Genomic landscape of pleural and peritoneal mesothelioma tumours

Background Malignant pleural and peritoneal mesotheliomas are rare malignancies with unacceptable poor prognoses and limited treatment options. The genomic landscape is mainly characterised by the loss of tumour suppressor genes and mutations in DNA repair genes. Currently, data from next-generation sequencing (NGS) of mesothelioma tumours is restricted to a limited number of cases; moreover, data comparing molecular features of mesothelioma from the pleural and peritoneal origin with NGS are lacking. Methods We analysed 1113 pleural mesothelioma and 355 peritoneal mesothelioma samples. All tumours were sequenced with the FoundationOne® or FoundationOne®CDx assay for detection of substitutions, insertion–deletions, copy-number alterations and selected rearrangements in at least 324 cancer genes. Results This analysis revealed alterations in 19 genes with an overall prevalence of at least 2%. Alterations in BAP1, CDKN2A, CDKN2B, NF2, MTAP, TP53 and SETD2 occurred with a prevalence of at least 10%. Peritoneal, compared to pleural mesothelioma, was characterised by a lower prevalence of alterations in CDKN2A, CDKN2B and MTAP. Moreover, we could define four distinct subgroups according to alterations in BAP1 and CDKN2A/B. Alterations in Hedgehog pathway-related genes (PTCH1/2 and SUFU) and Hippo pathway-related gene (NF2) as well as KRAS, EGFR, PDGFRA/B, ERBB2 and FGFR3 were detected in both cohorts. Conclusion Here, we report the molecular aberrations from the largest cohort of patients with mesothelioma. This analysis identified a proportion of patients with targetable alterations and suggests that molecular profiling can identify new treatment options for patients with mesothelioma.


INTRODUCTION
Mesothelioma is a highly aggressive and lethal disease of the membrane lining of the serous cavities of the pleura, peritoneum, pericardium, and the tunica vaginalis testes. Malignant pleural mesothelioma is closely linked to asbestos and develops 30-40 years after initial exposure [1]. In comparison to pleural mesothelioma, peritoneal mesothelioma has only been modestly linked to asbestos [2]. Even though asbestos has been forbidden and strictly controlled in many countries, the global incidence of mesothelioma will rise in the coming years due to continued use in emerging economies and its long latency period.
Mesothelioma still has an unacceptably poor outcome, and there is great need for new therapeutic options. Since 2004, the first-line treatment consists of systemic combination therapy with pemetrexed and platinum-based chemotherapy or inclusion in clinical trials. Just recently, treatment with immune checkpoint inhibitors showed promising results and prolonged survival compared to standard chemotherapy treatment [3,4]. The combination of ipilimumab and nivolumab was approved by the FDA in 2020 but is not yet a standard of care in most countries. In the later line setting, mesothelioma remains an orphan disease with very limited treatment options and other treatments such as multi-tyrosine kinase inhibitors sunitinib or sorafenib showed limited efficacy [5,6], possibly related to the lack of patient stratification. Treatment allocation based on molecular features has been successful for advanced lung cancer, where targeted treatment options, even when applicable in 1-2% of cases, have improved patient outcomes [7]. This underlines the importance and potential of such alterations in mesothelioma, even if rare.
The genomic profile of mesothelioma reveals a low protein-coding mutation rate [8], with alterations of tumour suppressor genes and of epigenetic modifiers. The most commonly occurring alterations are deletions of the tumour suppressor genes BAP1, NF2, and CDKN2A [9]. Despite similarities of cellular features and common genetic alterations, pleural and peritoneal mesothelioma might be related to different risk factors; therefore, understanding the molecular features, which differentiate or associate these two diseases is crucial to give insights to the pathogenesis and treatment of these rare diseases. The aim of this study was to examine the prevalence of targetable alterations in pleural and peritoneal mesothelioma in a large cohort of cases; we report the mutational profile of 1468 mesothelioma patients, including the comparison of pleural and peritoneal mesothelioma profiled using NGS.

MATERIALS AND METHODS Sequencing
Comprehensive genomic profiling (CGP) was performed with hybridcapture-based next-generation sequencing on formalin-fixed and paraffinembedded (FFPE) clinical specimen in a Clinical Laboratory Improvement Amendments-certified (CLIA), College of American Pathologists-accredited (CAP), New York State-approved laboratory (Foundation Medicine, Cambridge MA, more precise method description including percentage of tumour cell content, allele frequency, number of reads, copy-number detection as published in ref. [10]). CGP results included were from 1113 pleural mesothelioma and 355 peritoneal mesothelioma tumour tissue samples from patients in the United States sequenced through December 2020 as a part of routine clinical care. All tumours were sequenced with the FoundationOne ® or FoundationOne ® CDx test for detection of substitutions, insertion-deletions, copy-number alterations and selected rearrangements in at least 324 cancer genes. Samples included in this study were sequenced to a median depth of 818× and mean depth of 782.2×. All known or likely pathogenic alterations were included in this study including heterozygous and homozygous somatic and germline alterations. Microsatellite instability was called on at least 95 loci [11], and tumour mutational burden (TMB) was calculated on at least 0.8 Mb as described in [12]. Tumours with high TMB were defined as those with at least 10 mutations per megabase. Approval for this study, including a waiver of informed consent and a HIPAA waiver of authorisation, was obtained from the Western Institutional Review Board (Protocol No. 20152817).

Analysis
For the analysis of prevalence, the 95% confidence interval was determined using binomial error estimation. Co-occurrence and mutual exclusivity were analysed using Fisher's exact test with multiple hypothesis test correction by Benjamini-Hochberg. Statistical significance was defined as P < 0.05. Gene prevalence in molecular subgroups was analysed by chisquare with multiple hypothesis correction by Benjamini-Hochberg.

Cohort description
In total, 1113 pleural mesothelioma and 355 peritoneal mesothelioma samples from patients diagnosed through December 2020 were included in the analysis (total 994 male patients and 542 female patients) with 52.7% (n = 187) female, 47.3% male (n = 168) in the peritoneal group and 29.4% (n = 327) female, 70.6% (n = 786) male in the pleural group. Mesothelioma patients with other subtypes such as testis, pericardium and mesothelioma not otherwise specified were not included in this analysis.
Landscape of genomic alterations in pleural and peritoneal mesothelioma Mesothelioma is characterised by chromosomal abnormalities such as loss of chromosome regions and tumour suppressor genes [13], unlike other malignancies such as melanoma and lung cancer where the gain of function mutations in oncogenes are mainly present [14].
Due to no available clinical data, we studied the survival, histology and sex priorities of the four different subgroups in the TCGA mesothelioma cohort (n = 82) [18,19]. The subgroups with only CDKN2A alteration and the subgroup with CDKN2A/BAP1 alteration have the worst survival. Interestingly, the subgroup with no CDKN2A or BAP1 alteration survive the longest (Supplementary Fig. 2A). We could not detect differences in histology and sex according to the different alterations ( Supplementary Fig. 2B, C).
Mesotheliomas harbour fewer somatic mutations compared to other tumour entities [13] with less than 5% of mesothelioma tumours having a high TMB and an average of 1.7 mutations per Mb [8]. Here, we detected no differences in the TMB between pleural and peritoneal patients ( Supplementary Fig. 3) and very few patients with a high TMB (defined as ≥10 mutations/Mb), 15 (1.35%) patients in the pleural and 5 (1.41%) patients in the peritoneal group.
Alterations in genes of the cell-cycle machinery and its control mechanisms The most altered gene in pleural mesothelioma is CDKN2A, which encodes for inhibitors of the cyclin-dependent kinases 4 and 6 (CDK4/6) involved in cell-cycle regulation [20]. Therefore, CDKN2A deletions result in the loss of inhibitory components of the cellcycle regulation leading to pro-mitotic signals. In this cohort of pleural mesothelioma, 45% of sample harboured copy-number deletions, 1.3% had variants and 1.6% had rearrangements in CDKN2A. Compared to pleural mesothelioma, deletions in CDKN2A/B were significantly less frequent in peritoneal mesothelioma. Preclinical studies of CDK4/6 inhibitors showed encouraging results in mesothelioma cell lines and xenografts [21], but with limited efficacy in a Phase II clinical trial (NCT02187783). CDKN2A deletions appear often with co-deletions in the adjacent gene MTAP (Fig. 1). MTAP encodes a protein involved in the adenosine monophosphate and in methionine synthesis and is altered in several malignancies [22]. We detected CDKN2A deletions in 48.2% of pleural and in 25.9% of peritoneal mesothelioma. Earlier studies showed a homozygous deletion of CDKN2A in 74% of mesothelioma samples and co-deletion of MTAP in 91% of these cases [23]. MTAP deletions can lead to sensitivity to PRMT5 or MAT2a inhibitors, which showed only limited efficacies in MTAP-deficient solid tumours [24].
Alterations in genes controlling genome integrity Dysregulated cell cycle, genomic errors and lack of control mechanisms are hallmarks of cancer. TP53, tumour suppressor gene is activated upon cell stress and DNA damage and mediates cell-cycle arrest, apoptosis and the transition from S to G1 phase [25]. In previous publications, TP53 was found to be mutated in 8-19% of mesothelioma tumours [9,13]. In our study, TP53 alterations were detected in 17.8% in pleural mesothelioma and in 14.9% of peritoneal mesothelioma patients. A possible therapeutic approach to p53-altered tumours is the use of Hsp90 inhibitors. Hsp90 stabilises mutated p53, which in turn can inhibit the function of wild-type p53 leading to deficient DNA damage response and to sensitisation to chemotherapy [26]. Inhibition of Hsp90 with ganetespib in an unselected patient group lead to partial responses in 52 % of the patients when given in combination with chemotherapy [27].
In this study, several alterations in other genes involved in responses to DNA damage, such as BAP1, PBRM1, BRCA2 and CHEK2 were detected. BAP1 is a deubiquitinylating enzyme involved in gene transcription, DNA damage repair and cellcycle control mechanisms [28]. BAP1 alterations are correlated with environmental carcinogen exposure and BAP1 knockout mice showed increased incidence of mesothelioma after chronic asbestos exposure [29]. In this cohort, BAP1 was one of the most frequently altered genes and was found in 45.0% of pleural mesothelioma and 47.9% of peritoneal mesothelioma ( Fig. 1 and Supplementary Tables 1-3).
BRCA1/2 complexes are proteins involved in homologous recombination to suppress genetic instability. Tumours with defects in homologous recombination repair may be susceptible to PARP inhibitors including tumours with BAP1-deficiency [30]. In our cohort, BRCA2 alterations were detected in 2.5% of pleural mesothelioma and in 1.7% of peritoneal mesothelioma tumours, while a recent study detected BRCA2 alterations in 3 out of 37 pleural mesothelioma patients [31]. A recent Phase IIa clinical trial showed some activity of PARP inhibitors in mesothelioma patients with BAP1 and BRCA2 alterations (MiST1 trial) [32].

Epigenetic modifiers
Mutations occurring in chromatin remodelling complexes are one of the most frequent alterations in different cancers. Genes of the switch/sucrose non-fermentable (SWI/SNF) complex, a tumour suppressor complex, have been previously described to be altered in mesothelioma [33]. Such complex is constituted by 12-15 subunits, including ARID1A, SMARCA4, ARID1B, ARID2, and PBRM1. In pleural and peritoneal mesothelioma, deletions in the 3p21 region, where PBRM1 is located, were frequently  detected. In our cohort, PBRM1 was altered in 6.8% of pleural mesothelioma and in 15.8% peritoneal tumours. In comparison, in the TCGA dataset (cBioportal), with only pleural mesothelioma cases, 7% of the patients harbour an alteration in the PBRM1 gene [19]. In recent studies, alterations in PBRM1 were not detected [34] or overestimated [35] due to the small number of tumours analysed. PBRM1 alterations are also associated with a less immunogenic tumour microenvironment and resistance to immunotherapy [36]. Moreover, PBRM1-defective cancer cells were sensitive to PARP and ATR inhibitors in preclinical models [37]. Thus, further studies are needed to determine the prognostic effect of PBRM1 alterations in patients receiving immunotherapy and its possible value for patient's stratification. Beside its function in chromatin remodelling, the SWI/SNF complex is also a regulator of nucleosome positioning and can regulate gene transcription. The activity of the SWI/SNF complex is opposed by the polycomb repressor complex (PRC), including its subunit enhancer of zeste homologue 2 (EZH2). EZH2 is responsible for trimethylation of the histone H3 at Lys 27 (H3K27me3) and functions as an epigenetic regulator of transcription and is often overexpressed in cancer. Tazemetostat, an EZH2 inhibitor, is currently under evaluation in different Phase I and II clinical trials in tumours with SMARCB1, SMARCA4 and EZH2 mutations with positive results in patients with different solid tumours [38].
In our cohort, SETD2, a histone methyltransferase, was mutated in 10.3% of pleural and 10.4% of peritoneal mesothelioma tumours. Interestingly, WEE1 inhibitors were shown to selectively target SETD2 mutated cells and are currently in Phase II clinical trials with solid tumours and SETD2 loss [39].
DNMT3A, a histone methyltransferase important for de novo methylation of cytosine residues at CpG sites, is important for regulating gene expression in healthy cells. DNMT3A was shown to be overexpressed in mesothelioma cell lines and its expression correlates with worse prognosis in pleural mesothelioma patients [40]. In our cohort, DNMT3A was altered in 4.1% of pleural mesothelioma and in 3.9% of peritoneal mesothelioma cases. Today, DNMT inhibitors such as decitabine and azacytidine, are approved for myelodysplastic syndrome (MDS) and acute myeloid leukaemia and currently under investigation for solid tumours with considerably high toxicity [41].
Similarly, TET2 functions in DNA demethylation and is frequently mutated in haematologic malignancies but seldom in solid tumours [42]. It controls chemokine and PD-L1 expression, lymphocyte infiltration, and cancer immunity [43]. Thus, TET2 loss might represent a potential a biomarker for predicting the efficacy of response to anti-PD-1/PD-L1 immunotherapy. Here, we detected TET2 mutations in 4.2% of pleural mesothelioma samples and 2.5% of peritoneal mesothelioma patients. TET2 alterations were not found in the TCGA dataset [19].
TERT is the catalytic subunit of telomerase and is often reactivated in cancer. In our cohort of pleural and peritoneal mesothelioma patients, the TERT promoter was mutated in 7.6% and in 4.7% of the cases, respectively. In mesothelioma, promoter mutations were detected frequently in the non-epithelial subgroup and are associated with a more aggressive disease and with reduced survival in MPM [44].

Targetable alterations for approved drugs
Receptor tyrosine kinases and KRAS. We have previously reported the case of a patient with peritoneal mesothelioma harbouring an ALK translocation. The patient underwent targeted treatment and achieved a very good partial response [45]. Based on this case, we specifically evaluated the occurrence of ALK alterations in this cohort of patients. In 0.36% of pleural mesothelioma patients and 1.13% of peritoneal mesothelioma patients, we detected ALK rearrangements and short variant alterations (Fig. 4c). Even if rare, targetable alterations of KRAS, EGFR, PDGFRA/B, ERBB2 and FGFR3, were also detected (Fig. 4a). Specifically, ten patients harboured the G12C mutation in KRAS, which could be treated with small molecule inhibitors sotorasib or adagrasib [46,47] (Supplementary  Table 10 and Supplementary Fig. 4).
Hedgehog pathway. Alterations in Hedgehog pathway genes, specifically PTCH1, PTCH2, SUFU, GLI1 were detected. The hedgehog signalling pathway is well conserved and is an important mediator in development, tissue homoeostasis and embryogenesis. Suppressing Hedgehog signalling reduced cell viability in MPM cells [48], and treatment with vismodegib in a rat model of mesothelioma reduced the expression of target genes such as GLI1, HHIP and PTCH1 [49]. Here, we describe alterations in Hedgehog signalling genes PTCH1 (1.2%) and SUFU (0.8%) in the whole mesothelioma cohort (Fig. 4). Even though alterations in the Hedgehog signalling pathway are rare, these can be targeted through commercially available drugs. Hippo pathway. Another gene involved in cell-cycle regulation is the tumour suppressor gene NF2, encoding merlin. In our study, NF2 was altered in 32.8% of pleural mesothelioma and 26.5% of peritoneal mesothelioma cases (Supplementary Table 1). NF2 can regulate the Hippo pathway via signalling trough YAP and TAZ. Inactivation of NF2 can lead to hyperactivated YAP and to uncontrolled cell proliferation, thus, YAP inhibitors are currently tested in cell lines and could serve as an option in patients with uncontrolled YAP signalling [50]. NF2 also plays a role in cell-cycle regulation and mTORC1 signalling; inactivation of NF2 led to PAK1-induced increase of cyclin D1 and consequently to increased cell proliferation [51]. Rapamycin, an mTOR inhibitor, can inhibit proliferation of NF2-altered mesothelioma cells in vitro [52]. However, mTOR inhibitors such as everolimus had limited clinical activity in a Phase II clinical trial in an unselected mesothelioma cohort [53]. Currently, second and third-generation mTOR inhibitors and PI3K/mTOR inhibitors are under investigation in other malignancies [54].

DISCUSSION
This is the largest analysis formed on mesothelioma samples profiled by NGS so far. Thank to this broad number of cases we could identify four subtypes of mesothelioma according to their molecular genetic alterations in CDKN2A, CDKN2B and BAP1. We also detected the occurrence of rare genomic alterations, which can be targeted with FDA-approved or experimental drugs. Indeed, for several tumour types, patient stratification allows personalised treatment approach and improved survival.
Currently, mesothelioma patients are classified according to their histological subtype, however, due to major differences in their genomic features, genomic profiling may add value for treatment stratification and might refine future treatment strategies. Moreover, genomic subgroups might be due to different tumorigenesis mechanisms as already described in breast cancer, where oncogenic alterations lead to different tumour types with different clinical outcomes (reviewed in ref. [55]). Hmeljak et al. performed clustering of molecular data using two different algorithms and defined four different subgroups with different survival properties [9]. In our study, we identified four distinct subgroups of patients based on their genomic background ( Fig. 3 and Tables 1 and 2). Those different genomic subgroups might have different clinical characteristics and response to treatment, and this finding warrants further investigations. Preclinical models of mesothelioma confirm that specific alterations can lead to different outcomes, for example, CDKN2A loss leads to poor outcomes [56] and BAP1 alterations in combination with NF2 and CDKN2A/B to rapid disease onset [17].
We defined four different groups with significantly different prevalence of alteration in BAP1 and CDKN2A genes. For example, alterations in PBRM1 and SETD2 were mainly associated to Group 3 (BAP1 alteration only), while TP53 alterations occurred mainly in Group 4 (no CDKN2A/BAP1 alteration) in pleural mesothelioma. Patients with a combined loss of TP53 and RB1 might even have worse prognoses as described in different malignancies [57,58]. Interestingly, the patients in Group 4 had a worse survival in the TCGA cohort compared to Group 3 with only BAP1 alteration, which might be due to the significantly higher mutation rate in TP53 and RB1 ( Supplementary Fig. 2). However, further studies are needed to understand the differences in therapy response in these molecular subgroups.
The most common alterations occurred in BAP1, NF2 or deletions in CDKN2A/B, in line with the Catalogue of Somatic Mutations in Cancer (COSMIC) database [59] and in cBioportal [18]. Compared to previously reported data, we detected a much higher frequency of BAP1, CDKN2A and NF2 alterations in our cohort and this might be due to a larger cohort size. In our cohort, NF2 is altered in 32.8% of pleural and 26.5% of peritoneal mesothelioma, compared to other studies with alterations in about 20% of the cases [13,31,34,60]. Similarly, BAP1 alterations were more frequently detected (45.0% pleural and 47.9% peritoneal mesothelioma), compared to the studies from Bueno et al. and Quetel et al. (with 23% and 24.4%, respectively) [13,60]. However, this prevalence is consistent with what has been recently observed in low-risk mesothelioma patients [15]. Interestingly, deletions in CDKN2A/B occur more frequently in pleural compared to peritoneal mesothelioma, as previously described [61,62]. A possible explanation is that chromosome arm deletions are associated with asbestos exposure and, therefore, a possible different causality for peritoneal compared to pleural mesothelioma. Few cases showed an increased TMB, therefore it is not possible to discriminate if co-occurring alterations were truly enriched and functionally important, rather than just passenger by-products of any mutated gene.
In addition, this analysis revealed alterations in 19 genes with a prevalence higher than 2%. Moreover, co-alterations showed the clustering of different groups of patients with mutations in BAP1, PBRM1, SETD2; or in MYC, TERT, RB1 and TP53 or in CDKN2A/B, NF2 and BAP1. Despite the occurrence of alterations in several genes and co-mutations, cases with increased tumour mutational burden were rare, but leave the option for immunotherapy with immune checkpoint inhibitors in some patients. A broad analysis of the mutation spectra might become a stratification factor for patients receiving immunotherapy. Analysing mutations in genes correlated with response to treatment is of major importance for patient selection. In lung cancer, patients with STK11 mutations showed resistance to anti-PD-1 inhibitors [63], and deletions in CDKN2A were negatively correlated with T-cell markers in different cancer types [64]. These findings warrant further investigations about differences in tumour development and possibly responses to treatment. Based on our data, targeted treatments might also become a possible approach for mesothelioma, as alterations in hedgehog pathway-related genes (e.g., PTCH1/2 and SUFU) and hippo pathway-related genes (particularly NF2) were found. Moreover, we detected alterations in KRAS, EGFR, PDGFRA/B, ERBB2, FGFR3 and ALK. Even if those alterations are rare, their detection and the use of targeted treatments can change patient outcomes. We have recently shown that targeting a STRN-ALK rearrangement with an ALK-inhibitor in a patient with peritoneal mesothelioma, could lead to prolonged response [45] and Popat et al. presented a case with PTCH1 alteration where vismodegib treatment led to good response [65].
A major limitation of our study is the lack of matching clinical data about histology, outcome, treatment allocation and response to treatment; this was not possible due to the retrospective nature of this analysis. Nevertheless, clinical data from the TCGA cohort show significant differences in overall survival between the four defined groups but no differences in sex and histology. These findings might help understand mechanisms related to tumour development and provide novel insights into genomic subgroups for patient stratification for clinical trials. A recently opened clinical trial (MiST) is the first to stratify patients according to their molecular alterations, where BRCA1 mutated/BAP1 negative patients receive PARP inhibitors, while CDKN2A negative patients a CDK4/6 inhibitor or immunotherapy depending on their PD-L1 expression. However, in this study, the specified molecular alterations were measured by IHC assays, no NGS was performed for patient stratification, which might open the possibility of more tailored treatment approaches. In addition, the advantage of NGS lays in its cost-effectiveness, in delivering reliable results without interpersonal differences for interpretation, and it can save material compared to multi-step testing strategies.

CONCLUSION
Precision medicine, including high-throughput genomic screening, has tremendously improved the outcome of patients with a high prevalence of mutations, in particular in patients with lung cancer, breast cancer and melanoma. Nevertheless, also in rare malignancies and in malignancies with low numbers of somatic mutations, targetable mutations need to be analysed in order to enlarge treatment options for difficult-to-treat cancers. Our results indicate, that molecular analysis for mesothelioma should be implemented.

DATA AVAILABILITY
All relevant data are provided in the manuscript. Due to HIPAA requirements, FMI is not consented to share individualised patient genomic data, which contains potentially identifying or sensitive patient information. FMI is committed to collaborative data analysis, and we have well-established, and widely utilised mechanisms by which investigators can query our core genomic database of >500,000 de-identified sequenced cancers to obtain aggregated datasets.