ANN ARBOR—Notoriously wily, the influenza virus keeps public health experts guessing about when new strains will emerge and whether existing vaccines will be effective against them.

In an attempt to better understand the dynamics of influenza outbreaks, a team of theoretical ecologists at the University of Michigan has developed a new model that incorporates information on how the virus evolves with knowledge of the epidemiology of the disease.

While the model underscores the complexities that make outwitting the ever-changing virus so challenging, it also lays the groundwork for eventually teasing out patterns that may aid in predicting and controlling influenza outbreaks. The new model is presented in the Dec. 22 issue of the journal Science.

The first step in developing the model was to consider how the flu virus evolves, said first author Katia Koelle, who received her doctorate under the direction of U-M theoretical ecologist Mercedes Pascual and is now a postdoctoral fellow at The Pennsylvania State University.

” We know that the influenza virus evolves very rapidly” that’s why people can get re-infected with the virus within ten years of a previous infection,” she said. But there are two distinct patterns of evolution to consider: the underlying genetic mutations that lead to the development of different strains, and the physical changes in the virus that determine how our immune systems respond to the various strains.

The immune response is triggered by proteins called antigens on the surface of the virus. When a person is infected with the virus, the body produces antibodies that recognize the antigens. Strains of virus with similar antigens all elicit the same response, but viruses with significantly different antigens can evade the immune system and cause illness. Especially large flu outbreaks occur when new viral strains, with new antigens, appear.

Previous models have assumed that the greater the gene-level differences between two strains of flu virus, the more likely they were to have significantly different antigens. But after scrutinizing another research group’s data on a particular antigen called hemagglutinin, the authors rejected that assumption. They found that ” antigenically equivalent” strains may have considerable genetic differences, while strains that are nearly identical at the genetic level may have quite different antigenic properties.

” Our model does away with the idea that greater genetic differences imply greater antigenic differences,” Koelle said. Instead, the model considers the fact that certain genetic changes significantly affect antigen shape (and charge)” which antibodies use to recognize antigens” while others are essentially neutral, producing no significant changes in antigen shape. Borrowing the concept of neutral networks from the field of complex systems, the researchers developed a model in which cross-immunity between strains depends not on how similar or different they are genetically but on how much their antigen shapes differ. In their model, viral strains can be grouped into ” antigenic clusters” ” groups of strains with similarly shaped antigens and similar immunological properties.

The model helps explain the ups and downs of influenza outbreaks, Koelle said. When a new antigenic cluster first emerges, it has the advantage over established clusters, because its novel antigens can evade people’s immune systems. In such a situation, large outbreaks of flu are likely. The new cluster flourishes, eventually driving out the other clusters. As neutral genetic changes begin to accumulate in the new cluster, it becomes more genetically diverse, but its antigenic properties don’t change. Then, inevitably, a genetic change occurs that does result in a significant antigen shape change, and a new cluster is born. Repeated over time, this cycle produces a boom-and-bust pattern in the genetic diversity of circulating strains and explains some of the variation in year-to-year outbreak sizes.

The model’s results have implications for vaccination strategies, Koelle said. ” In years when there’s no cluster transition, using a vaccine strain that’s related to a strain in the current antigenic cluster should provide a lot of cross-immunity” the vaccine should work relatively well.”

However, the model also suggests that predicting when new clusters will arise and what their antigens will look like” information that could help public health experts stay one step ahead of the virus” is extremely difficult.

” It’s difficult, but it’s not impossible,” said Pascual, an associate professor of ecology and evolutionary biology. ” The next step will be working from our model with additional data to look for regular patterns that may help us cut through the complexity.”

In addition to Koelle and Pascual, coauthors include U-M doctoral student Sarah Cobey and Penn State population biologist Bryan Grenfell. The researchers received funding from the James S. McDonnell Foundation, the Penn State Center for Infectious Disease Dynamics and a National Science Foundation doctoral fellowship.

Katia KoelleSarah CobeyMercedes Pascual