Recombination vs. Reassortment

Here is a story in today's Wired News about Henry Niman and his ideas about viral evolution in the Avian Flu (H5N1).  While the text of the story is a bit unclear about the difference between recombination and reassortment, one of the associated images is quite nice.  This is yet another take on the specific mechanisms of viral evolution.  The figure defines reassortment as the emergence of a new strain via the replacement of whole genes from another (related) virus, and defines recombination as the insertion of fragments of genes into a new viral strain from another genome, potentially from the host.

Ignoring what labels are used, it seems the important point is that there may be two mechanisms for introduction of new sequences into an influenza viral genome; 1) inclusion of whole genes into a segmented genome or 2) insertion of gene fragments from another strain or species within a given viral gene.

Niman seems to think that not only is there evidence that the current H5N1 strain is evolving via the second mechanism, but that this is also the origin of the Spanish Flu (see my post "The Spanish Flu Story"), despite the fact that there appears to be a historically low occurrence of homologous recombination in negative sense RNA viruses (see my post "A Confluence of Concerns").

Reverse Genetics for H5N1 Vaccine

Just as I was banishing my ignorance about how the forthcoming trial vaccine for H5N1 was produced, a trio of excellent articles from the Wall Street Journal and Fortune landed in my Inbox, facilitating my education.  The short story is that the virus grown in chicken eggs as the source of an attenuated vaccine is not actually H5N1.  The genes that cause the virus to be so fatal to eggs have been replaced with genes from less virulent strains, while the HA protein on the surface of the virus is modified so that it is more stable.

Alas, because the WSJ doesn't allow you to look at their list of stories in the print edition without a subscription, I can't even provide links to the stories.  Fortune, evidently, is more forthcoming.  Here are the titles, etc;

"A primer on the Threat of Avian Flu...", by Gautam Naik, and "Avian Flu Poses Challenge to Global Vaccine Industry...", by David Hamilton and Gautam Naik; both are from the 28 Feb, 2005 issue of the WSJ.  "The Coming War Against Bird Flu", by David Stipp, will appear in the 7 March, 2005 issue of Fortune.

The upshot of the three articles is that the vaccine is produced in sterile chicken eggs via a recombinant virus that is a modified version of H5N1.  This strategy requires a large number of those eggs, which are not easy to come by, and produces a vaccine that prompts the production of antibodies against a virus that may, or may not, be similar to the wild type H5N1.  That is what human trials will have to determine.

Thus my initial concerns (here and here) about this issue were not so far off target.  Stipp's article does an excellent job describing the production of the vaccine, and associated challenges.  It is pretty clear we need to come up with alternative means of producing vaccines, preferably rapid synthetic approaches that are deployable from a distributed infrastructure.

UPDATE (7 March 2005): I stumbled over this article in The Scientist, "H5N1 vaccine strain in a week", from 29 January 2004, which opens;

A prototype vaccine strain of the H5N1 flu virus causing havoc in Asia will probably be ready next week, John Wood of the UK National Institute for Biological Standards and Control (NIBSC) told The Scientist today (January 29). However, months of other hurdles remain before it may be ready for public health use.

The article describes several genetic manipulations of the H5N1 strain that will make it easier to produce in chicken eggs, beginning with the removal of, "a stretch of 4 or 5 basic amino acids at the hemagglutinin cleavage site that allows the virus to replicate in every organ of a chicken's body, rather than respiratory and gut tissue normally infected".

The article cites Klaus Stohr as saying, "The H5N1 virus kills chicken eggs, the normal medium for growing flu vaccine viruses, so the WHO laboratories are using reverse genetics to lower the pathogenicity of the virus to chickens and to get a high yield in the egg cultures", and describes the additional genetic manipulations; "Using other lab strain flu plasmids containing the other components of the viral genome, the team will then reassort the pieces into a nonpathogenic vaccine strain." 

Finally, the article suggests that, "Sufficient amounts of safety-tested prototype vaccine virus will probably be available for the necessary 1 to 2 months of clinical trials in the next 4 weeks".  The date of this article, again, was 29 January, 2004.

Looks like we are well on our way, circa January 2004, to producing a lovely vaccine against a bug that doesn't actually exist in nature.  We clearly need an alternative to attenuated (or killed) whole virus vaccines.  When I have time, I will post what I have been learning about DNA vaccines.

Detecting and Diagnosing Avian Flu

"Bird fly outbreaks may go unnoticed in humans", a news piece in the 26 February, 2005 issue of New Scientist, reports that human cases of Avian Flu (H5N1) may be misdiagnosed.  Several patients in SE Asia have presented with symptoms unusual for the flu, and only after death did they test positive for the virus.  The piece also reports that the WHO is "analysing blood samples from people in areas affected by h5N1 to see how many carry antibodies against the virus".

The difficulty here is that it can take up to several weeks (say, one to three, depending on the etiology of the bug) for the adaptive immune response to produce antibodies against a pathogen.  It appears that people are dying within that time frame, which means that testing for antibodies is unlikely to be a useful diagnostic tool, at least given standard assay sensitivities.  Using reverse-transcriptase PCR (RT-PCR), it may be possible to detect the RNA genome of the bug, but clinical PCR is a true art.  It is often quite difficult to see anything via PCR in a clinical sample, unless you can really clean it up via purification.  That purification, however, particularly in the case of RNA, tends to reduce the sensitivity of the assay by removing or destroying the target nucleic acids before the amplification step.

Unrealistic H5N1 Vaccine Expectations?

I still haven't been able to determine what magical means will be used to produce a vaccine against the H5N1 strain of Avian Flu.  Press is very thin on how production and testing of the vaccine is going.  Yet policy decisions are being made based on the notion that the vaccine will be available in quantity soon.

A press release on the CIDRAP site from the World Health Organization notes that WHO will probably recommend governments start stockpiling vaccines against H5N1.  The release also cites unnamed "U.S. officials" who say that clinical trials of vaccines from Chiron and Sanofi-Pasteur are supposed to start soon, while also noting that, "H5N1 may not match the pandemic strain, the vaccine's shelf life of up to 2 years is relatively short, and, because companies have not yet begun clinical trials, licensing of the vaccine is months away."

And in another release, the CIDRAP site quotes Michael Osterholm, who is director of the University of Minnesota Center for Infectious Disease Research and Policy, "We don't have a pandemic strain of vaccine yet, and we don't have any idea whether any of the vaccines to date would be efficacious."  In the Technological Challenges to Vaccine Development section of the Pandemic Influenza overview at CIDRAP, we find; "Highly pathogenic avian strains cannot be grown in large quantities in eggs because they are lethal to chick embryos."

To the extent that we should trust the popular press on this issue, as part of a short story on What You Need to Know About Avian Flu, the 9 February, 2004 issue of Business Week states, "Vaccines are usually produced in chicken eggs, but H5N1 is lethal to fertilized eggs."

Yet a 24 February, 2005 story on Newsday.com says, "Two million doses of vaccine are being stored in bulk form for possible emergency use and to test whether it maintains its potency," while 8000 doses are, "nearly ready to be shipped to the National Institute for Allergy and Infectious Diseases for clinical trials."

Perhaps the vaccine about to enter trials is from source other than chicken eggs?

In 10 February, 2005 testimony before the The Committee on Government Reform, Jesse L. Goodman, Director of the Center for Biologics Evaluation and Research at the FDA, while describing how the Department of Health and Human Services will spend roughly a billion dollars over the next few years on influenza related activities, said; "While work remains to obtain sufficient vaccine yields and evaluate cell-based vaccines for their safety and effectiveness, moving from an egg-based production to a cell-culture production can potentially shorten the time needed to produce vaccine as well as decrease the risk of contamination inherent in egg-based production."  That is, there isn't yet a functional alternative to using chicken eggs to produce vaccine.

So what gives?  I can only speculate that details about the vaccine are being closely held until more is learned about how it behaves in humans.  But with so many sources suggesting the vaccine can't be grown in eggs, I have to wonder what tricks Chiron and Sanofi-Pasteur have come up with to produce it in bulk.  Perhaps it is a low yield process and they have concentrated the virus produced from a much larger number of eggs?

I wish someone would come out and clearly explain where the vaccine is coming from and how it is produced.  The issues of what infrastructure exists to make vaccines, how much can be made, and whether it will be effective are quite critical for charting our course as we prepare for a potential pandemic.

Avian Flu Uncertainties

(First, here is the NIH Focus on the Flu site.  Decent general info there.)

Klaus Stohr is the chief of the World Health Organization's global influenza program.  He is worried that we are overdue for a flu pandemic.  In this profile in The Lancet, he is attributed with the observations that flu pandemics occur on average every 27 years, that the last one hit 37 years ago in 1968, and that between 2 and 7 million people could die in the next pandemic.

As a veterinarian and influenza specialist, Dr. Stohr obviously knows a lot more about flu bugs than do I.  However, his statistics may need a second look, particularly for incidents in the past 100 years.  Arnold Monto notes in his New England Journal of Medicine perspective "The Threat of an Avian Influenza Pandemic" (27 Jan 2005) that, "There have been three influenza pandemics during the past century -- in 1918, 1957, and 1968."  It is true that the average interval between these three events is just under 30 years.  I don't know how many data points Dr. Stohr is working with, but the width of the distribution, in this case, is hardly even computable for pandemics this century.  The interval between events is just as likely to be 40 years as it is 30 (not so comforting, I admit).  In any event, given the state of modern medicine, travel, and sanitation (and the variability in all those things across the globe) nobody should be drawing firm statistical conclusions from the three most recent data points.  The point of this is that because this bug is not behaving as expected, perhaps we should reevaluate our expectations.

How much do we really know about pandemic strains?  Perhaps a good place to start is examining how similar the three 20th century strains were.  Not very, I am beginning to think.  Although each were a novel type A virus of avian origin, Monto observes that, "In 1957 and 1968, the new viruses had components of previous human viruses as well as avian viruses...it was determined retrospectively that in both cases, there had been a reassortment of avian and human genes -- most likely the result of the coinfection of a host by two different viruses."  Monto then notes that the 1918 strain appears to have resulted from mutation in an avian strain (see my post, The Spanish Flu Story).  So, we are down to two pandemic strains, out of only three total, that arose through the historically low probability process of reassortment (see the end of my post, A Confluence of Concerns).  The numbers aren't looking good for deriving general principles about potential pandemic flu strains.

Adding to the confusion is the fact that, according to Monto, "The genetic characteristics of [H5N1] are still completely avian; neither mutation nor the sharing of genetic material with a human virus has taken place."  (I don't entirely understand this statement in light of assertions that H5N1 is becoming more pathogenic in poultry -- how else would this occur than by mutation?  Or is recombination amongst avian strains the assumed mode of increase pathogenesis?)  Klaus Stohr himself, in a 27 Feb, 2005 editorial in NEJM, "Avian Influenza and Pandemics -- Research Needs and Opportunities", wonders;

Why has H5N1 not reassorted with a human influenzavirus?  It certainly has had ample opportunity to do so...Unprotected workers [destroying infected poultry have] had intense exposure, as did health care workers.  Virologic surveillance has demonstrated the concurrent circulation of human viruses.  Hence, one conclusion is tempting: if H5N1 could reassort, it should have done so by now.  The explanation may lie in sheer statistical luck.

Hmmm.  That's not so satisfying. 

The last thing I want to do here is undermine the efforts of experts to understand what is going on and to try to prevent a pandemic.  However, I can't square public statements about the risk we face with what data I find in the literature.  There is definitely a troublesome lack of information about how flu bugs work, how the evolve, and what we might do to stop them, particularly with vaccines.  This press release, dated 27 May 2004, from the National Institutes of Allergy and Infectious Disease, says vaccines against H5N1 will be made by Chiron and Aventis Pasteur using the traditional chicken egg method.  While I have informally heard that H5N1 is so lethal that it kills chicken embryos before they can produce an adequate amount of virus to use as a vaccine, I still haven't been able to confirm whether or not it is technologically possible to produce an H5N1 vaccine this way.

So what do we do?  Stohr, again; "Substantial gaps in knowledge remain, making the ability of science to guide policy imperfect at a critical time."

Indeed.

CIDRAP on Flu Bugs

The Center for Infectious Disease Research And Policy (CIDRAP), at the University of Minnesota, has an excellent web site for those interested in H5N1 and Pandemic Flu strains.  The site also covers Biosecurity, Bioterrorism, Food Safety, BSE, and SARS, amongst others, though I have yet to peruse those topics.

The section on "Pandemic Influenza" briefly mentions the problem that vaccines for H5N1 can't be made via the usual high-technology route of infecting chicken eggs because the virus kills the chick embryos too quickly.  I've heard this before, but can't find any references in my stash of PDFs.  Does anyone know of a decent paper/website/NY Times story that explains this in some detail?  Similarly for a review of efforts to grow virus in mammalian cell culture?

It's really quite embarrassing that we are stuck using century-old technology to combat these viruses.

The Spanish Flu Story

(UPDATE 15 Feb 06: Because so many people are finding their way to this post from Google and other search engines, I have reorganized the text to make it easier to read.)

Extending my earlier post "A Confluence of Concerns", on the potential for an epidemic from Avian Flu H5N1 and similarities between its emergence and the 1918 Flu:

James Newcomb at bio-era pointed me to a recent paper exploring the possible origins of the 1918 Spanish Flu.  In "A hypothesis: the conjunction of soldiers, gas, pigs, ducks, geese and horses in Northern France during the Great War provided the conditions for the emergence of the "Spanish" influenza pandemic of 1918-1919", Oxford et al. explore the hypothesis that this killer flu strain emerged at a large British Army camp in France during the Great War.

At the outset, the authors note that;

Four of the eight genes of influenza have now been sequenced and there is no clear genetic indication of why this virus was so virulent, though the NS1 gene-product may have played a role. Therefore, we need to examine the particular circumstances of 1918, such as population movements and major events of the time. Obviously, the unique circumstance of that period was the Great War. Could the special circumstances engendered in the war itself have allowed or caused the emergence, evolution and spread of a pandemic virus?

They go on to compile molecular, epidemiological, and historical evidence related to conditions in and around the base at Etaples, in Boulonge, which housed soldiers on the way to the front as well as large numbers of wounded brought by train directly from the front each night.  In particular, Oxford et al. note that more than one million soldiers moved through the camp by November 1917, with symptoms consistent with the flu appearing there as early as December 1916.  The camp is described as overcrowded, with the 100,000 troops quartered there housed in tents and temporary barracks.  There were numerous pigs, fowl, and horses in the vicinity, some of which were prepared for food by the troops themselves.  Finally, a great many of the troops in the area had been exposed to chemical weapons, some of them now known to be mutagenic.  That is, a large number of soldiers were living in very rough conditions, many of them with respiratory systems compromised by gas attacks, amidst animals known to carry viruses that jump to humans or recombine with viruses that we host.

So the conditions were ripe for more than one virus to be proliferate in immune compromised patients (taking the lungs as a component of the immune system), a necessary condition for recombination to take place within humans.  However, I find it particularly interesting that many of the gas weapons used in that area are mutagenic.  The authors note that no one has looked into the possibility that mustard gas, or any of the other weapons as far as I can tell, can "accelerate mutations in viruses such as influenza".

They conclude;

The evidence presented for 'seeding' of the 1918-1919 influenza pandemic up to 2 years earlier and the lack of a Chinese/Far East origin contains lessons for the future. In terms of advance planning for the next influenza pandemic, it should be recognised that it could emerge anywhere in the world when particular combinations of factors arise. The epicentre could be Hong Kong but it could equally be Saudi Arabia, Pakistan, Uruguay and other South American countries, Africa, Thailand and even some regions of modern day Europe. Influenza pandemic surveillance could be increased in all these regions.

So even if we don't see H5N1 emerge in Southeast Asia in the next year or so, that doesn't mean a strain that originates there won't become a problem elsewhere at a later date.  As for whether the conditions to create a killer strain in tsunami stricken regions are similar enough to the camp at  Etaples, it is probably not possible to draw many firm conclusions.  If a malaria outbreak occurs, then we may be in for trouble.  Yet the root cause of the transformations that brought the 1918 strain into being are still unclear; was it a recombination event or a series of mutations?  There are a number of papers that demonstrate that a key gene from the 1918 strain contain regions very similar to those in a strain that infects pigs.

However, the question of mutation or recombination seems to hinge on the assumptions used to construct models of the the lineage of the virus.  The origin of the hemagglutinin gene (HA) is, in particular, critical to sorting out how the bug came about because HA is the protein that enables viruses to bind to host cells and initiate infection.  It is also the primary viral target for the host immune system.  Thus, acquisition of HA domains that fool the human immune system, either by mutation or recombination, make viruses more effective in infecting us, and figuring out how those changes came about may help us understand the causes and likelihood of future outbreaks.

In "Questioning the Evidence for Genetic Recombination in the 1918 "Spanish Flu" Virus", Worobey et al., conclude that, "Phylogenetic analysis of [HA] gene sequences has indicated that the [1918 strain] was more closely related to the human lineage than to the swine or avian influenza lineages of the H1N1 subtype", and that, "The apparent recombination...results from difference in the rate of evolution between HA1 and HA2 -- a difference present only in human influenza A viruses".  The group who published the tree supporting porcine origin, or more specifically, recombination between human and porcine flu strains, maintains their position in a response published with Worobey et al.

Finally, Reid et al. analyzed both the HA gene and the neuraminidase gene (NA) from the 1918 strain, concluding that its HA and NA genes were avian in nature, and that the virus had been been adapting within a mammalian host for at least several years preceding 1918.  They also note that pigs evidently came down with the same flu in the fall of 1918, which seems to indicate pigs got it from us, not the other way around.

I can't say that any of this helps me sort out whether current conditions in SE Asia mean we are particularly at risk of pandemic strain of Avian Flu emerging soon.  If nothing else, it is clear that we need to put more effort into understanding the evolution of RNA viruses, in particular.  And precisely because it is unclear whether the 1918 strain emerged due to mutation -- perhaps aided by the use of mutagenic chemical weapons on the battlefields of France -- or just plain old recombination, we need to do whatever is possible to reduce the chance of other diseases, such as malaria, producing conditions conducive to the spread of flu bugs in Asia.

GENETICS PRIMER:

First a few words worth of primer on terms.  Mutation is an alteration of nucleic acid sequences, caused by mistakes in enzymatic copying, ionizing radiation, or chemicals, within the genome of an individual member of a species.  Recombination is an exchange of genetic material between individuals of the same species at the level of individual strands of nucleic acids. For example, Strain A and Strain B of a virus may coincidentally infect the same cell, thereby creating the opportunity for a strand from each to recombine to form a new, hybrid sequence, resulting in Strain C that combines features of A and B.  Reassortment is an exchange of chromosomes between strains, which is particularly relevant for segmented viruses.  Flu viruses can pick up human or avian sequences by swapping whole chromosomes, and vice versa. (Edited for clarity, 3 March, 2020.)

A Confluence of Concerns

I've started wondering about the worrisome overlap of countries affected by the recent tsunami in southeast Asia and reported instances of human infection with Avian Flu (H5N1).  I'm not the only one, fortunately, who is thinking about this.  Henry Niman (who knows much more virology than I) is keeping a list of interesting stories on this over at recombinomics (see the "in the news" section).  There is a 70% mortality rate in humans, which is certainly frightening, and there are now confirmed cases of human-to-human transmission between people living in close quarters, but I think it is important to delve a little deeper into what may be going there.

It isn't yet clear what changes would be necessary in the virus to make it the cause of a true pandemic.  Even the causes of the 1918 "Spanish Flu" are still under debate.  The great concern for H5N1 is that it will recombine with a strain that already easily infects humans.  This has long thought to be the way the Spanish Flu became so deadly, but recently some debate has emerged along the lines that mutation rates in some areas of the hemaggluttinin gene (HA) were accelerated instead.  That is, mutation within the genome, rather than recombination, may have created enough variation to result in the virus that killed tens of millions of people.  In order for recombination to operate, two different virus strains must simultaneously infect the same cell, providing the opportunity to mix their genes.  However, it turns out (see below) that homologous recombination among RNA viruses appears to be a low probability event.

For his part, the good Dr. Niman is quite firm about the role of recombination;

This is the key issue on the influenza pandemic.  The 1918 H1N1 virus gained its lethality by recombining, not reassorting.  The same thing has happened with H5N1.  The H5N1 in Thailand and Vietnam have already picked up pieces of genes that are not in any other H5N1 isolates.  These polymorphisms are found in mammalian isolates such as humans and pigs (and the 1918 isolate had polymorphisms normally found in humans and pigs).

Needless to say, we will never know exactly how the 1918 strain came to be.  But its transformation into a pandemic strain is of definite interest today.

There are stories running around that the 1918 flu was the result of a peculiar set of circumstances. [UPDATE: See my post The Spanish Flu Story.] I have only heard this story as hearsay, so if anyone knows where it came from give a yell (hopefully it isn't from an obvious book I should have read).  Essentially, the story blames the 1918 Flu on World War I.  Large numbers of wounded troops were being removed from disease ridden conditions on the battlefield, and then moved through various hospitals, with the most grievously ill and wounded becoming ever more concentrated along the way.  It is argued (not by me) that this provided a remarkable opportunity for the virus to thrive and evolve amidst a large number of immune suppressed patients.  As the sick and wounded were moved from hospital to hospital, they may have carried flu variants with them, and when introduced into a new ward inoculated the patients already present with new strains.  Whether or not this story is an accurate rendition of the origin of the 1918 strain, it does get the brain ticking over.

What I find particularly troublesome in current events is the confluence of the H5N1 infections with a potential malaria outbreak resulting from conditions brought about by the tsunami.  There are two potential things that must happen in order for H5N1 to become truly dangerous to large populations.  The first is that it must find initial purchase in humans in order to replicate itself, and the second is that it must replicate in sufficient numbers and diversity to produce a more virulent strain.  The former is already happening on a small scale, as the human to human transmission cases illustrate.

But it is a virtual certainty that more people have been exposed to the virus than have become ill.  The immune systems of those who have escaped illness have been able to fight off the bug.  This means H5N1 hasn't had much of a chance to adapt itself to humans as hosts.  But what happens if H5N1 has the opportunity to infect large numbers of immune suppressed (or immune challenged) people?  I fear that this may come to pass if a malaria epidemic does strike areas affected by the tsunami.  H5N1 may thrive in such conditions, and whether its genome is altered by mutation or by recombination with other strains, variation and selection will definitely both be operating.  The parallels to the hypothetical origin of the 1918 flu are alarming, particularly in the context of modern rapid travel.

It would be nice if our knowledge of epidemiology and molecular biology could help us understand the probability of H5N1 becoming a pandemic-causing strain.  But as far as I can tell, we just don't know enough yet.  The furthest I have gone down this road is reading (and digesting as much as I could) a paper entitled, "Phylogenetic analysis reveals a low rate of homologous recombination in negative-sense RNA viruses," by Chare et al, in the Journal of General Virology (2003, 84, 2691-2703).  This is a bioinformatic study of 79 gene sequence alignments from 35 negative sense RNA viruses, including the Spanish Flu. 

I can do no better to explain this paper than to quote from it;

Overall, our study reveals that recombination is unlikely to be a frequent process in negative-sense RNA viruses, with only a few clear-cut examples in the 79 gene sequence alignments studied here. While we were unable to estimate precise recombination rates from our analyses, it is clear that these rates must be lower than those of mutation, which is not the case in some other viruses. Indeed, the absence of any detectable recombination in 20 of 35 negative-sense RNA viruses suggests that they may be entirely clonal organisms, although this will clearly need to be confirmed with much larger sequence data sets.

It is important to reiterate this is essentially a theoretical study based on historical data.  The authors performed no experiments.  However nice our stories, making testable predictions and doing experiments are the only way we can get close to the truth.  If our models were better maybe we could get at a decent prediction for the behavior of H5N1.  Perhaps in turn this would enable a bit of practical planning in the field, as well as an estimate of the economic consequences of action and inaction.  The best we can do in this case is probably to marshall relevant historical, economic, and scientific stories, and perhaps combine this with some savy scenario planning.  But when it comes to nailing down details, we may just have to wait and see in this case.

We've picked up this story as an internal research project at Bio-Economic Research Associates.  If you are interested in contributing, or in supporting a more concentrated effort, let me know.