Evolution and Divergence of H3N8 Equine Influenza Viruses Circulating in the United Kingdom from 2013 to 2015

Equine influenza viruses (EIV) are a major cause of acute respiratory disease in horses worldwide and occasionally also affect vaccinated animals. Like other influenza A viruses, they undergo antigenic drift, highlighting the importance of both surveillance and virus characterisation in order for vaccine strains to be kept up to date. The aim of the work reported here was to monitor the genetic and antigenic changes occurring in EIV circulating in the UK from 2013 to 2015 and to identify any evidence of vaccine breakdown in the field. Virus isolation, reverse transcription polymerase chain reaction (RT-PCR) and sequencing were performed on EIV-positive nasopharyngeal swab samples submitted to the Diagnostic Laboratory Services at the Animal Health Trust (AHT). Phylogenetic analyses were completed for the haemagglutinin-1 (HA1) and neuraminidase (NA) genes using PhyML and amino acid sequences compared against the current World Organisation for Animal Health (OIE)-recommended Florida clade 2 vaccine strain. Substitutions between the new isolates and the vaccine strain were mapped onto the three-dimensional structure protein structures using PyMol. Antigenic analyses were carried out by haemagglutination inhibition assay using a panel of post-infection ferret antisera. Sixty-nine outbreaks of equine influenza in the UK were reported by the AHT between January 2013 and December 2015. Forty-seven viruses were successfully isolated in eggs from 41 of the outbreaks. Only three cases of vaccine breakdown were identified and in each case the vaccine used contained a virus antigen not currently recommended for equine influenza vaccines. Nucleotide sequencing of the HA and NA genes revealed that all of the viruses belonged to the Florida clade 2 sub-lineage of H3N8 EIV. Phylogenetic and sequence analyses showed that the two sub-populations, previously identified within clade 2, continued to circulate and had accrued further amino acid substitutions. Antigenic characterisation using post-infection ferret antisera in haemagglutination inhibition assays however, failed to detect any marked antigenic differences between the isolates. These findings show that Florida clade 2 EIV continue to circulate in the UK and support the current OIE recommendation to include an example of Florida clade 2 in vaccines.


Introduction
Equine influenza virus (EIV) is a major cause of acute respiratory disease-of a highly contagious nature-in horses and other equids.It is endemic in most parts of the world and can cause severe disruption to the racing and breeding industries.Like other influenza A viruses, EIV has two surface glycoproteins, haemagglutinin (HA) and neuraminidase (NA), which perform essential roles in virus entry and exit.HA is responsible for binding to host cell sialic acid receptors and promotes membrane fusion [1,2].NA has sialidase activity, which is important for virus exit and release, but has also been suggested to have an early role in virus entry through the removal of decoy receptors in the host's respiratory tract [3].HA is a major target for neutralising antibodies and so is an important component of commercial vaccines.The importance of EIV NA to immunity however, is currently unknown, although antibodies to human influenza NA have been shown to contribute to protection [4].NA has been the target of second-generation antiviral drugs to inhibit its function and prevent virus propagation.Although not routinely used in horses, NA inhibitors have been shown to have potential for the treatment of EIV infection [5,6].
H3N8 EIV was first isolated in 1963 during a widespread outbreak in the United States and has continued to circulate ever since [7][8][9].It has been responsible for causing major outbreaks of disease among horses around the world, including in the UK in 1979, 1989 and 2003 as well as in South Africa in 2003, Japan and Australia in 2007, India in 2008-2009 and, more recently, in South America in 2012 [10][11][12][13][14][15][16][17][18].Thought to have originated from an avian source in South America, the virus has since adapted to the horse and diverged into antigenically and genetically distinct lineages and sub-lineages [8,19,20].The American and Eurasian lineages emerged during the 1980s and since then the American lineage has further evolved into the Kentucky, South American and Florida sub-lineages [20].In the early 2000s, the Florida sub-lineage diverged into clades 1 (FC1) and 2 (FC2), and has been the dominant circulating sub-lineage ever since.Although not fully restricted geographically, FC2 viruses have predominately been isolated in Europe and Asia, whilst the majority of FC1 viruses have been isolated in North America [8,9,15,[21][22][23][24][25][26].
Since being made compulsory in the racing industry in the early 1980s, vaccination has been widely used for the control of EIV.However, like other influenza A viruses, EIV undergoes antigenic drift through the accumulation of amino acid substitutions over time in HA, which has led to vaccine breakdown in the past [13,27,28].EIV vaccines therefore need to be updated periodically and a formal process of vaccine strain selection, overseen by the World Organisation for Animal Health (OIE) has been in place for several years.An expert surveillance panel (ESP) makes vaccine strain recommendations based on genetic, antigenic and epidemiological data that are collected and reported to the panel on an annual basis [29].Current recommendations are to include both FC1 and FC2 strains in vaccines [30].Surveillance data are therefore crucial to this process.Our aim was to continue to monitor the genetic and antigenic changes occurring in EIV circulating in the UK.Here we present the HA and NA sequences, as well as antigenic characterisation of viruses isolated between 2013 and 2015 from disease outbreaks in the UK.We show that FC2 viruses have continued to diverge from the OIE-recommended vaccine strain into two sub-groups, with multiple amino acid substitutions occurring in both HA and NA.

Results
Between January 2013 and December 2015, 69 outbreaks of equine influenza in the UK were reported by the Animal Health Trust (AHT).These affected 28 counties in England, 8 in Scotland and 2 in Wales, with multiple outbreaks in some areas.Diagnoses were made by either qRT-PCR or nucleoprotein (NP)-ELISA performed on extracts from nasopharyngeal swabs submitted to the Diagnostic Laboratory Services at the AHT.In total, 84 samples from the 69 outbreaks tested positive.Virus was successfully isolated in embryonated hens' eggs from 47 positive samples corresponding to 41 outbreaks.Apart from three isolates, all of the viruses were isolated from samples taken from either unvaccinated animals (n = 45), or animals whose vaccination status was out of date (n = 2).Ayrshire/1/13 and Ayrshire/2/13 were isolated from horses that had received a booster vaccination with ProteqFlu TE, a Canarypox vaccine expressing the HA from Ohio/03 and Newmarket/2/93, 6 months previously.These horses had both received the primary course of vaccinations 5-7 months before the booster.Buckinghamshire/1/14 was isolated from a horse that had a complete vaccination history and had received the last annual booster vaccination with ProteqFlu eight months previously, although this horse had received a primary course and previous booster vaccinations with Duvaxyn IE-T/IE-T Plus containing Suffolk/89, Newmarket/1/93 and Prague/56.However, the antibody titres of these vaccinated animals were unknown at the time of infection.Details of the virus isolates are summarised in Table 1.

Genetic Analyses-HA
The nucleotide sequence encoding HA was determined for each of the virus isolates by Sanger dideoxynucleotide sequencing and the resulting sequences were deposited in the GISAID EpiFlu™ database [31] (Table 1).For phylogenetic analyses, nucleotide sequences from strains representative of the different H3N8 EIV lineages were retrieved from the Genbank and GISAID databases.Due to a lack of available full-length HA sequences, a maximum likelihood phylogenetic tree was constructed using the nucleotide sequences for HA1 only (Figure 1).Distinct clusters within the phylogenetic tree corresponded to the pre-divergent, Eurasian and American lineages, as well as the two clades of the Florida sub-lineage.All of the sequences from the 2013-2015 isolates were located within the FC2 group, in the expanded section of the tree, however two distinct sub-groups were evident separating strains encoding the A144V substitution (144-group) from those with the I179V substitution (179-group; HA1 residues are numbered from the serine residue downstream of the predicted signal peptide).Only two of the isolates from the UK, Ayrshire/1/13 and Ayrshire/2/13 were within the I179V cluster, together with German (North-Rhine Westphalia/1/14) and Italian (Rome/1/14) strains isolated in 2014.The remaining UK isolates from 2013-2015 were grouped with strains isolated in previous years that encoded the A144V substitution (Figure 1).The more recent 2015 isolates formed a distinct group from the 2014 and early 2015 isolates (Figure 1).A Chinese isolate from 2013, Xuzhou/1/13 (Genbank accession number KF806985.1),clustered with other Asian strains from 2011, suggesting that another sub-group of FC2 may be circulating in Asia (Figure 1).
The derived amino acid sequences for the full-length HA from the 2013-2015 UK isolates were compared to the current OIE-recommended FC2 vaccine strain A/eq/Richmond/1/07 and representative UK isolates from 2011 to 2012 (Figure 2).As in the phylogenetic tree (Figure 1) the sequences divided into two separate groups, one containing the strains encoding the A144V substitution and the other containing those with the I179V substitution (Figure 2).All of the isolates encoded the amino acid changes P103L, V112I (except Shropshire/8/13) and E291D in HA1 compared to Richmond/1/07.The majority of the 144-group also encoded V300I in HA1 and L187M in HA2 (Figure 2).Of the 2015 isolates, 9 out of 14 encoded an additional substitution, T192K, within antigenic site B located at the top of the HA trimer.The most recent of these isolates contained a further three substitutions: V267I in HA1 and T43A and L187I in HA2 (Figure 2).These were the four isolates that formed the distinct group in the phylogenetic tree (Figure 1).The two 179-group isolates, Ayrshire/1/13 and Ayrshire/2/13, shared the same additional amino acid substitutions in HA1 as North-Rhine Westphalia/1/14 and Rome/1/14: I46T, (I179V), T192K and I282V (Figure 2).Unlike the 144-group, there were no substitutions within HA2 of the Ayrshire/13 viruses compared to Richmond/1/07 (Figure 2).None of the amino acid substitutions observed in either group affected glycosylation sites within HA when compared to the recommended vaccine strain.The genetic similarities between the two sub-groups and the recommended vaccine strain are shown in Table S1.
The locations of the individual amino acid substitutions for each sub-group (144 and 179) were mapped on the three-dimensional structure of the equine H3 HA (Figure 3A,B) [32] compared against the OIE-recommended vaccine strain A/eq/Richmond/1/07.For the 144-group, two changes within antigenic sites (A144V-site A and T192K-site B) were visible on the globular head of HA1 (Figure 3A).The E291D substitution was at the base of the globular head of HA1, whilst the two HA2 substitutions T43A and N154S were visible on the surface of the stalk structure (Figure 3).Four of the HA1 substitutions mapped to internal regions of the structure (P103L, V112I, V267I and V300I/F) and are shown on a monomeric ribbon structure (Figure 3).Two further substitutions residing at the C-terminal domain of HA2, L187M/I and G204S, could not be mapped as they were not present in the crystal structure.
The amino acid substitutions observed in the 179-group were also mapped on the equine HA structure (Figure 3B).In addition to the changes that were common to both the 144-and 179-groups when compared to A/eq/Richmond/1/07, one substitution (I46T) was visible on the surface at the base of the globular head of HA1 and two substitutions (I179V and I282V) mapped to internal regions of the structure and are shown on a monomeric ribbon structure (Figure 3).older strains in black.The Florida clade 2 World Organisation for Animal Health (OIE)-recommended vaccine strain A/eq/Richmond/1/07 is shown in bold capitals.The HA1 sequences for the non-UK strains Xuzhou/1/13, Rome/1/14 and North-Rhine Westphalia/1/14 were obtained from Genbank (accession numbers KF806985.1,KR534268.1 and KJ538149.1 respectively).[9].The HA1 sequence for A/eq/Rome/1/14 was obtained from Genbank (accession number KR534268.1)and is included for comparison.

Genetic Analyses-NA
As for HA, the nucleotide sequence encoding NA was determined for each of the virus isolates and deposited in the GISAID EpiFlu™ database [31] (Table 1).A maximum likelihood phylogenetic tree was constructed using NA sequences downloaded from the Genbank and GISAID databases (Figure 4).The tree topology was similar to the HA tree and all of the 2013-2015 isolates reported here clustered within FC2 (Figure 4).The 2013-2015 isolates also grouped in a similar pattern to the HA tree, with the most recent 2015 isolates forming a distinct sub-group (Figure 4).The Ayrshire/13 179-group viruses were positioned separately to the other 2013-2015 isolates and were most similar to the NA sequence from the 2009 UK isolate Dorset/09 (Figure 4).
The derived amino acid sequences for NA from the 2013-2015 UK isolates were compared to the current OIE-recommended vaccine strain A/eq/Richmond/1/07 and representative UK isolates from 2011 and 2012 (Figure 5).As expected from the tree (Figure 4), and like the HA sequences (Figure 2), the sequences divided into two separate groups based on the 144 and 179 substitutions observed in HA.There were five amino acid substitutions common to all of the 144-group viruses: H25N, R109K, I410V, R415R and K434S (Figure 5).Isolates from late 2013, as well as the first isolate from 2014 (North Yorkshire/1/14), shared a substitution at position 44 (N44S), however this was not present in the other 2014 or 2015 isolates and appeared to be replaced by viruses encoding the G42C substitution (Figure 5).The four latest isolates from 2015, which formed a sub-group in the tree (Figure 4), had two further substitutions: T265I and A359V (Figure 5).The Ayrshire/13 isolates shared the I410V substitution with the 144-group and the 179 virus East Renfrewshire/2/11, but did The amino acid substitutions observed in the 179-group were also mapped on the equine HA structure (Figure 3B).In addition to the changes that were common to both the 144-and 179-groups when compared to A/eq/Richmond/1/07, one substitution (I46T) was visible on the surface at the base of the globular head of HA1 and two substitutions (I179V and I282V) mapped to internal regions of the structure and are shown on a monomeric ribbon structure (Figure 3).

Genetic Analyses-NA
As for HA, the nucleotide sequence encoding NA was determined for each of the virus isolates and deposited in the GISAID EpiFlu™ database [31] (Table 1).A maximum likelihood phylogenetic tree was constructed using NA sequences downloaded from the Genbank and GISAID databases (Figure 4).The tree topology was similar to the HA tree and all of the 2013-2015 isolates reported here clustered within FC2 (Figure 4).The 2013-2015 isolates also grouped in a similar pattern to the HA tree, with the most recent 2015 isolates forming a distinct sub-group (Figure 4).The Ayrshire/13 179-group viruses were positioned separately to the other 2013-2015 isolates and were most similar to the NA sequence from the 2009 UK isolate Dorset/09 (Figure 4).amino acid substitutions observed in the NA sequences affected glycosylation sites when compared to the recommended vaccine strain.In addition, none of the residues which form the active site or are involved in the catalytic function of the neuraminidase were affected.The H275Y mutation responsible for resistance to the antiviral drug oseltamivir was also not present in these viruses.The genetic similarities between the two sub-groups and the recommended vaccine strain are shown in Table S1.The derived amino acid sequences for NA from the 2013-2015 UK isolates were compared to the current OIE-recommended vaccine strain A/eq/Richmond/1/07 and representative UK isolates from 2011 and 2012 (Figure 5).As expected from the tree (Figure 4), and like the HA sequences (Figure 2), the sequences divided into two separate groups based on the 144 and 179 substitutions observed in HA.There were five amino acid substitutions common to all of the 144-group viruses: Pathogens 2017, 6, 6 9 of 16 H25N, R109K, I410V, R415R and K434S (Figure 5).Isolates from late 2013, as well as the first isolate from 2014 (North Yorkshire/1/14), shared a substitution at position 44 (N44S), however this was not present in the other 2014 or 2015 isolates and appeared to be replaced by viruses encoding the G42C substitution (Figure 5).The four latest isolates from 2015, which formed a sub-group in the tree (Figure 4), had two further substitutions: T265I and A359V (Figure 5).The Ayrshire/13 isolates shared the I410V substitution with the 144-group and the 179 virus East Renfrewshire/2/11, but did not share any of the other substitutions.These viruses had five distinct substitutions compared to Richmond/1/07 and the other viruses: V22I, H66Y, I376V, D396G and K451E (Figure 5).None of the amino acid substitutions observed in the NA sequences affected glycosylation sites when compared to the recommended vaccine strain.In addition, none of the residues which form the active site or are involved in the catalytic function of the neuraminidase were affected.The H275Y mutation responsible for resistance to the antiviral drug oseltamivir was also not present in these viruses.The genetic similarities between the two sub-groups and the recommended vaccine strain are shown in Table S1.In order to map the amino acid substitutions onto the three-dimensional structure of NA, the amino acid numbering of the protein was adjusted to correspond to the H5N1 NA protein structure database file 2HTY (Figure S1) [33].Substitutions within the membrane anchor and stalk regions could not be mapped as these are not included in the solved protein structure.Of the 144-group substitutions, R109K and T434S were visible on the top of the globular head with the 109 substitution forming a ring at the centre of the tetramer (Figure 6).The T265I and D386N changes were visible on In order to map the amino acid substitutions onto the three-dimensional structure of NA, the amino acid numbering of the protein was adjusted to correspond to the H5N1 NA protein structure database file 2HTY (Figure S1) [33].Substitutions within the membrane anchor and stalk regions could not be mapped as these are not included in the solved protein structure.Of the 144-group substitutions, R109K and T434S were visible on the top of the globular head with the 109 substitution forming a ring at the centre of the tetramer (Figure 6).The T265I and D386N changes were visible on the side of the molecule, whilst the I410V and K415R mapped to the base of the globular head and were adjacent to one another (Figure 6).The A359V substitution observed in the most recent 2015 isolates did not map on the surface of the structure.Of the 179-group changes, the D396G substitution mapped to the top of the molecule, I376V and K451E to the side and I410V to the base of the tetramer (Figure 6).isolates from the 144-and 179-sub-groups and the OIE-recommended vaccine strain A/eq/Richmond/1/07.Residues are numbered according to the N8 sequence but shown on the structure of N1 [33].Top, side and base views of the neuraminidase tetramer are shown for each sub-group.The 144-group change A359V is not shown, and N4S, A13T, H25N, G42C and N44S changes could not be mapped.179-group changes V22I and H66Y could not be mapped.

Antigenic Characterisation
Isolates from each of the years reported here, and representative of each of the HA sequences, were characterised by haemagglutination inhibition (HI) assay against a panel of post-infection ferret antisera raised against representative strains from the Eurasian and American lineages (including the Kentucky and FC1 and FC2 sub-lineages).The panel included both of the OIE-recommended vaccine strains (Richmond/1/07 and South Africa/4/03) and antisera raised against viruses with 144 and 179 substitutions.The geometric mean titres (GMT) of duplicate assays are shown in Table 2.The titres achieved using the antiserum raised to the Eurasian lineage virus Newmarket/2/93 against the 2013-2015 isolates were the lowest, with titres to most strains 8-to 32-fold lower than the homologous virus, supporting the ESP recommendation in 2010 to replace Eurasian lineage strains in vaccines with a Florida clade 2 strain (Table 2) [34].For all of the 2013-2015 isolates the highest titres were observed with either the Richmond/1/07 antiserum or the Ayrshire/1/13 antiserum, or both, ranging from 362 to 1024 (Table 2).The highest titres against both of Figure 6.Structure of the neuraminidase (NA) protein [33] (RCSB Protein Data Bank, accession number-2HTY) showing changes between 2013 (green dot), 2014 (red dot) and 2015 (blue dot) isolates from the 144-and 179-sub-groups and the OIE-recommended vaccine strain A/eq/Richmond/1/07.Residues are numbered according to the N8 sequence but shown on the structure of N1 [33].Top, side and base views of the neuraminidase tetramer are shown for each sub-group.The 144-group change A359V is not shown, and N4S, A13T, H25N, G42C and N44S changes could not be mapped.179-group changes V22I and H66Y could not be mapped.

Discussion
Since its emergence in 1963, H3N8 equine influenza has diverged into two distinct lineages and three sub-lineages, with Florida clades 1 and 2 predominating in recent years.The current recommendation to include a representative from both FC1 and FC2 in vaccines was first made by the OIE in 2010 [34].Examples of both clades have previously been isolated from horses in the UK, however only FC2 has been detected in the UK since 2010 [9,35].
The majority of EIV-positive samples submitted to the equine influenza surveillance programme at the AHT were from either unvaccinated horses or those that did not have up-to-date vaccination records.Only three cases of EIV infection in vaccinated horses were detected over the three years and in each case the vaccines used did not comply with the OIE recommendation to contain an FC2 strain.
Analysis of virus isolates from 2013 to 2015 revealed that, as in previous years, clade 2 viruses continued to circulate in the UK, whilst no clade 1 viruses were isolated during this time.Sequence analysis split the isolates into two sub-groups of clade 2, 144-group and 179-group, as reported previously, for both the HA and NA sequences [9].Whilst the majority of UK isolates belonged to the 144-group, sequences published on Genbank from Germany and Italy showed that viruses belonging to the 179-group were also circulating in Europe during this time.The HA and NA sequences of the two sub-groups had diverged further, with 7 amino acid substitutions in HA and 12 in NA between the most recent viruses.Interestingly, the T192K substitution in antigenic site B observed in the 2013 179-group viruses was also present in more recent 144-group isolates from 2015.This would also suggest that the T192K substitution in the 144-group occurred independently from the same change in the 179-group.Despite two changes in antigenic sites, antigenic analysis using post-infection ferret antisera did not reveal any marked difference between the two groups.However, a weakness of the HI assay is that only substitutions appearing close to the receptor binding site are likely to interrupt antibody-binding activity that block binding of HA1 to sialic acid receptors on red blood cells.Other substitutions, which may alter binding of neutralising antibodies at other regions of the HA, for example antibodies that bind to the stem region of the HA [36,37], would not necessarily be detected using this assay.Although not in the stem region, the A144V substitution has previously been shown to affect neutralisation titres of post-infection equine antisera raised against an Argentinian lineage vaccine strain (La Plata/93) in an egg-based virus-neutralisation (VN) assay [38].Moreover, horses vaccinated with the Argentinian lineage virus had increased temperatures and shed more virus following challenge with an A144V virus compared to horses vaccinated with a clade 2 virus lacking the A144V substitution [39].Whilst the HI titres against the A144V virus were similar between the vaccine groups, the VN titres were reported to be in the order of 8-fold lower [39].This suggests that the A144V substitution has an antigenic effect, which is not detected by post-infection ferret antisera, or indeed equine antisera in the HI assay.
Analysis of the NA sequences from the 2013-2015 isolates also showed divergence between the 144-and 179-group viruses, with at least 10 amino acid differences between the two sub-groups.The antigenic effects of these substitutions in NA are currently unknown and further work is needed to determine the role of antibodies to NA in the horse.
The continuing divergence of clade 2 viruses in Europe into the 144-and 179-groups may lead to the circulation of two antigenically distinct sub-lineages.A third group of clade 2 viruses appears to be circulating in Asia.Sequences available on Genbank from Chinese (2013) and Mongolian isolates (2011) show that these viruses share a different substitution at position 144 of HA (A144T) compared to the European clade 2 viruses.Careful monitoring of these three sub-groups is needed to ensure that appropriate strains are included in vaccines to protect against these groups, although at present there is no evidence to suggest that the current recommended strain (Richmond/1/07) would not provide adequate protection against either of the sub-groups circulating in Europe.
In conclusion, FC2 has continued to diverge, with the 144-group viruses predominating in the UK and the 179-group viruses in Europe, however there is a shortage of sequence data available from Europe for recent years.There appears to be three sub-groups within FC2: UK (A144V), Europe (I179V) and Asia (A144T).There is no evidence of significant antigenic drift away from the recommended vaccine strains.The recommendations to include FC2 as well as FC1 therefore still stand.The viruses have acquired further mutations in both HA and NA, therefore careful monitoring is required to identify antigenic drift at an early stage.

Haemagglutination Inhibition (HI) Assay
Post-infection ferret antisera from AHT archives were used to conduct antigenic characterisation of untreated viruses by HI assay.The antisera were treated with periodate and heat [41] and serially-diluted across a 96-well plate.The viruses were diluted to 4 HA units and incubated with the diluted antisera at room temperature for 30 min before 1% chicken blood was added and plates incubated at 4 • C for 45 min.HI titres were recorded as the highest dilution of antisera to show full haemagglutination inhibition.Assays were carried out in duplicate and geometric mean titres calculated.

Figure 1 .
Figure 1.HA1 phylogenetic tree.Phylogenetic analysis of HA1 nucleotide sequences encoded by EIV.A maximum likelihood tree was created using PhyML version 3. Bootstrap values obtained after 100 replicates are shown at major nodes.Phylogenetic groups are shown by continuous bars on the left, A-pre-divergent, B-Eurasian, C-American (Kentucky and Argentinian), D-Florida sub-lineage clade 2, E-Florida sub-lineage clade 1.The Florida sub-lineage clade 2 group (D) has been enlarged and positioned to the right with a dashed line indicating its branch position in the tree on the left.Amino acid substitutions within clade 2 of the Florida sub-lineage are indicated at branch points or in brackets-HA1 substitutions are in normal text, HA2 substitutions are italicised.Sequences are coloured by outbreak date for the years 2013 (green), 2014 (red) and 2015 (blue) with

Figure 1 .
Figure 1.HA1 phylogenetic tree.Phylogenetic analysis of HA1 nucleotide sequences encoded by EIV.A maximum likelihood tree was created using PhyML version 3. Bootstrap values obtained after 100 replicates are shown at major nodes.Phylogenetic groups are shown by continuous bars on the left, A-pre-divergent, B-Eurasian, C-American (Kentucky and Argentinian), D-Florida sub-lineage clade 2, E-Florida sub-lineage clade 1.The Florida sub-lineage clade 2 group (D) has been enlarged and positioned to the right with a dashed line indicating its branch position in the tree on the left.Amino acid substitutions within clade 2 of the Florida sub-lineage are indicated at branch points or in brackets-HA1 substitutions are in normal text, HA2 substitutions are italicised.Sequences are coloured by outbreak date for the years 2013 (green), 2014 (red) and 2015 (blue) with older strains in black.The Florida clade 2 World Organisation for Animal Health (OIE)-recommended vaccine strain A/eq/Richmond/1/07 is shown in bold capitals.The HA1 sequences for the non-UK strains Xuzhou/1/13, Rome/1/14 and North-Rhine Westphalia/1/14 were obtained from Genbank (accession numbers KF806985.1,KR534268.1 and KJ538149.1 respectively).

Figure 2 .
Figure 2. Amino acid substitutions within haemagglutinin (HA) between the OIE-recommended Florida clade 2 vaccine strain A/eq/Richmond/1/07 and strains isolated in the UK between 2013 and 2015.Isolates are grouped as belonging to either the 144-or 179-group, ordered by outbreak date and coloured by year: 2013-green, 2014-red, 2015-blue.HA1 residues are numbered from the serine residue downstream of the predicted signal peptide.Amino acid identity to A/eq/Richmond/1/07 is shown with a dot.Examples of strains from 2011 to 2012 are included to allow comparison with Woodward et al. (2014)[9].The HA1 sequence for A/eq/Rome/1/14 was obtained from Genbank (accession number KR534268.1)and is included for comparison.

Figure 2 .
Figure 2. Amino acid substitutions within haemagglutinin (HA) between the OIE-recommended Florida clade 2 vaccine strain A/eq/Richmond/1/07 and strains isolated in the UK between 2013 and 2015.Isolates are grouped as belonging to either the 144-or 179-group, ordered by outbreak date and coloured by year: 2013-green, 2014-red, 2015-blue.HA1 residues are numbered from the serine residue downstream of the predicted signal peptide.Amino acid identity to A/eq/Richmond/1/07 is shown with a dot.Examples of strains from 2011 to 2012 are included to allow comparison with Woodward et al. (2014)[9].The HA1 sequence for A/eq/Rome/1/14 was obtained from Genbank (accession number KR534268.1)and is included for comparison.

Figure 3 .
Figure 3. Structure of A/eq/Richmond/1/07 H3 HA [32] (RCSB Protein Data Bank, accession number-4UO0) showing changes between 2013 (green dot), 2014 (red dot) and 2015 (blue dot) isolates from the 144 (A) and 179 (B) sub-groups and the OIE-recommended vaccine strain A/Eq/Richmond/1/07.Three views of HA structure are shown-side view, top view and a monomeric ribbon structure highlighting internal changes which are not visible on the trimeric structure.Changes are shown in purple and are labelled with amino acid substitution.The 144-group HA2 changes L187M/I (2013-2015) and G204S (2014 only) could not be mapped.

Figure 3 .
Figure 3. Structure of A/eq/Richmond/1/07 H3 HA [32] (RCSB Protein Data Bank, accession number-4UO0) showing changes between 2013 (green dot), 2014 (red dot) and 2015 (blue dot) isolates from the 144 (A) and 179 (B) sub-groups and the OIE-recommended vaccine strain A/Eq/Richmond/1/07.Three views of HA structure are shown-side view, top view and a monomeric ribbon structure highlighting internal changes which are not visible on the trimeric structure.Changes are shown in purple and are labelled with amino acid substitution.The 144-group HA2 changes L187M/I (2013-2015) and G204S (2014 only) could not be mapped.

Figure 4 .
Figure 4. Phylogenetic tree of neuraminidase (NA) nucleotide sequences encoded by EIV.A maximum likelihood tree was generated using PhyML version 3, and phylogenetic groups are shown: A-pre-divergent, B-Eurasian, C-American (Kentucky and Argentinian), D-Florida clade 2, E-Florida clade 1. Bootstrap values are shown at major nodes.Amino acid substitutions within FC2 are indicated at branch points.Sequences are coloured by outbreak date for the years 2013 (green), 2014 (red) and 2015 (blue) with older strains in black.The FC2 OIE-recommended vaccine strain A/eq/Richmond/1/07 is shown in bold capitals.

Figure 4 .
Figure 4. Phylogenetic tree of neuraminidase (NA) nucleotide sequences encoded by EIV.A maximum likelihood tree was generated using PhyML version 3, and phylogenetic groups are shown: A-pre-divergent, B-Eurasian, C-American (Kentucky and Argentinian), D-Florida clade 2, E-Florida clade 1. Bootstrap values are shown at major nodes.Amino acid substitutions within FC2 are indicated at branch points.Sequences are coloured by outbreak date for the years 2013 (green), 2014 (red) and 2015 (blue) with older strains in black.The FC2 OIE-recommended vaccine strain A/eq/Richmond/1/07 is shown in bold capitals.

Figure 5 .
Figure 5. Amino acid substitutions within NA between the OIE-recommended Florida sub-lineage clade 2 vaccine strain A/eq/Richmond/1/07 and strains isolated in the UK between 2013 and 2015.Isolates are grouped as belonging to either the 144 or 179 HA1 group, are ordered by outbreak date and coloured by year as in Figures 1, 2 and 4: 2013-green, 2014-red, 2015-blue.Residues are numbered from the methionine residue at the start of the predicted signal peptide and have not been adjusted to correspond to N1 or N2 numbering.Amino acid identity to A/eq/Richmond/1/07 is shown with a dot.Examples of strains from 2011 to 2012 are included to allow comparison with Woodward et al. (2014) [9].

Figure 5 .
Figure 5. Amino acid substitutions within NA between the OIE-recommended Florida sub-lineage clade 2 vaccine strain A/eq/Richmond/1/07 and strains isolated in the UK between 2013 and 2015.Isolates are grouped as belonging to either the 144 or 179 HA1 group, are ordered by outbreak date and coloured by year as in Figures 1, 2 and 4: 2013-green, 2014-red, 2015-blue.Residues are numbered from the methionine residue at the start of the predicted signal peptide and have not been adjusted to correspond to N1 or N2 numbering.Amino acid identity to A/eq/Richmond/1/07 is shown with a dot.Examples of strains from 2011 to 2012 are included to allow comparison with Woodward et al. (2014) [9].

Pathogens 2017, 6 , 6 10 of 16 Figure 6 .
Figure 6.Structure of the neuraminidase (NA) protein[33] (RCSB Protein Data Bank, accession number-2HTY) showing changes between 2013 (green dot), 2014 (red dot) and 2015 (blue dot) isolates from the 144-and 179-sub-groups and the OIE-recommended vaccine strain A/eq/Richmond/1/07.Residues are numbered according to the N8 sequence but shown on the structure of N1[33].Top, side and base views of the neuraminidase tetramer are shown for each sub-group.The 144-group change A359V is not shown, and N4S, A13T, H25N, G42C and N44S changes could not be mapped.179-group changes V22I and H66Y could not be mapped.

Table 1 .
Virus isolates from outbreaks of equine influenza virus (EIV) in the UK from 2013 to 2015 including locations, detection methods and GISAID accession numbers.

Table 2 .
Haemagglutination inhibition titres of equine influenza virus strains using ferret antisera.