Viruses were isolated from specimens obtained from dead wild birds (Appendix Table) by inoculating embryonated eggs at the laboratory of the Agriculture, Fisheries and Conservation Department of the Hong Kong SAR Government. Viruses were identified by real time–PCR and by standard hemagglutination-inhibition (HI) tests using a panel of World Organization for Animal Health’s Avian Influenza Reference Laboratory antisera (Veterinary Laboratory Agency, Weybridge, UK) as previously described (14,23,24). All virus isolation was conducted in biosafety level 3 facilities. Details of the avian influenza (H5N1) surveillance program in Hong Kong for dead wild birds, including pathologic findings and diagnostic testing, are reported separately (17).

Antigenic characterization of the influenza viruses (H5N1) was carried out by HI assay using 5 ferret polyclonal antisera, as previously described (24). The ferret antisera were provided by St Jude Children’s Research Hospital (Memphis, TN, USA) (duck/Hunan/101/2004 and muscovy duck/Vietnam/1455/2006) and by the Centers for Disease Control and Prevention (Atlanta, GA, USA) (Anhui/1/2005, Indonesia/5/2005, Indonesia/CDC357/2006, Vietnam/1203/2004, and whooper swan/Mongolia/244/2005). The HI assay started at a serum dilution of 1:40.

To understand the evolutionary history of avian influenza viruses (H5N1) isolated from wild birds in Hong Kong, we conducted whole genome sequencing of 29 avian influenza viruses (H5N1) that were isolated from dead wild birds in 2006–2008. All 8 gene segments of these viruses were characterized and phylogenetically analyzed. These data were compared with the virus sequence data for an additional 18 influenza viruses (H5N1) isolated from dead wild birds in Hong Kong in 2004–2008, with virus sequence data for the 2 viruses obtained from chickens smuggled into Hong Kong in 2006, and with all other available sequence data from the NCBI Influenza Virus Resource (25).

Viral RNA extraction, cDNA synthesis, PCR, and sequencing were carried out as described previously (19). Sequences were assembled and edited with Lasergene version 7.2 (DNASTAR, Madison, WI, USA). Se-Al version 2.0 was used for alignment and residue analysis (http://tree.bio.ed.ac.uk/software/seal). The program MrModeltest version 2.2 was used to determine the appropriate DNA substitution model and rate heterogeneity (26). The generated model was used in all subsequent analyses. Neighbor-joining trees were constructed with PAUP* version 4.0b (27), and Bayesian analysis was conducted with MrBayes version 3.1.2 (28) by using 2 replicates of 1 million generations with 6 chains, sampling every 100 generations. The convergence of chains and the estimation of burn-in were assessed using Tracer version 1.4 (http://beast.bio.ed.ac.uk). Estimates of the phylogenies were calculated by performing 1,000 neighbor-joining bootstrap replicates, and Bayesian posterior probabilities were calculated from the consensus of 18,000 trees after excluding the first 2,000 trees as burn-in. The full-genome sequences of 29 influenza viruses (H5N1) obtained in this study are available from GenBank under accession nos. CY036042–CY036273.

From early 2004 through June 2008, most isolates of influenza virus (H5N1) from dead wild birds were detected during the cooler months (i.e., from December to the following February) (Appendix Table). Almost all positive samples of influenza virus subtype H5N1 were isolated from a variety of dead wild birds, including falcons, egrets, herons, and various passerine species. On 2 occasions, influenza virus (H5N1) was isolated from smuggled chickens (Appendix Table).

To understand the molecular epidemiology of the viruses isolated from the dead birds, we conducted phylogenetic analysis of the hemagglutinin (HA), neuramindase (NA), and each of the 6 internal gene segments of the viruses, along with the Gs/GD-like HPAI viruses (H5N1) isolated from different regions of Hong Kong. In the HA gene tree, the wild bird viruses fell into 2 main groups, either clade 2.3.2 or 2.3.4, with the exception of 1 virus in clade 9 (Figure). The phylogenetic placement of these viruses corresponds well with the known evolution of the influenza virus subtype H5N1 that has been documented in Asia.

The isolate detected in early 2004 (peregrine falcon/HK/D0028/2004) clustered into clade 9, which includes viruses isolated from both poultry and migratory ducks during 2003–2005 in China (Figure). The 3 viruses in clade 2.3.2 (grey heron/HK/728/2004, grey heron/HK/837/2004, Chinese pond heron/HK/18/2005) that were isolated in late 2004/early 2005 were most closely related to viruses detected in southern China and Vietnam during the same period. However, 29 of 31 viruses isolated in early 2006 and early 2007 were closely related to clade 2.3.4 viruses (represented by Dk/Fujian/1734/2005), corresponding with the time of emergence and predominance of this virus lineage (Figure, Appendix Table). The wild bird viruses isolated from May 2007 through March 2008 belonged exclusively to clade 2.3.2, with the exception of the clade 2.3.4 virus peregrine falcon/HK/2142/2008. These clade 2.3.2 viruses (H5N1) were most closely related to previous isolates from dead wild birds from Hong Kong (peregrine falcon/HK/5211/2006 and peregrine falcon/HK/1143/2007) and to isolates from Japan and Russia (whooper swan/Hokkaido/1/2008, whooper swan/Akita/1/2008 and Ck/Primorje/1/2008) (Figure).

Phylogenetic analysis of the NA of these isolates showed a similar phylogenetic relationship to that observed for the HA (data not shown). These findings show that influenza viruses (H5N1) detected each winter from 2004 through 2007 were genetically distinct and belonged to different sublineages or clades, suggesting that multiple introductions occurred during the past 4 years.

Phylogenetic analyses of the internal gene complex showed that the viruses from dead wild birds in Hong Kong belonged to different subtype H5N1 genotypes (Appendix Table). The virus peregrine falcon/HK/D0028/2004 clustered with those genotype Z viruses isolated from poultry in mainland China during the same period, and the 3 viruses in clade 2.3.2 that were isolated in late 2004 and early 2005 belonged to genotype V2 (22). The 17 viruses in clade 2.3.4 that were detected in early 2006 belonged to either genotype V (n = 6) or G (n = 10). The viruses isolated from January through June 2007, both clades 2.3.2 and 2.3.4, were mostly genotype V (n = 15), although 3 genotype V1 viruses were also detected (Appendix Table). Genotypes V1 and V2 are reassortments of genotype V that have incorporated novel PB2 and PB1 genes (22). Eight viruses isolated from November 2007 through March 2008 also belonged to genotype V, and 2 isolates (little egret/HK/8550/2007 and peregrine falcon/HK/2142/2008) were novel reassortments. The genetic diversity of these viruses confirms the multiple introductions of influenza viruses (H5N1) to Hong Kong.

Two thirds (12/18) of the clade 2.3.2 viruses were isolated from nonpasserine hosts, mostly species of egret, heron, and raptors (Appendix Table). In contrast, 3 (11%) of 28 viruses in clade 2.3.4 were isolated from nonpasserine hosts, excluding the viruses from the 2 chickens. Although inconclusive, this pattern suggests that clade 2.3.2 viruses may have an adaptation that enables them to infect and cause disease in nonpasserine species more easily than in other bird species.

All 29 viruses characterized were highly pathogenic with variations of the multibasic cleavage site in the HA molecule. However, all clade 2.3.4 viruses had a Gln→Leu substitution at position –9 from the cleavage site (LRERRRK-RG), a factor consistent with previous reports (18). The receptor-binding pocket of the HA1 retains amino acid residues 222-Gln and 224-Gly (H5 numbering used throughout) that preferentially bind to α-2,3-NeuAcGal receptors (29–31). Other amino acid residues relevant to receptor-binding sites were identical to those of HK/156/1997 and Gs/GD-like viruses in most isolates. However, all clade 2.3.2 viruses characterized had an HA Ser129Leu substitution, a factor previously observed in both clade 1 and 2 viruses (8,32). The clade 2.3.2 virus grey heron/HK/3088/2007 also had a Lys212Arg substitution (30).

In the NA amino acid sequences, all isolates characterized had 274-His, indicating sensitivity to oseltamivir (33). One virus (common buzzard/HK/9213/2007) had a Ser31Asn substitution in the M2 protein, a change that may confer resistance to the adamantanes and that has been present in all Clade 1 viruses characterized to date. This substitution has also been sporadically detected in other H5N1 lineages (34). No amantadine-resistant mutations were observed in the remaining isolates. None of these viruses have the Lys627 residue commonly found in Qinghai-like (clade 2.2) viruses (6).

Two of the clade 2.3.4 representative viruses (Japanese white-eye/HK/1038/2006 and scaly-breasted munia/HK/2572/2007) showed good reactivity against the clade 2.3.4 antiserum, but peregrine falcon/HK/2142/2008 was markedly less reactive with a >4-fold reduction in titer (Table). Also, peregrine falcon/HK/2142/2008 was poorly reactive against all ferret antisera tested. The pattern of reactivity of the clade 2.3.2 viruses from dead wild birds was similar to that of the homologous virus (muscovy duck/Vietnam/1455/2006) (Table).