Bioinformatics Analysis Identified Key Molecular Changes in Bladder Cancer Development and Recurrence

Background and Objectives: Bladder cancer (BC) is a complex tumor associated with high recurrence and mortality. To discover key molecular changes in BC, we analyzed next-generation sequencing data of BC and surrounding tissue samples from clinical specimens. Methods. Gene expression profiling datasets of bladder cancer were analyzed online. The Database for Annotation, Visualization, and Integrated Discovery (DAVID, https://david.ncifcrf.gov/) was used to perform Gene Ontology (GO) functional and KEGG pathway enrichment analyses. Molecular Complex Detection (MCODE) in Cytoscape software (Cytoscape_v3.6.1) was applied to identify hub genes. Protein expression and survival data were downloaded from OncoLnc (http://www.oncolnc.org/). Gene expression data were obtained from the ONCOMINE website (https://www.oncomine.org/). Results. We identified 4211 differentially expressed genes (DEGs) by analysis of surrounding tissue vs. cancer tissue (SC analysis) and 410 DEGs by analysis of cancer tissue vs. recurrent tissue cluster (CR analysis). GO function analysis revealed enrichment of DEGs in genes related to the cytoplasm and nucleoplasm for both clusters, and KEGG pathway analysis showed enrichment of DEGs in the PI3K-Akt signaling pathway. We defined the 20 genes with the highest degree of connectivity as the hub genes. Cox regression revealed CCNB1, ESPL1, CENPM, BLM, and ASPM were related to overall survival. The expression levels of CCNB1, ESPL1, CENPM, BLM, and ASPM were 4.795-, 5.028-, 8.691-, 2.083-, and 3.725-fold higher in BC than the levels in normal tissues, respectively. Conclusions. The results suggested that the functions of CCNB1, ESPL1, CENPM, BLM, and ASPM may contribute to BC development and the functions of CCNB1, ESPL1, CENPM, and BLM may also contribute to BC recurrence.


Introduction
Bladder cancer (BC) is a common urogenital cancer, with an estimate of 80,470 new cases and 17,670 deaths in the United States in 2019 [1]. Bladder cancer patients are often diagnosed by cystoscopy for diagnostic testing prompted by haematuria. Approximately 80% of urinary bladder tumors are superficial papillary lesions but also can be multifocal and exhibit a tendency for recurrence: remaining tumors may invade the bladder wall and lead to distant metastases [2]. Treatment for BC includes transurethral resection of bladder tumor (TURBT), chemotherapy, or vaccine-based therapy directed to the bladder, cystectomy, radiotherapy, and chemotherapy [3]. However, BC is a complex disease associated with a high recurrence rate and high mortality, and its biology remains poorly understood [4].
ere are several important risk factors for BC, such as cigarette smoking, occupational chemical exposure (especially to aromatic amines), water arsenic level, Schistosoma haematobium infection, and radiation therapy for pelvic malignancies [5]. Previous studies identified aspects of the molecular mechanism of BC development and recurrence. BC has been genetically associated with mutations of two genes, fibroblast growth factor receptor 3 (FGFR3, for low-grade, noninvasive papillary tumors), and tumor protein P53 (TP53, for high-grade, muscle-invasive tumors) [6]. Treatment with drugs targeting mutations in genes such as FGFR3, VEGF, signal transducer and activator of transcription 3, and CD24 has all shown preclinical activity [4]. Next-generation sequencing (NGS) has drastically increased the understanding of cancer processes including BC, and analyses of these data can provide insight into effective diagnostic and therapeutic BC treatments [7,8].
ere are significant BC molecular profiling data [9][10][11][12]. Researchers have explored screening of urine to detect DNA mutations as an alternative for urine cytology as a tool for the noninvasive detection and surveillance of BC [13]. Additionally, the analysis of frequently mutated genes in BC may suggest potential targets for personalized treatment and predict treatment response [8]. However, to date, it has been difficult to identify key genes related to BC from NGS data. To discover key molecules active in BC, we analyzed BC data from microarray experiments and NGS sequencing data of clinical specimens. Our results suggested CCNB1, ESPL1, CENPM, BLM, and ASPM may contribute to BC development and recurrence.

Identifying Differentially Expressed Genes.
To analyze the microarray data, we compared the gene expression between 58 normal tissues surrounding cancer and 165 primary bladder cancer samples to identify genes involved with tumorigenesis, and gene expression comparison between 165 primary bladder cancer and 23 recurrent samples was also performed to screen genes that promote tumor recurrence. Differentially expressed genes were screened by adjusted p value or p value and fold change (FC). For comparison between surrounding tissue and cancer tissue, differentially expressed genes were restricted by adjusted p value <0.05 and |FC| > 4, and we defined these genes cluster SC (surrounding tissue vs. cancer tissue). For comparison between cancer tissue and recurrent tissue, differentially expressed genes were restricted by p value <0.05 and |FC| > 2, and we defined these genes cluster CR (cancer tissue vs. recurrent tissue).

Merging Data.
We proposed two methods to process the clusters SC and CR: (1) tumorigenesis and recurrence were promoted by the same genes or proteins, the overlap between SC and CR were the key genes, and overlap genes were analyzed to perform Gene Ontology and KEGG pathway analysis and retrieve interacting genes; (2) tumorigenesis and recurrence were contributed by different genes, we would find key genes from clusters SC and CR individually, and SC and CR genes were individually analyzed to perform Gene Ontology and KEGG pathway analysis and retrieve interacting genes. For method 1, Venny 2.1.0 (http:// bioinfogp.cnb.csic.es/tools/venny/index.html) was used to identify overlapping differentially expressed genes between SC and CR. e upregulated and downregulated genes were measured, respectively.

Gene Ontology and KEGG Pathway Analysis.
e Database for Annotation, Visualization, and Integrated Discovery (DAVID, https://david.ncifcrf.gov/) was used to perform Gene Ontology (GO) functional and KEGG pathway enrichment analyses. p < 0.05 was considered as statistically significant.

Retrieving Interacting Genes.
Search Tool for the Retrieval of Interacting Genes (STRING) is an online tool (https://string-db.org) designed to integrate information by consolidating known and predicted protein-protein association data. Molecular Complex Detection (MCODE) in Cytoscape software (Cytoscape_v3.6.1) was applied to screen hub genes. All identified differentially expressed genes described above were analyzed. e top 20 hub genes with connection degree >10 were selected.

Statistical Analysis.
Clinical information was analyzed by SPSS 18.0 (IBM Corporation, Armonk, NY). A Cox regression model was conducted to perform univariate and multivariate analyses. e gene expressions were analyzed by GraphPad Prism 7.0. p < 0.05 is considered to reveal a statistically significant difference.

Results
Analysis was performed using data for 58 normal tissues surrounding cancer, 165 primary bladder cancer samples, and 23 recurrent cancer samples. We identified 4211 differentially expressed genes (DEGs) by analysis of surrounding tissue vs. cancer tissue (SC analysis) and 410 DEGs by analysis of cancer tissue vs. recurrent tissue cluster (CR analysis). ere were 1657 and 258 upregulated DEGs in cluster SC and cluster CR, respectively, and 2514 and 152 individually downregulated DEGs in cluster SC and cluster CR. A comparison of these sets of genes revealed 148 overlap genes, including 91 upregulated and 57 downregulated DEGs ( Figure 1). We next analyzed these genes by performing two kinds of functional analysis.

Gene Ontology and KEGG Pathway Analysis.
In the first analysis, the 91 upregulated and 57 downregulated genes that were differentially expressed in both the comparison of cancer and surrounding tissues and the comparison of cancer and recurrent cancer tissues were analyzed using the Database for Annotation, Visualization, and Integrated Discovery (DAVID; https://david.ncifcrf.gov/). Gene Ontology (GO) functional and KEGG pathway enrichment analyses were performed. GO function analysis revealed enrichment of these DEGs in functions related to the cytoplasm and nucleoplasm. ere is an enrichment of genes involved with protein binding and protein kinase binding, regulating cell division, DNA replication, and cyclin-dependent protein serine/threonine kinase activity. KEGG pathway analysis indicated that the identified DEGs are mainly enriched in the PI3K-Akt signaling pathway, microRNAs related to cancer, and the cell cycle. e 15 most enriched classes based on GO function analysis and the eight most enriched KEGG pathways are listed in Table 1.
In the second analysis, we focused on the DEGs identified by the comparison of cancer and surrounding tissues or those identified by the comparison of cancer and recurrent cancer samples. Analysis of DEGs from the surrounding tissue vs. cancer tissue comparison should reflect key genes participating in tumorigenesis or bladder cancer development. GO function analysis of these genes found high enrichment of functions related to extracellular exosomes, extracellular space, and extracellular matrix. Protein binding, heparin binding, and integrin binding are the main functions of these genes, which participate in cell adhesion, extracellular matrix organization, and aging. KEGG pathway analysis indicated enrichment of these genes in HTLV-I infection, Staphylococcus aureus infection, and focal adhesion (Table 2). We next analyzed the DEGs identified by the comparison of cancer and recurrent cancer samples, which should include genes related to bladder cancer recurrence. GO function analysis revealed enrichment of these genes in functions related to the cytoplasm, cytosol, and nucleoplasm, and analysis of molecular function showed enrichment in protein binding. e most relevant enriched biological processes are angiogenesis and the G1/S transition of the mitotic cell cycle, and KEGG pathway analysis indicated enrichment of these genes in cancer pathways, the PI3K-Akt signaling pathway, and cell cycle ( Table 2).

Hub Gene Analysis.
We used STRING for investigating and integrating interaction between proteins. Data were exported for further analysis by Cytoscape. We defined the top 20 genes with the highest degree of connectivity as the hub genes. For method 1, 20 hub genes are shown in Figure 2(a). Also, hub genes in clusters SC and CR are shown in Figures 2(b) and 2(c).

Clinical Analysis.
Kaplan-Meier analysis was performed for the identified hub genes using the DAVID website. We defined the 20 genes with the highest degree of connectivity as hub genes and determined hub genes for the SC comparison and for the CR comparison. For the 20 hub genes identified in the SC analysis, JUN and CDK6 were associated with the overall survival of bladder cancer patients (Figures 3(j) and 3(o)). High JUN expression increased the risk of death by 40% relative to low JUN expression (p � 0.041), and high CDK6 expression increased the risk of death by 50% compared to low CDK6 expression (p � 0.013). Overall survival analysis of other hub genes did not exhibit statistical significance for high and low expressions (Figures 3(a)-3(i), 3(k)-3(n), and 3(p)-3(t)).
We also determined 20 hub genes for the CR analysis. None of these hub genes were associated with overall survival (Supplement Figure 1). We next analyzed the hub genes and their association with disease-free survival (DFS) instead of overall survival. In this analysis, we found an association of CDK6 with DFS of bladder cancer patients (Supplement Figure 2).
We then downloaded the raw data from OncoLnc for further analysis. Cox regression revealed that CCNB1, ESPL1, CENPM, BLM, and ASPM are related to overall survival (Supplement Table 1). Of these, CCNB1, ESPL1, CENPM, and BLM were identified as hub genes from cluster CR, and ASPM was identified as a hub gene from cluster SC (Supplement Table 1).

Gene Expression in BC.
e expressions of CCNB1, ESPL1, CENPM, BLM, ASPM, and two other genes (JUN and CDK6) associated with bladder cancer patient overall survival are shown in Figure 4

Discussion
In this analysis, we defined differentially expressed genes for the SC comparison of surrounding tissue vs. cancer tissue and for the CR comparison of cancer tissue vs. recurrent tissue and considered the identified DEGs contributing to BC development and contributing to BC recurrence, respectively. Genes found in both SC and CR analyses affect both BC development and recurrence, and key genes identified in either SC analysis or CR analysis but not in both analyses are genes that affect either BC development or recurrence, respectively. GO function analysis discovered DEGs are mainly enriched in cytoplasm and nucleoplasm for both clusters, and KEGG pathway analysis indicated high enrichment of DEGs in the PI3K-Akt signaling pathway. We found that CCNB1, ESPL1, CENPM, BLM, and ASPM may be associated with BC development, and CCNB1, ESPL1, CENPM, and BLM may be associated with BC recurrence. It was interesting that our analysis revealed four genes, CCNB1, ESPL1, CENPM, and BLM, which are associated with both BC development and recurrence. Although JUN and CDK6 were not associated with BC development or recurrence, they may be prognostic factors for overall survival (Figure 3(j), 3(o)). e p value was unadjusted for tumor recurrence, and without a correction for multiple tests, the results are meaningful but not conclusive for recurrent tumors.
Among the identified genes, we found CCNB1 was 4.8fold more highly expressed in BC compared to the level in normal tissues (p = 3.86E− 13). CCNB1 is an important cell cycle protein and is a key regulator of the G2/M checkpoint. High levels of CCNB1 usually lead to cell immortalization, resulting in aneuploidy, which contributes to chromosomal instability and is related to the aggressive nature of certain cancers [14]. e involvement of CCNB1 with BC was demonstrated previously [15][16][17][18][19].
ree bioinformatics analyses indicated that CCNB1 was a key gene in BC, consistent with our findings [17][18][19]; however, other hub genes reported previously such as KIF4A, TPX2, BUB1B, CDK1, ISG15, KIF15, RAD54L, and TRIP13 were not identified in our analysis. CCNB1 has been positively correlated with cell proliferation, invasion, and migration [20]. Gene expression profiling in 102 patients with non-muscleinvasive BC identified an association of CCNB1 with disease recurrence [16], and other analyses showed a positive correlation of CCNB1 with pathological stage and metastasis [20]. Cytological experiments may be required to confirm the function of CCNB1 in BC cells.
Our analysis discovered ESPL1 was expressed at a level 5.0-fold higher in BC than the level in normal tissues (p � 5.92E − 20). ESPL1, also known as extra spindle poleslike 1 protein or separin, plays a central role in chromosome segregation by cleaving the cohesin complex at the onset of anaphase, and altered ESPL1 activity is correlated with aneuploidy and cancer [21]. Genomic analysis of transitional     cell carcinoma (TCC) by both whole-genome and wholeexome sequencing of 99 individuals with TCC found frequent alterations in ESPL1 [22]. ESPL1 expression was negatively correlated with gastric adenocarcinoma pathologic stage progression, and the high expression of ESPL1 was significantly correlated with favorable outcomes [23]. In contrast, ESPL1 functions as an oncogene rather than as an antioncogene in breast cancer [24]. Further work is required to resolve the conflicting roles of ESPL1 in cancer and determine its function in BC. CENPM was also identified as a key gene associated with BC. CENPM showed an 8.7-fold higher expression in BC compared to the levels in normal tissues (p � 5.91E− 26). A study comparing the effects of garlic extracts and cisplatin for the treatment of BC identified 515 common anticancer genes, including CENPM. BC patients with low expression of CENPM showed significantly better progression-free survival than those with high expression of CENPM [25]. CENPM encodes centromere protein M, which is a component of the CENPA-NAC (nucleosome-associated) complex. e complex plays a central role in the assembly of kinetochore proteins, mitotic progression, and chromosome segregation [26]. us, we speculated that CENPM may be an important gene in BC development and recurrence.
BLM participates in DNA replication and repair and plays an important role in the maintenance of genome stability [27,28]. Mutations altering BLM function are associated with highly elevated cancer susceptibility [29]. Its roles in BC are unknown, and our research suggests BLM function may be related to BC development and recurrence. e expression of BLM was 2.1-fold higher in BC than the level in normal tissues (p � 5.19E− 14), but further research will be required to uncover the underlying mechanisms.
ASPM is the only gene that we found involved that was associated with BC development but not recurrence (Supplement Table 1). ASPM exhibited a 3.7-fold higher expression level in BC than the level in normal tissues (p � 2.56E− 13). Abnormal spindle-like microcephaly-associated protein is encoded by ASPM and is involved in mitotic spindle regulation and the coordination of mitotic processes [30]. Recently, another study reported significant overexpression of ASPM in bladder cancer that was associated with invasive pathological characteristics [31]. ese results support our findings linking ASPM function to BC. ere are some limitations of this analysis that are worth noting. First, this research was based on data from a single gene array, so the inclusion of other expression data would strengthen the conclusions. Second, altered expression levels of these genes in BC have not been verified by biological methods, so additional experiments to knock down or overexpress these genes should be conducted. Finally, a major drawback of this study is insufficient evidence to suggest changes at the protein level, since the analysis was based only on mRNA expression data and protein interactions were predicted by STRING.
In conclusion, our study suggested CCNB1, ESPL1, CENPM, BLM, and ASPM may be associated with BC development, and CCNB1, ESPL1, CENPM, and BLM may be associated with BC recurrence. e functions of most of these candidate genes have not been the focus of previous studies of BC, and their functions in this cancer should be verified by in vivo and in vitro experiments.

Data Availability
No data were used to support this study.

Disclosure
Qingke Chen, Jieping Hu, and Jun Deng are co-first authors.

Conflicts of Interest
e authors declare that they have no conflicts of interest.