Highly Pathogenic Avian Influenza A(H5N1) Virus Struck Migratory Birds in China in 2015

Approximately 100 migratory birds, including whooper swans and pochards, were found dead in the Sanmenxia Reservoir Area of China during January 2015. The causative agent behind this outbreak was identified as H5N1 highly pathogenic avian influenza virus (HPAIV). Genetic and phylogenetic analyses revealed that this Sanmenxia H5N1 virus was a novel reassortant, possessing a Clade 2.3.2.1c HA gene and a H9N2-derived PB2 gene. Sanmenxia Clade 2.3.2.1c-like H5N1 viruses possess the closest genetic identity to A/Alberta/01/2014 (H5N1), which recently caused a fatal respiratory infection in Canada with signs of meningoencephalitis, a highly unusual symptom with influenza infections in humans. Furthermore, this virus was shown to be highly pathogenic to both birds and mammals, and demonstrate tropism for the nervous system. Due to the geographical location of Sanmenxia, these novel H5N1 viruses also have the potential to be imported to other regions through the migration of wild birds, similar to the H5N1 outbreak amongst migratory birds in Qinghai Lake during 2005. Therefore, further investigation and monitoring is required to prevent this novel reassortant virus from becoming a new threat to public health.

Apart from sporadic cases, widespread infections and deaths of wild birds from HPAIV infection have never been reported before 2005. However, a HPAIV of the H5N1 subtype killed thousands of bar-headed geese (Anser indicus), great black-headed gulls (Larus ichthyaetus), and brown-headed gulls (Larus brunnicephalus) in Qinghai Lake, China during May 2005 9,12 . In 2006, a Qinghai-like Clade 2.2 virus re-emerged in Qinghai Lake and caused more infections in wild birds, including bar-headed geese and great black-headed gulls. The Qinghai-like Clade 2.2 virus was found to possess a high genetic relationship with viruses isolated from other countries on the migratory flyway of wild birds 4 , suggesting that the migration of wild birds played an important role in circulating H5N1 HPAIV viruses between the different avian populations. The virus eventually spread into other parts of Asia, as well as Europe and Africa, resulting in the deaths of millions of birds, and hundreds of cases in humans 9,[13][14][15]2 .
In January of 2015, the deaths of approximately 100 migratory wild birds (mainly whooper swans and pochards) in the Sanmenxia Reservoir Area of China were investigated. Located approximately 1150 kilometers (km) east of Qinghai Lake, this location is an important wintering habitat for whooper swans and various species of wild ducks including pochards (Aythya ferina) migrating to China from Mongolia and Siberia. The deaths sparked concerns of another HPAIV outbreak being spread by migratory wild birds.
To investigate this outbreak, samples were taken from sick and dead wild birds, and H5N1 virus was identified as the causative agent. In this study the gene evolution, phylogenetics, virulence to chickens and mammals, as well as receptor binding properties of these H5N1 viruses are presented, in addition to its potential spread along the migratory flyway of wild birds.

Results
Virus isolation. Samples were collected from sick or dead birds, in addition to the surrounding environment, in the Sanmenxia Reservoir Area of China on January 4 th -5 th , 2015. All samples were inoculated in 10-day-old, embryonated chicken eggs, which were found to have died within 24~41 hours after inoculation. Allantoic fluids tested positive by a hemagglutination (HA) assay with titers between 128-256 reciprocal dilutions. These suggested that this virus is highly pathogenic to poultry. H5N1 influenza virus was confirmed and Newcastle disease virus (NDV) was ruled out by HA inhibition (HI) assay and RT-PCR (specific primer sequences for both viruses are available from the corresponding author upon request). In addition, no bacteria were detected in the organs and tissues of the sick or dead birds when the samples were plated onto LB agar.
A total of 12 influenza A (H5N1) viruses were isolated, with ten isolates from whooper swans, one from a dead pochard, and one isolate from 12 environmental samples containing the feces of apparently healthy birds.
Genetic identity analysis. The H5 and N1 sequences of the 12 viruses were found to possess high genetic identities with each other. To avoid redundancy, six of 12 viruses (Tables 1 and 2) were selected to further analyze the gene evolution of the Sanmenxia H5N1 viruses isolated from wild birds. Genetic analysis showed that all eight gene segments of the novel H5N1 viruses shared between 99.3~100% nucleotide identity with each other.
Sequence alignments and BLAST search in the GISAID and NCBI Flu database showed that the genomes of the Sanmenxia H5N1 viruses showed the highest genetic homology to a previously described H5N1 virus (A/Alberta/01/2014, Alberta2014). Alberta2014 was responsible for causing the first human H5N1 infection in North America, and was likely an imported case from Asia 20 . The nucleotide identities of the novel virus isolates with Alberta2014 ranged between 98.7 to 99.6%, and protein identities ranged between 98.2~100%. These new isolates also possessed high nucleotide (91.6~98.7%) and amino acid (97.6~100%) identities to another mammalian H5N1 strain (A/tiger/Jiangsu/01/2013, Tiger2013), which was isolated from a tiger that had died from respiratory failure in 2013 21 (Table 1). Phylogenetic analysis. To further analyze the genetic evolution of these novel H5N1 viruses, phylogenetic analysis of the whole genomes was performed with reference H5N1 strains. The HA gene of the new isolates was found to belong to Clade 2.3.2.1c (Figs 1A and S1). Notably, the topology of Clade 2.3.2.1c formed a "ladder-like" phylogenetic structure, and H5N1 strains causing the Sanmenxia outbreak formed a novel independent "ladder" (Fig. 1A). The other distinct "ladder-like" branches include the 2009-2011 sub-clade circulating in Asia and Eastern Europe, and the 2012-2014 sub-clade circulating in Eastern Asia. Therefore, Clade 2.3.2.1c has been circulating in Eastern Asia since 2009, undergoing a lineage replacement every 2-3 years. However, the gap between the sub-clade 2.3.2.1 in 2008 and the recent 2015 sub-clade is quite large, as indicated by branch length (Fig. 1A), but it is currently unknown whether these variations in gene sequences influence viral antigenicity/pathogenicity. Consistent with results obtained from the comparison of sequence identities, isolates from the 2015 outbreak were most closely related to a H5 lineage circulating in China from 2013 in the HA phylogenetic   tree, including Alberta2014 and Tiger2013 (Fig. 1A). Similar topologies were also observed in the phylogenetic trees constructed using the NA, NP, MP, NS, PA, and PB1 gene sequences (Figs S2-S7).
The PB2 genes of the novel isolates and Alberta2014 were found to cluster together with previously circulating H9N2 viruses in China (Fig. 1B). Interestingly, the PB2 gene of Tiger2013 was also not derived from the H5 lineage (Fig. S8), but rather it belonged to an AIV gene pool that mainly existed in wild birds (Fig. S8), confirming the results of a previous study 21 . Therefore, the novel isolates from Sanmenxia and Alberta2014 belong to the same genotype, with the HA gene from Clade 2.3.2.1c, and the PB2 gene from H9N2 origin. The Tiger2013 isolate belongs to a different genotype, with the HA gene from Clade 2.3.2.1c, and the PB2 gene from an AIV gene pool.
Genetic signature analysis. The genetic signatures of the Sanmenxia Clade 2.3.2.1c-like H5N1 isolates were analyzed for their potential biological characteristics in order to lay the groundwork for future experiments regarding virus pathogenicity, host adaptation and drug sensitivity. The results showed that the novel isolates identified in this outbreak possessed multiple basic amino acids at the cleavage sites of HA, RERRRKR/G (Table 2), which is the typical molecular signature for HPAIVs 22 . In addition, the HA proteins of these isolates had the amino acids Q226 and G228 (H3 numbering) in the receptor-binding site. This suggested that these viruses favor binding to avian-like receptors, but the N158D and T160A substitutions in these viruses may enhance their capacity to also bind human-type receptors [23][24][25] . Amino acids Q591, E627 and D701 were observed in the PB2 protein, suggesting that these novel H5N1 isolates are not yet fully adapted to infect mammals.
In light of the close genetic relationship between the H5N1 viruses isolated from wild birds and mammals, the genetic signatures of these viruses were also compared with Alberta2014 and Tiger2013. Interestingly, there are only five amino acid mutations in the HA proteins between H5N1 virus isolates from wild birds and human/tiger (Tables 2 and 3). The D225R substitution was found only in Alberta2014, which may influence viral receptor binding properties, according to the results of a previous study 26 .
Furthermore, the NA stalk of these novel H5N1 viruses possessed a 20 amino acid deletion (positions 49 to 68). Past studies had shown that AIVs with this deletion demonstrated better adaptation to terrestrial birds after they were introduced from wild birds, and showed increased virulence in mammals 27,28 . Moreover, these H5N1 isolates had a C-terminal (aa 58-90) truncated PB1-F2 protein with only 57 amino acids, which has also been shown to contribute to their virulence in mammals 29,30 . Hence, the novel H5N1 isolates may be highly pathogenic to mammals based on the molecular characteristics on NA stalk and PB1-F2, even though they possess the avian adaptation residues with Q591, E627 and D701 on the PB2 protein.
The NA-inhibitor resistance mutation H274Y was not found in the NA antigen of these novel H5N1 reassortants. However, the V27I substitution was observed in the M2 protein, indicating that these H5N1 isolates are sensitive to NA inhibitors, but may be resistant to M2 inhibitors 2 .

Pathogenicity in birds and mice.
To further characterize the virulence of these new H5N1 viruses, one representative strain (whooper swan/HN/SMX4/2015) was selected for studies in chickens and mice. All ten specific-pathogen free (SPF) chickens inoculated with whooper swan/HN/SMX4/2015 died 1~3 days post-infection (d.p.i.) and the intravenous pathogenicity index (IVPI) was 2.79 (  In addition, the novel H5N1 virus was found to cause severe pathological lesions in the non-surviving wild whooper swans and the pochard. Viral encephalitis, severe lung damage with congestion and hemorrhage, and necroticans enteritis can be readily observed from the histopathology results (Fig. 2).
Furthermore, whooper swan/HN/SMX4/2015 was also shown to be highly pathogenic in mice, with a median lethal dose (MLD 50 ) of 10 3.0 TCID 50 (median tissue culture infective dose) ( Fig. 3 and Table 4). The virus not only replicated in the lungs during 3~7 d.p.i. with high lung virus titers (LVTs), but was found to be present in the brains starting at 5 d.p.i. Interestingly, LVTs peaked at 5 d.p.i. before gradually decreasing, whereas the peak of virus replication in the brain was at 7 d.p.i. (Table 4). This suggests that the virus first replicated in the target organs (lungs), before bypassing the blood brain barrier to infect the brain.

Receptor binding properties.
In light of the highly pathogenic nature of the new H5N1 isolates, as well as the high genetic identities between these viruses and H5N1 of human origin, potential risks of these novel H5N1 viruses to humans was evaluated by a solid-phase receptor binding assay. The new H5N1 isolates were found to bind preferentially to the avian α -2,3-linked (3′ SLNLN) receptor over the human α -2,6-linked (6′ SLNLN) sialylglycan receptor (Fig. 4), suggesting that these novel isolates currently have a low risk for human infections.

Discussion
Genetic and phylogenetic analyses showed that this outbreak in whooper swans, pochards, and other wild ducks in the Sanmenxia Reservoir Area was caused by a H5N1 virus. These viruses were shown to be novel reassortants, possessing a Clade 2.3.2.1c HA gene and an H9N2-derived PB2 gene, and had the highest genetic homology with Alberta2014 and Tiger2013. The HA protein of these novel isolates also have the N158D and T160A amino acid substitutions, which were associated with at least two confirmed mammalian infections in the past 23,24 and highlights the potential risk to public health.
The novel isolates were found to be highly pathogenic to chickens and mice, and virus isolated from mouse brains after challenge. It should be noted that meningoencephalitis was also observed in the fatal human case with Alberta2014, an unusual outcome for infections with H5N1 HPAIV in humans 20 . Neurological symptoms were also noted in the non-surviving tiger infected with Tiger2013, with the heart, liver, spleen, lung, kidney, aquae pericardii, and cerebrospinal fluid all positive for H5N1 virus as detected by real-time RT-PCR 21 . This suggests that the novel Sanmenxia Clade 2.3.2.1c-like H5N1 viruses possesses tropism for the nervous system in several mammal species, and could pose a significant threat to humans if these viruses develop the ability to bind human-type receptors more effectively.
No domestic poultry breeding farms exist within 3 km of the Sanmenxia Reservoir Area and the majority of the Clade 2.3.2.1c strains are isolated from migratory birds (Fig. 1A). Therefore, the Sanmenxia Clade 2.3.2.1c-like H5N1 virus is likely spread by the long-distance migration of wild birds, especially ducks, which might have been infected by the virus from domestic poultry during stopover sites in their migration 6,10,11,31 . Whooper swans are known to breed in northern Eurasia, and winter in Europe and eastern Asia, such as China, Korea, and Japan 32 . There are approximately 20 000 whooper swans wintering in China, and more than half of these animals winter in the Sanmenxia Reservoir Area along with various ducks from the East Asian/Australasian and Central Asian flyways 33 . These birds arrive in the Sanmenxia Reservoir Area during early October for wintering, and then fly back to Mongolia and Siberia for breeding next spring (Figs 5 and 6).
It is believed that overlapping migratory flyways help circulate H5 HPAIVs amongst different bird species, and allow the spread of the virus across continents 8 . Recently sequential outbreaks of the H5N8 virus in domestic poultry in China, Korea, and Japan during 2010-2015 34,35 , Europe in 2014-2015 36 , and North America in 2014-2015 4,37,38 were considered to have occurred due to waterfowl migration 36,37,39 . Therefore, the migration of wild birds plays an important role in the transmission and spread of H5 HPAIVs, posing a severe risk to animal and human health. Further investigation is required to monitor the evolution and transmission of these novel Sanmenxia Clade 2.3.2.1c-like H5N1 isolates, in order to

Materials and Methods
Virus isolation and gene sequencing. The oropharyngeal and cloacal swabs, tissues and organs of the sick or dead birds, and feces from apparently healthy birds were collected for virus isolation. Swab samples and tissues were maintained at 4 °C in viral-transport medium and sample tubes, respectively. Before inoculation, swab samples with viral-transport medium were thoroughly mixed. The tissues and organs were homogenized using a Qiagen TissueLyser II machine (30 cycles/s; 4 min) in 1 ml of cold PBS under sterile conditions. The solid debris in the swab and tissue samples were pelleted by  Histopathologic analysis. The brain, lung, small intestine, spleen, kidney, and liver of non-surviving wild birds were fixed in 10% phosphate-buffered formalin, embedded in paraffin, cut into 5-mm-thick serial sections, and stained with hematoxylin and eosin (H&E) for histopathological evaluation as previously described 46 .

Studies in chickens.
To determine the pathogenicity of these novel virus isolates, the IVPI was determined according to the recommendations of the Manual of Diagnostic Tests and Vaccines for Terrestrial Animals 2014 47 . Fresh infective allantoic fluid with a HA titer of 128 was diluted 1/10 in sterile isotonic saline. 0.1 ml of the diluted virus was used for an intravenous challenge in ten 6-week-old SPF White Leghorn chickens.
Birds were examined at 24-hour intervals for 10 days. At each observation, each bird was scored 0 if normal, 1 if sick, 2 if severely sick, and 3 if dead. The judgment of sick and severely sick birds is a subjective clinical assessment. Typically, 'sick' birds demonstrate one of the following signs and those that are 'severely sick' demonstrate more than one of the following signs: respiratory involvement, depression, diarrhea, cyanosis of the exposed skin or wattles, edema of the face and/or head, nervous signs. The IVPI is the mean score per bird per observation over the 10-day period.
Studies in mice. Fourteen 6-to 8-week-old (~17 g) female BALB/c mice (Vital River Laboratory, Beijing, China) from each group were anesthetized with carbon dioxide (CO 2 ) and inoculated intranasally (i.n.) with 50 μ l of the 10-fold dilutions of virus in phosphate-buffered saline (PBS). Each group contains five mice and the MLD 50 value was determined by the Reed and Muench method 48 , and expressed as the TCID 50 value corresponding to 1 MLD 50 .
Nine mice were inoculated i.n. with of 50 μ l of 10 4.5 TCID 50 virus diluted in phosphate-buffered saline (PBS). Mock-infected control animals were inoculated i.n. with 50 μ l PBS. Three mice from each group were euthanized at 3, 5, and 7 d.p.i., in which the brains and lungs were collected and homogenized using a QIAGEN TissueLyser II machine (30 cycles/s, 4 min) in 1 ml of cold PBS under sterile conditions, respectively. Solid debris was then pelleted by centrifugation at 5,000 × g for 10 min, and the homogenates were used for virus titrations in MDCK cells. All of the mice were monitored daily for general behavior and clinical signs, including food intake, body weight, inactivity, and mortality for 14 days. Any mouse that lost ≥ 25% of its preinoculation body weight was euthanized.