Mutations in PB1, NP, HA, and NA Contribute to Increased Virus Fitness of H5N2 Highly Pathogenic Avian Influenza Virus Clade 2.3.4.4 in Chickens

H5Nx highly pathogenic avian influenza (HPAI) viruses of the A/goose/Guangdong/1/96 lineage continue to circulate widely, affecting both poultry and wild birds. These viruses continue to change and reassort, which affects their fitness for different avian hosts.

ABSTRACT

The H5N8 highly pathogenic avian influenza (HPAI) clade 2.3.4.4 virus spread to North America by wild birds and reassorted to generate the H5N2 HPAI virus that caused the poultry outbreak in the United States in 2015. In previous studies, we showed that H5N2 viruses isolated from poultry in the later stages of the outbreak had higher infectivity and transmissibility in chickens than the wild bird index H5N2 virus. Here, we determined the genetic changes that contributed to the difference in host virus fitness by analyzing sequence data from all of the viruses detected during the H5N2 outbreak and studying the pathogenicity of reassortant viruses generated with the index wild bird virus and a chicken virus from later in the outbreak. Viruses with the wild bird virus backbone and either PB1, NP, or the entire polymerase complex of the chicken isolate caused higher and earlier mortality in chickens, with three mutations (PB1 E180D, M317V, and NP I109T) identified to increase polymerase activity in chicken cells. The reassortant virus with the HA and NA from the chicken virus, where mutations in functionally known gene regions were acquired as the virus circulated in turkeys (HA S141P and NA S416G) and later in chickens (HA M66I, L322Q), showed faster virus growth, bigger plaque size, and enhanced heat persistence in vitro and increased pathogenicity and transmissibility in chickens. Collectively, these findings demonstrate an evolutionary pathway in which an HPAI virus from wild birds can accumulate genetic changes to increase fitness in poultry.

IMPORTANCE H5Nx highly pathogenic avian influenza (HPAI) viruses of the A/goose/Guangdong/1/96 lineage continue to circulate widely, affecting both poultry and wild birds. These viruses continue to change and reassort, which affects their fitness for different avian hosts. In this study, we defined mutations associated with increased virus fitness in chickens as the clade 2.3.4.4. H5N2 HPAI virus circulated in different avian species. We identified mutations in the PB1, NP, HA, and NA virus proteins that were highly conserved in the poultry isolates and contributed to the adaptation of this virus in chickens. This knowledge is important for understanding the epidemiology of H5Nx HPAI viruses and specifically the changes related to adaptation of these viruses in poultry.

INTRODUCTION

The natural reservoir of avian influenza (AI) viruses is wild aquatic birds, especially from the orders Anseriformes and Charadriiformes (1). AI viruses isolated from wild birds cause little or no disease in the natural host and generally cause no or mild disease when experimentally inoculated into chickens, with these viruses being classified as low pathogenicity (2). However, some low-pathogenicity H5 and H7 AI viruses can evolve in gallinaceous poultry and cause severe clinical signs and mortality, producing what is known as highly pathogenic avian influenza (HPAI) (3). Unlike other HPAI viruses, which are rarely found in wild birds, select subclades of the HPAI H5Nx A/goose/Guangdong/1/96 (Gs/GD) lineage have repeatedly spilled over from poultry to wild birds and have been maintained and spread through the migration of the infected wild birds (4).

The Gs/GD lineage H5N8 HPAI clade 2.3.4.4 viruses spread from Asia to North America in late 2014, likely through migratory aquatic birds because of overlapping breeding grounds of Asian and American birds. In North America, the H5N8 HPAI virus reassorted with North American wild bird AI viruses, which led to the emergence of H5N2 and H5N1 HPAI viruses (5). The H5N2 HPAI virus circulated in wild birds and backyard poultry along the Pacific flyway in the winter of 2014 to 2015, but in late winter and in the spring of 2015, the virus started transmitting to poultry in the Midwest and rapidly spread to a large number of turkey and chicken farms (6). The early wild bird index H5N2 virus was poorly adapted to chickens, showing low infectivity and transmissibility (7). Conversely, the later poultry H5N2 isolates showed increased infectivity and transmissibility in chickens and turkeys while still maintaining infectivity in mallards but with reduced virus shedding (8, 9).

The aim of this study was to understand the underlying genetic changes that contributed to the increased virus fitness observed with the H5N2 HPAI viruses in chickens compared to the early wild bird index H5N2 virus. We examined amino acid changes accumulated between the H5N2 index virus and the later poultry viruses by a time-scaled phylogenetic analysis including all the H5N2 virus isolates from the outbreak. Subsequently, in vitro and pathogenicity studies in chickens and mallards were conducted with reassortant viruses generated with the U.S. index H5N2 wild bird virus and an H5N2 virus isolated from a commercial chicken flock in the later stages of the poultry outbreak to determine the mutations causing increased adaptation to poultry.

RESULTS

Identification of potential amino acid changes associated with H5N2 HPAI virus adaptation in chickens as inferred by genetic analyses.

To identify potential molecular markers of adaptation of the H5N2 HPAI virus in chickens, a time-scaled Bayesian phylogenetic tree was generated by concatenating the eight genes of all the available H5N2 HPAI viruses from the North American outbreak (n = 269) (Fig. 1). The viral hosts (wild birds, turkeys, and chickens) were reconstructed to internal branches on the phylogenetic tree. Genetic groups were designated according to a previous study (6). Group 1 consisted of the early viruses mostly from wild birds in the Pacific flyway, and group 2 (2a, 2b, 2c, 2d, and 2e) consisted of the Midwest isolates. Most viruses belonging to group 2e were isolated in the later stage of the outbreak from chickens in Iowa.

FIG 1.

Bayesian phylogenetic tree of concatenated genes of the North American H5N2 HPAI viruses showing the amino acid changes found between A/Northern Pintail/Washington/40964/2014 (Dk) and A/Chicken/Iowa/13388/2015 (Ck). The inferred hosts with posterior probability of >0.5 are indicated with branches in blue (wild birds), red (chickens), and green (turkeys). Branch with mixed color had a posterior probability under 0.5. The index H5N2 virus (Dk) and later chicken virus (Ck) are shaped in dark blue and red circles, respectively. Highlighted thick branches represent where amino acid changes occurred between Dk and Ck and are listed according to branch numbers in the top left table. Twenty amino acid differences between Dk and Ck viruses examined in other H5N2 viruses during the outbreak are color coded on the right following amino acids of either Dk (blue), Ck (red), or other (yellow). The inset shows the percentages of amino acid changes between Dk and Ck found in H5N2 viruses isolated from wild birds, turkeys, or chickens.

Two H5N2 HPAI viruses, A/Northern Pintail/Washington/40964/2014 (Dk) (collected on 7 December 2014) and A/Chicken/Iowa/13388/2015 H5N2 (Ck) (collected on 26 April 2015), representing the U.S. wild bird index virus (group 1) and a later commercial chicken virus (group 2e), respectively, were chosen for amino acid analyses, since previous studies showed that the Dk virus was better adapted to mallards, whereas the Ck virus was better adapted to chickens (7, 8). The 20 amino acid differences between these two viruses are indicated on corresponding internal branches of the concatenated phylogenetic tree (Fig. 1). Three mutations (PB2 L386V, HA L8F, and NA E368K) were found in early H5N2 viruses from wild birds in the Pacific flyway. The PB2 V649I, M2 Q78R, and NS1 I176T changes were estimated to arise during the Pacific flyway wild bird/backyard poultry circulation. Host inference of a branch linking the Pacific flyway and Midwestern poultry outbreaks in which the NA R253K change occurred was not supported by posterior probability. Five amino acid changes (PB1 R215K, PA A337V, HA S141P, NS1 K217T, and NEP N60H) were estimated to occur during dissemination of the virus in turkey farms. Eight amino acid changes (PB1 E180D and M317V; PA A475T; HA L7P, M66I, and L322Q; NP I109T; and NA S416G) were conserved in viruses detected in the state of Iowa, with mixed host prediction in either chickens or turkeys.

The presence of the 20 amino acid differences between the Dk and Ck viruses was examined in all H5N2 viruses isolated in the United States, and the proportion of amino acids found in each position, was calculated for isolates from ducks, turkeys, or chickens (Fig. 1, inset). Three mutations (PB2 L386V, HA L8F, and NA E368K) were found in all H5N2 viruses regardless of the type of host (Fig. 1, branch 1), except in the reference Dk virus and two close relative viruses from the same species. In general, the ratio of amino acid distribution in each host type correlated with the order of occurrence of the other amino acid changes, as seen in the phylogenetic tree. First, the PB2 V649I (branch 2), M2 Q78R (branch 3), and NS1 I176T (branch 3) mutations, which arose during circulation in the Pacific flyway, were found in a high proportion in both turkey (99.4 to 100%) and chicken (96.9 to 98.5%) virus isolates. Second, the NA R253K (branch 4) mutation was found less commonly in wild bird viruses (25%), whereas it stayed at a high proportion in turkey viruses (99.4%) and chicken viruses (92.3%). Third, the five mutations that were inferred to arise in turkeys (branch 5, PA A337V; branch 6, PB1 R215K, HA S141P, NS1 K217T, and NEP N60H) were found in low proportions (15.0 to 22.5%) in wild bird viruses and high proportions in both turkey (89.3 to 89.9%) and chicken (86.2 to 89.2%) viruses. Fourth, the eight mutations that arose before virus introduction to farms in Iowa (branches 7 and 8, PB1 E180D and M317V; PA A475T; HA L7P, M66I, and L322Q; NP I109T; and NA S416G) had low proportions in turkey viruses (18.9 to 26.4%) and higher proportions in chicken viruses (53.8 to 73.8%). Seven of these changes (PB1 E180D and M317V; PA A475T, HA L7P, and L322Q; NP I109T; and NA S416G) were highly consistent upon divergence to group 2e, except HA M66I, where it remained unchanged in 17 out of 84 viruses (Fig. 1).

The PB1 and NP gene segments from the chicken H5N2 virus influence mortality and mean death times, whereas the HA and NA gene segments favor contact transmission among chickens.

To determine the potential role of gene segments in contributing to virus adaptation to chickens, we generated 10 reverse genetics (rg) viruses, including the recombinant rgDk and rgCk H5N2 HPAI viruses (Fig. 2). Experiments were set for chickens and mallards to compare pathogenesis between these viruses. The pathobiology of the H5N2 HPAI clade 2.3.4.4 viruses is not the same in chickens and mallards (7, 8), so the experiments were designed to highlight the differences expected in clinical presentation between the reassortant viruses in each species. Differences in mortality, virus shedding, and transmission were observed in the chickens and are summarized in Table 1. Survival curves are presented in Fig. 3. Detailed virus shedding data are shown as individual scatterplots in Fig. 4. Although most chickens showed little clinical disease before death (peracute disease), some presented typical clinical signs of HPAI, such as lethargy, ruffled feathers, and cyanotic combs. Consistent with our previous studies using the Dk and Ck viruses (7, 8), the rgCk-inoculated group (5/5 birds) showed higher mortality than the rgDk group (3/5 birds), with mean death times (MDTs) of 2.6 days and 4.3 days, respectively. Single-gene reassortants in the rgDk backbone carrying either the PB2 (rgDk/Ck-PB2) or PB1 (rgDk/Ck-PB1) gene segment from the rgCk virus also induced 100% mortality, with MDTs of 3.8 and 3.0 days, respectively; however, among the two groups, only the rgDk/Ck-PB1 group was significantly supported by the statistical comparison of survival curves. The same was observed with the reassortants carrying all three polymerase genes from rgCk (rgDk/Ck-POL) or the NP (rgDk/Ck-NP), with statistical support from the survival curve, presenting short MDTs of 2.2 and 2.0 days, respectively. Likewise, 100% mortality was observed in chickens inoculated with rgDk/Ck-HA.NA; however, the MDT in this group was longer than that in any other group (4.8 days). The remaining single-gene reassortants (rgDk/Ck-PA, rgDk/Ck-M, and rgDk/Ck-NS) failed to infect and cause death in all chickens in each group, with MDTs of 3.6, 4.0, and 4.1 days, respectively, for the birds that died.

FIG 2.

H5N2 viruses generated in this study. Virus genes are presented by squares in blue or red, corresponding to genes of the A/Northern Pintail/Washington/40964/2014 (Dk) or A/Chicken/Iowa/13388/2015 (Ck) viruses, respectively. Amino acid differences between the two viruses are color coded. H5 numbering was used for residues in HA, except for the two underlined residues being within the HA signal peptide.

TABLE 1.

Mortality, virus shedding, serology, and transmission to direct contacts of the H5N2 HPAI reassortant viruses in chickens

Virus	Mortalitya	Virus sheddingb (log virus copy no./ml)										Serologyc	No. of contact chickens infectedd
		1 dpi		2 dpi		3 dpi		4 dpi		5 dpi
		OP	CL	OP	CL	OP	CL	OP	CL	OP	CL
rgDk	3/5 (4.3)	2/5 (3.5)	0/5	2/5 (4.0)	2/5 (3.8)	3/5 (5.2)	3/5 (4.0)	3/3 (4.8)	1/3 (3.6)	2/2 (4.8)	1/2 (4.4)	0/2	0/3
rgDk/Ck-PB2	5/5 (3.8)	3/5 (4.6)	0/5	5/5 (5.4)	4/5 (4.6)	4/4 (6.8)	4/4 (5.4)	2/2 (8.1)	2/2 (6.6)				0/3
rgDk/Ck-PB1	5/5 (3.0)	4/5 (4.0)	0/5	5/5 (6.3)	5/5 (5.1)	2/2 (8.0)	2/2 (7.0)						0/3
rgDk/Ck-PA	4/5 (4.3)	3/5 (3.6)	1/5 (3.4)	4/5 (4.8)	5/5 (3.9)	5/5 (6.8)	4/5 (5.2)	1/2 (5.8)	1/2 (5.6)	0/1	1/1 (3.5)	0/1	0/3
rgDk/Ck-NP	5/5 (2.2)	5/5 (5.3)	0/5	5/5 (7.7)	5/5 (7.0)								0/3
rgDk/Ck-POL	5/5 (2.0)	5/5 (6.1)	5/5 (5.1)	5/5 (7.7)	5/5 (7.3)								0/3
rgDk/Ck-HA.NA	5/5 (4.8)	1/5 (3.9)	0/5	5/5 (5.4)	5/5 (4.0)	5/5 (5.7)	4/5 (4.2)	4/4 (6.3)	4/4 (4.6)	2/2 (7.6)	2/2 (5.3)		2/3e
rgDk/Ck-M	4/5 (4.0)	2/5 (3.8)	0/5	3/5 (4.4)	3/5 (4.0)	4/5 (5.0)	1/5 (3.5)	0/1	0/1	0/1	0/1	0/1	0/3
rgDk/Ck-NS	3/5 (3.3)	3/5 (4.1)	0/5	3/5 (4.8)	3/5 (4.3)	1/3 (4.7)	1/3 (4.3)	0/2	0/2	0/2	0/2	0/2	0/3
rgCk	5/5 (2.6)	5/5 (5.7)	2/5 (3.9)	5/5 (7.8)	5/5 (6.4)	3/3 (7.7)	3/3 (6.8)	-	-	-	-	-	0/3

^aNumber of dead birds/total number of birds inoculated (MDT of birds).

^bNumber of virus-positive birds/total number of birds sampled (log₁₀ virus copy number/ml).

^cNumber of HI-positive birds/total number of birds surviving (log₂ HI unit).

^dNumber of HI-positive or dead birds/total number of contact birds.

^eOne bird was found dead at 9 dpi with virus shedding (10^6.9/10^5.5 virus copy number in OP/CL). The other bird was HI positive (16 HI units).

FIG 3.

Survival curves for chickens infected with reassortant H5N2 HPAI viruses. Each group was compared to the rgDk group using the log-rank (Mantel-Cox) test with a Bonferroni correction for multiple comparisons. Shown are survival curves that differ significantly compared to rgDk (P ≤ 0.00279) on the left (A) and survival curves that had no significant difference compared to rgDK (P ≥ 0.04455) on the right (B). The rgTD16 group is shown in both curves for reference.

FIG 4.

Scatterplot of oropharyngeal (OP) and cloacal (CL) virus shedding detected by qRRT-PCR from chickens inoculated with reassortant H5N2 HPAI viruses in chickens (A) and mallards (B). Mean virus copy numbers are expressed with error bars. For statistical purposes, negative samples were given a value of limit of detection (10^2.8 virus copy number/ml). Different superscript lowercase letters denote significant difference for mean virus copy number per milliliter between groups (P < 0.05). Significant differences between groups were analyzed by the nonparametric Kruskal-Wallis test followed by Dunn’s post hoc test.

In general, higher virus shedding was observed with the reassortant viruses showing decreased MDT and high mortality compared to rgDk (Table 1 and Fig. 4). Chickens inoculated with rgCk, rgDk/Ck-NP, and rgDk/Ck-POL shed large amounts of virus by both the oropharyngeal (OP) and cloacal (CL) routes at 1 and 2 days postinoculation (dpi) compared to the other viruses. It is noted that chickens inoculated with rgDk/Ck-HA.NA showed slow onset of virus shedding, which helps explain the longer MDT observed in this group. Transmission to direct contact birds was only observed in this group (2 of 3 birds). Both contact birds shed virus via OP and CL routes, with one found dead at 9 dpi while the other was seropositive at the end of the experiment.

Less virus shedding in mallards infected with reassortants carrying gene segments from the chicken H5N2 virus.

To determine the pathobiology of the H5N2 HPAI viruses in waterfowl, mallards were inoculated with the same reassortant viruses used for the chicken experiment (Fig. 2) by following a different protocol based on previous studies (10). No clinical signs or mortality were observed in any of the mallards. All inoculated mallards became infected based on virus shedding and seroconversion (Table 2 and Fig. 4). Comparing the two recombinant viruses, mallards inoculated with rgCk showed significantly smaller amounts of OP virus shedding at 2 dpi than mallards infected with rgDk and CL virus shedding was detected in only one mallard inoculated with rgCK, whereas all mallards inoculated with rgDk shed virus by the CL route. No significant differences in virus shedding were observed between these two viruses at the remaining time points. As for the reassortant viruses, there were differences in the kinetics of virus shedding when replacing the PB2, PB1, PA, and M genes of rgDk with those of rgCk. Mallards inoculated with either rgDk/Ck-PB2, rgDk/Ck-PB1, rgDk/Ck-PA, or rgDk/Ck-M shed significantly smaller amounts of virus by the OP route at 2 dpi than mallards inoculated with rgDk. In addition, mallards inoculated with rgDk/Ck-PB2 or rgDk/Ck-M shed smaller amount of virus by the CL route at 2 dpi than mallards inoculated with the rgDk virus. No significant differences in virus shedding were observed between the viruses at 4, 7, and 10 dpi. Viral RNA was detected in all tissue samples from the mallards necropsied at 3 dpi, indicating similar systemic replication (Table 3). Due to the limited number of ducks tested (n = 2), virus titers could not be statistically compared.

TABLE 2.

Virus shedding and serology of the H5N2 HPAI reassortant viruses in mallards

Virus	Virus sheddinga (log virus copy no./ml)								Serologyb
	2 dpi		4 dpi		7 dpi		10 dpi
	OP	CL	OP	CL	OP	CL	OP	CL
rgDk	10/10 (6.7)	10/10 (4.5)	8/8 (5.3)	8/8 (5.0)	3/8 (3.1)	6/8 (3.5)	0/8	2/8 (3.0)	8/8 (3.8)
rgDk/Ck-PB2	10/10 (6.1)	10/10 (3.5)	8/8 (5.9)	8/8 (4.1)	8/8 (3.5)	4/8 (3.0)	2/8 (2.9)	3/8 (2.9)	8/8 (3.8)
rgDk/Ck-PB1	10/10 (5.8)	10/10 (4.6)	8/8 (6.3)	8/8 (5.0)	6/8 (3.5)	3/8 (3.2)	2/8 (3.0)	0/8	8/8 (3.8)
rgDk/Ck-PA	10/10 (6.1)	10/10 (4.4)	8/8 (5.5)	8/8 (4.9)	4/8 (3.2)	3/8 (3.1)	0/8	2/8 (3.0)	8/8 (4.3)
rgDk/Ck-NP	10/10 (6.4)	10/10 (4.4)	8/8 (5.9)	8/8 (5.4)	8/8 (3.3)	3/8 (3.1)	2/8 (2.9)	2/8 (3.0)	8/8 (3.8)
rgDk/Ck-POL	10/10 (6.5)	10/10 (4.1)	8/8 (6.0)	8/8 (4.2)	8/8 (3.5)	2/8 (3.0)	2/8 (2.9)	0/8	8/8 (3.8)
rgDk/Ck-HA.NA	10/10 (6.2)	10/10 (4.0)	8/8 (5.8)	8/8 (4.1)	8/8 (3.4)	3/8 (3.0)	0/8	0/8	8/8 (5.5)
rgDk/Ck-M	10/10 (6.1)	7/10 (3.5)	8/8 (5.3)	6/8 (3.7)	3/8 (3.0)	2/8 (3.0)	0/8	0/8	8/8 (4.0)
rgDk/Ck-NS	10/10 (6.7)	10/10 (4.3)	8/8 (6.1)	8/8 (5.0)	6/8 (3.2)	4/8 (3.4)	5/8 (3.1)	8/8 (3.5)	8/8 (4.3)
rgCk	10/10 (5.5)	1/10 (2.9)	8/8 (6.2)	7/8 (4.3)	5/8 (3.5)	4/8 (3.2)	2/8 (3.2)	0/8	8/8 (5.8)

^aNumber of virus-positive birds/total number of birds sampled (log₁₀ virus copy number/ml).

^bNumber of HI-positive birds/total number of birds tested (log₂ HI units).

TABLE 3.

Virus titers in brain, heart, spleen, lung, and muscle of mallards inoculated with H5N2 HPAI virus reassortants

Reassortant	Virus titer in tissuesa (log virus copy no./ml)
	Lung	Spleen	Brain	Heart	Muscle
rgDk	6.3/6.1	3.0/2.4	4.8/6.0	3.4/4.3	3.2/3.4
rgDk/Ck-PB2	4.4/5.4	3.5/3.5	3.3/5.0	3.7/3.7	4.6/2.9
rgDk/Ck-PB1	4.4/5.9	3.6/3.0	3.2/6.1	4.3/4.4	2.4/4.5
rgDk/Ck-PA	5.0/5.3	3.1/3.5	4.9/4.1	4.0/5.3	5.6/5.7
rgDk/Ck-NP	6.1/6.0	5.1/2.6	4.2/3.3	3.4/3.2	5.9/5.0
rgDk/Ck-POL	6.0/6.1	5.2/4.3	5.8/5.8	6.1/6.2	6.6/6.5
rgDk/Ck-HA.NA	2.7/4.2	2.4/4.5	6.9/5.2	3.8/4.0	3.4/3.3
rgDk/Ck-M	6.9/6.5	3.1/2.4	4.1/3.8	3.4/4.6	2.4/5.2
rgDk/Ck-NS	5.7/6.7	3.3/3.7	4.6/6.3	5.7/4.9	6.1/6.9
rgCk	3.8/4.8	4.3/5.4	3.4/4.1	2.6/5.8	2.4/2.9

^aTissues were taken from 2 mallards (mallard 1/mallard 2) at 3 dpi.

Virus growth kinetics and plaque-forming phenotype modulated by the surface gene segments.

To examine how the surface or polymerase proteins affect viral fitness in vitro, the virus growth kinetics of the rgCk and rgDk viruses were compared to each other and to the rgDk/Ck-HA.NA and rgDk/Ck-POL viruses in cells derived from chickens (DF-1) and ducks (CCL-141). Viral titers in cell culture supernatants were measured by plaque-forming assays on MDCK cells (Fig. 5). Irrespective of the cell line, the rgCk virus, followed by the rgDk/Ck-HA.NA virus, showed the fastest growth kinetics and reached higher peak virus titers than the two other viruses tested, rgDk and rgDk/Ck-POL (Fig. 4A and B). These observations matched the different abilities of these viruses to form plaques in MDCK cells, with the rgCk virus forming the largest plaques, followed by the rgDk/Ck-HA/NA virus. The polymerase gene segments in rgDk/Ck-POL had no effect on plaque size and, like the rgDk virus, formed pinpoint plaques.

FIG 5.

Virus growth kinetics and plaque sizes in chicken and duck cells. The rgDk (black), rgDk/Ck-HA.NA (pink), rgDk/Ck-POL (green), and rgCk (purple) viruses were inoculated to DF-1 (A) and CCL-141 (B) cells at a low dose (MOI of 0.00001). Supernatants were harvested at given time points and the number of PFU measured. Sizes of plaques were measured (C), and visible plaques are presented for each inoculum (D).

Increase in polymerase activity as mutations accumulated in the polymerase complex and in the NS genes of the H5N2 virus.

To better define the role of amino acid mutations in the polymerase complex of the Ck virus compared to the Dk virus, amino acid mutations were introduced in PB2, PB1, PA, and NP in the same order that was indicated by the phylogenetic tree. Polymerase assays were performed using an influenza reporter replicon encoding the Gaussian luciferase reporter flanked by the untranslated regions of the NS segment. Polymerase activity was assessed at two different temperatures, 33°C and 37°C (Fig. 6A). In the polymerase assay, we included the NS gene with the pertinent mutations, as it is known that mutations in NS1 and/or NEP can modulate selective translation of viral mRNA interacting with host RNA and proteins, directly or indirectly (11). Five sets of polymerase complex and NS genes were reconstructed following the mutations identified in the Bayesian phylogenetic tree between the Dk and Ck viruses. The L386V PB2 mutation by itself had marginal effects on the polymerase activity at either 33°C or 37°C. Adding the two mutations that possibly arose during transition of the virus in wild birds (V649I in PB2 and I176T in NS1) enhanced the polymerase activity about 4-fold compared to that of rgDk at 33°C but had no impact at 37°C. Adding one more mutation, A337V in PA, did not significantly increase polymerase activity at 37°C but increased the activity at 33°C compared to that of rgDk (Fig. 6A). Adding three mutations (R215K in PB1 and K217T and N60H in NS) improved the polymerase activity over the one from rgDk at 37°C but was not significantly different at 33°C. Subsequently, three mutations, E180D and M317V in PB1 and A475T in PA, kept the polymerase activity higher at 37°C and slightly higher at 33°C than rgDk. The I109T change in NP induced no significant increase in polymerase activity at 37°C compared to that of its preceding viral ribonucleoprotein complex (vRNP), but it decreased the activity at 33°C. Overall, the polymerase activity of the rgCk virus was higher than that from the rgDk virus at 37°C but was similar at 33°C.

FIG 6.

Polymerase activity of rgDk and rgCk and reconstructed polymerase mutants. DF-1 cells were transfected with plasmids expressing the vRNP and a plasmid for the synthesis of luciferase reporter RNA flanked with viral noncoding sequences. (A) Combinations of wild-type and reconstructed polymerase were tested with NS at 33°C or 37°C. (B) Each substitution in the polymerase complex was also tested separately, with no NS, at 37°C. Data shown are the means ± standard deviations from three independent experiments and normalized to the average for rgDk. The asterisk indicates a P value of <0.05 compared with the polymerase activity of the rgDk polymerase complex. **, P < 0.001; *, P < 0.05.

The effect of individual amino acid substitutions in the polymerase complex was also examined at 37°C (Fig. 6B). In testing the single-amino-acid substitutions in the polymerase complex, the NS gene was not included based on a previous study (12). Again, the polymerase activity of rgCk was higher than that for rgDk. As for single substitutions, three of them (PB1 E180D and M317V and NP I109T) upregulated the polymerase activity close to the level of rgCk. The slight increase observed with the PB1 R215K mutation was not significant. Polymerase activity was not affected with PB2 L386V, PB2 V649I, PA A337V, or PA A475T. Collectively, these observations suggest PB1 E180D and M317V and NP I109T contributed to increased polymerase activity at 37°C as the virus adapted to poultry.

Changes in the HA and NA affect temperature stability.

To determine if genetic changes between Dk and Ck contribute to temperature stability, the rgCk, rgDk, rgDk/Ck-HA.NA, and rgDk/Ck-POL viruses were incubated at different temperatures (20°C, 28°C, or 39°C), and the remaining viable viruses were measured at different times of incubation (0, 6, 24, 48, 72, and 120 h) (Fig. 7). The mean incubation time with a reduction of 99% of virus titer revealed that the rgCk and rgDk/Ck-HA.NA viruses were more stable at 39°C and 28°C than the rgDk and rgDk/Ck-POL viruses (linear regressions on virus titration plots are available upon request). The differences in the reduction time between the reassortant viruses indicate that mutations accumulated in HA and NA of the Ck virus conferred thermal stability at temperatures of ≥28°C, whereas mutations in the polymerase complex were not determinant for virus survival at the higher temperatures.

FIG 7.

Mean viral reduction time of recombinants at different temperatures. The rgDk (black), rgDk/Ck-HA.NA (pink), rgDk/Ck-POL (green), and rgCk (purple) viruses were diluted in phosphate-buffered saline and incubated at 20°C, 28°C, and 39°C. Reductions of viral titer (Rt) were measured in SPF eggs at the given time points using three different vials. Rt values correspond to the time required for a decrease in viral titer of 2 log₁₀ EID₅₀/ml. Significant differences in 99% reduction time for each virus are indicated (*, P < 0.05).

DISCUSSION

Similar to previous studies that compared the index H5N2 wild duck virus to later H5N2 viruses isolated from poultry (7, 8), the rgDk and rgCk viruses showed greater fitness in mallards and chickens, respectively. Mallards inoculated with the rgCk virus shed less virus than those inoculated with the rgDk virus. Although no single-gene segment substitution completely reduced early virus shedding to as low a level as rgCk, replacing either the PB2, PB1, PA, or M gene of the rgDk virus with those of the rgCK virus decreased virus shedding. In chickens, the rgCk virus caused higher and earlier mortality than the rgDk virus. This was also observed when the PB1, the NP, or all the polymerase complex segments of rgDk were replaced with those from the rgCk virus. Interestingly, replacing the HA and NA segments of rgDk with those of rgCk also caused 100% mortality, but this group had a longer MDT. These observations suggest that mutations in both the surface and internal gene segments are associated with increased virulence in chickens; however, in our controlled experimental setting, virus transmission to contacts was only observed with rgDk/Ck-HA.NA. Poor transmission in chickens in similar experiments was also observed with H5Nx Gs/GD lineage viruses (7, 8, 13) and other HPAI viruses (14, 15). However, under the same conditions, transmission was seen with low-pathogenicity avian influenza viruses that are well adapted to chickens, including H7N9, H9N2, and H5N2 viruses (16 and M. J. Pantin-Jackwood, unpublished results). One possible explanation for HPAI viruses not transmitting well under experimental conditions is the shorter time of exposure to HPAI virus due to the high early mortality. The longer survival of chickens infected with rgDk/Ck-HA.NA may have increased the chance of contact with uninfected chickens, consequently facilitating virus transmission. Another factor could be specific characteristics of the virus, including environmental stability, wherein transmission to contacts is enhanced with increased environmental hardiness of the virus particles. More studies are being conducted to better understand the transmission of AI viruses in chickens.

The HA and NA activity is important for viral entry and exit from cells as well as virus motility, affecting infectivity and virus transmission (17, 18). Bigger plaque size in MDCK cells and faster growth kinetics in avian cells were observed with rgCk and rgDk/Ck-HA.NA than the rgDk. These properties are consistent with the increased mortality observed in chickens inoculated with these viruses; however, these findings do not explain other observations, such as the longer MDT with rgDk/Ck-HA.NA and no transmission with rgCk. Further studies are also necessary to understand how virus replication in vitro correlates with pathogenicity and virus transmission in vivo.

Of the 20 amino acid differences found between rgDk and rgCk, 13 of them had an effect in vitro and/or in vivo in chickens and ducks, with many of these mutations previously identified in other studies as associated with virulence or found in motifs of known importance (Table 4). Of the four mutations in the HA acquired as the H5N2 virus circulated in turkeys and chickens, two have been previously associated with changes in virus phenotype (M66I and S141P). The S141P mutation in H5 HA was reported in both avian and human Eurasian H5Nx viruses (19, 20). The three-dimensional structure of a related clade 2.3.4.4. HA protein showed that residue 141 is located at the end of the 130-loop of the receptor binding site, and a proline at this position may confer receptor binding preference to turkeys (21). Another change in the HA gene is L322Q, which is in the second residue of the HA cleavage site. Viruses from this clade are the most widely spread group of H5Nx HPAI viruses in wild birds compared to other precedent clades that exclusively have glutamine at this position. The reversion to glutamine was acquired during the later H5N2 circulation in poultry. Further studies on HA cleavability in wild birds and poultry are needed.

TABLE 4.

Summary of amino acid changes in the H5N2 HPAI virus associated with adaptation in poultry

Segment	Amino acid substitutiona	In vivo findings	In vitro findings	Remark(s)	Reference(s)
PB2	L386V	Increased mortality in chickens and decreased virus shedding in mallards	n.a.c	Mutation located in the cap-binding site	50
	V649I		Increased polymerase activity at 33°C with NS1 I176T	Mutation located in the host-specific domain and is associated with increased virulence and transmissibility in mice and ferrets	51–55
PB1	E180D	Increased mortality in chickens and decreased virus shedding in mallards	Increased polymerase activity at 37°C with PB1 M317V	Mutations located in the nuclear localization motif	56
	R215K		n.a.		n.a.
	M317V		Increased polymerase activity at 37°C with PB1 E180D	Mutation located in the canonical RdRp domain; a similar hydrophobic change, M317I, was found in Mexican H5N2 and H7N3 HPAI viruses after circulation in poultry	15, 57, 58
NP	I109T	Increased mortality and delayed MDT in chickens	Increased polymerase activity at 37°C	Mutation located in the RNA binding subregion of the virus; reported to enhance virus gene transcription after the passage of a Gs/GD H5 HPAI virus in chicken brains	27, 59
HAb	M66I	Increased mortality but no effect on mean death time in chickens; increased virus transmission to contact chickens	Faster virus growth in avian cells and bigger plaque size in MDCK cells; increased thermal persistence at 28°C and 39°C	Mutation at vestigial esterase domain previously linked to acid stability and fusion pH	60, 61
	S141P			Mutation associated with antigenic drift and reduced host immune response in chickens	62
	L322Q			Leucine at this position is typical of clade 2.3.4 H5 HA genes	63
NA	E368K			Both mutations are located in proximity of the 2nd sialic acid binding site of NA, which is associated with neuraminidase activity and viral growth in chicken cells	64
	S416G
NS1	I176T	Increased duration of virus shedding in mallards	n.a.	n.a.	n.a.
	K217T		Increased polymerase activity at 33°C with PB2 V649I	K217T/N60H mutations in NS were experimentally selected as alpha-interferon sensitive residues through high-throughput random approaches	65

^aAmino acid changes between A/Northern Pintail/Washington/40964/2014 and A/Chicken/Iowa/13388/2015 H5N2 HPAI viruses.

^bH5 numbering was used by following the recommended numbering scheme (44).

^cn.a., not applicable.

The NA E368K and S416G mutations are located in the proximity of the hemadsorption site of this protein (22). Mutations of avian NA subtype N2 in this region are associated with HA and NA balance by regulating the second sialic acid binding site (23). Altogether, the changes in the HA and NA, particularly the five mutations located at functionally known regions (HA M66I, S141P, and L322Q and NA E368K and S416G), are likely associated with increased viral replication and pathogenicity in chickens and with increased growth in vitro.

The significance of the vRNP is well established in its role in mammalian host range restriction, primarily by determining the correlation between increased polymerase activity and virulence (24, 25). The contribution of the polymerase complex genes to AI virus virulence in avian species is well documented (26–33). The faster virus growth and higher virus titers observed with rgDk/Ck-POL than rgDk in vitro are consistent with the high virus replication, shorter MDT, and acute disease progress of the rgDk/Ck-POL virus in chickens. Because no difference in plaque size was observed with rgDk/Ck-M and rgDk/Ck-NS viruses compared to rgDk and rgDk/Ck-POL (data not shown), the bigger plaque sizes found with the rgCk virus than rgDk/Ck-HA.NA can be attributed to the enhanced polymerase activity when coupled with a well-balanced HA and NA activity.

The changes associated with an increase in virus polymerase activity at 33°C were acquired during the Pacific coast wild bird/backyard poultry virus circulation. The PB2 L386V mutation, which was found in all H5N2 viruses from the outbreak except the two Northern Pintail viruses, had no effect on polymerase activity. The increased polymerase activity at 33°C with PB2 V649I and NS1 I176T suggests that these changes are associated with adaptation of the viral polymerase to lower temperatures. Increased polymerase activity with V649I in PB2 at 33°C has also been reported in other studies as being related to increased virulence and transmissibility in mice and ferrets (Table 4). The Bayesian reconstruction inferred that the V649I change occurred during the Pacific coast wild bird/backyard poultry virus circulation, and the mutation was maintained in all viruses isolated from poultry during the U.S. H5N2 outbreak. The increased polymerase activity at low temperatures as it relates to virus fitness in different avian hosts and virus survival in different environments remains to be elucidated.

We also found that the polymerase activity increased at 37°C as we added the changes associated with the circulation of the virus in poultry. Substitution of PB1 (rgDk/Ck-PB1) improved the infectivity and shortened the MDT in chickens compared to that of the rgDk virus, which is consistent with the increased polymerase activity found with the E180D and M317V mutations at 37°C. When examining the amino acid usage in the U.S. H5N2 viruses, the PB1 R215K change was maintained after introduction of the virus to the Midwestern turkey farms (groups 2a, 2b, 2c, and 2e). The R215K change was also found in an isolate from a turkey farm in Ontario where a distinct introduction from wild bird dissemination occurred (34). The PB1 E180D and M317V mutations occurred before virus detection in the later cases in chickens (group 2e). A similar hydrophobic side chain change from methionine to an isoleucine at the same residue 317 in PB1 was found in other instances where HPAI viruses circulated in poultry (Table 4). The enhanced polymerase activity at a higher temperature (37°C) could explain the increased virus replication in avian cells and higher mortality in chickens. However, further analysis is needed to corroborate whether the increase in polymerase activity alone was enough to explain the phenotypical changes observed both in vivo and in vitro.

Not much is known about the role of residues 337 and 475 in PA, but the rgDk/Ck-PA virus was not able to infect all chickens and had an MDT similar to that of the rgDk virus. Additionally, the lack of a significant increase of the polymerase activity with the A337V and A475T changes in PA suggest that these two changes were not critical in adaptation of the H5N2 virus in chickens.

The single substitution of I109T in NP enhanced the activity of the rgDk polymerase. The I109T mutation was the only change found in NP, so the increased pathogenicity and shorter MDT in chickens inoculated with rgDk/Ck-NP indicate that this mutation is a molecular determinant for increased pathogenicity in chickens. The I109T mutation was mainly present in the late chicken isolates (group 2e), but the same mutation was also found in some viruses isolated in turkeys in group 2a. It is also noted that the same I109T mutation was identified in an early H5N2 virus incursion into a turkey farm followed by transmission to a chicken farm in British Columbia (34). The parallel mutation suggests that the I109T mutation is related to adaptation of the virus to poultry.

The NS modulates the virulence of influenza viruses by antagonizing the interferon-mediated antiviral host response in mammalian (35, 36) and avian (37–39) species. The longer duration of virus shedding of rgDk/Ck-NS in mallards compared to rgDk suggests that the three NS mutations are engaged in regulating the innate immune system in mallards but not in chickens. For example, downregulation of the innate immune system mediated by NS1 could affect interferon and RIG-I expression, which is species specific (40, 41).

The thermal persistence of rgDk/Ck-HA.NA was higher than that of rgDk, indicating that HA and NA mutations are associated with the virion stability at 28°C and 39°C. This could benefit survival of the virus outside the host, thereby increasing the opportunity of mechanical or environmental virus transmission in farm settings and in our experimental setting. Although the critical mutations associated with this outcome remain to be clarified, mutations in the HA and NA that accumulated after circulation of the virus in poultry are likely linked to the longer infectious period observed in layer farms in Iowa at the later stage of outbreak (42). In contrast, the mutations that accumulated in the polymerase complex proteins had no impact on virus persistence at the temperatures tested regardless of the enhanced polymerase activity.

Other mutations not addressed in this study also arose during circulation of the virus in poultry (6). Nine additional mutations were identified in viruses from turkey farms (PB1 R723L; HA L6I, P136S, and K234R; NP M105V and M105I; and NA T380I, V412A, and R430K), and a mutation in PB2 (K54R) was identified after further virus transmission to chicken farms in Iowa. The HA P136S mutation is located at a previously identified antigenic site (43), and the NP M105V mutation is known to enhance the polymerase activity of a Gs/GD lineage H5N1 virus in chicken cells (26). The role of the rest of the mutations as they relate to increased virus fitness in poultry remains to be determined.

To summarize, the mutations in the polymerase genes that most likely occurred during circulation of the H5N2 HPAI virus in turkeys (PB1 E180D and M317V and NP I109T) contributed to the increased polymerase activity at 37°C. Although body temperatures for gallinaceous species and waterfowl are similar and above the temperature tested, the increased polymerase activity and, consequently, increased virus replication, was reproduced in the chicken experiment, with rgDk/Ck-PB1, rgDk/Ck-NP, and rgDk/Ck-POL causing faster onset of disease and higher levels of virus shedding than rgDk. As observed with rgDk/Ck-HA.NA, the mutations in functionally known regions of the HA and NA genes, which occurred during virus circulation in turkeys (HA S141P and NA S416G) and chickens (HA M66I and L322Q), contributed to the increased virulence yet induced a longer MDT, which likely contributed to the transmission of the virus. These mutations in the PB1, NP, HA, and NA virus proteins were also highly conserved in the poultry isolates. These findings suggest that the H5N2 virus evolved to increase virulence in chickens by acquiring changes in the polymerase complex as well as the HA and NA during virus spread, first in turkeys and then in chickens. The knowledge acquired in these studies helps in identifying emerging wild bird AI viruses potentially infectious to domestic poultry by carrying mutations associated with adaptation to gallinaceous species and in understanding the genetic changes that occur as AI viruses adapt to poultry.

MATERIALS AND METHODS

Phylogenetic analysis.

All available full-genome sequences of the H5N2 HPAI clade 2.3.4.4 viruses isolated in North America were retrieved from the Influenza Research Database (n = 269). Bayesian relaxed-clock phylogenetic analysis was performed with a concatenated genome using the BEAST v1.10.4 program. Uncorrelated lognormal relaxed-clock models and Gaussian Markov random field (GMRF) Bayesian skyline coalescent tree prior were used in the analyses. The Hasegawa–Kishono–Yano and gamma site heterogeneity model was used as the nucleotide substitution model. A discrete-trait phylodynamic model was used to infer viral host, using an asymmetric substitution model with the Bayesian stochastic search variable selection and a strict-clock model. Host type (wild bird, turkey, and chicken) were defined in the model as discrete nominal categories. Three independent Markov chain Monte Carlo analyses were run for 100 million generations, with sampling every 10,000 steps, and then 10% of the initial chain was removed as burn-in. Three independent runs were combined to ensure that adequate effective sample size (200) was reached for relevant parameters. The maximum clade credibility trees were estimated from a posterior (posterior, >0.5) distribution of trees with TreeAnnotator v1.10.1 and visualized using FigTree v1.4.4. Genetic groups (1, 2a, 2b, 2c, 2d, and 2e) were classified by following the previous study (6), and corresponding viruses are indicated with the phylogenetic tree.

Generation of reassortant viruses.

Recombinant A/Northern Pintail/Washington/40964/2014 (rgDk) and A/Chicken/Iowa/13388/2015 H5N2 (rgCk) HPAI viruses, and reassortant viruses obtained by replacing genes from the wild bird virus with genes from the chicken virus, were generated (Fig. 2). These two strains were chosen based on the difference in virus fitness in mallards and chickens (7, 8) and have 20 amino acid differences between them (Fig. 1). The bidirectional influenza reverse genetics plasmid vector was constructed using human RNA polymerase I promoter (270 bp) and mouse RNA polymerase terminator (34 bp) sequences from pHH21 reverse genetics plasmid (45, 46) and human cytomegalovirus immediate-early promoter using pCMV-DSRed-Express plasmid as the backbone (Clontech Laboratories, Inc., USA). The DSRed gene was deleted and a cassette containing human RNA polymerase I promoter, BsmBI restriction enzyme site for insertion of influenza virus gene, and mouse RNA polymerase I terminator was inserted to obtain pCMV-BDRG vector and was previously evaluated for rescue of A/PR/8/1934 H1N1 virus (S. Jadhao and D. L. Suarez, unpublished data). Viral RNA segments from rgDk and rgCk HPAI viruses were reverse transcription-PCR (RT-PCR) amplified by using segment-specific oligonucleotides containing BsmBI restriction sites and cloned in the pCMV-BDRG vector. Transformed Escherichia coli cells were selected using kanamycin, and plasmids were purified and noncoding and coding sequences of the cloned viral gene segments were verified by Sanger nucleotide sequencing. Cells (293T; ATCC CRL-3216) in 6-well plates were transfected with 0.3 μg of each of the eight expression plasmids and 7.5 μl of TransIT-LT1 (Mirus Bio, Madison, WI) in 250 μl of Opti-MEM (Life Technologies, Grand Island, NY). At 12 h after transfection, chicken embryo fibroblast cells (ATCC CRL-12203) were seeded onto the transfected 293T cells. Five 10-day-old embryonated chicken eggs (ECEs) were inoculated with 100 μl of the supernatants harvested at 48 h after transfection. The allantoic fluid was harvested at 48 h after inoculation and titrated in ECEs. Sequences of the rescued viruses were confirmed by sequence-dependent amplification. All experiments with the HPAI viruses were conducted in biosafety level 3 enhanced facilities in accordance with procedures approved by the U.S. National Poultry Research Center (USNPRC) Institutional Biosecurity Committee.

Animal experiments.

Two bird experiments were conducted, one in chickens and one in mallards. The protocols used were chosen to reflect previously observed differences between infection with the viruses in chickens (clinical signs, virus shedding, and transmission) and in mallards (virus shed and replication in tissues). Three-week-old specific-pathogen-free (SPF) White Leghorn chickens (Gallus gallus domesticus) were obtained from the USNPRC in-house flocks. Mallard ducks (Anas platyrhynchos) were obtained at 1 day of age from a commercial hatchery and held for 2 weeks at the USNPRC. Serum samples were collected to confirm that the birds were serologically negative to AI virus by blocking enzyme-linked immunosorbent assay (FlockCheck avian influenza multiscreen antibody test; IDEXX Laboratories, Westbrook, ME, USA). Feed and water were provided with ad libitum access. Chickens (five birds per group) and mallards (10 birds per group) were inoculated by the intrachoanal route with 10⁷ 50% egg infectious doses (EID₅₀)/0.1 ml of virus. The inoculum titers were verified by back titration in ECEs. To evaluate virus transmissibility in chickens, three noninoculated hatch mates were added to each virus group at 1 day postinoculation (dpi). Oropharyngeal (OP) and cloacal (CL) swabs were collected from chickens at 1, 2, 3, 4, and 5 dpi and from mallards at 2, 4, 7, and 10 dpi. These schedules are based on our previous studies to best represent virus shedding kinetics in each bird species (7, 8). Swabs were suspended in 1.0 ml of brain heart infusion (BHI) with penicillin (2,000 U/ml; Sigma-Aldrich), gentamicin (200 μg/ml; Sigma-Aldrich), and amphotericin B (5 μg/ml; Sigma-Aldrich) and frozen at −80°C. To measure the tissue tropism of the viruses, two mallards were euthanized and necropsied at 3 dpi. Portions of lung, heart, brain, muscle, and spleen were collected and frozen at −80°C. Birds that survived were bled and euthanized at 14 dpi. Seroconversion was determined by hemagglutination inhibition (HI) assays using standard methods and homologous antigen (47). Birds were considered infected if virus shedding was detected and/or birds were seropositive at 14 dpi. This study and associated procedures were reviewed and approved by the USNPRC Institutional Animal Care and Use Committee.

Viral RNA quantification in swabs and tissues.

The presence of virus in OP and CL swabs and tissues was determined by quantitative real-time RT-PCR (qRRT-PCR). For swab virus quantification, RNA was extracted using a MagMAX-96 AI/ND viral RNA isolation kit (Ambion, Inc., Austin, TX) by following the manufacturer’s instructions. The qRRT-PCR targeting the influenza virus matrix gene (48) was conducted using the AgPath-ID one-step RT-PCR kit (Ambion, Inc., Austin, TX). Virus titers in tissue samples were determined by weighing, homogenizing, and diluting tissues in BHI to a 10% (wt/vol) concentration. Total RNA was extracted from the collected tissues using TRIzol LS reagent (Invitrogen, Carlsbad, CA) and the RNA Clean & Concentrator kit (Zymo Research, Orange, CA). To further standardize the amount of nonspecific RNA from tissue, the resulting total RNA extracts were quantified and diluted to 50 ng/μl. To avoid a bias in virus quantification caused by different infectivity of the viruses in ECEs, mean virus shedding was calculated in virus copy number using the qRRT-PCR results. The viral copy numbers in swab and tissue samples were extrapolated by using a standard curve correlating with a known number of matrix gene copies incorporated in the bidirectional plasmid used for the virus rescue. For statistical purposes, qRRT-PCR negative samples were given the value of the limit of detection (10^2.8 virus copy number/ml).

Viral growth kinetics and plaque size measurement.

Viral growth kinetics were examined in the DF-1 (chicken embryo fibroblast; ATCC CRL-12203) and CCL-141 (duck embryo fibroblast; ATCC CCL-141) lines. Monolayers from each cell type were trypsinized with 0.25% trypsin-EDTA (Invitrogen, Carlsbad, CA, USA) and counted with a hemocytometer. Titrated viral stocks of the rgDk, rgDk/Ck-HA.NA, rgDk/Ck-POL, and rgCk viruses were diluted and used to infect cells in triplicate at a multiplicity of infection (MOI) of 0.00001. Viruses were adsorbed for 40 min, and the inoculum was removed and washed with sterile phosphate-buffered saline (PBS). Supernatants were collected at 12, 24, 36, 48, and 60 h postinfection (hpi) for the chicken cells and at 24, 36, 48, 60, and 72 hpi for the duck cells. Titers were measured by plaque assay on the Madin-Darby canine kidney (MDCK; ATCC) monolayers. Plaque numbers and sizes were calculated by using ImageJ software ver. 1.51j8.

vRNP reconstitution assay.

Polymerase complex activity of the recombinant viruses was quantified as described previously (12). In addition, nucleotide substitutions resulting in amino acid changes were created using the QuikChange Lightning site-directed mutagenesis kit (Agilent Technologies, Santa Clara, CA) to represent intermediate vRNP inferred by the phylogenetic tree. Briefly, DF-1 cells were seeded in 12-well plates and transfected with 0.3 μg of the luciferase reporter plasmid (pMACkv-Gluc) and 0.3 μg of each bidirectional plasmid encoding PB2, PB1, PA, NP, and NS genes from rgDk and rgCk in different combinations using the TransIT-LT1 reagent (Mirus, Madison, WI). The cells were incubated at 33°C or 37°C for the wild-type and reconstructed polymerases and at 37°C for the single mutation polymerases. At 48 h posttransfection, supernatants from transfected cells were harvested and assayed with Pierce Gaussian luciferase glow assay (ThermoFisher, Waltham, MA). Relative polymerase activity was calculated as the ratio luminescence level of a given vRNP to the average luminescence level of rgDk vRNP for three independent experiments.

Viral titer reduction at different temperatures.

Stocks of rgDk, rgDk/Ck-HA.NA, rgDk/Ck-POL, and rgCk viruses were diluted 1:25 to 1:100 in phosphate-buffered saline to achieve an initial titer in the range of 7 to 8 log₁₀ EID₅₀/ml. Three different batches of the diluted viruses were incubated at 20°C, 28°C, and 39°C. A portion of the incubation was taken at 0, 6, 24, 48, 72, and 120 h after incubation and stored at −70°C until further virus titration in ECEs. Similar to virus stability tests under different conditions (49), linear regression was applied to estimate a 99% virus reduction time (Rt) for each batch; Rt values correspond to the time required for a decrease in viral titer of 2 log₁₀ EID₅₀/ml. Regression lines were calculated by GraphPad Prism 8 software.

Statistical analysis.

Statistical differences in mean HI titers, mean plaque size, mean polymerase activity, and mean Rt values were analyzed by Tukey one-way analysis of variance (ANOVA) using GraphPad Prism 8 software. Since the values of virus copy numbers were not normally distributed when assessed by the D’Agostino and Pearson test, significant difference between groups was analyzed by the nonparametric Kruskal-Wallis test followed by Dunn’s post hoc test (GraphPad Prism). A P value of <0.05 was set to be significant. To compare survival curves of chickens inoculated with the single or multiple reassortant viruses to that of rgDk virus, we performed the log-rank (Mantel-Cox) test with a Bonferroni correction for multiple comparisons. The P values below the Bonferroni-corrected threshold calculated (n = 0.006) were considered statistically significant.

Data availability.

Virus sequences of the reassortant viruses generated in this study were confirmed by sequencing. Each segment of the reassortant viruses was identical to its donor viruses as disclosed in GenBank database (accession numbers KX351784–KX351791 and KP307973–KP307980)