Date Published: April 8, 2016
Publisher: Public Library of Science
Author(s): William T. Harvey, Donald J. Benton, Victoria Gregory, James P. J. Hall, Rodney S. Daniels, Trevor Bedford, Daniel T. Haydon, Alan J. Hay, John W. McCauley, Richard Reeve, Scott E Hensley.
Determining phenotype from genetic data is a fundamental challenge. Identification of emerging antigenic variants among circulating influenza viruses is critical to the vaccine virus selection process, with vaccine effectiveness maximized when constituents are antigenically similar to circulating viruses. Hemagglutination inhibition (HI) assay data are commonly used to assess influenza antigenicity. Here, sequence and 3-D structural information of hemagglutinin (HA) glycoproteins were analyzed together with corresponding HI assay data for former seasonal influenza A(H1N1) virus isolates (1997–2009) and reference viruses. The models developed identify and quantify the impact of eighteen amino acid substitutions on the antigenicity of HA, two of which were responsible for major transitions in antigenic phenotype. We used reverse genetics to demonstrate the causal effect on antigenicity for a subset of these substitutions. Information on the impact of substitutions allowed us to predict antigenic phenotypes of emerging viruses directly from HA gene sequence data and accuracy was doubled by including all substitutions causing antigenic changes over a model incorporating only the substitutions with the largest impact. The ability to quantify the phenotypic impact of specific amino acid substitutions should help refine emerging techniques that predict the evolution of virus populations from one year to the next, leading to stronger theoretical foundations for selection of candidate vaccine viruses. These techniques have great potential to be extended to other antigenically variable pathogens.
Antigenic evolution of human influenza A viruses is characterized by rapid drift, with structural changes in antigenic epitopes allowing the virus to escape existing immunity. Consequently, seasonal influenza continues to impose a major burden on human health causing 250,000 to 500,000 deaths annually . Influenza vaccines, which remain the most effective means of disease prevention, currently comprise antigens from A(H1N1), A(H3N2) and B viruses predicted to elicit the most effective immune responses against circulating viruses in the forthcoming influenza season [1,2]. The continually evolving antigenic phenotype of influenza A viruses presents an ongoing challenge for vaccine virus selection, as effectiveness is greatest when vaccine components are antigenically similar to circulating viruses. The HA glycoprotein is the key antigenic determinant of influenza viruses  and consequently the most critical vaccine component. Neutralizing antibodies primarily bind to amino acids in protruding loops and helices that form defined antigenic sites [4,5], and substitutions that alter epitope structure can inhibit antibody binding and help the virus escape existing immunity . When the selective advantage conferred is sufficient, novel antigenic variants will replace circulating viruses. Hence, phylogenies of influenza A HA sequences are characterized by the presence of a single predominant trunk lineage, and short side branches, representing the rapid turnover of the influenza virus population [7,8].
We compiled HI assay data from former seasonal A(H1N1) viruses isolated between 1997 and 2009, comprising 19,905 individual measurements of cross-reactivity between 43 post-infection ferret antisera and 506 viruses. The HA1 sequences for all 506 viruses tested by HI in this dataset were used to generate a maximum clade credibility tree . Each of these 506 viruses was present in the HI dataset as test viruses, being tested against various antisera using the HI assay. Of these 506 viruses, 43 were also selected as reference viruses and used to raise antisera used in this HI dataset. HA1 trees for the complete set of 506 viruses and for the 43 reference strains only are shown with corresponding HI titers represented as a heat-map in Fig 1.
Using a modeling approach that integrated HA sequence data and HI antigenic data for over 500 viruses, we have identified substitutions responsible for the antigenic evolution of former seasonal influenza A(H1N1) viruses over a period of more than 10 years. We identified 18 substitutions at 15 amino acid positions in HA1: two that had high-impact on antigenicity (which individually can lead to a need to change vaccine virus), and 13 of lower impact, including some too low to be directly observable in routine HI assays. Substitutions at four of the fifteen amino acid positions occurred multiple times in the evolutionary history of the virus, consistently explaining observed antigenic changes. Antigenic cartography of the H1N1 viruses identified only three antigenic clusters, transitions between which can be explained by the two high-impact substitutions. Our more detailed findings could support our knowledge of the genetic basis of antigenic variation of the currently circulating A(H1N1)pdm09 viruses. Though these viruses have exhibited little antigenic variability so far, viruses with substitutions in the 153–157 region, including K153E which we specifically identify here (K154E in cited work), of the HA showing reduced reactivity to ferret antisera raised against the A/California/7/2009 vaccine virus have been noted [2,21]. Subsequent reverse genetics experiments have identified escape mutants possessing substitutions at position 153 including substitutions we detect here (K153E and K153G) .