Harvard University: Great virology, bad science writing

Harvard virologyHarvard University is home to some of the world’s finest virologists. But apparently they do not communicate with the writers at Harvard Magazine, where a botched story on the avian H5N1 influenza virus has just been published.

The problems begin with the first paragraph:

But when Dutch researchers recently created an even more deadly strain of the virus in a laboratory for research purposes, they stirred grave concerns about what would happen if it escaped into the outside world.

Readers of virology blog will know by now that the Dutch researchers did not make an ‘even more deadly strain of the virus’ – they made one that could be transmitted by aerosol, but which had lost its lethality.

The title of the article, ‘The Deadliest Virus’, presumably refers to the H5N1 virus that transmits by aerosol among ferrets. This title is simply wrong, because the virus is not deadly to ferrets.

The first paragraph also contains an equally egregious statement by epidemiologist Marc Lipsitch:

If you make a strain that’s highly transmissible between humans, as the Dutch team did, it could be disastrous if it ever escaped the lab.

Dr. Lipsitch seems to be saying that the Dutch group created an H5N1 virus that transmits among humans. As far as I know, ferrets are not humans.

The article is accompanied by a photograph of two scientists working in BSL4 suits. The legend reads:

The modified H5N1 virus could infect a billion people if it escaped a biocontainment lab like the Canadian facility shown above.

And later Lipsitch is quoted as saying:

It could infect millions of people in the United States, and very likely more than a billion people globally, like most successful flu strains do. This might be one of the worst viruses—perhaps the worst virus—in existence right now because it has both transmissibility and high virulence.

For Lipsitch to say that the virus is both transmissible and of high virulence in humans is a misrepresentation of the Dutch group’s findings. He seems to be making up numbers and scenarios.

Perhaps Dr. Lipsitch does not know that ferret studies are not predictive of how viruses will behave in humans. With so many virologists at Harvard, the writer could have checked Dr. Lipsitch’s statements. But he did not, and the result looks as foolish as the New York Times.

The risks and benefits of influenza H5N1 research

Both Nature and the New York Times have weighed in on the resumption of influenza H5N1 research. In an editorial from 23 January 2013, Nature opines that “Experiments that make deadly pathogens more dangerous demand the utmost scrutiny”:

As several critics point out, the assessments of the relative risks and benefits of such research remain restricted to largely qualitative arguments. The formal, quantitative risk assessment common in the nuclear power and other industries could have helped to nail down and quantify risks, and would have informed the debate better. One year on, an irreproachable, independent risk–benefit analysis of such research, perhaps convened by a body such as the World Health Organization (WHO), is still lacking.

The Times editors, who looked foolish in January 2012 after remarking that the H5N1 ferret transmission research should not have been done*, simply tow Nature’s line.

To clarify a point, the Fouchier and Kawaoka experiments on influenza H5N1 transmission did not make the virus more dangerous – they made it less dangerous for ferrets. How they affect the virus in humans is unknown.

I suspect that no one, not even WHO, has done a quantitative risk-benefit analysis of H5N1 research because it cannot be done. What basic research will reveal is frequently unknown – if the outcome could be predicted, then it would not be research. Scientists ask questions, and design experiments to answer them, but the results remain elusive until the experiments are done. How can the benefits be quantified if the outcome isn’t certain?

For example, one of the benefits of influenza H5N1 research is to understand what regulates aerosol transmission of the virus. It is without doubt an important question, but whether or not research will provide an answer is unknown. At best, we might identify the determinants of aerosol transmission in ferrets – but not in humans. I don’t know the solution to this problem –  should we simply assume that we will get answers to all the questions we ask? Should we conclude that H5N1 research will allow us to understand H5N1 transmission and pathogenesis, thereby leading to vaccines and antiviral drugs or novel therapies? In this case there is no doubt that the benefits of H5N1 research are very high, but I can’t put a number on it. Nature calls this a ‘qualitative’ argument. But if someone tried to make a quantitative risk-benefit analysis of H5N1 research it would be fiction.

What is the risk of influenza H5N1 virus research? Many influenza researchers feel that it is low, if work with infectious virus is carried out under the right containment conditions. Perhaps the more relevant question is what is the risk of releasing experimental results that could be used for nefarious purposes. Because H5N1 transmission experiments utilize animal models, the results cannot be directly extrapolated to humans. If a virus is isolated that transmits by aerosol among ferrets, it cannot be concluded that the same virus will transmit among humans. Also remember that gain of aerosol transmission among ferrets was accompanied by a loss of fitness – the altered virus did not cause lethal disease when transmitted by aerosol. It seems unlikely that these research findings could be used to successfully produce a biological weapon.

It seems unlikely that someone intent on producing an H5N1 biological weapon would base it on work done in ferrets, or any other animal model. Their solution would be to passage the virus in humans – an unethical experiment, but which one could imagine being done by unethical individuals. Even the outcome of this experiment would not be assured – no one knows if an H5N1 virus selected for aerosol transmission among humans would have high lethality.

I understand why the Times would ask for a cost-benefit analysis of basic scientific research – the editors are not scientists and do not understand the unpredictable nature of research. But I expected more from the science journal Nature. Have the editors who wrote this opinion forgotten how scientific research is done?

*Without having read the papers, the Times editors decided that the H5N1 ferret experiments should not have been done. When the papers were published we all learned that the modified H5N1 viruses were not lethal to ferrets.

Headline writers: Please take a virology course

Yesterday Denise Grady wrote in the New York Times about the end of the moratorium on influenza H5N1 virus research. The story headline read:

Research to resume on modified, deadlier bird flu

The Minneapolis Star Tribune reprinted Ms. Grady’s story with the following headline:

Studies will resume on deadly modified flu virus

Where do these headlines come from, outer space? The H5N1 viruses produced by Kawaoka and Fouchier, which transmit by aerosol among ferrets, are far less virulent than the parental H5N1 virus! Furthermore, the moratorium applied to all research on H5N1 virus, not just that related to these transmission experiments.

If most of the public obtains their virology information from the popular press, it is no wonder that much of the public distrusts these H5N1 experiments.

Yesterday I taught the first lecture of the 2013 version of my virology course (details forthcoming).  I told the students that one reason I want to teach virology is to enable them to understand why headlines like these are wrong.

Maybe some of my students will one day write the headlines and get them right.

End of moratorium on influenza H5N1 research

In early 2012 influenza virus researchers around the world decided to stop working on highly pathogenic avian influenza H5N1 virus. This decision came after work from the Fouchier and Kawaoka laboratories revealed the isolation of influenza H5N1 strains that can be passed among ferrets by aerosol. The moratorium on influenza H5N1 virus research has now been lifted, as described in a letter from influenza virologists to Science and Nature.

Lifting the embargo on H5N1 research is an important step forward for understanding what regulates influenza transmission. In my view it was an ill-conceived move, done to quell the growing concern over the adaptation of influenza H5N1 virus to aerosol transmission in ferrets. We now know that these viruses are not lethal for ferrets, and much of the outrage expressed about this work was misguided. In my view the moratorium has accomplished little other than delaying the conduct of important virology research.

According to the influenza virus researchers who signed on to the moratorium, its purpose was to:

…provide time to explain the public-health benefits of this work, to describe the measures in place to minimize pos- sible risks, and to enable organizations and governments around the world to review their policies (for example on biosafety, biosecurity, oversight, and communication) regarding these experiments.

An important consideration is the level of containment that will be required for studying influenza H5N1 transmission. WHO has released recommendations on risk control measures for H5N1 research, and individual countries will decided how to proceed. The US has not yet made a decision on the level of containment needed for H5N1 virus transmission research. Influenza virologists who participated in the moratorium have their own view:

We consider biosafety level 3 conditions with the considerable enhancements (BSL-3+) outlined in the referenced publications (11–13) as appropriate for this type of work, but recognize that some countries may require BSL-4 conditions in ac- cordance with applicable standards (such as Canada).

Their last statement forms the crux of the issue on H5N1 transmission research:

We fully acknowledge that this research—as with any work on infectious agents—is not without risks. However, because the risk exists in nature that an H5N1 virus capable of transmission in mammals may emerge, the benefits of this work outweigh the risks.

Origin of the H5N1 storm

ferretI still wonder why the influenza virus H5N1 ferret transmission studies generated such fear and misunderstanding among the public, the press, and even some scientists. I still cannot fully explain what transpired, but now that the papers have been published some new clues have emerged.

In my opinion, the main catalyst of the storm was the article Scientists brace for media storm around controversial flu studies by Martin Enserink. It began with the inflammatory statement ‘Locked up in the bowels of the medical faculty building here and accessible to only a handful of scientists lies a man-made flu virus that could change world history if it were ever set free’. Fouchier said that he created ‘probably one of the most dangerous viruses you can make’. Members of the NSABB were quoted as saying ‘I can’t think of another pathogenic organism that is as scary as this one’, and ‘This work should never have been done’.

This article presented a one-sided view because only Fouchier or NSABB members were quoted. I don’t understand why Fouchier made some of the statements that he did; perhaps he was quoted out of context. The NSABB members were on the way to restricting publication of the paper, so their views were clear. What Enserink did not do – what he should have done – was to speak with other virologists. This he could not do because the manuscript describing the work had not been made public. He violated a main tenet of journalism, to present both sides of the story.

With the publication of the Fouchier and Kawaoka papers, it became immediately apparent that all of the inflammatory statements in the Enserink article are wrong. For example, after 10 passages in ferrets, an altered H5N1 virus does transmit in the air among ferrets, but inefficiently and without killing the animals. Hardly one of the most dangerous viruses you can make.

To be fair, Enserink was not the first to report these findings. Fouchier presented the results of his H5N1 ferret transmission studies at a meeting in Malta in September 2011. A week later, New Scientist published an article on the findings entitled Five easy mutations to make bird flu a lethal pandemic. In the first paragraph, the author writes ‘…five mutations in just two genes have allowed the virus to spread between mammals in the lab. What’s more, the virus is just as lethal despite the mutations.’ A few paragraphs later: ‘The tenth round of ferrets shed an H5N1 strain that spread to ferrets in separate cages – and killed them.’ Both the title and these statements are all wrong. Fouchier’s published paper does not prove that five mutations are sufficient for aerosol transmission among ferrets, and the virus does not kill ferrets when it is transmitted through the air.  Also problematic is Fouchier quoted as saying that ‘The virus is transmitted as efficiently as seasonal flu’. Given the published data, and his comments at an ASM Biodefense Meeting in February 2012, I do not understand this statement.

It is easy to see how the misinformation in these two articles ignited the fear. Their stories were repeated by countless other publications without verifying whether or not they were correct, amplifying the false conclusions and spreading misinformation even further. Those of us who pointed out inconsistencies were dismissed as risk-takers. Even the New York Times without ever having seen the data – declared that the experiments should not have been done and the virus stocks should be destroyed.

A comparison of the original Fouchier manuscript with the version recently published in Science provides additional insight (I do not have the original version of the Kawaoka paper).  The first version of the manuscript was submitted to Science and reviewed by the NSABB, whose members recommended that it should be published in redacted form. Fouchier and colleagues then submitted a revised version which was reviewed by the NSABB, who then decided that the entire paper could be published.

Curiously, the titles of the paper are different. The original: Aerosol transmission of avian influenza H5N1 virus. The published version: Airborne transmission of influenza A/H5N1 virus between ferrets. The first is clearly ‘scarier’.

Another big difference between the two manuscripts is the length. A research article in Science is typically brief: it begins with a paragraph or two of background information, then delves right into the results. This is how the original Fouchier manuscript was constructed. In contrast, it is not until page five of the published version do we reach the data: the previous pages are filled with background material reminiscent of a review article. Included is information on the basic biology of influenza viruses, the functions of individual proteins, virulence, why the authors decided to do these studies, what is known to control host range and transmission, and the containment procedures that were undertaken. It is an impressive amount of background information, none of which was present in the original version.  The additional material does help to put the experiments in their proper context.

I found two examples of changes in the wording that I feel could make substantive changes in reader perceptions of the results.

In the abstract of the original manuscript, the authors wrote:

The virus acquired the ability to transmit via aerosols or respiratory droplets while remaining highly pathogenic to ferrets.

In the revised manuscript, this sentence has been changed:

None of the recipient ferrets died after airborne infection with the mutant A/H5N1 viruses.

The second version better represents the data. The first version is incorrect as stated because the virus is virulent only when inoculated intratracheally into ferrets, not after transmission by aerosols.

A second example occurs during discussion of the response of ferrets to infection with mutant H5N1 virus. In the original manuscript the authors note that after intransal inoculation, the animals have signs of disease but did not die. However, after intratracheal inoculation with the virus, all six ferrets died. Their conclusion:

 These data are similar as described previously for A/H5N1wildtype and thus do not point to reduced virulence (italics mine).

In the revised manuscript this sentence has been modified:

These data are similar to those described previously for A/H5N1wildtype in ferrets. Thus, although the airborne-transmissible virus is lethal to ferrets upon intratracheal inoculation at high doses, the virus was not lethal after airborne transmission.

The first version is misleading because it does not clearly state that virulence was assessed by intratracheal inoculation.

For the most part the same data are presented in the two versions of the manuscripts. A virologist would not draw different conclusions from the two manuscripts despite the longer introduction and the two modifications noted above. I remain puzzled as to why the first manuscript raised such a furor. I cannot believe that it was simply a consequence of overzealous writers and a few scientific overstatements.

Although the Kawaoka and Fouchier papers have been published, the effect of the H5N1 storm will linger for a long time. The moratorium on H5N1 transmission research continues, meaning that important questions cannot be answered. On 29 March 2012 the United States government issued its Policy for Oversight of Life Sciences Dual Use Research of Concern (pdf).  According to this new policy, seven different types of ongoing or proposed research (including transmission studies) on 15 different pathogens (including highly pathogenic avian influenza viruses) must now be reviewed by a committee for risk assessment and if the development of mitigation plans. Once the work is in progress, no deviations from proposed experiments are permitted without further review. I understand from a number of virologists that this policy has had a chilling effect on avian influenza virus research. As a consequence, this area of investigation is likely to substantially contract, depriving us of potentially important findings that could be useful in limiting influenza and other viral diseases.

We are in this position because access to the Fouchier and Kawaoka papers was restricted, and few could actually read them to understand exactly what was done. I can’t think of a better reason for unrestricted publication of scientific findings.

TWiV 190: The second ferret of the Apocalypse

On episode #190 of the science show This Week in Virology, Vincent, Alan, and Kathy review selection of influenza H5N1 viruses that can transmit among ferrets by aerosol.

You can find TWiV #190 at www.microbe.tv/twiv.

Influenza H5N1 virus versus ferrets, round two

H5N1 mutationsThe second of two papers on avian influenza H5N1 virus that caused such a furor in the past year was published today in the journal Science. I have carefully read the paper by Fouchier and colleagues, and I assure you that it does not enable the production of a deadly biological weapon. The results describe the requirements for airborne transmission of influenza viruses among ferrets, but it provides no information about human to human transmission. Failure to publish this work would have substantially restricted our understanding of influenza transmission.

The authors modified the HA protein of an Indonesian strain of influenza H5N1 virus so that it could attach to cell receptors in the ferret respiratory tract. They also added a change in one of the subunits of the viral RNA polymerase, called PB2 protein, that improves replication in mammalian cells (E627K). This H5N1 virus, with the amino acid changes HA Q222L, G224S and PB2 E627K, did not transmit through the air among ferrets.

In an attempt to select a virus with airborne transmissibility, the authors passed their modified H5N1 virus in ferrets. They inoculated a ferret intranasally with virus, waited 4 days, harvested virus from the respiratory tract, and infected the next animal. After ten ferret-to-ferret passages, the pool of viruses produced by the last animal contained mutations in all but one of the 8 viral RNA segments. The original alterations were present (HA Q222L and G224S, PB2 E627K) together with a new change in the HA protein, T156A. This amino acid change prevents the addition of a sugar group to the protein near the receptor binding site, thereby increasing virus binding to mammalian cell receptors.

After ten ferret to ferret passages, the modified H5N1 virus could transmit from one animal to another housed in neighboring cages, e.g. by the aerosol route. All viruses acquired by ferrets by this route had five amino acid changes in common:  the original three introduced by mutagenesis (HA Q222L and G224S, PB2 E627K) and two selected in ferrets, HA H103Y and T156A.

The modified H5N1 virus does not transmit with high efficiency among ferrets, and it is not lethal when acquired by aerosol transmission. For this reason, and because we do not know if the virus would transmit among humans, it would not be an effective biological weapon.

A minimum of five amino acid changes in H5N1 virus are required for aerosol transmission among ferrets. This conclusion is based on the observation that the viruses acquired by ferrets through aerosol infection all had five amino acid changes in common. The actual number could be higher. For example, one virus that was studied in more detail differed from the parent H5N1 virus by nine amino acid changes, and other mutations were identified in other isolates. Determining the exact number will require introducing mutations in various combinations into H5N1 virus and testing transmission in ferrets. At present these experiments cannot be because there is a moratorium on H5N1 transmission research.

How do these results compare with those of Kawaoka and colleagues? Those authors found that five amino acid changes in the H5 HA are needed for airborne transmission among ferrets. However, they used a different virus, the 2009 H1N1 pandemic virus with an H5 HA protein. The latter was modified so it could recognize mammalian receptors. They found that amino acid changes that shift the HA from avian to human receptor specificity reduce the stability of the virus. The amino acid changes HA N158D and T318I, which were selected during infection of ferrets, restore stability. The T318I change is near the HA fusion peptide, distant from the receptor binding site. [In the figure, changes identified by Kawaoka and colleagues are in red; black are those identified by Fouchier and colleagues].

It is quite possible that both Kawaoka and Fouchier independently found that virion stability is an important property of viruses that can be transmitted through the air among ferrets. I wonder if Fouchier’s alterations to the HA, Q222L and G224S, destabilized the protein, like those introduced by Kawaoka. This possibility is suggested by the presence of the HA T318I amino acid change that was selected in ferrets. Amino acid 156 is in the HA trimer interface and could confer stability to the protein (the viral HA protein is composed of three copies of one polypeptide; the interface of these  three proteins determines its stability). It is an hypothesis that can be easily tested (even with the moratorium on H5N1 transmission research).

The results demonstrate that 5 to 9 amino acid changes are sufficient to allow influenza H5N1 virus to transmit by the aerosol route among ferrets. The findings provide no information about aerosol transmissibility of H5N1 virus in humans. We cannot conclude from this work that a similar number of changes in H5N1 virus will allow transmission among humans. That information can only come from the study of a pandemic H5N1 strain (should such a virus ever emerge).

There is a great deal of good science in this paper, and I cannot imagine hiding it in a vault, or only providing it to certain individuals. I find the findings intriguing, and I am sure that other virologists will be similarly fascinated. One of them might do a seminal experiment on H5N1 transmission as a consequence. But that would never happen if the paper were not published.

The new manuscript is very different from the version submitted in 2011. That paper, in typical Science article style, contained only one paragraph of background information. The experimental findings are described tersely and with little explanation. In contrast, the first five pages of the revised manuscript read like a review article, with substantial detail on influenza virus biology, host range, and the precautions taken during conduct of the experiments. The experimental results are carefully explained. The style is considered and soothing, in contrast with the stark presentation of the original manuscript. I now understand why the NSABB changed their mind and decided to publish this version.

TWiV 182: One flu over the ferrets’ nest

On episode #182 of the science show This Week in Virology, Michael Imperiale joins the TWiV crew to discuss the recently published influenza H5N1 transmission paper and how it was viewed by the NSABB.

You can find TWiV #182 at www.microbe.tv/twiv.

Building the perfect bug

This past February I was interviewed by the Australian Broadcasting Company on the topic of the Fouchier and Kawaoka experiments on avian influenza virus H5N1. The video, Building the Perfect Bug, has been released by Journeyman Pictures and includes interviews with S.T. Lai, Laurie Garrett, Michael Osterholm, and Ron Fouchier (transcript available). It is far too alarmist for my taste, but both sides of the issue are represented.

The video includes sequences of me working in the cell culture laboratory. Note that I did wear a tie for my interview while Michael Osterholm did not.

TWiV 177: Live in Dublin

On episode #177 of the science show This Week in Virology, Vincent, Connor Bamford, Wendy Barclay, and Ron Fouchier discussed avian influenza H5N1 transmission experiments in ferrets and novel bunyaviruses at the 2012 Spring Conference of the Society for General Microbiology in Dublin, Ireland.

You can find TWiV #177 at www.microbe.tv/twiv.