In this study, cases of patients with GC who underwent resection were selected and retrospectively reviewed. A total of 459 cases were randomly selected for inclusion in this cohort from among 2495 patients who underwent gastrectomy at Seoul St. Mary’s Hospital (Korea) between January 2000 and December 2009. The cases were limited to patients with stage I to stage III cancer; gastrectomies were not initially performed in patients with stage IV disease. After excluding missing cases, 411 cases were selected for analysis.

The collected data included patient demographics, clinical findings, pathological diagnosis, and additional treatments, such as chemotherapy and specific regimens. Survival time was calculated from the day of surgery to the last known date when the patient was confirmed to be alive or recurrence-free. Data from medical records and from the National Cancer Center of Korea database were used if data from Seoul St. Mary’s Hospital were unavailable. Cancer stages were defined using the eighth edition of the American Joint Committee on Cancer (AJCC) criteria [16]. The tumor histologic grades were determined according to the AJCC Cancer Prognostic Factors Consensus Conference (grade 1, well-differentiated; grade 2, moderately differentiated; grade 3, poorly differentiated; and grade 4, undifferentiated) [16].

The specimens were collected with the approval of the Institutional Review Board of the Seoul St. Mary’s hospital, Catholic University of Korea (KC20RISI0503). The need for informed consent was waived due to the retrospective nature of the study.

Tissue microarrays (TMAs) were produced from formalinfixed, paraffin-embedded tumor tissues acquired from primary GC specimens, using a manual tissue arrayer (Quick-Ray Manual Tissue Microarrayer; Unitma Seoul, Korea).

TMA slides were prepared and stained after the surgical resection of GC lesions to assess protein expression (Manual tissue arrayer MTA-1; Beecher Instruments, Sun Prairie, WI, USA). Immunohistochemical staining was performed for ABCG2 and CD133.

Tissue sections (4-μm thick) were cut from paraffin-embedded blocks, followed by antigen retrieval and blocking of endogenous peroxidases after sample deparaffinization and rehydration. The sections were then incubated with rabbit polyclonal anti-ABCG2 monoclonal antibodies (dilution, 1:100, cat. no. sc-18841; Santa Cruz Biotechnology, Dallas, TX, USA). For the negative controls, the primary antibodies were substituted with normal rabbit IgG at the same concentration. Additional sections were immunostained using anti-CD133 antibodies (dilution, 1:650, GTX60471; GeneTex, San Antonio, TX, USA) at a 1:650 dilution. The sections were incubated at room temperature for 90 minutes with the anti-CD133 antibodies and overnight at 4°C. Positive controls were used as suggested by the supplier. The primary antibody was omitted from the negative control specimens. Normal positive control staining was performed using gastric epithelial tissue.

A semiquantitative scoring system, based on staining intensity, was used to score CD133 and ABCG2 expression. The staining intensity was graded from 0 to 3 (0, no expression; 1, weak; 2=moderate; 3=strong). The ABCG2 and CD133 immunohistochemical staining results, stratified by grade, are shown in Fig. 1. The expression levels of ABCG2 and CD133 were dichotomized as either high or low based on the median expression patterns (staining intensity >2).

Immunohistochemical staining scores were used to determine the levels of ABCG2 and CD133 expression, which were then assessed for association with disease severity and patient survival. Overall survival (OS) and recurrence-free survival (RFS) rates were estimated. OS and RFS were calculated as the number of years from randomization to the date of death or last contact.

Continuous data are presented means±standard deviations; categorical data are expressed as quantities and proportions. To evaluate the differences between groups, according to their ABCG2 and CD133 immunoreactivity, χ2 and two-sample in-dependent t-tests were used for categorical data and continuous variables, respectively. Cumulative patient survival curves, according to ABCG2 and CD133 immunoreactivity, were generated using the Kaplan–Meier method; between-group differences were compared using the log-rank test. Multivariate analyses were performed using the Cox proportional hazards regression model to identify factors associated with OS and RFS. Survival distribution was estimated using the Kaplan– Meier method. Univariate and multivariate Cox proportional hazards models were used to explore the associations between the biomarkers or clinical variables and OS. Multivariate models were adjusted for covariates. The score and likelihood ratio test p-values were used to test the significance of each covariate in the univariate and multivariate models, respectively.

The characteristics of the 411 patients expressing ABCG2 and CD133 included in the analysis are summarized in Table 1. The average age of the patients was 58.2±11.9 years, with males accounting for 67.9% of the included patients. The cancer stage of each patient was classified based on the TNM staging system using the AJCC criteria; 31.4%, 28.5%, and 40.1% of the patients had stage I, II, and III disease, respectively. Poorly differentiated tumors were observed in 61.6% of the patients, and the percentage of undifferentiated tumors was higher in tumors at higher stages (48.8%, 58.1%, and 73.9% for stages I, II, and III, respectively; p<0.001).

Total gastrectomies were performed on 36.0% of the patients, and the proportion of total gastrectomies increased as the disease stage increased (21.7%, 39.3%, and 44.8% for stages I, II, and III, respectively; p<0.001). Adjuvant chemotherapy was administered to 64.0% of the patients. Recurrence was observed in 23.1% of the patients; 38.9% of the patients died. Both OS and RFS trended downward as the cancer stage increased (Fig. 2).

ABCG2 expression was significantly and positively correlated with CD133 expression (R=0.559; p<0.001). High levels of ABCG2 expression were observed in 74.9% of patients; high CD133 expression was found in 80.5%. The clinicopathological characteristics, based on ABCG2 and CD133 expression, are summarized in Table 2. Both ABCG2 and CD133 expression levels were higher in patients with stage III disease than in those with stage I or II disease (p<0.001 and p=0.002, respectively) (Fig. 3). Lower ABCG2 expression was associated with higher nodal stages. Furthermore, low expression levels of both ABCG2 and CD133 were associated with poor differentiation.

Fig. 4 summarizes the OS and RFS rates associated with ABCG2 and CD133 expression levels. Patients with high CD133 expression showed worse OS and RFS outcomes than did the group with lower CD133 expression, although the difference was not statistically significant. In contrast, the group with high ABCG2 expression showed significantly improved outcomes than did the low ABCG2 expression group in terms of both OS (p=0.029) and RFS (p=0.006).

Tumor cells expressing ABCG2 and CD133 were divided into four groups based on their expression levels: ABCG2^low CD133^low, ABCG2^low CD133^high, ABCG2^high CD133^low, and ABCG2^high CD133^high. When assessing the OS and RFS for each group, significant differences were observed (p=0.014 and p< 0.001, respectively) (Fig. 5A). In the survival analysis (Fig. 4), high ABCG2 expression was associated with favorable outcomes, whereas high CD133 expression was associated with poor outcomes. When ABCG2 expression was high and CD133 expression was low, there was no statistically significant difference in outcomes between the groups. However, the ABCG2^highCD133^low group had the best outcomes, whereas patients with ABCG2^lowCD133^high cells had the worst OS and RFS outcomes (p=0.012 and p=0.003, respectively) (Fig. 5B).

Cox proportional hazards regression analysis was performed to examine multiple parameters. In the univariate analysis, stage, age, chemotherapy, and ABCG2 expression were associated with OS. After adjusting for these factors, stage, ABCG2 expression, and CD133 expression remained significantly associated with OS. Stage, chemotherapy, and ABCG2 expression were significant predictors of RFS. However, after multivariate analysis, only tumor stage, ABCG2 expression, and CD133 expression were significant predictors of RFS (Table 3); adjusted survival curves corresponding to stage and ABCG2 and CD133 expression levels are shown in Fig. 6.

To investigate the effect of ABCG2 expression on survival, a subgroup analysis was performed in patients who underwent systemic chemotherapy. The patients were divided into two groups based on whether they received systemic chemotherapy. Compared with those expressing low levels of ABCG2, the high ABCG2 expression patients exhibited better outcomes among individuals who received systemic chemotherapy, specifically demonstrating a significant improvement in RFS (p=0.039). Conversely, there was no statistically significant difference in the outcomes of patients who did not receive chemotherapy based on their ABCG2 expression levels. The results are shown in Fig. 7.