Using a comparative One Health approach to investigate the structural basis of antigenic variation among human and avian influenza viruses

Human influenza viruses are estimated to cause 3-5 million cases of severe illness globally each year resulting in 250,000 to 500,000 deaths. Avian influenza viruses (AIVs) represent a threat to global poultry production and to human health through both zoonotic infection and the pandemic potential of reassortant viruses. The WHO's Global Influenza Surveillance and Response System has conducted virological surveillance for over half a century, is responsible for genetic and antigenic characterisation of circulating, and supports the selection of influenza viruses for human influenza vaccine production. Vaccination of poultry has become one of the principal practices for control of the endemic disease in many countries and human vaccination would be a vital component of the response in a pandemic situation.

The effectiveness of vaccines that exist to protect against human and avian influenza viruses are threatened by the emergence of antigenic variants. Suboptimal vaccine matching due to antigenic drift can result in vaccine failure burdening human and animal health, causing significant economic losses, and threatening food security. Influenza vaccine efficacy is highly dependent on antigenic matching of vaccine seed to circulating strains, and influenza strains are characterized by antigenic drift over time, structural changes in B-cell epitopes facilitate escape from pre-existing immunity. There is, therefore, a pressing need to better understand the molecular basis of antigenicity of the major influenza antigen, haemagglutinin. This is particularly true for AIVs which typically have received less attention and consequently about which less is known. Meanwhile, extensive resources have been invested in the antigenic characterisation of human seasonal influenza viruses, in particular influenza A H1N1 and H3N2, to guide decisions on human seasonal influenza vaccine composition. The extensive knowledge and data from human viruses can inform more rigorous and biologically informed models to apply to AIVs where there is an opportunity for significant advances to be made using a comparative One Health approach.

A biophysically-informed structural model of the immunological reactivity and receptor binding avidity of these human and avian pathogens based on the structures and amino-acid sequences of their antigenic proteins will be built. This will be used to identify the underlying factors that affect virus cross-reactivity, and hence vaccine efficacy. This model will extend previous work through greater integration of information on 3-D protein structures and the biophysics determining antibody-antigen interactions. The extensive data already generated for human influenza A viruses will be collated and used to refine this model. This refined model will then be applied to AIV, transferring knowledge from the medical field to the veterinary field. Laboratory work will be used to experimentally validate the structural model of HA phenotypic variation. Studies of AIVs will offer a provide a better understanding of the influence of changes in receptor-binding avidity on cross-reactivity, which can in turn benefit the medical field, since this has become very difficult to measure for human influenza A H3N2 viruses.

We will then use these results to predict cross-reactivity for newly emerging viruses and thereby the likelihood of vaccine escape, and to propose vaccine seed strains most likely to be effective for control of novel AIV threats benefitting veterinary vaccinology. The improved knowledge of the structural basis of variation in HA phenotype will also benefit existing methods used to predict evolutionary trajectories of influenza viruses, aiding decision makers concerned with the selection of human influenza viruses. Additionally, the biophysical model will allow predictions of previously unobserved changes to viral proteins to be made.

Biophysical model: A Bayesian model of the genotype-antigenic phenotype relationship will extend previous models by (a) incorporating additional parameters to better describe virus-antibody and receptor binding; (b) accounting for 3-D haemagglutinin (HA) structure allowing spatial inference of the extent of epitopes; and (c) developing a matrix describing the general antigenic distance of amino acid pairs allowing inference across the HA structure.

Human influenza model refinement: The biophysical model will be applied to human influenza A H1N1 and H3N2 data consisting largely of titres haemagglutinin inhibition (HI) and virus neutralisation assays and genetic sequence data for HA1. This process will be used to test and refine the model, analysing existing HA structures to determine spatial relationships between residues within epitopes and binding sites, incorporating existing avidity analyses to parameterize the biophysical model, and tuning the substitution model using antigenic and sequence data.

Avian influenza data and modelling: Antigenic and genetic data from avian influenza subtypes will be collated and modelled. The human influenza virus model will now be refined on the avian data. The phenotypic impact of amino acid substitutions in the AIV HA will be characterised and contribution to vaccine escape will be assessed. The phenotypic impact of previously unobserved amino acid substitutions will also be predicted.

Model validation: Mutant recombinant viruses will be generated using site-directed mutagenesis and reverse genetics using available HA backbones. Recombinant viruses will be assessed using the HI assay performed with polyclonal antiserum and monoclonal antibodies. Changes in HI titre, relative to the backbone, will be attributed to changes in antigenicity and avidity. Changes in receptor binding avidity will be directly assessed using the receptor-binding destroying enzyme assay and surface biolayer interferometry.