Identification of Driver Genes and Key Pathways of Glioblastoma Shows JNJ-7706621 as a Novel Antiglioblastoma Drug

Sheng Zhong1,2, Bo Wu2, Xuechao Dong1, Yujuan Han2, Shanshan Jiang3, Ying Zhang2, Yang Bai1, Sean X. Luo4, Yong Chen1, Huimao Zhang5, Gang Zhao1


■ OBJECTIVE: The aim of this study is to identify novel targets of diagnosis, therapy, and prognosis for glioblas- toma, as well as to verify the therapeutic effect of JNJ- 7706621 regarding glioblastoma.
■ METHODS: The gene expression profiles of GSE42656, GSE50161, and GSE86574 were obtained respectively from the Gene Expression Omnibus database. The differentially expressed genes (DEGs) were identified with comparison between gene expression profiles of the glioblastoma tis- sues and normal tissues. Gene Ontology (GO), Kyoto Ency- clopedia of Genes and Genomes (KEGG) analysis and proteineprotein interaction (PPI) network analyses were performed. Quantitative reverse transcription polymerase chain reaction and survival curve analysis were also con- ducted to verify the correlation between expression of hub genes and prognosis. Moreover, in vitro, MTT assay, colony- forming assay, the scratch assay, and flow cytometry were performed to verify the therapeutic effect of JNJ-7706621.
■ RESULTS: AURKA, NDC80, KIF4A, and NUSAP1 were identified as hub genes after PPI network analysis. Differential expression of those genes was detected be- tween human normal glial cells and glioblastoma cells by quantitative reverse transcription polymerase chain reac- tion (P < 0.05), and the survival curve analysis showed that the patients with low expression of gene AURKA, NDC80, KIF4A, and NUSAP1 had a significant favorable prognosis (P < 0.05). In vitro assays showed that JNJ-7706621 inhibited glioblastoma cellular viability, proliferation, and migration via inducing glioblastoma cells apoptosis. ■ CONCLUSIONS: AURKA, NDC80, KIF4A, and NUSAP1 were significantly more highly expressed in glioblastoma cells than in human normal glial cell. Patients with low expression of those 4 genes had a favorable prognosis. JNJ-7706621 was a potential drug in treatment of patients with glioblastoma. Key words ■ Bioinformatics ■ Brain science ■ Drug treatment ■ Glioma ■ Prognosis INTRODUCTION lioblastoma multiforme, which comprises 30%e40% of all brain tumors, is one of the most aggressive diffuse neoplasms and the most common malignant tumor in the central nervous system; it is categorized into grade III/IV tumors according to the World Health Organization classifica- tion.1-3 About 3.19 per 100,000 persons are diagnosed with glio- blastoma multiforme annually, and the average age of these patients at diagnosis is 64 years.4 Combined and multiple therapies are generally adopted as the conventional treatment methods, including surgical resection, radiotherapy, and chemotherapy. Maximal surgical resection is recommended first when feasible, and radiotherapy and chemotherapy, such as temozolomide or carmustine wafer, are selected as concomitant and adjuvant therapies. In addition, repeat surgery is sometimes necessary to help control symptoms as well as to verify the diagnosis or to allow for intralesional chemotherapy with temozolomide or carmustin wafer.5 After these treatments, the disease still has a poor prognosis, with a median survival of less than 15 months, 2-year survival of 26%e33%, and 4%e5% survival at 5 years.6 Furthermore, the recurrence rate seems to be high; the diffuse neoplasm tends to recur after only a few months after surgical resection.7 The poor treatment outcome is unsatisfactory, and a precise and effective therapeutic strategy is needed to improve this situation. In recent years, the use of bioinformatics and microarray technology has enabled study of the initiation, progression, and metastasis of glioblastoma at the molecular level, which makes it possible to analyze the genetic alteration and molecular mecha- nisms in the development of glioblastoma.8,9 For instance, IGFBP-2 and CDC20 are reported to help with diagnosis based on the high correlation between the expression level of these 2 genes and glioblastoma.10 However, a lack of studies that show explicit molecular mechanisms of glioblastoma pathogenesis hinders comprehension of glioblastoma. This study aims to identify novel targets of diagnosis, therapy, and prognosis for glioblastoma, as well as to verify the therapeutic effect of JNJ-7706621. In this study, 3 messenger RNA microarray datasets (GSE42656, GSE50161, and GSE86574) involving glioblastoma were down- loaded from Gene Expression Omnibus, and then, those datasets were analyzed to identify differentially expressed genes (DEGs) by comparing gene expression profiles of the glioblastoma tissues and normal brain tissue samples. Subsequently, the mutual DEGs were screened with a Venn analysis, Gene Ontology (GO), and key pathways enrichment analysis were followed to depict the bio- logical process and molecular function of glioblastoma, and proteineprotein interaction (PPI) network analysis was then per- formed. In addition, quantitative reverse transcription polymerase chain reaction (qRT-PCR) and survival curve analysis were also performed to verify the differential expression and the correlation between expression of mutual hub genes and prognosis. Then, MTT assay, colony-forming assay, scratch assay, and flow cytometry were performed in vitro to verify the therapeutic effect of JNJ-7706621.11,12 JNJ-7706621, a dual cyclin-dependent kinase and aurora inhibitor, is a small molecular, fat-soluble, potent antitumor drug, which can easily cross the blood-brain barrier; its therapeutic effect has been verified in breast cancer and cervical carcinoma, for example. Recently, several phase 2 clinical trials regarding JNJ-7706621 have been conducted. JNJ-7706621 is a promising antiglioblastoma drug in future clinical practice. The framework and concise content of this study are shown in Figure 1. METHODS Microarray Data The gene expression profiles of GSE86574, GSE42656, and GSE50161 were obtained from Gene Expression Omnibus database ( Multiple sample sets were used to avoid race and clinical bias among different studies. Those profiles, which in total contain 49 glioblastoma samples and 31 normal samples, were provided on platforms GPL6947 (GSE42656) and GPL570 (GSE50161, GSE86574). GSE42656 included 5 glioblastoma samples and 8 normal samples, GSE50161 provided 34 glioblastoma samples and 13 normal samples, and GSE86574 contained 10 glioblastoma samples and 10 normal samples. Identification of DEGs The analysis was conducted based on raw data using GeneSpring software (version 11.5 [Agilent, Santa Clara, California, USA]) for 3 DEG groups to fit 3 gene expression profiles. The data were categorized with hierarchical clustering analysis, and the group glioblastoma and normal samples were identified. The probe quality control in GeneSpring was limited by virtue of principal component analysis, and probes with intensity values below the 20th percentile were filtered out using the “filter probesets by expression” option. Then, the DEGs were identified using a classic t test with a P value cutoff of <0.05 and a change >2-fold, which were applied for a statistically significant definition. Moreover, Venn plot analysis regarding DEGs was conducted among upre- gulated, downregulated, and total DEGs (http://bioinformatics.

GO and Pathway Enrichment Analysis of DEGs

The DAVID database (Database for Annotation, Visualization and Integrated Discovery, is an essential foundation that provides a comprehensive set of functional annotation tools to understand the biological meaning underlying many genes. GO is a useful method to analyze the biological process, molecular function, and cell component of genes. The Kyoto Encyclopedia of Genes and Genomes (KEGG) is a basis for gene function analysis and genomic information links. In this study, GO and KEGG pathway enrichment analysis were per- formed using DAVID for analysis of DEG functions. Gene Set Enrichment Analysis (GSEA) ( index.jsp) was performed to determine which set of genes showed statistical significance. This procedure was conducted to identify GO and KEGG pathway enrichment.

PPI Network Construction and Modules Selection

STRING (Search Tool for Retrieval of Interacting Genes, http://, an online database, provided PPI analysis for bioinformatic studies. Then, Cytoscape software was applied to screen hub genes and modules with MCODE (Molecular Complex Detection). In addition, function and pathway enrichment analysis of DEGs in modules was performed. This procedure was con- ducted to identify the hub genes and their degrees.

Cell Lines and Reagents

Human normal glial cells (HEB), glioblastoma cell lines (U87, U251, LN18, T98, SHG-44, U373), human umbilical vein endothelial cell (HUVEC), and human hepatocyte (HL7702) were received from the American Type Culture Collection. Those cell lines were cultured in Dulbecco’s modified Eagle’s medium (DMEM [GE Healthcare Life Sciences, HyClone Laboratories, Logan, Utah, USA]) supplemented with 10% fetal bovine serum (Gibco, Thermo Fisher Scientific, Waltham, Massachusetts, USA). An atmosphere of 5% CO2 and 95% air at 37◦C was maintained for cell line cultivation. JNJ-7706621 was purchased from Apexbio Inc., Houston, Texas, USA.

Real-Time qRT-PCR

To confirm the expression of aurora kinase A (AURKA), nuclear division cycle 80 (NDC80), kinesin super-family protein 4A (KIF4A), and nucleolar and spindle-associated protein 1 (NUSAP1) in glioblastoma cell lines and normal human glial cells, we per- formed qRT-PCR using FastStart Universal SYBR Green Master (ROX) (Roche Diagnostics) in a CFX96 Real-Time System (BioRad) according to the manufacturer’s instructions and expression levels were normalized to glyceraldehyde-3-phosphate dehydroge- nase. The 2—OOCt method was used for qRT-PCR data analysis.13 The primers of genes were listed as follows: AURKA sense, 50-GAGGTCCAAAACGTGTTCTCG-30; AURKA anti-sense, 50-ACAGG ATGAGGTACACTGGTTG-30; NDC80 sense, 50-CCTCTCCATG- CAGGAGTTAAGA-30; NDC80 anti-sense, 50-GGTCTCGGG-T CCTTGATTTTCT-30; KIF4A sense, 50-TACTGCGGTGGAGCAA- GAAG-30; KIF4A anti-sense, 50-CATCTGCGCTTGACGGAGAG-30; NUSAP1 sense, 50-AGCCCATCAATAAGGGAGGG-30; NUSAP1 anti- sense, 50-ACCTGACACCCGTTTTAGCTG-30.

Clinical Patients Datasets Used in This Study

The gene expression data of 325 patients (203 males and 122 females), with an average age of 43.38 years, were downloaded from the Chinese Glioma Genome Atlas ( Those patients were categorized into a high-expressed group and low-expressed group according to the expression level of the AURKA, NDC80, KIF4A, and NUSAP1 genes. We regarded progression-free survival (PFS) and overall survival (OS) as the prognostic outcome of patients with glioblastoma.

MTT Assay

The glioblastoma cells (U251, U87, LN18, T98, SHG-44, and U373) and human normal cells (HUVEC and HL7702) were plated into 96-well culture plates with a density of 500 cells/well and were treated with different doses of JNJ-7706621. The MTT reagent (Sigma, St. Louis, Missouri, USA) was dissolved in phosphate- buffered saline (5 mg/mL) to measure the viability of cells. On the day of measurement, medium was replaced on fresh DMEM supplemented with 10% fetal bovine serum and diluted MTT (1:10, 10% MTT), and incubated for 3.5 hours at 37◦C. Then, the incu- bation medium was removed and formazan crystals were dissolved in a 200-mL solution of DMSO. The ELx800 absorbance microplate reader (BioTek Instruments, Winooski, Vermont, USA) was applied to quantify the MTT reduction by measuring the light absorbance at 570 nm. Each test was repeated 4 times.

Colony-Forming Assay

The glioblastoma cells (U251 and LN18) were seeded in Petri dishes with a density of 50 cells/cm2. After 24 hours in culture, those glioblastoma cells were treated with different doses of JNJ- 7706621. After 10 days in vitro growth, colonies were counted and described according to Franken et al.14 Then, colonies were rinsed with phosphate-buffered saline, fixed in 4% para- formaldehyde, stained with 5% crystal violet for half an hour, and rinsed twice with water.

In Vitro Scratch Assay

The glioblastoma cells (U251) were cultured on 24-well Permanox plates. A 1-mL pipette tip across each well was used to create a consistent cell-free area. The loose cells were washed out gently using DMEM. Then, the cells were exposed to different doses of JNJ-770662. After the scratch and at 0, 12, and 24 hours, the im- ages of the scraped area were captured with phase contrast mi- croscopy. The remaining wounded area and the scratch width at 6 different points per image were measured.

Apoptosis Assays

The glioblastoma cells (LN18) in the log growth phase were seeded into 6-well plates with a density of 2 105 cells/well, and the cells were treated with different doses of JNJ-7706621. After culture for 24 hours, cells were harvested using Accutase detach- ment solution (Sigma, St. Louis, Missouri, USA) and Annexin-V- FITC/propidium iodide (PI) labeling was conducted according to the manufacturer’s instructions. The flow cytometer was applied to analyze the stained cells and the cells were calculated with FACSDiva version 6.2.

Cell Cycle Analysis

Cell cycle status was determined by measuring cellular DNA content by PI staining. Cells (LN18) in the log growth phase were seeded at a density of 2 105 cells per well in 6-well plates. After treatment with different doses of JNJ-7706621, cells were harvested and fixed in 70% cold ethanol at e20◦C overnight. The next day, the cells were treated with 100 mg/mL of ribonuclease and incu- bated at 37◦C for 30 minutes. Then, the cells were stained with 100 mg/mL of PI in the dark at room temperature for 30 minutes. The samples were subsequently analyzed with the FACScalibur flow cytometer; data were analyzed using ModFit LT 3.3 software.

Statistical Analysis

All statistic data were entered into SPSS 18.0 (SPSS Inc., Chicago, Illinois, USA) for analysis. An independent-samples t test was conducted to analyze quantitative data. P values <0.05 were set as the significance level. RESULTS Identification of DEGs Altogether, 1744 DEGs were picked up from GSE42656, of which 699 were upregulated and 1045 were downregulated. Among 418 DEGs found from GSE50161, 103 were upregulated and 315 were downregulated. A total of 1088 DEGs were identified from GSE86574, of which 307 were upregulated and 781 were down- regulated. A total of 162 mutual DEGs among those 3 datasets were identified by performing Venn plot analysis (Figure 2A), consisting of 48 upregulated genes and 115 downregulated genes. The detailed records of Venn analysis are shown in Supplementary Table 1. Functional and Pathway Enrichment Analysis The mutual upregulated and downregulated DEGs were uploaded to DAVID to gain further insight into those genes. The detailed results of GO and KEGG pathway analysis as well as GSEA analysis are shown in Table 1 and Figure 2B and F. The GO analysis results showed that the mutual upregulated DEGs were mainly associated with mitotic nuclear division, cell division, and transition of the mitotic cell cycle. For the mutual downregulated DEGs, the GO analysis results were primarily enriched in cellular function alternation, such as chemical synaptic transmission, neurotransmitter secretion, and neurotransmitter transport. In addition, the results of GSEA analysis and KEGG analysis indicated that the mutual DEGs were mainly enriched in cell cycle, oocyte meiosis. Module Screening from the PPI Network The previous 162 mutual DEGs among these 3 datasets were analyzed with the PPI network, and the hub genes were screened with degrees 36 based on the STRING database. Altogether 24 genes were identified as hub genes, including AURKA, NDC80, CDC20, KIF4A, NUSAP1, TTK, MELK, PBK, TOP2A, AURKB, BUB1, UBE2C, KIF11, CEP55, BIRC5, CCNB2, TPX2, CENPE, KIF2C, KIF20A, CENPF, KIF15, OIP5, and ASPM, as listed in well as the top 3 significant modules, as shown in Figure 3A. The functional annotation and enrichment of modules genes were also performed, as shown in Table 3. Enriched function analysis showed that genes in module 1 were primarily related to mitotic, cell division, and spindle; but in modules 2 and 3, genes were mainly enriched in presynaptic active zone, chloride transmembrane transport, and chloride channel complex. Validation of Common Hub Genes by qRT-PCR To validate the expression of AURKA, NDC80, KIF4A, and NUSAP1 in HEB and glioblastoma cells (U87, U251, LN18, and T98), qRT-PCR was performed. The results, presented in Figure 3B, showed a significant difference among those cell lines in that the AURKA, NDC80, KIF4A, and NUSAP1 genes were consistently expressed higher in HEB than in glioblastoma cell lines (P < 0.05). Moreover, among those glioblastoma cell lines, the levels of expression of the AURKA, NDC80, KIF4A, and NUSAP1 genes were slightly different. Survival Curve Analysis To identify the expression of the AURKA, NDC80, KIF4A, and NUSAP1 genes associated with prognosis of patients with glio- blastoma, survival curve analysis was conducted and the results are shown in Figure 4. With PFS as the prognostic outcome of patients with glioblastoma, the low-expressed patients showed significantly higher survival than did high-expressed patients in AURKA (hazard ratio [HR], 2.926; 95% confidence interval [CI], 2.186e3.918; P < 0.0001), NDC80 (HR, 4.106; 95% CI, 3.037e5.552; P < 0.0001), KIF4A (HR, 5.007; 95% CI, 3.695e6.784; P < 0.0001), and NUSAP1 (HR, 3.179, 95% CI, 2.368e4.267; P < 0.0001); and when OS was considered as the prognosis outcome of patients with glioblastoma, the percent survival of low-expressed patients was consis- tently greater than that of high-expressed patients in AURKA (HR, 2.937; 95% CI, 2.179e3.960; P < 0.0001), NDC80 (HR, 4.226; 95% CI, 3.101e5.759; P < 0.0001), KIF4A (HR, 5.034; 95% CI, 3.686e 6.875; P < 0.0001), and NUSAP1 (HR, 3.226; 95% CI, 2.386e4.361; P < 0.0001). Patients with low expression of the AURKA, NDC80, KIF4A, and NUSAP1 genes showed a significantly favorable prog- nosis (P < 0.05), accompanied by a higher percent survival. JNJ-7706621 Reduces Proliferation of Glioblastoma Cells To evaluate the sensitivity of glioblastoma cells to JNJ-7706621, the survival cells after treatment were calculated by MTT assay. As shown in Figure 5A and Supplementary Figure 1, after the augmentation of drug concentrations, the cellular viability (ratio to control) in cell lines U251, LN18, U87, T98, SHG-44, and U373 decreased significantly. However, JNJ-7706621 was relatively well tolerated for human normal cells HUVEC and HL7702, which still had a high cellular viability even when subjected to the highest dose. To determine the antiglioblastoma effects of JNJ-7706621 in glio- blastoma cells, we performed a colony-forming assay. The results showed fewer and smaller clonogenicities in Petri dishes with JNJ- 7706621 than with the control group (Figure 5B). The percentage of clone formation in controls was significantly higher than in drug groups (0.25 mmol/L, 1 mmol/L) (as shown in Figure 5D and E). JNJ-7706621 Inhibits Migration of Glioblastoma Cells The widths of scratched areas were measured after the scratch, after 12 hours, and after 24 hours, to research the migration of glioblastoma cells. In Figure 5C and F, the width of the scratched area was significantly smaller after 24 hours in the control group. However, there was only a slight decrease in JNJ-7706621 group. In addition, after 24 hours, the wounds in the control group were also significantly smaller than in the drug group. JNJ-7706621 Induces Arrest of Glioblastoma Cells in G2/M Phase and Apoptosis of Glioblastoma Cells To study the mechanism of the therapeutic effects of JNJ-7706621in glioblastoma, the glioblastoma cells were measured by flow cytom- etry after culture with different doses of JNJ-7706621 for 24 hours. The results showed that JNJ-7706621 contributed to the arrest of glio- blastoma cells in the G2/M phase. As shown in Figure 6C and D, after augmentation of the JNJ-7706621 dose, the number of cells in the G1 phase decreased, but the number of cells in G2/M increased. In addition, the apoptotic effect of JNJ-7706621 on glioblastoma cells was significant. As shown in Figure 6A, the percentage of cells with normal necrosis, late apoptosis, and early apoptosis was 87%, 4.52%, 7.09%, and 1.35%, respectively, in the control group; 55.2%, 1.23%, 27.3%, and 16.4% in the low-dose group; and 33.2%, 2.57%, 40.6%, and 23.6% in the high-dose group. Normal cells dominated in the control group; in the low-dose group, the percentage of late and early apoptosis cells increased; in the high-dose group, the apoptosis cells replaced the dominance of normal cells. DISCUSSION Despite advances in therapy of glioblastoma, including surgery resection, radiotherapy, chemotherapy, and even adjuvant sys- temic therapy, the prognosis of patients with glioblastoma has remained poor over the past decades.15 In recent years, increasing evidence has implied that cumulative multiple-step genetic change induces malignant progression of glioblastoma.3 Therefore, a comprehensive understanding of the molecular mechanisms, causes, and pathogenesis of glioblastoma is crucial for its diagnosis, therapy, and prognosis. In the present study, 49 glioblastoma samples and 31 formal samples were extracted from messenger RNA microarray datasets GSE50161, GSE42656, and GSE86574 for gene expression data. A total of 1088 DEGs, 1744 DEGs, and 418 DEGs were identified, respectively, from those 3 datasets. There were 48 mutual upre- gulated genes and 115 mutual downregulated genes among those 3 datasets, as shown by Venn plot. After GO analysis of abnormal expression genes, we detected that those upregulated genes were mainly associated with mitosis, such as cell division, mitotic nuclear division, transition of mitotic cell cycle, microtubule, and microtubule motor activity, which were closely related to cancer, whereas downregulated genes were primarily enriched in biological information transfer or cellular function alternation, including chemical synaptic transmission, neurotransmitter transport, synaptic vesicle membrane, g-ami- nobutyric acid A receptor activity, and calcium ion binding. This finding implied, consistent with previous studies, that the defec- tion of cell functions, especially mitotic division, played a main role in progression of tumor, as well as recession of normal cellular functions.16,17 Furthermore, the results of KEGG and GSEA analysis showed that the mutual upregulated and down- regulated DEGs were mainly enriched in cell cycle, oocyte meiosis, nicotine addiction, morphine synapse, p53 signaling pathway, and g-aminobutyric acidemediated synapse. The oocyte meiosis was first contacted to glioblastoma, and the mechanism was presumed to be related to progesterone-mediated oocyte maturation, which was reported to be associated with glioma pathogenesis,18,19 so we hypothesized that progesterone regulator drugs might work in preventing glioblastoma formation and progression. This theory also provided an explanation for the gender distribution of glio- blastoma. In previous studies, the organics included in tobacco, especially nicotine, were shown to be related to glioma, schizo- phrenia, and epilepsy, which suggested that smoking might in- crease the risks of glioblastoma.20 In addition, the disorders of immune function, which are highly related to formation of tumors, were reported to be induced by morphine exposure.21 This finding implies that morphine abuse might induce glioblastoma. With the aim of screening hub genes among DEGs identified in our previous work, the 162 mutual DEGs were analyzed with the PPI network base on the STRING database, and 24 genes were selected with high degrees, in particular AURKA, NDC80, KIF4A, and NUSAP1. AURKA, located on chromosome 20q13, is a serine/ threonine kinase, and drives various processes in the cell cycle, such as centrosome maturation and separation, assembly of bi- polar spindle, trigger of mitotic entry, and alignment of chro- mosomes in metaphase.22 AURKA was reported to be overexpressed in various malignancies by means of mitotic assembly checkpoint overrides and chemoresistance induction,23 including neuroblastoma, neuroendocrine prostate cancer, breast cancer, gastric and esophageal cancers, and chronic myeloid leukemia.24-28 In addition, the biological mechanisms of AURKA in glioma were reported to be associated with b-catenin stabili- zation, which suggested that the application of AURKA inhibitors might improve the prognosis of patients with glioblastoma.23 NDC80, as a mitotic regulator and a major element of outer kinetochore, formed the NDC80 complex with Nuf2, Spc24, and Spc25, which is a dumbbell-like heterotetramer and has been re- ported to drive functions in assembly checkpoint and chromosome segregation of mitosis regulation. The attachment between NDC80 complex and spindle microtubules, which was mainly enriched in proliferation and procession of cancer, was reported in previous studies.29 Moreover, overexpression of NDC80 was discovered in various tumors, such as colon cancer, malignant pleural mesothelioma, human hepatocellular carcinoma, and osteosarcoma.30-33 The curative effect of NDC80 inhibitors against glioma was reported, and that provided us with a new strategy in treatment of patients with glioblastoma.34 KIF4A, one of several KIFs, which ensured a correct order in mitosis by controlling spindle microtubule precisely, was implicated in spindle organization, chromosome alignment, and kinetochore microtubule dynamics.35 The ability to regulate the length of microtubule was a major mechanism for KIF4A to regulate mitotic and induce occurrence of various malignancies, such as breast cancer and human oral cancer.36-38 In addition, the expression of KIF4A of breast cancer cells was reported to be decreased by adriamycin, which was a potential drug to treat pa- tients with glioblastoma.39 NUSAP1, a microtubule-binding pro- tein and crucial regulator of normal cell cycle, drives spindle assembly during mitosis.40 Overexpression of NUSAP1 was reported to be related to cell multiplication and microtubules interaction, which implied the tumorigenicity of NUSAP1.41 Moreover, NUSAP1 was reported to be related to malignancies, such as oral squamous cell carcinoma, hepatic carcinoma, and breast cancer.42,43 We detected that these hub genes were all involved in assembly of spindle and microtubules. It prompted us to hypothesize that JNJ-7706621, a pan-aurora kinase and cyclin-dependent kinase family inhibitor drug, might have an antiglioblastoma effect, and this was verified in the following experiments. In this study, the NDC80, AURKB, KIF4A, CEP55, CENPE, KIF2C, and KIF20A genes were first shown to be involved in glioblastoma, which were precise diagnosis biomarkers, potential treatment targets, and prognosis markers for patients with glio- blastoma. The other genes included as hub genes in the present study provided verification of the connection between these genes and glioblastoma. qRT-PCR was performed to detect the differential expression level of the AURKA, NDC80, KIF4A, and NUSAP1 genes between HEB and glioblastoma cell lines (U87, U251, LN18, and T98). The results of qRT-PCR showed that the expression of those genes in HEB was significantly lower than in glioblastoma cell lines (P < 0.05). In addition, the relationship between expression level of AURKA, NDC80, KIF4A, and NUSAP1 and prognosis of patients with glioblastoma were also clarified in the present study by performing survival curve analysis. The results showed that the patients with low expression of AURKA, NDC80, KIF4A, and NUSAP1 showed a better prognosis in both PFS and OS (P < 0.05). This finding implied that the prognosis of patients with glioblastoma could be predicted by detecting the expression level of those 4 genes. Furthermore, the results of the present study provided biomarkers and targets, which could be applied in diagnosis and treatment of patients with glioblastoma for accurate therapy. The antiglioblastoma effects of JNJ-7706621 were evaluated with MTT assay, colony-forming assay, and scratch assay in vitro. In MTT assay, the cellular viability (ratio to control) in cell lines U251, LN18, U87, and T98 were shown to have a dose-dependent decrease when treated with JNJ-7706621, but there was only a slight impact on human normal cells HUVEC and HL7702. This finding implied that JNJ-7706621 was relatively nontoxic for hu- man normal cells. In colony-forming assays, the numbers and size of clonogenicities in the JNJ-7706621 group were significantly less than in the control group, which was consistent with results that the proliferation of glioblastoma cells was reduced by JNJ-7706621 in MTT assays and that the effects were dose dependent. In scratch assays, the wound widths in the control group decreased sharply with time and were significantly smaller than in the JNJ- 7706621 group after 24 hours. This finding implied that JNJ- 7706621 strongly inhibited migration of glioblastoma cells. In addition, apoptosis assays and cell cycle analysis were also measured by flow cytometry to explore the mechanism of the antiglioblastoma effect of JNJ-7706621. The percentages of cells in G2/M increased after augmentation of the JNJ-7706621 dose, as well as the apoptosis cells, showing that JNJ-7706621 could not only induce the arrest of glioblastoma cells in G2/M but also contribute to the apoptosis of glioblastoma cells.44 In addition, we assumed that the apoptosis of glioblastoma caused by the arrest of cell cycle in G2/M might induce biological changes, such as alteration of genes and expression of enzymes. However, the study still has some limitations. The therapeutic effect was assessed only in in vitro assays and it is also necessary to verify its effect in in vivo assays. In addition, the detailed mechanism needs to be investigated in further studies. CONCLUSIONS A total of 1088 DEGs, 1744 DEGs, and 418 DEGs were identified from the datasets GSE50161, GSE42656, and GSE86574, respectively. The GO, KEGG, and GSEA analysis showed that the enriched function and pathway in upregulated genes were mainly related to mitotic division and cell cycle. AURKA, NDC80, KIF4A, and NUSAP1 were screened as the main hub genes, which were significantly more highly expressed in glioblastoma cells than in HEB. Survival analysis suggested that patients with low expression of AURKA, NDC80, KIF4A, and NUSAP1 had a favorable prognosis. JNJ-7706621 is a promising drug in treatment of patients with glioblastoma. REFERENCES 1. Alifieris C, Trafalis DT. Glioblastoma multiforme: pathogenesis and treatment. Pharmacol Ther. 2015; 152:63-82. 2. Louis DN, Ohgaki H, Wiestler OD, von Deimling A, Figarella-Branger D, Cavenee WK, et al. The 2007 WHO classification of tumors of the central nervous system. Acta Neuropathol. 2007;114:97-109. 3. He WQ, Gu JW, Li CY, Kuang YQ, Kong B, Cheng L, et al. The PPI network and clusters analysis in glioblastoma. Eur Rev Med Pharmacol Sci. 2015;19:4784-4790. 4. Thakkar JP, Dolecek TA, Horbinski, Ostrom QT, Lightner DD, Barnholtz-Sloan JS, et al. Epidemi- ologic and molecular prognostic review of glio- blastoma. Cancer Epidemiol Biomarkers Prev. 2014;23: 1985-1996. 5. Gallego O. Nonsurgical treatment of recurrent glioblastoma. Curr Oncol. 2015;22:e273-281. 6. Batash R, Asna N, Schaffer P, Francis N, Schaffer M. Glioblastoma multiforme, diagnosis and treatment: recent literature review. Curr Med Chem. 2017;24:3002-3009. 7. Roy S, Lahiri D, Maji T, Biswas J. Recurrent glioblastoma: where we stand. South Asian J Cancer. 2015;4:163-173. 8. Huang Y, Tao Y, Li X, Chang S, Jiang B, Li F, et al. Bioinformatics analysis of key genes and latent pathway interactions based on the anaplastic thyroid carcinoma gene expression profile. Oncol Lett. 2017;13:167-176. 9. Ma M, Luo S, Zhou W, Lu L, Cai J, Yuan F, et al. Bioinformatics analysis of gene expression pro- files in B cells of postmenopausal osteoporosis patients. Taiwan J Obstet Gynecol. 2017;56:165-170. 10. Marucci G, Morandi L, Magrinix E, Farnedi A, Franceschi E, Miglio R, et al. Gene expression profiling in glioblastoma and immunohisto- chemical evaluation of IGFBP-2 and CDC20. Virchows Arch. 2008;453:599-609. 11. Danhier F, Ucakar B, Magotteaux N, Brewster ME, Préat V. Active and passive tumor targeting of a novel poorly soluble cyclin dependent kinase in- hibitor, JNJ-7706621. Int J Pharm. 2010;392:20-28. 12. Emanuel S, Rugg CA, Gruninger RH, Lin R, Fuentes-Pesquera A, Connolly PJ, et al. The in vitro and in vivo effects of JNJ-7706621: a dual inhibitor of cyclin-dependent kinases and aurora kinases. Cancer Res. 2005;65:9038-9046.
13. Li J, Wang F, Wang G, Sun Y, Cai J, Liu X, et al. Combination epidermal growth factor receptor variant III peptide-pulsed dendritic cell vaccine with miR-326 results in enhanced killing on EGFRvIII-positive cells. Oncotarget. 2017;8: 26256-26268.
14. Franken NA, Rodermond HM, Stao J, Haveman J, van Bree C. Clonogenic assay of cells in vitro. Nat Protoc. 2006;1:2315-2319.
15. Gilbert MR, Dignam J, Armstrong TS, Wefel JS, Blumenthal DT, Vogelbaum MA, et al. A randomized trial of bevacizumab for newly diagnosed glioblastoma. N Engl J Med. 2014;370: 699-708.
16. Urban´ska K, Sokolowska J, Szmidt M, Sysa P. Proliferative and apoptotic activity of glioblastoma multiforme cells cultured on in ovo model. In Vivo. 2014;28:541-548.
17. Telentschak S, Soliwoda M, Nohroudi K, Addicks K, Klinz FJ. Cytokinesis failure and suc- cessful multipolar mitoses drive aneuploidy in glioblastoma cells. Oncol Rep. 2015;33:2001-2008.
18. Zhou C, Teng WJ, Zhuang J, Liu HL, Tang SF, Cao XJ, et al. Analysis of the gene-protein inter- action network in glioma. Genet Mol Res. 2015;14: 14196-14206.
19. Liu M, Li X, Fan R, Liu X, Wang J. A systematic analysis of candidate genes associated with nico- tine addiction. Biomed Res Int. 2015;2015:313709.
20. Kurup RK, Kurup PA. Central role of hypotha- lamic digoxin in conscious perception, neuro- immunoendocrine integration, and coordination of cellular function: relation to hemispheric dominance. Int J Neurosci. 2002;112:705-739.
21. Zhang QY, Zhang M, Cao Y. Exposure to morphine affects the expression of endocannabinoid receptors and immune functions. J Neuroimmunol. 2012;247:52-58.
22. Nikonova AS, Astsaturov I, Serebriiskii IG, Dunbrack RL Jr, Golemis EA. Aurora-A kinase (AURKA) in normal and pathological cell growth. Cell Mol Life Sci. 2013;70:661-687.
23. Van Brocklyn JR, Wojton J, Meisen WH, Kellough DA, Ecsedy JA, Kaur B, et al. Aurora-A inhibition offers a novel therapy effective against intracranial glioblastoma. Cancer Res. 2014;74: 5364-5370.
24. Katsha A, Arras J, Soutto M, Belkhiri A, El- Rifai W. AURKA regulates JAK2-STAT3 activity in human gastric and esophageal cancers. Mol Oncol. 2014;8:1419-1428.
25. Yang J, Ikezoe T, Nishioka C, Udaka K, Yokoyama A. Bcr-Abl activates AURKA and AURKB in chronic myeloid leukemia cells via AKT signaling. Int J Cancer. 2014;134:1183-1194.
26. Schnepp RW, Khurana P, Attiyeh EF, Raman P, Chodosh SE, Oldridge DA, et al. A LIN28B-RAN- AURKA signaling network promotes neuroblas- toma tumorigenesis. Cancer Cell. 2015;28:599-609.
27. Mosquera JM, Beltran H, Park K, MacDonald TY, Robinson BD, Tagawa ST, et al. Concurrent AURKA and MYCN gene amplifications are har- bingers of lethal treatment-related neuroendo- crine prostate cancer. Neoplasia. 2013;15:1-10.
28. Zheng F, Yue C, Li G, He B, Cheng W, Wang X, et al. Nuclear AURKA acquires kinase- independent transactivating function to enhance breast cancer stem cell phenotype. Nat Commun. 2016;7:10180.
29. Suzuki A, Badger BL, Salmon ED. A quantitative description of Ndc80 complex linkage to human kinetochores. Nat Commun. 2015;6:8161.
30. Xing XK, Wu HY, Chen HL, Feng HG. NDC80 promotes proliferation and metastasis of colon cancer cells. Genet Mol Res. 2016;15. 10.4238/gmr.15028312.
31. Linton A, Cheng YY, Griggs K, Kirschner MB, Gattani S, Srikaran S, et al. An RNAi-based screen reveals PLK1, CDK1 and NDC80 as potential therapeutic targets in malignant pleural meso- thelioma. Br J Cancer. 2014;110:510-519.
32. Ju LL, Chen L, Li JH, Wang YF, Lu RJ, Bian ZL, et al. Effect of NDC80 in human hepatocellular carcinoma. World J Gastroenterol. 2017;23:3675-3683.
33. Xu B, Wu DP, Xie RT, Liu LG, Yan XB. Elevated NDC80 expression is associated with poor prog- nosis in osteosarcoma patients. Eur Rev Med Phar- macol Sci. 2017;21:2045-2053.
34. Gurzov EN, Izquierdo M. RNA interference against Hec1 inhibits tumor growth in vivo. Gene Ther. 2006;13:1-7.
35. Minakawa Y, Kasamatsu A, Koilke H, Higo M, Nakashima D, Kouzu Y, et al. Kinesin family member 4A: a potential predictor for progression of human oral cancer. PLoS One. 2013;8:e85951.
36. Bastos RN, Cundell MJ, Barr FA. KIF4A and PP2A- B56 form a spatially restricted feedback loop opposing Aurora B at the anaphase central spin- dle. J Cell Biol. 2014;207:683-693.
37. Wang H, Lu C, Li Q, Xie J, Chen T, Tan Y, et al. The role of Kif4A in doxorubicin-induced apoptosis in breast cancer cells. Mol Cells. 2014; 37:812-818.
38. Kim H, Fonseca C, Stumpff J. A unique kinesin-8 surface loop provides specificity for chromosome alignment. Mol Biol Cell. 2014;25:3319-3329.
39. Wang L, Wu HJ, Xu N, Wang CY, Ze HE, Shen CL. Inhibitory effect of adriamycin on expression of KIF4A in human breast cancer cells MCF-7. J Southeast Univ (Med Sci Edi). 2010;29:48-51.
40. Zhang M, Yang D, Liu X, Nakashima D, Koyama T, Miyamoto I, et al. Down-regulation of nucleolar and spindle-associated protein 1 (NUSAP1) expression suppresses tumor and cell proliferation and enhances anti-tumor effect of paclitaxel in oral squamous cell carcinoma. PLoS One. 2015;10:e0142252.
41. Kotian S, Banerjee T, Lockhart K, Huang K, Catalyurek UV, Parvin JD. NUSAP1 influences the DNA damage response by controlling BRCA1 protein levels. Cancer Biol Ther. 2014;15:533-543.
42. Zhang M, Yang DH, Liu X, Liu Y, Liang J. Expression and clinical significance of NUSAP1 in hepatical carcinoma. J Southeast Univ (Med Sci Edi). 2013;38:876-881.
43. Chen L, Yang L, Qiao F, Hu X, Li S, Yao L, et al. High levels of nucleolar spindle-associated pro- tein and reduced levels of BRCA1 expressionpredict poor prognosis in triple-negative breast cancer. PLoS One. 2015;10:e0120572.
44. Yu ZY, Xie GF, Zhou GT, Cheng Y, Zhang G, Yao G, et al. NVP-BE235, a novel dual PI3K-mTOP inhibitor displays anti-glioma activity and reduces chemoresistance to temozolomide in human gli- oma cell. Cancer Lett. 2015;367:58-68.