From January 2003 to December 2005, human gastric cancer samples were obtained from gastric cancer patients who underwent gastric surgery at the Department of Surgery, Taipei Veterans General Hospital. A total of 91 gastric cancer patients (including 59 males and 32 females, age from 33–84 years) were enrolled in this study and informed consent was obtained from all patients. Gastric tumors and adjacent gastric tissues were fixed overnight at 4°C with 4% neutral buffered paraformaldehyde, dehydrated, cleared with Histo-Clear II (National Diagnostics, Atlanta, GA, USA) and wax embedded. Five-micron sections were used for hematoxylin and eosin (H&E) staining and immunohistochemical staining. This study was approved by the Institutional Review Board of Tapei Veterans General Hospital and informed consent was provided by all subjects.

The localization of c-Met receptor and Jagged1 ligand in gastric tissues was detected using an avidin-biotin-peroxidase complex (ABC) technique according to a protocol established and used for gastric cancer specimens (2,3). Immunostaining of HGF/c-Met receptors in gastric cancer tissue was performed. Briefly, tissue sections were first microwaved in a sodium citrate buffer (10 mM, pH 6.0) and, were then treated with peroxidase blocking solution (3% H₂O₂) for 5–10 min. Serum blocking solution (Histostain SP kit; Zymed Laboratories Inc., South San Francisco, CA, USA) was applied to remove endogenous peroxidase activity and to reduce nonspecific background staining. The antibodies used for tissue sections were as follows: polyclonal rabbit anti-c-Met (C-28) (1:200; cat. no. sc-161; Santa Cruz Biotechnology Inc., Dallas, TX, USA) at a 1:500 dilution and polyclonal rabbit anti-human Jagged1 antibody (cat. no. sc8303; Santa Cruz Biotechnology) at a 1:50 dilution. Slides were incubated overnight at 4°C in a moist chamber. The tissue sections were subsequently treated with post-primary block reagent: NovoLink polymer detection system (cat. no. RE7140-K; Leica Biosystems, Wetzlar, Germany) for 10 min, streptavidin peroxidase-conjugated for 10 min, AEC substrate chromogen for 5–10 min, and then counterstained with hematoxylin for 5 min. Pre-immune rabbit IgG was used as a negative control. The distribution pattern of chromogen in tumor cells was examined by an experienced pathologist (A.F.-Y. Li).

The distribution of c-Met and Jagged1 proteins in human gastric cancer specimens was evaluated by a semi-quantitative system to calculate and estimate within the following arbitrary ranges: (−) no positive cells; (+) 1–25%; (++) 26–75%; and (+++) >75%.

Our previous study tested the Notch1 expression in SC-M1, AGS, NUGC3 and KATO III gastric cancer cell lines. The expression level of N1IC in SC-M1 gastric cancer cells was much more abundant than that in the other gastric cancer cell lines (3). Therefore, we chose SC-M1 cells for further experimentation. The overexpression of N1IC increased Notch activity in SC-M1 cells. Human gastric cancer SC-M1 cells were cultured in RPMI-1640 medium (Gibco; Thermo Fisher Scientific, Inc., Waltham, MA, USA) with 10% fetal bovine serum (FBS; HyClone Laboratory; GE Healthcare Life Sciences, Logan, UT, USA) and 0.1% gentamicin (Gibco; Thermo Fisher Scientific, Inc.). Our previous study showed that SC-M1 gastric cancer cell lines also express other Notch family members (Notch 1, 2, 3 and 4). N1IC expression was found to be significantly higher than the expression of other Notch family members (13,14). Therefore, we choose N1IC for further experimentation. To establish stable expression of an HA-N1IC fusion protein in SC-M1 cells (SC-M1/HA-N1IC), a pcDNA-HA-N1IC expression plasmid was transfected into SC-M1 cells, and expression levels were screened by western blot analysis according to a previous established protocol (3). For the control, the pcDNA3 vector was also transfected into the SC-M1 cell line. The SC-M1/HA-N1IC and SC-M1/pcDNA3 cell lines were selected with G418 (400 µg/ml; CalbioChem; EMD/Merck KGaA).

Human gastric cancer SC-M1, SC-M1/HA-N1IC and SC-M1/pcDNA3 cells were treated with HGF alone (0–25 ng/ml; Gibco; Thermo Fisher Scientific, Inc.), NS398 alone (50 µM; Cayman Chemical Company, Ann Arbor, MI, USA), or HGF in combination with NS398. HGF promotes gastric cancer cell growth, and NS398 is a COX-2-specific inhibitor that has anti-inflammatory and analgesic effects. To assess the effects on cell growth, the gastric cancer cell lines were treated with HGF and NS398 alone, or combined HGF and NS398 for 24, 48 or 72 h. To assess the effects on migration, cells were treated with HGF (25 ng/ml) for 64 h, and cell migration was observed at 8, 32 and 64 h. For western blot analysis, the gastric cancer cell lines were incubated with HGF (25 ng/ml), and protein samples were collected at 1, 2, 4 and 24 h to analyze c-Met, Jagged1 and COX-2 protein expression.

SC-M1 cells were seeded into a 96-well plate at a density of 1.2×10⁴ cells/well and incubated for 24 h. After treatment with HGF and/or NS398 for 24, 48, or 72 h, the medium was replaced with 100 µl of medium with 3-[4,5-dimethylthiazol-2-yl]-2,5-diphenyl-tetrazolium bromide (MTT). Then, the plate was incubated in a 37°C incubator. After discarding the medium, 100 µl dimethyl sulfoxide (DMSO) was added. The optical density was measured at an absorption wavelength of 570 nm by an ELISA reader.

The migration ability of the drug- or vehicle-treated gastric cancer cells was evaluated by wound healing migration analysis. Culture-inserts (Ibidi-cells in focus, Munich, Germany) were placed in 6-well dishes, and 8×10⁴ SC-M1 cells were seeded onto the inserts and incubated for 24 h. Inserts were removed and gently washed to remove floating cells. After addition of 0.5% FBS medium with or without HGF (25 ng/ml), images of cell migration were captured at 8, 32 and 64 h. The percentage of the migration area was compared to the initial area after removing the insert. The migration area of gastric cancer cells was measured by ImageJ version 1.46 software (National Institutes of Health, Bethesda, MD, USA).

Cell migration assays were analyzed using 24-well Transwell inserts with 5-µm pores (Corning Costar, Cambridge, MA, USA). Approximately 5×10⁴ cells were resuspended in 100 µl of low serum (0.1% FBS)-containing medium and loaded into the inserts with low serum-containing medium in the lower chamber. After incubating for 12 h, either DMSO or HGF in low serum-containing medium at indicated concentrations was added to the inserts, which were then immediately shifted to new wells with high serum (20% FBS)-containing medium to allow migration in a humidified 37°C incubator with 5% CO₂ for 24 h. SC-M1, SC-M1/HA-N1IC and SC-M1/pcDNA3 cells were then treated with HGF and/or NS398. After incubation for 24 h, the inserts were removed, fixed in ice-cold methanol for 20 min, and then stained using Liu staining solutions (Tonyar Biotech, Inc., Taoyuan, Taiwan). After staining, the cells that remained on the upper sides of the inserts were carefully wiped-off using rinsed cotton swabs. The migrated cells found on the bottom of the inserts were then imaged (at ×100 magnification) and counted. Migrated cells from 10 random fields of each inserts were counted by ImageJ version 1.46 software (National Institutes of Health) and normalized.

SC-M1 cells were treated with or without HGF (25 ng/ml) for 1, 2 or 4 h. The cells were mixed with NP-40 lysis buffer, and then added with mammalian protein extraction reagents (M-PER) in an ice bath. Cell lysates were collected after centrifugation at 23,220 × g for 10 min at 4°C, and cell lysates were heated at 90°C for 5 min. Protein concentrations were quantitatively determined using Bradford method. Protein (30 µg) samples were electrophoresed on 6% (for c-Met and Jagged1) or 7.5% (for COX-2) SDS-polyacrylamide gels. Each gel was then transferred to a polyvinylidene fluoride (PVDF) membrane (Immobilon-P; EMD Millipore, Bedford, MA, USA). The PVDF membrane was blocked with 5% non-fat milk for 30 min and probed with rabbit anti-c-Met antibody (C-28) (cat. no. sc-161; Santa Cruz Biotechnology) at a 1:250 dilution, rabbit anti-Jagged1 antibody (cat. no. sc8303; Santa Cruz Biotechnology) at a 1:500 dilution, or mouse anti-COX-2 antibody (cat. no. 160112; Cayman Chemical Company) overnight at 4°C. After washing with Tris-buffered saline containing 0.1% Tween-20 (TBS-T), the membranes were incubated with secondary anti-rabbit antibody (cat. no. 7074; Cell Signaling Technology, Danvers, MA, USA) at a 1:2,000 dilution. Enhanced protein bands were detected by a chemiluminescence kit (Amersham Life Science, Piscataway, NJ, USA).

All statistical analyses were carried out using SPSS software version 18.0 (SPSS, Inc., Chicago, IL, USA). Clinicopathological differences were compared with Chi-square tests, and the correlation between Jagged1 and c-Met expression in each specimen was estimated with linear regression modeling. A P-value <0.05 was considered to indicate a statistically significant result.

To evaluate the correlation between c-Met and Jagged1 in gastric cancer specimens, we collected samples from 91 gastric cancer patients in our study. The expression of c-Met and Jagged1 in gastric cancer tissue was analyzed by immunohistochemical staining of the 91 gastric cancer samples. While the positive c-Met expression rate was 71.4% (65/91), the positive Jagged1 expression rate was 76.9% (70/91). Patients with positive c-Met expression had poorer overall survival than patients with negative c-Met expression (P=0.043) (Fig. 1A). Our previous study showed that patients with positive Jagged1 expression also had poorer prognosis than patients with negative Jagged1 expression (3). Jagged1 expression was positively correlated with c-Met expression in the human gastric cancer tissues (P=0.004) (Fig. 1B). Taken together, these results indicate a positive correlation between the c-Met and Jagged1 signaling pathways.

To evaluate the effect of HGF on the Notch1/Jagged1 pathway, N1IC-overexpressing (SC-M1/HA-N1IC) and control (SC-M1/pcDNA3) human gastric cancer SC-M1 cells were used in our experiment. To reduce the influence of the growth factors in FBS, the concentration of FBS was tapered to 0.5%. Cells were treated with various concentration of HGF (0, 5 or 25 ng/ml) (−, + and ++, respectively in the figures) and/or NS398 (0 or 50 µM) (− and +, respectively in the figures). The cell growth of SC-M1/pcDNA3 cells was increased in proportion to the elevation of HGF concentration. However, this phenomenon was not observed in the SC-M1/HA-N1IC cells. The COX-2-specific inhibitor NS398 blocked cell growth in both SC-M1/HA-N1IC and SC-M1/pcDNA3 cells, and HGF-stimulated cell growth in SC-M1/pcDNA3 was also blocked by NS398 (Fig. 1C). These data showed that HGF can enhance the proliferation ability of gastric cancer with lower Notch1 (N1IC) activity. However, the effect of HGF treatment on proliferation in gastric cancer cells with constitutively overexpressed Notch1 (N1IC) was not obvious, such that the COX-2-specific inhibitor NS398 repressed the migration ability in gastric cancer cells with both lower and higher Notch1 (N1IC) activity.

Next, we investigated the migration ability of gastric cancer cells using wound healing migration and Transwell migration assays. In the wound healing migration study, SC-M1/HA-N1IC and SC-M1/pcDNA3 cells were treated with HGF (0 or 25 ng/ml). The relative migration area was increased after HGF treatment in SC-M1/pcDNA3 cells compared to HGF treatment in SC-M1/HA-N1IC cells (Fig. 2). Furthermore, we investigated the migration ability using a Transwell migration assay. Compared to the migration ability of SC-M1/pcDNA3 cells, SC-M1/HA-N1IC cell migration increased independently of HGF. The migration ability of SC-M1/pcDNA3 cells was increased after HGF treatment. However, there was no significant difference in the migration ability of SC-M1/HA-N1IC cells before and after HGF treatment (Fig. 3). NS398 treatment reduced the migration ability in the SC-M1/pcDNA3 and SC-M1/HA-N1IC cells (Fig. 4). These results suggest that HGF can enhance the migration ability of gastric cancers with lower Notch1 (N1IC) activity. However, the effect of HGF on gastric cancer cells with constitutively overexpressed Notch1 (N1IC) was not obvious. The COX-2-specific inhibitor NS398 can repress the migration ability in gastric cancer cells with both low and overexpressed N1IC.

First, we compared the expression of c-Met and, Jagged1 by western blot analysis in Notch1-overexpressing, control and WT SC-M1 gastric cancer cells. Jagged1 expression in SC-M1/WT and SC-M1/pcDNA3 cells was lower than that in SC-M1/HA-N1IC cells, but c-Met protein expression was lower in SC-M1/pcDNA3 and SC-M1/HA-N1IC cells than that in the SC-M1/WT cells. The expression of c-Met protein in the SC-M1/pcDNA3 cells was higher than that in the SC-M1/HA-N1IC cells. Reduced expression of c-Met protein was observed in Notch1-overexpressing SC-M1 cells (Fig. 5A). Second, we treated gastric cancer cells with HGF and examined the effects over time. After HGF (25 ng/ml) treatment for 1, 2 or 4 h, cell lysates were collected for western blot analysis. HGF stimulated the expression of Jagged1 and COX-2 proteins in the SC-M1/pcDNA3 cells in a time-dependent manner. However, Jagged1 and COX-2 proteins levels were not obviously increased in the SC-M1/HA-N1IC cells. A time-dependent increase of Jagged1 was observed in the HGF-stimulated SC-M1 cells but not in the Notch1-overexpressing SC-M1 cells (Fig. 5B). COX-2 expression was elevated in the SC-M1/WT and SC-M1/pcDNA3 cells after HGF treatment. However, no significant change in COX-2 expression was observed in the SC-M1/HA-N1IC cells (Fig. 5C). Taken together, these results imply that a feedback loop exists between the HGF/c-Met and Notch1/Jagged1 signaling pathways.