Highly Pathogenic H5 Influenza Viruses Isolated between 2016 and 2017 in Vietnamese Live Bird Markets

Routine surveillance in live poultry markets in the northern regions of Vietnam from 2016 to 2017 resulted in the isolation of 27 highly pathogenic avian H5N1 and H5N6 viruses of 3 different clades (2.3.2.1c, 2.3.4.4f, and 2.3.4.4g). Sequence and phylogenetic analysis of these viruses revealed reassortment with various subtypes of low pathogenic avian influenza viruses. Deep-sequencing identified minor viral subpopulations encoding variants that may affect pathogenicity and sensitivity to antiviral drugs. Interestingly, mice infected with two different clade 2.3.2.1c viruses lost body weight rapidly and succumbed to virus infection, whereas mice infected with clade 2.3.4.4f or 2.3.4.4g viruses experienced non-lethal infections.


Introduction
Highly pathogenic H5N1 influenza viruses of the A/goose/Guangdong/1/96-lineage emerged in 1996 and have spread throughout Asia, Europe, the Middle East, and parts of Africa and North America, causing more than 890 laboratory-confirmed human infections with a case fatality rate of approximately 50% (https://www.cdc.gov/flu/avianflu/ reported-human-infections.htm; accessed on 4 April 2023). These viruses have now evolved into 10 genetically distinct clades (clades 0-9) with multiple subclades.
In addition to their threat to public health, highly pathogenic H5 viruses have caused huge economic losses in the poultry industry. In Vietnam, highly pathogenic H5N1 influenza viruses were first reported in 2001 and were enzootic in poultry populations by the end of 2004. The early outbreaks in Vietnamese poultry populations were primarily caused by H5N1 viruses of clade 1. To control the spread of these viruses, mass poultry vaccination with a vaccine against a clade 1 H5N1 influenza virus was implemented in 2005 [1]. This vaccination campaign did not completely eradicate H5N1 viruses in Vietnam, and viruses of clade 1 continued to circulate in the southern part of the country where they evolved into subclades 1.  [2]). In 2010, clade 2.3.2 viruses were re-introduced into

Virus Isolation and Identification
In 2016-2017, 1765 oropharyngeal and cloacal swab samples were collected from apparently healthy birds in live bird markets in Northern Vietnam by the National Institute of Health and Epidemiology of Vietnam as part of routine surveillance activities. The samples were placed in a transport medium (DMEM containing 0.15% BSA, 100 IU/mL penicillin-streptomycin, 0.5 µg/mL amphotericin B, 100 µg/mL gentamicin, 20 µg/mL ciprofloxacin, and 0.02 M HEPES) for further processing.
First, we pooled the samples (up to 10 samples collected from the same species on the same day in the same location) into several batches and inoculated them into 9-11-day-old specific pathogen-free (SPF) eggs that were incubated at 35 • C for 24-48 h. The allantoic fluid was then tested for hemagglutination activity. For hemagglutination-positive batches, individual samples were inoculated into two SPF eggs each and tested for hemagglutination activity. Next, RNA was extracted from the hemagglutination-positive samples using the MagMAX TM -96 Viral RNA Isolation Kit (ThermoFisher, Waltham, MA, USA), and one-step reverse transcriptase polymerase chain reaction (RT-PCR) was performed using the Superscript III high-fidelity RT-PCR Kit (ThermoFisher, Waltham, MA, USA) and oligonucleotides to the conserved regions of the viral matrix (M) gene. The sequences of the primers are as follows: M-73F, 5 -GTCAGGCCCCCTCAAAGC; M-242R, 5 -CGTCTACGCTGCAGTCC. For influenza-positive samples, portions of the hemagglutinin (HA) and neuraminidase (NA) genes were Sanger-sequenced to determine the subtype.

Viral Genomic Sequence Analysis
For all highly pathogenic H5 viruses, the complete genomic sequence was determined with RT-PCR-amplifying extracted RNA with a mixture of oligonucleotides F1 (5 -GTTACGCGCCAGCAAAAGCAGG), F2 (5 -GTTACGCGCCAGCGAAAGCAGG), and R1 (5 -GTTACGCGCCAGTAGAAACAAGG). The amplified double-stranded cDNAs were purified with 0.45× volume of AMPure XP beads (Beckman Coulter) and fragmented with acoustic shearing (0.5-1 µg) on a Bioruptor Pico sonicator (Diagenode) with an average fragment size of 150 bp. Library preparation for next-generation sequencing used end repair, phosphorylation, A-tailing, and adaptor ligation NEBNext DNA library prep modules for Illumina (New England Biolabs), which was followed by additional PCR amplification. The Illumina MiSeq demultiplexed reads were processed and assembled using the Iterative Refinement Meta-assembler (IRMA v. 1.0.2) [18]. The internal gene sequences were obtained from the primary assembly, whereas the HA and NA sequences were obtained from the secondary assembly. We used the default IRMA FLU parameters with the following exceptions: LABEL was used for read sorting, the residual assembly factor was set to 400 for the secondary assembly, and reference elongation was prevented. All sequence data were preprocessed, the IRMA was run, and downstream analysis was conducted using version v0.9.9 of a snakemake workflow available at https://github.com/ IRI-UW-Bioinformatics/flu-ngs/releases/tag/v0.9.9 (accessed on 28 June 2022). Consensus sequences were generated with the IRMA.

Phylogenetic Analysis
Nucleotide sequences of the H5Nx (for HA), HxN1 (for N1-NA), HxN6 (for N6-NA), and HxNx (for PB2 and NS genes) viruses were downloaded from the NCBI Influenza Virus Database (www.ncbi.nlm.nih.gov) and the Global Initiative on Sharing Avian Influenza Data (GISAID; www.gisaid.org; Supplemental Table S1). The sequences were aligned using MAFFT (https://mafft.cbrc.jp/alignment/software/). Phylogenetic analysis was performed using the MEGA 7.0 software package, implementing the Tamura-Nei nucleotide substitution model and the maximum likelihood method. The tree topology was evaluated using 1000 bootstrap analyses.

Virus Stock Generation and Titration
The isolated, highly pathogenic H5 viruses were inoculated into three SPF eggs each (100 µL of 10 4 -10 5 -fold dilutions of virus). Following incubation at 35 • C for 24-48 h, the allantoic fluids were collected and combined. The stock virus titers were determined by performing plaque assays in Madin-Darby canine kidney (MDCK) cells.

Pathogenicity Studies in Mice
Six-week-old female BALB/c mice (Jackson Laboratory, Bar Harbor, ME, USA) were anesthetized with isoflurane and inoculated intranasally with 10 4 plaque-forming units (PFU) of each virus in a volume of 50 µL. Control mice were inoculated with 50 µL of PBS. Body weight and mortality were monitored daily for 14 days. Infected mice were euthanized if they lost more than 25% of their initial body weight.

Safety Statement
The isolation of the surveillance samples was conducted in an enhanced biosafety level 3 (BSL 3+) laboratory at the University of Wisconsin-Madison. The RNA inactivation protocol was approved by the University of Wisconsin-Madison's Institutional Biosafety Committee (IBC) after conducting a risk assessment in the Office of Biosafety. All experiments were approved by the University of Wisconsin-Madison's IBC. This manuscript was reviewed by the University of Wisconsin-Madison Dual Use Research of Concern (DURC) Subcommittee. This review was conducted in accordance with the United States Government September 2014 DURC Policy. The DURC Subcommittee concluded that the studies described herein do not meet the criteria for Dual Use Research of Concern (DURC).

Isolation and Identification of Avian Influenza Viruses from Birds in Vietnam in 2016-2017
We performed routine surveillance in live bird markets in several locations in northern Vietnam ( Figure 1) by collecting oropharyngeal and cloacal swab samples from apparently healthy chickens, Muscovy ducks, and other ducks. In total, 1765 samples were pooled into batches of up to 10 samples (collected on the same day, in the same location, and from the same species), amplified in SPF-embryonated chicken eggs, and tested for hemagglutination activity. For hemagglutination-positive batches, samples were amplified individually and tested for their hemagglutination activity. Sanger sequencing of the viral HA and NA segments was carried out to determine the influenza virus subtype. We isolated 136 low pathogenic viruses of several subtypes including H1N1, H3N2, H3N6, H3N8, H4N6, H6N2, H6N6, H9N2, H9N6, and H11N2. Importantly, we also detected 27 highly pathogenic avian H5 viruses (Table 1), all of which were deep-sequenced. The deep-sequencing data revealed viruses of the H5N1 and H5N6 subtypes and several samples with 'mixed' sequences, indicating infection with more than one virus (Table 1).

Phylogenetic Analysis of Highly Pathogenic H5 Viruses
To assess the genetic relationship of the isolated H5 viruses, we performed phylogenetic analysis for several viral genes ( Figure 2 and Figures S1-S4 Figure 2).
The N1-NA genes of the H5N1 viruses shared 96.5-100% identity at the nucleotide level and were similar to H5N1 viruses previously isolated in Vietnam (Supplemental Figure S1). The N6-NA genes of the H5N6 viruses shared 94.8-100% identity at the nucleotide level and fell into two larger phylogenetic groups (Supplemental Figure S2). We also performed phylogenetic analyses for the PB2 and NS1 genes whose gene products contribute to pathogenicity. The PB2 genes of the H5 viruses showed distinct diversity and formed three groups in the phylogenetic tree (Supplemental Figure S3). The genetic diversity of the PB2 segments was consistent with earlier findings of extensive reassortment between highly pathogenic avian H5 viruses and low pathogenic avian influenza viruses of the H6N6, H9N2, and other subtypes [16]. The NS1 genes of the H5 viruses were found to belong to Allele A and shared 95.4-100% identity at the nucleotide level (Supplemental Figure S4); they were similar to other H5 viruses isolated in Vietnam between 2012 and 2019.

Genetic Analysis of Highly Pathogenic H5 Virus Amino Acid Sequences
First, we compared the consensus sequences of the isolated highly pathogenic H5 viruses (Supplemental Tables S2 and S3). The viral HA protein serves as the receptorbinding protein and is the major viral antigen. All H5 viruses analyzed here encoded HA-222Q/224G (numbering of mature HA protein used throughout), which is known to confer avian-type receptor-binding specificity [19]. The HA cleavage site motif (at which HA is post-translationally cleaved by host proteases into HA1 and HA2) of highly pathogenic avian influenza viruses contains multiple basic amino acids. In our study, the motif RERRRKR/G (the forward slash indicates the cleavage site) was detected in all H5 viruses except for A/Muscovy duck/Vietnam/HN3790/2017 (H5N1; clade 2.3.2.1c), which encoded RERRKR/G (Supplemental Tables S2 and S3). Previous studies have shown that loss of the 154-156 glycosylation motif may enhance binding to α2,6-linked sialic acids and virus transmissibility via respiratory droplets among mammals [20,21]. All H5 viruses Viruses 2023, 15, 1093 6 of 14 characterized here lacked this glycosylation motif because they encoded alanine (A) at position 156; moreover, some of the viruses also lack an asparagine residue at position 154 (Supplemental Tables S2 and S3). The H5 subclades isolated here differed at several amino acid positions known to affect HA receptor-binding specificity, such as 189, 192, and 223 [22][23][24][25][26][27] (Supplemental Table S3); currently, the consequences of these differences for infection of mammals are not known.
The N1-NA genes of the H5N1 viruses shared 96.5%-100% identity at the nucleotide level and were similar to H5N1 viruses previously isolated in Vietnam (Supplemental Figure S1). The N6-NA genes of the H5N6 viruses shared 94.8%-100% identity at the nucleotide level and fell into two larger phylogenetic groups (Supplemental Figure S2). We also performed phylogenetic analyses for the PB2 and NS1 genes whose gene products contribute to pathogenicity. The PB2 genes of the H5 viruses showed distinct diversity and formed three groups in the phylogenetic tree (Supplemental Figure S3). The genetic diversity of the PB2 segments was consistent with earlier findings of extensive reassortment between highly pathogenic avian H5 viruses and low pathogenic avian influenza viruses of the H6N6, H9N2, and other subtypes [16]. The NS1 genes of the H5 viruses were found to belong to Allele A and shared 95.4%-100% identity at the nucleotide level (Supplemental Figure S4); they were similar to other H5 viruses isolated in Vietnam between 2012 and 2019.  The viral NA protein encodes a sialidase that facilitates virus release from infected cells. Deletions in the NA stalk region are frequently observed in avian influenza viruses adapted to terrestrial birds and may enhance pathogenicity in mammals (reviewed in [28]). The N1-NA proteins analyzed here possessed a 20-amino acid deletion at positions 49-68 (Supplemental Table S2), which has been observed in the H5N1 influenza virus since 2000 and enhances its pathogenicity in mallard ducks and mice (reviewed in [28]). Some, but not all, of the N6-NA proteins analyzed here possessed a shorter deletion of 11 amino acids (positions 58-68) (Supplemental Tables S2 and S3), which increases virus replication in mammalian and avian cells [29].
NA is the target of several antiviral compounds that block its enzymatic activity. H5N1 viruses with H274Y or N294S (N1 NA numbering) substitutions in NA are resistant to oseltamivir [30,31]. Moreover, viruses encoding N6 NA-R292K (possessing a full-length NA stalk) show reduced susceptibility to zanamivir and laninamivir, whereas the N6 NA-E119V+I222L substitutions reduce the inhibitory effect of oseltamivir [32]. None of the H5N1 or H5N6 strains analyzed here encoded these substitutions, suggesting that they are sensitive to laninamivir and oseltamivir.
The PB1 viral RNA segment encodes a second protein, PB1-F2, which modulates the inflammasome, type I IFN responses, and the induction of apoptosis; its function also affects the severity of bacterial infection (reviewed in Refs. [49,50] Tables S2 and S3).
The PA gene encodes another subunit of the viral polymerase complex. Several amino acid changes in PA have been reported to affect polymerase activity and/or pathogenicity in mice [52][53][54][55][56][57][58], but none of these amino acids was present in the viruses isolated here. However, all viruses analyzed possessed PA-383D, which confers increased polymerase activity to the H5N1 polymerase complex compared with a polymerase complex encoding PA-383N [59]. Most of the clade 2.3.4.4 viruses encoded PA-237E, which contributes to H5N1 pathogenicity in ducks [55] (Supplemental Tables S2 and S3), and all clade 2.3.2.1c viruses encoded PA-343S, which increases the pathogenicity of low pathogenic H5N1 virus in mice relative to a virus possessing PA-343A [57] (Supplemental Tables S2 and S3).
The nucleoprotein (NP) is the main structural component of the viral ribonucleoprotein complex, and its interaction with host factors also contributes to the host range of influenza viruses [60]. A previous study found that the NP-M105V substitution increased the pathogenicity of an H5N1 virus in chickens [61]. In our study, all viruses in clade 2.3.2.1c and one virus in clade 2.3.4.4f possessed NP-105V; the remaining viruses encoded isoleucine or methionine at this position (Supplemental Tables S2 and S3).
The matrix (M1) protein is the main structural component of influenza virions. H9N2 viruses of the G57 genotype are more pathogenic in chickens than H9N2 viruses of other genotypes, a phenotype that is conferred by the M1-37A, -95K, -224N, and -242N residues [62]. Several of the viruses characterized here encoded some of these amino acids, which may increase their pathogenicity (Supplemental Tables S2 and S3). In addition, all viruses characterized here encoded M1-30D and -215A, which increase H5N1 pathogenicity in mice relative to control viruses [63]. The M viral RNA segment also encodes the ion channel M2 protein, the target of the ion channel inhibitors amantadine and riman- tadine. Several amino acid changes in the ion channel confer resistance to ion channel inhibitors [64][65][66][67][68], but the viruses characterized here did not encode any of these amino acids, suggesting that they are sensitive to these drugs.
NS1 is a multifunctional viral protein that counteracts the innate immune response of the host, enabling the virus to replicate effectively during infection [69]. All viruses analyzed here lacked NS1 amino acids [80][81][82][83][84] (Supplemental Tables S2 and S3), a deletion that increases the pathogenicity of H5N1 viruses in mice [70]. The amino acids at positions 103 and 106 (equivalent to positions 98 and 101 in NS1∆80-84) are important for binding to CPSF30 [71,72], which affects the nuclear export of poly(A)-containing mRNA and thus influenza virus-mediated shutdown of host cell protein synthesis. The  Tables S2 and S3). Moreover, all viruses characterized here encoded NS1-42S, which confers increased pathogenicity in mice compared with viruses encoding proline at this position [75].

Analysis of Viral Subpopulations
Our analysis of viral subpopulations focused on sequence variants detected with a minimum frequency of 1% at the respective position (Supplemental Table S4). We focused on non-synonymous polymorphisms, that is, sequence variants that resulted in amino acid changes. Samples with 'mixed' HA and/or NA genes (indicative of infection with two different viruses) were eliminated from the analysis. With these parameters, no nonsynonymous polymorphisms were detected for PB1-F2 and NS2. Polymorphisms in PA, PA-X, and NP were mainly from A/duck/Vietnam/HN4042/2017 (with a frequency of 1-3%) (Supplemental Table S4).
Several non-synonymous polymorphisms were detected in the HA genes (Supplemental Table S4), some of which may affect the biological properties of the viruses. These include sites that form the rim of the receptor-binding domain (i.e., an HA-A125Y polymorphism detected for A/Muscovy duck/Vietnam/HN3790/2017), the receptor-binding pocket (i.e., an HA-R223H polymorphism detected for A/duck/Vietnam/HN3668/2017), the HA cleavage site (i.e., an HA-R327K polymorphisms detected for A/duck/Vietnam/HN4231/2017), or the HA transmembrane domain (i.e., an HA-A523T polymorphism detected for A/duck/ Vietnam/HN4231/2017 and A/duck/Vietnam/HN4240/2017). Two viruses (A/Muscovy duck/Vietnam/QN3253/2016 and A/Muscovy duck/Vietnam/QN3262/2016) encoded HA-A134S polymorphisms in 33.6% and 8.4% of sequence reads, respectively. This position is located in the 130-loop, and the A-to-T substitution has been shown to increase binding to α2,6-linked sialic acids [76].
Several non-synonymous polymorphisms were detected in the polymerase and NP genes and could affect viral replication. At amino acid position PB2-178, A/Muscovy duck/Vietnam/HN3656/2017 encoded threonine in 78.8% of sequence reads and alanine in 21.2% of sequence reads (Supplemental Table S4); interestingly, we previously found that amino acid changes at this position can affect virus replication (unpublished). A valine-toisoleucine substitution at position 591 of PB1 contributes to the temperature-sensitive phenotype of the live attenuated influenza A/Leningrad/134/17/57 strain [77]. Here, we detected a PB1-V591D subpopulation in 4.1% of sequence reads of A/duck/Vietnam/HN3680/2017 (Supplemental Table S4). PB1-A661T also contributes to the temperature-sensitive phenotype of the live attenuated influenza A/Leningrad/134/17/57 strain [78]; this amino acid polymorphism was detected in 2.3% of A/duck/Vietnam/HN4231/2017 sequence  Table S4). A/Muscovy duck/Vietnam/HN3801/2017 possessed a PB1-A652V subpopulation (Supplemental Table S4); this amino acid change has been detected in a mouse-adapted human pandemic H1N1 virus [79]. Methionine at PA-336, which confers the high replicative ability to pandemic H1N1 viruses [54], was detected in our study in a small subpopulation of A/Muscovy duck/Vietnam/HN3810/2017 sequence reads (Supplemental Table S4). NP-I365V confers decreased polymerase activity to the H7N9 virus [80] and was detected in 1.6% of sequence reads of A/duck/Vietnam/HK4042/2017 (Supplemental Table S4).

Discussion
Vietnam is one of the countries most affected by outbreaks of highly pathogenic H5 influenza viruses in poultry; in fact, these viruses are now enzootic in poultry populations throughout the country. Live bird markets, as one of the main places for poultry trading, are often the focus of active avian influenza virus surveillance activities. In this study, 27

Discussion
Vietnam is one of the countries most affected by outbreaks of highly pathogenic H5 influenza viruses in poultry; in fact, these viruses are now enzootic in poultry populations throughout the country. Live bird markets, as one of the main places for poultry trading, are often the focus of active avian influenza virus surveillance activities. In this study, 27 highly pathogenic H5 influenza viruses were isolated from Vietnamese live bird markets from 2016 to 2017. The isolated viruses belonged to clades 2. 3 HPAI H5 viruses have evolved rapidly through reassortment and mutations [16,84]. Here, we also found that some of the viruses characterized possessed genes that are most closely related to low pathogenic avian influenza viruses. For example, some of the PB2 genes were most closely related to low pathogenic avian H9N2 viruses, which are known to share gene segments encoding internal proteins with H5, H7N9, H10N8, and H3N8 viruses. Thus, despite attempts to control HPAI H5 viruses, they continue to reassort with low pathogenic avian influenza viruses, which may alter their phenotypes. In summary, our avian influenza virus surveillance in Vietnam resulted in the isolation of several HPAI viruses, some of which are lethal in mice and possess mutations that may affect their pathogenicity.

Informed Consent Statement: Not applicable.
Data Availability Statement: All sequences were deposited to GenBank with the following accession numbers: OQ546728 to OQ546795; OQ546869 to OQ546921; OQ547053 to OQ547096; OQ596733 to OQ596736; OQ596768 to OQ596783; OQ596785 to OQ596800; OQ596802 to OQ596819; and OQ600194; OQ603031 to OQ603034. Data output from the NGS sequence analysis pipeline are available in Supplemental Tables S2 and S3.