Memorial Sloan-Kettering Cancer Center's diagnostic molecular pathology laboratory is developing a number of targeted sequencing assays to test for actionable cancer mutations and plans to run the first patient samples early next year, following certification by the New York State Department of Health.
The first assay will focus on solid tumors. It will use Illumina's TruSeq Amplicon Cancer Panel, which includes 212 amplicons representing mutational hotspots in 48 genes, including BRAF, KRAS, and EGFR, and will run on the Illumina MiSeq.
In addition, the lab is working on an assay for acute myeloid leukemia and myeloproliferative neoplasms, combining RainDance Technologies' multiplex PCR platform and Illumina's MiSeq to analyze 30 genes.
By next fall, the lab also plans to develop a large panel of several hundred cancer genes, an assay that will use hybridization capture and the Illumina HiSeq 2500.
Cyrus Hedvat, director of the diagnostic molecular pathology lab at MSKCC, told Clinical Sequencing News that for almost two years, his lab has been using the Sequenom genotyping platform to assay about 60 actionable hotspot mutations in eight cancer genes – including BRAF, KRAS, EGFR, and PIK3CA − in patients with solid tumors, primarily lung cancer, colon cancer, and melanoma.
In total, his lab processes more than 8,000 patient samples a year, both solid tumors and hematological samples, on which it performs a variety of molecular tests, such as Sanger sequencing, real-time PCR, and the Sequenom assay.
Following the rapid technological developments in next-gen sequencing over the last few years, he and his colleagues decided that the next iteration of the multiplexed mutation assay should be sequencing-based. A year ago, the lab purchased an Illumina MiSeq, which Hedvat said seemed attractive because of the known quality of Illumina sequencing data, the availability of paired-end reads, the short run time of 27 hours, and the output of a gigabase per run, which fit the lab's needs.
In addition, the MiSeq would be able to support assay development on the HiSeq in the lab of Michael Berger, a researcher in MSKCC's department of pathology, work that might later be transferred to the diagnostic molecular pathology lab.
The Ion Torrent PGM was also available at the time of their MiSeq order, but its output was not high enough, it did not offer paired-end reads, and there was talk about issues with homopolymer regions, Hedvat said, adding that "in a clinical lab, we need the highest quality sequence we can possibly get."
He also did not consider Roche's 454 GS Junior, another available desktop sequencing platform, because of the high consumables costs and because he did not require the long reads that platform offers.
The lab received an early-access MiSeq in the fall of 2011 and designed a sequencing panel to look at 180 amplicons in 25 cancer genes, using the Illumina TruSeq Custom Amplicon assay. But there were a number of technical issues with the instrument, which Illumina eventually replaced with a new one, and due to mistakes in the assay design, the panel did not work.
As a result, the lab decided to try Illumina's TruSeq Amplicon Cancer panel, which had "most of the content" it wanted. The panel includes 212 amplicons from mutational hotspot regions of 48 cancer genes, of which a small number – Hevat said on the order of 20 or 30 – are considered to be actionable. "We tried their cancer panel, and it worked pretty well," he said, so the lab has been validating this assay and plans to submit it for approval by the New York State Department of Health by the end of this year.
This summer, the department released guidelines for somatic genetic variant detection using next-generation sequencing, which lay out the requirements for validation. Initial validation must include at least 40 patient samples "with a representative distribution of reportable genetic variants across all targeted amplicons," and positives must be confirmed by an independent method. At least 10 patient samples per amplicon or variant must be independently confirmed, either during the initial validation phase or during ongoing validation later on. After that, independent confirmation is no longer required.
Hedvat said his lab is performing a full validation – testing 10 positive samples – for its most common mutations in the KRAS, BRAF, and IDH1 genes first, and will validate the other amplicons on an ongoing basis.
Early next year, following certification by the health department, the lab plans to start running its first patient samples on the MiSeq, multiplexing around 16 to 24 samples per run. The assay is currently most appropriate for colon, melanoma, and glioblastoma samples but not yet for lung cancer because coverage for clinically relevant EGRF regions is still "suboptimal," Hedvat said. Until that issue is solved, lung cancer samples, which make up about two thirds of all samples assayed on the Sequenom, will remain on that platform.
Another potential issue is the amount of input DNA required for the assay. While the Sequenom assay only requires very little material –20 or 30 nanograms of DNA is sufficient – Hedvat said that going down to 100 nanograms in the Illumina assay, "you see much more inconsistency" than using more DNA. As a result, the Sequenom assay may remain the assay of choice if only very small amounts of patient DNA are available.
In parallel, Hedvat's lab is working with Ross Levine, a researcher in the human oncology and pathogenesis program at MSKCC, on a panel to profile patients with AML or myeloproliferative neoplasms, which they plan to develop by next spring. Earlier this year, Levine's team published a paper in the New England Journal of Medicine, showing the prognostic relevance of mutations in several genes for AML.
Their panel, which is in early development, includes 30 genes that will be PCR-amplified on the RainDance Thunderstorm platform and analyzed on the MiSeq. Hedvat's lab recently received an upgrade to the MiSeq that increases the system's output to up to about 8 gigabases per run, which he said will be necessary for this assay.
In addition, in collaboration with Berger's group, the diagnostic molecular pathology lab is working on a larger cancer panel for solid tumors that will include several hundred genes and use hybridization-based in-solution capture followed by sequencing on the Illumina HiSeq 2500. "We want to increase the panel so we can cover more genes and more exons within those genes, particularly with tumor suppressors. We want to be able to fully cover p53, fully cover p10, and we want to be able to grow the list," Hedvat said. The assay will also provide information about copy number variants and select rearrangements, in addition to point mutations.
The HiSeq 2500 is scheduled to arrive at the lab at the beginning of next year, and the plan is to have an assay developed by the fall. The instrument, which promises 120 gigabases of data in about 27 hours in rapid run mode, "makes sense to us," he said. "We can return samples and don't need to fill up eight lanes of a flow cell every time."
Berger's research lab has been developing a panel, using Roche NimbleGen in-solution capture, which currently comprises about 280 genes. It has had "good success" running 24 samples per lane on the HiSeq 2000, obtaining deep coverage, and is currently working on optimizing the capture conditions, coverage, and uniformity for a clinical version of the assay.
In addition, Berger's team, which has been using sequencing instrumentation in MSKCC's core facility, directed by Agnes Viale, has been working on sample prep methods to perform targeted sequencing from a variety of tumor specimens, including formalin-fixed paraffin-embedded samples and biopsies. They typically use 200 nanograms of DNA for the capture assay but have successfully gone down to as little as 15 nanograms from fine-needle aspirates of small-cell lung cancer. "We are starting a series of experiments to really probe how low we can go and still get adequate unique sequence coverage, in the neighborhood of at least 200x," he said.
Deep average coverage of 500x to 1,000x will be required for the clinical assay in order to bring low-coverage targets up to a minimum. Berger said he might also "trim the list [of genes] down to what makes sense in the clinical lab," enabling it to sequence even more deeply or pool more samples. For research projects, he has been reluctant to reduce the number of genes, though, because some projects depend on the presence of certain genes.
Berger has also collaborated with Foundation Medicine to examine the concordance between MSKCC's Sequenom assay and Foundation's targeted sequencing assay, and he presented data from that study at the American Association for Cancer Research annual meeting this spring (CSN 4/4/2012).
The study found results from both platforms to match well, and sequencing detected a number of actionable mutations that the Sequenom panel did not cover. "It was encouraging for us to see the technical concordance and also to see the potential for the added benefit by sequencing," Berger said. "Since then, we've really been pushing forward on the development of our own hybridization-based capture sequencing platform, akin to what [Foundation Medicine is] doing."
Developing their own platform, Berger said, allows them to use it for research as well as for clinical work, a significant advantage over outsourcing clinical testing. "By building the workflow and an analysis framework internally, you can build it large enough, so there can be extra capacity where you can run research samples as well," he said. "And even for the clinical samples, collecting and organizing this clinical data provides a research opportunity to establish the prevalence of different mutations in different tumor types, and to link them to clinical responses that we're seeing."
Hedvat explained that the diagnostic molecular pathology lab generally prefers to develop and run its own assays, especially if the testing volume is high. This allows them to customize assays to the needs of MSKCC's patients – for example, to develop assays not only for solid tumors but also for leukemias, or to include genes that can help enroll patients into clinical trials at Memorial.
He also said that MSKCC does not want to rely on providers whose assay is not yet certified by the New York State Department of Health – he said Foundation Medicine, like MSKCC itself, does not have certification yet, although it is able to accept samples from New York on a provisional basis. This has been an issue in the past, he said, when another company that did not have state approval had to stop testing, leaving MSKCC and several other clinical labs in the lurch.
Besides developing assays on the Illumina sequencing platforms, MSKCC researchers are also beginning to test Life Tech's Ion Proton, in collaboration with the New York Genome Center. Four early-access Proton instruments are being placed at MSKCC, which currently serves as the NYGC's Innovation Center, for technical validation by members of the NYGC (IS 8/7/2012).
Hedvat said that a sequencing platform must be stable for clinical applications, which he has found to be a challenge, given the changes Illumina has made to the MiSeq over the last year. Life Tech has likewise made many changes to its PGM, and has already outlined how the Proton will improve in the future.
"We don't necessarily want the latest and the greatest; we want something that we know works, that's been out for a while," he said. "I need a machine that's stable. They keep making all these improvements, and they are making us less confident in the stability of it because they keep changing the machine every six months."