Primary normal human oral keratinocytes (NHOKs) were prepared and maintained as previously reported (35). Briefly, NHOKs were isolated from human gingival tissue specimens obtained from healthy volunteers (age range, 20–30 years) who were undergoing oral surgery. NHOKs were cultured in keratinocyte growth medium containing 0.15 mM calcium and a supplementary growth factor bullet kit (KGM; Clonetics, San Diego, CA, USA). Keratinocytes in their second passage were used in the described experiments. Primary normal human dermal fibroblasts (NHDFs) were prepared and maintained as previously described (36). The HOK-16B line, human oral keratinocytes immortalized by transfecting cells with the cloned HPV-16 genome (37), was also cultured in KGM. The cancer cell lines SCC-4, SCC-9, FaDu and SiHa were purchased from the American Type Culture Collection (ATCC; Rockville, MD, USA). SCC-4 and SCC-9 cell lines were cultured in Dulbecco's modified minimum essential medium (DMEM)/Ham's F12 (Gibco-BRL, Bethesda, MD, USA) supplemented with 10% fetal bovine serum (FBS) and 0.4 µg/ml hydrocortisone. Among three oral cancer cell lines, CTHOK-16B-BaP and CTHOK-16B-DMBA, which were tumorigenically transformed by chronic exposure of the HOK-16B cells to benzo(a)pyrene and 7,12-dimethylbenz(a) anthracene, respectively (38), and Spt-HOK80, which was spontaneously transformed from primary human oral kera-tinocytes (39) were cultured in DMEM supplemented with 10% FBS and 0.4 µg/ml hydrocortisone. Cell lines such as CTHOK-16B-DMBA and Spt-HOK80 were established by our group in 1995 and 2010, respectively (38,39). HOK-16B-BaP cell line, the original name of which is CTHOK-16B-BaP, was first reported by our group in 1995 (38). Other oral cancer cell lines KOSCC-11, KOSCC-25A, KOSCC-25B, KOSCC-25C, KOSCC-25D, KOSCC-25E, KOSCC-33A and KOSCC-33B (40) were also cultured in DMEM supplemented with 10% FBS and 0.4 µg/ml hydrocortisone. The SiHa line was grown in DMEM supplemented with 10% FBS, while the FaDu line was grown in DMEM supplemented with 10% FBS and 0.4 µg/ml hydrocortisone.

The whole blood, whole saliva and OSCC tissue samples used in this study were obtained from OSCC patients who underwent surgical resection at the Department of Oral and Maxillofacial Surgery, Seoul National University Dental Hospital, and healthy controls who did not have signs or symptoms of inflammation of the gingival tissues and systemic diseases. In the present study, the oral hygiene of oral cancer patients was not taken into consideration. This was because OSCC patients with poor oral hygiene followed by chronic inflammation, including periodontitis, have high incidence of oral cancer clinically, making it difficult to exclude patients with inflammation. All procedures for obtaining human gingival tissue specimens, whole blood, whole saliva and cancer specimens were reviewed and approved by the Institutional Review Board on Human Subjects Research and the Ethics Committee at Seoul National University Dental Hospital.

Four hundred microliters of the whole saliva were used for RNA isolation. Saliva samples were extracted using the mirVana™ miRNA Isolation kit (Ambion, Austin, TX, USA) according to the manufacturer's instructions. For the initial lysis step, we used 1 ml of lysis/binding solution per 400 µl saliva sample. Total RNAs including small RNAs were isolated in whole blood and cells by miRNeasy Mini kit (Qiagen, Valencia, CA, USA) according to the manufacturer's instructions.

Cyanine 3-labeled complementary RNA was prepared from total RNA including miRNA (500 ng) using Agilent's Low RNA Input Linear Amplification kit (Agilent Technologies, Santa Clara, CA, USA). Labeled cRNA was applied to microarray (Human miRNA Microarray Release 19.0, 8×60K; Agilent Technologies) using Agilent's Gene Expression Hybridization kit. Hybridized microarray was washed using Agilent's Gene Expression Wash Buffer kit. The microarray chip was scanned using Agilent's DNA micro-array scanner, and the raw signal density was acquired from Feature Extraction software. The differential gene expression was classified by Benjamini-Hochberg's false discovery ratio method using GeneSpring GX12 software.

Complementary DNAs (cDNAs) were synthesized from total RNA including miRNA using TaqMan^® MicroRNA reverse transcription kit with gene-specific primers (Applied Biosystems, Foster City, CA, USA). Reverse transcription reactions (for final quantity or concentrations) contained 10 ng RNA samples, 1 mM of dNTPs, 1X RT primer, 1X RT buffer, 3.8 U of RNase inhibitor and 50 U of reverse transcriptase. The 15 µl reactions were incubated for 30 min at 16°C, 30 min at 42°C, 5 min at 85°C and then held at 4°C. Quantitative real-time PCR (RT-qPCR) quantification of miRNA expression was carried out using TaqMan^® MicroRNA assays kit (Applied Biosystems) according to the manufacturer's instructions. The primers used were: miR-146a-5p (ID, 000468) and RNU44 (ID, 001094) (Applied Biosystems). The 20 µl PCR included 1.33 µl RT product, 1X TaqMan^® Universal Master Mix II without UNG, and 1X TaqMan-primers mix (Applied Biosystems). Reactions were incubated in a 96-well plate at 95°C for 10 min, followed by 40 cycles of 95°C for 15 sec and 60°C for 1 min. The threshold cycle (Ct) was determined using default threshold settings. All experiments were done in quadruples. The RNU44 was used as a control to normalize miRNA input in the RT-qPCR assay.

The presence of 'high-risk' HPV (type 16, 18 and 33) DNA was determined by reverse transcription polymerase chain reaction (RT-PCR). Total RNA was reverse transcribed to cDNA, which was used for PCR amplification of HPV E5 and E6 (38 cycles) and GAPDH (28 cycles; denaturation at 95°C for 20 sec, annealing at 60°C for 10 sec and extension at 70°C for 10 sec). The primers (Table I) amplified the E5 and E6 genes of the three types of HPV. The PCR products were resolved on a 1.5% agarose gel, stained with ethidium bromide and visualized using a UV illuminator.

A non-specific control siRNA (Invitrogen, Carlsbad, CA, USA) and siRNAs against human TRAF6 (40 nM; Sigma-Aldrich, St. Louis, MO, USA) and FADD (40 nM; Santa Cruz Biotechnology, Santa Cruz, CA, USA) were used for gene silencing. To regulate miR-146a-5p expression in cells, specific miRVana mimic, inhibitor (Ambion) and each respective negative control (Thermo Fisher Scientific, Waltham, MA, USA; Applied Biosystems) were transfected using Lipofectamine RNAiMAX transfection reagent (Invitrogen). Cells were plated into 6-well plates at a density of 3×10⁵ cells/well, cultured overnight, and then transfected with 20 nM hsa-miR-146a-5p mimic or 20 nM hsa-miR-146a-5p inhibitor using 8 µl of Lipofectamine RNAiMAX transfection reagent. After 2 days, the transfected cells were harvested and analyzed by cell viability assay, TUNEL assay and immunoblotting.

The viabilities of cells were investigated using the EZ-Cytox Cell Viability Assay kit (water-soluble tetrazolium salt method; Daeil Lab Service Co., Ltd., Seoul, Korea). Cells (3×10³ cells/100 µl) were seeded onto a 96-well microplate, adapted overnight and then transfected with hsa-miR-146a-5p mimic or anti-miR hsa-miR-146a-5p inhibitor for 48 h at 37°C. The water-soluble tetrazolium salt reagent solution (10 µl) was added to each well, and the plate was incubated for 2 h at 37°C. The absorbance at 450 nm was then measured using a microplate reader.

Apoptosis was analyzed in situ by the TUNEL assay using the In Situ Cell Death Detection kit, Fluorescein (Roche Applied Science, Mannheim, German) according to the manufacturer's instructions. To ensure a representative count, each cell culture slide was divided into quarters, and at least two fields were photographed with Olympus BX50 fluorescence microscope. The number of apoptotic cells was measured by counting the TUNEL-positive cells. Average number and standard deviation were calculated from four independent experiments.

Cells (1×10⁶) were plated onto 60-mm dishes and then washed with ice-cold phosphate-buffered saline (PBS) and lysed with 150 µl RIPA buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, 1 mM PMSF, 2 mM Na₃VO₄ and 1 mM glycerol phosphate) containing a protease inhibitor cocktail. The proteins present in the RIPA lysates were denatured using SDS sample buffer, resolved by SDS-PAGE and electroblotted onto nitrocellulose membranes. The membranes were immunoblotted with primary antibodies against Smad4; FADD; p-FADD (Ser194); NF-κB p65; JNK; p-JNK (Thr183/Tyr185); p38; phospho-p38 (Thr180/Tyr182); Bcl-2 (Cell Signaling Technology, Danvers, MA, USA); Mad (Abcam, Cambridge, MA, USA); p53 (Millipore, Bedford, MA, USA); TRAF6; TGFβ1; p15; c-Jun; p-c-Jun (Santa Cruz Biotechnology); or actin (Sigma-Aldrich). All blots were then incubated with anti-rabbit horseradish peroxidase-conjugated secondary antibodies (Cell Signaling Technology). The signals were detected by electrochemiluminescence (iNtRON Biotechnology, Seongnam-si, Korea).

Comparison of the distribution of −ΔCT values between groups were analyzed by the language R (41), and ROC curves were obtained with the package ROCR (42). All other data are presented as the mean ± SD. The statistical analyses of data were performed using the Statistica 6.0 software package (StatSoft, Tulsa, OK, USA). Results were compared using analysis of variance tests. When significant differences were found, pairwise comparisons were performed using Scheffe's adjustment. Statistical significance was also calculated using a two-tailed Student's t-test. Differences with a P<0.05 were considered statistically significant.

To explore possible miRNAs involved in the transformation of normal oral keratinocytes, we analyzed the difference in miRNA expression between NHOKs and HPV-16-immortalized human oral keratinocytes, HOK-16B, by miRNA microarray with cut-off values of 2-fold increase or decrease. Twenty-three miRNAs were upregulated while 46 miRNAs were downregulated in immortalized human oral keratinocytes compared to NHOKs (Table II). Microarray also revealed miR-3934-5p, miR-365b-5p, miR-146a-5p and miR-3138 being most abun dant in immortalized human oral keratinocytes (Table II). On the basis of this, we hypothesized that miR-146a-5p expression is overexpressed in OSCCs and may underlie OSCC initiation and progression. Furthermore, the miR-146a-5p and its target genes were of interest, because miR-146a-5p regulates signal transduction of TGF-β, a potent antitumor cytokine, by repressing Smad4 in human cancer (43) and its biological function in carcinogenesis is not fully understood. To test our hypothesis, we first examined miR-146a-5p expression in 13 OSCC cell lines and other 2 SCC cell lines, and in matched 20 pairs of OSCC and normal tissues, 30 whole blood and 21 saliva samples from pre-operative OSCC patients and 10 healthy controls by a TaqMan miRNA assay. miR-146a-5p expression was significantly upregulated in all 13 OSCC cell lines examined compared to the normal oral keratinocytes; however, its expression in other 2 SCC cell lines established from pharynx and cervix carcinomas was extremely lower (~10%) than that observed in NHOKs (Fig. 1A). Analysis of matched OSCC-normal tissue pairs demonstrated an increase in miR-146a-5p expression in 65% (13/20) of the OSCCs compared to the matched normal tissues (Fig. 1B). Blood miR-146a-5p levels were significantly higher in OSCC patients compared to the healthy controls with mean −ΔCT values of 2.2300363 and 0.1399096, respectively (P=0.0007967) (Fig. 1C). Similarly, miR-146a-5p levels in saliva were significantly abundant in OSCC patients compared to the healthy controls with mean −ΔCT values of 0.7711158 and −2.7807178, respectively (P=0.00007308) (Fig. 1D). However, among the 18 OSCC patients whose whole blood and saliva were collected, each miR-146a-5p level in blood and saliva showed minimal evidence of a true correlation when tested by the Pearson method (correlation coefficient =−0.08920913, P=0.7248). To evaluate the diagnostic significance of miR-146a-5p in determining OSCC status, receiver operating characteristic (ROC) curves were obtained for each blood and saliva sample sets, and the area under the curves (AUC) was 0.8033333 and 0.9004329, respectively (Fig. 1C and D). Collectively, these data demonstrate that miR-146a-5p is highly upregulated in immortalized human oral keratinocytes and OSCC cell lines, and that miR-146a-5p levels in OSCC tissues, whole blood and whole saliva significantly increase in OSCC patients compared with healthy controls.

Because HPV-16 E5 oncogene is reported to modulate the expression of miR-146a-5p (44), we examined whether miR-146a-5p expression is correlated with HPV E5 expression in OSCC and SCC cell lines. We first examined whether cancer cell lines expressed 'high-risk' HPV E5 and E6 transcripts (type 16, 18 and 33) and compared to their cellular miR-146a-5p levels. HPV-16 E5 and E6 transcripts were detected in HOK-16, CTHOK-16B-BaP, CTHOK-16B-DMBA and SiHa cell lines (Fig. 1E). Other 12 OSCC or SCC cell lines, however, proved to be free of HPV DNA (data not shown). The miR-146a-5p expression was significantly upregulated in HOK-16B, CTHOK-16B-BaP and CTHOK-16B-DMBA, which expressed HPV-16 E5; however, its expression in SiHa, which expressed HPV-16 E5, was extremely lower (~10%) than that observed in NHOKs (Fig. 1A). Unexpectedly, our data support that miR-146a-5p expression is not necessarily correlated with HPV-16 E5 expression in human cancer cell lines.

In view of the observed widespread upregulation of miR-146a-5p expression in OSCC cell lines and OSCC clinical specimens, we assessed whether miR-146a-5p upregulation has associated proliferation potential in NHOKs, SCC-9, HEK293 and NHDFs. Compared to that in negative control-transfected cells, the proliferation of NHOKs and SCC-9 was significantly enhanced by introduction of exogenous miR-146a-5p (Fig. 2A), which was verified by a TaqMan miRNA assay (data not shown). Additionally, NHDFs responded with small but significant increase in cell proliferation. Distinctively, exogenous miR-146a-5p expression inhibited the proliferation of HEK293 cell line (Fig. 2A). Furthermore, we determined whether high miR-146a-5p expression is associated with apoptosis in NHOKs and SCC-9 using TUNEL assay. Exogenous miR-146a-5p expression caused inhibition of apoptosis in NHOKs, while it increased apoptosis in SCC-9 (Fig. 2B). Collectively, overexpression of miR-146a-5p by transfection of cells with miR-146a-5p mimic resulted in diverse effects on proliferation and apoptosis depending on the cell types.

To elucidate the underlying mechanisms associated with the different outcomes of miR-146a-5p on proliferation and apoptosis in different cell types, we examined the protein levels of several known miR-146a-5p target genes, such as TRAF6, SMAD4 and FADD and their downstream signaling molecules by immunoblotting. After transfection of cells with miR-146a-5p mimic, TRAF6 expression was invariably downregulated in all cell types tested; however, Smad4 expression remained constant in NHOKs and SCC-9, while it decreased in HEK293 and NHDFs (Fig. 2C). FADD expression was slightly downregulated only in HEK293 cells (Fig. 2C). Notably, exogenous miR-146a-5p expression downregulated phosphorylated FADD at Ser194 (p-FADD-Ser194) in NHOKs and HEK293 cells but not in SCC-9 and NHDFs (Fig. 2C). JNK and p38 are MAPKs at the downstream of TRAF6 in TGF-β signaling that are both activated by phosphorylation (27). Exogenous miR-146a-5p expression deactivated JNK in SCC-9, but had no discernable effects in NHOKs, HEK293 and NHDFs (Fig. 2C). Similarly, exogenous miR-146a-5p expression also deactivated p38 in SCC-9. These data suggest that miR-146a-5p hampers translation of its target genes in a cell type-specific fashion.

In the opposite approach, we suppressed endogenous miR-146a-5p expression in SCC-9 followed by functional assays. Suppression (~46%) of endogenous miR-146a-5p expression (Fig. 3A) by transfection of cells with miR-146a-5p inhibitor significantly inhibited cell proliferation compared to the negative control-transfected cells (Fig. 3B). SCC-9 rather than SCC-4 was selected for the further studies to clearly elucidate the effects of miR-146a-5p on cellular processes of OSCC cell lines in this study. It is important to finely control the expression levels of miR-146a-5p in the cell lines by its mimic or inhibitor. Since the basal miR-146a-5p expression level in SCC-4 was too high (Fig. 1A), 54% reduction of its expression by the miR-146a-5p inhibitor, as shown in Fig. 3A, seemed still high, making it ambiguous to clarify its reduced function. On the other hand, SCC-9 had relatively lower, but significantly higher expression level of miR-146a-5p compared to normal oral keratinocytes (Fig. 1A), making it more effective to examine the clear effect of the miR-146a-5p mimic or inhibitor transfection. Next, to assess the underlying mechanisms of miR-146a-5p inhibitor-induced inhibition of cell proliferation, we examined the protein level of TRAF6, one of the miR-146a-5p target genes. miR-146a-5p inhibitor induced upregulation of TRAF6 (Fig. 3C). Our data suggest that miR-146a-5p inhibitor inhibits proliferation of SCC-9 by modulating the expression level of TRAF6.

To assess the functional significance of the universal effect of exogenous miR-146a-5p expression, which led to TRAF6 downregulation, we silenced endogenous TRAF6 expression in exponentially proliferating NHOKs and SCC-9 by transfecting the cells with a specific small interfering RNA (siRNA). TRAF6 knockdown (Fig. 4A, upper panel) led to significantly increased cell proliferation (Fig. 4A, lower panel) and inhibited apoptosis of NHOKs (Fig. 4B), mimicking the effects of exogenous miR-146a-5p expression (Fig. 2A and B). By contrast, TRAF6 knockdown significantly decreased cell proliferation (Fig. 4D) and increased apoptosis (Fig. 4E) of SCC-9. Immunoblotting revealed that TRAF6 knockdown resulted in marked deactivation of JNK and p38 in SCC-9 (Fig. 4F), giving very similar results to those of the exogenous miR-146a-5p expression (Fig. 2C). However, TRAF6 knockdown did not affect the phosphorylation of JNK and p38 but decreased the amount of p-FADD-Ser194 in NHOKs (Fig. 4C). Overall, our data suggest two points; first, TRAF6 knockdown can be either proliferative or inhibitory in terms of cell proliferation depending on the cellular context, and second, the TRAF6-JNK and TRAF6-p38 channels can be constitutively active in SCC-9.

The phenotypically contradictory results of TRAF6 silencing imply antagonistic effectors downstream of TRAF6 signaling. Since JNK and p38 are activated by TRAF6 in certain circumstances, we defined their roles using specific kinase inhibitors targeting each of them. Treatment of NHOKs with SP600125, a JNK specific inhibitor, significantly reduced cell proliferation (Fig. 5A) and increased apoptosis (Fig. 5B) in a dose-dependent manner, as well as reduced proliferation of SCC-9, similarly (Fig. 5C). Immunoblotting revealed that SP600125 inhibited phosphorylation of JNK and its substrate c-Jun in a dose-dependent manner; compellingly, p-FADD-Ser194 was also reduced in SCC-9 (Fig. 5D). By contrast, SB203580, a p38 inhibitor, resulted in an insignificant reduction of proliferation of NHOKs (Fig. 5E) but reduced proliferation of SCC-9 (Fig. 5F). Unexpectedly, reduction of phosphorylated p38 was marginal, and an unanticipated upregulation of JNK and phosphorylated JNK was observed in SCC-9 by immunoblotting. JNK activation by SB203580 also increased phosphorylated c-Jun (Fig. 5G). Similarly, TGF-β1 activated JNK and c-Jun, but not p38 in exponentially proliferating NHOKs (Fig. 5H). These likely secondary effects could have dampened the effects of p38 suppression on cell proliferation. Nevertheless, these data suggest that JNK/c-Jun signaling pathway plays a key role in regulation of proliferation of SCC-9.

Since exogenous miR-146a-5p expression reduced the expression of p-FADD-Ser194 in NHOKs and HEK293 (Fig. 2C), we opted to investigate the functional relevance of p-FADD-Ser194 in cell proliferation. Proliferation was drastically inhibited by treatment of NHOKs and SCC-9 with NSC 47147 (Fig. 6A, lower panel), a kinase inhibitor that is known to inhibit FADD phosphorylation (45). Notably, expression of both FADD and p-FADD was suppressed (Fig. 6A, upper panel). Thus, to confirm the effects of simultaneous downregulation of both FADD and p-FADD-Ser194 expression, we suppressed their expression by transfection of cells with a FADD-specific siRNA. Unexpectedly, simultaneous silencing of both FADD and p-FADD-Ser194 (Fig. 6B, upper panel) did not result in cell proliferation blockage of NHOKs and SCC-9 (Fig. 6B, lower panel). Therefore, to verify these conflicting results, cells transfected with a FADD-specific siRNA were subsequently treated with NSC 47147. NSC 47147 was able to reduce cell proliferation in the absence of FADD and p-FADD-Ser194 expression, and the inhibition pattern of cell proliferation was strikingly similar to that of control siRNA-transfected cells (Fig. 6C). Immunoblotting revealed that NSC 47147 not only inhibited phosphorylation of FADD, but also inhibited phosphorylation of JNK and p38, suggesting a broad inhibition of MAPK pathway (Fig. 6D). Overall, our data suggest that FADD and p-FADD-Ser194 expression has a minimal effect on the proliferation of NHOKs and SCC-9.