Background
Malignant mesothelioma (MM) is an aggressive cancer caused by DNA damage in the mesothelial cells of the pleural and peritoneal cavities that principally results following asbestos exposure [
1‐
3]. There is a long latency between asbestos exposure and tumour development with periods of 30–50 years frequently reported. MM is almost universally fatal and median survival after diagnosis is short (9–12 months) [
1,
2].
Asbestos is one of only a few carcinogens, along with cigarette smoke, UV light and several others, that can be readily identified in individuals. Importantly, asbestos induces mesotheliomas in mice which are almost identical to their human counterpart in terms of pathology, immunology and clinical behavior, which is rare in mouse modeling of cancer [
4,
5]. Although this has made the mouse model a useful preclinical tool for studying mesothelioma treatments, until now there has been no detailed study of the genetic lesions associated with mouse mesothelioma to determine if such lesions also parallel the human disease. Detailed studies of the genetic lesions of human MM have only recently been published. On average, human MM tumours contain less than 1 mutation per million bases, which is lower than other malignancies associated with external carcinogens such as lung cancer and melanoma [
6‐
8]. A variety of somatic copy number variations (CNVs) and single nucleotide variations (SNVs) have been recurrently reported in several genes. Loss of
CDKN2A, located at 9p21, is reported in more than 70% of cases; loss of
NF2, located at 22q, is reported in around 40% of cases; and mutations in
BAP1, located at 3p21.1 are reported in 40–60% of cases [
9‐
13]. Recent large scale genomic analyses have also found recurrent, low frequency mutations in a number of additional genes including
LATS2 (22%)
, CUL1 (9%)
, TP53 (8%) and
SETD2 (8%) [
7,
8,
14]. Shared somatic mutations between MM tumours are rare, however several biological pathways are often reported as being dysregulated in human MM, including the Wnt, Hedgehog, Notch, Ras, p53, MAPK, mTOR and Hippo signaling pathways [
7,
15,
16].
Murine models of MM are an invaluable tool for the preclinical evaluation of disease pathogenesis and for developing novel treatment strategies [
17]. Recent studies have utilized mouse xenograft models and several genetically engineered mouse models to recapitulate the common mutations seen in human MM, such as
Nf2,
Cdkn2a and
Bap1 knockout models [
18‐
20]. However, such models only enable the study of MM in the context of the effect of the knocked out gene of interest. Our well established asbestos-induced wild-type murine MM model has the potential to offer useful molecular insights on the natural initiation and progression of MM in response to asbestos exposure, providing the opportunity for better understanding of pathogenesis, development of novel treatments and biomarker/signature discovery. Detailed characterisation of the genomic lesions underpinning the wild-type murine MM model have, as in other cancers, lagged behind the relevant human studies. This model has been extensively characterised at the phenotypic and morphological level on the BALB/c background [
5]. Gene expression has previously been characterised in the C57BL/6 strain [
21] and array-comparative genomic hybridisation (aCGH) studies have identified lesions in FVB/N mice [
22] however little is known about the mutational landscape of these wild-type tumours. We therefore undertook to characterize the somatic DNA lesions that underlie murine MM and characterise the mutational landscape using whole exome sequencing of fifteen independent murine MM tumour cell lines derived from three murine strains, comparing the mutations identified with those most often found in human MM.
Discussion
The asbestos-induced wild-type murine model of MM has the power to provide invaluable insights into the pathology of human MM, provided that the tumours are similar on a molecular basis between the species. Through whole exome sequencing of tumour cell lines developed from mesotheliomas from three murine strains, we have shown that the
Cdkn2a gene is consistently lost in the asbestos-induced wild-type murine model of MM. This parallels the situation in more than 70% of human MM tumours [
12]. Previously, in a genetically engineered MM mouse model, heterozygous for
Nf2
+/−, a high rate of homozygous deletion of
Cdkn2a was observed in MM tumours induced by asbestos exposure [
19]. In the human context
CDKN2A loss, deletion of
NF2 and/or mutation of
BAP1 has been reported in up to 80% of clinical cases [
8]. In our model we did not find mutations, copy number deletions or reduced mRNA expression of
Bap1, however, loss of BAP1 protein expression has been reported in up to 25% of cases where no genetic mutations were detected [
9]. Dysregulation of the Hippo pathway is often reported in human MM, typically through mutations or copy number aberrations of
NF2,
YAP and/or
LATS2 [
14,
48]. No mutations in
Lats2 were observed in this study, however 20% of samples in this study contained a copy number deletion of
Lats2. With post-transcriptional modification potentially accounting for loss of these tumour suppressors in the mouse asbestos induced MM model it is not possible to conclusively say that
Cdkn2a deletion is an independent driver of MM. However it seems that homozygous loss of the
Cdkn2a gene is sufficient for the development of asbestos-induced MM in the absence of other common MM driver mutations at the genetic and transcriptomic level in this model.
Though most human MMs develop in the pleura, the murine model used in this study developed in the peritoneum. Despite this difference, the developmental characteristics of murine MM faithfully mimic human MM [
4]. Pleural and peritoneal MM are different in terms of prognosis and treatment options, however the molecular features of peritoneal MM are not as well characterised as pleural MM, given the rarity of the diagnosis. It is not uncommon to have secondary pleural invasion from peritoneal MM and vice versa, and in these cases the distinction between the two types is clinically difficult [
49]. The histological feature of pleural and peritoneal MM are generally identical, and on this basis, the molecular characterisation of murine MM in this study will provide the necessary basis for future studies seeking to target specific molecular pathways for therapy.
The asbestos-exposed wild-type murine model of MM displayed a high proportion of chromosomal loss compared to gain, which is a characteristic feature of human MM. Human MM tumours display recurrent losses in 1p21-p22, 9p21-p22, 3p21, 6q15-q21, 17p13.1 and 22q12.2 [
50,
51]. Taking into account the locations of
BAP1,
NF2 and
CDKN2A in the human genome, the only orthologous region containing CNV aberration in this model was the region containing
Cdkn2a (4qC4). Previously we have shown using our MexTAg transgenic mouse model [
24] that the presence of exogenous SV40 large T antigen (Tag) performs a similar molecular role as
Cdkn2a loss [
21] In both the wild type and MexTAg model, asbestos-induced MM was not associated with reduced expression of other genes commonly down-regulated in human MM [
21].
To date no common oncogenic driver has been identified in MM, although amplification of the C-MYC locus in MM has been reported to occur in MM human tumour cell lines and may contribute to a malignant phenotype [
52]. In the current study, increased expression of c-Myc was observed, however no other recognized oncogenes were identified in significantly amplified regions across the samples. It is important to acknowledge that, despite these particular CNVs being previously reported in human MM, the selection pressure of culture conditions may account for some CNV in this study. Notably, however, there was no correlation between the number of CNVs and time in culture in this study.
Carcinogen-induced cancers such as melanoma and lung cancer typically show a high rate of mutation when compared to spontaneous cancers [
6‐
8]. Given that MM is also a carcinogen induced tumour, the average somatic mutation rate in human MM is lower than expected [
7,
8]. Here we show that the CBA and C57BL/6 strains of the wild-type murine model of MM display a comparable mutation rate to human MM. Interestingly, the BALB/c strain showed a wider range of mutation rates between the samples. Previously, germline mutations in a key DNA damage sensor gene,
Prkdc, have been reported in some colonies of BALB/c mice [
53]. These mutations were also present in the BALB/c mice used to generate the cell lines used in this study, possibly resulting in the current and previously reported genomic instability of tumour cell lines from this strain [
5] and may account for the higher rate of spontaneous mutation when exposed to asbestos.
Patterns of mutations can be used to identify mutational signatures which represent the underlying process of DNA damage, (for example tobacco smoking induces a mutational pattern that exhibits transcriptional strand bias for C > A mutations and CC > AA dinucleotide substitutions) [
6]. To date no asbestos-related mutation signature has been identified. The current exome sequencing study was underpowered to perform a full mutational signature analysis but did show a high rate of C > T and G > A mutations, consistent with previous studies in humans [
7,
8]. Whole genome sequencing of a larger cohort of samples will be required to fully answer this question in both humans and mice.
Given the genetically independent nature of MM tumours and the low likelihood of the same gene being mutated at different positions in more than one sample, the presence of different mutations in
Nkd1 in two BALB/c samples and
Cacna2d2 in samples from two different strains is informative. The mutation of
Nkd1 may implicate the role of Wnt signaling pathway regulation in the BALB/c strain. Genes involved in the Jak-STAT and MAPK signaling pathways were also altered in the murine model. These Wnt, Jak-STAT and MAPK pathways have been identified as being altered in many cancers, including MM. In humans,
CACNA2D2 is located 1.9 Mb upstream from
BAP1 in the region 3p21. However, in the mouse genome, the orthologs reside on different chromosomes. The importance of this gene in MM in undetermined, though it has been linked to tumorigenesis in prostate cancer [
46] and loss of expression has been demonstrated in non-small cell lung cancer [
54].
Acknowledgements
Not applicable.