Purpose:

The clinical standard treatment for patients with malignant pleural mesothelioma (MPM) includes a cisplatin-based chemotherapy, leading to reduction of tumor size in only a minority of patients. Predicting response to chemotherapy in patients with MPM by using a genetic marker would, therefore, enable patient stratification.

Experimental Design:

In this retrospective biomarker study, eligible patients had resectable MPM, measurable disease, and available primary MPM tissue. All patients underwent first-line treatment with cisplatin and pemetrexed, followed by surgery. Thorough molecular analysis was performed (whole-exome and targeted deep sequencing, and copy-number analyses), and also mechanistic in vitro data (viability assays, Western blots, and immunoprecipitation) using mesothelioma cell lines with and without siRNA-mediated BRCA1-associated protein 1 (BAP1) knockdown were provided.

Results:

In a training cohort of patients with MPM (n = 28), mutations or deletions of BAP1 each predicted resistance to chemotherapy in patients with primary MPM. The negative predictive value of BAP1 loss in patients with MPM was confirmed by amplicon sequencing and copy-number array technology in an independent test cohort (n = 39). Preliminary mechanistic studies using siRNA-based knockdown of BAP1 in MPM cell culture models along with immunoprecipitation assays confirmed chemoresistance in vitro, possibly through inhibition of apoptosis and transcriptional regulation of the BAP1/HCF1/E2F1 axis.

Conclusions:

Alterations in BAP1 in MPM were a negative predictor for response to chemotherapy and could possibly be used as a companion biomarker for treatment decision.

Translational Relevance

Malignant pleural mesothelioma (MPM) is a rare but fatal lung cancer. Because of the lack of many alternative treatment options, standard clinical therapy regimens include a platinum-based therapy in combination with an antifolate. However, chemotherapy shows effectiveness only in about one third of all patients with MPM, exposing two thirds of them to unnecessary and mostly severe side effects. In our project, we screened a cohort of 67 patients with MPM undergoing similar cisplatin-based treatment for a genetic marker predictive of response to chemotherapy and found that alterations in BRCA1-associated protein 1 (BAP1) were a negative predictor of MPM outcome. Using different MPM cell lines, we demonstrated that the absence of BAP1 in vitro is causative for cisplatin resistance. We therefore suggested that BAP1 mutational status could be used for patient stratification before chemotherapeutic treatment. In view of the recent FDA approval of ipilimumab plus nivolumab as first-line treatment for adults with unresectable MPM, neoadjuvant chemotherapy regimens for patients with resectable MPM are to be discussed in case of BAP1 alterations.

Malignant pleural mesothelioma (MPM) is a rare lung cancer entity that is still considered incurable. MPM arises from a thin layer of mesothelial cells lining the pleura and is mostly caused by inhalation of asbestos fibers accumulating in the pleura. Recent studies have concluded that MPM is caused by the chronic inflammatory process caused by the release of HMGB1 extra cellularly leading to the development of MPM over approximately 40 years (reviewed in ref. 1). The old hypothesis that asbestos fibers might mechanically cause genetic mutations by direct interference of asbestos fibers with the mitotic spindle and reactive oxygen release was in fact incorrect (2). Although asbestos is banned in many parts of the world, it continues to be produced in some countries, such as India, China, and Russia. Not only does the mining of asbestos continues, but its use is actually increasing exponentially in some countries, such as India and Africa (1). It is also still not completely banned in the United States (3, 4). As a consequence, the incidence of MPM is still rising (5). Of note, Comertpay and colleagues have shown that MPMs originate as polyclonal tumors and suggested that the carcinogenic field effect leads to several premalignant clones giving rise to polyclonal malignancies (6).

In the current histopathologic classification, MPMs are divided into epithelioid biphasic and sarcomatoid subtypes. The epithelioid subtype is the most prevalent and is associated with a better prognosis than the biphasic or sarcomatoid subtype. In general, the median life expectancy with palliative care is only around 7 months (7) and can be prolonged by multimodality treatment to more than 20 months (8). The multimodal therapy of patients with potentially resectable MPM usually consists of a cisplatin-based chemotherapy in combination with an antifolate, like pemetrexed, surgery, and radiotherapy (9). However, chemotherapy is only effective in around 30%–40% of the patients (10). As with other highly aggressive cancers, ipilimumab and nivolumab were recently approved by the FDA as first-line treatment for adult patients with unresectable MPM (11).

Identification of possible targets for therapy has been hampered by the genetic composition of MPMs and the high inter- and intratumoral heterogeneity (12). MPM is characterized by a high frequency of large chromosomal deletions and frequent mutations in mainly tumor suppressor genes, such as CDKN2A (p16), NF2 (Merlin), and BAP1, whereas mutations in oncogenes are rarely detected (13). Germline mutations in BAP1 are known to predispose for MPM (14), but also to several other cancers, such as renal cell carcinoma (1, 15, 16). BRCA1-associated protein 1 (BAP1) is a 729 amino acid ubiquitin carboxyl-terminal hydrolase that plays a role in cell-cycle progression, DNA damage response (DDR), histone modification, and apoptosis. The protein is found inside the nucleus, as well as in the cytoplasm (17–22).

Up to now, the cause of the intrinsic resistance of MPM to the cisplatin-based chemotherapy has been unclear. Recent publications suggest the involvement of several pathways based on in vitro and in vivo studies (23–25) or patient tissue analysis (26), but the results have so far not been translated into clinical use.

In this study, we genetically analyzed tissue samples of a cohort of patients with MPM and found that BAP1 mutations and deletions (summarized as “alterations”) are associated with resistance to cisplatin-based chemotherapy. We then present functional genetic evidence showing the causative role of decrease of BAP1 mRNA for cisplatin resistance using cell line isogenic models. On the basis of our data, we propose a model in which loss of BAP1 protein leads to downregulation of apoptosis via transcriptional downregulation of apoptotic genes that is probably regulated by interplay of BAP1, HCF1, and E2F1.

Patient cohort

For this retrospective, single-center nonrandomized biomarker study, tumor and normal tissue were collected from 67 patients between 1999 and 2015. All patients were treated with a neoadjuvant chemotherapy consisting of cisplatin and pemetrexed, followed by surgery [pleurectomy with decortication (P/D) or extrapleural pneumonectomy (EPP)] at the University Hospital Zurich (Zurich, Switzerland). The studies were conducted in accordance with the Declaration of Helsinki. All patients signed an informed consent and the study was approved by the Ethical Committee Zürich (KEK-ZH-2012-0094 and KEK-ZH-2015-0171). Response to chemotherapy was assessed in PET-CT scans according to the modified RECIST (mRECIST) criteria (27). Results of experiments were reported according the recommendations for tumor marker prognostic studies (REMARK).

Cell culture

The mesothelial cell line, Met5A (RRID:CVCL_3749), and the mesothelioma cell lines, NCI-H2052 (RRID:CVCL_1518), MSTO-211H (RRID:CVCL_1430), and NCI-H28 (RRID:CVCL_1555), were directly and newly purchased from the ATCC. The mesothelioma cell lines, JL-1 (RRID:CVCL_2080) and DM-3 (RRID:CVCL_2019), were purchased from DSMZ. All cell lines were kept in the media recommended by the manufacturer in a cell incubator at 37°C and 5% CO2. Authenticity of each newly acquired cell line was verified by fingerprinting. All cell lines were kept at low passage number (below 15) for the experiments and Mycoplasma negativity was confirmed by routine testing.

DNA and RNA isolation

For DNA isolation from formalin-fixed, paraffin-embedded (FFPE) patient tumor tissue, areas with high tumor content were marked on hematoxylin and eosin (H&E)-stained slides with a 2 μm tissue section by experienced pathologists (B. Vrugt and P.J. Wild). Afterwards, tissue cores (0.6 mm diameter) from FFPE blocks were taken from the marked area. If several blocks from the same tumor were available, several punches were taken and pooled for DNA extraction. From every patient, corresponding normal tissue from nonmalignant lymph nodes was punched as well. DNA was isolated using the Maxwell 16 FFPE Tissue LEV DNA Purification Kit (Promega) according to the manufacturer's instructions.

For DNA isolation from cell lines, pelleted cells were washed with PBS and the DNA was isolated using the Maxwell 16 Blood DNA Purification Kit (Promega) according to the manufacturer's instructions. The DNA was eluted in 50 μL nuclease-free water and stored at −20°C. The DNA was quantified using the Qubit 2.0 and the dsDNA HS Assay Kit (Thermo Fisher Scientific).

For RNA isolation from cell lines, pelleted cells were washed twice with PBS and the RNA was isolated using the Maxwell 16 Cell RNA Purification Kit (Promega) according to the manufacturer's instructions. The RNA was eluted in 50 μL nuclease-free water and stored at −80°C. The RNA was quantified using the Qubit 2.0 and the RNA HS Assay Kit (Thermo Fisher Scientific).

Tissue microarray construction and IHC

From the FFPE tissue blocks, H&E-stained slides of all specimens were inspected by an experienced pathologist (B. Vrugt) for identification of representative, tumor-rich areas for tissue microarray (TMA) construction. Afterwards, four tissue cores (diameter 0.6 mm) from each patient were taken from the respective areas and arranged in a new “recipient” block. Afterwards, 2 μm sections were cut from the “recipient” block (TMA). Antigen retrieval was performed with citrate buffer for 30 minutes at 100°C (for total caspase-3) or with Tris-EDTA-Borate at 100°C for 90 minutes (for BAP1 and cleaved caspase-3). Afterwards, the sections were incubated with the following primary antibodies: BAP1 (sc-28383, Santa Cruz Biotechnology, 30 minutes, 1:200), cleaved caspase-3 (Cell Signaling Technology, catalog no., 9661, RRID:AB_2341188, 1:400), and total caspase-3 (Cell Signaling Technology, catalog no., 9662, 1:000). The staining was performed on a BondIII (Leica, for BAP1) or Benchmark Discovery (Ventana, for cleaved caspase-3 and total caspase-3). The IHC detection kits used were the Bond Polymer Refine Detection Kit (BAP1) or ChromoMapKit with UltraMap anti-Rb horseradish peroxidase as secondary antibody (Roche, catalog no., 760-4311, RRID:AB_2811043, for cleaved and total caspase-3).

Amplicon panel resequencing

For targeted amplicon–based resequencing, a custom-designed MPM panel was used as described previously (12). For library preparation, 10 ng of DNA was used and processed according to the Ion AmpliSeq DNA Library Preparation user guide (Thermo Fisher Scientific). Template preparation and sequencing were performed according to the manufacturer's protocol (Ion PI HI-Q OT2 200 Kit and Ion PI Hi-Q Sequencing 200 Kit, Thermo Fisher Scientific). For alignment, variant calling, and the following filtering, we used Ion Reporter 5.0 (Thermo Fisher Scientific). All samples were sequenced at a median coverage of around 7,800 × (minimum, 2,870 ×). Details on the bioinformatics analysis can be found in the Supplementary Data.

Copy-number variation arrays

Array-based genome-wide copy-number analysis was conducted at IMGM Laboratories GmbH using OncoScan FFPE Microarrays (Affymetrix). For the assay, an input amount of 79.2 ng DNA was used. The molecular inversion probe processing was done according to the OncoScan FFPE Assay Kit Protocol (Affymetrix). The AGCC Viewer v4.2.1567 and the OncoScan Console v1.3.0.39 (Affymetrix) were used for data analysis. We performed the analysis of the copy-number variations (CNV) and the LOH with the Nexus Express for OncoScan v3.1 software. The CNVs were visualized using Circos software (RRID:SCR_011798).

Whole-exome sequencing

The exome sequencing was performed at Fasteris SA using the SureSelect human all exon V6 kit according to the manufacturer's instructions. Whole-exome sequencing data were aligned to the human reference genome (hg19) using bwa (28). Postprocessing of the alignment included removal of secondary alignments with samtools (RRID:SCR_002105), marking of PCR duplicates using Picard (RRID:SCR_006525), as well as base quality score recalibration and indel realignment using GATK 3.5 (RRID:SCR_001876). Strelka, Mutect, and Varscan 2 (RRID:SCR_000559, RRID:SCR_005109, and RRID:SCR_006849) were used to perform somatic variant calling. Only variants called by at least two of the callers were considered. Variant annotation was performed using SnpEff (RRID:SCR_005191) and SnpSift (RRID:SCR_015624). Tumor/normal concordance and sample contamination were estimated using Conpair. Afterwards, the variants that were found by at least two of the three variant callers were normalized on exon size to adjust for the basic mutational rate. Variants that were found in two or more samples were plotted in a heatmap.

Cell viability assays

For testing the cell viability to increasing concentrations of drugs, we first reversely transfected the cell lines with four pooled siRNAs targeting BAP1 (Qiagen) or the AllStars Negative Control siRNA (Qiagen). For this purpose, 24.5 μL OptiMEM Reduced Serum Media (Thermo Fisher Scientific) were mixed with 0.25 μL RNAiMAX Lipofectamine (Thermo Fisher Scientific) and 2.5 pmol siRNA per well and after 30 minutes pipetted into Nunc MicroWell 96-Well Optical-Bottom Plates (Thermo Fisher Scientific). Then, 6,000 cells in 75 μL of the cell line appropriate medium without antibiotics were added to the siRNA mix. The cell number was determined using a Neubauer counting chamber and Trypan Blue (Sigma) staining. After 24-hour incubation at 37°C, 5% CO2, the drugs were added in serial dilutions, including a nontransfected control and a transfected control treated with the respective solvent of the drug. All drugs were purchased from Selleckchem. Cisplatin was dissolved in dimethyl formamide (DMF), and all other drugs in DMSO. Drugs were serially 1:3 diluted in the solvent and then diluted in antibiotic-free media, to a maximum final concentration of 0.2% DMSO or DMF. Every concentration and the controls were tested in six technical replicates. After 48 hours, the supernatant was removed and the cells were incubated in 100 μL PrestoBlue Cell Viability Reagent (Thermo Fisher Scientific) diluted 1:10 in the appropriate medium for 1 hour, followed by fluorescence detection in an Infinite 200 Plate Reader (Tecan). Effectiveness of the siRNA knockdown was confirmed by RNA extraction, followed by reverse transcription to cDNA and qPCR (see below).

Gene expression analysis in cell lines

For the expression analysis, cell lines were reversely transfected. For this purpose, 163 μL OptiMEM Reduced Serum Media (Thermo Fisher Scientific) were mixed with 1.7 μL RNAiMax Lipofectamine (Thermo Fisher Scientific) and 17 pmol siRNA (either BAP1 siRNA pool or negative siRNA, see above) per transfection, incubated for 30 minutes at room temperature, and pipetted into 24-well plates. Afterwards, 8 × 104 cells in 500 μL appropriate medium without antibiotics were seeded on top. After 24-hour incubation at 37°C, 5% CO2, 10 μmol/L cisplatin dissolved in DMF and diluted in appropriate medium was added to the cells, as well as DMF as a control, to a final DMF concentration of 0.2% and incubated for another 48 hours. Afterwards, four wells of each treatment/transfection combination were pooled for protein extraction and two wells for RNA extraction.

Reverse transcription and qPCR

For reverse transcription, 200 ng of RNA was reverse transcribed using the high-capacity cDNA reverse transcription kit according to the manufacturer's instructions. For qPCR, all primers were designed using the primer3 software (RRID:SCR_003139). To ensure amplification of RNA only, intron spanning primers were chosen wherever possible. The complete list of primers is available on request. All primers, except for ACTB, were ordered from Microsynth AG. Primers for ACTB, which was used as a housekeeping control, were ordered as QuantiTect Primer Assay (Qiagen). A mastermix containing 5 μL SYBR Green PCR Master Mix (Thermo Fisher Scientific), 500 nm forward and reverse primers, and nuclease-free water was mixed with 10 ng of cDNA to a final volume of 10 μL per reaction. Every sample was analyzed in triplicates. Further details can be found in the Supplementary Data.

Protein extraction and Western blotting

Transfected and stimulated cells (see above) in 24-well plates were washed with PBS and then directly lysed with 100 μL per well cOmplete Lysis-M (Roche), transferred into microtubes, and centrifuged at 14,000 × g for 10 minutes. The supernatant containing the proteins was transferred into a fresh tube and used for protein concentration measurements using the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific) following the manufacturer's protocol and the Infinite 200 Plate Reader (Tecan). Afterwards, 20 μg of protein was mixed with NuPAGE LDS sample buffer (4 ×) containing 10% added β-mercaptoethanol and boiled for 5 minutes at 95°C. Afterwards, the denatured proteins were applied to a Mini-PROTEAN Precast Gel (Bio-Rad), as well as the Precision Plus Protein Dual Color Standard (Bio-Rad). After electrophoresis, proteins were blotted using the Trans-Blot Turbo Transfer Pack (Bio-Rad) and the included 0.2 μm polyvinylidene difluoride membranes. After blotting, the membrane was washed with TBS-T and blocked for 1 hour at room temperature with 5% nonfat milk powder in TBS-T. Afterwards, the membrane was washed three times with TBS-T and then incubated overnight at 4°C with the primary antibody (Supplementary Data). Afterwards, the membrane was washed four times with TBS-T, incubated for 10 seconds with Clarity Western ECL Substrate (Bio-Rad), and developed using the ChemiDoc XRS+ imaging system with the Image Lab software. Afterwards, the bands were washed three times with TBS-T, blocked again with 5% milk-TBS-T, and incubated for 1 hour at room temperature with an antibody against β-actin (Abcam, catalog no., ab8227, 1:1,000 in 5% milk-TBS-T) as loading control. Incubation with the secondary antibody and development of the membrane were conducted as described above.

Immunoprecipitation

H2052 cells were seeded in 6-well plates and grown in 37°C, 5% CO2 until 80% confluency. Afterwards, the lysis was conducted using 200 μL per well cOmplete Lysis-M (Roche), followed by centrifugation at 14,000 × g for 10 minutes. For immunoprecipitation, Dynabeads Protein G (Thermo Fisher Scientific) was used according to the manufacturer's protocol. In brief, 4 μg BAP1 antibody (sc28383, Santa Cruz Biotechnology) or mouse IgG1 kappa light chain (Abcam, catalog no., ab18443, RRID:AB_2736846) control antibody was coupled to 50 μL Dynabeads at room temperature for 10 minutes using a rotator. Afterwards, 300 μL lysate was added to the antibody-coupled beads and incubated on a rotator at 4°C overnight. Then, the beads were washed three times with cOmplete Lysis-M (Roche) and three times with PBS-T, eluted in the glycine buffer given in the protocol, and mixed with 4 × NuPAGE LDS Sample Buffer (Thermo Fisher Scientific) containing 10% added β-mercaptoethanol. Subsequently, the samples, as well as the controls, were loaded onto a gel and the Western blot analysis was conducted as described above.

Loss of BAP1 function is a driver in MPM development

We included 67 patients with histologically confirmed MPM in this study. All patients received three to four cycles of induction chemotherapy consisting of cisplatin and pemetrexed, followed by surgery, either EPP or P/D. Clinical data, including sex, histologic subtype, and radiological staging of chemotherapy response using mRECIST criteria (29), were available for all patients (Supplementary Table S1).

MPM is known to be intrinsically resistant to the inhibitory effects of chemotherapeutic treatment (30). Hence, we hypothesized that there might be a genetic determinant for the resistance that occurs early during the development of MPM, which is present in most or all of the tumor cells and that remains detectable after chemotherapy. To test this hypothesis, we selected 28 patients (training cohort), from whom FFPE tumor and matching normal tissue were available, and ultra-deep sequenced their genomic DNA derived from tumor tissue taken at different timepoints during therapy. These timepoints included the primary diagnostic biopsy (referred to as phase I), surgical specimen after induction chemotherapy (phase II), and relapse (phase III; Supplementary Fig. S1). To be able to trace mutations that might be only present in a subset of tumor cells due to intratumoral heterogeneity, we sampled and pooled the DNA from different locations within the tumor. Considering potential spatial genetic heterogeneity, by this approach, we ensured to get the best possible coverage of tumor subclones. Because these subclones might be present at only low allele frequencies, we decided to perform high-depth amplicon-based targeted sequencing to increase sensitivity and specificity (31). For this purpose, we used a custom-designed MPM panel for genes reported previously to be mutated in MPM (Supplementary Fig. S2; ref. 12). In 11 of 28 patients, somatic nonsynonymous mutations were undetectable in the exons of the 30 genes covered by the panel (Fig. 1A). However, in the remaining 17 patients, we detected somatic variants at varying allele frequencies mostly in NF2 (10/28, 36%), BAP1 (6/28, 21%), and TP53 (4/28, 14%). Interestingly, for some of the patients, for example, patient 3 and 95, the genetic composition of the tumor varied between the different phases, reflecting the spatial and temporal heterogeneity in MPM. For a better visualization of the data, we modeled the tumor evolution of patient 3 and 95 from the amplicon sequencing data (Fig. 1B). The models are based on the assumption that mutations of a high allele frequency are present in a higher fraction of tumor cells and are, therefore, of early clonal origin (32). Thus, in patient 3, the BAP1 mutation arose before the NF2 mutation and both mutations did not change during chemotherapy (Fig. 1B, left). In the relapsing phase, two other mutations in TERT and NF2 occurred. In patient 95, in contrast, TP53 was mutated rather early during MPM development, followed by an NF2 mutation (Fig. 1B, right). During chemotherapy, this composition of the tumor changed dramatically by the occurrence of another TP53 and NF2 variant in combination with the TP53 mutation found at the beginning. However, these subclones seem to have been removed during surgery, because we could only detect the TP53 and NF2 subclones from the diagnostic biopsy (phase I) in the relapse (phase III). In summary, these data show that the clonal composition in MPM may change during disease progression.

Figure 1.

Mutation status in patients with MPM during treatment. A, Mutated genes included in the amplicon panel sequencing are shown on the left and the protein positions of the detected nonsynonymous somatical mutations on the right. The samples derived from patients with MPM (“P”) and at different timepoints (“_1,” “_2,” and “_3”) are shown on the x-axis. The allele frequency is color coded. B, Graphical illustration of the clonal evolution of MPM patients 3 (left) and 95 (right). The allele frequency of the mutations was taken as a measure for emergence of the respective clone and is depicted in the width of the diagram.

Figure 1.

Mutation status in patients with MPM during treatment. A, Mutated genes included in the amplicon panel sequencing are shown on the left and the protein positions of the detected nonsynonymous somatical mutations on the right. The samples derived from patients with MPM (“P”) and at different timepoints (“_1,” “_2,” and “_3”) are shown on the x-axis. The allele frequency is color coded. B, Graphical illustration of the clonal evolution of MPM patients 3 (left) and 95 (right). The allele frequency of the mutations was taken as a measure for emergence of the respective clone and is depicted in the width of the diagram.

Close modal

When looking at the somatic mutations found in BAP1 (Fig. 1A), we could see that these were generally found at a high allele frequency in phase I (indicating an early clonal origin; ref. 32) and that these remain at a similar level throughout treatment. Most of the detected variants of BAP1 were terminating (3/6 patients) or frameshift (2/6 patients) mutations (referred to as truncating mutations). These are known to lead to a loss-of-function of the protein, because the localization into the nucleus is abolished (33). In only 1 patient (no., 75), we detected two different missense mutations predicted by PolyPhen-2 as “benign” (p.Ser585Phe; max. CADD score 20.5) and “probably damaging” (p.Leu553Val, max. CADD score 22.6). Taken together, we saw that somatic BAP1 mutations are an early event in MPM, they persist during chemotherapeutic treatment, and, thus, are a driver in MPM development.

Alterations in BAP1 are associated with cisplatin resistance

To find a genetic marker for resistance to cisplatin, we sequenced phase I FFPE tumor and normal tissue of 39 additional patients (Fig. 2C, test cohort). As the use of panel-based sequencing might exclude important variants that associate with chemoresistance due to the preselection of genes, we also conducted whole-exome sequencing in 20 patients (Supplementary Fig. S3). However, the latter only confirmed the most common variants found by the panel and did not yield any other relevant candidates for chemotherapy prediction.

Figure 2.

Copy number and mutational status of selected genes. A, Frequency of losses and gains in 55 patients with MPM detected by CNV arrays. B, Location of amplicon sequencing–detected somatic mutations in BAP1 (top scheme) and TP53 (bottom scheme). In BAP1, the catalytical UCH domain, the HCF binding motif (HBM), and the ULD, harboring two NLS, are shown, as well as the binding regions for important interactors. For TP53, the transactivation motif (TAM), the DNA-binding domain (DNA-BD), and the tetramerization motif are depicted. C, Heatmap summarizing the results from the CNV arrays and the amplicon panel sequencing for the genes included in the MPM panel. On the x-axis, the analyzed patients (P), as well as six MPM cell lines (right) and a mesothelial cell line (Met5A), are shown. D, Genetic alterations of BAP1, including small, large, and splice mutations, as well as germline mutations, detected by amplicon sequencing in 55 patients and the respective nuclear stainings, histotype, and mRECIST response status. PR, partial response.

Figure 2.

Copy number and mutational status of selected genes. A, Frequency of losses and gains in 55 patients with MPM detected by CNV arrays. B, Location of amplicon sequencing–detected somatic mutations in BAP1 (top scheme) and TP53 (bottom scheme). In BAP1, the catalytical UCH domain, the HCF binding motif (HBM), and the ULD, harboring two NLS, are shown, as well as the binding regions for important interactors. For TP53, the transactivation motif (TAM), the DNA-binding domain (DNA-BD), and the tetramerization motif are depicted. C, Heatmap summarizing the results from the CNV arrays and the amplicon panel sequencing for the genes included in the MPM panel. On the x-axis, the analyzed patients (P), as well as six MPM cell lines (right) and a mesothelial cell line (Met5A), are shown. D, Genetic alterations of BAP1, including small, large, and splice mutations, as well as germline mutations, detected by amplicon sequencing in 55 patients and the respective nuclear stainings, histotype, and mRECIST response status. PR, partial response.

Close modal

As MPM is known to harbor large chromosomal deletions (34), we also performed probe-based genome-wide copy-number microarrays of 55 patients from whom enough tumor DNA was available (Supplementary Table S2). The CNV arrays revealed recurrent losses especially in 3p21, 9p21, and 22q12, where BAP1, CDKN2A, and NF2 are located, respectively (Fig. 2A). Interestingly, 7p11 was one of the regions most commonly affected by a copy-number gain. Among others, the gene EGFR is located in that region, which might be a potential treatment option.

We then combined the data of the CNV arrays with the amplicon sequencing results (Fig. 2C). Notably, tumor suppressors are often seen to be biallelically inactivated by hemizygous loss of one allele and mutation of the other (e.g., BAP1 and NF2), or by homozygous deletions (e.g., in CDKN2A). In contrast to the other tumor suppressor genes, mutations in TP53 are classified as gain-of-function mutations (Fig. 2B; Supplementary Table S3). We also noticed a trend toward mutual exclusivity of TP53 and BAP1 mutations. Only in 1 patient (no., 75), missense mutations in BAP1 and TP53 were detected. When comparing the mutation status of the genes studied with survival, only the loss of CDKN2A was significantly associated with shorter survival (Supplementary Fig. S4). Next, we examined the relation between gene mutations and response to chemotherapy, determined by the evaluation of CT scans according to mRECIST (Material and Methods). We found that nonsynonymous mutations in BAP1 were significantly associated with decreased sensitivity to chemotherapy (Fig. 3A), defined as progressive (PD) or stable disease (SD; ref. 27). This negative predictive role of BAP1 mutations has never been shown before in patients with MPM or other types of cancer. Most of the detected variants in BAP1 were truncating mutations, frequently occurring in the C-terminal hydrolase domain, leading to a premature termination of the protein (Fig. 2B). All the truncating mutations found in our patients affected the N-terminal nuclear localization sequences (NLS; ref. 33). As BAP1 exerts most of its functions in the nucleus (35), truncating mutations are thought to result in a loss-of-function phenotype (36). Because mutations in BAP1 have been reported to occur more often in epithelioid MPMs, we also confirmed the predictive value of BAP1 somatic nonsynonymous variants in the epithelioid part of our cohort (Fig. 3D).

Figure 3.

Associations of BAP1 status with chemotherapy response based on mRECIST in the MPM cohort. Only data of analyzable samples are presented in cumulative bar charts. A and D, For BAP1 mutations, all patients (n = 67) were used for calculations. For BAP1 alterations (B and E), BAP1 nuclear staining (C and F), and BAP1 cytoplasmic staining (G and H), only the patients who were copy-number analyzed were taken into account (n = 55). The graphs in D–F and H show only the patients with epithelioid histotype. Response was defined by mRECIST as PD, SD, and partial response (PR). Significances were calculated using a two-sided Fisher exact test. I, Representative staining patterns for nuclear (“N”) and cytoplasmic BAP1 (“C”) immunoreactivity (bottom). The scale bar indicates 100 μm. cytopl., cytoplasmic; neg, negative; pos, positive.

Figure 3.

Associations of BAP1 status with chemotherapy response based on mRECIST in the MPM cohort. Only data of analyzable samples are presented in cumulative bar charts. A and D, For BAP1 mutations, all patients (n = 67) were used for calculations. For BAP1 alterations (B and E), BAP1 nuclear staining (C and F), and BAP1 cytoplasmic staining (G and H), only the patients who were copy-number analyzed were taken into account (n = 55). The graphs in D–F and H show only the patients with epithelioid histotype. Response was defined by mRECIST as PD, SD, and partial response (PR). Significances were calculated using a two-sided Fisher exact test. I, Representative staining patterns for nuclear (“N”) and cytoplasmic BAP1 (“C”) immunoreactivity (bottom). The scale bar indicates 100 μm. cytopl., cytoplasmic; neg, negative; pos, positive.

Close modal

Recent publications have shown that genetic loss of BAP1 can also occur via splice site mutations and intragenic deletions (37, 38). To see whether these types of BAP1 alterations are also associated with chemotherapy response, we first added the splice site mutations from the amplicon sequencing that were found in BAP1 (Fig. 2D). As the CNV arrays used for this study were not sensitive enough to call intragenic BAP1 deletions, we additionally used our amplicon sequencing data for detection of smaller deletions within the BAP1 gene with the CNVPanelizer R package. With this method, we found large (spanning several exons) intragenic deletions in BAP1 in 3 more patients (Supplementary Fig. S8). In addition, we used the amplicon coverage files of the BAP1 gene to manually calculate potential small intragenic deletions, finding further minute losses in 6 patients (Fig. 2D). As we did not confirm the small intragenic deletions by the use of another method (see Material and Methods), we excluded them from statistical analysis. However, including the splice site and large intragenic deletions, we could still see a significant association of BAP1 alterations with response to chemotherapy (Fig. 3B and E for epithelioid MPMs only).

As the routine diagnosis of MPM is rather based on the nuclear detection of BAP1 protein than on sequencing of the BAP1 gene, we wanted to see whether we could as well see an association of nuclear BAP1 with chemotherapy resistance. Therefore, a TMA, containing MPM tissue of 54 patients from whom CNV analysis had been performed, was stained with anti-BAP1 antibody. Loss of nuclear BAP1 immunoreactivity, however, was not found to be significantly associated with resistance to chemotherapy (Fig. 3C and F). However, as expected from the mutational data, we saw more patients with MPM with loss of nuclear BAP1 expression in the SD and PD group.

Recently, it was shown that most changes in nuclear BAP1 expression can be traced back to alterations in the BAP1 gene (37). Consequently, we had a closer look at this discrepancy of mutation and expression data. We first noticed that 2 patients had missense point mutations in BAP1 (no., 76 and 454) and still we could see expression of BAP1 in the nucleus (Fig. 2D). We, therefore, concluded that not all mutations in BAP1 necessarily lead to loss of protein expression in the nucleus, although the mutation in patient no. 76 (p.Ala92Val) was predicted to be deleterious by SIFT and PolyPhen-2 analysis (Supplementary Table S3).

Of note, loss of nuclear BAP1 expression was detected in 56% (30/54) of our MPM samples, whereas only 35% had BAP1 alterations (46% including unverified small intragenic deletions). Alterations that may have been missed with our approach of amplicon sequencing and CNV analysis were most likely intragenic tiny deletions (<3 kb; ref. 38). Nevertheless, it is possible that these changes, if not occurring in essential regions of the catalytical active ubiquitin carboxy-terminal hydrolase (UCH) domain or the middle domain (NORS, 240–598 amino acid) shown to bind to IP3R3 in the endoplasmic reticulum (ER; ref. 22), lead to a loss of protein expression in the nucleus, but not to a loss of function in the cytoplasm. We, therefore, decided to also assess the cytoplasmic immunoreactivity of BAP1 on the TMA, with no or weak cytoplasmic BAP1 expression considered negative and high expression considered positive (Fig. 3I). Indeed, we could find a highly significant association of positive BAP1 staining in the cytoplasm with response to chemotherapy in the total cohort (Fig. 3G), as well as in the subgroup of epithelioid MPMs (Fig. 3H).

Taken together, these findings suggest that patients with MPM who have BAP1 alterations at diagnosis do not benefit from chemotherapeutic treatment, which would warrant patient stratification based on BAP1 status before treatment decisions.

Decrease of BAP1 mRNA causes cisplatin resistance in vitro

We next sought to answer the question of whether alterations in BAP1 are causative or merely a predictive marker for chemotherapy resistance. Therefore, we performed siRNA-mediated knockdown of BAP1 (four pooled siRNAs) in four different BAP1-proficient cell lines, namely Met5A (virus-transformed nonmalignant mesothelial cell line), MSTO-211H, H2052, and DM-3 (mesothelioma cell lines). We verified the BAP1 status in these cell lines by using the same techniques as for the cohort sequencing (CNV arrays and targeted amplicon resequencing; Fig. 2C). We noticed a reduction of growth speed in MSTO-211H, Met5A, and DM-3, but not in H2052 cells (Fig. 4E), upon BAP1 siRNA transfection and knockdown. This phenomenon has already been described by various working groups (39, 40) and can be explained by the role of BAP1 in cell-cycle progression (41).

Figure 4.

Response of BAP1 siRNA-treated cell lines to increasing concentrations of cisplatin. A–D, Values were normalized on the DMF-treated control of BAP1 siRNA–treated and negative siRNA control–treated cells, respectively, in Met5A (n = 3), MSTO-211H (n = 3), H2052 (n = 5), and DM-3 (n = 3) cells. E, Both the negative and the BAP1 siRNA–transfected cells were normalized on the negative siRNA DMF-treated control to display the growth differences after 48-hour siRNA incubation. Differences in cell viability between negative control siRNA- and BAP1 siRNA–transfected cells were calculated with a two-sided Student t test (*, P < 0.05; ***, P < 0.001).

Figure 4.

Response of BAP1 siRNA-treated cell lines to increasing concentrations of cisplatin. A–D, Values were normalized on the DMF-treated control of BAP1 siRNA–treated and negative siRNA control–treated cells, respectively, in Met5A (n = 3), MSTO-211H (n = 3), H2052 (n = 5), and DM-3 (n = 3) cells. E, Both the negative and the BAP1 siRNA–transfected cells were normalized on the negative siRNA DMF-treated control to display the growth differences after 48-hour siRNA incubation. Differences in cell viability between negative control siRNA- and BAP1 siRNA–transfected cells were calculated with a two-sided Student t test (*, P < 0.05; ***, P < 0.001).

Close modal

When incubated with increasing concentrations of cisplatin, the Met5A mesothelial cell line showed a slight decrease in sensitivity to cisplatin in BAP1-knockdown cells compared with cells treated with negative control siRNA (neg siRNA; IC50, 2.33 vs. 1.65 μmol/L; Fig. 4A). The mesothelioma cell line, MSTO-211H, showed less sensitivity than the Met5a when treated with negative siRNA and cisplatin (IC50, 4.50 μmol/L; Fig. 4B), but turned even less sensitive upon BAP1 knockdown (IC50 > 20 μmol/L). The H2052 and the DM-3 cell lines were the most resistant ones (IC50, 12 and >20 μmol/L, respectively; Fig. 4C and D), but the cells lacking BAP1 still showed a decreased sensitivity toward cisplatin than the control cells. When repeating the experiments with MSTO-211H and H2052 and a combination of cisplatin and pemetrexed, mimicking the treatment used for patients with MPM, the cells transfected with BAP1 siRNA still showed a higher resistance to the combined treatment than the neg siRNA–transfected cells (Supplementary Fig. S5). As the H2052 cells transfected with BAP1 siRNA did not display growth impairment compared with the negative siRNA–treated cells and Met5A, we concluded that the chemoresistant effect of the decrease of BAP1 mRNA could not be explained by decreased proliferation rates and consequently less accumulation of DNA damage only, but rather by other cellular mechanisms, such as apoptosis (see below).

In summary, our data show that alterations in BAP1 are causative for cisplatin resistance in vitro, supporting the findings from our human MPM cohort and the potential role of BAP1 mutational status assessment for patient stratification.

As loss of BAP1 was shown to mimic BRCA deficiency in other cancers and can be therapeutically targeted via an approach using synthetic lethality (42), we used our BAP1 siRNA model to test whether a decrease BAP1 mRNA leads to increased sensitivity toward the PARP inhibitors, olaparib or veliparib. However, we rather detected a decrease in sensitivity toward the PARP inhibitors in cells with reduction of BAP1 mRNA than an increase (Supplementary Fig. S6). As BAP1 is known to play an important role in the DDR pathway (20, 36), we also tested a set of other DDR inhibitors, such as 10058-F4 (c-Myci), CGK-733 (ATM/ATRi), doxorubicin (TOP2i), NU7441 (DNA-PKi), veliparib (PARPi), and Pifithrin-μ (p53i), but none of them showed different efficiencies between BAP1 and negative siRNA–treated cells (Supplementary Fig. S6).

BAP1-knockdown cells show decreased levels of apoptosis

Because cisplatin is known to induce apoptosis via DNA double-strand breaks (43) and BAP1 has been implicated in the apoptosis pathway (22, 33), we hypothesized that BAP1 might mediate resistance to cisplatin via decreased levels of apoptosis. Therefore, we examined levels of cleaved caspase-3 and PARP, both markers for apoptosis. Indeed, Met5A and MSTO-211H showed a remarkable decrease of cleaved PARP and caspase-3 levels, whereas H2052 and another mesothelioma cell line, DM-3, already exhibit a high level of apoptosis resistance (Fig. 5A). Next, we asked whether we could confirm these results in vitro by measuring a reduction of cleaved caspase-3 levels in our patient cohort. Hence, we stained a TMA, including tissues from 53 patients from our cohort, with an anti-cleaved caspase-3 antibody. Remarkably, the BAP1 nonsynonymous coding mutations detected by the amplicon sequencing were significantly associated with low cleaved caspase-3 expression (Supplementary Fig. S7). When all BAP1 alterations (including splice site mutations and large intragenetic deletions) were compared with the caspase-3 staining results, only 1 patients with altered BAP1 showed high cleaved caspase-3 staining, in the remaining patients, caspase-3 only showed low expression, although this result did not reach statistical significance (Supplementary Fig. S7). Levels of cleaved caspase-3 expression were statistically not associated with mRECIST response to chemotherapy.

Figure 5.

Expression responses to BAP1 knockdown and cisplatin treatment. A and B, Protein expression was measured by Western blotting upon BAP1 siRNA–mediated knockdown or control siRNA transfection and treatment with 10 μmol/L cisplatin. β-Actin was used as loading control. C, Expression of selected genes in BAP1 siRNA- or negative siRNA control–transfected cells upon treatment with cisplatin. Expression values were compared against the negative siRNA- and DMF-treated control or against the negative siRNA- and cisplatin-treated control (*, P < 0.05; **, P < 0.01; ***, P < 0.001). CASP-3, caspase-3; cl.CASP-3, cleaved caspase-3.

Figure 5.

Expression responses to BAP1 knockdown and cisplatin treatment. A and B, Protein expression was measured by Western blotting upon BAP1 siRNA–mediated knockdown or control siRNA transfection and treatment with 10 μmol/L cisplatin. β-Actin was used as loading control. C, Expression of selected genes in BAP1 siRNA- or negative siRNA control–transfected cells upon treatment with cisplatin. Expression values were compared against the negative siRNA- and DMF-treated control or against the negative siRNA- and cisplatin-treated control (*, P < 0.05; **, P < 0.01; ***, P < 0.001). CASP-3, caspase-3; cl.CASP-3, cleaved caspase-3.

Close modal

A recent publication also showed decreased cleaved caspase-3 levels upon hemizygous deletion of BAP1 in vitro (22), linking the effect to the cytoplasmic function of BAP1 in the ER. However, when we investigated total caspase-3 levels in the cell lines, we noticed a great reduction of total caspase-3 levels upon BAP1 knockdown (Fig. 5A). In addition, we also detected a decrease in total amount of p53, an important regulator of cell survival and apoptosis. Therefore, the question arose whether these reductions in protein levels could be the result of transcriptional regulation. We, therefore, examined the levels of caspase-3 and TP53 mRNA upon BAP1 knockdown and treatment with or without cisplatin. All cell lines tested showed a strong reduction of caspase-3 levels after BAP1 knockdown (Fig. 5B). When additionally treated with cisplatin, only the Met5A cells were able to restore caspase-3 levels to the basal level. TP53 mRNA levels, in contrast, remained constant or showed slight up- or downregulation. These findings suggest that reduction of apoptosis by downregulation of (cleaved) caspase-3 is at least partially transcriptionally regulated.

In summary, our data suggest that loss of BAP1 function confers chemoresistance via inhibition of apoptosis and that this effect can be explained not only by the cytoplasmic role of BAP1, but probably also by its role as a gene regulator.

BAP1 might interact with HCF1 and E2F1 to enhance apoptosis

Given the transcriptional changes we detected upon BAP1 knockdown, we went on to identify the transcription factor interacting with BAP1 to regulate caspase-3 levels. Studies showed that most BAP1 is bound to HCF1 (19, 39). In addition, BAP1 was shown to enhance E2F1 cell-cycle–related target gene expression via HCF1 in uveal melanoma cells (44). Furthermore, E2F1 is known to enhance transcription of caspase-3 (45, 46) and that HCF1 directly binds E2F1 to enhance apoptosis (47). This led us to hypothesize that, BAP1 might control apoptosis in MPM cells by interacting with E2F1 and HCF1. Therefore, we investigated the expression of known E2F1 apoptosis–related target genes, such as E2F1 itself, APAF1, TP73, and CDKN1A(p21) (47). In our conditions, transcription of APAF1 remained more or less unchanged, but in the mesothelioma cell lines, MSTO-211H, H2052, and DM-3, we noticed a great reduction of TP73 and E2F1 mRNA and an increase of CDKN1A mRNA upon BAP1 knockdown (Fig. 5B), which is in line with previous findings studying the effects of HCF1 knockdown upon E2F1 target gene expression (47).

In a previous study in uveal melanoma cells, BAP1 was shown to act via HCF1 on E2F1 responsive promotors (44). Because BAP1 is known to directly interact with HCF1 (39) and HCF1 was shown to bind to E2F1 (47), we thought that there might as well be a direct interaction between BAP1, HCF1, and E2F1. Indeed, when performing preliminary immunoprecipitation assays with an anti-BAP1 antibody, we could detect HCF1 and E2F1, indicating that there might be an interplay between those proteins (Fig. 6A). We, therefore, hypothesized that these three proteins may interact in the transcription of apoptotic genes and that this interaction may be disrupted in the presence of BAP1 alteration (Fig. 6B).

Figure 6.

Interactions of BAP1. A, Immunoprecipitation (IP) of BAP1. Protein G Dynabeads were coupled with α-BAP1 and, matching to the BAP1 antibody, α-IgG kappa light chain isotype control and incubated with H2052 cell line lysate. Afterwards, the supernatant was used as control (flowthrough, lane 2) and the beads were washed six times (wash 6, lanes 3 and 4). Two different amounts of the respective immunoprecipitation products were run on the gel (lanes 5 and 6 compared with lanes 7 and 8). The IgG-coated beads showed some unspecific bands when incubated with the secondary antibodies, which ran higher (BAP1 development, lanes 6 and 8) or lower (E2F1 development, lanes 5–8) than the target protein. E2F1 protein in the immunoprecipitates is highlighted by the red arrow. B, Hypothesis of the emergence of cisplatin resistance due to BAP1 mutations or knockdown. CASP-3, caspase-3.

Figure 6.

Interactions of BAP1. A, Immunoprecipitation (IP) of BAP1. Protein G Dynabeads were coupled with α-BAP1 and, matching to the BAP1 antibody, α-IgG kappa light chain isotype control and incubated with H2052 cell line lysate. Afterwards, the supernatant was used as control (flowthrough, lane 2) and the beads were washed six times (wash 6, lanes 3 and 4). Two different amounts of the respective immunoprecipitation products were run on the gel (lanes 5 and 6 compared with lanes 7 and 8). The IgG-coated beads showed some unspecific bands when incubated with the secondary antibodies, which ran higher (BAP1 development, lanes 6 and 8) or lower (E2F1 development, lanes 5–8) than the target protein. E2F1 protein in the immunoprecipitates is highlighted by the red arrow. B, Hypothesis of the emergence of cisplatin resistance due to BAP1 mutations or knockdown. CASP-3, caspase-3.

Close modal

In summary, these results show that reduction of BAP1 mRNA and protein levels leads to altered expression of E2F1-regulated apoptosis genes, and, thus, resistance to chemotherapy, and that this effect may be regulated by a direct interaction of BAP1, HCF1, and E2F1.

Here, we retrospectively demonstrated in a human cohort of 67 patients with MPM that BAP1 alterations occur early during MPM development, as shown by sequencing of longitudinally obtained samples, and are negatively associated with response to chemotherapy. We verified that loss of BAP1 leads to resistance to cisplatin in vitro and that this resistance is mainly based on reduced apoptosis. In addition, we demonstrated that this inhibition of cell death cannot only be explained by the cytoplasmic role of BAP1, but also by its function as a transcriptional regulator. We, therefore, propose a model in which BAP1 interacts with HCF1 and E2F1 to control the expression of apoptosis-related genes. Our working hypothesis, based on preliminary data, is that the absence of the BAP1 protein in the nucleus disrupts a putative trimer formed by HCF1, BAP1, and E2F1, leading to transcriptional changes mediated by E2F, which is now released from the control of HCF1 and BAP1, resulting in reduced caspase-3 levels and consequently reduced apoptosis.

Although recent advances were made combining bevacizumab with the standard cisplatin/pemetrexed treatment, the efficacy of chemotherapy for patients with MPM is still limited, with maximal response rates ranging between 30% and 40% (10). As MPM shows intrinsic chemoresistance, we initially hypothesized that this might be caused by genetic alterations. To find such a predictive genetic marker, we aimed for highest possible homogeneity regarding the treatment of the patients included in this retrospective study. However, age-adjusted incidence for MPM in males in Switzerland is about 3.1 per 100,000 with around 180 new cases per year (48) and only about 60% of patients are eligible for chemotherapy followed by surgery (49). Therefore, only with contribution of several hospitals in Switzerland, we were able to build up a cohort of 67 patients that were all treated with cisplatin and pemetrexed as first-line therapy followed by surgery and from whom clinical data, such as mRECIST, overall survival (OS), and progression-free survival, were available. In addition, from all these patients, sufficient material from the diagnostic biopsy was available for DNA isolation, which usually contains very little tissue.

We sequenced tumor tissue sampled at different timepoints during the therapy, allowing us to monitor the change of dominating clones over the course of treatment, as it was shown before, for example, in chronic lymphocytic leukemia (50). Using a custom-designed MPM sequencing panel, we were able to identify mutations of BAP1 as evolutionary early mutations, as they occurred at a high allele frequency in the initial diagnostic biopsy and were retained in all timepoints sequenced (32), pronouncing the importance of these mutations also in regards of possible therapy targets (51). Our finding agrees well with the finding that germline mutations in BAP1 predispose patients to development of MPM (14), underlining the role of BAP1 mutations as a driving event. Most of the detected single-nucleotide variants in this study were frameshift and terminating mutations, they are considered as high-impact mutations leading to a truncated version of the protein. As a result, due to the missing NLS, the protein cannot translocate into the nucleus and exert its functions there (33).

Integration of the data from the panel sequencing and the CNV arrays revealed, as expected, frequent homozygous deletions of CDKN2A/p16, deletions and mutations of NF2 and BAP1, and less frequent copy-number gains. The frequencies of the coding mutations or homozygous deletions were 44% for CDKN2A, 24% for BAP1 (only counting nonsynonymous somatic mutations detected by panel sequencing), 20% for NF2, 9% for TP53, and 5% for LATS2, corresponding well with previously published observations (41, 52, 53), underlining the representativeness of our cohort.

A recent study combining several molecular techniques showed that the number of BAP1 mutations is underestimated using NGS or traditional CNV arrays only, because intragenic deletions cannot be directly detected by either of the methods (38). We, therefore, added splice site mutations, as well as possible intragenic biallelic deletions (estimated by taking into account the amplicon coverage) to the list of relevant mutations. Thereby, we reached a proportion of 45% BAP1 deletions and/or mutations, which is consistent with the recently reported one (38).

We showed that homozygous deletions in CDKN2A correlated with shorter OS, as reported previously (54). None of the other genes we tested showed a correlation with survival. Conflicting results have been published, with some groups showing a positive association of loss of expression of BAP1 and survival in MPM (55, 56), whereas other studies (37), as in the current one, could not find a significant correlation.

As described previously, the absence of BAP1 nuclear staining is a valuable diagnostic test for distinguishing benign (with positive nuclear staining) from malignant (with negative nuclear staining) mesothelial cells (reviewed in ref. 1). In general, nuclear and cytoplasmic BAP1 immunoreactivity can be seen in stromal cells and in tumor cells (with wild-type BAP1). All biallelic BAP1 mutations result in the absence of nuclear staining because either no BAP1 protein is present or the mutant BAP1 protein cannot enter the nucleus. As for the cytoplasm, two scenarios are possible in MPM. In most tumor cells that lack nuclear staining (e.g., mutant BAP1), there is no staining in the cytoplasm. However, sometimes mutant, biologically inactive BAP1 can accumulate in the cytoplasm, where it forms amyloid, and it can produce cytoplasmic staining without accompanying nuclear staining (reviewed in ref. 1).

Surprisingly, comparing the results from the panel sequencing and the CNV arrays with the mRECIST-classified response to chemotherapy, we saw a correlation of BAP1 coding nonsynonymous mutations (also in combination with splice site mutation and large intragenic deletions) with resistance to chemotherapy, a finding which, to our knowledge, has not been published for MPM before. The reason for the lack of previous reports on similar results might be that in other studies, mixed treatment cohorts were used (38, 52, 57), probably masking the effect of altered BAP1, or simply because chemotherapy response data were not available. When we tried to correlate BAP1 staining to chemotherapy response, we could also see more SD and PD patients with negative staining for nuclear BAP1, but this finding did not reach significance. However, when looking at the cytoplasmic fraction of BAP1, a significant association of high BAP1 expression with response to chemotherapy was found (Fig. 3G). For this reason, we think that a high proportion of the alterations in BAP1 that we could not detect using our methods lead to a loss of expression in the nucleus. However, the cytoplasmic role of BAP1 regulating apoptosis via IP3R3 stabilization (22) could be maintained.

Most of the mutations we found in BAP1 led to truncation of the protein, already affecting in most cases the UCH domain or the middle part of the protein (Fig. 2B). Therefore, these truncated forms of BAP1 do not contain a functional UCH37-like domain (ULD)-located NLS anymore, which is required for their function in the nucleus (33). In addition, all mutations we detected, except for the one in P46 (p.Tyr671Ter), probably also affect the cytoplasmic function of BAP1, because it was shown that a catalytically intact BAP1 and middle domain (NORS domain) are needed for IP3R3-mediated Ca2+ release and apoptosis induction (22). We, therefore, concluded that most of our detected mutations lead to a loss of function of BAP1 in the cytoplasm, as well as in the nucleus. Hence, we chose an siRNA-mediated knockdown approach to confirm the role of BAP1 in mediating chemoresistance in vitro. We could verify that BAP1 knockdown leads to increased insensitivity toward cisplatin. This effect cannot exclusively be explained by a slowed down cell cycle as it is seen for Met5A, MSTO-211H, and DM-3 and other cells lines (41). There are at least two reasons for this: first, the H2052 cells did not grow more slowly upon BAP1 knockdown and are still more resistant toward cisplatin upon BAP1 knockdown. Second, the Met5A cells display an extend of growth delay similar to the ones of the MSTO-211H and DM-3 cells (Fig. 4E) and at the same time, show the least increase in insensitivity upon BAP1 knockdown (Fig. 4A). Hence, we concluded that there must be another factor explaining the observed chemoresistance.

As BAP1 was shown to play a role in cell death (22, 58), we examined the levels of apoptosis. As expected, upon BAP1 knockdown and cisplatin treatment, we found reduced levels of the apoptosis markers, cleaved caspase-3 and cleaved PARP1, in the Met5A and MSTO-211H cell lines. In a study by Bononi and colleagues, using hemizygously BAP1-deleted fibroblasts, it was also shown that reduced levels of BAP1 led to impaired apoptosis, measured by reduced levels of cleaved caspase-3 (22). When we looked at total caspase-3 levels, we surprisingly also found them to be reduced in all cell lines tested. We, therefore, hypothesized that the total caspase-3 levels could be transcriptionally regulated. Indeed, in all cell lines we found strong downregulation of caspase-3 mRNA levels, and only the mesothelial cell line, Met5A, was able to restore some caspase-3 mRNA upon cisplatin treatment. This indicates that the apoptosis inhibiting effect could be enhanced in malignant cells. Taken together, we concluded that downregulation of apoptosis due to loss of BAP1 is not exclusively regulated by its cytoplasmic function.

As caspase-3 levels were shown to be directly regulated by E2F1 (45, 46), we considered E2F1 as a candidate transcription factor mediating the BAP1-regulated expression of apoptosis genes. E2F1 was shown to mediate the DDR and induction of apoptosis through interaction with HCF1 by controlling transcription of apoptotic genes (47). In the same study, the authors detected downregulation of TP73 and E2F1 upon HCF1 knockdown and concurrent upregulation of CDKN1A(p21) expression, which we could also detect in our experiments. The majority of BAP1 binds to HCF1 (19, 39) and BAP1 binds E2F1 responsive promotors of cell-cycle–regulating genes (44). On the basis of these observations, we think that regulation of apoptosis and response to cisplatin-based chemotherapy are mediated by interplay of BAP1, HCF1, and E2F1. This may also happen directly, as shown by our immunoprecipitation experiments. Interestingly, we could also detect lower p53 levels upon BAP1 knockdown (Fig. 5A). E2F1 was shown to interact with p53, stabilizing it, and biasing it toward apoptosis (59), which substantiates our hypothesis of BAP1 and E2F1 interaction in mediating apoptosis.

Platinum-based chemotherapy has been the approved standard of care for first-line unresectable MPM since 2004. Recently, ipilimumab plus nivolumab has been FDA approved as first-line treatment for adults with unresectable MPM. Data have been presented during the 2020 International Association for the Study of Lung Cancer's World Conference on Lung Cancer Virtual Presidential Symposium (11). A prespecified interim analysis of the phase III CheckMate 743 study of patients with previously untreated, unresectable MPM showed a 26% reduction in the risk of death with nivolumab plus ipilimumab versus chemotherapy (HR, 0.74; P = 0.002). At 2 years, 41% of patients treated with dual checkpoint inhibition were alive compared with 27% who received chemotherapy. This recent success of immune checkpoint inhibitors has renewed interest in immunotherapies, and in combining them with chemotherapy to achieve additive or synergistic clinical activity. Two major ways that chemotherapy promotes tumor immunity are (i) by inducing immunogenic cell death as part of its intended therapeutic effect and (ii) by disrupting strategies that tumors use to evade immune recognition (reviewed in ref. 60). Data about the interplay between chemotherapy and immunotherapy in MPM are missing. This is precisely why knowledge of the mechanisms of chemotherapy resistance is so important, especially because the group of patients with MPM not responding to immunotherapy may still receive chemotherapy.

In this study, we showed that somatic BAP1 mutations occur early in MPM development and are, therefore, present in the majority of tumor cells. Furthermore, we demonstrated that alterations in BAP1 lead to resistance to cisplatin. This implies that a large number of patients do not acquire the chemoresistance during treatment, but are intrinsically resistant. Therefore, loss of BAP1 probably leads to genetic instability, reduced DDR, and less apoptosis, ensuring a high genetic diversity, probably at the expense of slower tumor growth (61). This combination of slow growth, chemoresistance, and high heterogeneity will complicate the search of new therapeutic options. However, screening for BAP1 alterations could be used as a negative predictive marker for chemotherapy response, sparing probable nonresponding patients the serious side effects of a chemotherapy. We hope that the negative predictive impact of BAP1 loss will be verified in a large international cohort study, paving the way for its implication in the diagnostic routine for patient's benefit.

M.B. Kirschner reports grants from Polianthes Foundation and Innovation Pool, University Hospital Zurich during the conduct of the study. M. Meerang reports grants from Polianthes Foundation and Innovation Pool, University Hospital Zurich during the conduct of the study. W. Weder reports grants from Polianthes Foundation and Innovation Pool Grant during the conduct of the study. I. Opitz reports personal fees from AstraZeneca and Roche and grants from Medtronic outside the submitted work. P.J. Wild reports grants from Innovation Pool, University Hospital Zurich and Polianthes Foundation during the conduct of the study, and consulting fees and honoraria (personal/institutional) for lectures by Bayer, Janssen-Cilag, Novartis, Roche, MSD, Astellas Pharma, Bristol Myers Squibb, Thermo Fisher Scientific, Molecular Health, Sophia Genetics, Qiagen, and Astra Zeneca. No disclosures were reported by the other authors.

K. Oehl: Data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. B. Vrugt: Resources, data curation, formal analysis, validation, investigation, writing–original draft. U. Wagner: Data curation, formal analysis, visualization, writing–original draft. M.B. Kirschner: Resources, data curation, validation, investigation, writing–original draft. M. Meerang: Resources, data curation. W. Weder: Conceptualization, resources, project administration. E. Felley-Bosco: Resources, data curation. B. Wollscheid: Conceptualization, data curation, formal analysis, supervision, validation, methodology. K. Bankov: Writing–review and editing. M.C. Demes: Writing–review and editing. I. Opitz: Conceptualization, resources, data curation, funding acquisition, project administration. P.J. Wild: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing– original draft, project administration, writing–review and editing.

We thank Ruedi Aebersold for his helpful comments. We thank Susanne Dettwiler, Fabiola Prutek, and Peter Schraml from the Tissue Biobank at USZ, and Katrin Bankov from the Senckenberg Biobank at UKF for their excellent technical support. Martina Friess is thanked for providing the clinical data and Thomas Frauenfelder for the radiologic assessment of the patients. We also thank the KS Aarau, KS Winterthur, KS Baselland, KS Luzern, Stadtspital Triemli, Institute of Pathology Enge, Institute of Pathology of the University of Bern, and KS Zug for the great collaboration. This project was funded, in part, by Innovation Pool Grants provided by the University Hospital Zurich (to P.J. Wild) and by a Polianthes Foundation grant (to I. Opitz).

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.

1.
Carbone
M
,
Harbour
JW
,
Brugarolas
J
,
Bononi
A
,
Pagano
I
,
Dey
A
, et al
Biological mechanisms and clinical significance of bap1 mutations in human cancer
.
Cancer Discov
2020
;
10
:
1103
20
.
2.
Jaurand
MC
,
Fleury-Feith
J
. 
Pathogenesis of malignant pleural mesothelioma
.
Respirology
2005
;
10
:
2
8
.
3.
Frank
AL
,
Joshi
TK
. 
The global spread of asbestos
.
Ann Global Health
2014
;
80
:
257
62
.
4.
Lemen
RA
,
Landrigan
PJ
. 
Toward an asbestos ban in the United States
.
Int J Environ Res Public Health
2017
;
14
1302
.
5.
Robinson
BM
. 
Malignant pleural mesothelioma: an epidemiological perspective
.
Ann Cardiothorac Surg
2012
;
1
:
491
6
.
6.
Comertpay
S
,
Pastorino
S
,
Tanji
M
,
Mezzapelle
R
,
Strianese
O
,
Napolitano
A
, et al
Evaluation of clonal origin of malignant mesothelioma
.
J Transl Med
2014
;
12
:
301
.
7.
Opitz
I
. 
Management of malignant pleural mesothelioma—the European experience
.
J Thorac Dis
2014
;
6
:
S238
S52
.
8.
Kostron
A
,
Friess
M
,
Inci
I
,
Hillinger
S
,
Schneiter
D
,
Gelpke
H
, et al
Propensity matched comparison of extrapleural pneumonectomy and pleurectomy/decortication for mesothelioma patients
.
Interact Cardiovasc Thorac Surg
2017
;
24
:
740
6
.
9.
Van Schil
PE
,
Opitz
I
,
Weder
W
,
De Laet
C
,
Domen
A
,
Lauwers
P
, et al
Multimodal management of malignant pleural mesothelioma: where are we today?
Eur Respir J
2014
;
44
:
754
64
.
10.
Vogelzang
NJ
,
Rusthoven
JJ
,
Symanowski
J
,
Denham
C
,
Kaukel
E
,
Ruffie
P
, et al
Phase III study of pemetrexed in combination with cisplatin versus cisplatin alone in patients with malignant pleural mesothelioma
.
J Clin Oncol
2003
;
21
:
2636
44
.
11.
Baas
PS
,
Scherpereel
A
,
Nowak
A
,
Fujimoto
N
,
Peters
S
,
Tsao
A
, et al
First-line nivolumab + ipilimumab vs chemotherapy in unresectable malignant pleural mesothelioma: CheckMate 743
.
J Thorac Oncol
2020
;
15
:
E42
.
12.
Oehl
K
,
Kresoja-Rakic
J
,
Opitz
I
,
Vrugt
B
,
Weder
W
,
Stahel
R
, et al
Live-cell mesothelioma biobank to explore mechanisms of tumor progression
.
Front Oncol
2018
;
8
:
40
.
13.
Hylebos
M
,
Van Camp
G
,
van Meerbeeck
JP
,
Op de Beeck
K
. 
The genetic landscape of malignant pleural mesothelioma: results from massively parallel sequencing
.
J Thorac Oncol
2016
;
11
:
1615
26
.
14.
Testa
JR
,
Cheung
M
,
Pei
J
,
Below
JE
,
Tan
Y
,
Sementino
E
, et al
Germline BAP1 mutations predispose to malignant mesothelioma
.
Nat Genet
2011
;
43
:
1022
5
.
15.
Rai
K
,
Pilarski
R
,
Cebulla
CM
,
Abdel-Rahman
MH
. 
Comprehensive review of BAP1 tumor predisposition syndrome with report of two new cases
.
Clin Genet
2016
;
89
:
285
94
.
16.
Carbone
M
,
Arron
ST
,
Beutler
B
,
Bononi
A
,
Cavenee
W
,
Cleaver
JE
, et al
Tumour predisposition and cancer syndromes as models to study gene-environment interactions
.
Nat Rev Cancer
2020
;
20
:
533
49
.
17.
Sahtoe
DD
,
van Dijk
WJ
,
Ekkebus
R
,
Ovaa
H
,
Sixma
TK
. 
BAP1/ASXL1 recruitment and activation for H2A deubiquitination
.
Nat Commun
2016
;
7
:
10292
.
18.
Scheuermann
JC
,
de Ayala Alonso
AG
,
Oktaba
K
,
Ly-Hartig
N
,
McGinty
RK
,
Fraterman
S
, et al
Histone H2A deubiquitinase activity of the polycomb repressive complex PR-DUB
.
Nature
2010
;
465
:
243
7
.
19.
Yu
H
,
Mashtalir
N
,
Daou
S
,
Hammond-Martel
I
,
Ross
J
,
Sui
G
, et al
The ubiquitin carboxyl hydrolase BAP1 forms a ternary complex with YY1 and HCF-1 and is a critical regulator of gene expression
.
Mol Cell Biol
2010
;
30
:
5071
85
.
20.
Yu
H
,
Pak
H
,
Hammond-Martel
I
,
Ghram
M
,
Rodrigue
A
,
Daou
S
, et al
Tumor suppressor and deubiquitinase BAP1 promotes DNA double-strand break repair
.
Proc Natl Acad Sci U S A
2014
;
111
:
285
90
.
21.
Dey
A
,
Seshasayee
D
,
Noubade
R
,
French
DM
,
Liu
J
,
Chaurushiya
MS
, et al
Loss of the tumor suppressor BAP1 causes myeloid transformation
.
Science
2012
;
337
:
1541
6
.
22.
Bononi
A
,
Giorgi
C
,
Patergnani
S
,
Larson
D
,
Verbruggen
K
,
Tanji
M
, et al
BAP1 regulates IP3R3-mediated Ca(2+) flux to mitochondria suppressing cell transformation
.
Nature
2017
;
546
:
549
53
.
23.
Kopecka
J
,
Salaroglio
IC
,
Righi
L
,
Libener
R
,
Orecchia
S
,
Grosso
F
, et al
Loss of C/EBP-β LIP drives cisplatin resistance in malignant pleural mesothelioma
.
Lung Cancer
2018
;
120
:
34
45
.
24.
Moody
HL
,
Lind
MJ
,
Maher
SG
. 
MicroRNA-31 regulates chemosensitivity in malignant pleural mesothelioma
.
Mol Ther Nucleic Acids
2017
;
8
:
317
29
.
25.
Xu
D
,
Liang
SQ
,
Yang
H
,
Luthi
U
,
Riether
C
,
Berezowska
S
, et al
Increased sensitivity to apoptosis upon endoplasmic reticulum stress-induced activation of the unfolded protein response in chemotherapy-resistant malignant pleural mesothelioma
.
Br J Cancer
2018
;
119
:
65
75
.
26.
Mairinger
FD
,
Schmeller
J
,
Borchert
S
,
Wessolly
M
,
Mairinger
E
,
Kollmeier
J
, et al
Immunohistochemically detectable metallothionein expression in malignant pleural mesotheliomas is strongly associated with early failure to platin-based chemotherapy
.
Oncotarget
2018
;
9
:
22254
68
.
27.
Byrne
MJ
,
Nowak
AK
. 
Modified RECIST criteria for assessment of response in malignant pleural mesothelioma
.
Ann Oncol
2004
;
15
:
257
60
.
28.
Li
H
. 
Aligning sequence reads, clone sequences and assembly contigs with BWA-MEM
.
arXiv:1303.3997 [q-bio.GN]
.
29.
Byrne
MJ
,
Nowak
AK
. 
Modified RECIST criteria for assessment of response in malignant pleural mesothelioma
.
Ann Oncol
2004
;
15
:
257
60
.
30.
Remon
J
,
Reguart
N
,
Corral
J
,
Lianes
P
. 
Malignant pleural mesothelioma: new hope in the horizon with novel therapeutic strategies
.
Cancer Treat Rev
2015
;
41
:
27
34
.
31.
Shi
W
,
Ng
CKY
,
Lim
RS
,
Jiang
T
,
Kumar
S
,
Li
X
, et al
Reliability of whole-exome sequencing for assessing intratumor genetic heterogeneity
.
Cell Rep
2018
;
25
:
1446
57
.
32.
Gerstung
M
,
Beisel
C
,
Rechsteiner
M
,
Wild
P
,
Schraml
P
,
Moch
H
, et al
Reliable detection of subclonal single-nucleotide variants in tumour cell populations
.
Nat Commun
2012
;
3
:
811
.
33.
Ventii
KH
,
Devi
NS
,
Friedrich
KL
,
Chernova
TA
,
Tighiouart
M
,
Van Meir
EG
, et al
BRCA1-associated protein-1 is a tumor suppressor that requires deubiquitinating activity and nuclear localization
.
Cancer Res
2008
;
68
:
6953
62
.
34.
Borczuk
AC
,
Pei
J
,
Taub
RN
,
Levy
B
,
Nahum
O
,
Chen
J
, et al
Genome-wide analysis of abdominal and pleural malignant mesothelioma with DNA arrays reveals both common and distinct regions of copy number alteration
.
Cancer Biol Ther
2016
;
17
:
328
35
.
35.
Carbone
M
,
Yang
H
,
Pass
HI
,
Krausz
T
,
Testa
JR
,
Gaudino
G
. 
BAP1 and cancer
.
Nat Rev Cancer
2013
;
13
:
153
9
.
36.
Ismail
IH
,
Davidson
R
,
Gagne
JP
,
Xu
ZZ
,
Poirier
GG
,
Hendzel
MJ
. 
Germline mutations in BAP1 impair its function in DNA double-strand break repair
.
Cancer Res
2014
;
74
:
4282
94
.
37.
Nasu
M
,
Emi
M
,
Pastorino
S
,
Tanji
M
,
Powers
A
,
Luk
H
, et al
High incidence of somatic BAP1 alterations in sporadic malignant mesothelioma
.
J Thorac Oncol
2015
;
10
:
565
76
.
38.
Yoshikawa
Y
,
Emi
M
,
Hashimoto-Tamaoki
T
,
Ohmuraya
M
,
Sato
A
,
Tsujimura
T
, et al
High-density array-CGH with targeted NGS unmask multiple noncontiguous minute deletions on chromosome 3p21 in mesothelioma
.
Proc Natl Acad Sci U S A
2016
;
113
:
13432
7
.
39.
Machida
YJ
,
Machida
Y
,
Vashisht
AA
,
Wohlschlegel
JA
,
Dutta
A
. 
The deubiquitinating enzyme BAP1 regulates cell growth via interaction with HCF-1
.
J Biol Chem
2009
;
284
:
34179
88
.
40.
Sacco
JJ
,
Kenyani
J
,
Butt
Z
,
Carter
R
,
Chew
HY
,
Cheeseman
LP
, et al
Loss of the deubiquitylase BAP1 alters class I histone deacetylase expression and sensitivity of mesothelioma cells to HDAC inhibitors
.
Oncotarget
2015
;
6
:
13757
71
.
41.
Bott
M
,
Brevet
M
,
Taylor
BS
,
Shimizu
S
,
Ito
T
,
Wang
L
, et al
The nuclear deubiquitinase BAP1 is commonly inactivated by somatic mutations and 3p21.1 losses in malignant pleural mesothelioma
.
Nat Genet
2011
;
43
:
668
72
.
42.
Kovac
M
,
Blattmann
C
,
Ribi
S
,
Smida
J
,
Mueller
NS
,
Engert
F
, et al
Exome sequencing of osteosarcoma reveals mutation signatures reminiscent of BRCA deficiency
.
Nat Commun
2015
;
6
:
8940
.
43.
Roos
WP
,
Kaina
B
. 
DNA damage-induced cell death: from specific DNA lesions to the DNA damage response and apoptosis
.
Cancer Lett
2013
;
332
:
237
48
.
44.
Pan
H
,
Jia
R
,
Zhang
L
,
Xu
S
,
Wu
Q
,
Song
X
, et al
BAP1 regulates cell cycle progression through E2F1 target genes and mediates transcriptional silencing via H2A monoubiquitination in uveal melanoma cells
.
Int J Biochem Cell Biol
2015
;
60
:
176
84
.
45.
Hauck
L
,
Kaba
RG
,
Lipp
M
,
Dietz
R
,
von Harsdorf
R
. 
Regulation of E2F1-dependent gene transcription and apoptosis by the ETS-related transcription factor GABPgamma1
.
Mol Cell Biol
2002
;
22
:
2147
58
.
46.
Nahle
Z
,
Polakoff
J
,
Davuluri
RV
,
McCurrach
ME
,
Jacobson
MD
,
Narita
M
, et al
Direct coupling of the cell cycle and cell death machinery by E2F
.
Nat Cell Biol
2002
;
4
:
859
64
.
47.
Tyagi
S
,
Herr
W
. 
E2F1 mediates DNA damage and apoptosis through HCF-1 and the MLL family of histone methyltransferases
.
EMBO J
2009
;
28
:
3185
95
.
48.
Nicer database
. 
Annual incidence rates. Foundation National Institute for Cancer Epidemiology and Registration
. 
2014
.
49.
Opitz
I
,
Friess
M
,
Kestenholz
P
,
Schneiter
D
,
Frauenfelder
T
,
Nguyen-Kim
TDL
, et al
A new prognostic score supporting treatment allocation for multimodality therapy for malignant pleural mesothelioma: a review of 12 years' experience
.
J Thorac Oncol
2015
;
10
:
1634
41
.
50.
Nadeu
F
,
Delgado
J
,
Royo
C
,
Baumann
T
,
Stankovic
T
,
Pinyol
M
, et al
Clinical impact of clonal and subclonal TP53, SF3B1, BIRC3, NOTCH1, and ATM mutations in chronic lymphocytic leukemia
.
Blood
2016
;
127
:
2122
30
.
51.
McGranahan
N
,
Swanton
C
. 
Clonal heterogeneity and tumor evolution: past, present, and the future
.
Cell
2017
;
168
:
613
28
.
52.
Guo
G
,
Chmielecki
J
,
Goparaju
C
,
Heguy
A
,
Dolgalev
I
,
Carbone
M
, et al
Whole-exome sequencing reveals frequent genetic alterations in BAP1, NF2, CDKN2A, and CUL1 in malignant pleural mesothelioma
.
Cancer Res
2015
;
75
:
264
9
.
53.
Zauderer
MG
,
Bott
M
,
McMillan
R
,
Sima
CS
,
Rusch
V
,
Krug
LM
, et al
Clinical characteristics of patients with malignant pleural mesothelioma harboring somatic BAP1 mutations
.
J Thorac Oncol
2013
;
8
:
1430
3
.
54.
McGregor
SM
,
McElherne
J
,
Minor
A
,
Keller-Ramey
J
,
Dunning
R
,
Husain
AN
, et al
BAP1 immunohistochemistry has limited prognostic utility as a complement of CDKN2A (p16) fluorescence in situ hybridization in malignant pleural mesothelioma
.
Hum Pathol
2017
;
60
:
86
94
.
55.
Farzin
M
,
Toon
CW
,
Clarkson
A
,
Sioson
L
,
Watson
N
,
Andrici
J
, et al
Loss of expression of BAP1 predicts longer survival in mesothelioma
.
Pathology
2015
;
47
:
302
7
.
56.
Arzt
L
,
Quehenberger
F
,
Halbwedl
I
,
Mairinger
T
,
Popper
HH
. 
BAP1 protein is a progression factor in malignant pleural mesothelioma
.
Pathol Oncol Res
2014
;
20
:
145
51
.
57.
Bueno
R
,
Stawiski
EW
,
Goldstein
LD
,
Durinck
S
,
De Rienzo
A
,
Modrusan
Z
, et al
Comprehensive genomic analysis of malignant pleural mesothelioma identifies recurrent mutations, gene fusions and splicing alterations
.
Nat Genet
2016
;
48
:
407
.
58.
Sime
W
,
Niu
Q
,
Abassi
Y
,
Masoumi
KC
,
Zarrizi
R
,
Kohler
JB
, et al
BAP1 induces cell death via interaction with 14-3-3 in neuroblastoma
.
Cell Death Dis
2018
;
9
:
458
.
59.
Polager
S
,
Ginsberg
D
. 
p53 and E2f: partners in life and death
.
Nat Rev Cancer
2009
;
9
:
738
48
.
60.
Emens
LA
,
Middleton
G
. 
The interplay of immunotherapy and chemotherapy: harnessing potential synergies
.
Cancer Immunol Res
2015
;
3
:
436
43
.
61.
Jacqueline
C
,
Biro
PA
,
Beckmann
C
,
Moller
AP
,
Renaud
F
,
Sorci
G
, et al
Cancer: a disease at the crossroads of trade-offs
.
Evol Appl
2017
;
10
:
215
25
.

Supplementary data