The liver cancer tissues used in this study were collected from 62 patients with hepatocellular carcinoma who were hospitalized at the Third Affiliated Hospital of Sun Yat-sen University (Guangzou, China), after obtaining written informed consent signed by each patient from February, 2015 to February, 2016 (Table I). The sample collection processes and following analysis were approved by the Ethics Committee of the Third Affiliated Hospital of Sun Yat-sen University prior to the surgery. The clinical samples were confirmed by experienced clinical pathologists before further analysis. The liver cancer cell lines, HepG2, Huh7 and SK-Hep1 were obtained from the Cell Bank of Chinese Academy of Sciences (Shanghai, China). The cell lines, MHCC97-H, MHCC97-L and SMCC7721, and the normal liver cell line, L-02, were purchased from the American Type Culture Collection (ATCC; Manassas, VA, USA). The MHCC-LM3 cells were provided by the Liver Cancer Institute of Zhongshan Hospital, Fudan University (Shanghai, China). The liver cancer cells were cultured in DMEM supplemented with 0.1% fetal bovine serum (FBS) at 37°C with 5% CO₂. For the tumorigenesis assay, we used 12 BALB/c nude mice (4 weeks old, female, weighing 14–16 g) obtained from the Shanghai Laboratory Animal Center, Chinese Academy of Sciences (Shanghai, China). The mouse were housed in a controlled environment (temperature, 20–26°C; humidity, 40–70%; 12-h light/dark cycle). The mouse experiments in this study was performed under the Affidavit of the Approval of Animal Ethical and Welfare of the Laboratory Animal Center of the Third Affiliated Hospital of Sun Yat-sen University.

Total RNA samples from the hepatocellular carcinoma tissues and the cell lines were extracted using the PureLink™ RNA Mini kit (cat. no. 12183018A; Thermo Fisher Scientific, Inc., Waltham, MA, USA) following the manufacturer's instructions. The tissues were fully homogenized in liquid nitrogen, and the fine powder was used for RNA extraction. The synthesis of cDNA with ~2 µg of total RNA was completed using the High-Capacity cDNA Reverse Transcription kit (cat. no. 4368813; Thermo Fisher Scientific, Inc.) according to the manufacturer's instructions. Quantitative (real-time) polymerase chain reaction (qPCR) was carried out to determine the relative gene transcript or circular RNA levels using the SYBR Select Master Mix kit (cat. no. 4472908; Applied Biosystems, Foster City, CA, USA) following the manufacturer's instructions. For the quantification of circNA levels, divergent primers were designed, and opposite-directed pairs of primers were used to assess the gene transcript level. The GAPDH gene was used as the internal control. Each quantitative analysis was performed at least 3 times for statistical analysis. For the identification of the ligation site of circCDK13, the PCR products were analyzed by Sanger sequencing. Briefly, the purified double-stranded cDNA was repaired at the end and added poly-A ends and then ligated. The PCR method was then used for the amplification of the cDNA. The library of cDNA was then detected using Agilent 2100 system software and quantified by qPCR. The library was then sequenced using the illumina Hiseq 2000 platform (Illumina, San Diego, CA, USA). The sequences of the primers used in this study were as follows: CDK13 exon 5 forward, 5′-AGATGCTTTGGATTTCAAGAAGG-3′; CDK13 exon 2 reverse, 5′-TGACGACTTCTGGATCTAGACAATC-3′; CDK13 exon 6 forward, 5′-CCACTTACACCAAGCATAGGA-3′; CDK13 exon 9 reverse, 5′-CTTCTTTATCTTCAGGCAGCAT-3′; and GAPDH forward, 5′-GAGTCAACGGATTTGGTCGT-3′; and reverse, 5′-GACAAGCTTCCCGTTCTCAG-3′.

The establishment of liver cancer cells overexpressing circRNA was conducted according to the protocols described in a previous study (41). Briefly, the cDNA of circCDK13 was synthesized by the Nuclee Biotechnology Company (Guangzhou, China) and cloned into the lentiviral expression vector, pLVX-IRESneo (Clontech Laboratories Inc., San Francisco, CA, USA; pLVX-IRESneo was used as the control vector). The cDNA sequence encoding exons 2–5 of the CDK13 gene, together with the additional sequence 500 bp upstream and 300 bp downstream of the non-linear splice site, were used in this study. Following construction, direct sequencing was performed to verify the sequences. Subsequently, the construct was used to transfect the liver cancer cells with Lipofectamine 2000 reagent (Invitrogen, Carlsbad, CA, USA). The expression of circular RNA in transfected cells was confirmed by RT-qPCR using specific primers pairs. At 24 h following transfection, the following experiments were carried out.

The proliferative potential of the liver cancer cells was determined using the (3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium, inner salt) (MTS) assay using the MTS Assay (Cell Proliferation) (Colorimetric) kit (cat. no. ab197010; Abcam, Cambridge, UK) according to the manufacturer's instructions. Briefly, the cells transfected with lentiviral expression vectors were counted under a microscope, and 100 µl cell suspension was seeded into each well of a 96-well plate (1×10⁴ cells/well) and cultured at 37°C with a suitable level of CO₂. Following adherence, 10 µl MTS solution were added at 24, 48 and 72 h. The absorbance at 490 nm (OD490) was measured using a spectrophotometer (ELx800; BioTek Instruments, Inc., Winooski, VT, USA) following culture at 37°C for 4 h. At least 3 biological and technical replicates were analyzed.

The effects of the circRNA on cell cycle progression were evaluated by measuring the percentages of cells in the G0/G1, M and G2 phases of the cell cycle using the Cell Cycle kit (cat. no. F559763; Forevergen, Guangzhou, China) following the manufacturer's instructions. Briefly, the liver cancer cells were washed 3 times with PBS solution, suspended by digestion with trypsin, and washed and collected by centrifugation at 300 × g for 1 min. After cell counting, 1×10⁶ cells were fixed with 70% ice ethanol at −20°C overnight, collected by centrifugation at 500 × g for 5 min, and then washed twice with PBS solution. The cell pellets were mixed and resuspended in 500 µl cell cycle staining solution, cultured at 37°C for 30 min, and finally detected by flow cytometry (FACSCalibur flow cytometer; BD Biosciences, San Jose, CA, USA). To evaluate statistical significance, at least 3 biological and technical replicates were performed for this assay.

The Transwell culture system was used to determine the effects of the circRNA overexpression on the liver cancer cell migratory capacity. The cultured liver cancer cells transfected with the expression plasmids were digested into a single-cell suspension with trypsin and counted under a microscope (IM-35; Carl Zeiss Inc., Thornwood, NY, USA). The cell density was adjusted to 1×10⁶/ml using serum-free medium. The upper chambers were filled with a 100 µl cell suspension, and the lower chambers were filled with complete medium containing 10% FBS. The cells were then cultured at 37°C for 24 h in a humidified atmosphere with 5% CO₂. The cells in the lower chambers were then fixed with 4% paraformaldehyde for 10 min, washed once with PBS solution, stained with crystal violet for 10 min, washed again with PBS solution, and finally counted and photographed with a Leica DC300F digital microscope (Leica Camera AG, Solms, Germany). Three biological replicates were performed for the statistical analysis of the cell proliferative capacity.

The invasion capacity of the liver cancer cells was also measured using the Transwell culture system. Specifically, a layer of Matrigel Basement Membrane Matrix (BD Biocoat; BD Biosciences, Bedford, MA, USA) was laid on the inner sides of the Transwell chambers according to the manufacturer's instructions, to mimic the in vivo extracellular matrix environment. The liver cancer cells were suspended by trypsin digestion, and the cell density was determined by counting under a microscope and diluted to 1×10⁶/ml. The upper and lower chambers were filled with a 100 µl cell suspension and 600 µl culture medium, respectively, and then maintained at 37°C for 24 h. The cancer cells in the lower chambers were then fixed with 4% paraformaldehyde for 15 min, washed once with PBS solution, stained with 1% crystal violet solution for 10 min, washed again with PBS solution, and finally counted and photographed. Three biological replicates were evaluated for statistical analysis.

The expression patterns of the liver cancer cells overexpressing circCDK13 were analyzed by microarray assay as previously described (42). Briefly, total RNA samples were extracted from the liver cancer cells using the PicoPure™ RNA Isolation kit (cat. no. KIT0204; Applied Biosystems) following the manufacturer's instructions. cDNA synthesis was carried out using the Agilent in situ Hybridization Kit plus (Agilent Technologies, Wilmington, DE, USA). The complementary double-stranded cDNA labeled with Cy3 or Cy5 was then hybridized and subjected to microarray analysis using Agilent Human G4131F 4×44 K. The spot intensities were then determined and analyzed using ImaGene 8.0 software and Nexus Expression software (BioDiscovery, El Segundo, CA, USA). Gene spots showing differences exceeding 1.5-fold were defined as differentially expressed genes and selected for subsequent clustering analysis. The heatmap and enrichment analysis of the signaling pathways were completed by searching against the Koyto Encyclopedia of Genes and Genomes (KEGG) database (www.genome.jp/kegg/pathway.html).

The abundances of several major signaling components were measured by western blot analysis as previously described (42). After the cells were washed with PBS solution, total proteins were lysed in RIPA buffer (50 mM Tris HCl, pH 7.4; 150 mM NaCl; 1% NP-40; 0.5% sodium deoxycholate), extracted from the liver cancer cells and boiled with sample loading buffer at 100°C for 5 min. BCA assay was used to quantify the total protein. Approximately 30 mg proteins was loaded and separated by 12% SDS-PAGE, transferred onto PVDF membranes, and incubated with 5% lipid-free milk solution, primary antibodies and secondary antibodies. The blots were finally developed with ECL solution (Amersham™) and exposed to film. The primary antibodies used in this study included anti-AKT1/2/3 (cat. no. ab106693), anti-AKT (phospho; cat. no. ab106693), anti-STAT3 (cat. no. ab68153), anti-STAT3 (phospho Y705) (cat. no. ab76315) and anti-GAPDH (cat. no. ab8245) antibodies, which were purchased from Abcam. The secondary antibody horseradish peroxidase-conjugated goat anti-mouse or anti-rabbit secondary antibodies (cat. no. 115-035-044) were purchased from Jackson ImmunoResearch (West Grove, PA, USA). For statistical quantification, 3 biological repeats were performed, and GAPDH was used as the internal standard.

The liver cancer cells in logarithmic growth phase were collected by washing with PBS solution and digestion with trypsin, and resuspended in PBS solution. The cell density was adjusted to 5×10⁷/ml. Nude mice were subcutaneously injected with 100 µl Sk-hep-1 NC or Sk-hep-1 circCDK13 cell suspension (6 mice in each group) into the left armpit. The tumor volumes were calculated each week using the following formula: Tumor volume (mm³) = tumor length × tumor width × tumor width/2. The tumors were finally weighed and photographed 45 days later.

Significant differences between multiple groups were analyzed by the one-way ANOVA with the Least Significant Difference (LSD) test, and differences between 2 groups were analyzed by the Student's t-test using the SPSS 18.0 software package. A significant difference was defined by a P-value <0.05.

By bioinformatics searches against the circBase database (29) and several large-scale investigations using deep sequencing (30–32), the circRNA circCDK13 was produced from 4 exons encoded by CDK13, as shown in Fig. 1A. The ligation of the second and fifth exons of the CDK13 transcript formed the circular structure of circCDK13 (Fig. 1A). For the investigation of the circCDK13 in the context of liver cancer, the expression of circCDK13 in the cancer tissues from 2 patients with hepatocellular carcinoma was examined by RT-qPCR with 2 different primer pairs targeting the genomic DNA and circRNA, respectively. Our results revealed the circular form of CDK13 RNA, circCDK13, in the clinical samples from these 2 patients with hepatocellular carcinoma (Fig. 1B). The ligation site of circCDK13 was identified by Sanger sequencing of the PCR products, further confirming the presence of the circular from of circCDK13 (Fig. 1C). Subsequently, we measured the expression levels of circCDK13 in 7 liver cancer cell lines, and found that the circCDK13 levels in these cancerous cells were all significantly lower than those in the control L-02 cells, supporting the potential involvement of circCDK13 in liver cancer progression (Fig. 1D). The decrease in circCDK13 expression was also observed in a large number of patients with hepatocellular carcinoma by RT-qPCR (Fig. 1E). The observed expression levels of circCDK13 in both liver cancer tissues and cell lines suggested that circCDK13 might play regulatory roles during the initiation and development of hepatocellular carcinoma.

To investigate the possible roles of circCDK13 during liver cancer development, we established SK-Hep-1 and Huh7 cells overexpressing circCDK13 using the lentiviral expression system as described in the Materials and methods. According to the results shown in Fig. 1D, the expression of circCDK13 differed significantly between the different cell lines. The expression of circCDK13 in the HepG2 and Huh7 cells was higher than that in the other cells, while the expression of circCDK13 was lowest in the SK-Hep-1 cells. Thus, the HepG2, Huh7 and SK-Hep-1 were the candidate cells. However, after screening, we found that the invasive ability of the HepG2 cells was weak, and these cells were thus not suitable for following experiments. Hence, the Huh7 and SK-Hep-1 cells were selected for use in further experiments. By RT-qPCR analysis, we observed that the expression levels of circCDK13 in the SK-Hep1 and Huh7 cells were markedly elevated following transfection with the overexpression plasmid compared with the negative control (Fig. 2A). Moreover, the proliferation rates of both the SK-Hep1 and Huh7 cells overexpressing circCDK13 were significantly lower than those of the control group (Fig. 2B), indicating the role of circCDK13 in regulating liver cancer cell proliferation during liver cancer development. Additionally, the percentages of liver cancer cells in the G0/G1, M and G2 phases of the cell cycle were analyzed by flow cytometry, and we observed that a greater number of circCDK13-overexpressing liver cancer cells compared with the negative controls were in the G0/G1 phase of the cell cycle, while the number of cells in the M phase was decreased by circCDK13 overexpression (Fig. 2C). These results indicated that the expression of circCDK13 played an important role in the pathology of liver cancer by modulating cell proliferation and cell cycle progression.

The dysregulation of cell migration and aberrant cell invasion are characteristics of liver cancer. For the further verification of the role of circCDK13 in the pathological processes of liver cancer, the migratory and invasive capacity of the SK-Hep-1 and Huh7 cells overexpressing circCDK13 were evaluated in this study. As shown in Fig. 3A, the results of Transwell culture assay revealed that the increase in circCDK13 expression markedly suppressed the migratory rates of both the SK-Hep1 and Huh7 cells compared with those of the control group (Fig. 3A). Similar to the migration assay, we found that the invasive capacity of the SK-Hep-1 and Huh7 cells overexpressing circCDK13 was also significantly suppressed compared with the control group (Fig. 3B). In both the migration and invasion assays, the Sk-Hep-1 cells overexpressing circCDK13 exhibited a highly significant decrease in cell numbers compared with the control group (Fig. 3). The markedly suppressed migratory and invasive abilities of the human liver cancer cells directly confirmed that this newly identified circRNA molecule is as an essential factor for liver cancer development as it suppresses cell migration and invasion, as well as cell proliferation and cell cycle progression.

To investigate the molecular mechanisms underlying the effects of circCDK13 on liver cancer cell proliferation, migration and invasion, transcriptome analysis using a microarray assay was performed to compare the differences in expressional profiles between the liver cancer cells overexpressing circCDK13 and the control group. In total, >4,000 genes were found to be significantly differentially expressed in the liver cancer cells overexpressing circCDK13 compared with the negative control, including 3,513 upregulated and 1,671 downregulated genes (Fig. 4A). Moreover, our KEGG pathway analysis revealed that these differentially expressed genes were significantly enriched in the JAK/STAT, PI3K/AKT and focal adhesion signaling pathways (Fig. 4B). Specifically, transcriptome analysis revealed that several genes in the JAK/STAT and PI3K/AKT pathways were differentially expressed in the liver cancer cells with an enhanced circCDK13 expression (Fig. 5A). The results from transcriptome and bioinformatics analysis strongly indicated that the functions of circCDK13 in regulating liver cancer cell proliferation, migration and invasion were mediated by multiple signaling pathways. Furthermore, the levels of phosphorylated AKT (p-AKT) and phosphorylated STAT3 (p-STAT3) proteins were markedly downregulated in both the Huh7 and SK-Hep-1 cells overexpressing circCDK13, directly confirming that the JAK/STAT and PI3K/AKT pathways were regulated by circCDK13 in liver cancer cells (Fig. 5B). These results revealed that circCDK13 regulates various signaling pathways in the liver cancer pathological processes, and JAK/STAT and PI3K/AKT may be two of the major mediating signaling pathways.

For further confirmation of the role of circCDK13 in liver cancer progression, a tumorigenicity assay was carried out by transplanting Sk-Hep-1 cells overexpressing circCDK13 or negative control cells into nude mice. The tumors formed from the liver cancer cells overexpressing circCDK13 were much smaller than those derived from the negative control group (Fig. 6A). Consistently, the tumor volumes in the experimental groups with an elevated circCDK13 expression were also significantly smaller compared with those in the negative control (Fig. 6B). Additionally, markedly decreased tumor weights were observed in the experimental group of mice transplanted with liver cancer cells overexpressing circCDK13 (Fig. 6C). The significantly suppressed tumor-forming capacity of the liver cancer cells overexpressing circCDK13 strongly supports our hypothesis that circCDK13 functions as a novel circRNA molecule associated with liver cancer cell proliferation and migration during liver cancer development and progression.