Differential gene expression analysis in glioblastoma cells and normal human brain cells based on GEO database.

The differentially expressed genes between glioblastoma (GBM) cells and normal human brain cells were investigated to performed pathway analysis and protein interaction network analysis for the differentially expressed genes. GSE12657 and GSE42656 gene chips, which contain gene expression profile of GBM were obtained from Gene Expression Omniub (GEO) database of National Center for Biotechnology Information (NCBI). The 'limma' data packet in 'R' software was used to analyze the differentially expressed genes in the two gene chips, and gene integration was performed using 'RobustRankAggreg' package. Finally, pheatmap software was used for heatmap analysis and Cytoscape, DAVID, STRING and KOBAS were used for protein-protein interaction, Gene Ontology (GO) and KEGG analyses. As results: i) 702 differentially expressed genes were identified in GSE12657, among those genes, 548 were significantly upregulated and 154 were significantly downregulated (p<0.01, fold-change >1), and 1,854 differentially expressed genes were identified in GSE42656, among the genes, 1,068 were significantly upregulated and 786 were significantly downregulated (p<0.01, fold-change >1). A total of 167 differentially expressed genes including 100 upregulated genes and 67 downregulated genes were identified after gene integration, and the genes showed significantly different expression levels in GBM compared with normal human brain cells (p<0.05). ii) Interactions between the protein products of 101 differentially expressed genes were identified using STRING and expression network was established. A key gene, called CALM3, was identified by Cytoscape software. iii) GO enrichment analysis showed that differentially expressed genes were mainly enriched in 'neurotransmitter:sodium symporter activity' and 'neurotransmitter transporter activity', which can affect the activity of neurotransmitter transportation. KEGG pathway analysis showed that the differentially expressed genes were mainly enriched in 'protein processing in endoplasmic reticulum', which can affect protein processing in endoplasmic reticulum. The results showed that: i) 167 differentially expressed genes were identified from two gene chips after integration; and ii) protein interaction network was established, and GO and KEGG pathway analyses were successfully performed to identify and annotate the key gene, which provide new insights for the studies on GBN at gene level.

Introduction

As the most malignant type of astrocytic tumors, the recurrence and mortality rates of GBM are extremely high (1). Studies have found that the molecular mechanisms of primary glioblastoma (GBM) and secondary GBM were different (2). Primary GBM is caused by the overexpression of epidermal growth factor receptor (EGFR), while secondary GBM is caused by the mutations of p53 (3). Due to the differential expression of a large number of genes in GBM, conventional biomolecular methods cannot be used to demonstrate the pathogenesis of GBM. Gene expression profile chip, which can measure the expression levels of a large number of genes, is an ideal approach for the analysis of molecular mechanism of GBM (4). In recent years, more and more gene expression profile data become available, and the use of bioinformatics to analyze gene expression profile data has become a new research hotspot (5). In this study, bioinformatics methods were used to analyze the data of gene expression profiles with an expectation of analyzing the differentially expressed genes between GBN and normal human brain cells, so as to provide new insights for the studies on the pathogenesis of GBM.

Materials and methods

Gene expression profile data

Data of gene chip GSE12657 and GSE42656 were obtained from GEO database. GSE12657 was from Neuropathology in the Department of Medicine at Imperial College London with 7 cases of GBM patients as experimental group and 5 cases of normal samples as a control group. GSE42656 was from Neuroscience and Trauma at Barts and the London School of Medicine and Dentistry with 5 cases of GBM patients as experimental group and 8 cases of normal samples as a control group. This study was approved by the Ethics Committee of Xiangyang No. 1 People's Hospital, Hubei University of Medicine. Signed written informed consents were obtained from the patients and/or guardians.

Raw data preprocessing and screening and integration of differentially expressed genes. Affymetrix Expression Console and RMA algorithm were used for quality control, standardization and log2 conversion for the raw data of gene chips. Microarray data analysis package (Linear Models for Microarray Data, Limma) in ‘R’ software was used to screen the differentially expressed genes from raw data of two gene chips. Gene integration of differentially expressed genes identified from two gene chips was performed using RobustRankAggreg.

Gene Ontology (GO) enrichment analysis

DAVID and the plug-in unit ‘Bingo’ of Cytoscape software (San Diego, CA, USA) were used for GO enrichment analysis and functional annotation after gene integration. Database for Annotation, Visualization and Integration Discovery (DAVID) analysis, DAVID network software (NIH, Bethesda, MD, USA) contains almost all major public bioinformatics resources. DAVID can be used to annotate gene-related biological mechanisms using standardized gene terminology.

DAVID knowledge base is designed to facilitate high-throughput gene functional analysis. DAVID provides a wide range of heterogeneous annotation data in a centralized location for a given gene list. DAVID enriches the biological information for individual genes. DAVID knowledge base can be downloaded from the following website: https://david.ncifcrf.gov/.

KEGG pathway analysis. KEGG pathway analysis and functional annotation for differentially expressed gene were performed using KOBAS 3.0 software (Peking University, Beijing, China).

KOBAS is the first software to use the hypergeometric distribution method to determine the significance of pathway enrichment. KOBAS has been successfully used in the study of different organisms such as plants, animals and bacteria. KOBAS server can be accessed at https://kobas.cbi.pku.edu.cn.

Protein interaction network analysis

STRING software (STRING 10.0; European Molecular Biology Laboratory, Heidelberg, Germany) was used to analyze the protein-protein interaction (PPI) of differentially expressed genes. PPI refers to the forming of protein complex by two or more protein molecules through non-covalent bonds. STRING can be accessed at https://string-db.org/.

Results

Screening of differentially expressed genes

A total of 702 differentially expressed genes were identified from gene chip GSE12657, and 548 genes were significantly upregulated and 154 genes were significantly downregulated (p<0.01, fold-change >1). In gene chip GSE42656, 1,854 differentially expressed genes were identified, and 1,068 genes were significantly upregulated and 786 genes were significantly downregulated (p<0.01, fold-change >1). After gene integration, 167 differentially expressed genes including 67 downregulated genes and 100 upregulated genes were identified. Those genes showed significantly different expression levels in GBM compared with normal human brain cells (p<0.05).

GO enrichment analysis

The list of differentially expressed genes was submitted to DAVID Bioinformatics Resource Network (https://david.ncifcrf.gov/) with OFFICIAL-GENE-SYMBOL and Gene List were selected. All other parameters were default.

Differentially expressed genes were mainly enrich in ‘neurotransmitter:sodium symporter activity’ and ‘neurotransmitter transporter activity’, which can affect the activity of neurotransmitter transportation (Fig. 1).

Figure 1.

Results of GO enrichment. Abscissa is the enriched GO, and ordinate is the number and ratio of the differentially expressed genes. Different colors represent different GO classes, namely Molecular function, Biological process, and Cellular component. GO, Gene Ontology.

KEGG pathway analysis

KEGG pathway analysis and functional annotation were performed using KOBAS 3.0 software. Four key KEGG pathways included: ‘Dopaminergic synapses’, ‘MAPK signaling pathway’, ‘Glyoxylate and dicarboxylate metabolism’ and ‘Protein processing in endoplasmic reticulum’ (Table I).

Table I.

Results of KEGG pathway analysis.

Term	Count	P-value	FDR
hsa04141:Protein processing in endoplasmic reticulum	3	0.000309962	0.01239848
hsa04728:Dopaminergic synapse	2	0.004768424	0.095368478
hsa04010:MAPK signaling pathway	2	0.017081905	0.138117097
hsa00630:Glyoxylate and dicarboxylate metabolism	1	0.022356127	0.138117097
hsa03410:Base excision repair	1	0.026161426	0.138117097
hsa04130:SNARE interactions in vesicular transport	1	0.026920764	0.138117097
hsa04962:Vasopressin-regulated water reabsorption	1	0.034482696	0.138117097
hsa03420:Nucleotide excision repair	1	0.036740165	0.138117097
hsa05030:Cocaine addiction	1	0.038242305	0.138117097
hsa04978:Mineral absorption	1	0.040491267	0.138117097

Term, enriched KEGG; count, the number of differentially expressed genes of each term; P-value, enrichment statistical P-value; FDR, P-value after correction.

Protein interaction network analysis. Thirty outstanding proteins were identified through PPI analysis of STRING software. SNAP25, SYP, NAPA, TUBB2A and TUBB4A proteins were relatively more important. As the most important protein, CALM3 connected 17 nodes (Figs. 2 and 3).

Figure 2.

The diagram of protein interaction network. Circle represents the gene, and lines represent the protein interaction between the genes, and the information inside the circle describes protein structure: small nodes, protein of unknown 3D structure; large nodes, some 3D structure is known or predicted; a red line indicates the presence of fusion evidence; a green line, neighborhood evidence; a blue line, coocurrence evidence; a purple line, experimental evidence; a yellow line, text mining evidence; a light blue line, database evidence; a black line, coexpression evidence.

Figure 3.

Core protein histogram. The vertical coordinates is the gene name, and the horizontal coordinates is the number of adjacent genes, and the height represents the number of lines connected genes.

Discussion

As the most common and most aggressive diffuse glioma (6), GBM is the most malignant type of astrocytic tumor. GBM develops from cortex and shows infiltrative development. GBM can simultaneously affect several lobes (7). GBM is characterized by heterogeneity of morphology, genetics and gene expression (8). In recent years, with the rapid development of bioinformatics, the use of gene expression profiles to explore the relationship between gene differential expression and disease development has attracted more and more attention. However, those data have not been comprehensively investigated (9). In this study, gene chip data were used to identify the differentially expression genes between GBM and normal human brain cells through enrichment analysis and protein interaction analysis with the expectation of exploring the pathogenesis of GBM. As the most common endogenous primary brain tumor in adults, GBM represents the most common type of diffuse glioma (10). Central Brain Tumor Registry of the United States (CBTRUS) reported that GBM mainly affect patients between 75 and 84 years, and the incidence of GBM showing an increasing trend and incidence in white males is ever higher (11). GBM is rare in children and only account for ~3% of all primary brain and CNS tumors. The five-year survival rate is about 12% for children and less than 5% for adults (12). In spite of the achievement in the treatment of GBM, the survival rate of patients is still very low (13). GBM exists in cerebral cortex and showed strong invasion ability, so the course of disease is short and average survival period is only ~14 months (14). Without treatment, GBM patients cannot survive more than 2 months, so the development of effective diagnosis and treatment methods is always needed (15). Previous studies have shown that the development of GBM is very complex and is related to the abnormal expression of proto-oncogenes or tumor suppressor genes, which can lead to abnormal activation or dysregulation of intracellular signaling pathways (16).

In this study, GEO public database was used. The ‘limma’ package was used to analyze and integrate data. A total of 167 differentially expressed genes were identified. GO enrichment analysis and KEGG pathway analysis showed that the differentially expressed genes were mainly involved in neurotransmitter transporter activity, neurotransmitter:sodium symporter activity, cellular processes, solute: sodium symporter activity, monoplast processes, protein processing in endoplasmic reticulum, dopaminergic synapses, MAPK signal transduction pathway and glyoxylate and dicarboxylate metabolism. Neurotransmitter transporter activity and protein processing in endoplasmic reticulum are the two most outstanding pathways. However, CALM3 protein is the most influential protein on GBM in the network formed by CALM3, SNAP25, SYP, NAPA, TUBB2A, TUBB4A, DYNC1I1, GRIA1, STXBP1, ANK3, GOT1, GAD2, PPP3CB, SNAP91, AMPH, ATP2A2 and other genes. CALM3 gene was reported to be closely associated with long QT syndrome (17). More studies are needed to investigate the mechanism of the roles of CALM3.

Bioinformatics is a new discipline that combines biological science and computer science (18). Major bioinformatics tools were used in this study to identify the differentially expressed genes. Public database of gene chip data was used in this study, which significantly reduced the use of financial and material resources. Based on the strict inclusion criteria, the most reliable gene chip data were selected to avoid errors.

This study is limited by the small sample size. Gene expression in GBM can be altered by certain factors (19), and the small sample size failed to cover different races and regions, which can affected the gene expression in GBM (20).

In this study, CALM3 gene was proved to be related to the protein processing and transporter activity in GBM. Our future study will focus on those pathways. Studies on GBM at gene level are rare. Therefore, more studies are needed to improve the diagnosis, treatment and prognosis of GBM.