BAP1 and YY1 regulate expression of death receptors in malignant pleural mesothelioma

Malignant pleural mesothelioma (MPM) is a rare, aggressive, and incurable cancer arising from the mesothelial lining of the pleura, with few available treatment options. We recently reported that loss of function of the nuclear deubiquitinase BRCA1-associated protein 1 (BAP1), a frequent event in MPM, is associated with sensitivity to tumor necrosis factor–related apoptosis-inducing ligand (TRAIL)–mediated apoptosis. As a potential underlying mechanism, here we report that BAP1 negatively regulates the expression of TRAIL receptors: death receptor 4 (DR4) and death receptor 5 (DR5). Using tissue microarrays of tumor samples from MPM patients, we found a strong inverse correlation between BAP1 and TRAIL receptor expression. BAP1 knockdown increased DR4 and DR5 expression, whereas overexpression of BAP1 had the opposite effect. Reporter assays confirmed wt-BAP1, but not catalytically inactive BAP1 mutant, reduced promoter activities of DR4 and DR5, suggesting deubiquitinase activity is required for the regulation of gene expression. Co-immunoprecipitation studies demonstrated direct binding of BAP1 to the transcription factor Ying Yang 1 (YY1), and chromatin immunoprecipitation assays revealed BAP1 and YY1 to be enriched in the promoter regions of DR4 and DR5. Knockdown of YY1 also increased DR4 and DR5 expression and sensitivity to TRAIL. These results suggest that BAP1 and YY1 cooperatively repress transcription of TRAIL receptors. Our finding that BAP1 directly regulates the extrinsic apoptotic pathway will provide new insights into the role of BAP1 in the development of MPM and other cancers with frequent BAP1 mutations.


Edited by Eric Fearon
Malignant pleural mesothelioma (MPM) is a rare, aggressive, and incurable cancer arising from the mesothelial lining of the pleura, with few available treatment options. We recently reported that loss of function of the nuclear deubiquitinase BRCA1-associated protein 1 (BAP1), a frequent event in MPM, is associated with sensitivity to tumor necrosis factor-related apoptosis-inducing ligand (TRAIL)-mediated apoptosis. As a potential underlying mechanism, here we report that BAP1 negatively regulates the expression of TRAIL receptors: death receptor 4 (DR4) and death receptor 5 (DR5). Using tissue microarrays of tumor samples from MPM patients, we found a strong inverse correlation between BAP1 and TRAIL receptor expression. BAP1 knockdown increased DR4 and DR5 expression, whereas overexpression of BAP1 had the opposite effect. Reporter assays confirmed wt-BAP1, but not catalytically inactive BAP1 mutant, reduced promoter activities of DR4 and DR5, suggesting deubiquitinase activity is required for the regulation of gene expression. Co-immunoprecipitation studies demonstrated direct binding of BAP1 to the transcription factor Ying Yang 1 (YY1), and chromatin immunoprecipitation assays revealed BAP1 and YY1 to be enriched in the promoter regions of DR4 and DR5. Knockdown of YY1 also increased DR4 and DR5 expression and sensitivity to TRAIL. These results suggest that BAP1 and YY1 cooperatively repress transcription of TRAIL receptors. Our finding that BAP1 directly regulates the extrinsic apoptotic pathway will provide new insights into the role of BAP1 in the development of MPM and other cancers with frequent BAP1 mutations.
Malignant pleural mesothelioma (MPM) is a rare and aggressive cancer that arises from the mesothelial lining of the lungs and is commonly associated with occupational exposure to asbestos. There are currently no curative therapies. Standard first-line treatment is combination chemotherapy consisting of an antifolate and a platinum agent that offers only a modest survival benefit (1). Advances in the understanding of MPM tumor biology have led to the development of multiple novel targeted agents currently in preclinical and clinical development. Many of these therapies lack a biomarker for activity, and results so far have not delivered an effective clinical therapy (2).
We have previously demonstrated that loss of BAP1 function results in sensitivity to the death receptor (DR) agonist recombinant tumor necrosis factor (TNF)-related apoptosisinducing ligand (rTRAIL) (23). TRAIL is a member of the TNF cytokine superfamily. It activates the extrinsic apoptotic pathway by binding to either of two DRs, DR4 or DR5, which leads to the recruitment of the adaptor protein Fas-associated protein with death domain and caspase-8 to form the deathinducing signaling complex (24). Once formed, catalytic subunits of caspase-8 are cleaved and activate downstream effector caspases triggering apoptosis (25,26). Activation of this pathway by TRAIL is specific to cancer cells; however, the mechanism of this selectivity is poorly understood (27,28). Several therapeutic DR agonists including rTRAIL and agonistic DR4/5 antibodies have been developed (29)(30)(31). Clinical trials of such agents to date have demonstrated broad tolerability but unfortunately limited therapeutic benefit (32). Potential reasons include the suboptimal pharmacokinetics of ‡ These authors contributed equally to this work. * For correspondence: Sam M. Janes, s.janes@ucl.ac.uk; Neelam Kumar, neelam.kumar@alumni.ucl.ac.uk.
compounds, resistant cell populations, and the lack of a targeting biomarker (33). Novel DR agonists with improved pharmacokinetics are in development, and potential biomarkers such as BAP1 are emerging (34,35). We have extensively validated the association between loss of BAP1 function and increased sensitivity to rTRAIL in in vitro, in vivo, and ex vivo models (23). Here, we set out to delineate the mechanisms underlying this association. We hypothesize that BAP1 activity modulates expression of proteins of the extrinsic and intrinsic apoptosis pathways with an increase in proapoptotic protein expression in the absence of BAP1 activity. We demonstrate that both BAP1 activity and rTRAIL sensitivity correlate with expression of the DRs, DR4 and DR5, at the transcriptional level. As BAP1 lacks DNA-binding sites, we searched for the transcriptional factor that cooperates with BAP1 to modulate expression of DR4 and DR5 identifying the polycomb group protein Ying Yang 1 (YY1).

Loss of BAP1 activity correlates with increased DR4 and DR5 expression and increased rTRAIL sensitivity
We have previously shown that MPM cells with loss of BAP1 function are more sensitive to treatment with rTRAIL (23). To determine the mechanism underlying this, we investigated the expression of DRs, DR4 and DR5, the levels of which are known to significantly contribute to TRAIL response (36,37), and nuclear BAP1 expression, a surrogate for wt-BAP1 status (7). Immunohistochemical analysis of human tissue microarrays (TMAs) (88 cores from 32 patients) (Fig. 1A) demonstrated a significant correlation between loss of nuclear BAP1 expression and higher DR4 and DR5 expression (Figs. 1B and S1). This correlation was further supported by immunohistochemistry in primary MPM tissue samples collected as part of the MSO1 clinical trial (NCT00075699) (38); samples that lacked nuclear BAP1 also showed elevated levels of DR4 and DR5 (Fig. S2, B and C). Interestingly, when we used antibodies against cytokeratin 5 (CK5) and calretinin to confirm the areas of mesothelioma, we observed higher expression of DR4 and DR5, where CK5 or calretinin is expressed. This suggests that DR4 and DR5 are expressed in mesothelioma cells but not in surrounding stromal tissue (Fig. S2A). The expression of DRs on cancer cells but not nontransformed cells, including stromal tissue, is an existing theory for the selectivity of rTRAIL and other DR agonists for cancer cells, which our data support (36).
We further confirmed the correlation of loss of BAP1 activity and high DR4 and DR5 expression in a panel of MPM cell lines. Immunoblot analysis of MPM cell lines (seven BAP1 mutants and seven wt-BAP1) overall demonstrated a higher level of DR4 and DR5 expression in BAP1-mutant versus wt-BAP1 cell lines ( Fig. 1, C and D). The wt-BAP1 cell line, CRL2081, however expressed a high level of DR4 and was found to be rTRAIL sensitive. The wt-BAP1 cell line H2803 expressed a high level of DR5, yet remained rTRAIL resistant. It cannot simply be inferred, therefore, that expression levels of DR4 or DR5 alone determine rTRAIL sensitivity in these wt-BAP1 cells. Indeed, the apoptotic pathway consists of dozens of proteins, many of which are mutated in cancer cells. We hypothesize that it is the balance of proapoptotic and antiapoptotic factors that determine TRAIL sensitivity, of which DR4 and DR5 are likely to be dominant but not fully determinant. Indeed, we observed additional heterogeneous changes in expression of 20 other proteins involved in the extrinsic and intrinsic apoptosis pathways; however, they did not directly correlate with the mutational status of BAP1 or rTRAIL sensitivity (Fig. S3).
We have previously shown that strong nuclear BAP1 expression is highly correlated with rTRAIL resistance in human early passage and unsequenced MPM cultures (Meso-banK UK) (23,(39)(40)(41). Here, in further support of a correlation between loss of BAP1 activity and increased DR4 and DR5 expression and rTRAIL sensitivity, immunoblot analysis revealed that DR4 and DR5 expression was higher in MPM cultures with loss of nuclear BAP1 expression, and these cells were more sensitive to rTRAIL treatment (Figs. 1E and S4). Flow cytometry analysis also showed higher surface expression of DR4 and DR5 in MPM cultures with loss of nuclear BAP1 expression (Fig. 1F). Taken together, our data demonstrate strong inverse correlations between BAP1 expression and DR4 and DR5 expression, which may underlie the ability of BAP1 to determine rTRAIL sensitivity.
TRAIL has been documented in some cells to induce antiapoptotic, rather than proapoptotic, pathways. Therefore, we investigated expression of antiapoptotic proteins following treatment with rTRAIL (42)(43)(44)(45)(46). We examined cellular FLICElike inhibitory protein, a catalytically inactive caspase-8 homolog that competes with caspase-8, inhibitors of apoptosis proteins (cellular inhibitor of apoptosis 1 and 2), mitogenactivated protein kinase, and NFκB pathways that enhance proliferation and induce cellular inhibitor of apoptosis proteins (36). We saw no induction of these proteins, excluding this as a mechanism of TRAIL resistance in BAP1-mutant cells (Fig. S5).

Loss of BAP1 function increases DR4 and DR5 expression in malignant but not in nontransformed cells
To further investigate the relationship between BAP1 and DR4 and DR5 expression, we knocked down BAP1 expression in a BAP1-wt MPM cell line using lentiviral shRNA constructs. BAP1 knockdown significantly increased expression of both DR4 and DR5 ( Fig. 2A). DR5 has two isoforms; the expression of both was found to increase with BAP1 knockdown ( Fig. 2A). It is not understood if there is a difference in function between these isoforms (47,48). We confirmed that BAP1 knockdown resulted in increased sensitivity to rTRAIL in these cells (Fig. 2B). Induction of cleaved caspase-8 and cleaved PARP was observed only in BAP1 knockdown cells indicating apoptosis activation only in the absence of BAP1. To examine the effect across additional tumor types, we next knocked down BAP1 in two wt-BAP1 CCRCC cell lines. This also resulted in increased expression of DR4 and DR5 and increased sensitivity to rTRAIL (Fig. 2, C and D). An additional shRNA clone confirmed these results in these two CCRCC lines and in a wt-BAP1 MPM cell line (Figs. 2C and S6). BAP1 knockdown also lead to increased DR4 and DR5 mRNA levels ( Fig. 2E) indicating that the effect of BAP1 on DR4 and DR5 expression is at the transcriptional level. Significantly, BAP1 knockdown in human lung fibroblasts and human bronchial epithelial cells (HBECs) did not affect expression of DR4 and DR5 or sensitivity to rTRAIL suggesting this effect is specific to malignant cells (Fig. S7).

BAP1 negatively regulates transcription of DR4 and DR5
To test if BAP1 DUB activity is required for transcriptional regulation of DR4 and DR5 expression, we next transduced a BAP1-null early passage MPM cell line, Meso-8T, with a lentiviral construct expressing wt-BAP1 or BAP1 with an inactivating mutation in the DUB site C91A-BAP1 or A95D-BAP1. Various mutations at C91 have been reported and shown in COSMIC (https://cancer.sanger.ac.uk/ cosmic) (49). A95D is a naturally occurring mutation in MPM tumors in patients (6). Transduction with wt-BAP1 but not C91A-BAP1 resulted in a decrease in DR4 and DR5 expression (Fig. 3A). Flow cytometry confirmed a decrease in surface expression of DR4 and DR5 in cells transduced with wt-BAP1 but not C91A-BAP1 or A95D-BAP1 (Fig. 3B). Cell survival assays confirmed that transduction with wt-BAP1, but not C91A-BAP1 or A95D-BAP1, resulted in a significant reduction in rTRAIL sensitivity (Fig. 3C). Concordantly, we saw decreased activation of caspase-8, caspase-3, and reduced PARP cleavage in wt-BAP1-transduced cells relative to C91A-BAP1-transduced cells when treated with rTRAIL ( Fig. 3A) reflective of reduced activation of the extrinsic apoptotic pathway in the presence of wt-BAP1. Quantitative PCR (qPCR) analysis demonstrated that DR4 and DR5 mRNA expressions were both decreased in cells transduced with wt-BAP1 relative to those transduced with C91A-BAP1 suggesting that regulation of DR4 and DR5 expression by catalytically active BAP1 is at the transcriptional level (Fig. 3D). These results were confirmed in a further MPM cell line, H28, which harbors a BAP1 splice site mutation commonly found in MPM tumors (Fig. S8) (6). We have also previously confirmed reduced DR4 and DR5 expression in C91A BAP1-transduced relative cells to BAP1-wt-transduced H226 cells using flow cytometry analysis (23). Subsequently, we tested the effect of BAP1 on DR4 and DR5 transcription more directly. Meso-8T cells were transduced with lentiviral vectors with luciferase reporters under the control of DR4 or DR5 promoters (50). These reporter cells were also transduced with either wt-BAP1 or A95D-BAP1. Cells transduced with wt-BAP1 displayed a significantly lower luciferase activity than those transduced with A95D-BAP1 or the parental cell line reflecting decreased DR4 and DR5 transcriptional activity in the presence of functional BAP1 (Fig. 3E). Together, the aforementioned results support that the DUB activity of BAP1 mediates transcriptional repression of DR4 and DR5. To test whether this in turn determines rTRAIL sensitivity, we used two complementary approaches. First, we knocked down DR4 or DR5 in wt-BAP1 H2869 MPM cells transduced with BAP1 shRNA. BAP1 knockdown increased the sensitivity of H2869 cells to rTRAIL as expected (Fig. 3F). Interestingly, DR5, but not DR4 knockdown, in shBAP1-H2869 cells abolished the effect of BAP1 knockdown, resulting in rTRAIL resistance (Fig. 3F). Second, we knocked down DR4 or DR5 in the BAP1-null, rTRAIL-sensitive Meso-8T cell line. DR5 knockdown only slightly decreased rTRAIL sensitivity, but DR4 knockdown reduced it to a similar level as transduction with wt BAP1 (Fig. 3G). These data are in line with previous reports showing preferential use of one of the two receptors by distinct cell types (31). For example, hematological cancers seem to prefer DR4 for induction of apoptosis (51,52), whereas solid tumors appear to exhibit heterogeneity in DR preference (31,53,54).

YY1 negatively regulates transcription of DR4 and DR5
As BAP1 does not bind to DNA directly (5), we aimed to identify transcription factors that bind to the promoter regions of DR4 and DR5. Bioinformatics analysis of 2000 nucleotides of the promoter region of DR4 and DR5 was conducted. From Relative mRNA expression was normalized to beta-2-microgloblin (B2M) expression. Data are shown as the mean ± SD of two experiments performed in triplicates. *p < 0.05; **p < 0.01. E, reporter assay for promoter activities of DR4 and DR5 in parental Meso-8T cells transduced with a luciferase reporter under the control of DR4 or DR5 promoter and cells further transduced with wt-BAP1 or A95D-BAP1. Firefly luciferase/Renilla luciferase ratios were determined as relative luciferase activities. Data are shown as the mean ± SD of two experiments (n = 6 in each experiment). *p < 0.05; **p < 0.01. F, cell viability assay of wt-BAP1 H2869 cells transduced with empty vector (EV) or shBAP1 following treatment with a dose range of rTRAIL (0-1000 ng/ml) for 72 h and for shBAP1 cells further transduced with DR4 (shDR4) or DR5 shRNA (shDR5) following the same treatment. G, cell viability assay of parental BAP1 null Meso-8T early passage MPM cells transduced with wt BAP1 (wt-BAP1) or DR4 (shDR4) or DR5 shRNA (shDR5) following treatment with a dose range of rTRAIL (0-1000 ng/ml) for 72 h. +, ++; lentiviral titer. Error bars represent the SD. BAP1, BRCA1-associated protein 1; rTRAIL, recombinant tumor necrosis factor-related apoptosis-inducing ligand.
On the trail of BAP1 and YY1 candidates identified (Fig. S9), YY1 was selected for further analysis as it has previously been shown to negatively regulate DR5 expression in prostate cancer (55,56). Furthermore, YY1 has been shown to bind directly to BAP1, with the C-terminal region of BAP1 essential for this interaction, forming a complex capable of regulating gene expression (57). YY1 knockdown with two different shRNA clones in wt-BAP1 MPM and CCRCC cells resulted in increased expression of both DR4 and DR5 without affecting steady-state levels of BAP1 (Figs. 4A and S10). shRNA knockdown of BAP1 in MPM cells also did not affect steady-state levels of YY1 (Fig. S11A). In addition, we did not observe any difference in YY1 expression based on BAP1 mutational status and BAP1 expression level (Fig. S11B). qPCR analysis confirmed increased mRNA expression of DR4 and DR5 in cells transduced with YY1 shRNA (Fig. 4B). YY1 knockdown also significantly increased sensitivity to rTRAIL and the DR5 agonist Medi3039 in MPM and CCRCC cells (Fig. 4C) (58). We also determined if YY1 is able to regulate DR4 and DR5 expression in the absence of BAP1. We knocked down YY1 in BAP1-mutant MPM cell lines and BAP1-null early passage MPM cells and assessed the expression of DR4 and DR5 and rTRAIL sensitivity. Neither DR4/DR5 expression nor TRAIL sensitivity increased in the YY1 knockdown cells (Fig. 4, D and E) in these BAP1-mutant cells, unlike in BAP1wt cells, suggesting that BAP1 is required for DR4/5 regulation by YY1. These data demonstrate that YY1, in addition to BAP1, modulates expression of DR4 and DR5. As YY1 and BAP1 have been shown to form a complex capable of regulating gene expression, it is likely that this complex regulates DR4 and DR5 expression (57). BAP1 and YY1 act at DR4 and DR5 promoters to facilitate transcriptional repression BAP1 has been shown to form a ternary complex with YY1 and host cell factor 1 (HCF-1) in HeLa cells (57). Through its coiled-coil motif, BAP1 directly interacts with the zinc fingers of YY1, whereas HCF-1 interacts with the middle region of YY1 and is essential for the formation of the ternary complex in vivo (57). Therefore, we aimed to determine if BAP1 and YY1 also interact directly in MPM and CCRCC cells. Protein extracts from H2818, MPP89, and Caki-1 cells were coimmunoprecipitated using anti-YY1 antibody or immunoglobulin G (IgG) as a control. Immunoblot confirmed the interaction of endogenous YY1 with BAP1 (Fig. 5A). To verify the specificity of these results, we compared results of coimmunoprecipitation (co-IP) assay in BAP1-null MPM cells (Meso-8T) that were transduced with wt-BAP1 or a control vector alone. A strong interaction of YY1 and BAP1 was detected only in cells transduced with wt-BAP1 but not the control vector, confirming the specificity of the YY1-BAP1 interaction (Fig. 5B). Here, co-IP assay demonstrates that YY1 also interacts with HCF-1 in wt-BAP1 CCRCC cells (Fig. S12). However, we have previously shown that MPM cells expressing BAP1 that lacks the binding domain for HCF-1 are not significantly different in their TRAIL sensitivity compared with cells expressing wt-BAP1 (23). This suggests that the HCF-1/BAP1 interaction does not determine TRAIL sensitivity and is unlikely to be involved in DR regulation.
In addition to physical interactions, we sought to examine the functional interaction between YY1 and BAP1. As BAP1 does not have a DNA-binding domain, but directly interacts with the transcriptional repressor YY1, we hypothesized that BAP1 and YY1 are recruited to the promoter regions of DR4 and DR5. Chromatin immunoprecipitation (ChIP) assays were performed with antibodies for BAP1, YY1, or IgG as a control. The immunoprecipitated DNA was analyzed with probes for DR4 or DR5 by qPCR in Meso-8T cells transduced with wt-BAP1, C91A-BAP1, or a control vector. Both BAP1 and YY1 were enriched in the promoter regions of DR4 and DR5 in cells transduced with wt-BAP1 but not the control vector (Fig. 5C). Interestingly, BAP1 and YY1 were also enriched in these promoter regions in cells transduced with C91A-BAP1 indicating BAP1 and YY1 are recruited to these promoter regions regardless of DUB activity. This finding is consistent with previous reports that catalytically inactive BAP1 is also recruited to FoxK2-binding regions (59). Catalytically inactive BAP1 has also previously been shown to form a complex with YY1 (57). Taken together, we show that BAP1 and YY1 are recruited at the promoters of TRAIL receptors and are necessary to initiate transcriptional regulation of TRAIL receptors.

Discussion
We have recently reported that loss of BAP1 function is a predictive biomarker for rTRAIL sensitivity in cancer (23). In this study, we delineate the underlying molecular mechanism. We demonstrate BAP1 and the transcriptional regulator YY1 act at the promoter regions of DR4 and DR5, where they facilitate transcriptional repression of DR4 and DR5, which requires BAP1 DUB activity. Decreased cell surface expression of DR4 and DR5 and reduced activation of the apoptotic pathway in turn mediates rTRAIL resistance in wt-BAP1 cells. Conversely, increased cell surface expression of DR4 and DR5 in BAP1-mutant cells mediates the observed increased sensitivity to rTRAIL. Various mechanisms of resistance to rTRAIL and other DR agonists have been suggested (36). Evidence supports that low expression of DR4 and DR5 because of mutations, promoter methylation, constitutive endocytosis, or deficient transport to the cell surface is important (36,(60)(61)(62). Indeed, strategies to enhance the efficacy of rTRAIL treatment, such as a combination with chemotherapeutic drugs, have been demonstrated to mediate these effects through increased DR expression (25). Our results are consistent with these data and support the centrality of DR expression in TRAIL therapeutics. YY1 inhibition has previously been shown to upregulate DR5 expression and enhance rTRAIL sensitivity in prostate cancer and B-non-Hodgkin's lymphoma cells (55,56,63). Here however, we show that YY1 is involved in the transcriptional regulation of both DR4 and DR5 and is enriched at the promoters of both DR4 and DR5 when BAP1 is present. BAP1 is known to form multiprotein complexes including as many as ten partners, which in turn determine the precise targets of its DUB activity (5,64). It has previously shown that BAP1 forms a multiprotein complex with YY1 and the transcriptional cofactor HCF-1 (57). Although not investigated here, further work might identify additional cofactor(s) that direct BAP1 and YY1 to the DR4 and DR5 promoters. We have previously shown that mutation of the additional sex combs-like (ASXL)-binding site on BAP1 and ASXL1 knockdown also increases rTRAIL sensitivity (23). The BAP1-ASXL1 complex is a polycomb repressor DUB complex capable of deubiquitination of histone 2A at lysine 119 (H2A119Ub), a process that modulates expression of the polycomb genes (5). Interestingly, YY1 has also been shown to interact with polycomb proteins (57,65). It may therefore be that YY1 interacts with both BAP1 and ASXL1 to modulate DR expression through the deubiquitination of H2A119Ub or that BAP1 and YY1 form an alternate complex with a different target that modulates histone and chromatin structure DR expression.
YY1 and BAP1 may be involved more widely in the transcriptional regulation of the TNF receptor superfamily. Nitric oxide has been shown to inhibit YY1 binding to the Fas promoter resulting in Fas upregulation and cell sensitization to Fas ligand-induced apoptosis in prostate cancer (66). YY1 has also been shown to suppress the Fas promoter activity in Bnon-Hodgkin's lymphoma and colon cancer (67,68). We have also previously demonstrated that BAP1 knockdown sensitizes MPM cells to Fas ligand and TNF alpha (23).
Although BAP1 was originally identified as a tumor suppressor gene, accumulating evidence has revealed roles in multiple clinically targetable pathways (69)(70)(71)(72). Indeed, we have proposed BAP1 expression to be a stratifying biomarker for sensitivity to DR agonists (23), and our work here provides a biological rationale for this. The current study also demonstrates that YY1 knockdown enhances the sensitivity to TRAIL and DR5 agonist. YY1 is overexpressed in many types of cancer, and high expression correlates with poor clinical outcomes and resistance to chemotherapy and immunotherapy making it an attractive therapeutic target (73,74). Thus, targeting the BAP1-YY1 axis may be an additional novel therapeutic strategy in TRAIL therapeutics.

Experimental procedures
Cell culture All cancer cell lines were obtained from the Wellcome Trust Sanger Institute except the H226 line that was kindly gifted by Dr P. Szlosarek (Barts Cancer Institute). Cancer cell lines were cultured in RPMI1640, Dulbecco's modified Eagle's medium (DMEM), or DMEM and nutrient mix 12 medium (DMEM:F12) supplemented with 10% fetal bovine serum (FBS), penicillin/streptavidin, and sodium pyruvate. Early passage human mesothelioma cells were obtained from Mes-obanK UK (39) and cultured in RPMI1640 medium supplemented with 5% FBS, 25 mM Hepes, penicillin/streptavidin, and sodium pyruvate. Primary human lung fibroblasts (kind gift from Dr R. Chambers at University College London) were cultured in DMEM media supplemented with 10% FBS and penicillin/streptavidin in an incubator with 10% CO 2 (75). Experiments were conducted on cells between passage 6 and 8. Primary HBECs were obtained from endobronchial biopsies with patient consent as previously described (76). Ethical approval was obtained through the National Research Ethic Committee (reference 06/Q0505/12). All studies involving human subjects abide by the Declaration of Helsinki principles. HBECs were cultured in bronchial epithelial growth medium (Lonza) on top of 3T3-J2 mouse embryonic fibroblast feeder cells inactivated by mitomycin-C treatment (0.4 μg/ml; Sigma-Aldrich).

2,3-bis-(2-methoxy-4-nitro-5-sulfophenyl)-2H-tetrazolium-5carboxanilide cell viability assay
Cells were seeded in 96-well plates in 100 μl media per well at a density of 40,000 cells/ml 1 day prior to treatment with soluble rTRAIL (PeproTech) or MEDI3039 (MedImmune). 2,3-bis-(2-methoxy-4-nitro-5-sulfophenyl)-2H-tetrazolium-5carboxanilide reagent and the activation solution (AppliChem, LLC and Akron Biotech) were mixed and added to the cells at the end of treatment. The plate was returned to a CO 2 incubator to incubate for 2 h, and the absorbance at a wavelength of 490 nm was measured using a microplate reader. Relative cell viability was calculated as a fraction of viable cells relative to untreated cells.

Immunoblotting
Cells were lysed in radioimmunoprecipitation lysis buffer (Sigma-Aldrich) with protease inhibitors (cOmplete, Mini Protease Inhibitor Cocktail; Roche) on ice to extract protein. About 30 μg of protein samples were separated by SDS-PAGE and transferred onto nitrocellulose membranes using iBlot2 Dry Blotting System (Thermo Fisher Scientific). Membranes were incubated with specific primary antibodies, washed, incubated with secondary antibodies, and visualized using an ImageQuant LAS 4000 imaging system (GE Healthcare). Anti-BAP1 antibody was purchased from Santa Cruz Biotechnology, and anti-cellular FLICE-like inhibitory protein antibody was purchased from Enzo Life Sciences. All the other antibodies were purchased from Cell Signaling Technology. Quantification of bands was performed using ImageJ (Image Processing and Analysis in Java; the National Institutes of Health).

IP
Cells were lysed in IP buffer containing 0.2% NP-40, 20 mM Tris-HCl (pH 7.4), 150 mM NaCl, 10% glycerol, and protease inhibitors. The lysates were incubated overnight with gentle rocking with anti-YY1 antibody (ab38422; Abcam) or IgG (2729; Cell Signaling Technology). Protein-A magnetic beads (Pierce Biotechnology and Thermo Fisher Scientific) were added, and incubation was continued for 1 h. The beads were washed with IP buffer, and proteins were eluted from the beads by heating with SDS sample buffer. Proteins were separated by SDS-PAGE, and immunoblotting was performed as described previously with anti-BAP1 antibody (sc-28383; Santa Cruz).

Lentivirus production and cell transfection
Lentiviral particles were produced by cotransfection of 293T cells with construct plasmids and the packaging plasmids pCMV-dR8.74 and pMD2.G (kind gifts from Dr Adrian Thrasher, University College London) using a DNA transfection reagent jetPEI (Source BioScience). The viral particles were concentrated by ultracentrifugation at 17,000 rpm (SW28 rotor, Optima LE80K Ultracentrifuge; Beckman) for 2 h at 4 C. To determine the titers of prepared lentivirus, 293T cells were transduced with serial dilutions of viruses in the presence of 8 μg/ml polybrene, and protein expression was assessed by flow cytometry and immunoblotting.

RT-qPCR
Total RNA was extracted from the cells using SV Total RNA Isolation System (Promega) according to the manufacturer's instructions. Complementary DNA was synthesized using iScript Reverse Transcription Supermix for RT-qPCR (Bio-Rad Laboratories). qPCR was performed using TaqMan probes (DR4: Hs00269492_m1; DR5: Hs00366278_m1; beta-2-microglobulin: Hs00187842_m1) and TaqMan Gene Expression Master Mix (Life Technologies) as per the manufacturer's protocol. Relative expression of DR4 and DR5 was calculated using comparative CT method with a reference gene, beta-2-microglobulin.

Immunohistochemistry
Tumor biopsies taken from patients with MPM in the MS01 trial (NCT00075699) were stored as formalin-fixed paraffinembedded blocks or as unstained mounted sections as previously described (38). The TMA slides containing tumor samples from patients with MPM were obtained from MesobanK UK. All studies involving human subjects abide by the Declaration of Helsinki principles. To assess expression of DR4, DR5, CK5, and calretinin, samples were first incubated in the oven at 60 C for 30 min, then deparaffinized and rehydrated using an automated tissue processor (Tissue-Tek). Antigen retrieval was achieved by immersion in 10 mM citric acid buffer (pH 6.0) at 95 C for 15 min. After washing with PBS and blocking with 2.5% normal goat serum, samples were incubated with primary antibody: anti-DR4 (1:500 dilution, ab8414; abcam), anti-DR5 (1:500 dilution, ab8416; abcam), anti-calretinin (1:200 dilution, NCL-L-CALRET-566; Leica Biosystems), anti-keratin 5 (BioLegend; catalog no.: 905501, 1:500 dilution) in 1% bovine serum albumin/ 4% serum overnight at 4 C. Samples were incubated with ImmPRESS polymer reagent (VECTOR Laboratories) for 30 min and stained with ImmPACT Nova RED (VECTOR Laboratories). H&E staining was carried out using an automated tissue processor (Tissue-Tek). Staining for BAP1 was performed as described before using anti-BAP1 antibody (1:150 dilution, sc-28282; Santa Cruz Biotechnology) (38). Images were acquired using a NanoZoomer 2.0HT whole slide imaging system (Hamamatsu Photonics). Histology and nuclear BAP1 assessment were performed by two consultant pathologists. Intensity of DR4 and DR5 expression was assessed blindly by three independent observers and scored as follows (no staining = 0; low staining = 1; medium staining = 2; and strong staining = 3).

Bioinformatical analysis
To identify the common transcription factors that potentially regulate these genes, the 2000 nucleotide sequence of the promoter regions of DR4 and DR5 are entered into Human Core-Promoter Finder (http://rulai.cshl.org/tools/genefinder/ CPROMOTER/human.htm).

Statistical analysis
Data were evaluated using the statistical analysis and indicated with p values. p < 0.05 was considered statistically significant. Using Prism 8 (GraphPad Software, Inc), Student's t test was performed to analyze differences between two groups, whereas one-way ANOVA was used to determine the differences between three or more independent groups.
For the statistical analysis of TMAs, linear mixed modeling was used to account for multiple samples per patient, including the patient ID as a random effect. Linear mixed models were implemented using the Bioconductor Ime4 and ImerTest packages. Pairwise t test confirmed that there was no systematic bias between the score of different observers.

Data availability
All data are contained within the article.
Supporting information-This article contains supporting information.