CircZDBF2 up-regulates RNF145 by ceRNA model and recruits CEBPB to accelerate oral squamous cell carcinoma progression via NFκB signaling pathway

Among three OSCC cell lines (SCC9, SCC15 and SCC25), SCC9 and SCC15 cells were obtained from American Type Culture Collection (ATCC, Manassas, VA, USA) while SCC25 cells were from Procell Life Science & Technology Co., Ltd. (Wuhan, China). A healthy human oral keratinocyte (HOK) cell line was purchased from Binsui Biotechnology Co., Ltd. (Shanghai, China). All the cell lines were cultivated in Dulbecco's Modified Eagle Medium (DMEM) (Gibco, Gaithersburg, MD, USA) and supplemented with 10% fetal bovine serum (FBS) (Gibco, USA) and 100 U/ml penicillin/streptomycin under 5% CO₂ at 37 °C. For circRNA characterization, 3 U/mg of RNase R from Epicentre Technologies (Madison, WI) and 2 mg/ml of Actinomycin D (Act D) from Sigma-Aldrich (St. Louis, MO) were used.

To knock down circZDBF2 or RNF145 expression, short hairpin RNAs (shRNAs) specifically targeting circZDBF2 (sh-circZDBF2-1/2/3) or RNF145 (sh-RNF145-1/2/3), together with the negative control (NC) were synthesized from GenePharma (Shanghai, China). In addition, pcDNA3.1/circZDBF2, pcDNA3.1/RNF145 and pcDNA3.1/CEBPB, along with their negative control pcDNA3.1 obtained from GenePharma were constructed respectively for the overexpression of genes. The miR-362-5p mimics/inhibitor, miR-500b-5p mimics/inhibitor and their negative controls (NCs) (mimics-NC/inhibitor-NC) were designed via Ribobio (Guangzhou, China). Cell transfection was taken using Lipofectamine 2000 (Invitrogen Corp., Carlsbad, CA, USA). Forty-eight hours later, cells were collected for further investigations.

Total RNA was extracted in SCC9 and SCC15 cells by the application of TRIzol® reagent (Takara, Japan) in line with the protocols of supplier. Prime Script™ RT Master Mixture (Takara 11141ES10, Japan) or TaqMan® MicroRNA RT kit (4366596, Applied Biosystems™, Foster City, CA, USA) was utilized for RNA reverse transcription (RT). Then, qPCR was implemented with the qRT-PCR Kit (QR0100-1KT, Sigma-Aldrich, USA) by using StepOnePlus™ Real-time PCR Systems (Applied Biosystems™). Relative expression levels were calculated using the 2^−ΔΔCt method, with U6 small nuclear RNA (U6) as the endogenous control to normalize miRNA expression and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) as the endogenous control to normalize mRNA/circRNA expression. Related primer sequences were exhibited in Additional file 8: Table S1.

Transfected SCC9 and SCC15 cells harvested at the logarithmic growth phase were incubated into 96-well plates (5 × 10³ cells/well) with complete culture medium at 37 ℃ for 24, 48, 72 h. Then, cells were cultivated with 10 μL of CCK-8 reagent (ab228554, Abcam, UK) for 1 h. The absorbance at 450 nm was assayed by microplate reader.

Transfected SCC9 and SCC15 cells were prepared in 6-well plates (800 cell/well) for 14 days of colony formation. The culture medium was discarded and the cells were washed with phosphate buffer saline (PBS) for two times. The colonies were fixed with methanol and dyed by 0.5% crystal violet solution. Finally, colonies with no less than 50 cells were counted manually.

Serum-free medium containing transfected SCC9 and SCC15 cells was planted on the top of 24-well transwell chambers (5 × 10³ cells/well) along with Matrigel, and the lower chambers were loaded with complete medium. Twenty-four hours later, the medium was discarded and non-invaded cells were removed by a cotton swab. The bottom of the chamber was fixed by methanol solution for 15 min and dyed with crystal violet for 10 min. The cells that had invaded were counted at 5 randomly chosen visual fields under a microscope (Smart zoom 5, Zeiss).

Cell samples were seeded in 6-well plates (1 × 10⁶ cells/well) for reaching 100% cell confluence, and then treated with 200-μL pipette tip. The scratched cells were removed, the serum-free medium was added, and then the cells were cultivated in an incubator at 37 °C with 5% CO₂. The width of 3–5 randomly selected spots at 24 h was recorded and the distance of wound closure was analyzed.

Total protein extracted from SCC9 and SCC15 cells was isolated by Radio Immunoprecipitation Assay (RIPA) lysis buffer and then subjected to isolation through electrophoresis using sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE;10%). The samples transferring on polyvinylidene fluoride (PVDF) membranes were blocked via 5% nonfat milk and then subjected to incubation with primary antibodies against ZO-1 (ab276131, Abcam, UK), E-cadherin (ab40772, Abcam, UK), N-cadherin (ab76011, Abcam, UK), Vimentin (ab92547, Abcam, UK), p50 (ab32360, Abcam, UK), p65 (ab16502, Abcam, UK), IκBɑ (#9242, Cell Signaling Technology, China), and GAPDH (China Kangcheng Biotechnology Co., Ltd.) as the internal control. After being rinsed in Tris-buffered saline-tween (TBST), samples were incubated with secondary antibody labeled with horseradish peroxidase (HRP) and finally exposed to electrochemiluminescent (ECL) luminescent liquid (Santa Cruz Biotechnology, Santa Cruz).

The circZDBF2-specific RNA FISH probe procured from Ribobio was used for cellular analysis as instructed by provider. The fixed cell samples were rinsed in PBS, and then air-dried, followed by the hybridization with FISH probe. Cell nuclei were stained using 4',6-diamidino-2-phenylindole (DAPI) solution purchased from Beyotime (Shanghai, China), and the fluorescent images were captured by GLomax20/20 fluorescence microscope (Promega).

RIP analysis was made via Magna RIP™ RNA-Binding Protein Immunoprecipitation Kit (Millipore, Bedford, MA) based on user guides. SCC9 and SCC15 cells were lysed by RIP buffer and cell lysates were collected. Magnetic beads were conjugated with argonaute2 antibody (anti-AGO2; ab186733, Abcam, UK) or immunoglobulin G antibody (anti-IgG; ab205718, Abcam, UK). After rinsing, the precipitated RNA went through qRT-PCR analysis.

ChIP was conducted utilizing EZ-ChIP chromatin immunoprecipitation kit (Millipore, Bedford, USA) in accordance with user guides. SCC9 and SCC15 cells were cross-linked in 4% paraformaldehyde, sonicated into chromatin fragments of 200–1000-bp and incubated with the antibodies against CEBPB (ab264305, Abcam, UK), with Anti-IgG served as the negative control. The specific primers for the RNF145 promoter were used, and precipitated DNA was detected by qRT-PCR.

Based on the protocols of supplier, we utilized the Pierce Magnetic RNA–Protein Pull-Down Kit (Thermo Fisher Scientific, Waltham, MA) to perform this assay. Biotin-labeled circZDBF2 (Bio-circZDBF2) or RNF145 (Bio-RNF1452) probes were incubated with cell extracts and streptavidin magnetic beads, with Bio-NC as the negative control. Finally, the RNA complexes bound to beads were analyzed by qRT-PCR. In this study, three experimental groups (Bio-NC, Bio-circZDBF2/RNF145-WT and Bio-circZDBF2/RNF145-Mut) were constructed respectively to evaluate the binding capacity between circZDBF2/RNF145 and miR-362-5p/miR-500b-5p.

The wild-type (WT) and mutant (Mut) miR-362-5p or miR-500b-5p binding sites within circZDBF2 sequence or RNF145 3′UTR (3′ untranslated region) were sub-cloned into pmirGLO dual-luciferase vector, and pmirGLO-circZDBF2-WT/Mut or pmirGLO-RNF145 3’UTR-WT-Mut plasmids were hence constructed. The pmirGLO plasmids were co-transfected with miR-362-5p mimics or miR-500b-5p mimics and their NC mimics in SCC9 and SCC15 cells. Forty-eight hours later, the luciferase activity was estimated with Dual-Luciferase Reporter Assay System (Promega Corporation, Fitchburg, WI).

Total of 2 × 10⁶ SCC9 cells transfected with sh-NC or sh-circZDBF2-1 were injected into the male BALB/c nude mice (6–8-week old; total number = 8 mice; n = 4 mice/group; Slac Laboratories, Shanghai, China). The animal assay was performed strictly in line with the protocol approved by the Ethical Committee of Affiliated Jinling Hospital, Medical School of Nanjing University, with the approval numbered 2020DZGZRZX-078. Tumor volume was calculated every 3 days following the formula: volume = length × width²/2. Four weeks later, mice were sacrificed after which tumors were excised and weighed.

Paraffin-embedded tissues from in vivo tumor growth assay were first fixed by 4% paraformaldehyde, embedded in paraffin and cut into consecutive sections at 4-μm thick for IHC assay. Then sections were incubated with anti-Ki67 (ab15580, Abcam, UK) at 4 °C overnight, and then with secondary antibody for 30 min. Images were taken using an Olympus microscope (Olympus Corporation, Japan).

Each experiment was performed for three times in this research. Statistical analysis was made via SPSS 23.0. Mean ± standard deviation (SD) was used to show the results. After validating normal distribution via shapiro test and homogeneity of variance via Levene test, data were analyzed via Student’s t-test or one-way analysis of variance (ANOVA) for difference analyses. P < 0.05 was of great importance for statistically significant difference.

First of all, we discovered that hsa_circRNA_002141 (namely hsa_circ_0002141, here called circZDBF2) was highly expressed in oral cancer cell lines (HSC3, Sa3 and SAS) compared to human normal oral keratinocytes (HNOKs) according to GSE145608 (Additional file 1: Fig. S1A). The discovery interested us to probe into the probable role of circZBF2 in OSCC. Before that, circZDBF2 expression was assessed in OSCC cell lines (SCC9, SCC15, SCC25) and a healthy human oral keratinocyte (HOK) cell line. Results from qRT-PCR showed that circZDBF2 was overexpressed in OSCC cells (especially in SCC9 and SCC15 cells) compared with normal cells (Fig. 1A). Convergent primers were constructed to amplify linear ZDBF2 and divergent primers to amplify circZDBF2, with cDNA (complementary DNA) and gDNA (genomic DNA) extracted from cells as templates. It was shown that circZDBF2 was amplified from cDNA only by divergent primers (Fig. 1B). Further, unlike linear ZDBF2 mRNA, circZDBF2 showed no evident change in OSCC cells after being treated with RNase R (Fig. 1C), suggesting the circular structure of circZDBF2. Moreover, we added Act D (a transcription inhibitor) in SCC9 and SCC15 cells and found that circZDBF2 was more stable than liner ZDBF2 (Fig. 1D). These results verified the cyclization of circZDBF2. We further analyzed the cellular distribution of circZDBF2 by FISH assay, which indicated that circZDBF2 existed in the nucleus as well as the cytoplasm of OSCC cells (Fig. 1E). To sum up, circZDBF2 with a loop structure is with high expression in OSCC cells.

By transfecting sh-circZDBF2-1/2/3 plasmids into SCC9 and SCC15 cells, we obtained the reduced circZDBF2 expression (Additional file 1: Fig. S1B), and sh-circZDBF2-1/2 with better interference efficiency was chosen for further investigations. According to the results of CCK-8 assay, circZDBF2 knockdown substantially reduced the optical density (OD) value at 450 nm (Fig. 2A), implying that cell viability was repressed by silencing circZDBF2. Further, it was manifested from colony formation assay that cell proliferative capability was suppressed upon circZDBF2 depletion, as the number of cell colonies was obviously declined in the sh-circZDBF2-transfected groups (Fig. 2B). In addition, we performed transwell assay and found that the down-regulation of circZDBF2 caused a remarkable reduction in the number of invaded cells (Fig. 2C). Moreover, data of wound healing assay indicated that the relative wound width was increased in cells with circZDBF2 down-regulation (Fig. 2D), suggesting that cell migration was suppressed by circZDBF2 inhibition. Next, it was found that EMT phenotype was repressed by circZDBF2 knockdown (Fig. 2E). Western blot as well as qRT-PCR data further proved that circZDBF2 down-regulation elevated ZO-1 and E-cadherin expression while caused a decline on N-cadherin and Vimentin expression (Fig. 2F), demonstrating that EMT process was significantly hindered by circZDBF2 depletion in OSCC cells.

At the same time, we detected the effect of circZDBF2 up-regulation on OSCC cells after confirming the overexpression of circZDBF2 in SCC9 and SCC15 cells by transfecting pcDNA3.1-circZDBF2 (Additional file 1: Fig. S1C). Results of CCK-8 assay as well as colony formation assay manifested that circZDBF2 upregulation promoted cell viability and proliferation (Additional file 2: Fig. S2A, B). Further, cell invasion and migration was facilitated by up-regulating circZDBF2, as evidenced by transwell as well as wound healing assay (Additional file 2: Fig. S2C, D). Lastly, it was proved that EMT process was accelerated in the pcDNA3.1-circZDBF2 transfected cells (Additional file 2: Fig. S2E, F). Taken together, circZDBF2 accelerates the malignant cell behaviors in OSCC.

The competing endogenous RNA (ceRNA) network is a famous post-transcriptional regulatory mechanism of RNA interaction in cancers [20]. CircRNAs have been demonstrated to act as ceRNAs to sponge miRNAs and then modulate their downstream target genes, thereby accelerating or inhibiting cancer progression [21]. The result of AGO2-RIP assay manifested that circZDBF2 was enriched in anti-AGO2 group (Fig. 3A, B), suggesting that circZDBF2 can function as a ceRNA in OSCC. Next, we continued to discover the possible miRNAs of circZDBF2. Through ENCORI (http://​starbase.​sysu.​edu.​cn/​index.​php) database (low stringency), two miRNAs (miR-362-5p and miR-500b-5p) had the possibility to combine with circZDBF2 (Fig. 3C). As shown from RNA pull down assay, the biotinylated circZDBF2-WT probe abundantly pulled down miR-362-5p and miR-500b-5p (Fig. 3D). Moreover, the overexpression of miR-362-5p or miR-500b-5p in SCC9 and SCC15 cells reduced the luciferase activity of circZDBF2-WT but did not affect circZDBF2-Mut group (Fig. 3E, F). In short, circZDBF2 directly combined with miR-362-5p and miR-500b-5p in OSCC cells.

Through ENCORI database with the specific condition (strict stringency; 10 cancer types), three possible mRNAs were observed to be shared by miR-362-5p and miR-500b-5p, which were RNF145, SAMD4A (sterile alpha motif domain containing 4A) and ATP2B4 (ATPase plasma membrane Ca2+ transporting 4) (Fig. 4A, B). Among them, only RNF145 was declined by circZDBF2 silencing, while the expression of the other two mRNAs was not affected (Fig. 4C). Therefore, we selected RNF145 for later experiments. qRT-PCR data further proved that RNF145 expression was declined by miR-362-5p overexpression or miR-500b-5p overexpression in SCC9 and SCC15 cells (Fig. 4D), which supported the ceRNA mode. Next, the binding capacity between RNF145 and miR-362-5p/miR-500b-5p was evidenced by RNA pull down result (Fig. 4E). Subsequently, it was discovered that the up-regulation of miR-362-5p or miR-500b-5p reduced the luciferase activity of RNF145 3’UTR-WT rather than the mutant group (Fig. 4F, G). It was shown that circZDBF2 knockdown did not affect miR-362-5p or miR-500b-5p expression (Additional file 3: Fig. S3A). Further, RNF145 expression inhibited by circZDBF2 knockdown was partially rescued by the co-inhibition of miR-362-5p and miR-500b-5p in OSCC cells (Fig. 4H), suggesting that circZDBF2 might have another pathway to regulate RNF145 level. Overall, RNF145 is the target gene of miR-362-5p and miR-500b-5p in OSCC cells.

Since miR-362-5p and miR-500b-5p could not completely rescue the regulatory effect of circZDBF2 on RNF145, we speculated that circZDBF2 may have other pathways to regulate RNF145. We had verified that circZDBF2 was also distributed in the nucleus of OSCC cells, and it was manifested through dual luciferase reporter gene experiments that circZDBF2 up-regulation elevated the luciferase activity of RNF145 promoter (Fig. 5A), so we suspected that circZDBF2 in the nucleus might regulate RNF145 transcription. We utilized UCSC (http://​genome.​ucsc.​edu/​) website to predict the possible transcription factors for RNF145 and found that MYC (MYC proto-oncogene), E2F6 (E2F transcription factor 6), FOXA1 (forkhead box A1), FOXA2 (forkhead box A2), FOS (proto-oncogene c-Fos), ZNF263 (zinc finger protein 263), GABPA (GA binding protein transcription factor subunit alpha), USF1 (upstream transcription factor 1), TCF12 (transcription factor 12), TFAP2A (transcription factor AP-2 alpha) and CEBPB may bind to RNF145 (Fig. 5B). In accordance with the existing studies, we discovered that MYC [22], E2F6 [23], FOXA1 [24] and CEBPB [25] expressed at a high level in OSCC cell lines. Therefore, we selected these four transcription factors for further screening. Through qRT-PCR, RNF145 expression was only declined by CEBPB down-regulation rather than the other candidates (Additional file 4: Fig. S4A). Further, RNA–Protein Interaction Prediction (RPISeq; http://​pridb.​gdcb.​iastate.​edu/​RPISeq/​) was applied to predict the potential interaction between circZDBF2 and CEBPB (Fig. 5C). Based on these results, we speculated that circZDBF2 affected RNF145 transcription through interacting with CEBPB. We up-regulated CEBPB expression via transfecting the pcDNA3.1-CEBPB plasmids into OSCC cells (Additional file 1: Fig. S1D), and it was uncovered that CEBPB overexpression significantly elevated the luciferase activity of RNF145 promoter (Fig. 5D). Data of ChIP assay further showed that RNF145 promoter was enriched by anti-CEBPB (Fig. 5E), which further proved the binding of CEBPB to RNF145 promoter. Next, JASPAR (http://​jaspar.​genereg.​net/​) database was utilized to predict the binding sites of CEBPB to RNF145 promoter, and three reliable CEBPB-binding sites (Site#1/2/3) were exhibited (Additional file 4: Fig. S4B). Further, it was revealed that CEBPB overexpression increased the luciferase activity of RNF145 promoter-WT, RNF145 promoter/Site#2-Mut and RNF145 promoter/Site#3-Mut while RNF145 promoter/Site#1-Mut group was not affected (Fig. 5F), indicating that CEBPB bound to RNF145 promoter at Site 1. Next, the interactivity between CEBPB and circZDBF2 was certified through RNA pull down and RIP assays (Fig. 5G, H). Interestingly, we disclosed that after the knockdown of circZDBF2, the enrichment of RNF145 promoter precipitated by CEBPB antibody decreased significantly (Fig. 5I). Furthermore, it was shown that RNF145 expression was repressed by circZDBF2 depletion or increased by CEBPB overexpression, and the decreased RNF145 expression caused by circZDBF2 knockdown was countervailed by the addition of CEBPB (Fig. 5J). In short, circZDBF2 enhances RNF145 transcription by recruiting CEBPB.

It was shown from qRT-PCR assay that RNF145 expressed at a higher level in OSCC cells than in the normal HOK cell line (Additional file 5: Fig. S5A). We separately inhibited RNF145 expression via sh-RNF145-1/2/3 transfection or elevated RNF145 expression via pcDNA3.1-RNF145 transfection before conducting loss- and gain-of-function experiments (Additional file 1: Fig. S1E, F). Through CCK-8 along with colony formation assay, it was uncovered that the down-regulation of RNF145 obviously repressed cell proliferation, while the overexpression of RNF145 promoted cell proliferation (Additional file 5: Fig. S5B, C and Additional file 6: Fig. S6A, B). Subsequently, cell invasion and migration was suppressed by silencing RNF145, while accelerated by overexpressing RNF145 (Additional file 5: Fig. S5D, E and Additional file 6: Fig. S6C, D). Likewise, we discovered that the knockdown of RNF145 exerted an inhibitory effect on EMT phenotype, while up-regulation of RNF145 facilitated EMT process (Additional file 5: Fig. S5F, G and Additional file 6: Fig. S6E, F). Altogether, RNF145 exerts a tumor-facilitating function in OSCC.

Finally, the regulatory effect of circZDBF2/RNF145 axis on OSCC progression was testified through rescue functional assays. We observed that the proliferation of OSCC cells was suppressed by the transfection of sh-circZDBF2-1, while such effect was totally recovered by the co-transfection of sh-circZDBF-1 and pcDNA3.1-RNF145 (Fig. 6A, B). Next, results of transwell assay as well as wound healing assay manifested that cell invasion and migration inhibited by circZDBF2 knockdown was fully rescued by RNF145 up-regulation (Fig. 6C, D). Moreover, RNF145 overexpression counteracted the inhibitory effect of silencing circZDBF2 on EMT process (Fig. 6E, F). In a word, circZDBF2 facilitates the progression of OSCC via up-regulating RNF145.

It has been reported that ring finger protein 183 (RNF183) can induce the activation of protein p65 in the NFκB signaling pathway to regulate the transcription of IL-8 [26]. IL-8 has been shown to promote the tumorigenesis of OSCC [27, 28]. RNF145 and RNF183 are homologous proteins, so we speculated that RNF145 can also induce p65 so as to activate the NFκB signaling pathway and regulate the transcription of IL-8. Through western blot along with qRT-PCR, we found that the down-regulation of RNF145 repressed the protein level of p65 while increased that of IκBɑ. By contrast, the protein level of p65 was promoted while that of IκBɑ was inhibited by the up-regulation of RNF145 (Additional file 7: Fig. S7A, B). It was then shown by qRT-PCR that IL-8 expression was declined in sh-RNF145-transfected OSCC cells while increased in pcDNA3.1-RNF145-trnasfected cells (Additional file 7: Fig. S7C, D). Further, it was demonstrated from luciferase reporter assays that the luciferase activity of IL-8 promoter was elevated by the upregulation of wild-type RNF145 rather than mutant RNF145 (Additional file 7: Fig. S7E), suggesting that RNF145 promoted the transcription of IL-8 in OSCC. In a word, RNF145 activates NFκB pathway to transcriptionally facilitate IL-8 expression.

We detected the influence circZDBF2 may exert on OSCC tumor growth by xenotransplantation assay. Results manifested that tumor growth tendency was slower in the group with circZDBF2-silenced OSCC cells than in the negative control group (Fig. 7A). Besides, the tumor size along with tumor weight was decreased in the sh-circZDBF2 transfection groups (Fig. 7B). The expression of Ki67 (the nuclear antigen representing cell proliferation) was detected by immunohistochemistry, which showed that the positivity of Ki67 was decreased by circZDBF2 downregulation (Fig. 7C). Furthermore, the expression of RNF145, p65 and IL-8 was decreased in tumors with inhibited circZDBF2 expression (Fig. 7D). In conclusion, circZDBF2 accelerates in vivo OSCC tumor growth by regulating RNF145, p65 and IL-8.