Background
Cholangiocarcinoma is an aggressive malignancy that frequently occurs at the biliary tract, with an unfavorable prognosis [
1]. Intrahepatic cholangiocarcinoma (ICC) is rare and is regarded as the least common subtype of cholangiocarcinoma that arises from the epithelial cells of the intrahepatic bile ducts [
2]. The dominant risk factors for the pathogenesis of ICC consist of cirrhosis, chronic hepatitis B and C, alcohol use, diabetes, or even obesity [
3]. The early combination of different treatment modalities has been proposed to be beneficial for aggressive variants of ICC [
4]. However, the longer the time of diagnosis is delayed, the more likely the ICC lesion will undergo a loco-regional extension around the adjacent normal tissues, thus resulting in a poor prognosis [
5]. Identifying prognostic factors for ICC is therefore critical for the development of effective treatments for ICC.
The methylation of DLEC1 cilia and flagella associated protein is engaged in a favorable clinical outcome and prognosis in patients with small duct ICC [
6]. However, CD90 expression contributed to lymph node metastasis and thereby leads to a poor prognosis in patients with ICC [
7]. The expression of VEGFR-3 in ICC also promotes the angiogenesis of lymph, thereby resulting in an unfavorable prognosis [
8]. Long non-coding RNAs (lncRNAs) are a multiple class of RNAs engaged in various biological processes [
9]. LncRNAs can also serve as both oncogenes and suppressive genes, making them viable targets for tumorigenesis [
10]. Competing endogenous RNA (ceRNA) is a newly-emerged regulatory network, on a large scale across the transcriptome with expansion to the genetically functional information in the human genome, which exerts pivotal functions in cancer [
11]. Interestingly, ceRNAs indicate a novel modulation on lncRNAs and genes that exert crucial functions on cancer pathogenesis by binding with microRNAs (miRs) in cholangiocarcinoma [
12].
In this study, we adopted the transcriptome sequencing data of the lncRNAs, messenger RNA (mRNA) and miR from the Sequence Read Archive (SRA), and Cancer Genome Atlas (TCGA) database, respectively. Differentially expressed lncRNAs (DElncRNAs), DEmiRs and DEmRNAs were identified and adopted to construct an lncRNA-miR-mRNA ceRNA network. ICC-associated DEmRNAs were adopted to construct the protein–protein interaction (PPI) network. The expression of the top 6 genes was identified in the hub module provided by the PPI network, followed by GO and KEGG pathway enrichment analyses. The potential function of genes in the hub module of the ceRNA network was finally subjected to the Gene Set Enrichment Analysis (GSEA). Thus, the purpose of this study was to identify genetic alterations that underlie the prognostic factors of ICC.
Methods
Data collection and preprocessing
The original transcriptome sequencing data of the lncRNAs from human ICC and adjacent normal tissues were retrieved until September 2019 from the SRA database (
https://www.ncbi.nlm.nih.gov/sra/) and were then subjected to the SRP126672 dataset. The SRP126672 dataset was composed of 30 ICC tissues and 27 adjacent normal tissues. FastQC and Trimmomatic applications were adopted to quality-control and filter the original sequencing data. LncRNAs were quantified based on the Genome Research Project of Encyclopedia of DNA Elements (GENCODE) (GRCh37) catalog (
http://www.gencodegenes.org/). The RNA transcriptome sequencing data for ICC were downloaded from TCGA database. MiR sequencing and mRNA sequencing data were downloaded using a data transfer management tool (provided by GDC Apps) (
https://tcga-data.nci.nih.gov/). The miR sequencing data and mRNA sequencing data both consisted of 33 ICC tissues and 8 normal tissues, respectively. In addition, the GSE26566 dataset for ICC, which was downloaded from the Gene Expression Omnibus (GEO) database, included 10 samples of ICC and 10 non-tumor liver samples to determine the expression of the hub gene.
Identification of DEGs
The RNA-seq original data of ICC tissues and adjacent normal tissues were corrected, and normalized, with their expression calculated. DElncRNAs were screened using a DESeq 2 package. The adjusted standard was |log 2 (fold change [FC])| > 2, p < 0.05. In addition, the edgeR software package was used to screen DEmiRs and DEmRNAs with thresholds of |log 2 (fold change [FC])| > 2 and FDR < 0.01.
Construction of lncRNA-miR-mRNA ceRNA network
The ceRNA network was constructed based on the DElncRNAs, DEmiRs and DEmRNAs. The following databases: DElncRNAs, DEmiRs and DEmRNAs, and the miRcode (
http://www.mircode.org/) were adopted to predict the markedly downregulated lncRNAs targeted by upregulated miRs and markedly elevated lncRNAs targeted by repressed miRs in ICC. DEGs with the correct trends and targeting relationships served as candidate genes. Next, the TargetScan (
http://www.targetscan.org/), miRDB (
http://www.mirdb.org/) and miRTarBase online databases (
http://mirtarbase.mbc.nctu.edu.tw/php/index.php) were employed to predict markedly downregulated mRNAs targeted by upregulated miRs and markedly elevated mRNAs targeted by repressed miRs in ICC. The mRNAs with the correct trends in intersection among the three databases served as candidate genes. The predicted lncRNA-miR and miR-mRNA were combined to construct the lncRNA-miR-mRNA ceRNA network. Finally, the Cytoscape v3.6.1 software was adopted to visualize and map out the whole constructed network.
GO and KEGG pathway enrichment analysis
To elucidate the potential biological processes of DEmRNAs related to the ceRNA network in the development of ICC, the DAVID database (
https://david.ncifcrf.gov/) was used to perform a GO enrichment analysis of DEmRNAs by setting the default parameters. The GO function, enriched by
p < 0.05, was considered significant among the available transcriptome sequencing data. In order to understand the potential pathways of DEmRNAs involved in the ceRNA network, the KOBAS database (
http://kobas.cbi.pku.edu.cn/index.php) was employed to perform a KEGG pathway enrichment analysis on DEmRNAs, in which the significance of the KEGG pathway was evaluated at
p < 0.001.
Construction of the PPI network and module analysis
The interaction between DEGs identified key genes in modules involved in the development of ICC, with a combined score of > 0.4 for a PPI network to be considered as the threshold. The PPI information of DEmRNAs was obtained from the STRING database (
http://www.string-db.org/) and a PPI network was subsequently built using the Cytoscape v3.6.1 software. Lastly, the top six key genes in the hub module were obtained from the PPI network using the MCC network topology belonging to the cytoHubba plug-in in Cytoscape.
Association analysis of hub gene-related network and prognosis of ICC patients
ICC patients were arranged into 2 groups, the high RNA expression group and the low RNA expression group, according to the median expression value of the RNA. Both the Kaplan–Meier method and the log-rank test were used to determine the relationships between DElncRNAs, DEmiRNAs and DEmRNAs (belonging to the ceRNA network), as well as the overall survival (OS) curve of patients. The level of p < 0.05 was considered statistically significant.
Validation of expression of hub genes in ICC
The ICC gene expression dataset GSE45001, from the GEO database, was used to verify the expression of hub genes. The gene expression dataset GSE26566 was also used for the prediction of the hub gene function.
GSEA of the hub genes in ICC
To better elucidate the potential function of the hub genes in the ceRNA network, a GSEA was performed. According to the median expression of hub genes in the RNA sequencing data, 33 ICC samples from the TCGA database were assigned into the high expression group and the low expression group. The reference gene set is the annotated c2.cp.kegg.v7.0.symbols.gm gene set in the Molecular Signature Database (MSigDB), and the critical criterion was p < 0.05.
Discussion
The present study is still in its preliminary research phase pertaining to the prognostic factors for patients with ICC. To gain more insight into the molecules involved in the prognosis of patients with ICC, we analyzed the transcriptome sequencing data of the lncRNAs, mRNA and miR to construct a lncRNA-miR-mRNA ceRNA network, where 60 co-expressed DEmRNAs associated with ICC were identified. The main notable findings in the current study are that the expression of FOS, IGF2, FOXO1, and NTF3 was diminished, but the expression of IGF1R was enhanced in ICC tissues, compared with that of normal adjacent tissues. In addition, these five hub genes might regulate the development of ICC by targeting the “TGF-β signaling pathway”, “the hedgehog signaling pathway”, “retinol metabolism”, or “type II diabetes mellitus”.
FOS proteins are characterized by a leucine zipper motif and a basic region with a helix-turn-helix motif that binds to DNA and serve as an oncogene and as transcription factors by binding on the DNA sequences [
23]. The expression of FOS has been proposed to be positively related with the expression of c-Myb in colorectal cancer cells, while the expression of c-Myb is repressed in colorectal cancer tissues, suggesting that expression of FOS is also downregulated in colorectal cancer tissues [
24], which is concordant with the current study. The elevation of FOS-like antigen 1 expression is positively correlated with the progression of perihilar cholangiocarcinoma [
25]. More importantly, a prior research illustrated that FOS was involved in gene expression regulated by TGF-β [
26]. Furthermore, the tumor-promoting effects of TGF-β signaling pathway have been reported in cholangiocarcinoma [
15].
IGFs (IGF1 and IGF2) expedite glucose metabolism with their availability modulated by IGF binding proteins, and function as prognostic factors for type 1 diabetes [
27]. The induction of IGF2 is also partially involved in the proliferation and survival of rhabdomyosarcoma cells [
28]. Interestingly, IGF2 has been reported to be methylated in ICC compared to extrahepatic cholangiocarcinoma [
29]. IGF2 has been elucidated to be associated with retinol metabolism [
30]. Furthermore, Liu et al. observed that retinol metabolism was implicated in cholangiocarcinoma development [
19].
IGF1 maintains the phenotype of the tumor and allows the transformed murine pheochromocytoma cells to survive [
31]. The inhibition of IR/IGF1R reduced the epithelial-mesenchymal transition and cancer stem cell-like traits in ‘resistant cells’ of cholangiocarcinoma [
32]. The elevated expression of IGF1R was observed in tumor necrosis factor-related apoptosis-inducing ligand (TRAIL)-resistant gastric cancer cells, thus enhancing TRAIL resistance in gastric cancer cells [
33]. According to genes encoding proteins related to insulin receptors, IGF1R is able to stimulate renal cancer cells [
34]. It has been elucidated that IGFIR activates the Hedgehog signaling pathway in growth-plate (GP) chondrocytes [
35]. Research conducted by Guo et al. revealed that the activation of the Hedgehog signaling pathway is involved in proliferation, migration and EMT progression of cholangiocarcinoma cells [
36].
FOXO1 transcription factors orchestrate various cell types that are important in the host response [
37]. Moreover, FOXO1 has been described to assume a pivotal role in tumor initiation, progression and metastasis [
38]. A prior research indicated that the downregulation of FOXO1 elevated tumorigenesis and invasion of prostate cancer cells [
39]. FOXO1 expression was reduced in patients with type II diabetes mellitus, and that the downregulation of FOXO1 induces insulin resistance states that qualitatively and quantitatively mimic the function of adipocytes from patients with type II diabetes mellitus [
40]. NTF3, belonging to the neurotrophic factor family which encompasses nerve growth factor, brain-derived neurotrophic factor, and neurotrophic factor 4/5, has become a key mediator of neuronal development in early neurogenesis and throughout adulthood [
41]. Moreover, the correlation between NTF3 and diabetes mellitus has been identified [
42]. It has been shown that insulin resistance is a main determinant of the carcinogenic effect of type II diabetes mellitus [
43]. Notably, Lee et al. elaborated that diabetes mellitus was a risk factor for ICC [
44]. In fact, a literature reported that type II diabetes mellitus could increase the risk of ICC by 80%, and that the increase in ICC incidence and mortality observed over the past 3 decades was similar to that of type II diabetes mellitus and metabolic syndrome [
22].
Conclusion
In conclusion, our research identified several novel genetic alterations and pathways associated with the prognosis of ICC, as well as the potential roles of miRs, lncRNAs and mRNAs in the development of ICC via bioinformatic analysis. Based on the ceRNA network, we also discovered that FOS, IGF1R, IGF2, FOXO1, and NTF3 might target the “TGF-β signaling pathway”, “the hedgehog signaling pathway”, “retinol metabolism”, or “type II diabetes mellitus” pathways respectively, thereby modulating the subsequent development of ICC. A substantial insight gained from the understanding of the molecules involved in the prognosis of ICC contributed to an increased efficacy of the available treatments for patients with ICC. However, further studies with more selective candidate genes and a larger sample size are required to clearly define their roles in the development of ICC. Additional studies also need to be performed to examine the underlying mechanisms of FOS and TGF-β signaling pathway.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.