Pathogenicity of the H1N1 influenza virus enhanced by functional synergy between the NPV100I and NAD248N pair

By comparing and measuring covariations of viral protein sequences from isolates of the 2009 pH1N1 influenza A virus (IAV), specific substitutions that co-occur in the NP-NA pair were identified. To investigate the effect of these co-occurring substitution pairs, the V100I substitution in NP and the D248N substitution in NA were introduced into laboratory-adapted WSN IAVs. The recombinant WSN with the covarying NPV100I-NAD248N pair exhibited enhanced pathogenicity, as characterized by increased viral production, increased death and inflammation of host cells, and high mortality in infected mice. Although direct interactions between the NPV100I and NAD248N proteins were not detected, the RNA-binding ability of NPV100I was increased, which was further strengthened by NAD248N, in expression-plasmid-transfected cells. Additionally, the NAD248N protein was frequently recruited within lipid rafts, indirectly affecting the RNA-binding ability of NP as well as viral release. Altogether, our data indicate that the covarying NPV100I-NAD248N pair obtained from 2009 pH1N1 IAV sequence information function together to synergistically augment viral assembly and release, which may explain the observed enhanced viral pathogenicity.


Introduction
The genome of influenza A virus (IAV), a member of the Orthomyxoviridae family, consists of a segmented negative-sense single-strand RNA, and changes in the influenza viral genome are frequent events due to the accumulation of mutations and reassortment of RNA segments [1]. In addition to seasonal flu, IAV pandemic infections occasionally occur, with severe impacts on public health and society. There have been extensive efforts using various approaches to understand the nature of influenza pandemics. Among them, sequence comparison and reverse genetics using recombinant virus in murine or primate systems have been useful for identifying key RNA segments or sequence substitutions that contribute to influenza viral pathogenicity.
The 2009 pandemic H1N1 influenza A virus (2009 pH1N1 IAV) emerged and spread rapidly [2]. Although the 2009 H1N1 pandemic flu appeared to be mild compared to the 1918 H1N1 Spanish flu or the 1968 H3N2 Hong Kong flu, it exhibited atypical pathological potential distinct from that of seasonal influenza, with an exceptionally rapid rate of spread [3]. Sequence comparison between 2009 pH1N1 IAV and other pandemic strains has been performed to characterize amino acid substitutions contributing to pathogenicity. For example, the HA2 E47K substitution in hemagglutinin (HA) of the 2009 pH1N1 IAV reduces the pH threshold for membrane fusion, conferring the virus with thermal stability and infectivity, which partially explains its rapid spread and adaptation to humans [4]. Similarly, 2009 pH1N1 IAV substitutions in neuraminidase (NA) allow for low-pH stability [5]. Amino acid substitutions in influenza virus PB2, PB1-F2, NP, and NS1 have also been assessed with regard to their contributions to enhancing viral propagation and infectivity [6][7][8][9]. However, none of these single mutations in 2009 pH1N1 IAVs sufficiently explains the pathogenicity of the 2009 pandemics. Notably, M2 protein channel activity enhanced 2009 pH1N1 IAV infectivity by protecting against premature HA cleavage and preserving membrane fusion competence. Additionally, the mutant form of NP exhibits selectively reduced NA expression, thereby indirectly increasing influenza virus fitness [10]. These findings indicate that two or more noninteracting proteins may also elicit concerted actions and that identifying sequence variations in viral or host factors that synergistically contribute to enhanced infection pathogenicity is important.
During evolution, molecules that perform similar functions tend to coevolve. In addition to physically interacting proteins, noninteracting proteins with similar functions are also expected to coevolve [11][12][13]. Therefore, we utilized a methodology to search for coevolution of amino acid sequence variations in 2009 pH1N1 IAV strains. In this report, we explore the synergistic effects of NP and NA substitutions that may contribute to the enhancement of influenza viral pathogenicity.

Identification of covarying NP and NA pair RNA segments
To identify pairs of genetic traits that may have functioned to augment influenza viral pathogenicity, we examined covariance in sequence information from the viral transcripts of pH1N1 IAV isolates obtained during the 2009 pandemic season. This identification is based on the fact that genes that function together tend to covary during the course of evolution [11,14]. In total, the sequences of 417 pH1N1 virus isolates collected in 17 countries were examined (from the Influenza Sequence & Epitope Database, ISED [15]). The number of different amino acid sequences between any chosen gene pair among the 417 isolates was counted, and the covariance of sequence variations among 10 viral proteins were compared in pairs using Pearson's correlation analysis (Fig 1A). Among 45 total pairwise comparisons, the covariance value of the NP and NA pair was exceptionally high (Fig 1B and S1 Table). These results suggest that if one virus isolate has sequence alternations in NP proteins, then it is likely that this isolate contains a similar degree of sequence alternation in the NA protein as well.
We next assessed whether any specific amino acid substitutions co-occurred in NP and NA pairs of the 2009 pH1N1 IAVs. For this purpose, amino acid sequences from the 417 isolates were compared with those of seasonal H1N1 influenza viruses obtained from 1958 to 2008 (from ISED [15]), and variabilities at each position were separately mapped ( Fig 1C). In general, relatively high variability was observed in NP and NA sequences from seasonal H1N1 influenza strains. In contrast, NP and NA sequences from isolated 2009 pH1N1 pools were relatively invariable, except for a couple of residues. For NP of the isolated 2009 pH1N1 IAVs, the 100 th and 373 th positions were highly variable. Valine (V) was observed as the 100 th residue of NP of the seasonal H1N1 IAVs, whereas an isoleucine (I) was present at this position in 75% of the 2009 pH1N1 isolates. For residue 373, threonine (T) was dominant NP of in isolated 2009 pH1N1 IAVs, but alanine (A) was most frequent in the seasonal H1N1 NP sequence ( Fig  1D). Highly variable residues at the 106 th and 248 th positions were also observed for NA of the isolated 2009 pH1N1 IAVs: approximately 26% of the 2009 H1N1 isolates had V instead of I at the 106 th position of NA, and 71% had an aspartic acid (D) instead of an asparagine (N) at the 248 th position (Fig 1D).
First, we measured the virus titer (TCID 50 /ml) of the recombinant WSN viruses and the relative number of virus particle present based on M gene quantitation. To evaluate the replication kinetics of the recombinant WSN viruses, A549 cells were infected with the same amounts of recombinant WSN (10 TCID 50 /ml, 0.01MOI), and intracellular viral loads were measured at 3, 6, 12, and 24 hours postinfection. The relative virus particle numbers used in this experiment were similar in range, except for WSN viruses with the NP V100I and NA D248N mutations (Fig 2A), which were approximately one-fifth lower. To compare the differential kinetics of viral growth, viral loads at 6, 12, and 24 hours postinfection for each virus were separately normalized to the viral load at 3 hours postinfection for each virus. Although relative amount of intracellular viral genome was lower with NP V100I NA D248N influenza virus infection compared to NP WSN NA WSN influenza virus infection, WSN carrying the NP V100I NA D248N combination exhibited a high replication rate during the 24-hour period compared to the parental or other recombinant WSN viruses (Fig 2B and 2C). This result indicates that the presence of NA D248N in addition to NP V100I significantly enhanced relative infectivity compared to that of the parental WSN virus. Although infection with either NP WSN NA D248N WSN or NP V100I NA WSN WSN resulted in relatively higher levels of type I interferons (type I IFNs) and inflammatory cytokines, infection with NP V100I NA D248N WSN led to enormously high levels of cytokines, which cannot be explained by the additive effect of the dual NP and NA substitutions ( Fig 2D). We postulate that cytokine production might be indirectly enhanced by the increased viral production of NP V100I NA D248N influenza viruses. In addition, the recombinant NP V100I NA D248N WSN virus was also more toxic to host cells than was the parental or other single-mutant WSN virus (Fig 2E-2G). These findings indicate that a direct or indirect functional relationship may exist between NP V100I and NA D248N , possibly promoting influenza viral pathogenicity.

Recombinant NP V100I NA D248N WSN exhibits enhanced pathogenicity in vivo
We next evaluated the pathogenic effect of the NP V100I NA D248N WSN virus in BALB/c mice. Mice were intranasally infected with recombinant WSN influenza viruses (100 μl, 2000 TCID 50 /ml), and their survival rates were observed over a two-week period. For comparison, groups of mice were also infected with the NP V100I NA WSN or NP WSN NA D248N WSN virus. Infection with the NP V100I NA D248N WSN virus caused high mortality, with every mouse tested dying by 9 days postinfection (d.p.i.) (Fig 3A). In contrast, greatly reduced mortality was found with infection of either NP V100I NA WSN or parental WSN, with over 40% of NP WSN NA D248N WSN-infected mice being dead at 15 d.p.i. (Fig 3A). Mice infected with either NP V100I NA WSN or the parental WSN experienced weight loss but recovered, though infection with NP WSN NA D248N WSN resulted in nearly 30% weight loss at 15 d.p.i. (Fig 3B).
To better understand the pathogenicity of NP V100I NA D248N WSN, we next investigated the kinetics of viral replication and IFNβ production at an early time in the lungs of infected mice to evaluate viral replication alone and ignore factors influencing mouse survival. Viral titers obtained from the lungs of mice infected with NP V100I NA D248N recombinant WSN were significantly higher than those of mice infected with the parental WSN virus (Fig 3C). Furthermore, mice infected with the recombinant NP V100I NA D248N WSN virus were unable to clear the virus and produced higher levels of IFNβ than did mice infected with the parental WSN virus ( Fig 3D). Altogether, these data strongly indicate that in contrast to control or singlemutant viruses, the NP V100I NA D248N WSN virus exhibits strong pathogenicity in vivo.

The NP V100I and NA D248N combination enhances the binding of viral RNA by NP
In contrast to parental or single NP V100I or NA D248N mutant WSN viruses, the replication rate of NP V100I NA D248N WSN was strongly increased (Fig 2A). To exclude possible variations at the entry steps of infection, we measured accumulation of the intracellular viral genome at 24 hours relative to the intracellular viral genome at 6 hours postinfection ( Fig 4A). In this case, NP V100I NA D248N WSN again exhibited the highest viral genome accumulation, confirming that this double mutant virus shows enhanced viral production compared to that of the parental or single-mutant virus. These results suggest that NP V100I and NA D248N might function together to increase the overall fitness of infectious influenza viruses.
To address the molecular mechanism of functional synergy between the NP V100I and NA D248N proteins, we first assessed whether physical interactions between the NP and NA proteins are affected by these substitutions. To this end, mammalian expression plasmids carrying NP WSN , NP V100I , NA WSN , or NA D248N were generated separately, and their interactions were assessed in 293T cells. However, no physical interactions between these proteins were observed (S3 Fig). Therefore, we next hypothesized that the function of one protein might be indirectly aided by the other protein in the presence of the designated substitution. NP binds to viral RNA and is a component of the viral ribonucleoprotein (vRNP) complex [16,17], and V100I is located near the RNA-binding domain [18] (S4 Fig). We first examined whether the RNAbinding potential of NP is altered by this substitution using an RNA pulldown assay with biotin conjugated to WSN NS vRNA. Cell lysates from 293T cells transfected with NP WSN or NP V100I expression plasmids were incubated with labeled WSN NS vRNA, and the RNA-bound form of the NP protein was visualized by an immunoblot assay. In contrast to weak NP WSN binding, enhanced vRNA binding to NP V100I was detected (Fig 4B and 4C). Interestingly, vRNA binding by the NP protein, either NP WSN or NP V100I , was further increased when NA D248N was coexpressed (Fig 4D and 4E), indicating that NA D248N is able to aid NP protein-vRNA interactions. Next, an RNA-IP assay was performed in which total lysates of A549 cells infected with various forms of recombinant WSN viruses were incubated with an anti-NP antibody followed by analysis of the bound form of viral RNAs. Although NP V100 protein binding to viral RNA appeared to increase, the results were not significant. However, NP V100I binding to viral RNA significantly increased when was NA D248N also present ( Fig 4F). These data,  To better understand the altered function of NP proteins by substitution or coexpression with NA D248N , we examined viral genome replication because NP has been reported to function in viral genome replication steps along with other PB1/PB2/PA RNA polymerase complex components [16]. 293T cells were cotransfected with NP WSN , NP V100I , NA WSN , or NA D248N along with an eGFP reporter plasmid carrying NS vRNA, and the viral RNA-dependent RNA polymerase (RdRP) activity of NP was assessed by measuring the mean fluorescent intensity of GFP ( Fig 4G). Under all conditions tested, the RdRP activity of the NP protein was not significantly altered by single-amino acid substitutions at the 100 th position or in combination with NA D248N expression.

The NA D248N protein is recruited to the lipid raft compartment
Herein, we provide evidence that vRNA-NP assembly as well as viral production are synergistically enhanced when both NP V100I and NA D248N are present, indicating that the gain of function effect by NA D248N substitution influences the function of NP V100I , ultimately promoting viral production. During assembly of newly synthesized influenza viral proteins and vRNPs, the HA and NA proteins are independently targeted toward lipid rafts [19,20], which is also required for the apical targeting of vRNPs [21]. In transfected 293T cells, a large amount of NA D248N protein was observed in cellular compartments containing lipid raft markers (S5 Fig). Therefore, we hypothesized that intracellular trafficking of the vRNP complex to lipid rafts is affected by NA D248N , which may explain the enhanced viral production of recombinant NP V100I NA D248N WSN. To test this hypothesis, we measured the lipid raft trafficking of NA in recombinant virus-infected A549 cells. Similar to plasmid transfection, more NA protein was observed in the lipid raft compartment of cells infected with NA D248N WSN (Fig 5A). In this case, the input expression of NA was also increased after NA D248N WSN infection. Furthermore, statistically significant enhancement of NA localization in the lipid raft compartment was observed in cells infected with NP V100I NA D248N WSN virus (Fig 5B). These data suggested that NA D248N mutation causes increased NA levels, which could account for the enhanced NA lipid raft association.
To determine the functional relationship between enriched NA at lipid rafts and the enhanced RNA-binding of NP, cells were treated with a myriocin, a serine palmitoyltransferase inhibitor that depletes sphingolipid synthesis and inhibits lipid raft formation, and then NP protein binding to RNA was measured. Depletion of the lipid raft compartment in myriocin-treated cells was confirmed by observing cholesterol reduction at the plasma membrane using FITC-conjugated cholera toxin-B, and we found that NA association with lipid rafts was reduced by myriocin treatment (S6 Fig). Cells were infected with either NP WSN NA WSN WSN or NP V100I NA D248N WSN virus, and NP-mediated viral RNA binding was compared in the lipid raft-depleted condition. In myriocin-treated cells, NP binding to viral RNA was measured by calculating the ratio between the amount of intracellular viral RNA at 24 h.p.i. versus that at 6 h.p.i. Each bar represents three independent experiments. (B and D) A biotin-labeled NS vRNA pulldown assay was performed using 293T cell lysates transfected with the indicated plasmids (48 hours). Nonlabelled NS vRNA was used as a competitor. (C and E) Western blotting data represent one result from three or four repeated independent experiments. (F) RNA immunoprecipitation of NP proteins was performed using A549 cells infected with the indicated WSN influenza A viruses (10 TCID 50 /ml, 0.01 MOI, 24 hours). Relative amounts of NP-bound vRNAs were analyzed by qRT-PCR. (G) 293T cells were transfected with expression plasmids carrying PB2, PB1, and PA along with the designated NP and NA. To measure RdRP activity, equal amounts of the NS-GFP reporter plasmid were cotransfected. The mean fluorescence intensities (MFI) of GFP-expressing cells measured by FACS are shown. (C and D) The blots shown are cropped from different parts of the same membrane after stripping. The dot graph was obtained from five independent experiments, and each individual experiment was normalized to the control. The mean and standard error were analyzed by one-way ANOVA followed by the Newman-Keuls posttest in GraphPad Prism 5. ns, not significant; � p<0.05; and �� p<0.01. significantly reduced with both NP WSN NA WSN WSN and NP V100I NA D248N WSN infection (Fig 5C). In addition to virus infection, reduced NP binding to vRNA was observed in myriocin-treated, expression-plasmid-transfected cells (Fig 5D). Taken together, these data indicate

Discussion
In this report, we demonstrate that functional synergy between the covarying NP V100I and NA D248N mutation pairs of the 2009 pH1N1 IAV explains the enhanced pathogenicity of this influenza virus. A single-amino acid substitution at the 100 th residue of the NP protein increased its vRNA binding potential, which was significantly enhanced when the NA D248N protein was present. The ability of the NA protein to be recruited to the lipid raft compartment, which directly influences viral particle release, was also augmented with a single-amino acid change at the 248 th residue of the NA protein. Therefore, our data indicate that the covarying NP V100I and NA D248N pair functions to boost viral component assembly and virus particle release, which may explain the increased production of influenza virus (Fig 6).
We utilized amino acid sequence information from 2009 pH1N1 IAV isolates to search for covarying protein pairs based on the assumption that proteins that function together have a high probability of coevolution. Although we interpreted high covariation as a sign of a functional relationship, covariation can be affected by many factors in addition to the pressure to coevolve, and geographic distance, evolutionary distance between viruses, or gene lengths can influence our results. We consider two possible scenarios if geographic distance dominates impacts on covariation: 1) covariation among all genes should be very low, as approximately 70% (291/417) of examined viral sequences were obtained from a single country (USA); or 2) the total covariation score should be high because correlation according to distance is very significant, even among viruses from the USA. In general, our covariation analysis revealed relatively low correlation coefficients (between -0.104 and 0.253), except for the NP-NA pair (0.623). The length of a sequence is another potential factor that can affect the probability of random variation. To address this issue, the similarity value was normalized to the total number of sequences. Analysis of the covariation results indicated no correlation with sequence length: the longest genes (PB2, PB1) did not show the highest correlation, and the shortest genes (NS1, NS2, M2) did not show in relatively small correlations among themselves or with other genes (S1 Table). Based on this interpretation, we postulated that the influence of sequence length on covariation was minimal, if at all. Physically interacting proteins coevolve to complement structural changes at the interaction surface. When coevolutionary selection pressure is applied to pairs of interacting proteins, amino acid substitutions in one protein often affect the rates of substitutions in interacting partners. By measuring correlating variability of amino acid sites among proteins, structural as well as phylogenetic coevolution can be assessed [12,14]. Although most molecular coevolution studies have focused on interacting proteins, accumulating evidence indicates that noninteracting protein pairs with similar functions also coevolve. In fact, on a proteome-wide scale, analyses of evolutionary rate covariation (ERC) along with physically and genetically interacting proteins and noninteracting but functionally related proteins have revealed significant ERC [13]. Although biases and limitations are present, diverse analytical tools for measuring amino acid covariation have been developed and applied, leading to breakthrough studies for predicting protein structure and interactomes [12,22] In general, measuring covariation is extremely useful when employing a molecular approach to understand virus pathogenicity. For example, covariation between mutations in the variable site of the V3 loop of the human immunodeficiency virus type 1 envelope protein has been assessed to identify significant covarying mutations with structural or functional relationships [23]. In addition, covariation of eight amino acids in the capsid protein of Hibiscus chlorotic ringspot virus (HCRSV) has been identified as the genetic change responsible for the loss of virulence after serial passaging [24]. The amino acid covariance of the hepatitis C virus (HCV) viral coding region in infected patients has been measured before treatment to link covarying positions with differential responses to therapy [25], and genome-wide analysis of HCV covariance networks has been applied to identify amino acid pairs predicted to contribute to therapy failure as well as hub positions with antiviral target potential. In contrast to specific amino acid site-wise level comparison analysis, we measured the similarity between entire sequences, which distinguishes our approach from other more typical methods. In this study, we examined the amino acid covariance of viral proteins of the 2009 pH1N1 influenza virus to identify genetic factors contributing to the pathogenicity of the 2009 pandemic. A total of 417 whole-genome sequences from independent influenza virus isolates were compared to assess covariance, and covarying pairs of the influenza viral proteins NA and NP were identified. However, our analysis has some potential shortcomings. Among the 417 examined isolates, approximately 70% of the information was obtained from a single country (USA, low geographic distance), which suggests that our data analysis of predominant amino acid sequences at variable residues in NP or NA might be affected by limited sequence information.
We demonstrate that WSN with the NP V100I NA D248N pair was highly pathogenic. However, a low percentage of viruses with the NP V100I NA D248N pair were found among 2009 pH1N1 IAVs, with a naturally occurring frequency of only 4% (Fig 2A). Such an apparent contradiction can also be explained by considering the relatively mild pathogenicity of 2009 pandemic influenza. We postulate that highly pathogenic 2009 pH1N1 IAVs may have failed to adapt among the host population but that rather mildly pathogenic 2009 pH1N1 IAVs were well adapted and dominant. It is likely that the most pathogenic virus eliminated its host too rapidly to propagate, resulting in a small population of infected hosts. Alternatively, the pathogenic NP V100I and NA D248N combination may have caused a phenotype that prevented spread, yet we cannot present a possible mechanism that supports this hypothesis.
Influenza NP is a single-stranded RNA (ssRNA)-binding protein that encapsidates the viral RNA genome to form a ribonucleoprotein particle. NP also interacts with several viral and host proteins to elicit multiple functions during the virus replication cycle of influenza infection [18]. NP participates in viral RNA transcription processes through interactions with the RdRP complex and viral packaging; it also interacts with host factors, such as the human chaperone heat shock protein 40, to modulate PKR activation or inhibit IRF3 activation [26]. Both the sequence comparison and reverse genetics approaches applied to identify substitutions responsible for the pathogenicity observed in the 2009 pH1N1 influenza outbreak have mainly been concentrated on the membrane proteins HA and NA, with a limited number of studies reporting the role of NP. In this study, we observed that the single-amino acid substitution of V to I at the 100 th position near the RNA-binding domain enhanced NP-mediated RNA-binding ability but not RdRP activity. Interestingly, the increased RNA binding, or rather RNP formation, resulted in enhanced virus production, especially when a covarying amino acid substitution was present in the NA protein, affecting lipid raft recruitment.
During the final stage of virus assembly, the packaged RNA-protein complex associated with the intracellular membrane translocates to the apical side of the lipid raft-rich plasma membrane, and this is a critical step for virus release [27]. We observed that a transfected NA (D248N) protein or recombinant virus with an NA (D248N) substitution was recruited to the lipid raft compartment more frequently than was the control protein or virus, respectively. The D248 residue is located in the extracellular domain of the NA protein, and no reports thus far suggest a role for lipid raft trafficking within this region [19,20]. Therefore, we hypothesize that physical interactions between NA and host factors within the lumen of vesicular endosomal organelles, such as the trans-Golgi apparatus, might be disrupted by this amino acid substitution. Interestingly, galectin-1, which functions to sort apical glycoproteins or their apical trafficking, reportedly interacts with influenza virus envelope proteins and controls viral infectivity [28,29]. Thus, investigating whether the identified substitutions in the NA protein alter its interaction with host trafficking proteins is warranted.
Covariation analysis of 2009 pH1N1 influenza sequence information revealed covarying viral proteins and specific amino acid substitutions that synergistically influence the pathogenicity of infection. Although no obvious physical interactions were observed, we experimentally demonstrated that these covarying pairs of genetic factors function together to enhance viral production. Based on these results, we propose that covariation analysis is needed to screen genetic factors that do not interact but instead function together.

Identification of covarying gene pairs
Similarity scores of amino acid sequences were calculated for every pair of isolates for each influenza viral protein, and covariations of viral proteins were analyzed by Pearson's correlation coefficient (ρ). The amino acid sequences of influenza virus strains were aligned and compared using parsing methods in Biopython. The similarity score for a viral protein was calculated as a ratio of matched sequences between two strains among total amino acid sequences. Similarity scores were calculated for all possible combinatory pairs of pandemic strains per influenza protein with general programming using MATLAB.

Similarity strain i; strain j ð Þ ¼ Number of matched residues between strain i and strain j Total number of residues
The degrees of correlated mutations between viral proteins were analyzed by Pearson's correlation coefficients (ρ) of similarity scores of same-strain pairs using the standard function corr in MATLAB. r proteinA;proteinB ¼ Covariance ðsimilarity score of protein A; similarity score of protein BÞ s ðsimilarity of protein AÞ � s ðsimilarity of protein BÞ To identify and quantify degree of variability at each amino acid position in a protein, the RNA sequence was translated into the amino acid sequence and aligned using standard procedures in Biopython; analysis by general programming was performed using MATLAB. In addition, the number of strains that had different amino acids compared to the most frequent amino acid at each position was counted.
The sequence information for the 417 2009 pH1N1 IAVs and the sequence information for the 267 1948 to 2008 seasonal H1N1 IAVs were obtained from the ISED on December 2011.

In vitro virus production and virus infection
Recombinant WSN influenza A virus was generated using an in vitro virus production method [30]. A total of 12 plasmids, pCAGGS WSN PB2, pCAGGS WSN PB1, pCAGGS WSN PA, pCAGGS WSN NP 0/14, pHH21 polI WSN PB2, pHH21 polI WSN PB1, pHH21 polI WSN PA, pHH21 polI WSN NP, pHH21 polI WSN HA, pHH21 polI WSN NA, pHH21 polI WSN NS, and pHH21 polI WSN M (kindly provided by Drs. Kawaoka and Neumann, Madison-Wisconsin Univ., USA), were cotransfected into 293T cells using the calcium phosphate method. For NA or NP substitutions, the pHH21 polI WSN NP or pHH21 polI WSN NA expression plasmids with designated changes were generated. Twelve hours after transfection, cells were washed with PBS and further incubated in DMEM containing 10% FBS for 48 hours. The supernatants were harvested, and the recombinant viruses were isolated by sucrose density gradient (30%/60%) centrifugation. Recombinant viruses were then propagated in chicken embryonated eggs for 48 hours, and the TCID 50 /ml values of the recombinant viruses in A549 cells were calculated.
Recombinant WSN influenza A virus was infected into A549 cells at the indicated titer (TCID 50 /ml). Three hours later, the infected cells were washed with PBS to remove semi-infectious nonbound virus and incubated for the indicated time.

Mouse experiments
BALB/cOlaHsd (BALB/c) mice were obtained from Harlan Olac (Bicester, UK). After being intraperitoneally (i.p.) anesthetized with ketamine (100 mg/kg; Yuhan, Korea) and xylazine hydrochloride (10 mg/kg; Bayer, Belgium) in PBS, the 6 or 7 mice from each group were intranasally (i.n.) injected with WSN influenza A virus (2000 TCID 50 /ml, 100μl). This experiment was carried out for 15 days, and mice body weight and mortality were checked and monitored every 2 days from the 5th day after the initial virus infection (n = 27). To assess mortality, mice that lost more than 35% of their initial body weight were immediately euthanized by CO 2 inhalation or cervical dislocation (n = 14). At the end of the experiment, the mice that survived were also euthanized the same way (n = 13). For preparation of the lungs of infected mice, lungs were extracted after PBS perfusion and kept frozen at -80˚C. The isolated lungs were homogenized in PBS, and the supernatants were collected by centrifugation at 13,200 rpm for 30 min at 4˚C. To measure titers, A549 cells were treated with homogenate supernatants from the virus-infected lungs. For mRNA expression, total RNA was prepared from 200 μl of tissue homogenate.

Ethics statement
The

Crystal violet assay & apoptosis analysis
A549 cells in a 24-well plate (2 × 10 4 cells/well) were infected with recombinant WSN IAV. After the indicated time, the cells were washed with PBS and fixed with 4% paraformaldehyde (Sigma) for 10 min at room temperature (RT). After staining with 2% crystal violet (Sigma) (dissolved in 50% ethanol), the dye crystals were dissolved in a 1% SDS solution at 37˚C for 10 min and analyzed using an EZ Read 400 instrument (Biochrom, UK).
For measuring apoptosis, A549 cells (5 ×10 5 ) were infected with the indicated recombinant WSN IAV for 12 or 36 hours. The cells were stained with Annexin V (BD bioscience) and propidium iodide (Sigma) in 10 mM HEPES (pH 7.4), 140 mM NaCl, and 2.5 mM CaCl 2 for 20 min at RT and analyzed using a FACS caliber flow cytometer (BD Biosciences).

Total RNA isolation and real-time quantitative PCR
Total RNA was extracted using RNAiso plus (Takara Bio), and 1 μg was reverse transcribed with ImProm-II Reverse Transcriptase (Promega) using random primers (Promega). PCR analysis of the cDNA was performed using a StepOne Plus Real-Time PCR machine (Applied Biosystems). The 18s rRNA or hGAPDH gene was used as an internal control.

RNA-IP and RNA pulldown assay
RNA-IP was performed as previously described [31]. Briefly, A549 cells (2 × 10 6 ) were lysed in Polysome Lysis Buffer (PLB) (100 mM KCl, 5 mM MgCl 2 , 10 mM HEPES (pH 7.0), and 0.5% Nonidet P-40) supplemented with protease inhibitors and 100 U/ml RNAse inhibitor. After centrifugation at 13,000 rpm, the precleared supernatants were incubated with the anti-NP antibody and protein A/G beads for 6 hours at 4˚C. After 6 rounds of washing with PLB, the RNA in the bound fractions was eluted using RNAiso plus.

Preparation of lipid raft-enriched membrane fractions
Lipid raft-enriched fractions were obtained using a modified TX-100 extraction method [32]. Cells were suspended in prechilled TNE buffer (50 mM Tris-HCl (pH 7.5), 1 mM EDTA, and 150 mM NaCl) with protease inhibitors. After brief sonication, the samples were added to TNE buffer containing 1% TX-100 at 4˚C for 30 minutes. The samples were then centrifuged at 13,200 rpm (Eppendorf, 5415R) for 30 minutes to separate the soluble and insoluble fractions.

Statistics
One-way ANOVA followed by the Newman-Keuls posttest was used to compare differences between three or more independent experimental groups in all cases. A p value <0.05 was considered statistically significant, and individual biological experiments consisted of individual cell culture and virus infection. In general, three or more repeated individual experiments were used for each condition, which was sufficient to determine statistically significant differences.