Here, we report a detailed analysis of a large outbreak of H5N1 avian influenza virus occurring in migratory waterfowl from late April through June 2005 in the Qinghai Lake region of western China. In contrast to previous reports of viruses isolated during this outbreak (2
), our studies reveal a marked heterogeneity among the causative viruses. For example, sequence analyses of 15 viruses isolated from six different avian species showed that at least four genotypes were responsible for the outbreak. All of the viruses were highly lethal to chickens, and seven of the eight test viruses replicated systemically and were highly lethal in mice. We also found that the viruses isolated early in the outbreak possessed a typical avian virus signature amino acid at position 627 of PB2, Glu, unlike later isolates, which had Lys at this position (2
). Moreover, these index viruses possessed a phylogenetically distinct PB2 gene compared with those of other Qinghai isolates. This suggests that the virus introduced into the Qinghai Lake waterfowl population may have possessed the index virus-like PB2 gene but that during the outbreak, it acquired a PB2 gene with Lys at position 627. Alternatively, the newly introduced virus may have already possessed PB2 with Lys at this position, but viruses with the mutant type of PB2 were not detected until 10 May.
A distinct temporal pattern of infection of different avian species by the H5N1 viruses was also apparent (Fig. ). Bar-headed geese were the first species to be affected, followed by brown headed gulls and great black-headed gulls about 10 days later (13 May) and then by ruddy shelducks and great cormorants after another 10 days (24 and 25 May). The time between the detection of small numbers of deaths (13 and 16 May) and the detection of considerably larger numbers of deaths (24 and 26 May) of ruddy shelducks and great cormorants was also about 10 days. These findings could be interpreted to indicate stepwise introduction of the virus into different avian species in the lake. Our sequence analyses revealed that at least three genotypes of H5N1 viruses were circulating among bar-headed geese, while the viruses isolated from great black-headed gulls, brown-headed gulls, great cormorants, and whooper swans were similar to each other and belonged to only one of the genotypes found in bar-headed geese (Fig. ). Viruses representing genotypes A and B were isolated from the bar-headed geese that died early during the outbreak and were likely not spread to other species. We speculate that viruses of genotype D may also have been present in bar-headed geese at the beginning of the outbreak but were not identified because of the limited number of dead birds analyzed.
The origin of the virus responsible for the Qinghai Lake outbreak remains unclear. The disease was first recognized in bar-headed geese (Fig. and ), suggesting at least two possible mechanisms for the introduction of the H5N1 virus into wild-bird populations by this species using the lake as a habitat. One possibility is that the virus was carried to the lake by other wild birds not susceptible to H5N1 infection and was then transmitted to bar-headed geese. Another possibility is that bar-headed geese infected elsewhere were the species that brought the virus to Qinghai Lake, presumably via the East Asian-Australian flyway or the Central Asian-Indian flyway. If the first scenario is correct, the virus should have been transmitted to all susceptible species at the same time, including brown-headed gulls and great black headed gulls, which congregate with bar-headed geese in the islet where H5N1 virus-infected bar-headed geese were found. The fact that the disease was identified in these two species of gulls approximately 10 days after the discovery of fatal cases of H5N1 infection among bar-headed geese supports the second scenario.
Importantly, a genotype C virus, which was found in multiple species in Qinghai Lake, was responsible for the wild-bird outbreak of H5N1 infection in Mongolia and Russia in August 2005 and also caused major outbreaks in chickens in the Liaoning Province and Inner Mongolia in October and November 2005 (Fig. ; see Fig. S2 in the supplemental material), suggesting that viruses of this genotype may be more pernicious than those of other genotypes. These findings call for intensive surveillance of wild migrating birds as biologic vectors that possibly spread H5N1 viruses over a wide range of territories.
It is important that viruses of genotype C possess Lys at position 627 in PB2, unlike any other avian viruses, and that this residue is found in some human H5N1 (7
) and H7N7 (5
) isolates as well as in the virus responsible for the Spanish influenza pandemic (21
). Thus, it is worrisome that H5N1 viruses with a trait associated with human adaptation have entered into migrating waterfowl populations.
In the present study, seven of eight test viruses replicated systemically and killed mice. Among these seven lethal viruses, five have a lysine at position 627 of the PB2 protein and one has asparagine at position 701 of this molecule. Although both of these changes are associated with high virulence in mice (6
), our findings indicate the existence of additional mutations that contribute to the virulence of avian H5N1 viruses in mammals.
Several different animal models have been used to evaluate the virulence of avian H5N1 influenza viruses in mammals. Maines et al. demonstrated the general equivalence of mice and ferrets for assessing the pathogenic potential of H5N1 viruses isolated from chickens and humans in Thailand and Vietnam (13
). Experimental infection of cynomolgous macaques with the index H5N1 virus from the 1997 outbreak in Hong Kong resulted in acute respiratory distress syndrome and multiple-organ dysfunction, which was similar to findings in humans (18
). By contrast, the clinical signs produced in rhesus macaques by infection with two of the Qinghai Lake isolates were quite mild. Although the A/duck/Guangxi/35/01 virus did cause systemic infection and symptoms of influenza-like illness, the extent of the disease as judged by both its clinical symptoms and histopathology was milder than that reported previously by Rimmelzwaan et al. (18
). This discrepancy may reflect the experimental procedures used by Rimmelzwaan et al. and our group. While we intranasally infected macaques with 2 ml of virus in fluid, Rimmelzwaan and colleagues used 5 ml of viral fluid, applying 4 ml intratracheally, 0.5 ml to the tonsils, and 0.25 ml to each of the conjunctiva, which would be expected to induce more severe disease than that seen in our study.
In conclusion, the H5N1 viruses that caused a massive outbreak of lethal disease among wild birds at Qinghai Lake in western China represent a phylogenetically and biologically heterogeneous group reiterating the features of H5N1 viruses now circulating in nature. The fact that viruses with a PB2 mutation associated with human adaptation of avian viruses are circulating in migratory waterfowl and that an avian H5N1 virus was capable of causing systemic infection in primates is worrisome. Moreover, migratory waterfowl may possibly spread these viruses over a wide range of territories. If viruses with the ability to replicate systemically in primates establish in migratory waterfowl, there would be an even more critical need for increased surveillance of poultry and the development of control measures.