Journal of Cancer Research and Therapeutics

ORIGINAL ARTICLE
Year
: 2018  |  Volume : 14  |  Issue : 7  |  Page : 1638--1643

Probing pathway-related modules in invasive squamous cervical cancer based on topological centrality of network strategy


Xiu-Hua Fu1, Yu-Fang Wu2, Fang Xue3,  
1 Department of Obstetrics, Binzhou People's Hospital, Binzhou 256610, China
2 Department of Gynecology, Binzhou People's Hospital, Binzhou 256610, China
3 Department of Gynecology and Obstetrics, Jinan Maternity and Child Care Hospital, Jinan 250001, P.R. China

Correspondence Address:
Fang Xue
Department of Gynecology and Obstetrics, Jinan Maternity and Child Care Hospital, No. 2 Jianguo Xiaojingsan Road, Jinan 250001
P.R. China

Abstract

Objective: Our work aimed to identify pathway-related modules and hub genes involved in invasive squamous cervical cancer (SCC) based on topological centralities analysis of networks. Materials and Methods: To determine the functional modules changed in SCC, functional enrichment analyses were performed for differentially expressed genes (DEGs) between invasive SCC samples and normal controls. Then, co-expression network was constructed using EBcoexpress approach based on the DEGs. Moreover, pathway-related modules were probed from the global co-expression network based on pathway genes and their adjacent genes. Finally, topological centralities for co-expression network and pathway-related subnetworks were carried out to explore hub genes and significant pathway-related functional modules. Results: Functional analyses revealed that DEGs mainly involved in three biological processes (metabolic process, cellular process, and cellular component organization) and 8 significant pathways. Furthermore, the co-expression network with 659 nodes and 1087 edges and 8 pathway-related modules were obtained. Topological centralities indicated two significant modules (cell cycle and base excision repair pathway-related modules), in which the common hub gene ARFGAP3 showed the most significant importance. Conclusions: The bioinformatics elucidation of certain pathway-related modules and hub genes might be beneficial to understand the molecular pathogenesis and reveal their potential as novel molecular markers of SCC to a great extent.



How to cite this article:
Fu XH, Wu YF, Xue F. Probing pathway-related modules in invasive squamous cervical cancer based on topological centrality of network strategy.J Can Res Ther 2018;14:1638-1643


How to cite this URL:
Fu XH, Wu YF, Xue F. Probing pathway-related modules in invasive squamous cervical cancer based on topological centrality of network strategy. J Can Res Ther [serial online] 2018 [cited 2022 Sep 29 ];14:1638-1643
Available from: https://www.cancerjournal.net/text.asp?2018/14/7/1638/187352


Full Text



 Introduction



Cervical cancer is the second most common incident cancer and the third leading cause of cancer-related death among women worldwide.[1] More than 90% of patients with cervical cancer belong to squamous cervical cancer (SCC).[2] Previous studies mainly focused on the individual genes related to SCC. With the rapid development of high-throughput technology, numerous protein interactions have been screened out, whereas a nice bit of significant and important interactions have not been tested such as vital genes in certain significant pathways.[3] Mining and analyzing functional modules of the comprehensive networks or subnetworks could contribute to the resolution of this type of difficulty.[4] Network analysis actually brings a great convenience for studying unknown connections for incomplete networks. Moreover, functional modules might integrate the most closely related proteins or genes through their interactions. Thus, in this study, we identified the functional modules from the differential co-expression network based on gene expression levels and pathways involved in SCC and defined the modules as pathway-related modules. Our study might shed light on the understanding of pathogenesis and therapy of SCC.

 Materials and Methods



Data source

In this study, the microarray expression profile of human SCC and normal squamous cervical epithelial samples was recruited from ArrayExpress database (http://www.ebi.ac.uk/arrayexpress/), under the access number of E-GEOD-7803 which was contributed by Zhai et al.[5] This microarray was based on the platform of the GPL96 (A-AFFY-33-Affymetrix GeneChip Human Genome HG-U133A [HG-U133A]). In the dataset, there were a total of 41 samples including 10 normal squamous cervical epithelial samples, 7 high-grade squamous intraepithelial lesions, 21 invasive SCC samples, and 3 additional test samples. In this study, only 10 normal samples and 21 invasive SCC samples were recruited for further analysis.

The annotations for probes were obtained from the manufacturer documentation, and the original information on all conditions was subjected to data preprocessing. Background correction, quantiles data normalization, and probe match values of datasets were conducted by the robust multiarray average algorithm.[6] Then, the data were screened by the feature filter method. Each probe was mapped to 1 gene, and a probe was discarded if it did not match any gene.

Differentially expression analysis

Since differentially expressed genes (DEGs) are highly associated with the pathogenesis of disease, the Limma package was employed to determine DEGs between invasive SCC patients and normal controls.[7] First, the expression estimates were performed for the raw data, and then the datasets were constructed to the linear model. Only the genes with the fold change value |logFC| larger than 1.5 and P < 0.05 were referred to as DEGs.

Functional analysis of differentially expressed genes

Gene ontology (GO) analysis was performed using BiNGO plugin coming from the popular open source tool Cytoscape for the DEGs. GO terms were picked out with an adjusted P < 0.01 calculated by Expression Analysis Systemic Explorer test. In addition, to understand the biochemistry pathways involved in pathogenesis of invasive SCC, a recognized and comprehensive Kyoto Encyclopedia of Genes and Genomes (KEGG) database was utilized for the pathway enrichment analysis.[8] In this study, an online tool DAVID was employed to perform the KEGG enrichment analysis.[9] KEGG pathways under the threshold of gene count >2 and P < 0.01 were considered as significant terms.

Co-expression network construction

The incorporated co-expression network presents superior ability in describing the pairwise gene–gene relationships and brings a great convenience for researching their function.[10] In our work, a package named EBcoexpress was successfully utilized for performing various aspects of correlation analysis of DEGs in SCC. A powerful biological graph visualization tool, Cytoscape software was applied to construct the co-expression network.[11] In the network, the nodes correspond to genes and correlations strength between two genes are denoted as edges.

Pathway-related modules detection

A major goal of network analysis is to identify functional modules. In our work, pathway-related modules were obtained from the co-expression network to probe significant genes and modules which may play a vital role in the development and progression of SCC. To achieve this objective, we entailed genes in the significant pathways to the global co-expression network and then extracted these pathway-related subnetworks consisted of pathway genes and their adjacent genes. Eventually, the pathway-related modules were proposed based on topological analysis. In this work, two algorithms were implemented to calculate the topological parameter of the pathway-related module; one summed topological parameter value of genes in the corresponding pathway-related module; one calculated the mean parameter value of genes in the corresponding pathway-related module.

Topological analysis

Topological centrality approaches were used to a complex network through quantifying centrality and connectivity in this paper. Some researchers proved that the network centrality measure could be applied to find essential and key nodes that modulated the propagation of functional influences within the co-expression network.[12] We performed centrality analysis in the complex networks on the local scale (degree), and the global scale (closeness, stress, and betweenness).

We first describe several preliminary definitions to better understand the topological centrality indexes. Let an undirected graph G = (V, E), and V represents nodes in the network and E stands for edges indicating the relationships between the actors. A path between node s and t was referred to as a sequence of edges, and the sum of the weights of edges was the length of a path. d(s, t) is the shortest path connecting s and t in G. σst denotes the number of shortest paths between s and t. σst (v) is the total number of the shortest paths between s and t passing through the vertex v.

Degree centrality

Degree centrality is the simplest topological index corresponding to the number of nodes directly connected to a given node, and thus quantifies the local topology for every node.[13] This index is very useful in static graphs, where we are interested in gaining vertices that possess the most adjacent nodes. In the network, nodes with high degree usually suggest central regulatory roles, which are defined as hub genes. In this work, genes with degree ≥15 were considered as hub genes.

Closeness centrality

Closeness centrality is also a node centrality index. The closeness of a node v can be obtained via calculating the sum of the shortest path between v and other nodes in the graph. Once this result is obtained, its reciprocal is computed. The higher values reveal a significantly positive meaning for node proximity. Closeness centrality, Cc(v), was defined as the reciprocal of the average shortest path length:

[INLINE:1]

Stress centrality

Stress centrality, another centrality index, calculates the number of all the shortest paths which pass through a node.[14] A “stressed” node means a node which is traversed by a high number of shortest paths. However, a node with high-stress value does not indicate that the node v is crucial to maintain the connection between nodes whose paths are passing through it. The stress Cs(v) was counted as:

[INLINE:2]

Betweenness centrality

Betweenness centrality is similar to the stress index but more elaborated and informative. It refers to a shortest path enumeration-based metric in graphs for determining how the neighbors of a node are interconnected and is the ratio of the node in the shortest path between two other nodes v 1 and v 2.[15] Thus, if the path is the only one connecting v 1 and v 2, the node n will gain a higher betweenness. CB(v) value ranges between 0 and 1 and is calculated as follows:

[INLINE:3]

 Results



Differentially expressed genes analysis

We used the Limma package in R to observe the DEGs between invasive SCC patients and normal controls. With the cutoff criteria of |logFC| >1.5 and P < 0.05, a total of 833 DEGs were obtained including 358 up-regulated genes and 475 downregulated genes.

Gene ontology functional enrichment and Kyoto Encyclopedia of Genes and Genomes pathway enrichment

To get further insights into the DEGs functionally, we carried out the GO functional enrichment and pathway enrichment analysis. GO analysis indicated that these DEGs mainly enriched in 120 functional terms under the criterion of P < 0.01. The relationship of GO terms under P < 1.0E-10 was shown in [Figure 1]. Apparently, DEGs were enriched in three biological processes including metabolic process, cellular process, and cellular component organization.{Figure 1}

Moreover, we performed pathway enrichment analysis via mapping DEGs to the KEGG database, and the tool DAVID was used to select the obvious dysregulated pathways with P < 0.01 and gene count >2 as the threshold. According to the enrichment analysis results, these DEGs were significantly enriched in 8 pathways [Table 1]. The top 3 enriched pathways were cell cycle (P = 4.64E-17), DNA replication (P = 6.93E-08), and p53 signaling pathway (P = 8.42E-06).{Table 1}

Co-expression network construction and topological analysis

In our study, the co-expression network was built using the Cytoscape software based on 833 DEGs in SCC. There were 659 nodes representing genes and 1087 edges standing for the interaction between two co-expressed genes in the correlation co-expression network [Figure 2]. Degree centrality analysis showed that a total of 7 hub genes (TMEM97, KIF4A, MMP12, ARFGAP3, SERPINB13, PRIM1, and POT1) were identified under the criterion of degree ≥15 [Table 2].{Figure 2}{Table 2}

Pathway-related modules probing and topological analysis

Based on eight significant pathways, we obtained eight pathway-related subnetworks or modules. To further investigate the biological functions of these modules, topological centralities analyses (degree, stress, closeness, and betweenness) of these modules were performed. By calculating the sum of topological parameter, as shown in [Figure 3], module 1 possessed the highest degree of 548, closeness of 23.72, stress of 2,612,364, and betweenness of 0.026. When calculating the mean value, module 7 possessed the highest degree of 6.42, closeness of 0.275, stress of 43,985, and module 1 had the highest betweenness of 0.021 [Figure 4]. Meanwhile, module 1 had the most number of nodes. In this work, module 1 was cell cycle pathway-related module, and module 7 was base excision repair pathway-related module.{Figure 3}{Figure 4}

 Discussion



Cervical cancer has become the third leading cause of cancer-induced death in women worldwide and over 90% belongs to SCC.[1] However, the SCC pathogenesis is still unclear. Commonly, the alteration of one or more functional pathways could widely affect the normal differentiation and cause tumors.[16] In this paper, we got a total of 8 pathway-related modules and 7 hub genes by accessing centralities analyses (degree, stress, closeness, and betweenness centrality) for co-expression network and subnetwork. We identified the most two significant pathway-related modules: cell cycle pathway-related module and base excision repair pathway-related module. Moreover, the hub gene ARFGAP3 existed in both of the significant pathway-related modules indicating a more significant tendency to associate with SCC.

In the present study, two significant functional modules involved in SCC were screened, especially cell cycle functional module. Cell cycle is the series of events that take place in a cell leading to cell division and duplication. The previous study had indicated that the dysregulation of the cell cycle components might cause the cell to multiply uncontrollably, further leading to tumor formation,[17] including SCC.[18],[19] Conesa-Zamora et al. suggested that several cell cycle-related markers showed good performance in the diagnosis of SCC.[18] While these cell cycle phase-specific markers did not appear to predict disease grade, stage, or outcome.[19] Several studies indicated that cell cycle pathway and cell cycle related-molecular could be considered as the targets in cancer treatment.[20],[21],[22]

Base excision repair pathway, an evolutionarily conserved process in the eukaryotes including mammals, is responsible for maintaining the genomic integrity by eliminating most endogenous base lesions and abnormal bases. Moreover, base excision repair pathway also plays a vital role in repairing the DNA single strand break.[23] Generally, the occurrence of cancer is suspected from the accumulation of inherited and somatic mutations in oncogenes and tumor suppressor genes. Previous evidences have shown the correlation between CC risk and several oncogenes such as TLR9,[24]HPV-16,[25] and PTEN.[26] The presence of susceptibility variants of oncogenes might contribute to the significant difference of this pathway-related module. However, the significant function of base excision repair pathway in prevention of disease still remains unclear. In our paper, we put forward the idea that the base excision repair pathway may be closely involved in the incidence of ovarian cancer for the first time based on the topological centrality analysis of network strategy.

In our study, the hub gene ARFGAP3 was present in both of cell cycle and base excision repair pathway-related module, indicating a more significant tendency to associate with SCC. ADP-ribosylation factors are critical regulators of vesicular trafficking pathways and act at multiple intracellular sites. ARFGAPs, ADP-ribosylation factor-GTPase-activating proteins, are proposed to contribute to site-specific regulation. ARFGAP3 is a member of the ARFGAP family, which is essential for COPI coat assembly on the Golgi membrane and interacts with ARF1.[27],[28]ARFGAP3 has been reported to be regulated by the agonist phosphatidylinositol 4,5-diphosphate and antagonist phosphatidylcholine.[29] Moreover, ARFGAP3 plays an important role in the regulation of cellular secretion and affects cellular adhesion such as other ARFGAP family members.[28],[29] A present research shows that ARFGAP3 is a novel androgen-regulated gene and its overexpression can promote prostate cancer cell proliferation and migration in collaboration with paxillin by cell cycle progression at the G1/S phase transition.[30] Understanding how ARFGAP3 regulates proliferation and migration may lead to the development of novel therapeutic targets for cancer treatment as well as possible diagnostic markers for tumor aggressiveness.

In the light of the preliminary study, we successfully confirmed two functional modules (cell cycle and base excision repair pathway-related module) and hub gene ARFGAP3 in SCC based on topological centralities of the co-expression network. These functional modules and hub gene might be ideal underlying biomarkers for diagnosing and treating SCC.

Financial support and sponsorship

This study was supported by the Science and Technology Development Project of Jinan (200705) from Health and Family Planning Commission of Jinan Municipality.

Conflicts of interest

There are no conflicts of interest.

References

1Torre LA, Bray F, Siegel RL, Ferlay J, Lortet-Tieulent J, Jemal A. Global cancer statistics, 2012. CA Cancer J Clin 2015;65:87-108.
2Shah AS, Scheele WH, Glant MD, Fugère P. Raloxifene HCl is not stimulatory in the endometrium as assessed by the blaustein criteria and an estrogenicity scoring system. Prim Care Update Ob Gyns 1998;5:167.
3Nibbe RK, Chowdhury SA, Koyutürk M, Ewing R, Chance MR. Protein-protein interaction networks and subnetworks in the biology of disease. Wiley Interdiscip Rev Syst Biol Med 2011;3:357-67.
4Wu Y, Jing R, Jiang L, Jiang Y, Kuang Q, Ye L, et al. Combination use of protein-protein interaction network topological features improves the predictive scores of deleterious non-synonymous single-nucleotide polymorphisms. Amino Acids 2014;46:2025-35.
5Zhai Y, Kuick R, Nan B, Ota I, Weiss SJ, Trimble CL, et al. Gene expression analysis of preinvasive and invasive cervical squamous cell carcinomas identifies HOXC10 as a key mediator of invasion. Cancer Res 2007;67:10163-72.
6Irizarry RA, Hobbs B, Collin F, Beazer-Barclay YD, Antonellis KJ, Scherf U, et al. Exploration, normalization, and summaries of high density oligonucleotide array probe level data. Biostatistics 2003;4:249-64.
7Naydenov AV, Vassoler F, Luksik AS, Kaczmarska J, Konradi C. Mitochondrial abnormalities in the putamen in Parkinson's disease dyskinesia. Acta Neuropathol 2010;120:623-31.
8Kanehisa M, Goto S. KEGG: Kyoto encyclopedia of genes and genomes. Nucleic Acids Res 2000;28:27-30.
9Huang DW, Sherman BT, Tan Q, Collins JR, Alvord WG, Roayaei J, et al. The DAVID gene functional classification tool: A novel biological module-centric algorithm to functionally analyze large gene lists. Genome Biol 2007;8:R183.
10Cokus S, Rose S, Haynor D, Grønbech-Jensen N, Pellegrini M. Modelling the network of cell cycle transcription factors in the yeast Saccharomyces cerevisiae. BMC Bioinformatics 2006;7:381.
11Shannon P, Markiel A, Ozier O, Baliga NS, Wang JT, Ramage D, et al. Cytoscape: A software environment for integrated models of biomolecular interaction networks. Genome Res 2003;13:2498-504.
12Prifti E, Zucker JD, Clément K, Henegar C. Interactional and functional centrality in transcriptional co-expression networks. Bioinformatics 2010;26:3083-9.
13Haythornthwaite C. Social network analysis: An approach and technique for the study of information exchange. Libr Inf Sci Res 1996;18:323-42.
14Fekete SP, Kaufmann M, Kröller A, Lehmann K. A New Approach for Boundary Recognition in Geometric Sensor Networks. arXiv Preprint cs/0508006; 2005. Avaialble from: http://www.arxiv.org/abs/cs/0508006. [Last accessed on 2015 Aug 12].
15Barthelemy M. Betweenness centrality in large complex networks. Eur Phys J B Condens Matter Complex Syst 2004;38:163-8.
16Mohseny AB, Hogendoorn PC. Concise review: Mesenchymal tumors: When stem cells go mad. Stem Cells 2011;29:397-403.
17Champeris Tsaniras S, Kanellakis N, Symeonidou IE, Nikolopoulou P, Lygerou Z, Taraviras S. Licensing of DNA replication, cancer, pluripotency and differentiation: An interlinked world? Semin Cell Dev Biol 2014;30:174-80.
18Conesa-Zamora P, Trujillo-Santos J, Orantes-Casado FJ, Ortiz-Reina S, Torres-Moreno D, Pérez-Guillermo M. Analysis of performance characteristics of five cell cycle-related immunohistochemical markers and human papillomavirus genotyping in the diagnosis of cervical squamous cell carcinoma precursor lesions. Anal Quant Cytol Histol 2012;34:49-55.
19Kuku S, Proctor I, Loddo M, Kadalayil L, KhoshZaban M, Ledermann J, et al. Do cell-cycle phase-specific markers predict disease grade, stage, and outcome in cervical carcinoma? Int J Gynecol Cancer 2015;25:1066-72.
20Rao RN. Targets for cancer therapy in the cell cycle pathway. Curr Opin Oncol 1996;8:516-24.
21Newman DJ, Cragg GM, Holbeck S, Sausville EA. Natural products and derivatives as leads to cell cycle pathway targets in cancer chemotherapy. Curr Cancer Drug Targets 2002;2:279-308.
22Hosgood HD 3rd, Menashe I, Shen M, Yeager M, Yuenger J, Rajaraman P, et al. Pathway-based evaluation of 380 candidate genes and lung cancer susceptibility suggests the importance of the cell cycle pathway. Carcinogenesis 2008;29:1938-43.
23Gassman NR, Wilson SH. Micro-irradiation tools to visualize base excision repair and single-strand break repair. DNA Repair (Amst) 2015;31:52-63.
24Mu X, Zhao J, Yuan X, Zhao X, Yao K, Liu Y, et al. Gene polymorphisms of toll-like receptor 9 -1486T/C and 2848G/A in cervical cancer risk. Int J Gynecol Cancer 2015;25:1173-8.
25Boumba LM, Assoumou SZ, Hilali L, Mambou JV, Moukassa D, Ennaji MM. Genetic variability in E6 and E7 oncogenes of human papillomavirus type 16 from congolese cervical cancer isolates. Infect Agent Cancer 2015;10:15.
26Harima Y, Sawada S, Nagata K, Sougawa M, Ostapenko V, Ohnishi T. Mutation of the PTEN gene in advanced cervical cancer correlated with tumor progression and poor outcome after radiotherapy. Int J Oncol 2001;18:493-7.
27Kartberg F, Asp L, Dejgaard SY, Smedh M, Fernandez-Rodriguez J, Nilsson T, et al. ARFGAP2 and ARFGAP3 are essential for COPI coat assembly on the Golgi membrane of living cells. J Biol Chem 2010;285:36709-20.
28Luo JQ, Liu X, Frankel P, Rotunda T, Ramos M, Flom J, et al. Functional association between Arf and RalA in active phospholipase D complex. Proc Natl Acad Sci U S A 1998;95:3632-7.
29Liu X, Zhang C, Xing G, Chen Q, He F. Functional characterization of novel human ARFGAP3. FEBS Lett 2001;490:79-83.
30Obinata D, Takayama K, Urano T, Murata T, Ikeda K, Horie-Inoue K, et al. ARFGAP3, an androgen target gene, promotes prostate cancer cell proliferation and migration. Int J Cancer 2012;130:2240-8.