High Myeloperoxidase Positive Cell Infiltration in Colorectal Cancer Is an Independent Favorable Prognostic Factor

Background Colorectal cancer (CRC) infiltration by adaptive immune system cells correlates with favorable prognosis. The role of the innate immune system is still debated. Here we addressed the prognostic impact of CRC infiltration by neutrophil granulocytes (NG). Methods A TMA including healthy mucosa and clinically annotated CRC specimens (n = 1491) was stained with MPO and CD15 specific antibodies. MPO+ and CD15+ positive immune cells were counted by three independent observers. Phenotypic profiles of CRC infiltrating MPO+ and CD15+ cells were validated by flow cytometry on cell suspensions derived from enzymatically digested surgical specimens. Survival analysis was performed by splitting randomized data in training and validation subsets. Results MPO+ and CD15+ cell infiltration were significantly correlated (p<0.0001; r = 0.76). However, only high density of MPO+ cell infiltration was associated with significantly improved survival in training (P = 0.038) and validation (P = 0.002) sets. In multivariate analysis including T and N stage, vascular invasion, tumor border configuration and microsatellite instability status, MPO+ cell infiltration proved an independent prognostic marker overall (P = 0.004; HR = 0.65; CI:±0.15) and in both training (P = 0.048) and validation (P = 0.036) sets. Flow-cytometry analysis of CRC cell suspensions derived from clinical specimens showed that while MPO+ cells were largely CD15+/CD66b+, sizeable percentages of CD15+ and CD66b+ cells were MPO−. Conclusions High density MPO+ cell infiltration is a novel independent favorable prognostic factor in CRC.


Introduction
Outgrowth and progression of human colorectal cancers (CRC) are driven by gene mutations and microsatellite instability tumor inherent characteristics [1,2], and by the interaction of cancer cells with microenvironmental stimuli provided by non-transformed cells [3,4]. In particular, cytokine and chemokine environment and infiltration by immunocompetent cells significantly influence CRC outcome [5][6][7][8].
Infiltration by activated CD8+ memory T cells and expression of IFN-c gene within CRC were convincingly shown to be associated with favorable prognosis [5,7]. Furthermore, we and others have shown that FOXP3+ immune cell infiltration independently predicts improved survival in CRC [9,10].
Granulocytes have largely been disregarded by tumor immunologists [18]. However, recent studies, mainly performed in experimental models, suggest that neutrophil granulocytes might prevent metastatic cancer progression [19]. Furthermore, they were suggested to undergo cytokine driven differentiation into N1 and N2 cells endowed with anti-and pro-tumor properties, respectively [20,21]. These findings have led to a resurgent interest in granulocyte infiltration in cancer [22][23][24].
In previous work, we showed that CRC infiltration by CD33+/ HLA-DR2/CD16+ myeloid cells is associated with improved patient survival [13]. Based on these phenotypic features, we hypothesized that CRC could be infiltrated by granulocytes with a favorable prognostic significance.
Myeloperoxidase (MPO) is a lysosomal enzyme produced in high amounts by neutrophilic granulocytes (NG) [25], especially during their early maturation phase. MPO catalyzes the production of hypochlorous acid from hydrogen peroxide and chloride anion and oxidizes tyrosine to tyrosyl radicals. Both hypochlorous acid and tyrosyl radicals are cytotoxic to a variety of microorganisms. Notably, MPO is also involved in the induction of granulocyte apoptosis following activation [26,27].
In a small series of CRC samples (n = 67), it has been shown that MPO+ cell infiltration is significantly higher in CRC than in normal colon mucosa [28]. However, prognostic relevance of CRC infiltration by MPO+ cells has not been addressed so far.
CD15, also known as Lewis X and stage-specific embryonic antigen 1, is a carbohydrate adhesion molecule expressed on mature neutrophils, mediating phagocytosis and chemotaxis [29]. Importantly, CD15 expression has been detected in tumor cells and found to correlate with poor prognosis in head and neck, gastric and lung cancers [30][31][32]. In CRC, expression of CD15 on tumor cells was shown to occur during progression to metastatic stages [33] and to be associated with high incidence of recurrences and poor survival [34,35]. However, the prognostic value of CRC infiltration by CD15+ immune cells has not been explored.
Here we show for the first time that a subgroup of CRC is characterized by a high infiltration by MPO+ and CD15+ positive cells. Most importantly, high MPO+ cell density in CRC is independently associated with favorable prognosis.

Ethics Statement
Written consent has been given from the patients for their information to be stored in the hospital database and used for research. The use of this clinically annotated TMA for research was approved by the corresponding Ethics Committee of the University Hospital of Basel (Ethikkommission beider Basel) and the ex vivo analyses were approved by the Institutional Review Board (63/07). For freshly excised clinical specimens included in this study written consent has been given from the patients undergoing surgical treatment at Basel University Hospital.

Tissue Microarray Construction
The TMA used in this work was constructed by using 1420 nonconsecutive, primary CRCs, as previously described [36]. Briefly, formalin-fixed, paraffin-embedded CRC tissue blocks were obtained. Tissue cylinders with a diameter of 0.6 mm were punched from morphologically representative areas of each donor block and brought into one recipient paraffin block (30625 mm), using a semiautomated tissue arrayer. Each punch was made from the center of the tumor so that each TMA spot consisted of at least a 50% of tumor cells. One core per case was used.

Clinicopathological Features
Clinicopathological data for the 1420 CRC patients included in the TMA were collected retrospectively in a non-stratified and non-matched manner. Annotation included patient age, tumor  diameter, location, pT/pN stage, grade, histologic subtype,  vascular invasion, border configuration, presence of peritumoral  lymphocytic inflammation at the invasive tumor front and diseasespecific survival (table 1). Tumor border configuration and peritumoral lymphocytic inflammation were evaluated according to Jass using the original H&E slides of the resection specimens corresponding to each tissue microarray punch [37]. The number of lymph nodes evaluated ranged between 1 and 61 with mean and median of 12 and 11, respectively. MMR status was evaluated by IHC according to MLH1, MSH2 and MSH6 expression [38]. Based on this analysis, the TMA included 1031 MMR-proficient tumors and 194 MMR-deficient tumors.
Overall survival was defined as primary endpoint. Follow-up data were available for 1379 patients with mean/median and IQR event-free follow-up time of 67.7/68 and 45-97 months.

Flow Cytometry Analyses
Following Institutional Review Board approval (63/07), tissues from surgically removed CRC and adjacent normal mucosa were minced, centrifuged, and resuspended in RPMI 1640 medium supplemented with 5% foetal calf serum, 2 mg/ml collagenase IV, 0.1 mg/ml hyaluronidase V, and 0.2 mg/ml DNAse I (Sigma Aldrich, Basel, Switzerland). Following a 1 hour digestion, cell suspensions were filtered and centrifuged. For phenotypic analysis of surface markers, cells were stained with mAbs for 15 minutes on ice in PBS, washed once with PBS 0.5% FCS, 0.5 M EDTA buffer and fixed in lysis buffer from BD Bioscience (1:10). Samples were then permeabilized in BD fixation/permeabilization buffer. For intracellular staining, an anti-MPO reagent or an isotype matched negative control antibody were added for a 15 min incubation at room temperature. After a PBS wash, cells were suspended in wash buffer and analyzed by flow cytometry using a 2-laser BD FACSCalibur (Becton Dickinson, San Jose, CA). Results were analyzed by Cell Quest (Becton Dickinson, San Jose, CA) and Flow Jo (Tree Star, Ashland, OR) softwares.

Statistical Analysis
Cut-off scores used to classify CRC with low or high MPO+ or CD15+ infiltration were obtained by regression tree analysis, evaluating sensitivity and false positive rate for the discrimination of survivors and non-survivors, on all tumor samples [40]. Specific scores were set at 60 cells/TMA-punch for MPO+ and at 46 cells/ TMA-punch for CD15+ infiltration. Chi-Square or Fisher's Exact tests were used to determine the association of MPO+ and CD15+ Tumor grade Stage IV metastasis (%) 88 (6.5) infiltration and clinicopathological features. Univariate survival analysis was carried out by the Kaplan-Meier method and log rank test. The assumption of proportional hazards was verified for both markers by analyzing the correlation of Schoenfeld residuals and the ranks of individual failure times. Any missing clinicopathological information was assumed to be missing at random. Subsequently, MPO+ and CD15+ cell infiltration data were entered into multivariate Cox regression analysis and hazard ratios (HR) and 95% confidence intervals (CI) were used to determine associations with survival time. The multivariate Cox regression analysis was performed with 975 cases since missing values were excluded from the model. Spearman's rank correlation was used to analyze the correlation between MPO+, CD15+, CD16+, CD68+, CD8+ and FOXP3+ cell infiltration. Statistical analyses were performed using R i386 Version 2.15.2 (http://www.R-project.org). Data reporting was performed according to the REMARK criteria [41].

CRC Infiltration by MPO+ and CD15+ Cells: Detection and Association with Clinicopathological Features
MPO+ and CD15+ cells could be successfully identified in the TMA under investigation by specific mAbs. Figure 1A-D shows representative stainings of CRC with low and high MPO+ or CD15+ cell infiltration.
Out of 1420 CRC individual specimens, MPO expression could be evaluated in 1225 samples including 1031 MMR-proficient and 194 MMR-deficient CRC. CD15 expression was evaluable in 1191 CRC specimens, including 817 MMR proficient and 191 MMR deficient samples. Numbers of samples evaluated for each feature are indicated in absolute numbers and percentages in table 2. Dropouts were due to loss of punches during TMA staining preparation or missing information, and usually accounted for ,15% of data (tables 1-2).
MPO+ and CD15+ infiltration, as evaluated by absolute cell numbers, was significantly higher in MMR deficient than in MMR proficient CRC (median: 30 cells/punch in deficient vs. 21 cells/ punch in proficient CRC P = 0.007 and 9 cells/punch in deficient Regression tree analysis defined cut-off scores for MPO+ and CD15+ CRC infiltrating cells detected in individual punch biopsies (n = 60 and n = 46, respectively) were used to assess clinicopathological correlations.

Prognostic Significance of MPO+ and CD15+ Cell Infiltration in the CRC Microenvironment
Median survival time was 50 and 46 months for patients with high or low MPO+ cell density, respectively. High MPO+ cell infiltration was significantly (P = 0.0003) associated with better prognosis (0.59 HR, 95%CI: 0.45-0.74), as compared to tumors with low MPO+ cell infiltration in univariate Cox regression analysis. Upon splitting of the cohort in a test and a validation set, high score MPO+ cell infiltration was still associated with significantly improved survival (P = 0.038 and P = 0.002, respectively; figure 3A-B). Several randomizations of the overall cohort were tried and all results were found to be comparable.
In univariate analysis survival was also increased in case of high score CD15+ cell infiltration (P = 0.051, figure 3C). A combination analysis however, showed that MPO is the dominant marker associated with improved prognosis, without relevant additive benefit provided by CD15 positivity ( figure 3D).
Most importantly, in multivariate analysis, high score MPO+, but not CD15+, cell infiltration was independently associated with favorable prognosis after adjusting for several known prognostic factors such as age, sex, T stage, N stage, tumor grade, vascular invasion, tumor border configuration and microsatellite stability (P = 0.004; table 3). Also in the two stratified collectives the effect of MPO+ cell infiltration on survival of patients with CRC remained significant (P = 0.048 and P = 0.036 in the testing and validation set, respectively).

Discussion
To the best of our knowledge, this is the first study identifying MPO+ neutrophil granulocyte tumor infiltration as an independent favorable prognostic factor in CRC.
Myeloid cell infiltration is known to promote tumor growth and to be associated with poor prognosis in a variety of human cancers [42]. In particular, tumor associated macrophages have been indicated as obligate partners for tumor progression and metastasis formation [43]. Granulocyte infiltration has also been found to be associated with poor prognosis in different tumors including lung cancers and renal and hepatocellular carcinoma [44][45][46][47]. In this context, CRC might represent an interesting exception [48,49]. Indeed, controversial data have been reported on the prognostic significance of macrophage infiltration in CRC [50][51][52][53][54].
Neutrophil infiltration has been found to be increased in the transition from normal to dysplastic and cancerous mucosa [55]. Furthermore, CRC infiltration by CD66b+ cells has recently been proposed to be associated with adverse prognosis [56].
In previous work we showed that CRC infiltration by CD16+ cells correlates with improved survival [13]. This marker is expressed in neutrophils, in a subset of macrophages and in NK cells [57,58]. Indeed, NK cell infiltration in CRC is negligible [13,14] and devoid of prognostic significance [59]. Since CRC infiltrating CD16+ cells are also CD11b+/CD33+/HLA-DR-, in this study we focused on the analysis of cells expressing MPO and CD15 granulocyte markers.
In univariate analysis high density CRC infiltration by cells expressing either marker was associated with improved survival. However, following adjustment for multiple comparisons carried out to compensate for the exploratory nature of this analysis this favorable prognostic relevance was maintained for MPO+ but not for CD15+ cell infiltration.
Importantly, in accord with a previously published report [55] we observed that MPO+ and CD15+ cells preferentially colonized CRC tissues while they poorly infiltrated normal colon mucosa, suggesting that they might be specifically recruited to the tumor microenvironment. Interestingly, MPO+ cell infiltration was higher in MMR-deficient than in MMR-proficient CRC as previously suggested by a study conducted with 67 samples from a limited number of cancers (n = 35) [28].
Flow-cytometric analysis of digested paired normal and cancerous tissues indicates that in both healthy mucosa and CRC infiltrating MPO+ cells are CD15+/CD16+/CD66b+/ HLA-DR-, consistent with a granulocytic lineage. Most importantly however, we observed substantial percentages of CD66b+/ MPO-cells infiltrating CRC. These data might explain the discrepancies between our findings and a previous report focusing on CD66b+ CRC infiltrating cells [56]. Notably, neutrophils with similar phenotypic characteristics have also been found to infiltrate head and neck cancers [60].
Similarly, the presence of CD15+/MPO2 cells in healthy mucosa and cancerous tissues might explain the differential prognostic relevance of these markers. Notably however, MPO gene expression was undetectable in CRC or normal mucosa specimens (data not shown) consistent with a mature granulocyte nature of MPO+ infiltrating cells [61].
MPO activity has been suggested to contribute to the pathogenesis of degenerative diseases, including atherosclerosis, multiple sclerosis and Alzheimer disease [62]. Furthermore, high MPO activity or MPO+ cell infiltration have been detected in esophageal [63], and gynecological cancers [64,65] and in CRC [28,66,67], but their prognostic impact was not analyzed.
Despite early reports documenting their ability to mediate tumor cell cytotoxicity [68] granulocytes have largely been neglected by tumor immunologists [18]. However experimental models in the past have indicated that colorectal cancer cells transfected with G-CSF gene can be rejected by tumor infiltrating neutrophils upon interaction with IFN-c producing T cells [69]. Granulocytes were also shown to express co-stimulatory molecules and to be able to present antigens [70,71], thus suggesting the possibility of a role in the initiation of antigen specific antitumor responses.
More recently, evidence of the ability of granulocytes to inhibit lung metastasis formation in an experimental breast cancer model was provided [19]. Furthermore, the capacity of granulocytes to undergo TGF-b dependent polarization into N1 and N2 functional profiles, characterized by anti-or pro-tumoral properties, respectively, similarly to macrophages, has also been documented [20].
The molecular background underlying CRC infiltration by MPO+ cells and its prognostic significance is unclear. These phenomena could reflect chemokine production by activated T cells, and therefore indirectly result from ongoing antitumoral adaptive responses. Alternatively, they might be related to the production of granulocyte attracting chemokines by tumor cells. At least CXCL8 (IL-8) is known to be produced by CRC cells. However, its production was suggested to be associated with increased angiogenesis and tumor dissemination [72]. On the other hand, we and others have previously shown that GM-CSF, promoting granulocyte maturation and survival, can also be produced by CRC cells [73,74].
Our results contribute to the characterization of the complex features inherent in gut microenvironment and with CRCimmune system interaction [75]. Further research is warranted to clarify molecular mechanisms underlying the independent prognostic impact of MPO+ cells in CRC. Importantly, here we show that CRC infiltrating MPO+ cells express CD16 Fcc intermediate affinity receptor. The ability of granulocytes to mediate antibody dependent cellular cytotoxicity (ADCC) is debated. However, the availability of novel therapeutic mAb with glycoengineered Fc fragments characterized by increased affinity for Fcc receptors [76] might lead to a reevaluation of the effector significance of granulocytes. Within this framework, it is tempting to speculate that neutrophil infiltration should be included in current prognostic models for CRC [77] and that it might represent an important novel stratification factor for randomization in specific clinical trials.