The studies involving human participants were approved by the Ethics Committee of the Stomatology Hospital of Guangzhou Medical University (Approval No. KY2019026). Thirty primary OSCC specimens and adjacent tissue specimens were obtained from the Department of Oral and Maxillofacial Surgery, Stomatology Hospital of Guangzhou Medical University. All OSCC specimens were pathologically diagnosed as OSCC, and the adjacent tissue specimens were obtained ≥2 cm from the resection margin, which were pathologically diagnosed to have no precancerous or reactive changes histologically. All specimens were frozen immediately in liquid nitrogen after resection.

Immunohistochemical staining (IHC) was performed on 5 human OSCC tumor tissues and paired adjacent specimens. Paraffin-embedded tissues were cut into 4-µm sections. The samples were dewaxed, dehydrated, and then treated with 3% H₂O₂ solution to inhibit the endogenous peroxidase activity. Antigen retrieval was carried out by microwave heating in sodium citrate buffer (pH 7.8). Slides were then blocked with 5% BSA solution. Tissue sections were further incubated with primary antibody against CHRDL1 (1:150 dilution, Zen-Bio) at 4°C for 12 h and then washed and incubated with secondary antibody (1:100 dilution, ZSJB-Bio) at 37°C for 1 h. Finally, slides were incubated with diaminobenzidine (DAB, ZSJB-Bio) and counterstained with hematoxylin.

IHC slices were scanned by a digital tissue slice scanner (3D HISTECH, Magyarország). The Servicebio image analysis system was used to score the slides for positivity. Positivity grading was first performed (I) as follows: negative without staining, 0 points; weak positive/light yellow, 1 point; medium positive/brownish yellow, 2 points; strong positive/brown, 3 points; then, the cumulative weak, medium, and strong positive area, tissue area, and integrated optical density (IOD) value of positive in the measurement area were calculated. The histochemistry score (H-score) is the output of a histological scoring method for immunohistochemistry. The number of positive cells in each section and their staining intensity were converted into corresponding values for semiquantification of the tissue staining, as follows: H-Score = ∑(pi×i) = (percentage of weak intensity area ×1) + (percentage of moderate intensity area ×2) + (percentage of strong intensity area ×3), where pi represents the percentage of the pixel area of positive signal, and i represents positive grade (38, 39).

The human tongue squamous cell carcinoma cell line CAL27 (ATCC, ATCC^® CRL-2095™) was purchased from ATCC, while the normal oral epithelial keratinocyte line HOK was purchased from AULU (Guangdong, China). CAL27 and HOK cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) (Gibco, Waltham, MA, USA). The complete cell culture medium contained 10% fetal bovine serum (FBS, Gibco) and 1% penicillin/streptomycin (Gibco). The serum-free cell culture medium for carcinoma cells was prepared as DMEM containing 1% penicillin/streptomycin for a number of subsequent experiments (e.g., wound-healing assay, cell migration, and invasion assays). The cells were cultured in a humidified incubator at 37°C and 5% CO₂ incubator.

Commercial lentivirus (lenti-CHRDL1 and shRNA-CHRDL1; OBIO, Shanghai, China) was utilized to overexpress or silence CHRDL1. An empty carrier lentivirus (lenti-NC/shRNA-NC) was used as a negative control. CAL27 cells were infected with the lentivirus (MOI = 60) and screened with puromycin (2 µg/ml) for 15 days. Quantitative reverse transcription-polymerase chain reaction (RT-qPCR) and Western blotting (WB) were used to validate the overexpression/knockdown level of CHRDL1.

Total RNA from the clinical specimens was extracted using TRIzol Reagent (Invitrogen, USA) according to the manufacturer’s instructions. Total RNA from the cells was extracted using an RNA extraction kit (19221ES50, Yeasen, China). The quantity and quality of RNAs were detected by A260/A280 with a spectrophotometer (NanoDrop 2000, Thermo Fisher Scientific, Waltham, MA, USA). Two micrograms of total RNA was used to synthesize cDNA (11141ES60, Yeasen, China). RT-qPCR was performed using a SYBR Green qPCR kit (11199ES08, Yeasen, China). Relative mRNA expression was normalized to that of the internal GAPDH control. The primer sequences are listed in Supplementary Table S1 . The relative expression of targeted genes was calculated by the 2^−ΔΔCt method. Each test was repeated at least three times.

Total proteins were isolated from cell samples through precooled cell lysis buffer (Cell Signaling Technology, Danvers, MA, USA) with protease inhibitor and phosphatase inhibitors (RayBiotech, Guangzhou, China) and quantified by a BCA protein assay kit (SE248351, Thermo Fisher, USA). Equal protein extracts (30 µg) were separated by SDS–PAGE (PG112, Epizyme, China) and transferred to polyvinylidene fluoride membranes (Millipore, Burlington, MA, USA). After incubation with the primary antibody and the secondary antibody, the target protein was visualized by chemiluminescence using an ECL kit (P001AS, Beyotime, China). The antibodies used in the Western blot assay are listed as follows: CHRDL1 (1:500, Thermo Fisher, USA), Fibronectin (1:1,000, Proteintech, China), E-cadherin (1:1,000, Proteintech, China), E-cadherin (1:1,000, Proteintech, China), Cytokeratin 18 (1:1,000, Abcam, UK), β-actin (1:1,0000, Cell Signaling Technology, USA), JNK (1:1,000, Cell Signaling Technology, USA), p-JNK (1:1,000, Cell Signaling Technology, USA), ERK1/2 (1:1,000, Cell Signaling Technology, USA), p-ERK1/2 (1:1,000, Cell Signaling Technology, USA), p38 (1:1,000, Cell Signaling Technology, USA), p-p38 (1:1,000, Cell Signaling*nbsp;Technology, USA), and goat anti-rabbit IgG H&L (HRP) antibody (1:10,000, Abcam, UK). β-actin served as the internal control to calculate the relative expression of the targeted proteins.

For the wound-healing assay, transfected CAL27 cells were first incubated and cultured with complete medium. A scratch was made after a confluent monolayer of cells was formed. Afterward, the cells were washed with phosphate-buffered saline (PBS) and cultured with serum-free medium. Images were taken at 0 and 24 h after wound generation. The wound-healing areas were assessed by ImageJ to calculate the wound-healing rate. Wound-healing rate% = [Area at 0 – Area at 1]/Area at 0 × 100% (Area at 0 is the area of the wound measured immediately after scratching, and Area at 1 is the area of the wound measured at 1 h after scratching). The wound area was measured at 0 and 24 h. Each test was repeated at least three times.

CAL27 cells were seeded at a density of 1 × 10⁵ in serum-free DMEM in the upper wells of Transwell chambers (8 μm pore size, Corning, New York, NY, USA), while the lower wells were filled with complete medium containing 20% FBS. After 48 h, cells in the upper layer were removed with a swab. The cells on the bottom membrane were fixed with 4% paraformaldehyde and stained with crystal violet. Five random visual fields were photographed for each group, and the experiments were repeated three times.

CAL27 cells were seeded at a density of 2 × 10⁵ in serum-free DMEM in the upper wells of chambers that were coated with 60 µl of Matrigel (200 μg/µl, Corning) and incubated for 2 h at 37°C, while the lower wells were filled with complete medium containing 20% FBS. Cells in the upper layer were removed with a swab after 36 h of culture, and cells on the bottom membrane were fixed with 4% paraformaldehyde and stained with crystal violet. Each group was photographed in five randomized visual fields, and the experiments were repeated three times.

A Cell Adhesion Detection kit (BB-48120, BestBio Science, China) was used according to the manufacturer’s instructions to analyze the adhesion ability of CAL27 cells. Briefly, the coating buffer (100 µl) was transferred to each well in a 96-well plate and incubated overnight at 4°C. Cells (5×10⁴) were inoculated into each well of the coated 96-well plate and allowed to adhere for 1.5 h at 37°C. After the cells were washed three times with PBS, cell imaging was performed. The optical density value at 450 nm was measured to calculate the cell adhesion rate. Each test was repeated at least three times.

CAL27 cells were seeded at a density of 1,000 in a confocal dish. After 24 h, cell monolayers were fixed with 4% paraformaldehyde and stained with rhodamine-conjugated phalloidin (40737ES75, Yeasen, China) according to the manufacturer’s instructions. Confocal images were captured using a confocal laser scanning microscope (LEIKATCSSP8, Leica, Germany).

Total RNA was extracted from transfected CAL27 cells (OV-CHRDL1/OV-NC and sh1-CHRDL1/sh-NC) using TRIzol reagent (Invitrogen Life Technologies, USA) following the manufacturer’s protocol. RNA concentration and purity were checked with a Nanodrop 2000 instrument (Thermo Fisher Scientific, USA). After RNA quality control was performed, the libraries for next-generation sequencing were prepared using the TruSeqTM RNA Sample Prep Kit (Illumina, USA) according to the manufacturer’s instructions. Sequencing was performed by Shanghai Origingene Biopharm Technology Co., Ltd. (Shanghai, China). After quality control of the original data, the high-quality sequencing data were compared with the designated reference genome. The expression values were calculated by the StringTie tool, and the tDESeq algorithm was applied to filter the differentially expressed genes. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses were performed to reveal the involved pathways.

The animal experiments were conducted in accordance with the guidelines approved by the Institutional Animal Care and Use Committee of the First Affiliated Hospital of Guangzhou Medical University of China (Approval number was 2017-086).

Four-week-old male BALB/c-nude mice were purchased from GemPharmatech (Jiangsu, China). All animal experiments were carried out following the guidelines set by the Institutional Animal Care and Use Committee of the First Affiliated Hospital of Guangzhou Medical University of China. Nude mice were randomly divided into four groups. CAL27 cells (1×10⁶) were suspended in 0.1 ml PBS and injected via the tail vein into animals that had been anesthetized with 3% isoflurane. After 9 weeks, the nude mice were anesthetized and euthanized by cervical dislocation. Lungs were collected, weighed, rinsed, fixed, and embedded in paraffin blocks and sectioned for H&E staining to confirm metastasis. Three slices were acquired from each lung sample. The number of pulmonary tumor nodules were counted under the dissecting microscope. The averaged lung colonies from 3 slices were calculated for each lung sample.

Cells (4,000) were cultured in 96-well plates, and cell proliferation was detected using a Cell Counting Kit-8 (Dojindo, Kumamoto, Japan) at 1, 3, and 5 days of culture according to the manufacturer’s instructions. EdU staining was performed to assess cell proliferation ability by the Cell-Light EdU imaging kit (C0071S, Beyotime, China) following the manufacturer’s instructions. A colony formation assay was performed. One thousand transfected CAL27 cells were seeded in six-well plates and cultured for 8 days in complete medium. The colonies were further fixed with 4% paraformaldehyde and stained with crystal violet. Colonies that contained more than 50 cells were counted. Each test was repeated at least three times.

The data were analyzed using statistical software (SPSS Statistics 16.0). The data are presented as the means and standard deviations. Student’s t-test was used for comparisons of normally distributed data. The Mann–Whitney U test was used for nonnormally distributed data. A p-value <0.05 was considered statistically significant.

The data from the Gene Expression Profiling Interactive Analysis database indicated that the expression of CHRDL1 was downregulated in HNSC tissues compared to normal tissues ( Figure 1A ). We further collected samples from 30 OSCC patients, and qRT-PCR indicated that the expression of CHRDL1 was significantly decreased in OSCC tissues compared to adjacent tissues ( Figure 1B ). A lower expression level of CHRDL1 was also detected in the common OSCC cell line CAL27 than in the normal cell line HOK ( Figure 1C ). IHC analysis of tissue samples showed that CHRDL1 was downregulated in OSCC tissues. Representative IHC images revealed that CHRDL1 protein was primarily distributed in the epithelial structure of OSCC tissues and adjacent tissues. However, there was little staining of CHRDL1 in either the tumor stroma in OSCC or subepithelial tissues in the adjacent normal tissues ( Figures 1D, E ). These data demonstrated that CHRDL1 was typically negatively correlated with OSCC development.

The role of CHRDL1 in the progression of OSCC was explored in the CAL27 cell line. CHRDL1 was overexpressed in the CAL27 cell line by lentivirus infection ( Figures 2A–C ). Overexpression of CHRDL1 markedly suppressed the wound-healing and migration abilities of CAL27 cells ( Figures 2D–G ). Furthermore, CHRDL1 overexpression significantly decreased the number of invading CAL27 cells ( Figures 2F, G ). The cell adhesion assay showed a lower adhesion rate in the CHRDL1 overexpression (OV-CHRDL1) group than in the control group (NC-CHRDL1) ( Figures 2H, I ). However, the EdU assay, colony formation assay, and CCK-8 assay showed no significant difference between the OV-CHRDL1 group and the NC-CHRDL1 group, indicating that overexpression of CHRDL1 did not affect the proliferation of CAL27 cells ( Supplementary Figure 1 ). These results suggested that upregulation of CHRDL1 inhibited metastasis rather than proliferation in the CAL27 cell line.

To determine whether silencing CHRDL1 promoted malignant biological behaviors in the CAL27 cell line, two stable CHRDL1-knockdown CAL27 cell lines (sh1-CHRDL1 and sh2-CHRDL1) and their control group (sh-NC) were established. Although CHRDL1 mRNA levels were significantly decreased in both stable CHRDL1-knockdown cell lines ( Figure 3A ), the protein level was reduced only in the sh1-CHRDL1 cell line and was obviously increased in the sh2-CHRDL1 cell line ( Figure 3B, C ). Thus, the sh1-CHRDL1 CAL27 cell line was used for the following experiments. We demonstrated that CHRDL1 silencing markedly promoted the wound-healing and migration abilities of CAL27 cells ( Figures 3D–G ). The sh1-CHRDL1 CAL27 cells also exhibited enhanced invasion ( Figures 3D–G ) and adhesion ability ( Figures 3H, I ) compared to sh-NC CAL27 cells. Intriguingly, proliferation assays also showed no significant difference between the sh1-CHRDL1 group and the sh-NC group ( Supplementary Figure 1 ). These findings confirmed that CHRDL1 silencing promoted the progression of CAL27 cells.

EMT has been widely reported to increase OSCC metastasis (40, 41). Compared to sh-NC cells, sh1-CHRDL1 cells became fusiform and fibroblast-like in shape and lost their polygonal epithelial morphology ( Figure 4A a). The sh1-CHRDL1 CAL27 cells also became more dispersed and showed loose cell-to-cell contact compared to sh-NC CAL27 cells under high cell density conditions ( Figure 4A b). Rhodamine phalloidin staining was further performed to detect F-actin. Actin-rich membrane protrusions called filopodia (white arrow) were clearly seen at the periphery of sh1-CHRDL1 CAL27 cells but not sh-NC CAL27 cells ( Figure 4A c). To further examine the EMT-like phenotype of sh1-CHRDL1 CAL27 cells, the expression of epithelial and mesenchymal markers, including cytokeratin 18, E-cadherin, N-cadherin, and fibronectin, was examined by WB ( Figures 4B, C ). The protein expression levels of E-cadherin and cytokeratin 18, epithelial differentiation markers, were significantly upregulated in OV-CHRDL1 CAL27 cells compared to OV-NC CAL27 cells. The protein expression levels of the mesenchymal differentiation markers N-cadherin and fibronectin were significantly downregulated in OV-CHRDL1 CAL27 cells compared to OV-NC CAL27 cells. In contrast, the protein expression levels of E-cadherin and cytokeratin 18 were 1.21- and 2.17-fold downregulated, respectively, in sh1-CHRDL1 CAL27 cells compared to sh-NC CAL27 cells, while the protein expression levels of N-cadherin and fibronectin were 0.64- and 0.53-fold upregulated, respectively, in sh1-CHRDL1 CAL27 cells compared to sh-NC CAL27 cells. These results suggested that CHRDL1 silencing in CAL27 cells led to the EMT phenotype, while overexpression of CHRDL1 in CAL27 cells inhibited the process. Thus, CHRDL1 knockdown may induce metastasis in CAL27 cells by activating EMT.

To identify genes differentially expressed specifically in the context of high versus low CHRDL1 levels, RNA sequencing was performed in OV-NC/OV-CHRDL1 and sh-NC/sh1-CHRDL1 CAL27 cells. A fold-change >2.0 or <−2.0 and p _adj < 0.05 were chosen as cutoff criteria and visualized using the heatmap. A total of 163 upregulated and 179 downregulated genes in OV-CHRDL1 CAL27 cells compared to OV-NC cells were identified by gene expression analysis ( Figure 5A , left). In addition, 1067 genes were upregulated and 952 genes were downregulated in sh1-CHRDL1 CAL27 cells compared to sh-NC cells ( Figure 5A , right). Then, Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses were performed on the differentially expressed genes (DEGs) identified in OV-CHRDL1 CAL27 cells and sh1-CHRDL1 CAL27 cells. The top 10 most significant GO biological processes (BP) are shown in Figure 5B . Cell morphogenesis and regulation of cell adhesion were identified in sh1-CHRDL1 CAL27 cells, consistent with the in vitro results. The top 10 enriched KEGG pathways are shown in Figure 5C . Interestingly, MAPK signaling was enriched in both OV-CHRDL1 CAL27 cells and sh1-CHRDL1 CAL27 cells. Thus, MAPK signaling pathway components were further probed by WB ( Figures 5D, E ). The protein expression of phosphorylated JNK (p-JNK) was upregulated, and that of phosphorylated ERK (p-ERK1/2) and p38 (p-p38) was downregulated, in OV-CHRDL1 CAL27 cells compared to OV-NC CAL27 cells. In contrast, the protein expression of p-JNK was downregulated and that of p-ERK and p-p38 was upregulated in sh1-CHRDL1 CAL27 cells compared to sh-NC CAL27 cells. These results indicated that CHRDL1 may influence the MAPK pathway in CAL27 cells to regulate cell metastasis.

To determine the metastatic effect of CHRDL1 in vivo, OV-NC/OV-CHRDL1 and sh-NC/sh1-CHRDL1 CAL27 cells were injected into 4-week-old male BALB/c nude mice via the tail vein, and their lung colonization ability was analyzed. Samples were collected after 9 weeks to determine the number of pulmonary tumor colonies and weight changes. As shown in Figure 6A , the number of lung tumors (red arrow) derived from OV-CHRDL1 CAL27 cells was significantly decreased compared to that derived from OV-NC CAL27 cells. The number of lung tumors (red arrow) derived from sh1-CHRDL1 CAL27 cells was significantly increased compared to that derived from sh-NC CAL27 cells. The lung weights of the mice injected with OV-CHRDL1 CAL27 cells were lower than those of the mice injected with OV-NC CAL27 cells. In contrast, the lungs from the mice injected with sh1-CHRDL1 CAL27 cells weighed more than those of the mice injected with sh-NC CAL27 cells ( Figure 6B ). Furthermore, hematoxylin-eosin staining of lung tissues showed that mice injected with OV-CHRDL1 CAL27 cells had fewer lung colonies, while mice injected with sh1-CHRDL1 CAL27 cells showed more lung colonies, compared to the control group ( Figure 6C ). These results demonstrated that CHRDL1 silencing could promote tumor metastasis and enhance lung colonization in vivo, while CHRDL1 overexpression could restrain tumor metastasis and enhance lung colonization in vivo.