This study was performed under a protocol approved by the Institutional Review Boards of Affiliated Stomatological Hospital of Nanjing Medical College and all examinations were performed after obtaining written informed consents.

A total of 232 OSCC patients hospitalized in Department of Oral and Maxillofacial Surgery, Stomatological Hospital of Nanjing Medical College from January 2008 to January 2014 were included in this study. None of the patients received preoperative radiotherapy or chemotherapy. Each patient was pathologically diagnosed as squamous cell carcinoma and graded according to WHO criteria by two pathologists. All patients were followed up at least 24 months and the complete clinicopathological data including recurrence and metastasis were collected.

These OSCC samples were sent to Jiangsu Key Laboratory of Biological Cancer Therapy, Xuzhou Medical College (Xuzhou, China) to manufacture the OSCC TMA. The array dot diameter was 1.5 mm, and each dot represented a tissue spot from one individual specimen that was selected and pathologically confirmed.

Immunohistochemistry of OSCC TMA was performed with the streptavidin-peroxidase (Sp) method using a standard Sp kit (Zhongshan Biotech, Beijing, China). The TMA slide was incubated with monoclonal mouse anti-RUNX3 antibody (1:250, Abcam, USA) overnight at 4°C, and diaminobenzidine (DAB; Zhongshan Biotech, China) was used to produce a brown precipitate. Negative controls were obtained by substituting primary antibodies with non-immune serum. The immunoreactivity was assessed blindly by two independent observers using Zeiss Imager Z1 microscope, and the image was collected by AxioCam MRc5 camera. The expression of RUNX3 was graded as positive when 5% of tumor cells showed immunopositivity, while the biopsies with <5% were considered negative (18).

Human OSCC cell lines HN6 and Cal27 were purchased from the Shanghai Institute of Biochemistry and Cell Biology, Chinese Academy of Sciences (Shanghai, China). Human umbilical vascular endothelial cells (HUVECs) were obtained from KeyGen Biotech (Nanjing, China). HN6 and Cal27 cells were cultured in RPMI-1640 medium supplemented with 10% fetal calf serum (Invitrogen, Shanghai, China), and HUVECs were cultured in DMEM medium supplemented with 10% fetal calf serum. Cells were placed at 37°C in humidified incubator with 95% air, 5% CO₂.

The pFlag-control and pFlag-RUNX3 expression plasmids were obtained from Jiangsu Key Laboratory of Biological Cancer Therapy, Xuzhou Medical College (Xuzhou, China). Transfection of the pFlag-control and the pFlag-RUNX3 plasmids into HN6 and Cal27 was carried out using Lipofectamine 2000 transfection reagent (Invitrogen) following the manufacturer's protocol.

Cells were seeded in 6-well plates at a density of 5×10⁴ cells/well and cultured to confluence. The cell monolayer was serum-free starved overnight. Confluent cell monolayer was then scraped with a yellow pipette tip for scratch wounds and washed twice with PBS. After 24-h incubation, the cell images were captured in the same position. The wound areas were evaluated by AxioVision 4.8 software.

Cell migration was determined using Transwell migration assay (8-µm pore size; Cell Biolabs). Cells were seeded into the upper chamber in serum-free medium at a density of 5×10⁴. After a 12-h incubation at 37°C, cells in the upper chamber were carefully removed with a cotton swab. Cells traversed the membrane were fixed in methanol and stained with leucocrystal violet. Images were taken with an inverted microscope. Five random selected fields of view were captured and the invasive cells were counted.

The Transwell filter inserts were coated with Matrigel (BD Biosciences, NJ, USA). Transfected 0.5×10⁵ HN6 cells and 1×10⁵ Cal27 cells were seeded in serum-free medium in the upper chamber. After a 24-h incubation at 37°C, cells at the top of the Matrigel were gently removed with a cotton swab. Invasive cells at the bottom of Matrigel were fixed in methanol, stained with leucocrystal violet and counted.

Transfected HN6 and Cal27 cells (1×10⁶) were cultured in 6-well plate with serum-free medium for 24 h. The medium was collected as a conditioned medium. Cellular proliferation was performed using Cell counting kit-8 (CCK-8) purchased from Beyotime Institute of Biotechnology (Nanjing, China). Briefly, 2×10⁴ HUVECs were suspended in conditioned medium or vehicle control and seeded at a density of 2×10⁴ per well in a 96-well plate. After incubation at 37°C for 24 h, cell proliferation was detected according to the manufacturer's instructions.

HUVECs were starved overnight and then seeded at a density of 2×10⁴ per well in a 48-well plate pre-coated with Matrigel. After incubation at 37°C for 30 min, the medium was replaced with conditioned medium or vehicle control. After a 24-h incubation, three random selected fields of view were captured. Quantified evaluation of tube formation was obtained from measuring the length of tube-like structures. The experiments were repeated three times.

Cells were washed with PBS and lysed in radioimmunoprecipitation lysis buffer (Beyotime, China) containing 1 mM phenylmethylsulfonyl fluoride based on the manufacturer's instructions. The cell protein (10 µg) was separated on a 12% SDS-polyacrylamide gel. The protein was then transferred to nitrocellulose membrane and incubated overnight at 4°C with the following antibodies: mouse anti-RUNX3 (1:2,000, Abcam), mouse anti-VEGF (1:1,000, Abcam), rabbit anti-MMP9 and mouse anti-β-actin (1:1,000, Boster Biotechnology, China). Membranes were then washed in PBS and incubated with secondary antibody (goat anti-rabbit and goat anti-mouse IgG) for 2 h and then washed in PBST (PBS containing Tween-20) at room temperature. After that, the membrane was stained by coloration fluid which contains 10 ml alkaline phosphatase buffer, 33 µl BCIP, and 66 µl NBT. Finally, the protein bands were detected using SuperSignal West Pico Chemiluminescent Substrate (Thermo Fisher Scientific, Waltham, MA, USA) and exposed to Kodak X-ray film. Three independent trials of each experiment were carried out.

HN6 and Cal27 cells were seeded in 6-well plates at a density of 1×10⁶ cells per well. Then, cells were transfected with pFlag-control and pFlag-RUNX3 with serum-free medium. The supernatants were collected 24 h after transfection. MMP-9 and VEGF concentration were determined using Quantikine ELISA kits according to the manufacturer's instructions (R&D Systems, MN, USA).

Statistical analysis was performed with SPSS 20 software (SPSS Inc., IL, USA) and data were expressed as the mean ± SD. The associations between RUNX3 staining and the clinicopathologic parameters were evaluated by two-sided Fisher's exact and χ² tests. The correlation between statue of RUNX3 expression and patient's survival was analyzed by Kaplan-Meier survival analysis. For migration and invasion assays, endothelial cell tube formation and CCK-8 cell proliferation assays, and data of western blotting and ELISA, Student's t-test was used. Differences were considered significant when P<0.05.

We first determined whether RUNX3 expression was changed in human OSCC. Immunohistochemistry of the OSCC TMA was performed to detect the expression of RUNX3 protein and the distribution of RUNX3-positive cells. We observed RUNX3 staining displayed different patterns: 93 (40.1%) cases presented with positive nuclear staining and 85 (36.6%) cases with positive cytoplasmic staining, and only 54 (23.2%) cases showed underexpression of RUNX3 protein with undetectable or low levels of staining (Fig. 1A). The cytoplasmic staining, termed as ‘RUNX3 mislocalization’ in the cytoplasm, suggested RUNX3 protein was inactive in a non-functional form as a tumor suppressor (7,10). Both underexpression and mislocalization of RUNX3 protein were aberrant and represented RUNX3 gene impairment in cancer cells; hence we analyzed these two patterns together as negative expression.

Based on TMA results, we further investigated the relationship between RUNX3 expression and the clinicopathological parameters of OSCC. We found that positive expression of RUNX3 was more frequent in T1/T2 compared with T3/T4 cases (P=0.0021). Also, we found that negative expression of RUNX3 was significantly correlated with lymph node statue and clinical stage (P<0.0001 and P=0.000, respectively). The correlations between expression of RUNX3 with other clinicopathological variables, including patient age, sex, invasion stage and histological grade, were not significant (Fig. 1B and Table I).

We then analyzed whether the RUNX3 expression was associated with the survival of patients by Kaplan-Meier survival analysis. The cases of clinical stage I were excluded because of favorable prognosis and then 110 cases (among 156 cases of clinical stage II–IV) with at least 60 months follow-up were chosen for prognosis analysis. Our data revealed that negative expression of RUNX3 in OSCC correlated with both 5-year overall and disease-specific patient survival (P=0.0348 and P=0.0301, respectively, log-rank test; Fig. 1C).

To determine the effects of RUNX3 restoration on OSCC cell migration and invasion, we transiently transfected HN6 and Cal27 cells with pFlag-control and pFlag-RUNX3 plasmids. Twenty-four hours after transfection, RUNX3 protein was significantly overexpressed in cancer cells (Fig. 2A and B). The protein detection of RUNX3 by western blotting showed two bands, and this could be due to protein mislocalization, which was reported in a previous study (7). The results of wound healing and Transwell assays showed that RUNX3 tansfected cells showed less migration ability compared with vehicle control (Fig. 2C-E). Moreover, the invasion assay proved that RUNX3 restoration inhibited cell invasive ability of HN6 and Cal27 cells in Matrigel-coated Transwell (Fig. 2F and G).

To further determine the effect of restored RUNX3 expression on angiogenic potential of human OSCC cells, we used the supernatant of HN6 and Cal27 cells as conditioned medium to culture the endothelial cell line HUVECs. We observed that conditioned medium from HN6 and Cal27 cells transfected with pFlag-RUNX3 inhibited proliferation of HUVECs compared with those of pFlag-control transfected cells (Fig. 3A and B). In tube formation assay, the average tube length of HUVECs cultured with supernatant from pFlag-RUNX3 transfected HN6 and Cal27 cells was significantly decreased in contrast with that of vehicle control, as shown in Fig. 3C and D.

To investigate the mechanisms of RUNX3 regulating invasion and angiogenesis, we performed western blotting and ELISA to detect the MMPs and VEGF levels in OSCC. We found expression of MMP-9 and VEGF protein was downregulated significantly in HN6 and Cal27 cells transfected with pFlag-RUNX3 (Fig. 4A-C). Also, the restoration of RUNX3 in HN6 and Cal27 cells led to a significant reduction of MMP-9 and VEGF secretion in conditioned medium (Fig. 4D and E).