Equine-Like H3 Avian Influenza Viruses in Wild Birds, Chile

Since their discovery in the United States in 1963, outbreaks of infection with equine influenza virus (H3N8) have been associated with serious respiratory disease in horses worldwide. Genomic analysis suggests that equine H3 viruses are of an avian lineage, likely originating in wild birds. Equine-like internal genes have been identified in avian influenza viruses isolated from wild birds in the Southern Cone of South America. However, an equine-like H3 hemagglutinin has not been identified. We isolated 6 distinct H3 viruses from wild birds in Chile that have hemagglutinin, nucleoprotein, nonstructural protein 1, and polymerase acidic genes with high nucleotide homology to the 1963 H3N8 equine influenza virus lineage. Despite the nucleotide similarity, viruses from Chile were antigenically more closely related to avian viruses and transmitted effectively in chickens, suggesting adaptation to the avian host. These studies provide the initial demonstration that equine-like H3 hemagglutinin continues to circulate in a wild bird reservoir.

Since their discovery in the United States in 1963, outbreaks of infection with equine influenza virus (H3N8) have been associated with serious respiratory disease in horses worldwide. Genomic analysis suggests that equine H3 viruses are of an avian lineage, likely originating in wild birds. Equine-like internal genes have been identified in avian influenza viruses isolated from wild birds in the Southern Cone of South America. However, an equine-like H3 hemagglutinin has not been identified. We isolated 6 distinct H3 viruses from wild birds in Chile that have hemagglutinin, nucleoprotein, nonstructural protein 1, and polymerase acidic genes with high nucleotide homology to the 1963 H3N8 equine influenza virus lineage. Despite the nucleotide similarity, viruses from Chile were antigenically more closely related to avian viruses and transmitted effectively in chickens, suggesting adaptation to the avian host. These studies provide the initial demonstration that equine-like H3 hemagglutinin continues to circulate in a wild bird reservoir.
across the central region of Chile. These samples were collected by using sterile flocked swabs (Copan Italia S.P.A., https://www.copangroup.com) and stored in 1-mL universal transport media tubes (Copan Italia S.P.A). They were then transported at 4°C to the Faculty of Veterinary Science of the University of Chile (Santiago, Chile) and stored at -80°C until analysis.
RNA extraction and quantitative reverse transcription PCR were performed at St. Jude Children's Research Hospital (Memphis, TN, USA) as described (13). In brief, RNA was extracted from 50 µL of swab specimen by using the Mag Max-96 IA/ND Viral RNA Isolation Kit (Life Technologies, https://www. thermofisher.com) on a Kingfisher Flex Magnetic Particle Processor (ThermoFisher Scientific, https:// www.thermofisher.com). Quantitative reverse transcription PCR was performed on a CFX96 Real-Time PCR System with the 4x TaqMan Fast Virus Master Mix (ThermoFisher Scientific) and primers and probes specific for the influenza A matrix gene (14). Samples with a cycle threshold value <38 where considered having positive results. Viral isolation was attempted on 9-day old specific pathogen-free (SPF) embryonated chicken eggs as described (15).
To identify the host species, genetic barcoding was performed by using primers that amplified the cytochrome oxidase I gene as described (16). Sequencing was performed by using either Sanger sequencing and universal oligonucleotide primer sets as described (17) or by deep sequencing on an Illumina MiSeq System (https://www.illumina.com) as described (18). Sequences were assembled by using CLC Genomic Workbench Version 9 (http:// webapp.cabgrid.res.in/biocomp/CLCBio/clc-bio. html). The H3 sequences from Chile used in this study have been deposited into GenBank (accession nos. KX101146, KY644162, MH675632, MH499154, MK163999, and MK164010).

Genetic and Phylogenetic Analysis
Phylogenetic analysis included sequences from avian and equine hosts downloaded from the National Center for Biotechnology Information (Bethesda, MD, USA) Influenza Virus Database. BLAST analysis (https://blast.ncbi.nlm.nih.gov/Blast.cgi) was performed to ensure that sequences from Chile not have higher homology to influenza A virus in other hosts. Representative sequences were selected by clustering similar sequences using CD-HIT-EST (19).
Maximum-likelihood trees were constructed by using the DECIPHER and PHANGORN packages in R (20,21). Sequences were aligned by using DECIPHER with the AlignTranslation function. The modelTest function was used to evaluate which nucleotide substitution with and without gamma-distributed rate variation among sites (γ) and invariant sites (I) is the best fit model. The generalized time reversible (GTR) + γ 4 + I model was best fitting by using Bayesian information criteria. The phylogenetic tree was constructed according to a standard protocol using phangorn (21). The starting neighborjoining tree was constructed by using the distance matrix, and maximum-likelihood trees were generated from the starting tree by using the GTR + γ 4 + I model with stochastic branch rearrangement. The bootstrap method was used to determine the confidence of the tree topology with 1,000 replicate trees. The maximum-likelihood tree was saved with the bootstrap percentage in the Newick format (https:// evolution.genetics.washington.edu) and visualized by using the ggtree package (22). Alternatively, the maximum-likelihood tree was constructed by using RaxML with a GTR + γ distribution with 1,000 bootstraps (23). The tree was visualized in FigTree version 1.4.4 (http://tree.bio.ed.ac.uk).
Clocklikeness of the resulting trees was investigated by using TempEst to determine if the molecular-clock assumption holds for the gene segment (24). The NS1 gene segment had no temporal signal and was not further analyzed. The H3, PA, and NP gene segments had positive, linear time vs. root-to-tip distance trend and a moderate scatter with correlation coefficients from 0.55 to 0.7. These gene segments were further analyzed to estimate the age of the most recent common ancestor.
We estimated tMRCAs by using tip date sampling and Bayesian Markov chain Monte Carlo analysis with the BEAST 1.10 package (25). Dates were estimated by using coalescent constant size prior with 3 partitions for the 3 codons using the maximum-likelihood estimation tree as the starting tree. Simulations were run using either the GTR or Hasegawa-Kishino-Yano substitution model and either the strict clock or the uncorrelated relaxed clock model to determine the robustness of the date estimates (26). The substitution and clock model had minor effects on the divergence date estimates. The maximum clade credibility trees (MCC) generated by the treeannotator package included with BEAST is reported. Figures were visualized and annotated by using the ggtree package (22).

Hemagglutination Inhibition Assay
We tested 2 equine H3 viruses [A/equine/ Uruguay  (28). In brief, 25 µL of serum was treated overnight with receptor-destroying enzyme (Denka Seiken, Co., Ltd., https://denka-seiken.com) and serially diluted 2-fold in 25 µL of phosphate-buffered saline in duplicate in a 96 well v-bottom plate. Each homologous virus was adjusted to 4 hemagglutination units, and 25 µL were added to the serum dilutions and incubated for 30 min at 4°C. Finally, 50 µL of 0.5% chicken erythrocytes were added to each of the wells, and the plate was incubated at 4°C for 30 min, after which results of the assay were read.

In Vitro Infections
A549 and MDCK cells were infected at a multiplicity of infection of 0.01 for 1 hour at 37°C. Cells were washed 3 times to remove unbound virus, and infected cells were cultured in appropriate medium con-taining 0.075% bovine serum albumin and 1 µg/mL l-1-tosylamido-2-phenylethyl chloromethyl ketonetreated trypsin. Aliquots of culture supernatants were collected at 6, 16, 24, 48, and 72 hours postinfection (hpi) and immediately stored at -80°C for the determination of virus titers by 50% median tissue culture infectious dose assay in MDCK cells as described (29).

Chicken Transmission Study
The chicken experiment was performed as described (30). In brief, 6-week-old SPF leghorn chickens (3/ group) were separated into 4 experimental groups: (A/Equine/NY/2/2016 H3N8, A/red-fronted coot/ Chile/5/2013 H3N6, A/red-gartered coot/Chile/ C16030/2016 H3N4, and A/red-fronted coot/ Chile/5/2013 H3N6) and inoculated with 10 6 50% egg infective dose/0.5 mL virus by intraocular, intranasal, and intratracheal routes. These chickens were placed in direct contact with virus-naive chickens (n = 9/group) 24 hours later. All chickens were observed daily for clinical signs of illness, such as body weight loss, labored breathing (including upper respiratory signs, such as coughing and sneezing), and diarrhea. To assess virus shedding, cloacal and tracheal swab specimens were collected at 3, 5, 7, 9, and 12 dpi. Swab specimens were stored in viral transport medium at -80°C until virus titration. Viral titers were established by using the method of Reed and Munch by performing 50% median tissue culture infectious dose assay analysis of the swab specimen inoculum with 50 µL of allantoic fluid and 50 µL of 0.5% chicken erythrocytes (29). All birds were euthanized at day 21 pi, and 1 mL blood was collected of each bird to check for seroconversion by HI. Chicken serum samples were tested with their corresponding homologous virus, according to their assigned group.

Internal Gene Phylogenetic Nucleotide Analysis
The nucleotide similarity with the EIV H3 HA prompted us to conduct a detailed phylogenetic analysis. Like HA, the H3 viruses from Chile had NS1 (Appendix Figure 3), NP (Appendix Figure 4), and PA (Appendix Figure 5)  from Canada (A/blue-winged teal/ALB/651/1978 (H6N2), GenBank accession no. ABB18989) that was more like these sequences from South America than other AIVs, suggesting that these gene fragments have spread into North America through bird migration, but did not establish a major reservoir to be consistently identified during surveillance. The NA, matrix, polymerase basic 1, and polymerase basic 2  gene segments showed greater similarities to typically avian origin gene segments (Appendix Figures  6, 7) and were not further analyzed. Bayesian molecular clock analysis was performed for the NP and PA gene segments to estimate the age of tMRCA between the 1963 H3 EIV and sequences from Chile (Table 1; Figure 4, panels B, C). The tM-RCA of the NP gene segment between most of the sequences from Chile and the 1963 H3N8 EIV is 1947 (95% credible interval 1936-1955) (Appendix Figure  8). The tMRCA for the PA gene is 1948 (95% credible interval 1938-1956) (Appendix Figure 9).
The NS1 segment had insufficient temporal signal for molecular clock analysis. Overall, our phylogenetic analysis suggests that many of the gene segments collected from wild birds in South America were distinct from those collected from wild birds in North America; the H3, PA, NP, and NS1 genes were more closely related to the 1963 H3N8 EIV, supporting the hypothesis that the 1963 H3N8 EIV originated from wild birds in South America. Our results are consistent with those of Rimondi et al., who reported that viruses from birds from Argentina also carry PA and NP sequences with homology to the EIV (12). These sequences are closely related to the highlighted sequences from Chile. Rimondi et al. also estimated the tMRCA for PA to be 1943, and the tMRCA for NP to be 1951, which were similar to our estimates.

Antigenicity and Biologic Properties
The H3 HA from Chile and the 1963 EIV H3 HA also show amino acid similarity, having 91.5% amino acid identity to A/equine/Uruguay/1/1963. Thus, antigenicity was assessed by using the HI assay with ferret antisera against equine viruses or viruses from Chile available for these studies. The H3 viruses from Chile were not inhibited by ferret-generated antisera against A/equine/Miami/1963 or A/equine/New York/2016 viruses ( Table 2). The equine viruses also failed to cross-react with ferret antisera generated against A/yellow-billed pintail/Chile/C2014/2015 virus ( Table 2). Antisera against the avian virus from Chile or equine virus did not inhibit H3 IAVs representing the 3 genetic wild bird clusters identified (Appendix Figure 1). These studies highlight that the H3 viruses from Chile are antigenically unique from equine and other wild bird origin H3 IAVs.
To compare biologic properties, we assessed virus growth in mammalian cells and infections in vivo. The viruses from Chile and 1963 EIV H3 viruses replicated to similar titers and kinetics in MDCK cells (except at 48 hpi) when the eq/Miami/63 virus replicated to significantly higher titers compared with the other H3 viruses (p<0.05) ( Figure 5, panel A). There was no major difference between eq/ Miami/63 virus and the control A/California/2009 H1N1 virus. Replication kinetics and titers were significantly different in human A549 cells in which eq/Miami/63 virus again replicated to significantly higher titers at 24 hpi and 48 hpi (p<0.05) ( Figure  5, panel B). The IAVs from Chile had lower overall titers and slower kinetics compared with those of eq/Miami/63 virus. This finding is highlighted by area under the curve analysis ( Figure 5, panel C). In contrast to the results with eq/Miami/63 virus, the IAVs from Chile had higher viral titers from 24 hpi to 72 hpi than eq/Uruguay/63 virus, which is genetically similar to eq/Miami/63 virus (p<0.05) ( Figure  5, panel A). A/red-footed Coot/Chile/5/2013 virus also replicated to higher titers than the eq/Uruguay/63 virus in A549 cells (p<0.05) ( Figure 5, panel  B). The reasons for the differential replication kinetics between the genetically similar equine viruses is unknown but was consistent. Attempts to replicate the viruses in EQKC3 and NBL6 equine cell lines were not successful.

Discussion
In these studies, we isolated several H3 influenza viruses from wild birds in Chile that contain HA, NP, PA, and NS segments that are genetically similar to the 1963 EIVs, suggesting that the direct descendants of the virus that originated the EIV pandemic continue to circulate in wild birds in Chile. Previous AIV surveillance studies in South America have identified internal genes that were putatively related to the ancestral virus that originated the 1963 equine pandemic (9,13,31); however, we have identified an EIV-like HA in wild birds. Although phylogenetic nucleotide sequence analysis places the H3 viruses from Chile in a monophyletic group together with equine viruses, amino acid composition of the HA and antigenic properties show that viruses from Chile are antigenically unique and can infect and transmit in poultry. However, to better interpret the antigenic data, future studies could benefit from the  use of antigenic cartography to learn about the antigenic evolution of this subtype (32). Although H3 is the most common subtype found in wild birds from North America (33), typical avian origin H3 subtypes are seldom recovered in South America. To date, only 4 isolates in Peru and 2 in Chile, all resembling contemporary North American AIVs, have been described (31,34). This finding could be related to the lack of surveillance throughout Latin America or low-level circulation of H3 viruses in wild birds (35). In contrast, we found that the H13 and H16 subtypes comprised up to 54% of the overall subtype diversity of AIVs from Chile deposited in GenBank during the same surveillance period (2013-2017), which could be caused by a biased surveillance effort in the order Charadriiformes. However, if we only consider isolates obtained from waterfowl in South America, the abundance of the H3 subtype (8.5%) is below the relative abundance of more common AIV subtypes, such as H1 (10.6%), H4 (10.6%), H5 (17%), and H7 (14.9%), which illustrates rarity of H3 subtype virus in wild birds in South America.
Global trade of thoroughbred horses from South America carrying the original 1963 H3N8 EIV, and not migratory birds, was responsible for the spread of these avian-origin gene segments (36). In comparison, although independent avian-to-equine transmission of lineage H3N8 AIV from Asia to horses was described in eastern Asia in the late 1980s, the epizootic event that followed was self-limited and died out after a few years (37). Subsequent active surveillance of wild birds in Mongolia has shown that AIVs carrying several gene segments closely related to H3N8 EIV from Asia are still circulating in wild birds, similar to our findings in Chile (38). This finding suggests that avian-to-equine transmission of H3 influenza A viruses is not an uncommon event, but that posterior sustained transmission is more limited. The conditions and key genetic signatures that facilitated the species-jump and rapid adaptation from waterfowl to horses in South America and eastern Asia remain unknown. This limitation reflects our inability to estimate whether the EIV-like HA reassorted together with the PA/NP/NS1 or in a separate event on the basis of our data. However, it is not uncommon to find wild waterfowl next to free-ranging horses in South America, which enables repeated transmission events that might lead to emergence of a new virus strain with pandemic potential in horses.