Genetic diversification of H5N1 highly pathogenic avian influenza A virus during replication in wild ducks

Waterfowl, including ducks, are natural reservoirs for influenza A virus, since infected hosts are usually asymptomatic (Webster et al., 1992). However, this virus–host equilibrium has changed recently: highly pathogenic avian influenza A virus (HPAIV) subtype H5N1 with genotype Z, which is geographically widespread throughout Asia, Africa and Europe, has been causing severe mortality in a variety of waterfowl and domestic poultry since 2002 (Bragstad et al., 2007; Ellis et al., 2004), indicating establishment of a new host–virus relationship in birds (Li et al., 2004; Sturm-Ramirez et al., 2004).

According to a recent OIE report (OIE, 2010), avian H5N1 HPAI has become endemic in some areas; e.g. China, Indonesia, Vietnam and Egypt. Ducks have been significant reservoirs for the emergence of H5N1 outbreaks and the spread of H5N1 to domestic poultry and humans in these areas (Chen et al., 2004; Gilbert et al., 2006). However, the role of ducks in H5N1 diversification has not been completely elucidated.

Currently, circulating genotype Z viruses show a broader range of tissue tropism and, as a result, more complex pathogenicity in ducks than previous influenza virus strains (Hulse-Post et al., 2007; Sturm-Ramirez et al., 2004). Such pathogenic behaviour substantially enhances the possibility that a particular mutation might be selected and expanded in a specific organ or tissue during a systemic infection in ducks (viral compartmentalization). However, there have been few reports on the emergence and selection of new influenza virus variants in ducks (Hulse-Post et al., 2005). Newly emerging variants in infected hosts can be easily detected by analysing viral sequences cloned directly from infected hosts. Such an analysis for infected ducks may provide a clue to understanding the potential role of wild ducks in the dynamics of H5N1 evolution.

We report here that a variety of H5N1 variants, with appreciably distinct phenotypes, were compartmentalized in different organs of wild ducks. Our observations highlight a possible role for wild ducks as a source of new H5N1 variants in HPAI outbreaks and ongoing epidemics.

During an H5N1 outbreak in Northern Egypt from February to March 2007 (OIE, 2010), we collected samples from three dead ducks (designated D1–D3) and three dead chickens (designated C1–C3) with severe clinical signs characteristic of HPAI. We sampled four organs: trachea, lung, brain and liver. The H5N1 viral RNA copy number in each organ homogenate was measured, as described previously (WHO, 2007), to investigate viral replication patterns in the infected birds. The results showed that viral RNA copy numbers in non-respiratory organs (brain and liver) were comparable to those in respiratory organs (trachea and lung) in all of the birds tested, indicating that these had been systemic infections: the mean log₁₀ viral RNA copy number µg⁻¹ total homogenate RNA±sd was 4.3±0.5 in trachea, 4.3±2.0 in lung, 5.0±2.1 in brain and 3.9±0.2 in liver.

Previous reports (Hiromoto et al., 2000; Takemae et al., 2010) showed that mutant viruses were selected during influenza virus isolation and passage in both Madin–Darby canine kidney (MDCK) cells and embryonated chicken eggs, so these procedures may misrepresent the virus population in infected individuals. Therefore, we performed direct genetic analysis of organ homogenates from infected birds. RNAs extracted from these homogenates were reverse-transcribed with Superscript Ш Reverse Transcriptase (Invitrogen) and Uni-12 primer (Hoffmann et al., 2001). PCR amplification was performed with Platinum Taq DNA Polymerase High Fidelity (Invitrogen) and primers specific for both ends of each H5N1 influenza virus gene segment (Hoffmann et al., 2001; WHO, 2007). PCR products were cloned into a TA cloning vector (TOPO TA Cloning kit; Invitrogen) and transfected into TOP10 cells (Invitrogen). Escherichia coli clones were picked and the complete nucleotide sequence of each viral protein-coding region was analysed as described previously (Watanabe et al., 2007).

For each virus segment, the nucleotide sequences of 10 molecular clones were aligned to determine the dominant sequence of all eight virus segments in the organs of each bird. A total of 1920 clones (10 clones of eight virus segments from four organs of six birds) were analysed. Genetic analysis revealed that all eight virus segments had >99 % nucleotide identity with those of the H5N1 viruses circulating in Egypt at the time of this investigation, and confirmed the viruses to be genotype Z. In particular, viral genomes in the trachea samples from all of the birds had <6 nt differences per kb relative to each other, as expected from previous reports that H5N1 was introduced in Egypt by a single genotype Z virus infection (Cattoli et al., 2009), with the initial infection site being the upper respiratory tract of birds (Sturm-Ramirez et al., 2004, 2005). These results confirmed a close phyogenetic relationship between the original genotype Z virus and the viruses in the infected birds studied here.

In this study, we considered the trachea to be the primary site of infection and lung, brain and liver to be secondary sites of infection based on the presumed viral infection route. Although histopathological analysis of duck and chicken organs infected in nature found more severe pathological changes in trachea and lungs than in brain and liver (Supplementary Fig. S1, available in JGV Online), these data could not indicate the timing of pathological changes in these organs post-infection. To investigate H5N1 variants in each bird, we compared the sequences of the viral genes from the trachea to those from the other organs and determined the genetic variation in viruses infecting different organs in individual birds. The results showed that the viral sequences were heterogeneous in ducks, but homogeneous in chickens. The total number of nucleotide changes in the influenza virus genomes from each bird was 18 in D1, 52 in D2 and 22 in D3, but only four in C1 and zero in C2 and C3 (Fig. 1, left panel). The total number of the amino acid changes was two in D1, 10 in D2 and six in D3, but there were no amino acid changes in the virus genomes from the three chickens (Fig. 1, right panel). These results indicated that H5N1 virus diversified during replication in ducks but not in chicken samples, suggesting the emergence and compartmentalization of H5N1 variants in ducks.

Therefore, we focused on the amino acid changes in viruses from different organs in the three ducks (Fig. 2). For viruses from the trachea, the amino acid sequences were the same for all three ducks and the same as those of H5N1 viruses circulating in Egypt at the same time as the samples in this study were obtained. In contrast, for sequences from the brain and liver, amino acid changes were found in viral non-structural proteins PB1-F2, PA and NS1 in the three ducks. These changes have not been reported to date in H5N1 viruses isolated from birds in Egypt (the National Center for Biotechnology Information, NCBI). For viruses from the lung, one duck had multiple amino acid changes in HA. These multiple changes also have not been previously reported (NCBI). These results indicated that ducks can generate H5N1 variants with novel amino acid substitutions. Remarkably, the mutations were mostly conservative changes and located in the functional domains of each viral protein.

Since the sequence of each segment in Fig. 2 is the dominant sequence based on data from 10 clones, it was interesting to note that most multiple mutations were in the same clone; e.g. a clone from the lung of D3 that had one of the HA amino acid changes usually also had the other two HA amino acid changes, and a clone from the liver of D2 that had one of the PA amino acid changes usually also had the other PA amino acid change. We also noted that clones from both brain and liver from two of the ducks (D2 and D3) had both a PB1-F2 L82S and a PA V423I amino acid sequence change. These results suggested that the viruses have undergone specific selective pressure during replication in ducks and acquired a phenotype distinct from the original virus. Indeed, novel sequence variants were not detected in the trachea of any of the ducks, indicating that variants were <10 % of the initial virus population, suggesting a strong selective pressure in ducks. In addition, of the 10 residues with non-synonymous mutations in the three ducks, only two (HA residues 120 and 140) are known to be associated with antigenic drift (Kaverin et al., 2007), but the pathogenicity of the other mutations in H5N1 in ducks or other species has not been described.

To determine whether new H5N1 variants that have emerged in ducks can acquire a biologically distinct phenotype during infection of a single duck, we tried to isolate viruses from organ homogenates of each duck and eventually were able to isolate four viruses by single passage in embryonated eggs. The ducks and organs from which these four viruses were isolated were the trachea of D1, the brain of D2, the liver of D2 and the lung of D3. Each virus had the amino acid sequences characteristic of the organ from which it was isolated, as shown in Fig. 2: the virus from the trachea had no amino acid changes from the original virus; the virus from the lung had HA 97N-120N-140K; the virus from the brain had PB1 F2-82S, PA 132M-432I and NS1 193V-204N; and the virus from the liver had PB1 F2-82S, PA 132M-432I and NS1 95K-159L. No other mutations were identified in these four viruses.

Based on their sequences, the virus from the trachea was designated the original virus and the three other viruses, from the lung, brain and liver, were designated variants 1, 2 and 3, respectively. We studied the growth of these viruses in cell cultures derived from several duck, canine and human tissues by infecting cells at an m.o.i. of 0.001 or 0.1 and assaying viral growth for 72 h. The original and variant viruses had similar growth kinetics in duck cells from four tissues, but significantly different growth kinetics in canine cells and human cells from two tissues (Fig. 3a). We then carried out plaque assays to investigate the morphology of the plaques produced on MDCK cells. Interestingly, variant 2 showed smaller plaques than the other viruses (Fig. 3b), in agreement with its growth kinetics. These findings indicated that the emergence of new H5N1 viruses in ducks was associated with distinct replication properties in vitro.

Quasispecies provide a source of variants for natural selection, ensuring evolutionary flexibility (Domingo et al., 1998). Our results on diversification of H5N1 virus and compartmentalization of variants in wild ducks suggest a potential path for H5N1 evolution. These results are also in agreement with a previous report that antigenically and biologically distinct variants were selected in ducks during a single experimental infection (Hulse-Post et al., 2005). In addition, our sequence analyses found that certain amino acid substitutions were selected in H5N1 non-structural proteins in infected wild ducks. Amino acid substitutions in specific genes (82S in PB1-F2, and 132M and 432I in PA) were found in H5N1 viruses infecting different ducks. These results imply that mutants may emerge as an adaptation or escape process in ducks during the course of an H5N1 systemic infection. However, although the growth kinetics of H5N1 variants were different than that of the original virus in some types of cells (Fig. 3a), acquisition of new amino acids during an infection in ducks did not necessarily produce a fitter phenotype (i.e. better growth) in cells from other species. The original virus was passaged up to five times in a panel of primary cell cultures from duck organs to see whether the virus would acquire the same mutations seen in viruses in wild ducks. However, the viruses exhibited no additional mutations in the culture system. Thus, the question whether viruses were selected for reduced pathogenicity in ducks, as suggested previously (Hulse-Post et al., 2007), should be investigated in a future in vivo study. In conclusion, our findings show the potential role of wild ducks in generating H5N1 variants that may serve as a source of new viruses producing new outbreaks and epidemics.

The mechanism underlying viral compartmentalization in ducks, but not in chickens, has not been fully elucidated. However, we speculate that the ducks may be infected for a longer time before death than chickens and, therefore, a virus could go through more replication cycles and mutations in ducks than in chickens. Although the novel amino acid substitutions found in this study were compartmentalized mainly in non-respiratory organs from which they are unlikely to be excreted efficiently to continue transmission, they may be carried from infected birds by interspecies predation (Keawcharoen et al., 2004; Thiry et al., 2007). Such exceptional transmission could accelerate virus evolutionary dynamics (Webster et al., 1992). Even if the H5N1 variants studied here had not acquired a significantly distinct phenotype during infection of a single bird, the appearance of new variants could facilitate further transmission, accumulation of mutations, and eventual pathogenicity in birds and even humans. However, the limited number of birds in this study makes it difficult to draw strong conclusions. Additional surveillance of wild ducks and characterization of the viruses isolated from them are needed to further elucidate the evolution of H5N1 virus in nature.