3.1 The mRNA expression and protein expression of CAV1

To check the CAV1 expression in different cancers, RNA sequencing data in TCGA was mined. The differential CAV1 mRNA expression between tumor and adjacent tissues were shown in Fig. 1A. CAV1 mRNA expression was significantly higher in low-grade glioma (LGG) (Fig. 1B), the combination of LGG and glioblastoma multiforme (GBMLGG) (Fig. 1C) as well as glioblastoma multiforme (GBM) (Fig. 1D), compared with adjacent tissues. CAV1 protein expression was verified by immunohistochemistry (IHC) in adjacent tissues and tumor tissues (Fig. 1E with antibody HPA049326 and Fig. 1F with antibody CAB003791). To check the relationship between CAV1 expression and the pathological progression, we checked the CAV1 expression in different stages of LGG and GBMLGG, As shown in Fig. 1G (LGG) and H (GBMLGG), the expression of CAV1 is significantly higher in the higher grade of glioma. Since GBM refers to the Grade 4 glioma, the expression of CAV1 in GBMLGG data equals the GBM data. We checked the GBM data in UALCAN database too, as shown in Fig. 1I, the protein expression of CAV1 is much higher in the GBM than that in the normal group. The expression of CAV1 is higher in glioma malignancy than other kinds of cells in the tumor microenvironment, single cell sequencing of GSE131928 (Fig. J–M,) indicates that CAV1 is highly expressed in all different malignancy cells including AC-like malignant, OPC-like malignant, MES-like malignant and NPC-like malignant cells. Another single-cell sequencing data of GSE 148842 (Fig. N–Q) indicates CAV1 is highly expressed in the malignant cells and malignant cells occupy the majority of cell lineages in all the clusters. The expression profile of CAV1 in CGGA data shows similar result as TCGA (Supplementary Fig. 1).

Fig. 1

The mRNA expression and protein expression of CAV1. A mRNA expression of CAV1 in pan-cancers. B The differentiated mRNA expression of CAV1 in TCGA-LGG, total 1675 samples (1152 normal + 523 tumor samples). C The differentiated mRNA expression of CAV1 in TCGA-GBMLGG, total 1846 samples (1152 normal + 689 tumor samples + 5 tumor adjacent samples). D The differentiated mRNA expression of CAV1 in TCGA-LGG, total 1323 samples (1152 normal + 166 tumor samples + 5 tumor adjacent samples). E The protein expression of CAV1 in normal brain tissues (left) and glioma (right) with antibody HPA049326. F The protein expression of CAV1 in normal brain tissues (left) and glioma (right) with antibody CAB003791. G. The expression of CAV1 in different stages in LGG. H The expression of CAV1 in different stages of GBMLGG. I The protein expression of CAV1 in GBM and normal, data from UALCAN (can.path.uab.edu/). J–M. Single cell sequencing of GBM sample GSE 131928. N–Q Single cell sequencing of GBM sample GSE148842. *p value < 0.05; **p value < 0.01; ***p value < 0.001. OS: overall survival, DFS: disease-free survival

3.2 The clinical correlation between CAV1 expression and glioma patients’ survival

We explored the correlation between the high and low expression groups of CAV1 subsequently. As shown in Fig. 2A (LGG), Fig. 2B (GBMLGG), the overall survival (OS) of the CAV1-high expressed group is much shorter than that of the CAV1-low expressed group. In the GBM group in Fig. 2C, the survival difference is not as obvious as the previous that in the previous two stages.

Fig. 2

The clinical correlation of CAV1 expression in glioma. A–C. OS between and high-low expression groups of CAV1 gene in KM databases of LGG (A), GBMLGG (B) and GBM (C) patients. D–F ROC curve established the efficiency of CAV1 mRNA expression level on distinguishing LGG tumor (D), GBMLGG (E), and GBM (F) from non-tumor tissue. X-axis represents false positive rate, and Y-axis represents true positive rate. G–I The ROC curve using CAV1 as an indicator of LGG (G), GBMLGG (H) and GBM (I) were explored. J–L The calibration curve using CAV1 as an indicator of LGG (J), GBMLGG (K) and GBM (L) was checked. M, N The univariate (M) and multivariate regression (N) analysis of CAV1 and other clinicopathologic parameters with OS in LGG patients were explored. *p value < 0.05; **p value < 0.01;***p value < 0.001. OS: overall survival; DFS: disease-free survival

The risk prediction model is characterized by the result of the discrimination and calibration. The receiver-operating characteristic (ROC) curve is applied for assessing the model discrimination and it is the most popular graphical method for assessing the classification accuracy of a diagnostic biomarker. It was used to analyze the effectiveness of CAV1 mRNA expression level AUC on distinguishing glioma tissues from normal issues. The AUC of CAV1 was 0.822 for LGG (Fig. 2D), 0.855 for GBMLGG (Fig. 2E), and 0.963 for GBM (Fig. 2F), suggesting that CAV1 could serve as a biomarker to distinguish glioma from non-tumor tissue. In time-to-event studies, the subject's event result is time-dependent, therefore, a new time-dependent extension of ROC curve is an important estimator. As shown in Fig. 2G–I, CAV1 as an indicator in LGG (2G), GBMLGG(2H) and GBM (2I) was explored. Calibration is a critical component for the reliability, accuracy, and precision of prediction models, we checked the calibration of CAV1 in the 1-year survival, 3-year survival and 5-year survival of LGG (Fig. 2J) and GBMLGG (Fig. 2K), for GBM (Fig. 2L), only 1-year and 3-year survival data is available, probably due to the short survival for GBM patients. The univariate (Fig. 2M) and multivariate regression (Fig. 2N) analysis of CAV1 and other clinicopathologic parameters with OS in LGG patients were explored. All the above were checked in CGGA database and similar results were shown in Supplementary Fig. 2.

3.3 The analysis of differentiated genes in CAV1-high and -low groups in glioma patients

To explore the function of CAV1 in glioma, we divided the glioma patients into CAV1-high and CAV1-low groups and mined out the differentially expressed genes (DEGs) (Fig. 3A, B, Supplementary Table 1). The Gene Ontology (GO) and (Kyoto Encyclopedia of Genes and Genomes) KEGG pathway analysis (Fig. 3C) indicates that the CAV1-high groups are enriched in immunoglobulin complex, the complement activation, etc., and the CAV1-low groups are enriched in neuroactive ligand-receptor interaction (Fig. 3D). The Gene Set Enrichment Analysis (GSEA) indicates that the high expressed genes are enriched in Reactome Signaling by Interleukins and the low expressed genes are enriched in Neutrophil Degranulation (Fig. 3E, F, Supplementary Table 2). Then we conducted the combinational analysis of GO/KEGG and the LogFC. The GO enrichment and the highly expressed genes are shown in Fig. 3G and 3H (GO:0006959 humoral immune response, GO:0006958: complement activation, GO: 0019814 immunoglobin complex, GO:0003823 antigen binding).

Fig. 3

The analysis of differentiated genes in CAV1-High and -low groups in glioma patients. A The volcano plot shows the DEGs in CAV1-high and -low groups. B The Rank of differentially expressed genes. C The GO and KEGG analysis of high expressed genes in CAV1-high groups. D The GO and KEGG analysis of low expressed genes in CAV1-low groups. E The GSEA analysis or the DEGs. F–H The GSEA analysis in ridge plot of DEGs. G The combinational analysis of GO/KEGG and LogFC of CAV1-high groups

3.4 CAV1 expression has a positive correlation with the innate immune cell infiltration

Since the analysis in Fig. 3 indicated that the DEGs in CAV1-high groups are enriched in the immune response and immune infiltration, we then analyzed the expression of CAV1 and the immune infiltration in different gliomas. The stroma score and the immune score were calculated by applying Estimation of Stromal and Immune cells in Malignant Tumors using the Expression data (ESTIMATE) algorithm based on gene expression MATRIX of glioma patients in the TCGA database. Stroma scores indicate the stroma quantity in the extracellular matrix, the immune scores indicate the infiltration of immune cells in the tumor and the final ESTIMATE scores are used to deduce the tumor purity. As shown in Fig. 4A–C, the expression of CAV1 is significantly positively associated with the stroma score, immune score and ESTIMATE score (r > 0.5), and the GBMLGG type shows the highest correlation. The immune cells infiltration shows that the correlation between the expression of CAV1 and the infiltration of dendritic cells (DCs), macrophages and neutrophils are significantly based on the database from TCGA and GTEX (Fig. 4D), we then checked only TCGA via the ssGSEA analysis and found macrophages, neutrophils and eosinophils are the top infiltration immune cells and has the positive correlation with the expression of CAV1 (Fig. 4E). The expression and CAV1 and the enrichment of macrophages, neutrophils and eosinophils in GBMLGG were checked, as shown in Fig. 4F, CAV1 positively correlates with the enrichment of above immune cells. Therefore, the innate cell enrichment and infiltration in gliomas are much higher than the adaptive cells. Immune checkpoint targeting is a promising cancer immunotherapy and has been studied a lot in gliomas, the relationship between the expression of CAV1 and the immune checkpoints expression was explored. CAV1 expression is positively related with VTCN1 and TNFR18, TNFR14 in LGG patients’ samples (Fig. 4G). In a word, the immune infiltration analysis indicates that the expression of CAV1 correlates with immune infiltration positively and significantly.

Fig. 4

CAV1 expression has a positive correlation with the innate immune cell infiltration. A The correlation between the expression of CAV1 and stroma score. B The correlation between the expression of CAV1 and immune score. C The correlation between the expression of CAV1 and ESTIMATE score. D The correlation of the expression of CAV1 and immune cell infiltration, data from TCGA + GTEX. E The correlation of the expression of CAV1 and immune cells infiltration, data from GSEA via ssGESA analysis. F. The correlation between the expression of CAV1 and the enrichment of macrophage, neutrophil and eosinophil in GBMLGG. G The correlation between CAV1 expression and the immune checkpoints

3.5 CAV1 expression positively relates to glioma cancer invasion and stemness

Since CAV1 has been demonstrated to promote pancreatic cancer invasion and metastasis [27], we would like to check its role in glioma cancer invasion and metastasis. Vimentin (VIM) is one of the human intermediate filament proteins and is required for the plasticity of mesenchymal cells under normal physiological conditions and for the migration of cancer cells that have undergone epithelial-mesenchymal transition [28]. We checked the correlation between CAV1 and VIM expression and found CAV1 expression positively significantly correlates with the expression of VIM (Fig. 5A). Besides VIM, the basement membrane also promotes cancer cell invasion and metastasis, COL4A1 and COL4A2 are typical basement membranes ubiquitously expressed on cells [29]. The expression of CAV1 positively correlates with COL4A1 (Fig. 5B) and COL4A2 (Fig. 5C). More metastasis biomarkers such as TGM2 (Fig. 5D), GBP1 (Fig. 5E) and IGFBP7 (Fig. 5F) were checked too, and all of their expressions have a positive correlation with the expression in CAV1 in all different types of gliomas. Moreover, we also checked the protein–protein interaction (PPI) network of CAV1 and found it is highly connected with Caveolin-2 and EGFR (Fig. 5G), which are important factors for tumor progression. Therefore, we propose that the high expression of CAV1 may enhance glioma cancer cell invasion and progression.

Fig. 5

CAV1 expression is positively relates to glioma cancer cell migration. A The relationship between the expression of CAV1 and vimentin in LGG, GBMLGG and GBM. B The relationship between the expression of CAV1 and COL4A1 in LGG, GBMLGG and GBM. C The relationship between the expression of CAV1 and COL4A2 in LGG, GBMLGG and GBM. D The relationship between the expression of CAV1 and TGM2 in LGG, GBMLGG and GBM. E The relationship between the expression of CAV1 and GBP1 in LGG, GBMLGG and GBM. F The relationship between the expression of CAV1 and IGFBP7 in LGG, GBMLGG and GBM. G The PPI network of CAV1 was analyzed by string analysis. H The relationship between the expression of CAV1 and cancer stemness of pan-cancers

Cancer stem cells (CSCs) are self-renewing cells that facilitate tumor initiation, promote metastasis, and enhance cancer therapy resistance [30]. Glioma cancer stemness is highly related to its progression, drug resistance and recurrence. Therefore, we explored the relationship between CAV1 and glioma stemness from RNA sequencing data in TCGA and GTEX, and found that the expression of CAV1 is positively related to the glioma stemness in pan-cancer analysis (Fig. 5H).

3.6 The methylation status of CAV1 and its relationship with patient’s survival time

DNA methylation regulates gene expression and cancer progression [31]. Then we checked the methylation status of CAV1 in cancer and normal samples with the consideration of several factors including gender, pathological stages, copy numbers, etc. (Supplementary Fig. 3A). The cluster analysis of the methylation of CAV1 in four different transcripts was analyzed, As shown in Supplementary Fig. 3B (LGG) and Fig. 3C (GBM), the methylation level in GBM is higher than that in the LGG, there is no significant differences among the four different transcripts. We also checked the methylation status of CAV1 promoters and the correlation between the methylation level of CAV1 and the glioma patient’s survival time. As shown in Fig. 6A, the DNA methylation of CAV1 is very low and the expression of CAV1 is quite high. The same results were found in CGGA data too (Supplementary Fig. 4). The methylation level is negatively associated with the glioma pathological stages (Supplementary Fig. 4). Patients with lower methylation of CAV1 shows shorter survival (Fig. 6B, C and Supplementary Fig. 5).

Fig. 6

The methylation status of CAV1 and its relations with patients’ survival time. A The dot plot shows the expression of CAV1 and its methylation status. B The correlation between the methylation of CAV1 (at the location of Shore, cg-07838272) and patients’ survival, methylation probes cg-07838272. C The correlation between the methylation of CAV1 (at the location of TSS 200-Island, cg-07964538) and patients’ survival. D WGBS data of nine cell lines including two glioma cell lines (GB-1 and SF126) shows the methylation level in their DNA promoters

According to our Whole Genome Bisulfite Sequencing (WGBS) of four cancer cells that are resistant to oxidative phosphorylation (786-O, CFPAC-1, GB-1, and SF126) and five types of cancer cells that are sensitive to OXPHOS inhibition (NCI-H82, G-401, MDA-MB-453, WSU-DLCL2, and SW48), the methylation of CAV1 in two glioma cancer cell lines (GB-1 and SF126) are very low compared to that in the resistant cell line group (Fig. 6D).

3.7 High expression of CAV1 renders the glioma cells OXPHOS inhibition

Previous studies indicated that CAV1 is involved in the modulation of cancer metabolism and glycolytic activities [32]. In our RNA-sequencing data, we found CAV1 is much higher expressed in the OXPHOS-inhibition resistant groups than that in the sensitive groups (Fig. 7A). We used OXPHOS inhibitor Gboxin to treat all 57 cancer cells including glioma cancer cell GB-1 and SF-126 and found that both GB-1 and SF-126 are resistant to Gboxin, their IC50s are among the highest in these 57 cancer cells (Fig. 7B, Supplementary Table 3). Further analysis has shown that the expression of CAV1 is positively correlated with the IC50 of Gboxin for GB-1 and SF-126 (Fig. 7C). We verified the above result by Cell Viability Assay of SF-126, the cell viability doesn’t change significantly after Gboxin treatment in different dosage (Fig. 7D), indicating that SF-126 is resistant to Gboxin. We also explored CAV1 expression in OXPHOS-inhibition resistant and sensitive cells from CCLE and found that CAV1 is highly expressed in the OXPHOS-inhibition resistant groups than that in the sensitive groups (Fig. 7E, F). In the Depmap PRISM database, cell viability of more different glioma cells treated by another classical OXPHOS inhibitor Antimycin A is available, the CERES score of more glioma cells treated by Antimycin A is above zero (Fig. 7G, Supplementary Table 4), indicating that there are more glioma cancer types are resistant to the OXPHOS inhibition besides the cells lines GB-1 and SF-126 we checked.

Fig. 7

High expression of CAV1 renders the glioma cells OXPHOS inhibition. A Relative RNA expression of CAV1 in 9 cancer cells that we did RNA-sequencing, among which GB-1 and SF-126 are the human glioma cells. B The IC50 of all 57 cancer cell lines after treatment of Gboxin (the OXPHOS inhibitor) for 72 h. Glioma cancer cells SF-126 and GB-1 were marked. C The correlation between the expression of CAV1 and the OXPHOS-resistant cancer cell’s IC50. D Cell Viability test for SF-126 human glioma cancer cell line under Gboxin treatment in different dosages in 72 h. E The relative expression of CAV1 of OXPHOS-inhibition resistant cancer cell line, data from CCLE. F The relative expression of CAV1 of OXPHOS-inhibition sensitive cancer cell line, data from CCLE. G DepMap PRISM data shows more glioma cells under treatment with OXPHOS inhibitor Oligomycin A. The CERES scores were marked. H The GO/KEGG analysis of high-expressed genes in OXPHOS-inhibition groups. I The correlation between the expression of CAV1 in LGG gliomas with key OXPHOS genes including mt-CO1, mt-CO2 and mt-CO3. J The correlation between the expression of CAV1 in GBMLGG gliomas with key OXPHOS genes including mt-CO1, mt-CO2 and mt-CO3. K The correlation between the expression of CAV1 in GBM gliomas with key OXPHOS genes including mt-CO1, mt-CO2 and mt-CO3. L The correlation between the expression of CAV1 and HK2 in gliomas. M The correlation between expression of CAV1 and GLUT3 (SLC2A3) in gliomas. *p value < 0.05; **p value < 0.01;***p value < 0.001

To check whether CAV1 is the essential gene for all glioma cancer cell’s viability, we mined the DepMap data for the whole-genome CRISPR knocking out CAV1 in 67 glioma cancer cell lines and found the CERES scores of 45 cell lines are above zero and 22 cell lines are below zero (Supplementary Table 5), indicating that CAV1 are essential for some of the glioma cells not for all and there is high heterogeneity of gliomas.

When we conduct the GO/KEGG analysis for the highly expressed genes in the OXPHOS-resistant groups, we found the most significant pathways are the “collagen-containing extracellular matrix” in the biological process (BP) and “positive regulation of cell adhesion” in the molecular function (MF) (Fig. 7H). Considering promoting cell migration is one of the important functions of CAV1, and the RNA-sequencing data of resistant cells including two glioma cancer cell lines, we propose that the high expression of CAV1 may render the glioma cells OXPHOS inhibition and promotes cancer cell progression. To further verify the hypothesis, we checked the expression of CAV1 and the key OXPHOS pathway genes including MT-CO1, MT-CO2 and MT-CO3, and found that the expression of CAV1 is negatively related to the expression of MT-CO1, MT-CO2 and MT-CO3 in LGG and GBMLGG (Fig. 7I–K), but the correlations between CAV1 and OXPHOS genes in GBM is not as strong as that in LGG and GBMLGG. Meanwhile, we checked the expression of CAV1 with the key glycolysis pathway genes HK2 and GLUT3(SLC2A3) in gliomas and found that the expression of CAV1 is positively related to HK2 (Fig. 7L) and GLUT3 (Fig. 7M).