Abstract

   Glioblastoma (GBM) resistance to chemoradiotherapy is a major factor
   contributing to poor treatment outcomes. This resistance markedly
   affects the effectiveness of surgery combined with chemoradiotherapy
   and leads to post‐surgical tumor recurrence. Therefore, exploring the
   mechanisms underlying chemoradiotherapy resistance in GBM is crucial
   for understanding its progression and improving therapeutic options.
   This study found that moesin (MSN) acts as a key promotor of
   chemoradiotherapy resistance in glioma stem cells (GSCs), enhancing
   their proliferation and stemness maintenance. Mechanistically, MSN
   activates the downstream PI3K/mTOR signaling pathway, driving the
   proneural‐to‐mesenchymal transition (PMT) in GSCs. This process
   enhances the repair of DNA damage caused by radiotherapy (RT) and
   temozolomide (TMZ), thereby increasing the resistance of GSCs to
   chemoradiotherapy. Additionally, GNE‐317, a small molecule drug capable
   of crossing the blood‐brain barrier, specifically inhibits MSN and
   suppresses the activation of downstream PI3K/mTOR signaling.
   Importantly, the combination of GNE‐317 with RT and TMZ exhibits a
   strong synergistic effect both in vivo and in vitro, achieving better
   efficacy compared to the traditional combination of RT and TMZ. This
   study not only advances understanding of the mechanisms underlying
   chemoradiotherapy resistance in GBM but also provides a promising new
   approach for enhancing treatment outcomes.

   Keywords: chemoradiotherapy resistance, glioblastoma, GNE‐317, MSN,
   PI3K/mTOR, proneural‐to‐mesenchymal transition
     __________________________________________________________________

   The study identifies MSN as a key regulator of proneural‐to‐mesenchymal
   transition (PMT) in GSCs, driving resistance to radiotherapy (RT) and
   temozolomide (TMZ) in glioblastoma (GBM). However, GNE‐317 specifically
   inhibits MSN, reverses PMT, resensitizes cells to therapy, and shows
   promise as a targeted strategy to overcome chemoradiotherapy resistance
   in GBM.

   graphic file with name ADVS-12-2412517-g007.jpg

1. Introduction

   Glioblastoma (GBM) represents the predominant and most virulent primary
   malignancy affecting the central nervous system (CNS), ranking as a
   principal factor in patient deaths.^[ [54]^1 , [55]^2 ^] Despite
   maximal surgical resection followed by chemoradiotherapy for patients
   with newly diagnosed GBM, tumor recurrence is nearly inevitable and
   treatment options are limited.^[ [56]^3 , [57]^4 ^] The median overall
   survival for these individuals ranges below 15 months, while for
   recurrent glioblastoma, it is ≈2 to 9 months.^[ [58]^5 ^] The
   challenges in GBM treatment arise not only from the limitations of safe
   surgical resection margins but also from resistance to adjuvant
   chemoradiotherapy, including radiotherapy (RT) combined with
   temozolomide (TMZ).^[ [59]^6 ^] Therefore, understanding the mechanisms
   underlying chemoradiotherapy resistance is critical for improving
   current treatments and developing more effective therapeutic
   strategies.

   Chemoradiotherapy resistance in GBM is driven by several factors, with
   tumor heterogeneity and cellular plasticity playing key roles in
   treatment resistance and recurrence.^[ [60]^7 ^] Large‐scale genomic
   analyses of histologically similar GBM samples have identified three
   major molecular subtypes: proneural (PN), classical (CL), and
   mesenchymal (MES).^[ [61]^8 ^] Notably, the PN and MES subtypes differ
   markedly in prevalence, prognosis, and treatment response; the PN
   subtype is often linked to favorable outcomes and treatment sensitivity
   while the MES subtype is associated with poor prognosis and greater
   resistance to therapy.^[ [62]^9 ^] Additionally, GBMs contain
   populations of glioma stem cells (GSCs), which exhibit self‐renewal and
   multipotent differentiation capabilities^[ [63]^10 ^] and are key
   drivers of tumor proliferation, recurrence, and resistance to
   chemoradiotherapy.^[ [64]^11 ^] Furthermore, GSCs expressing markers of
   the PN and MES transcriptomic subtypes can be derived from GBMs.^[
   [65]^12 ^] However, GSCs expressing markers of the CL transcriptomic
   subtypes have not been convincingly identified.^[ [66]^13 ^] Similarly,
   GSCs derived from these GBM subtypes display corresponding
   characteristics, with MES GSCs being more resistant to treatment.^[
   [67]^14 , [68]^15 , [69]^16 ^] Therefore, GSCs are an ideal model for
   studying GBM therapies. Current research indicates that as GBM
   progresses and recurs, GSCs undergo proneural‐to‐mesenchymal transition
   (PMT), enhancing their resistance to therapeutic stressors.^[ [70]^17 ,
   [71]^18 , [72]^19 ^] As a result, the PMT process is considered to be a
   significant driver of chemoradiotherapy resistance and tumor
   recurrence.

   Several potential mechanisms have been implicated in the initiation of
   PMT. For example, one groundbreaking study identified CEBPB and STAT3
   as key synergistic regulators of mesenchymal transformation.^[ [73]^20
   ^] The PMT process also encompasses stimulation of both NF‐KB and TAZ
   signaling cascades.^[ [74]^9 , [75]^21 ^] Furthermore, interactions
   between tumor‐associated macrophages and GBM cells within the tumor
   microenvironment can promote PMT.^[ [76]^22 ^] However, despite
   extensive research, the molecular mechanisms underlying PMT in GSCs
   remain poorly understood. Targeting the key factors of the PMT process
   to enhance the efficacy of RT and TMZ represents a promising strategy
   for overcoming chemoradiotherapy resistance in GBM.

   In this study, we identified MSN as a key gene upregulated in patients
   with chemoradiotherapy‐resistant GBM. Moreover, we demonstrated that
   MSN not only promoted GSC proliferation and stemness but also increased
   resistance to RT and TMZ. Furthermore, MSN was closely associated with
   the MES subtype, and silencing MSN in GSCs reversed PMT by inhibiting
   the PI3K/mTOR pathway. Importantly, through a small‐molecule drug
   screen, we identified GNE‐317 as a PI3K/mTOR inhibitor capable of
   crossing the blood‐brain barrier (BBB), which not only effectively
   suppressed MSN expression but also exhibited a strong synergistic
   effect with RT and TMZ, significantly enhancing the efficacy of these
   combined therapies.^[ [77]^23 ^]

2. Results

2.1. MSN is Identified as a Key Factor Causing Resistance to
Chemoradiotherapy in GBM

   Standard chemoradiotherapy offers limited improvement in GBM patient
   prognosis, with its overall effectiveness remaining suboptimal (Figure
   [78]1A). To quantify the impact of chemoradiotherapy on GBM, we used
   data from patients in The Cancer Genome Atlas (TCGA) GBM database who
   had received standard chemoradiotherapy and had an outcome of death.
   These patients were divided into chemoradiotherapy‐resistant and
   chemoradiotherapy‐sensitive groups based on their survival time
   (Figure [79]1B). Differential analysis between the resistant and
   sensitive groups revealed a set of genes associated with
   chemoradiotherapy resistance (Figure [80]1C). Next, we treated GBM
   cells in vitro with RT and TMZ for 48 h and classified them into
   chemoradiotherapy‐resistant and ‐sensitive cells based on their
   apoptosis rates (Figure [81]1D; Figure [82]S1, Supporting Information).
   Differential analysis of the two groups identified a gene signature (n
   = 189) linked to chemoradiotherapy resistance in GBM cells
   (Figure [83]1E). By overlapping the two gene sets, we identified MSN as
   a gene associated with chemoradiotherapy resistance in patients with
   GBM and in GBM cells (Figure [84]1F). Additionally, in public
   databases, MSN also played a significant role in gliomas, particularly
   in high‐grade gliomas such as GBM (Figure [85]1G,H).

Figure 1.

   Figure 1
   [86]Open in a new tab

   MSN may be a key factor in chemoradiotherapy resistance in GBM. A)
   Kaplan‐Meier survival curves showing the association between standard
   chemoradiotherapy and prognosis in GBM patients in the TCGA GBM
   database. (Other therapy, n = 137; Only radiotherapy, n = 53; Only
   chemotherapy, n = 10; Chemoradiotherapy, n = 325). P values were
   calculated using the log‐rank test. ^** p < 0.01, ^*** p < 0.001. B)
   Scatter plot of patients who received standard chemoradiotherapy and
   had an outcome of death (n = 266), divided into sensitivity (n = 76)
   and resistance (n = 172) groups based on survival time. Data are shown
   as mean ± SEM. C) Volcano plot comparing resistant and sensitive
   patients. Red indicates chemoradiotherapy‐resistant genes, and yellow
   indicates chemoradiotherapy‐sensitive genes. The x‐axis represents
   log2(fold change), and the y‐axis represents ‐log10(p value). D)
   Quantification of apoptosis levels in one normal cell line and six GBM
   cell lines after control, RT, or TMZ treatment. Data are shown as mean
   ± SD. ^* p < 0.05, ^*** p < 0.001, ns, p > 0.05; two‐tailed unpaired
   t‐test. E) Volcano plot showing differential genes between resistant
   and sensitive cells. Red indicates chemoradiotherapy‐resistant genes,
   and yellow indicates chemoradiotherapy‐sensitive genes. F) Venn diagram
   showing chemoradiotherapy‐resistant genes common to both GBM cells and
   GBM patients. G) MSN expression in different grades and subtypes, and
   its association with survival outcomes in the glioma database (TCGA
   GBMLGG, CGGA, and Rembrandt). Bar chart data are shown as mean ± SD. ^*
   p < 0.05, ^*** p < 0.001, ns, p > 0.05; two‐tailed unpaired t‐test. For
   survival analysis, p values were calculated using the log‐rank test. H)
   MSN expression and prognosis in GBM patients with different IDH1 and
   MGMT status. Bar and box chart data are shown as mean ± SD. ^* p
   < 0.05, ^*** p < 0.001; two‐tailed unpaired t‐test. For survival
   analysis, p values were calculated using the log‐rank test.

   To further validate our findings, MSN expression was assessed using
   glioma tissue microarrays (TMAs) (Figure [87]2A). The findings
   demonstrated that MSN exhibited elevated expression in GBM tissues and
   correlated with unfavorable outcomes (Figure [88]2B,C). We then
   categorized patients with GBM who had received standard
   chemoradiotherapy and had an outcome of death into
   chemoradiotherapy‐resistant and ‐sensitive groups based on the median
   survival time (Figure [89]2D,E). We found that MSN was highly expressed
   in the chemoradiotherapy‐resistant group and that these patients had a
   poorer prognosis than those in the chemoradiotherapy‐sensitive group
   (Figure [90]2F,G). These results suggest that MSN promotes
   chemoradiotherapy resistance in patients with GBM. We also investigated
   the role of MSN using GSC models. First, we selected GSC924 and GSC628
   cells, which have high MSN expression, as positive cells for our
   experiments (Figure [91]2H). Subsequently, MSN was knocked down in
   GSC924 and GSC628 cells (Figure [92]2I,J). After MSN knockdown, half
   maximal inhibitory concentration (IC50) values for RT and TMZ treatment
   were significantly reduced, further inhibiting GSC growth
   (Figure [93]2K; Figure [94]S2A, Supporting Information). Furthermore,
   in the MSN knockdown group, RT and TMZ treatment significantly reduced
   the DNA damage repair capacity of GSCs, resulting in notable apoptosis
   (Figure [95]2L–O; Figure [96]S2B–E, Supporting Information). All
   experimental results suggest that MSN enhances chemoradiotherapy
   resistance in both GBMs and GSCs.

Figure 2.

   Figure 2
   [97]Open in a new tab

   MSN enhances chemoradiotherapy resistance in GBM patients and GSCs. A)
   Grade information and HE staining of glioma TMAs. Black circles
   indicate TMA spots excluded due to hemorrhage. B) Representative images
   and quantification of MSN expression in TMAs, grouped by glioma grades.
   Data are shown as mean ± SEM. ^*** p < 0.001; two‐tailed unpaired
   t‐test. C) Kaplan‐Meier survival curves for patients stratified by
   median MSN expression levels in grade IV glioma (GBM). Censored data
   are represented by vertical lines. P values were calculated using the
   log‐rank test. D) Bar chart showing the survival status of GBM patients
   after standard chemoradiotherapy. E) Scatter plot showing patients who
   received standard chemoradiotherapy and had an outcome of death,
   divided into chemoradiotherapy‐sensitive and
   chemoradiotherapy‐resistant groups based on median survival time. F)
   Scatter plot showing MSN expression levels in
   chemoradiotherapy‐sensitive and chemoradiotherapy‐resistant groups.
   Data are shown as mean ± SEM. ^* p < 0.05; two‐tailed unpaired t‐test.
   G) Kaplan‐Meier survival curves for chemoradiotherapy‐sensitive and
   chemoradiotherapy‐resistant groups. P values were calculated using the
   log‐rank test. H) Bar graph showing relative mRNA expression levels of
   MSN in normal tissues and GSCs by qPCR. Data are shown as mean ± SEM. n
   = 3 independent experiments. I) qPCR and J) immunoblot analysis of MSN
   expression in GSCs transduced with non‐targeting shRNA (shNT) or MSN
   shRNA (shMSN). qPCR data are shown as mean ± SEM. n = 3 independent
   experiments. ^*** p < 0.001; two‐tailed unpaired t‐test. K) Cell
   viability assay of GSC924 transduced with shNT or shMSN, treated with
   different doses of RT or varying concentrations of TMZ for 48 h. n = 3
   biological independent samples. The black dashed line represents the
   IC50 value. ^* p < 0.02, ^** p < 0.01, ^*** p < 0.001; two‐way ANOVA
   followed by Tukey's multiple comparison test. L) Representative images
   and M) quantification of γ‐H2AX staining in GSC924 with or without RT
   treatment (6 Gy, 48 h) or TMZ (200 µm, 48 h). Scale bar: 10 µm. Data
   are shown as mean ± SEM. n = 5 independent experiments. ^*** p < 0.001,
   ns, p > 0.05; two‐tailed unpaired t‐test. N) Immunoblot analysis of
   γ‐H2AX and cleaved Caspase‐3 expression in GSC924 with or without RT (6
   Gy, 48 h) or TMZ (200 µm, 48 h) treatment. O) Relative mRNA levels of
   DNA repair‐related genes in GSC924 with or without RT (6 Gy, 48 h) or
   TMZ (200 µm, 48 h) treatment. Data are shown as mean ± SEM. n = 3
   independent experiments. ^* p < 0.05, ^** p < 0.01, ^*** p < 0.001;
   two‐tailed unpaired t‐test.

2.2. MSN Supports the Growth and Maintenance of Stemness in Glioma Stem Cells

   Given MSN's role in chemoradiotherapy resistance, we explored its
   effects on GSC‐driven tumor progression both in vitro and in vivo. In
   vitro, silencing MSN in GSC924 and GSC628 inhibited proliferation,
   reduced cell viability, and impaired self‐renewal capacity, as
   demonstrated by diminished neurosphere formation and results from
   limiting dilution assays (Figure [98]3A–E). In contrast, MSN
   overexpression enhanced both the growth and stemness of GSC727‐1, which
   are typically negative cells with low MSN expression (Figure [99]3F–I).
   Given MSN's essential function in GSCs, potentially facilitating
   aggressive expansion, we investigated the consequences of MSN
   disruption on GSC‐driven tumor growth in vivo. GSCs (GSC924 and GSC628)
   expressing firefly luciferase, combined with either MSN shRNA or
   control shRNA (shNT), were transplanted into the brains of
   immunocompromised mice. Bioluminescence imaging revealed that MSN
   knockdown significantly inhibited GSC‐driven tumor growth
   (Figure [100]3J–K; Figures [101]S3A,B, Supporting Information).
   Consequently, mice with xenografts from GSCs expressing shMSN exhibited
   significantly longer survival than did control mice (Figure [102]3L,
   Figure [103]S3C, Supporting Information). Together, these findings
   highlight that MSN is indispensable for GSC proliferation and
   maintenance of stemness, both in vitro and in vivo, driving tumor
   growth in GBM.

Figure 3.

   Figure 3
   [104]Open in a new tab

   MSN is required for GSC proliferation and self‐renewal. A) Cell
   viability assay of GSCs transduced with shNT or shMSN. n = 6 (GSC924)
   or n = 6 (GSC628) biological independent samples. Data are shown as
   mean ± SD. ^*** p < 0.001, two‐way ANOVA followed by Tukey's multiple
   comparison test. B) Two independent shRNAs targeting MSN reduced the
   growth of GSC924 and GSC628 compared to shNT, as measured by cell
   count. Data are shown as mean ± SD from 5 independent experiments. ^* p
   < 0.05, ^** p < 0.01, ^*** p < 0.001, ns, p > 0.05; two‐way ANOVA
   followed by Tukey's multiple comparison test. C) Representative images
   of tumorspheres from GSCs transduced with shNT or shMSN. Scale bar:
   150 µm. D) Quantification of tumorsphere numbers and average clone
   diameter in GSCs transduced with shNT or shMSN. n = 16 for tumorsphere
   number, n = 5 biological independent cell cultures for average clone
   diameter. Data are shown as mean ± SEM. ^*** p < 0.001; two‐tailed
   unpaired t‐test. E) In vitro extreme limiting dilution analysis of
   tumorsphere formation in GSCs expressing shNT or shMSN. n = 16
   biological independent cell cultures. ^*** p < 0.001 by ELDA analysis.
   F) Immunoblot analysis of MSN expression in GSC727‐1‐Vector and
   GSC727‐1‐MSN‐overexpression (MSN‐OE) through lentiviral infection. G)
   Cell viability assay, H) representative images of tumorspheres, and I)
   extreme limiting dilution analysis of tumorsphere formation in GSC727‐1
   expressing vector or MSN‐OE. J) Representative images and K)
   quantification of relative luciferase on days 7, 14, 21, and 28
   post‐transplantation; bioluminescence is measured in p^−1s^−1cm^2‐1sr.
   GSC924: shNT (n = 5), shMSN‐1 (n = 5), shMSN‐2 (n = 5). Data are shown
   as mean ± SEM. ^* p < 0.05, ^*** p < 0.001; one‐way ANOVA followed by
   Tukey's multiple comparison test. L) Kaplan‐Meier survival curves of
   mice bearing GSC924‐derived xenografts expressing shNT or shMSN. ^*** p
   < 0.001, log‐rank test. GSC924: shNT (n = 10), shMSN‐1 (n = 10),
   shMSN‐2 (n = 10).

2.3. High‐MSN Chemoradiotherapy Resistant GBM Subclusters are Localized to
MES‐Like Regions

   Given the heterogeneity of GBM and the varying degrees of
   chemoradiotherapy resistance, we employed single‐cell sequencing to
   characterize resistant GBM cell clusters. To more precisely identify
   the gene expression patterns of different tumor subclusters, we need a
   tool that can provide full‐length transcript information and detect
   low‐abundance gene expression at single‐cell resolution. Therefore, we
   used Smart‐seq2 single‐cell sequencing data from 73 regions across 14
   patients (Figure [105]4A; Figure [106]S4A,B, Supporting Information).
   Following data quality control and the use of predicted copy number
   variation (CNV) and marker gene expression, we constructed a Gliomap
   with information on four factors: grade, patient, non‐tumor, and glioma
   (Figure [107]4B–F; Figure [108]S5A–D, Supporting Information). Among
   glioma cells, 90.3% were from WHO grade IV patients, 0.7% were from
   gliosarcoma (GS), and 9% were from WHO grade II patients
   (Figure [109]4G). We found that MSN was highly expressed in high‐grade
   glioma cells (both GBM and GS) (Figure [110]4H–I). Therefore, we
   divided the glioma clusters into high and low groups based on MSN
   expression levels (Figure [111]4J). Next, using the GBM
   chemoradiotherapy resistance signature (n = 189) defined in
   Figure [112]1E, we performed gene set enrichment analysis across 13
   glioma clusters. Based on the results of this enrichment, we classified
   the glioma cells into chemoradiotherapy‐resistant and
   chemoradiotherapy‐sensitive groups (Figure [113]4K). As expected, the
   high MSN group showed a significant overlap with the
   chemoradiotherapy‐resistant group (Figure [114]4L). Thus, we were able
   to identify high MSN/chemoradiotherapy‐resistant and low
   MSN/chemoradiotherapy‐sensitive glioma subclusters (Figure [115]4M).
   Moreover, compared to the low MSN‐sensitive subclusters, the cells in
   the high MSN‐resistant subclusters were almost exclusively derived from
   GBM (Figure [116]4N). Intriguingly, the high MSN‐resistant subclusters
   are characterized by enhanced survival signaling, metabolic
   reprogramming (including glycolysis and fatty acid metabolism), and
   resistance to apoptosis. These features are driven by key pathways such
   as PI3K/AKT/mTOR, hypoxia/HIF‐1, NF‐kB, and TGF‐β/Smad, collectively
   highlighting their pivotal role in GBM chemoradiotherapy resistance
   (Figure [117]S6A–D, Supporting Information).

Figure 4.

   Figure 4
   [118]Open in a new tab

   At the single‐cell level, MSN is highly correlated with
   chemoradiotherapy resistance. A) Basic information about single‐cell
   sequencing. B) Cell filtering excludes non‐single cells using
   DoubletFinder. C) t‐SNE analysis of all cells, showing 19 significant
   cell clusters color‐coded and labeled accordingly. D) RNA‐derived
   single‐cell CNV information. Cell clusters were classified as CNV^− and
   CNV^+ based on CNV score. E) Dot plot of cell clusters, showing shared
   and specific marker genes, dividing clusters into non‐tumor 1,
   non‐tumor 2, glioma, and lung cancer metastasis (LC‐meta) cells. F)
   t‐SNE visualization of the Gliomap, including patient data, grades (WHO
   II, Gliosarcoma (GS), and WHO IV), and subdivisions of non‐tumor and
   glioma subclusters. The axis outside the circular plot displays the log
   scale of total cell numbers for each cell type (level‐3 annotation). G)
   Pie chart showing the proportion of glioma cells originating from
   patients of different grades. H) Distribution of MSN expression on the
   original t‐SNE coordinates for 14 glioma subclusters. I) Scatter plot
   showing MSN expression levels in patients with low (WHO II) and high
   grades (GS and WHO IV). The black line represents the mean value. ^***
   p < 0.001, two‐tailed unpaired t‐test. J) Bar chart showing MSN
   expression across 14 glioma subclusters, categorized as high MSN (mean
   + SEM) and low MSN (mean – SEM) groups. K) Scatter plot showing
   enrichment scores of the chemoradiotherapy‐resistance signature,
   categorizing glioma subclusters into chemoradiotherapy‐resistant (mean
   + SEM) and chemoradiotherapy‐sensitive (mean – SEM) groups. L) Venn
   diagram showing high‐MSN chemoradiotherapy‐resistant and low‐MSN
   chemoradiotherapy‐sensitive glioma subclusters. M) Distribution of
   high‐MSN‐resistance and low‐MSN‐sensitivity groups on the original
   t‐SNE coordinates. N) Pie charts showing the proportions of cells from
   low‐MSN chemoradiotherapy‐sensitive and high‐MSN
   chemoradiotherapy‐resistant glioma subclusters in glioma patients of
   different grades.

   The spatial distribution of cells within tumor tissue also determines
   the functions they exert. Therefore, we integrated single‐cell
   sequencing with spatial transcriptomics of GBM and mapped the high
   MSN/resistant GBM subclusters to spatial locations within the tumor
   (Figure [119]5A,B). Using this approach, we identified regions
   potentially associated with chemoradiotherapy resistance
   (Figure [120]5C). To explore the anatomical locations of these regions,
   we performed a joint analysis using the Ivy Glioblastoma Atlas Project
   (Ivy GAP) database.^[ [121]^24 ^] The results showed that, compared to
   the sensitive regions, the resistant regions were primarily located in
   the microvascular proliferation (MP) and pseudopalisading cells (PSEU)
   regions (Figure [122]5D). These regions not only exhibit intense
   hypoxia, increased angiogenesis, and abnormal blood vessel networks,
   but according to previous studies, they are also regions where MES GSCs
   are frequently enriched.^[ [123]^25 , [124]^26 , [125]^27 ^] This
   observation was further validated using our spatial transcriptomic data
   (Figure [126]5E,F). The above results indicate a significant
   association between MSN and the GBM MES subtype, suggesting that MSN
   may promote chemoradiotherapy resistance by maintaining the
   characteristics of the MES subtype.

Figure 5.

   Figure 5
   [127]Open in a new tab

   Combined analysis indicates that high‐MSN chemoradiotherapy‐resistant
   cell subclusters are predominantly distributed in the MES‐like cells
   region. A) Schematic of integrated analysis combining glioma
   single‐cell sequencing and GBM spatial transcriptomics. B) The heatmap
   shows the matching between scRNA clusters and spatial transcriptomics
   data. In the scRNA clusters, red clusters indicate high‐MSN
   chemoradiotherapy‐resistant groups, while blue clusters represent
   low‐MSN chemoradiotherapy‐sensitive groups. C) Spatial distribution of
   high‐MSN chemoradiotherapy‐resistant and low‐MSN
   chemoradiotherapy‐sensitive subclusters. D) The bar chart represents
   the enrichment scores of high‐MSN chemoradiotherapy‐resistant and
   low‐MSN chemoradiotherapy‐sensitive groups in different regions of the
   Ivy GAP database. CT: Cellular Tµmor; IT: Infiltrating Tµmor; LE:
   Leading Edge; MP: Microvascular Proliferation; PSEU: Pseudopalisading
   cells. Data are presented as mean ± SD. Spatial distribution of E) 4
   cell states scores defined by scRNA‐seq and F) MES marker.

2.4. MSN Triggers the PMT, Enhancing GBM Resistance to Chemoradiotherapy

   We further investigated the role of MSN in MES‐subtype GBM. First,
   using the TCGA GBM database, we observed that MSN was highly expressed
   in the MES‐subtype GBM, and minimally expressed in the PN‐subtype
   (Figure [128]6A). Moreover, the high MSN group was significantly
   enriched in the VERHAAK_MES_2017 pathway, whereas the low MSN group was
   enriched in the VERHAAK_PN_2017 pathway (Figure [129]6B,C).^[ [130]^8
   ^] Additionally, in patients with GBM, correlation analysis revealed
   that MSN expression was positively correlated with MES pathway‐related
   and marker genes (CD44, YKL40, TIMP1, and TGFB1), whereas it was
   significantly negatively correlated with PN pathway‐related and marker
   genes (OLIG2, DLL3, SOX2, ASCL1) (Figure [131]6D).^[ [132]^28 ,
   [133]^29 , [134]^30 ^] In summary, these results indicate that MSN
   contributes to the promotion of PMT. We then examined the expression of
   MSN, MES, and PN markers in GSCs. Consistent with previous
   observations, the results showed that MSN is highly expressed in MES
   GSCs (Figure [135]6E). To further explore whether MSN influences PMT,
   we designed a CD44 (MES marker) reporter system to monitor changes in
   the state of MES GSCs (Figure [136]6F). The results showed that MSN
   knockdown in MES GSCs led to a decrease in MES marker gene expression,
   but an increase in the expression of PN marker genes (Figure [137]6G,H;
   Figure [138]S7A,B, Supporting Information). Additionally, the key
   transcriptional regulators of the MES subtype, such as, CEBPB, NFKB,
   STAT3, and YAP/TAZ signaling, were also weakened (Figure [139]6I).
   These data suggest that MSN may promote the PMT process by regulating
   key transcriptional regulators of the MES subtype.

Figure 6.

   Figure 6
   [140]Open in a new tab

   MSN promotes the proneural‐to‐mesenchymal transition to resist
   chemoradiotherapy. A) The scatter plot shows MSN expression levels
   across different GBM subtypes in the TCGA GBM database. Data are
   presented as mean ± SD. ^*** p < 0.001, two‐tailed unpaired t‐test. B)
   The scatter plot divides patients into high‐MSN (Mean + SEM) and
   low‐MSN (Mean – SEM) groups based on MSN expression levels. C) GSEA
   enrichment analysis of high‐MSN and low‐MSN groups in the TCGA GBM
   database. D) The left heatmap shows the correlation analysis between
   MSN and subtype signature genes, while the right pie chart represents
   the correlation analysis between MSN and subtype marker genes. E) qPCR
   analysis is used to determine the subtypes of different GSCs. F)
   Schematic of the CD44 reporter system. G) Representative images and
   quantitative analysis of CD44 reporter system luciferase in GSCs
   transduced with shNT or shMSN. n = 5 (GSC924) or n = 5 (GSC628)
   biological independent samples. Data are presented as mean ± SEM. ^***
   p < 0.001, two‐tailed unpaired t‐test. H) Immunoblot analysis of
   subtype markers in GSCs transduced with shNT or shMSN. I) Immunoblot
   analysis of key MES transcriptional regulators and pathways in GSCs
   transduced with shNT or shMSN. J) Multiplex immunofluorescence staining
   of TMAs in GBM, dividing GBM patients into MES or PN subtypes. The left
   side shows the immunohistochemical staining of MSN. K) The bar chart
   shows the MSN H‐score in MES and PN subtype patients. Data are
   presented as mean ± SEM. ^** p < 0.01, two‐tailed unpaired t‐test. L)
   Kaplan–Meier curves of survival for different subtypes of patients.
   Censored data is represented by vertical lines in the graph. P value is
   calculated using the log‐rank test. M) The bar chart represents the
   number of cases from patients who received standard chemoradiotherapy
   and had an outcome of death in MES and PN subtypes. N) Kaplan–Meier
   curves of survival for patients selected in M), stratified by subtype
   in GBM TMAs. P value is calculated using the log‐rank test.

   Subsequently, we performed a multiplex immunofluorescence analysis of
   TMAs from patients with GBM. After excluding low‐quality spots, we
   classified the patients into MES and PN subtypes based on the
   expression of MES and PN marker genes (Figure [141]S7C, Supporting
   Information). Combined with an immunohistochemical analysis of MSN
   expression, we observed that MSN was highly expressed in the MES
   subtype of GBM and was associated with poorer prognosis in these
   patients (Figure [142]6J–L). Finally, we conducted a survival analysis
   of patients with MES or PN subtype GBM who received standard
   chemoradiotherapy, and for whom the outcome was death. The results
   showed that patients with the MES subtype and high MSN expression had
   poorer survival outcomes and were more resistant to chemoradiotherapy
   than patients with the PN subtype and low MSN expression
   (Figure [143]6M,N). Collectively, these in vitro and clinical findings
   strongly suggest that MSN enhances chemoradiotherapy resistance in GBM
   by promoting PMT, highlighting MSN as a potential therapeutic target
   for MES subtype GBM.

2.5. GNE‐317 Enhances Sensitivity to RT and TMZ in GSCs by Reversing PMT via
MSN Inhibition

   Recent advancements in computer‐aided drug design have accelerated drug
   discovery, reducing both time and cost.^[ [144]^31 ^] Using
   high‐throughput screening methods, we identified 13 small‐molecule
   drugs that targeted MSN‐expressing cells in GBM (Figure [145]7A,B).
   However, among these small‐molecule drugs, only GNE‐317 has been
   validated to cross the BBB (Figure [146]7C).^[ [147]^32 ^] Analysis of
   the TCGA GBM database suggests that MSN may influence chemoradiotherapy
   resistance via the PI3K/mTOR signaling pathway (Figure [148]S8A–E,
   Supporting Information). Interestingly, GNE‐317 is also a specific
   inhibitor of the PI3K/mTOR pathway. Importantly, the inhibition of MSN
   in GSCs by GNE‐317 occurs primarily at the protein level, with minimal
   effects observed at the RNA level (GNE‐317: 5 µM) (Figure [149]7D,E).
   Moesin protein is a member of the ERM (Ezrin‐Radixin‐Moesin) protein
   family. Its structure is composed of three main parts: an N‐terminal
   FERM domain, a C‐terminal domain, and an α‐helical region connecting
   the two.^[ [150]^33 , [151]^34 ^] In its inactive state, moesin adopts
   a “closed” conformation, where the FERM domain and the C‐terminal
   domain interact via a head‐to‐tail interaction. When PIP[2] binds to
   the FERM domain, it disrupts the “closed” conformation, exposing the
   C‐terminal domain. This exposure allows phosphorylation of a threonine
   residue in the C‐terminal domain. These two events together lead to the
   activation of moesin, enabling it to perform its functions.^[ [152]^35
   , [153]^36 , [154]^37 ^] Furthermore, molecular docking analysis
   revealed that the GNE‐317 directly binds to the FERM domain of moesin
   protein (aa:1‐296) (Figure [155]7F). Therefore, leveraging the amino
   (‐NH2) group in the molecular structure of GNE‐317, we constructed
   biotin‐GNE‐317 beads (Figure [156]S8F, Supporting Information). In
   subsequent pull‐down assays, we found that the biotin‐GNE‐317 beads
   could bind to the moesin N‐terminal recombinant protein (aa:1–346), and
   this binding was competitively inhibited by the prior addition of free
   GNE‐317 (Figure [157]7G). The above results demonstrate that GNE‐317
   can directly bind to the FERM domain of the moesin protein. Based on
   the above results, we speculate that when GNE‐317 replaces PIP[2] in
   binding to the FERM domain of the moesin protein, moesin may become an
   open but inactive abnormal protein. This abnormal protein might be
   recognized by the body and degraded (Figure [158]7H). Furthermore, we
   observed that polyubiquitinated moesin decreased following treatment
   with GNE‐317 but increased significantly upon the addition of the
   proteasome inhibitor lactacystin (Figure [159]7I). This suggests that
   the abnormal moesin protein bound to GNE‐317 is degraded through
   ubiquitin‐mediated proteolysis. Therefore, we selected GNE‐317 as the
   drug to target MSN in GBM. We subsequently found that MSN
   overexpression in PN GSCs (GSC727‐1) promoted PMT and enhanced DNA
   damage repair following RT and TMZ treatment, thereby driving
   chemoradiotherapy resistance (Figure [160]7J–M). Furthermore, MSN was
   found to independently induce AKT and mTOR phosphorylation, activating
   the PI3K/mTOR pathway without reliance on GNE‐317 (Figure [161]7N–Q).
   However, treatment with GNE‐317 reversed these effects in GSC727‐1
   (Figure [162]7J–O) and suppressed the activation of the key
   transcriptional regulators of the MES subtype driven by MSN
   overexpression (Figure [163]7R). Finally, we found that solely
   inhibiting the PI3K or mTOR signaling pathway could partially suppress
   the PMT process and weaken DNA damage repair. However, its
   effectiveness is limited compared to dual inhibitors like GNE‐317
   (Figures S‐T).

Figure 7.

   Figure 7
   [164]Open in a new tab

   GNE‐317 enhances sensitivity to RT and TMZ by inhibiting MSN in GSCs.
   A) The scatter plot shows 33 glioma cells divided into high‐MSN (Mean +
   SEM) and low‐MSN (Mean – SEM) groups based on MSN expression levels. B)
   The volcano plot shows differential IC50 values of 141 drugs between
   high‐MSN and low‐MSN group cells. Red dots represent drugs sensitive to
   MSN. The x‐axis and y‐axis represent log[2] (fold change) and –log[10]
   (p value), respectively. C) The heatmap shows the relative IC50 value
   of 13 drugs sensitive to MSN in high MSN and low MSN groups. Red drug
   indicates the ability to cross the blood‐brain barrier. D) Immunoblot
   and E) qPCR analyses of the inhibitory effect of different
   concentrations of GNE‐317 on MSN at protein and RNA level in GSCs. F)
   Molecular docking analysis (3D Binding model) of Moesin protein
   (PDB:8CIU) with GEN‐317. GEN‐317 is shown as brick red sticks. The key
   residues are shown as blue sticks. Hydrogen bonds are shown as yellow
   dashed lines. Cation‐Pi interaction is shown as a green dashed line. G)
   Pull‐down assay demonstrates the protein interaction between GNE‐317
   and the recombinant N‐terminal FERM domain of Moesin. Biotin‐GNE‐317
   beads. Biotin‐GNE‐317 beads were constructed by conjugating
   biotinylated GNE‐317 to streptavidin‐coated beads. Free GNE‐317 refers
   to the addition of unbound GNE‐317 to recombinant of Moesin prior to
   pull‐down assays to neutralize binding. H) Hypothetical schematic
   diagram of Moesin activation under normal conditions and its inhibition
   by GNE‐317. I) Immunoprecipitation (IP) and immunoblot (IB) analyses of
   MSN polyubiquitination in GSC727‐1 overexpressing Flag‐tagged MSN, with
   or without GNE‐317 treatment. GSC727‐1 were then treated with 10 µM
   lactacystin (lacta) for 5h. Ub, ubiquitin. J) Immunoblot analysis of
   subtype marker genes in GSC727‐1 (Vector or MSN overexpression) treated
   with GNE‐317 (5 µm). K) Relative mRNA levels of DNA repair‐related
   genes in GSC727‐1 (Vector, MSN‐OE, MSN‐OE+GNE‐317 (5 µm)) with or
   without RT (6 Gy, 48 h) or TMZ (200 µm, 48 h) treatment. Data are shown
   as mean ± SEM. n = 3 independent experiments. ^* p < 0.05, ^** p
   < 0.01, ^*** p < 0.001; two‐tailed unpaired t‐test. L) Representative
   images and M) quantification of γ‐H2AX staining on GSC727‐1 (vector,
   MSN‐OE, MSN‐OE+GNE‐317) with or without RT treatment (6 Gy, 48 h) or
   TMZ (200 µm, 48 h). Scale bars, 10 µm. Data are represented as means ±
   SEM. n = 5 independent experiments. ^*** p < 0.001, ns, p >0.05;
   two‐tailed unpaired t‐test. N) Immunoblot and O) quantification
   analysis of downstream PI3K/mTOR signaling pathway in GSC727‐1 (Vector
   or MSN overexpression) treated with GNE‐317 (5 µm). P) Immunoblot and
   Q) quantification analysis of downstream PI3K/mTOR signaling pathway in
   GSC924 transduced with shNT or shMSN. R) Immunoblot analysis of key MES
   transcriptional regulators and pathways in GSC727‐1 (vector control or
   MSN overexpression) treated with GNE‐317 (5 µm). S) Immunoblot analysis
   of subtype marker genes in GSC727‐1 cells (vector control or MSN
   overexpression) treated with the PI3K inhibitor (LY294002, 10µm), the
   mTOR inhibitor (Torin1, 10 nm), or GNE‐317 (5 µM). T) Relative mRNA
   levels of DNA repair‐related genes in GSC727‐1 (Vector, MSN‐OE,
   MSN‐OE+LY294002 (10µm), MSN‐OE+Torin1 (10 nm), MSN‐OE+GNE‐317 (5 µm)).
   Data are shown as mean ± SEM. n = 3 independent experiments. ^* p
   < 0.05, ^** p < 0.01, ^*** p < 0.001; two‐tailed unpaired t‐test.

   We also investigated whether GNE‐317 exhibits a synergistic effect with
   the first‐line treatments RT and TMZ in MES GSCs. Combination matrices
   of cell inhibition and synergy scores (based on the ZIP model)
   demonstrated strong synergy in GSC924 cells with the combination
   treatment (Figure [165]8A). Considering optimal inhibitory efficiency
   and drug toxicity, we selected RT 8 Gy with GNE‐317 1 µm and TMZ 100 µm
   with GNE‐317 1 µm as the optimal combinations. In vitro, cell viability
   assays further demonstrated that the combined treatment not only
   exhibited stronger growth inhibitory effects but also significantly
   enhanced the sensitivity of cells to RT and TMZ (Figure [166]8B). Given
   that the first‐line clinical treatment for GBM is a combination of RT
   and TMZ, with RT or TMZ rarely used alone, we designed in vivo
   experiments using only a combined RT and TMZ treatment group (TMZ:
   20 mg^−1kg, RT: 2 Gy, GNE‐317: 40 mg^−1kg, administered five times per
   week for two consecutive weeks) (Figure [167]8C).^[ [168]^38 ^] Our
   results showed that the addition of GNE‐317 to the standard RT and TMZ
   treatment led to a notable inhibition of GSC growth in vivo and
   significantly prolonged mouse survival (Figure [169]8D–F). In
   conclusion, GNE‐317 exhibited robust synergy with RT and TMZ, markedly
   enhancing the sensitivity of MES GSCs to RT and TMZ. These findings
   suggest that the combination of GNE‐317 and standard chemoradiotherapy
   offers a promising new therapeutic strategy for improving outcomes in
   patients with GBM.

Figure 8.

   Figure 8
   [170]Open in a new tab

   GNE‐317 can reduce resistance to RT or TMZ in GSCs. With the indicated
   concentrations of GEN‐317 and RT/TMZ for 48 h. A) Cell inhibition
   matrices (left), synergy scores (middle), and 3D ZIP model (right) were
   calculated using the ZIP model in GSC924. White boxes and black crosses
   mark the optimal combination. B) Cell viability assay of GSCs treated
   with different treatments (RT: 8 Gy, TMZ: 100 µM, GNE‐317: 1 µm). n = 6
   biological independent samples. Data are shown as means ± SD. P values
   are in the table on the right, two‐way ANOVA analysis followed by
   Tukey's multiple test. C) A schematic of the experimental procedure for
   the GSC924 orthotopic xenograft model. Tumor‐bearing nude mice were
   treated with specified methods (DMSO (CON), RT, TMZ, GNE‐317, RT +TMZ,
   and RT +TMZ + GNE‐317) by gavage for 2 weeks, respectively. D)
   Representative images and E) quantification analysis of relative
   luciferase on days 10, 17, and 24 are shown. n = 5 for each group. Data
   are represented as means ± SEM. ^* p < 0.05, ^*** p < 0.001, ns, p
   > 0.05; one‐way ANOVA with Tukey's method for multiple comparisons. F)
   Kaplan–Meier survival curves of mice bearing GSC924‐derived xenografts
   treated with different treatments. ^*** p < 0.001, ns, p > 0.05;
   log‐rank test. n = 10 biological independent samples.

3. Discussion

   GBM is a highly aggressive brain tumor with a poor prognosis, primarily
   due to its resistance to standard first‐line therapies such as RT and
   TMZ. This resistance leads to recurrence and poses challenges for
   long‐term treatment. In this study, we identified MSN as a key driver
   promoting the PMT process in GSCs by activating the PI3K/mTOR signaling
   pathway. The resulting MES GSCs exhibit resistance to chemoradiotherapy
   by enhancing DNA damage repair following RT and TMZ treatments, thereby
   sustaining GBM malignancy and promoting tumor recurrence.

   Moesin (MSN), a membrane‐cytoskeletal linker protein, is critical for
   maintaining cell cortex organization and plays a pivotal role in cell
   morphogenesis and differentiation.^[ [171]^35 ^] Numerous studies have
   demonstrated that moesin is overexpressed in various tumors and plays a
   crucial role in their progression.^[ [172]^39 , [173]^40 ^] Moesin is
   suggested to be a potential marker for epithelial‐mesenchymal
   transition (EMT) in both breast and pancreatic cancers.^[ [174]^41 ,
   [175]^42 , [176]^43 ^] In addition, Moesin has been found to activate
   the Wnt/β‐catenin pathway and the PI3K/mTOR signaling pathway, serving
   as a marker for glioma progression.^[ [177]^44 ^] This provides
   theoretical support for the finding that MSN can activate the PI3K/mTOR
   signaling pathway in our study. Then, through a comprehensive analysis
   of the Cancer Cell Line Encyclopedia (CCLE) and Genomics of Drug
   Sensitivity in Cancer (GDSC) databases, we identified GNE‐317, a
   PI3K/mTOR inhibitor that crosses the blood‐brain barrier and
   specifically targets Moesin. By reversing the MSN‐Mediated PMT process,
   GNE‐317 synergizes with RT and TMZ to significantly enhance the
   sensitivity of GBM cells to chemoradiotherapy. (Figure [178]S9,
   Supporting Information).

   RT and TMZ are standard first‐line treatments for patients with GBM.
   TMZ is an alkylating agent that induces DNA damage by adding methyl
   groups to guanine at the O6 position, leading to DNA mismatches,
   persistent DNA double‐strand breaks (DSBs), and ultimately cell
   death.^[ [179]^45 , [180]^46 ^] Meanwhile, RT generates DNA damage
   primarily through ionizing radiation, leading to the formation of DSBs,
   which, if unrepaired, can trigger cell death.^[ [181]^47 , [182]^48 ^]
   However, extensive research has shown that glioma stem cells (GSCs)
   contribute significantly to resistance against both TMZ and RT.^[
   [183]^49 , [184]^50 ^] On the one hand, GSCs possess strong tumorigenic
   potential and the ability to maintain stemness, which supports
   continued tumor growth and therapy resistance.^[ [185]^51 ^] On the
   other hand, the PMT in GSCs is considered a hallmark of GBM recurrence
   and multi‐therapy resistance.

   The mechanisms driving PMT have drawn considerable interest. For
   instance, it has been shown that this PMT is mediated by C/EBPβ
   activation in GSCs after RT. Other studies have suggested that
   suppressing ASCL1 expression or elevating NDRG1 level in PN‐GSCs can
   likewise trigger PMT, further contributing to chemoradiotherapy
   resistance.^[ [186]^15 ^] Additionally, the tumor microenvironment
   (TME) contributes to PMT, with macrophage‐derived factors, such as
   oncostatin M (OSM), promoting the MES state and resistance to
   chemoradiotherapy.^[ [187]^52 ^] Moreover, our previous research
   indicated that CEBPB+ GBM subclusters can promote M2 macrophage
   polarization and that these M2 macrophages further contribute to the
   malignant progression of MES GBM cells, thereby enhancing their
   resistance to chemoradiotherapy.^[ [188]^22 ^] Together, these findings
   suggest that PMT is driven by a multifaceted network of intrinsic and
   extrinsic factors. In this study, we found that MSN can activate the
   PI3K/mTOR signaling pathway in GSCs, thereby upregulating the
   expression of key transcription factor networks and signaling pathways
   of the MES subtype (including CEBPB, NFKB, STAT3, and YAP/TAZ
   signaling), which triggers the PMT process and leads to resistance to
   RT and TMZ both in vitro and in vivo. This discovery provides novel
   insights into the mechanisms underlying PMT and highlights the
   potential of MSN as a therapeutic target to overcome treatment
   resistance in GBM.

   We focused our efforts on addressing chemoradiotherapy resistance in
   GBM by using adjuvant pharmacotherapy, specifically through repurposing
   existing small‐molecule drug inhibitors, for several key reasons.
   First, the presence of GSCs, particularly MES GSCs, drives GBM
   heterogeneity and resistance to treatment. Small‐molecule inhibitors
   can specifically target pathways underlying MES GSC therapy resistance,
   such as enhanced DNA repair. Second, GBM does not appear to be driven
   by a single pathway, as demonstrated by the lack of success in most
   monotherapy‐based clinical trials for this malignancy.^[ [189]^53 ^] By
   combining small‐molecule inhibitors with existing treatments such as RT
   and TMZ, it is possible to achieve synergistic effects that enhance
   therapeutic efficacy and overcome resistance. Third, reversing the PMT
   process with small‐molecule drugs capable of crossing the BBB to
   re‐sensitize GBM to chemoradiotherapy is a highly feasible therapeutic
   strategy. In this study, we identified GNE‐317, a small molecule
   capable of targeting MSN and reversing the PMT process. However,
   previous studies have shown that GNE‐317 exhibits varying inhibitory
   effects across different GBM cell lines. For example, while Chani
   et al. reported no significant effect on GL261 cells, Laurent et al.
   found that GNE‐317 notably inhibited the growth of U87, GS2, and GBM10
   cells.^[ [190]^38 , [191]^54 ^] These findings suggest that GNE‐317 may
   have specific targeting effects, consistent with our data showing that
   GNE‐317 had a significantly lower IC50 in high MSN GBM cells, such as
   U87 and T98G cells, which are typically highly resistant to
   chemoradiotherapy (Figure [192]1D and [193]7C). This further
   demonstrates that GNE‐317 may serve as a specific inhibitor of the MES
   subtype with high MSN expression to enhance GBM sensitivity to
   chemoradiotherapy.

4. Conclusion

   Our investigation revealed MSN functions as a crucial regulator of PMT
   in GSCs, driving resistance to RT and TMZ in GBM. We further
   demonstrated that GNE‐317 specifically targets MES GSCs with high MSN
   expression, reversing the PMT process and resensitizing these cells to
   chemoradiotherapy. Our analysis also showed that the variability in
   GNE‐317′s effects across GBM cell lines is likely due to differences in
   MSN expression. Specifically, GNE‐317 demonstrated a lower IC50 value
   in MES GSCs with high MSN expression. These findings underscore the
   potential of GNE‐317 as a targeted inhibitor of MES in GBM and offer an
   encouraging therapeutic strategy to overcome chemoradiotherapy
   resistance and enhance treatment outcomes for patients with GBM.

5. Experimental Section

Data Accessibility

   Bulk RNA‐sequencing data (TCGA, CGGA, Rembrandt, and Ivy GAP databases)
   were obtained from GlioVis ([194]http://gliovis.bioinfo.cnio.es/). The
   Gene Expression Omnibus (GEO, [195]http://www.ncbi.nlm.nih.gov/geo/)
   provided the scRNA‐seq data for glioma ([196]GSE117891), and spatial
   transcriptomics (ST) data were obtained from the 10X Genomics platform
   ([197]https://www.10xgenomics.com/cn). Drug sensitivity data for GBM
   cell lines were retrieved from the Genomics of Drug Sensitivity in
   Cancer (GDSC, [198]www.cancerRxgene.org), while transcriptional
   profiles for these cell lines were obtained from the Cancer Cell Line
   Encyclopedia (CCLE, [199]https://portals.broadinstitute.org/ccle/).

scRNA‐Seq Data and Spatial Transcriptomics Data Process

   scRNA‐seq and ST data were analyzed utilizing the R package Seurat
   (version 4.1.0). For scRNA‐seq, the R package DoubletFinder was first
   used to remove non‐single cells based on the optimal pK value, then
   filtered out low‐quality cells with fewer than 200 total feature RNAs
   and with genes expressed in fewer than three cells. Dimensional
   reduction was conducted through principal component analysis (PCA)
   (npcs = 30), succeeded by t‐distributed stochastic neighbor embedding
   (tSNE) algorithmic processing (dims = 1:10). Initial analysis utilizing
   FindNeighbors and FindClusters functions revealed nineteen distinct
   cell groupings. Characteristic genes were identified in each grouping
   through the FindAllMarkers function (only.pos = TRUE, min.pct = 0.25
   and logfc.threshold = 0.5). Further, each cluster was annotated
   manually using CNV values predicted by the inferCNV package (version
   1.8.0) along with marker genes. CNV calculations for all cell
   categories were derived from single‐cell sequencing data, implementing
   a 0 cutoff and 0.2 noise filter. Expression data underwent
   re‐standardization per sample, constraining values between ‐1 and 1.
   Each cell's CNV score was ascertained through quadratic summation of
   CNV regions.^[ [200]^55 ^] Non‐tumor cells and glioma cells were then
   further subdivided into 10 and 14 subclusters, respectively. Spatial
   transcriptomics data were processed according to the developer's
   recommended guidelines
   ([201]https://satijalab.org/seurat/articles/spatial_vignette.html).
   Finally, data normalization was performed using the SCTranform
   function, followed by dimensionality reduction using principal
   component analysis and tSNE algorithms. Clustering was achieved using
   the default resolutions of the first 30 principal components.

Integrating scRNA‐Seq and Spatial Transcriptomic Data

   A joint analysis of scRNA‐seq and spatial transcriptomic data was
   performed using the integration features provided by the Seurat
   package. First, dimensionality reduction was preprocessed and performed
   on the scRNA‐seq and spatial transcriptomic data as described above.
   The FindTransferAnchors function was then used to identify anchors
   between scRNA‐seq and spatial transcriptomic data (normalization.method
   = “SCT”), and apply the TransferData function to calculate prediction
   scores between them (dim = 1:30). Finally, the prediction scores of
   different single‐cell subclusters were obtained across various spatial
   clusters.

Gene Set Enrichment Analysis (GSEA) and Pathway Analysis

   First, the Limma package was used to perform differential gene
   expression analysis between the high and low‐MSN groups. The 296 genes
   highly regulated were then used by MSN (Log[2] (fold change) > 0.5,
   ‐Log[10] (p value) > 10) for the gene set enrichment analysis (GSEA
   software, version 4.1.0); pathway enrichment analysis was executed
   utilizing the R package clusterProfiler (version 4.0.5).

Molecular Docking

   A molecular docking model of Moesin was established with GNE‐317. The
   molecular structure of moesin was obtained from the RCSB Protein Data
   Bank ([202]http://www.rcsb.org) and processed using Schrödinger Protein
   Preparation Wizard ([203]https://www.schrodinger.com). The LigPrep
   module with the OPLS3e force field was utilized to perform ionization
   and minimization of GNE‐317′s structure. The preprocessed moesin and
   GNE‐317 underwent standard precision docking through the Glide module
   in the ligand‐docking component.

Cell and Culture Condition

   All cells were kept in a moisture‐controlled chamber at 37 °C under 5%
   CO[2] conditions. Commercial ATCC lines (NHA, A172, A1207, LN18, T98G,
   U251, and LN229) were grown in DMEM (Gibco) enriched with 10% fetal
   bovine serum (Mei5bio) and 1% penicillin‐streptomycin (Bioss).
   Glioblastoma tissue from GSCs (GSC428‐1, GSC428‐4, PGSC8, GSC3, GSC924,
   and GSC628) originated from surgical resections of glioblastoma
   patients treated at the Department of Neuro‐Oncology and Neurosurgery
   within Tianjin Medical University Cancer Institute and Hospital,
   Tianjin, China. The ethics board of Tianjin Medical University Cancer
   Institute and Hospital authorized all human tissue protocols
   (Ek2023099). Each participant provided written consent. Patient‐derived
   xenograft models served as sustainable GSC sources. STR profiling
   confirmed the authenticity of tumor models. The GSC populations were
   expanded as spheroid formations in neurobasal medium (Gibco) containing
   1% B27 supplement (Gibco), 2 mm L ‐glutamine, 1 mm sodium pyruvate,
   along with 10 ng‐1 mL bFGF and EGF (Gibco).

Realtime‐qPCR (RT‐qPCR) Analysis

   Total mRNA was procured and cleaned utilizing an RNA isolation kit
   (SparkJade) following the protocol provided by the manufacturer. An
   amount of 500 ng of messenger RNA underwent reverse transcription to
   generate cDNA employing the UEIris RT mixture with DNase components (Us
   EVERBRIGHT) in a T20 thermocycler (LongGene). RT‐qPCR analysis was
   performed utilizing Universal SYBR Green qPCR Supermix (Us EVERBRIGHT)
   components in a 7900‐thermocycler system (Applied Biosystems). The
   cycling protocol consisted of 32 cycle rounds with three phases: 95 °C
   lasting 30 s, 60 °C lasting 10 s, and 72 °C lasting 30 s. Expression
   values were normalized to 18S, and each RT‐qPCR was performed in
   triplicate for data analysis. Gene‐specific primers are listed in Table
   [204]S1 (Supporting Information).

Immunoblot Analysis

   Harvested cells underwent lysis utilizing RIPA buffer (Thermo
   Scientific) supplemented with phosSTOP phosphatase inhibitor (Roche)
   and protease inhibitor mixture (Sigma). Protein separation was
   accomplished through SDS‐PAGE (NuPAGE Bis‐Tris gel, Invitrogen)
   followed by NC membranes (Millipore) transfer. The membranes underwent
   blocking in 5% (wt/vol) non‐fat milk dissolved in TBS containing 0.5%
   (vol/vol) Tween‐20, followed by incubation with primary antibodies
   targeting Tubulin (1:5000, EASYBIO), MSN (1:5000, Abcam), γ‐H2AX
   (1:1000, CST), TP53 (1:1000, CST), Caspase‐3 (1:1000, Proteintech), AKT
   (1:2000, ABclonal), p‐ATK (1:1000, ABclonal), mTOR (1:2000, ABclonal),
   p‐mTOR (1:1000, ABclonal), YKL40 (1:2000, Abcam), CD44 (1:2000,
   Proteintech), SOX2 (1:1000, Proteintech), OLIG2 (1:2000, Abcam),
   Ubiquitin (1:1000, ABclonal), CEBPB (1:1000, Santa cruz), NFKB p65
   (1:5000, ABclonal), p‐NFKB p65 (1:10000, ABclonal), STAT3 (1:2000,
   ABclonal), p‐STAT3 (1:10000, ABclonal), YAP/TAZ (1:1000, CST) at 4 °C
   overnight. Following triple TBST washes, the membranes underwent room
   temperature incubation with HRP‐conjugated secondary antibodies:
   anti‐mouse IgG (CST, 7076), anti‐rabbit IgG (CST, 7074), and anti‐goat
   IgG (EASYBIO, BE0103) in 5% milk for 1 h. Signal detection employed
   luminol reagent (Millipore, WBKLS), with image acquisition performed
   using a molecular imager (BLT PHOTON TECHNOLOGY, GV6000 PLUS) and
   subsequent analysis via GV6000 M2 software. For the quantitative
   analysis of immunoblots, the Gels tool was used in ImageJ software to
   analyze bands. Three independent biological experiments were performed,
   and the quantified values were subjected to statistical analysis.

Human Tissue Microarray (TMA) Immunostaining and Multiplexed
Immunofluorescence Assays

   The TMA consisted of 101 tissue spots. Among these, 27 spots were
   derived from normal brain tissues obtained from autopsy specimens. The
   remaining spots consisted of glioma samples (WHO grades 1–4) collected
   from patients at Tianjin Medical University Cancer Institute and
   Hospital. Ethical approval for all procedures involving using human
   tissues was granted by the ethics committee of Tianjin Medical
   University Cancer Institute and Hospital (Ek2023099). Furthermore,
   informed consent was secured from all tissue donors. For
   immunostaining, the MSN antibody (1:50, Abcam) was employed to detect
   signals. A semi‐quantitative evaluation of tissue staining was
   performed by converting the proportion of positive cells and their
   staining intensities into corresponding numerical values. The H‐Score
   was determined using the following formula: H‐Score = ∑(pi × i) =
   (percentage of weak intensity area × 100) + (percentage of moderate
   intensity area × 200) + (percentage of strong intensity area × 300).^[
   [205]^56 ^] To identify the molecular subtype of each patient's GBM,
   FFPE slides of the tissue samples underwent multiplexed
   immunofluorescence and multispectral imaging using a PANO Multiplex IHC
   kit (Panovue), targeting CD44, YKL40, SOX2, and OLIG2. The staining
   procedure adhered to a standard protocol that entailed sequential
   incubation with primary and secondary antibodies (CD44: 1:8000, Abcam;
   YKL40: 1:1000, Abcam; SOX2: 1:400, Abcam; OLIG2: 1:500, Oasis Biofarm),
   followed by tyramide signal amplification (TSA). Nuclear staining was
   achieved using DAPI. Multispectral images were captured utilizing using
   the Mantra System (PerkinElmer). Tissue spots exhibiting detachment
   exceeding 30% were excluded from subsequent analysis.

Cell Viability and Treatment

   For the cell viability assay, the viability of cells was assessed on
   designated days following cell seeding, utilizing the Cell Counting
   Kit‐8 (TargetMol, USA, C0005) Assay Kit, in accordance with the
   manufacturer's instructions. GSCs (2000) were seeded into each well of
   a 96‐well plate, followed by incubation for 24 h for pre‐cultivation.
   Analysis was conducted on the specified days post‐seeding employing an
   enzyme‐linked immunosorbent assay reader. For treatments, TMZ and
   GNE‐317 were dissolved in dimethyl sulfoxide (DMSO), with the DMSO
   concentration strictly maintained below 0.1%. GSCs were exposed to
   varying drug and RT doses as dictated by the experimental protocol. At
   predetermined intervals, CCK‐8 assays were performed to evaluate the
   effects of the treatments.

Tumorsphere Formation Assays and Limiting Dilution Assays

   For the tumorsphere formation assay, 1000 GSCs were seeded into
   individual wells of a 96‐well plate. Tumorspheres were enumerated seven
   days post‐seeding. For the limiting dilution assays, GSCs were plated
   in a well of 96‐well plates at predefined densities (0, 1, 5, 10, 20,
   50 cells), with 16 replicates for each density. After seven days, the
   presence and count of tumorspheres in each well were documented and
   analyzed using the ELDA software
   ([206]http://bioinf.wehi.edu.au/software/elda/).

Plasmid and Lentiviral Transduction

   Lentiviral plasmids for MSN shRNA knockdown (shMSN‐1, shMSN‐2), MSN
   overexpression, and nonspecific control sequences (CON054) were
   obtained from Genechem (Shanghai, China). The vector elements
   hU6‐MCS‐CMV‐Puromycin were utilized for MSN shRNA knockdown, while
   Ubi‐MCS‐3FLAG‐SV40‐BSD was employed for MSN overexpression. Lentiviral
   particles were generated in 293T cells through the use of PAX2 and
   PMD2G helper plasmids (Addgene) in DMEM. For lentiviral transduction,
   GBM cells were infected with lentiviruses expressing the shMSN, MSN
   overexpression, or control sequences for 48 h before being subjected to
   subsequent analysis. For the CD44 reporter system, two luciferase
   constructs were transfected into the GSCs. Firefly‐luciferase, driven
   by the CMV promoter, served as an internal control, while
   NanoLuc‐luciferase driven by the CD44 promoter was used to assess CD44
   expression levels. After transfection, fluorescence signals were
   detected separately using substrates specific for F‐luciferase or
   N‐luciferase.

DNA Damage Detection

   To evaluate DNA damage in cells following RT and TMZ treatments, a
   γ‐H2AX immunofluorescence assay was conducted. Cells were fixed using
   4% paraformaldehyde for 15 min at room temperature and subsequently
   permeabilized with 0.1% Triton X‐100 for 10 min. Following PBS washes,
   blocking was performed with 5% BSA for 1 h. Cells were then incubated
   overnight at 4 °C with a primary antibody targeting γ‐H2AX (1:200,
   CST). On the following day, the cells were washed and treated with a
   fluorescently labeled secondary antibody for 1 h at room temperature.
   Nuclei were counterstained with DAPI. Confocal microscopy (Zeiss, LSM
   880) was utilized to capture images and assess DNA damage.
   Quantification of γH2AX foci was performed using the ImageJ software.
   Immunofluorescence images were first processed to enhance contrast and
   remove background noise. Individual γH2AX foci were identified and
   counted using the “Analyze Particles” function in ImageJ after
   appropriate thresholding. For each group, 5 randomly selected images
   were analyzed to determine the number of γH2AX foci per cell.
   Quantified values were subjected to statistical analysis.

Animal Experiments and Treatment

   All animal procedures were sanctioned by approval from the Animal
   Ethics and Welfare Committee of the Tianjin Medical University Cancer
   Institute and Hospital, Tianjin, China (Approval No. AE‐2022111,
   Ek2020157). Four‐week‐old female nude mice obtained from Beijing
   SiPeiFu Biotechnology Co., Ltd, were utilized in these studies. The
   mice were housed under a controlled environment with a 12 h light/12 h
   dark cycle, a temperature range of 20–26 °C, and humidity maintained
   between 30–70%, and were provided with ad libitum access to food. For
   all in vivo animal models, the mice were first weighed, and selected
   those with similar body weights for the experiments. These mice were
   then randomly assigned to the experimental and control groups, followed
   by the orthotopic implantation of the corresponding GBM cells. The
   specific procedure was as follows: first, anesthesia was administered,
   and the mice were positioned on a stereotactic injection apparatus to
   facilitate tumor implantation surgery. A burr hole was drilled into the
   right cerebral cortex of each mouse after exposing the head.
   Firefly‐luciferase‐expressing GBM cells (GSC924 cells (2 × 10^5) or
   GSC628 cells (2 × 10^5)) were implanted into the right cerebral cortex
   of nude mice at a depth of 3.5 mm, and the incision was sutured. The
   size of the orthotopic tumor was monitored weekly via the
   bioluminescence channel of the IVIS Spectrum optical imaging system.
   The relative luciferase signal was quantified by comparing the
   luciferase signal at each detection time point for each mouse to its
   signal at the first detection, resulting in the relative luciferase
   signal to monitor tumor growth in vivo. For in vivo treatment, TMZ at a
   dose of 20 mg^−1 kg was dissolved in a solvent composed of 40%
   polyethylene glycol‐300 (PEG‐300), 10% DMSO, 5% Tween‐80, and 45%
   saline. For intragastric administration, GNE‐317 at a dose of 40 mg^−1
   kg was prepared in a solution of 0.5% methylcellulose and 0.2% Tween‐80
   (MCT). RT was administered using fractionated doses (2 Gy x 5)
   delivered through a ^60Co teletherapy unit equipped with a custom gig
   and validated dosimetry. According to the standard protocol, treatments
   were performed five times per week.

Drug Combination Assay

   GSC924 was treated with varying concentrations and doses of treatments
   to investigate whether GNE‐317 exhibits a synergistic effect with RT or
   TMZ. The detailed procedure was as follows: First, an equal number of
   GSC cells were seeded in 96‐well plates. The cells were then treated
   with different concentrations of the drugs as follows: Group 1: RT
   (Gy): 0, 4, 8, 12, 16 and GNE‐317 (µm): 0, 1, 2, 5, 10, 15; Group 2:
   TMZ (µm): 0, 100, 200, 400, 1000 and GNE‐317 (µm): 0, 1, 2, 5, 10, 15.
   After 48 h of combined treatment, cell viability was assessed using the
   CCK8 assay. The results were normalized to the control group
   (untreated), and the inhibition rate of the drug combinations was
   calculated. Finally, the synergistic effect was quantified by
   calculating the synergistic score using the ZIP model in SynergyFinder
   based on the inhibition rates.^[ [207]^57 ^]

Statistical Analysis

   All grouped data were presented as mean ± SEM or mean ± SD. Comparisons
   between groups were conducted using one‐way ANOVA or Student's t‐test.
   For Kaplan‐Meier survival analyses, statistical significance was
   determined via the Wilcoxon test or log‐rank test. To assess the
   correlation between samples, two‐sided Pearson correlation coefficients
   were employed to evaluate sample correlations. All statistical analyses
   were carried out using Microsoft Excel 2019, GraphPad Prism(versions 8
   and 9), or R version 4.0.5, with statistical significance defined as p
   < 0.05 (^*** p < 0.001; ^** p < 0.01; ^* p < 0.05; ns, p > 0.05).
   Additional details are provided in each figure legend. For all results,
   excluding those derived from the public databases, similar findings
   were obtained from three independent experiments.

Conflict of Interest

   The authors declare no conflict of interest.

Supporting information

   Supporting Information
   [208]ADVS-12-2412517-s001.docx^ (47.4MB, docx)

Acknowledgements