NEW YORK – Experts say that reopening the US economy will require a massive scale-up in COVID-19 testing, contact tracing, and isolation of cases to reign in new outbreaks, which will include testing of asymptomatic individuals at risk of being infected. High-throughput sequencing-based diagnostic tests currently in development by academic and commercial groups could play an important role in providing this additional capacity.
And while antibody tests can tell if someone has previously been infected with the virus, and is likely to have developed some degree of at least temporary immunity, only diagnostic testing that is widely available can detect new infections quickly.
According to an estimate by the Harvard Global Health Institute, published in collaboration with National Public Radio last week, the US should have at least 900,000 SARS-CoV-2 tests per day available by mid-May, based on the size of the outbreak in early May. However, the reality looks much different, and the daily testing average over the past week has only reached about 280,000, according to the COVID Tracking Project.
The number of methods available for diagnostic COVID-19 tests has grown in the past few weeks, with recent Emergency Use Authorization authorizations from the US Food and Drug Administration for a CRISPR-based assay from Sherlock Biosciences and an antigen test from Quidel, but the vast majority of tests is currently still conducted by quantitative PCR (real-time PCR).
Despite the wide availability of qPCR instruments from different manufacturers in clinical laboratories today, the technology has inherent limits when it comes to scaling. According to one expert's estimate, a 384-well qPCR thermocycler can process a maximum of 640 patient tests per day if run in two 10-hour shifts, and increases in testing volume would require the installation of additional instruments, which are costly and may not be available.
Next-generation sequencers, with their ability to produce massive amounts of data per run and to multiplex thousands of barcoded samples, could offer a possible solution, provided that enough samples can be collected, shipped, and prepared to match their throughput. Despite their longer turnaround time compared to a qPCR assay – on the order of eight hours to several days – NGS machines could immensely boost the number of samples processed over time.
Several groups have been working on sequencing-based testing solutions, and many are planning to submit their designs to the FDA for emergency use authorization in the near future.
These fall into two broad categories: On the one hand are testing schemes that, like existing qPCR tests, only look at one or a couple of small regions in the viral genome and use the sequencer as a kind of detector that provides a yes-or-no answer. This approach is explored by a number of laboratories — including the University of California, Los Angeles; the University of Washington; the HudsonAlpha Institute for Biotechnology; the New York Genome Center; Ginkgo Bioworks; and Helix — that are working with a method called SwabSeq, developed by synthetic biology startup Octant.
A similar massively parallel diagnostic sequencing assay called Dx-Seq is under development by a group at Pennsylvania State University, which is currently validating it with collaborators on clinical samples. A third approach, called LAMP-Seq, developed by researchers at the Broad Institute, combines loop-mediated isothermal amplification with sequencing and also promises massive scalability.
On the other end of the spectrum are sequencing-based COVID-19 tests that cover the entire coronavirus genome, or even go beyond it by including the genomes of other respiratory viruses in a panel. While these tests can't provide the same throughput as the more targeted sequencing assays, they can get around some of the supply chain issues of qPCR tests. In addition, they can provide valuable insights into the origin of an infection or about co-infections. This approach is pursued by groups that include Baylor College of Medicine, Fulgent Genetics, Helix, HudsonAlpha, and customers of Illumina and IDbyDNA, who have co-developed a respiratory sequencing panel.
Octant, a synthetic biology-based drug discovery company in the San Francisco Bay Area, put its SwabSeq protocol out on the web more than a month ago. According to Sri Kosuri, the firm's cofunder and CEO, Octant had originally developed the high-throughput RNA amplicon sequencing method for its own work on human cell cultures but realized that it could be easily repurposed for COVID-19 detection. "We've never done diagnostics in the past," said Kosuri, who is currently on leave from his post as associate professor of chemistry and biochemistry at UCLA to work full-time at the company. And rather than turning the company into a COVID-19 diagnostic lab, he and his colleagues are helping others to implement the SwabSeq assay in their labs.
SwabSeq uses a single-step RT-PCR reaction of a few amplicons, followed by sequencing, and does not require RNA extraction. It is highly scalable – up to 10,000 samples per day without automation and 100,000 samples per day or more with automation – and has an estimated consumables cost of under $1 per sample. A single Illumina MiSeq run could process 500, a NextSeq run 5,000, and a NovaSeq run with S2 flow cells 50,000 samples.
While it takes about 12 hours to perform, the test has good sensitivity, with a limit of detection of one to six molecules per test, according to the firm. The current design uses just one viral amplicon and one human control amplicon, but it could be modified to include more viral amplicons.
At the heart of the method lies an internal standard, a spike-in RNA molecule that has the same sequence as the viral amplicon except for six bases, so the sequencer can easily tell them apart. "We put a known number of molecules into each well, usually a few hundred, and then we run the assay," Kosuri explained. This turns the test into a ratiometric assay, allowing the amount of virus in the sample to be quantified without the need for normalization.
Octant has shipped the spike-in RNA to about a dozen sites, a mix of academic and commercial labs, and more than half of them have shown that it works in their hands, including with clinical samples, Kosuri said. His team has organized a Slack group to discuss the scale-up of COVID-19 testing, with a sub-channel for SwabSeq, holding weekly online user meetings to discuss updates.
Collaboration and sharing of data and ideas between the research groups has been exceptional, he said. For example, Lior Pachter, a computational biologist at Caltech, recently "rewrote our entire computational backend," Kosuri said, speeding it up significantly.
From a technical viewpoint, "there is no reason why [SwabSeq] won't work at scale," he said. "I think the real challenge is going to be logistics." Collecting 10,000 swabs for a single sequencing run, for example, will not be an easy feat when many samples today are still nasopharyngeal swabs that need to be taken by a healthcare professional.
But the assay, despite its name "SwabSeq," could in principle also work with saliva samples. Also, Jay Shendure's group at the University of Washington has been testing it with dry nasal swabs, a sampling alternative his team recently described in a BioRxiv preprint.
Jason Gehring, a postdoctoral fellow in Shendure's group who has been working on the SwabSeq assay, said that the dry swabs, which require no RNA extraction, "could get around a lot of the bottlenecks of testing without even using sequencing."
"We see the dry swabs as a stepping stone to get our throughput really pretty high," he explained. Once dry swabs or other sampling methods make sample collection no longer a limiting factor, and qPCR machines become the bottleneck, then the SwabSeq assay could kick in and increase throughput even further.
Most likely, this will be the case when testing goes beyond symptomatic patients and includes healthcare and essential workers, or other businesses testing their workforce on a regular basis. "If we want to reopen the economy, we need additional testing of asymptomatic people to clear populations of the disease, and also, potentially, for surveillance as well," Gehring said. "It is diagnostic testing. You want to be able to report the results to the patients, but it's a different goal than the clinic."
One challenge with SwabSeq, he said, is that the same amplicon is sequenced in every sample. As a result, the sequencer detects the same color across the entire image and has trouble telling individual DNA clusters apart, which reduces the sequence quality. However, this problem has been known for many years and there are "a lot of tricks to mitigate that issue," he said.
Gehring said the UW team eventually plans to seek regulatory authorization for the SwabSeq assay but is still making improvements to the test and continues to collect data about its performance. For example, it plans to look at multiple viral regions instead of just one amplicon, which might increase the sensitivity of the assay.
The HudsonAlpha Institute for Biotechnology is also looking into implementing SwabSeq for diagnostic testing. "It really is a very elegant assay," said Shawn Levy, CSO of HudsonAlpha Discovery, and has good sensitivity. "We're quite enthusiastic about the results so far." The key motivation for using SwabSeq, he said, is that it is sensitive enough to detect the virus and extremely scalable at the same time, at a low cost per sample.
His lab, which in parallel is developing a more comprehensive diagnostic viral sequencing assay, is "deep into the optimization" phase but is not rushing to put in an EUA application, though he said the plan is to do so within weeks rather than months. He added that other labs are close to submitting pre-EUA or EUA applications for sequencing-based COVID-19 tests. "We're waiting and seeing what the comments are."
In the meantime, HudsonAlpha's clinical lab has already implemented a qPCR-based COVID-19 diagnostic test that uses the CDC assay design. While that test is not broadly offered, Levy said, it is important for validating the NGS-based tests. It is certainly faster – a few hours compared to at least eight hours for an NGS test – but a sequencing test "could potentially do thousands of samples."
He agreed that sample collection will be the limiting factor, and his lab is looking into alternate sample types, including at-home collection, as well as extraction-free processing. "That's by far the biggest challenge, the front-end sample logistics," he said, "and, realistically, another reason why we're not necessarily rushing forward in terms of trying to build huge scale and huge capacity. But we want to at least be familiar with what works, what the sensitivities are, what the scalability of those assays is, so we can be prepared for things coming in the future."
Companies with existing sequencing capacity are also interested in developing high-throughput diagnostic COVID-19 assays and have taken notice of SwabSeq. Last week, for example, synthetic biology firm Ginkgo Bioworks put out a white paper on "How to Deploy Millions of COVID-19 Tests Per Day" that lists different test technologies.
According to Keith Robison, a principal scientist at Ginkgo, the firm is interested in developing a sequencing assay that could provide higher throughput than qPCR assays, while avoiding the supply chain issues associated with them. "We're still at the broad front stage where we're trying a lot of different approaches," he said. The firm is also in the process of getting its laboratory CLIA-certified and has filed a pre-EUA with the FDA "just to start understanding what the landscape is," since there are currently no sequencing-based COVID-19 assays that have an EUA.
While Ginkgo is not planning on morphing into a diagnostics company, it is serious about providing a diagnostic NGS test for COVID-19 and plans to submit an EUA at some point. It already has expertise processing thousands of samples per day, he added, though so far not clinical samples.
"We're still trying to figure out what's the right strategy, what's the best way to test large numbers of people," Robison said, and SwabSeq is one approach the company is considering. "It's almost like taking a qPCR assay and converting it into a barcoded sequencing assay," he said.
He agreed that sample collection is an unsolved issue. "It is no good to have a huge testing capacity if your sample collection capacity is one percent of that," he said.
Likewise, population and consumer genomics firm Helix has been looking to harness its fleet of Illumina sequencers for COVID-19 testing, banking on its expertise with running clinical NGS-based tests at scale. "We are looking at both the Octant protocols and a whole viral genome approach," said Helix Cofounder and Chief Business Officer Justin Kao in an email. The focus is on highly sensitive diagnostic testing to enable "back to work" applications, he said, with viral surveillance as a secondary goal. Meanwhile, the company is also bringing a qPCR-based COVID-19 test online to support its health system partners. Helix plans to submit an EUA application for its RT-PCR test first, following by the NGS test, and has been in dialog with the FDA about these.
In addition, the company has been looking into saliva self-collection, an approach with which it has a lot of experience, to potentially boost the number of samples for testing. In collaboration with Renown Health and the University of California, San Diego, Helix recently completed a study comparing saliva and nasopharyngeal swabs in a community diagnostic cohort and a convalescent cohort.
"Unfortunately, our data shows that saliva is significantly less sensitive than NPS for diagnosing mildly symptomatic patients in the community setting and for monitoring recovering patients for return to work," Kao said, possibly due to a lower viral titer in saliva or in less symptomatic, yet contagious, patients. "Therefore, we do not believe at this time that saliva will be the answer and are actively working on other self-administered solutions such as anterior nasal swabs," he said.
Howard Salis, an associate professor of biological and chemical engineering at Penn State University, and his group have been developing another high-throughput diagnostic sequencing assay for COVID-19, called Dx-Seq. The assay, a first version of which is described in a white paper, requires RNA extraction and combines barcoded reverse transcription, PCR, and sequencing to test up to 19,200 patient samples in a single workflow. It generates three amplicons: two viral ones and a human control. Following the RT reaction, the cDNA is pooled and amplified in a single PCR reaction to produce a sequencing library. Thus, the test requires just a single PCR machine and a next-gen sequencer.
Since putting out the protocol, Salis said his team has experimentally validated it in several runs and is now clinically validating it with collaborators. It is also preparing a preprint. "Besides the massive scalability of the assay itself, we're also putting great effort into developing and validating automated workflows using Opentrons robotic workstations," he said in an email. "Within a single test facility, automation is really the only way to process 19,200 samples each and every day."
Meanwhile, researchers led by Feng Zhang and Aviv Regev at the Broad Institute have developed yet another high-throughput sequencing-based diagnostics COVID-19 assay, called LAMP-Seq.
A month ago, they posted a BioRxiv preprint describing the assay, which uses reverse-transcription loop-mediated isothermal amplification (RT-LAMP) of parts of the viral genome with sample-specific barcodes, followed by sequencing of pooled samples.
The approach requires a single heating step and no RNA extraction and has an estimated reagents cost of under $7 per sample, which could be further reduced through optimized production of the key enzyme, according to the authors. "Most importantly, this approach is predicted to be scalable to hundreds of thousands of samples per day per sequencing facility," they wrote. What still needs to be established is the sensitivity of the RT-LAMP reaction in swab samples and the stability of the products during shipment, they added. "Given the urgency of the present need, we hope that sharing this protocol early, even in the absence of clinical validation data, will spur additional development and parallel testing."
The developers also set up an online forum where interested researchers and collaborators can discuss the assay. In an email, lead author Jonathan Schmid-Burgk said that he and his colleagues are currently validating LAMP-seq in collaboration with researchers at University Hospital Bonn in Germany, using paired dry swabs, and will update their preprint shortly with the results. "We have a promising protocol but have to further optimize it," he said, before promoting it for clinical use or applying for FDA EUA. "Our current goal is to prototype large-scale testing in Germany," he added.
Amid these various efforts to harness next-gen sequencing for high-throughput COVID-19 diagnostics, the Broad Institute's own CLIA-certified sequencing lab remains committed to testing by qPCR – at least for now.
In March, the lab quickly transformed itself into a high-throughput COVID-19 testing facility, adapting the CDC's qPCR protocol and implementing it on its liquid handling robots. The lab now has a daily capacity of around 10,000 tests but only conducts about 3,000 to 4,000 tests per day because of a lack of incoming samples, according to Stacey Gabriel, senior director of the Broad's genomics platform. "A switch to sequencing would not solve our bottlenecks. It's just a different way to detect," she said.
At least for her lab, moving to a sequencing-based test is currently not warranted because it would not change the lack of samples. "There are all kinds of things pre-lab that are really what stop our capacity from being fully utilized," she said, including the availability of swabs, how tests are ordered, and who they are ordered for. "Hopefully that's going to change over time."
As hospitals are bringing online their own tests, she said, her lab has seen more demand for testing frontline and essential workers more systematically and has started doing a lot of testing in nursing homes in Massachusetts.
It is also exploring back-to-work testing for different industries but has not implemented that yet. "We're having a lot of discussions right now, including with our own institute, and many others across the state of Massachusetts," she said. The lab is also working with genetic testing company Color to help expand access to testing further.
"For now, our qPCR test with automation and the supply chain seems scalable to meet our needs. It's a really robust assay and we're sticking with that for now and keep an eye on how other things evolve," she said. "Maybe it seems ironic to people because we're known as an NGS lab. But we were really just trying to build what was the most practical [test] and what was needed at the time, and it seems to be working very well."