A novel circular RNA circ-LRIG3 facilitates the malignant progression of hepatocellular carcinoma by modulating the EZH2/STAT3 signaling

RNase R was purchased from Epicentre (Farmingdale, NY, USA), actinomycin D purchased from RayStarBio (HangZhou, Zhejiang, China) was used to assess the half-life. The specific STAT3 small-molecule inhibitor C188–9 was obtained from StemMed, Ltd. (Monmouth, NJ, USA). And the STAT3 activator colivelin was purchased from Santa Cruz Biotechnology (sc-361,153). The primary antibodies including anti-p-STAT3 (#9145), anti-STAT3 (#9139), anti-EZH2 (#5246), anti-GAPDH (#2118) and anti-Ki-67 (#9449) were all purchased from Cell Signaling Technology (Danvers, MA, USA).

A total of 130 paired surgically excised HCC and adjacent normal tissues from Henan Provincial People’s Hospital (Zhengzhou, China) were immediately collected into liquid nitrogen to preserve the integrity of the RNA. In addition, plasma samples from 36 HCC patients and 36 healthy controls were also collected to test the non-invasive diagnostic value of circ-LRIG3. None of the patients received preoperative radiotherapy or chemotherapy, and they all provided written informed consent. This study was conducted with the approval of the ethics committee of Henan Provincial People’s Hospital. Six HCC cell lines including SNU-423, HepG2, Hep3B, Huh7, SMMC-7721 and MHCC-97 L and one normal LO2 hepatocytes were all purchased from ATCC and cultured in DMEM medium containing penicillin/streptomycin and 10% fetal bovine serum.

Total RNA was classically extracted by Trizol reagent (Invitrogen, CA, USA) and dissolved in DEPC water. Then, RNA was reverse transcripted into cDNA using All-in-One™ First-Strand cDNA Synthesis Kit (GeneCopoeia, MD, USA), followed by quantification using BlazeTaq™ SYBR Green qPCR Mix 2.0 (GeneCopoeia). The primer sequences are shown in Table S3. GAPDH was used as the internal reference. The relative expression was calculated using the comparative Ct (2^-ΔΔCt) method. For determining the junction sequence of circ-LRIG3, the PCR product was purified and subjected to Sanger sequencing.

The FISH assay was performed by using RNA FISH Kit (GenePharma, Shanghai, China) according to manufacturer’s protocol with FAM-labeled circ-LRIG3 probe against the junction site of circ-LRIG3 (GenePharma). The nucleus was stained with DAPI. The fluorescence signal was observed, photographed and merged using confocal laser microscope system.

To overexpress circ-LRIG3, pLC5-ciR-GFP (#GS0108) vector containing one paired reverse complementary sequence (RCS) provided by Geneseed (Guangzhou, China) was used, in which the full-length of circ-LRIG3 (chr12: 59277301–59,308,117) was inserted into pLC5-ciR vector between EcoRI and BamHI restriction sites, followed by packaging into the lentivirus and infection into HepG2 cells. The stable circ-LRIG3-overexpressing cells were selected by 0.8 μg/mL puromycin. To knock down circ-LRIG3, two antisense oligonucleotide (ASO) against circ-LRIG3 junction site were designed and synthesized by RiboBio (Guangzhou, China). Then, the ASO was transfected into SMMC-7721 cells using riboFECT™ CP transfection solution (RiboBio) based on manufacturer’s protocol.

The CCK-8 assay was used to assess cell viability. In brief, 1500 HCC cells were seeded into the 96-well plate. After 24, 48 and 72 h, the medium was added with 10 μL CCK-8 solution (Dojindo, Kumamoto, Japan) for 2.5 h, followed by detection of the absorption value using a microplate spectrophotometer. Besides, the DNA synthesis rate was assessed by using EdU Staining Proliferation Kit provided by RiboBio based on manufacturer’s protocol. The positive cells were observed and photographed using confocal laser microscope system. For detecting cell apoptosis, the Annexin V-PE/7-ADD Apoptosis Detection Kit (BD Biosciences, CA, USA) was used, in which HCC cells were incubated with 5 μL Annexin V-PE and 5 μL 7-ADD in 500 μL binding buffer, followed by analyzing apoptotic cells using a flow cytometer (BD Biosciences).

The upper chamber of the trans-well insert with pore size of was coated with (invasion) or without (migration) Matrigel (Corning) and was placed in a 37 °C incubator for 2.5 h. Then, HCC cells were added into the upper chamber, and the lower chamber was added with 600 μL DMEM complete medium. 20 h later, the trans-well was fixed with formaldehyde for 5 min, permeabilized by methanol for 5 min, and was then stained by 0.5% crystal violet for 10 min. The invaded/migratory cells were counted under a light microscope at least 5 random fields.

For Western blot, whole cell extract was prepared by lysing cells in NP-40 lysis buffer. Then, proteins were resolved by 10% SDS-PAGE gel and transferred onto PVDF membrane. After incubating with corresponding primary and secondary antibodies, the membrane was strictly washed using TBST, and the chemiluminescent signals were developed using Clarity™ Western ECL Substrate (Bio-Rad, CA, USA). The Co-IP assay was carried out as previously described [17]. The IHC assay was performed by using Histostain-Plus IHC Kit (NeoBioscience, Guangzhou, China) as per Manufacturer’s instructions.

For the RNA pull-down assay, the biotinylated probe against circ-LRIG3 junction site was designed and synthesized by RiboBio. Then, the indicated HCC cell lysates were prepared using the NP-40 lysis buffer and incubated with above probe at 26 °C for 2 h with agitation. 50 μL streptavidin magnetic beads (Invitrogen) were pre-washed with 20 mM Tris (pH = 7.5) twice and added into above complex, incubating at 26 °C for 30 min with agitation. After strict washing, the proteins enriched by circ-LRIG3 were eluted for Western blot analysis. The RIP assay was performed using Magna^RIP RNA-Binding Protein Immunoprecipitation Kit (Millipore, MA, USA) based on the manufacturer’s instructions with additional UV crosslink step as previously described [17].

HCC cells were treated with formaldehyde for 10 min, followed by addition with glycine for 5 min. Then, cells were harvested and digested with a micrococcal nuclease for 20 min at 37 °C, making DNA break into 200 ~ 1000 bp fragments. The indicated antibodies were incubated with above cell lysates at 4 °C overnight with agitation. The second day, the protein A agarose (Santa Cruz Biotechnology) was prepared and added into above antibody-target protein-DNA complex, and incubated for 1 h at 26 °C with agitation. Lastly, the enriched DNA fragments were purified and collected for PCR analysis.

The Cignal STAT3 Reporter (luc) Kit (SA Biosciences, CA, USA) was used to assess STAT3 transcriptional activity according to manufacturer’s instructions. For determining the effect of STAT3 on transcriptional activity of circ-LRIG3 promoter, the full-length of circ-LRIG3 promoter with wild-type or mutant STAT3 binding site was inserted into pGL3-basic vector (Promega, WI, USA), followed by transfection into HepG2 and SMMC-7721 cells treated with colivelin or c188–9. The Dual-Luciferase Reporter Assay System (Promega) was applied to detect luciferase luminescence.

The stable circ-LRIG3-overexpresing HepG2 cells were subcutaneously or tail vein injected into age-matched male BALB/c mice, followed by growth for 5 weeks under specific-pathogen-free conditions. During this period, the mice were intraperitoneally injected with C188–9 (50 mg/kg) daily for 3 weeks. At the end of observation, the mice were killed by cervical dislocation, and the subcutaneous tumors were weighed and subjected to hematoxylin&eosin (H&E), IHC and TUNEL (Beyotime, Beijing, China) staining. For mice injected with cells via the tail vein, the lung tissues were collected for counting surface metastases under a dissecting microscope, followed by H&E staining. All experimental procedures in our study were approved by the Institutional Animal Care and Use Committee of Henan Provincial People’s Hospital.

Comparisons between two groups were analyzed by Student’s t test or Chi-square test as appropriate. ROC and Kaplan Meier-plotter were used to determine the diagnostic and prognostic values of circ-LRIG3, respectively. The link between circ-LRIG3 and p-STAT3 expression in HCC tissues was analyzed by Pearson Correlation Coefficient. All P-values were two-sided unless otherwise specified. P < 0.05 is indicative of statistical significance.

We collected 130 pairs of HCC and paracancerous normal tissues to conduct qRT-PCR. The results showed that circ-LRIG3 was significantly upregulated in HCC (Fig. 1a). Similarly, increased expression of circ-LRIG3 was observed in HCC cells in comparison to human normal LO2 hepatocytes (Fig. 1b). Sanger sequencing verified that circ-LRIG3 is produced by the back splicing of pre-LRIG3 mRNA exon 2 ~ 11, and its mature full-length is 1080 bp (Fig. 1c). Then, we designed divergent and convergent primers to amplify circ-LRIG3 and linear LRIG3, respectively (Fig. 1d). As shown in Fig. 1e, circ-LRIG3 was only amplified in cDNA, but not in gDNA. Further, we treated LO2 and SMMC-7721 cells with 5 μg/ml actinomycin D, and found that circ-LRIG3 had a half-life of more than 24 h (Fig. 1f). Besides, circ-LRIG3, not linear LRIG3, was highly resistance to RNase R (Fig. 1g). These suggest that circ-LRIG3 is a bona fide circRNA. Clinically, HCC patients with high circ-LRIG3 expression had larger tumor size, more vascular invasion, higher edmondson’s grade and later TNM stage than patients with low circ-LRIG3 expression (Table S1). Moreover, high circLRIG3 was positively correlated with short overall and disease-free survival time (Fig. 1h), and it was identified as an independent risk prognostic factor (Table S2). Of note, the qRT-PCR results showed that circLRIG3 was also increased in HCC plasma (Fig. 1i), and the area under the ROC curve (AUC) was 0.8681 (95%CI: 0.7843 ~ 0.9519) (Fig. 1j), implying that circLRIG3 is an excellent indicator for HCC diagnosis. The FISH assay showed that circLRIG3 was mainly localized in the nucleus (Fig. 1k).

To investigate the biological function of circ-LRIG3 in HCC cells, we stably overexpressed circ-LRIG3 in HepG2 cells relatively lowly expressing circ-LRIG3 by using pLC5-ciR lentiviral vector (Fig. 2a, b). And we knocked down circ-LRIG3 in SMMC-7721 cells relatively highly expressing circ-LRIG3 by using antisense oligonucleotide (ASO) targeting the junction site of circ-LRIG3 (Fig. 2c, d). Neither pLC5-circ-LRIG3 nor ASO-LRIG3 affected linear LRIG3 expression (Fig. 2b, d). The CCK-8 results showed that overexpression of circ-LRIG3 increased HepG2 cell viability, while knockdown of circ-LRIG3 decreased SMMC-7721 cell viability (Fig. 2e). Likewise, more DNA synthesis was observed in circ-LRIG3-overexpressing HepG2 cells (Fig. 2f), while less DNA synthesis was observed circ-LRIG3-silenced SMMC-7721 cells (Fig. 2g) as compared with their respective control cells. The migration and invasion of cells were enhanced after ectopic expression of circ-LRIG3 (Fig. 2h), and circ-LRIG3 depletion resulted in an opposite effect (Fig. 2i). In addition, the flow cytometry results showed that circ-LRIG3 overexpression reduced the number of apoptotic cells, whereas circ-LRIG3 depletion increased the number of apoptotic cells (Fig. 2j). More importantly, we found that overexpression of circ-LRIG3 in normal LO2 cells could also increase cell viability and invasion (Fig. S1). These above functional data indicate that circ-LRIG3 is a carcinogenic circRNA in HCC.

To determine how circ-LRIG3 promotes HCC progression, we performed RNA sequencing in control and circ-LRIG3-silenced SMMC-7721 cells, and found that circ-LRIG3 significantly affected the downstream genes of STAT3 signaling (Fig. 3a). The gene-set enrichment analysis (GSEA) results confirmed the the close connection between circ-LRIG3 and STAT3 signaling (Fig. 3b). Then, we performed qRT-PCR assay to verify RNA sequencing results, as shown in Fig. 3c, the levels of FAS, SOCS1, DUSP5, CXCL1 and STAM2 were notably increased in circ-LRIG3-overexpressing HepG2 cells in comparison to control cells, and depletion of circ-LRIG3 in SMMC-7721 cells led to an opposite trend (Fig. 3d). Further, overexpression of circ-LRIG3 markedly increased, while knockdown of circ-LRIG3 reduced STAT3 transcriptional activity (Fig. 3e). Consistently, the phosphorylation level of STAT3 (p-STAT3) significantly increased after circ-LRIG3 overexpression, and decreased after circ-LRIG3 silencing, but the total STAT3 level remained unchanged (Fig. 3f). Besides, we performed p-STAT3 immunohistochemistry (IHC) staining in HCC tissues, and found that its level was strongly positively associated with circ-LRIG3 level (r = 0.716) (Fig. 3g). Functionally, when HepG2 cells were treated with C188–9, a specific STAT3 small-molecule inhibitor, the enhanced cell viability and invasion caused by circ-LRIG3 overexpression were effectively abolished (Fig. 3h, i). These findings demonstrate that circ-LRIG3 functions via activating STAT3 signaling in HCC.

Given that EZH2 can directly bind STAT3 to increase STAT3 methylation and subsequent phosphorylation [18], we wondered whether circ-LRIG3 activates STAT3 through EZH2. The RPISeq online tool result showed that circ-LRIG3 and EZH2 are highly likely to interact (RF classifier = 0.91, SVM classifier = 0.92) (http://​pridb.​gdcb.​iastate.​edu/​) (Fig. 4a). We then designed biotinylated circ-LRIG3 probe to perform RNA pull-down assay, the circ-LRIG3 probe was verified to effectively enrich circ-LRIG3, but not linear LRIG3 (Fig. 4b). As shown in Fig. 4c, EZH2 was abundantly enriched by circ-LRIG3 probe in HepG2 cells, and this enrichment was significantly increased by circ-LRIG3 overexpression. Likewise, the interaction between circ-LRIG3 and EZH2 was also observed in SMMC-7721 cells, and circ-LRIG3 depletion evidently decreased this phenomenon (Fig. 4d). To confirm above results, we performed RNA immunoprecipitation (RIP) assay, and found that circ-LRIG3 was significantly immunoprecipitated by anti-EZH2 antibody as compared with negative anti-IgG antibody (Fig. 4e). Importantly, circ-LRIG3 overexpression increased STAT3 methylation and phosphorylation, whereas these effects were completely blocked after treatment with EZH2 siRNA or GSK-126, a highly selective EZH2 inhibitor (Fig. 4f). Similarly, EZH2 siRNA or GSK-126 effectively abrogated the enhanced malignant phenotype induced by circ-LRIG3 overexpression in HepG2 cells (Fig. 4g). These indicate that EZH2 is required for circ-LRIG3-induced STAT3 activation. In additions, the RPISeq online tool also showed that circ-LRIG3 and STAT3 are highly likely to interact (RF classifier = 0.81, SVM classifier = 0.88) (Fig. 4h), and this prediction was confirmed by RNA pull-down assay using biotin-labeled circ-RIG3 probe (Fig. 4i) and RIP assay using anti-STAT3 antibody (Fig. 4j) in both HepG2 and SMMC-7721 cells. More importantly, the Co-IP results showed that endogenous EZH2 and STAT3 were directly bound, whereas this interaction was almost disappeared after circ-LRIG3 knockdown (Fig. 4k), suggesting that circ-LRIG3 is critical for the binding between EZH2 and STAT3. And the immunofluorescence (IF) results showed that circ-LRIG3 was co-located with EZH2 and p-STAT3 in the nucleus (Fig. 4l), and this phenomenon was decreased by circ-LRIG3 depletion (Fig. 4l).

Given that STAT3 is a critical transcription factor, we then test whether circ-LRIG3 is also regulated by STAT3. After treatment of HepG2 and SMMC-7721 cells with colivelin, a specific STAT3 activator, the expression levels of circ-LRIG3 were remarkably increased, while treatment with C188–9 resulted in an opposite effect (Fig. 5a). Through analyzing the JASPAR database, two STAT3 binding motifs were found on circ-LRIG3 promoter (Fig. 5b, c). We then performed chromatin immunoprecipitation (ChIP) coupled PCR assay, the results showed that p-STAT3 was abundantly enriched on site 2 adjacent to the transcription initiation site (TSS), but not on site 1 and negative control region (Fig. 5d, e). Further, we performed luciferase reporter assay using pGL3-basic vector with wild-type or mutant STAT3 binding site (Fig. 5f), the results showed that colivelin and C188–9 significantly increased and decreased wild-type circ-LRIG3 transcriptional activity, respectively, whereas mutant of site 2, not site 1, completely blocked above effects (Fig. 5g). These data reveal that p-STAT3 directly binds to circ-LRIG3 promoter and activates its transcription.

Lastly, we tested whether circ-LRIG3 was functional in vivo via subcutaneous injection of HepG2 cells into nude mice. As shown in Fig. 6a, b, mice bearing tumors of circ-LRIG3-overexpressing cells had larger tumor volume and weight than mice bearing tumors of control cells. Moreover, more Ki-67 and p-STAT3 positive cells, while less apoptotic cells were observed in circ-LRIG3-overexpressing group as compared with control group (Fig. 6c). Further, we established lung metastasis model by caudal vein injection of circ-LRIG3-overexpressing HepG2 cells into nude mice, the results showed that circ-LRIG3 overexpression notably increased the number of lung metastasis nodules (Fig. 6d). Importantly, when nude mice were treated with C188–9, the increased tumor size and lung metastasis nodules caused by circ-LRIG3 overexpression were restored to control levels (Fig. 6a-d).