Bromodomain inhibition exerts its therapeutic potential in malignant pleural mesothelioma by promoting immunogenic cell death and changing the tumor immune-environment

ABSTRACT Systemic treatment of malignant pleural mesothelioma (MPM) is moderately active for the intrinsic pharmacological resistance of MPM cell and its ability to induce an immune suppressive environment. Here we showed that the expression of bromodomain (BRD) proteins BRD2, BRD4 and BRD9 was significantly higher in human primary MPM cells compared to normal mesothelial cells (HMC). Nanomolar concentrations of bromodomain inhibitors (BBIs) JQ1 or OTX015 impaired patient-derived MPM cell proliferation and induced cell-cycle arrest without affecting apoptosis. Importantly, BBIs primed MPM cells for immunogenic cell death, by increasing extracellular release of ATP and HMGB1, and by promoting membrane exposure of calreticulin and ERp57. Accordingly, BBIs activated dendritic cell (DC)-mediated phagocytosis and expansion of CD8+ T-lymphocyte clones endorsed with antitumor cytotoxic activity. BBIs reduced the expression of the immune checkpoint ligand PD-L1 in MPM cells; while both CD8+ and CD4+ T-lymphocytes co-cultured with JQ1-treated MPM cells decreased PD-1 expression, suggesting a disruption of the immune-suppressive PD-L1/PD-1 axis. Additionally, BBIs reduced the expansion of myeloid-derived suppressor cells (MDSC) induced by MPM cells. Finally, a preclinical model of MPM confirmed that the anti-tumor efficacy of JQ1 was largely due to its ability to restore an immune-active environment, by increasing intra-tumor DC and CD8+ T-lymphocytes, and decreasing MDSC. Thereby, we propose that, among novel drugs, BBIs should be investigated for MPM treatment for their combined activity on both tumor cells and surrounding immune-environment.


Introduction
Malignant pleural mesothelioma (MPM) is an asbestos-related cancer characterized by an extremely long latency. Current classification is based on three main histological subtypes, epithelioid, sarcomatoid and biphasic, having respectively better, worse and intermediate prognosis. 1 Since MPM is usually diagnosed in advanced stages, chemotherapy usually remains the only therapeutic option, but it is only modestly effective, with a median overall survival of approximately 12 months. 1 This limited efficacy is also ascribed to the immune-evasive attitude of MPM that is characterized by a low antigenicity and by an immune-suppressive environment. [2][3][4] Molecular classification of MPM has lagged behind compared to other cancer types. Two recent high-throughput genomic analyses 5,6 and provisional data from The Cancer Genome Atlas TCGA (https://tcga-data.nci.nih.gov) indicate that MPM has a generally low mutational burden. 7 On the other hand, sporadic observations indicate that many genes involved in epigenetic modifications, such as BAP1, NF2, SPOP, NUTM1, LATS1/2 SMARCA4/1, SETDB1, SETD2, can be deleted, mutated or amplified, 7,8 while druggable kinases are not generally altered, thus limiting the use of existing targeted therapies.
BET-Bromodomain Inhibitors (BBIs) represent a new class of drugs that modulate the epigenetic and the transcriptional program of cancer cells exerting a very potent therapeutic action in several hematological and solid tumors. [9][10][11] Interestingly, it has been recently demonstrated that the BBI OTX015 decreases MPM cell proliferation by reducing c-Myc expression and delays MPM tumor growth with an efficacy comparable to standard chemotherapy. 12 However, the functional interactions between BBI and the host immune system of mesothelioma (MPM) tumors as well as the ability of BBIs to alter the immunogenicity of MPM cells remain therapeutically unexplored. Here, we show that BBIs act as multitasking agents that are able to interfere with MPM cell growth and to convert an immune-suppressive to an immune-active environment.

Results
BRDs are amplified or overexpressed in primary MPM samples and BBI treatment reduces cell growth of patient-derived MPM cells We first interrogated through the cBioPortal 13,14 publicly available TCGA data of 87 MPM samples (MESO). Interestingly, BRD2, BRD3, BRD4 and BRD9 were either amplified or up-regulated in 6, 2, 9 and 13 cases, respectively (n D 87; Fig. 1A). Collectively, BRDs were up-regulated in 28/87 (32%) MPM samples. Thereby we extended BRD expression analysis to our series of 15 primary MPM samples (Tables S1 and S2). BRD2, BRD4 and BRD9 were significantly upregulated in tumors compared to primary not-transformed human mesothelial cells (HMC; Fig. 1B). Consistently with the high expression of BRDs in MPM, both BBIs JQ1 and OTX015 impaired cell proliferation in a dose-dependent manner in all histological subtypes of patient-derived MPM cells ( Fig. 2A  and B, Fig. S1 A and B). Importantly, a concentration of 250 nM of BBIs was sufficient to interfere with cell cycle progression (Fig. 2C, Fig. S1C, Fig. S2A and B). However, the anti-proliferative activity of JQ1 was not associated to apoptosis (Fig. 2D), and OTX015 treatment was accompanied by a modest increase in cell death (about 15%; Fig. S1D).

BBIs induce immunogenic cell death (ICD) along with adaptive immune response against MPM cells
Since inhibitors of chromatin-associated enzymes and BRDs can exert their therapeutic action also by modulating tumor  cell immunogenicity 15,16 we investigated this aspect in our primary patient-derived MPM cells under BBI treatment. Intriguingly, JQ1 and OTX015 increased the release of ATP (Fig. 3A, Fig. S3A) and High Mobility Group Protein 1 (HMGB1; Fig. 3B, Fig. S3B) in the extracellular supernatant of MPM cells, as well as the exposure of the "eat-me signals" calreticulin (CRT; Fig. 3C, Fig. S3C) and ERp57 (Fig. 3D, Fig. S3D), without affecting these parameters in non-transformed HMC. All these findings are typical of immunogenic cell death (ICD), a process that promotes an anti-tumor adaptive response followed by expansion of T lymphocytes 17,18 with an increased percentage of cytotoxic CD8 C CD107 C cells 19 and secretion of IFN-g. 17,18 Accordingly, BBIs significantly increased the DC-mediated phagocytosis of patient-derived MPM cells, which were more resistant to phagocytosis than HMC in untreated condition (Fig. 3E, Fig. S3E). As we previously observed, 20 proliferation of co-cultured CD8 C T-lymphocytes was negatively affected by MPM cells respect to normal HMC (Fig. 3F, Fig. S3F). This low expansion was associated with a lower IFN-g secretion (Fig. 3G) and percentage of CD8 C CD107 C cells (Fig. 3H), after co-culture with DC that had phagocytized HMC or MPM cells. Conversely, BBIs increased all these parameters in patient-derived MPM cells, without significant differences across histotypes (     Table 1, Table S3). These data suggested that patient-derived MPM cells primed immune cell populations to an immune-suppressive rather than an immune-active environment. Interestingly, treatment with either JQ1 or OTX015 counteracted the immune-suppressive potential of MPM cells, increasing CD8 C T-lymphocytes (Fig. 4A, Table 1, Table S3) and decreasing Gr-MDSC and Mo-MDSC ( Fig. 4B and C, Table 1, Table S3).
The expression of the immune checkpoints on CD8 C and CD4 C T-lymphocytes (e.g. PD-1, CTLA-4, TIM-3 and LAG-3), plays a key role in MPM-induced immune suppression. 4 Indeed, CD8 C T-lymphocytes co-cultured with patient-derived MPM cells showed an increased expression of PD-1, CTL-4 and LAG-3 ( Fig. 5A and B, Table 2). Notably, JQ1-treated MPM cells reduced the proportion of PD-1and LAG-3 positive CD8 C ( Fig. 5A and B, Table 2) and CD4 C (Table S4) T-lymphocytes. Although PD-L1 and LAG-3 were expressed at higher levels in MPM than in HMC, JQ1 reduced both markers at levels comparable to HMC cells (Fig. 5C, Table 3). Treatment of MPM cells with OTX015 induced the same modulation of immune checkpoints on CD8 C and CD4 C T-lymphocytes (Tables S5 and S6), as well as in tumor cells (Table S7).

JQ1 reduces tumor growth and immunosuppressive tumorinfiltrating cells in vivo
The efficacy of BBI was finally evaluated against murine MPM AB1 cells, implanted in syngeneic immunocompetent or immunodeficient Balb/C mice. AB1 derived tumors grew more rapidly in nude Balb/C mice than in immunocompetent hosts ( Fig. 6A and B). JQ1 was particularly effective in restraining tumor growth in immunocompetent animals ( Fig. 6A and B), suggesting that BBI activity could be partly ascribed to modulation of the immune system. Accordingly, flow cytometry analysis of intra-tumor immune infiltrate revealed that JQ1 increased the amount of DC (Fig. 6C, Fig. S4A) and CD8 C Tlymphocytes (Fig. 6D, Fig. S4B), and reduced the amount of Gr-MDSC (Fig. 6E, Fig. S4C) and Mo-MDSC (Fig. 6F,  Fig. S4C). The production of IFN-g from draining lymph nodes was also increased (Fig. 6G). Notably, JQ1 did not elicit signs of myelosuppression or liver, kidney and heart toxicity (Table S8).

Discussion
Resistance to conventional chemotherapy, lack of effective targeted therapies, low antigenicity of MPM and its ability to induce an immune-suppressive environment suggest that novel therapeutic strategies, including epigenetic drugs, should be explored to treat MPM patients. Histone deacetylase and DNA methyltransferase inhibitors have been evaluated in clinical trials in combination with chemotherapy, obtaining only a low rate of partial response associated to a high degree of toxicity. 21 Since BBIs have well documented activities on both tumor and immune system cells, we hypothesized that they could represent a novel potential therapeutic option in MPM. Data released from TCGA and analysis of our series of primary MPM samples indicate that several BRD members are overexpressed in MPM compared to HMC. In a screening of more than 650 cancer cell lines treated with JQ1, cells were classified as "JQ1-sensitive" if their IC 50 was lower than 1 mM. 11 Since JQ1 reduced MPM cell proliferation and induced cell cycle arrest at nanomolar concentrations, MPM cells may be reasonably considered sensitive to the drug. Differently from previous reports that tested JQ1 in the range of 0.5-5 mM, 10,11 in this work the reduction of cell proliferation was not paralleled by increased apoptosis. Accordingly, also the treatment at nanomolar concentrations with OTX015 was associated to a very modest apoptotic index. It can be argued that the induction of apoptosis is a concentration-dependent event and that 250 nM of BBIs is below a putative "pro-apoptotic" threshold. Alternatively, it can be hypothesized that different mechanisms of cell death are involved. ICD is a process that makes dying tumor cells visible to the immune system, following stress events such as chemotherapy or radiotherapy that induce endoplasmic reticulum (ER) stress and/or alter autophagy mechanisms. This is associated to ATP and HMGB1 release in the extracellular environment, and to the exposure on the cell surface of ERresiding proteins, such as calreticulin and ERp57. All these signals contribute to recruitment and activation of local DC to remove dying cancer cells. 18 MPM cells are known to be refractory to chemotherapy-induced ICD. 19 Of note, both JQ1 and OTX015 BBIs overcame such refractoriness and induced a typical ICD signature in MPM cells, increasing DC-mediated phagocytosis and the subsequent expansion of anti-tumor CD8 C T-lymphocytes characterized by cytotoxic activity. It has already been reported that JQ1 activates antigen-presenting cells against melanoma, 22 a tumor with high immunogenicity. Our results are particularly relevant because MPM is a poorly immunogenic tumor. 2,3 Moreover, MPM-infiltrating DC are defective in presenting tumor antigens and inducing a CD8 Cmediated anti-tumor response. 23 Interestingly, both BBIs spared not-transformed mesothelial cells from ICD. The differential expression of BRD between HMC and MPM cells may explain such selectivity, and may represent an advantage for using BBIs in MPM.
MPM is associated to an immune-suppressive rather than immune-active micro-environment, as documented by increased amount of anergic CD4 C and CD8 C T-lymphocytes, Treg and MDSC. [24][25][26] Our experimental data from MPM/ PBMC co-cultures well fit with the findings from the analysis of the immune infiltrate in murine models and MPM patients. Indeed, compared to HMC, patient-derived MPM cells decreased the amount of CD8 C T-lymphocytes, and increased the amount of Treg, Gr-MDSC and Mo-MDSC. Importantly, BBIs modified two critical cell populations in the immuneenvironment associated to MPM. First, BBI-treated MPM cells showed an increase in CD8 C T-lymphocytes. Second, BBIs significantly reduced the percentage of Gr-MDSC and Mo-MDSC that are critical in sustaining MPM progression. 27 Since active anti-MPM CD8 C T-lymphocytes induce the apoptosis of MDSC, 28 BBIs likely induced a virtuous circle: by increasing MPM cell immunogenicity and priming it for ICD, the drugs activate anti-tumor CD8 C clones that eliminate MDSC; in turn, the reduction of MDSC rescues the   4 Consistently, we found that both CD4 C and CD8 C T-lymphocytes co-cultured with patient-derived MPM cells had increased expression of PD-1, CTLA-4 and LAG-3 compared to lymphocytes co-cultured with HMC. PD-L1 and LAG-3 were also more expressed on MPM cells than on HMC. The prognostic implications of PD-1/PD-L1 axis in MPM are well characterized. 29 For instance, PD-L1 expressing MPM have a worst prognosis 29,30 and are characterized by an increased number of immune-suppressive Treg and anergic PD-1/TIM-3-positive CD4 C and CD8 C T-lymphocytes. 31 Interestingly, PD-L1 is a direct target of BRD4 and JQ1 has been identified as the most effective BBI in reducing PD-L1 transcription in ovarian cancer and increasing the anti-tumor activity of CD8 C T-lymphocytes. 32 Also in our experimental MPM models, JQ1 and OTX015 significantly reduced PD-L1 expression in MPM cells, as well as PD-1 expression in co-cultured CD8 C and CD4 C T-lymphocytes. The disruption of this biological circuit may further contribute to overcome the immune anergy induced by MPM cells.
The high expression of BRD members may underlie the efficacy of BBIs in our series of primary MPM samples. Our results indicate that the broad activity of BBIs seen in MPM models is not related to specific clinical or histological features of MPM patients from which they were derived. However, we acknowledge that our MPM series, although representative of the three histotypes and of the main clinical and pathological features of MPM, are rather limited and can potentially lead to data over interpretation. Expansion of this collection of MPM patientderived models will help to identify potential unresponsive tumors and characterize the molecular bases of refractoriness to BBIs.
A limitation of this study may be related to the challenge of MPM cells with PBMC of healthy donors. On one hand, our results may provide useful indications about the alterations induced by MPM on a healthy immune system and about the rescuing activity of BBIs. However, our work cannot predict the effect of BBIs on the immune infiltrate of MPM patients that is known to change during MPM progression. 33 To partially overcome this limitation, we measured the effects of JQ1 on local immune system in a preclinical model of MPM. JQ1 resulted significantly more effective against MPM growing in immunocompetent rather than in immunodeficient animals. These results suggest that a significant fraction of JQ1 effect was due to the restoration of a proper anti-tumor immune activity. Murine MPM growth is characterized by a first phase of progressive increase of Treg cells that suppress T-lymphocytes functions, followed by a second phase of progressive increase of MDSC. 27 MDSC are well detectable within MPM of untreated Balb/C mice, suggesting that our model mirrors an advanced stage of MPM. The intratumor immune infiltrate profiling of JQ1-treated animals recapitulates the events induced by the drug in ex vivo assays, i.e. the increase of DC and CD8 C T-lymphocytes, and the reduction of Gr-MDSC and Mo-MDSC. The high ratio of CD8 C T-lymphocytes/MDSC observed in JQ1-treated animals indicates a clear shift from an immune-suppressive to an immuneactive environment. Indeed, the higher production of IFN-g from tumor-draining lymph nodes confirmed the presence of active cytotoxic CD8 C T-lymphocytes. Collectively, our results well reconcile with the experimental observation that JQ1, in combination with histone deacetylase inhibitors, fosters a T cell-mediated anti-tumor immunity against nonsmall cell lung cancer. 34 In summary, we demonstrated that BBIs induce the reduction of MPM cell proliferation and the reversion of the MPM-induced immune-suppression. Strategies combining epigenetic drugs and immunotherapy 35 are under Table 1. Immune phenotype analysis of immune cells co-cultured with HMC and MPM cells treated with JQ1.
Cell type

Immune-phenotype of primary MPM cells
The DDCT method was employed to analyze the data. HuPO expression was used to normalize the results.

Cell proliferation analysis
For long-term cell proliferation assays, 2 £ 10 3 cells were seeded in each well of a 24-well plate in complete growth media containing the indicated JQ1 or OTX015 concentrations. After 10 days, medium was aspirated, cells were fixed and stained with 5% w/v crystal violet solution in 66% v/v methanol, washed and analyzed under bright field Olympus IX73 microscope (Olympus Corporation), equipped with the CellSense Dimension imaging system (10 £ objective; 10 £ ocular lens). For short-term proliferation assay, cells were plated in 96-well plates at a density of 2 £ 10 3 per well. Proliferation was evaluated by CellTiter-Glo (Promega). Proliferation at day 0 was considered as 100%; the results were expressed as percentage of proliferation vs day 0. HMC and MPM cells (epithelioid: epi; biphasic: bip; sarcomatoid: sar; n D 4/histotype), incubated for 6 days in fresh medium (ctrl) or with 250 nM JQ1, were co-cultured with PBMC. After this incubation time, CD3 C CD8 C T-lymphocytes were isolated and analyzed by flow cytometry. Results are expressed as means §SEM percentage of cells positive for the indicated markers. Ã p < 0.05; ÃÃÃ p < 0.001: MPM vs HMC ctrl; p < 0.01; p < 0.001: JQ1-treated vs untreated MPM cells.  Phagocytosis DC were generated from peripheral blood samples obtained from healthy donors provided by Blood Bank of AOU Citt a della Salute e della Scienza, Torino, Italy (#DG-767/2015), as previously reported. 19 The percentage of phagocytized cells at 4 C was less than 5% than the phagocytized cells at 37 C (not shown). Phagocytosis rate was expressed as phagocytic index, calculated as previously reported. 37 T-lymphocyte proliferation in HMC/MPM co-cultures 1 £ 10 6 /ml human PBMC, isolated from buffy coats of healthy donors (Blood Bank, Citt a della Salute e della Scienza di Torino Hospital, Torino, Italy) by centrifugation on Ficoll-Hypaque density gradient, were treated with anti-CD3 (1:2 000, mouse clone OKT3, BioLegend) and anti-CD28 (1:500, mouse clone 37.51, BioLegend) antibodies, to induce the specific proliferation of T-lymphocytes. PBMC were co-cultured at an effector/target (HMC or MPM cells) ratio of 10:1 for 6 days. The proliferation of T-lymphocytes was assessed by adding 1 mCi of [ 3 H]thymidine (PerkinElmer) 18 h before the end of the co-cultures, then harvesting the plates and counting the radioactivity.

T-lymphocyte activation
After MPM cell phagocytosis, DC were washed and co-cultured 10 days at a 1:5 ratio with autologous T-lymphocytes, isolated by immuno-magnetic sorting with the Pan T Cell Isolation Kit (Miltenyi Biotec.). The percentage of CD8 C CD107 C T-lymphocytes, indicative of active anti-tumor cytotoxic T-lymphocytes, was determined by flow cytometry. 19 IFN-g in the culture supernatant of CD8 C T-cells co-cultured with DC or in the supernatant of tumor-draining lymph nodes was measured with the Human IFN-g DuoSet Development Kit (R&D Systems). Results were expressed as ng/mg cell proteins or pg/ml, according to the respective calibration curves.

Immune phenotyping
PBMC isolated from buffy coats as indicated above, were incubated for 6 days with HMC or MPM cells, then harvested, washed and re-suspended in PBS containing 5% v/v FBS. A three-and four-color flow cytometry was performed on 1 £  In vivo tumor growth, immune-environment analysis and hematochemical parameters 1 £ 10 7 AB1 cells, mixed with 100 mL Matrigel, were injected subcutaneously in 6 weeks-old female immunocompetent or nude Balb/C mice (Charles River Laboratories), housed (5 per cage) under 12 h light/dark cycle, with food and drinking provided ad libitum. Tumor growth was measured daily by caliper and was calculated according to the equation (LxW 2 )/2, where L D tumor length; W D tumor width. When the tumor reached the volume of 50 mm 3 (day 15 after injection), mice were randomized into 2 groups, treated intraperitoneally twice a week for 3 consecutive weeks, as follows: 1) control group, treated with 0.1 ml saline solution; 2) JQ1 group, treated with 0.1 ml JQ1 (in 1:10 sterile saline solution/DMSO solution; final dosage: 50 mg/kg). Tumor volumes were monitored daily by caliper and animals were euthanized with zolazepam (0.2 ml/kg) and xylazine (16 mg/kg) at the end of treatment. Tumor were excised, cut into 1 mm 3 -pieces and digested (in DMEM medium containing 1 mg/ml collagenase and 0.2 mg/ml hyaluronidase) for 1 h at 37 C. The material was filtered in a syringe using a 70 mm-cell strainer to obtain a single cell suspension, and washed in DMEM. Infiltrating immune cells were collected by centrifugation on Ficoll-Hypaque density gradient and subjected to immune phenotyping by flow cytometry. Draining lymph nodes were collected, homogenized for 30s at 15 Hz, using a TissueLyser II device (Qiagen) and centrifuged at 12 000 £ g for 5 minutes. The supernatant was used to measure the amount of IFN-g. The hemocromocytometric analyses were performed with a UniCel DxH 800 Coulter Cellular Analysis System (Beckman Coulter Inc.) on 0.5 ml of blood collected immediately after mice sacrifice; lactate dehydrogenase, aspartate aminotransferase, alanine aminotransferase, alkaline phosphatase, creatinine, creatine phosphokinase were measured using kits from Beckman Coulter Inc. Animal care and experimental procedures were approved by the Bio-Ethical Committee of the Italian Ministry of Health (#122/2015-PR).

Statistical analysis
All data in the text and figures are provided as means §SEM. The results were analyzed by a one-way analysis of variance (ANOVA), using Statistical Package for Social Science (SPSS) software (IBM SPSS Statistics v.19). p < 0.05 was considered significant.

Disclosure of potential conflicts of interest
No potential conflicts of interest were disclosed.
Italian Ministry of University and Research (EX60% Funding 2015 to SN and CR, EX60% Funding 2016 to RT); Italian Ministry of Health (to LR and GVS); Fondazione Cassa di Risparmio di Torino (to CR); University of Turin, Progetti Ateneo 2016, Compagnia di San Paolo (to RT). ICS is recipient of PhD scholarships from the Italian Institute for Social Security (INPS). The funding institutions had no role in the study design, data collection and analysis, or in writing the manuscript.