Any diagnostic test worth its salt has precision on its side. It can pinpoint the presence of this condition, but not that one; it can, when used for an infectious disease, distinguish between microbes that look very much the same.
For most of the pandemic, that exactness has been a major asset for the hundreds of tests that detect SARS-CoV-2. But the discerning nature of most tests has also opened up a weak point for the coronavirus to exploit. With the virus mutating into new and concerning variants, a few of the tests designed to recognize its original iteration are now getting duped. What was once a singular target has split itself off into many, many bull’s-eyes, each a little different from the next, and we’re having trouble taking aim.
This isn’t yet a crisis, and perhaps it never will be. Most tests are still performing very well against all known versions of the coronavirus, and researchers have work-arounds that will buttress others against the virus’s shape-shifting. But keeping tabs on test performance is crucial, especially while the pandemic rages on in many parts of the world. “From a global perspective, it’s absolutely necessary to do what we can to make sure we have a good grip on the virus and its mutations,” Neha Agarwal, the associate director of the PATH Diagnostics Program, an organization that has been tracking SARS-CoV-2 tests, told me. As long as it has hosts to infect, the coronavirus will continue to rejigger its genes, which means that test manufacturers will need to closely track the virus’s movements and tailor their products to follow it. More than a year out from the time when experts first got eyes on the virus, tests are in need of a touch-up.
Most of the coronavirus tests deployed in the United States detect specific stretches of RNA, the genetic material of the virus’s genome, usually chosen because they’re unique to SARS-CoV-2 (or at least the virus family it’s in). When the tests fail, it’s because they’re picky. These molecular tests search the genomic manuscript with about as much precision as the Ctrl+F function on a computer, which means that even single-letter typos—that is, simple RNA mutations—can discombobulate them.
According to the FDA, almost none of these tests is actually pinging back variant-related false negatives, with perhaps the exception of the Accula, made by Mesa Biotech. A document from the company says the test can occasionally be stumped when it encounters mutations in a gene called N (which produces the nucleocapsid protein), leading it to mistakenly say that no virus is present at all. But that’s an extreme case. Three other molecular tests identified by the FDA as being affected by mutations are still able to at least partially register the pathogen. (At least two more recently identified by researchers may soon join the list of tests whose detective powers are weakened, but not obliterated, by variants.)
One test on the FDA’s list, Thermo Fisher Scientific’s TaqPath, targets a segment of the S coronavirus gene (which encodes the spike protein). A snippet of that segment is missing from several variants of concern—including the very contagious Alpha (B.1.1.7), the dominant form of the coronavirus in the U.S.—rendering S effectively invisible to the TaqPath. But most molecular tests, including the TaqPath, have a de facto insurance policy: They generally scan the genome for multiple RNA segments at a time—two, three, sometimes more—making it nearly impossible for the virus to elude the test’s scrutiny entirely. The TaqPath, for instance, detects two additional gene segments outside of S, both of which are intact in Alpha, and will still spit out a positive result.
A slightly different set of issues is now playing out with antigen tests—a type of rapid test that can usually be done outside a lab—which detect coronavirus proteins. While molecular tests essentially scan genomes letter by letter for precise spelling, tests that search for proteins work more like a reader skimming words for overall meaning. Typos might slip by unnoticed, making antigen tests tougher to flummox with minor mutations. However, while molecular tests typically have multiple targets, antigen tests tend to have only one, usually the nucleocapsid protein, which makes them more “brittle,” says Alex Greninger, the assistant director of the clinical-virology laboratories at the University of Washington Medical Center.
In a recent paper, not yet published in a scientific journal, Greninger and his colleagues found that a common nucleocapsid-hunting antigen test called the Sofia, made by Quidel, might not recognize a very small fraction of coronavirus variants, incorrectly marking infected samples as virus-free. Greninger told me that the test-confounding mutation is present in less than 0.5 percent of SARS-CoV-2 genomes cataloged to date, so the test itself is fine for now. But the mistake it’s making isn’t necessarily an anomaly. Another recent study, also not yet peer-reviewed, claims a similar issue with a test called the PanBio, made by Abbott. The PanBio isn’t available in the U.S., but it’s similar to another test made by Abbott, the BinaxNOW, that has been authorized by the FDA.
(I reached out to the manufacturers behind several of the affected tests. Quidel and Mesa Biotech did not respond to repeated requests for comment. Manoj Gandhi, Thermo Fisher’s senior medical director for genetic-testing solutions, said the company was aware of the issue, but told me that the TaqPath wouldn’t need to be modified, because of its “built-in redundancy.” Kim Modory, a spokesperson for Abbott, defended its tests’ performance, noting that the company “conducted a thorough analysis of the new variants, and we are confident that our tests remain effective at detecting these strains.”)
Even if they’re rare, false negatives due to variants can have a ripple effect on our ability to contain the virus. Antigen tests have already stirred controversy because they aren’t as good as molecular tests at identifying infections, especially in people who don’t have symptoms or in whom the virus is somewhat scarce. False negatives could endanger the people who take these tests, and those around them, by wrongly reassuring them that they are infection-free. They can also help variants fly under the radar, skewing our sense of which versions of the virus are blazing through a given community. A positive antigen test usually means the coronavirus is present, but negatives are having a harder and harder time saying anything at all.
For that reason, many of the experts I talked with are pinning their hopes on the molecular approach to keep testing going as the coronavirus continues to splinter into new lineages. Now that scientists are clued in to the many distinctive patterns that certain variants produce on molecular tests, they’re using them as quick, preliminary screens to tease apart different versions of the virus. When the TaqPath picks up two familiar genes but fails to detect S, for example, researchers may flag the specimen as an Alpha candidate. What was initially viewed as a diagnostic fumble has become a valuable litmus test for mutation. “We took advantage of the failure,” Esther Babady, the medical director of the microbiology service at Memorial Sloan Kettering Cancer Center, told me. Other researchers, including Yale’s Chantal Vogels, have used similar strategies to develop tests that can tell Alpha from Beta (B.1.351) and Gamma (P.1). A few companies, including Thermo Fisher, are marketing products that can do the same.
The signatures that variants are registering on tests might someday make it possible for doctors to regularly deliver variant-specific diagnoses to their patients, instead of the binary “coronavirus” or “no coronavirus” conversations we deal in now. Those data have traditionally been funneled straight into public-health efforts to track the virus at the community level. But IDing certain variants that could elude antibody-based therapies could help guide treatments for individual patients as well. Still, experts stressed that standard coronavirus tests aren’t a reliable way to fully vet variants, which can be done only by sequencing the pathogen’s entire genome. Sequencing is also how researchers confirm the existence of new variants that might otherwise be mistaken for others or escape notice entirely. Testing and genome surveillance are inextricably linked: We can’t find variants if we’re not looking for them; we can’t design tests compatible with variants that we’re not aware of.
The arms race between virus and test isn’t a new story. The mutational capers of SARS-CoV-2 are actually pretty mild compared with the wild rides pathogens have taken us on before. The rise of the H1N1 flu virus in 2009 bamboozled several tests, leaving researchers and companies scrambling to retool their products, some of which failed entirely against the strain. Across the board, the experts I talked with were optimistic about our ability to target SARS-CoV-2 variants, especially given the dizzying array of tests American laboratories have to choose from. Even if a few eventually fail their way into obsoletion, we will have plenty of backups; manufacturers will keep tinkering with their tests to keep them relevant. “Everything is cool, honestly,” Greninger said. “For the foreseeable future, we’re going to have a panoply of diagnostics.”
More tests are already on their way, and molecular- and antigen-test manufacturers are intent on designing them to be long-lasting. For one, they’re focusing on bits of the virus’s genome that are less likely to mutate. Now that millions of SARS-CoV-2 genomes have been sequenced, identifying more steadfast regions of RNA should be easier. The tag-teaming approach—looking for multiple targets at once—is also a buffer against mutations. “The more targets there are, the more resilient the test is against changes,” Melissa Miller, the director of the clinical-microbiology laboratory at the University of North Carolina’s School of Medicine, told me. Doing this also helps tests detect just one pathogen at a time, in the same way that plugging several super-specific keywords into Google will narrow a search. Cepheid, whose two-gene Xpert tests are among those listed by the FDA as “impacted by SARS-CoV-2 mutations,” is adding a third target to its product to make it a bit more variant-proof, according to David Persing, the company’s chief medical and technology officer.
Still, it’s hard to ignore the game of whack-a-mole we’ve locked ourselves into. No test can be completely impervious to evolution’s hijinks. The coronavirus has infected more than 170 million people around the world, and will be with us for some time yet; it has had countless opportunities to morph into new forms, far faster than any test could be whipped up in a lab. The nature of tests—which search for pathogens after they’ve invaded the body—forces their design to be reactionary. The virus barrels on; tests give chase.
Some of those mutational moves will be zigzaggy and unpredictable. But experts assured me that researchers and companies are keeping close watch, tracking the virus’s many modifications and ensuring that tests are still up to snuff, similar to what’s been done for years with the flu. Testing is, at its core, about keeping the virus in our line of sight—even through a costume change or two.