Making Contact with Sequencing's Fourth Generation

Abstract

BioTechniquesVol. 50, No. 2 Special News FeatureOpen AccessMaking Contact with Sequencing's Fourth GenerationJeffrey PerkelJeffrey PerkelSearch for more papers by this authorPublished Online:28 Jun 2018https://doi.org/10.2144/000113608AboutSectionsPDF/EPUB ToolsAdd to favoritesDownload CitationsTrack Citations ShareShare onFacebookTwitterLinkedInRedditEmail Today's next-generation sequencing systems rely on optical detection to decode nucleotide sequences. But as Jeffrey Perkel reports, a new generation of optics-free sequencing instruments could result in longer reads, faster runs, and lower costs.Next-generation DNA sequencing certainly had a season for the ages in 2010. Genomes were felled by the technology on a weekly basis, it seemed, including the turkey, the strawberry, Xenopus, and the Neanderthal—and hundreds upon hundreds of humans. Most recently, the technology earned worldwide attention for its breathtaking deciphering of the Haitian cholera epidemic.The technologies underlying those successes represent just a handful of companies: Illumina, Complete Genomics, Helicos, Roche/454 Life Sciences, Life Technologies, and most recently, Pacific Biosciences, whose first instruments rolled out to beta testers in November. Though each of these companies' sequencers operates on a different principle, they have one thing in common: all rely on optical detection of nucleotide incorporation. Such detection schemes require high-priced components such as lasers, cameras, lenses, and the like, not to mention complicated image-based base-calling algorithms. They also require specialty reagents to run, adding significantly to the cost of ownership.Yet there's more than one way to sequence a genome, as they say, and not all of them involve optics. In December, Life Technologies officially launched the first commercial sequencing instrument to eschew optical detection. The Personal Genome Machine (PGM), the product of Life Technologies' recent $375 million acquisition of Ion Torrent, monitors nucleotide incorporation electrochemically. It's what Chad Nusbaum—co-director of the Genome Sequencing and Analysis Program at the Broad Institute of Harvard and MIT—calls a “post-light” sequencing instrument, and it's likely just the first of many. Indeed, waiting in the wings, other companies are pursuing detection strategies based on nanopores and electron microscopy.Promising longer reads, single-molecule detection, smaller and simpler instruments, and lower costs, these nonoptical methods could fundamentally change the sequencing landscape. According to Harvard geneticist George Church, human genome sequences still cost at least $7000 to produce, but in order to be clinically useful, “probably should get down to below $700. There's probably another factor of 10 at least to be squeaked out.” Plus, he says, there are new niches to explore that current instrumentation cannot handle—cell phone–sized, real-time environmental monitors, for instance. “That seems far-fetched,” says Church, “but where we are right now would have seemed far-fetched a few years ago.” But these new technologies will first have to prove they can make the hop from the drawing board to the marketplace.Sensing the ProtonsThe first post-light sequencing system has already made that hop. The heart of the Ion PGM is a semiconductor chip studded with over one million wells, each containing one template and a DNA polymerase. As nucleotide triphosphates flow over the chip one at a time, the system registers incorporation events by the concomitant release of a proton.“Essentially, we've made the smallest pH meter in the world,” says Jay Therrien, vice president of commercial operations for next-gen sequencing at Life Technologies.According to Maneesh Jain, vice president of marketing and business development at Ion Torrent, the Ion PGM's initial “314” chip will contain 1.4 million sensors and cost $250 (plus another $250 per run for reagents). Out of the gate, that chip will generate at least 100,000 reads per run with an average length of 100 base pairs and a consensus accuracy (on 6× coverage) of 99.99%. That's 10 million bases per run: “significantly greater” than what 5 or 10 classic ABI 3730s could produce, says Therrien, but a mere fraction of what an ABI SOLiD could churn out. A newer “316” chip, slated for release in 2011 with some 7 million sensors, should boost throughput to 100 million bases per run. And the throughput can go up from there, says Jain, as semiconductor fabrication and molecular methods improve. In December, Life Technologies announced three $1 million “Grand Challenges” to get the ball rolling. The goal: enlisting researchers and programmers to improve Ion Torrent's speed, throughput, and accuracy.Ion Torrent's Ion 314 semiconductor sequencing chip will initially generate at least 100,000 reads per run with an average length of 100 base pairs.Courtesy of Ion Torrent.With a price tag of $50,000, about one-tenth that of its higher-throughput next-gen cousins, the PGM could “democratize” next-gen sequencing, according to the Broad Institute's Nusbaum. “There are a lot of people in the world who don't have access to ‘heavy lifting' sequencers,” Nusbaum says. “They simply have no opportunity to raise the money to buy an Illumina or SOLiD machine, and even if they did, they wouldn't be able to support it on the budget that they have.”But it's about more than just price, says Nusbaum; by enabling small research groups to sequence on demand, the PGM would allow almost anyone to become a mini-sequencing core facility. “The democratization isn't so much of a democratization about price, it's democratization of access.”The system doesn't provide complete democratization, of course: Even the 316 chip won't produce enough sequence for applications such as whole human genome sequencing, says Nusbaum, whose facility is beta-testing the device. Instead, the system is appropriate for specific small-scale applications like microbial genome sequencing, microbial transcriptome analyses, and targeted and amplicon sequencing. At the Broad, the Institute's four PGMs are used for library quality control and accelerating technology development cycles. “Why do I care about this machine right now?” Nusbaum asks.“The short answer is that it's fast—very fast.” In just a couple of hours, he says, and for just a few hundred dollars, researchers at the Broad can quality-test sequencing libraries that normally take a week and a half or longer—as well as thousands of dollars—to run.“It's not glamorous, but it's something that the high-throughput centers really want to do,” he says. “We spend a lot of our money sequencing really complex pools, so everything we do, every step that we have, we want to put quality control in place. This is a good thing for that.”Sequencing's Electron MicroscopeOther in-development methods dispense with the sequencing-by-synthesis approach altogether. One alternative—sequencing single molecules directly using transmission electron microscopy (TEM)—is being developed at Halcyon Molecular in Redwood City, CA and ZS Genetics in North Reading, MA.It won't be easy. The technique's resolution is sufficient to resolve individual DNA bases, but DNA is more or less transparent to TEM; the nuclei present in nucleic acids simply aren't dense enough to reflect the instrument's electron beam. EM-based sequencing therefore requires a contrast agent. ZS Genetics uses an enzyme-driven DNA amplification step to incorporate high–atomic number (Z) labels, distinguishing bases by their differential contrast in the EM. Halcyon uses chemical reactions to incorporate such high-Z agents as Os-BIPY and “monofunctionalized cisplatin variants,” tagging molecules with one label at a time and leveraging base complementarity rules and bioinformatics to stitch the resulting single-base patterns into a complete sequence, says the company's Chief Visionary Officer, Kent Kemmish.An artist's representation of DNA passing through a grapheme nanopore.Courtesy of Paul Montie Design.The resulting sequence reads could be among the longest of any method, says Church, who advises 18 next-generation sequencing players. Kemmish says Halcyon is targeting reads in excess of 100,000 bases apiece. ZS is aiming for the 10,000–20,000 nucleotide range, says president William Glover. “We think we will end up with a system that's taking 3–10 [images] per second, and will give 10–50 megabases per hour—and that would be one of our earlier systems,” Glover says.At those lengths, EM-derived sequences could enable de novo sequencing even of highly repetitive regions that traditionally have been refractory to such methods, says Glover—everything from a 4-kb stretch of the herpes simplex virus to the vast major histocompatibilty region on human chromosome 6. The problem with current short-read technologies, Glover explains, is that eukaryotic repeat elements typically are longer than the reads, making de novo assembly impossible. A long-read technology, though, can collect enough sequence to span the repeats, thereby enabling unambiguous alignment. “If you could give me thousands-of-bases-long reads of even modest accuracy, those would be very powerful for looking at genome structure,” Nusbaum says.But sequencing by EM is not a trivial exercise; the technology was tried and abandoned in the 1960s and '70s, over one key problem: spreading. Samples aren't simply poured onto the TEM sample grid like agar plates; instead, the DNA strands must be arrayed just so to ensure base-to- base spacing is uniform enough to provide reliable reads—to know, for instance, that a gap of x angstroms corresponds precisely to y nucleotides. “There is some art to it,” Glover says. But both companies now claim to have overcome this problem. Halcyon, for instance, has developed a technique Kemmish calls “molecular threading.” “We're able to take individual DNA molecules and actually pull them into air and place them with nanometer precision onto arbitrary substrates,” he explains.Though neither firm has actually published or publicly released proof-of-principle data, both claim to be making good progress. “We are happy with our internal data,” Glover says.The ‘Hole' SequenceOf all the post-light methods, perhaps none generates more excitement than nanopore sequencing. In development by Oxford Nanopore Technologies, NABsys Inc., and most recently, a partnership between Roche and IBM, nanopore sequencing has made significant strides over the past 18 months or so, says Nusbaum.Oxford is pursuing two parallel strategies: “Exonuclease sequencing” (for which the company has entered into a commercialization agreement with Illumina), based on exonuclease digestion of a single-stranded template into nucleotides that are fed into a nearby protein nanopore in a lipid membrane, and “strand sequencing,” which is like feeding thread through the eye of a needle. NABsys employs a variant of this latter strategy, identifying regions of known sequence by hybridizing short nucleic acid probes to a single-stranded template and monitoring when those duplexes pass through a solid-state semiconductor pore. In both cases, detection occurs by monitoring the blockage of current through the pore as individual bases pass through.In 1996, Harvard cell biologist Daniel Branton and David Deamer of the University of California, Santa Cruz, first demonstrated the ability to detect DNA as it was threaded through nanometer-sized protein pores based on its impact on ion current. In that paper, the pair listed five conditions required to convert such a system into a sequencer (1); in 2010, Branton says, these conditions were finally met.In one key recent study, Slaven Garaj, with Branton and Jene Golovchenko of Harvard and Jing Kong of MIT, demonstrated that a one- to two-atom–thick sheet of graphene (a novel form of carbon whose discovery won the 2010 Nobel Prize in Physics) could function as a DNA nanopore, thereby solving one problem Branton and Deamer had noted: “the limiting aperture of the channel must have appropriate dimensions to reflect the presence of only one nucleotide at a time” (1). Other nanopores, such as the protein nanopores in development at Oxford and the solid-state pores at NABsys, are relatively thick—at least 5 nm long, according to a 2008 review—meaning that “at least 10–15 nucleotides of ssDNA” may occupy the pore simultaneously (2). As Garaj explains, the problem is like running your finger along a strand of pearls: With a “very tiny finger,” the pearls can be distinguished easily. With big, thick fingers, “you cannot really see what the shape of each pearl is,” he says.A second study, published in December by Mark Akeson and colleagues at UCSC, addresses another problem. Polynucleotides can zip through nanopores at up to a million bases per second, but this is far too rapid for them each to be read. Akeson's team used the phi29 DNA polymerase to “ratchet” a single-stranded molecule through an alpha-hemolysin pore at about 25 bases per second. “You're getting a situation where you can introduce single bases at the [pore's] constriction point sequentially, in a time scale that is compatible with the technology,” says a spokesperson for Oxford Nanopore, which collaborates with Akeson.Other recent studies address alternative sequencing and detection strategies. In September, a team led by Jens Gundlach of the University of Washington detailed a new genetically tweaked protein nanopore, MspA, whose “constricted region” (i.e., the “finger” that interrogates the nucleotide “pearls”) is 10 times narrower than alpha-hemolysin. The authors used that pore to develop a novel sequencing strategy called “duplex interrupted,” which leverages the fact that dsDNA pauses the analyte DNA in the pore long enough for individual bases to be read. Using that method, the authors then sequenced a five-nucleotide-long stretch of DNA.And in November, Stuart Lindsay at Arizona State University published data suggesting that bases could be distinguished in the context of an intact molecule by quantum mechanical electron tunneling across the pore, rather than by blocking current through the pore. That's the theory, anyway; Lindsay's study didn't actually use a nanopore, but rather a functionalized gold surface and probe to scan DNA molecules as they drifted into the narrow gap between these two contacts. Now, says Lindsay, comes the hard part: integrating such electrodes with a nanopore and making them work. “That's obviously the focus of the current research,” he says.Nevertheless. for Nusbaum, these and other studies have combined to advance nanopore sequencing from the realm of the fantastical to the practical. What's left, effectively, is systems integration. “Never before have I felt like saying nanopores could work. Nobody is really sequencing DNA yet by a nanopore, but I think it's been shown that it should be doable.”If and when that happens, he adds, “it could be a real game-changer. It will have the cheapness, speed, and simplicity of an Ion Torrent, but it may also have potential to have the kind of yield that you're used to having with the massively parallel technologies.” Branton says he anticipates someone will put all the pieces together into a “first primitive version” within the year.Of course, there's a long road between prototype and commercialization, and it won't be easy to break into a field already flush with technical powerhouses. So, for all the excitement surrounding post-light sequencing, don't count existing technologies out just yet, says Church; Life Technologies, Illumina, 454, and the like still have some juice left in them, too.Still, Church cites five niches that could help a new technology establish a beachhead: cheaper instruments, lower costs per run, longer reads, faster speed, and portability. “But to really capture the market,” he says, “it's going to be dollars per base pair.”