The value of influenza aerosol transmission experiments

ferretA Harvard epidemiologist has been on a crusade to curtail aerosol transmission experiments on avian influenza H5N1 virus because he believes that they are too dangerous and of little value. Recently he has taken his arguments to the Op-Ed pages of the New York Times. While Dr. Lipsitch is certainly entitled to his opinion, his arguments do not support his conclusions.

In early 2013 Lipsitch was the subject of a piece in Harvard Magazine about avian influenza H5N1 virus entitled The Deadliest Virus.  I have previously criticized this article  in which Lipsitch calls for more stringent H5N1 policies. More recently Lipsitch published an opinion in PLoS Medicine in which he called for alternatives to experiments with potential pandemic pathogens. We discussed this piece thoroughly on This Week in Virology #287.  The arguments he uses in both cases are similar to those in the OpEd.

The Times OpEd is entitled Anthrax? That’s not the real worry. The title is a reference to the possible exposure to anthrax bacteria of workers at the Centers for Disease Control. Even worse than anthrax, argues Lipsitch, would be accidental exposure to a pathogen that could transmit readily among humans. He then argues that such a pathogen is being created in laboratories that study avian influenza H5N1 transmission.

Lipsitch tells us ‘These experiments use flu strains like H5N1, which kills up to 60 percent of humans who catch it from birds.’ As an epidemiologist Lipsitch knows that this statement is wrong. The case fatality ratio for avian H5N1 influenza virus in humans is 60% – the number of deaths divided by the cases of human infections that are diagnosed according to WHO criteria. The mortality rate is quite different: it is the number of fatalities divided by the total number of H5N1 infections of humans. For a number of reasons the H5N1 mortality ratio in humans has been a difficult number to determine.

Next Lipsitch incorrectly states that the goal of experiments in which avian influenza H5N1 viruses are given the ability to transmit by aerosol among ferrets is ‘to see what gives a flu virus the potential to create a pandemic.’ The goal of these experiments is to identify mechanistically what is needed to make an avian influenza virus transmit among mammals. Transmission of a virus is required for a pandemic, but by no means does it assure one. I do hope that Lipsitch knows better, and is simply trying to scare the readers.

He then turns to the experiments of Kawaoka and colleagues who recently reconstructed a 1918-like avian influenza virus and provided it with the ability to transmit by aerosol among ferrets. These experiments are inaccurately described. Lipsitch writes that the reconstructed virus was ‘both contagious and comparably deadly to the 1918 flu that killed tens of millions of people worldwide’. In fact the reconstructed virus is less virulent in ferrets than the 1918 H1N1 virus that infected humans. In the same sentence Lipsitch mixes virulence in ferrets with virulence in humans – something even my virology students know is wrong. Then he writes that ‘Unlike experiments with anthrax, creating such flu strains in the lab presents a danger that affects us all, because once it is out, such a strain would be extremely hard to control.’ This is not true for the 1918-like avian influenza virus assembled by the Kawaoka lab: it was shown that antibodies to the 2009 pandemic H1N1 influenza virus can block its replication. The current influenza virus vaccine contains a 2009 H1N1 component that would protect against the 1918-like avian influenza virus.

The crux of the problem seems to be that Lipsitch does not understand the purpose of influenza virus transmission experiments. He writes that ‘The virologists conducting these experiments say that by learning about how flu transmits in ferrets, we will be able to develop better vaccines and spot dangerous strains in birds before they become pandemic threats.’ This justification for the work is wrong.

Both Kawaoka and Fouchier have suggested that identifying mutations that improve aerosol transmission of avian influenza viruses in ferrets might help to detect strains with transmission potential, and help vaccine manufacture. I think it was an error to focus on these potential benefits because it detracted from the real value of the work, to provide mechanistic information on what allows aerosol transmission of influenza viruses among mammals.

In the Kawaoka and Fouchier studies, it was found that adaptation of H5N1 influenza virus from avian to mammalian receptors lead to a decrease in the stability of the viral HA glycoprotein. This property had to be reversed in order for these viruses to transmit by aerosol among ferrets. Similar stabilization of the HA protein was observed when the reconstructed 1918-like avian influenza virus was adapted to aerosol transmission among ferrets. It is not simply coincidence when three independent studies come up with the same outcome: clearly HA stability is important for aerosol transmission among mammals. This is one property to look for in circulating H5N1 strains, not simply amino acid changes.

Lipsitch mentions nothing about the mechanism of transmission; he focuses on identifying mutations for surveillance and vaccine development. He ignores the fundamental importance of this work. In this context, the work has tremendous value.

The remainder of the Times OpEd reminds us how often accidents occur in high security biological labortories. There are problems with these arguments. Lipsitch cites the emergence of an H1N1 influenza virus in 1977 as ‘escaped from a lab in China or the Soviet Union’. While is seems clear that the 1977 H1N1 virus probably came from a laboratory, there is zero evidence that it was a laboratory accident. It is equally likely that the virus was part of a clinical trial in which it was deliberately administered to humans.

Lipsitch also cites the numerous incidents that occur in American laboratories involving select agents. I suggest the reader listen to Ron Fouchier explain on TWiV #291 how a computer crash must be recorded as an incident in high biosecurity laboratories, but does not lead to the release of infectious agents.

Lipsitch clearly feels that the benefits of aerosol transmission research do not justify the risks involved. I agree that the experiments do have some risk, but it is not as clear cut as Lipsitch would suggest. Although ferrets are a good model for influenza virus pathogenesis, like any animal model, they are not predictive of what occurs in humans. An influenza virus that transmits by aerosol among ferrets cannot be assumed to transmit in the same way among humans. This is the assumption made by Lipsitch, and it is wrong.

I agree that transmission work on avian H5N1 influenza virus must be done under the proper containment. Before these experiments can be done they are subject to extensive review of the proposed containment and mitigation procedures. There is no justification for the additional regulation proposed by Lipsitch.

In my opinion aerosol transmission experiments on avian influenza viruses are well worth the risk. We know nothing about what controls aerosol transmission of viruses. The way to obtain this information is to take a virus that does not transmit by aerosol, derive a transmissible version, and determine why the virus has this new property. To conclude that such experiments are not worth the risk not only ignores the importance of understanding transmission, but also fails to acknowledge the unpredictable nature of science. Often the best experimental results are those which were never anticipated.

Lipsitch ends by saying that ‘There are dozens of safe research strategies to understand, prevent and treat pandemic flu. Only one strategy — creating virulent, contagious strains — risks inciting such a pandemic.’ Creating a virulent strain is not part of the strategy. Lipsitch conveniently ignores the fact that Fouchier’s H5N1 strain that transmits by aerosol among ferrets is not virulent when transmitted by that route. And of course we do not know if these strains would be transmissible in humans.

I am very disappointed that the Times chose to publish this OpEd without checking Lipsitch’s statements. He is certainly entitled to his own opinion, but he is not entitled to his own facts.

Unusual mortality pattern of 1918 influenza A virus

1918 influenza mortalityThe 1918 influenza pandemic was particularly lethal, not only for the very young and the very old (as observed for typical influenza), but unexpectedly also for young adults, 20 to 40 years of age (pictured). It has been suggested that the increased lethality in young adults occurred because they lacked protective immunity that would be conferred by previous infection with a related virus. Reconstruction of the origins of the 1918 influenza virus provides support for this hypothesis.

Analysis of influenza virus genome sequences using a host-specific molecular clock together with seroarchaeology (analysis of stored sera for the presence of antibodies to influenza virus) indicates that the 1918 H1N1 virus arose ~1915 by reassortment of an avian influenza virus with an H1 virus that had previously emerged around 1907. The 1918 virus acquired the HA gene from the 1907 virus, and the NA gene and internal protein genes from an avian virus. This 1918 virus also infected pigs, in which descendants continue to circulate; however the human 1918 virus was displaced in 1922 by a reassortant with a distinct HA gene.

Seroarchaeology and mortality data indicate that an influenza pandemic in 1889-1893 was caused by an influenza H3N8 virus. This virus appears to have circulated until 1900, when it was replaced by a H1N8 virus (the N8 gene originating from the previously circulating H3N8 virus).

How do these events explain the unusual mortality pattern of the 1918 influenza A virus? High mortality among 20-40 year old adults might have been a consequence of their exposure to the H3N8 virus that circulated from 1889-1900. This infection provided no protection against the 1918 H1N1 virus. Protection of other age groups from lethal infection was likely a consequence of childhood exposure to N1 or H1 containing viruses (this may also have resulted in the lower than usual mortality in the elderly population). Influenza is typically highly lethal in very young children due to lack of immunologic memory.

These observations suggest that childhood exposure to influenza virus is a key predictor of virulence of a pandemic strain. Antibodies against the stalk of the HA protein protect against severe disease, but only within groups of HA subtypes (HA groups are determined by phylogenetic analysis). In 1918, antibodies against a group 2 HA subtype virus (H3) did not protect against severe disease caused by a group 1 HA subtype virus (H1). Childhood exposure might also determine mortality of seasonal influenza. For example, the high virulence of currently circulating H3N2 influenza viruses in those older than 65 years might be a consequence of infection with an H1N1 virus at a young age.

This logic can also explain mortality caused by influenza H5N1 and H7N9 viruses. Most fatalities caused by H5N1 viruses (the H5 is a group 1 HA) have been in individuals who were infected as children with an H3 virus (group 2 HA). Most fatalities caused by H7N9 viruses (group 2 HA) have occurred in individuals who were infected as children with H1N1 or H2N2 viruses (group 1 HA).

The practical consequence of this work are clearly stated by the authors:

Immunization strategies that mimic the apparently powerful lifetime protection afforded by initial childhood exposure might dramatically reduce mortality due to both seasonal and novel IAV strains.

TWiV 282: Tamiflu and tenure too

On episode #282 of the science show This Week in Virology, the TWiV team reviews a meta-analysis of clinical trial reports on using Tamiflu for influenza, and suggestions on how to rescue US biomedical research from its systemic flaws.

You can find TWiV #282 at www.microbe.tv/twiv.

Virology question of the week: why a segmented viral genome?

influenza-reassortmentThis week’s virology question comes from Eric, who writes:

I’m working on an MPH and in one of my classes we are currently studying the influenza virus. I’d forgotten that the genome is in 8 separate parts. Curious, I’ve been searching but can’t find any information as to why that is?

What evolutionary advantage is conferred by having a segmented genome?

Terrific question! Here is my reply:

It’s always hard to have answers to ‘why’ questions such as yours. We answer these questions from a human-centric view of what viruses ‘need’. We might not be right. But I’d guess there are at least two important advantages of having a segmented RNA genome.

Mutation is an important source of RNA virus diversity that is made possible by the error-prone nature of RNA synthesis. Viruses with segmented genome have another mechanism for generating diversity: reassortment (illustrated).

An example of the evolutionary importance of reassortment is the exchange of RNA segments between mammalian and avian influenza viruses that give rise to pandemic influenza. The 2009 H1N1 pandemic strain is a reassortant of avian, human, and swine influenza viruses.

Having a segmented genome is another way to get around the limitation that eukaryotic mRNAs can only encode one protein. Viruses with segmented RNA genomes can produce at least one protein per segment, sometimes more. There are other ways to overcome this limitation – for example by encoding a polyprotein (picornaviruses), or producing subgenomic RNAs (paramyxoviruses).

Other segmented viral genomes include those of reoviruses, arenaviruses, and bunyaviruses.

There are various ways to achieve genetic variation and gene expression, and viruses explore all aspects of this space.

Attenuated influenza vaccine enhances bacterial colonization of mice

attenuated influenzaInfection with influenza virus is known to increase susceptibility to bacterial infections of the respiratory tract. In a mouse model of influenza, increased bacterial colonization was also observed after administration of an infectious, attenuated influenza virus vaccine.

Primary influenza virus infection increases colonization of the human upper and lower respiratory tract with bacteria, including Streptococcus pneumoniae and Staphylococcus aureus. Such infections may lead to complications of influenza, including pneumonia, bacteria in the blood, sinusitis, and ear infections.

One of the vaccines available to prevent influenza is an infectious, attenuated preparation called Flumist. To determine if a vaccine such as Flumist increases susceptibility to bacterial infection, the authors created their own version of the vaccine (illustrated) in which the six RNA segments encoding internal proteins were derived from the A/Puerto Rico/8/34 (H1N1) strain (allowing replication in mice), and the HA and NA proteins were derived from A/Hong Kong/1/68 (H3N2). In addition, mutations were introduced into the viral genome that are important for the safe and protective properties of Flumist. For simplicity we’ll call this virus ‘live attenuated influenza virus’, or LAIV.

Mice were inoculated intranasally with a strain of S. pneumoniae known to colonize the nasopharynx, followed 7 days later by LAIV or wild type influenza virus. Inoculation with either virus similarly increased the bacterial levels in the nasopharynx, and extended the time of colonization from 35 to 57 days. In mice that were given only bacteria and no influenza virus, the inoculated bacteria were cleared beginning 4 days after administration. The more extensive and extended colonization of virus-infected mice was not associated with overt disease.

Administration of LAIV or wild type virus 7 days before bacteria also resulted in excess bacterial growth in mice. Similar results were obtained using S. aureus. Administration of S. pneumoniae up to 28 days after virus also lead to excess bacterial growth, despite clearance of the viruses around 7 days after vaccination.

All mice died when they were vaccinated with wild type influenza virus followed 7 days later by a sublethal dose of a highly invasive strain of S. pneumoniae. In contrast, pretreatment with LAIV lead to no disease or death of any mice.

It is not known if these findings in a mouse model directly apply to humans. However, because Flumist reduces influenza virus replication, it is associated with a decrease in secondary bacterial infections. It is possible that, after administration of LAIV to humans, there is an increase in bacterial colonization of the respiratory tract. Upper respiratory tract symptoms are a known adverse effect of LAIV, and it is possible that these might be related to increased bacterial loads. It is important to emphasize that use of LAIV is not associated with severe upper or lower tract disease.

These findings are important because they show that a mouse model could be used to understand why influenza virus infection leads to increased bacterial colonization of the respiratory tract. It will be important to determine the precise mechanisms by which influenza virus infection, and the associated virus and immune-mediated alteration to the respiratory tract, allows enhanced bacterial colonization. At least one mechanism, which we discussed on episode #62 of This Week in Microbiology, involves the disruption of biofilms, allowing bacteria to enter the bloodstream.

The alphanumeric soup known as influenza

Robert Herriman, co-founder of The Global Dispatch, recently started a radio show called Outbreak News This Week. Robert calls the show “Your source for all the news about worms and germs”. He covers the latest news and information about infectious diseases and often includes interviews with expert guests. The show can be heard live Saturday mornings at 7:30 am EST on on The Tan Talk Radio Network: 1340 AM WTAN Clearwater, 1350 AM WDCF Dade City and 1400 AM WZHR Zephyrhills. You can also listen online.

I have been a frequent guest on Robert’s Outbreak News This Week, most recently this past Saturday, when we had a broad-ranging conversation about influenza virus. We also managed to squeeze in a few words about my favorite virus, poliovirus, and India’s success in remaining polio-free for three years. Listen below.

 

TWiV 267: Snow in the headlights

On episode #267 of the science show This Week in Virology, Vincent, Alan, Rich and Kathy review a protease essential for influenza pathogenesis in mice, and directionality of rhinovirus RNA exit from the capsid.

You can find TWiV #267 at www.microbe.tv/twiv.

Cutting through mucus with the influenza virus neuraminidase

influenza virusNeuraminidase is one of three different viral proteins embedded in the lipid membrane of influenza virus (NA is blue in the illustration at left). This enzyme has a clear and proven role in virus release from cells. NA is also believed to be important during virus entry, by degrading the mucus barrier of the respiratory tract and allowing virus to reach cells. This role is supported by the finding that treatment of mucus-covered human airway epithelial cells with the NA inhibitor Tamiflu substantially suppresses the initiation of infection.  Further evidence comes from the recent finding that influenza virus binds to sialic acids in mucus and that NA cleaves these sugars to allow infection.

The mucus layer of the respiratory tract has a defensive role because it contains soluble glycoproteins that are rich in sialic acids, which are cell receptors for some viruses. When influenza virions enter the respiratory tract, they are thought to be trapped in mucus where they bind sialic acids, preventing infection of the underlying cells. The role of NA in penetration of the mucus layer was studied in frozen sections of human tracheal and bronchial tissues, in which the mucus layer is preserved. Influenza A virions bind to this mucus layer; the interaction is blocked when the tissues are first treated with a bacterial neuramindase, which removes sialic acids from glycoproteins. These observations indicate that the mucus layer of human airway epithelial cells contain sialic acid-containing decoys that bind influenza A viruses.

Human salivary mucins, which approximate the mucus of the human respiratory tract, can protect cultured cells from influenza virus infection. The protective effect can be achieved by simply adding the mucins to cells. The inhibitory effect is dependent on the sialic acid content of the mucins: fewer cells are infected when higher concentrations are used. In contrast, porcine salivary mucins do not substantially reduce influenza virus infection. The type of sialic acids and the virus strain also determine the extent of protection.

To determine the role of the viral NA in mucin-mediated inhibition, Tamiflu was mixed with virus before infection of mucus-coated cells. The presence of Tamiflu increased the inhibition of infection caused by mucins, indicating that the sialic acid-cleaving (sialidase) activity of NA is needed to overcome inhibition by human salivary mucins. When influenza virions are incubated with human salivary mucins linked to beads, sialic acids are cleaved from mucins, and this enzymatic activity is inhibited by Tamiflu. Human salivary mucins inhibit NA by binding to the active site of the enzyme.

These studies establish a clear role for the influenza viral NA in bypassing the defenses of the mucosal barrier. An important message is that not all strains of influenza virus are equally inhibited by mucus; presumably this property is one of many that determines viral virulence. The balance between how tightly the viral HA binds to sialic acids, and how well the NA cleaves them, is probably one predictor of how well we our protected by our mucus.

 

TWiV 263: Game of clones

On episode #263 of the science show This Week in Virology, Ben tenOever joins the TWiV team to reveal the winner of his contest in which influenza viruses carrying different interferon-stimulated genes vie against one another in mice.

You can find TWiV #263 at www.microbe.tv/twiv.

Changing influenza virus neuraminidase into a receptor binding protein

neuraminidaseThe hemagglutinin (HA) and neuraminidase (NA) glycoproteins of the influenza virus particle serve distinct functions during infection. The HA binds sialic acid-containing cellular receptors and mediates fusion of the viral and cell membranes, while the NA removes sialic acids from glycoproteins. Apparently this division of labor is not absolute: influenza viruses have been identified with NA molecules that serve as receptor binding proteins.

An influenza virus was created that could not bind sialic acid by introducing multiple mutations into the HA gene. This mutant virus was not expected to be infectious, but nevertheless did propagate to moderate titers in cell culture. A single amino acid change was identified in the NA protein of this virus: G147R, which is just above the active site of the enzyme (illustrated; active site marked with green spheres). Passage of the virus in cell culture produced a virus that multiplied to higher titers; improved growth was caused by a K62E change in the HA stalk. The results of site-directed mutagenesis showed that the G147R change allowed the NA protein to serve the receptor binding function normally provided by HA. It is not clear how the HA change leads to improved growth of the G147 virus.

Although the G147R NA can serve as receptor binding protein, the HA is still required for fusion: abolishing this activity by mutation or by treatment with a fusion-blocking antibody did not allow virus growth.

The influenza NA protein is an enzyme (sialidase) that cleaves sialic acids from cellular and viral proteins. The G147R NA is active as a sialidase, and this activity can be blocked by the antiviral compound oseltamivir, which is an NA inhibitor. Treatment of G147R-containing virus with oseltamivir also blocked virus binding to cells. Virus-like particles that contain G147R NA but not HA can attach to sialic acid-containing red blood cells. This attachment can be reversed by oseltamivir. After binding to red blood cells, these virus-like particles slowly fall off, a consequence of NA cleaving sialic acid receptors. These observations indicate that the G147R NA binds to sialic acids at the active site of the enzyme, and cleaves the same receptor that it binds.

Treatment of cells with a bacterial sialidase that removes a broad range of sialic acids only partially inhibits G147R NA-mediated binding to cells. In contrast, growth of wild type influenza virus is completely blocked by this treatment. Therefore the receptor recognized by G147R NA is not the same as that bound by wild type virus.

Changing the influenza virus NA to a receptor binding protein is not simply a laboratory curiosity: the G147R NA change was found in 31 of 19,528 NA protein sequences in the Influenza Virus Resource. They occur in seasonal H1N1 viruses that circulated before 2009, in the 2009 swine-origin pandemic H1N1 virus, and in avian H5N1 viruses. The presence of this change in phylogenetic clusters of seasonal H1N1 and chicken H5N1 sequences suggests that they are also found in circulating viruses, and are not simply sequence errors or the product of passage in the laboratory.

These observations emphasize the remarkable flexibility of the influenza viral glycoproteins in their ability to switch receptor binding function from HA to NA. They might also have implications for vaccines, whose effectiveness are thought to depend largely on the induction of antibodies that block the function of HA protein. The work underscores the importance of serendipity in science: the HA receptor binding mutant virus was originally produced as a negative control for a different experiment.