Recent insights emerging from malignant mesothelioma genome sequencing

In this issue of JTO, Lo Iacono et al. present a retrospective study performed on 123 formalin-fixed, paraffin-embedded (FFPE) biopsies from 123 malignant mesothelioma (MM) patients, using next generation sequencing (NGS) with a commercially available library of genes that are frequently mutated in cancers. Biopsies were sequenced with the Ion Torrent platform and positive results were validated by Sanger sequencing¹. Two limitations of this approach, acknowledged by the Authors, are: 1) Since BAP1 and NF2 were not represented in the commercial library, their mutations status was analyzed using a different custom-made library; 2) the particular targeted NGS methodology chosen in this paper specifically detects small gene mutations, while this methodology cannot detect large genetic losses, gene amplifications and translocations, or epigenetic changes.

In a parallel study recently published in Cancer Research², Guo et al. performed whole exome sequencing on DNA and matched blood samples of 22 MM frozen biopsies from 22 patients collected in the operating room, integrated with the analysis of somatic copy number alterations (SCNAs). Exome capture libraries were sequenced on the Illumina HiSeq platform². In both studies the biopsies were from patients previously treated with chemotherapy (100% of patients Lo Iacono et al.; 41% Guo et al.). Therefore, the well-documented mutagenic effects of chemotherapy on the cancer genome³ may have contributed to some of the mutations detected. These manuscripts^{1, 2} present the first landscape view of the somatic genomic alterations in MM. Because the researchers^{1, 2} used different experimental approaches, an exact comparison of the results is not possible. However, concordant data developed by independent research teams based in Europe and in the US, using different experimental strategies, provide reassurance on the reliability of these results. In fact, these two studies, which used robust statistical algorithms for the mutation analysis, show minor discrepancies and several corroborating findings.

In both studies the number of non-synonymous gene mutations was smaller than in other cancers. Lo Iacono et al. found that 20/52 “cancer” genes studied harbored variations in 25/123 (20%) of FFPE biopsies (including intronic, synonymous, non synonymous and regulative mutations). These mutations were clustered in the two main “p53/DNA repair” (TP53, SMACB1, and BAP1) and “PI3K-AKT” (PDGFRA, KIT, KDR, HRAS, PIK3CA, STK11 and NF2) pathways.

Guo et al. instead sequenced “only” the exome, but looked for all possible genes, and found 490 mutated genes, of which 447 (97%) were mutated only in one biopsy, and found an average of 23 mutations per biopsy (range 2–51).

Several factors may explain the different mutation frequency detected: the different source of DNA (FFPE¹ versus frozen biopsies²), different platforms, study design and algorithms used, and the fact that Lo Iacono et al.¹ sequenced also the intronic regions (except for BAP1 and NF2). The genes most commonly mutated in both studies were BAP1, NF2, and CDKN2A.

Both studies detected frequent somatic non-synonymous BAP1 mutations in 41%² and 58%¹ of MMs respectively, confirming previous findings that detected somatic BAP1 mutations in 22–61% of MMs^4–6. These mutations result in stop codons that produce truncated BAP1 proteins lacking the nuclear localization sequence (NLS) or are mutations within the catalytic subunit that impair BAP1 auto-deubiquitination, which is required for nuclear localization⁷. Therefore, these mutations are predicted to result in BAP1 proteins that cannot migrate to the nucleus and which may have aberrant de-ubiquitinase activity in the cytoplasm⁸. Accordingly, Lo Iacono et al. found that 52% of 116 MM biopsies stained for nuclear BAP1 – an indication of normal BAP1 activity – while 48% did not, an indication of mutated BAP1¹. Nuclear BAP1 staining correlated with presence/absence of DNA mutations (p=0.001). Both studies^{1, 2} report frequent mutations of NF2, encoding Merlin, a component of the Hippo signaling pathway⁹, with about 50% frequency, a value comparable with previous reports^{10, 11}. Surprisingly, 92% of the specimens analyzed by Lo Iacono et al., showed NF2 expression by immunohistochemistry¹. The Authors propose that the genetic variations detected might deregulate NF2 without affecting protein expression and stability. Alternatively, the NF2 genetic mutations detected are often of minor biological significance.

Lo Iacono et al. emphasize the possible role of mutations found in the PIK3CA gene encoding the catalytic subunit of PI3K in favoring tumor progression. However, the Authors acknowledge the need to study larger cohorts before conclusions can be drawn.

Guo et al. performed an integrative pathway analysis of somatic mutations and focal SCNAs: most recurrent alterations were in the MAPK and Wnt signaling pathways, and in the cell cycle, with recurrent alterations of CUL1, CDKN2A and TP53, the last two matching the results obtained by Lo Iacono et al..

In summary, in spite of the different study design and methodology, these two NGS analyses of the MM genome reveal that inactivating mutations occur randomly and are rarely shared among MM biopsies, with the exception of BAP1^{1, 2} and to a lesser extent NF2, CDKN2A^{1, 2} and possibly CUL1².

These results are in agreement with a large body of research that led to the conclusion that driver mutations (i.e., gene mutations present in all tumor cells in most cancers of the same type) are rare; a finding that significantly complicates the attempt to develop target therapies. In fact, the main goal of NGS tumor studies is to identify somatic driver mutations that would become potential therapeutic targets and/or clinical biomarkers. Instead, the emerging picture indicates that each single tumor has its own specific sets of genetic alterations. In addition to inter-tumor genetic heterogeneity, there is also significant intra-tumor genetic heterogeneity as branched evolutionary growth generates genetic diversity in several tumor sub-clones. Thus, a single biopsy taken at a given time point is unlikely to be representative of the full spectrum of tumor genetic alterations. For example, over 60% of mutations detected by NGS in clear cell carcinomas of the kidney were not present in all tumor areas sampled¹². In the same study, mTOR mutations were found in seven of eight primary sites, but in none of three metastases¹². The studies of Lo Iacono et al. and of Guo et al. did not explore the issue of intra-tumor genetic heterogeneity. Based on studies in other tumor types¹², and the recent findings that MMs are heterogeneous from start because they originate as polyclonal malignancies¹³, it is easy to predict that MMs will also show marked intra-tumor heterogeneity, further complicating attempts to develop molecular therapies that may benefit a large number of patients.

However, and in spite of these considerations, the studies of Lo Iacono et al. and Guo et al. have succeeded in identifying recurrent genetic alterations in MMs, that may be “actionable”, BAP1 being the most common. BAP1 is a nuclear ubiquitin carboxy-terminal hydrolase (UCH), associated with multiprotein complexes regulating key cellular pathways, including the cell cycle, cellular differentiation, cell death, gluconeogenesis and DNA repair⁸. When mutated in the germline, carriers develop MMs, uveal and cutaneous melanomas, renal and cholangiorcarcinomas, other malignancies, and often several cancers in combination^{5, 8}. Moreover, several studies point at BAP1 as harboring the putative driver mutations for sporadic (non-genetically related) MMs. We⁵ and Bott et al.⁴, initially reported — in independent and parallel studies using Sanger sequencing from MM biopsies — that 22% and 23% of MM biopsies contained somatic BAP1 mutations. Yoshikawa et al.⁶ reported that 61% of cell cultures derived from MM biopsies contained BAP1 mutations. Arzt et al. found that 60% of MM did not stain for nuclear BAP1, suggestive of inactivating mutations¹⁴. To address the discrepancy in the frequency of BAP1 mutations detected in different studies, we used an integrated genomic approach to study frozen MM biopsies which included Sanger sequencing, Multiplex Ligation-Dependent Probe Amplification (MLPA), cDNA sequencing, copy number analyses, methylation studies of the BAP1 promoter and immunohistochemistry. We found that 14/22 (63.6%) MM biopsies contained BAP1 mutations/inactivation. None of these methodologies alone was able to capture all inactivating BAP1 mutations. Thus, studies using methodology based exclusively on one type of molecular approach, such as Sanger sequencing^{4, 5} or NGS^{1, 2}, will underestimate the percent of MMs carrying BAP1 mutations. However, BAP1 nuclear staining was detected only in the 8 specimens demonstrated to contain wild-type BAP1 by the integrated genomic approach described above. Therefore, immunohistochemistry appeared to be capable of capturing the whole array of possible mechanisms of BAP1 inactivation¹⁵.

Several findings underscore the apparent “driver” role of BAP1 in MMs and point at BAP1 as a potentially useful target: 1) Multiple studies, including these two recent NGS^{1, 2} studies, found that BAP1 is frequently mutated in MM; 2) The persistence of BAP1 mutations in early pre-malignant lesions as well as in MM biopsies established in cell culture, suggesting that BAP1 mutations are an early event and that MM cells do not select against BAP1 mutations⁸; 3) The finding that, with rare exceptions, BAP1 nuclear staining is either detected in 100% of MM cells, or it is not detected at all¹⁵. Moreover, frequent somatic mutations in BAP1 are present in several malignancies⁸, making therapies to restore BAP1 activity in tumors relevant to many cancer patients.