Patients and data collection.
Our study included CSF and tumor samples from 44 patients who were treated for glioma at Qilu Hospital of Shandong University from November 2017 to October 2019. The whole course CSF sEV samples collected preoperatively were defined as ‘pre’ , and the postoperative CSF samples were defined as ‘p1’ to ‘p6’ according to the collecting series. The 12 cases of non-tumorous brain tissues were obtained from the cortex of decompressive surgery patients with brain trauma or hypertensive intracerebral hemorrhage between November 2018 and April 2019 from the Department of Neurosurgery of the Qilu Hospital of Shandong University, the Second Hospital of Shandong University, and the 5th People's Hospital of Jinan Shandong University. And 3 normal CSF (nor) were obtained from shunt procedures of normal pressure hydrocephalus (NPH) patients between February 2018 and December 2018 from the Department of Neurosurgery of the Qilu Hospital of Shandong University. All glioma patients had received surgical treatment microscopically. All patients authorized the informed consent, and this study was conducted in accordance with institutional ethical standards, the Declaration of Helsinki, and national and international guidelines. Ethics approval was obtained from the Clinical Research Ethics Committee of Qilu Hospital Shandong University.
Magnetic resonance imaging.
All patients underwent brain MRIs as indicated by the standard of care with standard sequences including axial T1-weighted, T2-weighted FLAIR and contrast T1-weighted images. Brain MRIs were reviewed by an experienced neuroradiologist without knowledge of the CSF sEV sequencing results. The tumor regions were semi-automatically segmented slice by slice using 3D Slicer (www.slicer.org) on FLAIR and contrast T1-weighted sequence. To truncate outlier intensities, bias correction was applied on FLAIR and contrast T1-weighted sequence images to compensate for intensity non-uniformities using N4 algorithm (32) implemented in "extrantsr" package and then the intensities of images were normalized using "WhiteStripe" package in R which conducts white stripe normalization procedure (33). Tumor regions registration was performed by using "lesymap" package. The anatomical image of the patient was registered to the same geometric space as the template, the ICBM 152 2009c Nonlinear atlas, with the tumor regions being excluded. The transformation matrices obtained from the above registration were applied to the tumor region, and the tumor region was brought in template space. The tumor burden (sum of the products of the diameters, SPD), radiographic progression and presence or absence of radiographic signs of tumor spread to subependymal, pial and leptomeningeal sites (CSF type) was according to the previous publications (34).
Isolation and characterization of sEV from CSF.
sEVs were isolated from CSF by differential centrifugation, according to the previous publications (35). After removing cells and other debris by centrifugation at 2000 g for 30min, the supernatant was centrifuged at 12,000 g for 45 min to remove shedding vesicles and the other vesicles with bigger sizes. The supernatant was centrifuged at 110,000 g for 70 min and re-suspended in 10mL PBS. Finally, the suspension was re-centrifuged at 110,000 g for 70 min (all steps were performed at 4 C); sEV were collected and re-suspended in 50μL PBS for further trans-omics RNA sequencing. The transmission electron microscopy assay of CSF sEV pellet was examined and photographed with an FEI Tecnai spirit TEM T12(FEI Tecnai Spirit 120 kv), according to the previous publications (35).
RNA library preparation and sequencing.
Total RNA from tissues and sEV of CSFs was isolated by using TRIzol Reagent (Invitrogen) according to the manufacturer’s instructions. RNA quality and quantity were assessed using a Nanodrop 2000 spectrophotometer (Thermo Fisher Scientific) and Agilent 2100 bioanalyzer (Agilent Technologies). Short-chain RNAs (miRNAs) and long-chain RNAs (mRNAs, lncRNAs and circRNAs) libraries were prepared by using NEBNext® Multiplex Small RNA Library Prep Set for Illumina® (NEB) and NEBNext® UltraTM RNA Library Prep Kit (NEB), respectively. Long-chain RNA and miRNA sequencing was separately performed using the HiSeqX and HiSeq2500 platform (Illumina) according to the Illumina standard protocol by Beijing Novel Bioinformatics Co., Ltd. (https://en.novogene.com/).
Quantification of transcripts abundance.
Clean reads were obtained after removal of reads containing adapters, reads containing ploy-N and low-quality reads from the raw Illumina sequencing reads. For long-chain RNAs, human reference genome and annotation files were downloaded from the genome website (NCBI/UCSC/Ensembl). Then, clean reads were aligned against the reference genome using HISAT2. To quantify the gene expression level, HTSeq was used to count the read numbers mapped for each gene. For miRNAs, clean reads were aligned against the human reference database (miRbase, http://www.mirbase.org/) using Bowtie and exact matches to known mature miRNA sequences in miRBase were counted. The Transcripts Per Million (TPM) of each miRNA was calculated based on the miRNA read counts. CIRI2 and Find_circ, the circRNA identification algorithms, were employed to predict circRNAs.
Differential expression analysis.
Normalization and differential expression analysis between glioma and control tissues as well as in sEV between preoperative (pre) and control CSF (p1 and nor) were performed by using the DEseq2 R package. The P-value was calculated and corrected for multiple testing using the Benjamini-Hochberg method. Normalized expression boxplots and volcano plots were generated by ggplot2 and ggpubr R package. Sample clustering, principal components analysis (PCA), hierarchical clustering were performed by using hcluster, prcomp and pheatmap R package, respectively.
The miRNA target prediction.
Based on the differentially expressed mRNAs, lncRNAs and circRNAs between glioma and control tissues, target mRNA, lncRNA and circRNA of differentially expressed miRNAs in sEV of preoperative CSF (pre) and glioma tissue were obtained by these following steps. The target mRNAs were obtained using Targetscan (http://www.targetscan.org/), miRDB (http://www.mirdb.org/), miRTarBase (http://mirtarbase.mbc.nctu.edu.tw/) which were further filtered by Pearson Correlation Coefficient (PCC) analysis between the expression levels of mRNA and miRNA in glioma tissues (PCC< -0.3 and P <0.05); the target lncRNAs were obtained by DIANA-LncBasev2.0 (http://carolina.imis.athena-innovation.gr/diana_tools/web/index.php?r=lncbasev2/index-predicted) and StarBase 2.0 (http://starbase.sysu.edu.cn/starbase2/index.php); the target circRNAs were obtained by miRDB (http://www.mirdb.org/) and StarBase 2.0 (http://starbase.sysu.edu.cn/starbase2/index.php). Finally, a competitive endogenous RNA (ceRNA) network was constructed by using Cytoscape 3.7.0 software after the PCC analysis of expression levels between lncRNA and mRNA (PCC >0.3 and P <0.05) as well as circRNA and mRNA (PCC >0.3 and P <0.05) in glioma tissues.
Pathway enrichment analysis.
Pathway enrichment analysis was done on target genes of up- and down-regulated miRNAs in preoperative CSF sEV (pre) compared with glioma tissue using the clusterProfiler R package. Firstly, symbol gene IDs were converted to Entrez gene IDs. Then, Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis was implemented.
Cell culture.
P3 cell line was kindly provided by Prof. Rolf Bjerkvig, University of Bergen, which was isolated from human glioblastoma tissue. P3 cells were maintained in neurobasal medium (NBM) supplemented with GlutaMAX (2mM), B-27(1×), penicillin/streptomycin (1×), heparin(32 IE/ml), EGF(20 ng/ml) and FGF2(20 ng/ml). U87MG, U251, A172 (Chinese Academy of Sciences Cell Bank) and LN229 cells (ATCC) were cultured in DMEM (Sigma) supplemented with 10% FBS (Thermo Fisher Scientific). All the cell lines were incubated at 37℃ with 5% CO2 and 95% air. All cells were authenticated by short tandem repeat (STR) profiling and routinely tested for mycoplasma contamination.
Peripheral blood mononuclear cells (PBMC) and monocyte isolation
PBMCs were isolated from the venous blood of healthy donors and patients using lymphocyte separation medium (LTS1077, TBD, China). Blood was separated using standard density gradient centrifugation (30 minutes at 500 g at 21°C), and the PBMC layer was carefully transferred to another tube. CD14 + monocytes were separated from PBMCs using CD14 MicroBeads (Miltenyi Biotec, Germany) according to the manufacturer's protocol. Purity (>95%) was confirmed by flow cytometry.
MDSC induction
PBMCs (5 x 10 5) from healthy donors were cultured in 12-well plates with 0.5 ml of exosome-depleted RPMI 1640 complete medium. CSF- or glioma cell-derived exosomes were added to the culture medium and co-cultured with PBMCs for 72h. PBMCs were examined via flow cytometry following the co-culture.
Exosomal miRNA transportation assay
Cy3-labeled miR-1298-5p was purchased from GenePharma. GDEs were isolated from the supernatant of U87MG and P3 cells transfected with Cy3-labeled miR-1298-5p and used to
treat monocytes. Two days later, the monocytes were collected and stained with DAPI
and FITC-phalloidin.
Exosome uptake assay
Exosomes were labeled with PKH67 (Sigma-Aldrich, USA) according to the manufacturer's protocol. PKH lipophilic dyes are fluorescent and their aliphatic domains intercalate into lipid bilayers, such that exosomes stained with PKH67 can be visualized via fluorescence microscopy. The PKH67 labeled exosomes were incubated with monocytes. Once the PKH67 labeled exosomes are internalized by monocytes, the fluorescence signal from PKH67 can be observed within the recipient cells. Briefly, exosomes were reconstituted in 50 µl PBS before 1 ml of Diluent C was added. Four microliters of PKH67 dye was added to 1 ml of Diluent C before being added to the exosomes. The samples were mixed gently for 4 min. To neutralize the excess dye, the PKH67-labeled exosome solution was mixed with 3 ml 0.5% BSA and centrifuged at 100,000 g for 1 hour. The exosome pellet was resuspended in PBS and added to the culture medium of human monocytes. The monocytes were then fixed and examined under a fluorescence microscope.
T cell suppression assay
CD14-depleted PBMCs were labeled with 2.5 μM CFSE and stimulated with coated anti-CD3 (1 μg/ml, eBioscience) and soluble anti-CD28 (1 μg/ml, eBioscience) antibodies. Exosome-induced monocytes were cultured with these cells at a ratio of 1:2 in a U-bottom 96-well plate. Three days later, the cells were stained with anti-CD8-APC antibody and analyzed for CFSE dilution.
Induction of M2 Macrophages.
PBMCs were isolated from the blood of healthy donors as previously described (9). Blood was separated using standard density gradient centrifugation (30 minutes at 500g at 21°C, LTS1077, TBD). PBMCs were extracted from the interphase. CD14+ cells were selected using magnetic CD14-positive beads (Miltenyi Biotec, 130-050-201). To differentiate these monocytes into monocyte-derived macrophages, CD14+ cells were cultured in RPMI 1640 media (Thermo Fisher Scientific) supplemented with 10% FBS and 100ng/ml M-CSF (PeproTech). Six days later, the cells were cultured in 0.5ml Opti-MEM containing 4μl Lipofectamine3000 and 20pmol mimics. The Opti-MEM was replaced with complete RPMI 1640 medium 6 hours later. Two days later, cells were stained with anti-CD206-APC (Invitrogen, 17-2069-42). Flow cytometry was performed using the Beckman Coulter Gallios and data was analyzed using FlowJo software.
RNA extraction and quantitative reverse-transcription (qRT-PCR)
Total cell RNA was extracted using RNA-Quick Purification Kit (ESscience Biotech, China) according to the manufacturer’s protocol. ReverTra Ace qPCR RT Master Mix (Toyobo, Japan) was used to synthesize cDNA following the manufacturer’s instruction. qRT-PCR was performed with SYBR Green PCR Master Mix (Applied Biosystems, Foster City, USA). Expression data of microRNA and mRNA were normalized to the internal controls U6 and GAPDH, respectively. The relative expression levels were calculated using the ΔΔCt method. The sequence of primers are shown in Supplementary Table 6.
Western blotting
Whole-cell protein was extracted from glioma cells and MDSCs in RIPA buffer (Thermo Fisher Scientific, USA) and centrifuged at 12,000 rpm for 20 min. A BCA kit (Thermo, Waltham, MA, 23228) was used to measure the protein concentration. After immunoblotting, the proteins were transferred to a nitrocellulose membrane and incubated with specific antibodies. The following primary antibodies were used: β-actin (Proteintech, 60008-1-Ig), CyclinD1 (Cell Signaling Technology, 2978), P27 (Cell Signaling Technology, 3686), CDK6 (Cell Signaling Technology, 3136), p-AKT (Cell Signaling Technology, 4060), AKT (Cell Signaling Technology, 4691), SETD7 (Cell Signaling Technology, 2813), hnRNPA2/B1 (Cell Signaling Technology, 9304), Phosphorylated NF-κB p65 (S536) (Cell Signaling Technology, 3033), NF-κB p65 (Cell Signaling Technology, 8242).
Small interfering RNA, miR mimics and adenovirus vector transfection
Control microRNAs, miR-1298-5p mimics, were purchased from GenePharma (Shanghai, China). si-setD7, si-MSH2 and control siRNAs were purchased from RiboBio (Guangzhou, China). All sequences are listed in Supplementary Table 6. For microRNA and siRNA transfection, cells were incubated in 6-well plates overnight and transfected with LipofectamineTM 3000 reagent (Thermo Fisher Scientific, USA) according to the manufacturer’s protocol. The miR-1298-5p overexpression and control lentiviruses were synthesized by Genechem (Shanghai, China). The knockdown/overexpression efficiency of the siRNAs and virus are available in the supplementary materials.
CCK8 assay
Cell Counting Kit-8 (CCK-8) was used to measure cell viability according to the manufacturer’s instructions (Beyotime, China). Cells were seeded at 5,000 cells per well into 96-well plates and cultured at 37°C with different treatments. CCK-8 solution (10 µl) was added at 24, 48 and 72 h. Following incubation for 2 h, the absorbance at 450 nm (OD450) was measured using a Multimode Plate Reader (PerkinElmer, USA).
Cell cycle analysis
Glioma cells were and stained with Propidium iodide (PI) in the presence of RNase A for 15 min. Flow cytometer (BD Biosciences) was used to perform cell cycle analysis according to the protocol.
Ethynyl-2'-deoxyuridine (EdU) cell proliferation assay
EdU assay kit (Ribobio, China) was used to test the cell proliferation ability according to the manufacturer’s instructions. Glioma cells were seeded into wells of poly-l-ornithine precoated 12-well plates. Cells were then incubated with 200 μl of 5-ethynyl-20-deoxyuridine for 2 h at 37℃. Nuclei were counterstained with Hoechst 33342. Representative images were obtained with a Leica inverted fluorescence microscope.
Dual-luciferase reporter assay
The dual luciferase reporter plasmids (SETD7 WT/MUT and MSH2 WT/MUT) were designed and synthesized by GenePharma (Shanghai, China). HEK-293T cells were seeded in 96-well plates overnight (2×104 / well). For 3’UTR tests, the dual luciferase reporter plasmids (0.1 μg/ well) were co-transfected with miR-Nc and miR-1298-5p mimics (20 nM×0.5 μl/well). Approximately 48 hours after transfection, the cells were subjected to luciferase activity analysis using a Dual-Luciferase Reporter Assay System (Promega) following the manufacturer’s instructions.
Exosome isolation
Cells were cultured in DMEM supplemented with 10% exosome-depleted FBS under normoxic (21% O2) or hypoxic (1% O2) conditions. Exosomes were isolated from cell culture supernatant as previously described for later analysis.
Electron microscopy and qNano
Isolated exosomes were examined using Transmission Electron Microscopy (TEM) as previously described. qNano (Izon Sciences Ltd, NZ) was used for exosome particle size and concentration analysis.[18]
Cytokine assay
Cell culture medium was collected 72 hours after the indicated treatment. The secretion of TNF-β was detected by ELISA (Proteintech, USA) and the NO was measured using Greiss regent, according to the manufacturer’s instructions.
RNA Binding Protein Immunoprecipitation
The RIP assays were performed using an EZ-Magna RIP kit (Millipore). Lysates of 1 × 107 glioma cells obtained using complete RIP lysis buffer were immunoprecipitated with RIP buffer containing anti-hnRNPA2B1 antibody–conjugated magnetic beads (Abcam). The precipitated RNAs were analyzed by qRT-PCR. Mouse IgG was used as the negative controls.
RNA pull-down assay
The miR-1298-5p-binding proteins were examined using RNA pull-down assays according to the instructions of the Pierce Magnetic RNA-Protein Pull-Down Kit (ThermoFisher Scientific). Biotinylated miR-1298-5p and control sequences were synthesized by GenePharma (Shanghai, China). The cell lysate obtained using a Pierce IP Lysis Buffer (Thermo Fisher Scientific) was incubated overnight with biotinylated miR-1298-5p, followed by precipitation with streptavidin magnetic beads. The retrieved protein was eluted from the RNA-protein complex and analyzed by western blotting.
Animal studies
Luciferase labeled and stably transfected U87MG cells overexpressing miR-1298-5p or vector were injected into the brains of randomly grouped 4-week BALB/c nude mice (5×10 5 /mouse) to build the orthotopic xenograft model. Bioluminescence imaging was used to image the mouse brains every 5 days after glioma cell implantation. Next, we randomly chose 5 mice in each group and euthanized them on the same day (10 d). The brains were fixed with paraformaldehyde for further study. The remaining mice (5/group) were kept until death for survival analysis. All procedures that involved mice were approved by and under the requirements of the Animal Care and Use Committee of the Qilu Hospital of Shandong University.
Statistical analysis
The cut-off value between high and low miR-1298-5p expression was set as the expression level of a median sample. Survival analysis was performed using the Kaplan-Meier method and comparisons were done using the log-rank test. The one-way ANOVA test or Student’s t test were used for all other data comparisons using GraphPad Prism 8. All data are presented as the mean ± standard error and P-values < 0.05 were considered statistically significant.