A dataset GSE97332 of circRNAs from 7 pairs of human HCC tumor tissue and normal hepar tissue was obtained from the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/geo/). And the expression profile of circRNAs in normal hepar tissue and HCC tissue was analyzed by GEO2R. P < 0.05 and │log2 (Fold Change)│ > 2 were used as the threshold for screening out the circRNAs with differential expressions. Volcano plots were used to show the upregulated and downregulated circRNAs in the dataset. Heat maps were used to show the 10 circRNAs that were significantly up-regulated and the 10 circRNAs that were significantly down-regulated in HCC tissues.

34 pairs of HCC tissues and matched adjacent tissues were collected from HCC patients receiving surgery in the Hospital. Immediately after surgical resection, tissue samples were stored in liquid nitrogen until further analysis. None of the patients receive chemotherapy or radiotherapy before the operation. This research, with signed informed consent from the patients, was endorsed by the Ethics Committee of the Hospital. The experiments concerning human tissues were conducted in line with the Declaration of Helsinki.

Human HCC cell lines (HepG2, SNU423, and SNU475) were obtained from American Type Culture Collection (ATCC, Rockville, MD, USA). A normal hepatocyte line (HL-7702) and human HCC cell lines (Huh-7 and SMMC-7721) were purchased from the Type Culture Collection of Chinese Academy of Sciences (Shanghai, China). All of the cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM, Sigma, St. Louis, MO, USA) with 10% fetal bovine serum (FBS, Invitrogen, Carlsbad, CA, USA) at 37°C and in 5% CO₂.

The total RNA was extracted from tissues and cells by a TRIzol kit (Invitrogen, Carlsbad, CA, USA) and then reversely transcribed into cDNA by a Primescript™ RT kit (TaKaRa, Dalian, China) and random primers or oligo (dt)₁₈ primers. A Mir-X™miRNA First-Strand Synthesis kit (Clontech, Shanghai, China) was used for the reverse transcription of miRNA. Then, qRT-PCR was performed on ABI 7300 rapid real-time PCR system (Applied Biosystems, Foster City, CA, USA) with a SYBR^®Premix Ex TaqTM II kit (Takara, Dalian, China) with U6 and GAPDH as the internal references. The relative expression was calculated by the 2^-ΔΔCt method. Primer sequences are detailed in Table 1 . In RNase R treatment assay, the total RNA (2 μg) was incubated at 37°C for 20 min with or without 3 U/μg RNase R. Then circ_0000098 expression was detected by qRT-PCR, respectively, with GAPDH as the control. For determining the subcellular localization of circ_0000098, A PARIS™ kit (Thermofisher Scientific, Waltham, MA, USA) was used for subcellular fractionation. Then Circ_0000098 expression in cytoplasm and nucleus of the cells was examined by qRT-PCR, with GAPDH and U6 as the cytoplasmic control and nuclear control, respectively.

Huh7 and SMMC-7721 cells were treated with 2 μg/mL actinomycin D (Amyjet, Wuhan, China) for 0, 4, 8, 12, and 24 h. After the cells were harvested, the expression levels of circ_0000098 and SLC30A7 mRNA were detected by qRT-PCR to verify the stability of circ_0000098.

Empty plasmid vector (Vector), circ_0000098 overexpression plasmid (circ_0000098-OE), siRNA negative control (si-NC), siRNAs targeting circ_0000098 (si-circ_0000098#1 and si-circ_0000098#2), miRNA mimics control (miR-NC), miR-1204 mimics, miR-1204 inhibitors (miR-1204-in), negative controls (miR-in), siRNAs targeting ALX4 (si-ALX4), and its negative control (si-NC) were available from GenePharma Co., Ltd. (Shanghai, China). The cells were transfected with Lipofectamine™ 3000 (Invitrogen, Carlsbad, CA, USA). After 48 h, the transfection efficiency was detected by qRT-PCR.

Cell proliferation was detected by a cell counting kit-8(CCK-8) kit (Beyotime, Jiangsu, China). To be specific, Huh7 and SMMC-7721 cells were transferred in a 96-well plate at a density of 2×10³ cells/well. At different time points (0, 1, 2, 3, 4, and 5 d), 100 µL of serum-free medium and 10 µL of CCK-8 solution were added to each well, followed by the incubation at 37°C for 1 h. Next, a ELX-800 spectrometer microplate reader (Bio-Tek Instruments, Winooski, VT, USA) was used to measure the absorbance at the wavelength of 450 nm.

The cells were inoculated in a 6-well plate at a density of 1×10⁵ cells/well. When cells reached 80%~90% confluence, a sterile pipette was adopted to make a cell-free scratch. The the cells were cultured with serum-free medium. Wound healing was monitored at 0 and 24 h by an inverted optical microscope (Nikon, Tokyo, Japan).

Cell migration and invasion were evaluated with Transwell chamber (pore size 8.0μm; Millipore, Billerica, MA, USA). The transfected Huh7 and SMMC-7721 cells (4×10⁵ cells) were resuspended in 200 µL of serum-free DMEM and then added into the upper chamber of Transwell chamber, and the bottom chamber was added with 500 μL of complete medium containing 10% FBS. After 24 h, the cells remaining on the upper membrane surface were removed. Moreover, the cells on the bottom surface were fixed in 4% paraformaldehyde and stained with 0.1% crystal violet (Beyotime, Jiangsu, China). Images were captured by an inverted optical microscope, and the number of cells in five random fields of view was counted. Matrigel (diluted 1: 9, Corning Incorporated, Corning, NY, USA) was used to cover the membrane in invasion assay, but it was not used during the process of migration assay.

The wild type (WT) or mutant type (MUT) of circ_0000098 or ALX4 3’-UTR sequence was amplified and cloned into the psiCHECK-2 vector (Promega, Madison, WI, USA) to construct luciferase reporter plasmid (circ_0000098-WT/MUT or ALX4 3’-UTR-WT/MUT). The reporter plasmids, together with miR-1204 mimic or miR-NC, were co-transfected into Huh7 and SMMC-7721 cells by Lipofectamine™ 3000. After 48 h, the luciferase activity was detected by the dual-luciferase reporter analysis kit (Promega, Madison, WI, USA).

RIP assay was conducted with a Magna RIP RNA binding protein immunoprecipitation kit (Millipore, Billerica, MA, USA). Huh7 and SMMC-7721 cells were lyzed by RIP lysis buffer, and 100 μL of whole-cell extract was incubated with RIPA buffer (Beyotime, Shanghai, China) containing magnetic beads coupled with human anti-Argonaute2 (Ago2) antibody (Millipore, Billerica, MA, USA) at 4°C 8 h. Normal mouse IgG (Millipore, Billerica, MA, USA) was employed as a negative control. The samples were rinsed with washing buffer and then incubated with proteinase K at 55°C for 30 min to separate RNA- protein complexes from magnetic beads. Ultimately, the immunoprecipitated RNA was extracted and analyzed by qRT-PCR.

The total proteins were extracted by RIPA lysis buffer (Beyotime, Shanghai, China), with the protein concentration determined by a BCA protein assay kit (Thermo Fisher Scientific, Rockford, IL, USA). The equal amount of proteins (30 μg/lane) were separated by SDS-PAGE and then transferred to PVDF membrane (Millipore, Billerica, MA, USA). After being blocked with 5% skimmed milk at ambient temperature for 1 h, the membrane was incubated overnight at 4°C with primary antibodies including anti-ALX4 (SC-33643, 1: 1000, Santa Cruz Biotechnology, Shanghai, China), anti-E-cadherin (ab76055, 1:1000, Abcam, Shanghai, China), anti-N-cadherin (ab280375, 1:1000, Abcam, Shanghai, China), anti-Vimentin (ab16700, 1:1000, Abcam, Shanghai, China), and anti-GAPDH (ab9484, 1: 1000, Abcam, Shanghai, China). After the membranes were washed by TBST, the membranes were incubated with horseradish peroxidase (HRP)-coupled secondary antibody (ab205719, 1: 2000, Abcam, Shanghai, China) for 0.5 h at room temperature. Ultimately, Clarity Max™Western ECL substrate (Bio-Rad, Hercules, CA, USA) was adopted to develop the protein bands.

The in vivo study were approved by the Animal Care and Use Committee of Yantai Affiliated Hospital of Binzhou Medical University. BALB/c nude mice (6 weeks old, male) were used to establish the lung metastasis model. 1×10⁶ Huh7 cells with circ_0000098 depletion or the control cells were injected into the tail vein of each mouse (10 mice per group). After 3 weeks, the mice were killed, and lung tissues were harvested. Subsequently, the lung tissues were fixed, embedded in paraffin and sliced, and hematoxylin and eosin (H&E) staining was performed to detect the metastatic nodules.

All experiments were performed in triplicate. SPSS version 22.0 software (SPSS Inc., Chicago, IL, USA) was adopted for statistical analysis, with the data recorded as mean ± standard deviation (SD). Fisher’s exact test was employed to analyze the relationship between circ_0000098 expression and clinicopathological features. Student’s t-test was used to analyze the difference between two groups. One-way ANOVA was used to analyze the differences among the multiple groups. Pearson correlation analysis was employed to study the correlation among circRNA, miRNA, and mRNA expressions. P<0.05 denoted that the difference was of statistical significance.

GEO database (GSE97332) was adopted to identify differentially expressed circRNAs in HCC tissues compared with normal tissues adjacent to HCC. According to the criteria (log₂|fold change |>2, P<0.05), there were 149 differentially expressed circRNAs in HCC tissues, of which 98 circRNAs were up-regulated and 51 circRNAs were down-regulated ( Figure 1A ). Circ_0000098 expression in HCC tissue was lower than that in normal tissues ( Figure 1B ). Circ_0000098 was derived from exons 3, 4, 5, 6, 7, and 8 of SLC30A7 gene ( Figure 1C ) (14). qRT-PCR indicated that circ_0000098 expression was markedly lower in HCC tissues than that in adjacent tissues ( Figure 1D ). Also, circ_0000098 expression in HCC cell lines, compared with that in HL-7702 cells, was significantly down-regulated ( Figure 1E ). To verify the circular structure of circ_0000098, we carried out RNase R treatment experiments in Huh7 and SMMC-7721 cells and found that circ_0000098 was resistant to RNase R while the expression of GAPDH mRNA was decreased significantly after RNase R treatment ( Figure 1F ). Furthermore, random primers or oligo (dT)₁₈ primers were used to perform reverse transcription with total RNAs extracted from Huh7 and SMMC-7721 cells. The results showed that when oligo (dT)₁₈ primers were used, the relative expression of circ_0000098 was significantly lower than when random primers were used, while the expression of linear SLC30A7 mRNA did not change significantly ( Figure 1G ). This indicated that circ_0000098 did not contain 3’ polyadenylated tail. In addition, actinomycin D assay demonstrated that circ_0000098 had a longer half-life time in Huh7 and SMMC-7721 cells compared with linear SLC30A7 mRNA ( Figure 1H ). This indicated that circ_0000098 was more stable than linear SLC30A7 mRNA. Besides, the Fisher’s exact test showed that the low expression of circ_0000098 was associated with the larger tumor size of HCC patients ( Table 2 ). The results suggested that circ_0000098 might be related to the progression of HCC.

To pinpoint the biological function of circ_0000098 in HCC, we transfected Huh7 and SMMC-7721 cells with two siRNAs targeting circ_0000098 to construct cell models of circ_0000098 knockdown ( Figure 2A ). CCK-8, wound healing, and Transwell assays showed that knocking down circ_0000098 significantly promoted the proliferation, migration, and invasion of Huh7 and SMMC-7721 cells ( Figures 2B – E ). Western blot assay verified that knockdown of circ_0000098 decreased the expression level of E-cadherin and up-regulated the expression levels of N-cadherin and Vimentin ( Figure 2F ). On the other hand, we transiently transfected the circ_0000098 overexpression plasmid into Huh7 and SMMC-7721 cells. Overexpression of circ_0000098 significantly inhibited proliferation, migration, invasion, and EMT of Huh7 and SMMC-7721 cells ( Supplementary Figures 1A – F ). These data implied that circ_0000098 modulated the malignant biological behaviors of HCC cells in vitro.

CircRNAs feature prominently in the cancer progression by adsorbing miRNAs (15). To reveal the hidden mechanism of circ_0000098 in regulating the HCC progression, we first analyzed the subcellular distribution of circ_0000098 and observed that circ_0000098 was predominantly located in the cytoplasm of HCC cells ( Figure 3A ). By searching the Circular RNA Interactome database (https://circinteractome.irp.nia.nih.gov) (16), we found putative binding sites between circ_0000098 and miR-1204, miR-1248, miR-1276, miR-136, miR-140-3p, miR-155, miR-183, miR-203, miR-337-3p, miR-346, miR-369-5p, etc. We mainly focused on exploring the role of miR-1204 ( Figure 3B ). Interestingly, high expression of miR-1204 expression was associated with the vascular invasion of the enrolled HCC patients ( Table 3 ). Dual-luciferase reporter assay indicated that as against the control group, the up-regulation of miR-1204 expression markedly decreased the luciferase activity of circ_0000098-WT reporter, but no significant change was observed on the luciferase activity of circ_0000098-MUT reporter ( Figure 3C ). In addition, RIP analysis showed that circ_0000098 and miR-1204 were markedly enriched in Ago2 group as against the IgG group ( Figure 3D ). In comparison with the si-NC group, miR-1204 expression in Huh7 and SMMC-7721 cells with circ_0000098 knockdown was observably increased ( Figure 3E ). Additionally, miR-1204 expression in HCC tissues and cell lines, compared with that in adjacent tissues or HL-7702 cells, was markedly elevated ( Figures 3F, G ). In addition, miR-1204 expression was negatively correlated with circ_0000098 expression in HCC tissues ( Figure 3H ). Overall, these data indicated that circ_0000098 could directly bind to miR-1204 and restrain its expression.

To study whether circ_0000098 partakes in regulating the progression of HCC through sponging miR-1204, we transfected si-NC+miR-in, si-NC+miR-1204-in, si-circ_0000098#1+miR-in, or si-circ_0000098#1+miR-1204-in into Huh7 and SMMC-7721 cells, respectively. qRT-PCR showed that miR-1204 expression was markedly declined in si-NC+miR-1204-in group as against si-NC+miR-in group, while miR-1204 expression level in si-circ_0000098#1+miR-in group was higher; miR-1204 expression in si-circ_0000098#1+miR-1204-in group, compared with that in si-circ_0000098#1+miR-in group, was decreased observably ( Figure 4A ). CCK-8, wound healing, Transwell, and Western blot assays found that in comparison with si-NC+miR-in group, the transfection of miR-1204 inhibitors markedly impeded the proliferation, migration, invasion, and EMT of Huh7 and SMMC-7721 cells; as against si-circ_0000098#1+miR-in group, the transfection of miR-1204 inhibitors weakened the promotion of circ_0000098 knockdown on proliferation, migration, invasion, and EMT of HCC cells ( Figures 4B – F ).

To elaborate on the potential mechanism of circ_0000098/miR-1204 axis in HCC, we predicted the potential target gene of miR-1204 by the TargetScan tool (http://www.targetscan.org/) (17) and discovered that there were complementary binding sites between miR-1204 and ALX4 3’-UTR ( Figure 5A ). Notably, low expression of ALX4 expression was associated with the vascular invasion of the enrolled HCC patients ( Table 4 ). Dual-luciferase reporter assay confirmed that miR-1204 overexpression markedly inhibited the luciferase activity of ALX4-WT reporter in Huh7 and SMMC-7721 cells, but the luciferase activity of ALX4-MUT reporter was not significantly changed ( Figure 5B ). Next, we studied the effects of the circ_0000098/miR-1204 axis on ALX4 expression and found that knocking down circ_0000098 inhibited ALX4 mRNA and protein expressions compared with the control while the transfection of miR-1204 inhibitors partially reversed these effects ( Figures 5C, D ). In addition, ALX4 mRNA and protein expressions were observably declined in HCC tissues and cells compared with that in adjacent tissues or HL-7702 cells ( Figures 5E – H ). Pearson’s analysis showed that ALX4 mRNA expression was negatively correlated with miR-1204 expression and positively with circ_0000098 expression in HCC tissues ( Figures 5I, J ). Next, we transfected miR-in, miR-1204-in, miR-1204-in+si-NC, and miR-1204-in+si-ALX4 into Huh7 and SMMC-7721 cells, and the successful transfection was confirmed by Western blot ( Supplementary Figure 2A ). We demonstrated that, the inhibitory effects of inhibiting miR-1204 on the proliferation, migration, invasion, and EMT of Huh7 and SMMC-7721 cells were reversed by ALX4 knockdown ( Supplementary Figures 2B – F ), and these data implied that miR-1204 promoted the malignant biological behaviors of HCC cells via repressing ALX4. Collectively, these results substantiated that ALX4 was a downstream target of miR-1204 in HCC cells, and its expression was negatively modulated by miR-1204 and positively by circ_0000098.

To further consolidate that circ_0000098 suppressed HCC progression, Huh7 cells with circ_0000098 knockdown and the control cells were respectively injected into the tail vein of the nude mice to establish a lung metastasis model. H&E staining of the lung tissues of the mice showed that, the lung metastasis of circ_0000098 knockdown group was much severer than that in the control group ( Figures 6A, B ). In addition, we found that compared with the control group, miR-1204 was highly expressed and ALX4 mRNA was lowly expressed in the lung tissues of mice in the circ_0000098 knockdown group ( Figures 6C, D ). In summary, these data indicate that circ_0000098 may inhibit the progression of HCC by suppressing metastasis.