Identification of DEGs between ICC tissues and adjacent normal tissues
To identify DEmRNAs, DEmiRs and DElncRNAs in ICC, the SRP126672 was composed of 30 ICC tissues and 27 adjacent normal tissues downloaded from SRA database, followed by Fastqc and trimmomatic applications and then DElncRNAs were screened using the DESeq2 package. A total of 912 DElncRNAs were obtained, with 524 upregulated and 388 downregulated DElncRNAs (Fig. 1A,B). The different expression of DElncRNAs could help distinguish ICC tissues and 27 adjacent normal tissues determined by principal component analysis (Fig. 1C). DEmiR and DEmRNA sequencing data (counts) of ICC and adjacent normal tissues were obtained from the TCGA database, followed by the edgeR package to perform differential analysis. Finally, 66 DEmiRs (38 upregulated and 28 downregulated) (Fig. 1D,E) and 5,522 DEmRNAs (3158 upregulated and 2364 downregulated) (Fig. 1F,G) were obtained.
The lncRNA-miR-mRNA ceRNA network in ICC
To better understand the role of lncRNAs and miRs in the ceRNA network of ICC tissues, a lncRNA-miR-mRNA ceRNA network was established. Initially, 66 DElncRNAs targeted by with 66 DEmiRs, and 31,452 lncRNA-miR pairs were predicted. DEGs with correct trend and targeted relationships serve as candidate genes. The selected candidate genes contain 66 lncRNA-miR pairs including 18 markedly up-regulated DElncRNAs, 19 down-regulated DElncRNAs, 10 significantly up-regulated miRs, and 2 significantly down-regulated miRs.
In addition, 16457, 7052, and 2081 target mRNAs were predicted from the databases TargetScan, miRDB, and miRTarBase respectively. Then, 10 downregulated candidate DEmiRs were intersected with 2364 downregulated DEmRNAs, yielding 43 shared genes (Fig. 2A), and 2 downregulated candidate DEmiRs were intersected with 3158 upregulated DEmRNAs, yielding 17 share genes (Fig. 2B). Next, the shared genes serve as candidate genes and generated 136 miR-mRNA pairs including 43 significantly down-regulated and 17 significantly up-regulated DEmRNAs. A total of 37 DElncRNAs, 12 DEmiRs, and 60 DEmRNAs were used to construct the ceRNA network. Based on these lncRNA-miR pairs and miR-mRNA pairs, a ceRNA network consisting of 37 lncRNA nodes, 12 miR nodes, and 60 mRNA nodes in ICC was constructed (Fig. 2C).
GO and KEGG pathway enrichment analysis on DEmRNAs
To identify the biological functions and pathways of the 60 DEmRNAs in the ceRNA network, we used the DAVID database for GO function enrichment analysis and the KOBAS database for KEGG pathway enrichment analysis. GO enrichment analysis showed that DEmRNAs related to biological processes were mainly enriched in GO terms including the insulin receptor signaling pathway, positive regulation of cell proliferation, cell respiration and other items (p < 0.05). DEmRNAs related to cellular component were most closely related to mitochondrial inner membrane (p < 0.05). DEmRNAs related to molecular function were mainly enriched in GO terms of chemoattactant activity, growth factor activity, and transcriptional coactivator binding (p < 0.05) (Fig. 3A). KEGG analysis showed that DEmRNAs were significantly enriched in pathway such as cancer pathways, calcium signaling pathways, cytokin-cytokine receptor interaction, MAPK singling pathway and proteoglycans in cancer (p < 0.001) (Fig. 3B). Coherently, 60 significantly different DEmRNAs exert pivotal role in the occurrence and development of ICC.
Identification of 6 genes in the hub module in PPI network
PPI network was constructed based on DEmRNAs, involving 60 nodes and 41 edges (Fig. 4A). To identify the first six genes in the hub module in the PPI network, the MCC network topology in the cytoHubba plug-in was used to determine the intimacy of DEmRNAs. Fos proto-oncogene, AP-1 transcription factor subunit (FOS), insulin like growth factor 1 receptor (IGF1R), hepatocyte growth factor (HGF), insulin like growth factor 2 (IGF2), forkhead box O1 (FOXO1), and neurotrophin 3 (NTF3) in the hub module were obtained (Fig. 4B). Finally, the lncRNA-miR-hub gene subnetwork was constructed (Fig. 4C), including 86 ceRNA regulatory modules.
Validation of FOS, HGF, IGF2, FOXO1, NTF3 and IGF1R expression in ICC
To better validate our analysis, the expression levels of 6 genes in the hub module were in 33 ICC tissues and 8 adjacent normal tissues. Concordant with our previous analysis, results (Fig. 5A) demonstrated that FOS, HGF, IGF2, FOXO1, and NTF3 remarkably downregulated while IGF1R potently rose in ICC tissues (logFC |> 2, FDR < 0.01) compared to adjacent normal tissues. FOS, IGF2, and IGF1R were randomly selected and subjected to the gene expression dataset GSE45001 of ICC, and results exhibited the same trend (Fig. 5B-D). Taken together, these results validated our prediction.
GSEA prediction of biological pathways related to FOS, IGF2, and IGF1R
To better identify the biological pathways related to FOS, IGF2, and IGF1R, we classified 33 IHC samples from the TCGA database into the high expression group and low expression group. The reference gene set is the annotated c2.cp.kegg.v6.2.symbols.gmt gene set in the MSigDB. Results revealed that the ICC tissues in the high expression group of FOS, IGF2 or IGF1R were markedly enriched in “N-glycan biosynthesis”, “α-linolenic acid metabolism” or “type II diabetes” respectively (all p < 0.01) (Fig. 6A-C), suggesting the regulatory roles of FOS, IGF2, and IGF1R in ICC may be exerted by these pathways.