Tumors routinely attract and co-opt macrophages to promote their growth, angiogenesis, and metastasis. Macrophages are also the key effector cell for mAb therapies. Here we report that the tumor microenvironment creates an immunosuppressive signature on tumor-associated macrophages (TAM), which favors expression of inhibitory rather than activating Fcγ receptors (FcγR), thereby limiting the efficacy of mAb immunotherapy. We assessed a panel of TLR and STING agonists (a) for their ability to reprogram macrophages to a state optimal for mAb immunotherapy. Both STINGa and TLRa induced cytokine release, modulated FcγR expression, and augmented mAb-mediated tumor cell phagocytosis in vitro. However, only STINGa reversed the suppressive FcγR profile in vivo, providing strong adjuvant effects to anti-CD20 mAb in murine models of lymphoma. Potent adjuvants like STINGa, which can improve FcγR activatory:inhibitory (A:I) ratios on TAM, are appealing candidates to reprogram TAM and curb tumor-mediated immunosuppression, thereby empowering mAb efficacy. Cancer Res; 77(13); 3619–31. ©2017 AACR.
Tumors can suppress the innate and adaptive arms of the immune system through regulation of myeloid cells (1, 2). Central to this suppressive capacity is the regulation of macrophages. Macrophages that have differentiated through interaction with tumor cells play a key role in exerting local immunosuppression and promoting tumor metastasis, neoplastic invasion of ectopic tissue, and angiogenesis (3). Although the description of macrophage activation is currently contentious, these tumor-promoting macrophages have been proposed to be akin to IL4/13–stimulated, anti-inflammatory “M2” macrophages generated during wound healing that orchestrate Th2 responses and promote tissue repair and remodeling (4). Polarization of TAM toward an LPS/IFNγ–activated, inflammatory “M1” state provides the potential to arrest these tumor-promoting activities and alleviate immunosuppression (5). Reagents capable of achieving this have the capacity to provide antitumor effects, particularly as an adjuvant to immunotherapy, and are keenly sought in cancer therapy (6).
Toll-like receptor agonists (TLRa) are potent stimulators of innate immunity, acting as “danger” signals that elicit phenotypic, secretory, and transcriptomic changes in macrophages consistent with immune activation (7, 8). Several studies have shown that TLRa can provide adjuvants effects in human cancers; TLR9a CpG was shown to be feasible, safe, and powerful to induce objective clinical response in lymphoma patients (9), and TLR7a imiquimod was effective in the treatment of vulvar intraepithelial neoplasia (10). Efficacy of TLRa in combination with mAb therapy in animal models has also been explored. Agonistic anti-CD40 mAb combined with imiquimod induced a systemic antitumor CD8+ T-cell and type I IFN response, significantly delaying the growth of implanted tumors and prolonging animal survival in models of mesothelioma (11) and melanoma (12). In our own studies (13), we have shown that TLR3a polyinosinic:polycytidylic acid (Poly I:C) augmented the agonistic activity of anti-CD40 mAb dependent upon the upregulation of activating Fc-gamma receptors (FcγR). Although TLRa-based reagents have been investigated in combination with antibody immunotherapy in humans (14), and such studies have motivated ongoing trials, they are yet to be successfully translated into the clinic (15).
Cyclic dinucleotides are a new class of immune adjuvants being evaluated in preclinical studies and in an early-phase clinical trial in patients with advanced/metastatic solid tumors (NCT02675439).They signal via stimulator of interferon genes (STING), which is crucial for sensing DNA viruses (16, 17). Cytosolic DNA activates STING, leading to phosphorylation of IRF3 via tank-binding kinase 1 (TBK1), and subsequent transcription of type I IFN genes (17–19). In vivo studies have shown that STING−/− or IRF3−/− mice fail to prime T cells against tumor antigens and do not reject immunogenic tumors (20), emphasizing a critical role for host STING in immune sensing of tumors through dendritic cell (DC) activation and T-cell priming (21, 22).
Here we document a comprehensive analysis of the efficacy of STING and TLR agonists in promoting macrophage proinflammatory activation, and tumoricidal function in combination with mAb immunotherapy. Certain TLRa were highly potent at activating both human and murine macrophages and augmenting antibody-dependent cellular phagocytosis (ADCP) in vitro, but did not elicit similar potency in murine in vivo models of normal and malignant B-cell depletion. However, STINGa were potent both in vitro and in vivo, crucially reversing lymphoma-mediated immunosuppression and providing protection in tumor-bearing mice where immunotherapy alone failed. STINGa, but not TLRa, were subsequently shown to effectively reverse the suppressive effects induced by the lymphoma on macrophage FcγR expression, the principal immune effector cells in vivo.
Materials and Methods
Clinical samples and ethics
Ethical approval was obtained by Southampton University Hospitals NHS Trust from Southampton and South West Hampshire Research Ethics Committee. Informed consent was provided in accordance with the Declaration of Helsinki. Chronic lymphocytic leukemia (CLL) samples were from Human Tissue Authority licensed University of Southampton, Cancer Sciences Unit Tissue Bank and leukocyte cones from Southampton General Hospital National Blood Service.
Mice were bred and maintained in local facilities and experiments approved by the local ethical committee under Home Office license PPL30/2964. Experiments conformed to the Animal Scientific Procedure Act (UK). hCD20Tg, γ-chain−/− and FcγR-null mice (23, 24) have been described with genotypes confirmed by PCR and/or flow cytometry.
In vivo B-cell depletion (adoptive transfer) assays
Splenocytes from target hCD20Tg (T) and nontarget (NT) wild-type (unless otherwise specified) mice were labeled with 5 μmol/L and 0.5 μmol/L CFSE (Invitrogen), respectively, mixed (1:1), and intravenously injected into recipients (5–8 × 106 cells/mouse). Two doses of adjuvant [TLR1/2a Pam3CSK4: 10–100 μg, TLR2/4a LPS: 10 μg, TLR3a Poly I:C: 100 μg, TR7/8a R848: 2.5 μg, DMXAA: 400 μg, type I IFN: 10,000 IU] were administered at 24 and 48 hours intraperitoneally, followed by Ritm2a (25–50 μg) or isotype control intravenously. For BALB/c mice, DMXAA was administered only once (300 μg at 24 hours). Spleen was harvested 16–20 hours after Ritm2a administration, splenocytes stained with anti-mouse CD19 APC (eBioscience), and assessed for T:NT ratio. IFNAR-blocking mAb (clone MAR1-5A3; Leinco) was given at 500 μg i.p. as indicated in the figure legend.
Generation and polarization of human monocyte-derived macrophages and murine bone marrow–derived macrophages
Human monocyte-derived macrophages (hMDM) and murine bone marrow–derived macrophages (mBMDM) were generated as described in refs. 23, 25. For hMDM polarization, cells were stimulated with TLRa or STINGa (Invivogen) for 48 hours [LPS: 50 ng/mL, recombinant human (rh) IFNγ: 2 ng/mL (PeproTech), rhIL4: 10 ng/mL (PeproTech), rhIL13: 10 ng/mL (PeproTech), TLR1/2: Pam3CSK4 0.1 μg/mL, TLR3: Poly I:C 40 μg/mL, TLR4: Monophosphoryl Lipid A (MPLA): 5 μg/mL, TLR5: Flagellin: 125 ng/mL, TLR7/8: R848 1 μmol/L, 2′2-, 2′3′-, 3′3′-cGAMPs: 10–50 μg/mL; type I IFN: 5–100 ng/mL]. For polarization of mBMDMs, cells were stimulated with recombinant murine (rm) IFNγ [2 ng/mL (Peprotech)], rmIL4 [10 ng/mL (Peprotech)], rmIL13 [10 ng/mL (Peprotech)], and TLRa or STINGa overnight (similar concentration as in hMDM).
Phenotypic analysis of hMDM/mBMDM and calculation of FcγR activatory:inhibitory ratio
Human and murine FcγR staining is described elsewhere (26). Fluorescently conjugated mAbs were from BD Biosciences, AbDSerotec, eBioscience, or made in-house. hMDMs were stained with anti-human CD40 Alexafluor (AF)488 (Clone ChiLob 7/6), CD38 AF488 (Clone AT 13/5), both in-house, and CD11b PE (eBioscience). mBMDMs/Splenocytes were stained with anti-mouse CD11b PE, Ly6C PerCpCy5.5, Ly6G PeCy7 (eBioscience), and F4/80 APC (AbD Serotec). Samples were acquired on FACScalibur/canto II (BD Biosciences) and data analyzed with FCS express (DeNovo Software). FcγR activatory:inhibitory (A:I) ratio for hMDMs was calculated as: MFI for FcγRI × FcγRIIA × FcγRIII/FcγRIIB (FcγRI × FcγRIII × FcγRIV/FcγRII for mBMDMs) giving a value of x for NT macrophage. The results for each test condition were then divided by x so that unstimulated macrophages received a ratio of 1.
hMDM and mBMDM phagocytosis assay and phagocytic index
Phagocytosis assay was performed as described in refs. 23 and 25. hCD20 transgenic murine B cells and human CLL cells were used as targets for mBMDM and hMDM, respectively. Phagocytic index was calculated by dividing the percentage of phagocytic macrophages under the test condition by the percentage of phagocytic macrophages seen in the unstimulated condition.
Supernatant from macrophage cultures was harvested and levels of cytokines (IFNγ, TNFα, IL12p70 and IL6) assessed by MSD V-Plex assay (Meso Scale Discovery) according to the manufacturer's instructions. Type I IFNs, IFNα and IFNβ, were measured by ELISA kit (PBL Assay Science).
BCL1 lymphoma model and therapy
On day 0, 8- to 12-week-old female WT or FcγR null BALB/c mice were injected in tail vein with 1 × 104 BCL1 tumor cells. DMXAA (300 μg) and anti-CD20 18B12 (200 μg, produced in-house from patented published sequences) were administered as indicated in Fig. 7A. Anti-CD8 antibody YTS169 (500 μg, in-house) was injected intraperitoneally on day 0, 5, 10, and 14. Tumor-bearing mice were culled humanely before reaching terminal endpoint.
Chemokine and cytokine gene expression
RNA was purified from mice spleen using Qiagen RNAeasy Mini Kit. RNA (500 ng) was used to synthesize cDNA using RT2 First Strand Kit and gene expression was assessed using RT2 Profiler PCR mouse cytokine and chemokine array kit (Qiagen).
Statistical analysis was performed using GraphPad Prism. To compare differences between the experimental groups, Student t test, Wilcoxon, paired or unpaired t test analyses were performed. Kaplan–Meier curves were produced and analyzed by Log rank (Mantel–Cox) test. A P value <0.05 was considered significant at the 95% confidence interval. Asterisks denote statistical significance (*, P < 0.05; **, P < 0.01; ***, P < 0.001; and ****, P < 0.0001).
Tumors can generate a suppressive microenvironment with a low FcγR A:I ratio on macrophages, leading to mAb therapy resistance
To explore how tumors can regulate their microenvironment in vivo, we performed adoptive transfer experiments (24) in mice harboring the syngeneic mouse lymphoma, BCL1 (27). The ability of the anti-human CD20 mAb, Ritm2a to deplete adoptively transferred target human CD20 transgenic (hCD20 Tg) B cells was examined (24). Whereas depletion of target cells was efficient in mice lacking tumor, in mice implanted with even low numbers (1 × 104) of BCL1 cells 9 days prior to mAb administration, depletion was completely abrogated (Fig. 1A). As the lymphoma cells themselves lack the hCD20 target and are not deleted in this system, this directly demonstrates the ability of the tumor cells to elicit an immunosuppressive microenvironment. Analysis of cytokine and chemokine gene expression showed that inoculation of tumor upregulated expression of IL10 and IL21 (Supplementary Fig. S1A). IL10 has previously been directly implicated as a major immunosuppressive component of BCL1 lymphoma (28, 29) and IL21 has recently been identified to induce B cells to produce IL10 (30). Previously, we have shown that mice depleted of macrophages are unable to eliminate target B cells (24), implicating macrophages as the most probable cells affected by this suppressive tumor microenvironment. We and others have shown that FcγR expression profiles on these cells are important for determining mAb efficacy (24, 25, 31) Therefore, next we investigated the differences in FcγR expression and A:I ratio (32) induced on splenic macrophages following BCL1 inoculation. Tumor became detectable in the spleen only after day 14 postinoculation (Fig. 1B) with numbers significantly correlating with spleen weight (Fig. 1C). Increasing tumor presence corresponded to a significant decrease in expression of the activatory FcγRs I and III on splenic macrophages. Over the same time-course, FcγRIV was marginally elevated, while the inhibitory FcγRII first showed an initial (day 7), small reduction and was then rapidly and substantially upregulated, approximately 4-fold (Fig. 1D). This resulted in an overall, statistically significant suppression of macrophage FcγR A:I ratio by day 14 (Fig. 1E), which persisted to day 21. Further evidence of the ability of the tumor to manipulate macrophage activation status was observed by measuring F4/80, CD40, and CD11b (Supplementary Fig. S1B). These data illustrate the profound effects that the tumor can elicit on macrophage FcγR expression profiles and phenotypic markers culminating in a substantially lowered A:I ratio and resistance to antibody-mediated target cell depletion (Fig. 1A); even when very few tumor cells were detectable (e.g., at day 7 and 14). Notably, and in accordance with their redundancy in mediating target cell clearance in the adoptive transfer model (Supplementary Fig. S1C–S1E), monocytes and neutrophils did not show such profound changes in their FcγR A:I ratios (Fig. 1E).
The inhibitory FcγRII has been previously shown to suppress antibody-mediated depletion (33). We therefore assessed the contribution of elevated FcγRII levels to the defective mAb depletion in our system. We found that FcγRII knockout mice harboring BCL1 lymphoma treated with anti-CD20 mAb demonstrated enhanced (∼50%) deletion of target cells compared with WT mice (Fig. 1F). These results show that a significant proportion, but not all of the effects of the tumor immune suppression, is due to elevations in FcγRII, and strategies to enhance FcγR A:I would likely augment mAb activity.
FcγR changes induced in mBMDMs by TLRa and STINGa
Having established that FcγR profiles and A:I ratios on TAM were the key determinants of mAb-mediated target cell deletion, we sought to identify suitable TLR and STING ligand reagents capable of modifying these properties. The majority of TLRa tested did not show any activity with mBMDMs, either in terms of modulating FcγR expression (Fig. 2A), increasing the FcγR A:I ratio (Fig. 2B), or augmenting phagocytosis (Fig. 2C). The sole exception was TLR3a (Poly I:C), which typically showed an increase in FcγR A:I ratio coincident with an increase in phagocytosis (Fig. 2B and C). In contrast to TLRa, both human (2′2′-, 2′3′- and 3′3′- cGAMP) and murine (DMXAA) STINGa showed increases in expression of activating FcγRs I, III, and IV with no changes in FcγRII (Fig. 2A), culminating in an increased FcγR A:I ratio (Fig. 2B), and significant increases in phagocytosis (Fig. 2C; Supplementary Fig. S2). Notably, we observed a correlation between phagocytic activity of macrophages with their FcγR A:I ratio (Fig. 2D, R2 = 0.48, P = 0.0041).
Stimulation of mBMDMs with STINGa also induced potent type I IFN-α and -β responses, with TLR3a again being the only TLRa to induce the secretion of these cytokines (Fig. 2E).
Phenotypic and cytokine changes induced in hMDMs stimulated with TLRa or STINGa
We next examined whether our observations pertaining to mouse macrophages could be translated to human. We first established that hMDMs stimulated with TLRa displayed changes in a range of phenotypic markers; CD40, CD80, CD14, MHCII, CD206, and CD11b that were akin to changes induced by LPS/IFNγ treatment (Supplementary Fig. S3). We then investigated responses to a broader panel of both TLR and STINGa for their ability to induce phenotypic changes in hMDMs in relation to the representative markers CD40, CD38, and CD11b. CD40 and CD38 are considered markers of immune cell activation and are generally upregulated on macrophages following stimulation with LPS (34, 35). Accordingly, CD40 and CD38 were significantly upregulated in hMDMs stimulated with LPS/IFNγ and decreased following IL4/13 treatment. Although the integrin αmβ2 (CD11b) is generally considered a pan-macrophage marker, its expression was significantly decreased with LPS/IFNγ and increased with IL4/13 (Fig. 3A and B).
hMDMs stimulated with both TLRa and STINGa showed a profile resembling LPS/IFNγ–stimulated macrophages with significant increases in CD40 seen with TLRa-1/2, -4, -5, and -7/8 and STINGa 2′2′- and 2′3′-cGAMP. Likewise, TLRa-1/2 and STINGa 2′2′- and 2′3′-cGAMP resulted in increased CD38. Most TLRa resulted in decreased CD11b akin to LPS/IFNγ–treated macrophages, with significant changes seen with TLRa-1/2, -4, and -7/8. The other TLRa or STINGa induced subtle changes in those markers, which were reflective of steady-state, nonpolarized macrophages, demonstrating that within the spectrum of polarization, macrophages stimulated with TLRa and STINGa mostly display an inflammatory LPS/IFNγ–activated profile (Fig. 3A and B).
Consistent with an activated proinflammatory profile, supernatant from hMDMs stimulated with TLRa contained elevated IFNγ, IL6, IL12p70, and TNFα, comparable with supernatant from LPS/IFNγ–stimulated macrophages (Fig. 3C). However, such cytokine responses were not seen following stimulation with STINGa; instead supernatant from STINGa 2′2′- and 2′3′-cGAMP–treated macrophages displayed a type I IFN cytokine profile with release of both IFNα and IFNβ (Fig. 3D), thereby demonstrating a divergence in STING and TLR inflammatory signaling with respect to cytokine induction in human myeloid cells.
FcγR changes and phagocytic activity of hMDMs stimulated with TLRa or STINGa
We subsequently examined the expression pattern of activating (FcγRI, -IIA, -III) and inhibitory (FcγRIIB) FcγR on hMDMs stimulated with TLRa or STINGa (Fig. 4A) and calculated the A:I ratio (Fig. 4B). TLRa-1/2, -5, and -7/8 showed an increase in the expression of activating FcγRs and decrease in inhibitory FcγRIIB, resulting in an increased A:I ratio (Fig. 4A and B), similar to LPS/IFNγ–polarized macrophages. STINGa 2′2′ and 2′3′-cGAMP also resulted in an increase in the expression of activatory receptors FcγRIIA and FcγRIII, and a decrease in FcγRIIB, resulting in a statistically significant increase in A:I ratio (Fig. 4B; Supplementary Fig. S4A). In contrast, we observed a decrease in A:I ratio with IL4/13 polarized macrophages, largely mediated through an upregulation in FcγRIIB.
The phenotypic changes, cytokine profile, and FcγR changes of hMDMs treated with TLRa and STINGa were indicative of proinflammatory tumoricidal effectors. We therefore assessed whether these cells also displayed augmented functional activity in terms of their ability to phagocytose antibody-opsonized CLL cells (Fig. 4C; Supplementary Fig. S4B). Macrophages incubated in the presence of IL4/13 showed a decrease in phagocytic index while LPS/IFNγ–stimulated macrophages maintained a lower or similar phagocytic profile compared with unstimulated macrophages, perhaps due to the induction of higher levels of F-actin via the IFNγ-mediated mechanism involving PI3K reported previously (36). TLRa-1/2, 5, and -7/8 and all STINGa-stimulated hMDMs showed a significant increase in their ability to phagocytose CLL cells (Fig. 4C), correlating with their A:I ratios (Fig. 4D, R2 = 0.34, P = 0.0007). It is not clear whether the contrasting activities of TLR3a in hMDMs and mBMDMs, both in terms of ADCP and type I IFN production, relate to differences in TLR signaling pathways in human and murine cells, but cell type- and species-specific responses to TLR3 stimulation in human and murine effector cells have been previously reported (37).
STINGa, but not TLRa, enhance mAb-mediated target B-cell depletion in the presence of the tumor microenvironment
Having established the relative abilities of TLRa and STINGa to augment FcγR-mediated cell depletion with both murine and human macrophages in vitro, we sought to determine their efficacy in vivo. Consistent with our findings with mBMDMs, among the panel of TLRa assessed, only TLR3a significantly increased the mAb-mediated depletion of target B cells in adoptive transfer assays in vivo. Approximately 40% of target B cells were deleted by Ritm2a alone, whereas in mice primed with TLR3a approximately 60% of target B cells were depleted following Ritm2a administration (Fig. 5A). Intriguingly, the TLRa-1/2 that was potent with hMDMs in vitro failed to enhance target B-cell depletion in mice in vivo and appeared to be inhibitory. We next investigated the differences in FcγR expression and A:I ratio induced on splenic macrophages following TLRa-1/2 and -3 treatment (Fig. 5B). No significant change in FcγR A:I ratio was observed in macrophages from TLR1/2a–treated mice, while TLR3a induced a significant increase in FcγR A:I ratio, likely explaining the elevated depletion with this reagent.
While TLR3a mediated a modest increase in the depletion of adoptively transferred target B cells in these experiments, the murine STING ligand DMXAA mediated a dramatic, approximately 90%, depletion of target B cells in combination with Ritm2a (Fig. 5C). We also observed a statistically significant, 4-fold increase in the FcγR A:I ratio on splenic macrophages of mice that were primed with DMXAA compared with naïve mice (Fig. 5D) demonstrating that among all reagents assessed, DMXAA was the most potent in both upregulating the A:I ratio and depleting target B cells.
The adoptive transfer experiments discussed above provided a robust model to study target B-cell depletion in the absence of any complexity arising from a suppressive tumor microenvironment. However, when BCL1 cells were inoculated into recipient mice 1 week prior to adoptive transfer, the most potent TLRa Poly I:C failed to induce mAb target cell deletion, while DMXAA retained its efficacy even in the presence of the tumor (Fig. 5E). This DMXAA effect on Ritm2a activity was an antibody Fc–dependent process as FcR γ chain KO mice, devoid of activatory FcγRs, were unable to delete target B cells in the same setting (Fig. 5E). When splenic macrophages from TLR3a- or DMXAA-treated tumor-bearing mice were assessed for FcγR changes, we found that DMXAA, but not TLR3a, was able to completely reverse the suppressive effect of the tumor on the macrophage FcγR A:I ratio (Fig. 5F). Hence, the observed changes in FcγR A:I ratios provide an explanation as to why TLR3a loses its efficacy in tumor-bearing hosts, but DMXAA retains its ability to enhance mAb-mediated B-cell depletion even within an immunosuppressive tumor microenvironment. We were also able to exclude that this augmented deletion of target cells in the presence of tumor microenvironment was the result of any direct cytotoxic effect of DMXAA, as the percentage of both tumor and nontumor B cells retrieved 24 hours after DMXAA injection into BCL1-bearing mice was equivalent to that of mice not treated with DMXAA (Fig. 5G). Notably, administration of DMXAA induced upregulation of type I IFN gene expression and also reduced expression of both IL10 and IL21 that were upregulated following inoculation of the tumor (Supplementary Fig. S5).
Type I IFN elicits macrophage FcγR skewing and augments mAb target cell deletion
In both human and murine cells, STINGa induced type I IFN (Figs. 2 and 3; Supplementary Fig. S5). Type I IFN is a key downstream mediator of STING activation (18), and therefore we interrogated whether type I IFNs were also able to modulate FcγR A:I ratio changes and target B-cell deletion. We found that in hMDMs, type I IFN induced upregulation of FcγRIIA and III, leading to an increase in A:I ratio and phagocytosis of target CLL cells, in a manner similar to that mediated by STINGa (Fig. 6A). Likewise, administration of type I IFN to mice led to a significant increase in splenic macrophage FcγR A:I ratio, mostly mediated by an increase in FcγRIV, as seen with DMXAA (Fig. 6B). mBMDMs incubated in the presence of type I IFN also phagocytosed target cells more effectively, with a phagocytic index comparable with STINGa-treated mBMDMs (Fig. 6B). Furthermore, Ritm2a-mediated deletion of target B cells in adoptive transfer assays were also significantly enhanced in mice that were primed with type I IFN prior to mAb administration (Fig. 6C). Most importantly, the target B-cell–deleting ability of DMXAA was completely abrogated in mice that received IFNAR-blocking antibody, further strengthening the role of type I IFN in downstream mechanisms governing the STING-mediated effects (Fig. 6D).
STINGa DMXAA enhances mAb immunotherapy in BCL1 lymphoma in an FcγR-dependent manner
Finally, having established that STING activation was capable of overcoming a suppressive lymphoma microenvironment to delete normal B cells in a transplant system, we assessed whether the same approach could provide effective deletion of the malignant B cells themselves in a therapy model. BCL1-bearing mice were treated with anti-mouse CD20 mAb (18B12; ref. 38) in the presence or absence of DMXAA as depicted in Fig. 7A. When we monitored FcγR modulation in the therapy model, we observed that the tumor substantially downregulated activatory FcγRI and caused a dramatic increase in the inhibitory FcγRII. Both of these changes were reversed by DMXAA, which also induced a 4-fold increase in FcγRIII and IV (Fig. 7B); allowing DMXAA to reverse the A:I changes induced by the lymphoma microenvironment (Fig. 7C). This was reflected in the survival of experimental animals: anti-CD20 and DMXAA monotherapies produced modest therapeutic effects, whereas ≥90% of mice that were primed with DMXAA prior to anti-CD20 administration were effectively cured, surviving for longer than 100 days (Fig. 7D). This significant enhancement in tumor protection over anti-CD20 (median survival 36 days; P < 0.0001) or DMXAA alone (median survival 32 days; P < 0.0001) produced by the combination therapy was lost when mice were devoid of FcγRs (FcγR-null; median survival 30 days; Fig. 7E) showing that modulation of FcγRs is crucial to the combination effect. In contrast, the induction of an adaptive cytotoxic CD8+ T-cell response is not required as 100% of mice receiving the combination therapy following CD8+ T-cell depletion were also cured (Fig. 7F).
TAMs are typically immunosuppressive and have the capacity to negatively impact on anticancer immunotherapy strategies. Although deletion of TAM could overcome this issue, it will also impede tumor cell clearance with direct targeting antibodies, which rely on macrophages for their mode of action (24, 31) and are an important component of lymphoma treatment. Therefore, an alternative approach to restrict or prevent tumor growth and yet retain or provide an increased capacity of the TAM to deliver immunotherapy treatments is highly desired (39).
Here we identify that lymphoma cells can elicit a microenvironment that is profoundly suppressive, leading to resistance of mAb-mediated deletion of target cells and therapy. Lowering of the TAM FcγR A:I ratio was a cardinal feature of the immunosuppressive tumor microenvironment illustrated here, achieved through a reduction of the activating FcγR and elevation of the inhibitory FcγRII. Almost all mAbs used in the clinic to date, with the exception of “true blockers,” such as anti-PD1, rely on the appropriate FcγR engagement for their optimal therapeutic activity (40, 41). Recalibration of the FcγR expression profile and A:I ratio is proposed to be one of the most important mechanisms by which mAb function can be modified. Clearly, tipping the balance in the favor of higher A:I ratios may be beneficial in cancer immunotherapy treatments employing direct targeting mAbs. With regard to anti-CD20 mAb and lymphoma, these changes appear to be more important in the macrophage population, as both monocytes and neutrophils were relatively unaffected by the tumor.
We demonstrate that the majority of TLRa tested polarized hMDMs toward an activated phenotype, including an enhanced FcγR A:I ratio, induced a broad, proinflammatory cytokine response, and augmented their mAb-mediated phagocytic activity; all parameters desirable for antitumor immunity and yet had more modest effects on mBMDMs. The reasons for this are currently unclear but may reflect established species-specific differences (37). Alternatively, these differences may be attributed to the alternative progenitors from which the macrophages were generated (i.e., peripheral blood monocytes versus bone marrow precursors). Conversely, STINGa elicited more reproducible changes in FcγR, and type I IFN production, in both hMDMs and mBMDMs, effectively augmenting ADCP and performed consistently in murine assays both in vitro and in vivo. FcγR modulation is likely to be an important mechanism behind the activity of STINGa in vivo as its activity was completely abrogated in γ chain−/− mice lacking activating FcγR. Furthermore, the enhanced B-cell–depleting activity induced by STINGa was directly correlated with an enhanced A:I ratio, both in nontumor and tumor-bearing adoptive transfer assays.
Following its engagement, STING stimulates the transcription of numerous innate immune genes, including type I IFN with the capacity to augment macrophage activation (42). Our observations here indicate that type I IFN, together with FcγR modulation, rather than classical myeloid proinflammatory cytokine production, may play an important role in augmenting STINGa-based mAb immunotherapy. This point is of interest, as it is known that after engagement of certain cyclic and double-stranded DNA molecules, STING elicits its effects on macrophages in a manner different to, and apparently independently of, other DNA recognizing receptors, such as TLR9 (17). Generation of similar responses in terms of FcγR modulation, A:I ratio changes, enhanced phagocytosis, in vivo deletion, and the inability of STINGa to delete target cells in the presence of IFNAR-blocking antibodies indicate that type I IFN largely govern the downstream effects of STING agonism observed here, which are conserved between mice and humans.
The ability of STINGa, but not TLRa, to eliminate malignant B cells in vivo mirrors how well they reverse the lower FcγR A:I ratio in the macrophage population. Previously, we have shown that the inhibitory FcγRIIB helps tumor cells to escape deletion by rituximab (43, 44). Here we observed a significant improvement in deletion of target B cells in FcγRII knockout mice harboring lymphoma, which suggests that some but not all of the immune suppression mediated by the tumor is due to the upregulation of FcγRII on macrophages, and strengthens the hypothesis that target cell deletion is a composite function of FcγR A:I ratio. Therefore, adjuvants, such as STINGa, capable of eliciting such FcγR A:I ratio changes can be exploited in mAb immunotherapy. Importantly, and in marked contrast to TLRa, we have also demonstrated a commonality of response to STING agonism between mouse and human macrophages strongly supporting the translational nature of our observations. Some studies have suggested that STINGa can directly eradicate malignant B cells and mediate tumor regression (45) or disrupt tumor vasculature (46) leading to tumor necrosis, but we did not observe any direct cytotoxic killing of tumor cells in our study. Although disruption of tumor vasculature has been observed in certain subcutaneously grown tumor models and endothelial cells may be directly affected by DMXAA, we and others (47) have established that DMXAA primarily acts to induce macrophage activation and modulation of intratumoral macrophage phenotype to augment immunotherapy. Presumably, the discrepancies are attributed to the type of STINGa, differences in dosing strategy, and the models used in the studies.
A principal finding of our study was the impressive potency of STINGa in augmenting macrophage activation and mAb-mediated effector mechanisms both in vitro and in vivo. Although the STINGa DMXAA showed very impressive effects in the murine lymphoma model studied here, early human cancer trials using DMXAA failed (48). Subsequent studies revealed that this was due to an inability of DMXAA to activate human STING (49). Since then, novel STINGa capable of engaging human STING have been generated and demonstrated potent adjuvant effects in the radiotherapeutic management of pancreatic cancer (50) and we have, for the first time, demonstrated its efficacy in a lymphoma model in conjunction with immunotherapy, in a manner dependent on FcγR A:I ratio changes, and in a context where TLRa failed. Studies should now be directed to the development, selection, and formulation of cyclic dinucleotide analogues specific to human STING, which deliver enhanced pharmacokinetics and defined potency.
Disclosure of Potential Conflicts of Interest
A.J. Steele has received speakers bureau honoraria from Portola Pharmaceuticals. M.S. Cragg is a consultant at Bioinvent International, reports receiving a commercial research grant from Bioinvent International, and has received speakers bureau honoraria from Baxalta. S.A. Beers reports receiving other commercial research support from Bioinvent International and is a consultant/advisory board member for Astex Pharmaceuticals.
Conception and design: L.N. Dahal, M.J. Glennie, M.S. Cragg, S.A. Beers
Development of methodology: L.N. Dahal, L. Dou, A. Earley, A.J. Steele, S.A. Beers
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): L.N. Dahal, L. Dou, K. Hussain, R. Liu, A. Earley, S. Murinello, F. Forconi
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): L.N. Dahal, L. Dou, K. Hussain, A. Earley, A.J. Steele, M.J. Glennie, M.S. Cragg, S.A. Beers
Writing, review, and/or revision of the manuscript: L.N. Dahal, F. Forconi, A.J. Steele, J.L. Teeling, M.J. Glennie, M.S. Cragg, S.A. Beers
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K.L. Cox, I. Tracy, F. Forconi, P. Duriez, D. Gomez-Nicola
Study supervision: M.S. Cragg, S.A. Beers
We thank patients and volunteers who donated specimens, Mark J. Shlomchik for hCD20Tg mice, Sjef Verbeek for FcγR-null mice, colleagues from the Cancer Sciences Unit and antibody production team, Biomedical Research Facility for animal husbandry, I. Henderson and K.N. Potter for collection and characterization of CLL samples.
This work was supported by grants from Bloodwise (12050 to M.S. Cragg), and CRUK (A12343 to S.A. Beers). This work was also supported by the Southampton ECMC grant C24563/A15581 and a CRUK Southampton Centre grant C34999/A18087.
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.