br CircFAT e expression was downregulated in
3.2. CircFAT1(e2) expression was downregulated in GC tissues and cell lines and predicted better prognosis
To explore the functions of circFAT1(e2) in GC, we first examined its expression in GC tissues and cell lines by RT-PCR analysis. It was demonstrated that circFAT1(e2) expression was significantly down-regulated in GC tissues compared to normal tissues (Fig. 2A), and its expression was lower in metastasis tissues (n = 7) than non-metastasis tissues (n = 31) (P < 0.05, Fig. 2B). Moreover, circFAT1(e2) expres-sion was remarkably decreased in six GC cell lines compared with the normal human gastric epithelial mucosa cell line GES-1 (Fig. 2C). FISH was used to further evaluate the expression of circFAT1(e2), and results showed that its expression was significantly lower in GC tumor tissues than in normal tissues (P < 0.05, Fig. 2D and E). There was no cor-relation between circFAT1(e2) expression and age, or gender. However, circFAT1(e2) expression was significantly correlated with the pN status, pM status and clinical stage (P < 0.05, Table S2).
3.3. Overexpression of circFAT1(e2) significantly inhibited GC cell proliferation, invasion, and migration
To explore the functions of circFAT1(e2) in GC cells, we established a recombinant plasmid containing the exon2 of FAT1 and CMV pro-moter to generate GC cell lines that stably overexpressed circFAT1(e2) (Fig. 3A). The relative expression of circFT1 was significantly increased in transfected MGC-803 and MKN-28 Pifithrin-α (PFTα) (P < 0.05, Fig. 3B). Cell growth curves demonstrated that overexpression of circFAT1(e2) sig-nificantly inhibited cell proliferation in MGC-803 and MKN-28 cells (Fig. 3C and D). Colony formation assay suggested that overexpression of circFAT1(e2) remarkably reduced the number of colonies of MGC-
803 and MKN-28 cells (P < 0.05, Fig. 3E-H). Moreover, transwell as-says indicated that overexpression of circFAT1(e2) significantly atte-nuated the invasive and migratory abilities of MGC-803 and MKN-
28 cells (P < 0.05, Figs. S2A–2D). We further explored the eﬀects of enforced expression of circFAT1(e2) on regulating tumor growth in vivo. MGC-803 cells stably transfected with circFAT1(e2) or NC were subcutaneously injected into BALA/c mice, the tumor volume and weight were measured 0, 7, 14, 21, and 28 days after injection. De-creased tumor volume and weight of the xenografts were revealed in circFAT1(e2) transfectants compared with the NC group (P < 0.05, Fig. 3I-K). Immunohistochemistry (IHC) analysis suggested that the tumors formed from circFAT1(e2) transfected MGC-803 cells exhibited weaker Ki-67 staining than the tumors derived from NC transfected cells (P < 0.05, Fig. 3L).
3.4. circFAT1(e2) upregulated RUNX1 expression by directly binding to miR-548g in GC cell lines
To address the subcellular location of circFAT1(e2) in MGC803 cells, RT-PCR and FISH were performed in our study. The results showed that circFAT1(e2) was localized in both the cytoplasm (55%) and nucleus (45%) in MGC-803 cells (P < 0.05, Fig. 4A and S1B). It has been demonstrated that circRNAs may serve as competing RNAs to bind miRNAs in the cytoplasm. Therefore, we speculated that circFAT1(e2) may target miRNAs in MGC-803 cells. The circFAT1(e2)-miRNAs interaction analysis based on TargetScan (http://www. targetscan.org/vert_71/) showed the predicted targeted miRNAs of circFAT1(e2) in MGC-803 cells (Supplementary Excel-3) and that miR-548g had five binding sites for circFAT1(e2) (Fig. 4B). To validate the interaction between miR-548g and circFAT1(e2), wild type or mutant targeted sites of miR-548g in circFAT1(e2) were cloned into pGL3. Results from the dual luciferase reporter assay indicated that miR-548g mimics significantly attenuated the luciferase activity driven by wild type circFAT1(e2) (P < 0.05, Fig. 4C). Additionally, miR-548g in-hibitors remarkably enhanced the luciferase activity driven by wild type circFAT1(e2) (P < 0.05, Fig. 4D), whereas the mutant cir-cFAT1(e2) was not aﬀected by miR-548g. RNA precipitation (RIP) was performed to further determine the interaction between miR-548g and circFAT1(e2) using a miR-548g specific probe. As expected, we found a specific enrichment of circFAT1(e2) and miR-548g in the miR-548g