CSF1/CSF1R Signaling Inhibitor Pexidartinib (PLX3397) Reprograms Tumor-Associated Macrophages and Stimulates T-cell Infiltration in the Sarcoma Microenvironment

Tumor-associated macrophages (TAMs) represent a substantial proportion of all tumor-infiltrating immune cells in the tumor microenvironment (TME). TAMs have a dominant role as orchestrators of tumor-related inflammation toward tumor growth and progression and are potential new therapeutic targets (1, 2). The tumor-promoting function of TAMs is based on their ability to secrete proangiogenic and growth factors, as well as to potently suppress T-cell effector function by releasing immunosuppressive cytokines (1, 3, 4). Two opposing phenotypes, termed classically activated (or M1-like) and alternatively activated (or M2-like), have been associated with anti- and protumoral functions (1, 2). Clinically, M2-like TAMs are associated with high tumor grade and poor prognosis in many carcinomas and sarcomas (1, 2, 5). TAM-targeting treatment strategies are under early-phase investigation in various cancer types (2, 6), but not in sarcoma, although transdifferentiating M2-like macrophages to M1-like macrophages has great potential as a new sarcoma therapy.

Despite major advances in multimodality therapies for sarcomas, more than one-third of patients show a poor response to conventional therapy, resulting in subsequent recurrence and a poor prognosis (7, 8). The overall survival for osteosarcoma is 60%–70% at five years (9), but less than 30% for patients who present with metastatic disease (7). Similarly, the median overall survival is around one year for metastatic soft-tissue sarcomas (10), and approximately 10% of patients are alive at five years (11, 12), highlighting the need for innovative therapeutic approaches.

Colony-stimulating factor 1 (CSF1) plays a significant role in the recruitment of peripheral blood monocytes to the TME, differentiation into macrophages, and polarization of macrophages toward an M2-like phenotype via binding to the CSF1 receptor (CSF1R; ref. 13). Preclinical studies have shown that CSF1R blockade may inhibit tumor growth in several cancer models through the elimination or polarization of TAMs (14, 15). PLX3397 (pexidartinib), a potent CSF1R inhibitor, was recently approved by the Food and Drug Administration (FDA) to treat symptomatic tenosynovial giant cell tumor that expresses CSF1 (16, 17, 18). Although several preclinical and clinical trials have been ongoing for various malignant tumors (17, 19, 20), data regarding PLX3397 in sarcomas are limited (21).

In this study, we investigated the efficacy and safety of PLX3397 in our established in vitro and in vivo sarcoma TAM model. We demonstrate that CSF1/CSF1R blockade using PLX3397 suppresses survival, migration, and M2 polarization of sarcoma TAMs in vitro and significantly inhibits osteosarcoma growth and metastatic spread in vivo. Furthermore, we found that PLX3397 administration resulted in not only depleting TAMs but also altering T-cell infiltration in the TME. This preclinical trial demonstrates that PLX3397 is a promising immunotherapeutic agent and would have immediate clinical implications for sarcomas, which should be explored for further clinical development.

The murine osteosarcoma cell line LM8 (RCB1450), fibrosarcoma cell line NFSa (RCB0282), mesenchymal stem cell line KUM5 (RCB2322), and fibroblastic connective tissue cell line LAG (RCB2758) were purchased from the Riken Cell Bank. These cell lines have C3H mouse origin, and the LM8 cell line, a highly metastatic cell line, was derived from Dunn osteosarcoma (22). LM8 clones expressing luciferase and dTomato, named LM8-Luc, were generated with a lentiviral vector (Supplementary Methods and Supplementary Fig. S1). All cell lines were tested negative for Mycoplasma contamination (CycleavePCR Mycoplasma Detection Kit, Takara Bio). Cell lines were maintained in Dulbecco's modified Eagle's medium (DMEM; Life Technologies, Inc.), containing 10% heat-inactivated fetal bovine serum (FBS; Life Technologies), 1% penicillin (100 U/mL), and streptomycin (100 mg/mL; Life Technologies). The cells were maintained under 5% CO₂ in a humidified incubator at 37°C.

LM8 and NFSa cells were seeded at 1 × 10⁷ cells per T175 cell culture flask in 30 μL of DMEM containing 10% FBS and 1% penicillin and streptomycin. After 48 hours, the supernatant was collected, passed through a 0.2-μm vacuum filter (VWR), and stored at −20°C.

Cytokine array analysis was performed using 500 μL of freshly stored TCM and the Mouse Cytokine Array Panel A (R&D Systems), according to the manufacturer's protocol. IL34 quantification was performed using 50 μL of freshly stored TCM and the Mouse IL34 Quantikine ELISA Kit (R&D Systems), according to the manufacturer's protocol.

Bone marrow cells were isolated and harvested from the femur, tibia, and humeri of 6–8-week-old C3H/HeJ female mice (Jackson Laboratories, cat. #JAX:000659, RRID:IMSR_JAX:000659), as previously described (23) with the Hospital for Special Surgery's (HSS) Institutional Animal Care and Use Committee (IACUC) approved protocol (#2016-0039). Bone marrow cells were differentiated into BMDMs with α-Minimum Essential Medium (α-MEM; Life Technologies) supplemented in 10% FBS, 1% penicillin and streptomycin, and 10% conditioned media (CM; equivalent to 140 ng/mL CSF1) from the CSF1 overproducing cell line CMG (23, 24) or TCM from cultured LM8 or NFSa for five days until confluent.

The CSF1R inhibitor PLX3397 was purchased (Bioactive Molecular Research), and its analytical profile was further confirmed in our laboratory and compared with previously reported data (16). For in vitro studies, a 10 mmol/L stock of PLX3397 was formulated in dimethyl sulfoxide (DMSO), which was used as the vehicle control. For in vivo studies, PLX3397 was formulated in 20% DMSO at a concentration of 5 and 10 mg/kg.

Pazopanib (MedChemExpress, cat. #HY-10208, RRID: N/A, Monmouth Junction), anti-mouse CSF1 monoclonal antibody (mAb; BioCell, BE0204), rat IgG1 isotype control (BioCell, BE0290), anti-mouse CSF1R mAb (BioCell, BE0213), and rat IgG2a isotype control (BioCell, BE0089) were purchased and used according to the manufacturer's protocol.

All animal studies were approved by HSS's IACUC (#2016-0039). Six- to 8-week-old female C3H/HeJ mice (Jackson Laboratory) were anesthetized by exposure to 3% isoflurane on day 0 and subsequent days. On day 0, the mice were anesthetized with 3% isoflurane, and the right leg was disinfected with 70% ethanol. A 100 μL volume of solution containing 1 × 10⁶ LM8-Luc cells was orthotopically injected into the proximal tibia (25).

Individual mice were retro-orbitally injected with 100 μL of PLX3397 (5 or 10 mg/kg) or control phosphate-buffered saline (PBS) per injection on days 7 and 14 after inoculation of LM8-Luc cells. Each experimental condition included five animals per group. Photons from firefly luciferase were counted on days 5, 12, 19, and 22 using the IVIS imaging system (Xenogen) following intraperitoneal injection of d-luciferin (150 mg/kg; Promega). Following the evaluation of the primary sites, chest lesions of mice were analyzed individually if the spillage/leakage of photons from the primary sites was suspected. Data were analyzed using LivingImage software (version 4.7.2; Xenogen). On day 22, at the end of the experiment, the primary tumor and lung of each animal were resected at necropsy for histologic analysis. Blood tests, body weight measurement, and a histologic examination of the liver were performed for drug toxicity assessment in each mouse.

All tumors resected from mouse primary tumors and lungs were fixed with 10% buffered formalin before paraffin embedding. A rabbit polyclonal anti-CD68 antibody (Boster Bio; cat. #PA1518, RRID:N/A), a rabbit monoclonal anti-CD8 antibody (Cell Signaling Technology; cat. #98941), and a rat monoclonal anti-FOXP3 antibody (Thermo Fisher Scientific; cat. #14-5773-82, RRID:AB_467576) were used in concentrations of 5, 4.8, and 5 μg/mL, respectively. The tissue sections were blocked for 30 minutes in 10% normal goat serum, 2% bovine serum albumin in PBS. A 5-hour incubation with the primary antibody was followed by a 32-minute incubation with biotinylated goat anti-rabbit IgG (1:200 dilution; Vector Laboratories; cat. #PK4001, RRID:AB_2336810). Sequential incubations were performed with Secondary Antibody Blocker, Blocker D, Streptavidin-HRP, and Tyramide Alexa Fluor 488 (Life Technology; cat. #B40932). Histology slides were digitally scanned with Pannoramic Flash (3DHistech) using 20×/0.8NA objective. Snapshots were taken using CaseViewer Software (3DHistech).

All statistical analyses were performed using Prism (GraphPad Prism version 5; San Diego) or SPSS Statistics version 24 (IBM, Armonk, NY, USA). The unpaired t test or one-way ANOVA, corrected for multiple comparisons as appropriate, was used to determine the significance of any differences between experimental groups. Results were shown as means ±SEM or ±SD as indicated. The Kaplan–Meier method and the log-rank test were used to analyze metastasis-free survival. Metastasis-free survival was defined as the time interval from tumor inoculation until initial metastasis detected by IVIS imaging. The value of P < 0.05 was considered to be statistically significant.

Additional details are in the Supplementary Methods.

To establish an in vitro TAM model of sarcomas, we differentiated bone marrow cells in the presence of CM from the CSF1–producing CMG cell line (24, 26) and control mesenchymal stem cell line KUM5, and TCM from osteosarcoma LM8 cells and fibrosarcoma NFSa cells. To comprehensively profile the production of cytokines/chemokines from CMG, KUM5, LM8, and NFSa cells, we performed cytokine array analysis. We found striking differences among the cytokine/chemokine profiles in the CM from KUM5, CMG, LM8, and NFSa cells (Fig. 1A). The highest production of CSF1 was observed in CM from LM8, followed by NFSa and CMG cells (Fig. 1B). Various other cytokines/chemokines, including IFNγ, chemokine (CXC) ligand (CXCL)-1, chemokine (CC motif) ligand (CCL)-2, CCL-5, CXCL-12, and tissue inhibitor of the metalloproteinase (TIMP)-1 were also produced from CMG cells (Fig. 1A and 1B). However, except for IFNγ and TIMP-1, most of the above were highly produced by LM8 and NFSa cells compared with CMG cells (Fig. 1B). We also identified differences in production levels of cytokines/chemokines including IFNγ-Inducible Protein (IP)-10, CXCL-1, CCL-2, CXCL-9, CXCL-2, and CXCL-12 between LM8 and NFSa cells (Fig. 1B). We further investigated the secretion of IL34, another known ligand for CSF1R, but detected no IL34 secretion from these cells (Supplementary Fig. S1B). Based on these results, we hypothesized that CM from these cell lines would have distinct and differential effects on the proliferation and polarization of macrophages.

Mouse BMDMs were obtained by flushing the long bones of C3H/HeJ mice, which share the same genetic background as the KUM5, LM8, and NFSa cells, and expanded in cultures supplemented with CM from KUM5, CMG, LM8, and NFSa cells. Significantly increased numbers of BMDMs were observed with CM from CMG, LM8, and NFSa cells compared with KUM5 cells (Supplementary Fig. S1C), indicating the proliferation of BMDMs was largely dependent on CSF1. Morphologically, these cells had a spindle-shaped, fibroblast appearance after expansion, which was slightly different depending on the CM used. BMDM-CMG had shorter dendritic extensions, and BMDM-LM8 and -NFSa were more spindle-shaped, but BMDM-LM8 had longer dendritic extensions (Fig. 2A). Macrophage differentiation and polarization are induced with increased mitogen-activated protein kinase (MAPK)/extracellular signal-regulated kinase (ERK) pathway, characterized by the activation of ERK1/2 (27). When purified murine CSF1 (50 ng/mL) was added to the BMDMs, phosphorylation of ERK1/2 (pERK1/2) was enhanced with a peak stimulation after 10 minutes of treatment (Fig. 2B). Similarly, enhanced pERK1/2 was also detected when TCM/LM8 and TCM/NFSa were added to BMDMs at a 50/50 ratio with culture media (Fig. 2B).

To investigate the polarization of BMDMs toward an M2/TAM-like phenotype, we verified the established surface marker expression by flow cytometry and the gene-expression profiles characteristic of M1- and M2-like phenotypes. Flow cytometry analysis revealed BMDM-CMG had a high frequency (94%) of positivity for the M2 marker CD206, indicating polarization toward an M2-like phenotype (Fig. 2C). Yet, BMDM-LM8 and BMDM-NFSa had a relatively more heterogeneous population with 64% and 71% CD206⁺ within CD45⁺CD11b⁺ populations, respectively (Fig. 2C). However, the mean fluorescence intensity (MFI) of CD206⁺/CD80⁺ showed no significant difference among BMDM-CMG, BMDM-LM8, and BMDM-NFSa (Fig. 2C), indicating that these BMDMs are polarized toward M2-like phenotype. Reverse transcription quantitative-PCR (RT-qPCR) analysis revealed BMDM-LM8 and BMDM-NFSa had higher mRNA expression of M1-like genes, including IL1β, iNOS, and CD80 (Supplementary Fig. S1D) than BMDM-CMG, which could be attributed to the GM-CSF secreted from LM8 and NFSa cells (Fig. 1B) (28, 29). Yet, BMDM-NFSa showed greater mRNA expression of M2-like genes, including CD206 and CCL-2, than BMDM-CMG (Supplementary Fig. S1D). These data confirmed TCM/LM8 and TCM/NFSa were able to polarize BMDMs into an M2-like phenotype, while CM/CMG differentiated BMDMs toward a pure M2-like phenotype (2, 30).

Malignant tumors largely recruit circulating monocytes and macrophages into the TME (1, 2). Therefore, we investigated whether BMDMs were recruited under stimulation of CM/CMG, TCM/LM8, and TCM/NFSa by transwell assays. BMDMs were able to migrate through the transwell chamber in response to CM/CMG, TCM/LM8, and TCM/NFSa (Fig. 2D). Notably, the chemotaxis of BMDMs was enhanced by 1.5- to 2-fold change in response to TCM/LM8 and TCM/NFSa as compared with CM/CMG (Fig. 2D). On the other hand, the migration of control LAG cells, mouse fibroblastic mesenchymal cells, which have markedly low expression of CSF1R compared with BMDM-CMG (1: 15,731; P < 0.0001; Supplementary Fig. S1E), was not promoted even under CSF1 stimulation (Supplementary Fig. S1E). These results indicate CM/CMG, TCM/LM8, and TCM/NFSa can recruit macrophages, where TCM/LM8 and TCM/NFSa were more effective than CSF1.

Survival and polarization of macrophages are dependent on CSF1R signaling (5). To validate this contention and gain further molecular insight in this process, we exploited PLX3397, a potent inhibitor of CSF1R signaling. Before investigating the effect of PLX3397 on macrophages, we confirmed that the CSF1R expression was markedly higher in BMDM-LM8 and BMDM-NFSa than in LM8 and NFSa sarcoma cells (Supplementary Fig. S1F). PLX3397, at doses of 50 nmol/L, 100 nmol/L, and 250 nmol/L strongly inhibited the phosphorylation of ERK1/2 levels in Western blots of treated BMDM-CMG, -LM8, and -NFSa cells, that were stimulated by 10 minutes of exposure to CSF1, TCM/LM8, and TCM/NFSa, respectively (Fig. 3A).

Next, we tested if CSF1R inhibition affected macrophage polarization using flow cytometry analysis and RT-qPCR. Flow cytometry data revealed the PLX3397 dose dependently decreased the CD45⁺CD11b⁺CD206⁺ cell populations in BMDM-CMG, BMDM-LM8, and BMDM-NFSa cells (Fig. 3B and Supplementary Fig. S1G). At 500 nmol/L PLX3397, the CD206⁺/CD80⁺ MFIs of all the above cell types were decreased by more than 50% (Fig. 3B). The most significant effect was observed in BMDM-LM8 cells, where the CD206⁺/CD80⁺ cell population (M2-like phenotype) shifted from 95% in untreated BMDM-LM8 cells to 58% in BMDM-LM8–treated cells. There was a 65% decrease in CD206⁺/CD80⁺ MFI (Fig. 3B). Aligning with the flow cytometry data, the RT-qPCR analysis showed that PLX3397 reduced the gene-expression levels of the M2-related markers, CD206 and CCL-2, in a dose-dependent manner (Fig. 3C). In contrast, the gene-expression levels of M1-related genes IL1β, iNOS, and CD80 were increased dose dependently by PLX3397 (Fig. 3C). These data suggest CSF1R inhibition induces gene-expression changes in TAMs and reduces M2 macrophage polarization in bone and soft-tissue sarcoma models.

Next, we investigated the effect of CSF1R inhibition on the viability and chemotaxis of TAMs. Increasing doses of PLX3397 significantly reduced cell viability in a dose-dependent manner (Fig. 3D). The effect on macrophage viability was most striking in BMDM-CMG (Fig. 3D). PLX3397 treatment decreased the CSF1–mediated chemotaxis of BMDM-CMG, -LM8, and -NFSa by migration assays, which was enhanced by CM/CMG, TCM/LM8, and TCM/NFSa (Fig. 3E). After PLX3397 treatment for one hour, the cellular chemotaxis of BMDM-TAMs through the transwell membrane was significantly reduced in a dose-dependent manner (Fig. 3E).

The above results were further verified by using anti–CSF–1 and anti–CSF1R mAbs. Both reduced survival and chemotaxis of sarcoma TAMs, comparable to the effect of PLX3397. They promoted the depolarization away from the M2-like phenotype, although less than PLX3397 did (Supplementary Fig. S2). By comparison, the multikinase inhibitor pazopanib, which strongly targets VEGFR-1, VEGFR-2, and VEGFR-3 and mildly targets PDGFR-β, c-kit, FGFR-1, and CSF1R, reduced survival of sarcoma TAMs, but showed a limited effect on polarization and chemotaxis compared with PLX3397 (Supplementary Fig. S2). Collectively, CSF1/CSF1R inhibition using the blocking mAb and pazopanib confirmed the reduction in survival, depolarization, and migration of sarcoma TAMs, encouraging the preclinical test of PLX3397 to inhibit the development and progression of sarcoma metastasis.

We first determined the ability of LM8 cells infected with a lentiviral vector expressing firefly luciferase to monitor the development of primary tumor at the orthotopic site and its spontaneous metastasis to the lung in the C3H/HeJ mice (22). We transplanted 1.0 × 10⁶ LM8-Luc cells into the right proximal tibia. Tumor growth was analyzed by following the increase in luciferase activity determined as bioluminescence in both primary and lung metastatic sites. Five days after the orthotopic transplantation of the tumor cells, primary tumors were macroscopically detectable in all mice, but no signals were detected in the lungs (Supplementary Fig. S3A). On day 12 after tumor cell transplantation, the first detectable metastasis in some of the mice was observed (Supplementary Fig. S3B). By day 19 after tumor cell transplantation, all animals showed signals in their lungs (Supplementary Fig. S3C). By the end of three weeks following the initial tumor cell transplantation, lung metastases become detectable by luciferase imaging and verified by the bioluminescence of luciferase in the lungs at necropsy. The lung metastatic foci were histologically verified by hematoxylin and eosin (H&E) staining (Supplementary Fig. S3D).

The therapeutic effect of PLX3397 on the primary and spontaneous lung metastasis of osteosarcoma was investigated in the same mouse model. At the end of week one, luciferase activity was detected at the primary lesion in each group (Supplementary Fig. S4A). Tumor-bearing mice (5 per group) received therapeutic retro-orbital doses of 5 mg/kg (low-dose group) or 10 mg/kg (high-dose group) of PLX3397. Control mice received PBS injections. These were administered 7 and 14 days after the orthotopic transplantation of the LM8-Luc cells. Tumor size was monitored weekly after sarcoma inoculation or drug administration, which revealed significantly suppressed growth in high-dose PLX3397 compared with low-dose PLX3397 or PBS cohorts (Fig. 4A and B). The development of a primary tumor and lung metastasis was monitored weekly by bioluminescence imaging. On day 12, one control mouse (20%) developed a signal presumptively in the pulmonary area, suggesting metastasis into the lung (Supplementary Fig. S4B). However, no signal was detected in the lung in the PLX3397-treated groups (Supplementary Fig. S4B). On day 19, luciferase activity in the high-dose PLX3397-treated tumors was significantly weaker than those in the low-dose PLX3397 group and control group (Fig. 4C and D). None of the mice in the high-dose PLX3397 group showed the metastatic signal in the lung (100% inhibition), compared with one mouse (20% tumor-free) in the control group and two mice (40% inhibition) in the low-dose PLX3397 group (Fig. 4E and Supplementary Fig. S4C). The presence of metastases in the resected lung was also confirmed by the post necropsy after all mice were euthanized at the end of three weeks (day 22). All mice in the control group and low-dose PLX3397 group showed relatively strong signals from the resected lung, whereas relatively low-intensity signals were detected in three mice (60%) in the high-dose PLX3397 group (Supplementary Fig. S4D). We observed smaller-sized metastatic osteosarcomas in the resected lungs of high-dose PLX3397 compared with those of low-dose PLX3397 or PBS-treated mice by microscopy (Fig. 4F). The Kaplan–Meier analysis revealed high-dose PLX3397-treated mice showed relatively prolonged metastasis-free survival among the three groups (P = 0.014, log-rank test; Fig. 4G).

No apparent systemic effects and weight loss were detected following drug administrations (Supplementary Fig. S5A). Considering the hepatotoxicity and renal toxicity reported in the human use of this drug, we also evaluated them at baseline and on day 22, measuring aspartate aminotransferase (AST), alanine aminotransferase (ALT), and creatinine (CREA). We observed a trend of an increase in the levels of AST and ALT in two out of five mice (40%) treated with high-dose PLX3397 (Supplementary Fig. S5B and 5C). The levels of lactate dehydrogenase, alkaline phosphatase, total bilirubin, total protein, albumin, cholesterol, and creatinine levels did not show any significant variations among the treatment groups (Supplementary Fig. S5D–S5J). The histopathologic examination of the liver revealed no evidence of necrosis, fibrosis, steatosis, inflammation, or biliary changes in any of the three groups (Supplementary Fig. S6A).

To understand how CSF1R inhibition elicited a potent antitumor response in vivo, we compared the flow-cytometric data of the dissociated tumor cells from the PLX3397-treated mice to that of PBS-treated control. We observed a significantly decreased percentage of dTomato⁺ cells in the tumors of high-dose PLX3397-treated mice, presumably an altered composition of tumor cells versus the cells of TME (Fig. 5A and Supplementary Fig. S7A). Contrarily, tumor-infiltrating CD45⁺ myeloid cells were significantly increased in the tumors of high-dose PLX3397-treated mice compared with those of PBS-treated tumors (Fig. 5A and Supplementary Fig. S7A). Under the same experimental conditions, we observed significantly decreased CD45⁺CD11b⁺ macrophages in the PLX3397-treated tumors (Fig. 5A and Supplementary Fig. S7A). In a comparison of M1- and M2-like phenotypes between treatment groups, the accumulation of both CD45⁺CD11b⁺CD206⁺ M2-like phenotype and CD45⁺CD11b⁺CD80⁺ M1-like phenotype was decreased with the PLX3397 treatment (Fig. 5A and B and Supplementary Fig. S7A). However, the MFI of CD206⁺/CD80⁺ was significantly decreased in the high-dose PLX3397-treated tumors (Fig. 5A), indicating the systemic administration of 10 mg/kg of PLX3397 resulted in a polarization of TAMs against the M2-like phenotype. High-dose PLX3397 systemic treatment also resulted in a polarization against M2 direction in normal tissues such as liver and lymph nodes (Supplementary Fig. S6B and 6C).

CSF1R inhibition alters the TME composition of other immune cells (3, 4). We also investigated the comparative levels of other immune cell types in osteosarcoma tumors by flow cytometry. Although the accumulation of CD45⁺CD3⁺ T lymphocytes remained the same between treated versus the control (Fig. 5C and Supplementary Fig. S7B), there was an increased subpopulation of CD8⁺ T cells in high-dose PLX3397-treated tumors (Fig. 5B and C). Yet, the accumulation of CD4⁺ T cells remained unaltered (Fig. 5C and Supplementary Fig. S7B). Similarly, we also observed the unchanged quantity of CD45⁺CD19⁺ B cells among the treatment groups (Fig. 5C and Supplementary Fig. S7B.) Microscopically, there were fewer CD68⁺ macrophages (Fig. 5E and Supplementary Fig. S7B) in the high-dose PLX3397-treated tumors. Conversely, we observed increased CD8⁺ T cells in the high-dose PLX3397-treated tumors (Fig. 5E and Supplementary Fig. S7B). Additionally, the number of FOXP3⁺ regulatory T cells was significantly decreased in the high-dose PLX3397-treated tumors (Fig. 5D,–E, and Supplementary Fig. S7B). These findings in all of the above cell types were observed in the osteosarcoma microenvironment at both primary (Fig. 5E) and metastatic sites (Supplementary Fig. S7B). Collectively, systemic administration of PLX3397 reduced the infiltrating TAMs and FOXP3⁺ regulatory T cells and increased the CD8⁺ T-cell infiltration in the TME of osteosarcoma.

We describe the preclinical evidence of the efficacy of PLX3397, a potent CSF1R inhibitor, in a sarcoma model. PLX3397 inhibited M2 polarization, cellular proliferation, and chemotaxis of macrophages in our established in vitro sarcoma TAM models. The antitumor effect of PLX3397 was confirmed in the preclinical model using osteosarcoma-bearing mice. A recent report by Smeester and colleagues described antitumor effects of PLX3397 for osteosarcoma, but their investigations focused on the effect on tumor cells (31); thus, the effect of PLX3397 on the TAMs/immune cells in the osteosarcoma microenvironment has been unclear. We found markedly high CSF1R expression in sarcoma TAMs compared with tumor cells of osteosarcoma and fibrosarcoma, indicating the dominant effect of PLX3397 on TAMs. Our preclinical findings are the first delineation of the role of a small molecule (PLX3397) in modulating the sarcoma microenvironment of primary and metastatic sites and retarding osteosarcoma progression.

PLX3397 is a small molecule designed to block the CSF1R with restricted kinome targets approved for the treatment of adult patients with symptomatic tenosynovial giant cell tumor (17, 18). Preclinical tests demonstrated the antitumor effect of PLX3397 in mice with lung, breast, and prostate cancer, melanoma, and gastrointestinal stromal tumor (GIST; refs. 14, 15). PLX3397 is also being investigated in clinical trials, either as monotherapy or in combination with other drugs, for glioblastoma, melanoma, and metastatic breast cancer (18). Furthermore, PLX3397 enhanced the efficacy of chemotherapy (32, 33) or radiotherapy (15, 34). The interaction of this therapy with the immune system is suggested in a phase I dose-escalation study of PLX3397, among various advanced cancer histologies in which markedly reduced subsets of circulating monocytes (CD14^dim/CD16⁺) were found (35). Preliminary data from a phase I study of combined cabiralizumab (anti–CSF1R) and anti–PD-1 mAbs reported responses in 4/31 (13%) of advanced pancreas cancer patients (36, 37). The prevalence of TAMs in osteosarcoma and correlation with poor outcome suggests that CSF1R inhibitors such as PLX3397 are promising agents for clinical development with immunomodulators in patients with this orphan disease.

We identified striking differences in cytokine secretion from distinct sarcoma cell lines, including CSF1, CCLs, IP-10, and CXCLs in vitro. These differences likely underlie the variations in response to PLX3397 among macrophage populations differentiated in the presence of TCM from these sarcoma cells. For example, BMDM chemotaxis in response to TCM/LM8 and TCM/NFSa was remarkably enhanced compared with pure CSF1. These results indicated cytokines/chemokines other than CSF1 in TCM/LM8 and TCM/NFSa can also promote macrophage recruitment. Evidence suggests CCL2, overexpressed in a wide range of tumors, is involved in the recruitment of macrophages and lack of CCL2 signaling reduces macrophage infiltration (38). CXCL-1 is also known to recruit macrophages into the TME (39). The presence of these chemokines may explain the residual BMDMs infiltrating through the membrane even in the presence of PLX3397 (Fig. 3E). Notably, studies suggest the chemotaxis of M1- and M2-like macrophages is fine-tuned by different chemokines in the process of inflammation associated with tumorigenesis (40). Regarding macrophage survival, our findings demonstrated the most striking effect of PLX3397 with the use of pure CSF1, followed by TCM-LM8 and TCM-NFSa (Fig. 3D). These differences in macrophage viability might also be attributed to the differential expression of cytokines included in the TCM/LM8 and TCM/NFSa. Evidence indicates CCL-2, CCL-3, and CCL-14 stimulate proliferation of TAMs and activate normal macrophage polarization and differentiation into TAMs (41), which partially explain the differences in expression of M1- and M2-related markers among BMDM-CMG, BMDM-LM8, and BMDM-NFSa. These data indicated the possible benefit of targeting chemokines in addition to CSF1R blockade in depleting TAMs for bone and soft-tissue sarcomas. IL34 secretion was not detected from LM8 and NFSa, indicating that CSF1/CSF1R signaling, rather than IL34/CSF1R signaling, plays a major role in producing sarcoma TAMs.

In addition to the depletion of TAMs in our osteosarcoma model, systemic PLX3397 resulted in decreased FOXP3⁺ regulatory T cells and increased CD8 T lymphocytes within the osteosarcoma microenvironment. A positive feedback loop between FOXP3⁺ regulatory T cells and M2-like macrophages was recently identified (42). M2-polarized macrophages in the tumor environment promote the differentiation of CD4⁺CD25⁻ T cells into regulatory T cells which, in turn, skew the differentiation of monocytes toward M2-like macrophages (42). Thus, the decreased FOXP3⁺ regulatory T cells in vivo can be explained by the depletion of M2-polarized macrophages. The finding regarding the number of CD8 T cells was consistent with recent findings in the preclinical testing for multiple carcinomas (3, 4, 43), in which CSF1R blockade led to an increased number of CD8 T cells with approximately 1.5- to 5.5-fold change. Similarly, another therapeutic approach targeting TAMs by inhibition of the myeloid PI3K isoform, which induces proinflammatory gene expression in TAMs, showed an antitumor effect in mouse models of carcinoma and melanoma, by increasing CD8 T-cell infiltration without directly affecting T-cell activation or cytotoxicity (44). Several mechanisms can be envisioned for the CD8 T-cell recruitment after macrophage depletion. First, a change in the adhesive environment, which CD8 T cells encounter in the TME by CSF1R inhibition, might explain the increased infiltration of CD8 T cells, because macrophages express numerous adhesion molecules for T cells (43). In inflammatory settings, for example, macrophages are shown to form conjugates with T cells (43). Second, a change of chemokine levels in the TME might explain increased CD8 T-cell migration. Evidence suggests increased levels of CCL2, CXCL9, and CXCL10 following CSF1R inhibition contribute to CD8 T-cell trafficking to tumors (4, 43, 45), although the mechanism underlying the increase of these inflammatory chemokines following the depletion of TAMs remains to be elucidated. Furthermore, recent evidence has shown CSF1R blockade improves checkpoint immunotherapy by enhancing CD8 and CD4 T-cell activities (4, 43, 46). Further preclinical and clinical testing of combination therapy CSF1R blockade and immune-checkpoint inhibitors could be an important step for the development of a novel therapeutic strategy for sarcomas.

In summary, we present the antitumor effect of a potent CSF1R kinase inhibitor, PLX3397, in our established preclinical model of sarcomas. CSF1R blockade in the in vitro TAM model resulted in reduced viability and chemotaxis of macrophages and polarization from M2-like to a more M1-like phenotype. Systemic administration of PLX3397 in osteosarcoma model mice suppressed primary tumor growth and spontaneous lung metastasis, improving metastasis-free survival. Additionally, the effects of PLX3397 were not limited to depleting TAMs but also decreased FOXP3⁺ regulatory T cells and increased CD8 T-cell migration and infiltration into tumors. Based on current evidence, PLX3397 would provide a promising immunotherapeutic approach to many cancer types encompassing carcinoma, melanoma, and sarcoma, encouraging the combined use with immune-checkpoint blockade therapy in addition to conventional chemotherapy and radiotherapy.

No disclosures were reported.

T. Fujiwara: Conceptualization, resources, data curation, software, formal analysis, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. M.A. Yakoub: Formal analysis and methodology. A. Chandler: Data curation, software, formal analysis, validation, investigation, and methodology. A.B. Christ: Conceptualization, data curation, software, investigation, and methodology. G. Yang: Resources, formal analysis, visualization, and project administration. O. Ouerfelli: Conceptualization, resources, supervision, methodology, writing–original draft, writing–review and editing. V.K. Rajasekhar: Conceptualization, data curation, formal analysis, validation, investigation, visualization, methodology, writing–review and editing. A. Yoshida: Resources, data curation, formal analysis, methodology, project administration, writing–review and editing. H. Kondo: Resources, data curation, formal analysis, methodology, project administration, writing–review and editing. T. Hata: Resources, data curation, formal analysis, methodology, project administration, writing–review and editing. H. Tazawa: Resources, data curation, formal analysis, methodology, project administration, writing–review and editing. Y. Dogan: Conceptualization, resources, software, investigation, and project administration. M.A.S. Moore: Data curation, software, formal analysis, and project administration. T. Fujiwara: Resources, data curation, formal analysis, methodology, project administration, writing–review and editing. T. Ozaki: Resources, data curation, software, formal analysis, funding acquisition, methodology, writing–original draft, project administration, writing–review and editing. E. Purdue: Resources, software, formal analysis, investigation, and methodology. J.H. Healey: Conceptualization, formal analysis, supervision, funding acquisition, writing–review and editing.

The authors thank Dr. Afsar Barlas and his laboratory members, the MSK Molecular Cytology Core Facility, and the MSK Flow Cytometry Core Facility. We also thank Dr. Umeshkumar K. Bhanot and his laboratory members, the MSK Pathology Core Lab, for the pathological evaluation of primary and metastatic tumors and organs in the therapeutic test using animals. We also thank Jessica Massler and Dagmar Schnau for their editorial assistance. This study was supported in part by the Major Family Fellowship, the Reindeer Run Research Fund, Nanotechnology Center, Memorial Sloan Kettering Cancer Center Project #302 (J.H. Healey); NCI P30 CA008748 Cancer Center Support Grant (Dr. O. Ouerfelli) and NCI R50 CA243895 (Dr. O. Ouerfelli); a grant-in-aid for overseas research fellowships from the Yasuda Medical Foundation (2018; T. Fujiwara); and a grant-in-aid for overseas research fellowships from the Japan Society for the Promotion of Science (201860336; T. Fujiwara).

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.