Sage Journals: Discover world-class research

Abstract

BACKGROUND:

Bladder cancer is one of the most common genitourinary malignancies, with a high rate of recurrence and progression. The prognosis for patients with bladder cancer, especially muscle-invasive bladder cancer, remains poor despite systemic therapy.

OBJECTIVE:

To explore the underlying disease mechanisms and identify more effective biomarkers for bladder cancer.

METHODS:

Weighted gene co-expression network analysis (WGCNA) and protein-protein interaction (PPI) network analysis were applied to identify hub genes correlated with the bladder cancer progression. Survival analyses were then conducted to identify potential biomarkers correlated with the prognosis of bladder cancer. Finally, validation and analysis of these potential biomarkers were conducted by a series of bioinformatics analyses.

RESULTS:

Based on the results of weighted gene co-expression network analysis and protein-protein interaction network analysis, ten hub genes closely correlated with bladder cancer progression were identified in the relevant module. Survival analyses of these genes indicated that elevated expressions of six potential biomarkers (COL3A1, FN1, COL5A1, FBN1, COL6A1 and THBS2) were significantly associated with a worse overall survival. Furthermore, these 6 potential biomarkers were validated in association with the progression of bladder cancer. Bladder cancer samples with higher expression of these genes were most significantly enriched in gene set associated with ECM-receptor interaction.

CONCLUSIONS:

This study identified several biomarkers associated with bladder cancer progression and prognosis. As novel findings, these may have important clinical implications for diagnosis, treatment and prognosis prediction.

Keywords

Bladder cancer weighted gene co-expression network analysis (WGCNA)biomarkers progression prognosis

1. Introduction

Bladder cancer is the 7th most commonly diagnosed cancer in the male population worldwide [1, 2]. In China, there were approximately 80,500 new cases and 32,900 deaths from this malignancy in 2015 [3]. Approximately 75% of newly diagnosed cases are non-muscle-invasive bladder cancer (NMIBC) with the remaining being muscle-invasive bladder cancer (MIBC) [4]. For patients with NMIBC, the treatment with a relatively low mortality is complete resection of the tumor followed by Bacillus Calmette-Guérin (BCG) intravesical immunotherapy or chemo- therapy [5, 6]. However, patients with noninvasive disease suffer from a high rate of recurrence and progression. Reported 5 year rates of NMIBC recurrence range from 50% to 70%, and 5 year rates of progression to MIBC range from 10% to 30% [7]. For patients with MIBC, radical cystectomy with neoadjuvant chemotherapy is usually recommended [8]. However, MIBC is frequently associated with metastatic disease and has a high rate of mortality. The prognosis of patients with MIBC remains poor despite systemic therapy [8, 9]. Early diagnosis with individualized treatment and follow-up is the key to successful outcomes. Therefore, it is urgent to explore the potential mechanisms involved in cancer progression and find more effective biomarkers for early diagnosis, treatment and prognosis.

With the rapid development of genome-wide sequencing, more tumor-related genes are quickly being detected, leading to breakthroughs in cancer research [10]. Gene expression profiles such as microarray and RNA sequencing have been widely used to identify biomarkers associated with bladder cancer progression. However, the high degree of interconnection between genes has been often ignored in many studies, although genes with similar expression patterns may be functionally related. Weighted gene co-expression network analysis (WGCNA), a systems biology algorithm, is proposed to describe correlation patterns among genes across microarray samples, and correlation relationship between clusters of highly co-expressed genes or modules and external sample traits [11]. Briefly, WGCNA can identify clusters (modules) of highly co-expressed genes as well as screening of hub genes and modules related to clinical traits. Correlation networks facilitate network based gene screening methods that can be used to identify candidate biomarkers or therapeutic targets. WGCNA has been widely applied to screen hub genes and modules related to clinical traits in cancers, including bladder cancer [12, 13, 14, 15].

Protein-protein interactions (PPIs), commonly understood as physical contacts with molecular docking between proteins that occur in a cell or in a living organism in vivo [16], are emerging as an attractive class of molecular targets for treatment [17]. STRING (Search Tool for the Retrieval of Interacting Genes), a biological database of known and predicted protein-protein interactions, can construct a protein-protein interaction (PPI) network, in which the proteins are the nodes and the interactions are the edges [18]. Hubs that are “highly connected” in a PPI network are more likely to be essential proteins [19].

Thus, in the present study, we attempted to construct the weighted co-expression network as well as the PPI network in order to identify biomarkers significantly associated with the progression and prognosis of bladder cancer and to distinguish NMIBC and MIBC.

2. Materials and methods

2.1 Data collection

A flow diagram of this study protocol is shown in Fig. 1. Gene expression profiles and related clinical data were downloaded from Gene Expression Omnibus (GEO) database (http://www.ncbi.nlm.nih.gov/geo/). Dataset GSE83586, which includes 307 bladder cancer samples, was used as a training set to construct co-expression networks in this study, and six samples without complete clinical data were excluded from subsequent analysis. Another independent dataset, GSE13507, was also downloaded, of which 165 primary bladder cancer samples were used as a test set to validate our results.

Figure 1.

The flow diagram of this study. Data preparing, processing, analysis and validation were shown in the flow diagram.

2.2 Data preprocessing

Normalized data were downloaded from the GEO database. Cluster analysis was conducted according to the distance between different samples in Pearson’s correlation matrices and average linkage to assess microarray quality.

2.3 Differentially expressed genes (DEGs) screening

The “limma” [20] package in R software was used to screen for DEGs between NMIBC samples and MIBC samples. FDR (false discovery rate) $<$ 0.05 and $|$ log2 fold change (FC) $|$ $>$ 0.263 were chosen as the cut-off criterion.

2.4 Weighted co-expression network construction and module detection

The “WGCNA” package in R was used to construct a DEGs co-expression network. A Pearson’s correlation coefficient matrix was then computed for all pair-wise genes. Next, a weighted adjacency matrix was constructed using the power function $a_{ij}=s_{ij}^{\beta}$ ( $\beta$ is a soft-thresholding parameter that emphasized high correlations at the expense of low correlations). Subsequently, the adjacency was transformed into a topological overlap matrix (TOM) and the corresponding dissimilarity (1-TOM) was also calculated. Next, average linkage hierarchical clustering was performed according to the TOM-based dissimilarity measure. In this study, we chose a minimum module size of 30 for the genes dendrogram and some highly similar modules with the correlation higher than 0.8 were merged together. It must be noted that the grey module contains a set of unassigned genes that do not belong to any of the modules.

2.5 Functional and pathway enrichment analysis

The “clusterProfiler” [21] package in R was used to conduct gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis of genes in each module. False discovery rate (FDR) $<$ 0.05 was set as the cut-off criterion.

2.6 Identification of relevant module by module significance

After relating modules to clinical traits, Module Significance (the correlation between the module eigengene and the trait) was calculated. In general, the module with the absolute Module Significance ranked first or second among all the selected modules was considered as the one related to clinical trait. Thus, by comparing the Module Significance, we could identify which module was highly related to a certain clinical trait.

2.7 Identification of hub genes in relevant module

In this study, hub genes in weighted co-expression network were defined by Module Membership (the correlation between the gene expression profile and the module eigengene) $>$ 0.8 and Gene Significance (the correlation between the gene and the trait) $>$ 0.2. Furthermore, all genes in the relevant module were uploaded to the STRING database [22] in order to construct a protein-protein interaction network, and hub genes represented genes with a large degree and were hypothesized to play a key role in the network. Genes with a connectivity degree of $\geqslant$ 22 were defined as hub genes in the PPI network in this study. Common genes identified in both weighted co-expression network and PPI network were regarded as hub genes that were closely correlated with bladder cancer progression.

Figure 2.

Determination of soft-thresholding power in the weighted gene co-expression network analysis (WGCNA) and identification of modules associated with clinical traits. (A) The left panel shows the scale-free fit index for various soft-thresholding powers ( $\beta$ ). The left panel shows the mean connectivity for various soft-thresholding powers. (B) Dendrogram of all differentially expressed genes clustered based on a dissimilarity measure (1-TOM). (C) Distribution of average gene significance and errors in the modules associated with the tumor stage of bladder cancer.

2.8 Survival analysis and validation of candidate biomarkers

Potential candidate biomarkers for bladder cancer were identified using overall survival analyses of the hub genes based on TCGA data in the GEPIA database [23] (http://gepia.cancer-pku.cn). After this, the test set GSE13507 was analyzed to compare expressions of candidate biomarkers between different pathological stages of bladder cancer, and the statistical significance was analyzed by one-way ANOVA. Moreover, the receiver operating characteristic (ROC) curves of these genes were also plotted based on microarray data from GSE13507. Lastly, the database, GEPIA, was used to validate expressions of candidate genes between different pathological stages of bladder cancer.

2.9 Gene set enrichment analysis (GSEA)

The test set GSE13507 was also used to conduct GSEA [24] (http://software.broadinstitute.org/gsea/ index.jsp) of hub genes. To this end, the 165 bladder cancer samples in this dataset were divided into two groups according to the median expression of their hub genes (high vs. low). Annotated gene sets c2.cp.kegg.v6.1.symbols.gmt was chosen as the reference gene sets. Cut-offs of FDR $<$ 0.05 and gene limits $\geqslant$ 80 were set.

Figure 3.

Overall survival analyses of the hub genes in bladder cancer (based on TCGA data in GEPIA). (A) COL3A1, (B) FN1, (C) COL5A1, (D) FBN1, (E) COL6A1, and (F) THBS2.

3. Results

3.1 DEGs screening, co-expression network construction and relevant module identification

After data preprocessing and quality assessment, expression matrices were obtained from the training set GSE83586. Under the threshold of FDR $<$ 0.05 and $|$ log2FC $|$ $>$ 0.263, a total of 611 DEGs (275 up-regulated and 366 down-regulated) were selected to construct a co-expression network (Supplementary Table 1). Three hundred and one samples with related clinical data were included in co-expression analysis (Supplementary Fig. 1). We used the “WGCNA” package in R to construct a DEGs co-expression network. In this study, a power of $\beta=$ 9 was selected as the soft-thresholding (Fig. 2A) and a total of three real modules were subsequently identified (Fig. 2B). The modules were correlated with tumor stage, and the blue module was identified as the relevant module by comparing the Module Significances between modules (Fig. 2C).

3.2 Functional and pathway enrichment analysis

Mechanisms correlated with the progression of bladder cancer were explored with GO and KEGG pathway enrichment analysis. GO and KEGG pathway enrichment analysis results showed that DEGs in the blue module were enriched in GO terms and pathways associated with the ECM (Supplementary Tables 2 and 3); DEGs in the brown module were enriched in GO terms and pathways associated with immunity (Supplementary Tables 4 and 5); DEGs in the turquoise module were enriched in GO terms and pathways associated with fatty acid metabolism (Supplementary Tables 6 and 7).

3.3 Identification of hub genes in relevant module

Based on Gene Significance $>$ 0.2 and Module Membership $>$ 0.8, there were 60 hub genes identified within our weighted co-expression network (Supplementary Table 8). As to the PPI network, hub genes represented genes with a large degree and were hypothesized to play a key role in the network. Under the cut-off of connectivity degree $\geqslant$ 22, 14 genes were identified as hub genes in the PPI network (Supplementary Table 9). Eventually, as shown in Table 1, 10 common genes were identified in both the weighted co-expression network and PPI network. These genes were considered to be hub genes closely correlated with bladder cancer progression and were selected for further analysis.

Table 1
Ten hub genes that were closely correlated with bladder cancer progression

Symbol	Gene	Module	Connectivity
	significance	membership	degree
COL1A1	0.356938	0.878407	45
COL3A1	0.330875	0.899094	39
MMP2	0.251273	0.858248	32
FN1	0.416872	0.845032	32
SPARC	0.304999	0.898514	29
COL5A2	0.346804	0.933838	26
COL5A1	0.33074	0.932825	24
FBN1	0.271694	0.877338	23
COL6A1	0.278278	0.891926	22
THBS2	0.329662	0.826138	22

Figure 4.

Boxplots of candidate biomarkers across different pathological stages in the test set GSE13507. The boxplots show the medians and dispersions of the samples of different pathological stages for each candidate biomarker. (A) COL3A1, (B) FN1, (C) COL5A1, (D) FBN1, (E) COL6A1, and (F) THBS2. $p$ -values are the results of independent sample $t$ -test between pathological stage I and II/III/IV and one-way ANOVA for different pathological stages.

3.4 Survival analysis and validation of candidate biomarkers

In order to identify potential candidate biomarkers for bladder cancer, overall survival analyses were performed based on TCGA data. As shown in Fig. 3, higher expressions of six hub genes (COL3A1, FN1, COL5A1, FBN1, COL6A1 and THBS2) were significantly associated with a worse overall survival, and these six hub genes were chosen as candidate biomarkers for further validation. In the test set, GSE13507, significant differences were detected for each candidate biomarker expression across the pathological stages of bladder cancer and all these genes were effective for distinguishing NMIBC and MIBC (Fig. 4). Moreover, ROC curves also indicated that all candidate biomarkers were effective for distinguishing NMIBC and MIBC (Fig. 5). In addition, based on TCGA data, we also identified significant differences for all candidate biomarkers among different pathological stages of bladder cancer (Fig. 6).

Figure 5.

The receiver operating characteristic (ROC) curves of candidate biomarkers based on the test set GSE13507. (A) COL3A1, (B) FN1, (C) COL5A1, (D) FBN1, (E) COL6A1, and (F) THBS2. Areas under the curve (AUC) statistics are used to evaluate the diagnostic efficiency of the hub genes to distinguish between NMIBC and MIBC.

Figure 6.

Validation of the correlation between the expression levels of candidate biomarkers and the pathologic stage of bladder cancer (based on TCGA data in GEPIA). (A) COL3A1, (B) FN1, (C) COL5A1, (D) FBN1, (E) COL6A1, and (F) THBS2.

3.5 Gene set enrichment analysis (GSEA)

To characterize the potential mechanisms of influence on overall survival of the above six candidate biomarkers, GSEA was conducted. Using the cut-off criterion FDR $<$ 0.05 and gene limits $\geqslant$ 80, 10 gene sets were commonly enriched in the samples with all these genes highly expressed (Supplementary Table 10). Within the gene set enrichment analysis, the gene set associated with “ECM-receptor interaction” was most significantly enriched.

4. Discussion

Bladder cancer is one of the most common cancers with a high rate of recurrence and progression. The prognosis for patients with bladder cancer, especially muscle-invasive bladder cancer, remains poor. Therefore, it is essential to understand the potential mechanisms correlated with the progression of bladder cancer and identify more effective biomarkers for this malignancy.

WGCNA is a recently developed method for the analysis of gene expression data able to identify candidate biomarkers or therapeutic targets. As for bladder cancer, WGCNA has been successfully applied to screen hub genes in studies [25]. Yuan et al. found that over-expression of COL3A1 referred a poor prognosis in bladder cancer via co-expression analysis [14]. Li et al. identified several candidate biomarkers correlated with the TNM staging and overall survival of BC patients using WGCNA [15]. Similarly, Gaballah revealed several gene modules and hub genes associated with the tumor stage of bladder cancer [26]. In addition, Zhang et al. performed single-cell RNA-seq on human squamous cell carcinoma of the bladder and uncovered some hub genes by the WGCNA method [27]. Deng et al. also detected hub genes associated with bladder cancer by constructing differential co-expression networks [28]. In this study, we attempted to perform weighted co-expression network analysis as well as PPI network analysis in order to identify biomarkers significantly associated with the progression and prognosis of bladder cancer and to distinguish NMIBC and MIBC. Eventually, 10 common genes identified in both the weighted co-expression network and PPI network were considered hub genes for further analysis. Our survival analyses of the hub genes indicated that enhanced expressions of six hub genes (COL3A1, FN1, COL5A1, FBN1, COL6A1 and THBS2) were significantly associated with worse overall survival. Furthermore, these survival-associated hub genes were validated in association with the progression of bladder cancer and could distinguish NMIBC and MIBC.

COL3A1, COL5A1 and COL6A1 are members of the collagen family. Collagens are the most abundant components of extracellular matrix (ECM); they provide the structural integrity and tensile strength of human tissues and organs [29]. In the context of cancer biology, collagens regulate the physical and the biochemical properties of the tumor microenvironment, which modulate cancer cell polarity, migration and signaling [30]. COL3A1 encodes the pro-alpha1 chain of type III collagen, a fibrillar collagen that is found in extensible connective tissues. Some studies have reported that COL3A1 could be targeted to regulate cell migration and invasion as well as influence cancer progression [31, 32]. Similarly, COL5A1 encodes an alpha chain of type V collagen, which is found in tissues containing type I collagen and appears to regulate the assembly of heterotypic fibers composed of both type I and type V collagen. Cheon et al. found that COL5A1 was regulated by TGF-beta 1 signaling which could promote metastasis and contribute to poor overall survival for patients with serous ovarian cancer [31, 32]. COL6A2 encodes the alpha 1 subunit of type VI collagen as major structural component of microfibrils. COL6A1 over-expression has been found to be a metastasis-associated gene in cancers [34, 35]. FN1 encodes fibronectin, which is involved in cell adhesion and migration processes, including embryogenesis, wound healing, blood coagulation, host defense, and metastasis. Xu et al. reported that over-expression of FENDER, a long non-coding RNAs that could bind to polycomb repressive complexe 2 and/or Trithorax group/MLL protein complexes, suppressed invasion and migration of gastric cancer cells in vitro by down-regulating FN1 and MMP2/MMP9 expression [36]. FBN1 also encodes a member of the fibrillin protein family. Wang et al. demonstrated that FBN1 could promote ovarian cancer metastasis through the p53 and SLUG-associated signaling [37]. Furthermore, several studies have demonstrated that FBN1 could serve as a cancer biomarker [38, 39]. The protein encoded by THBS2 belongs to the thrombospondin family and may modulate cell surface properties of mesenchymal cells and be involved in cell adhesion and migration. Cheon et al. reported that THBS2 could promote metastasis in patients with serous ovarian cancer contribute to poor OS in patients with serous ovarian cancer [33]. Finally, Hsu et al. also revealed that THBS2 could be a potentially useful salivary marker for the detection of oral cavity squamous cell carcinoma [40].

GSEA was employed to explore the potential mechanisms influencing overall survival of the above 6 candidate biomarkers. Our results indicated that bladder cancer samples with relatively increased expression of these genes were most significantly enriched in gene set associated with “ECM-receptor interaction”. In summary, we can assume that these six hub genes influence the progression and prognosis of bladder cancer probably by regulating the tumor microenvironment.

Finally, highly co-expressed genes were connected and grouped into modules by WGCNA, thus, different modules included different functionally related genes. In order to better define the potential mechanisms correlated with the progression of bladder cancer, we performed functional and pathway enrichment analysis of genes in each module. Functional and pathway enrichment analysis indicated that DEGs between NMIBC and MIBC were enriched in GO terms and pathways associated with the ECM, immunity and fatty acid metabolism. The ECM is a complex meshwork composed of approximately 300 proteins that are important regulators of cell proliferation, survival, differentiation, and migration [41, 42]. However, ECM is commonly deregulated and becomes disorganized in cancers, which leads to generation of a tumorigenic microenvironment and affects cancer progression [43, 44, 45, 46]. Furthermore, cancer development and its response to therapy are strongly influenced by innate and adaptive immunity, which either promote or attenuate tumorigenesis and can have opposing effects on the therapeutic outcome. Interestingly, antitumor immune responses activated by tumor-specific antigens and tumor-associated antigens could fight tumors, while chronic inflammation can promote tumor development, progression, and metastatic dissemination, as well as treatment resistance [47]. Immune checkpoint therapy, which could enhance antitumor immune responses, has led to important clinical advances and provided a new weapon against cancer [48]. As to the fatty acid metabolism, epidemiologic studies reported that the high level of dietary fatty acids intake had been implicated in the development of several human malignancies [49]. There is increasing evidence that cancer cells show specific alterations of fatty-acid metabolism, which can affect numerous cellular processes, including cell growth, proliferation, differentiation and motility [50]. In addition, fatty-acid metabolism is now associated with clinically aggressive tumor behavior and tumor-cell growth and survival, which may offer novel therapeutic strategies for cancer treatment [51]. Thus, we can assume that the ECM, immunity and fatty acid metabolism play important roles in bladder cancer progression.

5. Conclusion

This study used weighted gene co-expression analysis to construct a gene co-expression network, identify and validate network hub genes associated with the progression and poor prognosis of bladder cancer. Eventually, six potential biomarkers, including COL3A1, FN1, COL5A1, FBN1, COL6A1 and THBS2, were identified and validated in association with the progression and poor prognosis of bladder cancer probably by regulating the tumor microenvironment. However, further studies are still needed to confirm these observations.

Footnotes

Acknowledgments

The free use of GEO, GEPIA and TCGA databases were acknowledged.

Supplementary data

The supplementary files are available to download from http://dx.doi.org/10.3233/CBM-181940.

References

Siegel

R.L.

Miller

K.D.

and Jemal

, Cancer statistics, 2018, CA Cancer J Clin 68 (2018), 7–30.

Torre

L.A.

Bray

Siegel

R.L.

Ferlay

Lortet-Tieulent

and Jemal

, Global cancer statistics, 2012, CA Cancer J Clin 65 (2015), 87–108.

Chen

Zheng

Baade

P.D.

Zhang

Zeng

Bray

Jemal

X.Q.

and He

, Cancer statistics in China, 2015, CA-A Cancer Journal for Clinicians 66 (2016), 115–132.

Burger

Catto

J.W.F.

Dalbagni

Grossman

H.B.

Herr

Karakiewicz

Kassouf

Kiemeney

L.A.

La Vecchia

Shariat

and Lotan

, Epidemiology and risk factors of urothelial bladder cancer, European Urology 63 (2013), 234–241.

Babjuk

Bohle

Burger

Capoun

Cohen

Comperat

E.M.

Hernandez

Kaasinen

Palou

Roupret

van Rhijn

B.W.

Shariat

S.F.

Soukup

Sylvester

R.J.

and Zigeuner

, EAU guidelines on non-muscle-invasive urothelial carcinoma of the bladder: update 2016, Eur Urol 71 (2017), 447–461.

Chang

S.S.

Boorjian

S.A.

Chou

Clark

P.E.

Daneshmand

Konety

B.R.

Pruthi

Quale

D.Z.

Ritch

C.R.

Seigne

J.D.

Skinner

E.C.

Smith

N.D.

and McKiernan

J.M.

, Diagnosis and treatment of non-muscle invasive bladder cancer: AUA/SUO guideline, J Urol 196 (2016), 1021–1029.

Kamat

A.M.

Hahn

N.M.

Efstathiou

J.A.

Lerner

S.P.

Malmstrom

Choi

Guo

C.C.

Lotan

and Kassouf

, Bladder cancer, Lancet 388 (2016), 2796–2810.

Witjes

J.A.

Lebret

Comperat

E.M.

Cowan

N.C.

De Santis

Bruins

H.M.

Hernandez

Espinos

E.L.

Dunn

Rouanne

Neuzillet

Veskimae

van der Heijden

A.G.

Gakis

and Ribal

M.J.

, Updated 2016 EAU guidelines on muscle-invasive and metastatic bladder cancer, European Urology 71 (2017), 462–475.

Youssef

R.F.

and Lotan

, Predictors of outcome of non-muscle-invasive and muscle-invasive bladder cancer, Scientific World Journal 11 (2011), 369–381.

10.

Stratton

M.R.

Campbell

P.J.

and Futreal

P.A.

, The cancer genome, Nature 458 (2009), 719–724.

11.

Langfelder

and Horvath

, WGCNA: an R package for weighted correlation network analysis, BMC Bioinformatics 9 (2008), 559.

12.

Giulietti

Righetti

Principato

and Piva

, LncRNA co-expression network analysis reveals novel biomarkers for pancreatic cancer, Carcinogenesis (2018).

13.

Zhou

X.G.

Huang

X.L.

Liang

S.Y.

Tang

S.M.

S.K.

Huang

T.T.

Z.N.

and Wang

Q.Y.

, Identifying miRNA and gene modules of colon cancer associated with pathological stage by weighted gene co-expression network analysis, Onco Targets Ther 11 (2018), 2815–2830.

14.

Yuan

Shu

Chen

Qian

Wang

Qian

Zhu

Cao

Xie

Xiao

and Wang

, Overexpression of COL3A1 confers a poor prognosis in human bladder cancer identified by co-expression analysis, Oncotarget 8 (2017), 70508–70520.

15.

Liu

Meng

Yin

Fang

Huang

Cao

Weng

Zeng

and Wang

, Identification of biomarkers correlated with the TNM staging and overall survival of patients with bladder cancer, Frontiers in Physiology 8 (2017).

16.

De Las Rivas

and Fontanillo

, Protein-protein interactions essentials: Key concepts to building and analyzing interactome networks, Plos Computational Biology 6 (2010).

17.

Zinzalla

and Thurston

D.E.

, Targeting protein-protein interactions for therapeutic intervention: a challenge for the future, Future Medicinal Chemistry 1 (2009), 65–93.

18.

Szklarczyk

Morris

J.H.

Cook

Kuhn

Wyder

Simonovic

Santos

Doncheva

N.T.

Roth

Bork

Jensen

L.J.

and von Mering

, The STRING database in 2017: quality-controlled protein-protein association networks, made broadly accessible, Nucleic Acids Res 45 (2017), D362–D368.

19.

Jeong

Mason

S.P.

Barabasi

A.L.

and Oltvai

Z.N.

, Lethality and centrality in protein networks, Nature 411 (2001), 41–42.

20.

Ritchie

M.E.

Phipson

Law

C.W.

Shi

and Smyth

G.K.

, limma powers differential expression analyses for RNA-sequencing and microarray studies, Nucleic Acids Res 43 (2015), e47.

21.

Wang

Han

and He

, clusterProfiler: an R package for comparing biological themes among gene clusters, Omics-A Journal of Integrative Biology 16 (2012), 284–287.

22.

Szklarczyk

Franceschini

Wyder

Forslund

Heller

Huerta-Cepas

Simonovic

Roth

Santos

Tsafou

K.P.

Kuhn

Bork

Jensen

L.J.

and von Mering

, STRING v10: protein-protein interaction networks, integrated over the tree of life, Nucleic Acids Res 43 (2015), D447–D452.

23.

Tang

Kang

Gao

and Zhang

, GEPIA: a web server for cancer and normal gene expression profiling and interactive analyses, Nucleic Acids Research 45 (2017), W98–W102.

24.

Subramanian

Tamayo

Mootha

V.K.

Mukherjee

Ebert

B.L.

Gillette

M.A.

Paulovich

Pomeroy

S.L.

Golub

T.R.

Lander

E.S.

and Mesirov

J.P.

, Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles, Proc Natl Acad Sci U S A 102 (2005), 15545–15550.

25.

Giulietti

Occhipinti

Righetti

Bracci

Conti

Ruzzo

Cerigioni

Cacciamani

Principato

and Piva

, Emerging biomarkers in bladder cancer identified by network analysis of transcriptomic data, Frontiers in Oncology 8 (2018).

26.

Gaballah

H.H.

, Integration of gene coexpression network, GO enrichment analysis for identification gene expression signature of invasive bladder carcinoma, Transcriptomics: Open Access 4 (2016).

27.

Zhang

Hou

Zou

Liu

and Wu

, Single-cell analyses of transcriptional heterogeneity in squamous cell carcinoma of urinary bladder, Oncotarget 7 (2016), 66069–66076.

28.

Deng

Zhu

and Huang

, Mining the bladder cancer-associated genes by an integrated strategy for the construction and analysis of differential co-expression networks, BMC Genomics 163 (2015).

29.

van der Rest

and Garrone

, Collagen family of proteins, Faseb J 5 (1991), 2814–2823.

30.

Gilkes

D.M.

Semenza

G.L.

and Wirtz

, Hypoxia and the extracellular matrix: drivers of tumour metastasis, Nat Rev Cancer 14 (2014), 430–439.

31.

Qiu

Sun

Deng

Guo

Cao

Wang

Zou

Zhang

Jing

Ling

Xie

and Zhang

, miR-29a/b enhances cell migration and invasion in nasopharyngeal carcinoma progression by regulating SPARC and COL3A1 gene expression, Plos One 10 (2015).

32.

Zhao

Shi

Zhang

Xie

Guo

Zhang

Liu

Shen

Wang

Zhang

and Zhou

, Let-7d suppresses growth, metastasis, and tumor macrophage infiltration in renal cell carcinoma by targeting COL3A1 and CCL7, Molecular Cancer 13 (2014).

33.

Cheon

Tong

Sim

Dering

Berel

Cui

Lester

Beach

J.A.

Tighiouart

Walts

A.E.

Karlan

B.Y.

and Orsulic

, A collagen-remodeling gene signature regulated by TGF-beta signaling is associated with metastasis and poor survival in serous ovarian cancer, Clinical Cancer Research 20 (2014), 711–723.

34.

Chiu

Chang

Lee

and Liao

, Quantitative secretome analysis reveals that COL6A1 is a metastasis-associated protein using stacking gel-aided purification combined with iTRAQ labeling, Journal of Proteome Research 10 (2011), 1110–1125.

35.

Blanco

M.A.

LeRoy

Khan

Aleckovic

Zee

B.M.

Garcia

B.A.

and Kang

, Global secretome analysis identifies novel mediators of bone metastasis, Cell Research 22 (2012), 1339–1355.

36.

Huang

Xia

Liu

Sun

Yin

Chen

Han

Zhang

Kong

and Shu

, Decreased expression of the long non-coding RNA FENDRR is associated with poor prognosis in gastric cancer and FENDRR regulates gastric cancer cell metastasis by affecting fibronectin1 expression, Journal of Hematology & Oncology 7 (2014).

37.

Wang

Liu

Yang

Yin

Wang

Meng

Zang

and Yang

, Fibrillin-1, induced by Aurora-A but inhibited by BRCA2, promotes ovarian cancer metastasis, Oncotarget 6 (2015), 6670–6683.

38.

Zhang

Ota

Shridhar

Chien

and Kuang

, Network-based survival analysis reveals subnetwork signatures for predicting outcomes of ovarian cancer treatment, Plos Computational Biology 9 (2013).

39.

Guo

Song

Zhang

Xia

and Dang

, Detection of hypermethylated fibrillin-1 in the stool samples of colorectal cancer patients, Medical Oncology 30 (2013).

40.

Hsu

Peng

Liu

Chang

and Wu

, Secretome profiling of primary cells reveals that THBS2 is a salivary biomarker of oral cavity squamous cell carcinoma, Journal of Proteome Research 13 (2014), 4796–4807.

41.

Naba

Clauser

K.R.

Hoersch

Liu

Carr

S.A.

and Hynes

R.O.

, The matrisome: in silico definition and in vivo characterization by proteomics of normal and tumor extracellular matrices, Mol Cell Proteomics 11 (2012), M111.014647.

42.

Hynes

R.O.

, The extracellular matrix: Not just pretty fibrils, Science 326 (2009), 1216–1219.

43.

Chang

H.Y.

Nuyten

D.S.

Sneddon

J.B.

Hastie

Tibshirani

Sorlie

Dai

Y.D.

Van’T

V.L.

Bartelink

van de Rijn

Brown

P.O.

and van de Vijver

M.J.

, Robustness, scalability, and integration of a wound-response gene expression signature in predicting breast cancer survival, Proc Natl Acad Sci U S A 102 (2005), 3738–3743.

44.

Chang

H.Y.

Sneddon

J.B.

Alizadeh

A.A.

Sood

West

R.B.

Montgomery

Chi

J.T.

van de Rijn

Botstein

and Brown

P.O.

, Gene expression signature of fibroblast serum response predicts human cancer progression: similarities between tumors and wounds, PLoS Biol 2 (2004), E7.

45.

Gilkes

D.M.

Semenza

G.L.

and Wirtz

, Hypoxia and the extracellular matrix: drivers of tumour metastasis, Nat Rev Cancer 14 (2014), 430–439.

46.

Weaver

V.M.

and Werb

, The extracellular matrix: A dynamic niche in cancer progression, Journal of Cell Biology 196 (2012), 395–406.

47.

Shalapour

and Karin

, Immunity, inflammation, and cancer: an eternal fight between good and evil, Journal of Clinical Investigation 125 (2015), 3347–3355.

48.

Sharma

and Allison

J.P.

, The future of immune checkpoint therapy, Science 348 (2015), 56–61.

49.

La Vecchia

de Groh

Negri

Morrison

and Mery

, Dietary transfatty acids and cancer risk, Eur J Cancer Prev 20 (2011), 530–538.

50.

Santos

C.R.

and Schulze

, Lipid metabolism in cancer, Febs J 279 (2012), 2610–2623.

51.

Kuhajda

F.P.

, Fatty-acid synthase and human cancer: New perspectives on its role in tumor biology, Nutrition 16 (2000), 202–208.

Identification of biomarkers associated with progression and prognosis in bladder cancer via co-expression analysis

Abstract

BACKGROUND:

OBJECTIVE:

METHODS:

RESULTS:

CONCLUSIONS:

Keywords

1. Introduction

2. Materials and methods

2.1 Data collection

2.3 Differentially expressed genes (DEGs) screening

2.4 Weighted co-expression network construction and module detection

2.5 Functional and pathway enrichment analysis

2.6 Identification of relevant module by module significance

2.7 Identification of hub genes in relevant module

2.9 Gene set enrichment analysis (GSEA)

3.1 DEGs screening, co-expression network construction and relevant module identification

3.2 Functional and pathway enrichment analysis

3.3 Identification of hub genes in relevant module

Table 1 Ten hub genes that were closely correlated with bladder cancer progression

4. Discussion

5. Conclusion

Footnotes

Acknowledgments

Supplementary data

References

Table 1
Ten hub genes that were closely correlated with bladder cancer progression