Four novel biomarkers for bladder cancer identified by weighted gene coexpression network analysis

Received: 21 November 2018 Revised: 27 February 2019 Accepted: 6 March 2019 DOE: 10.1002p28546 ORIGINAL RESEARCnes-highdearee of intra modole connecHvity WILEYa Phypliology ARTICLE Four novel biomarkers for bladder cancer identified by weighted gene coexpression network analysis Zi-Xin Guo | Xiao-Ping Liu Sheng Li124 | Yu-Jia Feng | Ying-Jie Zhao3 Xin Yan120 Tong-Zu Liu a cadtin /dsase can be ideiel anatrdy ocurring gere in uohich Department of Urology, Zhongnan Hospital Abstract Bladder cancer (BC) is one of the most malignancies in terms of incidenceCo e recurrence worldwide. The aim of this study is to find out novel and proplostic dvdarn of Wuhan University, Wuhan, China Department of Biological Repositories Zhongnan Hospital of Wuhan University Wuhan, China rst. we identified 258 differentialy expressed biomarkers for patients with BC. genes by using GSE1991 from Gene Expression Omnibus database. pecond, a total of 33 modules were identified by constructing a coexpression network by using weighted gene coexpression network analysisandyellow module was regarded as the Department of Urology, Zunyi Medical University, Zunyl China Human Genetics Resource Preservation Center of Hubei Province, Wuhan, China wnere Corespondence Sheng Li Department of Biological Repositories, Zhongnan Hospital of Wuhan University, 169 Donghu Road 430071 key module. Furthermore, by constructing protein-protein interaction networks, we preliminarily picked out 13 genes. Among them four hub genes (CCNB1, KIF4A TPX2 and TRIP13) were eventually identified by using five different methods (survival analysis, one-way analysis of variance, the Spearman correlation analysis receiver operating characteristic curve, and expression value comparison), which Wuhan, China Email: lisheng-znyv@whuedu.cn Funding information were significantly correlated with the prognosis of BCc. The validation of transcrip- National Natural Science Foundation of China Grant/Award Number: 81802541: Zhongnan Hospital of Wuhan University Science Technology and Ienovation Seed Fund, Grant Award Number: ampy2017050: 351 Talent tional and translational levels made sense (based on Oncomine and The Human Protein Atlas database). Moreover,functional enrichment analysis suggested that all- the hub genes played crucial roles in chromosome segregation, sister chromatid segregation, nuclear chromosome segregation mitotic nuclear division nuclear division, and organelle fission during cell mitosis. In addition, three of the hub genes Project of Wuhan University (Luojia Young Scholars: S (KIF4A, TPX2, and TRIP13) might be potential targets of cancer drugs according to the results of the genetical alteration. In conclusion, this study indicates that four hub genes have great predictive value fon the prognosis of BC, and may contribute to the exploration of the further and more in-depth research of BC KEYWORDS bladder cancer, coexpression, histologic grade, hug genes, WGCNA 1 INTRODUCTION cancers (Siegel et al. 2018). For diagnosis, cystoscopy and biopsy are still the gold standard (Emerson & Cheng, 2005). Unfortunately, the Bladder cancer (BC) is the most common malignancy oft the urinary average age for BC at diagnosis is 65 years, which means that most system (Ebrahimi et al, 2019). There are 437,422 new cases annually BCs are diagnosed at an advanced stage (Shadpour, Emami & in the world according to a recent research (Ebrahimi et al, 2019 Haghdani, 2016. Many patients often lost the optimal chance for Although in the country with health facilities well developed, BC most effective treatment. As for the prognosis of BC. the situation was not optimistic. Five-year survivai rate for BC patients was reported as low as 50-70% (Hussein et al, 2016) Moreover, it has causes a lot of troubles (Siegel, Miller, & Jemal, 2018). For example, in the USA, BC is the 13th most common cause of deaths among all 0 2019 Wley Periodicals, Inc. 90738C J Cell Physiol 2019234 19073-19087 wileyonlinelibrary.com/journal/jcp Thee is a So-70% chace that Pe boto Patients suruiue only up to Syrs

1907s YAN T AL Cellular PhysiologyWILEY- 2.6 Construction of protein-protein interaction (PPI) networks addition, we validated the messenger RNA (mRNA)-level and translational-level expressions of the hub genes based on the Oncomine (http://www.oncomine.org/: Rhodes et al., 2004) and By means of the Search Tool for the Retrieval of Interacting Genes The Human Protein Atlas database (https//www.proteinatlas.org/ Uhlén et al, 2015) (STRING: Szklarczyk et al, 2015), we constructed the PPI networks of common DEGs and genes in hub module. Parameters setting network scoring: degree cutoff 2 cluster finding: node score 2.9 Genetical alteration of hub genes cutoff 0.2. k-core-2 and maximum depth 100. In this study, we calkulated the degree of genes by network analyzer (a tool in Visualization and analysis of cancer genomic data sets can be realized Cycoscape software https//cytoscape.org. Genes with degree by using CBio Cancer Genomics Portal (http://www.cbioportal.org/ Cerami et al, 2012 Gao et al, 2013). In the present study, greater than or equal to 10 were considered to be hub genes in the PPl network. CBioPortal was used to explore the genetic alterations of the hub genes and the relationships between genes and drugs 2.7 Identification of hub genes 2.10 Investigation of the associations between the clinical features of patients with BC and the hub gene expression levels In the present study, a key module was chosen Hub genes in coexpression network were identified under the threshold of IMMI 0.50 and IGSI 0.20. The common hub genes in coex- pression network, PPI network of key module, and PPi network of Based on GSE13507, we evaluated the median of hub DEGS were picked out for follow-up analysis. Then we used five genes expression levels. After that, 165 BCs from GSE13507 different methods to identify hub genes among these genes using were divided into two groups. The associations between the two GEO data sets (GSE13507 and GSE31684). Survival analysis clinicopathological features of BC patients and the hub gene was performed by R package “survival” (Therneau, 2015) using expression levels were analyzed by the Version 21.0) test through SPSS (IBM GSE13507, and we split 165 BCs into two groups based on genes expression (high group, n 82: low group, n-83). This package also generated Kaplan-Meier survival curve. The one-way 2.11 Gene ontology (GO) and kyoto encyclopedia of genes and genomes (KEGG) pathway enrichment analysis analysis of variance (ANOVA) test and the Spearman correlation analysis were performed using GSE13507 and GSE31684. Both of the two analyses were performed using SPSS (IBM, Armonk, NY. Version 21.0. Meanwhile, by means of R package “plotROC We performed GO (Ashburner et al. 2000) and KEGG pathway (Sachs, 2017), receiver operating characteristic curve (ROC) enrichment analysis (Kanehisa & Goto. 2000) for DEGS and genes in analysis was performed.dn GSE13507, we calculated the area key module by using R package “cluster Profiler (Yu, Wang Han, & under curve (AUC) to distinguish BC samples from normal tissues He, 2012). In this study, we only showed the results of biological process (BP) and KEGG. Gene sets at pe005 were considered to be In GSE31684, we used AUC to differentiate BC of high grade from BC of low grade Alter that, we compared the genes ignificantly enriched expression levels between BCs and normal bladder tissues using GSE13507 and TCGA BLCA data The boxplots were drawn using 2.12 Gene set enrichment analysis (GSEA) and guilt of association of hub genes R package “gestatsplot” (Patil, & Powell, 2018) and gene expression profiling interactive analysis (GEPIA: Tang et al 2017) Genes satisfied the conditions (p<005 in all analyses and With the same method we mentioned before, the 165 BC samples AUC2080) were considered to be hub genes in the study. An from GSE13507 were classified into two groups. GSEA (Subramanian upset plot was also performed using R package “UpSetR et al 2005) was conducted between the two groups. Signaling (Conway, Lex, & Gehlenborg. 2017) to overlap genes in these pathways reached the standards (nominal p<005: IES >06gene five analyses. Moreover, the Pearson correlation between hub size 100, FDR 25 % ) were considered significant in the present genes and marker of proliferation Ki-67 (MKi67) were performed study. Also, we performed batch Spearman correlation analysis of based on the TCGA-BLCA data. Venn diagram was performed by hub genes using the TCGA-BLCA data. Correlation coefficient online tool Venn-diagrams (http://bioinformatics.psb.ugent.be/ absolute values were calculated, and we selected the top 500 genes beg/tools/venn-diagrams) of each hub gene using R packages “dplyr” and “tidyr.” Moreover, functional enrichment analysis was performed by using “clusterPro- filer.” According to the results, we predicted the lurking functions of hub genes. We called this method guilt of association Validation of hub genes 2.8 T stage (Ta. T1. T2. T3. and T4) boxplots and tumor grade Then, we compared the results between GSEA and guilt of (low and high) boxplots were performed using “ggstatsplot. In association

YAN AL 19079 bla Physiclogy WILEY Kaplan-Meer Curve for BC overall survival Kaplan-Meler Cuve for BC overall survival (a) (b) co 180 P0012 o00- 150 150 Tme Time Number at risk Number at risk Highg 20 High gro l 12 150 150 Tme Tme Kaplan-Meier Curve tor BC overal survival Kaplan-Meier Curve tor BC overal survival (c) (d) 180 100 P-000 P0006 180 Time Time Number at risk Number at risk 150 Tme FIGURE 4 Survacal analysis-of the association between the expression levels of hub genes and overall survival time in BC (based on GSE13507). (a) CCNB1 (b) KIF4A (c) TPX2, and (d) TRIP13. BC: bladder cancer (Color figure can be viewed at wileyonlinelibrary.com 3.4 Validation and genetical alteration of hub genes. These results made the hub genes we screened out reliable. As for genetical alteration, four hub genes altered in 159 (39 % ) of 412 genes patients (Figure 6b). As shown in Figure 6a, TRIP13 altered most Based on GSE13507 and GSE31684, the grade and stage boxplots of hub genes were shown in Figure 57. In addition, mRNA expression (23%) and the main type was mRNA upregulation. A network containing 58 genes (three real hub genes and 55 most variant genes) is shown in Figure 6c. TPS3 was significantly vital in the network. As for the relationship between anticancer drugs and hub genes, we found CCNB1 was the target of cancer drugs. But there was no drug levels were all significantly higher in BC tissues than those in normal bladder tissues (Figure $8c), which was suggested by Oncomine database. Figure S9 showed the translational-level expression of hub ad ung e d ng d ng god ung