Tumor-Specific Bacteriophages Induce Tumor Destruction through Activation of Tumor-Associated Macrophages¹

Despite intensive research and introduction into clinical practice of multiple approaches of therapy and prevention, cancer is the second leading cause of death in Western countries. Cancer immunotherapy (i.e., the activation of the patients’ own immune system to combat cancer) has evolved during the past decades as an alternative modality of cancer treatment. More recently, this approach has also been applied to models of autoimmunity (1, 2). Unfortunately, the success of cancer immunotherapy has, thus far, been limited. To improve the efficacy of cancer immunotherapy, alternative strategies that use the host immune response need to be explored and characterization of novel molecular and cellular mechanisms involved in tumor immune evasion is critical.

In animal models and humans, tumors develop several strategies to escape recognition of the immune system and to actively suppress the immune response. The mechanisms of tumor immune evasion include down-regulation of molecules involved in Ag presentation and processing (3, 4, 5, 6), expression of factors promoting survival (7, 8, 9), or secretion of immunomodulatory factors within the tumor microenvironment, such as cytokines (10). Other important factors released by tumor cells or stroma are MCP-1, macrophage CSF, macrophage migration inhibitory factor, and vascular endothelial growth factor (11, 12), which may recruit tumor-associated macrophages (TAMs).⁴

In contrast to the tumoricidal M1 (classically activated) macrophages which produce NO, IL-12, and TNF (13), TAMs are M2-polarized (alternatively activated) and suppress tumor-specific T cell activation and proliferation through secretion of factors such as IL-10 (14), TGF-β, and IDO (15). The presence of TAMs has been shown to correlate with poor prognosis in several cancers (e.g., melanoma and breast cancer) (16, 17) and TAMs were demonstrated to promote tumor growth by secretion of proangiogenic factors (18).

Given these observations, targeting of TAMs represents a promising strategy for cancer immunotherapy. Studies in animal models have demonstrated that both ablation of TAMs using a DNA vaccine (19) and switching TAMs from the M2 to the M1 phenotype through applying a combination of CpG and an anti-IL-10 receptor Ab (20) significantly prolonged survival, the latter also resulting in tumor protection and antitumor immunity. However, macrophage activation phenotype switching is still a somewhat controversial issue, as conflicting reports have been provided (21, 22) and thus efficacious means of altering the properties of TAMs must be elucidated to improve tumor therapy.

We previously described a novel experimental immunotherapeutic approach to treat mouse melanoma tumors based on the use of tumor-specific phage display particles (23). Bacteriophages, or phages, are a family of viruses that can only infect bacteria. By insertion of specific DNA sequences encoding peptides or proteins into the genes encoding the phage capsid proteins, fusion proteins are produced, which are incorporated into the phage particle during assembly. This was first demonstrated by Smith in 1985 (24) and phages have since then been widely used to construct peptide/protein (25) and Ab libraries (26) of up to billions of variants. Through various selection protocols, such as biopanning, phage display libraries offer the possibility of rapid selection and identification of peptides and proteins with high binding specificity for a huge diversity of targets, including tumor Ags.

Phages are highly immunogenic and are known to induce immune responses directed to their naive coat proteins (27). Immune responses can also be induced by phage DNA. Administration of M13 phage DNA induced IFNs and protected mice against a vaccinia virus infection, most likely due to the presence of CpG motifs in their genome (28). Phages are generally considered to have no intrinsic tropism for mammalian cells. However, it was reported already in 1940 that phages possess antitumor activity in mice and rabbits (29). More recently, it was demonstrated that a substrain of phage T4 could significantly reduce growth of experimental s.c. melanoma and lung cancer in mice when administered i.p (30) and also inhibit the formation of lung metastases (31).

Based on our recent finding that tumor-specific phage treatment results in neutrophil infiltration and tumor regression and other reports that TLR agonists effectively activate TAMs and possibly also drive the conversion of the M2 to the M1 phenotype (32), we hypothesized that activation of TAMs might be the mode of action of treatment with tumor-specific phages. In the present work, we demonstrate that bacteriophages, in a TLR-dependent manner, efficiently modify the tumor microenvironment, polarizing TAMs toward an M1 phenotype, which subsequently promotes the influx of neutrophils with enhanced tumoricidal capacity.

To investigate the ability of phages to induce secretion of IFN-γ, single-cell suspensions were prepared from C57BL/6 or MyD88^−/− mouse spleens. Erythrocytes were removed by ammonium chloride lysis and the splenocytes were cultured in DMEM (BioWhittaker) supplemented with 2 mM l-glutamine (Sigma-Aldrich), 10 mM HEPES, 1% nonessential amino acids, 5 × 10⁻⁵ M 2-ME, 25 μg/ml gentamicin, and 10% heat-inactivated FCS (Invitrogen) at 1 × 10⁶ splenocytes/well in a 96-well plate. Wild-type phages, crude fractions, or preparations cleared of endotoxin at 5 × 10¹⁰ pfu/150 μl were added to the cells to a total volume of 300 μl/well. After 48 h incubation at 37°C, culture supernatants were collected and cytokine content was measured using an IFN-γ ELISA kit (Mabtech) according to the manufacturers’ instructions.

All experiments using mice were approved by the Swedish National Board for Laboratory Animals. Female MyD88^−/−, which lack signaling through TLRs (33), and 6- to 10-wk-old C57BL/6 mice were obtained from Taconic M&B and housed at the Microbiology and Tumor Biology Center (MTC) at Karolinska Institute. Mice were injected s.c. in the right flank with 5 × 10⁴ B16-F10 cells. Once the tumors were palpable (∼5 mm diameter), endotoxin-free tumor-specific phage (WDC-2; Ref. 23), 1 × 10¹¹ pfu in 100 μl PBS, was injected s.c. peritumorally. Twenty-four hours post treatment, tumors were excised. For histological examination, tumors were fixed in 4% buffered formalin phosphate (Apoteksbolaget) for 6–8 h and transferred to 70% ethanol. Sections (5 μm) from paraffin-embedded tumors were placed on glass slides. Deparaffinization, rehydration, and H&E staining were performed using standard procedures. Neutrophils were characterized by identification of segmented nuclei. For low-density array (LDA), tumors were stored in RNAlater (Qiagen) for later use.

Tumor-specific phage-treated tumors were subjected to RNA isolation using the RNeasy kit (Qiagen) according to the manufacturer’s instructions. After isolation, the RNA concentration was measured by spectroscopy (NanoDrop). cDNA was synthesized using Superscript II (Invitrogen) reverse transcriptase as per the manufacturer’s instructions. For the TaqMan LDA assay, 5 μl of cDNA was mixed with 45 μl of RNase-free water and 50 μl of 2× Universal TaqMan Mastermix (Applied Biosystems). The mixture was applied to an LDA plate (Applied Biosystems) containing 45 different macrophage target genes including three housekeeping genes (18S, GAPDH, and HPRT). The plate was spun for 2 min at 330 × g and was sealed with a manual sealer (Applied Biosystems). The RT-PCR analysis was performed in a 7900 HTA (Applied Biosystems) using the default cycling conditions 2 min at 50°C, 10 min at 95°C, 40 cycles of 15 s at 95°C, and 60 s at 60°C. Fluorescent signals were collected during the annealing temperature and CT threshold was set at the point where the sample signals were exponential. Data were analyzed with the comparative CT method and sample data was normalized using the geometric mean of two housekeeping genes (GAPDH and HPRT), which had a correlation factor higher than 0.9. All experiments were related to a negative control.

The B16-F10 mouse melanoma cell line was obtained from American Type Culture Collection. Cells were cultured in complete medium (CM) (DMEM supplemented with 2 mM l-glutamine, 10 mM HEPES, 25 μg/ml gentamicin, and 10% heat-inactivated FCS). Primary mouse macrophages were collected by flushing the peritoneal cavities of 8- to 10-wk-old female C57BL/6 mice with cold PBS. After collection, the cells were washed twice in PBS, resuspended in complete medium, and plated on 24-well plates. Nonadherent cells were washed off after 24 h and adherent cells were cultured for 7 additional days. Cells were stained with a F4/80 Ab to verify the mouse macrophage phenotype. To induce the TAM phenotype, macrophages were cultured in CM conditioned with 30% medium from B16-F10 melanoma cultures for 48 h.

Mouse macrophages derived from C57BL/6 and MyD88^−/− mice were plated at 2 × 10⁶ cells/well in 24-well plates. Wild-type M13 bacteriophages (New England Biolabs) were propagated in Escherichia coli according to standard procedures and purified by PEG/NaCl precipitation. Phage preparations were rendered endotoxin free using Triton X-114 phase separation as described (23). Phages (1 × 10¹¹ pfu) were added to macrophage or TAM cultures. Twenty-four hours later, supernatants were collected and cytokine content was measured using the cytometric bead array Mouse Inflammation Kit (BD Biosciences). Data collection and analysis was performed using a FACSCalibur flow cytometer and the FCAP array software version 1.0.0 (BD Biosciences).

Mouse macrophage and TAM cultures were subjected to RNA isolation and cDNA synthesis was performed as described above. For analysis of cytokine gene expression, the following primers were used: IFN-γ forward, GCT-TTA-ACA-GCA-GGC-CAG-AC; IFN-γ reverse, GGA-AGC-ACC-AGG-TGT-CAA-GT; TGF-β forward, CAC-AGT-ACA-GCA-AGG-TCC-TTG-C; TGF-β reverse, AGT-AGA-CGA-TGG-GCA-GTG-GCT; IL-12 forward, ATG-ACC-CTG-TGC-CTT-GGT-AG; IL-12 reverse, GAT-TCT-GAA-GTG-CTG-CGT-TG; IL-10 forward, GT-GAA-AAT-AAG-AGC-AAG-GCA-GTG; IL-10 reverse, AT-TCA-TGG-CCT-TGT-AGA-CAC-C; TNF-α forward, AAT-GGC-CTC-CCT-CTC-ATC-AGT; TNF-α reverse, CT-ACA-GGC-TTG-TCA-CTC-GAA. Real time PCR was performed on Applied Biosystem’s ABI Prism 7700 Sequence Detection System using the SYBR Green PCR Master Mix (Applied Biosystems), and normalized using L32 expression values.

Mouse macrophages and TAMs were stimulated with bacteriophages as described above. Macrophages were detached using 200 μl of Accutase (PAA Laboratories) and washed in PBS. To analyze the induction of surface marker expression, cells were double stained with either anti-H2-K^b/I-A^b or anti-CD80/CD86 Abs (FITC/PE) (all from BD Biosciences). Data collection and analysis was performed using a FACSCalibur flow cytometer and CellQuest Pro software (BD Biosciences).

Mouse spleens were collected and splenocytes were tested for their migratory capacity against medium from C57BL/6- and MyD88^−/−-derived macrophage and TAM cultures stimulated with bacteriophages as described. Medium from nonstimulated macrophages, normal CM, B16.F10 medium, and medium with bacteriophages served as controls. In brief, medium was transferred to the lower chamber of a 96-well chemotaxis chamber (Neuroprobe), and a polycarbonate 5-μm pore filter (Neuroprobe) was inserted. Mouse neutrophils were isolated using the Anti-Ly-6G microbead kit (Miltenyi Biotec) and overlain (2 × 10⁵ cells) in the upper chamber and migration was allowed to take place for 4 h at 37°C. Next, the neutrophil suspension in the upper chambers was replaced with an equal volume of 2 mM EDTA in PBS and the migration chamber was placed at 4°C for another 30 min. After removal of all liquid from the upper chambers and washing of all cells from the upper part of the polycarbonate filter, migrated cells in the lower chamber were collected by centrifugation and counted.

B16-F10 melanoma cells were cultured at 2 × 10⁴ cells/well in flat-bottom 96-well plates. Medium from C57BL/6- and MyD88^−/−-derived macrophage and TAM cultures, before or after stimulation with bacteriophages, was added to the B16-F10 cells. In another set of experiments, B16-F10 cells were cocultured with neutrophils or both in the presence or absence of medium from stimulated macrophages or TAMs (1:10 target:effector ratio). Mouse neutrophils were isolated using the Anti-Ly-6G microbead kit (Miltenyi Biotec). After 24 h incubation, cells were harvested and the number of apoptotic tumor cells was determined by flow cytometry analysis using 7-AAD positivity (BD Biosciences).

To test the potential role of TLRs in triggering the immune response observed following tumor-specific phage treatment, we used MyD88^−/− mice, which cannot transduce TLR-mediated signals. Splenocytes from wild-type or MyD88^−/− mice were plated and stimulated with either LPS-free or crude phage preparations for 48 h. ELISA analysis revealed that IFN-γ secretion following bacteriophage stimulation was completely abrogated in MyD88^−/− splenocyte cultures (Fig. 1,A). Mice received s.c. injections of B16-F10 cells and tumor-specific phages were administered when tumors were palpable. Tumors were subjected to histological examination 24 h post treatment. As expected, administration of tumor-specific phages caused a massive infiltration of neutrophils and subsequent necrosis in tumors from wild-type mice (Fig. 1,B) whereas in MyD88^−/− mice, no or modest infiltration was observed (Fig. 1 C).

For a more detailed analysis of the effects of tumor-specific phage treatment on the tumor microenvironment, we performed gene expression profiling using a LDA customized for detection of 30 genes preferentially expressed by APCs and encompassing both M1 and M2 markers. The data analysis revealed that tumor-specific phage treatment in C57BL/6 mice resulted in up-regulation of several inflammatory cytokines such as TNF-α, IL-6, and IL-1β (14-, 14-, and 80-fold, respectively). A 9-fold increase of the costimulatory marker CD80 was also observed (Fig. 2). No up-regulation of genes, with the exception of IL-1 receptor antagonist (10-fold), was detected in MyD88^−/− mice after treatment with tumor-specific phages (Fig. 2). Down-regulated genes included matrix metalloprotease 14 (19-fold) and CD163 (6-fold). The M2-polarized tumor microenvironment before phage treatment was verified by an 8-fold higher expression of Arginase-1 compared with NO synthase 2 (data not included). In phage-treated tumors, this balance shifted toward a more M1-polarized milieu as demonstrated by a 14-fold increase of NO synthase 2 (Fig. 2). Other up-regulated genes included TLRs, MMPs, adhesion molecules, and Fc receptors (Supplementary Fig. 1).⁵

The data strongly indicate that TLR triggering play a major role for the phage-mediated tumor regression observed after tumor-specific phage treatment. As TLR signaling is a distinct feature of APCs, and macrophages (TAMs) comprise the majority of APCs in tumor stroma, we elaborated on the effect of bacteriophages on this cellular population.

To further investigate the effects of bacteriophages on TAMs, an in vitro system for differentiation of primary peritoneal macrophages to M2-polarized counterparts was developed. When primary macrophages were cultured with conditioned medium from B16-F10 melanoma cells, we observed an M2-like phenotype characterized by up-regulation of immunosuppressive genes such as IL-10 and TGF-β and down-regulation of the type-1 cytokines IL-12 and IFN-γ (Fig. 3,A). The addition of bacteriophages drastically altered the gene expression profile in TAM cultures to that of the M1 phenotype (Fig. 3 B) and levels were similar to those of normal macrophages stimulated with bacteriophages (Supplementary Fig. 2A).

To investigate cytokine expression at the protein level, C57BL/6 and MyD88^−/− TAM cultures were treated with bacteriophages and culture supernatants were analyzed for cytokine content. Phage administration resulted in 246- and 314-fold increases of the M1 type cytokines TNF-α and IL-6, respectively, compared with nonstimulated controls. A 7-fold increase of the chemokine CCL2 (MCP-1) was also observed. Changes in IL-10 secretion were minor (1.5-fold increase) (Fig. 3,C). Corresponding values for the analyzed cytokines from MyD88^−/− TAMs were 24-, 10-, 2-, and 1.2-fold increases, respectively (Fig. 3,C). A similar pattern of up-regulation of secreted cytokines was observed for normal macrophages (Supplementary Fig. 2B). Furthermore, both macrophages (data not included) and TAMs exhibited an increased expression of MHC class I/II and CD80 (Fig. 3 D) following bacteriophage stimulation. However, no increase in CD86 expression was observed. In preliminary experiments, the same phenomenon was also observed on bone marrow-derived mouse dendritic cells, with the exception that these also up-regulated CD86 after bacteriophage stimulation (data not included).

These data indicate that, regardless of macrophage activation status, bacteriophages are able to induce expression of cytokines which could alter the immunosuppressive tumor microenvironment in favor of an M1 phenotype and promote tumor toxicity. Overall, these data corroborate well with the in vivo LDA data. In addition, bacteriophages can stimulate APCs to express costimulatory molecules, which could also facilitate T cell priming within an immunosuppressive milieu.

In an attempt to explain the massive infiltration of neutrophils associated with tumor-specific phage treatment and whether the infiltration is an effect mediated by phages alone or by phage-stimulated macrophages, transmigration assays were performed. Medium from phage-stimulated TAMs attracted, although not significantly (p = 0.07, t test), higher levels of neutrophils compared with medium from nonstimulated TAMs and culture medium spiked with bacteriophages (Fig. 4). No enhanced migration was observed using medium derived from phage-stimulated MyD88^−/− TAMs (Fig. 4). These results clearly demonstrate that phages alone are not able to attract neutrophils to tumors and that the neutrophil recruitment is dependent on factors secreted by stimulated macrophages.

The effect of factors secreted by bacteriophage-stimulated macrophages on neutrophil cytotoxicity was tested. B16-F10 cells were cultured in medium from either nonstimulated or bacteriophage-stimulated TAMs in the absence or presence of neutrophils. We observed a clear increase in cell death of B16 cells cultured with medium from phage stimulated TAMs compared with medium from nonstimulated counterparts (Fig. 5,A). Addition of neutrophils significantly increased the amount of apoptotic cells and the increase was further enhanced when neutrophils were added to tumor cells in the presence of medium from stimulated TAMs (Fig. 5,A). No increase in tumor cell apoptosis was observed when using medium derived from phage-stimulated MyD88^−/−-derived TAMs (Fig. 5 B). These results demonstrate that although neutrophils are able to kill tumor cells without any support from macrophages, their cytotoxicity is enhanced after exposure to factors secreted by M13 bacteriophage-stimulated TAMs.

In the present study, we evaluated the underlying mechanisms of the phage-induced tumor destruction observed in our previous study (23). The mechanism of how bacteriophages trigger the immune system is still unclear. However, our results suggest that the mechanism is dependent on TLR signaling since cytokine production in MyD88^−/− splenocyte cultures and neutrophil infiltration within tumors in MyD88^−/− mice is lacking. Although deficiency in neutrophil migration in MyD88^−/− mice has been reported previously (34), no studies have demonstrated that bacteriophages may stimulate immune responses through TLR engagement. Gene profiling using a LDA further elucidated the involvement of TLR signaling in our system as no or very low up-regulation of inflammatory gene expression was observed in phage-treated tumors from MyD88^−/− mice. Furthermore, the LDA analysis demonstrated a shifted balance in arginine metabolism in favor of the Arginase-1 pathway in tumors before phage treatment, which is indicative of M2 polarization (35). Interestingly, in C57BL/6 mice following phage treatment, we observed a 19-fold down-regulation of MMP-14, which is suggested to be the rate-limiting protease responsible for the invasive activity of tumor cells (36). The absence of MMP-14 down-regulation in the tumors of MyD88^−/− mice indicates that these tumors are largely unaffected by the phage therapy, while the down-regulation in the C57BL/6 group may suggest a loss of tumor cell viability.

As macrophages (TAMs) are the major TLR-expressing cell population in tumors and can comprise >50% of the tumor stroma (37), a more detailed investigation of this cellular population was conducted. In an attempt to induce the M2 phenotype in vitro, primary mouse macrophages were cultured with B16-F10-conditioned medium. Compared with normal macrophages, these cells exhibited an M2 phenotype characterized by an up-regulation of IL-10 and TGF-β and down-regulation of IL-12 and IFN-γ. This concords with another report in which peritoneal macrophages cultured in B16-conditioned medium exhibited an impaired response to TLR4 signaling reflected by reduced TNF-α and NO production (38), which is reminiscent of M2 polarization.

Macrophage phenotype had no impact on the response to bacteriophage stimulation because both naive macrophages and TAMs expressed highly elevated levels of the proinflammatory genes IL-12, TNF-α, and IFN-γ. Cytokine expression at the protein level supports these results and also indicates that the secretion is impaired in the absence of the MyD88 adaptor protein. In addition, these results concur well with the LDA results, although one cannot rule out that cells other than TAMs may be responsible for some of the gene expression observed in vivo. Indeed, a similar trend was observed in preliminary experiments using dendritic cells. These findings are in agreement with other studies in which TLR ligands such as CpG and/or cytokines (e.g., IL-10 and IL-12) were successfully used to induce an M1 phenotype in alternatively activated M2 macrophages (20, 32, 39). Given these observations, it is apparent that bacteriophages are equally potent immunomodulators of macrophage activation state, but importantly they also possess specificity that allows homing to the tumor, thus minimizing the risk for systemic side effects. Furthermore, after coculture with phages, normal macrophages and TAMs, as well as bone marrow-derived dendritic cells, up-regulated molecules associated with Ag presentation and T cell costimulation. Such events may further improve the treatment efficacy through induction of tumor-specific adaptive cellular responses. In fact, we have observed that subsequent to tumor-specific phage therapy and complete tumor regression mice are protected against tumor rechallenge in a CD8⁺ T cell-dependent manner (our unpublished data).

Tumor-specific phage treatment is associated with a rapid infiltration of neutrophils into the tumor and subsequent tumor necrosis. It is well established that upon classical activation macrophages can destroy tumors through secretion of NO. Nevertheless, in our experimental setting only a moderate secretion of NO by macrophages was observed following bacteriophage stimulation (data not shown). The increase in tumor cell death in B16-F10 cultures after addition of medium from phage-stimulated TAMs was also equivalent to that apparent in control cultures with medium from nonstimulated TAMs. Thus, we do not consider the tumor regression by tumor-specific phage treatment to be directly mediated by either macrophages or macrophage-derived factors.

Our in vitro studies revealed that neutrophils, in a TLR-dependent manner, selectively migrated in response to factors secreted by TAMs and normal macrophages following phage stimulation, and that in the presence of such factors they exhibited an enhanced tumoricidal activity. Macrophages alone did not yield enhanced killing, indicating that neutrophils are the primary effector arm mediating tumor regression in this model. These findings agree with previous strategies using neutrophils in cancer immunotherapy. It was demonstrated that tumor cells engineered to secrete various cytokines such as TNF-α and IL-2 create a microenvironment which promotes neutrophil infiltration and subsequent tumor rejection (40). However, the tumor-specific phage based approach to direct neutrophils to tumor sites offers a greater potential as a treatment modality because the neutrophil recruitment does not require any in vitro manipulation of tumor cells.

In summary, the present study describes a close interplay between different compartments of innate immunity leading to the eradication of palpable B16 tumors. The bacteriophage-mediated tumor destruction occurs in a TLR-dependent manner. We also demonstrate that bacteriophages are able to switch the immunosuppressive phenotype of TAMs to a M1 phenotype, thereby triggering an innate antitumor response and stimulating up-regulation of molecules which facilitate priming of specific antitumor immunity. Furthermore, mouse neutrophils selectively migrate toward factors secreted by phage-stimulated macrophages and in such an environment, exhibit an increased tumoricidal activity. Strategies using tumor-targeted phage particles in cancer immunotherapy therefore warrant further investigation.

We thank Liss Garberg and Margareta Rodensjö at the Cancer Centre Karolinska pathology unit for conducting tissue sectioning and stainings.

The authors have no financial conflict of interest.