November 2011
Feature Story

Anne Paxton

There may be debate about whether DNA sequencing really has advanced more rapidly than any human technology in history. But as a benchmark, consider that if aviation had progressed as much since the first transcontinental flight as DNA sequencing has progressed since just 2006—when next-generation technology came on the scene—a flight from Los Angeles to New York would be akin to teleportation, which so far exists only in science fiction.

Along those same lines, some pathologists and scientists involved in next-generation DNA sequencing have advice on how to get ready for the next five years: Fasten your seatbelts.

Under Moore’s Law, computing power and affordability have been reliably predicted to double every two years. DNA sequencing is outstripping that pace. As next-generation sequencing (NGS) continues to edge aside Sanger sequencing of DNA, exponential leaps forward in speed and exponential reductions in cost every six months have become routine.

When it was launched in 1990, the 13-year $3 billion Human Genome Project was many people’s introduction to future shock. “That just started the ball rolling,” says David I. Smith, PhD, professor of laboratory medicine and pathology and chair of the Technology Assessment Group at the Mayo Clinic, Rochester, Minn. “The real revolution began in 2006 when the first of the next-generation sequencing machines, Roche’s 454 machine, became available. And this is the biggest revolution I’ve seen in my scientific career.” In fact, he compares it to a technological “singularity,” an advance “that will so drastically change everything, you can’t imagine what things will look like on the other side.”

Dr. Smith predicts that next-generation sequencing will move from the research laboratory to the clinical laboratory, and begin to reshape the diagnosis and treatment of inherited disease, cancer, infectious disease, and perhaps chronic disease, as early as next year. Nevertheless, to make that transition, he and other DNA sequencing experts warn, the industry and laboratories have a few hurdles to clear.

While next-generation sequencing is available in several commercial forms from Illumina, Life Technologies, Pacific Biosciences, and Roche’s 454 Life Sciences, what the instruments have in common is the use of massively parallel processing, the kind associated with supercomputers. “In each instrument, the DNA sample is fragmented into many pieces and all of those pieces are sequenced simultaneously,” says Gregory Storch, MD, professor of pediatrics, medicine, and molecular microbiology, Washington University, and director of laboratory medicine in pediatrics, St. Louis Children’s Hospital. “This makes it possible to generate much, much more sequence at a lower cost per base in the sequence.”

Even in 2007, this massively parallel technology could sequence simultaneously a half million bases at a time. And since then, computing speed has increased eight- to 10-fold per year. “The human genome is 3.3 gigabases. In 2007 the best machine in the world could churn out, with one run, one gigabase of DNA sequence. Now we are three orders of magnitude better,” Dr. Smith says. Just around the corner is the next upgrade of the HiSeq 2000 from Illumina, which will be able to sequence 1.13 terabases.

The cost is dropping just as astronomically, he adds. “We used to measure the cost of sequencing a million base pairs; now we’re measuring similar costs but for a billion base pairs.”

This dramatically evolving technology will open new clinical doors. “With first-generation DNA sequencing, we could complete the human genome, but we certainly could not think about practicing genomic medicine clinically,” says Jim Versalovic, MD, PhD, professor of pathology and immunology, Baylor College of Medicine, and head of the Department of Pathology, Texas Children’s Hospital, Houston. “Next-generation sequencing is allowing us to tackle genomics and apply genomics to medicine.”

“NGS is going to enable many forms of diagnostic testing that we’ve never considered studying before,” says Timothy J. Triche, MD, PhD, developer of the Center for Personalized Medicine at Children’s Hospital Los Angeles, and professor of pathology and pediatrics, Keck School of Medicine, University of Southern California. Technically, NGS technology is restricted to research use only (RUO) or investigational use only (IUO), but CLIA-certified clinical labs often use RUO or IUO technology to develop laboratory-based tests, so-called LDTs. Many such tests have been developed using a variety of methods, from immunohistochemistry in the past to NGS now. These tests are restricted to the lab that develops the test.

“The key thing is whether NGS is used to develop a laboratory-developed test. LDTs frequently utilize non-FDA approved technologies, of which NGS is an example. Diagnostic immunohistochemistry, for example, generally uses antibodies chosen by the pathologist that are often labeled RUO, and the pathologist generally evaluates the performance of the antibody as part of his or her medical interpretation. The FDA published draft guidance last June that reagents and platforms labeled RUO or IUO should not be supplied to clinical labs that perform diagnostic tests that use these reagents or platforms. If that guidance is implemented as currently stated, it will certainly complicate the use of NGS for diagnostic testing.”

Life Technologies says it plans to submit a 510(k) device application for its next-generation instruments to the FDA as early as next year, with hopes for approval in 2013. “We expect in the next three to five years that next-gen sequencing will really be used widely in pathology for both diagnosis and to guide treatment,” says Mark Stevenson, president and chief operating officer of Life Technologies.

Whether next generation will be able to live up to the hype in the clinical arena remains to be seen. “These are early days in the exploration of clinical applications,” Dr. Storch says. “In oncology they may unfold quite rapidly, while in infectious disease it may be a slower process of understanding the relative strengths and weaknesses of PCR versus NGS. I don’t think NGS is going to revolutionize the world of infectious diseases overnight, but it’s an extremely powerful technology that will have a major impact over the next several years.”

At this stage, panels of tests using NGS are becoming available, Life Technologies’ Stevenson says. “We think it’s more actionable to use the genes we already know about than to go to the whole genome where there are a lot of things we don’t yet understand. The research community may be doing the whole genome, but for the pathology lab, I think it’s easier to start by doing panels of assays where we know what a particular gene means.”

Life Technologies, for example, recently released what it’s calling the Ampliseq Cancer Panel. “It has 46 genes together on a panel with 739 mutations. So you have all the common genes like EGFR, KRAS, BRAF, and in one shot you can analyze the hot spots you would expect in a few of the major forms of cancer.” Rather than just order one test for KRAS, a researcher could order and look at all the pathways—not only KRAS but other well-known markers, Stevenson says.

The first clinical applications are most likely to be laboratory-developed tests, says Dr. Triche, who has two major grants from the Department of Defense that have been responsible for his laboratory’s embrace of next-generation technology. “We are planning to get our first LDT out the door at the end of the year if at all possible. That will be retinoblastoma gene testing.”

The test is relevant to the clinical program at Children’s Hospital Los Angeles because the hospital has so many retinoblastoma patients. “They need to be tested for an inherited defective gene rather than an acquired defect, because the inherited gene patients are more at risk of developing a second malignancy. We estimate that one out of seven patients who look like they have sporadic retinoblastoma actually have familial, and we’re developing the test to determine which ones are which.” In an ideal world, Dr. Triche says, both the retinoblastoma test and next-generation sequencing would be declared ready for CAP/CLIA certification by the end of this year. “That’s our goal anyway.”

Retinoblastoma is far beyond the capability of routine Sanger sequencing, Dr. Triche says. “Because the gene covers about 250,000 bases of the human genome, and there are 27 exons and a whole bunch of introns that really should be looked at, most people don’t bother to look at all that material. Our goal is to leverage the inherent strength of NGS to cover the whole thing.” With NGS, the cost could be as low as $600 or $700 for the entire gene in two days, compared with as much as $4,000 for a conventional test that sequences only parts of the gene. Now that Life Technologies has just announced the third generation of its Ion Torrent semiconductor chip technology by the end of the year, those speeds may increase dramatically, placing same-day turnaround time within the realm of possibility, he says.

Instruments like the Ion Torrent can sequence only a fraction of the entire genome. However, Dr. Triche believes there is good reason not to sequence the whole genome. “It’s common to hear, once the whole genome is sequenced, that out of 3.3 billion bases at least 3.3 million don’t match. Buried in that may be disease-associated mutations, but you now have to figure out which ones they are out of the 3.3 million. So people are just overwhelmed trying to figure out which mutations actually matter. Everybody is groaning under the analytic burden of separating fish from fowl.”

Sequencing only the exome (all the protein-encoding exons in a genome) is considered an efficient alternative, since the exome is only about one percent of the genome but includes an estimated 90 percent of all mutations with large effects. But Dr. Triche questions it. “I’m not really sure anyone really knows what the exome is. There’s a real risk you’re going to miss important mutations and other abnormalities.”

The current thinking among some scientists is that only a certain number of genes have ever been documented as mutated in cancer. “The largest number I’ve seen is 500,” says Dr. Triche. “So now a lot of people are interested in looking at just those genes—the low-hanging fruit or most likely suspects.” Using a kit like Ion Torrent’s new AmpliSeq, “you could look at all the commonly mutated genes in all forms of cancer in one fell swoop in about a day. So the analytic problem becomes very tractable.” At Children’s Hospital, once the retinoblastoma test is validated, “we will be increasingly looking at panels of genes implicated in childhood cancer and presumably other cancers, so we’ll go from a single gene mutation analysis to multiple gene panels sequencing analysis,” Dr. Triche says.

Thomas Jefferson University in Philadelphia was interested in performing next-generation sequencing for two applications when it acquired a Roche 454 Junior, says Stephen C. Peiper, MD, professor and chair of the Department of Pathology, Anatomy, and Cell Biology. “We were interested in targeted sequencing, where we capture exons and then sequence specific genes, and in deep sequencing, also called amplicon sequencing. And NGS gives a different readout from what we do with traditional [Sanger] sequencing.”

One gene he considers a useful target of NGS is JAK2. “It has a mutation at specific sites that is associated with several diseases called myeloproliferative neoplasms, like polycythemia vera and idiopathic myelofibrosis. Typically, blood or bone marrow is tested for the JAK2 mutation, but they aren’t really quantitative tests, so they have a chance of error.” Next-generation sequencing, however, can tell exactly how much JAK2 is in the patient’s chromosomes and blood cells, and exactly what percent of the cells are mutated. “So in polycythemia vera, you can see whether the patient is responding to therapy by having a lower number of cells circulating with mutations, or whether the number is expanding or getting worse.”

While the clinical laboratory is currently not offering next-generation sequencing, he says, “It’s our intention to do so over the next year, starting with KRAS and BRAF.” Eventually the lab will look at a more complex mutational profile of 15 or 20 genes in many cancers. “There are a dozen centers right now that are starting to do that, picking many targets to look at for mutations. Some of that is being done in translational research mode, but I think that’s going to move forward.”

However, beyond such panels of tests, Dr. Peiper thinks it’s important to proceed cautiously into clinical applications. “Before we start thinking about how we’ll do whole genomes and whole exomes and use the full capability of next-generation technology, we should consider how the technology can help us. I think it becomes almost a self-correcting process, where, as we use the technology, we will understand what its strengths and limits are, and that will guide us into fuller implementation.”

Researchers at Washington University are studying dozens of patients with acute myeloid leukemia, glioblastoma multiforme, and several other tumors, Dr. Storch says. They are taking DNA from the malignant cells and doing whole genome sequencing, then for the same patients sequencing the normal genome from a skin biopsy. In each patient they compare the malignant genome with the normal genome to detect somatic mutations, which are those mutations present in the tumor DNA but not the germline DNA obtained from the skin biopsy. “This would have been inconceivable without next-generation sequencing; it would have cost millions and millions of dollars,” Dr. Storch says.

Mayo Clinic is working on a clinical test based on next-generation sequencing that may be available in 2012. The goal of the test, which is being developed under the leadership of Matthew Ferber, PhD, is to sequence simultaneously 22 genes known to be involved with hereditary colorectal cancer at a cost that is currently less than the cost of using Sanger sequencing to sequence just one of these genes. “Once this hereditary colorectal cancer panel is up and running,” Dr. Smith says, “there are then many other additional gene panels, such as one for cardiovascular disorders, that could be developed. By examining the complexities of testing a small number of genes, we hope that Mayo and the clinical community will be much better poised to tackle whole exomes and then eventually whole genome sequences.”

As for cancer treatment, sequencing the genome, then sequencing the expressed portion and figuring out if it’s methylated, another type of modification, are all going to be part of routine care. “And once we know what’s up with the cancer cell, we’ll have AIDS-like cocktails as cancer cures very quickly.”

Some patients are already showing up at Mayo Clinic with their complete genome sequenced, Dr. Smith says. “Not everyone is going to do it, but the number of people around the world that have been sequenced is up to 10,000 or 20,000, and we estimate next year it will be about 200,000 to 300,000.”

One of the first clinical applications Dr. Smith expects from next-gen technology is an inherited disease assay out of the National Center for Genomic Research. “They had a test where they pulled out 450 genes, each one responsible for a relatively rare, recessive, very bad disease. You have a young couple take the test and you ask the question, Do they have a mutation in any of the same genes? If they’re clear, they’re clear.” It’s the same model, he says, that was effective in the Ashkenazi Jewish population for wiping out Tay-Sachs disease, but much of the PCR and sequencing for that gene was conducted before the arrival of NGS.

Because of the speed and enormous capability of next-gen sequencing, broader uses in dealing with cancer are in store, says Jeff Boyd, PhD, executive director of the Cancer Genome Institute at Fox Chase Cancer Center, Philadelphia. “We’re talking about two hours to run through a very large amount of data, and unmatched sensitivity, detecting a signal down to the five percent to 10 percent range. And that’s very important in the cancer context, because if you are dealing with biopsy specimens or resections or other types of cancer material, a large number of cells likely to be non-cancer cells will dilute your signal for the mutations of interest.”

His institute intends to use NGS on essentially any patient available who has metastatic cancer to identify relevant mutations and guide treatment. The list of identifiable mutations and the list of drugs to treat them will lengthen rapidly, Dr. Boyd predicts. “The deeper we look, the more we are finding that beyond the mutations well known to be associated with certain tumors, there are mutations that haven’t been traditionally associated with those tumors.” For example, the BRAF mutation is known to be linked to melanoma, and novel treatments for it are coming out. “But we’re also finding that if you look hard enough, you see that a small proportion of non-small cell lung cancers also have BRAF mutations.” That’s the payoff of using technology with the sensitivity of next-gen sequencing, he adds.

Fox Chase is now engaged in clinical trials to demonstrate these links, and its plans are ambitious. “We fully expect to have validated the Ion Torrent instrument on paraffin-embedded material, and we will be ready to see patients very early in the new calendar year. And I firmly believe we will be among the first institutions to offer a product like this that actually does have a substantive impact on choosing cancer therapy for patients,” says Dr. Boyd.

With the Ion Torrent’s panel-based approach, Fox Chase has been able to hone in on appropriate targets. “We’re only looking at about 50 cancer-related genes that are most relevant to clinical intervention, not the whole exome,” he says, noting that the majority of the genome and even most of the exome is of little utility in terms of targeted therapy. So while the pharmaceutical industry works on identifying new small molecules or antibodies that target genes known to be mutated and to be druggable in human cancers, research using NGS will be showing that some of the new drugs are in fact effective in targeting genes not previously known to be druggable. “It’s just a matter of the research moving the needle on the record.”

What Fox Chase research is revealing is that most cancers are relatively complex in genetic architecture. “At this point, it’s hard to say that one or another type of common cancer is going to be more or less amenable to personalized therapy,” Dr. Boyd says. In five to 10 years, he predicts, patients at major academic medical centers and cancer centers will have their genome sequences in hand when they come in with a diagnosis. “It will certainly not replace anytime soon standard pathology in terms of diagnosis, but it will become a standard supplement, as we get away from concepts like ovarian cancer and liver cancer based on site of origin, and we move toward category A cancer, or category B and C cancer, based on molecular architecture, as opposed to how the cancer appears under a microscope.” He describes this transformation as a huge paradigm shift.

Beyond cancer and inherited diseases, NGS is opening up tantalizing new horizons for study of the relatively little explored human microbiome—the communities of microbes that colonize the human body and outnumber human cells by 10 to one. It’s a field called metagenomics, the study of the microbiome plus the human genome. “Metagenomics allows us to look at hundreds of bacteria in the intestine, reproductive tract, oral cavity, and so on—any body surface that’s connected somehow to the exterior. They are full of different microbes—bacteria, viruses, yeasts, and fungi,” says Dr. Versalovic, who is a principal investigator within the Human Microbiome Project. Most of these microbes have never been successfully cultured. But with the extraordinary sensitivity of NGS, “we can begin to sequence and determine changes in patients that may be associated with different disease states.” In fact, NGS is essential for metagenomics, he says.

Clinical applications are in development. “At Texas Children’s Hospital in the pathology department, we are beginning to develop new diagnostic tests based on new findings on the human microbiome. For example, we have identified particular bacterial DNA sequences that are related to a chronic GI condition called irritable bowel syndrome.” The project is also looking at obesity and metabolic syndrome and such disorders as inflammatory bowel disease and ulcerative colitis, as well as the vaginal microbiome in women and the urethral microbiome in men.

Dr. Versalovic is optimistic that the microbiome can be modified. “It will be much easier to talk about modifying the metagenome than about modifying the human genome. There have been efforts for years to sell gene therapy to change the genetic composition of human beings and it’s very difficult to do. The microbiome is a more tractable problem.”

In the area of infectious disease, other labs are already using NGS to discover a large number of microbial agents, Dr. Storch says. “NGS applied to human genomes will also help us get a much better handle on the subtle genetic variations that may make people more or less susceptible to a wide range of infectious agents. Right now, we take a person with pneumococcal pneumonia and just treat them for their infection, but in the future we’ll look for several mutations that might make a person more susceptible to bacterial infection and might have consequences for managing their care.”

Dr. Storch became involved in the Human Microbiome Project when a request for proposals was issued. “At the time I was doing a study trying to determine viral causes of fever in young children without a source. I realized there was a bias in the Human Microbiome Project toward bacterial infections, so I put in an application to use NGS to detect viruses, and it was funded.”

In his study, children who came to the ER with fever without a source—that is, with no respiratory or gastrointestinal symptoms—were enrolled and tested using PCR assays for a number of known viruses considered possible causes. “We are also doing next-generation sequencing on the very same samples—blood, nasopharyngeal secretions, and stool—to see if we can find other viruses, including previously undiscovered ones,” Dr. Storch says.

Through NGS, the study team did find a large number of viruses in the patient samples—some not expected. “This illustrates one of the potential advantages of NGS if applied to infectious disease diagnosis. The doctor sees a patient with possible infection and does a PCR test. But it has to be for specific microbes, and there could be more possibilities than that. NGS is potentially unbiased and could reveal other pathogens besides the ones in the PCR panel.” At current prices, both the extensive panel of PCR tests used in the study and NGS would be considered too expensive to be practical for these kinds of patients, Dr. Storch says. “But the cost of NGS is going down as we talk. So in two to five years, it might be manageable.”

From a regulatory point of view, it’s not all clear sailing ahead. “There are potential obstacles,” says Life Technologies’ Stevenson. “At the moment the industry and the FDA are working closely on how we can deploy this technology robustly into a clinical setting. The first agreement will be around analytical and quality standards so that we can compare data with reference to standard sequencing. The second issue will be to describe how we’ll do clinical utility, because in these kinds of trials with NGS, we will be able to read genomes and variations of genomes that may be best done by analytical standard rather than having to prove clinical utility.”

The regulatory situation is a “clouded scenario” right now, says Eric Kolodziej, PhD, senior vice president of global regulatory affairs for Roche Diagnostics. “The FDA wants to bring forward an IVD solution. There’s no doubt that is its ultimate goal. A number of companies, including Roche, have been actively engaged in discussion with the FDA to define a path for how to analytically validate the technology and deal with the variety of platforms that exist, as well as the challenges related to bioinformatics. In the meantime, what’s going on is the LDTs really rule the roost in the space for sequencing, and they’re governed largely by CLIA labs. So the FDA also has the issue of how to resolve the oversight of LDTs or homebrews to ensure an even playing field.”

The pathway in Europe is much more straightforward, Dr. Kolodziej notes. “We sell these kits or reagents to the labs, and they develop their own reimbursable tests. There’s not as much confusion in the oversight.”

The FDA brought many industry, research, and medical leaders together June 23 for a public meeting on this issue, and it will hold another such meeting before making decisions, but the instrument manufacturers express concern. “We expect this technology to have significant demand in this area. But it depends exactly on how the FDA goes forward,” says Thomas Schinecker, PhD, president of Roche’s 454 Life Sciences and life cycle leader, sequencing. “And that’s very unclear at the moment.”

Dr. Versalovic says FDA regulation of diagnostic tests is going to have to be reconsidered. He worries that required validation and verification could hobble the progression of next-gen technology: “I think we need to think very hard about how we validate these diagnostic tests. We are struggling with new mental models and approaches.”

Right now, he notes, “we can spend months validating a test for a specific mutation in a specific disease or virus we’re looking for. As we go forward, if we spend the same amount of time validating tests for each individual gene in 2015 or 2020, I think we’ll never be able to translate this technology into new advances in patient care.”

One of the factors pushing the development of panels is the danger of too much data. “People are designing tests to get specific things they want answers to, and they don’t really want to know the rest,” Dr. Triche explains. “This is important because the FDA is starting to ask, Should you report every variant you find, or not? If you found something not disease associated and later found that mutation was disease associated and there were bad outcomes, are you liable for not reporting it? Maybe the thing to do is just ask specific questions and report only on those. For example: Is the p53 gene mutated, yes or no?” From the FDA’s standpoint, he believes, that’s probably a much better idea.

Most people don’t fully comprehend the implications of having a full genome sequence, Mayo’s Dr. Smith says. “There’s information in there that makes sense and information that makes essentially no sense. For example, when they sequenced James Watson’s genome as part of the Human Genome Project, he had five lethal homozygous mutations; theoretically he should have been dead. So there are complexities we don’t yet understand.” If somebody offered to sequence Dr. Triche’s entire genome for free, he says he would turn them down: “Many people are beginning to realize it’s no bargain to have a bunch of genomic sequencing data without the sophisticated analysis required to understand it.”

Even for smaller sequencing projects, though, the technology is way in advance of our ability to deal with the data or with the implications of the data, Dr. Smith says. “The true crunch point is definitely bioinformatics. Not only do we need more biostatisticians developing more tools to make sense of the data, we need a new generation of researchers trained in informatics. And when we sequence about a million people and start correlating that with who’s sick and who’s not sick, and longevity, we may start having a better understanding of what it begins to mean.”

Dr. Versalovic, too, is concerned that the available analytical tools are no match for the firehose of data next-gen sequencing is bringing in. “NGS has not yet solved the data analysis problem. It’s actually made it worse, because the problem is the deeper we go right now, we can be spending months to analyze that data,” he points out. “It boils down to how we can distinguish signal from noise, and decide which fragments of information, out of this sea of data, will be important to make a diagnosis and manage the treatment of patients.”

But bioinformatics is a rapidly developing area, Dr. Versalovic notes. “In September a new board-certified specialty was announced by the American Board of Medical Specialties: clinical informatics. And pathologists are eligible for this.” Dr. Storch says that at Washington University there is a genome institute with funding from the National Institutes of Health, including a large computing center and bioinformatics people who are devoted only to understanding how to interpret NGS data.

To truly understand all the mutations, scientists are probably going to need a couple of decades, says Life Tech’s Stevenson. “But we can get started now on better treatments. Our understanding of stomach ulcers and what causes them wasn’t as great 30 years ago as now, but it didn’t stop us from starting to treat ulcers.”

The College plans to include clinical applications of NGS in a laboratory accreditation checklist, to a small degree by June 2012, and to develop appropriate proficiency testing challenges for NGS as well as standards and guidelines in conjunction with groups like the American College of Medical Genetics and the Association for Molecular Pathology. At the June FDA meeting, the CAP urged the agency to actively engage pathologists as NGS moves from research settings to clinical use.

That is happening because of a confluence of events letting us understand cancer and other diseases much better than we used to, says Dr. Triche. “We have multiple pieces of the puzzle all coming together very rapidly, and we have wonderful new tools and increasing knowledge of how to analyze the data. It’s just amazing how fast the field is moving. It’s an exciting era.”