Tumor and adjacent tissues (2 cm distant from the tumor margin) were collected from OSCC patients who underwent surgical resection of the primary lesions at Hainan General Hospital (Haikou, Hainan). All of the patients underwent neither chemotherapy nor radiotherapy, and written informed consent was provided by all patients prior to participation. After radical resection, all tissues were snap-frozen in liquid nitrogen and immediately stored at −80°C until use. The histopathological quality of all tissue specimens was independently validated by two licensed pathologists, including the adjacent tissues (histopathological quality of all adjacent tissues were without cancer cells). After histopathological vetting, a total of 23 paired tumor and adjacent tissues were finally included in this study. The microarray chip assay was performed in 3 pairs of OSCC vs. adjacent tissues. The remaining 20 paired tissues were used for verification using quantitative real-time PCR (qPCR). The clinicopathological data of 20 patients is provided in Table I. This study was approved by the Ethics Committee of Hainan General Hospital.

The human OSCC cell line, TSCC1, was obtained from the Cell Bank of Type Culture Collection of Chinese Academy of Sciences (Shanghai, China). TSCC1 cells were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS) (Invitrogen; Thermo Fisher Scientific, Inc.) and maintained in a humidified air atmosphere containing 5% CO₂ at 37°C.

Total RNA was extracted from the tissue samples using TRIzol reagent (Invitrogen; Thermo Fisher Scientific, Inc.) following the manufacturer's instructions. A NanoDrop 2000 spectrophotometer (NanoDrop; Thermo Fisher Scientific, Inc.) was used to measure the quantity (ng/ml) and purity (260/280 and 260/230 ratios) of the RNAs. Specifically, OD260/OD280 ratios between 1.8 and 2.1, while OD260/OD230 ratios >1.8 were deemed acceptable. RNA integrity and DNA contamination were then assessed through electrophoresis on a denaturing agarose gel.

Sample labeling and array hybridization were performed according to the manufacturer's protocol (Arraystar Inc.). Briefly, Rnase R (Epicentre, Inc.) was used to digest total RNAs in order to degrade any linear RNAs and enrich circular RNAs. These enriched circular RNAs were amplified and transcribed into fluorescent cRNA using a random priming method (Arraystar Super RNA Labeling Kit; Arraystar). The labeled cRNAs underwent purify by RNeasy Mini Kit (Qiagen) and quantitation by NanoDrop ND-1000 (Thermo Fisher Scientific, Inc.). A total of 1 µg of each labeled cRNA was fragmented by adding 5 µl 10 X blocking agent and 1 µl of 25X fragmentation buffer, and then the mixture was heated at 60°C for 30 min, and finally diluted by adding 25 µl 2X hybridization buffer. Subsequently, 50 µl of hybridization solution was dispensed into the gasket slide and applied to the circRNA expression microarray slide. The slides were incubated for 17 h at 65°C in an Agilent Hybridization Oven (Agilent Technologies, Inc.). The hybridized arrays were washed, fixed and scanned using the Agilent Scanner G2505C (Agilent Technologies, Inc.).

The raw data were extracted by importing the scanned images by Agilent G2565C Microarray Scanner into Agilent Feature Extraction software (version 11.0.1.1; Agilent Technologies, Inc.). R software package limma package (http://bioconductor.riken.jp/packages/3.0/bioc/html/limma.html) was used to carry out a series of data processing including quantile normalization of raw data. Subsequently, differentially expressed circRNAs between the tumor and normal groups were evaluated using t-test, and the P-values were corrected for False Discovery Rate (FDR) by Benjamini-Hochberg (BH) procedure. CircRNAs exhibiting fold changes ≥1.5 and FDR P-values <0.05 were considered as significantly differentially expressed.

Hierarchical Clustering was used to show the distinguishable circRNA expression pattern among samples. Scatter plots were performed to evaluate differentially expressed circRNAs with statistical significance between two groups. Volcano Plot filtering was utilized to visualize the significantly differential circRNAs between each pair wise comparison. Hierarchical Clustering, Scatter plots and Volcano Plot filtering were performed using Agilent Feature Extraction software (version 11.0.1.1; Agilent Technologies, Inc.)

Gene Ontology (GO) enrichment analysis and Kyoto Encyclopedia of Genes and Functional Categories analysis were performed with the DAVID online tool (https://david.ncifcrf.gov/). GO analysis was performed to annotate genes meaningfully in terms of their biological processes (BP) and molecular functions (MF). The -log10 (P-value) yields an enrichment score representing the significance of GO term enrichment among differentially expressed genes. Functional Categories analysis was performed to determine the involvement of target genes in the different biological pathway.

Based on fold change >2, we selected the top 20 including 10 most upregulated circRNAs, and all 10 downregulated circRNAs for subsequent validation. Briefly, total RNA was extracted from 20 pairs of OSCC samples and matched adjacent tissues using TRIzol reagent. The cDNAs were synthesized with the Prime-Script RT Reagent kit (TakaraBiotechnology) from 500 ng of total RNA. The qPCR was performed with the ABI 7300 PCR instrument using the SYBR Green (Takara Bio Inc., Dalian, China) detection method with divergent primers, as listed in Table II. The thermal cycling parameters used for amplification were as follows: denaturation at 94°C for 2 min, followed by 40 cycles at 94°C for 20 sec, 58°C for 20 sec and 72°C for 30 sec. The reaction was stopped by incubation at 75°C for 5 min. GAPDH was used as an internal control. The relative gene expression levels were analyzed by the 2^−ΔΔCq method (18).

To construct the circRNA_102459 overexpression plasmid, the full-length circRNA_102459 cDNA sequence was amplified and cloned into the pcircRNA 1.2 vector (Invitrogen; Thermo Fisher Scientific, Inc.) and sequenced, named as pcRNA-circRNA_102459 (circRNA_102459), while the mock plasmid served as the control vector. The specific small interfering RNAs (siRNAs) targeting circRNA_043621(si-circRNA_043621) and scrambled siRNA control were designed and synthesized by RiboBio Co., Ltd. (Guangzhou, China). All of these plasmids and oligonucleotides were transfected into TSCC1 cells using Lipofectamine 2000 (Invitrogen; Thermo Fisher Scientific, Inc.) according to the manufacturer's instructions. At 48 h after transfection, cells were harvested and used for the in vitro experiments.

Cell growth status was measured using Cell Counting Kit-8 (CCK-8) reagent (Beyotime, China) according to the manufacturer's instructions. Briefly, 1×10⁵ cells were seeded into the wells of a 96-well plate, each group of TSCC1 cells was collected at 48 h after transfection. Then, 10 µl CCK-8 solution was added into each well and incubation was carried out for 2 h at 37°C in darkness. Finally, the absorbance in each well at 450 nm was measured using a microplate reader (Bio-Tek, USA). The experiment was conducted in three separate wells for each sample, and performed in triplicate.

Cell proliferation was determined by the Cell-Light EdU DNA Cell Proliferation Kit (RiboBio Co., Ltd.) according to the manufacturer's instructions. Briefly, the transfected cells were seeded in triplicate in 96-well plates at a density of 3×10³ cells per well, and then exposed to 50 mM EdU for 2 h at 37°C. After collection, the cells were fixed with 4% formaldehyde at room temperature for 20 min and incubated with 100 µl of 1X Apollo^® reaction (Ribobio, Guangzhou, China) for 30 min. The cells were subsequently incubated with DAPI detecting liquid for 10 sec followed by microscopic observation using a fluorescent microscope (magnification ×200; Olympus Corp., Tokyo, Japan). The percentage of the EdU-positive relative to the DAPI-positive cells was calculated by ImageJ software (ImageJ bundled with 64-bit Java 1.8.0 112; National Institutes of Health, Bethesda, MD, USA). The experiment was conducted in three separate wells for each sample, and performed in triplicate.

For cell cycle analysis, 5×10⁵ transfected cells were harvested by trypsinization, washed two times with PBS, and then fixed in 75% ice-cold ethanol overnight at 4°C. The fixed cells were resuspended in PBS and stained with propidium iodide (PI) (50 µg/ml) containing RNase (0.1 µg/ml). Cell cycle distribution analysis was performed using FACSan flow cytometry (BD Biosciences, San Jose, CA, USA). Data are presented as the percentages of cells in the G0/G1, S and G2/M stages.

Apoptosis analysis was performed using Annexin V-FITC Kit (Abcam, Cambridge, UK). In brief, transfected cells were harvested by trypsinization and resuspended in binding buffer. Then the cells were stained with Annexin V-FITC and PI for 15 min at room temperature. Cell apoptotic rates were assessed by FACSan flow cytometry (BD Biosciences).

Following cell transfection, cells at the density of 3×10⁵ were plated in 6-well plates and further cultured for 12 h. Then cells were washed twice with pre-cooled PBS after the upper culture solution was carefully discarded. Subsequently, cells were fixed with 4% paraformaldehyde for 10 min and washed with distilled water, followed by staining with Hoechst 33258 (Thermo Fisher Scientific, Inc.) for 20 min in the dark. Then, the cells were washed with distilled water, after which the morphology of cells was observed under a fluorescence microscope (magnification ×200; Olympus Corp.).

Total protein was isolated from cells using RIPA Lysis Buffer (Thermo Fisher Scientific, Inc.). After collecting the supernatants, the protein concentration was determined with a BCA protein assay kit (Bio-Rad Laboratories, Inc.). Equal amounts (20 µg) of total protein were separated by 12% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and transferred onto PVDF membranes (Millipore). Then, the membranes were blocked with 5% non-fat dry milk in 1X TBS containing 0.05% Tween-20 at room temperature for 1 h. Next, the membranes were incubated with primary antibodies against MAPK (dilution 1:1,000, ab205926; Abcam), PI3K (dilution 1:500; cat. no. ab70912; Abcam), p-PI3K (dilution 1:800, ab182651; Abcam), AKT (dilution 1:1,000; cat. no. ab8805; Abcam), p-AKT (dilution 1:800; cat. no. ab38449; Abcam), Bcl-2 (dilution 1:1,000; cat. no. ab32124; Abcam), Bax (dilution 1:1000; cat. no. ab32503; Abcam) at 4°C overnight and GAPDH (dilution 1:10,000; Cell Signaling Technology) used as internal control. Afterwards, the membranes were incubated with HRP-conjugated secondary antibodies (cat. nos. sc-516102 or sc-2357; Santa Cruz Biotechnology) for 2 h at room temperature. The protein signal was detected and visualized using an enhanced chemiluminescence reagent (Thermo Fisher Scientific, Inc.). Densitometric quantification was determined using Bio-Rad Quantity One software v.4.6.3 (Bio-Rad Laboratories, Hercules, CA, USA).

The fold-change of each circRNA was computed from the profile difference between the cancer and control groups, and the significance was analyzed with a paired t-test. All the in vitro data are expressed as the mean ± standard deviation (SD) from at least triplicate trials. Statistical analyses were performed by GraphPad Prism 5 (GraphPad Software, Inc., La Jolla, CA, USA). The statistical significance for multi-group comparisons was assessed using one-way ANOVA test with Tukey's test. A P-value <0.05 was considered to indicate statistical significance.

To study the potential involvement of circRNAs in OSCC, a high throughput circRNA microarray assay was performed in the three pairs of human OSCC and matched normal tissue. Fig. 1 is a hierarchical cluster displaying the circRNA expression pattern in the OSCC and adjacent tissues. The scatter plot of circRNA expression profile was used to describe the variation in circRNA expression between OSCC and adjacent tissues (Fig. 2A). The statistical significance of the differentially expressed circRNAs between the OSCC and adjacent tissues was additionally visualized using volcano plot filtering (Fig. 2B). Based on the filter criteria (fold-change ≥1.5 and a P-value <0.05), a total of 213 significantly differentially expressed circRNAs were identified, including 124 upregulated and 89 downregulated circRNAs in the OSCC tissues compared with adjacent tissues. Considering that false positives can be caused by multi-comparisons, we used the FDR method to adjust the P-values. After FDR correction, the top 20 differentially expressed circRNAs, including 10 significantly upregulated and 10 significantly downregulated circRNAs were identified, which are listed in Table III.

To explore the expression of circRNAs on a more functional level, we performed GO enrichment analysis for the genes involved with the circRNAs that were found to be differentially expressed in the circRNA microarray results. The count number >2 and FDR <0.05 were chosen as cut-off criteria. The results showed that the most significant enriched GO terms in the biological processes (BP) included ‘MAPK cascade’, ‘cell signaling’ and ‘apoptotic process’ (Fig. 3A). The most significant enriched GO terms in molecular functions (MF) included ‘oxygen binding’, ‘PI3K activity’ and ‘Heme binding’ (Fig. 3B). Functional categories analysis revealed that pathways such as ‘phosphoprotein’ and ‘alternative splicing’ were related to the differentially expressed circRNAs (Fig. 3C).

To verify that the differentially expressed circRNAs discovered through the microarray were accurately identified, we selected 20 potentially significant circRNAs involved in the MAPK and PI3K signaling pathways, including 10 upregulated and 10 downregulated circRNAs for validation by qPCR in 20 pairs of human OSCC and matched normal tissues. As expect, the expression patterns of the 10 downregulated (Fig. 4A, P<0.001) and 10 upregulated circRNAs (Fig. 4B, P<0.001) were consistent with the microarray data. Notably, hsa_circRNA_102459 and hsa_circRNA_043621 displayed the lowest and highest expression in the 20 OSCC tissues relative to the normal tissues, respectively, which may play a crucial role in the development and progression of OSCC. In addition, circRNA-miRNA network prediction was conducted, and the results revealed that there are numerous miRNAs that are able to interact with hsa_circRNA_102459 and hsa_circRNA_043621 (Fig. S1).

Table I shows the correlation between circRNA_ 043621/circ_102459 and the clinicopathological characteristics of the 20 OSCC patients. Specifically, there were no obviously significant correlation between expression and patient age and sex (P>0.05). However, clinicopathological characteristics, including clinical stage, tumor differentiation and lymph node metastasis presented significant difference in regards to the expression of circRNA_043621 and circRNA_102459 (P<0.05, Table I).

Next, we performed gain-of-function and loss-of-function assays to evaluate the biological functions of circRNA_102459 and circRNA_043621 in OSCC cells, respectively. As shown in Fig. 5A, qPCR analysis demonstrated that following circRNA_102459 transfection the expression of circRNA_102459 was significantly elevated in the TSCC1 cells (P<0.001), while si-circRNA_043621 transfection remarkably reduced the expression of circRNA_043621 (P<0.001) in the TSCC1 cells. Subsequent CCK-8 assay revealed that upregulation of circRNA_102459 and downregulation of circRNA_043621 significantly suppressed the growth of TSCC1 cells (Fig. 5B, P<0.01). In addition, the linear RNA expression levels of circRNA_102459 and circRNA_043621 did not change various to overexpression and knockdown of circRNA_102459 and circRNA_043621 (Fig. S2). Furthermore, 5-ethynyl-2′-deoxyuridine (EdU) staining assay showed that the proliferation potential of TSCC1 cells was impaired upon upregulation of circRNA_102459 or downregulation of circRNA_043621 (Fig. 5C and D, P<0.001).

To determine whether the effects of circRNA_102459 or circRNA_043621 on the proliferation of OSCC cells are mediated by inhibition of cell cycle progression, flow cytometry was used to evaluate the cell cycle distribution in TSCC1 cells. As shown in Fig. 6A, circRNA_102459 overexpression or circRNA_043621 knockdown led to the significant accumulation of cells in the G0/G1-phase (P<0.01) and a significant decrease in cells in the S-phase (P<0.01) and G2/M-phase (P<0.05) compared with the corresponding control. Furthermore, the effects of circRNA_102459 or circRNA_043621 on apoptosis were assessed in TSCC1 cells. The percentage of cells undergoing apoptosis, including early apoptosis and late apoptosis, was significantly increased in the circRNA_102459 overexpression or circRNA_043621 knockdown group compared with the corresponding control (Fig. 6B, P<0.05). Apoptosis could also be observed when cells were stained with Hoechst 33258. Here, circRNA_102459 overexpression or circRNA_043621 knockdown induced morphological changes characteristic of apoptosis such as chromatin condensation and nuclear blebbing (Fig. 6C). Taken together, these data suggest that circRNA_102459 overexpression or circRNA_043621 knockdown induces G0/G1 phase arrest and enhances apoptosis in OSCC cells.

To explore the potential mechanism underlying the role of circRNA_102459 or circRNA_043621 in cell proliferation, the activation of the MAPK and PI3K/Akt pathways was assessed using western blotting. As shown in Fig. 7, circRNA_102459 overexpression or circRNA_043621 knockdown downregulated the expression of p38 MAPK. The PI3K/Akt pathway was suppressed by circRNA_102459 overexpression or circRNA_043621 knockdown, as revealed by decreased PI3K and Akt phosphorylation. In addition, we found that circRNA_102459 overexpression or circRNA_043621 knockdown obviously reduced anti-apoptotic Bcl-2 expression, but elevated pro-apoptotic Bax expression in the TSCC1 cells.