RNA sequencing data of GC, also known as stomach adenocarcinoma (STAD), including 211 normal samples and 408 tumor samples, were obtained from The Cancer Genome Atlas (TCGA,¹) database (Gu et al., 2021a).

The analysis of differentially expressed circRNAs in TCGA STAD tumor samples in comparison with normal samples was accomplished by using the limma package. Differentially expressed circRNAs’ P values were counted by t test during analysis. A cutoff criterion was set as | log2FC| ≥ 1 and P < 0.05 (Shi et al., 2018; Zhang et al., 2020a).

Gene Ontology (GO) is used for gene annotation, including three categories: cell component (CC), biological process (BP), and molecular function (MF) (Gu et al., 2020c). The Kyoto Encyclopedia of Genes and Genomes (KEGG) is a database for associating gene sets with related pathways (Gu et al., 2021b). DAVID (Database for Annotation, Visualization and Integrated Discovery) is an online device to operate GO annotation and KEGG pathway enrichment analyses.

Package Molecular Complex Detection (MCODE) of Cytoscape software was employed to aid in the analysis of the most dominant clustering module. Then, DAVID was adopted to analyze the GO term that was enriched by DEGs in different modules.

The samples of GC patients, as well as the corresponding normal tissues, were collected from our hospital. Each participant has provided the written informed consent, and our Hospital Ethics Committee has approved all the procedures. GES-1, BGC-823, NCI-N87, SGC7901, and AGS cells were used. In Dulbecco modified eagle medium (Gibco, NY, United States) supplemented with 10% fetal bovine serum (Gibco, United States) at 37°C in a 5% CO₂ incubator, all cell types were cultured.

Trizol reagent (Invitrogen, United States) was employed to obtain total RNA of GC cell and tissues. Then, the cDNA was obtained from total RNA using the Bestar^TM qPCR RT kit. Based on the guidance of the manufacturer, RT-PCR reactions were applied using the ABI 7500 system with Bestar^TM qPCR MasterMix.

The siRNA against circITGA7, miR-1417 mimics, anti-miR1417, and controls were produced by Genepharm. The sequence of circITGA7 was cloned into the pcDNA3.1 vector to obtain overexpressed circITGA7. In six-well plates, SGC7901 and AGS cells were cultured, and then they were transfected with siRNA using LipoFiter transfection reagent (Hanbio) following the guidance of the manufacturer.

The insertion of the circITGA7 and MTDH-3′ UTR fragment that contained the wild-type (WT) or mutant (MUT) miR-1471–binding site into the pGL3 vector (Promega, United States) was completed to construct the circITGA7 WT or MUT reporter vector.

For the dual-luciferase reporter assay, SGC7901 and AGS cells were seeded and then cotransfected with miRNAs and vector reporter plasmids with Lipofectamine 2000 (Invitrogen). Then, according to previous reports, dual-luciferase reporter assay was employed to evaluate luciferase activity. Renal luciferase activity was normalized to firefly luciferase activity.

Cell migration assays were conducted with 8-μm pore size Transwell chamber (Corning). Cancer cells (2 × 10⁴) in complete medium were added to the lower chamber, and those in 100 μL serum-free medium were added to the upper chamber. After being incubated for 24 h, the migrated cells were counted under a microscope according to a previous report.

A 24-well Transwell chamber was used to conduct the cell invasion assay. Transfected cells (5 × 10⁴ cells/well) were suspended in serum-free medium and then added to superluminal or matrix gel-coated Transwell inserts. The medium, which contained 10% fetal bovine serum, was added to the lower chamber. After 48 h of incubation, the migrated cells were counted under a microscope according to a previous report.

At a density of 2 × 10³ cells/well, GC cells were seeded. CCK-8 reagent (Dojindo Chemical Laboratory, Kumamoto, Japan) was added 3 h before the end of the experiment. The absorbance was measure with a microplate reader at 450 nm (Gu et al., 2020a).

In this study, the data are presented as mean ± SEM (Pan et al., 2017). GraphPad Prism version 7.0 software was employed to operate two-way analysis of variance or Student t test (Chen et al., 2021). P value, which was less than 0.05, was taken as having statistical significance.

To investigate the specific role of ITGA7 in the development of GC, we first evaluated how ITGA7 expressed in GC using TGGA database, and it was revealed that ITGA7 was expressed lowly in STAD (Figure 1A). Besides, in many other tumors, such as rectum adenocarcinoma, colon adenocarcinoma, and breast invasive carcinoma (BRCA), ITGA7 was also lowly expressed (Figure 1B). However, ITGA7 was highly expressed in glioblastoma multiforme (GBM) tissues in comparison with adjacent normal tissues.

In order to further understand the functional association of differentially expressed circRNAs in STAD, we used DAVID software to perform GO analysis and KEGG analysis. As Figure 2A shows, the results demonstrated that the significantly enriched GO terms in BP included muscle contraction, muscle system process, muscle cell differentiation, heart process, heart contraction, regulation of heart contraction, regulation of blood circulation, actin-mediated cell contraction, cell–substrate junction assembly, and cell–substrate junction organization. The significantly enriched GO terms in CC contained contractile fiber, focal adhesion, cell–substrate junction, myofibril, collagen-containing extracellular matrix, cell leading edge, sarcomere, I band, Z disk, and sarcolemma. The significantly enriched GO terms in MF comprised actin binding, tubulin binding, cation channel activity, extracellular matrix structural constituent, metal ion transmembrane transporter activity, calcium-channel activity, extracellular matrix structural constituent, calcium ion transmembrane transporter activity, and structural constituent of muscle. Furthermore, KEGG analysis demonstrated the association of these differentially expressed circRNAs with cGMP-PKG signaling pathway, the vascular smooth muscle contraction, axon guidance, aldosterone synthesis and secretion, oxytocin signaling pathway, adrenergic signaling in cardiomyocytes, focal adhesion, renin secretion, insulin secretion, calcium signaling pathway, dilated cardiomyopathy, cortisol synthesis and secretion, circadian entrainment, mitogen-activated protein kinase signaling pathway, pancreatic secretion, extracellular matrix–receptor interaction, hypertrophic cardiomyopathy, salivary secretion, purine metabolism (Figure 2B).

Being one of the features of the protein–protein interaction (PPI), the network module could also provide specific and significant biological information. As shown in Figure 3, there were in total nine modules detected and constructed by MCODE. Then, we further performed GO term enrichment analysis of some module. The results showed that module 1 was involved in divalent inorganic cation transport, divalent metal ion transport, and calcium ion transport; module 2 was involved in muscle system process and muscle contraction; module 3 was involved in muscle system process, striated muscle contraction, and muscle contraction; module 4 was involved in protein-containing complex disassembly, response to extracellular stimulus, and response to nutrient levels; module 5 was involved in anti-inflammatory response favoring Leishmania parasite infection, Leishmania parasite growth and survival, and ADORA2B-mediated anti-inflammatory cytokine production; module 6 was involved in class A/1 (rhodopsin-like receptors), G alpha (q) signaling events, and neuroactive ligand–receptor interaction; module 7 was involved in laminin interactions, NABA basement membranes, and PID integrin1 pathway.

To explore the role of circRNAs in GC, we first compared the expressions of circITGA7 between normal tissues and GC tissues. The data showed that compared to normal tissues, circITGA7 expressed significantly lower in GC tissues (Figure 4A). Furthermore, we examined the expressions of circITGA7 in GC cell lines (BGC-823, AGS SGC7901, and NCI-N87), and one healthy gastric epithelial cell (GES-1) was set as negative control. qRT-PCR assays confirmed that in healthy gastric cells, the expression levels of circITGA7 were lower in GC cells lines, especially in SGC7901 and AGS cells (Figure 4B).

We regulated high circITGA7’s expression in SGC7901 and AGS cells to explore its role in GC. In order to detect the circITGA7 expression, qRT-PCR was applied. The results showed that compared to control cells, the expressions of circITGA7 were remarkably downregulated in cells transfected with siRNA (Figure 5A) and upregulated in cells transfected with circITGA7 overexpression plasmid (Figure 5B). Then, it was shown that cell proliferation in circITGA7-overexpressed SGC7901 and AGS cells was significantly decreased (Figures 5C,D). Similarly, in circITGA7 lowly expressed cells, cell proliferation was significantly increased (Figures 5E,F). In Transwell assays, we added si-ITGA7 in SGC7901 and AGS cells to conduct invasion and migration assays, respectively. The results demonstrated that in circITGA7-downregulated cells, the abilities to invade and migrate were enhanced (Figures 5G,H).

To further explore the potential mechanism of circITGA7 in GC, we predicted the target miRNAs of circITGA7. We used miRBase² and circInteractome³ to analyze the potential targets of circITGA7. The ceRNA network analysis showed circITGA7 interacted with four miRNAs (miR-1471, miR-370-3p, miR-3187-3p, and miR-198). Thus, we examined the expressions of these four miRNAs in AGS cells transfected with siRNA or circITGA7-overexpressed plasmid, and the results are shown in Figures 6A,B. MiR-1471 expressed differently than the other three miRNAs. Next, to explore the interaction between miR-1471 and circITGA7 in SGC7901 and AGS cells, dual-luciferase reporter assay was conducted. Results revealed that miR-1471 mimics significantly attenuated the luciferase activity driven by circITGA7 WT in both AGS and SGC7901 cells, whereas miR-1471 mimics did not attenuate those driven by circITGA7 Mut (Figures 6C,D). Then, we observed the expression of miR-1471 in SGC7901 and AGS cells transfected with sicircITGA7. The data showed miR-1471 had an increased expression (Figure 6E). We also observed the expression of circTIGA7 in miR-1471–overexpressed SGC7901 and AGS. The results showed a decreased circITGA7 expression (Figure 6F). Therefore, there existed a negative relation between miR-1471 and circITGA7.

According to the results of TargetScan 7.2⁴, MTDH was identified as miR-1471’s target gene. We have established the plasmids of MTDH WT and MUT recombinant luciferase plasmid to verify the interaction between miR-1471 and MTDH. Using dual-luciferase reporter experiments, we revealed that miR-1471 mimics remarkably reduced the expression of MTDH 3′ UTR-WT–driven luciferase activity in AGS cells (Figure 7A), whereas anti–miR-1471 enhanced luciferase activity in SGC7901 cells (Figure 7B). Then, we detected the levels of MTDH in SGC7901 and AGS cells transfected with miR-1471 and anti–miR-1471, respectively. The assays indicated that there existed a negative relation between MTDH and miR-1471 in GC cells (Figures 7C,D). We explored the effects of MTDH on SGC7901 and AGS cells treated with circITGA7 siRNAs to better clarify the significant function of MTDH in GC cell lines. MTDH partially reversed the downregulation of circITGA7 in AGS cell (Figure 7E), as well as reversed upregulation of miR-1471 in SGC7901 cell (Figure 7F) induced by circITGA7 siRNAs. In functional assays, we found that circITGA7 overexpression’s inhibition of SGC7901 cell proliferation activity was reversed with MTDH (Figure 7G). Therefore, it could be verified that MTDH had a positive relation with circITGA7 in GC cells, and circITGA7 affected GC progression through miR-1471/MTDH axis.