Combined features based on MT1-MMP expression, CD11b + immunocytes density and LNR predict clinical outcomes of gastric cancer

The records of GC patients who underwent surgery with a curative intent at the Department of Oncology, Zhongnan Hospital of Wuhan University (Wuhan, China) between December 2002 and February 2011 were reviewed. Major demographic and clinico-pathological characteristics were available. No patients received neoadjuvant chemotherapy before surgery. TNM stage was determined according to the 7th edition UICC/AJCC TNM system [18]. By the most recent follow-up on May 31, 2012, the median follow-up was 59.5 (range: 16.8-102.3) months. A total of 108 (58.7%) patients died. Overall survival (OS) was defined as the interval from the date of surgery to GC-related death. Any recurrence in abdomino-pelvic cavity was defined as local-regional recurrence; and recurrence via blood flow was defined as distant metastasis, such as liver metastasis and lung metastasis. Written informed consent was obtained from the patients and the study protocol was approved by the ethics committee of Zhongnan Hospital of Wuhan University. The study was undertaken according to the ethical standards of the World Medical Association Declaration of Helsinki.

All hematoxylin and eosin (HE)-stained slides were examined by independent reviewers who were not aware of the clinical characteristics or clinical outcomes. Three tissue microarrays (TMAs) were constructed. Briefly, two cores were taken from each representative tumor tissue (1.5 mm in diameter for each core). Then, three TMAs sections with 184 tumor tissues were constructed (in collaboration with Shanghai Biochip Company Ltd., Shanghai, China).

The primary antibodies for MT1-MMP (sc101451, Santa Cruz, USA, dilution 1/100), CD11b + (ab52478, Abcam, UK, dilution 1/100), the corresponding horseradish peroxidase (HRP) conjugated secondary antibody (ab97265, ab97200, Abcam, UK, dilution 1/300), and diaminobenzidine (DAKO, Denmark) were obtained and validated for labeling.

The immunohistochemistry protocols were described previously [19]. In short, three TMAs sections were deparaffinised with xylene thenrehydrated through three changes of alcohol. Endogenous peroxidase activity was blocked by 0.3% hydrogen peroxide for 10 min. Antigen retrieval was performed by microwave treatment in 0.01 M Tris-EDTA buffer (pH 9.0) for 20 min. Slides were incubated for 1 h with the primary antibodies for MT1-MMP (sc101451, Santa Cruz, USA, dilution 1/100), CD11b + (ab52478, Abcam, UK, dilution 1/100). After washing with Tris-buffered saline (TBS, pH 7.4), sections were incubated with b the corresponding horseradish peroxidase (HRP) conjugated secondary antibody (ab97265, ab97200, Abcam, UK, dilution 1/300) for a further 50 min. Antigen–antibody reaction was visualised with 3,30-diaminobenzidine (DAKO, Denmark). After counterstaining with haematoxylin, sections were dehydrated through ascending alcohols to xylene and mounted. Positive staining controls were performed in parallel with paraffin sections of normal human tonsil. Negative control was performed by omitting the primary antibody.

The slides were examined under Olympus BX51 fluorescence microscope equipped with an Olympus DP72 camera (Olympus Optical Co., Ltd., Tokyo, Japan). Panorama of each labeled core was obtained based on 4 photographs under low-power magnification (×100) with high resolution by DP-BSW software (Olympus Optical Co., Ltd., Tokyo, Japan). Identical settings were used for every photograph, so as to minimize the selection bias. The MT1-MMP expression and CD11b + immunocytes density were digitally evaluated by Image-Pro Plus v6.2 software (Media Cybernetics Inc, Bethesda, MD). To keep results reliable and comparable, a uniform setting for all the slides was applied for the reading of each antibody staining. Integrated optical density (IOD) of all the positive staining of MT1-MMP in each photograph was measured, and its ratio to total area of each photograph was calculated as MT1-MMP expression. CD11b⁺ immunocytes density was recorded as the number of positive cells per tissue surface unit in square millimeters. Cut points for MT1-MMP density was explored by “the best cut-off approach by log-rank test” [20]. And the 25th percentile value was defined as the cut-off point to determine the MT1-MMP negative and positive expression subgroups in this study. For the CD11b + immunocytes density, the cut-off point for the definition of subgroups (low and high CD11b + immunocytes density) was the median value.

Statistical analyses were performed with SPSS software (version 18.0, SPSS Institute, Chicago, IL). The Pearson χ² test or Fisher’s exact test was used to compare qualitative variables. Kaplan-Meier analysis was used for survival analysis and significance among patients’ subgroups was calculated by log rank test. The Cox regression model was used to perform multivariate analysis. Logistic regression was used to assess the influence of binary factors. Receiver operating characteristic curve (ROC) analysis was used to determine the predictive value of the parameters. Two sided P < 0.05 was considered as statistically significant.

Among the 184 cases included in this study were 132 (71.7%) males and 52 (28.3%) females, ranging in age from 25 to 85 yr (Mean ± SD: 57.9 ± 12.9 yr). The main demographic and clinico-pathological characteristics were presented in Table 1.

Immunohistochemistry was performed in all GC TMAs and the result of each specimen was obtained for images-based digital analysis (Figure 1A-1D). MT1-MMP staining was mainly in the cytoplasm or on the cell membrane of tumor cells. Most of the stromal cells were negative, although sporadic positive staining on these cells was also observed (Figure 1E, 1G). CD11b + immunocytes mainly infiltrated into the juncture of cancer nest and stromal (Figure 1F, 1H). The level of MT1-MMP density and CD11b + immunocytes density were presented in Table 1. The 25th percentile of MT1-MMP density was 0.00186 and 138 (75.0%) patients were documented as MT1-MMP positive according to the abovementioned classification criteria. The median value of CD11b + immunocytes were 257 cells per tissue surface unit in square millimeter.

There were 77 (55.8%) and 15 (32.6%) patients with high CD11b + immunocytes density in MT1-MMP positive (n = 138) and negative (n = 46) subgroups, respectively (P = 0.006). The likelihood of MT1-MMP expression was decreased significantly for low CD11b + immunocytes density subgroup (odds ratio (OR) = 0.383 [95%CI: 0.190-0.774], P = 0.007, Logistic regression). The MT1-MMP expression and CD11b + immunocytes density were correlated with the clinico-pathological features as summarized in Table 2. MT1-MMP expression was correlated with small tumor size (P = 0.007), tumor at distal stomach (P = 0.027) and increased recurrence risk (P = 0.001), but not correlated with age, gender, histological grade, serosa invasion, lymph node status, TNM stage and sites of tumor recurrence (P > 0.05 for all). CD11b + immunocytes density was correlated with lymph node status (P = 0.015), LNR (P = 0.018) and TNM stage (P = 0.011), but not correlated with age, gender, tumor size, tumor location, histological grade, serosa invasion and recurrence risk (P > 0.05 for all).

For 184 cases, the median OS was 26.8 (range 0.8-102.3) months, and the 1-, 3-, and 5-yr survival rate was 79.3%, 50.5% and 37.4%, respectively (Figure 2A). As expected, those traditional factors were associated with GC patients’ OS, such as tumor size, histological grade, lymph node status, serosa invasion and TNM stage (P < 0.05 for all). In addition, both high CD11b + immunocytes density and MT1-MMP negative expression were correlated with better OS (P value was 0.031 and 0.015, respectively.) (Figure 2B, 2C). The discrimination ability in OS was increased in LNR subgroups (P < 0.001) compared to lymph nodes status subgroups (P = 0.015) (Figure 2D, 2E ), and ROC analysis showed that, in terms of predicting risk of death, the classification by LNR (Area under the curve: 0.659 [95%CI: 0.579-0.739], P < 0.001) had better performance than positive lymph nodes number (Areas under the curve: 0.568 [95%CI: 0.484-0.653], P = 0.114). All results indicated that MT1-MMP positive, low CD11b + immunocytes density and high LNR were risk factors related to poor prognosis (Figure 2, Table 3).

As we proposed above, combined features based on the number of risk factors were explored to improve prediction of GC prognosis (Table 3). Thus, patients were divided into two subgroups according to the number of risk features: group I, less than two risk factors were observed; and group II, two or three risk factors were observed. Combined analysis showed that the OS of patients in group II was significantly shorter than patients in group I (P < 0.001) (Figure 2F).

Factors showing significance by univariate analysis were integrated into multivariate Cox proportional hazards analysis. In this study, LNR, MT1-MMP, CD11b and combined features were highly correlated. Therefore, two separate multivariate models were generated to avoid the multicollinearity among those variables. LNR, MT1-MMP expression and CD11b + immunocytes density were independent prognostic factors for OS after excluding other confounding factors (P < 0.05 for all) Moreover, the combined features were independent prognostic factor, too (HR = 3.818 [95%CI: 2.223-6.557], P < 0.001) (Table 4).

ROC analysis was implemented to further evaluate the prognostic performance of the four independent factors in this study. Combined features would be better to predict the clinical outcomes of GC patients compared to other factors (Area under the curve: 0.689 [95%CI: 0.609-0.768], P < 0.001) (Figure 3).