Antigenic evolution of H3N2 influenza A viruses in swine in the United States from 2012 to 2016

Abstract Background Six amino acid positions (145, 155, 156, 158, 159, and 189, referred to as the antigenic motif; H3 numbering) in the globular head region of hemagglutinin (HA1 domain) play an important role in defining the antigenic phenotype of swine Clade IV (C‐IV) H3N2 IAV, containing an H3 from a late 1990s human‐to‐swine introduction. We hypothesized that antigenicity of a swine C‐IV H3 virus could be inferred based upon the antigenic motif if it matched a previously characterized antigen with the same motif. An increasing number of C‐IV H3 genes encoding antigenic motifs that had not been previously characterized were observed in the U.S. pig population between 2012 and 2016. Objectives A broad panel of contemporary H3 viruses with uncharacterized antigenic motifs was selected across multiple clades within C‐IV to assess the impact of HA1 genetic diversity on the antigenic phenotype. Methods Hemagglutination inhibition (HI) assays were performed with isolates selected based on antigenic motif, tested against a panel of swine antisera, and visualized by antigenic cartography. Results A previously uncharacterized motif with low but sustained circulation in the swine population demonstrated a distinct phenotype from those previously characterized. Antigenic variation increased for viruses with similar antigenic motifs, likely due to amino acid substitutions outside the motif. Conclusions Although antigenic motifs were largely associated with antigenic distances, substantial diversity among co‐circulating viruses poses a significant challenge for effective vaccine development. Continued surveillance and antigenic characterization of circulating strains is critical for improving vaccine efforts to control C‐IV H3 IAV in U.S. swine.


| INTRODUC TI ON
Influenza A virus (IAV) is an important respiratory pathogen of both humans and swine. Vaccination is the main strategy employed to control the morbidity and mortality associated with IAV illnesses in both hosts. Although vaccine platforms, formulations, and strain selection processes differ between host species, all widely utilized IAV vaccines primarily target the hemagglutinin (HA) protein. While IAV vaccine strain selection for seasonal human vaccines is a globally coordinated event led by the World Health Organization with strains publicly announced twice per year, 1 veterinary vaccine companies are not required to report virus strain names on vaccine labels or product inserts included in commercially available vaccines. Fully licensed vaccines for commercial use in the United States include two killed multivalent vaccines, an alphavirus-vectored vaccine, and a newly licensed live-attenuated influenza A virus vaccine (LAIV). [2][3][4] Furthermore, swine producers may use farm or company-specific custom autogenous vaccines to protect their swine herds from IAV in swine (IAV-S). 5,6 These factors limit the ability to evaluate vaccine strain matching and efficacy from a national perspective.  9 It is unclear if the antigenic evolution of these recent human-like HA's will follow similar patterns as the C-IV H3 IAV-S or if this genetically and antigenically distinct human-like H3 HA will displace the endemic C-IV H3 HA in the United States, but at present time, the two H3 IAV-S clades cocirculate. Previous studies in search of molecular determinants responsible for antigenic drift in human H3 IAV identified seven amino acid positions (145, 155, 156, 158, 159, 189, and 193; H3 numbering throughout) in the HA protein that largely determined the antigenic phenotype. 10,11 A similar study with H3 IAV-S found that six of the seven positions (145, 155, 156, 158, 159, and 189) implicated in human IAV antigenic evolution were also important for the antigenic phenotype of C-IV H3 IAV-S, and phenotypic differences were observed among co-circulating swine IAV. 12 Analysis by antigenic cartography revealed distinct antigenic groupings of viruses, termed antigenic clusters, and labeled by different colors for visualization.
The antigenic clusters were associated with specific combinations of amino acids at the six positions, and, thus, the combinations of these key positions were referred to as an "antigenic motif." For example, the Cyan antigenic cluster was comprised of viruses encoding at least four antigenic motifs (NHNNYR/NHNDYR/NNNDYR/ NHSYR), and the Red antigenic cluster was comprised of at least four antigenic motifs (NYNNYK/NYHNYK/NYNNHK/NHNNYK).
Site-directed mutagenesis at these six amino acid positions in a prototype C-IV H3 IAV-S strain confirmed that these six positions played a key role in defining the antigenic phenotype. 13 Trends in antigenic motif patterns over time revealed a predominance of viruses encoding Cyan antigenic cluster motifs in 2009, followed by a steady decline. The emergence of viruses encoding Red antigenic cluster motifs was observed in 2010, followed by sustained circulation through 2016, and the emergence of viruses encoding Green antigenic motifs in 2013 (previously "light green" 13 ). However, the previous study reported that 23% of virus isolates collected from 2009 to 2015 encoded antigenic motifs that had yet to be antigenically characterized. These uncharacterized H3 IAV-S likely represented additional antigenic diversity.
In this study, we selected contemporary C-IV H3 IAV-S isolates in the US based on the observed expanding genetic diversity of the HA gene and an increase in antigenic motif patterns. The antigenic phenotype of 50 C-IV H3 IAV-S collected between 2012 and 2016 was characterized using hemagglutination inhibition (HI) assay data generated with swine antisera and visualized with antigenic cartography. The selected viruses contained uncharacterized motifs to determine the impact of amino acid substitutions at the six key sites as well as viruses with previously characterized antigenic motifs to validate previous observations. A maximum-likelihood phylogeny was inferred from the protein alignment using FastTree (v2.1) with default settings with a JTT+CAT model of molecular evolution. 17 Each HA protein sequence was assigned to one of six clades within C-IV (clades A-F) following Kitikoon et al, 18 or to the recently emerged human-like clade. 9 C-IV H3N2 IAV-S (n = 1007) were grouped by antigenic motif and at least three strains were selected for analysis from uncharacterized antigenic motif groups. For each selected motif, a strain was chosen for both high and low similarity to the motif group consensus HA sequence. A third strain was selected to assess common substitution patterns within the given antigenic motif group if present.

| Sequence analyses
Additional strains were selected with HAs that encoded less frequently detected antigenic motifs. A total of 50 C-IV H3N2 IAV-S were selected as antigens for hemagglutination inhibition (HI) assays and/or antiserum production (Table S1).

| Antiserum production
Swine antiserum was produced by immunizing two pigs as previously described. 12 For use in HI assays, sera were incubated at 37°C overnight with receptor-destroying enzyme (RDE(II); Denka Seiken, Tokyo, Japan). After the addition of 0.85% saline (w/v) the following morning, sera were incubated at 56°C for 45 min to deactivate the RDE, followed by adsorption with 50% turkey red blood cells at 4°C to remove any additional nonspecific inhibitors of HA.

| Hemagglutination inhibition assays
Standard HI assays were performed with turkey red blood cells, and fold reduction values for endpoint titers were calculated by dividing the homologous geometric mean titer (GMT) for each pair of sera by the heterologous GMT of each test antigen.

| Antigenic cartography
HI data generated in this study were merged with a subset of H3 IAV-S HI data generated previously by Lewis et al using the same methods described herein (Table S2). 12 Antigenic relationships were visualized in multi-dimensional space using antigenic cartography. 12,19,20

| Contemporary C-IV IAV-S strains encoded antigenic motifs not previously characterized
Although there were 73 unique antigenic motifs detected, 90% of the HA sequences encoded one of the 20 most frequently detected motifs (904/1007) ( Table 1). The two most frequently detected motifs in H3 IAV-S over the five-year timespan were NYNNYK and KYNNYK, corresponding to the previously defined Red and Green antigenic clusters, respectively. 12

| Uncharacterized antigenic motifs represented additional antigenic diversity
Twenty-nine viruses encoding ten previously uncharacterized antigenic motifs from three distinct phylogenetic clades of C-IV H3 IAV-S were selected for characterization (Table S1). The HI results were used to generate an antigenic map (Figure 2A), and strains encoding the same archetypal motif generally clustered together in the antigenic map with one exception described below. F I G U R E 1 Temporal frequency of H3 antigenic clusters. (A) Temporal frequency of H3 antigenic clusters prior to this study. (B) Temporal frequency of H3 antigenic clusters following this study. Cluster designations and coloring follow the color scheme used previously by Lewis et al. 12 Strains denoted "Other" encode outlier antigenic motifs of low prevalence. Strains encoding an antigenic motif not yet phenotypically characterized are denoted as "Uncharacterized", while "New" strains encode an antigenic motif characterized in this study

| Antigenic motif alone was not sufficient to explain intra-cluster diversity
To determine whether variation outside of the 6 amino acid motif positions contributed to intra-cluster drift, we selected strains encoding the most prevalent antigenic motifs corresponding to the Red and Green antigenic clusters with collection dates from 2012 to 2016. Eight viruses encoding an NYNNYK antigenic motif and one virus encoding an NHNNYK motif were chosen from the Red antigenic motifs (Table S1, Figure 3). Twelve viruses encoding a KYNNYK antigenic motif were similarly tested among viruses encoding Green antigenic motifs (Table S1, Figure 3). All newly characterized viruses mapped relatively near previously tested viruses encoding the same motif, but with demonstrable intra-cluster variation (up to 4.8 AU for Red and up to 6.3 AU for Green) ( Figure 3A). To study temporal intra-cluster drift, A/swine/Pennsylvania/ A01076777/2010 (PA/10) and A/swine/Illinois/A01327903/2012 (IL/12) were chosen as reference antigens because they were the earliest antigens characterized within the Red and Green antigenic clusters, respectively. The antigenic distances of each virus to its respective cluster predecessor were plotted to observe intra-cluster diversity over time ( Figure 3B). Virus strains within the Green antigenic cluster demonstrated greater intra-cluster variation and antigenic distance from the cluster representative strain over a four-year period than virus strains belonging to the Red antigenic cluster over a six-year period.

| Antigenic phenotype was not restricted to monophyletic clades
The number of substitutions in the HA1 region between each pair of antigens was plotted against the antigenic distance between each F I G U R E 2 Antigenic phenotype of strains encoding a previously uncharacterized motif. (A) Three-dimensional antigenic map of strains encoding a previously uncharacterized motif. Viruses encoding identical antigenic motifs are grouped (dotted circles) and labeled. Predominant antigenic clusters from 2009 to 2016, with cluster representative viruses denoted by an asterisk (*), are visualized for reference (the dominant antigenic motif is indicated for each colored phenotype). (B) Antigenic distance from Cyan (MN/09), Red (NY/11), and Green (IA/14) cluster representative strains. The 3 antigenic unit (AU) line denotes an 8-fold loss in HI cross-reactivity, the cutoff typically used in human H3 IAV antigenic studies to define significant antigenic drift pair ( Figure 5A). There was a positive linear association between the number of HA1 amino acid differences and antigenic distance (r = 0.51, P < 0.0001, Pearson's correlation); however, there was a large amount of unexplained variation, supporting the proposition that certain amino acids have a disproportionate impact on antigenic phenotype. We observed marked variability among pairs of antigens that differed by five or less sites in the HA1 region ( Figure 5B), often implicating amino acid positions of known importance (Table S3). The distribution of antigenic clusters among the C-IV clades A-F was annotated on a maximum-likelihood tree of H3 genes to assess whether genetic clade 12 was associated with antigenic phenotype ( Figure 5C). Putative antigenic clusters of viruses collected 2012-2016 were distributed widely across the C-IV clades, with no single antigenic phenotype populating a single given clade, providing further evidence that a small number of amino acid positions described by the antigenic motif disproportionately affect phenotype, and similarity at these positions is not restricted to monophyletic clades. In addition, these data demonstrate that a high number of antigenically diverse H3 strains are co-circulating in U.S. swine. F I G U R E 3 Antigenic evolution within antigenic clusters. (A) Three-dimensional antigenic map of strains encoding a Red or Green antigenic motif. Newly characterized viruses encoding KYNNYK (bright green) or NYNNYK (bright red), along with a previously characterized Green virus (pale green) and previously characterized Red (pale red) and Cyan (cyan) cluster viruses. (B) Intra-cluster antigenic distance from Red (PA/10) and Green (IL/12) cluster predecessors across the study time frame. One antigenic unit (AU) is equal to a twofold loss in crossreactivity F I G U R E 4 Cross-reactivity of swine sera raised against antigenic cluster representatives. Relative fold reduction in heterologous strains from antisera raised to MN/09 (Cyan), NY/11 (Red), and IA/14 (Green). A ≥ 8-fold reduction in the heterologous GMT from the homologous reaction is considered a significant loss in cross-reactivity by the sera. The dominant antigenic motif is indicated for colored phenotypes Contemporary viruses encoding Red and Green antigenic motifs were phenotypically similar to older strains encoding the same motif, but intra-cluster variation was observed within these two clusters.

| D ISCUSS I ON
Despite such intra-cluster antigenic variation within the Red and Green clusters between 2012 and 2016, we found that high titer antisera raised to early cluster representatives still cross-reacted with viruses encoding the same antigenic motif. Some antigenic clusters were maintained over the study time frame, particularly the Red antigenic cluster, predominant since 2009. 13 The sustained transmission of antigenically distinct viruses over the last seven years may be explained by a lack of long-lived population immunity in swine herds.
In addition to non-standardized strain composition in vaccines, other factors such as a relatively short generation time for pigs, shared facilities by pigs of different ages and immune status, and longdistance transport of swine via domestic routes and from Canada may all hinder the acquisition of effective population immunity to strains of a particular antigenic cluster. [21][22][23] In addition to the sustained circulation of H3N2 viruses in the swine population and introduction of new human H3N2 to swine, there is a continued risk of transmission of H3N2 from swine back to the human population, exemplified by the H3N2 variant (H3N2v) cases reported in recent years. 24 IAV-S will be replaced by the more recent human-like introduction, or if viruses from multiple genetic clades will co-circulate similar to what is seen with the multiple lineages and clades of H1 IAV-S. 27 At the current time, the two H3 lineages continue to co-circulate. H3 C-IV viruses remain a useful tool to study influenza antigenic drift due to antigenic heterogeneity, apparent plasticity at the antigenic sites, and availability of isolates through the USDA IAV-S repository.
Ultimately, a better understanding of antigenic evolution of influenza A viruses will help inform vaccine strain selection for more effective vaccines and identify swine strains that potentially pose higher risks when they spill back over to the human population.

ACK N OWLED G EM ENTS
We are grateful to the pork producers, swine veterinarians,