Novel LRRK2 mutations and other rare, non-BAP1-related candidate tumor predisposition gene variants in high-risk cancer families with mesothelioma and other tumors

Abstract

There is irrefutable evidence that germline BRCA1-associated protein 1 gene (BAP1) mutations contribute to malignant mesothelioma (MM) susceptibility. However, BAP1 mutations are not found in all cases with evidence of familial MM or in other high-risk cancer families affected by various cancers, including MM. The goal of this study was to use whole genome sequencing (WGS) to determine the frequency and types of germline gene variants occurring in 12 MM patients who were selected from a series of 141 asbestos-exposed MM patients with a family history of cancer but without a germline BAP1 mutation. WGS was also performed on two MM cases, a proband and sibling, from a previously reported family with multiple cases of MM without the inheritance of a predisposing BAP1 mutation. Altogether, germline DNA sequencing variants were identified in 21 cancer-related genes in 10 of the 13 probands. Germline indel, splice site and missense mutations and two large deletions were identified. Among the 13 MM index cases, 6 (46%) exhibited one or more predicted pathogenic mutations. Affected genes encode proteins involved in DNA repair (ATM, ATR, BRCA2, BRIP1, CHEK2, MLH3, MUTYH, POLE, POLE4, POLQ and XRCC1), chromatin modification (ARID1B, DNMT3A, JARID2 and SETD1B) or other cellular pathways: leucine-rich repeat kinase 2 gene (LRRK2) (two cases) and MSH4. Notably, somatic truncating mutation or deletions of LRRK2 were occasionally found in MMs in The Cancer Genome Atlas, and the expression of LRRK2 was undetectable or downregulated in a majority of primary MMs and MM cell lines we examined, implying that loss of LRRK2 expression is a newly recognized tumor suppressor alteration in MM.

Introduction

Malignant mesothelioma (MM) is an aggressive cancer generally associated with asbestos exposure. Individuals with MM, especially the pleural form of this disease, have a poor prognosis owing to the dearth of opportunities for early surgical intervention and relatively ineffective chemotherapies (1,2). MM causes about 3200 deaths annually in the USA, and the incidence of MM is expected to increase by 5–10% per year over the next two decades in Europe (1,2).

In 2011, we reported germline mutations of the BRCA1-associated protein 1 gene (BAP1) in two families, dubbed L and W, with high incidence of MM and other cancers (3). Immunohistochemical analysis of MMs from these families revealed a lack of BAP1 nuclear expression and only weak cytoplasmic BAP1 staining (3). Interestingly, in the L family, two family members had uveal melanoma (UM), one of whom later developed MM. Notably, Harbour et al. (4) had previously uncovered inactivating somatic mutations of BAP1 in ~85% of metastasizing UMs, and one of their patients had a germline mutation in BAP1, suggesting the existence of a tumor susceptibility allele in this individual. Moreover, in the same issue of Nature Genetics as our report (3), Wiesner et al. (5) described inactivating germline BAP1 mutations in two families with multiple benign melanocytic tumors; furthermore, some of the affected individuals developed cutaneous melanomas or an occasional UM, and one family member was later diagnosed with peritoneal MM (6). Since then, there have been many other reports documenting germline BAP1 mutations in families with MM and various other cancers (7–10). Collectively, these findings suggested the existence of a single BAP1 tumor predisposition syndrome (BAP1-TPDS) in which affected families are predisposed primarily to MM, UM, cutaneous melanoma, benign melanocytic tumors, clear cell renal cell carcinoma and basal cell carcinoma (11–13). In a comprehensive assessment of the clinical phenotype of BAP1 variant-carrying families worldwide, the classically described core tumor spectrum for the BAP1-TPDS has been expanded to include meningioma and cholangiocarcinoma based on the associated much higher incidence of these two rare neoplasms than in the general population as well as the molecular evidence from tumors of these carriers (11–13). Other unconfirmed tumors in the BAP1-TPDS are thought to include breast cancer, non-small cell lung adenocarcinoma and neuroendocrine carcinoma (12). Biallelic inactivation of BAP1 has been documented in multiple tumors from these high-risk BAP1 families, implying that BAP1 acts as a classical tumor suppressor gene. Consistent with these riveting genetic discoveries, BAP1 has also been found to exhibit tumor suppressor activity both in cell-based transfection assays and in vivo studies with genetically engineered mice (14). Furthermore, tumor suppression requires both nuclear localization and BAP1 deubiquitinase activity (15).

Despite the identification of germline mutations in BAP1 as predisposing to MM and other cancers, there are a number of recent reports indicating that this gene may not account for all examples of familial MM or for all high-risk cancer families with at least one MM (16–20). For example, in one study, an Italian family with eight confirmed cases of MM did not exhibit a predisposing BAP1 mutation (20). In another study, we sequenced the entire BAP1 gene in blood samples from 150 MM cases with a history of asbestos exposure and a past personal or family history of one or more of the cancers reported in BAP1-TPDS families. These cases all had high ‘cancer signal’ scores, based on an objective scoring system designed to identify MM individuals with a strong family history of cancers, including at least some known to be part of the BAP1-TPDS, particularly MM, cutaneous and UM and renal cell carcinoma (18). Sequencing of BAP1 in these 150 MM cases uncovered nine patients (6%) with germline mutations affecting the coding sequence of the BAP1 gene (18). Here, whole genome sequencing (WGS) was performed on germline DNA from 12 of the remaining 141 MM cases without BAP1 mutations (18). These 12 MM cases were selected from families having a high overall incidence of cancer generally, not necessarily a personal or family history strongly indicative of BAP1-TPDS. In addition, WGS was also performed on two additional MM cases from the Italian family that lacked a germline BAP1 mutation (20).

Results

Mutated candidate genes were selected for further evaluation based on several criteria: (1) DNA variants present at <0.05% allele frequency in the general population, as determined using the ExacDB database and a Combined Annotation Dependent Depletion tool for scoring the deleteriousness of SNPs and insertion/deletion variants in the human genome (CADD) score greater than 20; (2) the candidate gene has been previously implicated in cancer either somatically or hereditarily, or is known to be connected with a cellular process thought to be a hallmark of cancer (21); (3) greater priority given to mutations known to affect protein-coding sequence (e.g., nonsense, indel and splice site mutations); and (4) genes mutated in more than one family. Mutations in selected candidate genes were confirmed by Sanger sequencing. Based on these criteria, germline variants were identified in 21 genes (1 involving heterozygous loss of the entire gene) that were considered to be the most promising candidates (Tables 1 and 2). Eleven of the candidate genes [ATM, ATR, BRCA2, BRIP1, CHEK2, MLH3, MUTYH, POLE, DNA polymerase epsilon 4, accessory subunit (POLE4), POLQ and X-ray repair cross-complementing 1 (XRCC1)] encode proteins that have roles in DNA repair, whereas four (ARID1B, DNMT3A, JARID2 and SETD1B) encode proteins involved in chromatin modification. In several MM cases, three or four candidate tumor susceptibility gene variants were identified in the same individual. Interestingly, three of the genes shown in Table 1 were found to be mutated in more than one proband: CHEK2 (cases ABS3383 and ABS2640), leucine-rich repeat kinase 2 gene (LRRK2) (cases ABS2640 and 946-P) and POLQ (cases ABS3460 and ABS3505). Each of these same three genes, as well as several others, was found to be mutated/deleted somatically in pleural MM tumors in The Cancer Genome Atlas (TCGA) database. The germline mutations in our cases consisted of indel (predicted frameshift), splice site and missense mutations as well as large exonic deletions. Among the 13 index cases reported here, 8 (61.5%) had DNA sequencing variants that are predicted to be pathogenic mutations (e.g., indels), which are described later.

Table 1.

Candidate genes with germline DNA sequencing variants identified by WGS

Symbol	Gene	Function	Mutation	Samples
APC	APC regulator of Wnt signaling pathway	Tumor suppressor that antagonizes the Wnt signaling pathway	Missense	ABS3425
ARID1B	AT-rich interaction domain 1B	Component of the SWI/SNF chromatin remodeling complex in cell cycle activation	Missense	ABS3460
ATM	ATM serine/threonine kinase	Double-strand break DNA damage sensor	Missense	R88-T
ATR	ATR serine/threonine kinase	Single-strand break DNA damage sensor	Missense	ABS3481
BRCA2	BRCA2 DNA repair associated	Homologous recombination-mediated repair of double-strand DNA breaks	Indel	946-P
BRIP1	BRCA1 interacting protein C-terminal helicase 1	Member of RecQ DEAH helicase family that complexes with BRCA1 for double-strand break DNA repair	Missense	ABS3505
CBFA2T3	CBFA2/RUNX1 translocation partner 3	Transcription factor-binding protein that facilitates co-repressor recruitment	Missense	ABS3505
CHEK2	Checkpoint kinase 2	Blocks cell cycle progression in response to DNA damage; linked to Li-Fraumeni syndrome and confers a predisposition to sarcomas, breast cancer and brain tumors	Missense Large del.	ABS3383 ABS2640
DACT2	Disheveled-binding antagonist of β-catenin 2	Tumor suppressor inhibiting Wnt/β-catenin pathway	Missense	ABS2640
DNMT3A	DNA methyltransferase 3 alpha	CpG DNA methylation; interacts with EZH2, a histone methyltransferase	Indel	ABS3425
JARID2	Jumonji and AT-rich interaction domain containing 2	DNA-binding protein that complexes with PRC2 to repress transcription	Missense	ABS3481
LRRK2	Leucine-rich repeat kinase 2	Leucine-rich repeat kinase whose germline activating missense mutation causes Parkinson’s disease-8	Missense Splice site	ABS2640 946-P/R88-T
MLH3	MutL homolog 3	Repair of DNA mismatches	Missense	ABS3572
MSH4	MutS protein homolog 4	Meiotic homologous recombination	Indel	ABS2406
MUTYH	MutY DNA glycosylase	DNA glycosylase involved in oxidative DNA damage repair	Splice site	ABS2813
POLE	DNA polymerase epsilon, catalytic subunit	Involved in DNA repair; missense germline mutation causes colorectal polyposis	Missense	ABS3383
POLE4	DNA polymerase epsilon subunit 4	Accessory component of the DNA polymerase epsilon complex involved in DNA repair	Gene del.	ABS3425
POLQ	DNA polymerase theta	Mediates microhomology-mediated end-joining repair of double-strand DNA breaks	Nonsense Missense	ABS3460 ABS3505
RHBDF2	Rhomboid 5 homolog 2	Regulates secretion of several EGFR ligands; activating germline mutations causes esophageal cancer and hyperkeratotic skin lesions	Missense	ABS3505
SETD1B	SET domain containing 1B	Histone H3 K4 trimethylation (active transcription)	Missense	ABS3460
XRCC1	X-ray repair cross-complementing 1	Repair of DNA single-strand breaks through base excision repair pathway	Indel	ABS3460

Table 2.

Variants observed in germline DNA from 14 MM patients

Patient	Gene(s)	Variant(s)	Predicted Protein Change(s)	Type of variant(s)	ExAC_All %; CADD score
ABS2406	MSH4	NM_002440.4:c.719dupT	NP_002431.2:p.Ile240fs	Indel	Novel; 35
ABS2640	CHEK2   DACT2   LRRK2	NM_007194:c.909-2028_1095+330del5395 NM_214462.3:c.533C>T NM_198578.3:c.6055G>A	NP_009125.1:p.M304Lfs*16 NP_999627.2:p.Ser178Leu NP_940980.3:p.Gly2019Ser	Del. (exons 9-10) Missense Missense	Novel; NA Novel; 25.9 0.0004; 32
ABS2813	MUTYH	NM_001128425.1:c.389-1G>A	Frameshift / truncation	Splice site	Novel; 33
ABS3425	DNMT3A   POLE4   APC	NM_022552.4:c.2631delC Whole gene deletion NM_000038.5:c.1642T>G	NP_072046.2:p.Ser689fs No protein expression NP_000029.2:p.Leu548Val	Indel Whole gene del. Missense	Novel; 35 Novel; NA Novel; 22.7
ABS3460	POLQ   XRCC1   SETD1B   ARID1B	NM_199420.4:c.3589C>T NM_006297.2:c.175delG NM_015048.1:c.2554C>T NM_017519.2:c.2405C>T	NP_955452.3:p.Arg1197* NP_006288.2:p.Asp59fs NP_055863.1:p.Arg852Cys NP_059989.2:p.Ser802Leu	Nonsense Indel Missense Missense	1.65E-05; 35 8.24E-06; 35 5.11E-05; 24 0.0004; 32
946-P	BRCA2   LRRK2	NM_000059.3:c.657_658del NM_198578.3:c.5314_5317+6delAAAGGTAAGG	NP_000050.2:p.Thr219fs Frameshift/truncation	Indel Indel at splice site	6.12E-05; 22 Novel; 35
R88-T	LRRK2   ATM	NM_198578.3:c.5314_5317+6delAAAGGTAAGG NM_000051.4:c.1837G>T	Frameshift/truncation NP_000042.3:p.Val613Leu	Indel at splice site Missense	Novel; 35 4.26E-05; 25.5
ABS3383	CHEK2   POLE	NM_007194:c. 1451C>T NM_006231.2:c.177G>C	NP_009125.1:p.Pro484Leu NP_006222.2:p.Lys59Asn	Missense Missense	0.0001; 29.7 4.12E-05; 21.7
ABS3481	ATR   JARID2	NM_001184.4:c.4351C>T NM_004973.4:c.3682G>A	NP_001175.2:p.Arg1451Trp NP_004964.2:p.Val1228Met	Missense Missense	0.0003; 26 6.59E-05; 23.7
ABS3505	POLQ   BRIP1   CBFA2T3   RHBDF2	NM_199420.4:c7024_7026TAA>AAT NM_032043.3:c.2220G>T NM_005187.5:c.956G>A NM_024599.5:c.1726A>G	NP_955452.3:p.Leu2342Ile NP_114432.2:p.Gln740His NP_005178.4:p.Cys319Tyr NP_078875.4:p.Ile576Val	Missense Missense Missense Missense	0.0002; 24.2 0.0005; 23.7 1.76E-05; 23.7 2.02E-05; 21.6
ABS3572	MLH3	NM_001040108.2:c.G3455G>A	NP_001035197.1:p.Arg1152His	Missense	6.59E-05; 24.1
ABS2586	No significant genes identified
ABS3444	No significant genes identified
MC7010	No significant genes identified

NA, not applicable.

M‌M case ABS2406

This individual was found to have an indel mutation involving a member of the mismatch repair (MMR) family: MSH4 (c.719dupT; p.Ile240fs) as well as a non-frameshift deletion of 18 bp in SMARCB1 (c.56_73del; p.20_25del). The deletion in SMARCB1 is predicted to affect residues located within the DNA-binding domain of the protein. In addition to MM, case ABS2406 had a rhabdomyosarcoma and a Schwannoma, and her daughter also developed more than one tumor (Fig. 1A).

Figure 1 .

Pedigrees of three families in which a proband (arrows) with MM has a predicted pathogenic germline mutation. (A) Family of patient ABS2406, who was found to have an indel mutation involving MSH4 (c.719dupT; p.Ile240fs). In addition to MM, this proband had a rhabdomyosarcoma and a Schwannoma. (B) Family of MM patient ABS2640m who had a 5395 bp deletion of CHEK2 exons 9 and 10 (c.909-2028_1095 + 330del5395; p.M304Lfs^*16) was predicted by several WGS structural variation analysis programs and confirmed by Sanger sequencing of a PCR product encompassing the junction created by the deletion (C). This individual also had a germline pathogenic kinase-activating mutation of LRRK2 (c.6055G>A; p.Gly2019Ser).

M‌M case ABS2640

An unusual alteration of CHEK2 was found in case ABS2640 (Fig. 1B). A large (5395 bp) deletion of CHEK2 exons 9 and 10 (c.909-2028_1095 + 330del5395; p.M304Lfs^*16) was predicted by several WGS structural variation analysis programs (Manta, Delly and GRIDSS), and we confirmed this by Sanger sequencing of a PCR product encompassing the novel junction created by the deletion (Fig. 1C). This individual also had a germline kinase-activating mutation (c.6055G>A; p.Gly2019Ser) of LRRK2, which encodes leucine-rich repeat kinase 2. Additionally, we found a missense mutation (c.533C>T; p.Ser178Leu) in DACT2 (disheveled antagonist of beta catenin 2), a tumor suppressor gene whose protein product inhibits the Wnt/β-catenin pathway.

M‌M case ABS2813

In this MM patient (Fig. 2A), a germline splice site mutation, c.389-1G>A, was present in the MutY DNA glycosylase gene, MUTYH. This mutation is predicted to lead to a frameshift and protein truncation. The proband has a father with MM as well as a son with cutaneous melanoma, but unfortunately, samples from either relative were not available for genetic testing.

Figure 2 .

Pedigrees of families in which a proband (arrows) with MM has one or more predicted pathogenic germline alterations. (A) Family of patient ABS2813, who had a germline pathogenic MUTYH splice site mutation, c.389-1G>A. The proband’s father also had MM and his son had cutaneous melanoma. (B) Family of patient ABS3425, who has had both peritoneal MM and basal cell carcinoma, and has a DNMT3A indel mutation (c.2631delC; p.Ser689fs) as well as an ~269 kb heterozygous germline chromosomal deletion at chromosome 2p12 (NC_000002.12:g.(74816393_75085246)del of the GRCh38.p13 reference genome).

M‌M case ABS3425

In this individual, proband (II-2) in the pedigree is shown in Figure 2B, and an indel mutation (c.2631delC; p.Ser689fs) was observed in the DNA methyltransferase 3A gene, DNMT3A. This mutation is predicted to cause a frameshift. This patient has had both peritoneal MM and basal cell carcinoma, and multiple members of his family have been affected by various carcinomas. Interestingly, we also found an APC missense mutation (c.1642T>G; p.Leu548Val) in this same proband. Notably, the proband’s mother (I-2 in the pedigree in Fig. 2B) and grandmother had colorectal cancer, but DNA samples from these two individuals were not available to test whether they also harbored the same germline APC mutation. The DNMT3A and APC mutations were not present in the DNA isolated from saliva of a brother (II-3) who had thyroid cancer, indicating that these genes did not play a role in this family member’s cancer. Besides these two mutations, several structural variation software analysis programs (Manta, Delly, Lumpy and SvABA) uncovered a large (~269 kb) heterozygous germline chromosomal deletion at chromosome 2p12 [NC_000002.12:g.(74816393_75085246)del of the GRCh38.p13 reference genome] in case ABS3425. This heterozygous deletion encompasses all or part of several protein-coding genes: POLE4, hexokinase 2 (HK2) and tachykinin receptor 1 (TACR1) (Fig. 3, top). Sanger sequencing of a PCR product using primers flanking the subsequent junction sequence confirmed the predicted deletion (Fig. 3, bottom).

Figure 3 .

Predicted structure and sequencing confirmation of the large germline deletion at chromosome 2p12 in patient ABS2813. Mapping of the heterozygous ~269 kb germline deletion encompasses POLE4, HK2 and part of TACR1 (top). Sanger sequencing of a PCR product of genomic DNA, using primers flanking the consequent junction sequence confirming the predicted deletion (bottom).

M‌M case ABS3460

This MM patient harbored sequencing variants in four notable cancer-related genes: POLQ, XRCC1, SETD1B and ARID1B. Remarkably, his family history included multiple and varied carcinomas, and both he (II-5) and his brother (II-6) had MM and one or more other cancers (Fig. 4). Moreover, case ABS3460 developed both pleural and peritoneal epithelioid MM. The nonsense mutation in POLQ (c.3589C>T; p.Arg1197^*) represents an inactivating mutation, and the alteration of XRCC1 (c.175delG; p.Asp59fs) consisted of an inactivating indel mutation. In addition, germline missense mutations were identified in the SET domain containing 1B gene, SETD1B (c.2554C>T; p.Arg852Cys), which encodes a component of the histone methyltransferase complex that produces trimethylated histone H3 on K4, and in ARID1B (c.2405C>T; p.Ser802Leu), a gene that encodes a protein that is part of the SWI-SNF chromatin remodeling complex involved in cell cycle activation. Notably, annotation of ARID1B protein post-translational modifications (www.phosphosite.org) provided evidence that Serine 802 can be phosphorylated; thus, loss of this serine in case ABS3460 could have a consequential effect on ARID1B protein function.

Figure 4 .

Family pedigree of patient ABS3460, who has had both pleural and peritoneal epithelioid MM as well as basal cell carcinoma and squamous cell carcinoma. The proband (arrow) had a brother who had MM and liver cancer, and there have been several other carcinomas in this family. The proband has germline inactivating mutations in both POLQ (c.3589C>T; p.Arg1197^*) and XRCC1 (c.175delG; p.Asp59fs) as well as missense mutations in SETD1B (c.2554C>T; p.Arg852Cys) and in ARID1B (c.2405C>T; p.Ser802Leu) that were not predicted to be pathogenic.

M‌M cases R88-T and 946-P

Lastly, in a previously reported Italian family with multiple cases of MM (Fig. 5A) (22), WGS uncovered a germline deletion at a splice site (c.5314_5317 + 6delAAAGGTAAGG) in the LRRK2, which is predicted to lead to a frameshift and protein truncation. This mutation in LRRK2 was identified in DNA from peripheral blood lymphocytes of family member III-5 (Fig. 5C) as well as in DNA isolated from tumor tissue from family member III-2, the latter containing a mixture of MM and normal stromal cells. In both cases, the mutation was present in ~45% of the WGS reads, indicating heterozygosity of the mutation. In addition, DNA from a macrodissected MM tumor sample obtained from individual III-1 was also available, and this sample also harbored the LRRK2 mutation. Notably, the mutation in this macrodissected specimen was in a homozygous (or hemizygous) state, indicating loss of heterozygosity (LOH) (Fig. 5D). The LRRK2 mutation was not present in DNA isolated from normal formalin-fixed paraffin-embedded (FFPE) tissue in family member III-3 (Fig. 5B), who did not have MM. Interestingly, an inactivating indel mutation of BRCA2 (c.657_658del; p.Thr219fs) was found in the germline of individual III-5 (946-P) but not in the two MM tumor samples from her sister, III-2 (R88-T). This was unforeseen, given that III-5 (946-P) did not develop breast cancer, whereas III-2 (R88-T) did. MM tumor tissue from III-5 had the BRCA2 mutation (data not shown). When we performed Sanger sequencing of the gene in the MM tumors of other family members, we found that BRCA2 was WT in III-01 but mutant in III-07, who developed breast cancer. Since the germline BRCA2 mutation was observed in only two (III-5, III-07) of the four family members tested, who developed MM, it is unclear whether this mutation also contributes to MM susceptibility in some members of this family.

Figure 5 .

Novel alteration of a leucine-rich repeat kinase gene, LRRK2, associated with familial MM. (A) Partial pedigree of an asbestos-exposed family with multiple cases of pleural MM (22). WGS was performed on peripheral blood sample 946-P from family member III-5 and on MM tumor specimen R88-T from individual III-2. (B) Germline deletion at a splice site in the LRRK2 gene (c.5314_5317 + 6delAAAGGTAAGG), found in family member III-5. The alteration causes a predicted frameshift and protein truncation. (C) Sanger sequencing did not identify this alteration in DNA isolated from normal FFPE tissue from family member III-3, who did not develop MM. (D) However, the LRRK2 mutation was identified in DNA isolated from MM tissue from family members III-2 and III-1, the latter examined by Sanger sequencing. Analysis of the macrodissected tumor specimen from case III-1 revealed that the mutant LRRK2 allele was present in a hemizygous or homozygous state, indicative of LOH.

Pedigrees from the remaining seven families without predicted pathogenic germline mutations are presented in Supplementary Material, Figures S1 and S2. In four of these families, germline missense variants were found in one to four cancer-related genes (Supplementary Material, Fig. S1), although none of these variants was predicted to be pathogenic. In the other three families, no candidate gene variants were identified in germline DNA from the probands (Supplementary Material, Fig. S2).

With the availability of TCGA data on pleural MM samples (23), we used the cBioPortal Cancer Genomics software to identify the germline and somatic mutations affecting the candidate genes we identified herein. No germline mutations were found in the TCGA dataset, whereas a small percentage of somatic mutations and other alterations affected our candidate genes were identified among the 82 MM patients for which genomic data were available (Supplementary Material, Fig. S3). The most frequently altered genes, BRIP1, LRRK2 and RHBDF2, occurred in 3–6% of the TCGA cases. Notably, there were two deep deletions and a truncating mutation in LRRK2. In addition, there were three or four amplifications of RHBDF2 and BRIP1, respectively. Altogether, two deep deletions, truncating mutations and/or gene fusions were observed in CHEK2, POLQ and DNMT3A.

To further assess the importance of LRRK2 in MM generally, we examined its protein expression in a set of human primary pleural tumors and pleural tumor-derived cell lines. Ten of 16 (62.5%) MM cell lines and 7 of 12 (58%) primary MM tumors had undetectable or substantially downregulated expression of LRRK2 compared with immortalized LP9 mesothelial cells and other MM tumors (Fig. 6).

Figure 6 .

Immunoblot depicting expression of LRRK2 in human pleural MM cell lines (A) and primary MM tumors (B). LRRK2 protein levels are absent or substantially downregulated in a majority of samples compared with immortalized LP9 human mesothelial cells. Expression of control proteins β-actin (ACTB) and GAPDH, respectively, are shown for comparison.

Discussion

Many of the genes with germline variants that we identified have been reported in other MM patients from high-risk cancer families, although with different alterations (Supplementary Material, Table S1) (19,24–28). Among the affected genes, germline mutations of APC, ATM, BRCA2, CHEK2 and DNMT3A have each been reported by more than one other group. The findings presented here, as well as in several other recent studies (19,29–34), demonstrate that not all MM families, high-risk cancer families with at least one MM or apparently sporadic MM cases with cancer-related germline mutations involve the BAP1 gene and that it is likely that there are other genes associated with MM susceptibility. We have access to the DNA and disease history of a large, unique cohort of high-risk cancer families with at least one member having MM. We utilized WGS technology to sequence the entire genome of 14 MM patients from 13 cancer families that do not harbor germline BAP1 mutations. We identified mutations in a number of candidate cancer-related genes that may have contributed to the high incidence of cancer, including MM, in these families. Notably, several of these candidate genes were found to be mutated in more than one family. In particular, different mutations were discovered in the POLQ, CHEK2 and LRRK2 genes in unrelated individuals.

Altogether, 10 of the 13 probands exhibited germline DNA sequencing variants in cancer-related genes, 6 of which had one or more predicted pathogenic mutations; 2 of these same 6 cases also had either a large deletion encompassing two exons of a gene or an entire gene. Three of these six cases had two different predicted pathogenic changes. Five of these 9 probable deleterious alterations affected genes that encode proteins involved in DNA repair (CHEK2, MUTYH, POLE4, POLQ and XRCC1), and the other 7 genes encode proteins involved in chromatin modification (ARID1B, DNMT3A, JARID2, SETD1B and SMARCB1) or other cellular pathways: LRRK2 (two cases) and MSH4. CHEK2 encodes cell cycle checkpoint kinase 2, a DNA repair signaling kinase downstream of ATM and ATR. The 5395 bp deletion of CHEK2 exons 9 and 10 seen in MM case ABS2640 (Fig. 4A) has been previously reported in some European families, and interestingly, this alteration was shown to be associated with a predisposition to breast or ovarian cancers (35,36). In addition, a germline missense mutation of CHEK2 was observed in MM case ABS3383, although this variant has been classified in ClinVar as having uncertain significance. However, other germline mutations in CHEK2 have been shown to increase risk of familial breast cancer up to 4-fold (37). MUTYH, which was involved in a germline splice site mutation in MM case ABS2813, encodes a DNA glycosylase that is involved in oxidative DNA damage repair. Individuals harboring germline homozygous, inactivating mutations in MUTYH, have an increased risk for developing MUTY-associated polyposis (MAP) in gastrointestinal organs and the uterus. Cancer risk in carriers of a heterozygous MUTYH mutation (which accounts for about 1% of the Caucasian population) is uncertain, but it may be as much as two to three times higher than in the general population (38,39). POLE4, which was heterozygously deleted in the germline of MM case ABS3425, encodes a subunit of the DNA polymerase epsilon polymerase, an enzyme involved in DNA replication and repair. In inbred C57BL/6 mice, Pole4 homozygous mutant animals are embryonic lethal, but FVB/sv129 outbred strains were viable with a lower than expected Mendelian ratio (40). The surviving outbred mice had developmental defects of the skeleton and a high incidence of lymphomas. Both a nonsense mutation in POLQ and an inactivating mutation in XRCC1 were observed in MM Case ABS3460. POLQ encodes the DNA polymerase theta (Polθ) protein, which is involved in the error prone, microhomology-mediated end-joining repair of double-strand breaks. Polθ is normally expressed at low levels in normal tissues, but it is highly expressed in tumors such as homologous recombination-deficient breast and ovarian cancers (41). How an inactivating mutation of POLQ might contribute to cancer susceptibility is presently unknown. The XRCC1 protein is a scaffold protein essential in base excision and single-strand break repair of DNA (42). XRCC1-deficient cells exhibit hypersensitivity to various mutagens, and certain XRCC1 polymorphisms have been implicated in reduced genomic stability and increased breast cancer risk (42).

Although sequencing variants were found in five genes (ARID1B, DNMT3A, JARID2, SETD1B and SMARCB1) that encode proteins that participate in chromatin modification processes, a predicted pathogenic mutation was found only for DNMT3A (case ABS3425). The protein product of the DNMT3A gene, DNA methyltransferase 3 alpha, is normally involved in CpG DNA methylation and epigenetic silencing of target genes. Interestingly, DNMT3A interacts with EZH2, a histone methyltransferase, which has been previously reported to be elevated as a result of BAP1 loss in MM (43). Additionally, inactivating germline DNMT3A mutations have been implicated in the rare DNMT3A overgrowth syndrome, which is characterized by tall stature and other physical anomalies in affected individuals. Analysis of RNA-seq and survival findings from TCGA’s MM database revealed that MM patients whose tumors had low DNMT3A mRNA expression levels had a significantly better overall survival rate. How germline mutations in DNMT3A might predispose to MM is unknown.

In case ABS2406, an MSH4 indel mutation results in a predicted frameshift mutation (p.Ile240fs). Although a member of the MMR family, MSH4 has not been found to play a role in MMR, but instead, is involved in homologous recombination during meiosis (44). However, one study determined that MSH4 does play a role in maintaining genomic stability through its ability to suppress non-homologous end-joining double-strand break repair (45). Interestingly, a missense mutation in MSH4 was found to co-segregate with multiple family members who collectively had three gliomas and two schwannomas (46). ABS2406 also possess a non-frameshift deletion within SMARCB1. The gene encodes a protein that is part of the SWI-SNF chromatin remodeling complex, and the deletion is predicted to affect residues located within the DNA-binding region (p.20_25del). Germline mutations of SMARCB1 are known to be associated with schwannomatosis and rhabdoid TPDS (47). The occurrence of a rhabdoid sarcoma and schwannoma in our index case mirrors another study where an individual with a germline deletion of SMARCB1 developed both of these types of cancers (48). Whether the development of MM in ABS2406 is related to the mutation in SMARCB1 or in combination with the MSH4 mutation remains unknown.

One highly noteworthy candidate gene discovered in our investigation is LRRK2, which incurred a deletion at a splice site in an asbestos-exposed Italian family with multiple members afflicted by pleural MM (Fig. 5A). LRRK2 encodes a kinase that is involved in oxidative stress, inflammation and autophagy. In a macrodissected MM tumor sample from this family, we were able to demonstrate homo- or hemizygosity of the LRRK2 splice site mutant allele and the loss of the remaining wild-type (LOH), suggesting that the mutant gene acts as a driver for MM in this unique asbestos-exposed family with an unusually high penetrance of MM.

Even among heavily exposed asbestos workers, the incidence of MM is only about 5% (49). As is the case for BAP1 (3,11,14,50,51), germline mutation of LRRK2 may make individuals highly susceptible to the carcinogenic effects of asbestos. Ten different pathogenic missense mutations in this gene have been described in Parkinson’s disease (52), and germline LRRK2 G2019S missense mutations have been found in ~10% of individuals with Parkinson's disease (53). Interestingly, epidemiological studies have indicated that LRRK2 G2019S carriers have an increased risk of developing cancer, including hormone-related neoplasms (prostate and breast carcinomas), colon and kidney carcinomas, as well as meningioma (53), the latter two also part of the BAP1-TPDS tumor spectrum (11–13). LRRK2’s role in predisposing to various cancers may be owing to its involvement in the DNA damage response. Treatment of mouse embryonic fibroblasts (MEFs) with the DNA-damaging agent, adriamycin, resulted in phosphorylation of the LRRK2 protein at several sites (54). In contrast, LRRK2 phosphorylation was not increased in Atm knockout MEFs, indicating that LRRK2 is downstream of Atm. In addition, induction of p53 and p21 expression caused by adriamycin treatment was suppressed when LRRK2 was silenced by siRNA (54). Another group demonstrated that LRRK2 can phosphorylate p53, leading to its translocation to the nucleus and induction of p21 gene expression (55). The LRRK2 mutation identified in our family is unique in that it is a 10 bp deletion encompassing the end of exon 36 as well as the adjacent splice site and intron. It is unknown what cDNA and protein product the mutation would produce, if any. Since we observed loss of the wild-type allele of LRRK2 in an MM tumor (Fig. 5D), we hypothesize that LRRK2 may act as a tumor suppressor gene in this context. Supporting this is a recent study that found a striking reduction in LRRK2 mRNA expression in ~40% of human lung adenocarcinomas, with reduced LRRK2 expression being significantly associated with worse survival as well as signatures of less-differentiated disease and immunosuppression. The investigators also determined that Lrrk2 knockout mice were highly susceptible to carcinogen-induced lung adenocarcinomas (56). Notably, our immunoblot analyses demonstrated frequent downregulation or loss of LRRK2 protein expression in the majority (61% overall) of pleural MM tumors and MM cell lines (Fig. 6). Collectively, our data suggest that in addition to being a candidate MM tumor susceptibility gene, loss of LRRK2 expression is a newly recognized common tumor suppressor alteration in MM.

The utilization of structural variation detection programs, such as Manta and Delly, has allowed us to identify large-scale deletions of the CHEK2 and POLE4 genes from the WGS data. These alterations would have been highly unlikely to be discovered if exome sequencing was performed instead of WGS or if only standard mutation detection programs were used to analyze the WGS data. The deletion that encompasses POLE4 has not been reported yet to the best of our knowledge, and the incidence and its association with disease susceptibility may be of significant interest.

Owing to the lack of blood and/or tumor samples from the relatives of most of the MM probands presented here, it has not been possible to confirm the involvement of most of the candidate genes in other tumors in the probands’ families. Thus, future work will include analysis of these genes in a larger cohort of MM patients from high-risk cancer families. In addition, functional studies of the predicted pathogenic mutations can be evaluated in vivo to assess their role in MM susceptibility.

Current evidence indicates that genetically susceptible individuals are at elevated risk of MM when exposed to asbestos. Owing to the presence of asbestos in older buildings and its persistent use in developing countries and in some settings in the USA, as well as the long latency of MM development, MM will continue to be a health burden for decades. Therefore, understanding how inherited mutations in genes can lead to the increased risk for developing MM is highly important. Identifying such high-risk individuals will allow for the implementation of proactive measures, such as biannual ultrasound or MRI imaging, annual physical examinations and serum marker monitoring and education-related self-awareness among high-risk individuals when presented with disease symptoms.

Materials and Methods

Patients and samples

Twelve US or Canadian cases of MM were from a series of 141 MM patients with a family history of cancer but with no germline mutation of BAP1 (18). All 12 cases had a known history of asbestos exposure and were identified via one or more of the following: (1) patient health care providers, (2) a patient support organization (Mesothelioma Applied Research Foundation) or (3) independent medical evaluations for medical–legal purposes. All study participants provided informed consent for their participation, and the protocol was approved by the Institutional Review Board at the Wake Forest School of Medicine. The original 141 MM cases were selected based on a personal or family history of one or more of the cancers that were previously reported in BAP1-TPDS families (18); however, for the present study of 12 selected cases, the main criterion was a high overall incidence of cancer generally, not necessarily a personal or family history strongly indicative of BAP1-TPDS. Study entry criteria consisted of a pathology report, including immunohistochemical staining confirming the diagnosis of MM, from a CLIA-certified US or Canadian laboratory.

Three Italian cases of MM (two analyzed by WGS, and one examined by Sanger sequencing) were from an asbestos-exposed family with multiple cases of pleural MM without inheritance of a predisposing BAP1 mutation (22). Between 1987 and 2012, six women and two men developed pleural MM in one generation (generation III). In addition to the eight confirmed MMs, two female family members in generation II had pleural cancers (highly suspected to be MM, but unconfirmed), without radiological evidence of a primary tumor in the lung or elsewhere (22). The kindred had known exposure to crocidolite asbestos in the domestic setting, as documented by transmission electron microscopy in several family members.

For all of the US and Canadian MM cases, DNA was isolated from blood using standard techniques. For the Italian MM family, DNA for WGS was isolated from peripheral blood in one MM case and from an OCT-embedded sample containing a mixture of MM tumor and normal tissue. A third DNA sample from the Italian family was obtained from a macrodissected tumor of another family member with MM.

Next-generation sequencing

WGS of genomic DNA isolated from blood was performed by Novogene Corp. using their Illumina HiSeq X Ten platform with paired-end 150 bp reads, with approximately 30–50× coverage. FASTQ files generated from the runs were processed by Novogene, with mapping to the human reference genome (b37) done using the Burrows-Wheeler Aligner. Read alignment Binary Alignment Map (BAM) files (comprehensive raw data of genome sequencing) were generated after SAMtools sorting and Picard marking of duplicates. GATK was used to call single nucleotide polymorphisms (SNPs) and small insertions/deletions (small indels) from the BAM files. ANNOVAR was used to annotate the variants, which was then scored using the CADD program. Structural variants (SVs) consisting of large deletions, insertions, duplications, inversions and translocations, were determined using Manta, Delly, SvABA and Lumpy software. Annotations of the Manta and Delly SVs were done using the AnnotSV program supplemented with gene enhancer data from GeneHancer (57).

Sequence analysis

Primers were designed encompassing the locations of mutations to amplify 100–300 bp products. PCR was performed using the Fast Cycling PCR kit (Qiagen) according to the manufacturer’s recommended protocol with the following conditions: 95°C for 5 min; 35 cycles of 96°C for 5 s, 60°C for 5 s and 68°C for 45 s; 72°C for 1 min. PCR products were gel-purified and Sanger-sequenced with the same primers used for PCR. cDNA and protein mutation nomenclature standardized by the Human Genome Variation Society (HGVS, http://www.hgvs.org/mutnomen) was used to describe the mutations observed.

Immunoblot analysis

Protein lysates from pleural MM cell lines were prepared using RIPA cell lysis buffer supplemented with 2 mM PMSF. Pleural MM tumor protein lysates were prepared by pulverizing frozen tumor pieces in liquid nitrogen, using a mortar and pestle and then by disrupting the cells in 1X cell lysis buffer from Cell Signaling (Danvers, MA) that was supplemented with 2 mM PMSF. All protein lysates were incubated for 30 min on ice followed by centrifugation for 20 min at 4°C. Bradford reagent was used to measure protein concentrations. Then, 30 μg cell lysates were loaded into Bis-Tris gels (Invitrogen) and were transferred onto PVDF membranes (Millipore). Membranes were blocked with 5% non-fat milk in TBS buffer with Tween 20 for 1 h, followed by incubation with primary antibodies at 4°C overnight. After washing, membranes were incubated with secondary antibody at room temperature and were further washed for three times. The antibodies used for immunoblotting were from Antibodies, Inc. (Davis, CA): anti-LRRK2/Dardarin (C-terminus), clone N241A/34, #75-253) and Santa Cruz Biotechnology (Dallas, TX): anti-β-actin (ACTB, sc-47 778) and anti-GAPDH (sc-32 233).

Funding

Funding was provided by the Mesothelioma Applied Research Foundation (to J.A.O.), National Cancer Institute (CA175691 and CA-06927 to J.R.T.) and Local No. 14 Mesothelioma Fund of the International Association of Heat and Frost Insulators & Allied Workers (to J.R.T.).