This study complies with the Declaration of Helsinki and was approved by the Medical Ethics Committee of the Second Xiangya Hospital of Central South University. All clinical tissue samples used for immunohistochemistry (IHC) and real-time PCR were obtained from OSCC patients who underwent surgical treatment at the Stomatological Center of the Second Xiangya Hospital of Central South University from 2013 to 2015. All patients did not receive radiotherapy and chemotherapy before and after surgery. The pathological characteristics of OSCC patients are shown in Table 1. The OSCC cell lines SCC4 and CAL27 were obtained from the Center for Molecular Medicine, Xiangya Hospital, Central South University (Changsha, China).

SCC4 and CAL27 cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM)/high glucose containing 10% fetal bovine serum (FBS) and 1% penicillin–streptomycin. The cells were cultivated in a humidified 5% CO₂ incubator.

TCGA (https://tcga-data.nci.nih.gov/tcga/) is a public database, and the use of its datasets does not require ethical approval. The GSE dataset (GSE138206) was downloaded from the Gene Expression Omnibus (GEO) (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE138206). In this study, the mRNA expression profile data and clinical follow-up data for bioinformatics analysis were downloaded from TCGA database HNSCC dataset (only the data of 330 OSCC samples and 32 matched normal oral mucosa samples were retained) and GEO database (GSE138206). Among the 330 OSCC patients, 260 patients had complete clinical follow-up data, which could be used for further survival analysis and Cox regression analysis. Based on N grade and M grade, we further divided these 260 OSCC patients into the metastasis-positive group (n ≥ 1 or M = 1) and the metastasis-negative group (N = 0, M = 0).

In this work, only genes with a false discovery rate (FDR) < 0.05 and |log2 FC| ≥1 were selected as differentially expressed genes (DEGs) (Szklarczyk et al., 2011). Gene set enrichment analysis (GSEA) was performed with Kyoto Encyclopedia of Genes and Genomes (KEGG) gene sets (c2) or oncogenic signature gene set (c6) collections of the Molecular Signature Database v7.0 (Mootha et al., 2003; Subramanian et al., 2005). To further explore the mechanisms underlying BTC-mediated EMT, protein–protein interaction (PPI) network analysis was performed using the Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) (version 11.0, https://www.string-db.org/cgi/input?sessionId=bX0xOZT5Yfvc&input_page_show_search=on) (Patil and Nakamura, 2005).

Formalin-fixed paraffin-embedded (FFPE) tissues were embedded in paraffin. We used BTC as the primary antibody (Proteintech, Wuhan, China). Methods experiment was conducted according to histological and immunohistochemical analysis in our last study (Yao et al., 2020). The score for each slide was calculated on the basis of staining intensity and the percentage of positive cells. Immunostaining intensity was divided into four grades: 1, negative; 2, weak; 3, moderate; and 4, strong. The percentage of positively stained cells was also divided into four grades: 1, <5%; 2, 5–35%; 3, 35–75%; and 4, >75%.

For RNA extraction, cells were plated in a six-well plate in advance. When the cells were allowed to grow to 80–100% confluence, the medium was removed, and the cells were washed with PBS. Then, 1 ml TRIzol (Invitrogen, Carlsbad, CA) was added to each well for 5 min at room temperature. The complementary DNA synthesis and quantitative real-time polymerase chain reaction were performed as described previously (Yao et al., 2020). The reverse transcription was performed according to the manufacturer’s protocol for the PrimeScript RT reagent kit (Takara, Japan). Also, GAPDH was used to normalize the expression. The primers used for real-time PCR were as follows: BTC: F: 5′-CCT​GGG​TCT​AGT​GAT​CCT​TCA-3′, R: 5′-CTT​TCC​GCT​TTG​ATT​GTG​TGG-3’; GAPDH: F: 5′-GTC​TCC​TCT​GAC​TTC​AAC​AGC​G-3′, R: 5′-ACC​ACC​CTG​TTG​CTG​TAG​CCA​A-3′.

Semiquantitative real-time PCR was performed using the SYBR Premix Ex Taq II kit (Bio-Rad, California, United States) on a Bio-Rad PCR system.

293T cells in good condition and the logarithmic growth phase were used for lentivirus packaging. The cell density was controlled at approximately 80% during transfection. High-quality plasmids (PMD2.g, PAX2, Flag-plvx-sbp, and Flag-plvx-sbp-btc) were used for transfection. Puromycin was added for selection, 3 and 4 days after transfection.

A 15% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) gel was prepared the night before the experiment. The cells were plated in a six-well plate in advance, and when they reached 80% confluence, they were used for protein extraction. The original medium was removed, and the cells were washed twice with PBS. Then, 1X SDS was added, and the lysate was collected with a cell scraper and transferred to a 1.5-ml EP tube. The tube was placed at 100°C for 5 min, and the protein concentration was assessed. The same amount of total protein (20 μg per well) was separated by 15% SDS-PAGE at 80 V for 30 min and then 120 V for 60 min. Then, the proteins were transferred onto polyvinylidene fluoride membranes (Merck Millipore) at 300 mA for 60 min. Skim milk was applied to seal the blots, and the following primary antibodies were applied at 4°C overnight: AKT, pAKT (Ser 473), pAKT (Thr 308) E-cadherin, N-cadherin, vimentin (1:1000, Cell Signaling Technology, Shanghai, China), and BTC (1:1000, Proteintech, Wuhan, China). After washing three times with PBST, the secondary antibody was applied for 1 h at room temperature, and after washing three times with PBST again, the bands were visualized utilizing enhanced chemiluminescence (ECL) reagents (Pierce, United States).

The cells were seeded in a six-well plate at 25–30% confluence and were cultured at 37°C and 5% CO₂. After 3 days, a confluent cell monolayer (95–100%) was ensured, and the medium was removed from the culture dishes and washed with PBS two times. A sterile P-200 pipette tip was used to scratch the cell monolayer in a straight line in the center of the dishes. Then, the dishes were cleaned with PBS, followed by adding 2ML DMEM. Defining exact positions and focal planes, image acquisition was taken throughout 0–36 h. Quantitative data analysis was performed with Open-source software (ImageJ/Fiji).

The Matrigel was prepared at 4°C. Transwell inserts (BD Bioscience, San Jose, CA) with a filter of 8 μm were prepared for transwell cell migration/invasion assay. Matrigel of 50 μL was coated or uncoated downside surface of the transwell membrane and incubated at 37°C for 30 min for gelling. A total of 5 × 104 cells were seeded into the upper chamber with or without Matrigel, and 500 μL complete medium was added to the lower chamber. After 24-h or 48-h incubation, the cells were fixed with methanol and stained with crystal violet. Then, cells on the top surface of the membrane were wiped off. Four random fields were photographed in a phase-contrast microscope using 10 × objective. Quantitative data analysis was performed with Open-source software (ImageJ/Fiji).

CCK-8 Kit was applied to evaluate cell proliferation ability (KeyGEN BioTECH, Jiangsu, China). A total of 3 × 103 cells were seeded in 96-well plates with 100 μL medium for each well. After 24, 48, 72, and 96 h, the medium in each well was removed, and then 10 μg CCK-8 solution and 100 μL medium were added into the well for 1 h in a dark environment. After that, absorbance was measured at 450 nm using a PerkinElmer’s EnSpire Multilabel Plate Reader.

All analyses were performed with SPSS 25.0 (SPSS Inc., United States), R (Version 4.0.3), and GraphPad Prism (version 8.0, GraphPad Software, Inc., United States). Statistical significance was determined by Student’s t-test and analysis of variance comparisons. Overall survival was assessed using the Kaplan–Meier method, and the differences in survival between the groups were compared using log-rank tests. The effect of clinicopathological factors on survival was determined with univariate and multivariate Cox proportional hazards models. Data from three independent experiments are presented as the mean ± SD. Differences with a p-value of <0.05 were considered statistically significant. *, **, and *** indicate p < 0.05, p < 0.01, and p < 0.001, respectively.

Through the analysis of the mRNA expression profile data in TCGA database, we found that a total of 4,626 differentially expressed genes (DEGs) were identified between 330 OSCC samples and 32 matched normal tissue specimens. Among these DEGs, 3,027 genes were upregulated in OSCC tumor tissues, whereas 1,599 genes were suppressed (Figure 1A). To further screen out the metastasis-related genes involved in OSCC, 260 patients with complete clinical follow-up data from TCGA database were divided into a metastasis-positive group (n = 146) and a metastasis-negative group (n = 113). After differential mRNA expression analysis, 43 DEGs were screened out, among which 25 genes showed low expression and 18 genes were highly expressed in the metastasis-positive group (Figures 1B,C). Furthermore, multivariate regression analysis revealed that BTC, IFNK, CCBE1, NKX2.2, VGLL2, and AC022075.2 were independent prognostic factors of OSCC among 43 DEGs (Figure 1D). Furthermore, BTC, the only tumor-suppressor gene with the largest fold change among IFNK, CCBE1, NKX2.2, VGLL2, and AC022075.2, was selected for further study, and the low expression of BTC was significantly correlated with the poor prognosis (in terms of OS) of OSCC (Figure 1E). These results indicate that BTC might be a promising biomarker for OSCC development.

To further validate the role of BTC in the malignant progression of OSCC, we used IHC assays to examine the expression of BTC in 38 tissues of OSCC patients and their paired normal samples. The results showed that the expression level of BTC was remarkably lower in OSCC tissues than normal samples. In particular, the BTC expression level was significantly decreased in the lymph nodes with metastatic cancer of OSCC tissues when compared to tumor specimens (Figures 2A,B). Moreover, mRNA expression of BTC was in accordance with the results of IHC after examining 5 OSCC and paired normal samples (Figure 2C). The low expression of BTC was significantly correlated with the poor prognosis (in terms of their paired normal samples) of OSCC (Figure 2D). Furthermore, we analyzed the associations between the clinicopathological variables and BTC expression in 38 OSCC patients, and we found that the BTC expression was related to histological grade, N classification, tumor stage, and metastasis (Table 1). Through Cox proportional hazards models, we found that the low expression of BTC was an independent risk factor for prognosis in OSCC (Table 2). To further verify these results, we analyzed the expression profile data of 330 OSCC samples and 32 normal oral mucosa samples from TCGA database and found that compared to that in normal tissues, the BTC expression level in tumor tissues was low (Figures 2E,F). Furthermore, we analyzed the correlation between BTC expression and the clinical prognosis-related factors of the patients. BTC expression showed a significant correlation with patient age, tumor stage, histological grade, and lymph metastasis status (Figures 2G–L), especially in patients with lymph metastatic OSCC whose BTC levels were significantly reduced. Altogether, these results indicate that BTC may play a vital role in the malignant progression of OSCC, especially in metastasis.

In order to explore the biological function of BTC in OSCC, we stably overexpressed BTC in SCC4 and CAL27 OSCC cells by lentivirus (Figure 3A). The CCK-8 assay indicated that overexpression of BTC significantly decreased cell proliferation as compared to control groups (**p < 0.01) (Figures 3B,C). Moreover, the effects of BTC on regulation of OSCC cell migration were determined by a wound healing assay. As shown in Figure 3D, the ability of cell migration was obviously inhibited after overexpression of BTC in SCC4 and CAL27 cells. In addition, the transwell assay exhibited a consistent tendency in which SCC4 and CAL27 cell migration and invasion were reduced in the overexpressed BTC group (Figures 3E,F). Collectively, these results suggest that upregulation of BTC alleviated cell growth, migration, and invasion in OSCC.

Emerging studies have shown that EMT is an essential process that contributes to tumor invasion and metastasis in OSCC (Pastushenko and Blanpain, 2019; Chen et al., 2020). Therefore, we investigated the role of BTC in EMT of OSCC. We divided OSCC patients in TCGA database into two groups (BTC low-expression group and BTC high-expression group). As shown in Figure 4A, the BTC high-expression group had significantly decreased EMT scores when compared to BTC low-expression group. Through GSEA, the function of BTC was mainly enriched to EMT-related processes, such as tight junctions, focal adhesion, adherens junctions (Figure 4B). Moreover, whether in TCGA database or the GEO dataset (GSE 138206), we found that 17 EMT-related markers were differentially expressed between BTC low- and high-expression groups, among which 11 genes (AGER, CDH2, FH1, MMP2, SNAI1, SNAI2, TWIST 1, TWIST 2, VIM, ZEB1, and ZEB2) were related to epithelial functions and 6 genes (EPCAM, MAL2, CLDN4, CDH1, CDH3, and ST14) were associated with mesenchymal ability (Figure 4C). We also found that the expression of BTC was correlated with the expression level of multiple EMT markers, such as E-cadherin (R = 0.4, p < 0.001) (Figure 4D), N-cadherin (R = −0.39, p < 0.001) (Figure 4E), and vimentin (R = −0.43, p < 0.001) (Figure 4F).

To investigate the mechanisms underlying BTC-mediated EMT, we used STRING to construct a PPI network and found that BTC is a hub gene that can interact with AKT in the protein regulatory network of OSCC (Figure 4G). To further explore whether BTC regulated the EMT process via the PI3K-AKT signaling pathway in OSCC cells, Western blot analysis was performed to show that upregulation of BTC inhibited the activation of PI3K and the phosphorylation of AKT in ser473 and thr308. Furthermore, the suppression of PI3K-AKT signaling was accompanied by the altered protein expression of EMT-related markers, among which the expression of E-cadherin was significantly upregulated, whereas the expression of N-cadherin and vimentin was significantly downregulated (Figure 4H).