Antigenic mapping of the hemagglutinin of the H9 subtype influenza A viruses using sera from Japanese quail (Coturnix c. japonica)

ABSTRACT Influenza A viruses (FLUAV) of the H9N2 subtype are zoonotic pathogens that cause significant economic damage to the poultry industry. Vaccination to prevent and control H9N2 infections in poultry is widely employed in the Middle East and Asia. We used phylogenetics and antigenic analysis to study the antigenic properties of the H9 hemagglutinin (HA) using sera produced in Japanese quail (Coturnix c. japonica). Consensus HA1 sequences were generated to capture antigenic diversity among isolates. We constructed chimeric H9N2 viruses containing the HA1 of each consensus sequence on a constant isogenic backbone. The resulting viruses were used to generate antisera from quail, a common and significant minor poultry species whose anti-HA response profiles remain poorly defined. Antigenic maps were generated by plotting the cross-hemagglutination inhibition (HI) data from the panel of quail sera against the chimeric constructs and 51 H9 field isolates. The chimeric antigens were divided into four different antigenic profiles (cyan, blue, orange, and red). Site-directed mutagenesis analysis showed nine amino acid positions of antigenic relevance. Substitutions at amino acid positions 149, 150, and 180 (H9 HA numbering) had relatively significant impact on HI activity using quail sera. Substitutions E180A and R131K/E180A led to the most significant antigenic change transitions. This study provides insights into the antigenic profile of H9 FLUAVs, with important implications for understanding antigenic drift and improving vaccine development for use in minor poultry species. IMPORTANCE Determining the relevant amino acids involved in antigenic drift on the surface protein hemagglutinin (HA) is critical to understand influenza virus evolution and efficient assessment of vaccine strains relative to current circulating strains. We used antigenic cartography to generate an antigenic map of the H9 hemagglutinin (HA) using sera produced in one of the most relevant minor poultry species, Japanese quail. Key antigenic positions were identified and tested to confirm their impact on the antigenic profile. This work provides a better understanding of the antigenic diversity of the H9 HA as it relates to reactivity to quail sera and will facilitate a rational approach for selecting more efficacious vaccines against poultry-origin H9 influenza viruses in minor poultry species.

To prevent and control H9N2 virus infections in poultry, several countries in Asia and Middle East have resorted to vaccination programs (4)(5)(6)(7)(8)(9)(10).Antigenic drift of H9 FLUAVs is readily observed in the field, likely a combination of natural evolution and vaccine use (4)(5)(6)(7)(8)(9)(10).Near and around the receptor binding site, the globular head HA1 portion of the H9 HA contains two partially overlapping antigenic sites.These sites have been defined previously using mouse monoclonal antibodies (mAbs) and are known as sites I and II or, more recently, as sites H9-A and H9-B, respectively (11)(12)(13)(14)(15). Site H9-A is immunodomi nant compared to site H9-B (12,16).A limited set of the most prominent poultry-adapted Eurasian lineages from specific regions have been examined antigenically (11)(12)(13)(17)(18)(19).Most antigenic analyses of H9N2 viruses have been performed using chicken sera and, to a lesser extent, ferret sera, but not with sera from minor poultry species such as quail.Japanese quail have been suggested as key players in the genesis of influenza viruses with respiratory tract tropism (20,21).Quail show wide distribution in the respiratory tract of both avian-like (SAα2.3)and human-like (SAα2.6)sialic acid receptors, which may have contributed to the emergence of the poultry-adapted H9N2 strains with human-like receptor preference (22,23).Anti-H9 sera have been raised by different approaches and regimes, which act as confounding factors to assess antigenicity faithfully (16,(24)(25)(26)(27). Immunization approaches have included either live virus challenge or most typically inactivated/adjuvanted viruses in either single or prime and boost infection or vaccina tion.Despite the absence of a standardized approach for sera production, these analyses have shown some significant clues about the antigenic makeup of the H9 HA.Combined with studies using mouse mAbs, a cluster of amino acids has been shown to affect the antigenic profile of the HA, namely those at positions 72, 74,121,131,135,150,180,183,195,198,216,217,249,264,276,288, and 306 (H9 numbering throughout the manuscript) (16,25,26,28,29).Further analyses on the contributions of each of these and alternative positions to antigenicity/receptor binding avidity are discussed later in the context of this report's findings.
To broaden the understanding of the antigenic diversity of HAs of H9 FLUAVs, we included strains from the American and Eurasian lineages.Starting from an initial phylogenetic analysis of nucleotide sequences corresponding to the HA1 region of the HA, we identified 18 clades utilizing sequence information of strains from 1966 to 2020.Analyses of these clades led to the selection of 10 consensus sequences that largely embodied the amino acid diversity within each H9 lineage/sub-lineage/sub-sub-lineage.The 10 HA1 sequences were used to generate chimeric H9 HA gene segments carrying a constant HA2 portion derived from the prototypic strain gf/HK/WF10/1999 (H9N2) (WF10) (30,31).The chimeric HA constructs were subsequently used for reverse genetics.To better understand the H9 HA antigenic makeup in the context of neutralizing responses in minor poultry, Japanese quail were challenged with the chimeric H9 HA viruses.Anti-H9 quail sera were used to perform hemagglutination inhibition (HI) assay and antigenic cartography (14,32).These analyses showed H9 HA antigens positioned in four antigenic clusters in the antigenic map, with additional outliers.Viruses carrying amino acid substitutions at relevant antigenic positions were generated to explain cluster transitions.These results provide new insights into the antigenic evolution of H9N2 influenza viruses and offer new opportunities to improve vaccine development.

Phylogenetic analysis, consensus sequences, and antigenically relevant amino acids on H9 HA
Using the H9 HA1 region, a maximum likelihood phylogenetic tree was established based on nucleotide sequences from isolates between 1966 and 2016 and then updated with sequences up to 2020.The phylogenetic analysis allowed the identification of different clades (h9.1.1 to h9.4.2).Consensus sequences were generated for each clade, n = 10 (Fig. 1).The percentage of amino acid identity ranged from 83.1% (h9.2.3 vs h9.3.9) to 98.4% (h9.3.3 vs h9.3.4).The number of amino acid differences in the HA1 region between the consensus sequences and the HA of the prototypic h9.3), and 48 (h9.3.9),respectively (Fig. 2).Chimeric HA constructs were used for reverse genetics in the WF10 backbone.In addition to the wild-type (WT) WF10 strain, 8 out of the 10 chimeric HA constructs resulted in viable H9N2 viruses.No virus rescue was obtained for the chimeric HA representing the h9.2.3 and h9.2.4 clades.Analysis of the HA1 portion of the consensus viruses and the closest relative from a subset of field viruses showed high similarity (Fig. 2).For WF10, the closest relative was A/qa/HK/G1/97 (98.4%); for h9.4.

Growth kinetics in vitro and replication and transmission in quail of chimeric H9 viruses
Growth kinetics evaluation in vitro in Madin-Darby canine kidney (MDCK) cells showed that all chimeric H9 viruses (Fig. 3A) achieved titers from 6.7 to 7.7 log 10 TCID 50 eq/mL except the h9.1.1 consensus virus (5 log 10 TCID 50 eq/mL at 96 h).Replication and transmission of consensus chimeric of H9 viruses were evaluated (Fig. 3B) in groups of quail (nine groups, n = 6 per group) that were directly inoculated (DI) with the H9N2 chimeric viruses (one chimeric virus per group) or the WF10 wild-type virus.At 24 h post-inoculation, naive quail were introduced in direct contact (n = 6 per group, CONTACT) with the corresponding DI group.We chose Japanese quail (Coturnix c. japonica) as a relevant minor poultry host of H9 FLUAVs (20,30).Virus replication was monitored in tracheal swabs at different times post-inoculation/post-contact (Fig. 3B).The wild-type h.9.4.1 (WF10) virus and consensus chimeric viruses carrying a poultryadapted HA1 region (h9.4.3, h9.3.9, h9.3.7,h9.3.5, h9.3.4,and h9.3.3)replicated and transmitted in quail, although there were clear variations both in the number of quail with active virus replication and/or days in which quail were positive.Chimeric H9 viruses carrying prototypical wild bird-derived HA1 regions showed distinct patterns: the h9.2.2 virus carrying the HA1 from a wild bird Eurasian-origin consensus showed decreasing levels of replication in DI birds as the infection progressed and with four out of six CONTACT quail being virus positive only on the last day of testing (6 days post-contact).In contrast, the h9.1.1 wild bird American consensus virus showed replication below levels of detection and no evidence of transmission.

HI responses against consensus clade viruses in quail
The DI quail (nine groups, n = 6 per group) were used to generate antisera against the chimeric HA consensus viruses (Fig. 4A).At 14 days post-inoculation (14 dpi), quail were boosted subcutaneously with inactivated/adjuvanted preparations of each virus.At 28 dpi, quail were terminally bled, and two independent pooled sera were generated (three birds per pool).We analyzed the seroconversion to the homologous virus in inoculated quail after prime and boost by HI assays.HI titers between 40 and 5,120 against the homologous viruses were observed (Table 1).The highest homologous HI titers were obtained for h9.3.3 and h9.3.9, with a titer of 5,120 in each case.HI titers of 2,560-5,120 were observed for h9.3.4,while a titer of 2,560 was obtained for h9.4.2.For the h9.3.7 and WF10 viruses, HI titers of 1,280-2,560 were observed.The h9.1.1 and h9.2.2 groups were the exception, with HI titers of 80-160 and 40-160, respectively, which are considerably lower than the other consensus viruses and consistent with poor virus replication during prime.Taken together, the homologous HI data show high levels of neutralizing antibodies against the different consensus viruses, except h9.1.1 and h9.2.2, which elicit poor antibody responses in the quail model.

Antigenic analysis of H9 HA
Using the antigenic cartography platform, the cross-HI data were merged and visualized by generating maps in which the spheres represent antigens and the squares represent the sera, distributed into space.Antigenic distances between antigens in the map are expressed in antigenic units (AU; one AU corresponds to a twofold dilution of antiserum in the HI assay).Dimensional analysis of the HI data set led to lower error yield in the 3D maps, though 2D maps were selected for better visualization, given that the relation ship between consensus antigens remained unvaried.The antigens were grouped into four different clusters (Fig. 4A).We used three AU or a ≥8-fold loss in cross-reactivity, as defined for the human seasonal vaccine strain update (WHO recommendation), as the threshold of significant antigenic difference.The WT WF10 HA prototypic h9.4.1 antigen (cyan) was 3.4 AU from the h9.4.2 antigen (blue).The h9.3.3, h9.3.4,h9.3.5, and h9.3.7 antigens (blue) clustered antigenically very close to each other (<0.3AU) and with 1.3, 1.6, 1.3, and 1.4 AU from the h9.4.2 blue antigen, respectively.The h9.3.9 antigen (orange) was 4.5 AU from the h9.3.7 consensus (blue), the closest phylogenetic relative, and 5.1 AU from the h9.4.2 blue antigen.The distance between WT WF10 HA prototypic h9.4.1 antigen (cyan) and the h9.3.9 antigen (orange) was 4.1.The h9.1.1 and h9.2.2 consensus antigens (red) showed relatively close antigenic relationships (2.9 AU from each other), but distances between h9.1.1 and WF10 (cyan), h9.4.2 (blue), and h9.3.9 (orange) antigens were 4.0, 5.3, and 8.1, respectively.It must be noted that the robustness of positioning of h9.1.1 and h9.2.2 must be interpreted cautiously due to the relatively low inherent antigenicity/immunogenicity compared to the rest of the consensus antigens.
We expanded these analyses to 48 additional field strains (Fig. 4C), bringing the panel to 51 field strains (Table S2).The analysis of other consensus viruses and the antigenic distances from their closest relatives (Fig. 2) revealed similarities between genetic and antigenic properties except for h9.3.3 and h9.3.9 and their respective closest relatives (Table 2).Distances between h9.3.3 and A/ck/Sichuan/5/1997 were 5.7 AU, while distances between h9.3.9 and A/dk/Hunan/1/2066 were 4.9 AU placing consen sus viruses and closest relatives in different clusters.The remaining consensus viruses showed good correlation with their closest relatives with distances between h9.1.1-A/rt/NewJersey/AI11-1946/2011, h9.2.2-A/ma/Li13384/2010, h9.3.5-A/ck/HK/SF3/99,h9.3.7-A/dk/Hunan/1/2006, and h9.4.2-A/ck/Pakistan/47/2003 of 0.9 AU, 2.4 AU, 0.5 AU, 1.1 AU, and 0.8 AU, respectively.From the 51 field isolates evaluated (Fig. 4C), 11 fell Except for the orange antigenic h9.3.9 antigen, all other antigens that showed sera reactivity but did not fall into an antigenic cluster are shown in gray as outliers.Specific viruses are denoted by codes shown in Table 2.   within the red cluster, 26 within the blue cluster, 4 within the cyan cluster, and none in the orange cluster (Table 2).Due to the low reactivity of the red-cluster consensus viruses (h9.1.1 and h9.2.2) in HI assays compared to the rest of the consensus antigens, field isolates of the red cluster were removed from the map (Fig. 4C).The h9.3.9 antigen was antigenically distinct from the rest of the h9.were classified as outliers as they were >3.0 in AU distance from any of the consen sus antigens (gray; Fig. 4; Table 2).H9s with <40 HI titers against any of the antisera were considered to have low to no cross-reactivity against any of the antisera and were removed from the antigenic analysis (Table 2).We observed mismatching between phylogenetic and antigenic analyses among viruses within the h9.3 and h9.4 lineages, mostly poultry isolates.Both h9.3 and h9.4 phylogenetic lineages contained the most antigenically variable strains, which fell under different clusters (and some were outliers).
The A/qa/UAE/302/2001 (b18, Fig. 4C) HA antigen was equally distant from h9.4.2 and WF10 antigens with 2.1 AU of distance in both cases (Table 2).Taken together, the results provide an antigenic map of the H9 HA using consensus and wild-type HA sequences probed with quail sera.

DISCUSSION
The HA plays a pivotal role in the antigenicity of FLUAV as it is the major target of neutralizing antibodies and subject to positive selection.Phylogenetics combined with antigenic analysis is the basis for human, avian, and equine influenza vaccine selection (34).Antigenic cartography facilitates the understanding of FLUAV antigenic drift by visualizing HI data as a spatial relationship between antigens in a map (35)(36)(37).We captured the antigenic diversity of dissimilar H9 viruses, underscored by the ability of Resolution of these issues requires further studies beyond the scope of this report.Nevertheless, the chimeric consensus approach was powerful enough to dissect the contribution of HA1 region of H9 in antigenicity in the context of an isogenic HA2 region.For antigenic characterization, boost immunizations with inactivated-whole virus adjuvant formulations were performed in quail at 14 dpi to increase HI titer levels of poorly immunogenic antigens.Quail antibody responses to H9 FLUAV mimicked what was previously reported in the literature for chicken sera (12,16).The synthetic consensus viruses aligned antigenically with representative H9 prototype field strains, supporting the notion that the globular head in the HA is the major target of the neutralizing antibody response.WF10 h9.4.1 and A/qa/HK/G1/97 (G1 prototype strain), which are phylogenetically related, showed also antigenic similarity (cyan cluster).These two antigens clustered separately from h9.4.2 (blue cluster), which showed strong cross-reactivity with most poultry isolates from the Middle East and Asia (Table 2).Similarly, h9.3.3 and h9.3.4 consensus antigens demonstrated strong cross-reactivity with their respective prototype lineages A/ck/HK/G9/1997 (G9, h9.3.3-like) and A/dk/HK/ Y280/1997 (Y280, h9.3.4-like).The strong HI cross-reactivity of the H9 field isolates against the heterologous clade-specific consensus antisera also supported the antigenic map results.Interestingly, consensus clades h9.3.3 through h9.3.7 and h9.4.2 showed similar antigenic phenotypes despite their genetic differences.The h9.3.9 (orange cluster) antigenic properties differed significantly from the rest of the h9.3 consensus viruses, with the highest reaction against its homologous sera (HI titer: 5,120) and marginal cross-reactivity with heterologous sera.Strikingly, the percentage of identity between the h9.3.9 consensus HA and the closest relative (A/dk/Hunan/1/2006) was 93.3%, being the lowest observed among the different clades, perhaps exposing a gap in sequence availability from the online databases.Nonetheless, sequence comparison between h9.3.9 and A/dk/Hunan/1/2006 revealed differences in key positions such as G72E, R146Q, N149G, N183D, and M217Q (Fig. 6 and 8) (11,12,15,16,33) which may account for the antigenic differences despite the close phylogenetic relationships.Few other H9 field isolates fell outside the three AU radius from any consensus antigen, despite the intermediate level of reactivity against the antisera panel.These observa tions highlight the significant impact of a few amino acid changes in modulating HI activity (35,(38)(39)(40) and reiterate the importance of antigenic cartography in correcting phylogenetic predictions.
Levels of HI titers were consistent with the virus' ability to replicate in quail.Perhaps, not surprisingly, the poultry-adapted chimeric viruses and particularly the wild-type WF10 virus were the most efficient in replication and transmission in quail.In contrast, limited replication and transmission were observed with the h9.2.2 wild bird Eurasianorigin chimeric virus, whereas no replication was observed with the h9.1.1 wild bird American-origin virus.It can be speculated that the replication and transmission of the recombinant consensus viruses were affected by the chimeric nature of the HA segment.However, these observations are in agreement with previous observations on the limited capacity for replication and transmission in quail of non-poultry-adapted wild bird-origin FLUAVs (20,21).
The impact of single or double amino acid substitutions was less clear (Fig. 7  and 8).The E180A-h9.4.2 single mutant (Fig. 7C) and the R131K/E180A-h9.4.2 double mutant (Fig. 7E) showed the strongest effect, with antigens positioning at <3.0 AU from the cyan and blue antigenic cluster.These observations suggest a role for position 180 since the R131K single mutant had minimal effect on HI activity compared to the WF10 h9.4.1 HA (Fig. 7A).Consistent with these observations, a previous report using the strain A/chicken/Shanghai/F/98 (H9N2) determined position 180 as directly responsible for antigenic drift (29).Variability at position 180 was also reported in field isolates from Morocco between 2018 and 2019, reinforcing a preponderant role of position 180 in evading pre-existing immunity (45).Consistent with the Morocco study, molecular characterization of H9N2 viruses from local markets in southern China also revealed a potential role of position 180 (and other positions) on antigenic properties (27).Spatiotemporal dynamics analysis from live-poultry markets in China has shown selection pressure in positions 146, 150, and 180 (46).A role of position 180 has been suggested also for the cross-species barrier where the 180V mutation favors the replication of H9N2 in mice (47).Other studies have attributed antigenic modulation to several HA residues without including position 180 (26,48).The latter is consistent with the idea that additional positions within the HA can modulate the antigenic properties, which is consistent with the findings in this report where eight or nine substitutions were introduced (Fig. 5 and 6).A previous report also described the role of position 217 in H9 antigenicity.However, in the global scale analysis, position 217 alone is insufficient for an antigenic cluster transition suggesting modest effects on antigenicity (28).Position 183 was also recently suggested as a modulator of the antigenic properties and overall replication of H9N2 viruses (49).This is consistent with the results observed between WF10 h9.4.1 (cyan) and h9.4.2 (blue).Antigenically relevant positions such as 180 and 217 have also been shown to affect receptor binding avidity (28,50,51), as it has position 216 (3,23,52).
Despite the remarkable plasticity of the H9 HA of WF10, reversions were observed in 5 out of 19 mutants, suggesting that tolerability of changes in antigenically relevant amino acids may be context dependent and likely encompass compensatory substitu tions (39,53).In addition, we identified a set of non-cross-reactive strains (Table 2; Fig. 4C) whose initial sequence information would predict to fall in at least one of the antigenic clusters described.These strains included A/dk/HK/448/1978, A/qa/Saudi A/489_46v08/2006, A/ck/NKorea/99020/99, and A/ma/Eng/7798_6499/2006.The strain A/ck/Tun/345/2011 with an HA1 region almost identical to A/ck/Tun/812/2012 in key amino acid signatures, failed to show cross-reactivity with members of the blue cluster, suggesting the involvement of other potentially relevant epitopes.
Most studies of H9N2 antigenicity in poultry involve the use of chicken sera but not sera from minor land-based poultry species, such as quail.Japanese quail have been suggested as key players in the genesis of influenza viruses with respiratory tract tropism (20,21).Quail are also more susceptible to H9N2 infection than chickens (30).In addition, quail show wide distribution in the respiratory tract of avian-like (SAα2.3)and human-like (SAα2.6)sialic acid receptors, which may have contributed to the emergence of the current poultry-adapted H9N2 strains with human-like receptor preference (54).Thus, quail might have played a role as an intermediate host between wild aquatic birds and poultry in the emergence of H9N2 strains with altered host range (22,23).The antigenic analyses using quail antisera provide significant insights into anti-HA responses in a relevant poultry species for influenza replication and evolution.The current literature shows different approaches employed for the antisera generation, including live virus inoculation and inactivated/adjuvanted virus vaccination to study antigenicity of the HA of influenza viruses.Still, none have used quail sera as a model (16,(24)(25)(26)(27).The results validate using the quail model to study the antigenicity of H9N2 as well as other viral properties such as virus replication, pathogenesis, and transmission.Although the results provide novel insights into the antigenic properties of FLUAV of the H9 subtype on a global scale, some limitations must be noted.The initial phylogenetic analysis for generating the consensus sequences was performed in 2016.As H9N2 viruses continue to evolve with inherent animal and public health risks, further studies are needed to better dissect the role of amino acid substitutions on the HA that modulate host range, replication, pathogenesis, transmission, and antigenicity.
In conclusion, phylogenetics was used to generate consensus on H9 viruses encompassing their natural diversity.We demonstrated that these consensus H9 viruses were biologically active, capable of triggering an immune response associated with the generation of neutralizing antibodies, and manifested important distinctive biologic characteristics driven only by their differences in the HA1 domains.Using this system, we explored antigenicity and modulation of HI profiles using antisera obtained from quail.The sera obtained allowed us to narrow down antigenically relevant amino acids, as many as nine for h9.4.2 (at positions 72, 131, 135, 180, 186, 188, 198, and 217) and six for h9.3.9 (127, 131, 180, 182, 183, and 217) to as few as one (E180A), to produce antigenic cluster transitions.The results are relevant to pave the way for a better understanding of the molecular signatures of antigenicity in H9 viruses, facilitating a rational approach for selecting more efficacious vaccines against poultry-origin H9 influenza viruses.

Database and phylogenetic analysis of HA sequences
H9 HA sequences were obtained from the Influenza Research Database, the Bacterial and Viral Bioinformatics Resource Center, and the Global Initiative on Sharing All Influenza Data (55,56).The initial phylogenetic analysis was performed on 984 global representative H9 avian isolates from 1966 to 18 March 2016 and was used to build the H9 consensus sequences presented in this study.The phylogenetic analysis was then updated on 14 July 2020, and included 1,316 manually curated sequences.The amino acid frequencies were analyzed using the protein sequence variant analysis tool provided by Scop3D (57).HA sequences were mapped to the A/gf/HK/WF10/1999 (WF10), GenBank accession #AY206676 (30), reference sequence using Geneious (version 10.2.3, Auckland, New Zealand).H9 HA1 sequences spanning the period from 1966 to 2020 were manually pruned to remove truncated and or repetitive sequences.An amino acid alignment was generated using default settings in MUSCLE (58).The numbering of HA corresponds to the mature H9 HA.All known key antigenic sites were considered in the phylogenetic algorithm using optimization with GARLI (59).A maximum likelihood tree was inferred using RAxML v.8.1.24(60) with a general time-reversible substitution model with gamma-distributed rate variation among sites, followed by GARLI for branch optimization.A starting tree was generated using parsimony methods with the bestscoring tree, and statistical support was obtained using the rapid bootstrap algorithm.Initially, 18 consensus sequences were produced, representative of genetic variations within phylogenetic groups.Of these, 10 consensus sequences were selected (Fig. 1).

Antigenic characterization
Standard HA and HI assays were performed as previously described (66).Before HI testing, sera were heat inactivated at 56°C for 30 min and adsorbed with 50% chicken red blood cells (RBCs) to remove nonspecific inhibitors of hemagglutination.Sterile PBS was added, allowing the sera to reach a final dilution of 1:10.Then, sera were transferred to 96-well plates and serially diluted twofold in 25 µL of sterile PBS and mixed with four HAU/25 µL of each virus.The virus-sera mixture was incubated for 15 min at room temperature and then added 50 µL per well of 0.5% chicken RBCs (100 µL final volume per well).The HI activity was determined after 45 min of incubation.

Antigenic cartography
The HI data using quail sera (Table S2) were analyzed separately and merged through the Antigenic Cartography Macros (ACMACS) website (https://acmacs-web.antigenic-car tography.org)as previously described (67,68).HI data sets were subject to a dimensional analysis in all dimensions (2D, 3D, 4D, and 5D) with 2,000 optimizations and an auto matic minimum column basis parameter to identify which model best fits this data set.Antigens that exploited no to low (<40) reactivity against the entire antisera panel were removed from the analysis and annotated.The distance between the spheres (antigens) and antisera (squares) is inversely correlated to the log 2 titer measured by the HI assay.One antigenic unit is the equivalent of a twofold loss/gain in HI activity.Clusters were initially established by applying the Ward method of hierarchical clustering.Within these, reference antigens were selected based on their biological significance, and clusters were adjusted to enclose antigens exclusively within a three AU radius from these selected reference antigens.We used three AU or a ≥8-fold loss in cross-reactivity, as defined by the WHO recommendation to update human seasonal vaccine strains, as the threshold of significant antigenic difference.

Site direct mutagenesis
The site-directed mutagenesis kit (Thermo Fisher, Waltham, MA) generated single and double amino acid substitutions in the WF10 HA gene segment following manufacturer conditions.Plasmid sequences were confirmed by Sanger sequencing.

FIG 1
FIG 1 Global phylogenetic analysis of H9N2 FLUAV.Maximum likelihood phylogeny of 1,316 H9 avian HA1 nucleotide sequences from the database of Global Initiative on Sharing All Influenza Data and Influenza Research Database updated 14 July 2020 generated with RaxML followed by GARLI branch length optimization.Nodes at the end of each branch are color-coded based on the geographic origin of each isolate.Amino acid substitutions using one-letter code and numbering based on H9 HA mature sequence are shown.Highlighted in black are H9 sub-lineages chosen to generate consensus HA1 region sequences and to produce chimeric H9 HA constructs with a constant HA2 region.Sub-lineages that were unsuccessful in reverse genetics are shown in gray.The h9.4.1 consensus is represented by the prototypic virus A/gf/HK/WF10/1999 (H9N2).

FIG 4
FIG 4 Antigenic maps using quail sera against H9 viruses.(A) Antigenic map with spheres representing consensus viruses and squares representing the different antisera.Viruses are highlighted and colored by respective clusters (cyan, red, blue, and orange).Antigenic unit (AU) distances between representative antigens from each cluster are shown next to dashed gray lines connecting them.(B) Antigenic map with spheres representing consensus viruses + prototypical strains (QA/HK/G1/07, DK/HK/Y280/97, and CK/HK/G9/97) and squares representing the different antisera.Viruses are highlighted and colored by respective clusters (cyan, red, blue, and orange).AU distances between representative antigens from each cluster and prototypic strains are shown next to dashed gray lines connecting them.(C) Antigenic map with spheres representing consensus viruses + field isolates (n = 46) and squares representing the different antisera.Viruses are highlighted and colored by respective clusters (cyan, red, blue, and orange).AU distances between representative antigens from each cluster are shown next to dashed gray lines connecting them.Except for the orange antigenic h9.3.9 antigen, all other antigens that showed sera reactivity but did not fall into an

TABLE 2
Cluster location and antigenic distances of consensus viruses and field isolates (Continued on next page)

TABLE 2
Cluster location and antigenic distances of consensus viruses and field isolates (Continued) Determined by antigenic analysis through Antigenic Cartography Macros (ACMACS).bOne unit of antigenic distance is equal to a twofold difference in the HI assay.All virus strains are of the H9N2 subtype except where noted.Viruses with Black label were non-cross-reactive against the panel of antisera.Dashes were used when no closest centroid was assigned.Animal species acronyms dl, duck; ck, chicken; ty, turkey; ph, pheasant; qa, quail; mal, mallard; rt, ruddy turnstone; sh, shorebird; av, avian.

TABLE 3
Antigenic distances between consensus viruses and different field isolates (Continued on next page)

TABLE 3
Antigenic distances between consensus viruses and different field isolates (Continued)

TABLE 4
Summary of amino acid substitutions for each mutant and antigenic distances.Dashes denote viruses that were not generated.Cyan synthetic consensus viruses to induce HI responses that recognize genetically related antigens from the field.We successfully obtained 8 out of 10 chimeric H9 viruses.Despite multiple attempts, representative chimeric viruses from the clades h9.2.3 and h9.2.4 were not viable.As the HA evolves in concert with other viral gene segments, it is possible that we encountered incompatibility issues within the HA segment itself or altered interactions with other viral proteins.

TABLE 5
Antigenic distances between consensus viruses and the different mutants