|Home | About | Journals | Submit | Contact Us | Français|
Reprogramming of immunosuppressive tumor microenvironment (TME) by targeting alternatively activated tumor associated macrophages (M2TAM), myeloid-derived suppressor cells (MDSC), and regulatory T cells (Tregs), represents a promising strategy for developing novel cancer immunotherapy. Prostaglandin E2 (PGE2), an arachidonic acid pathway metabolite and mediator of chronic inflammation, has emerged as a powerful immunosuppressor in the TME through engagement with one or more of its 4 receptors (EP1-EP4). We have developed E7046, an orally bioavailable EP4-specific antagonist and show here that E7046 has specific and potent inhibitory activity on PGE2-mediated pro-tumor myeloid cell differentiation and activation. E7046 treatment reduced the growth or even rejected established tumors in vivo in a manner dependent on both myeloid and CD8+ T cells. Furthermore, co-administration of E7046 and E7777, an IL-2-diphtheria toxin fusion protein that preferentially kills Tregs, synergistically disrupted the myeloid and Treg immunosuppressive networks, resulting in effective and durable anti-tumor immune responses in mouse tumor models. In the TME, E7046 and E7777 markedly increased ratios of CD8+granzymeB+ cytotoxic T cells (CTLs)/live Tregs and of M1-like/M2TAM, and converted a chronic inflammation phenotype into acute inflammation, shown by substantial induction of STAT1/IRF-1 and IFNγ-controlled genes. Notably, E7046 also showed synergistic anti-tumor activity when combined with anti-CTLA-4 antibodies, which have been reported to diminish intratumoral Tregs. Our studies thus reveal a specific myeloid cell differentiation-modifying activity by EP4 blockade and a novel combination of E7046 and E7777 as a means to synergistically mitigate both myeloid and Treg-derived immunosuppression for cancer treatment in preclinical models.
Checkpoint inhibitors have produced impressive therapeutic effects in cancer patients, but thus far, only a fraction of patients benefit from this class of cancer immunotherapies. It is therefore of great interest to develop novel immunotherapies and combinations to increase patient responsiveness. In this regard, reprogramming of the immunosuppressive TME by targeting myeloid-derived suppressor cells (MDSC), M2TAMs, and Tregs represents an important approach to developing novel cancer immunotherapies.
One immunosuppressive mechanism used by various tumors is production of PGE2,1 a lipid product of the arachidonic acid metabolism cascade and an inhibitory damage-associated molecular pattern (iDAMP) in the TME.2 PGE2 controls a large variety of biologic functions, through its binding to E-type prostanoid receptors (EP1 to EP4). The relative contribution of each of these EP receptors in mediating the immunosuppressive effect of PGE2 has not been fully elucidated and some reports even show opposing activities of EP receptors on immune function.3 EP4, a Gαs protein coupled receptor expressed mainly on myeloid cells, T lymphocytes, and tumor cells4 has emerged as a major contributor to PGE2-mediated enhancement of tumor survival pathways and also as a suppressor of innate and adaptive anti-tumor immune responses, by triggering cancer-promoting chronic inflammation.5 This chronic inflammation is orchestrated by PGE2 through its activity on a variety of immune cells present within the TME including macrophages,6-8 DCs,9 CD4+ T helper cells,10-12 CTLs,13 and Tregs.14 Given the broad nature of PGE2 biologic activities and the differential involvement of EP receptors in mediating these activities, specific targeting of EP4 signaling should provide a more focused and possibly safer approach to anti-cancer therapy than inhibitors of the entire pathway leading to PGE2 production.
The TME often contains a large number of myeloid cells, mainly MDSCs and TAMs, whose presence is associated with poor prognosis in a substantial proportion of solid tumors.15 Both monocytic MDSCs (mMDSCs), defined as CD11b+Ly6c+Ly6G−, and granulocytic (gMDSCs) defined as CD11b+Ly6c−Ly6G+ are found in the TME and peripheral lymphoid tissues of animal models. MDSCs exert immunosuppressive activity mainly through arginase and production of reactive oxygen species. The less numerous but longer-lived mMDSCs have been shown to be more immunosuppressive on a per cell basis in the TME than the more abundant gMDSCs.16 While CD11b+Ly6cloMHCIIhi M1-like macrophages, or classically activated macrophages, are activated by IFN-γ and sterile or infectious injury, and are tumoricidal17; alternative activated M2TAMs, defined as CD11b+Ly6c−MHCIIlo, support tumor progression and metastasis through remodeling of extracellular matrix and release of factors that promote cell proliferation, angiogenesis and migration.18 PGE2 has been shown to favor the immunosuppressive M2TAM phenotype,19 therefore, blocking PGE2 activity in the TME should reduce M2 polarization. Other important mediators of immunosuppression within the TME are Treg cells which normally control key aspects of self-tolerance. In the TME, Tregs contribute to immunosuppression by maintaining immune tolerance to the tumor and opposing cytotoxic T cell activity.20 Treg cells express high levels of CD25 and CTLA-4 and are defined by their expression of the transcription factor Foxp3. Administration of denileukin diftitox, an IL-2-diphtheria toxin fusion protein, has been shown to diminish the number of CD25highCD4+ Treg cells in patients.21-23 Similarly, a key anti-tumor property of anti-CTLA-4 antibodies has been attributed to their ability to reduce Tregs as well.24,25 We thought to identify an agent or a combination that can simultaneously target both intratumoral myeloid cell differentiation and Treg viability to efficiently restore the anti-tumor immunity.
In this report, we evaluated myeloid cell-modifying effects of EP4 antagonism in vitro and in vivo, examined pharmacologically the concurrent targeting of EP4 signaling and Treg viability in tumor-bearing immunocompetent mice, and investigated the mechanisms driving the combined anti-tumor activity of these agents through a combination of genetic, molecular and cellular phenotypic analyses. To selectively antagonize PGE2-EP4 signaling we used E7046 (also known as ER-886406),26 a clinical stage novel and orally bioavailable small molecule EP4 antagonist (NCT02540291). To reduce Tregs we used E7777, a fusion protein comprising human IL-2 and fragments of diphtheria toxin, representing an improved-purity version of denileukin diftitox. We show here a specific myeloid cell differentiation and anti-tumor activity of E7046 by antagonism of EP4 signaling and a highly potent anti-tumor activity of the novel E7046 and E7777 combination in preclinical cancer models by synergistic inhibition of both myeloid and Treg–induced TME immunosuppression.
Mice with floxed EP4 gene were acquired from the Breyer Lab of Vanderbilt University27 and ER-Cre mice were purchased from Taconic. Both parent strains were on the C57BL/6 background and were crossed to create mice that were homozygous for the floxed EP4 gene and heterozygous for the ER-Cre gene. The colony was maintained and breeding performed at Charles River Laboratories (Wilmington, MA). To produce the knockout of EP4, the mice were orally dosed once daily with tamoxifen (Sigma, St. Louis) 1 mg/kg formulated in sunflower oil. Mice were dosed for 5 consecutive days to generate complete knockout. To verify the efficacy of the knockout, blood and tissue samples were collected and analyzed by qPCR for EP4 expression. BALB/c, C57BL/6, and A/J mice were purchased from Jackson Laboratory (Bar Harbor, ME). Nude mice were obtained from Charles River Laboratories (Wilmington, MA). Experiments were conducted with 6–10 week-old mice, unless otherwise specified. All mice were kept under specific pathogen-free conditions in the Life Science Facility of AIM/Eisai Inc. (Andover, MA), and animal procedures were conducted under conditions approved by the Institutional Animal Care and Use Committee following the guidance of the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC).
MDA-MB-231 and COLO-205 human cancer cell lines were obtained from ATCC (American Type Culture Collection, Manassas, VA). HMVEC were obtained from Lonza. Mouse cell lines CT26, LL2, 4T1, EMT6, SAI/N, and were obtained from ATCC, mouse cell line Pan02 was obtained from the DCTD-NCI (Frederick, MD) and mouse cell line H22 from GenScript (Piscataway, NJ). All cell lines were cultured according to vendor recommended growth media with 10–15% FBS and conditions (37°C and 5% CO2). Cell lines from early passage stocks (less than 5) were stored at -180°C in the Eisai Biobank. Cell lines were used in assays before reaching 15 passages in culture. Cell line authentication was conducted by the Short Tandem Repeat (STR) profiling method at ATCC.
E7046 was formulated in 0.5% methylcellulose (MC) and administered by oral gavage at 100 mg/kg (H22 model) or 150 mg/kg (all other in vivo studies). For in vitro studies, E7046 was dissolved in DMSO to form a 30 mM stock solution, which was further diluted for specific experiments. E7777, supplied as a sterile 156 µg/mL stock solution, was diluted in saline for injection. For in vivo cell-depletion studies, in vivo Plus anti-mouse CD4 (clone GK1.5) and anti-CD8 (clone 53–6.72) antibodies from BioXCell (West Lebanon, NH) were administered i.p. As controls, in vivo Plus Rat IgG2b (clone LTF-2) and Rat IgG2a isotype control (clone 2A3) (BioXCell) were injected in the control mice. In vivo Plus anti-PD1 (clone RMP1–14), anti-CTLA-4 (clone9H10) antibodies and Syrian hamster IgG control (Bi XCell) were diluted in sterile PBS and injected i.p. PGE2, SC-51089, ER-880696, and L-798106 were dissolved in DMSO at 10 mM stock concentration.
For the LL2 tumor growth in EP4−/− mice, EP4 wild type (WT) n = 10, and EP4−/− n = 9 mice were injected subcutaneously (s.c.) with 1 × 105 cells/mouse in 100 µL sterile HBSS. Tumor growth was monitored by tumor volume measurement every 3–4 d throughout the experiment. Tumor volume was calculated by the formula of a prolate ellipsoid: (Lx2W)/2 where L and W are the respective orthogonal length and width measurements (mm).
For the tumor isograft efficacy studies, 6-week old female BALB/c mice were implanted with cancer cells: 1 × 105 CT26 or 4T1 cells or 8 × 105 H22 cells per mouse s.c., or 1 × 105 EMT6 cells in the mammary fat pad. C57BL/6 mice were implanted s.c. with 1 × 106 Pan02 cells per mouse, and A/J mice were implanted s.c. with 2–3 mm3 SAI/N tumor fragments. To investigate the role of T cells in the anti-tumor response, 6 week old female nude mice (which lack T cells) were injected s.c. with 1 × 105 CT26 cells. When tumors reached approximately 50–100 mm3, tumor–bearing mice were randomly assigned to vehicle or treatment groups, and treatment regimens began. E7046 was administered per oral (p.o.) as a 100 or 150 mg/kg suspension in 0.5% MC, daily for 21 d (QDx21). For the combination studies, E7777 was administered intravenously (i.v.) at 2.5 µg/mouse in saline, as 2 to 3 doses injected one week apart (Q7Dx2–3). Tumor volumes and body weights were recorded 2–3 times a week. For comparison with current immunotherapies, in addition to vehicle control and E7046 + E7777 groups, mice were assigned to anti-PD1 antibodies or anti-mouse PD-1 + anti-mouse CTLA4 antibodies treatment groups. Anti-PD-1 and anti-CTLA-4 antibodies (1 mg/mL), were administered i.p. in 100 µL, 3 times 4 d apart (Q4Dx3) for a total of 300 µg each. Isotype controls were administered i.p. at 1 mg/mL to the control group. For the CD4+T and CD8+T lymphocyte depletion, anti-mouse CD4 or anti-mouse CD8 antibodies or their isotype controls were administered in 100 µL, i.p. at 2.5 mg/mL every 4 days, for a total of 4 injections per mouse (1 mg). To deplete macrophages, clodronate-containing or control liposomes (Encapsula Nanosciences) were administered i.p. at 1 mg/mouse, every other day for a total of 7 injections. For the live H22 cell challenge, mice whose initial tumors had completely and stably disappeared were paired with age-matched naïve controls and injected s.c. with 8 × 105 H22 cells, on the opposite side from the original implantation site. Tumor volumes and body weights were measured 2 times weekly throughout all experiments.
Bone marrow (BM) cells were flushed from femurs of BALB/c mice using sterile CM. Freshly harvested (BM) cells (0.5 × 106) were differentiated in the presence of 20 ng/mL recombinant mouse GM-CSF (Peprotech), ± PGE2 (10 nM), at 37°C, for 8 d. CM + fresh GM-CSF ± PGE2 was changed on days 3 and 6. After in vitro differentiation, cells were analyzed by flow cytometry. For certain experiments, CT26, 4T1 cell supernatants, and/or EP1 (SC-57089), EP2 (ER-880696), EP3 (L-798106), or EP4 (E7046) antagonists at 1 µM, were added to the BM cells. To assess the effect of differentiated BM cells on T cell proliferation, mouse BM cells differentiated as described above were co-cultured for 72 hours with anti-CD3/CD28 Dynabeads-stimulated and CFSE (1 µM)-stained T cells. T cell proliferation was assessed by CFSE dilution using flow cytometry.
Monocytes were purified from PBMCs isolated from healthy volunteer donors or cancer patients using a monocyte separation kit (Miltenyi), and placed in culture with recombinant human GM-CSF (20 ng/mL) + human IL-4 (500 U/mL) ± PGE2 (10 nM) ± E7046 (10, 100, 300, 1000, or 3000 nM) for 7 d. Differentiated myeloid cells were identified by surface marker phenotyping by flow cytometry. For the suppression of T cell proliferation, monocytes differentiated as described above were then co-cultured with previously purified autologous T cells (Miltenyi T cell separation kit) at a 1:4 ratio for 96 h. To assess proliferation, T cells were stained with CFSE at 1 µM (Life Technologies) before co-culture with APCs. T cells were stimulated with antiCD3/CD28 Dynabeads according to the manufacturer's protocol (Life Technologies). T cell proliferation was assessed by CFSE dilution using flow cytometry.
Tumor samples were collected in liquid nitrogen, thawed on ice, then minced and lysed in lysis buffer containing a cocktail of protease inhibitors (Roche). Lysates were incubated for 1 h at 4°C on a rotator, sonicated at 4°C for 5 min with intermittent on/off every 30 sec, and centrifuged (6700 x g) for 10 min at 4°C. Protein concentration was determined in the supernatants using the BCA assay (Pierce). Mouse cytokines and chemokines were measured in CT26 tumor lysates on days 3, 7, and 14 of E7046 treatment, using the Proinflammatory Panel 1 (mouse) multiplex assay kit (K15048D) and the Sector Imager 2400 from MSD according to the manufacturer's instructions.
FcR-blocked BM-derived APCs, differentiated human PBMC-isolated monocytes, murine splenocytes, or tumor cells were stained for flow cytometry while on ice for 20 min. To obtain single cell suspensions for analysis, tumors were excised from killed mice, cut into small pieces, and digested in medium containing a mix of 1 mg/mL type IV collagenase, 0.25 mg/mL type I collagenase (Worthington Biochemical Corporation), and 0.02 mg/mL DNAse (Roche). Digestion was performed in a 37°C-heated shaker for 30 min, followed by passage through a 70-µm filter. Surface-staining antibodies were added for 30 min after the FcR blocking antibodies. The following antibodies were obtained from BD Biosciences: anti-mouse CD45 (B6.F11), -CD11b (M1/70), -Ly6G (1A8), anti-human CD86 (2331), and -CD16 (3G8). The following Biolegend-manufactured antibodies were used for cell staining: anti-mouse CD11c (N418), anti-human CD1a (HI149), -CD163 (GH/61), and -CD206 (15–2). The following antibodies were purchased from eBioscience: anti-mouse CD4 (GK1.5), -CD8α (53–6.7), -CD25 (PC61.5), - F4/80 (BM8), -MHCII (M5/114.15.2), -PD-L1 (MIH5) –Ly6c (HK1.4), -Foxp3 (FJK-16s), -granzyme B (NGZB), -Gr1 (RB6–8C5). Zombie NIR fixable viability dye was used to determine cell death (Biolegend). Intracellular staining for granzyme B and Foxp3 was performed after surface staining, fixation, and subsequent permeabilization following the manufacturer's protocol. Flow cytometry acquisition was conducted on FACS Canto instruments (I and II), (Becton Dickinson) and data were analyzed using FlowJo software (Tree Star, Inc.).
Total RNA from tumor tissues was isolated using RNeasy Mini Kit (Qiagen) as per the manufacturer's instructions. RNA concentrations were determined by Nanodrop (ND-1000). Hybridization targets were prepared by means of the Affymetrix GeneChip® WT PLUS Reagent Kit using 500 ng of total RNA. The Affymetrix Expression Console Software was used to perform the normalization of the cel files by the Signal Space Transformation (SST) algorithm. Comparison of the treatment groups with vehicle were analyzed in the Affymetrix Transcriptome Analysis Console (TAC) software which calculated the differential gene expression, relative fold change and p-values, using the default settings. Of the resulting gene signatures, those meeting a threshold of > 1.5-fold change and a p-value <0.05, were uploaded into Ingenuity Pathway Analysis (IPA) for further analysis. qPCR was conducted using the pre-configured TaqMan® Low Density Mouse Immune Array (Life Technologies). cDNA was generated from total RNA using SuperScript VILO Master Mix (Life Technologies). Relative quantification was calculated using the Comparative CT method. For both GeneChip and qPCR data visualization, Clustvis28 was used.
IHC analysis was performed on tumors from CT26-tumor bearing BALB/c mice (n = 3/group) that had received vehicle or 300 mg/kg p.o. daily of E7046 for 14 d. Tumors were excised, formalin fixed, and embedded in paraffin (FFPE). Sections of FFPE tissue blocks were stained with hematoxylin and eosin, and IHC was performed post antigen retrieval using early-T cell specific rabbit anti-mouse CD3 monoclonal antibody (ThermoFisher Scientific, clone SP7). Percentages of CD3 positive T cells were measured on an Aperio image analyzer using Scan Scope software.
Differences between mice treated with E7046 and control mice were analyzed by the 2-tailed Student t test and expressed as the mean ± SEM. For in vivo studies involving multiple treatment groups, significance was determined by 2-way ANOVA followed by Tukey's multiple comparisons test. Differences in population frequencies as determined by flow cytometry were analyzed using GraphPad Prism 6 software and significance determined by one-way ANOVA followed by Tukey's multiple comparisons test. P < 0.05 was considered significant for all analyses.
E7046 is a newly developed and highly specific EP4 antagonist with a > 1500 fold selectivity for EP4 relative to EP4 closest homolog, the PGE2 receptor EP2 (Fig. 1A, Supplementary Fig. S1).26 To evaluate a specific role of PGE2-EP4 signaling in pro-tumor myeloid cell activation, we assayed effects of PGE2 with or without E7046 and antagonists specific for EP1–3 in a modified BioMap assay system.29 This system consists of co-cultures of primary human PBMCs, or purified macrophages, with endothelial cells in conditions that stimulate TCR, or activate either M2TAM or M1 macrophages, allowing a systematic examination of numerous immune readouts including leukocyte proliferation, cytotoxicity, and production of cytokines and chemokines, adhesion molecules, and enzymes (Supplementary Fig. S2A). In these cellular systems neither E7046 nor the EP1–3 antagonists alone altered immune cell functions or survival in the absence of exogenous PGE2 (Supplementary Fig. S2B and data not shown). However, a TME-comparable concentration of PGE2 significantly suppressed the activation of type 1 macrophages, evidenced by markedly reduced soluble TNF-α and CD40, and promoted type 2 macrophage activity by increasing IL-6 secretion (Supplementary Fig. S2C). Importantly, EP4 antagonist E7046 was a potent inhibitor of PGE2 activity in these systems; the EP2 antagonist had much weaker immunomodulatory activity (Supplementary Fig. S2C), and EP1 and EP3 antagonists showed no significant activity (data not shown). TNF-α cytokine secretion was reduced the most by PGE2;’ E7046 largely restored TNF-α levels in both PBMC and macrophage systems at the concentrations tested (Fig. 1B). Compared to macrophage systems, much less pronounced changes were detected for PGE2 and E7046 in TCR-stimulated co-culture systems (second panel of Supplementary Figure S2B,C). The results together indicate a specific role of PGE2-EP4 signaling among the 4 EP receptors in promoting the activation of pro-tumor M2-like macrophages and concomitant suppression of anti-tumor M1-like macrophages.
Since PGE2 levels correlate with accumulation of immunosuppressive MDSCs and M2TAMs in human tumors and4 we next examined the effects of PGE2-EP4 signaling in the differentiation and function of monocytes toward antigen-presenting cells, using human monocytes in the presence GM-CSF and IL-4 (Fig. 1C). PGE2 (10 nM) showed potent inhibitory activity on GM-CSF/IL-4-induced DC formation. E7046 reduced in a dose-dependent manner the PGE2-biased monocyte differentiation toward CD1a−CD16+ macrophages and promoted differentiation toward CD1a+CD16− DCs (Fig. 1D, ,E).E). PGE2 inhibited expression of the activation marker CD86 and promoted expression of CD163 and CD206 markers, both of which are characteristically associated with the M2-like phenotype (CD163+CD206+CD86−); this effect was dose-dependently inhibited by E7046 (Fig. 1D, ,E).E). Furthermore, PGE2-differentiated human monocytes exerted an immunosuppressive function toward autologous T cell proliferation in response to TCR stimulation, and this effect was also reduced by E7046 in a dose-dependent manner (Fig. 1F). Together, these results demonstrate that selective pharmacological blockade of EP4 signaling by E7046 was able to largely abolish PGE2-mediated pro-tumor myeloid cell differentiation and immunosuppressive activities.
To assess whether EP4 antagonism by E7046 has similar specific activity on mouse myeloid cells, we used murine BM-derived haematopoietic cells induced to differentiate into CD11c-expressing DCs (Supplementary Fig. S3A). E7046 antagonism of PGE2 effects on BM cell differentiation rescued T cell proliferation in response to TCR activation in a dose-dependent manner (Supplementary Fig. S3B). Similar to the observations in human co-culture systems, when we compared the effects of EP1-EP3 antagonists with E7046 on mouse BM cell differentiation, only EP4 antagonism rescued the CD11c+ DC differentiation from BM cells that PGE2 exposure suppressed. This indicated that PGE2 mediates its inhibitory effects on mouse myeloid cell differentiation toward DC through EP4 (Supplementary Fig. S3C,D). Furthermore, the myeloid-modifying effects of PGE2 could be replicated by the addition of culture supernatants from PGE2-producing CT26 and 4T1 mouse cell lines (Supplementary Fig. S3E, F). Addition of E7046 to these supernatants also sufficed to rescue DC differentiation from BM cells (Supplementary Fig. S3E), suggesting that PGE2 was the major soluble inhibitory factor secreted by the cancer cells. Finally, we implanted PGE2-producing Lewis lung carcinoma (LLC) tumors into EP4 inducible conditional knockout mice in which removal of EP4 was achieved by crossing the EP4flox/flox mice with ER-cre transgenic mice followed by tamoxifen administration. Fig. 2A shows a significant delay in the growth of LLC tumors implanted in EP4 receptor-deficient compared with wild type control mice, indicating that PGE2-EP4 signaling in host stromal cells plays an important role in promoting tumor progression.
Taken together, these data demonstrate that PGE2 signaling through EP4 plays a dominant immunosuppressive role in myeloid cell differentiation and activation in both human and mouse cells and that E7046, by selectively blocking EP4, inhibits these PGE2 activities.
Given the important role of myeloid cells in defining a pro- or anti-tumor immune microenvironment, we next assessed the anti-tumor activity of E7046 in syngeneic mouse cancer models. E7046 has a circulating half-life of approximately 4 hours in mice; daily oral administration of 150 mg/kg E7046 was found to inhibit the growth of multiple syngeneic tumor models: fibrosarcoma SaI/N, pancreatic PAN02, colon CT26, and breast EMT6 and 4T1 (Fig. 2B-E). The highest anti-tumor activity, accompanied by 25% tumor complete regression, was observed in the SaI/N model (Fig. 2B). This anti-tumor activity was dose-dependent (Fig. 2E) and relied on T cells, since tumor-bearing nude mice did not respond to E7046 treatment (Fig. 2G). To confirm and refine this result, CD8+ T cell depletion, but not CD4+ T cell depletion, completely abrogated the anti-tumor effect of E7046 in immunocompetent mice (Fig. 2F). Furthermore, macrophage depletion by clodronate-containing liposomes impaired tumor growth, but addition of E7046 had no further effect, suggesting that myeloid cells, especially macrophages and MDSC, were the primary cellular targets of E7046 in vivo (Fig. 2H). Consistent with the immune-mediated anti-tumor activity, E7046 at concentrations up to 100 µM, did not show any direct anti-proliferative activity against cancer cells of human or mouse origin, or of human endothelial cells in vitro (Supplementary Fig. S4). These results indicate a broad anti-tumor efficacy of EP4 blockade by E7046, which is mediated by both myeloid and CD8+ T cells.
The dual myeloid- and CD8+ T cell–dependent anti-tumor activity of E7046 is consistent with the finding that E7046 promoted DC development and function and inhibited M2 macrophage development and function in vitro (Fig. 1 and Supplementary Fig. S2). To understand the in vivo immunomodulatory activity of E7046, we characterized the immune cell composition of tumors from CT26 tumor-bearing mice treated with E7046 or vehicle. We found that, compared with vehicle, E7046 treatment of 7 d significantly reduced the proportions of intra-tumoral CD45+CD11b+Ly6c+Gr-1lomMDSC and increased those of MHCIIhi M1-like macrophages while also decreasing MHCIIlo M2-like macrophages (Fig. 3A,,B).B). Analysis of TME cytokines revealed dynamic changes in levels of pro- and anti-tumor cytokines and chemokines. Compared to vehicle treatment, E7046 increased the myeloid cell-derived anti-tumor cytokine TNF-α, and reduced the levels of pro-tumor CXCL1, IL-10, and IL-6 (Fig. 3C). While TNF-α levels increased and IL-10 levels decreased early after E7046 administration, on day 3, IL-6 and CXCL1 levels were reduced on day 14. No significant changes in levels of IL-1β were noted. In the spleen, administration of E7046 for 14 d significantly decreased MDSC numbers (Fig. 3D), but did not alter percentages of Treg cells (Fig. 3E). In the tumors, numbers of T cells were significantly increased after 14 d of E7046 treatment (Fig. 3F). These data indicate that in vivo the effects of E7046 in EP4 signaling blockade are both systemic, affecting the myeloid cells in both tumor and spleen, and dynamic: reducing the intra-tumoral immunosuppressive MDSCs and M2-like TAMs relatively early after treatment initiation, and promoting intra-tumoral T cell accumulation later during treatment. In addition, E7046 influences the function of TME cells by favoring production of anti-tumor cytokines. These results further implicate immune modulation in the anti-tumor mechanism of E7046 and confirm T cell involvement in the anti-tumor response.
Because E7046 did not impact Treg cell populations (Fig. 3E), we thought to combine it with the Treg-reducing agent E7777 to modulate myeloid cells and Tregs simultaneously. Similarly to the Treg-reducing effect of IL-2-diphtheria toxin fusion protein in patients,22,23 intravenous administration of E7777 resulted in a significant reduction in numbers of Tregs, but not CD8+ T cells, in mouse spleen (Supplementary Fig. S5). To interrogate the effect of the E7046 and E7777 combination, we chose the hepatocellular carcinoma H22 model, which is rich in both macrophages and Treg cells. Administration of E7777 alone on a Q7Dx3 schedule was effective and rendered 30% of the mice tumor-free, while 100 mg/kg of E7046 alone showed limited activity; only 10% of the treated mice had experienced a complete regression at the end of the study (Fig. 4A,,B).B). Importantly, treatment of H22 tumor-bearing mice with E7046 + E7777 using the same drug doses and schedules as each single agent, was highly effective and rendered 80% of the mice tumor-free (Fig. 4A,,B).B). These tumor-free mice were kept for an additional 30 d without further treatment. No tumor growth was detected in any mice, indicating a durable tumor regression resulting from the combination treatment. Subsequent re-challenge of the combination-cured mice with live H22 cells in the absence of any further treatment resulted in complete tumor rejection, indicating that E7046 + E7777 combination treatment induced effective immunological memory in those animals. In contrast, in all naïve control mice implanted with H22 tumor cells at this time, tumors grew aggressively (Fig. 4C). The combination of E7046 and E7777 was further tested in the moderately immunogenic CT26 and poorly immunogenic 4T1 models. In the CT26 model, combination therapy using the same drug administration regimen resulted in a nearly complete tumor growth inhibition. This activity was largely abolished by CD8+ T cell depletion using anti-CD8 antibodies (Fig. 4D). In the 4T1 model, where anti-CTLA-4 antibodies alone or a combination of anti-CTLA-4 and anti-PD-1 antibodies had shown only very limited activity, the combination of E7046 and E7777 was clearly more active than the checkpoint blockade combination (Fig. 4E). Treg cell reduction has been identified as a major anti-tumor activity of anti-CTLA-4 antibodies in preclinical models.24,25 Indeed, we observed a combined anti-tumor effect of E7046 and anti-CTLA-4 antibodies compared with each agent alone in this same 4T1 model accompanied by 12.5% tumor-free mice (Fig. 4F), providing additional evidence that EP4 antagonism combined with a Treg cell reduction treatment is a functional strategy to enhance anti-tumor activity. Collectively, these results demonstrate for the first time that combined EP4 blockade and Treg reduction is an effective means of controlling tumor growth and even of rejecting established tumors.
To understand the mechanism of the anti-tumor immune activity of the E7046 + E7777 combination, we performed cellular immunotyping of CT26 tumors and spleens isolated from treated mice by flow cytometry (Fig. 5, Supplementary Fig. S5). Both types of tissues were isolated 24 hours after the animals had received the second dose of E7777, which was after 7 d of dosing with E7046.
As expected, the CD4+ T cell frequencies were decreased in the TME of animals treated with E7777 or the combination of E7046 + E7777, as a result of reduced frequencies of intra-tumoral Treg cells (Fig. 5A,,E).E). The reduction of Tregs was associated with significantly increased apoptosis of these cells, an effect less evident in non-Treg CD25+Foxp3−CD4+ T cells (Fig. 5B,,CC,,F).F). In line with the result in Fig. 3F, CD8+T cell percentages were significantly higher in response to E7046 alone (Fig. 5G). Consistent with the enhanced pharmacological efficacy of the combination, we detected significantly higher frequencies of granzyme B+CD8+ T cells in the tumors of combination-treated mice compared with vehicle or single agent-treated groups (Fig. 5D,,I).I). Given the CD8+ T cell dependency of the anti-tumor effect of the combination (Fig. 4D), we expected enhanced percentages of intra-tumoral CD8+ T cells in response to combination treatment. Further analysis found that a large proportion of CD8+ T cells in the combination treatment group was positive for the live/dead dye while CD8+ T cells in the E7777 single agent-treated group were minimally positive for the live/dead dye. This indicated that the CD8+ T cells in the combination group were bona fide functioning CTLs dying while exerting their killing activity (Fig. 5H,,I).I). The ratio of intra-tumoral granzyme B+CD8+ T cells (CTLs)/live Treg cells was also significantly higher in the group receiving combination treatment than in the vehicle control and single-agent groups, strongly indicative of an ongoing anti-tumor immune response (Fig. 5J).
Analysis of the intra-tumoral CD11b+ myeloid cells showed that the M1-like Ly6clo-expressing macrophages acquired low expression of the granulocytic marker Ly6G in response to E7046 + E7777 combination, while Ly6chiLy6G− mMDSC frequencies slightly decreased and Ly6cloLy6Ghi gMDSC frequencies increased to some extent (Fig. 6A and data not shown). Thus, based on Ly6c and Ly6G expression, we could distinguish 2 macrophage populations, the M1-like Ly6cloLy6Glo which was significantly better represented in the combination treatment group, and M2-like Ly6c−Ly6G− TAMs which were reduced by E7046 + E7777 combination compared with vehicle treatment (Fig. 6B,,C).C). These cells were analyzed separately for MHCII and PD-L1 expression. The percentage of PD-L1lo and MHCIIhi M1-like TAMs was significantly higher in the combination treatment compared with vehicle group while 30–50% of M2-like TAMs were PD-L1hiMHCIIlo regardless of treatment (Fig. 6D). The ratio of M1-like/M2-like macrophages in the tumor was significantly increased by E7046 + E7777 treatment compared with vehicle or single agent treatment (Fig. 6E) indicating that the E7046 + E7777 combination is able to alter the TAM populations by favoring the anti-tumor M1-like over pro-tumor M2-like TAMs.
To further investigate the mechanism, expression of immune genes was determined by microarray and TaqMan® Low Density Array (TLDA) in CT26 tumors isolated from control mice and mice that had received daily oral doses of E7046 for 20 days, or 3 doses of E7777 once every 7 days, or a combination of both. We considered as significant gene expression changes ≥ 1.5-fold and P<0.05. RNA expression results show that, while E7046 or E7777 alone elicited modest gene expression changes compared with vehicle control, E7046 + E7777 treatment strongly upregulated numerous genes involved in antigen presentation, adhesion, Th1 cell differentiation, activation of cytotoxic mechanisms, Cxcr3 and Ccr5 ligand chemokines, but also tumor and myeloid-produced factors and known T cell co-inhibitory receptors (Fig. 7, Supplementary Fig. S6). The modest gene expression changes in response to EP4 antagonism or Treg cell reduction alone was expected given that neither E7046 nor E7777 are transcriptional regulators; however, their combination showed synergistic activity on gene expression, both in the number of genes and the amplitude of the response (Fig. 7A,,B).B). Pathway analysis of the altered immune gene expression revealed that the genes significantly up- or downregulated in response to treatment belong to multiple pathways involving immune functions (Supplementary Fig. S6A). Gene expression analysis indicated M1-like macrophage activation: upregulation of genes encoding MHC type I and II proteins (H2-T24, H2-Eb1, H2-M3), antigen processing and presentation machinery: class II trans-activator (Ciita), transporter 1 ATP-binding cassette (Tap1), proteasome subunit β type 8 (Psmb8), TAP binding protein (Tapbp), solute carrier family 11a1 (Slc11a1), innate immune sensing proteins such as Z-DNA binding protein 1 (Zbp1), and functional molecules such as inducible nitric oxide synthase 2 (Nos2) (Fig. 7C, Supplementary Fig. S6B). These effects were paralleled by a concerted decrease in M2TAMs and the associated immunosuppressive functions marked by downregulation of scavenger receptor gene CD163, Fc receptor-like S (Fcrls), stabilin 1 (Stab1), mannose receptor (Mrc1), and other M2TAM markers and secreted proteins, including lymphatic vessel endothelial hyaluronan receptor 1 (Lyve1), and coagulation factor XIII A subunit (F13a1)30-33 (Fig. 7C). Of note, Ptgs2, which encodes Cox2 (and thus, controls the cascade leading to PGE2 production) was not regulated by E7046 or by the combination (Supplementary Fig. S6B). The E7046 + E7777 combination also enhanced transcription of markers of M1-like macrophage and DC activation (CD40), numerous NK activating receptors such as natural killer cell group 7 sequence (NKg7), killer cell lectin-like receptor subfamily k 1 (Klrk1) encoding NKG2D, CRACC (Slamf7), and STAT1 as well as IFN-γ target genes involved in DC maturation and NK cell activation. In the same class, C-type lectin domain family 7a (Clec7a), family with sequence similarity 26 (Fam26f), oligoadenylate synthetase 2 (Oasl2), and NK-lytic associated molecule NKLAM (Rnf19b) (Fig. 7C, Supplementary Fig. S6B).34-36 Finally, E7046 + E7777 treatment induced immune cell activation and effector function within the tumor, as evidenced by increased mRNA expression of T cell receptors (CD8α, CD8β), inflammatory chemokines (Ccl5, Ccl3, Cxcl9, Cxcl10, Cxcl11, Cxcl13), CTL promoters such as membrane-spanning 4-domains, subfamily A, member 4B (Ms4a4b), Th1 cytokines and transcription factors (Ifn-γ, IL-2, Tnf, Tbx21), classical IFN-γ stimulated genes (Stat1, Irf1, Igtp, Fam26f), adhesion molecules (Icam1, Lfa-1), T cell activation markers (Icos, CD40L, CD28, CD27), and molecules directly involved in tumor cell recognition and cytotoxic effector function (Granzyme A, Perforin, Dnam-1, Nklam) (Fig. 7C, Supplementary Fig. S6B).37,38
Overall, transcriptional analysis suggested a potent switch from chronic to acute inflammation mediated by activation of the Stat-1/Irf-1 pathway, similar to an immune response to intracellular viral pathogens. Notably, the powerful intratumoral immune activation that we observed in response to the combination treatment also caused upregulation of genes encoding immunosuppressive molecules such as Pdcd1 (encoding PD-1), Havcr2 (Tim-3), Ido1, Pdcd1lg1 (PD-L1), Pdcd1lg2 (PD-L2), and Ctla-4. MDSC markers CD36 and Arg1 were upregulated to a lesser extent (Fig. 7C, Supplementary Fig. S6B). Together with the above described TME immune cellular profiling results, the molecular profiling data indicated a highly synergistic and effective immunomodulatory activity in the TME by the E7046 and E7777 combination, which offers a mechanistic explanation for the potent anti-tumor activity and long lasting immune memory response.
We have shown that PGE2-EP4 signaling plays a specific and important role in driving pro-tumor myeloid cell differentiation and function, and that pharmacological blockade of EP4 signaling by E7046, a novel EP4 antagonist in early clinical development, is able to largely reverse PGE2-induced myeloid immunosuppression and thus promote anti-tumor immune responses. The specific role of EP4, but not other EP receptors, was unexpected because both EP2 and EP4 have been reported by others to contribute to PGE2-mediated activity on immune cells.39 This discrepancy may be explained by differential cellular and tissue expression of EP2 and EP4, and also by the incomplete specificity of agents used in earlier studies. Our results are consistent with a recent finding that selective deletion of EP4 in macrophages effectively inhibited intestinal adenoma development and growth in colon cancer-prone APCmin/+ mice.5 The significant reduction of tumor growth in inducible EP4−/− compared with WT mice supports the important role of host cell EP4 in tumor progression. Importantly, though myeloid cells appear to be the major direct target cell population for E7046, in vivo anti-tumor activity of E7046 requires CD8+ cytolytic T cells but not CD4+ T cells, suggesting a unique mechanism of action of E7046 in which reduced myeloid cell-based immunosuppression facilitates a cytolytic CD8+ T cell-dependent anti-tumor activity. Accumulation of T cells in E7046-treated compared with vehicle-treated tumors supports this notion. Further studies will be required to show the full extent of in vivo activation of these cells in response to E7046.
Because E7046 affects M2TAMs and mMDSCs but not Treg cells, we hypothesized that by combining EP4 signaling blockade with Treg cell reduction we would obtain an enhanced immunomodulatory anti-tumor effect. Indeed, we identified a strong synergy in inducing tumor regression and/or growth control in multiple animal models when both agents were administered simultaneously. Immunophenotyping at both cellular and molecular levels demonstrated a strong synergistic effect of the combination in the initiation of acute inflammation and immune activation for an effective anti-tumor response. The granzyme B+ CTL effectors detected in CT26 tumors which are contributing to this response also stained positive for live/dead dye which is an indication of their short-lived nature as well as their possible increased susceptibility to E7777 targeting. Notably, the combination of E7046 + E7777 was more effective than the combination of anti-PD-1+anti-CTLA-4 dual checkpoint blockade therapy in the “cold” 4T1 breast cancer model, which is characterized by strong myeloid cell-mediated immunosuppression and is typically refractory to T cell checkpoint inhibitors. This finding indicates that PGE2-EP4 signaling blockade coupled to Treg reduction would be a synergistic anti-tumor combination in general. In support of this notion, combination of E7046 and anti-CTLA-4 antibodies, known to reduce Treg cells in both cancer patients and preclinical models,24,25,40 was superior to either agent alone.
E7046 + E7777 treatment activated the innate immune system which led to an “inflamed” TME evidenced by expression of genes encoding Ccl5/RANTES, macrophage inflammatory protein-1α (MIP-1α/Ccl3), Cxcl9, Cxcl10, and Cxcl11, previously found associated with tumor destruction.41 These chemokines are well known for their anti-tumor roles in recruiting and activating a wide array of immune cells mediated through a common receptor Cxcr3,41 also upregulated by the combination. Cxcl13, encoding a B and T cell chemoattractant, was the most upregulated chemokine gene in response to combination treatment. Although previously associated with a Th2 response,42 a recent study has found that expression of Cxcl13 is strongly associated with B, follicular helper T, Th1, and cytotoxic T cells, and also correlates with a significantly prolonged disease-free survival time in both preclinical models and colorectal cancer patients.43 E7046 + E7777 downregulated the expression of Ccl12, a chemokine gene reported to be expressed in MHCIIloTAMs30 and the pf4 gene (platelet factor 4 or Cxcl4), which hints at platelets as additional targets of the combination. Interestingly, like PGE2, CXCL4 has been shown to bias human monocyte differentiation into CD1a− DCs which were less capable than CD1a+ DCs in stimulating allogeneic T cells to proliferate.44 It is conceivable then that EP4 antagonism combined with Treg reduction is able to block myeloid cell differentiation at multiple levels.
While cellular and molecular immunophenotyping demonstrated a robust reprogramming of both innate and adaptive anti-tumor immunity in response to E7046 + E7777 combination therapy, it also induced expression of genes encoding immunosuppressive molecules such as ido1, arg1, PD-1, PD-L1, and Tim3. Though PD-1 expression has not always been associated with the exhausted T cell phenotype,45 our results suggest that exhausted T cells and their ligand-expressing immunosuppressive cells are present within the treated tumors and thus, addition of checkpoint inhibitors to this treatment should further enhance the anti-tumor response. In this context, EP4 was found expressed on terminally exhausted CTLs and its expression has been correlated with PD-1 expression in conditions involving persistent antigen stimulation such as chronic viral infections and cancer.10,46,47 The co-expression of EP4 and PD-1 on exhausted T cells raises the intriguing possibility that dual blockade of EP4 and PD-1 may limit T cell exhaustion. PD-L1 expression on TAMs as well as on tumor cells has been found associated with resistance to immunotherapies but is also dependent on CD8+ T cell presence.48
In conclusion, we have shown here that a specific EP4 blockade by E7046 reversed the PGE2-induced myeloid-mediated immunosuppression and that E7046 + E7777 synergistically mitigated both myeloid and Treg-derived immunosuppression within the TME, to promote robust anti-tumor immune responses. Either E7046 alone or E7046 + E7777 combination therapy may benefit cancer patients with tumors enriched in myeloid and Treg cells. These approaches are the subject of on-going further investigation.
No potential conflicts of interest were disclosed.
We greatly acknowledge Angel Patel, Yiming Jiang, Haili Zhang, Wei Li, Hui Fang, Farid Benayoud, Hans Hansen, Noel Taylor, Shannon McGrath, Muzaffar Akram, Barbara Kalinowska, Dinesh Chandra, Mark Matijevic, Marc Lamphier, and Donna Kolber-Simonds for their excellent technical support. We thank Amy Siu, Yvonne Van Gessel, Chean Eng Ooi, Ozlem Ataman, Ilian Tchakov, John Wang, Frank Fang, Janna Hutz, Akihiko Koyama, Nadeem Sarwar, Yasuhiro Funahashi, Toshimitsu Uenaka, and Takashi Owa for scientific discussion and comments.