Genetic and biological characterization of H9N2 avian influenza viruses isolated in China from 2011 to 2014

The genotypes of the H9N2 avian influenza viruses have changed since 2013 when almost all H9N2 viruses circulating in chickens in China were genotype 57 (G57) with the fittest lineage of each gene. To characterize the H9N2 variant viruses from 2011 to 2014, 28 H9N2 influenza viruses were isolated from live poultry markets in China from 2011–2014 and were analyzed by genetic and biological characterization. Our findings showed that 16 residues that changed antigenicity, two potential N-linked glycosylation sites, and one amino acid in the receptor binding site of the HA protein changed significantly from 2011–2014. Moreover, the HA and NA genes in the phylogenetic tree were mainly clustered into two independent branches, A and B, based on the year of isolation. H9N2 virus internal genes were related to those from the human-infected avian influenza viruses H5N1, H7N9, and H10N8. In particular, the NS gene in the phylogenetic tree revealed genetic divergence of the virus gene into three branches labeled A, B, and C, which were related to the H9N2, H10N8, and H7N9 viruses, respectively. Additionally, the isolates also showed varying levels of infection and airborne transmission. These results indicated that the H9N2 virus had undergone an adaptive evolution and variation from 2011–2014.

To understand the evolutionary characteristics of the H9N2 avian influenza viruses from 2010-2014, molecular changes, the antigenic residues, potential glycosylation sites, and receptor-binding sites in HA from 28 H9N2 strains isolated from 2011 to 2014 were analyzed. Moreover, the infection and airborne transmission from eight H9N2 isolates in this study were identified.

Ethics statement
The specific-pathogen-free (SPF) chickens and chicken embryos used in this study were purchased from Beijing Center for Laboratory Animals. Procedures involving the care and use of animals were approved by the Jiangsu Administrative Committee for Laboratory Animals (permission number SYXK-SU-2007-0005) and complied with the Jiangsu Laboratory Animal Welfare and Ethics guidelines of the Jiangsu Administrative Committee of Laboratory Animals.

Sample collection and viral isolation
From 2011 to 2014, tracheal or cloacal swabs collected from live poultry markets in China, including Jiangsu, Anhui, Heilongjiang, and Fujian, were placed into phosphate buffered saline (PBS) containing 2000 units/mL penicillin, 2 mg/mL streptomycin, 50 μg/mL gentamycin, and 1000 units/mL mycostatin. After freezing, the supernatants were collected through a 0.22 μm pore size filter after centrifugation at 1000 g for 10 min at 4˚C. Next, 200 μL of the supernatant was inoculated in the allantoic cavities of 10-day-old SPF embryonated chicken eggs. After 72 h of incubation at 37˚C, the viruses were tested for HA activity. The isolated viruses were identified as H9 subtype by hemagglutination inhibition (HI) assay using antisera against H9 subtype, H5 subtype avian influenza viruses (AIV), and Newcastle disease viruses (NDV). The 50% egg infective doses (EID 50 ) for the isolates were determined by propagating serial dilutions of viruses in eggs (0.2 mL per embryo) and calculated as described by Reed & Muench [23]. The isolate HI assays were performed with an initial dilution of 1:2 as previously described [24].

Genetic and phylogenetic analysis
Total viral RNA was extracted from allantoic fluids infected with purified isolates using the TRIZol Reagent (Invitrogen Life Technologies, Carlsbad, CA, USA). Viral RNA was reversetranscribed (PrimeScriptIIReverse Transcriptase, Takara Biotechnology, Dalian, China) into cDNA using 12-base universal primers for influenza A viruses (U12 A/G: AGCG/AAAAGC AGG). The gene fragments were amplified by polymerase chain reaction (PrimeSTAR Max DNA Polymerase, Takara Biotechnology, Dalian, China) after reverse transcription using specific primers, sequenced, and assembled by Takara Biotechnology (Dalian, China) [25,26]. The sequences were edited using the Lasergene software package (DNASTRA Inc., Madison, WI, USA). Phylogenetic trees were generated using the MEGA6.0 software suite. The neighbor-joining method with 1000 bootstrap replicates was performed for the phylogenetic and molecular evolutionary analysis. The phylogenetic lineages of all eight genes were defined by gene phylogeny as previously described [26].

Key site analysis
To investigate the antigenic drift, the mutation levels for each of 18 reported antigenic residues from 28 isolates were quantified by comparison with 1081 full-length H9N2 HA sequences deposited in GenBank from 2010 to 2015 (date to Jan 1, 2016). The three commercial vaccine strains SH/F/98, SD/6/96, and GD/SS/94 were used as controls.
The potential N-linked glycosylation sites (PNLGSs) in HA were predicted using the NetN-Glyc 1.0 Server online software (http://www.cbs.dtu.dk/services/NetNGlyc/). Some key residues from all gene fragments were also investigated.

Antigenic analysis of the isolates
To investigate the antigenic relationship between the H9N2 isolates and two commonly used vaccine strains (SH/F/98 and GD/SS/94), the nine H9N2 isolates (SD/C9QH/11, JS/JT12/11, JS/YZ640/12, JS/YZ618/12, JS/TM58/13, JS/TM59/13, JS/JT95/13, AH/WB/14 and JS/TM71/ 1) and two commercial vaccine strains (SH/F/98 and GD/SS/94) were selected according to the genotype of the isolates and the year of isolation. The antiserum of each strain was made, then the HI titer of each straun were detected as following description. Briefly, 3-week-old SPF chickens were immunized twice by subcutaneous injection of 0.3 mL of oil-emulsion of inactive whole virus vaccine of the indicated virus, which was inactivated by adding 0.2% formalin (v/v) for 24h at 37˚C. The antisera were collected from vaccinated SPF chickens in two weeks after the booster vaccination and used to characterize the antigenicity of the 28 H9N2 isolates and the two commercial vaccine strains. The HI assay was expressed as the reciprocal of the highest serum dilution in which HA was inhibited. Then, antigenic cartography was performed with the Antigen Map program (http://sysbio.cvm.msstate.edu/AntigenMap), which uses matrix completion multidimensional scaling to map HI titers in two dimensions [27].

Replication and transmission of H9N2 influenza virus isolates in chickens
To study the viral replication and pathogenicity of the H9N2 influenza virus in chicken, groups of six 3-week-old SPF chickens were orally, intranasally, or intratracheally inoculated with 0.2 mL of 10 6 EID 50 of each virus or a PBS control. Tissue samples (trachea and lung) from the inoculated chickens were collected at 3 days and 5 days post-inoculation (dpi). Briefly, three SPF chickens were euthanatized using CO 2 asphyxiation at designated times, and half of the tissues were harvested, washed, and ground into 20% (w/v) suspension in 1 mL PBS. Virus titers in the trachea and lung were determined in 10-day-old SPF embryonated chicken eggs.
The other half of the tissues were fixed with 10% neutral buffered formalin, embedded in paraffin, sectioned at 5 mm, and processed for staining with hematoxylin and eosin for histopathological examination. These experiments were repeated three times.
For studying the viral transmission, fifteen 3-week-old SPF chickens were divided into five groups: (i) inoculated group (three chickens), (ii) direct contact group (three chickens), (iii) airborne contact group (three chickens), (iv) PBS control group (three chickens), (v) inoculated group for sera (three chickens). To prevent cross-contamination, the chickens from the experimental groups (inoculated group and direct contact group were in the same poultry isolator) were housed in a separated poultry isolator after fumigation with the formaldehydepotassium permanganate. The infected group was inoculated orally, intranasally, and intratracheally with 10 6 EID 50 of the indicated virus. The airborne contact group was placed in a poultry isolator adjacent to the infected group with a distance of 50 cm between poultry isolators. At day 3 and 5 post-inoculation, tracheal and cloacal swabs from chickens were collected in 1 mL of PBS containing antibiotics and, following one freeze-thaw cycle, were centrifuged at 3000 rpm for 10 min. Of the resulting supernatant, 0.2 mL were taken, and the EID 50 titers of the tracheal and cloacal swabs collected from the indicated passaged virus were determined by serial titration of viruses in SPF embryonated chicken eggs using the Reed & Muench method [23]. Sera collected from each experimental bird at 21 dpi were tested by HI assay to check the antibody in the serum against the indicated viruses [28]. The experiment was repeated three times.

Viral isolation and identification
Allantoic fluids infected with tracheal or cloacal swabs taken from chickens were harvested from embryonated SPF chicken eggs. The viruses were identified as H9 subtype by HI assay using chicken H9 antiserum and were identified as N2 via PCR using specific primers. The isolated viruses did not react with H5 or Newcastle disease virus antiserum. Twenty-eight H9N2 strains from chickens in live poultry markets were isolated in this study. HA genes from the 28 H9N2 isolates and seven other genes (genes neuraminidase (NA), polymerase basic 2 (PB2), polymerase basic 1 (PB1), polymerase acidic (PA), nucleoprotein (NP), matrix (M), nonstructural (NS)) from the 9 representative isolates for the 28 H9N2 isolates were sequenced and submitted to GenBank under the accession numbers MG280611-MG280630 and MG280639-MG280709 (Table 1).

Phylogenetic analysis
The phylogenetic analysis included the HA nucleotide sequences from the 28 field isolates and the other 42 reference strains, which include 18 reference strains from different lineages, three vaccine strains, three G57 reference strains, and 18 field strains isolated from 2010-2015. The results showed that HA genes from all 28 H9N2 isolates were clustered into the BJ/94-like lineage. According to the year of isolation, the viruses in the phylogenetic tree could mainly be clustered into two independent branches labeled A ( (Fig 1a).

Antigenic cross-reactivity of the isolates
Results from a reciprocal HI assay demonstrated that antisera against the SD/C9QH/11, JS/ TM58/13, JS/JT95/13, or AH/WB/14 strains had low reactivity with most of the isolates and the two vaccine strains, SH/F/98 and GD/SS/94. The two vaccine strains and most of the isolates reacted well with the antisera against JS/TM71/14 or JS/TM59/13 ( Table 2). The antiserum against JS/JT12/11 only has reactivity against the vaccine strain SH/F/98 (Table 2). Further AntigenMap analysis showed that all 28 field isolates and the two vaccine strains, SH/ F/98 and GD/SS/94, were divided into five independent antigenic groups 1-5 (Fig 4). Group-1

Molecular analysis
The mutations for each of the 18 reported antigenic residues were analyzed by comparison with 1081 full-length H9N2 HA sequences deposited in GenBank (date to Jan 1, 2016). In comparison to three commercial vaccine strains (SH/F/98, SD/6/96, and GD/SS/94), the 18 antigenic residues in the HA protein of the isolates were changed significantly from 2011-2014, including H66Q, G90E, S127R, S145N, D153G, Q164K/R, N167G/S/K, A168D/N, E181G, A198T/V, T200R, N201D, D216E, T220I, Q235M, R254K, N256D, and S283R ( Fig  5). In addition, the antigenic residue mutations S145N and T220V, only appeared in moderate isolates, while the mutations S145D and T220I, were dominant mutations in the published sequences in GenBank (Fig 5). The PNLGSs in HA from the 28 isolates and three vaccine strains were predicted. The results showed that between 7 and 9 PNLGSs were found in these strains,  Compared with the three vaccine strains, the receptor-binding sites (RBS) with amino acids Y109, W161, T163, A191, L202, and Y203 were conserved among the 28 isolates, though a mutation occurred at position A198. Among the 28 isolates, 11 strains possessed an A198V Table 2. Cross-reactivity between H9N2 influenza viruses in hemagglutination inhibition assays. mutation, and 8 strains possessed an A198T mutation. The 198T position began to be dominant in 2013, and the proportion of 198V or 198A decreased yearly since 2012 or 2011, respectively (Fig 6b). The amino acid sequence, 146GTSKA150, in the left edge of the receptor binding pocket was conserved among 28 isolates; however, compared with the amino acid sequence, 232NGQQGR237, in the right edge of the receptor binding pocket in the three vaccine strains, the amino acid sequence, 232NGLQGR237, was found in 10 isolates, and the amino acid sequence, 232NGLMGR237, was observed in 18 isolates. The proportion of 234LQ235 sequences decreased rapidly since 2011; whereas, 234LM235 became predominant since 2013 (Fig 6c). Additionally, the 333PSRSSR#GLF341 cleavage site (H9 numbering) was observed in 23 field isolates, and 333PSKSSR#GLF341 was observed in five isolates. The 334SR335 sequence was predominant in published sequences from 2011-2015 (Fig 6 and Table 4).  (Table 5). Some critical residues in internal genes contributing pathogenicity or infectivity were analyzed, and the results showed that 627E and 701D in PB2 were observed among nine isolates (701N in PB2 was observed only in the virus JS/TM59/13), which are characteristics of lowpathogenic AIVs. In contrast, D253N and Q591K in PB2 were not observed, which could increase AIV pathogenicity [30]. 588V in PB2 was observed in the JS/TM59/13 virus, which could increase the AIV pathogenicity in mammals [31]. In this study, 95K, 224N, and 242N in the M1 protein were observed among 9 isolates, 37A in the M1 protein was observed in eight isolates, and 21G in the M2 protein was observed in 9 isolates, all of which could increase AIV infections [32]. 227K in the NS protein was observed in nine isolates, which also improves the pathogenicity of virus [33]. The nine representative isolates all contained amantadine-resistance mutations at position 13 (proline) in the PB1 protein, position 15 (isoleucine) in the M1 protein, position 55 (phenylalanine) in the M2 protein, and adamantine-resistance mutation at positions 31 (asparagine) in the M2 protein [14,34,35]. Both HA 381K and PA 627L in one H9N2 AIV were found to be important for airborne transmissibility among chickens [36], and all strains, except for the SH/F/98 and FJ/SN02/14 strains, possessed both HA 381K and PA 627L. These results from the molecular analysis showed that the field isolates had evolved in the direction of stronger infectivity, pathogenicity, drug resistance, and transmission.

Replication and transmission of H9N2 influenza virus isolates in chickens
To investigate the replication and transmission of H9N2 isolates in chickens, EID 50 counts from oropharyngeal and cloacal swabs, tracheal and lung homogenates, and tracheal and lung histological specimens were conducted on 3 or 5 dpi. The EID 50 results from the tracheal and lung homogenates on 3 and 5 dpi showed that all isolates replicated better in tracheae than in lungs. In particular, the JS/JT12/11, ZJ/618/12, JS/YZ640/12, JS/TM58/13, and JS/JT95/13 strains reached higher EID 50 on 3 and 5 dpi than other strains in the tracheae. The AH/WB/14 and JS/TM71/14 strains reached lower EID 50 on 5 dpi in the tracheas than on 3 dpi, while no SD/C9QH/11 virus was detected on 5 dpi in the trachea. Only the ZJ/618/12 strain was detected in lung homogenates on 3 and 5 dpi, and the JS/TM58/13 or JS/JT95/13 viruses were detected in lung homogenates on 3 dpi or 5 dpi, respectively, while little or no virus was detected in the lung homogenates on 3 and 5 dpi in the groups infected with the rest of the isolates (Fig 7b). These results showed that the isolates could replicate better in the upper respiratory tract than in the lower respiratory tract. The results from the viral pathogenicity testing in chickens showed that different degrees of lesions were observed in the lungs and tracheae of chickens infected with the isolates. Severe lesions were still observed in trachea infected with the ZJ/618/12, JS/YZ640/12, and JS/TM58/13 strains on 5 dpi, while more severe lesions were observed in tracheae infected with the JS/JT12/11, SD/C9QH/11, ZJ/618/12, JS/JT95/13, JS/ TM71/13, and AH/WB/14 strains on 3 dpi than on 5 dpi. The isolates, except for JS/TM58/13, all possess different tissue tropism in the trachea or lungs. Interestingly, although the EID 50 for most of the isolates from the lungs was low, lesions in the lungs were severe on 3 and 5 dpi (Fig  7a and 7c). The results from the airborne transmission experiments showed distinct airborne transmission among isolates, which may be related to a better ability to shed viruses in the trachea than in the cloaca. The exposed chickens in the JS/JT12/11, ZJ/618/12, JS/YZ640/12, JS/ JT95/13, and JS/TM71/14 groups did not shed viruses in the cloaca, and the chickens in the other exposed groups only shed limited viruses. Overall, compared to the trachea, the exposed chickens shed limited or no viruses in the cloaca (Table 6).

Discussion
H9N2 subtype avian influenza viruses have caused great economic losses to the poultry industry. The H9N2 virus is recognized as a donor for internal genes to other AIV subtypes, Genetic and biological characterization of H9N2 avian influenza viruses including the H5N1, H7N9, and H10N8 viruses, which can infect humans, causing a potential threat to public health. The H9N2 virus contains a negative, single-stranded RNA molecule that is susceptible to mutation and reassortment during the replication process, and its surface and internal proteins are divided into distinct lineages [26]. Based on HA, the H9N2 AIVs in Eurasia include the following three lineages: BJ/94-like, G1-like, and Y439/97-like. BJ/94-like and G1-like viruses have become the predominant strains in mainland China since the mid-1990s [37]. However, from 2010-2013, a fittest genotype (G57) containing a PB2 gene from dk/ST/04, which is clustered closely to an H7N7 human isolate, A/Netherlands/219/2003 [26], an M gene from the quail origin qa/HK/G1/97, which was closely related to the emerging H7N9 virus [16,32], an HA gene from ck/JS/00, and the remaining genes from SH/F/98 emerged with changed antigenicity and improved adaptability in chickens [16]. Currently, the G57 virus is the predominant strain in China since 2013 [16]. In this study, 28 H9N2 strains were isolated from live poultry markets in China, including Jiangsu, Anhui, Heilongjiang, and Fujian, from 2011 to 2014. Three G57-type H9 viruses https://doi.org/10.1371/journal.pone.0199260.t004 Genetic and biological characterization of H9N2 avian influenza viruses  isolated in different years (ZJ/HJ/07, SD/sd01/10, and WZ/614/13) and some field strains isolated in recent years were selected for phylogenetic analysis. The HA segment of the 28 isolates belonged to G57-like viruses. Additionally, the locations of all the strains in the phylogenetic tree are closely related to the year of isolation, and the later the isolation, the farther they are located from the original G57 ZJ/HJ/07 strain in the phylogenetic tree. Specifically, the NA gene has gradually evolved into an undefined lineage. In comparison to other lineages of each gene, the fittest and most advantaged one of each gene may help H9N2 viruses to survive easily in hosts. Additionally, our findings also showed that 16 antigenic residues, two PNLGSs, and one RBS residue changed significantly from 2011-2013, which might be related to the emergence of the G57 lineage. It is generally believed that the RSSR sequence at the HA cleavage site is a characteristic of low-pathogenic AIVs [38]. In this study, we found that the KSSR sequence at the HA cleavage site is also characteristic of low-pathogenic AIVs. The R or K residue at position 335, both of which are alkaline amino acids, may be involved in HA cleavage. Changes in the HA receptor-binding site and its nearby potential glycosylation sites also affect the binding of the virus to host cells and therefore the pathogenicity of the influenza viruses [39]. Although some PNLGSs in isolates potentially have HA sequences increased or decreased from 2011-2013, most published sequences still possessed eight PNLGSs, suggesting that the eight glycosylation sites may be more stable in current H9N2 viruses. The changes in the PNLGSs are closely related to the year of isolation, which is similar to the antigenic residues. Monitoring PNLGSs sites and antigenic residues in circulating viral isolates may facilitate the prediction of mutants with epidemic potential. The virulence, infectivity, and transmission of the H9N2 virus are affected by multiple factors, particularly genetic factors. In this study, most isolates possessed residues in critical sites that increase the infectivity and transmission. Subsequently, animal experiments showed that the isolates caused different degrees of histopathological changes concerning host tissue tropism and better replication in the upper respiratory tract. Better airborne transmission and replication in the host can create an opportunity for H9N2 viruses to survive easily.
Currently, H9N2 viruses are circulating in China without an effective vaccination strategy. H9N2 viruses escape selection pressure from virus-host interactions by evolution, which may have directly resulted in the emergence of an epidemic G57 virus. To understand the evolution of the H9N2 virus from 2011-2013 when the G57 virus became predominant, we performed a molecular analysis of H9N2 isolates in this study. From 2011-2013 changes in antigenic Table 6. Airborne transmission study of 9 representative isolates. The number of chickens shedding/total number of chickens.

A Strains
SD/C9QH/11 group JS/JT12/11 group JS/YZ618/12 group JS/ YZ640/12 group   Trachea  Cloacal  Trachea  Cloacal  Trachea  Cloacal  Trachea  Cloacal   Virus shedding on day  3  5  3  5  3  5  3  5  3  5  3  5  3  5  3  residues, PNLGSs, and some critical residues might have contributed to the increased H9N2 virus survival and evolution. These changes might help the G57 virus to become the predominant strain and could potentially cause enhanced infectivity and transmission by the circulating H9N2 virus.

Conclusions
In summary, our findings suggest that the H9N2 virus surface glycoproteins have undergone evolution and variation from 2011-2014, which might be related to the G57 virus outbreaks and emergence since 2013. The H9N2 virus internal genes were closely related to those from the human-infected avian influenza viruses H5N1, H7N9, and the emerging influenza virus H10N8. In particular, some NS genes from H9N2 viruses were closely related to the H10N8 and H7N9 viruses, respectively.