A total of 74 GBM tissue samples were retrospectively collected and analyzed from patients who underwent surgical resection at the Affiliated Hospital of North Sichuan Medical College (Nanchong, Sichuan, China) between September 2022 and March 2024. The cohort included 45 male and 29 female patients, with an age range of 40-80 years. Diagnosis was established by senior associate chief physicians and chief physicians in the pathology department based on typical histopathological features of GBM according to the World Health Organization (WHO) Classification of Tumors of the CNS, 5th edition (2021) (25). As part of routine pathological evaluation, immunohistochemical analyses, including glial fibrillary acidic protein, oligodendrocyte transcription factor 2, Ki-67 proliferation marker, tumor protein p53 (p53), α-thalassemia/mental retardation syndrome X-linked and isocitrate dehydrogenase 1 (IDH1; R132H), were performed. Notably, all 74 cases were confirmed as IDH-wildtype based on IDH1 immunohistochemistry and exhibited characteristic histopathological features of GBM, such as microvascular proliferation and/or necrosis. No IDH-mutant astrocytomas were included in the cohort. In addition, molecular pathological testing was performed in 32 patients as part of routine clinical evaluation. All four molecular assays, including O6-methylguanine-DNA methyltransferase promoter methylation, telomerase reverse transcriptase promoter mutation, epidermal growth factor receptor (EGFR) amplification and combined chromosome 7 gain/chromosome 10 loss (+7/-10), were completed for all 32 patients. However, as the diagnosis for the entire cohort was already definitively established through the integrated IDH-wildtype status and histopathological hallmarks, these additional molecular data were not used as inclusion criteria for the present study. None of the tumors met WHO 2021 GBM diagnostic criteria based on molecular features alone, because all cases showed grade 4 histopathological hallmarks. Inclusion criteria were as follows: i) Patients with a primary diagnosis of GBM confirmed by histopathological examination; ii) availability of sufficient tumor tissue for subsequent analyses; and iii) complete clinical and pathological information. Exclusion criteria included: i) Recurrent GBM; ii) prior chemotherapy or radiotherapy before surgical resection; and iii) insufficient or poor-quality tissue samples unsuitable for analysis. Additionally, adjacent normal brain tissues (NBTs) were collected from 21 patients with GBM and used as normal controls. All surgical resections were performed by senior neurosurgeons with extensive clinical experience. NBTs were obtained from regions located at least 2 cm away from the tumor margin, ensuring accurate intraoperative differentiation between tumor tissue and non-tumoral brain tissue. The collected adjacent samples were evaluated microscopically by senior associate chief physicians and chief physicians in the pathology department using standard hematoxylin and eosin (H&E) staining. The absence of tumor infiltration was determined based on established histological criteria, including the absence of hypercellularity, nuclear atypia, mitotic figures, microvascular proliferation, or necrosis. Only tissues judged to be free of overt tumor infiltration were included as NBT controls. Notably, no molecular or genetic testing was performed to further confirm the absolute normality of these adjacent tissues, which remains a limitation of the current study. All enrolled patients and their families provided informed consent. The study was approved by the Clinical Ethics Committee of the Affiliated Hospital of North Sichuan Medical College (approval no. 2024020).

Human GBM cell lines (U87, U251, A172 and T98G) were obtained from Keycell Biotech (Wuhan) Co., Ltd. (cat. nos. U87-QS-H103, U251-QS-H007, A172-QS-H021 and T98G-QS-H034). The HA1800 astrocyte cell line was supplied by Xiamen Immocell Biotechnology Co., Ltd. (cat. no. HA1800-IM-H437). The U87 cell line corresponds to the U87 MG ATCC version (cat. no. HTB-14) and its precise origin is unknown. The identity of the U87 cell line was authenticated by STR profiling prior to use. The cells were cultured in Dulbecco's Modified Eagle's Medium (DMEM; Keycell Biotech (Wuhan) Co., Ltd.), supplemented with 10% fetal bovine serum (FBS; INTL KANG) and 1% penicillin-streptomycin solution [Keycell Biotech (Wuhan) Co., Ltd.]. The cells were then incubated at 37°C in a humidified incubator with 5% CO₂ (Shanghai Yiheng Technology Co., Ltd.).

The small interfering RNAs (siRNAs) investigated in the present study included si1-circRAD18, si2-circRAD18, circRAD18-negative control (NC), miR-1231 mimic, miR-1231 mimic NC, miR-1231 inhibitor and miR-1231 inhibitor NC, all designed and synthesized by General Biosystems (Anhui) Co., Ltd. The siRNAs targeting circRAD18 were specifically designed against the back-splice junction sequence unique to circRAD18. The mature sequence of miR-1231 was 5'-GUGUCUGGGCGGACAGCUGC-3'. The LUC7L2 overexpression plasmid VP051-CMV-MCS-EF1-ZsGREEN-T2A-PURO (OE-LUC7L2) and its corresponding empty vector (OE-NC) were constructed by Wuhan Huibao Bio-Technology Co., Ltd. U87 and U251 cell lines were used for transfection. In 96-well plates, cells in the logarithmic growth phase with optimal viability were plated at a density of 5×10³ cells per well and incubated overnight at 37°C in a 5% CO₂ incubator. Transfection was performed when the cells reached 80% confluency. For each well, 0.5 μl of si1-circRAD18, si2-circRAD18, circRAD18-negative control (NC), miR-1231 mimic, miR-1231 mimic NC, miR-1231 inhibitor or miR-1231 inhibitor NC at a concentration of 20 μM was used for transfection. According to the manufacturer's instructions, Lipofectamine^® 2000 (Invitrogen; Thermo Fisher Scientific, Inc.) was used to transfect the cells. Subsequent experiments were performed 48 h after transfection. Short hairpin RNAs (shRNAs), including sh1-circRAD18, sh2-circRAD18 and sh-circRAD18-NC, were produced by Wuhan Bio-Tower Biotechnology Co., Ltd. The pLVX-ShRNA2-Puro plasmid (Wuhan Bio-Tower Biotechnology Co., Ltd.) was used to construct a shRNA expression vector. An empty vector lacking the circRAD18 sequence was used as an NC. To produce lentiviruses, co-transfection of the shRNA expression vector (10 μg) and the packaging plasmids (15 μg) was performed in 293T cells from Keycell Biotech (Wuhan) Co., Ltd. when the cells reached 80% confluency. For lentiviral infection, U87 cells were plated in 6-well plates at a density of 5×10⁴ cells per well and cultured in complete medium overnight. A viral suspension with a multiplicity of infection of five was added to each well and incubated at 37°C. At 48 h after infection, the culture medium was replaced with fresh medium containing 2 μg/ml puromycin (Shanghai Yuanye Bio-Technology Co., Ltd.) for selection of stably transduced cells. The medium was changed every two to three days to remove dead cells and maintain selective pressure until stable transduction was indicated by the formation of resistant colonies. Subsequently, stably transduced cells were maintained in medium containing 2 μg/ml puromycin for routine culture. Table I lists the oligonucleotide sequences used for cell transfection.

TRIzol^® reagent (Invitrogen; Thermo Fisher Scientific, Inc.) was used to extract RNA from tissues and cell lines according to the manufacturer's protocol. For cell samples, RNA was extracted from 1×10⁶ cells. HiScript^® II Q Select RT SuperMix for qPCR (Vazyme Biotech Co., Ltd.) was used to reverse-transcribe circRAD18 and LUC7L2 mRNA into cDNA after RNA concentration and purity were assessed, according to the manufacturer's protocol. The miRNA 1st Strand cDNA Synthesis Kit (by tailing A; Vazyme Biotech Co., Ltd.) was used to reverse-transcribe miR-1231 according to the manufacturer's protocol. The cDNA was mixed with 2X Q3 SYBR qPCR Master Mix (ToloBio) for RT-qPCR, which was performed using specific primers on a ViiA 7 Real-Time PCR System (Applied Biosystems; Thermo Fisher Scientific, Inc.) in accordance with the manufacturer's instructions. The thermocycling conditions were as follows: Initial denaturation at 95°C for 10 min, followed by 40 cycles of denaturation at 95°C for 10 sec and annealing/extension at 60°C for 60 s. Subsequently, melting curve analysis was performed by heating at 95°C for 15 sec, cooling to 60°C for 60 sec and reheating to 95°C for 15 sec to verify amplification specificity. All experiments were performed with three independent biological replicates, and each sample was analyzed in triplicate. Relative gene expression was calculated using the 2^−ΔΔCq method (26), with GAPDH, β-actin and U6 as internal controls. The reverse primers for miR-1231 were purchased from Vazyme Biotech Co., Ltd. and other primers are listed in Table II.

Total RNA and genomic DNA (gDNA) were extracted from U87 cells using TRIzol^® reagent (Invitrogen; Thermo Fisher Scientific, Inc.) and a universal gDNA kit [Tiangen Biotech (Beijing) Co., Ltd.]. A total of 5 μg of RNA was collected and RNase R-treated and mock-treated groups were established. The cells were treated with 20 U/μl of RNase R (IVDShow Biotechnology Huailai Co., Ltd.) for 30 min at 37°C. A 20 μl reaction system was prepared as follows: Template RNA (2.5 μg), 10X Reaction Buffer (2 μl), RNase R (2 μl) and RNase-free ddH₂O (adjusted to a final volume of 20 μl). Following reverse transcription, RT-qPCR was performed using convergent and divergent primers. The amplified products were then analyzed using agarose gel electrophoresis. Table III lists the sequences of the primers, which were acquired from Beijing Tsingke Biotech Co., Ltd.

Cell climbing slides were used for FISH. Following a 20-min fixation in 4% paraformaldehyde [Keycell Biotech (Wuhan) Co., Ltd.] at room temperature, the cells underwent three 5-min washes with phosphate-buffered saline [PBS; Keycell Biotech (Wuhan) Co., Ltd.]. The slides were permeabilized with proteinase K (20 μg/ml) (Wuhan Servicebio Technology Co., Ltd.) for 2 min. After digestion, the slides were given three 5-min rinses with distilled water, followed by three 5-min washes with PBS. The slides were pre-hybridized by incubating them for 1 h at 37°C in pre-hybridization buffer. The hybridization buffer containing the first probe was used to perform the hybridization, which was then incubated overnight at 37°C in a humidified chamber. Post-hybridization washes were conducted with saline-sodium citrate (SSC) buffer (Wuhan Servicebio Technology Co., Ltd.) at 37°C, using the following steps: 2XSSC for 10 min, 1XSSC for 5 min twice and 0.5XSSC for 10 min. If non-specific hybridization occurred, an additional formamide wash was performed. The procedure was repeated for the second probe with the same pre-hybridization, hybridization and post-hybridization steps. After hybridization, 4,6-diamidino-2-phenylindole (DAPI; Beyotime Biotechnology) was used to counterstain the nuclei for 5 min in the dark. This was followed by washing and mounting with an anti-fade medium. An Olympus fluorescence microscope (Olympus Corporation) was used to observe and photograph the slides. All FISH probes used in the present study were commercially designed and synthesized by Sangon Biotech Co., Ltd. as 5'-Cy3-labeled oligonucleotide probes. The circRAD18 probe was 24 nt in length, and the miR-1231 probe was 20 nt in length. The probe sequences are listed in Table IV.

A 96-well plate was used to culture the transfected cells at a density of 5×10³ cells per well. After that, the medium was mixed with 10 μl of CCK-8 reagent [Keycell Biotech (Wuhan) Co., Ltd.] and incubated for 1 h. The absorbance of the cells at 450 nm at 24, 48 and 72 h after transfection was then measured using a microplate reader (BIOBASE). Cell viability was evaluated by analyzing these data.

Initially, U87 and U251 cell lines were plated at 1×10⁶ cells/well until confluent on 6-well plates. Next, a 200-μl pipette tip was used to make uniformly sized scratches in the cell monolayers. Each wound was imaged at 0, 12 and 24 h using an inverted microscope (Nikon Corporation) following cell washing with PBS to eliminate fragmented cellular debris. During the wound healing assay, the cells were maintained in serum-free DMEM [Keycell Biotech (Wuhan) Co., Ltd.]. No drugs were used in this assay.

The Transwell assay involved applying 100 μl of Matrigel (1 mg/ml; Corning, Inc.) to the upper chamber of the insert and incubating it for 1 h at 37°C. The upper chamber was used to seed glioma cells in the exponential growth phase. DMEM [Keycell Biotech (Wuhan) Co., Ltd.] was applied to the upper and lower chambers, together with 10% FBS (INTL KANG) and 1% penicillin-streptomycin [Keycell Biotech (Wuhan) Co., Ltd.]. The cells in the lower chamber were fixed with 70% ice-cold ethanol for 1 h and then stained with 0.5% crystal violet (Wuhan Servicebio Technology Co., Ltd.) at room temperature for 20 min. The number of invasive glioma cells was counted using an inverted microscope (Nikon Corporation).

Total protein was extracted from cells using RIPA lysis buffer supplied by Wuhan Servicebio Technology Co., Ltd. Protein concentration was measured using a BCA test kit (GBCBIO Technologies Inc.). Standard protocols were followed to separate equivalent amounts of protein (40 μg per lane) using 5% stacking gel and 10, 12 or 15% separating gels, depending on the molecular weight of the target protein, and transfer them onto PVDF membranes (MilliporeSigma). The membranes were blocked for 2 h at room temperature using 5% skimmed milk and then incubated with the primary antibodies overnight at 4°C. The next day, the membranes were incubated for 2 h at room temperature after secondary antibodies were added. ECL substrate (Wuhan Servicebio Technology Co., Ltd.) was then used to visualize the protein bands and capture the image. Densitometric analysis was performed using Image-Pro Plus software (version 6.0.0.260; Media Cybernetics, Inc.). Details of the primary and secondary antibodies, including their dilutions, are provided in Table V.

For the construction of the luciferase reporter gene, the 3'UTR sequences of circRAD18 or LUC7L2 containing the predicted and mutated binding sites were inserted into the psiCHECK-2 vector (BIOFENG). The constructs included circRAD18-WT, circRAD18-MUT, LUC7L2 3'UTR-WT and LUC7L2 3'UTR-MUT. Then, using Lipofectamine^® 2000 (Invitrogen; Thermo Fisher Scientific, Inc.), the luciferase reporter gene was co-transfected into U87 cells with either a miR-1231 mimic or a miR-1231 mimic NC. The cells were incubated at 37°C in a 5% CO₂ incubator, and the medium was replaced with normal culture medium after 6 h. After 48 h of transfection, the dual-luciferase reporter gene assay kit (Beyotime Biotechnology) was used to quantify luciferase activity according to the manufacturer's instructions.

Sangon Biotech Co., Ltd. synthesized the biotin-labeled circRAD18 and NC probes. RNA pull-down was performed using the RNA-Binding Protein Immunoprecipitation Kit (MilliporeSigma). U87 cells were collected by centrifugation at 215 × g for 5 min at 4°C and then lysed. The lysates were centrifuged at 18,757 × g for 10 min at 4°C. For each reaction, 100 μl cell lysate was mixed with 200 μl RIP washing buffer and 10 μg probe. The circRAD18 and control probes were incubated alongside the cell lysate overnight at 4°C. To create circRNA-probe-bead complexes, 50 μl pre-washed magnetic bead suspension was incubated with 300 μl probe-complex solution for 1 h at room temperature after hybridization. After bead incubation, the samples were briefly centrifuged at 61 × g for 1 min and the complexes were then washed six times with RIP washing buffer. Purified RNA attached to the beads was evaluated by RT-qPCR following bead washing using a washing buffer. Table VI lists the probe sequences.

A total of 18 six-week-old male BALB/c nude mice (body weight 18-20 g) were obtained from Hunan SJA Laboratory Animal Co., Ltd. Mice were housed in a specific pathogen-free animal facility using individually ventilated cages supplied by Suzhou Fengshi Experimental Animal Equipment Co., Ltd. The housing conditions were maintained at a controlled temperature of 21.0-25.0°C and relative humidity of 40-70%, under a 12-h light/dark cycle with an air exchange rate of ≥15 times per h. Mice had ad libitum access to standard laboratory chow (Wuhan Wanqian Jiaxing Biotechnology Co., Ltd.) and sterilized purified water. U87 cells were divided into three groups. After lentiviral infection, nude mice's axillae were subcutaneously injected with sh-circRAD18-NC, sh1-circRAD18 and sh2-circRAD18 cells. Once tumors had formed, tumor volume was recorded every 3 days and growth curves were plotted. After 4 weeks, mice were euthanized by intraperitoneal injection of 1.25% tribromoethanol at a dose of 250 mg/kg, followed by cervical dislocation after confirmation of deep anesthesia. Tumors were then excised and weighed. Tumor volume was calculated using the formula: volume=0.5x length × width². Animal experiments were performed according to the principles and procedures approved by the Animal Experimentation Ethics Committee of Affiliated Hospital of North Sichuan Medical College (approval no. 202410171).

Dehydration of the tissue samples was achieved using a graded ethanol series (China National Pharmaceutical Group Corporation), followed by clearing with xylene (China National Pharmaceutical Group Corporation) and embedding in paraffin (China National Pharmaceutical Group Corporation) at 60°C. Tissue sections were cut at 3 μm using a microtome (Leica Microsystems GmbH), placed on slides and allowed to dry for 3 h at 60°C. Following xylene deparaffinization, graded ethanol was used to rehydrate the slides and distilled water was used to rinse them. 0.01 M Tris-EDTA [pH 9.0; Keycell Biotech (Wuhan) Co., Ltd.] was used for antigen retrieval on a hot plate for 15 min, followed by cooling to room temperature. Endogenous peroxidase activity was blocked with 3% hydrogen peroxide (China National Pharmaceutical Group Corporation) for 15 min at room temperature, then incubated with normal goat serum (Wuhan Boster Biological Technology, Ltd.) for 30 min at room temperature. The sections were incubated overnight at 4°C with rabbit polyclonal anti-LUC7L2 antibody (Novus Biologicals; cat. no. NBP2-33621; dilution, 1:200). Following a wash, the sections were incubated with an HRP-labeled goat anti-rabbit/mouse secondary antibody provided in an immunohistochemical detection kit (Dako; Agilent Technologies, Inc.; cat. no. K5007) for 30 min at 37°C. DAB chromogen from the same kit was used for signal detection, and the reaction was stopped using deionized water. Mayer's hematoxylin [Keycell Biotech (Wuhan) Co., Ltd.] was used to counterstain the sections for 2 min at room temperature, after which they were dehydrated, cleared and mounted using neutral resin. Slides were imaged using an Olympus BX53 light microscope (Olympus Corporation) at magnifications of ×100 and ×400.

Ultracentrifugation was used to isolate the exosomes from the cell culture medium. Cell debris was removed using centrifugation at 400 × g for 15 min at 4°C. The supernatant was then collected using centrifugation at 10,000 × g for 30 min at 4°C. After obtaining the supernatant, it was ultracentrifuged once again for 70 min at 100,000 × g at 4°C. The exosomes were rinsed in PBS and ultracentrifuged at 100,000 × g for 70 min at 4°C after the top layer (supernatant) was removed. 100 μl PBS was used to resuspend the exosomes. Western blotting was used to validate exosome identification by detecting the exosome markers CD9 and CD63.

Copper grids were plasma-cleaned for 10-30 sec with the carbon-coated side facing up. Sealing film was placed on an ice-cold plate to cool. Grids were transferred onto the chilled sealing film, followed by sequential addition of 20 μl drops of the sample, ultrapure water and phosphotungstic acid (PTA) solution. After cooling the grids, samples and solutions for 1-2 min, the carbon-coated side was gently placed on the sample droplets for 1 min to allow adsorption. Excess liquid was removed by touching the grid with filter paper until dry, followed by an additional 10 sec of blotting. The grids were then placed on ultrapure water droplets, immediately lifted and blotted dry with filter paper. The grids were stained by placing on PTA droplets for 0.5-1 min, followed by blotting to remove excess PTA. Finally, grids were air-dried in a shaded area and observed using TEM (Tecnai G20 TWIN; FEI; Thermo Fisher Scientific, Inc.).

The NanoSight NS300 (Malvern Instruments, Ltd.) was used to examine the size distribution of exosomes. A camera captured the light scattered by exosomes under laser illumination, producing a video of their Brownian motion. The size of individual particles was tracked and analyzed using NTA software (version 3.3 Dev Build 3.3.104; NanoSight, Malvern Instruments, Ltd.).

The Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/geo/) provides information on the expression of circRNAs, miRNAs and mRNAs in GBM. The exoRBase database (http://www.exorbase.org:9000/) provides access to circRNA data from exosome-derived GBM cells. The Chinese Glioma Genome Atlas (CGGA) database (http://www.cgga.org.cn) provides clinical information, WHO grades and expression data for miR-1231 and LUC7L2, rated from low to high. The possible target miRNAs of circRAD18 were then predicted using the CircBank (http://www.CircBank.cn/) and CircInteractome (http://circinteractome.nia.nih.gov/) databases. Utilizing the TargetScan database (https://www.targetscan.org/), the mirDIP database (https://ophid.utoronto.ca/mirDIP/index_confirm.jsp) and the miRwalk database (http://mirwalk.umm.uni-heidelberg.de/), possible miR-1231 target mRNAs were predicted.

Statistical analysis was conducted using SPSS Statistics software (version 25.0; IBM Corp.). Statistical significance between groups was assessed using the t-test, Mann-Whitney U test, or one-way analysis of variance (ANOVA) followed by Bonferroni's post hoc test, as appropriate. Survival rates were analyzed using the Kaplan-Meier method. Results are presented as mean ± standard error of the mean (SEM). P<0.05 was considered to indicate a statistically significant difference.

The GEO database (GSE146463; https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE146463) was initially screened to investigate the difference in circRNA expression between GBM cells and NBTs. A fold change of >1 and a P<0.05 were used as screening criteria and it was found that 1,965 circRNAs were substantially expressed in GBM (Fig. 1A and B). Subsequently, the exoRBase database was searched and 511 circRNAs expressed in GBM-derived exosomes identified. In parallel, a total of 1,965 circRNAs were identified from GSE146463. Among them, 45 circRNAs were shared between the two datasets, whereas 1,920 and 466 circRNAs were specific to GSE146463 and exoRBase, respectively (Fig. 1C). Among the 45 circRNAs overlapping between GSE146463 and exoRBase, circRAD18 was selected for further investigation because it exhibited relatively high expression in GBM-derived exosomes compared with other tumor types (Fig. 1D). According to the circBase database, circRAD18 is derived from the RAD18 gene, with its splicing structure depicted in Fig. 1E. As circRNAs are resistant to RNase R digestion, RNase R treatment experiments were performed. The findings showed that whereas circRAD18 expression was mostly unaltered after RNase R treatment, linear RAD18 mRNA expression was markedly decreased (Fig. 1F). The effective amplification of circRAD18 from cDNA using divergent primers and its resistance to RNase R treatment were verified by RT-qPCR and agarose gel electrophoresis (Fig. 1G). These findings confirmed that circRAD18 adopts a circular structure. Furthermore, circRAD18 was found to be notably more abundant in GBM cell lines and tissues than in normal astrocyte HA1800 cells and NBTs (Fig. 1H and I). U87 and U251 cell lines, which exhibited the highest circRAD18 expression, were selected for subsequent experiments. Since the intracellular location of circRNAs is frequently associated with their function, FISH experiments combined with cytoplasmic/nuclear fractionation analyses showed that circRAD18 was predominantly located in the cytoplasm (Fig. 1J-L).

A total of two siRNAs that target circRAD18 were created and transfected into GBM cells to examine the function of circRAD18 in the development of GBM. RT-qPCR results confirmed that these two siRNAs effectively reduced circRAD18 expression (Fig. 2A) without affecting linear RAD18 mRNA levels (Fig. 2B). As demonstrated by CCK-8 assays, circRAD18 knockdown markedly reduced GBM cell proliferation (Fig. 2C and D). GBM cell migration was markedly impaired by circRAD18 knockdown, as shown by wound healing assays (Fig. 2E and F). Additionally, Transwell assays showed that knocking down circRAD18 reduced the invasive capacity of GBM cells (Fig. 2G). Furthermore, circRAD18 knockdown markedly decreased the amounts of invasion-related proteins in GBM cells, including MMP2, MMP9, MMP14, N-cadherin and vimentin, according to western blot analysis (Fig. 2H and I). According to the findings, circRAD18 contributes to GBM cell invasion, migration and proliferation.

Prior research has demonstrated that circRNAs primarily control gene expression by serving as miRNA sponges. Bioinformatic analysis via CircBank and CircInteractome databases was used to identify miRNA targets of circRAD18, selecting miR-1231, miR-1248, miR-556-5p and miR-626 for further validation (Fig. 3A). The analysis of the GEO database (GSE90603) identified 807 miRNAs with downregulated expression (fold change <1; P<0.05) in GBM cells (Fig. 3B), including miR-1231 (Fig. 3C). Additionally, research by Zhang et al (27) and Wang et al (28) showed that GBM is associated with under-expression of miR-1231. CGGA database data indicated that lower miR-1231 expression correlates with increased glioma malignancy, with lower miR-1231 levels corresponding to higher WHO grades (Fig. 3D). Additionally, patients with higher miR-1231 expression exhibited longer survival times (Fig. 3E). FISH experiments and cytoplasmic/nuclear fractionation analyses revealed that circRAD18 co-localizes with miR-1231 in the cytoplasm (Fig. 3F-H). miR-1231 mimics or a control were co-transfected into U87 cells using wild-type and mutant luciferase reporter vectors to confirm the binding location of circRAD18 with miR-1231. The outcomes demonstrated that cells transfected with a wild-type vector and treated with miR-1231 mimics exhibited a substantial decrease in luciferase activity, but cells transfected with a mutant vector did not exhibit this effect (Fig. 3I and J). Using a biotin-labeled circRAD18 probe, RNA pull-down experiments were performed in U87 cells to verify direct interaction between circRAD18 and miR-1231. Compared with the control oligonucleotide probe, the circRAD18 probe markedly increased pull-down efficiency and enriched miR-1231 (Fig. 3K). Subsequently, we examined miR-1231 expression after circRAD18 knockdown and found that miR-1231 was upregulated (Fig. 3L). These results suggest that circRAD18 interacts with miR-1231, playing a regulatory role in the process.

To identify the molecular targets of miR-1231, the TargetScan, miRDIP and miRWalk databases were used to predict 61 downstream targets (Fig. 4A). The GEO database (GSE4290) was then searched and 2,499 highly expressed mRNAs in GBM cells identified (fold change >1; P<0.05; (Fig. 4B), with 9 mRNAs overlapping the 61 predicted targets (Fig. 4C). The expression of these nine mRNAs across glioma grades was analyzed using the CGGA database and it was discovered that LUC7L2 expression was noticeably greater in high-grade gliomas than in low-grade gliomas (Fig. 4D), whereas the other candidate genes displayed lower expression levels or lacked a clear association with tumor grade (Fig. S1). Higher LUC7L2 expression was also associated with poorer survival outcomes (Fig. 4E). Studies have shown that miRNAs regulate target mRNAs by inhibiting transcription or inducing degradation through binding to the 3'-UTR (29-31). It was hypothesized that miR-1231 interacts with the 3'UTR of LUC7L2 mRNA to control GBM development. A dual-luciferase reporter experiment was employed to confirm that miR-1231 binds to LUC7L2. U87MG cells were co-transfected with either wild-type or mutant LUC7L2 3'-UTR plasmids and miR-1231 mimics. Luciferase activity was considerably decreased by co-transfection with the wild-type plasmid and miR-1231 mimics (Fig. 4F and G). Furthermore, transfection of miR-1231 mimics was found to decrease LUC7L2 protein levels substantially (Fig. 4H). These results implied that miR-1231 directly targets the gene LUC7L2.

The present study performed a series of experiments in U87 and U251 cells to explore whether circRAD18 is involved in GBM progression in association with the miR-1231/LUC7L2 axis. Using CCK-8, wound healing and Transwell assays, it was observed that miR-1231 inhibitor transfection enhanced GBM cell proliferation, migration and invasion (Fig. 5A-E), while knockdown of circRAD18 attenuated these effects (Fig. 5F-J). Western blot analysis revealed that LUC7L2 protein levels were notably reduced by circRAD18 knockdown (Fig. 6A). Transfection of miR-1231 inhibitor markedly increased LUC7L2 protein levels (Fig. 6B), while co-knockdown of circRAD18 partially reversed this effect in GBM cells (Fig. 6C). Furthermore, to evaluate the downstream role of LUC7L2, LUC7L2-overexpressing cell models were established (Fig. 6D-E). Overexpression of LUC7L2 promoted GBM cell proliferation, migration and invasion as assessed by CCK-8, wound healing and Transwell assays (Fig. 6F-J) and knockdown of circRAD18 partially attenuated these effects (Fig. 6K-O). Collectively, these experiments supported the involvement of a circRAD18/miR-1231/LUC7L2 regulatory axis in GBM cell proliferation, migration and invasion.

The U87 cell line was split into three groups: A NC group and two shRNA knockdown groups that targeted circRAD18 (sh1-circRAD18 and sh2-circRAD18) to assess the effect of circRAD18 on GBM cell growth in vivo. Then, cells were subcutaneously injected into the axillae of BALB/c nude mice to establish xenograft tumor models. The xenograft models exhibited a notable decrease in tumor growth rate, volume and weight in the circRAD18 knockdown group (Fig. 7A-E). RT-qPCR analysis revealed that in the circRAD18 knockdown group, circRAD18 expression decreased (Fig. 7F), miR-1231 expression increased (Fig. 7G) and LUC7L2 mRNA expression decreased (Fig. 7H) compared with the control group. LUC7L2 expression was much lower in the circRAD18 knockdown group, according to IHC (Fig. 7I). These findings suggested that circRAD18 facilitates malignant GBM cell development in vivo.

By searching the exoRBase database, high expression of circRAD18 was identified in exosomes derived from GBM cells. Next, the EVs were isolated from the culture supernatant of the HA1800 normal astrocytes and the U87 cells. TEM and NTA showed that the EVs were round, had a bilayer membrane and ranged in size from 50-120 nm (Fig. 8A and B). The exosomal markers CD63 and CD9 were shown to be positively expressed in the isolated EVs by western blot analysis (Fig. 8C). These findings indicated successful isolation of exosomes. circRAD18 expression was substantially greater in exosomes obtained from U87 cells than in exosomes derived from HA1800 cells, according to RT-qPCR analysis (Fig. 8D). Based on these results, exosomal circRAD18 could potentially serve as a biomarker for GBM.