No announcement yet.

eLife: Parallel Evolution Of Influenza Across Multiple Spatiotemporal Scales

  • Filter
  • Time
  • Show
Clear All
new posts

  • eLife: Parallel Evolution Of Influenza Across Multiple Spatiotemporal Scales

    eLife: Parallel Evolution Of Influenza Across Multiple Spatiotemporal Scales

    Credit NIAID


    We've an absolutely fascinating influenza study that looks back at the viral evolution of seasonal H3N2 in four immunocompromised patients sampled 10 years ago, and finds parallels between the mutations that occurred during their prolonged illnesses, and evolutionary changes to influenza viruses around the globe in the years that followed.
    While I let that sink in, a brief refresher on influenza evolution.
    Influenza viruses evolve via two well established routes; Antigenic drift & Antigenic Shift (reassortment).
    • Antigenic drift causes small, incremental changes in the virus over time. Drift is the standard evolutionary process of influenza viruses, and often come about due to replication errors that are common with single-strand RNA viruses (see NIAID Video: Antigenic Drift).
    • Shift occurs when one virus swap out chunks of their genetic code with gene segments from another virus. This is known as reassortment. While far less common than drift, shift can produce abrupt, dramatic, and sometimes pandemic inducing changes to the virus (see NIAID Video: How Influenza Pandemics Occur).

    In either case, most mutations/reassortments are evolutionary failures.

    They die out quickly because they are not as biologically `fit’ as the parental virus they must compete with. Only rarely does a mutation convey enough of an evolutionary advantage to allow it to become `fixed' in a host, and potentially transmitted onward.

    And while these mutations enable influenza viruses evade host immunity and avoid extinction, they are random, chance events. It is only through an absolutely astronomical number of rolls of the genetic dice that influenza viruses eventually stumble upon amore advantageous genome sequence.
    Today's study adds a new wrinkle to all of this, suggesting that while influenza's mutations are random - there are only a limited number of successful evolutionary pathways available to any given strain - and raises the possibility that future viral mutations may follow a (semi) predictable path.
    It is important to note that the original intent of this study was simply to examine how the flu evolves in an immunocompromised individual who can carry a flu infection for weeks or even months.
    What they found, quite unexpectedly, were several common mutations across 4 randomly selected cases. Even more surprising, these mutations heralded - by several years - evolutionary changes that would become fixtures in globally circulating seasonal flu viruses.
    First the abstract of the study, followed by some excepts from a press release by the Fred Hutchinson Cancer Research Center, where much of this work was conducted.
    Parallel evolution of influenza across multiple spatiotemporal scales

    Katherine S Xue1,2, Terry Stevens-Ayers3, Angela P Campbell3†,
    Janet A Englund4,5, Steven A Pergam3,6,7, Michael Boeckh3,6,7, Jesse D Bloom1,2*


    Viral variants that arise in the global influenza population begin as de novo mutations in single infected hosts, but the evolutionary dynamics that transform within-host variation to global genetic diversity are poorly understood. Here, we demonstrate that influenza evolution within infected humans recapitulates many evolutionary dynamics observed at the global scale.
    We deep-sequence longitudinal samples from four immunocompromised patients with long-term H3N2 influenza infections. We find parallel evolution across three scales: within individual patients, in different patients in our study, and in the global influenza population. In hemagglutinin, a small set of mutations arises independently in multiple patients. These same mutations emerge repeatedly within single patients and compete with one another, providing a vivid clinical example of clonal interference. Many of these recurrent within-host mutations also reach a high global frequency in the decade following the patient infections. Our results demonstrate surprising concordance in evolutionary dynamics across multiple spatiotemporal scales.

    DOI: 10.7554/eLife.26875.001

    We suggest that parallelism in HA evolution may emerge from the confluence of several evolutionary conditions (L?ssig et al., 2017). First, if selection acts concordantly across environments, it will favor a common set of beneficial mutations. Second, in a constrained evolutionary landscape, this set of beneficial mutations will be relatively small. Finally, given sufficiently large population sizes, high mutation rates, and time, these beneficial mutations will emerge and be selected to detectable frequencies. Our observation that similar mutations arise repeatedly within single patients, within multiple different patients, and at the global scale, suggests that at least some of these conditions may hold true.
    (Continue . . . )

    The entire (open access) study is well worth reading, as is the press release from the Fred Hutchinson Cancer Research Center, which provides a good lay explanation of the study and its findings.
    Date:June 27, 2017

    Source:Fred Hutchinson Cancer Research Center

    Summary: A new study has found that flu evolution within some individuals can hint at the virus's eventual evolutionary course worldwide. The study of 10-year-old flu samples also found the virus's evolution in individual transplant patients partially mirrors later global trends.
    A new study has found that flu evolution within some individuals can hint at the virus's eventual evolutionary course worldwide.

    Samples taken more than 10 years ago from people with unusually long flu infections -- and analyzed recently using modern genome sequencing methods -- revealed certain viral changes that matched global flu evolution trends several years later.

    The study, published in the journal eLife, tracked how flu evolved over time in four people who were especially vulnerable to unusually severe viral infections: bone marrow transplant patients. For people with healthy immune systems, a typical flu infection lasts about a week, said Fred Hutchinson Cancer Research Center evolutionary biologist and doctoral student Katherine Xue, first author on the study. So she and her colleagues at the Hutch, Seattle Children's Research Institute and the University of Washington studied viruses that originated from patients who received transplants and developed severe flu infections that lasted two or more months.

    These four patients were drawn from a group of nearly 500 transplant recipients who participated in a previous study led by Fred Hutch infectious disease researcher Dr. Michael Boeckh, also a co-author on this study. That large study began in 2005 to improve understanding about the sometimes devastating impact of respiratory viruses in this vulnerable population -- in fact, two of the four patients whose samples were analyzed in the current study went on to die of their infections.

    (Continue . . . )
    All of this work is in its infancy, and we are a long way from being able to predict what changes will occur in seasonal influenza 1, 2, or 5 years out. But this research suggests that flu's evolution may be less random, and more predictable, than we previously believed.

    One can't help but hope that Chinese researchers are taking note, and perhaps doing similar deep sequencing on H7N9 patients - who often shed the virus for weeks - to see if similar `common' mutations are turning up that might give us some early warning as to where that virus is headed.
    All medical discussions are for educational purposes. I am not a doctor, just a retired paramedic. Nothing I post should be construed as specific medical advice. If you have a medical problem, see your physician.

  • #2
    While the results should not be much of a surprise - given a natural selection model that postulates random mutation followed by winnowing based on fitness - there is a difference between the groups which the data set may help throw light on.

    While not explicitly covered in this investigation the difference lies in the winnowing. In the immuno-compromised group fitness is the ability to infect another cell in that host and replicate, a task further complicated for the general population by a step involving transmission to a new host. This additional step involves getting out of the host, surviving in the environment and getting in to the new host. These new tricks may have their own set of desired mutations which would have no fitness advantage to the billion plus in-host replications that precede a single host jump.
    These 'jumpers' are vanishingly rare and I previously described them as flu's Columbus & Magellan. Pushing the analogy further in the virus' case - for all but the immuno-compromised - the immune system kills off all of the virus in the host our explorers departed from. The only surviving genetic material is within them and, as new hosts are usually uninhabited, only these genes can restart the process. As a consequence any group that are not equipped to make the journey become extinct - regardless of how fit they were in the old world. This gives the rare explorer a genetic advantage great enough to counter-balance his rarity.

    In theory playing spot-the-difference between the mutations observed within the two groups should show which mutations are 'jump' specific. In practice, the data set's size only allowed for firm correlation between the two data sets and common strains in the immuno-compromised ?partially mirrors later global trends?. However its predictive abilities are less secure and here we are using a subset of that data so will be lucky to see even a weak signal. This could be solved with a larger data set, with size comes statistical significance.

    On Gain of Function (GOF) and Immunocompromised zoonosis.
    These are really all part of the same thing. It is the ability of the immunocompromised to keep the virus and the immune-system in a state of stalemate for long periods that has permitted this longitudinal study. For the rest of us it would have got us or we would have got it, either way it would have been over in weeks.

    What made them so good for this research is exactly what makes them ideal laboratories for anti-microbial resistance research, which is basically what they are doing. It is ideal for creating multi-drug resistance bacteria and virus, assuming the patient is lucky enough to have access to these drugs. Apart from this studies? transplant patients HIV and other pharmaceutical, or natural, causes of immunosuppression aid Multi-drug-resistance.
    The segue into GOF may be less obvious. In a comment, on another of Mike?s excellent posts ? on GOF, I argued for the continuation of Fuchier?s H7H9 research with the proviso that sequences be kept but ?improved? live virus not be kept. This is the equivalent to watching the mutations in the immune-compromised set but making sure nothing can get loose into the general population. In this research, and the H7N9 GOF, flu was being observed under selection pressure in the hope of learning which mutations kept popping up as possible forewarning of danger signs in the evolving wildtype consensus strand sequences. Unlike this H3N2 research the H7N9 selection pressure was directed, in that their experiment was designed to look for improved host-to-host transmission mutations, they were selecting for Columbus.
    If the H3N2 data did hint at ?jumpers? then comparison with Fuchier?s ?jumper? data might show mutation traits that remain stable across subtypes.
    Last edited by JJackson; July 4, 2017, 08:55 AM.


    • #3

      Thanks for the additional insight. I particularly like your description of `jumpers'.

      While the findings of this study initially surprised me, the more I thought about it, the more it seemed liike an almost inevitable result. There are only a limited number of evolutionary pathways that a virus can take and retain its biological advantage. Whether the number of viable pathways is small enough to make predictions (based on a much larger sampling) possible, is another matter.


      All medical discussions are for educational purposes. I am not a doctor, just a retired paramedic. Nothing I post should be construed as specific medical advice. If you have a medical problem, see your physician.


      • #4
        For those who have been following this thread I would recommend this old, but still very relevant, paper. ?Characterization of Multidrug-Resistant Influenza A/H3N2 Viruses Shed during 1 Year by an Immunocompromised Child?

        The full paper can be found at the link but I found it tricky to read so have included a resume.

        Some poor child was both immunocompromised and had H3N2 seasonal flu early in April 2005 and the disease progress was tracked over the course of a year (another longitudinal study).

        Oseltamivir& Zanamivir (Neuraminidase Inhibiters - NIs).
        They instituted a course of Oseltamivir but on the 22nd of Apr they found the child had become co-infected with H1N1 (which was NI susceptible). By the 10th of June she had developed a 274 fold increase in resistance (NA E59G,E119V & I222V) which had increased to 1000 fold by Aug. 16 (E119V & I222V) at which point Zanamivir is substituted and seems to clear the infection.

        Amantadine (M2 ion blocker).
        The addition of Amantadine therapy was initiated on 26th of July but 5 days later M S31N (a potent M2 blocker) had appeared.

        On the 15th of Oct. all treatment was stopped, due to a series of clear test, however by the 1st of Nov. it was back and Zanamivir/Amantadine combination treatment was re-introduced and the sample from the 4th showed a 193 fold decrease in NI efficacy (mix of WT & E119V) and S31N had remained on M2.
        Between Nov. 2005 to Apr. 2006 this combination therapy failed to clear the infection but no major Zanamivir resistance developed.

        What I would take from all this is that in this ? admittedly rather unusual patient ?

        M2 S31N appears rapidly (5 days), seems to have a negligible fitness penalty and does not have a secondary mutation which worsens its effect.

        NA E119V caused a 131 fold increase while I222V did little but together they produced a 1000 fold increase so there is obviously some kind of synergy. It took about 6 weeks to achieve this double mutation and in the month of no treatment that followed it seems to have been reverting to the wild type, implying some fitness penalty and, under Zanamivir treatment, to eventual decreased to almost no Oseltamivir resistance. Despite about 6 months of Zanamivir treatment the virus never managed to find any significant mutations. However it also never managed to clear the H2N3 infection.

        As a side note in this paper ?Evidence for Zanamivir Resistance in an Immunocompromised Child Infected with Influenza B Virus? ( ) a 1000 fold increase in Zanamivir resistance was detected within 2 weeks (R152K N2 numbering) however this child was so ill that treatment had to be stopped and, sadly, the child died two days later.

        While the two papers I have discussed here, and the E-life paper, all add pieces to the puzzle each only contributes a small amount of data to the total sequence database. All are very much ?special cases? and although they may give tantalising hints they may not be very good models for behaviour in patients with a normally functioning immune system. It was very much with this thread in mind that I wrote the second post in ?Influenza databases need root and branch reform? ( ) which looks at a rather different way of performing research based on a system that places sequence generation first. The post?s aim is to explain the rational for this change of emphasis.
        Last edited by JJackson; July 18, 2017, 04:13 PM.