Phylogenetic analyses of the H5 HA genes from the 2004 and 2005 outbreak showed 2 different lineages of HA genes, termed clades 1 and 2. Viruses in each of these clades are distributed in nonoverlapping geographic regions of Asia (Figure 1). The H5N1 viruses from the Indochina peninsula are tightly clustered within clade 1, whereas H5N1 isolates from several surrounding countries are distinct from clade 1 isolates and belong in the more divergent clade 2. Clade 1 H5N1 viruses were isolated from humans and birds in Vietnam, Thailand, and Cambodia but only from birds in Laos and Malaysia. The clade 2 viruses were found in viruses isolated exclusively from birds in China, Indonesia, Japan, and South Korea. Viruses isolated from birds and humans in Hong Kong in 2003 and 1997 made up clades 1´ and 3, respectively.

The HA genes from H5N1 viruses isolated from human specimens were closely related to HA genes from H5N1 viruses of avian origin; human HA gene sequences differ from the nearest gene from avian isolates from the same year in 2–14 nucleotides (<1% divergence). These findings are consistent with the epidemiologic data that suggest that humans acquired their infections by direct or indirect contact with poultry or poultry products (21).

Analysis of the amino acid sequences showed that both clades of H5 HAs from the 2004–2005 outbreak have a multiple basic amino acid motif at the cleavage site, a defining feature of highly pathogenic avian influenza viruses. Among all H5N1 isolates collected in east Asia since 1997, only those in clades 1, 1´, and 3 appear to be associated with fatal human infections (22,23). We compared amino acid sequences of HA from contemporary isolates (clades 1 and 2) with those of the fatal H5N1 infections in Hong Kong in 1997 and 2003 to identify changes that may correlate with patterns of human infection (Table 1). Thirteen polymorphic sites were identified when the HA1 from the 4 consensus sequences were compared. One change in the 2004–2005 viruses is serine 129 to leucine (S129L). This change affects receptor binding because S129 makes atomic contact with cellular sialoside receptors (24). A second structural change in HA was the A156T substitution, which resulted in glycosylation of asparagine 154 and is predicted to reduce its affinity for sialosides. This change is commonly associated with viral adaptation to terrestrial poultry and increased virulence for these birds (25–27).

Because of the heightened alert due to H5N1 infections in Vietnam during the first months of 2005, we examined the HA sequences for evidence of shared amino acid changes. The HA of viruses isolated in the first 3 months of 2005 showed several amino acid changes relative to 2004 viruses (Table 1). None of the changes in the HA were common to all the 2005 viruses, which suggests that these variant viruses are cocirculating independently in poultry. The most commonly observed changes are located within short distances of the receptor-binding site. For example, positions D94, L175, and T188 may modulate the interaction of Y91, H179, and L190 with sialosides. One of the isolates from a fatal infection in 2005 showed a substitution of serine 223 to asparagine, which is predicted to facilitate binding of sialosides commonly found in mammalian species (Table 1).

The phylogenetic tree of the NA genes resembled that of the HA genes, which indicates coevolution of these 2 envelope genes (Figure 2). NA genes of isolates from Thailand seem to have diverged to form a group distinct from that of genes from Vietnam viruses. As reported previously, the NA of HK/213/03 did not co-evolve with the HA genes (28). NA genes from human and related avian H5N1 isolates from 2003–2005 as well as clade 3 isolates were characterized by deletions in the stalk region of the protein (positions 49–68 for clades 1–2 and 54–72 for clade 3) (29). Deletions in the stalk of the NA are thought to increase retention of virions at the plasma membrane (30) to balance weaker binding of sialic acid receptors by the HA with newly acquired N154 glycosylation.

Neuraminidase inhibitors are effective antiviral drugs against human influenza viruses, and preclinical studies suggest a similar effectiveness against avian influenza in humans (5,31). The IC₅₀ of oseltamivir for the clade 1 and 2 NA of 2004–2005 isolates was <10 nmol/L, as compared to IC₅₀ values of 85 and 1,600 nmol/L for resistant H1N1 or H3N2 mutants used as controls (Table 3). Thus, NA of H5N1 isolates is sensitive to this class of antiviral agents.

The phylogenetic tree of the M genes resembled that of the HA genes, indicating coevolution of these genes (results not shown). The amino acid sequence of the M2 protein of clade 1 viruses as well as of HK/213/03 indicated a serine-to-asparagine substitution at residue 31 (S31N), known to confer resistance to adamantanes (including amantadine and rimantadine) (6). Clade 1 isolates from 2004 and 2005 cultured in the presence of 2 μg/mL rimantadine replicated as efficiently as in untreated cultures, whereas the replication of HK/483/97 was reduced to 1% of control values, indicating that all the currently circulating clade 1 isolates are resistant to adamantanes (data not shown).

A complete genetic characterization of circulating H5N1 viruses is critical to identify the possible incorporation of human influenza virus genes by reassortment. To this end, we analyzed the phylogeny of the internal protein coding genes. The PB2, PB1, and PA polymerase genes from 2003–2005 H5N1 isolates from humans constitute a single clade (data not shown) and have coevolved with the respective HA genes (Figure 1). No evidence of reassortment with polymerase genes from circulating H1N1 or H3N2 human influenza virus was found. The phylogenies of the NP and NS genes also supported the avian origin of these genes, indicating that all the genes from the human H5N1 isolates analyzed are of avian origin, which confirms the absence of reassortment with human influenza genes. Taken together, the phylogenies of the 8 genomic segments show that the H5N1 viruses from human infections and the closely related avian viruses isolated in 2004 and 2005 belong to a single genotype, often referred to as genotype Z (1).

Influenza vaccines whose HA are antigenically similar to circulating strains provide the highest level of protection from infection (32). H5N1 isolates collected in 2004 and 2005 analyzed by the HI test showed reactivity patterns that correlated with the 3 main clades of recent isolates identified in the HA gene phylogeny (Table 3 and Figure 1). Viruses from humans and birds in clade 1, represented by VN/1203/04, were found to constitute a relatively homogeneous and distinct antigenic group characterized by poor inhibition by ferret antisera to isolates from other clades (Table 2), in particular by the ferret antiserum raised to HK/213/03 (64-fold reduction compared to the homologous titer). The latter isolate was previously used to develop a vaccine reference strain in response to 2 confirmed H5N1 human infections in February 2003 (15). These HI results provided the motivation for the development of an updated H5N1 vaccine that would be antigenically similar to 2004–2005 human isolates. The antigenic similarity of VN/1203/04 and the closely related VN/1194/04 to the contemporaneous H5N1 isolates from humans (data not shown) prompted their selection for vaccine reference stock development.

Antigenic analysis of human isolates from 2005 provided evidence of antigenic drift among the most recently circulating H5N1 strains (Table 2). For example, VN/JPHN30321/05 showed a reduced HI titer against VN/1203/04 reference serum. This antigenic difference is correlated with 7 amino acid differences between the HA1 domain VN/1203/04 and VN/JPHN30321/05: R53K, N84D, D94N, K140R, L175M, K189R, and V219I (Table 1 and Table A1). Development of Candidate H5N1 Vaccine Reference Stocks Mass vaccination is the most effective approach to reduce illness and death from pandemic influenza. Inactivated influenza vaccines are manufactured from reassortant viruses obtained by transferring the HA and NA genes with the desired antigenic properties into a high-growth strain such as PR8 (33). However, reassortants with H5-derived HA with a polybasic cleavage site are potentially hazardous for animal health. Because the high pathogenicity of the H5N1 viruses in poultry, mice, and ferrets depends primarily on the polybasic cleavage site in the HA molecule, a derivative with a deletion of this motif was engineered in cloned HA cDNAs. Three high-growth reassortant influenza viruses were developed: NIBRG-14 (NIBSC), VN/04xPR8-rg (SJCRH), and VNH5N1-PR8/CDC-rg (CDC). These candidate vaccine strains, bearing mutant H5 HA, intact NA, and the internal genes from PR8, were generated by a reverse genetics approach (12,13,20,34) using Vero cells and laboratory protocols compatible with eventual use of the vaccine in human subjects (15,18). These 3 vaccine candidates were characterized genetically (nucleotide sequencing of HA and NA) and antigenically in HI assays to confirm that their antigenicity remained unchanged relative to the wildtype virus (Table 2). The candidate reference stocks had molecular and antigenic properties equivalent to parental H5N1 donor strains and lacked virulence in chicken, mouse, and ferret models.