The human gastric cancer cell lines, AGS, SGC-7901, BGC-823, MKN-45 and MGC-803, were obtained from the Key Laboratory of Medicine and Clinical Immunology of Jiangsu Province, the First Affiliated Hospital of Soochow University. The 293T cells (provided by Professor Yong Zhao, Institute of Zoology, Chinese Academy of Sciences, Beijing, China) were maintained in our central laboratory. All the cells were cultured in RPMI-1640 medium supplemented with 10% fetal bovine serum (FBS), 100 U/ml penicillin, 100 µg/ml streptomycin and 2 mM glutamine. All the cells were cultured at 37°C in a humidified atmosphere containing 5% CO₂.

The miR-17-92 cluster overexpression vector was constructed on the MSCV vector containing green fluorescent protein (GFP; from Professor Yong Zhao, Institute of Zoology, Chinese Academy of Sciences, Beijing, China). The MSCV-GFP-miR-17-92 or MSCV-GFP control plasmid was transfected into the Phoenix A packaging cells (provided by Professor Yong Zhao, Institute of Zoology, Chinese Academy of Sciences) using FuGENE HD transfection reagent (cat. no. 04709705001; Roche, Shanghai, China). Viral supernatants were collected and used to infect the MGC-803 cells. For obtaining stably expressing cell lines, the cells were selected in the presence of 5 µg/ml puromycin (cat. no. J593; Amresco, Beijing, China). A shRNA-carrying sequence targeting the TRAF3 gene (506-524: 5′-GATAAGGTGTTTAAGGATA-3′) was designed and synthesized by Invitrogen (Beijing, China). The shRNA-TRAF3 was subcloned into the pSilencer3.1-H1-neo plasmid (cat no. 5770; Thermo Fisher Scientific™, Shanghai, China). The recombinant pSilencer3.1-siTRAF3 plasmid or the pSilencer3.1-H1-neo control plasmid were then trans-fected into the MGC-803 cells using Lipofectamine 2000 (cat no. 12566014; Thermo Scientific™). To obtain stably transfected cell lines, the cells were selected in the presence of 600 ng/µl neomycin (cat. no. A1720-5G; Sigma, Beijing, China).

Total RNA (2 µg) was reverse transcribed with SuperScript M-MLV (cat. no. 28025013; Promega, Shanghai, China) according to the manufacturer’s instructions. All RT-qPCR reactions were performed with a LightCycler 480 system (Roche) in triplicate. Primers were designed using Primer-BLAST (https://www.ncbi.nlm.nih.gov/tools/primer-blast/; NCBI; PubMed) and synthesized from Invitrogen (Beijing, China). β-actin was used as an internal control. cDNA was amplified using 2× LC480 SYBR-Green Master Mix (cat. no. 04887352001; Roche). Data were analyzed with the Pfaffl method (27,28). The primer sequences were as follows: BIM forward, 5′-ACAGAGCCACAAGACAGGAGCCC-3′ and reverse, 5′-CGCCgGCAACTCTTGGGCGAT-3′; cyclin D1 (CCND1) forward, 5′-CACTTTCAGTCCAATAGGTGTAG-3′ and reverse, 5′-TTCTTCTTGACTGGCACG-3′; PTEN forward, 5′-TGGGGAAGTAAGGACCAGAGA-3′ and reverse, 5′-TGAGGATTGCAAGTTCCGCC-3′; Von Hippel-Lindau (VHL) forward, 5′-GCAGGCGTCGAAGAGTACG-3′ and reverse, 5′-CGGACTGCGATTGCAGAAGA-3′; PH domain and leucine rich repeat protein phosphatase 2 (PHLPP2) forward, 5′-GTACGCAAGGGAAAGACCCA-3′ and reverse, 5′-AGCAAGGGAGTATTGCCGTC-3′; TRAF3 forward, 5′-TGTAAATACCGGGAAGCCACA-3′ and reverse, 5′-TGCACTCAACTCGCTCCTCA-3′; β-actin reverse, 5′-GCTACGAGCTGCCTGACGG-3′ and reverse, 5′-TGTTGGCGTACAGGTCTTTGC-3′.

Total RNA (2 µg) was polyadenylated with ATP by E. coli poly (A) polymerase (PAP; cat. no. M0276L; New England Biolabs, Ltd., Beijing, China). Following phenol-chloroform extraction and ethanol precipitation, the polyadenylated RNA was reverse transcribed with M-MLV and universal RT primer sequence. The cDNA (1:5 dilution) was then amplified using 2× LC480 SYBR-Green Master Mix (Roche). Each sample was determined in triplicate. U6 expression was considered as an internal control. Data were analyzed using the Pfaffl method (27,29). The primer sequences were as follows: hsa-miR-17 5′-GCAAAGTGCTTACAGTGCAGGTAG-3′; hsa-miR-18, 5′-CCGGTAAGGTGCATCTAGT-3′; hsa-miR-19a, 5′-TGTGCAAATCTATGCAAAACTG-3′; hsa-miR-19b, 5′-GCATCCCAGTGTGCAAATCC-3′; hsa-miR-20, 5′-GGTAAAGTGCTTATAGTGCAGGTAG-3′; hsa-miR-92. 5′-ATTGCACTTGTCCCGGCCTGT-3′; RT primer, 5′-AACGAGACGACGACAGACTTTTTTTTTTTTTTTV-3′; U6, 5′-GCAAATTCGTGAAGCGTTCCATA-3′.

Whole-cell extracts (WCE), cytoplasmic extracts (CE) and nuclear extracts (NE) were prepared using suspension buffer (10 mM Tris-HCl, 0.1 M NaCl, 1 mM EDTA) and proteinase inhibitor (cat. no. 11697498001; Roche Diagnostics, Mannheim, Germany) according to standard procedures. The total protein was measured with a micro BCA protein assay. Proteins (50 µg) were fractionated on an 10% SDS-PAGE gel and then transferred onto nitrocellulose membranes. After blocking with 5% milk in PBS for 1 h at room temperature, the membranes were incubated with primary antibodies (Abs). After washing, the membranes were incubated with secondary Abs. Finally, proteins were detected using the Odyssey system (LI-COR Biosciences, Lincoln, NE, USA). Abs (1:200 dilution) against RelA (sc-372X), RelB (sc-226X), c-Rel (sc-70), p105/p50 (sc-7178X), p100/p52 (sc-298), inhibitor of kappa B (IκB-α; sc-1643), p-IκB-α (Ser32/36, sc-101713), cyclin D1 (sc-20044) and Lamin A/C (sc-20681) were purchased from Santa Cruz Biotechnology (Santa Cruz, CA, USA). Abs (1:1,000 dilution) against AKT (#4691), p-AKT (Ser473, #4060), p-AKT (Thr308, #2965), p-c-Raf (#9421), p-glycogen synthase kinase (GSK)-3β (#5558), p-PTEN (#9551), phosphorylated phosphoinositide-dependent protein kinase-1 (p-PDK1; #3438), ERK1/2 (#4695), p-ERK1/2 (#4370S), BIM (#2933), PTEN (#9559), VHL (#2738), TRAF3 (#4729), E-cadherin (#3195), claudin1 (#13255), epithelial cellular adhesion molecule (EpCAM; #14452), CK8/18 (#4546), N-cadherin (#13116) and Vimentin (#5741) were obtained from Cell Signaling Technology. Abs (1:1,000 dilution) against α-tubulin (AJ1034a) and β-actin (AO1215a) were purchased from Abgent (Suzhou, China). Secondary Abs (1:10,000 dilution) against IRDye 680CW (926-32222) and IRDye 800CW (926-32210) were obtained from LI-COR Biosciences (Lincoln, NE, USA).

NE preparation and TransAM assay were performed as previously described (30). The activity of individual NF-κB subunits was detected by an ELISA-based NF-κB family transcription factor assay kit (cat. no. 43296; Active Motif, Carlsbad, CA, USA). Briefly, nuclear extracts (2 µg) were incubated in a 96-well plate, which were immobilized NF-κB consensus oligonucleotides. The captured complexes were incubated with specific NF-κB primary Abs and subsequently detected with HRP-conjugated secondary Abs (included with the kit). Finally, the optical density (OD) value at 450 nm was measured by spectrophotometry (ELx800; BioTek Instruments, Winooski, VT, USA).

The cells were harvested and seeded in a 96-well plate at a density of 5×10³ cells per well with 100 µl of complete culture medium, and incubated overnight at 37°C in a 5% CO₂ humidified incubator. Cell growth rate was assessed using a cell counting kit-8 (cat. no. CK04; Dojindo, Kumamoto, Japan) at different time points (0, 24, 48 or 72 h). Briefly, CCK-8 (10 µl) was added to each well. Following incubation for 2 h at 37°C, the absorbance at 450 nm was detected using a plate reader (ELx800; BioTek Instruments). Cell growth was measured by the relative absorbance with the absorbency of culture medium deducted.

The cells were harvested and seeded in a 96-well plate at a density of 5×10³ cells per well with 100 µl of complete culture medium, and incubated overnight at 37°C in a 5% CO₂ humidified incubator. Cell viability assay was performed using CellTiter-Glo (cat. no. G7570; Promega, Shanghai, China) according to the manufacturer’s instructions. Briefly, CellTiter-Glo (100 µl) was added to each well. Following incubation for 10 min at room temperature, the luciferase activities were detected using a plate reader (ELx800; BioTek Instruments) CellTiter-Glo measurements were taken at individual time points (0, 24, 48 or 72 h) to track cell proliferation.

The cells were cultured on cover slides for 24, 48 and 72 h at 37°C overnight. TUNEL assay was performed according to the manufacturer’s instructions of the TUNEL system kit (Roche). In brief, the slides were fixed by 4% para-formaldehyde for 30 min and blocked with 3% H₂O₂ methanol solution for 10 min. The slides were then treated with 1% Triton PBS solution for 2 min. Subsequently, 50 µl of TUNEL reaction solution were added to incubate the cells on slides for 1 h at 37°C. Finally, the TUNEL signals were converted using peroxidase (POD) for 30 min at 37°C, and the sections were treated with DAB for 3 min. The results were observed under alight system microscope IX71 (Olympus, Tokyo, Japan).

For the scratch healing assays, THE cells were treated with 10 mg/ml mitomycin C (Sigma, Beijing, China) for 3 h. Subsequently, the cells were wounded using a 200 µl sterile pipette tip, washed 3 times with PBS, and RPMI-1640 with 10% FBS was added. The wound closure was imaged continuously for 24, 48 and 72 h under a magnification of x10 using the light System Microscope IX71 (Olympus).

Transwell chambers (cat. no. 3422, 8 µm, 24-well insert; Corning, Lowell, MA, USA) were used for the migration and invasion assay. In brief, 600 µl of 10% FBS-containing medium was added to the lower chamber and 1×10⁵ cells in 200 µl serum-free medium were added to the upper chamber. The cells were incubated for 24 h at 37°C, and the non-invading cells were removed. Finally, the insert membranes were fixed with 4% paraformaldehyde, stained with 0.1% crystal violet (C8470; Solarbio, Beijing, China) for 15 min at room temperature, and photographed under an inverted microscope (SystemMicroscope IX71; Olympus, Tokyo, Japan). For the invasion assay, the insert membranes were coated with diluted Matrigel (cat. no. 354234; BD Biosciences, San Jose, CA, USA). Subsequently, 1×10⁵ cells were added to the upper chamber. At least 3 randomly selected fields were observed, and the average cell number was calculated.

For the gelatinase zymography assay, the activities of matrix metalloproteinase (MMP)-9 and -2 were examined. Culture medium was loaded on an 8% SDS-PAGE gel in the presence of 0.1% gelatin under non-reducing conditions. Culture medium samples were not denatured prior to electrophoresis. Following electrophoresis, the gels were washed twice in 2.5% Triton X-100 for 30 min. The gels were then incubated in substrate buffer (50 mM Tris-HCl and 10 mM CaCl₂, pH 8.0) at 37°C overnight, and stained with 0.5% Coomassie blue R250 (50% methanol and 10% glacial acetic acid) for 30 min, and then de-stained. Upon renaturation of the enzyme, the gelatinases digested the gelatin in the gel to produce clear bands against an intensely stained background.

The 3′-untranslated region (UTR) of TRAF3 mRNA was amplified by PCR from genomic DNA of the MGC-803 human gastric cancer cell line and ligated into the pmirGLO dual-luciferase miRNA target expression vector (#E1330; Promega, Madison, WI, USA). The primer sequences of the 3′-UTR of TRAF3 mRNA were as follows: 5′-CTGGACATGTCAGCATGTTAAGT-3′ and 5′-CGAGGCTCCGTTCAGAATTG-3′. The mutant constructs were generated using a Site-Directed Mutagenesis kit (#SDM-15, Beijing SBS Genetech Co., Ltd., Beijing, China). The 293T cells were seeded in 24-well plates at a density of 5×10⁴ cells per well with 500 µl of complete culture medium, and incubated overnight at 37°C in a 5% CO₂ humidified incubator. The pmirGLO-TRAF3-3′-UTR-wt plasmid or pmirGLO-TRAF3-3′-UTR-mut plasmid was co-transfected into the 293T cells with the miR-17-92 plasmid or control plasmid using Lipofectamine 2000 reagent (#11668-027; Invitrogen/Thermo Fisher Scientific, Waltham, MA, USA), respectively. Luciferase and Renilla activities were evaluated 48 h following transfection using the Dual-GLO^® Luciferase Assay System (#E2920; Promega, Madison, WI, USA) according to the manufacturer’s instructions.

Four-week-old male BALB/c-nude mice were purchased from Shanghai Slac Laboratory Animal Co. Ltd., Shanghai, China. The animals were randomly classified into 2 groups as follows: the MGC-803-control group and the MGC-803-miR-17-92 group, and each group had 5 mice. A total of 5×10⁶ cells were resuspended in 200 µl PBS and then injected subcutaneously into the right flanks of the nude mice. Tumor volume was measured every 4 days using a digital caliper according to the following formula: TV (mm³) = 0.5 × length × width². All the mice were sacrificed at 32 days after the injection. The tumors were excised, photographed, measured, weighted and fixed in 10% neutralized formalin overnight. The tumor tissues were dehydrated, fixed in paraffin, and cut into 5-µm-thick sections for histopathological examination and immunohistochemistry (IHC). The sections were stained for TRAF3 (dilution 1:50) expression using standard protocols. The sections were developed with 3,3-diaminobenzine (DAB) and counterstained with hematoxylin. All procedures and animal experiments were approved by the Animal Care and Use Committee of Soochow University.

The commercial human gastric cancer tissue microarray was purchased from Shanghai Outdo Biotech Company from the National Human Genetic Resources Sharing Service Platform (2005DKA21300). It consists of 100 paired gastric cancer and adjacent normal tissues, and was used to evaluate the protein expression of TRAF3. Patient follow-up information was obtained from 2007 to 2014. The information was obtained from the microarray company.

The IHC assay of tissue microarray was performed using a standard peroxidase-based staining method. Tissue sections (5-µm-thick) were incubated in a dry oven at 60°C for 30 min and then deparaffinized in xylene for 10 min in triplicate, rehydrated with graded ethanol in 100, 95, 90, 80 and 70% ethanol for 5 min each. Antigen retrieval was then performed using 0.01 M citrate buffer (pH 6.0). Subsequently, the slides were treated with 3% hydrogen peroxide (H₂O₂) to block endogenous peroxidase. The slides were then blocked with 5% bovine serum albumin (BSA; Boster Bioengineering, Wuhan, China), incubated with mouse anti-TRAF3 antibody (#4729; Cell Signaling Technology, 1:50 dilution) overnight at 4°C. Subsequently, the slides were incubated with diluted secondary antibody (#7076; Cell Signaling Technology, 1:500 dilution) for 45 min at 37°C. Finally, the slides were treated with DAB and counterstained with hematoxylin for microscopic examination. The results were observed by 2 independent investigators under an Olympus microscope BX 51 (Olympus) and scored as follows: 0, No staining; 1+, light; 2+, moderate; 3+, strong, according to the intensity of the staining. The percentage of positively stained cells was scored as follows: 0, No staining; 1, <25% staining; 2, 26-50% staining; 3, 51-75% staining; and 4, >75% staining. The product of the intensity and extent grades ≥4 of positive cells was considered high expression, and the score of 0-3 of positive cells was regarded as low expression.

All the experiments were repeated at least 3 times. All quantitative data are presented as the means ± SD and analyzed using the Student’s t-test. The association between TRAF3 expression and the clinicopathological factors was estimated using the Chi-square test. Overall survival (OS) curves were plotted on the basis of the Kaplan-Meier method and analyzed using the log-rank test. All statistical analyses were performed using GraphPad Prism version 5.0. A P-value ≤0.05 was considered to indicate a statistically significant difference, and P-values ≤0.01 and ≤0.001 were considered to indicate highly significant differences.

The expression of the miR-17-92 cluster, including miR-17, miR-18, miR-19a, miR-19b, miR-20 and miR-92, was detected by RT-qPCR analyses of human gastric cancer cell lines, including AGS, SGC-7901, BGC-823, MKN-45 and MGC-803. As shown in Fig. 1, all of the 6 members of the miR-17-92 cluster could be clearly detected in individual human gastric cancer cell lines, albeit with different levels. The expression levels of the individual miR-17-92 members in the AGS cells were comparable to those in the SGC-7901 and BGC-823 cells. However, the expression levels of the miR-17-92 cluster in the MKN-45 and MGC-803 cells were significantly lower than those in the AGS, SGC-7901 and BGC-823 cells. The expression levels of miR-17 (P=0.0003), miR-18 (P=0.0041), miR-19a (P=0.0114), miR-19b (P=0.0008), miR-20 (P=0.0086) and miR-92 (P=0.0348) were significantly decreased approximately 2- to 4-fold in the MGC-803 cells, as compared with those in the AGS cells. It was indicated that the miR-17-92 expression levels in the MGC-803 cells were the lowest among the 5 human gastric cancer cell lines. Thus, the MGC-803 cell line was selected to perform the following experiments to explore the function of the miR-17-92 cluster in gastric cancer progression.

To explore the effects of the miR-17-92 cluster on the biological behaviors of the MGC-803 cells, the constructed miR-17-92 overexpression plasmid was transfected into the MGC-803 cells. The positive monoclones were selected upon exposure to puromycin (5 ng/µl) for 2 weeks. The formed monoclones were further verified for miR-17-92 expression by RT-qPCR analyses. The expression levels of each miR-17-92 family member were significantly increased in the MGC-803-miR-17-92 cells compared to those in the MGC-803-control cells, 4-fold for miR-17 (P=0.0109), 3-fold for miR-18 (P=0.0415), 6-fold for miR-19a (P=0.0111), 5-fold for miR-19b (P=0.0104), 4-fold for miR-20 (P=0.0058) and 2-fold for miR-92 (P=0.0169), indicating a successful introduction of the miR-17-92 cluster into the MGC-803 cells (Fig. 2A). Moreover, GFP was used as a marker for the transfection of the plasmids. As shown in Fig. 2B, both the established MGC-803-miR-17-92 and the MGC-803-control cells presented strong GFP signals observed under a fluorescence microscope (Fig. 2B). The GFP signals of the MGC-803-miR-17-92 and the MGC-803-control cells were measured by flow cytometry. The percentages of GFP-positive cells in the 2 established cell lines were 95.71 and 96.88%, respectively (data not shown), indicating the successful transfection of the plasmids. Moreover, the expression levels of certain well-known target genes of the miR-17-92 cluster, including BIM, PTEN, CCND1, PHLPP2 and VHL were significantly decreased in the MGC-803-miR-17-92 cells compared to those in the MGC-803-control cells (Fig. 2C). Consistently, the expression levels of BIM, PTEN, cyclin D1, PHLPP2 and VHL at the protein level were decreased, as detected by western blot analysis (Fig. 2D). Taken together, it was indicated that the MGC-803 gastric cancer cell line stably overexpressing the miR-17-92 cluster was successfully established.

To determine whether the overexpression of the miR-17-92 cluster modulates the proliferative activity of the MGC-803 cells, a CCK-8 assay was performed. As shown in Fig. 3A, the OD values at 450 nm in the miR-17-92 overexpression group (0.76±0.10 and 1.26±0.11) were significantly higher than those in the control group (0.37±0.06 and 0.61±0.08) at 48 and 72 h, respectively (both P<0.01). To further characterize the effects of miR-17-92 on the viability of the MGC-803 cells, a CellTiter-Glo cell viability assay was carried out. It was shown that the luciferase activities of the MGC-803-control cells were 37,098±6,833, 45,846±7,009 and 63,883±8,471 at 24, 48 and 72 h, respectively; while those in the MGC-803-miR-17-92 cells were 54,545±8,151, 86,085±5,774 and 102,027±11,564 (^**P<0.01, ^***P<0.001, Fig. 3B). To investigate whether the overexpression of miR-17-92 affected the apoptotic capability of the MGC-803 cells, a TUNEL assay was performed. As shown in Fig. 3C, both the MGC-803-control and MGC-803-miR-17-92 cells underwent apoptosis in a time-dependent manner. The apoptotic rates of the MGC-803-control cells were 14.67±1.28, 18.28±1.75 and 20.17±1.76%, while those of the MGC-803-miR-17-92 cells were 5.33±1.28, 7.11±1.07 and 7.97±1.25% at 24, 48 and 72 h, respectively (all P<0.01). Moreover, the cell cycle was analyzed by flow cytometry, and the results revealed no significant differences between the MGC-803-control and MGC-803-miR-17-92 cells (data not shown).

To further investigate whether miR-17-92 overexpression can promote cell growth in vivo, BALB/c-nude mice were injected subcutaneously into the right flanks with the MGC-803-control or MGC-803-miR-17-92 cells to establish a heterotopic xenograft tumor mouse model. A representative image of tumors from both the control and the miR-17-92 overexpression group is shown in Fig. 3D. The mean tumor volume in the miR-17-92 overexpression group reached >800 mm3, but was <400 mm³ in the control group at 32 days following implantation (Fig. 3E). The mean tumor weight in the miR-17-92 overexpression group (0.94±0.16 g) was much greater than that in the control group (0.26±0.08 g, P<0.001, Fig. 3F). Taken together, these results indicated that miR-17-92 overexpression in the MGC-803 cells increased the prolifera-tive activity and decreased the apoptosis of the gastric cancer cells in vitro, and promoted tumor xenograft growth in vivo.

As shown in Fig. 4A, the protein expression of AKT signaling molecules, including AKT, p-AKT-473, p-AKT-308, p-c-Raf, p-GSK-3β, p-PTEN and p-PDK1, were measured by western blot analyses. Total AKT expression was similar between the MGC-803-control and MGC-803-miR-17-92 cells. However, miR-17-92 over-expression induced the phosphorylation of AKT at Thr308, but not at Ser473 (Fig. 4A), which was due to the inhibition of PTEN expression (Fig. 2D). No significant changes were observed in the other AKT signaling molecules (Fig. 4A). As shown in Fig. 4B, the protein expression level of total ERK1/2 in the MGC-803-miR-17-92 cells was comparable to that in the MGC-803-control cells. However, the overexpression of the miR-17-92 cluster in the MGC-803 cells induced the phosphorylation of ERK1/2 (p-ERK1/2). Thus, miR-17-92 overexpression in the MGC-803 cells induced p-ERK1/2 expression without affecting total ERK1/2 expression.

Subsequently, to investigate whether miR-17-92 over-expression affects the activity of NF-κB signaling, western blot analysis was performed. As shown in Fig. 4C, the expression of RelA/p65 and p50, representing canonical NF-κB activity, was clearly increased in both the CE and NE fractions in the MGC-803-miR-17-92 cells compared to those in the MGC-803-control cells. The phosphorylation of IκB-α at Ser32/36 was induced by miR-17-92 overexpression in the MGC-803 cells (Fig. 4D), leading to the subsequent degradation of IκB-α. The expression of IκB-α, an important upstream regulator of the canonical NF-κB signaling, was markedly decreased in the MGC-803-miR-17-92 cells. Moreover, the expression of RelB and p52, representing non-canonical NF-κB activity, was slightly increased in the NE of the MGC-803-miR-17-92 cells. However, the expression of c-Rel in the MGC-803-miR-17-92 cells was similar to that in the MGC-803-control cells (Fig. 4C). To determine the exact contribution of the miR-17-92 gene cluster to the NF-κB DNA-binding capabilities in the MGC-803 cells, an ELISA-based NF-κB activity assay was performed. As shown in Fig. 4E, the DNA-binding capability of each NF-κB subunit including RelA, p50, RelB and p52 in the NE of the MGC-803-miR-17-92 cells was significantly increased as compared with that in the MGC-803-control cells. However, the DNA-binding capability of c-Rel was comparable between the 2 cell lines. Taken together, these results indicated that the introduction of the miR-17-92 cluster into the MGC-803 cells activated the AKT, ERK and NF-κB signaling pathways, likely contributing to the increased cellular proliferation.

As shown in Fig. 5A, the number of migrated cells in the miR-17-92 overexpression group was increased approximately 3-fold as compared with that of the control group (P<0.01). Cell motility was also evaluated by scratch wound healing assay. As shown in Fig. 5B, a scratched cell monolayer was formed and images were captured at 24, 48 and 72 h. It was shown that the MGC-803-miR-17-92 cells migrated from the edge towards the scratch center more rapidly than the MGC-803-control cells, indicating an enhanced migratory ability. Transwell invasion assay was performed to further examine the role of miR-17-92 in the invasive ability of the MGC-803 cells. As shown in Fig. 5C, the number of cells that invaded the Matrigel layer from the MGC-803-miR-17-92 group was increased approximately 4-fold compared with that of the MGC-803-control group (P<0.01). The results of the gelatin zymography assay revealed that MMP-9 activity was not detected in either of the established cell lines; however, MMP-2 activity was markedly increased in the MGC-803-miR-17-92 cells (Fig. 5D). The expression levels of key molecules involved in EMT, including E-cadherin, claudin1, EpCAM, CK8/18, N-cadherin and Vimentin, were examined by western blot analysis. The expression levels of claudin1 and EpCAM, representing epithelial markers, were decreased in the MGC-803-miR-17-92 cells compared to the MGC-803-control cells, while the expression levels of other epithelial markers, such as E-cadherin and CK8/18 were unaffected. The expression levels of N-cadherin and Vimentin, mesenchymal markers, were increased in the MGC-803-miR-17-92 cells compared with the MGC-803-control cells (Fig. 5E). Thus, EMT occurred when the miR-17-92 cluster was overexpressed in the MGC-803 cells. Collectively, these results indicated that the enforced introduction of the miR-17-92 cluster into the MGC-803 cells enhanced the migratory and invasive abilities of the cells, which was owing to the occurrence of EMT.

Of note, we focused on the TRAF3 gene, as it was found to be one of the predicted target genes of the miR-17-92 cluster using TargetScan Release 5.1 online software (http://www.targetscan.org/, Whitehead Institute for Biomedical Research, Cambridge, MA, USA) (data not shown). The protein expression of TRAF3 was significantly decreased in the MGC-803 cells overexpressing the miR-17-92 cluster, which was in line with the decreased mRNA level detected by RT-qPCR analysis (Fig. 6A and B). The deregulated NF-κB family members were observed in the cells overexpressing the miR-17-92 cluster. Compared with the control group, both the mRNA and protein expression levels of TRAF3 were decreased in the xenograft tumors in the miR-17-92 overex-pression group (Fig. 6C and D), which was consistent with the in vitro results.

To verify whether TRAF3 is truly a direct target of the miR-17-92 cluster, a luciferase assay was performed. The pmirGLO-TRAF3-3′-UTR-wild-type or pmirGLO-TRAF3-3′-UTR-mutant vector were co-transfected into 293T cells with the control or miR-17-92 plasmid. As shown in Fig. 6E, the atopic expression of the miR-17-92 cluster significantly reduced the relative luciferase activity of the wild-type 3′-UTR of TRAF3 (P<0.001), but not that of the mutant 3′-UTR of TRAF3. Taken together, it was confirmed that TRAF3 was a direct target of miR-17-92, which could downregulate TRAF3 expression by directly binding to the 3′-UTR of TRAF3.

We further investigated whether TRAF3 functions as a target gene of the miR-17-92 cluster in the MGC-803 cells. The MGC-803 cells were transfected with shRNA-TRAF3 or shRNA-control plasmid, respectively. As shown in Fig. 7A, both the mRNA and protein levels of TRAF3 were markedly lower in the MGC-803-siTRAF3 cells as compared with the MGC-803-sictrl cells, indicating a successful RNA interference (RNAi) of the TRAF3 gene. The expression levels of RelA and p50 were clearly increased in the NE of the MGC-803-siTRAF3 cells compared to that in the MGC-803-sictrl cells. Moreover, the expression levels of RelB and p52 were upregulated in both the CE and NE in the MGC-803-siTRAF3 cells compared to that in the control cells (Fig. 7B). To further determine the exact contribution of TRAF3 silencing to the NF-κB DNA-binding capabilities in the MGC-803 cells, an ELISA-based NF-κB activity assay was performed. In line with the results of western blot analysis, the average DNA-binding capability of RelA, p50, RelB and p52 in the NE of the MGC-803-siTRAF3 cells were significantly increased compared to that in the MGC-803-sictrl cells, while the DNA-binding capability of c-Rel remained unaltered (Fig. 7C). Therefore, TRAF3 silencing significantly activated not only canonical NF-κB activity, but also non-canonical NF-κB activity in the MGC-803 cells.

As shown by CCK-8 assay, the OD values at 450 nm in the MGC-803-siTRAF3 group (0.72±0.10, 1.71±0.20 and 2.14±0.17) were significantly higher than those in the MGC-803-sictrl group (0.64±0.13, 1.39±0.16 and 1.63±0.15) at 24, 48 and 72 h, respectively (all P<0.01, Fig. 7D). Moreover, a CellTiter-Glo cell viability assay was carried out. The relative luciferase activities of the MGC-803-sictrl cells were 110,886±17,428, 165,340±28,222 and 183,612±16,171 at 24, 48 and 72 h, respectively; while those in the MGC-803-siTRAF3 cells were 150,425±30,618, 224,387±31,175 and 241,320±24,341 at 24, 48 and 72 h, respectively (all P<0.01, Fig. 7E). Transwell migration and invasion assays were also performed. TRAF3 silencing markedly increased the number of migrated cells approximately 3-fold in the MGC-803-siTRAF3 group as compared with the control group (P<0.01, Fig. 7F). Moreover, the number of cells that invaded the Matrigel layer from the MGC-803-siTRAF3 group was increased approximately ut 3-fold compared with that of the corresponding control group (P<0.01, Fig. 7G). Taken together, these results indicated that TRAF3 silencing promoted the proliferation, and enhanced the migratory and invasive abilities of the MGC-803 cells in vitro.

The protein expression levels of TRAF3 in human gastric cancer tissues and adjacent normal tissues were detected by IHC analyses. As shown in Fig. 8A, the expression of TRAF3 could be clearly detected in both the nuclear and the cytoplasmic portions of the adjacent normal tissues. By contrast, the expression of TRAF3 was predominantly detected in the cytoplasm of the gastric tumor tissues. Moreover, the expression of TRAF3 in the gastric tumor tissues was significantly lower than that in the adjacent normal gastric mucosa, and there was a statistically significance (P=0.007) between the gastric tumor tissues and the adjacent normal tissues (Fig. 8B). As shown in Table I, the association between TRAF3 expression and the clinicopathological characteristics in the 100 paired gastric cancer patients was evaluated. The expression of TRAF3 was negatively associated with tumor diameter and the TNM stage of patients with gastric cancer (P<0.05). The expression of TRAF3 was not associated with the other clinicopathological characteristics examined, including age, sex, site of tumor, histological types and tumor differentiation (P>0.05).

OS curves were also plotted according to TRAF3 expression using the Kaplan-Meier method. Survival analysis indicated that the low TRAF3 expression group had a significantly shorter OS rate than that of the high TRAF3 expression group, and there was a statistically significance (P=0.002) between the low TRAF3 expression group and the high TRAF3 expression group (Fig. 8C). Taken together, these results indicated that TRAF3 expression was downregulated in gastric cancer tissues and that TRAF3 functions as an important prognosis factor in patients with gastric cancer.