Tissue samples were obtained from patients treated for MPM at the Department of Medical Oncology and Department of Thoracic Surgery between January 2007 and December 2014. The study has been approved by the Zurich Cantonal Ethics Committee (reference numbers StV 24-2005 and 29-2009). Written informed consent was obtained from all patients. Tumor samples obtained from diagnostic biopsies prior to neoadjuvant chemotherapy and the corresponding resection specimens were processed immediately for total RNA extraction using Qiagen RNAeasy kit (Qiagen, Hilden, Germany). In addition, another part of the tumor samples was embedded in OCT and immediately frozen at −80°C. The remaining tumor samples were fixed in 4% buffered formaldehyde for paraffin embedding.

Primary mesothelioma cell cultures were generated as previously described (11). In order to establish a model for chemoresistance, two primary cultures from chemotherapy-naive epithelioid MPM specimens were continually exposed to low doses of cisplatin and pemetrexed according to a previously published method of cisplatin-resistance in ovarian cancer cell lines (12). These two cultures originated from two MPM patients. The first primary culture was taken from a biopsy of a male 69-year-old patient (P95), who had an epithelioid MPM (Figures S1 and S2 in Supplementary Material) and was enrolled into the SAKK17/04 study (13). The patient, therefore, underwent extrapleural pneumonectomy after neoadjuvant chemotherapy but relapsed 15 months later. He died 27 months after the initial diagnosis. The second culture was established from a female, 50-year-old patient (P236) with papillary epithelioid tumor morphology (Figure S1 in Supplementary Material). The patient underwent partial pleurectomy (Figure S2 in Supplementary Material) but relapsed 19 months later and was operated again, followed by chemotherapy. Unfortunately, the patient died 3 years after the first surgery. In the two chemo-naive primary cultures, cells were treated with cisplatin at the initial dose of 3 and 1 nM pemetrexed. The treatment was conducted in three cycles, including 4 days of treatment and 3 days of recovery time. The concentration was doubled and the procedure was repeated until the concentration of cisplatin and pemetrexed reached 48 and 16 nM, respectively. Chemoresistance induction was evaluated based on the cytotoxicity to a subsequent exposure to cisplatin and pemetrexed. Doubling time was estimated as previously described (14). Primary cells were authenticated by DNA fingerprinting of short tandem repeat loci (Microsynth, Switzerland). Senescence-associated-β-gal activity was determined as previously described (15).

To extract RNA, Qiagen RNeasy kit was used. cDNA was determined as previously described (15). Selected gene expression [primers are listed in Table S1 in Supplementary Material and (15–17) analysis using MIQE compliant protocols (18)] was conducted as previously described.

Whole cellular protein extracts were prepared in 95°C Laemmli sample buffer and mechanically sheared by pressing few times through syringes (26 G). Protein concentration was determined using a Pierce™ 660 nm Protein Assay (Thermo Scientific). A total of 5 µg protein per extract was separated on denaturing 10–20% gradient SDS-PAGE gels. Proteins were transferred on PVDF transfer membranes (0.45 µm, Perkin Elmer). Membranes were probed with the following primary antibodies: anti-calretinin (Sigma, HPA007306), mouse anti-actin (#69100) from MP Biomedicals, N-Cadherin (BD Biosciences, 610920), YAP (Cell Signaling 4912), Mesothelin (Rockland Inc. 200-301-A88), GATA4 (C-20) (Santa Cruz sc-1237), p62 (Progen GP62-C), LC3B (Cell Signaling 2775S), p53 (DO-1) (Santa Cruz sc-126), Thy1 (H-110) (Santa Cruz sc-9163), and γ-H2AX (Millipore 05-636). Membranes were then incubated with the secondary antibody rabbit anti-mouse IgG-HRP (A-5420) from Antell, and goat anti-rabbit IgG-HRP (#7074) from Cell Signaling. The signals were detected by enhanced chemiluminescence (ECL™ Western Blotting Reagents, GE Healthcare) and detected on photosensitive film (Super RX Fuji x-Ray Film, Fujifilm). Proteins were quantitated with densitometry using Image J software (Version 1.42q, USA).

Primary cells were grown on 12-mm glass coverslips in 24-well plates. Cells were fixed for 10 min with 4% paraformaldehyde and permeabilized with 0.05% saponin for 5 min. The cells were then incubated over night at 4°C with Calretinin (N-term) (Sigma HPA007306), Podoplanin (D2-40) (Dako M3619), YAP1 (Cell Signaling 4912), Ki67 (Abcam ab8191), Phalloidin 488 (molecular probes A12379), Thy1 (H-110) (Santa Cruz sc-9163), and N-Cadherin from Dako (M3613) diluted in PBS containing 1% bovine serum albumin. Secondary antibodies, Alexa Fluor 488-conjugated goat anti-rabbit IgG (Life Technologies A11034) and Alexa Fluor 555-conjugated goat anti-mouse IgG (Life Technologies A21424) antibody was added for 1 h at RT. Nuclear DNA was stained using DAPI. Coverslips were mounted using Prolong Gold antifade reagent (Life Technologies). Images were acquired using an Olympus BX61 microscope (Schwerzenbach, Switzerland) equipped with an F-view camera for conventional fluorescence imaging. The image capture was controlled with the AnalySISPro software (Soft Imaging System, Münster, Germany).

Cells resuspended in ice-cold PBS/2mM EDTA were incubated with antibodies against Thy1 (H-110) (Santa Cruz sc-9163) at 4°C for 30 min in the dark. Secondary antibody (PE-labeled, Dianova) was incubated at 4°C for 30 min in the dark. After washing with PBS cells were resuspended in ice-cold PBS/2mM EDTA and fluorescence was measured on an Attune flow cytometer (Applied Biosystems) and analyzed with the Attune cytometric software v1.2.5 (Applied Biosystems).

For DNA isolation from formalin-fixed, paraffin-embedded (FFPE) tissue, two punches of 0.6 mm were taken from the tumor of each FFPE Block. DNA was extracted using the Maxwell 16 FFPE Tissue LEV DNA Purification Kit (Promega) according to the manufacturer’s protocol. Isolated DNA was diluted in nuclease-free water and the concentration was measured by fluorometric quantitation using the Qubit 2.0 fluorometer (Thermo Fisher Scientific).

For targeted amplicon-based sequencing, we developed a custom-designed Ion AmpliSeq MPM Panel (Thermo Fisher Scientific), covering the coding regions of 30 genes that were reported to be commonly mutated in MPM. 10 ng of either FFPE tumor or primary cell DNA were used as an input for library preparation, following the Ion AmpliSeq DNA Library Preparation user guide (Thermo Fisher Scientific). Template preparation and sequencing was performed according to the manufacturer’s protocol (Ion PI Hi-Q OT2 200 Kit User Guide and Ion PI Hi-Q Sequencing 200 Kit, Thermo Fisher Scientific). Alignment, variant calling, and filtering of the resulting data were performed with Ion Reporter 5.0 (Thermo Fisher Scientific). The filtering chain included removal of all synonymous mutations, variants that are found in the UCSC Common SNPs database, variants with an allele ratio below 10% or a read count below 50. Additionally, all mutations were visually inspected, using the Integrative Genomics Viewer Software (Broad Institute).

For array-based genome wide copy number analysis, OncoScan FFPE microarrays (Affymetrix) were conducted at IMGM Laboratories GmbH (Martinsried, Germany). An input amount of 79.2 ng DNA was used for the assay. Molecular inversion probe processing was carried out according to the OncoScan FFPE Assay Kit Protocol (Affymetrix). For data analysis, the AGCC Viewer v4.2.1567 and the OncoScan Console v1.3.0.39 were used (Affymetrix). Copy number variation (CNV) and loss of heterozygosity (LOH) analysis was performed using the Nexus Express for OncoScan v3.1 software. CNVs were visualized using Circos19. Oncoscan have been deposited in GEO (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE109305).

Data are expressed as mean ± SD. Statistical analysis was performed using Mann–Whitney U tests using StatView 5.0.1 (SAS institute) and t-test using GraphPad. Differences were considered statistically significant at p < 0.05.

Between 2007 and 2014, 235 surgical samples were obtained and processed, as described in Figure 1A. After preservation of tumor material for histopathological evaluation, remaining parts of the tumor were processed in order to extract RNA. Derived cDNA was then used for quality control assays, i.e., for calculation of the MPM score (15), allowing to determine that 82% of the samples contained at least 50% tumor cells. In case of large representative tumor tissue, additional parts of the fresh sample were used to establish primary cultures, either in the presence of serum (11) or in the absence of serum (17), for enzymatic analysis (15) or in situ hybridization (17). Successful primary cultures from “bona fide” tumor samples, defined as sufficiently growing cells for collecting RNA and total protein lysates, and preparing frozen stocks, was achieved in 68% of initial samples (159/235).

Primary cultures from two patients (P236 and P95) for whom samples were available at different stages of the disease (for nomenclature defining the different samples from these patients see Figure S2 in Supplementary Material) and where we had observed maintenance of epithelioid histology in one (P236) and epithelial to mesenchymal transition (EMT) in the second (P95), were selected for this proof of principle study. It is worth noting that EMT was evaluated on surgical specimens excluding any diagnostic bias introduced by single biopsy analysis. For these two patients, established primary cultures from different time points were compared with matching initial tumor samples by short tandem repeat analysis (Table S2 in Supplementary Material), Oncoscan CNV testing, and amplicon panel sequencing (Figure 1B). While few genetic alterations were present in P236, consistent with current knowledge on papillary mesothelioma (19), P95 was characterized by a widespread LOH including NF2, BAP1, and TP53 tumor suppressor genes, as well an additional truncating mutation in NF2 (c.553G > T/p.Glu185Ter, no Cosmic annotation) and a point mutation in TP53 (c.839G > C/p.p.Arg280Thr, cosm254987).

N-cadherin, mesothelin, and calretinin MPM biomarker mRNA and protein expression was determined (Figure 1C) and decreasing calretinin expression reflected the evolution of the tumor. A high level of YAP expression was observed in all cultures. Culture heterogeneity was evidenced in one of the patient (P236_cells) by immunofluorescence analysis, revealing that not all cells stained for podoplanin (D2-40), Ki-67, nuclear YAP, or nuclear calretinin, while all cells expressed N-cadherin (Figure 1D). Changes in mesothelin, calretinin and podoplanin gene expression levels were also evaluated in tumor samples (Figure 1E). Differences observed in vivo for selected mRNAs involved either in differentiation (HES1, RUNX2), EMT (GREM1, CTGF), senescence (PUMA, LOX, PAI1), drug transport (SLC22A4, ABCC1), cell cycle (CALB2 (20), CCND1, CDKN2A), response to oxidative stress or hypoxia (HO-1, CAIX, GLUT1), or MPM-associated surface proteins (PDPN, THY1, CD276, TNFRSF10B) were weakly (R² = 0.2641) but significantly correlated with differences observed in primary cultures in patient P236 (Figure 1F). This supports the idea of using primary cultures for the functional investigation of mechanisms underlying tumor evolution.

Malignant pleural mesothelioma chemoresistance models have been described after 96 h exposure of human mesothelioma cell lines to high doses of pemetrexed (9) or after 15 cycles of combination of treatment with IC50 concentrations of cisplatin plus pemetrexed during 72 h followed by a recovery time in a rat cell line (21). We opted for a method that has been used in a study inducing cisplatin-resistance to ovarian cancer cell lines (12) (Figure 2A) to avoid acute induction of senescence-associated growth arrest observed using high doses of pemetrexed (9). Using this approach, we successfully generated a chemoresistant model for primary cells derived from patient P236 (Figure 2B) but not for primary cells derived from patient P95 (Figure S3 in Supplementary Material). This was accompanied by a higher basal chemoresistance in primary cells derived from patient P95 (IC50 2.5 vs. 1.5 µM in the primary culture from patient P236). Doubling time in the cultures exposed or not to cisplatin/pemetrexed did not differ in either of the two patients (Figure S4 in Supplementary Material) excluding that growth kinetics were the underlying mechanism of resistance.

We have previously described the occurrence of senescence in MPM patients treated with chemotherapy (15). Because senescence is a stress response involving multiple effector mechanisms, including DNA damage response (22, 23), senescence-associated secretion phenotype (SASP) (24–28) and autophagy (29), we analyzed markers involved in these pathway. Consistent with development of chemoresistance in the cisplatin/pemetrexed primary culture from patient P236, we observed increased levels of senescence-associated β-gal activity (Figure 3A), autophagy markers LC3B-II (53% increase) and p62 (41% increase) and appearance of SASP-associated (28) GATA-4 (Figure 3B), and H2AX phosphorylation (more than sevenfold increase, Figure 3C). Importantly, basal levels of senescence-associated β-gal activity were higher in patient P95, and did not differ in cisplatin/pemetrexed treated cultures. In the same patient the expression of GATA-4 was already detected in the control primary culture and did not increase further in the cisplatin/pemetrexed culture. Increased levels of the other markers in the cisplatin/pemetrexed compared to control culture were smaller compared to changes observed in patient P236. All in all, this data confirms that chemotherapy can induce resistance via induction of senescence but also indicates that the cancer genetic background per se is likely responsible for senescence and resistance due to genomic instability.

In order to get more insight into events occurring during the development of chemoresistance in primary cultures from patient P236 exposed to cisplatin/pemetrexed, we analyzed the mRNA expression of autophagy/senescence-associated IL-6 and PAI-1 (15, 30) and chemoresistance-development associated Thy1 (31) at three different time points during culture treatment (Figure 4A) and in the original primary cultures. Consistent with increased level of SASP-associated GATA4, we observed upregulation of IL-6 at the first two time points and in the primary culture obtained in the progressing tumor. PAI-1 expression followed a similar pattern. THY1 was decreased at the first time point then it increased, as also observed in the progressing tumor. We had identified Thy1 N-glycoprotein on the cell surface of mesothelioma cells in MPM tumors (32), therefore, we analyzed Thy1 positive cells by immunofluorescence in both, control and cisplatin/pemetrexed adapted primary cultures and original primary cultures from the two specimens obtained from the tumor. We observed that Thy1 was expressed at increased levels in primary cultures obtained during cisplatin/pemetrexed adaptation (Figure 4B, upper panel) and as well during tumor progression (Figure 4B, lower panel), although to a lower extent. For quantification, we performed flow cytometric analyses (Figure 4C) and observed a 10-fold increase of expression of Thy1 surface protein in cisplatin/pemetrexed exposed primary cells at the third time point, confirming our immunofluorescence data. In addition, we detected Thy1 protein expression by WB (data not shown) in cisplatin/pemetrexed adapted cells. The observation associating higher levels of Thy1 with tumoral cell progression is consistent with Thy1 high expression levels with worst overall survival in TCGA (Figure S5 in Supplementary Material).

In the process of trying to better understand mechanisms regulating Thy1 expression, we realized that we had used primers (15) based on UCSC Genome Browser on Human Feb. 2009 Assembly, which detect both THY1 and long non-coding RNA (lncRNA) RP11-334E6.12, a lncRNA in antisense orientation compared to THY1 and which partially overlaps with THY1 3′UTR. We, therefore, redesigned primers recognizing individually THY1 and lncRNA RP11-334E6.12, and observed a strong correlation between lncRNA RP11-334E6.12 and THY1 expression (Figure 4D), which is consistent with correlated expression between these two gene expression reported in G.Tex (33) (Figure S6 in Supplementary Material). Taken together, this data indicate an enrichment of THY1 positive cells during chemoresistance development and a possible role for lnc RP11-334E6.12 in such an increase.