Background
Circular RNAs (circRNAs) are a newly identified class of noncoding RNAs (ncRNAs) that have recently elicited increased attention [
1,
2]. CircRNAs are formed by back-splicing of a downstream splice donor site to an upstream splice acceptor site, thus producing a covalently closed RNA molecule. Lacking 5′ caps and 3′ poly (A) tails, circRNAs have not received significant attention for a long time; recently, with the application of high-throughput sequencing, a growing number of circRNAs have been unveiled [
3]. Studies on circRNAs have revealed that they are structurally stable, presumably because their lack of free ends is resistant to exonuclease activity, which enables circRNAs to serve as a new class of diagnostic or prognostic biomarkers of diseases. Furthermore, emerging lines of studies have revealed that some circRNAs play important roles in physiological and pathological conditions including malignant tumors, and they may provide new potential therapeutic targets [
4].
Gastric cancer (GC) is the fifth most common cancer and the fourth-leading cause of cancer-related deaths worldwide [
5] and is thus a global cancer burden. Although diagnosis and treatment have improved over the last decades, the prognosis remains poor and the 5-year survival rate remains low in patients with GC [
6]. Therefore, the discovery of effective biomarkers and therapeutic targets is of great importance. Competing endogenous RNA (ceRNA) refers to RNAs that sequester or sponge miRNAs to regulate mRNA transcripts containing common miRNA recognition elements (MREs) [
7]. Recent studies reported that a large amount of conserved MREs present in circRNAs [
8], making circRNAs a new research hotspot in the field of ceRNA. CircRNAs have been found to participate in various biological processes in GC by working as ceRNAs [
9‐
11]. However, the identification and function of circRNAs in GC still require further investigation.
In the current study, data mining and bioinformatics analysis were combined to identify novel circRNAs and to investigate the underlying mechanism of circRNA-miRNA-mRNA network in GC (Additional file
2: Figure S1). First, we collected GC-related circRNA microarray and RNA-Seq datasets from the Gene Expression Omnibus (GEO) databases and PubMed publications and screened overlapped differentially expressed circRNAs (DECs). Then we validated these overlapped DECs by sanger sequencing and RNase R treatment, and verified their expression via reverse transcription-quantitative polymerase chain reaction (RT-qPCR) using GC samples. Further subcellular localization analysis and AGO2-binding sites mining indicated that these DECs function as ceRNAs. After predicting the sponge miRNAs of the DECs and miRNA target genes, we constructed circRNA-miRNA-mRNA networks; Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis on the target genes were conducted to investigate the potential pathogenesis of GC. Then, a protein–protein interaction (PPI) network with target genes was set up and hubgenes were determined. Moreover, we performed a connectivity map (CMap) analysis based on the hubgenes to identify potential bioactive compounds, which provides possible alternatives and chemotherapeutics for the treatment of GC patients.
Discussion
CircRNAs that were first discovered in eukaryotic cells in 1979 [
31] has been recently re-recognized and elicited attention. Unlike the formation of traditional linear RNA, circRNAs display a unique covalently closed circular form, which is responsible for their high stability and resistance to exonucleases. In addition, previous studies have proved their sequence conservation, abundant presence in exosomes and plasma, and spatial and cell-type specificity [
32]. Thereby, circRNAs have great potential to serve as a biomarker for early diagnosis and prognosis of many human diseases including cancers. With the development of high-throughput sequencing and bioinformatics analysis, mounting circRNAs were found and confirmed to be involved in the regulation of diversified biological processes in different cancer types [
33,
34]. In GC, some circRNAs, such as circHECTD1 [
35] and circFAT1 [
36], have been reported to be involved in GC cell proliferation, migration and invasion, and thereby exert their tumor-promoting or tumor-suppressing effects. However, the identification and function of circRNAs in GC still require further exploration.
In this study, we collected microarray and RNA-Seq datasets of GC from GEO databases and PubMed publications and screened six downregulated circRNAs (hsa_circ_0000390, hsa_circ_0000615, hsa_circ_0001438, hsa_circ_0002190, hsa_circ_0002449 and hsa_circ_0003120). Following sanger sequencing, RNase R treatment and RT-qPCR analysis, the six DECs were validated and their decreased expression in GC tissues was confirmed. Previous studies have revealed that hsa_circ_0000615 acts as miRNA sponge to exert different roles in different solid tumors [
37‐
40]; for example, hsa_circ_0000615 upregulates Sp1 expression by adsorbing miR-150-5p, and thereby promoting the proliferation and metastasis of nasopharyngeal cancer cells [
38]. However, until now studies about the effect of hsa_circ_0000615 and the other five DECs on GC have been rarely reported. Further studies are imperative to determine their potential roles in GC.
As ceRNA, circRNA harboring MREs can sponge miRNAs, which suppress the miRNA activity and result in alteration of expression level of miRNA target genes [
41,
42]. Subcellular localization analysis and AGO2-binding sites mining in this study indicate that the six DECs in the cytoplasm probably sponge miRNAs. To further determine whether the above six DEC act as ceRNAs in GC, their corresponding MREs were predicted with two online tools, CircInteractome and Circbank. The former web tool forecasts MREs via Targetscan algorithm, which predicts MREs by surveying for 7-mer or 8-mer complementarity to seed region and the 3′ end of each miRNA [
16,
43]. The latter web tool predicts MREs based on two different algorithms including miRanda [
44] and Targetscan [
43]. We selected miRNAs predicted by both CircInteractome and Circbank as target miRNAs of the six DECs, and collectively, 36 circRNA-miRNA interactions composed of 6 circRNAs and 33 miRNAs were identified.
After intersecting the target genes of the aforementioned 33 miRNAs and DEGs in GC from TCGA, 320 overlapped target genes were acquired to establish circRNA-miRNA-mRNA regulatory networks, providing an evidence of the ceRNA functional mechanism of the six DECs in GC. The KEGG pathway analysis indicated that these target genes were related to two critical tumor-associated signaling pathways, “MAPK signaling pathway” [
27,
28] and “PI3K-AKT signaling pathway” [
29,
30]. To comprehensively reveal potential relationships of target genes, we developed a PPI network using the 320 target genes and extracted 15 hubgenes (ATF3, BTG2, DUSP1, EGR1, FGF2, FOSB, GNAI1, GNAO1, GNAZ, GNG7, ITPKB, ITPR1, JUNID, NR4A3, PRKCB); and the differential expressions of the 15 hubgenes were subsequently validated with GC tissues from TCGA and with our GC samples. As previously demonstrated, some of the fifteen genes serve crucial roles in GC [
45‐
47]. For example, Tang et al. have demonstrated that FOSB was significantly decreased in GC tissues, consistent with the results of this study, and moreover, downregulated expression of FOSB was correlated with poor prognosis for GC patients [
46].
With the development of novel agents in recent years, survival outcomes of GC patients have improved [
48]. However, the overall prognosis of patients with advanced gastric cancer remains poor, and more effective drugs against GC are needed. Therefore, CMap analysis of the fifteen hubgenes was performed to explore available compounds for the treatment of GC. Based on the genome-wide expression profiling of gene transcripts technology, CMap presents a data-driven and reliable approach for identifying new drugs or repositioning existing drugs [
49]. Three chemicals (vorinostat, TSA, and astemizole) were determined and validated as the therapeutic options for GC. As histone deacetylase (HDAC) inhibitor, vorinostat alters the level of histone and nonhistone protein acetylation and thereby regulates gene expression, cell proliferation, angiogenesis and cell survival [
50]. Vorinostat is FDA-approved for the treatment of cutaneous T cell lymphoma [
51] and has been investigated in diverse clinical trials as a potential mono- or combination-drug therapy for solid tumors including GC [
52,
53]. TSA is another kind of typical HDAC inhibitor used mainly in laboratory experiments. TSA is known to induce cell cycle arrest and apoptosis in different cancer cell lines, and its antitumor effect in solid tumors including GC has been illuminated previously [
54]. Astemizole is an old anti-histamine that can target important proteins involved in the cancer progression, namely, ether à-go-go (Eag1) and Eag-related gene potassium channels [
55], thus inhibiting tumor cell proliferation [
56]. Moreover, previous evidence has revealed that Eag1 is expressed in several human tumor cell lines, including those from GC [
57]. However, its anti-GC effect have not been elucidated presently. In this study, we found that it potentially serves as therapeutic agent for GC.
This study provides a basis for exploring the pathogenesis and treatment strategy of GC from the circRNA-miRNA-mRNA network perspective and provides more evidence for three bioactive compounds (vorinostat, TSA, astemizole) as anti-GC agents. A recent study has presented perspectives on circRNA‐miRNA‐mRNA networks to explore the pathogenesis and therapy for pancreatic ductal adenocarcinoma (PDAC) [
58], and another study reported similar insights into the pathogenesis and therapy of HCC from the circRNA–miRNA–mRNA network view [
59]. However, these perspectives or insights needs further wet lab investigations. Nevertheless, combined with our findings, these studies further our understanding of tumors from the perspective of circRNA‐related ceRNA networks.
Conclusions
In conclusion, through an integrated analysis of microarray and RNA-Seq data, RT-qPCR and computational biology, six circRNA-miRNA-mRNA regulatory networks were established and these ceRNA networks uncovered six circRNAs (hsa_circ_0000390, hsa_circ_0000615, hsa_circ_0001438, hsa_circ_0002190, hsa_circ_0002449 and hsa_circ_0003120) that might function as ceRNA to play important roles in GC. Additionally, three bioactive compounds (vorinostat, TSA, astemizole) obtained from the CMap analysis were identified as therapeutic agents for GC. Our study provides a new insight for further exploration of the pathogenesis and therapy strategies of GC from the circRNA-miRNA-mRNA network perspective.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.