Next-generation sequencing data (HTSeq-FPKM) and clinical variables were obtained from TCGA (Wang et al., 2016). In addition, data for 1,018 glioma cases generated with the Illumina HiSeq platform were obtained from the CGGA (Zhao et al., 2021). Apart from that, a total of 112 glioma cases enrolled from the Department of Neurosurgery (Huanhu Hospital, Tianjin, China) were used for validation.

In TCGA, the differential expression of CENPN was analyzed using GEPIA, a web database containing abundant normal specimens from GTEx database (Tang et al., 2019). Additionally, GSE16011 (Gravendeel et al., 2009), a dataset from the Gene Expression Omnibus database (GEO), was used for validation. The relationships between CENPN and clinical variables were explored using the beeswarm R package. The survival, survminer and ROC packages were used to plot receiver operating characteristic (ROC) curves and Kaplan–Meier curves.

GSEA was conducted based on the well-established gene sets of c2.cp.kegg.v7.2.symbols.gmt and c5.all.v7.2.symbols.gmt in GSEA 4.0 software. ssGSEA was used to transform the 29 immune signatures into scores for each glioma sample (Subramanian et al., 2005; Bindea et al., 2013; Zhou et al., 2020). Additionally, the CIBERSORT deconvolution algorithm was performed to evaluate the immune infiltration of 22 tumor-infiltrating immune cells (TILs) (Newman et al., 2015). The TIL signatures were downloaded from CIBERSORT (https://cibersortx.stanford.edu/). Relevant R packages were obtained from Bioconductor (http://bioconductor.org/) and CRAN (http://cran.r-project.org/). The bioinformatics analyses were conducted using R software.

The human glioblastoma cell lines LN229 and U251 were purchased from Beijing Beina Chuanglian Biotechnology Institute and cultured in DMEM containing 10% fetal bovine serum (FBS, Gibco, Invitrogen, CA, United States). The siRNAs targeting CENPN mRNA were produced by GenePharma. The sense sequences of these siRNAs were as follows: siCENPN-i, GCG​UGC​AAG​UAU​CAG​UGA​UTT; siCENPN-ii, GAC​CCU​UUG​UUA​CUC​AAA​UTT; and scramble (siScr), UUC​UCC​GAA​CGU​GUC​ACG​UTT. siRNA transfections were performed using Lipofectamine 2000 according to the manufacturer’s instructions (Thermo Fisher Scientific).

TRIzol reagent was used for RNA extraction, and the SureFireRT kit (06-104; Abgen) was applied for RNA reverse transcription. Then, real-time PCR was conducted on a LightCycler 480 II (Roche). The sequences of designed PCR primers were as follows: GAPDH forward, 5′-CAA​TGA​CCC​CTT​CAT​TGA​CC-3′ and reverse, 5′-GAC​AAG​CTT​CCC​GTT​CTC​AG-3′; CENPN forward, 5′- TGG​ATG​AGA​CTG​TTG​CTG​A -3′ and reverse, 5′- ACTTGCACGCTTTTCCTC -3′. The PCR conditions were 95°C for 5  min, followed by 40 cycles at 95°C for 10  s and 60°C for 1 min. Relative quantification was calculated by the 2^−ΔΔCt method.

Cells were lysed using RIPA buffer following the manufacturer’s protocols (Solarbio Co., Beijing, China). Protein (50 μg/lane) was added to the 12% SDS-PAGE gel and then transferred to a membrane (PVDF, Merck Millipore). The protein blots on the membrane blocked by BSA were incubated with antibody against CENPN (1:500 dilution; Affinity Biosciences, Jiangsu, China; DF2315) or β-tubulin (1:500 dilution; CST, United States) overnight. After incubation with HRP-linked goat anti-rabbit IgG (CST, United States) for 60 min, the membrane was exposed to enhanced chemiluminescence reagent.

Immunohistochemistry assays were conducted by the pathology department of Huanhu Hospital. Immunostaining of CENPN was performed using a rabbit polyclonal anti-CENPN antibody (1:500 dilution; Affinity Biosciences, Jiangsu, China; DF2315). The positive slides were recognized by integrated scoring, which was performed independently by two experienced pathologists. The glioma tissues were divided into high and low expression groups based on the expression density of CENPN, as previously reported (Xiao et al., 2015).

5 × 10³ cells per well were seeded in 96-well plates and then transfected with siCENPN. Cells were incubated with CCK-8 reagent (K009-500, ZETA) for 30–60 min at room temperature at 0, 24, 48 and 72 h time points. Eventually, the absorbance changes were observed at 450 nm on a molecular devices microplate reader.

For the wound-healing assay, the bottom of plate was scraped by a pipette tip when 100% cell confluence. Then, wound closure (%) was observed at 0 and 24 h by ImageJ. For the Transwell assay, 5 × 10⁴ cells with or without CENPN knockdown were seeded in the upper chamber (8 µm pore size, Corning, Cambridge, United States) coated with Matrigel (BD Biosciences) while 500 µl of medium containing 10% FBS was placed in the lower chamber. Twenty-four hours later, the upper chamber was immersed in 4% paraformaldehyde for 20 min and then subjected to 20 min crystal violet staining. Eventually, the number of penetrating cells was calculated randomly by three fields of view (FOVs). A colony formation assay was conducted to examine the glioma proliferation ability, with 500 cells per well seeded in six-well plates. Fourteen days later, the cells were fixed with 4% formaldehyde and subjected to crystal violet staining.

Cell apoptosis assay (FITC Annexin V Apoptosis Detection Kit, BD Biosciences, United States) was conducted according to the manufacturer’s protocol, as previously described (Zhou et al., 2021). For the cell cycle assay, glioma cells were harvested 48 h following transfection and then fixed with precooled 75% alcohol for 3 h. Subsequently, the cells were incubated with 0.5 ml of PI/RNase reagent (BD Biosciences, United States) at room temperature for 20 min and then analyzed by flow cytometry.

Cases with incomplete information were removed from the corresponding analysis. Student’s t-test and one‐way analysis of variance (ANOVA) were used to test for significant differences between two groups or multiple groups, respectively. For CENPN protein expression, the chi-square test was employed. Data are presented as the mean ± standard deviation (SD) of at least three independent experiments. p < 0.05 was viewed statistically significant. All statistical tests were two-sided. R software v3.6.2. and GraphPad Prism 8 were used for the statistical analyses.

In the present study, 1833 glioma cases from the TCGA, CGGA and Huanhu datasets were included. The clinical features of these datasets are summarized in Table 1. Pathology reports containing the WHO grade, IDH1 mutation status and other information on the patients in the Huanhu dataset were provided by the Department of Pathology of Huanhu Hospital. The patients in the Huanhu cohort consisted of 44 females and 68 males, and their age ranged from 24 to 82 years old. Detailed information on the patients in the Huanhu cohort is shown in Supplementary Table S1.

The mRNA expression levels of CENPN in glioma were assessed based on GEPIA and GSE16011, a glioma dataset obtained from the Gene Expression Omnibus database. Compared with the values in normal samples, the CENPN mRNA level was markedly higher in the glioma samples (Figures 1A,B).

To explore the CENPN expression profiles in glioma, the clinical features of glioma patients obtained from TCGA and CGGA were analyzed. As demonstrated in Figures 1C,E–G, CENPN was highly expressed in patients older than 41 years and patients with advanced grades, histopathologies and wild-type IDH1 in both TCGA and CGGA, although no association was found between CENPN expression and gender (Figure 1D). In TCGA, CENPN expression increased with higher Karnofsky Performance Status scores (KPS ≥ 80) (Figure 1H). In CGGA, some other clinical variables (radiotherapy, 1p19q codeletion, chemotherapy and recurrent status) were also analyzed (Figure 1I-L). These results revealed that CENPN upregulation could predict unfavorable glioma.

Kaplan-Meier plots were graphed to estimate the prognostic value of CENPN expression. As depicted in Figure 2A, CENPN high expression was associated with worse survival of patients in TCGA. ROC analyses demonstrated that CENPN could be a fine predictor for patient survival: 1 year (AUC = 0.800), 3 years (AUC = 0.807) and 5 years OS (AUC = 0.786) (Figure 2B). These results were verified in CGGA dataset (Figures 2C,D).

Furthermore, univariate and multivariate Cox regression analyses were employed to assess the independent prognostic roles of CENPN. In TCGA, univariate regression showed that patients with CENPN-high had a shorter overall survival time (hazard ratio (HR): 2.56, 95% CI [2.05–3.20], p < 0.001) (Figure 2E). The remaining clinical features associated with OS, such as age, grade and histology, are shown in Figure 2E. The multivariate regression model demonstrated that CENPN was independently correlated with patient OS (Figure 2F, HR = 1.73, 95% CI [1.28–2.33], p < 0.001). Moreover, these results were also confirmed in CGGA dataset (Figures 2G,H), indicating that CENPN could be an independent risk factor for glioma.

The KEGG analysis based on the GSEA of genes in the high CENPN expression group in TCGA showed that some biological pathways were enriched, including the cell cycle, P53 signaling pathway and apoptosis. Moreover, some pathways related to immunity were also enriched, such as leukocyte transendothelial migration, Fc gamma r-mediated phagocytosis and antigen processing and presentation. The GO analysis indicated that immune-related biological processes, such as immunoglobulin production, leukocyte homeostasis and negative phagocytosis regulation, were enriched in the cohort with CENPN-high (Figure 3A). Similar results were also verified in the CGGA database (Figure 3B). GSEA results suggested that CENPN was associated with multiple molecular mechanisms of glioma development.

To further explore the role of CENPN in the immune microenvironment, ssGSEA was applied to assess the enrichment scores of immune-related pathways, immune cells and immune-related functions grounded on the 29 gene sets derived from previous studies (Subramanian et al., 2005; Bindea et al., 2013; Zhou et al., 2020). In TCGA, the heatmap (Figure 3C) shows that some immune-related functions were positively correlated with CENPN expression. APC coinhibition, APC costimulation, chemokine receptor (CCR), cytolytic activity, immune checkpoint, inflammation promoting, human leukocyte antigen (HLA), parainflammation, major histocompatibility complex (MHC) class I, type I interferon (IFN) responses, T cell coinhibition and T cell costimulation were enriched in CENPN-high patients. Additionally, several immune cell types, such as CD8⁺ T cells, macrophages and regulatory T cells (Tregs), were infiltrated in glioma cases with CENPN-high, while neutrophils, dendritic cells (DCs) and natural killer cells (NK cells) showed opposite trends. Meanwhile, similar results were also obtained in CGGA database (Figure 3D). These results indicated that CENPN has a close relationship with immunomodulation of glioma.

To systematically estimate the relationship between CENPN and tumor-infiltrating lymphocytes (TILs), the CIBERSORT algorithm method was employed to infer the fractions of 22 immune cell types in tumor samples based on gene expression profiles. Consistent with the ssGSEA results, CD8⁺ T cells, follicular helper T cells, Tregs, gamma delta T cells, M0, M1 and M2 macrophages in TCGA and follicular helper T cells and M0 macrophages in CGGA were enriched in the CENPN-high group. However, memory resting CD4⁺ T cells, NK cells activated, dendritic cells and mast cells, monocytes and eosinophils in TCGA and CD4 naïve T cells, monocytes, mast cells and NK cells activated in CGGA were enriched in the CENPN-low group (Figure 4). These findings indicated an important role of CENPN in the glioma immune microenvironment.

Our study further evaluated the protein level of CENPN in 112 glioma specimens by IHC. We found that CENPN was upregulated in 72.3% (81/112) of glioma tissues (Supplementary Table S1). Representative IHC images of samples with grades II, III and IV are displayed in Figure 5A. CENPN expression was positively correlated with some adverse outcome clinicopathological variables, such as age, WHO grade and histology (Figure 5B). However, no correlation was found between CENPN expression and 1p19q codeletion or P53 mutation.

To explore the functions of CENPN in glioma, LN229 and U251 cells were transfected with two siRNAs (siCENPN-i and siCENPN-ii) to silence CENPN transcription. The CENPN siRNAs resulted in significant declines in both mRNA and protein levels (Figures 6A,B). The CCK-8 assay showed that transfection with siCENPN significantly reduced the proliferation of the cells (Figure 6C). The colony formation assay displayed a similar result (Figure 6D). Furthermore, wound healing and Transwell assays showed that downregulation of CENPN suppressed the migration and invasion ability of glioma cells compared with that in the siScr group (Figures 6E,F). In addition, we found that higher apoptotic percentages were detected in the siCENPN groups than in the siScr groups (Figure 6G). However, the cell cycle assay showed that there were no differences in the fractions of cells in the G0/G1, G2/M, and S phases after CENPN suppression compared with those subjected to siScr (Supplementary Figure S1). These results suggest that CENPN could promote malignant phenotypes of glioma cells in vitro.