This article is mentioned in:

Introduction

Glioma is a highly aggressive and malignant brain tumor that originates from glial cells (1) and provides structural and functional support to neurons. Glioblastoma multiforme (GBM) is the most common and fatal type of primary brain tumor (2), accounting for ~50% of all gliomas (3). Common signs include cognitive decline and changes in personality or behavior. As the tumor expands, it puts increased pressure on the surrounding brain tissue, leading to neurological deficits (4). Despite extensive research and advancements in medical science, the prognosis for patients diagnosed with glioma remains poor (5).

The aberrant reactivation of epithelial-mesenchymal transition (EMT) is associated with the malignant characteristics of tumor cells during cancer progression and metastasis (6). Studies have indicated that centromere protein F (CENPF) serves a pivotal role in tumor metastasis, as it can promote EMT progression in hepatocellular carcinoma and pancreatic cancer (7,8). Through in vitro experiments, Huang et al (9) reported that the CENPF/CDK1 signaling pathway facilitated the progression of adrenocortical carcinoma by regulating the G2/M phase of the cell cycle. Furthermore, Han et al (10) reported that in vitro CENPF modulates the proliferation, apoptosis and cell cycle of thyroid carcinoma cells, impacting tumor growth in mice.

Significant constituents of the kinesin family include kinesin superfamily protein 20A (KIF20A) and kinesin superfamily protein 4A (KIF4A). A previous study reported that the upregulation of KIF20A promotes tumor proliferation and invasion in renal clear cell carcinoma, with associations with adverse clinical outcomes (11). Huang et al (12) identified KIF20A as a prognostic marker in patients with estrogen receptor-positive breast cancer receiving tamoxifen adjuvant endocrine therapy. KIF4A enhances cell proliferation and migration via Hippo signaling, predicting poor prognosis in esophageal squamous cell carcinoma (13). Jin and Ye (14) reported that KIF4A promotes ovarian cancer cell proliferation and inhibits apoptosis by upregulating BUB1 mitotic checkpoint serine/threonine kinase (BUB1) expression. Moreover, marker of proliferation Ki-67 (MKI67) is already a widely utilized proliferation marke (15). Hu et al (16) reported its upregulation in uterine leiomyosarcoma, suggesting its potential as a diagnostic biomarker. Meng et al (17) also reported that KIF20A stimulates the expression of MKI67, promoting the growth and metastasis of bladder cancer.

Characterized by rapid growth and invasion into surrounding brain tissue (18), glioma is a devastating brain tumor (19). With limited treatment options and high recurrence rates, glioma poses significant challenges to patients, caregivers and health care professionals alike. Ongoing research and advancements in the field of neuro-oncology will offer hope for improved management and outcomes in the battle against this formidable malignancy.

Materials and methods

Data origination

The TCGA-glioma data were downloaded from the Cancer Genome Atlas database (TCGA; http://www.cancer.gov/ccg/research/genome-sequencing/tcga). GSE111260 (20) and GSE16011 (21) series profiles from the Gene Expression Omnibus database (GEO; http://www.ncbi.nlm.nih.gov/geo/). From TCGA, gene expression data from 1,097 glioma samples and 5 adjacent normal samples was obtained. GSE111260 comprises 67 glioma samples and 3 control samples, whereas GSE16011 includes 276 glioma samples of all histologies and 8 control samples. In each dataset, differentially expressed genes (DEGs) were subsequently screened using the R ‘limma’ package (version 3.44.1; The R Foundation) with the following screening requirements: False discovery rate (FDR) <0.05 and |log2-fold change (FC)|>1, where log2FC <-1 indicates downregulation and log2FC >1 indicates upregulation. The ‘ggplot2’ package (version 3.3.5) in R (version 4.0.2; The R Foundation) was used to design the volcano maps (22), and the overlapping DEGs were checked and displayed using Venn diagrams (http://bioinformatics.psb.ugent.be/webtools/Venn/).

Functional analyses and protein-protein interaction (PPI) network construction

Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses were performed using the R software packages clusterProfiler (version 4.12.6), enrichplot (version 1.24.4) and ggplot2 (The R Foundation; FDR <0.05). Biological process, molecular function and cellular component are the three separate branches of GO. The PPI network was then created using the CytoHubba plug-in (version 0.1; http://apps.cytoscape.org/apps/cytohubba) and the Search Tool for the Retrieval of Interacting Genes (version 12.0; http://string-db.org/) in the Cytoscape program (version 3.10.2; http://cytoscape.org/). Subnetworks of the overlapping DEGs were generated using the Molecular Complex Detection (MCODE; version 2.0.3; http://apps.cytoscape.org/apps/mcode) 1 and 2 algorithms.

Development and validation of the prognostic signature model

In the univariate Cox proportional hazard regression analysis, the R package c060 (version 0.2–4; The R Foundation) was used, and the stability selection approach was used to further restrict the scope. Using the R package glmnet (version 2.0–16; The R Foundation) and the genes chosen in earlier rounds, a least absolute shrinkage and selection operator (LASSO) Cox model was used to construct a prognostic model. To establish the ideal LASSO penalty parameter value, a 10-fold cross-validation was performed. Patients with gliomas in the TCGA cohort were separated into high- and low-risk groups according to the median risk score. The survival status of patients in the two groups is presented in scatter plots. A heatmap was created using ‘pheatmap’ software (version 1.0.12; http://rdrr.io/cran/pheatmap/) to display the differential expression of hallmark genes between groups. The best risk score cut-off was assessed and a Kaplan-Meier overall survival (OS) curve was produced. Finally, for thorough analyses using receiver operating characteristic curve analysis, the area under the curve (AUC) values of the 1st, 3rd and 5th years were computed.

Gene Set Cancer Analysis (GSCA)

The present study assessed the changes in the expression of 16 prognostic signature genes, namely, single nucleotide variants (SNVs) and copy number variations (CNVs), in lower-grade gliomas (LGGs) and GBMs using GSCA (23). The top 10 mutated genes were chosen for further study, and their waterfall plots displayed detailed information.

Development of a prognostic nomogram

The prognostic importance of the top 10 mutated genes, grade, age and sex were evaluated for their prognostic importance using univariate and multivariate regression analysis. Subsequently, using important factors (CENPF, KIF20A, KIF4A, MKI67 and age), a predictive nomogram was created. The performance of the nomogram in predicting the 1-, 3- and 5-year OS times of patients with glioma was assessed using a calibration chart.

Tumor Immune Estimation Resource (TIMER)

The TIMER (https://cistrome.shinyapps.io/timer/) (24) provides a systematic study of the prevalence of immune infiltrates in a range of cancers. The TIMER scores were used to evaluate the relationships between immune cells (CD4+ T cells, B cells, CD8+ T cells, macrophages, neutrophils and myeloid dendritic cells) and the expression of prognostic hub genes (25). These investigations led to the identification of crucial genes.

The Human Protein Atlas (HPA)

The HPA (26) uses a combination of omics technologies to demonstrate every human protein. The present study detected the protein level of CENPF using the HPA and the mRNA level using Gene Expression Profiling Interactive Analysis (GEPIA; http://gepia.cancer-pku.cn/) in the glioma and normal groups. In addition, the effects of CENPF expression on primary and recurrent glioma, OS and disease-specific survival (DSS) status in patients with glioma were assessed. A log-rank test was used to assess the effect of gene expression on survival.

Cell culture and transfection

Human U87 MG cells (U87; glioblastoma of unknown origin) were purchased from the American Type Culture Collection (ATCC; cat. no. HTB-14). Human U251 MG cells (U251) were purchased from Shanghai Anwei Biotechnology Co., Ltd. (cat. no. AW-CELLS-H0379). The cells used in the present study were subjected to short tandem repeat analysis. In RPMI-1640 media (cat. no. PM150110; Procell Life Science & Technology Co., Ltd.) supplemented with 10% FBS (cat. no. 164210; Procell Life Science & Technology Co., Ltd.), the human glioma cell lines U87 and U251 were grown and incubated at 37°C with 5% CO2.

Small interfering (si)RNAs targeting CENPF (si-CENPF-1, 2 and 3) were generated and synthesized by Hanbio Biotechnology Co., Ltd. The siRNA sequences used were as follows: si-CENPF-1, 5′-GCGCAGAAUCAAGAGCUAA-3′; si-CENPF-2, 5′-CCCAAGAGAAUGGGACUCUUA-3′; si-CENPF-3, 5′-GCGAGUCAGAUCAAGGAGAAU-3′; and si-negative control (si-NC), 5′-UUCUCCGAACGUGUCACGUTT-3′. Lipofectamine™ 2000 (Invitrogen™; Thermo Fisher Scientific, Inc.) was used to transfect these nucleotides into U87 and U251 cells following the manufacturer's instructions. U87 and U251 cells were divided into three groups: si-NC (negative control), si-CENPF-1, si-CENPF-2, and si-CENPF-3. According to the grouping, each well was transfected with 1 µg of the corresponding siRNA and incubated for 5 h at 37°C for transfection. Subsequently, U87 and U251 cells were incubated for an additional 48 h at 37°C before further experimentation.

Reverse transcription-quantitative (q)PCR

Using TRIzol® reagent (Invitrogen; Thermo Fisher Scientific, Inc.), total RNA was extracted from the U87 or U251 cells as directed by the manufacturer. The GoScript™ Reverse Transcription (RT-PCR) Kit (cat. no. A2790; Promega Coporation) was utilized to transcribe 2 µg total RNA into complementary DNA (cDNA). The reaction parameters were as follows: Iincubation at 37°C for 10 min, followed by 42°C for 45 min and then 70°C for 5 min, after which the mixture was cooled on ice for 5 min. Subsequently, the following components were added: 4 µl of GoScript™ 5X reaction buffer, 1.7 µl MgCl2 (final concentration of 2 mM), 1 µl 0.5 mM dNTPs, 0.3 µl ribonuclease inhibitor (20 units), 1 µl reverse transcriptase and distilled deionized water (ddH2O) to achieve a final volume of 15 µl. After thorough mixing, the samples were incubated at 42°C for 60 min, followed by inactivation at 70°C for 15 min. cDNA was subjected to qPCR using ChamQ Universal SYBR qPCR Master Mix (cat. no. Q711-02; Vazyme Biotech Co., Ltd.). The thermocycling conditions were as follows: An initial denaturation at 95°C for 2 min, followed by 40 cycles of denaturation at 95°C for 30 sec, annealing at 95°C for 10 sec and a final extension at 60°C for 30 sec. GAPDH was used as the internal control. The 2−ΔΔCq approach was used to assess the relative fold changes in expression (27). The sequences of the primers used included: CENPF forward, 5′-AAAGAAACAGACGGAACAACTG-3′ and reverse, 5′-CCAAGCAAAGACCGAGAACT-3′; and GAPDH forward, 5′-TGAAGGTCGGAGTCAACGGATTTGG-3′ and reverse, 5′-GGAGGCCATGTGGGCCATGAG-3′.

Western blotting (WB)

RIPA buffer containing 1 mM PMSF was used to lyse the total protein of the cells (Beyotime Institute of Biotechnology). A Pierce™ BCA protein assay kit (Thermo Fisher Scientific, Inc.) was used to assess the protein concentration. Protein samples (30 µg) were separated using 10% SDS-PAGE and transferred to a polyvinylidene difluoride membrane. Following an overnight incubation at 4°C with diluted primary antibodies, the membrane was blocked indoors for 1 h in 5% nonfat milk at room temperature. The primary antibodies used for WB were as follows: anti-CENPF (1:1,000; cat. no. Ab5; Abcam), anti-p21 (1:1,000; cat. no. ab109520; Abcam), anti-CDK1 (1:1,000; cat. no. ab265590; Abcam), anti-vimentin (1:1,000; cat. no. ab92547; Abcam), anti-E-cadherin (1:10,000; cat. no. ab40772; Abcam) and anti-GAPDH (1:8,000; cat. no. ab128915; Abcam) antibodies. Horseradish peroxidase-labelled secondary antibodies (1:5,000; cat. no. ab205718; Abcam) were then applied to the membrane for 1 h indoors at room temperature. The ChemiDoc™ Touch Imaging System (Bio-Rad Laboratories, Inc.) was used to capture the signal after it had been visualized using ECL reagent (cat. no. KGC4902; Nanjing KeyGen Biotech Co., Ltd.).

Cell proliferation and colony formation assays

In 96-well plates with 100 µl culture media, 1,000 U87 or U251 cells per well were cultured. Cell Counting Kit-8 (CCK-8) reagent (10 µl; Beyotime Institute of Biotechnology) was added to each well and incubated for 2 h. The colorimetric absorbance at 450 nm was measured using an enzyme marker (Molecular Devices, LLC).

A total of 1.5×103 treated U87 or U251 cells were plated three times onto 6-well plates for the colony formation assay. The inoculated cells were cultured for another 14 days at 37°C with medium renewal every 3 days. Subsequently, the U87 and U251 cells were washed with PBS and fixed at room temperature with 1 ml of 4% paraformaldehyde (cat. no. P0099; Beyotime Institute of Biotechnology) to a final concentration of 2% for 15 min. The U87 or U251 cells were then washed again with PBS. The formed colonies were subsequently stained with 0.5% crystal violet (cat. no. C0121; Beyotime Institute of Biotechnology) for 5 min at room temperature. The number of cell colonies was quantified using ImageJ software (version 3.0; National Institutes of Health), and the colony formation rate was calculated. Images were captured under a light microscope (Olympus, Tokyo, Japan).

Transwell migration and invasion assays

Migration and invasion assays were performed using well plates with an 8-µm pore size filter insert (Corning, Inc.) with or without diluted Matrigel (precoated for 1 h at 37°C; Becton Dickinson and Company). The upper compartment was filled with U87 or U251 cells (5×104/well) in medium without serum, and the lower chamber contained RPMI-1640 medium supplemented with 10% FBS. The cells were incubated at 37°C for 48 h before being immobilized and stained with DAPI (Beyotime Institute of Biotechnology) for 10 min at room temperature. Cells in the lower chamber were subsequently counted in five arbitrary regions using a light microscope.

Cell cycle assay

U87 or U251 cells were harvested using 0.05% trypsin (MilliporeSigma) for digestion and washed with pre-cooled PBS. The treated U87 or U251 cells (1×106) were collected and fixed with 75% ethanol at −20°C for 3 h. The cells were washed twice with PBS after the ethanol was removed and then resuspended in 1 ml DNA staining solution and 10 µl permeabilization solution (cat. no. CCS012; Multi Sciences Biotech Co., Ltd.) in the dark for 30 min at room temperature. A CytoFLEX S flow cytometer (Beckman Coulter, Inc.) was used for analysis using the FACS LSR II system (BD Biosciences).

Establishment of animal models

The Ethical Committee of the Second Affiliated Hospital of Anhui Medical University (Hefei, China) approved the animal experiments in the present study (approval no. LLSC20230730). The experiments were performed in accordance with the Guide for the Care and Use of Laboratory Animals (GB/T 35892-2018; Standardization Administration of the People's Republic of China) (28). U251 cells transfected with si-NC or si-CENPF were harvested at a concentration of 1×107 cells/ml. A total of 12 male BALB/c nude mice (4–5 weeks old; 15–22 g) were purchased from Beijing Vital River Laboratory Animal Technology Co., Ltd. Mice were housed under controlled environmental conditions (temperature, 22±2°C; humidity, 55±10%; 12-h light/dark cycle) and had free access to standard laboratory food and water. Prior to the start of the experiment, the animals were acclimatized to the laboratory conditions for 6 days to minimize physiological changes related to stress. The mice were randomly divided into two groups (n=6 in each group): the NC group and the si-CENPF (injected with CENPF knockdown) group. The transfected U251 cells (1×106) were first injected subcutaneously into the posterior flanks of the mice. A digital caliper was used once a week to measure the tumor diameters. After 28 days, the mice were euthanized by cervical dislocation following 4% isoflurane anesthesia. The tumor xenografts were excised, imaged and weighed (g). The following humane endpoints were established: Tumor diameter, >2.0 cm; weight loss, >20%; and overall poor condition. In the present study, no mouse reached the humane endpoint.

Statistical analysis

The results from ≥3 separate tests are presented as the mean ± standard deviation. For statistical analysis, IBM SPSS Statistics for Windows, version 17.0 statistical software (IBM Corp.) was used. Significant differences between groups were evaluated using the unpaired Student's t test. One-way ANOVA followed by Tukey's post hoc test was used for multiple comparisons. The Wilcoxon rank-sum test was employed to analyze the mRNA levels of CENPF between glioma and normal tissue samples. Spearman's correlation analysis was used to assess the correlation between CENPF, the G2M checkpoint and EMT markers in glioma. P<0.05 was considered to indicate a statistically significant difference.

Results

Identification of DEGs from the TCGA glioma, GSE111260 and GSE16011 datasets

After a series of bioinformatics analyses, 5,944 up- and 919 downregulated DEGs were identified from TCGA glioma samples, 601 up- and 970 downregulated DEGs from GSE111260 and 1,203 up- and 1,311 downregulated DEGs from GSE16011 (Fig. 1A-C). Subsequently, using a Venn diagram, 273 overlapping up- and 213 overlapping downregulated DEGs were identified (Fig. 1D and E) and analyzed using functional analyses. According to the GO terms, the overlapping DEGs were enriched in the terms ‘mitotic spindle organization’ and ‘neurotransmitter transport’ (Fig. 2A). The enriched KEGG pathways were related to pathways such as oocyte meiosis, the synaptic vesicle cycle and the cell cycle (Fig. 2B). In addition, the PPI network of the overlapping genes revealed the interactions between genes (Fig. 2C). The genes in the subnetworks (Fig. 2D and E) generated by MCODE 1 (23 nodes) and 2 (10 nodes) were chosen for further study.

Figure 1. Identification of DEGs. DEGs identified from (A) TCGA-glioma, (B) GSE111260 and (C) GSE16011 samples (orange indicates upregulation and purple indicates downregulation). (D) Overlapping upregulated DEGs from TCGA-glioma, GSE111260 and GSE16011. (E) Overlapping downregulated DEGs from TCGA-glioma, GSE111260 and GSE16011. DEGs, differentially expressed genes; TCGA, The Cancer Genome Atlas.

Figure 2. Functional enrichment analysis and PPI networks of overlapping DEGs. Enriched (A) GO terms and (B) KEGG pathways of overlapping DEGs. (C) PPI network of overlapping DEGs. The red dots are genes in the subnetworks generated by MCODE 1 and 2 algorithms. Subnetwork of genes analyzed by MCODE (D) 1 and (E) 2 algorithms. PPI, protein-protein interaction; DEGs, differentially expressed genes; GO, Gene Ontology; KEGG, Kyoto Encyclopedia of Genes and Genomes; MCODE, Molecular Complex Detection.

Screening of 16 prognostic signature genes

On the basis of the 33 genes identified by the MCODE algorithm, LASSO-Cox analysis was used to select the optimal threshold parameter (λ=16) for the risk score model (Fig. 3A and B). Patients with glioma were subsequently classified into high (n=332) and low (n=333) risk score groups (Fig. 3C), with the number of deaths increasing from low to high risk scores. Moreover, 16 prognostic signature genes were obtained in which the expression increased from low to high risk scores. The OS probability was markedly greater in the low-risk group compared with that in the high-risk group (Fig. 3D), and the model had the best performance in year 3 comprare with that in year 1 and 5 (Fig. 3E).

Figure 3. Construction of prognostic risk score model for signature genes. (A) Plots of the 10-fold cross-validation error rates. (B) Least absolute shrinkage and selection operator-Cox regression analysis was performed of 33 glioma-related genes identified by Molecular Complex Detection algorithms. The optimal threshold parameter (λ=16) was chosen for constructing the risk score model. (C) Patients with glioma were stratified into high-risk (n=332) and low-risk (n=333) score groups based on their individual risk scores. The distribution of risk score in patients (top), the survival status of high- and low-risk groups (middle) and the expression heatmap of 16 signature genes (bottom) were analyzed. (D) Overall survival probability between the high-risk and low-risk score groups. (E) Receiver operating characteristic curves indicate that the model had the best predictive performance in the year 3. HR, hazard ratio; CI, confidence interval; AUC, area under the curve.

Top 10 mutated genes identified by GSCA

To identify the most frequently mutated genes, the online database GSCA was first searched for genes with CNV and SNV in LGG and GBM, respectively. MKI67, centrosomal protein 55 (CEP55), kinesin family member 11 (KIF11) and DLG-associated protein 5 (DLGAP5) were the top mutated genes according to CNV % analysis, especially in GBM (Fig. 4A). KIF20A, KIF4A, BUB1, CENPF and MKI67 demonstrated a notable mutation frequency in both LGG and GBM, according to SNV % analysis (Fig. 4B). Furthermore, missense mutations were demonstrated to account for the vast majority of gene mutations in glioma, with most mutations occurring in SNPs (Fig. 4C). In glioma, point mutations frequently convert base C to base T and base C to base A (Fig. 4C). Additionally, the top 10 genes with the greatest mutation rate were identified (Fig. 4D), and different mutations in these genes were revealed in 48/52 glioma samples, accounting for 92.31% of the total. The distribution and locations of missense mutations in the top 10 mutated genes in GBM and LGG were also identified (Fig. S1).

Figure 4. Top 10 mutated genes identified by Gene Set Cancer Analysis. (A) CNV % and (B) SNV % heatmap of 16 signature genes in LGG and GBM. (C) Genomic alteration of 16 signature genes in glioma. (D) CNV of top 10 mutated genes in LGG and GBM. CNV, copy number variation; LGG, lower-grade gliomas; GBM, glioblastoma multiforme; SNV, single nucleotide variant.

Prognostic hub genes screened via the nomogram

Combining the mutated genes and clinical factors, univariate and multivariate Cox analyses were performed to screen individual variables, including CENPF, KIF20A, KIF4A, MKI67 and age (Fig. 5A and B). Subsequently, a prognostic nomogram was constructed using the aforementioned variables (Fig. 5C) and its predictive ability was the greatest at year 1 (Fig. 5D). These findings indicate that CENPF, KIF20A, KIF4A and MKI67 could be promising biomarkers for glioma prognosis.

Figure 5. A total of four prognostic hub genes were screened in a nomogram. (A) Univariate and (B) multivariate Cox regression analysis was performed to identify significant variables in glioma. (C) A prognostic nomogram was constructed, integrating the significant variables (CENPF, KIF20A, KIF4A, MKI67 and age). (D) Predictive ability of the prognostic nomogram was assessed by a 45° diagonal line to predict the 1-, 3- and 5-year survival status. CENPF, centromere protein F; KIF20A, kinesin superfamily protein 20A; KIF4A, kinesin superfamily protein 4A; MKI67, marker of proliferation Ki-67; CI, confidence interval.

Immunoassay of prognostic hub genes and immune cells

The TIMER scores of immune cells in the high- and low-expression groups were assessed for CENPF, KIF20A, KIF4A and MKI67. The results revealed that all immune cells had high TIMER scores in the high-score groups, and myeloid dendritic cells had the highest scores in each boxplot (Fig. 6). Notably, the difference in TIMER scores between the high- and low-CENPF expression groups was more pronounced compared with those in the KIF20A, KIF4A and MKI67 gene groups. The aforementioned results indicate that CENPF is a key gene.

Figure 6. Association between prognostic hub genes and immune cells. TIMER scores of immune cells in high- and low- (A) CENPF, (B) KIF4A, (C) KIF20A and (D) MKI67 groups. **P<0.01; ***P<0.001. TIMER, Tumor Immune Estimation Resource; CENPF, centromere protein F; KIF4A, kinesin superfamily protein 4A; KIF20A, kinesin superfamily protein 20A; MKI67, marker of proliferation Ki-67.

Expression and survival analysis of CENPF in glioma

By using the HPA and GEPIA databases, it was demonstrated that the protein and mRNA levels of CENPF were both markedly higher in the glioma group than in the normal group (Fig. 7A and B). Moreover, low CENPF expression was significantly associated with an improved survival probability compared with high CENPF expression, particularly in primary glioma (Fig. 7C and D). Similarly, low CENPF expression was significantly associated with improved OS and DSS probabilities compared with high CENPF expression(Fig. 7E and F). These findings indicate that CENPF may be an oncogene in glioma.

Figure 7. Expression and survival analysis of CENPF in glioma. (A) Protein level of CENPF in glioma and normal tissue from the Human Protein Atlas. (B) mRNA level of CENPF in glioma and normal tissue from Gene Expression Profiling Interactive Analysis. Effect of CENPF expression on the survival probability in (C) primary and (D) recurrent glioma, and on the (E) OS and (F) DSS survival probability in glioma. **P<0.01. CENPF, centromere protein F; OS, overall survival; DSS, disease-specific survival; HR, hazard ratio; CI, confidence interval; WHO, World Health Organization.

CENPF knockdown inhibits glioma proliferation and metastasis and induces G2 arrest in vitro

CENPF was subsequently knocked down in U87 and U251 cells, and the results of PCR and WB revealed that si-CENPF-3 had the greatest knockdown efficiency (Fig. 8A and B). By performing CCK-8 (Fig. 8C and D), colony formation (Fig. 8E) and Transwell (Fig. 8F and G) assays, it was demonstrated that CENPF knockdown significantly inhibited the proliferation, invasion and migration of glioma cells, in comparison with controls. Moreover, it was demonstrated that CENPF was positively correlated with the G2M checkpoint (Fig. 9A). A cell cycle assay revealed that the glioma cells in the G2 phase markedly increased in the si-CENPF-3 group (Fig. 9B), and si-CENPF-3 was significantly associated with reduced p21 and CDK1 levels, in comparison with controls (Fig. 9C and D). Therefore, the results indicate that CENPF knockdown could induce G2 arrest in glioma.

Figure 8. Effects of CENPF knockdown on glioma proliferation and metastasis in vitro. (A) Detected by PCR, si-CENPF-3 demonstrated the best knockdown efficiency in glioma cell lines. (B) Western blot results confirmed the efficient reduction of CENPF protein levels in U87 and U251 cells after si-CENPF-3 treatment. Effect of si-CENPF-3 on cell proliferation was assessed using the Cell Counting Kit-8 assay in (C) U87 and (D) U251 cells. (E) Colony formation assay of glioma cell lines with si-CENPF-3. CENPF knockdown suppressed the (F) invasion and (G) migration of glioma cells. *P<0.05. CENPF, centromere protein F; si, small interfering; NC, negative control; OD, optical density.

Figure 9. CENPF knockdown induces G2 arrest in the cell cycle of glioma cells. (A) Positive correlation between CENPF expression and G2M checkpoint in glioma cells. (B) Flow cytometry analysis demonstrated that CENPF knockdown using si-CENPF-3 increased the proportion of cells in the G2 phase in both U87 and U251 cells. (C) Protein levels of p21 and CDK1 in U87 and U251 cells following si-CENPF-3 treatment as detected by western blotting. (D) Reduced protein levels of p21 and CDK1 in U87 and U251 cells following si-CENPF-3 treatment, as shown by western blotting semi-quantitative analysis. ****P<0.0001. CENPF, centromere protein F; TPM, transcripts per million; si, small interfering; NC, negative control.

CENPF suppresses the progression of glioma by regulating the EMT pathway

Spearman's correlation analysis demonstrated that CENPF was positively correlated with EMT markers (Fig. 10A), indicating that the EMT pathway may be involved in the mechanism of CENPF in glioma. Furthermore, the PCR and WB assays revealed significantly decreased Vimentin and elevated E-cadherin in both U87 and U251 cells of the si-CENPF group compared with those of the si-NC group (Fig. 10B-E). Therefore, the findings indicate that CENPF suppressed the progression of glioma by regulating the EMT pathway.

Figure 10. CENPF regulates the EMT pathway in glioma cells. (A) Spearman's correlation analysis revealed a positive relationship between CENPF expression and EMT markers in glioma cells. mRNA expression levels of EMT markers (Vimentin and E-cadherin) were analyzed by PCR in (B) U87 and (C) U251 glioma cells with si-CENPF-3. (D) Protein levels of Vimentin and E-cadherin were assessed by western blot in U87 and U251 glioma cells with si-CENPF-3. (E) Western blotting semi-quantitative analysis of Vimentin and E-cadherin protein levels in U87 and U251 glioma cells with si-CENPF-3. *P<0.05; ****P<0.0001. CENPF, centromere protein F; EMT, epithelial-mesenchymal transition; si, small interfering; NC, negative control.

CENPF knockdown blocks the tumorigenesis of glioma in vivo

In the constructed animal models, tumor tissues were collected and measured. The sizes of the tumors in the si-CENPF group were notably smaller than those in the si-NC group (Fig. 11A). In addition, the tumor mass in the si-CENPF group was significantly lower than that in the si-NC group (Fig. 11B). These findings confirm that CENPF knockdown suppressed the tumorigenesis of glioma in mice.

Figure 11. CENPF knockdown blocks the tumorigenesis of glioma in vivo. (A) Comparison of tumor size and (B) mass in si-NC and si-CENPF groups. ***P<0.001. CENPF, centromere protein F; si, small interfering; NC, negative control.

Discussion

Currently, the diagnosis of glioma typically involves imaging techniques such as MRI or CT (29), followed by a biopsy to confirm the presence of malignant glial cells. Standard treatment for glioma (30) usually involves a combination of surgical resection, radiation therapy and chemotherapy (31). However, this type of brain tumor tends to develop quickly and infiltrate surrounding tissues, essentially prohibiting thorough surgical removal. The rapid and invasive growth of glioma is attributed to its highly proliferative glial cells (32), which are characterized by an increased capacity for angiogenesis and resistance to apoptosis (33). Therefore, early detection, personalized treatment strategies and innovative therapeutic interventions hold the key to enhancing the overall survival and quality of life of patients with glioma.

In the present study, glioma DEGs were identified using data from the TCGA, GSE111260 and GSE16011 datasets. The overlapping DEGs obtained from these datasets were found to be significantly enriched in spindle, Microtubule motor activity, Cell cycle and Synaptic vesicle cycle. Through further application of the MCODE 1 and 2 algorithms, 33 genes associated with prognosis were identified, from which 16 prognostic signature genes were selected in the risk score model for gene mutation analysis. Subsequently, the top 10 mutated genes was used to construct a prognostic nomogram. Finally, four key prognostic hub genes were identified, namely, CENPF, KIF20A, KIF4A and MKI67. These genes exhibited the potential to serve as valuable prognostic biomarkers in glioma.

Members of the kinesin superfamily of motor proteins include KIF4A and KIF20A. KIF4A functions as a motor protein that is based on microtubules and is related to the organization and dynamics of the mitotic spindle (34), which are essential for proper cell division and genomic stability (35). Hou et al (36) demonstrated that KIF4A enhances cell proliferation and promotes colorectal cancer development by promoting cell cycle progression both in vitro and in vivo. Additionally, Hou et al (37) reported that KIF4A overexpression enhances the proliferation and migration of hepatocellular carcinoma cells, whereas KIF4A knockdown reduces cell proliferation and migration, suggesting a potential role for KIF4A in mediating tumorigenesis and progression. Jin and Ye (14) reported that KIF4A regulates the expression of BUB1, inhibiting apoptosis and promoting ovarian cancer progression. Zhang et al (38) reported that Rac1/Cdc42 transcriptional suppression by KIF4A, which causes cytoskeletal reorganization in glioma cells, promotes the formation of gliomas. KIF20A is involved primarily in regulating microtubule dynamics during cell division (39) and is essential for proper cytokinesis and cell cycle progression (40). Previous research linked the overexpression of KIF20A with several cancers. KIF20A was first discovered to be overexpressed in pancreatic ductal adenocarcinoma (PDAC), and its knockdown in PDAC cell lines severely inhibited cell growth (41). Further studies in human liver cancer cell lines have also reported elevated levels of KIF20A, whereas it is undetectable in normal human liver cells (42). Yan et al (43) demonstrated that KIF20A RNAi inhibited the viability of gastric cancer (GC) SGC7901 cells. Peptides derived from KIF20A used alone as immunotherapy vaccines or in combination with other peptides/chemotherapy drugs have achieved notably higher OS rates in GC treatment (44,45). Copello and Burnstein (46) reported that KIF20A promotes progression to castration-resistant prostate cancer by activating the androgen receptor via autocrine mechanisms. Research has reported that elevated KIF20A levels are associated with poor prognosis in patients with GBM (47).

In addition, MKI67 is a nuclear protein associated with cell proliferation. Under normal conditions, MKi67 shows cortical nucleolar localization during interphase and is recruited to condensed chromosomes during mitosis (48). The MKi67 gene is located on chromosome 10q25-ter and primarily encodes two MKi67 isoforms (345 and 395 kDa) (49,50). The expression of the MKi67 protein can be assessed in the nuclei of cells in the G1, S, G2 and mitotic phases but not in the quiescent G0 phase (51). The expression of MKI67 is widely used as a biomarker for assessing cell proliferation in several types of cancers (52), including glioma (53). High expression of Ki67 in cancer cells can be considered a prognostic predictor for cancer (54). Substantial evidence supports the role of MKi67 in cancer diagnosis (55). In a study involving patients with liver hepatocellular carcinoma (LIHC) who underwent surgery, high MKi67 expression in cancer tissues was reported to predict increased tumor grade and early cancer recurrence (56). Moreover, MKi67 staining has been widely used to predict postoperative survival rates and even survival rates after liver transplantation in patients with LIHC (57).

CENPF is a human gene that encodes the centromere protein F, a crucial component of the kinetochore complex. It has been extensively studied in several fields, including cell biology, cancer research and genomics, and is associated with the prognosis of patients with non-small cell lung cancer and prostate cancer (58,59). Additionally, overexpression of CENPF is associated with poor prognosis and bone metastasis in patients with breast cancer (60). Moreover, in hepatocellular carcinoma, high CENPF levels are associated with poor prognosis and aggressive tumor behavior (61). Moreover, one study analyzed genomic data from patients with glioma and identified CENPF as one of the notably amplified genes in tumor samples (62). However, to fully understand the molecular processes of CENPF in glioma, further study is necessary.

Through immunoassays targeting hub genes, CENPF emerged as a key gene in the present study, and its significant association with glioma prognosis was revealed. High CENPF expression was associated with the poor prognosis of patients with glioma. In in vitro and in vivo assays, CENPF was observed to be upregulated in glioma, and its knockdown inhibited glioma progression and metastasis, as demonstrated by data from public databases. Collectively, these findings establish CENPF as an oncogene in glioma. Furthermore, given the close relationship of CENPF with the cell cycle, flow cytometry was used to assess its impact on cell cycle regulation in glioma. The results demonstrated that CENPF knockdown led to G2 arrest in the cell cycle.

EMT is a complex biological process (63). In glioma, EMT promotes cell invasion into surrounding brain tissue and facilitates their ability to migrate to distant sites, contributing to tumor spread and metastasis (64). As the EMT pathway may be involved in the mechanism of CENPF in glioma, the present study assessed the levels of EMT markers in glioma cells with CENPF knockdown. The results revealed decreased vimentin and elevated E-cadherin levels, suggesting that CENPF promotes glioma development by regulating the EMT pathway.

The present study has several limitations. Whilst in vitro and in vivo experiments were performed to assess the role of the CENPF gene in glioma, additional experiments are needed to confirm its mechanisms. The present study also used multiple public databases and mouse models to analyze the role of CENPF in glioma but lacked validation in clinical samples. In future studies, more clinical data should be collected, including clinical samples, clinical characteristics and survival data to perform further validation of the expression level and prognostic value of CENPF in glioma clinical samples.

In conclusion, the present study assessed the molecular landscape and potential prognostic biomarkers in glioma, a highly aggressive and lethal brain tumor. Through joint analysis of the TCGA, GSE111260 and GSE16011 datasets, 486 genes associated with glioma were identified. Comprehensive bioinformatics analyses, including PPI networks, risk score models, gene mutation analyses and diagnostic models, revealed potential prognostic biomarkers for glioma (CENPF, KIF20A, KIF4A and MKI67). CENPF was significantly upregulated in glioma and was associated with poor patient prognosis. In vitro functional experiments demonstrated that CENPF promotes the proliferation and metastasis of glioma cells, promoting glioma progression through the regulation of the EMT pathway. In vivo experiments indicated that downregulation of CENPF expression inhibits tumor progression in glioma. Overall, the present study contributes to the understanding of glioma biology and provides a basis for further investigations and the development of personalized approaches for glioma diagnosis, treatment and prognosis.

Supplementary Material

Supporting Data

Acknowledgements

Not applicable.

Funding

Funding: No funding was received.

Availability of data and materials

The data generated in the present study may be requested from the corresponding author.

Authors' contributions

JL, LL, GZ, ZY, DZ and BZ contributed to the study conception and design. Material preparation, data collection and analysis were performed by JL, LL, GZ and ZY. The first draft of the manuscript was written by DZ and BZ, and all authors commented on previous versions of the manuscript. JL and BZ confirm the authenticity of all the raw data. All authors have read and approved the final version of the manuscript.

Ethics approval and consent to participate

The animal experiments in the present study were approved by The Ethical Committee of the Second Affiliated Hospital of Anhui Medical University (Hefei, China; approval no. LLSC20230730}.

Patient consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.

References