IFN-γ signaling drives multiple immunosuppressive pathways and poor clinical outcome in glioblastoma

Abstract

Background

Glioblastoma (GBM) is a highly aggressive brain tumor characterized by a profoundly immunosuppressive tumor microenvironment and limited response to immunotherapy. While interferon-γ (IFN-γ) signaling is classically associated with antitumor immunity, sustained activation of this pathway has been linked to immune dysfunction in several cancers. However, its tumor-intrinsic role in shaping immunosuppression in GBM remains poorly understood.

Methods

We integrated transcriptomic and immunohistochemical analyses with clinical data from GBM patients to assess associations between IFN-γ signaling, patient survival, immune cell infiltration, and immune checkpoint expression. Genetic and pharmacological perturbation of IFN-γ signaling was performed to dissect pathway-specific effects in GBM-derived cellular models. Functional immune responses were evaluated through co-culture assays with human peripheral blood mononuclear cells (PBMCs) or isolated natural killer (NK) cells.

Results

IFN-γ signaling was found significantly upregulated in GBM and associated with poor patient survival. While the activation of this pathway reduced tumor cell growth, it simultaneously induced a broad immunoregulatory transcriptional program. In turn, high IFN-γ activity was associated with increased myeloid infiltration, impaired NK and T cell cytotoxicity, and upregulation of multiple inhibitory immune checkpoints. In this context, targeting a single checkpoint, such as PD-1, failed to restore antitumor immunity. By contrast, genetic ablation of STAT1, but not IRF-1, effectively reversed IFN-γ-induced immunosuppression and restored immune effector function, identifying STAT1 as a central mediator of this phenotype. Pharmacological inhibition of JAK1/2 recapitulated STAT1 loss in vitro, suppressing IFN-γ-dependent immune inhibitory programs and restoring cytotoxic activity.

Conclusions

These findings demonstrated that tumor IFN-γ signaling drives a STAT1-dependent immunosuppressive state in GBM that functionally impairs antitumor immunity. Our study defines a dominant regulatory axis linking IFN-γ signaling to immune escape mechanisms in GBM and provides a mechanistic framework for future studies exploring therapeutic modulation of this pathway to enhance immunotherapy response.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12916-026-05128-9.

Background

Glioblastoma (GBM) is the most prevalent primary cancer of the central nervous system, accounting for approximately half of all malignant brain tumors in adults [1]. It is a highly aggressive cancer and commonly refractory to temozolamide and novel therapies, with a median survival time of ~ 15 months following diagnosis [2]. Immunosuppression is a hallmark of GBM, orchestrated by tumor-derived factors and the surrounding brain microenvironment, which impairs both local and systemic immunity [3–6]. Immunosuppression is further exacerbated by standard treatments, including corticosteroids, temozolomide, and radiation therapy, ultimately leading to reduced immunotherapy efficacy and decreased overall survival [6–10].

Multiple mechanisms contribute to this immunosuppressive landscape, including the tumor’s location behind the blood-brain barrier [11], the recruitment of immunosuppressive cells, such as regulatory T cells and tumor-associated macrophages [5, 11, 12], and the low infiltration of cytotoxic natural killer (NK) and CD8 + T cells, which are essential for mounting effective antitumor immunity [13, 14]. GBM cells also overexpress immunosuppressive mediators such as IDO-1 [15, 16], TGF-β [17, 18], and IL-10 [19]. Similarly, tumor-infiltrating lymphocytes (TILs) in GBM display high levels of inhibitory immune checkpoints, including PD-1, CTLA-4, LAG-3, Tim-3, BTLA, and TIGIT [13, 20, 21]. Nevertheless, immune checkpoint inhibitors (ICIs) in monotherapy with anti-PD-1/PD-L1 or anti-CTLA-4 have shown limited success in GBM [22–25]. Notably, we have recently shown that ILT2 functions as a relevant inhibitory immune checkpoint in GBM, with its expression correlating with poor patient survival. Moreover, ex vivo blockade of ILT2 yielded promising therapeutic potential [26].

Interferon-gamma (IFN-γ) is a key pleiotropic cytokine mainly produced by activated NK cells, CD4 + Th1 cells, and CD8 + cytotoxic T cells that regulates the activity of multiple immune cells, thereby coordinating their response to infections [27–29]. IFN-γ exerts dual and context-dependent roles in cancer [30]. It has a critical antitumor role, preventing both tumor initiation and progression [31–35]; however, in advanced tumors, aberrant IFN-γ signaling can paradoxically drive immune evasion and tumor progression [36–40].

In GBM, IFN-γ signaling is associated with proliferation and a mesenchymal phenotype primarily through STAT1 activation [41], while showing extensive crosstalk with several oncogenic pathways such as EGFR and PI3K [42–44]. It also facilitates myeloid cell infiltration and suppresses adaptive immunity via upregulation of IDO-1, TGF-β, chemokines, and PD-L1 [45–47]. Accordingly, IFN-γ-related gene signatures are associated with a worse prognosis in glioma and are linked to immune checkpoints overexpression [45, 48, 49].

Based on previous evidence supporting the involvement of IFN-γ signaling in immunosuppression and therapy resistance [50–53], we hypothesized that this pathway plays a crucial role in regulating immune evasion in GBM. Specifically, we proposed that targeting the IFN-γ/JAK/STAT1 axis is a promising strategy to reprogram the GBM immune microenvironment and overcome immunotherapeutic resistance. To address this, we integrated transcriptomic and immunohistochemical data from GBM patient cohorts, revealing that IFN-γ pathway activation correlates with poor prognosis and immunosuppressive remodeling of the tumor microenvironment (TME). Functional studies confirmed that IFN-γ signaling promotes immune escape through the upregulation of multiple inhibitory molecules that impair antitumor responses. Our results identify STAT1 as a key mediator of this process and support JAK/STAT pathway inhibition as a promising therapeutic strategy in GBM.

Altogether, this work provides mechanistic insight into IFN-γ-induced immunosuppression in GBM and identifies the IFN-γ pathway as a prognostic biomarker that could help to stratify patients. We also propose that IFN-γ inhibition may be a promising immunotherapeutic approach to enhance immune responses and improve clinical outcomes in this devastating disease.

Methods

In silico data analysis

RNA sequencing data from patients with GBM were obtained from The Cancer Genome Atlas (TCGA) Firehose Legacy dataset, published by the Broad Institute [54–57]. This dataset includes clinical annotations, transcriptomic profiles, and summary visualizations from 619 tumor samples corresponding to 606 GBM patients. Additional transcriptomic data were retrieved using GEPIA2 [58], which integrates 163 GBM samples from TCGA with 207 normal brain tissue samples from the GTEx project for comparative gene expression analysis. Differential gene expression analysis was performed using DESeq2, with significant changes defined as log₂FC > 1 or < -1 and an adjusted p-value (adj-P) < 0.01. Pathway enrichment analysis was conducted based on the Molecular Signatures Database (MSigDB), with hallmark pathways ranked by normalized enrichment score (NES) (-log₁₀(q-value)). To assess IFN-γ signature activity, the relative expression of the corresponding genes was analyzed, and patients were stratified based on median expression values, as previously described by Qian et al. [45], Ayers et al. [59], and MSigDB Hallmark IFN-γ Response signature [60]. Additionally, an immune response-related gene set was analyzed in these samples based on Hu et al. [61]. To analyze gene expression and IFN-γ signature scores across gliomas of different World Health Organization (WHO) grades, data were retrieved from the Gliovis platform [62] (TCGA), including 226 grade II, 244 grade III, and 150 grade IV glioma samples. The immune infiltration was estimated by deconvolution analysis using TIMER 2.0 based on the xCell 2.0 algorithm [63]. Normalized RNA expression data from low-grade glioma (LGG) or GBM and normal brain tissue, along with survival information, were retrieved from GEPIA2 [58]. Venn diagrams were generated with Venny 2.0 [64].

Patients and tissue samples

Formalin-fixed paraffin-embedded (FFPE) tissue samples from surgically resected GBM tumors at the Hospital Universitario Central de Asturias (HUCA), Spain, were obtained from the Asturias Biobank (BioPA, PT22/0007). Ethical approval was obtained from CEImPA (2023.5), and informed consent was obtained from all patients. Clinical characteristics of the cohort, including age, survival, tumor location, and laterality in relation to IRF-1 and STAT1 immunohistochemistry, are summarized in Additional file 1: Table S1, while molecular features are summarized in Additional file 1: Table S2.

Immunohistochemistry

FFPE GBM samples were analyzed by immunohistochemistry using the EnVision FLEX Mini Kit and the Dako Autostainer system (Agilent), following the manufacturer’s protocol. Briefly, 3 μm paraffin-embedded sections were deparaffinized, rehydrated, and subjected to heat-induced epitope retrieval (HIER) at 95 °C for 20 min in a pH 9 buffer using the PT Link Pre-Treatment Module. Endogenous peroxidase activity was quenched with the EnVision FLEX Peroxidase-Blocking Reagent for 5 min, followed by a 60-minute incubation with Protein Block (serum-free buffer). Tissue sections were then stained with polyclonal antibodies against human STAT1 (Cell Signaling Technology, #14994) at 1:200 dilution and IRF-1 (Cell Signaling Technology, #8478) at 1:200 dilution and counterstained with hematoxylin. Antibody specificity was verified by performing negative control staining in which the primary antibody was omitted, and tissue sections were incubated with the secondary antibody alone. All specimens were independently evaluated by a pathologist.

Genomic DNA extraction and methylation analysis

Genomic DNA from patients with GBM was extracted, and the following analyses were performed: Methylation-specific PCR (MSP) was conducted using primers described by Esteller et al. [65]. to determine the methylation status of the MGMT gene promoter. Gene copy alterations in TP53, PTEN, CDKN2A, EGFR, and PDGFRA (located at 17p13.1, 10q23.3, 9p21, 7p12, and 4q12, respectively), along with probes targeting regions on 12q, 14q, and the TERT promoter, were analyzed using MLPA with the P105-E1 kit (MRC-Holland). MLPA results were interpreted as follows: gene amplification was classified as low (amplification ratio 1.2-2), moderate [2–20], or high (> 20). Loss of heterozygosity (LOH) was defined as a ratio < 0.7, while homozygous deletion was defined as a ratio < 0.25. For IDH1 and IDH2 mutation screening, the presence of the R132 mutation in the IDH1 gene and the R172 mutation in the IDH2 gene was assessed using MLPA with the P088 kit (MRC-Holland). 1p/19q codeletion analysis was also performed. Loss of chromosomal arms 1p and 19q was evaluated by MLPA using the P088 kit (MRC-Holland).

Cell culture

LN-18, LN-229, T98G, and HEK293T cells were obtained from the American Type Culture Collection (ATCC) and cultured in DMEM supplemented with 10% heat-inactivated FBS, 1 mM sodium pyruvate, 2 mM L-glutamine, 100 U/mL penicillin, and 10 µg/mL streptomycin. Cells were maintained at 37 °C in a humidified atmosphere with 5% CO₂. Cas9-expressing LN-18 and LN-229 cell lines were generated by lentiviral transduction using pKLV2-EF1a-Bsd2Cas9-W (Addgene, #67978) and selected with 4 µg/mL blasticidin (Gibco) for at least one week. When indicated, specific experiments included treatments with IFN-γ (Peprotech #300-02) or ruxolitinib (MedChemExpress).

Western blot

Whole-cell lysates from LN-18, LN-229, T98G, sgCTR vector, and knock-out (KO) cells were prepared using RIPA lysis buffer (Thermo Scientific, 89900) supplemented with dithiothreitol (DTT) (Thermo Scientific) and protease-phosphatase inhibitor cocktail (Sigma Aldrich, 78430). Lysate mixes were kept at 4 °C in rotation for 30 min and then centrifuged at maximum speed for 10 min at 4 °C. Then, supernatants were collected, and protein concentration was determined by the Pierce BCA assay (Thermo Scientific, 23225). Proteins were resolved on SDS-PAGE by using NuPAGE™ 4 to 12% Bis-Tris gels (Life Technologies) with MOPs running buffer and transferred to Trans-Blot^® Turbo™ Nitrocellulose membranes (Bio-Rad Laboratories). Membranes were blocked for 1 h in TBS containing 0.1% (v/v) Tween-20 and 0.5% (w/v) BSA. Primary antibodies against IRF-1 (Cell Signaling Technology, #8478), STAT1 (Cell Signaling Technology, #14994), phospho-STAT1 (Cell Signaling Technology, #9167), β-actin (Sigma, #A2228), EGFRvIII (Invitrogen, #MA5-36216), and GAPDH (Invitrogen, #MA5-35235) were incubated overnight at 4 °C at a 1:1,000 dilution in TBS solution. Secondary antibodies, goat anti-rabbit IRDye 800CW or anti-mouse IRDye 680RD (IRDye, LICOR, at 1:10,000 dilution) for 1 h at room temperature. Membranes were washed in TBS-Tween 0.1%, and immunocomplexes were detected and measured using Odyssey^® Fc Imager (LICOR Biosciences) and analyzed with Image Studio Lite software (LICOR Biosciences).

Viability assay

LN-18, LN-229, and T98G cells (5000 cells/well) were seeded in a 96-well plate, and the next day, the cells were treated with IFN-γ at different concentrations. Cell viability was assessed using the make-to-stock (MTS) kit (CellTiter 96 Aqueous One Solution Cell Proliferation Assay from Promega, Madison, WI, USA) at 72 h. Data is presented as the percentages of four replicates. The absorbance was measured using a Synergy HT plate reader (BioTek, Winooski, VT, USA).

Proliferation assay

LN-18, LN-229, and T98G cells were stained with Carboxyfluorescein Succinimidyl Ester (CFSE) cell tracker (ChemCruz) at day 0. The following day, cells were treated with 10 ng/mL of IFN-γ for 24, 48, and 72 h. All samples were collected simultaneously, and proliferation was measured by the dilution of CFSE in the treated conditions compared to the control by flow cytometry.

Apoptosis and cell death

LN-18, LN-229, and T98G cells were treated with 10ng/mL of IFN-γ for 48 h. At this time, cells were treated with temozolomide (MedChemExpress) at 200 µM or 500 µM, maintaining the stimulation of IFN-γ for 48 h. Apoptosis and cell death were measured by flow cytometry using 3,3’-dihexyloxacarbocyanine iodide (DIOC 6,3) (Sigma Aldrich) and propidium iodide (PI) (Immunostep), respectively. Both supernatants (containing floating and adherent cells) were collected for flow cytometry analysis.

RNA sequencing

LN-18 and LN-229 cells were cultured with 10 ng/mL IFN-γ for 48 h. Total RNA was then extracted in three biological replicates using the GeneJET RNA Purification Kit (Thermo Scientific). All steps following RNA extraction, including mRNA enrichment, library preparation, sequencing, and primary bioinformatic analysis, were performed by Novogene, Inc. Differential gene expression analysis was performed using DESeq2, and significant changes were defined as log₂FC > 1 or < -1 and an adjusted p-value (adj-P) < 0.01. Gene set enrichment analysis (GSEA) was conducted based on the Molecular Signatures Database (MSigDB), with hallmark pathways ranked by normalized enrichment score (NES) (-log₁₀(q-value)).

Flow cytometry

Cell suspensions treated with IFN-γ, ruxolitinib, or the combined treatment were stained for flow cytometry. The fluorochrome-tagged monoclonal antibodies (mAbs) used in phenotyping GBM cell lines were: anti-HLA-I (-A, -B, -C) (Biolegend, PE, clone W6/32 #B394406), anti-HLA-II (-DR, DP-, DQ-) (Biolegend, Alexa-Fluor 647, clone Tü39 #B394686), anti-PD-L1 (Biolegend, PE, clone 29E.2A3 #387142), and anti-LGALS9 (Biolegend, PE, clone 9M1-3 #B370901). The remaining samples were then analyzed using CytoFLEX (Beckman Coulter) and FlowJo software (BD Biosciences).

Isolation of PBMCs and NK cells

Buffy coats from healthy donors were obtained from the Centro Comunitario de Sangre y Tejidos de Asturias following the Declaration of Helsinki. Peripheral blood mononuclear cells (PBMCs) were isolated by density gradient centrifugation using Ficoll (Biowest) and cultured in RPMI 1640 medium supplemented with 10% heat-inactivated FBS, 1 mM sodium pyruvate, 2 mM L-glutamine, 100 U/mL penicillin, and 10 µg/mL streptomycin. Unless otherwise specified, cultures were maintained at 37 °C with 5% CO₂ and supplemented with 100 ng/mL IL-2 (Peprotech). NK cells were purified from PBMCs by negative selection using the MojoSort™ Human NK Cell Isolation Kit (Biolegend, 480054), following the manufacturer’s instructions. Isolated NK cells were cultured at a density of 2 × 10⁶ cells/mL for 36 h in IL-2-supplemented RPMI 1640, either alone or in the presence of individual antibodies or their combination. Antibodies used for NK cell receptor blockade experiments included anti-ILT2 (clone HP-1, kindly provided by Miguel López-Botet, Universitat Pompeu Fabra), anti-LAG-3 (MedChemExpress), anti-PD-1 (Biolegend clone EH12.2H7), and anti-TIM-3 (MedChemExpress), each at a concentration of 10 µg/mL.

Cytokine analysis

GBM cells were treated or not with 10 ng/mL of IFN-γ for 48 h. Then supernatants were aspirated, and cells were washed twice with PBS 1x and cocultured at 5:1 effector-to-target (E: T) with PBMCs. Different assays were performed to measure cytokine production. For LEGENplex and ELISA, PBMCs were in contact with tumor cells for 24 h until they were stimulated with 20 ng/mL of phorbol 12-myristate 13-acetate (PMA) and 500 ng/mL of ionomycin for 24 h. Then, supernatants were collected. LEGENDplex Human CD8/NK Panel (Biolegend #741187) was used to estimate the concentrations of selected anti-inflammatory and proinflammatory cytokines in the PBMCs: GBM co-culture supernatants. The LEGENDplex assay was performed according to the manufacturer’s procedure. Samples were acquired on a CytoFLEX (Beckman Coulter) with a 488-nm laser and a 536/40 (BP) filter for the fluorochrome PE and a 638-nm laser and 675/20 (BP) filter for the fluorochrome APC. Results were analyzed using LEGENDplex Data Analysis Software (version 2023-02-15 (64759)). The concentration of each cytokine was determined using a standard curve generated during the assay. The human IFN-γ ELISA (Peprotech #900-K27) kit was performed with the same supernatants following the manufacturer’s instructions.

Immune synapse analysis

Isolated NK cells were stained with Tag-it violet (Biolegend #B390313) at 5 µM at a density of 1 × 10⁶ cells/ml following the manufacturer’s instructions. NK cells were then co-cultured with ZsG + GBM cells, either pretreated or not with IFN-γ (10 ng/mL) for 48 h, at a 1:1 (E: T) ratio for 4 h in suspension. Double-positive ZsG+/Tag-it Violet+ cells detected by flow cytometry were interpreted as immune synapses.

Immune cell degranulation, activation, and cytotoxicity assays

GBM cell lines were pretreated with vehicle, IFN-γ (10 ng/mL), ruxolitinib (1 or 10 µM), or the combination for 48 h prior to co-culture with PBMCs or isolated NK cells at the indicated effector-to-target (E: T) ratios. For degranulation assays, PBMCs were cocultured with GBM cells at a 5:1 E: T ratio for 4 h. After 1 h of coculture, 2 µM Monensin (BioLegend #B297750) and fluorochrome-labeled anti-CD107a mAb (BioLegend, PE, clone HYA3, #B321484) were added to block intracellular protein transport and detect degranulation, respectively. At the end of the assay, cells were stained with anti-CD3 mAb (BioLegend, FITC, clone UCHT1, #B358130 or BioLegend, APC, clone OKT3, #B363504, anti-CD56 mAb (BioLegend, APC, clone QA17A16, #B445634 or BioLegend, FITC, clone 5.1H11, #B351042) and anti-CD8 mAb (Immunostep, APC/750, #420086) to distinguish NK cells (CD56⁺CD3⁻) and CD8⁺ T cells (CD3⁺CD8⁺), followed by flow cytometric analysis.

For activation assays, PBMCs were co-cultured with GBM cells at a 25:1 E: T ratio for 48 h. Cells were then stained using the same gating strategy as above, together with anti-CD4 (Immunostep, CFBlue, #6012025) and the early activation marker anti-CD69 mAb (Biolegend, PE, clone FN50, #B490357) to identify activated CD4⁺ T cells (CD3⁺CD4⁺), CD8⁺ T cells and NK cells populations and analyzed by flow cytometry.

For cytotoxicity assays, isolated NK cells were stained with Tag-it Violet (following the manufacturer’s instructions) and after 4 h of coculture at a 5:1 E: T ratio, propidium iodide (Immunostep) was added to identify dead target cells by flow cytometry.

Generation of GBM knock-out and EGFRvIII models

The day before transfection, HEK293T cells were seeded in 6-well plates at 40–60% confluence. In the case of GBM KO cell lines, transfection was carried out using a mixture containing 0.75 µg VSV.G, 1.5 µg pLP1, 0.75 µg pLP2, and 1 µg of the vector of interest, combined with 1:2 (µg DNA:µL) ratio of Lipofectamine (Thermo Scientific #11668027) per well in Opti-MEM (Thermo Scientific). After 16 h, the medium was replaced with 2 mL of complete DMEM. At 48 h post-transfection, the supernatant containing lentiviral particles was collected, filtered through 0.45 μm membranes, and either used immediately or aliquoted and stored for future transductions. Cas9-expressing cell lines were transfected with an empty vector (sgCTR) or with guide RNAs (gRNAs) targeting exon sequences of STAT1 or IRF-1 (Addgene #67975) using polybrene at 8 µg/mL (Sigma Aldrich, #TR-1003-G). Two independent KO cell populations were established for each gene using distinct gRNAs.

STAT1 KO1 gRNA: CTGATTAATGATGAACTAG.

STAT1 KO2 gRNA: AGGTCATGAAAACGGATGG.

IRF-1 KO1 gRNA: AACTCCCTGCCAGATATCG.

IRF-1 KO2 gRNA: TAATTCCAACCAAATCCCG.

For the EGFRvIII-mutant model, lentiviral transfection was performed using a mixture containing 3 µg psPAX2, 1.5 µg pMD2.G, and 4 µg of the vector of interest, using a 1:2 (µg DNA:µL) ratio of Lipofectamine (Thermo Scientific #11668027) per well in Opti-MEM (Thermo Scientific).

The EGFRvIII construct was designed using the VectorBuilder platform and contains an in-frame deletion of exons 1–8 of the EGFR gene under the control of the EF1A promoter, along with GFP and a puromycin resistance cassette (Use ID: VB251007-1200jkp to retrieve detailed information about the vector on vectorbuilder.com). The corresponding empty vector, lacking the EGFRvIII deletion, was used as a control (CTR). Lentiviral particle collection and transduction were performed as previously described.

KO cells were selected using blasticidin and puromycin, and mutant cells were selected using puromycin. Editing efficiency was confirmed at the protein level by Western blot analysis.

Bispecific T-cell engager antibody generation

The day before transfection, HEK293T cells were seeded in 6-well plates at 40–60% confluence. The bispecific T-cell engager (BiTE) targeting CD3 and EGFRvIII (αCD3;αEGFRvIII) was produced by transfecting cells with 2 µg of the pBITE-EGFRvIII-F269-OKT3 vector (Addgene #190681) [66] using PEI (MedChemExpress #478248) at a 1:4.5 (µg DNA:µL) ratio in DMEM without antibiotics or FBS. Six hours post-transfection, the medium was replaced with complete DMEM, and cells were maintained for 7 days to allow BiTE production. Supernatants containing BiTEs were then collected, filtered through 0.45 μm membranes, aliquoted, and stored until use. For functional assays, 200 µL of BiTE-containing supernatant was pre-incubated with 1 × 10⁶ PBMCs (2 × 10⁶ cells/mL) for 30 min prior to co-culture with GBM cell lines.

Statistical analysis

For in silico mRNA expression analysis of individual genes, a one-way ANOVA test was used. For in silico analyses of survival, the log-rank (Mantel-Cox) test was applied, and patient groups were categorized based on the average expression of each gene. Survival data from GBM patients treated at HUCA were analyzed using IBM SPSS Statistics. Overall survival, measured in days, was assessed using Kaplan-Meier curves and the log-rank test. Log-transformed survival data were also used for statistical modeling to assess differences between patient subgroups. Patients were stratified based on positive or negative staining, and survival differences were evaluated using univariate and multivariate Cox proportional hazards regression models. Clinical characteristics were compared according to IRF-1 and STAT1 expression status. Categorical variables were analyzed using Pearson’s chi-squared test. Continuous variables are expressed as mean values and were compared between groups using the independent-samples Student’s t-test. For in vitro assays, statistical analysis was performed using GraphPad Prism 8 (GraphPad Software, San Diego, CA). Data distribution was assessed using the Shapiro-Wilk test. For normally distributed data, comparisons between two independent groups were performed using the unpaired Student’s t-test, while paired t-tests were used for intra-group comparisons. For non-normally distributed data, the Mann–Whitney U test (for unpaired samples) or the Wilcoxon signed-rank test (for paired samples) was applied. Results are presented as mean ± standard error of the mean (SEM), unless otherwise specified. A p-value less than 0.05 was considered statistically significant. Statistical significance is indicated as follows: p < 0.05 (*), p < 0.01 (**), and p < 0.001 (***).

Results

The IFN-γ pathway is upregulated and correlates with poor survival in GBM

To gain novel insights into the mechanisms involved in GBM immunosuppression, we analyzed publicly available transcriptomic data from patients with GBM from the TCGA dataset through the GEPIA2 platform. Comparative gene expression analysis was performed using 163 GBM tumor samples from TCGA and 207 normal brain tissue samples from the GTEx project. Differential expression analysis of samples from patients with GBM revealed that 5,221 genes were significantly upregulated whereas 2,438 were significantly downregulated (log2FC > 1 or <-1, padj < 0.01) (Fig. 1A). Pathway enrichment analyses of the upregulated genes revealed multiple pathways associated with cell proliferation, apoptosis, epithelial-to-mesenchymal transition (EMT), and immune regulation (Fig. 1B). Notably, immune-related pathways accounted for approximately 40% of the total enriched pathways, with the IFN-γ response gene set showing the highest enrichment (Fig. 1B). In agreement, genes encoding key members of the IFN-γ pathway: IFNGR1, IFNGR2, JAK1, STAT1, and IRF-1, were found significantly upregulated in LGG and/or GBM compared to normal brain tissue (Fig. 1C and Additional file 2: Fig. S1A). Particularly, STAT1 and IRF-1 genes, which encode key transcription factors that mediate the main cellular responses induced by IFN-γ signaling, were significantly upregulated in both LGG and GBM (Fig. 1C). There was a strong positive correlation between the expression of these two genes (Pearson r = 0.58, p < 0.0001), and high levels of either STAT1 or IRF-1 genes were associated with higher tumor grade (Additional file 2: Fig. S1B-D) and poorer overall survival in patients with GBM (Fig. 1D). Similarly, transcriptomic profiling of LGG and GBM patients using multiple independent IFN-γ-related gene signatures (Additional file 2: Fig. S1E) demonstrated a consistent association between IFN-γ pathway activation and patient overall survival (Fig. 1E-F, Additional file 2: Fig. S1F-G). The IFN-γ score reported by Qian et al. [45]. was significantly elevated in both LGG and GBM patients (Fig. 1E), whereas the signatures from Ayers et al. [59]. and MsigDB [60] were specifically enriched in GBM (Additional file 2: Fig. S1F-G). Across all signatures, higher IFN-γ activity correlated with advanced disease stage and poorer overall survival (Fig. 1F, Additional file 2: Fig. S1F-H), suggesting that persistent IFN-γ signaling is associated with glioma progression and unfavorable clinical outcome.

Fig. 1.

IFN-γ response genes are upregulated in glioblastoma (GBM) and associated with patient prognosis. (A) Volcano plot of differentially expressed genes (DEGs) from bulk RNA-seq of GBM patients compared to normal brain tissue (log₂FC > 1, adjusted p < 0.01). The number on top indicates the total number of DEGs. Selected IFN-γ-related genes are highlighted. (B) Pathway enrichment analysis of upregulated genes using the Hallmarks MSigDB database. (C) Whisker box plot showing mRNA expression levels of the key IFN-γ signaling regulators STAT1 and IRF-1 in low-grade glioma (LGG, n = 518) and glioblastoma (GBM, n = 163) (red boxes) compared with normal brain tissue (control, n = 207) (grey boxes). (D) Kaplan Meier survival analysis categorized by mRNA levels of the indicated genes in LGG and GBM, comparing high (n = 338) and low (n = 338) expression groups based on the average gene expression. (E) IFN-γ gene signature score expression and corresponding Kaplan-Meier survival analysis (F), stratified as in (D). G, H) Immunohistochemical staining of IRF-1 (n = 229) (G) and STAT1 (n = 243) (H) in GBM tumor sections. Representative images show negative (a), low (b), moderate (c), and high (d) staining (scale bar = 100 μm). Bottom panels show Kaplan-Meier survival analysis comparing patients with positive and negative expression of IRF-1 or STAT1. Data in panels (A-F) are from the TCGA, GEPIA2 database. HR, hazard ratio. *p < 0.05

To further assess the clinical relevance of STAT1 and IRF-1 in GBM, we performed immunohistochemical analyses in a cohort of patients with GBM surgically treated at HUCA (Fig. 1G-H, Additional file 1: Tables S1-S2). Cytoplasmic IRF-1 expression was detected in 138 of 229 cases (60.3%), whereas 91 tumors (39.7%) were IRF-1-negative (Additional file 1: Table S1). Similarly, STAT1 expression was observed in 164 of 243 GBM samples (67.5%), whereas 79 tumors (32.5%) were STAT1-negative (Additional file 1: Table S1). No significant associations were identified between IRF-1 or STAT1 expression and clinicopathological variables including patient age, tumor location, or tumor laterality (Additional file 1: Table S1). Similarly, no significant associations were identified between either IRF-1 or STAT1 expression and recurrent genomic alterations such as PTEN deletion/loss of heterozygosity, CDK4 amplification, CDKN2A loss of heterozygosity, CDKN2A deletion, MDM2 amplification, PDGFRA amplification, or MGMT promoter methylation (Additional file 1: Table S2).

In contrast, IRF-1 expression was significantly associated with EGFR gene amplification (67.3% vs. 32.7% in IRF1-positive and IRF1-negative tumors, respectively; p = 0.04) (Additional file 1: Table S2). STAT1 expression positively correlated with IRF-1 expression (Pearson correlation = 0.36, p < 0.001), and a trend toward association was observed between STAT1 expression and TP53 loss of heterozygosity (p = 0.06) (Additional file 1: Table S2).

Importantly, patients with IRF-1-positive tumors showed a shorter mean overall survival than those lacking IRF-1 expression (406 vs. 521 days), although this difference did not reach statistical significance. Consistently, Kaplan-Meier analysis revealed a trend toward poorer survival in IRF1-positive patients (p = 0.06) (Fig. 1G). In contrast, STAT1-positive patients exhibited a markedly shorter mean overall survival than STAT1-negative patients (419 vs. 625 days; p = 0.012) (Additional file 1: Table S1). Accordingly, Kaplan-Meier analysis confirmed a significant association between STAT1 expression and poorer overall survival (p < 0.001) (Fig. 1H).

Together, these findings indicate that the IFN-γ pathway is significantly upregulated in patients with GBM and is associated with a worse clinical outcome, suggesting a relevant role of IFN-γ signaling in the pathogenesis of this disease.

IFN-γ reduces the proliferation and increases the apoptosis of GBM cells

To assess the functional consequences of IFN-γ signaling in GBM cells, we performed western blot analyses, which showed that the basal levels of STAT1 and IRF-1 proteins were markedly low in all GBM cell lines tested (LN-18, LN-229, and T98G cells). Nevertheless, the expression of both proteins and also STAT1 phosphorylation were strongly induced upon IFN-γ treatment (10 ng/mL) in all cell lines, indicating that the IFN-γ canonical pathway is functional and can be effectively activated in these cells (Fig. 2A). Higher concentrations of IFN-γ (> 100 ng/mL) were found to significantly reduce cell viability in a dose-dependent manner (Fig. 2B). In addition, IFN-γ stimulation at non-cytotoxic concentrations (10 ng/mL) for 24–72 h significantly inhibited cell proliferation (Fig. 2C). Notably, IFN-γ at this dose (10 ng/mL) enhanced the pro-apoptotic and cytotoxic effects of temozolomide, with T98G cells displaying the highest sensitivity whereas LN-18 cells were less responsive to the combined treatment (Fig. 2D).

Fig. 2.

IFN-γ mediates antiproliferative and proapoptotic effects in GBM. (A) Western blot analysis of IRF-1, total STAT1, and phosphorylated STAT1 in GBM cells, control or treated with 1 and 10 ng/mL IFN-γ for 48 h. β-Actin was used as a control. (B) Cell viability measured by MTS assay at 72 h after treatment with increasing doses of IFN-γ. (C) CFSE-based proliferation assay analyzed by flow cytometry at 24, 48, and 72 h after stimulation with 10 ng/mL IFN-γ (relative MFI ± SEM). (D) Apoptosis and cell death analysis by flow cytometry following treatment with IFN-γ alone or combined with temozolamide (TMZ) using DiOC and propidium iodide (PI) staining in GBM cell lines. ***p < 0.001, **p < 0.01, *p < 0.05

IFN-γ regulates multiple immune-related signaling pathways in GBM cells

To dissect the IFN-γ-driven transcriptional program in GBM, we performed RNA-seq-based differential gene expression profiling of IFN-γ-treated LN-18 and LN-229 cells and integrated these data with transcriptomic profiles from GBM patients (TCGA Firehose Legacy GBM cohort) stratified according to multiple independent IFN-γ-related gene signatures, with particular emphasis on the glioma-specific IFN-γ signature defined by Qian et al. [45]. (Fig. 3A). RNA-seq analysis revealed that after IFN-γ stimulation, 548 and 689 genes were significantly upregulated in LN-18 and LN-229 cells, respectively (log₂FC > 1, adjusted p < 0.01; Fig. 3B-C). Pathway enrichment analysis of the upregulated genes indicated a substantial overlap between both GBM cell lines, with the most enriched pathways related to immune response regulation, apoptosis, and EMT processes (Fig. 3D-E). The IFN-γ-induced expression of immunoregulatory molecules, including MHC class I and II, galectin-9, and PD-L1, identified by RNA-seq analysis, was validated by flow cytometry in both LN-18 and LN-229 cell lines (Fig. 3F-G). Notably, many of the genes induced by IFN-γ in GBM cell lines were also upregulated in GBM patients with high IFN-γ signature scores (Additional file 2: Fig. S2A-C). Pathway enrichment analyses revealed a remarkable overlap between the biological processes enriched in IFN-γ-stimulated GBM cell lines and those associated with high IFN-γ scores in patient tumors. In both cases, top enriched pathways included IFN-γ response, inflammatory response, TNFα/NF-κB signaling, IL6-JAK-STAT3 signaling, and adaptive immune resistance-related programs (Fig. 3D-E, Additional file 2: Fig. S2D-E). This robust intersection defines a core set of IFN-γ-responsive genes consistently upregulated across cell lines and patient tumors, confirming the clinical relevance of our in vitro findings. The shared signature included genes encoding classical and nonclassical MHC class I molecules (HLA-A, -B, -C, -E, -F), MHC class II molecules (HLA-DR, -DP, -DQ), adhesion molecules (CEACAM1) and costimulatory molecules such as TNFSF4 and CD40LG (Fig. 3H, Additional file 2: Fig. S2 F-G). Notably, GBM patients with high IFN-γ score also showed significant overexpression of multiple immunosuppressive molecules and immune checkpoints, such as HAVCR2 (TIM-3) and its ligand LGALS9 (galectin-9), CD274 (PD-L1), PDCD1LG2 (PD-L2), TNFRSF14 (HVEM) and its ligand TNFSF14 (LIGHT), CTLA4, TIGIT or IDO1 (Fig. 3H, Additional file 2: Fig. S2F-G).

Fig. 3.

IFN-γ signaling of GBM cells promotes an immunosuppressive phenotype and TME. A) Schematic overview of the experimental and in silico workflow employed to define the IFN-γ-dependent transcriptional program in GBM. B, C) Volcano plots displaying DEGs in IFN-γ-stimulated LN-18 (B) and LN-229 (C) cells compared to control conditions (RNA-seq analysis). Genes with log₂ fold change > 1 or <-1, and adjusted p < 0.01 were considered significantly differentially expressed. Upregulated DEGs are highlighted in red, and downregulated DEGs in blue. D, E) Pathway enrichment analysis of upregulated genes after IFN-γ-stimulation in LN-18 (D) and LN-229 (E) cells. Enrichment plots represent hallmark gene sets from the MSigDB Database, ranked by NES (-log₁₀(q value), with positive NES values indicating upregulation. F, G) Flow cytometric analysis of classical MHC class I molecules (HLA-A, -B, -C), MHC class II molecules (HLA-DR, DP-, DQ-), PD-L1 and LGALS9 expression in LN-18 (F) and LN-229 (G) cells following 48 h of IFN-γ stimulation (10 ng/mL). Histograms illustrate expression levels, while quantification graphs show the relative mean fluorescence intensity (MFI) normalized to control conditions (set as 1) (relative MFI ± SEM) from at least three independent replicates. H) Heatmaps showing the Spearman correlation between immune response-related gene set expression [61] or xCell 2.0-inferred immune cell infiltration and high IFN-γ signaling score in GBM patients stratified using the Qian et al. glioma IFN-γ signature (TCGA Firehose Legacy cohort).**p < 0.01, *p < 0.05

To further investigate how IFN-γ signaling influences the composition of the GBM TME, immune cell infiltration was estimated by deconvolution analysis of transcriptomic data from the TCGA Firehose Legacy GBM cohort. High IFN-γ score was associated with increased infiltration of myeloid cells and negatively associated with NK and CD8⁺ T cell infiltration (Fig. 3H, Additional file 2: S2F-G). Collectively, these findings support the biological relevance of the IFN-γ-driven transcriptional program identified in GBM cells and highlight a role for IFN-γ signaling in shaping an immunoregulatory landscape associated with an immunosuppressive TME in GBM.

IFN-γ-activated GBM cells establish a positive feedback loop that increases IFN-γ secretion and reduces NK cell cytotoxicity

Next, we analyzed the effect of IFN-γ signaling in the regulation of antitumor immunity. GBM cells pre-treated with IFN-γ (10 ng/mL for 48 h) or untreated control cells were co-cultured with PBMCs, and the secretion of key cytokines (IL-2, IL-4, IL-6, IL-10, IL-17, TNF-α, and IFN-γ) was assessed. Co-culture with IFN-γ-stimulated GBM cells led to increased production of several cytokines in a cell line-dependent manner (Additional file 2: Fig. S3A). Notably, IFN-γ was the only cytokine produced by immune cells whose levels were significantly increased by co-culturing with IFN-γ-stimulated GBM cell lines (Fig. 4A and Additional file 2: Fig. S3A). This induction of IFN-γ secretion was further confirmed by ELISA (Additional file 2: Fig. S3B-C). These findings suggest that interactions between IFN-γ-stimulated tumor cells and immune cells amplify IFN-γ production, potentially establishing a self-amplifying positive feedback loop that reinforces IFN-γ-driven signaling within the TME.

Fig. 4.

IFN-γ signaling in GBM cells amplifies IFN-γ production and impairs NK cell cytotoxicity. (A) Dot plot quantification of IFN-γ concentration in coculture supernatants by LEGENDplex™. (B) Flow cytometry analysis of GBM (ZsGreen+) and NK (Tag-it Violet+) cells, immune synapse formation. C, D) Flow cytometry analysis of CD107a expression in PBMCs cocultured for 4 h with control or IFN-γ-pretreated GBM cells (relative MFI ± SEM). Representative contour plots and histograms showing CD107a⁺ NK cells in cocultures with control (grey) or IFN-γ-pretreated (red) GBM cells. E, F) Specific lysis (%) of LN-18 (E) and LN-229 (F) GBM cells by NK cells after 4 h of coculture, assessed by flow cytometry using propidium iodide staining. Representative contour plots and histograms are shown; paired dot plots represent individual donor responses. G, H) NK cells isolated from PBMCs were treated with individual ICIs (anti-ILT2, anti-LAG-3, anti-PD-1, or anti-TIM-3) or left untreated, and cocultured for 4 h with GBM cells, either untreated or pre-stimulated with IFN-γ. The same analysis strategy as in panels (E-F) was applied. ***p < 0.001, **p < 0.01, *p < 0.05

To determine the effect of IFN-γ on GBM-NK cell interactions, immune synapse formation was analyzed. Pre-stimulation of LN-18 and LN-229 cells with IFN-γ significantly increased their adhesion to NK cells (p = 0.0078 and p = 0.0160, respectively) (Fig. 4B), consistent with the IFN-γ-induced upregulation of adhesion molecules and chemokines on tumor cells such as ICAM-1 or CXCL10 observed in our transcriptomic analysis (Additional file 2: Fig. S3D-E). However, despite this increased binding, CD107a degranulation assays and cytotoxic analysis showed that induction of IFN-γ signaling in GBM cells significantly reduced the NK cell-mediated degranulation and cytotoxicity (p = 0.0025 for LN-18 and p = 0.0046 for LN-229) (Fig. 4C-F).

Analysis of patient datasets revealed that several inhibitory immune checkpoints (including LILRB1, HAVCR2, LAG3, and PDCD1) positively correlated with the IFN-γ score as well as STAT1 and IRF-1 expression in patients with GBM (Additional file 2: Fig. S3F). Furthermore, elevated levels of these genes are associated with reduced survival of patients with GBM (Additional file 2: Fig. S3G), suggesting that multiple inhibitory checkpoint pathways induced by IFN-γ signaling may contribute to immune evasion in GBM. To functionally evaluate the contribution of these pathways, we next assessed the role of these immune checkpoints ligands, including MHC class I molecules (ligands of ILT2, encoded by LILRB1 gene), MHC class II molecules (ligands of LAG-3), PD-L1 (ligand of PD-1, encoded by PDCD1 gene), and Galectin-9 (ligand of TIM-3, encoded by HAVCR2 gene), in the reduction of NK cell cytotoxicity against in IFN-γ-stimulated LN-18 and LN-229 cells. Blockade of PD-1, TIM-3, or LAG-3 with monoclonal antibodies completely failed to restore NK cell cytotoxic activity against IFN-γ-stimulated LN-18 and LN-229 cells, and only ILT2 blockade partially restored NK cell cytotoxicity (Fig. 4G-H). These findings suggest that IFN-γ signaling induces a complex and redundant network of immune checkpoints and immunosuppressive pathways, and targeting a single inhibitory axis is unable to fully restore the antitumor immune response.

Inhibition of STAT1, but not IRF-1, restores the cytotoxic immune response against GBM cells

To dissect the specific role of STAT1 and IRF-1 in IFN-γ signaling, we generated CRISPR-Cas9 knockout (KO) models using two individual sgRNAs per target in LN-18 and LN-229 cells. Western blot analysis confirmed the effective abrogation of IFN-γ-stimulated IRF-1 expression in IRF-1 KO cells, and the reduction of STAT1 expression and phosphorylation in STAT1 KO cells (Fig. 5A-B). Notably, STAT1 KO also impaired IFN-γ-induced IRF-1 expression, whereas IRF-1 KO did not affect STAT1 phosphorylation (Additional file 2: Fig. S4A). Consistent with a broader disruption to IFN-γ signaling, STAT1 KO more efficiently suppressed the IFN-γ-mediated upregulation of HLA class I and PD-L1 molecules in comparison with IRF-1 KO (Additional file 2: Fig. S4B-C). Concordantly, IRF-1 KO failed to reverse IFN-γ-induced suppression of NK cell activity, as assessed by CD107a degranulation and cytotoxicity assays, whereas STAT1 KO fully restored NK cell cytotoxic function (Fig. 5C-F). Furthermore, IFN-γ also reduced the degranulation capacity of CD8 + T cells, although to a lesser extent than that observed in NK cells (Additional file 2: Fig. S4D). Consistently, STAT1 deficiency also rescued this impairment in CD8 + T-cell degranulation, further supporting a broader role for STAT1 in mediating IFN-γ-driven immune dysfunction (Additional file 2: Fig. S4D).

Fig. 5.

STAT1 is a key regulator of IFN-γ-mediated immune resistance mechanisms in GBM. A, B) Western blot analysis of IRF-1, total STAT1, and phosphorylated STAT1 in LN-18 (left) and LN-229 (right) cell lines. sgCTR, STAT1 knockout (STAT1 KO), and IRF-1 KO cells were treated with or without IFN-γ (10 ng/mL) for 24 and 48 h. C-F) Functional immune assays using GBM cells pre-treated for 48 h as described in (A), followed by co-culture with PBMCs (C, E) or NK cells (D, F) at a 5:1 effector-to-target (E: T) ratio for 4 h. NK cell degranulation was measured by CD107a surface expression (C, E), and cytotoxicity was assessed by propidium iodide (PI) staining (D, F), measured by flow cytometry. Graphs show the relative CD107a MFI (± SEM) or the percentage of specific lysis (%) (± SEM). G) Volcano plots displaying DEGs in IFN-γ-stimulated STAT1 KO cells versus IFN-γ-stimulated sgCTR cells in LN-18 and LN-229 cells. Genes with log₂FC > 1 and adjusted p < 0.01 were considered significantly differentially expressed. Upregulated genes are shown in red, downregulated in blue. Venn diagram showing the overlap of upregulated differentially expressed genes (DEGs) in IFN-γ-stimulated LN-18 (blue) and LN-229 (pink) cells, and downregulated DEGs in IFN-γ-stimulated STAT1 KO cells (yellow for LN-18, green for LN-229). H) Heatmap representation of selected STAT1-regulated genes from the overlapping gene set identified in (G). ***p < 0.001, **p < 0.01, *p < 0.05

To define the molecular mechanisms underlying this immune rescue, we analyzed the transcriptomic changes in STAT1-deficient cells. Transcriptional profiling revealed that STAT1-dependent genes were coordinately upregulated, encompassing multiple immune-modulating components (Fig. 5H). These include classical and non-classical HLA class I/II molecules and antigen-processing components (TAP1, TAP2, CD74, CIITA, among others). In addition, STAT1 controlled the expression of several cytokine- and chemokine-related genes, immunomodulatory molecules such as LGALS9, IDO1, and TNFRSF14, adhesion molecules including ICAM1 and MUC1, and canonical IFN-γ-responsive transcription factors such as IRF1 (Fig. 5H). Altogether, these findings establish STAT1 as a critical mediator of IFN-γ-driven immunosuppression in GBM, orchestrating a broad transcriptional program including antigen processing and presentation, immunoregulatory molecules and immune-interaction networks that may shape tumor recognition and compromise the activity of both innate and adaptive antitumor immune cells.

Pharmacological targeting of JAK1/2 with ruxolitinib boosts anti-GBM immunity

To determine whether pharmacological inhibition of upstream IFN-γ signaling could recapitulate the effects of STAT1 ablation, we next evaluated ruxolitinib, a clinically approved JAK1/2 inhibitor. Western blot analysis demonstrated that treatment with non-toxic doses of ruxolitinib (1 and 10 µM) completely abolished downstream JAK1/2 signaling in IFN-γ-stimulated LN-18 and LN-229 cells as evidenced by the reduction of IRF-1 expression as well as STAT1 expression and phosphorylation (Fig. 6A). Furthermore, flow cytometry analyses revealed that ruxolitinib treatment prevented IFN-γ-induced expression of inhibitory immune checkpoint ligands, including HLA class I and II and PD-L1, in both GBM cell lines (Fig. 6B-D). Moreover, cytotoxic assays demonstrated that ruxolitinib pretreatment completely restored NK cell-mediated cytotoxicity against IFN-γ-stimulated GBM cells (Fig. 6E).

Fig. 6.

Inhibition of IFN-γ-induced JAK-STAT signaling by ruxolitinib enhances GBM cell susceptibility to NK- and T cell- mediated immune responses. A) Western blot analysis of IRF-1, total STAT1, and phosphorylated STAT1 in LN-18 (left) and LN-229 (right) GBM cells treated for 48 h with IFN-γ (10 ng/mL) alone or in combination with ruxolitinib (1 or 10 µM). B-D) Flow cytometric assessment of surface expression of classical MHC class I (HLA-A, -B, -C), MHC class II (HLA-DR, -DP, -DQ), and PD-L1 in LN-18 (B) and LN-229 (C) cells under the same treatment conditions. Representative histograms (D) and quantification (B, C) of normalized mean fluorescence intensity (mean MFI ± SEM; untreated = 1) are shown. E) NK cell-mediated cytotoxicity assays using pretreated GBM cells co-cultured with NK cells at a 5:1 effector-to-target ratio (E: T) for 4 h. Cell death was quantified by PI staining and flow cytometry, shown as specific lysis (%) (mean ± SEM). F) Western blot confirmation of EGFRvIII expression in engineered GBM cell lines. GAPDH served as loading control. G-I) PBMC activation induced by αEGFRvIII;αCD3 BiTEs following co-culture with GBM cells at an E: T ratio of 25:1 for 48 h. (G) Graphical abstract illustrating αEGFRvIII;αCD3 BiTE-mediated recognition of EGFRvIII-expressing tumor cells and engagement of T cells via CD3. (H) Representative flow cytometry plots of CD69 expression in NK cells, CD8⁺ T cells, and CD4⁺ T cells following co-culture with EGFRvIII-expressing GBM cells. (I) Quantification of CD69 MFI ± SEM in immune cell subsets from all co-culture conditions, including control (empty vector) and EGFRvIII-expressing GBM cells. ***p < 0.001, **p < 0.01, *p < 0.05

To evaluate the functional consequences of STAT1 pathway inhibition by ruxolitinib across a broader range of immune effector populations, particularly T cells, we established a model that enabled antigen-specific recognition of GBM cells beyond NK cell-mediated responses. To this end, EGFRvIII-expressing LN-18 and LN-229 cell lines were generated and validated by Western blot analysis (Fig. 6F). We then produced BiTEs targeting EGFRvIII and CD3 (αEGFRvIII;αCD3, pBITE-EGFRvIII-F269-OKT3), allowing CD3⁺ T cells to specifically recognize and engage EGFRvIII-positive GBM cells and thereby trigger antitumor immune responses (Fig. 6G). Co-culture of PBMCs with EGFRvIII-expressing LN-18 or LN-229 cells in the presence of αEGFRvIII;αCD3 BiTEs resulted in robust immune cell activation, as evidenced by increased CD69 expression compared with control GBM cells lacking EGFRvIII expression (Fig. 6H-I). This effect was most pronounced in CD8⁺ and CD4⁺ T cells, highlighting antigen-specific EGFRvIII recognition and BiTE-mediated activation, whereas NK-cell activation was observed independently of EGFRvIII expression (Fig. 6I).

Importantly, ruxolitinib treatment further enhanced the activation of NK cells as well as CD8⁺ and CD4⁺ T cells, reversing the suppressive effects of IFN-γ on immune cell activation (Fig. 6H-I). These findings indicate that inhibition of the IFN-γ/STAT1 axis not only restores NK-cell function but also enhances the activation of adaptive immune populations, underscoring a broader role for STAT1 signaling in shaping antitumor immune responses in GBM and supporting pharmacological targeting of this pathway as a promising strategy to reinvigorate antitumor immunity.

Discussion

GBM is a highly aggressive malignancy characterized by extensive intratumoral heterogeneity, resistance to standard therapies, and a profoundly immunosuppressive TME, all of which contribute to its poor clinical outcome. This immunosuppressive environment is shaped by alterations in both myeloid and lymphoid compartments, including impaired cytotoxic immune responses and dysfunctional effector lymphocytes [2, 67–69]. In this study, we identify tumor-intrinsic IFN-γ signaling as a central regulator of immune suppression in GBM and demonstrate that sustained activation of this pathway promotes immune dysfunction through STAT1-dependent mechanisms.

The integration of transcriptomic and immunohistochemical analyses reveals that, compared with normal brain tissues, IFN-γ signaling is significantly upregulated in GBM and is associated with adverse clinical outcomes. These findings are consistent with prior reports linking high IFN-γ scores to immune exhaustion and poor ICI responsiveness across multiple cancers [40, 45, 49, 59, 70, 71]. In this context, our data support the use of IFN-γ-related signatures as potential prognostic and stratification biomarkers in GBM.

The underlying mechanisms implicated in the sustained activation of IFN-γ signaling in GBM have not yet been fully elucidated. Potentially, it may involve multiple mechanisms such as genetic alterations, epigenetic modifications, and crosstalk with other oncogenic signaling pathways such as EGFR and PI3K [51–53, 72, 73]. Supporting these associations, our immunohistochemical analyses showed that IRF-1 expression positively correlates with EGFR amplification. Furthermore, STAT1 expression is positively associated with TP53 loss of heterozygosity. Notably, both EGFR amplification and TP53 alterations are recurrent genetic events in GBM, as reported in large-scale genomic datasets such as TCGA [74]. Together, these findings suggest a potential convergence between IFN-γ signaling and major oncogenic and tumor suppressor pathways in GBM, which may reflect a functional interplay between oncogenic programs and immune regulatory mechanisms in this tumor type. In addition, it has been reported that GBM cells might produce and respond to IFN-γ, creating an autocrine loop that sustains or amplifies pathway activity [75]. In line with this, we observed that interaction between immune cells and IFN-γ-primed GBM cells led to enhanced IFN-γ production, thus supporting the involvement of this autocrine loop in amplifying IFN-γ signaling. Nevertheless, the complete elucidation of the underlying mechanisms warrants further investigation.

Importantly, IFN-γ signaling has been reported to exert a dual and context-dependent role in several cancer types [30, 76]. While IFN-γ can induce direct antitumor effects on GBM cells by reducing proliferation and inducing apoptosis [41, 45–49], our data indicate that it simultaneously induces a potent immunosuppressive program in tumor cells, characterized by the upregulation of multiple inhibitory molecules, including IDO1, PD-L1, and ligands for receptors such as PD-1, ILT-2, LAG-3, and TIM-3, ultimately abrogating antitumor immune cytotoxicity. This immunosuppressive state is further shaped by profound alterations in the TME. IFN-γ-high tumors are associated with increased infiltration of myeloid populations and reduced cytotoxic lymphocyte activity, with CD8⁺ T cells or NK cells displaying an exhausted phenotype and limited effector function. Collectively, these findings suggest that although IFN-γ can activate antitumor programs at the tumor-cell level, this is insufficient to overcome the dominant immunosuppressive TME in GBM. This complex balance may also explain the limited efficacy of ICIs in this disease. Consistently, large phase III clinical trials evaluating PD-1/PD-L1 or CTLA-4 blockade have failed to demonstrate significant survival benefit as monotherapy in most GBM patients [22–24], likely due to low cytotoxic lymphocyte infiltration and strong immunosuppressive signals [77–81]. In agreement with this, we show that individual blockade of PD-1, LAG-3, or TIM-3 does not restore NK cell-mediated cytotoxicity against GBM cells, and only ILT2 blockade showed a significant effect, supporting the need to target upstream regulatory mechanisms to fully reinstate antitumor immunity.

STAT1 is a crucial transcription factor that acts as a downstream central effector of IFN-γ and a driver of tumor immunosuppression, progression, invasion, and poor prognosis in GBM [82–85]. Our results show that genetic deletion of STAT1 reverses the IFN-γ-induced immunosuppressive program and restores antitumor immune activity, whereas IRF-1 silencing fails to rescue immune cytotoxicity. These findings position STAT1 as a critical mediator of IFN-γ-driven immune escape and a potential therapeutic vulnerability in GBM.

In this context, the JAK1/2 inhibitor ruxolitinib, already in clinical use, has emerged as a potential therapeutic candidate. Our data indicate that treatment with non-toxic doses of ruxolitinib effectively suppressed IFN-γ-induced signaling and prevented IFN-γ-mediated upregulation of multiple immunosuppressive ligands. Functionally, ruxolitinib pretreatment restored NK- and T- cell-mediated cytotoxicity against IFN-γ-stimulated GBM cells. These findings demonstrate that IFN-γ-driven immune suppression in GBM is reversible and critically dependent on intact JAK1/2-STAT1 signaling. Consistently, JAK inhibition has shown potential to enhance ICI efficacy in multiple ICI-resistant tumor types [51–53, 86–88], and preliminary evidence in recurrent GBM further supports its translational relevance [19, 89].

Taken together, our findings identify STAT1 as a key mediator of IFN-γ-driven immune evasion in GBM, orchestrating the coordinated upregulation of multiple immunosuppressive pathways. The ability of pharmacological JAK1/2 inhibition to reverse this phenotype and restore antitumor immune activity highlights the therapeutic relevance of the IFN-γ-STAT1 axis. While further in vivo investigation is required, our findings underscore the relevance of exploring the therapeutic potential of JAK inhibitors in tumors such as GBM, where IFN-γ signaling contributes to profound immunosuppression, poor survival outcomes, and low response to ICI-based immunotherapy.

Conclusions

Our study identifies IFN-γ signaling as a key driver of immunosuppression and poor prognosis in GBM. Activation of the IFN-γ/STAT1 axis promotes multiple immune evasion mechanisms, limiting the efficacy of single-checkpoint blockade therapies. Importantly, in vitro genetic or pharmacological inhibition of this pathway reverses IFN-γ-mediated immunosuppression. These findings provide a mechanistic framework that warrants further in vivo validation and open new avenues for therapeutic strategies aimed at improving patient stratification and immunotherapy efficacy in GBM.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

GBM

Glioblastoma

IFN-γ

Interferon-γ

PBMCs

Human peripheral blood mononuclear cells

NK

Natural killer cells

TILs

Tumor-infiltrating lymphocytes

ICIs

Immune checkpoint inhibitors

TME

Tumor microenvironment

TCGA

The Cancer Genome Atlas

HUCA

Hospital Universitario Central de Asturias

KO

Knock-out

EMT

Epithelial-to-mesenchymal transition

Author contributions

Conceptualization: C. A-G, F. H-P, S. G. Investigation: C. A-G, A. M-P, C. S-B, R. G-D, F. H-P, S. G. Resources: I. F-V, J M. G-P Formal Analysis: C. A-G, I. F-V, J M. G-P, F. H-P, S. G. Original Draft Preparation: C. A-G, F. H-P, S. G. Supervision: F. H-P, S. G. Funding acquisition: J M. G-P, C. S-B, S. G. All authors read and approved the final manuscript.

Authors’ social media handles

The Tumor Immunology Group is present on social media X: @immunofighters,

BlueSky: @immunofighters.bsky.social.

Funding

This work was supported by a Spanish grant from Instituto de Salud Carlos III (PI23/01576 and PI22/00167) cofounded by the European Union; by the Government of the Principality of Asturias through the Agency for Science, Business Competitiveness, and Innovation of the Principality of Asturias and co-financed by the European Union through the Grants for Research Groups of Organizations of the Principality of Asturias, under file number IDE/2024/000745 and IDE/2024/000778, Asociación Xti Xtod@s, Fundacion LAIR and “Programa de ayudas Fundación Alimerka a la investigación en Ciencias de la Salud”. C.A-G is a recipient of the 2022 predoctoral grant from the Scientific Foundation of the Spanish Association Against Cancer (PRDAS222516AGUI), A.M-P is a recipient of the 2022 ISPA intramural predoctoral contract (ITM-PREDOC-22), R.G-D is a recipient of Intramural ISPA-Janssen grant and postdoctoral contract from Instituto Universitario de Oncología del Principado de Asturias and F.H-P is a recipient of Miguel Servet contract from ISCIII (CP24/00064). The funding agencies had no involvement in the design of the study, data collection, analysis, interpretation, manuscript writing, or the decision to submit the article for publication.

Data availability

The original contributions presented in this study are included in the article and its additional material. Further inquiries can be directed to the corresponding authors. RNA-sequencing data are available in the Gene Expression Omnibus (GEO) repository under accession number GSE341302.

Declarations

Ethics approval and consent to participate

The study was approved by the institutional review board of the Principado de Asturias ethics committee (CEImPA 2023.5) according to the Declaration of Helsinki.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.