By Monica Heger
Over the last two years, researchers have shown that whole-genome sequencing can identify numerous mutations in cancer cell lines and tumors, and several large-scale projects, including the National Institutes of Health’s Cancer Genome Atlas, are now using next-generation sequencing to characterize the genomes of more than 20 different types of cancer.
But some researchers are questioning the clinical utility of such projects.
Last week at the Cambridge Healthtech Institute's XGen Congress meeting in San Diego, several researchers, among them Larry Loeb, a professor of biochemistry and pathology at the University of Washington, said in presentations and a roundtable discussion that until significant improvements are made in sequencing technology, sequencing cancer genomes would not yield the kind of data that would allow researchers to develop new therapies.
Others, like Nicole Cloonan, senior research officer in Sean Grimmond's lab at the Queensland Centre for Medical Genomics at the University of Queensland in Australia, think the sequencing projects are worth the effort and resources. However, Cloonan and the Grimmond team, whose work is part of the International Cancer Genome Consortium, are sequencing not only the cancer genome, but also the transcriptome and methylome, and trying to look for relevant pathways as opposed to specific genes and mutations.
At the conference, University of Washington's Loeb said that cancer is too heterogeneous for sequencing to uncover therapeutically relevant mutations. "Sequencing large numbers of cancer genomes is not the best use of the technology," he said.
Loeb pointed to the sequencing of an acute myeloid leukemia tumor by Washington University researchers in 2008 as an example of cancer's complexity and the failure of whole-genome sequencing to uncover clinically relevant data.
For instance, although that study identified eight mutated genes, the Wash U team was unable to find any of them in 187 other patients tested (see In Sequence 11/11/2008).
Loeb cited the heterogeneity of cancer and the ubiquity of random mutations in tumor cells as the major reasons why sequencing large numbers of cancer genomes is not a fruitful endeavor.
"There are thousands of mutations all over the place, and you're not going to find new drug targets," Loeb said. Even among individuals with the same type of cancer, it can manifest itself in different ways. "We're just sitting on the tip of the iceberg."
Right now, researchers are only able to detect clonal mutations in a tumor, he said, but there are also subclonal mutations — mutations found in a minority of tumor cells — and "deeper down the iceberg are random mutations," he added. Loeb said it is the random mutations that make up the majority of mutations in a cancer tumor.
Accurate single-molecule sequencing will be able to detect the random mutations, he said. But until that technology is demonstrated, sequencing tumor genomes will have very limited clinical use, said Loeb.
In a roundtable discussion at the meeting, Harris Lewin, director of the Institute for Genomic Biology at the University of Illinois, said that the data from current cancer genome-sequencing studies was not providing the clinically relevant information that researchers initially hoped — although a couple of cancer sequencing studies have identified potential biomarkers (see In Sequence 9/9/2008 and 2/23/2010) — but thought the problem could be fixed.
Lewin said the main problems are that current sequencing technology is not able to find large rearrangements, which he said will be more informative about cancer progression than other types of mutations like SNPs, and there is not a good map against which to align cancer genomes.
While some researchers have detected rearrangements, such as the group from the Wellcome Trust Sanger Institute, a team from Johns Hopkins, and another research group from UCLA, Lewin said large rearrangements of highly repeated areas were still not being detected.
Lewin said longer, accurate reads — such as a scaffold that spanned an entire chromosome, or reads several kilobases in length — as well as a comprehensive map, would help make sense of the data and also be able to detect the large rearrangements. "Longer reads may allow us to assemble long enough scaffolds to get chromosomal maps," said Lewin. "But until we get that, we'll miss the details of the evolution [of cancer]."
He said that the paired-end and mate-pair sequencing strategies were missing the big segmental duplications. He thought for cancer, it might even be necessary to have individual maps against which to align cancer genomes so you could see which mutations contribute to the disease.
Graeme Suthers, the chair of the Genetics Advisory Committee at the Royal College of Pathologists of Australasia thought that sequencing formalin-fixed cancer tissue samples would have more relevance than sequencing tumors from current patients.
With formalin-fixed samples, there is a history of patient data, he said, so you can do a retrospective study comparing the sequence with actual patient outcomes. He said that information could help determine if there are mutations or genes that influence responsiveness to treatment or survival rates.
As for sequencing clinical samples, "we haven't demonstrated that what we observe carries any relevance in the clinical world," he said.
Meanwhile, other researchers such as the University of Queensland's Cloonan, think that sequencing not just the cancer genome, but also its transcriptome and methylome will help in pinpointing the pathways involved instead of specific mutations.
The University of Queensland team is sequencing the genomes, transcriptomes, and methylomes of 500 tumor/normal pairs, including 350 pancreatic and 150 ovarian cancer samples, on Life Technologies' SOLiD system. The study is part of the International Cancer Genome Consortium. So far, they have completed 10 samples, and are in the process of analyzing two of them.
"The ultimate goal" is to figure out "what treatment will work best for you, given your mutational load," Cloonan said.
For the whole-genome sequencing aspect, the team is using a 50-base mate-paired sequencing strategy, and achieving about 25-fold coverage of both the tumor and matched normal genomes.
For the methyl-seq, they are using Invitrogen's Methyl Miner kit to capture the methylated portions of the DNA, and using a 50-base fragment read sequencing protocol, generating around 20 million reads per sample.
The team is also using a 50-base fragment read sequencing strategy for the transcriptome sequencing portion, as well as testing SOLiD's paired-end sequencing. They are sequencing both RNA and microRNAs and generating about 100 million reads per sample.
"We're taking a more holistic approach," said Cloonan. "At the end of the day, there's not going to be one gene that is going to cure cancer." Instead, she said the group is more interested in identifying the common pathways. "We want to determine what pathways are activated or repressed," she said.
"We're collapsing data into pathways rather than looking at individual genes," which may not be uniformly mutated even in patients with the same type of cancer, Cloonan added. She thinks that looking at the pathways involved will help predict how effective a specific drug regimen will be, and she is also hoping to find novel pathways and biomarkers for new drug targets.
While there was disagreement over the utility of sequencing cancer genomes, the researchers who presented last week did agree that single-molecule sequencing would be a valuable tool, and would help make cancer sequencing clinically relevant. Cloonan said that single-molecule sequencing will allow you to look more deeply at cancer tumors and detect mutations more accurately, but, she added, "that's not to say there's not value in what we're doing now."