Identification and validation of ion channels-related mRNA prognostic signature for glioblastomas Authors: Chao Huang, Xue-Bin Yu, Yong-Zhi Zhou, Wu-Qiao Bao Published: 2024 ABSTRACT Glioblastomas (GBM) is a kind of malignant brain tumor with poor prognosis. Identifying new biomarkers is promising for the treatment of GBM. The mRNA-seq and clinical data were obtained from The Cancer Genome Atlas and the Chinese Glioma Genome Atlas databases. The differentially expressed genes were identified using limma R package. The prognosis-related genes were screened out and a risk model was constructed using univariate, least absolute shrinkage and selection operator, and multivariate Cox analysis. Receiver operating characteristic curve was used to assess the efficiency of model. Kaplan–Meier survival curve was applied for the survival analysis. Mutation analysis was conducted using maftools package. The effect of immunotherapy was analyzed according to TIDE score, and the drug sensitivity analysis was performed. The Gene Ontology, Kyoto Encyclopedia of Genes and Genomes, and Gene Set Enrichment Analysis enrichment analyses were performed for the functional analysis. The regulatory network was constructed by STRING and Cytoscape software. RT-qPCR was performed to validate the expression of 3 hub genes in vitro. A risk model was constructed based on 3 ion channels related genes (gap junction protein beta 2 [GJB2], potassium voltage-gated channel subfamily h member 6 [KCNH6], and potassium calcium-activated channel subfamily n member 4 [KCNN4]). The risk score and hub genes were positively correlated with the calcium signaling pathway. Patients were divided into 2 groups based on the risk score calculated by 3 signatures. The infiltration levels of T cell, B lineage, monocytic lineage, and neutrophils were increased in high risk group, while TIDE score was decreased. IC50 of potential drugs for GBM treatment was elevated in the high risk group. Furthermore, GJB2, KCNH6, and KCNN4 were oncogenic, and GJB2 and KCNN4 were upregulated, while KCNH6 was downregulated in high risk group and GBM cells. The regulatory network showed that KCNH6 was targeted by more miRNA and transcription factors and KCNN4 interacted with more drugs. We constructed a three-signature risk model, which could effectively predict the prognosis of GBM development. Besides, KCNH6 and KCNN4 were respectively considered as the targets of molecular targeted treatment and chemotherapy. 1. Introduction Glioma is the most prevalent primary malignant tumor originating from the central nervous system, and it is characterized by high recurrence and mortality rates.[1,2] The survival probability of patients diagnosed with glioma greatly depends on the glioma grade.[3] According to the histology characteristics, glioma is usually classified into grade II to IV.[1] The grade II gliomas were considered as low-grade glioma (LGG), and grade III was an anaplastic glioma, and grade IV glioma was considered as high-grade glioma or glioblastomas (namely GBM).[4,5] The survival time of LGG and anaplastic glioma could be longer than 5 years, but the survival time of GBM is shorter than 1 year.[1] Previous studies have proved that targeted therapy was useful and promising for tumor treatment,[6,7] but most molecular targeted drugs failed in the treatment of GBM because of their high invasiveness and drug resistance.[5] Thus, identifying more biomarkers and targeted drugs was meaningful for the GBM treatment. Ion channels are pore-forming transmembrane proteins, which could regulate the passive movement of various ions,[8] and the main types of ion channels include potassium (K+), calcium (Ca2+), chloride (Cl‐), and sodium (Na+).[9] Previous studies have proven that the abnormally expressed ion channel-related genes were usually observed in tumor tissues,[10] and the transmembrane transport of ions plays an essential role in cellular processes, including cell proliferation, migration, and apoptosis.[11] Moreover, the ion channel genes could predict a multiple cancer prognosis, including breast cancer, lung adenocarcinoma, and prostate cancer.[12–14] Nevertheless, little research was conducted on the prediction value of ion channel-related genes in GBM development. Thus, in this study, we tried to identify some biomarkers and establish a risk model, which could predict the prognosis of GBM. Besides, we also tried to explore the potential mechanism of ion channels in the development of GBM. 2. Methods 2.1. Data collection The mRNA expression data and corresponding clinical data were downloaded from The Cancer Genome Atlas (TCGA, https://portal.gdc.cancer.gov/) database. Totally 169 patients were diagnosed with GBM, and 534 patients were diagnosed with LGG. A dataset including 139 GBM samples, from the Chinese Glioma Genome Atlas (http://www.cgga.org.cn/), was used as the validation set. The ion channel-related genes and immune escape-related genes were obtained from previous studies.[15,16] The differentially expressed genes (DEGs) between the 2 groups were identified using the limma R package. The P-value < .05 and the fold change ≥ 1.5 were the threshold of significant genetic difference. The results were visualized using the ggplot2 R package. 2.2. The construction and validation of the GBM prognosis risk model After the DEGs were screened out, the univariate Cox analysis was performed to select genes which were closely related to the prognosis of GBM. Then, the least absolute shrinkage and selection operator Cox analysis was executed to shrink and select variables using the glmmet R package. Next, the hub genes with independent prediction values and corresponding coefficients were determined using multivariate Cox regression. Furthermore, we constructed the risk model based on the expressions and coefficients of hub genes. The receiver operating characteristic (ROC) curve was used for the validation of the risk model using the timeROC package in R software. Finally, some clinical characteristics and risk scores were integrated to construct a nomogram for the clinical application using rms and survival packages. 2.3. Genomic alteration analysis The simple nucleotide variation data of GBM patients were obtained using the “TCGAbiolinks” package, and the tumor mutation burden was calculated using the maftools package. The mutation landscape of GBM patients was visualized by the waterfall function of the maftools package. 2.4. The immune infiltration and immune treatment analysis The proportions of 22 immune cell subtypes in the TCGA data were evaluated by the CIBERSORT R package which contained the marker genes of 22 immune cells as same as the data file from CIBERSORTX (https://cibersortx.stanford.edu/). Besides, the results were visualized using the ggplot2 package. 2.5. Survival analysis After the samples were divided into different groups, the Kaplan–Meier survival analysis was performed in R software (version 4.1.0) using the survival package, and the results were visualized using the survminer R package. 2.6. The function enrichment analysis After the data of gene expression was imported into R software, the enrichGO and enrichKEGG function of clusterProfiler R package was applied to perform Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis. Besides, the GO and KEGG terms with P < .05 were considered as statistically significant, and the results were visualized using ggplot2. The Gene Set Enrichment Analysis (GSEA) analysis was conducted in GSEA software (version 4.3.2). The number of permutations was 1000. The Enrichment statistic was weight. The pathways significantly enriched were adopted as adjusted P < .05. 2.7. Drug sensitivity analysis We utilized the R package “pRRophetic”[17] to estimate the IC50 values of potentially administered chemotherapeutic drugs for GBM. Differential analyses of the IC50 values between different groups were performed to evaluate potential differences in treatment efficacy. 2.8. The selection of hub genes and construction of regulatory network Firstly, we constructed the protein–protein interaction network using the STRING database (https://string-db.org/). Then, the interaction between every core gene was imported into Cytoscape. After the miRNA, transcription factor (TF), and drugs-targeted core genes were predicted by the CyTargetLinker tool, the regulatory network was constructed. 2.9. RT-qPCR Total RNA was extracted from the cells using the TRIzol reagent (Thermo Fisher Scientific, Waltham, MA). The RNA was converted to cDNA and qRT-PCR was performed using EXPRESS One-Step Superscript™ qRT-PCR (1178101K, Thermo Fisher Scientific, Waltham, MA) in 7300 PCR system (Thermo Fisher Scientific). The primers were shown in Table 1. The following cycling conditions were applied: 95 or 5 minutes, followed by 40 cycles of 95 °C for 20 seconds and 60 °C for 30 seconds. For each sample, qPCR assays were conducted in triplicate in a 10 mL reaction volume. b-Actin served as an internal control to normalize the expression of PLAU. The 2-Ct method was employed to calculate the relative expression of TPBG mRNA. Table 1The prime sequence used in this study.GeneForward (5’–3’)Reverse (5’–3’)GJB2TGGACCTACACAAGCAGCATCTTCCCAGGCGTTGCACTTCACCAGKCNH6GCTCAGCGATCAGGACGAATCACTTGACATAGGTGGTGCGGAAGTTGKCNN4GCCATGCTGCTGCGTCTCTACGGCGGAAGCGGACTTGATTGAGGJB2 = gap junction protein beta 2, KCNN4 = potassium calcium-activated channel subfamily n member 4, KCNH6 = potassium voltage-gated channel subfamily h member 6. The prime sequence used in this study. GJB2 = gap junction protein beta 2, KCNN4 = potassium calcium-activated channel subfamily n member 4, KCNH6 = potassium voltage-gated channel subfamily h member 6. 2.10. Statistical analysis In this study, all statistics analyses were performed in SPSS 25, GraphPad Prism (version 10.1.2) and R software (version 4.3.0). Student t test was used to compare the differences between the 2 groups. Log-rank test was used to compare the survival probability in the Kaplan–Meier curves. Pearson correlation was used to analyze the linear relationship between 2 variables. The P-value < .05 was considered statistically significant. 3. Results 3.1. The identification and functional enrichment of DEGs We compared the expression profile between LGG and GBM and 7453 DEGs were identified including 4474 upregulated genes and 2979 downregulated genes (Fig. 1A). Besides, 207 ion channel-related DEGs were determined using a Venn plot (Fig. 1B). The GO enrichment analysis showed that these genes were enriched in calcium ion transport, chloride transmembrane transport and potassium ion transmembrane transport in the aspect of BP, and they were mainly involved in the function of monoatomic channel activity, gated channel activity and monoatomic cation channel activity (Fig.1C). Additionally, these genes mainly performed the function in the cation channel complex, monoatomic ion channel complex, and postsynaptic membrane (Fig. 1C). Furthermore, the KEGG results exhibited that the major enriched pathways were the neuroactive ligand–receptor interaction, cAMP signaling pathway, and calcium signaling pathway (Fig. 1C). The determination of hub genes and enrichment analysis. (A) The volcano plots showed the DEGs between LGG and GBM. (B) The Venn plot displayed the intersection of DEGs and ion channel genes. (C) GO and KEGG enrichment analyses were performed on the 207 overlapping genes. DEGs = differentially expressed genes, GBM = glioblastomas, GO = Gene Ontology, LGG = low-grade glioma, KEGG = Kyoto Encyclopedia of Genes and Genomes. 3.2. The construction and validation of IC-related risk model At first, we identified 22 genes related to the overall survival (OS) in GBM based on the intersection of DEGs and ion channel genes and univariate Cox analysis. Then, these genes were used for the least absolute shrinkage and selection operator regression analysis based on 10-fold cross-validation. As shown in Figure 2A, after λ = 0.16 was determined as the optimal value, and 6 genes with nonzero coefficients were obtained (Fig. 2B). Finally, 3 genes, namely gap junction protein beta 2 (GJB2), potassium voltage-gated channel subfamily h member 6 (KCNH6), and potassium calcium-activated channel subfamily n member 4 (KCNN4), with independent prediction value and their coefficients were determined using multivariate Cox analysis, and the risk model was constructed based on their expression and coefficients as follow: risk score = 0.008* exp (GJB2) + 0.103* exp (KCNH6) + 5.059* exp (KCNN4). The survival analysis showed that the survival probability and time of the high risk group were significantly lower than that of the low risk group (Fig. 2C). The ROC curve displayed that the risk model had good discrimination with AUC of 0.75, 0.76 and 0.77 for 1-, 2- and 3-year survival (Fig. 2D). We then validated the model’s performance in an external dataset. As shown in Fig. 2E, GBM patients with high risk scores showed poorer prognoses than those with low risk scores (P < .001). Additionally, AUC values at 1-, 2-, and 3-year survival for the risk model were 0.62, 0.66, and 0.64, respectively (Fig. 2F). The construction and validation of the risk model. (A) The 10-fold cross-validation for parameter selection in the LASSO model. (B) Six core genes with nonzero coefficient in GBM. (C) The K–M plot compared the OS of TCGA-GBM patients between high and low risk groups. (D) The ROC curve of predicting prognosis in GBM based on the TCGA-GBM dataset. (E) K–M curve for the OS of GBM patients between high and low risk groups based on CGGA dataset. (F) The ROC curve of predicting prognosis in GBM based on CGGA dataset. CGGA = Chinese Glioma Genome Atlas, GBM = glioblastomas, GO = Gene Ontology, K–M = Kaplan–Meier, OS = overall survival, LASSO = least absolute shrinkage and selection operator, ROC = receiver operating characteristic, TCGA = The Cancer Genome Atlas. 3.3. The mutation landscape of high and low risk groups of GBM patients By plotting mutation waterfall maps of different risk groups, the correlation between risk score and mutations of some genes could be clearly observed. From Fig. 3, missense mutation was the most common type of mutation for most genes. Additionally, it can be seen that PTEN, TP53, and EGFR had the highest mutation frequencies in the high risk group (Fig. 3A), while the genes with the top 3 mutation frequencies in low risk group were TP53, TTN, and EGFR (Fig. 3B). Comparing with gene mutation between 2 groups, it was obvious that mutation frequency of AHNAK2, FLG, PIK3R1, and PIK3CA were significantly increased in the high risk group. The mutation landscape of GBM patients in the high and low risk groups. (A) The mutation landscape of GBM patients with high risk score. (B) The mutation landscape of GBM patients with low risk score. GBM = glioblastomas. 3.4. The different functions between high and low risk groups Next, we identified the DEGs between low risk group and the high risk group (Fig. 4A). The GO enrichment analysis revealed that DEGs were enriched in transport, response to chemical and cellular development process, and they performed functions in the intrinsic component of member, extracellular region, and endomembrane system (Fig. 4B). In addition, the function of them mainly included signaling receptor activity, molecular transducer activity, and signaling receptor binding (Fig. 4B). Moreover, the KEGG results exhibited that DEGs were involved in the cytokine–cytokine receptor interaction, calcium signaling pathway, and PI3K–Akt signaling pathway (Fig. 4C). The functional enrichment of DEGs between low and high risk groups. (A) The volcano plot showed the DEGs between low and high risk groups. (B) The GO enrichment results of DEGs between low and high risk groups. (C) The KEGG enrichment results of DEGs between low and high risk groups. DEGs = differentially expressed genes, GO = Gene Ontology, KEGG = Kyoto Encyclopedia of Genes and Genomes. Combined with the KEGG results based on ion channel-related DEGs, calcium signaling pathway was the shared pathway, we then focused on the calcium signaling pathway and analyzed its association with the risk score. As shown in Figure S1A, Supplemental Digital Content, http://links.lww.com/MD/O77, the risk score was significantly positively related to the calcium signaling pathway. Additionally, we analyzed the correlation of 3 hub genes with the calcium signaling pathway. Except for GJB2, KCNH6, and KCNN4 were both significantly positively associated with this pathway (Figure S1B–D, Supplemental Digital Content, http://links.lww.com/MD/O77). These results revealed that indicating that the calcium signaling pathway may be more active in patients with higher risk scores or higher expression levels of KCNH6 and KCNN4. 3.5. The analysis of immune microenvironment and drug sensitivity As shown in Fig. 5A, the immune infiltration levels of T cell, B lineage, monocytic lineage, and neutrophils were significantly increased in high risk group compared with the levels in the low risk group. Besides, to better assess the value of the risk model in predicting immunotherapy response, we evaluated the TIDE scores in GBM patients based on the TCGA data (Fig. 5B). The results revealed that the TIDE score was lower in the high risk group. Besides, the TIDE score was significantly and negatively related to the infiltration of T cells, B lineage, monocytic lineage, and neutrophils (Fig. 5C–F). In addition, Figure 5G exhibited that neutrophils were closely related to most immune escape-related genes which were differentially expressed in the high and low risk groups. The immune microenvironment analysis. (A) The infiltration level of immune cells with significant change. (B) The TIDE score in low and high risk groups. (C–F) The correlation between TIDE score and 4 immune cells. (G) The association of TIDE and 4 immune cells with immune escape-related genes. To explore the relationship between risk scores and GBM medications, we compared differences in common potential therapeutic agents for GBM between different risk groups. The results showed that, except for Afatinib_1377, other drugs including Bryostatin 1_197, Midostaurin_153, Ponatinib_155, Afatinib_1032, and Elesclomol_1031 had significantly higher IC50 in the high risk group (Figure S2A, Supplemental Digital Content, http://links.lww.com/MD/O78). This suggests that drug resistance is stronger in the high risk group, which may lead to a worse prognosis for drug therapy. Subsequently, the drug sensitivity of potential drugs for GBM between different expression groups of 3 hub genes was analyzed. As shown in Figure S2B, Supplemental Digital Content, http://links.lww.com/MD/O78, only Afatinib_1377’s sensitivity was elevated in the GBJ2 high expression group. In the KCNH6 high expression group, sensitivities of Bryostatin 1_197, Midostaurin_153, and Ponatinib_155 were significantly higher than those in the KCNH6 low expression group (Figure S2C, Supplemental Digital Content, http://links.lww.com/MD/O78). As for KCNN4, IC50 values of Ponatinib_155 and Afatinib_1032 were higher in the high expression group than in the low expression group (Figure S2D, Supplemental Digital Content, http://links.lww.com/MD/O78). These results indicated that GBM patients in the high risk group and those with high expression of hub genes exhibit stronger drug resistance to these drugs, which may lead to poorer prognosis in drug therapy. 3.6. The expression and survival analysis of hub genes Next, we explore the role of hub genes in the development of GBM. Comparing the expression of 3 hub genes in GBM and LGG, we found that the expression of GJB2 and KCNN4 were significantly upregulated in GBM, while KCNH6 expression was significantly downregulated (Fig. 6A). The same results were observed in the external dataset, with GBM patients having significant higher GJB2 and KCNN4 expression (Fig. 6B). KCNH6 was downregulated in GBM but with no significance (Fig. 6B). The survival analysis displayed that the high expressions of GJB2, KCNH6, and KCNN4 were closely related to poor prognosis (Fig. 6C–E). To validate these findings, expression levels of 3 hub genes were detected in vitro. Compared to those in the normal cells, levels of GJB2 and KCNN4 were elevated in the GBM cells while KCNH6 levels were reduced in the GBM cells (Fig. 6F). These results were coincidence with the bioinformatic analysis. The expression and prognosis value of GJB2, KCNH6, and KCNN4. (A) The expression of GJB2, KCNH6, and KCNN4 in GBM and LGG based on the TCGA-GBM dataset; ***P < .001. (B) The expression of GJB2, KCNH6, and KCNN4 in GBM and LGG based on the CGGA dataset; ****P < .0001. (C) Survival curve of OS in high and low expression of GJB2 in GBM. (D) Survival curve of OS in high and low expression of KCNH6 in GBM. (E) Survival curve of OS in high and low expression of KCNN4 in GBM. (F) Expression levels of GJB2, KCNH6 and KCNN4 in normal cells HEB and GBM cells U251. CGGA = Chinese Glioma Genome Atlas, GBM = glioblastomas, GJB2 = gap junction protein beta 2, KCNN4 = potassium calcium-activated channel subfamily n member 4, KCNH6 = potassium voltage-gated channel subfamily h member 6, LGG = low-grade glioma, OS = overall survival, TCGA = The Cancer Genome Atlas. 3.7. The functional enrichment of hub genes To further explore the potential mechanism of hub genes, the GSEA enrichment analysis was performed. The results exhibited that GJB2 was enriched in IL6/JAK/STAT3 signaling pathway, inflammatory response and TNFA signaling via NFKB (Fig. 7A–C). KCNH6 was mainly involved in hedgehog signaling, KRAS signaling and interferon gamma response (Fig. 7D–F). Furthermore, KCNN4 was enriched in epithelial-mesenchymal transition, PI3K/Akt/mTOR signaling and mTORC1 signaling (Fig. 7G–I). The enrichment analysis of hub genes using hallmark gene set. (A–C) The GSEA enrichment of GJB2 using a hallmark gene set. (D–F) The GSEA enrichment of KCNH6 using a hallmark gene set. (G–I) The GSEA enrichment of KCNN4 using a hallmark gene set. GJB2 = gap junction protein beta 2, GSEA = Gene Set Enrichment Analysis, KCNN4 = potassium calcium-activated channel subfamily n member 4, KCNH6 = potassium voltage-gated channel subfamily h member 6. 3.8. The regulatory network of hub genes Finally, we explored the regulatory network of hub genes. From Figure 8A, it could be seen that GJB2, KCNH6, and KCNN4 could not interact directly and KCNH6 had the most miRNA targets and TF targets. Besides, the DrugBank database was used to predict the correlation between proteins and drugs and construct the interaction network. The drug–gene network displayed that KCNN4 could target most compounds compared with KCNH6 and GJB2 (Fig.8 B). The regulatory network of hub genes. (A) The miRNA–mRNA–TF interaction network of hub genes. (B) The drug–mRNA interaction network of hub genes. The red circle represents the hub genes; the gray circle represents the protein related to the hub genes; the yellow circle represents miRNA; the gray circle represents the transcription factor; the purple circle represents the drugs. TF = transcription factor. 4. Discussion Currently, molecular targeted therapy is considered an effective and promising treatment for cancers.[18] Besides, it has been reported that ion channels were involved in proliferation, metastasis and apoptosis in various cancers, which suggested the physiological intracellular effects were the potential targets for cancer treatments.[14,19] Therefore, the prognosis value of ion channels related genes in the diagnosis and prediction of cancers received increasing attention. In our study, we constructed and validated a risk model based on 3 ion channels related genes, which could effectively predict the prognosis of GBM. The mutation landscape showed that the mutation ratio of PETN and PIK3R1 was both increased in the high risk group, and the DEGs in high and low risk groups were involved in the calcium signaling pathway and PI3K–Akt signaling pathway. Ca2+ is involved in processes such as cell proliferation, migration, invasion, and angiogenesis and has been proven to be a therapeutic target for GBM.[20] Pei et al pointed out that increased intracellular Ca2+ activates nitric oxide synthase to generate nitric oxide, affecting the progression of GBM.[21] In our study, we found that the calcium signaling pathway was positively related to risk score. As for the PI3K–Akt signaling pathway, it was activated caused by the mutation and loss of PETN.[22,23] Additionally, PIK3R1 could encode p85α, one of the regulatory subunits of PI3K, and the mutation of PIK3R1 also caused the activation of the PI3K–Akt signaling pathway through disrupting the iSH2 domain.[24] These results indicated that the PI3K–Akt signaling pathway was induced by mutation of PIK3R1 and PTEN and played a vital role in the tumor deterioration of GBM. Besides, Karanovic et al demonstrated that the mutation of PIK3R1 led to immune dysregulation.[25] Our results identified that the infiltration levels of T cell, B cell, monocyte, and neutrophils were all significantly increased in the high risk group, and the level of monocyte was the highest. Although most immune cells could inhibit the development of tumor cells, some subtypes of immune cells performed pro-oncogenic functions in tumors. It has been proven that regulatory B cells and regulatory T cells (Tregs) could negatively regulate the immune response through inducing the production of pro-inflammation cytokines and suppressing the activation of immune cells.[26,27] In addition, the high level of monocytes and neutrophils could produce more reactive oxygen species and reactive nitrogen species to induce the conformational change of CD8 T cells, which resulted in the dysfunction of CD8 T cells.[28] On the other hand, monocytes and neutrophils could induce the differentiation of Tregs through secreting cytokines, which inhibited the antitumor responses.[29] According to our analysis, we found that the TIDE score of the high risk group was higher than that of the low risk group, which indicated that the immune escape potentiality was lower in the high risk group and the immunotherapy outcomes of patients with high risk scores were more effective. The correlation analysis revealed the TIDE score was most closely correlated to the infiltration level of monocytes, and monocytes were closely related to most immune escape-related genes. Thus, we inferred that monocytes may inhibit the immune escape by inducing differentiation of Tregs. GJB2, namely connexin 26, belongs to the gap junction protein family[30] and is involved in cell adhesion.[31] Many connexins can function as “hemichannels,” and abnormal activity of these proteins may lead to diseases. Previous studies claimed that GJB2 played an essential role in the proliferation and apoptosis.[30] Lu et al found that GJB2 was related to ion channels and was a prognostic biomarker for lung adenocarcinoma.[32] Besides, Yang et al demonstrated that GJB2 could lead to drug sensitivity through activating the PI3K–Akt signal pathway.[31,33] A recent study reported that GBJ2 knockdown influenced tunneling nanotube pathways, which control intercellular communication in GBM.[34] KCNN4 is a member of the Ca2+-activated potassium channel family.[35] KCNN4 gene encodes the Ca2+-regulated K channel K+Ca3.1 that participates in regulating membrane potential oscillations.[36] Franco et al demonstrated that aggressive migration in acidic pH of GBM stem cell line in vitro is independent of KCa3.1 ion channels.[37] However, it has been reported that KCNN4 was involved in the migration of various GBM cells in vitro experiments.[35] Besides, D’Alessandro et al identified that the treatment of TRAM-34, a high affinity blocker of KCNN4, inhibited the proliferation of GBM cells, and co-treated with TRAM-34 and temozolomide, a kind of chemotherapy for GBM patients, could obviously increase the survival time of GBM samples.[38] KCNH6, also known as ERG2, is a member of the ether-a-go-go-related gene channel genes belonging to the voltage-gated potassium (Kv) channels family.[39] Arcangeli et al identified EGR channel could mediate the integrin signal pathway which played an essential role in the migration and invasion of tumor cells.[40,41] Besides, SUZUKI and Takimoto found that the suppression of HERG channels led to a decrease in a number of cells in the G2/M phase.[42] In our study, we found that GJB2 and KCNN4 expressions were upregulated in the high risk group, while KCNH6 expression was downregulated. However, the survival analysis revealed that they were all oncogenic. Correlation analysis revealed that except for GJB2, both KCNH6 and KCNN4 were positively associated with the calcium signaling pathway. Moreover, we found that GJB2 and KCNH6 were positively and respectively related to the IL6–JAK–STAT3 and PI3K–Akt–mTOR signal pathway, and KCNN4 was negatively related to the INF-γ response. These results indicated that highly expressed GJB2 may promote the proliferation of tumor cells and induce differentiation of immune cells in GBM through activating IL6–JAK–STAT3.[43] Additionally, lowly expressed KCNH6 may promote the activation of INF-γ response, which may suppress the apoptosis of GBM cells and promote immune escape.[44,45] Furthermore, except for the PI3K–Akt–mTOR signal pathway, upregulated KCNN4 may also promote migration and invasion of GBM cells through inducing EMT.[46,47] The regulatory network exhibited that KCNH6 was targeted by more miRNA and TF, and KCNN4 interacted with more drugs. Furthermore, drug sensitivity analysis showed that the high expression groups of GJB2, KCNH6, and KCNN4 had higher IC50 values for certain drugs, such as Bryostatin 1_197, Midostaurin_153, Ponatinib_155, and Afatinib_1032. These results indicate that the high expression groups of GJB2, KCNH6, and KCNN4 have a stronger resistance to these drugs, which may partially explain why patients in these groups have poorer prognosis. In conclusion, we constructed a risk model based on 3 ion channels related genes, which could effectively predict the prognosis of GBM development. The further analysis found that KCNH6 and KCNN4 were respectively considered as the targets of molecular targeted treatment and chemotherapy. However, there are limitations in this study. The results of this study were primarily obtained based on the bioinformatics analysis, and the application of the 3 hub genes in the clinical treatment of GBM requires further experimental research. Additionally, this study has identified some potential signaling pathways, but the regulatory role of hub genes in these pathways in GBM also warrants further investigation. Author contributions Conceptualization: Chao Huang, Xue-Bin Yu. Data curation: Chao Huang, Wu-Qiao Bao. Formal analysis: Chao Huang, Xue-Bin Yu. Funding acquisition: Chao Huang. Investigation: Yong-Zhi Zhou. Methodology: Xue-Bin Yu, Yong-Zhi Zhou. Supervision: Wu-Qiao Bao. Writing – original draft: Chao Huang, Xue-Bin Yu, Yong-Zhi Zhou, Wu-Qiao Bao. Writing – review & editing: Wu-Qiao Bao. Supplementary Material