William Check, PhD
|Dr. Corless offers two custom NGS oncogene panels, one for non-small cell lung cancer and a second for GIST. “We decided pretty early on that this approach has value,” he says.
Flanders and Swann, the popular British comedic duo of the 1960s, had a song about a mythical entity called a Wompom, a creature of seemingly universal utility:
“From the streets all full of houses”
“To the buttons on your trousers”
“With a Wompom you have everything you need.”
Next-generation sequencing is anything but mythical. Yet NGS is rapidly emerging as a kind of medical Wompom, with a potential to be ubiquitously useful for basic discovery of disease-causing genetic variants and for clinical application to patient diagnosis—when used knowledgeably.
Christopher Corless, MD, PhD, professor of pathology at Oregon Health and Science University, is offering two NGS-based oncogene panels to guide targeted cancer therapies. “The main advantage of next-generation sequencing is throughput,” Dr. Corless says. “We can sequence a whole lot more DNA at a reasonable cost.” Despite the value of NGS, its adoption must be approached with eyes wide open. “If you get into this, be prepared to make a substantial investment in time and effort to deal with the data,” Dr. Corless cautions. “It is not plug and play.” A second concern is financial. “There are a lot of people who see this as a gold mine,” he says. “There is kind of a gold rush going on to get into it. Given the uncertainty around reimbursement, one needs to get into this area carefully.”
Illustrating the basic discovery aspect of NGS in oncology, and the first stages of clinical application of those discoveries, is work on pancreatic cancer by Ralph Hruban, MD, and colleagues at Johns Hopkins University School of Medicine. “Next-generation sequencing has fundamentally changed our approach to gene discovery, especially for cancer,” says Dr. Hruban, professor of pathology and oncology and director of the Sol Goldman Pancreatic Cancer Research Center at Johns Hopkins. “We can now search entire exomes and genomes for mutations.” In the pancreas cancer work, this enhanced scope has led to discovery of a number of new genetic alterations. “Discovery of the oncogenes ATRX and DAXX in pancreatic neuroendocrine tumors was made possible through next-generation sequencing,” Dr. Hruban says. These genes define a new cancer pathway. Work with NGS also demonstrated the involvement of the mTOR pathway, which could be a target for therapy.
“Our challenge now is applying this knowledge to patient care,” Dr. Hruban notes. Most clinical sequencing is done now with the Sanger method. “In short order, application of next-generation sequencing to clinical samples in a CLIA-certified setting will change the landscape of clinical testing,” Dr. Hruban predicts. Already commercial clinical testing is being offered for PALB2, one of the familial pancreatic cancer genes discovered by whole exome sequencing.
In the field of medical genetics, the Sequenom Center for Molecular Medicine (CMM) is offering for patient care a noninvasive DNA-based blood test to identify trisomies, MaterniT21 Plus, that is done with NGS on circulating cell-free fetal DNA in maternal plasma. Detecting the small amount of extra DNA from chromosome 21, 18, or 13 in maternal blood is possible only because of the extremely high sensitivity of NGS. “The primary advantage for the MaterniT21 Plus test is that it is noninvasive,” Mathias Ehrich, MD, Sequenom’s vice president of R&D, says. Trisomy testing is performed now with amniocentesis and karyotyping. Of the rapid adoption of MaterniT21 Plus, Dr. Ehrich says: “Not much convincing is necessary. The majority of physicians are excited to offer this technology to their patients.”
Arnold Cohen, MD, chairman of obstetrics and gynecology at Albert Einstein Medical Center in Philadelphia, is enthusiastic about the test. Dr. Cohen calls NGS “breakthrough technology, a game-changer.” With regard to detecting trisomies, he says the best way is noninvasively on fetal DNA from the mother’s circulation: “There is no risk of losing the pregnancy. It is only in the last couple of years that technology gave us that ability.”
A more complex application of NGS in medical genetics is being offered by the UCLA Clinical Genomics Center: whole exome sequencing to detect causative genes for rare Mendelian conditions in individual patients and families. “We are trying to infer a protein sequence change compatible with a child’s diagnosis,” says Stanley F. Nelson, MD, professor in the Departments of Human Genetics, Pathology and Laboratory Medicine, and Psychiatry. “Having the facility to do whole exome sequencing on patients who present for clinical diagnosis opens up a new avenue,” Dr. Nelson adds. Extensive analysis is needed to match genes with conditions because, Dr. Nelson says, “Rare Mendelian genetic diseases can present in very heterogeneous ways. The same phenotype can be caused by multiple genes, including a subset not even described yet.”
Implicit in Dr. Nelson’s last comment is that the clinical test also does discovery work. “In just the first few months that we have offered the test we discovered a new gene for cardiomyopathy and one for seizure and autism syndrome,” Dr. Nelson says, but cautions, “We haven’t yet proven those associations.” They are searching for additional families that carry the genes.
Dr. Nelson offers another caution—about doing NGS in a clinical setting. “It is not at all simple and the nuances are not to be taken lightly,” he told CAP TODAY. In particular, he emphasizes crucial differences between NGS used in a research mode and NGS applied to patient diagnosis. “This is true in terms of how we sequence, the depth of sequencing, and the base positions we are able to call with a high degree of certainty,” he says. Although confident that NGS technology will continue to improve, he says: “We can easily see a world where sequencing is very accurate but interpretation is still a very complex process, given that each person has a couple of hundred very rare protein-damaging DNA variants.”
Pathologists performing NGS testing foresee a strong role for their specialty. “Pathologists with molecular training will play a key role” in the application of NGS, Dr. Hruban says. “It’s the ability to integrate molecular understanding with traditional histopathology. I don’t see molecular displacing histopathology but supplementing it.” In lung cancer, for instance, EGFR and KRAS oncogenes are sought only in adenocarcinomas. “So first we do standard pathology, then add molecular.”
“We [pathologists] need to involve ourselves now to understand this new technology and to use it,” says Karen Kaul, MD, PhD, chair of molecular pathology and director of the molecular diagnostics laboratory at Evanston (Ill.) Hospital, NorthShore University HealthSystem. But the typical pathologist is unlikely to be doing this type of testing in-house anytime soon, she says. “Setting up next-generation sequencing is a huge investment at this point. I think the terrain will change pretty quickly in the next few years,” says Dr. Kaul, clinical professor of pathology at the University of Chicago Pritzker School of Medicine. Changes will be made in instruments, reagents, and software, making it more practical for clinical use.
Even so, she says, the current level of training in pathology residencies, and even molecular genetics pathology fellowships, is not enough preparation to set up next-generation sequencing as a clinical laboratory test. “The complex data analysis needed at present requires a skill set beyond what most fellowships now offer,” Dr. Kaul says. As kits, automation, and software tools are developed, it will become easier to bring NGS in-house. “We need to be ready to use this to extend our diagnostic skills.”
A description of the first next-generation sequencer, Life Sciences’ 454, was published in 2005 (Margulies M, et al. Nature. 2005;437:376–380). The 454 offered an increase of several orders of magnitude in sequencing speed relative to the Sanger method, an increase that was made possible by miniaturization of sequencing chemistries, enabling massively parallel sequencing reactions to be carried out on an unprecedented scale (Rothberg JM, Leamon JH. Nat Biotechnol. 2008;1117–1124). Several variations on the 454 have been devised, and more rapid and more accurate instruments are now available.
Smaller instruments suitable for sequencing small sets of genes have since been introduced; they’re ideal for clinical applications such as the panels Dr. Corless offers. He uses an Ion Torrent PGM platform to screen for mutations across 23 genes clinically important in non-small cell lung cancer; a second panel covers oncogenes involved in gastrointestinal stromal tumors. Both are custom panels developed in his lab using Ion Torrent/Life Technologies’ AmpliSeq mutliplex PCR technology. Additional cancer-specific panels are being developed and validated.
“We decided pretty early on from our initial experience that this approach has value,” Dr. Corless says. “We figured out in 2010 that we had to make the jump and purchased three of Ion Torrent’s PGMs. We have been collaborating with them in beta testing reagents and protocols.” Dr. Corless offers the tests through the Knight Diagnostic Laboratories (of which he is chief medical officer), a CLIA-certified facility that is separate from the university’s research genomics core and whose establishment was made possible in part through a substantial gift from Nike chairman Phil Knight and his wife, Penny.
Of course, oncogenes can also be screened by other multiplex PCR technologies. Dr. Corless cites two advantages of NGS. First, NGS can detect rare mutations in oncogenes that are not detected by allele-specific PCR approaches. Second, “With sequencing we can pick up mutations in tumor suppressor genes. We can find variations virtually anywhere that might knock out gene function. Of all genes mutated in cancer,” Dr. Corless points out, “the majority are tumor suppressor genes.”
Two major types of resources are needed to do NGS: personnel who have qualifications to work with data files and determine what is a real mutation and what is background, and personnel who can understand the mutations or variants identified, put them into clinical context, and communicate the information to clinicians. “Those two sets of workers are not necessarily the same people,” Dr. Corless notes.
Underlying the whole exome and whole genome sequencing that Dr. Hruban and his colleagues do is the principle that development of cancer has two aspects: inherited mutations and those acquired during a person’s life (somatic mutations). Familial predisposition genes can be identified by sequencing genomes from families in which multiple members have been diagnosed with pancreatic cancer. Somatic genetic alterations can be found by comparing sequences in cancer tissue to normal tissue from the same person. The Johns Hopkins group has done all of these types of analyses.
To generate a discovery set of genetic alterations, they extensively analyzed the exomes of 24 people with pancreatic cancer (Jones S, et al. Science. 2008; 1801–1806). For familial predisposition genes they sequenced the genomes or exomes of 38 affected patients from 16 families with at least three cases of pancreatic cancer to identify mutations that segregated with the disease (Roberts NJ, et al. Cancer Discov. 2012;2:41–46). In two families, all six sequenced individuals with pancreatic cancer carried mutations in ATM, a gene first found in patients with ataxia telangiectasia.
In another project, they used NGS to sequence the whole exomes of 10 nonfamilial pancreatic neuroendocrine tumors (PanNETs), extending their findings to 58 additional tumors. In 16 percent of the tumors they found a mutation in genes of the mTOR pathway, which might make these tumors susceptible to mTOR inhibitors, such as rapamycin (Jiao Y, et al. Science. 2011;331:1199–1203). Interestingly, a large clinical trial published at about the same time found that some patients with PanNETs responded to the mTOR inhibitor everolimus. “We haven’t yet matched up treatment success with mTOR pathway mutations,” Dr. Hruban says. But that’s a logical thing to do.
In work of this type, Dr. Hruban says, “There is a huge amount of human effort to interpret the data.” Interpretation requires two elements. First, when looking at millions of data points, one must develop appropriate infrastructure and computer programs to handle the large dataset. Second, Dr. Hruban says, “We must develop the knowledge base to interpret the clinical significance of the findings. Determining whether a change in a gene has a clinical effect will be quite difficult.” To quantify that challenge, he notes that a human exome has on average almost 36,000 variants, 45 percent of which are not in a SNP database and about 100 of which can cause loss of function.
Sequenom CMM first offered the MaterniT21 test in October 2011, upgrading it to the Plus version at the start of 2012. The test is offered to women with high-risk pregnancies: family history, age over 35 years, abnormal ultrasound for nuchal translucency, or abnormal first trimester serum screen. Dr. Ehrich explains that the test uses NGS to sequence a very short portion of every DNA fragment in the maternal bloodstream to determine which chromosome it comes from. (It is not necessary to distinguish fetal from maternal fragments, which comprise about 90 percent of circulating cell-free DNA.) A fetus that carries an extra chromosome 21 also releases a bit of extra genetic material from chromosome 21. “It is a very small percentage,” Dr. Ehrich says, “but that is why we use next-generation sequencing. It clearly differentiates between normal and abnormal with 99.9 percent specificity and 99.1 percent sensitivity.” Compared with amniocentesis, MaterniT21 Plus detects slightly more abnormalities, 2.7 percent versus 2.2 percent, and trisomy 21’s, 2.1 percent versus 1.1 percent.
Sequenom CMM’s goal is to perform 50,000 MaterniT21 Plus tests this year. Its current run rate extrapolates to a higher annual volume than that, testifying to its rapid acceptance. The test is done on Illumina’s HiSeq 2000 machine; Sequenom CMM has the capacity to do 100,000 tests per year. Turnaround time is seven business days, and out-of-pocket cost to patients is $235.
Dr. Cohen, chair of ob/gyn at Albert Einstein, notes that MaterniT21 Plus is the culmination of a trend. “Testing for trisomy 21 has become better and better with time,” he says. When he started in practice 30 years ago, about 60 percent of fetuses with Down syndrome could be detected using serum screening plus amniocentesis. “In the last 10 years nuchal translucency plus the blood test raised this to 90 to 95 percent sensitivity,” Dr. Cohen says. “Now with a negative MaterniT21 test, we can exclude Down syndrome with greater than 99 percent assurance. It also has a high positive predictive value.” A positive test is confirmed with amniocentesis or chorionic villus sampling.
Could an NGS-based test for trisomies become first-line? “I don’t see any scientific reason why a sequencing test shouldn’t work on a low-risk population,” Dr. Cohen says. “It just needs to be verified with data. If the price comes down, it will be for everybody.”
Dr. Nelson and his UCLA colleagues started offering whole exome sequencing, which they call Clinical Exome Sequencing, in January 2012. Using an Illumina HiSeq 2000, they sequence 95 percent to 96 percent of all protein coding bases, including all known Mendelian genes, looking for de novo point mutations. Preferred testing mode includes the affected child and both parents. “We can do about 40 such complete diagnoses per month with a rather small crew and a typical turnaround time of eight weeks,” Dr. Nelson says.
To indicate the extent of the challenge in handling the sequence data, Dr. Nelson says that for each sample, they sequence the complete exome, roughly 30 million bases, each of which is over-sampled 130 times—65 readings per base from the mother’s genome and the same from the father’s. The child’s sequences are compared with those of the parents. “If we have a one percent error rate, many hundreds of bases will be wrong,” Dr. Nelson says. With careful data processing, however, “we can have an average likelihood of error less than one in a million.” (Some regions do not sequence well, such as repetitive regions or gene families.) All of these sequence data are processed through a clinical pipeline for cleaning data. “So we feel confident they are accurate,” Dr. Nelson says. To get an acceptable overall accuracy, they discard the less reliable data.
“That is what takes time,” Dr. Nelson says, “determining which base reads are unlikely.” In a typical flow, it takes several days to make DNA libraries from the three family members and quality check them. Sequencing takes about 10 days, after which the sequence data are put onto a server. “It takes about a week to do the complex process of essentially taking each one of a couple hundred reads and finding it in the genome,” Dr. Nelson continues. “Once we have all sequences aligned, among the 30 million bases there will be about 20,000 variants. Of those the vast majority, greater than 95 percent, are polymorphisms. Some may cause disease, but most often we are searching for de novo or rare variants. We will typically see 1,000 or so variants not observed before, which make up a unique map for that patient.” Finally, they look among the 4,000 or so Mendelian disease genes to see whether any variation is “of sufficient interest”—that is, whether it might be related to that patient’s diagnosis.
“We are fairly early in this game,” Dr. Nelson says, “so we do this examination in a rather thorough but relatively permissive fashion. We probably overlink disease genes and mutations.” Possible links are presented to a Genomics Data Board consisting of Dr. Nelson, geneticist Eric Villain, MD, molecular pathologist Wayne Grody, MD, PhD, a bioinformatics researcher who organizes the data and presents them, and other interested individuals. Genetic counselors hear all discussion so they can present the information to the family. “This is a very popular meeting,” Dr. Nelson says. “There is considerable variation among humans, and even how Mendelian variants present is highly variable, so the meeting itself is highly educational for all of us.” Data review at the board meetings is a critical part of the process of interpreting sequence data. “It requires informatics experts and people who know genetics well. In some cases we have been able to clarify the diagnosis rather dramatically,” Dr. Nelson says.
They are encouraging pediatric specialty physicians who are not getting an accurate or adequate molecular diagnosis to order this test, he says. Indications might be autosomal dominant diseases in a child under one year that present as developmental delay or a presentation consistent with Duchenne muscular dystrophy. “There are sometimes point mutations in the gene for Duchenne’s,” Dr. Nelson says, “and often duplications or deletions as well.” Traditionally, sequential analysis is performed for these two types of genetic variation. Using NGS, Dr. Nelson says, “We can often do both analyses in one test and return the result in a shorter time at about comparable cost.” Clinical exome sequencing for mother, father, and child costs $6,500.
Currently, diagnosis of presumed genetic conditions consists of ordering a gene sequence based on phenotypic studies. If that result is not helpful, another gene may be chosen. “Clinical presentation is not really guiding the diagnosis well,” Dr. Nelson says, “so the physician is testing individual genes one at a time. Consequently, it often takes many months to years to give people a diagnosis.” Some patients arrive at UCLA after having spent tens of thousands of dollars. “With clinical exome sequencing,” he says, “we can short circuit that process substantially. We believe that genetic diagnosis by clinical exome sequencing with early implementation is more cost-effective and has a higher diagnostic yield than the current approach.” He admits that this conclusion is “still anecdotal,” saying, “It would be nice to prove it.”
“We need to begin to learn about it now,” Dr. Kaul, speaking as a practicing pathologist, says of NGS. “We will need to have this information [from NGS] to determine management of patients for many, if not eventually most, diseases.” For pathologists, the first use of NGS may be to identify molecular abnormalities in tumors that dictate the most effective treatment. “There are a variety of levels at which pathologists can get involved,” she says. Institutions will need to determine when and whether to bring these tests in-house or send them out, and the answer may change over time as manufacturers begin to offer reagent kits on their various platforms, she says. Whether from in-house or send-out testing, the information that NGS provides will be incorporated into the pathology report to help guide oncologists. “Eventually we may use whole genome sequencing, though right now for clinical purposes we will more likely be doing targeted analysis,” she says.
A number of laboratories, like Dr. Corless’, are already offering oncogene panels, she notes, which requires technology, experience, and understanding. “People talk about the $1,000 genome,” she says, “but what is not often addressed is the $10,000 or more of information processing that makes the data useful.”
Becoming familiar with the diagnostic power of NGS appears to stimulate those doing it to think big. Dr. Nelson, for one, says he would like to launch clinical genome sequencing as inexpensively as clinical exome sequencing.
Dr. Ehrich says that the technology used for the MaterniT21 Plus test is genomewide technology: “It can be used to interrogate the entire genome. Our next logical step is detection of Y chromosome material, then sex chromosome aneuploidy, such as Turner’s syndrome.” He also describes what he calls “blue sky research.” He and his colleagues published a paper earlier this year showing that NGS can be used to detect microdeletions, or subchromosomal fetal copy number variants (Jensen TJ, et al. Clin Chem. 2012;58:1148–1151). Based on this demonstration, Dr. Ehrich says, “Ultimately, the technology has the capability to generate something akin to a molecular karyotype,” though this application is “very far in the future.”
At Knight Cancer Institute, Dr. Corless foresees their eventually doing whole exome and even whole genome sequencing. But first they have issues to address. “One is that not all clinicians we serve are ready to receive these huge amounts of information. Some are not even so keen on getting data from a panel of 20 genes. They think more in terms of just two genes,” he says. “So we need to get clinicians up to speed on handling this new information.” Moreover, many new disease-causing variants are likely to be identified with whole genome or whole exome sequencing. “We don’t have the pharmaceuticals to target all those variants,” Dr. Corless says.
This much is clear: “The potential power of next-generation sequencing for clinical testing is substantial,” Dr. Hruban says. But it does pose challenges, such as how to sort and interpret all the data. “It will be a while before this technology is brought fully to the clinic,” he concedes, “but the issues are all addressable.”
William Check is a medical writer in Ft. Lauderdale, Fla.