For most of 2020, the coronavirus that causes COVID-19 jumped from human to human, accumulating mutations at a steady rate of two per month—not especially impressive for a virus. These mutations have largely had little effect.

But recently, three distinct versions of the virus seem to have independently converged on some of the same mutations, despite being thousands of miles apart in the United Kingdom, South Africa, and Brazil. (A mutation is a genetic change; a variant is a virus with a specific set of mutations.) The fact that these mutations have popped up not one, not two, but now three times—that we know of—in variants with unusual behavior suggests that they confer an evolutionary advantage to the virus. All three variants seem to be becoming more common. And all three are potentially more transmissible.

“Anytime when you have mutations that come up independently of each other in multiple places, it’s really a sign,” says Vineet Menachery, a coronavirus researcher at the University of Texas Medical Branch. Now scientists are scrambling to figure out if and how these mutations might give the viruses an edge.

It’s still early, and data on the variant in Brazil are particularly sparse. In addition to sharing certain mutations, though, these variants simply have a large number of mutations, some unique to each variant. Gaining a whole suite of mutations quickly should be a very uncommon event. But with the virus so widespread right now, very uncommon events will happen—and will happen more than once. The usual two-per-month mutation rate may undersell how the coronavirus can mutate in unusual situations. “It’s a little bit of a wake-up call,” Kristian Andersen, a microbiologist at Scripps Research, told me.

Read: The coronavirus is evolving before our eyes

The role of each individual mutation is still unclear, but a particular mutation in the spike protein called N501Y is noteworthy because all three variants have it. The spike protein is how the coronavirus enters cells, and N501Y is in an especially important region called the receptor-binding domain, which latches on to the cell. An N501Y mutation may make the spike protein stickier, allowing it to bind to and enter cells more readily. Such a virus could become more transmissible. On the plus side, however, the mutation doesn’t seem to affect immunity from vaccines.

Here’s how to read the names of the mutations, by the way: Proteins are made up of building blocks called amino acids. N501Y means that the 501st amino acid was originally an N, which stands for the amino acid asparagine, but has been changed to a Y, which stands for tyrosine.

N501Y is not unique to these three variants, though; it’s been found in a number of sequences around the world. What is unusual about these three variants is that they also have an additional constellation of other mutations in other parts of the virus. A change in a variant’s behavior, such as increased transmissibility, is probably “due to not just one mutation, but multiple mutations,” says Emma Hodcroft, a molecular epidemiologist at the University of Bern. The U.K. variant has more than a dozen other mutations, which have not been scrutinized as much as N501Y. But the variant’s increased transmissibility is looking more certain: It’s growing more prevalent not just in the U.K., but also in Ireland and Denmark, two other countries that regularly sequence large numbers of samples. The CDC recently warned that it’s likely to become the dominant variant in the United States by March.

(Scientists have given all three variants more specific names, but they have not, alas, standardized them yet. The U.K. variant is also known as B.1.1.7, and 20I/501Y.V1, and VOC 202012/01. The South Africa variant is sometimes called B.1.351 or 20C/501Y.V2. The Brazil variant is known as P.1 and 20J/501Y.V3.)

The South Africa and Brazil variants also have a second and third mutation in common in the spike’s receptor-binding domain: E484K and K417. Scientists know a little bit more about the E484K mutation. It switches a negatively charged amino acid for a positively charged one; it’s like flipping a magnet. This likely changes the spike protein’s shape as it is binding to a cell, but this change seems to work in synergy with the N501Y mutation, Andersen said. These mutations, possibly along with others, may make the virus better at binding to cells.

Read: Why aren’t we wearing better masks?

But the South Africa and Brazil variants might have an additional advantage. A recent study suggests that viruses with the E484K mutation might be better at evading antibodies from the blood plasma of recovered COVID-19 patients. Some viruses with this mutation could become a little better at reinfecting people or even infecting vaccinated people.

This one mutation alone is unlikely to render immunity from previous infections or vaccines totally ineffective, though. With current vaccines, “you have more than enough antibody, and even if you cut that amount in half, you still have more than enough antibody to control the virus,” Menachery told me. “If the new variant reduces the efficacy … by 50 percent, you still have a lot of protection there.” Studies are ongoing to figure out exactly how much this mutation affects vaccines, but it does suggest that vaccine makers might need to update their shots if more mutations like E484K accumulate over a period of years. This is already done every year with the flu shot, and the current mRNA COVID-19 vaccines can be updated especially quickly, in as little as six weeks, according to the manufacturers.

Scientists now wonder whether the variants in South Africa and Brazil are spreading precisely because they have this slight advantage in overcoming previous immunity. Both variants were originally found in parts of the countries that have had high levels of COVID-19 infection—especially in Manaus, Brazil, where an especially large proportion of people have already had the virus. (One December study says 76 percent, which is probably an overestimate, but the region’s high COVID-19 death toll suggests that it indeed had a huge outbreak in 2020.) The South Africa variant is becoming dominant in the country; the situation in Brazil is less clear because less data exist, but Manaus is currently experiencing another big surge of COVID-19. Menachery said he doesn’t think previous immunity is necessarily a reason for these variants to become more common, especially because South Africa isn’t as close to herd immunity. Better transmissibility is already an advantage.

Read: The next phase of vaccination will be even harder

But others sketched this plausible, though still hypothetical, scenario: The variants may have evolved in immunocompromised patients who were infected with the virus for months. Normally, Hodcroft says, “your immune system is going to town on it. It’s really trying to beat it up.” But immunocompromised patients mount weaker immune responses. “It becomes almost like a training course for how to live with the human immune system,” she says. That may be why these variants have so many new mutations at once, as if a year or two of evolution has been compressed into months. This is probably quite rare, but with tens of millions of infections around the globe, rare things will show up.

A variant could emerge, then, from the training ground of a chronic infection, with mutations that make the virus better at binding to cells and thus more transmissible. This may be what happened with the U.K. variant. It could also emerge slightly more capable of reinfection. This may be what’s happening in Brazil, where there are already two documented cases of reinfection with the new variant. In a place where many people have already been infected with COVID-19, a variant that is just a little better at evading preexisting immunity will have an advantage. These reinfections might not be serious, and they still might not be the norm, but over time, that variant will win out. The coronavirus is in a constant arms race against our immune system. It will keep evolving.

That means our vaccines may need to evolve with it. But the United States is sequencing only a tiny percentage of its COVID-19 cases. (Standard COVID-19 diagnostic tests probe a few regions of the virus genome, but they don’t sequence the whole thing.) “San Diego is one of the places in the country we’re doing well, and we’re sequencing 2 percent of cases. It’s laughable compared to the U.K. and Denmark,” Andersen said. “And we need to change that.” The sequencing data, when they are collected, are fragmented across individual labs all over the country. What the U.S. needs, Andersen said, is a federal mandate for genomic surveillance. That’s the only way for the U.S. to keep abreast of an ever-changing virus.