In total, 56 pairs of HCC tumor and adjacent liver tissues were obtained from surgical resections of patients at the First Affiliated Hospital of Guangzhou University of Chinese Medicine (Guangzhou, China). Patients received no chemo- or radio-therapy prior to surgery. Among them, six pairs of tissues (Supplementary Table S1) were used for circRNA-sequencing. 50 pairs of tissues (Supplementary Table S2) were used for qRT-PCR validation and exploring the correlation between circFGGY expression and overall survival of patients after hepatectomy. The use of clinical samples and materials for research purposes were approved by the Ethics Committee of the First Affiliated Hospital of Guangzhou University of Chinese Medicine [ethical approval number ZYYECK (2019) 008].

Total RNA was extracted using TRIzol Reagent (GLPBIO, Montclair, CA, United States). Reverse transcription for mRNA or circRNA was performed using Evo M-MLV RT Premix (Accurate Biology, Hunan, China) and cDNA amplification was performed using SYBR Green Premix Pro Taq HS qPCR Kit (Accurate Biology, China). Reverse transcription and cDNA amplification for miRNA were performed using miDETECT A Track miRNA qRT-PCR Starter Kit (RiboBio, Guangzhou, China) according to the manufacturer’s instructions. The primers are listed in Additional file: Supplementary Table S3.

The total protein of HepG2 and MHCC97H cells was extracted using RIPA lysis buffer (Beyotime, Shanghai, China), and equal amounts of protein lysates were separated on 10% SDS-PAGE gels. The separated protein bands were transferred onto polyvinylidene fluoride membrane. The membranes were incubated overnight at 4°C with the following primary rabbit-antibodies: Smad7 (1:1,000), E-cadherin (1:5,000), N-cadherin (1:5,000), Vimentin (1:5,000), GAPDH (1:5,000) and β-actin (1:2000) (Proteintech, Wuhan, China). After washing with TBST (15 min × 3), the membranes were incubated for 1.5 h at room temperature with anti-rabbit IgG, HRP-linked secondary antibody (1:3,000, Cell Signaling Technology, MA, United States). Then, the membranes were washed with TBST again (10 min × 4). Finally, the immunoreactive bands were visualized using the Immobilon® Western Chemiluminescent HRP Substrate (Millipore, Billerica, MA, United States).

HepG2 cells were planted into six-well plates (3 × 10⁵ cells/well). After 24 h, cells were treated with 5 μg/ml Actinomycin D (Cayman Chemical, Ann Arbor, MI, United States) or DMSO and collected at indicated time points. Total RNA (2 μg) was incubated with 3 U/μg of RNase R (Geneseed, Guangzhou, China) for 20 min at 37°C and 10 min at 70°C. After treatment with Actinomycin D or RNase R, the expression of circFGGY and FGGY mRNA were analyzed by qRT-PCR.

The probe for circFGGY was 5′-cy3-AGATTAGAACACTTTCCCGCATCC-3′. Fluorescent In Situ Hybridization Kit (RiboBio, China) was used in this assay. The HepG2 cells were seeded on the cell culture slide within 24-well plate (3 × 10⁴ cells/well). After 24 h incubation, the cells were fixed with 4% paraformaldehyde, hybridized with circFGGY-cy3 probes and the nuclei were stained with DAPI staining solution according to the manufacturer’s instructions. The images were acquired using a confocal microscope (Leica TCS SPE Ⅱ, Wetzlar, Germany).

The circFGGY-specific-ASO was designed and synthesized by RiboBio Co., Ltd. (Guangzhou, China). The overexpression plasmid of circFGGY was constructed by Shanghai Lianfeng Co., Ltd. (Shanghai, China).

The cell proliferation ability was assessed by cell counting kit-8 (CCK-8) assay and 5-Ethynyl-20-deoxyuridine (EdU) assay. The transfected cells were plated in a 96-well plate (4,000 cells/well). Then, on the indicated day, 10 μl of CCK-8 reagent (GLPBIO, United States) was added directly to the culture medium. After incubating for 1.5 h at 37°C, the absorbance (450 nm) was measured by Spectrophotometer (Thermo Fisher Scientific, Waltham, MA, United States).

EdU assays were performed using a Cell-Light EdU In Vitro Kit (RiboBio, China) according to the manufacturer’s instructions. The transfected cells were plated in a 24-well plate (5 × 10⁴ cells/well). Then, the cells were incubated with 50 μM EdU for 2 h, and stained with ApolloDye Solution and Hoechst33342. The cell proliferative ability was assessed based on the percentage of Edu-positive cells.

The diluted Matrigel Basement Membrane Matrix (50 μg/100 μl, Corning, Glendale, Arizona, United States) was added to the upper surfaces of transwell inserts (Corning, United States) and incubated at room temperature for 1 h. Transfected cells (1 × 10⁵) in 100 μl of serum-free DMEM medium were added to the inside of the transwell inserts chamber. A total of 600 μl of DMEM medium supplemented with 10% FBS was added into the lower chamber. After incubation for 24 h, the cells that did not invade were scraped off and invaded cells were fixed with 4% paraformaldehyde and stained with 0.5% crystal violet solution (Beyotime, China). The invasive ability of cells was assessed based on the number of the invaded cells.

PmirGLO-SMAD7 and pmirGLO-circFGGY Dual-Luciferase Reporter Vector (HedgehogBio, Shanghai, China) was constructed. The mutant strand was mutated in the putative binding site for miR-545-3p in SMAD7 3′-UTR region and circFGGY sequence. The wild-type (WT) or mutant SMAD7/circFGGY luciferase reporter plasmids (1 μg/ml) were co-transfected into MHCC97H cells with miR-545-3p mimics or NC-mimics (100 nM), respectively. Dual Luciferase Reporter Gene Assay Kit (Beyotime, China) was applied to assess the Firefly and Renilla luciferase activity 24 h after transfection. Data were acquired using the Multilabel Reader (PerkinElmer, Waltham, MA, United States).

The research was conducted in accordance with the internationally accepted principles for laboratory animal use and care as found in the Guideline for the Care and Use of Laboratory Animals published by the National Institutes of Health (NIH publication #85–23, revised in 1985). All animal procedures were approved by the Guangzhou University of Chinese Medicine Animal Ethics Committee (animal’s review number is: 20210721005). The five-weeks-old male BALB/c nude mice were purchased from Guangdong Medical Experimental Animal Center (Guangzhou, China), housed under Specific Pathogen Free Experimental Animal Center of Guangzhou University of Chinese Medicine. MHCC97H cells with circFGGY stably overexpressed or the NC cells (5 × 10⁶ cells in 0.2 ml of PBS) were injected to the axillary region on the right side (n = 5/group).

After 2 weeks, the circFGGY overexpressed groups were subcutaneously injected with miR-545-3p agomir or NC-agomir (Designed and synthesized by RiboBio). After 4 weeks, the mice were sacrificed, and the xenograft tumors were excised and analyzed by immunohistochemistry and hematoxylin and eosin (HE) staining. For imaging tumors in live animals, mice were anaesthetized with 2% isoflurane and injected intraperitoneally with 100 µl of the D-Luciferin sodium solution (15 mg/ml, MedChemExpress, NJ, United States). After 10 min, images were acquired with the in vivo imaging system (NightOwl II LB 983, Berthold, Germany).

GraphPad Prism version 8 (GraphPad Software, San Diego, CA) or SPSS 19 (SPSS Inc., Chicago, IL, United States) was used for data analysis. The Quantitative data were expressed as mean ± standard deviation (SD). Qualitative data were expressed by ratios. Student’s t test, chi-squared test, one-way ANOVA, two-way ANOVA, Wilcoxon signed-rank test, Pearson correlation analysis and log-rank test were performed, depending on the purpose, data types and grouping. In addition, Dunnett’s multiple comparisons test was performed after one way ANOVA. The selection of statistical methods is described in the figure legends. Statistical significance was indicated by a p value less than 0.05 (*p < 0.05, **p < 0.01, ***p < 0.001). All experiments were repeated for at least three times.

To examine the expression of circRNAs in HCC, six pairs of human HCC clinical samples and adjacent normal tissues were analyzed using circRNA microarray assays. A total of 8,934 distinct circRNAs were detected. Among them, 5,587 circRNAs have been recorded in circBase (Glazar et al., 2014) (Figure 1A). As the heatmap (Figure 1B) and volcano plot (Figure 1C) showed, 95 circRNAs were differentially expressed in tumor tissues and adjacent normal tissues. Among them, 73 circRNAs were upregulated (adjusted p value < 0.05, lgFC ≥ 1) while 22 were downregulated (adjusted p value < 0.05, lgFC ≤ -1).

It was found that circFGGY was downregulated in the HCC tumor samples compared with the adjacent normal tissues, which has not been reported previously. The expression of circFGGY was validated in various HCC cell lines and HCC clinical samples. The expression of circFGGY was significantly decreased in HCC cell lines including HepG2, Hep3B, MHCC97L, MHCC97H and Huh7 compared to that of the normal liver cell line L02 (Figure 1D). To confirm our findings, the level of circFGGY was quantified in 50 paired HCC and adjacent normal tissues by qRT-PCR (Figure 1E). In line with the RNA-seq results, the level of circFGGY was dramatically lower in tumor tissues than that in adjacent normal tissues. The Kaplan-Meier’s survival curves revealed that the HCC patients with low circFGGY expression had poor overall survival [(OS) p < 0.05] after hepatectomy (Figure 1F). Furthermore, the correlations between circFGGY expression and other clinical characteristics of 50 patients with HCC were analyzed (Table 1). BCLC stage and Edmondson’s grade were correlated with the expression of circFGGY. These results revealed that CircFGGY is downregulated in human hepatocellular carcinoma tissues and cells and might be associated with the pathogenesis of HCC.

To confirm the circular structure and characteristics of circFGGY, we firstly designed a pair of divergent primers that were specific for circFGGY and a pair convergent primer for FGGY mRNA. According to circBase, circFGGY was consisted of FGGY exons 3–5 and the specific back splicing site of circFGGY were examined by Sanger sequencing (Figure 2A). The PCR for cDNA and gDNA template using the divergent and convergent primers were performed respectively. The results showed that FGGY was amplified from the two kinds of DNA template, while circFGGY only existed in the cDNA template (Figure 2B).

Moreover, circFGGY was tolerated to RNase R, indicating that circFGGY is a circular RNA (Figure 2C). Furthermore, the half-life of circFGGY and FGGY mRNA in HepG2 cells treated with actinomycin D (5 ug/ml), a transcription inhibitor, was evaluated. The result showed that circFGGY was more stable than FGGY mRNA (Figure 2D). In addition, the results of qRT-PCR and in situ hybridization against circFGGY showed that circFGGY mainly distributed in the cytoplasm (Figures 2E,F). In brief, circFGGY was a circular and stable transcript that was generated from back-splicing of the FGGY gene transcript.

To investigate the biological functions of circFGGY in HCC, we conducted a series of experiments both in vitro and in vivo. Firstly, we constructed a circFGGY-overexpression plasmid and anti-sense oligonucleotide (ASO) targeting circFGGY specific back splicing sequence. According to the expression of circFGGY in various HCC cell lines, MHCC97H with low expression of circFGGY was transfected with circFGGY-overexpression plasmid, while HepG2 with high expression of circFGGY was transfected with circFGGY-ASO (Supplementary Figure S1A). We successfully upregulated the expression of circFGGY in MHCC97H cells and knocked down the expression of circFGGY in HepG2 cells (Supplementary Figure S1B) Then, the circFGGY-transfected cells were subjected to CCK8 assays, colony-formation assays, wound healing assays and transwell invasion assays to respectively test their cell growth, colony formation, migration and invasion properties. As expected, overexpression of circFGGY led to inhibition of growth, colony formation, migration and invasion in MHCC97H cells. In contrast, circFGGY silence promoted the cell growth, colony formation, migration and invasion in HepG2 cells (Figures 3A–D).

As showed above, circFGGY inhibited the growth and invasion of HCC cells. Therefore, we examined the expression of the biomarkers of EMT in HCC cells transfected with circFGGY overexpression vector or ASO-circFGGY. The results showed that circFGGY significantly promoted the expression of E-cadherin and decreased the expression of N-cadherin and Vimentin, indicating the inhibition of EMT (Figure 3E and Supplementary Figure S1C).

To further confirm the role of circFGGY in vivo, MHCC97H cells were stably transduced with lentiviral vectors containing circFGGY sequence. Subcutaneous injection of circFGGY–overexpressing MHCC97H cells showed decreased tumor growth rate. Strikingly, the tumor volume and weight at endpoint in mice of Ov-circFGGY group were about 0.52 ± 0.17 cm³ and 0.20 ± 0.07 g respectively, significantly lower than that in mice of NC-group which were about 1.66 ± 0.18 cm³ and 0.50 ± 0.05 g (Figures 3F–I). IHC analyses showed that the Ki-67 expression was dramatically decreased after overexpressing circFGGY (Figure 3J and Supplementary Figure S1D). These results suggested that circFGGY was involved in the regulation of cell growth, invasion and EMT in HCC.

It is well known that the TGFβ signaling pathway plays a critical role in the process of EMT. We investigated the RNA expression of the important molecule in the TGFβ signaling pathway. We found that circFGGY regulated RNA expression of Smad7 (Figure 4A) but not smad2, smad3, smad4, smad6 or FGGY, the host gene of circFGGY (Supplementary Figure S2A). The protein expression of Smad7 was also significantly increased in the cells overexpressed circFGGY and decreased in the cells with silence of circFGGY (Figure 4B, Supplementary Figure S2B).

As previously reported, circRNAs primarily functioned as miRNA sponges to regulate gene expression (Yu et al., 2018; Huang et al., 2020; Ren et al., 2021). We tried to find out the potential common miRNA target for Smad7 and circFGGY, which is associated with liver cancer. Through the Venn diagram analysis, it was found that miR-545-3p was the potential target (Figure 4C). To confirm this prediction, a dual-luciferase assay was performed. Targetscan (http://www.targetscan.org/vert_72/) was used to predict the binding sites of miR-545-3p and circFGGY/Smad7. miR-545-3p mimics or NC mimics were co-transfected with the circFGGY or Smad7-3′UTR luciferase reporters into the MHCC97H cells. Compared with the NC-mimics, miR-545-3p significantly reduced the luciferase reporter activity in the wild-type group of circFGGY and Smad7, but the luciferase activity did not significantly change after the target sites of miR-545-3p were mutated (Figure 4D). By performing in situ hybridization assays, we confirmed that circFGGY was colocalized with miR-545-3p in the cytoplasm (Figure 4E).

To explore the relation between Smad7 and circFGGY/miR-545-3p, we examined the RNA expression of miR-545-3p and Smad7 in the 50 paired HCC and adjacent normal tissues. Quantitative RT-PCR analysis showed that the miR-545-3p expression was higher, while the Smad7 expression was significantly lower in the HCC tissues than that in the adjacent normal tissues of these 50 patients with HCC (Figure 4F). Through Pearson correlation coefficients analysis, we found that the mRNA expression of Smad7 was positively correlated with the expression of circFGGY and negatively correlated with the expression of miR-545-3p (Figure 4G).

Moreover, miR-545-3p did not show significant changes after overexpressing or silencing of circFGGY (Supplementary Figure S2C). Similarly, circFGGY did not show significant changes after overexpression or knock-down of miR-545-3p (Supplementary Figure S2D). These findings suggested that circFGGY and miR-545-3p was not abolished by each other. Collectively, our data supported that circFGGY functions as a sponge of miR-545-3p, and upregulates the expression of Smad7.

According to a previous report, miR-545-3p was a tumor promoter in HCC (Changjun et al., 2018). Thus, we hypothesized that circFGGY inhibited the proliferation and metastasis of HCC by protecting Smad7 from digestion by miR-545-3p in HCC. To test this hypothesis, we upregulated and downregulated the expression of miR-545-3p through transfecting miR-545-3p mimics or inhibitors (Figure 5A). Then, the proliferation, migration and invasion of cells, as well as the expression of E-cadherin and Smad7 were assessed. As expected, the proliferation, migration and invasiveness of cells were increased in HepG2 transfected with the miR-545-3p mimics and decreased in HepG2 transfected with the miR-545-3p inhibitors, compared with the NC-group (Figures 5B–D). The mRNA and protein expression of Smad7, as well as the protein expression of E-cadherin were downregulated upon overexpression of miR-545-3p mimics and upregulated upon knock-down of miR-545-3p (Figures 5E,F).

We next explored whether circFGGY inhibits the invasiveness of HCC by sponging miR-545-3p. Functional experiments showed that the overexpression of miR-545-3p alleviated the inhibition of cell proliferation migration, and invasion of HCC cells caused by circFGGY overexpression (Figures 6A–C). Meanwhile, the mechanism-related experiments revealed that circFGGY promoted the mRNA and protein expression of Smad7 and reduced the level of EMT biomarkers, and the expression of these molecules was rescued by overexpression of miR-545-3p (Figures 6D,E).

To further explore the influence of circFGGY/miR-545-3p/Smad7 axis in tumor growth in vivo, we established a mouse HCC model. It was found that the suppression of xenograft tumor growth by overexpressing circFGGY was reversed by miR-545-3p agomir (Figures 7A–D). In line with in vitro study, the expression of E-cadherin and Smad7 was dramatically increased and that of Ki-67, N-cadherin and Vimentin was significantly decreased in the xenograft tumor with overexpression of circFGGY, which could be rescued by miR-545-3p agomir (Figures 7E,F and Supplementary Figure S2E).

Collectively, these observations demonstrate that circFGGY inhibits cell growth, invasion and EMT of HCC via regulating the miR-545-3p/Smad7 axis.