RETRACTED ARTICLE: Circ-DENND4C up-regulates TCF4 expression to modulate hepatocellular carcinoma cell proliferation and apoptosis via activating Wnt/β-catenin signal pathway

HCCLM3, Huh7, HepG2 and Hep3B were purchased as the human liver cancer cell lines, and THLE-3 served as a normal liver cell line. All cell lines were obtained from Cell Biology of the Chinese Academy of Sciences (Shanghai, China). All above cells were cultured in DMEM medium (Invitrogen, Carlsbad, CA) containing 100 U/ml penicillin, 10% fetal bovine serum (FBS) and 100 mg/ml streptomycin (all, Invitrogen) at 37°C with 5% CO₂.

Lipofectamine 2000 (Invitrogen, Carlsbad, CA, USA) was used to transfect cells with indicated plasmids. Cells were put in the six-well plate (2 × 10⁵ cells/well) for 24-hour cultivation. Lipofectamine RNAiMAX reagent (Invitrogen) was utilized for shRNA transfection in line with the protocols of suppliers. HepG2 and Huh7 cells were subjected to transfection with 2 μg of shRNA for two days. qRT-PCR was used to estimate transfection efficiency.

After transfection, cells were cultured in 6-well plates (3 × 10⁵ cells/well) and collected after 10 min of centrifugation. Afterwards, cells were rinsed in PBS and then subjected to propidium iodide (PI) dying in the dark. Then, cell proportions in G0/G1, S, and G2/M phases were examined by flow cytometry (BD Biosciences) as guided.

One Step TUNEL Apoptosis Assay Kit (Beyotime, Jiangsu, China) was used to estimate cell apoptosis. In short, cells in 6-well plates (1 × 10⁵ cells/well) were processed with the TUNEL kit, and then subjected to DAPI (Beyotime) staining. By fluorescent microscope (Olympus), apoptotic cells was imaged.

In order to test the localization of circ-DENND4C, RNAscope Multiplex Fluorescent Reagent Kit v2 (Advanced Cell Diagnostics, Inc., Newark, NJ, USA) was utilized to conduct FISH assay. Briefly, 4% paraformaldehyde was adopted to fix cells (1 × 10⁴) for half an hour. After digested by protease III prior, cells were subjected to take hybridization by the specific probes of target RNA at 40 °C for two hours. TSA plus Cyanine3, HRP-labeled oligos, amplifier and preamplifier were subjected to the hybridization by turn at 40 °C. Hoechst was adopted to stain the nuclei. Images were obtained by Zeiss LSM 710 confocal microscope (Jena, Germany).

Magna RIP™ RNA-Binding Protein Immunoprecipitation Kit (Millipore, Billerica, MA) was adopted for conducting the RIP assays in 1 × 10⁷ cells. Ago2 antibody was utilized to conduct the RIP assay and qRT-PCR was utilized to detect the co-precipitated RNA. IgG was seen as a negative control.

We initially detected circ-DENND4C expression in HCC cell lines (HCCLM3, HepG2, Huh7 and Hep3B) and normal human liver epithelial cell line (THLE-3). As manifested in Fig. 1a, the expression of circ-DENND4C was notably higher in HCC cells than that in THLE-3 cells. HepG2 and Huh7 cells were used to conduct following assays for that they exhibited highest expression of circ-DENND4C. Then, we started to verify the circular characteristic of circ-DENND4C. Convergent primers were designed to amplify DENND4C while divergent primers were designed to amplify circ-DENND4C. cDNA and gDNA were applied as the templates. Gel electrophoresis revealed that circ-DENND4C was amplified by divergent primers in only cDNA while DENND4C was amplified by convergent primers in both cDNA and gDNA (Fig. 1b). Next, we treated HepG2 and Huh7 cells with actinomycin D (ActD). The results disclosed that circ-DENND4C was not impacted while the level of linear DENND4C was obviously decreased under ActD treatment (Fig. 1c). Above results verified the abnormal expression and circular characteristic of circ-DENND4C in HCC. Thus, we continued to explore the function of circ-DENND4C in HCC. Before functional assays, knockdown efficiency of circ-DENND4C was verified (Fig. 1d). Then, we disclosed that circ-DENND4C knockdown decreased the proliferation of HCC cells (Fig. 1e, f, Additional file 1: Figure S1A). Further, flow cytometry analysis showed that silenced circ-DENND4C induced cell cycle arrest (Fig. 1g). In TUNEL assay, circ-DENND4C depletion elevated the apoptosis rates of HCC cells (Fig. 1h, Additional file 1: Figure S1B). Later, the levels of cell cycle-related proteins (Cyclin D1, CDK4) and apoptosis-associated proteins (Bcl-2, Bax) were tested in sh-circ-DENND4C#1/2 transfected cells. The results demonstrated that circ-DENND4C deficiency decreased the levels of Cyclin D1, CDK4 and Bcl-2 while increased that of Bax (Fig. 1i). Besides, transwell assay indicated the weakened HCC cell invasive ability in response to circ-DENND4C depletion (Fig. 1j). It was well known that tumor formation was associated with high differentiation of stem cells. Thus, the mRNA and protein levels of OCT4, SOX2 and NANOG (biomarkers of stemness) were examined by qRT-PCR and western blot. Results implied that OCT4, SOX2 and NANOG levels were all decreased in HCC cells transfected with sh-circ-DENND4C#1/2 (Fig. 1k, l). All data suggested that circ-DENND4C facilitated HCC cell proliferation, cell cycle, stemness, invasion and repressed HCC cell apoptosis.

Then, we sought the potential downstream pathways of circ-DENND4C in HCC. Activators of Hedgehog pathway, NOTCH pathway, Wnt/β-catenin pathway and PI3K/AKT pathway were used. The result revealed that only LiCl (activator of Wnt/β-catenin pathway) could rescue the suppressive effect of circ-DENND4C knockdown on HCC cell proliferation (Fig. 2a). Then, we detected levels of Wnt/β-catenin pathway associated proteins upon circ-DENND4C knockdown. It was disclosed in western blot analysis that protein levels of p-GSK3β, c-Myc, β-catenin and MMP7 were all reduced in cells transfected with sh-circ-DENND4C#1/2 (Fig. 2b). Later, western blot analysis and Immunofluorescence assay confirmed that circ-DENND4C knockdown could hinder the translocation of β-catenin into nucleus (Fig. 2c, d). Next, rescue assays were conducted to explore the influence of Wnt/β-catenin pathway on HCC cells with circ-DENND4C depletion. Colony formation assay disclosed that the treatment of LiCl could counteract the suppressive effect of silenced circ-DENND4C on HCC cell proliferation (Fig. 2e). Further, the inhibitory role of circ-DENND4C knockdown in cell cycle was also neutralized by treating LiCl (Fig. 2f). TUNEL assay displayed that LiCl abolished the facilitative effect on HCC cell apoptosis caused by circ-DENND4C depletion (Fig. 2g, Additional file 1: Figure S1C). It was also observed that the effect of circ-DENND4C deficiency on proteins related to cell cycle and apoptosis was abrogated after the treatment of LiCl (Fig. 2h). Besides, LiCl treatment recovered the weakened HCC cell invasion caused by circ-DENND4C knockdown (Fig. 2i). Further, treatment of LiCl countervailed the expression changes of OCT4, SOX2, NANOG at both mRNA and protein levels in circ-DENND4C-inhibited cells (Fig. 2j, k). Thus, we summarized that circ-DENND4C regulated HCC cell growth, invasion and stemness via activating Wnt/β-catenin signaling.

To further explore the molecular mechanism of circ-DENND4C in HCC, we first detected the subcellular localization of circ-DENND4C. It was disclosed that circ-DENND4C was mainly located in cytoplasm of HCC cells (Fig. 3a, b). Cytoplasmic distribution of circ-DENND4C indicated the post-transcriptional regulation of circ-DENND4C in HCC cells. Since ceRNA mechanism is a typical post-transcriptional network, we supposed that circ-DENND4C might serve as a ceRNA in HCC. By using “starBase” [15], 9 miRNAs were identified under the screening condition that strict stringency of CLIP data and at least 5 supported AGO CLIP-seq experiments. Next, it was displayed that only miR-195-5p and miR-6838-5p were significantly pulled down by circ-DENND4C biotin probe, while other not (Fig. 3c). The following RIP assay indicated that circ-DENND4C co-existed with miR-195-5p rather than miR-6838-5p in anti-Ago2 group (Fig. 3d). Meanwhile, we discovered that the expression of miR-195-5p was down-regulated in HCC cells while that of miR-6838-5p didn’t exhibit difference (Fig. 3e). In this regard, miR-195-5p was then proposed as the downstream of circ-DENND4C in HCC. Further, we found the binding sequences of miR-195-5p to circ-DENND4C and mutated these sequences as well (Fig. 3f). Besides, the overexpression efficiency of miR-195-5p was verified in qRT-PCR analysis (Fig. 3g). As revealed in luciferase reporter assay, miR-195-5p overexpression notably decreased the luciferase activity of circ-DENND4C-wt vectors while that of circ-DENND4C-mut vectors was not affected (Fig. 3h). Similarly, RNA pull down assay further verified the interaction of circ-DENND4C with miR-195-5p at predicted binding sequences (Fig. 3i). In conclusion, circ-DENND4C functioned as a sponge of miR-195-5p in HCC.

To support ceRNA hypothesis, we explored the target genes of miR-195-5p. According to the result of “starBase”, TCF4 was predicted as target gene of miR-195-5p. More importantly, TCF4 has been widely reported to activate Wnt/β-catenin pathway in many cancers, including HCC [16‐18]. We first observed that TCF4 was notably overexpressed in HCC cells compared with normal THLE-3 cells (Fig. 4a). Then, RIP assay was conducted to verify the interaction between miR-195-5p and TCF4 mRNA. The result disclosed that both miR-195-5p and TCF4 were significantly enriched in RNA-induced silenced complexes (RISCs) (Fig. 4b). Next, the binding sites between TCF4 3′UTR and miR-195-5p were disclosed and we mutated the binding sequences in TCF4 (Fig. 4c). RNA pull down assay revealed that TCF4 was notably pulled down by biotinylated miR-195-5p-wt while had no response to biotinylated miR-195-5p-mut (Fig. 4d). Further, we down-regulated the expression of miR-195-5p in HCC cells as verified by qRT-PCR analysis (Fig. 4e). Subsequently, luciferase reporter assay was carried out and results revealed that luciferase activity of wild type TCF4 reporter was significantly decreased by miR-195-5p overexpression and enhanced by miR-195-5p knockdown. However, the luciferase activity of mutant type TCF4 showed no change in cells transfected with miR-195-5p-mimics/inhibitor (Fig. 4f). Further, we examined the effect of miR-195-5p overexpression or circ-DENND4C knockdown on TCF4 expression in HCC. As expected, TCF4 mRNA and protein expressions were both down-regulated by miR-195-5p overexpression or by circ-DENND4C knockdown in HCC cells (Fig. 4g, h). Thus, we concluded that TCF4 was downstream of circ-DENND4C/miR-195-5p axis in HCC.

Finally, rescue assays were conducted to examine the effects of TCF4 on circ-DENND4C knockdown-mediated HCC cell development. Immunofluorescence staining and colony formation assay disclosed that TCF4 overexpression restored the suppressive effect of circ-DENND4C depletion on HCC cell proliferation (Fig. 5a, b). Furthermore, circ-DENND4C silencing-mediated inhibition on cell cycle was also recovered by TCF4 overexpression (Fig. 5c). TUNEL assay disclosed that TCF4 up-regulation restored the facilitating effect of circ-DENND4C depletion on HCC cell apoptosis (Fig. 5d). Through western blot analysis, overexpressed TCF4 rescued the effect of silenced circ-DENND4C on the levels of proteins involved in cell cycle and apoptosis (Fig. 5e). Moreover, TCF4 overexpression reversed the weakened HCC cell invasion caused by circ-DENND4C depletion (Fig. 5f). Additionally, the expression levels of OCT4, SOX2 and NANOG decreased by knockdown of circ-DENND4C were further normalized by TCF4 up-regulation at both mRNA and protein levels (Fig. 5g, h). In sum, circ-DENND4C mediated HCC cell growth, invasion and stemness via up-regulating TCF4.

Subsequently, we investigated whether circ-DENND4C regulate tumor growth by increasing TCF4 expression. At first, we subcutaneously injected HepG2 cells transfected with sh-NC, sh-circ-DENND4C#1 or sh-circ-DENND4C#1+pcDNA3.1-TCF4 into nude mice. As a result, we observed that the size of subcutaneous tumors was remarkably decreased by circ-DENND4C knockdown but was further normalized in sh-circ-DENND4C#1+pcDNA3.1-TCF4 group (Fig. 6a). Accordingly, we recorded that the tumor growth was obstructed in circ-DENND4C-depleted group but recovered after TCF4 upregulation (Fig. 6b). Also, the weight of tumors was evidently reduced in sh-circ-DENND4C#1 group but was further restored in the group of sh-circ-DENND4C#1+pcDNA3.1-TCF4 (Fig. 6c). IHC assay unveiled that TCF4 overexpression could offset the inhibitory effect of circ-DENND4C depletion on the positivity of Ki67 and PCNA (two well-known proliferation markers) in above tumors (Fig. 6d). All experimental data suggested that circ-DENND4C accelerated HCC progression by targeting miR-195-5p/TCF4 axis and activating Wnt/β-catenin pathway (Fig. 6e).