Abstract
Glioblastoma multiforme (GBM) is a highly malignant primary brain tumor with a poor prognosis. Reactive oxygen species that accumulate during tumorigenesis can cause oxidative stress (OS), which plays a crucial role in cancer cell survival. Clinical and transcriptome data of TCGA-GBM dataset from UCSC Xena database were analyzed. Consensus clustering analysis was conducted to identify OS-related molecular subtypes for GBM. The immune infiltrate level between subtypes were characterized by ESTIMATE algorithm. Differentially expressed genes (DEGs) between subtypes were screened by DESeq2 package. Two OS-related molecular subtypes of GBM were identified, and cluster 2 had poorer overall survival and higher immune infiltration levels than cluster 1. Enrichment analysis showed that 54 DEGs in cluster 2 were significantly enriched in cytokine/chemokine-related functions or pathways. Ten hub genes (CSF2, CSF3, CCL7, LCN2, CXCL6, MMP8, CCR8, TNFSF11, IL22RA2, and ORM1) were identified in GBM subtype 2 through protein-protein interaction network, most of which were positively correlated with immune factors and immune checkpoints. A total of 55 small molecule drugs obtained from drug gene interaction database (DGIdb) may have potential therapeutic effects in GBM subtype 2 patients. Our study identified 10 hub genes as potential therapeutic targets in GBM subtype 2 patients, who have poorer overall survival and higher immune infiltration levels. These findings could pave the way for new treatments for this aggressive form of brain cancer.
Keywords: differentially expressed genes, enrichment analysis, glioblastoma multiforme, oxidative stress, RNA-seq
1. Introduction
Glioblastoma multiforme (GBM) is the most common primary malignant brain tumor in adults, accounting for 57% of all gliomas and 48% of primary malignant tumors of the central nervous system.[1] Current treatments for GBM, including maximal tumor resection, adjuvant radiation therapy, and chemotherapy, only marginally improve patient survival.[2] The median survival of patients is only 12 to 15 months, and the 5-year survival rate is <7%, mainly due to limited drug uptake by tumor cells and tumor resistance to chemotherapy.[3] To better understand GBM, it is classified according to histopathological and molecular characterization, such as IDH mutation status, 1p/19q deletion, O6-methylguanine-DNA-methyltransferase promoter methylation, and the epidermal growth factor receptor variant III amplification. These indicators are clinically significant for diagnosis, identifying effective treatments, and assessing prognosis.[4]
Although there have been promising findings regarding hub genes and pathways in diagnosing and treating GBM, developing new therapies for GBM faces significant challenges. These challenges primarily arise from biological factors such as the blood-brain barrier and the unique tumor immune microenvironment.[5] However, a ray of hope lies in the field of nanomedicine, which has the potential to overcome the blood-brain barrier and deliver drugs directly to the brain tumor site.[6] Therefore, there is an urgent need to develop more effective treatments and identify new targets that affect cellular metabolism in GBM, enhance responsiveness to drugs, maximize patient survival, and improve patient quality of life and prognosis.[7]
Oxidative stress (OS) is a pathological phenomenon that occurs when the organism is exposed to harmful endogenous or exogenous factors or when there is an imbalance in oxidant-antioxidant synthesis.[8] High OS levels can increase reactive oxygen species (ROS) in cells, which can cause DNA damage, protein oxidation, lipid peroxidation, and alter their function.[9] Studies have shown that OS is associated with various diseases, including diabetes,[10] malignancy,[11] cardiovascular disease,[12] and neurological disorders.[13] In glioma cells, OS can promote tumor cell progression and drug resistance by increasing ROS levels and damaging mitochondrial DNA.[14] Although the role of ROS in GBM is less studied, regulating OS and ROS status may serve as potential targets for glioma therapy. Therefore, we aimed to identify OS-related subtypes and biomarkers in GBM through the Cancer Genome Atlas Program (TCGA) database. Additionally, we performed immune cell infiltration, tumor microenvironment, and drug prediction analyses in GBM subtypes with poor overall survival.
2. Methods
2.1. Data source
The transcriptome and survival data of TCGA-GBM, comprising 167 GBM cases, were acquired from the UCSC Xena database (http://xena.ucsc.edu). GeneCards (http://www.genecards.org), an exhaustive database offering a comprehensive compendium of annotative information about human genes, was queried using a specifically designed algorithm to obtain 168 OS-related genes with a relevance score >15.[15] This allowed for a stringent and targeted data retrieval, focusing on relevant gene entities.
Identification of OS-related molecular subtypes using consensus clustering analysis
The expression matrix of 168 OS-related genes was meticulously extracted from the GBM transcriptome. This extraction involved complex computational algorithms and a series of data transformation and normalization processes to ensure accuracy. The consensus clustering analysis was performed using the R package “ConsensusClusterPlus (1.56.0)” to identify OS-related molecular subtypes of GBM patients.[16] This analysis entailed intensive computational processing and applied a consensus clustering algorithm, which is a sophisticated iterative technique involving multiple resampling and clustering runs to ensure robustness and stability of the identified clusters. The main aim was to group GBM patients based on their gene expression profiles related to OS. This provided a deeper, multi-dimensional insight into the underlying biological processes (BP).
2.2. Overall survival analysis
The overall survival analysis was executed using the R package “survival (3.2.13).”[17] This analysis implemented the Cox proportional-hazards model and Kaplan–Meier survival curves to compare the overall survival of patients in different molecular subtypes. This allowed for a nuanced evaluation of the prognostic value of the identified subtypes.
2.3. Immune microenvironment analysis of OS-related molecular subtypes
To understand the immune infiltrate level between the 2 identified molecular subtypes, the immune score, stromal score, ESTIMATE score, and tumor purity were calculated using the ESTIMATE algorithm embedded in the R package “estimate.”[18] The immune and stromal scores were determined by analyzing specific gene expression characteristics of immune and stromal cells. High-throughput computational analysis was performed, followed by statistical tests to evaluate the differences in immune/stromal/ESTIMATE scores and tumor purity between the 2 clusters. The purpose of this analysis was to identify possible differences in the immune microenvironment between the 2 subtypes.
2.4. Identification of differentially expressed genes (DEG) between OS-related molecular subtypes
The identification of DEGs between the 2 OS-related molecular subtypes was enabled using the R package “DESeq2 (1.32.0).”[19] This rigorous process involved leveraging the power of DESeq2 internally developed algorithms to conduct differential expression analysis based on a model using the negative binomial distribution. An adjusted P value cutoff of <.05, and an absolute log2 fold-change (FC) >1 were set up as the significant thresholds. Subsequent to selecting DEGs, a volcano plot and heatmap were created to transform large-scale transcriptomic data into visually interpretable formats. The volcano plot was used to simultaneously display the fold-change and significance of genes, while the heatmap presented the relative expression of DEGs in a color-coded format, emphasizing their distribution and disparity between the 2 molecular subtypes. The purpose of this analysis was to identify genes that are differentially expressed between the 2 subtypes and may be responsible for the observed differences in clinical outcomes and immune infiltration levels.
2.5. Functional enrichment analysis of DEGs
To ascertain the functional roles of DEGs, gene ontology (GO) term analysis and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis were performed using the R packages “clusterProfiler (4.0.2)”[20] and “enrichplot (1.12.2),” respectively. These analyses incorporated advanced bioinformatic algorithms to interpret the biological theme among gene clusters, offering insights into the DEGs’ proposed biological roles at 3 levels: BP, cellular component (CC), and molecular function. GO terms or KEGG pathways with adjusted P values < .05 were considered statistically significant, accurately identifying significantly enriched terms and pathways among the group of DEGs.
2.6. Construction of protein-protein interaction (PPI) network and identification of hub genes
A PPI network was built using the Search Tool for the Retrieval of Interacting Genes (STRING) online database, with interactions requiring a minimum interaction score over 0.4.[21] This necessitated the application of sophisticated predictive algorithms to identify potential PPI based on an extensive array of sources, including experimental data, computational prediction methods and public text collections. The PPI network was visualized using Cytoscape software (3.8.2), which synthesizes vast amounts of biomolecular interaction data into a readable, visually compelling map. Hub genes were identified using the cytoHubba plug-in of Cytoscape, which ranks nodes in a network using sophisticated topological analysis. This process involved advanced computations based on the node degree connectivity, revealing the most connected nodes in the network. The purpose of this analysis was to identify key genes associated with the molecular subtypes.
2.7. Correlation analysis of immune factors, immune checkpoints with hub genes
Spearman correlation coefficient analysis, a nonparametric measure of statistical dependence between 2 variables, was employed to ascertain correlations between immune factors and the previously identified hub genes. Critical data for this analysis, incorporating immune-suppressive factors, immune-stimulatory factors, and chemokines, was sourced from the TISIDB database.[22] In a parallel procedure, the expression correlation of immune checkpoints with hub genes was calculated, utilizing comprehensive statistical calculations to explore potential associations herein. The purpose of this analysis was to explore the relationship between hub genes and immune factors and checkpoints.
2.8. Identification of the potential drugs of GBM subtype 2
The Drug Gene Interaction database (DGIdb) was utilized to predict potential drug or molecular compounds that interacted with hub genes.[23] This preemptive analysis hinged on comprehensive data retrieval from the DGIdb and meticulous computational work to establish potential drug-gene interactions. A drug-gene interaction network was constructed and subsequently visualized using Cytoscape software. The purpose of this analysis was to identify potential drugs for patients with GBM subtype 2.
2.9. Statistical analysis
All statistical analysis were performed using the R programming language (version 3.6.3) - a powerful tool for statistical computing and graphics. All analysis adhered to a rigorous standard, where a P value threshold of <.05 was set to denote statistical significance.
3. Results
3.1. Identification of molecular subtypes related to OS
The expression matrix of OS-related genes in GBM was extracted from the TCGA database. Consensus clustering analysis was conducted on the expression matrix of these genes to identify molecular subtypes related to OS in GBM. The matrix plots in Figures 1A to 1D display the consensus clustering results for different k values (k = 2 to k = 5). The cumulative distribution function plot in Figure 1E shows that the optimal k value was determined as 2 based on the highest relative change. As a result, all samples were classified into 2 clusters, cluster 1 (containing 93 samples) and cluster 2 (containing 74 samples). The Kaplan–Meier survival curves for the 2 clusters are shown in Figure 1F, which indicates that patients in cluster 2 had poorer overall survival than those in cluster 1 (log-rank test, P < .05).
Figure 1.
Two subtypes in glioblastoma multiforme (GBM). (A) Consensus clustering matrix when k = 2. (B) Consensus clustering matrix when k = 3. (C) Consensus clustering matrix when k = 4. (D) Consensus clustering matrix when k = 5. (E) Cumulative distribution function plot with k valued 2 to 5. K = 2 was chosen as the optimal k value. All samples were divided into 2 groups, with cluster 1 containing 93 samples and cluster 2 containing 74 samples. (F) Kaplan–Meier survival curves were plotted according to 2 different clusters (P = .041).
3.2. Distinct differences in tumor microenvironment scores among GBM subtypes
Subtype-specific differences were observed in the tumor microenvironment scores of patients with GBM. Cluster 2 exhibited significantly higher immune scores (Fig. 2A), stromal scores (Fig. 2B), and ESTIMATE scores (Fig. 2C) compared to cluster 1. Conversely, tumor purity (Fig. 2D) was significantly lower in cluster 2 than in cluster 1. These findings suggest that the molecular subtypes related to OS exhibit distinct patterns of immune infiltration, stromal activation, and tumor purity within the tumor microenvironment. Cluster 2 may represent a subtype of “hot” tumor and cluster 1 may represent a subtype of “cold” tumor.
Figure 2.
Tumor microenvironment scores of patients with different subtypes of GBM, including immune score (A), stromal score (B), ESTIMATE score (C), and tumor purity (D). ***P < .001. GBM = glioblastoma multiforme.
3.3. Identification of DEGs between OS-related molecular subtypes
A total of 54 DEGs were identified between the 2 OS-related molecular subtypes, wherein 51 genes were up-regulated and 3 genes were down-regulated in cluster 2 (see Figure S1, Supplemental Content, http://links.lww.com/MD/L708; http://links.lww.com/MD/L709; which illustrates differential genes between clusters 1 and 2). The expressions of the DEGs are shown in Figure 3.
Figure 3.
Heat map of differentially expressed genes (DEGs) between cluster 1 and cluster 2 groups. The colors represent the gene expressions. The redder represents the higher gene expressions, and the bluer represents the lower expressions.
3.4. GO and KEGG enrichment of DEGs
The GO and KEGG enrichment analysis showed that DEGs were significantly enriched in cytokine/chemokine-related functions and pathways (Table 1). Regarding BP, DEGs were enriched in cytokine-mediated signaling pathway, cell chemotaxis, and humoral immune response (Fig. 4A). Concerning CC, DEGs were significantly enriched in vesicle lumen, cytoplasmic vesicle lumen, and secretory granule lumen (Fig. 4B). For molecular function, DEGs were significantly enriched in signaling receptor activator activity, cytokine activity, and chemokine receptor binding (Fig. 4C). In the KEGG pathway enrichment analysis, DEGs were mainly enriched in cytokine/chemokine-related pathways such as cytokine-cytokine receptor interaction, viral protein interaction with cytokine and cytokine receptor, and chemokine signaling pathway (Fig. 4D).
Table 1.
GO and KEGG enrichment analysis for DEGs between the 2 oxidative stress-related molecular subtypes of GBM.
| Ontology | ID | Description | GeneRatio | BgRatio | pvalue | p.adjust |
|---|---|---|---|---|---|---|
| BP | GO:0019221 | cytokine-mediated signaling pathway | 10/49 | 486/18800 | 4.04264E-07 | 0.000314922 |
| BP | GO:0071222 | cellular response to lipopolysaccharide | 6/49 | 217/18800 | 2.03959E-05 | 0.002269775 |
| BP | GO:0071219 | cellular response to molecule of bacterial origin | 6/49 | 229/18800 | 2.76159E-05 | 0.002689101 |
| BP | GO:0071216 | cellular response to biotic stimulus | 6/49 | 256/18800 | 5.14752E-05 | 0.004455464 |
| BP | GO:0050730 | regulation of peptidyl-tyrosine phosphorylation | 6/49 | 261/18800 | 5.73099E-05 | 0.004464445 |
| BP | GO:0060326 | cell chemotaxis | 6/49 | 315/18800 | 0.000160885 | 0.00997763 |
| BP | GO:0006959 | humoral immune response | 6/49 | 317/18800 | 0.000166507 | 0.00997763 |
| BP | GO:0032496 | response to lipopolysaccharide | 6/49 | 333/18800 | 0.000217316 | 0.010580584 |
| BP | GO:0002237 | response to molecule of bacterial origin | 6/49 | 354/18800 | 0.00030179 | 0.013347099 |
| BP | GO:0018108 | peptidyl-tyrosine phosphorylation | 6/49 | 373/18800 | 0.000398729 | 0.014091039 |
| BP | GO:0018212 | peptidyl-tyrosine modification | 6/49 | 376/18800 | 0.000416038 | 0.014091039 |
| BP | GO:0001819 | positive regulation of cytokine production | 6/49 | 475/18800 | 0.001404042 | 0.035282205 |
| BP | GO:0070098 | chemokine-mediated signaling pathway | 5/49 | 89/18800 | 3.43455E-06 | 0.001022439 |
| BP | GO:1990868 | response to chemokine | 5/49 | 97/18800 | 5.25001E-06 | 0.001022439 |
| BP | GO:1990869 | cellular response to chemokine | 5/49 | 97/18800 | 5.25001E-06 | 0.001022439 |
| BP | GO:0019730 | antimicrobial humoral response | 5/49 | 122/18800 | 1.60828E-05 | 0.002088085 |
| BP | GO:0097529 | myeloid leukocyte migration | 5/49 | 229/18800 | 0.000316338 | 0.013347099 |
| BP | GO:0030595 | leukocyte chemotaxis | 5/49 | 236/18800 | 0.000363233 | 0.013474221 |
| BP | GO:0006953 | acute-phase response | 4/49 | 48/18800 | 7.2827E-06 | 0.001134645 |
| BP | GO:0030593 | neutrophil chemotaxis | 4/49 | 106/18800 | 0.000166414 | 0.00997763 |
| BP | GO:0071347 | cellular response to interleukin-1 | 4/49 | 111/18800 | 0.000198721 | 0.010580584 |
| BP | GO:0002526 | acute inflammatory response | 4/49 | 113/18800 | 0.000212829 | 0.010580584 |
| BP | GO:0071621 | granulocyte chemotaxis | 4/49 | 128/18800 | 0.000342673 | 0.013347099 |
| BP | GO:1990266 | neutrophil migration | 4/49 | 128/18800 | 0.000342673 | 0.013347099 |
| BP | GO:0070555 | response to interleukin-1 | 4/49 | 141/18800 | 0.000494376 | 0.016046605 |
| BP | GO:0097530 | granulocyte migration | 4/49 | 154/18800 | 0.000688788 | 0.02146262 |
| BP | GO:0050731 | positive regulation of peptidyl-tyrosine phosphorylation | 4/49 | 190/18800 | 0.001501389 | 0.035698501 |
| BP | GO:0002573 | myeloid leukocyte differentiation | 4/49 | 210/18800 | 0.002163501 | 0.043636869 |
| BP | GO:0070374 | positive regulation of ERK1 and ERK2 cascade | 4/49 | 220/18800 | 0.002560274 | 0.048270506 |
| BP | GO:0007596 | blood coagulation | 4/49 | 221/18800 | 0.002602518 | 0.048270506 |
| BP | GO:0002548 | monocyte chemotaxis | 3/49 | 70/18800 | 0.00080561 | 0.024137314 |
| BP | GO:0061844 | antimicrobial humoral immune response mediated by antimicrobial peptide | 3/49 | 79/18800 | 0.001144835 | 0.030752632 |
| BP | GO:0042509 | regulation of tyrosine phosphorylation of STAT protein | 3/49 | 83/18800 | 0.001320494 | 0.034288835 |
| BP | GO:0007260 | tyrosine phosphorylation of STAT protein | 3/49 | 87/18800 | 0.00151226 | 0.035698501 |
| BP | GO:2001237 | negative regulation of extrinsic apoptotic signaling pathway | 3/49 | 97/18800 | 0.002065556 | 0.043488318 |
| BP | GO:0045639 | positive regulation of myeloid cell differentiation | 3/49 | 99/18800 | 0.002189375 | 0.043636869 |
| BP | GO:0045618 | positive regulation of keratinocyte differentiation | 2/49 | 19/18800 | 0.001106219 | 0.030752632 |
| BP | GO:0048245 | eosinophil chemotaxis | 2/49 | 19/18800 | 0.001106219 | 0.030752632 |
| BP | GO:0072677 | eosinophil migration | 2/49 | 23/18800 | 0.001625835 | 0.037250747 |
| BP | GO:0042730 | fibrinolysis | 2/49 | 24/18800 | 0.001770692 | 0.038315813 |
| BP | GO:0045672 | positive regulation of osteoclast differentiation | 2/49 | 24/18800 | 0.001770692 | 0.038315813 |
| BP | GO:0045606 | positive regulation of epidermal cell differentiation | 2/49 | 27/18800 | 0.002240661 | 0.043636869 |
| BP | GO:0050817 | coagulation | 4/49 | 226/18800 | 0.002820956 | 0.050743111 |
| BP | GO:0007599 | hemostasis | 4/49 | 227/18800 | 0.002866106 | 0.050743111 |
| BP | GO:0071356 | cellular response to tumor necrosis factor | 4/49 | 229/18800 | 0.002957891 | 0.051204384 |
| BP | GO:0045684 | positive regulation of epidermis development | 2/49 | 32/18800 | 0.00314011 | 0.052355128 |
| BP | GO:0050900 | leukocyte migration | 5/49 | 384/18800 | 0.003158782 | 0.052355128 |
| BP | GO:0050727 | regulation of inflammatory response | 5/49 | 394/18800 | 0.003525391 | 0.05721416 |
| BP | GO:0071346 | cellular response to interferon-gamma | 3/49 | 118/18800 | 0.003599049 | 0.05721754 |
| BP | GO:0034110 | regulation of homotypic cell-cell adhesion | 2/49 | 35/18800 | 0.00374816 | 0.058396327 |
| BP | GO:0046887 | positive regulation of hormone secretion | 3/49 | 122/18800 | 0.003952104 | 0.058631617 |
| BP | GO:0030851 | granulocyte differentiation | 2/49 | 36/18800 | 0.003962062 | 0.058631617 |
| BP | GO:0034612 | response to tumor necrosis factor | 4/49 | 249/18800 | 0.003989057 | 0.058631617 |
| BP | GO:0010951 | negative regulation of endopeptidase activity | 4/49 | 251/18800 | 0.004103966 | 0.059203514 |
| BP | GO:0001580 | detection of chemical stimulus involved in sensory perception of bitter taste | 2/49 | 37/18800 | 0.004181524 | 0.059225588 |
| BP | GO:0045616 | regulation of keratinocyte differentiation | 2/49 | 38/18800 | 0.004406518 | 0.061297811 |
| BP | GO:0010466 | negative regulation of peptidase activity | 4/49 | 262/18800 | 0.004776303 | 0.065276144 |
| BP | GO:0050913 | sensory perception of bitter taste | 2/49 | 41/18800 | 0.005114395 | 0.068691619 |
| BP | GO:0050912 | detection of chemical stimulus involved in sensory perception of taste | 2/49 | 43/18800 | 0.005613438 | 0.072855749 |
| BP | GO:0140353 | lipid export from cell | 2/49 | 43/18800 | 0.005613438 | 0.072855749 |
| BP | GO:0032103 | positive regulation of response to external stimulus | 5/49 | 442/18800 | 0.005723007 | 0.072855749 |
| BP | GO:0034341 | response to interferon-gamma | 3/49 | 140/18800 | 0.005798532 | 0.072855749 |
| BP | GO:0002920 | regulation of humoral immune response | 2/49 | 45/18800 | 0.006133914 | 0.075846335 |
| BP | GO:0030195 | negative regulation of blood coagulation | 2/49 | 48/18800 | 0.006954317 | 0.084647073 |
| BP | GO:1900047 | negative regulation of hemostasis | 2/49 | 49/18800 | 0.007238255 | 0.086747704 |
| BP | GO:2001236 | regulation of extrinsic apoptotic signaling pathway | 3/49 | 153/18800 | 0.007405644 | 0.087409037 |
| BP | GO:0050819 | negative regulation of coagulation | 2/49 | 52/18800 | 0.008121093 | 0.094422853 |
| BP | GO:0050832 | defense response to fungus | 2/49 | 53/18800 | 0.00842562 | 0.096522911 |
| BP | GO:0070372 | regulation of ERK1 and ERK2 cascade | 4/49 | 311/18800 | 0.008673637 | 0.097924106 |
| BP | GO:0043410 | positive regulation of MAPK cascade | 5/49 | 491/18800 | 0.008824661 | 0.098205873 |
| BP | GO:0002763 | positive regulation of myeloid leukocyte differentiation | 2/49 | 55/18800 | 0.009049882 | 0.099293772 |
| CC | GO:0034774 | secretory granule lumen | 5/52 | 322/19594 | 0.001604495 | 0.021630018 |
| CC | GO:0060205 | cytoplasmic vesicle lumen | 5/52 | 325/19594 | 0.001671128 | 0.021630018 |
| CC | GO:0031983 | vesicle lumen | 5/52 | 327/19594 | 0.001716668 | 0.021630018 |
| CC | GO:0035580 | specific granule lumen | 4/52 | 62/19594 | 2.19547E-05 | 0.001383144 |
| CC | GO:0042581 | specific granule | 4/52 | 160/19594 | 0.000854559 | 0.021630018 |
| CC | GO:0072562 | blood microparticle | 3/52 | 147/19594 | 0.006989466 | 0.044033633 |
| CC | GO:0034364 | high-density lipoprotein particle | 2/52 | 27/19594 | 0.002323846 | 0.024400387 |
| CC | GO:0034358 | plasma lipoprotein particle | 2/52 | 36/19594 | 0.004107893 | 0.032349661 |
| CC | GO:1990777 | lipoprotein particle | 2/52 | 36/19594 | 0.004107893 | 0.032349661 |
| CC | GO:0032994 | protein-lipid complex | 2/52 | 39/19594 | 0.004807198 | 0.033650383 |
| CC | GO:1904724 | tertiary granule lumen | 2/52 | 55/19594 | 0.00937694 | 0.053704294 |
| CC | GO:0031093 | platelet alpha granule lumen | 2/52 | 67/19594 | 0.013681747 | 0.071829171 |
| MF | GO:0048018 | receptor ligand activity | 11/52 | 489/18410 | 9.37357E-08 | 2.5695E-06 |
| MF | GO:0030546 | signaling receptor activator activity | 11/52 | 496/18410 | 1.08189E-07 | 2.5695E-06 |
| MF | GO:0005125 | cytokine activity | 10/52 | 235/18410 | 9.38406E-10 | 8.91486E-08 |
| MF | GO:0005126 | cytokine receptor binding | 9/52 | 272/18410 | 6.2058E-08 | 2.5695E-06 |
| MF | GO:0001664 | G protein-coupled receptor binding | 5/52 | 288/18410 | 0.001289457 | 0.013610932 |
| MF | GO:0008009 | chemokine activity | 4/52 | 49/18410 | 1.09137E-05 | 0.00020736 |
| MF | GO:0042379 | chemokine receptor binding | 4/52 | 71/18410 | 4.78064E-05 | 0.000756935 |
| MF | GO:0070851 | growth factor receptor binding | 4/52 | 139/18410 | 0.000636102 | 0.008632811 |
| MF | GO:0004866 | endopeptidase inhibitor activity | 4/52 | 180/18410 | 0.001659479 | 0.015765049 |
| MF | GO:0030414 | peptidase inhibitor activity | 4/52 | 187/18410 | 0.001907662 | 0.01647526 |
| MF | GO:0061135 | endopeptidase regulator activity | 4/52 | 194/18410 | 0.002180507 | 0.01726235 |
| MF | GO:0061134 | peptidase regulator activity | 4/52 | 230/18410 | 0.004018461 | 0.025364311 |
| MF | GO:0005539 | glycosaminoglycan binding | 4/52 | 234/18410 | 0.004271884 | 0.025364311 |
| MF | GO:0004867 | serine-type endopeptidase inhibitor activity | 3/52 | 98/18410 | 0.002675931 | 0.018158106 |
| MF | GO:0019955 | cytokine binding | 3/52 | 141/18410 | 0.007390222 | 0.041161546 |
| MF | GO:0045236 | CXCR chemokine receptor binding | 2/52 | 18/18410 | 0.001163066 | 0.013610932 |
| MF | GO:0004806 | triglyceride lipase activity | 2/52 | 26/18410 | 0.002435131 | 0.017795186 |
| MF | GO:0048020 | CCR chemokine receptor binding | 2/52 | 47/18410 | 0.00779903 | 0.041161546 |
| MF | GO:0008201 | heparin binding | 3/52 | 168/18410 | 0.011894226 | 0.059471131 |
| KEGG | hsa04060 | Cytokine-cytokine receptor interaction | 11/27 | 295/8164 | 8.87173E-10 | 4.25843E-08 |
| KEGG | hsa04061 | Viral protein interaction with cytokine and cytokine receptor | 7/27 | 100/8164 | 2.43185E-08 | 5.83645E-07 |
| KEGG | hsa04657 | IL-17 signaling pathway | 5/27 | 94/8164 | 1.20125E-05 | 0.0001922 |
| KEGG | hsa04630 | JAK-STAT signaling pathway | 5/27 | 162/8164 | 0.000163955 | 0.001967458 |
| KEGG | hsa04062 | Chemokine signaling pathway | 5/27 | 192/8164 | 0.000361762 | 0.003472915 |
| KEGG | hsa04610 | Complement and coagulation cascades | 3/27 | 85/8164 | 0.002660037 | 0.021280297 |
| KEGG | hsa05323 | Rheumatoid arthritis | 3/27 | 93/8164 | 0.003433903 | 0.022536547 |
| KEGG | hsa05150 | Staphylococcus aureus infection | 3/27 | 96/8164 | 0.003756091 | 0.022536547 |
BP = biological process, CC = cellular component, DEGs = differentially expressed genes, GBM = glioblastoma multiforme, GO = gene ontology, KEGG = Kyoto encyclopedia of genes and genomes, MF = molecular function.
Figure 4.
DEGs enrichment analysis. Top 10 Gene ontology (GO) enrichment bubble graphs for biological processes (BP) (A), cellular components (CC) (B), molecular functions (MF) (C), and KEGG enrichment bubble graphs (D). DEGs = differentially expressed genes, KEGG = Kyoto encyclopedia of genes and genomes.
3.5. Identification of hub genes by PPI network
To study if there were protein interactions among these DEGs, a PPI network was constructed based on the STRING database (Fig. 5A). According to node degree, the top 10 genes were considered hub genes of cluster 2 in GBM, namely colony stimulating factor 2 (CSF2), colony stimulating factor 3 (CSF3), C-C motif chemokine ligand 7 (CCL7), lipocalin 2 (LCN2), C-X-C motif chemokine ligand 6 (CXCL6), matrix metallopeptidase 8 (MMP8), C-C motif chemokine receptor 8 (CCR8), TNF superfamily member 11 (TNFSF11), interleukin 22 receptor subunit alpha 2 (IL22RA2) and orosomucoid 1 (ORM1). For better visualization, the interaction of hub genes was re-built using Cytoscape software (Fig. 5B). The degrees of hub genes are shown in Figure 5C.
Figure 5.
Protein-protein interaction (PPI) network and gene screening. (A) PPI network of DEGs. (B) Top 10 hub gene with highest degree. (C) The degree of hub genes. DEGs = differentially expressed genes.
3.6. Correlation analysis of immune factors and immune checkpoints with hub genes
The results of expression correlation analysis indicated that most immune factors and immune checkpoints were positively correlated with the hub genes (Figs. 6A-D). Among them, immunosuppressive factor IL10, immunostimulatory factor IL2RA, chemokine CCL2, and immune checkpoint leukocyte associated immunoglobulin like receptor 1 (LAIR1) were most positively correlated with hub gene CCL7. However, immunosuppressive factor and immune checkpoint V-set domain containing T cell activation inhibitor 1 (VTCN1), immune checkpoint lymphocyte activating 3 (LAG3), immunostimulatory factor TNF receptor superfamily member 13C (TNFRSF13C) and lymphotoxin alpha (LTA), and chemokine CXCL17 and CCL25 were negatively correlated with some hub genes.
Figure 6.
The correlation of immune factors and immune checkpoints with hub genes. (A–D) The correlation between hub genes and immune factors, including chemokines (A), immunostimulatory factors (B), immunosuppressive factor (C), immune checkpoints (D). Red represents positive correlations, and blue represents negative correlations. *P < .05; **P < .01; ***P < .001.
3.7. Identification of the potential drugs of GBM subtype 2
In the drug-gene interaction network, a total of 55 drugs or molecular compounds and corresponding genes (CSF2, MMP8, TNFSF11, and CSF3) were retrieved from the DGIdb database. No drugs or molecular compounds were found to correspond to other hub genes. CSF2 was found to be regulated by 41 drugs or molecular compounds, such as ACLARUBICIN, IDARUBICIN, and PROCARBAZINE. MMP8, TNFSF11, and CSF3 were found to be regulated by 7, 4, and 4 drugs or molecular compounds, respectively (Fig. 7).
Figure 7.
An interaction network of hub gene-small molecule drug. The blue nodes represent hub genes, and the white nodes represent small molecule drugs.
4. Discussion
OS has been shown to have a significant impact on glioma, with high levels of ROS affecting the function of glioma cells.[24] In this study, we identified 2 molecular subtypes of GBM, with patients in cluster 2 having a poorer prognosis and higher immune score. GO and KEGG enrichment analysis showed that DEGs were significantly enriched in cytokine/chemokine-related functions and pathways. Through PPI network, we identified ten hub genes (CSF2, CSF3, CCL7, LCN2, CXCL6, MMP8, CCR8, TNFSF11, IL22RA2, and ORM1) that could serve as potential therapeutic targets for GBM subtype 2 patients.
ROS play a crucial role in regulating cell survival, apoptosis, proliferation, and intracellular environmental stability through normal cellular metabolism. When there is an imbalance in ROS, glioma cells can experience OS, which is associated with glioma cell development and drug resistance.[25] Previous research has also highlighted the critical value of OS as a tumor marker, influencing intra- and intercellular molecular heterogeneity within tumors and producing immunomodulatory escape or resistance to conventional therapies.[26] The dynamic nature of its gene expression pattern may contribute to early diagnosis and assessment of GBM.[27]
In our study, PPI network analysis identified 10 hub genes (CSF2, CSF3, CCL7, LCN2, CXCL6, MMP8, CCR8, TNFSF11, IL22RA2, and ORM1) that could be potential targets for immune-targeted therapy in GBM. These genes have been previously reported to be associated with GBM, and targeting the signaling pathways they are involved in may be a promising approach for treating GBM. For example, CSF2 stimulates the recruitment of microglia and polarization of macrophages, which was suggested to be linked to tumor progression and unfavorable prognosis in GBM.[28] Consequently, targeting macrophage infiltration holds promise as a potential strategy to enhance outcomes for GBM patients. Additionally, CCL7 mediates immune cell recruitment and regulates the formation of the tumor immune microenvironment.[29] Upregulated chemokines such as CCL7, CXCL13, and CCL18 potentially promoted GBM tumor progression and affected poor prognosis in GBM patients.[30] LCN2 acts as an immunomodulator and is involved in various biological responses and pathophysiological processes, including OS.[31] Additionally, LCN2 can affect the development and progression of multiple malignancies. In prostate cancer, the invasive ability of tumor cells is positively correlated with LCN2 expression, promoting epithelial-mesenchymal transition (EMT) and increasing tumor cell invasiveness.[32] TNFSF11 (RANKL), a member of the tumor necrosis factor family, regulates osteoclast development and bone metabolism.[33] Kim et al found that RANKL could regulate the glioma tumor microenvironment to promote glioma cell invasion and blocking RANKL-RANK signaling can retard GBM cell growth and prolong patient survival time.[34]
Based on the hub genes identified, we also screened several small molecule drugs for their potential therapeutic effects in GBM. For example, Letrozole has been studied as a treatment for GBM and may inhibit cell proliferation by inhibiting hERG potassium channels.[35] Epigallocatechin gallate (EGCG) is another promising drug, as it can inhibit the mRNA and protein expression of MGMT and enhance the cytotoxicity of TMZ in GBM.[36] DOXYCYCLINE, a commonly used antibiotic, has been reported to possess anti-tumor properties and the ability to reduce inflammatory mediator secretion in macrophage.[37] ACLARUBICIN had been investigated as a potential chemotherapeutic agent in rats.[38] IDARUBICIN is lipophilic and cytotoxic in glioma cell lines.[39] PROCARBAZINE treatment after radiation therapy at the time of initial diagnosis resulted in longer progression-free survival.[40] These drugs may exert their anti-inflammatory actions by inhibiting key signaling pathways involved in inflammation and immune modulation.
Despite these findings, our study has some limitations. Firstly, it is important to note that our research primarily relies on the analysis of RNA-seq data for the identification of a novel molecular subtype of GBM. While our bioinformatics analysis pipeline is comprehensive and rigorous, it is essential to recognize that it lacks direct experimental validation. Due to the large-scale nature of RNA-seq data, conducting ex vivo experiments to validate our findings within the scope of this study is not feasible. Therefore, it is crucial to interpret our results with caution and consider them as exploratory findings that warrant further experimental investigation. Secondly, additional clinical databases were not included for external validation. Lastly, the effectiveness of the screened small molecule drugs still requires further investigation.
5. Conclusions
In conclusion, our study aimed to identify molecular subtypes related to OS in GBM and characterize their differences in immune infiltration levels and overall survival. We successfully identified 2 GBM subtypes with distinct characteristics, where cluster 2 showed poorer overall survival and higher immune infiltration levels compared to cluster 1. Through PPI network, we identified ten hub genes (CSF2, CSF3, CCL7, LCN2, CXCL6, MMP8, CCR8, TNFSF11, IL22RA2, and ORM1) that could serve as potential therapeutic targets for GBM subtype 2 patients. Additionally, we found 55 small molecule drugs from the DGIdb database that may have therapeutic effects on GBM subtype 2 patients. Utilizing nanomedicine to deliver therapeutic agents directly to the site of the brain tumor may be a more efficient treatments. These findings highlight the importance of exploring alternative treatment options for GBM subtype 2 patients and provide insights into developing personalized therapies in the future.
Acknowledgments
The authors sincerely thank Bo Li for his valuable contributions to the statistical analysis and interpretation of the data presented in this manuscript, which greatly contributed to the accuracy and reliability of our findings.
Author contributions
Conceptualization: Xiaohong Hou.
Investigation: Xiaohu Li.
Methodology: Xiaohong Hou.
Validation: Yong Yu.
Visualization: Xiaohu Li.
Software: Yong Yu.
Supervision: Xuecheng Ge.
Writing – original draft: Guanyou Huang.
Writing – review & editing: Hongchuan Gan.
Supplementary Material
Abbreviations:
- BP
- biological process
- CC
- cellular component
- DEGs
- differentially expressed genes
- DGIdb
- drug gene interaction database
- GBM
- glioblastoma multiforme
- GO
- gene ontology
- KEGG
- Kyoto encyclopedia of genes and genomes
- OS
- oxidative stress
- PPI
- protein-protein interaction
- ROS
- reactive oxygen species
The datasets generated during and/or analyzed during the current study are publicly available.
Due to the public availability of data on the GEO database, ethical approval or informed consent was not required for this study.
The authors have no conflicts of interest to disclose.
Supplemental Digital Content is available for this article.
The work was supported by Science and Technology Foundation of Guizhou Provincial Health Commission, China [gzwkj2022-348].
How to cite this article: Huang G, Hou X, Li X, Yu Y, Ge X, Gan H. Identification of a novel glioblastoma multiforme molecular subtype with poor prognosis and high immune infiltration based on oxidative stress-related genes. Medicine 2024;103:7(e35828).
Contributor Information
Guanyou Huang, Email: hgy024410@yeah.net.
Xiaohong Hou, Email: 1549460390@qq.com.
Xiaohu Li, Email: 1057874261@qq.com.
Yong Yu, Email: 695046174@qq.com.
Xuecheng Ge, Email: 42241413@qq.com.
References
- [1].Ostrom QT, Gittleman H, Truitt G, et al. CBTRUS Statistical Report: primary brain and other central nervous system tumors diagnosed in the United States in 2011-2015. Neuro Oncol. 2018;20(suppl_4):iviv1–iv86. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [2].Taylor OG, Brzozowski JS, Skelding KA. Glioblastoma multiforme: an overview of emerging therapeutic targets. Front Oncol. 2019;9:963. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [3].Bilmin K, Kujawska T, Grieb P. Sonodynamic therapy for gliomas. perspectives and prospects of selective sonosensitization of glioma cells. Cells. 2019;8:1428. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [4].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. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [5].Tan AC, Ashley DM, Lopez GY, et al. Management of glioblastoma: state of the art and future directions. CA Cancer J Clin. 2020;70:299–312. [DOI] [PubMed] [Google Scholar]
- [6].Neganova ME, Aleksandrova YR, Sukocheva OA, et al. Benefits and limitations of nanomedicine treatment of brain cancers and age-dependent neurodegenerative disorders. Semin Cancer Biol. 2022;86(Pt 2):805–33. [DOI] [PubMed] [Google Scholar]
- [7].Bahadur S, Sahu AK, Baghel P, et al. Current promising treatment strategy for glioblastoma multiform: A review. Oncol Rev. 2019;13:417. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [8].Pisoschi AM, Pop A. The role of antioxidants in the chemistry of oxidative stress: a review. Eur J Med Chem. 2015;97:55–74. [DOI] [PubMed] [Google Scholar]
- [9].Krawczynski K, Godlewski J, Bronisz A. Oxidative stress-part of the solution or part of the problem in the hypoxic environment of a brain tumor. Antioxidants (Basel). 2020;9:747. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [10].Ighodaro OM. Molecular pathways associated with oxidative stress in diabetes mellitus. Biomed Pharmacother. 2018;108:656–62. [DOI] [PubMed] [Google Scholar]
- [11].Jelic MD, Mandic AD, Maricic SM, et al. Oxidative stress and its role in cancer. J Cancer Res Ther. 2021;17:22–8. [DOI] [PubMed] [Google Scholar]
- [12].Steven S, Frenis K, Oelze M, et al. Vascular inflammation and oxidative stress: major triggers for cardiovascular disease. Oxid Med Cell Longev. 2019;2019:7092151. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [13].Salim S. Oxidative stress and the central nervous system. J Pharmacol Exp Ther. 2017;360:201–5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [14].Soon BH, Abdul Murad NA, Then SM, et al. Mitochondrial DNA mutations in Grade II and III glioma cell lines are associated with significant mitochondrial dysfunction and higher oxidative stress. Front Physiol. 2017;8:231. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [15].Stelzer G, Rosen N, Plaschkes I, et al. The GeneCards suite: from gene data mining to disease genome sequence analyses. Curr Protoc Bioinformatics. 2016;54:1 30 31–31 30 33. [DOI] [PubMed] [Google Scholar]
- [16].Wilkerson MD, Hayes DN. ConsensusClusterPlus: a class discovery tool with confidence assessments and item tracking. Bioinformatics. 2010;26:1572–3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [17].Lei C, Chen W, Wang Y, et al. Prognostic prediction model for glioblastoma: a metabolic gene signature and independent external validation. J Cancer. 2021;12:3796–808. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [18].Yoshihara K, Shahmoradgoli M, Martinez E, et al. Inferring tumour purity and stromal and immune cell admixture from expression data. Nat Commun. 2013;4:2612. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [19].Love MI, Huber W, Anders S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol. 2014;15:550. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [20].Yu G, Wang LG, Han Y, et al. clusterProfiler: an R package for comparing biological themes among gene clusters. OMICS. 2012;16:284–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [21].Szklarczyk D, Gable AL, Lyon D, et al. STRING v11: protein-protein association networks with increased coverage, supporting functional discovery in genome-wide experimental datasets. Nucleic Acids Res. 2019;47(D1):D607–13. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [22].Ru B, Wong CN, Tong Y, et al. TISIDB: an integrated repository portal for tumor-immune system interactions. Bioinformatics. 2019;35:4200–2. [DOI] [PubMed] [Google Scholar]
- [23].Freshour SL, Kiwala S, Cotto KC, et al. Integration of the drug-gene interaction database (DGIdb 4.0) with open crowdsource efforts. Nucleic Acids Res. 2021;49(D1):D1144–51. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [24].Ostrowski RP, Pucko EB. Harnessing oxidative stress for anti-glioma therapy. Neurochem Int. 2022;154:105281. [DOI] [PubMed] [Google Scholar]
- [25].Silber JR, Bobola MS, Blank A, et al. The apurinic/apyrimidinic endonuclease activity of Ape1/Ref-1 contributes to human glioma cell resistance to alkylating agents and is elevated by oxidative stress. Clin Cancer Res. 2002;8:3008–18. [PubMed] [Google Scholar]
- [26].Da Ros M, De Gregorio V, Iorio AL, et al. Glioblastoma chemoresistance: the double play by microenvironment and blood-brain barrier. Int J Mol Sci . 2018;19:2879. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [27].Lu D, Yang N, Wang S, et al. Identifying the predictive role of oxidative stress genes in the prognosis of glioma patients. Med Sci Monit. 2021;27:e934161. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [28].Sielska M, Przanowski P, Pasierbinska M, et al. Tumour-derived CSF2/granulocyte macrophage colony stimulating factor controls myeloid cell accumulation and progression of gliomas. Br J Cancer. 2020;123:438–48. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [29].Lee YS, Cho YB. CCL7 Signaling in the tumor microenvironment. Adv Exp Med Biol. 2020;1231:33–43. [DOI] [PubMed] [Google Scholar]
- [30].Zhu Z, Zhang X, Yu Z, et al. Correlation of Tim-3 expression with chemokine levels for predicting the prognosis of patients with glioblastoma. J Neuroimmunol. 2021;355:577575. [DOI] [PubMed] [Google Scholar]
- [31].Ferreira AC, Sousa N, Bessa JM, et al. Metabolism and adult neurogenesis: towards an understanding of the role of lipocalin-2 and iron-related oxidative stress. Neurosci Biobehav Rev. 2018;95:73–84. [DOI] [PubMed] [Google Scholar]
- [32].Ding G, Fang J, Tong S, et al. Over-expression of lipocalin 2 promotes cell migration and invasion through activating ERK signaling to increase SLUG expression in prostate cancer. Prostate. 2015;75:957–68. [DOI] [PubMed] [Google Scholar]
- [33].Ono T, Hayashi M, Sasaki F, et al. RANKL biology: bone metabolism, the immune system, and beyond. Inflamm Regen. 2020;40:2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [34].Kim JK, Jin X, Sohn YW, et al. Tumoral RANKL activates astrocytes that promote glioma cell invasion through cytokine signaling. Cancer Lett. 2014;353:194–200. [DOI] [PubMed] [Google Scholar]
- [35].Shugg T, Dave N, Amarh E, et al. Letrozole targets the human ether-a-go-go-related gene potassium current in glioblastoma. Basic Clin Pharmacol Toxicol. 2021;128:357–65. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [36].Xie CR, You CG, Zhang N, et al. Epigallocatechin gallate preferentially inhibits O(6)-Methylguanine DNA-Methyltransferase expression in glioblastoma cells rather than in nontumor glial cells. Nutr Cancer. 2018;70:1339–47. [DOI] [PubMed] [Google Scholar]
- [37].Cazalis J, Bodet C, Gagnon G, et al. Doxycycline reduces lipopolysaccharide-induced inflammatory mediator secretion in macrophage and ex vivo human whole blood models. J Periodontol. 2008;79:1762–8. [DOI] [PubMed] [Google Scholar]
- [38].Lu W, Wan J, Zhang Q, et al. Aclarubicin-loaded cationic albumin-conjugated pegylated nanoparticle for glioma chemotherapy in rats. Int J Cancer. 2007;120:420–31. [DOI] [PubMed] [Google Scholar]
- [39].Boogerd W, Tjahja IS, van de Sandt MM, et al. Penetration of idarubicin into malignant brain tumor tissue. J Neurooncol. 1999;44:65–9. [DOI] [PubMed] [Google Scholar]
- [40].Buckner JC, Shaw EG, Pugh SL, et al. Radiation plus procarbazine, CCNU, and vincristine in low-grade glioma. N Engl J Med. 2016;374:1344–55. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.