Integrative Molecular Characterization of Malignant Pleural

Author Manuscript Published OnlineFirst on October 15, 2018; DOI: 10.1158/2159-8290.CD-18-0804 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

1 Integrative Molecular Characterization of Malignant Pleural

2 Mesothelioma 3 4 Julija Hmeljak1,37#, Francisco Sanchez-Vega2#, Katherine A. Hoadley3, Juliann Shih4, 5 Chip Stewart3, David Heiman3, Patrick Tarpey5, Ludmila Danilova6, Esther Drill7, Ewan 6 A. Gibb8, Reanne Bowlby9, Rupa Kanchi10, Hatice U. Osmanbeyoglu11, Yoshitaka 7 Sekido12, Jumpei Takeshita13, Yulia Newton14, Kiley Graim14, Manaswi Gupta4, Carl M. 8 Gay15, Lixia Diao10, David L. Gibbs16, Vesteinn Thorsson16, Lisa Iype16, Havish 9 Kantheti17, David T. Severson18, Gloria Ravegnini19, Patrice Desmeules20, Achim A. 10 Jungbluth21, William D. Travis21, Sanja Dacic22, Lucian R. Chirieac23, Françoise 11 Galateau-Sallé24, Junya Fujimoto25, Aliya N. Husain26, Henrique C. Silveira27, Valerie W. 12 Rusch28, Robert C. Rintoul29, Harvey Pass30, Hedy Kindler31, Marjorie G. Zauderer32, 13 David J. Kwiatkowski33, Raphael Bueno18, Anne S. Tsao15, Jenette Creaney34, Tara 14 Lichtenberg35, Kristen Leraas35, Jay Bowen35, TCGA Research Network, Ina Felau36, 15 Jean Claude Zenklusen36, Rehan Akbani10, Andrew D. Cherniack4, Lauren A. Byers18, 16 Michael S. Noble4, Jonathan A. Fletcher22, Gordon Robertson9, Ronglai Shen7, Hiroyuki 17 Aburatani13, Bruce W. Robinson34*, Peter Campbell5*, Marc Ladanyi1*$ 18 19 #joint first authors 20 *TCGA analysis working group co-chairs 21 $corresponding author, lead contact 22 23 Author affiliations 24 1 Molecular Diagnostics Service, Memorial Sloan Kettering Cancer Center, 1275 York 25 Ave., 10065, New York, NY. 26 27 2 Marie-Josée and Henry R. Kravis Center for Molecular Oncology, Memorial Sloan 28 Kettering Cancer Center, 1275 York Ave., 10065, New York, NY. 29 30 3 Department of Genetics, Lineberger Comprehensive Cancer Center, University of 31 North Carolina at Chapel Hill, Chapel Hill, NC 27599, USA. 32 33 4 The Eli and Edythe L. Broad Institute of Massachusetts Institute of Technology and 34 Harvard University, Cambridge, MA 02142, USA. 35 36 5 Cancer Genome Project, Wellcome Trust Sanger Institute. Hinxton, Cambridgeshire, 37 UK.

Downloaded from cancerdiscovery.aacrjournals.org on September 27, 2021. © 2018 American Association for Cancer Research. Author Manuscript Published OnlineFirst on October 15, 2018; DOI: 10.1158/2159-8290.CD-18-0804 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

1 2 6 The Sidney Kimmel Comprehensive Cancer Center at Johns Hopkins University, 3 Baltimore, MD 21287, USA. 4 5 7 Department of Epidemiology and Biostatistics, Memorial Sloan Kettering Cancer 6 Center, New York, NY 10065, USA 7 8 8 GenomeDx Biosciences, Vancouver, Canada. 9 10 9 Canada’s Michael Smith Genome Sciences Centre, BC Cancer Agency, Vancouver, BC 11 V5Z 1L3, Canada 12 13 10 Department of Bioinformatics and Computational Biology, The University of Texas 14 MD Anderson Cancer Center, Houston, TX 77030, USA. 15 16 11 Computational Systems Biology, Memorial Sloan Kettering Cancer Center, 1275 17 York Ave., 10065, New York, NY. 18 19 12 Division of Cancer Biology, Aichi Cancer Center Research Institute, Nagoya, Aichi, 20 Japan. 21 22 13 Genome Science Division, The University of Tokyo, Tokyo, Japan. 23 24 14 Department of Biomolecular Engineering and Center for Biomolecular Science and 25 Engineering, University of California, Santa Cruz, Santa Cruz, CA 95064, USA 26 27 15 Department of Thoracic Head and Neck Medical Oncology, The University of Texas 28 MD Anderson Cancer Center, Houston, TX 77030, USA. 29 30 16 Institute for Systems Biology, Seattle, WA. 31 32 17 The University of Texas at Dallas 33 34 18 Division of Thoracic Surgery, The Lung Center and International Mesothelioma 35 Program, Brigham and Women's Hospital, Boston, MA. 36 37 19 Department of Pharmacy and Biotechnology, University of Bologna, Bologna, Italy. 38 39 20 Department of Pathology, Quebec Heart and Lung Institute, Quebec, Canada. 40 41 21 Department of Pathology, Memorial Sloan Kettering Cancer Center, 1275 York Ave., 42 10065, New York, NY. 43 44 22 Department of Pathology, University of Pittsburgh Medical Center, Pittsburg, PA. 45 46 23 Department of Pathology, Brigham and Women's Hospital, Boston, MA.

1 2 24 MESOPATH. Cancer Center Leon Berard, Lyon, France. 3 4 25 Department of Translational Molecular Pathology, The University of Texas MD 5 Anderson Cancer Center, Houston, TX 77030, USA. 6 7 26 Department of Pathology, University of Chicago, Chicago, IL. 8 9 27 Molecular Oncology Research Center, Barretos Cancer Hospital, Barretos, Sao Paulo, 10 Brazil. 11 12 28 Department of Surgery, Memorial Sloan Kettering Cancer Center, 1275 York Ave., 13 10065, New York, NY. 14 15 29 Department of Oncology, University of Cambridge, Cambridge, UK. 16 17 30 Department of Cardiothoracic Surgery, NYU Langone Medical Center, New York, 18 NY. 19 20 31 Department of Medicine, Section of Hematology/Oncology, University of Chicago 21 Medical Center and Biological Sciences, Chicago, IL. 22 23 32 Department of Medicine, Memorial Sloan Kettering Cancer Center, 1275 York Ave., 24 10065, New York, NY. 25 26 33 Division of Pulmonary Medicine, Brigham and Women's Hospital, Boston, MA. 27 28 34 School of Medicine and Pharmacology, University of Western Australia, Nedlands, 29 Australia. 30 31 35 The Research Institute at Nationwide Children's Hospital, Columbus, OH, USA. 32 33 36 National Cancer Institute, Bethesda, MD, USA. 34 35 37 Currently affiliated with The Company of Biologists, Histon CB24 9 LF, UK. 36 37 38 Running Title 39 Integrative Molecular Characterization of Mesothelioma 40 41 42 Keywords 43 Mesothelioma, TCGA, BAP1, SETDB1, VISTA, MTAP

1 2 3 4 5 Financial Support 6 This work was supported by the following grants from the USA National Institutes of 7 Health: U54 HG003273, U54 HG003067, U54 HG003079, U24 CA143799, U24 8 CA143835, U24 CA143840, U24 CA143843, U24 CA143845, U24 CA143848, U24 9 CA143858, U24 CA143866, U24 CA143867, U24 CA143882, U24 CA143883, U24 10 CA144025, P30 CA016672 and P30 CA008748. Whole exome sequencing and analysis 11 was funded by the British Lung Foundation grant APG11-7 (‘The genomic architecture of 12 mesothelioma’). 13 14 Contact Information 15 Marc Ladanyi 16 Department of Pathology, Room S-801 17 Phone: 212-639-6369 18 Fax: 212-717-3515 19 [email protected] 20 21 Conflict of Interest 22 Andrew Cherniak and Matthew Meyerson received funding from Bayer AG. 23 Ewan Gibb is an employee of GenomeDX Biosciences, INC. Harvey Pass’ 24 spouse is on the Speakers Board for Genentec, and Genentec has performed 25 molecular analyses of a separate MPM cohort for Dr. Pass’ work. 26 27 28 Total Figures + Tables: 7 29 Word Count: 5120 (approx) 30 31

1 Abstract 2 Malignant pleural mesothelioma (MPM) is a highly lethal cancer of the lining of the chest 3 cavity. To expand our understanding of MPM, we conducted a comprehensive integrated 4 genomic study, including the most detailed analysis of BAP1 alterations to date. We 5 identified histology-independent molecular prognostic subsets, and defined a novel 6 genomic subtype with TP53 and SETDB1 mutations and extensive loss of 7 heterozygosity. We also report strong expression of the immune checkpoint gene VISTA 8 in epithelioid MPM, strikingly higher than in other solid cancers, with implications for the 9 immune response to MPM and for its immunotherapy. Our findings highlight new 10 avenues for further investigation of MPM biology and novel therapeutic options. 11 12 13 Statement of Significance 14 15 Through a comprehensive integrated genomic study of 74 malignant pleural 16 mesotheliomas (MPM), we provide a deeper understanding of histology-independent 17 determinants of aggressive behavior, define a novel genomic subtype with TP53 and 18 SETDB1 mutations and extensive loss of heterozygosity and discovered strong 19 expression of the immune checkpoint gene VISTA in epithelioid MPM. 20 21

1 Introduction 2 3 Malignant pleural mesothelioma (MPM) is a cancer of the mesothelial cells lining 4 the pleural cavity. It was rare until the widespread use of asbestos in the mid-20th 5 century (1). Although reduction and strict regulation of asbestos use may be leading to a 6 leveling off in new cases in Western countries, its long latency, together with continued 7 use of asbestos in non-Western countries, ensures that MPM remains a global problem 8 (2). MPM is almost universally lethal, with only modest survival improvements in the past 9 decade (3), suggesting that standard treatment is reaching a therapeutic plateau. 10 Elucidating oncogenic genomic alterations in MPM is therefore essential for therapeutic 11 progress (4-7). Frequent copy number loss and recurrent somatic mutations in BAP1, 12 NF2, CDKN2A, and others have been identified (4-6), yet no targeted therapies 13 exploiting these alterations have emerged. 14 15 To expand our understanding of the molecular landscape and biological subtypes 16 of MPM, and provide insights that could lead to novel therapies, we have conducted a 17 comprehensive, multi-platform, genomic study of 74 MPM samples, as part of The 18 Cancer Genome Atlas (TCGA). Here, we report how these integrated analyses define 19 prognostic subsets of MPM, characterize a new near-haploid molecular subtype, and 20 identify novel potential therapeutic targets. 21 22

23 Results 24 Cohort Description 25 26 We studied 74 samples of primary MPM from patients with no prior systemic 27 therapy. This cohort was predominantly male (82%), with a median age of 64 years, and 28 tumors were of mostly epithelioid histology (70%), a typical profile for MPM. Asbestos 29 exposure history was positive in 62%, negative in 18%, and unavailable or unknown in 30 the remainder. Demographic and clinical details are provided in Supplementary Tables 31 S1A and S1B, as well as Supplementary Fig. S1. 32 We performed comprehensive molecular profiling, including exome sequencing, 33 copy number arrays (Supplementary Fig. S2), mRNA sequencing (Supplementary 34 Fig. S3), non-coding RNA profiling, DNA methylation (Supplementary Fig. S4) and

1 reverse-phase protein arrays (RPPA; Supplementary Fig. S5). Methods and detailed 2 results of individual analyses are provided in Supplementary Sections 1-13. 3 4 Landscape of somatic mutations and copy number alterations 5 6 Whole exome sequencing (WES) revealed a somatic mutation rate of <2 non- 7 synonymous mutations per megabase in all samples except for an outlier case with a 8 mutation rate of 8 non-synonymous mutations per megabase (Fig. 1A). This places 9 MPM at the low end of somatic mutation burden among cancers (8). The outlier tumor 10 with a ten-fold higher mutation rate showed a distinctive pattern of C>T mutations 11 occurring almost exclusively at CpG dinucleotides (Fig. 1B). This relatively 12 hypermutated tumor harbored a homozygous nonsense mutation in MSH2, which would 13 suggest that the tumor lacked mismatch repair capacity but the observed mutational 14 spectrum was atypical for mismatch repair deficiency (9). Otherwise, the observed 15 mutational spectrum was similar across patients and lacked distinctive or novel features 16 (Fig 1C). Signatures of smoking- or APOBEC-induced mutagenesis were not observed. 17 Asbestos has been proposed to cause genotoxicity via DNA breaks and secondary 18 oxidative damage (10). However, the mutational spectrum and local sequence context 19 was not significantly different between cases with or without known asbestos exposure 20 (chi-squared, P=0.3); while this negative finding should be viewed with caution given the 21 limitations of the dataset, it is in agreement with prior studies (5). 22 23 We sought to identify genes mutated significantly above the background rate 24 using MutSig2CV. Significantly mutated genes (SMGs) at an FDR < 0.05 included BAP1, 25 NF2, TP53, LATS2 and SETD2, all known cancer genes in MPM (Fig. 1A, 26 Supplementary Table S1C). All five genes showed high rates of nonsense, frameshift 27 and splice site mutations, highlighting them as targets of inactivation consistent with their 28 functions as tumor suppressors. Mutations in these five SMGs did not show associations 29 with histories of asbestos exposure or smoking. For validation in an independent cohort 30 that also used a different algorithm to define SMGs, we compared the results of our 31 SMG analysis with the SMGs previously identified in 99 MPM exomes using the MUSIC 32 algorithm (5) and found a strong overlap for SMGs at an FDR < 0.05 in both studies 33 (Fig. 1D). Notably, among lower confidence SMGs (FDR < 0.15 but > 0.05), only

1 SETDB1, which may define a novel subtype of MPM (discussed below), was identified in 2 both analyses (Fig. 1D). 3 4 The somatic copy number alteration (SCNA) landscape was characterized by 5 frequent recurring focal and arm-level deletions, but no recurring amplifications, 6 consistent with the notion that MPM development is driven primarily by loss of tumor 7 suppressor genes, rather than by activation of classic oncogenic driver genes 8 (Supplementary Section 4, Fig. S2). Focal deletions affected several tumor suppressor 9 genes known to be altered in MPM (5), most notably CDKN2A with deletions (defined as 10 deep, likely homozygous) in 36 (49%) and losses (defined as shallow, possibly single- 11 copy) in 5 (7%) samples. Likewise, NF2 deletions were confirmed in 25 (34%) and 12 losses in 30 (40%) samples; as many of the latter harbored mutations of the remaining 13 allele, evidence of biallelic NF2 inactivation was common (Fig 1A). CDKN2A deletions 14 often encompass MTAP, the adjacent gene on 9p21 (11), which encodes 15 methylthioadenosine phosphorylase, whose deficiency has recently been reported to 16 lead to reduced PRMT5 enzymatic activity and heightened sensitivity to its 17 pharmacologic inhibition (12,13). Co-deletion of CDKN2A and MTAP, associated with 18 low levels of mRNA expression of both genes, was observed in 20 cases 19 (Supplementary Fig. S2). Whilst no correlation with overall survival was observed for 20 BAP1 status, loss of CDKN2A was strongly associated with shorter overall survival (Cox 21 proportional hazards, P=7.3x10-6), as previously shown (14,15). 22 23 Finally, an analysis for genomically-integrated viral sequences, including SV40, 24 was negative (Supplementary Section 8), as was a screen of exome and RNA 25 sequencing data for evidence of EWSR1 and ALK fusions, recently reported in rare 26 cases of MPM (16,17) and peritoneal mesothelioma (18), respectively. As well, no 27 activating mutations in the canonical MAPK or PI3/AKT signaling pathways were 28 identified in this cohort. 29 30 Comprehensive analysis of BAP1 status in MPM 31 32 BAP1, encoding a nuclear deubiquitinase, is the most frequently mutated cancer 33 gene in MPM, both in our dataset and others (5), and is also recurrently inactivated in 34 clear cell renal carcinoma, uveal melanoma, and cholangiocarcinoma (19). Somatic

1 BAP1 mutations have germline counterparts that define the BAP1 hereditary cancer 2 syndrome (20). In contrast to other cancers, BAP1 inactivation in MPM does not 3 correlate with adverse outcomes. Because its role in MPM biology remains unclear, we 4 compared BAP1 inactivated and wild-type cases across multiple platforms. First, to 5 better segregate cases according to BAP1 status, we performed a comprehensive 6 analysis of inactivating alterations through a detailed review of SNVs, small and large 7 indels, whole gene deletions, and structural variants; this showed the overall prevalence 8 of BAP1 alterations to be 57% (Fig. 1A, Supplementary Table S2A), in line with recent 9 studies (21). Most MPM with inactivating mutations in BAP1 (25/26, 96%) also had 10 concurrent loss of heterozygosity on chromosome 3p21.1, supporting a classic two-hit 11 tumor suppressor mechanism. Overall, BAP1 status was defined as follows: 32 samples 12 with no evidence of BAP1 alteration, 6 with a single (heterozygous) mutation or deletion 13 and 36 with biallelic inactivation. No germline mutations in BAP1 were identified in this 14 cohort. 15 16 BAP1 alterations showed non-random patterns of co-occurrence with mutations 17 in other key MPM cancer genes (Fig. 2A). As expected, mRNA expression levels of 18 BAP1 itself were reduced in the presence of genomic BAP1 inactivation or loss (Fig 2B). 19 We also observed an inverse correlation of BAP1 alterations and SCNAs. BAP1-altered 20 tumors had fewer chromosome arm gains and losses (Fig. 2C, median 9.5 vs. 15.5, 21 Mann-Whitney, P<0.01), as well as fewer focal SCNAs (Supplementary Fig. S2). 22 23 Since BAP1-mediated deubiquitination of histones and transcriptional proteins is 24 thought to regulate gene expression, we compared mRNA expression patterns between 25 tumors that were wild-type for BAP1 and those with inactivation of one or both BAP1 26 alleles. We identified 1,324 differentially expressed genes with a FDR <0.01, of which 27 75% were downregulated in the BAP1-inactivated samples. As previously noted in 28 experimental models (22), BAP1-inactivated tumors in our cohort had lower mRNA 29 expression of several HOXA genes, including HOXA5 and HOXA6, but no difference in 30 EZH2 mRNA expression or its PRC2 partners EED and SUZ12. 31 32 As BAP1 is known to regulate the ubiquitination and hence stability of several 33 transcriptional proteins, we examined the association of BAP1 status with inferred 34 activity of transcription factors (TF) using a recently developed computational strategy

1 that integrates phosphoproteomic and transcriptomic data with predicted transcription 2 factor binding sites (23). We used this algorithm to assess significant differences in 3 inferred TF activities between BAP1-inactivated and wild-type MPM (satisfying FDR- 4 corrected P<0.01, t-test), linking BAP1 status to altered activity of TFs. Indeed, many 5 TFs identified in this analysis had highly significant associations with BAP1 status (Fig. 6 2D and 2E; Supplementary Table S2B). 7 8 In particular, YY1 had a significantly reduced inferred activity in BAP1-inactivated 9 samples (Fig 2F). BAP1 forms a ternary complex with HCF1 and YY1, which acts as 10 transcriptional repressor of genes involved in cell proliferation (24). YY1 has also been 11 shown to regulate or interact with other TFs whose inferred activity was also altered in 12 our analysis, such as MAX and EGR2. Moreover, there is evidence that YY1 interacts 13 with EZH2, the catalytic subunit of PRC2, and is required for its function in gene 14 silencing (25). Thus, to a degree, the transcriptional consequences of BAP1 inactivation 15 could be attributed to changes in recruitment of specific TFs to their target genes, either 16 directly, as in the loss of ternary YY1-HCF1-BAP1 complexes, or indirectly, as in the 17 case of TFs that are themselves downstream targets of YY1 (Supplementary Table 18 S2C). 19 20 Notably, BAP1-inactivated samples demonstrated a significantly increased 21 inferred activity of IRF8 (Fig. 2G, Supplementary Table S2D), a hematopoietic lineage 22 TF involved in interferon signaling and dendritic cell differentiation. Both pathways were 23 upregulated in our BAP1-based supervised gene mRNA expression analysis (see 24 above). Moreover, PARADIGM analysis confirmed that the IRF activation pathways are 25 upregulated in BAP1-inactivated samples. IRF8 has been implicated in the biology of 26 CD103-positive dendritic cells whose antigen-presenting function is highly effective at 27 stimulating cytotoxic T cells in the tumor microenvironment (26). Taken together, these 28 observations suggest an association between BAP1 status and perturbed immune 29 signaling in MPM that warrants further exploration. 30

1 MPM with genomic near-haploidization, a novel subtype 2 3 Allele-specific copy number analysis of WES data using the FACETS algorithm 4 (27) identified three cases with striking genome-wide LOH that affected more than 80% 5 of the genome, which was confirmed in SNP6.0 copy number array data (Fig. 3A and 6 3B, Supplementary Fig. S2). This phenomenon, well-described in acute lymphoblastic 7 leukemia (28), occurs when the nascent cancer clone loses one copy of nearly all 8 chromosomes (which we term ‘genomic near-haploidization’, GNH) followed by 9 duplication of the remaining chromosomes. To better assess the prevalence of this 10 unusual genomic phenomenon in MPM, we screened 80 samples from a Japanese 11 International Cancer Genome Consortium (ICGC) MPM cohort, identifying two additional 12 cases (Fig. 3B), representing a combined prevalence of 5/154 (3.2%). Neither the three 13 TCGA cases, nor the two ICGC cases, had deletions or point mutations in BAP1, 14 PBRM1 or SETD2. Remarkably, all 5 cases had inactivating point mutations in, or 15 homozygous deletion of, SETDB1, which encodes a histone methyltransferase involved 16 in gene silencing, representing the only SETDB1-mutated MPMs in the combined 17 TCGA-ICGC cohort. Additionally, 4 of 5 cases with evidence of GNH had driver 18 mutations in TP53. Although near-haploid cases have been reported in other cancers 19 (e.g. acute lymphoblastic leukemia, chondrosarcoma, adrenocortical carcinoma), the 20 strong association of TP53 and SETDB1 co-mutation with widespread LOH has not 21 been previously reported in other cancers (28,29). 22 23 To further define the clinical features of this novel subset, we obtained data on 16 24 additional cases with near-haploid karyotypes from a large independent series of MPM 25 studied by cytogenetic analysis of short-term primary cultures (Supplementary Section 26 2.2; Note: cytogenetic analysis can only identify genomic near-haploidization cases 27 which have not undergone genome duplication). Although most of the genome in our 5 28 cases showed LOH, chromosomes 5 and 7 strikingly retained heterozygosity (Fig. 3C). 29 These two chromosomes also remained disomic in the cases with near-haploid 30 karyotypes in the independent cytogenetic series (Supplementary Table S3). Based on 31 this combined set of 21 near-haploid MPM, a distinctive profile of genomic and clinical 32 features emerged, demarcating a novel molecular subtype of MPM. Strikingly, most 33 patients with MPM showing genomic near-haploidization were female (M:F=1:4), in 34 sharp contrast to the overall TCGA-ICGC MPM cohort, where males greatly

1 predominated (M:F>4:1; p<0.01, Fisher’s exact test). No difference in distribution of 2 histological types was observed in the GNH subset. 3 4 Since MPM with evidence of GNH have typically undergone a genome-doubling 5 event, we were able to assess the timing of this relative to point mutations (30,31). 6 Essentially, mutations acquired before genome duplication would be present on both 7 copies of the duplicated chromosome, whereas those occurring after would be 8 heterozygous. In all three TCGA cases with genomic near-haploidization, the majority of 9 the point mutations were present on both copies of duplicated chromosomes, suggesting 10 the genome duplication occurred relatively late in the evolution of the cancer. In three 11 cases, the TP53 driver mutation was homozygous, suggesting it occurred before 12 genome doubling. In contrast, the SETDB1 mutations were present on only one of the 13 two copies of chromosome 1 in two of the three cases, suggesting they occurred after 14 LOH and genome duplication (Fig. 3D). Overall, these data suggest a model in which 15 TP53 mutations occur early, and presumably permit the steady (or catastrophic) loss of 16 chromosomes. It seems likely that genome reduplication occurs after achieving near- 17 haploidy, with haploinsufficiency of SETDB1 arising later. 18 19 Integrative Multi-Platform Analysis Defines Novel Prognostic Subsets 20 21 While the current classification of MPM into epithelioid, sarcomatoid and biphasic 22 histologies is prognostically useful, there remains variability in clinical features and 23 patient outcomes within histological subtypes. Previous analyses (5,7) based on mRNA 24 expression alone have defined unsupervised clusters largely recapitulating these 25 histologic distinctions. To find out whether multi-platform molecular profiling might 26 provide additional resolution to define prognostic subsets of MPM, we performed 27 integrative clustering across multiple assay platforms using two algorithms: iCluster (32) 28 and PARADIGM (33). Both identified four distinct integrated subtypes of MPM. There 29 was a strong concordance in subtype assignments between the two algorithms (Fig. 4A, 30 Supplementary Figs. S6, S7), especially for the best (cluster 1) and worst (cluster 4) 31 prognosis clusters, indicating that integration of molecular data can identify distinct 32 subgroups of MPM, independent of the specific statistical methodology. Survival was 33 significantly different across the 4 clusters (P<0.0001 for either algorithm; Fig. 4B, 34 Supplementary Fig. 7). This survival difference remained significant (P=0.008) after

1 adjusting for histology (epithelioid vs. non-epithelioid) and CDKN2A homozygous 2 deletion (Fig. 4C, Supplementary Fig. S6), a known molecular prognostic factor in 3 MPM (14,15). While they did provide additional independent prognostic information, the 4 iClusters nonetheless did show correlation with consensus histology (p=0.002), with 5 iCluster cluster 1 being enriched for epithelioid histology (similar finding for Paradigm 6 cluster 1). They also exhibited differences in immune cell infiltrates (Supplementary 7 Table S4A), as discussed below, but showed no significant correlation with clinical 8 variables such as T stage, N stage, asbestos exposure history, or smoking 9 (Supplementary Table S4B). Molecularly, these tumors had low SCNA, relatively few 10 CDKN2A homozygous deletions (11%), and a high level of methylation (Supplementary 11 Fig. S4). All but one (95%) had BAP1 alterations: 26% had homozygous deletions and 12 53% had heterozygous loss with mutations. 13 14 The poor prognosis cluster (cluster 4; red) had a high score for epithelial- 15 mesenchymal transition (EMT) based on gene expression (P<0.001; Figs. 4D), which 16 was distinguished by high mRNA expression of VIM, PECAM1 and TGFB1, and low 17 miR-200 family expression. These tumors also displayed MSLN promoter methylation 18 and consequent low mRNA expression of mesothelin, a marker of differentiated 19 mesothelial cells, as noted previously in sarcomatoid MPM and the sarcomatoid 20 components of biphasic MPM (15,34). Overall, this poor prognosis cluster also showed 21 enrichment of LATS2 mutations (30% compared to 4% in the rest of the cohort) and 22 CDKN2A homozygous deletions (66%). Moreover, this cluster showed higher AURKA 23 mRNA expression, higher leukocyte fraction (based on methylation), and elevated 24 mRNA expression of E2F targets, G2M checkpoints, and DNA damage response genes. 25 PI3K-mTOR and RAS/MAPK signaling was upregulated, based on both mRNA and 26 protein expression (Supplementary Fig. S7). Additionally, several miRNAs were 27 differentially expressed between the good and poor prognostic clusters, including miR- 28 193a-3p, which has been proposed as a potential tumor suppressor (35). Finally, a 29 comparison of immune gene mRNA expression signatures (36) across the four clusters 30 revealed a significantly higher score for the Th2 cell signature in the poor prognosis 31 Cluster 4 compared to the other clusters (Fig. 4E, Supplementary Table S4A). 32 Coincidentally, it has been reported that Th2 cytokines secreted by immune cells upon 33 exposure to asbestos may promote MPM (37). The analyses of other immune signatures 34 are shown in Supplementary Fig. 5.

1 While biphasic and sarcomatoid MPM are more aggressive, there remains a 2 need for improved risk stratification of epithelioid MPM, for which clinical outcomes are 3 more heterogeneous. Therefore, we conducted an integrative clustering analysis 4 restricted to epithelioid MPM. The results for the 4-cluster epithelioid-only solution were 5 highly similar to the 4-cluster all-MPM solution (Fig. 5A), with only 7 of the 52 epithelioid 6 samples reassigned to other clusters. This stability indicates that the features driving the 7 all-MPM clustering are largely independent of histology. The epithelioid-only clusters 8 share many of the features defining the corresponding clusters in the all-MPM solution 9 (Fig. 5B). The survival analysis also paralleled the all-MPM solution, with cluster 1 10 having the best outcomes and cluster 4 having the worst (Fig. 5C). PARADIGM analysis 11 of the epithelioid-only subset confirmed upregulation of AURKA mRNA expression in the 12 poor prognosis epithelioid-only cluster 4 (Fig. 5D), corroborating the results from the all- 13 MPM analysis. 14 15 Finally, we sought to independently validate the clinical correlations of clusters 16 identified in the TCGA epithelioid cases using mRNA expression profiles from two 17 published studies: 211 MPM analyzed by RNA-sequencing (5) and 52 MPM samples 18 analyzed by mRNA expression microarrays (14). Specifically, we assigned each mRNA 19 expression profile to one of the integrative clusters based on the rules derived from the 20 TCGA mRNA dataset. For the larger validation cohort (henceforth referred to as Bueno), 21 we restricted our analysis to epithelioid samples and used the epithelioid-only gene 22 signature to cluster samples. We found that the epithelioid-only samples assigned to 23 iCluster 1 (good prognosis) had significantly better survival, even after adjusting for age 24 (P= 3.9x10-4; Fig. 5E and Supplementary Fig. S6). In the smaller cohort (referred to as 25 Lopez-Rios), patient numbers were too small to split by histology. However, this analysis 26 provided independent validation of the survival differences for the four all-MPM clusters 27 (P=0.01; Supplementary Fig. S6). Taken together, these results suggest that the 28 prognostically relevant molecular profiles defined by our analysis are robust and 29 reproducible, and could be potentially used to improve risk stratification of patients with 30 epithelioid MPM. The core mRNA gene lists are provided in Supplementary Table 4C- 31 G which also include reduced classifiers based on methylation, lncRNA, miRNA, and a 32 reduced classifier combining mRNA and methylation data to facilitate practical 33 application of these data and independent validation of these clusters. 34 35

1 Analysis of noncoding RNAs in MPM 2 3 We next assessed two types of noncoding RNAs not extensively studied in MPM 4 that may also represent a source of robust biomarkers: lncRNAs, which often show 5 higher expression specificity for cell type than coding genes, and miRNAs, which are 6 relatively stable in biological fluids, so potentially suitable as noninvasive biomarkers. 7 8 In both the TCGA and Bueno (5) cohorts, lncRNAs returned four unsupervised 9 consensus subtypes that were associated with 5-year survival (P=1.4x10-10 and 10 P=1.1x10-3 respectively). The TCGA lncRNA subtypes were highly concordant with 11 iCluster (P=8x10-14) and PARADIGM (P=2x10-21) subtypes, and were associated with 12 EMT scores (P=2.5x10-6) (Figs. 4A, 6A-E; Supplementary Table S5A.1). For both 13 cohorts, lncRNAs that were differentially abundant between the good-prognosis subtype 14 and other samples included those associated with cancer (e.g. H19, LINC00152, 15 MEG3), or with MPM in particular: NEAT1 and SNHG8 (38), and GAS5 (39) (Fig. 6F,G; 16 Supplementary Table S5B.3-6). We noted that a number of lncRNAs distinguished the 17 good-prognosis cluster in both TCGA and Bueno cohorts (Supplementary Table S5B.1- 18 2,5-6). 19 20 Unsupervised clustering based on miRNA mature strands resolved five 21 consensus subtypes in the TCGA cohort (Fig. 6H); these were associated with 5-year 22 survival (P=7x10-4), iCluster (P=1x10-12) and PARADIGM (P=3x10-21) clusters, and 23 associated with EMT scores (P=2x10-7) (Fig. 6H-K; Supplementary Table S5A.2). For 24 the good-survival cases, the miRNA subtypes were strikingly concordant across multiple 25 analysis platforms, and many cancer-associated miRNAs were differentially abundant in 26 the good-survival cluster (Fig. 6L; Supplementary Table S5B.3-4). Taken together, 27 these results suggest that lncRNAs and miRNAs may be important predictors of survival 28 in MPM. 29 30 Epithelial-Mesenchymal Transition and VISTA expression 31 32 Because mRNA expression of EMT-associated genes was a key differentiating 33 feature between prognostically distinct integrative clusters, we performed a detailed 34 analysis of EMT in MPM using a previously published EMT-related mRNA signature

1 (40). Increasing EMT scores significantly correlated with clusters defined by integrative 2 algorithms (iCluster and PARADIGM), as well as with individual genomic features 3 including miRNA, lncRNA, methylation, RPPA and overall gene expression (Fig 7A). Of 4 all tumor types included in the TCGA pan-cancer analysis, MPM had the second-highest 5 average EMT score (Fig. 7B) (after soft tissue sarcomas), consistent with previous 6 reports that EMT is a frequent phenomenon in MPM (7,41). 7 8 While EMT score positively correlated with the mRNA expression of many 9 immune regulatory genes, such as OX40L, TGFB1, CD276, OX40 and PD-L2 (p<0.001) 10 (Fig. 7A and 7C), mRNA expression of VISTA (42), a negative immune checkpoint 11 regulator primarily expressed on hematopoietic cells (43), was strongly inversely 12 correlated with EMT score (Fig. 7C), and was expressed at levels higher than in any 13 other TCGA tumor type analyzed (Fig. 7D). In the MPM cohort, VISTA mRNA levels 14 were highest in the epithelioid subtype (Fig. 7C, E). Using Regulome Explorer 15 (Supplementary Section 13), we found that VISTA mRNA expression was highly 16 correlated with mesothelin mRNA expression (Spearman correlation=0.81; P=6.3x10-19), 17 but not with mRNA expression of PD-1 or PD-L1. Moreover, there was no significant 18 correlation between overall mutation burden and VISTA expression levels (P=0.64). 19 20 VISTA (V-domain Ig suppressor of T cell activation; also known as c10orf54, PD- 21 1H, and B7-H5) is a member of the B7 family of negative checkpoint regulators, 22 expressed on the surface of several immune cell types. It can function both as receptor 23 and ligand (44). As a ligand, VISTA is present on the surface of antigen presenting cells 24 (APC), and inhibits early-stage T cell activation (45). The normal mesothelium has APC 25 properties (46), which are retained upon malignant transformation (47,48). We thus 26 performed immunohistochemical staining of two epithelioid TCGA MPM cases, as well 27 as normal and reactive mesothelium (Supplementary Section 14), to define the cellular 28 compartment expressing VISTA in MPM tumor samples. Remarkably, VISTA protein 29 expression was not restricted to infiltrating immune cells, but was present in tumor cells 30 in MPM, as well as in normal and reactive mesothelium (Fig 7F,G), suggesting that its 31 expression in epithelioid MPM may reflect retention of APC properties in this more 32 differentiated subset of MPM. 33

1 Discussion 2 3 Our comprehensive integrative analysis of 74 cases of MPM further defines a cancer 4 driven primarily by inactivation of tumor suppressor genes. Indeed, we confirm the high 5 frequency of BAP1 inactivation by mutation and copy number loss, as well as recurrent 6 inactivating alterations in CDKN2A, NF2, TP53, LATS2 and SETD2. In addition to this 7 landscape of known loss of function events, we have genomically characterized a novel 8 molecular subtype of MPM accounting for approximately 3% of MPM in our datasets, 9 defined by evidence of genomic near haploidization and recurrent TP53 and SETDB1 10 mutations, with a distinctive clinical phenotype showing female predominance and 11 younger age at diagnosis. Our findings should facilitate systematic clinical studies of this 12 subset to better define its survival and its association with asbestos exposure, which so 13 far appears weak or unclear. Isolated cases with similar molecular profiles (genomic 14 near-haploidization and SETDB1 mutation) have been anecdotally reported (5,49), but 15 had not, until now, been recognized as a distinct molecular subtype of MPM. The 16 genomic data in our cases support a model in which early TP53 mutations are 17 permissive for a loss of chromosomes to a near-haploid state, followed by genome 18 reduplication and SETDB1 inactivation. As H3K9 methylation by SETDB1 is a repressive 19 chromatin mark (50), it is tempting to speculate that inactivation of SETDB1 allows a 20 general increase in transcription activity in a cancer cell that has sacrificed nearly half of 21 its genome, presumably including several genes that are imprinted or otherwise 22 monoallelically expressed. Why chromosomes 5 and 7 are spared from the LOH in these 23 MPM remains mysterious – interestingly, the same two chromosomes have recently 24 been shown to also retain heterozygosity in thyroid Hürthle cell carcinomas with 25 evidence of genomic near haploidization (51). Overall, the identification of this novel 26 subset of MPM highlights how, as the proportion of asbestos-related MPM plateaus and 27 hopefully begins declining in Western countries, cases of possibly different etiology may 28 become more apparent. 29 30 A better understanding of the determinants of aggressive behavior and predictors of 31 poor clinical outcomes in MPM remains an unmet clinical need. To this end, integrative 32 analyses with iCluster and PARADIGM revealed a set of molecular features that define 33 MPM subsets with better and worse prognosis, and might point to candidate therapeutic 34 targets. For instance, cases in the poor prognosis subset showed higher aurora kinase A

1 mRNA expression, consistent with prior studies (14,52). Treatment of MPM cell lines 2 with an aurora kinase inhibitor leads to growth arrest (53), and several aurora kinase 3 inhibitors are under investigation in MPM patients. AURKA mRNA expression could be 4 used to help identify poor prognosis epithelioid MPM patients so that they could be 5 directed towards experimental therapies early in their treatment course. It is presently 6 unclear whether aurora kinase inhibitors will be active in MPM and the ongoing phase II 7 study of alistertib (NCT02293005) includes all patients irrespective of molecular 8 signature. While the value of targeting the AURKA pathway with currently available 9 clinical compounds in MPM remains unknown, mRNA expression of AURKA is a 10 prognostic marker that could be used in clinical practice to help stratify patients with 11 epithelioid MPM. 12 Additionally, the poor prognosis group also exhibited upregulation of PI3K and 13 mTOR signaling pathways. Preclinical studies have reported this finding, leading to 14 clinical trials with a low response rate (2%) and no significant survival benefit in the 15 salvage setting (54). These data suggest that combination therapies with mTOR 16 inhibition may be necessary. Our analysis of epigenetic alterations revealed some 17 associations between BAP1 status and DNA methylation (Supplementary Table S6). A 18 recently established functional link between BAP1 and EZH2 (22) provided the rationale 19 for the recently completed clinical trial of the EZH2 inhibitor tazemetostat in BAP1-null 20 MPM (NCT028602). By contrast, no relevant findings resulted from our analysis of viral 21 and microbial sequences (Supplementary Table S7). Finally, another genomically- 22 driven potential vulnerability in MPM is very frequent CDKN2A deletion associated with 23 co-deletion of MTAP, the latter recently shown to metabolically lead to impaired activity, 24 and therefore sensitization to further inhibition, of the arginine methyltransferase PRMT5 25 (12,13), an inhibitor of which is currently being evaluated in a phase I trial 26 (NCT02783300). 27 28 Harnessing current immunotherapy approaches to improve outcomes of patients with 29 MPM is an area of intense clinical interest. While our study confirms that MPMs show a 30 low tumor mutation burden and therefore may present a more challenging setting for 31 immunotherapy, a remarkable and novel finding of the present study is that of strong 32 expression of the immune checkpoint gene VISTA in epithelioid MPM, on the tumor cells 33 themselves, unlike other cancer types where it is more often expressed on infiltrating 34 reactive cells (42). VISTA is a member of the B7 family of negative checkpoint regulators

1 that is expressed primarily on infiltrating tumor macrophages, and whose immune 2 restraining effects on T cells may be similar to those of PD1 (43,55). However, unique 3 structural features mean that VISTA can repress activation of T cells both as a ligand 4 present on the surface of APC cells, and as a receptor on the surface of T cells (56). 5 Because VISTA is expressed on MPM cells, and its mRNA expression levels do not 6 correlate with overall mutation load, our results raise the possibility that VISTA 7 expression may be restraining anti-tumor immune responses in a subset of MPM cases. 8 As we find that non-neoplastic mesothelium also expresses VISTA protein, we speculate 9 that VISTA expression in MPM is retained or possibly selected for by immune pressure. 10 This is consistent with previous publications demonstrating that MPM is an 11 ‘immunogenic’ tumor (57), including recent trials showing some responses to immune 12 checkpoint blockade therapy (58,59). Indeed, VISTA has recently been reported as a 13 possible compensatory immune inhibitory pathway in prostate cancers that fail to 14 respond to ipilimumab (60). Our findings thus provide both a rationale and a candidate 15 biomarker for clinical trials of emerging anti-VISTA therapy (42,44) (NCT02812875) in 16 epithelioid MPM. Taken together, our findings point to new lines of investigation into the 17 biology of MPM with the potential to lead to new therapeutic strategies. 18 19 Materials and Methods 20 21 TCGA Project Management has collected necessary human subjects documentation to 22 ensure the project complies with 45-CFR-46 (the “Common Rule”). The program has 23 obtained documentation from every contributing clinical site to verify that IRB approval 24 has been obtained to participate in TCGA. Such documented approval may include one 25 or more of the following: 26 27 ● An IRB-approved protocol with Informed Consent specific to TCGA or a 28 substantially similar program. In the latter case, if the protocol was not TCGA- 29 specific, the clinical site PI provided a further finding from the IRB that the 30 already-approved protocol is sufficient to participate in TCGA. 31 ● A TCGA-specific IRB waiver has been granted. 32 ● A TCGA-specific letter that the IRB considers one of the exemptions in 45-CFR- 33 46 applicable. The two most common exemptions cited were that the research 34 falls under 46.102(f)(2) or 46.101(b)(4). Both exempt requirements for informed 35 consent, because the received data and material do not contain directly 36 identifiable private information.

1 ● A TCGA-specific letter that the IRB does not consider the use of these data and 2 materials to be human subjects research. This was most common for collections 3 in which the donors were deceased. 4 5 6 A detailed description of the sample acquisition and pathology review process, as well as 7 the experimental and computational methods used for the different analyses presented in 8 our study is provided as Supplementary Sections 1-13. 9 10 Author Contributions 11 12 Co-chairs: Marc Ladanyi, Peter J. Campbell, Bruce W. S. Robinson 13 Manuscript Coordinator: Julija Hmeljak 14 Analysis Coordinator: Ina Felau 15 Data Coordinator: Francisco Sanchez-Vega 16 Project Manager: Ina Felau 17 Copy-number: Andrew D. Cherniack, Juliann Shih 18 mRNA: Katherine A. Hoadley, 19 Methylation: Luda Danilova 20 miRNA: Reanne Bowlby, Gordon Robertson, Ewan Gibb 21 Exomes: Manaswi Gupta, Chip Stewart, Patrick Tarpey 22 RPPA: Rupa Kanchi, Rehan Akbani 23 Transcription factor analysis: Hatice U. Osmanbeyoglu 24 25 Integrative analyses: 26 PARADIGM: Yulia Newton, Kiley Graim, 27 Firehose: David Heiman, Michael Noble 28 iCluster: Esther Drill, Ronglai Shen; David T. Severson, Raphael Bueno 29 Regulome explorer: Lisa Iype, Vésteinn Thorsson, David L. Gibbs, 30 EMT analysis: Lauren A. Byers, Carl Gay, Lixia Dao 31 32 Pathology Subgroup: William D. Travis, Sanja Dacic, Lucian Chirieac, Françoise 33 Galateau-Sallé, Junya Fujimoto, Aliya Husain, Marc Ladanyi 34 VISTA IHC: Patrice Desmeules, Achim Jungbluth 35 Clinical Subgroup: Marjorie Zauderer, Anne Tsao, David J. Kwiatkowski, Hedy Kindler, 36 Harvey Pass, Jenette Creaney, Valerie Rusch, Robert C. Rintoul, Tara Lichtenberg, 37 Bruce W. S. Robinson, Nathan Pennell, Henrique Silveira, 38 Genomic near-haploidization validation datasets: Gloria Ravegnini, Jonathan 39 Fletcher, Hiroyuki Aburatani 40 41 The TCGA research network contributed collectively to this study. M.L., B.W.R.S. and 42 P.J.C. oversaw data analyses and interpretation. J.H., F.S.-V., M.L., P.J.C., M.Z., R.S., 43 B.W.S.R. and A.T. wrote the manuscript. 44 45 The members of The Cancer Genome Atlas Research Network are Hiroyuki Aburatani, 46 Rehan Akbani, Adrian Ally, Pavana Anur, Joshua Armenia, J. Todd Auman, Miruna 47 Balasundaram, Saianand Balu, Stephen B. Baylin, Carmen Behrens, Rameen 48 Beroukhim, Craig Bielski, Tom Bodenheimer, Moiz S. Bootwalla, Jay Bowen, Reanne 49 Bowlby, Denise Brooks, Raphael Bueno, Lauren Averett Byers, Flávio M. Cárcano,

1 Rebecca Carlsen, Andre L. Carvalho, Andrew D. Cherniack, Dorothy Cheung, Lucian 2 Chirieac, Juok Cho, Eric Chuah, Sudha Chudamani, Carrie Cibulskis, Leslie Cope, 3 Daniel Crain, Jenette Creaney, Erin Curley, Sanja Dacic, Ludmila Danilova, Assunta De 4 Rienzo, Timothy DeFreitas, John A. Demchok, Noreen Dhalla, Lixia Diao, Esther Drill, 5 Ina Felau, Michael Feldman, Martin L. Ferguson, Jonathan A. Fletcher, Junya Fujimoto, 6 Junya Fujimoto, Shiro Fukuda, Stacey B. Gabriel, Francoise Galateau Sallé, Jianjiong 7 Gao, Johanna Gardner, Julie M. Gastier-Foster, Carl M. Gay, Nils Gehlenborg, Mark 8 Gerken, Gad Getz, Ewan A. Gibb, David L. Gibbs, Chandra Goparaju, Kiley Graim, 9 Benjamin Gross, Guangwu Guo, Manaswi Gupta, Seiki Hasegawa, David Haussler, D. 10 Neil Hayes, David I. Heiman, Zachary Heins, Julija Hmeljak, Katherine A. Hoadley, 11 Robert A. Holt, Alan P. Hoyle, Aliya Husain, Carolyn M. Hutter, Lisa Iype, Stuart R. 12 Jefferys, Steven J.M. Jones, Corbin D. Jones, Rupa S. Kanchi, Katayoon Kasaian, 13 Jaegil Kim, Hedy Kindler, Nobuyuki Kondo, Thomas Krausz, Ritika Kundra, Kozo 14 Kuribayashi, David J. Kwiatkowski, Marc Ladanyi, Phillip H. Lai, Peter W. Laird, Michael 15 S. Lawrence, Darlene Lee, Kristen M. Leraas, Tara M. Lichtenberg, Pei Lin, Jia Liu, 16 Wenbin Liu, Eric Minwei Liu, Laxmi Lolla, Adhemar Longatto-Filho, Yiling Lu, Yussanne 17 Ma, Dennis T. Maglinte, David Mallory, Marco A. Marra, Michael Mayo, Sam Meier, 18 Jonathan Melamed, Shaowu Meng, Matthew Meyerson, Piotr A. Mieczkowski, Gordon 19 B. Mills, Richard A. Moore, Cesar Moran, Scott Morris, Lisle E. Mose, Andrew J. 20 Mungall, Karen Mungall, Takashi Nakano, Rashi Naresh, Yulia Newton, Michael S. 21 Noble, Angelica Ochoa, Hatice Osmanbeyoglu, Joel S. Parker, Harvey I. Pass, Joseph 22 Paulauskis, Nathan A. Pennell, Robert Penny, Charles M. Perou, Todd Pihl, Nilsa C. 23 Ramirez, Doris M. Rassl, Gloria Ravegnini, Glen Reid, Rui M. Reis, Sheila M. Reynolds, 24 David Rice, William G Richards, Robert C. Rintoul, Jeffrey Roach, A. Gordon Robertson, 25 Valerie Rusch, Sara Sadeghi, Gordon Saksena, Francisco Sanchez-Vega, Chris 26 Sander, Ayuko Sato, Cristovam Scapulatempo-Neto, Jacqueline E. Schein, Nikolaus 27 Schultz, Steven E. Schumacher, Tanguy Seiwert, Yoshitaka Sekido, David T Severson, 28 Candace Shelton, Troy Shelton, Ronglai Shen, Robert Sheridan, Yan Shi, Juliann Shih, 29 Yuichi Shiraishi, Ilya Shmulevich, Henrique C. S. Silveira, Janae V. Simons, Payal 30 Sipahimalani, Tara Skelly, Heidi J. Sofia, Matthew G. Soloway, Paul Spellman, Chip 31 Stewart, Josh Stuart, Qiang Sun, Jumpei Takeshita, Angela Tam, Donghui Tan, Roy 32 Tarnuzzer, Kenji Tatsuno, Barry S Taylor, Nina Thiessen, Eric Thompson, Vesteinn 33 Thorsson, William D. Travis, Anne Tsao, Tohru Tsujimura, Federico Valdivieso, David J. 34 Van Den Berg, Nico van Zandwijk, Umadevi Veluvolu, Luciano S. Viana, Douglas Voet, 35 Yunhu Wan, Zhining Wang, Jing Wang, Joellen Weaver, John N. Weinstein, Daniel J. 36 Weisenberger, Matthew D. Wilkerson, Lisa Wise, Ignacio Wistuba, Tina Wong, Ye Wu, 37 Shogo Yamamoto, Liming Yang, Marjorie G. Zauderer, Jean C. Zenklusen, Jiashan 38 Zhang, Hailei Zhang, Hongxin Zhang, and Erik Zmuda. 39 40 41 Acknowledgements 42 43 We thank Drs. Elaine Mardis and Matthew Meyerson for invaluable feedback, and Dr. 44 Lee Spraggon for helpful comments and assistance with figures. 45 46 This work was supported by the following grants from the USA National Institutes of 47 Health: U54 HG003273, U54 HG003067, U54 HG003079, U24 CA143799, U24 48 CA143835, U24 CA143840, U24 CA143843, U24 CA143845, U24 CA143848, U24 49 CA143858, U24 CA143866, U24 CA143867, U24 CA143882, U24 CA143883, U24 50 CA144025, P30 CA016672 and P30 CA008748. Whole exome sequencing and analysis

1 was funded by the British Lung Foundation grant APG11-7 (‘The genomic architecture of 2 mesothelioma’). 3

1 2 REFERENCES 3 4 1. Sekido Y. Molecular pathogenesis of malignant mesothelioma. Carcinogenesis 5 2013;34(7):1413-9 doi 10.1093/carcin/bgt166.

6 2. Leong SL, Zainudin R, Kazan-Allen L, Robinson BW. Asbestos in Asia. 7 Respirology 2015;20(4):548-55 doi 10.1111/resp.12517.

8 3. Yap TA, Aerts JG, Popat S, Fennell DA. Novel insights into mesothelioma 9 biology and implications for therapy. Nat Rev Cancer 2017;17(8):475-88 doi 10 10.1038/nrc.2017.42.

11 4. Bott M, Brevet M, Taylor BS, Shimizu S, Ito T, Wang L, et al. The nuclear 12 deubiquitinase BAP1 is commonly inactivated by somatic mutations and 3p21.1 13 losses in malignant pleural mesothelioma. Nat Genet 2011;43(7):668-72 doi 14 10.1038/ng.855.

15 5. Bueno R, Stawiski EW, Goldstein LD, Durinck S, De Rienzo A, Modrusan Z, et al. 16 Comprehensive genomic analysis of malignant pleural mesothelioma identifies 17 recurrent mutations, gene fusions and splicing alterations. Nat Genet 18 2016;48(4):407-16 doi 10.1038/ng.3520.

19 6. Guo G, Chmielecki J, Goparaju C, Heguy A, Dolgalev I, Carbone M, et al. Whole- 20 exome sequencing reveals frequent genetic alterations in BAP1, NF2, CDKN2A, 21 and CUL1 in malignant pleural mesothelioma. Cancer Res 2015;75(2):264-9 doi 22 10.1158/0008-5472.CAN-14-1008.

23 7. de Reynies A, Jaurand MC, Renier A, Couchy G, Hysi I, Elarouci N, et al. 24 Molecular classification of malignant pleural mesothelioma: identification of a 25 poor prognosis subgroup linked to the epithelial-to-mesenchymal transition. Clin 26 Cancer Res 2014;20(5):1323-34 doi 10.1158/1078-0432.CCR-13-2429.

27 8. Martincorena I, Campbell PJ. Somatic mutation in cancer and normal cells. 28 Science 2015;349(6255):1483-9 doi 10.1126/science.aab4082.

29 9. Alexandrov LB, Jones PH, Wedge DC, Sale JE, Campbell PJ, Nik-Zainal S, et al. 30 Clock-like mutational processes in human somatic cells. Nat Genet 31 2015;47(12):1402-7 doi 10.1038/ng.3441.

32 10. Chew SH, Toyokuni S. Malignant mesothelioma as an oxidative stress-induced 33 cancer: An update. Free Radic Biol Med 2015;86:166-78 doi 34 10.1016/j.freeradbiomed.2015.05.002.

35 11. Illei PB, Rusch VW, Zakowski MF, Ladanyi M. Homozygous deletion of CDKN2A 36 and codeletion of the methylthioadenosine phosphorylase gene in the majority of 37 pleural mesotheliomas. Clin Cancer Res 2003;9(6):2108-13.

38 12. Kryukov GV, Wilson FH, Ruth JR, Paulk J, Tsherniak A, Marlow SE, et al. MTAP 39 deletion confers enhanced dependency on the PRMT5 arginine

1 methyltransferase in cancer cells. Science 2016;351(6278):1214-8 doi 2 10.1126/science.aad5214.

3 13. Mavrakis KJ, McDonald ER, 3rd, Schlabach MR, Billy E, Hoffman GR, deWeck A, 4 et al. Disordered methionine metabolism in MTAP/CDKN2A-deleted cancers 5 leads to dependence on PRMT5. Science 2016;351(6278):1208-13 doi 6 10.1126/science.aad5944.

7 14. Lopez-Rios F, Chuai S, Flores R, Shimizu S, Ohno T, Wakahara K, et al. Global 8 gene expression profiling of pleural mesotheliomas: overexpression of aurora 9 kinases and P16/CDKN2A deletion as prognostic factors and critical evaluation 10 of microarray-based prognostic prediction. Cancer Res 2006;66(6):2970-9 doi 11 10.1158/0008-5472.CAN-05-3907.

12 15. Dacic S, Kothmaier H, Land S, Shuai Y, Halbwedl I, Morbini P, et al. Prognostic 13 significance of p16/cdkn2a loss in pleural malignant mesotheliomas. Virchows 14 Arch 2008;453(6):627-35 doi 10.1007/s00428-008-0689-3.

15 16. Panagopoulos I, Thorsen J, Gorunova L, Micci F, Haugom L, Davidson B, et al. 16 RNA sequencing identifies fusion of the EWSR1 and YY1 genes in mesothelioma 17 with t(14;22)(q32;q12). Genes, chromosomes & cancer 2013;52(8):733-40 doi 18 10.1002/gcc.22068.

19 17. Desmeules P, Joubert P, Zhang L, Al-Ahmadie HA, Fletcher CD, Vakiani E, et al. 20 A Subset of Malignant Mesotheliomas in Young Adults Are Associated With 21 Recurrent EWSR1/FUS-ATF1 Fusions. The American journal of surgical 22 pathology 2017;41(7):980-8 doi 10.1097/PAS.0000000000000864.

23 18. Hung YP, Dong F, Watkins JC, Nardi V, Bueno R, Dal Cin P, et al. Identification 24 of ALK Rearrangements in Malignant Peritoneal Mesothelioma. JAMA Oncol 25 2018;4(2):235-8 doi 10.1001/jamaoncol.2017.2918.

26 19. Luchini C, Veronese N, Yachida S, Cheng L, Nottegar A, Stubbs B, et al. 27 Different prognostic roles of tumor suppressor gene BAP1 in cancer: A 28 systematic review with meta-analysis. Genes Chromosomes Cancer 29 2016;55(10):741-9 doi 10.1002/gcc.22381.

30 20. Carbone M, Ferris LK, Baumann F, Napolitano A, Lum CA, Flores EG, et al. 31 BAP1 cancer syndrome: malignant mesothelioma, uveal and cutaneous 32 melanoma, and MBAITs. J Transl Med 2012;10:179 doi 10.1186/1479-5876-10- 33 179.

34 21. Nasu M, Emi M, Pastorino S, Tanji M, Powers A, Luk H, et al. High Incidence of 35 Somatic BAP1 alterations in sporadic malignant mesothelioma. J Thorac Oncol 36 2015;10(4):565-76 doi 10.1097/JTO.0000000000000471.

37 22. LaFave LM, Beguelin W, Koche R, Teater M, Spitzer B, Chramiec A, et al. Loss 38 of BAP1 function leads to EZH2-dependent transformation. Nat Med 39 2015;21(11):1344-9 doi 10.1038/nm.3947.

1 23. Osmanbeyoglu HU, Toska E, Chan C, Baselga J, Leslie CS. Pancancer 2 modelling predicts the context-specific impact of somatic mutations on 3 transcriptional programs. Nat Commun 2017;8:14249 doi 10.1038/ncomms14249.

4 24. Yu H, Mashtalir N, Daou S, Hammond-Martel I, Ross J, Sui G, et al. The ubiquitin 5 carboxyl hydrolase BAP1 forms a ternary complex with YY1 and HCF-1 and is a 6 critical regulator of gene expression. Mol Cell Biol 2010;30(21):5071-85 doi 7 10.1128/MCB.00396-10.

8 25. Boyd NH, Morgan JE, Greer SF. Polycomb recruitment at the Class II 9 transactivator gene. Mol Immunol 2015;67(2 Pt B):482-91 doi 10 10.1016/j.molimm.2015.08.003.

11 26. Broz ML, Binnewies M, Boldajipour B, Nelson AE, Pollack JL, Erle DJ, et al. 12 Dissecting the tumor myeloid compartment reveals rare activating antigen- 13 presenting cells critical for T cell immunity. Cancer Cell 2014;26(5):638-52 doi 14 10.1016/j.ccell.2014.09.007.

15 27. Shen R, Seshan VE. FACETS: allele-specific copy number and clonal 16 heterogeneity analysis tool for high-throughput DNA sequencing. Nucleic Acids 17 Res 2016;44(16):e131 doi 10.1093/nar/gkw520.

18 28. Holmfeldt L, Wei L, Diaz-Flores E, Walsh M, Zhang J, Ding L, et al. The genomic 19 landscape of hypodiploid acute lymphoblastic leukemia. Nat Genet 20 2013;45(3):242-52 doi 10.1038/ng.2532.

21 29. Zheng S, Cherniack AD, Dewal N, Moffitt RA, Danilova L, Murray BA, et al. 22 Comprehensive Pan-Genomic Characterization of Adrenocortical Carcinoma. 23 Cancer Cell 2016;29(5):723-36 doi 10.1016/j.ccell.2016.04.002.

24 30. Nik-Zainal S, Van Loo P, Wedge DC, Alexandrov LB, Greenman CD, Lau KW, et 25 al. The life history of 21 breast cancers. Cell 2012;149(5):994-1007 doi 26 10.1016/j.cell.2012.04.023.

27 31. Carter SL, Cibulskis K, Helman E, McKenna A, Shen H, Zack T, et al. Absolute 28 quantification of somatic DNA alterations in human cancer. Nat Biotechnol 29 2012;30(5):413-21 doi 10.1038/nbt.2203.

30 32. Shen R, Olshen AB, Ladanyi M. Integrative clustering of multiple genomic data 31 types using a joint latent variable model with application to breast and lung 32 cancer subtype analysis. Bioinformatics 2009;25(22):2906-12 doi 33 10.1093/bioinformatics/btp543.

34 33. Vaske CJ, Benz SC, Sanborn JZ, Earl D, Szeto C, Zhu J, et al. Inference of 35 patient-specific pathway activities from multi-dimensional cancer genomics data 36 using PARADIGM. Bioinformatics 2010;26(12):i237-45 doi 37 10.1093/bioinformatics/btq182.

38 34. Tan K, Kajino K, Momose S, Masaoka A, Sasahara K, Shiomi K, et al. 39 Mesothelin (MSLN) promoter is hypomethylated in malignant mesothelioma, but

1 its expression is not associated with methylation status of the promoter. Hum 2 Pathol 2010;41(9):1330-8 doi 10.1016/j.humpath.2010.03.002.

3 35. Williams M, Kirschner MB, Cheng YY, Hanh J, Weiss J, Mugridge N, et al. miR- 4 193a-3p is a potential tumor suppressor in malignant pleural mesothelioma. 5 Oncotarget 2015;6(27):23480-95 doi 10.18632/oncotarget.4346.

6 36. Bindea G, Mlecnik B, Tosolini M, Kirilovsky A, Waldner M, Obenauf AC, et al. 7 Spatiotemporal dynamics of intratumoral immune cells reveal the immune 8 landscape in human cancer. Immunity 2013;39(4):782-95 doi 9 10.1016/j.immuni.2013.10.003.

10 37. Maki Y, Nishimura Y, Toyooka S, Soh J, Tsukuda K, Shien K, et al. The 11 proliferative effects of asbestos-exposed peripheral blood mononuclear cells on 12 mesothelial cells. Oncol Lett 2016;11(5):3308-16 doi 10.3892/ol.2016.4412.

13 38. Wright CM, Kirschner MB, Cheng YY, O'Byrne KJ, Gray SG, Schelch K, et al. 14 Long non coding RNAs (lncRNAs) are dysregulated in Malignant Pleural 15 Mesothelioma (MPM). PloS one 2013;8(8):e70940 doi 16 10.1371/journal.pone.0070940.

17 39. Renganathan A, Kresoja-Rakic J, Echeverry N, Ziltener G, Vrugt B, Opitz I, et al. 18 GAS5 long non-coding RNA in malignant pleural mesothelioma. Mol Cancer 19 2014;13:119 doi 10.1186/1476-4598-13-119.

20 40. Mak MP, Tong P, Diao L, Cardnell RJ, Gibbons DL, William WN, et al. A Patient- 21 Derived, Pan-Cancer EMT Signature Identifies Global Molecular Alterations and 22 Immune Target Enrichment Following Epithelial-to-Mesenchymal Transition. Clin 23 Cancer Res 2016;22(3):609-20 doi 10.1158/1078-0432.CCR-15-0876.

24 41. Fassina A, Cappellesso R, Guzzardo V, Dalla Via L, Piccolo S, Ventura L, et al. 25 Epithelial-mesenchymal transition in malignant mesothelioma. Mod Pathol 26 2012;25(1):86-99 doi 10.1038/modpathol.2011.144.

27 42. Lines JL, Sempere LF, Broughton T, Wang L, Noelle R. VISTA is a novel broad- 28 spectrum negative checkpoint regulator for cancer immunotherapy. Cancer 29 immunology research 2014;2(6):510-7 doi 10.1158/2326-6066.CIR-14-0072.

30 43. Nowak EC, Lines JL, Varn FS, Deng J, Sarde A, Mabaera R, et al. 31 Immunoregulatory functions of VISTA. Immunol Rev 2017;276(1):66-79 doi 32 10.1111/imr.12525.

33 44. Deng J, Le Mercier I, Kuta A, Noelle RJ. A New VISTA on combination therapy 34 for negative checkpoint regulator blockade. J Immunother Cancer 2016;4:86 doi 35 10.1186/s40425-016-0190-5.

36 45. Wang L, Rubinstein R, Lines JL, Wasiuk A, Ahonen C, Guo Y, et al. VISTA, a 37 novel mouse Ig superfamily ligand that negatively regulates T cell responses. J 38 Exp Med 2011;208(3):577-92 doi 10.1084/jem.20100619.

1 46. Shaw TJ, Zhang XY, Huo Z, Robertson D, Lovell PA, Dalgleish AG, et al. Human 2 Peritoneal Mesothelial Cells Display Phagocytic and Antigen-Presenting 3 Functions to Contribute to Intraperitoneal Immunity. Int J Gynecol Cancer 4 2016;26(5):833-8 doi 10.1097/IGC.0000000000000697.

5 47. Mutti L, Valle MT, Balbi B, Orengo AM, Lazzaro A, Alciato P, et al. Primary 6 human mesothelioma cells express class II MHC, ICAM-1 and B7-2 and can 7 present recall antigens to autologous blood lymphocytes. Int J Cancer 8 1998;78(6):740-9.

9 48. Creaney J, Dick IM, Leon JS, Robinson BW. A Proteomic Analysis of the 10 Malignant Mesothelioma Secretome Using iTRAQ. Cancer Genomics Proteomics 11 2017;14(2):103-17 doi 10.21873/cgp.20023.

12 49. Kang HC, Kim HK, Lee S, Mendez P, Kim JW, Woodard G, et al. Whole exome 13 and targeted deep sequencing identify genome-wide allelic loss and frequent 14 SETDB1 mutations in malignant pleural mesotheliomas. Oncotarget 15 2016;7(7):8321-31 doi 10.18632/oncotarget.7032.

16 50. Keniry A, Gearing LJ, Jansz N, Liu J, Holik AZ, Hickey PF, et al. Setdb1- 17 mediated H3K9 methylation is enriched on the inactive X and plays a role in its 18 epigenetic silencing. Epigenetics Chromatin 2016;9:16 doi 10.1186/s13072-016- 19 0064-6.

20 51. Ganly I, Makarov V, Deraje S, Dong Y, Reznik E, Seshan V, et al. Integrated 21 genomic analysis of Hürthle Cell cancer reveals oncogenic drivers, recurrent 22 mitochondrial mutations, and unique chromosomal landscapes. Cancer Cell 23 2018;34(2):256-270.e5. doi: 10.1016/j.ccell.2018.07.002. 24 25 52. Suraokar MB, Nunez MI, Diao L, Chow CW, Kim D, Behrens C, et al. Expression 26 profiling stratifies mesothelioma tumors and signifies deregulation of spindle 27 checkpoint pathway and microtubule network with therapeutic implications. Ann 28 Oncol 2014;25(6):1184-92 doi 10.1093/annonc/mdu127.

29 53. Crispi S, Fagliarone C, Biroccio A, D'Angelo C, Galati R, Sacchi A, et al. 30 Antiproliferative effect of Aurora kinase targeting in mesothelioma. Lung Cancer 31 2010;70(3):271-9 doi 10.1016/j.lungcan.2010.03.005.

32 54. Ou SH, Moon J, Garland LL, Mack PC, Testa JR, Tsao AS, et al. SWOG S0722: 33 phase II study of mTOR inhibitor everolimus (RAD001) in advanced malignant 34 pleural mesothelioma (MPM). J Thorac Oncol 2015;10(2):387-91 doi 35 10.1097/JTO.0000000000000360.

36 55. Liu J, Yuan Y, Chen W, Putra J, Suriawinata AA, Schenk AD, et al. Immune- 37 checkpoint proteins VISTA and PD-1 nonredundantly regulate murine T-cell 38 responses. Proc Natl Acad Sci U S A 2015;112(21):6682-7 doi 39 10.1073/pnas.1420370112.

40 56. Flies DB, Han X, Higuchi T, Zheng L, Sun J, Ye JJ, et al. Coinhibitory receptor 41 PD-1H preferentially suppresses CD4(+) T cell-mediated immunity. J Clin Invest 42 2014;124(5):1966-75 doi 10.1172/JCI74589.

1 57. Chee J, Robinson BW, Holt RA, Creaney J. Immunotherapy for Lung 2 Malignancies: From Gene Sequencing to Novel Therapies. Chest 3 2017;151(4):891-7 doi 10.1016/j.chest.2016.10.007.

4 58. Alley EW, Lopez J, Santoro A, Morosky A, Saraf S, Piperdi B, et al. Clinical 5 safety and activity of pembrolizumab in patients with malignant pleural 6 mesothelioma (KEYNOTE-028): preliminary results from a non-randomised, 7 open-label, phase 1b trial. Lancet Oncol 2017;18(5):623-30 doi 10.1016/S1470- 8 2045(17)30169-9.

9 59. Scherpereel A, Mazieres J, Greillier L, Dô P, Bylicki O, Monnet I, et al. Second- 10 or third-line nivolumab (Nivo) versus nivo plus ipilimumab (Ipi) in malignant 11 pleural mesothelioma (MPM) patients: Results of the IFCT-1501 MAPS2 12 randomized phase II trial. Journal of Clinical Oncology 13 2017;35(18_suppl):LBA8507-LBA doi 10.1200/JCO.2017.35.18_suppl.LBA8507.

14 60. Gao J, Ward JF, Pettaway CA, Shi LZ, Subudhi SK, Vence LM, et al. VISTA is an 15 inhibitory immune checkpoint that is increased after ipilimumab therapy in 16 patients with prostate cancer. Nat Med 2017;23(5):551-5 doi 10.1038/nm.4308. 17 18 19 Figure Legends 20 21 Figure 1: Genomic and clinical features of the TCGA MPM cohort. A: iCOMUT 22 plot describing clinical and molecular features of the TCGA MPM cohort. Each column 23 represents an individual case, whilst rows represent clinical and molecular features. Samples 24 are grouped based on MPM histological type. Red arrowhead indicates the hypermutated 25 case. Copy-number alterations are defined as follows: "Deletion" is a deep loss, possibly a 26 homozygous deletion, "Loss" is a shallow loss (possibly heterozygous deletion), "Gain" 27 indicates a low-level gain, and "Amplification" is a high-level amplification. Individual genes 28 shown include significantly mutated genes and selected additional genes of interest. B: 29 Observed mutational spectrum of the hypermutator case (TCGA-UD-AAC1) with 375 non- 30 silent mutations. C: Observed mutational spectrum of the remaining other 73 MPM cases in 31 the TCGA cohort (cohort average: 36 non-silent mutations/patient). D: Comparison of SMGs 32 between the present study and that of Bueno et al, 2016. 33 34 Figure 2: Supervised comparisons between BAP1-inactivated and wild-type 35 MPM. A: BAP1 inactivation by copy number loss or mutation across the cohort, along with 36 mutations in 6 other genes (left). These genes were selected as those with more than 3 non- 37 silent mutated tumors. Large genes with more than 3kb coding regions (TTN, FAT4, MGA) 38 are unlikely to be functional in cancer and were excluded. The bar plot (right) shows the 39 Fisher-exact test p-values for mutual exclusivity and co-occurrence relative to BAP1. Only 40 SETD2 and NF2 approach significant co-occurrence with BAP1 inactivation. The 1-tail Fisher 41 p-values for co-occurrence and the Bejamini-Hochberg FDR’s are p=0.04, FDR=0.15 for 42 SETD2 and p=0.05, FDR=0.15 for NF2. We detected MALAT1 (aka NEAT2) "RNA"

1 mutations in 4 BAP1-inactivated samples, but this does not reach significance (p=0.05, 2 FDR~0.27). B: Normalized BAP1 gene mRNA expression levels in the wild-type, 1 hit and 2 3 hit subgroups. C: Box plot demonstrating a significantly lower frequency of arm-level losses 4 in BAP1-inactivated tumors (BAP1 2) compared to wild-type (BAP1 0). D: Inferred 5 transcription factor (TF) activities significantly associated with BAP1 inactivation (FDR<0.01). 6 E: Volcano plot with mean inferred TF activity difference in BAP1 inactivated and BAP1 wild- 7 type patients plotted on the x-axis, and false discovery rate (FDR)-adjusted significance from 8 t-test plotted on the y-axis (–log10 scale). TFs significantly associated with BAP1 inactivation 9 status (FDR<0.01) are colored in orange. F, G: Box plots with differential inferred activities of 10 YY1 (F) and IRF8 (G), two biologically relevant transcription factors. The target genes on 11 which inferences for YY1 (427 genes) and IRF8 (248 genes) activity were based are 12 provided in Supplementary Table S2. 13 14 Figure 3. MPM cases with genomic near-haploidization. A: Whole exome 15 sequencing-based LOH profiling with the FACETS algorithm revealed three MPM samples 16 with genome-wide LOH. B: Allelic copy number plots of genome-wide LOH cases in the 17 TCGA and Japanese ICGC cohorts. X-axis and Y-axis are the chromosome locations, and 18 the ratio of an allelic copy number of tumor sample to that of matched normal control 19 (lymphocyte), respectively. Red line shows the higher allele and blue line shows the lower 20 allele. C: Near-haploid metaphase cell derived from the BWH genome-wide LOH cohort, 21 stained by Giemsa, showing loss of one copy of all chromosomes except 5, 7 and X. D: 22 Allelic copy number plot of a representative TCGA case with biallelic inactivation of TP53 23 and monoallelic mutation of SETDB1, suggesting the latter occurred after the LOH and 24 genome duplication events. 25 26 Figure 4: Integrative analysis of 74 MPM. A: Concordance between integrative 27 (PARADIGM and iCluster) and platform-specific unsupervised clustering results. Clusters are 28 color-coded and ranked based on survival (dark blue indicates best survival, whilst red and 29 orange marks the worst surviving subgroup). B: Kaplan-Meier plot of the integrative 30 subgroups reveals distinct outcomes. C: Cox regression analysis demonstrates significant 31 associations of the molecular subtypes with patient survival, even upon adjusting for 32 histology, age, and CDKN2A status. D: iCluster identified 4 integrative subgroups with 33 distinct BAP1 alteration (defined as mutation and/or copy number alteration), TP53 mutation, 34 CDKN2A status, copy number alteration, DNA methylation, mRNA, miRNA and lncRNA 35 mRNA expression profiles. E: Comparison of Th2 cell immune gene mRNA expression 36 signature across the four integrated clusters. 37 38 Figure 5: Integrative analysis of epithelioid MPM. A: Comparison of cluster 39 assignments between the epithelioid-only and full cohort integrative analyses demonstrating 40 good concordance, with only 7 cases being reassigned to another cluster. B: Integrative 41 clustering analysis applied to cases with epithelioid histology. C: Kaplan-Meier plot of the 42 epithelioid-only integrative subgroups. D: PATHMARK analysis revealed differentially active 43 molecular pathways that define the poor prognosis epithelioid-only subgroup. E: Validation of 44 the TCGA epithelioid subtypes in an independent cohort of 141 epithelioid MPM (Bueno et 45 al.) confirming the protective effect of molecular features that define iCluster 1. 46

1 Figure 6: Non-coding RNA Subtypes and differential abundance for lncRNAs 2 and miRNAs in the TCGA and Bueno cohorts. a-g) LncRNA subtypes and differential 3 abundance. A: Top to bottom: Normalized abundance heatmap for a 4-subtype solution, 4 then a silhouette width profile (Wcm) calculated from the consensus membership matrix; 5 clinical and molecular covariates, with p-values from Fisher exact, Chi-square or Kruskal 6 tests; and profiles of RNA-seq-based EMT scores and leukocyte fraction. B: Distribution of 7 purity estimated by ABSOLUTE, with a Kruskal p-value. C: Distribution of RNA-seq-based 8 EMT scores, with a Kruskal p-value. D: Kaplan-Meier plot for overall survival, with a log-rank 9 p-value. E: Kaplan-Meier plot for overall survival, with a log-rank p-value for a 4-subtype 10 solution for the Bueno cohort. F,G: lncRNAs that were differentially abundant (SAM 2-class 11 unpaired analysis, FDR < 0.05) between the better-survival lncRNA subtype and all other 12 samples, for the TCGA cohort (F) and the Bueno cohort (G). The largest 15 positive and 15 13 negative fold changes are shown; blue triangles mark lncRNAs that were in these gene sets 14 in results for both cohorts. Text to the right of each barplot gives means-based fold changes, 15 mean abundance in the target then the other samples, and the cytoband for the gene. See 16 also Supplementary Table S5A. H-L) microRNA mature strand subtypes and differential 17 abundance in the TCGA cohort. H: Top to bottom: Normalized abundance heatmap for a 18 5-subtype solution, then a silhouette width profile (Wcm) calculated from the consensus 19 membership matrix; clinical and molecular covariates, with p-values from Fisher exact, Chi- 20 square or Kruskal tests; and profiles of RNA-seq-based EMT scores and leukocyte fraction. 21 l: Distribution of purity estimated by ABSOLUTE, with a Kruskal p-value. J: Distribution of 22 RNA-seq-based EMT scores, with a Kruskal p-value. K: Kaplan-Meier plot for overall 23 survival, with a log-rank p-value. L: miRNA mature strands that were differentially abundant 24 between the better-survival lncRNA subtype and all other samples, for the TCGA cohort. The 25 largest 15 positive and 15 negative fold changes are shown. Text to the right of each barplot 26 gives means-based fold changes, mean abundance in the target then the other samples, and 27 the cytoband(s) for the mature strand. See also Supplementary Table S5B. 28 29 Figure 7. MPM is enriched for both EMT and mRNA expression of immune 30 targets. A: Unsupervised analysis identifies correlations between EMT and multiple 31 platforms. Tumors are ordered from left to right according to increasing EMT score. 32 Numbered color bars indicate group assignments (clusters) from other data 33 types. Statistically significant correlations are shown between EMT score and (starting at top) 34 integrative multi-dimensional analyses on both iCluster and PARADIGM platforms, along 35 with mRNA, miRNA and lncRNA clusters, methylation status and consensus histology. 36 Lower panels illustrate significant correlations between these clusters and selected miRNAs, 37 proteins, immune target genes. B: Spectra of EMT scores across different tumor types. 38 Mesothelioma is the second most mesenchymal cancer type after sarcoma in 31 tumor types 39 analyzed. Despite most MPM tumors having undergone EMT, a broad range of EMT scores 40 were observed across mesothelioma cases, which corresponded to a large extent with 41 histologic subtype. C: Waterfall plot illustrating the correlation between EMT and immune 42 target genes. D: Plot highlighting VISTA gene mRNA expression, which is highest in MPM, 43 across all TCGA tumor types. E: Box plot indicating VISTA mRNA expression levels in 44 individual histological types of the TCGA MPM cohort. The highest mRNA expression levels 45 were observed in the epithelioid subtype (Wilcoxon rank-sum, p=2e-7), whilst sarcomatoid 46 MPM had the lowest mRNA expression (p=0.017). Red arrows indicate two epithelioid cases

1 that were examined by immunohistochemistry, TCGA-SC-A6LQ-01 (1) and TCGA-SC- 2 A6LM-01 (2). F: Immunohistochemical staining for VISTA (Rabbit monoclonal anti-VISTA 3 antibody, clone D1L2G, 0.1 μg/mL, Cell Signaling Technology, Danvers, MA, USA) in normal 4 mesothelial lining from pleura (*) and benign pleuritis with reactive mesothelial proliferation 5 (**), and 2 TCGA MPM cases, TCGA-SC-A6LQ-01 (1) and TCGA-SC-A6LM-01 (2); images 6 captured at 100x magnification. These results confirm high protein expression of VISTA on 7 tumor cells in epithelioid MPM. G: VISTA immunohistochemistry. VISTA protein is 8 expressed both in tumor cells (red arrows) and in infiltrating inflammatory cells (black arrows) 9 in the epithelioid MPM case TCGA-SC-A6LQ-01. Image captured at 200x magnification. 10 11 12 13 14

Downloaded from cancerdiscovery.aacrjournals.org on September 27, 2021. © 2018 American Association for Cancer Research. Mutation Rate Gene Mutation Alteration Copy Number Clinical Parameters A Probability B 0.25 0.00 0.05 0.10 0.15 0.20 # SCNAs # Mutations Downloaded from C>A 20 40 Author manuscriptshavebeenpeerreviewedandacceptedforpublicationbutnotyetedited. 10 20 Author ManuscriptPublishedOnlineFirstonOctober15,2018;DOI:10.1158/2159-8290.CD-18-0804 C>G Hypermutator: TCGA-UD-AAC1 0 0 16p13.3 (RBFOX1) Asbestos exposure 13q14.11 ( 9p21.3 (CDKN2A) 6q25.1 (LATS1) 3p21.1 (BAP1) cancerdiscovery.aacrjournals.org 22q12.2 (NF2) C>T 10q24.32 Histology N_Stage SETDB1 T_Stage 16q24.2 10p15.3 1p36.31 11q23.2 PBRM1 LATS2) PTCH1 SETD2 1p21.3 LATS1 LATS2 15q14 MSH2 BAP1 TP53 Sex NF2 Mutations per Mb Age 0 1 2 3 T>A 378 mutations T>C T>G on September 27, 2021. © 2018American Association for Cancer Research.

Probability C 0.00 0.01 0.02 0.03 0.04 0.05 0.06 C>A C>G Figure 1 All othersamplescombined C>T T>A 36 mutations/patient Average: T>C T>G D SMGs from TCGA cohort LATS2 FDR < 0.05 -log10(q) -log10(q)

Copy Number Alteration Mutation Rate T_Stage Age Histology T4 T3 T2 T1b T1a T1 TX Diffuse malignantmesothelioma-NOS Sarcomatoid Biphasic Epithelioid Above 80 40 to80 Below 40 non synonymous synonymous NA Amplification Gain Loss Deletion No Change SETDB1 SETD2 TP53 NF2 BAP1

SMGs from Buenocohort N_Stage NX N3 N2 N1 N0 0.05 < FDR <0.2

Gene Mutation Asbestos exposure Sex DDX51 RYR2 ULK2 DDX3X RNA Nonstop_Mutation Translation_Start_Site Intron Nonsense Frameshift Splice Site Missense Other NonSyn In-frame INDEL Syn No Mutation male female No Yes [Not Available] [Unknown] B C A

−log10 FDR E 0 2 4 6 MALAT1 MAX SETD2 LATS2 BAP1

BAP1 normalized gene expression TP53

Number of Arms NF2 −3 0 10 20 30 40 −2 −1 0 1 2 MEF2 p <0.01 Downloaded from BAP1 wild−type

A UD-AAC7 BAP1 wild−type CEBP Eect dierence ( TTF1 MQ-A4LV

ESRRA 3U-A98H SMAD4 -0.1 A ZN-A9VV

NFE2L1 3U-A98F Author manuscriptshavebeenpeerreviewedandacceptedforpublicationbutnotyetedited. SO MEIS1 EN1 LK-A4NZ CEBPG Author ManuscriptPublishedOnlineFirstonOctober15,2018;DOI:10.1158/2159-8290.CD-18-0804 GABP

X9 SC-A6LQ TS-A7OZ YY1

S UD-AABZ A

TAT UT-A88D ADD1 UT-A97Y 4 A A TF3 ZN-A9VO CTR1 ZN-A9VW BAP1 2hits BAP1

BAP1 an 3H-AB3T cancerdiscovery.aacrjournals.org

A TS-A7OU

0 3U-A98E p =6.9e−13 MQ-A4LM anyhitvs. wild-ype) 3H-AB3U TS-A7PB y hit

P UT-A88C AX8 IRF1 NR3C1 UT-A88E

TA 3H-AB3S

L1 LK-A4NW

FO LK-A4O2 IRF8 SP3 XO1 LK-A4O6 TF

MTF1 LK-A4O7 AP2 0.1 MQ-A4KX EGR2

A MQ-A6BQ IRF9 SC-A6LM

D TS-A7P1 TS-A7P3 TS-A8AI TS-A8AS UD-AAC4 UD-AAC6 IRF9 YS-A95C ZN-A9VQ SP3 ZN-A9VS ZN-A9VU on September 27, 2021. © 2018American Association for Cancer EGR2 MQ-A6BR Research. IRF8 SC-A6LP SC-AA5Z Figure 2 F TAL1 LK-A4O5 YY1 inferred activity YS-AA4M NR3C1 BAP1 wild−type −0.2 −0.1 0.0 0.1 0.2 0.3 UD-AAC1 FO XO1 3U-A98I FDR=2.847e−05 UT-A88G PAX8 ZN-A9VP LK-A4O0 MTF1 MQ-A6BL TS-A8AV IRF1 MQ-A6BS TF AP2A 3U-A98J SH-A7BD ADD1 3H-AB3M 3H-AB3X ACTR1A 3U-A98D

BAP1 anyhit CEBPG 3U-A98G LK-A4NY EN1 MQ-A4LC MQ-A4LJ ESRRA MQ-A4LP NQ-A638 YY1 SC-A6LN MEIS1 SC-A6LR SH-A7BC GABPA SH-A9CT SH-A9CU SO X9 TS-A7OY S TAT TS-A7P0 TS-A7P6 G ATF3 TS-A7P8 TS-A8AY IRF8 inferred activity NFE2L1 YS-A95B 0 MEF2A Fisher exacttest −0.2 0.0 0.2 0.4 BAP1 SMAD4 -log10(p) FDR=2.6e−07 TTF1 1 CEBPA * wild−type MAX * 2 * FDR<0.25 BAP1 BAP1 Inactivated Other Splice site Nonsense Missense Frameshift In-frame indel Histological Type BAP1 TF activity BAP1 Exclusive NOS biphasic sarcomatoid epithelioid Co-occurence 1-hit wild-type 2-hits −0.4 −0.2 0 0.2 0.4 status anyhit Author Manuscript Published OnlineFirst on October 15, 2018; DOI: 10.1158/2159-8290.CD-18-0804 A iCluster 1 iClusterAuthor 2 manuscriptsiCluster have 3 beeniCluster peer 4 reviewed and accepted for publication but have not yet been edited. B TCGA-MQ-A6BR (TCGA case 1) 1 4

2 3 3

4 2

6 1

7 Ratio (tumor/normal)

8 0 9 Chr 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 10 TCGA-SC-A6LP (TCGA case 2) Chromosome 11 4 12 13 14 3 15 16

17 2

18 cohort TCGA

20 21 1

22 Ratio (tumor/normal) A8AI AB3S AAC4 0 3H Chr 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 TCGA 3U A98I TCGA TS TCGA 3U A98J TCGA 3U A98F TCGA TS A7P3 TCGA TS A7P0 TCGA TSTCGA A7P6 TS A7P8 TCGA 3U A98E TCGA TS A7P1 TCGA TS A8AY TCGA TS A8AV TCGA UT A88E TCGA LK A4O4 TCGA LK A4O2 TCGA 3U A98H TCGA YS A95B TCGA LK A4O0 TCGA LK A4NZ TCGA 3H AB3T TCGA UT A97Y TCGA LK A4O6 TCGA 3U A98D TCGA LKTCGA A4O5 LK A4O7 TCGA UT A88C TCGA UT A88D TCGA LK A4NY TCGA TSTCGA A7PB TS A8AS TCGA 3U A98G TCGA SC A6LP TCGA 3H AB3X TCGA YS A95C TCGA TCGA NQ A638 TCGA TS A7OY TCGA ZN A9VV TCGA UT A88G TCGA ZNTCGA A9VP 3H AB3U TCGA SCTCGA A6LN TS A7OZ TCGA MQ A4LJ TCGA SCTCGA A6LR SC AA5Z TCGA ZN A9VS TCGA SH A9CT TCGA SC A6LQ TCGA MQ A4LV TCGA ZN A9VU TCGA ZN A9VO TCGA SHTCGA A7BD TS A7OU TCGA ZN A9VQ TCGA SH A7BC TCGA 3H AB3M TCGA SC A6LM TCGA UD TCGA UD AAC1 TCGA UD AAC6 TCGA UD AAC7 TCGA UD AABZ TCGA SH A9CU TCGA MQ A6BL TCGA YS AA4M TCGA MQ A4LP TCGA LK A4NW TCGA MQ A4LC TCGA MQ A6BS TCGA ZN A9VW TCGA MQ A4KX TCGA MQ A4LM TCGA MQ A6BR TCGA-SC-A6LP TCGA-UD-AAC1TCGA MQ A6BQ TCGA-UD-AAC1 (TCGA case 3) TCGA-MQ-A6BR 4

C 1 Ratio (tumor/normal)

0 Chr 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22

ICGC-K2F2-A45 (ICGC case 1) 4 1 2 3 4 5 3

1 Ratio (tumor/normal) 0 6 7 8 9 10 11 12 Chr 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 ICGC-K2F2-H60 (ICGC case 2) 4

3 ICGC cohort 13 14 15 16 17 18 2

1 Ratio (tumor/normal)

Chr 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 19 20 21 22 X Y TP53 NF2 Homozygous, Homozygous, 2 copies 2 copies D TCGA-SC-A6LP

SETDB1 ACOT12; RELN Heterozygous, ABLIM3 Heterozygous, 1 copy Heterozygous, 1 copy 2 copies Downloaded from cancerdiscovery.aacrjournals.org on September 27, 2021. © 2018 American Association for Cancer Research.Figure 3 Author Manuscript Published OnlineFirstAdjusted on Rand October Index 15, * 2018; DOI:CLUSTERS 10.1158/2159-8290.CD-18-0804 HISTOLOGY Author manuscripts have been peerPARADIGM reviewed andvs. others: accepted for publication but have not yet been edited. A PARADIGM 1 EPITHELIOID iCluster 0.61 2 BIPHASIC lncRNA 0.60 3 SARCOM ATOID mRNA 0.41 4 NOS miRNA 0.25 Methylation 0.13 5 0.10 CNV 6 RPPA (n=52) 0.06 Histology

B D iCluster BAP1 alt <0.001 TP53 mut 0.06 CDKN2A <0.001 hom del High EMT Low CNA Low BAP1 alt CDKN2A hom del Low MSLN High DNA meth Low DNA meth Low CLDN1 High AURKA High BAP1 alt LATS2 mut

DNA copy number

C DNA methyla on Cox regression with iCluster, Histology and CDKN2A status

Characteristic HR (95% CI) P iCluster group (ref iCluster 1) iCluster 4 3.58 (1.42, 9.01) iCluster 3 1.17 (0.45, 3.04) 0.008 iCluster 2 1.33 (0.55, 3.19) Histology (ref Epithelioid) mRNA expression Non-Epithelioid 2.47 (1.33, 4.58) 0.005 CDKN2A (ref WT) Homozygous dele tion 2.72 (1.32, 5.61) 0.005

ncRNA expression E Th2 cells - (Kruskal-Wallis p= 7.6e-07)

miRNA expression 0.05 0.10 0.15 Immune score (Bindea signatures)

iCluster1 iCluster2 iCluster3 iCluster4 p=0.01 p=0.0080 p=0.0002 p=0.0001 Downloaded from cancerdiscovery.aacrjournals.orgp=6.5e-7 on September 27, 2021. © 2018 American Association for Cancer Research.Figure 4 TP53 mut TP53 CDKN2A CDKN2A BAP1 alt B A hom hom del iCluster

miRNA expression lncRNA expression mRNA expression DNA methylation DNA copy number iCluster (n=74) ALL-MPM CASES FROM EPITHELIOID

High miR-361 Low Low CNA DNA meth Downloaded from

Author manuscriptshavebeenpeerreviewedandacceptedforpublicationbutnotyetedited. EPITHELIOID-ONLYiCluster

Low DNA LowDNA meth Author ManuscriptPublishedOnlineFirstonOctober15,2018;DOI:10.1158/2159-8290.CD-18-0804 BAP1 4 4 3 2 1 No

loss

0 0 0 0 14 1

cancerdiscovery.aacrjournals.org CDKN2A hom del 0 0 0 13 0 2

miR-126 LOXL2 0 0 9 0 1 3

9 3 0 3 4

<0.001 <0.001 0.30

on September 27, 2021. © 2018American Association for Cancer Figure 5 Research. C E D TCGA cohort Bueno etal.cohort A lncRNAs D TCGA (5 yr) E Bueno (5 yr) H miRNAs K TCGA (5 yr) 1.0 1.0 1.0 1 (20/14) 1 (50/31) 1 (14/9) 2 (18/14) 2 (60/52) 2 (9/7) 3 (8/7) 0.8 3 (19/16) 0.8 3 (56/49) 0.8 miR-29abc 4 (20/16) PVT1 4 (17/17) 4 (45/28) 193a 5 (23/22) H19 0.6 p= 1.4e-10 0.6 0.6 LINC00152 p= 1.1e-3 221/222 0.4 0.4 22-3p 0.4 1 4 let-7i-3p 0.2 0.2 0.2 2,3 151a Propo rtion surviving 99b Propo rtion surviving Propo rtion surviving 0.0 0.0 0.0 p= 6.6e-4 100-5p 0 10 20 30 40 50 60 rowscaled log10(RPM+1) 0 10 20 30 40 50 60 rowscaled log10(FPKM+1) MEG3 0 10 20 30 40 50 60 200 family Time (months) Overall su rvival (mo) 30b,d Time (months) Author Manuscript Published OnlineFirst on October 15, 2018; DOI: 10.1158/2159-8290.CD-18-0804 Xq27.3 group

Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. 0 RP1-78O14.1 17~92a group LINC01105 mean mean mean mean FPKM FPKM L RPM RPM F FC 1 2/3/4 band FC 1 2 to 5 band(s) 14q32 group RP11 263K4.5 59.1 50 0.8 12q21.31 0.05, fs miR 100 5p 2.4 29471 12152 11q24.1 LINC01550 6.9 12 1.7 14q32.2 miR 203a 2.3 728.7 318.9 14q32.33 LINC00657 RP11 879F14.2 4.9 5.5 1.1 18q21.33 21-5p miR 193a 3p 2.3 43.3 19 17q11.2 RP11 297J22.1 4.8 14.4 3 2q14.2 miR 29b 2 5p 2.2 59.9 26.9 1q32.2 CTC 573N18.1 3.7 6.6 1.8 5q35.3 142 miR 148b 3p 2.2 227.3 104 12q13.13

AC017104.6 3.3 16.1 4.9 2q37.1 10, q-val 146b miR 652 3p 2.2 29.2 13.5 Xq23 PAX8-AS1 AC067959.1 3 8 2.6 2p24.1 miR 193a 5p 2.1 1528 728 17q11.2 RP11 638I2.6 2.7 16.6 6.2 14q32.2 199ab miR 378c 2.1 28 13.5 10q26.3 C8orf37 AS1 2.7 10.4 3.9 8q22.1 miR 361 3p 2 323.6 158.1 Xq21.2 GATA6-AS1 RP11 19E11.1 2.6 6.7 2.6 2q14.2 143-3p miR 1180 3p 2 56.1 27.6 17p11.2 NEAT1 CDC37L1 AS1 2.2 5.6 2.6 9p24.1 miR 1287 5p 2 48.9 24.4 10q24.2 MALAT1 AC124789.1 2.1 46.8 21.8 17q12 212 miRs mean RPM miR 29c 3p 1.9 2124 1093 1q32.2 197 lncRNAs mean >= 2, q<=0.05, f score 0.25 1 RAD51-AS1 SERPINB9P1 2.1 5.7 2.6 6p25.2 1 let 7i 3p 1.9 111 57.7 12q14.1 RP11 359M6.1 2.1 19.5 9.4 12q21.2 miR 29b 3p 1.8 1742 967.9 1q32.2,7q32.3 LINC01133 2.1 53.1 25.8 1q23.2 miR 497 5p 1.7 53.6 31.7 17p13.1 1 4 3 2 Wcm 5 4 1 2 3 Wcm 0 PCED1B AS1 1.4 15.7 21.5 12q13.11 0 miR 27a 3p 1.9 698.9 1319 19p13.13 N stage p= 0.26 LINC00662 1.4 6 8.6 19q11 N stage miR 23a 3p 1.9 2061 3953 19p13.13 PAX8 AS1 1.5 19.1 28.2 2q14.1 p= 0.18 miR 199b 3p 2 1352 2686 9q34.11 T stage 0.70 CTD 3252C9.4 1.5 7.4 11.3 19p13.12 T stage 0.770 miR 199a 3p 2 1356 2698 1q24.3,19p13.2 Asbestos 0.22 CTB 193M12.5 1.6 4.4 7 16p13.11 Asbestos 0.038 miR 625 3p 2 52.9 108.2 14q23.3 CTC 459F4.3 1.8 3.7 6.6 19q11 miR 181a 2 3p 2.1 352.1 727.3 9q33.3 Histology 2.3E-03 Histology 0.022 MORT ZNF667 AS1 1.8 5.6 10.3 19q13.43 miR 24 2 5p 2.1 14.3 30.3 19p13.13 lncRNA H19 2.6 396.6 1021 11p15.5 miRNA miR 143 3p 2.7 11126 29778 5q32 mRNA 3.5E-17 LINC01116 3.4 3.1 10.6 2q31.1 miR 199a 5p 2.8 553 1556 1q24.3,19p13.2 SF TA1P 4 5.9 23.6 10p14 lncRNA 1.4E-12 miR 145 5p 2.9 297.6 857.4 5q32 iCluster 7.7E-14 RP11 54H7.4 5.1 1.2 6.1 13q33.3 mRNA 1.2E-14 miR 214 5p 2.9 14.4 41.5 1q24.3 PARADIGM 2.1E-21 XIST 16.5 1.3 21.9 Xq13.2 iCluster 1.3E=12 miR 143 5p 3.2 10.6 33.5 5q32 miRNA 1.4E-12 RP11 417E7.1 25.1 1.5 37.8 6q27 miR 217 4.8 13.9 66.3 2p16.1 RP11 865I6.2 108.9 0.1 7.7 8q13.2 PARADIGM 3.0E-13 miR 1 14.2 6.3 90 18q11.2,20q13.33 SCNA 0.011 GS1 600G8.5 179.8 0.2 28.3 Xp22.2 SCNA 2.6E-04 miR 133a 3p 17.2 2.7 45.8 18q11.2,20q13.33 DNA methylation 0.025 DNA methylation 1.6E-04 5 0 0 10 20 20 40 60 40 RPPA — In TCGA and RPPA — 15 Purity p= 0.054 Bueno results p= 1.0E-7 Fold change Purity Fold change TWIST2 TWIST2 C10orf54 (VISTA) C10orf54 (VISTA) CD274 (PD-L1) mean CD274 (PD-L1) PDCD1 (PD-1) (FPKM) PDCD1 (PD-1) SNAI1 G FC 4, 1/2/3 band SNAI1 Consensus TWIST1 RP11 263K4.5 9.1 28.6, 3.1 12q21.31 TWIST1 N stage T stage Asbestos histology ZEB1 LINC00689 5.3 4.2, 0.8 7q36.3 ZEB1 N0 T1 Yes Epithelioid SNAI2 RP11 863P13.6 4.9 3.9, 0.8 16q24.2 SNAI2 ZEB2 AC067959.1 3 5.7, 1.9 2p24.1 ZEB2 N1 3 T2 4 No Biphasic 8 EPB41L4A AS2 2.8 2.8, 1 5q22.2 8 LINC00885 2.7 2.7, 1 3q29 NX TX Unknown Sarcomatoid 4 4 EMT score LINC01550 2.6 6, 2.3 14q32.2 EMT score 0 p= 2.5E-6 p= 1.7E-07 NOS LINC00473 2.4 6.8, 2.8 6q27 0 1 1 LINC01105 2.4 24.2, 10.3 2p25.2 mRNA Leukocyte RP11 79N23.1 2.2 3.3, 1.5 6p12.1 Leukocyte lncRNA fraction p= 9.4E-3 AC034220.3 2.1 18, 8.7 5q31.1 fraction p= 1.1E-6 0 RP11 379F4.6 2.1 3.5, 1.7 3q25.32 0 iCluster miRNA DNA n= 20 17 19 18 P WAR6 2 7.7, 3.8 15q11.2 n= 23 20 14 9 8 PARADIGM RPPA SCNA methylation LINC00545 2 2.9, 1.5 13q12.3 RP11 379F4.7 2 20.6, 10.3 3q25.32 1 1 1 High AC012146.7 1.9 2.4, 4.5 17p13.2 2 2 2 CTB 193M12.5 2 2.8, 5.4 16p13.11 Int B p= 0.054 C p= 2.5E-6 AC093850.2 2 4.2, 8.5 2q35 I p= 1.0e-7 J p= 1.7e-7 3 3 3 Low 1.0 8 RP11 14N7.2 2.1 8.5, 17.9 1q21.1 1.0 8 RP11 482H16.1 2.3 1.1, 2.5 2p16.1 4 4 4 7 LINC01436 2.3 2.5, 5.9 21q22.12 7 5 5 0.8 MIR155HG 2.5 1.7, 4.3 21q21.3 0.8 6 H19 3.1 223.8, 685.3 11p15.5 6 NA 6 0.6 5 AC133644.2 3.1 0.9, 3 2p11.2 0.6 5 SF TA1P 7.1 1.9, 13.8 10p14 0.4 4 RP11 54H7.4 8 0.7, 5.9 13q33.3 0.4 4 Pu rity DISC1FP1 17.1 0.3, 5.1 11q14.3 Pu rity 3 RP11 417E7.1 19.8 0.9, 16.9 6q27 3 EMT score 0.2 GS1 600G8.5 51 0.3, 14.4 Xp22.2 0.2 EMT score 2 RP11 865I6.2 73.8 0.1, 6.1 8q13.2 2 0.0 0.0 0 60 1 2 3 4 1 2 3 4 40 20 1 2 3 4 5 1 2 3 4 5 Fold change

Downloaded from cancerdiscovery.aacrjournals.org on September 27, 2021. © 2018 American Association for Cancer Figure 6 Research. A EMT Score B EMT Score Dist ribution in TCGA Pan-Cancer samples Histology P: 2.99e−05 SARC (N=206) MESO (N=74) iCluster P: 5.57e−08 GBM (N=169) PCPG (N=183) PARADIGM P: 2.9e−08 SKCM (N=471) Author Manuscript Published OnlineFirst on October 15, 2018; DOI: 10.1158/2159-8290.CD-18-0804ACC (N=78) mRNA Author manuscripts have been peer reviewed andP: 1.37e−08 accepted for publicationLGG but (N=532) have not yet been edited. UCS (N=57) miRNA P: 3.64e−10 KIRC (N=516) UVM (N=80) lncRNA P: 6.38e−07 PAAD (N=157) TGCT (N=155) Methylation P: 1.26e−02 BRCA (N=1090) LIHC (N=371) hsa−miR−200a (P=1.14e−13) THYM (N=120) miRNA hsa−miR−200b (P=2.18e−11) LUSC (N=487) hsa−miR−429 (P=5.95e−09) S TAD (N=400) HNSC (N=516) RP PA E−Cadhe rin (P=3.17e−04) LUAD (N=512) TGFB1 (P=4.04e−09) OV (N=273) OX40L (P=1.25e−08) KIRP (N=286) OX40 (P=7.7e−07) ESCA (N=183) CHOL (N=36) B7−H3 (P=8.84e−06) BLCA (N=404) PDL2 (P=4.75e−05) THCA (N=508) ADORA2A (P=3.09e−04) READ (N=90) UCEC (N=174) TIM3 (P=0.003) C OAD (N=280) CD4 (P=0.003) PRAD (N=494) IL6 (P=0.006) KICH (N=65) CESC (N=301) CXCR4 (P=0.007) −5 0 5

Immune Targets PDL1 (P=0.017) CTLA4 (P=0.019) EMT score CD137 (P=0.026) ICOSLG (P=0.003) VIS TA (P=4.39e−08) Cluster Consensus histology 1 epithelioid 2 biphasic −2 −1 0 1 2 3 sarcomatoid 4 NOS 5 2 3 4 5 6 7 D VIS TA Expression In TCGA Pan-Cancer Samples C 0.6 P<0.001 MESO P<0.01 LGG P<0.05 KIRC PAAD 0.3 P>0.05 HNSC SARC GBM 0.0 THYM S TAD THCA CESC −0.3 CHOL ESCA BLCA

Pearson correlation LUAD −0.6 OV LUSC

IL6 KIRP CD4 PD1 IL10 IL1A TIM3 ICOS PDL2

LAG3 ACC PDL1 OX40 CCL2 VIS TA B7−H3 OX40L

CTLA4 CD137

TGFB1 TGCT CXCR4 ICOSLG C OAD

ADORA2A READ LIHC SKCM PRAD E PCPG 14 1 UCS BRCA UCEC KICH 13 UVM 5.0 7.5 10.0 12.5 VIS TA expression 12 F

11 VISTA log2(exp) 10 2

biphasic NOS epithelioid sarcomatoid PLEURA G NON-NEOPLASTIC TCGA MPM CASES TCGA

Downloaded from cancerdiscovery.aacrjournals.org onFigure September 7 27, 2021. © 2018 American Association for Cancer Research. Author Manuscript Published OnlineFirst on October 15, 2018; DOI: 10.1158/2159-8290.CD-18-0804 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

Integrative Molecular Characterization of Malignant Pleural Mesothelioma

Julija Hmeljak, Francisco Sanchez-Vega, Katherine A. Hoadley, et al.

Cancer Discov Published OnlineFirst October 15, 2018.

Updated version Access the most recent version of this article at: doi:10.1158/2159-8290.CD-18-0804

Supplementary Access the most recent supplemental material at: Material http://cancerdiscovery.aacrjournals.org/content/suppl/2018/10/13/2159-8290.CD-18-0804.DC1

Author Author manuscripts have been peer reviewed and accepted for publication but have not yet been Manuscript edited.

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://cancerdiscovery.aacrjournals.org/content/early/2018/10/13/2159-8290.CD-18-0804. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.