|Home | About | Journals | Submit | Contact Us | Français|
The highly pathogenic (HP) influenza viruses H5 and H7 are usually nonpathogenic in mallard ducks. However, the currently circulating HP H5N1 viruses acquired a different phenotype and are able to cause mortality in mallards. To establish the molecular basis of this phenotype, we cloned the human A/Vietnam/1203/04 (H5N1) influenza virus isolate that is highly pathogenic in ferrets, mice, and mallards and found it to be a heterogeneous mixture. Large-plaque isolates were highly pathogenic to ducks, mice, and ferrets, whereas small-plaque isolates were nonpathogenic in these species. Sequence analysis of the entire genome revealed that the small-plaque and the large-plaque isolates differed in the coding of five amino acids. There were two differences in the hemagglutinin (HA) gene (K52T and A544V), one in the PA gene (T515A), and two in the PB1 gene (K207R and Y436H). We inserted the amino acid changes into the wild-type reverse genetic virus construct to assess their effects on pathogenicity in vivo. The HA gene mutations and the PB1 gene K207R mutation did not alter the HP phenotype of the large-plaque virus, whereas constructs with the PA (T515A) and PB1 (Y436H) gene mutations were nonpathogenic in orally inoculated ducks. The PB1 (Y436H) construct was not efficiently transmitted in ducks, whereas the PA (T515A) construct replicated as well as the wild-type virus did and was transmitted efficiently. These results show that the PA and PB1 genes of HP H5N1 influenza viruses are associated with lethality in ducks. The mechanisms of lethality and the perpetuation of this lethal phenotype in ducks in nature remain to be determined.
Wild aquatic birds are natural reservoirs of all known influenza A viruses (8, 22, 40). All 16 influenza A virus subtypes are perpetuated in this reservoir, in which they are usually nonpathogenic. The highly pathogenic (HP) H5 and H7 influenza subtypes are unique in their abilities to evolve rapidly, acquire additional basic amino acids at the connecting peptide of hemagglutinin (HA), and become highly pathogenic to gallinaceous poultry (chickens, quail, pheasants, etc.) (21, 26, 29) after transmission to these species. In contrast, these viruses have been nonpathogenic in ducks with some exceptions that include the A/Tern/South Africa/61 (H5N3) subtype (1) and an H7N1 virus that was pathogenic for Muscovy ducks in Italy in 2000 (2). This pattern changed in 2002 when the H5N1 virus that had emerged in 1997 killed the majority of exotic waterfowl in Penfold Park and Kowloon Park in Hong Kong, including tufted ducks, flamingos, geese, swans, and many other species (7, 36).
The importance of ducks in the emergence of the HP H5N1 virus and its spread to domestic poultry and humans in Asia is now well documented. After the initial detection of H5N1 influenza viruses in geese in Guangdong in 1996 and their spread to ducks in coastal southern China (3), the H5N1 viruses spread through live poultry markets and killed 6 of 18 virologically diagnosed humans in Hong Kong (34). Despite the culling of all poultry in Hong Kong and the probable eradication of the index genotype, additional novel genotypes emerged. The Z genotype became dominant (12) and spread to Thailand, Vietnam, Cambodia, and Laos.
The spread of H5N1 virus to chickens and to humans in Thailand corresponded to the distribution of free-grazing ducks (10). Despite their extensive lethality to multiple waterfowl species, the H5N1 viruses varied in their pathogenicity to ducks (20, 37), and the majority of grazing duck flocks infected with HP H5N1 in Thailand showed limited or no disease signs (35). The continuing importance of waterfowl in the evolution of H5N1 clades and subclades is documented by the detection of HP H5N1 virus in apparently healthy ducks and geese during most months of 2004 and 2005 in southern China (5). It appears likely that ducks played a role in the emergence of A/Bar-headed Goose/Qinghai/1A/05 (H5N1) virus and its subsequent spread to Europe, Africa, and India because this virus was lethal to geese but not to mallard ducks (4). The reemergence of HP H5N1 in ducks and geese in Vietnam in the period from July to August 2006, despite the extensive use of an H5N1 vaccine in domestic poultry again points to the key role of ducks in the continuing resurgences of HP H5N1.
The molecular changes associated with the unusual lethality of HP H5N1 viruses to ducks have not been identified. HP H5N1 lethality has been associated with multiple basic amino acids in the HA gene in chickens and with lysine at residue 627 of the PB2 gene in mice (14, 19, 26, 31). These changes were present in the HP H5N1 viruses that killed chickens and spread to humans in Asia between 1997 and 2001, yet the viruses were benign in ducks (33). Recent studies of HP H5N1 viruses from 2004 in ferrets and mice revealed high pathogenicity to be a complex phenotype dependent on both the virus and the host and involving a complex of the polymerase genes and the NS gene (30). To further elucidate the roles of these genes, we biologically cloned the human A/Vietnam/1203/04 (H5N1) virus and detected plaque-forming isolates that differed in their size and their pathogenicity to ducks. The molecular differences between the highly pathogenic and nonpathogenic isolates were associated with the polymerase PA and PB1 genes.
The H5N1 influenza A virus A/Vietnam/1203/04 (VN/1203) sample was provided by the World Health Organization collaborating laboratories and stored in the repository of St. Jude Children's Research Hospital. The virus was propagated in 10-day-old embryonated chicken eggs and handled at St. Jude Children's Research Hospital in biosafety level 3+ facilities approved by the United States Department of Agriculture and Centers for Disease Control and Prevention. Previously, the eight gene segments of the VN/1203 virus were cloned into a dual-promoter plasmid, pHW2000 (30). The VN/1203 plasmids were manipulated by using QuikChange site-directed mutagenesis kits (Stratagene) to generate a sequence identical to that of the A/Vietnam/1203/04 small-plaque virus isolate (VN/1203 SS). Each of the resulting five plasmids had one point mutation: HA K52T, HA A544V, PA T515A, PB1 K207R, or PB1 Y436H. For use in the luciferase assay, the VN/1203 PB1 plasmid was also manipulated to include both point mutations observed for the VN/1203 SS plaque virus (at positions 207 and 436). Five reverse genetic (rg) viruses differing from the VN/1203 wild-type virus by a single point mutation were generated by DNA transfection as described previously (17). Briefly, 293T and Madin-Darby canine kidney (MDCK) cells were cocultured and transfected with 1 μg of each of the eight plasmids and 18 μl of transit LTI (Panvera, WI) in a total volume of 1 ml of OPTIMEM-I (Gibco, NY). The supernatant was removed, and 100 μl was injected into the allantoic cavity of 10-day-old embryonated chicken eggs. After 48 h, the allantoic fluid was harvested, RNA was extracted and analyzed by reverse transcription-PCR (RT-PCR), and each viral segment was partially sequenced to confirm the identity of the virus.
All eight viruses utilized in this study were sequenced either fully (VN/1203 wild-type virus and SS and large [LL] plaque viruses) or partially (rg viruses). Viral RNA was isolated from allantoic fluid by using an RNeasy kit (QIAGEN). Gene segments were amplified by RT-PCR using the universal primer set for influenza A viruses (18). Viral cDNA and template cDNA were sequenced by the Hartwell Center for Biotechnology at St. Jude Children's Research Hospital.
Studies with mallards were performed as described previously (20). Briefly, two 4-week-old mallard ducks (Anas platyrhynchos) were inoculated with 106 50% egg infective doses (EID50s) of each stock virus in a 1-ml volume (0.5 ml was applied to the cloaca, 0.2 ml to the trachea, and 0.1 ml each to throat, nares, and eyes). Four hours later, the inoculated birds were placed in a new cage with two uninfected ducks, sharing food and drinking water. All birds were observed daily for 21 days. Birds that exhibited severe disease signs were euthanized and recorded as having died on the following day. Tracheal and cloacal swabs were collected every other day starting on day 3 postinoculation (p.i.) until virus was no longer isolated from embryonated chicken eggs (13, 36). The infectivity of positive samples was measured by determining the EID50. The limit of detection was 1 log10 EID50/ml. For the intravenous (i.v.) inoculation experiments (intravenous pathogenicity index [IVPI]), 4-week-old mallard ducks or 6-week-old specific pathogen-free chickens were injected with 106 EID50 of virus in a 0.1-ml volume. Ducks and chickens were observed for mortality over a 10-day period.
One-year-old ferrets seronegative by HA inhibition for exposure to influenza B, H1N1, H3N2, and H5N1 viruses were obtained from Marshall's Farms. For each virus, two ferrets were anesthetized with isoflurane and inoculated intranasally with 106 EID50 of virus. Temperatures were recorded daily via subcutaneous implantable temperature transponders, and clinical signs and weight were monitored. On days 3, 5, and 7 p.i., ferrets were anesthetized intramuscularly with ketamine (25 mg/kg of body weight), and 0.5 ml of sterile phosphate-buffered saline (PBS) was introduced into each nostril. Virus titrations in the wash fluid were determined for 10-day-old embryonated chicken eggs and expressed as log10 EID50/ml.
Groups of 9 or 10 6-week-old BALB/c female mice (obtained from Jackson Laboratory) were anesthetized with isoflurane and inoculated intranasally with 100 EID50 of virus. Weight and mortality were recorded daily.
The EID50 was determined in duplicate by injecting 100 μl of 10-fold dilutions of virus into the allantoic cavities of 10-day-old embryonated chicken eggs. The eggs were incubated at 37°C for 48 h, and hemagglutination activity was assayed. MDCK cells were infected with 10-fold dilutions of the wild-type, plaque-purified, and reverse genetic viruses, incubated at 37°C for 1 h, and then washed and overlaid with infection medium (minimal essential medium [MEM] with 0.3% bovine serum albumin [BSA] and 1 μg/ml tosylsulfonyl phenylalanyl chloromethyl ketone [TPCK]-trypsin). The 50% tissue culture infective dose (TCID50) was determined by hemagglutination assay after incubation at 37°C for 3 days. Both the EID50 and TCID50 values were calculated by the method described by Reed and Muench (27). To determine the multistep growth curves, MDCK cells were infected with VN/1203 wild-type, SS, or LL virus at a multiplicity of infection of 0.001 PFU/cell. After incubation, the cells were overlaid with infection medium and incubated at 37°C. Supernatant was collected at 12, 24, 36, 48, 60, and 72 h p.i. and assayed for hemagglutination activity.
Plaque purification of viruses was performed as described previously (15). Briefly, confluent MDCK cells were incubated for 1 h at 37°C with 10-fold dilutions of virus. The cells were then washed and overlaid with MEM containing 0.3% BSA, 0.9% Bacto agar, and 1 μg/ml of TPCK-trypsin and incubated at 37°C. After 3 days, plaques were picked to be further purified or stained with 0.1% crystal violet solution containing 10% formaldehyde.
On day 6 p.i., two mallards inoculated with each virus were humanely euthanized, and lung, brain, heart, and pancreas were collected, weighed, placed in sterile PBS (wt/vol ratio, 1:1), and homogenized. Homogenate titrations were determined in 10-day-old embryonated chicken eggs. For histologic examination, lung, heart, pancreas, and brain were fixed in 10% neutral buffered formalin for a minimum of 24 h. Fixed tissues were processed routinely, embedded in paraffin, cut into 4-μm sections, stained with hematoxylin and eosin, and evaluated microscopically by a veterinary pathologist.
The luciferase assays were performed as described previously (30). All experiments were performed in triplicate. Briefly, 293 T cells were transfected with 2 μg of luciferase reporter plasmid (16) and a mixture of PB2, PB1, PA, and NP plasmids in quantities of 1, 1, 1, and 2 μg, respectively. The VN/1203 wild-type plasmids were mixed together with or without a plasmid carrying the point mutation (e.g., the VN/1203 wild-type PB2, PB1, and NP plasmid mixture plus the rg VN/1203 PA T515A plasmid). The VN/1203 wild-type NP and PB2 plasmids, the PA T515A plasmid, and the PB1 plasmid with the double mutation (PB1 K207R and Y436H) were used to quantify the polymerase activity of the VN/1203 SS plaque virus. Cell extracts were harvested 24 h posttransfection and added to 500 μl of lysis buffer. Luciferase activity was then assayed with a luciferase assay system (Promega) and read on a BD Monolight 3010 luminometer (BD Biosciences).
Previously, we found that less pathogenic variants were selected for in mallards inoculated with H5N1 viruses (20). When the original stock of VN/1203 virus was plaque purified on MDCK cells, two distinct types of plaques (large and small) were observed. Both large and small plaques were passaged a second time on MDCK cells, and their pathogenicity in mallards was tested. In the first experiment, the original sample (VN/1203 wild type) caused neurological signs in the inoculated and contact mallards and the deaths of three of the four ducks (Table (Table1).1). The large-plaque virus, VN/1203 LL, caused the death of one duck and severe neurological signs in the other three ducks, whereas VN/1203 SS virus caused clinical signs in only one duck and no mortality. All of the ducks, including the contact birds, were infected. Mallards inoculated with the VN/1203 wild-type and VN/1203 LL viruses shed virus for 13 days; those inoculated with VN/1203 SS shed for only 7 days. When we repeated the experiment, some differences were observed. Inoculation with the wild-type virus caused results similar to those in the first experiment (Table (Table1).1). The VN/1203 LL virus resulted in the death of one inoculated and one contact duck, and three of the four ducks showed signs of disease. The VN/1203 SS virus caused no signs of disease and no mortality. Wild-type virus was shed until day 5 p.i., VN/1203 LL virus was shed until day 5 p.i., and VN/1203 SS virus was shed until day 3 p.i.
We next investigated whether the route of infection would affect the pathogenicity of the VN/1203 SS virus. Mortality was observed after i.v. inoculation with VN/1203 wild-type virus (10/10 deaths) and VN/1203 LL virus (3/8 deaths) but not after i.v. inoculation with VN/1203 SS virus (0/10 deaths) (Table (Table1).1). To determine whether lower virus replication explained the decreased pathogenicity of the VN/1203 SS virus, virus titrations shed from the trachea and cloaca were determined (Fig. (Fig.1).1). The titers are the averages obtained from one infected duck and one contact duck on day 3 p.i. Wild-type virus was shed in the highest quantity, and VN/1203 LL virus and VN/1203 SS virus were shed at similar levels from both the trachea and cloaca. Therefore, replication efficiency was intact, and all contact ducks were infected, indicating transmission of virus.
To determine if the viruses were all pathogenic for chickens, the IVPI was determined. Chickens were inoculated intravenously with VN/1203 wild-type, VN/1203 SS plaque, and VN/1203 LL plaque viruses. All of the viruses had an IVPI score of 3.0, and all of the inoculated chickens (n = 10) died within 24 h postinoculation.
Two ferrets were inoculated with 106 EID50 of each virus. All ferrets inoculated with VN/1203 wild-type or VN/1203 LL virus died (Fig. (Fig.2A)2A) after losing >20% of their body weight (Fig. (Fig.2B).2B). These findings are similar to those observed previously with VN/1203 wild-type virus (11) and with reverse genetically derived VN/1203 virus (30). Interestingly, VN/1203 SS virus caused no mortality, although the ferrets became very ill and lost >20% of their body weight (Fig. 2A and B) before beginning to recover and gain weight after day 10 p.i. Nasal wash titers on days 3 and 5 p.i. were very similar for all three viruses (Fig. (Fig.2C).2C). By day 7 p.i., only VN/1203 SS virus was being shed, and its titers remained high. Although the titers varied, all of the viruses replicated to a similar mean level in the upper respiratory tract, and the decreased mortality caused by VN/1203 SS virus did not reflect reduced replication.
We next inoculated BALB/c mice with a low dose of virus (approximately 100 EID50). The VN/1203 wild-type and VN/1203 LL viruses caused mortality in 4/9 mice, while the VN/1203 SS virus caused no mortality. Not all of the inoculated mice lost weight, indicating that not all were infected with this low dose of virus.
After determining that the VN/1203 LL and VN/1203 SS viruses differed in their pathogenicity, we sequenced their genomes to determine what changes might be responsible for this difference. Table Table22 shows the coding changes found for the three viruses. The VN/1203 LL virus differed from the VN/1203 wild-type sequence in one coding base pair change in the HA gene. The VN/1203 SS virus had five coding changes (two base changes in the HA gene, two in the PB1 gene, and one in the PA gene) compared to the wild-type virus sequence.
To determine which of the five amino acid changes in the VN/1203 SS virus contributed to its reduced pathogenicity in mallards, we used the eight-plasmid reverse genetics system (17). We used the previously generated eight-plasmid set encoding individual genes of A/Vietnam/1203/04 (30). The rg VN/1203 virus was tested for pathogenicity in mallards. In the first experiment, 2/4 ducks died, and 3/4 showed clinical signs of disease. In the second experiment, 1/4 ducks died, and all four showed clinical signs. Therefore, the rg VN/1203 was less pathogenic to mallards than the VN/1203 wild-type virus but had a pathogenicity profile similar to that of the VN/1203 LL virus. To further test the pathogenicity in mallards, IVPI tests were performed using the VN/1203 wild-type virus and the rg VN/1203 virus. The VN/1203 wild-type virus had an IVPI score of 2.32. In comparison, the rg VN/1203 virus had an IVPI score of 1.76.
Salomon et al. (30) previously reported that rg VN/1203 virus caused high mortality in ferrets and mice. That study's results were similar to those obtained with the VN/1203 wild-type (11) and the VN/1203 LL viruses tested in this study.
After determining that the pathogenicity patterns of the rg VN/1203 and the VN/1203 LL viruses were similar, we used site-directed mutagenesis to make point mutations in the rg VN/1203 genes that mimic those in the VN/1203 SS virus. Five reverse genetic viruses were constructed, each containing one point mutation in the rg VN/1203 genome, as follows: rg VN/1203 HA K52T, rg VN/1203 HA2 A544V, rg VN/1203 PA T515A, and rg VN/1203 PB1 K207R or rg VN/1203 PB1 Y436H.
Several early studies showed that the amino acids in the surface proteins play a major role in determining the pathogenic phenotype of avian influenza viruses (25, 28, 29, 39, 41). We therefore focused on the changes in the HA gene as the most likely causes of the decreased pathogenicity of the VN/1203 SS virus. The two HA mutant reverse genetic viruses (rg VN/1203 HA K52T and rg VN/1203 HA2 A544V) were tested. The rg VN/1203 HA K52T virus killed 3 of 4 inoculated mallards in the first experiment; in the survivor, cloudy eyes were the only sign of disease, and virus was shed until day 7 p.i. (Table (Table1).1). One of four mallards inoculated with rg VN/1203 HA2 A544V died; the rest showed no clinical signs and shed virus until day 7 p.i. In the second experiment, rg VN/1203 HA K52T resulted in the death of only one duck; the remaining three recovered and shed virus until day 7 p.i. (Table (Table1).1). The rg VN/1203 HA2 A544V virus again resulted in the death of only 1 duck. The surviving three ducks showed signs ranging from cloudy eyes to neurologic disorders and shed virus until day 7 p.i. All of the contact ducks in both experiments shed virus, indicating efficient transmission. Virus titers from the trachea and cloaca showed that both viruses replicated well and were released into the environment (Fig. (Fig.1).1). There were large differences between individual mallards in the amount of virus shed (up to 3 log units in the rg VN/1203 HA K52T group), reflecting both individual variability and differences between inoculated and contact ducks.
The two reverse genetic viruses were also tested in mice. Death occurred in 7/9 mice inoculated with rg VN/1203 HA K52T and in 2/9 mice inoculated with rg VN/1203 HA A544V. The rg VN/1203 HA K52T virus caused more weight loss and more severe clinical signs. Because both of these viruses were lethal to mice and mallards; they were not tested in ferrets.
Both changes in the HA gene slightly reduced the mortality observed for mallards. In mice, the rg VN/1203 HA K52T virus was more pathogenic, although both viruses caused mortality. Therefore, the HA gene alone cannot explain the reduced pathogenicity of the VN/1203 SS virus.
We next inoculated mallards with reverse genetic viruses containing the remaining three amino acid differences. The rg VN/1203 PA T515A virus caused no mortality in either the first or second experiment (Table (Table1).1). The only clinical sign of illness was cloudy eyes in some ducks. All of the contact ducks were infected and continued to shed virus until days 5 to 7 p.i. This experiment demonstrated that the single T515A mutation in the PA gene can reduce the pathogenicity of the virus. To determine the effect of delivery via a different route, we inoculated five ducks intravenously with rg VN/1203 PA T515A virus. Three of the five ducks died. Therefore, the rg VN/1203 PA T515A could still be lethal to mallards.
The next two mutations evaluated occurred in the PB1 gene. The rg VN/1203 PB1 K207R virus caused the mortality of 1/4 birds in the first experiment and 3/4 birds in the second (Table (Table1).1). In the first experiment, the two inoculated ducks became ill, had cloudy eyes, and appeared depressed but recovered completely, whereas the contact birds became severely ill, and one died. In the second experiment, all inoculated birds became ill, and the one surviving bird showed neurologic signs. All of the contact birds were infected, and all birds shed virus until day 5 to 7 p.i. When inoculated i.v., the rg VN/1203 PB1 K207R virus resulted in the death of 4/5 birds. These results indicate that the K207R mutation in the PB1 gene is not responsible for the decreased pathogenicity of the VN/1203 SS plaque virus.
The final reverse genetic virus tested had a Y436H point mutation in the PB1 gene. This virus caused no mortality in either experiment (Table (Table1).1). All of the ducks remained healthy in the first experiment; in the second, two inoculated birds became slightly ill (cloudy eyes and depression), while the two contact birds remained healthy. Interestingly, one of the contact ducks in the first experiment did not start shedding virus until day 7 p.i. In the second experiment, one of the contact ducks never shed virus, indicating that the duck had not become infected. Therefore, the mutation at position 436 of the PB1 gene appears to interfere with transmission of the virus. After i.v. inoculation, the rg VN/1203 PB1 Y436H virus caused the death of 1/5 mallards, indicating that it retained the capacity to cause mortality.
Virus titers were determined for the tracheal and cloacal swabs of ducks inoculated with the three reverse genetic viruses with polymerase gene mutations (Fig. (Fig.1).1). Interestingly, ducks inoculated with the rg VN/1203 PB1 Y436H virus were shedding virus (except for one duck in the second experiment that never shed virus). When the swab medium was diluted from both inoculated and contact ducks to determine the viral titers, the levels were below detectable (10 EID50/ml) for the rg VN/1203 PB1 Y436H virus.
Death occurred in 3/10 mice inoculated with rg VN/1203 PA T515A, in 1/10 mice inoculated with rg VN/1203 PB1 K207R, and in 0/10 mice inoculated with rg VN/1203 PB1 Y436H. As observed for the above studies, not all of the inoculated mice lost weight and, therefore, not all were infected. Only 3/10 mice inoculated with the rg VN/1203 PB1 Y436H virus lost weight. These results indicated that only the Y436H mutation in the PB1 gene was responsible for the decreased pathogenicity of VN/1203 SS in mice and that the mutation at position 515 affected pathogenicity only in mallards.
Parallel pathogenicity tests were conducted in ferrets. Both ferrets inoculated with rg VN/1203 PA T515A lost weight and became visibly ill; one died, and one remained ill until the end of the study (Fig. 3A and B). In comparison, rg VN/1203 PB1 Y436H caused no mortality, but both ferrets became ill, lost weight, and had an increased temperature until day 5 p.i., when they began to recover rapidly. Both of these reverse genetic viruses replicated in the upper respiratory tract of the ferrets (Fig. (Fig.3c).3c). These results were similar to those observed for mice: only the mutation at position 436 of the PB1 gene was individually linked to decreased pathogenicity.
The virus plaque morphology in MDCK cells was characterized. The yields of all eight VN/1203 wild-type, plaque-purified, and reverse genetic viruses in MDCK cells were comparable (106 to 107 PFU/ml). However, their plaque morphologies differed (data not shown). The VN/1203 wild-type, VN/1203 SS, VN/1203 LL, and rg VN/1203 HA K52T viruses all grew both large and small plaques on MDCK cells. In comparison, the rg VN/1203 HA A544V, rg VN/1203 PA T515A, rg VN/1203 PB1 K207R, and rg VN/1203 PB1 Y436H viruses grew only small plaques. Therefore, plaque size was not correlated with the pathogenicity observed for mallards, mice, or ferrets. This result differs from those previously reported (30). We further assayed the growth of the VN/1203 wild-type, VN/1203 SS, and VN/1203 LL viruses in multiple replication cycles in MDCK cells (data not shown). No differences were observed, indicating that the viruses had similar levels of replication efficiency.
To determine whether the differences observed for pathogenicity reflected differences in tissue tropism in mallards, virus titers were determined for the brain, heart, pancreas, and lungs of two inoculated birds on day 6 p.i. We hypothesized that the lethal viruses would replicate in the brain and that the less pathogenic viruses would not. As shown in Fig. Fig.4,4, detectable titers of all of the viruses were found in brain, heart, and lungs. The VN/1203 SS virus had the lowest titer in each of these organs, whereas the VN/1203 wild-type virus had the highest titers and was the only virus detected in pancreas. All of the other viruses had similar titers in brain and lung. The VN/1203 LL and rg VN/1203 PB1 Y436H viruses showed higher titers in the heart, but this difference was not statistically different. All of the viruses replicated to similar levels in the tissues tested, indicating that the differences in pathogenicity did not reflect differences in viral replication in these organs.
Histopathologic analysis of the lung, heart, pancreas, and brain of these ducks showed changes in all of these organs (data not shown). The severity and frequency of lesions varied with the virus' virulence. Findings were characterized by lymphoplasmocytic and histiocytic inflammation and included meningoencephalitis, myocarditis, pancreatitis, and lymphoplasmocytic pulmonary infiltrates. In ducks inoculated with the least virulent viruses, myocarditis was never observed, and pancreatitis, pulmonary infiltrates, and meningoencephalitis were minimal and sporadic. As virulence increased, mononuclear inflammation was consistently observed for all examined organs. Higher virulence was correlated with more severe meningoencephalitis and more extensive myocarditis and myocardial necrosis.
A viral untranscribed region-driven luciferase reporter gene assay was performed to compare the polymerase transcription/replication activity levels of the wild-type polymerase complex (PB2, PB1, and PA), the VN/1203 SS polymerase complex, and the polymerase complex with a single mutation (16, 30). The VN/1203 wild-type polymerase complex had statistically approximately twofold higher luciferase activity (relative light units [RLUs]) than that of the VN/1203 SS plaque polymerase complex (analysis of variance, t test, P < 0.05) (Fig. (Fig.5).5). Each of the point mutations in the polymerase complex (PA T515A, PB1 K207R, and PB1 Y436H) caused a decrease in luciferase activity, indicating that each contributes to the lower activity in the VN/1203 SS virus polymerase complex. Interestingly, the rg VN/1203 PB1 K207R polymerase complex induced a level of luciferase activity similar to that of the rg VN/1203 PB1 Y436H complex (2.76 versus 2.71 RLU, respectively). The virus with the mutation at position 207 was still lethal to mallards and mice, whereas the virus with the mutation at position 436 was not. Therefore, the decreased polymerase activity was not directly related to decreased pathogenicity of the VN/1203 virus.
A/Vietnam/1203/04 (H5N1) virus isolated from a human and grown in chicken embryos produced a heterogeneous virus population that formed two types of plaques in MDCK cells, differing in size and in pathogenicity for ducks, ferrets, and mice (summarized in Table Table3).3). Large-plaque isolates, like the wild-type virus, were highly pathogenic in ducks, ferrets, and mice, whereas the small-plaque isolate was nonpathogenic in ducks, mice, and ferrets. Sequence analysis of the entire genome revealed five encoded amino acid differences between these two isolates. We utilized reverse genetics to investigate which residues in the large-plaque (high-pathogenicity) virus would convert the phenotype to low pathogenicity in ducks, mice, and ferrets.
The rg construct of the A/Vietnam/1203/04 (H5N1) virus behaved essentially like the large-plaque virus. The rg wild-type construct killed fewer ducks than the original virus but made all ducks extremely ill and behaved phenotypically like the large-plaque isolate. The slight differences in mortality in the ducks probably reflect the biological heterogeneity of the outbred wild mallard ducks used in the studies.
Neither of the HA gene mutations (K52T or A544V), when inserted into the wild-type virus, markedly affected the properties of the rg viruses; neither change prevented mortality in mallard ducks or in mice. The HA2 construct (A544V) killed fewer ducks but was transmitted efficiently to contacts, and the birds had neurological disease and cloudy eyes. These results show that the HA gene is not individually responsible for the low pathogenicity of the small-plaque viruses in ducks and mice.
The mutations in the PA (T515A) and PB1 (Y436H) genes abolished the pathogenicity of the constructs in ducks. However, these constructs still replicated in ducks, were transmitted to contacts, and were able to kill ducks if inoculated intravenously. The latter finding indicates that an early step in infection was probably affected.
It is noteworthy that the virus with the Y436H mutation in PB1 was not transmitted as efficiently as other viruses in ducks; virus was detected late in the contact ducks (7 days postexposure), and virus was not detected in tracheal or cloacal samples from multiple inoculated and contact ducks. Therefore, the Y436H mutation in PB1 compromises transmissibility.
Pathogenicity studies in ferrets and mice were previously performed, comparing two different virus isolates from the same human patient (A/Vietnam/1203/04 [H5N1] and A/Vietnam/1204/04 [H5N1]). These two viruses differed by eight amino acids, including amino acid position 627 in the PB2 gene (K or E), yet both viruses were highly pathogenic in mice and ferrets (23). These two isolates from the same patient also contained the R207K and H436Y differences in the PB1 gene that were found in the present study.
The A/Vietnam/1203/04 (H5N1) virus with a lysine at residue 627 of PB2 was 40-fold more virulent in mice than the A/Vietnam/1204/04 (H5N1) virus that possesses a Glu at this residue. All of the viruses used in this study had a lysine at residue 627 of the PB2 gene. While the A/Vietnam/1203/04 (H5N1) strain containing the Lys627 was slightly more pathogenic in mice, both isolates (possessing either Glu or Lys at residue 627) were highly pathogenic in ferrets. Thus, the molecular requirements for high viral pathogenicity differ depending on the host.
It is noteworthy that a single amino acid change in the PA (T515A) gene converted a lethal virus to a nonlethal virus in ducks but retained high pathogenicity for mice and ferrets. Little is known about the molecular basis of transmissibility, but the PB1 Y436H change appears to have made the virus less fit. However, the small-plaque variant itself was transmitted efficiently in ducks, suggesting that other molecular interactions can overcome the problem.
Although the changes in the PA (T515A) and PB1 (Y436H) genes reduced lethality in ducks inoculated by natural routes, both of these viruses killed ducks when they were injected intravenously. This finding indicates differences in compatibility of replicative precursors in cells available at the surface of the upper respiratory and digestive tracts and after systemic exposure by intravenous injection. Thus, host cell tropism contributes to the complexity of pathogenesis.
Although we found that differences in lethality in ducks were associated with differences in the PA (T515A) and PB1 (Y436H) genes, we did not identify the mechanism involved. Our evidence suggests that the viruses with these changes replicate to the same levels in the same organs. Thus, the wild-type virus, the small-plaque isolate, and the large-plaque isolate had indistinguishable replication levels both in ducks and in ferrets. Possible explanations are (i) that the differences are immunological and cytokine based or (ii) that these are differences in polymerase activity and that additional testing is needed to determine whether it takes longer for the virus to get to different tissues.
The association of high pathogenicity in ducks with the polymerase genes (PA and PB1) is in keeping with earlier findings that the polymerase genes contain important determinants of pathogenicity (9, 11, 14, 24, 30, 32, 38). However, the residues involved differ between hosts. As mentioned above, residue 627 of PB2 is important in determining pathogenicity in mice but not in ferrets. Here we have shown that a single change in the PA gene (T→A) reduces pathogenicity for ducks but not for mice or ferrets. It is clear that high pathogenicity involves a complex of viral and host genes, but in most cases, the polymerase genes in concert with the NS and HA genes play a role; the connection between this role and the type of host is unresolved.
While H5N1 viruses are highly pathogenic in some duck species (e.g., tufted ducks), the highly pathogenic phenotype appears to be selected against in others, such as the mallard and the grazing ducks of Thailand and Vietnam. Highly pathogenic avian influenza viruses (H5 and H7) must over the centuries have spread to ducks, yet the influenza virus phenotype of high pathogenicity has not previously been perpetuated in wild aquatic birds. It seems unlikely that in the long term, HP H5N1 viruses will be perpetuated in the aquatic bird reservoir. However, the continuing isolation of HP H5N1 viruses from domestic ducks since 1996 is of concern, as the domestic duck (3, 5, 6) may perpetuate H5N1 virus that is nonpathogenic in these ducks but that remains pathogenic for other species.
The present study provides the first insight into the high pathogenicity of H5N1 viruses in ducks, but much remains to be done. An important unresolved question is whether the HP H5N1 virus is currently perpetuated in migrating or in domestic waterfowl. Detailed prospective surveillance of both domestic and wild waterfowl is needed to answer this question.
These studies were supported by U.S. Public Health Service grants AI95357 and CA21765 and by the American Lebanese Syrian Associated Charities (ALSAC).
The exchange of H5N1 viruses between southeast Asian countries and our laboratory was facilitated by the World Health Organization. We thank David Walker and Ashley Baker for excellent technical assistance. We thank Carol Walsh for manuscript preparation and Sharon Naron for editing the manuscript.
Published ahead of print on 6 June 2007.