Distribution pattern of tumor associated macrophages predicts the prognosis of gastric cancer

Purpose As mayor biomarkers in tumor microenvironment (TME), tumor associated macrophages (TAMs) of gastric cancer (GC) still needs further studies in terms of the number and distribution pattern. Methods Herein, tissue microarrays (TMA) incorporating 494 GC surgical samples in duplicate were stained for TAMs infiltration analysis. TAMs number was counted according to the locations, including infiltrating macrophages in cancer nest (MC), in invasive front (MF) and in stroma (MS). Correlations between TAMs number, distribution pattern and clinic-pathological features and survival analyses were performed. Results Infiltrating macrophages number in GC tissues was much higher than that in peritumoral tissues. TAMs number was not significantly correlated with the overall survival (OS). TAMs distribution pattern could be categorized into MC or MF/MS dominant pattern, and correlated with histological grade (P =0.001). The median OS of MF/MS dominant pattern (22.1, 95%CI: 23.5-28.9) was significantly shorter than that of MC dominant pattern (25.6, 95%CI: 28.5-35.6) (P =0.002). By receiver operating characteristic curve (ROC) analysis, the predictive value of TAMs distribution pattern was superior to histological grade and pM stage, but inferior to pN and TNM stage. Conclusions TAMs distribution pattern could be an independent prognostic factor for the OS of GC patients, and patients with MF/MS dominant pattern had worse outcomes.


INTRODUCTION
Tumor microenvironment (TME) plays an important role in cancer progression and metastasis [1,2]. Within TME, distinct immune cells are recruited by cancerderived signals and mutually interact with cancer cells [3]. Tumor associated macrophages (TAMs), the most abundant immune-related stromal cells [4], act as key orchestrators in TME, by directly attacking cancer cells, or promoting cancer progression by suppressing antitumor immunity, or inducing angiogenesis [5].
Significant advances have been made in TAMs studies regarding their impacts on clinical outcomes. The clinical significance of TAMs can be influenced by the www.impactjournals.com/oncotarget/ Oncotarget, 2017, Vol. 8, (No. 54), pp: 92757-92769 Research Paper number, phenotypes and distributions at each pathological stage [6]. Studies in gastric cancer (GC) have shown that higher number of TAMs is associated with worse prognosis [7,8]. However, other reports have found that a higher level of TAMs infiltration results in a better outcome [9]. Therefore, it becomes controversial that TAMs emerge as significant but opposite predictors of survival for GC [10,11]. These conflicting results could be due to the fact that most studies pay attention to the ratios of TAMs with different phenotypic features [12,13], or ignorance of TAMs distributions by simply focusing on the number.
TAMs at different locations within the tumor may have different impacts on GC progression [14]. TAMs infiltration into tumor stroma has significant clinical relevance in GC, indicating the importance of not only studying the number but also studying the locations [15]. TAMs infiltration at invasive front could influence cancer metastasis through epithelial-mesenchymal transition (EMT) mechanism [16]. In turn, the degree of cell-to-cell contact may also influence the balance between protumoral and antitumoral properties of macrophages [17]. Some studies have also suggested that GC TAMs in different locations play different roles in relation to angiogenesis, stromal reaction, and prognosis [18]. Taken together, these results indicate that TAMs number and distributions are crucial factors to impact the coevolution between cancer cells and TAMs.
Therefore, this study evaluated both the number and locations of TAMs, especially the distribution pattern. TAMs number was counted at three locations, including cancer nest (MC), invasive front (MF), and stroma (MS). By comparison of TAMs number, TAMs distribution pattern could be defined as MC and MF/MS dominant pattern. Correlations of TAMs number and distribution pattern with GC clinical outcomes were both evaluated.
Among 204 GC cases in which TAMs presented both in cancer and peritumoral tissues, there was a trend towards significant higher number of TAMs in tumor tissues than peritumoral tissues (P <0.001) ( Figure 2D).

Correlations between total TAMs number and clinic-pathological features and OS
The relationship between total TAMs number and major clinic-pathological characteristics were studied in 494 GC patients (Table 1). Total TAMs number was significantly correlated with pathological types (P =0.003), serosa invasion (P =0.007) and TNM stage (P =0.009), but not significantly correlated with age, gender, tumor location, histological grade, lymph node metastasis, distant metastasis (P >0.05 for all).

TAMs locations and distribution pattern
According to the criteria on TAMs locations, 494 patients could be evaluated for MC, 319 patients could be evaluated for both MC and MF, and 296 patients could be evaluated for all MC, MF and MS. Major clinicpathological and survival information among the three databases were comparable (Supplementary Table 1). Detailed analyses were performed on 296 patients to study the impact of MC number, MF/MS number, and TAMs distribution pattern on GC prognosis.
The flowchart and detailed exclusion criteria was shown in Figure 4A. Representative photos of MC, MF, and MS were shown in Figure 4B-4D.

Correlations between MC, MF/MS number and clinic-pathological features and OS
Out of 296 GC patients, correlations between total TAMs number, MC number, MF/MS number, and major clinic-pathological characteristics were studied ( Table 2). The median value of total TAMs number was 25 (range: 0-171). Total TAMs number was significantly correlated with serosa invasion (P =0.019), but not significantly correlated with age, gender, tumor location, histological grade, lymph node metastasis, distant metastasis, and TNM stage (P >0.05 for all). MC number was significantly correlated with tumor location (P =0.018), and histological grade (P =0.007), but not significantly correlated with age, gender, serosa invasion, lymph node metastasis, distant metastasis (P >0.05 for all). MF/MS number (median value: 13, range: 0-166) was higher than MC number (median value: 6, range: 0-126), the difference was statistically significant (P <0.001). MF/ MS number was significantly correlated with histological grade (P =0.045), and lymph node metastasis (P =0.037), but not significantly correlated with age, gender, tumor location, serosa invasion, distant metastasis, and TNM stage (P >0.05 for all).

Uni-and multivariate analysis and ROC analysis of TAMs distribution pattern
The univariate analysis of clinic-pathological factors and TAMs distribution pattern regarding OS was also conducted. In univariate analysis, traditional clinicpathological features (such as histological grade, pT stage, pN stage and TNM stage), and MF/MS distribution pattern (P =0.002) were associated with OS (Table 3).
These factors were also integrated into multivariate Cox proportional hazards analysis. In multivariate The number and distribution of TAMs were analyzed and three TAMs locations including MC, MF, MS were found out. In virtual, one field could be divided into three parts, including cancer nest, cancer invasive front, and cancer stroma. The number of TAMs was evaluated for all locations. (B1) When the distance between two cancer nests were less than 50μm, such two cancer nests were considered just one big cancer nest. The TAMs in those areas were named as MC, MC=Total TAMs. (B2) The TAMs in stroma and nest which stay from the junction of stroma and nest less or equal to 25μm were named as MF. When the distance between two cancer nests were just equal to 50μm, there were only MC and MF, MC+MF =Total TAMs. (B3) When the distance between two cancer nests were more than 50μm, such two cancer nests were considered as two distinct cancer nests. The TAMs in stromal which stay from the neighboring cancer nest more than 25μm were named as MS. MC+MF+MS =Total TAMs. (B4) In this study, there were 296 patients could evaluate MC, MF and MS simultaneously. analysis, significant factors correlating with OS were pN stage (P =0.008), pM stage (P =0.020), TNM stage (P <0.001), and TAMs distribution pattern (hazard ratio, HR =2.177 [95%CI: 1.449-3.270], P <0.001) ( Table 4).
ROC analysis was conducted to further evaluate the prognostic performance of histological grade, pT stage, pN stage, pM stage, TNM stage, and MF/MS distribution pattern. As shown in Figure

DISCUSSION
Inflammation is one of the significant hallmarks of cancer, and influences the coevolution between cancer cells and TME [19]. Within the TME, TAMs recruited by tumor-derived signals might be the most important inflammation cells [20,21]. Evaluating both the number and distributions could promote a comprehensive understanding of TAMs significance during GC progression.
In this study, TAMs number in GC tissues was much higher than that in peritumoral tissues. This result, in accordance with some previous studies, proved that TAMs were recruited and aggregated within tumor tissues, then mutually interacted with tumor cells [22]. In fact, TAMs also correlated with tumor progression. TAMs number had been identified as an independent prognostic factor in several cancer types, such as breast cancer and colorectal cancer [23,24]. Herein, total TAMs number was higher in GC patients with no serosa invasion, early TNM stages, and in survival patients, which indicated TAMs may be protective factors of GC. To investigate the exact prognostic significance of TAMs, detailed analyses were focused on total TAMs number and OS, and no correlations were found. Such contradictory results have also been reported previously [25,26]. Some studies considered TAMs as protumoral through induction of    angiogenesis and suppression of antitumor immunity [27], while other reports found that a dense TAMs infiltration positively influenced prognosis in GC [9]. To explore such contradiction, we studied TAMs from aspects of both the number and histological distributions. Based on the spatial position relationships between TAMs and tumor cells, MC and MF/MS number were counted and analyzed. The distribution characteristic of TAMs had also been mentioned previously [28,29]. Su et al. elucidated mutual interactions between cancer cells and TAMs at tumor invasive front [28]. Additionally, the close vicinity of TAMs and tumor neo-vessels was observed in tumor stroma, and constituted tumor invasion unit [29]. On the basis of these morphological studies, quantitative    [30]. Taken together, TAMs number might not be a significant prognostic marker related to GC OS. Herein further analysis was developed using TAMs distribution pattern through a definition by systematical integration of TAMs number and histological distributions, which is crucial to reflect the coevolution between tumor cells and TME [31]. In this study, GC patients with MF/MS dominant pattern had worse clinical outcomes than MC dominant pattern, indicating MF/MS as main contributors to promote GC progression. Moreover, the predictive value of TAMs distribution pattern was verified by ROC analysis. Theoretically, the heterogeneity and plasticity were hallmarks of macrophages [32]. TAMs could undergo phenotypic switch from the classically activated macrophages (also known as M1 macrophages) to the alternatively activated macrophages (M2 macrophages). M1 macrophages were described as pro-inflammatory to display tumor-resistant effects, while M2 macrophages often associated with tumor-promoting properties. Consequently, TAMs gained potentials to promote cancer cell motility in invasion areas, to promote metastasis in stromal and perivascular areas, and to stimulate angiogenesis in avascular and peri-necrotic hypoxic areas [33]. Therefore, TAMs among different sites in GC tissues might represent distinct significances and prognostic values [34]. Owing to vigorous cell-to-cell contacts among TAMs, cancer cells and other activated stromal cells, TAMs along tumor invasive front are of great significance. At tumor invasive front, TAMs might correlate with various signaling pathways [35], undergo phenotypically switch from M1 to M2 [36], and finally promote the acquisition of specific pathological features of cancers such as immunosuppression, neovascularization and modification of extracellular matrix [37]. Pinto et al. have shown that TAMs with M2 phenotype were recruited by multiple chemoattractant and accumulated at the interface between cancer nest and stroma [38], thereby raising a hypothesis that the main constituent of MF/MS was M2 macrophages, which had been verified in some studies and correlated with poor prognosis [39]. With these in consideration, MF/MS could be a predominant and decisive factor, while MC might be a less important factor on GC prognosis. TAMs distribution pattern integrated both MC and MF/MS into consideration, and could be an independent prognostic factor of GC. The advantage of TAMs distribution pattern was due to the significance of researching TME and complex quantitative analyses [40]. However, limitations also existed by optimizing single biomarker, thus demonstrating TAMs distribution pattern inferior to TNM stage. To make sense of the overall landscape of TAMs, combined analyses with other biomarkers may provide new insights for deep understandings [41].
In summary, TAMs number merely correlates with several unfavorable clinic-pathological features, but has no significant prognostic value. Both the number and distributions should be taken into consideration for TAMs evaluations. Promisingly, TAMs distribution pattern could be a new factor related to GC OS.

Study population and database
The records of patients who underwent surgical resection of GC from December 2002 to February 2011 were reviewed. Major demographic and clinicpathological characteristics were retrieved. The tumor type, histologic grade, depth of invasion, number of lymph nodes retrieved, and number of lymph nodes with metastases were re-confirmed histologically. Inclusion and exclusion criteria were defined as follows. Patients were included when histology confirmed adenocarcinoma of the stomach and the survival data were available. Patients were excluded when distant metastasis has been diagnosed before surgery, histology identified a pathological type other than adenocarcinoma, R1 resection, no lymph node was retrieved or histopathological and survival data were incomplete. No patients receive neoadjuvant chemotherapy. In this study, 329 (66.6%) patients received adjuvant chemotherapy. In stage III patients (n = 316), 217 (68.7%) patients received adjuvant chemotherapy, and only 69 patients received more than six cycles of adjuvant chemotherapy. TNM stage was determined according to the 7th edition UICC/AJCC TNM staging system. Overall survival (OS), defined as the duration from operation to GC-related death or last follow-up, was used for prognosis evaluation. The primary endpoint of this study was OS, and patients alive at the last follow-up were recorded as censored events.

Ethics statement
Written informed consent was obtained from the patients with the study protocol approved by the ethics committee of Zhongnan Hospital of Wuhan University. The study was undertaken in accordance with the ethical standards of the World Medical Association Declaration of Helsinki.

Tissue microarrays (TMAs) and immunohistochemistry (IHC)
TMAs have been constructed for this study. Briefly, two cores were taken from each representative tumor tissue and peritumoral tissue (at least 50mm away from the tumor border). Then, sixteen TMAs sections with 494 tumor tissues (988 cores, 2mm each core) and 237 peritumoral tissues (474 cores, 2mm each core) were constructed (in collaboration with Shanghai Biochip Company Ltd., Shanghai, China).
Routine IHC method was performed for the staining of TAMs. The primary antibody was mouse anti-human monoclonal antibody against macrophages (ab22506 [MAC387], Abcam, UK, dilution 1/100), with corresponding horseradish peroxidase (HRP) conjugated secondary antibody (ab97265, Abcam, UK, dilution 1/300). The reaction products were visualized with diaminobenzidine (DAB, DAKO, Denmark). Then the slides were evaluated by two senior pathologists, who were blinded to the patients' clinical features and outcomes. A consensus was achieved using a multi-headed microscope in case of discrepancy. In brief, one TMA core with at least 4 standard-compliant vision fields (magnification, ×200) per patient was considered to be adequate, with no focus on hotspots.

Image acquisition and the classification criteria of TAMs locations and distribution patterns
The digital images of infiltrating macrophages staining were captured under Olympus BX51 fluorescence microscope equipped with Olympus DP72 camera (Olympus Optical Co., Ltd., Tokyo, Japan) at ×200 magnification. Identical settings were used for every photograph, so as to minimize the selection bias.
To assess the role of TAMs in GC progression, the number and the distribution of TAMs were evaluated. First, the positive cells were counted in at least four high power fields (hpfs, ×200 magnification) each core, and the average number of two cores (eight hpfs) from the same patient were recorded as the number of TAMs. Second, TAMs geographic distributions were assessed, aiming to uncover the exact role of TAMs in different areas. Theoretically, each field could be divided into three different regions including cancer nest, cancer invasive front and cancer stromal. Correspondingly, the TAMs in GCs could be classified into three distinct patterns including infiltrating macrophages in cancer nest (MC), infiltrating macrophages in invasive front (MF) and infiltrating macrophages in stromal (MS). Such classification represented the spatial position relationship between macrophages and cancer cells.
Then, the number of MC, MF and MS were recorded, respectively. Briefly, MC+MF+MS=Total TAMs. Notable, MF and MS may be zero. Actually, not each patient has three locations simultaneously, but may have one superiority location. Thus, some patients may have MC only, some patients may have MC and MF, and some patients may have MC, MF and MS simultaneously. In this study, there were 494 patients could be evaluated for MC, 319 patients could be evaluated for both MC and MF, and only 296 patients could be evaluated for all MC, MF and MS. Finally, total of 296 patients containing three locations of TAMs were classified into two distribution patterns, including MC dominant pattern and MF/MS dominant pattern. Investigators were blind to the clinicpathological data and clinical outcomes of GC patients. Cut points of TAMs number for subgroup analysis were explored by the median value.

Statistical analysis
Statistical analyses were performed with IBM SPSS 19.0 software (SPSS Institute, Chicago, IL). The correlations between the number of infiltrating TAMs and clinic-pathological parameters were calculated with the chi-square test. Survival probabilities between subgroups were analyzed with the Kaplan-Meier method, by use of the log-rank test for univariate analyses, and by use of Cox regression model for multivariate analyses. ROC analysis was used to determine the predictive value of the parameters. Two-sided P value of <0.05 was considered to be statistically significant.