The nuclear factor κB (NF-κB) signaling pathway is important in cancer-related inflammation and malignant progression. Here, we describe a new role for NF-κB in cancer in maintaining the immunosuppressive phenotype of tumor-associated macrophages (TAMs). We show that macrophages are polarized via interleukin (IL)-1R and MyD88 to an immunosuppressive “alternative” phenotype that requires IκB kinase β–mediated NF-κB activation. When NF-κB signaling is inhibited specifically in TAMs, they become cytotoxic to tumor cells and switch to a “classically” activated phenotype; IL-12high, major histocompatibility complex IIhigh, but IL-10low and arginase-1low. Targeting NF-κB signaling in TAMs also promotes regression of advanced tumors in vivo by induction of macrophage tumoricidal activity and activation of antitumor activity through IL-12–dependent NK cell recruitment. We provide a rationale for manipulating the phenotype of the abundant macrophage population already located within the tumor microenvironment; the potential to “re-educate” the tumor-promoting macrophage population may prove an effective and novel therapeutic approach for cancer that complements existing therapies.
Within most human and mouse cancers there is a significant macrophage population, attracted to the tumor microenvironment by cytokines and chemokines such as CSF-1 and CCL2 (1, 2). Macrophages are plastic cells; their phenotype depends on their anatomical location and the physiological or pathological context. Classically activated macrophages (also called M1) and “alternatively” activated macrophages (M2) represent two extremes in the spectrum of the macrophage phenotype (3). Tumor-associated macrophages (TAMs) closely resemble “alternative” (M2) macrophages (4). M2 macrophages produce high amounts of IL-10 but not IL-12, express scavenger receptors, and exhibit antiinflammatory and tissue repair functions (5). In contrast, M1 macrophages, activated by microbial products or IFN-γ, produce large amounts of proinflammatory cytokines, express high levels of MHC molecules, and are potent killers of pathogens and tumor cells (3). Despite their intrinsic antitumor potential as both tumoricidal cells and the major antigen-presenting cells present in tumors, ablation of macrophage function or their infiltration into experimental tumors inhibits cancer growth and metastasis (1, 6, 7). Due to the large TAM population in many tumors, an attractive therapeutic approach would be to increase their tumoricidal activity and attempt to promote antitumor immunity.
Recent studies in mouse models of colon and liver cancer have defined an important role for NF-κB activation in driving cancer-associated inflammation (8, 9). Our data describe an additional role for NF-κB in maintaining the immunosuppressive TAM phenotype. We have previously shown that malignant epithelial cells polarize macrophages to an M2-like phenotype (10, 11). Here, we show that malignant epithelial cells drive NF-κB activation in TAMs in a way that maintains their immunosuppressive phenotype.
The experiments described here are based on the hypothesis that NF-κB signaling is the central mechanism that maintains the alternative phenotype of TAMs. To investigate this we have used two different approaches to block NF-κB activity specifically in macrophages, both of which target IκB kinase (IKK)β, the major activator of NF-κB. Mouse bone marrow–derived macrophages (BMDMs) were infected with a recombinant adenovirus expressing a dominant-negative inhibitor of IKKβ (Adv-IKKβDN). IKKβ expression was also targeted in mice harboring a “floxed” Ikkβ allele (IKKβf/f) with adenovirus expressing Cre recombinase (Adv-Cre) to generate IKKβ-null macrophages (IKKβΔ) (12). Our data show that targeting IKKβ activity in BMDMs co-cultured with malignant tumor cells or TAMs isolated from established tumors reverses their tumor-promoting activity and promotes an antitumor M1 phenotype.
Results And Discussion
We previously showed that co-culture of BMDMs with human and mouse ovarian cancer cells led to increased tumor cell invasion (10, 11). In the first series of experiments described here, we investigated the macrophage signaling pathways that drive tumor cell invasion using co-cultures of BMDMs and syngeneic ID8 mouse ovarian cancer cells. Our first observation was that co-culture with macrophages deficient in MyD88 and IL-1R but not Toll-like receptor (TLR)2 or TLR4 inhibited, rather than increased, invasiveness of tumor cells in vitro (Fig. 1 A). Downstream of IL-1R and MyD88, we found that the ability of macrophages to enhance tumor cell invasiveness was dependent on NF-κB activation. Specific inhibition of IKKβ signaling and NF-κB activation by expression of IKKβDN or targeted deletion of IKKβ (IKKβΔ) in BMDMs prevented these cells from enhancing tumor cell invasiveness in vitro (Fig. 1 A and Fig. S1 A). We also isolated CD11b+ TAMs from ID8 tumor-bearing mice and infected them in vitro with Adv-IKKβDN before co-culture with ID8 cells. As with BMDMs, inhibition of IKKβ activity in TAMs also prevented tumor cell invasion (Fig. 1 A).
Previous literature has suggested that the mechanism by which inhibition of NF-κB prevented macrophage-induced tumor cell invasion might be due to decreased production of chemotactic factors by macrophages (8). But in fact we found that co-culture with IKKβ-targeted BMDMs or TAMs isolated from established tumors increased tumor cell apoptosis, as measured by caspase 3/7 activation (Fig. 1 B). In addition, inhibition of IKKβ activity in BMDMs or TAMs enhanced macrophage-mediated tumoricidal activity in vitro (Fig. 1 C).
We have previously shown that co-culture with ID8 cells polarizes macrophages to an M2-like phenotype: IL-10high, IL-12low, increased scavenger receptor expression, and TNF-αhigh (10, 11). NF-κB inhibition by expression of IKKβDN or targeted deletion of IKKβ (IKKβΔ) in BMDMs and TAMs isolated from ID8 tumors reduced production of IL-10 and TNF-α upon co-culture with tumor cells in vitro (Fig. 2 A and Fig. S1 B); however, IKKβ inhibition increased production of the antitumor cytokine IL-12 (Fig. 2 A and Fig. S1 B). The IL-12high IL-10low phenotype of IKKβ-targeted macrophages was associated with decreased expression of arginase-1 and elevated inducible nitric oxide (NO) synthase (NOS2) expression (Fig. 2 B and Fig. S1 C). Furthermore, increased NOS2 expression correlated with an increase in NO levels in the co-culture supernatant (Fig. 2 A and Fig. S1 B). These data suggested that inhibition of IKKβ signaling in BMDMs co-cultured with ID8 cells, and TAMs isolated from established tumors, switched macrophages from an M2 to a “classically” activated M1 phenotype.
As Stat1 is an essential transcription factor for IL-12, MHC class II, and NOS2 expression (13), we analyzed Stat1 activation in macrophages co-cultured with tumor cells. When IKKβ-targeted macrophages were co-cultured with ID8 tumor cells, they showed increased tyrosine phosphorylation of Stat1 (pY-Stat1, Tyr701), which correlated with increased NOS2 expression (Fig. 2 C). NOS2 has been directly implicated in macrophage-mediated tumoricidal activity (14), and several in vitro studies have demonstrated that NO donors are cytotoxic and proapoptotic to tumor cells (15, 16). We performed 111In-release cytotoxicity assays in the presence of a selective NOS2 inhibitor, 1400W, and demonstrated that inhibition of NOS2 rescues the cytotoxic activity of IKKβ-targeted BMDMs or TAMs isolated from established tumors (Fig. 2 D and Fig. S1 D). These data suggest that IKKβ-mediated suppression of Stat1 activation inhibits the M1 phenotype and tumoricidal activity of tumor-polarized macrophages.
To investigate the in vivo implications of these observations, we adoptively transferred IKKβ-targeted macrophages into mice bearing established ovarian tumors to test if “re-educating” macrophages through IKKβ inhibition could affect tumor progression. ID8 ovarian cancer cells that expressed firefly luciferase were injected i.p. into mice. After 5 wk, these mice had malignant ascites and established tumors, with deposits throughout the peritoneum (Figs. 3 A and 4 A, and Fig. S2) (11). We then adoptively transferred 5 × 106 WT-, IKKβΔ-, or IKKβDN-expressing BMDMs or TAMs isolated from established tumors and infected in vitro with the IKKβDN adenovirus (TAMDN). Tumor progression was assessed in situ by bioluminescence imaging on days 0, 7, and 14 after adoptive transfer (Figs. 3 A and 4 A, and Fig. S2). Transfer of IKKβ-targeted BMDMs or TAMs into tumor-bearing mice led to a significant (P < 0.01) decrease in tumor burden when compared with control macrophages (Figs. 3 B and 4 B, and Fig. S2). After 14 d, luciferase levels were still significantly lower in the IKKβ-targeted group compared with the untreated or control groups (Figs. 3 B and 4 B, and Fig. S2). We next investigated whether the increase in NOS2 expression in IKKβ-targeted macrophages was associated with enhanced tumoricidal activity in vivo. 4 h after adoptive transfer of IKKβ-targeted BMDMs or TAMs, there were significantly higher NO levels in the peritoneum compared with control groups (Figs. 3 C and 4 C, and Fig. S2; P < 0.01); however, this was not sustained and NO levels were no different 24 h after transfer of macrophages, despite the sustained effect on tumor growth (Figs. 3 C and 4 C, and Fig. S2). To analyze the TAM phenotype in vivo, we measured IL-10, IL-12, and TNF-α levels in ascitic fluid 14 d after macrophage transfer (Figs. 3 D and 4 D, and Fig. S2). Mice receiving IKKβ-targeted macrophages showed a significant change in cytokine production with a switch to an IL-10low IL-12high and TNF-αlow profile (Figs. 3 D and 4 D, and Fig. S2; P < 0.01). CD11b+ ascitic macrophages from these mice expressed increased IL-12p40 and decreased arginase-1 mRNA (Figs. 3 D and 4 D, and Fig. S2) and showed higher MHC class II surface expression compared with control macrophages (Figs. 3 E and 4 E, and Fig. S2). These data show that targeting IKKβ activity in BMDMs prevents their polarization to an M2 phenotype by the tumor microenvironment in vivo. In addition, we show that inhibition of IKKβ in TAMs isolated from established tumors switches their phenotype from tumor-promoting M2 to antitumor M1.
Despite a sustained antitumor effect upon transfer of IKKβ-targeted macrophages in vivo, there was only a transient increase in NO production. This suggested that a different mechanism might be responsible for the long-term inhibition of tumor growth. Of particular interest was a significant (P < 0.01) increase in NK cells (Fig. 5 A and Fig. S3) in ascites from mice adoptively transferred with IKKβ-targeted macrophages. We therefore reasoned that an IL-12–mediated increase in NK cells in the peritoneum may contribute to the tumoricidal effect of targeting IKKβ in TAMs. To investigate this hypothesis, we treated mice with a neutralizing IL-12p40 antibody before adoptive transfer of macrophages. Neutralizing IL-12 in our model rescued the increased recruitment of NK cells in mice receiving IKKβ-targeted macrophages and restored tumor growth to similar levels as in controls (Fig. 5, A–C, and Fig. S3). The p40 subunit of IL-12 is shared with the related proinflammatory cytokine IL-23 (17). To rule out a role for IL-23, we used a neutralizing antibody against IL-23p19. Neutralizing IL-23p19 had no effect on tumor growth (Fig. 5 C and Fig. S3). In further experiments, we recapitulated NK cell–mediated tumoricidal activity ex vivo using purified spleen NK cells. DX5+ NK cells were isolated and treated with ascitic fluid from the adoptive transfer experiments described above. When NK cells were stimulated with ascites from mice receiving IKKβ-targeted macrophages, they showed a significant increase in tumoricidal activity (Fig. 5 D and Fig. S3). This increase in cytotoxicity was rescued with a neutralizing antibody to IL-12p40, whereas neutralizing IL-23p19 had an additive effect (Fig. 5 D and Fig. S3).
The data described above show that IKKβ signaling in macrophages maintains their alternative tumor-promoting phenotype. The experiments described in Fig. 1 show that IL-1R/MyD88 signaling in macrophages is important in driving tumor cell invasion, suggesting that IL-1 may be the soluble mediator of NF-κB activation in our co-culture system. To investigate the role of IL-1R/MyD88 signaling in macrophage polarization in vivo, we adoptively transferred BMDMs deficient in MyD88, TLR2, TLR4, or IL-1R into ID8 tumor–bearing mice. Mice treated with IL-1R−/− or MyD88−/− macrophages had significantly less tumor burden after 14 d (S4A). In addition, ascitic fluid from mice treated with IL-1R−/− or MyD88−/− macrophages contained lower levels of IL-10 but higher levels of IL-12p70 compared with control groups (S4B). A similar switch toward an M1 phenotype was observed in the resident ascitic CD11b+ population with an IL-12p40high arginase-1low expression profile (S4C).
In summary, these data show that inhibition of IKKβ activity specifically in TAMs is able to reverse their tumor-polarized phenotype and re-educate them to actively kill tumor cells by both direct tumoricidal activity through NO production and the promotion of NK cell–mediated antitumor immunity through IL-12.
The studies described above reveal a new role for IKKβ in cancer-related inflammation. We demonstrate a previously unknown function for IKKβ in maintaining the alternative phenotype of TAMs and negative regulation of macrophage tumoricidal activity. In malignant disease, macrophages are recruited to the tumor from the circulation and become polarized to an alternative immunosuppressive phenotype, characterized by high expression of IL-10, TNF-α, and arginase-1, but low NOS2, IL-12, and MHC II expression. There is a large macrophage population in many tumors, and our data suggest that we may be able to harness their intrinsic antitumor potential to promote tumor regression. During the 1990s there were clinical trials of in vitro–activated autologous macrophages, using either IFN-γ alone or in combination with LPS; however, this approach had only limited success (18, 19). We have shown convincingly that targeting IKKβ in TAMs reverses their tumor-promoting activity and can drive tumor regression in vivo. These studies suggest that IKKβ inhibitors could represent a powerful therapeutic tool in cancer therapy. Besides the direct effects on antiapoptotic signaling in malignant cells, IKKβ inhibition may also modify the inflammatory microenvironment to promote an antitumor immune response.
Materials And Methods
Cell lines and reagents.
The mouse ovarian cancer cell line ID8 (provided by K. Roby, University of Kansas, Kansas City, KS) (20) was cultured as described previously (10). Lentiviral infection of ID8 cells with luciferase reporter construct was performed as described previously (21). In vitro luciferase activity was assayed in 0.5 × 106 ID8 cells in triplicate and according to the manufacturer's instructions (Promega). ID8–macrophage co-cultures were performed as described previously (22).
Generation of mouse macrophages.
Mouse macrophages were derived from the bone marrow of WT, TLR2, TLR4, MyD88, and IL-1R knockout C57BL/6 mice or IKKβ floxed mice (IKKβf/f) and cultured for 7 d in DMEM medium supplemented with 10 ng/ml recombinant mouse M-CSF (R&D Systems). TAMs were isolated from the peritoneal cavity of mice in which ID8 ovarian cancers had been growing i.p. for 10 wk.
BMDMs or CD11b+-selected in vivo–polarized TAMs were infected with recombinant adenovirus expressing a dominant-negative inhibitor of IKKβ (IKKβDN, TAMDN). These viruses are E1/E3 deleted, belong to the Ad5 serotype, and have been used in other studies (23, 24). IKKβf/f BMDMs were infected with recombinant adenovirus expressing Cre. IKKβf/f BMDMs infected with Adv-GFP were used as a control. Infections were performed at day 4 of macrophage differentiation in the presence of recombinant mouse M-CSF (R&D Systems).
Immunoblotting was performed with NOS2 (Santa Cruz Biotechnology, Inc.), β-actin (Sigma-Aldrich), NF-κB p65, phospho–NF-κB p65 (Ser536), Stat1, phospho-Stat1 (Tyr701), IKKβ and phospho-IKKβ (Ser181), and IκBα and phospho-IκBα (Ser32; all Cell Signaling Technology).
DNA-binding activity of transcription factor NF-κB p50/p65.
NF-κB activation was measured using the TransFactor assay (Millipore) (11).
Adoptive transfer of macrophages into mice bearing established tumors.
8-wk-old C57BL/6 female mice (B&K Universal) were injected i.p. with 106 ID8-Luc cells, and tumors were allowed to establish for 35 d. 5 × 106 macrophages were transferred adoptively i.p. on day 35. Neutralizing antibodies against IL-12p40/70 (17.8; R&D Systems), IL-23p19 (G23-8; eBioscience), or the respective control antibody were injected i.p. at 100 μg/mouse twice weekly. To assess tumor progression, luciferase activity was measured in cell lysates. All mice were maintained in the Biological Services Unit, Institute of Cancer and the CR-UK Clinical Centre, Barts and The London School of Medicine, and used according to established institutional guidelines under the authority of a UK Home Office project license (Guidance on the Operation of Animals, Scientific Procedures Act 1986).
Tumor progression was assessed in situ by bioluminescent imaging as described previously (21).
The phenotype of transferred macrophages was assessed by FACS analysis on a FACScan flow cytometer and analyzed using Cellquest software (Becton Dickinson).
For 111In-release assay, a cytotoxicity assay was performed for 24 h with macrophages. Radioactivity was measured by counting 1 min in a γ counter. For 51Cr-release assays, NK cell cytotoxicity was defined as lysis of target cells after 4 h. Ascite samples from tumor-bearing mice treated with adoptive transfer of modified macrophages were concentrated under vacuum for 30 min at 37°C. In vitro stimulation with ascites was performed by adding 10 μg/ml concentrated total ascites at the start of the 4-h cytotoxicity assay. Neutralizing antibodies against IL-12p40/70 (17.8; R&D Systems) or IL-23p19 (G23-8; eBioscience) or the respective control antibody were added to the release assay at 10 μg/ml.
The percentage of cytotoxic activity was calculated as the specific release of radioactivity: % cytotoxicity = [(experimental isotope − spontaneous isotope)/total isotope − spontaneous isotope)] × 100.
To determine caspase-3/7 activity (Promega), cytokine levels in cell culture supernatant or ascite ELISAs (R&D Systems) were used according to the manufacturer's instructions.
Experiments were performed in triplicate, and representative data are shown. Results were tested for statistical significance using Student's t test with Welch's correction or the Mann-Whitney test with GraphPad Prism Version 4.0c software.
Online supplemental material.
Fig. S1 shows the results of the NF-κB activity assay, the in vitro profile for in vivo–generated TAMs under co-culture. Fig. S2 summarizes the in vivo adoptive transfer of IKKβΔ macrophages and their respective controls. Fig. S3 shows the effect of neutralizing IL12p40 and IL-23p19, and NK cell data. Fig. S4 shows ex vivo analysis of tumor burden in mice adoptively transferred with macrophages from MyD88, TLR2, TLR4, and IL-1R knockout mice.
© 2008 Hagemann et al. This article is distributed under the terms of an Attribution–Noncommercial–Share Alike–No Mirror Sites license for the first six months after the publication date (see http://www.jgp.org/misc/terms.shtml). After six months it is available under a Creative Commons License (Attribution–Noncommercial–Share Alike 3.0 Unported license, as described at http://creativecommons.org/licenses/by-nc-sa/3.0/).
T. Hagemann and T. Lawrence contributed equally to this work.
The IKKβDN-expressing adenovirus was kindly provided by Rainer de Martin, Department of Vascular Biology and Thrombosis Research, Medical University Vienna, Austria. The MyD88−/− bone marrow was a kind gift from Prof. Siamon Gordon, Sir Wiliam Dunn School, Oxford, UK. The IL-1R−/− bone marrow was a kind gift from Prof. Nancy Rothwell, Faculty of Life Sciences, University of Manchester, Manchester, UK.
This research was supported by grants from the Medical Research Council (to T. Hagemann, T. Lawrence, and I. McNeish), Cancer Research UK (to T. Hagemann, H. Kulbe, I. McNeish, and F.R. Balkwill), Charitable Foundation of Bart's and The London (to S.C. Robinson, T. Lawrence, and T. Hagemann), Wellcome Trust International Research Fellowship (to T. Lawrence), Centocor Inc. (to K.A. Charles), and HEFCE (to F.R. Balkwill).
F.R. Balkwill has received funding and in the past has acted as a consultant for Centocor, which supplies anti–TNF-α antibodies commercially. K.A. Charles was funded by Centocor. The authors have no other conflicting financial interests.