If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Malignant pleural mesothelioma is a disease primarily associated with exposure to the carcinogen asbestos. Whereas other carcinogen-related tumors are associated with a high tumor mutation burden, mesothelioma is not. We sought to resolve this discrepancy.
Methods
We used mate-pair (n = 22), RNA (n = 28), and T cell receptor sequencing along with in silico predictions and immunologic assays to understand how structural variants of chromosomes affect the transcriptome.
Results
We observed that inter- or intrachromosomal rearrangements were present in every specimen and were frequently in a pattern of chromoanagenesis such as chromoplexy or chromothripsis. Transcription of rearrangement-related junctions was predicted to result in many potential neoantigens, some of which were proven to bind patient-specific major histocompatibility complex molecules and to expand intratumoral T cell clones. T cells responsive to these predicted neoantigens were also present in a patient’s circulating T cell repertoire. Analysis of genomic array data from the mesothelioma cohort in The Cancer Genome Atlas suggested that multiple chromothriptic-like events negatively impact survival.
Conclusions
Our findings represent the discovery of potential neoantigen expression driven by structural chromosomal rearrangements. These results may have implications for the development of novel immunotherapeutic strategies and the selection of patients to receive immunotherapies.
Malignant pleural mesothelioma (MPM) is characterized by exposure to the carcinogen asbestos, although some cases develop after therapeutic radiation or are inherited due to loss of function mutations in BRCA1 associated protein 1 (BAP1).
More recent studies that used next-generation sequencing (NGS) to evaluate single nucleotide variation mutations in MPM reported a very low mutation burden overall.
This finding was unexpected because other tumors associated with carcinogenic exposure such as malignant melanoma and NSCLC typically have a high mutation burden from ultraviolet radiation and tobacco exposure, respectively.
Since a high tumor mutation burden and an increase in predicted neoantigens has been linked to favorable responses to immune checkpoint inhibitors of programmed cell death 1 ligand 1 (PD-L1) or its receptor programmed death 1 (PD-1), and given that response rates are similar between patients with MPM or NSCLC in second or later lines of therapy treated with these inhibitors, we hypothesized that chromosomal rearrangements that may not be detected by conventional exon-based paired-end sequencing techniques may result in expression of additional neoantigens in MPM.
Mate-pair sequencing (MPseq) differs from standard NGS approaches by tiling the whole genome with larger fragments (2 to 5 kb) to reliably detect structural variants such as insertions, deletions, and rearrangements.
We previously used MPseq to determine the lineage relationships of multifocal lung cancer, to identify a chromoplectic ALK rearrangement in an inflammatory myofibroblastic tumor, to discover recurrent t(6:7)(p25.3;q32.3) translocations in ALK-negative anaplastic large cell lymphomas, and for other purposes.
Recently, we have also integrated MPseq with RNA sequencing (RNAseq) to identify transcribed chromosomal rearrangements in peripheral T cell lymphomas.
To resolve the reported discrepancy between the numerous chromosomal abnormalities detected by cytogenetics and the low mutation burden detected by standard exome sequencing in MPM, we leveraged MPseq and RNAseq to detect transcribed structural variants. We hypothesized that this method would predict expression of rearrangement-related peptides with neoantigenic potential and define a novel mechanism of tumor immunogenicity that may not be detected with standard sequencing approaches.
Methods
Patients
Twenty-eight cases of frozen MPM specimens that were collected through the Mayo Clinic Thoracic Specimen Registry or routine clinical care were identified, retrieved, and reviewed. These specimens were obtained before the administration of any systemic therapy. The diagnosis of MPM and the subtype of each case were verified by a pathologist (J.C.C. or M.C.A.). Patient characteristics are presented in Supplemental Table 1.
Tissue Extraction and NGS
Tumor DNA and RNA were extracted independently from each fresh frozen clinical specimen using the DNeasy Blood and Tissue kit (Qiagen, Germantown, Maryland, #69504) or RNeasy Plus Mini kit (Qiagen, #74134), respectively, following established protocols. Total RNA and DNA yields and concentrations were determined by Qbit. Indexed libraries for MPseq (1 μg DNA) and RNAseq (100 ng total RNA) were generated using the Nextera Mate-Pair Kit (Illumina, California, FC-132-1001) or the TruSeq RNA Access Library Prep Kit (Illumina, RS-301-2001), as previously described, following the manufacturer’s instructions.
The read-to-reference-genome-mapping algorithm was modified to map both mate-pair (MP) reads across the whole genome. Discordant mapping MP reads covered by at least five associates were identified for further analysis. Concordant mapping MP reads were used to determine frequency coverage levels across the genome.
We recently published (1) how MPseq provides comparable sensitivity to the state-of-the-art chromosomal microarray technology for the detection of copy number variations, with the benefit of improved breakpoint resolution; (2) how use of an internal database of polymorphisms is used to mask germline events; and (3) that display of MPseq results with a U plot which provides enhanced resolution of rearrangements and copy number variations than some other commonly used displays.
Transcriptome analyses were performed in the R/Bioconductor environment (https://www.bioconductor.org/). Paired-end sequence fragments were aligned by the TopHat aligner using the latest genome reference file (hg38) and the ensemble annotation database. Sorted “sam” files were input to the htseq program to calculate expression levels of genes in each sample. Finally, the “.count” files from the previous step were used by the edgeR program to generate a normalized expression matrix for all 28 samples. The data presented are from the 22 specimens with accompanying MPseq data.
Validation of Rearrangements and Fusion Transcripts
Primers spanning the detected fusion junctions were used in DNA and cDNA polymerase chain reaction (PCR) validations (25-μL reaction volumes, 50-ng Template, 35 cycles) using EasyA high-fidelity polymerase (Stratagene, La Jolla, California, #600404). DNA PCR validations were performed on tumor DNA and a pooled human genomic DNA control (Promega, #G304A, Madison, Wisconsin). A total of 700 ng of total RNA was reverse transcribed into cDNA using Random Hexamers (Invitrogen, Carlsbad, California, #N8080127) and Superscript III (Invitrogen, #18080093). An unrelated case, ME025, was used as a negative control for RNA validations. BetaActin control PCRs were performed using standard primers. Junctions were confirmed in both DNA and cDNA with Sanger Sequencing of PCR products. DNA/cDNA was electrophoresed on a 4% agarose gel and visualized with ethidium bromide.
HLA Typing
HLA typing was performed computationally directly from RNAseq data. An HLA-specific reference genome consisting of class I HLA sequences was created from the IPD-IMGT/HLA Database (release 3.29.0). This reference file was used as input to a TopHat alignment of the RNAseq reads. All reads that aligned to the HLA sequences were taken and curated to remove all secondary alignments, low-quality mapping reads, replicates, and reads that contained insertions or deletions. The remaining primary, high-quality reads were used to create a sequence alignment for each position in the HLA being typed. Any position found to have less than 90% consensus for a particular nucleotide was considered a position of allelic variation. Using these positions of allelic variation, haplotype phasing was performed to produce consensus sequences for the two HLA alleles. Only positions with high read depth coverage (more than 1000) were used for final typing. The final typing results provided a full list of HLA sequences that matched the consensus sequence with 100% sequence identity.
MPseq-RNAseq-HLA Matching Neoantigenicity
Potentially altered peptides that could include expressed neoantigens were first identified by integrating DNA junction data, derived from MPseq, with expression data, derived from RNAseq data. MPseq and RNAseq sequences were mapped by using Binary Indexing Mapping Algorithm (BIMA) and mapped with SVAtools and junctions were identified from each case.
An algorithm in R was developed to go through each junction and uncover pair-ends from the corresponding RNAseq data file that span the two regions. To accommodate for splicing the region, windows were 1 Mb. A lower threshold of at least two independent pair ends spanning the regions was required. The paired-end sequences from RNAseq were aligned with each other to develop a consensus sequence, which then was translated to amino acid sequence. The aberrant part of that sequence was provided to the step below to predict the neoantigens.
Antigen Presentation Prediction Methods
Prospective neoantigens were identified through application of software tools that predict proteasome cleavage, tapasin binding protein (TAP) transport, binding of peptide in HLA molecules, and immunogenicity. For each peptide product inferred from an unambiguous reading frame, constitutive proteasome cleavage and TAP transport was predicted for each 9- to 11-mer using the 2013-02-22 version of the Immune Epitope Database (IEDB) MHC-I Processing Prediction tool.
Binding affinity of each peptide to the individual’s HLA molecules was predicted using NetMHC-4.0 and relative binding affinity was assessed using the percentile rank measure.
Binders were defined as having a binding rank in the 98th percentile, and strong binders in the 99.5th percentile. Immunogenic reactivity with T cell receptors was predicted using the class I immunogenicity predictor hosted by the Immune Epitope Database (IEDB), which compares the residues at key contact positions with those known to engage with T cell receptors in immunogenic epitopes.
Antigens presented in HLA class I molecules are derived from cytosolic proteins that are cleaved into short peptide sequences in the proteasome, and then loaded into the HLA molecules by TAP. The sites at which proteasomes tend to cleave are relatively conserved, and multiple cleavage site prediction tools have been developed. The IEDB hosts a major histocompatibility complex (MHC) class I processing prediction tool that predicts proteasome cleavage sites and TAP binding affinity, which returns a processing score that estimates the relative quantity of peptide yielded by these processes (http://tools.iedb.org/processing/). It further incorporates HLA:peptide binding predictions, using a user-specified tool. The processing score for this analysis was performed using NetMHC.
Immunogenicity
IEDB also hosts a tool that predicts the likelihood that an HLA:peptide complex will be recognized by a T cell receptor (TCR) based on amino acid properties, position of the residue in the peptide, and HLA allele (http://tools.iedb.org/immunogenicity/). This was developed based on observations that larger and more aromatic residues tended to increase the likelihood of TCR interaction in a consistent manner, namely that particular features at different residue positions favored T cell interaction. Higher ratings indicate prediction of more TCR interaction, although the authors show that simply having a positive immunogenicity score is sufficient. The data set used to train this tool contained only 9-mer peptides, although it extends its predictions to larger sizes as well.
Peptide Synthesis
Peptides NYLETTSDF, NYLETTSDFHF, and CYGETYQNI were synthesized by Fmoc solid phase methods on preloaded Wang resin (Anaspec, Inc., Fremont, California) by Mayo Clinic’s Medical Genome Facility’s Peptide Synthesis Services. Each peptide chain was assembled from Nα-Fmoc protected amino acids on a Liberty Blue Microwave-Assisted Peptide Synthesizer (CEM Corp., Pennsauken, New Jersey) according to the manufacturer’s coupling and deprotection protocols. After synthesis each peptide was deprotected and removed from its resin support by acidolysis with a solution of trifluoroacetic acid containing 2.5% water (v/v) and 2.5% triisopropylsilane (v/v) and 2.5% 3,6, dioxa-1,8-octanedithiol (v/v) for 30 minutes at 42°C. The crude peptide was then purified by preparative reversed-phase high-performance liquid chromatography (LC) using an aqueous acetonitrile gradient containing 0.1% trifluoroacetic acid (v/v) on a reverse-phase C18 column (Phenomenex Jupiter 15μ; 250 × 21.2 mm). The mass weight was verified by LC-electrospray ionization mass spectrometry on an Agilent 6224 TOF LC/mass spectrometry instrument. Peptide homogeneity was confirmed by analytical reversed-phase high-performance liquid chromatography and was 99.9%, 95.5%, and 98.2% for NYLETTSDF, NYLETTSDFHF, and CYGETYQNI, respectively.
MHC Peptide Binding Assay
Peptides NYLETTSDF, NYLETTSDFHF, and CYGETYQNI were assembled with allele HLA-A*24:02 and analyzed using an MHC-peptide binding assay (ProImmune REVEAL, ProImmune Ltd., Oxford, United Kingdom) to determine their level of incorporation into MHC molecules. Binding to MHC molecules was compared to that of a known T-cell epitope, a positive control peptide, with very strong binding properties. A binding score for each MHC-peptide complex was calculated by comparison to the binding of the relevant positive control. Experimental standard error was obtained by triplicate positive control binding experiments.
TCR Sequencing
TCR profiling was performed per protocol with ImmunoSEQ (Adaptive Biotechnologies, Seattle, Washington, hsTCRβ Kit) as we have done previously.
The number of productive T cell clones was used to determine Pielou’s Evenness Index () to understand whether T cell clones were equally distributed amongst specimens.
Productive Clonality is the inverse of Pielou’s Evenness Index and ranges from 0 to 1, where 0 represents an even distribution of clones and 1 represents an uneven distribution of clones (also known as clonal expansion).
Peripheral Blood Mononuclear Cell Preparation and Cryopreservation
Peripheral blood mononuclear cells (PBMCs) were isolated by Ficoll density gradient centrifugation. PBMCs were cryopreserved in liquid nitrogen at 10 million PBMCs/mL in Roswell Park Memorial Institute (RPMI) solution containing 10% (v/v) DMSO, 110 mg/mL human serum albumin, and 2.7 mg/mL HEPEs.
Interferon-γ Enzyme-Linked Immunospot
In 96-well plates, 2.5 × 105 PBMCs per well plus different stimuli (media alone or media and one of the following peptides at 10 mg/mL: irrelevant cyclin D1 peptide control, MELLLVNKLKWNLAA, LASELREGF, YSTARALYL, KSDPYSTARYH, TALLPPAAL, TARYHSTLL, KSDPYSTAL, FFSLASFKM, DPYSTARAL) were added in 200 μL media (5% human sera + RPMI) and incubated at 37°C for 24 hours. Each sample was done in duplicate. Enzyme-linked immunospot (ELIspot) plates (Millipore, Billerica, Massachusetts) were coated with 10 μg/mL interferon gamma (IFN-γ) capture antibody (MabTech USA, Mariemont, Ohio) and incubated overnight. After 24 hours, the ELIspot plates were washed with phosphate-buffered saline (PBS) and blocked with media for 2 hours. The PBMCs and supernant were transferred from the 96-well plates to the ELIspot plates and incubated at 37°C for 24 hours. After washing with PBS containing 0.05% Tween-20, 2 μg/mL of biotinylated secondary antibody for IFN-γ (MabTech USA) was added and the plates were incubated for 2 hours at 37 °C followed by another wash. Next, 1 μL of Streptavidin-horseradish peroxidase (BD Pharmingen, San Diego, California) per milliliter of 10% fetal bovine serum in PBS was added and the plates were incubated for 1 hour at room temperature. For the final wash, plates were first washed with PBS containing 0.05% Tween-20, followed by washing with PBS. Plates were developed by adding 20 μL of 3-amino-9-ethyl-carbazole (AEC) chromogen per milliliter of AEC substrate (Sigma-Aldrich, Allentown, Pennsylvania) and the reaction was stopped with water. After drying overnight, the plates were read on an AID EliSpot reader (San Diego, California), which provides quantitative spot information based on the number of stimulated cells that secrete IFN-γ.
The Cancer Genome Atlas
We analyzed the mesothelioma cohort from The Cancer Genome Atlas (TCGA) with an algorithm (CTLPScanner) that uses a sliding window scan statistic to identify events consistent with chromothriptic-like pattern (CTLP) detection.
Our cutoffs for CTLP detection of copy number variation of 20 or more and a log10 likelihood ratio of 8 or more were derived from a clinical training set previously described, and are the default values for the CTLPScanner.
The “survival” package in R was used to create the Kaplan-Meier survival figure. The Cox proportional hazards model was used to assess the effects of histology and CTLP on survival using JMP Pro 13.0.0 (SAS Institute Inc. Cary, North Carolina). The Fisher’s exact test with a 2-tailed p value was used to compare proportions of cases with CTLPs between histologists.
Immunohistochemistry for CD3
Immunohistochemistry for CD3 was performed as we have done previously.
Temporal and spatial discordance of programmed cell death–ligand 1 expression and lymphocyte tumor infiltration between paired primary lesions and brain metastases in lung cancer.
Blocks were sectioned at 5 μ. Deparaffinization and immunohistochemistry staining were performed online. Staining for CD3 was performed on the Ventana Benchmark XT (Ventana Medical Systems, Tucson, Arizona). CD3, Mouse Monoclonal (Clone LN10, Leica, Buffalo, Illinois, #NCL-L-CD3-565) was diluted 1/250 and incubated for 15 minutes at 37°C. OptiView DAB (Ventana Medical Systems) was used for detection. Normal tonsil was used as positive control and normal tonsil without primary antibody was used as a negative control. The number of CD3+ tumor-infiltrating lymphocytes were counted and averaged over three high-powered fields.
Statistics and Data Presentation
To compare the frequency of chromosomal rearrangements, novel gene fusions or predicted neoantigens between epithelioid and biphasic or sarcomatoid subtypes of MPM we used the Mann-Whitney test. The correlation between PD-L1 expression (normalized expression determined by RNAseq) with genomic features and survival were determined with nonparametric Spearman correlation and were two-tailed. The numbers of spots detected by ELIspot were first compared with a one-way analysis of variance. Because of the significant finding, multiple comparisons between the control and each peptide were made with Dunnett’s test. Multiplicity adjusted p values were used to determine significance. Survival was estimated with the Kaplan-Meier methods for the TCGA mesothelioma dataset. PRISM 7 for Windows (GraphPad, La Jolla, California) was used for all tests and their related graphs unless otherwise described.
Results
To evaluate whether chromosomal rearrangements result in expression of potential neoantigens in MPM, we analyzed 28 frozen MPM specimens collected at Mayo Clinic by MPseq and RNAseq. Twenty-two of these specimens had sufficient DNA for analysis by MPseq and 28 specimens had sufficient RNA for analysis by RNAseq. We identified numerous chromosomal rearrangements in the majority of specimens. Among these 22 specimens there were 1535 chromosomal rearrangements (median 41, range 3 to 298; Table 1) that resulted in junctions or novel fusions of noncoding DNA or genes. Six-hundred thirty-seven of these rearrangements (median 22, range 5 to 103) resulted in novel fusions of genes. Junction-guided analysis of RNAseq expression data predicted that these junctions could result in the expression of 179 novel amino acid sequences (median 5, range 0 to 51) that are potential neoantigens (Fig. 1). There were no significant differences in the frequency of chromosomal rearrangements (p = 0.75), novel gene fusions (p = 0.57), or predicted neoantigens (p = 0.19) between epithelioid (n = 11) and biphasic or sarcomatoid (n = 10) subtypes of MPM. The predicted neoantigens were unique to each case. There was no correlation between PD-L1 (normalized expression determined by RNAseq) and chromosomal rearrangements (Spearman r = 0.26, p = 0.28), novel gene fusions (Spearman r = 0.41, p = 0.09) or predicted neoantigens (Spearman r = 0.38, p = 0.11). As expected, there was a negative correlation with PD-L1 expression and survival (Spearman r = -0.79, p < 0.0001).
Table 1Summary of Patterns of Chromoanagenesis
Specimen #
Junctions
Chromothripsis
Chromoplexy
ME015
3
0
0
ME016
4
0
0
ME025
7
0
0
ME017
11
0
0
LU-3127
18
0
0
LU-3396
22
0
0
ME021
22
0
0
ME027
30
0
0
ME026
37
0
0
LU-3023
38
2
0
LU-2951
40
0
0
LU-1304904895
40
0
1
LU-2687
58
2
1
LU-2578
74
0
1
LU-1416305913
83
2
1
ME018
103
2
0
ME028
120
4
1
ME020
125
2
1
LU-0179
128
2
1
ME022
136
2
1
LU-2232
138
4
1
LU-0727
298
1
0
Specimens are ordered by increasing number of junctions (genomic rearrangements) detected by mate-pair sequencing and their associated patterns of chromoanagenesis if present.
Figure 1Genomic event summary. The unfilled circles represent the number of mate-pair sequencing–detected junctions (rearrangements), the black circles represent the number rearrangements involving genes, and the green circles represent the number of peptides that are potential neoantigens.
We further characterized the abnormalities detected by MPseq and observed that many rearrangements were part of a pattern of chromoanagenesis consistent with chromothripsis (present in 10 of 22 specimens, 45%), chromoplexy (present in 9 of 22 specimens, 41%), or both (present in 7 of 22 specimens, 32%; Table 1).
In 9 of 10 cases with chromothripsis, there were multiple chromothriptic events. Twelve of 22 (55%) specimens contained at least one chromoanagenic event with 0 to 4 events of chromothripsis (median = 0, mean = 1) and 0 to 1 (median 0, mean = 0.4) events of chromoplexy in each specimen. Chromothripsis most commonly involved chromosomes 2, 3, 6, 7, and 17, whereas chromoplexy involved multiple chromosomes including rearrangements between chromosomes 8 and 19, 2 and 9, or others (Supplemental Fig. 1). In some cases, chromoplectic events were associated with chromothripsis (Fig. 2, Supplemental Figs. 2 to 23). These chromosomal rearrangements were associated with amplifications and deletions of many chromosomal segments, affecting genes such as CDKN2A and NF2 at similar rates as reported by others (Supplemental Table 2 and Supplemental Fig. 24). The genes most frequently involved with chromosomal rearrangements were RBFOX1, PARK2, PTPRD, CTNNA3, and ANKS1B (Table 2).
Figure 2Genome plot for ME022. In this genome plot of specimen ME022 the chromosomes are plotted in order by size as numbered near the margins. Curved pink lines represent intrachromosomal rearrangements wheres light green lines represent interchromosomal rearrangements. Deletions are represented in red and amplifications are represented in blue. Accordingly, the multiple pink lines on chromosome 7 represent chromothripsis and the green lines between chromosomes 7 and 11 represent chromoplexy. CNV, copy number variation.
Table 2Summary of Genes Most Commonly Involved With Rearrangements
Gene Name
N (%)
Chromosome
RBFOX1
11 (50)
16
PARK2
4 (18)
6
PTPRD
4 (18)
9
CTNNA3
4 (18)
10
ANKS1B
4 (18)
12
ABGL4
3 (14)
1
SMYD3
3 (14)
1
LRP1B
3 (14)
2
FHIT
3 (14)
3
PTPRG
3 (14)
3
ANK2
3 (14)
4
MMS22L
3 (14)
6
BBS9
3 (14)
7
AUTS2
3 (14)
7
SLC24A2
3 (14)
9
MLLT3
3 (14)
9
LINC01239
3 (14)
9
LINCO2
3 (14)
9
CDH13
3 (14)
16
MACROD2
3 (14)
20
The genes most commonly involved with rearrangements in the 22 MPM specimens as determined by mate-pair sequencing and their chromosomal locations are summarized.
To determine whether transcription of chromosomal rearrangement-related junctions have neoantigenic potential, we focused on case ME018 because of its intermediate number of rearrangements in relation to the other specimens. We used NetMHC-4.0 to predict whether any of the expressed junctions could encode peptides that could be presented by the patient’s HLA molecules. A total of 1146 9-mer, 10-mer, and 11-mer peptides were identified from six protein products of expressed junctions with at least 10 RNAseq reads (Supplemental Table 3). Of these peptides, 75 were predicted to bind to ME018’s specific HLA molecules (HLA-A*02:05, HLA-A*24:02, HLA-B*15:01, and HLA-B*35:01) (Supplemental Figs. 25 to 29), yielding 87 distinct epitopes. Seventeen of the epitopes had a predicted HLA-peptide binding affinity in the 99.5th percentile, indicating that these peptides were potentially strong binders. The predicted binders were primarily sourced from the NOD1, CREB5, and CADPS2 protein products, with 34, 16, and 15 binding peptides derived from each protein, respectively (Supplemental Fig. 30). Nine peptides were predicted to form epitopes with multiple HLA molecules. The median proteasome and TAP processing score for these peptides was 1.48, which was significantly higher than the median of 0.82 for all 1146 peptides. The median immunogenicity score was 0.013, suggesting that prohibitively hydrophilic side chains at the contact residues are minimal. Given the significance of these predictions, we validated that CADPS2 was rearranged by PCR and Sanger sequencing using primers that spanned the fusion junctions in both DNA (Fig. 3) and RNA (Supplemental Table 4).
Figure 3Validation of CADPS2 rearrangement in ME018. The junction plot in the top left presents the mate-pair reads supporting the intrachromosomal rearrangement between CADPS2 and a nongenic region on chromosome 7. Red and blue dots represent mate-pair reads mapping to the positive or negative DNA strands, respectively, with coverage levels across each position presented in gray-shaded regions, with single (1N) and two-copy levels (2N) indicated (green). Gene regions are presented in red with exons numerated. On the top right of this figure, this rearrangement was confirmed by polymerase chain reaction with primers spanning the breakpoint in both DNA and RNA in ME018, but not in negative controls (pooled human genomic DNA for DNA validation, and case ME025 which lacked this rearrangement for RNA validation). In the bottom of this figure, this rearrangement was confirmed by Sanger sequencing of the DNA.
To evaluate whether the candidate rearrangement-related peptides with neoantigenic potential bind patient-specific HLA molecules, the 9 mer of two candidate peptides (NYLETTSDF, CYGETYQNI) and the 11 mer (NYLETTSDFHF) of one candidate peptide were selected based on their predicted immunogenicity. These candidate peptides resulted from the rearrangement of noncoding DNA with CADPS2 or noncoding DNA with NOD1 (Supplemental Tables 5 to 9). An MHC-peptide binding assay was performed and showed that all three peptides bound HLA-A*24:02 very well or as well as that of a high affinity T-cell epitope–positive control (Supplemental Table 10).
To determine the effects of chromosomal rearrangements on tumor-infiltrating T-cell clonality, we used TCR sequencing. There was a correlation between the number of T-cell clones and predicted neoantigens (r2 = 0.35, p = 0.03) but not the number of chromosomal rearrangements (r2 = 0.26, p = 0.08) or rearrangements involving genes (r2 = 0.27, p = 0.07). For each specimen we also calculated a productive clonality score which represents the inverse of Pielou’s Evenness Index and provides a measurement of clonal distribution. There was a positive correlation between productive clonality with chromosomal rearrangements (r2 = 0.34, p = 0.04), rearrangements involving genes (r2 = 0.37, p = 0.03), and predicted neoantigens (r2 = 0.31, p = 0.04). In other words, an increase in predicted neoantigens resulting from chromosomal rearrangements was correlated with clonal expansion of tumor-infiltrating T cells.
We prospectively collected the PBMCs and tumor from an additional patient treated at our institution. The tumor was analyzed as per the 22 cases above and 86 potential neoantigens were predicted. We then used eight of these peptides that correspond to 20 epitopes (cross-presented on 1-4 HLA alleles, with a median of 2 epitopes per peptide) with neoantigenic potential for an IFN-γ ELIspot assay. This assay suggested that T cells responsive to these neoantigens were present in this patient’s PBMCs (Supplemental Fig. 31, Supplemental Table 11).
Finally, we analyzed the copy number segmentation data of the mesothelioma cohort (n = 87) in TCGA for chromothripsis using the CTLPScanner.
This method was selected due to the lack of MPseq data available for TCGA. In brief, the CTLPScanner method identifies recurring patterns of copy number switching in the genome, termed chromothripsis-like patterns (CTLPs). In the mesothelioma cohort, we observed 27, 22, 15, and 23 patients with 0, 1, 2, and 3 or more CTLPs, respectively. There were similar rates of any CTLPs in epithelioid (68%) and nonepithelioid (70%) cases (p = 1.0), and similar rates of 3 or more CTLPs in epithelioid (25%) and nonepithelioid (30%) cases (p = 0.62). Subjects with 3 or more CTLPs had a significantly worse prognosis than others (hazard ratio [HR]: 2.002, p = 0.006) (Fig. 4). We used a proportional hazards model to assess the effects of histology and CTLPs on survival from this dataset. Both CTLPs (HR: 2.2, p = 0.006) and histology (HR: 1.9, p = 0.013) had a significant effect on survival. Surprisingly, mesothelioma was one of only two cancer subtypes of the 34 found in TCGA to have an HR greater than 2 (the other being uterine corpus endometrial carcinoma) with respect to CTLPs and poor survival, and one of six those had a statistically significant association (p < 0.05) between CTLPs and survival.
Figure 4Kaplan-Meier survival curve for mesothelioma by chromothripsis-like pattern (CTLP) detection. Subjects were classified by the presence of 0 to 2 CTLPs (n = 64), or 3 or more CTLPs (3+; n = 23). Survival was significantly worse for patients with 3+ CTLPs (hazard ratio: 2.002, p = 0.006).
We used MPseq to identify that chromosomal rearrangements are common in MPM and frequently in a pattern of chromoanagenesis such as chromoplexy or chromothripsis. MPseq-guided analysis of RNAseq expression data revealed that these chromosomal rearrangements commonly result in the expression of junctions of genes and noncoding DNA with neoantigenic potential. An increase in predicted neoantigens was correlated with clonal expansion of tumor-infiltrating T cells. Some of these novel peptide sequences were predicted and proven to bind to patient-specific HLA molecules. Furthermore, T cells responsive to these predicted neoantigens were present in a patient’s circulating T cell repertoire. Finally, patients with multiple CTLPs had poor survival in the TCGA dataset. Overall, these data suggest that potential neoantigens may result from genomic rearrangements in addition to the more commonly described single nucleotide variants (SNVs).
Since BAP1 is commonly deleted in MPM but not always detected by exome-based NGS approaches, others interrogated chromosomal region 3p21 which harbors BAP1 using a high-density comparative genomic hybridization array.
Multiple biallelic deletions that alternated with copy number changes consistent with chromothripsis were identified. We also identified multiple events consistent with chromothripsis throughout MPM genomes that most frequently affected chromosomes 6, 7, and 17; however, we did not identify as many deletions specifically involving BAP1 possibly because high-density comparative genomic hybridization array can potentially identify smaller deletions (<3 kb) than MPseq based on the design of the array and the library preparation for MPseq.
On the other hand, MPseq can profile the whole genome and identify other patterns of chromoanagenesis including chromoplexy. For illustration of the sensitivity of MPseq to identify rearrangements, a recent study only found 43 gene fusions in 22 specimens with RNAseq in comparison to the 1535 rearrangements and 637 gene fusions we identified in the same sample size.
Our data also build upon recent findings that gene fusions may result in loss of copy or function of tumor suppressor genes such as BAP1, NF2, and others.
We most frequently identified involvement of RBFOX1 with chromosomal rearrangements. The significance of RBFOX1 remains uncertain in MPM, but it belongs to a family of proteins that regulate alternative mRNA splicing and its absence in the brain may increase susceptibility to seizures.
Many therapeutic approaches are driven by the inhibition of immune checkpoints such as PD-L1 which is commonly expressed in MPM and may affect the tumor microenvironment.
Over the last few decades there have been many attempts to target nonmutant cancer-germline or cancer testis antigens such as NY-ESO-1 and MAGE-A3 in a variety of tumors
; however, the availability of affordable sequencing technologies has facilitated more recent efforts to identify and target tumor-specific neoantigens resulting from point mutations.
For example, targeting of a KRAS G12D mutation in a patient with metastatic colorectal cancer with adoptive T-cell transfer resulted in a partial response.
Insertions and deletions (indels) that result from frameshifts have been shown to generate approximately three-times as many high-affinity neoantigens than nonsynonymous SNVs. These frameshifts were predictive of response to PD-1 inhibition in separate cohorts of patients with melanoma.
Although MPM has a relatively low nonsynonymous tumor mutation burden, our finding of rearrangement-related peptides with neoantigenic potential may promote the immunogenicity of MPM in a similar fashion as frameshift indel high-affinity neoantigens.
A weakness of the present work is the lack of sufficient tissue and germline DNA for accurate determination of tumor mutation burden for assessment of its relationship to chromoanagenesis in MPM. Also, although we have provided indirect evidence of these predicted neoantigens with the presence of T cells that recognize them in the ELIspot assay, direct evidence of computationally predicted neoantigens is an area of ongoing work of ours. Regardless, given the more recent efforts to develop personalized cancer immunotherapies, rearrangement-related peptides with neoantigenic potential provide an opportunity for tumor-specific vaccination strategies in MPM and other malignancies.
In conclusion, our data suggest that neoantigen expression may be driven by structural chromosomal rearrangements in mesothelioma. These results may have implications for the development of novel immunotherapeutic strategies and the selection of patients to receive immunotherapies.
Acknowledgements
The authors thank Bobbi-Ann Jebens for her assistance with the submission of this manuscript. This work was supported by Leah and Richard Lommen, the National Institutes of Health [NIH K12 CA90628] and Mayo Clinic’s Center for Individualized Medicine’s Biomarker Discovery Group.
The authors thankful the patients who provided specimens for use in this work, and Bobbi-Ann Jebens for her assistance submitting this manuscript.
Temporal and spatial discordance of programmed cell death–ligand 1 expression and lymphocyte tumor infiltration between paired primary lesions and brain metastases in lung cancer.
Disclaimer: Dr. Mansfield’s institution has received honoraria for his participation on advisory boards from AbbVie and Genentech. Dr. Peikert has been on the advisory board for Epizyme. The remaining authors declare no conflict of interest.
In this issue of the Journal of Thoracic Oncology, Mansfield et al. report the occurrence of interchromosomal or intrachromosomal rearrangements with a frequent pattern of chromothripsis or chromoplexy in 28 treatment-naive patients with malignant mesothelioma (MM), irrespective of the different tumor subtypes.1 The analysis was conducted by a mate-pair sequencing (MPseq)/RNA sequencing approach. Mansfield et al.1 identified chromosomal rearrangements associated with amplification and deletions of numerous chromosomal segments affecting several genes, including cyclin dependent kinase inhibitor 2A gene (CDKN2A), neurofibromin 2 gene (NF2), and BRCA1 associated protein 1 gene (BAP1) copy losses, as reported in previous studies of MM.
33179-4&id=gr1.jpg){kind=link}
33179-4&id=gr2.jpg){kind=link}
33179-4&id=gr3.jpg){kind=link}
33179-4&id=gr4.jpg){kind=link}