Genetic relationship between poultry and wild bird viruses during the highly pathogenic avian influenza H5N6 epidemic in the Netherlands, 2017–2018

Summary In the Netherlands, three commercial poultry farms and two hobby holdings were infected with highly pathogenic avian influenza (HPAI) H5N6 virus in the winter of 2017–2018. This H5N6 virus is a reassortant of HPAI H5N8 clade 2.3.4.4 group B viruses detected in Eurasia in 2016. H5N6 viruses were also detected in several dead wild birds during the winter. However, wild bird mortality was limited compared to the caused by the H5N8 group B virus in 2016–2017. H5N6 virus was not detected in wild birds after March, but in late summer infected wild birds were found again. In this study, the complete genome sequences of poultry and wild bird viruses were determined to study their genetic relationship. Genetic analysis showed that the outbreaks in poultry were not the result of farm‐to‐farm transmissions, but rather resulted from separate introductions from wild birds. Wild birds infected with viruses related to the first outbreak in poultry were found at short distances from the farm, within a short time frame. However, no wild bird viruses related to outbreaks 2 and 3 were detected. The H5N6 virus isolated in summer shares a common ancestor with the virus detected in outbreak 1. This suggests long‐term circulation of H5N6 virus in the local wild bird population. In addition, the pathogenicity of H5N6 virus in ducks was determined, and compared to that of H5N8 viruses detected in 2014 and 2016. A similar high pathogenicity was measured for H5N6 and H5N8 group B viruses, suggesting that biological or ecological factors in the wild bird population may have affected the mortality rates during the H5N6 epidemic. These observations suggest different infection dynamics for the H5N6 and H5N8 group B viruses in the wild bird population.

H5 viruses of different neuraminidase (NA) subtypes (Lee, Bertran, Kwon, & Swayne, 2017). Of these reassortant viruses, the clade 2.3.4.4 H5 viruses further evolved into four genetic groups, named A to D (Lee et al., 2016). Outbreaks in poultry in the Netherlands were caused by H5N8 group A viruses in 2014 , and H5N8 group B viruses in 2016 . In December 2017, an outbreak of HPAI H5N6 was detected on a commercial poultry farm in the Netherlands.
We recently showed that this virus is a novel reassortant of the H5N8 clade 2.3.4.4 group B viruses , which were first detected at the Russia-Mongolia border in May 2016. The H5N6 virus obtained novel polymerase basic 2 (PB2) and NA segments derived from Eurasian low pathogenic avian influenza (LPAI) viruses. After the first outbreak in poultry, H5N6 viruses were detected on two additional commercial poultry farms and two hobby holdings between December 2017 and March 2018. In this period, H5N6 virus was also detected in several dead wild birds that were tested in the passive wild bird surveillance programme in the Netherlands. Infected wild birds were also found in several other European countries (Germany, United Kingdom, Ireland, Denmark, Sweden, Finland and Slovakia).
Wild bird mortality caused by H5N6 viruses was limited compared to that observed during the H5N8 epidemic in  in 2014, no dead wild birds infected with the H5N8 group A virus were detected. These observations suggest striking differences in the pathogenicity of these HPAI H5 viruses in wild birds.
After March 2018, the virus was not detected in wild birds or poultry in the Netherlands for several months. However, in August two dead wild birds infected with H5N6 virus were found again.
Detection of the virus in late summer suggests long-term circulation of H5N6 viruses in the local wild bird population. Alternatively, it may be a novel incursion of H5N6 virus due to the start of fall migration of wild birds to Europe. In this study, we analysed the genetic relationship between the viruses isolated from poultry and wild birds. This analysis will provide insight in whether the farms were infected by farm-to-farm transmission, or by separate introductions from wild birds. In addition, the analysis will provide insight in the origin of the H5N6 virus that was detected in a wild bird in late summer. Finally, the pathogenicity of the H5N6 virus in ducks was compared to that of the H5N8 viruses that caused epidemics in 2014 and 2016. These combined results will provide more insight in the HPAI H5N6 epidemic, which affected both wild birds and poultry in the Netherlands in 2017-2018.

| Virus detection and subtyping
Viral RNA was extracted from tracheal or cloacal swabs from dead wild birds using the MagNa Pure 96 (Roche). For commercial poultry farms, pools of five samples from clinically affected chickens or ducks were used. Tracheal and cloacal swabs were pooled separately.
The samples were tested in a matrix gene real-time PCR, which detects all avian influenza (AI) virus subtypes, as described previously . The positive samples were then subtyped using a H5-specific real-time PCR (Slomka et al., 2007), as recommended by the European Union reference laboratory. The sequence of the hemagglutinin (HA) cleavage site and the NA subtype was determined by Sanger sequencing .

| Complete genome sequencing and analysis
Viral RNA was purified using the High Pure Viral RNA kit (Roche), and amplified using universal eight-segment primers and directly sequenced, as described previously .
Briefly, purified amplicons were sequenced at high coverage (average >1000 per nucleotide position) using the Nextera library preparation method and Illumina MiSeq paired-end 150 base pairs sequencing. Quality control-passed sequence reads were mapped using the ViralProfiler-Workflow, an extension of the CLC Genomics Workbench (Qiagen, Germany), as previously described . The consensus sequences generated in this study were submitted to the GISAID database. The GISAID accession numbers are listed in Table 1.

| Phylogenetic network
The eight-gene-segment alignments were manually concatenated to generate a single alignment that was used to construct phylogenetic networks using the median-joining method implemented in the program NETWORK, as described previously (Bataille, van der Meer, Stegeman, & Koch, 2011). This model-free method uses a parsimony approach based on pairwise differences to connect each sequence to its closest neighbour, and allows creation of internal nodes ('median vectors'), which could be interpreted as unsampled or extinct ancestral genotypes to link the existing genotypes in the most parsimonious way (Bandelt, Yao, Bravi, Salas, & Kivisild, 2009). In the analysis, we also included sequences from three wild bird viruses that were isolated in research programmes in the Netherlands (NL1-3); however, a forth sequence was excluded because of a reassortment of the polymerase acidic (PA)gene segment. The sequences of publically available Dutch and European sequences that were used in this study are listed in Table 1. a Sequences generated in a previous study, see reference (8). b We acknowledge the authors, originating and submitting laboratories of the sequences from GISAID's EpiFlu™ Database on which this research is based. All submitters of data may be contacted directly via the GISAID website www.gisaid.org.

| Detection of HPAI H5N6 in poultry and wild birds
The first introduction of HPAI H5N6 into a commercial poultry In the passive wild bird surveillance programme, diagnostic testing for AI is performed for wild birds found dead. Between 1 November 2017 and 1 April 2018, 281 dead birds were tested, of which 13 birds tested positive for HPAI H5N6 (Table 2).
Early in the epidemic, mostly birds of the family Anatidae

| Genetic analysis of the H5N6 epidemic
We previously determined the complete genome sequences of HPAI H5N6 viruses isolated from the first outbreak on the commercial poultry in Biddinghuizen and three wild birds (wb 1-3, Table 2) found dead nearby this farm ). In the current study, we performed deep sequencing to determine the complete genome sequences of H5N6 viruses that were detected later in 2017-2018 in commercial and hobby poultry holdings, and in dead wild birds.
To analyse the genetic relationship between viruses isolated from commercial poultry, a median-joining network analysis was performed ( Figure 2). We identified 47 nucleotide differences between the genomes of the viruses detected at the farm in Biddinghuizen Therefore, likely the farms were infected by separate introductions from wild birds.
The genetic relationship between H5N6 viruses isolated from poultry and wild birds was also studied. In this network analysis, we included publically available sequences of other European H5N6 virus isolates (Table 1). The analysis identified several wild bird viruses related to the virus isolated from the farm in Biddinghuizen ( Figure 3). The genetically most closely related wild bird virus (wb 2) contained eight nucleotide differences, and was isolated from a tufted duck. This dead bird was found 2 days after detection of the virus at the farm (distance 9 km) ( Table 3). Two mute swans (wb 1 and 3) that were found dead in the central lake area near Biddinghuizen on the same day also carried highly similar viruses.
For the two hobby holding, viruses isolated from three different poultry species were sequenced (Table 1)

| Pathogenicity of the H5N6 virus
The pathogenicity of the H5N6 virus was determined in Pekin ducks, and compared to that of the H5N8 viruses detected in the

| D ISCUSS I ON
In December 2017, the Netherlands was first to report HPAI H5N6 in Europe . Later in 2017-2018, the H5N6 virus was also detected in wild birds in several other European countries.
In the Netherlands, three commercial poultry farms and two hobby holdings were infected with H5N6 viruses. H5N6 viruses were detected in 15 dead wild birds tested in the passive wild bird surveillance programme. We previously showed that this virus is a novel reassortant of the HPAI H5N8 clade 2.3.4.4 group B virus that obtained new PB2 and NA segments derived from Eurasian LPAI viruses F I G U R E 3 Median-joining network showing the genetic relationship between HPAI H5N6 viruses isolated from commercial poultry farms (red), hobby holdings (green) and from dead wild birds tested in the passive surveillance programme (dark blue) in the Netherlands. Also shown are viruses isolated from wild birds in other (research) programmes in the Netherlands (light blue), and other European countries (grey). Predicted median vectors are shown in yellow. The virus isolates used for this analysis are listed in Table 1, which also provides the GISAID accession numbers  However, no wild bird viruses genetically related to outbreaks 2 and 3 were isolated. Late in summer, two wild birds found dead in the central lake area of the Netherlands tested positive for HPAI H5N6.
Genetic analysis showed that this "late" virus shares a common ancestor with the virus detected in outbreak 1, although these viruses differ at 57 nucleotide positions. As this virus was detected almost 9 months after outbreak 1, this suggests a nucleotide substitution rate of around 5.7 × 10 −3 substitutions/site/year. This calculated substitution rate is within the range previously estimated for HPAI H5 viruses (Fourment & Holmes, 2015;Rejmanek, Hosseini, Mazet, Daszak, & Goldstein, 2015). This suggests that the virus most likely showed that the pathogenicity of the three viruses for galliform birds was similar, for all viruses an IVPI score of 3.0 was measured .  prey were infected (Kleyheeg et al., 2017). These predators were most likely infected by feeding on infected preys. During the H5N6 epidemic, the virus was mainly detected in residential birds (mute swans), and later again in birds of prey. Furthermore, genetic analysis provided indications for long-term circulation of H5N6 viruses in the local wild bird population. These observations suggest different infection dynamics for the H5N6 and H5N8 group B viruses in the wild bird population in the Netherlands.

ACK N OWLED G EM ENTS
We thank Sandra Venema for technical assistance, José Gonzales for the programme for plotting maps and Armin Elbers for helpful discussion. We acknowledge the NVWA and GD for excellent cooperation during the outbreak, and the sample preparation/diagnostics departments of WBVR for handling all samples. This work was funded by the Dutch Ministry of Agriculture, Nature and Food Quality (project WOT-01-003-012). We acknowledge the authors and submitting laboratories of the sequences from the Global Initiative on Sharing All Influenza Data (GISAID) EpiFlu Database (www.gisaid.org).

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest.