CHICAGO — Two labs that have been using Sequenom's MassArray instrument for cancer genotyping of formalin-fixed, paraffin-embedded samples reported this week that they have conducted comparison studies with next-generation sequencing panels and found the results to be concordant.
Both groups — the Diagnostic Molecular Pathology Laboratory at Memorial Sloan-Kettering Cancer Institute and the Knight Diagnostic Laboratories at Oregon Health and Sciences University — have been running custom cancer genotyping panels on the mass spectrometry-based MassArray platform since 2009. In presentations at this week's annual meeting of the American Association for Cancer Research, representatives from the labs discussed initial comparisons of the Sequenom panels with NGS-based systems.
In MSKCC's case, the team used Foundation Medicine's 189-gene sequencing test, which the company runs on the Illumina HiSeq platform. OHSU, meantime, used Life Technologies' Ion Torrent PGM and AmpliSeq Cancer panel. Both groups looked at FFPE samples, which typically pose a challenge for sequencing-based studies because the fixing process damages DNA.
Neither lab compared the costs of NGS-based testing against the Sequenom panels, however; nor do they have immediate plans to switch platforms.
Nevertheless, the early results from both groups suggest that next-gen sequencing can not only provide results that are as good as existing testing platforms, but can also provide information about additional mutations that may be useful for clinical decision-making.
MassArray vs HiSeq
Michael Berger of MSKCC said that his lab's MassArray cancer genotyping panel includes 91 oncogenic mutations in eight oncogenes and that the center has used it to profile more than 6,000 tumors since implementing it in 2009.
He said that the lab was interested in seeing if a sequencing-based approach could improve upon genotyping.
"In theory," sequencing is a promising alternative, he said, because it has the ability to detect mutations "across entire exons rather than genotyping specific precompiled point mutations and hot spots," and could also provide information on tumor suppressor genes, copy number alterations, and structural rearrangements. Furthermore, the high coverage of Foundation's approach — it sequences to between 700x and 1000x — should have good sensitivity for low allele frequency alterations.
However, he noted, "before using next-generation sequencing clinically, one needs to demonstrate the accuracy of the method and its compatibility with formalin-fixed, paraffin-embedded tissue, which is routinely encountered in clinical labs."
For the comparison, which MSKCC conducted as part of a collaboration with Foundation Medicine, the team used 71 surgically resected formalin-fixed, paraffin-embedded tumors that had previously been genotyped with the Sequenom panel. These included 42 lung tumors, 26 colorectal cancer samples, and three melanoma tumors.
They used at least 250 ng of DNA per case and performed sequencing library construction and hybridization-based capture of 3,230 exons and 37 intronic intervals per sample. Foundation sequenced on the HiSeq to average coverage of more than 750x and then analyzed the data for single nucleotide variants and small insertions and deletions.
Berger said the team saw "high concordance" between Sequenom and HiSeq. The two platforms had 60 mutation calls in common, while Sequenom had two unique calls and Foundation had six.
Berger said these "occasional discrepancies" could have been due to different sensitivities of the two technologies, but intratumor heterogeneity may have also played a role. In one case, the Sequenom assay detected an exon 19 deletion in EGFR, but there were only 10 sequencing reads out of 1,200 that contained that deletion, which is below the 1 percent threshold that Foundation uses for reporting, Berger said. Likewise, sequencing picked up a dinucleotide substitution in KRAS, but the Sequenom assay is not designed to detect that type of mutation.
There was "very high sensitivity" with both methods for low-frequency mutations, he said. Mutant allele frequencies in concordant calls were as low as 4 percent with both platforms.
Furthermore, sequencing revealed a number of mutations that the Sequenom assay does not test, including 73 somatic mutations and 30 loss-of-function variants in a range of tumor suppressor genes. "All of these we think may hold possible implications for the selection of approved or experimental targeted therapies for these patients," Berger said.
Berger told Clinical Sequencing News that MSKCC's plans regarding next-generation sequencing for clinical use are "still under development."
MassArray vs. Ion PGM
Like MSKCC, OHSU implemented Sequenom-based genotyping of tumor samples in 2009, explained Christopher Corless, medical director of the university's Knight Diagnostic Laboratories.
The lab offers two Sequenom panels: A leukemia panel that screens for 370 mutations in 31 genes, which is "run on every leukemia case that comes into [the] university," and a newer solid tumor panel that tests for 643 mutations in 53 genes. The lab currently runs around 1,200 cases per year on the MassArray platform, he said.
While the Sequenom technology is "robust" and works well with FFPE samples, one drawback is that "it can only go after hotspots," Corless said, so his group decided to test the Ion PGM in an effort to find something "that will allow us to go more deeply and across a greater spectrum of genes."
OHSU took delivery of the system last spring but was initially frustrated by the NimbleGen hybridization capture for FFPE samples.
"We found that it was a little tricky," he said. "You have to shear the DNA as part of the hybridization process and some samples in paraffin blocks are of poor quality, so you have to adjust the shearing to get the right-sized fragments."
The lab found this process to be "a real pain," he said, because it could take up to four days of "hard labor" before the samples were even ready to sequence — "not fast enough" for clinical use.
In order to address this, the OHSU team began collaborating last fall with Ion Torrent on the company's AmpliSeq Cancer panel, a single-tube primer pool that targets 190 amplicons in oncogenes and tumor suppressor genes, including whole-exon coverage of KRAS, BRAF, and EGFR.
Corless said that the panel allows the lab to get from sample extraction to final sequence in 48 hours, "which is what we'd like to see in a clinical setting." The protocol also requires only 10 ng of starting material, which allows the group to go right from laser microscopy microdissection to sequencing.
In a validation study of 45 FFPE tumor samples that were previously run on the Sequenom platform, the results were "comparable" across the two platforms, which was "encouraging," given the "entirely different" primer sets and platforms, he said.
Using a four-plex format with the Ion 316 chip, the PGM detected all 53 known mutations included in the Sequenom assay, as well as 26 new mutations in known oncogenes. However, of the 19 known indels in the samples, the PGM called two exactly, flagged five as point mutations, and missed 12 completely.
As a result, Corless noted that with the current generation of Ion's software, "if you are concerned about the possibility of a deletion in a specific region or a specific gene, you need to go in and look at it manually to make sure it's not missed." He noted, however, that this is not "a big surprise because indels are a problem for virtually all the next-gen approaches."
In addition, there were a few "aberrations," such as 54 "variants" that were called in both tumor and normal samples, though Corless attributed this to the PGM's known "struggle" with homopolymers and said that the next version of the software will filter such calls out.
The OHSU lab ultimately set the lower limit of detection of the software at 8 percent mutant allele frequency, which gave it 100 percent sensitivity for the 53 point mutations in the assay, and 94 percent specificity. He noted that this threshold worked well with the tumors in the study, which had a minimum of 40 percent tumor content, but said he wouldn't be comfortable using samples with 5 percent tumor content.
While the AmpliSeq Cancer panel identified more mutations than the Sequenom panel, it's still just a "hotspot" approach, Corless said. The group has ordered a custom panel from Ion Torrent with overlapping amplicons that will provide more comprehensive coverage, he said.
Like Berger, Corless said his lab is still considering its options going forward, but said it's likely that it will replace the MassArray with a next-gen sequencing platform some time in the next year.
Have topics you'd like to see covered in Clinical Sequencing News? Contact the editor at btoner [at] genomeweb [.] com.