IQGAP3 bridges matrix stiffness with glioma stem cell maintenance and radioresistance by stabilizing SOX2

Po Zhang; Weichi Wu; Tengfei Huang; Xujia Wu; Donghai Wang; Huairui Yuan; Suchet Taori; Fanen Yuan; Frank P Vendetti; Rui Wang; Tingting Duan; Hailong Mi; Huan Li; Kailin Yang; Daqi Li; Ahmed Habib; Briana C Prager; Ryan C Gimple; Pascal O Zinn; Kalil Abdullah; Christopher Bakkenist; Junran Zhang; Qiulian Wu; Jeremy N Rich

doi:10.1038/s41467-026-74058-0

. 2026 Jun 6;17:7231. doi: 10.1038/s41467-026-74058-0

IQGAP3 bridges matrix stiffness with glioma stem cell maintenance and radioresistance by stabilizing SOX2

Po Zhang ^1,^2,^#, Weichi Wu ^1,^2,^3,^#, Tengfei Huang ^1,^2,³, Xujia Wu ^1,², Donghai Wang ^1,^2,³, Huairui Yuan ^1,², Suchet Taori ^1,², Fanen Yuan ^1,², Frank P Vendetti ⁴, Rui Wang ^1,^2,³, Tingting Duan ^1,², Hailong Mi ^1,², Huan Li ^1,^2,³, Kailin Yang ⁵, Daqi Li ^1,^2,³, Ahmed Habib ^1,⁶, Briana C Prager ⁷, Ryan C Gimple ⁸, Pascal O Zinn ^1,⁶, Kalil Abdullah ^1,⁶, Christopher Bakkenist ^1,⁴, Junran Zhang ⁹, Qiulian Wu ^1,^2,³, Jeremy N Rich ^1,^3,^10,^11,^✉

PMCID: PMC13396378 PMID: 42251046

Abstract

Drivers of therapeutic resistance in cancer include evolving tumor cell heterogeneity and the tumor microenvironment (TME). We find that increased matrix stiffness promotes radioresistance in glioblastoma (GBM) and maintains tumor cell hierarchies. Differential gene expression reveals that stiff matrices induce expression of IQGAP3 (IQ Motif Containing GTPase Activating Protein 3) through YAP1 and TEAD transcription factors in GBM stem cells (GSCs). IQGAP3 promotes GSC self-renewal and survival upon radiation treatment through binding and stabilization of core stem cell transcription factor, SOX2. Targeting IQGAP3 reduces SOX2 protein levels in vitro and in vivo, increasing GSC radiosensitivity and inhibiting tumor growth. Structure-function drug screening of FDA-approved agents blocking IQGAP3-SOX2 binding identifies trimetrexate as a brain penetrant pharmacologic disruptor of IQGAP3 function in radioresistance, sensitizing GSCs to radiotherapy. These results identify molecular underpinnings for biomechanical promotion of cancer stem cell maintenance and therapeutic resistance, informing therapeutic strategies to augment efficacy of radiotherapy.

Subject terms: CNS cancer, Cancer stem cells, Cancer microenvironment

In glioblastoma, matrix stiffness has been linked to poor prognosis and resistance to therapy. Here, the authors demonstrate that matrix stiffness drives radioresistance via upregulation of IQGAP3 promoting cancer cell stemness via SOX2 stabilization and show that blocking this interaction resensitises glioblastoma cells to radiotherapy.

Introduction

GBM represents the most prevalent primary intrinsic cerebral malignancy, accounting for 50.1% of all primary intrinsic brain malignant tumors¹. Standard-of-care for GBM includes surgery, chemoradiation, and adjuvant chemotherapy. Unfortunately, therapeutic resistance is essentially universal with median survival less than 2 years². GBM was originally designated as glioblastoma multiforme due to its striking heterogeneity, including spatial variation of morphologic and radiographic features of vascular proliferation and pseudopalisading necrosis. The GBM tumor microenvironment (TME) is exceptionally complex, reflecting the underlying heterogeneity of normal brain that displays unique aspects in the vasculature and immune system. Normal brain is soft and has a unique extracellular matrix (ECM) consisting of hyaluronic acid, tenascins, and lecticans with near absence of other matrix proteins found in most tissues. In nervous system development, brain ECM guides neurons and blood vessels to targeted destinations. Cancer has been described as the “wound that does not heal” with recent identification of injury responses in SOX2-positive neural crest-like cells as the origin of gliomas³. Biophysical properties evolve in most cancers upon tumor progression and recurrence after treatment with increased fibrosis and calcification. These changes are associated with tumor invasion, metastasis, and cancer stemness in several cancers⁴. ECM stiffness of normal brain tissue is typically less than 200 Pa with variation across brain regions, while stiffness of human low- and high-grade gliomas (LGG and HGG, respectively) increases, ranging from 100 to 10⁴ Pa⁵. Rigid ECM portends a poor prognosis in glioma patients⁶. Stiffening is primarily driven by the excessive deposition, cross-linking, and altered alignment of key components, like collagen and fibronectin, often exacerbated by the actomyosin-mediated contractile forces exerted by stromal cells, such as cancer-associated fibroblasts (CAFs).

In addition to diversity of the TME, GBM tumor cells display striking spatial and temporal variation, containing cellular hierarchies with self-renewing GSCs at the apex. GSCs contribute to tumor invasion, immune evasion, angiogenesis, and therapeutic resistance, suggesting that targeting GSCs may improve treatment of GBM^7–10. However, effective GSC targeting strategies have been challenging to develop, in part due to the heterogeneity and plasticity of GSCs with differential responses to molecularly targeted therapies^11–13. The interplay between tumor cells and TME is multi-dimensional, with biophysical properties integral to cancer biology and proposed as a source for new strategies for oncology treatment¹⁴.

Radiotherapy is a cornerstone of nonsurgical treatment for GBM, which is typically administered concomitantly with the chemotherapeutic agent temozolomide (TMZ) as the current standard-of-care, but radioresistance is universal with most patients suffering recurrence within 2 years^15,16. GSCs are relatively radioresistant through activation of the DNA damage response (DDR)⁷. Befitting an association with therapeutic resistance, ECM is stiffer in recurrent GBMs than newly diagnosed tumors⁶. Matrix stiffness induces chemotherapy resistance in tumor cells^17,18, but less is known about the role in tumor cell radioresistance. Three-dimensional analysis of tumor genetics and expression confirm spatial variation of GBMs¹⁹. Given that GSCs are radioresistant, we explored the interplay between ECM stiffness and GSC radioresistance.

In this work, we demonstrate that stiff matrix within GBM tissues sustains stemness and radioresistance in GSCs by inducing the expression of IQGAP3 protein, which binds and stabilizes the core GSC transcription factor, SOX2. Furthermore, we identify a specific inhibitor of IQGAP3-SOX2 binding, which reduces the radioresistance of GSCs and improves the therapeutic efficacy of GBM treatment.

Results

Matrix stiffness maintains GSCs

ECM stiffness varies across regions of GBMs with central regions having the stiffest ECM and areas contiguous to normal brain tissue possessing the softest^20,21. The mechanosensitive ion channel PIEZO regulates mitosis and tissue stiffness of Drosophila and human gliomas⁵. Prior studies have cultured tumor cells on engineered matrices to examine the effect on GBM growth²². Based on the regional variation of GSCs and promotion of tumor cell growth on stiff matrices, we interrogated the relationship between GSC frequency and regional variation of ECM stiffness in surgical biopsy specimens from patients diagnosed with GBM. First, we dissociated human GBM tissues from different regions, then employed flow cytometry using PIEZO1 as a marker of ECM stiffness⁵ and CD133 and SOX2 to identify GSCs²³ (Fig. 1a). We validated the specificity of the PIEZO1 antibody (Supplementary Fig. 1a–e). Central tumor regions had the highest frequency of PIEZO1⁺ cells (Fig. 1b, c and Supplementary Fig. 2e, f) and CD133⁺/SOX2⁺ cells (Fig. 1d, e). Analysis of RNA sequencing data from different regions (e.g., normal brain tissue, tumor edge, and tumor central region) of GBM²⁴ revealed ECM stiffness and stemness score highest in the central region (Supplementary Fig. 2a, b and Supplementary Data 1). Immunofluorescence staining of human GBM tissue sections showed expression of SOX2 protein correlated with the expression of ECM stiffness-related proteins (PIEZO1, phosphorylated FAK (p-FAK), and phosphorylated MLC2 (p-MLC2))^5,6 between regions (Fig. 1f, g).

Fig. 1 — a Graphical illustration of matrix stiffness and stem cell detection in different GBM regions. GBM: Glioblastoma. **b–e** Flow cytometry gating (b, d) and quantification (c, e; n = 6 GBM samples) show regional differences. (c, *** p = 0.0005; e, *** p = 0.0001). f,g Immunofluorescence (f) and quantification (g, n = 18 fields/6 samples) of stiffness markers (p-MLC2, PIEZO1, p-FAK) and SOX2. (g, **** p < 0.0001). h Workflow using hCD147 (tumor cells), CD133/SOX2 (GSCs) to test ECM stiffness effects on GSC stemness. i, j Gating strategy (i) and statistical quantification (j, n = 5 mice per group) of flow cytometry. (j, * p = 0.0209, **** p < 0.0001). k Immunoblot of GSCs, DGCs, and NSCs cultured on matrices with different stiffness levels for 48 hours. GSCs: Glioma Stem Cells. DGCs: Differentiated Glioma Cells. NSCs: Neural Stem Cells. **l–o** GSCs cultured on diffusion-based polyacrylamide stiffness gradient gels were subjected to immunofluorescence staining (l) and statistical quantification (m, n = 6 randomly selected fields per group) as well as quantification of EdU incorporation (n, o; n = 6 randomly selected fields per group). Stronger green fluorescence indicates higher matrix stiffness. (m, o; **** p < 0.0001). p, q, Proliferation of GSCs on different stiffness substrates. (n = 4 biologically independent samples). (p, q; **** p < 0.0001). Immunoblots are representative of three independent experiments with similar results. Data are presented from three independent experiments (m, o). Data: mean ± s.d. (c, e, g, j, m, o, p, q). Statistics: two-tailed unpaired t-test (c, e, g, m, o), one-way ANOVA (j), two-way ANOVA (p, q). Source data are provided as a Source Data file.

To explore the influence of ECM stiffness on the stemness maintenance of GSCs, we conducted in situ tumorigenesis experiments on immunodeficient mice using patient-derived GSCs (MES20). Upon tumor formation, tumors were resected and digested into single cells. Cells were divided into three conditions: (1) uncultured tumor cells directly subjected to flow cytometry; (2) tumor cells cultured in hard matrix gel (Young’s modulus ~10 kPa)⁶, a stiffness comparable to GBM tissues, for 5 days and then examined; (3) tumor cells were cultured in a soft matrix gel (Young’s modulus ~500 Pa) for 5 days and then subjected to flow cytometry. Human CD147 was utilized to enrich human-derived tumor cells (Fig. 1h)²⁵. Compared with the original tumor cells, culture in hard matrix gel increased CD133⁺ and SOX2⁺ cell populations, whereas culture in soft matrix gel decreased those populations (Fig. 1i, j and Supplementary Fig. 2g).

To explore the sensitivity of different GBM tumor cell populations to mechanical stimuli, we interrogated a large-scale single-cell RNA sequencing (scRNA-seq) dataset²⁶ and distinguished GSCs and their differentiated glioma cells (DGCs) counterparts leveraging a logistic regression classifier²⁷ (see Methods) (Supplementary Fig. 2c). GSC markers correlated with markers of mechanical stimuli (Supplementary Fig. 2d). Flow cytometry of human GBM surgical specimens revealed that the proportion of PIEZO1⁺ cells was higher in CD133⁺ and SOX2⁺ cells than in CD133^- and SOX2^- cells (Supplementary Fig. 2h–l). GSCs, matched DGCs²⁸, or neural stem cells (NSCs) were cultured in gels with a gradient of matrix stiffness for 48 h, then matrix stiffness-related proteins were detected. GSCs displayed greater sensitivity to mechanical stimuli than DGCs and NSCs (Fig. 1k). As matrix stiffness increased, markers of stiffness (p-FAK and PIEZO1) and stemness (OLIG2 and SOX2) increased in GSCs (Supplementary Fig. 2m).

To assess the influence of different matrix stiffness on the proliferative ability of GSCs, we fabricated diffusion-based polyacrylamide stiffness gradient gels (Young’s modulus ~0.5-10 kPa), then added green fluorescent particles to the stiff portion of the matrix gel to measure stiffness²⁹. GSCs cultured for 48 h in this gradient gel underwent immunofluorescence staining, revealing that GSCs grown on stiff matrix gel had higher SOX2 protein levels (Fig. 1l, m and Supplementary Fig. 2n, o) and 5-ethynyl-2’-deoxyuridine (EdU) incorporation (Fig. 1n, o and Supplementary Fig. 2p, q) than those grown on soft matrix. However, no differences in SOX2 expression (Supplementary Fig. 3a–d) or EdU incorporation (Supplementary Fig. 3e–h) were observed in DGCs cultured on the gradient gel. Increased matrix stiffness enhanced the proliferative capacity of GSCs (Fig. 1p, q). Collectively, these results demonstrate that mechanical stiffness maintained GSCs and that GSCs were more sensitive to mechanical stimulation than differentiated progeny and normal stem cells.

Matrix stiffness induces IQGAP3 in GSCs

To discover essential factors involved in GSC responses to stiffness, we cultured GSCs in gels of different matrix stiffnesses for 48 h and conducted RNA sequencing (Fig. 2a and Supplementary Data 2), revealing that genes overexpressed in GSCs cultivated on stiff matrix gels were primarily associated with cell proliferation and stemness (Supplementary Fig. 3i–k and Supplementary Data 2). To prioritize targets, we analyzed sequencing data of the core region (stiff) and the edge region (soft) of GBM tumors³⁰ (Fig. 2b and Supplementary Data 3), then overlapped these genes to reveal 123 genes that were highly expressed in GSCs cultured in stiff matrix gels and highly expressed in GBM cores (Fig. 2c and Supplementary Data 4). As GSCs responded to matrix stiffness differentially compared to DGCs and NSCs, we refined the 123 genes to identify targets preferentially expressed in GSCs relative to DGCs and NSCs (GSC vs. DGC³¹ and GSC vs. NSC³²) (Fig. 2d) (Supplementary Data 5). Three genes -- IQGAP3, WDR62, and NCAPG2 — were induced in GSCs cultured under stiff conditions, expressed in the tumor core, and differentially expressed in GSCs (Fig. 2e).

a,b Volcano plots show RNA-seq DEGs in stiff/soft GSCs and GBM core/edge tissues (GSE153746) with |log2FC | > 0.585, FDR < 0.05; red/blue dots denote up/downregulated genes in stiff groups. c Venn analysis identifies 123 overlapping upregulated genes. d, e volcano plots of GSC vs DGC/NSC datasets (GSE89623, GSE140441) using these 123 genes, followed by Venn filtering for GSC-highly expressed genes. f, RT-qPCR confirms elevated IQGAP3, NCAPG2 and WDR62 mRNA in stiff GSCs. (MES20, **** p < 0.0001; 3028, *** p = 0.0004). g, Immunoblotting shows their protein levels under varying stiffness. **h, i**, IHC staining (h) and quantification (i, n = 6 samples per group, ** p = 0.0017) of IQGAP3 across distinct GBM regions. j, Schematic for identifying IQGAP3-regulating transcription factors. **k, l**, Proteomic heatmap (k) and immunoblot (l) validation screen IQGAP3 promoter-binding proteins. m, Immunoblot and Co-IP show stiffness-induced YAP1 nuclear translocation and YAP1-TEAD1 binding. IP, immunoprecipitation. **n-q**, Immunofluorescence (n, o) and quantitative (p, q; MES20, n = 106 (soft), n = 86 (stiff); 3028, n = 100 (soft), n = 90 (stiff), **** p < 0.0001) analysis reveal increased nuclear YAP1 in stiff GSCs. r, ChIP-qPCR confirms TEAD1/YAP1 binding to the IQGAP3 promoter (**** p < 0.0001). **s-v**, RT-qPCR (s and u) and immunoblotting (t and v) demonstrate that YAP1/TEAD1 mediate stiffness-dependent IQGAP3 upregulation in GSCs. (s, u; **** p < 0.0001). Immunoblots are representative of three independent experiments with similar results. Data are presented from three independent experiments (f, i, p - s and u). Data are shown as mean ± s.d. (f, i, p, q, r, s, u). Statistics: two-tailed unpaired t-test (i, p, q); one-way ANOVA (s, u); two-way ANOVA (f, r). Transcriptomic and proteomic data were obtained from independent biological replicates. Statistical analyses and P values for Fig. 2a, b, d are available in the Source Data file. Source data are provided as a Source Data file.

To confirm these screening results, we cultivated GSCs in matrix gels of increasing stiffness, then conducted RT-qPCR and immunoblotting. Stiff matrices induced IQGAP3 levels to a much greater degree compared to WDR62 and NCAPG2 (Fig. 2f, g). Immunohistochemical (IHC) staining on different regions of human GBM tissue showed that IQGAP3 was more abundant in GBM core regions compared to edges (Fig. 2h, i and Supplementary Fig. 3l). Immunofluorescence staining of human GBM tissues showed that IQGAP3 protein was concentrated in regions with high p-MLC protein expression (Supplementary Fig. 3m, n). GSCs cultured on gradient gels were subjected to immunofluorescence staining 48 hours later; IQGAP3 was highly expressed in stiff matrix regions (Supplementary Fig. 4a, b).

To explore the molecular regulation by which mechanical stimulation induced IQGAP3 expression in GSCs, we measured levels of IQGAP3 promoter-binding proteins under different levels of matrix stiffness. The DNA domain sequence of the IQGAP3 promoter underwent DNA pull-down, then silver staining, which indicated that levels of IQGAP3 promoter-binding proteins in GSCs cultivated on stiff matrix gel were higher than on soft matrix gel (Supplementary Fig. 4c). Proteomic analysis (Fig. 2j and Supplementary Data 6) showed that TEAD transcription factors, transcription-related proteinases, and histone-modifying enzymes were all enriched in the IQGAP3 promoter region (Fig. 2k). DNA pull-down followed by immunoblotting revealed that the TEAD1 protein exhibited strong binding to the IQGAP3 promoter region (Fig. 2l). Activation of TEAD proteins depends on binding to YAP1 upon nuclear entry³³. Through nuclear and cytoplasmic fractionation, co-immunoprecipitation (co-IP) (Fig. 2m) and immunofluorescent staining (Fig. 2n–q and Supplementary Fig. 4d, e) showed that more YAP1 entered nucleus in GSCs cultured on stiff matrix than on soft matrix. Binding of YAP1 to the TEAD1 protein increased after nuclear entry (Fig. 2m). Chromatin immunoprecipitation followed by quantitative PCR (ChIP-qPCR) showed that YAP1 and TEAD1 binding to the IQGAP3 promoter occurred preferentially in GSCs cultured on stiff matrix (Fig. 2r). To confirm YAP1 and TEAD1 regulation of IQGAP3, we separately knocked down the expression of YAP1 and TEAD1 in GSCs, showing that IQGAP3 mRNA and protein levels decreased with either YAP1 or TEAD1 loss in GSCs cultivated on stiff matrix but not on soft matrix (Fig. 2s–v and Supplementary Fig. 4f-i). IQGAP3 knockdown in GSCs cultured on stiff matrix gels demonstrated that IQGAP3 knockdown slightly increased cytoplasmic YAP1 and decreased nuclear YAP1 in GSCs (Supplementary Fig. 4j).

IQGAP3 maintains GSCs

IQ motif–containing GTPase-activating proteins (IQGAPs) are scaffolding proteins that regulate multiple cellular processes, including growth factor receptor signaling, cytoskeletal rearrangement, adhesion, and proliferation, and are highly expressed in multiple cancers³⁴. To determine the role of IQGAP3 in GSCs, we cultivated GSCs, DGCs, and NSCs in stiff matrix gel and detected IQGAP3 protein, showing IQGAP3 protein levels were higher in GSCs than other cell types (Fig. 3a, b). Chromatin immunoprecipitation (ChIP)-sequencing (ChIP-seq)³⁵ showed enrichment of acetylation of histone H3 on lysine 27 (H3K27ac) on the IQGAP3 promoter in GSCs (Supplementary Fig. 5a), indicating that IQGAP3 was actively transcribed in GSCs. Analysis of IDH-wildtype (IDH wt) glioma data from the Chinese Glioma Genome Atlas (CGGA) database indicated that IQGAP3 positively correlated with stemness-associated markers CD15 and CD133 (Supplementary Fig. 5b, c). Targeting IQGAP3 in GSCs reduced SOX2 expression and increased GFAP protein expression, suggesting induction of differentiation (Fig. 3c) with decreased GSC viability (Fig. 3d) and EdU incorporation (Fig. 3e and Supplementary Fig. 5d). Depleting IQGAP3 diminished GSC self-renewal measured by impaired sphere formation using extreme limiting dilution (Fig. 3f) and reduced sphere size (Fig. 3g and Supplementary Fig. 5e). By contrast, loss of IQGAP3 did not suppress DGC or NSC proliferation (Supplementary Fig. 5f–i), suggesting the specific requirement for IQGAP3 in GSCs.

Fig. 3 — a, b Immunoblotting (a) and immunofluorescence (b) detected IQGAP3, the differentiation marker GFAP, and the stemness marker SOX2 in indicated cells. GSCs: Glioma Stem Cells. DGCs: Differentiated Glioma Cells. NSCs: Neural Stem Cells. c Immunoblots show altered SOX2 and GFAP levels following IQGAP3 knockdown in GSCs; shCONT served as a control. d, e Cell viability (d, n = 4 biologically independent samples) and quantification of EdU incorporation (e, n = 3 randomly selected fields per group) in GSCs with or without IQGAP3 knockdown. (d, **** p < 0.0001) (e; MES20(* p = 0.0102 and * p = 0.0222), 3028(*** p = 0.0002), 387(*** p = 0.0002)). f, g Limiting dilution (f) and sphere formation (g) (n = 50 spheres) confirmed impaired self-renewal upon IQGAP3 depletion. (f, g; **** p < 0.0001). **h–l** In vivo xenograft models: H&E staining (h), bioluminescence imaging (i) and quantification (j, n = 5 mice per group), tumor growth curves (k, n = 5 mice per group) and Kaplan - Meier survival analyses (l, n = 6 mice per group) show suppressed tumor progression and prolonged survival after IQGAP3 knockdown. (j, ** p = 0.0014 and ** p = 0.0030; k, **** p < 0.0001; l, *** p = 0.0006). **m, n**, IHC staining (m) and statistical quantification (n) analysis of IQGAP3 in designated tissues. IHC: Immunohistochemical. (Normal brain, n = 8; Low-grade glioma, n = 19; High-grade glioma, n = 30; Recurrent glioblastoma, n = 10). (n, **** p < 0.0001). o, Kaplan - Meier survival curves of CGGA (primary, gloma_IDH_wt) based on IQGAP3 mRNA expression. The top 50% and the bottom 50% are defined as high and low groups, respectively. IDH_wt: IDH_wild_type. (o, **** p < 0.0001). Immunoblots are representative of three independent experiments with similar results. Data are presented from three independent experiments (e, f, g). Data: mean ± s.d. (d, e, g, j, k, n). Statistics: one-way ANOVA (e, g, j, n), two-way ANOVA (d, k). Two-tailed likelihood-ratio test for f. Log-rank test for l and o. CGGA: Chinese Glioma Genome Atlas. Source data are provided as a Source Data file.

In vivo tumor growth is a gold standard for tumor initiation. GSCs transduced with either a non-targeting control shRNA (shCONT) or one of two, non-overlapping shRNAs targeting IQGAP3 (shIQGAP3) were orthotopically xenografted into immunocompromised mice. Knockdown of IQGAP3 in GSCs reduced tumor volumes and prolonged the survival of tumor-bearing mice (Fig. 3h–l and Supplementary Fig. 5j–p). Immunofluorescence staining of xenograft tumor sections showed that knockdown of IQGAP3 expression decreased the number of Ki67⁺ cells (Supplementary Fig. 5q–s).

Immunohistochemical staining of normal brain tissue, low- and high-grade glioma, and recurrent GBM showed that the expression of IQGAP3 protein was lowest in normal brain tissue and highest in recurrent GBM tissue (Fig. 3m, n). To investigate the expression profile of IQGAP3 across different cell types in GBM tissues, we analyzed scRNA-seq data^36,37. IQGAP3 was predominantly expressed in tumor cells (Supplementary Fig. 6a–c). IQGAP3 mRNA levels negatively correlated with the survival of patients (Fig. 3o).

Matrix stiffness promotes radiation resistance in GSCs through IQGAP3

To identify downstream functions of IQGAP3 in GSCs, we silenced IQGAP3 in GSCs using shRNA, cultured the cells on stiff matrix gel, and performed RNA sequencing (Supplementary Fig. 7a). Gene Set Enrichment Analysis (GSEA) (Fig. 4a) and Gene Ontology (GO) analysis (Supplementary Fig. 7c, d) demonstrated downregulation of pathways related to DNA damage repair and cell cycle regulation in IQGAP3 knockdown conditions (Supplementary Data 7). Reanalysis of RNA-sequencing data of GSCs cultured on stiff matrix revealed upregulation of genes related to radiation resistance and cell cycle (Supplementary Fig. 3i). As GSCs preferentially activate DNA damage responses to promote radioresistance⁷, we hypothesized that matrix stiffness and IQGAP3 contribute to GSC radioresistance.

a GSEA analysis of pathways associated with differentially expressed genes in GSCs with or without IQGAP3 knockdown. NES: Normalized Enrichment Score. b GSEA analysis of pathways associated with differentially expressed genes in post-irradiation GSCs cultured on stiff versus soft substrates. c, d Alkaline comet assay assesses time-dependent DNA damage in 2 Gy-irradiated GSCs on soft/stiff matrices (c), with comet-tail cell ratio quantification (d, n = 6 randomly selected fields per group, ** p = 0.0077 and ** p = 0.0025). Gy: Gray. IR: Ionizing Radiation (X-ray). e, f EdU assay (e) and quantification show proliferation (f, n = 3 randomly selected fields per group, ** p = 0.0034, * p = 0.0407 and *** p = 0.0004) of IR-treated GSCs under different stiffness. g, h Comet assay (g) and quantification (h, n = 6 randomly selected fields per group, **** p < 0.0001) evaluate DNA damage post-IR in IQGAP3-knockdown vs. control GSCs on stiff matrices. i Immunoblotting shows changes in γ-H2AX after radiotherapy in GSC MES20 (stiff) cells, with or without IQGAP3 knockdown. j, k Immunofluorescence (j) and quantification (k, n = 100 cells per group, **** p < 0.0001) of γ-H2AX intensity at 48 h post - IR in stiffness gradient - cultured GSCs with or without IQGAP3 knockdown. l Proliferation of IR-treated GSCs (with/without IQGAP3 knockdown) on soft/stiff matrices (n = 4 biological replicates, **** p < 0.0001). **m–o** Immunoblot (m), IF quantification (n, n = 100 cells/group, **** p < 0.0001) and proliferation assays (o, n = 4 biological replicates, **** p < 0.0001) in IR-exposed GSCs with IQGAP3 overexpression on soft matrices. EV, empty vector. Immunoblots are representative of three independent experiments. Data are from three independent experiments (d, f, h, k, n) and presented as mean ± s.d. (d, f, h, k, l, n, o). Statistics: two-tailed unpaired t-test (n); one-way ANOVA (f); two-way ANOVA (d, h, k, l, o). shIQGAP3 denotes shIQGAP3.1 and shIQGAP3.2. Statistical details for Fig. 4a, b are provided in the Source Data file. Source data are provided as a Source Data file.

In assessing radioresistance responses, we irradiated (2 Gy dose) GSCs cultured on either stiff or soft matrix gel and performed RNA sequencing 48 hours following irradiation (Supplementary Fig. 7b). GSEA (Fig. 4b) and GO (Supplementary Fig. 7e, f) indicated upregulation of DNA damage repair and cell cycle regulation-related genes in GSCs cultured on the stiff matrix gel (Supplementary Data 8). To confirm these results in vitro, we conducted Comet assays, revealing that the rate of DNA damage repair in GSCs cultured on stiff matrix gel was greater compared to GSCs cultured on soft matrix gel following irradiation (Fig. 4c, d and Supplementary Fig. 8a, b). Concurrently, EdU⁺ incorporation in GSCs following irradiation increased as the matrix stiffness increased (Fig. 4e, f and Supplementary Fig. 8c, d). To test whether these results were specific to GSCs, we irradiated DGCs cultured on either stiff or soft matrix. GSCs exhibited greater levels of radioresistance, with the least γ-H2AX foci and the lowest level of cleaved PARP and cleaved Caspase 3 protein on stiff matrix as assessed by western blot (Supplementary Fig. 8h). These results indicate that matrix stiffness regulates radioresistance through a GSC-specific mechanism.

Next, we explored the role of IQGAP3 in regulating GSC radioresistance through matrix stiffness. Following IQGAP3 knockdown, we cultured GSCs on stiff matrix and conducted Comet assays following 2 Gy irradiation. Persistence of comet tails increased with IQGAP3 knockdown, indicating elevated levels of DNA damage (Fig. 4g, h and Supplementary Fig. 8e, f). In an orthogonal experiment, IQGAP3 knockdown increased γ-H2AX foci in radiated GSCs on stiff matrix (Fig. 4i and Supplementary Fig. 8g, i–l). When GSCs were cultured on gels of graded stiffness, GSCs on stiffer regions had fewer γ-H2AX foci compared to GSCs grown in softer regions. However, after IQGAP3 knockdown, γ-H2AX expression in GSCs on stiff matrix was not appreciably different from those grown on soft matrix (Fig. 4j, k and Supplementary Fig. 8m, n), indicating selective dependency on IQGAP3. Following irradiation, IQGAP3 knockdown inhibited growth of GSCs grown on stiff matrix more than those grown on soft matrix gel, indicating stiffness-dependent influence of IQGAP3 on GSC viability (Fig. 4l and Supplementary Fig. 8o). Reciprocally, IQGAP3 overexpression in irradiated GSCs on soft matrix decreased γ-H2AX foci (Fig. 4m, n and Supplementary Fig. 8p) and increased cell viability (Fig. 4o). Taken together, loss- and gain-of-function demonstrate that IQGAP3 is essential in regulating GSC radiation resistance via responses to matrix stiffness.

To investigate the impact of IQGAP3 on DNA repair pathways, we knocked down IQGAP3 expression in GSCs cultured on stiff matrix gels. The results demonstrated that IQGAP3 knockdown reduced the protein levels of phosphorylated BRCA1 (p-BRCA1), phosphorylated CHK1 (p-CHK1), RAD51, and phosphorylated DNA-PKcs (p-DNA-PKcs) (Fig. 5a). To explore the effect of IQGAP3 on foci formation, size, and structure, we performed immunofluorescence staining, revealing that targeting IQGAP3 impaired the resolution of γH2AX foci without affecting γH2AX foci accumulation (Fig. 5b, c), while inhibiting RAD51 foci formation (Fig. 5d, e). Combined with ionizing radiation, IQGAP3 knockdown in GSCs increased the number and area of γH2AX foci (Fig. 5f, g).

Fig. 5 — a Immunoblotting revealed altered expression levels of proteins related to the DNA damage repair pathway following IQGAP3 knockdown. **b–e** MES20 GSCs were transduced with non-targeting (shCONT) or IQGAP3-targeting (shIQGAP3) shRNA, then cultured on stiff matrix gels under control (untreated) or 2 Gy irradiation conditions. γ-H2AX foci formation was assessed by immunofluorescence over 1–24 h (b). Immunofluorescence staining showed RAD51 foci changes at different time points with/without IQGAP3 knockdown in stiff matrix gel-cultured MES20 GSCs (e). γ-H2AX foci formation and RAD51 foci were quantified in (c) and (d) (c,d n = 100 cells per group, **** p < 0.0001). GSC: Glioma Stem Cell. Gy: Gray. IR: Ionizing Radiation (X-ray). f,g GSCs (MES20 and 3028) were transduced with IQGAP3-targeting shRNA (shIQGAP3) or a non-targeting control shRNA (shCONT). After transduction, cells were cultured on stiff matrix gels, irradiated (2 Gy), and immunostained for γ-H2AX at 12 h and 48 h post-irradiation to assess DNA damage foci formation. Representative images are shown; experiments were repeated three times independently with similar results. The magnified views on the right show the regions marked by the white boxes in the main images. Scale bars: 5 μm (main images), 1 μm (magnified regions). Immunoblots are representative of three independent experiments with similar results. In c and d data are presented from three independent experiments. Data: mean ± s.d. (c,d). Statistics: two-way ANOVA (c,d). shIQGAP3: shIQGAP3.1 and shIQGAP3.2. IR: Ionizing Radiation (X-ray). Source data are provided as a Source Data file.

IQGAP3 regulates radioresistance of GSCs by altering SOX2 protein levels

To determine molecular mechanisms governing IQGAP3’s role in GSC radioresistance, we conducted co-immunoprecipitation (co-IP) followed by proteomic mass spectrometry to analyze IQGAP3-interacting proteins when cultured on stiff matrix (Supplementary Fig. 9a). AIF1L, SOX2, and COBL interacted with IQGAP3 and were highly expressed in GSCs compared to DGCs²⁸ (Fig. 6a and Supplementary Data 9). Co-IP revealed that the binding of SOX2 to IQGAP3 was the most pronounced among these target gene products (Fig. 6b and Supplementary Fig. 9b). In GSCs cultured on stiff matrix, IQGAP3 and SOX2 colocalized (Supplementary Fig. 9c). Knockdown of IQGAP3 in GSCs cultured on stiff matrix decreased SOX2 protein levels without changes in mRNA levels (Fig. 6c, d and Supplementary Fig. 9d). SOX2 protein levels were elevated in GSCs cultured on stiff matrix (Fig. 1k) without obvious changes at mRNA levels (Supplementary Fig. 9e). IQGAP3 knockdown decreased SOX2 protein levels in tumor cells of GSCs-derived xenografts in immunocompromised mice, as measured by immunofluorescence (Fig. 6e–h).

Fig. 6 — a Proteomic (IQGAP3 IP vs. IgG) and RNA-seq (GSC vs. DGC, GSE54791) identified IQGAP3-interacting proteins highly expressed in GSCs; red dots indicate these proteins. GSC: Glioma Stem Cell. DGC: Differentiated Glioma Cell. IP: Immunoprecipitation. b Immunoblot of IQGAP3 and SOX2 protein immunoprecipitates in GSCs cultured on stiff matrix gel. c Immunoblot of GSCs cultured on stiff matrix gel with or without IQGAP3 knockdown. d RT-qPCR analysis of GSCs cultured on stiff matrix gel with or without IQGAP3 knockdown (**** p < 0.0001). **e–h** Representative images (e and g) and quantifications (f and h, n = 9 randomly selected fields examined across 3 mice per group, **** p < 0.0001) of SOX2 (green) and IQGAP3 (red) in brain sections derived from the indicated xenografts. DAPI (blue) was used to stain the nuclei. i SOX2 ChIP-seq heat maps in GSCs with or without IQGAP3 knockdown. j SOX2 ChIP-qPCR in GSCs (CDK1, ** p = 0.0060; others, **** p < 0.0001). k, RT-qPCR analysis in GSCs with or without IQGAP3 knockdown (**** p < 0.0001). l Immunoblot of GSCs with or without IQGAP3 knockdown ± SOX2 overexpression post-IR. m,n Immunofluorescence (m) and quantification (n, n = 100 cells per group, **** p < 0.0001) of GSCs after IR exposure. Immunoblots are representative of three independent experiments with similar results. In (d), (f), (h), (j), (k) and (n), data are presented from three independent experiments. Data: mean ± s.d. (d,f,h,j,k,n); Statistics: one-way ANOVA (f,h,n); two-way ANOVA (d,j,k). ChIP: chromatin immunoprecipitation. shIQGAP3: shIQGAP3.1 and shIQGAP3.2. IR: Ionizing Radiation (X-ray). Source data are provided as a Source Data file.

As SOX2 functions as a transcription factor, we investigated SOX2 binding to target gene promoters using ChIP-seq on GSCs grown in stiff matrix. SOX2 binding to DNA target gene promoters decreased after IQGAP3 knockdown in GSCs (Fig. 6i). Four genes previously associated with radioresistance and cell proliferation^38–40 -- CDK1, CDK2, CCND1, and BCL2 – showed decreased ChIP-seq peak signals upon IQGAP3 knockdown (Supplementary Fig. 9f). SOX2 ChIP-qPCR validated these results in vitro (Fig. 6j and Supplementary Fig. 9g). IQGAP3 knockdown in GSCs also decreased mRNA levels of these target genes (Fig. 6k and Supplementary Fig. 9h).

To demonstrate that IQGAP3 regulates GSC radioresistance by modulating SOX2 protein levels, we rescued the effects of IQGAP3-knockdown in GSCs by overexpressing SOX2 (OE-SOX2) and then treated with radiation. As expected from above, IQGAP3 knockdown in GSCs increased γ-H2AX levels that were rescued upon concurrent SOX2 overexpression (Fig. 6l–n and Supplementary Fig. 9i, j). These findings suggest that IQGAP3 regulates GSC radioresistance by altering SOX2 protein levels.

IQGAP3 stabilizes SOX2 in the cytoplasm of GSCs

As IQGAP3 regulated SOX2 primarily at the protein level (Fig. 6c, d), we hypothesized that IQGAP3 impacts SOX2 protein degradation. Therefore, we performed a time-course experiment in which we treated GSC cultures grown on stiff matrix gel with cycloheximide (CHX, 50 μg/ml), a protein synthesis inhibitor. IQGAP3 knockdown in GSCs increased the SOX2 degradation rate when treated with CHX (Fig. 7a, b). To determine the mechanism by which SOX2 was being degraded, we treated GSCs with either MG132 (10 μM for 10 h), a proteasome inhibitor, or chloroquine (10 μM for 10 h), an inhibitor of the lysosomal protein degradation. MG132 rescued the reduction of SOX2 protein caused by IQGAP3 knockdown, whereas chloroquine had minimal effect (Fig. 7c and Supplementary Fig. 10a), suggesting that SOX2 was being degraded via the proteasome. Concurrently, we observed that targeting IQGAP3 in GSCs increased SOX2 protein ubiquitination (Fig. 7d). MG132 treatment similarly rescued SOX2 protein levels in wildtype GSCs cultured on soft matrix (Supplementary Fig. 10b) and enhanced SOX2 protein ubiquitination in GSCs cultured on soft matrix (Supplementary Fig. 10c).

IQGAP3 and SOX2 proteins bound primarily in the cytoplasm (Supplementary Fig. 9c). To understand the direct molecular relationship between IQGAP3 and SOX2, we performed nuclear and cytoplasmic fractionation in GSCs grown on stiff matrix gel and observed that SOX2 protein in both the cytoplasm and nucleus decreased after IQGAP3 knockdown (Fig. 7e and Supplementary Fig. 10d, e). GSCs cultured on stiff matrix gels, compared to soft matrix gels, displayed greater levels of SOX2 protein in both the cytoplasmic and nuclear fractions (Supplementary Fig. 10h). Next, we treated GSCs with MG132 following IQGAP3 knockdown and performed western blotting after nuclear and cytoplasmic fractionation. Compared to control GSCs, IQGAP3 knockdown increased cytoplasmic SOX2 protein levels in GSCs, but not in the nucleus (Fig. 7f and Supplementary Fig. 10f, g). These findings suggest that targeting IQGAP3 induces degradation of SOX2 protein in the cytoplasm of GSCs, reducing the nuclear SOX2 levels.

Based on our observation that IQGAP3 knockdown increased SOX2 ubiquitination (Fig. 7d), we hypothesized that IQGAP3 stabilizes SOX2 in the cytoplasm by competing with E3 ubiquitin ligases. We knocked down four E3 ubiquitin ligases known to target SOX2 in other cellular contexts (UBR5, CUL4A, TRIM26, and WWP2)^41–44 and assessed SOX2 protein levels in GSCs. WWP2 (WW domain-containing protein 2) knockdown increased SOX2 protein levels (Fig. 7g–j), indicating that WWP2 may be partially responsible for SOX2 degradation in GSCs. Consistent with our results that SOX2 protein was degraded in the cytoplasm, WWP2 knockdown increased SOX2 protein in the cytoplasm of GSCs (Fig. 7k and Supplementary Fig. 10i, j). To confirm that IQGAP3 and WWP2 compete for binding to SOX2, we transfected HA-tagged WWP2 and mCherry-tagged SOX2 into HEK293T cells with varying levels of transfected Flag-tagged IQGAP3. Following mCherry immunoprecipitation, increasing the amount of Flag-IQGAP3 decreased the interaction between WWP2 and SOX2 measured by western blot (Fig. 7l). WWP2 knockdown in GSCs cultured on stiff matrix reduced SOX2 ubiquitination levels caused by IQGAP3 silencing (Fig. 7m). These results suggest that IQGAP3 in the cytoplasm competes with WWP2 for binding to SOX2, thereby preventing the degradation of SOX2 (Fig. 7n).

Trimetrexate is an IQGAP3 inhibitor that augments radiotherapy responses against GBM

To lay the foundation to translate our findings into a therapeutic paradigm, we interrogated FDA-approved drugs that could be repurposed to target IQGAP3 function in SOX2 stabilization and overcome GSC radioresistance. First, we generated truncated IQGAP3 fragments based on its six domains, then individually overexpressed each fragment and mCherry-labeled SOX2 in HEK293T cells. Co-IP experiments demonstrated that the region of IQGAP3 spanning amino acids 150-670 interacted with SOX2 (Fig. 8a). In silico drug screening based on this binding region of IQGAP3 and SOX2 was performed using an FDA-approved drug library of 3607 drugs. Ten compounds with the highest Lidock scores were selected for in vitro analysis (Fig. 8b–d and Supplementary Data 10). As radiation therapy continues to remain the most efficacious non-surgical treatment modality for GBM, and we have implicated IQGAP3 in promoting GSC radioresistance, we sought to combine our drug treatment with radiation therapy to augment treatment response. We treated GSCs grown on stiff matrix gel with or without ionizing radiation (IR) and with vehicle control or with one each of the ten selected drugs, then performed cell viability assays (Fig. 8e and Supplementary Fig. 11a). Trimetrexate consistently displayed the greatest anti-GSC efficacy (Fig. 8e), so we prioritized trimetrexate for further drug studies.

We first measured the IC₅₀ values of trimetrexate treatment on GSCs cultured on stiff matrix at baseline or with IR. Trimetrexate treatment combined with IR resulted in a lower IC50 (Fig. 8f, g). At the biochemical level, in silico analysis identified residues R324 and E358 as essential for the binding of IQGAP3 to Trimetrexate (Fig. 8h). The 1–670 amino acid fragment of IQGAP3 was purified in vitro and subjected to Surface Plasmon Resonance (SPR) analysis with trimetrexate. Results showed that trimetrexate directly bound to this IQGAP3 fragment with a dissociation constant (KD) of 1.353 µM (Fig. 8i and Supplementary Data 11). We then generated an IQGAP3 mutant by substituting glutamic acid (Glu, E) at position 358 with alanine (Ala, A) (E358A). SPR assays revealed that the E358A mutant exhibited a KD value of 540 μM for trimetrexate, representing a ~400-fold reduction in binding affinity compared to the wild-type fragment, thus confirming the critical role of Glu358 in IQGAP3-Trimetrexate interaction (Fig. 8j). Trimetrexate shifted the thermal stability of IQGAP3 protein, confirming direct interaction between IQGAP3 and trimetrexate (Fig. 8k and Supplementary Fig. 11b).

GSCs cultured on a stiff matrix gel with trimetrexate and IR induced greater γ-H2AX levels than when each therapy was administered alone (Fig. 8l and Supplementary Fig. 11c). GSCs cultured on stiff matrix gel and treated with trimetrexate demonstrated greater synergistic tumor inhibition with IR than GSCs cultured on soft matrix gel associated with differential expression of IQGAP3 (Fig. 8m–n and Supplementary Fig. 11d). Comparison of IC50 values for trimetrexate between GSCs cultured on soft matrices and GSCs, DGCs, and NSCs cultured on stiff matrices demonstrated that trimetrexate exhibited the lowest IC50 in GSCs grown on stiff matrices (Fig. 8o).

Treatment of GSCs with trimetrexate reduced SOX2 protein levels but not IQGAP3 protein levels (Fig. 8p). We hypothesized that the docking of trimetrexate with IQGAP3 affected the interaction between the IQGAP3 and SOX2 proteins. We overexpressed mCherry-SOX2 and Flag-IQGAP3 in HEK293 cells, then treated with varying concentrations of trimetrexate for 6 hours, followed by separate Flag and mCherry immunoprecipitation. As the concentration of trimetrexate increased, the binding of SOX2 to IQGAP3 decreased (Fig. 8q and Supplementary Fig. 11e), which we confirmed in GSCs (Supplementary Fig. 11f). Concurrent treatment of GSCs with trimetrexate and MG132 increased SOX2 protein levels (Fig. 8r). Trimetrexate treatment increased SOX2 ubiquitination (Fig. 8s), suggesting that trimetrexate inhibits the interaction between IQGAP3 and SOX2, promoting the degradation of SOX2. To evaluate the therapeutic efficacy of trimetrexate, temozolomide (TMZ), and radiotherapy (RT) alone or in combination, GSCs (MES20) were cultured on stiff Matrigel, followed by treatment with temozolomide (TMZ, 200 μM), radiotherapy (RT, 2 Gy), and/or trimetrexate (100 nM). The therapeutic efficacy of the TMZ plus trimetrexate combination regimen was superior to that of either agent alone. However, this combined effect was weaker than that of the RT plus trimetrexate regimen. The triple-combination regimen of RT + TMZ + trimetrexate exhibited the most potent therapeutic efficacy among all treatment groups (Supplementary Fig. 11g).

To translate our in vivo findings, we orthotopically xenografted GSCs into the brains of immunocompromised mice. Tumor-bearing mice were treated with trimetrexate administered via intraperitoneal injection (i.p.) and a single dose of ionizing radiation (IR) at 2 Gy (Fig. 9a). Combination treatment of trimetrexate and IR prolonged survival of tumor-bearing mice and reduced tumor burden as compared to monotherapy controls alone (Fig. 9b–d and Supplementary Fig. 12a–d). No increases in toxicity in vivo were detected (Supplementary Fig. 12e–g). To investigate the effects of trimetrexate treatment on SOX2 protein levels and DNA damage in tumor cells in vivo, mice were administered trimetrexate at a dose of 10 mg/kg per injection for 5 consecutive days. After drug administration, the mice were subjected to radiotherapy (RT). Twenty-four hours post-RT, immunofluorescence staining was performed on xenograft tumors. The results demonstrated that trimetrexate treatment led to a reduction in SOX2 protein expression and an increase in γH2AX levels in tumor cells (Fig. 9e–h).

Fig. 9 — a Schematic diagram illustrating the in vivo experiment involving Trimetrexate and IR. GSCs: Glioma Stem Cells. b H&E-stained brain sections at indicated times. IR: Ionizing Radiation (X-ray). **c,d** Kaplan-Meier survival curves (n = 5 per group). **e–h** MES20 GSC-bearing mice received vehicle or trimetrexate (10 mg/kg/day, 5 days) followed by radiotherapy (RT). Brains collected 24 h post-RT for immunofluorescence. Representative images (e,f) and quantification of SOX2 (g, n = 6 fields, *** p = 0.0003) and γ-H2AX (h, n = 100 cells, **** p < 0.0001). IF: Immunofluorescence. i Trimetrexate concentration in brain tissue. j HEK293T cells with mCherry-SOX2, HA-WWP2, Flag-IQGAP3 treated with trimetrexate (0, 200, 400 nM, 6 h). Immunoblots of input and IP. IB: Immunoblot. k MES20 GSCs were transduced with mCherry-SOX2, HA-WWP2, and increasing amounts of Flag-IQGAP3, then cultured on soft Matrigel. Cell lysates were used for input western blots or immunoprecipitated with anti-mCherry antibody. l Western blot analysis identified IQGAP3 domains mediating interaction with SOX2 in MES20 GSCs, following overexpression of mCherry-SOX2 and Flag-IQGAP3 (domain-specific constructs). m Western blot analysis assessed trimetrexate’s effect on IQGAP3-SOX2 interaction in MES20 GSCs overexpressing mCherry-SOX2 and Flag-IQGAP3. Immunoblots are representative of three independent experiments with similar results. Data are presented from three independent experiments (j,h). Data are presented as mean ± s.d. (g,h). Statistics: two-tailed unpaired t-test (g,h); Log-rank test (c,d). IP: immunoprecipitation. Source data are provided as a Source Data file.

To investigate the final accumulated concentration of trimetrexate in tumor tissues of xenograft models in our therapeutic regimen, we measured trimetrexate concentrations in brain tissues using liquid chromatography-tandem mass spectrometry (LC-MS/MS) (see Methods for details). The results showed that trimetrexate concentrations in brain tissues reached 260–540 nM (Fig. 9i). To confirm that these concentrations effectively interfere with the SOX2/WWP2 interaction, we transduced 293 T cells with Flag-IQGAP3, HA-WWP2, and mCherry-SOX2 constructs. The transduced cells were then treated with trimetrexate at different concentrations (0, 200, and 400 nM) for 6 h, followed by protein extraction for co-immunoprecipitation (Co-IP) assays. Increasing trimetrexate concentrations attenuated the binding between SOX2 and IQGAP3 proteins while enhancing the interaction between SOX2 and WWP2—supporting an on-target molecular effect of trimetrexate (Fig. 9j). To validate that IQGAP3 regulates the binding of SOX2 to WWP2 and that trimetrexate modulates this process, we overexpressed Flag-IQGAP3, HA-WWP2, and mCherry-SOX2 individually in GSCs and repeated the Co-IP assay, which confirmed the regulatory role of IQGAP3 and the modulatory effect of trimetrexate (Fig. 9k–m). In clinical practice, radiotherapy is typically administered in 30 fractions of 2 Gy over 6 weeks. These fractionated doses are essential to induce DNA damage repair mechanisms and promote tumor cell death. To enhance the synergistic effect between trimetrexate and radiotherapy, we increased the number of radiation fractions and divided the mice into four groups: 1) DMSO control group; 2) Trimetrexate monotherapy group; 3) Radiotherapy alone group (three fractions of 2 Gy each); 4) Trimetrexate plus radiotherapy combination group (three fractions of 2 Gy each, administered on days 1, 3, and 5 post-completion of trimetrexate treatment, respectively). The results demonstrated that increasing the radiation fraction number combined with trimetrexate treatment prolonged the overall survival of mice bearing xenograft tumors (Fig. 10a–g).

Fig. 10 — a Schematic diagram illustrating the in vivo experiment involving trimetrexate and IR. GSCs: Glioma Stem Cells. **b–g** Representative in vivo bioluminescence imaging (b and d) and Tumor growth curve from in vivo bioluminescence analysis of mice bearing the indicated xenografts (c and e n = 5 mice per group). Kaplan - Meier survival curves of mice bearing the indicated xenografts (f and g, n = 5 mice per group). (c *** p = 0.0003, **** p < 0.0001; e, ** p = 0.0013, **** p < 0.0001). Data are presented as mean ± s.d. (c,e). Two-way ANOVA (c,e). Log-rank test (f, g). RT: Radiotherapy. Source data are provided as a Source Data file.

Discussion

The tumor microenvironment of the ECM is crucial for the occurrence and progression of malignant tumors and serves as a niche for the growth of cancer stem cells (CSCs). However, due to the heterogeneity of tumor cells, different types of CSCs have distinctly different requirements for the stiffness of the extracellular matrix^45–47. Changes in ECM structure and stiffness are considered hallmarks of cancer, and alterations in ECM stiffness have been linked with tumor recurrence⁴⁸ and treatment resistance⁴⁹. The increase in ECM stiffness is not a static state but a dynamic result of accelerated ECM turnover. MMP-2 and MMP-9, alongside cross-linking enzymes like Lysyl Oxidase (LOX), are key players in establishing a reciprocal feedback loop between ECM mechanical alterations and cellular responses. In high-stiffness environments, these MMPs are coordinately upregulated to facilitate the structural reorganization required for tumor invasion and metastasis⁵⁰. Increased stiffness is often driven by excessive collagen deposition. Collagen fragments (degradation products) act as signaling molecules to further induce the expression of MMP-2 and MMP-9, often through the activation of VEGF or CXCR4 pathways⁵¹. ECM stiffening promotes chemotherapeutic drug resistance via multiple mechanisms⁴⁸. Compared with cells adjacent to stiff ECM, tumor cells located in soft ECM regions have lower DNA double-strand break repair efficiency, resulting in DNA defects and increased sensitivity to genotoxic drugs⁵². Although the effect of matrix stiffness on chemotherapy effectiveness and resistance has been well studied, the role of matrix stiffness underlying radiotherapeutic responses in GBM remains unclear.

Here, we found that stiff matrix regions spatially correlate with stem-cell rich areas in human GBM, although the stiffness in vivo is far more complex than the homogeneous stiffness of cell culture matrices. Notably, the GBM core region, while classified as a high-stiffness area, exhibits inherent variation (heterogeneity) in extracellular matrix stiffness. This in vivo mechanical heterogeneity likely explains why 15-20% of genes are highly expressed in both the GBM core region and in vitro stiff matrices. Mechanical stimulation in stiff matrix induced expression of IQGAP3 in GSCs, leading to stabilization of SOX2 in the cytoplasm, and increased SOX2-mediated transcription of stemness and treatment resistance-related pathways in the nucleus (Supplementary Fig. 13). IQGAP3 promotes tumor progression in other cancers⁵³. However, the mechanistic role of IQGAP3 in brain tumors remains to be further elucidated.

Yes-associated protein (YAP1) is a mechanosensitive transcriptional regulator that plays a role in the pathogenesis of cancer and other diseases⁵⁴. YAP1 responds to physical signals by translocating from the cytoplasm to the nucleus. In tumor cells, YAP1 enters the nucleus and activates the TEAD family of transcription factors³³. YAP1-TEAD interactions promote CSC maintenance⁵⁵, radioresistance of tumor cells⁵⁶, epithelial-mesenchymal transition (EMT), and increased migratory ability⁵⁴. Here, we found that mechanical stimulation in the ECM increased nuclear entry of YAP1 in GSCs. Within the nucleus, YAP1 bound to TEAD1, activating it. TEAD1 directly bound to the promoter region of IQGAP3, increasing IQGAP3 protein expression in GSCs.

Upregulation of the transcription factor SOX2 maintains GSC stemness and cancer cell growth and treatment resistance^57,58. Thus, SOX2 serves as an attractive target in GBM. SOX2 is also expressed in adult NSCs and neural progenitor cells (NPCs) in neurogenic regions, such as the subventricular zone (SVZ), the subgranular zone (SGZ) of the hippocampus, and the ependymal lining of the adult central canal^59,60. SOX2 knockout in NPCs in vitro reduces their differentiation into neurons⁶¹. In vivo depletion of SOX2-positive adult stem cells leads to signs of premature aging in mice, such as kyphosis, hair whitening, and reduced fat mass, indicating that SOX2 plays a vital role in maintaining adult stem cells⁶². Thus, direct targeting of SOX2 may induce severe side effects that may counteract the advantages of SOX2-specific cancer treatment. SOX2 is a small transcription factor that is difficult to target directly⁵⁸. As such, identifying upstream or downstream regulators of SOX2 that are more easily targetable is essential. We discovered that ECM stiffness influences the degradation of SOX2. GSCs cultured in soft stiffness increased SOX2 protein degradation, while in regions of high stiffness, IQGAP3 competitively bound to SOX2, reducing the degradation of SOX2 protein. GSCs appear enriched in regions of stiff ECM, while neural precursor cells survive in a relatively soft ECM^5,63,64. Therefore, selective targeting of IQGAP3 may result in GSC-specific killing while sparing side effects caused by damage to normal brain. Matrix stiffness promotes SOX2 transcription by activating TAZ in CSCs⁴⁵, which is highly consistent with our findings. In our study, we clarified the mechanism by which matrix stiffness regulates SOX2 in GSCs.

Although the ECM of primary IDH-mutant GBM is relatively soft, the ECM stiffness of recurrent IDH-mutant gliomas increases compared to the primary tumor and is equally as hard as the ECM found in primary IDH-wildtype GBM⁶. Standard-of-care treatment for GBM includes surgical resection, chemoradiotherapy, and adjuvant chemotherapy. The TME evolves post-treatment, including deposition and remodeling of ECM molecules, ultimately resulting in a stiffer ECM⁶. As GSCs contribute to GBM recurrence⁷, we posit that blocking the response of mechanical stimulation in GSCs can augment existing therapy for GBM. We found that IQGAP3 was a key factor regulated by matrix stiffness in GSCs and that targeting IQGAP3 enhances the radiosensitivity of GSCs.

IQGAP3 expression was elevated in high-grade and recurrent gliomas compared to low-grade gliomas and normal brain samples. IQGAP3 informed poor prognosis in publicly available patient cohorts. Using an in silico drug screen, we identified trimetrexate as an inhibitor of IQGAP3 function through disruption of its binding to SOX2. Trimetrexate is a folate antagonist that has antitumor activity in leukemia and solid tumor models following oral, intravenous, or intraperitoneal administration^65,66. Trimetrexate crosses the blood-brain barrier⁶⁶ and is highly lipophilic, allowing for cellular uptake. Trimetrexate inhibition of DHFR depletes intracellular tetrahydrofolate pools, which are essential for de novo purine and thymidylate synthesis⁶⁷. Given that GSCs are highly metabolic and depend heavily on one-carbon metabolism for rapid proliferation and epigenetic maintenance⁶⁸, the disruption of this pathway by trimetrexate leads to growth arrest and apoptosis.

Translating trimetrexate as a IQGAP3-targeting agent warrants further investigation. As a known folate antagonist, trimetrexate has other effects on cellular metabolism, which contribute to systemic toxicity. At conventional doses, the primary toxicity in mice is reversible, but trimetrexate causes neurotoxicity at high doses, suggesting the need to weigh the clinical utility of drug treatment in the brain with potential adverse effects⁶⁶. Therefore, while trimetrexate served as a valuable proof-of-concept tool to validate the therapeutic potential of disrupting the IQGAP3-SOX2 interaction, its clinical use may be limited by its off-target effects.

An alternative translational approach may leverage the rational design and development of highly specific, low toxicity small molecule inhibitors targeting IQGAP3 itself or targeting its SOX2 binding domain. Such a selective drug could exploit the GSC-specific niche defined by the stiff ECM and high IQGAP3 expression, achieving synergistic efficacy with standard-of-care radiotherapy while minimizing the side effects associated with general metabolic disruption or direct SOX2 inhibition. Targeting IQGAP3 with trimetrexate displayed combinatorial benefit with radiotherapy to inhibit GBM growth in vitro and in vivo. Taken together, our study reveals the role of matrix stiffness in maintaining the stemness and radioresistance of GSCs through the IQGAP3-SOX2 axis. Targeting key regulators of pathways involved in GSC maintenance and response to mechanical stimuli in the ECM may serve as a therapeutic strategy for GBM patients, provided that more specific IQGAP3 inhibitors are developed.

Methods

This study complies with all relevant ethical standards and has been approved by the Ethics Committee and Institutional Review Board at the University of Pittsburgh, Tongji Medical College, Huazhong University of Science and Technology, and Case Western Reserve University.

Human glioma and non-tumorous brain tissue

All pathological glioma samples, adjacent brain tissues, and non-tumorous brain tissues (from traumatic brain injury) used in this study were obtained from excess samples resected during neurosurgical procedures at the University of Pittsburgh, Tongji Medical College, Huazhong University of Science and Technology. All samples were from both genders as sex and gender were identified as insignificant factor in our analysis. Written informed consent was obtained from all participants. All patient studies were conducted in accordance with the Declaration of Helsinki.

Glioblastoma stem cell derivation and cell culture

Glioblastoma (GBM) tissues were obtained from excess surgical resection samples from patients at Case Western Reserve University. Written informed consent was secured from all patients, and the study adhered to institutional review board-approved (Protocol #090401). Neuropathologists examined and confirmed diagnosis and grade for all samples. Patient-related studies complied with the Declaration of Helsinki.

GSCs (GSC387, MES20, 3028, 23, and 456) were derived from human specimens as previously described^69,70. To reduce cell culture-related artifacts, patient-derived xenografts served as renewable sources for GSCs. The NSC11 line (hNSC11, Alstem) was derived from human induced pluripotent stem cells (iPSCs). Human neural progenitors (HNP1, HN60001, ArunA Biomedical) were fully differentiated and obtained as adherent cells from the hESC WA09 line. ENSA (ENStem-A, Millipore) are human embryonic stem-derived neural progenitors. All GSC and NSC lines were cultured in Neurobasal medium (Gibco, 21103049) supplemented with B27 without vitamin A (Gibco, 12587010), 20 ng/mL recombinant human EGF (R&D Systems, 236-EG-01M), 20 ng/mL recombinant human bFGF (R&D Systems, 4114-TC-01M), sodium pyruvate (Gibco, 11360070), GlutaMAX (Gibco, 35050061), and streptomycin–penicillin (Gibco, 15140122). Cultures were maintained at 37 °C with 20% oxygen and 5% carbon dioxide. Matched serum-derived glioblastoma cells (DGCs) were cultured in DMEM (Gibco, 11995065) supplemented with 10% FBS (Gibco, 26140079) to preserve differentiation status. HEK293T cells, purchased from ATCC (CRL-3216), were maintained in DMEM with 10% FBS. Short tandem repeat (STR) analyses were conducted annually to authenticate the identity of tumor models used in the study. Mycoplasma testing was also performed at least once a year to ensure cultures remained uncontaminated. All experimental procedures complied with applicable regulatory standards.

Xenografts

All animal experiments were conducted in accordance with a protocol approved by the Institutional Animal Care and Use Committee (IACUC) at the University of Pittsburgh (Protocol #21049014). Mice were housed in specific-pathogen-free conditions, maintained at an ambient temperature of 20–26 °C with 30–70% humidity, and subjected to a 12-hour light–dark cycle. A maximal tumor size of 15 mm in any direction was not exceeded in any experiment. Both male and female mice were used in studies. Sex was considered in the study design by including both genders, but the research focus was on the biological effects of GSC intracranial xenografts, and the experimental outcomes (including tumor growth, neurological symptom development, and histological changes) were not associated with sex. No significant sex-specific differences were observed in any measured endpoints, so sex was not included as a variable in the final data analysis. For intracranial xenograft studies, healthy NSG mice (NOD. Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ, strain 005557, Jackson Laboratory), aged 4–6 weeks, were randomly selected for intracranial injections. A total of 157 mice (64 males and 93 females) were used in the intracranial xenograft experiment, with male and female mice randomly allocated to experimental groups. Detailed information regarding the sex distribution of mice is provided in the corresponding source data. A total of 10,000 GSCs were implanted into the right cerebral cortex at a depth of 3.5 mm using standard procedures. Mice were housed under veterinary supervision, and their status was closely monitored. Animals were observed for the development of neurological symptoms, including hunched posture, altered gait, lethargy, and weight loss. Upon the onset of such signs, mice were euthanized, and their brains were harvested. Brains were fixed in 4% paraformaldehyde, processed into paraffin-embedded sections, and analyzed histologically using H&E staining.

In vivo therapy

Trimetrexate (MCE, HY-10373) treatment was initiated 7 days after intracranial GSC injection in mice. The drug was administered intraperitoneally at a dose of 10 mg/kg once daily for 5 consecutive days, followed by radiotherapy. Mice were irradiated with 2 Gy per fraction using a Precision SmART+ animal irradiator. Imaging for brain targeting was performed with a 2.0 mm aluminum (Al) filter, while therapeutic irradiation was delivered using a 0.3 mm copper (Cu) treatment filter and a 10 mm fixed circular collimator. Radiation doses were administered via dual-opposed lateral beams, with each beam delivering half of the total dose per fraction.

Conventional polyacrylamide hydrogels

Glass-bottom dishes (Cellvis, D35–14-1-N and D60-30-1-N) were prepared for gel attachment by treating them with 100～200 μl of Bind-Silane solution for 20 minutes at room temperature. The Bind-Silane solution consisted of 3-(trimethoxysilyl)propylmethacrylate (7.15% v/v, Sigma-Aldrich, M6514) and acetic acid (7.15% v/v) dissolved in absolute ethanol. Following treatment, the solution was removed, and the glass surfaces were washed twice with ethanol. The dishes were then allowed to dry completely before further use. To prepare homogeneous hydrogels with a constant Young’s modulus, predefined ratios of 40% (w/v) acrylamide (Sigma-Aldrich, A4058) and 2% (w/v) N,N-methyl-bis-acrylamide (Sigma-Aldrich, M1533) were combined in PBS and mixed gently using a vortex. The final concentrations were optimized to achieve the target Young’s modulus, as outlined in Supplementary Data 12⁷¹. Dichlorodimethylsilane (DCDMS) (Sigma-Aldrich, 440272) was evenly applied to the surface of each cover glass to ensure complete coverage. The reaction was allowed to proceed for up to 5 minutes. Excess DCDMS was then removed with a Kimwipe, followed by rinsing the cover glass with distilled water for 1 minute. The solution was then pipetted onto the glass-bottom dish, and a circular coverslip was carefully placed over the droplet. Polymerization was allowed to proceed for approximately 30 minutes at room temperature. Once polymerization was complete, the gel was immersed in PBS for about 5 minutes. The top coverslip was then gently removed, and the gel was washed twice with PBS to eliminate any residual acrylamide. Hydrogels with continuous 2D stiffness gradients were generated following a previously established protocol⁷². In summary, acrylamide prepolymer solutions with stiffness values of 0.5 kPa and 10 kPa were prepared. Fluorescent microspheres (0.2 μm, 505/515 nm, Invitrogen, F8811) were added to the 10 kPa solution at a final concentration of approximately 1.0 × 10¹¹/ml. Upon initiating polymerization, the two solutions were allowed to diffuse together on a glass-bottom dish, covered with a glass coverslip. This process produced a gradient where the density of microspheres exhibited a linear correlation with the substrate’s Young’s modulus. Before use, the hydrogels were activated using a solution containing 0.2 mg/ml Sulfo-SANPAH (Thermo Fisher Scientific, 22589) dissolved in 50 mM HEPES buffer (pH 8.5). A volume of 500 μl of this solution was applied to the surface of each hydrogel. The hydrogels, along with the solution, were exposed to UV light (365 nm) for 10 minutes at a distance of approximately 3 inches to activate the Sulfo-SANPAH. Following activation, the hydrogels were rinsed three times with 50 mM HEPES to remove any residual compounds. Finally, the hydrogels were solidified by incubating them overnight at 37 °C in a solution of 0.1 mg/ml Collagen I (Sigma-Aldrich, CLS354231) prepared in 50 mM HEPES buffer. Lastly, hydrogels were rinsed with PBS and UV sterilized.

Immunoblotting

Cells were lysed in RIPA buffer (50 mM Tris pH 7.4, 150 mM NaCl, 1% NP-40, 0.5% sodium deoxycholate, 0.1% SDS) containing protease inhibitor cocktail. For immunoblotting, 30-50 μg proteins were subjected to SDS-PAGE and then transferred to polyvinylidene difluoride membrane. Blots were first blocked with 5% fat-free milk, then incubated with primary antibodies and appropriate horseradish peroxidase conjugated secondary antibodies. Protein signal was detected with Clarity Western ECL Substrate. Detailed information about the antibodies used can be found in Supplementary Data 13. Uncropped and unprocessed scans of all blots are provided in the Source Data file with molecular weight markers visible.

Cytoplasm and nucleus fractionation

The cytoplasmic and nuclear fractions were isolated using the Nuclear Extract Kit (Active Motif, 40010) following the manufacturer’s instructions. To assess the completeness of nucleoplasm separation, the cell lysate was mixed with Trypan Blue stain in a 1:1 ratio and examined under a microscope.

Immunofluorescence and immunohistochemistry staining

For IF analyses, cells were grown on matrigel-coated coverslips, and fixed with 4% PFA for 15 min followed by permeabilization with 0.1% Triton X-100 in PBS for 15 min. After washing with PBS for three times, cells were blocked in PBS buffer with 5% goat serum for 1 h. The relevant primary antibodies diluted in blocking buffer were then incubated at 4°C overnight, followed by secondary antibodies incubation for 1 h at room temperature according to manufacturer’s instructions. Cryosectioned tumor-bearing brain tissues were permeabilized and blocked in PBS buffer with 0.1% Triton X-100 and 5% goat serum. Sections were then incubated with primary antibodies at 4°C overnight and appropriate secondary antibodies for 1 h at room temperature. For IHC staining, paraffin-embedded sections were deparaffinized and underwent antigen retrieval in sodium citrate buffer in a microwave oven for 15 min. Sections were permeabilized and blocked in PBS buffer with 0.1% Triton X-100 and 5% goat serum. Sections were then incubated with IQGAP3 antibody at 4°C overnight and appropriate secondary antibody for 1 h at room temperature and performed as manufacturer’s instructions. Detailed information about the antibodies used can be found in Supplementary Data 13.

RNA extraction and quantitative real-time PCR

Total RNA was isolated using the TRIzol Reagent (Life Technologies, 15596018) and the Direct-zol RNA Miniprep Kit (Zymo Research, R2052). Reverse transcription was performed with the High-Capacity cDNA Reverse Transcription Kit (Thermo Fisher Scientific, 4368814). Quantitative real-time PCR was carried out using the SYBR Green Master Mix (Thermo Fisher Scientific, 4309155) on a CFX96 Touch Real-Time PCR Detection System (Bio-Rad). Data were normalized to ACTB expression levels. The primers utilized in this study are provided in Supplementary Data 13.

Neurosphere formation and proliferation assay

The neurosphere formation capacity was evaluated using in vitro extreme limiting dilution assays. Cells were plated at decreasing densities (50, 40, 30, 20, 10, and 1 cell per well) in a 96-well plate. The presence and quantity of neurospheres in each well were assessed 7–14 days post-plating. Data analysis was performed using the software available at http://bioinf.wehi.edu.au/software/elda. CellTiter-Glo (Promega) was used to measure cell viability at specified times.

Click-iT EdU incorporation assay

Cell proliferation was assessed using the Click-iT EdU Cell Proliferation Kit (Thermo Fisher Scientific, C10339). Cells were incubated with 10 μM EdU for 2 hours, and images of five random fields per condition were captured. The percentage of EdU-positive cells was calculated by comparing the number of EdU-positive cells to the total number of DAPI-positive cells, with analysis performed using ImageJ software.

Chromatin immunoprecipitation followed by PCR

ChIP assays were performed using the Magna ChIP A/G Chromatin Immunoprecipitation Kit (Millipore, 17-10085). Cells were cross-linked with 1% formaldehyde for 10 minutes, and the reaction was quenched with 0.125 M glycine. Nuclei were isolated, lysed, and sonicated to fragment chromatin into 200–500 bp fragments. The sheared chromatin was diluted and incubated overnight with 5 μg of the specific antibody or IgG, followed by binding with magnetic protein A/G beads (Thermo Fisher Scientific, 26162). After a series of washing steps, chromatin was eluted, cross-links were reversed, and ChIP DNA was purified for PCR analysis. Details of the primers and antibodies used are provided in Supplementary Data 13.

Comet assay

DNA damage repair was evaluated using single-cell gel electrophoresis under alkaline conditions, following the protocol provided with the kit (Cell Biolabs, STA-350).

ChIP-seq analysis

For chromatin Immunoprecipitation sequencing analysis (ChIP-Seq), GSC cells were cross-linked with 1% paraformaldehyde for 10 minutes, followed by quenching with glycine for 5 minutes at room temperature. After cell lysis, the fixed chromatin was sonicated to generate DNA fragments ranging from 200 to 600 bp. Immunoprecipitation was performed using a SOX2 antibody (Thermo Fisher Scientific, MA1-014) and Magna ChIP Protein A/G Magnetic Beads (Sigma-Aldrich, 16-663). For library preparation, 1 to 10 ng of immunoprecipitated DNA and the NEBNext Ultra DNA Library Prep Kit for Illumina were used, and the samples were subjected to sequencing on the NovaSeq X Plus. Raw ChIP-seq reads were aligned to the human genome (hg38) using Bowtie2 (version 2.5.3) with default settings. Quality control was performed using FastQC (version 0.12.0), and adapter trimming was done with Trimmomatic (version 0.39). Read coverage was analyzed using deepTools (version 3.5.5) and visualized in the IGV software. Peak calling was performed with MACS2 (version 2.1.2) using default parameters and a p-value cutoff of 1 × 10-4. Heatmaps and metagene plots were generated using the computeMatrix function from deepTools.

RNA sequencing and data analysis

Total RNA was extracted and prepared for library construction using the Illumina TruSeq Stranded Total RNA Library Prep Kit at Novogene. Libraries were sequenced with paired-end 150-bp reads. Raw FASTQ reads were processed using Trim Galore for quality trimming. Transcript mapping was performed with HISAT2 against the human reference genome (hg38). SAMtools was utilized for sorting, indexing, and format conversion from SAM files. Quantification and differential expression analysis were conducted using featureCounts and DESeq2. Differentially expressed genes (DEGs) were identified with a fold change threshold at 1.5 and an adjusted P-value of <0.05. Gene Ontology (GO) enrichment analysis and gene set enrichment analysis (GSEA) were performed using clusterProfiler. For scoring the activities of individual pathways, single sample GSEA (ssGSEA) was conducted using the GSVA package with the ssGSEA method in R. Matrix stiffness score was inferred using genes that have been reported or predicted to be associated with matrix stiffness⁴⁸. Cancer stemness score⁷³ and pan stemness score⁷⁴ were inferred using published gene set (Supplementary Data 14).

Single-cell RNA sequencing reanalysis

The scRNA-seq data of 65,655 cells from 28 early-passage GSC cultures derived from 24 patients and 14,207 malignant tumor cells from seven GBM²⁶ were used for reanalysis. The predetermined annotation and PCA coordinates were according to the original data and code. We first calculated the DEGs (|Log2 fold change | >2, adjusted P < 0.05) between in vitro GSC culture and in vivo tumor cells by Findmarkers (Seurat package) in R. To determine GSCs and DGCs from 14,207 malignant tumor cells, a logistic regression classifier was calculated as previously described²⁷. Specifically, on the training set, the logistic regression classifier was trained using the best hyperparameter determined from fivefold cross-validation using the caret package in R based on the DEGs. We repeated the 80–20 train–test split randomly (stratified) 30 times and tested the accuracy and stability of these models. We found that the model was robust with high accuracy, and we picked the best-performing model (highest test accuracy) to predict the entire dataset. The correctly classified tumor cells were reassigned as DGCs (n = 13,226) and the misclassified tumor cells were reassigned as GSCs (n = 981). The activity of sensing mechanical stimuli was inferred using AUCell (version 1.20.2) based on GOBP: Detection of mechanical stimulus. Scores were normalized between 0 and 1 by subtracting the minimum and dividing by the range, and two-tailed unpaired t-test was used for comparison.

Co-immunoprecipitation analysis

Cells were lysed in IP buffer (Thermo Fisher Scientific, 87788) supplemented with a protease inhibitor cocktail (Thermo Fisher Scientific, A32953). After lysis, cell debris was removed via centrifugation. To minimize non-specific binding, the lysates were pre-incubated with an IgG antibody matching the species of the immunoprecipitation antibody and protein A/G magnetic beads (Thermo Fisher Scientific, 88802) for 2 h at 4 °C. The pre-cleared lysates were then incubated overnight at 4 °C with the designated primary antibody and protein A/G beads, anti-Flag M2 beads (Sigma-Aldrich, M8823), or anti-mCherry agarose (Proteintech, rta). The resulting immunoprecipitates were washed three times with IP buffer, boiled, and subsequently analyzed by immunoblotting following standard protocols. Details of the antibodies used are provided in Supplementary Data 13.

Mass Spectrometry analysis

A total of 10 samples were analyzed, including biological duplicates (n = 2 per condition) of the IQGAP3 immunoprecipitation (IQGAP3 IP) group, IgG control group, and DNA pull-down samples (Control, Region#1, Region#2). All MS data were derived from independent biological replicates (n = 2 per experimental condition) to ensure the reliability of the results. Samples were prepared according to the standard protocols for co-immunoprecipitation (co-IP) analysis and DNA pull-down assay, respectively. Liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis was performed using a nanoElute2 ultra-high-performance liquid chromatography (UHPLC) system (Bruker Daltonics, Bremen, Germany) coupled to a timsTOF Pro2 mass spectrometer (Bruker Daltonics), which was equipped with a CaptiveSpray nanoelectrospray source. A 2 µl aliquot of the peptide digest (approximately 200 ng of peptides) was loaded onto a capillary C18 column (25 cm length, 75 μm inner diameter, 1.6 μm particle size, 120 Å pore size; IonOpticks Aurora Gen3, Fitzroy, VIC, Australia). Peptide separation was conducted at 55°C with a 60-min gradient at a flow rate of 300 nl/min. The mobile phase consisted of mobile phase A (MPA: 0.1% formic acid, FA) and mobile phase B (MPB: 0.1% FA in acetonitrile, ACN). A linear gradient from 2% to 35% MPB was applied for 60 min, followed by a 5-min wash at 95% MPB and re-equilibration at 2% MPB for 6 min (total run time: 71 min). The timsTOF Pro2 mass spectrometer was operated in data-dependent PASEF (Parallel Accumulation-Serial Fragmentation) mode, collecting full-scan mass spectra over a range of m/z 100 to 1700. Ion mobility resolution was set to 0.60–1.60 V·s/cm with a ramp time of 100 ms. Ten PASEF MS/MS scans were acquired per 1.1-second cycle. The active exclusion time was set to 0.4 min, and the intensity threshold for MS/MS fragmentation was 2.5×10⁴. Low m/z ions and singly charged ions were excluded from PASEF precursor selection, and MS/MS spectra were acquired using ramped collision energy as a function of ion mobility. Ion mobility-mass spectrometry (IM/MS) raw files were processed using MSFragger (version 3.6). MS/MS spectra were searched against the human UniProt/Swiss-Prot database (or a custom database containing IQGAP3 and its common interactors). The search parameters were set as follows: precursor ion mass error of 10 ppm, fragment ion mass error of 20 ppm, and a maximum of two missed cleavages. Carbamidomethylation of cysteine was designated as a fixed modification, while oxidation of methionine and acetylation of protein N-termini were set as variable modifications. The false discovery rate (FDR) was capped at 0.01 (1%) at the peptide-spectrum match (PSM) level to ensure the validity of peptide identification. Protein quantitation was performed using peptide abundances. Differential expression analysis was conducted in R (version 4.0.3) using the DEP package (version 1.30.0), where Log₂fold changes (Log₂FC) and adjusted p-values (via Benjamini–Hochberg FDR correction) were calculated to identify differentially expressed proteins. The MS proteomics raw data have been deposited in the ProteomeXchange Consortium via the PRIDE partner repository with dataset identifiers PXD072725 and PXD072748.

DNA pull-down assay

Ten million cells were cultured, and nuclei were isolated using the nuclear fractionation method described above. IP lysis buffer (Thermo Fisher Scientific, 87788) was added to the nuclear pellet and sonication was performed. The mixture was then centrifuged at 4 °C, and the supernatant was collected. The supernatants were incubated with 8 μg of biotinylated DNA overnight at 4 °C. Following incubation, anti-biotin beads (Thermo Fisher Scientific, 65001) were added to the samples for 2 to 4 h in a rotating mixer at 4 °C. The pulled-down proteins were eluted for mass spectrometry or western blotting analysis. The primer information used for constructing the biotinylated DNA is listed in Supplementary Data 13.

Flow cytometry analysis

Brain tumors were dissected, minced, and digested using 5 mg/ml collagenase Type II (Gibco, 17101-015) and 0.1 mg/ml recombinant DNase I (Sigma-Aldrich, 4536282001) at 37 °C for 30 min. The resulting cell suspension was passed through a 40 µm filter to remove clumps and ensure single-cell suspensions. Brain cell suspensions were separated using a 30% Percoll gradient at 700 g for 30 min at 4 °C. Red blood cells were lysed using ACK Lysing Buffer (Thermo Fisher Scientific), and the remaining cells were resuspended in staining buffer (BD Biosciences, 554656). For human GBM tissues, cells were first stained with the Piezo1 antibody (Novus Biologicals, NBP1-78446SS, 1:100) on ice for 45 min, followed by incubation with a secondary antibody (A-11008, Thermo Fisher Scientific, 1:200) on ice for an additional 45 min. The samples were then stained with an anti-human CD133 antibody (BD Biosciences, 566594, 1:100) on ice for 45 min. For tumor tissues derived from GSC xenografts, cells were first stained with anti-human CD147 (BioLegend, 306204, 1:200) and anti-human CD133 on ice for 45 min.

Following fixation and permeabilization (eBioscience, 88-8824-00), the samples were stained for SOX2 (BioLegend, 656111, 1:200) and incubated on ice for 45 minutes. Flow cytometry analysis was performed using the LSR Fortessa (BD) system, and data were processed with FlowJo software (v.10). Details of the antibodies used are provided in Supplementary Data 13.

Virtual screening docking

The 3D crystal structure of IQGAP3 was obtained via AlphaFold2 (https://alphafold.ebi.ac.uk/). The protein was further refined in Discovery Studio 2.5 by removing all water molecules, adding hydrogen atoms, removing protein polymorphs, and supplementing the structure with non-intact amino acid residues. Ligands were downloaded and prepared by rectifying their bond angles and bond orders and were subsequently minimized using the CHARMm force field. For drug screening, Lipinski’s rule of five (ROF) was initially used. Lipinskis five rules are predicted by using the molecular properties in the computational small molecule protocol in Discovery Studio 2.5, including molecular weight (MWT) ≤ 500, hydrogen bonded acceptor ≤ 10, hydrogen-bonded donor ≤ 5, and ClogP ≤ 5 (or MLogP > 4.15). respectively. The screened ligands were subjected to docking analysis. The initial screening was performed in the Libdock module of Discovery Studio 2.5 (Dassault Systems BIOVIA). LibDock is a rigid docking method with an algorithm based on molecular dynamics annealing. All molecules from the last step were docked into the site of the SOX2 binding domain, and individual Libdock analysis were performed. Then, all compounds were ranked according to the Libdock scores, and the top-ranked compounds were selected for precise docking by CDock.

Protein expression

The amino acid sequence of the human IQGAP3 (1-670) protein was optimized using the MaxCodon™ Optimization Program (V13). The optimized IQGAP3 (1-670) genes were then inserted into pET30a expression vectors. Integration was confirmed through enzyme digestion and sequencing. The expression vector was transformed into BL21 (DE3) competent cells, which were evenly spread on LB plates containing 50 μg/mL kanamycin sulfate. These plates were incubated overnight at 37 °C in an inverted position. A single clone was selected from the transformed plates and inoculated into 4 mL of LB medium (supplemented with 50 μg/mL kanamycin sulfate). The culture was grown until the OD600 reached 0.5–0.8, after which isopropyl β-D-1-thiogalactopyranoside (IPTG) was added to a final concentration of 0.2 mM to induce protein expression. Induction was performed at 15 °C for 16 hours. Following induction, the culture was centrifuged at 12,000 rpm for 5 minutes to collect the cells. The supernatant was discarded, and the pellet was resuspended in PBS solution. SDS-PAGE loading buffer was added to the sample, which was then heated at 100 °C for 10 minutes and centrifuged. For further processing, the bacterial cells were lysed ultrasonically in a buffer containing 20 mM Tris (pH 8.0), 300 mM NaCl, 20 mM imidazole, and 1% Triton. The inclusion bodies were washed with 50 mM Tris (pH 8.0), 300 mM NaCl, 1% Triton, and 20 mM imidazole buffer. Simultaneously, the Ni-IDA affinity chromatography column was equilibrated. The target protein was eluted using buffers with varying concentrations of imidazole, and the collected fractions were analyzed by SDS-PAGE. High-purity fractions, verified via Ni-IDA affinity chromatography, were pooled and placed into a treated dialysis bag. Dialysis was performed against a buffer containing 1× PBS (pH 7.4), 4 mM GSH, 0.4 mM GSSG, 0.4 M L-arginine, and 1 M urea for renaturation. The protein was subsequently dialyzed into a storage buffer (1× PBS, pH 7.4) for 6–8 h. After renaturation and dialysis, the supernatant was filtered through a 0.22 μm filter, aliquoted, and stored at −80 °C for future use.

Surface plasmon resonance (SPR) assay

IQGAP3 (1-670) protein, with an isoelectric point of 5.15, was immobilized on a CM5 chip through amine coupling. Initially, the protein samples were diluted to 20 μg/ml in 10 mM sodium acetate buffers at pH 5.0, pH 4.5, and pH 4.0, with a flow rate of 10 μL/min for pre-concentration experiments. The results indicated that the coupling was most effective in the 10 mM sodium acetate buffer at pH 4.0. Consequently, for the formal coupling experiment, IQGAP3 was diluted to 50 μg/ml in 10 mM sodium acetate buffer at pH 4.0. The coupling was performed in the ‘specify contact time and flow rate’ mode, with a contact time of 450 seconds and a flow rate of 10 μL/min, resulting in a final immobilization level of 15000 RU. During the interaction experiments, the reaction temperature was maintained at 25 degrees Celsius and the flow rate was set at 30 μL/min. Real-time subtraction of the reference channel was utilized to eliminate nonspecific binding and buffer-induced volume effects. PBS-P+ buffer (pH 7.4) containing 5% DMSO was used as the running buffer. Different concentrations of the small molecule Trimetrexate (0.1, 0.2, 0.39, 0.78, 1.56, 3.125, 6.25, 12.5, 25, 50, 100, 250, 500 μM) were prepared and their binding to IQGAP3 (1-670) was evaluated using SPR. The sample solutions were injected onto the chip surface for 60 seconds at a flow rate of 30 μL/min. After each binding reaction, a natural dissociation of 60 seconds allowed the signal to return to baseline. The experiment included a zero concentration and one repeated concentration. Equilibrium and kinetic constants were calculated using a 1:1 binding model (Global fits) with BIA evaluation software.

In vivo detection of trimetrexate concentration

Glioblastoma stem cells (GSCs) were intracranially implanted into mice. Starting seven days after tumor implantation, trimetrexate (10 mg/kg) or vehicle control (dimethyl sulfoxide, DMSO) was administered via intraperitoneal injection for 5 consecutive days. Six hours after the 5th and final dose, ipsilateral brain tissues (the side of GSC implantation) were harvested, and trimetrexate concentrations in these tissues were measured using liquid chromatography-tandem mass spectrometry (LC-MS/MS).

Cellular thermal shift assay (CETSA)

To investigate whether Trimetrexate functions as a direct ligand for the IQGAP3 protein, a Cellular Thermal Shift Assay (CETSA) was performed based on a previously published protocol⁷⁵. Briefly, HEK293T cells (~10⁷ cells) overexpressing IQGAP3 were pretreated with 200 nM Trimetrexate for 6 hours before proceeding with the CETSA protocol. Cells from each experimental group were harvested, washed once with ice-cold PBS, and resuspended in 1.5 mL of PBS containing a protease inhibitor cocktail. The cell suspension was divided into nine 0.2-mL PCR tubes, each containing 100 µL. The samples were heat-shocked at designated temperatures for 3 min using a Bio-Rad T100 Thermal Cycler to induce protein denaturation, followed by immediate cooling to room temperature for 3 min. Next, the samples underwent three freeze-thaw cycles using dry ice and a thermal cycler set to 25 °C to lyse the cells. The lysates were centrifuged at 10,000 × g for 20 min at 4 °C to separate cell debris and precipitated or aggregated proteins. The resulting supernatants were then mixed with 4× Sample Buffer (Thermo Fisher Scientific, NP0008) and boiled for Western blot analysis.

Public glioma patient datasets

Public glioma databases available on GlioVis (http://gliovis.bioinfo.cnio.es) were utilized to analyze the mRNA expression and correlation of IQGAP3, CD133, CD15, and GFAP. For survival analysis, only IDH-wild-type glioma samples were included, and the data were evaluated using Kaplan–Meier survival curves and the log-rank test.

Statistics and reproducibility

Sample sizes were not predetermined using statistical methods; however, they are consistent with those reported in previous studies^25,27. Mice used for in vivo treatments were randomly assigned to experimental groups, and cells for non-mouse experiments were similarly randomized. Data collection and analysis were not conducted blind to experimental conditions, and no data were excluded from the analyses. Statistical analyses were performed using R, Python, or Prism 9 software, as detailed in the figure legends. Although data distribution was assumed to be normal, this assumption was not formally tested. Unless specified otherwise, data in the figures are presented as mean ± standard deviation (s.d.). For all statistical tests, a p-value of <0.05 was considered indicative of statistical significance unless stated otherwise. Sex was considered in the study design by including both male and female mice, but no sex-specific analysis was performed because the study was not designed to detect sex differences.

Materials availability

All newly generated plasmids described in this study are available from the corresponding author upon reasonable request.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Supplementary information

Supplementary Information^{(6MB, pdf)}

41467_2026_74058_MOESM2_ESM.pdf^{(75.1KB, pdf)}

Description of Additional Supplementary Files

Supplementary Data 1^{(13.6KB, xlsx)}

Supplementary Data 2^{(2MB, xlsx)}

Supplementary Data 3^{(1.7MB, xlsx)}

Supplementary Data 4^{(10.4KB, xlsx)}

Supplementary Data 5^{(2.8MB, xlsx)}

Supplementary Data 6^{(311.6KB, xlsx)}

Supplementary Data 7^{(1.1MB, xlsx)}

Supplementary Data 8^{(1.3MB, xlsx)}

Supplementary Data 9^{(128.1KB, xlsx)}

Supplementary Data 10^{(48.5KB, xlsx)}

Supplementary Data 11^{(254KB, xlsx)}

Supplementary Data 12^{(9.4KB, xlsx)}

Supplementary Data 13^{(17.5KB, xlsx)}

Supplementary Data 14^{(14.4KB, xlsx)}

Reporting Summary^{(143.4KB, pdf)}

Transparent Peer Review file^{(14.1MB, pdf)}

Source data

Source Data^{(13MB, zip)}

Acknowledgements

We thank the Mass Spectrometry Core and the Flow Cytometry Core Facility at the University of Pittsburgh. We are grateful to Professors Xiaobing Jiang, Dongsheng Guo, Baofeng Wang, and Xingjiang Yu, as well as Drs. Chao Song and Zeyang Wan from Tongji Medical College, Huazhong University of Science and Technology for their kind help and support.

Author contributions

P.Z., W.W., and J.N.R. conceived the project, designed the overall experiments, analyzed data and wrote the paper. P.Z., W.W., T.H., X.W., D.W., H.Y., S.T., F.Y., R.W., T.D., H.M., H.L., D.L., and Q.W. performed the experiments. F.P.V. and C.B. performed animal radiotherapy. A.H., P.O.Z., and K.A. provided clinical glioma samples. K.Y., B.C.P., J.Z., and R.C.G. participated in experimental design and revised the manuscript.

Peer review

Peer review information

. Nature Communications thanks Nils Cordes, Eddy Pasquier, Annabelle Ballesta and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

This work was supported by the National Institutes of Health (NIH) under grant numbers CA283998, CA197718, CA238662, CA268634, NS134724, NS136424, and NS103434 (to J. N. Rich), and by the Defense Health Agency under grant number HT9425-23-1-0689 (to J. N. Rich). J. N. Rich is also supported by the American Cancer Society Lisa Dean Moseley Foundation Cancer Stem Cell Consortium. All other authors declare no relevant funding.

Data availability

The publicly available RNA-seq data used in this study are available in the Gene Expression Omnibus (GEO) under accession codes GSE59612, GSE153746, GSE89623, GSE140441, GSE54791, and GSE291583. The publicly available single-cell RNA-seq data used in this study are available in the GEO under accession codes GSE182109 and GSE131928. The ChIP-seq data generated in this study have been deposited in the GEO under accession code GSE291362. The mass spectrometry proteomics data generated in this study have been deposited in the ProteomeXchange Consortium via the PRIDE partner repository under dataset identifiers PXD072725 and PXD072748. The processed RNA-seq and proteomics data are available in the Supplementary Datas of this article. Public glioma patient data are available from GlioVis (https://gliovis.bioinfo.cnio.es/). Source data are provided with this paper.

Code availability

No custom code was generated for this study.

Competing interests

The authors declare no competing interests.

Footnotes

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

These authors contributed equally: Po Zhang, Weichi Wu.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-74058-0.

References

1.Ostrom, Q. T. et al. CBTRUS Statistical Report: primary brain and other central nervous system tumors diagnosed in the United States in 2015-2019. Neuro Oncol.24, v1–v95 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
2.Tan, A. C. et al. Management of glioblastoma: state of the art and future directions. CA Cancer J. Clin.70, 299–312 (2020). [DOI] [PubMed] [Google Scholar]
3.Hamed, A. A. et al. Gliomagenesis mimics an injury response orchestrated by neural crest-like cells. Nature, 10.1038/s41586-024-08356-2 (2025). [DOI] [PMC free article] [PubMed]
4.Erickson, A. E., Lan Levengood, S. K., Sun, J., Chang, F. C. & Zhang, M. Fabrication and characterization of chitosan-hyaluronic acid scaffolds with varying stiffness for glioblastoma cell culture. Adv. Health. Mater.7, e1800295 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
5.Chen, X. et al. A feedforward mechanism mediated by mechanosensitive ion channel PIEZO1 and tissue mechanics promotes glioma aggression. Neuron100, 799–815.e797 (2018). [DOI] [PubMed] [Google Scholar]
6.Miroshnikova, Y. A. et al. Tissue mechanics promote IDH1-dependent HIF1alpha-tenascin C feedback to regulate glioblastoma aggression. Nat. Cell Biol.18, 1336–1345 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
7.Bao, S. et al. Glioma stem cells promote radioresistance by preferential activation of the DNA damage response. Nature444, 756–760 (2006). [DOI] [PubMed] [Google Scholar]
8.Murat, A. et al. Stem cell-related “self-renewal” signature and high epidermal growth factor receptor expression associated with resistance to concomitant chemoradiotherapy in glioblastoma. J. Clin. Oncol.26, 3015–3024 (2008). [DOI] [PubMed] [Google Scholar]
9.Quail, D. F. et al. The tumor microenvironment underlies acquired resistance to CSF-1R inhibition in gliomas. Science352, aad3018 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
10.Zhou, B. B. et al. Tumour-initiating cells: challenges and opportunities for anticancer drug discovery. Nat. Rev. Drug Discov.8, 806–823 (2009). [DOI] [PubMed] [Google Scholar]
11.Safa, A. R., Saadatzadeh, M. R., Cohen-Gadol, A. A., Pollok, K. E. & Bijangi-Vishehsaraei, K. Glioblastoma stem cells (GSCs) epigenetic plasticity and interconversion between differentiated non-GSCs and GSCs. Genes Dis.2, 152–163 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
12.Vieira de Castro, J., Goncalves, C. S., Hormigo, A. & Costa, B. M. Exploiting the complexities of glioblastoma stem cells: insights for cancer initiation and therapeutic targeting. Int J Mol Sci21, 10.3390/ijms21155278 (2020). [DOI] [PMC free article] [PubMed]
13.Gimple, R. C., Yang, K., Halbert, M. E., Agnihotri, S. & Rich, J. N. Brain cancer stem cells: resilience through adaptive plasticity and hierarchical heterogeneity. Nat. Rev. Cancer22, 497–514 (2022). [DOI] [PubMed] [Google Scholar]
14.Nia, H. T., Munn, L. L. & Jain, R. K. Physical traits of cancer. Science370, 10.1126/science.aaz0868 (2020). [DOI] [PMC free article] [PubMed]
15.Stupp, R. et al. Radiotherapy plus concomitant and adjuvant temozolomide for glioblastoma. N. Engl. J. Med.352, 987–996 (2005). [DOI] [PubMed] [Google Scholar]
16.Obrador, E. et al. Glioblastoma therapy: past, present and future. Int J Mol Sci25, 10.3390/ijms25052529 (2024). [DOI] [PMC free article] [PubMed]
17.Deville, S. S. & Cordes, N. The extracellular, cellular, and nuclear stiffness, a trinity in the cancer resistome-a review. Front Oncol.9, 1376 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
18.Rizzuti, I. F. et al. Mechanical control of cell proliferation increases resistance to chemotherapeutic agents. Phys. Rev. Lett.125, 128103 (2020). [DOI] [PubMed] [Google Scholar]
19.Mathur, R. et al. Glioblastoma evolution and heterogeneity from a 3D whole-tumor perspective. Cell187, 446–463.e416 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
20.Sohrabi, A. et al. Microenvironmental stiffness induces metabolic reprogramming in glioblastoma. Cell Rep.42, 113175 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
21.Mahaffey, B. J. et al. The prognostic effect of mechanical, ultrastructural, and ECM signatures in glioblastoma core and rim. APL Bioeng.8, 036101 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
22.Marhuenda, E. et al. Glioma stem cells invasive phenotype at optimal stiffness is driven by MGAT5 dependent mechanosensing. J. Exp. Clin. Cancer Res.40, 139 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
23.Lathia, J. D., Mack, S. C., Mulkearns-Hubert, E. E., Valentim, C. L. & Rich, J. N. Cancer stem cells in glioblastoma. Genes Dev.29, 1203–1217 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
24.Gill, B. J. et al. MRI-localized biopsies reveal subtype-specific differences in molecular and cellular composition at the margins of glioblastoma. Proc. Natl. Acad. Sci. USA111, 12550–12555 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
25.Wu, X. et al. Threonine fuels glioblastoma through YRDC-mediated codon-biased translational reprogramming. Nat. Cancer5, 1024–1044 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
26.Richards, L. M. et al. Gradient of Developmental and Injury Response transcriptional states defines functional vulnerabilities underpinning glioblastoma heterogeneity. Nat. Cancer2, 157–173 (2021). [DOI] [PubMed] [Google Scholar]
27.Yuan, H. et al. Lysine catabolism reprograms tumour immunity through histone crotonylation. Nature617, 818–826 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
28.Suva, M. L. et al. Reconstructing and reprogramming the tumor-propagating potential of glioblastoma stem-like cells. Cell157, 580–594 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
29.Isomursu, A. et al. Directed cell migration towards softer environments. Nat. Mater.21, 1081–1090 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
30.Bastola, S. et al. Glioma-initiating cells at tumor edge gain signals from tumor core cells to promote their malignancy. Nat. Commun.11, 4660 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
31.Pollak, J. et al. Ion channel expression patterns in glioblastoma stem cells with functional and therapeutic implications for malignancy. PLoS One12, e0172884 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
32.Ciechomska, I. A., Gielniewski, B., Wojtas, B., Kaminska, B. & Mieczkowski, J. EGFR/FOXO3a/BIM signaling pathway determines chemosensitivity of BMP4-differentiated glioma stem cells to temozolomide. Exp. Mol. Med52, 1326–1340 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
33.Zhao, B. et al. TEAD mediates YAP-dependent gene induction and growth control. Gene Dev.22, 1962–1971 (2008). [DOI] [PMC free article] [PubMed] [Google Scholar]
34.Monteleon, C. L. et al. IQGAP1 and IQGAP3 serve individually essential roles in normal epidermal homeostasis and tumor progression. J. Invest Dermatol135, 2258–2265 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
35.Mack, S. C. et al. Chromatin landscapes reveal developmentally encoded transcriptional states that define human glioblastoma. J. Exp. Med216, 1071–1090 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
36.Abdelfattah, N. et al. Single-cell analysis of human glioma and immune cells identifies S100A4 as an immunotherapy target. Nat. Commun.13, 767 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
37.Neftel, C. et al. An integrative model of cellular states, plasticity, and genetics for glioblastoma. Cell178, 835–849.e821 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
38.Vlatkovic, T., Veldwijk, M. R., Giordano, F. A. & Herskind, C. Targeting Cell Cycle Checkpoint Kinases to Overcome Intrinsic Radioresistance in Brain Tumor Cells. Cancers14, 10.3390/cancers14030701 (2022). [DOI] [PMC free article] [PubMed]
39.Madhavan, K. et al. Venetoclax cooperates with ionizing radiation to attenuate diffuse midline glioma tumor growth. Clin. Cancer Res. 28, 2409–2424 (2022). [DOI] [PubMed] [Google Scholar]
40.Bhat, K. P. L. et al. Mesenchymal differentiation mediated by NF-kappaB promotes radiation resistance in glioblastoma. Cancer Cell24, 331–346 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
41.Zhao, J. et al. Luteolin attenuates cancer cell stemness in PTX-resistant oesophageal cancer cells through mediating SOX2 protein stability. Pharm. Res.174, 105939 (2021). [DOI] [PubMed] [Google Scholar]
42.Fang, L. et al. A methylation-phosphorylation switch determines Sox2 stability and function in ESC maintenance or differentiation. Mol. Cell55, 537–551 (2014). [DOI] [PubMed] [Google Scholar]
43.Cui, C. P. et al. Dynamic ubiquitylation of Sox2 regulates proteostasis and governs neural progenitor cell differentiation. Nat. Commun.9, 4648 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
44.Mahlokozera, T. et al. Competitive binding of E3 ligases TRIM26 and WWP2 controls SOX2 in glioblastoma. Nat. Commun.12, 6321 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
45.Liu, X. et al. Niche stiffness sustains cancer stemness via TAZ and NANOG phase separation. Nat. Commun.14, 238 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
46.Lv, J. et al. Cell softness regulates tumorigenicity and stemness of cancer cells. EMBO J.40, e106123 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
47.Liu, J. et al. Soft fibrin gels promote selection and growth of tumorigenic cells. Nat. Mater.11, 734–741 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
48.Jiang, Y. et al. Targeting extracellular matrix stiffness and mechanotransducers to improve cancer therapy. J. Hematol. Oncol.15, 34 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
49.Mohammadi, H. & Sahai, E. Mechanisms and impact of altered tumour mechanics. Nat. Cell Biol.20, 766–774 (2018). [DOI] [PubMed] [Google Scholar]
50.Palumbo, A., Jr. et al. Esophageal cancer development: crucial clues arising from the extracellular matrix. Cells9, 10.3390/cells9020455 (2020). [DOI] [PMC free article] [PubMed]
51.Palmieri, D. et al. Procollagen I COOH-terminal fragment induces VEGF-A and CXCR4 expression in breast carcinoma cells. Exp. Cell Res. 314, 2289–2298 (2008). [DOI] [PubMed] [Google Scholar]
52.Deng, M. et al. Extracellular matrix stiffness determines DNA repair efficiency and cellular sensitivity to genotoxic agents. Sci Adv6, 10.1126/sciadv.abb2630 (2020). [DOI] [PMC free article] [PubMed]
53.Song, F., Dai, Q., Grimm, M. O. & Steinbach, D. The Antithetic Roles of IQGAP2 and IQGAP3 in Cancers. Cancers15, 10.3390/cancers15041115 (2023). [DOI] [PMC free article] [PubMed]
54.Moroishi, T., Hansen, C. G. & Guan, K. L. The emerging roles of YAP and TAZ in cancer. Nat. Rev. Cancer15, 73–79 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
55.Mo, J. S., Park, H. W. & Guan, K. L. The Hippo signaling pathway in stem cell biology and cancer. EMBO Rep.15, 642–656 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
56.Lacombe, J. & Zenhausern, F. Effect of mechanical forces on cellular response to radiation. Radiother. Oncol.176, 187–198 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
57.Weina, K. & Utikal, J. SOX2 and cancer: current research and its implications in the clinic. Clin. Transl. Med3, 19 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
58.Huser, L., Novak, D., Umansky, V., Altevogt, P. & Utikal, J. Targeting SOX2 in anticancer therapy. Expert Opin. Ther. Targets22, 983–991 (2018). [DOI] [PubMed] [Google Scholar]
59.Ferri, A. L. et al. Sox2 deficiency causes neurodegeneration and impaired neurogenesis in the adult mouse brain. Development131, 3805–3819 (2004). [DOI] [PubMed] [Google Scholar]
60.Ellis, P. et al. SOX2, a persistent marker for multipotential neural stem cells derived from embryonic stem cells, the embryo or the adult. Dev. Neurosci.26, 148–165 (2004). [DOI] [PubMed] [Google Scholar]
61.Cavallaro, M. et al. Impaired generation of mature neurons by neural stem cells from hypomorphic Sox2 mutants. Development135, 541–557 (2008). [DOI] [PubMed] [Google Scholar]
62.Vilas, J. M. et al. Adult Sox2+ stem cell exhaustion in mice results in cellular senescence and premature aging. Aging Cell17, e12834 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
63.Guilak, F. et al. Control of stem cell fate by physical interactions with the extracellular matrix. Cell Stem Cell5, 17–26 (2009). [DOI] [PMC free article] [PubMed] [Google Scholar]
64.Segel, M. et al. Niche stiffness underlies the ageing of central nervous system progenitor cells. Nature573, 130–134 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
65.Bertino, J. R., Sawicki, W. L., Moroson, B. A., Cashmore, A. R. & Elslager, E. F. 2,4-diamino-5-methyl-6-[(3,4,5-trimethoxyanilino)methyl]quinazoline (tmq), a potent non-classical folate antagonist inhibitor-I effect on dihydrofolate reductase and growth of rodent tumors in vitro and in vivo. Biochem Pharm.28, 1983–1987 (1979). [DOI] [PubMed] [Google Scholar]
66.Jackson, R. C. et al. Biochemical pharmacology of the lipophilic antifolate, trimetrexate. Adv. Enzym. Regul.22, 187–206 (1984). [DOI] [PubMed] [Google Scholar]
67.Li, H. et al. Folylpolyglutamate synthetase inactivation in relapsed ALL induces a druggable folate metabolic vulnerability. Drug Resist Updat77, 101141 (2024). [DOI] [PubMed] [Google Scholar]
68.Liu, X. et al. PHGDH activation fuels glioblastoma progression and radioresistance via serine synthesis pathway. J. Exp. Clin. Cancer Res. 44, 99 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
69.Qiu, Z. et al. Transcription elongation machinery is a druggable dependency and potentiates immunotherapy in glioblastoma stem cells. Cancer Discov.12, 502–521 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
70.Jin, X. et al. Targeting glioma stem cells through combined BMI1 and EZH2 inhibition. Nat. Med23, 1352–1361 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
71.Tse, J. R. & Engler, A. J. Preparation of hydrogel substrates with tunable mechanical properties. Curr. Protoc. Cell Biol.Chapter 10, Unit 10 16 (2010). [DOI] [PubMed] [Google Scholar]
72.Barber-Perez, N. et al. Mechano-responsiveness of fibrillar adhesions on stiffness-gradient gels. J Cell Sci133, 10.1242/jcs.242909 (2020). [DOI] [PMC free article] [PubMed]
73.Miranda, A. et al. Cancer stemness, intratumoral heterogeneity, and immune response across cancers. Proc. Natl. Acad. Sci. USA116, 9020–9029 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
74.Palmer, N. P., Schmid, P. R., Berger, B. & Kohane, I. S. A gene expression profile of stem cell pluripotentiality and differentiation is conserved across diverse solid and hematopoietic cancers. Genome Biol.13, R71 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
75.Jafari, R. et al. The cellular thermal shift assay for evaluating drug target interactions in cells. Nat. Protoc.9, 2100–2122 (2014). [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Information^{(6MB, pdf)}

41467_2026_74058_MOESM2_ESM.pdf^{(75.1KB, pdf)}

Description of Additional Supplementary Files

Supplementary Data 1^{(13.6KB, xlsx)}

Supplementary Data 2^{(2MB, xlsx)}

Supplementary Data 3^{(1.7MB, xlsx)}

Supplementary Data 4^{(10.4KB, xlsx)}

Supplementary Data 5^{(2.8MB, xlsx)}

Supplementary Data 6^{(311.6KB, xlsx)}

Supplementary Data 7^{(1.1MB, xlsx)}

Supplementary Data 8^{(1.3MB, xlsx)}

Supplementary Data 9^{(128.1KB, xlsx)}

Supplementary Data 10^{(48.5KB, xlsx)}

Supplementary Data 11^{(254KB, xlsx)}

Supplementary Data 12^{(9.4KB, xlsx)}

Supplementary Data 13^{(17.5KB, xlsx)}

Supplementary Data 14^{(14.4KB, xlsx)}

Reporting Summary^{(143.4KB, pdf)}

Transparent Peer Review file^{(14.1MB, pdf)}

Source Data^{(13MB, zip)}

Data Availability Statement

No custom code was generated for this study.

[CR1] 1.Ostrom, Q. T. et al. CBTRUS Statistical Report: primary brain and other central nervous system tumors diagnosed in the United States in 2015-2019. Neuro Oncol.24, v1–v95 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR2] 2.Tan, A. C. et al. Management of glioblastoma: state of the art and future directions. CA Cancer J. Clin.70, 299–312 (2020). [DOI] [PubMed] [Google Scholar]

[CR3] 3.Hamed, A. A. et al. Gliomagenesis mimics an injury response orchestrated by neural crest-like cells. Nature, 10.1038/s41586-024-08356-2 (2025). [DOI] [PMC free article] [PubMed]

[CR4] 4.Erickson, A. E., Lan Levengood, S. K., Sun, J., Chang, F. C. & Zhang, M. Fabrication and characterization of chitosan-hyaluronic acid scaffolds with varying stiffness for glioblastoma cell culture. Adv. Health. Mater.7, e1800295 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR5] 5.Chen, X. et al. A feedforward mechanism mediated by mechanosensitive ion channel PIEZO1 and tissue mechanics promotes glioma aggression. Neuron100, 799–815.e797 (2018). [DOI] [PubMed] [Google Scholar]

[CR6] 6.Miroshnikova, Y. A. et al. Tissue mechanics promote IDH1-dependent HIF1alpha-tenascin C feedback to regulate glioblastoma aggression. Nat. Cell Biol.18, 1336–1345 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR7] 7.Bao, S. et al. Glioma stem cells promote radioresistance by preferential activation of the DNA damage response. Nature444, 756–760 (2006). [DOI] [PubMed] [Google Scholar]

[CR8] 8.Murat, A. et al. Stem cell-related “self-renewal” signature and high epidermal growth factor receptor expression associated with resistance to concomitant chemoradiotherapy in glioblastoma. J. Clin. Oncol.26, 3015–3024 (2008). [DOI] [PubMed] [Google Scholar]

[CR9] 9.Quail, D. F. et al. The tumor microenvironment underlies acquired resistance to CSF-1R inhibition in gliomas. Science352, aad3018 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR10] 10.Zhou, B. B. et al. Tumour-initiating cells: challenges and opportunities for anticancer drug discovery. Nat. Rev. Drug Discov.8, 806–823 (2009). [DOI] [PubMed] [Google Scholar]

[CR11] 11.Safa, A. R., Saadatzadeh, M. R., Cohen-Gadol, A. A., Pollok, K. E. & Bijangi-Vishehsaraei, K. Glioblastoma stem cells (GSCs) epigenetic plasticity and interconversion between differentiated non-GSCs and GSCs. Genes Dis.2, 152–163 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR12] 12.Vieira de Castro, J., Goncalves, C. S., Hormigo, A. & Costa, B. M. Exploiting the complexities of glioblastoma stem cells: insights for cancer initiation and therapeutic targeting. Int J Mol Sci21, 10.3390/ijms21155278 (2020). [DOI] [PMC free article] [PubMed]

[CR13] 13.Gimple, R. C., Yang, K., Halbert, M. E., Agnihotri, S. & Rich, J. N. Brain cancer stem cells: resilience through adaptive plasticity and hierarchical heterogeneity. Nat. Rev. Cancer22, 497–514 (2022). [DOI] [PubMed] [Google Scholar]

[CR14] 14.Nia, H. T., Munn, L. L. & Jain, R. K. Physical traits of cancer. Science370, 10.1126/science.aaz0868 (2020). [DOI] [PMC free article] [PubMed]

[CR15] 15.Stupp, R. et al. Radiotherapy plus concomitant and adjuvant temozolomide for glioblastoma. N. Engl. J. Med.352, 987–996 (2005). [DOI] [PubMed] [Google Scholar]

[CR16] 16.Obrador, E. et al. Glioblastoma therapy: past, present and future. Int J Mol Sci25, 10.3390/ijms25052529 (2024). [DOI] [PMC free article] [PubMed]

[CR17] 17.Deville, S. S. & Cordes, N. The extracellular, cellular, and nuclear stiffness, a trinity in the cancer resistome-a review. Front Oncol.9, 1376 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR18] 18.Rizzuti, I. F. et al. Mechanical control of cell proliferation increases resistance to chemotherapeutic agents. Phys. Rev. Lett.125, 128103 (2020). [DOI] [PubMed] [Google Scholar]

[CR19] 19.Mathur, R. et al. Glioblastoma evolution and heterogeneity from a 3D whole-tumor perspective. Cell187, 446–463.e416 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR20] 20.Sohrabi, A. et al. Microenvironmental stiffness induces metabolic reprogramming in glioblastoma. Cell Rep.42, 113175 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR21] 21.Mahaffey, B. J. et al. The prognostic effect of mechanical, ultrastructural, and ECM signatures in glioblastoma core and rim. APL Bioeng.8, 036101 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR22] 22.Marhuenda, E. et al. Glioma stem cells invasive phenotype at optimal stiffness is driven by MGAT5 dependent mechanosensing. J. Exp. Clin. Cancer Res.40, 139 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR23] 23.Lathia, J. D., Mack, S. C., Mulkearns-Hubert, E. E., Valentim, C. L. & Rich, J. N. Cancer stem cells in glioblastoma. Genes Dev.29, 1203–1217 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR24] 24.Gill, B. J. et al. MRI-localized biopsies reveal subtype-specific differences in molecular and cellular composition at the margins of glioblastoma. Proc. Natl. Acad. Sci. USA111, 12550–12555 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR25] 25.Wu, X. et al. Threonine fuels glioblastoma through YRDC-mediated codon-biased translational reprogramming. Nat. Cancer5, 1024–1044 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR26] 26.Richards, L. M. et al. Gradient of Developmental and Injury Response transcriptional states defines functional vulnerabilities underpinning glioblastoma heterogeneity. Nat. Cancer2, 157–173 (2021). [DOI] [PubMed] [Google Scholar]

[CR27] 27.Yuan, H. et al. Lysine catabolism reprograms tumour immunity through histone crotonylation. Nature617, 818–826 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR28] 28.Suva, M. L. et al. Reconstructing and reprogramming the tumor-propagating potential of glioblastoma stem-like cells. Cell157, 580–594 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR29] 29.Isomursu, A. et al. Directed cell migration towards softer environments. Nat. Mater.21, 1081–1090 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR30] 30.Bastola, S. et al. Glioma-initiating cells at tumor edge gain signals from tumor core cells to promote their malignancy. Nat. Commun.11, 4660 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR31] 31.Pollak, J. et al. Ion channel expression patterns in glioblastoma stem cells with functional and therapeutic implications for malignancy. PLoS One12, e0172884 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR32] 32.Ciechomska, I. A., Gielniewski, B., Wojtas, B., Kaminska, B. & Mieczkowski, J. EGFR/FOXO3a/BIM signaling pathway determines chemosensitivity of BMP4-differentiated glioma stem cells to temozolomide. Exp. Mol. Med52, 1326–1340 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR33] 33.Zhao, B. et al. TEAD mediates YAP-dependent gene induction and growth control. Gene Dev.22, 1962–1971 (2008). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR34] 34.Monteleon, C. L. et al. IQGAP1 and IQGAP3 serve individually essential roles in normal epidermal homeostasis and tumor progression. J. Invest Dermatol135, 2258–2265 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR35] 35.Mack, S. C. et al. Chromatin landscapes reveal developmentally encoded transcriptional states that define human glioblastoma. J. Exp. Med216, 1071–1090 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR36] 36.Abdelfattah, N. et al. Single-cell analysis of human glioma and immune cells identifies S100A4 as an immunotherapy target. Nat. Commun.13, 767 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR37] 37.Neftel, C. et al. An integrative model of cellular states, plasticity, and genetics for glioblastoma. Cell178, 835–849.e821 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR38] 38.Vlatkovic, T., Veldwijk, M. R., Giordano, F. A. & Herskind, C. Targeting Cell Cycle Checkpoint Kinases to Overcome Intrinsic Radioresistance in Brain Tumor Cells. Cancers14, 10.3390/cancers14030701 (2022). [DOI] [PMC free article] [PubMed]

[CR39] 39.Madhavan, K. et al. Venetoclax cooperates with ionizing radiation to attenuate diffuse midline glioma tumor growth. Clin. Cancer Res. 28, 2409–2424 (2022). [DOI] [PubMed] [Google Scholar]

[CR40] 40.Bhat, K. P. L. et al. Mesenchymal differentiation mediated by NF-kappaB promotes radiation resistance in glioblastoma. Cancer Cell24, 331–346 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR41] 41.Zhao, J. et al. Luteolin attenuates cancer cell stemness in PTX-resistant oesophageal cancer cells through mediating SOX2 protein stability. Pharm. Res.174, 105939 (2021). [DOI] [PubMed] [Google Scholar]

[CR42] 42.Fang, L. et al. A methylation-phosphorylation switch determines Sox2 stability and function in ESC maintenance or differentiation. Mol. Cell55, 537–551 (2014). [DOI] [PubMed] [Google Scholar]

[CR43] 43.Cui, C. P. et al. Dynamic ubiquitylation of Sox2 regulates proteostasis and governs neural progenitor cell differentiation. Nat. Commun.9, 4648 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR44] 44.Mahlokozera, T. et al. Competitive binding of E3 ligases TRIM26 and WWP2 controls SOX2 in glioblastoma. Nat. Commun.12, 6321 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR45] 45.Liu, X. et al. Niche stiffness sustains cancer stemness via TAZ and NANOG phase separation. Nat. Commun.14, 238 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR46] 46.Lv, J. et al. Cell softness regulates tumorigenicity and stemness of cancer cells. EMBO J.40, e106123 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR47] 47.Liu, J. et al. Soft fibrin gels promote selection and growth of tumorigenic cells. Nat. Mater.11, 734–741 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR48] 48.Jiang, Y. et al. Targeting extracellular matrix stiffness and mechanotransducers to improve cancer therapy. J. Hematol. Oncol.15, 34 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR49] 49.Mohammadi, H. & Sahai, E. Mechanisms and impact of altered tumour mechanics. Nat. Cell Biol.20, 766–774 (2018). [DOI] [PubMed] [Google Scholar]

[CR50] 50.Palumbo, A., Jr. et al. Esophageal cancer development: crucial clues arising from the extracellular matrix. Cells9, 10.3390/cells9020455 (2020). [DOI] [PMC free article] [PubMed]

[CR51] 51.Palmieri, D. et al. Procollagen I COOH-terminal fragment induces VEGF-A and CXCR4 expression in breast carcinoma cells. Exp. Cell Res. 314, 2289–2298 (2008). [DOI] [PubMed] [Google Scholar]

[CR52] 52.Deng, M. et al. Extracellular matrix stiffness determines DNA repair efficiency and cellular sensitivity to genotoxic agents. Sci Adv6, 10.1126/sciadv.abb2630 (2020). [DOI] [PMC free article] [PubMed]

[CR53] 53.Song, F., Dai, Q., Grimm, M. O. & Steinbach, D. The Antithetic Roles of IQGAP2 and IQGAP3 in Cancers. Cancers15, 10.3390/cancers15041115 (2023). [DOI] [PMC free article] [PubMed]

[CR54] 54.Moroishi, T., Hansen, C. G. & Guan, K. L. The emerging roles of YAP and TAZ in cancer. Nat. Rev. Cancer15, 73–79 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR55] 55.Mo, J. S., Park, H. W. & Guan, K. L. The Hippo signaling pathway in stem cell biology and cancer. EMBO Rep.15, 642–656 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR56] 56.Lacombe, J. & Zenhausern, F. Effect of mechanical forces on cellular response to radiation. Radiother. Oncol.176, 187–198 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR57] 57.Weina, K. & Utikal, J. SOX2 and cancer: current research and its implications in the clinic. Clin. Transl. Med3, 19 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR58] 58.Huser, L., Novak, D., Umansky, V., Altevogt, P. & Utikal, J. Targeting SOX2 in anticancer therapy. Expert Opin. Ther. Targets22, 983–991 (2018). [DOI] [PubMed] [Google Scholar]

[CR59] 59.Ferri, A. L. et al. Sox2 deficiency causes neurodegeneration and impaired neurogenesis in the adult mouse brain. Development131, 3805–3819 (2004). [DOI] [PubMed] [Google Scholar]

[CR60] 60.Ellis, P. et al. SOX2, a persistent marker for multipotential neural stem cells derived from embryonic stem cells, the embryo or the adult. Dev. Neurosci.26, 148–165 (2004). [DOI] [PubMed] [Google Scholar]

[CR61] 61.Cavallaro, M. et al. Impaired generation of mature neurons by neural stem cells from hypomorphic Sox2 mutants. Development135, 541–557 (2008). [DOI] [PubMed] [Google Scholar]

[CR62] 62.Vilas, J. M. et al. Adult Sox2+ stem cell exhaustion in mice results in cellular senescence and premature aging. Aging Cell17, e12834 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR63] 63.Guilak, F. et al. Control of stem cell fate by physical interactions with the extracellular matrix. Cell Stem Cell5, 17–26 (2009). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR64] 64.Segel, M. et al. Niche stiffness underlies the ageing of central nervous system progenitor cells. Nature573, 130–134 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR65] 65.Bertino, J. R., Sawicki, W. L., Moroson, B. A., Cashmore, A. R. & Elslager, E. F. 2,4-diamino-5-methyl-6-[(3,4,5-trimethoxyanilino)methyl]quinazoline (tmq), a potent non-classical folate antagonist inhibitor-I effect on dihydrofolate reductase and growth of rodent tumors in vitro and in vivo. Biochem Pharm.28, 1983–1987 (1979). [DOI] [PubMed] [Google Scholar]

[CR66] 66.Jackson, R. C. et al. Biochemical pharmacology of the lipophilic antifolate, trimetrexate. Adv. Enzym. Regul.22, 187–206 (1984). [DOI] [PubMed] [Google Scholar]

[CR67] 67.Li, H. et al. Folylpolyglutamate synthetase inactivation in relapsed ALL induces a druggable folate metabolic vulnerability. Drug Resist Updat77, 101141 (2024). [DOI] [PubMed] [Google Scholar]

[CR68] 68.Liu, X. et al. PHGDH activation fuels glioblastoma progression and radioresistance via serine synthesis pathway. J. Exp. Clin. Cancer Res. 44, 99 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR69] 69.Qiu, Z. et al. Transcription elongation machinery is a druggable dependency and potentiates immunotherapy in glioblastoma stem cells. Cancer Discov.12, 502–521 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR70] 70.Jin, X. et al. Targeting glioma stem cells through combined BMI1 and EZH2 inhibition. Nat. Med23, 1352–1361 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR71] 71.Tse, J. R. & Engler, A. J. Preparation of hydrogel substrates with tunable mechanical properties. Curr. Protoc. Cell Biol.Chapter 10, Unit 10 16 (2010). [DOI] [PubMed] [Google Scholar]

[CR72] 72.Barber-Perez, N. et al. Mechano-responsiveness of fibrillar adhesions on stiffness-gradient gels. J Cell Sci133, 10.1242/jcs.242909 (2020). [DOI] [PMC free article] [PubMed]

[CR73] 73.Miranda, A. et al. Cancer stemness, intratumoral heterogeneity, and immune response across cancers. Proc. Natl. Acad. Sci. USA116, 9020–9029 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR74] 74.Palmer, N. P., Schmid, P. R., Berger, B. & Kohane, I. S. A gene expression profile of stem cell pluripotentiality and differentiation is conserved across diverse solid and hematopoietic cancers. Genome Biol.13, R71 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR75] 75.Jafari, R. et al. The cellular thermal shift assay for evaluating drug target interactions in cells. Nat. Protoc.9, 2100–2122 (2014). [DOI] [PubMed] [Google Scholar]

PERMALINK

IQGAP3 bridges matrix stiffness with glioma stem cell maintenance and radioresistance by stabilizing SOX2

Po Zhang

Weichi Wu

Tengfei Huang

Xujia Wu

Donghai Wang

Huairui Yuan

Suchet Taori

Fanen Yuan

Frank P Vendetti

Rui Wang

Tingting Duan

Hailong Mi

Huan Li

Kailin Yang

Daqi Li

Ahmed Habib

Briana C Prager

Ryan C Gimple

Pascal O Zinn

Kalil Abdullah

Christopher Bakkenist

Junran Zhang

Qiulian Wu

Jeremy N Rich

Abstract

Introduction

Results

Matrix stiffness maintains GSCs

Fig. 1. Matrix stiffness maintains GSCs.

Matrix stiffness induces IQGAP3 in GSCs

Fig. 2. Matrix stiffness induces IQGAP3 in GSCs.

IQGAP3 maintains GSCs

Fig. 3. IQGAP3 maintains GSCs.

Matrix stiffness promotes radiation resistance in GSCs through IQGAP3

Fig. 4. Matrix Stiffness promotes radiation resistance in GSCs through IQGAP3.

Fig. 5. IQGAP3 regulates DNA damage repair in GSCs cultured on stiff matrigel.

IQGAP3 regulates radioresistance of GSCs by altering SOX2 protein levels

Fig. 6. IQGAP3 regulates radioresistance of GSCs by altering SOX2 protein levels.

IQGAP3 stabilizes SOX2 in the cytoplasm of GSCs

Fig. 7. IQGAP3 stabilizes SOX2 in the cytoplasm of GSCs.

Trimetrexate is an IQGAP3 inhibitor that augments radiotherapy responses against GBM

Fig. 8. Trimetrexate is an IQGAP3 inhibitor that augments radiotherapy responses against GSCs.

Fig. 9. Combined therapeutic efficacy of trimetrexate and radiotherapy against GSC xenografts.

Fig. 10. Fractionated radiotherapy combined with trimetrexate extends mouse survival.

Discussion

Methods

Human glioma and non-tumorous brain tissue

Glioblastoma stem cell derivation and cell culture

Xenografts

In vivo therapy

Conventional polyacrylamide hydrogels

Immunoblotting

Cytoplasm and nucleus fractionation

Immunofluorescence and immunohistochemistry staining

RNA extraction and quantitative real-time PCR

Neurosphere formation and proliferation assay

Click-iT EdU incorporation assay

Chromatin immunoprecipitation followed by PCR

Comet assay

ChIP-seq analysis

RNA sequencing and data analysis

Single-cell RNA sequencing reanalysis

Co-immunoprecipitation analysis

Mass Spectrometry analysis

DNA pull-down assay

Flow cytometry analysis

Virtual screening docking

Protein expression

Surface plasmon resonance (SPR) assay

In vivo detection of trimetrexate concentration

Cellular thermal shift assay (CETSA)

Public glioma patient datasets

Statistics and reproducibility

Materials availability

Reporting summary

Supplementary information

Source data

Acknowledgements