Introduction

Gliomas are the most prevalent primary malignant tumors in the adult central nervous system. According to the World Health Organization (WHO) classification established in 2016, glioma are divided into grades I to IV based on the histological criteria1. With the discovery of many molecular biomarkers, glioma types were updated in the fifth edition of the WHO Classification of Tumors of the Central Nervous System (WHO CNS5) in 20212. Radiotherapy, chemotherapy and surgical resection are still the major clinical treatment strategies, yet patients with glioma obtain relative-short survival period. Targeted treatment and immunotherapy have achieved dramatic successes against many solid tumors, such as non-small-cell lung cancer (NSCLC), breast cancer, hepatocellular carcinoma (HCC), colorectal cancer3,4,5,6. However, there are no substantial progress in glioma with targeted therapy and immunotherapy. A phase III randomized clinical trial targeting epidermal growth factor receptor (EGFR) indicated no overall survival (OS) benefit from depatuxizumab mafodotin in newly diagnosed glioblastoma (GBM) with EGFR amplification7. Rindopepimut (also known as CDX-110), a vaccine targeting the EGFR deletion mutation EGFRvIII, failed to improve survival in newly diagnosed GBM patients with EGFRvIII expression8. Nivolumab, a PD-1 inhibitor, also failed to indicate an OS advantage in patients with GBM although the safety profile was consistent with that in other solid tumors (CheckMate-143)9. The failure of these clinical trials highlights the urgent need for novel glioma therapies and underscores the importance of tumor microenvironment.

Genetic mutations, such as isocitrate dehydrogenase (IDH), have been applicable to divide categories of glioma as long ago as 2016 WHO classification10. Beyond IDH mutation11 additional glioma biomarkers, including loss of chromosomes 1p and 19q, EGFR amplification or mutation12 TP53 mutation13 phosphatase and tensin homolog (PTEN) deletion14 O6-methylguanine-DNA methyltransferase (MGMT) methylation15,16 have been established as criteria for diagnosis, classification and prognostic prediction. These gene variations lead to abnormality of signaling pathway activation, ultimately impacting the occurrence and development of glioma. Recently, the status of each signaling pathway by accumulated genetic mutations could be evaluated using the individual pathway mutation perturbation (iPMP) calculation17. Here, we explored the cumulative effects of genetic mutations in signaling pathways on patient prognosis and identify specific pathway abnormalities significantly impacting survival in glioma patients. Meanwhile, potential reasons of poor survival were researched from clinical factors, tumor immune microenvironment (TIME) and signaling regulation levels.

Results

Pathway activation abnormalities by genetic mutations reduced survival in glioma

We obtained clinical information of 1148 cases from the TCGA glioma dataset. There were 816 patients with paired genetic mutation, RNA sequencing, clinical and survival data. As shown in Supplementary Table 1, the cohort contained 506 (62.0%) low-grade glioma (LGG, grade I-II) cases and 310 (38.0%) high-grade glioma (HGG, grade III-IV) cases. Patients with IDH1 mutation accounted for half of all patients. We calculated activity scores for 123 candidate signaling pathways for each case using the PMAPscore package and analyzed the correlation between pathway activity and survival using univariate Cox regression analysis in the glioma cohort from the TCGA dataset. Seventy-six pathways showed a significant correlation with progression-free survival (PFS) (Supplementary Fig. 1). These pathways were further analyzed using multivariate Cox regression analysis, identifying three pathways significantly correlated with PFS (Fig. 1A). The three signaling pathways contained autophagy-animal, focal adhesion and neuroactive ligand-receptor interaction. Based on the three signaling pathways, we constructed a biological pathway score (BPS) model and calculated BPS according to BPS = 0.09*autophagy-animal + 0.11*focal adhesion + 0.22*neuroactive ligand-receptor interaction in the glioma cohort from the TCGA dataset. As shown in Fig. 1B, an optimal cutoff value (BPS = 0.45) was identified, dividing the cohort into low/high BPS groups. The cutoff value also was used in the subsequent research. Between the low and high BPS groups, we found that PFS and OS exhibited significant difference in the glioma cohort from the TCGA dataset (Fig. 1C and D). The low BPS group was beneficial for PFS and OS in the cohort.

Fig. 1
Fig. 1The alternative text for this image may have been generated using AI.
Full size image

Correlation of pathway abnormality by accumulated genetic mutations and survival in glioma. (A) Multivariate Cox regression analysis of signaling pathways in the glioma cohort from the TCGA dataset. (B) Distribution of BPS in the glioma cohort from the TCGA dataset. An optimal cutoff value (BPS = 0.45) was used in the research. (C and D) Kaplan–Meier survival curves of PFS (C) and OS (D) comparing the low/high BPS cohorts with glioma from the TCGA dataset. (E) Kaplan–Meier survival curves of PFS comparing the low/high BPS cohorts with glioma from Hui Zhou Central People’s Hospital. (F and G) Kaplan–Meier survival curves of PFS (F) and OS (G) comparing the low/high BPS cohorts with glioma from the cBioPortal dataset.

To validate the results, we utilized two independent glioma cohorts from Hui Zhou Central People’s Hospital and cBioPortal (Glioma mskcc 2019). The clinical features of these cohorts were showed in Table 1 and Supplementary Table 2, respectively. In our hospital cohort, patients with low BPS had significantly better PFS than those with high BPS (Fig. 1E). A similar result was also observed in the glioma cohort from the cBioPortal dataset (Fig. 1F). Furthermore, the low BPS cases were beneficial for OS, which was also demonstrated in the cohort from the cBioPortal dataset (Fig. 1G). These results indicated that BPS could predict patient survival, with low BPS associated with a survival benefit across glioma cohorts.

Table 1 Clinical characteristics of the glioma cohort from Hui Zhou central people’s Hospital.
Full size table

Genetic mutations and clinical factors impacting survival in low/high-BPS glioma patients

Assessment of BPS was on the basis of the three PFS-related pathways in the glioma cohort. The signature, including autophagy-animal, focal adhesion and neuroactive ligand-receptor interaction, was consist of 672 genes (Supplementary Table 3). We displayed top 10 genes by mutation rate in the three pathways in the glioma cohort from the TCGA dataset. In the autophagy-animal signaling pathway, mutation frequencies of PTEN, PIK3R1 and PIK3CB genes in the low BPS group were significantly lower than in the high group (Fig. 2A). In the focal adhesion signaling pathway, mutation frequencies of PTEN, EGFR, PIK3R1, COL6A3, COL2A2, and LAMA1/2 genes were also reduced in the low BPS cases compared to the high BPS cases with glioma (Fig. 2B). In the neuroactive ligand-receptor interaction signaling, mutation rates of GRIN2A and DRD5 genes were enhanced in the high BPS patients with glioma (Fig. 2C). These differential genes from three signaling pathways exhibited different pathway scores between the low and high BPS cases with glioma, and the pathway scores in the high BPS group were significantly enriched (Fig. 3A). Analysis of the top 20 mutated genes in the glioma cohort from the TCGA cohort exhibited that variation rates of many genes, such as TP53, IDH1, PTEN, TNN, EGFR and ATRX, were markedly different between the BPS low and high groups (Fig. 3B). Surprisingly, mutation frequencies of TP53, IDH1 and ATRX were higher in the low group.

Fig. 2
Fig. 2The alternative text for this image may have been generated using AI.
Full size image

Features of genetic mutations in three abnormal signaling pathways in low/high BPS patients with glioma. Landscapes of top 10 gene mutations in autophagy-animal (A), focal adhesion (B) and neuroactive ligand-receptor interaction (C) in the glioma cohort from the TCGA dataset. p value was determined by the Chi-Square test. *p < 0.05; **p < 0.01; ***p < 0.001.

Fig. 3
Fig. 3The alternative text for this image may have been generated using AI.
Full size image

Features of genetic mutations in low/high BPS patients with glioma. (A) Distribution of the scores of three signaling pathways in low/high BPS patients with glioma from the TCGA dataset. (B) Landscape of top 20 genes of mutation frequencies in the glioma cohort from the TCGA cohort. p value was determined by the Chi-Square test. *p < 0.05; **p < 0.01; ***p < 0.001.

We next studied correlations between BPS and clinical factors in the low and high patients with glioma. As shown in Supplementary Table 4, patients in the low-BPS group were significantly younger than those in the high-BPS group within the TCGA cohort. Low-BPS patients had a higher proportion of LGG diagnoses and IDH1 mutations (Supplementary Table 4). The similar results were found in the glioma cohorts from our hospital and the cBioPortal dataset (Table 2 and Supplementary Table 5).

Table 2 Clinical characteristics of the low/high BPS patients with glioma from Hui Zhou central people’s Hospital.
Full size table

Tumor immune microenvironment in the low/high BPS patients with glioma

To evaluate the mechanisms underlying poor survival in the high BPS patients with glioma, TIME was explored in the glioma cohort from the TCGA dataset. We surprisingly discovered that many immune-activated cells, such as activated CD4 T cell, activated CD8 T cell, activated dendritic cell, macrophage and natural killer cell, as well as immune-suppressive cells, such as myeloid-derived suppressor cell (MDSC) and regulatory T cell (Treg), were present at higher levels in the high BPS patients compared to the low patients in the glioma cohort (Fig. 4A). In contrast, activated B cell and eosinophil in the low BPS patients with glioma were more abundant (Fig. 4A). The cancer-immunity cycle analysis also revealed that both anti-tumor immune (IGS1-5) and immune suppression (IGS6, IGS8) in the high BPS group were stronger than in the low group (Fig. 4B). These results suggested that tumor-infiltrating immune cells were abundant yet suppressed in the high BPS patients, which led to poor survival compared to the low BPS patients with glioma.

Fig. 4
Fig. 4The alternative text for this image may have been generated using AI.
Full size image

Tumor immune microenvironment in the low/high BPS patients with glioma. (A) ssGSEA analyzed tumor immune-cell infiltration in the low/high BPS cohorts with glioma. (B) Cancer-immunity cycle analysis in the low/high BPS cohorts with glioma. IGS1, existence of T-cell immunity in the tumor; IGS2, tumor antigenicity; IGS3, priming and activation; IGS4, trafficking and infiltration; IGS5, recognition of tumor antigens; IGS6, absence of inhibitory cells; IGS7, absence of checkpoint expression; and IGS8, absence of inhibitory molecules. (C) Expression of 69 immune checkpoint genes (ICGs) in the low/high BPS cohorts with glioma. (D) Expression of PD-L1 and PD-1 in the low/high BPS cohorts with glioma. (E and F) Representative immunofluorescence imaging (E) and quantitative analysis (F) in the low and high BPS patients. Scale bar, 50 μm. Statistics based on the two-tailed Mann-Whitney U test. *p < 0.05; **p < 0.01; ***p < 0.001.

We next investigated expression differences of 69 immune checkpoint genes (ICGs) between the low and high BPS groups. These ICGs contained butyrophilin (BTN, 5 genes), killer cell immunoglobulin-like receptors (KIR, 7 genes), major histocompatibility complex (MHC) class I (6 genes), MHC class II (11 genes) and other ICGs (40 genes) families. As shown in Fig. 4C, BTN family genes (including BTN3A1 and BTN2A2) in the high BPS group were higher expression. There was no significant difference in 7 genes of KIR family between two groups. Sixteen-gene expressions of MHC family were more in the high BPS patients with glioma, which was consistent with the result of IGS5 in Fig. 4B. Recognition ability of tumor antigens in the high BPS group was stronger than in the low group. But expressions of CD279 (PD-1), CD274 (PD-L1) and HAVCR2 (TIM-3) genes were also markedly enhanced in the high BPS patients with glioma (Fig. 4C and D). Immunofluorescence imaging indicated that protein levels of PD-1 and PD-L1 were increasing in the high BPS cases compared to the low cases (Fig. 4E and F). Overall, these results demonstrated that tumor-infiltrating immune cells were abundant but suppressed in the high BPS patients, which likely contributes to the poor survival outcomes observed in this subgroup.

Differential expression genes and potential application in the low/high BPS patients with glioma

We further explored differential expression genes (DEGs) between the low and high BPS groups in the glioma cohort from the TCGA dataset. We found that expressions of 451 genes were up-regulating and expressions of 607 genes were down-regulating in the low BPS patients with glioma (Fig. 5A). KEGG enrichment analysis indicated that these DEGs markedly enriched in cell cycle, ECM-receptor interaction, complement and coagulation cascades, neuroactive ligand-receptor interaction and nicotine addiction (Fig. 5B). Meanwhile, proliferation relation signaling pathways, such as cell cycle, cellular senescence, p53 signaling pathway and transcriptional misregulation in cancer, were enriched in the high BPS group (Fig. 5C). This enrichment represents one potential mechanism underlying the poor survival in high BPS patients with glioma.

Fig. 5
Fig. 5The alternative text for this image may have been generated using AI.
Full size image

DEGs and potential application in the low/high BPS patients with glioma. (A) Volcano plot showing profile of DEGs between the low and high BPS patients with glioma. 451 genes were up-regulating and 607 genes were down-regulating in the low group. (B) KEGG pathway enrichment of DEGs between the low and high BPS patients with glioma. (C) GSEA in cell cycle, cellular senescence, ECM-receptor interaction, neuroactive ligand-receptor interaction, p53 signaling pathway and transcriptional misregulation in cancer between the low and high BPS patients with glioma. (D) Volcano plot showing profile of DEGs between the glioma patients with propofol or sevoflurane. 153 genes were up-regulating and 527 genes were down-regulating in the propofol group. (E) GSEA in neuroactive ligand-receptor interaction between propofol and sevoflurane groups with glioma.

As surgical resection is still one of main therapeutic strategies for glioma, anesthesia is necessary during surgery. In the glioma cohort from the TCGA dataset, we found DEGs significantly enriched neuroactive ligand-receptor interaction in the low patients with glioma (Fig. 5C), who obtained beneficial survival (Fig. 1B and C). We investigated anesthetic effects using the glioma cohort (GSE179004) with anesthesia information (propofol vs. sevoflurane). There were 680 DEGs between the propofol and sevoflurane groups and 153-up genes and 527-down genes were observed in the propofol group compared to the sevoflurane group (Fig. 5D). Excitedly, DEGs dramatically enriched neuroactive ligand-receptor interaction in the sevoflurane group (Fig. 5E). These results suggested that the low BPS patients with glioma might benefit from sevoflurane anesthesia.

Discussion

The study explored that predictive ability of biological pathway abnormalities by accumulated genetic mutations for outcomes in glioma patients. We screened three PFS-related signaling pathways, including autophagy-animal, focal adhesion and neuroactive ligand-receptor interaction, constructed and validated the prediction prognosis model. Patients with low BPS exhibited significantly longer PFS and OS. Age, tumor subtype and IDH1 status were important factors of better survival in the low BPS patients compared to the high patients. TIME results indicated that immune checkpoints such as PD-1, PD-L1 and TIM-3 were strikingly expressed in the high group although high levels of tumor immune infiltration were observed. Meanwhile, DEGs in the high BPS group were markedly enriched in proliferation relation signaling pathways, such as cell cycle, cellular senescence, p53 signaling pathway and transcriptional misregulation in cancer. These results from clinical factors, TIME and signaling regulation comprehensively revealed reasons of poor PFS in the high BPS patients with glioma.

While gene variation affects occurrence, development, treatment and prognosis of glioma, most studies generally focus on the mutations of driver genes. The present research explored the cumulative effects of genetic mutations in each signaling pathway on prognosis of patients with glioma, identifying three PFS-relative pathways, including autophagy-animal, focal adhesion and neuroactive ligand-receptor interaction. Autophagy is generally recognized as tumor-suppressive18 inhibited by activating mutations of oncogenes (such as PIK3CA) or inactivation mutations of suppressor genes (such as PTEN)19. Meanwhile, mutations in suppressor gene PTEN and oncogene EGFR and PIK3CA activate the focal adhesion pathway, promoting tumor cell motility, proliferation and survival20. In addition, defects in the neuroactive ligand-receptor interaction pathway was demonstrated to have a poor prognosis in GBM patients21. Thus, abnormalities in these signaling pathways by accumulated genetic mutations were closely linked with survival of patients with glioma. Our study sufficiently demonstrated that the BPS model based on three pathways effectively predicts prognosis of patients in glioma.

Immune cells execute their immune surveillance function primarily through cell migration. The TIME in glioma is highly complex. In the present study, ssGSEA and cancer-immunity cycle analyses revealed that the high BPS group exhibits elevated levels of tumor immune infiltration. However, the TIME in this group simultaneously demonstrates a highly immunosuppressive state. Previous studies have implicated the focal adhesion signaling pathway in glioma patient survival and its influence on the TIME status22. Notably, the high-risk group for focal adhesion signaling also shows significantly increased infiltration of immune cells22. We further found a significantly higher proportion of HGG in the high BPS group. This higher HGG proportion is one of the major factors contributing to the increased immune cell infiltration observed in the high BPS group. This result aligns with prior studies indicating that the degree of tumor-infiltrating immune cells increases with ascending tumor grade23. HGGs, particularly GBM, are characterized by an immunosuppressive microenvironment24. Key mechanisms underlying this immunosuppression involve the aggregation of MDSCs, glioma-associated macrophages/microglia (GAMs), and Tregs25. Within the high BPS group, the levels of MDSCs, macrophages, and Tregs were significantly higher than in the low BPS group. This elevated presence of immunosuppressive cell populations was a primary cause of the inhibitory TIME in the high BPS group. Although MDSCs constitute only a minority (4–8%) of CD45 + cells in GBM, they play a central role in promoting immune evasion, driving tumor progression, angiogenesis, invasion, and metastasis25. Furthermore, we observed significantly increased expression of immune checkpoint molecules, such as PD-1, PD-L1, and TIM-3, in the high BPS group. The combination of these factors (the increased proportion of HGGs, the heightened infiltration of immunosuppressive cells, and the elevated expression of immune checkpoints) constituted the primary reason for the paradoxical state observed in the high BPS group: high levels of tumor-infiltrating immune cells coexisting with a profoundly immunosuppressive microenvironment.

Immunotherapy has achieved dramatic successes against many solid tumors3,4,26,27 and is written into the Chinese Society of Clinical Oncology (CSCO) and National Comprehensive Cancer Network (NCCN) guidelines. However, glioma remains obviously absent from the list of approvals. This discrepancy correlates with the unique TIME of glioma. Tumor-infiltrating immune cells display low numbers relative to other tumor types because of the blood-brain barrier (BBB) limiting the infiltration of circulating immune cells28,29. Meanwhile, immune cells were suppressed by immune inhibitory cells and immune checkpoint blockade29. The present study indicated that many tumor-infiltrating immune cells in the high BPS patients with glioma were more levels than in the low cases, but immune inhibitory cells (such as MDSC and regulatory T cell) and ICGs (such as PD-1, PD-L1 and TIM-3) were also elevated in the high patients. These results revealed one of reasons that the high BPS patients with glioma exhibited poor survival, suggesting the high BPS patients might be beneficial from immunotherapy. Unfortunately, PD-1 inhibitor (CheckMate-143) or their combination with radiotherapy (CheckMate-498) have failed to demonstrate a survival advantage in GBM patients9,30. Thus, how to remove barriers to successful immune checkpoint blockade is a key of immunotherapy and a future goal of clinical trials in glioma.

Mutations in IDH, TP53 and ATRX have been implicated as important markers of diagnostic categories about glioma10,31,32. Glioma patients with IDH mutation have a better outcome than those with wild-type IDH genes33. In the low BPS patients with glioma, the high frequency of IDH1 mutation was observed. Meanwhile, mutation rates of EGFR and PTEN were lower in the low BPS cases with glioma compared to the high cases. Younger age and a lower HGG rate in the low BPS cohort were also one of key factors of favorable PFS. In addition, DEGs were enriched in the proliferation-related signaling pathways in the high BPS group and in the neuroactive ligand-receptor interaction pathway in the low group. Enrichment of the neuroactive ligand-receptor interaction pathway exhibits a beneficial prognosis in glioblastoma patients21. These results comprehensively revealed reasons of poor PFS in the high BPS patients with glioma.

Sevoflurane and propofol are two popular anesthetics used during glioblastoma surgery. Sevoflurane is one of the most widely used anesthetics for the induction and maintenance of general anesthesia in surgical patients, providing hypnosis, amnesia, analgesia, akinesia, and autonomic blockade during surgical and procedural interventions34. Sevoflurane exhibited neuroprotective effects on adult brains across various brain disorders, including traumatic injuries, hypoxic conditions, and ischemia-reperfusion injuries35. Previous study indicated that sevoflurane can more effectively prevent the development of glioma after surgery than propofol36. In the present study, we found that DEGs dramatically enriched neuroactive ligand-receptor interaction in the sevoflurane group compared to the propofol group. These results suggested that the low BPS patients with glioma might benefit from sevoflurane during period of anesthesia.

Conclusion

In conclusion, signaling pathway abnormalities by accumulated genetic mutations restricted survival of patients with glioma, potential mechanisms of which comprehensively explored from clinical factors, TIME and signaling regulation. These results provide a comprehensive explanation for poor PFS in high-BPS glioma patients.

Materials and methods

Patients and clinical data

The study enrolled 45 patients with glioma from Hui Zhou Central People’s Hospital between January 2019 and January 2023. This research was approved by the Ethics Committee of Hui Zhou Central People’s Hospital (ID: kyll2024270) and treatment was performed by the Chinese society of clinical oncology (CSCO) guidelines. All patients provided their written informed consent to participate in this research. These patients obtained target exon sequencing data and extended follow-up data was retrospectively assembled with medical ethics committee approval.

A glioma cohort was obtained from the cancer genome atlas (TCGA) (12 datasets) (https://xenabrowser.net/datapages/). An 816-glioma cohort was screened, which obtained paired DNA mutation, RNA sequencing, clinical and survival information. A validation cohort with 493-glioma cases from cBioPortal Glioma (MSK, Clin Cancer Res 2019) (https://www.cbioportal.org/) was used, which contained paired DNA mutation, clinical and survival information. In addition, a glioma cohort with propofol or sevoflurane was studied (GSE179004; https://www.ncbi.nlm.nih.gov/geo/).

Biological pathway assessment and risk model construction

Biological pathway scores by accumulated genetic mutations were assessed according to the Xiangmei Li’s study17. 123 signaling pathways of each patient were calculated using the “PMAPscore” R package. Screening signaling pathways were correlated with PFS using univariate and multivariate Cox regression analysis and constructed a BPS model using the TCGA cohort. The BPS of each patient was calculated as follow: BPS=∑(Coefficient×SignalingScore). The Coefficient and SignalingScore represented the risk coefficient and score of each signaling pathway, respectively. The BPS model was validated using the clinical cohort from Hui Zhou Central People’s Hospital and the glioma mskcc 2019 cohort from cBioPortal.

Signaling enrichment and immune infiltration analysis

DEGs were analyzed between the low and high BPS groups in the TCGA cohort. Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis37,38,39 and Gene set enrichment analysis (GSEA) analysis were performed40. Single-sample gene set enrichment analysis (ssGSEA) evaluated the extent of infiltration of 28 immune cell types in each patient. Cancer-Immunity Cycle evaluated the interaction of immune cells and cancer cells41.

Multiplex Immunofluorescence

Multiplex immunofluorescence was performed as our previous study42. The formalin-fixed paraffin-embedded (FFPE) samples of tumor tissues were stained using the Opal Seven-color IHC Kit (NEL797B001KT; PerkinElmer, USA). Primary antibodies contained CD8 (ZA-0508, clone SP16; Zsbio), CD68 (ZM-0060, clone KP1; Zsbio), programmed cell death-1 (PD-1; Clone UMAB199; Zsbio) and programmed cell death-ligand 1 (PD-L1; Clone E1L3N; Cell Signaling Technology). Cell nucleus was stained by diamidino-phenyl-indole (DAPI). The sections were scanned and analyzed using PerkinElmer Vectra (Vectra 3.0.5; PerkinElmer, Massachusetts, USA).

Statistical analysis

Statistical analysis was conducted by R software (version 4.2.3). PFS-relative signaling pathways were screened using univariate and multivariate Cox regression analysis. PFS and OS analysis were performed using Kaplan-Meier analysis (log-rank test). Clinical feature analysis was performed using the Chi-Squared test. Comparison between 2 groups was performed using the two-tailed Mann-Whitney U test. A p-value of < 0.05 was considered as statistically significant.