Address reprint requests to Dr I. N. Buhtoiarov, Department of Human Oncology, The University of Wisconsin School of Medicine and Public Health, K4/450 Clinical Science Center, 600 Highland Avenue, Madison, WI 53792–4672, USA. Email: email@example.com
Dr A. L. Rakhmilevich, Department of Human Oncology, The University of Wisconsin School of Medicine and Public Health, K4/413 Clinical Science Center, 600 Highland Avenue, Madison, WI 53792–4672, USA. Email: firstname.lastname@example.org Senior author: Dr A. L. Rakhmilevich
We explored the mechanisms of class B CpG-oligodeoxynucleotide-induced antitumour effects against weakly immunogenic tumours. Treatment with CpG-oligodeoxynucleotide 1826 (CpG) induced similar antitumour effects in B16 melanoma-bearing immunocompetent C57BL/6 mice and T-cell-deficient severe combined immunodeficient (SCID) mice, and NXS2 neuroblastoma-bearing T-cell-depleted A/J mice. Both macrophages (Mφ) and natural killer (NK) cells from CpG-treated C57BL/6 mice could mediate cytotoxicity in vitro, suggesting that these cell types might control tumour growth in vivo. However, CpG treatment of SCID/beige mice or T-cell-depleted and NK-cell-depleted A/J mice still induced antitumour effects in vivo, arguing against a major role of NK cells in the antitumour effects of CpG in the absence of T cells. In contrast, CpG treatment of interferon-γ knockout (IFN-γ–/–) C57BL/6 mice resulted in no antitumour effects in vivo and no Mφ-mediated tumoristasis in vitro despite unaltered cytolytic function of NK cells in vitro. Moreover, Mφ inactivation by silica substantially reduced CpG-induced suppression of tumour growth in vivo, revealing an important role of Mφ in CpG-induced antitumour effects. The in vitro tumouritoxicity by CpG-stimulated Mφ (CpG-Mφ) correlated with tumour cell mitochondria dysfunction and involved nitric oxide (NO), tumour necrosis factor-α (TNF-α) and IFN-γ, whereas interleukin-1α (IL-1α), IL-1β, IFN-α, TNF-related apoptosis-inducing ligand and Fas ligand played insignificant roles in CpG-Mφ tumouritoxicity. Taken together, our results indicate that the growth control of weakly immunogenic tumours during CpG-immunotherapy is mediated predominantly by Mφ, rather than T cells or NK cells.
Immune reactions against cancer cells are frequently based on the induction of adaptive immunity through activation of cytotoxic CD8+ T cells capable of recognizing tumour antigens in the context of their association with the major histocompatibility complex (MHC) class I molecules. T-cell activation can be augmented by additional stimulation of cells of the innate immune system with so-called pathogen-associated molecular patterns (PAMP), as has been shown both experimentally and clinically.1,2 Enhancing effects of PAMP on T-cell immunity are usually indirect and believed to act through either facilitating presentation of the tumour antigens to T cells,1,2 or through altering the tumour microenvironment, allowing the cells of adaptive immunity to migrate into the tumour to mediate tumour cell killing.3 Nevertheless, many cancer patients still succumb to progressive disease, even those patients who have received forms of immunotherapy that induce circulating T cells able to selectively recognize tumour antigen ex vivo.4 Immune escape by down-regulation of the MHC class I apparatus is one of several potential reasons that induction of specific T-cell responses to tumour antigen may not provide clinically meaningful tumour growth control.5
Induction of direct tumoritoxic effects by the cells of the innate immune system might help to induce antitumour effects through different recognition and destructive pathways than those used by T cells.6 Macrophages (Mφ), the effector cells of the innate immune system, can play an important role in antitumour immune reactions.7 While involved in tumour development and progression,8 Mφ can also be activated to destroy cancer cells via both cell contact-dependent and -independent mechanisms.9,10 We have recently documented that stimulation of Mφ via their CD40 molecules using an agonistic anti-CD40 monoclonal antibody (mAb; αCD40) induces them to be tumoritoxic in vitro and in vivo.11,12 These Mφ-mediated antitumour effects could be synergistically augmented by two distinct PAMP, the bacterial lipopolysaccharide (LPS) and the class B CpG-oligodeoxynucleotides (CpG-ODN), the synthetic analogues of bacterial DNA containing unmethylated cytidine-phosphate-guanosine (CpG) motifs.11,13 Notably, neither Mφ stimulation nor Mφ-mediated antitumour effects induced by αCD40 required the presence of T cells or natural killer (NK) cells because the level of retardation of tumour growth in fully immunocompetent C57BL/6 mice was similar to that observed in T-cell-deficient severe combined immunodeficient (SCID) mice, or T-cell-deficient, B-cell-deficient and cytotoxic NK-cell-deficient SCID/beige mice.11–13
While testing a combination of αCD40 plus CpG-ODN1826 (referred to as CpG in subsequent text), a prototypic class B CpG-ODN (CpG-B), for induction of synergistic antitumour effects in vivo, we found that treatment with CpG alone was able to induce in vivo growth retardation of weakly immunogenic B16 melanoma tumours.13 CpG-B has been recently demonstrated to induce regression of highly immunogenic tumours through the engagement of both innate and adaptive immune systems.14 In contrast, antitumour immune reactions induced by CpG-A do not appear to require CD8+ T cells, and therefore they were suggested to be mediated primarily by NK cells.15,16 In this study we characterized the mechanisms of CpG-B-induced antitumour immune effects against weakly immunogenic murine tumours. We show here that cytotoxic Mφ rather than T cells or NK cells play a central role in antitumour reactions against weakly immunogenic tumours during immunotherapy with CpG-B.
Materials and methods
Mice and tumour cell lines
C57BL/6 and A/J mice (Harlan Sprague Dawley, Madison, WI), interferon-γ knockout (IFN-γ–/–) C57BL/6 mice (The Jackson Laboratory, Bar Harbor, ME) and CB-17, CB-17 SCID and CB-17 SCID/beige mice (Taconic, Germantown, NY, or Charles Rivers, Wilmington, MA) were housed, cared for and used in accordance with the Guide for Care and Use of Laboratory Animals (NIH publication 86-23, 1985; National Institutes for Health, Bethesda, MD). Murine B16 melanoma, YAC-1 thymoma and L5178Y lymphoma cell lines were grown in RPMI-1640 complete medium, and the NXS2 murine neuroblastoma cell line was grown in Dulbecco's modified Eagle's medium complete medium at 37° in a humidified 5% CO2 atmosphere, as previously described.11,13
In vivotumour model and CpG-immunotherapy
Either 1 × 105 B16 cells or 2 × 106 NXS2 cells were injected subcutaneously (s.c.) in 0·05 ml Ca2+/Mg2+-free phosphate-buffered saline (PBS; Biosource, Rockville, MD) in the right abdominal flank in syngeneic C57BL/6 and IFN-γ–/– C57BL/6 or A/J mice, respectively, as well as in immunodeficient SCID or SCID/beige mice. Mice with palpable tumours received 0·1 mg of CpG: endotoxin-free completely phosphorothioate-modified CpG1826, 5′-TCCATGACGTTCCTGACGTT-3′(Coley Pharmaceuticals, Wellesley, MA) intraperitoneally (i.p.) in 0·5 ml PBS on days 5, 8, 11 and 14 after tumour implantation, unless otherwise indicated. Mice in the control groups received 0·5 ml PBS.13 Antitumour effects were evaluated by measuring the tumour size twice a week until day 17–24 after tumour implantation, when control mice were killed because of their moribund status. Tumour size (mm3) was calculated according to the formula: (π/6) × tumour length × (tumour width2).13
In vivoeffector cell depletion
A/J mice were depleted of T cells by i.p. injection with a mixture of 0·3 mg αCD4 (GK1.5) and 0·3 mg αCD8 (2.43), or were given 0·6 mg control rat immunoglobulin G (Sigma, St Louis, MO) on days 2, 5, 8, 11 and 14 after tumour implantation, as previously described.13 NK cells in A/J mice were depleted by giving them i.p. 0·05 ml anti-asialo-GM1 antibody (WAKO Pure Chemicals, Richmond, VA) or control rabbit immunoglobulin G (Sigma), on the above schedule and as previously described.13 Polymorphonuclear neutrophils (PMN) were depleted by αGR-1 antibody (RB-6), 0·3 mg i.p. on the above schedule and as previously described.13 The effectiveness of depletion of T cells and PMN was confirmed by flow cytometry, and of NK cells in the 4-hr 51Cr cytotoxicity assay against NK-cell-sensitive YAC-1 cells.12,17,18 Mφ were inactivated by i.p. injection of 30 mg of silicon dioxide (Sigma) in 0·5 ml PBS on days −1, 3, 7, 11 and 15 relative to tumour cell implantation, as previously described.13
In vitroMφ-mediated tumoristasis
Peritoneal cells were obtained by peritoneal cavity lavage from mice 3 days after treatment with 0·1 mg CpG1826 or non-CpG1982 (TCCAGGACTTCTCTCAGGTT, Sigma-Genosys). In preliminary experiments, no difference in Mφ stimulatory capacity of control non-CpG1982 and PBS (diluent) was observed; therefore, 0·5 ml PBS was used as control for CpG1826 in all subsequent experiments. Total peritoneal cells were seeded in 96-microwell cell-culture clusters at 3 × 106 cells/ml, 0·1 ml/well. After 90 min, non-adherent cells were removed from the culture by repeated pipetting followed by aspiration of the cell culture medium. This protocol yields a relatively pure population of mature Mφ, based on 95–97% expression of CD97 (F4/80).11 The resultant adherent PBS- and CpG-stimulated Mφ (PBS-Mφ and CpG-Mφ, respectively) (1·2 × 105 to 1·5 × 105/well) were thereafter incubated with tumour cells (1 × 104/well) for 24–48 hr in medium with or without 10 ng/ml LPS. In selected experiments, 10 μg/ml functional grade blocking antibodies to tumour necrosis factor-α (TNF-α; MP6-XT3), IFN-γ (R4-6A2), interleukin-1α (IL-1α; ALF-161), IL-1β (B122), TNF-related apoptosis-inducing ligand (TRAIL; N2B2), Fas ligand (FasL; MFL3) (eBioscience, San Diego, CA.), IFN-α (F18, Serotec, Raleigh, NC), and 5 mm of the inducible nitric oxide synthase (iNOS) inhibitor l-nitro-arginine-methyl esterase (l-NAME, Sigma), alone or in combination, were added into the cultures at the start of coculture with tumour cells. For the last 6 hr, tumour cells were pulsed with 1 μCi/well [3H]thymidine ([3H]TdR), and retained radioactivity was counted by γ-scintillation of total cells. Under these conditions Mφ incorporate negligible amounts of [3H]TdR, enabling the [3H]TdR to reflect the level of proliferation of the tumour cells.11 Results are presented as mean counts per 5 min for triplicate wells ± SEM.
Nitric oxide detection
PBS-Mφ and CpG-Mφ were cocultured in vitro with tumour cells in medium with or without 10 ng/ml LPS for 48 hr. Nitrite accumulation in the cell-culture supernatants was determined using the Griess reagent (Sigma).11
NK cytotoxicity assay
Spleens were collected from mice 3 days after treatment with PBS, CpG or mouse recombinant IL-2 (TECIN; Hoffmann-La Roche, Inc., Nutley, NJ, 50 000 U for 3 days, i.p.) and processed to a single-cell suspension. Erythrocytes were lysed by hypotonic shock. The 4-hr 51Cr-release cytotoxicity assay against YAC-1 cells was performed as described elsewhere.18 Results are presented as per cent of lysis of 5 × 103 YAC-1 cells cultured with splenocytes at multiple effector-to-target ratios (100 : 1, 50 : 1, 25 : 1).
Assay for tumour cell apoptosis
L5178Y cells (1 × 104 cells/well/0·2 ml) were cultured with or without PBS-Mφ or CpG-Mφ in medium with or without 10 ng/ml LPS for 24 hr. Total cells were harvested by pipetting, and stained for Mφ with αF4/80-APC or isotype-matched control mAb. Next, the cells were incubated for 15 min at 37° in 5% CO2 in PBS containing 40 nm of the redox potential-sensitive dye 3,3-dihexyloxacarbocyanine iodide (DiOC6(3); Sigma), which accumulates in the mitochondria of viable cells.19 After repeated washing, flow cytometry analysis of F4/80– L5178Y cells was performed on a FACScan flow cytometer with CellQuest software (BD, San Jose, CA) as described elsewhere.19
A two-tailed Student's t-test was used to determine the significance of differences between experimental and relevant control values within one experiment. The non-parametric Wilcoxon rank sum test (two-sided) was applied to determine difference between results of several similar experiments.
T cells and NK cells are not essential for antitumour effects during CpG-B-immunotherapy
The mechanisms of antitumour effects induced by CpG, when used alone as immunotherapy against weakly immunogenic tumours, are not fully understood. We found that systemic CpG-immunotherapy of tumour-bearing, immunocompetent C57BL/6 mice resulted in potent retardation of growth of B16 tumours (Fig. 1a). Similar antitumour effects were documented in B16 tumour-bearing SCID mice (Fig. 1b), suggesting that T cells are not essential for induced antitumour effects. In the absence of T cells, NK cells and Mφ are two major effector cell populations that might be mediating this antitumour effect.
Both NK cells and Mφ were activated in CpG-treated C57BL/6 mice. Three days after a single-dose CpG treatment, NK cells mediated 48% cytotoxicity of YAC-1 cells at an effector-to-target ratio of 100 : 1, compared to 20% killing by spleen cells from PBS-treated mice (Fig. 1c). Similarly, CpG-Mφ could substantially suppress the proliferation of B16 tumour cells in vitro. This effect was further enhanced by LPS (Fig. 1d). In contrast to CpG-Mφ, PBS-Mφ did not mediate any significant tumoristasis, even in the presence of LPS. LPS itself did not have any suppressive effect on tumour cell proliferation either. Hence, these results indicate that in vivo stimulation with CpG induces cytotoxic properties in Mφ. In agreement with this, supernatants from tumour cell–CpG-Mφ cultures contained 20 μm nitrite, an important metabolite of NO. In the presence of LPS, production of nitrite by CpG-Mφ increased up to 56 μm, whereas supernatants from cultures of tumour cells alone or tumour cells with PBS-Mφ in the absence or presence of LPS did not have any detectable nitrite (data not shown). When harvested 1, 2 or 3 days after treatment of mice with CpG, the Mφ obtained on day 3 demonstrated the best antitumour effects in vitro (not shown).
Surprisingly, despite enhanced NK activity in vitro by splenocytes from CpG-treated C57BL/6 mice, depletion of NK cells had no significant impact on CpG-induced antitumour effects in vivo. This was shown by depleting the NK cells in tumour-bearing mice with anti-asialoGM1 antibody (Fig. 2a) and by using SCID/beige mice (Fig. 2b). We used a model of NXS2 neuroblastoma cells growing in A/J mice to apply the regimen and conditions for NK-cell and T-cell depletion that we had established in a previous study with the NXS2 model and αCD40 (Fig. 2a).12 In addition, we wanted to confirm the antitumour effects of CpG against B16 tumour that are shown in Fig. 1(a) by using a distinct tumour model in a different strain of mice. Depletion of NK cells with anti-asialo-GM1 antibody, similar to depletion of CD4+ T cells and CD8+ T cells (90·5% and 93·7% reduction, respectively, as determined by flow cytometry), had no impact on the inhibition of growth of NXS2 tumours (Fig. 2a). In this experiment, CpG-immunotherapy (at this dose and schedule) of NK-cell-depleted or T-cell-depleted mice uniformly led to complete tumour inhibition; this effect was identical to that seen in CpG-treated animals without depletion of T cells and NK cells. As absence of T cells does not preclude involvement of NK cells in the antitumour effects, and vice versa, we repeated the experiment in B-cell-deficient, T-cell-deficient and cytolytic NK-cell-deficient SCID/beige mice and obtained similar results (Fig. 2b). Under these circumstances, the likely effector cells mediating antitumour effects in the absence of T cells and NK cells are cytotoxic Mφ. In this regard, Mφ from CpG-treated SCID/beige mice showed more cytostasis of B16 tumour cells in vitro in the presence of LPS than did Mφ from PBS-treated mice (Fig. 2c). In addition, these CpG + LPS-Mφ from SCID/beige mice produced significant amounts of nitrite at concentrations comparable to control CB-17 and C57BL/6 CpG + LPS-Mφ (data not shown).
Cellular mechanisms of in vivo antitumour effects induced by CpG-immunotherapy
The above experiments suggested that CpG-Mφ could mediate antitumour effects in the absence of T cells and NK cells. We further explored the role of Mφ in CpG-induced tumour growth control by implanting B16 tumours in IFN-γ–/– C57BL/6 mice because IFN-γ was previously shown to be important for the activation of Mφ with CpG-B.20 As expected, CpG-treatment of IFN-γ–/– C57BL/6 mice did not induce any substantial retardation of B16 tumour growth (Fig. 3a). In contrast, the wild-type IFN-γ+/+ C57BL/6 mice developed very strong antitumour effects. In concert with in vivo effects, CpG-Mφ from IFN-γ–/– C57BL/6 mice were less effective in suppressing the proliferation of B16 cells in vitro, even in the presence of LPS (Fig. 3b), compared to CpG-Mφ from IFN-γ+/+ C57BL/6 mice. However, NK cells from CpG-treated IFN-γ–/– C57BL/6 mice were still able to lyse YAC-1 cells at a level similar to that of splenocytes from CpG-treated IFN-γ+/+ C57BL/6 animals (Table 1; part A). Hence, our results confirm the essential role of IFN-γ in CpG-B-mediated activation of Mφ, as well as the importance of Mφ in CpG-induced antitumour effects.
Table 1. Role of IFN-γ and Mφ in the activation of NK cells during CpG-immunotherapy
Cytotoxicity (%; E : T)
Part A: NK-cell cytolytic activity remains unaffected in IFN-γ–/– C57BL/6 mice. Splenocytes from PBS-treated or CpG-treated IFN-γ+/+ C57BL/6 or IFN-γ–/– C57BL/6 mice were tested in vitro in a 4-hr cytotoxicity assay against 51Cr-pulsed YAC-1 cells. Results are presented as percentage of YAC-1 cells lysed by splenic NK cells. E : T, effector to target ratio.
Part B, Inactivation of Mφ with silica results in reduction of NK cytolytic activity in CpG-treated, but not IL-2-treated, mice. Splenocytes from naive, IL-2-treated mice (50 000 U/0·5 ml PBS i.p. on days − 2, − 1 and 0) and CpG-treated mice (100 μg/0·5 ml PBS i.p. on day 0) were tested for cytolysis of 51Cr-pulsed YAC-1 cells 3 days after treatment with CpG or after the last injection of IL-2. Selected groups of mice (n = 3 mice/group) were injected i.p. with silica on days − 7, − 3 and 1. Results are presented as percentage of YAC-1 cells lysed by splenic NK cells during a 4-hr coculture at various effector-to-target ratios.
IL-2 + PBS
IL-2 + silica
CpG + silica
We next attempted to inactivate Mφin vivo by treatment with silica, an approach that is selective for functional depletion of Mφ and other phagocytes.21 CpG-immunotherapy of B16 tumour-bearing C57BL/6 mice had a potent inhibitory effect on the tumours, whereas CpG-treatment of silica-treated mice led to near abrogation of the antitumour effect (Fig. 3c). Thus, data presented in Fig. 3(a,c) indicate that Mφ are the primary effector cells involved in CpG-induced control of B16 tumour growth. Notably, the cytolytic function of splenic NK cells from the silica-treated, CpG-treated mice was also decreased compared to the silica-untreated CpG-treated mice (Table 1; part B), suggesting that CpG-Mφ were involved in the activation of NK cells.21 To exclude non-specific damage of NK cells by silica, we injected recombinant IL-2, which directly stimulates NK cells, into mice that were either treated with silica or not. For comparison, separate groups of mice were injected with CpG. If NK cells were non-specifically damaged by silica, the cytolytic activity of IL-2-stimulated NK cells in the silica-treated mice would be attenuated relative to that in silica-untreated mice. The results in Table 1 (part B) show that NK cells from both silica-depleted and undepleted, IL-2-treated mice were similarly activated to lyse YAC-1 cells in vitro. In contrast, NK activity in CpG-treated mice was substantially reduced by silica treatment. These results suggest that NK cells were activated indirectly by CpG in vivo via the in vivo effects of CpG on Mφ. Altogether, these data demonstrate that growth control of weakly immunogenic B16 melanoma during CpG-immunotherapy is achievable even in the absence of T cells and NK cells via the induction of tumour-inhibitory Mφ.
Molecular mechanisms of antitumour effects by CpG-Mφin vitro
CpG-Mφ secrete a number of molecules that may be important for their biology and function.22 However, the relative role of these factors in CpG-induced Mφ-mediated tumoritoxicity has not yet been established. To address this question, we attempted to neutralize Mφ-derived cytotoxic factors, including TNF-α, IFN-γ, IFN-α, IL-1α, IL-1β, FasL, TRAIL and NO, in tumour cell–CpG-Mφ cultures at the initiation of Mφ–tumour cell coculture. As shown in Fig. 4(a), in the absence of any inhibitors, CpG + LPS-Mφ potently suppressed L5178Y tumour cell proliferation. PBS-Mφ, with or without in vitro stimulation with LPS, did not inhibit proliferation of the tumour cells in this experiment, and the results are not shown on the graph. Selective blocking of IFN-γ or NO caused a small to moderate reversal of tumour cell cytostasis mediated by CpG-Mφ with LPS. Inhibition of NO production (with l-NAME) had a greater impact on protecting tumour cells from CpG-Mφ with LPS than blocking of any other individual factor. In contrast, targeted neutralization of TNF-α, IFN-α, IL1-α, IL1-β, FasL, or TRAIL caused no detectable reversal of Mφ-mediated tumoristasis (Table 2). Combining αTNF-α, αIFN-γ and l-NAME together caused a 53% reduction in the CpG + LPS-Mφ-mediated tumoristasis. Simultaneous neutralization of all listed factors by the full complex of inhibitors (referred to as ‘Complex’) led to even greater (78%) reduction in tumour cytostasis by CpG + LPS-Mφ.
Table 2. Role of different Mφ-derived factors in cytotoxic and secretory function of CpG-Mφin vitro
Cytotoxicity c (%)
Reduction of cytotoxicity d (%)
Nitrite production n (μm)
Reduction of nitrite production e (%)
PBS-Mφ or CpG-Mφ from C57BL/6 mice were cocultured with tumour cells for 48 hr in medium with 10 ng/ml LPS in the absence or presence of indicated neutralizing mAbs and inhibitors. Antitumour effects were measured in a [3H]TdR incorporation assay, as described in the Materials and methods. In parallel, the culture supernatants were tested for NO metabolites in the Griess colorimetric assay. PBS-Mφ (either with or without LPS) had no suppressive effect on tumour cells, and therefore those results are not shown in the table. Similarly, effects by CpG-Mφ in the absence of LPS stimulation are not included in the table. The table is a summary of six separate experiments. Cytotoxicity, c, (as a percentage) towards tumour cells was calculated as c = (a–b)/a × 100%, where a is [3H]TdR counts emitted by tumour cells cultured alone, i.e. without Mφ, and b is counts from tumour cells cultured with Mφ. Reduction of cytotoxicity, d, (as a percentage) was calculated as d = (cm– cf)/cm × 100%], where cm is cytotoxicity (%) in medium alone, i.e. in the absence of any neutralizing agent, and cf is cytotoxicity in the presence of the neutralizing factor or inhibitor. Reduction of NO production, e, was calculated as e= (nm − nf)/nm × 100%, where nm is μm nitrate in medium alone and nf is μm nitrate in the presence of neutralizing factor or inhibitor. 1P > 0·05, 2P < 0·05 in the non-parametric Wilcoxon rank sum test (two-sided).
We have previously shown that tumour cells exposed to cytotoxic Mφ can die through apoptosis as revealed by morphological alterations in the tumour cell membranes.11,13 To characterize the mechanism of CpG-Mφ-mediated apoptogenic effects, we measured the mitochondrial membrane potential (ΔΨm)19 of L5178Y lymphoma cells cultured with CpG-Mφ. L5178Y tumour cells grow in suspension, allowing their retrieval for flow cytometry analysis without associated physical or chemical damage from the harvesting method. In addition, we sought to determine in this assay the roles of CpG + LPS-Mφ-derived cytotoxic factors that were found to be important for tumoristasis in vitro (Fig. 4a and Table 2). The results show that 51% of L5178Y cells had dissipated ΔΨm when cultured with CpG + LPS-Mφ for 24 hr (Fig. 4b). In contrast, minimal mitochondrial damage was seen in these same L5178Y cells when cocultured with LPS alone, PBS-Mφ, PBS + LPS-Mφ or CpG-Mφ; the percentage of L5178Y cells with dissipated ΔΨm varied from 2·1% for cells grown in medium alone to 13% for the cells exposed to CpG-Mφ (data not shown). Adding αTNF-α or αIFN-γ or l-NAME resulted in the partial protection of the tumour cells, as revealed by a lower percentage of tumour cells with dissipated ΔΨm (34, 32 and 30%, respectively, Fig. 4B). Combining αTNF-α, αIFN-γ and l-NAME resulted in dramatic protection from CpG + LPS-Mφ, causing the percentage of tumour cells with mitochondrial dysfunction to reach control levels (8%). Using the complete ‘Complex’ of inhibitors in tumour cell–CpG + LPS-Mφ cultures did not further reduce the percentage of cells with dysfunctional mitochondria (8%) but did further protect the ability of L5178Y cells to incorporate [3H]TdR (not shown).
Analyses of the antitumour properties of immunostimulatory CpG-ODN represent a rapidly expanding field of experimental and applied immunology. Based on the oligodeoxynucleotide backbone and the nucleotide bases flanking the CpG motifs, at least three classes of CpG-ODN can be distinguished structurally and functionally.23 Depending upon the class of CpG-ODN and the immunotherapeutic strategy, different kinds of immune effectors may become involved in antitumour responses. Thus, CpG-B was originally shown to directly stimulate B cells,24,25 dendritic cells,25 PMN26 and Mφ25,27in vitro, as well as to support the proliferation and survival of T cells.28,29 In contrast, in vitro stimulation of NK cells could be better achieved with CpG-A.30 Class C CpG-ODN combines the functional properties of both CpG-A and CpG-B.23
In experimental models of cancer immunotherapy, CpG-B have been successfully used as adjuvants together with tumour vaccines, augmenting the induction of tumour-specific T-cell-mediated immune responses.31 However, when CpG-B were administered as a monotherapy, the antitumour effects and effector mechanisms varied depending on the CpG-ODN sequence and dosage used, the tumour immunogenicity and localization, and the treatment mode, i.e. systemic treatment versus local delivery into the tumour vicinity. Thus, Baines et al.32 and Weigel et al.33 demonstrated that systemic monotherapy with CpG-B induced predominantly T-cell-mediated antitumour effects, whereas intra- or peri-tumour delivery of CpG-B was shown to induce antitumour effects through the activation of components of both the innate (Mφ, NK cells, dendritic cells) and the adaptive (CD4+ T cells and CD8+ T cells) immune systems.14,34,35 In addition, several studies using intra- or peri-tumour injections of CpG-B highlighted primary roles for NK cells, apparently activated through indirect mechanisms, in antitumour effects.35,36 However, some of these findings may not be conclusive, because anti-NK1.1 mAb, which is specific for C57BL/6, FVB/N and NZB mouse strains,37 would not be expected to effectively deplete NK cells in A/J mice.36 Finally, at least three studies indicated the involvement of Mφ in the antitumour effects of locally administered CpG-B38 and even CpG-A,39 alone or in combination with other immunomodulators40 and in the absence of T cells, B cells and cytotoxic NK cells. Thus, despite evidence of the involvement of several types of immune effectors in the antitumour effects of CpG-ODN, the relative contributions of T cells, NK cells and Mφ in the antitumour effects following systemic CpG-B administration to mice bearing weakly immunogenic syngeneic tumours have not been analysed in a comprehensive study.
In this study we extended our previous findings describing the synergy between CD40 ligation and CpG13 by comparing the roles of different effector cells in immune effects against weakly immunogenic tumours following systemic therapy with CpG. The results demonstrate, to the best of our knowledge for the first time, that Mφ play the central role in CpG-induced antitumour effects. While T cells and NK cells could be activated and therefore involved, they are not essential for the CpG-induced tumour growth retardation. Thus, almost identical levels of tumour growth inhibition were achieved in CpG-treated B16 melanoma-bearing C57BL/6 mice and SCID mice, indicating that T cells were not essential in these antitumour effects. Although NK cells were activated upon CpG treatment to destroy tumour cells in vitro, the elimination of NK cells from NXS2 tumour-bearing mice by anti-asialo-GM1 antibody did not lead to a reduction of tumour growth suppression. This observation was reinforced by CpG-immunotherapy of B16 melanoma-bearing SCID/beige mice, lacking both T cells and effector NK cells (Fig. 2b). The level of tumour growth inhibition was almost identical in C57BL/6 mice (71·2%Fig. 1a), SCID mice (76·5%, Fig. 1b) and SCID/beige mice (79·8%, Fig. 2b), as measured on day 17 after tumour cell implantation. Statistical analyses of these comparisons showed no significant differences in the antitumour effects induced by CpG: P = 0·07 for Fig. 1(a) versus Fig. 2(b); P = 0·34 for Fig. 1(a) versus Fig. 1(b); and P = 0·32 for Fig. 1(b) versus Fig. 2(b). While PMN might also be activated by CpG, they too are not essential for this CpG-induced antitumour effect, based on our studies depleting B16-melanoma-bearing C57BL/6 mice of PMN with αGr1 mAb (data not shown). As in the B16 model, neither T cells nor NK cells were found essential in CpG-induced antitumour effects in mice with NXS2 neuroblastoma. Thus A/J mice depleted of T cells or NK cells were still able to reject the tumour (Fig. 2a). In agreement with this, the CpG-induced antitumour effects were not associated with the induction of memory T-cell antitumour immunity because by day 35 (i.e. 21 days after CpG-treatment had concluded in Fig. 2a) the NXS2 tumours recurred in all T-cell-depleted A/J mice and in four of the six mice from NK cell-depleted and T-cell/NK-cell-undepleted groups (data not shown). Furthermore, tumour re-challenge into all tumour-free mice in Fig. 2(a) on day 35 after initiation of the experiment resulted in rapid tumour progression and death (data not shown). Together, the results with cell subset depletion suggest that CpG-Mφ are the cells that can control tumour growth in the absence of T cells and NK cells. Indeed, Mφ from CpG-treated SCID/beige mice could suppress B16 cell proliferation in vitro at a level similar to that of CpG-Mφ from control CB-17 mice, especially in the presence of small amounts of LPS (86·5% versus 91·4%, P = 0·1001, Fig. 2c). Similar to αCD40 in our previous study,11 CpG reproducibly activated Mφ, which exhibited varying degrees of direct in vitro antitumour effects and consistent priming to LPS. Several previous reports demonstrated synergistic effects between CpG-B and LPS, e.g. in NO production.41 We extended these findings by demonstrating their synergy in activating Mφ to mediate antitumour effects.
The central role of Mφ in CpG-induced antitumour effects was addressed in experiments with IFN-γ–/– C57BL/6 mice. It was previously demonstrated that IFN-γ is essential for the activation of Mφ with ligands to various stimulatory receptors, including the TLR9 ligand for CpG.11,13,20 As expected, minimal antitumour effects were induced by CpG in IFN-γ–/– C57BL/6 mice in contrast to the wild-type IFN-γ+/+ C57BL/6 mice, which demonstrated potent retardation of tumour progression (Fig. 3a). In agreement with the in vivo data, IFN-γ–/– CpG-Mφ mediated significantly reduced cytotoxicity in vitro in comparison to IFN-γ+/+ CpG-Mφ (97·5% versus 55·1%, Fig. 3b). The remaining cytotoxic activity (∼40%) could be explained by the existence of IFN-γ-independent autocrine and paracrine pathways of Mφ stimulation, involving IL-12, TNF-α, etc.42 Significant attenuation of CpG-Mφ-mediated cytotoxicity in the absence of IFN-γ could explain the absence of antitumour effects in vivo in IFN-γ–/– C57BL/6 mice, despite the unaltered cytotoxic function of NK cells (Table 1). This striking dissociation of alterations of the cytotoxic functions of NK cells and Mφ in IFN-γ–/– C57BL/6 mice unveils several fine aspects of the immune network between components of innate immunity and confirms the non-essential role of NK cells in antitumour reactions induced by CpG-immunotherapy against weakly immunogenic tumours.
The role of Mφ in the CpG-induced antitumour effects was confirmed by Mφ inactivation with silica. Silica treatment alone had no significant impact on tumour progression, whereas it virtually abrogated the in vivo antitumour effects of CpG against B16 melanoma (Fig. 3c). Interestingly, the tumoritoxic function of NK cells from CpG-treated, silica-depleted mice was reduced to the level of cytotoxicity of NK cells from control, CpG-untreated mice, suggesting that silica treatment could prevent CpG-induced activation of NK cells as well. However, a direct toxic effect of silica on splenic NK-cell function was ruled out by treating mice with silica and recombinant IL-2 which directly activates NK cells. These IL-2-activated NK cells from both silica-treated and untreated mice showed similar levels of tumour cell lysis in vitro, confirming the indirect mechanism of suppression of CpG-induced NK-cell activation in silica-treated mice (Table 2) and suggesting the importance of Mφ in CpG-induced activation of NK cells.43 As CpG-B does not directly stimulate NK cells, these results suggest that CpG activates Mφ, which in turn activate NK cells, similar to our observations of αCD40 therapy.44 Activation of NK cells by CpG-Mφ through the release of IL-12 may account for the unaltered NK-cell cytolytic function in IFN-γ–/– C57BL/6 mice in our experiments because production of IL-12 precedes IFN-γ secretion in the course of Mφ stimulation44 and is unaffected in IFN-γ–/– C57BL/6 mice.42 Dendritic cells can also secrete IL-12 in response to CpG-B stimulation.45 Thus, dendritic cell-derived IL-12 might account for the retained cytolytic activity (∼ 40%) of NK cells from Mφ-depleted and CpG-treated mice at the 100 : 1 effector-to-target ratio (Table 1). Whether IL-12 is the only soluble factor participating in indirect activation of NK cells during the course of immunotherapy with CpG-B remains to be determined, as does the role of Mφ, dendritic cells and B cells in IL-12 production following direct stimulation with CpG-B. The results showing that silica treatment, but not NK-cell depletion, abrogates the in vivo antitumour effect of CpG demonstrate that Mφ are the major antitumour effector cells of CpG therapy in our models.
Whereas the potential role of cytotoxic Mφ as a cancer immunotherapy has been studied for decades, the mechanisms of Mφ-mediated tumoritoxicity in vivo remain poorly understood. Different approaches for activating cytotoxic Mφ provide unique profiles of cytotoxic factors capable of inducing tumour cell death. CpG-B has been recently demonstrated to induce the expression of several factors22 that might be involved in CpG-Mφ-mediated tumour cell killing. We explored the relative role of some of these factors, IFN-α, IFN-γ, IL-1α, IL-1β, TNF-α, NO, FasL and TRAIL, in in vitro tumoristasis by CpG-Mφ. Whereas the antitumour activity of CpG-Mφ in medium was highly variable [from 89·2% of tumour proliferation inhibition in Fig. 1(d) to 0·6% in Fig. 4(a), average 45·1 ± 25·6% in six similar independent experiments], this antitumour activity was reproducible and nearly complete (96·2 ± 1·21%, Table 2) when CpG-Mφ were cocultured with tumour cells in the presence of low concentrations (10 ng/ml) of LPS. Among the above listed cytotoxic factors, blocking of NO production by CpG + LPS-Mφ had the most significant inhibitory impact on tumoristasis in vitro. Whereas blocking of TNF-α and IFN-γ alone did not result in a statistically significant decrease of CpG + LPS-Mφ-mediated cytotoxicity (Table 2), combining their blocking with l-NAME led to a dramatic decrease of antitumour effects in vitro, suggesting either additive or synergistic effects between these molecules. Indeed, the presence in medium of either αTNF-α or αIFN-γ mAb led to a significant decrease of NO in these studies (data not shown), which appears to be the major cytotoxic factor in this model. In this respect, the role of IFN-γ is more pronounced than the impact of TNF-α. Interestingly, simultaneous blocking of αTNF-α and αIFN-γ in the tumour cell–Mφ cultures resulted in the inability of CpG + LPS-Mφ to produce NO, and was also associated with a substantial decrease in the tumour cell proliferation inhibition in vitro, similar to that induced by the triple combination of αTNF-α + αIFN-γ + l-NAME (data not shown). In addition, our preliminary results show that culturing CpG-Mφ in serum from the same CpG-treated mice resulted in the induction of NO (yielding production of ∼20 μm nitrite, which is comparable to the amounts of nitrite produced by CpG-Mφ cultured in medium containing 10 ng/ml LPS). Adding αTNF-α plus αIFN-γ mAbs abrogated the stimulatory effect of the serum from CpG-treated mice (data not shown). These results suggest that TNF-α and IFN-γ play a role in vivo as endogenous stimulants supporting the tumoritoxic function of CpG-Mφ. Thus, it is possible that a potent immunostimulatory effect of LPS on CpG-Mφ observed in vitro reflects the immunoactivating effect of endogenous TNF-α and IFN-γ produced in response to CpG in vivo. IL-1α, IL-1β, IFN-α, FasL and TRAIL appeared to make minimal contributions to the overall tumoritoxic effects because blocking of these molecules as single factors had only negligible inhibitory effects (Table 2). However, combined blocking of these five cytotoxic factors along with NO, TNF-α and IFN-γ had the greatest inhibitory effect on the ability of CpG + LPS-Mφ to suppress B16 and L5178Y cell proliferation, suggesting that at least one of these factors is acting additively or synergistically with NO, TNF-α and/or IFN-γ. It is not clear at this time if the local activation of peritoneal Mφ with CpG in vivo can trigger enough soluble effector molecules to mediate systemic antitumour effects, if systemically injected CpG can induce these molecules in Mφ infiltrating subcutaneous tumours, or if in vivo administration of CpG also induces other Mφ-mediated tumoritoxic pathways that are important in vivo, yet not evaluated in these in vitro assays.
We have recently demonstrated that tumour cells exposed to tumoritoxic Mφ express several morphological signs of apoptosis, such as tumour cell membrane depolarization and phosphatidylserine expression, as well as alteration of membrane integrity associated with increasing membrane permeability to the DNA-binding dye 7-actinomycin-D.11,13 However, these signs are non-specific and could be observed in cells undergoing programmed death via various mechanisms.46,47 Some of the factors secreted by Mφ have been shown to be involved in mediating caspase-independent death via damage of mitochondria.47 We assessed tumour cells cultured with or without Mφ for their mitochondrial function by staining with DiOC6(3). After 24 hr of coculture in vitro, 51% of tumour cells cultured with CpG + LPS-Mφ demonstrated dissipation of ΔΨm that could be prevented by blocking TNF-α, IFN-γ and NO in combination in tumour cell–Mφ culture medium. Notably, blocking of any one of these factors caused similar incomplete protective effects on the tumour cells, whereas blocking of all three factors simultaneously resulted in complete protection of tumour cells from mitochondrial damage. In this assay, the presence of neutralizing antibodies to IL-1α, IL-1β, IFN-α, FasL and TRAIL, in addition to the inhibitors of TNF-α, IFN-γ and NO, did not further reduced the number of tumour cells with dysfunctional mitochondria. However, the parallel [3H]TdR assay revealed a similar, but not identical, pattern of protection of the tumour cells' ability to uptake [3H]TdR. In particular, blocking of NO, TNF-α or IFN-γ had unequal effects on [3H]TdR incorporation, with the most potent protection observed with the NO inhibitor (data not shown). The complete complex of inhibitors to all tested cytotoxic factors had the maximal protective effect on [3H]TdR incorporation function. Thus, the alteration of mitochondria appears to be only one of several possible mechanisms by which antitumour Mφ can damage tumour cells. Whether this mechanism is uniform for all kinds of tumoritoxic Mφ, or whether it is unique to CpG-Mφ, is currently under investigation in our laboratory.
In conclusion, this study documents the essential role of Mφ in mediating CpG-B-induced antitumour effects for the tumours studied here and addresses the mechanisms involved. Whether the results from these murine models can be translated to other weakly immunogenic murine tumours or to human cancers, remains to be determined. Infiltration of human cancer lesions with Mφ has an impact on prognosis, treatment efficacy and survival.7,8 The presence of these Mφ in human tumours also provides a rationale for attempting to activate their antitumour potential as active immunotherapy. Whereas resting human Mφ have been shown in some studies to be negative for TLR9,48 several other reports suggest that human Mφ express mRNA for TLR949 and can be activated with CpG-B to mediate antibacterial functions.50 Our study suggests that activating tumoritoxic Mφ might be an additional goal for clinical immunotherapies using CpG-ODN, and this potential could be evaluated in ongoing and proposed clinical trials.
We are grateful to Drs Jacquelyn A. Hank, Jacek Gan, Hillary D. Lum, Erik Johnson, Paul J. Bertics and Gideon Berke for helpful discussions of the results of this study. This work was supported by NIH-NCI grants CA87025 and CA32685 (to P.M.S. and A.L.R.), by grants from the Midwest Athletes Against Childhood Cancer (MACC) Fund (to I.N.B., P.M.S. and A.L.R.) and by a grant from The Cure Kids Cancer Coalition (CKCC) (to I.N.B. and P.M.S.).