Introduction

Avian influenza A viruses of the H9N2 subtype were first isolated in turkeys in the United States in 1966 [20]. Now, this subtype of virus has been circulating worldwide in multiple avian species and endemic in poultry populations across Eurasia [1, 5, 55]. It is noteworthy that H9N2 influenza viruses in poultry have occasionally been transmitted to mammalian species, including humans and pigs [8, 39], since this subtype was first reported to be isolated from pigs in Hong Kong and from patients with influenza-like illness in Guangdong Province of China in 1998 [17, 32, 39]. Several serological surveys suggest that human H9N2 infections could be more prevalent than has been reported, and possible human-to-human transmission cannot be completely excluded [8, 17, 22]. Moreover, human H9N2 infections produce a typical human-influenza-like illness that can easily be overlooked [8, 32], so they may have a greater opportunity to adapt to humans and acquire the ability of transmission in humans. Therefore, H9 influenza virus, along with H5 and H7 viruses, are in the World Health Organization’s (WHO) list of leading candidates that may potentially cause another human influenza pandemic.

During the last two decades, phylogenetic analysis has revealed that H9N2 viruses have undergone extensive inter- and intrasubtype reassortments to generate multiple novel genotypes with gene segments from different H9N2 lineages or influenza A viruses of other subtypes [13, 55]. Reassortment is an important mechanism for the generation of a pandemic influenza strain. The pandemic influenza viruses of 1957, 1968 and 2009 (pH1N1) emerged through genetic reassortment of human viruses with swine and/or avian viruses [21, 46]. In addition, the highly pathogenic H5N1 viruses in Hong Kong in 1997 and the novel H7N9 viruses in the current outbreak in China are also reassortants that derived viral genes from different subtypes of avian influenza viruses [10, 12, 33]. More importantly, a recent study showed that H9 reassortants between avian H9N2 virus and the 2009 pH1N1 virus exhibit higher pathogenicity to mice than either of the parental viruses [48]. It is also documented that H9N2 surface genes reassorted with pH1N1 internal genes can be transmitted efficiently via respiratory droplets, creating a clinical infection similar to human influenza virus infections in the ferret model [27].

In addition to reassortment, mutation is another important mechanism for producing a new pandemic influenza virus. It is likely that the 1918 pandemic H1N1 virus derived all eight of its genes from an avian virus and then evolved by accumulating mutations during adaptation in a mammalian host [41]. Some of the mutations in viral genes that have been implicated in the adaptation of avian influenza viruses to mammals have been identified and vary according to the virus strain and host [40, 47, 49]. The HA mutation Q226L in the receptor-binding site (RBS) has been shown to facilitate preferential binding to human-like α2, 6-linked sialic acid (SA2,6) receptors [36, 51] and is regarded as one of the key elements for successful infection of humans by influenza viruses. Interestingly, an increasing number of contemporary avian H9N2 viruses contain L at position 226 of the HA protein [36, 53]. Furthermore, recent research has demonstrated that L226-containing H9N2 viruses replicate efficiently in human airway epithelial cells and are more likely to be transmitted in ferrets, although no aerosol transmission has been observed [52, 53]. Therefore, avian H9N2 viruses may impose threats to human health in the future by reassortment or mutation

In August 2008, a novel reassortant H9N2 avian influenza virus (DK1), which contains a human-like PB2 segment, was isolated from apparently healthy domestic mallard ducks in Jiangsu Province of Eastern China [56]. Although the human-like PB2 segment contains 627 K, which is recognized as a critical virulence determinant of H9N2 avian influenza virus in mice [30], the DK1 virus is nonpathogenic for mice. To determine whether the virulence of the DK1 virus could be increased by mutation in a mammalian host, we serially passaged the virus in mice and tested the mouse-adapted virus (DK1-MA) obtained from the passage 18 population for its pathogenicity in mice and genomic sequence mutations. We found that the DK1-MA virus is highly pathogenic to mice and accumulates eight mutations in six of the ten viral proteins, highlighting the potential threat of H9N2 viruses to humans.

Materials and methods

Virus and animals

A/duck/Jiangsu/01/2008 (H9N2) (DK1) virus was isolated from apparently healthy domestic mallard ducks in Jiangsu Province of eastern China in January 2008 [56]. Viruses were plaque purified twice in Madin-Darby canine kidney (MDCK) cells and propagated once in specific-pathogen-free (SPF) embryonated chicken eggs. Its infectivity titer was determined by calculating the 50 % egg infectious dose (EID50). The virus preparation was stored at −70 °C until use. Female BALB/c mice (6-8 weeks old) were used for virus adaptation and evaluation of virulence in this study. All animal experiments were approved by the Jiangsu Administrative Committee for Laboratory Animals and complied with the guidelines of laboratory animal welfare and ethics of Jiangsu Administrative Committee of Laboratory Animals. All experiments involving live viruses and animals were carried out in negative-pressure isolators with HEPA filters in a biosafety level 3 (BSL3) animal facility at Yangzhou University, in accordance with the institutional biosafety manual.

Adaptation of H9N2 avian influenza virus in mice

Adaptation was carried out by lung-to-lung passaging, a classical method described previously [7]. Briefly, mice were lightly anesthetized with pentobarbital natricum (40–50 mg/kg) and inoculated intranasally with 30 µl of allantoic fluid containing wild-type H9N2 virus. Three days post-inoculation (p.i.), lungs of the virus-inoculated mice were harvested and homogenized in 1 ml of phosphate-buffered saline (PBS) containing 100 U penicillin and 100 µg streptomycin per ml, and 30 µl of the centrifuged supernatant was used as the inoculum for the next passage. After a total of 18 passages, virus present in the lung homogenate was cloned twice by plaque purification in MDCK cells to create a mouse-adapted mutant, named DK1-MA, and the DK1-MA virus was passaged once in the allantoic cavities of 10-day-old chicken eggs for 72 h at 37 °C to prepare a virus stock. The 50 % mouse lethal dose (MLD50) was determined for the passage 5, 10, 15, 18 viruses and the DK1-MA virus by intranasal inoculation of groups of five mice with 10-fold serial dilutions of viruses in PBS. The mice were monitored for 14 days. MLD50 was calculated and expressed in EID50 units.

Mouse studies

To evaluate the virulence of the mouse-adapted viruses, groups of 6-week-old female BALB/c mice were anesthetized with pentobarbital natricum and inoculated intranasally with viruses at the indicated doses in 30 µl PBS or mock inoculated with PBS to serve as controls. Body weight and survival of mice were recorded daily for 14 days. Mice that showed severe symptoms or greater than 25 % weight loss were euthanized and scored as dead for humane reasons.

To evaluate the replication of the viruses, groups of 6-week-old female BALB/c mice were inoculated intranasally with 105 EID50 of the indicated viruses. To determine virus replication in the lungs of infected mice, three mice in each group were euthanized at 1, 2, 3, and 5 days p.i. Whole lungs were removed and homogenized in 1 ml of PBS for virus titration in 10-day-old embryonated eggs as described previously [29]. To determine the tissue distribution of virus in mice, samples of heart, kidney, lung, liver, spleen and brain were collected at 3 and 5 days p.i., and whole tissues were homogenized in 1 ml of PBS for virus titration. Viral titers were determined in 10-day-old embryonated eggs and expressed as mean log10EID50/ml ± standard deviation (SD). The limit of virus detection was 0.8 log10EID50/ml.

Lung injury was assessed by testing lung water content and differential leukocyte counts in bronchoalveolar lavage (BAL) fluid of mice infected with 105 EID50 of the indicated viruses. To determine lung water content, three mice per group were weighed and sacrificed at 1, 3, and 5 days p.i. The whole lungs were surgically dissected, blotted dry, and weighed immediately (wet weight). The lung tissue was then dried in an oven at 80 °C for 72 h and reweighed to determine the dry weight. The lung wet/dry weight ratios were calculated for each animal to assess tissue edema as described previously [28]. To determine differential leukocyte counts, BAL cells were obtained from three euthanatized mice in each group at 1, 3, and 5 days p.i. according to the protocol described by Majeski et al. and Nick et al. [35, 38]. In brief, the lungs were lavaged twice with a total 1 ml saline (4 °C) through the endotracheal tube. The recovery rate of BAL fluid was >90 % for all tested animals. After the amount of fluid recovered was recorded, an aliquot of BAL fluid was diluted 1:1 with 0.01 % crystal violet dye and 2.7 % acetic acid for leukocyte staining and erythrocyte hemolysis. The leukocytes in BAL fluid were counted using a haemacytometer under a microscope. The remaining BAL fluid was centrifuged for 10 min at 300 × g. Differential cell counts were determined by Wright staining of a spun sample, on the basis of morphological criteria under a light microscope, with evaluation of at least 200 cells per slide. All slides were counted twice by different observers blinded to the status of the animal.

Sequence analysis

Viral RNA was extracted from virus-infected allantoic fluid using an RNeasy Mini Kit (QIAGEN) according to the manufacturer’s protocol. Reverse transcription was performed using the Uni12-primer (AGC AAA AGC AGG) by standard methods. After reverse transcription, PCR amplification of cDNA was done using previously described primers [19]. After agarose gel electrophoresis, PCR products were excised and purified using a gel extraction kit (Tiangen, Beijing, China), and sequence analysis was done by Sangon Biotechnology Co., Ltd., (Shanghai, China). Nucleotide and deduced amino acid sequences were compiled, edited and aligned using the Lasergene sequence analysis software package (DNASTAR, Madison, WI, USA). The GenBank accession numbers for the DK1 segments are KF142478 to KF142485, and those for the DK1-MA segments are KF881737 to KF881744.

Results

Adaptation of H9N2 avian influenza virus to mice

A novel natural reassortant H9N2 influenza virus (A/duck/Jiangsu/1/2008, DK1) [56] is able to infect but is otherwise avirulent in mice. To increase its virulence in mice, the DK1 virus was serially passaged in BALB/c mouse lungs. Virulence was assayed by determining the MLD50 after 5, 10, 15 and 18 passages of the total population of virus obtained directly from lung extracts without further culturing (Table 1). The parental virus was completely avirulent for mice (MLD50 > 107.17 EID50). However, after five passages, the MLD50 for the population of virus in lung extracts had decreased by >102-fold; and after 15 passages, by >103-fold. Furthermore, after 18 passages, the virus had an MLD50 of 102.5 EID50. According to the criteria that a virus with an MLD50 value of <103.0 EID50 is regarded as highly pathogenic to mice [24], the passage 18 virus population had acquired key mutations that profoundly affect virulence. A mouse-adapted virus (DK1-MA) was obtained from the passage 18 population by two rounds of plaque purification in MDCK cells.

Table 1 Virulence change of H9N2 virus during mouse adaptation
Full size table

Pathogenicity of DK1-MA virus in mice

The virulence of the DK1-MA variant was compared with its parental virus, DK1, in mice. We observed that DK1-MA was >104.8-fold more virulent than DK1 virus, with an MLD50 of 102.37 EID50. Another study was conducted to compare morbidity and mortality in mice infected with these viruses. Mice infected with DK1-MA exhibited clinical signs of disease, including decreased activity, huddling, heavy/labored breathing, hunched posture, and ruffled fur, and began to lose weight at day 4 p.i., even at a low inoculation dose of 103.0 EID50 (Fig. 1b). All of the mice in the DK1-MA-infected group showed obvious weight loss and died by day 5 p.i. at a dose of 105.0 EID50 and by day 9 p.i. at a dose of 103.0 EID50 (Fig. 1d). DK1-infected mice displayed only slight weight reduction over a period of 14 days and started to gain weight on day 6 p.i., even at a high dose of 106.0 EID50 (Fig. 1a). All of these mice survived the infection during a 14-day observation period (Fig. 1c).

Fig. 1
figure 1

Pathogenicity of DK1 and DK1-MA viruses in mice. Six-week-old female BALB/c mice (n = 5/group) were inoculated intranasally with different doses of virus (103–106 EID50 of DK1 or 101–106 EID50 of DK1-MA) or mock inoculated (Mock). (a-b) Morbidity was assessed by measuring weight changes over a 14-day period and is graphed as a percentage of the animals’ weight on the day of inoculation (day 0). The average body weight of mice infected with DK/1 (a) or DK-MA (b) is shown. (c-d) Mortality in the DK1 (c) and DK1-MA (d) groups was determined by counting the surviving mice

Full size image

Replication of DK1-MA virus in mice

Female BALB/c mice were inoculated intranasally with 105 EID50 of DK1 and DK1-MA viruses, and the levels of viral replication in mouse lungs were compared. As shown in Fig. 2, the DK1-MA virus replicated to a high titer that was 103.21-fold higher than that of the DK1 virus as early as 1 day p.i., and it sustained higher levels of replication than the DK1 virus throughout the course of infection. The peak viral titer of DK1-MA virus was observed on day 3 p.i., reaching 108.17 EID50/ml, versus a virus titer of 106.0 EID50/ml at 3 day p.i., and reaching peak yield with 106.6 EID50/ml at 5 days p.i. for DK1. These results indicated that the DK1-MA virus grew faster and to significantly higher titers than the DK1 virus, although both viruses could replicate in mouse lungs. To investigate the distribution of virus in different tissues of infected mice, we measured titers of total homogenates from heart, liver, spleen, lung, kidney, and brain on days 3 and 5 p.i. (Table 2). No infectious virus was detected in any organ other than the lung from any of the DK1-infected mice. In contrast, DK1-MA virus was sporadically detected in heart, liver, spleen, and kidney, indicating that the DK1-MA virus had partly gained the ability to replicate in other organs as well as in the lungs. Taken together, our findings showed that the mouse-adapted virus replicated more efficiently than wild-type virus in mouse tissues.

Fig. 2
figure 2

Replication kinetics of DK1 and DK1-MA viruses in mouse lungs. Groups of three mice were inoculated intranasally with 105 EID50 of DK1 or DK1-MA virus, and lungs were collected at 1, 2, 3 and 5 day p.i. for virus titration in eggs. The average of each group is shown, with error bars representing the SD. **, P < 0.01 compared with the value for the DK1 group

Full size image
Table 2 Viral distribution of DK1 and DK1-MA viruses in different tissues in mice
Full size table

DK1-MA virus elicits severe pulmonary edema

Lung wet/dry weight ratios were determined to assess whether the different levels of virulence of DK1-MA and DK1 viruses were related to a difference in ability to elicit pulmonary edema (Fig. 3a). In the DK1-MA-virus-infected group, the lung wet/dry weight ratios did not change significantly on day 1 p.i. but were dramatically elevated on days 3–5 p.i., with increases in ratios of 1.6- and 1.5-fold compared to the mock-infected group. By contrast, there were no statistically significant differences in this parameter between the DK1 infection group and the mock-infected group at any of the time points measured.

Fig. 3
figure 3

Lung wet/dry weight ratios of virus-infected mice. The means ± SD (n=3) at each time point are shown. * indicates p < 0.05 and ** indicates p < 0.01 compared with the mock group

Full size image

Increased numbers of BAL cells in lungs of mice infected with DK1-MA virus

To better characterize the cellular components in the lungs following virus infection, total and differential cell counts in BAL fluid were determined for infected mice on days 1, 3, and 5 p.i. (Fig. 4). The total number of BAL cells recovered from DK1-MA-virus-infected mice increased as early as 1 day p.i., with cell numbers increasing 2.4- to 5.5-fold by day 5 p.i. (Fig. 4a). By contrast, the total BAL cell count in fluid recovered from DK1-infected mice did not indicate an increase in infiltrates at 1 day p.i., but an increase was observed on day 3 p.i. Furthermore, the lungs of mice infected with DK1-MA virus had significantly higher numbers of BAL cells at each time point measured than were found in the DK1 infection groups. During DK1-MA virus infection, peak cellular infiltration in the BAL samples on days 3 and 5 p.i. was associated with a substantial increase in the percentage of neutrophils and lymphocytes compared with that for DK1-infected mice (Fig. 4b, c). These results implied that DK1-MA virus infection induced a predominant neutrophil and lymphocyte infiltrate increase into the lungs that may have contributed to the pathogenesis of the enhanced disease associated with DK1-MA virus infection.

Fig. 4
figure 4

Total and differential cell counts in BAL fluid. Groups of three mice were inoculated intranasally with 105 EID50 of DK1 or DK1-MA virus and BAL specimens were collected at 1, 3 and 5 days p.i. to determine cell counts. Total numbers of BAL cells (a) and percentages of neutrophils (b), lymphocytes (c) and macrophages (d) of DK1- or DK-MA-virus-infected mice are shown as means ± SD. Dotted lines indicate values from mock-inoculated animals. * indicates p < 0.05 and ** indicates p < 0.01 compared with the DK1 virus infection group

Full size image

Sequence analysis

To identify the molecular basis for the enhanced virulence acquired by an H9N2 virus during mouse adaptation, the complete genomes of both the DK1 and DK1-MA viruses were sequenced. Comparison of full genome sequences between DK1 and DK1-MA viruses revealed eight amino acid substitutions, which were mapped to the PB2, PA, HA (H3 numbering used throughout the text), NP, NA, and M genes (Table 3). However, no amino acid substitutions were observed in the PB1 or NS gene. To find out whether the amino acid exchanges observed in DK1-MA also occur in other H9N2 strains, we analyzed the H9N2 sequences deposited in the Influenza Research Database (data were obtained from the National Institute of Allergy and Infectious Diseases database available online at http://www.fludb.org) (Table 3). About one-fifth and one-seventh of the avian isolates share common residues with the mouse-adapted virus in the M1 (M1-166V) and NP (NP-105M) proteins, respectively, whereas only a limited number of human or swine isolates possess the same residues as DK1-MA at these positions, indicating that the DK1-MA virus acquired amino acids normally found in avian viruses during adaptation in mouse lungs. However, HA-80L, HA-193N, NA-27A, PB2-404F, PA-3D, and PA-225S are highly conserved among most avian, swine, and human isolates, although some isolates have mutations at those positions, as does DK1-MA. Therefore, the amino acids found at these positions in DK1-MA were unique to this mouse-adapted virus, which may contribute to its increased virulence.

Table 3 Amino acid differences between the wild-type strain DK1 and the mouse-adapted strain DK1-MA
Full size table

Discussion

Gene mutation and reassortment are key mechanisms for the evolution of influenza viruses, which may allow avian influenza viruses to cross species barriers to infect humans or other mammals and increase virulence [9, 10, 33]. Avian influenza viruses, particularly the H9N2 subtype, which has infected humans and undergone genetic changes [8, 13, 36], should not be overlooked as potential pandemic threats. The present results showed that the novel reassortant H9N2 avian influenza virus, which bears the PB2 segment from human influenza virus [56], could adapt well in mouse lungs and mutate to become highly pathogenic for mice after eighteen serial passages. The DK1-MA virus caused signs of severe disease and resulted in 100 % mortality at a low inoculation dose of 103 EID50. In contrast, infection with the DK1 virus did not cause death or clinical signs of illness.

The primary feature of viruses with adaptive mutations is that they increase in prevalence in the population because of improved replicative fitness [3]. Moreover, most of the previous studies found that enhanced replication is involved in the adaptation of avian influenza viruses to mammals [14, 47]. In the current study, the DK1 virus could be detected only at low titers in the lung of infected mice. However, the DK1-MA virus grew faster and to significantly higher titers in mouse lungs and was detected in other organs as well. All these data support the viewpoint that efficient replication is an important and characteristic prerequisite for high virulence of avian influenza A virus in mice.

Severe pulmonary edema and inflammatory cellular infiltrate are involved in the disease during severe influenza infection [54]. We observed that DK1-MA infection could elicit significantly higher pulmonary edema, while no difference in lung water content was observed between the DK1-infected group and the control group. In addition, the DK1-MA virus resulted in significantly higher of numbers inflammatory cells in BAL than DK1. Moreover, the percentages of neutrophils and lymphocytes in BAL cells increased substantially during DK1-MA infection when compared with DK1. The elevated levels of neutrophils and lymphocytes were also found in the BAL samples of mice infected with highly pathogenic H5N1 viruses and reconstructed pandemic 1918 H1N1 virus [50, 54]. Neutrophils are primary mediator/effector cells involved in producing acute lung injury, and the activated neutrophils release free radicals, inflammatory mediators, and proteases, which lead to severe lung lesions [2, 18]. Lymphocytes are also known to play a critical role in the development of bronchiolitis [16, 23]. Accordingly, enhanced invasion by pulmonary neutrophils and lymphocytes may be associated with the severity of DK1-MA virus infection in mice.

The pathogenesis of influenza A viruses involves multiple genes, and molecular determinants may differ among animal species [9, 25, 40, 42]. However, several molecular determinants have already been identified that govern the virulence of naturally occurring and adapted strains, such as the amino acids at the cleavage site of HA [44], the length of the NA stalk [57], the mutations involved in the ability of NS1 proteins to restrict the induction of the host interferon response [31], and specific amino acid substitutions in the ribonucleoprotein (RNP) complex [14, 43, 47]. In this study, comparison of the predicted amino acid sequences of DK1-MA and DK1 viruses demonstrated eight substitutions involving six of the ten viral proteins (HA, NA, PB2, PA, NP, M1), and none of these changes had been recognized to be related to increased virulence or replication efficiency, with the exception of the N193D mutation in the HA protein [26]. However, all of these substitutions except for HA-L80F were located in recognized functional regions that are involved in interaction of viral proteins or between the virus and host factors. The HA-N193D mutation occurs in the membrane-distal region of the HA receptor binding site and has been shown to result in reduced affinity for 6′-sialyl substrates [26], which may be relevant in that the avian α2,3 receptor is the dominant sialic acid species in the mouse respiratory tract [15]. The NA-A27T mutation resides in the amino-terminal transmembrane domain (TMD) [11, 45], and the M1-A166V mutation resides in the carboxyl-terminal RNP-binding region (165-252aa) [4]. Four mutations (PB2-F404L, PA-D3V, PA-S225R, and NP-D105G) in the RNP complexes were all located in the known functional regions, including the host 7-methyl guanosine cap-binding domain of the PB2 protein, the amino-terminal domain and second nuclear localization sequence (NLS) of the PA protein, and the first segment of the NP body domain [6, 34, 37, 40]. Therefore,multiple amino acid substitutions might contribute to the increased virulence of the H9N2 avian influenza virus during adaptation in mice.

In summary, our data show that a duck-origin H9N2 influenza virus can evolve to be highly pathogenic for mice after serial passage in mouse lungs, which might suggest the potential of H9N2 avian influenza viruses for adaptation to replication and high virulence in other mammals, including humans. After adaptation, multiple amino acid substitutions occurred in specific functional regions of the viral genome, resulting in the rapid growth of H9N2 virus, severe pulmonary edema, excessive inflammatory cellular infiltration in lungs, and death in mice. We are currently undertaking a reverse genetics approach to explore the roles of specific amino acid substitutions in the increased pathogenicity of H9N2 influenza viruses in mice.