Identification of potential biomarkers and candidate small molecule drugs in glioblastoma

Glioblastoma (GBM) is a most common and aggressive malignant brain tumor, accounting for 16% of all primary brain and central nervous system neoplasms [1]. The mean survival of GBM is approximately 14.6 months, and GBM is one of the most challenging malignancies to treat due to its high heterogeneity, high recurrence rate, and diffusing invasiveness [2]. Despite extensive efforts to explore novel therapies, the survival of GBM has not markedly improved. Therefore, it is necessary to develop effective treatment options. Currently, gene therapy, molecularly targeted therapy, and immunotherapy are promising treatment approaches [3].

Extensive studies have reported the biomarkers and drug targets for GBM treatment. Previous study indicated that genes such as estrogen receptor 2 (ESR2), ELOVL fatty acid elongase 6 (ELOVL6), iroquois homeobox 3 (IRX3), PDZ binding kinase (PBK), centromere protein A (CENPA), and kinesin family member 15 (KIF15) were significantly associated with the prognosis of GBM, suggesting that these genes might be potential targets for GBM treatment [4, 5]. Additionally, drugs like triple-drug therapy (bevacizumab, irinotecan, and temozolomide) had benefit effect on recurrent GBM [6]. However, different studies often yield diverse results and the molecular mechanism of GBM pathogenesis has not been entirely elucidated. Thus, it is desperately required to explore novel biomarkers and small drug molecules.

Currently, the microarray gene expression research has been performed to uncover the molecular mechanism of various cancers. The mRNA data are collected from two databases, including Gene Expression Omnibus (GEO) and The Cancer Genome Atlas (TCGA). The GEO database can be applied to identify the differentially expressed genes (DEGs), explore molecular signal and its correlation, and analyze gene regulation network [7]. However, due to the limited samples, the analysis results of a single microarray dataset may be biased and unreliable. Hence, integrated analysis of multiple datasets can improve the accuracy and reliability of the results, thus obtain a comprehensive discovery of DEGs in tumors.

In the present study, three microarray datasets related GBM were selected for further study, the raw data of mRNA profile were downloaded from GEO database and integrated bioinformatics analyses of three data sets were conducted. The overlapping DEGs were identified by the intersection of three datasets. Then, the DEGs associated with GBM prognosis were screened using TCGA database. Functional enrichment analysis was performed to understand the biological functions of these DEGs. We also established a protein–protein interaction (PPI) network to screen hub genes. Thereafter, the mRNA and protein expression level of hub genes were respectively verified by using UALCAN online tool and Human Protein Atlas (HPA) database. Finally, the small molecule drugs of GBM were explored by connectivity map (CMAP) database. The flow chart of this study protocol is shown in Fig. 1.

Although significant breakthrough in GBM treatment programs, including surgery, molecular therapy, and drug treatment, the prognosis for GBM patients remains poor and unchanged over the last 30 years [21]. Therefore, revealing the etiology and molecular mechanism of GBM might play important role in the diagnosis and treatment of tumor. In this study, bioinformatics analysis was used to screen the potential hub genes associated with GBM. By integration analysis of three GEO datasets of GBM, 613 overlapping DEGs were identified, among these, 78 DEGs were significantly associated with the OS of GBM. The GO analysis showed that these DEGs was mainly enriched in trans-synaptic signaling; and the KEGG pathways enrichment analysis indicated that DEGs were significantly involved in longevity regulating pathway. PPI analysis revealed that CETN2, MKI67, ARL13B, SETDB1, CALN1, ELAVL3, ADCY3, SYN2, SLC12A5, and SOD1 with high degree of connectivity were selected as hub genes. For CETN2, MKI67, ARL13B, and SETDB1, patients with high expression experienced a worse OS, while high expression of CALN1, ELAVL3, ADCY3, SYN2, ARL13B, SLC12A5, and SOD1 were associated with better overall survival among patients with GBM. To validate the results of bioinformatics analysis, we evaluated the mRNA and protein expression levels of hub genes by using TCGA and HPA databases. The results showed the same gene expression trend as observed in the GEO database, which further confirmed the accuracy of our findings. Specially, CETN2, MKI67, ARL13B, SETDB1, and CALN1 might be potential biomarkers for screening high-risk patients with GBM. Furthermore, the small molecular drugs analysis showed that cycloserine and 11-deoxy-16,16-dimethylprostaglandin E2 might as potential therapeutic drugs for GBM.

In this study, GO analysis revealed that trans-synaptic signaling was the significantly enriched term for DEGs, which was consistent with previous study [22]. During the process of synaptogenesis, the glycans could modulate trans-synaptic signaling [23]; interestingly, glycans served important role in cancer progression and treatment. Glycosylation resulted in a variety of functional changes in glycoproteins, including adhesion molecules and cell surface receptors, such as e-cadherin and integrin. Notably, these changes conferred distinctive phenotypic characteristics connected with cancer cells [24]. Bassoy et al. observed that the sensitivity of glioma cells to cytotoxic lymphocytes might increase with the decrease of glycan surface expression [25]. Besides, metabotropic glutamate receptors, which involved in synaptic signaling, also participated in the transformation of multiple cancer types, such as GBM, breast cancer, and melanoma skin cancer [26]. These findings suggested that trans-synaptic signaling might play a vital role in the pathogenesis of GBM.

PPI analysis showed that CETN2 was a hub gene, as well as the mRNA and protein expression levels of it were over-expressed in the GBM tissues. CETN2 was a member of the calcium-binding protein family, and caltractin played a fundamental role in structure and function of the microtubule-organizing center [27]. In addition, CETN2 was also involved in nucleotide excision repair that was linked with the risk of cancer [28]. Accumulating evidences demonstrated that CETN2 was identified in various types of cancers. Huan et al. revealed that CETN2 was associated with invasive ductal carcinoma of the breast, and might be potential biomarker for breast cancer [29]. It was reported that the down-regulated of CETN2 might have tumor suppressive function in bladder cancer [30]. Similarly, we found that the low expression level of CETN2 was significantly related to better survival of GBM patients. However, no studies have reported the potential mechanism of CETN2 in the initiation and progression of GBM. Hence, the mechanism of how CETN2 contributed to the GBM still need further research.

Meanwhile, we also found MKI67 was closely related to the prognosis of GBM. The protein encoded by MKI67 was necessary for cellular proliferation. Hou et al. showed that the down-regulated of MKI67 could suppress cell growth in the hepatocellular carcinoma cell [31]. Laible et al. indicated that MKI67 was a biomarker of breast cancer [32]. Meanwhile, MKI67 was connected with nuclear features and the survival of GBM [33, 34]. In this study, MKI67 was up-regulated in GBM tissues, and GBM patients with a low MKI67 expression level displayed longer survival. Györffy et al. showed MKI67 was a prognostic factor in breast carcinoma [35]. Taken together, we speculated that MKI67 played vital roles in GBM progression and might serve as a molecular target for GBM treatment.

ARL13B was also involved in the GBM development, and the protein level of ARL13B was higher in tumor samples than in normal samples. Casalou et al. confirmed that breast cancer was promoted by ARL13B, which was connected with cancer cell migration and invasion [36]. Another gene, SETDB1 regulated histone methylation, gene silencing, and transcriptional repression [37]. SETDB1 mediated Akt methylation promoted its k63-linked ubiquitination and activation, leading to tumorigenesis [38]. Meanwhile, the oncogenic role of SETDB1 has been reported in GBM [39], which was further supported our findings. In addition, CALN1 was another hub gene in PPI analysis. CALN1 encoded a protein with high similarity to the calcium-binding proteins of the calmodulin family. CALN1 might influence the invasion and migration of osteosarcoma cell line, and it was also associated with the survival of osteosarcoma [40]. We found the high expression level of CALN1 was related to the poor prognosis of GBM. ELAVL3 was one of neuronal-specific RNA-binding proteins (Hu antigens), which was recognized by anti-hu antibody in the serum of patients with paraneoplastic encephalomyelitis and sensory neuropathy [41]. Delgado-López et al. revealed that the expression of ELAVL3 was increased in the GBM tissues [42]. Unfortunately, we found ELAVL3 was down-regulated in the tumor samples, this discrepancy required further study. ADCY3 could catalyze the formation of cyclic adenosine monophosphate [43]. Hong et al. indicated that ADCY3 was overexpressed in the gastric cancer tissues and promoted cell proliferation, migration, as well as invasion [44]. In this study, we found the high expression level of ADCY3 with worse overall survival. Additionally, we observed SLC12A5 was closely involved in the development of GBM. Verhaak et al. found that SLC12A5 was a common biomarker of GBM [45], which was consistent with our results. Furthermore, we observed that SOD1 was closely relevant to the prognosis of GBM. The protein encoded by SOD1 bound copper and zinc ions, and SOD1 was responsible for destroying superoxide free radicals in the body. Kato et al. demonstrated that the expression level of SOD1 was significantly changed in the GBM [46]. Gao et al. indicated that GBM with low expression level of SOD1 had better response to radiotherapy [47]. In the present study, patients with low expression of SOD1 experienced a better prognosis. Taken together, these genes served vital role in the development of GBM. However, our study was performed based on the bioinformatics analysis, further experimental studies must be conducted to understand the potential effect of key genes in the GBM pathogenesis.

Based on the small molecular drugs analysis, we determined a set of small molecule drugs that had potential to reverse the abnormal gene expression changes of GBM. Among these, cycloserine and 11-deoxy-16,16-dimethylprostaglandin E2 showed highly significant negative correlation and might serve as potential drugs for GBM treatment. Cycloserine is a cyclic analog to d-alanine, which can target alanine racemase and d-alanine ligase, thereby preventing the formation of bacterial cell walls [48]. Recently, cycloserine has been widely used in tuberculosis treatment, but no research has focused on the potential role of it in GBM. In addition, previous study revealed that 11-deoxy-16,16-dimethylprostaglandin E2 (DDM-PGE2) protected proximal renal tubular epithelial cells from potent nephrotoxicity-induced cell damage by exerting anti-oxidative stress [49]. Meanwhile, it also protected against oncotic cell death which induced by H(2)O(2) and iodoacetamide [50]. Similarly, the relationship between DDM-PGE2 and GBM was not investigated. Given the emergence of these small molecules drugs in silico, further studies that explore the potential effects of them on GBM are imperative and will contribute to the study on new therapeutic drugs for GBM.

Despite studies devoted to investigate the molecular mechanisms of GBM development, integrated studies based on multiple datasets are rare. In the present study, 10 hub genes were identified for the first time in GBM by integrated bioinformatics analysis; meanwhile, the mRNA and protein expression levels of them were verified by using TCGA and HPA databases. Importantly, we also screened the putative therapeutic agents for GBM. This study comprehensively analyzed the pathogenesis of GBM, which provided certain guiding significance for the diagnosis and treatment of this disease. Although the clinical value of these genes and drugs in GBM has not been reported in previous study, the importance of them should not be underestimated.

In summary, with the integrated bioinformatics analysis of three GBM-related gene expression profiles, we identified 10 key genes connected with pathogenesis and prognosis of GBM. These hub genes might serve as novel diagnostic and treatment biomarkers of GBM, which might conduct to elucidate the molecular mechanism of the occurrence and progression of GBM. Additionally, a series of small molecule drugs which could reverse the abnormal gene expression of GBM were identified. Our work may provide powerful evidence for the genomic individualized treatment of GBM.