For LN229 cell replication, adhere the cells to glass coverslips until they reach 30% confluency, then fix them with 4% formaldehyde in phosphate-buffered saline (PBS) for 10 minutes at room temperature. After washing with PBS, increase membrane permeability by treating with 0.1-0.5% Triton X-100 in PBS for 5 minutes. Block nonspecific binding by incubating with 5% bovine serum albumin (BSA) or serum for 30 minutes. Next, incubate with the primary antibody, diluted in blocking solution, for 1 hour at room temperature or overnight at 4°C. After washing with PBS, a fluorescent secondary antibody was applied for 1 hour in the dark, followed by a final PBS wash (41–43). Mount the coverslip and examine the cellular signals using a fluorescence microscope (44, 45).

Apoptosis was assessed using the Annexin V-FITC kit from BD Biosciences, USA. Cells were incubated with Annexin V-FITC for 15 minutes, followed by a 5-minute incubation with propidium iodide (PI), both in the dark. Flow cytometric analysis was performed with BD Biosciences equipment, and data were analyzed using FlowJo software.

Cell invasion was evaluated using Matrigel-coated Transwell inserts. LN229 cells (5 x 10^5 cells/ml) were seeded in the upper compartment and incubated for 36 hours at 37°C in 5% CO₂. After incubation, cells adhering to the upper membrane were fixed with 4% formaldehyde, stained with crystal violet, washed with PBS, and examined microscopically. The invasion was quantified by counting cells that migrated through the membrane in five random fields (46). A wound-healing assay was conducted to study the effect of disulfidptosis-Tex on cell migration. A scratch was made in the monolayer at 0 hours, detached cells were removed with PBS, and images were taken after 36 hours for analysis (47).

The mitochondrial membrane potential (Ψm) was evaluated using the ‘Ψm Assay JC-1 Kit’ (Solarbio, M8650, China), employing JC-1 as a fluorescent probe. When the membrane potential is high, JC-1 accumulates within the mitochondrial matrix, leading to the emission of red fluorescence. Conversely, at reduced potentials, JC-1 forms monomers that emit green fluorescence.

The Reactive Oxygen Species (ROS) levels were measured using the Reactive Oxygen Species Assay Kit (Solarbio, CA1410, China) with DCFH-DA as the fluorescent probe. ROS converts the non-fluorescent DCFH to fluorescent DCF, which is then analyzed to determine intracellular ROS concentrations.

Transcriptomic and comprehensive clinical data for the TCGA-GBM cohort were sourced from the GDC portal (https://portal.gdc.cancer.gov/). The study focused on entries that provided both extensive clinical records and transcriptomic data (48, 49). Additionally, the CGGA database was used for whole-genome expression profiles with corresponding clinical information for GBM (50).

Using Seurat package version 4.2.0, the pre-filtered single-cell dataset was imported (51, 52). Data normalization was performed using the ‘NormalizeData’ function. Post-normalization, genes with significant variation were identified by balancing average expression levels and dispersion metrics. The ‘FindClusters’ function, a graph-based clustering tool using a modularity optimization algorithm from shared nearest neighbors, delineated 19 distinct clusters from 33 principal components at a resolution of 0.2. Differentially expressed genes (DEGs) in each cluster were determined using ‘FindAllMarkers’ with default settings in Seurat.

Cell Communication Profiling via single-cell analysis ligand-receptor interactions among various cell types were analyzed to identify unique signaling pathways (53). The ‘CellChat’ tool quantified and estimated the probability of intercellular signaling interactions, applying default parameters with a significance threshold of P ≤ 0.05 and adjustments for multiple testing using the Benjamini-Hochberg procedure (54). We also assessed the expression of ten disulfidptosis-Tex-related genes across different glioblastoma cell types using the AUCell scoring method (55, 56). Twelve cell types identified in the scRNA data were analyzed, categorizing cells with AUCell scores above 0.3 as high disulfidptosis-Tex activity and those with lower scores as reduced activity.

Disulfidptosis-Tex’s role in glioblastoma was assessed using ‘ssGSEA’ to calculate gene enrichment scores in individual samples (57–59). Using ‘surv_cutpoint’, samples were categorized into low or high disulfidptosis-Tex enrichment groups. Intersection analysis identified cell types with significant disulfidptosis-Tex activity and contrasted gene enrichment between the groups. T cell exhaustion-linked DEGs in GBM were identified. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses with ‘clusterProfiler’ revealed pathways enriched among these DEGs (60–62).

The prognostic relevance of disulfidptosis-Tex-associated DEGs in glioblastoma was evaluated using univariate Cox regression to analyze their correlation with patient overall survival (OS) (63–65). Genes with a P-value < 0.05 were selected for further analysis (66). The study used 268 tumor samples, split into training and validation cohorts in a 7:3 ratio, with 106 samples in the latter. A prognostic model was developed using the LASSO Cox regression method via the ‘glmnet’ R package (67, 68), refining the list of potential genes.

Differential expression between high- and low-risk glioblastoma groups was analyzed using the ‘limma’ R package (69, 70). Gene Set Enrichment Analysis (GSEA) of log2 FC-ranked genes was performed with ‘clusterProfiler’ (71–73), and functional differences were examined using ‘GSVA’ (74, 75). Results were visualized with ‘pheatmap’.

ssGSEA analyses were performed using the ‘gsva’ package in R to evaluate immune pathway activities in the study’s samples, utilizing established molecular markers (76). As an enhancement of GSEA, ssGSEA calculates enrichment scores for individual gene set pairs across different samples (77). These scores reflect the coordinated regulation of genes, either upregulated or downregulated, within specific gene sets for each sample. Gene expression data from all GBM samples were used to compute enrichment scores for 28 distinct immune cell types, derived from the TISIDB database (http://cis.hku.hk/TISIDB/index.php) (78). Variations in immune cell infiltration levels between low- and high-risk groups were visualized using the ‘ggplot2’ R package (79, 80).

Statistical analyses were performed using R software (version 4.1.3) and GraphPad Prism 8.0. A P-value of <0.05 was considered statistically significant (81). In the graphs, the symbols *, **, and *** represent P-values of <0.05, <0.01, and <0.001, respectively.

Single-cell RNA sequencing has significantly enhanced our understanding of the cellular composition of glioblastoma. In this study, we utilized the single-cell RNA sequencing dataset GSE173278 to explore the genes exhibiting elevated expression linked to disulfidptosis-Tex ( Figure 1 ). The analysis of single-cell transcriptomes revealed 19 distinct clusters across 29,543 cells, as shown in the UMAP plot ( Figure 2A ). Cell surface markers were also used to identify 12 unique cell subtypes, including dendritic cells, central memory T cells, and macrophages ( Figures 2B, D ). Further investigation focused on disulfidptosis-Tex-associated differential expression genes (DEGs) expression patterns within these subtypes ( Figure 2C ).

Figure 3A illustrates the identification of 205 cells exhibiting active disulfidptosis-Tex. The UMAP representation of these cells revealed a significant prevalence of endothelial cells (ECs) ( Figure 3B ). Notably, ECs showed the strongest association with disulfidptosis-Tex, indicating a marked accumulation of this marker within tumor-associated ECs.

Pseudotime analysis of endothelial cells (ECs) in glioblastoma revealed the key role of disulfidptosis-Tex genes in tumor progression ( Figure 3C ). Five distinct transcriptional states were identified along the trajectory ( Figure 3D ). Further analysis of these genes showed their involvement in ‘angiogenesis regulation’ and ‘extracellular matrix organization’ ( Figure 3E ).

To investigate the roles of various cell populations in glioblastoma, we conducted an analysis of intercellular communication, which revealed significant interactions between glioblastoma cells and immune cells, including central memory T cells and macrophages ( Figure 4A ). Both outgoing and incoming signals were examined, alongside relevant ligand-receptor pairs across 12 distinct cell types. Key signaling pathways identified in this analysis included SPP1, PTN, MK, PSAP, GRN, and MIF ( Figure 4B ). Further analysis of the signaling pairs highlighted the PSAP-GPR37 pathway as the predominant interaction in endothelial cells, facilitating communication with oligodendrocytes and dendritic cells ( Figure 4C ). Signaling pathways from oligodendrocytes and macrophages to endothelial cells were also identified ( Figure 4D ). PSAP was found to be expressed across all 12 cell types ( Figure 4E ), while GPR37L1 was predominantly present in corticotroph cells, and GPR37 showed the highest expression in oligodendrocytes ( Figure 4F ).

In glioblastoma, endothelial cells with the highest disulfidptosis-Tex activity were characterized by 3,890 differentially expressed genes (DEGs) compared to other cell types ( Figure 5A ). A comparison between glioblastoma tissues and normal controls revealed 2,200 DEGs, with a heatmap showing the top five upregulated and downregulated genes ( Figure 5B ). Stratification based on disulfidptosis-Tex activity identified an additional 4,255 DEGs, with the top five genes in the high-activity group also highlighted ( Figure 5C ). An intersection analysis of these groups identified 143 key genes ( Figure 5D ). Enrichment analysis focused on Gene Ontology (GO) and KEGG pathways. GO analysis showed significant enrichment in processes such as synapse organization and urogenital system development, along with molecular functions like glycosaminoglycan binding and ECM structural components ( Figure 5E ). KEGG pathway analysis revealed notable enrichment in ECM-receptor interactions ( Figure 5F ).

A LASSO-Cox regression analysis was performed to assess the impact of disulfidptosis-Tex genes on glioblastoma survival, resulting in a model consisting of 11 key genes ( Figure 5G ). This model successfully stratified patients into high- and low-risk groups, with the high-risk group demonstrating higher mortality rates and upregulated prognostic genes ( Figure 5H ). Kaplan-Meier survival curves further confirmed that the high-risk group had a worse prognosis compared to the low-risk group ( Figure 5I ). The model’s ability to predict patient outcomes was evaluated using ROC curves, achieving AUC values of 0.780 for 5-year survival in the training set ( Figure 5J ) and 0.841 in the validation set ( Figure 5K ). Notably, SMC4 showed a strong negative correlation with T cell exhaustion genes ( Figure 5L ).

We investigated the role of SMC4, a key gene in the disulfidptosis-Tex-related prognostic model, in the migratory behavior of glioblastoma cells. Wound-healing assays showed that silencing SMC4 reduced cell migration at 36 hours compared to controls ( Figures 6A, B ). While transwell assays at 24 hours showed no significant differences between the SMC4-silenced and control groups ( Figure 6C ), at 48 hours, the migration of SMC4-knockdown cells was significantly lower than in the control group ( Figure 6D ).

Gene Set Enrichment Analysis (GSEA) was used to explore the underlying mechanisms of the 11-gene disulfidptosis-Tex model. The results revealed significant enrichment in six key pathways, including the T cell receptor signaling pathway, in the high-risk subgroup ( Figure 7A ). Additionally, GSVA was performed using the same MsigDB pathway data, identifying five pathways with the most pronounced differences between high- and low-risk subgroups. A heatmap of these variations revealed a notable enrichment of drug metabolism processes in glioblastoma patients with low disulfidptosis-Tex activity ( Figure 7B ). These findings suggest that disrupting disulfidptosis-Tex could influence drug metabolism and sensitivity in glioblastoma cells.

The relationship between immune cell infiltration and tumor progression was demonstrated by analyzing 28 immune cell types in both high- and low-risk subgroups, revealing a link with disulfidptosis-Tex ( Figure 8A ). Notably, activated and central memory CD8 T cells showed a significant negative correlation ( Figure 8B ). Immune cell infiltration differed markedly between risk groups, especially in regulatory T cells ( Figure 8C ). Additionally, strong correlations were found between prognostic genes and specific immune cells ( Figures 9A-I ). SMC4 was positively correlated with activated CD4 T cells ( Figure 9D ) and negatively correlated with CD56dim NK cells ( Figure 9E ). These findings suggest that disulfidptosis-Tex influences the infiltration of CD56dim NK cells and various T cell subsets, emphasizing the role of these genes in modulating tumor microenvironment interactions.

To better understand the prognostic value of the risk score signature in predicting patient outcomes, we examined mutations in glioblastoma-specific genes, focusing on the 20 most frequently mutated genes. TP53 mutations were the most common across both subgroups, followed by PTEN mutations ( Figures 10A, B ). We also evaluated whether risk scores could predict chemotherapeutic responses in glioblastoma patients. Clinical trials were performed testing drugs such as Dinaciclib, Bortezomib, and Docetaxel ( Figure 10C ). The results indicated that patients with higher risk scores exhibited increased sensitivity to these drugs, suggesting that they may be promising treatment options for high-risk glioblastoma patients.

We also evaluated drug sensitivity using Dactinomycin, Bortezomib, and Docetaxel. The results showed that combining si-SMC4 with these drugs significantly reduced the invasiveness of LN299 cell lines compared to single-agent treatments ( Figures 11A, B ), highlighting the role of disulfidptosis-Tex in glioma metastasis. Interestingly, all the drugs tested were found to upregulate PD-L1 expression in LN299 cells ( Figure 11C ). Further investigation revealed a significant increase in PD-L1 expression when LN299 cells were treated with the si-SMC4-drug combination ( Figure 11D ).

To evaluate mitochondrial dynamics, LN299 cells were stained with JC-1 and divided into two groups: control and SMC4-interfered. Flow cytometry analysis revealed a significant reduction in mitochondrial membrane potential in the SMC4-interfered cells compared to controls ( Figure 12A ). Additionally, there was a marked increase in reactive oxygen species (ROS) accumulation in these cells ( Figure 12B ). These results suggest that SMC4 plays a regulatory role in maintaining mitochondrial integrity.

Our analysis showed a positive correlation between SMC4 expression in gliomas and CD4 T cell activation. To explore this further, we co-cultured SMC4 knockdown LN299 cells and control cells with T cells. The results indicated that LN299 cells with high SMC4 expression significantly increased the proportion of CD4 T cells ( Figures 13A-C ). However, SMC4 depletion led to a 13.3% reduction in CD4 T cell activation ( Figures 13D-F ) and a 9.62% decrease in CD8 T cell activation ( Figures 13G-I ). These findings suggest that SMC4 expression in glioma cells influences T cell cytotoxicity.