The raw data of TNC expression in NPC tissues and normal nasopharyngeal tissues were downloaded from the GEO database (ncbi.nlm.nih.gov/geo), including two datasets, GSE53819 (23) and GSE13597 (24). In addition, the GSE103611 dataset (25) was downloaded from the GEO database, which included data of gene expression between 24 NPC non-metastatic tissues and 24 NPC metastatic tissues. An NPC tissue microarray (TMA) was purchased from Guilin Fanpu Biotech (http://www.fanpu.com/aboutus/aboutus.htm). There were 150 patients in the TMA, including 115 NPC cases and 35 nasopharyngitis (NPG) cases. The inclusion and exclusion criteria were as follows: i) None of the patients had received antitumor therapy prior to biopsy; ii) the pathological data of patients with NPC were complete. Among the patients with NPC, there were 78 men and 37 women. The median age of patients with NPC was 48.3 years (age range, 24–78 years), and the average age was 49.47±7.32 years. The present study was approved by the Clinical Research Ethics Committee of Renmin Hospital of Wuhan University [Wuhan, China; approval no. 2020(421)] and written informed consent was provided by all patients prior to the study start.

TNC expression was detected in patients in the TMA via IHC analysis. The TMA was dewaxed and hydrated after heating in a thermostat at 60°C for 20 min. For antigen retrieval, the TMA was placed in sodium citrate buffer (10 M, pH 6.0) and heated at 95°C for 10–15 min. Normal goat serum blocking solution (cat. no. C0265; Beyotime Institute of Biotechnology) was added dropwise to the TMA at room temperature for 20 min to block the endogenous enzymes and antibodies in the tissue. The TMA was incubated with anti-TNC at 37°C for 30 min (1:100; cat. no. YT6187; ImmunoWay Biotechnology Company) to detect TNC expression. A non-immune IgG antibody (1:500; cat. no. C0265; Beyotime Institute of Biotechnology) was used as the negative control. Tissues stained with the fluorescently labeled antibodies were observed under a fluorescent microscope (magnifications, ×100 and ×400).

The percentage of TNC positive cells was calculated and scored as follows: 1, 0–9; 2, 10–50 and 3, >50%. The immunoreactive scores of TNC were scored as follows: 0, negative; 1, weak; 2, moderate and 3, strong. The scores were multiplied and graded as follows: -, 0; +, 1–3; ++, 4–6 and +++, 7–9. Tissues graded as ‘−’ and ‘+’ were considered as the low expression TNC group, while ‘++’ and ‘+++’ were considered as the high TNC expression group (26).

The WGCNA R package 1.69 (Bioconductor) was used to construct the WGCNA on the GSE53819 dataset, which consisted of 18 patients with NPC and 18 healthy individuals (21). Genes with an expression >0.1 were screened in GSE53819 and used for the construction. Samples were clustered using hierarchical clustering by the average method to detect outliers before an appropriate soft threshold power was selected to finish standard scale-free networks. Subsequently, gene dendrogram and module identification were performed through the construction of adjacency and topological overlap matrix (TOM). Highly similar modules were merged by implementing the clustering of modular eigengenes, and the association between each module eigengene and TNC was calculated. The internal connectivity and module membership of each gene in the module that was most highly associated with TNC was used to screen for hub genes meeting the following criteria: Gene and specified module significance >0.2, modular membership value >0.8 and q<0.01.

Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses (https://www.genome.jp/kegg) of hub genes were performed using the Molecular Signatures Database (gsea-msigdb.org/gsea/msigdb/annotate.jsp) (27). In addition, gene set enrichment analysis (GSEA) for the differentially expressed genes between the high and low TNC expression groups was performed using the same database.

The human NPC cell lines, 5–8F and 6-10B, were purchased from the Laboratory of Molecular Tumor Pathology, Southern Medical University (Guanagzhou). Cells were maintained in DMEM (Thermo Fisher Scientific, Inc.) supplemented with 10% fetal bovine serum (FBS, Gibco; Thermo Fisher Scientific, Inc.), at 37°C with 5% CO₂.

Lentivirus transduction particles (Shanghai GeneChem Co., Ltd.) containing small interfering RNA (siRNA) targeting TNC and negative control siRNA were transduced into 5-8F and 6-10B cells. A total of 5×10⁴ cells/ml cell suspension with complete medium (90% RPMI1640 media + 10% Gibco Fetal Bovine Serum) was prepared and 2 ml cell suspension/well was inoculated into 6-well plates. Cells were incubated at 37°C for 16–24 h until they reached 20–30% confluence. During transduction, 1 ml of medium was discarded, and 40 µl of HiTransG infection solution (Shanghai GeneChem Co., Ltd.) and 40 µl of lentivirus solution (The MOI was 20 according to the pre-infection experiment and the virus titer was 1×10⁸ TU/ml) were added, and cells were cultured at 37°C. After 16 h, the medium was replaced with fresh medium. Transduction efficiency was observed via light and fluorescence microscopy (magnification, ×100 times) after 72 h. The successfully transduced cells had red fluorescence as there were GFP labels in the lentivirus particles. The stable cells were screened using puromycin (2 µg/ml) (Zhejiang Genom Co., Ltd.) for 14 days.

Cell proliferation was assessed via the Cell Counting Kit-8 (CCK-8) assay (Dojindo Molecular Technologies, Inc.) and the EdU cell proliferation assay (cat. no. C10310-1; Guangzhou RiboBio Co., Ltd.).

For the CCK-8 assay, 5-8F and 6-10B cells were seeded into 96-well plates at a density of 4×10³ cells/well. Cell proliferation was assessed after 0, 24, 48 and 72 h. A total of 100 µl reagent, containing 10 µl CCK-8 solution and 90 µl serum-free media, was added to each well and the plates were incubated for 1 h. The absorbance values were measured at a wavelength of 450 nm, using a microplate reader (PerkinElmer, Inc.), and the proliferation curve was plotted.

For the EdU cell proliferation assay, 5-8F and 6-10B cells were seeded into 6-well plates at a density of 1×10⁵ cells/well and cultured until they reached 80-90% confluence. Cells were subsequently labeled using Cell-Light™ EdU Apollo^® 567 (Guangzhou RiboBio Co., Ltd.), according to the manufacturer's protocol, and observed under an automatic fluorescence microscope (magnification, ×100).

For the wound healing assay, 5–8F and 6-10 B cells were seeded into 6-well plates. When the cells confluence reached 80–90%, the cell monolayers were scratched using a 100 ul sterile pipette tip, washed twice with PBS and incubated with FBS-free medium for 24 h at 37°C. Cells were observed in five randomly selected fields using an optical microscope (magnification, ×100). The distance between the two cell edges was measured and compared with the original distance of the gap using ImageJ 1.8.0 software (National Institutes of Health), and this was considered the wound healing rate.

50,000 5-8F or 6–10 B cells in 200 µl FBS-free media were plated in the upper chambers of Transwell plates (Corning, Inc.), which were pre-coated with Matrigel for the invasion assay for 2 h at 37°C (10 mg/ml), but not the migration assay. A total of 600 µl supplemented media was added to the lower chambers. Following incubation for 24 h at 37°C, the migratory and invasive cells were stained with 0.1% purple crystal for 60 min at room temperature and counted in five randomly selected fields using a light microscope (magnification, ×200 times).

Total RNA was extracted using TRIzol^® reagent (Invitrogen; Thermo Fisher Scientific, Inc.) from 5-8F or 6–10 B cells and reverse transcribed (RT) into cDNA using the Super RT Reverse Transcriptase reagent kit (CoWin Biosciences). In RT process, reaction system including RNA was incubated at 42°C for 60 min to synthesize cDNA, then incubated at 80°C for 10 min to inactivate the reverse transcriptase and terminate the RT reaction. qPCR was subsequently performed using SYBR^® Green MasterMix (Thermo Fisher Scientific, Inc.) on a 7500 Fast Real-Time PCR system (Applied Biosystems; Thermo Fisher Scientific, Inc.), according to the manufacturer's instructions. The thermocycling conditions for qPCR were as follows: Thermal denaturation for 3–5 min at 95°C; amplification program for 30 sec at 95°C, 60°C for 30 sec, cycle number was 40; final extension at 72°C for 5–10 min; after all PCR reactions are completed, the temperature was maintained at 4–12°C. The following primer sequences were used for qPCR: TNC forward, 5′-ACTGACTCAGAAGCCTTGG-3′ and reverse, 3′-ATTTCTGGCACTTTCTCGC-5′; cadherin 1 (CDH1) forward, 5′-AGCACCTTCCATGACAGACCC-3′ and reverse, 5′-AGAACGCATTGCCACATACAC-3′; CDH2 forward, 5′-CATCATCATCCTGCTTATCCTTGT-3′ and reverse, 5′-GGTCTTCTTCTCCTCCACCTTCTT-3′; vimentin (VIM) forward, 5′-ATCGTGATGCTGAGAAGTTTCG-3′ and reverse, 5′-TCTGGATTCACTCCCTCTGGTT-3′; snail family transcriptional repressor 2 forward, 5′-TTTTGCACTGGTATTTCTTTACATC-3′ and reverse, 5′-CCCTGGTTGCTTCAAGGACAC-3′; and GAPDH forward, 5′-TGACTTCAACAGCGACACCCA-3′ and reverse, 5′-CACCCTGTTGCTGTAGCCAAA-3′. Relative expression levels were calculated using the 2^−ΔΔCq method (28) and normalized to the internal reference gene GAPDH.

5-8F and 6–10 B cells were collected and lysed using RIPA lysis buffer (Thermo Fisher Scientific, Inc.) containing protease and phosphatase inhibitors (Roche Applied Science). The supernatants were collected via centrifugation at 14,000 × g for 20 min at 4°C. Total protein was quantified using the BCA protein assay kit. Proteins were resolved using 10% SDS-PAGE, transferred onto PVDF membranes (mass of protein loaded was 20 ug/lane) and blocked with 5% skimmed milk for 1 h at room temperature. The membranes were incubated with primary antibodies (all dilutions 1:1,000) against TNC (cat. no. YT6187; ImmunoWay Biotechnology Company), E-cadherin (cat. no. 14472; Cell Signaling Technology, Inc.), N-cadherin (cat. no. 13116; Cell Signaling Technology, Inc.), VIM (cat. no. 5741; Cell Signaling Technology, Inc.), Slug (cat. no. 9585; Cell Signaling Technology, Inc.), protein kinase B (Akt; cat. no. 4691; Cell Signaling Technology, Inc.), phospho-(p)-Akt (cat. no. 4060; Cell Signaling Technology, Inc.), 70 kDa ribosomal protein S6 kinase 1 (P70S6K, cat. no. 2708; Cell Signaling Technology, Inc.), p-P70S6K (cat. no. 9204; Cell Signaling Technology, Inc.) and β-actin (cat. no. 4970; Cell Signaling Technology, Inc.) overnight at 4°C. Following the primary incubation, membranes were incubated with secondary antibodies (dilution 1:20,000; cat. no. 926-32211; LI-COR Biosciences) for 1 h at room temperature. The membranes were scanned using an Odyssey infrared imaging system (LI-COR Biosciences), and semi-quantitative analysis was performed using ImageJ 1.8.0 software (National Institutes of Health).

Male BALB/c nude mice (n=16; 5 weeks old; 16–18 g) were purchased from Beijing Weitonglihua Laboratory Animal Technology Co., Ltd (https://www.vitalriver.com). Mice were fed under specific pathogen-free conditions at constant temperature (20–25°C) and humidity (40–70%), with ad libitum access to water and food and a 12 h light/dark cycle. A total of 16 mice were randomly divided into four groups (shTNC 5-8F, shNC 5-8F, shTNC 6-10B and shNC 6-10B). Mice were anesthetized using isoflurane (Shenzhen biomart lifescience Co., Ltd, http://cdmocmo.ribobio.com/en/contact-us) with an induction concentration of 4-5% and a maintenance concentration of 1–3% after 1 week of acclimatization. The right forelimb of nude mice was subcutaneously inoculated with TNC knockdown 5–8F cells, control 5-8F cells, TNC knockdown 6-10B cells and control 6-10B cells. A total of 3.0×10⁶ cells were injected per mouse, which was suspended in 200 µl normal saline. Tumor grew on the fifth day, and the weight, tumor length (L) and width (W) were measured every other day. The maximum tumor size reached 15 mm 21 days after cell injection. The maximum tumor volume was 0.755 mm³. Mice were sacrificed via cervical dislocation following anesthesia using 4–5% isoflurane. The tumors were surgically removed, and tumor volumes were calculated using the following equation: Tumor volume=0.52×LxW². All animal experiments were approved by the Ethics Committee of Animal Experiments of Renmin Hospital of Wuhan University [Wuhan, China; approval no. 2020(421)], and performed in compliance with the laboratory animal management and guidelines.

Statistical analysis was performed using GraphPad Prism version 7.0 (GraphPad Software, Inc.). All experiments were performed in triplicate and data are presented as the mean ± standard deviation. A two-tailed unpaired Student's t-test was used to compare differences between two groups, while one-way ANOVA followed by Tukey's post hoc test were used to compare difference between multiple groups. Fisher's exact two-tailed test was used to analyze TNC protein level difference between NPG and NPC tissues and its association with clinicopathological characteristics of NPC tissues. Pearson's correlation analysis was performed to analyze the correlation between genes. P<0.05 was considered to indicate a statistically significant difference.

To assess the transcriptional and expression levels of TNC in NPC, NPC gene datasets were downloaded from the GEO database. As presented in Fig. 1A, TNC transcription levels were significantly higher in the NPC tissues compared with the normal nasopharyngeal tissues, based on the GSE53819 (P=0.006) and GSE13597 (P=0.01) gene expression profiles. Fig. 1B presents TNC expression in 24 NPC metastasis tissues and 24 NPC non-metastasis tissues. The results demonstrated that TNC mRNA levels were significantly higher in NPC metastasis tissues compared with non-metastasis tissues (P=0.033). The distribution of TNC mRNA levels in NPC gene chip is presented in Table I. Representative images of TNC staining in NPC and NPG tissues with classification of ‘−’, ‘+’, ‘++’ and ‘+++’ are presented in Fig. 1C. Tissues graded as ‘−’ was considered as the negative expression TNC group, while ‘+’, ‘++’ and ‘+++’ were considered as the positive TNC expression group. The results demonstrated that 87.8% of NPC tissues and 12.9% of NPG tissues positively expressed TNC (P<0.0001; Table II).

To further investigate the clinicopathological significance of TNC in NPC, the association between TNC expression and the clinicopathological characteristics of patients with NPC was analyzed. As presented in Table III, TNC protein expression was not significantly associated with age and sex of patients with NPC (P>0.05). However, TNC expression was significantly associated with T staging and N staging. The results demonstrated that patients at T3-T4 stages had higher TNC protein expression levels than those at T1-T2 stages (P=0.001). In addition, patients at N2-N3 stages had higher TNC protein expression levels than those at N0-N1 stages (P<0.0001). T staging represents the size and scope of invasion of NPC, while N staging refers to the involvement of regional lymph nodes (29). Taken together, these results suggest that TNC is associated with the malignant progression of NPC.

WGCNA is based on two hypotheses: i) Genes with similar expression patterns may be functionally relates, co-regulated or in the same pathway and ii) gene networks conform to a scale-free distribution (22). Based on these two points, the gene network can be divided into different modules according to the similarity of expression to find the key genes (30–32). Gene expression data were analyzed to remove outliers, and no samples were excluded. A total of 36 samples were included in the WGCNA (Fig. S1). The samples were grouped into different clusters in the sample dendrogram and the trait heatmap is presented in Fig. S2A. The scale R² value and power value (β) decide the independence and average connectivity (22). Fig. S2B and C demonstrate that when β=7, the scale was R²=0.89. Therefore, soft thresholding with β=7 was used to achieve and analyze recognizable co-expression gene modules in datasets. As a result, 15 gene co-expression modules were identified in different colors based on a TOM-based dissimilarity measure (Fig. S2D). The interactions of these co-expression modules were analyzed via Pearson's correlation coefficient, and every module was independent of each other (Fig. S2E). A higher degree of module correlation is indicated by a darker background. The first principal component gene in each module was identified as the feature vector gene representing the overall gene expression level (Fig. 2A). Red squares along the diagonal represent the meta-modules. Module eigengenes were clustered in the dendrogram (Fig. S3A). Modules below the red line indicate that the correlation was >0.8, and these were merged. Fig. 2B presents the association between the module eigengenes and TNC expression. The light cyan module was closely associated with TNC expression (r=0.56; P=0.005). Fig. S3B presents the heatmap of the expression volume of the light cyan module and the histogram of the feature vector gene. The expression of the feature vector gene was highly correlated with the expression of genes in the entire module. The internal connectivity was calculated, and the module membership of each gene was used to identify hub genes that were highly correlated with TNC expression. The internal connectivity measures the position of the gene in the module, and the module membership indicates which module the gene belongs to. The intermodular connectivity was significantly correlated with the module membership (P=0.00001) (Fig. S4). There were three criteria to screen for hub genes, including a gene and specified module significance of >0.2, modular membership value of >0.8 and q<0.01. A total of 52 genes that met the above criteria were identified. Fig. 2C presents the overlapping topological heatmap of these hub genes. In the heatmap, the darker the color, the higher the degree of topological overlap.

To demonstrate the functions of hub genes in the light cyan module, GO and KEGG enrichment analyses in the Molecular Signatures Database were performed. The top 10 terms of biological process (BP), cellular component (CC) and molecular function (MF) in GO analysis are presented in Fig. 3A-C. Hub genes were primarily enriched in the ‘locomotion’, ‘cell motility’, ‘biological adhesion’, ‘enzyme-linked receptor protein signaling pathway’, ‘regulation of cellular component movement’, ‘embryo development’, ‘cellular component morphogenesis’, ‘positive regulation of locomotion’, ‘transmembrane receptor protein tyrosine kinase signaling pathway’ and ‘extracellular structure organization’ in BP (Fig. 3A). In CC, the terms that hub genes were primarily enriched in included ‘endoplasmic reticulum’, ‘anchoring junction’, ‘cell junction’, ‘Golgi apparatus’, ‘cell-substrate junction’, ‘cell surface’, ‘endoplasmic reticulum lumen’, ‘extracellular matrix’, ‘cell leading edge’ and ‘ruffle’ (Fig. 3B). In MF, hub genes were primarily enriched in the terms ‘signaling receptor binding’, ‘protein-containing complex binding’, ‘growth factor binding’, ‘cell adhesion molecular binding’, ‘integrin binding’, ‘cytokine binding’, ‘platelet-derived growth factor receptor binding’, ‘fibroblast growth factor binding’, ‘extracellular matrix binding’ and ‘actin dependent ATPase activity’ (Fig. 3C). The hub genes were enriched in the KEGG pathway of ‘focal adhesion’, ‘regulation of actin cytoskeleton’, ‘ECM receptor interaction’, ‘pathways in cancer’, ‘arrhythmogenic right ventricular cardiomyopathy arch’ and ‘small cell lung cancer’ (Fig. 3E). Also, the hub genes in the light cyan module were enriched for Hallmark gene sets using the Molecular Signatures database. There were 31 Hallmark gene sets enriched, and the top 10 are presented in Fig. 3D, amongst which EMT was the most enriched.

Based on the results of WGCNA and the hallmark enrichment analysis, it is suggested that TNC may activate EMT to promote NPC malignant progression by altering the components of the ECM and influencing cell adhesion. Consequently, stable TNC knockdown in NPC cells was established, which was verified via western blot and RT-qPCR analyses (Fig. 4A and B). The expression levels of the EMT-related markers, VIM, CDH2, Slug and CDH1 were detected in the knockdown and control cells via western blot and RT-qPCR analyses (Fig. 4C and D). The results demonstrated a negative association between TNC expression and the epithelial phenotype, and a positive association between TNC expression and the mesenchymal phenotype. Collectively, these results suggest that suppression of TNC inhibits EMT progression of NPC cells.

The role of TNC in the proliferation and metastasis of NPC cells was assessed. The CCK-8 and EdU assays were performed to assess cell proliferation. The results of the CCK-8 assay demonstrated that TNC knockdown significantly decreased the proliferation of both 5-8F and 6-10B cells (P<0.05) (Fig. 5A). In the EdU assay, TNC knockdown suppressed cell proliferation, with lower proportions of EdU-positive cells in both cell lines (Fig. 5B).

The effect of TNC on NPC cell migration and invasion was determined via the wound healing and Transwell invasion and migration assays. As presented in Fig. 5C, wound closure was notably higher in the control group compared with the TNC knockdown group, which suggests that TNC knockdown inhibited cell migration. In the Transwell migration assay, TNC knockdown significantly decreased the number of migratory cells compared with the control group (P<0.05; Fig. 5D). Similarly, in the invasion assay, TNC knockdown decreased the number of invasive cells in the experimental group compared with the control group. Taken together, these results suggest that TNC knockdown inhibits the proliferation, migration and invasion of NPC cells.

To investigate the potential mechanism by which TNC expression promotes EMT and cell proliferation in NPC cells, the expression of proteins associated with the mTOR signaling pathway was detected via western blot analysis in NPC cells transfected with shNC or shTNC. The results demonstrated that TNC knockdown significantly decreased the expression levels of p-P70S6K and p-AKT in both 5-8F and 6-10B cells (P<0.05; Fig. 6A and B). Collectively, these results suggest that TNC promotes EMT, at least partially via the mTOR signaling pathway.

To determine whether TNC expression affected tumor growth in vivo, a xenograft mouse model was established by injecting shTNC or shNC NPC cells into nude mice. The results demonstrated that injection of cells in which TNC expression was knocked down reduced tumor growth in the xenograft mouse model compared with mice injected with shNC-transfected 5-8F and 6-10B cells (Fig. 7A-C). In addition, there were no significant differences in the weight of mice between the different groups (Fig. 7D). Taken together, these findings suggest that TNC expression regulates tumor growth in vivo.