The mesothelioma and other neoplasia cell lines were established from pleural fluids of patients in our laboratory.11 All cell lines were maintained in RPMI-1640 medium (Gibco) supplemented with 2 mM L-glutamine, 100 IU/mL penicillin, 0.1 mg/mL streptomycin and 10% heat-inactivated fetal calf serum (FCS) (Gibco) and cultured at 37°C in a 5% CO₂ atmosphere. The primary peritoneal mesothelial cells, MES-F, were purchased from Tebu-bio biosciences and cultured according to the manufacturer’s recommendations. Meso 34 NanoLuc cells were obtained after transfection of Meso 34 cells with pNL2.1[Nluc/Hygro] (Promega). After 48 hours, selection was performed using hygromycine (Invitrogen) (125 µg/mL) for 2 weeks. Expression of NanoLuc was assessed by seeding cells at 5×10³ cells per well of white-walled 96-well plate (Corning). Twenty-four hours later, after a wash with phosphate-buffered saline (PBS), coelenterazine (3.5 µM) (Interchim) was added and the luminescence signal was recorded after 10 min for 1 s using a Mithras LB 940 microplate analyzer (Berthold Technologies).

MPM primary cell lines were established at “Functional Genomics of Solid Tumors” laboratory, Paris, from surgical resections, pleural biopsies, or malignant pleural fluids of confirmed MPM cases, obtained from several French hospitals with patient’s consents. Most of them were used in several previous studies showing their relevance to MPM primary tumors. Genetic alterations in key genes of mesothelial carcinogenesis (CDKN2A, BAP1, NF2, LATS2 and TP53) and C1/C2 subtypes of the molecular classification were determined in this MPM series.12 13 Normal mesothelial cells were cultured from surgical resection of blebs from patients with spontaneous pneumothoraxes. Gene expressions were determined using cultures at low-passage number.

PEs from patients with a suspected mesothelioma were aseptically collected by thoracocentesis at the Laënnec Hospital (St-Herblain, France) between 1998 and 2016. Samples were centrifuged at 1000×g in a Heraeus Multifuge for 20 min at +4°C and supernatants were aliquoted and stored at −80°C. Serum samples were also collected at the Laënnec Hospital, aliquoted and stored at −80°C. Diagnoses were established by both fluid cytology and immunohistochemical staining of pleural biopsies performed by the pathology department at Laënnec Hospital (St-Herblain, France) and then externally confirmed by Mesopath, the French panel of pathology experts for the diagnosis of mesothelioma. All recruited patients had received no prior anticancer therapy and gave signed informed consent. All the collected samples and the associated clinical information were registered in a database (DC-2011-1399) validated by the French ministry of research.

The collection of MPM tumor samples and normal pleura was previously described.14 Transcriptome microarray data were available from 63 tumor samples of the same collection15; ArrayExpress database: accession codes E-MTAB-6877.

IL-34 and M-CSF titrations were performed with the Human IL-34 DuoSet ELISA and the Human M-CSF Quantikine ELISA kit (both from R&D Systems), respectively, following the manufacturers’ recommendations. The supernatants of MCTS were aliquoted and stored at −80°C until use. Cytokines were measured using the LEGENDplex Human M1/M2 macrophage panel (BioLegend) according to the manufacturer’s recommendations.

Total RNA was isolated using the Nucleospin RNAII Kit according to the manufacturer’s protocol (Macherey-Nagel). One microgram of total RNA was reverse-transcribed using Moloney murine leukemia virus reverse transcriptase (Invitrogen). Real-time PCR (RT-PCR) was carried out using an Mx3005P thermocycler (Stratagene). PCR was performed using QuantiTect Primer Assays (Qiagen) and the RT² Real-Time SYBR-Green/ROX PCR Mastermix (Qiagen), according to the manufacturer’s instructions. The relative amount of the target RNA was determined using the MxPro software, by comparison with the corresponding standard curve for each sample performed in duplicate. Each transcript level was normalized by division with the expression values of the acidic ribosomal phosphoprotein P0 housekeeping gene (RPLP0), used as an internal standard.

For gene expression analysis on MPM primary cell lines and frozen tumor samples, total RNA was isolated using Trizol (Thermo Fisher Scientific) or AllPrep DNA/RNA/miRNA Universal kit (Qiagen) according to the manufacturer’s protocol. 1.5 µg of total RNA was reverse-transcribed using High-Capacity cDNA Reverse Transcription Kit (Thermo Fisher Scientific). RT-PCR was carried out using ABI Prism 7900HT Real-Time PCR System. PCR was performed using Taqman assays (Thermo Fisher Scientific) according to the manufacturer’s instructions. IL34 and CSF1 transcript levels were normalized by the mean of the expression values of the five housekeeping genes Ribosomal 18S, ACTB, CLTC, GAPDH and TBP (-ΔCt). The following Taqman assays have been used: IL34 (Hs01050926_m1), CSF1 (Hs00174164_m1), 18S (Hs03928990_g1), ACTB (Hs01060665_g1), CLTC (Hs00964504_m1), GAPDH (Hs02758991_g1) and TBP (Hs00427620_m1).

All RNAseqv2 samples from the The Cancer Genome Atlas (TCGA)-MESO dataset (n=87 patients) are available on the Broad’s Genome Data Analysis Center (http://gdac.broadinstitute.org/). Gene expressions as RNA-seq by expectation maximization values (RSEM values) were analyzed. Clinical data for these samples were downloaded from FireBrowse (http://firebrowse.org; version 2018_02_26 for MESO).

Meso 34 cells were mixed with or without monocytes from healthy donors obtained by elutriation (DTC Core Facility, Nantes Hospital)16 at a ratio of 2:1 in 96-well U bottom plates NUNCLON SPHERA (Thermo Fisher Scientific) and in a volume of 180 µL of complete culture medium. The plates were centrifuged 2 min at 800×g and incubated at 37°C in a 5% CO₂ atmosphere for 3 days.

Multicellular tumor spheroids (MCTSs), constituted at formation of 20×10³ cells, were fixed with 4% paraformaldehyde (Electron Microscopy Sciences) for 24 hours at room temperature (RT). After one wash in PBS, MCTSs were included in HistoGel (Microtech, Thermo Fisher Scientific). Then, immunohistochemical analysis was performed using standard techniques by the Cellular and Tissue Imaging Core Facility of Nantes University (MicroPICell). The anti-CD163 antibody (Invitrogen) was used at 1/100 and the anti-CD14 antibody (Abcam) was used at 0.5 µg/mL. The revelation was performed using Leica Bond Polymer Refine Detection (Leica). Pictures were obtained using a NanoZoomer 2.0HT (Hamamatsu).

MCTSs, constituted of 20×10³ cells, were collected, washed one time in PBS and fixed in paraformaldehyde 4% (Electron Microscopy Sciences) for 48 hours at RT. MCTSs were washed once with PBS and permeabilized for 24 hours with PBS containing 2% Triton X-100 at RT. This solution was removed and then MCTSs were incubated with a solution of PBS containing 1% BSA, 0.2% Triton X-100, Hoescht 5 µg/mL (Sigma-Aldrich) and anti-CD163-Alexa Fluor 647 (BD Biosciences) for 48 hours at 4°C. Two steps of washing were performed with PBS containing 3% NaCl and 0.2% Triton X-100 for 2 hours at RT. Finally, MCTSs were resuspended into a RapiClear solution (SunJin Lab) and observed with a confocal microscope (Nikon A1R Si).

MCTSs (n=24) were collected, washed once in PBS and incubated with Trypsin 0.05% EDTA (Gibco) for approximately 10 min. Each 3 min, trypsin was removed and diluted in the culture medium to preserve detached cells and new trypsin was added on the residual MCTS to optimize dissociation of MPM cells and macrophages. The cell suspension was centrifuged at 800×g for 60 s in an Eppendorf Minispin. Cells were washed with PBS and stained with an anti-CD14-PE (clone REA-599, Miltenyi Biotec), an anti-CCR2-BV605 (clone K036C2, Biolegend), an anti-HLA-DR-FITC (clone G46-6, BD Pharmingen) and an anti-CD163-alexa647 (clone GH/61, BD Pharmingen), in RPMI-1640 with 10% FCS for 30 min at 4°C. IgG1_Κ-FITC (clone MOPC-21, BD Pharmingen), IgG2a-BV605 (clone MOPC-173, Biolegend), REA control-PE (clone REA293, Miltenyi Biotec) and IgG1_Κ-alexa647 (clone MOPC-21, BD Pharmingen) isotypes were used as controls. Samples were washed twice and then resuspended in PBS. Sample acquisition was performed using an LSR-Fortessa X-20 cytometer (Becton Dickinson). Results were analyzed with the DIVA Software (Tree Star).

Meso 34 NanoLuc MCTSs, constituted of 5×10³ cells including 30% of monocytes, were obtained as described above. Then, 20×10³ cells of the previously described HLA-A*0201/MUC1(950–958)-specific CD8+ T cell clones were added.17 Cytotoxicity of the CD8 T cell clone toward Meso 34 NanoLuc cells was evaluated by measuring nanoluciferase activity released in MCTS supernatants following cell lysis as follows. After 24 hours, 45 µL of medium was collected and light emission at 480 nm was measured immediately after addition of 5 µL of coelenterazine at 30 µM using Mithras LB 940 microplate analyzer (Berthold Technologies).

The estimation of the abundance of immune cell populations infiltrating MPM was done by using Microenvironment Cell Population Counter (MCP-counter) software on the gene expression dataset.15–18 Comparisons were performed using parametric paired t-test or Kruskal-Wallis test followed by the Dunn’s post hoc test. Log-rank Mantel-Cox test was used for survival analyses. Correlations were evaluated using non-parametric Spearman test. All statistical analyses were performed using GraphPad Prism (Prism V.6 for Windows) except univariate and multivariate Cox regression analysis that was performed using R statistical software.

In order to evaluate the involvement of M-CSF and IL-34 in MPM, we measured the expression of these cytokines in our collection of PE from patients. This collection is constituted of 96 MPM, 105 other neoplasia and 26 benign pleural effusions (BPEs) (table 1).

First, we observed that all the effusion types tested contained detectable level of M-CSF (figure 1A). There was no significant difference in the expression of M-CSF between MPM, other neoplasia and BPE groups. There was also no significant difference between malignant and benign PE (see online supplementary figure S1A). When considering subgroups, sarcomatoid mesothelioma (SM) seemed to express more M-CSF than the others (see online supplementary figure S1B). Regarding IL-34, MPM and other neoplasia groups had a higher expression level compared with BPE. Expression of IL-34 was significantly higher in malignant PE compared with BPE (see online supplementary figure S1C). However, a strong dispersion in the values of IL-34 was observed inside groups and subgroups (figure 1B and see online supplementary figure S1C), particularly for malignant PE. Indeed, less than 50% of the samples were positive for IL-34. Interestingly, the SM subgroup presented 71.4% of positive samples, whereas the other subgroups presented less than 50% of positive samples (figure 1C). Figure 1D shows that, in MPM samples, M-CSF levels were significantly higher in IL-34 positive PE compared with IL-34 negative PE.

Then, we evaluated the prognostic value of M-CSF and IL-34 in PE of patients with MPM. Overall survival data were available for 74 patients and global median survival for patients with MPM was 349 days. Patients were separated in two groups according to the M-CSF mean values and according to positivity for IL-34. Patients with M-CSF levels above mean presented a significant lower survival than the others (238 days vs 387 days; p=0.0395) (figure 1E). Patients with IL-34 positive PE presented a significant lower survival than the others (256 days vs 387 days; p=0.0044) (figure 1F). A multivariate analysis was performed including histologic subtype, IL-34 and M-CSF as parameters. We found IL-34 score to be an independent prognostic factor in our MPM cohort (online supplementary figure S2), whereas histologic subtype and M-CSF did not reach significance (p=0.243 and p=0.218, respectively).

To complete our study, we analyzed the mRNA expressions of CSF1 and IL34 in a collection of MPM tumors (n=178) and normal pleura (n=26) biopsies. A slightly lower level of CSF1 gene expression was observed in MPM tumors compared with normal pleura (p=0.0456) but no difference in the IL34 gene expression (figure 2A,B, respectively). By performing a correlation study using transcriptomic data obtained on 63 samples of the previous MPM tumor biopsy collection, we observed that CSF1 expression was positively associated with the expression of its receptor CSF1R, whereas it was not the case with IL34 expression (see online supplementary figure S3A and B). Moreover, CSF1 expression was associated with the infiltration of the tumor by immune cells of the monocytic lineage determined by the MCP-Counter tool18 on transcriptomic data (figure 2C). More precisely, CSF1 expression was correlated with the expression of tumor-associated macrophage markers, particularly of the ‘M2-like’ macrophage subset, such as CD163 and CD14 (figure 2C), and with a tendency for IL10 (data not shown). These correlations were not observed with the expression of IL34 (figure 2D). These results were confirmed using data from the TCGA database and a positive correlation was observed between CSF1 expression and IL10 (see online supplementary figure S3C-D, table S1 and online supplementary figure S4).

In order to focus only on malignant cells, we measured the mRNA expression of CSF1 and IL34 in a collection of samples of primary MPM cells (n=69) and normal mesothelial cells (MCs; n=4). Only the expression of the IL34 gene was significantly higher in MPM cells compared with MC (figure 3A,B). In this collection of primary tumor cells, CSF1 and IL34 mRNA expressions were not associated with histologic subtypes of MPM (see online supplementary figure S5). Higher expression of IL34 was also observed in a collection of MPM cell lines (n=30), established from PE of patients with MPM, compared with primary mesothelial cells (n=7) or other neoplasia cell lines (n=7) also established from PE of patients (see online supplementary figure S6). Interestingly, we observed that overexpression of IL34 in primary MPM cells was strongly associated with genetic alterations, consisting mainly in biallelic deletions of the CDKN2A gene and weakly with mutations in the NF2 gene, whereas CSF1 expression was independent of the mutational status (figure 3C–F). No other significant association was found between IL34 or CSF1 expression and genetic alterations in BAP1, LATS2 or TP53 genes or with C1/C2 subtypes of the molecular classification.12

With the objective to be closer to the physiopathological situation, we developed models of cell culture in three dimensions, namely, multicellular tumor spheroids (MCTSs). Two MPM cell lines with spontaneous capacity to form MCTS were used. The characteristics of these cell lines are provided in online supplementary figure S7 and online supplementary table S2. Figure 4A shows that coculture of Meso 34 with monocytes led to the formation of Mo-MCTS that contain CD14+ and CD163+ cells. In MCTS constituted only of Meso 34 cells, there was neither CD14 nor CD163 labeling. Using confocal microscopy, we observed CD163-positive cells (pink) inside the Mo-MCTS after a transparisation procedure (figure 4B). In order to improve the characterization of macrophages obtained, we used flow cytometry. MCTSs were dissociated and cells were labeled with antibodies anti-CCR2, anti-HLA-DR, anti-CD14 and anti-CD163 to determine the M1-like or M2-like phenotype. Monocytes obtained by elutriation were CCR2 high, CD14 high, CD163 low and HLA-DR mild (online supplementary figure S8A‒D). As controls, monocytes from elutriation were incubated with GM-CSF to obtain M1-like macrophages, characterized by a CCR2 low, CD14 low, CD163 low and HLA-DR mild expression, or with M-CSF to obtain M2-like macrophages, characterized by a CCR2 low, CD14 high, CD163 high and HLA-DR mild expression (see online supplementary figure S8). In MCTS, myeloid cells expressed HLA-DR and CD14, whereas mesothelioma cells did not express HLA-DR and expressed low level of CD14 (see online supplementary figure S9A). Approximately 85% of the myeloid cells were CD14 high and CD163 high (figure 4C and online supplementary figure S9B), corresponding probably to M2-like macrophages, and approximately 15% of the myeloid cells were CD14 mild and CD163 low, corresponding probably to M1-like macrophages (figure 4C and online supplementary figure S9B). This suggests that in MCTS, monocytes differentiated mainly into ‘M2-like’ macrophages. To confirm this hypothesis, we measured the mRNA expression of MAFB, CD14, CD163 and IL10. Expressions of all these macrophage markers were higher in MCTS made of Meso 34+ monocytes than in MCTS containing only Meso 34 cells (figure 4D). No major change in the mRNA transcription of the modulators of the immune response PDL-1, PDL-2, GITRL and OX40L was observed in the presence of macrophages (see online supplementary figure S10). Similar results were obtained with Meso 13 (see online supplementary figure S9C and D, figure S11). Analysis of the cytokines secreted by MCTS in culture supernatants showed that in the presence of monocytes, levels of IL-6, IL1-RA, IL-10 and IP-10 (figure 4E), tumor necrosis factor alpha (TNFα) and IL-1β (see online supplementary figure S12) were higher than in condition with Meso 34 alone. This set of cytokines allowed us to discriminate M1 from M2 macrophages, as shown in online supplementary figure S12. The profile of cytokines expressed in MCTS supernatants was similar to the one of M2 macrophages: absence of IL-12, presence of high amount of IL-6, IP-10 and presence of IL-10. Altogether, these results suggest that the coculture of MPM cells with monocytes as MCTS led to the differentiation of monocytes into ‘M2-like’ macrophages.

We evaluated the involvement of the CSF-1R pathway in this process by adding a CSF-1R inhibitor, GW2580, during the formation of the MCTS. Figure 5A shows that treatment with GW2580 reduced the expression of macrophage markers MAFB, CD14, CD163 and IL10 by approximately 50%. These observations were associated with a decrease of the levels of IL-6, IL-1RA, IL-1β, TNFα, IL-10 and IP-10 in the culture supernatants of MCTS (figure 5B and see online supplementary figure S13).

In order to evaluate the immunological impact of the presence of macrophages in MCTS, we measured the cytotoxic activity of a tumor antigen-specific CD8⁺ T cell clone toward MPM cells. We used an HLA-A2*0201/MUC1(950–958)-specific T cell clone that recognizes the HLA-A2+MUC1+Meso 34 MPM cell line.17 The coculture of Meso 34 NanoLuc MCTS with the CD8⁺ T cell clone induced a release of luciferase activity in the culture supernatants of MCTS demonstrating the lysis of the MPM cells (figure 6A). This cytotoxicity was correlated with the number of specific CD8⁺ T cell used. In the presence of macrophages in the MCTS, the activity of the CD8⁺ T cell clone was reduced by approximately 60% (figure 6A). Inhibition of CSF-1R by GW2580 restored significantly the cytotoxicity of the CD8⁺ T cells from 40% up to 75% (figure 6B).