Last March, during the first wave of the pandemic, Adriana Heguy set out to sequence coronavirus genomes. At the time, New York City’s hospitals were filling up, and American testing capacity was abysmal; the focus was on increasing testing, to figure out who had the virus and who didn’t. But Heguy, the director of the Genome Technology Center at N.Y.U. Langone Health, recognized that diagnostic tests weren’t enough. Tracking mutations in the virus’s genetic code would be crucial for understanding it. “No one was paying attention to the need for sequencing,” Heguy told me recently. “I thought, I can’t just sit here and not do anything.” Within weeks, her team had sequenced hundreds of samples of the virus collected in New York City and published a paper with three key findings: the virus had been circulating in the city for weeks before the lockdown; most cases had come from Europe, not China; and the variant infecting New Yorkers carried a mutation, D614G, that scientists soon confirmed made it far more contagious than the original virus isolated in Wuhan.
Heguy’s efforts were prescient. The world is now confronting a growing number of coronavirus variants that threaten to slow or undo our vaccine progress. In recent months, it’s become clear that the virus is mutating in ways that make it more transmissible and resistant to vaccines, and possibly more deadly. It’s also clear that, at least in the United States, there is no organized system for tracking the spread or emergence of variants. As Heguy sees it, the U.S. has more than enough genome-sequencing expertise and capacity; the problem is focus. “Efforts in the U.S. have been totally scattered,” she said. “There’s no mandate to do it in a timely fashion. The government is kind of like, Let us know if you find something.” Funding has also been a major constraint. “It boils down to money,” Heguy said. “With money, I could hire a technician, another scientist, get the reagents and supplies I need.” Because of their better-organized efforts, other countries have been more successful in identifying new versions of the virus: “The reason the U.K. variant was identified in the U.K. is that the U.K. has a good system for identifying variants.” The U.K. has, for months, sequenced at least ten per cent of its positive tests. “If you’re doing ten per cent, you’re not going to miss things that matter,” Heguy said. “If a variant becomes prevalent, you’ll catch it.”
Heguy’s lab sequences ninety-six samples a week—as many as will fit onto a single sample plate, which has eight rows and twelve columns. The process—receiving, preparing, sequencing, and analyzing samples, then reporting the results—takes time and resources, and diverts attention from other research. “Mostly we do this out of a sense of moral obligation,” Heguy told me. “This feeling that the country shouldn’t be left in the dark.” As we enter what seems to be the endgame of the pandemic, tracking and analyzing variants—which could fill hospitals and reduce the effectiveness of therapies and vaccines—is more important than ever.
To understand coronavirus variants, you need to understand a little about viral biology and, more specifically, about how the fragments of RNA and protein from which viruses are made go about replicating. SARS-CoV-2, the coronavirus that causes COVID-19, has about thirty thousand letters of RNA in its genome. These letters, or “bases,” are like the architectural plans for the virus’s twenty-nine proteins, including the “spike” protein that it uses to enter cells. Once inside a cell, the virus hijacks the cellular machinery, using it to make copies of itself. Because the machinery is good but not perfect, there are occasional errors. SARS-CoV-2 has a mechanism that checks the new code against the old code; still, it’s possible for the substitution, deletion, or addition of an amino acid to evade this proofreading. If the errors don’t arrest the replication process completely, they sneak into the next generation. Most mutations don’t meaningfully change a protein’s structure or function. Sometimes, however, one of these accidental experiments “works.” A variant has been created—a virus with a slightly different design.
In the time that SARS-CoV-2 has troubled humans, it’s accumulated innumerable mutations. Those that matter have one of two key features: they either help the virus latch onto and enter cells more easily, or they allow it to better evade tagging and destruction by the immune system. Today, scientists are following three variants of particular concern: B.1.1.7, originally detected in the U.K.; B.1.351, from South Africa; and P.1, from Brazil. Predictably, variants seem to have emerged more quickly in countries with rampant viral spread—places where the virus has had more chances to replicate, mutate, and hit upon changes that confer an evolutionary advantage. The U.K.’s B.1.1.7 variant has spread to more than eighty countries and has been doubling every ten days in the U.S., where it is expected to soon become the dominant variant. Its key mutation is called N501Y: the name describes the fact that the amino acid asparagine (“N”) is replaced with tyrosine (“Y”) at the five-hundred-and-first position of the spike protein. The mutation affects a part of the spike that allows the virus to bind to cells, making the variant some fifty per cent more transmissible than the original; new evidence also suggests that people infected with it have higher viral loads and remain infectious longer, which could have implications for quarantine guidelines.
Both the B.1.351 and P.1 variants carry the N501Y mutation. They also have another, more dangerous mutation, known as E484K: a substitution of glutamate (“E”) for lysine (“K”) at the spike protein’s four-hundred-and-eighty-fourth position. This mutation diminishes the ability of antibodies—both naturally acquired and vaccine-generated—to bind to and neutralize the virus. Last month, South Africa halted use of the vaccine produced by AstraZeneca, citing evidence that it offers minimal protection against the B.1.351 variant that is now dominant in that country; a monoclonal antibody drug from Eli Lilly is also inactive against it. In the U.S., a number of homegrown variants are beginning to circulate, including some with the antibody-evading E484K mutation; in the U.K., B.1.1.7 has, in some cases, also acquired the mutation, becoming more like the South African and Brazilian variants.
There’s growing concern that B.1.351 and P.1 can infect people who’ve already had COVID-19. The city of Manaus, in Brazil, has faced a viral surge this winter, even though some three-quarters of its population is thought to have been infected by the original virus in the fall—a level at which herd immunity is believed to settle in. This suggests that the antibodies produced by the original virus have struggled to neutralize its successor. Lab tests examining blood from immunized people have shown that the Pfizer-BioNTech and Moderna vaccines—which are effective against the U.K. variant—tend to produce fewer antibodies that fight the South African and Brazilian variants. It’s not yet clear how this will affect real-world protection: the vaccines still elicit huge numbers of antibodies—probably more than enough to neutralize the virus—and they stimulate other parts of the immune system, such as T cells, that weren’t assessed in the blood tests. At least for now, a degree of uncertainty is inevitable.
How worried should we be about the variants? They pose a challenge, but, compared to the original vaccine-development effort, it’s small. Pfizer-BioNTech and Moderna have said that they can develop booster shots within six weeks that work against these variants; Moderna has already started working on one that targets the South African version. From a scientific perspective, developing variant-specific vaccines is a straightforward proposition—one simply swaps the new genomic material for the old. Testing, manufacturing, and distribution could still take months. But the F.D.A. has released guidance designed to streamline the approval process for coronavirus boosters, indicating that it will review them using roughly the same approach it employs for annual flu shots. This means that the new vaccines will likely be tested in small trials of several hundred people, as opposed to the larger randomized trials that were needed for initial approval of the vaccines. Instead of following trial subjects for months to see if they develop COVID-19, researchers will be able to use a blood test to determine if they are mounting an adequate immune response to the variant. The U.S. regulatory apparatus is evolving with the virus.
On January 6, 2020, Jason McLellan, a structural biologist at the University of Texas at Austin, was in Park City, Utah, waiting in a ski shop for his new boots to be heat-molded. His phone rang; it was Barney Graham, the deputy director of the Vaccine Research Center at the National Institute of Allergy and Infectious Diseases. McLellan had previously collaborated with Graham on projects to study the molecular structure of viruses such as RSV and MERS-CoV. After the conversation, McLellan sent his team a text: “Barney is going to try and get the coronavirus sequence out of Wuhan, China. He wants to rush a structure and vaccine. You game?”
In the coming weeks, McLellan and his team determined the structures of key proteins in the new coronavirus. They learned that SARS-CoV-2 had an “unstable” spike protein, capable of changing shape when it attaches to cells, and sometimes before. The immune system makes more effective antibodies against the initial, “prefusion” version of the protein. The trick, therefore, was to lock the protein in that state. Drawing on their work with MERS-CoV, McLellan and Graham introduced two mutations to stabilize the spike protein. Every successful COVID-19 vaccine developed in the United States works by presenting the immune system with the “locked” proteins that McLellan and Graham devised; the paper describing their work, published online last February in the journal Science, has been cited nearly four thousand times.
McLellan has been tracking coronavirus mutations and how they change the structure of the spike protein. For much of 2020, he told me, the protein seemed to accumulate a few mutations a month. Then, in December, variants began to emerge with as many as twelve mutations simultaneously. “We were like, Wow, how did this one variant get so many mutations all of a sudden?” he said. McLellan hypothesizes that, in addition to the usual factors—the passage of time, uncontrolled viral spread—certain individuals vastly accelerate the rate of mutation. “Some people aren’t able to eliminate the virus for a long time—sixty days, a hundred days,” McLellan said. “They mount enough of an immune response to not die, but not enough to get rid of the virus. That creates selective pressure. There’s an evolutionary experiment going on inside these people. The virus emerges with a bunch of changes, some of which improve its fitness.” Such individuals become not superspreaders but supermutators.
A growing body of evidence suggests that persistent infection within a person can greatly accelerate the speed with which the virus mutates. Last year, in Boston, a forty-five-year-old man with an autoimmune condition contracted the coronavirus. The man suffered labored breathing, fatigue, abdominal pain, a fungal infection, and extensive bleeding throughout both lungs and was admitted to the hospital six times; he was given the usual COVID-19 therapies—remdesivir, monoclonal antibodies, steroids—as well as other powerful immunosuppressants to treat complications of his autoimmune condition. All the while, his compromised immune system struggled to clear the infection. In total, he experienced a five-month illness. He died a hundred and fifty-four days after he was diagnosed, with the virus still circulating in his body.
Genetic analyses conducted at different points during the man’s illness revealed that the virus in his system had accrued a startling number of mutations. Dozens of genomic letters had changed or been deleted. The genes encoding the spike protein account for thirteen per cent of the virus’s genome, but had accumulated nearly sixty per cent of the observed changes, with most of these occurring in a region that allows the protein to bind to its receptor. Many scientists now suspect that the B.1.1.7 variant, which surfaced with nearly two dozen concurrent mutations in the U.K., emerged from immunosuppressed COVID-19 patients who were treated with therapies that exerted further selective pressure on the virus. (The South African variant, by contrast, appears to have evolved more gradually, suggesting that population spread was its dominant mutational force.)
Like all viruses, SARS-CoV-2 will continue to evolve. But McLellan believes that it has a limited number of moves available. “There’s just not a lot of space for the spike to continue to change in ways that allow it to evade antibodies but still bind to its receptor,” he said. “Substitutions that allow the virus to resist antibodies will probably also decrease its affinity for ACE-2”—the receptor that the virus uses to enter cells. Recently, researchers have mapped the universe of useful mutations available to the spike’s receptor-binding area. They’ve found that most of the changes that would weaken the binding ability of our antibodies occur at just a few sites; the E484K substitution seems to be the most important. “The fact that different variants have independently hit on the same mutations suggests we’re already seeing the limits of where the virus can go,” McLellan told me. “It has a finite number of options.”
Over time, SARS-CoV-2 is likely to become less lethal, not more. When people are exposed to a virus, they often develop “cross-reactive” immunity that protects them against future infection, not just for that virus, but also for related strains; with time, the virus also exhausts the mutational possibilities that might allow it to infect cells while eluding the immune system’s memory. “This is what we think happened to viruses that cause the common cold,” McLellan said. “It probably caused a major illness in the past. Then it evolved to a place where it’s less deadly. But, of course, it’s still with us.” It’s possible that a coronavirus that now causes the common cold, OC43, was responsible for the “Russian flu” of 1889, which killed a million people. But OC43, like other coronaviruses, became less dangerous with time. Today, most of us are exposed to OC43 and other endemic coronaviruses as children, and we experience only mild symptoms. For SARS-CoV-2, such a future could be years or decades away.
For now, tracking and analyzing variants remains vital. In July, a report on the state of genomic sequencing in the U.S., published by the National Academies of Sciences, Engineering, and Medicine, concluded that “genome sequence data are patchy, typically passive, and reactive in the United States.” Last year, the federal government organized two efforts to increase genetic surveillance; neither was particularly effective, and, in January, the U.S. sequenced less than one per cent of all positive coronavirus tests—placing it thirty-eighth in the world, behind Gambia, Vietnam, and Thailand, by proportion of tests analyzed. President Joe Biden has announced a two-hundred-million-dollar investment to bolster the country’s sequencing infrastructure; the C.D.C. has indicated that it hopes to sequence twenty-five thousand samples a week in the near future; and Biden’s COVID-19 relief plan, which passed the Senate on Saturday and will likely be signed into law later this coming week, will provide nearly two billion dollars to strengthen the country’s genomic-sequencing efforts.
Still, these improvements are yet to come. In January, New York City, where I practice, sequenced, on average, just fifty-five samples a day. In hopes of expanding its capacity, the city has convened a consortium of research institutions and is seeking to identify more partners. Much of the resulting effort will likely run through the N.Y.C. Pandemic Response Lab, created by Opentrons, a Brooklyn-based robotics company, whose technology is used to automate research functions and efficiently process samples in labs around the world. Since September, PRL has focussed on diagnostic testing; now it is turning its attention to sequencing, as well. In recent weeks, it has tracked the spread of the U.K. variant and identified New York’s first case of the South African variant. The lab has more than doubled its sequencing capacity every week for the past month and plans to expand its testing and sequencing efforts to cities around the country.
Effective vaccines, emerging variants, expanding testing—what does it all add up to? In September, I wrote about two models of infectious-disease control that can help us think about the fight against COVID-19. On the one hand, there’s the silver-bullet model, typified by the eradication of polio: vaccines for that disease were so effective that, within a few years, we had extinguished it entirely in the U.S. On the other hand, there’s the incremental, multipronged approach, which was used to tamp down tuberculosis. There is no silver-bullet vaccine for TB; instead, the disease has been beaten back slowly, over a long period, using a series of interventions, including better sanitation, contact tracing, masking, and therapies. In the days after we learned of the spectacular efficacy of the COVID-19 vaccines from Pfizer-BioNTech and Moderna, the polio model felt within reach. To an extent, it still is: universal vaccination would drastically reduce the damage of COVID-19, even if it doesn’t stamp out the coronavirus completely. But, given how easily SARS-CoV-2 spreads, how entrenched the virus has become, and how many people are skeptical of vaccines, the TB model remains relevant. We live in a liminal state, requiring progress on both fronts. Now, variants have further complicated the story.
Confronting the variants, we should be cautious but hopeful. They are a worrying development but not a devastating blow. Every coronavirus vaccine available in the U.S. appears likely to prevent the more concerning consequences of infection—severe illness, hospitalization, death—even for the new variants. (In South Africa, where B.1.351 dominates, Johnson & Johnson’s vaccine prevented a hundred per cent of COVID-19 deaths a month after inoculation.) Vaccinated people, therefore, should feel confident in the protection they’ve gained, and in the knowledge that booster shots, should they become necessary, can quickly be developed and approved. Even for those who have been inoculated, the risk of illness has not been, and may never be, eliminated—but it remains vastly lower than it was before vaccination, despite the new variants in our midst.
For millions of unvaccinated Americans, however, the variants pose a heightened danger. More transmissible variants mean that activities such as travel, shopping, socializing, and dining carry a higher risk of infection; if individuals infected by variants do become ill, they may be less likely to benefit from existing therapies. This spring, people who haven’t been vaccinated—the vast majority of Americans—have reason to be concerned. The variants may well provoke another viral surge, especially as governors rush to reopen states and discontinue mask mandates. With the addition of a third coronavirus vaccine, the U.S. should have enough supply to immunize every American adult by the end of May. The emergence of variants is a reason to strengthen, not weaken, public-health measures—surveillance, masks, distance, isolation—until widespread vaccination has been achieved.