Mutations during the Adaptation of H9N2 Avian Influenza Virus to the Respiratory Epithelium of Pigs Enhance Sialic Acid Binding Activity and Virulence in Mice

ABSTRACT

The natural reservoir for influenza viruses is waterfowl, and from there they succeeded in crossing the barrier to different mammalian species. We analyzed the adaptation of avian influenza viruses to a mammalian host by passaging an H9N2 strain three times in differentiated swine airway epithelial cells. Using precision-cut slices from the porcine lung to passage the parental virus, isolates from each of the three passages (P1 to P3) were characterized by assessing growth curves and ciliostatic effects. The only difference noted was an increased growth kinetics of the P3 virus. Sequence analysis revealed four mutations: one each in the PB2 and NS1 proteins and two in the HA protein. The HA mutations, A190V and T212I, were characterized by generating recombinant viruses containing either one or both amino acid exchanges. Whereas the parental virus recognized α2,3-linked sialic acids preferentially, the HA190 mutant bound to a broad spectrum of glycans with α2,6/8/9-linked sialic acids. The HA212 mutant alone differed only slightly from the parental virus; however, the combination of both mutations (HA190+HA212) increased the binding affinity to those glycans recognized by the HA190 mutant. Remarkably, only the HA double mutant showed a significantly increased pathogenicity in mice. In contrast, none of those mutations affected the ciliary activity of the epithelial cells which is characteristic for virulent swine influenza viruses. Taken together, our results indicate that shifts in the HA receptor affinity are just an early adaptation step of avian H9N2 strains; further mutational changes may be required to become virulent for pigs.

IMPORTANCE Swine play an important role in the interspecies transmission of influenza viruses. Avian influenza A viruses (IAV) of the H9N2 subtype have successfully infected hosts from different species but have not established a stable lineage. We have analyzed the adaptation of IAV-H9N2 virus to target cells of a new host by passaging the virus three times in differentiated porcine respiratory epithelial cells. Among the four mutations detected, the two HA mutations were analyzed by generating recombinant viruses. Depending on the infection system used, the mutations differed in their phenotypic expression, e.g., sialic acid binding activity, replication kinetics, plaque size, and pathogenicity in inbred mice. However, none of the mutations affected the ciliary activity which serves as a virulence marker. Thus, early adaptive mutation enhances the replication kinetics, but more mutations are required for IAV of the H9N2 subtype to become virulent.

INTRODUCTION

High genetic variability is a hallmark in the epidemiology of influenza A viruses. Mutations in the surface glycoproteins hemagglutinin (HA) and neuraminidase (NA) are the driving forces of antigenic drift, whereas reassortment of gene segments in progeny viruses is the basis for major antigenic shifts. These genetic changes not only are efficient means of immune evasion but also facilitate interspecies transmission (1–3).

Wild waterfowl is considered the major natural reservoir of influenza A viruses (4, 5). In different bird populations, especially in waterfowl, virus strains have evolved that are classified into 16 different hemagglutinin subtypes, H1 to H16, and nine different neuraminidase subtypes, N1 to N9 (6, 7). Some strains have succeeded in crossing the species barrier to terrestrial birds, poultry, and different mammalian species, e.g., pigs, humans, and horses, and have become established in the new host (8–12). Interspecies transmission may occur directly or via an intermediate host. Pigs have been discussed as a mixing vessel (13) from which influenza A viruses may be transmitted to humans, like the 2009 pandemic H1N1 viruses (14–17).

Apart from the subtypes H1N1, H2N2, and H3N2 that established stable lineages in humans, avian viruses of different subtypes may cause sporadic zoonotic infections, e.g., H5N1, H7N9, and H9N2 viruses (8, 18, 19). Viruses of the last subtype have been reported from poultry since 1990 (20, 21) and have become widespread in recent years. By now, they are circulating worldwide in multiple avian species and have repeatedly infected mammalian species, including pigs and humans. H7N9 viruses that appeared in 2013 and infected more than 600 humans are reassortants with the gene segments coding for internal proteins being derived from an H9N2 strain (18, 22).

Adaptation of influenza viruses to a new host is accompanied by a number of mutations especially in the viral polymerase and in the hemagglutinin (HA), one of the viral surface glycoproteins. Adaptive mutations in the HA protein often affect the receptor-binding site (RBS) (23–27). Sialic acids present on cellular surface glycoproteins or glycolipids serve as receptor determinants (15, 28). Avian strains generally have a preference for binding to α2,3-linked sialic acids, whereas swine and human viruses preferentially recognize α2,6-linked sialic acid, which is predominant in the upper airways of pigs and humans (29–31).

To identify mutations that are associated with the adaptation of avian viruses to growth in pigs, we passaged an avian H9N2 virus in the target cells of the new host, i.e., the differentiated porcine airway cells. The avian strain was passaged three times in precision-cut lung slices (PCLS), an ex vivo culture system of differentiated airway epithelial cells. The virus after the third passage displayed a shorter growth cycle and differed from the parental virus by one mutation each in the PB2 and NS1 proteins plus two mutations in the HA protein. The latter two mutations were characterized in more detail by generating HA mutants carrying one or two mutations. Those exchanges broadened the receptor-binding activity, because the respective hemagglutinins were able to recognize not only α2,3-linked sialic acids but also α2,6- and α2,8/9-linked sialic acids. Moreover, the HA mutants showed different phenotypes in several culture systems and an enhanced pathogenicity in mice. However, none of the mutants had an increased replication efficiency in porcine PCLS, suggesting that further adaptive mutations are required for H9N2 viruses to become virulent for pigs.

(This work was performed by W.Y. and D.P. in partial fulfillment of the requirements for doctoral degrees from the University of Veterinary Medicine Hannover.)

RESULTS

Virus passaging in differentiated airway epithelial cells results in a shortening of the growth cycle.

The target cells for mammalian influenza viruses are respiratory epithelial cells. In a previous study, we established precision-cut lung slices (PCLS) as a culture system for differentiated respiratory epithelial cells from the porcine lung that, upon infection by swine influenza viruses (SIV), reflects the viral virulence properties (32, 33). To investigate the adaptation of H9N2 avian influenza viruses to growth in the respiratory epithelium of pigs, we passaged an avian influenza virus, A/chicken/Saudi Arabia/CP7/98 (H9N2), three times (P1, P2, and P3) in PLCS. Virus of each passage was analyzed for its ciliostatic effect (Fig. 1A) and growth characteristics (Fig. 1B). The parental virus (P0) causes only partial ciliostasis, i.e., only approximately 50% of the epithelium covering the airway in the microscopic field showed ciliary activity (32). This characteristic was maintained in P1 to P3 viruses. As the P2 virus did not differ from the other viruses, only the first and the last virus passage is shown (Fig. 1A). In this respect, H9N2 viruses resembled low-virulence SIV. In contrast, infection of PCLS by virulent SIV results in complete ciliostasis (33). A difference was noted only for replication efficiency. As shown in Fig. 1B, there was no difference in the maximum titer between P1 to P3 viruses, which was about 10⁶ PFU/ml, similar to the value reported for the parental virus (32). However, the growth cycle of the P3 virus was shorter than that of the viruses from the previous passages (Fig. 1B). For P1 and P2, it took until about 48 h postinfection (hpi) to reach the maximum titer, whereas that peak value was found for the P3 virus already at about 36 hpi. The P1 virus did not differ from the P2 virus; therefore, only the P2 and P3 viruses are shown in Fig. 1B.

FIG 1.

Growth characteristics and ciliostatic effect of H9N2 virus passaged three times in precision-cut slices from the porcine lung. (A) Ciliostatic effect of H9N2 influenza virus during the first (P1) or third (P3) round of infection in porcine PCLS. The ciliary activity was determined at different times p.i. by microscopic inspection of the slices. As the P2 virus did not differ from the other viruses, only the first and the last virus passage is shown. (B) Growth curve of the P2 and P3 H9N2 influenza viruses on porcine PCLS. The titer of infectious virus released into the supernatant of infected PCLS after different times p.i. was determined by plaque titration. The P1 virus did not differ from P2 virus; therefore, only the P2 and P3 viruses are shown. Each experiment was repeated three times. Results are expressed as means ± SEM, and significance (*, P < 0.05; **, P < 0.01; ***, P < 0.001) was determined using one-way ANOVA and Tukey's multiple-comparison test.

The genomic RNAs of P0 to P3 viruses differ by four mutations.

The genomic RNAs of the parental (P0) and passaged viruses (P1, P2, and P3) were subjected to sequence analysis. Apart from a point mutation in PB2, G685R in P3 virus, and a point mutation in NS1, R200K in P1 virus, two amino acid exchanges were detected in the HA protein: mutation A190V occurred at passage P1 and T212I at passage P3. The HA mutations were designated according to the H3 HA numbering. The amino acids commonly found in H9N2 viruses at these positions are shown in Table 1. Each monomer of the HA molecule is composed of a globular domain and a stem domain. As illustrated in Fig. 2, the receptor-binding site (RBS) in the globular domain is composed of three major structural elements: a 220 loop (residues 221 to 228), a 130 loop (residues 134 to 138), and a 190 helix (residues 188 to 194). In addition, some highly conserved residues (Tyr98, Trp153, His183, and Tyr195) form the base of the pocket (35). As shown in Fig. 2, residue 190 is located within the receptor-binding site, whereas amino acid 212 has a more distant position within the globular domain of the hemagglutinin. As indicated by arrowheads, the E190V exchange appears to render the receptor-binding site of the mutants more accessible for ligands compared to the parental HA protein. As mutations within the HA protein are considered to be crucial for host tropism and interspecies transmission, the importance of those two HA mutations were analyzed further. Recombinant viruses were generated from another parental H9N2 virus (A/chicken/Emirates/R66/2002). To reveal the relationship of R66 and A/chicken/Saudi Arabia/CP7/98 (H9N2) to established H9N2 lineages (36), we performed a phylogenetic analysis for each viral gene. To this end, we included sequences of the two H9N2 prototype strains A/quail/Hong Kong/G1/1997 and A/Duck/Hong Kong/Y280/1997, along with sequences from representative strains belonging to one of the four monophyletic groups A, B, C, and D (36). Whereas all genes of A/chicken/Saudi Arabia/CP7/98 (H9N2) belong to cluster D, all R66 genes originate from cluster C (Fig. 3). The R66 virus was used because it was available for immediate use. Furthermore, we reasoned that phenotypic analysis of the above-mentioned mutations in the distantly related H9N2 strain R66 would show that the respective functional changes are not strictly dependent on the genetic backbone of the A/chicken/Saudi Arabia/CP7/98 virus. Recombinant viruses were generated to contain either one (HA190 or HA212) or both (HA190+HA212) HA mutations.

TABLE 1.

Percentage of H9N2 viruses with each amino acid at positions 190 and 212 in the HA protein

Amino acid	Result by host species of the different virus strainsa
	Avian		Human		Swine
	Countb	Relative frequency (%)	Countb	Relative frequency (%)	Countb	Relative frequency (%)
Position 190c
A	1,331	34.6	4	17.4	8	40.0
T	1,377	35.8	4	17.4	6	30.0
V	697	18.1	7	30.4	4	20.0
E	406	10.5	6	26.1	2	10.0
D	15	0.4	2	8.7
K	13	0.3
I	7	0.2
M	2	0.1
S	2	0.1
Total	3,850	100.0	23	100.0	20	100.0
Position 212c
T	2,369	61.6	16	69.6	20	100.0
I	1,183	30.8	6	26.1
A	9	0.2	1	4.3
V	282	7.3
N	2	0.1
S	1	0.0
Total	3,846	100.0	23	100.0	20	100.0

^aThe H9 protein sequences were downloaded from the Influenza Research Database (34) and GISAID. Duplicate, gapped, or truncated sequences were excluded prior to analysis. Two amino acid exchanges occurred during adaptation to PCLS: A190V at passage P1 and T212I at passage P3.

^bNumber of virus strains with the respective amino acid at position 190 or 212.

^cH3 numbering.

FIG 2.

Three-dimensional structure of the HA protein of parental and mutant viruses. Structural elements of the receptor-binding site are illustrated: the 190 helix, 130 loop, and 220 loop are shown in orange. (A) HA surface representation of parental virus. (B) HA surface representation of mutant virus HA190. (C) HA surface representation of mutant virus HA212. (D) HA surface representation of mutant virus HA190+HA212. The crystal structure of swine H9 hemagglutinin (PDB entry 1JSD) was selected as the template (78). All structural figures were generated with PyMOL (79).

FIG 3.

Phylogenetic analysis of PB2, PB1, PA, HA, NP, NA, M, and NS genes. Neighbor-joining consensus trees of nucleotide sequences from R66 (highlighted in yellow), A/chicken/Saudi Arabia/CP7/98 (H9N2) (highlighted in yellow), A/quail/Hong Kong/G1/1997, A/Duck/Hong Kong/Y280/1997, and sequences representing one of the four monophyletic groups A, B, C, or D (36), supported by 100 bootstrap samples. From each group, one strain was marked by a prefix indicating the cluster (CA, CB, CC, and CD) and highlighted in another color. For each sequence downloaded, the GISAID accession number is provided after the strain name. Numbers at nodes are bootstrap values.

Glycan array analysis reveals an extended sialic acid binding specificity of the HA mutants.

The sialic acid binding activity of parental and mutant R66 viruses was characterized by performing a glycan array analysis. A collection of 88 glycan structures was used comprising oligosaccharides with the following terminal disaccharides: NeuNAc-α2,3-Gal, NeuNAc-α2,6-Gal/GalNAc/GlcNAc, and NeuNAc-α2,8/9-NeuNAc (Fig. 4E). As shown in Fig. 4A, among 42 glycans containing α2,3-sialic acids, the majority were recognized by the parental H9N2 R66 virus. We detected no virus binding to five biantennary structures with a sialic acid residue on only one of the two branches. Only some of the glycans containing α2,6- or α2,8/9-linked sialic acids were recognized by parental R66 virus; as indicated by the fluorescent intensity (y axis), this binding was generally weak compared to the interaction with oligosaccharide structures containing α2,3-linked sialic acids. The 190V mutation did not have a major impact on the recognition of α2,3-linked sialic acid. However, binding to glycans containing α2,6- or α2,8/9-linked sialic acids was greatly enhanced (Fig. 4B, orange bars). All of the respective glycans were recognized, although with different affinities. Those glycans with α2,6- or α2,8/9-linked sialic acids that interacted with parental virus were recognized by the HA190 mutant with higher affinity. The T212I mutation had a much less dramatic effect. As shown in Fig. 4C, compared to the parental virus, the HA212 mutant bound to a few more glycans (shown in green bars), and those recognized by the parental virus were bound with higher affinity. The most efficient binding was observed with the double mutant HA190+HA212 (Fig. 4D); like HA190, it recognized all glycans with α2,6- or α2,8/9-linked sialic acids, but binding was characterized by higher affinity (shown in blue bars). Our results indicate that the HA190 mutation has a major effect on the binding activity by extending the spectrum of sialic acids recognized to the α2,6/8/9-linkage types.

FIG 4.

Glycan array analysis of parental and mutant viruses. (A to D) Parental and mutant viruses were subjected to glycan array analysis. (E) Glycans used for the array. a.u., arbitrary units.

Mutant viruses differ in plaque size.

To compare the replication of parental and mutant viruses, we applied different culture systems, MDCK II cells, newborn pig trachea (NPTr) cells, and PCLS. When we analyzed the replication of the recombinant R66 viruses in MDCK II cells, no difference between parental virus and mutant viruses was detected in the growth curve (not shown). However, a significant difference was detected in the plaque morphology (Fig. 5A). The HA212 mutant showed a large plaque size similar to that of the parental R66 virus. In contrast, the plaques induced by the HA190 mutant were much smaller. The small plaque size was maintained in the double mutant (HA190+HA212). The differences in plaque size were statistically significant (Fig. 5B). These results indicate that the point mutation at position 190 is responsible for the change in plaque morphology. There is some heterogeneity in the plaque size of the HA190 mutant. We think this is not due to a mixed virus population, because the mutant viruses were generated by reverse genetics and passaged only twice. In control samples that were sequenced prior to the mouse experiment described below, there was no indication of a mixed population. Rather, the heterogeneity may be due to the fact that cells were not synchronized at the time of infection.

FIG 5.

Plaque assay on MDCK II cells. (A) Plaque morphology. The parental virus H9N2-R66 and the recombinant viruses that contain the respective mutated proteins showed a different plaque morphology on MDCK II cells. (B) Plaque size of parental and mutant viruses on MDCK II cells. The plaque size of the mutant viruses is indicated as x-fold increase compared to the parental virus. Results are expressed as means ± SEM, and significance (***, P < 0.001) was determined using one-way ANOVA and Tukey's multiple-comparison test.

HA mutations 190V and 212I enhance viral replication in different culture systems.

In order to analyze the importance of the individual mutations, NPTr cells and PCLS were infected by parental R66 virus or recombinant viruses that contain the respective mutated proteins. The replication efficiency was determined by titration of the infectious virus in the supernatant at different time points after infection. As shown in Fig. 6A, at 24 hpi, the HA190 mutant had grown to a titer that was significantly higher than that of the other three viruses on NPTr cells. By 48 and 72 hpi, the titer of the double mutant, HA190+HA212, approached that of the HA190 mutant. While the HA212 mutant had a slight growth advantage at 24 hpi over the parental R66 virus, the titers of both viruses were similar at later time points and significantly lower than those of the other two viruses, HA190 and HA190+HA212. From these results, we conclude that the HA190 mutation had an enhancing effect on R66 virus replication in NPTr cells, and this effect was attenuated to some extent by the HA212 mutation.

FIG 6.

Growth characteristics of parental and mutant viruses on NPTr cells and PCLS. (A) Growth curve of parental and mutant viruses on NPTr cells. The titer of infectious virus released into the supernatant of infected NPTr cells at different times p.i. was determined as the TCID₅₀. Each experiment was repeated three times. (B) Percentage of fluorescent HA protein on NPTr cells. The area of the NPTr cell surface positive for red fluorescent HA protein was analyzed by applying analySIS 3.2 software (Soft Imaging System) to quantify HA attachment. (C) Growth curve of parental and mutant viruses on PCLS. The titer of infectious virus released into the supernatant of infected PCLS at different times p.i. was determined as the TCID₅₀. Each experiment was repeated three times. Results are expressed as means ± SEM, and significance (*, P < 0.05; **, P < 0.01; ***, P < 0.001) was determined using one-way ANOVA and Tukey's multiple-comparison test.

To find out whether the differences in growth characteristics of the four viruses are related to the interaction with the cell surface receptors, we performed a virus binding assay. For this purpose, NPTr cells were incubated with parental or mutant viruses at 4°C for 1 h. The HA protein of bound virions was detected by indirect immunofluorescence (not shown). Subsequently, the area of the NPTr cell surfaces positive for red fluorescent HA protein was analyzed by applying analySIS 3.2 software (Soft Imaging System) to quantify HA attachment. The percentage of fluorescent HA protein on NPTr cells indicated that virus binding was most efficient in the case of the HA190 mutant and the double mutant H190+HA212 (Fig. 6B).

In infection studies with PCLS, parental and mutant viruses did not differ in their effect on ciliary activity, indicating that neither of the mutants had an increased ciliostatic effect compared to the parental virus (not shown).

As far as the growth characteristics in PCLS are concerned (Fig. 6C), the HA212 mutant grew at both 24 and 48 hpi to titers that were significantly increased compared to those of the parental virus. The other two mutants, HA190 and HA190+HA212, showed somewhat increased titers at 24 hpi but not at later time points.

The HA double mutation increases pathogenicity in mice.

To investigate the effect of mutations in the HA protein in vivo, C57BL/6J mice were infected intranasally with 2 × 10⁵ focus-forming units (FFU) (Fig. 7A). Mice infected with the parental virus, or with viruses containing a single mutation (HA190 or HA212), showed moderate body weight losses. Remarkably, mice infected with the double mutant (HA190+HA212) showed significantly increased body weight losses compared to wild-type-infected animals from day 2 to 5 (Fig. 7A).

FIG 7.

Infection of mice by parental and mutant viruses. Female mice at the age of 8 to 12 weeks were infected with 2 × 10⁵ FFU of parental or mutant viruses, and body weight was monitored for 14 days p.i. (A) Body weight change in infected mice. Data represent mean values as a percentage of starting weight ± SEM. Stars indicate the level of significance of the double mutant (green) compared to R66-infected mice (black) from a post hoc pairwise t test (after having established significance between groups by ANOVA). (B) Antibody response to R66 infection. Eye blood was taken from infected mice on day 14 p.i., and sera were analyzed for influenza-specific IgG antibodies in an indirect ELISA. Noninfected C57BL/6J mice served as a negative control (neg co). Sera of mice showing body weight loss represent the positive control (pos co). Each data point represents the means from one serum tested twice. All mice exhibited an antibody response after infection. (C) Lung titers on day 3 p.i. in C57BL/6J mice. Lungs were homogenized and viral load was titrated at least twice for each sample. Each symbol represents one mouse. Mean values (gray) ± SEM are depicted. Statistical analysis was performed using nonparametric Mann-Whitney U test. The dotted line marks the detection limit. NS, nonsignificant; *, P < 0.05; **, P < 0.01; ***, P < 0.001.

It is important to note that mice that did not show a body weight loss were infected with the indicated virus, because all infected mice exhibited a virus-specific IgG antibody response to influenza A virus at day 14 postinfection (p.i.) (Fig. 7B).

To determine the viral load, on day 3 p.i. lung homogenates were prepared. Viral lung titers were highest in animals infected with H9N2-R66 (Fig. 7C). While infection with HA212 did not result in a statistically significant change in titer (P value of 0.0815), HA190 as well as the double mutant virus HA190+HA212 induced a significantly reduced viral load (P values of 0.0019 and 0.009, respectively) (Fig. 7C).

DISCUSSION

H9N2 influenza viruses circulate in poultry populations worldwide (37–40), and isolation from pigs has been reported (41, 42). Human infections by H9N2 strains have been observed occasionally (8, 43, 44). However, those cases might be considerably more frequent according to epidemiological studies (45, 46). To analyze the adaptation potential of H9N2 strains to growth in the target cells of a new host, we chose precision-cut lung slices (PCLS) as a culture system for differentiated airway epithelial cells. Replication of swine influenza viruses in PCLS was found to reflect the virulence properties of the respective virus strains (33). Following three passages in porcine PCLS, the P3 virus eventually differed from P0 to P2 viruses by a shorter growth cycle but not by enhanced replication efficiency or ciliostasis. Thus, adaptation manifested first in accelerated growth.

From the finding that the growth kinetics changes occurred at passage 3, one might conclude that the two mutations that were first detected in P3 virus are also responsible for the phenotypic change. However, the detailed characterization of the HA mutations indicated that the mutations are interrelated and affect each other in their phenotypes. The A190V mutation that was first detected in P1 virus had a major effect on the sialic acid binding activity of the HA protein. In contrast to the hemagglutinin of the parental virus that mainly recognized glycans containing α2,3-linked sialic acids, the HA190 mutant was able to interact with sialic acids present in α2,3, α2,6, and α2,8/9 linkages. This binding characteristic is consistent with lectin staining data obtained with PCLS that indicated an abundant presence of α2,6-linked sialic acids on both ciliated and mucus-producing cells, whereas α2,3-linked sialic acids were less prominent and restricted to ciliated cells (32). The A190V exchange at the edge of the receptor-binding site may result in a slight opening of the receptor-binding pocket that may allow the HA protein to recognize sialic acids in different linkage types. A wider receptor-binding site compared to that of avian HAs has also been reported for the HA protein of the strain A/swine/Hong Kong/9/98 (47), an H9N2 virus isolated from pigs. Interestingly, this virus also contained a valine residue at position 190, while avian H9N2 viruses most often contain an alanine or a threonine at this position (Table 1). In the A/swine/Hong Kong/9/98 virus, the widening of the receptor-binding site involved the Q226L mutation that enabled the respective HA protein to interact with both α2,3- and α2,6-linked sialic acids. Binding to sialic acids in each of these two linkage types is also characteristic for the double mutant analyzed in our study, although our mutants do not contain a leucine at position 226. They have retained the Gln-226 that is associated in avian hemagglutinins with a narrow receptor-binding site and a preferential recognition of α2,3-linked sialic acids. It appears that the T212I mutation detected in P3 and located a distance from the receptor-binding site resulted in a conformational change that had an effect on the sialic acid binding activity similar to that of a Q226L mutation. The enhanced binding activity of the HA190 and HA190+HA212 mutants was also evident in some of the biological assays. Viruses with the A190V mutation or the double mutation were superior to the parental virus and the HA212 mutant in both binding and replication efficiency in a porcine tracheal cell line (NPTr). Interestingly, recent avian H9N2 viruses isolated in live poultry markets in China were reported to preferentially recognize α2,6-linked sialic acids. In these viruses, the binding to human-like sialic acid receptors has been attributed to Q226L and I155T mutations and not to a 190V residue (48). However, avian H9N2 viruses with a valine at residue 190 showed enhanced binding affinity and replicated more efficiently in mice (49). In our study, the A190V+T212I double mutant also showed an increased virulence in mice. This finding underscores the importance of changes in the receptor-binding site of avian viruses for successful transmission to mammalian species.

A more differentiated picture arises when precision-cut lung slices are applied. Compared to the parental virus, the HA190 and HA190+HA212 mutant viruses were more efficient in the amount of infectious virus released but only at an earlier time point in the replication (24 hpi). Surprisingly, the highest virus titers were determined at all time points for the HA212 mutant. This result indicates that enhanced binding activity is not necessarily paralleled by an increased replication efficiency. A somewhat improved binding activity, as is the case for the T212I mutant virus, may be more favorable for infection of differentiated respiratory epithelial cells. Efficient infection by influenza viruses requires well-balanced sialic acid binding and neuraminidase activity (50). For H1 and H7 hemagglutinin, it has been shown that mutations enhancing the sialic acid binding activity may be detrimental for the fusion activity (51). Depending on the sialic acids on the cell surface, the optimal balance between sialic acid binding and neuraminidase activity may vary between different cell types. This may explain why the double mutation in the HA increased the pathogenicity in mice but not the virulence in PCLS.

Mice infected with the HA190+HA212 mutant exhibited increased body weight loss compared to mice infected with the parental R66 virus. However, this phenotype was not associated with a higher viral load; in fact, the HA190 mutant and the double mutant showed the lowest virus titers in mouse lungs. This effect may be explained by an enhanced inflammatory response that impairs virus replication and, in the case of the HA190+HA212 mutant, results in increased pathogenicity (52–54). The murine airway epithelium is known to contain predominantly α2,3-linked sialic acids (55). Therefore, it may be surprising that the two mutants that have extended their sialic acid binding activity to the α2,6 and α2,8/9 linkage types differ from the parental virus in their effect on mice. However, it should be noted that the mutants not only have a broader sialic acid binding activity but also show an increased affinity for the α2,3 linkage type. This change in the binding activity is especially pronounced in the double mutant and may allow the virus to infect a subset of cells that are resistant to infection by the parental virus. In this way, the HA190+HA212 mutant may induce an inflammatory response that is responsible for the observed body weight loss.

Taken together, our results show that HA mutations may have quite diverse effects on the replication of influenza viruses depending on the culture system used. As the function of the HA protein depends not only on the presence of receptor determinants but also on the interplay with the neuraminidase and on the ability to induce membrane fusion, an increased binding activity may enhance or decrease virus entry and, thus, infection efficiency. However, it may also just impede virus release and thus result in a smaller plaque size. In complex culture systems such as PCLS or animal experiments, minor cell types may be involved that are not well characterized. Because of these considerations, the effect of the mutations is not predictable and has to be elucidated by appropriate experiments.

Apart from amino acid exchanges in the hemagglutinin, mutations in the PB2 protein are considered to be essential for adaptation of avian influenza viruses to mammalian hosts (24, 56). PB2 is a major polymerase determinant for the virulence and host range of influenza viruses (45). In particular, an E627K mutation has been reported to be sufficient for several avian viruses to replicate efficiently in mammalian hosts (57, 58). Our P1 to P3 viruses retained E627; however, the increased growth kinetics of P3 virus was paralleled by a G685R mutation in PB2 which may have contributed to the enhanced replication efficiency. Thus, the initial mutations that occurred during adaptation of avian influenza viruses to porcine airway epithelial cells did not include the E627K mutation. Furthermore, this mutation is also not essential for rendering avian virus pathogenic for mice, as has been demonstrated by our analysis of the HA double mutation. The importance of the two mutations in the PB2 and NS proteins has to be analyzed in future studies.

The four mutations that occurred during the initial three passages in porcine PCLS resulted in an increased growth kinetics but not in an enhanced ciliostatic effect. It will be interesting in the future to adapt the avian H9N2 virus further by additional passages in porcine PCLS. From our results it may be that more passages resulting in additional mutations are required for adaptation of avian viruses to acquire properties of virulent swine strains. The actual adaptation to growth in pigs is likely to take even longer, as the whole organism has more defense mechanisms than the PCLS. Therefore, it is not very likely that avian viruses succeed in adaptation to a new host and in founding a new virus lineage. Even though pigs are discussed as mixing vessels for transmission of avian influenza viruses to humans, so far avian influenza viruses have succeeded only once in establishing a new virus lineage in swine after the first isolation of swine influenza virus (59). The H1N1 lineage of viruses that appeared in Europe in 1979 was of avian origin (60, 61). However, even partly adapted viruses may be a health risk if they acquire some gene segments from other viruses by reassortment. This risk is documented by hundreds of patients infected by currently circulating H7N9 and H5N1 viruses (45) which contained some or all internal gene segments from an avian H9N2 virus. Therefore, adaptation of influenza viruses to new hosts remains a key issue.

MATERIALS AND METHODS

Cells.

Human embryonic kidney 293T (HEK 293T) cells, Madin-Darby canine kidney (strain MDCK II) (62) cells, and a cell line derived from newborn pig trachea (NPTr) were maintained in Eagle's minimal essential medium (EMEM) supplemented with 10% fetal calf serum (Biochrom AG, Berlin, Germany). The cells were incubated in a humidified atmosphere containing 5% CO₂ at 37°C.

Virus.

Avian influenza virus A/chicken/Saudi Arabia/CP7/98, a low-pathogenicity avian influenza virus (LPAI) of the H9N2 subtype, was provided by Hans-Christian Philipp (Lohmann Tierzucht, Cuxhaven, Germany). The virus was propagated as described previously (32), and virus stocks were stored at −80°C.

PCLS.

PCLS were prepared from lungs of 3-month-old healthy crossbred pigs as described previously (32, 33, 63, 64).

The viability of the cells of PCLS was analyzed by monitoring ciliary activity under a light microscope (Zeiss Axiovert 35) equipped with an ORCA C4742-80 digital camera (Hamamatsu) and SIMPLE-PCI analysis software (Compix Imaging Systems). Each bronchus in the microscopic field was virtually divided into 10 segments, each of which was monitored for the presence or absence of ciliary activity (32, 33). Slices which showed 100% ciliary activity at the beginning were selected for the experiments. From selected samples, the slices were analyzed for bronchoconstriction by addition of 10⁻⁴ M methacholine (acetyl-ß-methylcholine chloride; Sigma-Aldrich) as described previously (32, 64). The integrity of the cells was also determined by applying a Live/Dead viability/cytotoxicity assay kit (FPBE4710; Fluo Probes) (32, 33).

Passaging of avian H9N2 virus in porcine PCLS.

Avian influenza virus A/chicken/Saudi Arabia/CP7/98 (H9N2) was passaged three times in PLCS. At first, PCLS were infected by A/chicken/Saudi Arabia/CP7/98 (H9N2) at 1 × 10⁴ PFU/ml (500 μl/slice) with acetylated trypsin (1 μg/ml). Supernatants from virus-infected slices were collected at 72 hpi and used for the next passage. The P1, P2, and P3 viruses were characterized by analyzing (i) the amount of infectious virus released into the supernatant of infected cells, (ii) the growth curve of the viruses, and (iii) the ciliostatic effect.

Sequencing of viral RNA genomes.

Sample preparation and sequencing using a 454 Genome Sequencer FLX (Roche, Mannheim, Germany) were done as described previously (65). Briefly, PCR products generated from viral genomic RNA were pooled in equimolar amounts. Sequencing libraries were prepared according to the manufacturer's instructions using a 454 rapid library preparation kit (Roche) and barcoded rapid library adaptors (Roche). The final size-selected libraries were quantified with a KAPA library quantification for the Roche 454 titanium kit (KAPA Biosystems, Cape Town, South Africa). For emulsion PCR, 0.08 copies per bead were used as the input. Sequencing was performed using Titanium chemistry (Roche). The obtained raw reads were assembled using the 454 assembler software Newbler, v2.6 (Roche).

Since the aim of the sequencing was to detect fixed mutations at the consensus level, the sequencing, although carried out with a second-generation sequencing instrument, was not performed to a depth enabling population analyses. The achieved sequence depth would not fulfill the minimal requirements to detect low-frequency mutations according to standards set forth in, e.g., reference 66. However, sensitive and reliable detection would be necessary, since the selection of mutations occurs quickly, as shown, for instance, in reference 67. Therefore, the focus was put on the analysis of mutations under positive selection.

Recombinant viruses.

Eight plasmids (pHW2000-R66-PB2, pHW2000-R66-PB1, pHW2000-R66-PA, pHW2000-R66-NP, pHW2000-R66-NA, pHW2000-R66-M, pHW2000-R66-NS, and pHW2000-R66-HA), encoding the individual viral RNA segments for influenza A virus A/chicken/Emirates/R66/2002 (H9N2) (R66) (GenBank accession numbers CY076720 to CY076727), were applied to generate recombinant viruses (68, 69). To generate a mutant virus, site-directed mutagenesis was performed on the pHW2000-R66-HA plasmid using the QuikChange lightning site-directed mutagenesis kit (Agilent Technologies). The mutated plasmids were completely sequenced to rule out unwanted mutations. Recombinant viruses containing the respective mutated proteins were rescued using the eight-plasmid DNA transfection system (70). A/chicken/Emirates/R66/2002 (H9N2) and mutants derived from it were propagated in MDCK II cells overlaid with serum-free EMEM containing 1 μg/ml acetylated trypsin (Sigma-Aldrich, Munich, Germany). For experiments with mice, viruses were grown on 10-day-old SPF (specific-pathogen-free) embryonated chicken eggs and sequenced by next-generation sequencing (Illumina) to ensure correct insertion of mutations. These experiments were performed under biosafety level 2 conditions.

Phylogenetic analysis.

All eight genes of the strains R66 and A/chicken/Saudi Arabia/CP7/98 (H9N2) were subjected to phylogenetic analysis by using Geneious 9.0.5 software for alignment (global alignment by Muscle alignment) and subsequent tree calculation (neighbor joining, Tamura Nei distance, 100 bootstraps) (71 and http://www.geneious.com). For tree calculations, all nucleotide sequences for all H9N2 strains quoted in Fusaro et al. were downloaded from GISAID (Global Initiative on Sharing All Influenza Data; EpiFlu database; www.gisaid.org; accessed 1 December 2016). Gapped or truncated sequences were removed, resulting in 33 PB2, 33 PB1, 35 PA, 38 HA, 36 NP, 36 NA, 37 M, and 34 NS sequences.

Plaque assay.

MDCK II cells were grown in EMEM containing 5% fetal calf serum on a 6-well plate for 1 day. The cells were washed twice with phosphate-buffered saline (PBS) and infected with serially diluted viral suspensions in EMEM containing acetylated trypsin (1 μg/ml). For virus adsorption, cells were kept on a shaker in 5% CO₂ at 37°C. After 1 h, the overlay medium containing Avicel microcrystalline cellulose RC 591 (2.5%; FMC Biopolymer, Brussels, Belgium), EMEM with glutamine (GIBCO BRL Life Technologies), and bovine serum albumin fraction V (0.2%; AppliChem) was added in a volume of 3 ml. After incubation for 2 to 3 days in 5% CO₂ at 37°C, the cells were fixed and stained with a formaldehyde solution containing 1% crystal violet (32). The plaque size was calculated using the ImageJ/Fiji software.

Virus infection and titration.

PCLS were infected by different viruses at 1 × 10⁵ 50% tissue culture infective dose (TCID₅₀)/ml (500 μl/slice) with acetylated trypsin (1 μg/ml). NPTr cells were grown in EMEM containing 5% fetal calf serum on a 6-well plate for 1 day. The NPTr cells were infected by different viruses at 1 × 10⁵ TCID₅₀/ml (500 μl/well) with acetylated trypsin (1 μg/ml). Supernatants from virus-infected and uninfected control slices or cells were collected at different time points (0, 8, 24, 48, and 72 hpi), and the samples were stored at −80°C. For virus titration or infectivity analysis, virus titers were determined by endpoint titration on MDCK II cells in 96-well plates as described previously (72). Briefly, for each sample, 10-fold serial dilution steps were performed. From each dilution, 100 μl/well was added onto confluent MDCK II cells in a 96-well plate, and every sample had 6 replicates. Plates were incubated for 72 h, and the wells were visually analyzed for virus-induced cytopathogenic effects (33).

HA binding activity.

NPTr cells were incubated with parental or mutant viruses at 1 × 10⁶ TCID₅₀/ml (500 μl) at 4°C for 1 h. The cells then were washed three times with cold PBS. HA protein was detected by indirect immunofluorescence analysis. For detection of HA protein, an anti-H9N2 rabbit polyclonal antibody at a 1:1,000 dilution was used, followed by a red fluorescent anti-rabbit IgG secondary antibody [Alexa Fluor 568 anti-rabbit IgG(H+L) antibody (Life Technologies)]. Anti-H9N2 rabbit polyclonal antibody was provided by Wolfgang Garten (Philipps-Universität Marburg, Germany). All antibodies were diluted in 1% bovine serum albumin and incubated with the cells for 1 h at room temperature in a humid incubation chamber. Finally, all of the samples were stored at 4°C until examination under a Nikon Eclipse Ti-S inverse immunofluorescence microscope equipped with a 10×/0.30 Plan Fluor objective (Nikon). Subsequently, the area of the NPTr cells surface positive for red fluorescent HA protein was analyzed by applying analySIS 3.2 software (Soft Imaging System) to quantify HA attachment (73).

Glycan array analysis. (i) Preparation of viruses.

Viruses were inactivated as described previously (32).

(ii) Glycan microarray fabrication.

In total, 88 α2,3/α2,6/α2,8/α2,9-sialic acids with linker modifications were printed on to N-hydroxysuccinimide-coated glass slides (Schott, Germany) by a robotic printing arrayer (Biodot, USA) (74, 75). Each carbohydrate was printed at 100 μM with two replicas per well, 16 wells per slide. The slides were stored in a dry box before being used.

(iii) Virus binding assay.

Prior to virus binding onto the glycan array, hemagglutination assays with 0.5% turkey red blood cells were performed to ensure the binding affinity of the different viruses. The glycan array was blocked with blocking buffer (Superblock; Thermo Fisher Scientific, USA) for 1 h and washed two times. Oseltamivir carboxylate-treated viruses were incubated with the glycan array for 1 h (76). Unbound viruses were washed away with PBST (0.05% Tween 20 in PBS), and the remaining viruses bound on arrays were incubated with polyclonal rabbit anti-H9 antibody, followed by a goat anti-rabbit IgG secondary antibody (Alexa Fluor 647; Jackson ImmunoResearch, USA). The arrays were sequentially washed with PBST and deionized water. Slides were spin dried for scanning with a GenePix 4300 (Molecular Devices, USA). Fluorescent intensity was analyzed with a GenePix Pro 7.0 (Molecular Devices, USA) and illustrated with Prism 6.0 (GraphPad, USA). The fluorescent intensity is a measure of the relative binding affinity of the respective viruses.

(iv) Structure determination.

Three-dimensional models of the HA protein of parental and mutant viruses were generated by the web-based homology modeling server SWISS-MODEL (77). The crystal structure of swine H9 hemagglutinin (PDB entry 1JSD) was selected as the template (78). All structural figures were generated with PyMOL (79).

Animal experiments. (i) Ethics statement.

All experiments in mice were approved by an external committee according to the national guidelines of the animal welfare law in Germany (BGBl. IS. 1206, 1313, and 1934). The protocol used in these experiments has been reviewed by an ethics committee and approved by the relevant authority, the Niedersächsisches Landesamt für Verbraucherschutz und Lebensmittelsicherheit, Oldenburg, Germany (permit number 3392 42502-04-13/1234). Female C57BL/6J mice (8 to 12 weeks old) were purchased from Janvier (France) and maintained under SPF conditions at the Central Animal Facilities at the HZI, Braunschweig, Germany.

(ii) Titration of infectious virus.

The infectivity of virus used for infection of mice was determined as described previously (80, 81). Virus titers were determined using a focus-forming unit (FFU) assay. The different unit size is due to the fact that the in vivo studies were performed in a different laboratory.

(iii) Infection of mice.

C57BL/6J mice (female, 8 to 12 weeks of age) were anesthetized by intraperitoneal injection of ketamine-xylazine solution (5 mg/ml ketamine [WDT, Garbsen, Germany] and 1 mg/ml xylazine [CP-Pharma, Burgdorf, Germany] in sterile 0.9% NaCl [WDT, Garbsen, Germany]) with a dose adjusted to the individual's body weight. Virus was diluted in sterile PBS to a dose of 2 × 10⁵ FFU, and mice were infected by intranasal application. Subsequently, body weight and survival were monitored for 14 days. From all animals, eye blood was taken on day 14 postinfection and serum was prepared.

(iv) Enzyme-linked immunosorbent assay (ELISA).

A 96-well plate was coated with 50 μl/well of A/chicken/Emirates/R66/2002 (H9N2) (5 × 10⁴ FFU in PBS) and incubated overnight at room temperature. After three wash steps with 100 μl PBST, 80 μl blocking buffer (BB) (0.05% Tween 20 and 5% FCS in PBS) was added for 1 h at 37°C and again washed thrice with 100 μl PBST. Mice sera were diluted 1:150 in BB, and 50 μl/well was added in duplicates. After 2 h of incubation at 37°C, wells were washed thrice with 100 μl PBST, and a goat-anti-mouse IgG (diluted 1:1,000 in BB; KPL) was added (50 μl/well) for 2 h at 37°C. Wells were again washed thrice with 100 μl PBST, and 50 μl/well of SureBlue TMB peroxidase substrate (KPL) was added for ∼3 min until a color change occurred. The TMB stop solution (KPL) then was added and the optical density was measured at 450 nm.

(v) Preparation of lung homogenates.

On day 3 p.i., lung homogenates were prepared using a FastPrep-24 instrument (MP Biomedicals). For this, lungs were homogenized in 1 ml PBS plus 0.1% bovine serum albumin in a lysing matrix D tube (MP Biomedicals) for 30 s at 6.5 m/s and then spun down for 10 min at 200 × g at 4°C to obtain the supernatant.

Statistical analysis.

If not stated otherwise, experiments were performed at least three times and results were expressed as means with standard errors of the means (SEM). Data were analyzed by one-way analysis of variance (ANOVA) and Tukey's comparison test using GraphPad Prism 5 software. A P value of <0.05 was considered significant.

For comparison of body weight loss in infected mice, ANOVA, to test for significant differences between groups, and a post hoc pairwise t test (Benjamini-Hochberg procedure for multiple test corrections) were performed using the R statistical software package (82).