NK Cells Control Tumor-Promoting Function of Neutrophils in Mice

The tumor microenvironment consists of not only transformed cancer cells, but also various stromal cells that control tumor progression (1, 2). Of such tumor stromal cells, innate and adaptive immune cells and their effector molecules regulate tumor development with an immune editing process (3–5). Natural killer (NK) cells contribute to immunological surveillance of development and subsequent growth of cancer cells through a variety of effector mechanisms, including perforin/granzymes or death receptor dependent cytotoxicity and antitumor cytokine production (6–8). Besides their direct antitumor effector function, NK cells also control subsequent adaptive immune responses (9–11).

In contrast to antitumor immune responses, tumor-promoting immune responses, often regarded as an inflammation, contribute to malignant cancer progression (2, 12, 13). The inflammatory immune cells involved in such tumor-promoting immunity include innate cells as well as adaptive lymphocytes. Such cells produce a variety of inflammatory cytokines, chemokines, and growth factors (2, 13). Neutrophils, abundant circulating innate inflammatory cells, respond to infection or tissue damage (14–16). Neutrophils mediate host defense through multiple functions including phagocytosis, reactive oxygen species (ROS) production or release of granules containing antimicrobial peptides and proteases (14–17). Neutrophils also aid tumor-promoting immune responses (18–20). Indeed, clinical studies support the use of circulating neutrophil counts or neutrophil-to-lymphocyte ratios (NLRs) as a predictive biomarker for poor outcome or distant metastasis in cancer patients (21–25). The function of neutrophils may be converted from tumor-suppressing to tumor-promoting in the tumor-bearing state. Indeed, tumor-associated neutrophils (TANs) can be classified into antitumor (N1) or tumor-promoting (N2) subpopulations (26). Although N2 TANs were driven by the presence of TGFβ (26), IFNβ polarized TANs toward N1 phenotype (27). These studies suggest that an improved understanding of how TANs switch from the antitumor to tumor-promoting state will aid efforts to control cancer progression.

Although both NK cells and neutrophils are components of tumor stroma, there is no indication of whether the interaction between NK cells and neutrophils is involved in tumor control. In this study, we demonstrate that NK cells control tumor-promoting inflammation through the functional modification of neutrophils. NK cells control the tumor-promoting function of neutrophils via an IFNγ-mediated mechanism. Therefore, tumor progression in an NK cell–depleted host is diminished when the IL17A–neutrophil axis is absent. In NK cell–depleted mice, neutrophils acquired a tumor-promoting phenotype as seen in their VEGF-A upregulation to promote tumor growth and angiogenesis. The VEGFR inhibitor suppressed tumor growth in NK cell–depleted mice through mechanisms dependent upon neutrophils. Furthermore, the systemic neutropenia induced by an antimetabolite treatment showed an anticancer effect only in mice lacking NK cells. Thus, NK cells likely control the tumor-promoting and angiogenic function of neutrophils.

Wild-type C57BL/6 (WT) mice were purchased from CLEA Japan, Inc.. IFNγ^–/– (IFNγ KO) and IL17^–/– (IL17 KO) mice on B6 background were kindly provided by Dr. Y. Iwakura (Tokyo University of Science, Chiba, Japan) and maintained at Laboratory Animal Research Center, Institute of Medical Science, University of Tokyo. In some experiments, groups of mice were treated with anti-CD4 (clone GK 1.5), anti-CD8 (clone 53.6.2), anti-NK1.1 (clone PK136), or anti-asialo-GM1 (asGM1; Wako Chemicals) on days −3 and −1, anti-CXCR3 (clone CXCR3-173, Bio X Cell) on days −1, 0, 2, 4, and 6 or anti-Ly6G (clone 1A8, Bio X Cell) on day −1 and 3 (where day 0 is the day of primary tumor inoculation). Antibodies against CD4, CD8, and NK1.1 were purified from hybridoma cells. In some experiments, mice were treated with SU5416 [AdooQ BioScience, 30 mg/kg intraperitoneally (i.p.)] for 7 days from the day of tumor inoculation. To induce systemic neutropenia, mice were injected intravenously (i.v.) with 150 mg/kg 5-fluorouracil (5-FU) at day 0 relative to the tumor implantation. All experiments were approved and performed according to the guidelines of the Animal Care and Use Committee of the Graduate School of Pharmaceutical Sciences of the University of Tokyo, the Care and Use of Laboratory Animals of University of Toyama, and the Animal Care and Use Committee of Institute of Medical Science of the University of Tokyo.

MCA205, a methylcholanthrene-induced murine fibrosarcoma cell line was kindly provided by Dr. S.A. Rosenberg (National Cancer Institute, Bethesda, MD) and maintained in complete RPMI 1640 medium. MCA205-Luc2 cells stably expressing luciferase with a CMV promoter were established as previously described (28). Briefly, MCA205 cells (5 × 10⁵/well) were seeded in a 6-well plate and transfected with pGL4.50 vector using Lipofectamine 2000. The cells were selected with Hygromycin B (200 μg/mL) and cloned by limiting dilution. In vitro and in vivo characterization of MCA205-Luc2 cells is shown in Supplementary Fig. S1.

The mouse Lewis lung carcinoma cell line (3LL) was kindly provided by Dr. Kazuyoshi Takeda (Juntendo University), and the luciferase-expressing 3LL cell line (3LL-Luc2) was prepared as previously described (29). pGL4.50 [luc2/CMV/Hygro] vector, and D-luciferin were obtained from Promega. Lipofectamine 2000 was purchased from Invitrogen. Hygromycin B was obtained from Nacalai Tesque. Those cell lines were kept in culture no more than a month and have not been reauthenticated.

MCA205-Luc2 cells or 3LL-Luc2 cells were inoculated subcutaneously (s.c.). To obtain bioluminescent images, mice were injected with D-luciferin (Promega, 150 mg/kg i.p.) and luminescence was measured with an in vivo imaging system (IVIS Lumina II) 10 minutes after the D-luciferin injection. Analyses of regions of interest were performed using Living Image 4.2 Software (Caliper Life Science) to determine the light emitted from the tumor. For each mouse, all values were determined as photons per second (photon/sec).

MCA205-Luc2 cells (10⁵) were inoculated s.c. Seven days after inoculation, tumor tissues were dissected, minced, and digested with 2 mg/mL collagenase (Roche Diagnostics GmbH) and 0.1 mg/mL DNase I (Roche Diagnostics GmbH) in serum-free RPMI 1640 for 1 hour at 37°C. Samples were further homogenized through wire mesh. For collecting tumor culture supernatant, the tumor homogenates were cultured in complete RPMI 1640 media for 24 hours and the cell-free supernatant were harvested. For flow cytometry analysis, cells were first preincubated with anti-CD16/32 (2.4G2) to avoid nonspecific binding of antibodies to FcγR. The cells were then incubated with a saturating amount of fluorophore-conjugated monoclonal antibody (mAb). Antibodies against CD3ϵ (2C11), NK1.1 (PK136), CD11b (M1/70), Gr-1 (RB6-8C5), and Ly6G (1A8) were purchased from Biolegend, eBioscience, or Tombo Bioscience. Flow cytometry analysis was performed with a FACS Canto (BD Bioscience) and the data were analyzed with FlowJo software (Tree Star).

Growth factor–reduced Matrigel (BD Biosciences) containing MCA205-Luc2 cells (10⁵) was s.c. inoculated and 1% Evans blue was injected intravenously into mice at 7 days after the inoculation. Ten minutes later, Matrigel plugs were removed after perfusion by saline and incubated with formamide for 48 hours. The amount of Evans blue dye was quantified by the absorbance at 620 nm and normalized to the weight of Matrigel.

To isolate tumor-infiltrating neutrophils, tumor samples were harvested 7 days after the inoculation and CD11b⁺ Ly-6G⁺ neutrophils were isolated by cell sorting (>90 % purity, FACS Aria Special Order, BD Bioscience). To isolate peritoneal neutrophils, mice were injected i.p. with 7.5% casein solution (1 mL) and cells were collected from the peritoneum 24 hours after casein injection. Ly6G⁺ neutrophils were purified using a Mojosort cell separation kit (Biolegend). Briefly, cells were first preincubated with CD16/32 (2.4G2) mAb to avoid nonspecific binding of antibodies to FcγR. The cells were then incubated with a saturating amount of biotin-conjugated Ly6G mAb and streptavidin-conjugated magnet beads. Ly6G⁺ neutrophils were isolated using a Mojosort magnet. Peritoneal neutrophils (10⁶ cells) were cultured with tumor supernatant (1:1 dilution with fresh media) for 6 hours. Tumor supernatant was prepared as described above mentioned and the blocking anti-IFNγ (clone XMG1.2, 10 μg/mL) was added in some experiments. Gene expression of neutrophils was measured using real-time RT-PCR. RT-PCRs were performed with the following primers: ICAM-1 (f) 5 ″-CAATTCACACTGAATGCCAGCTC-3″ (r) 5″-CAAGCAGTCCGTCTCGTCCA-3″; TNFα (f) 5″-TTGTCTACTCCCAGGTTCTCT-3″ (r) 5″- GAGGTTGACTTTCTCCTGGTATG-3″; CCL-2 (f) 5″-CTCACCTGCTGCTACTCATTC-3″ (r) 5″-ACTACAGCTTCTTTGGGACAC-3″; VEGF-A (f) 5″-GTGCACTGGACCCTGGCTTTA-3″ (r) 5″-GGTCTCAATCGGACGGCAGTA-3″. To standardize the cDNA, the housekeeping gene GAPDH was tested with primer pairs: (f) 5″-AAATGGTGAAGGTCGGTGTG-3″; (r) 5″-TGAAGGGGTCGTTGATGG-3″. The protein expression of VEGF-A was determined by the specific ELISA kit (RayBiotech) according to the manufacturer's instruction.

For frozen section preparation, tumor tissues were immediately fixed with 1% paraformaldehyde in 0.1 mol/L phosphate buffer (pH 7.4) at 4°C, and then transferred to PBS. Tissues were soaked in 30% PBS-buffered sucrose, and then the entire tissue was embedded in optimal cutting temperature (OCT) compound (Sakura Finetek) and frozen at −20°C. The frozen sections were cut on a cryostat (CM 3050S-IV; Leica Microsystems). The sections were incubated at 4°C overnight with monoclonal anti-CD31 (1:100 MAB1398Z Millipore). For immunofluorescence, the Alexa-Fluor488–conjugated secondary antibodies (Jackson) were used at 1:500 dilutions. Nuclei were stained with Hoechst 33258 (Nacalai Tesque). The immunofluorescent images were randomly obtained by a confocal microscope (TCS-SP5; Leica Microsystems).

All data were obtained from a group of 6 to 9 mice and are representative of at least two independent experiments. Data were analyzed for statistical significance using the Student t test. P values less than 0.05 were considered statistically significant.

In order to determine the role of NK cells in controlling cancer cell growth, we monitored the luminescence of MCA205-Luc2 cells implanted in control B6 mice (control) or B6 mice treated with antibodies that deplete immune effector cells. In the absence of NK cells, but not CD4⁺ or CD8⁺ T cells, the in vivo growth of MCA205-Luc2 cells was significantly promoted compared with control B6 mice (Fig. 1A). Such NK cell–dependent tumor control was seen for up to 10 days after the tumor inoculation. T-cell control, particularly control by CD8⁺ cells, dominated later time points (data not shown). To further understand the effector pathway by which NK cells control cancer cell growth, we tested IFNγ-deficient (IFNγ KO) mice. As shown in Fig. 1B, the in vivo growth of MCA205-Luc2 cells was promoted in IFNγ KO mice as in NK cell–depleted mice. NK cell depletion did not further enhance MCA205-Luc2 cell growth in IFNγ KO mice. To examine whether NK cells systemically or locally control in vivo growth of MCA205-Luc2 cells, we blocked the CXCR3 chemokine receptor which regulates tumor infiltration of migratory CD27^hi NK cells (30, 31). As shown in Fig. 1C, the in vivo blocking of CXCR3 showed similar results in the growth of MCA205-Luc2 cells to NK cell–depleted mice. CXCR3 blocking did not further enhance MCA205-Luc2 cell proliferation in the NK cell–depleted mice. These results suggest that NK cells control cancer cell growth through mechanisms dependent on IFNγ and CXCR3.

The cytokine IL17A supports the tumor-promoting immune response and counteracts the antitumor function of IFNγ (32–34). We found that NK-cell depletion did not affect the in vivo growth of MCA205-Luc2 cells in IL17A-deficient (IL17 KO) mice (Fig. 2A). This result indicates that the promotion of in vivo cancer cell growth in the absence of NK cells required IL17A. As IL17A attracts neutrophils to inflammatory sites (35) and neutrophils are a component of tumor-promoting inflammation (18–20), we next determined whether neutrophils are involved in the promotion of cancer cell growth in NK cell–depleted mice. We used anti-Ly6G (clone 1A8) to deplete neutrophils as previously described (36). As shown in Fig. 2B, the promotion of MCA205-Luc2 cell growth was diminished by neutrophil depletion in the absence of NK cells. This contribution of neutrophils to cancer cell growth in NK cell–depleted mice was confirmed in a 3LL-Luc2 model (Fig. 2C). Neither MCA205 cells nor 3LL cells in mice with intact NK cells showed differences in cancer cell growth upon neutrophil depletion (Fig. 2B and C), suggesting that neutrophils are involved in controlling cancer cell growth only in the context of NK cell absence. In order to understand whether NK cells control tumor neutrophil accumulation, we examined the infiltration of neutrophils into MCA205-Luc2 tumors by flow cytometry. Although we found that neutrophil infiltration into the tumor sites was impaired in IL17 KO mice, NK-cell depletion did not affect the proportion of tumor infiltrating neutrophils in either WT and IL17 KO mice (Fig. 3A and B). Collectively, these results indicate that neutrophils promote in vivo cancer cell growth in the absence of NK cells and the quality, but not quantity, of neutrophils at the tumor site can be controlled by NK cells.

Tumors contain tumor-associated neutrophils (TAN), which are identified with either antitumor (N1) or tumor-promoting (N2) phenotypes (26). In order to examine the phenotypic difference of TANs with or without NK cells, we isolated and characterized TANs. Although we did not see any difference in the maturation status of TANs as determined by their nuclear segments (Fig. 4A), TANs in the absence of NK cells upregulated expression of VEGF-A mRNA as well as other antitumor (ICAM-1, TNFα) or protumor (CCL2) markers (Fig. 4B). Naïve peritoneal neutrophils upregulated expression of VEGF-A mRNA upon culture with the supernatant of an NK cell–depleted MCA205 tumor tissue culture (Fig. 5A), suggesting that NK cells control VEGF-A expression of TANs by modulating soluble factor(s) within the tumor microenvironment. To determine whether IFNγ directly inhibits VEGF production (37, 38), we tested the role of NK cells and IFNγ in causing neutrophils to express VEGF-A. Although the blocking of IFNγ did not affect neutrophil VEGF-A expression upon coculture with tumor tissue culture supernatants (Fig. 5B), the supernatant of IFNγ KO tumors upregulated neutrophil VEGF-A expression regardless of NK-cell status (Fig. 5C). Expression of VEGF-A protein within tumors was elevated in the absence of NK cells, although addition of anti-Ly6G treatment abrogated the effect (Fig. 5D). These results imply that NK cells control neutrophil VEGF-A expression and that endogenous IFNγ controls neutrophil VEGF-A switching, likely by an indirect mechanism. We then assessed whether the absence of NK cells affects in vivo tumor angiogenesis by using a Matrigel plug assay. As shown in Fig. 6A, vascularization toward the MCA205/Matrigel plug was enhanced in the absence of NK cells. Depletion of neutrophils in NK cell–depleted mice eliminated the effect on angiogenesis. We also observed that blood vessel area and diameter increased in the absence of NK cells as determined by the immunohistochemistry staining of anti-CD31 within MCA205/Matrigel plug. Such blood vessel formation was dependent on neutrophils (Fig. 6B–D). Those data indicate that NK cells control tumor angiogenesis through neutrophils.

To demonstrate an involvement of VEGF-dependent angiogenesis in tumor growth of NK cell–depleted mice, we used the tyrosine kinase inhibitor SU5416 to block the VEGF pathway in the presence or absence of NK cells and neutrophils. Treatment with SU5416 inhibited in vivo growth of MCA205-Luc2 cells only in NK cell–depleted mice (Fig. 7). This selective antitumor effect was neutrophil dependent, because the effect was blocked by anti-Ly6G treatment. Thus, NK cells control the neutrophil angiogenic switch by regulating VEGF-A expression.

In order to test a clinical application of our findings, we examined whether systemic neutropenia could control the growth of tumors in mice with impaired or absent NK cells. The antimetabolite 5-FU acts as an anticancer drug by targeting DNA replication (39) and inducing systemic neutropenia (40). A single treatment of 5-FU reduced the population of neutrophils in the peripheral blood, bone marrow, spleen, and tumor (Fig. 8A). At this dose of 5-FU, we did not see any antitumor effect as shown in Fig. 8B. In contrast, the same dose of 5-FU showed an anticancer effect in mice with depleted NK cells. These data suggest that chemotherapy with 5-FU may control tumor malignancy both through inducing systemic neutropenia and through direct antitumor effects when NK cells are impaired.

Although NK cells are known for their antitumor immune responses, their role in tumor-promoting immune responses remains unknown. In this study, we demonstrated that the tumor-infiltrating NK cells control in vivo cancer cell growth through mechanisms dependent on IFNγ and CXCR3, in concert with our previous results (31). We further show that the IL17A–neutrophil axis is involved in such promotion of in vivo cancer cell growth when NK cells are absent. NK cells suppressed expression of VEGF-A in tumor-infiltrating neutrophils. When NK cells were absent, neutrophil-dependent tumor-induced angiogenesis was promoted. An inhibitor of VEGFR suppressed tumor growth in NK cell–depleted mice in a manner dependent on neutrophils. Treatment with the antimetabolite 5-FU controlled cancer cell growth by two mechanisms: (i) inducing systemic neutropenia and (ii) direct antitumor effects when NK cells are absent. Thus, NK cells function in angiogenic switching of neutrophils to control tumor-promoting immune responses within the tumor microenvironment.

In addition to the suppressive arm of the immune response to cancer, known as cancer immune surveillance or immunoediting, evidence indicates that an inflammatory tumor microenvironment contributes to cancer promotion or malignant progression (2, 12, 13). Of the immune cells involved in tumor-promoting immune responses, neutrophils are innate immune cells that often promote tumor-associated inflammation (13, 41). Infiltration of TANs correlates with poor prognosis in both preclinical and clinical tumor settings (19, 20, 42, 43). TANs display plasticity in their function either as an antitumor (N1) or as a tumor-promoting (N2) phenotype. TGFβ can lead TANs to the tumor-promoting N2 phenotype (26, 27). Although we observed a higher production of TGFβ in NK cell–depleted tumors, we did not see any change in the maturation status of TANs in NK cell–depleted tumors as determined by their nuclear segmentation. Instead, TANs in the absence of NK cells upregulated expression of VEGF-A among the N2-associated molecules. Because the culture supernatant of NK cell–depleted tumors also upregulated VEGF-A expression in naïve peritoneal neutrophils, we hypothesized that a soluble factor produced within the tumor microenvironment would be responsible for polarization of TANs in the absence of NK cells. Corresponding with the suppressive role of IFNγ in VEGF expression (37, 38), we found that the NK-cell IFNγ regulated VEGF expression on neutrophils by an unknown mechanism. Although NK-cell IFNγ can block migration of neutrophils to inflammation sites (44, 45), NK-cell depletion did not affect the ability of neutrophils to infiltrate tumors, suggesting that NK cells control the tumor-promoting, but not tumor-infiltrating functions of neutrophils. The mechanism by which NK cells control the tumor-promoting function of neutrophils depends on IFNγ. Endogenous type I IFNs such as IFNβ also regulate tumor angiogenesis and growth in a mouse model through neutrophils (27).

Two components of the innate immune system, neutrophils and NK cells, were found in this study to counter-regulate tumor progression. In their anti-inflammatory role, NK cells regulate neutrophil function through the NKG2A receptor pathway in a colitis model (46). Human NK cells trigger neutrophil apoptosis through activating NK-cell receptor NKp46- and Fas-dependent pathways (47, 48). The NK cell–produced cytokines, IFNγ and TNFα, convert neutrophils from tumor promoting to tumor suppressing (49). TANs also influence tumor control by coordinating with other components of immune cells through their ability to suppress antitumor immunity (43, 50, 51). Although a suppressive function of TANs in the activity of antitumor CD8⁺ T cells was reported (43), we did not see any difference in the status of tumor-infiltrating CD8⁺ T cells as determined by their effector or memory marker CD62L/CD44 expression or activation marker PD-1 expression (Supplementary Fig. S2). Moreover, NK cells, but not CD8⁺ T cells, temporarily and dominantly control in vivo cancer cell proliferation by modifying neutrophil function through an IFNγ-dependent mechanism. Further study is required to ascertain the relevance of our findings to later stage tumors. Nevertheless, we conclude that TANs in the absence of NK cells promote tumors rather than suppress antitumor CD8⁺ T-cell responses.

Regarding the clinical implication of our findings, neutrophil-to-lymphocyte ratios correlate with prognosis for cancer patients (21–25). We demonstrated that systemic neutropenia induced by 5-FU treatment controlled cancer cell proliferation separately from its direct antitumor effect in NK cell–depleted mice. In alignment with our preclinical findings, a clinical study indicates that peripheral NK-mediated cytotoxicity is negatively associated with survival of patients treated with 5-FU (52). We propose that a therapeutic approach involving functional impairment of neutrophils in cancer patients may improve their prognosis, particularly for tumors that lack NK-cell activity or accumulation. Our findings offer an opportunity to improve cancer therapy and predict the therapeutic outcome in the context of NK cell–neutrophil counterregulation.

No potential conflicts of interest were disclosed.

Conception and design: K. Ogura, H. Tahara, Y. Hayakawa

Development of methodology: K. Ogura, S. Yamamoto, M. Sasahara, Y. Hayakawa

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K. Ogura, M. Sato-Matsushita, S. Yamamoto, Y. Iwakura, H. Tahara

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K. Ogura, S. Yamamoto, M. Sasahara, Y. Hayakawa

Writing, review, and/or revision of the manuscript: K. Ogura, S. Yamamoto, H. Tahara, Y. Hayakawa

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K. Ogura, S. Yamamoto, T. Hori, M. Sasahara, I. Saiki, Y. Hayakawa

Study supervision: S. Yamamoto, I. Saiki, Y. Hayakawa

This work was partly supported by grants from Grant-in-Aid for Scientific Research (C) 26430158, the MEXT and Extramural Collaborative Research Grant of Cancer Research Institute, Kanazawa University (Y. Hayakawa).

We are grateful to Satoru Yokoyama and Naofumi Mukaida for their discussion, and Setsuko Nakayama and Asuka Asami for their technical assistance.

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.