The Limits of Avian Flu Studies in Ferrets

Science  03 Feb 2012:
Vol. 335, Issue 6068, pp. 512-513
DOI: 10.1126/science.335.6068.512

To publish or not to publish, that is not the question—or at least not the ultimate one. For the last few months, scientists, policymakers, and the public have intensely debated whether journals should reveal details of experiments by two labs that have engineered H5N1 avian influenza viruses that easily transmit among mammals—and could thus potentially spark a human flu pandemic. But there's another difficult quandary in this tug of war between scientific freedom and bioterror fears. Both research teams relied on experiments with ferrets. How concerned should people be that what happened in Mustela putorius furo with these mutant viruses will apply to humans?

Researchers widely agree that the ferret provides the best model to study influenza transmission in humans: This weasel-like animal uses the same cellular receptors for the virus, and strains that infect people spread among ferrets and cause similar symptoms. “We should not say, ‘Well, it's just a ferret; let's forget about the results,’” says Ron Fouchier, a virologist at Erasmus MC in Rotterdam, the Netherlands, who led one of the groups behind the controversial studies. But even Fouchier emphasizes that, in many ways, the ferret model “is pretty shaky” and can lead to misleading conclusions, as it did with a study by his group when the 2009 pandemic H1N1 (pH1N1) virus emerged. “We always have to be really, really critical about what we do,” Fouchier says. “We cannot say [because] it transmitted in ferrets, it must be transmissible in humans.”

False alarm.

This ferret study triggered fears that the 2009 pandemic H1N1 was far deadlier than the seasonal strain.


The H5N1 virus, which decimates chicken flocks, first surfaced in humans in 1997 but has rarely spread from person to person. The World Health Organization has tracked cases since 2003, and as of 24 January, 344 of the 583 people who had confirmed infections died from H5N1. That startlingly high 59% mortality likely overestimates the actual lethality of the virus, as many nonfatal or subclinical infections surely go undetected. Still, influenza investigators have feared that if this virus mutated and began spreading easily among people, it could wreak havoc, as we have no immunity to the avian hemagglutinin—the H5—glyco-protein on the viral surface that enables the virus to establish an infection.

The hubbub started at an influenza meeting in September, when Fouchier described how his lab had introduced five mutations to create a highly lethal H5N1 that passes between ferrets like “seasonal” strains of the virus that routinely infect humans. Fouchier and his co-workers have a paper accepted by Science about their experiment; they are still deciding what to include, in the wake of a December recommendation by the U.S. government's National Science Advisory Board for Biosecurity (NSABB) to redact methodological and other details. Similarly, Yoshihiro Kawaoka of the University of Wisconsin, Madison, and the University of Tokyo has an accepted manuscript on hold at Nature that describes how his lab combined H5N1 with pH1N1 to make a mutant that easily spreads between ferrets. The new virus effectively stitches the H5—which has properties that make it especially deadly—into the pandemic strain.

Few specifics are known about the work in either lab, but Kawaoka revealed an unexpected detail in a commentary online in Nature on 25 January, underscoring the confusion that swirls around these findings and the ferret model: His virus didn't kill the animals. So if nature or bioterrorists created a similar virus—or if it accidentally escaped from the lab, as some fear—does the ferret model predict it would have little chance of causing mayhem in humans? “Severity of infection, including lethality, is a complex biologic phenomenon involving pre-existing immunity to pathogens, individual susceptibility, pathogen dose, etc.,” said Kawaoka in an e-mail reply to Science. “With all of these variables it is impossible to declaratively answer this question.”

Kawaoka has firsthand experience with the difficulty of connecting the dots between ferrets and humans. During the relatively mild 2009 pandemic with H1N1, when many researchers worried that the newly discovered mashup of avian-swine-human influenza viruses might kill millions of humans who had little immunity to the bird component, his lab published a ferret study online in Nature on 13 July 2009 to help clarify the threat. Seasonal strains of influenza uniformly spread between ferrets and rarely cause serious harm—which is a key reason the model has become so popular. In a comparison with seasonal strains, Kawaoka's group found that pH1N1 was more pathogenic in ferrets and caused “appreciable pathology” that was “reminiscent of infections with highly pathogenic H5N1 influenza viruses.” A headline in the press release from the University of Wisconsin, Madison, read “Study suggests H1N1 virus more dangerous than suspected.” Science had published a similar ferret study ( online a couple of weeks earlier (2 July 2009) by Fouchier's group that reached the same conclusions.

In another published experiment, the Fouchier group went a step further. They compared pH1N1, seasonal H1N1, and H5N1 in ferrets and found that the pandemic strain was indeed more lethal than seasonal H1N1 (see graph) and less than H5N1. Fouchier now says they made “a big mistake” by concluding that “pandemic H1N1 reaches further down the airway than seasonal flu, and, therefore, it's likely to become a severe pandemic.”

Conditional probability.

Many lab variables can alter results of ferret transmission experiments.


As it turns out, neither team factored in that their young ferrets, by experimental design, had no influenza immunity. Long-lived humans, in contrast, acquire immunity from both strains in circulation and vaccines. The pH1N1 caused the least amount of harm in older people who had been exposed to variants of the 1918 pandemic strain, which shared similarities with the new one. But many children, who had scant preexisting immunity to flu, were hospitalized with lower respiratory disease. “The clinical picture we saw in the hospital was quite similar to what we saw in our ferret model,” Fouchier says. But of course that's in hindsight.

The choice of influenza strains used in ferret experiments can also lead to conflicting results, as happened in a study by the U.S. Centers for Disease Control and Prevention (CDC) when pH1N1 emerged. In tandem with its online publication of the Dutch experiment in ferrets, Science ran a report ( from CDC researchers that demonstrated that different strains of pH1N1 did not spread to all their ferrets. This “lack of efficient respiratory droplet transmission suggests that additional virus adaptation in mammals may be required to reach the high-transmissible phenotypes observed with seasonal H1N1 or the 1918 pandemic virus,” they concluded. In the end, pH1N1 infected humans throughout the world with similar efficiency to seasonal strains.

CDC's Ruben Donis, who did not participate in that study but has done other flu experiments with ferrets, stresses that even a given strain represents a swarm of viruses, as flu mutates rapidly within each host. The specimen taken from a human may also differ if it is taken from the lung versus the nose. “There's all this fuzziness in every experiment we do,” Donis says.

Further complicating analyses, these experiments typically involve few ferrets, as they require complex handling procedures and large cages in often tight biocontained facilities. “Go and look for the confidence intervals or the p-value [in ferret studies],” Donis says. “Nobody shows statistics. The statisticians laugh when you tell them you used three ferrets.” This can often make potentially important results uninterpretable—which may help explain why a recent study he led, published online in Virology on 5 November 2011, did not spark an international uproar. In that experiment, his team created a mutant H5N1 that spread from infected ferrets to one of two animals in an adjacent cage.

“We have to be really, really careful to interpret our data in ferret transmission in a quantitative way,” Fouchier says. “You cannot say if you got two out of four trans missions that your virus is 50% transmissible.” His group's new study has little to say about the kinetics and efficiency of transmission of its mutant virus in humans. “I can't tell if this virus were released now to the world whether it would transmit efficiently,” he stresses. “The only thing we say is it can be airborne.”

What's more, as Peter Palese and Taia Wang of the Mount Sinai School of Medicine in New York City point out in a perspective published online 25 January in the Proceedings of the National Academy of Sciences, some strains spread between ferrets but not humans. And pathogenicity of H5N1 clearly differs between the species, too. “Very few strains of H5N1 in the ferret are lethal,” says influenza researcher and NSABB member Robert Webster of St. Jude Children's Research Hospital in Memphis, Tennessee.

In the pig, which shares the same receptors for influenza viruses used by both humans and birds, H5N1 is wimpier still, never causing serious illness. “You could argue that the ferret in the case of H5N1 is an outlier; it's just too susceptible to severe disease,” CDC's Donis says. “Maybe the pig is a better model for H5N1 because, in reality, many, many people are exposed to H5N1 and very few get sick.”

Laboratory conditions can alter outcomes, too, says Kanta Subbarao, who studies flu in ferrets and mice at the U.S. National Institute of Allergy and Infectious Diseases. Factors such as infecting the “index” animal via nose drops versus the trachea, the amount of virus used, the genetic differences in this outbred species, and the age of ferrets all can affect results. For technical reasons, no one quantifies the aerosolized droplets that presumably spread the virus in transmission experiments. “My take-home message when I speak of animal models for influenza is we should look at more than one model,” Subbarao says.

Many influenza researchers hope the Fouchier and Kawaoka data can ultimately help clarify the factors that allow flu to spread in humans. “We desperately need to understand transmissibility,” says Webster, who is glad the new studies were done. “People are starting to say these scientists are irresponsible for doing this work. We're never going to understand what transmissibility is unless we can work in the best model, and the best model we've got is this ferret—unless you would like to volunteer.”


Navigate This Article