Dual role of exosomal circCMTM3 derived from GSCs in impeding degradation and promoting phosphorylation of STAT5A to facilitate vasculogenic mimicry formation in glioblastoma

Abstract

Background: Glioblastoma (GBM) is characterized by abundant neovascularization as an essential hallmark. Vasculogenic mimicry (VM) is a predominant pattern of GBM neovascularization. However, the biological functions of circRNAs prompting VM formation in GBM remains unclarified. Methods: The circular RNA circCMTM3 was identified through high-throughput sequencing and bioinformatics analysis. The expression of circCMTM3 in exosomes in glioma tissues and cells was verified via RT-qPCR and FISH. In vitro and in vivo assays, such as EdU, MTS, Transwell, and tube formation assays were performed to investigate functional roles of circCMTM3. Meanwhile, in situ tumorigenesis assay were implemented to explore the influences of circCMTM3 on the GBM progression. Additionally, RNA pull-down, RIP, ChIP, and dual-luciferase reporter gene assays were executed to confirm the underlying regulation mechanism of circCMTM3. Results: CircCMTM3, as a novel circular RNA, was packaged into exosomes derived from glioblastoma stem cells (GSCs), which facilitates the phenotypic transition of differentiated glioma cells (DGCs) to VM. Mechanistically, exosomal circCMTM3 is internalized by DGCs and disrupt the ubiquitination degradation of STAT5A and STAT5B by E3 ubiquitin ligase CNOT4. Additionally, through molecular scaffold function of circCMTM3, STAT5A is activated and triggers transcriptional regulation of target genes including the pro-vasculogenic factor CHI3L2 and the RNA-binding protein SRSF1. Subsequently, circCMTM3/STAT5A/SRSF1 positive feedback loop sustainably enhances VM formation and accelerates tumor progression in GBM. Conclusion: Exosomal circCMTM3 possessing growth factor-mimetic property activates the JAK2/STAT5A pathway via non-canonical manner, and promotes VM formation in GBM. The molecular communications between GSCs and DGCs offers a therapeutic strategy for targeting the neovascularization of GBM.

Keywords: CNOT4; STAT5A; circRNA; exosome; glioblastoma stem cells; vasculogenic mimicry.

Figure 1

Exosomal circCMTM3 is upregulated in GBM and correlated with poor prognosis. A Schematic diagram of the acquisition and analysis of exosomal circRNAs data from patient-derived GSCs and DGCs. B Heatmap displaying the z-scores value of circRNAs differentially expressed in exosomes derived from GSCs and DGCs respectively. C Volcano plot depicting the log₂ (fold change) of circRNAs in the two types of exosomes mentioned above. Grey dashed lines represent the cutoff value, which is P. adj value < 0.05 and | log₂(fold change) |> 1. Downregulated (purple) and upregulated (red) circRNAs in exosomes are color-coded. D Rank of differentially expressed exosomal circRNAs according to values of P. adj and log₂FC. E A schematic representation detailing the genomic characteristics of circCMTM3 (hsa_circ_0008450). The upper panel depicted the genomic location of the parental gene with its exons structure and the back-splicing site, as identified through Sanger sequencing, is displayed at the bottom of the panel. F Agarose gel electrophoresis of RT-qPCR assays showing the expression of circCMTM3 amplified from templates of GSC01 (upper panel) and GSC03 (lower panel) using divergent and convergent primers. G, H The RNA expression levels of circCMTM3 and CMTM3 in GSC01 (G) and GSC03 (H) after RNase R treatment. I Representative fluorescence images of in situ hybridization detection of circCMTM3 expressional location in GSC01 and GSC03. Scale bar = 50 μm. J, K The transmission electron microscopy analysis (J) and nanoparticle tracking analysis (K) displaying the morphologic characteristics and size distribution of GDEs. Scale bar = 100 nm. L Western blotting analysis of exosomal markers CD9, CD81 and TSG101 in GDEs. M The circCMTM3 expression level in glioma (n = 70) and normal brain (n = 10) tissues-derived exosomes. N The circCMTM3 expression difference in gliomas tissues-derived exosomes with different malignant grades. O Kaplan-Meier survival curve for all glioma patients with high and low exosomal circCMTM3 expression. Data are presented as means ± SD (three independent experiments). *p < 0.05; **p < 0.01; ***p < 0.001; ns, no significance.

Figure 2

Exosomal circCMTM3 can be internalized and expressed in DGCs to promote VM formation in vitro. A Representative fluorescence images of DGCs after incubating with PKH26-labeled GDEs. Scale bars = 50 μm. B RT-qPCR analysis of circCMTM3 expression in DGCs treated with different groups of GDEs. C Quantification of the EdU positive cells of DGCs in different GDEs treatment groups. D Representative images of EdU assays showing the proliferation of DGC01 after incubating with different groups of GDEs. Scale bars = 100 μm. E, F Representative images (E) and quantification (F) of the Transwell assay of DGCs treated with different groups of GDEs. Scale bars = 50 μm. G The migration ability of DGC01 is detected by HoloMonitor and visualized in Hstudio (n = 5) in different GDEs treatment groups. Scale bars = 50 μm. H Quantification of relative migration distance of DGCs by monitor visualization. I Western blotting analysis of VM markers MMP2, VE-Cardherin and Vimentin. J-L Representative images (J) and quantification (K, L) of the tube formation assay of DGCs treated with different groups of GDEs. Scale bars = 100 μm. Data are presented as means ± SD (three independent experiments). *p < 0.05; **p < 0.01; ***p < 0.001; ns, no significance.

Figure 3

Exosomal circCMTM3 targets and interacts with CNOT4 in DGCs. A Venn chart of exosomal circCMTM3 downstream target screening. B Silver staining image of RNA pull-down assay with circCMTM3 and control probes in DGCs. C LC-MS/MS spectrum showing the CNOT peptides pulled down by circCMTM3 probe. D 3D schematic diagram predicting the interaction between circCMTM3 and CNOT4 via HDOCK. E, F Representative images of immunofluorescence staining (E) and line chart of fluorescence signal positioning analysis (F) showing the colocalization of circCMTM3 (red) and CNOT4 (green) in DGCs. Scale bar = 100 μm. G-J RIP assays showing anti-CNOT treatment leaded to exosomal circCMTM3 enrichment in DGCs by incubating with different groups of GDEs. K, L Western blot analysis after RNA pull-down assay to investigate the interaction between exosomal circCMTM3 and CNOT4 in DGCs. M Heatmap of RNA-protein interaction binding strength between circCMTM3 and CNOT4 via the CatRAPID algorithm (top) and the diagrams of domain structure of CNOT4 and Flag-tagged CNOT4 truncations (bottom). N Left, western blot analysis showing the expression of full length or CNOT4 truncations from DGCs transfected with the indicated vectors; Right, western blot analysis revealing the enriched CNOT4 truncations pulled down by circCMTM3 probe. O RIP assays displaying enrichment levels of circCMTM3 by anti-Flag in DGCs transfected with the truncated mutant vectors. P RT-qPCR assays showing the expression of CNOT4 in DGCs treated with different groups of GDEs. Q, R western blot assays revealing the expression of CNOT4 in DGCs after incubating with different groups of GDEs. Data are presented as means ± SD (three independent experiments). *p < 0.05; **p < 0.01; ***p < 0.001; ns, no significance.

Figure 4

CNOT4 induces ubiquitination and degradation of STAT5A/B in DGCs. A Venn diagram showing integrated CatRAPID, mass spectrometry, TFs and UbiBrowser screening results. B, C Surface diagram of the docking model and interfacing residues between CNOT4 and STAT5A proteins (B) as well as CNOT4 and STAT5B (C) by GRAMM-X. D-G Western blotting analysis after Co-IP assays to assess the interaction of CNOT4 and STAT5A (D, E) as well as CNOT4 and STAT5B (F, G). H, I Western blotting analysis to confirm STAT5A and STAT5B expression after CNOT4 downregulation (H) or overexpression (I) in DGCs. J, K Western blotting assays showing STAT5A and STAT5B expression in CNOT4-silenced DGC01 (J) or overexpressed DGC01(K) treated with or without MG-132 (50 μM) for 6 h. L-O Western blotting assays showing STAT5A and STAT5B expression at different time nodes in CNOT4-silenced DGC01 (L, M) or overexpressed DGC01(N, O) by treating with CHX (50 μg/ml). P-S Quantitative analysis revealing half-life time(t_1/2) of STAT5A (P, Q) and STAT5B (R, S) expression by regulating CNOT4 in DGCs. T-W Ubiquitination assays showing the STAT5A and STAT5B ubiquitination levels followed by CNOT4 silencing (T, U) or overexpressed (V, W) in DGC01with MG132 treatment (50 μM) for 6 h. Data are presented as means ± SD (three independent experiments). *p < 0.05; **p < 0.01; ***p < 0.001; ns, no significance.

Figure 5

Exosomal circCMTM3 competitively binds to CNOT4 and prevents from STAT5A/B degradation in DGCs. A Schematic diagram of the regulatory mode reflecting circCMTM3 impacting on STAT5A/B expression. B Co-IP assay displaying the binding domain of CNOT4 responsible for its interaction with STAT5A/B in DGCs transfected with the truncated mutant vectors. C, D Western blotting assays showing STAT5A and STAT5B expression in DGCs with treatment of different groups of GDEs. E-H Co-IP assays illustrating interaction efficiency of CNOT4 and STAT5A (E, F) as well as CNOT4 and STAT5B (G, H) in DGCs under the condition of exosomal circCMTM3 overexpression. I-L Western blotting analysis showing STAT5A and STAT5B expression in CNOT4-overexpressed DGCs with treatment by GDEs containing upregulated circCMTM3 and CHX (50 μg/ml) (I, K), meanwhile, quantitative analysis on STAT5A and STAT5B expression half-life time (t_1/2) reflecting degradation rates (J, L). M, N Western blotting assays showing STAT5A (M) and STAT5B (N) expression in CNOT4-overexpressed DGC01 treated with or without MG-132 (50 μM) after incubating with circCMTM3-upregulated GDEs. O-R Ubiquitination assays showing the STAT5A and STAT5B ubiquitination levels in DGC01 with MG132 treatment (50 μM) followed by CNOT4 overexpression combined with circCMTM3-upregulated GDEs treatment (O, Q) or CNOT4 and exsomal circCMTM3 silenced simultaneously (P, R). S, T In vivo ubiquitination assays of polyubiquitin chains of STAT5A (S) and STAT5B (T) in DGCs transfected with mutant ubiquitin plasmids at the K6, K11, K48, and K63 sites. U, V Ubiquitination assays showing polyubiquitin chains assembly of STAT5A (U) and STAT5B (V) in DGCs transfected with wild-type or K6R, K48R and K6/48R-mutant ubiquitin plasmids. Data are presented as means ± SD (three independent experiments). *p < 0.05; **p < 0.01; ***p < 0.001; ns, no significance.

Figure 6

Exosomal circCMTM3 serves as a molecular scaffold activating JAK2/STAT5A pathway in DGCs. A, B Western blotting analysis showing protein levels of p-STAT5A (S726), total STAT5A, p-STAT5B (S731) and total STAT5B in DGCs after incubating with circCMTM3-upregulated or silencing GDEs. C Schematic diagram illustrating circCMTM3 as a molecular scaffold leading to rearrangement of the spatial positions of JAK2, STAT5A, and STAT5B. D-F RNA pull-down assay to confirm the interaction between exosomal circCMTM3 and JAK2 (D), STAT5A (E), and STAT5B (F) respectively in DGC01. G-L RIP assays showing anti-JAK2 (G, H), anti-STAT5A (I, J) and anti-STAT5B (K, L) treatment leaded to exosomal circCMTM3 enrichment in DGC01 by incubating with different groups of GDEs. M-R Co-IP assays displaying interaction efficiency of JAK2 and STAT5A as well as JAK2 and STAT5B in DGCs under the condition of exosomal circCMTM3 overexpression (M-O) and knockdown (P-R). S Illustration of functional fragments of circCMTM3 (upper panel) and corresponding deletion mutants (down panel) predicted and divided by the CatRAPID tool. T Western blot analysis revealing the affinity between different circCMTM3 fragment probe and STAT5A, STAT5B and JAK2 proteins via RNA pulled down assays. U Western blotting analysis illustrating protein levels of p-STAT5A (S726), total STAT5A, p-STAT5B (S731) and total STAT5B in DGCs treated by different mutant circCMTM3-riched GDEs. Data are presented as means ± SD (three independent experiments). *p < 0.05; **p < 0.01; ***p < 0.001; ns, no significance.

Figure 7

STAT5A transcriptionally upregulates the expression of the pro-vasculogenic factor CHI3L2. A GSEA based on bulk RNA-seq data of TCGA-glioma with STAT5A expressional difference. B Upset plot illustrating the number of genes obtained by differential analysis based on STAT5A expression and GSVA score of angiogenesis pathway in TCGA, CGGA325 and CGGA693 glioma samples. C Six candidate genes were screened from the overlap of the TCGA-STAT5A-UP, TCGA-ANGIO-UP, CGGA693-STAT5A-UP, CGGA693-ANGIO-UP, CGGA325-STAT5A-UP and CGGA325-ANGIO-UP data. The different polygons represent the correlation of candidates and STAT5A expression levels in TCGA, CGGA325 and CGGA693 respectively. D, E Western blotting assays showing CHI3L2 expression in DGC01 and DGC03 treated with or without Stafia-1 (20 μM) after incubating with circCMTM3-upregulated GDEs. F RT-qPCR assays displaying the mRNA expression of CHI3L2 in DGCs intervened with or without Stafia-1 (20 μM) after treatment with circCMTM3-overexpressed GDEs. G Schematic diagram showing the promoter region of CHI3L2 is divided into 11 parts for ChIP assays to validate the transcriptional activity of STAT5A. H Quantitative analysis of ChIP assay indicating the STAT5A binding regions in the CHI3L2 promoters under the condition of exosomal circCMTM3 overexpression with or without Stafia-1intervention. I, J ChIP-qPCR detecting H3K4me (I) and H3K27ac (J) enrichment in the CHI3L2 promoters in DGC01 with different group GDEs treatment and Stafia-1or DMSO intervention. K Schematic representation of the binding sites for STAT5A in the CHI3L2 promoters obtained from HOCOMOCO (left). Sequences of the wild-type (WT) and mutated (MT) binding sites in the promoter regions used in the luciferase reporter plasmids (right). L Dual luciferase reporter assay indicating that exosomal circCMTM3 increased the activity of the wild-type CHI3L2 promoter but had no effect on the activity of the mutated binding sites. Stafia-1 presenting transcriptional inhibition in DGC01. Data are presented as means ± SD (three independent experiments). *p < 0.05; **p < 0.01; ***p < 0.001; ns, no significance.

Figure 8

SRSF1 maintains the stability of exosomal circCMTM3 and is transcriptionally upregulated by STAT5A in DGCs. A SRSF1 motif sequence identified by RBP suite database. B Schematic representation illustrating RNA secondary stem-loop structures of circCMTM3 predicted by cRNAsp12. The characteristic binding site sequence interacted with SRSF1 is exhibited within the enlarged box on the left of the panel. C, D RT-qPCR assays displaying the RNA expression of exosomal circCMTM3 in DGCs intervened with GDEs in SRSF1-silencing (C) and overexpressed (D) DGCs. E-H RIP assays showing anti-SRSF1 treatment caused exosomal circCMTM3 enrichment in DGC01 (E, G) and DGC03 (F, H) by treatment with different groups of GDEs. I, J RNA pull-down assay to validate the interaction between exosomal circCMTM3 and SRSF1 in DGC01 (I) and DGC03 (J). K-N RNA dynamic assays showing the half-life of exosomal circCMTM3 in SRSF1- overexpressed (K, L) or silencing (M, N) DGCs followed by actinomycin D treatment. O Schematic illustration displaying CA-STAT5A domain structure of CA-STAT5A carrying mutations at two amino acids sites. P RT-qPCR assays showing the mRNA expression of SRSF1 in DGCs via CA-STAT5A overexpression and intervention by Stafia-1 or DMSO. Q, R Western blotting assays showing SRSF1 expression in DGC01(q) and DGC03 (r) treated with or without Stafia-1 after CA-STAT5A overexpression. S Schematic representation of the binding sites for STAT5A in the SRSF1 promoters and matched mutant sequences for Dual-luciferase reporter assays. T, U The Dual-luciferase reporter assays revealing the luciferase promoter activities of SRSF1 in DGC01 (T) and DGC03 (U) with or without CA-STAT5A overexpression and treatment by Stafia-1 or DMSO. V, The ChIP qPCR showing the enrichment difference of SRSF1 promoter sequence via anti-STAT5A treatment in DGCs with CA-STAT5A overexpression and treatment by Stafia-1 or not. Data are presented as means ± SD (three independent experiments). *p < 0.05; **p < 0.01; ***p < 0.001; ns, no significance.

Figure 9

Exosomal circCMTM3 promotes GBM malignant progression via facilitating VM formation in vivo. A Schematic illustration of treatments for nude mice intracranially implanted with patient-derived GSCs. B, C Representative H&E staining (B) and correspond quantification (C) displaying tumor size in brain slices from different experimental groups (n = 5), Scale bars=1mm. D Kaplan-Meier analysis of mice from the indicated groups (n = 5). E Double staining showing the VM formation assessed by PAS and anti-CD31 immunohistochemical staining in tumor tissue. Red arrows indicating the VM tubular structures with PAS⁺/CD31^-. Scale bars = 20 μm. F Assessment of VM scores in each group. G Representative H&E and immunohistochemical images demonstrating the morphological characteristics of GBM and expression of Ki-67, p-STAT5A, p-STAT5B, CHI3L2 and SRSF1 of tumor tissues from mice in different groups. Scale bar =50μm. Data are presented as means ± SD (three independent experiments). *p < 0.05; **p < 0.01; ***p < 0.001; ns, no significance.