Depletion of Tumor-Associated Macrophages with a CSF-1R Kinase Inhibitor Enhances Antitumor Immunity and Survival Induced by DC Immunotherapy

With the implementation of checkpoint inhibition therapy for several malignancies, immunotherapy has become an effective strategy for some patients in the treatment of advanced cancer (1). Although a subgroup of advanced disease stage patients benefits and shows prolonged overall survival to these therapies, the majority of patients do not respond or eventually relapse. Efforts have been made to characterize the mechanisms behind immunotherapy efficacy, leading to the distinction among inflamed, immune-sensitive tumors, and non-inflamed tumors associated with immunologic exclusion or ignorance (2–4). Blocking antibodies to the programmed cell death protein-1 (PD-1) appear to be most effective in patients whose tumors have a preexisting immune cell infiltrate, which can be enhanced by immunotherapy (5). The next breakthrough in cancer immunotherapy will be finding ways to sensitize non-inflamed and resistant-inflamed tumors to therapy, thereby increasing response rates and survival in advanced stage cancer patients (2, 3).

The dismal prognosis of malignant mesothelioma has not improved in the last decades despite implementation of conventional cancer treatment modalities (6). Malignant mesothelioma is often characterized by a prominent stromal component, dominated by macrophages (7). These tumor associated macrophages (TAM) often display an “alternatively activated” (“M2”) immune inhibitory phenotype characterized by the production of interleukin 10 (IL10) and surface expression of CD206 and inhibitory molecules such as PD-L1 (8). Furthermore, TAMs have been shown to be critical regulators of angiogenesis and they are closely linked to local tumor outgrowth and pleural fluid-mediated immune suppression in malignant mesothelioma patients (9–11). Furthermore, a high M2-TAM to CD8⁺ T-cell ratio correlates with a poor prognosis in malignant mesothelioma patients (9, 12).

TAMs are dependent on macrophage-colony stimulating factor (M-CSF) for their survival, proliferation, and recruitment towards tumors (8, 13). In addition, M-CSF promotes an M2 phenotype (14). M-CSF in the tumor and circulation is correlated with poor survival in several solid tumor types (15). For these reasons, different approaches to inhibit the M-CSF/M-CSFR pathway have been designed in order to deplete TAMs (16, 17). These molecules can be subdivided into antibodies targeting M-CSF, the M-CSFR, and small-molecule tyrosine kinase (TK) inhibitors with variable specificities for related kinases downstream from the receptors c-kit and FLT3 (15). The advantages of TK inhibitors over antibody-mediated therapies include (i) their ability to affect both murine and human M-CSFR–signaling, improving translatability across species, (ii) inhibition of autocrine M-CSF/M-CSFR-signaling, and (iii) the absence of rebound monocytopoeiesis after cessation of therapy due to intact receptor-mediated internalization of M-CSF which is TK-independent (13). PLX3397 (pexidartinib) is a clinically tested inhibitor of M-CSF-receptor (M-CSFR) and c-kit tyrosine kinase. PLX3397 is safe and effective in reducing TAMs in several solid tumor types (17, 18). Whether PLX3397 is effective in depleting TAMs in mesothelioma, and possibly capable of restoring antitumor immunity in these tumors is currently unknown.

Therapies dependent on dendritic cells (DC) and T cells circumvent aberrant antigen presentation and the formation of ineffective immune responses in cancer, possibly explaining their efficacy over conventional tumor vaccines (19). We have previously demonstrated that vaccination is a safe and effective way to generate functional antitumor immunity and clinical responses in malignant mesothelioma patients (20, 21). We hypothesize that TAM-mediated immune suppression could limit DC-vaccination efficacy and that depletion of TAMs would improve response rates and response durability in malignant mesothelioma models.

Here, we show that PLX3397 is effective in depleting TAMs but that monotherapy does not improve survival in murine mesothelioma models. Combining M-CSFR inhibition with DC vaccination (DC therapy) to induce effective antitumor immune responses improved survival such that mice were protected from disease following tumor rechallenge. This therapeutic synergy may improve response rates and survival with immunotherapy.

Female 8–12-week-old BALB/c (H-2d) mice (Envigo) and CBA/J mice (Janvier) were housed under specific pathogen-free conditions at the animal care facility of the Erasmus MC, Rotterdam. Experiments were approved by the local Ethical Committee for Animal Welfare and complied with the Guidelines for the Welfare of Animals in Experimental Neoplasia by the United Kingdom Coordinating Committee on Cancer Research (UKCCCR) and by the Code of Practice of the Dutch Veterinarian Inspection. The AB1 cell and AC29 mesothelioma cell lines were kindly provided by Professor Bruce W.S. Robinson (Queen Elizabeth II Medical Centre, Nedlands, Australia). These cell lines were derived from tumors induced by crocidolite asbestos injected i.p. into CBA/J and BALB/c mice (22). Both cell lines were obtained in the years 2003–2005, aliquoted and stored at −190°C.

Tumor cells were cultured from early passages (max. 5 passages following cell line acquisition) in RPMI 1640 medium containing 25 mmol/L HEPES, Glutamax, 50 g/mL gentamicin, and 5% (v/v) fetal bovine serum (FBS) (all obtained from Gibco) in a humidified atmosphere and at 5% CO₂, in air. For culture, either culture flasks or CellSTACKs (Corning Life Sciences) were used to reach appropriate tumor cell frequencies for injection. AB1 and AC29 cells were passaged once or twice a week to a new flask by treatment with 0.05% trypsin, 0.53 mmol/L EDTA in phosphate buffered saline (PBS, all Gibco). At every 8–10 passages, cell lines were tested for mycoplasma contamination by PCR and remained negative. At the start of the experiment, CBA/J or BALB/c mice were i.p. injected with either 20 × 10⁶ AC29 cells or 0.5 × 10⁶ AB1 cells respectively, dissolved in PBS, or with PBS as control as was described previously (23). Mice were scored using the body condition score, killed if profoundly ill and scored as a death in survival analysis.

AB1 cell line-derived tumor lysate was prepared as previously described (23). Briefly, cells were suspended in PBS and frozen in liquid nitrogen and disrupted by four cycles of freeze-thawing followed by sonication for 3 × 10 seconds with an amplitude of 10 μm, using a Soniprep 150 ultrasonic disintegrator equipped with a microtip (Sanyo Gallenkamp BV) on ice. Cell lysate was aliquoted and stored at −80°C.

DCs were generated following an adapted protocol (Lutz and colleagues 1999) as previously described (23). Bone-marrow derived cells seeded in 100-mm Petri dishes (day 0) and cultured in 10 mL DC Culture Medium [DC-CM]: RPMI 1640 containing glutamax-I (Gibco) supplemented with 5% (v/v) FBS, 50 mol/L mercaptoethanol (Sigma-Aldrich), 50 g/mL gentamicin (Invitrogen), and 20 ng/mL recombinant murine granulocyte macrophage-colony-stimulating factor (GM-CSF, kindly provided by Prof. B. Lambrecht, VIB Ghent, Belgium). Cells were cultured in a humidified atmosphere at 5% CO₂ in air. At day 3, 10 mL of fresh DC-CM was added. On day 6, 10 mL of each plate was replaced with 10 mL of fresh DC-CM. After 9 days of culture, AB1 cell lysate was added to the DC cultures, to the equivalent of three AB1 cell-equivalents per DC. After 8 hours, 10 g/mL CpG (ISS-ODN 1668, Invitrogen) was added to the culture to allow complete maturation while incubated overnight. The next day, DCs were harvested by and purified by Lympholyte-Mammal (Cedarlane) density gradient centrifugation, the interphase was washed three times in PBS and resuspended at a concentration of viable cells. The quality of the DC preparation was determined by cell counting, morphologic analysis, and cell surface marker expression by flow cytometry, as previously described (23).

Mature, lysate pulsed-DCs were generated as demonstrated above and cultured for 48 hours in GM-CSF DC-TCM and varying concentrations of PLX3397 (provided by Plexxikon Inc. as part of a material transfer agreement; ref. 17) or the vehicle (Dimethyl sulfoxide, DMSO, Sigma-Aldrich) in 6-well plates. PLX3397 was reconstituted in DMSO to reach a stock concentration of 20 mmol/L, which was then diluted to a concentration range of 0.1 to 10 μmol/L in DC-TCM. After 48 hours, cells were harvested from the 6-well plates and analyzed for viability and surface expression of immune markers using multicolor flow cytometry.

On day 0, BALB/c or CBA/J mice were inoculated intraperitoneally with AB1 or AC29 tumor cells, respectively. On day 10, mice were treated with either 2–3 × 10⁶ DCs dissolved in 200 μL PBS, or PBS alone. This time point was determined to be most optimal in previous experiments (23). In specified cases, treatment with either DC therapy alone or in combination with PLX3397 commenced at day 15 to allow for enough tumor to be collected several days later and prepare a window to detect possible changes between mono- and combination-treated tumors. Also, depending on the treatment arm, mice were fed ad libitum PLX3397-containing chow or control chow of equal nutritional value and consistency until the end of the experiment or prior to rechallenge. The optimal dose to effectively deplete TAMs was determined by Plexxikon and has been further corroborated in different tumor models to be 290 mg PLX3397/kg chow (delivering daily doses of approximately 45 mg/kg; ref. 24). On day 15 post-tumor cell injection, prior and after rechallenge, blood was obtained via tail vein extraction. During the remainder of the follow-up period, mice were examined daily for evidence of illness caused by overt tumor growth.

Tumor biopsies were embedded in Tissue-Tek II optimum cutting temperature medium (Miles), snap-frozen, and stored at −80°C. Tissue sections (6 μm) were cut on a cryostat (Cryostar NX70, Thermo Fisher Scientific). Frozen sections were warmed to room temperature, fixed with acetone for 10 minutes and rinsed in PBS. Sections were incubated in peroxidase blocking solution (0.1% H2O2 and 0.1% sodium azide in PBS) for 30 minutes and rinsed with PBS. Slides were placed in a semi-automatic stainer (Sequenza) and incubated in 1:10 diluted normal Goat serum (CLB) for 10 minutes and subsequently for 60 minutes with the diluted primary Abs, followed by rinsing in PBS for 5 minutes and incubation for 30 minutes with diluted secondary Abs. Double-immunostaining was carried out using antibodies supplied in Supplementary Table S1. Binding of antibody was detected using alkaline phosphatase- (AP-) or peroxidase-conjugated goat anti-rat (Sigma-Aldrich) and Naphtol-AS-MX-phosphate (0.30 mg/mL; Sigma-Aldrich) + new fuchsine (160 mg/mL in 2 mol/L HCl; Chroma-Gesellschaft) or AEC [0.1 mol/L NaAc + 1% AEC stock (100 mg AEC in 10 mL DMF), Vectorlabs], respectively, were used as substrate. The specificity was checked using a protein concentration-matched non-relevant rat antibody and PBS. Finally, the sections were rinsed in distilled water and mounted in vecta mount (Vector). Slides were scanned using a Nanozoomer 2.0 HT (Hamamatsu).

Scanned IHC slides were viewed using NDP-viewing software (Hamamatsu) at 20× and 40× magnification and regions of interest were captured and imported into ImageJ software (NIH). Colors were separated, thresholds were installed to select for positive cells and these were depicted as percentage of total area. For each sample, 5 random areas (including tumor rim and center) at 20× magnification were selected, analyzed and averaged. CD8-positive cells were well demarcated and counted (average of 5 random tumor areas per sample) to be expressed as cells/mm².

Single-cell suspensions were generated from the spleens, blood, and tumors of mice from each group. All tissues were either weighed in a microbalance in case of tumors and spleens, or volume determined for blood. Briefly, spleens were aseptically removed and mechanically dispersed over 100-μm nylon mesh cell strainer (BD Biosciences) followed by erythrocyte lysis using osmotic lysis buffer (8.3% NH₄Cl, 1% KHCO₃, and 0.04% Na₂EDTA in Milli-Q). Blood was collected in EDTA tubes (Microvette CB300, Sarstedt) and subsequently lysed. Tumors were collected and dissociated using a validated tumor dissociation system (Miltenyi Biotec). Cell suspensions were filtered through a 100-μm nylon mesh cell strainer (BD Biosciences) and counted in trypane blue with a hemocytometer using the Burker–Turk method.

For measurements of cytokine production in lymphoid cells by flow cytometry, cells were restimulated for 4 hours at 37°C using PMA and ionomycin supplemented with GolgiStop (BD Biosciences). For assessing cytokine production by myeloid cells, cells were subjected to 4 hours incubation with Golgistop. For cell surface marker staining, cells were washed with FACS-wash (0.05% NaN₃, 2% BSA in PBS) and Fc II/III receptor blocking was performed using anti-mouse 2.4G2 antibody (1:300; kindly provided by L. Boon, Bioceros, Utrecht, the Netherlands). After the blocking procedure, antibodies (Supplementary Table S1) for cell surface staining were added into each sample and placed on ice for 30 minutes. Cells were washed in FACS-wash followed by a PBS wash, and then stained for viability using fixable LIVE/DEAD aqua cell stain (Thermo Fisher Scientific, 1:200). After two additional washes with FACS-wash, cells were either measured or in case of intracellular staining; fixated, permeabilized, and stained using Fix/Perm buffer (in case of nuclear protein staining, eBioscience) or 4% PFA and 0.5% saponin (in case of cytokine stainings, Sigma-Aldrich). Antibodies were stained for 30 minutes in case of the PFA/Saponin protocol and 60 minutes for the intranuclear staining protocol, on ice in the dark. A fixed number of counting beads (Polysciences Inc.) was added prior to data acquisition to determine the absolute amount of cells. Data were acquired using an LSR II flow cytometer (BD Biosciences) equipped with three lasers and FACSDiva software (BD Biosciences) and analyzed by FlowJo (Tree Star, Inc.) software V10.1.

In order to detect tumor-antigen specific CD8⁺ T cells in the CBA/J model, dextramers specific for potentially dominant WT1 and Mesothelin epitopes were constructed. Peptides for WT1 and Mesothelin were selected based on freely available tools (IEDB Analysis Resource) predicting likelihood of processing, MHC-I (H2-Kk) binding affinity and immunogenicity of the chosen peptides. The peptide sequences for WT1 and Mesothelin selected were SENHTAPIL, and QEATLLHAV, respectively. Dextramers were labeled to APC or PE fluorochromes and ordered from Immudex. Dextramers were stained independently from one another following company instructions on cells pretreated in RPMI containing 50 nmol/L Dasatinib for 30 minutes at 37°C to limit TCR internalization. FMO and dextramer binding-CD4⁺ T cells were used to determine background signal and set cutoff limits.

Biotin-labeled cRNA derived from AB1 and AC29 (n = 3 samples) was hybridized to the Mouse Genome 430 2.0 Array according to the manufacturer's instructions (Affymetrix); data were analyzed with BRB-ArrayTools (version3.7.0, National Cancer Institute) using Affymetrix CEL files obtained from GCOS (Affymetrix To examine the quality of the various arrays, the R package affyQCreport for generating QC reports was run starting from the CEL files. All created plots indicated a high quality of samples and an overall comparability. Raw intensity values of all samples were normalized by RMA normalization (Robust Multichip Analysis; background correction and quantile normalization) using Partek version 6.4 (Partek Inc.). Genes involved in myeloid cell chemotaxis and survival, and expression of clinically relevant tumor antigens were manually selected and plotted for both AB1 and AC29 cell lines. The GEO accession number for this microarray is GSE97150.

Data are expressed as medians with interquartile range. Comparisons between groups were made using the Mann–Whitney U test for independent samples, or the Wilcoxon signed rank test in case of paired samples. When correlations were depicted, Spearman's rank correlation test was performed to test for statistical significance. A two-tailed value of P < 0.05 was considered statistically significant. Survival data were plotted as Kaplan–Meier survival curves, using the log-rank test to determine statistical significance. Data were analyzed using GraphPad Prism software (Graphpad, V5.01)

In the past, we have shown that applying DC therapy in syngeneic and orthotopic mesothelioma mouse models is effective for assessing antitumor T-cell responses and evaluating treatment efficacy (23, 25). The AB1 and AC29 mesothelioma cell lines are injected intraperitoneally (i.p.) in BALB/c and CBA/J mice, respectively. Compared with the BALB/c model, the CBA/J model has a more pronounced TAM-dependent phenotype (Supplementary Fig. S1A), which may be explained by the increased expression of M-CSF and CCL-2, being key TAM-homing and survival factors as determined by microarray (accession number: GSE97150, Supplementary Fig. S1B). In line with this M2-TAM–dominant phenotype, CBA/J mice injected with AC29 mesothelioma cells develop ascites, paralleling disease heterogeneity also seen in mesothelioma patients (22, 26, 27).

Here, we used the M-CSFR–inhibitor PLX3397 to target TAMs in vivo. To confirm that there was no direct effect of PLX3397 on the murine mesothelioma cell lines, we analyzed expression of M-CSFR and c-kit using flow cytometry and tissue microarray analysis. We found negligible levels of both molecules (Supplementary Fig. S1B). To exclude any direct effects of PLX3397 on maturated tumor-lysate pulsed DCs to be administered in the DC therapy protocol, we cultured these cells in vitro with increasing concentrations of PLX3397 or vehicle alone, then measured cell viability and expression of relevant surface markers by flow cytometry (Supplementary Fig. S2). As previously reported, GM-CSF–cultured cells are composed of heterogeneous populations of DCs and macrophages, which we also identified in our system (Supplementary Fig. S2; ref. 28). Effects of PLX3397 on DC viability or surface expression of MHC-II, CD86, PD-L1, M-CSFR, and c-kit were negligible (Supplementary Fig. S2A). This resistance of DCs to PLX3397 could be explained by the rapid downregulation of M-CSFR surface expression following maturation using unmethylated CpG (Supplementary Fig. S2B-S2C).

Mice were left untreated or were treated at day 10, after solid tumors had established, using three different regimens: (i) PLX3397 was administered in chow and continued for the duration of the experiment, (ii) DCs were injected i.p. as monotherapy, or (iii) mice were subjected to double therapy (Fig. 1A and B). PLX3397 monotherapy failed to improve survival in both mesothelioma tumor models (Fig. 1C and D). DC therapy prolonged survival only in the AC29 tumor model. Combination therapy, however, enhanced survival in both tumor models, indicating therapeutic synergy (Fig. 1C and D).

To assess whether PLX3397 was effective in depleting TAMs, we focused on the CBA/J mesothelioma model, because this is more TAM-dependent. IHC analysis showed a decrease in F4/80⁺ TAMs, CD31⁺ endothelial cells, and CD206⁺ (M2) cells in PLX3397-treated mice compared with control mice (Fig. 2A–G). TAMs are crucial for tumor angiogenesis and concomitant ascites production in metastatic ovarian carcinoma (29). Indeed, we found a correlation between CD31⁺ endothelial and TAM areas in these tumors and a decrease in ascites volume in the PLX3397-treated mice (Fig. 2D and E). Tumors of untreated mice contained a CD8⁺ T-cell infiltrate which was not further enhanced following TAM depletion (Fig. 2H).

Taken together, these findings show that M-CSFR inhibition effectively reduced TAMs, neo-angiogenesis, and ascites in mesothelioma mouse models, but did not improve local CD8⁺ T-cell infiltration.

We next sought to investigate the potential mechanisms that lead to enhanced survival in DC therapy only and combination immunotherapy arms in the CBA/J model. To this end, we extracted blood 5 days after start of treatment and analyzed the circulating immune compartment using flow cytometry (Supplementary Fig. S3). DC therapy caused an increase in CD8⁺ T-cell numbers with CD4⁺ Foxp3⁺ T-helper–cell numbers remaining unchanged, whereas proliferation, as determined by Ki-67 expression, was increased in both T-cell subsets (Fig. 3A–D; Supplementary Fig. S4A). Numbers of T-regulatory cells (Treg) were decreased following DC therapy, a decline amplified by the addition of PLX3397 treatment, resulting in an improved CD8⁺ T-cell/Treg-ratio (Fig. 3C; Supplementary Fig. S4B). DC therapy, but not the addition of PLX339, increased the frequency of short-lived effector cells (SLEC), as determined by KLRG1 expression, from almost undetectable to approximately 20% of total CD8⁺ T cells (Fig. 3E). In addition to proliferation, an increased proportion of CD8⁺ T cells in the combination therapy–arm coexpressed CD4, which has previously been associated with increased CD8⁺ T-cell effector functions (Fig. 3D and F; ref. 30, 31). These CD4/CD8 double-positive cells were also highest in proliferation in all treatment arms as demonstrated by increased proportions of Ki-67⁺ cells (Supplementary Fig. S4C–S4D).

Whereas DC therapy primarily influenced lymphocyte dynamics and phenotype, PLX3397 therapy predominantly affected myeloid subsets (Fig. 4). Granulocyte and total monocyte numbers were increased by DC therapy but only granulocyte numbers were increased by PLX3397 as mono- or combination therapy (Fig. 4A and B). Monocytes can be subdivided into classical Ly6Chi- and nonclassical Ly6Clow monocytes with each subset having different functions and migration patterns in blood (32, 33). Dissecting monocyte subsets in our model revealed nearly complete depletion of nonclassical (Ly6C^low) monocytes in the PLX3397-treated arms (from ∼40 to 5–10 cells/μL blood), whereas classical (ly6C^hi) monocytes numbers were slightly increased from ∼50 to 75 cells/μL (Fig. 4C and D). Approximately 90% of nonclassical monocytes expressed PD-L1 in untreated conditions which was decreased to 20% to 40% of monocytes by M-CSFR-inhibition (Fig. 4F). Classical monocytes were rarely positive for PD-L1 (1%–5%), which was similar in DC therapy and combination therapy-treated mice (Fig. 4E).

Overall, the observed synergy between therapies was illustrated in the CBA/J model by an improved CD8⁺ T-cell phenotype primarily in response to DC therapy, and a decrease in PD-L1⁺ nonclassical monocytes due to PLX3397 treatment. Similar patterns in blood immune cell dynamics could be discerned in the AB1 tumor model; however, numbers of mice were limited at day 5 after start of treatment (Supplementary Fig. S5).

To examine the tumor microenvironment (TME) for the effects of treatment and to relate these to changes in immune cell composition observed in spleen and blood, we sacrificed mice at day 15 in our CBA/J model. Five days after therapy began, we observed a similar tumor burden between the different treatment groups (Fig. 5A). TME immune composition at day 15, however, differed between treatment groups in line with interim analyses in blood and findings in end stage disease tumors of diseased mice (Supplementary Fig. S6A). Both IL-10⁺ (M2) and IL10⁻ (M1) TAMs were diminished by PLX3397 as monotherapy or in combination with DC therapy (Fig. 5B). We also further characterized peripheral blood nonclassical monocytes as being higher in IL10 and PD-L1 expression and lower in surface MHC-II expression, further establishing their immune suppressive phenotype (Supplementary Fig. S6B).

Confirming earlier results, the numbers of CD8⁺ tumor-infiltrating lymphocytes (TIL) were not altered following PLX3397 monotherapy, and these TILs displayed an exhausted phenotype (34), defined as PD-1 positive, lymphocyte-activation gene 3 (LAG-3) positive, and interferonγ (IFNγ) negative (Fig. 5C). DC therapy, in contrast, enhanced the number of non-exhausted CD8⁺ T cells in the tumor. This did not differ between DC therapy only and combination therapy-treated mice (Fig. 5C). To investigate whether this was a kinetics-related phenomenon, we treated tumor-bearing mice on day 15 and analyzed the tumor 8 days later to allow for proper T-cell infiltration. In addition, we sought to identify T-cell clones specific for the clinically meaningful tumor antigens Mesothelin and Wilms Tumor 1 (WT1) to assess whether CD8⁺ T cells with these tumor specificities were enhanced by DC therapy and possibly enriched or improved in phenotype by coadministration of PLX3397. Both tumor antigens were highly expressed by the AC29 mesothelioma cell line (Supplementary Fig. S1B). T cells specific for these antigens were present in tumors from mice in all treatment groups, with an increase in WT1-specific but not Mesothelin-specific CD8⁺ T cells in DC therapy-treated mice (Fig. 5D). There were prominent differences in CD8⁺ T-cell phenotype in combination therapy-treated mice compared with those receiving DC therapy only, as determined by a higher proportion of PD-1–negative and –intermediate expressing cells (Supplementary Fig. S7). These cells are capable of eliciting potent antitumor functions (35–37). In line with this less-exhausted phenotype is the lower expression of coinhibitory molecules TIM3 and LAG3, and lower expression of CD8 indicative of enhanced activation (Supplementary Fig. S7B; ref. 38). Similar shifts in PD-1–expression were detected in both the Mesothelin- and WT1-specific T cells (Fig. 5E; Supplementary Fig. S7D).

In summary, we observed a PLX3397-dependent reduction in TAM frequency with a concurrent DC therapy dependent increase in tumor infiltrating CD8⁺ T cells whose phenotype was improved upon TAM-depletion as shown by a decrease in proportion of PD-1-high CD8⁺ T cells associated with T-cell exhaustion.

To investigate whether the pro-inflammatory phenotype present at day 5 following therapy persisted in treated mice, blood was extracted 3 months later (day 107) from surviving mice (treated with DC therapy, either as a monotherapy or in combination with PLX3397). At this time point, nonclassical monocyte frequencies were low but this was less pronounced as on day 15 (Fig. 6A). The percentage of PD-L1–positive nonclassical monocytes, however, was decreased and was negligible on classical monocytes (Fig. 6B). Prior to rechallenge, CD8⁺ T-cell numbers were increased in treated mice compared with untreated tumor naïve mice, but proliferation and effector phenotype had returned to a resting state (Fig. 6C and D).

To test whether surviving CBA/J mice (Fig. 1D) were free of tumor and protected against a second tumor encounter, they were rechallenged with mesothelioma in parallel with age-matched tumor-naïve mice (Fig. 6E). Three days prior to rechallenge, PLX3397 treatment was terminated and blood was again extracted 8 days after rechallenge and recall responses were evaluated. After rechallenge, all tumor naïve mice reached their humane endpoint due to high tumor burden, whereas DC therapy-treated and combination therapy-treated mice remained disease free (Fig. 6F). CD8⁺ T-cell numbers were higher in protected mice after challenge compared with control mice but similar in both treatment groups, whereas CD4⁺ T-helper cells were lower in the combination therapy-treated mice (Fig. 6G). CD8⁺ and CD4⁺ T-cell proliferation was higher in DC+PLX3397-treated mice compared with DC-only mice, indicating superior memory responses after dual immunotherapy (Fig. 6H).

Taken together, these findings demonstrate that targeting the TME and simultaneously directing the immune system to combat cancer can lead to durable responses in mesothelioma-bearing mice and improved recall responses.

We have shown that combination therapy using two safe anticancer treatment strategies being DC therapy and M-CSFR inhibition through the small molecule inhibitor PLX3397, results in improved overall survival and superior antitumor immune reactivity in mesothelioma mouse models. TAM-depletion in itself was insufficient to improve antitumor T-cell responses. DC-monotherapy improved survival in mice although most mice still progressed after treatment, paralleling findings in patients who develop resistance to immunotherapy (20, 21, 39). Only when PLX3397 treatment was combined with DC vaccination did we observe that the decrease in TAMs was associated with an increase in CD8⁺ T-cell numbers and functionality, both in the TME and in the circulation. In particular, our data show that CD8⁺ T cells, including WT1- and Mesothelin-specific CD8⁺ T cells, in the TME of combination therapy-treated mice expressed lower levels of key inhibitory molecules PD-1, LAG3, and TIM3, associated with exhaustion, on their cell surface. Our findings are in line with previous reports in malignant mesothelioma patients where a high ratio of CD8⁺ T cells to M2 TAM was related to better survival (9, 12). Combining DC therapy with M-CSFR–blockade may be a safe and effective strategy to increase this clinically relevant ratio and improve disease prognosis.

TAMs are a major leukocyte subset in the stroma of mesothelioma and other cancers and capable of potently suppressing endogenous or treatment-induced antitumor T-cell responses (8, 11). Depletion of this immune suppressive cell type may, thus, render previously immune resistant or escaped tumors sensitive to checkpoint inhibition (40), chemo- and radiotherapy (41, 42), and cellular therapies such as adoptive T-cell transfer (43) and now also DC therapy. Use of chemo(radio)therapy or PD-1 checkpoint inhibitors either come with burdensome side effects or are most effective in boosting a preexisting antitumor T-cell response (3, 4). DC therapy effectively induces novel antitumor immune responses which can then be sustained by targeting the generally immune suppressive TME (23, 44).

M-CSFR inhibition depletes TAMs in multiple tumor models but its effects on survival as monotherapy or combined with other therapies are variable (16, 18, 41, 45). For example, when M-CSFR inhibition was combined with a tumor vaccine or with certain chemotherapeutic drugs, efficacy of these therapies was abrogated, questioning the role of TAMs in tumor behavior and response to these therapies (16, 45, 46). On the other hand, glioblastoma (GBM)-bearing mice treated with the M-CSFR–inhibitor BLZ945 as monotherapy showed improved survival (16). Although resistance to monotherapy ensued in the majority of mice, tumors regressed due to skewing of M2 to M1 TAMs rather than TAM depletion. In our models, M-CSFR inhibitor monotherapy alone did not prolong survival despite gross changes in stromal composition. These differences in drug efficacy may be explained by TAM-dependency of certain therapies, mechanism of M-CSFR inhibition, or timing of therapy and tumor stage and type (46). In the GBM model, for example, TAM skewing was dependent on glioblastoma cell-derived factors including granulocyte-macrophage colony stimulating factor (GM-CSF) (16). GM-CSF is often used to culture pro-inflammatory M1 macrophages and DCs whereas M-CSF cultured macrophages display a M2 phenotype (14, 28, 47). Other tumor types such as mesothelioma, both in patient and in mice, produce considerable amounts of M-CSF but much less GM-CSF (Supplementary Fig. S1; ref. 48), providing a likely explanation for the depletion rather than skewing of TAMs in our models.

Although CD8⁺ TILs were present in mesothelioma tumors of untreated mice, they did not increase in frequency or functionality following TAM-depletion and maintained an exhausted, dysfunctional phenotype. After induction of an immune response using DC therapy in late stage disease, functional TIL frequencies increased. Survival was unaltered in BALB/c mice, and, although prolonged in the CBA/J model, those mice showed disease progression or recurrence during follow-up. Tumor responses were only durable in the majority of mice when DC therapy was applied in a TAM-deficient TME. We could detect this synergy between therapies in T-cell proliferation and phenotype in blood during therapy and after rechallenge, and in the tumor where T-cell phenotype was improved following TAM depletion. Which immune suppressive property of TAMs is key to induction of T-cell exhaustion in our model remains to be full elucidated, but most likely involves multiple factors (e.g., IL10, arginase, PD-L1; ref. 8, 49).

Uncertainty remains about the ontogeny of TAMs in solid tumors. We appreciate that classical (Ly6Chi) monocytes are the main precursors of inflammatory DCs and macrophages including TAMs, but note incongruence regarding the sequence of events that precede M2-TAM development (50, 51). Classical monocytes are obligatory precursors of nonclassical monocytes; the functions and migration patterns of both cell types differ, with nonclassical monocytes being crucial for endothelial and tissue integrity (32, 33). Nonclassical monocytes are particularly dependent on M-CSF for their survival (13, 52). Nonclassical monocytes and TAMs share a similar immune suppressive phenotype and are most sensitive to M-CSFR inhibition, suggesting a relationship between these cell types. Future research will have to further delineate the contribution of nonclassical monocytes to TAMs in cancer.

In conclusion, we have demonstrated that therapeutic efficacy of M-CSFR inhibition as monotherapy is limited in mesothelioma. However, TAM depletion combined with an effective DC-mediated antitumor T-cell response is capable of producing durable tumor responses and functional antitumor immunity.

J.G. Aerts reports receiving commercial research grant from Amphera, has ownership interest (including patents) in dendritic cell immunotherapy with an allogenic lysate, and is a consultant/advisory board member for AstraZeneca, Bristol-Myers Squibb, Boehringer Ingelheim, Eli Lilly and Company, MSD, and Roche. No potential conflicts of interest were disclosed by the other authors.

Conception and design: F. Dammeijer, L.A. Lievense, K. Bezemer, R.W. Hendriks, J.G. Aerts

Development of methodology: F. Dammeijer, L.A. Lievense, K. Bezemer, J.P. Hegmans, J.G. Aerts

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): F. Dammeijer, L.A. Lievense, J.G. Aerts

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): F. Dammeijer, L.A. Lievense, J.G. Aerts

Writing, review, and/or revision of the manuscript: F. Dammeijer, L.A. Lievense, T. van Hall, R.W. Hendriks, J.G. Aerts

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): F. Dammeijer, M.E. Kaijen-Lambers, M. van Nimwegen, K. Bezemer, J.P. Hegmans

Study supervision: R.W. Hendriks, J.G. Aerts

The authors would like to acknowledge Plexxikon for distributing to us PLX3397 as part of a material transfer agreement. We thank R. Bouzid and the Erasmus MC Animal Facility staff for their support with the animal experiments.

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.