This story was originally published May 15.
Testing the waters for how next-generation sequencing could be used to guide individual therapy selection in cancer, researchers at Thomas Jefferson University have conducted a proof-of-principle study in colorectal adenocarcinoma using the 454 GS Junior.
The study "demonstrates the power and feasibility of this approach for practical aspects of diagnosis," said Stephen Peiper, chair of the department of pathology, anatomy, and cell biology at the university, who presented some results during a webinar on next-generation sequencing and translational research last week. The webinar was organized by the journal Science and sponsored by Roche and 454.
While challenges remain, "mutational profiling by next-generation sequencing is ideally suited for the diagnostics of the future, including the potential selection of combinations of targeted therapies," he said.
Following the pilot study, his team will begin setting up next-gen sequencing in the hospital in July in order to validate the results, prior to offering sequencing panels clinically, he told Clinical Sequencing News. While Peiper declined to disclose whether the lab is considering platforms other than the GS Junior for use in the clinic, he noted that "automation and IT/bioinformatics, as well as reagent costs, are critical" in this decision.
Existing companion diagnostics for cancer drugs, both those that have been approved by the US Food and Drug Administration and laboratory-developed tests, have single targets and use in situ hybridization, immunohistochemistry, real-time PCR, or Sanger sequencing. "My prediction is that many of these tests will be rolled into next-generation sequencing platforms," Peiper said during the webinar.
Such high-throughput profiles would be able to "detect more complex mutational signatures where we may show eligibility for combinations of experimental therapies, or show tumor variation," he said.
To test the feasibility of next-gen sequencing as a diagnostic, his team developed a gene panel for colorectal adenocarcinoma. The panel includes 275 exons from 51 genes that represent mutational hotspots in this type of cancer and was composed using existing companion diagnostics, codified guidelines, and other existing tests that have clinical and diagnostic utility.
For example, it includes certain exons of four members of the EGF receptor signaling pathway — KRAS, BRAF, NRAS, and PIK3CA — mutations in which predict poor response to anti-EGF receptor therapy.
It also contains all exons of four DNA mismatch repair genes, MLH1, PMS2, MSH2, and MSH6, that are associated with Lynch syndrome, a subset of hereditary non-polyposis colorectal cancer. These genes were included because a professional group has recommended testing for Lynch syndrome in all patients diagnosed with colorectal cancer. While a positive result in the tumor will not replace a diagnosis by testing a germline sample in a CLIA lab, the panel could serve as a "very efficient way" for prescreening patients for mutations in those genes, Peiper said.
The panel further includes regions of other genes that are commonly mutated in colorectal cancer, such as P53, APC, and RET.
The proof-of-principle study was designed to show the ability of next-gen sequencing to detect known mutations in KRAS, BRAF, and DNA mismatch repair genes, Peiper told CSN, and involved at least nine cases.
Gene panels are preferable over exome sequencing, Peiper said during the webinar, because the coverage is higher and because it makes more sense to focus on actionable mutations than to generate a lot of data and then only analyze some of it.
Using whole-genome sequencing for clinical tests, he believes, is still far off. "Even if it becomes cheap, I think the bioinformatics takes a significant amount of time, and again, there is the curation and generation of a great deal of data that we don't know what to do with," he said.
To enrich the genes for the proof-of-principle study, his lab used Roche NimbleGen in-solution capture. It then sequenced them on the 454 GS Junior, for which they were early-access users. For FFPE samples, they sequenced with 200-base reads, and for fresh tissue, they used reads between 400 and 450 bases.
The long reads of the 454 platform made the transition from Sanger sequencing easier, Peiper said. "It gave us more of a sense of what we are sequencing, rather than having a black box and getting bioinformatics back."
Other criteria they used in choosing the platform was easy bioinformatics and cost. Going forward, sample prep automation will also be "a critical issue" for platform choice, he said.
The DNA to be tested was isolated from four to six 10-micrometer sections from formalin-fixed paraffin-embedded samples. Peiper said it is critical that tissue areas enriched in tumor cells are selected by a pathologist. The DNA yield varies depending on the tumor tissue size, with two slides of a tumor-enriched area of 0.5 cm2 to 1 cm2 providing between 1 and 5 micrograms.
In the future, he and his colleagues will increasingly sample tumors at different sites in order to be able to detect tumor heterogeneity, he said. Guidelines for doing that still need to be established; for example, if immunohistochemistry indicates great tumor heterogeneity, more tumor sites would be sequenced.
The turnaround time for the test, from obtaining the sample to getting data, is about a week. While this is longer than Sanger sequencing, he said that the wait is not an issue because patients frequently need to recover from surgery before they start therapy. With added sample prep automation, the test will likely be able to meet the turnaround time required "without significant problems," he said.
The largest cost of the test right now is consumables, about $1,500 per sample including quality control, exon capture, and sequencing.
The researchers tested their panel on patient samples that had previously been tested for KRAS and BRAF mutations in a clinical lab and were able to detect all mutations by next-gen sequencing that had been identified in the diagnostic lab.
Of note, in one sample, immunohistochemistry detected a deficiency in the mismatch repair proteins but next-gen sequencing found wildtype sequences for all four mismatch repair genes, suggesting that the defect results from epigenetic silencing.
The team did not report any results from the proof-of-principle study to patients and their physicians, since all samples were de-identified. For clinical use, they will consent patients for the portion of the test "that is not front-line care," Peiper told CSN. "One logical approach is to offer the option for being informed if there is a relevant treatment option based on the testing."
In the future, Peiper said during the webinar, next-gen sequencing will likely be one of several tools in the diagnosis of cancer. "The final classification will include both pathologic and genomic features, including a mutational signature, pathway activation, some indication of target availability and selection, as well as tumor heterogeneity."
And while next-gen sequencing will likely "replace a lot of other molecular technologies" because it is getting cheaper and more powerful, he noted that other methods will not disappear. Immunohistochemistry, for example, is "pretty cost-effective" and enables researchers to look at individual cells, so it is likely going to remain.
Validating next-gen sequencing tests for clinical use will remain a bottleneck, he said, and "is going to take much more work than the current generation of molecular tests." Because the technology is error-prone, there needs to be discussion, for example, on what accuracy is required and what coverage is acceptable.