Gastric cancer cell lines AGS and NCI-N87 were purchased from American Type Culture Collection (ATCC, Manassas, VA, USA) and cultured in RPMI-1640 GlutaMAX (Gibco, Thermo-Fisher Scientific, Waltham, MA, USA) with 10% fetal bovine serum and penicillin-streptomycin (Sigma-Aldrich, St. Louis, MO, USA) at the concentration of 1% (v/v). All cells were kept in a hypoxic chamber (Invivo200, UK) in a humidified atmosphere with 5% CO₂, 1% O₂, and 94% N₂ and a constant temperature of 37˚C.

For total protein extraction from the tumor samples, the tissues were crushed on dry ice to protect from denaturation. Lysis buffer (2.5 μL/mg) was added to the sample and then the tissues were homogenized on ice and sonicated, before incubation for 30 minutes on ice. Protein extraction from the cancer cell lines was accomplished by adding lysis buffer and incubating on ice for 30 minutes following washing of the cells with phosphate-buffered saline (PBS). Protein samples from both the tissues and cells were then centrifuged at 12,000 rpm for 10 minutes at 4°C. The supernatant was subsequently collected and stored at −80°C. Later, the protein concentrations were measured using a bicinchoninic acid kit (AR0146; Boster Biological Technology Co. Ltd., Wuhan, China). Equal amounts of each protein sample were electrophoresed on a sodium dodecyl sulfate polyacrylamide gel. Then the proteins were transferred onto a polyvinylidene fluoride membrane and the membranes were blocked with 5% skim non-fat milk solution in tris-buffered saline with Tween-20. The primary and secondary antibodies used were as follows: anti-HIF-1α antibody (1:1,000; ab51608, Abcam, Cambridge, MA, USA), anti-Bcl3 antibody (4 µg/mL; ab27780, Abcam), anti-Atg5 antibody (1:10,000; ab109490, Abcam), anti-LC3 A/B antibody (2 µg/mL; ab128025, Abcam), anti-vascular endothelial growth factor (VEGF) antibody, (1:5,000; ab32152, Abcam), anti-matrix metallopeptidase 9 (MMP-9) antibody, (1:1,000; ab38898, Abcam), anti-glyceraldehyde 3-phosphate dehydrogenase antibody (1:1,000; ab181602, Abcam), horseradish peroxidase-conjugated goad anti-rabbit (1:5,000; ab6721, Abcam), anti-mouse (1:5,000; ab6728, Abcam), and donkey anti-goat (1:5,000; ab6885, Abcam) antibodies. Protein bands were eventually visualized using an ECL kit (RPN2232, Amersham, Little Chalfont, UK). Quantification was conducted with the Image J software.

AGS and NCI-N87 was seeded in six-well plates. The confluent cell monolayer was scraped with a pipette tip in the middle of the well. After 24 hours incubation, the cell migration was captured with a DM2500 bright field microscope (LEICA, Wetzlar, Germany) and the migration distance was measured using the ImageJ software.

Cells (2×10³/well) were seeded in six-well plates with 6 parallel wells set for each experimental group. Briefly, 6 hours post cell attachment was recorded as the 0 hour and the cells were cultured for 30 hours. Then 100 μL mixture of medium and Cell Counting Kit-8 reagent (9:1) was added, and the cells were incubated in the incubator with 5% CO_2, 1% O₂ and 94% N₂ at 37°C. A microplate reader was used to read the absorbance value at 450 nm.

Immunofluorescence staining was used to confirm the expression of LC3 in the cells. Cells were cultured in different conditions for 48 hours. Then, the cells were washed with PBS and fixed for 20 minutes at 25°C with 4% paraformaldehyde. The cells were permeabilized with 0.4% TritonX-100 for 10 minutes, washed with PBS for thrice and then blocked with 2% bovine serum albumin in PBS for 1 hour at 37°C. Antibodies against LC3 (in 1% blocking solution) was added to the cells and incubated overnight at 4°C. After 5-min washes with PBS, fluorescein-conjugated secondary antibody was added (in 1% blocking solution) and the cells were incubated for 1 hour. Finally, the stained cells were mounted with DAPI (1 μg/mL for 7–10 minutes) to stain the nuclei. Following 3 additional 10 minutes washes, the samples were examined and analyzed with a fluorescence microscope (Olympus, Tokyo, Japan)

BALB/c nude mice purchased from the Beijing Vital River Laboratory Animal Technology Co., Ltd. (Charles River Laboratories, Wilmington, MA, USA) were used for the tumorigenesis experiments. Female BALB/c (4 weeks old and weighing about 15 g) mice were used. Fifteen mice were separated into 5 groups. A total of 5×10⁶ AGS cells (transfected with si-RNA-Bcl-3 [si-Bcl-3] or not) resuspended in 200 μL PBS were injected subcutaneously into nude mice. Intraperitoneal injection of 5-fluorouracil (5-FU, 50mg/kg daily for 7 days)/rapamycin (Rapa) were given to mice. After a 30-day observation period, the experimental mice were euthanized and the tumors were dissected for further experiments. All the animal experiments procedures in this study were approved by the Animal Ethics Committee of Committee of the Beijing Friendship Hospital (BFH-20180306) . All the animal protocols were conducted strictly in accordance with the guidelines of International Guiding Principles for Animal Research.

The paraffin-embedded tumor sections were incubated with anti-Ki67 (Abcam, Cambridge, UK). Immunohistochemistry was performed as previously described [9].

TUNEL staining was performed to detect apoptotic cell death in situ in the tumors according to the manufacturer's instructions (TUNEL apoptosis detection kit: Upstate, Lake Placid, NY, USA). Relative TUNEL-positive rate is displayed as the ratio of brown staining within the nucleus of the apoptotic cells.

In this study, data are expressed as mean±standard deviation, and paired t-test was performed for comparison with the control groups. All statistical calculations were performed using the SPSS10.0 (SPSS Inc., Chicago, IL, USA) software. A significant difference was determined by P-value <0.05 and P-value <0.01.

HYP is a key feature in the microenvironment of solid tumors. Elevated level of HIF-1α is often found in solid tumors, which correlates with cancer-related death [210]. Therefore, we subjected the gastric cancer cell lines AGS and NCI-N87 to hypoxic conditions, and an in vitro HYP cell model was established to mimic the microenvironment in solid tumors. Cells were grouped into control, HYP 6 hours, HYP 12 hours, HYP 24 hours according to the HYP treatment duration. The HIF-1α and Bcl-3 protein levels were determined at different time points by western blot analysis. The results showed that HIF-1α and Bcl-3 protein levels gradually increased with increasing exposure to HYP (Fig. 1A and B), and differences between the groups were significant (P<0.05). The results indicate that expression of Bcl-3 is induced by HYP. The increase of Bcl-3 may be positively correlated with gastric cancer development.

To understand the mechanism of Bcl-3 in gastric cancer cells, si-Bcl-3 vectors were constructed to knockdown the expression of Bcl-3. As shown in Fig. 2A, the protein level of Bcl-3 in the 3 si-Bcl-3 (si-Bcl-3-1, si-Bcl-3-2, and si-Bcl-3-3) groups were significantly decreased compared to the control group (P<0.05), while the si-negative control (si-NC) group showed no obvious change compared to control group. These results confirm the knockdown efficacy of three si-Bcl-3 small interfering RNAs.

Cells were subsequently divided into 4 groups as follows: control, HYP, si-Bcl-3+HYP, and si-NC+HYP. Cell migration in the four groups was evaluated using wound healing assay. Cells in the HYP group showed increased migration rate compared to the control group (Fig. 2B, P<0.05), while cells in the si-Bcl-3+HYP group showed decreased migration rate compared to the HYP group (P<0.05). In addition, si-Bcl-3 significantly increased the chemosensitivity of gastric cancer cells to oxaliplatin (OXA), 5-FU, and irinotecan (IRI) compared to their treatment alone (Fig. 2C-E, P<0.05). These results suggest that Bcl-3 knockdown reduced the migration and increased the chemosensitivity of gastric cancer cells.

To assess the effect of si-Bcl-3 on autophagy, autophagy-related proteins were analyzed following Bcl-3 knockdown. As shown in Fig. 3A and B, the protein level of HIF-1α was significantly increased following HYP, while the increase was inhibited by si-Bcl-3 in gastric cancer cells (P<0.05). As shown in Fig. 3A, C, and D, the protein level of Atg5 and LC3II were also induced by HYP and suppressed by si-Bcl-3. Interestingly, the protein expression level of Bcl-3 showed similar expression pattern as Atg5/LC3II (Fig. 3A and E). In addition, the LC3 immunofluorescence assay further confirmed that si-Bcl-3 inhibited LC3 protein level induced by HYP (Fig. 3F and G, P<0.05). These results indicate that Bcl-3 knockdown decreases HYP-induced protective autophagy in gastric cancer cells.

To further emphasize that Bcl-3 knockdown suppresses migration and 5-FU resistance were mediated by autophagy, autophagy inducer Rapa was introduced. As shown in Fig. 4A, under hypoxic conditions, si-Bcl-3 clearly inhibited cell migration, while treatment with Rapa significantly increased cell migration. As expected, si-Bcl-3 reduced Rapa-induced cell migration. Moreover, under the hypoxic conditions, si-Bcl-3 significantly decreased resistance to 5-FU (Fig. 4B, P<0.05). In addition, si-Bcl-3 reduced the expression of Atg5 and LC3II. Further, si-Bcl-3 inhibited Rapa-induced Atg5 and LC3II expression (Fig. 4C-E, P<0.05). Taken together, these results suggest that under hypoxic conditions Bcl-3 knockdown reduces migration and 5-FU resistance of gastric cancer cells by decreasing autophagy.

To further clarify the role of Bcl-3 on migration and chemosensitivity of gastric cancer cells in vivo, a xenograft tumor model was constructed in nude mice. As depicted in Fig. 5A, compared to the control group, the tumor sizes in 5-FU+si-Bcl-3, 5-FU and 5-FU+si-Bcl-3-NC groups were significantly decreased (P<0.05), while those in 5-FU+si-Bcl-3+Rapa showed minimal difference compared to the control group. Interestingly, compared to the 5-FU group, 5-FU+si-Bcl-3 showed reduced tumor sizes (P<0.05). As shown in Fig. 5B-H, compared to the control group, the relative expression of Bcl-3, HIF-1α, VEGF, MMP-9, Atg5, and LC3II were clearly downregulated in the 5-FU+si-Bcl-3, 5-FU, and 5-FU+si-Bcl-3-NC group (P<0.05), while 5-FU+si-Bcl-3+Rapa group showed minimal difference compared to the control group. Moreover, compared with 5-FU group, the 5-FU+si-Bcl-3 group showed reduced expression of these proteins (P<0.05). In addition, compared to the control group, Ki67 expression was downregulated in 5-FU+si-Bcl-3, 5-FU, and 5-FU+si-Bcl-3-NC groups (Fig. 5I, P<0.05), but increased following Rapa treatment. Importantly, the apoptotic rates were increased in the 5-FU+si-Bcl-3, 5-FU, and 5-FU+si-Bcl-3-NC groups compared to the control group (Fig. 5J, P<0.05), but increased following Rapa treatment. These results indicate that knockdown of Bcl-3 suppresses the in vivo growth of gastric cancer through inhibiting autophagy.