Gene fusions during the early evolution of mesothelioma correlate with impaired DNA repair and Hippo pathways

Malignant pleural mesothelioma (MPM), a rare cancer a long latency period (up to 40 years) between asbestos exposure and disease presentation. The mechanisms coupling asbestos to recurrent somatic alterations are poorly defined. Gene fusions arising through genomic instability may create novel drivers during early MPM evolution. We explored the gene fusions that occurred early in the evolutionary history of the tumor. We conducted multiregional whole exome sequencing (WES) of 106 samples from 20 patients undergoing pleurectomy decortication and identified 24 clonal nonrecurrent gene fusions, three of which were novel (FMO9P‐OR2W5, GBA3, and SP9). The number of early gene fusion events detected varied from zero to eight per tumor, and presence of gene fusions was associated with clonal losses involving the Hippo pathway genes and homologous recombination DNA repair genes. Fusions involved known tumor suppressors BAP1, MTAP, and LRP1B, and a clonal oncogenic fusion involving CACNA1D‐ERC2, PARD3B‐NT5DC2, and STAB2‐NT5DC2 fusions were also identified as clonal fusions. Gene fusions events occur early during MPM evolution. Individual fusions are rare as no recurrent truncal fusions event were found. This suggests the importance of early disruption of these pathways in generating genomic rearrangements resulting in potentially oncogenic gene fusions.

5][6] Genomic analysis of somatic copy number alterations (SCNAs) and targeted sequencing in MPM initially identified BAP1, CDKN2A/B, and NF2 as tumor suppressor genes frequently lost in MPM. 7Subsequent whole exome and transcriptome analyses has increased the number of genes found to be recurrently mutated in MPM, with the implication that these genes are important driver loci where mutation drives the development of the tumor. 8,9ven the extensive heterogeneity within tumors, distinguishing individually rare driver loci from nontumorogenic passenger loci can remain a challenge.One approach is based on the concept that a genomic alteration is more likely to be a driver event if it occurs early in tumor development.A multiregional approach, where tumor genomes are sampled from different physical regions of the same tumor, allows identification of genomic alterations occurring clonally, that is across all samples, and therefore likely to have occurred early in tumor development before differentiation in space and in genomic alteration content.In our MEDUSA multiregional exome study, we have determined the relative timing of somatic mutations and SCNAs in a series of tumors, and shown key early mutational events are common in MPM, including loss of chromosome 22q (which carries NF2). 10 Furthermore, we have shown that that SCNA burden correlates with a high neutrophil: lymphocyte and platelet: lymphocyte ratio, both indicators of a systemic inflammatory response. 11ne fusions are an established mechanism of how structural variation in genomes can lead to a change in function, and many gene fusions are known to have an important role in cancer initiation and progression. 12The well-known recurrent driver fusion BCR-ABL1 in chronic myeloid leukemia (CML) has shown to have a substantial impact on clinical and treatment outcome. 13,14The chimeric fusion codes for a protein that drives CML progression via aberrant tyrosine kinase activity of ABL1 which signals and activates downstream oncogenic pathways. 15R-ABL1 was the first chimeric oncogenic protein that was targeted with a tyrosine kinase inhibitor. 16Other examples include IGH-MYC predominantly seen in Burkitt's lymphoma 17 and TMPRSS2-ERG and TMPRSS2-ETV4 fusions in prostate cancers. 18As well as providing information on the molecular mechanism of disease, and being potential drug targets, gene fusions can potentially encode tumor-specific antigens with the potential to be targets for cancer immunotherapy. 19though gene fusions are associated with activating oncogenes, most rearrangements reported so far in MPMs involve tumor suppressor genes and noncoding DNA. 8,20An analysis of the transcriptome of 216 MPMs along with matched normal samples using RNASeq found 43 recurrent gene fusions in 22 samples, confirmed using reverse transcription PCR. 8These fusions mostly involved the tumor suppressor genes NF2, BAP1, PTEN, PBRM1, and SETD2, and the function of the encoded proteins was likely to be disrupted.This study highlighted an alternative route for detecting tumor suppressor loss, for example the inactivation of NF2 as a result of an inversion event generating GSTT1-NF2 fusion.This alteration does not alter the gene dosage and would have been undetected without gene fusion analysis.Fusions involving EWSR1 have been found in a subset of perotineal mesotheliomas, including EWSR1-ATF1 21 and EWSR1-YY1 22 but not in MPM.Gene fusion transcripts have been identified in large scale scans of RNASeq data from MPM tumors, 23,24 but the lack of timing information and, in most cases, the lack of matching nontumor data, means the interpretation of these observations is unclear.
Identifying a gene fusion resulting from a SCNA in a tumor genome or fusion transcript from transcriptome data is not sufficient evidence to show that this is a driver alteration, indeed it is likely that most gene fusions are stochastic events and are passenger mutations 25 especially in tumors with a high burden of somatic structural variants.Here, we use multiregional sampling of individual tumors, combined with matched nontumor samples, to identify gene fusions occurring early in the evolution of MPM.By using matched nontumor exome data, multiple exomes of the same tumor, transcriptome data and PCR validation, we aimed to robustly identify early gene fusions more likely to be driver events that help drive MPM progression.

| Ethical and governance approvals
The MEDUSA study was approved by NHS ethics committees under the reference 4/LO/1527 14/EM/1159.

| Tissue acquisition and processing
All patients had a confirmed histological diagnosis of MPM before recruitment into the MEDUSA cohort and were undergoing routine surgery involving extended pleurectomy decortication. 10During surgery, samples were collected consistently from the same distinct sites of the tumor including (1) apex, (2) pericardium, (3) anterior costophrenic recess, (4) posterior costophrenic recess, and (5) oblique fissure.
Genomic DNA was extracted from peripheral blood and tumor tissue using a QiAamp mini kit (Qiagen).Total RNA from 100 μm sections of formalin-fixed, paraffin-embedded (FFPE) tissue was isolated using RNAprep pure tissue Kit from TIANGEN (DP431).RNA concentration was measured using Qubit RNA Assay Kit (Life Technologies, CA, USA) and RNA integrity was assessed using the Bioanalyzer 2100 system (Agilent Technologies, CA, USA).

| Whole exome sequencing
The GRCh37 reference genome was used throughout this project, and all genomic locations refer to that reference genome.Using the Agilent SureSelect Human All ExonV6 kit (Agilent), the WES library was prepared using 1 μg of genomic DNA derived from each tumor and matched peripheral blood samples.The genomic DNA was sheared into fragments of around 180-280 bp using the hydrodynamic shearing system (Covaris) and fragments were then hybridized with biotinylated probes and captured by Streptomycin-coated magnetic beads to enrich Exonic regions via amplification.This is followed by paired end sequencing using the Illumina NovaSeq 6000 platform with a coverage of 276Â at Novogene Ltd.After passing quality control, the reads were aligned to the genome assembly using Burrows-Wheeler Aligner (bwa-0.7.17).A combination of tools including Sambamba (v0.6.7), 26 Picard (v2.18.9),FastQC (v0.11.5) and SAMtools ref (v1.8) were used for data filtering and quality control.

| RNA sequencing
2 μg RNA per sample was used as input material for the RNA sample preparations.Sequencing libraries were generated using NEBNext ® UltraTM RNA Library Prep Kit for Illumina ® (NEB, USA) following manufacturer's recommendations and index codes were added to attribute sequences to each sample.Briefly, mRNA was purified from total RNA using poly-T oligo-attached magnetic beads.After fragmentation, the first strand cDNA was synthesized using random hexamer primer followed by the second strand cDNA synthesis using dTTP.
Remaining overhangs were converted into blunt ends via exonuclease/polymerase activities.After adenylation of 3 0 ends of DNA fragments, NEBNext Adaptor with hairpin loop structure were ligated to prepare for hybridization.In order to select cDNA fragments of preferentially 150-200 bp in length, the library fragments were purified with AMPure XP system (Beckman Coulter, Beverly, USA).The concentration of each library was measured with real-time PCR.Pools of the indexed library were then prepared for cluster generation and PE150 sequencing on Illumina NovaSeq 6000 at Novagene Ltd.

| Gene fusion detection
Meerkat detects gene fusions from the input aligned sequence file in BAM format by spanning each discordant read pair across the fusion partners. 27With discordant read pair support, we can only indirectly infer that there is a breakpoint somewhere between the spanning regions.Split read mapping refines the breakpoint regions by local alignment.The structural variants (SVs) undergo a filtering pipeline to remove germline variants present in the reference and matched normal BAM files.A prediction was discarded if the fusion event is smaller than 100 bp or larger than 1 Mb, or the fusion event is mapped to simple or satellite repeats, or the fusion event predicted has an extensive homology of more than 40 bp at the breakpoint junction, or the identity at the breakpoint is less than 20 bp for fusion events involving inter-chromosomal translocation.
Fusion events either contains more than 25% nonuniquely mapped reads, >3 discordant read pairs and > soft-clipped reads across the predicted fusion breakpoint in the normal matched genome were also discarded.Finally, intergenic-intergenic fusions, fusion events between paralogous genes, and fusions mapping to repetitive regions of the genome were removed, and truncal fusion events were selected.To confirm somatic status, fusion events were inspected using the Integrative Genome Browser (IGV).
Delly (version 0.8.7) 28 requires a joint input of tumor and matched normal sequencing data (BAM files), an indexed reference genome file and a list of genomic regions excluded from SV calling, such as centromeres and telomeres.The software predict SVs based on paired-end reads which can identify insert size distribution and read orientation, and split read support which provides single nucleotide resolution to characterize fusion breakpoints and investigate microhomologies or micro insertions.There are multiple filtering steps in between command runs in order to exclude any unmapped contigs and reads aligned to the centromere and telomere regions, as well as to remove any germline SVs detected in the matched normal exome.The output BCF (binary variant format) file containing the predicted somatic SVs, is converted to a VCF (variant calling format) file.To confirm somatic status, fusion events were inspected using the Integrative Genome Viewer (IGV). 29

| Gene fusion detection from RNA sequencing
The STAR aligner (v.2.7.9a) 30 generates chimeric alignments from RNAseq data using a reference genome index file.The chimeric alignments are recorded in an BAM file, which is then used by Arriba (v.2.1.0) 31to predict fusion events by searching for split and pairedend (discordant) reads.Fusion predictions were filtered for somatic fusions by discarding those detected in matched RNAseq data from whole blood.Next, likely recurrent artifacts were removed such as fusions predicted from read-throughs, noncanonical splicing or internal tandem duplication.

| Statistics
Comparison of SCNA counts in truncal gene fusion positive and negative tumors used a two-tailed Fisher's exact test implemented in R v.4.1.0.

| Identification of truncal gene fusions
An overview of the approach used to detect gene fusions from exome data is shown in Figure 1.Exomes from 20 patients were analyzed JAMA ET AL.
3 of 12 using Meerkat, with 14 patients having exomes from four sampling sites and peripheral blood, and 6 patients (MED1, MED23, MED24, MED27, MED34, and MED37) with exomes from five sampling sites and peripheral blood.After filtering and visual inspection of the bam sequence alignments using Integrated Genomics Viewer (IGV), 23 truncal fusion events involving at least one gene were detected across five patients (Table 1), with no truncal fusions seen in more than one patient (Supplementary Table S2).For the 23 truncal fusion events detected, nine were a consequence of chromosomal translocations, six a consequence of inversions, five a consequence of duplications and three a consequence of deletions.Of the 23 truncal fusions identified using Meerkat, we found 21 by using Delly to independently identify SV breakpoints (Table 1).In addition, another SLC39A1 fusion (ATP1A4-SLC39A1) was also observed in patient 23, but only predicted in 3/5 regional tumor samples.An exception to the truncal rule (must be detected in 4-5 regional samples) was applied in this instance to further study the various SVs affecting the gene SLC39A1 (Table 1).To validate our gene fusions, we selected these 24 fusion events across different tumors, and 22 (92%) were confirmed by genomic PCR across the breakpoint and Sanger sequencing (Table 1, Supplementary Table S1), including the ATP1A4-SLC39A1 fusion not being initially identified as truncal.
To expand the scope of our analysis, we integrated publicly available data from a range of cancer types from CBioPortal (www. cbioportal.org).We extracted information on gene fusions, copy number alterations, and nonsynonymous mutations, and focused on our 22 validated fusion events from MPMs in this study.Twelve of our fusion events in MPMs had one fusion gene partner previously reported to be involved in a gene fusion in tumors, and seven fusions having both gene partners previously reported as a gene fusion in tumors (Figure 2).The genes HSD17B4, EEF2, TBC1D32, GOLM1, and PARD3B have been reported as canonical fusions in tumors. 32We found fusions involving BAP1 and MTAP in patient MED24, with both fusions acting as a clonal second hit for these gene.Both BAP1 and MTAP fusion genes have been previously reported in MPM, but none of the other fusions found had previously been reported in MPM.The majority of the genes involved in the fusions identified here (22/29)   have been previously reported to be altered by copy number The fusion breakpoints were predicted using Delly and Meerkat, using whole exome sequencing files as input.Inter-and intra-chromosomal (Chr) structural variants (SV) generated the truncal fusion events such as inversions (INV), deletions (DEL), tandem duplications (DUP) and translocations (T).Fusion events were inspected using integrative genome browser (IGV) to confirm somatic (S) status.In the "PCR and Sanger" column, positive and negative findings are indicated as "+" and "À."Only MED6 RNA-seq data were available, hence N/A denotes no data available.We also investigated whether the second copy of the fusion gene partners are altered via inspecting copy number (CN) and single nucleotides variants (SNV) data corresponding to each patient (data from reference 10).Truncal and branch events are represented as the "Tr" and "Br," respectively.
alterations or nonsynonymous mutations in MPM, 9,33 except ARMC2, GOLM1, B3GALT2, OR2W5, LONP2, PARD3B, and SYDE2, which may represent novel genes affected in MPM (Figure 2).Only three of the genes involved in these fusions have been previously identified by Cosmic (v95) as being tumor suppressor genes (BAP1) or oncogenes (CACNA1D, LRP1B), further suggesting that the remaining fusions may highlight potentially novel genes involved in MPM.
For 13 of the 20 patients, matched multiregional RNASeq data was available.A total of 950 fusion transcripts were identified in these 13 patients, of which 32 fusions were truncal, ranging from 0 to 8 fusions per patient.16 of 32 were detected in corresponding WES profiles for each MPM using IGV (Supplementary Table S3) and were also confirmed as somatic due to the absence of fusion event in the matched normal.Delly found all the truncal fusion predictions from the RNA seq data, in contrast to Meerkat, which detected 14 of the 16 with only 4 truncally (Table 2).Combining genomic and transcriptomic data, we found 36 tumor-specific novel truncal fusions across nine patients (MED6, MED8, MED18, MED23, MED24, MED27, MED33, MED34, and MED37) (Tables 1 and 2).
In our set of truncal fusions, the SLC39A1-ATP1A2 and ATP1A4-SLC39A1 fusion genes detected in patient MED23 are of particular interest.SLC39A1 encodes a zinc transporter and is a known tumor suppressor gene controlled by YAP, part of the Hippo pathway. 34The Hippo pathway is known to be dysregulated in mesothelioma and therefore loss of this gene is potentially a driver of mesothelioma. 35cause the two fusion events breakpoints in SLC39A1 are slightly different, and the two fusion events involve breakpoints in different members of the ATP1A family, two distinct inversions on both homologous chromosomes is possible.However, one event involving a single inversion followed by a deletion on one chromosome, is more likely (Figure 3).
Although no gene fusions were recurrent across patients, patient MED6 showed distinct rearrangements affecting the HSD17B4 gene on both alleles, implicating HSD17B4 as a novel tumor suppressor gene in MPM (Figure 4).This interpretation was supported by transcript data which identified the truncated transcripts predicted by the exome analysis.The gene encodes the enzyme 17-beta-hydroxysteroid-dehydrogenase IV which is involved in fatty acid β-oxidation metabolism in peroxisomes and steroid metabolism. 36,37Its role in fatty acid metabolism potentially links it to feroptosis, a form of programmed cell death that plays a crucial role in tumor suppression.1, Supplementary Table S4).

| Mechanism of gene fusion formation
As the truncal gene fusions were found in a subset of patients analyzed (8 out of a total of 20 patients), we considered whether there was a particular molecular lesion or pattern correlated with, and possibly explaining, the presence or absence of truncal gene fusions in particular tumors.For example, BAP1 mutation is a common driver of mesothelioma, and BAP1 is a key component of the homologous recombination pathway, so we hypothesized that these eight truncal fusion positive patients (MED6, MED23, MED24, MED27, MED34, MED18, MED33, and MED8) might all have mutations in BAP1 and therefore a likely deficiency in homologous recombination (HR) repair, thereby promoting NHEJ-mediated repair of double-strand breaks.
From the 14 of 20 patients where SCNA data were available from a previous study, 10 we identified six of the eight truncal fusion positive tumors showed truncal BAP1 loss from genomic analysis, in which five of eight (MED24, MED18, MED33, MED27 and MED34) harbored a double-hit BAP1 alteration.In comparison, only two of the six truncal Genes involved in the validated malignant pleural mesothelioma (MPM) truncal fusions.
For the genes affected by fusions in our cohort, the grid shows the extent of somatic alteration in MPM and across other tumor types as well as known cancer genes.Data from references 9,32,33 and COSMIC v91.
fusion negative patients show BAP1 loss, but this difference is not significant, possibly because of small numbers (p = 0.2) (Supplementary Table S5).

| Clonal fusions are associated with impaired loss of homologous recombination DNA repair and Hippo pathway inactivation
In order to explore the basis of the difference between the truncal gene fusion-positive and truncal gene fusion-negative patients further, we compared truncal gene losses in the two groups to look for enrichment for particular pathways.This increases statistical power by analyzing changes across multiple genes in a pathway, rather than individual genes.Using the gene list from the KEGG database (hsa03440), 33 out of 39 genes involved in HR were lost truncally, with enrichment in the eight truncal fusion positive patients ( p = 0.01, two sided Fisher's exact test) (Figure 5) (Supplementary Table S5).Analysis of the breakpoints suggests that this loss of HR gene function in accompanied by a switch to NHEJ (Supplementary Table S4).Although clonal alterations do affect some NHEJ genes (KEGG hsa03450), there is no difference in the number of NHEJ genes affected between truncal gene fusion positive and negative patients ( p = 0.08, two sided Fisher's exact test).We also analyzed the Hippo pathway genes (KEGG hsa04390) in a similar manner and many more truncal alterations were found in truncal gene fusion positive patients than negative ( p < 0.0001, two sided Fisher's exact test).Together this suggests that impaired Hippo pathway leading to DNA damage response 38 and impaired HR are responsible for an early, truncal gene fusion positive phenotype.
The above analysis is a correlation, and it could be argued that the high rate of gene fusions and high rate of gene loss are simply due to tumors from gene fusion positive tumors having a high number of SCNAs, as both gene fusions and gene loss are caused by SCNAs.However, there is no difference between genefusion-positive and gene-fusion-negative tumors in the proportion of the genome affected by truncal SCNAs (p = 0.95, Wilcoxon rank sum test, data from Reference 10).Alternatively, the apparent enrichment of genes in particular pathways could be due to more genes of those pathways being present in the genome.If this was the case, however, we would expect there to be no relationship with time, in that the enrichment shown in later SCNA/SNV events would be the same as the enrichment shown in early events.Because we can distinguish early and late events in tumor evolution by clonal/ subclonal analysis of the multiple regions of the tumor, we can test this by testing for enrichment when subclonal events are also included.For genes involved in HR and the Hippo pathway, this enrichment disappears when subclonal events are included, with no difference between truncal gene fusion positive patients and truncal gene fusion negative patients (Supplementary Table S5).This strongly argues that early disruption of HR genes and Hippo pathway genes causes extensive further gene fusion events.
T A B L E 2 Truncal fusions identified in the MPM cohort from RNAseq.Note: The fusion breakpoints were predicted using Arriba, using RNA sequencing files as input.Inter-and intra-chromosomal (Chr) structural variants (SV) generated the truncal fusions including inversions (INV), deletions (DEL), tandem duplications (DUP) and translocations (T).Truncal fusion calls from Arriba were compared to fusion data obtained from SV callers, Meerkat and Delly.Fusion events were inspected using integrative genome browser (IGV) to confirm somatic (S) status.We also investigated whether the second copy of the fusion gene partners are altered via inspecting copy number (CN) and single nucleotides variants (SNV) data corresponding to each patient.Truncal and branch events were denoted as "Tr" and "Br," respectively.

| DISCUSSION
In this study we analyzed exomes from multiple regions of malignant pleural mesothelioma in 20 patients to determine the extent and nature of gene fusion events that occurred early in the evolution of the tumor.Our results reveal the presence of BAP1 and MTAP fusions, consistent with previous studies in the field (Figure 2).Both of these fusions act as clonal second mutational hits to these genes in HSD17B4 encoding for an enzyme involved in steroid degradation in the peroxisome.These genes have been previously reported to be involved in tumor progression.Several fusion events act as second mutational hits to genes which have either copy number loss or a somatic single nucleotide mutation on the other allele (eg, 17/35 genes identified in whole exome sequencing, Table 1).This emphasizes the importance of searching for fusion events when identifying potential cancer tumor suppressor loci, as gene fusion events can be missed by standard nucleotide and copy number analyses.We also identified a novel potentially oncogenic fusion (CACNA1D-ERC2) and an in-frame PARD3B-NT5DC2 fusion in a subset of MPM samples using exome sequencing, although we were unable to validate the transcription of these fusions.
There are limitations to this study.First, the number of patients is small so our power is limited to drawing only broad conclusions from our data.Extending patient cohorts and data sharing is critical particularly for rare tumor types such as MPM.Second, our analysis uses whole exome sequencing rather than whole genome sequencing, meaning that only a small proportion of the genome is sequenced at high coverage and many fusion events will be missed to due low coverage of sequence reads.Nevertheless, it is noticeable that at 200Â coverage of the exome there are enough off-target sequence reads to identify at least some intergenic events, as 19 of 36 fusions identified involved a breakpoint in an intergenic region.These are potentially important, as they will inactivate the gene involved in the fusion.
Our key findings are that each patient had a unique constellation of truncal gene fusions, and truncal gene fusions occur only in a subset of MPMs, generating two classes of truncal gene fusion positive and negative tumors.For the gene fusion positive tumors, early truncal disruption by SCNAs are enriched in genes involved in the Hippo pathway and homologous repair genes.Our model is that, in some patients, early loss of key genes involved in DNA damage sensing and homologous repair causes a reliance on NHEJ and alt-EJ mechanisms to repair SV.These subsequent SVs can cause further loss of tumor suppressor genes or formation of oncogenic gene fusions.The overall burden of truncal SCNA does not differ between gene fusion positive and gene fusion negative tumors, so it is not clear why apparently similar levels SCNA should generate gene fusions in some tumors but not others.Our hypothesis is that this reflects the different sizes or different natures of SV alterations, so that truncal-gene fusion positive tumors either have smaller SCNAs, or more copy number neutral SVs, generated by NHEJ and alt-EJ mechanisms.A larger dataset with whole genome sequence is needed to reliably establish a more complete view of the SV landscape to test this idea.
Our results have clinical therapeutic implications.Because disruption of the Hippo pathway activates the YAP protein, small molecule inhibition of YAP has important therapeutic potential in MPM.
Our results show that a subset of patients with high numbers of fusions may be especially disrupted in the Hippo pathway and therefore could be particularly responsive to such treatments.At present, NHEJ and alt-EJ inhibitors are currently in clinical and preclinical trails, 39,40 and gene fusion positive patients may show increased vulnerabilities to these inhibitors, allowing another approach to MPM treatment.

| CONCLUSION
This study shows the benefit of multiregional sampling of tumors to detect genomic events occurring clonally across the tumor, and therefore implying that they occurred early in evolution.This has allowed us to identify potentially oncogenic gene fusions, and to construct a model of mesothelioma evolution for a subset of tumors consistent with our data based on the early occurrence of genomic F I G U R E 5 Truncal alterations in HR and Hippo pathway genes enriched in truncal fusion positive malignant pleural mesothelioma (MPMs).Patients IDs are in rows and the genes in the particular pathway are in columns.Truncal alterations are inferred if all the regional samples of a tumor have the same alteration, red denotes copy number gain, blue indicates copy number loss and yellow denotes somatic SNV.Source data from Zhang et al. 10 .
changes.The ability to deconstruct the timing of genomic events in a tumor has important therapeutic and diagnostic implications, particularly for mesothelioma, where the long latency and late diagnosis of the tumor remain a challenge for successful treatment of this cancer.

AUTHOR CONTRIBUTIONS
Funding acquisition: Edward J. Hollox and Dean A.

A
duplex PCR assay was designed for each potential fusion, with one primer pair designed based on the chromosomal orientation of each fusion partner and the other primer pair amplifying an unrelated region of the human genome acting as a positive control.Primer sequences were constructed using Primer3 (version 0.4.0) and flanked a distance of 0.5 kb from the proximal and distal position of the breakpoint junction to ensure the PCR product spanned the predicted breakpoints for each fusion event.Following standard PCR (1.5 mM MgCl 2 , Tris-ammonium sulfate buffer) and agarose gel electrophoresis, amplicons were extracted from the agarose gel using a Monarch ® DNA Gel Extraction Kit (New England Biolabs) and Sanger sequenced using standard approaches (Eurofins Genomics [Cologne] and Source Biosciences [Nottingham]).

1 B3GALT2 1
Overview of study.T A B L E 1 Truncal fusions identified in the MPM cohort from WES. Patient Meerkat truncal fusion predictions Breakpoint location Chr SV Observed using Delly IGV PCR and Sanger Observed in RNA-seq CN loss SNV MED6 igs-5 0

F I G U R E 3
Validation of an inversion generating a ATP1A2-SLC39A1 fusion gene.(A) Genomic Sanger sequence of breakpoint 1 of patient MED23.(B) Genomic Sanger sequence of breakpoint 2 of patient MED23.(C) Inferred structure of rearrangement based on breakpoint location and orientation.these two patients.However, we did not identify any potentially clonal truncal gene fusions involving the NF2 gene, which has been reported in other studies. 8.Additionally, we did not observe any common recurrent gene fusions in our sample population.However, other clonal events merit further exploration, such as an in-frame PARD3B-STAB1 fusion, as well as truncal biallelic alteration of F I G U R E 4 Validation of two fusion events disrupting HSD17B4.(A) Genomic Sanger sequence of HSD17B4 inversion breakpoint 1 in patient MED6.(B) HSD17B4 inversion breakpoint allele 1 inferred from RNASeq data.(C) Genomic Sanger sequence of HSD17B4 inversion breakpoint 1 in patient MED6.(D) HSD17B4 deletion breakpoint allele 2 inferred from RNASeq data.