To analyze the differences between GSCs and GBM cells, gene microarray data were obtained from GSE74304 (9) and GSE124145 (10) in the NCBI GEO database (http://www.ncbi.nlm.nih.gov/geo/). Both gene profiles were from the GPL570 platform [(HG-U133_Plus_2) Affymetrix Human Genome U133 Plus 2.0 Array], a typical array for detecting mRNA expression levels. GSE74304 contained data from two patient-derived GSC lines and corresponding GBM cell lines, and GSE124145 contained data from two human GSC lines (X01 and X03), one human GBM tissue and the GBM U251 cell line. Each sample contained two or three replicates. Thus, a total of 10 GSC samples and 12 GBM samples were available for experiments. Profiles filter criteria were as follows: GPL570 platform, publication date between 2015 and 2019, no previous drug treatment and ≥500 DEGs.

DEGs were identified using the GEO2R online tools (https://www.ncbi.nlm.nih.gov/geo/geo2r) between two GSC lines and two GBM cell lines in GSE74304, and two GSC lines and GBM tissue or cell lines in GSE124145 with an adjusted P<0.05 and ǀlog (fold-change) FCǀ >2. Overlapping genes were obtained by uploading the DEGs from the two datasets to Venn software online (http://bioinformatics.psb.ugent.be/webtools/Venn/). The DEGs with log FC >0 were considered as upregulated genes and those with log FC <0 were considered as downregulated genes in GSCs compared with in GBM cells.

GO and KEGG pathway analysis was performed via the DAVID website (https://david.ncifcrf.gov/) to identify significantly enriched molecular functions, cellular components, biological processes and biological pathways [P<0.05; enrichment score = -lg (P-value)]

PPI analysis among overlapping DEGs was performed using the STRING database (https://string-db.org/), and network diagrams were constructed via Cytoscape software (v3.6.1; http://www.cytoscape.org/) (maximum number of interactors =0; confidence score ≥0.4). The hub genes were obtained by module analysis with the Molecular Complex Detection (MCODE) plugin in Cytoscape according to the following criteria: Node score cut-off, 0.2; degree cut-off, 2; k-core, 2; MCODE scores, ≥5; and maximum depth, 100.

To validate the hub genes through survival analysis, the GEPIA2 website (http://gepia2.cancer-pku.cn/#index) was used, an online database for analysis of RNA expression levels, based on the Genotype-Tissue Expression Projects and The Cancer Genome Atlas (11). Kaplan-Meier survival curves were analyzed using the log-rank test. Log-rank P-value, hazard ratio (HR) and HR P-value are displayed in the figures.

The U87 cell line (glioblastoma of unknown origin) was purchased from the American Type Culture Collection (ATCC^® HTB14™). U87 cells were maintained in DMEM/F12 (Corning, Inc.) supplemented with 10% FBS (Gibco; Thermo Fisher Scientific, Inc.) and 1% penicillin/streptomycin (Gibco; Thermo Fisher Scientific, Inc.) and cultured at 37°C with 5% CO₂. GSCs were maintained in stem cell medium as previously described (9,12,13). After being cultured at 37°C in a humidified 5% CO₂ incubator for 12 h, U87 cells were infected with negative control lentivirus (GV341; empty vector) or prolyl 4-hydroxylase subunit α2 (P4HA2) overexpression lentivirus at a multiplicity of infection of 5 for 12–16 h. The virus was purchased from Shanghai GeneChem Co., Ltd. P4HA2 siRNA-1 (sense, 5′-GAACCAAGUACCAGGCAAUTT-3′ and antisense, 5′-AUUGCCUGGUACUUGGUUCTT-3′), siRNA-2 (sense, 5′-GCAGCAUAUCACAGGGUUATT-3′ and antisense, 5′-UAACCCUGUGAUAUGCUGCTT-3′), siRNA-3 (sense, 5′-GCAAGUGGGUCUCCAAUAATT-3′ and antisense, 5′-UUAUUGGAGACCCACUUGCTT-3′) and scrambled negative control (sense, 5′-UUCUCCGAACGUGUCACGUTT-3′ and antisense, 5′-ACGUGACACGUUCGGAGAATT-3′) were synthesized by OBiO Technology (Shanghai) Corp., Ltd. After being cultured at 37°C in a humidified 5% CO₂ incubator for 12 h, U87 cells were transfected with siRNAs (50 nM) using Lipofectamine^® 3000 (Thermo Fisher Scientific, Inc.) according to the manufacturer's protocol for 12 h at 37°C. Subsequent experiments were performed 48 h after lentivirus infection or 24 h after siRNA transfection.

Total RNA was extracted using TRIzol^® reagent (Thermo Fisher Scientific, Inc.), cDNA synthesis was performed with a RevertAid First Strand cDNA Synthesis kit (Thermo Fisher Scientific, Inc.) according to the manufacturer's protocol and qPCR was performed with Universal SYBR Green Master Mix (Roche Diagnostics) according to the manufacturer's protocol using the following PCR conditions: 95°C for 10 min, followed by 40 cycles of 95°C for 30 sec, 60°C for 30 sec and 72°C for 40 sec, followed by a final elongation step of 10 min at 72°C. P4HA2 expression was normalized to GAPDH expression calculated using the 2-ΔΔCq method (14). The primer sequences were as follows: P4HA2 forward, 5′-GCCAAAGCCCTGATGAGACT-3′ and reverse, 5′-GCTCCATCCACAACACCGTA-3′; and GAPDH forward, 5′-TCATCATCTCTGCCCCCTCT-3′ and reverse, 5′-GTGATGGCATGGACTGTGGT-3′.

For EdU incorporation, a Cell-Light EdU Apollo567 In Vitro kit (Guangzhou RiboBio Co., Ltd.) was used for EdU and Hoechst staining according to the manufacturer's instructions. EdU-positive cells were analyzed under a fluorescence microscope (Axio Scope. A1; Zeiss AG) using ×20 magnification.

U87 cells were harvested 48 h after infection or 24 h after transfection and then fixed with ice-cold 75% ethanol at −20°C for 24 h. After fixation, cells were stained with PI/RNase Staining Buffer (BD Biosciences) and incubated for 30 min at room temperature. Cell cycle analysis was analyzed using the BD FACSCalibur system (BD Biosciences) with ModFit LT v3.3.11 software (Verity Software House, Inc.).

A total of 1,000 U87 cells were plated in 96-well plates and cultured at 37°C with virus infection or siRNA transfection. After culture for 0, 1, 2, 3 and 4 days, 10 µl CCK8 reagent (Dalian Meilun Biology Technology Co., Ltd.) was added into each well and incubated at 37°C for 2 h. The optical density (OD) was examined at a wavelength of 450 nm (OD 450 nm).

Infected or transfected cells were seeded in 6-well plates and incubated for 24 h until reaching 90% confluence. The monolayer was scratched with a 1-ml pipette tip and washed with PBS to remove cell debris. Cells were then cultured in DMEM/F12 supplemented with 2% FBS. The wound closure was measured using ImageJ software (v1.52a; National Institutes of Health) in photographs taken using a light microscope (Nikon Corporation) at ×10 magnification at the time of wounding (0 h), and then 24 and 48 h after wounding.

For migration assays, 3×10⁴ infected or transfected cells in DMEM/F12 were plated in the upper chambers of 24-well Transwell plates (pore size, 8 µm; Corning, Inc.), and the lower chambers were filled with 600 µl DMEM/F12 with 10% FBS. After being cultured at 37°C for 24 h, the non-migrating cells on the upper surface were removed with a cotton swab, and the migrated cells were fixed with 4% formaldehyde at room temperature for 20 min and then stained with 0.2% crystal violet at room temperature for 45 min. The pictures were taken using the AMAFD1000-EVOS™ FL Auto Imaging System (Thermo Fisher Scientific, Inc.) at ×20 magnification.

The data were presented as the mean ± SEM. Statistical analysis of the data was performed using GraphPad Prism 8.0 (GraphPad Software, Inc.) on at least three independent samples for each experiment. Two-tailed unpaired Student's t-test was used for comparison between two groups, and two-way ANOVA followed by Sidak's post-hoc test was used for comparing multiple groups. P<0.05 was considered to indicate a statistically significant difference.

A total of 10 GSC samples and 12 GBM samples were used in the present study. Through GEO2R online tools, 793 DEGs were extracted from GSE74304, with 381 upregulated genes and 412 downregulated genes, and 2,162 DEGs from GSE124145, with 531 upregulated genes and 1,631 downregulated genes in GSCs compared with in GBM cells. Overlapping genes were obtained with Venn software online. A total of 183 DEGs were screened, including 41 upregulated genes (log FC >0) and 142 downregulated genes (log FC <0) in GSCs compared with in GBM cells (Fig. 1A-C).

A total of 41 upregulated and 142 downregulated DEGs were analyzed via the DAVID website with the screening criteria of P<0.05 for upregulated genes, and both P<0.05 and enrichment score >3 for downregulated genes. The results of GO analysis revealed that: i) For molecular functions, upregulated DEGs were particularly enriched in the term ‘3′-5′ DNA helicase activity’, and downregulated DEGs were particularly enriched in the terms ‘collagen binding’, ‘integrin binding’, ‘heparin binding’ and ‘extracellular matrix binding’; ii) for cellular components, upregulated DEGs were enriched in the terms ‘nucleoplasm’ and ‘bicellular tight junction’, whereas downregulated DEGs were enriched in the terms ‘extracellular region’, ‘extracellular space’, ‘extracellular matrix’, ‘endoplasmic reticulum lumen’, ‘external side of plasma membrane’, ‘proteinaceous extracellular matrix’, ‘plasma membrane’ and ‘collagen trimer’; iii) for biological processes, upregulated DEGs were significantly enriched in the terms ‘negative regulation of transcription from RNA polymerase II promoter’, ‘brain development’, ‘regulation of auditory receptor cell differentiation’, ‘mitotic nuclear division’, ‘cellular response to cholesterol’, ‘spinal cord motor neuron differentiation’, ‘cell fate determination’, ‘labyrinthine layer blood vessel development’, ‘regulation of smoothened signaling pathway’, ‘DNA replication and cartilage condensation’, and downregulated DEGs were enriched in the terms ‘extracellular matrix organization’, ‘cell adhesion’, ‘leukocyte migration’, ‘negative regulation of sequence-specific DNA binding TF activity’, ‘response to drug’, ‘negative regulation of apoptotic process’, ‘collagen fibril organization’, ‘angiogenesis’, ‘positive regulation of cell migration’, ‘cellular response to hypoxia’ and ‘positive regulation of gene expression’ (Fig. 1D and F). KEGG analysis results demonstrated that upregulated DEGs were particularly enriched in the term ‘cell cycle’, whereas downregulated DEGs were particularly enriched in the terms ‘ECM-receptor interaction’, ‘hematopoietic cell lineage’, ‘TGF-β signaling pathway’, ‘PI3K-Akt signaling pathway’, ‘focal adhesion’, ‘FoxO signaling pathway’, ‘protein digestion and absorption’, ‘HIF-1 signaling pathway’, ‘p53 signaling pathway’ and ‘hypertrophic cardiomyopathy’ (Fig. 1D and E).

To construct the PPI network, the data for the 183 overlapping DEGs obtained from the STRING database were imported into Cytoscape software. The interaction network contained 41 upregulated genes and 142 downregulated genes (Fig. 2A). For further analysis of the key genes among them, the Cytoscape MCODE plugin was applied, and 21 hub genes were obtained, all of which were downregulated DEGs (Fig. 2B). Subsequently, the association between the expression levels of these genes and survival time in patients with GBM was validated with GEPIA2 analysis (Fig. 3). The survival curves revealed that low expression levels of four hub genes, P4HA2 [log-rank P=0.0011; P(HR)=0.0012], TGF-β induced [TGFBI; log-rank P=0.0082; P(HR)=0.0079], integrin subunit α3 [ITGA3; log-rank P=0.021; P(HR)=0.021] and thrombospondin 1 [THBS1; log-rank P=0.048; P(HR)=0.046], were associated with significantly prolonged survival time in patients with GBM (Fig. 3). The present study then focused on P4HA2 since it had the smallest P-value.

GSCs were generated from the U87 cell line, a human malignant glioblastoma cell line, as previously described (9,12,13). RT-qPCR was used to detect P4HA2 expression. In agreement with the bioinformatics analysis results, P4HA2 expression was significantly higher in U87 cells than in GSCs (Fig. 4A). Subsequently, U87 cells were infected with empty vector (Ctrl) or P4HA2 vector (P4HA2) expression lentiviruses. The RT-qPCR results revealed that P4HA2 mRNA expression was significantly increased after P4HA2 overexpression (Fig. 4B). To explore the effects on GBM cell proliferation, CCK8, EdU incorporation and cell cycle analyses were performed. As shown in Fig. 4C, the overexpression of P4HA2, compared with the control, induced significant increases in cell proliferation from day 1. The EdU-positive cells also significantly increased after P4HA2 viral transfection (Fig. 4D and E). Cell cycle analysis by flow cytometry aims to calculate the proportion of cells at different stages of mitosis based on the DNA content, so as to evaluate the proliferative ability of cells. Therefore, this method was used to further confirm the effect of P4HA2 on U87 cell proliferation. According to the amount of DNA, the whole cell cycle can be divided into G₀/G₁, S and G₂/M phases (15). DNA in the G₀/G₁ phase is diploid, while it is tetraploid in the G₂/M phase and in between in S phase (16,17). The ratio of cells in S phase and G₂/M phase is usually used as an index to assess the cell proliferation state (15,17). As shown in Fig. 4F-H, P4HA2 overexpression promoted the entry of cells into S phase and significantly increased the proportion of cells in S+G2 phase, thus explaining the robust proliferation of U87 cells. Overall, the present results demonstrated that P4HA2 increased the proliferation of GBM cells.

Since metastasis is the main malignant characteristic of GBM, wound healing and Transwell assays were conducted to investigate the effect of P4HA2 on U87 cell migration. In the wound healing assays, the scratch width narrowed significantly in the P4HA2 group from 24 h after transfection, thereby indicating that cells with P4HA2 overexpression exhibited increased migratory ability (Fig. 5A and B). In agreement with this finding, Transwell assays also demonstrated that the overexpression of P4HA2 promoted the migration of U87 cells compared with cells in the control group (Fig. 5C and D).

To further determine the function of P4HA2 in GBM, P4HA2 was knocked down in U87 cells with siRNA transfection. First, three siRNAs were designed targeting P4HA2; siRNA-1 (hereafter called siRNA) was selected for further experiments, since it exhibited the highest knockdown efficiency, as verified by RT-qPCR (Fig. 6A). To investigate the influence of P4HA2-knockdown on GBM, CCK8, EdU incorporation and cell cycle analyses were performed to examine cell proliferation, and wound healing and Transwell assays were performed to examine cell migration. As shown in Fig. 6B, P4HA2-knockdown induced significant decreases in cell proliferation from day 2, compared with the control group. The EdU-positive cells also significantly decreased after siRNA transfection (Fig. 6C and D). The results of cell cycle analysis by flow cytometry indicated that P4HA2-knockdown promoted G₁ phase arrest and delayed entry into S phase (Fig. 6E-G). Moreover, both the wound healing (Fig. 7A and B) and Transwell assays (Fig. 7C and D) revealed that P4HA2-knockdown significantly suppressed the migratory ability of U87 cells. Thus, P2HA2 overexpression promoted the proliferation and migration of GBM cells, while P4HA2-knockdown inhibited these processes.