The GSC23 line was a gift from the Soochow University Stem Cell Research Group, and these cells were cultured in serum-free medium (Gibco, USA). Then, we added serum to the medium to induce GSC23 differentiation into glioma cells. A172, U251 and SVG cells were procured from Shanghai Institutes for Biological Sciences and grown in RPMI 1640 medium (Gibco, USA) supplemented with 10% fetal bovine serum (FBS) (Gibco, USA), 100 ng/mL streptomycin, and 100 U/mL penicillin (Gibco, USA). All cultures incubated at 37 °C and 5% CO2.
Exosome isolation and identification
Exosomes were extracted from cell culture supernatants by ultrahigh-speed centrifugation. Briefly, cell culture supernatants were differentially centrifuged (300g for 10 min, 1000g for 20 min, or 10,000g for 30 min). After filtration, the supernatant was ultracentrifuged at 100,000g for 180 min. Then, the precipitates were subjected to two phosphate-buffered saline (PBS)-washing steps, resuspended in PBS and stored at − 80 °C. Exosomes were visualized by scanning electron microscopy and confirmed by protein concentration determination, and the expression levels of CD9, CD63, Annexin, and Calnexin were assessed using Western blotting.
Scanning electron microscopic observation and measurement
Exosomal morphologies were visually assessed by scanning electron microscopy. The GSC23 and glioma cell supernatants were collected and extracted by ultrahigh-speed centrifugation, and a small amount of white sediment was observed at the bottom of the tube after centrifugation. Then, it was dissolved in 1 × PBS, fixed with 2.5% glutaraldehyde and dehydrated. Finally, we used a scanning electron microscope for observation. The samples were diluted with PBS, and 100 μL dilution was added to the sample plate. An Izon particle analyzer and a high-resolution tunable resistance pulse were used to detect the particle diameter.
Total RNA was extracted from GSC23/GSC23-differentiated cells with TRIzol (Life Technologies, ™, Carlsbad, CA, USA). Before constructing RNA-seq libraries, both the Ribo-Zero rRNA Removal Kit (Illumina, San Diego, CA, USA) and CircRNA Enrichment Kit (Cloud-seq, USA) were employed to attain rRNA removal and circRNA enrichment. The RNA-seq libraries were developed through pretreatment of RNAs with the TruSeq® Stranded Total RNA Library Prep Kit (Illumina,™, San Diego, CA, USA). All libraries were denatured into single-stranded DNA molecules, placed onto flow cells, amplified in situ as clusters and consequently sequenced (150 cycles) on an Illumina HiSeq™® 4000 Sequencer (Illumina,™, San Diego, CA, USA).
Microarray data and RNA sequencing data
Microarray datasets were retrieved from the GEO database. CircRNA sequencing was conducted in the Experimental Center of the Second Affiliated Hospital of Soochow University.
Expression profile assessments
CircRNA expression profiling for GSC23- and GSC23-differentiated cells (3 replicates each) was obtained by circRNA microarray technology. Cell circRNA expression difference analysis was performed using the limma package in R, and |logFC|> 1 and P < 0.05 were considered statistically significant.
According to gene ceRNA theory, highly expressed circRNAs may release miRNAs from their target mRNAs, thus promoting mRNA expression. Therefore, specifically overexpressed circRNAs in GSCs may act on miRNAs in glioma and result in negative regulation, thereby indirectly promoting abnormally high mRNA expression in glioma cells. To assess the interactions between circRNAs in GSCs and miRNAs in glioma cells, we first used the GEOquery package in R to determine the miRNA expression profile in the GSE25632 dataset in the GEO database (5 normal controls and 82 glioma tissues). The miRNA expression profile was annotated through the GPL8179 platform, and differential analysis was performed with the limma package in R. |logFC|> 1 and P < 0.05 were considered statistically significant.
To further explore the mRNA regulated by miRNA in glioma cells and specifically highly expressed circRNAs in GSCs, we used the GEOquery package in R to determine the mRNA expression profile from the GSE103227 dataset in the GEO database (including 5 normal controls and 5 glioma tissues). Differential analysis was performed with the limma package in R, with |logFC|> 1 and P < 0.05 being considered statistically significant.
CircRNA-miRNA-mRNA regulatory network development
We used the miRNA-circRNA module on the StarBase v3.0 website (http://starbase.sysu.edu.cn) to predict the targeted relationship between the miRNAs and the specifically highly expressed circRNAs in GSCs. Then, a Venn diagram was used to determine the intersection between miRNAs specifically targeting highly expressed circRNAs in GSCs and miRNAs that are specifically expressed at low levels in glioma cells. Thus, we identified target miRNAs and used Cytoscape 3.8.2 to draw the circRNA-miRNA regulatory network of specifically highly expressed circRNAs in GSCs that may participate in glioma gene regulation.
We used the miRTarBase website (http://mirtarbase.mbc.nctu.edu.tw/php/index.php) for target correlation prediction between miRNAs and mRNAs in a regulatory network of circRNAs that are highly expressed in GSCs and glioma miRNAs. Then, a Venn diagram was used to determine the intersection between mRNAs specifically targeting miRNAs with low expression in glioma cells and mRNAs that are specifically highly expressed in glioma cells; thus, we identified target mRNAs and used Cytoscape 3.8.2 to draw the miRNA-mRNA regulatory network of specifically highly expressed circRNAs in GSCs that may participate in glioma gene regulation.
The circRNA-miRNA-mRNA regulatory network was developed by combining circRNA-miRNA/miRNA-mRNA pairs. Ultimately, network visualization was obtained through Cytoscape 3.8.2.
Functional enrichment analysis
The R-based ‘Cluster Profiler package’ was employed for GO/KEGG pathway enrichment analysis of mRNA in the miRNA-mRNA regulatory network, followed by visualization of the results through Cytoscape 3.8.2. In addition, all such GO/KEGG outcomes were collected using R studio/R scripting language, and placed a criterion that GO analysis-P value would be < 0.05, and the P and Q values of KEGG analysis would be < 0.05.
Construction of the PPI network, module and immune-correlation analysis
We used CGGA data and Kaplan–Meier survival analysis to delve into deeper assessment of the interactions of mRNA expression levels and the prognosis of glioma patients. Consequently, glioma poor prognosis-related mRNAs were submitted to the STRING version 11.0 (https://string-db.org/) database to develop the PPI network, species were restricted to Homo sapiens with medium confidence > 0.4, and isolated targets were removed. Then, the PPI network map was imported into Cytoscape 3.8.2 software for visualization, and the top 10 hub genes were analyzed by means of the cytoHubba plug-in. We further combined the glioma data in the GTEx and TCGA databases to compare expression level profiles for these 10 hub genes (glioma and healthy tissue). The Molecular Complex Detection (MCODE) app was employed for screening hub gene modules within this PPI network. Then, we used the GSVA R package to analyze the relationship between these 10 mRNAs and the degree of immune cell infiltration.
Quantitative reverse transcription polymerase chain reaction (qRT-PCR)
Total RNA was isolated with TRIzol® (Invitrogen™, USA). The quality of total RNA was assessed using a NanoDrop® 1000 Spectrophotometer (Thermo Fisher Scientific™, USA). First-strand cDNA was prepared using the QuantiTect® Reverse Transcription Kit (QIAGEN™, USA). Actual RT-qPCR runs were conducted through qPCR SYBR Green Mix® (Bio-Rad™, USA) and the ABI 7500® platform (Applied Biosystems™, USA). PCRs were conducted using three technical replicates/sample, with resulting data assessed through the 2−ΔΔCt methodology . GAPDH served as a normalizing/reference gene for circRNA, mRNA and circRNA, while U6 served as a reference for circRNA. Primers consisted of GAPDH: 5′-AGAAGGCTGGGGCT CATTTG-3′ (forward) and 5′-AGGGGCCATCCACAG TCTTC-3′ (reverse); U6: 5′-CTCGCTTCGGCAGCACA-3′ (forward) and 5′-AACGCTTCACGAATTTGCG T-3′ (reverse).
Cell treatment and transfection
SiRNAs targeting circRNA-Serpine2, miR-124-3p/KIF20A mimics and inhibitors were procured through GenePharma (JiangSu, China). Such oligonucleotides/vectors underwent transient transfection using Lipofectamine 3000® (Invitrogen™, USA) following the manufacturer's protocols. Regarding circRNA-Serpine2 knockdown, GSC23 cellular aliquots were transfected with 20 nM siRNA against circRNA-Serpine2. MiR-124-3p/KIF20A overexpression or knockdown was achieved with 20 nM miR-124-3p/KIF20A mimic or inhibitor in A172 cells. GSC23, A172 and U251 cells underwent transduction for 24 h. Consequently, qPCR was conducted to validate altered circRNA/miRNA/mRNA expression within stabilized cells.
Cellular lysis was conducted using RIPA® buffer (CWBio™, Beijing, China). Lysates were consequently placed into loading buffer and denatured (100 °C for 10 min). The resulting solutions were exposed to sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE), followed by transfer onto polyvinylidene difluoride (PVDF) membranes (Millipore™, Billerica, Massachusetts, USA). Following a 60-min blocking step, the membranes were placed into an incubator overnight at 4 °C in tandem with selected primary antibodies (Abcam™, Cambridge, MA, USA). Consequently, the relevant secondary antibody (CST™, Danvers, MA, USA) was introduced, and the membrane was kept at room temperature for 120 min. Proteomic loads were assessed using Immobilon Western Chemiluminescent HRP Substrate® (Millipore™, USA). The GAPDH signal was used as a loading control.
Cell proliferation assay
For clone formation assays, transfected cells were seeded within 6-well plates and grown in medium with 10% FBS at 37 °C/5% CO2. Following a 10-day incubation period, cells were stained using 0.1% crystal violet (Beyotime™, Beijing, China), followed by manual colony counting.
Cell invasion assay
Cell invasion assays were carried out using 24-well Transwells (8 μm, Corning, USA) coated with Matrigel (BD, USA). A172 cells were grown within circ-siRNAs, miRNA inhibitors and mRNA mimics according to the manufacturer’s instructions. Overall, 1 × 105 cells in 500 μL DMEM (1% FBS) were added to the upper chamber, and 750 μL DMEM (10% FBS) was added to the lower chamber. After incubation for 48 h, Matrigel and cells in the upper chamber were removed. Cells on the lower surface of the membrane were fixed in 4% paraformaldehyde and stained with 0.5% crystal violet. The invasive cells were imaged using an inverted microscope (Nikon, Japan) and quantified in five random fields per well. Each trial had three independent experiments.
Cell migration assay
A172 cells were grown within 6‐well plates together with circ-siRNAs, miRNA inhibitors or mRNA mimics according to the manufacturer’s instructions. Seventy-two hours later, serum-free DMEM was introduced to the plates for 12 h. All cells were seeded in marked plates and subjected to three consecutive wash steps using sterile PBS, and monolayers were scratched (plate central region) using 200-μL pipettor tips. After incubation under normal conditions for 24 h, the cells were evaluated microscopically. All assays were performed on three separate occasions.
A172 cells were harvested, washed in PBS and incubated with an Annexin V-FITC Apoptosis Detection Kit® (Beyotime Biotech™, Haimen, China) 48 h after transfection. Annexin V-FITC acted as a cell-staining agent, followed by resuspension into binding buffer (190 μL) prior to the introduction of 10 μL of PI (20 μg/mL). Consequently, the cells were incubated (15 min/dark conditions/RT) and then assessed through flow cytometry and FACSDiva® software (Version 6.2). Cell types were separated/grouped into viable, necrotized, and apoptotic cells, and the percentages of apoptotic cells within individual groups were determined.
Ten BALB/c nude murines (with 200 μL of cell suspension per mouse, 21 days old) were segregated in a random manner into two groups. In the control group, A172 cells were subcutaneously introduced within the right anterior flank of each mouse continuously for a five-day period. Concomitantly, circ-Serpine2 knockout exosomes were collected and injected across subcutaneous tumor circumferences at 0, 6, 12, 18, and 24 days post A172 injection. In the experimental group, mice were subcutaneously treated with A172 cells and GSC23 exosomes as described above. Day 6, 12, 18, 24 and 30 postsubcutaneous injection of A172 cells served as assessment time points, where tumor size was measured through calipers, and tumor volume (mm3) was assessed accordingly: length × width2/2.
Thirty days posttreatment, all mice were euthanized (pentobarbital sodium by intraperitoneal injection, 150–200 mg/kg). Murine tumors were removed and weighed, with sections collected, hematoxylin/eosin stained and placed for observation under microscopy measures.
Statistical analysis was conducted using SPSS 21.0® (Chicago, IL, USA). Major circRNA dysregulations were assessed through R. Limma packages/FDR filtering were employed for comparative analyses. A P-value < 0.05 and absolute fold change ≥ 2 were deemed to confer statistical significance. The chi-square test was used to assess the interactions of mRNA expression level and clinical characteristics.