Scanning hemagglutinin to predict the future of flu

From the Bloom Lab, Basic Sciences Division and Bedford Lab, Vaccine and Infectious Disease Division

Every year the World Health Organization (WHO) is tasked with the unenviable duty of deciding on the exact composition of the annual flu vaccine.  The influenza virus is rapidly evolving, tweaking the amino acid composition of the proteins that cover its surface partly to avoid recognition by the human immune system.  The surface proteins, hemagglutinin (HA) and neuraminidase (NA), are constantly evolving in response to this pressure.  Each flu season, a handful of variants emerge from the thousands circulating in the human population and dominate next year’s flu outbreak.  The WHO needs to be able to predict next year’s dominant strains based on this year’s circulating viruses.  Using historical data on abundance of major circulating strains and an imperfect understanding of the evolutionary dynamics of the virus, the WHO’s predictions sometimes miss the mark yielding less effective vaccines.  Each year’s flu vaccine is designed to neutralize the three influenza variants thought most likely to cause human suffering and death.

The laboratory of Dr. Jesse Bloom in the Basic Sciences Division and Dr. Trevor Bedford’s group in the Vaccine and Infectious Disease Division report a comprehensive “deep” look at the influenza virus’ evolutionary potential in the journal Proceedings of the National Academy of Sciences.  This comprehensive look may enable WHO scientists and epidemiologists to better predict which influenza variants are more likely to dominate the next flu season.  The authors, led by graduate students Juhye Lee (Bloom lab) and John Huddleston (Bedford lab), sought to determine the ability of each of HA’s 566 amino acid positions to tolerate substitution by each of the 19 other possible amino acids.  In order to obtain results relevant to recent flu outbreaks, the study used the amino acid sequence of the HA that was part of the 2010-2012 flu vaccine as the starting point.  The authors used a tissue culture method of selection to determine the relative ability of a variant to replicate through multiple rounds of infection.  To do this, they created a library of 10,754 (19 X 566) individual HA point mutants. 

Influenza strain trajectories over time.
Influenza strain trajectories over time. Provided by the Bloom Lab.

The library containing DNA encoding the mutant HA proteins was split into four pools.  The first pool was sequenced prior to selection to determine the relative abundance of each HA variant.  The remaining pools were subjected to identical rounds of selection in MDCK dog kidney cells as hosts to provide three independent replicates.  Viral RNA from the three replicates of the selection was sequenced to determine the relative fitness of each variant and the inherent variability of the measurement.  The results presented a treasure trove of information.  For starters, a lot is known about HA’s ability to accommodate mutations just from the analysis of current and past influenza strains.  Some positions are invariant, such as cysteine residues that form intra-protein disulfide bonds needed for proper folding and stability.  The selection process correctly identified many residues that are never mutated in nature.  At the other extreme are residues that are often mutated, for example residues that span epitopes recognized by antibodies.  Again, the selection process was able to identify many of the tolerant positions in HA.  In addition, the process identified many previously untested residues that were either invariant or highly variable.

The HA protein can be thought of as composed of two domains, the head, which protrudes out, away from the virus particle and is exposed to the immune system, and the stalk, which spans from the head to the virion’s membrane and is less exposed.  Previous work with an HA variant from a virus from the 1930’s suggested that the head domain is much more tolerant to amino acid substitution than the stalk.  Lee and colleagues found that in their HA, the head and stalk were able to tolerate substitutions to an equal extent.  The difference may be due to subtle differences inherent to the two variants, the 2010’s HA is an H3 subtype while the 1930’s HA is an H1.  A major goal of the study was to determine which mutations are likely to persist in nature and thus pose a threat.  To determine whether the laboratory selection process mimicked natural selection, the authors looked at past HA sequences to see whether mutations that were deemed deleterious in the laboratory ever persisted in nature.  The answer was largely “no”.  The laboratory selection accurately predicted mutations that were unlikely to persist and dominate in human outbreaks.  Conversely, mutations that persist in the laboratory also tend to be found more frequently in natural isolates.

The study by Lee and colleagues represents the first comprehensive, functional analysis of the evolvability of each residue in the influenza HA protein.  The results suggest that laboratory determination of fitness can provide accurate information about the future evolutionary trajectory of influenza outbreaks giving the WHO important new tools in the annual vaccine selection process.


Research was supported by the National Institutes of Health, Howard Hughes Medical Institute, Simons Foundation, Burroughs Welcome Foundation and Pew Charitable Trusts.

Lee J.M., Huddleston J., Doud M.B., Hooper K.A., Wu N.C., Bedford T., Bloom J.D. 2018. Deep mutational scanning of hemagglutinin helps predict evolutionary fates of human H3N2 influenza variants. Proceedings of the National Academy of Sciences. 28;115(35):E8276-E8285.