The RNA-sequencing (RNA-seq) and matching clinical data of 600 patients (patients with incomplete data were deleted) with glioma in TCGA were obtained from the Genomic Data Commons (https://gdc.cancer.gov/, accessed on 3 September 2022), and 657 patients were used as a validation set that gained from the public CGGA database (http://www.cgga.org.cn/, accessed on 3 September 2022). The Tumor IMmune Estimation Resource (TIMER, http://TIMER.cistrome.org/, accessed on 8 September 2022) database was used to analyze the interrelation of CDC42 expression between different tumors and controls. The GTEx (https://www.gtexportal.org) database was used to supplement normal controls of the corresponding tissues. The Clinical Proteome Tumor Analysis Consortium (CPTAC, https://proteomics.cancer.gov) database was adopted to supplement and discover the CDC42 total protein expression level in tumors and controls. The c5.go.v2022.1.Hs.symbols and c2.cp.kegg.v2022.1.Hs.symbols.gmt datasets were obtained from the Molecular Signatures Database (MSigDB, https://www.gsea-msigdb.org/, accessed on 8 September 2022).

The TIMER 2.0 database is a comprehensive resource (Li et al., 2020). TCGA serves as a background database that includes gene expression data for 33 tumors, and CDC42 was submitted in the “Gene_DE” column with an affiliated exploration module of the TIMER 2.0 website. Then, CDC42 expression differences between tumors and adjacent tumor-free tissues for 33 cancer types from the TCGA database were analyzed. Considering the lack of normal tissue controls in the TCGA database, we supplemented the expression of CDC42 in TCGA tumors using the web-based tool Gene Expression Profiling Interactive Analysis (GEPIA, http://geppia.cancer-pku.cn/, accessed on 8 September 2022) (Tang et al., 2019) with matched TCGA normal group and GTEx data as controls. The following parameters were set as follows: log₂FC cutoff of 0.5 and a p-value of 0.05. UALCAN (http://ualcan.path.uab.edu, accessed on 8 September 2022) is an interactive portal (Chandrashekar et al., 2017) to analyze CDC42 protein expression in tumor and normal samples using data from the Clinical Proteome Tumor Analysis Consortium (CPTAC, http://ualcan.path.uab.edu/Analysis~prot.html).

The profiles of relevant clinicopathological and molecular biological features of CDC42 expression in gliomas were analyzed from TCGA and CGGA databases, including age, sex, pathology, isocitrate dehydrogenase (IDH) mutation status, 1p/19q codeletion status, O6-methylguanine-DNA methyltransferase (MGMT) promoter methylation status, and the World Health Organization (WHO) grade. The relationship between various clinicopathological features and CDC42 expression was shown by heatmaps and box plots drawn by “pheatmap” (version 1.0.12) and “ggplot2” (version 3.4.2) R packages. The significance of the dissimilarity was examined by the unpaired t-test and a one-way ANOVA.

The Kaplan–Meier (KM) curve was portrayed to estimate the prognostic value of CDC42 in glioma patients through TCGA and CGGA databases. With median survival as the cutoff value, patients were separated into two groups of low and high levels, and the results were visualized utilizing the “survminer” (version 0.4.9), “survival” (version 3.3-1), and “ggplot2” (version 3.4.2) R package. To evaluate the reliability of CDC42 expression in predicting survival, the receiver operating characteristic (ROC) curves were established using the “timeROC” (version 0.4) R package. Independent prognostically relevant factors were evaluated through univariate Cox and multivariate Cox proportional hazard regression models. The outcomes were displayed using the forest diagram produced by the R software.

The CDC42-combination proteins were obtained from the STRING online tool (https://string-db.org, accessed on 8 September 2022) (von Mering et al., 2003), and the functional annotations of each binding protein were downloaded. The CDC42 were entered in the “Proteins by name” query, and “Homo sapiens” were selected for organisms. Furthermore, the following main parameters were confirmed: network type (“full STRING network”), required score [“medium confidence (0.400)”], and size cutoff (“no more than 20 interactors”). The results were imported into the Cytoscape (Shannon et al., 2003) (version 3.7.2) software, and we applied the maximal clique centrality (MCC) algorithm to screen out the top four genes most closely related to CDC42. Based on the median CDC42 expression, samples were split into two groups in TCGA and CGGA databases, respectively. Differential expression genes (DEGs) were identified with the LIMMA (Ritchie et al., 2015) (version 3.52.4) R package after being analyzed (log₂FC > 1.5, p < 0.01). GO and KEGG analyses were carried out using the “clusterProfiler” (version 4.4.4), “org.Hs.e.g.db” (version 3.15.0), “ggplot2” (version3.4.1), and “enrichplot” (version 1.16.2) R packages. In biological process (BP), cell component (CC), molecular function (MF), and KEGG, barplot and dot plot charts illustrated the most significant pathways in upregulated DEGs.

Using the R packages “GSVA” (Hanzelmann et al., 2013) (version 1.44.5), “limma,” and “GSEABase” (version 1.58.0), ssGSEA was utilized to evaluate the RNA-seq data of 29 critical immune gene sets (Guo et al., 2022) from each glioma sample. Immune-related gene set scores between the two CDC42 expression groups were shown with the “pheatmap” package. The variances between both groups were demonstrated using the horizontal box plot manufactured by the “ggplot2” R package. We calculated the percentage of 22 types of tumor-infiltrating immune cells (TIICs) in the glioma microenvironment by applying the cell-type identification by estimating relative subsets of RNA transcripts (CIBERSORT) deconvolution algorithm (https://cibersortx.stanford.edu/) (Newman et al., 2015) to further explore the relationship between immune cell infiltration and CDC42 expression in gliomas. To illustrate the distribution of TIICs in the CDC42 high and low expression groups, boxplot and barplot were utilized by R software. The correlation between CDC42 and the relevant immune checkpoint was studied, and boxplot and scatter plots were employed by R software.

All statistical analyses were performed using R v4.2.1 and SPSS v.25.0 software (IBM Corp.). In the analysis of the correlation of clinicopathological features, the significance of the distinction between the two groups was tested by an unpaired t-test. One-way ANOVA was performed on multiple groups of samples to assess significant differences among three groups and above. The Kaplan–Meier survival analysis was performed to compare the overall survival of glioma patients with high and low CDC42 expression. The log-rank test was implemented to evaluate group differences. The Spearman correlation analysis was utilized to assess the correlation between groups of immune cells or immune-related pathways. All statistical tests were two-sided, and the cutoff for statistical significance was set at a p-value of <0.05.

TIMER2.0 was applied to study the mRNA expression levels of CDC42 for the entire 33 tumors in the TCGA database to explore the difference in CDC42 expression between tumors and adjacent tumor-free tissues. The outcome demonstrated that the expression rates are relatively high in breast invasive carcinoma (BRCA, p = 0.00073), cholangiocarcinoma (CHOL, p = 2.26e-09), esophageal carcinoma (ESCA, p = 6.60e-05), head and neck squamous cell carcinoma (HNSC, p = 1.63e-09), liver hepatocellular carcinoma (LIHC, p = 6.34e-11), and stomach adenocarcinoma (STAD, p = 0.00026) compared with the corresponding tumor-free tissues (Figure 2A). Interestingly, we did not find differences in expression between CDC42 normal tissue and tumors in some tumor types, such as GBM and LGG. We further supplemented the analysis using the GTEx database to consider the lack of normal sample controls in the TCGA database.

The GEPIA online tool was used to analyze tumors in the TCGA database and corresponding TCGA and GTEx normal tissues as controls. We found that CDC42 was highly expressed in large B-cell lymphoma (DLBC, p < 0.05), glioblastoma multiforme (GBM, p < 0.05), lower grade glioma (LGG, p < 0.05), ovarian cancer (OV, p < 0.05), pancreatic adenocarcinoma (PAAD, p < 0.05), testicular germ cell tumors (TGCTs, p < 0.05), and thymoma (THYM, p < 0.05) than normal tissues (Figure 2B).

The normal metabolism of tissues and various life activities is closely related to protein functions. The results of the CTPAC online tool implied that CDC42 protein was enriched in clear cell renal cell carcinoma (ccRCC, p = 1.7e-08), hepatocellular carcinoma (HCC, p = 1.0e-05), head and neck cancer (HNSC, 5.3e-18), and endometrioid cancer (UCEC, p = 1.5e-09) compared to normal tissues (Figure 2C). This further provided important evidence for the high expression of CDC42 in most tumor tissues. All evidence manifested that this molecule may be a promising marker for tumor-targeted therapy.

We then investigated the association between CDC42 expression levels and several clinicopathological features of glioma in a total of 1,257 glioma patients (Table 1) from the TCGA and CGGA databases, respectively. CDC42 showed distinct expression levels in glioma about their clinicopathological characteristics, such as IDH mutation status, MGMT promoter methylation status, 1p/19q codeletion status, pathological classification, and the WHO grade, which were analyzed in the TCGA database (Figure 3A) and validated in the CGGA database (Figure 3B). We found that CDC42 was highly enriched in high-grade glioma patients (Figures 3F, J), IDH wild-type glioma patients (Figures 3C, G), non-methylated glioma patients (Figures 3D, H), and non-codeletion glioma patients (Figures 3E, I) in TCGA and CGGA databases. The clinical prognoses of the patient and the degree of tumor malignancy are inextricably linked with the pathological classification of glioma (Ostrom et al., 2014). We discovered that dangerous pathological classifications of gliomas, such as glioblastoma, had high levels of CDC42 expression (Supplementary Figure 1).

According to an analysis of the KM curve based on the TCGA and CGGA databases, patients with high CDC42 expression (median survival 639 days) in the TCGA (Figure 4A) database had considerably shorter overall survival compared to patients with low CDC42 expression (median survival 3,519 days). This conclusion was verified in the CGGA database (Figure 4B). The ROC curves for 1, 2, and 3 years of survival were established, and the effectiveness of CDC42 in predicting glioma patient survival was confirmed by the area under curve (AUC) in TCGA (AUC at 1 year: 0.785, AUC at 2 years: 0.814, and AUC at 3 years: 0.812) (Figure 4C) and CGGA (AUC at 1 year: 0.617, AUC at 2 years: 0.653, and AUC at 3 years: 0.660) (Figure 4D) databases. CDC42 expression was a predictive factor in univariate and multivariate Cox regression studies, independent of other known prognostic markers, such as WHO grade, age at diagnosis, IDH mutation, 1p/19q codeletion, and MGMT promoter methylation (Tables 2, 3). In univariate Cox regression analysis, the forest plot demonstrated that high expression of CDC42 was linked to a shorter overall survival (HR = 4.518, 95% CI 3.417–5.881, p < 0.001) in the TCGA database, and a large number of other clinicopathologic variables, including WHO grade III (compared to grade II), grade IV (compared to grade III), age, IDH status, and 1p19q codeletion status, also showed significant relationships with overall survival (Figure 5A). Subsequently, multivariate Cox regression revealed that poor overall survival was associated with high expression of CDC42 (HR = 1.721, 95% CI 1.141–2.596, p < 0.01), WHO grade III (compared to grade II) (HR = 2.186, 95% CI 1.299–3.680, p < 0.01), WHO grade IV (compared to grade III) (HR = 2.714, 95% CI 1.343–5.488, p < 0.01), and older age (HR = 1.058, 95% CI 1.041–1.074, p < 0.01), while IDH mutation (HR = 0.353, 95% CI 0.202–0.618, p < 0.001) was connected with a beneficial outcome (Figure 5B). Similar results were displayed by the CGGA database (Figures 5C, D).

A total of 20 CDC42-bound proteins were screened through the STRING online database (Supplementary Figure 2A). We imported the results into the Cytoscape software and identified the top four proteins (PAK1, ARHGDIA, IQGAP1, and WAS) that were most closely associated with CDC42 using applying the MCC algorithm (Supplementary Figure 2B). Overactivation of CDC42 can lead to an exaggerated of p21-activated kinases (PAKs), which is crucial in triggering the growth of tumors (Radu et al., 2014). Zhang et al. (2023) discovered that the tripartite motif containing 56 (TRIM56) acts through the IQGAP1-CDC42 signaling axis to promote glioma cell migration and invasion. This evidence suggests that CDC42 plays a role in the development of glioma. To further investigate the biological functions and molecular mechanism connection between CDC42 and tumors, differential analysis of genes between the high and low groups of CDC42 expression was performed using the “limma” package. We screened (log₂FC = 1.5, p < 0.01) 1,011 and 226 DEGs associated with high expression of CDC42 in TCGA and CGGA databases for further GO and KEGG analyses. The top 20 upregulated and downregulated DEGs were shown with heatmaps (Supplementary Figures 3, 4). Barplots and dot plots were drawn to observe the enrichment results using the R software. GO analysis of TCGA and CGGA databases manifested that DEGs related to CDC42 high expression were intimately related to immune and inflammatory responses, chemotaxis, extracellular region, and chromatin (Figures 6A, B). KEGG enrichment analysis showed that CDC42 was associated with signaling pathways such as cytokine–cytokine receptor interaction, neutrophil extracellular trap formation, and transcriptional misregulation in cancer (Figures 6C, D). These results indicated that CDC42 may be crucial for immune and inflammatory responses and disease regulation in glioma.

Based on the 29 fully established immune-associated gene sets (Supplementary Table 1), ssGSEA was applied to examine the association between CDC42 expression and immune cell types as well as immune-related pathways or functions in the TCGA database. The heat map indicated that CD8⁺ T cells, Treg cells, T helper (Th) cells, and macrophages were significantly infiltrated in CDC42 hyper-expression glioma samples, yet natural killer (NK) cells showed the opposite result (Figure 7A). In addition, immune-related pathways and functions such as type I and type II interferon (IFN) response, major histocompatibility complex (MHC) class I molecules, T cell co-inhibition and co-stimulation, marker genes for immune checkpoints, para-inflammatory factors, antigen-presenting cell (APC), co-inhibition and co-stimulation, and chemokine receptor (CCR) activation (Figure 7B) were involved with CDC42 high expression.

To further investigate the association between immune cells in glioma and CDC42 expression, we computed the infiltration ratio of 22 types of TIICs in the 2 groups of CDC42 high and low expression using the CIBERSORT algorithm. M0, M1, and M2 macrophages (R = 0.143, p < 0.001, R = 0.108, p < 0.01 and R = 0.289, p < 0.0001, respectively), neutrophils (R = 0.258, p < 0.0001), T cell CD4⁺ memory (resting and activated, R = 0.139 and R = 0.201, respectively, p < 0.001), Treg cells (R = 0.23, p < 0.0001), and B cells' naïve cells (R = 0.193, p < 0.0001) were highly infiltrated in CDC42 high expression group (Figures 7C, D). T cell CD4⁺ naive (R = −0.254, p < 0.0001), B cells memory (R = −0.219, p < 0.0001), NK cells active (R = −0.217, p < 0.0001), eosinophils (R = −0.166, p < 0.0001), and monocytes (R = −0.109, p < 0.01) exhibited the opposite result (Figure 7E).

Immune cells express a class of immunosuppressive molecules known as immune checkpoints, which can control the level of immunological activation. Immune evasion of tumors can be accomplished by upregulating immune checkpoints (Di Giacomo et al., 2019). Immune checkpoint inhibitors (ICIs) have been applied in a range of cancer as a form of cancer immunotherapy based on this mechanism (Carlino et al., 2021). The effectiveness of anti-programmed cell death 1 (PD1) immunotherapy, a promising pathway (Wang et al., 2019) in the treatment of gliomas, is correlated with the expression of the programmed death ligand 1 (PD-L1) and the tumor immune microenvironment cell components (Yang et al., 2021; Yao et al., 2021). In TCGA and CGGA databases, we compared the expression levels of PD1, PD-L1, and 6 additional immune checkpoints in the CDC42 high and low expression groups. B and T lymphocyte attenuator (BTLA), CD200 receptor 1 (CD200R1), PD-L1, B7 homolog 3 (B7-H3), CD70, cytotoxic T lymphocyte-associated protein 4 (CTLA4), indoleamine-pyrrole 2,3-dioxygenase 1 (IDO1), and PD1 (Diesendruck and Benhar, 2017) expression were significantly upregulated (p < 0.0001) in the CDC42 high expression glioma patients (Figures 8A, B). Through correlation analysis, we discovered that the expression levels of CDC42 and PD-L1 in the TCGA (R = 0.48, p < 0.0001) and CGGA (R = 0.49, p < 0.0001) datasets exhibited a certain positive association (Figures 8C, D). Treg cell infiltration in glioma inhibits CD8⁺ T lymphocytes (CTLs) antitumor activity and may mediate resistance to ICIs (Amoozgar et al., 2021). We investigated the associations between CDC42 expression and several Treg cell markers, and we discovered that CDC42 expression was positively correlated with CD4 (R = 0.575, p < 0.0001), IL2RA (CD25, R = 0.551, p < 0.0001), and IL7R (CD127, R = 0.513, p < 0.0001) expression levels in the TCGA database (Figures 9A, C, E). The results were validated in the CGGA database (Figures 9B, D, F). All evidence suggests that CDC42 may be a target for glioma immunotherapy.