Regulation of Abnormal Expression of hsa_circRNA_0127523 on Proliferation, Migration and Invasion and mRNA Expression of miR-515-5p in Oral Squamous Cell Carcinoma

Background: Oral squamous cell carcinoma (OSCC) is one of the most common cancer. Circular RNA (circRNA) play an important role in many biological processes of occurrence and development of cancers. However, the mechanism of circRNA in OSCC remains unclear. We successfully established a Chinese hamster oral squamous cell cancer animal model and discovered the differential expression of non-coding RNAs. In this study, we investigated the functional roles of a signicantly up-regulated circRNA (hsa_circ_0127523) in OSCC. Methods: DMBA-induced malignant-transformed buccal sac tissue and normal tissue of Chinese hamster were analyzed by the second-generation circRNA sequencing. The top 8 different expression circRNAs in sequencing results were further conrmed by qRT-PCR. The RNAi, CCK-8, transwell migration and invasion assays were conducted to investigate the behavioral functions of hsa_circ_0127523 in OSCC cells. Results: In this study, we found total of 3485 circRNAs were differentially expressed. GO and KEGG analyses showed that differently expressed circRNAs function and pathway. The axis relationship of circRNA-miRNA was predicted by miRanda. Silencing hsa_circ_0127523 signicantly suppressed the proliferation, migration and invasion abilities of OSCC cells. Furthermore, bioinformatics databases suggested that hsa_circ_0127523 might function by sponging miR-515-5p to regulate the expression of TRIP13. It was subsequently identied that hsa_circ_0127523 may inhibit miR-515-5p expression and increase TRIP13 expression in the OSCC cells. Conclusions: In summary, the results of the present study revealed that OSCC tissues have abundant circRNAs and we rstly explore the role of the hsa_circ_0127523 in the behavior regulation of OSCC cells. Its demonstrating that hsa_circ_0127523 may be a potential biomarker and therapeutic OSCC. Oral squamous cell carcinoma; circRNA: circular RNA; miRNA: microRNA; lncRNA: long non-coding RNA; ceRNA: competitive endogenous RNA; STEM: Scanning Transmission Electron Microscopy;RPM: read per million; GO: gene ontology; KEGG: Kyoto Encyclopedia of Genes and Genomes pathway; MREs: miRNA response elements; siRNA: small interfering RNA; si-NC: negative control siRNA; qRT-PCR: quantitative real-time polymerase chain reaction; GAPDH: glyceraldehyde-3-phosphate dehydrogenase; CCK-8:Cell Counting Kit-8; OD: optical density; MF:molecular function; CC: cellular component; BP process; TRIP13:Thyriod hormone receptor interacting protein 13; CDK: Cyclin-dependent kinase; CDKI: cyclin-dependent kinase inhibitor.

Non-coding RNA is a type of RNA transcript that cannot encode proteins after transcription, included microRNA (miRNA), long non-coding RNA (lncRNA) and circular RNA (circRNA) [8,9]. They constitutes a competitive endogenous RNA (ceRNA) regulatory network. CircRNA is a novel special non-coding RNA, compared to traditional RNA, it's a covalently closed-loop RNA with no 3' polyadenylation tail and 5' cap, which is resistant to ribonuclease degradation and stably present in the cytoplasm and nucleus [10][11][12][13]. Besides, circRNA exhibit tissue-speci c and highly conservative which is another new hot research in the non-coding RNA family following the reporting of miRNA and lncRNA. There are four mechanisms about circRNA: the rst is by acting as miRNA "molecular sponge" to regulate gene expression, the second is by regulating gene transcription, the third is interaction with protein and the fourth is participating in protein translation [10,14,15]. At present, there are researches on relevant mechanisms that Zhang et al has found circular AKT3 RNA encoded a novel tumor suppressor protein by competing with active phosphoinositide-dependent kinase-1 inhibits glioblastoma tumorigenity [16]. In addition, circ_SMARCA5 regulates glioblastoma multiforme through sponge SRSF1 protein [17]. What's more, Xu et al has con rmed that the circTADA2As/miR-203a-3p/SOCS3 pathway modulates the progression and metastasis of breast cancer, and this pathway might be a promising target for the development of therapeutics for breast cancer [18].
Based on the above considerations, circRNA plays crucial roles in cancer by functioning as miRNA sponges and regulating mRNA translation. Therefore, we established dynamic oral squamous cell carcinogenesis animal model using the Chinese hamsters, and acquired circRNA differential expression pro ling of OSCC by the next generation high-throughput sequencing, veri ed several signi cant differentially expression circRNA molecules which consistent with sequencing results. Additionally, we screened circRNA hsa_circ_0127523 (circMAN2A1) was markedly overexpressed by combine highthroughput sequencing with PCR veri cation. According to reports from related studies, Wang et al has found circDOCK1 affects the apoptosis of oral squamous cell carcinoma through the circDOCK1/miR-196a-5p/BIRC3 axis. The knockdown of the expression of circDOCK1 with small interfering RNA led to the increase of apoptosis. In addition, the protein expression level of BIRC3 was decreased when interfering with circDOCK1 expression in oral cancer cells of CAL27, SCC9 and SCC25 [19]. Hence, we explored the role of hsa_circ_0127523 in vitro OSCC cells. We designed and synthesized two siRNAs for the sequence near the cleavage site of hsa_circ_0127523, which were transiently transfected them into the OSCC cell lines to detect the expression levels of downstream target genes and proteins.

Construction of animal model
Chinese hamsters were bred in a barrier animal laboratory facility of the Center for Experimental Animal of Shanxi Medical University [SCXK (Jin) 2017-0001]. Animals were individually housed in cages with standard laboratory conditions at 23°C-25°C, 45%-55% humidity, controlled environment with a 12-hour light/dark cycle. They were divided into three groups: control group (n=30), negative control group (n=12) and experimental (model) group (n=48). The control group was not any treated, the negative control group was coated with acetone and the model group was coated with acetone-dissolved 9, 10-Dimethy1-1, 2-Benzanthracence (DMBA) [20]. Animal experiments were performed at the Experimental Animal Center of Shanxi Medical University and carried out strictly in accordance with the operating rules formulated by the Institutional Animal Care and Use Committee of Shanxi Medical University (IACUC 2017-016). In the course of experiment, the Chinese hamsters were fasted for 2 hours after the application, and the rest was free to eat and drink. According to the 3R principle (Reduction, Replacement and Re nement) used by the experimental animals, humane care is given. The model was created with DMBA which applied every other day. Anatomical samples were performed regularly at weeks 6, 9, 12, 15, 18, and 21: the left and right cheek pouches of the Chinese hamster were taken, which one of two buccal sac tissues were marked and placed in liquid nitrogen for freezed immediately, and then at -80°C in the refrigerator which stored in preparation for RNA extraction; one quarter buccal sac tissues were xed in 4% paraformaldehyde for 24-72 hours, then subjected to alcohol gradient dehydration, para n embedding; one quarter buccal sac tissues were quickly placed in a glutaraldehyde xative solution and xed at 4°C (Maximum storage time of one month). It is noteworthy that the Chinese hamsters were anesthetized with 0.3% pentobarbital sodium solution (1.0mL / 100g body weight ratio to determine the dosage) by intraperitoneal injection before taking the buccal sac tissues, and sodium pentobarbital dose was added to euthanize after the buccal capsule was collected.
HE staining to detect pathologic structure The para n-embedded tissues were subjected to alcohol gradient and xylene dewaxing treatment, 4 m longitudinal sections were stained with hematoxylin solution for 3 minutes, then rinsed in distilled water twice. Immerse the slice in the 1% acid ethanol (1% HCl in 70% ethanol) for 1 minute and rinse with distilled water. Then the sections were stained with eosin solution for two minutes and thirty seconds and followed by rinsing in distilled water and dehydration with graded alcohol and clearing in xylene. Then the pathological results were observed under a uorescence microscope (IX70 Olympus).

Ultrastructural observation by Scanning Transmission Electron Microscopy (STEM)
The buccal sac tissue xed by glutaraldehyde was immersed in phosphate buffer solution, then immobilized in osmium tetroxide solution, dehydrated by ethanol gradient, and embedded in epoxy resin No. 812. After polymerization, the buccal sac tissue was sliced by ultrathin slice machine, stained with uranyl acetate and lead citrate, and observed by transmission electron microscopy under a microscope.

RNA extraction and quality test
Total RNAs were extracted from frozen tissues and OSCC cells using miRNA easy Mini Kit External content QIAzol Lysis Reagent , according to the manufacturer's instructions. Then the RNA qualities were evaluated by NanoDrop ND-1000 spectrophotometer (Thermo Fisher Scienti c, Waltham, MA, USA). The RNA purity is quali ed by the OD260/OD280 ratio, which ranges from 1.8 to 2.0, and RNA integrity was assessed by 1% agarose gel electrophoresis analysis. Total miRNA was extracted from the OSCC cells using miRNA extraction and puri cation kits (Transgen, Beijing, China), preserved at -80℃ before using.

Library construction and sequencing
After RNA extraction and quality test, a quantity of 1 μg total RNA of tissues was constructed library, using TruSeq Small RNA Library Prep Kit for Illumina following the manufacturer's protocol. Removed ribosomes RNAs, enriched circRNAs were short fragment by using fragmentation buffer. First strand cDNA was synthesized using reverse transcriptase with random hexamer primers. Second strand cDNA was synthesized by DNA polymerase I, dNTP, RNase H and buffer. Next, the cDNA fragments were puri ed with AMPure XP beads. The puri ed double-stranded cDNA was repaired end, poly(A) added and ligated to Illumina sequencing connector, then, AMPure XP beads were used to select the fragment size, the target fragment was recovered by agarose gel electrophoresis. The obtained fragment was subjected to PCR ampli cation and enrichment, the cDNA library was constructed. The constructed cDNA library was quanti ed using Qubit2.0 and diluted to 1.5 ng/μL. Which was tested by Agilent 2100 Bioanalyzer and ABI StepOne plus Real-time PCR System. After passing the quality inspection, the effective concentration of the library was quanti ed accurately by qPCR (the effective concentration of the Library > 2nM) to ensure the quality of the library. Sequenced using Illumina Hiseq2500 SR50 (Hangzhou, China). Each sample was blasted to the genome by quality-controlled PE clean reads data and the Chinese hamster reference genome (GCF_000223135.1_GriGri_1.0_genomic.fa, NCBI). The comparison software uses the BWA-MEM algorithm required by CIRI, a circular RNA identi cation tool. The algorithm is fast and accurate and is suitable for long sequence genome alignment.
CircRNA expression standardization and data analysis All raw reads were obtained from the sequencing machines. The absolute expression of circular RNA is generally measured by detecting the number of sequences aligned to the head-to-tail locus, but since the total amount of sequencing of each sample library is different, absolute quanti cation will result in a preference for expression, so standardization is required. The RPM (read per million) method commonly used in data analysis was used to normalize the expression level of circular RNA. According to the screening threshold of RPM>0.1 in the literature, a total of 3485 circRNAs were obtained that satis ed the expression requirements. Differential expression analysis of model and control groups was performed using edgeR package. We identi ed circRNAs with log 2 (Fold change)>2 or log 2 (Fold change) <2 and Pvalue <0.05 as signi cant differentially expressed circRNAs which are up-regulated or down-regulated respectively, then circRNAs were annotated with Cricetulus griseus genome. Those cannot be annotated were de ned as novel circRNAs.

Functional enrichment and pathway analysis
During functional enrichment analysis of differentially expressed circRNAs, the online analysis tool-Metascape (http://metascape.org/gp/index.html#/main/step1) was utilized to perform gene ontology (GO) enrichment analysis. Kyoto Encyclopedia of Genes and Genomes pathway (KEGG) analysis associated genes of differentially expressed circRNA.
Integrated analysis of circRNAs-miRNAs in Chinese hamster tissue The target relationship of annotated circRNAs with miRNAs were theoretically predicted by conserved seed-matching sequence using miRanda (v3.3a) analysis, which the maximum binding energy required is less than -20, and there are at least two miRNA binding sites on a circular RNA sequence. The analysis showed all circRNAs contained their respective miRNA response elements (MREs). To better understand the circRNA-miRNA network was visualized by Cytoscape 3.01.

Cells culture
The human cell lines CAL27 and Tca8113 were purchased from Procell and Boster Biological Technology (Wuhan, China). OSCC cell lines were cultured in DMEM medium (Boster Biological Technology, Wuhan, China) supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin. All cells were cultured at 37℃ with 5% CO 2 .

RNA interference and transient cell transfection
The small interfering RNA (siRNA) and negative control siRNA (si-NC) were synthesized by GenePharma (Shanghai, China) to inhibit the expression of hsa_circ_0127523 in CAL27 and Tca8113 cells. Si-hsa_circ_0127523 siRNA (si-hsa_circ_0127523-1:sense: 5 -ACAUACAGGGAGUGAAACCTT-3 ; antisense: Total RNA was reverse transcribed to cDNA by random hexamer primers (Takara, Tokyo, Japan). SYBR green PCR Master Mix (Takara, Tokyo, Japan) was used to examine circRNA, miRNA and mRNA expression, circRNAs and mRNAs using glyceraldehyde-3-phosphate dehydrogenase (GAPDH) as an endogenous control, miRNAs using U6 as an endogenous control, every group at least three tissues and repeat at least three times, which monitored by the ABI StepOne Plus Real-time PCR System (ABI 7300). The melting curve insure the speci city of primers. Relative circRNAs, miRNAs and mRNAs expression were calculated according to the 2 -ΔΔCT method. The gene sequence of buccal sac tissues and OSCC cells were derived from the sequencing results and UCSC database, and primers were designed using Oligo 6.0 software. The primers were synthesized by Shanghai Sangon Biotech. The primers was applied to qRT-PCR assay are showed in Table1-1 and Table1-2.
Cell Counting Kit-8 (CCK-8) assay The proliferation of CAL27 and Tca8113 cells were detected by the CCK-8 assay. Infected cells were plated into 96-well plates at 3×10 3 cells/well, ve replicate wells were used for each analysis. At 0, 24, 48 and 72 hours (times after seeding OSCC cells at 1, 2, 3 and 4 days respectively), according to the manufacturer's guidelines, 10 μL of CCK-8 liguid (Boster Biological Technology, Wuhan, China) was added adherence and incubation for 2 hours. The optical density (OD) was detected by a microplate reader at 450 nm.

Transwell migration and matrigel invasion assays
The migration and matrigel invasion assays were performed using 24-well transwell plates, using transwell chamber and pre-coated matrigel transwell chamber respectively (Corning, USA). The infected cells were suspension in DMEM without FBS and added to the upper chambers (2×10 4 cells in 200 μL medium), while the lower chambers were added DMEM including 10% FBS (700 μL) and incubated for 48 hours. Compared with the upper chambers of migration assay, the upper chamber of matrigel invasion assay was added with matrigel. After incubation 48 hours, the cells and matrigel in upper chambers were removed, the cells which migration or invasion were xed 30 minutes in 4% paraformaldehyde and stained 1 hour with 0.1% crystal violet, taken photos and counted.

Statistical analysis
All data were statistical analyses by SPSS 17.0 (USA) and GraphPad Prism 5.0 software (USA). Demonstrate signi cant difference circRNAs from high-throughput sequencing were analyzed by t-test between model and control groups. The results of PCR, CCK-8, migration and invasion were presented as mean ± standard error of the mean (SEM). Statistically signi cant data (P<0.05) were determined by Student t-test. The pairwise comparison between the multiple groups of samples was performed by the LSD method in one-way analysis of variance (ANOVA). *P<0.05, **P <0.01 and ***P <0.001 were considered as statistical signi cance.

HE staining of buccal sac tissues observe pathological structure
Under the light microscope of normal buccal sac tissue, the spine cell layer of normal buccal pouch tissue was thin, the basal layer cells were clearly arranged neatly and the nail processes were not obvious. The histological observations of the blank control group and the negative control group were consistent and no abnormal changes occurred ( Fig. 1A and B). The pathological structure of simple hyperplasia tissues group and abnormal hyperplasia tissues group is mild hyperemia of the buccal mucosa, pathological observation of atrophic cells thickening, basal layer cell proliferation and obvious changes, nail process is more obvious than the normal group ( Fig. 1C and D). At cancer group, local leukoplakia lesions and erosion, pathological observation of spinous layer cells in a single or clustered keratosis, basal layer cell thickened, most of the basal cells abnormally proliferated, a large parts of cell polarity disappeared and arranged scatteredly (Fig. 1E).

Transmission ultrastructure observation
In ultrastructural observation, the shape of cell is closely arranged, the cell nucleus shape is regular, and the morphology of each organelle is normal and the desmosome is abundant which in normal group ( Fig.  2A). Compared with the normal group, the morphology of the nucleus in some lesions changed, nucleoli edge gathered and desmosomes were relatively few in simple hyperplasia group (Fig. 2B); nuclei enlarged and condensed irregularly, desmosomes decreased mitochondrial edema in abnormal hyperplasia group (Fig. 2C); epithelial cells undergo canceration, cell shape changes, nuclear-cytoplasmic ratio decreases, and cancer cells invade connective tissue in squamous cell carcinoma group (Fig. 2D).
Based on the World Health Organization Cancer Diagnostic Criteria [21], oral precancerous lesion and oral squamous cell carcinoma are known to be induced in animal models after 9 and 15 weeks of drug treatment. Therefore, HE staining and ultrastructural observation of animal tissue samples con rmed the successful establishment of Chinese hamster animal model of OSCC. Thus, 12 samples were screened to high throughput sequencing, of which 3 samples were normal group, and the other 9 samples were simple hyperplasia group (n=3), abnormal hyperplasia group (n=3) and squamous cell carcinoma group (n=3).

CircRNA expression pro le in OSCC and bioinformatic analysis
Heatmap of sample correlation intutively re ects the difference between the two groups of samples and whether there is good sequencing stability between the three replicates of a treatment group. The darker the color, the stronger the correlation (Fig. 3). Hierarchical clustering showed that circRNA expression pattern was distinguishable between simple hyperplasia group, abnormal hyperplasia group and carcinoma group (Fig. 4). The expression analysis also revealed that a series of circRNAs was differentially expressed in simple hyperplasia tissues, abnormal hyperplasia tissues and carcinoma tissues compared with control tissues (Fig. 5). A total of 3485 distinct circRNAs were identi ed in these buccal sac tissues. Among these, a little part of circRNAs were novel (n=229). Eighty-nine circRNAs were identi ed to obviously express differentially between model and control groups, among which 38 circRNAs were up-regulated (log 2 (Fold change) >2, P-value <0.05) and 51 were down-regulated (log 2 (Fold change) <2, P-value <0.05), which 20 novel circRNAs.

Predicted functions and pathways of differential expression circRNAs
Since there is no annotation information for circRNA at present, and the annotation information of the source gene is relatively complete, we have functional annotation of the source gene of the circRNA. We select differentially expressed circRNA by statistical analysis, and enrich the source genes of differentially expressed circRNA by using GO and KEGG databases. The GO enrichment analysis revealed these differently expressed circRNAs were involved in several vital molecular function (MF), cellular component (CC) and biological process (BP) (Fig. 6). The KEGG analysis revealed that the host genes of differently expressed circRNA in different pathways (Wnt signaling pathway, PI3K-Akt signaling pathway, MAPK signaling pathway TGF-β signaling pathway, ErbB and mTOR signaling pathway and so on) (Fig. 7).
Prediction of circRNA-miRNA interaction and network visualization Target miRNAs were predicted from the most signi cant circRNAs screened. Based on the maximum score predicted by miRanda (3.3a), miRNAs were screened by positional position and maximum binding energy. The results were displayed in the form of network graph by Cytoscape 3.01 (Fig. 8).

Validation of differential expression circRNAs
To validate the differential expression circRNAs from the high-throughput sequencing results, which by quantitative real-time PCR (qRT-PCR) in tissue samples including the tissues that from high-throughput sequencing analysis. These include four up-regulated (cgr_circ_Gnaq, cgr_circ_Znf827, cgr_circ_Man2a1 and cgr_circ_Pdcd4) and four down-regulated (cgr_circ_Zeb1, cgr_circ_Me1, cgr_circ_Ankrd12 and cgr_circ_Akap7) circRNAs. The results showed that eight circRNA molecules were signi cantly dysregulation expression in simple hyperplasia tissues, abnormal hyperplasia tissues and carcinoma tissues (Fig. 9), which are the same as the results of sequencing (in Table2).
Elevated transfection e ciency of si-hsa_circ_0127523-1 and si-hsa_circ_0127523-2 by qRT-PCR in OSCC cells Hsa_circ_0127523 is encoded by the MAN2A1 gene and located on chromosome 5 at chr5:109103235-109159538 and has a spliced sequence length of 1731 bp from UCSC and circBank database. The transfected cells were incubated for 48 hours, and the expression level of hsa_circ_0127523 was the gold standard for detecting the transfection e ciency. Our results showed that the expression level of hsa_circ_0127523 in the experimental group (Transient transfection si-hsa_circ_0127523-1 and si-hsa_circ_0127523-2 in OSCC cells) was signi cantly lower than that in the negative control group (Transient transfection si-NC in OSCC cells) (P<0.05), indicating that the interference e ciency was achieved (Fig. 12B).

Si-hsa_circ_0127523 inhibits proliferation ability of OSCC cells in vitro
The effect of silencing hsa_circ_0127523 (si-hsa_circ_0127523-1 and si-hsa_circ_0127523-2) on the proliferation of CAL27 and Tca8113 cells was detected by CCK-8 assay, and the growth curve was drawn (Fig. 10). Our results showed that, the cell proliferation ability of CAL27 (Fig. 10A) and Tca8113 (Fig. 10B) cells after 3 and 2 days of silence respectively in hsa_circ_0127523 was signi cantly inhibited, compared with the control cells (P<0.05).

Si-hsa_circ_0127523 inhibits migration and invasion of OSCC cells in vitro
To examine the potential role of hsa_circ_0127523 in OSCC cell migration and invasion, we performed transwell assays in CAL27 and Tca8113 cells transfected with si-NC, si-hsa_circ_0127523-1 and si-hsa_circ_0127523-2. Compared with the negative control group, the results showed that si-hsa_circ_0127523-1 and si-hsa_circ_0127523-2 signi cantly inhibited migration (Fig. 11A) and matrigel invasion (Fig. 11B) of CAL27 and Tca8113 cells after 24 hours of culture (P<0.05), indicating hsa_circ_0127523 promoted the migration and invasion of CAL27 and Tca8113 cells.
Hsa_circ_0127523 may interact with miRNA and mRNA In order to explore the mechanism and function of hsa_circ_0127523, investigated the potential interaction between circRNAs and miRNAs. By the online tools of CircInteractome, miRanda and Targetscan, we found that hsa_circ_0127523 may interact with hsa-miR-1200, hsa-miR-338-3p and miR-515-5p (Fig. 12A). Among them, the latest literature reports that miR-515-5p directly targeted TRIP13 3'-UTR and negatively correlated with TRIP13 mRNA expression in prostate cancer [22]. So, we examined the expression level of miR-515-5p and TRIP13 by qRT-PCR, found miR-515-5p was up-regulated and TRIP13 was down-regulated when silencing hsa_circ_0127523 (Fig. 12C, D). In addition, we found and tested the gene expression levels of Notch1 and p21 which by combine online software with literature.
Also, the results showed that Notch1 and p21 were down-regulated when silenced hsa_circ_0127523 ( Fig.  12E and F).

Discussion
OSCC is one of the most common malignant tumors in the head and neck cancer. Its incidence is increasing year by year in the world, and the population of OSCC tends to be younger, which poses a serious threat to human health and life. Therefore, early detection and treatment of OSCC should not be delayed. So we build a dynamic animal model of OSCC in different stages, which can make up for the di culty of obtaining samples in different clinical stages. In order to better prognosis treatment, it is important to elucidate the pathogenesis of OSCC. Non-coding RNA as a research hotspot in oncology, plays an important role in OSCC.
At present, the non-coding RNAs studied mainly include miRNA, lncRNA and circRNA [23]. Among them, miRNA and lncRNA are more mature, and circRNA is a hotspot in recent years. Compared with linear miRNA and mRNA, circRNA as a disease diagnostic marker has the unique advantages of high content and stable structure. Many studies on miRNAs and lncRNAs have been carried out and have demonstrated that these RNAs are related to oral squamous cell carcinoma proliferation, migration and invasion. Studies have shown that the miR-504 inhibiting cell proliferation, migration and invasion by regulating the expression of CDK6 in the process of OSCC [24]. Chen et al found that miRNA-10a promotes cancer cell proliferation in oral squamous cell carcinoma by regulating GLUT1 and promoting glucose metabolism [25]. The literature reports that the lncRNA LUCAT1 might promote OSCC migration and invasion by regulating PCNA [26]. Moreover, some factors also regulate the function of ncRNA, for instance, Chen et al has pointed out NKILA has been shown to bind to the nuclear factor-kB (NF-kB)/IkB complex, which masks the phosphorylation sites of IkB, thus preventing its degradation [27]. It is well known that circRNA is stably present in vivo and vitro due to its covalently closed-loop structure. Consequently, using circRNA as a therapy target would better than miRNA and lncRNA.
Based on the above considerations, we established and validated oral squamous cell carcinoma animal model using Chinese hamster. Then we used RNA-seq technology to analyze the expression pro les of circRNA in tissues of OSCC at three different pathological stages and normal control tissues, and screen signi cantly different circRNAs in OSCC tissues. Our results pointed out that the expression of circRNA is abundant in twelve samples, including 3485 existing circRNA molecules and 229 novel circRNA molecules, 38 circRNAs were up-regulated while 51 were down-regulated in model group compared with control group. We selected four up-regulated and four down-regulated circRNAs for qRT-PCR validated that accordance with the high-throughput sequencing results. To the best of our knowledge, most of circRNAs were differentially expressed, indicating that these differentially expresses circRNAs might be implicated in the incidence and progression of OSCC. GO function and KEGG pathway analysis of the host genes of the signi cantly different circRNAs were used to predict related functions through bioinformatics. Based on the correlation between circRNA and its host gene and its known function, GO annotation analysis results show that GO term is mainly enriched in epithelial cell proliferation, actinbased cell projection and protein domain speci c binding; KEGG further analysis shows that KEGG term is mainly enriched in Wnt, PI3K-Akt, MAPK, TGF-β, p53 and mTOR signaling pathways, revealing that the host genes of circRNAs may participate in the regulation of OSCC development through related molecular functions or signaling pathways. Besides important function and pathway, more and more studies have revealed that the mechanism of circRNAs sponge miRNAs [28,29]. Finally, 35 abnormally expression circRNAs were selected to predict the circRNA-miRNA regulatory network from the perspective of ceRNA, which the potential biological functions were analyzed from the perspective of possible molecular mechanisms and revealed the important regulatory functions of circRNA in OSCC pathology. The above bioinformatics prediction analysis shows that these signi cantly different circRNAs have important regulatory roles in the pathology of OSCC, and are involved in promoting or inhibiting the occurrence and development of OSCC.
Due to the conservation and tissue speci city of circRNA, we screened up-regulated expressed cgr_circ_MAN2A1 in tissues by high-throughput sequencing. Therefore, we combined with bioinformatics databases (circBase, circBank and UCSC databases) and bioinformatics analysis to screen the gene of MAN2A1 related circRNA, it was found that hsa_circ_0127523 is obtained by cleaving the pre-mRNA of MAN2A1 and the full-length sequence of hsa_circ_0127523 was gotten. By detecting the expression level of the gene hsa_circ_0127523, it was found that the expression level of hsa_circ_0127523 in oral squamous cell carcinoma cell lines was signi cantly higher than that of normal cells. It is necessary to further study whether hsa_circ_0127523 plays an important regulatory role in OSCC. According to Ouyang et al research reports that circRNA_0109291 is highly expressed in oral squamous cell carcinoma and the authors tested its functions and mechanisms by design to interfere with the circRNA_0109291 in OSCC [30]. Then, the behavioral function of hsa_circ_0127523 in CAL27 and Tca8113 cell lines shows that silencing hsa_circ_0127523 inhibited the proliferation, migration and invasion of OSCC cells in vitro.
It is well known that circRNAs play crucial roles in cancer by functioning as miRNA sponges, and affect the stability of target mRNA and regulate mRNA expression level [31,32]. The latest literatures report the mechanism of circRNA to miRNA is a competitive endogenous mechanism, also called sponge mechanism [33], and the expression of circRNA and mRNA is positively correlated, the expression of circRNA and mRNA is negatively correlated with the expression of miRNA, which formed "circRNAs-miRNAs-mRNAs" target regulation network [34][35][36]. Studies have shown that hsa_circRNA_100533 act as a competitive endogenous RNA (ceRNA) and regulates the expression of GNAS by sponging has-miR-933, thereby modulating the proliferation, migration and apoptosis of OSCC [37]. By using a circRNA microarray, Gao et al demonstrated that circ-PKD2 as a ceRNA regulates APC2 expression by sponge miR-204-3p and participates in OSCC cell proliferation, migration and invasion, induced apoptosis and cell cycle arrest [38].
In our next study, the interaction network of hsa_circ_0127523 showed that it was most likely to regulate the expression of miR-515-5p according to the bioinformatics analysis. Further interaction analysis showed that the target gene of miR-515-5p most likely to affect the TRIP13 gene regulation. Therefore, we detected the expression of miR-515-5p and TRIP13 at the gene level after hsa_circ_0127523 was silenced, and found that the expression level of miR-515-5p decreased and TRIP13 increased after silencing hsa_circ_0127523. Moreover, Zhang et al found that TRIP13 as a direct target of miR-515-5p by binding to it's 3'-UTR, which abolish the role of miR-515-5p in suppressing cell proliferation, migration and invasion that the miR-515-5p-TRIP13 axis may be a brightening prognostic marker and therapeutic target for prostate cancer [22]. TRIP13 (Thyriod hormone receptor interacting protein 13) is a nuclear factor, which is a protein-coding gene from the perspective of gene type. It is mainly involved in meiosis, cell proliferation and metastasis and is expressed in normal and cancerous tissues. Recent studies have found that it's related to tumorigenesis, development and prognosis [39,40]. Kurita et al [41] found through research that TRIP13 is highly expressed in cancer tissues of colorectal cancer patients and the deletion of the TRIP13 gene can inhibit cancer cell proliferation, migration and invasion, which may become a potential target for colorectal cancer treatment. In addition, TRIP13 gene is abnormally highly expressed in head and neck malignancies, multiple myeloma, chronic lymphocytic leukemia and lung adenocarcinoma [42][43][44], which may be related to the occurrence and development of malignant tumors.
There are no reports about the study of TRIP13 in OSCC. Therefore, this study hopes to provide valuable reference for clinical treatment from the perspective of TRIP13. Our study forecasted the sponge mechanisms of hsa_circ_0127523/miR-515-5p/TRIP13 axis in OSCC, which veri ed the differential expression level that are coincidence with competitive endogenous mechanism. In addition, the expression level of Notch1 and p21 were abnormal when inhibited the expression level of hsa_circ_0127523 in OSCC cells, current studies have suggested that hsa_circ_0127523 may regulate the occurrence and development of OSCC through Notch pathway. Role of TRIP13 has been reported in the literature, and related research validation has also suggested that TRIP13 gene affects the expression levels of Notch1 and p21 in the Notch pathway in ovarian cancer [22,45]. Analysis by Yoshida et al reported that Notch1 plays a vital role in the development and progress of OSCC. Based on deep sequencing of the cancer genome, research has also suggested the anti-cancer effect of Notch1. Notch plays an indispensable role in OSCC because Notch is considered to play an oncogene role in a subset of OSCC and also has a suppressive effect [46]. Lv et al has found that miR-495 restricts the proliferation and invasion of OSCC cells by directly targeting Notch1. Notch1 knockdown shows an inhibitory effect similar to that induced by miR-495 overexpression in OSCC cells [47]. Cyclin-dependent kinase (CDK) and cyclin-dependent kinase inhibitor (CDKI) can directly control the cell cycle. p21 is a negative regulator of cyclin-dependent kinases and is a negative checkpoint regulator of the cell cycle in this function [48].
Previous studies have shown that p21 is over expressed in lung cancer and gastrointestinal cancer cells [49,50].The results of Zhang's study show that the 5-year survival rate of OSCC patients with high p21 expression is signi cantly reduced. Kaplan-Meier analysis and Cox proportional hazard model showed that lymph node metastasis and p21 expression were independent prognostic factors of OSCC [51], which are consistent with our ndings that silenced the hsa_circ_0127523, the TRIP13 expression level was decreases and the Notch1 decreases in the Notch pathway.
Increasing studies suggest that circRNA has a vital part in a variety of tumors as a tumor regulator. As reported by Su that hsa_circ_0055538 could regulate the malignant biological behavior of OSCC through the p53/Bcl-2/caspase signaling pathway [52]. And it has been reported that circDOCK1 suppresses cell apoptosis via inhibition of miR-196a-5p by targeting BIRC3 in OSCC [53]. Chen et al found that circRNA_100290 was bound to miR-29 family members directly and co-expressed with CDK6 in uenced proliferation in oral cancer cell [54]. In non-small cell lung cancer, Huang et al demonstrated that hsa_circ_0001946 may serve as a novel biomarker for the diagnosis and prediction of platinum-based chemosensitivity in patients with NSCLC [55]. Sang et al has identi cated that hsa_circ_0025202 served an anti-oncogenic role in HR-positive breast cancer, could suppress tumor growth and enhance tamoxifen e cacy, and it could be exploited as a novel marker for breast cancer [56]. Based on these, we propose a view that circRNA may participate in development and prognosis of OSCC and our study showed that hsa_circ_0127523/miR-515-5p/TRIP13 by modulated Notch pathway may play an important role in OSCC cancer development. However, hsa_circ_0127523 has not been reported in the any disease, so it is concluded that hsa_circ_0127523 may be a potential biomarker of OSCC. In the next study, we will further elucidate the regulatory mechanism of hsa_circ_0127523/miR-515-5p/TRIP13 in OSCC.

Conclusion
In conclusion, our study provided evidence that oral circRNAs are differentially expressed in response to OSCC. In the expression signatures, we uncovered the circRNAs expression pro le in three dynamic stages by comparing it with normal tissue. Additionally, eight dysregulated circRNAs expression level in tissue were veri ed by qRT-PCR, which is consistent with the sequencing result. Additionally, we con rmed hsa_circ_0127523 was interfered signi cantly affected the proliferation, migration and invasion abilities of OSCC cells. More important, we predicted and con rmed hsa_circ_0127523 inhibit miR-515-5p expression and increase TRIP13 expression in the OSCC cells which provide potential axis for the next step diagnosis of OSCC and novel genetic insights, and the speci c mechanism still needs further research. Hsa_circ_0127523 may be a potential marker and therapeutic target for the diagnosis of OSCC.

Consent for publication
Not applicable.

Availability of data and materials
All data generated or analyzed during this study are included in this published article.

Competing interests
All of the authors have announced that there was no con ict of interest in this work.      Heatmap of sample correlation. The darker the color, the stronger the positive or negative correlation, and the number in the grid is the correlation coe cient.  CircRNA of GO function analysis. We select differentially expressed circRNA by statistical analysis, and enrich the source genes of differentially expressed circRNA by using GO databases. GO annotation analysis results show that GO term is mainly enriched in epithelial cell proliferation, actin-based cell projection and protein domain speci c binding.

Figure 7
The KEGG pathway enrichment in the differently expression circRNAs (partly). This pathway map indicate cancer-related Wnt, PI3K-Akt, MAPK, TGF-β, p53 and mTOR signaling pathways. The red and green frames represent the gene were up and down regulated respectively.