The Interplay Between Neutrophils and CD8⁺ T Cells Improves Survival in Human Colorectal Cancer

Granulocytes account for 50%–70% of leukocytes in humans. They represent a first-line defense against bacterial and fungal infections (1, 2). However, clinical and prognostic relevance of granulocyte infiltration in human cancers is debated (1–3). A number of studies suggest that high granulocyte/lymphocyte ratios in peripheral blood are associated with poor prognosis in different malignancies (4). Furthermore, myeloid cells of the granulocytic lineage at different maturation stages were shown to represent sizeable subsets of myeloid-derived suppressor cells (MDSC), promoting tumor growth and inhibiting cancer-specific adaptive responses (5).

More recently, the possibility that neutrophils might promote antitumor immune responses of clinical relevance has started to be explored (2). In particular, the ability of neutrophils to polarize into N1 and N2 functional profiles, similarly to macrophages, has been documented in experimental models (6, 7). Furthermore, tumor “educated” neutrophils were shown to elicit antimetastatic effects (8), and interaction of hepatocyte growth factor (HGF) with its receptor MET was suggested to play a key role in the recruitment of neutrophils mediating antitumor activities (9). Earlier studies indicated that production by tumor cells of G-CSF or granulocyte-macrophage colony-stimulating factor (GM-CSF), promoting neutrophil survival and activation, could induce adaptive antitumor immune responses and regression of established tumors based on neutrophil–T-cell interaction (10, 11). Furthermore, the ability of early-stage lung cancer–infiltrating neutrophils to support T-cell proliferation and antitumor responses has been demonstrated (12, 13). However, their prognostic significance was not addressed.

Colorectal cancer represents the third cause of cancer-related mortality worldwide. Tumor–node–metastasis (TNM) staging, routinely used to identify patients eligible for different treatments is frequently ineffective in predicting colorectal cancer clinical course (14).

Clinical relevance of the composition of tumor infiltrate in colorectal cancer has been extensively investigated. Colorectal cancer infiltration by CD8⁺ and memory T cells has been consistently associated with favorable prognosis (15, 16). The specificity of these cells is largely unclear. Recognition of differentiation antigens or neoantigens (17) expressed by tumor cells was reported. Alternatively, bystander effects related to T-cell responses against antigens from microbial commensal could also be hypothesized. Interestingly, expression of activation markers by colorectal cancer–infiltrating lymphocytes correlates with prolonged survival (18).

The role of the innate immune system is unclear. NK-cell infiltration is modest and apparently devoid of prognostic significance (19). Although tumor infiltration by myeloid cells is associated with poor prognosis in a variety of cancers (20), macrophage infiltration has been suggested to correlate with favorable prognosis in colorectal cancer (21).

The role of neutrophils has not been comparably explored. We previously observed that colorectal cancer infiltration by CD16⁺ myeloid cells correlates with favorable outcome (22). Similarly to neutrophils, these cells are HLA-class II and largely myeloperoxidase (MPO)+ (23). Data from other groups suggest that high tumor infiltration by CD66b⁺ neutrophils may correlate with either benign or poor prognosis in patients with colorectal cancer. In a cohort of East Asian patients (n = 229), neutrophil infiltration was associated with severe prognosis (24). Moreover, neutrophil infiltration in colorectal cancer–derived lung metastases has been suggested to be associated with severe prognosis following surgical excision (25). However, neutrophil infiltration in colorectal cancer was reported to be associated with responsiveness to 5-fluorouracil (5-FU) treatment (26). Thus, clinical significance of tumor-associated neutrophils (TAN)-infiltrating colorectal cancer is still unclear and underlying functional mechanisms remain to be elucidated.

We have analyzed the prognostic significance of colorectal cancer–infiltrating CD66b⁺ neutrophils by using a clinically annotated tissue microarray (TMA) including >650 cases. In addition, we have comparatively evaluated the phenotypes of neutrophils from healthy and cancerous colon tissues and peripheral blood from patients and healthy donors (HD). Their ability to support adaptive immune responses was specifically addressed. Finally, the prognostic relevance of the association of neutrophils with CD8⁺ T cells infiltrating colorectal cancer microenvironment was explored.

The TMA used in this work was constructed using >650 nonconsecutive, formalin-fixed, and paraffin-embedded primary colorectal cancer samples, from the tissue BioBank of the Institute of Pathology of the University Hospital Basel (Switzerland; refs. 18, 22). A semiautomated tissue arrayer was used to transfer punches of a 0.6-mm diameter from tissue blocks onto glass slides. Punches derived from tumor centers and consisted of at least 50% tumor cells. Clinical–pathologic data for patients included in the TMA are summarized in Supplementary Tables S1 and S2. Use of clinical information was approved by the local ethical authorities.

TMA slides were incubated with primary antibodies specific for CD8, CD16, MPO (18, 22, 23), and CD66b (clone G10F5, Biolegend). Secondary staining and negative controls were performed as described previously (18, 22, 23). Colorectal cancer infiltration by cells expressing defined markers was scored by experienced pathologists.

Established human colorectal cancer cell lines (DLD1, HCT116, SW480, HT29, and SW620) were purchased from European Collection of Authenticated Cell Cultures (ECACC). DLD1 and HCT116 were cultured in RPMI1640 supplemented with 10% FBS, GlutaMAX-I, nonessential amino acids (NEAA), 100 mmol/L sodium pyruvate, 10 mmol/L HEPES (all from Gibco). HT29 cells were cultured in McCoy 5A medium (Sigma-Aldrich), supplemented with 10% FBS, GlutaMAX-I, and kanamycin (Gibco). SW480 and SW620 cells were cultured in L15 medium (Sigma-Aldrich), supplemented with 10% FBS, GlutaMAX-I, and kanamycin (Gibco). Absence of mycoplasma contamination was verified by PCR, prior to experimental procedures.

Clinical specimens from consenting patients undergoing surgical treatment at Basel University Hospital St. Claraspital, Basel, Switzerland, and Ospedale Civico (Lugano, Switzerland), were obtained according to procedures approved by local ethical commissions. Tumor tissues and corresponding tumor-free mucosa fragments were embedded in optimal cutting temperature compound for further histologic evaluation or enzymatically digested by using an enzyme cocktail including 200 U/mL collagenase IV (Worthington Biochemical Corporation) and 0.2 mg/mL DNAse I (Sigma-Aldrich for 1 hour at 37°C) to obtain single-cell suspensions, as described previously (27).

TANs were isolated from tumor cell suspensions by positive selection of CD66b⁺ cells with antibody-coated microbeads according to the manufacturer's instructions (Miltenyi Biotec, code: 130-104-913). Heparinized peripheral blood was collected from patients with colorectal cancer prior to surgery or from HDs, and density-gradient centrifugation was performed. Sedimented fractions containing high-density neutrophils were washed and treated with dextran 4% (T500, Pharmacia) in saline solution and residual erythrocytes in supernatants were lysed by using lysis buffer (Miltenyi Biotec). Peripheral blood neutrophils (PBN) were further enriched by positively removing contaminating cells, using the EasySep Human Neutrophil Enrichment Kit (StemCell Technologies). Purity of isolated PBN and TAN was evaluated by flow cytometry upon staining for the neutrophils/myeloid markers CD66b, CD16, myeloperoxidase (MPO), and CD11b and exceeded 98% and 80%, respectively, in cells used in functional assays. Average percentages of apoptotic cells in PBN and TAN suspensions used in functional assays, as measured by Annexin V/PI staining (BioLegend), did not exceed 5% and 20%, respectively. Peripheral blood and tumor-infiltrating CD8⁺ lymphocytes (TIL) were isolated from peripheral blood mononuclear cells (PBMC) obtained by gradient centrifugation or digested tumor specimens, respectively, by using anti CD8-coated magnetic beads (Miltenyi Biotec), as described previously (28).

Colorectal cancer cells from established cell lines were cultured in the presence or absence of neutrophils untreated or previously treated for 1 hour with IFNγ or fMLP (Sigma) at different ratios and tumor cell proliferation was assessed by ³H-Thymidine incorporation (³H-TdR). In specific experiments, induction of apoptosis in tumor cells was tested by Annexin V/PI staining.

Cell suspensions from colorectal cancers and tumor-free mucosa, and peripheral blood of HDs or patients with colorectal cancer, were stained with fluorochrome-conjugated antibodies specific for human CD66b, CD16, CD11b (BioLegend) and CD54, CXCR1, CXCR2 (BD Biosciences). Alternatively, cells were fixed and intracellular staining was performed with antibodies specific for MPO (23). Stained cells were analyzed by FACSCalibur flow cytometer (BD Biosciences), using FlowJo software (Tree Star).

Following CD66b, CD16, and intracellular MPO-specific staining, cells were washed and resuspended in PBS supplemented with 0.5% FBS and 5 mmol/L EDTA, prior to processing through ImageStream, Mark II Imaging Flow Cytometer (Amnis, EMD Millipore). Analysis was performed using IDEAS software (Amnis, EMD Millipore) and neutrophils from colorectal cancer tissue, healthy mucosa, and peripheral blood were identified on the basis of brightfield morphology, granularity, and CD66b expression.

PBNs and TANs, obtained from HD and colorectal cancer patients, were cocultured with CD8⁺ T cells from autologous peripheral blood or tumor specimens. For costimulation experiments, 96-well flat-bottom culture plates were coated overnight with anti-CD3 mitogenic mAb (TR66, a gift of Dr. Lanzavecchia, Bellinzona, Switzerland), or UCHT-1 (eBiosciences), at suboptimal concentrations ranging between 0.5 and 5 μg/mL depending from hybridoma and lot. Neutrophils and CD8⁺ T cells at a 0.5 10⁶/mL concentration each were cultured in RPMI1640 medium supplemented with GlutaMAX I, HEPES, sodium pyruvate, nonessential amino acids, antibiotics (all from Gibco), and 5% AB serum (Blood Bank, Kantonsspital, Basel, Switzerland), thereafter referred to as complete medium, in the presence of anti-CD28 (1 μg/mL, BD Pharmingen). After 24-hour incubation, expression of CD69 early T-cell activation marker was evaluated by flow cytometry. T-cell proliferation was measured by assessing carboxyfluorescein succinimidyl ester (CFSE, Invitrogen) dilution in labeled CD8⁺ T cells following 72-hour culture by flow cytometry (28). IFNγ release in culture supernatants was assessed by using commercial ELISA Kits (BD Biosciences). When indicated, coculture experiments were performed by using Transwell plates (Corning), or in the presence of anti-CD11a (BioLegend) or control reagents.

Colorectal cancer sections were fixed with formalin 4% for 15 minutes at room temperature and blocked with 2% goat serum diluted in PBS containing 0.3% Triton X-100 for 1 hour at room temperature. They were then incubated with rabbit polyclonal anti human CD8 (Abcam) or rabbit polyclonal anti human CD45RO (BiorByt) and mouse monoclonal anti human CD66b (Biolegend) for 1 hour at 37°C. Slides were washed with PBS and incubated for 1 hour at room temperature with goat anti-mouse Alexa Fluor 488 and anti-rabbit 546–conjugated antibodies (Invitrogen). Nuclei were counterstained with 4,6-diamidino-2-phenyldole (DAPI, Invitrogen). Sections were examined using Olympus BX61 fluorescence microscope (Olympus) and images were captured with 10× and 20× magnification using an F-View II camera (Olympus) and AnalySiS software (Soft Imaging System GmbH).

Statistical significance of differential expression of activation markers, cytokine release, and cell proliferation was analyzed by Student t and Wilcoxon/Mann–Whitney tests, as appropriate.

Associations with survival were explored using the Cox proportional hazards regression model. Cut-off values used to classify colorectal cancer with low or high immune cell infiltration were obtained by regression tree analysis (rpart package). On the basis of this calculation and on the test evaluability, threshold value for CD66b⁺ infiltration was set at 10 cells per punch. After dichotomization, Kaplan–Meier curves were plotted, and compared by log-rank test.

Kruskal–Wallis and Jonckheere–Terpstra tests were used to determine the association of CD66b⁺ and CD8⁺ cell infiltration and clinical–pathologic features depending on continuous or discrete nature of the variable. Any missing clinical–pathologic information was assumed to be missing at random. Subsequently, CD66b⁺ and CD8⁺ cell infiltration data were entered into multivariate Cox regression analysis with clinical–pathologic variables and HRs and 95% confidence intervals (CI) were used to determine prognostic effects on survival time.

Data were analyzed using the Statistical Package Software R (Version 3.2.4, www.r-project.org). P values <0.05 were considered statistically significant.

Ethikkommission Nordwest- und Zentralschweiz (EKNZ), no. 2014-388.

In previous studies, we had observed that colorectal cancer infiltration by MPO⁺ cells is associated with favorable prognosis (23). However, this enzyme is produced by different cells of the myeloid lineage. Therefore, to precisely identify TANs in colorectal cancer, we stained a TMA including >650 colorectal cancer with a mAb recognizing CD66b, a classical neutrophil marker (26). Colorectal cancer–infiltrating CD66b⁺ cells could be efficiently detected in punches from fixed, paraffin-embedded tissues (Fig. 1A and B). In this cohort of patients, high colorectal cancer infiltration by CD66b⁺ cells, as dichotomized by using a cut-off value (n = 10) obtained by regression tree analysis (see “Materials and Methods”), was associated with increased overall survival (OS; Fig. 1C, P = 0.0001). Similar results were observed when CD66b⁺ cell infiltration was analyzed by dichotomizing data using median (n = 5) or mean (n = 16.5) values as cutoff (P = 0.0003 and 0.001, respectively) or by using nondichotomized continuous log 10–transformed CD66b⁺ cell infiltration values (P < 0.0001). Interestingly, high-density CD66b⁺ cell infiltration was significantly associated with pT1–2 stage (P = 0.027), pN0 stage (P = 0.001), clinical stage (P < 0.0001), absence of vascular invasion (P = 0.005), and “pushing” (18) tumor borders (P = 0.028; Supplementary Table S1).

Prompted by data supporting their prognostic significance, we investigated phenotypic profiles of neutrophils infiltrating colorectal cancer, adjacent healthy mucosa, and autologous peripheral blood.

Colorectal cancer were characterized by a significantly (P = 0.0009) higher infiltration by CD66b⁺ cells, as compared with autologous healthy mucosa (Fig. 2A), although wide variations were observed from patient to patient. Imagestream analysis indicated that TAN and mucosa-associated neutrophils included CD66b⁺ cells with variable expression of MPO and CD16 (Fig. 2B).

To obtain accurate quantitative data, phenotypic profiles of tissue-infiltrating neutrophil were analyzed by flow cytometry in comparison to PBNs. Representative examples are shown in Fig. 2C. In keeping with previous reports (22, 23), we observed that sizeable percentages of tissue-infiltrating CD66b⁺ neutrophils express MPO and CD16 to lower extents as compared with autologous PBN (Fig. 2C and D). On the basis of this background, and considering that MPO and CD16 are also expressed by cells other than neutrophils, we explored the relative prognostic significance of the expression of these markers in the TMA under investigation. We observed that colorectal cancer infiltration by CD66b⁺ cells is associated with improved OS also in the absence of a concomitantly high CD16⁺ or MPO⁺ cell infiltration (Fig. 2E and F). Furthermore, in the presence of a high CD66b⁺ cell infiltration, presence or absence of concomitant CD16⁺ or MPO⁺ high-density infiltration did not significantly modify survival curves (P = 0.75 and 0.79, respectively).

Expression of other markers was also investigated. CD66b and CD11b are expressed in similarly high percentages of TANs and PBNs (data not shown), whereas CXCR1 and CXCR2 are expressed to lower extents in TANs, as compared with autologous PBNs (Fig. 2G). This phenotypic profile is shared by neutrophils infiltrating adjacent autologous healthy mucosa (Fig. 2D and G). However, higher percentages of TANs express CD54, as compared with autologous PBNs and mucosa-infiltrating neutrophils (Fig. 2G).

Notably, phenotypic profiles of PBNs from patients with colorectal cancer and HD are similar (Supplementary Fig. S1).

Data from TMA analysis consistent with an antitumor potential of colorectal cancer infiltration by neutrophils prompted us to explore possible mechanisms of action. Direct effects on colorectal cancer cells were first considered (29). However, short life span, and relatively low numbers of cells recovered from clinical specimens hampered routine use of TANs in these functional assays. Therefore, these experiments were performed by using PBNs from patients with colorectal cancer and HDs. Coculture in the presence of granulocytes did not decrease proliferation (Supplementary Fig. S2A and S2B) nor induced apoptosis (data not shown) in a panel of colorectal cancer cell lines. Prior treatment of neutrophils with IFNγ or fMLP also failed to impact on viability and proliferation potential of cocultured colorectal cancer cells (Supplementary Fig. S2C).

Alternative mechanisms of action underlying favorable prognostic significance of TANs in colorectal cancer might be related to their ability to exert indirect antitumor effects, mediated by other cell subsets. Colorectal cancer infiltration by CD8⁺ T cells has widely been reported to associate with favorable prognosis (15), although their antigen specificity is largely unclear. Cytokines released by activated T cells, including GM-CSF and IFNγ, are able to activate neutrophils and to prolong their survival (30). More recently, neutrophils infiltrating early-stage lung cancers, but not their peripheral blood counterpart, were shown to promote T-cell response to anti-CD3 triggering (12, 13).

Initial studies suggested that CD66b⁺ granulocytes frequently colocalize with CD8⁺ and CD45RO⁺ T lymphocytes within tumor tissues (Fig. 3A and B). On the basis of these findings, we tested the ability of TANs derived from enzyme-digested colorectal cancer specimens to modulate responses of autologous peripheral blood CD8⁺ T cells to anti-CD3 triggering. Upon addition of TANs to CD8⁺ lymphocyte cultures, a significantly (P = 0.006) increased expression of CD69 early activation marker induced by suboptimal concentrations of anti-CD3 mAb in the presence of anti CD28 mAb was observed. Furthermore, importantly, IFNγ release in these cultures was also significantly enhanced (P = 0.01; Fig. 3C). Consistent with data from experiments with TANs, we observed that interaction with PBNs from patients (Fig. 3D) and HD (Fig. 3E) resulted in significant increases in CD69 expression and IFNγ release by autologous CD8⁺ lymphocytes upon stimulation with suboptimal concentrations of anti-CD3 mAb and anti-CD28 mAb. Representative flow cytometry plots are reported in Supplementary Fig. S3B and cumulative data are reported in Fig. 3D and E. T-cell proliferation, as assessed by CFSE dilution at 72 hours, was also significantly enhanced. Representative flow cytometry profiles are shown in Supplementary Fig. S3D, whereas cumulative data are reported in Fig. 3E. In contrast, these costimulatory effects were undetectable in T cells activated with optimal mitogenic concentrations of anti-CD3 mAb (Supplementary Fig. S3A).

Neutrophil-mediated costimulation critically required cell-to-cell contact as it was not observed in experiments performed in Transwell plates (Fig. 4A; Supplementary Fig. S3B). Furthermore, blocking of CD11a on CD8⁺ T cells, preventing binding to CD54/ICAM-1 expressed by neutrophils, significantly (P = 0.015) inhibited elicitation of costimulatory functions (Fig. 4B; Supplementary Fig. S3C). Notably, CD54/ICAM-1 expression appeared to be upregulated in neutrophils upon culture in the presence of resting and activated CD8⁺ T cells (Fig. 4C). Furthermore, coculture with activated CD8⁺ T cells improved neutrophil viability (Fig. 4D).

These data indicate that untreated TAN and PBN are able to costimulate CD8⁺ T cells, and that these effects are detectable in suboptimal activation conditions.

Favorable prognosis in colorectal cancer has been repeatedly associated with tumor infiltration by “memory” T lymphocytes (15, 16). To further characterize neutrophil-mediated costimulation of CD8⁺ T cells, we evaluated phenotypic profiles of lymphocytes activated by optimal anti-CD3 concentrations and CD28 in the presence or absence of granulocytes for 5 days, thus, beyond time points usually considered for detection of maximal proliferation (31). Remarkably, peripheral blood CD8⁺ T-cell stimulation in the presence of PBN resulted in significantly increased percentages of “central” memory cells expressing a CD45RO⁺/CD62L⁺ phenotype. Representative examples and cumulative data are reported in Fig. 5A. However, in the presence of TAN, these effects were only detectable in four of seven experiments performed with cells from different patients (data not shown).

Similar experiments were also performed by using TAN and autologous tumor-derived CD8⁺ TIL. These cells are characterized by phenotypic profiles different from autologous peripheral blood CD8⁺ T cells. In particular, significantly higher percentages of CD8⁺ TIL, as compared with peripheral blood CD8⁺ T cells, do express CD69 (71.9% ± 19.1% vs. 1.6% ± 0.6%, n = 4, P = 0.008) or CD45RO (64.5% ± 25.7% vs. 22.5 ± 10%, n = 4, P = 0.02), consistent with a locally “activated” state. However, we observed that the percentage of CD62L⁺ cells was markedly increased in anti-CD3/CD28–stimulated cultures performed in the presence of autologous TAN, as compared with cultures performed in their absence, in CD8⁺ TILs derived from three out of four tumors tested, whereas it was identical in a fourth. Accordingly, IFNγ release upon anti CD3/CD28 stimulation in cultures of CD8⁺ TIL from two out of three different tumor specimens tested was higher in the presence than in the absence of autologous TAN. Representative data and cumulative results are shown in Fig. 5B and C.

“In vitro” mechanistic results urged us to analyze potential prognostic significance of combined colorectal cancer infiltration by neutrophils and CD8⁺ T cells. Colorectal cancer infiltration by CD66b⁺ cells was characterized by weak, but significant, correlation with CD8⁺ T-cell infiltration (P < 0.001). These findings prompted us to investigate the prognostic significance of combined colorectal cancer infiltration by both CD66b⁺ neutrophils and CD8⁺ T cells. In our cohort (Supplementary Table S2), 50% of the tumors (325/652) were characterized by poor CD8⁺ and CD66b⁺ cell infiltration. Although 39% (259/652) and 23% (149/652) of cases showed evidence of high CD66b⁺ or CD8⁺ T-cell infiltration, respectively, a concomitantly high CD66b⁺ and CD8⁺ infiltrate was detectable in 12% of colorectal cancer samples (81/652). Colorectal cancer samples infiltrated by both CD66b⁺ and CD8⁺ cells displayed favorable prognosis, whereas cancers with low CD66b⁺ and CD8⁺ cell infiltration were characterized by poor prognosis (Fig. 6A, P < 0.0001). Most interestingly, the favorable prognostic significance of CD8⁺ colorectal cancer infiltration was significantly (P = 0.011) enhanced by a concomitant infiltration by CD66b⁺ neutrophils (Fig. 6). Accordingly, colorectal cancer samples with concomitant high infiltration by CD66b⁺ and CD8⁺ T cells were more frequently characterized by pN0 stage (absence of nodal metastases P = 0.03), and a more frequent “pushing” tumor border (P = 0.038; Supplementary Table S2).

Several models with additive inclusion of single clinical–pathologic data were also tested. Age and gender of the patients and pT or pN stages of the tumors did not affect the significant prognostic impact of CD66b⁺ and CD8⁺ cell infiltration. However, when vascular invasion or invasive margins were added to the model, colorectal cancer infiltration by CD66b⁺ and CD8⁺ cells lost its independent prognostic value (data not shown).

The role of neutrophils in tumor immunobiology and the prognostic significance of neutrophil infiltration in cancer tissues are controversial (1, 2). Early studies have documented a direct cytotoxic potential of neutrophils against defined tumor cell lines (29). However, granulocytes have also been shown to actively contribute to the generation of microenvironmental conditions favoring tumor growth, particularly in cancers associated with chronic inflammation. Their ability to degrade extracellular matrix and to promote angiogenesis has been shown to play critical roles in tumor progression (32). More recently, neutrophils were shown to accumulate in premetastatic niches with pro- or antitumor functions in different experimental models (33, 34). Importantly, the ability of neutrophils at different maturation stages to suppress immune responses has been clearly demonstrated in experimental model (35, 36). However, the phenotypic and functional characterization of human MDSC of the granulocytic lineage has not been elucidated in comparable detail (37). Recent studies suggest that local microenvironmental conditions might result in the polarization of neutrophils toward pro- or antitumor functional states (38), possibly characterized by different physical and functional profiles (36). Remarkably, depending on anatomical locations and histologic origins, human cancers may be characterized by highly diverse microenvironmental conditions, potentially impacting on the clinical significance of granulocyte infiltration.

In this study, we report that the analysis of a large clinically annotated TMA including over 600 colorectal cancer reveals that CD66b⁺ cell infiltration is associated with favorable prognosis. CD66b is expressed by neutrophils and eosinophils. However, in keeping with previously published data (39), we observed that >90% of CD66b⁺ colorectal cancer–infiltrating cells were neutrophils. Although colorectal cancer appears to be infiltrated to a larger extent than autologous adjacent healthy tissue, phenotypic profiles of colorectal cancer–infiltrating neutrophils largely match those detectable in healthy mucosa–infiltrating cells. However, in agreement with data regarding early lung cancer–infiltrating neutrophils (12, 13), TANs appear to express CD54 to higher extents as compared with neutrophils infiltrating autologous adjacent healthy mucosa, consistent with an “activated” phenotypic profile.

Mechanisms potentially underlying the favorable prognostic significance of colorectal cancer infiltration by CD66b⁺ cells were investigated in detail. Our results indicate that coculture with autologous neutrophils enhances T-cell receptor–triggered activation of CD8⁺ T cells and may promote the expansion of a lymphocyte subset characterized by the expression of “memory” markers. The relevance of these findings to colorectal cancer immunobiology is indirectly supported by the colocalization of neutrophils and both CD8⁺ and CD45RO⁺ T cells in colorectal cancer tissues. Furthermore, most importantly, colorectal cancer concomitantly infiltrated by neutrophils and CD8⁺ T cells are characterized by a significantly more favorable prognosis, as compared with tumors displaying high CD8⁺ but low CD66b⁺ cell infiltration. High neutrophil infiltration “per se,” in the absence of concomitant CD8⁺ lymphocytes, provided a more modest, but still highly significant, prognostic advantage as compared with colorectal cancer where neither high-density CD8⁺ nor CD66b⁺ infiltrates were detectable. Therefore, absence of lymphocyte infiltration or infiltration by “unfavorable” T cells might prevent a full elicitation of antitumor effects of neutrophils.

These data may suggest that the favorable prognostic significance of infiltration by neutrophils in colorectal cancer might, at least in part rely on their interaction with CD8⁺ T cells, possibly based on costimulatory mechanisms, as suggested by our “in vitro” results. Admittedly, literature reports based on “in vitro” studies with human cells in this area are controversial (30, 40, 41). However, in our experience, different techniques, and, in particular, the use of human, as opposed to calf serum, might largely account for the observed discrepancies.

The potential relevance of these results might obviously extend beyond colorectal cancer immunobiology. Neutrophil–CD8⁺ T-cell interactions might indeed be operational in a wider range of conditions, thus supporting the notion of a highly effective cooperation between innate and adaptive immune responses.

Remarkably, similar results have also emerged from studies conducted in experimental models (42) and in clinical settings, including early-stage lung cancer (12, 13) and autoimmune and infectious disease (31). Underlying molecular mechanisms have not been fully clarified. In particular, cell–cell interactions mediated by OX40, CD58, CD59, and their ligands have been proposed (12, 31). Alternatively, a role for ROS release has also been suggested (42). Our data suggest that CD11a/CD54 interaction powerfully contributes to the elicitation of the costimulatory effects of neutrophils on CD8+ T-cell activation. Nevertheless, further research is warranted to obtain additional mechanistic insights.

Overall, an important limitation in studies on TANs is represented by their short life span, high sensitivity to enzymes necessary for the generation of single-cell suspensions from clinical specimens and relatively low numbers, preventing the routine performance of functional studies. However, the consistency of “in vitro” results data emerging by using TANs, and PBNs from patients with colorectal cancer and HDs together with the clinical data, appears to support the notion of a potentially high relevance of neutrophil–T-cell interaction on tumor sites.

Our study suffers from a number of additional limitations largely inherent in research requiring the use of clinical materials. Reported clinical data stem from a retrospective study. However, performance of prospective studies, currently being planned, is delayed by overall survival rates in patients with colorectal cancer frequently exceeding 50% at 5 years following surgery. Furthermore, repeated biopsies of metastatic sites are usually not included in routine clinical procedures. Therefore, the performance of longitudinal studies addressing the role of neutrophils in different stages of tumor progression is problematic. Finally, numbers of neutrophils and CD8⁺ T cells which may be obtained from freshly excised colorectal cancer are usually modest and barely amenable to standard cellular immunology assays, particularly in autologous settings. For instance, direct tumor cytotoxicity studies could only performed with resting or activated peripheral blood neutrophils from patients and HDs and we were unable to separate sufficient numbers of high/normal and low-density granulocytes (36) from tumor suspensions.

Remarkably, our results also appear to suggest functional discrepancies between PBN and autologous TAN as regarding, for instance, the capacity of expanding “central memory” cells. These data urge further research aimed at clarifying whether these discrepancies are due to functional impairments of TAN in specific tumor microenvironments (7) or to tumor infiltration by CD66⁺ cell subpopulations of different functional significance (36, 37).

Our study also poses a number of additional important questions. The fact that in a variety of tumors other than colorectal cancer, neutrophil infiltration has been suggested to be associated with poor prognosis (35, 43, 44) raises the issue of the specificities inherent in neutrophil infiltration in colorectal cancer. The microenvironment of these tumors presents a variety of peculiar characteristics. Similar to other cancers (45), colorectal cancer infiltration by “memory” CD8⁺ T cells has been shown to be associated with favorable prognosis. However, tumor infiltration by cells expressing FOXP3, a classical regulatory T-cell marker, also appears to correlate with a less severe clinical course (46). Furthermore, at difference with tumors of different histologic origin (20) colorectal cancer infiltration by macrophages is also associated with favorable prognosis (21). Most remarkably, colorectal cancer cells have previously been shown (47) to produce GM-CSF, possibly enhancing viability and functions of granulocytes eventually recruited within tumor microenvironment.

Data regarding neutrophil infiltration in colorectal cancer from East Asia patients (24) appear to contradict our results. Genetic background may play an important role in this context. However, additional factors might be involved in determining the prognostic relevance of neutrophil infiltration. Colorectal cancer oncogenesis is typically characterized by an early increase in bacterial translocation from the gut lumen (48) and gut microbiome composition has recently been shown to decisively impact on chemo- and immunotherapy outcome (49). Considering the key role of neutrophils in the response to bacteria, it is tempting to speculate that microbiome composition might decisively affect their ability to actively participate to antitumor immune response. Thus, differences in human gut microbiome in different geographic areas (50) might also be of importance in the evaluation of the prognostic significance of neutrophil infiltration in colorectal cancer. Within this context, our data also urge studies aimed at the clarification of mechanisms favoring the recruitment of granulocytes within colorectal cancer tissues.

In conclusion, our study, providing clear evidence of the prognostic significance of concomitant infiltration by CD8⁺ cells and neutrophils in colorectal cancer, represents an additional example of clinically relevant interaction between noncancerous cells from the tumor microenvironment. Furthermore, it clearly identifies neutrophils as key players in colorectal cancer immunobiology.

L. Tornillo reports receiving speakers bureau honoraria from AMGEN Switzerland AG. No potential conflicts of interest were disclosed by the other authors.

Conception and design: V. Governa, V. Mele, L. Tornillo, R. Rosso, G. Iezzi, G.C. Spagnoli

Development of methodology: V. Governa, G.C. Spagnoli

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): V. Governa, E. Trella, L. Tornillo, F. Amicarella, E. Cremonesi, M.G. Muraro, H. Xu, R Droeser, S.R. Däster, M. Bolli, R. Rosso, L. Terracciano

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): V. Governa, E. Trella, V. Mele, L. Tornillo, E. Cremonesi, M.G. Muraro, S. Eppenberger-Castori, L. Terracciano, G. Iezzi, G.C. Spagnoli

Writing, review, and/or revision of the manuscript: V. Governa, V. Mele, F. Amicarella, R. Droeser, S.R. Däster, D. Oertli, S. Eppenberger-Castori, G. Iezzi, G.C. Spagnoli

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): V. Governa, R. Droeser, S.R. Däster, G.C. Spagnoli

Study supervision: D. Oertli, G.C. Spagnoli

We thank Marco Cassatella for advice on neutrophils purification and activation techniques, Elisabetta Padovan, Luca Quagliata, Jean Pieters, Alfred Zippelius, Nina Khanna, Pascal Forrer, Marco Lepore, Christian Hirt, and Paul Zajac for critical discussions.

This work was partially funded by an SNF grant (to G.C. Spagnoli), an SNF Professorship grant (to G. Iezzi), and a grant from Swiss Cancer League to (R.A. Droeser). V. Governa was funded by a grant from the Swiss Centre for Applied Human Toxicology (SCAHT) and Sophienstiftung zur Förderung der klinischen Krebsforschung, Zürich, Switzerland, and Novartis Foundation.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.