Feature SOURCE: STRUCTURAL DATA FROM K. SHEN & J. LUBAN K. SHEN FROM DATA STRUCTURAL SOURCE: The spike of SARS-CoV-2 has a common (circled) that seems to shift the protein from a closed (left) to an open (right) form. MAKING SENSE OF CORONAVIRUS Different SARS-CoV-2 strains haven’t yet had a major impact on the course of the pandemic — but they might in future. By Ewen Callaway

hen COVID-19 spread around started scouring thousands of coronavirus In April, Korber, Montefiori and others the globe this year, David genetic sequences for mutations that might warned in a preprint posted to the bioRxiv Montefiori wondered how the have changed the ’s properties as it made server that “D614G is increasing in frequency deadly virus behind the pan- its way around the world. at an alarming rate”1. It had rapidly become demic might be changing as it Compared with HIV, SARS-CoV-2 is chang- the dominant SARS-CoV-2 lineage in Europe passed from person to person. ing much more slowly as it spreads. But one and had then taken hold in the United States, Montefiori is a virologist who mutation stood out to Korber. It was in the Canada and Australia. D614G represented a has spent much of his career gene encoding the spike protein, which helps “more transmissible form of SARS-CoV-2”, Wstudying how chance mutations in HIV help it virus particles to penetrate cells. Korber saw the paper declared, one that had emerged as to evade the immune system. The same thing the mutation appearing again and again in a product of natural selection. might happen with SARS-CoV-2, he thought. samples from people with COVID-19. At the These assertions dismayed many scientists. In March, Montefiori, who directs an 614th amino-acid position of the spike pro- It wasn’t clear that the D614G viral lineage was AIDS- research laboratory at Duke tein, the amino acid aspartate (D, in biochem- more transmissible, or that its rise indicated University in Durham, North Carolina, con- ical shorthand) was regularly being replaced anything unusual, they said. But alarm spread tacted Bette Korber, an expert in HIV evolution by glycine (G) because of a copying fault fast across the media. Although many news and a long-time collaborator. Korber, a compu- that altered a single nucleotide in the virus’s stories included researchers’ caveats, some tational biologist at the Los Alamos National 29,903-letter RNA code. Virologists were call- headlines declared that the virus was mutat- Laboratory (LANL) in New Mexico, had already ing it the D614G mutation. ing to become more dangerous. In retrospect,

174 | | Vol 585 | 10 September 2020 | Corrected 16 September ©2020 Spri nger Nature Li mited. All rights reserved. ©2020 Spri nger Nature Li mited. All rights reserved.

2 mm Montefiori says he and his colleagues regret GLOBAL SPREAD even potentially beneficial mutations might describing the variant’s rise as “alarming”. The By the end of June, the D614G mutation was found not flourish. “As far as the virus is concerned, word was scrubbed from the peer-reviewed in almost all SARS-CoV-2 samples worldwide. every single person that it comes to is a good 2 3 mm version of the paper, published in Cell in July . 1.0 piece of meat,” says William Hanage, an epi- The work sparked a frenzy of interest in demiologist at the Harvard T. H. Chan School 3 mm D614G. Even those who were sceptical that of Public Health in Boston, Massachusetts. 0.8 the mutation had changed the virus’s prop- “There’s no selection to be doing it any better.” erties agreed that it was intriguing, because of its meteoric rise and ubiquity. For months, 0.6 Faster spread? that lineage has been found in almost all When Korber saw the rapid spread of D614G, sequenced samples of SARS-CoV-2 (see ‘Global she thought she might have found an example 0.4 spread’). “This variant now is the pandemic. of meaningful natural selection. The muta-

As a result, its properties matter,” wrote of D614G* Frequency tion caught her eye because of its position in Nathan Grubaugh, a viral epidemiologist at 0.2 the spike protein, which is a major target for the Yale School of Public Health in New Haven, ‘neutralizing’ that bind to the virus Connecticut, and two colleagues in a Cell essay and render it non-infectious. And with 3 0 on Korber and Montefiori’s findings . Jan Feb Mar Apr May Jun the mutation were also rising in frequency in So far, the upshot of this work is less clear 2020 more than one part of the world. *Number of samples = 52,292

SOURCE: REF. 8 REF. SOURCE: than Montefiori and Korber’s preprint sug- D614G was first spotted in viruses collected gested. Some experiments suggest that mistakes. A typical SARS-CoV-2 virus accu- in China and Germany in late January; most viruses carrying the variant infect cells more mulates only two single-letter mutations per scientists suspect the mutation arose in China. easily. Other work has revealed possible good month in its — a rate of change about It’s now almost always accompanied by three news: the variant might mean that half that of influenza and one-quarter that of mutations in other parts of the SARS-CoV-2 can target SARS-CoV-2 more easily. But many HIV, says Emma Hodcroft, a molecular epidemi- genome — possible evidence that most D614G scientists say there remains no solid proof that ologist at the University of Basel, Switzerland. viruses share a common ancestor. D614G has a significant effect on the spread of Other genome data have emphasized D614G’s rapid rise in Europe drew Korber’s the virus, or that a process of natural selection this stability — more than 90,000 isolates attention. Before March — when much of the explains its rise. “The jury’s out,” says Timothy have been sequenced and made public (see continent went into lockdown — both unmu- Sheahan, a coronavirologist at the University www.gisaid.org). Two SARS-CoV-2 viruses tated ‘D’ viruses and mutated ‘G’ viruses were of North Carolina at Chapel Hill. “This muta- collected from anywhere in the world differ by present, with D viruses prevalent in most of tion might mean something, or it might not.” an average of just 10 RNA letters out of 29,903, the western European countries that geneti- Researchers still have more questions says Lucy Van Dorp, a computational geneticist cists sampled at the time. In March, G viruses than answers about coronavirus mutations, at University College London, who is tracking rose in frequency across the continent, and by and no one has yet found any change in the differences for signs that they confer an April they were dominant, reported Korber, SARS-CoV-2 that should raise public-health evolutionary advantage. Montefiori and their team1,2. concerns, Sheahan, Grubaugh and others Despite the virus’s sluggish mutation rate, But natural selection in favour of G viruses say. But studying mutations in detail could researchers have catalogued more than 12,000 isn’t the only, or even the most likely, explana- be important for controlling the pandemic. mutations in SARS-CoV-2 . But scien- tion for this pattern. The European dominance It might also help to pre-empt the most tists can spot mutations faster than they can of G variants could be simply down to chance worrying of mutations: those that could help — if, for instance, the mutation happened to the virus to evade immune systems, vaccines “This variant now is the be slightly more common in the viruses that or therapies. arrived in Europe. A small number of indi- pandemic. As a result, its viduals seem to be responsible for most of Slow change properties matter.” the virus’s spread, and an early, chance tilt in Soon after SARS-CoV-2 was detected in China, favour of G viruses could explain the lineage’s researchers began analysing viral samples and apparent takeover now. Such ‘founder effects’ posting the genetic codes online. Mutations — make sense of them. Many mutations will are common in viruses, especially when they most of them single-letter alterations between have no consequence for the virus’s ability spread unchecked, as SARS-CoV-2 did in much viruses from different people — allowed to spread or cause disease, because they do of Europe until mid- to late March. researchers to track the spread by linking not alter the shape of a protein, whereas those Korber and her colleagues tried to rule out closely related viruses, and to estimate when mutations that do change are more a founder effect, by showing in their April pre- SARS-CoV-2 started infecting humans. likely to harm the virus than improve it (see print1 that D614G rose to dominance quickly Viruses that encode their genome in RNA, ‘A catalogue of coronavirus mutations’). “It’s in Canada, Australia and parts of the United such as SARS-CoV-2, HIV and influenza, tend much easier to break something than it is to States (an exception was Iceland, where to pick up mutations quickly as they are cop- fix it,” says Hodcroft, who is part of G viruses present early in its outbreak were ied inside their hosts, because enzymes that (https://nextstrain.org), an effort to analyse overtaken by D viruses). Analysing hospitali- copy RNA are prone to making errors. After SARS-CoV-2 genomes in real time. zation data from Sheffield, UK, the team found the severe acute respiratory syndrome (SARS) Many researchers suspect that if a mutation no evidence that viruses carrying the mutation virus began circulating in humans, for instance, did help the virus to spread faster, it probably made people any sicker. But those infected it developed a kind of mutation called a dele- happened earlier, when the virus first jumped with G viruses seemed to have slightly higher tion that might have slowed its spread4. into humans or acquired the ability to move levels of viral RNA in their noses and mouths But sequencing data suggest that coronavi- efficiently from one person to another. At a than did those with D viruses. ruses change more slowly than most other RNA time when nearly everyone on the planet is Many scientists weren’t convinced that viruses, probably because of a ‘proofreading’ susceptible, there is likely to be little evolution- D614G’s rise was remarkable — or all that rele- enzyme that corrects potentially fatal copying ary pressure on the virus to spread better, so vant to the pandemic. “I thought that preprint

Nature | Vol 585 | 10 September 2020 | 175 ©2020 Spri nger Nature Li mited. All rights reserved. ©2020 Spri nger Nature Li mited. All rights reserved.

Feature A CATALOGUE OF CORONAVIRUS MUTATIONS absolute clarity. Some studies show that Various mutations have been detected in SARS-CoV-2 genomes, including the most prevalent one, D614G. certain mutations to the spike protein in the The virus’s genetic code has just under 30,000 nucleotides of RNA, or letters, that spell out at least 29 genes. Middle East respiratory syndrome (MERS) The most common mutations are single-nucleotide changes. virus can cause more-severe disease in mice The coloured areas represent A key area is the RNA that — yet other mutations in the protein show sections of the viral genome that encodes the virus’s spike protein,

encode one or many genes. which helps it to enter cells. very little effect in people or in camels, the (2020) , 603–611

SARS-CoV-2 likely reservoir for human MERS infections, 396 genome says Stanley Perlman, a coronavirologist at the

Nucleotide 0 15,000 30,000 University of Iowa in Iowa City. LANCET position The D614G The clearest sign that D614G has an effect mutation is now on the spread of SARS-CoV-2 in humans present in nearly A di”erent kind of comes from an ambitious UK effort called the all viral sequences. 30,000 Three other mutations mutation, in which COVID-19 UK Consortium, which are found in nearly all 382 nucleotides of D614G viruses. One is RNA are deleted has analysed genomes of around 25,000 viral in a viral enzyme from the genome, samples. From these data, researchers have targeted by drugs has been linked to identified more than 1,300 instances in which such as remdesivir. a milder form of COVID-19. a virus entered the and spread, 20,000 Another mutation in the spike RBD allows including examples of D- and G-type viruses. the virus to escape A team led by Andrew Rambaut, an evo- Lab experiments have recognition by some lutionary biologist at the University of flagged a mutation in the neutralizing antibodies. receptor binding domain It was common in Edinburgh, UK, epidemiologist Erik Volz, (RBD) of the spike protein. sequences from at , and biologist 10,000 This change boosts protein Scotland, UK, but expression and is found in hasn’t been detected Thomas Connor at Cardiff University, stud- some virus samples. for months. ied the UK spread of 62 COVID-19 clusters seeded by D viruses and 245 by G viruses7. The

Number of viral samples in which mutationNumber of viral detected researchers found no clinical differences in people infected with either virus. However, 0 G viruses tended to transmit slightly faster 0 5,000 10,000 15,000 20,000 25,000 30,000 than lineages that didn’t carry the change, and Nucleotide position

formed larger clusters of infections. Their esti- ET AL . E. YOUNG REFS 2, 11, 12; B. DORP ET AL . (HTTP://GO.NATURE.COM/3GSRNH6); L. VAN SOURCES: was incredibly premature,” says Sheahan. In laboratory tests, “all of us agree that D to mates of the difference in transmission rates Montefiori says his and Korber’s perspective G is making the particles more infectious”, says hover around 20%, Volz says, but the true value on D614G was shaped by their work on HIV, Jeremy Luban, a virologist at the University of could be a bit higher or lower. “There’s not a which has found that even seemingly insig- Massachusetts Medical School in Worcester. large effect in absolute terms,” says Rambaut. nificant mutations can have a profound effect But these studies come with many caveats It’s possible that D614G is an adaptation on how the immune system recognizes that — and their relevance to human infections is that helps the virus to infect cells or compete virus. “We were alarmed by it, and we need to unclear. “What’s irritating are people taking with viruses that don’t carry the change, while see if it’s having an effect on vaccines,” he says. their results in very controlled settings, and altering little about how SARS-CoV-2 spreads saying this means something for the pandemic. between people or through a population, Rush of lab studies That, we are so far away from knowing,” says Rambaut says. “This might be a bona fide To examine further whether D614G made the Grubaugh. The pseudoviruses carry only the adaptation to humans or some human cells,” virus more transmissible, Montefiori gauged coronavirus spike protein, in most cases, and agrees Grubaugh, “but that doesn’t mean any- its effects under laboratory conditions. He so the experiments measure only the ability thing changes. An adaptation doesn’t have to couldn’t study the natural SARS-CoV-2 virus in of these particles to enter cells, not aspects of make it more transmissible.” his lab, because of the biosafety containment their effects inside cells, let alone on an organ- Grubaugh thinks that D614G has received required. So he studied a genetically modified ism. They also lack the other three mutations too much attention from scientists, in part form of HIV that used the SARS-CoV-2 spike that almost all D614G viruses carry. “The bot- because of the high-profile papers it has protein to infect cells. Such ‘pseudovirus’ par- tom line is, they’re not the virus,” says Luban. garnered. “Scientists have this crazy fasci- ticles are a workhorse of labs: they Some labs are now working with infectious nation with these mutations,” he says. But enable the safe study of deadly pathogens such SARS-CoV-2 viruses that differ by only the single he also sees D614G as a way to learn about as the Ebola virus, and they make it easy to test amino acid. These are tested in laboratory cul- a virus that doesn’t have much in the way of the effects of mutations. tures of human lung and airway cells, and in lab genetic diversity. “The virologist in me looks The first team to report pseudovirus exper- animals such as ferrets and hamsters. For labs at these things and says it would be a lot of fun iments on D614G, in June, was led by Hyeryun with the experience and the biosafety capabil- to study,” he says. “It creates this whole rabbit Choe and Michael Farzan, virologists at the ities to manipulate viruses, “this is like bread- hole of different things you can go into.” Scripps Research Institute in La Jolla, California5. and-butter kind of work”, says Sheahan. The He’ll have company. Intense study of D614G Several other teams have posted similar studies first of those studies, led by researchers at the should help to explain how SARS-CoV-2 fuses on bioRxiv (Montefiori’s experiments, and those University of Medical Branch in Galves- with cells, says Luban — a process that might of another collaborator, appeared in the Cell ton, was reported in a 2 September preprint6. It be blocked by drugs or targeted by a vaccine. paper2). The teams used different pseudovirus found that viruses with the mutation were more In an updated version of their pseudovirus systems and tested them on various kinds of infectious than were D viruses in a human lung experiments posted on bioRxiv on 16 July8, cell, but the experiments pointed to the same cell line and in airway tissues, and that mutated Luban’s team used cryo-electron microscopy to conclusion: viruses carrying the G mutation viruses were present at greater levels in the analyse the structure of spike proteins bearing infected cells much more ably than did D viruses upper airways of infected hamsters6. the D614G change. The spike protein is com- — up to ten times more efficiently, in some cases. Even these experiments might not offer prised of three identical peptides in an ‘open’

176 | Nature | Vol 585 | 10 September 2020 ©2020 Spri nger Nature Li mited. All rights reserved. ©2020 Spri nger Nature Li mited. All rights reserved.

or ‘closed’ orientation. Previous research has THE MUTATION THAT LOOSENS THE SPIKE PROTEIN suggested that at least two of the three pep- Spike proteins on SARS-CoV-2 bind to receptors on human cells, helping the virus to enter. A spike protein is made tides need to be open for the viral particle to up of three smaller peptides in ‘open’ or ‘closed’ orientations; when more are open, it’s easier for the protein to bind. 9 The D614G mutation — the result of a single-letter change to the viral RNA code — seems to relax connections fuse with the cell membrane , and Luban’s team between peptides. This makes open conformations more likely and might increase the chance of infection. found that viruses carrying the G spike variant Bases RNA base sequence …CAGGAUGUUAAC… …CAGGG UGUUAAC… A: adenine were much more likely to be in this state (see Mutation C: cytosine ‘The mutation that loosens the spike protein’). Amino acid Gln Asp Val Asn Gln Gly Val Asn G: guanine (Q) (D) (V) (N) (Q) (G) (V) (N) U: uracil Computational modelling work by Montefiori and Korber, led by Korber’s LANL colleague San- Spike protein drasegaram Gnanakaran, came to the same con- clusion10. “It looks like this molecular machine is primed to go in a way that D is not,” Luban says. D614G No escape from antibodies — yet Most available evidence suggests that D614G Virus surface doesn’t stop the immune system’s neutralizing antibodies from recognizing SARS-CoV-2, as Montefiori had worried. That might be because All peptides closed One open Two open Three open the mutation is not in the spike protein’s recep- tor-binding domain (RBD), a region that many Likelihood of infection

SOURCE: STRUCTURAL DATA FROM K. SHEN & J. LUBAN K. SHEN FROM DATA STRUCTURAL SOURCE: neutralizing antibodies target: the RBD binds to the cell-receptor protein ACE2, a key step in antibodies, but his team’s results suggest that single viral mutation, for example. Cocktails the virus’s entry to cells. such changes are possible. “It is a possibility, of monoclonal antibodies, each of which can But evidence is emerging that other muta- but by no means a certainty, that the virus will recognize multiple regions of the spike protein, tions could help the virus to avoid some anti- acquire mutations that change its susceptibility might lessen the odds that such a mutation bodies. A team led by virologists Theodora to antibodies and immunity,” says Bloom. will be favoured through natural selection, Hatziioannou and Paul Bieniasz, at Rockefeller Based on experience with other corona- researchers say. Vaccines arouse less concern University in New York City, genetically modi- viruses, that might take years. Studies of on this score because, like the body’s natural fied the vesicular stomatitis virus — a livestock common-cold coronaviruses, sampled across immune response, they tend to elicit a range pathogen — so that it used the SARS-CoV-2 multiple seasons, have identified some signs of antibodies. spike protein to infect cells, and grew it in the of evolution in response to immunity. But the It’s even possible that the D614G change presence of neutralizing antibodies. Their pace of change is slow, says Volker Thiel, an could make the virus an easier target for vac- goal was to select for mutations that enabled RNA virologist at the Institute of Virology and cines, Montefiori’s team found in a study posted the spike protein to evade antibody recogni- Immunology in Bern. “These strains remain to bioRxiv in July13. Mice, monkeys and humans tion. The experiment generated spike-protein constant, more or less.” that received one of a number of experimental mutants that were resistant to antibodies taken With most of the world still susceptible to RNA vaccines, including one being developed from the blood of people who had recovered SARS-CoV-2, it’s unlikely that immunity is by drug maker Pfizer in New York City, pro- from COVID-19, as well as to potent ‘monoclo- currently a major factor in the virus’s evolu- duced antibodies that proved more potent at nal’ antibodies that are being developed into tion. But as population-wide immunity rises, blocking G viruses than D viruses. therapies. Every one of the spike mutations With G viruses now ubiquitous, the finding is was found in virus sequences isolated from “It is a possibility that the “good news”, says Montefiori. But as a scientist patients, report Hatziioannou, Bieniasz and who has watched HIV mutate to elude many their team — although at very low frequencies virus will acquire mutations vaccines developed against it, he remains that suggest positive selection is not yet mak- that change its susceptibility wary of the potential of SARS-CoV-2 to evade ing the mutations more common11. to antibodies and immunity.” humanity’s responses. Luban agrees: “We need Other scientists are trying to stay ahead of to keep our eyes open for additional changes.” SARS-CoV-2’s evolution by predicting which mutations are likely to be important. Jesse whether through infection or vaccination, a Ewen Callaway writes for Nature from London. Bloom, an evolutionary virologist at the steady trickle of immune-evading mutations 1. Korber, B. et al. Preprint at bioRxiv https://doi. Fred Hutchinson Cancer Research Center in could help SARS-CoV-2 to establish itself per- org/10.1101/2020.04.29.069054 (2020). Seattle, Washington, led a team that created manently, says Sheahan, potentially causing 2. Korber, B. et al. Cell 182, 812–827 (2020). nearly 4,000 mutated versions of the spike mostly mild symptoms when it infects individ- 3. Grubaugh, N., Hanage, W. & Rasmussen, A. Cell 182, 794–795 (2020). protein’s RBD, and measured how the alter- uals who have some residual immunity from a 4. Muth, D. et al. Sci. Rep. 8, 15177 (2018). ations affected the expression of the spike previous infection or vaccination. “I wouldn’t 5. Zhang, L. et al. Preprint at bioRxiv https://doi. protein and its ability to bind to ACE2. Most be surprised if this virus is maintained as a org/10.1101/2020.06.12.148726 (2020). 6. Plante, K. et al. Preprint at bioRxiv https://doi. of the mutations had no effect on or hin- more common, cold-causing coronavirus.” But org/10.1101/2020.09.01.278689 (2020). dered these properties, although a handful it’s also possible that our immune responses 7. Volz, E. M. et al. Preprint at bioRxiv https://doi. 12 org/10.1101/2020.07.31.20166082 (2020). improved them . Some of these mutations to coronavirus infections, including to 8. Yurkovetskiy, L. et al. Preprint at bioRxiv https://doi. have been identified in people with COVID-19, SARS-CoV-2, aren’t strong or long-lived enough org/10.1101/2020.07.04.187757 (2020). but Bloom’s team found no signs of natural to generate selection pressure that leads to sig- 9. Wrapp, D. et al. Science 367, 1260–1263 (2020). 10. Mansbach, R. A. et al. Preprint at bioRxiv https://doi. selection for any of the variants. “Probably the nificantly altered virus strains. org/10.1101/2020.07.26.219741 (2020). virus binds to ACE2 about as well as it needs to Worrisome mutations could also become 11. Weisblum, Y. et al. Preprint at bioRxiv https://doi. right now,” he says. more common if antibody therapies aren’t org/10.1101/2020.07.21.214759 (2020). 12. Starr, T. N. et al. Cell 182, 1295–1310 (2020). The researchers didn’t test whether any of the used wisely — if people with COVID-19 receive 13. Weissmann, D. et al. Preprint at bioRxiv https://doi. mutations allow the virus to thwart the action of one antibody, which could be thwarted by a org/10.1101/2020.07.22.20159905 (2020).

Nature | Vol 585 | 10 September 2020 | 177 ©2020 Spri nger Nature Li mited. All rights reserved. ©2020 Spri nger Nature Li mited. All rights reserved.

Correction This News Feature gave the wrong location for the Los Alamos National Laboratory. The lab is in Los Alamos, not Santa Fe.

©2020 Spri nger Nature Li mited. All rights reserved.