By Monica Heger
This article was originally published July 30.
Researchers from Genentech, Affymetrix, and Pennsylvania State University have used a variety of methods, including mismatch repair detection technology, sequencing on Roche's 454 GS FLX, and array-CGH, to identify more than 2,500 somatic mutations in approximately 1,500 coding genes from 441 breast, lung, ovarian, and prostate cancer tumors.
Of the 2,576 somatic mutations they identified, 95 percent are novel. They also found 37 recurrent mutations in 13 different genes.
In the study, published last week in Nature, the Genentech team also reported a number of mutations in known drug targets such as protein kinases and G-protein-coupled receptors. Combining their analysis of somatic mutations with copy number analysis allowed them to identify possible tumor suppressors and oncogenes.
Somasekar Seshagiri, senior scientist in molecular biology at Genentech and the study's senior author, attributed the fact that they found so many novel mutations to the sheer number of tumors they analyzed.
"Studies published so far either had a small number of tumors analyzed for presence of mutations in the 20,000 coding genes or they looked at a larger number of [tumors] and sequenced a smaller number of genes," he said. "We have looked at a reasonably larger number of tumors, covering different tumor types for mutation in 1,507 genes."
Nallasivam Palanisamy, an assistant professor at the University of Michigan who has used transcriptome sequencing to study tumors (IS 6/8/2010), said that the fact that the group found so many novel mutations was striking. "It really underscores the power of this technology," he said. With mismatch repair detection, "you are directly cloning the gene of interest, so the sensitivity of finding mutations is higher."
Mismatch repair detection uses the DNA repair machinery of Escherichia coli to detect mismatches in human DNA with high specificity and sensitivity. The technique can detect variations that are as low as 1 percent in abundance, according to the authors. In the study, Sesahgiri's team used a tiling array from Affymetrix to read the variants identified by the mismatch repair detection technique, but he said that the mismatch repair products could easily be sequenced.
Seshagiri likened MRD to exon capture, except that the technique also fixes and then amplifies the variants, making it more likely that they will be correctly identified. He said that MRD combined with sequencing may even yield better results than exon capture and sequencing, although the two methods have not been compared side by side.
Palanisamy agreed that MRD plus sequencing would enrich for mutations, making them easier to call. However, he said it was unclear whether sequencing would need to be done following MRD. In cases where it is "difficult to make the calls for some of the mutations detected by MRD then sequencing would help to enrich the target generated by MRD and, depending on the coverage, mutations can be easily identified," he said.
However, he added, until the methods are compared directly, it will be difficult to tell whether "it is always necessary to combine MRD with additional methods."
Seshagiri added that the combination of technologies, and particularly the addition of array-CGH to detect copy number variation, allowed them to pick up mutations that may have been missed by using just one technology alone.
"Maybe neither the mutational frequency nor the amplification is high, but combined they become relevant and important in the context of cancer," Seshagiri said.
The addition of array-CGH also allowed the team to distinguish between potential oncogenes and potential tumor suppressors. "If genes are deleted and mutated they tend to be tumor suppressors, and if they're amplified and mutated, they tend to be oncogenes," he said.
The team also used 454 sequencing following PCR amplification of coding exons to analyze 18 genes in 50 additional lung squamous carcinoma samples. The sequencing validated the mismatch repair detection, and also uncovered an additional 56 somatic mutations. The mutation frequency was comparable to the original screen, including the novel cancer genes the team identified.
The Genentech team focused the study on genes with known or suspected roles in cancer. "We focused in on things we could target and things on which we could build therapeutics," Seshagiri said.
Interestingly, they identified a number of G-protein-coupled receptors, which are known drug targets, but had not previously been suspected to play a large role in cancer. They identified 87 mutated genes coding for G-protein-coupled receptors, out of 156 that they analyzed.
"Given the frequency of mutations that we're finding, they deserve a second look," Seshagiri said. Although he cautioned that first, "the biological roles of these mutations will have to be elucidated."
The team also found a number of mutated protein kinases, which are also druggable. Out of the 230 protein kinases they studied, they found 315 mutations in 157 of them.
While the paper identified a number of potential drug targets, it also added more evidence to the heterogeneity of tumors. Among the druggable mutations, each was present in only a small proportion of the samples, suggesting that numerous, individualized drugs will be needed to treat most cancers.
"The diverse spectrum of mutational and genomic changes from large-scale sequencing studies, including this, show that each tumor is unique even within a given type and subtype," the authors wrote.