Low Polymerase Activity Attributed to PA Drives the Acquisition of the PB2 E627K Mutation of H7N9 Avian Influenza Virus in Mammals

The emergence of the PB2 E627K substitution is critical in the mammalian adaptation and pathogenesis of AIV. H7N9 AIVs that emerged in 2013 possess a prominent ability in gaining the PB2 E627K mutation in humans. Here, we demonstrate that the acquisition of the H7N9 PB2 E627K mutation is driven by the low polymerase activity conferred by the viral PA protein in human cells, and four PA residues are collectively involved in this process. Notably, the H7N9 PA protein leads to significant dependence of viral polymerase function on human ANP32A protein, and Anp32a knockout abolishes PB2 E627K acquisition in mice. These findings reveal that viral PA and host ANP32A are crucial for the emergence of PB2 E627K during adaptation of H7N9 AIVs to humans.

important. Amino acid mutations in HA are known to play important roles in promoting efficient binding to human-type receptors, thereby facilitating the replication and transmission of influenza virus in humans (2)(3)(4)(5). Using a ferret model, Lakdawala et al. showed that transmissible influenza viruses with human-type receptor preference are rapidly selected in the soft palate, where long-chain ␣2,6-sialic acids predominate on the nasopharyngeal surface (6). Different mutations in PB2 have been identified to contribute to the adaptation of influenza viruses to mammalian hosts; these mutations include E627K (7,8), D701N (9, 10), T271A (11), Q591R/K (12,13), and E158G (14).
The key role of the PB2 E627K substitution in the mammalian adaptation of AIVs has been well documented, having been shown to enhance polymerase activity, virus replication, pathogenicity, and transmission of AIVs in humans and other mammals (2,7,8,(15)(16)(17)(18)(19). However, not all AIVs acquire the PB2 E627K mutation upon infection of humans, such as many H5N1 viruses (20,21). Therefore, understanding the biological mechanism involved in the occurrence of this substitution is important to gain insights into the adaptation and pathogenesis of AIVs in mammals.
In early 2013 in China, a novel H7N9 AIV crossed the species barrier and caused the first human infection (22). To date, five epidemic waves of human infection with H7N9 viruses have caused 1,567 cases, with a fatality rate of approximately 39% (23). Moreover, highly pathogenic H7N9 viruses possessing a multibasic cleavage site motif in HA were detected in 2017 and caused severe threats to the poultry industry and human health (18,(24)(25)(26). Remarkably, the PB2 of H7N9 viruses easily acquires the PB2 E627K or D701N adaptive mutation during replication in mammalian hosts (18,22,27). In particular, most H7N9 human isolates acquire the PB2 E627K substitution (18). The prominent ability of H7N9 AIVs to acquire the PB2 E627K substitution upon infection of humans provides us with an excellent opportunity to decipher the underlying mechanism by which AIVs rapidly acquire the PB2 E627K substitution during their adaptation to mammalian hosts, including humans.
H7N9 AIV is a triple reassortant virus, with all six internal genes being derived from H9N2 AIVs (19,28). In this study, we generated a series of reassortant or mutant viruses in the background of an avian 2013 H7N9 virus, harboring genes or mutations from an avian H9N2 virus, and monitored the residue phenotype of PB2 627 during viral passages in mice. We found that the low polymerase activity attributed to the H7N9 viral PA protein is the intrinsic force driving the emergence of the PB2 E627K mutation during the replication of H7N9 AIVs in mammals, a process that also involves interplay with host factors, such as ANP32A.

Different avian influenza viruses have different capabilities to acquire the PB2 E627K mutation during replication in MDCK cells and mice.
To investigate whether the H7N9 and H9N2 viruses have the same capability to obtain the PB2 E627K mutation during replication in a mammalian host, we first selected an H7N9 avian strain, A/pigeon/Shanghai/S1421/2013 [PG/S1421(H7N9)], and five H9N2 avian strains and passaged them in MDCK cells. As shown in Table 1, two H9N2 viruses, A/chicken/ Henan/5/1998 [CK/5(H9N2)] and A/chicken/Guangxi/9/1999 [CK/9(H9N2)], isolated in the 1990s (29) did not acquire the PB2 E627K mutation even after being passaged six times in MDCK cells. Three other H9N2 viruses and the PG/S1421(H7N9) virus obtained the PB2 E627K mutation at passages 3 to 6 in MDCK cells (Table 1). The phenotypic emergence of PB2 E627K mutation was confirmed as occurring during virus passage because deep sequencing did not identify any subpopulations of PB2 627K in the original stocks of PG/S1421(H7N9) (see Table S1 in the supplemental material) and CK/SC197(H9N2) and CK/SC324(H9N2) virus (30).
Next, we determined the residue phenotype of PB2 627 for PG/S1421(H7N9) and the two H9N2 strains, CK/5(H9N2) and CK/9(H9N2), during in vivo passages in mice. PG/S1421(H7N9) quickly acquired the PB2 E627K mutation on day 5 after one passage in mice, CK/9(H9N2) acquired the PB2 E627K mutation after the second passage, whereas for CK/5(H9N2), the PB2 627E residue was stably maintained even after four passages in mice (Table 1). CK/5(H9N2), a representative of viruses that maintain a stable PB2 627E during replication in mammals, was selected with PG/S1421(H7N9) to form a model virus pair to investigate the viral factors that drive the emergence of the PB2 E627K mutation during adaptation of H7N9 viruses to mammalian hosts.
The low polymerase activity attributed to PA drives the emergence of the mammalian-adaptive PB2 E627K mutation. We used PG/S1421(H7N9) as the backbone to generate reassortant viruses, each containing one gene derived from CK/ 5(H9N2), as described previously (31). The rescued single-gene reassortant viruses were intranasally inoculated into mice to monitor their ability to acquire the PB2 E627K mutation. As shown in Fig. 1A, the reassortants exhibited variable replicative abilities in the lungs of infected mice: the reassortant containing CK/5(H9N2) PB1 replicated to higher titers than parental PG/S1421(H7N9); the replication titers of reassortants containing the PB2, PA, HA, NA, or M protein of CK/5(H9N2) were lower than that of PG/S1421(H7N9); and the reassortants containing the NP or NS of CK/5(H9N2) replicated poorly in the lungs of infected mice with no virus detected on day 5 postinfection (p.i.). We then sequenced the PB2 gene of the reassortant viruses recovered from the lungs of the infected mice. The reassortants containing the PB2, HA, NA, or M protein of CK/5(H9N2) acquired the PB2 E627K mutation on day 5 p.i. in mice, just like parental PG/S1421(H7N9) (Fig. 1B). The reassortant virus containing CK/5(H9N2) PB1 obtained a PB2 D701N mutation on day 5 p.i. (Fig. 1B). In contrast, the reassortant virus bearing CK/5(H9N2) PA was stable during replication in mice with no mutations acquired in PB2 (Fig. 1B). The recovered PG/S1421-CK/5PA(H7N9) virus was then reinoculated into mice to identify potential adaptive mutations in PB2. Again, no mutations occurred in PB2 of PG/S1421-CK/5PA(H7N9) recovered from the second-passage mice (Fig. 1B). Deep sequencing of PG/S1421-CK/5PA(H7N9) recovered from the lungs of the secondpassage mice demonstrated that the PB2 E627K mutation was not acquired (Table S1). These results indicate that the PA gene carried by PG/S1421(H7N9) is responsible for driving the emergence of the PB2 E627K mutation in mice and that replacing PG/ S1421(H7N9) PA with CK/5(H9N2) PA abolishes the acquisition of the adaptive PB2 E627K mutation in the mammalian host.
The PA protein is a component of the viral ribonucleoprotein (RNP) complex, which is important for virus replication and virulence (32)(33)(34)(35)(36). We therefore performed a minigenome assay to determine the viral polymerase activity of four RNP combinations possessed by PG/S1421(H7N9), CK/5(H9N2), and two single PA reassortants in avian DF-1 cells. As shown in Fig. S1, the polymerase activity of PG/S1421(H7N9) (i.e., 7 PB2 7 PB1 7 PA 7 NP ) was much higher than that of CK/5(H9N2) (i.e., 9 PB2 9 PB1 9 PA 9 NP ) (the gene segments derived from the parental H7N9 virus were designated "7," and those derived from the parental H9N2 virus were designated "9"), which may partially explain the prevalence of H7N9 viruses in avian species. Notably, the polymerase activity of 7 PB2 7 PB1 9 PA 7 NP was dramatically decreased compared with that of 7 PB2 7 PB1 7 PA 7 NP , whereas the polymerase activity of 9 PB2 9 PB1 7 PA 9 NP was slightly increased compared with that of 9 PB2 9 PB1 9 PA 9 NP . These data suggest that the PG/S1421(H7N9) PA is better able to confer high polymerase activity in an avian host, especially in the H7N9 virus background.
Four amino acids in the N-terminal PA domain are critical in mediating the acquisition of PB2 E627K. To explore the specific domain of PG/S1421(H7N9) PA that mediates the emergence of the PB2 E627K mutation of H7N9 viruses, we created two PA chimeras in which domain 1-252 or domain 253-716 of PG/S1421(H7N9) PA was replaced with that of CK/5(H9N2) PA ( Fig. 2A). The RNP combination containing domain 1-252 of CK/5(H9N2) PA (i.e., 7 PB2 7 PB1 7 9(1-252)PA 7 NP ) possessed significantly higher polymerase activity than that of 7 PB2 7 PB1 7 PA 7 NP in HEK293T cells, whereas the activity of the RNP combination containing domain 253-716 of CK/5(H9N2) PA (i.e., 7 PB2 7 PB1 7 9(253-716)PA 7 NP ) was as low as that of 7 PB2 7 PB1 7 PA 7 NP (Fig. 2B). Two chimeric PA viruses were then rescued in the background of PG/S1421(H7N9) and evaluated in mice. The chimeric PA virus containing domain 1-252 of CK/5(H9N2) PA replicated in mice without acquiring the PB2 E627K mutation (Fig. 2C). In contrast, the chimeric PA virus bearing domain 253-716 of CK/5(H9N2) PA acquired the PB2 E627K mutation during replication in mice (Fig. 2C). These results demonstrate that the N-terminal 1-252 domain of PA (PAN) is critical in mediating the acquisition of the PB2 E627K substitution during the replication of H7N9 virus in mammalian hosts.
Next, we constructed another four PA chimeras containing amino acids 1 to 60, 1 to 86, 1 to 120, or 1 to 191 of CK/5(H9N2) PA in the backbone of PG/S1421(H7N9) PA to Schematic diagram of PA chimeras between PG/S1421(H7N9) and CK/5(H9N2) as well as PA mutants in the background of PG/S1421(H7N9). (B) The effects of PA chimeras or mutants on viral polymerase activities in the background of PG/S1421(H7N9) in a minigenome assay. Four RNP expression plasmids (PG/S1421-PB2, -PB1, -NP, and wild-type, chimeric, or mutant PA) were transfected into HEK293T cells together with pHH21-SC09NS F-Luc and pRL-TK reporters. Thirty hours later, the cells were harvested for luciferase assays. The values are standardized to the polymerase activity of PG/S1421(H7N9) (100%). The polymerase activity of PG/S1421(H7N9) bearing the PB2 E627K mutation was also included as a control. P values were determined by using a two-tailed unpaired Student's t test compared to the RNP complex bearing PG/S1421(H7N9) PA. NS, not significant. (C) Sequencing of the PB2 627 domain during the replication of PG/S1421(H7N9) viruses bearing wild-type, chimeric, or mutant PAs in mice. Viral RNAs were isolated from lung homogenates of infected mice (n ϭ 3) on day 5 p.i., and the PB2 segment was amplified by RT-PCR and sequenced to monitor adaptive mutations. The number of mice harboring the indicated PB2 residue is shown in parentheses. NA, virus was not recovered; red arrows indicate the appearance of the PB2 E627K mutation. Data are representative of three independent experiments (means Ϯ SDs) (B).
H7N9 PA determines the sensitivity of PB2 627E polymerase to variations in human ANP32A expression. Host factors ANP32A and ANP32B have been found to play an important role in viral genome transcription and replication of influenza virus (41)(42)(43). In comparison to ANP32B, ANP32A is reported to be involved in determining the host restriction of avian-like viral polymerase (PB2 627E) in human cells, demonstrating that avian ANP32A (avANP32A), due to unique sequence features, is able to effectively support the avian-like polymerase function compared with the human ANP32A (huANP32A) (41,44,45). However, whether huANP32A directly plays a role in the acquisition of the PB2 E627K substitution during the mammalian adaptation of AIVs has not been resolved. To address this point, we first determined the binding affinity of huANP32A for the trimeric polymerase complex of PG/S1421(H7N9) viruses bearing different PAs. To this end, we performed a GST pulldown assay in HEK293T cells that were transfected with a plasmid expressing GST or GST-ANP32A, together with plasmids for the expression of the viral polymerase (PG/S1421 PB2, PG/S1421 PB1, and different PAs). HuANP32A efficiently interacted with polymerase complexes containing CK/5(H9N2) PA, PG/S1421-CK/5PA 1-191 , or PG/S1421-PA 142R-147V-171V-182L , whereas it bound less efficiently to the polymerase complex containing PG/S1421(H7N9) PA (Fig. 4A). We then tested the polymerase activity of different RNP complexes when huANP32A was overexpressed or knocked down. Interestingly, although the interaction between huANP32A and the polymerase complex containing PG/S1421(H7N9) PA was relatively weak, overexpression of huANP32A in HEK293T cells (Fig. 4B) enhanced the polymerase activity of the PG/S1421(H7N9) RNP complex (by approximately 1.4-fold) (Fig. 4C). In contrast, the polymerase activities of the RNP complexes bearing CK/ 5(H9N2) PA, PG/S1421-CK/5PA 1-191 , or PG/S1421-PA 142R-147V-171V-182L were relatively unaffected by huANP32A overexpression (Fig. 4C). We also knocked down huANP32A expression by using short interfering RNA (siRNA) in HEK293T cells (Fig. 4D and  Fig. S4A). Significantly, knockdown of huANP32A expression led to an over 60% reduction in the polymerase activity of the PG/S1421(H7N9) RNP complex, whereas the polymerase activities of PG/S1421(H7N9) RNPs bearing CK/5(H9N2) PA, PG/S1421-CK/ PG/S1421-PAmut stands for the construct PG/S1421-PA 142R-147V-171V-182L ; data are representative of three (C, E, and G) or two (I) independent experiments (means Ϯ SDs); P values were determined by using a two-tailed unpaired Student's t test compared to the RNP complex bearing PG/S1421 PA (C, E, and G) and by using multiple t tests compared to viral titers in control A549 cells (I). NS, not significant. 5PA 1-191 , or PG/S1421-PA 142R-147V-171V-182L were much less affected by huANP32A knockdown (Fig. 4E). These results demonstrate that the diversity among different PAs influences the sensitivity of the H7N9 PB2 627E polymerase to up-or downregulation of human ANP32A expression in human cells. Furthermore, the H7N9 PA protein is responsible for the low activity of PB2 627E polymerase, being highly sensitive to variations in human ANP32A expression.
We also performed a minigenome assay for the PG/S1421(H7N9) RNP bearing the PB2 627K mutation and found that its polymerase activity was unchanged when the expression of huANP32A was knocked out (Fig. S5A). Moreover, the depletion of ANP32A protein did not have any adverse effects on the growth of PB2 627K mutant virus in A549 cells (Fig. S5B). These results indicate that once the viral polymerase activity is compensated for by the PB2 E627K mutation, the polymerase function and virus growth are no longer influenced by huANP32A in human cells.
The impaired H7N9 PB2 627E polymerase activity incurred by ANP32A depletion drives the acquisition of adaptive PB2 mutation. We demonstrated that the polymerase activity and growth of PG/S1421(H7N9) are dramatically impaired by the depletion of ANP32A protein in human cells. To determine how the PG/S1421(H7N9) virus behaves under such conditions in vivo, we generated Anp32a Ϫ/Ϫ C57BL/6J mice by use of CRISPR/Cas9-mediated gene targeting. A targeting construct was designed to delete exons 2 to 4 of Anp32a (Fig. 5A). Deletion of Anp32a was confirmed by sequencing, and the absence of ANP32A protein from Anp32a Ϫ/Ϫ mice was confirmed by Western blotting of lung extracts (Fig. 5B). The Anp32a Ϫ/Ϫ mice and matched wild-type (WT) controls were intranasally inoculated with PG/S1421(H7N9), and mouse lungs were collected on day 5 p.i. for virus isolation. The titers of PG/S1421(H7N9) virus in Anp32a Ϫ/Ϫ mice were dramatically decreased compared with those in WT mice (Fig. 5C). Sequence analysis confirmed that the PB2 E627K mutation was acquired by virus in the lungs of the infected WT mice (Fig. 5D). Notably, PG/S1421(H7N9) replication was not detected in one of the five Anp32a Ϫ/Ϫ mice, PB2 627E was maintained in viruses recovered from two Anp32a Ϫ/Ϫ mice, and a PB2 D701N mutation emerged in viruses in the other two Anp32a Ϫ/Ϫ mice (Fig. 5D). These results indicate that the impaired polymerase activity caused by the depletion of ANP32A protein significantly reduced the replication of H7N9 PB2 627E virus in Anp32a Ϫ/Ϫ mice. The absence of ANP32A protein abolished the acquisition of the PB2 E627K mutation and instead forced the virus down an alternative adaptive pathway to acquire the compensatory PB2 D701N mutation.

DISCUSSION
Introduction of adaptive mutations in viral polymerase proteins is critical for AIVs to cross the species barrier to infect and kill humans. The role of the PB2 E627K mutation in mammalian adaptation has been well established for a wide range of AIVs, such as the H5N1, H7N7, H7N9, and H10N8 subtypes (7,22,47,48). In this study, we used the H7N9 Viral PA Drives the PB2 E627K Mutation ® H7N9 virus model, which easily acquires the PB2 E627K mutation when it infects humans (18), to investigate the molecular basis for the emergence of PB2 E627K during human adaptation. Our results show that the polymerase activity of PG/S1421(H7N9) is significantly lower than that of CK/5(H9N2) in HEK293T cells, which is the opposite of our findings in avian DF-1 cells. Importantly, we found that the low polymerase activity attributed to the PA protein of H7N9 AIVs in human cells is the intrinsic driving force to acquire the PB2 E627K mutation during replication in wild-type mice. Moreover, the H7N9 PA protein is responsible for the sensitivity of the viral polymerase activity and growth to variations in the expression of the huANP32A protein in human cells. The impaired polymerase activity of H7N9 AIV due to the knockout of ANP32A reduces virus replication in Anp32a Ϫ/Ϫ mice. Under such conditions, ANP32A depletion abolishes the acquisition of the PB2 E627K mutation, thereby forcing the virus to acquire the alternative adaptive PB2 D701N mutation. Our data thus reveal that the prominent ability of H7N9 AIVs to acquire the mammalian-adaptive PB2 E627K mutation is driven by the intrinsic low polymerase activity attributed to the viral PA protein, which also involves interplay with mammalian ANP32A.
The PA protein has been well documented to be involved in the pathogenesis of influenza virus (32)(33)(34)(35)(36). It is separated into two domains linked by a protease-sensitive peptide (49). The crystal structure of the N-terminal PA domain has been resolved for several influenza viruses (50,51). The amino acid position 142 is correlated with the pathogenicity of AIV (52,53) and is part of a loop connecting ␣5 and ␤4, residue 147 is a part of sheet ␤4, and residues 171 and 182 are localized at helix ␣6. To better view the location of these four critical PA residues, we mapped them in the structure of the viral polymerase complex (54) (see Fig. S6 in the supplemental material). Although (C) Replication of PG/S1421(H7N9) in Anp32a Ϫ/Ϫ mice. Wild-type (WT) and Anp32a Ϫ/Ϫ C57BL/6J mice (n ϭ 5 per genotype) were inoculated with 10 6 EID 50 of PG/S1421(H7N9). Organs were collected on day 5 p.i. and titrated for virus infectivity in eggs. The data are presented as means Ϯ SDs for organ samples of five mice. P values were determined by using multiple t tests (one unpaired t test per row). (D) Phenotype of PB2 residues 627 and 701 during replication of PG/S1421(H7N9) virus in WT or Anp32a Ϫ/Ϫ mice (n ϭ 5). Viral RNAs were isolated from lung homogenates of infected mice on day 5 p.i., and the PB2 segment was amplified by RT-PCR and sequenced to monitor adaptive mutations arising during passage in mice. /, virus was not detected in one of the five Anp32a Ϫ/Ϫ mice. these four PA residues are not structurally adjacent to the PB2 627 residue, they could possibly affect the interaction of PA with other polymerase subunits and/or host proteins. By examining these four PA residues in AIVs of subtypes H1 through H16, we found that most AIVs possess the PG/S1421(H7N9)-like constellation (i.e., 142K, 147I, 171I, and 182M) and that few H9 AIVs (0.12%) had the CK/5(H9N2)-like constellation (i.e., 142R, 147V, 171V, and 182L) (Table S3). Of note, the four PG/S1421(H7N9)-like PA residues are primarily responsible for the low viral polymerase activity in human cells. Acquiring the potent polymerase substitution PB2 E627K is an effective way to overcome the species barrier.
A previous study by Long et al. focused on how the avian polymerase is restricted in human cells and identified that avANP32A can strongly support PB2 627E polymerase activity because it possesses an additional 33 amino acids compared with huANP32A (41). Domingues and Hale identified a SUMO-interacting-motif-like sequence unique to avANP32A that promotes PB2 627E polymerase activity (45). Baker et al. further demonstrated that avANP32A restores RNP complex assembly for avian polymerase in human cells by enhancing RNA synthesis (44). However, these studies did not explore the role of huANP32A in the emergence of the PB2 E627K mutation during the replication of AIVs under unfavorable conditions in human cells. In the present study, we found that H7N9 PB2 627E polymerases bearing different PAs show diversity in their sensitivity to the up-or downregulation of huANP32A expression. When PG/S1421(H7N9) PA was replaced with CK/5(H9N2) PA, PG/S1421-CK/5PA 1-191 , or PG/S1421-PA 142R-147V-171V-182L , the polymerase activity increased, abrogating the need to acquire the PB2 E627K mutation during virus replication in mice. The increased polymerase activity was accompanied by reduced sensitivity to changes in huANP32A expression even though the interaction between the polymerase and huANP32A was enhanced. In contrast, the low polymerase activity of PG/S1421(H7N9) conferred by the viral PA protein was highly sensitive to the depletion of huANP32A protein although their interaction was comparatively weak, suggesting that the H7N9 PB2 627E polymerase can efficiently exploit human ANP32A. The impaired PB2 627E polymerase activity due to the depletion of ANP32A led to reduced virus replication and created the need for adaptive mutations in Anp32a Ϫ/Ϫ mice. Moreover, the depletion of ANP32A abolished the acquisition of the adaptive PB2 E627K mutation, forcing the virus to use an alternative adaptive pathway to acquire the PB2 D701N mutation.
The activity of PG/S1421(H7N9) polymerase bearing the PB2 627K mutation was not affected by the depletion of ANP32A protein in human cells. The replication of the PB2 627K mutant virus was also unchanged in this context. These findings imply that huANP32A has less of a role in maintaining the high polymerase activity of H7N9 PB2 627K virus. The enhanced polymerase activity of H7N9 PB2 627K virus is most likely mainly conferred by the attribute of the PB2 627K residue, which forms a contiguous basic face in the PB2 627 domain that is important for RNA binding and polymerase function in mammalian cells (55).
The introduction of CK/5(H9N2) PB1 into the PG/S1421(H7N9) backbone led to the appearance of the adaptive mutation PB2 D701N (9,10,(38)(39)(40), during the replication of PG/S1421-CK/5PB1(H7N9) in mice. This finding suggests that the acquisition of the PB2 D701N mutation may correlate with the origin of the viral PB1 gene. Furthermore, ANP32A depletion diverts the H7N9 polymerase adaptation from acquiring the PB2 E627K mutation, instead forcing the polymerase adaptation into an alternative pathway of acquiring PB2 D701N. Our data suggest that the host factors involved in the adaptation of the avian polymerase may not be exclusive. Previous studies have shown that the PB2 D701N mutation is responsible for enhanced binding with importin ␣, leading to increased virus replication in mammalian cells (56,57). However, whether importin ␣ is involved in the emergence of the PB2 D701N mutation in the background of H7N9 viruses remains to be further investigated.
In summary, our study provides new insights into the biological mechanism employed by H7N9 AIVs to acquire the PB2 E627K substitution in order to adapt and become pathogenic in humans. We discovered that the low polymerase activity of H7N9 AIVs in human cells, an attribute conferred by the viral PA protein, is the intrinsic driving force in the emergence of the adaptive PB2 E627K mutation, which also involves mammalian ANP32A. The findings thus enhance our understanding of the adaptation and pathogenesis of AIVs in humans and other mammals.  (29,30). The H7N9 viruses A/pigeon/Shanghai/S1421/2013 [PG/S1421(H7N9)] and A/chicken/Guangdong/SD008/2017 [CK/SD008 (H7N9)] were reported in previous studies (18,19). All H7N9 and H9N2 viruses were propagated in embryonated chicken eggs.

MATERIALS AND METHODS
Plasmids. The construction of plasmids for virus rescue was performed as described previously (31). The PB2, PB1, PA, and NP genes of the respective influenza viruses, as well as human ANP32A (huANP32A), were cloned into the mammalian expression vector pCAGGS. GST-tagged ANP32A was constructed in pCAGGS with a GST tag at the N terminus and an SV40 large-T antigen nuclear localization signal (NLS) at the C terminus. Chimeric PA genes, single or multiple point mutants of the PA genes, and the PB2 E627K mutant (generated by using a PCR approach) were cloned into vectors pCAGGS and pHH21. Plasmid pHH21-SC09NS F-Luc and paviPolI-T-Luc, for the expression of a virus-like RNA bearing the firefly luciferase gene under the control of the human or avian RNA polymerase I promoter, were reported previously (34,37). All constructs were sequenced to ensure the absence of unwanted mutations. The primer sequences used for cloning are available upon request.
Passage of H9N2 and H7N9 viruses in MDCK cells. Confluent MDCK cells were inoculated for 1 h with the indicated H9N2 and H7N9 viruses at a multiplicity of infection (MOI) of 0.01. The cells were then washed and maintained in 1ϫ MEM (0.3% BSA, 0.5 g/ml N-tosyl-L-phenylalanyl chloromethyl ketone [TPCK]-treated trypsin) at 37°C. After 48 h of incubation, the virus supernatants were collected for sequencing and the next passage in MDCK cells.
Generation of reassortant viruses by reverse genetics. Reassortant viruses were generated by using the reverse genetics system as described previously (31). The rescued viruses were detected by using a hemagglutination assay and were fully sequenced to ensure the absence of unwanted mutations.
Quantification of viral RNA species. A549 cells grown in 6-well plates were infected with the indicated viruses at an MOI of 2. Total RNA was extracted by using an RNeasy Plus Mini Kit (Qiagen) at 6 and 8 h postinfection (p.i.) according to the manufacturer's instructions. Relative quantities of viral NP genomic RNA (vRNA), cRNA, and mRNA were determined by RT-qPCR using a tagged-primer system as described previously (37).
GST pulldown assay. HEK293T cells grown in 10-cm dishes were transfected with the indicated plasmids (PG/S1421 PB2 and PG/S1421 PB1, 10 g each; different PAs, 4 g each; GST or GST-ANP32A, 6 g) by using the Lipofectamine LTX and Plus reagents. At 48 h posttransfection, cells were lysed with NP-40 buffer (Beyotime Biotechnology). After centrifugation, the cleared lysates were incubated with glutathione-Sepharose 4B resin (GE Healthcare) at 4°C for 3 h. After three washes with cold PBS, the resin-bound proteins were separated by SDS-PAGE and detected by using a standard Western blotting procedure.
Liang et al.