The ten human SACC tissues and ten paired healthy submandibular gland (SMG) tissues (age, 42-68 years; female:male=6:4) used for the tissue micro-array were collected from patients with ACC at the Peking University School and Hospital of Stomatology (Beijing, China) between 2015.08 and 2016.04. The remining 52 human SACC (age, 31-75 years; female:male=26:26) and 38 separate healthy SMG tissues used to analyse miR-103a-3p and TPD52 expression were collected between 2010.07 and 2018.07 at the Peking University School and Hospital of Stomatology. Patients had not undergone chemotherapy or radiation therapy, and the study was approved and followed the rules of the Ethics Committee of Peking University School and Hospital of Stomatology (permit no. PKUSSIRB-201522040). According to the relative gene expression of miR-103a-3p in 52 patients with SACC examined by reverse transcription-quantitative PCR (RT-qPCR), 52 patients were classified into the high or low miR-103a-3p group depending on the median miR-103a-3p relative gene expression. When the miR-103a-3p expression level of SACC tissue was higher than the median miR-103a-3p expression, the SACC tissue was classified into the high miR-103a-3p group, otherwise the SACC tissue was classified into the low miR-103a-3p group.

The SACC-83 cell line originated from ACC tissue from a patient with SACC in November 1983 (32). The SACC-LM cell line had enhanced lung metastatic features and was isolated following injection of SACC-83 cells into the tail vein of immunodeficient mice (33,34). The SACC-83 and SACC-LM cell lines were collected by the author SLL and kept at Peking University School and Hospital of Stomatology. SACC-83 and SACC-LM cells were cultivated in RPMI-1640 medium supplemented with 10% FBS (both from Gibco; Thermo Fisher Scientific, Inc.). The primary epithelial cells of the SMG (SMG-E or pSMG) cells were derived from the sublingual gland of a 4-year-old boy in November 2016, with cell cultivation performed as previously described (35).

Exosomes isolation from cells supernatants was performed using ultracentrifugation and sucrose cushion. Cell supernatants were subjected to consecutive centrifugation at 300 × g for 10 min, 2,000 × g for 30 min and 10,000 × g for 30 min to remove cellular debris and large vesicles at 4°C. The supernatants were then passed through a Centrifugal Filter (100K; EMD Millipore) to concentrate the solution, followed by a 30% sucrose/deuterium oxide (D2O) cushion. After gradient centrifugation at 100,000 × g for 70 min using an Optima L-90K Ultracentrifuge (Beckman Coulter, Inc.) at 4°C, the exosome-enriched sucrose/D2O was then re-suspended in PBS and the retained exosomes were stored at -80°C. All the isolation procedures were performed at room temperature.

The microarray analysis of tissues (ten human SACC tissues vs. ten paired SMG tissues) and cell exosomes (SACC-83 cells vs. SACC-LM cells) was performed by Shanghai Biotechnology Corporation.

Total RNA was extracted from tissues or cells using TRIzol^® reagent (Invitrogen; Thermo Fisher Scientific, Inc.) according to the manufacturer's protocol. Reverse transcription was performed with the GoScript™ Reverse Transcription System (cat. no. A5001; Promega Corporation) according to the manufacturer's protocol. The procedure used for reverse transcription was as follows: 25°C for 5 min, 42°C for 60 min, 70°C for 15 min followed by hold at 4°C. PCR reactions were performed on an ABI 7500 Sequence Detection System (Thermo Fisher Scientific, Inc.) using FastStart Universal SYBR Green Master (ROX) reagent (Roche Diagnostics GmbH). The thermocycling procedure used for amplification was as follows: Initial denaturation at 50°C for 2 min, 95°C for 10 min, followed by 40 cycles at 95°C for 15 sec and 60°C for 1 min. Dissociation curves were used to assess amplification specificity. cDNA was synthesized for RT-qPCR using the following primer: miR-103a-3p, 5'-GTCGTATCCAGTGCAGGGTCCG AGGTATTCGCACTGGATACGACTCATAG-3'. RT-qPCR was conducted using the following primers: miR-103a-3p- forward (F), 5'-CGCGAGCAGCATTGTACAGGG-3' and reverse (R), 5'-ATCCAGTGCAGGGTCCGAGG-3'; U6-F, 5'-CGAATTTGCGTGTCATCCT-3' and R, 5'-GCTTCGGC AGCACATACTAA-3'; TPD52-F, 5'-TCGGAAGAGGAGCA GGAAGAGC-3' and R, 5'-AGATGTTGCTGTCACGTCTT GCC-3'; and GAPDH-F, 5'-CCTGCCGTCTAGAAACCTG-3' and R, 5'-AGTGGGTGTCGCTGTTGAAGT-3'. Relative gene expression was calculated using the 2^-ΔΔCq method (36).

For miR-103a-3p overexpression and interference, miR-103a-3p mimics or miR-103a-3p inhibitor and their negative controls (NCs) were purchased from Guangzhou RiboBio Co., Ltd. The miR-103a-3p mimics and mimic NC were composed of a double-strand structure. The miR-103a-3p mimics sequences were as follows: Sense, 5'-AGCAGCAUUGU ACAGGGCUAUGA-3' and antisense, 3'- UCGUCGUAACAUG UCCCGAUACU-5'. The miR-103a-3p mimic NC sequences were as follows: Sense, 5'-UUUGUACUACACAAAAGUA CUG-3' and antisense, 3'-AAACAUGAUGUGUUUUCAU GAC-5'. The miR-103a-3p inhibitor sequence was: 5'-UCAUAG CCCUGUACAAUGCUGCU-3'. The miR-103a-3p inhibitor NC sequence was: 5'-CAGUACUU UUGUGUAGUACAAA-3'. To knockdown or overexpress TPD52, two small interfering (si)RNAs specific for TPD52 (siTPD52) and NC siRNAs were purchased from Guangzhou RiboBio Co., Ltd., and a recombinant vector overexpressing TPD52 (TPD52-OE) GV146-CMV-MCS-IRES-EGFP-SV40-neomycin and an empty vector were purchased from Shanghai GeneChem Co., Ltd. The TPD52 siRNA sequences were as follows: 5'-GAC TCTGTCTCAAGTGTTA-3' (siRNA1) and 5'-GCGGAAACT TGGAATCAAT-3' (siRNA2). The siRNA control sequence was: 5'-TTTCTCCGAACGTGTCACG-3'. SACC-83 and SACC-LM cells were transfected with miR-103a-3p mimics, miR-103a-3p inhibitor, TPD52 siRNA, the TPD52-OE vector or the empty vector and their NCs using Lipofectamine^® 2000 reagent (Invitrogen; Thermo Fisher Scientific, Inc.) according to the manufacturer's protocol. miR-103a-3p mimics, TPD52 siRNA and their NC were transfected into cells at a final concentration of 50 nM, and miR-103a-3p inhibitor and inhibitor NC at a final concentration of 100 nM. The plasmid of transfection was 2 µg/well/well in 6 well in 6-well -well plates. The transfec plates. The transfections were performed at 37°C for 48 h when the cells grew up to 50-60% confluence. Subsequent experimentations were performed 48 h post-transfection.

For the wound healing assay, SACC-83 and SACC-LM cells were transfected 2 days in 6-well plate (50-60% confluence), then the transfected cells were digested, counted and seeded into a 96-well plate (3×10⁴ per well). Wound areas were measured using an IncuCyte instrument (Essen Bioscience, Inc.). Briefly, ~3×10⁴ tumour cells in RPMI-1640 medium supplemented with 10% FBS were seeded into 96-well plates. After reaching confluence, the cells were cultured in serum-free RPMI-1640 medium, and after 12 h, a reproducible and uniform scratch wound was made using a WoundMaker™ (Essen BioScience, Inc.) in all wells. After washing three times with PBS, the cells were incubated in medium without FBS at 37°C for 48 h and the migration width change in the scratch wound was observed using an IncuCyte instrument at x10 magnification (light microscope). Migration widths were analysed by IncuCyte™ ZOOM (version 2016A; Essen BioScience, Inc.).

SACC-83 and SACC-LM cells transfected with miR-103a-3p mimics, miR-103a-3p inhibitor, TPD52 siRNA or TPD52-OE vector were seeded into the upper chamber of cell culture inserts (pore size, 8 µm; Falcon; Corning, Inc.) at a density of 8×10⁴ cells/well. The cells in the upper chamber were cultured in RPMI-1640 medium without serum. The lower chamber contained 600 µl RPMI-1640 medium supplemented with 20% FBS. Then, the chamber was incubated at 37°C under a humidified atmosphere containing 5% CO₂ for 19 h. After the incubation, the cells on the lower surface of the insert were stained with a 1% crystal violet solution for 10 min after fixation with 95% ethanol for 30 min, and the process of fixation and stain were performed at room temperature. Then, cells on the lower surface were imaged under a BX51 fluorescence microscope (Olympus Corporation) at x20 magnification. In total, six fields were randomly selected in each group for counting and statistical analysis. The statistical analysis was performed by GraphPad Prism (version 7.00; GraphPad Software, Inc.). Every experiment was repeated independently ≥3 times.

Transfected cells were lysed in RIPA protein lysis buffer (Beyotime Institute of Biotechnology). The protein concentration was measured with a Pierce™ Bicinchoninic Acid Protein Assay kit (Thermo Fisher Scientific, Inc.), following the manufacturer's protocol. Each protein sample (40 µg/lane) was separated by 10-12% SDS-PAGE and transferred onto PVDF membranes. The membranes were blocked with 5% non-fat milk in TBST (20 mM Tris, 137 mM NaCl and 0.1% Tween-20, pH 7.4) for 1 h at room temperature. Then, the PVDF membranes were incubated with primary antibodies in blocking buffer overnight at 4°C and followed by an incubation with horseradish peroxidase-conjugated goat anti-rabbit (1:100,000; cat. no. ZB-2301; OriGene Technologies, Inc.) or anti-mouse (1:100,000; cat. no. ZB-2305; OriGene Technologies, Inc.) IgG antibodies for 1 h at room temperature. The primary antibodies used were anti-β-actin (1:1,000; cat. no. TA-09; OriGene Technologies, Inc.), anti-TPD52 (1:1,000; cat. no. ab182578; Abcam), anti-E-cadherin [1:1,000; cat. no. 3195; Cell Signalling Technology, Inc. (CST)], anti-N-cadherin (1:1,000; cat. no. 13116; CST), anti-vimentin (1:1,000; cat. no. 5741; CST), anti-Snail (1:1,000; cat. no. 3879; CST) and anti-Slug (1:1,000; cat. no. 9585; CST). The immunocomplexes were visualized with a SuperEnhanced chemiluminescence detection kit (CWBIO). All bands were quantified using ImageJ (version 1.8.0; National Institutes of Health), and three independent experiments with three biological replicates each were performed.

The pEZX-MT05 plasmid was used to construct the TPD52 3'UTR luciferase reporter gene plasmid (iGene Biotechnology, Co., Ltd.). Wild-type (wt) or mutant (mut) 3'UTRs of TPD52 were cloned into the downstream sites of the pEZX-MT05 vector (iGene Biotechnology, Co., Ltd.). Next, 50 nM miR-103a-3p mimics or NC and 2.5 µg TPD52-wt or TPD52-mut were co-transfected into SACC cells using Lipofectamine^® 2000 (Invitrogen; Thermo Fisher Scientific, Inc.) according to the manufacturer's protocol. After 48 h, the cell culture medium was collected and the activities of Gaussia Luciferase (GLuc) and Secreted Alkaline Phosphatase (SEAP) in the culture medium were assayed using a Secrete-Pair™ Luminescence Assay kit (cat. No. SPDA-D010; GeneCopoeia, Inc.) according to the manufacturer's protocol. The Gaussia Luciferase (Gluc) and Secreted Alkaline Phosphatase (SEAP) for each sample were measured using a Centro LB 960 Microplate Luminometer (Titertek-Berthold). Using SEAP signal as an internal standard control, signal normalization (ratio of GLuc and SEAP activities) eliminated the impact of transfection efficiency variations and made the normalized GLuc activities of samples of comparison more accurately reflect the true biological events. The ratio of luminescence intensities (RLU, relative light unit) of the GLuc was calculated over SEAP. Then, the normalized GLuc activity (GLuc/SEAP ratio) was compared with all samples.

To elucidate the role of miR-103a-3p in SACC lung metastasis in vivo, 19 female NOD/SCID mice (age, 6 weeks; weight 20-25 g) were purchased from the Beijing Vital River Laboratory Animal Technology Co., Ltd. Animals were maintained under specific pathogen-free conditions at room temperature (20-26°C) with a humidity level of 50-60%, a 12-h light/dark cycle and food and water ad libitum in the animal facility of Peking University School and Hospital of Stomatology. All efforts were made to minimize animal suffering. All animal assays were performed following approval from the Peking University Institutional Animal Care and Use Committee (Beijing, China; permit no. LA2015099). SACC-83 cells transfected with miR-103a-3p mimics or NC (1×10⁶ cells/100 µl/mouse) were injected into the tail vein of NOD/SCID mice. During the whole experiment, the animals were carefully monitored at least twice a week. The maximum percentage of body weight loss observed in mice was <12% from start to endpoint. The mice would be euthanized if they showed serious symptoms including, but not limited to, breathing difficulties, weight loss (>20% of body weight), vocalization, irritability, hunching, stationary, ruffling or poor grooming. The time to detect the tumour metastases in vivo was determined according to our previous experiment (34). After 8 weeks, the mice injected with miR-103a-3p mimics (n=9) or mimic NC (n=10) transfected cells were sacrificed by cervical dislocation, and the procedure was performed out of sight of the other mice. The mortality of the mouse was verified by respiratory arrest and cardiac cessation. Subsequently, the lung tissues were collected from these mice. A maximum of two macroscopic nodules were observed in one lung tissue and the maximum diameter of a nodule was 4 mm. Then, the lung tissues were fixed in 4% paraformaldehyde for 24 h at room temperature, embedded in paraffin and cut into 5-µm thick sections. These sections were subjected to haematoxylin and eosin (H&E) staining to analyse the number of lung metastasis nodules (light microscope; x40 magnification). Sections were deparaffinised and rehydrated with xylene I, xylene II and xylene III for 20 min each. Then, absolute ethanol I and absolute ethanol II for 10 min each and finally 95, 90, 80 and 70% ethanol for 5 min. The sections were wash with deionized water three times for 3 min each time, then stained with haematoxylin for 2-3 min. Next, the sections were rinsed with deionized water for 10 min and stained with eosin for 1 min. Gradient dehydration was performed with 70, 80, 90, 95 and 100% ethanol for 2, 2, 3, 3 and 10 min, respectively, and then with xylene I, xylene II and xylene III for 10 min each. The sections were sealed with neutral gum and placed in a ventilated room overnight. All the procedures were performed at room temperature.

Statistical analyses, including two-tailed unpaired Student's t-test, one-way ANOVA with Bonferroni post-test correction, χ² test and Pearson correlation coefficient analysis, were conducted with SPSS (version 20.0; SPSS, Inc.). Data are presented as the mean ± standard deviation of three independent experiments. These results were repeated in ≥3 independent experiments. P<0.05 was considered to indicate a statistically significant difference.

To identify dysregulated miRNAs in SACC, the expression of miRNAs in SMG vs. SACC tissues was compared by microarray. GraphPad Prism was used to create the heat map generated from ten human SACC and ten paired SMG tissues, which demonstrated that miR-103a-3p ranked 11th among the top 15 upregulated miRNAs (Fig. 1A).

To evaluate the biological function of miR-103a-3p in SACC, its expression was measured in 38 SMG and 52 SACC tissues. miR-103a-3p expression was significantly higher in SACC tissues compared with healthy tissues (Fig. 1B). In addition, miR-103a-3p expression was significantly higher in SACC cells compared with SMG-E (Fig. 1C). Furthermore, SACC-LM cells had increased miR-103a-3p expression compared with SACC-83 cells. The expression of miR-103a-3p was also examined in SACC cell exosomes, and the results indicated that miR-103a-3p was significantly upregulated in SACC-LM cell exosomes (Fig. 1D).

The clinicopathological features of 52 patients with SACC were assessed and the results are presented in Table I. All the patients with SACC were classified into the high or low miR-103a-3p group depending on the median miR-103a-3p expression. High miR-103a-3p expression was associated with the local regional recurrence and lung metastasis. However, no significant associated was identified between miR-103a-3p expression and the other clinicopathological features, including age, sex, tumour size, clinical stage, site, lymph node metastasis, perineural invasion and pathological type. Collectively, these results indicated that miR-103a-3p serves as an oncogene in SACC.

To investigate the function of miR-103a-3p, miR-103a-3p mimics and a mimic NC were transfected into SACC-83 cells (Fig. 2A), and then the effect of miR-103a-3p on cell migration was assessed using a Transwell migration assays. The Transwell migration assay results demonstrated that significantly more cells migrated in the miR-103a-3p mimic group compared with the NC group (Fig. 2B). Subsequently, miR-103a-3p expression was knocked down to assess its migration effects in SACC-LM cells using a miRNA inhibitor (Fig. 2C). The Transwell assay results indicated that significantly fewer cells migrated in the miR-103a-3p inhibitor group compared with the inhibitor NC group (Fig. 2D). The effects of miR-103a-3p on cell proliferation were also assessed using a Cell Counting Kit-8, and it was found that miR-103a-3p had no significant effect on the proliferation of SACC cells (Data S1; Fig. S1).

The effect of miR-103a-3p on cell migration were also examined using wound healing assay. Compared with the NC group, the migratory distance of SACC-83 cells transiently transfected with miR-103a-3p mimics was significantly wider (Fig. 2E). Furthermore, migration was slower in the miR-103a-3p inhibitor group compared with the control group (Fig. 2F). It was also demonstrated that the mimic-transfected SACC-83 cells had a fibroblast-like morphology (Fig. 3A), while knockdown of miR-103a-3p in SACC-LM cells resulted in an epithelial-like morphology (Fig. 3B).

Next, the changes in EMT markers were evaluated. In the miR-103a-3p mimic group, the mesenchymal cell markers N-cadherin, vimentin, Slug and Snail were significantly upregulated, while the epithelial cell marker E-cadherin was significantly downregulated (Fig. 3C). When miR-103a-3p was inhibited, E-cadherin was significantly upregulated, and N-cadherin, Slug, Snail and Vimentin were significantly downregulated (Fig. 3D). Therefore, these results suggested that miR-103a-3p can affect cell function via the EMT process and that miR-103a-3p plays a role in promoting SACC metastasis.

To investigate the effect of miR-103a-3p on SACC lung metastasis in vivo, miR-103a-3p mimic- and mimic NC-transfected SACC-83 cells were injected into the tail veins of NOD/SCID mice. These mice were sacrificed after 8 weeks, and the lung tissues were collected, fixed, sectioned and subjected to H&E staining. The mice injected with mimic- transfected SACC-83 cells had more visible tumour nodules compared with those with mimic NC-transfected SACC-83 cells. (Fig. 3E). Furthermore, the number and extent of lung tumour nodules was significantly greater in the mice injected with miR-103a-3p mimic-transfected cells compared with those injected with mimic NC-transfected cells (Fig. 3F). Thus, these results indicated that miR-103a-3p promotes SACC lung metastasis in vivo.

Next, the possible mechanism via which miR-103a-3p promotes the migration of SACC cells was investigated. The potential interaction of miR-103a-3p with target mRNAs was analysed using different prediction tools, including TargetScan (http://www.targetscan.org/vert_72/), miRTarBase (http://mirtarbase.mbc.nctu.edu.tw/php/index.php) and miRDB (http://mirdb.org/). TPD52 was identified as a target gene of miR-103a-3p (Fig. 4A). To determine whether TPD52 is a direct target of miR-103a-3p, wt and mut 3'UTRs of TPD52 were inserted downstream of the luciferase reporter vector (Fig. 4A). The luciferase reporter assay results demonstrated that miR-103a-3p overexpression significantly reduced luciferase activity in the TPD52-wt group (Fig. 4B) but not in the TPD52-mut group, indicating that miR-103a-3p can directly target TPD52.

RT-qPCR and western blotting results indicated that TPD52 mRNA and protein expression levels were decreased in miR-103a-3p mimic-transfected SACC-83 cells and increased in miR-103a-3p inhibitor-transfected SACC-LM cells (Fig. 4C and D). Moreover, the expression of TPD52 was measured in the 38 SMG and 52 SACC tissues by RT-qPCR. It was identified that TPD52 expression was significantly higher in healthy tissues compared with SACC tissues (Fig. 4E), which is controversial with the known function of miR-103a-3p. SACC-83 cells expressed significantly higher levels of TPD52 compared with SACC-LM cells (Fig. 4F). In addition, the relationship between miR-103a-3p and TPD52 was assessed by RT-qPCR and it was demonstrated that there was a significantly weak negative correlation between miR-103a-3p and TPD52 (Fig. 4G). Therefore, it was speculated that miR-103a-3p promoted cancer migration by targeting TPD52.

To investigate whether miR-103a-3p acts as an important cancer promoter by targeting TPD52, the function of TPD52 was evaluated in SACC cells. TPD52 was overexpressed using plasmid transfection. After transfection with a TPD52 recombinant plasmid, RT-qPCR and western blotting results indicated that the mRNA and protein expression levels of TPD52 were significantly increased in TPD52-OE vector-transfected SACC-LM cells (Fig. 5A and E). The effects of TPD52 on the proliferation of SACC cells were also assessed using a Cell Counting Kit-8 assay, and it was found that TPD52 had no significant effect on the proliferation of SACC cells (Fig. S2).

Subsequently, the function of TPD52 in tumour migration was investigated, and the wound healing assay results demonstrated that TPD52 overexpression significantly suppressed the migration of SACC-LM cells (Fig. 5B). In addition, TPD52 expression was downregulated by transfecting SACC-83 cells with two TPD52 siRNAs. RT-qPCR and western blotting results identified significant decreases in TPD52 mRNA and protein expression levels in TPD52 siRNA-transfected SACC-83 cells (Fig. 5C and F). It was also found that TPD52 knockdown promoted SACC-83 cell migration in wound healing assays (Fig. 5D). Furthermore, western blotting results indicated that TPD52 overexpression decreased the expression levels of the mesenchymal markers N-cadherin and Snail, but increased that of the epithelial marker E-cadherin (Fig. 5E). When TPD52 was knocked down, there were significant increases in N-cadherin and Snail but a significant decrease in E-cadherin (Fig. 5F). These data suggested that TPD52 inhibits the migration of SACC cells.

The associations between clinicopathological variables and TPD52 mRNA expression were examined in patients with SACC, and the results are presented in Table SI. All patients with SACC were classified into high-TPD52 and low-TPD52 groups depending on the median TPD52 mRNA expression. Low TPD52 mRNA expression was associated with a high incidence of lung metastasis and a high rate of perineural invasion. However, no significant associations were observed between TPD52 expression and the other clinicopathological features.

To further verify whether the promoting effects of miR-103a-3p on SACC cell migration and EMT are mediated by targeting TPD52, cell phenotype changes were examined after overexpressing TPD52 in miR-103a-3p mimic-transfected SACC cells. The wound healing assay results suggested that TPD52 overexpression significantly decreased the migratory distance of miR-103a-3p mimic-transfected SACC cells compared with the mimics + vector groups (Fig. 6A). Furthermore, the Transwell migration assay results demonstrated a inhibitory effect in miR-103a-3p mimic SACC cells overexpressing TPD52 compared with the mimics + vector group (Fig. 6B).

Western blotting results suggested that overexpressing TPD52 rescued the EMT effect of miR-103a-3p over-expression, as indicated by decreased expression of the mesenchymal markers N-cadherin, Vimentin, Snail and Slug, and increased expression of the epithelial marker E-cadherin in miR-103a-3p mimic-transfected SACC cells (Fig. 6C). These findings further indicated that overexpression of miR-103a-3p promotes SACC cell migration and EMT by targeting TPD52. TPD52 may also negatively regulate its own expression. Moreover, knockdown of TPD52 inhibited the expression of miR-103a-3p in SACC-83 cells (Fig. 6D), while TPD52 overexpression promoted the expression of miR-103a-3p in SACC-LM cells (Fig. 6E). The schematic diagram demonstrated the role of miR-103a-3p and TPD52 in SACC lung metastasis (Fig. 6F). miR-103a-3p promoted lung metastasis by targeting TPD52, and TPD52 moderated the expression of miR-103a-3p in turn.