The human gastric cancer MKN-74 cell line, established at the National Cancer Center, was maintained in Dulbecco's modified Eagle's medium/F12 (Sigma-Aldrich; Merck KGaA, Darmstadt, Germany) supplemented with 10% heat-inactivated fetal bovine serum (FBS; Thermo Fisher Scientific, Inc., Waltham, MA, USA) at 37°C in an atmosphere containing 5% CO₂.

A total of 98 samples were obtained from patients who underwent surgery at the Department of Digestive Surgery and Surgical Oncology, Yamaguchi University Graduate School of Medicine (Yamaguchi, Japan) from January 2004 to December 2008. The patients were diagnosed with gastric cancer with the tumor invaded through the muscularis propria or deeper. No patients had received pre-operative therapy. Pathological staging was performed according to the Japanese Classification of Gastric Carcinoma (14th edition) based on the Japanese Gastric Cancer Association TNM staging system (11). The clinicopathological characteristics of the patients are summarized in Table I. The study protocol was approved by the Institutional Review Board for Human Use at Yamaguchi Hospital (Yamaguchi, Japan). Written informed consent for the present study was obtained from all patients prior to surgery.

Flow cytometric analysis was performed to confirm the expression of NKG2DLs on the MKN-74 gastric cancer cell and NKG2D on the surface of the NK and T cells. BD FACSFlow (BD Biosciences, San Jose, CA, USA) was used for buffer of flow cytometric analysis. ULBP1 against MKN-74 was stained using phycoerythrin (PE)-conjugated mouse anti-ULBP1 antibody (dilution, 1:10; cat no. FAB1380P; R&D Systems, Inc., Minneapolis, MN, USA), and ULBP2/5/6 against MKN-74 was stained using allophycocyanin (APC) -conjugated mouse anti-ULBP2/5/6 (dilution, 1:10; cat no. FAB1298A; R&D Systems, Inc.). CD3 against NK and T cells was stained using v450-conjugated mouse anti-CD3 (dilution, 1:20; cat no. 560351; BD Biosciences), CD8 against NK and T cells was stained using APC-Cy7-conjugated mouse anti-CD8 (dilution, 1:20; cat no. 557834; BD Biosciences), CD16 against NK and T cells was stained using v500-conjugated mouse anti-CD16 (dilution, 1:20; cat no. 561394; BD Biosciences), and CD56 against NK and T cells was stained using PE-CF594-conjugated mouse anti-CD56 (dilution, 1:20; cat no. 562289; BD Biosciences). All samples were maintained at 4°C for 30 min. MKN-74, mononuclear cells, activated T cells and activated NK cells were stained to identify each cell surface markers. Following 3 washes with BD FACSFlow buffer, the cells were resuspended in phosphate-buffered saline (PBS) with 2% FBS at a density of 2×10⁵ cells/100 µl. Appropriate isotype control antibodies were stained in the aforementioned manner.

Samples were analyzed with a BD LSRFortessa X-20 instrument (BD Biosciences), and data were analyzed with FlowJo software version ×10.0.7r2 (Tree Star, Inc., Ashland OR, USA). The relative fluorescence intensity (RFI) was calculated as follows: RFI = [(MFI) of molecule) - (MFI of corresponding isotype control)]/MFI of corresponding isotype control, where MFI is the mean fluorescence intensity.

Mononuclear cells (LP_241; Cellular Technology, Ltd., Cleveland, OH, USA) were activated to generate NK cells by addition of 5 µl/ml/well of mixed CD2 and CD335 beads with 500 U/ml/well IL-2 (all from Miltenyi Biotec GmbH, Bergisch Gladbach, Germany) for 2 h. For T-cell activation, the mononuclear cells were stimulated with 2 µg/ml/well anti-CD3 agonistic antibodies (eBioscience; Thermo Fisher Scientific, Inc.) and 1 µg/ml/well anti-CD28 agonistic antibodies (BioLegend, San Diego, CA, USA) with 100 U/ml/well IL-2 (Miltenyi Biotec GmbH) for 2 h.

Paraffin-embedded tissue blocks were cut into 5 µm sections, mounted on saline-coated slides, and subsequently dewaxed and rehydrated using xylene and graded alcohol washes. Antigen retrieval was performed by microwaving in Target Retrieval solution (pH 9; Dako; Agilent Technologies, Inc., Santa Clara, CA, USA). Endogenous peroxidase was blocked at room temperature (RT) for 15 min with 0.3% hydrogen peroxide (Wako Pure Chemical Industries, Ltd., Osaka, Japan) in PBS containing 0.1% sodium azide. Subsequent to washing twice in PBS, the sections were blocked with Protein Block Serum-Free (Dako; Agilent Technologies, Inc.) at RT for 30 min. The primary antibodies were then added. Anti-ULBP1 antibodies (cat no. HPA007547; 1:100; Sigma-Aldrich; Merck KGaA), anti-NKG2D antibodies (cat no. ab36136; 1:100; Abcam, Cambridge, UK), anti-CD56 antibodies (cat no. B159; 1:50), and anti-CD8 antibodies (cat no. SK1; 1:50) (both from BD Biosciences) were used for IHC according to the manufacturers' protocol. Briefly, following incubation with primary antibodies at 4°C for 15 h, the slides were washed twice in PBS, incubated with horseradish peroxidase-conjugated rabbit/mouse antibodies with original concentration at RT for 30 min (K406311-2; EnVision Plus mouse/HRP system; Dako; Agilent Technologies, Inc.), incubated with Dako-Chromogen solution, and washed in de-ionized water. Following background staining with Mayer's hematoxylin, the sections were dehydrated through ascending alcohols to xylene and mounted. To confirm the specificity of antibodies, negative control slides were incubated with mouse IgG1 monoclonal antibodies (ab81032; 1:100; Abcam). IHC results were evaluated by imaging under low-power magnification (×100) in eight fields of view to identify regions containing positive immunoreactivity. Immunostaining was additionally evaluated at a high-power (×400) magnification (ECLIPSE E200 and Nikon 1 V2 Micro imaging system; Nikon, Tokyo, Japan). Microscopic analysis of ULBP1 and NKG2D expression was performed by three blinded independent observers.

Evaluation of ULBP1 staining in the cancer cells was performed semi-quantitatively, as described previously by Loos et al (12) Evaluation of ULBP1 staining was based on the area and intensity of the staining. Scoring system for area of staining is as follows: Score 1, <33% of cancer cells; score 2, 33–66% of cancer cells; score 3, >66% of cancer cells. Scoring for staining intensity is as follows: Score 1, absent/weak staining; score 2, moderately intense staining and score 3, strong staining. Each section was given a final grade derived from the combination of the area and intensity scores. The final score was calculated by adding scores of intensity and area. Sections with a final score of 3 were classified as exhibiting low ULBP1 expression (ULBP1 low), whereas sections with a final score of >3 were classified as exhibiting high ULBP1 expression (ULBP1 high).

Absolute numbers of NKG2D-positive cells in 1 µm invasive tumor core biopsies were counted manually using an eyepiece reticule by three blinded independent individuals. The number of NKG2D-positive cells around the tumors was counted using a computerized image analysis system composed of an Olympus DP70 CCD camera on an Olympus AX70 light microscope (Olympus, Tokyo, Japan). Under magnification, ×400, there were ≥8 independent and intact computerized microscopic fields for the duplicates of each patient sample. A total of four independent microscopic fields (magnification, ×400), representing the densest lymphocytic infiltrates, were selected for each patient sample to ensure representativeness and homogeneity. The scores for four fields were averaged to calculate the final number for one computerized microscopic field at magnification, ×400 (0.0768 mm²/field). The evaluation of NKG2D-positive cells was performed by three independent blinded observers. Discrepancies in enumeration, within a range of 5%, were re-evaluated and a consensus decision was made. The ratio of NKG2D-positive cells/tumor-infiltrating lymphocytes (TILs) was calculated for each specimen. The median value was selected as the cut-off for defining TIL subgroups (median, 0.093). High and low ratios of NKG2D were termed NKG2D high and NKG2D low, respectively.

Differences were analyzed using Fisher's exact tests. Actuarial OS rates and recurrence-free survival rates were calculated by the Kaplan-Meier method and analyzed by the log-rank test. Univariate analyses were performed using Fisher's exact tests. Multivariate analyses were performed to determine factors affecting overall survival (OS) based on the Cox proportional hazards regression model. A secondary analysis was performed to assess the associations among the expression of NKG2DLs/NKG2D and clinicopathological characteristics by multivariate analysis based on the proportional hazards regression model. The activation of T cells and NK cells was analyzed by Mann-Whitney tests. For comparisons of individual variables, paired-sample t-tests were performed. Two-tailed P<0.05 was considered to indicate a statistically significant difference. All statistical analyses were performed with EZR version 1.35 (Saitama Medical Center, Jichi Medical University), a graphical user interface for R (The R Foundation for Statistical Computing, v.2.13.0), and more specifically a modified version of R commander (v.1.6–3) designed to add statistical functions frequently used in biostatistics.

In the flow cytometric analysis, cell surface expression of ULBP1 and ULBP2/5/6 in MKN-74 cells was constitutively high (Fig. 1A). By contrast, cell surface expression of MICA and MICB was not detected in MKN-74 cells (data not shown).

The cell surface expression of NKG2D on CD3⁺ T cells was significantly increased following T-cell activation (MFI, 621) compared with the expression (MFI, 243) prior to T-cell activation [P=0.0286, n=4, standard deviation (SD) = 378; Fig. 1B]. Similarly, cell surface expression of NKG2D on CD3⁺CD8⁺ T cells was significantly higher following T-cell activation (MFI, 1234) compared with expression (MFI, 515) prior to T-cell activation (P=0.0286, n=4, SD=719; Fig. 1C). Cell surface expression of NKG2D on CD16⁺CD56⁺ NK cells was significantly higher following NK cell activation (MFI, 974) compared with expression (MFI, 276) prior to NK cell activation (P=0.0283, n=4, SD=698; Fig. 1D).

ULBP1 expression was observed in the cell membrane and cytoplasm. Fig. 2A and B show representative images of ULBP1-high staining in gastric cancer (magnification, ×100 and ×400, respectively). Fig. 2C and D show isotype control staining in gastric cancer (magnification, ×100 and ×400, respectively). Of the 98 samples examined in the analysis, 70 were positive for ULBP1 expression.

NKG2D was expressed on the membrane of mononuclear cells around the gastric tumors. Fig. 2E and F reveal high NKG2D expression in TILs with (magnification, ×100 and ×400, respectively). Fig. 2G and H demonstrate isotype control staining in TILs (magnification, ×100 and ×400, respectively) in the same sections as those used in Fig. 2E and F. Of the 98 samples, 19 were positive for NKG2D.

Among the NKG2D-expressing tissue sections, a total of 6 samples were stained for CD8 and CD56 and were used as representative sections. Fig. 2I-L displays CD8 or CD56 expression in the same samples as those analyzed in Fig. 2E-H. In all tissues, the majority of NKG2D-expressing cells were CD8⁺ T cells.

Univariate analysis indicated that NKG2D expression in mononuclear cells was associated with lymphatic invasion (P=0.0028) and postoperative recurrence (P=0.0048; Table II). No significant associations were observed between ULBP1 expression and pathological findings or postoperative recurrence (Table II).

Survival curves following surgery were compared between high and low ULBP1 and NKG2D expression groups. A total of 70 patients were positive for ULBP1, whereas 28 patients were negative for ULBP1 (Table II). By contrast, a total of 19 and 79 patients were positive and negative for NKG2D, respectively (Table II). There were no significant differences in OS observed according to ULBP1 expression (P=0.334; Fig. 3A). However, patients with NKG2D expression were indicated to have significantly longer OS compared with patients without NKG2D expression (P=0.0217; Fig. 3B). Combined expression of NKG2D and ULBP1 was also analyzed and are as follows: ULBP1-high/NKG2D-high (n=12), ULBP1-high/NKG2D-low (n=58), ULBP1-low/NKG2D-high (n=7), and ULBP1-low/NKG2D-low (n=21). With regards to OS, the findings indicated that the ULBP1-high/NKG2D-high group had the best prognosis, and the ULBP1-low/NKG2D-low group had the poorest prognosis (P=0.0096; Fig. 3C).

Multivariate analysis was used to determine the hazard ratios for OS conferred by ULBP1, NKG2D, sex, lymph node metastasis, lymphatic invasion and venous invasion. NKG2D expression and lymph node metastasis were identified to be independent prognostic factors for OS (P=0.0431 and P=0.0030, respectively; Table III).