We leveraged a combination of two high-throughput next-generation sequencing technologies to define the alterations that occur at the nucleotide level in a single MPM tumor. The copy number variations, the SNV profiles, as well as the many rearrangements observed indicate that the number of alterations in this single tumor is quite substantial both in terms of polymorphism of the normal genome and in tumor-specific mutations. At the depth sequenced here, many more rearrangement type mutations than point mutations were identified and validated. This finding, in combination with a previously known propensity for aneuploidy in MPM, is consistent with the hypothesis that MPM is more likely to be associated with larger-scale chromosomal rearrangements than point mutations. This might explain why specific oncogenes or tumor suppressors have not yet been found for MPM. Obvious caveats are that this pilot work focused on tumor from a single patient and that the depth of sequencing may possibly be biased towards discovery of rearrangement-type mutations. However, data from other re-sequencing projects 
as well as from our transcriptome sequencing project 
support the notion that some solid tumors have fewer point mutations than others. Also consistent with previous work is the observation that all four point mutations previously discovered in the transcriptome of this tumor 
were due to a novel SNP present in one allele and a deletion in the complementary allele.
Rearrangements, translocations, deletions, insertions, and inversions have been associated with other cancers, particularly leukemias and sarcomas. Specific rearrangements have been clinically used for diagnosis and even to direct therapy as in the case of the BCR-ABL fusion protein, the target of the drug Gleevec 
. Given that most of the rearrangements we discovered involve chromosomes 8, 17, and 21, it is intriguing to speculate that a similar phenomenon may occur in MPM. At the very least, it highlights these areas of the genome as important for additional study.
Rearrangements have been implicated in many other solid cancers as well 
. Thus, the potential roles of rearrangements as oncogenic driver mutations and therapeutic targets in the solid tumor genome may approach that of somatic point mutations, at least for some tumors. However, unlike point mutations, rearrangements have been difficult to define precisely at the nucleotide level or to compare across many samples prior to the advent of massively parallel sequencing approaches. Therefore, the observations and the discovery methods developed herein support a role and provide a strategy for direct unbiased genomic sequencing for the precise identification, prioritization, and validation of rearrangements in solid tumors.
Limitations associated with individual massively parallel sequencing technologies include sensitivity versus specificity, the creation of artifacts, as well as the biases exhibited in certain stretches. It has been previously reported that each of the technologies used here has a large number of false positive artifacts suggesting that stand alone methods may not be sufficiently accurate to use independently for this purpose 
. Some biases that have been observed include the propensity for errors in homopolymer stretches, under representation bias against A/T rich sequence regions, and creation of rearrangement artifacts due to false primary priming during PCR expansion 
. While greater depth of sequencing and innovative filtering techniques may improve upon some of the limitations encountered in this analysis, artifacts inherent in the method of sequencing will not be ameliorated with deeper sequencing. Therefore, the most powerful rationale for ruling in alterations comes from a comparison and agreement of at least two orthologous technologies. The need for the combination of two different next-generation sequencing technologies is further supported by the recent findings of Pleasance et al. who reported deep sequencing of a small cell lung cancer cell line genome 
. These investigators identified 134 coding exon somatic mutations from among 22,910 somatic substitutions as well as 58 genomic rearrangements. At this number of putative mutations, rapid validation of all true mutations using traditional Sanger sequencing is not practical.
Although, clearly, more patient samples are needed to identify the most frequent genomic alterations that play a functional role in MPM, one may examine the data, postulate mechanisms of action, and make inferences relative to putative pathways involved in disease etiology. For example, many chemotherapeutic drugs are largely ineffective in the treatment of MPM, including methotrexate which targets the DHFR
gene product 
. The fact that there is a substantial amplification of the DHFR
locus in the tumor of this chemotherapy naïve patient may explain why 80% of MPM patients are resistant to methotrexate and similar drugs 
. The 30 validated tumor specific rearrangements were directly within 17 genes and were just upstream from the promoters of 35 other genes (see File S2
), most of which have not been previously described in cancer. However, several could plausibly be involved in oncogenesis as they are known to affect receptor signaling, signal transduction, cell proliferation, and apoptosis. Candidate genes for further study include a growth factor (NRG3
), two membrane receptors (EPHA6
), two signaling molecules (MAP2K6
), a putative transcription factor (TANC2
), as well as a potassium channel protein and modulator (KCNJ12
Particularly intriguing is the discovery that the absence of DPP10
expression in MPM is associated with poor clinical outcome, i.e., survival. DPP10
is located on the long arm of chromosome 2 (2q12.3–2q14.2), close to DPP4
and FAP, and it extends over 1 Mb of genomic DNA. In DPP10
, the serine residue critical to the active site of other DPP (dipeptidyl peptidase) family members is replaced by a glycine residue, such that DPP10
lacks dipeptidyl-peptidase activity 
binds specific voltage-gated potassium (K+) channels and alters their expression and biophysical properties 
. Alternate transcriptional DPP10
splice variants, encoding 2 different isoforms (i.e., the “long” and “short” isoforms), have been partially characterized within NCBI's RefSeq database. In all the MPM samples expressing DPP10
, only the short isoform has been detected (data not shown). Northern Blot analysis revealed strong DPP10
expression in brain, pancreas, spinal cord, and adrenal glands. Less expression was found in placenta, liver, and airways (trachea). Although Northern Blot analysis did not show DPP10
expression, RT-PCR detected low DPP10
expression in lung 
It is unknown what functional role DPP10
might play in MPM tumor cells, but if it modulates K+ channel function it could conceivably influence tumor cell survival or growth, as K+ channels are known in general to play important roles in regulating cell proliferation, cell cycle progression, and apoptosis 
has also been linked to asthma by several association studies of linkage and fine mapping 
. In addition, DPP10
is part of the “DPP-IV activity and/or structure homologues” (DASH) molecules, which have deregulated expression in multiple human cancers determining their pathobiological relevance in carcinogenesis.
Given the variable expression patterns, known and inferred physiological functions, and likely role of MAP2K6
, and MTAP
in tumor cells, it is a priority to determine whether the differential microarray gene expression patterns that were observed in a larger number (n
40) of additional MPM samples can be replicated in additional samples and linked to any of the genomic disruptions observed in the current study. Specifically, we are testing the hypothesis that MPM samples lacking expression of these genes harbor similar genomic disruptions. MAP2K6
is a kinase that phosphorylates p38 leading to apoptosis. Hence, loss of expression (i.e., function) would inhibit a cell death signaling mechanism. Given the level of aneuploidy in the MPM tumor, a mutation resulting in a Chr17:21 translocation that interrupted CCDC46
was notable. This gene product is likely involved in DNA replication/repair based on conserved protein domain analysis 
, and a lack of expression would likely promote tumorigenesis. In addition, mutations in this family of genes are known to be associated with gross chromosomal aberrations such as those evident in many MPM tumors, including the sample analyzed in this study. This mutation could potentially represent a driver mutation for the tumor given the incidence of rearrangement mutations 
. Finally, the function and physiological role (if any) of MTAP
alternatively spliced transcript is unknown, but it is frequently deleted in many types of cancer 
making its presence in 95% of MPMs (i.e., 38/40 tumors) an intriguing avenue for additional exploration.
In addition to rearrangements, we discovered three novel point mutations in three different genes (NKX6-2
is a transcription factor with known positive and negative regulatory activities in development and differentiation 
and has been postulated to be a tumor suppressor for some types of brain tumors (e.g., oligodendrogliomas) 
codes for cadherin, part of an integral membrane protein family affecting calcium-dependent cell adhesion. Inappropriate CDH8
expression has been linked to a subset of renal cell carcinomas 
codes for a DNA-binding protein that interacts with NFKB
1 and is a candidate oncogene currently being evaluated in hematologic and solid tumors 
This analysis demonstrates that there are many major, tumor-specific chromosomal rearrangements in this single MPM tumor and that clues discovered in the sequence of one tumor can lead to the discovery of genes that might be inactivated in other ways in related tumors. Ultimately, functional analysis, deeper sequencing, and genotyping in additional specimens will be required to identify true driver mutations in MPM and better define specific targets for therapy. We find that comparison of tumor to normal sequences for each case and using a second sequencing strategy, at least at the validation phase, are essential given the large number of polymorphisms and sequencing artifacts observed in MPM. Even now, the cost of sequencing and bioinformatics analysis such as described herein is too costly for widespread clinical application. However, third generation sequencing technologies are becoming available at more affordable prices. We envision that within the next few years the high-throughput analysis of multiple cancer genomes will become a reality and ultimately part of the clinical diagnostic pipeline.