Gastric cancer cell lines (MKN28, KATOIII) were cultured in RPMI 1640 medium supplemented with 10% fetal bovine serum, 100 µg/ml streptomycin and 100 unit/ml penicillin and incubated at 37℃ in a humidified atmosphere containing 5% CO₂ in air (referred to as the normoxic condition).

To obtain the intestinal proteins, cells and tissues were homogenized in RIPA buffer (20 mM Tris, 1 mM EDTA, 255 mM sucrose, pH 7.1). The supernatants were used. The protein concentrations were determined using a PIERCE Protein Assay Regent (Pierce, Rockford, IL, USA). Each protein sample was mixed with Laemmli sample buffer (Bio-Rad, Hercules, CA, USA). Western blots were performed using 20 µg of total protein. Protein samples were analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis in 10% acrylamide gels. The proteins were transferred onto nitrocellulose membranes (Protran, Whatman, Dassel, Germany), which were blocked in 5% non-fat dry milk in TBS buffer for 1 hour. Then, the nitrocellulose membrane was incubated with a polyclonal antibody against E-selectin (Biovision, Inc., Milpitas, CA, USA, 1 : 1,000) and ICAM-1 (SANTACRUZ, Santa Cruz Biotechnology, Inc., Santa Cruz, CA, USA, 1 : 100) diluted in blocking solution overnight at 4℃. Afterwards, the membranes were rinsed three times with TBS, 0.1% Tween-20 for 10 minutes, and re-incubated for 1 hour at room temperature in blocking buffer with 1 : 2,500 diluted horseradish peroxidase-conjugated secondary antibody, and then washed three times for 10 minutes. Protein detection was performed using the electrochemiluminescence Plus system (Amersham Biosciences, GE Healthcare, Pittsburgh, CT, USA). Optical density analysis of proteins was performed using Image J software (download directory at "imagej.nih.gov/ij/download/."). Image J is a public domain Java image processing program produced by National Institute of Health. The negative control was the zero signal density, which was the amount of E-selectin protein in KATOIII cells. The reference level was an amount of protein expression in normal stomach tissue determined as one. The amount of protein expression levels in other gastric cancer tissue and cells were measured by the proportion with the amount of density in normal stomach tissue level.

We studied 157 patients with gastric cancer who underwent gastrectomy at Korea University Hospital from 2003 to 2006. The mean follow-up period was 35.8 months (range, 15~69 months). The clinical data, including gender, age, and cancer recurrence including distant metastasis were obtained from medical records and the data on tumor size, depth of tumor invasion, lymph node metastasis, Lauren classification, lympho-vascular invasion and perineural invasion were obtained from pathology reports. Before designing this study, we received the approval of the Institutional Review Board in Korea University Hospital during the process of specimen and data collection (Approval number: AS0825-001).

Tissue micro-array was made for the primary tumors of all patients. H&E stained slides of primary tumor tissue were reviewed and central portion and invasion front of tumor was selected. Tumor tissue at the marked portion was aspirated from the paraffin block by Quik-Ray™ SYSTEM (Unitma Co., Ltd., Seoul, Korea, Core needle biopsy). Two cores were cut from each donor block for the tissue microarray. Five tissue array blocks were prepared and each contained about 60 tumor cores. Sections of 5 mm thickness were cut and processed for ICAM-1 and E-selectin immunostaining. Sections were stained for ICAM-1 and E-selectin antigen using the avidin-biotin-peroxidase complex method. We used ICAM-1 rabbit monoclonal antibody (Epitomics, Inc., Burlingame, CA, USA) and E-selectin polyclonal antibody (Biovision, Inc.) as primary antibodies. After deparaffinization in xylene and washing in ethanol, sections were heated in 0.01 M sodium citrate buffer (pH 6) at 95℃ for 20 minutes to unmask hidden epitopes.

After cooling to room temperature and wash in 0.01 M phosphate-buffered saline, the slides were processed in DAKO EnVsion™+/HRP kit according to the manufacturer's instructions (Copenhagen, Denmark). The ICAM-1 monoclonal antibody was diluted 1 : 100 in phosphate-buffered saline and E-selectin polyclonal antibody was diluted 1 : 1,000, and then they were reacted with tissue specimens at 37℃ for 30 minutes. Diaminobenzidine was used as a chromogen, followed by hematoxylin counterstaining. Normal rabbit IgG was substituted for primary antibody as the negative control. For positive controls, staining patterns of endothelial cells or leukocytes were checked. The specimens were examined under the microscope. When the tumor specimens were reacted with anti-ICAM-1 and E-selectin antibody, they showed various staining patterns. The degree of positive staining of tumor cells was represented as an approximate percentage of positive cells. As Yashiro et al.(18) indicated, for analyses, specimens were considered positive if more than 50% of the cancer cells expressed ICAM-1 and E-selectin. Specimens with less than 50% staining of cancer cells were classified into the ICAM-1 and E-selectin-negative groups.

The statistical significance of the correlation between ICAM-1 expression and clinicopathologic features was evaluated by chi-square test and independent t-test. Survival curves were constructed according to the Kaplan-Meier method. For statistical differences between curves, a P-value was calculated using the log-rank test. A Cox regression model was used to examine independent prognostic factors in patients. All statistical analyses were performed using SPSS 13.0 for Windows (SPSS Inc., Chicago, IL, USA) and a P-value<0.05 was considered statistically significant.

Western blotting was performed using gastric cancer tissues as well as corresponding tissues from non-cancerous regions and two gastric cancer cell lines, MKN28 and KATO-III. ICAM-1 expression was noted in normal stomach tissues and the level of expression was increased as much as four times in gastric cancer tissues compared with normal stomach tissues. However, E-selectin expression was more prominent in normal stomach tissues and it was decreased in gastric cancer tissues. In cultured gastric cancer cell lines, MKN28 and KATO-III, the level of expression of ICAM-1 was increased, especially in MKN28, but the level of expression of E-selectin was not increased (Fig. 1).

From 157 gastric cancer tissues, nearly 50% (n=79) tissues were positive for ICAM-1 and 54% (n=85) were positive for E-selectin by immunohistochemical staining (Fig. 2). There was a reverse relationship between the positivity of ICAM-1 expression and positivity of E-selectin expression (Table 1).

In our analysis, there was no statistically significant relationship between the expression of E-selectin and clinicopathologic factors and prognosis in these gastric cancer patients.

Statistical analysis showed that the ICAM-1 positive group had a larger tumor size, advanced depth of tumor invasion, positive lymph node involvement and distant metastasis, namely advanced cancer stage. And hence, the expression of ICAM-1 was significantly related with the positivity of perineural invasion (Table 2).

In the relationship between ICAM-1 and cancer recurrence, ICAM-1 was significantly correlated with cancer recurrence and when divided according to the recurrence pattern, it was mostly associated with peritoneal seeding (56.3%, 18/32 cases) (Table 3).

In survival analysis, ICAM-1-positive group showed a poor 3-year survival than the ICAM-1-negative group. (54.9% vs. 85.9%) and the disease-free survival result was also the same (Fig. 3).

In analysis of prognostic factors in these patients, ICAM-1 positive expression, diffuse type, depth of tumor invasion, lymph node involvement, distant metastasis, lymphovascular invasion and perineural invasion were proved to be negative prognostic factors in univariate analysis. However, in multivariate analysis, depth of tumor invasion and distant metastasis were proved to be independent prognostic factors (Table 4).