Sage Journals: Discover world-class research

Abstract

BACKGROUND:

Cholangiocarcinoma (CCA) is the most common biliary malignancy worldwide. However, the molecular mechanisms of its tumorigenesis and progression are still largely unclear.

OBJECTIVE:

This study aimed to explore the hub genes and pathways associated with CCA prognosis by coexpression analysis.

METHODS:

A coexpression network complex was constructed using the top 20% most variant genes in the GSE89748 dataset to find modules associated with prognosis related clinical trait-histology. The hub genes in the clinically significant modules were defined as candidates if they were common in both the coexpression network and protein-protein interaction (PPI) network. Afterwards, survival analysis, expression level analysis and a series of bioinformatic analysis were used to validate the hub genes.

RESULTS:

Twenty-five modules were obtained, and the cyan, light cyan and red modules regarded as closely associated with histology were selected. Subsequently, combining the PPI network complexes and coexpression networks, we screened 20 candidates. After expression and survival analysis, 10 real hub genes (LIMA1, HDAC1, ITGA3, ACTR3, GSK3B, ITGA2, THOC2, PTGES3, HEATR1 and ILF2) were finally identified. Additionally, functional enrichment analysis revealed that the hub genes were mainly enriched in cell cycle-related pathways.

CONCLUSIONS:

Overall, this study identified 10 hub genes and cell cycle-related pathways were closely related to CCA development, progression and prognosis, which may contribute to CCA diagnosis and treatment.

Keywords

Cholangiocarcinoma bioinformatics analysis weighted gene coexpression network analysis (WGCNA)prognosis

Figure 1.

Flow chart of data collection, processing, analysis and validation.

1. Introduction

Cholangiocarcinoma (CCA) is the most common biliary tract malignancy and the second most prevalent primary hepatic malignancy after hepatocellular carcinoma (HCC) [1]. CCA combined with HCC was reported to be the third leading cause of cancer-related deaths in China in 2015 [2]. Moreover, despite having a lower CCA incidence rate, the age-adjusted incidence rates of CCA have increased in the United States and Western Europe during the last several decades [3]. At present, the commonly used treatment approaches for CCA include surgical resection, chemotherapy, molecularly targeted therapies and liver transplantation [4, 5]. However, the 5-year overall survival rate of CCA is still dismal, especially in advanced CCA [6]. In addition, the current serum biomarkers used for CCA diagnosis are not sensitive or accurate, and there is a lack of knowledge about CCA development, progression and prognosis [7, 8, 9]. Consequently, a better understanding of the molecular mechanisms of CCA tumorigenesis and progression is critical and may provide new insight into CCA therapy.

Figure 2.

Clustering dendrogram of 55 samples with proper clinical data. The clustering was based on the top 20% genes in expression in GSE89748. The colour intensity was proportional to sex, age, histology, status and survival days.

Figure 3.

Determination of soft-thresholding power in the WGCNA. (A) Analysis of the scale-free fit index for various soft-thresholding powers ( $\beta$ ). (B) Analysis of the mean connectivity for various soft-thresholding powers. (C) Histogram of connectivity distribution when $\beta=$ 5. (D) Checking the scale free topology when $\beta=$ 5.

In recent years, gene expression analysis based on high-throughput platforms has been widely used for detecting significant expression changes in thousands of genes and identifying promising biomarkers for the diagnosis and therapy of cancers [10]. Some studies have found many differentially expressed genes (DEGs) that play critical roles in CCA development, progression and prognosis and could be regarded as potential diagnostic and therapeutic targets. Betesh et al. [11] identified fucosylated Fetuin-A as an effective serum marker in the surveillance of people at risk of developing of CCA. Through multiplexed mutational profiling, Zhu et al. [12] demonstrated that KRAS and IDH1 were the most common mutations noted in intrahepatic cholangiocarcinoma (ICC) and were associated with the clinicopathological features of ICC. Weighted gene coexpression network analysis (WGCNA) is a powerful tool that can be used in data exploration or to find modules containing highly correlated genes. It has been widely used for finding hub genes in many malignancies. For instance, Tang et al. [13] used WGCNA and identified five core genes, the dysregulation of which was closely associated with the progression and prognosis of breast cancer. In this study, to further understand the molecular mechanisms of CCA development, progression and prognosis, we used WGCNA to explore modules related to clinical characteristics.

2. Materials and methods

2.1 Dataset collection and processing

The workflow of this study is shown in Fig. 1. The CCA-associated dataset GSE89748 submitted by Jusakul Apinya and generated on the GPL10558 platform (Illumina HumanHT-12 V4.0 Expression BeadChip) was downloaded from the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih. gov/geo/). The dataset included data on 72 tumour samples [14]. The clinical traits of the tumour samples were obtained from the supplementary files of the study, and samples with the following histological results were excluded: Papillary carcinoma, adenosquamous carcinoma, TBA and N/A. After downloading the raw Series Matrix Files of the dataset and the Annotation SOFT table of the platform, the probe names in the raw Series Matrix Files were transferred into related gene names using Perl 5.0 (available online: http://www.perl.org/). The top 20% of the most variant genes were selected for further analysis.

Figure 4.

Identification of modules associated with the clinical features of cholangiocarcinoma. (A) Dendrogram of all the top 20% genes in expression clustered based on a dissimilarity measure (1-TOM). (B) Heatmap of the correlation between module eigengenes and clinical features of cholangiocarcinoma, the red modules represent positive correlation between the clinical traits and genes in modules, and the green modules represent negative correlation. (C) Distribution of average gene significance and errors in the modules related to histology of cholangiocarcinoma.

2.2 Coexpression network construction and identification of clinical modules

The methods used to construct coexpression networks and to identify clinically relevant modules have been described in previous studies [13, 15, 16]. The “WGCNA” package for the statistical software R (version 3.5.1, https://www.r-project.org/) was used to construct a coexpression network. After the calculation of all pairwise genes, the power of $\beta=$ 5 (scale free $R^{2}=$ 0.92) was selected (Fig. 3) to guarantee the correlations between genes and penalize weak correlations. Subsequently, to assess the network connection of a gene, the adjacency was transformed into a topological overlap matrix (TOM) [17]. Finally, according to the TOM-based dissimilarity measure, average linkage hierarchical clustering was performed to classify the genes with similar expression profiles into gene modules, the minimum size of which was 30 [18].

Modules related to clinical characteristics were identified using two approaches. First, in the principal component analysis of each gene module, module eigengenes (MEs) were defined as the first principal component of the module. All genes in a module were represented by the expression of the MEs in that module. Next, we identified the most relevant module by calculating the correlation between the MEs and clinical traits. Second, in the linear regression between clinical traits and gene expression, the gene significance (GS) was the log10 transformation of the P-value of each gene (GS $=$ lgP) and the module significance (MS) was considered as the average GS of all genes in the module. The Pearson’s correlation method and average linkage method were used to cluster the samples. The screened modules with higher absolute MS values were regarded as those more strongly correlated with clinical traits. Since histology was regarded as a prognostic factor for CCA [19, 20], the top three modules closely related to histology were selected for further analysis.

2.3 GO and KEGG pathway enrichment analysis

The molecular and functional characteristics of the genes in the hub modules were analysed using the R package “ggplot2” [21] (version 3.5.1) and the online biological tool The Database for Annotation, Visualization and Integrated Discovery (DAVID) 6.8 (available online: https://david.ncifcrf.gov/). Afterwards, the online biological tool KOBAS 3.0 (available online: http://kobas.cbi.pku.edu.cn) and the “ggplot2” R package were employed for KEGG pathway enrichment analysis of the genes. $P<$ 0.05 was considered to have statistical significance during the analysis.

Figure 5.

Gene ontology and Signaling pathway enrichment analysis. GO analysis of cyan (A), light cyan (B) and red (C) module genes. KEGG pathway enrichment analysis of cyan (D), light cyan (E) and red (F) module genes. The X-axis shows the rich factors of each term and the Y-axis shows the description of the GO and KEGG pathway terms.

2.4 Identification of hub genes

After the identification of the top three modules closely related to histology, hub genes were selected from each module using the following criteria: 1) genes with high within-module connectivity were considered hub genes; consequently, module membership $>$ 0.8 was chosen as the selection criteria; 2) since the mean GS of all three modules $>$ 0.2, genes with GS $>$ 0.2 in these modules were selected. Furthermore, using the Search Tool for the Retrieval of Interacting Genes/Protein (STRING) database (available online: http://string-db.org) and Cytoscape software (version 3.6.1), three protein-protein (PPI) network complexes were constructed to identify hub genes in the selected modules, and nodes with connectivity $>$ 2 (total edges/total nodes) were considered hub genes in the PPI network complexes [22]. The genes connected with over 5, 6 and 5 nodes were defined as hub genes in the cyan, light cyan and red PPI network complexes, respectively. Finally, the common genes from both of these methods were considered candidates for further analysis and validation.

Figure 6.

Identification of the hub genes. PPI network complexes of genes in cyan (A), light cyan (B) and red (C) modules. Scatter plot of module eigengenes in cyan (D), light cyan (E) and red (F) modules. Identification of hub gene candidates in cyan (G), light cyan (H) and red (I) modules.

2.5 Validation of the hub genes

To validate the relationships between the identified hub genes and CCA prognosis, survival analysis was performed. X-tile was employed to find the optimal cut-off value. The expression level, status and survival days were imported into the software, and the optimal points were calculated by the system. Afterwards, the R package “survival” [23] was used to perform Kaplan-Meier analysis [24]. To validate the expression and diagnostic efficacy of the hub genes in external datasets, CCA RNA sequencing data were downloaded from The Cancer Genome Atlas (TCGA) database (available online: https://cancergenome.nih.gov/), and box plots were drawn using the online tool E Chart (available online: http://www.ehbio.com/ImageGP/). The R package “pROC” [25] was used to draw the receiver operating characteristic (ROC) curves, and a larger area under the curve (AUC) represented a greater accuracy of the diagnosis. Finally, the Human Protein Atlas database (available online: https://www.proteinatlas. org/) was used to validate the immunohistochemistry results of the hub genes.

3. Results

3.1 Genes processing

The gene expression profile dataset GSE 89748, containing 72 CCA samples, was downloaded from the NCBI GEO database. A total of 31371 genes were obtained, and 6274 (the top 20% most variant genes) of them were used for WGCNA.

3.2 Construction of the weighted gene coexpression network and identification of key modules

After screening, 55 samples with qualified histological results were selected for coexpression construction (Fig. 2). The average linkage hierarchical clustering identified 25 modules, and the cyan, light cyan and red modules were the top three modules closely related to histology. Consequently, these three modules were regarded as the modules connected with CCA prognosis and were selected for further analysis (Fig. 4).

3.3 Gene Ontology (GO) and KEGG analysis of the hub genes in the hub modules

According to the results of GO analysis, the genes in the cyan module were mainly enriched in protein binding, nucleus, cytoplasm, nucleoplasm and cytosol. In the light cyan module, the genes were mainly enriched in protein binding, cytoplasm, cytosol, extracellular exosome and plasma membrane. In the red module, the genes were significantly enriched in ;protein binding, cytosol, extracellular exosome, membrane and nuclear region of cytoplasm (Fig. 5A–C). In KEGG analysis, gene expression, signal transduction, metabolism of proteins, metabolism and post-translational protein modification were the top five pathways in which the genes in the cyan module were significantly enriched. In the light cyan module, the genes were mainly enriched in signal transduction, metabolism, disease, post-translational protein modification and metabolism of proteins. The genes in the red module were significantly enriched in signal transduction, metabolism, immune system, gene expression and development biology (Fig. 5D–F).

3.4 Identification of hub genes

Based on the criteria described above ( $|$ GS $|>$ 0.2 and $|$ MM $|>$ 0.8), 11, 14 and 32 genes with high connectivity were selected as hub genes in the cyan, light cyan and red modules, respectively. Furthermore, we constructed three PPI network complexes using genes in these modules, containing 186 nodes and 538 edges (cyan module), 171 nodes and 534 edges (light cyan module), and 214 nodes and 595 edges (red module). Genes with connectivity $>$ 2 (total edges/total nodes) were considered core genes in each PPI network complex. Consequently, 20 genes, TSPAN8, LIMA1, HDAC1, ITGA3, ACTR3, GSK3B, KPNA1, ITGA2, THOC2, PTGES3, ILF2, SDHC, TFB2M, COPA, VPS72, NUP133, HEATR1, FLAD1, POLR3C and MRPL24, were regarded as potential candidates to be analysed and validated (Fig. 6). The

Figure 7.

Survival analysis of real hub genes in the dataset GSE89748. (A) LIMA1, (B) HDAC1, (C) ITGA3, (D) ACTR3, (E) GSK3B, (F) ITGA2, (G) THOC2, (H) PTGES3, (I) HEATR1 and (J) ILF2. Red lines represent high expression of hub genes and green lines represent low expression.

Figure 8.

Expression level of the real hub genes in tumour and normal samples. (A) LIMA1, (B) HDAC1, (C) ITGA3, (D) ACTR3, (E) GSK3B, (F) ITGA2, (G) THOC2, (H) PTGES3, (I) HEATR1 and (J) ILF2. Red points represent normal samples and green points represent tumour samples. Two-tailed Student’s t-tests were used to evaluate the statistical significance of differences.

details for the WGCNA and PPI network complex hub genes are displayed in Table S1.

3.5 Validation of real hub genes

In the survival analysis of the 20 candidate genes, the optimal cut-off values were obtained using X-tile (Fig. S1-2, Table S2), and 13 genes, TSPAN8, LIMA1, HDAC1, ITGA3, ACTR3, GSK3B, KPNA1, ITGA2, THOC2, PTGES3, COPA, HEATR1 and ILF2, were found to be negatively related to the prognosis of CCA (Fig. 7). Moreover, to validate the expression and diagnostic efficacy of the hub genes in external datasets, the TCGA dataset was analysed, and 10 genes were discovered to be significantly dysregulated in tumour samples compared to normal samples, including LIMA1, HDAC1, ITGA3, ACTR3, GSK3B, ITGA2, THOC2, PTGES3, HEATR1 and ILF2 (Fig. 8, Fig. S2). The ROC curves of these genes also showed higher diagnostic efficiency for CCA and non-tumour samples (Fig. S3). In addition, based on the Human Protein Atlas database, all 10 genes exhibited significantly higher protein levels in CCA samples than in normal samples (Fig. 9).

Figure 9.

Immunohistochemistry of the ten hub genes based on the Human Protein Altas. (A) LIMA1, (B) HDAC1, (C) ITGA3, (D) ACTR3, (E) GSK3B, (F) ITGA2, (G) THOC2, (H) PTGES3, (I) HEATR1 and (J) ILF2.

4. Discussion

CCA is the most common biliary malignancy arising from the epithelial lining of the bile duct [26]. Due to its silent clinical characteristics, difficulties in early diagnosis and limited therapeutic approaches, the prognosis of CCA is still dismal [6]. The detection of the disease at an early stage is crucial for patients to receive surgical resection, which is regarded as the single curative treatment. Nevertheless, the diagnosis of CCA is still challenging because of the lack of specific biomarkers in serum and/or biopsy samples and imaging methods [27, 28, 29]. Consequently, the identification of novel diagnostic and therapeutic biomarkers of patients with CCA is an important future direction. Recent advances in high-throughput technologies, including gene chips and mass-spectrometric techniques [30, 31], provide novel insight into the molecular understanding of CCA that can increase the identification of potential biomarkers for clinical use.

In the present study, a CCA-related dataset, GSE 89748, encompassing the molecular and clinical data of 72 cholangiocarcinoma samples, was downloaded from the NCBI GEO database. Using WGCNA, a coexpression network complex was constructed using 6274 genes, and 25 modules were screened. Among them, three modules, cyan, light cyan and red, were closely associated with histology and were regarded as modules containing hub genes. Subsequently, 20 candidates were discovered in both the coexpression and PPI networks, 13 of which were found to be related to the prognosis of CCA, and 10 genes, LIMA1, HDAC1, ITGA3, ACTR3, GSK3B, ITGA2, THOC2, PTGES3, HEATR1 and ILF2 were finally identified as real hub genes dysregulated in CCA samples and were associated with CCA prognosis.

HDAC1, also known as histone deacetylase 1, belongs to the class I histone deacetylases, which also contain HDAC2, 3 and 8 [31]. HDAC1 located in the nucleus, as an epigenetic factor, could antagonize the acetylation status of histone and nonhistone proteins [33]. It was reported that HDAC1 was dysregulated in various types of malignancies. In patients with breast cancer, Tang et al. [34]. discovered that the mRNA level of HDAC1 was overexpressed in tumour tissues. In addition, they also found that HDAC1 could activate Snail/IL-8 signalling and promote the proliferation and migration of cancer cells. In lung cancer patients, a recent meta-analysis revealed that the dysregulation of HDAC1 was related to poorer differentiation grade, which is commonly regarded as a reflection of poor prognosis in cancer [33]. Moreover, the relationship between the dysregulation of HDAC1 and dismal prognosis of lung cancer was also proven in this study. In liver cancer, Ser et al. [35] found that HDAC1 was overexpressed in HCC samples, and in addition to a higher number of cancer-specific deaths, its dysregulation was also related to poor differentiation grade, multiple lesions and satellite nodules. Several studies have focused on the role of HDAC1 in CCA prognosis. In a study of thirsty-five patients, Morine et al. [36] found that the overexpression of HDAC1 was associated with various malignant tumour characteristics, including higher tumour stage, vascular invasion and lymph node metastasis. They also proved that the dysregulation of HDAC1 was related to shorter disease-free survival and overall survival. In a recent study, Zhou et al. [37] discovered that the histone deacetylase inhibitor HC toxin could suppress the proliferation and migration of CCA cells in vivo and could thus be a potential candidate for CCA chemotherapeutic drugs. Undoubtedly, HDAC1 seems to be a crucial biomarker in CCA development and progression, and further studies should be conducted to reveal the related mechanisms.

Integrin alpha 2 (ITGA2) and Integrin alpha 3 (ITGA3) are both members of the integrin family of proteins, serving as cell surface adhesion proteins, and they play crucial roles in cell-cell interactions and developmental and pathological processes [38]. Both genes are aberrantly expressed in various cancers and are associated with poor prognosis of cancers [39, 40, 41, 42, 43]. For instance, in pancreatic cancer, the dysregulation of ITGA3 was discovered to be associated with poorer histological grades, later tumour stages, and shorter disease-free survival and overall survival [43]. To reveal the related mechanisms, Gong et al. [44] demonstrated that the overexpression of UCA1-ITGA2 could promote the proliferation, migration and invasion of pancreatic cancer cells. In a recent study, Ren et al. [45] proved that via the activation of the STAT3 signalling pathway, ITGA2 could upregulate PD-L1 expression and thus promote cancer cell proliferation and migration, which might provide a new insight into immune therapy for cancer. In CCA patients, ITGA3 was found to be overexpressed in CCA samples and was related to many malignant clinical traits, including tumour size, stage and lymph node metastasis [46]. In an in vitro experiment, although negatively related to CCA cell migration and invasion, Huang et al. [46] proved that ITGA3 could promote CCA cell proliferation by regulating the G1 phase of cell cycle progression. Considering the crucial roles of ITGA2 and ITGA3 in cancer development and progression, it is clear that they could be potential diagnostic biomarkers and therapeutic targets for CCA, and further explorations should be performed.

The dysregulation of the remaining core genes has also been discovered to be associated with cancer development, progression or poor prognosis. LIMA1, also known as EPLIN, is an actin-related molecule that is dysregulated in various cancers via adherens junction [47] and post-translational modifications [48], LIMA1 can influence cancer cell proliferation and invasion, which increases the metastatic probability of tumours [49]. GSK3B is a serine/threonine kinase that participates in various processes, including metabolism, neurological diseases and oncogenesis [50]. Recent studies proved that GSK3B plays two core roles in cancer development and progression: i) induction of NF- $\kappa$ B activation; and ii) prevention of EMT pathways [51]. THOC2 was proven to be an oncogene in melanoma and was associated with the poor prognosis of these patients; it could promote tumour cell proliferation and migration via the cAMP signalling pathway [52]. HEATR1 was found to play a crucial role in ribosome biogenesis, and the dysregulation of HEATR1 could cause impaired ribosome biosynthesis and lead to the activation of the p53-dependent cell cycle checkpoint, which might be associated with the genesis of cancer [53]. In pancreatic ductal adenocarcinoma ILF2 was found to regulate cancer cell proliferation and promote tumorigenesis by regulating cyclin members such as cyclin E and G1 cyclin [54].

In the GO analysis, protein binding was the pathway in which the genes in the three modules were mainly enriched. This result was consistent with the related functions of the hub genes, including LIMA1, GSK3B ITGA2 and ITGA3. Furthermore, in the KEGG analysis, the majority of the genes in the three modules were enriched in signal transduction, metabolism-related pathways, post-transduction modification and gene expression, which suggested that the genes were closely related to the mitotic cell cycle process. Cell cycle dysregulation has been demonstrated to be associated with the development and progression of various types of cancers, including CCA [55, 56]. Chen et al. [57] demonstrated that the downregulation of KLHL21 could alter the P21/P27/Cyclin E1 protein, trigger G0/G1 cell cycle arrest and prevent CCA cell migration and invasion by restraining the activation of the Erk signalling pathway, which might be a novel therapeutic target. In addition, several studies have shown that many drugs, such as telmisartan and cryptotanshinone, could induce cell cycle arrest and apoptosis in CCA and are potential therapeutic agents for CCA [58, 59]. Clearly, more attention should be paied to these cell cycle-related pathways and genes to explore the mechanism of CCA development and progression and CCA treatment.

There are still several limitations in this study. First, although a relatively large number of CCA samples were analysed, and an external TCGA dataset was used to validate the expression and diagnostic efficacy of the hub genes, more studies with larger samples sizes are needed to validate the findings of this study. Second, although the immunohistochemistry results of the hub genes were evaluated in this study, they were based only on the information from an online database. Consequently, further experiments should be performed to validate the results in this study and to reveal the mechanisms of these hub genes in CCA progression and prognosis.

In summary, by using coexpression analysis and a series of other bioinformatics approaches, we identified 10 real hub genes, LIMA1, HDAC1, ITGA3, ACTR3, GSK3B, ITGA2, THOC2, PTGES3, HEATR1 and ILF2, involved in the development, progression and prognosis of CCA. Cell cycle-related pathways such as signal transduction, metabolism related pathways, post-transduction modification and gene expression, were enriched in the hub genes. These study results enrich our understanding of CCA development and progression and may help us improve CCA therapy.

Footnotes

Acknowledgments

This study was funded by the National Natural Science Foundation of China (No. 81770566).

Conflict of interest

The authors declare no conflict of interest.

Supplementary data

The supplementary files are available to download from http://dx.doi.org/10.3233/CBM-190038.

References

Everhart

J.E.

and Ruhl

C.E.

, Burden of digestive diseases in the United States Part III: Liver, Biliary Tract, and Pancreas, Gastroenterology 136 (2009), 1134–44.

Chen

Zheng

Baade

P.D.

Zhang

Zeng

Bray

et al., Cancer Statistics in China, 2015. CA: A Cancer Journal for Clinicians 66 (2016), 115–32.

Blechacz

, Cholangiocarcinoma: Current Knowledge and New Developments, Gut and Liver 11 (2017), 13–26.

Rizvi

Khan

S.A.

Hallemeier

C.L.

Kelley

R.K.

and Gores

G.J.

, Cholangiocarcinoma — Evolving Concepts and Therapeutic Strategies, Nature Reviews Clinical Oncology 15 (2017), 95–111.

Goldaracena

Gorgen

and Sapisochin

, Current Status of Liver Transplantation for Cholangiocarcinoma, Liver Transplantation 24 (2018), 294–303.

Blechacz

Komuta

Roskams

and Gores

G.J.

, Clinical Diagnosis and Staging of Cholangiocarcinoma, Nat Rev Gastroenterol Hepatol 8 (2011), 512–22.

Jusakul

Cutcutache

Yong

C.H.

Lim

J.Q.

Huang

M.N.

Padmanabhan

et al., Whole-Genome and Epigenomic Landscapes of Etiologically Distinct Subtypes of Cholangiocarcinoma, Cancer Discovery 7 (2017), 1116–35.

Levy

Lymp

Angulo

Gores

G.J.

Larusso

and Lindor

K.D.

, The Value of Serum CA 19-9 in Predicting Cholangiocarcinomas in Patients with Primary Sclerosing Cholangitis, Dig Dis Sci 50 (2005), 1734–40.

Charatcharoenwitthaya

Enders

F.B.

Halling

K.C.

and Lindor

K.D.

, Utility of serum tumor markers, imaging, and biliary cytology for detecting cholangiocarcinoma in primary sclerosing cholangitis, Hepatology 48 (2008), 1106–17.

10.

Kulasingam

and Diamandis

E.P.

, Strategies for discovering novel cancer biomarkers through utilization of emerging technologies, Nat Clin Pract Oncol 5 (2008), 588–99.

11.

Betesh

Comunale

M.A.

Wang

Liang

Hafner

Karabudak

et al., Identification of fucosylated fetuin-a as a potential biomarker for cholangiocarcinoma, PROTEOMICS – Clinical Applications 11 (2017), 1600141.

12.

Zhu

A.X.

Borger

D.R.

Kim

Cosgrove

Ejaz

Alexandrescu

et al., Genomic Profiling of intrahepatic cholangiocarcinoma: Refining prognosis and identifying therapeutic targets, Annals of Surgical Oncology 21 (2014), 3827–34.

13.

Tang

Kong

Cui

Wang

Zhang

Gong

et al., Prognostic genes of breast cancer identified by gene co-expression network analysis, Frontiers in Oncology 8 (2018).

14.

Seol

M.A.

Chu

I.S.

Lee

M.J.

G.R.

Cui

X.D.

Cho

B.H.

et al., Genome-wide expression patterns associated with oncogenesis and sarcomatous transdifferentation of cholangiocarcinoma, BMC Cancer 11 (2011), 78.

15.

Zhou

Cheng

Jiang

Liu

Zhang

Liu

et al., Ten hub genes associated with progression and prognosis of pancreatic carcinoma identified by co-expression analysis, International Journal of Biological Sciences 14 (2018), 124–36.

16.

Langfelder

and Horvath

, WGCNA: An R Package for Weighted Correlation Network Analysis, BMC Bioinformatics 9 (2008), 559.

17.

Botia

J.A.

Vandrovcova

Forabosco

Guelfi

D’Sa

Hardy

et al., An additional K-means clustering step improves the biological features of WGCNA gene Co-expression networks, BMC Syst Biol 11 (2017), 47.

18.

Ravasz

Somera

A.L.

Mongru

D.A.

Oltvai

Z.N.

and Barabasi

A.L.

, Hierarchical organization of modularity in metabolic networks, Science 297 (2002), 1551–5.

19.

Liu

Wang

Zhang

Long

Zhang

et al., Declined preoperative aspartate aminotransferase to neutrophil ratio index predicts poor prognosis in patients with intrahepatic cholangiocarcinoma after hepatectomy, Cancer Res Treat 50 (2018), 538–50.

20.

Akgul

Bagante

Olsen

Cloyd

J.M.

Weiss

Merath

et al., Preoperative prognostic nutritional index predicts survival of patients with intrahepatic cholangiocarcinoma after curative resection, J Surg Oncol 118 (2018), 422–30.

21.

Wickham

, Ggplot2: Elegant Graphics for Data Analysis, 2009.

22.

Morris

J.H.

Demchak

and Bader

G.D.

, Biological Network Exploration with Cytoscape 3, Curr Protoc Bioinformatics 47 (2014), 8–13.

23.

Terry

M.T.

and Patricia

M.G.

, Modeling survival data: Extending the Cox Model Springer, New York, NY; 2000.

24.

Camp

R.L.

Dolled-Filhart

and Rimm

D.L.

, X-Tile: A new bio-informatics tool for biomarker assessment and outcome-based cut-point optimization, Clinical Cancer Research: An Official Journal of the American Association for Cancer Research 10 (2004), 7252–9.

25.

Robin

Turck

Hainard

Tiberti

Lisacek

Sanchez

J.C.

et al., PROC: An open-source package for r and s+ to analyze and compare ROC Curves, BMC Bioinformatics 12 (2011), 77.

26.

Torre

L.A.

Bray

Siegel

R.L.

Ferlay

Lortet-Tieulent

and Jemal

, Global Cancer Statistics, 2012, CA: A Cancer Journal for Clinicians 65 (2015), 87–108.

27.

Rahnemai-Azar

A.A.

Weisbrod

Dillhoff

Schmidt

and Pawlik

T.M.

, Intrahepatic cholangiocarcinoma: Molecular markers for diagnosis and prognosis, Surgical Oncology 26 (2017), 125–37.

28.

Weber

S.M.

Ribero

O’Reilly

E.M.

Kokudo

Miyazaki

and Pawlik

T.M.

, Intrahepatic cholangiocarcinoma: Expert consensus statement, Hpb 17 (2015), 669–80.

29.

Bridgewater

Galle

P.R.

Khan

S.A.

Llovet

J.M.

Park

J.W.

Patel

et al., Guidelines for the diagnosis and management of intrahepatic cholangiocarcinoma, J Hepatol 60 (2014), 1268–89.

30.

Sheynkman

G.M.

Shortreed

M.R.

Frey

B.L.

Scalf

and Smith

L.M.

, Large-scale mass spectrometric detection of variant peptides resulting from nonsynonymous nucleotide differences, J Proteome Res 13 (2014), 228–40.

31.

Vogelstein

Papadopoulos

Velculescu

V.E.

Zhou

Diaz

L.J.

and Kinzler

K.W.

, Cancer genome landscapes, Science 339 (2013), 1546–58.

32.

Buchwald

Kramer

O.H.

and Heinzel

, HDACi–targets beyond chromatin, Cancer Lett 280 (2009), 160–7.

33.

Cao

L.L.

Song

Pei

Liu

Wang

and Jia

, Histone deacetylase HDAC1 expression correlates with the progression and prognosis of lung cancer: A meta-analysis, Medicine (Baltimore) 96 (2017), e7663.

34.

Tang

Ding

Huang

Luo

Qing

Zhang

et al., HDAC1 triggers the proliferation and migration of breast cancer cells via upregulation of Interleukin-8, Biol Chem 398 (2017), 1347–56.

35.

Ler

S.Y.

Leung

C.H.

Khin

L.W.

G.D.

Salto-Tellez

Hartman

HDAC1 et al. and HDAC2 independently predict mortality in hepatocellular carcinoma by a competing risk regression model in a southeast asian population, Oncol Rep 34 (2015), 2238–50.

36.

Morine

Shimada

Iwahashi

Utsunomiya

Imura

Ikemoto

et al., Role of histone deacetylase expression in intrahepatic cholangiocarcinoma, Surgery 151 (2012), 412–9.

37.

Zhou

Chen

K.E.

Xiao

Duan

Huang

et al., Histone deacetylase inhibitor screening identifies HC toxin as the most effective in intrahepatic cholangiocarcinoma cells, Oncology Reports 35 (2016), 2535–42.

38.

Barczyk

Carracedo

and Gullberg

, Integrins, Cell Tissue Res 339 (2010), 269–80.

39.

Chuang

Y.C.

H.Y.

Lin

Y.L.

Tzou

S.C.

Chuang

C.H.

Jian

T.Y.

et al., Blockade of ITGA2 induces apoptosis and inhibits cell migration in gastric cancer, Biol Proced Online 20 (2018), 10.

40.

Guo

Moses-Gardner

Huang

Smith

E.R.

and Moses

M.A.

, ITGA2 as a potential nanotherapeutic target for glioblastoma, Sci Rep 9 (2019), 6195.

41.

Desgrosellier

J.S.

and Cheresh

D.A.

, Integrins in cancer: Biological implications and therapeutic opportunities, Nat Rev Cancer 10 (2010), 9–22.

42.

Koshizuka

Hanazawa

Kikkawa

Arai

Okato

Kurozumi

et al., Regulation of ITGA3 by the Anti-Tumor miR-199 family inhibits cancer cell migration and invasion in head and neck cancer, Cancer Sci 108 (2017), 1681–92.

43.

Jiao

Liu

Chen

and Liu

, ITGA3 serves as a diagnostic and prognostic biomarker for pancreatic cancer, Onco Targets Ther 12 (2019), 4141–52.

44.

Gong

Xiong

Zhang

and Yu

, Coexpression of UCA1 and ITGA2 in pancreatic cancer cells target the expression of mir-107 through focal adhesion pathway, J Cell Physiol 234 (2019), 12884–96.

45.

Ren

Zhao

Sun

Meng

Wang

et al., Overexpressed ITGA2 promotes malignant tumor aggression by up-regulating PD-L1 expression through the activation of the STAT3 signaling pathway, J Exp Clin Cancer Res 38 (2019), 485.

46.

Huang

Kong

Zhang

Zhou

Yan

et al., High Expression of ITGA3 promotes proliferation and cell cycle progression and indicates poor prognosis in intrahepatic cholangiocarcinoma, Biomed Res Int 2018 (2018), 2352139.

47.

Meng

and Takeichi

, Adherens junction: Molecular architecture and regulation, Cold Spring Harb Perspect Biol 1 (2009), a2899.

48.

Han

M.Y.

Kosako

Watanabe

and Hattori

, Extracellular signal-regulated kinase/mitogen-activated protein kinase regulates actin organization and cell motility by phosphorylating the actin cross-linking protein EPLIN, Mol Cell Biol 27 (2007), 8190–204.

49.

Collins

R.J.

Jiang

W.G.

Hargest

Mason

M.D.

and Sanders

A.J.

, EPLIN: A fundamental actin regulator in cancer metastasis? Cancer Metastasis Rev 34 (2015), 753–64.

50.

Kockeritz

Doble

Patel

and Woodgett

J.R.

, Glycogen synthase Kinase-3-An overview of an over-achieving protein kinase, Curr Drug Targets 7 (2006), 1377–88.

51.

Edderkaoui

Chheda

Soufi

Zayou

R.W.

Ramanujan

V.K.

et al., An Inhibitor of GSK3B and HDACs kills pancreatic cancer cells and slows pancreatic tumor growth and metastasis in mice, Gastroenterology 155 (2018), 1985–98.

52.

Zhou

Liu

Zhang

Chen

Wang

et al., Knockdown THOC2 suppresses the proliferation and invasion of melanoma, Bioengineered 10 (2019), 635–45.

53.

Turi

Senkyrikova

Mistrik

Bartek

and Moudry

, Perturbation of RNA Polymerase I Transcription Machinery by Ablation of HEATR1 Triggers the RPL5/RPL11-MDM2-p53 Ribosome Biogenesis Stress Checkpoint Pathway in Human Cells, Cell Cycle 17 (2018), 92–101.

54.

Wan

Gong

Liu

Wang

et al., NF45 overexpression is associated with poor prognosis and enhanced cell proliferation of pancreatic ductal adenocarcinoma, Mol Cell Biochem 410 (2015), 25–35.

55.

Malumbres

and Barbacid

, Cell cycle, CDKs and cancer: A changing paradigm, Nature Reviews Cancer 9 (2009), 153–66.

56.

Briggs

C.D.

Neal

C.P.

Mann

C.D.

Steward

W.P.

Manson

M.M.

and Berry

D.P.

, Prognostic molecular markers in cholangiocarcinoma: A systematic review, Eur J Cancer 45 (2009), 33–47.

57.

Chen

Song

Xuan

Zhao

et al., Inhibition of KLHL21 prevents cholangiocarcinoma progression through regulating cell proliferation and motility, arresting cell cycle and reducing erk activation, Biochem Biophys Res Commun 499 (2018), 433–40.

58.

Samukawa

Fujihara

Oura

Iwama

Yamana

Tadokoro

et al., Angiotensin receptor blocker telmisartan inhibits cell proliferation and tumor growth of cholangiocarcinoma through cell cycle arrest, Int J Oncol 51 (2017), 1674–84.

59.

Wang

Song

Jiang

et al., Cryptotanshinone induces cell cycle arrest and apoptosis through the JAK2/STAT3 and PI3K/Akt/NFkappaB pathways in cholangiocarcinoma cells, Drug Des Devel Ther 11 (2017), 1753–66.

Identification of hub genes and pathways in cholangiocarcinoma by coexpression analysis

Abstract

BACKGROUND:

OBJECTIVE:

METHODS:

RESULTS:

CONCLUSIONS:

Keywords

2.1 Dataset collection and processing

2.3 GO and KEGG pathway enrichment analysis

3. Results

3.1 Genes processing

3.2 Construction of the weighted gene coexpression network and identification of key modules

3.3 Gene Ontology (GO) and KEGG analysis of the hub genes in the hub modules

3.4 Identification of hub genes

Footnotes

Acknowledgments

Conflict of interest

Supplementary data

References