A Clever New Way to Predict Next Year's Flu

A study suggests an unusual strategy to make vaccines more effective.

Flu evolves remarkably fast.

Consider the example of H3N2, one of two major subtypes of flu that cause trouble every winter. In some cases, a single extremely well-adapted variant of the H3N2 virus can replace all other H3N2 viruses on Earth over just a few years. Then once enough humans become immune to it, the whole cycle begins anew.

This constant turnover is why the flu vaccine changes every year. Scientists usually have to predict a flu season’s dominant variants months in advance, so that vaccine manufacturers can make enough doses in time. Sometimes, those predictions are quite off. In 2015, for example, the vaccine was only 23 percent effective against that year’s circulating H3N2 variant.

In search of new ways to understand flu evolution, a group of scientists in Seattle decided to try something unusual. They didn’t bother to look at ordinary people sick with the flu. They instead decided to track how H3N2 viruses change in people with weakened immune systems, who come down with the flu for months at a time. Surprisingly, the mutations that arose in these patients ended up being some of the same ones that dominated global trends years later. Just four patients were microcosms for the greater world when it came to flu evolution.

The study began when Jesse Bloom, a biologist who studies flu viruses at the Fred Hutchinson Cancer Research Center, went downstairs for coffee and ran into Steven Pergam, a doctor at the same center. Bloom was interested in studying long-term flu infections, and Pergam mentioned that they happened to have a a freezer of snot samples from long-term flu patients. A decade ago, researchers at Fred Hutch had run a clinical trial on flu in cancer patients with bone-marrow transplants. The samples had been sitting in storage the whole time.

“We realized now that we have these powerful deep sequencing methods, we might be able to to repurpose these samples,” says Katherine Xue, a graduate student in Bloom’s lab. Deep sequencing is a technique for identifying rare mutations. Since the flu virus is evolving all the time, there are many different mutations among the millions of viruses in any single patient. If you go through the sequence only a few times in the sample, you’re only going to pick up the average virus. If you go through it at least 200 times, like Xue and her colleagues did, you’re going to see those rare mutations.

At first, the team didn’t know what to expect. The four patients in their study obviously didn’t have typical medical histories—given the cancer, bone-marrow transplant, and later use of the antiviral medication Tamiflu to combat their flu infections. But the team started to see some of the same mutations in the patients. The fact that these samples were old turned out to be fortuitous: The researchers could compare the mutations in these patients to flu variants that circulated years later. And again, they found similarities. For example, a particular mutation in three patients that altered the virus’s outer protein shell would become common in H3N2 viruses around the world by 2015.

To be clear, this is not because Seattle is some flu epicenter and these patients spread their H3N2 virus to the rest of the world. Rather, it suggests evolutionary pressures for the viruses in these individuals are the same pressures that apply on a global scale.

One way to think about this, says Katia Koelle, a virologist at Duke who was not involved with the research, is that the human immune system is always out to get flu viruses. The immune system makes antibodies that stick to the lollipop-shaped proteins jutting out of the virus’s shell. This neutralizes the virus. But if the very tip of the lollipop protein gets mutated, then the virus can escape the immune system. So that explains why viruses in these patients and around the world are acquiring mutations in the same places. Other molecular constraints make it so that it’s also often the same mutation in the same place.

Koelle’s lab has deep sequenced flu viruses in healthy people, who are sick for only a few days. The virus evolves in them, too. But, she says, “it’s hard for a new variant to get to high frequency in a short amount of time.” In other words, a beneficial new mutation may arise, but it may not have time to that variant to become common enough to spread  another person before your immune system clears the infection.

It’s not just immunocompromised cancer patients whose infections could help predict the future of flu. Elodie Ghedin, a parasitologist and virologist at New York University, noted that children, pregnant women, and people with obesity also tend to have longer flu infections. Deep sequencing in infected people, especially in those groups, could help scientists monitor emerging mutations. “This is a point I have been making for years in meetings, especially when discussing prediction and vaccine selection, but it has not gotten much traction,” she says. An early problem was the cost of deep sequencing, but the plunging cost of all DNA sequencing has make the technique much more practical.

Current flu surveillance typically uses regular sequencing, so you only see the common variant. With deep sequencing, you’ll find a lot of variants, many of them which are rare and will stay rare. But if you deep sequence enough patients—and the right ones—you could find the that variants start showing up in multiple patients and maybe later end up dominating the world.