NEW YORK (GenomeWeb News) – In the early, online version of the Proceedings of the National Academy of Sciences last night, researchers from the University of Washington and the Fred Hutchinson Cancer Research Center described the approach that they used to sequence and find new mutations in the exomes of nearly two-dozen advanced prostate cancers.
The researchers tracked down the new coding mutations — around 200 per exome, on average — by comparing the exomes of advanced prostate cancers grown as mouse xenografts with germline mutations found in large sets of genome and exome sequences. In a few of the samples, researchers found thousands of non-synonymous coding changes, a hyper-mutation pattern that they suspect may stem from mutations to DNA repair genes in the affected tumors.
Despite the collection of mutations detected in the exomes, though, the team found relatively few genes that were frequently mutated in the prostate cancers tested.
"[O]ur results indicate that point mutations arising in coding regions of advanced prostate cancers are common but, with notable exceptions, very few genes are mutated in a substantial fraction of tumors," senior author Jay Shendure, a genome sciences researcher at the University of Washington, and co-authors wrote.
Some genome-wide studies of prostate cancer have been done in the past, including a study published in Nature in February by researchers from the Broad Institute and elsewhere who did whole-genome sequencing of tumor-normal pairs from seven individuals to find mutations and genetic rearrangements in primary prostate cancers.
For the current study, researchers focused on more advanced prostate cancers, using a Roche NimbleGen solution-based hybrid approach to nab coding sequences from 23 prostate cancer tumors grown in mice xenograft models.
"The idea was to find mutations in prostate cancer, including those that might be involved in metastasis itself," Shendure told GenomeWeb Daily News, explaining that advanced cancers are generally not as well studied as primary tumors from a genomics standpoint.
The samples originated from 16 fatal tumors that metastasized and three high-grade primary carcinomas. The human tumors were grown in immunocompromised mice — a system that made it possible to generate relatively pure samples of these tumors in a system where contaminant sequences mostly come from mice.
Using the Illumina GAIIx or HiSeq platform, the researchers generated sequence covering 26.6 million bases of coding sequence to an average of 100 times coverage for eight samples and 36.6 million bases of coding sequence at 140 times depth, on average, in the remaining 15 samples.
Since matched normal tissues were not available for most of the prostate cancers tested, the team compared the tumor exomes to 1,000 Genomes pilot data and sequence data for around 2,000 exomes sequenced at the University of Washington, removing sequence variants found in these sequences to get a collection of private germline variants and somatic mutations for the tumor exomes.
"We didn't have access to normal tissues for the vast majority of these xenografts," Shendure said. "So we speculated that databases of germline variations had gotten sufficiently deep at this point that we could take advantage of them to filter out the vast majority of germline variation, thereby enriching for somatic variation in what was left."
Most of the non-synonymous coding changes detected in the prostate cancer exomes were tumor specific. Although there were some genes that were mutated in multiple tumor samples, including GPC6, DLK2, SDF4, and the p53 tumor suppressor gene TP53, the researchers found that the frequency of these mutations was typically quite low.
"There don't appear to be genes, at least in the set we're targeting, that are recurrently mutated to high frequency, with the possible exception of p53," Shendure said. "We do see other genes that, at least by our preliminary analyses, appear to be mutated at rates above background," he said, but added that determining whether those mutations are recurrent will require replication in follow-up studies.
In three of the exomes, though, researchers found far more mutations: between 2,700 and more than 4,000 non-synonymous coding changes.
For one of these hyper-mutated tumors — which contained a preponderance of guanine to adenine and cytosine to thymine transitions — the team directly compared xenograft and patient metastasis samples, demonstrating that the excess coding mutations were present prior to xenografting.
"Our initial concern was that this was something that happened while the tumor was passaging through mice," Shendure said. "But, in fact, it clearly was associated with the tumor itself."
While more research will be needed to explore the cause and consequences of this hyper-mutation, the team speculated that it might be due to mutations in DNA polymerase or repair genes — a notion supported by the observation that some of these samples harbored mutations to DNA repair genes such as MSH6.
The researchers suspect that while the hyper-mutated tumors contain more point mutations than other prostate cancers, they might have fewer genetic rearrangements, Shendure noted, though that too requires further follow-up study.
By examining the mutation patterns in tumor pairs derived from the same individual and grown as xenografts in mice subjected to castration therapy, which deprives the tumor of the male androgen hormone, the team also found clues about genes and pathways that render prostate cancers castration-resistant.
In particular, they found that mutations to genes in the Wnt signaling pathway seemed to coincide with the switch from tumors that were castration-sensitive to those that had evolved to grow without androgen.
The researchers are in the process of doing additional studies to learn more about the mutations involved in advanced stages of prostate cancer, including targeted re-sequencing studies focusing on genes implicated in the current study in a large set of prostate cancer metastases.