‐free phosphate‐buffered saline (PBS) at 37°C. Then, 50 μL of the mixture was dropped onto a microscope slide and allowed to solidify at 4°C for 10 min. The slides were then immersed in freshly prepared lysis buffer and incubated overnight at 4°C. After lysis, the slides were washed twice with PBS. An alkaline unwinding buffer was prepared using 0.4 g NaOH, 250 μL of 200 mM EDTA, and double‐distilled water to a final volume of 50 mL, then cooled to room temperature. The slides were incubated in this buffer for 20 min at room temperature in the dark to unwind the DNA and expose single‐strand breaks, facilitating DNA migration during electrophoresis. The electrophoresis buffer was prepared by mixing 8 g NaOH and 2 mL of 500 mM EDTA in 1 L of double‐distilled water and cooled to 4°C. Slides were placed in a horizontal electrophoresis tank filled with 850 mL of the cold alkaline buffer and subjected to electrophoresis at 25 V / 300 mA for 30 min. After electrophoresis, the slides were removed, and excess buffer was gently blotted with filter paper. Slides were washed twice with PBS and immersed in 70% ethanol for 5 min, followed by a final PBS rinse. Slides were then air‐dried in a 37°C oven for 10–15 min. Each slide was stained with 50 μL of SYBR Green solution in the dark for 30 min. After a PBS rinse, slides were observed and imaged under a fluorescence microscope (Nikon eclipse ti2). The tail moment was calculated using the OPENCOMET plugin for Fiji (RRID: SCR_002285). Acyl‐Biotin Exchange Assay The ABE assay was performed to detect protein palmitoylation as previously described. Lysis buffer was freshly prepared by mixing 10 mL of base lysis buffer with 100 μL PMSF and 100 μL protease/phosphatase inhibitor cocktail (IKM1010, Solarbio). N‐ethylmaleimide (NEM, 04260, Sigma) was dissolved in anhydrous ethanol to prepare a 2 M stock solution. For each sample, 20 mL of lysis buffer was mixed with 0.5 mL of 2 M NEM to block free thiol groups. Cells were lysed in 500 μL of the above buffer and incubated on ice for 10 min. Protein concentrations were quantified using the BCA assay to ensure equal protein input across groups (minimum of 500 μg per sample). Samples were incubated with 5 μg of primary antibody overnight at 4°C with gentle agitation. The next day, 50 μL of Protein A/G magnetic beads (88802, Thermo Fisher Scientific) were added to each sample and incubated for 1 h at 4°C with rotation. Tubes were gently inverted or vortexed briefly to mix, and magnetic separation was used to collect beads. The beads were then resuspended in 600 μL of lysis buffer supplemented with 10 mM NEM. From this suspension, 200 μL of bead‐bound lysate was transferred to a new microcentrifuge tube and labeled as the hydroxylamine (HAM, 159417, Sigma) (–) group; 300 μL of lysis buffer containing 10 mM NEM was added to this tube. The remaining 400 μL of bead‐bound lysate was mixed with 100 μL of the same buffer and labeled as the HAM (+) group. Both groups were incubated on ice for 10 min. Beads were washed once with 0.5 mL of buffer. HAM elution buffer (1 M HAM, freshly prepared in lysis buffer) was added to the HAM (+) samples, while lysis buffer alone was added to the HAM (–) controls. Samples were incubated on a rotator at room temperature for 45 min. Biotin‐BMCC (21900, Thermo Fisher Scientific) was prepared by dissolving 2.1 mg in 0.5 mL of DMSO to yield an 8 mM stock solution. Beads were washed again with lysis buffer using magnetic separation, then incubated with 0.5 mL of biotin‐BMCC working solution (5 μM in lysis buffer) at 4°C for 45 min with gentle mixing. Biotinylated proteins were then incubated with streptavidin–agarose beads (SA10004, Thermo Fisher Scientific) to capture palmitoylated proteins via the strong biotin–streptavidin interaction. Bound proteins were subsequently eluted and purified for downstream analysis such as immunoblotting. Micro‐PET/CT Imaging of Mice The indicated GSC 267 cells (5 × 105, 10 uL) were injected into the brains of BALB/c nude mice (4–5 weeks old). PET imaging was performed 14 days after tumor cells inoculation. The tumor‐bearing mice were deprived of food and water for 8–10 h before micro‐PET/CT imaging. The mice were subsequently injected with approximately 300 μCi of 1 8F‐FDG via the tail vein, followed by dynamic PET imaging acquisition every 5 min using a PET/CT/FMT system (MadicLab PSA094). All animals were anesthetized with isoflurane (2%–3% in 1 L/min oxygen) and placed on a heating pad to maintain body temperature throughout the imaging procedure. Respiratory rate and signs of distress were continuously monitored. Data acquisition began immediately following intravenous tracer injection. Micro‐PET/CT images were analyzed using PMOD software (version 4.0; PMOD Technologies Ltd., Zurich, Switzerland). Each PET image was manually co‐registered to the Mirrione mouse brain template using rigid body transformation, and the resulting parameters were applied to the corresponding PET datasets. Inveon Research Workplace software was used to quantify the percentage of injected dose per gram (%ID/g) and the SUV. The SUVmax data were plotted and analysed. Seahorse XF Glycolysis Stress Test The Seahorse XF Glycolysis Stress Test was performed using the Agilent Seahorse XF Glycolysis Stress Test Kit (103020‐100), following the manufacturer's protocol. 1 day before the assay, the Seahorse XFe96 Analyzer (Agilent Technologies) was powered on for equilibration, and the sensor cartridge was hydrated overnight in Seahorse XF Calibrant. XF96 cell culture microplates were coated with poly‐L‐ornithine (PLO, 1:10 dilution) and fibronectin (FN, 1:40 dilution) at 4°C overnight. On the day of the experiment, Seahorse XF Base Medium was supplemented with 2 mM L‐glutamine and adjusted to pH 7.4 to prepare the assay medium. GSCs were resuspended in assay medium and seeded into the coated XF96 plate at 1 × 104 cells/well in 50 μL volume. The plate was incubated in a 37°C non‐CO2 incubator for 25–30 min to allow attachment, followed by the addition of 130 μL pre‐warmed assay medium per well and further incubation for 15–25 min. The following compounds were prepared and loaded into the injection ports of the hydrated sensor cartridge: Port A: 20 μL of 100 mM glucose (final concentration: 10 mM), Port B: 22 μL of 10 μM oligomycin (final concentration: 1 μM), Port C: 25 μL of 500 mM 2‐deoxy‐D‐glucose (2‐DG) (final concentration: 50 mM). After calibration, the sensor cartridge was loaded onto the cell plate, and the assay was conducted using the Seahorse XFe96 Analyzer. Following the assay, cells were lysed, and protein concentration was quantified for normalization. Data were analyzed using Wave software (Seahorse Wave Controller 2.6). Co‐Immunoprecipitation Assay Co‐IP was performed using the Pierce Co‐IP Kit (88804, Thermo Fisher Scientific) according to the manufacturer's instructions. GSC cells or GBM tissues were lysed in 500 μL of RIPA buffer supplemented with phenylmethanesulfonyl fluoride (PMSF; ST506, Beyotime), followed by centrifugation at 12 000 × g for 15 min at 4°C. The resulting supernatants were incubated with magnetic beads and specific antibodies at 4°C for 12 h. Beads were subsequently washed four times with Co‐IP buffer, and the bound proteins were eluted and subjected to Western blot assay or used for mass spectrometry. Lactate Dehydrogenase Activity Assay LDH enzymatic activity was measured using the Lactate Dehydrogenase Activity Assay Kit (MAK066, Sigma‐Aldrich) according to the manufacturer's protocol. To generate the standard curve, 0, 2, 4, 6, 8, and 10 μL of 1.25 mM NADH standard solution were added to a 96‐well plate, and the volume in each well was adjusted to 50 μL with LDH assay buffer. GSCs from different treatment groups were resuspended in LDH assay buffer at a concentration of 1 × 106 cells per 500 μL. The suspension was centrifuged at 10 000 × g for 15 min at 4°C to remove insoluble debris. Then, 10 μL of the supernatant was transferred to a 96‐well plate, and the volume was brought up to 50 μL with assay buffer. Next, 50 μL of reaction mix—consisting of 48 μL LDH assay buffer and 2 μL substrate mix—was added to each well, yielding a total reaction volume of 100 μL. After 2–3 min of pre‐incubation, absorbance at 450 nm (A450) was measured using a microplate reader. The plate was incubated at 37°C, and A450 was recorded every 5 min in the dark. Measurements continued until the absorbance approached or exceeded the maximum value of the NADH standard curve. LDHA activity was calculated by determining the change in absorbance (ΔA450) over time and referencing the standard curve. Enzymatic activity was expressed as the amount of NADH consumed (or generated) per unit time. In Vitro LDHA Enzymatic Activity Assay Plasmids encoding LDHAWT‐Flag and LDHAC163A‐Flag were transfected into 293T cells. 48 h post‐transfection, cells were lysed and LDHA proteins were immunoprecipitated using anti‐Flag affinity beads. Bound proteins were eluted with elution buffer containing 20 mM HEPES, 150 mM NaCl, 1 mM EDTA, and 200 μg/mL 3×Flag peptide (pH 7.5). For the enzyme activity assay, 1 μg of purified LDHA protein was added to a reaction buffer consisting of 0.2 M Tris‐HCl (pH 7.4), 30 mM pyruvate, and 2 mM NADH. The decrease in NADH absorbance was monitored at 340 nm using a microplate reader. For kinetic analysis, reactions were performed in 0.1 M Tris‐HCl buffer, pH 7.4, containing 0.2 mM NADH and increasing concentrations of pyruvate: 0.05, 0.1, 0.2, 0.5, 1, 2, 5 and 10 mM. Reactions were initiated by adding equal amounts of purified LDHA protein. Initial velocities were calculated from the linear decrease in absorbance at 340 nm during the early phase of the reaction. Michaelis‐Menten curves were fitted by nonlinear regression to determine Km and Vmax values. Molecular Dynamics Simulation MD simulations were conducted to evaluate the structural stability and dynamic behavior of LDHA in complex with pyruvic acid, both in its wild‐type form and after palmitoylation at Cys163. All simulations were performed using GROMACS v2022.03 (RRID: SCR_014565) with the AMBER99SB‐ILDN force field. The initial structures were obtained from molecular docking results and converted into GROMACS‐compatible formats. The protein–ligand complexes were solvated in a TIP3P water model with a minimum distance of 1.2 nm from the box edge. To mimic physiological conditions, 0.154 M NaCl was added and counterions were used to neutralize the system. Ligand parameters were generated using GAFF via AmberTools22, with RESP charges derived from Gaussian 16 W calculations. Energy minimization was carried out using the steepest descent algorithm. Short‐range electrostatic and van der Waals interactions were truncated at 1.0 nm, and long‐range electrostatics were calculated using the Particle Mesh Ewald (PME) method. All bonds involving hydrogen atoms were constrained using the LINCS algorithm. Temperature and pressure were maintained using V‐rescale and Berendsen barostat, respectively. The systems underwent 100 ps of initial equilibration in an NVT ensemble, followed by 100 ps in an NPT ensemble. Production MD simulations were performed for 200 ns at 300 K and 1 bar. Simulation trajectories were analyzed for root‐mean‐square deviation (RMSD), root‐mean‐square fluctuation (RMSF), B‐factor, and the C‐terminal to N‐terminal distance using built‐in GROMACS tools. Free energy landscapes were generated using g_sham, and binding free energies were estimated using the MM/PBSA method via MMPBSA.py v16.0. Exosome Isolation Conditioned media from GSC cells cultures were collected and subjected to sequential centrifugation at 4°C: first at 2000 × g for 30 min and then at 12 000 × g for 45 min to remove cellular debris. The resulting supernatant was ultracentrifuged at 100 000 × g for 70 min at 4°C to pellet exosomes. Exosome pellets were resuspended in PBS and stored at –80°C, avoiding repeated freeze–thaw cycles to preserve integrity. PBMC Isolation and Induction of Macrophages Peripheral blood mononuclear cells (PBMCs) were isolated from the peripheral blood of three healthy volunteers as previously described []. Briefly, the PBMCs were isolated using standard density gradient centrifugation (2000 rpm, 30 min) from venous blood. Within 15 min of collection, CD14+ monocytes were then purified from PBMCs using magnetic CD14 microbeads (130‐050‐201, Miltenyi Biotec) according to the manufacturer's instructions. The isolated monocytes were cultured in RPMI 1640 complete medium supplemented with 100 ng/mL M‐CSF (300‐25‐50UG, Peprotech) for 7 days to induce differentiation into macrophages. Virtual Screening of Small Molecule Inhibitors Targeting LDHA Cys163 Virtual screening was performed to identify small‐molecule inhibitors targeting the Cys163 site of human LDHA (PDB ID: 5W8J, 1.55 Å). Protein preparation was carried out using the Protein Preparation Wizard in Schrödinger Maestro 12.8 with hydrogen addition and energy minimization (OPLS_2005, RMSD cutoff: 0.30 Å). A receptor grid was generated around Cys163 (box size: 20 Å3). Four compound libraries from MedChemExpress were screened: HY‐L028 (∼0.9K), HY‐L901 (50K), HY‐L0055V (50.2K), and HY‐L001V (24.4K). Compounds were processed using LigPrep (RRID: SCR_016746) to generate energy‐minimized 3D structures. Docking was conducted using the Glide module within the Virtual Screening Workflow. HY‐L028 was screened directly using extra precision (XP) mode. HY‐L901 and HY‐L0055V underwent three‐stage docking (HTVS‐SP‐XP), retaining the top 10% at each stage, while HY‐L001V used a 15% threshold. The top 200 ranked compounds from each library were manually curated based on docking scores and structural features. The top hit, HY‐N0148 (docking score: −9.448), was further analyzed in PyMOL (RRID: SCR_000305). Six key hydrogen bonds were identified with residues VAL270, ARG171, ASN164, and ARG269. Both 2D and 3D interaction diagrams were generated to visualize the binding conformation. Lactate Quantification Assay Lactate levels were measured using the Lactate Assay Kit (E‐BC‐K044, Elabscience,) according to the manufacturer's instructions. A lactate standard curve was generated by preparing a series of standard dilutions. GSCs from different treatment groups were resuspended in PBS at a concentration of 1 × 106 cells per 300 μL. After centrifugation at 10 000 × g for 10 min at 4°C, the supernatants were collected and kept on ice for analysis. An aliquot of the supernatant was used to determine protein concentration using the BCA protein assay for normalization. For the assay, 5 μL of each sample or standard solution was added to a 96‐well plate, followed by the addition of 100 μL of enzyme working solution. Subsequently, 20 μL of Reagent III was added, and the plate was incubated at 37°C for 10 min. Then, 180 μL of Reagent IV was added to each well and mixed thoroughly. The absorbance at 530 nm was measured using a microplate reader. Lactate concentrations were calculated based on the standard curve generated from known lactate standards and normalized to protein content where appropriate. Protein Purification and Surface Plasmon Resonance Analysis The Homo‐LDHA in pet28a plasmid was obtained from GenePharma and transformed into Escherichia coli BL21 cells (TransGen Biotech, China). Transformed strains were cultured in LB medium containing 0.1 mg/mL kanamycin at 37°C with shaking at 200 rpm. Protein expression was induced with 0.5 mM isopropyl‐β‐D‐thiogalactopyranoside (IPTG; Solarbio, China), and LDHA protein was purified using a Protein Purification Kit (CoWin Bio, Cambridge, MA, USA). Surface plasmon resonance (SPR) analysis was conducted using the SensiQ Pioneer platform (ForteBio, Fremont, CA, USA). PCH biosensor chips (19‐0129, ForteBio) were first activated with 1 mM NiCl2, followed by immobilization of LDHA at a concentration of 50 μg/mL. Binding of the small molecule candidates were assessed, and response units (RU) were recorded to quantify binding activity. Data were normalized to control values and analyzed using Qdat software (ForteBio), and sensorgrams were used to generate binding curves. Cellular Thermal Shift Assay Cellular Thermal Shift Assay (CETSA) was performed with minor modifications from previously described protocols []. Briefly, cells were trypsinized, washed, and resuspended in PBS. Aliquots (50 μL, approximately 1 × 106 cells) were heated at temperatures ranging from 37°C to 73°C in 4°C increments for 3 min using a thermocycler. Cells were then lysed by three freeze–thaw cycles in liquid nitrogen, followed by incubation on ice for 20 min. Lysates were centrifuged at 20 000 × g for 10 min at 4°C, and the resulting soluble fractions were mixed with 5× SDS loading buffer for subsequent immunoblot analysis. Single‐Cell RNA Sequencing To evaluate the role of LDHA palmitoylation inhibition by Rutin in remodeling the immunosuppressive tumor microenvironment of GBM in vivo, Rutin was administered via oral gavage five times per week starting 7 days after GBM cell implantation. scRNA‐seq was conducted following the protocol described in previous studies []. Briefly, freshly dissociated single‐cell suspensions were loaded onto the microfluidic chip of the Chip A Single Cell Kit v2.0 (S050100201, MobiDrop), and droplets were generated using the MobiNova‐100 system (A1A40001). Each cell or nucleus was encapsulated in a droplet containing a gel bead linked to millions of oligonucleotides with unique barcodes. After light‐induced cleavage by the MobiNovaSP‐100 (A2A40001), oligonucleotides diffused into the reaction mix, allowing mRNA capture via oligo (dT) sequences on the beads. Barcoded cDNAs were synthesized by reverse transcription, amplified, and used for library construction with the High Throughput Single Cell 3’ RNA‐Seq Kit v2.0 and 3’ Single Index Kit. Libraries were sequenced on the Illumina NovaSeq 6000 platform. Cell Ranger was used to align scRNA‐seq data to the reference genome and quantify individual transcripts for each cell. Seurat (v4.1.0) was then employed to remove low‐quality cells based on user‐defined criteria, followed by data downsampling and clustering. Finally, t‐distributed stochastic neighbor embedding (t‐SNE) was applied for two‐dimensional visualization of cellular distributions. GBM Organoids Construction The methods were adapted from a previous study on GBOs []. Fresh surgical specimens obtained from GBM patients were mechanically minced and cultured in 6‐well plates containing 2 mL of GBO medium per well. The GBO medium consisted of a 1:1 mixture of DMEM/F12 (A4192001, Gibco) and Neurobasal (21103049, Gibco), supplemented with 1% GlutaMax (35050061, Gibco), 1% non‐essential amino acids (NEAAs), 1% Penicillin‐Streptomycin, 1% N2 supplement (A1370701, Gibco), and 1× B27 supplement (17504044, Gibco). Cultures were maintained on an orbital shaker at 140 rpm in a sterile incubator set to 37°C with 5% CO2. Within 2–3 weeks, tissue fragments typically developed into spherical organoid structures suitable for analyses. Statistics All statistical analyses were performed using GraphPad Prism version 10 (GraphPad Software, RRID: SCR_002798). Analyses were performed with blinding to group allocation. Data are presented as mean ± standard deviation (SD). For comparisons between two groups, a two‐tailed unpaired Student's t‐test was used. For comparisons among multiple groups, one‐way analysis of variance (ANOVA) followed by Tukey's post hoc test was applied. Statistical significance was set at p < 0.05, and all statistical tests were two‐tailed. *p < 0.05, **p < 0.01, and ***p < 0.001.