Researchers at the Scripps Translational Science Institute are investigating a range of technical issues involved in using next-generation sequencing to help guide therapy for cancer patients — including the pros and cons of exome sequencing over whole-genome sequencing, as well as the impact of tumor heterogeneity on sequencing results.
The group recently completed a single-patient pilot study that involved exome and whole-genome sequencing of tumor and normal samples from a 27-year-old patient with stage 4 metastatic colorectal cancer and plans to expand its efforts into a larger pilot that will involve at least 10 patients, Samuel Levy, director of genomic sciences at STSI, told Clinical Sequencing News.
The goal, he said, is to develop a sequencing approach that is "both comprehensive and accurate, which is actually quite a challenge." While there are numerous logistical, ethical, and regulatory issues associated with the adoption of sequencing for clinical applications, Levy said that his team is focused for now on addressing the "practical aspects of applying sequencing that go above and beyond a genomics project."
Next-generation sequencing "is emerging more and more as an approach to supplement how patients [with cancer] are treated, so we wanted to try and iron out the technical aspects of it … so we can really feel confident that as a methodology, it works," he said.
Levy discussed the single-patient colorectal cancer study at the Scripps Future of Genomic Medicine Conference last month. It was already known that the patient had a KRAS mutation that ruled out treatment with cetuximab (ImClone's and Bristol-Myers Squibb's Erbitux), so the Scripps team, in collaboration with the patient's oncologist, decided to use sequencing to determine whether there might have been other mutations that would assist in determining a treatment strategy.
Scripps carried out whole-genome sequencing to 40x coverage of the patient's tumor and blood DNA and also used Agilent's SureSelect technology to perform exome sequencing at 300X coverage. The high depth of the exome sequencing was due to the fact that the group was trying out a new Illumina HiSeq and thought it would be a good idea to dedicate a whole lane to one exome, Levy said in an interview following the conference.
A comparison of the two data sets revealed that both methods missed a handful of variants in protein-coding regions — the whole-genome approach identified 35 single nucleotide variants within exons and the exome sequencing identified about 69, of which only 28 overlapped.
Levy said that the whole-genome approach likely missed some mutations because it wasn't deep enough, even at 40x coverage. Most filtering methods require a minimum number of reads to support a variant call, and the fact that some regions of the genome had lower coverage, combined with the heterogeneous nature of tumor tissue, could mean that there simply weren't enough reads at the given position to make the call for some mutations, Levy said.
In terms of the exome data, Levy said that if the target-capture step doesn't capture every region with equal efficiency, it will likely miss some mutations, regardless of the depth of coverage. For example, if there are significant rearrangements in the tumor genome, it's possible that the capture probes wouldn't be able to pick up those regions.
"Each technique has its biases," Levy said. "Clearly, the loci that are missed in either approach are basically due to insufficient data, or data that was on the edge of being on a threshold for us to be able to call a variant or not."
Ideally, he said, "the global approach is what we would like to adopt. That would mean whole-genome sequencing, no capture methodology, and keep it as least biased as possible." On the other hand, he said, exome sequencing wins out in terms of cost and data-analysis requirements.
One drawback with exome sequencing, however, is that current sequencing platforms are designed for high-throughput whole-genome sequencing and aren't tailored to sequence one exome at a time cost-effectively. "You can easily sequence from 8 exomes to 64 exomes, depending on what sequencing platform you have and how you want to barcode samples, but none of them efficiently, at the same cost, sequence a single sample," Levy said.
The alternative is for labs to wait until they get enough samples to load up a full sequencing run, "but that doesn't help the patient," he noted.
Levy said that he and his colleagues are still evaluating the advantages and drawbacks of each approach. "These technologies will differ in the way in which they work and the consistency with which they produce results," he said. "But what we really want to understand is, 'Does it affect the outcome?'"
Work on answering this question is still ongoing.
Sequencing the tumor genome of the colorectal cancer patient revealed a number of mutations in genes that are known to be associated with cancer, and for which there are inhibitors either on the market or in clinical trials, including PIK3CA, MEK1, and MEK2.
Based on these findings, the researchers informed the patient's oncologist that one possible course of action would be to use targeted MEK inhibitors in parallel with PIK3CA inhibitors. As it turned out, the patient had already been placed on the standard course of treatment for KRAS mutation-positive colorectal cancer patients, which is to avoid cetuximab and move directly into chemotherapy with the so-called Folfox cocktail and then follow that up with bevacizumab (Genetech's Avastin).
The Scripps team also found that the patient had a germline mutation in the ERCC1 gene that indicated he would respond well to chemotherapy, but there would likely be onset of neuropathy.
This prediction was ultimately validated. The patient responded to the treatment, but suffered from mild neuropathy, Levy said.
[ pagebreak ]
Even though sequencing did not impact the actual course of treatment for this first patient, Levy noted that it was reassuring to know that the data would have been able to provide a better idea of what to expect from the chemotherapy — a benefit of genome sequencing that some have overlooked in favor of identifying potential targets for protein inhibitors.
"All of us in the field are looking for targeted drugs — this is the gene that's upregulated, so this is the protein you need to target — but most people are being treated with chemotherapy for the most part, and having just a better way of applying that tool would actually be very useful from a genomics perspective," he said.
"What we'd like to be able to say is, 'You may have an adverse physical response to this, and you may even have neuropathy, but guess what? It's going to benefit you.' That is a very different thing than not providing reasons why we think therapies will work."
Levy said that in a separate project, the Scripps team is working with another oncologist to determine whether it will be possible to use sequencing data to guide treatment with chemotherapy. "Chemotherapy cocktails are also specific inhibitors," he noted. "They may inhibit many more targets, but they are still specific to some degree. So having an understanding of how to apply those would actually be very useful."
The Scripps researchers are also examining the effect of tumor heterogeneity on sequencing data. Since tumors exhibit a high degree of variability, researchers know that all the DNA collected from a tumor isn't from mutated cells, but they don't know the exact proportion of these mutated cells within any given sample, and how that might impact sequencing results.
In order to get a better handle on the extent of mutational change across a tumor, Levy and colleagues sampled and sequenced six different loci within a single colon adenocarcinoma. They found that each locus harbored different numbers of mutations, with some regions of the tumor exhibiting many unique variants and the others bearing relatively few.
Levy said that his team is hoping to extend this work into additional tumor types in order to develop a sampling strategy that captures the most relevant mutations in a given tumor. One aspect of the study, he noted, is to "determine whether the ability to provide a molecular profile changes if you sample in different locations, and by how much." More importantly, the researchers want to determine whether these mutational differences would have any impact on the recommended course of treatment for cancer patients.
The Scripps researchers would also like to use this information to determine how many reads are necessary to call a variant based on the ratio of tumor and normal tissue in a given region of a tumor. If they can get a better handle on the fraction of DNA from a given sample that comes from mutated cells as opposed to normal cells, "you might be able to tailor how you apply these filters of what fraction of reads you'd expect to be mutated," Levy said.
As the Scripps researchers look to scale up their efforts to sequence more cancer patients, Levy said they expect to discover additional issues that will need to be addressed. In the meantime, they plan to apply a range of methods for each patient, including exome sequencing, whole-genome sequencing, and SNP-based genotyping.
"Depending on the requirements for the patient and what the oncologist sees as being important in terms of time, we're adapting our method," he said.
One area the group is interested in exploring is the impact of whole-genome data on treatment decisions.
"We went from testing for certain genes that are frequently mutated to looking at everything, and the question is, 'How does learning all the mutations really help?' I think that's going to be the next big challenge," he said.
"Coming up with a strategy that would really leverage that information to help with treatment is partly science and it's partly driven by the clinical requirement, and I think that's a wonderful place for us to be working right now," he added.
Have topics you'd like to see covered in Clinical Sequencing News? Contact the editor at btoner [at] genomeweb [.] com.