Inhibition of CSF-1 Receptor Improves the Antitumor Efficacy of Adoptive Cell Transfer Immunotherapy

Established solid tumors consist of both transformed neoplastic cells and nontransformed host cells such as stromal cells, lymphocytes, dendritic cells, macrophages, and myeloid-derived suppressor cells (MDSC). In order to escape immune responses, tumor cells manipulate the surrounding tumor microenvironment by producing cytokines that suppress cytolytic T cells and recruit immunosuppressive cells (1–3). Colony-stimulating factor 1 (CSF-1) is a cytokine frequently produced by several cancers, including melanoma (4, 5). The secreted CSF-1 binds to the tyrosine kinase receptor CSF-1 receptor (CSF-1R) on the myeloid cells, which results in increased proliferation and differentiation of myeloid cells into type M2 macrophages and MDSCs, and their recruitment into tumors (6, 7). The M2-polarized macrophages and MDSCs use several mechanisms to induce an immunosuppressive tumor environment, such as the release of arginase I or inducible nitric oxide synthase, leading to T-cell inhibition (1). Therefore, an immunosuppressive tumor milieu mediated by CSF-1 may limit the antitumor activity of tumor immunotherapy and lead to low response rates (3).

In prior studies, we have established a BRAF^V600E mutant murine melanoma cell line, SM1, to provide a relevant model of melanoma in fully syngeneic immunocompetent mice (8). BRAF^V600E is the driver oncogene in approximately 50% of human melanomas (9). Besides being driven by the BRAF^V600E oncogene, SM1 has multiple genomic aberrations in a pattern similar to 108 human melanoma cell lines based on results of high-density single-nucleotide polymorphism/copy number alteration arrays. It includes amplification of oncogenic BRAF^V600E and of the microphthalmia-associated transcription factor, and a deletion of CDKN2A. In the SM1 model, adoptive cell transfer (ACT) of melanoma-targeted T cells induces antitumor responses that are enhanced by the addition of the BRAF inhibitor vemurafenib (8). In addition, this mouse model has been used to test the antitumor effects of several immune modulating antibodies, including anti-CTLA4, anti-PD-1, anti-TIM3, and agonistic anti-CD137 (41BB; ref. 10). Only anti-CD137 had significant antitumor activity alone or in combination with BRAF inhibitor therapy, suggesting that there may be factors in the tumor microenvironment that inhibit the effector arm of the immune system.

PLX3397 is a potent tyrosine kinase inhibitor that is selected for its ability to inhibit CSF-1R. It is currently in clinical development as a single agent or in combination therapy for the treatment of patients with glioblastoma, breast cancer, and other cancers through its inhibition of the CSF-1R, and also acute myelogenous leukemia through its inhibition of FLT3-ITD. In preclinical models, CSF-1R inhibitors including PLX3397 have been reported to inhibit the immunosuppressive tumor milieu and facilitate immune responses to cancer (11–14). We hypothesized that a combination of PLX3397 and ACT immunotherapy would improve the tumor microenvironment through the inhibition of immunosuppressive myeloid cells, resulting in better T-cell antitumor functions. Our results demonstrate significantly enhanced efficacy of the combined treatment, mediated by decreasing TIMs and increasing activated tumor-infiltrating lymphocytes (TIL) compared with either of the single treatment groups.

C57BL/6 mice, OT-1 transgenic mice (The Jackson Laboratories), pmel-1 (Thy1.1) transgenic mice (kind gift from Dr. Nicholas Restifo, Surgery Branch, National Cancer Institute, Bethesda, MD), and NOD/SCID/γ chain^null (NSG) mice (NOD.Cg-Prkdc^scidIl2rg^tm1Wjl/SzJ; The Jackson Laboratory) were bred and kept under defined-flora pathogen-free conditions at the Association for Assessment and Accreditation of Laboratory Animal Care International-approved animal facility of the Division of Experimental Radiation Oncology, University of California, Los Angeles (UCLA), Los Angeles, CA, and used under the UCLA Animal Research Committee protocol #2004-159. The B16 murine melanoma cell line was obtained from the American Type Culture Collection and maintained in Dulbecco's Modified Eagle Medium (Mediatech, Inc.) with 10% fetal calf serum (FCS; Omega Scientific) and 1% penicillin, streptomycin, and amphotericin (Omega Scientific). The SM1 murine melanoma was generated from a spontaneously arising tumor in BRAF^V600E mutant transgenic mice as previously described (15). SM1 was maintained in RPMI (Mediatech) with 10% FCS (Omega Scientific), 2 μmol/L l-glutamine (Invitrogen) and 1% penicillin, streptomycin, and amphotericin. SM1-OVA was generated by the stable expression of ovalbumin (OVA) through lentiviral transduction as previously described (15). PLX3397 was obtained under a materials transfer agreement with Plexxikon Inc.. PLX3397 was dissolved in dimethyl sulfoxide (DMSO; Fisher Scientific). For in vivo studies, PLX3397 was dissolved in DMSO, and then in a suspension made by dilution into an aqueous mixture of 0.5% hydroxypropyl methyl cellulose and 1% polysorbate (PS80; Sigma-Aldrich). Of note, 100 μL of the suspended drug was administered by daily oral gavage into mice at 50 mg/kg when tumors reached 3 mm in diameter. For macrophage depletion studies, 1 mg of clodronate (Clodrosome) was injected intraperitoneally (i.p.) every 5 days. For antibody-mediated depletion studies, 250 μg of anti-CD8 antibody, 200 μg of anti-CSF-1, or isotype control antibody (BioXCell) was injected i.p. every 3 days.

Murine melanoma cells (5 × 10³ cells per well) and activated C57BL/6 splenocytes (5 × 10⁴ cells per well) were seeded on 96-well flat-bottomed plates with 100 μL of 10% FCS media and incubated for 24 hours. Graded dilutions of PLX3397 or DMSO vehicle control, in culture medium, were added to each well in triplicate and analyzed by using a tetrazolium compound MTS-based colorimetric cell proliferation assay (Promega).

B16, SM1-OVA, or SM1 cells were implanted subcutaneously in C57BL/6 mice, and when tumors reached 5 mm in diameter, mice were conditioned for ACT with a lymphodepleting regimen of 500 cGy of total body irradiation. Then they received 2 × 10⁵ or 1 × 10⁶ OVA_257-264 peptide-activated OT-1 splenocytes or gp100_25-33 peptide-activated pmel-1 splenocytes intravenously as previously described (15). In both cases, the ACT was followed by 3 days of daily intraperitoneal administration of 50,000 IU of interleukin (IL)-2. Tumors were followed by caliper measurements three times per week.

SM1 tumors, lungs, blood, bone marrow, and spleens were harvested from mice. Tumors and lungs were further digested with collagenase (Sigma-Aldrich). TIMs obtained from digested SM1 tumors were stained with antibodies to Gr-1, CD11b, F4/80, MHCII (eBiosciences), and Ly6C (BD Biosciences). TILs were stained with antibodies with CD3, Thy1.1 (BD Biosciences), CD4, and CD8 (eBiosciences) and analyzed with a LSR-II or FACSCalibur flow cytometers (BD Biosciences), followed by Flow-Jo software (Tree-Star) analysis as previously described (16). Intracellular IFN-γ staining was done as previously described (16).

Staining was performed as previously described (15). Briefly, sections of OCT (Sakura Finetek) cryopreserved tissues were blocked in donkey serum/PBS and incubated with primary antibodies to Gr-1 (BD Biosciences) or F4/80 (Abcam), followed by secondary donkey anti-rat antibodies conjugated to DyLight488 (Jackson Immunoresearch Laboratories). Negative controls consisted of isotype-matched rabbit or rat immunoglobulin G in lieu of the primary antibodies listed earlier. 4′,6-Diamidino-2-phenylindole (DAPI) was used for the visualization of nuclei. Immunofluorescence images were taken in a fluorescence microscope (Axioplan-2; Carl Zeiss Microimaging).

OT-1 or pmel-1 splenocytes were retrovirally transduced to express firefly luciferase as previously described (15), and used for ACT. Bioluminescence imaging (BLI) was performed with a Xenogen IVIS 200 Imaging System (Xenogen/Caliper Life Sciences) as previously described (15).

Data were analyzed with GraphPad Prism (version 5) software (GraphPad Software). A Mann–Whitney test or ANOVA with Bonferroni posttest was used to analyze experimental data. Survival curves were generated by the actuarial Kaplan–Meier method and analyzed with the Jump-In software (SAS) with log-rank test for comparisons from the time of tumor challenge to when mice were sacrificed due to tumors reaching 14 mm in maximum diameter, or the end of the study period had been reached.

We tested the effects of single agent PLX3397 against SM1 using an in vitro MTS cell proliferation assay after 72 hours of exposure to rule out a potential direct antitumor effect of this CSF-1R inhibitor in the SM1 murine melanoma cell line. An IC₅₀ was not reached even at 1 μmol/L (Fig. 1A). In addition to its resistance in MTS assays, exposure of SM1 to PLX3397 at the range of concentrations between 10 nmol/L and 1 μmol/L showed no inhibition of the downstream mitogen–activated protein kinase signaling pathway (Fig. 1B). SM1 does not express either CSF-1R or c-kit by fluorescence-activated cell sorting (FACS) analysis and gene expression profiling (data not shown); these are the two main targets of PLX3397. SM1 also produces high levels of CSF-1 protein and its level is not affected by PLX3397 at 1 μmol/L (Supplementary Fig. S1).

We then ruled out that PLX3397 had a potential detrimental effect against T cells. Increasing concentrations of PLX3397 did not negatively alter the viability of murine splenocytes after 72 hours of treatment (Fig. 1C). In contrast, 1 μmol/L of PLX3397 was enough to downregulate pErk level and to kill cells that were dependent on CSF-1/CSF-1R for growth such as myeloid cells and microglia cells (12, 17). Therefore, because single-agent PLX3397 did not affect the viability of either SM1 cells or murine splenocytes, this supports SM1 as a permissive model for testing the effects of PLX3397 to improve the antitumor activity of ACT immunotherapy.

Lymphodepleted C57BL/6 mice with established subcutaneous SM1-OVA tumors received ACT of splenocytes obtained from OT-1 mice expressing an OVA-specific T-cell receptor (TCR). We titrated the application of immunotherapy to provide a suboptimal antitumor effect that was similar to the antitumor effect of single agent PLX3397 so that the effect of combination could be revealed (Fig. 2A). The combined therapy for PLX3397 and OT-1 TCR transgenic ACT demonstrated superior antitumor effects compared with either therapy alone in duplicate experiments and improved overall survival (Fig. 2B; Supplementary Fig. S2a). As the OVA model is based on the recognition of a foreign antigen, we further confirmed the results in the pmel-1 ACT model that is based on transgenic T cells with a TCR recognizing gp100, a murine melanosomal antigen endogenously expressed by SM1 (Fig. 2C; ref. 8). In replicate studies, the combined therapy with pmel-1 ACT and PLX3397 also had superior antitumor response compared with either single agent therapy alone and improved survival (Fig. 2D; Supplementary Fig. S2a). To also test the general applicability of combining PLX3397 and pmel-1 ACT, another murine melanoma model, B16 was used. Combined therapy also demonstrated superior antitumor response compared with either therapy alone in duplicate experiments (Supplementary Fig. S2b).

To analyze whether PLX3397 changed the magnitude of TIMs, including macrophages and MDSCs, we analyzed their presence in tumors by immunofluorescence. There was a decrease in the quantity of F4/80(+) macrophages in both the PLX3397 single agent group and the combined group compared with vehicle or ACT single treatment groups (Fig. 3A). Analysis of tumors, spleens, lungs, blood, and bone marrow from mice treated with PLX3397 demonstrated that the effects of PLX3397 were specific for intratumoral macrophages as opposed to a systemic depletion of macrophages. There was a local (tumor) decrease in quantity of F4/80(+) CD11b(+) macrophages but no significant decrease systemically (Fig. 3B and D; Supplementary Fig. S3). To further characterize the phenotype of the remaining macrophages, MHCII expression was analyzed by FACS. There was a shift in macrophage phenotype from MHCII^low to MHCII^hi upon treatment with PLX3397 (Fig. 3C and D; Supplementary Fig. S3).

The level of MDSCs in SM1 tumors was first analyzed by immunofluorescence. The MDSCs were present at a low level in tumors and their level seemed to be unaltered by ACT or PLX3397 treatment (Fig. 4A). To better enumerate the magnitude and distribution of MDSCs in vivo, we analyzed their presence in tumors, spleens, lungs, blood, and bone marrow by flow cytometry. Confirming the immunofluorescence data, there was no change in the already low (∼4%–6%) baseline quantity of Gr-1(+) CD11b(+) MDSCs following treatment with PLX3397 (Fig. 4B and D; Supplementary Fig. S3). We then examined whether PLX3397 altered the ratio between the two recognized subsets of MDSCs, the polymorphonuclear MDSCs (PMN-MDSC, Gr-1^hi Ly6C^low), and the monocytic MDSCs (MO-MDSC, Gr-1^low Ly6C^hi). There was no significant difference between PLX3397 treated and nontreated groups in MDSC subsets, just as there was no change in total MDSCs (Fig. 4C and D; Supplementary Fig. S3).

In order to confirm the role of CSF-1 in the activity of PLX3397, PLX3397 treatment was combined with anti-CSF-1 antibody in mice that received ACT with pmel-1 splenocytes. There was no improvement in antitumor activity in the PLX3397, pmel-1 ACT, and anti-CSF-1 antibody treatment groups, compared with mice receiving PLX3397 and pmel-1 ACT therapy with an isotype control antibody (Fig. 5A), suggesting that the effects of PLX3397 and anti-CSF-1 antibody-mediated depletion might be overlapping. To determine if the target of PLX3397 was macrophages, mice with established SM1 tumors were treated with PLX3397 in combination with clodronate, an agent that depletes macrophages. There was no enhanced antitumor response in the combined treatment group of PLX3397 and clodronate compared with either single treatment group (Fig. 5B). These studies suggest that PLX3397 mediated its effects through inhibition of immunosuppressive CSF-1–responding intratumoral macrophages.

To analyze whole animal T-cell distribution in the presence or absence of PLX3397, we genetically labeled the adoptively transferred OT-1 T cells with firefly luciferase transgene for in vivo BLI. There was an increased expansion, in vivo distribution to tumor, and tumor targeting by adoptively transferred OT-1 T cells when mice were treated with PLX3397 (Fig. 6A). The quantitative analysis of T-cell-associated luciferase activity in mice treated with OT-1 ACT combined with PLX3397 showed increased accumulation within OVA antigen-matched tumors over time (Fig. 6B; Supplementary Fig. S4a). Furthermore, we repeated BLI using pmel-1 T cells also labeled with the firefly luciferase transgene and used for ACT. Again, PLX3397 increased the expansion of pmel-1 T cells distributed to gp100 positive antigen-matched SM1 tumors (Fig. 6C). Mice that received pmel-1 ACT combined with PLX3397 also demonstrated an increased intratumoral accumulation of luciferase activity to tumors over time (Fig. 6D; Supplementary Fig. S4b). We then analyzed the activation state of TILs by detecting cytokine production. In two replicate experiments, TILs collected from mice treated with the combination of pmel-1 ACT and PLX3397 showed a higher ability to secrete IFN-γ and to respond to short term ex vivo restimulation with the gp100 antigen (Fig. 6E). Therefore, treatment with PLX3397 increased both the number and functionality of adoptively transferred antitumor antigen-specific T cells.

Single agent PLX3397 treatment demonstrated a weak but reproducible antitumor response compared with vehicle control (Figs. 2B, C, and 5B). We tested if this antitumor activity was mediated by endogenous cytotoxic CD8 T cells. SM1 tumors were implanted in C57BL/6 mice without irradiation and received PLX3397 in combination with anti-CD8 antibody. The depletion of CD8+ cells abrogated the antitumor activity of single agent PLX3397 (Fig. 7A). To further test the role of immune cells in the antitumor activity of PLX3397, immunodeficient NSG mice were implanted with SM1 tumors and dosed with PLX3397. In these immunodeficient mice there was no antitumor activity of PLX3397 compared with mice receiving vehicle control (Fig. 7B). Finally, we tested whether the antitumor effects of PLX3397 was also dependent on cytotoxic CD8 T cells in the pmel-1 ACT model. C57BL/6 mice with established SM1 tumors were treated with a combination of PLX3397, anti-CD8 antibody, and pmel-1 ACT. There was no antitumor activity of pmel-1 ACT combined with PLX3397 in mice that received CD8-depleting antibody (Fig. 7C). Collectively, these studies highlight the role of CD8+ T cells as effectors of the antitumor activity of PLX3397 in the SM1 murine melanoma model.

Tumor immunotherapy with cytokines such as IFN or IL-2, immune modulating antibodies blocking the cytotoxic T-leukocyte-associated antigen-4 (CTLA4) or the programmed cell death receptor 1 (PD-1) or its ligand PD-L1, and the adoptive transfer of tumor antigen-specific T cells, result in tumor responses in patients with advanced cancers, and melanoma in particular. These immunotherapy-induced tumor responses are frequently extremely durable and can last years. However, their low response rate has been one of the obstacles remaining to overcome (3, 18). A potential explanation is that intratumoral immunosuppressive myeloid cells may inhibit immunotherapy responses via different mechanisms (1). These myeloid lineage cells represent various distinct and heterogeneous populations of immunosuppressive cells, including MDSCs and macrophages. By suppressing these myeloid cells with CSF-1R blockade using PLX3397, we hypothesized that T-cell function within tumors may improve. Hence, the ideal treatment would be to suppress the myeloid cells and increase the number of effector T cells at the same time. SM1 tumors grow very rapidly when implanted in C57BL/6 mice, such that mice often need to be sacrificed within 2 to 3 weeks of tumor implantation, limiting the ability to induce an immune response with immune modulating antibodies (8, 10). To overcome this, and provide the opportunity for a fully effective immune response in combination with CSF-1R blockade, we have used the ACT of T cells expressing TCRs recognizing a specific tumor antigen on the tumor cells. The scientific rationale posits that inhibition of CSF-1R signaling in immunosuppressive TIMs will improve the intratumoral milieu by taking away immunosuppressive factors that limit the antitumor activity of cytotoxic T lymphocytes.

We explored the potential mechanisms by which PLX3397 improves the antitumor effect of ACT. The myeloid cells infiltrating SM1 tumors are mainly macrophages. Our data demonstrate that PLX3397 decreases the quantity of macrophages locally in tumor. Although the classification between M1 and M2 polarized macrophages is still controversial, MHCII is used as a marker to distinguish them (19). With PLX3397, there is a skewing of the population of MHCII^low to MHCII^hi macrophages, consistent with a previous report (13). By decreasing the presence of immunosuppressive MHCII^low macrophages, PLX3397 likely facilitates the intratumoral trafficking of adoptively transferred lymphocytes and their antitumor functions. Our studies also suggest that macrophage inhibition alone by PLX3397 is not sufficient for an antitumor response. An immune response is indeed needed for the antitumor effect of PLX3397 in the SM1 model because this effect requires CD8 T cells. Therefore, the main beneficial effects of PLX3397 in this model are derived from the ability to improve T-cell effector functions indirectly through the inhibition of intratumoral immunosuppressive macrophages.

Reports have shown that different tumor models such as prostate and lung carcinoma models are MDSC-dominant and these MDSCs are heavily infiltrated in tumors and systemic organs such as spleen, blood, and bone marrow (14, 20). By contrast, our SM1 model has few MDSCs infiltrating the tumor or accumulating in the spleen over time. This is consistent with a recent report that shows patients with melanoma have far less MDSCs than other cancers and also compared with many nonmelanoma murine tumor models (21). In addition, prior reports have shown that CSF-1R blockade therapy reduced the number of myeloid cells including MDSCs and macrophages in both tumors and systemic organs (4, 14). In contrast, in our model, CSF-1R blockade with PLX3397 mainly targeted the more abundant macrophages instead of MDSCs, suggesting the role of CSF-1 may be tumor model-dependent. Different tumor models, genetic backgrounds, or treatments may induce different growth factors or cytokines in the tumor microenvironment. As myeloid cell tumor infiltration is a complex process regulated by different pathways, it is possible that different tumor models and tumors from different patients producing different cytokines like CSF-1, IL-34, or CCL2 result in attraction of different myeloid lineages leading to differential responses to CSF-1R blockade (22–25). For example, it has been reported that under hypoxic conditions, induced expression of hypoxia-inducible factor-1α results in the expression of CXCR4 and SDF-1. The SDF-1/CXCR4 axis is another pathway that mediates the recruitment of TIMs to tumors (26, 27). Future immunotherapy trials may thus benefit from molecular profiling-based stratification of patients in regards to their tumor microenvironment cytokines.

In conclusion, combined therapy with the CSF-1R inhibitor PLX3397 and TCR-based ACT immunotherapy results in superior antitumor effects than single agent treatment in a murine model of melanoma. The antitumor activity is mediated by inhibition of the myeloid cell-mediated immunosuppressive tumor microenvironment so that more lymphocytes infiltrate into the tumor with enhanced functionality.

G. Bollag is the CEO of Plexxicon. B. L. West is an employee of Plexxikon. No potential conflicts of interest were disclosed by the other authors.

Conception and design: S. Mok, R.C. Koya, J. Xu, L. Wu, A. Ribas

Development of methodology: S. Mok, R.C. Koya, L. Robert, A. Ribas

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Mok, R.C. Koya, C. Tsui, J. Xu, L. Robert, B.L. West, G. Bollag, A. Ribas

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Mok, R.C. Koya, J. Xu, T.G. Graeber, A. Ribas

Writing, review, and/or revision of the manuscript: S. Mok, R.C. Koya, L. Wu, B.L. West, G. Bollag, A. Ribas

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S. Mok, C. Tsui

Study supervision: R.C. Koya, A. Ribas

This work was funded by the NIH (grants P01 CA168585, P50 CA086306, and R21 CA169993), the Seaver Institute, the Louise Belley and Richard Schnarr Fund, the Wesley Coyle Memorial Fund, the Garcia-Corsini Family Fund, the Bila Alon Hacker Memorial Fund, the Fred L. Hartley Family Foundation, the Ruby Family Foundation, the Jonsson Cancer Center Foundation, the Eli & Edythe Broad Center of Regenerative Medicine and Stem Cell Research at UCLA, and the Caltech–UCLA Joint Center for Translational Medicine (A. Ribas and T.G. Graebar).

T.G. Graebar is supported by an American Cancer Society Research Scholar Award (RSG-12-257-01-TBE), the Caltech–UCLA Joint Center for Translational Medicine, the National Center for Advancing Translational Sciences UCLA CTSI Grant UL1TR000124, and a Concern Foundation's CONquer CanCER Now Award.

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.