LncRNA SOX21-AS1 is associated with poor prognosis and immunomodulation in glioblastoma

Introduction

Glioblastoma multiforme (GBM) is the most aggressive type of astrocytic glioma in terms of aggressiveness, with a poor prognosis (1). The clinical management of GBM necessitates an integrated approach that combines surgical intervention, radiotherapy, and chemotherapeutic regimens. Despite this multifaceted strategy, the therapeutic efficacy remains limited. A considerable number of patients with glioma experience extensive intracerebral invasion and recurrence following conventional oncological treatments, while distant metastasis of glioblastoma is extremely rare; the extensive intracerebral invasion makes safe complete surgical resection often impossible and is almost always lethal (2). Immunotherapy, targeted therapy, and other novel therapies have recently received increasing attention. Targeted immunotherapy has shown significant efficacy in delaying GBM progression and improving the median survival time while exhibiting good tolerance (3-6). Currently, there is a paucity of GBM-specific therapeutic targets. Consequently, the identification of efficacious therapeutic targets for GBM is critical to enhance patient prognostic outcomes and treatment efficacy.

Long non-coding RNAs (lncRNAs) are non-coding transcripts that have a length exceeding 200 nucleotides. These molecules are implicated in a diverse array of biological processes, including cellular proliferation, migration, and apoptosis. Moreover, lncRNAs are fundamental in the initiation and progression of numerous cancer types, playing a pivotal role in tumorigenesis (7,8). The regulation of gene expression levels is governed by various biological mechanisms, such as chromatin remodeling and cell cycle control (9). Additionally, lncRNAs influence the clinical features and outlook of individuals with GBM and are linked to how well patients respond to immunotherapy and their ability to resist radiotherapy, which is essential in deciding the results of treatment. In addition to serving as a possible diagnostic marker, lncRNAs could also serve as promising targets for drugs in the therapy of GBM (10-12). However, GBM-specific reliable diagnostic/prognostic biomarkers remain unknown.

This study hypothesizes that specific lncRNAs play an important role in the prognosis and immune regulation of GBM, and may serve as potential biomarkers. To verify this hypothesis, this study used bioinformatics methods to screen lncRNAs associated with GBM prognosis, built GBM prognostic risk prediction models based on the screened lncRNAs, and analyzed the mechanism of action of key lncRNAs, especially their potential roles in immune regulation. Through these analyses, we expect to provide new insights into the diagnosis and treatment of GBM. We present this article in accordance with the TRIPOD reporting checklist (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1-2684/rc).

Methods

Data acquisition

GBM gene expression data from The Cancer Genome Atlas (TCGA) were obtained from the University of California, Santa Cruz (UCSC) Xena database at https://xenabrowser.net/. The RNA-Seq data, which included 167 tumor tissues and 5 normal tissues, were converted to the log₂[fragments per kilobase of transcript per million mapped reads (FPKM) + 1] format. Furthermore, TCGA TARGET Genotype-Tissue Expression (GTEx) data were downloaded from the UCSC Xena database to extract brain tissue information. The gtf annotation file was downloaded from GENCODE (https://www.gencodegenes.org). The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Differentially expressed genes (DEGs)

The DEGs were identified using the “limma” R-package filter based on TCGA-TARGET-GTEx’s lncRNA expression matrix; Genes with P<0.05 and log₂fold change (FC) greater than 1-fold were selected as DEGs. Volcano plots were drawn for DEGs. Patient survival information for GBM was acquired from the UCSC Xena database. Through one-way Cox regression analysis, lncRNAs linked to GBM prognosis were discovered, considering those with P<0.05 as significant for prognosis. The intersection of GBM-associated differentially expressed lncRNA encoding genes with GBM prognosis-associated lncRNA encoding genes was used to identify GBM-associated differentially expressed prognostic lncRNA encoding genes.

Development and verification of a predictive model

After obtaining GBM-associated differentially expressed prognostic lncRNA encoding genes, we used the “glmnet” R package for least absolute shrinkage and selection operator (LASSO) model construction. The analysis was performed with a fixed random seed (2021) and Cox regression type, incorporating 10-fold cross-validation repeated for 10 iterations to enhance model generalizability. A penalty parameter (λ) was implemented to minimize overfitting by reducing partial likelihood deviance. Model behavior was visualized through variable trajectories and risk distribution patterns. The optimal λ value (λ.min) was determined based on the minimum cross-validated error, while λ.1se corresponded to the most parsimonious model within one standard error of the optimal performance. Then, for each candidate factor in the constructed model, a proportional hazards (PH) test and variance inflation factor (VIF) calculation were performed using the coxph function to assess collinearity. Factors that passed both the PH test and the VIF collinearity test were further modeled using quadratic terms. The accuracy and discriminative ability of the prediction model were evaluated using the Concordance Index and time-dependent receiver operating characteristic (ROC) curves. Risk scores (RS) were calculated according to the following formula:

RiskScore=∑iCoefficient(genei)∗mRNAExpression(genei)[1]

Patients were categorized into high- and low-risk groups based on the RS threshold corresponding to the maximum area under the curve (AUC) value (0.95) derived from time-dependent ROC analysis.

Prognostic impact of risk modeling on GBM

To calculate the overall survival (OS) of patients in both high- and low-risk groups, we used the “survive” R package and Kaplan-Meier analysis, and statistical significance was determined at P<0.01. DEGs were analyzed and heat maps were plotted for high and low risk groups. Furthermore, Cox multivariate analysis was conducted to analyze risk factors and identify independently predicted risk factors. Additionally, a forest plot was created.

Gene set enrichment analysis (GSEA)

GSEA on the high- and low-risk groups was performed using GSEA software (4.2.0) and Kyoto Encyclopedia of Genes and Genomes (KEGG), Gene Ontology (GO), and ImmuneSigDB datasets in Molecular Signatures Database (MSigDB). In addition, we examined the differences in biological functions between high- and low-risk groups, which were statistically significant at P<0.05.

Analyzing immune infiltration and immune checkpoints

In our study, we analyzed immune infiltration in high- and low-risk GBM patients using CIBERSORT and single‑sample gene set enrichment analysis (ssGSEA) algorithms, identifying immune cells that were differentially enriched. Moreover, we analyzed the correlation between risk factors and infiltration scores of different immune cells.

Identification of immune-related risk factors, prediction of target genes, and functional enrichment analysis

Our study used the ImmPort database to identify immune-associated genes that are correlated with risk factors. We analyzed TCGA-GBM transcriptome data using the co-expression method with set values (P<0.001 and Cor coefficient >0.6) to determine the expression correlation between lncRNAs and immune-associated genes. Additionally, intersections were incorporated with risk factors to obtain immune-related hub genes. Next, we conducted Kaplan-Meier prognostic analyses utilizing the R packages ‘survival’ and ‘survminer’. Next, the MEM web tool (https://biit.cs.ut.ee/mem/index.cgi#tutorial) was used to predict the target genes affected by these immune-related hub genes. The STRING (https://cn.string-db.org/) web tool was utilized for a proton-pump inhibitor (PPI) analysis, while the ‘clusterProfiler’ R package was used for GO and KEGG analyses. Statistical significance was found to be below 0.05, indicated as P<0.05.

Drug sensitivity analysis

Genomics of Drug Sensitivity in Cancer (GDSC) is the largest public repository for molecular markers of drug sensitivity and drug response in cancer cells, available at www.cancerRxgene.org. The pRRophetic algorithm was applied to predict the half-maximal inhibitory concentration (IC50) of 20 common anticancer drugs for each patient, and the Mann-Whitney U test was performed to compare the differences in IC50 values between the high- and low-risk groups. GSCALite is a web-based gene set cancer analysis platform that integrates cancer genomics data for 33 cancer types from TCGA, drug response data from GDSC and Cancer Therapeutics Response Portal (CTRP), and normal tissue data from GTEx into a full data analysis workflow for gene set analysis. We used the GSCALite platform to analyze the correlation between the expression of the 19 hub lncRNAs and drug sensitivity in the GDSC and CTRP databases, so as to verify their potential value as drug targets for GBM treatment.

Statistical analysis

R programming (version 4.1.1) was utilized for all statistical analyses. Students’ t-tests were used to assess the statistical significance of variables that follow normal distributions for two sets of continuous variables. Mann-Whitney U tests have been used to identify differences between variables that do not follow a normal distribution, and ROC curves have been plotted in R using the pROC package, calculating the AUC to determine the accuracy of the RS in predicting prognosis. Statistically significant effects were considered to be those with a P value of 0.05 or less.

Results

Acquisition of DEGs

The flow chart provides a concise summary of the construction process for the risk signal model associated with GBM prognosis in this study, as well as the subsequent validation methodologies (Figure 1A). Our study examined how lncRNA gene expression levels differ between GBM tissues and normal tissues using TCGA-TARGET-GTEx data. This dataset consisted of 689 tumor tissue samples and 1,153 normal tissue samples, encompassing 60498 genes (Table 1). Utilizing the lncRNA annotation from GENCODE, we identified 15,098 lncRNAs. Differential expression analysis was conducted using the limma R package, with a significance threshold set at a P value less than 0.05 and a log₂FC cutoff greater than 1. Applying these criteria, we detected 2,493 differentially expressed lncRNA genes, of which 1,076 were down-regulated and 1,417 were up-regulated in the dataset. Next, we identified 1,018 prognosis-related lncRNAs using one-way COX regression analysis and plotted volcano maps (Figure 1B) in GBM patients. We selected the intersection of 2,493 GBM-associated DEGs with 1,018 GBM prognosis-associated genes to obtain 217 glioblastoma-associated differentially expressed prognostic genes (Figure 1C).

Figure 1 Differentially expressed genes, genes associated with prognosis, and the establishment of LASSO models. (A) Flow chart of this study. (B) Volcano plot of the DEGs. (C) DEGs in TCGA are depicted as blue circles. Yellow circles indicate prognosis-related genes obtained from Cox regression analysis of TCGA, and the genes at the intersection are selected to obtain GBM-related differentially expressed prognostic genes. (D,E) LASSO-based regression models for differentially expressed prognostic genes; different colored lines represent various possible models. DEG, differentially expressed gene; GBM, glioblastoma multiforme; GSEA, gene set enrichment analysis; K-M, Kaplan-Meier; LASSO, least absolute shrinkage and selection operator; lncRNA, long non-coding RNA; ROC, receiver operating characteristic; TCGA, The Cancer Genome Atlas; UCSC, University of California, Santa Cruz.

Construction of risk prognostic models

Based on LASSO-Cox regression analysis and multifactor regression analysis on 217 differentially expressed lncRNAs (Figure 1D,1E), we identified 19 hub genes that significantly influenced prognosis in GBM patients (Table 2). Using these 19 hub genes, a risk model was constructed and patients were classified as high- and low-risk (Figure 2A). Mortality in high-risk groups was positively associated with RS, but not in low-risk groups (Figure 2B). Heat maps were also generated to show the differences between high- and low-risk expression patterns of the 19 hub genes (Figure 2C). In addition, the Kaplan-Meier analysis revealed a significant difference between patients at high- and low-risk regarding OS (Figure 2D). Notably, the continuous variable Cox regression multifactor analysis of these 19 hub genes with a forest plot showed that ENSG00000204802, RARA_AS1, LINC01956, CNIH3_AS1, LINC01127, HOXC13_AS, MIR1_1HG_AS1 and ENSG00000259869 were independent prognostic factors (Figure 2E). The genes RARA_AS1 and LINC01956, which had the most significant P value, were chosen and their expression levels were measured in both the high- and low-risk groups. In the high-risk group, a strong association between poor prognosis and expression of both genes was observed (Figure 2F,2G).

Figure 2 Risk modeling and prognosis analysis. (A) ROC curves. (B) Scatterplot of risk scores. (C) Heatmap displays the levels of expression for 19 central genes in the high- and low-risk groups. (D) Performing prognostic analysis by utilizing the Kaplan-Meier curve for the groups categorized as high- and low-risk. (E) Forest plot showing the results of the multifactorial analysis of the 19 hub genes. (F) Expression levels of the risk factor RARA-AS1 in both the high- and low-risk groups. (G) Expression levels of the risk factor LINC01956 in both the high- and low-risk groups. *, P<0.05; **, P<0.01; ***, P<0.001. AIC, Akaike information criterion; AUC, area under the curve; ROC, receiver operating characteristic.

GSEA

To further elucidate the potential biological functions and major signaling pathways of DEGs in the high- and low-risk groups, we performed GO function and KEGG pathway analyses. In biological processes, genes in the high-risk group were prominently enriched in immune regulatory pathways, including heightened responsiveness to interleukin-4, negative regulation of αβ T cell differentiation, and positive regulation of CD4⁺ αβ T cell differentiation (Figure 3A). Cellular component analysis highlighted the involvement of DEGs in lipid metabolism-related organelles and DNA replication machinery (Figure 3B). Molecular function annotations further demonstrated enrichment in TNF receptor superfamily binding, transcriptional regulation via RNA polymerase II-associated factors, and apoptosis-related activities (Figure 3C).

Figure 3 Gene enrichment analysis. (A-C) GSEA-GO analysis. (D) GSEA-KEGG analysis. (E) GSEA-immune correlation analysis. BP, biological process; CC, cellular component; GO, Gene Ontology; GSEA, gene set enrichment analysis; KEGG, Kyoto Encyclopedia of Genes and Genomes; MF, molecular function.

KEGG pathway analysis corroborated these findings, identifying adipocytokine signaling, cytosolic DNA sensing, and JAK/STAT signaling as key pathways dysregulated in high-risk patients (Figure 3D). Notably, these pathways collectively underscored a pronounced influence on lipid metabolism and immune-inflammatory crosstalk. Additionally, immune-specific pathway analysis revealed significant enrichment of DEGs in CD8⁺ T cell effector functions and regulatory T cell (Treg) activity, further emphasizing the interplay between metabolic reprogramming and immune evasion in high-risk GBM (Figure 3E). Together, these results suggest that high-risk tumors exhibit a unique molecular profile characterized by disrupted lipid homeostasis, aberrant transcriptional regulation, and immunosuppressive T cell dynamics, which may collectively drive aggressive disease progression.

Tumor immunoassay

To investigate the disparities in the extent of immune infiltration within glioblastoma tumor tissues between high- and low-risk groups, we employed the CIBERSORT algorithm and ssGSEA algorithm on the TCGA-GBM dataset. These methods were utilized to quantify the level of immune cell infiltration for both the high- and low-risk groups individually. The high-risk group had more naive B cells, M2 macrophages, and γδ T cells (Figure 4A). According to the ssGSEA analysis, high-risk patients had significantly higher effector memory CD4 T cells and Type 1 T helper cells. As compared with the high-risk group, the low-risk group had fewer activated dendritic cells, central memory CD8 T-cells, natural killer cells, and plasmacytoid dendritic cells (Figure 4B).

Figure 4 Immune infiltration analysis. (A) 22 immune cells were analyzed using CIBERSORT to determine differences between high- and low-risk groups. (B) A comparison of 28 immune cell types as determined by ssGSEA. (C) Expression levels of 16 crucial immunomodulatory factors in high- and low-risk groups. *, P<0.05; **, P<0.01; ***, P<0.001; NS, not significant. ssGSEA, single‑sample gene set enrichment analysis.

Furthermore, we assessed the differential expression of pivotal immune checkpoint genes between the high- and low-risk groups. The analysis revealed that the expression of programmed cell death 1 (PDCD1) and T Cell Immunoglobulin and ITIM Domain (TIGIT) was markedly elevated in the high-risk group, whereas the expression of CD274 [programmed death-ligand 1 (PD-L1)] exhibited a downward trend (Figure 4C).

Analysis of immune-related hub genes, examination of immune infiltration, and analysis of target genes

We have characterized the expression profiles of immune cells in high- and low-risk GBM patients and investigated the expression trends of key immune checkpoint molecules. To better understand the relationship between the 19 hub genes and immune regulation, we identified immune-related lncRNAs in the GBM transcriptome data and an immune-related set through co-expression analysis, then compared them with the 19 hub genes. As a result, we discovered five central genes associated with immunity, specifically DLEU1, ENSG00000259704, ENSG00000249109, LINC1574 and SOX21-AS1. DLEU1 and SOX21-AS1 exhibited low expression, while the remaining three genes displayed high expression in the high-risk group (Figure 5A).

Figure 5 Expression and pathway analysis of immune-related hub genes. (A) Expression levels of DLEU1, ENSG00000259704, ENSG00000249109, LINC1574 and SOX21-AS1 in groups with high and low risk. (B) Correlation analysis of DLEU1, ENSG00000259704, ENSG00000249109, LINC1574 and SOX21-AS1 with the Spearman of CD274. (C) Spearman correlation analysis of ENSG00000259704 and SOX21-AS1. (D) Kaplan-Meier analysis of ENSG00000259704 and SOX21-AS1 expression levels on prognosis of GBM patients. (E) PPI analysis of target genes of SOX21-AS1. (F) GO analysis showing target genes of SOX21-AS1. (G) MF enrichment analysis of SOX21-AS1 target genes. BP, biological process; FPKM, fragments per kilobase of transcript per million mapped reads; GBM, glioblastoma multiforme; GO, Gene Ontology; MF, molecular function; PPI, proton-pump inhibitor.

Among the 16 notable immunomodulatory factors, CD274 exhibited reduced levels in the high-risk group. Moreover, we examined the relationship between CD274 expression level and hub genes related to immune function. ENSG00000259704 expression was negatively correlated with CD274 (Figure 5B), while SOX21-AS1 expression was positively correlated with CD274 (Figure 5C).

Based on our investigation of ENSG00000259704 and SOX21-AS1 expression in GBM patients, Kaplan-Meier analysis indicates that ENSG00000259704 is associated with a worse prognosis in patients with GBM, consistent with its high expression in high-risk patients. Nevertheless, the detrimental prognosis was once linked to the lowered expression of SOX21-AS1, which aligns with its decreased expression in the high-risk cohort (Figure 5D).

CD274 expression, a molecule that controls the immune system, is present on T-cell surfaces and is notably elevated in the low-risk category. This high expression of CD274 confirms the association of SOX21-AS1 with a favorable prognosis. This implies that it also impacts the immune system within the tumor environment and influences various biological processes that have a substantial impact on the prognosis of GBM. Furthermore, PPI analysis showed that SOX21-AS1 significantly regulates the interaction network of target genes (Figure 5E), primarily involved in morphogenesis (Figure 5F). The analysis of MF enrichment indicated that these genes had a significant impact on the activity of cell membrane protein receptors (Figure 5G).

To address potential systematic errors from GBM transcriptomic heterogeneity, we correlated key findings with the three major GBM subtypes (classical, proneural, mesenchymal) using the TCGA cohort. SOX21-AS1 expression varied significantly across subtypes: it was lowest in the mesenchymal subtype, with no difference between classical and proneural subtypes (Figure S1A). Consistent with mesenchymal GBM’s known high immune infiltration, immune checkpoints CD274 and PD-1 were highest in the mesenchymal subtype (Figure S1B,S1C). The 19-lncRNA RS was significantly elevated in the mesenchymal subtype and high-risk patients were predominantly enriched in this subtype (Figure S1D,S1E). SOX21-AS1 and CD274 expression showed a significant positive correlation (Figure S1F). The expression heatmap of the 19 hub lncRNAs clearly distinguished subtypes and risk groups, with high-risk samples clustering primarily in the mesenchymal subtype (Figure S1G).

Drug sensitivity analysis of different clinical variables

We estimated the treatment response of patients under various clinical factors by determining the IC50 to investigate drug sensitivity. Utilizing drug sensitivity data procured from the GDSC database, we employed the Mann-Whitney U test to ascertain the variations in sensitivity to various anticancer agents among distinct patient cohorts. We subsequently kept the top 20 medications with significant variations among various subcategories and displayed the outcomes (Figure 6). In addition, we analyzed the drug sensitivity of risk-associated hub genes in the GDSC dataset and the CTRP dataset (Figure S2), and the results showed that risk-associated hub genes could be used as drug targets for glioblastoma-related therapy.

Figure 6 Drug sensitivity analysis of different clinical variables. Patients with GBM exhibit different sensitivity (IC50 values) to anticancer drugs. (A) Lenalidomide; (B) gefitinib; (C) OSI.906; (D) VX.702; (E) CCT007093; (F) ABT.888; (G) ABT.263; (H) GW.441756; (I) Nutlin.3a; (J) BIBW2992; (K) X681640; (L) PD.0332991; (M) SL.0101.1; (N) SB590885; (O) nilotinib; (P) BMS.708163; (Q) bosutinib; (R) AICAR; (S) AZD6244; (T) BIRB.0796. ***, P<0.001. est, estimated; GBM, glioblastoma multiforme; IC50, half-maximal inhibitory concentration.

Discussion

In the realm of glioblastoma research, there is a growing emphasis on the role of lncRNAs. These lncRNAs are integral to the pathogenesis of glioblastoma, where they modulate genomic activity, transcriptional processes, post-translational modifications of proteins, and cellular communication. Elucidating these mechanisms is essential for deepening our understanding of glioma biology and for the development of innovative therapeutic approaches that specifically target the aberrant lncRNA networks within glioma cells (13). In this study, we systematically analyzed the TCGA dataset to explore the prognostic value of lncRNAs in GBM. Through differential expression and Cox regression analyses, we identified 217 lncRNAs significantly associated with GBM prognosis, of which 19 were selected to construct the prognostic model.

Cox regression analysis of these 19 core lncRNAs found that ENSG00000204802, RARA_AS1, LINC01956, CNIH3_AS1, LINC01127, HOXC13_AS, MIR1_1HG_AS1 and ENSG00000259869 were independent prognostic factors for GBM. Notably, RARA_AS1 and LINC01956 were significantly upregulated in the high-risk group and were closely associated with poor prognosis. RARA_AS1 is a chromosome 17q21.2-based antisense RNA that potentially contributes to sepsis development in children (14). Additionally, it plays a significant part in infectious disease-related shock (15). In GBM, RARA_AS1 is abundantly expressed, potentially influencing GBM cell growth and movement, as evidenced in our findings (16). In addition, LINC01956 might play a critical role in bladder cancer and clear cell renal cell carcinoma development (17,18), indicating LINC01956 might be involved in cancer pathogenesis. However, none of the studies have evaluated the association between LINC01956 expression and GBM so far. Thus, our findings reveal the therapeutic potential of LINC01956 for GBM.

In order to understand the relationship between the possible biological roles of lncRNAs in the model, the impacted cellular components, and the exerted molecular functions, GO and KEGG analyses were used to analyze the enriched genes. The significantly enriched biological processes and molecular pathways provided mechanistic details for understanding GBM pathology. The lncRNA-related genes primarily affected the metabolism of lipids, DNA replication, and transcription regulatory processes related to apoptosis and had negative regulatory effects on T-cell differentiation. In addition, they might contribute to GBM progression through pathways such as the adipocytokine pathway and the JAK/STAT pathway. GBM, known for its aggressive nature and poor prognosis, is attributed to the dysfunction of DNA repair mechanisms within GBM and the capacity of tumor cells to adapt their metabolism and cell cycle for sustained cellular proliferation (19-21). Notably, fatty acid (FA) metabolism in GBM cells is steered to maintain their growth and tumor progression; GBM cells dynamically store excess lipids in lipid droplets to prevent lipotoxicity and maintain tumor growth (22). The JAK/STAT pathway becomes active to transmit signals from cytokines, resulting in cell growth, specialization, and programmed cell death. Different diseases, such as tumors, are advanced by its activation (23). Also crucial to the aggressive behavior of GBM is the JAK/STAT pathway (24,25). Notably, the differences in lipid metabolism revealed by our study suggest that lncRNAs might influence GBM pathogenesis by affecting the metabolism of GBM cells. Metabolic reprogramming is considered a hallmark of cancer, and in addition to the typical changes caused by the Warburg effect, abnormal lipid anabolism is a hallmark of malignant tumor progression, with recent studies indicating that enhanced lipid synthesis promotes the growth and metastasis of GBM cells (26-28). LncRNA regulates the synthesis and catabolism of cholesterol, FAs, triglycerides (TG) and phospholipids (PL) through multiple mechanisms, thus participating in the occurrence and development of pancreatic cancer, breast cancer and other tumors (29-31). Research exploring the role of lipid-related lncRNAs in GBM is limited, necessitating further investigation in the future.

The immune microenvironment plays a crucial role in the development and progression of GBM, as well as the prognosis of patients and the effectiveness of chemotherapy. A significant correlation exists between an abundance of M2 macrophages, neutrophils, Treg and Breg cells in the GBM microenvironment and poor prognosis (32-35). Breg cells could directly or indirectly induce differentiation of Treg cells, thus affecting antitumor immune function and mediating tumor immune escape (36-38). Through CIBERSORT and ssGSEA analyses, we further explored immune cell infiltration in tumor tissues, which suggested that naive B-cells, macrophage M2, and T-cells γδ were infiltrated to a higher extent in the high-risk group, while activated central memory CD8 T-cells, dendritic cells, natural killer cells, and plasma-like dendritic cells were infiltrated to a high extent in the low-risk group. In accordance with the results of the risk group analysis, lncRNA genes affect CD8⁺ T-cells and Treg immunoregulatory cells significantly. This further affirms the crucial involvement of T cells in the immune response against GBM.

Targeted therapy against immune checkpoints has become an important approach in tumor immunotherapy. Modulating inhibitory signals to T-cells through immune checkpoint molecules could greatly enhance the survival rate of patients with resistant tumors. Thus, we examined the levels of 16 crucial immunomodulatory factors in groups with varying levels of risk. As a result, CD274 expression is decreased and PDCD1 and TIGIT expression are elevated in high-risk patients. The results we obtained align with earlier studies indicating that inhibiting CD274, PDCD1, and TIGIT could be a successful strategy for treating GBM (39-41). Next, we screened immune-related hub genes and evaluated their relationship with CD274. Notably, the expression of the immune-related hub gene SOX21-AS1 was positively correlated with CD274. CD274, a recently identified inhibitory molecule from the B7 group, is found to be expressed abnormally on the outer layer of numerous cancerous cells (42). Engaging with the T-cell surface receptor PD-1, tumor cells hinder the immune response of T-cells, ultimately aiding the tumor in evading the immune system (43-45). In this study, the positive regulatory relationship between SOX21-AS1 and CD274 suggested that SOX21-AS1 might participate in GBM immune microenvironment remodeling by regulating CD274 expression.

In addition, studies have found that the SOX21-AS1 gene is abnormally expressed in several tumor types, like breast cancer, cervical cancer, pancreatic cancer, and GBM. This gene is implicated in the advancement and progression of cancer (46-49). Our results show that low SOX21-AS1 expression is associated with poor prognosis in patients, suggesting that high SOX21-AS1 expression is associated with good prognosis, and some other bioinformatics studies have found the same results (50-52). In GBM, studies have shown that the expression of SOX21-AS1 gene is up-regulated in glioma tissues, and the knockdown of SOX21-AS1 gene in vitro can inhibit tumor growth by inhibiting cell proliferation and invasion in vivo (49). This is inconsistent with our results. We further performed PPI and functional enrichment analysis of SOX21-AS1 target genes, suggesting that SOX21-AS1 may affect tumor progression by regulating cell morphogenesis and membrane receptor activity. This suggests that SOX21-AS1 may regulate multiple biological processes and affect patient outcomes as a whole. The transcriptomic subtype analysis further validated the biological relevance of our 19-lncRNA signature. Consistent with prior reports that link the mesenchymal subtype to inferior prognosis and robust immune infiltration, our data demonstrated that the mesenchymal subtype exhibited the lowest SOX21-AS1 expression, the highest levels of CD274/PDCD1, and the most significant enrichment of high-risk patients.

To evaluate the clinical significance of our model, we used drug sensitivity data from the GDSC database to predict the sensitivity of samples of different clinical variables to common anticancer drugs, and the results showed that the IC50 value of a variety of drugs, including lenalidomide and gefitinib, for GBM patients was significantly increased, suggesting that their chemotherapy resistance was enhanced. indicating a decrease in drug sensitivity. This may explain the poorer prognosis of GBM patients, which further emphasizes the importance of individualised treatment for tumor patients. In addition, we analyzed the drug sensitivity of the risk-related hub gene, and the results showed that the risk-related hub gene could be used as a drug target for glioblastoma-related therapy.

Although this study revealed the multi-dimensional role of lncRNA in GBM, there are still limitations: based only on TCGA transcriptome data, protein level verification is lacking, and the specific mechanism of gene interaction with the immune system is not yet clear. And we were not able to cover a broader population to validate the generalization power of the risk model, a limitation that could affect the applicability of the model across different patient cohorts. Future experimental studies are needed to further verify the function of these lncRNAs and their feasibility as therapeutic targets.

Conclusions

This study disclosed five immune-related target genes as potential prognostic biomarkers for GBM. Particularly, SOX21-AS1 might become a new crucial target for exploring the pathogenesis and treatment of GBM. These discoveries offer a theoretical foundation for clinical application and indicate future research directions to verify the practical application value of these genes in clinical settings.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1-2684/rc

Peer Review File: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1-2684/prf

Funding: This work was supported by the Basic Scientific Research Business Expenses Project for Undergraduata Colleges and Universities, funded by the Department of Education of Liaoning Province (No. LJ212410160048).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1-2684/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Louis DN, Perry A, Wesseling P, et al. The 2021 WHO Classification of Tumors of the Central Nervous System: a summary. Neuro Oncol 2021;23:1231-51. [Crossref] [PubMed]
2. Tan AC, Ashley DM, López GY, et al. Management of glioblastoma: State of the art and future directions. CA Cancer J Clin 2020;70:299-312. [Crossref] [PubMed]
3. Newman JP, Wang GY, Arima K, et al. Interleukin-13 receptor alpha 2 cooperates with EGFRvIII signaling to promote glioblastoma multiforme. Nat Commun 2017;8:1913. [Crossref] [PubMed]
4. Wen PY, Reardon DA, Armstrong TS, et al. A Randomized Double-Blind Placebo-Controlled Phase II Trial of Dendritic Cell Vaccine ICT-107 in Newly Diagnosed Patients with Glioblastoma. Clin Cancer Res 2019;25:5799-807. [Crossref] [PubMed]
5. Singh S, Dey D, Barik D, et al. Glioblastoma at the crossroads: current understanding and future therapeutic horizons. Signal Transduct Target Ther 2025;10:213. [Crossref] [PubMed]
6. Li J, Ross JL, Hambardzumyan D, et al. Immunopathology of Glioblastoma. Annu Rev Pathol 2026;21:135-62. [Crossref] [PubMed]
7. Yadav G, Kulshreshtha R. Metastasis associated long noncoding RNAs in glioblastoma: Biomarkers and therapeutic targets. J Cell Physiol 2022;237:401-20. [Crossref] [PubMed]
8. Carlevaro-Fita J, Rahim A, Guigó R, et al. Cytoplasmic long noncoding RNAs are frequently bound to and degraded at ribosomes in human cells. RNA 2016;22:867-82. [Crossref] [PubMed]
9. Simon L, Probst AV. Maintenance and dynamic reprogramming of chromatin organization during development. Plant J 2024;118:657-70. [Crossref] [PubMed]
10. Shree B, Tripathi S, Sharma V. Transforming Growth Factor-Beta-Regulated LncRNA-MUF Promotes Invasion by Modulating the miR-34a Snail1 Axis in Glioblastoma Multiforme. Front Oncol 2021;11:788755. [Crossref] [PubMed]
11. Yadav B, Pal S, Rubstov Y, et al. LncRNAs associated with glioblastoma: From transcriptional noise to novel regulators with a promising role in therapeutics. Mol Ther Nucleic Acids 2021;24:728-42. [Crossref] [PubMed]
12. Zhao Z, Liu M, Long W, et al. Knockdown lncRNA CRNDE enhances temozolomide chemosensitivity by regulating autophagy in glioblastoma. Cancer Cell Int 2021;21:456. [Crossref] [PubMed]
13. Doghish AS, Mahmoud A, Abd-Elmawla MA, et al. Innovative perspectives on glioblastoma: the emerging role of long non-coding RNAs. Funct Integr Genomics 2025;25:43. [Crossref] [PubMed]
14. Wu Y, Yin Q, Zhang X, et al. Long Noncoding RNA THAP9-AS1 and TSPOAP1-AS1 Provide Potential Diagnostic Signatures for Pediatric Septic Shock. Biomed Res Int 2020;2020:7170464. [Crossref] [PubMed]
15. Zhang X, Cui Y, Ding X, et al. Analysis of mRNA‑lncRNA and mRNA‑lncRNA-pathway co‑expression networks based on WGCNA in developing pediatric sepsis. Bioengineered 2021;12:1457-70. [Crossref] [PubMed]
16. Huang Y, Deng S, Jiang Q, et al. LncRNA RARA-AS1 could serve as a novel prognostic biomarker in pan-cancer and promote proliferation and migration in glioblastoma. Sci Rep 2023;13:17376. [Crossref] [PubMed]
17. Wang S, Chai K, Chen J. A novel prognostic nomogram based on 5 long non-coding RNAs in clear cell renal cell carcinoma. Oncol Lett 2019;18:6605-13. [Crossref] [PubMed]
18. Zhong W, Qu H, Yao B, et al. Analysis of a Long Non-coding RNA associated Signature to Predict Survival in Patients with Bladder Cancer. Cureus 2022;14:e24818. [Crossref] [PubMed]
19. Lim YC, Ensbey KS, Offenhäuser C, et al. Simultaneous targeting of DNA replication and homologous recombination in glioblastoma with a polyether ionophore. Neuro Oncol 2020;22:216-28. [PubMed]
20. Kou Y, Geng F, Guo D. Lipid Metabolism in Glioblastoma: From De Novo Synthesis to Storage. Biomedicines 2022;10:1943. [Crossref] [PubMed]
21. Minami JK, Morrow D, Bayley NA, et al. CDKN2A deletion remodels lipid metabolism to prime glioblastoma for ferroptosis. Cancer Cell 2023;41:1048-1060.e9. [Crossref] [PubMed]
22. Cheng X, Geng F, Pan M, et al. Targeting DGAT1 Ameliorates Glioblastoma by Increasing Fat Catabolism and Oxidative Stress. Cell Metab 2020;32:229-242.e8. [Crossref] [PubMed]
23. Hu Q, Bian Q, Rong D, et al. JAK/STAT pathway: Extracellular signals, diseases, immunity, and therapeutic regimens. Front Bioeng Biotechnol 2023;11:1110765. [Crossref] [PubMed]
24. Delen E, Doğanlar O. The Dose Dependent Effects of Ruxolitinib on the Invasion and Tumorigenesis in Gliomas Cells via Inhibition of Interferon Gamma-Depended JAK/STAT Signaling Pathway. J Korean Neurosurg Soc 2020;63:444-54. [Crossref] [PubMed]
25. Mao P, Wang T, Du CW, et al. CXCL5 promotes tumorigenesis and angiogenesis of glioblastoma via JAK-STAT/NF-κb signaling pathways. Mol Biol Rep 2023;50:8015-23. [Crossref] [PubMed]
26. Zhong Y, Geng F, Mazik L, et al. Combinatorial targeting of glutamine metabolism and lysosomal-based lipid metabolism effectively suppresses glioblastoma. Cell Rep Med 2024;5:101706. [Crossref] [PubMed]
27. Kao TJ, Lin CL, Yang WB, et al. Dysregulated lipid metabolism in TMZ-resistant glioblastoma: pathways, proteins, metabolites and therapeutic opportunities. Lipids Health Dis 2023;22:114. [Crossref] [PubMed]
28. Eyme KM, Sammarco A, Jha R, et al. Targeting de novo lipid synthesis induces lipotoxicity and impairs DNA damage repair in glioblastoma mouse models. Sci Transl Med 2023;15:eabq6288. [Crossref] [PubMed]
29. Yu Y, Dong JT, He B, et al. LncRNA SNHG16 induces the SREBP2 to promote lipogenesis and enhance the progression of pancreatic cancer. Future Oncol 2019;15:3831-44. [Crossref] [PubMed]
30. Qin Y, Hou Y, Liu S, et al. A Novel Long Non-Coding RNA lnc030 Maintains Breast Cancer Stem Cell Stemness by Stabilizing SQLE mRNA and Increasing Cholesterol Synthesis. Adv Sci (Weinh) 2022;9:e2204046. [Crossref] [PubMed]
31. Zhang ZD, Hou XR, Cao XL, et al. Long non‑coding RNAs, lipid metabolism and cancer Exp Ther Med 2023;26:470. (Review). [Crossref] [PubMed]
32. Bao L, Li X. MicroRNA-32 targeting PTEN enhances M2 macrophage polarization in the glioma microenvironment and further promotes the progression of glioma. Mol Cell Biochem 2019;460:67-79. [Crossref] [PubMed]
33. DiDomenico J, Lamano JB, Oyon D, et al. The immune checkpoint protein PD-L1 induces and maintains regulatory T cells in glioblastoma. Oncoimmunology 2018;7:e1448329. [Crossref] [PubMed]
34. Yee PP, Wei Y, Kim SY, et al. Neutrophil-induced ferroptosis promotes tumor necrosis in glioblastoma progression. Nat Commun 2020;11:5424. [Crossref] [PubMed]
35. González-Tablas Pimenta M, Otero Á, Arandia Guzman DA, et al. Tumor cell and immune cell profiles in primary human glioblastoma: Impact on patient outcome. Brain Pathol 2021;31:365-80. [Crossref] [PubMed]
36. Wang Z, Lu Z, Lin S, et al. Leucine-tRNA-synthase-2-expressing B cells contribute to colorectal cancer immunoevasion. Immunity 2022;55:1748. [Crossref] [PubMed]
37. Ishigami E, Sakakibara M, Sakakibara J, et al. Coexistence of regulatory B cells and regulatory T cells in tumor-infiltrating lymphocyte aggregates is a prognostic factor in patients with breast cancer. Breast Cancer 2019;26:180-9. [Crossref] [PubMed]
38. Zhang Y, Wu J, Zhang H, et al. The Regulation between CD4(+)CXCR5(+) Follicular Helper T (Tfh) Cells and CD19(+)CD24(hi)CD38(hi) Regulatory B (Breg) Cells in Gastric Cancer. J Immunol Res 2022;2022:9003902. [Crossref] [PubMed]
39. Lee AH, Sun L, Mochizuki AY, et al. Neoadjuvant PD-1 blockade induces T cell and cDC1 activation but fails to overcome the immunosuppressive tumor associated macrophages in recurrent glioblastoma. Nat Commun 2021;12:6938. [Crossref] [PubMed]
40. Raphael I, Kumar R, McCarl LH, et al. TIGIT and PD-1 Immune Checkpoint Pathways Are Associated With Patient Outcome and Anti-Tumor Immunity in Glioblastoma. Front Immunol 2021;12:637146. [Crossref] [PubMed]
41. Li T, Xu D, Ruan Z, et al. Metabolism/Immunity Dual-Regulation Thermogels Potentiating Immunotherapy of Glioblastoma Through Lactate-Excretion Inhibition and PD-1/PD-L1 Blockade. Adv Sci (Weinh) 2024;11:e2310163. [Crossref] [PubMed]
42. Gou Q, Dong C, Xu H, et al. PD-L1 degradation pathway and immunotherapy for cancer. Cell Death Dis 2020;11:955. [Crossref] [PubMed]
43. Ma W, Gilligan BM, Yuan J, et al. Current status and perspectives in translational biomarker research for PD-1/PD-L1 immune checkpoint blockade therapy. J Hematol Oncol 2016;9:47. [Crossref] [PubMed]
44. Banta KL, Xu X, Chitre AS, et al. Mechanistic convergence of the TIGIT and PD-1 inhibitory pathways necessitates co-blockade to optimize anti-tumor CD8(+) T cell responses. Immunity 2022;55:512-526.e9. [Crossref] [PubMed]
45. Geels SN, Moshensky A, Sousa RS, et al. Interruption of the intratumor CD8(+) T cell:Treg crosstalk improves the efficacy of PD-1 immunotherapy. Cancer Cell 2024;42:1051-1066.e7. [Crossref] [PubMed]
46. Sheng XY, Wang CH, Wang CF, et al. Long-Chain Non-Coding SOX21-AS1 Promotes Proliferation and Migration of Breast Cancer Cells Through the PI3K/AKT Signaling Pathway. Cancer Manag Res 2020;12:11005-14. [Crossref] [PubMed]
47. Zhang X, Zhao X, Li Y, et al. Long noncoding RNA SOX21-AS1 promotes cervical cancer progression by competitively sponging miR-7/VDAC1. J Cell Physiol 2019;234:17494-504. [Crossref] [PubMed]
48. Yu D, Zhao Z, Wang L, et al. SOX21-AS1 activated by STAT6 promotes pancreatic cancer progression via up-regulation of SOX21. J Transl Med 2022;20:511. [Crossref] [PubMed]
49. Gai SY, Yuan ZH. Long non-coding RNA SOX21-AS1 promotes cell proliferation and invasion through upregulating PAK7 expression by sponging miR-144-3p in glioma cells. Neoplasma 2020;67:333-43. [Crossref] [PubMed]
50. Xie P, Yan H, Gao Y, et al. Construction of m6A-Related lncRNA Prognostic Signature Model and Immunomodulatory Effect in Glioblastoma Multiforme. Front Oncol 2022;12:920926. [Crossref] [PubMed]
51. Wang Z, Ji X, Gao L, et al. Comprehensive In Silico Analysis of a Novel Serum Exosome-Derived Competitive Endogenous RNA Network for Constructing a Prognostic Model for Glioblastoma. Front Oncol 2021;11:553594. [Crossref] [PubMed]
52. Paul Y, Thomas S, Patil V, et al. Genetic landscape of long noncoding RNA (lncRNAs) in glioblastoma: identification of complex lncRNA regulatory networks and clinically relevant lncRNAs in glioblastoma. Oncotarget 2018;9:29548-64. [Crossref] [PubMed]