**Well-Di**ff**erentiated Papillary Mesothelioma of the Peritoneum Is Genetically Distinct from Malignant Mesothelioma**

**Raunak Shrestha 1,2,3 , Noushin Nabavi 1,4 , Stanislav Volik <sup>1</sup> , Shawn Anderson <sup>1</sup> , Anne Haegert <sup>1</sup> , Brian McConeghy <sup>1</sup> , Funda Sar <sup>1</sup> , Sonal Brahmbhatt <sup>1</sup> , Robert Bell <sup>1</sup> , Stephane Le Bihan <sup>1</sup> , Yuzhuo Wang 1,2,4 , Colin Collins 1,2,\* and Andrew Churg 5,\***


Received: 17 May 2020; Accepted: 11 June 2020; Published: 13 June 2020

**Abstract:** Well-differentiated papillary mesothelioma (WDPM) is an uncommon mesothelial proliferation that is most commonly encountered as an incidental finding in the peritoneal cavity. There is controversy in the literature about whether WDPM is a neoplasm or a reactive process and, if neoplastic, whether it is a variant or precursor of epithelial malignant mesothelioma or is a different entity. Using whole exome sequencing of five WDPMs of the peritoneum, we have identified distinct mutations in *EHD1*, *ATM*, *FBXO10*, *SH2D2A*, *CDH5*, *MAGED1*, and *TP73* shared by WDPM cases but not reported in malignant mesotheliomas. Furthermore, we show that WDPM is strongly enriched with C > A transversion substitution mutations, a pattern that is also not found in malignant mesotheliomas. The WDPMs lacked the alterations involving *BAP1*, *SETD2*, *NF2*, *CDKN2A*/*B*, *LASTS1*/*2*, *PBRM1*, and *SMARCC1* that are frequently found in malignant mesotheliomas. We conclude that WDPMs are neoplasms that are genetically distinct from malignant mesotheliomas and, based on observed mutations, do not appear to be precursors of malignant mesotheliomas.

**Keywords:** well-differentiated papillary mesothelioma; WDPM; malignant mesothelioma; DNA sequencing; mutation

#### **1. Introduction**

Well-differentiated papillary mesothelioma (WDPM) is a morphologically distinctive papillary proliferation of mesothelial cells that is most commonly encountered as an incidental finding in the peritoneal cavity, and less often in the pleural cavity, pericardium, and tunica vaginalis. These lesions may be single or multiple but by definition do not invade the underlying stroma and usually behave in a benign or indolent fashion, sometimes persisting for many years [1]. However, the nature of WDPM is disputed, with theories ranging from a reactive non-neoplastic process to a benign tumor, to a variant and/or precursor of epithelial malignant mesotheliomas [2]. To add further confusion, unequivocal invasive malignant mesotheliomas can have areas that mimic WDPM. Since malignant mesotheliomas are aggressive tumors, the distinction from WDPM is important, but WDPMs are

sometimes treated with debulking cytoreductive surgery followed by hyperthermic intraperitoneal chemotherapy (HIPEC) as if they were mesotheliomas [3].

Genome-wide sequencing analyses of malignant mesotheliomas have revealed frequently observed genomic aberrations such as loss of function mutation and/or copy number alterations/deletion of *BAP1*, *SETD2*, *CDKN2A*, and *NF2* [4–6]. Studies analyzing WDPM using DNA sequencing technology are limited. Case studies have reported WDPMs with somatic mutation of *E2F1* [7], heterozygous loss of *NF2* [8], and germline *BAP1* mutation [9], which if correct would suggest that they may be variants of malignant mesothelioma. Nevertheless, using immunohistochemistry (IHC) and fluorescence in situ hybridization (FISH), Lee et al. demonstrated that, unlike in malignant mesothelioma, both *BAP1* and *CDKN2A* are intact and respective proteins are expressed in WDPMs [10]. More recently, Stevers et al. [11] performed genomic profiling of 10 WDPMs and found that they harbored *TRAF7* or *CDC42* mutually exclusive missense mutations. –

To shed further light on this question we performed an extensive genomic characterization of a cohort of five WDPMs of the peritoneum.

#### **2. Results**

#### *2.1. Histopathological Features of WDPM*

We assembled a cohort of five incidentally identified WDPM cases in the peritoneum detected during surgery for another process and all were solitary lesions. All of these five cases had the typical features described for WDPM [12], i.e., a papillary architecture with a single layer of covering bland mesothelial cells and myxoid cores in the papillae (Figure 1).

**Figure 1.** Histopathology of five WDPM cases used for the study. Microphotographs of histological features of WDPM stained using haematoxylin and eosin (H&E). The panel under the dotted box represents the magnified section of the photomicrographs at ×20. The lesion sites/sizes were peritoneum, site not specified, for cases WDPM-01 (3 mm), WDPM-02 (6 mm), WDPM-03 (4 mm), WDPM-04 mesentery (4 mm), and WDPM-05 omentum (4 mm).

#### *2.2. Mutational Landscape of WDPM*

– – We performed high-coverage whole exome sequencing of five WDPMs from formalin-fixed and paraffin embedded (FFPE) samples. We achieved a mean sequencing reads coverage of 87×–117×, with at least 20–45% of targeted bases having a coverage of 100× (Table S1). Due to papillary architecture, the tumor cellularity of the WDPM tissues was estimated to be about 50% (Table S2). Although the high coverage sequencing provides us an opportunity to detect higher proportions of mutations, the normal tissue admixture lowers the mutation detection sensitivity. To overcome this challenge, we implemented strict mutation filtering criteria as described in the Methods section and retained only high confident mutation calls for downstream analysis.

Analysis of the mutational patterns in WDPM revealed a strong enrichment of C > A transversion substitution mutation (Figure 2A). Using the software deconstructSigs [13], we evaluated the characteristic mutation patterns in WDPM against the mutational signature obtained from the COSMIC mutational signature database [14]. Intriguingly, we identified consistent patterns of nucleotide substitution mutation associated with WDPM. Notably, we found that mutational signature 24 is significantly operative in all five WDPM cases (Figure 2B). In addition to this, mutational signature 21 and 28 were also observed in the WDPM cases.

– **Figure 2.** Landscape of mutations in WDPM. (**A**) Mutational signature present in WDPM. (**B**) Proportional contribution of different COSMIC mutational signature per sample. (**C**) Mutation status in WDPM. Top seven most recurrent mutations are represented in the figure. The bar plot on the top panel represents the total number of mutations detected in the respective WDPM. (**D**–**H**) Plots showing mutation distribution and the protein domains for the corresponding mutated protein.

— *—* We identified 461 unique non-silent mutations across five WDPM samples affecting 297 unique protein coding genes (Table S3). Patient WDPM-04 had the highest mutation burden and WDPM-01 had the least. Two genes—*FBXO10* and *SH2D2A*—were mutated in all five WDPM cases, again

–

displaying consistent mutational patterns (Figure 2C). Missense mutation *EHD1*D147A in the dynamin protein domain was found in four cases (Figure 2C,D). The variant allele frequency (VAF) of *EHD1* was in the range 29–43%, indicating its likely clonal origin (given that the tumor cellularity of the WDPM tissues were estimated to be about 50%) (Figure S1). Notably, we identified missense mutation in DNA-damage response gene *ATM* in four cases (Figure 2C,E). All four cases harbored *ATM*K2303R located in the FRAP-ATM-TRRAP (FAT) domain in the ATM protein. The VAF of *ATM* was also in the range 25–30%, indicating its likely clonal origin (Figure S1). The gene encoding cadherin 5 (*CDH5*) harbored *CDH5*D714E mutations in its C-terminus cadherin protein domain in four cases (Figure 2C,F). The VAF of *CDH5* was also in the range 26–38%, indicating its likely clonal origin (Figure S1). We also identified missense mutation *FBXO10*C42F in four cases and *FBXO10*C26F in one case (Figure 2C,G). Both mutation variants of *FBXO10* were present in the F-box like protein domain. The VAF of *FBXO10* was also in the range 24–37%, indicating its likely clonal origin (Figure S1). Similarly, we identified missense mutation *SH2D2A*G155V in four cases and *SH2D2A*G155D in one case (Figure 2C,H). These variants were located in the SH2 protein domain. The VAF of *SH2D2A* in WDPM-04 was 69%, indicating the mutation to be clonal. The VAF of *SH2D2A* in the rest of the four WDPMs was in the range 37–47% (Figure S1). Furthermore, we also identified mutations in *MAGED1* and *TP73* in each of the four WDPM cases (Figure 2C).

#### *2.3. Copy Number Landscape of WDPM*

The aggregate copy number aberration (CNA) profile of WDPM is shown in Figure S2. We observed 278 CNA events across all samples (Table S4). The CNA resulted in alterations of about 4–14% of the protein-coding genomes in the WDPM. Patient WDPM-02 had a high copy number burden and WDPM-03 had the least copy number burden (Figure S2). Overall, copy number profiles of the WDPM did not show many alterations (Figure S3). Notably, we found copy number gain of *SETDB2* and *LAST2* and copy number loss of *SMARCA4* and *TRAF7* in WDPM-02. We also found copy number loss of cancer genes such as *CCNE1*, *MAF*, *MAFB*, *MYC*, *ZNF479*, and *MGMT* and copy number gain of *FOXA2*, *CDH10*, and *GPC5* in at least two WDPM cases.

#### *2.4. Signaling Pathways Dysregulated in WDPM*

To identify signaling pathways dysregulated by mutated genes in WDPM, we performed pathway enrichment analysis using the KEGG [15] pathway database (see Methods section). Our analysis revealed that WDPM mutations target different signaling pathways often dysregulated in cancer (Figure 3 and Table S5) such as pathways in cancer, focal adhesion, Vascular endothelial growth factor (VEGF) signaling, Janus kinases - signal transducer and activator of transcription (JAK-STAT) signaling, Wnt signaling, P53 signaling, apoptosis, etc. We found *CDH5* mutations target cell adhesion and the leukocyte migration pathway, *EHD1* mutations target endocytosis, *SH2D2A* mutations target the VEGF signaling pathway, *ATM* mutations target apoptosis and P53 signaling pathways, and *TP73* targets the neurotrophin signaling and P53 signaling pathways. This indicates that the mutations identified in WDPM cases might be relevant to pathogenesis of WDPM.

**Figure 3.** Signaling pathways dysregulated in WDPM. We performed pathway enrichment analysis using genes mutated in WDPM cases against the signaling pathways in the KEGG pathway database. The figure shows the top 20 pathways enriched with mutated genes in WDPM. Each circle represents a pathway, its size indicates the number of mutated genes targeting the pathway, and its color indicates the pathway enrichment score. The thickness of edges connecting two circles (pathways) is proportional to the number of mutated genes common between the two pathways. PI signaling: phosphatidylinositol signaling pathway.

#### *2.5. WDPM is Genetically Distinct from Malignant Mesothelioma*

Next, we compared the genomic profiles of WDPM with those of malignant peritoneal mesothelioma. For this, we leveraged the DNA sequencing data from two recently published peritoneal mesothelioma patient cohorts [6,16]. We first assessed the pattern of mutations in WDPM and peritoneal mesothelioma cases. Intriguingly, we observed that WDPM has a strong enrichment of C > A transversion substitution mutation (Figure 2A,B), whereas, peritoneal mesothelioma has strong enrichment of C > T transition substitution mutation (Figure S4). This mutational pattern in WDPM is different from those reported in pleural [4,5] or peritoneal [6] mesotheliomas.

Notably, we found WDPM specific mutations in *EHD1*, *FBXO10*, *CHD5*, *MAGED1*, *ATM*, and *TP73* genes that were absent in peritoneal mesothelioma (Figure 4A). Although mutations in *EHD1* and *ATM* genes were each observed in peritoneal mesothelioma, we did not find the WDPM-specific *EHD1* D147A , *EHD1* A465D, and *ATM*K2303R mutations in these cases. Interestingly, in WDPM, we did not find any of the mutations in *BAP1*, *SETD2*, *TP53*, *NF2*, *CDKN2A*, and *LAST1*/*2* frequently observed in malignant mesotheliomas (Figure 4A). We also did not find mutations in *TRAF7* or *CDC42* in WDPM, however, *TRAF7* mutations were observed in several peritoneal mesothelioma cases. Furthermore, we evaluated the differences in the copy number status of genes between WDPM and peritoneal mesothelioma. We did not find any copy number loss in gene characteristics of malignant mesotheliomas such as *BAP1*, *SETD2*, *PBRM1*, *SMARCC1*, *CDKN2A*/*B*, *LATS1*/*2*, and *NF2* (Figure 4B). *TRAF* copy number loss was observed in one WDPM case, whereas, several peritoneal mesothelioma cases harbored *TRAF7* copy number alteration.

**Figure 4.** Genomic alterations in WDPM and peritoneal mesothelioma. We compared the genomic alteration profile of the WDPM cases to the peritoneal mesothelioma patient cohorts from two recently published studies, Vancouver Prostate Centre (VPC) cohort [6] and American Association for Cancer Research (AACR) project Genomics Evidence Neoplasia Information Exchange (GENIE) cohort [16]. (**A**) Oncoplot showing differences in mutation pattern between WDPM and peritoneal mesothelioma. Each column in the figure represents an individual cancer sample. (**B**) Oncoplot showing the copy number aberration status of WDPM and peritoneal mesothelioma. Each column in the figure represents an individual cancer sample.

#### **3. Discussion**

In this study, we investigated the genomic alterations found in a cohort of five WDPMs. The tumors analyzed here are clinically typical of the setting in which WDPM is most commonly found, i.e., as an incidental lesion discovered during surgery for another process, and all lesions were morphologically characteristic WDPM.

Overall, our results suggest that WDPM are distinctive lesions with their own set of genomic alterations. Given the number of mutations and the nature of the mutations found, including at least one tumor suppressor gene, *TP73*, and several genes that may be associated with other types of malignancy (*ATM*, *CDH5*, *MAGED1*) [17–19], WDPM clearly appears to be a functionally benign neoplasm and not a reactive process. Further, it is clear that WDPM are genetically quite different from both peritoneal and pleural mesotheliomas. Indeed, our most important finding is the lack of alterations involving *BAP1*, *SETD2*, *NF2*, *CDKN2A*, *PBRM1*, and *SMARCC1* genes consistently mutated or deleted in malignant mesotheliomas.

We found consistent mutation patterns in five WDPMs with strong enrichment of C > A transversion substitution mutation and COSMIC mutational signature 24. The WDPMs harbored distinct mutations in *EHD1*, *FBXO10*, *CHD5*, *MAGED1*, *ATM*, and *TP73* genes either in all five or at least four out of five WDPM cases. The COSMIC mutational signature 24 has been shown to be commonly found in certain liver cancers with exposure to carcinogen such as aflatoxin [20]. However, these WDPMs were incidental findings during surgery and any prior exposure to carcinogens (either aflatoxin or asbestos) is extremely unlikely. Mutations and copy number changes in *CDH5* have been previously reported in mesotheliomas [21,22] but are uncommon events and were not present in any of our reference mesothelioma datasets (Figure 3). *CHD5* is known to promote intravasation and stimulates TGF-β driven epithelial–mesenchymal transition (EMT) [23]. *EHD1* regulates the endocytic recycling process. *EHD1* is known to play a key role in transportation of receptors from endosomes into the endocytic recycling compartment (ERC) and from the ERC to the plasma membrane [24]. Moreover, *EHD1* has been associated with cell proliferation, apoptosis, metastasis, and drug resistance in breast and non-small cell lung cancer (NSCLC) [25] but has not been reported to be abnormal in malignant mesotheliomas. *FBXO10* binds to the anti-apoptotic oncoprotein BCL-2 and promotes its degradation, thereby initiating cell death in lymphomas [26]. *SH2D2A* is known to be involved in T-cell activation [27]. Mutations in *FBXO10*, *SH2D2A*, and *TP73* has not been reported in any malignant mesotheliomas.

Our study confirms a lack of copy number alterations in *BAP1*, *SETD2*, *PBRM1*, *SMARCC1*, *CDKN2A*/*B*, *LATS1*/*2*, and *NF2*. Copy number loss of *BAP1*, *SETD2*, *PBRM1*, and *SMARCC1* is often observed in peritoneal mesothelioma [6,28]. Copy number loss of *BAP1*, *CDKN2A*/*B*, *LAST1*/*2*, and *NF2* is frequently found in pleural mesothelioma [4,5].

What is surprising in our results is the absence of the TRAF7 and CDC42 alterations reported by Yu et al. [7] and Stevers et al. [11] in WDPM and by the same group in adenomatoid tumors [29]. Alterations in TRAF7 have also been reported in malignant mesotheliomas [4,16,30]. However, this does not appear to be a case of tumor misclassification, since the lesion illustrated by Stevers et al. [11] is a very typical WDPM and is identical to the tumors analyzed here. The lesions analyzed by Stevers et al. [11] were also all incidental findings and 8/10 were solitary, as were ours, and the lesions for which they had follow up did not behave in a malignant fashion.

The exact reasons for the discrepancy between our study and those of Stevers et al. [11] are unclear. It is possible that the underlying populations are genetically different, particularly given the very large and diverse immigrant population in Vancouver, Canada. The analytical approach used in these two studies was also somewhat different. Stevers et al. [11] used a targeted panel consisting of 479 cancer-related genes (UCSF500 Cancer Panel) for sequencing (Illumina HiSeq 2500), whereas we used Ion AmpliSeq™ (Thermo Fisher Scientific, Waltham, MA, USA) exome sequencing which covers 18,961 genes (Ion Proton™). The overlap in the genes examined between these two studies is given in Figure S5A. Using a targeted panel provided Stevers et al. [11] an advantage to sequence a small number of genes at a high depth (average depth = 320×, range = 33×–722×), whereas we sequenced a large number of genes at a cost of sequencing depth (average depth = 102×). Stevers et al. [11] reported 21 mutations covering 10 genes in 10 WDPM cases, whereas we identified 461 mutations covering 297 genes in 5 WDPM cases. There is no overlap of the mutated genes reported in Stevers et al. [11] and our study (Figure S5B). In fact, the UCSF500 gene panel used by Stevers et al. [11] covered only 10 mutated genes reported by our study (Figure S5C). We note that, despite high sequencing depth, no mutations in *ATM* (which was examined in the UCSF500 panel) were reported by Stevers et al. [11], whereas we identified consistent *ATM*K2303R mutations in 4 out of 5 WDPM cases (Figure S5C). We did

identify a few low confidence *TRAF7* mutations, but these did not pass our mutation filtering criteria (see Table S6 and Appendix A for detailed information). These differences likely indicate genomic heterogeneity in WDPM and warrants further investigation in larger patient cohort settings. Once there are sufficient cases described with consistent results, it may be possible to use a genomic approach to decide whether an equivocal case is a WDPM or a malignant mesothelioma and to base treatment on such data.

#### **4. Materials and Methods**

#### *4.1. Patient Cohort Description and Tissue Procurement*

A cohort of incidentally identified WDPM tissues (n = 5) were assembled from the surgical pathology archives at the Vancouver General Hospital. This study was approved by the Institutional Review Board of the University of British Columbia and the Vancouver Coastal Health (REB No. H15-00902 and V15-00902).

#### *4.2. Whole Exome Sequencing*

DNA from marked FFPE tissue sections (5–10 µm in thickness, ~50% WDPM cellularity) were isolated using a truXTRAC FFPE DNA Kit with Covaris Adaptive Focused Acoustics® (AFA®) technology, which enables the removal of the paraffin from the FFPE tissue in SDS buffer while simultaneously rehydrating the tissue. The samples were treated with proteinase K 0.2 mg/mL (Roche) followed by overnight incubation at 55 ◦C. After post-incubation in proteinase K, the samples were treated with RNAse and DNA extracted as per the truXTRAC FFPE DNA extraction protocol (cat#: 520136, Covaris, Inc., Woburn, MA, USA). The amount of DNA was quantified using Qubit® dsDNA HS Assay (Thermo Fisher Scientific).

For Ion AmpliSeq™ (Thermo Fisher Scientific) exome sequencing, 100 ng of DNA was used as input for Ion AmpliSeq™ Exome RDY library preparation, a PCR-based sequencing approach using 294,000 primer pairs (amplicon size range 225–275 bp), which covers >97% of consensus coding sequence (CCDS) (Release 12), >19,000 coding genes, and >198,000 coding exons. Libraries were prepared, quantified using qPCR, and sequenced according to the manufacturer's instructions (Thermo Fisher Scientific). Samples were sequenced on the Ion Proton System using the Ion PI™ Hi-Q™ Sequencing 200 Kit and Ion PI™ v3 chip. Two libraries were run per chip for a projected minimum coverage of 40 million reads per sample.

#### *4.3. Single Nucleotide Variant Calling*

We used Torrent Server (Thermo Fisher Scientific) for mapping aligned reads to the human reference genome hg19 (Torrent Mapping Alignment feature). Variants were identified using a Torrent Variant Caller plugin with the optimized parameters for AmpliSeq exome sequencing (Thermo Fisher Scientific). The variant call format (VCF) files from all samples were annotated using ANNOVAR [31].

To account for the low tumor cellularity in the WDPM samples and the absence of the matched control samples, we used strict mutation calls filtering criteria. Mutations were retained if (a) allele frequency (AF) < 75%, (b) read quality pass > 50%, (c) average heterozygosity < 0.1, (d) mutation calls not present in dbSNP database. We filtered out all In-Dels from our variant calls. Non-silent exonic variants including non-synonymous single nucleotide variations (SNVs), stop-codon gain SNVs, stop-codon loss SNVs, splice site SNVs, and frameshift In-Dels in coding regions were retained if they were supported by more than 50 reads. Furthermore, putative variants were manually scrutinized on the Binary Alignment Map (BAM) files through Integrative Genomics Viewer (IGV) version 2.3.25 [32]. Furthermore, due to lack of matched germline control samples from the WDPM cases, we used genomic DNA samples from blood of a cohort of peritoneal mesothelioma patients as germline control samples. We filtered out any variants that were also present in these control samples [6]. In this way, we excluded any potential germline variants as well as false positive calls and obtained highly confident variants of

WDPM. Based on the variant allele frequency (VAF), the mutations identified in WDPM were clustered into different groups using the R-package Maftools [33].

#### *4.4. Copy Number Aberration (CNA) Calls*

Copy number changes were assessed using Nexus Copy Number Discovery Edition Version 8.0 (BioDiscovery, Inc., El Segundo, CA, USA). Nexus NGS functionality with the FASST2 Segmentation algorithm was used to make copy number calls (a circular binary segmentation/hidden Markov model approach). The significance threshold for segmentation was 5 × 10−<sup>6</sup> with a minimum of 3 probes per segment and a maximum probe spacing of 1000 between adjacent probes before breaking a segment. The log ratio thresholds for single copy gain and single copy loss were set at +0.2 and −0.2, respectively. The log ratio thresholds for homozygous gain/loss were set at +0.6 and −1.0, respectively. The tumor BAM files were processed and compared with BAM files from a normal tissue pool as reference control. Reference reads per CN point (window size) was set to 8000. We used the Genomic Identification of Significant Targets in Cancer (GISTIC) [34] algorithm in Nexus to identify significantly amplified or deleted regions across the genome. The amplitude of each aberration is assigned a G-score as well as a frequency of occurrence for multiple samples. The false discovery rate (FDR) q-value for the aberrant regions was set to a threshold of 0.15.

#### *4.5. Mutational Signature Analysis*

We used deconstructSigs [13] software, a multiple regression approach to statistically quantify the contribution of mutational signatures for each tumor. The 30 mutational signatures were obtained from the COSMIC mutational signature database [14]. Only non-silent mutations were used to obtain the mutational signatures. In brief, deconstructSigs attempts to recreate the mutational pattern using the trinucleotide mutation context from the input sample that closely resembles each of the 30 mutational signatures from the COSMIC mutational signature database. In this process, each mutational signature is assigned a weight normalized between 0 to 1 indicating its contribution. Only those mutational signatures with a weight more than 0.06 were considered for analysis.

#### *4.6. Pathway Enrichment Analysis*

The mutated genes were tested for enrichment against signaling pathways present in the KEGG [15] pathway database obtained from the Molecular Signature Database (MSigDB) v6.0 [35]. A hypergeometric test-based gene set enrichment analysis was used for this purpose (https://github. com/raunakms/GSEAFisher). A cut-off threshold of Benjamini–Hochberg (BH) corrected *p*-value < 0.01 was used to obtain the significantly enriched pathways. Only pathways that are enriched with at least three mutated genes were considered for further analysis.

#### *4.7. Peritoneal Mesothelioma Datasets*

We utilized DNA sequencing datasets of two publicly available patient cohorts of peritoneal mesothelioma—VPC cohort [6] and AACR Project GENIE Cohort [16]. We used mutation and copy number profiles from both datasets for comparison with the genomic profiles of WDPM cases. AACR GENIE project data, Version 5.0, were downloaded from https://www.synapse.org/#!Synapse: syn7222066.

#### **5. Conclusions**

We have shown that WDPM are genetically distinct from malignant mesotheliomas and in our hands have a characteristic pattern of C > A transversion substitution mutations; *EHD1*, *FBXO10*, *CHD5*, *MAGED1*, *ATM*, and *TP73* missense mutations; as well as enrichment of COSMIC mutation signature 24. Taken in conjunction with the data from Stevers et al., these findings further reinforce the idea that WDPM should not be treated in the same fashion as malignant mesotheliomas.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2072-6694/12/6/1568/s1, Figure S1. Distribution of variant allele frequency (VAF) in WDPM, Figure S2. Landscape of copy number alterations in WDPM, Figure S3. Copy number segments (log ratio) of WDPM samples, Figure S4. Mutational signature present in malignant peritoneal mesothelioma obtained from Shrestha et al., Genome Medicine 2019 [6], Figure S5. Comparison of Stevers et al. 2018 [11] with present study, Table S1. QC metric of whole exome sequencing, Table S2. Quality control statistics of WDPM samples, Table S3. Mutation profile of WDPM, Table S4. Copy number profile of WDPM, Table S5. Signaling Pathways dysregulated in WDPM, Table S6. Mutations in gene CDC42 and TRAF7 reported Stevers et al. 2018 [11] and the corresponding sequencing reads detected in WDPM cases in the present study.

**Author Contributions:** Conceptualization, R.S., N.N., C.C., and A.C.; methodology, R.S., N.N., and S.L.B.; software, R.S., S.V., S.A., and R.B.; validation, N.N., A.H., B.M., F.S., and S.B.; formal analysis, R.S. and N.N.; investigation, R.S., N.N., and A.C.; writing—original draft preparation, R.S., N.N., and A.C.; writing—review and editing, R.S., N.N., C.C., and A.C.; visualization, R.S.; supervision, S.L.B., Y.W., C.C., and A.C.; project administration, N.N., and S.L.B; funding acquisition, Y.W., C.C., and A.C. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by BC Cancer Foundation, Mitacs, WorkSafe BC, Canadian Institutes of Health Research (CIHR), and Terry Fox Research Institute, and R.S. and N.N. were funded by Mitacs Accelerate Awards.

**Conflicts of Interest:** The authors declare no conflicts of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **Appendix A**

Here we provide a detailed comparison between Stevers et al. [11] and the present study, which we label WDPM-VPC for convenience.

Stevers et al. [11] used a targeted panel of 479 cancer-related genes (UCSF500 Cancer Panel) for sequencing (Illumina HiSeq 2500 machine), whereas we used Ion AmpliSeq™ (Thermo Fisher Scientific, Waltham, MA, USA) exome sequencing, which covers 18,961 genes (Ion Proton™). The overlap in the genes examined between these two studies is given in Supplementary Figure S5A.

Using a targeted panel provided Stevers et al. [11] an advantage to sequence a small number of genes at a high depth (average depth = 320×, range = 33×–722×), whereas we sequenced a large number of genes at a cost of sequencing depth (average depth = 102×).

Stevers et al. [11] reported 21 mutations covering 10 genes in 10 WDPM cases, whereas we identified 461 mutations covering 297 genes in 5 WDPM cases. There is no overlap of the mutated genes reported in Stevers et al. [11] and our study (Supplementary Figure S5B). In fact, the UCSF500 gene panel used by Stevers et al. [11] covered only 10 mutated genes reported by our study (Supplementary Figure S5C).

Next, we analyzed the sequencing reads covering *CDC42* and *TRAF7* genes in the 5 WDPM cases in this study. Stevers et al. [11] reported two unique mutations in *CDC42* and six unique mutations in *TRAF7*. We focused on the corresponding gene regions in the WDPM cases in our study, as summarized in Supplementary Table S6.

We identified three unique very low confidence mutations in the *TRAF7* gene, one in WDPM-03 and two in WDPM-01 (Supplementary Table S6). In WDPM-03, only 16 reads (out of 111 reads) supported the *TRAF7*Y621D mutant allele. In WDPM-01, *TRAF7*N520SD was supported by 1 read (out of 102 reads), and *TRAF7*G536S was supported by 5 reads (out of 109 reads). Given that the tumor cellularity of the WDPM tissues was estimated to be about 50%, the *TRAF* mutations mentioned above were deemed very low confidence and hence did not pass our mutation filtering criteria. The rest of the *CDC42* and *TRAF7* mutated regions reported by Stevers et al. [11] were identified as wild type in the WDPM cases in this study.

Thus, within the experimental settings of our study, we do not find any high confidence mutations in *TRAF7* or *CDC42*.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Review* **Assessment of the Carcinogenicity of Carbon Nanotubes in the Respiratory System**

**Marcella Barbarino 1,2,\* and Antonio Giordano 1,2**


**Simple Summary:** Malignant mesothelioma is an aggressive cancer of the membranes covering the lung and chest cavity (pleura) or the abdomen (peritoneum), mainly linked to asbestos exposure. Asbestos is a proven human carcinogen but its use is far from being universally banned and the forecasts on the incidence of mesothelioma over the next several years are far from optimistic. Carbon nanotubes are a promising type of nano-materials used in the field of nanotechnology for a wide range of applications. However, the similarities between asbestos and CNTs have raised many concerns about their danger and are still the subject of intense research. Keeping in mind that the asbestos tragedy could have been prevented, the aim of this study is to review the recent scientific evidence on CNTs carcinogenicity.

**Abstract:** In 2014, the International Agency for Research on Cancer (IARC) classified the first type of carbon nanotubes (CNTs) as possibly carcinogenic to humans, while in the case of other CNTs, it was not possible to ascertain their toxicity due to lack of evidence. Moreover, the physicochemical heterogeneity of this group of substances hamper any generalization on their toxicity. Here, we review the recent relevant toxicity studies produced after the IARC meeting in 2014 on an homogeneous group of CNTs, highlighting the molecular alterations that are relevant for the onset of mesothelioma. Methods: The literature was searched on PubMed and Web of Science for the period 2015–2020, using different combinations keywords. Only data on normal cells of the respiratory system after exposure to fully characterized CNTs for their physico-chemical characteristics were included. Recent studies indicate that CNTs induce a sustained inflammatory response, oxidative stress, fibrosis and histological alterations. The development of mesothelial hyperplasia, mesothelioma, and lungs tumors have been also described in vivo. The data support a strong inflammatory potential of CNTs, similar to that of asbestos, and provide evidence that CNTs exposure led to molecular alterations known to have a key role in mesothelioma onset. These evidences call for an urgent improvement of studies on exposed human populations and adequate systems for monitoring the health of workers exposed to this putative carcinogen.

**Keywords:** malignant mesothelioma; carcinogenesis; asbestos exposure; carbon nanotubes

#### **1. Introduction**

Malignant pleural mesothelioma (MPM) is an aggressive cancer of the pleural membranes covering the lungs and is strongly linked to asbestos exposure. MPM generally manifests in an advanced stage after a latency period of 30–40 years following asbestos exposure.

Asbestos is a commercial term describing a group of specific silicate minerals forming bundles of long and thin mineral fibers that, because of their intrinsic characteristic of durability and resistance to chemicals, heat and electricity, were widely used in the late 1800s with the start of the Industrial Revolution. However, as early as 1898, lung damage was described in industry workers exposed to asbestos dust [1] and in the early 1900s,

**Citation:** Barbarino, M.; Giordano, A. Assessment of the Carcinogenicity of Carbon Nanotubes in the Respiratory System. *Cancers* **2021**, *13*, 1318. https://doi.org/10.3390/ cancers13061318

Academic Editors: Daniel L. Pouliquen and Joanna Kopecka

Received: 8 February 2021 Accepted: 11 March 2021 Published: 15 March 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

the first reports documenting fibrosis [2,3] and asbestosis [4] in asbestos-exposed workers were published. Only 30 years later (1935), the first association between asbestos and lung cancer was described [5,6] and it was another 10 years passed before asbestos exposure was correlated with pleural tumors, in the work of Wedler in 1943 [7] and the doctorate thesis of Wyres in 1946 [8]. In 1977 [9] and 1987 [10], the International Agency for Research on Cancer (IARC) concluded that asbestos is a human carcinogen and that the size and shape of the fibers influence the incidence of tumors. In 2006 [11] and 2009 [12] asbestos exposure was also correlated with an increased risk of other cancers, such as laryngeal and ovarian cancer.

Currently, even though asbestos is a known carcinogen, it is not banned in about 70% of the world (Figure 1). As such, more than 100 years after the recognition of asbestos as a carcinogenic agent, the case is not yet closed (http://ibasecretariat.org/chron\_ban\_list.php, accessed on 2 of October 2020). Indeed, it is important to note that countries that banned asbestos a quarter of a century ago are still contributing to the worldwide toll of more than 100,000 asbestos-related deaths per year [13]. As highlighted by Terracini [13], while banning asbestos is important, that alone does not create an asbestos-free environment. It will take a very long time to ban the use of asbestos worldwide, and it will take an even longer time to end up with an environment that is completely safe from the toxic effects of asbestos. For all of these reasons, the forecasts on the incidence of mesothelioma over the next several years are far from optimistic.

**Figure 1.** Timelines of the significant events leading to asbestos banning (**A**) and the available evidences of carbon nanotubes-induced toxicity (**B**).

It should also be considered that asbestos present in old constructions still represents a daily hazard to human health. There are numerous cases in which the presence of asbestos has been detected during the renovation or demolition of old buildings. The 9/11 terrorist attack in New York City to the World Trade Center, built in the 1970s, created extra exposure of asbestos, the impact of which will be known only in the coming years [14]. In the dense clouds of dust resulting from this tragic event, relevant quantities of carbon nanotubes (CNTs) produced by the high combustion temperatures were also found, along with other pollutants.

CNTs are nanomaterials composed of graphene sheets consisting of a series of carbon rings rolled into cylindrical fibers with an external measurement between 1 and 100 nm. Their fibrous particulate matter, similar to that of asbestos [15] has raised much concern about their safety for human health. In particular, growing evidence supports the idea that

inhaled nanomaterials of >5 µm and with a high aspect ratio (3:1), like rod-like carbon nanotubes resembling asbestos, may cause pleural disease including mesothelioma. In 2014, the International Agency for Research on Cancer (IARC) classified the first type of CNT, the long, rigid, needle-shaped Mitsui-7, as possibly carcinogenic to humans (Group 2B) [16], while in the case of other CNTs, it was not possible to ascertain their toxicity due to lack of evidence. It is also important to consider that, together with the lack of sufficient evidence supporting CNTs' carcinogenicity, their heterogeneity in chemical and physical structures makes it difficult to generalize the available results regarding their possible hazardous effects on human health.

The present review aims to provide an overview of the recent relevant toxicity studies produced after the IARC meeting in 2014 restricting analysis on a homogeneous group of CNTs: standard materials from the Joint Repository Center (JCR) and well-characterized commercial or in-house-made CNTs produced by catalytic carbon vapor deposition (CVD). Moreover, we review the data on mesothelial and lung cells since the respiratory system is considered the main route of exposure to asbestos and CNTs due to exposure during manufacturing process or to accidental exposure. Therefore, we exclude from our analysis CNTs produced for medical purposes, which are functionalized or modified and, consequently, results obtained from cancer or other models resembling a pathological status.

This review is structured following the IARC's parameters [17] for defining an agent as a human carcinogen: induces oxidative stress; induces chronic inflammation; induces epigenetic alterations; is genotoxic; alters DNA repair or causes genomic instability; causes immortalization; alters cell proliferation, cell death, or nutrient supply; acts as an electrophile either directly or after metabolic activation; is immunosuppressive; and modulates receptor-mediated effects.

#### **2. Methods**

The literature was searched on PubMed and Web of Science for the period 2015–2020, using different combinations of the following keywords: CNT, carbon nanotubes, SWCNT, MWCNT, single-walled carbon nanotubes, multi-walled carbon nanotubes, genotoxicity, DNA damage, epigenetic, oxidative stress, inflammation, immunosuppression, immortalization, and cytotoxicity. The language was restricted to English. Only data on normal cells of the respiratory system (pleural cells, lung cells, fibroblasts, and lung macrophages) after exposure to reference material (NM-400, NM-401, NM-402, and NM-403), SWCNTs, and MWCNTs synthetized by the CVD method and fully characterized for their physicochemical characteristics (length, diameter, agglomeration, and surface area) were included in the review.

#### **3. An Overview of Carbon Nanotubes**

Thirty years ago, the IBM researcher Don Eigler moved the first individual atom using a scanning tunnelling microscope. Despite that progress, Eigler has said he is not sure about when or even if his ideas for computing will bear fruit. It was Eigler who started the era of nanotechnology, the science that is able to create and manipulate materials at the nanoscale. Nano-sized materials, defined as having at least one dimension between 1 and 100 nm, include many types of materials, different in their physicochemical properties, and used in a great variety of applications [18]. Given the immense potential of nanotechnology, the global nanotechnology market has been estimated to reach 126.8 billion U.S. dollars by 2027 [19].

The big world of nanotechnology comprises various types of nanomaterial, all differing in their chemo-physical properties. CNTs are the most promising type of nanomaterials in the industry today. They are defined as nanotubes composed of carbon, consisting of one or more cylindrical graphene layers and are classified, on the basis of the number of graphene layers, as single- or multi-walled carbon nanotubes (respectively, SWCNTs and MWCNTs). Larger MWCNTs can contain hundreds of concentric layers.

As CNTs come to be used in a wider range of products, human exposure can take place through various routes, such as local (in medical applications, such as drug delivery, cancer therapy, medical diagnostics and imaging), environmental (industrial waste or accidentally released by the final product), or pulmonary (during occupational handling or accidental exposure). The work environment is actually thought to be the principal source of human exposure to CNTs during the phases of their production, as seen for example in laboratory handling and packaging of the final product, and in this case the most plausible route of exposure to manufactured nanomaterials remains pulmonary inhalation. The inhalation of particles during their synthesis is a significant concern in the growing nanotechnology field.

Despite different governmental organizations monitoring CNT exposure in workers, there are still no standards for defining the risk levels for CNT exposure. The method of monitoring CNTs in work environments involves measurement of Elemental Carbon (EC). The National Institute for Occupational Safety and Health (NIOSH, USA), based on quantification limits and not on studies in exposed workers, recommends an exposure level of 1 µg/m<sup>3</sup> elemental carbon (EC) [20]. This limit, which might not be representative of a safe exposure limit, has often been found to be much lower than those measured in various industries, ranging from 2.6 µg/m<sup>3</sup> to 45 µg/m<sup>3</sup> depending on the particular workplace analyzed (handling facilities, production areas, construction sites, offices, etc.) [21,22].

The pulmonary toxicity of fibrous materials such as asbestos has been demonstrated to result from deposition (thin fibers deposit in the lungs more efficiently than thick fibers) and tissue persistence ("biopersistence" is directly related to fiber length and inversely related to dissolution and fragmentation rates). CNTs have been demonstrated to deposit in human lungs and other organs. Lung biopsies of people exposed to the dense clouds of dust during the tragic events of 9/11 in New York City have shown the presence of CNTs produced by high combustion temperatures. The first adverse health effects diagnosed were pulmonary fibrosis, and bronchiolocentric parenchymal and granulomatous diseases [14].

Carbon nanotubes, although a sub-group in the immense word of nanomaterials, comprise various substances that differ from each other in length, size, diameter, impurities, and method used for synthesis and dispersion of the final product, among other characteristics. All of these characteristics impact their biological effects, and it is now recognized that generalized conclusions about CNTs should not be drawn by extrapolating data that are available on similar, but not identical, compounds. For these reasons, we focused our analysis on the results obtained using reference CNTs (NM-400, NM-401, NM-402, and NM-403) (https: //publications.jrc.ec.europa.eu/repository/bitstream/JRC91205/mwcnt-online.pdf accessed on 15 October 2020), with fully characterized commercial and in-house CNTs produced using the CVD method, which is currently one of the principal techniques used for CNT synthesis. Data regarding CNTs that had been chemically modified to alter their properties and data obtained in cancer cells were excluded from our analysis; this model is suitable for other purposes, such as drug-delivery studies, which are not the focus of this review.

We reported data relevant to assessing the potential adverse respiratory effects following the IARC's protocol for defining an agent as a human carcinogen [17]. For each group of characteristics, we analyzed data obtained from in vitro models of pleura, lung macrophages, and airway cells, from in vivo studies examining effects on the respiratory system, and from biological fluids collected from exposed workers, highlighting those results that could be relevant for mesothelioma onset.

#### **4. Carbon Nanotubes and the Hallmarks of Cancer**

*4.1. Oxidative Stress, Chronic Inflammation*

The oxidative potential of a particle is the intrinsic property to form reactive oxygen species (ROS). Generation of ROS and free radicals has been demonstrated to be involved in the molecular mechanisms leading to mesothelioma as well as other asbestos-related diseases. In cell-free systems, asbestos can generate free radicals and induce release of

inflammatory mediators such as cytokines, growth factors, reactive oxygen and nitrogen species in neutrophils, and alveolar macrophages for incomplete/frustrated phagocytosis of fibers. At cellular level, in asbestos exposed cells, inflammation, oxidative stress, and carcinogenesis has been associated with the alteration of the iron metabolism due to iron accumulation on fibers [23]. Similarly, iron impurities in CNTs have been demonstrated to participate in increased inflammation and oxidative stress in CNTs exposed mesothelial cells, in a "dose-dependent" manner [24–26].

At the molecular level, ROS may cause different injuries, such as gene mutations and structural alterations to the DNA, leading to deregulation in cell proliferation and apoptosis. Oxidative DNA damage is often characteristic of chronic inflammation, one of the main mechanisms underlying mesothelial transformation.

During the inflammation process, the cross-talk between inflammatory cells and damaged alveolar cells has been recognized to contribute to mesothelioma pathogenesis as well as other respiratory disease like lung fibrosis and lung cancer [27,28]. Lung fibrosis manifests with excessive deposition of collagen fibers in the extracellular matrix (ECM) and remodelling of the alveolar parenchyma, leading to a progressive loss of lung function. It includes a first acute inflammation phase where inflammatory cells infiltrate the tissue, secrete proinflammatory mediators (cytokines TNFα, IL1α, IL1β, IL6, chemokine CCL2, and fibrogenic growth factors TGF-β1 and PDGF-A), and collagen is deposited in the ECM. After this early response, granulomatous fibrotic foci deposits around the lesions are detectable. Activation of fibroblasts and formation of myofibroblasts (fibroblast-tomyofibroblast transition) and epithelial-to-mesenchymal transition (EMT) of alveolar type II cells are drivers of this process [29,30]. Lung fibrosis is one of the first documented injuries to lung described in asbestos-exposed workers 2,3 and the inflammatory process leading to fibrosis has been well characterized using long, needle-like Mitsui-7 MWCNT exposure in vivo [31,32]. The role of oxidative stress in CNTs-induced lung fibrosis was demonstrated through the use of the antioxidant N-Acetyl Cysteine, which interfered with NLRP3 inflammasome activation and generation of pulmonary fibrosis in mice [33].

In both asbestos- and MWCNT-exposed workers, markers of fibrosis, profibrotic inflammatory mediators and immune markers [21,34,35], as well as dysregulation in mRNAs and target genes linked to the activation of key pathways involved in several disease outcomes (e.g., cancer, respiratory and cardiovascular disease, and fibrosis) [36] have been found. Markers of oxidative stress and mitochondrial dysfunction have also been found in exposed workers [37].

The similarity between MWCNTs and asbestos due to their inflammatory and oxidative potential has been recently demonstrated in vivo with long MWCNTs (Nanostructured & Amorphous Materials, Houston, TX, USA; University of Manchester, Manchester, UK) and long fiber amosite asbestos instilled into the pleural cavity of mice. Exposure to long fibers but not to short fibers resulted in the development and progression of inflammatory lesions along the pleura and in the increase of markers of oxidative stress and genotoxicity. All exposed animals displayed pleural lesions (mesothelial hyperplasia and fibrosis), and chronic inflammation and, in 10–25% of animals exposed to long MWCNTs, the lesions progressed to pleural mesothelioma [38]. Different results were obtained with long NM-401 and Mitsui-7 MWCNTs. In this study, toxicity and inflammation were observed only in mice exposed to short MWCNTs (NM-400, NM-402, NM-403, and MWCNTs from CheapTubes, Brattleboro, VT, USA) [39].

However, other studies in vivo have demonstrated that both long and short industrial MWCNTs induced granulomatous changes in the lungs, development of pulmonary fibrosis, and inflammation accompanied by increase in vimentin, TGF-beta, IL-1b, IL-18, and cardiac fibrotic deposition [40–44]. Commercial short MWCNTs (tangled) (Graphistrength© C100; Arkema, France) showed prolonged TNF-α release in BAL of exposed rats associated with increased collagen staining [45].

Similar results were obtained with SWCNTs (Nikkiso Co., Ltd., Tokyo, Japan), showing strong persistent pulmonary inflammation [46]. The same group also demonstrated that

the shorter the length of SWCNTs is, the stronger the toxicity. Short SWCNTs (Nikkiso Co., LTD., Tokyo, Japan) with a length of 2.8 µm induced a weaker inflammatory response and pulmonary toxicity than those with a length of 0.4 µm [42].

It has also been demonstrated that chronic exposure to commercial short SWCNTs (CNI, Houston, TX, USA) induces tumor growth (subcutaneously injected) and metastasis to liver and lung through activation of EMT [47]. Cancer development (Bronchiolo-alveolar adenoma and carcinoma) was also found in 18% of mice exposed to a single intratracheal instillation of short SWCNT (Nikkiso Co., Ltd., Tokyo, Japan) [46].

For a long time, length has been considered a predictor of CNTs' adverse biological effects. However, even if this is true in some cases, many in vitro studies support the concept that the length of CNTs might not be a unique determinant of the biological response. Recently, shape and diameter have been correlated with accessibility to the macrophage interior subsequently affecting their degradation ability and, therefore, ROS production. Since alveolar clearance contributes to inhalation toxicity, the understanding of parameters predicting CNT toxicity is of crucial importance. This question has been challenged in many studies. Rigid, needle-shaped, long Mitsui-7 MWCNTs (diameter > 50 µm), which are poorly uptaken into phagosomes of alveolar macrophages, have been demonstrated to not induce ROS release. On the contrary, curved, straight, long and thin MWCNTs from different manufacturers, with diameters <20 µm which localize in vacuole-like compartments, have been demonstrated to generate intracellular ROS. For all the analyzed MWCNTs, increased levels of pro-inflammatory cytokines (IL-1α, IL-1β, MIP-1α, INF-γ, IL-18, MCP-1, and TNF-α) were found, implying that the inflammatory response might not be strictly related to the phagocytic ability of the macrophages [48]. ROS production from lung cells could be responsible for the inflammatory response of macrophages in the absence of phagocytic activity. While the rigid, straight, "needle-like" NM-401 MWCNTs, which are similar to Mitsui-7, are poorly uptaken by macrophages and do not cause an increase in NO production, lung fibroblast cells (V79) were demonstrated to be able to uptake NM-401, with 80% of fibers localized in endosomes, generating a consistent production of intracellular ROS [49]. Short NM-400 and NM-402 MWCNTs with a diameter <20 µm, are instead efficiently degraded by macrophages and induce an increase in NO accompanied by acute inflammation [50].

Markers of inflammation and oxidative stress were also studied in epithelial cells. Induction of oxidative stress have been described in lung epithelial cells exposed to NM-402 and NM-403 with values comparable to or higher than that of Mitsui-7 [51] while in BEAS-2B cells, a significant reduction in the levels of mRNA expression of pro-inflammatory cytokines IL-1β, IL-6 and IL-8 and an increase in the antioxidant HO-1 gene were found in long-term exposure (three weeks) to NM-403 [52]. However the authors associated these contradictory findings to the metal contaminants present in NM-403.

As a driver of lung fibrosis, the activation of the EMT program in lung epithelial cells by fibrous materials has been documented in four different studies in airway epithelial cells. Exposure to chrysotile asbestos, SWCNTs, Mitsui-7, and Mitsui-7-derived MWCNTs with the length reduced to 1.12 µm, at sub-toxic concentrations led to an increase in mesenchymal markers (α-smooth muscle actin, vimentin, metalloproteinases, and fibronectin), a decrease in epithelial markers (E-cadherin and β-catenin), and activation of the TGF-β–mediated signaling pathway [40,53,54].

Fibrogenic potential was also demonstrated with an in-house lung microtissue array device in airway epithelial cells exposed to non-toxic concentrations of short MWCNTs (CheapTubes.com accessed on 15th of October 2020) together with a significant increase in expression of the fibrogenic marker miR-21. These effects were not found in cells exposed to long MWCNTs [55].

All of the results reported above indicate that physico-chemical characteristics such as length and diameter could partially explain the different biological responses but, alone, might not be predictive of inflammatory response. Many variables such as the presence of CNTs of different lengths in the same preparation together with their heterogeneity

in experimental settings contribute to the difficulty in predicting their inflammatory and oxidative effects. Particularly in in vivo studies (Tables 1 and 2), different route of exposure and different endpoints analyzed have been used for the evaluation of pathological parameters. Even though studies comparing the inhalation and instillation of MWCNT showed that both methods induced pulmonary inflammation [56], inhalation is more powerful in inducing inflammation [57] and should be the preferred method for studies on accidental exposure during the manufacturing process since it recreates real situations better.

**Table 1.** Cancer development, histological changes, and inflammatory response observed in in vivo experiments with MWCNTs.


Abbreviations: "x": studies that have reported a relationship between these characteristics and exposure to the material.


**Table 2.** Cancer development, histological changes, and inflammatory response observed in in vivo experiments with SWCNTs.

Abbreviations: "x": studies that have reported a relationship between these characteristics and exposure to the material.

#### *4.2. Epigenetic Alterations*

It is well known that epigenetic changes in DNA and RNA play an important role in the regulation of gene expression by changing DNA accessibility to the cellular machinery, and switching on/off gene expression. As indicators of environmental insults, the study of epigenetics is a useful tool to understand disease-related mechanisms as well as serve as an indicator of disease risk. Among the epigenetic modifications affecting the genome, DNA methylation, the process by which a methyl group is added to carbon five in the cytosine pyridine ring forming 5-methylcytosine (5 mC) in DNA, is the most studied for the assessment of the potential hazard of fiber-like materials. Mesothelioma, as well other asbestos-related diseases, has been related to epigenetic changes, and the methylation changes of blood markers have been proposed as diagnostic and prognostic markers for mesothelioma [61–63]. In recent epidemiological studies in asbestos-exposed populations, a decrease in the levels of blood global 5-methylcytosine (5 mC) has been described in both healthy exposed workers and in those with benign asbestos-related disorders, confirming that global methylation could be a useful marker of asbestos exposure but, unfortunately, cannot be used as indicator of asbestos-related disease [64,65].

In MWCNT-exposed workers, changes in the methylation of specific genes mainly involved in DNA damage repair, cell cycle regulation, chromatin remodelling, and transcriptional repression (DNMT1, ATM, SKI and HDAC4 promoter) was described in a cross-sectional study [22]. Unlike with asbestos, no significant difference was found in total DNA methylation.

Hypermethylation of specific genes was also found in mice exposed to long MWCNTs (Nanostructured & Amorphous Materials, Houston, TX, USA; University of Manchester, UK) and long amosite fibers, which caused chronic inflammatory lesions or mesothelioma. Of particular importance is the epigenetic silencing of the *CDKN2A* locus, a well-known driver mutation in asbestos-induced mesothelioma, observed in mice exposed to both long MWCNTs and long amosite fibers [38].

Many in vitro studies have confirmed the methylation of specific genes. In 16HBE airway epithelial cells, in-house synthesized short MWCNTs and SWCNTs induced differentially methylated and expressed genes in cellular pathways related to DNA damage repair and cell cycle, with more pronounced effects in MWCNTs. No alteration of global DNA methylation was found [66]. An increased alteration on CpG sites after short -and long-term exposure has also been described for both benchmark short NM-400 MWCNTs and asbestos (CDKN1A and ATM among others) [66–70].

Together with specific gene methylation, other studies have also found a strong genome-wide DNA hypomethylation in airway epithelial cells (BEAS-2B and 16HBE) exposed to commercial short MWCNTs (CheapTubes, Brattleboro, VT, USA) and NM-400 and NM-401 [67–69,71].

It is important to note that most of the hypomethylated genes observed after two weeks of exposure to NM-401 became hypermethylated after four weeks of exposure [67], thus highlighting how time and particle type can trigger different and apparently discordant results.

In conclusion, many studies have demonstrated that change in methylation can be used as a marker of exposure to CNTs but heterogeneity of this class of nanomaterial does not allow for making generalizations. More studies are needed to expand our knowledge about epigenetic regulation of specific genes after CNT exposure. Given our current knowledge of asbestos, we know what genes are strictly linked to mesothelioma onset, and the results regarding epigenetic changes reported above suggest that CNTs could act via a similar mechanism.

#### *4.3. Genotoxicity, Alteration in DNA Repair, and Genome Instability*

Genotoxic effects can result from primary or secondary mechanisms. The first implies a direct interaction with the genetic material, the latter the oxidation of DNA by reactive oxygen/nitrogen species (ROS/RNS) generated during substance-induced inflammation. Both mechanisms could be involved in the genotoxic response elicited by MWCNTs.

Although CNTs are considered by IARC to be usually non-reactive and, for Mitsui-7 genotoxicity, have been demonstrated to act via secondary mechanisms, it cannot be excluded that defects in their structure occurring during the synthesis or functionalization could increase their reactivity [72,73]. Very recently, for long and short SWCNTs, the nucleus has been hypothesized to be the primary target site with DNA damage likely due to mechanical penetration [74].

Many studies, such as those described above, support the hypothesis that CNT genotoxicity could result from secondary mechanisms triggered by a strong inflammatory response and ROS release.

A genotoxicity study recently conducted in workers exposed to CNTs (unspecified manufacturer), revealed an 18.3% increase in telomere length and a 35.2% increase in mitochondrial DNA copy number from peripheral blood [75].

Asbestos-induced mesothelioma has been linked to polyploidization and aneuploidization, and MWCNTs seem to have similar adverse effects [76]. Chromosomal aberrations (polyploidy), and mitotic and chromosomal disruptions have been demonstrated for commercial MWCNTs (Hodogaya Chemical, Tokyo, Japan; Tokyo Chemical Industry, Tokyo, Japan; Showa Denko K.K, Tokyo, Japan), including MWCNT-7, with different length and shape (including straight fibrous, not straight fibrous (curved), and tangled MWCNTs) in Chinese hamster lung cell lines with straight fibrous being the more potent inducers of polyploidy. None of the seven MWCNTs analyzed caused structural chromosomal aberrations [76]. In the same model, NM-401 was found to be genotoxic, increasing HPRT mutant frequency [49].

In vivo experiments with long MWCNTs (Mitsui & Co. Ltd., Tokyo, Japan) showed a significant increase in DNA damage (comet assay) in the cells of lungs with straight MWCNTs but not with tangled MWCNTs. Moreover, straight MWCNTs caused an increase in DNA strand breaks in BAL cells collected after inhalation but not after pharyngeal aspiration [77]. DNA strand breaks were also observed after intratracheal instillation of straight NM-401 MWCNTs in the transgenic MutaTMMouse model. Moreover, both straight NM-401 and Mitsui-7 MWCNTs increased p53 expression predominantly in the area of fibrotic lesions (more pronounced for NM-401), and induced chronic inflammation and changes in the expression of genes linked to hallmarks of cancer. There was no evidence of a LacZ mutation [58].

Short commercial MWCNTs comprised of straight and tangled MWCNTs (Cheap-Tube, Brattleboro, VT, USA), were demonstrated to induce a dose- and time- dependent neutrophil influx in BAL and to cause DNA damage in the lungs of mice exposed by intratracheal instillation, with large MWCNTs diameter associated with increased genotoxicity (Analysis at 1, 28 and 92 days after exposure). All MWCNTs analyzed induced similar histological changes [60].

Another study using commercial short tangled MWCNTs (Graphistrength© C100; Arkema, France) did not disclose genotoxicity in lung cells or a microscopic change in the pleura. As the authors hypothesized, these effects could in part be ascribed to the

formation of agglomerates that are poorly uptaken by cells [45]. However, the lack of a positive control in the experimental setting could represent a weakness in the study.

Similar results have been seen in in vitro studies. Long-term exposure of primary human airway epithelial cells (SAECs) to commercial short SWCNTs (CNI, Houston, TX, USA), long Mitsui-7 MWCNTs (Mitsui & Co., Ltd., Tokyo, Japan) and Crocidolite, and mesothelial MeT-5A cells exposed to commercial long MWCNTs (Sigma-Aldrich, St Louis, MO, USA) have demonstrated a substantial increase in DNA damage in γH2A.X foci and p53 dysregulation [54,78].

Chromosome damage and chromosome mis-segregation have also been described in airway epithelial cells chronically exposed to sub-toxic doses of short NM-400 and NM-403 MWCNTs [52], while no primary DNA damage or oxidized DNA bases have been observed in short-term experiments with NM-400, NM-401, and NM-403 [50,79,80]

Contrasting results for Micronuclei (MN) formation assay were found in NM-401 exposed cells, according to the different methods used. With the cytokinesis-blocked micronucleus assay (CBMN), authors did not observe significant increases in the frequency of micronucleated binucleated cells or induction of DNA damage by the comet assay [81]. When analyzed by flow cytometry, NM-401 at 20 and 50 µg/mL were able to increase the MN formation [79]. No genotoxic effects with the CBMN assay were detected also for NM-400, NM-402, and NM-403 [81].

Bacterial reverse mutation tests and chromosomal aberration tests, according to the Organization for Economic Co-operation and Development (OECD) Guidelines for Testing of Chemicals, were conducted on straight, long, thin MWCNTs, revealing no structural or numerical chromosomal aberrations below a concentration of 50 µg/mL following shortterm exposure, both with and without metabolic activation [48]. However, this test is not suitable for studies with nanomaterials since they are not able to enter the bacterial cell wall, thus leading to the production of false-negative results.

Even though a definitive conclusion on the genotoxicity of CNTs is still impossible to draw, many results have indicated the presence of damaged DNA after exposure to CNTs. It is clear that for genotoxicity assessment, many variables, in addition to those mentioned previously, could interfere with the results. In particular, due to different responses in terms of DNA repair of different cell types, in vitro and in vivo models used represent a key factor together with the dose and time chosen for the analysis.

#### *4.4. Immortalization, Altered Cell Proliferation, Cell Death, or Nutrient Supply*

MWCNT-7 carcinogenicity has been demonstrated by different studies in mice in which the whole body has been exposed [82,83]. Nikkiso MWCNTs, which is similar to Mitsui-7, have also been demonstrated to induce pleural malignant mesothelioma and lung tumors in intratracheal instillation studies [59].

The transformation potential in vitro has been documented in different studies. After long-term exposure to commercial long MWCNTs (Sigma-Aldrich, St Louis, MO, USA), mesothelial MeT-5A cells showed features resembling a malignant transformation process and specifically an increase in cell proliferation and invasion capacity, morphology change, and DNA damage [78].

Similarly, after long-term exposure to short SWCNTs (CNI, Houston, TX, USA), Mitsui-7 (Mitsui & Co., Ltd., Tokyo, Japan) and Crocidolite asbestos, primary human small airway epithelial cells (SAECs) exhibited neoplastic and cancer stem cell-like properties, such as anchorage-independent colony formation, spheroid formation, anoikis resistance, and expression of cancer stem cell markers [54].

Altered cell proliferation was also described. Cell growth inhibition with benchmark NM-403 MWCNTs [52], and NM-400 and NM401 MWCNTs have been demonstrated in bronchial epithelial cells in long-term experiments [84] and, for NM401 and NM403, in short-term experiments without significant cytotoxicity [51].

Similar results were obtained with commercial short SWCNTs and MWCNTs (SES Research, USA; Heji, Hong Kong, China), in lung fibroblasts and in epithelial cells with short rod-like SWCNTs and straight MWCNTs showing higher toxicity [77,85,86].

Toxicity studies in macrophages mostly supported the hypothesis that rigidity and high diameters are as key factors underlying toxicity. Indeed, exposure to rigid, needleshaped Mitsui-7 MWCNTs, and Nikkiso and NM-401 MWCNTs all induce cytotoxicity in macrophage cells while NM-400 and NM-402 did not [48,50]. However, the opposite has also been described in rat alveolar macrophages acutely exposed to highly bent, lowdiameter NM-403 MWCNTs, which induced significant toxicity [87].

#### *4.5. Immunosuppression, Modulation of Receptor-Mediated Effects, and Electrophilicity*

Few data are available regarding the characteristics grouped below.

Available studies have demonstrated that CNTs can interact and activate the complement system, a key part of the immune system, and induce an early and sustained immunosuppressive response [44,88–90]. Moreover, it has been shown that SWCNT exposure in mice increases susceptibility to respiratory viral infections [91].

The ability of CNTs to act as an electrophile and then interact with cellular macromolecules, such as DNA, RNA, lipids, and proteins, has not been thoroughly investigated. It has been suggested that SWCNTs block K+ channel subunits by "plugging" the channel by virtue of the small diameter [92] and interact with TLR4 by hydrophobic interactions [93].

All of these studies suggest that the immunosuppression and modulation of the immune responses elicited by CNTs need further investigation. Indeed, an increased susceptibility to pathogens as well as immunosuppression could be a new and potentially significant mechanism of toxicity in humans.

#### **5. Discussion**

Nanotechnology is changing our world and is believed that it will improve our lives in the near future. CNTs are indeed remarkably valuable given their applications, ranging from drug delivery to electronics. Since we are at the beginning of the nanotechnology era, elucidation of the putative carcinogenicity of CNTs is also at the beginning. Intensive research is underway to understand their safety for human health and a remarkable data pool is being produced using different types of CNTs, models, methods, duration of exposure, amount of CNTs, and time points analyzed. While such heterogeneity is yielding many important results, it is, on the other hand, complicating the evaluation of the danger of CNTs. This situation well reflects the heterogeneity of this class of compounds as well as the different applications intended for their use, thereby making it particularly challenging to identify common features predicting their toxicity. It is not yet understood which aspects of carbon nanomaterials, e.g., surface areas, mass concentrations, lengths or a combination of these features or other factors, influence their toxicity. In addition, establishing criteria for preparation and dispersion, concentrations, models and methods to use, and also including reference materials, will undoubtedly play a crucial role in determining the reliability, reproducibility and comparability of data. In recent years, great improvements have been made in this direction and most non-human-based studies have reported a detailed description of the physiochemical characteristics of CNTs, the method used for their synthesis, the dispersion protocol and the percentage of the impurities present. However, despite these efforts, the lack of a complete characterization of CNT exposure in workers remains a crucial consideration. The type of CNTs varies both across companies and within them over time. Furthermore, in epidemiological studies, there is a high variability among instruments used for sampling and analysis of exposure, and there is still a low number of participants. All of these weaknesses, together with the lack of specific legislation addressing manufacturing processes for nanomaterials, make a direct comparison between studies difficult.

However, since the last IARC evaluation of CNT carcinogenicity, conducted in 2014, when enough evidence was available only for Mitsui-7, nine new studies have been performed on humans exposed to CNTs in the workplace, documenting markers of fibrosis, profibrotic inflammatory mediators, and immune markers [21,34,35,94]; epigenetic changes in genes related to DNA repair, cell cycle and repression of transcription [22]; deregulation in pathways and signaling networks linked to pulmonary and carcinogenic outcomes [36]; increase of oxidative markers in the exhaled breath condensates [37], increase in mtDNA copy number [75]; and development of respiratory allergies [95]. Recent findings in vivo have clearly indicated that CNTs induce a sustained inflammatory response and oxidative stress, and fibrosis and histological alterations have been documented in animals exposed to MWCNTs (Table 1) and SWCNTs (Table 2) by inhalation, aspiration, and tracheal instillation [32,44,58]. The development of mesothelial hyperplasia, mesothelioma, and lung tumors have been also described with SWCNTs and long fibers of both asbestos and MWCNTs [32,38,46,59] (Figure 2).

**Figure 2.** Hallmarks of cancer due to CNTs exposure in vivo and on human-based studies.

Less evidence is available for assessing cytotoxicity and genotoxicity and we are still far from reaching a consensus. It is nevertheless important to note that there are, however, new findings indicating DNA damage and gene-specific methylations after CNT exposure. In particular, the epigenetic silencing of the *CDKN2A* locus, a well-known driver mutation in asbestos-induced mesothelioma, has been documented in mice exposed to commercial long MWCNTs together with the loss of p16 and p19 protein expression [38].

In light of these recent studies analyzed, we agree with the need to evoke a global improvement of studies on exposed human populations as well as with the non-applicability of disproportionate precautionary measures of exposure control. However, considering the absence of any global agreement about the hazards of CNTs, we cannot take the risk creating another man-made tragedy like the case of asbestos where a century passed before its carcinogenicity was recognized, with many scientific papers defending its use to influence policy decisions on its hazards [13]. Moreover, years after its banning, we still have not achieved an asbestos-free environment and indeed the consequences thereof we still cannot predict.

Cancer is a multi-step process and, especially in the case of mesothelioma, it could takes years before it manifests itself. Fortunately, we are at the beginning of the CNT era and while we do not yet have data on the carcinogenicity of CNTs, we do have the opportunity to establish safe management of these materials. While we cannot precisely assess which modifications in the genome or in the epigenome will lead to mesothelioma onset, we do know that the long latency of malignant mesothelioma is sustained by decades of chronic inflammation in an aberrant microenvironment rich in ROS and the resulting oxidative DNA damage. We must carefully reflect on the data supporting the strong inflammatory potential of CNTs, similar to that of asbestos, as well as the data correlating CNT exposure with molecular alterations known to have a key role in mesothelioma onset

#### **6. Conclusions**

The heterogeneity of this class of substances is undoubtedly the main obstacle to reaching a consensus on their toxicity and more studies are needed to gain detailed knowledge on the effects of exposure to CNTs. We believe that future studies on CNTs toxicity must be assessed case-by-case and, on this premise, a new evaluation of the danger of CNTs for human health is urgently needed. We strongly support the need to create a repository of biological samples from CNT-exposed workers in order to monitor biologically relevant changes over time and to encourage research collaboration within different areas of expertise. In any case, an adequate system for monitoring the health of workers exposed to this putative carcinogen remains the basis on which to build future research.

**Author Contributions:** Conceptualization, writing—review and editing, M.B.; project administration, review, A.G. Both authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by funding from the TOMA institute, Italy. The APC was funded by University of Siena, Dept. of Medical Biotechnology, Siena, Italy.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


## *Review* **Ancillary Diagnostic Investigations in Malignant Pleural Mesothelioma**

**Alex Dipper \*, Nick Maskell and Anna Bibby**

Academic Respiratory Unit, University of Bristol, Bristol BS105NB, UK; Nick.Maskell@bristol.ac.uk (N.M.); anna.bibby@bristol.ac.uk (A.B.)

**\*** Correspondence: alex.dipper@bristol.ac.uk

**Simple Summary:** Malignant pleural mesothelioma (MPM) is a cancer affecting the covering of the lung (the pleura). This commonly causes a build-up of fluid around the lung, called a pleural effusion. Draining the pleural effusion can improve breathlessness and tests can be performed on the fluid. However, for most patients with MPM, a sample of tissue from the pleura, called a biopsy, is required in addition to make the diagnosis. Sometimes, due to medical conditions, frailty or personal preference, patients may not be able to have a biopsy. This review article discusses additional tests used in this situation to help doctors make a diagnosis of MPM. These techniques include tests on pleural fluid using "immunocytochemistry" methods, biomarkers and scans. Although, without a biopsy, no test in isolation can diagnose MPM, combining information from different types of tests and reviewing results among a specialist team can enable a consensus diagnosis.

**Abstract:** For a number of patients presenting with an undiagnosed pleural effusion, frailty, medical co-morbidity or personal choice may preclude the use of pleural biopsy, the gold standard investigation for diagnosis of malignant pleural mesothelioma (MPM). In this review article, we outline the most recent evidence on ancillary diagnostic tests which may be used to support a diagnosis of MPM where histological samples cannot be obtained or where results are non-diagnostic. Immunocytochemical markers, molecular techniques, diagnostic biomarkers and imaging techniques are discussed. No adjunctive test has a sensitivity and specificity profile to support use in isolation; however, correlation of pleural fluid cytology with relevant radiology and supplementary biomarkers can enable an MDT-consensus clinico-radiological-cytological diagnosis to be made where further invasive tests are not possible or not appropriate. Diagnostic challenges surrounding non-epithelioid MPM are recognised, and there is a critical need for reliable and non-invasive investigative tools in this population.

**Keywords:** malignant pleural mesothelioma; pleural effusion; biomarkers

#### **1. Introduction**

Arising predominantly from the pleural or peritoneal surface (less commonly the pericardium and tunica vaginalis), mesothelioma grows insidiously, often resulting in an advanced stage at clinical presentation. Whilst research into innovative treatment options is an active area of interest and brings new hope for patients, malignant pleural mesothelioma (MPM) remains relatively refractory to conventional therapies. Consequently, prognosis is poor, with a median survival of just 9.5 months and a 3-year survival rate of 12% [1,2].

An association with asbestos was first observed in 1960 in a case series of 33 patients with pleural mesothelioma from the Asbestos Hills in the Cape Province of South Africa [3]. Today, 85% of all mesotheliomas in males are attributable to occupational asbestos exposure, with para-occupational exposure being a recognised cause in women [4]. Despite a ban on asbestos products in 52 countries by 2010 [5], the long latency period from exposure to disease (typically 30–40 years) and continued unregulated use in countries such as

**Citation:** Dipper, A.; Maskell, N.; Bibby, A. Ancillary Diagnostic Investigations in Malignant Pleural Mesothelioma. *Cancers* **2021**, *13*, 3291. https://doi.org/10.3390/ cancers13133291

Academic Editors: Daniel L. Pouliquen and Joanna Kopecka

Received: 10 June 2021 Accepted: 18 June 2021 Published: 30 June 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

India, Brazil and Russia means that MPM continues to represent a significant global health concern, with an estimated burden of 38,400 cases per year worldwide [6].

Other aetiological mechanisms include genetic predisposition, with inherited germline mutations of the BRCA 1-associated protein (BAP1) gene (a tumour suppressor gene involved in modulation of transcription and DNA repair) identified amongst families with high incidence of mesothelioma in 2011 [7]. Exposure to other elongated mineral particles (including environmental exposure to erionite and fluoro-edenite in Turkey, USA and Mexico) and ionising radiation are also implicated [8]. Pathogenic mechanisms of carcinogenesis following asbestos fibre inhalation highlight a cycle of genetic and cellular damage with chronic inflammation [2,4,9–11].

Four main histological subtypes of MPM are described; epithelioid, sarcomatoid, biphasic and desmoplastic, with epithelioid associated with the most favourable prognosis (median survival of 13 months) and sarcomatoid the least (median survival 4 months) [12]. With no established role for surgical resection outside of clinical trials [1], histological diagnosis of MPM typically relies on biopsy samples. Thoracoscopic pleural biopsy is recommended as the gold standard for investigating an undiagnosed pleural effusion where the differential includes MPM, with diagnostic yields of 95% and higher [13]. Alternatively, where contrast-enhanced thoracic computed tomography (CT) demonstrates focal areas of abnormal pleura, image-guided needle biopsy may be employed to obtain tissue [8,14,15].

There is a cohort of patients for whom frailty, medical co-morbidity or personal choice preclude the use of invasive pleural biopsy. The 2020 UK National Mesothelioma Report showed the median age of patients diagnosed with pleural mesothelioma was 76 years and that over 20% of patients had stage IV disease at diagnosis [16]. Furthermore, a multinational population-based evaluation of 9014 patients demonstrated that more than half of those diagnosed with mesothelioma were aged 70 years or older [17]. Given demographic trends, the proportion of elderly patients will continue to rise over coming decades, with increasing comorbidity further complicated by advanced stage at disease presentation. Diagnostic approaches that are tolerable to and appropriate for patients of higher age or with significant comorbidity are increasingly necessary. Additionally, a proportion of patients who are considered suitable to undergo pleural biopsy at initial assessment go on to have a protracted diagnostic pathway, with repeated procedures yielding equivocal or non-diagnostic results.

Although international guidelines do not advocate cytology-based diagnoses of MPM in patients who are fit for further diagnostic tests [1], the importance of obtaining a diagnosis for frail patients who are unable to undergo invasive procedures to obtain a biopsy is no less significant. Confirmation of a diagnosis is important for future planning and to enable patients to access financial compensation. In some regions, a multi-disciplinary team (MDT) diagnosis based on cytological, radiological and clinical information is sufficient to avoid requirements for a post-mortem examination after death [18].

In this article, we will explore and outline the most up-to-date evidence on ancillary diagnostic tests currently available in clinical practice. We will focus on techniques which may be used to support a diagnosis of MPM from cytological specimens and other less invasive modalities, where histological samples cannot be obtained or where results may be non-diagnostic.

#### **2. Pleural Fluid (PF) Cytology**

Diagnostic thoracentesis is the primary means of obtaining PF for evaluation and is an essential step in the initial investigation of a unilateral pleural effusion [19]. Diagnostic cytology on PF can spare the patient more invasive investigations to obtain a tissue biopsy, reducing the risk of procedural complication with both cost and time saving in addition. However, the diagnostic yield of MPM from conventional PF cytology alone is highly variable, with sensitivity ranging from 16% to 73% [1]. In one study of 921 patients with an undiagnosed unilateral pleural effusion, fluid cytology was diagnostic in only 9 of 148 (6%) participants with MPM [20].

Several factors contribute to the wide range of sensitivities quoted. Whilst epithelioid cancers can shed malignant cells into pleural effusion fluid, this is rare in sarcomatoid subtypes. Cytological diagnosis is usually limited, therefore, to the epithelioid subtype. Heavy bloodstaining or rich inflammatory cell infiltrate may additionally reduce cellular yield in effusion specimens. Concentration techniques such as cell block and cytospin preparations can overcome these problems and enhance detection of malignant cells. Cell blocks can also provide a substrate on which adjunctive tests, including immunocytochemical and molecular techniques, can be applied. [8,21,22]

Cytologist experience is another important consideration, with cytopathology being a recognised subspecialty in its own right. For example, morphological appearances of benign reactive mesothelial cells can overlap with malignant cells, complicating diagnosis and demanding meticulous assessment. The volume of PF submitted for analysis may be an additional limitation [1,23], with the British Thoracic Society recommending that 20–40 mL should be sent for evaluation [19].

An important limitation on cytology-based diagnosis is the inability to determine tumour invasion into the lung or chest wall on the basis of PF cytology alone [21,24]. Cytological yield in epithelioid mesothelioma is, however, higher in the presence of visceral pleural invasion. In one study of 75 patients with epithelioid MPM, 37/45 (82%) with positive PF cytology at initial thoracentesis had evidence of visceral pleural invasion at local anaesthetic thoracoscopy (defined as masses, nodules, thickening or mixed appearance) compared with 9/30 (30%) patients having negative cytology, giving an odds ratio for an association between visceral pleural invasion and cytological positivity of 11.87 (95% confidence interval (CI): not stated; *p* < 0.001) [25].

#### **3. PF Immunohistochemistry (IHC) and Molecular Techniques**

Initial cytomorphology may be sufficient to confirm the presence of malignant cells in PF after routine staining and, in some cases, may confirm MPM. However, more often, ancillary techniques are required to discriminate benign from malignant mesothelial populations and to differentiate MPM from carcinoma or neoplasms of other origins (for example, melanoma). Recent advances in immunocytochemical and molecular testing have facilitated these diagnostic steps [22,26].

#### *3.1. Discriminating Benign from Malignant Mesothelial Populations*

Reactive mesothelial proliferation is a common mimic of MPM (and metastatic carcinoma) and has numerous causes, including infection, pulmonary infarction, trauma, autoimmune disease and drug reactions [27]. Cytomorphological features overlap with MPM and include high cellularity, numerous mitotic figures and cytologic atypia. The inability to evaluate tissue invasion in cytology-based specimens means that reactive mesothelial proliferation is more frequently documented in cytologic specimens than in tissue biopsies [24].

Certain immunocytochemical stains are more likely to be positive in benign mesothelial cell proliferation and other stains in malignant mesothelial proliferation. However, most IHC staining patterns do not reliably differentiate malignant from benign mesothelial proliferation. Desmin, reported previously to favour benign reactive mesothelium, shows positivity in up to 56% of mesotheliomas [28]. Similarly, whilst epithelial membrane antigen (EMA), p53 and insulin-like growth-factor 2 messenger ribonucleic acid (RNA) binding protein 3 (IMP-3) may support a diagnosis of malignancy, benign reactions can also stain positively for these markers [29]. Whilst positive staining with glucose transporter 1 (GLUT1) may have a higher specificity for malignant cell populations in pleural biopsy specimens [1], cytological studies demonstrate lower specificity, with 9/50 patients with benign reactive mesothelial proliferations demonstrating positive polyclonal GLUT-1 staining in one study [30] and 14/38 participants with benign effusions staining positive in another [31].

Detection of specific mesothelioma-associated genetic mutations can help confirm the presence of malignant cells. Loss of BAP1 can be demonstrated on IHC staining and is highly specific for malignancy, whilst fluorescent in situ hybridisation (FISH) can detect deletion of the *CDKN2A/P16* gene, commonly seen in MPM.

#### 3.1.1. BAP1 Loss

BAP1 is a nuclear ubiquitin hydrolase, which functions as a tumour suppressor, and is encoded by the BAP1 gene. It controls DNA repair, expression of genes related to cell cycle and cell proliferation. It can also induce cell death. Cells with reduced or absent BAP1 are unable to repair damaged DNA and cannot execute apoptosis. BAP1-mutant cells are therefore prone to malignant transformation [10].

Somatic mutation of the BAP1 gene in mesothelioma was first described in 2011, with mutations occurring in approximately 70% of epithelioid mesotheliomas [10]. Germline BAP1 mutation is less common, occurring in approximately 1–2% of MPM, usually in the context of the autosomal dominant BAP1 cancer predisposition syndrome [29,32]. Germline BAP1 loss is associated with earlier onset MPM tumours, as well as other BAP1-related malignancies such as uveal melanoma.

BAP1 loss (defined as absence of nuclear staining when a positive internal control is present on a slide) may occur by mutation, biallelic deletion or deletion/insertion [8] and is most reliably detected by IHC [32]. Cells expressing at least one wild-type copy of BAP1 retain IHC staining. Notably, even in tumours arising from germline BAP1 mutation, non-tumour cells express a single wild-type copy and hence produce a positive IHC response. To show loss of BAP1 immunoreactivity, both copies must be mutated, either by a combination of germline and somatic mutation events, as in BAP1 cancer syndrome, or by two somatic events in sporadic cancers [29].

Loss of BAP1 expression has been repeatedly validated in differentiating MPM from benign mesothelial populations and is now in routine use in many pathology laboratories. A recent meta-analysis identified 12 studies of 1824 patients (1016 with MPM), published between 2015 and 2017. The overall pooled sensitivity of BAP1 loss for malignant mesothelioma was 0.56 (95% CI: 0.50–0.62) and specificity 1.00 (95% CI: 0.95–1.00). The area under curve (AUC) was 0.72, indicating moderate diagnostic accuracy. Notably, all studies were of retrospective design, and only four included more than 100 participants. Heterogeneity was evident, with potential explanations including different cut-off values for BAP1 loss, inclusion of participants with pleural and peritoneal mesothelioma and variation in diagnostic accuracy across mesothelioma histological subtypes. For example, the sensitivity ranged from 0.07 (95% CI: 0.00–0.72) in sarcomatoid MPM to 0.74 (95% CI: 0.66–0.80) in epithelioid [33]. Offering additional explanation for this low diagnostic sensitivity, 30–40% of mesotheliomas have been shown to carry a wild-type BAP1 and therefore stain positively in a similar manner to benign lesions [10].

In a subgroup meta-analysis comparing the diagnostic performance of BAP1 loss in histology and cytology specimens, near identical sensitivity and specificity was observed. However, data from the 5 studies evaluating cytology specimens demonstrated reduced diagnostic accuracy with an AUC of 0.69 [33]. Studies of BAP1 loss in cytology specimens have, to date, been hindered by retrospective design, small sample size and the use of cytology specimens in subgroup analyses. Well-designed research is required to accurately determine the diagnostic potential of BAP1 loss in cytology specimens in order to improve current diagnostic pathways and potentially avoid the need for additional invasive procedures.

As a stand-alone test, BAP1 loss has moderate diagnostic sensitivity with excellent specificity for MPM. BAP1 loss is therefore reliable as a "rule in" for mesothelioma, but pleural malignancy cannot be excluded in its absence. Notably, BAP1 loss is uncommon in sarcomatoid and desmoplastic mesothelioma and is demonstrated in other malignancies including melanoma and renal cell carcinoma [34]. Superior diagnostic accuracy may be achieved in combination with other adjunctive tests.

#### 3.1.2. p16 Fluorescence In Situ Hybridization (FISH)

Homozygous deletion of the 9p21 locus is one of the most common genetic alterations in MPM. Its loss affects a cluster of genes, including p16 (also known as cyclin-dependent kinase inhibitor (CDKN)-2A), CDKN2B and methylthioadenosine phosphorylase (MTAP). p16/CDKN2A is a tumour suppressor gene that is present in all healthy cells. Its normal function results in the cessation of the cell cycle; hence, inactivation results in uncontrolled cell proliferation and tumour development.

Homozygous deletion of P16 can be detected using FISH in both cytological and histological specimens [35]; however, the diagnostic sensitivity for MPM is relatively low at 0.53 (95% CI: 0.35–0.70), despite gene profiling studies demonstrating p16/CDKN2A loss in up to 80% of MPM tumours. In part, the low sensitivity reflects variation in p16 deletion across the different MPM subtypes (90–100% loss in sarcomatoid variant compared with a 70% loss in epithelioid and biphasic), although other alterations that affect the 9p21 locus and cannot be detected by FISH also contribute [1,8,21,23,24,29].

An alternative approach, where histological specimens are available, is the application of IHC staining to determine p16 protein expression in cells, which could represent a more accessible ancillary test to laboratories where FISH cannot be performed [36]. However, the sensitivity to discriminate MPM from reactive mesothelial hyperplasia using p16 IHC in combination with BAP1 loss was 10% lower than those of more traditional FISH techniques in one study [35]. IHC techniques may be employed, in addition, to detect MTAP loss, distinguishing malignant from benign proliferations with a specificity of 100% and a sensitivity of 43% (increased to 79.5% when used in combination with BAP1 IHC) in cell block specimens from pleural effusions [37,38]. IHC for MTAP can also discriminate sarcomatoid MPM from fibrous pleuritis. A more recent multicentre evaluation of MTAP loss by IHC demonstrated a 78% sensitivity and a 96% specificity for CDKN2A homozygous deletion, suggesting it to be a reliable surrogate for CDKN2A FISH [39]; however, the use of MTAP is not yet recommended by international guidelines [1,8,15].

Overall, when used in isolation, both FISH and IHC techniques for p16 deletion are limited by low sensitivity. Consequently, whilst p16 deletion can confirm a suspected diagnosis of malignancy, failure to detect its loss does not exclude a diagnosis of MPM. However, combining testing for p16 loss with IHC for BAP1 loss has been shown to increase diagnostic sensitivity (combined sensitivity 0.76 (95% CI: 0.62–0.88)) [40]. Therefore, if BAP1 is intact or a sarcomatoid mesothelioma is suspected, additional testing with p16 FISH may strengthen diagnostic certainty [21] and help to discriminate benign from malignant mesothelial cell populations (Figure 1).

**Figure 1.** A suggested diagnostic approach where distinction of malignant from benign mesothelial proliferation is unclear on initial fluid cytology. BAP1 loss and p16 deletion support the diagnosis of MPM. MPM, malignant pleural mesothelioma; ICC, immunocytochemistry; FISH, fluorescence in-situ hybridization; BAP1, BRCA 1-associated protein.

#### *3.2. Distinguishing Mesothelioma from Carcinoma*

Distinguishing mesothelioma from other causes of malignant pleural effusion is critical in guiding therapeutic strategies and prognosis. Malignancies commonly metastasising to the pleura include lung cancers, breast and gastrointestinal carcinomas. Distinction between epithelioid MPM and carcinomas may be made on morphology and simple histochemical staining alone. As no one marker exhibits a 100% specificity, guidelines recommend a combination of at least two positive mesothelial markers (calretinin, cytokeratin 5/6, Wilms tumour 1 and D2-40) and at least two negative adenocarcinoma IHC markers (thyroid transcription factor 1 (TTF1), carcinoembryonic antigen (CEA) and Ber-EP4) (see Table 1) [1]. Positive markers of other tumour types should be used for differential diagnoses of metastatic carcinomas from other sources, such as hormone receptors in breast and ovarian cancer and PAX8 in renal cell carcinoma [1,15,24].

BAP1 loss may play a role in differentiating mesothelioma from carcinoma, with loss in 46/53 (87%) pleural and peritoneal mesotheliomas compared with 4/204 (2%) (*p*: <0.001) carcinomas in one study [41]. Further evaluation of the role of BAP1 loss in this context is required, however, before universal adoption is recommended.

#### *3.3. Distinguishing Mesothelioma from Other Malignant Cell Neoplasms*

Malignant pleural effusion may be the first presentation of an unknown primary cancer. In this setting, appropriate immunocytochemical panels often enable a precise diagnosis, starting with CK7 and CK20 staining [42]. Other differential diagnoses of MPM depend on histologic category, with epithelioid MPM requiring distinction from carcinomas, sarcomatoid MPM from sarcomas and other spindle cell neoplasms, mixed MPM from other mixed or biphasic tumours such as synovial sarcoma and desmoplastic MPM from fibrous pleuritis. Immunostain selection in this setting would depend on basic morphology [24].

Affirmative markers used in the evaluation of epithelioid MPM are of limited utility in sarcomatoid tumours. More usefully, cytokeratin markers, such as CAM5.2, are important in differentiating sarcomatoid MPM (positive staining) from sarcoma, which is usually keratin-negative [43]. D2-40 (podoplanin) can be used to differentiate sarcomatoid MPM from pulmonary sarcomatoid carcinoma (which also stains positively for TTF1, napsin and p40/p63). Synovial sarcoma can be confirmed by molecular testing for the X; 18 translocation [24].


**Table 1.** Immunohistochemical markers for differentiating tumour types in malignant pleural effusion [12,24]. Adapted from Bibby et al. [44].

Immunocytochemical markers are summarised in Table 1.

#### **4. Diagnostic Biomarkers**

Biomarkers present an attractive solution to diagnostic challenges posed by MPM, and consequently, a large number of studies have evaluated potential targets in serum, plasma, PF and exhaled breath. An ideal marker should be obtainable by minimally

invasive means and be sufficiently sensitive to detect most cases of MPM, whilst also being highly specific, to avoid false positive results and discriminate individuals with MPM from other pathologies. Protein biomarkers of interest include mesothelin, osteopontin and fibulin-3 [45].

#### *4.1. Mesothelin*

Identified in the early 1990s as a surface antigen on ovarian cancer cells, mesothelin is a glycoprotein thought to play a role in cell adhesion and signalling. The mesothelin gene, MSLN, encodes a precursor protein from which membrane-bound mesothelin and a soluble protein megakaryocyte potentiating factor (MPF) are formed. These are commonly referred to as "soluble mesothelin-related peptides" (SMRPs). In normal tissue, mesothelin is only found on mesothelial cells; hence, serum levels of SMRP are low. However, increased concentrations of SMRPs are found in serum samples of patients with ovarian and pancreatic cancers, in addition to mesothelioma. In 2003, Robinson et al. demonstrated that patients with MPM had significantly higher concentrations of serum SMRP than asbestos-exposed healthy controls, non-asbestos-exposed healthy controls and patients with non-mesothelioma malignant or inflammatory pleural disease. They reported a sensitivity of 84% (95% CI: 73–93) and a specificity of 100% (95% CI: 91–100) for MPM. SMRP concentrations were higher in patients with epithelioid tumours and in those with a large tumour bulk (maximum tumour width: >3 cm) [11,46,47]. In contrast, SMRP was less likely to be raised in people with sarcomatoid and biphasic disease; however, small study numbers and non-disclosure of histologic subtype in some studies mean that accurate sensitivity and specificity estimates are difficult to derive for these tumour subtypes [11].

Serum mesothelin has become the most widely studied diagnostic biomarker in MPM, with a meta-analysis in 2014 identifying 28 relevant publications, involving 7550 patients [48]. Pooled sensitivity and specificity estimates were found to be 0.61 and 0.87, respectively, lower than indicated in previous studies. This is mostly accounted for by heterogeneity across the included studies, although publication bias may also play a role. Heterogeneity arose from the use of various ELISA assays, different cut-off values and differences in participant characteristics (i.e., mesothelioma subtypes and choice of control groups). The negative likelihood ratio (NLR) value was 0.43, meaning if participants were serum-SMRP-negative, the probability of having MPM was still moderate at 43%. The authors reported that low sensitivity limited the added value of SMRPs but a positive result may be helpful in confirming MPM, with a positive likelihood ratio of 5.71 [48].

PF mesothelin has been studied as an alternative biomarker, as mesothelin is shed from mesothelioma tumour cells directly into pleural effusion fluid. In 2005, Pass et al. identified that SMRP levels were significantly higher in PF samples from 45 patients with MPM compared to 30 healthy controls [49]. In the first study to assess the clinical utility of PF SMRP, Davies et al. demonstrated levels were 10.9 times greater in patients with MPM compared to benign pleural disease and were highly reproducible [50]. They concluded that the measurement of PF mesothelin contributed valuable additional information to PF cytology alone, especially where initial cytology results were inconclusive. In a metaanalysis by Cui et al., pooled estimates of sensitivity were higher for PF SMRP than serum samples (0.79 compared to 0.61) with PF SMRP specificity remaining robust at 0.85 [48].

Although considered as the current "gold standard" biomarker for MPM in some international guidelines [15], neither serum nor pleural fluid mesothelin is recommended as diagnostic tests in isolation. With low sensitivity, a negative result adds little value and is a frequent finding in non-epithelioid disease. In contrast, a positive result increases the likelihood of mesothelioma; however, false positives are possible in benign inflammatory conditions such as benign asbestos pleural effusion (BAPE) or in the presence of impaired renal function [51]. Consequently, mesothelin testing should be considered as an adjunct in patients with suspicious or inconclusive cytology, who are unsuitable for or decline invasive diagnostic tests with a high pre-test probability of MPM [1,4,8]. Further research into the utility of biomarkers in MPM diagnosis and better understanding of markers of non-epithelioid disease may help to elucidate the role of this test in the diagnostic pathway.

#### *4.2. Other Diagnostic Biomarkers*

Osteopontin, a protein mediator of cell matrix interaction, cell signalling and tumour development, has been viewed as a promising biomarker for MPM, but results have been inconsistent. In a meta-analysis of six studies, the overall diagnostic sensitivity and specificity were 0.65 (95% CI: 0.6–7.0) and 0.81 (95% CI: 0.78–0.85), respectively. Notably, the majority of included studies evaluated serum and/or plasma osteopontin from frozen samples with uncertainty regarding the long-term stability of osteopontin in frozen specimens. Degradation of osteopontin during the freezing and defrosting process may explain the low detection rates of this protein in retrospective studies [52]. Similar to mesothelin, the clinical utility of osteopontin is limited by low sensitivity, and further understanding of its added diagnostic value in comparison to other biomarkers is required.

Fibulin 3, an extracellular matrix glycoprotein mediator of cell-to-cell and cell-tomatrix communication, is detectable in blood and PF with a small number of studies reporting varied outcomes on its potential as a biomarker for MPM. Initially promising, with a 97% sensitivity and a 95% specificity to determine MPM from other causes of pleural effusion in one study [53], subsequent analyses have suggested a sensitivity as low as 22% [54]. A questionable diagnostic value was highlighted by one study, with no difference in fibulin 3 levels in pleural effusion samples of patients with MPM and controls. Whilst plasma levels were higher in patients with MPM compared to in controls in a population in Sydney, this was not replicated in a cohort of patients studied in Vienna and the diagnostic accuracy was low (receiver operating curve analyses overall accuracies of 63.2% and 56.2% for correct diagnostic characterisation of MPM in the Sydney and Vienna cohort, respectively). The authors did, however, observe that low pleural effusion fibulin 3 levels were significantly associated with better survival [55]. A meta-analysis of 8 studies demonstrated a pooled diagnostic sensitivity of blood fibulin 3 of 0.87 (95% CI: 0.58–0.97) and a specificity of 0.89 (95% CI: 0.77–0.95) [56]. A subsequent meta-analysis of 7 studies demonstrated a lower overall sensitivity from pooled studies of blood and pleural effusion samples of 0.62 (95% CI: 0.45–0.77) and a specificity of 0.82 (95% CI: 0.73–0.89) [57]. Ultimately the value of fibulin 3 in diagnosing MPM remains unclear, with prospective validation studies ongoing [58].

#### **5. Imaging Techniques**

CT with contrast enhancement is the primary imaging modality used for diagnosis and staging of pleural malignancy and can identify the primary tumour, intrathoracic lymphadenopathy and extrathoracic spread [59]. Positive features of malignant pleural disease include circumferential pleural thickening, nodular pleural thickening, parietal pleural thickening of greater than 1 cm and mediastinal pleural involvement [60]. The diagnostic accuracies of CT for detection of pleural malignancy are 68–97% with specificities of 78–89% [1]. CT scanning is widely available and has high clinical utility. However, it has limited soft tissue differentiation, and early malignant disease with minor pleural thickening can be missed. Additionally, subtle invasion of certain structures may be challenging to identify, which has implications for the accuracy of staging. Timing of contrast and reporting of images by non-thoracic radiologists add further variability. Subsequently, 35–46% of patients with pleural malignancy will have a "benign" CT report in routine practice [61].

Differentiating mesothelioma from metastatic pleural malignancy can also be challenging. Parenchymal lung tumours with mediastinal or hilar lymphadenopathy may indicate metastatic pleural disease, whereas the presence of pleural plaques, involvement of the interlobar fissure and absence of lung parenchymal masses favour MPM [1]. It may be particularly difficult to differentiate MPM from pleural metastatic disease, if the tumour presents as a localised pleural or subpleural nodule, a localised anterior mediastinal mass

or involves the diaphragmatic pleura with liver invasion, especially in the absence of a pleural effusion [43].

Alternative imaging modalities have been proposed for use in MPM. Positron emission technology (PET)-CT combines high-resolution CT scanning with an injection of a metabolic tracer which accumulates at areas of metabolic activity. Uptake is assessed at regions of interest and reported as standard uptake values (SUV), with a threshold value of 2.0 reliably differentiating between benign and malignant disease [4]. A meta-analysis of 11 PET-CT studies reported a pooled sensitivity of 95% (95% CI: 92–97%) and a specificity of 82% (95% CI: 76–88%) for differentiating malignant from benign disease [62]. False positive results are common, however, particularly in the context of prior talc pleurodesis, active pleural infection, or indolent inflammation such as tuberculous pleuritis. PET-CT cannot distinguish MPM from metastatic pleural disease and, due to poor spatial resolution, has low sensitivity (78%) for extrapleural invasion [61]. Whilst lacking specificity to diagnose MPM routinely, PET-CT may provide functional information on pleural lesions, although it does not appear to be helpful in guiding choice of site for biopsy [63]. It is currently recommended only for staging patients in whom the presence of distant metastatic disease would alter treatment approach [1,8,15].

Magnetic resonance imaging (MRI) offers higher soft tissue contrast than CT, resulting in an increased sensitivity for chest wall and diaphragm invasion, higher contrast with adjacent effusion and higher inter-observer agreement [64]. The contrast enhanced perfusion augments sensitivity in detection of pleural malignancy, even where pleural thickening is minimal [64]. In addition to differentiating malignant from benign pleural disease, diffusion-weighted MRI (DWI-MRI) has distinguished between epithelioid and sarcomatoid MPM with a sensitivity of 60% and a specificity of 94% [1]. At present, the added value of MRI in equivocal or atypical CT scans is unclear, with prospective evaluation required, but, where available, MRI may be considered in difficult diagnostic cases to better delineate invasive disease [1,8,15].

#### **6. Future Directions**

The search for novel diagnostic biomarkers is expanding and encompasses multiple branches of medical science. Proteomic analysis has identified new panels of candidate biomarkers [65] with prospective multicentre evaluation of a novel assay ongoing [58]. Gene-expression-based classification has outperformed BAP1 and p16 FISH [40]. Deeper understanding of the genomic and epigenomic factors relevant to MPM may herald new diagnostic techniques that better distinguish MPM from other tumours [66–68]. Circulating plasma micro-RNA [69] and metabolomic profiling [70,71] of PF are other experimental areas of interest.

Whilst these studies may yield new markers which negate the requirement for invasive tissue sampling, all are limited currently to the research setting and are not yet available in clinical practice.

As the range of therapeutic options for MPM expands, the importance of genetic and molecular phenotyping of tumours to enable targeted treatment will increase. Currently, no marker is able to provide this level of personalised tumour phenotyping, so tissue biopsies are likely to remain the diagnostic gold standard for the foreseeable future.

To obtain tissue in patients fit to undergo invasive procedures, a "direct-to-LAT" approach (pathway stratification where selected patients proceed directly to local anaesthetic thoracoscopy (LAT) to obtain pleural biopsies) may be employed in patients where the pre-test probability of MPM is high and the anticipated yield from PF cytology is low [72]. However, a streamlined diagnostic approach is required for more frail patients and those who choose not to undergo pleural biopsy. Research to determine the combined value of the investigations discussed in this article is essential to formalise integrated non-invasive pathways for the diagnosis of MPM.

#### **7. Conclusions**

For patients in whom malignant pleural mesothelioma is suspected, tissue diagnosis remains the gold standard and is the only method that can confirm the presence of invasive disease. However, for those unable or unwilling to undergo tissue sampling, the low sensitivity of pleural effusion cytology can be augmented by incorporating ancillary techniques such as immunocytochemical markers to increase reliability [8]. No adjunctive test has a sensitivity and specificity profile to support use in isolation, but findings such as BAP1 loss can provide additional support for a suspected diagnosis if the pre-test probability is high. Where diagnoses remain challenging, even despite use of ancillary techniques, expert radiological review of disease distribution on imaging and occupational history of asbestos exposure are important considerations. Correlation of PF cytology with relevant radiology and supplementary biomarkers can enable an MDT-consensus clinico-radiological-cytological diagnosis to be made, where further invasive tests are not possible or not appropriate [18]. Diagnostic challenges surrounding non-epithelioid MPM are recognised, and there is a critical need for reliable and non-invasive investigative tools in this population.

**Author Contributions:** Conceptualization, A.D., N.M. and A.B.; writing—original draft preparation, A.D. All authors have read and agreed to the published version of the manuscript.

**Funding:** This review article received no external funding.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Acknowledgments:** The authors would like to thank Richard Daly for reviewing the pathology-based information within the manuscript.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


## *Article* **Liquid Biopsies from Pleural Effusions and Plasma from Patients with Malignant Pleural Mesothelioma: A Feasibility Study**

**Gabriele Moretti <sup>1</sup> , Paolo Aretini <sup>2</sup> , Francesca Lessi <sup>2</sup> , Chiara Maria Mazzanti <sup>2</sup> , Guntulu Ak 3,4 , Muzaffer Metinta¸s 3,4 , Cecilia Lando <sup>5</sup> , Rosa Angela Filiberti <sup>5</sup> , Marco Lucchi <sup>6</sup> , Alessandra Bonotti <sup>7</sup> , Rudy Foddis <sup>8</sup> , Alfonso Cristaudo <sup>8</sup> , Andrea Bottari <sup>1</sup> , Alessandro Apollo <sup>1</sup> , Marzia Del Re <sup>9</sup> , Romano Danesi <sup>9</sup> , Luciano Mutti <sup>10</sup> , Federica Gemignani 1,\* and Stefano Landi <sup>1</sup>**


**Simple Summary:** Patients with malignant pleural mesothelioma (MPM) often have to wait a long time before receiving a diagnosis. To contribute to the research on this neoplasm, we analyzed various samples of tumor biopsy and the relative liquid biopsies from both plasma and pleural fluid. We tested the possibility of obtaining information about the tumor in a quicker and less invasive way compared to the usual solid biopsy. We performed NGS on blood and tumor samples from patients and obtained a list of somatic mutations. With the digital droplet PCR technique, we tested the respective pleural fluids and plasma for the previously found mutations. We discovered that pleural fluid is a good proxy to obtain the mutational landscape of the MPM. We also tracked tumor DNA in plasma, leading to the idea that this could be used in a clinical setting to perform follow-ups of patients and monitor drug responses.

**Abstract:** Background: Malignant pleural mesothelioma (MPM) is a fatal tumor with a poor prognosis. The recent developments of liquid biopsies could provide novel diagnostic and prognostic tools in oncology. However, there is limited information about the feasibility of this technique for MPMs. Here, we investigate whether cancer-specific DNA sequences can be detected in pleural fluids and plasma of MPM patients as free circulating tumor DNA (ctDNA). Methods: We performed wholeexome sequencing on 14 tumor biopsies from 14 patients, and we analyzed 20 patient-specific somatic mutations with digital droplet PCR (ddPCR) in pleural fluids and plasma, using them as cancerspecific tumor biomarkers. Results: Most of the selected mutations could be detected in pleural fluids (94%) and, noteworthy, in plasma (83%) with the use of ddPCR. Pleural fluids showed similar levels of somatically mutated ctDNA (median = 12.75%, average = 16.3%, standard deviation = 12.3) as those detected in solid biopsies (median = 21.95%; average = 22.21%; standard deviation = 9.57),

**Citation:** Moretti, G.; Aretini, P.; Lessi, F.; Mazzanti, C.M.; Ak, G.; Metinta¸s, M.; Lando, C.; Filiberti, R.A.; Lucchi, M.; Bonotti, A.; et al. Liquid Biopsies from Pleural Effusions and Plasma from Patients with Malignant Pleural Mesothelioma: A Feasibility Study. *Cancers* **2021**, *13*, 2445. https:// doi.org/10.3390/cancers13102445

Academic Editors: Daniel L. Pouliquen and Joanna Kopecka

Received: 8 April 2021 Accepted: 11 May 2021 Published: 18 May 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

and their paired difference was weakly statistically significant (*p* = 0.048). On the other hand, the paired difference between solid biopsies and ctDNA from plasma (median = 0.29%, average = 0.89%, standard deviation = 1.40) was highly statistically significant (*p* = 2.5 × 10−<sup>7</sup> ), corresponding to the important drop of circulating somatically mutated DNA in the bloodstream. However, despite the tiny amount of ctDNA in plasma, varying from 5.57% down to 0.14%, the mutations were detectable at rates similar to those possible for other tumors. Conclusions: We found robust evidence that mutated DNA is spilled from MPMs, mostly into pleural fluids, proving the concept that liquid biopsies are feasible for MPM patients.

**Keywords:** malignant pleural mesothelioma; liquid biopsies; circulating tumor DNA; plasma; cancerspecific mutations; genomics; cancer biomarkers

#### **1. Introduction**

Malignant pleural mesothelioma (MPM) is a fatal cancer that arises from the mesothelial cells of the pleura. Asbestos exposure and the host's predisposing conditions (e.g., inherited mutations within *BAP1* or a chronic inflammatory state of the pleura [1]) play a role in the carcinogenesis of this neoplasm. Fibers are hypothesized to trigger a chronic inflammatory status, inducing a condition known as "frustrated apoptosis" of macrophages [2] and leading to increased production of oxygen reactive species, DNA damage, and cell proliferation, eventually initiating and promoting the malignant process [3,4]. The latency between exposure to asbestos fibers and diagnosis usually takes decades [5], and the first symptoms (which include, but are not limited to, chest pain, breathing difficulties, dyspnea, or increased abdominal volume) are common to other respiratory conditions [6], making a prompt diagnosis very difficult. Widely used imaging methods are not sufficient for the diagnosis of MPM. Thus, to achieve a reliable diagnosis, one needs to perform a biopsy through video-assisted thoracoscopy (VATS) [7,8], although this invasive procedure cannot be routinely used to assess the successive genetic changes.

Liquid biopsies (LBs) represent an innovative approach under development and consist of the analysis of genetic material extracted from body fluids. Events like apoptosis, necroptosis, and cell migration may result in the dispersion of tumor cells or their debris in the fluids surrounding the tumor mass [9]. Therefore, under these circumstances, it is possible to detect circulating tumor cells (CTCs), circulating cell-free tumor DNA (ctDNA), tumor proteins, and tumor-derived extracellular vesicles (tEVs, which include exosomes) in plasma, urine, or other body fluids [9]. Numerous studies have confirmed the possibility of gaining information on many kinds of tumors via blood samples. At first, CTCs were isolated and examined to get more insight into tumor progression and mutational history [10]. CTC phenotypic characterization and count can give hints on the tumor stage and expansion, whereas their DNA can provide information about the tumor mutational landscape [10]. Similarly, ctDNA could also be useful for LBs. In cancer patients, up to 1% of circulating nucleic acids are derived from tumor cells. The ability to isolate and analyze this DNA has made it possible to detect circulating mutations deriving from hepatocellular, breast, lung, and pancreatic carcinoma [11–14]. Evidence suggests the possibility of inferring or confirming the diagnosis of these tumors and performing clinical follow-ups by tracking the mutational load in response to therapies. In specific cases, such as lung adenocarcinoma, the monitoring of mutated ctDNA could provide important information to adjust a personalized therapy based on the use of anti-EGFR drugs [15,16]. On the other hand, other tumors (such as glioblastoma) are not equally capable of spilling ctDNA into the bloodstream [17], and the knowledge, in this regard, on MPM is limited. CtDNA from MPM patients has been analyzed in two previous studies. In 2012, higher DNA integrity was detected in cytologically negative pleural fluids (PFs) from 16 MPM patients (median = 1.2) compared to 23 noncancer patients (median = 0.8). The conclusion is that this biomarker, along with others (e.g., mesothelin), could improve the

specificity and sensitivity needed to discriminate MPM from non-MPM patients [18]. More recently, in 2018, 10 MPM patients were analyzed for ctDNA (half of them were treatmentnaïve). In this work, tumor biopsies were sequenced, and ctDNA was investigated in plasma samples via digital-droplets PCR (ddPCR). The authors showed that more than half of the treatment-naïve subjects showed positive droplets for mutated ctDNA in their blood samples, demonstrating the presence of tumor-specific mutations in circulating DNA [19]. However, the number of analyzed patients was limited, and not all of them showed mutated ctDNA from MPM in their bloodstream. Furthermore, no other fluids have been analyzed in the attempt to find an alternative approach to increasing the analysis' sensitivity. In order to fill the lack of knowledge on this topic, we analyzed a series of 14 MPM patients and carried out more systematic research on solid tumor biopsies, PF, and plasma withdrawn from the same patient. Thus, we could show that the share of somatically mutated cancer-specific DNA from PFs is similar to that detected in solid biopsies and that the same somatic mutations can also be detected, in tiny amounts, in the plasma of the same patient. Therefore, this feasibility study provides evidence that, in the future, PFs and plasma could constitute a valuable source of information, allowing for the diagnosis, follow-up, and stratification of MPM patients.

#### **2. Materials and Methods**

#### *2.1. Patients Cohorts*

We analyzed samples from patients diagnosed with MPM from three different hospital centers; we divided them into two groups based on the availability of blood samples.

Group GE consisted of 7 frozen tumor biopsies from San Martino Hospital in Genoa (Italy); each of them was associated with frozen samples of plasma and PF. Biopsies were about 1 mm<sup>3</sup> in size. PF, collected from patients' pleural effusions, and plasma samples were available in different amounts for each subject, ranging from 3 to 6 mL. Patients were diagnosed with MPM at an average age of 71 during the period 2002–2012. All of them were deceased at the moment of this analysis, having a median survival time since diagnosis of 4.6 months, with a minimum of 1 month and a maximum of 33.

Group PT (Pisa and Turkey) consisted of biopsies, frozen whole-blood, and frozen plasma samples from 2 patients (P) from the University Hospital of Cisanello in Pisa (Italy) and 5 patients (T) from Eski¸sehir Osmangazi University Hospital in Turkey in the period 2017–2019. Patients were diagnosed at a median age of 68. All of them were deceased at the time of the analysis, with a median survival time since diagnosis of 8.9 months and a minimum of 6.4 months and a maximum of 15.5 months. The size of the biopsies was about 1 mm<sup>3</sup> , and the volumes of whole blood and plasma were 2 and 1.5 mL, respectively. For patients T, a sample of PF was also available (2 mL). Patients' information is reported in Table S1.

#### *2.2. Sequencing and Filtering*

In order to discern somatic from germline mutations, whole-exome sequencing (WES) was carried out on solid biopsies and buffy coats withdrawn from the same patient of Group PT, while specific algorithms and filtering procedures were employed for the patients of Group GE. Genomic DNA was extracted using a PureLink™ Genomic DNA Mini Kit (Thermo Fisher Scientific; Waltham, MA, USA), following manufacturer protocol. This was used for both blood and tumor samples. The final DNA samples' concentration was measured with a Qbit3 (Thermo Fischer Scientific; Waltham, MA, USA). WES was performed on a NextSeq 550 (Illumina; San Diego, CA, USA) and the library was prepared using the kit from the same producer (Nextera DNA Flex Pre-Enrichment Library Prep and Enrichment). Sequencing indexes were also provided by the same manufacturer. Alignment of the resulting FASTA files was performed with Burrows-Wheeler Aligner software [20]. The calling of somatic mutations for the tumor samples was performed with VarScan [21], where paired blood was available; for the remaining cases, GATK tool Mutect2 [22] was used. The resulting single nucleotide variations (SNVs) were annotated

using the VEP online tool from the Ensembl portal (http://grch37.ensembl.org/Homo\_ sapiens/Tools/VEP/ accessed on 16 March 2019).

Since no blood was available for Group GE whereas it was available for the PT group, two alternative filtering procedures, FGE and FPT, were carried out.

FGE was carried out as follows. To maximize the chances of selecting a somatic mutation, we considered those with a ratio of alternative allele reads (i.e., alternative depth, AD) to total reads (i.e., total depth, TD) of 0.25 or lower. This is because such a ratio may originate from mutated tumor cells, whereas a higher ratio could indicate a homozygous or heterozygous mutation present in all the sample's cells, which is less likely somatic. Another parameter of FGE filtering excluded the mutations within noncoding regions. This was done to allow an easier interpretation of the functional consequence of the variation in the context of the carcinogenic progression. The last filter condition required mutations to have an AD greater than 20X and a minor allele frequency (MAF) in the population ≤10−<sup>4</sup> , according to gnomAD (https://gnomad.broadinstitute.org accessed on 16 March 2019). The former parameter ensures a good NGS quality, while the latter allows us to take into account the negative selection a mutation undergoes in the population, decreasing the possibility of it being germline.

FPT consisted, firstly, in the use of VarScan2, a software based on the statistical analysis of a coverage value for both reference and alternative bases, comparing that found in blood with that of the paired tumor. Then, further filtering was applied using the following criteria: (i) a minor allele frequency (MAF) <1% among Europeans, according to gnomAD, (ii) a read depth ≥20X, and (iii) AD = 0 in the blood sample.

From the final list of SNVs obtained with the filtering procedures, up to 5 mutations per patient were selected for further experimental validation. This last choice did not follow a strict criterion but was based on a variety of criteria that included (i) mutations present on a gene already filed for MPM within COSMIC or TCGA databases (https: //cancer.sanger.ac.uk/cosmic, https://www.cancer.gov/tcga accessed on 16 March 2019), (ii) the lack of any repeated sequence in the neighboring region of the SNV, and (iii) the lack of paralogues/gene families of the mutated genes.

#### *2.3. Validation and Biostatistical Analysis*

SNVs selected following WES were verified in tumor biopsies with an allele-specific oligonucleotide and a real-time quantitative PCR method (ASO–qPCR). For each SNV, the real-time curve obtained with mutation-specific primers was compared with the curve obtained with specific primers designed for the wild-type allele. The results were also compared to the same assay performed on DNA extracted from the whole blood of a healthy subject (reference). This analysis is not quantitative enough to measure the amount of mutated DNA. On the other hand, it is inexpensive and sensitive enough to verify the presence/absence of small amounts of mutated alleles among a plethora of wild-type alleles. Experiments were performed with a CFX96 thermal cycler (Bio-Rad; Hercules, CA, USA) using 5× HOT FIREPol® EvaGreen® qPCR Mix Plus (Solis Biodyne; Tartu, EE, Estonia) and custom oligonucleotides primers (Europhins Genomics, Louxemburg, LU). When ASO–qPCR confirmed the mutation in tumor DNA, we proceeded by using the more sensitive ddPCR for quantification. Thus, for each patient, ddPCR was applied on tumor samples as well as other available fluids.

Circulating DNA was extracted using a QIAamp Circulating Nucleic Acid Kit (Qiagen; Venlo, NL, Netherlands), and the concentrations were measured with a Qbit (Thermo Fisher Scientific; Waltham, MA, USA). DdPCR was used to measure and compare the amount of mutated DNA (using mutation-specific probes) in tumor, blood, PF, and plasma. DNA from a healthy individual and a blank (buffer only) were used as negative controls. We used a QX100 droplet generator to form the reaction droplets and a QX200 droplet reader (Bio-Rad; Hercules, CA, USA) to get the results. The PCR amplification reaction was performed in a T100 thermal cycler from the same manufacturer. For each SNV, we used a pair of TaqMan-like probes, each targeted either to the variant or the common allele, the former being labeled with FAM and the latter with HEX fluorophore. Probes and primers were designed using Bio-Rad's online probes design tool. The reaction mix used was Bio-Rad's ddPCR Supermix for Probes (No dUTP). Differences in the amount of mutated ctDNA from plasma or PF compared to that measured in solid biopsies (as reference) were evaluated with a paired Student's *t*-test analysis, following arcsin transformation for non-normally distributed data, and the nonparametric Wilcoxon signed-rank test.

#### **3. Results**

#### *3.1. GE patients, NGS Analysis*

For Group GE, NGS analysis yielded an average of 78.67 million reads per tumor, with a mean length of 122 bases. Across all samples, 87.1% of the reads were correctly aligned with the reference, with a mean mapping quality of 59.5 and an average coverage inside the exome regions of 97.7X. After the analysis with the Mutect2 tool, which computed all mismatches in the reads to find mutations, we obtained 97,826 to 123,405 variants, depending on the sample (Table 1). Of those variants, 6.9–10.8% were indels (insertion/deletions) and 89.2–93.1% were SNVs, with an average median coverage of 102X. The values of this parameter fitted a Laplace distribution with a mean of 41.5X.

Then, FGE was optimized to maximize the likelihood of selecting truly somatic mutations. Firstly, variants with the number of AD reads (the ones covering the alternative allele) lower than 25% (arbitrarily chosen), relative to TD (total depth), were positively selected. Between 1065 and 3053 variants passed this step, depending on the sample (Table 1). The second step of selection included only the SNVs within the coding regions (between 593 and 1342). A further step of the FGE procedure was carried out by excluding the variants with less than 20 reads covering the genomic position. Then, among the selected SNVs, those with a MAF ≥ 10−<sup>4</sup> (global, according to gnomAD, arbitrary threshold) were excluded as well. A number of SNVs, between 42 and 184, remained in the final list.

#### *3.2. In GE Patients, Selected Somatic Mutations Were Detected in ctDNA from Plasma and PFs*

A total of 25 mutations within 25 genes in 7 subjects were evaluated with ASO–qPCR in tumor biopsies, and 14 were confirmed by this method. The list encompassed *COL1A2*, *BACE2*, *MYBPC1*, *TRPC7*, *ARPP21*, *OR4K2*, *HIST1H2AD*, *OR5AC2*, *SZT2*, *AMPH*, *SPTAN1*, *NLGN1*, *DICER1*, and *FLI1*; most of them had already been reported as somatically mutated in the MPM patients, according to COSMIC or TCGA databases (Table 1). Four SNVs within *BAP1, LATS2, MUC16,* and *FLG*, the genes most frequently mutated in MPM, together with other 7 mutations in *POTEF*, *RAD50*, *FGFR1*, *UNC79*, *ERBB4*, *CSMD3*, and *CCNL2*, could not be confirmed by ASO–qPCR and were not investigated further with ddPCR.


**Table 1.** Selected mutations analyzed for Group GE after FGE filtering and relative ASO–qPCR and ddPCR results.

 MAF = minor allele frequency; TD = total depth; AD = alternative depth, MA = mutated allele; PF = pleural fluid; S = synonymous; M = missense; FS = frame shift; IF = in frame; SG = stop gain. a According to gnomAD (https://gnomad.broadinstitute.org, accessed on 16 March 2019), global frequency. b This SNV does not have a frequency in gnomAD (https://gnomad.broadinstitute.org accessed on 16 March 2019). c There is a nearby SNP, rs140004238 (G > A), with a global frequency of 3.98 × <sup>10</sup>−6, at 7:38516516 (+1bp). d There is a SNP, rs1349519137 (G > C), with a global frequency of 3.19 × <sup>10</sup>−5, at the same genomic position. e There is a nearby SNP, rs1377012777 (A > G), with a global frequency of 3.98×<sup>10</sup>−6, at 8:113988191 (+2 bp).

The measurements carried out with ddPCR on the 14 confirmed mutations showed that 3 (*COL1A2*-rs773494330, ID = 696; *SZT2*-rs760370909, ID = 2324; *DICER1*-rs775912475, ID = 2829) could not be detected in tumor biopsies with ddPCR, whereas positive results were obtained for 11 mutations found in the biopsies of 7 patients (3 patients were positive for 1 mutation and 4 patients were positive for 2 mutations), as reported in Table 1.

When PFs were analyzed, samples from 6 patients were available. One, ID = 696, could not be analyzed because the amplification failed several times, even after adopting alternative protocols for DNA extraction, suggesting the presence of unknown PCR inhibitors. Thus, only 10 mutations could be compared between tumor biopsies and PFs. Interestingly, for 7 of them, the share of mutated alleles measured in PFs was similar (ranging from 10.24% to 20.20%) to that measured in the respective tumor biopsies. The remaining three mutations showed a reduced amount; however, they were still detectable to a significant extent: *TRPC7*-rs566980923 (ID = 1148), 12.5% in tumor and 4.9% in PF; *AMPH*-COSM1673120 (ID = 2324), 15.1% (tumor) and 4.01% (PF); *NLGN1*-COSM479730 (ID = 2438), 21.95% and 1.59%, respectively (Table 1).

Interestingly, 9 out of 11 mutations of tumor biopsies were also detected in the ctDNA from plasma. Two, *TRPC7*-rs566980923 ID = 1148 and *NLGN1*-COSM479730 ID = 2438, were undetected, and this was in agreement with the low quantity already detected in the respective PF samples. Of the 9 detectable mutations, 2 (*MYBPC1*-rs752347381 ID = 1148 and *OR4K2*-rs757533510 ID = 2294) were likely germline. In fact, for these mutations, the percentage of the alternative allele in PF and plasma was about 25% and of a similar range to that measured in the tumor biopsies. However, as reported in Table 1, the remaining seven mutations were most likely somatic and showed a percentage ranging between 0.16% and 0.79%, whereas their corresponding share within the tumor biopsies ranged between 11.1% and 23.05%. The one showing the highest amount was *OR5AC2*-rs1021819573 (ID = 2294) with a percentage of 5.57 (it was 11.1 in the tumor). Thus, 6 out of 7 patients showed ctDNA in their plasma. Only patient ID = 1148 could not be traced using the selected mutations. Unfortunately, the analysis of the other patient's mutations could not be carried out because of the lack of additional vials of plasma.

#### *3.3. PT Patients, NGS Analysis*

NGS analysis on Group PT yielded an average of 80.66 million reads for each subject's tumor sample. The average read length was 98 bases. Across all samples, 70% of the reads were correctly aligned in the exome reference region, with a mapping quality of 54.9 and an average coverage of 72X. After the analysis with the software VarScan2 for each tumor–blood pair, we identified between 102,753 and 130,073 SNVs, of which 1948–3120 were marked as somatic (Table 2). The TD values had a median of 45 and fitted a Laplace distribution with a mean of 50X. The indels were not evaluated in our assays; however, they consisted of a share ranging from 8.89% to 14% of the total variations. FPT consisted of selecting mutations within coding regions, eliciting from 158 to 281 SNVs. Then, SNVs with TD < 20 and a population MAF > 1% (global according to gnomAD) were excluded. The resulting 54–104 SNVs were further filtered by including only mutations with AD = 0 in blood samples, yielding 2 to 42 mutations. Finally, among the available variants filtered for both groups, we selected two mutations per sample for further analyses, as specified in "Materials and Methods" (Tables 1 and 2).


#### **Table 2.**Selected mutations analyzed for Group PT after FPT filtering and relative ASO–qPCR and ddPCR results.

 MAF = minor allele frequency; TD = total depth; AD = alternative depth; PF = pleural fluid; S = synonymous; M = missense; SG = stop gain. ¥ Predicted by VarScan2 tool (DOI:10.1101/gr.129684.111); a according to gnomAD (https://gnomad.broadinstitute.orgaccessed on 16 March 2019), global frequency. b There is a SNP, rs770127999 (C > A), with a global frequency of 4.00×<sup>10</sup>−6, at the same genomic position.

#### *3.4. Selected Mutations for Group PT were also Detected in the ctDNA from Plasma and PF*

Fourteen mutations were analyzed with ASO–qPCR in Group PT, and twelve (two for each T patient and one for each P patient) could be confirmed in the tumor biopsies (*OR10A4*-rs547489107 and *NLRP6*-NM\_138329.2:c.403G > T were undetected), as reported in Table 2. Thus, we used ddPCR to measure the amount of mutated DNA within the tumor biopsies, and only one mutation (*SS18*-NM\_001007559.3:c.98A > G, of subject 01T) could not be detected. Of the remaining 11 mutations in 10 genes (*BAP1* occurred twice), we found that *BAP1*, *NF2*, *FAT1*, *JADE1*, and *FLT1* were already present in the COSMIC and TCGA databases for MPM patients. For eight variants, the percentage of mutated DNA analyzed was of a similar extent to that yielded by NGS, considering an expected 10% error rate. On the other hand, *JADE1*- rs775483821 (ID = 01T) had 7.48% of mutated DNA in ddPCR opposed to 24.53% of the NGS, whereas *BAP1*-COSM4411449(C > T) (ID = 02T) had 33.50% vs. 21.88% and *NF2*- NM\_016418.5:c.985A > T (ID = 03P) showed 15.85% vs. 27.03%, respectively. In PF, among the 11 mutations detected in tumor biopsies, 2 (*VIL1*- NM\_007127.3:c.2070C > T and *NF2*-NM\_016418.5:c.985A > T) could not be analyzed due to the lack of biological specimens of subjects 02P and 03P, whereas 1 (*FAT1* rs776531396; ID = 05T) was undetectable. The remaining 8 SNVs were detected with percentages compatible with those observed in tumor biopsies, ranging from 12.75% to 39.70%. The only exception was patient 04T, whose mutations (*FAM71B*- rs1404037352 and *CSMD2* rs770364421) had lower mutated allele relative abundance in PF compared with the tumor sample, namely, 9.7% against 30.80% and 8.95% against 25%, respectively.

The 11 mutations were also investigated in plasma samples. For patient 04T, we could not assay two mutations because of the lack of biological specimens. Of the remaining nine mutations, eight were also detectable in the patients' plasma, whereas *FAT1*- rs776531396 (ID = 05T) was undetectable, in agreement with the lack of detection in his PF. In plasma, the eight mutations could be detected, with percentages ranging from 0.14% to 2.68%. All these results are summarized in Table 2.

Considering both groups of patients and excluding the two mutations highly suspected to be of germline origin and the one not detected in solid biopsy, the percentages of mutated DNA detected in solid biopsies were higher than those detected in ctDNA from PFs: median = 21.95 vs. 12.75 (respectively); average ± st.dev = 22.21 ± 9.57 vs. 16.3 ± 12.3. This difference was statistically significant (*p* = 0.0237) when analyzed with Student's *t*test for paired data and not statistically significant when analyzed with nonparametric Wilcoxon's test (*p* = 0.0648). The difference was much greater when compared to ctDNA from plasma (median = 0.29, average 0.89 ± 1.40), providing a high statistical significance to the same statistical tests (*p* = 2.49 × 10−<sup>7</sup> and *p* = 3.2 × 10−<sup>4</sup> , respectively).

#### **4. Discussion**

In this study, we report a positive feasibility study that in MPM patients, ctDNA is present in PF at high concentrations and cancer-specific DNA can be detected in plasma, although at low percentages. Therefore, we provide evidence that LBs for patients with MPM is feasible, and this could represent a potentially important tool for the diagnosis, therapy, follow-up, and stratification of patients, especially with pleural effusions.

It is noteworthy that we ruled out the possibility of selected germline mutations, either by using stringent filtering procedures or by carrying out WES of the buffy coat, when available. Thus, the present data reinforce and extend the preliminary evidence reported by Hylebos et al. [19], where only 3 out of 10 patients presented ctDNA in plasma samples. In that study, only one mutation was assayed, and no PFs were available. In our study, we started from a selection of 39 total mutations, and 22 could be confirmed in solid biopsies, allowing further investigations in PFs and plasma. Since we considered these SNVs enough for our purposes, we did not pay further attention to the remaining 17 mutations. Likely, they could not be validated because of poor ASO–qPCR probe performance.

In two GE patients, two mutations showed high and similar percentages in tumor, PF, and plasma, strongly suggesting a germline origin. This result was not surprising because, despite the stringent filtering procedure we applied, GE patients' buffy coat was lacking, and WES could not be carried out. However, the remaining 20 were enough for investigating whether MPM patients could carry ctDNA in PF or plasma. With the exception of subjects 696 (PCR could not work for an inhibitor), 02P, and 03P (PFs not available), 16 out of 17 mutations could be detected in PF. The percentages of the mutated allele detected differed by about 7.5%, on average, from those found in the tumor biopsies, a value falling within the intrinsic error of NGS technology (Figure 1). This fact indicates that DNA extracted from PF is a good proxy for its counterpart obtained from the solid tumor. In the future, DNA from PF could be employed instead of the classical solid biopsies to gain insights on the cancer's mutational landscape with much less distress for the patient. Moreover, 15 out of 18 analyzable mutations were also detectable in plasma, with relative abundances varying from 0.14% to 5.57%. Since only a few milliliters of plasma were available from the biobank, we could not analyze a high number of DNA copies in plasma. It is conceivable that the analysis of higher amounts of DNA could have elicited positive results in the three negative cases as well. PF, and plasma, strongly suggesting a germline origin. This result was not surprising because, despite the stringent filtering procedure we applied, GE patients' buffy coat was lacking, and WES could not be carried out. However, the remaining 20 were enough for investigating whether MPM patients could carry ctDNA in PF or plasma. With the exception of subjects 696 (PCR could not work for an inhibitor), 02P, and 03P (PFs not available), 16 out of 17 mutations could be detected in PF. The percentages of the mutated allele detected differed by about 7.5%, on average, from those found in the tumor biopsies, a value falling within the intrinsic error of NGS technology (Figure 1). This fact indicates that DNA extracted from PF is a good proxy for its counterpart obtained from the solid tumor. In the future, DNA from PF could be employed instead of the classical solid biopsies to gain insights on the cancer's mutational landscape with much less distress for the patient. Moreover, 15 out of 18 analyzable mutations were also detectable in plasma, with relative abundances varying from 0.14% to 5.57%. Since only a few milliliters of plasma were available from the biobank, we could not analyze a high number of DNA copies in plasma. It is conceivable that the analysis of higher amounts of DNA could have elicited positive results in the three negative cases as well.

In two GE patients, two mutations showed high and similar percentages in tumor,

*Cancers* **2021**, *13*, 11 of 14

**Figure 1.** Mean levels of mutated DNA found in the samples from three sources: solid tumor, pleural fluid, and plasma. **Figure 1.** Mean levels of mutated DNA found in the samples from three sources: solid tumor, pleural fluid, and plasma.

> The fact that MPM is a locally spreading tumor on the pleural surface could provide a good explanation of the high amount of ctDNA detected in PFs and the low amount detected in plasma. We can hypothesize that the observed interindividual variability of ctDNA levels could be ascribed to the relative amounts of subclones tracked with the picked mutation, the aggressiveness of the subclone carrying the picked mutation, or to The fact that MPM is a locally spreading tumor on the pleural surface could provide a good explanation of the high amount of ctDNA detected in PFs and the low amount detected in plasma. We can hypothesize that the observed interindividual variability of ctDNA levels could be ascribed to the relative amounts of subclones tracked with the picked mutation, the aggressiveness of the subclone carrying the picked mutation, or to the mechanisms involved in the release of tumor DNA.

> the mechanisms involved in the release of tumor DNA. We foresee that the use of PF or plasma could be very important in the diagnosis process and for a noninvasive clinical follow-up of the patients. An earlier diagnosis could be carried out by integrating the results of ctDNA analysis with currently available biomarkers, such as serum soluble mesothelin levels, and other epigenetic biomarkers under research, such as the expression of the circulating microRNAs miR-16, miR-17, miR-126, miR-486 or CpG methylation at *CDKN2A* or *SFRP* genes [23]. In fact, once the tumor is characterized for its genetic background, specific mutations could be used to monitor the evolution of the disease, allowing early detection of its worsening before any clinical observation. The analysis of cancer-specific mutations through the use of LBs could also allow more accurate monitoring of responses to therapies. With our work, we enlighten the versatility of this method to obtain genetic information on MPM using PF and plasma We foresee that the use of PF or plasma could be very important in the diagnosis process and for a noninvasive clinical follow-up of the patients. An earlier diagnosis could be carried out by integrating the results of ctDNA analysis with currently available biomarkers, such as serum soluble mesothelin levels, and other epigenetic biomarkers under research, such as the expression of the circulating microRNAs miR-16, miR-17, miR-126, miR-486 or CpG methylation at *CDKN2A* or *SFRP* genes [23]. In fact, once the tumor is characterized for its genetic background, specific mutations could be used to monitor the evolution of the disease, allowing early detection of its worsening before any clinical observation. The analysis of cancer-specific mutations through the use of LBs could also allow more accurate monitoring of responses to therapies. With our work, we enlighten the versatility of this method to obtain genetic information on MPM using PF and plasma as starting materials.

> as starting materials. One limitation to the present study consisted of the limited clinical information available from the biobanks of the samples G and P. It could be hypothesized that the percentage of mutated copies would be higher in patients presenting the tumors at advanced stages,

conceivably with the idea of a higher extent of ctDNA released from largely spread tumors or metastases. At the present time, it is not possible to know whether the mutated DNA could also be detected in LBs from MPM patients with earlier stages of the disease. Future research should be encouraged to approach this task. However, we analyzed whether the amount of mutated DNA could correlate with patients' overall survival (a proxy of the tumor staging), and we could not find any statistically significant correlation. We could hypothesize that this is due to the low statistical power for this type of analysis or to the fact that all MPM patients are diagnosed at late stages. Given the possibility of gathering more clinical and histological data about the tumor, such as cell type, tumor stage, and treatment response, our results may prove even more useful in the clinical field.

#### **5. Conclusions**

In summary, this study showed that LBs are feasible in MPM, paving the way for novel tools in the clinical management of these patients. It has been figured out that once the profile of MPM's somatic mutations is fully achieved, the choice of the therapy, its effectiveness, and/or the occurrence of relapses can also be monitored by using PF and plasma as a source of ctDNA.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/article/10 .3390/cancers13102445/s1, Table S1: patients clinical data.

**Author Contributions:** G.M.: conceptualization, data curation, formal analysis, investigation, methodology, software, validation, writing. C.M.M.: methodology, supervision. P.A.: data curation, methodology, software. F.L.: investigation, methodology. G.A.: data curation, sample collection. M.M.: data curation, sample collection, supervision, writing and editing. C.L.: data curation, sample collection. R.A.F.: data curation, sample collection. M.L.: data curation, sample collection. A.B. (Alessandra Bonotti): data curation, sample collection. R.F.: data curation, sample collection. A.C.: data curation, sample collection. L.M.: writing and editing. A.B. (Andrea Bottari): data curation, formal analysis, investigation, methodology, software, validation. A.A.: data curation, formal analysis, investigation, methodology, software, validation. M.D.R.: validation. R.D.: validation, supervision. F.G.: investigation, project administration, supervision, writing and editing. S.L.: conceptualization, data curation, formal analysis, funding acquisition, investigation, project administration, supervision, visualization, writing and editing. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by Fondazione Pisa grant number 153/16.

**Institutional Review Board Statement:** The study was conducted according to the guidelines of the Declaration of Helsinki and approved by the Institutional Review Boards (or Ethics Committee) of Pi Ge Tu Eskisehir Osmangazy University (protocol code 24.11.2020/22) and Azienda Ospedaliero-Universitaria Pisana (dated 7 February 2013, protocol number 192/53).

**Informed Consent Statement:** Informed consent was obtained from all subjects involved in the study.

**Data Availability Statement:** Data is contained within the article or supplementary materia.

**Conflicts of Interest:** The authors declare there is no conflict of interest. The founding entity had no role in the writing of the manuscript and the performing of the study; in the collection, analyses, or interpretation of data; and in the decision to publish the results.

#### **Abbreviations**


#### **References**


*Article*
