Double-Stranded RNA Binding of Influenza B Virus Nonstructural NS1 Protein Inhibits Protein Kinase R but Is Not Essential To Antagonize Production of Alpha/Beta Interferon

Abstract

Expression of alpha/beta interferon (IFN-α/β) in virus-infected vertebrate cells is a key event in the establishment of a sustained antiviral response, which is triggered by double-stranded RNA (dsRNA) produced during viral replication. These antiviral cytokines initiate the expression of cellular proteins with activities that limit the replication and spread of the invading viruses. Within this response, the dsRNA-dependent protein kinase R (PKR) that is expressed at constitutive levels and upregulated by IFN-α/β acts as an important antiviral effector that can block the cellular translational machinery. We previously demonstrated that efficient replication of influenza B virus depends on the viral dsRNA-binding NS1 protein that inhibits the transcriptional activation of IFN-α/β genes. Here we tested the postulate that the viral NS1 protein counteracts antiviral responses through sequestering intracellular dsRNA by analyzing a collection of recombinant influenza B viruses. As expected, viruses expressing dsRNA-binding-defective NS1 proteins were strongly attenuated for replication in IFN-competent hosts. Interestingly, these virus mutants failed to prevent activation of PKR but could effectively limit IFN induction. Conversely, a mutant virus expressing the N-terminal dsRNA-binding domain of NS1 prevented PKR activation, but not IFN induction, suggesting an important role for the NS1 C-terminal part in silencing the activation route of IFN-α/β genes. Thus, our findings indicate an unexpected mechanistic dichotomy of the influenza B virus NS1 protein in the suppression of antiviral responses, which involves at least one activity that is largely separable from dsRNA binding.

Influenza A and B viruses are globally distributed pathogens that cause an acute severe respiratory disease. Despite vaccination campaigns and the availability of antiviral therapeutics, annual epidemics of influenza claim the lives of an estimated 10,000 individuals on average in Germany alone (70). The viruses belong to the Orthomyxoviridae family and are characterized by a segmented genome that consists of eight single-stranded RNAs of negative polarity (26). Both influenza virus types share many features with respect to replication strategy and protein functions. However, there are differences in the coding strategies of two gene segments (26) as well as expression of one type-specific polypeptide as represented by the type B-specific NB protein (50) and the proapoptotic PB1-F2 protein encoded by most influenza A viruses (5). Another important biological distinction is indicated by a narrow host spectrum for influenza B viruses that is largely restricted to humans, whereas influenza A viruses have numerous host species, including birds and a variety of other mammals, such as horses and pigs (65).

Efficient replication of influenza viruses and most other viruses necessitates suppression of antiviral responses mediated by the alpha/beta interferon (IFN-α/β) system, an important part of the innate immune responses of vertebrates (13, 15). Induction of the IFN-α/β system is orchestrated, in which the first wave of IFN-β leads to expression of antiviral proteins and the transcription factor IRF-7 (interferon regulatory factor 7) that in turn induces a secondary wave of IFN-α and IFN-β (21, 36, 56). The initial transcriptional induction of IFN-β genes is triggered by double-stranded RNA (dsRNA) molecules, a by-product of viral replication, that are recognized by the recently described RNA helicases RIG-I and MDA-5, both containing two N-terminal CARD-like domains and a dsRNA-binding C-terminal helicase domain (1, 35, 53, 67). RIG-I interacts with the newly identified MAVS protein (also known as IPS-1, VISA, and Cardif) that also contains a CARD-like domain and is localized in the outer mitochondrial membrane (24, 38, 48, 66). This interaction mediates the activation of the kinases TBK-1 (Traf family member-associated NF-κB activator-binding kinase 1) and IKKɛ (IκB kinase ɛ) that stimulate the latent key transcription factor IRF-3 (11, 37, 49). Phosphorylated IRF-3 forms a dimer and accumulates in the nucleus (31, 47). The coordinated assembly of IRF-3 and the nuclear coactivator CBP/p300 together with the transcription factors NF-κB and ATF2/c-Jun on the IFN-β gene promoter induces its transcription (10, 22, 25, 44, 59, 62, 63, 68). Secreted IFN-β binds to the IFN-α/β receptor, which leads to the formation of the heterotrimeric transcription factor ISGF-3 (interferon-stimulated gene factor 3) via signaling through the JAK/STAT pathway (45). ISGF-3 mediates the transcriptional upregulation of more than 100 IFN-stimulated genes, including the Mx proteins, 2′-5′ oligoadenylate synthetases, and the kinase PKR (protein kinase R) (46). PKR is activated by dsRNA and limits viral propagation through blocking cellular protein synthesis by sustained phosphorylation of the initiation factor eIF2α (46). Additionally, IFN-α/β connect innate and adaptive immune responses, as they also modulate the differentiation of dendritic cells, cross-presentation, and cross-priming, expression of costimulatory factors and major histocompatibility complex molecules and activation of NK cells (28, 29, 40).

The strategies applied by different virus families to counteract the antiviral response range from inhibition of the transcriptional activation of IFN genes to blocking the JAK/STAT pathway or direct targeting of antiviral proteins (13, 15, 52). The influenza A viruses express a nonstructural protein of 202 to 237 amino acids (aa) (A/NS1 protein) that binds single- and double-stranded RNA, inhibits the polyadenylation and splicing of cellular pre-mRNAs, and enhances translation (2, 6, 12, 16, 18, 32, 43, 60, 64). The A/NS1 protein also antagonizes IFN-β expression and is necessary for efficient viral replication in IFN-competent hosts (14, 41). This effect is mediated at least in part by its ability to inhibit activation of the transcription factors IRF-3/IRF-7, NF-κB, and ATF2/c-Jun (34, 51, 54, 61).

Influenza B virus expresses a structurally related NS1 protein of 281 aa (B/NS1) that has less than 20% sequence identity to the A/NS1 protein. B/NS1 does not inhibit pre-mRNA splicing (60), but it has the distinctive abilities of being able to bind to the interferon-stimulated gene 15 (ISG15) product and to inhibit its conjugation to cellular targets (69). Despite these differences, it binds to the same RNA targets as the A/NS1 protein through an N-terminal RNA-binding domain (60). We have shown previously that the activities of the B/NS1 protein are essential for efficient viral replication, which involve inhibition of both IRF-3 induction and IFN-β expression (7, 9). Further studies have indicated that the NS1 proteins of both influenza virus types also act as inhibitors of PKR. However, whereas several analyses showed that the A/NS1 protein antagonizes PKR activation in vivo (3, 17, 30), there is only indirect evidence for this activity for the B/NS1 protein, as it prevented a block of translation induced by dsRNA in reticulocyte lysate (60). In addition, it has been postulated for the A/NS1 protein that dsRNA sequestration is the mechanistic basis for the inhibition of IFN-β induction and PKR activation (17, 33, 61). In contrast, there is comparatively little knowledge about the impact of dsRNA sequestration for the influenza B virus life cycle.

In this report we took a reverse genetic approach to study the role of dsRNA binding by the influenza B virus NS1 protein in promoting viral replication and the inhibition of antiviral responses. To this end, we generated a set of isogenic influenza B viruses with mutations that either affected NS1 dsRNA binding or introduced a large C-terminal truncation. Analysis of these viral mutants demonstrated that inhibition of IFN-α/β production is largely independent of the dsRNA-binding activity of NS1 but critically relies on the presence of the C-terminal part of the protein. However, dsRNA binding was necessary to inhibit PKR activation and, hence, enable efficient viral replication. Thus, our findings reveal an unexpected mechanistic dichotomy of the NS1 protein to tackle antiviral responses induced by dsRNA.

MATERIALS AND METHODS

Cells and viruses.

293T and A549 cells were grown in Dulbecco's modified Eagle medium supplemented with 10% fetal calf serum, 2 mM l-glutamine, and antibiotics. Madin-Darby canine kidney (MDCK) type II cells were grown in minimal essential medium (MEM) supplemented with the same additives. The MDCK-C3 cell line that contains a stable integrate of a firefly luciferase gene controlled by the human IFN-β promoter has been described elsewhere (7). All cells were maintained at 37°C and 5% CO₂. Stocks of the recombinant influenza B/Lee wild-type virus were grown in the allantoic cavities of 11-day-old embryonated chicken eggs for 3 days at 33°C. The influenza B virus mutants ΔNS1, NS1 33/38, NS1 47/50, NS1 52/53/54, NS1 58/60/64, NS1 77/78, NS1 83/86, and NS1-104 were amplified in 6-day-old embryonated chicken eggs. The ΔNS1 virus was purified and further concentrated as described previously (7). To analyze viral replication, confluent MDCK cell cultures were infected at the indicated multiplicity of infection (MOI) and incubated for 3 days at 33°C in MEM containing 0.2% bovine albumin and 1 μg/ml tosylsulfonyl phenylalanyl chloromethyl ketone (TPCK)-treated trypsin. To analyze replication in 6- and 11-day-old embryonated chicken eggs, the eggs were inoculated with 10³ infectious units of the respective virus and incubated for 3 days at 33°C. Virus titers were determined by indirect immunofluorescence staining of infected MDCK cells with the NP-specific monoclonal antibody BM3149 (DBC Biermann, Bad Nauheim, Germany) as described previously (7). Titers were correspondingly expressed as fluorescence-forming units (ffu) per milliliter.

Construction of plasmids.

The plasmids pHW-Lee-NS-33/38, pHW-Lee-NS-47/50, pHW-Lee-NS-52/53/54, pHW-Lee-NS-58/60/64, pHW-Lee-NS-77/78, and pHW-Lee-NS-83/86 are derivatives of pHW-Lee-NS-XhoI (7) and were constructed with the QuikChange mutagenesis kit (Stratagene) by creating alanine exchange mutations in the NS1 reading frame at the indicated amino acid positions and novel recognition sites for the endonucleases BseMI, HindIII, NheI, PstI, SalI, XhoI, and SapI. Similarly, pHW-Lee-NS1-104 was prepared by the introduction of two consecutive translational stops downstream of amino acid codon 104. The integrities of the constructs were confirmed by DNA cycle sequencing by using an ABI Prism 3100 genetic analyzer (Applied Biosystems).

Transfection-mediated recovery of recombinant influenza B virus and RT-PCR analysis.

To generate influenza B viruses with recombinant NS segments, we transfected 10⁶ 293T cells with the plasmids pHW-Lee-PB2, pHW-Lee-PB1, pHW-Lee-PA, pHW-Lee-HA, pHW-Lee-NP, pHW-Lee-NA, pHW-Lee-M, and a pHW-Lee-NS construct (0.5 μg each) using Lipofectamine 2000 (Invitrogen) as previously described (7). At 72 h after transfection, the cell supernatant was inoculated into the allantoic cavities of 6-day-old chicken eggs to grow stocks of recombinant virus. To confirm the recovery of recombinant influenza B virus, the viral RNA was extracted from allantoic fluid by using the QIAampMinElute virus spin kit (QIAGEN) followed by reverse transcription-PCR (RT-PCR) amplification of the NS segment with the OneStep RT-PCR kit (QIAGEN). The presence of the introduced mutations and the absence of unexpected nucleotide alterations in the respective NS genes was shown by restriction analysis and cycle sequencing of amplified RT-PCR products.

Metabolic labeling of infected cells.

Confluent 5 × 10⁵ MDCK cells seeded in 22-mm dishes were either mock infected or infected with wild-type virus or with ΔNS1 or NS1 mutant virus at an MOI of 10. Cells were incubated in MEM containing 0.2% bovine albumin at 33°C for 7 h and were subsequently labeled for 1 h with 25 μCi of [³⁵S]methionine (Amersham, Braunschweig, Germany) diluted in methionine-free MEM. The cells were washed with ice-cold phosphate-buffered saline and lysed in 100 μl of radioimmunoprecipitation assay buffer (64). The extracts were subjected to 12.5% sodium dodecyl sulfate (SDS)-polyacrylamide gel electrophoresis, and the radiolabeled proteins were visualized by autoradiography.

Immunoprecipitation and immunoblot analysis.

To analyze the IRF-3/CBP interaction, 1 × 10⁷ A549 cells were mock infected or infected with wild-type virus or with ΔNS1 or NS1 mutant virus at an MOI of 1. Cell extracts were prepared 10 h postinfection (p.i.) in 1 ml lysis buffer (150 mM Tris-HCl, pH 7.5, 1% Nonidet P-40, 150 mM NaCl, 1 mM EDTA, 10 mM β-glycerophosphate, 1 mM Na₃VO₄, 1 mM Pefabloc) and cleared by centrifugation. The lysates were incubated at 4°C overnight with 1 μg IRF-3-specific monoclonal antibody SL12.1 (BD Pharmingen). Immunocomplexes were collected on protein G-agarose beads (Roche, Mannheim, Germany) and washed, and the precipitated proteins were dissolved in SDS sample buffer (17). The precipitated proteins were analyzed by immunoblotting with suitable primary (rabbit anti-IRF-3 [Zymed] and rabbit anti-CBP [A22; Santa Cruz Biotechnology]) and secondary horseradish peroxidase-conjugated antibodies by using an enhanced chemiluminescence protocol (Pierce). To analyze phosphorylation of PKR and eIF2α, lysates of infected A549 cells were prepared as described above and subjected to SDS-polyacrylamide gel electrophoresis. Phospho-PKR and total PKR were detected by immunoblot analysis with primary rabbit anti-phospho-PKR (Thr446) antibody (Cell Signaling) or mouse anti-PKR antibody 71/10 (Ribogene) and suitable secondary horseradish peroxidase-conjugated goat anti-rabbit or rabbit anti-mouse antibodies, respectively. Phospho-eIF2α and total eIF2α were detected with primary rabbit anti-phospho-eIF2α (Ser51) and anti-eIF2α antibodies (Cell Signaling). Viral proteins were stained on blots with an NP-specific mouse monoclonal antibody (DPC Biermann) or rabbit anti-B/NS1 serum, respectively.

Analysis of IFN-β- and IRF-3-dependent promoter activity and quantification of human IFN-β and IFN-α by ELISAs.

IFN-β promoter activity was analyzed in infected MDCK-C3 cells as described elsewhere (7). IRF-3-dependent gene expression was assessed by cotransfecting MDCK cells in 12-well plates with 50 ng p4x(PRD)I/III-Luc reporter plasmid and 5 ng pRL-TK by using Lipofectamine 2000 (Invitrogen) as described previously (58). Cells were analyzed for reporter activation after mock infection or infection (MOI = 1) for 14 h with wild-type, ΔNS1, or NS1 mutant virus using the dual luciferase assay (Promega). For quantification of secreted IFN-α/β, A549 cells were either mock infected or infected with wild-type, ΔNS1, or NS1 mutant virus at an MOI of 1. At 18 h p.i., 1 ml of supernatant was removed, and the levels of human IFN-β and human IFN-α were quantitated by enzyme-linked immunosorbent assays (ELISAs) (human IFN-β ELISA kit; FUJIREBIO Inc., Tokyo, Japan; human IFN-α ELISA kit; PBL Biomedical Laboratories).

IRF-3 translocation assay.

Nuclear translocation of IRF-3 was examined with a transfected enhanced green fluorescent protein (EGFP) fusion construct as described previously (9). Briefly, MDCK cells were transfected with pEGFP-C1-hIRF-3 and were either mock infected or infected with wild-type, ΔNS1, or NS1 mutant virus at an MOI of 1. The cells were fixed and stained for the viral NP protein at 8 h p.i. and analyzed by immunofluorescence microscopy. The percentage of cells with nuclear IRF-3 was calculated by counting 100 cells with clear positive NP staining in several different fields.

RESULTS

Generation of recombinant influenza B viruses with NS1 proteins that differ in their dsRNA-binding activity.

The N-terminal part (aa 1 to 93) of the influenza B virus NS1 protein binds to dsRNA with no apparent sequence specificity (60). We have shown previously that the dsRNA-binding activity of NS1 depends on several conserved clusters of basic amino acids within this domain: mutation of single clusters of basic amino acids to alanine residues resulted in NS1 proteins that had either normal, weakened, or no dsRNA-binding activity (Fig. 1A) (9). To study the impact of dsRNA binding of the NS1 protein on various aspects of the viral life cycle, we generated recombinant influenza B viruses expressing NS1 proteins with the previously characterized alanine replacements of basic amino acid clusters at amino acid positions 33 and 38 (33/38), 47/50, 52/53/54, 58/60/64, 77/78, and 83/86. The NS segments of the recombinant viruses carried in addition an engineered genetic tag site that allowed us to distinguish the corresponding cDNAs from the wild-type virus cDNA by cleavage with restriction endonucleases (Fig. 1B). The presence of the expected mutations in the NS segments was also verified by sequence analysis. Metabolic labeling of MDCK cells with [³⁵S]methionine during single-cycle infections showed that the amounts of viral proteins synthesized by the wild-type and mutant viruses were comparable, suggesting that none of the mutations caused a major defect in viral gene expression (Fig. 1C).

FIG. 1.

Generation of recombinant influenza B viruses with mutations in the dsRNA-binding domain of the NS1 protein. (A) The diagram indicates the positions of basic amino acid residues within the dsRNA-binding domain of the NS1 protein. NS1 proteins with alanine exchange mutations at the indicated positions had been shown to have strong (+), weak (+/−), or no (−) dsRNA-binding activity (9). The respective mutations were introduced by directed mutagenesis into the portion of the pHW-NS-XhoI plasmid that encodes the NS gene. Each of these constructs was used in an eight-plasmid system for recovery of recombinant influenza B mutant virus expressing an NS1 protein with alanine residues at the indicated positions. (B) RT-PCR analysis of the viral NS segments. After the extraction of viral RNAs of the wild-type (WT) virus and the NS1 mutant viruses, the NS segments were reverse transcribed and amplified by PCR. The amplified NS cDNAs were distinguished by digestion at the introduced HindIII, PstI, NheI, PstI, or BseMI site. The positions of size markers (in kilobases) are indicated on the left. (C) Metabolic labeling of proteins in virus-infected MDCK cells. Cells were either mock infected or infected with the recombinant wild-type or NS1 mutant viruses at an MOI of 10 and incubated at 33°C. At 7 h p.i., cells were metabolically labeled with [³⁵S]methionine for 1 h. Cell extracts were prepared and analyzed by SDS-polyacrylamide gel electrophoresis and autoradiography. The positions of the molecular mass markers (in kilodaltons) and the NP, NS1, and M1 proteins are indicated on the left and right, respectively.

dsRNA binding by the NS1 protein is not essential to inhibit induction of IRF-3 and production of IFN-α/β.

Critical steps for the transcriptional activation of the IFN-β gene include the nuclear accumulation of IRF-3 and the recruitment of the coactivator CBP to the IFN-β promoter. We have recently demonstrated that the NS1 protein of influenza B virus antagonizes activation of IRF-3 (9). Consequently, Fig. 2 shows that complexes of IRF-3 and CBP can be detected by coimmunoprecipitation in A549 cells after infection with the previously described influenza B virus in which the NS1 gene had been deleted (ΔNS1) (7) at 9 and 12 h p.i. In contrast, such complexes were not detected in wild-type virus-infected cells. Furthermore, we observed only in ΔNS1-infected cells the appearance of slowly migrating forms of IRF-3 and partial degradation of the transcription factor, which are known correlates of its activation (31).

FIG. 2.

Activation of IRF-3 in infected cells. (A) A549 cells were either mock infected or infected with wild-type (WT) virus or ΔNS1 virus at an MOI of 1. Cell extracts were prepared 6, 9, and 12 h p.i. and subjected to coimmunoprecipitation with a monoclonal IRF-3 antibody. The immunocomplexes were analyzed by immunoblotting using CBP-specific and IRF-3-specific antibodies. (B) MDCK cells were transfected with IRF-3 reporter plasmid and pRL-TK-luc for normalization. Twenty-four hours posttransfection, cells were mock infected or infected with wild-type (WT), ΔNS1, or NS1 mutant virus at an MOI of 1. Reporter activities were determined by the dual luciferase assay and are presented as the change in activation (n-fold) compared to the luciferase value of mock-infected cells. The graph shows average values of three experiments conducted in duplicate. Error bars indicate the standard deviations. (C) A549 cells were either mock infected or infected with wild-type, ΔNS1, or NS1 mutant viruses at an MOI of 1. Cell extracts were prepared 10 h p.i. and immunoprecipitated with an IRF-3-specific antibody followed by immunoblot analysis for CBP and IRF-3. Cell extracts were further analyzed for the presence of the viral NS1 protein. (D) MDCK cells expressing an EGFP-IRF-3 fusion protein were infected with wild-type, ΔNS1, or NS1 mutant viruses at an MOI of 1 and stained for viral NP protein at 8 h p.i. The percentage of cells with nuclear IRF-3 was calculated by counting cells with a clear NP signal and is indicated by the gray bars.

To study the influence of the dsRNA-binding capacity of the NS1 protein on its ability to inhibit the activation of IRF-3, we analyzed three different facets of this process in cells infected with the NS1 mutant viruses. Remarkably, this analysis showed that regardless of the dsRNA-binding capacity of their NS1 proteins, all mutants activated a purely IRF-3-driven reporter gene considerably less than the ΔNS1 virus did and were only slightly more active than the wild-type virus was (Fig. 2B). Consistent with this finding, the coimmunoprecipitation assay revealed that all of the mutant viruses induced IRF-3/CBP complexes only weakly in comparison to the ΔNS1 virus (Fig. 2C). Nearly identical amounts of IRF-3 were precipitated from the lysates, and infection efficiencies were similar as demonstrated by immunoblotting with NS1-specific antiserum (Fig. 2C). Remarkably, we observed that nuclear migration of EGFP-IRF-3 occurred only in a minority of cells infected with the wild-type virus or mutant viruses NS1 52/53/54 and 83/86 expressing dsRNA-binding NS1 proteins (Fig. 2D). In contrast, we detected nuclear EGFP-IRF-3 in 80% or more of the cells infected with ΔNS1 or the viruses expressing NS1 with inactivated dsRNA-binding activity (NS1-47/50, −58/60/64, −77/78) or diminished dsRNA-binding activity (NS1-33/38). However, it was recently reported that nuclear accumulation of IRF-3 is necessary but not sufficient for its biological activity in gene expression (52). Thus, these findings show that in essence all our NS1 mutant viruses are capable of blocking IRF-3-mediated antiviral responses.

We next tested the NS1 mutant viruses for their capacity to inhibit induction of IFN-α and -β genes. A reporter assay showed that compared to the ΔNS1 virus, the constructed mutants were only weak activators of the IFN-β promoter, irrespective of the dsRNA-binding capacities of their NS1 proteins (Fig. 3A). However, we noted a slightly stronger reporter activation by the basic amino acid cluster mutants than by the wild-type virus. The low level in promoter activation correlated directly with the cytokine expression level, as A549 epithelial cells infected with the NS1 mutant viruses secreted only minor levels of IFN-β and IFN-α as detected by ELISAs (Fig. 3B and C). These results demonstrate that dsRNA binding is not decisive for the IFN-antagonistic function of the NS1 protein.

FIG. 3.

dsRNA binding by the NS1 protein is not decisive for the inhibition of IFN-α/β induction. (A) MDCK-C3 cells that contain a stable firefly luciferase reporter gene under the control of the IFN-β promoter were infected with recombinant wild-type (WT), ΔNS1, or the NS1 mutant viruses at an MOI of 1. Luciferase activity was determined 8 h p.i. and is presented as the change in activation (n-fold) compared to the value for mock-infected cells. The graph shows average values of a typical experiment conducted in duplicate, which has been repeated five times. Error bars indicate the standard deviations. (B and C) A549 cells were mock infected or infected with wild-type, ΔNS1, or NS1 mutant viruses at an MOI of 1. Eighteen hours p.i., the supernatant was removed, and IFN-β (B) and IFN-α (C) levels were quantitated by ELISAs. The graphs show average values of three experiments conducted in duplicate. Error bars indicate the standard deviations. hu, human.

Loss of the dsRNA-binding capacity of the NS1 protein severely attenuates viral replication in IFN-competent hosts.

The finding that all constructed NS1 mutant viruses inhibited production of antiviral IFN-α/β raised the initial assumption that those mutants would replicate well in IFN-competent hosts. Therefore, we analyzed propagation of the NS1 mutant viruses in 6- and 11-day-old embryonated chicken eggs and MDCK cells. Due to their immature type I IFN system, 6-day-old embryonated chicken eggs are known to support increased replication of viruses that lack an IFN-antagonistic protein, such as NS1-deleted type A and B influenza viruses (7). In this host (6-day-old eggs), all NS1 mutant viruses replicated to titers similar to that of the wild-type virus, with the values ranging from 10⁷ to 10⁸ ffu/ml (Fig. 4). Interestingly, in 11-day-old eggs that have matured IFN inducibility, the dsRNA-binding-defective mutant viruses NS1 47/50, NS1 58/60/64, and NS1 77/78 were strongly attenuated for replication by 3 to 4 orders of magnitude, with maximal titers ranging from 6.1 × 10³ to 2.2 ×10⁴ ffu/ml. The virus expressing the NS1 33/38 mutant protein that has reduced dsRNA-binding activity showed an intermediate phenotype (1.5 × 10⁵ ffu/ml). In contrast, the mutants NS1 52/53/54 and NS1 83/86 that had retained full or partial dsRNA-binding activities replicated to wild-type virus titers in 11-day-old chicken eggs (Fig. 4). In addition, these recombinants displayed very similar relative replication characteristics when analyzed for multicyclic growth in MDCK cells that also have a functional IFN system (Table 1). These growth phenotypes led to the unexpected conclusion that efficient replication in IFN-competent hosts correlates with the dsRNA-binding capacity of the NS1 protein of the respective virus, but not with the capability to suppress expression of IFN-α/β genes.

FIG. 4.

Replication of recombinant influenza B viruses in embryonated chicken eggs. Six- and eleven-day-old embryonated chicken eggs were inoculated with 1,000 infectious units of wild-type (WT), ΔNS1, or NS1 mutant virus and incubated for 72 h at 33°C. Virus titers were determined as described in Materials and Methods. The indicated values represent the averages of at least three independent experiments. Error bars indicate the standard deviations.

TABLE 1.

Replication in MDCK cells^a

Expt no. (MOI)	Virus titer (ffu/ml)
	WTb	ΔNS1	NS1 33/38	NS1 47/50	NS1 52/53/54	NS1 58/60/64	NS1 77/78	NS1 83/86
Expt 1 (0.01)	1.2 × 107	<1.7 × 101	3.3 × 102	<8.3 × 101	3.5 × 106	8.3 × 102	<8.3 × 101	4.8 × 106
Expt 2 (0.01)	9.1 × 106	<1.7 × 101	1.7 × 102	<8.3 × 101	6.2 × 106	1.7 × 102	<8.3 × 101	2.0 × 106
Expt 3 (0.001)	3.1 × 107	NAc	<8.3 × 101	<8.3 × 101	1.4 × 106	8.3 × 101	<8.3 × 101	2.6 × 106
Expt 4 (0.001)	4.1 × 107	NA	<8.3 × 101	<8.3 × 101	2.7 × 106	<8.3 × 101	<8.3 × 101	2.7 × 106

^aConfluent monolayer cultures were infected at the indicated MOI and incubated for 3 days at 33°C. Virus titers were determined as described in Materials and Methods.

^bWT, wild type.

^cNA, not analyzed.

dsRNA binding by the NS1 protein is necessary to inhibit activation of PKR by influenza B virus.

The selective strong attenuation of the dsRNA-binding-defective virus mutants pointed to an important function of the NS1 protein related to dsRNA binding that is different from suppressing IFN induction. PKR is a well-known dsRNA-activated kinase with antiviral activity. After autophosphorylation, it phosphorylates the eukaryotic translation initiation factor eIF2α at Ser51, thereby blocking translation of cellular and viral mRNAs (46). To analyze whether there is a correlation between the dsRNA-binding capacity of the NS1 protein and PKR activation, we infected A549 cells with wild-type and mutant influenza B viruses. Analysis of the cell lysates by immunoblotting revealed that the wild-type virus and the NS1 52/53/54 and NS1 83/86 mutant viruses with retained dsRNA binding inhibited activation of PKR, whereas infection of the cells with the ΔNS1 virus and NS1 33/38, NS1 47/50, NS1 58/60/64, and NS1 77/78 mutant viruses induced phosphorylation of PKR at Thr446, which indicates PKR activation (Fig. 5A). The detection of Ser51-phosphorylated eIF2α confirmed the activity status of PKR after infection with the dsRNA-binding-defective virus mutants (Fig. 5C). Interestingly, the ΔNS1 virus stimulated PKR phosphorylation but had no effect on eIF2α phosphorylation (see Discussion). The total amounts of PKR and eIF2α were not significantly altered by any of the viruses (Fig. 5B and D), and their infection efficiencies were comparable, as demonstrated by immunoblotting with NP- and NS1-specific antibodies (Fig. 5E and F). These results establish a direct correlation between the dsRNA-binding capacity of the NS1 protein and the inhibition of PKR activation by the respective virus. Thus, the strong attenuation of the dsRNA-binding-deficient mutant viruses may at least partly be caused by the antiviral effect of PKR, suggesting that the wild-type NS1 protein enables efficient replication by sequestering dsRNA and thereby inhibiting activation of PKR.

FIG. 5.

The dsRNA-binding activity of the NS1 protein is necessary to inhibit activation of PKR. A549 cells were either mock infected or infected with wild-type (WT), ΔNS1, or NS1 mutant viruses at an MOI of 1. Cell extracts were prepared 10 h p.i. and were analyzed by immunoblotting with antibodies specific for phospho-PKR (P-PKR) (A), total PKR (B), phospho-eIF2α (P-eIF2α) (C), total eIF2α (D), viral NP protein (E), and viral NS1 protein (F).

The C-terminal domain of the NS1 protein is required for inhibition of IFN-α/β induction but is dispensable to prevent PKR activation.

We recently demonstrated that the ectopic expression of either the N- or C-terminal domain of the B/NS1 protein alone inhibited activation of the IFN-β promoter by Sendai virus (9). However, our finding that even dsRNA-binding-deficient NS1 proteins with mutations in the N-terminal domain had largely retained IFN-antagonistic activity in infected cells, suggested that the C-terminal domain of the NS1 protein plays a predominant role in this process. To test this hypothesis, we generated another influenza B virus that via the introduction of two stop codons expressed a truncated NS1 protein of 104 amino acids (NS1-104) that represented basically the dsRNA-binding domain and lacked the C-terminal amino acids 105 to 281. The NS segment of the NS1-104 virus carried an engineered genetic tag site, rendering the corresponding cDNA of the rescued mutant virus susceptible to cleavage by the restriction endonuclease EcoRV (Fig. 6A). A metabolic labeling analysis also confirmed that a protein with the molecular mass expected for the shortened NS1 protein was expressed in NS1-104 virus-infected cells (Fig. 6B).

FIG. 6.

Generation of a recombinant influenza B virus with a C-terminally truncated NS1 protein (NS1-104). (A) Within the plasmid pHW-Lee-NS-XhoI, the amino acid codons 105 and 106 of the NS1 open reading frame were mutated to TAG and TGA, generating two translational stop codons and introducing a genetic tag site. This construct was used along with the other seven plasmids encoding the residual viral gene segments for recovery of the recombinant influenza B NS1-104 virus. After the extraction of viral RNAs from the wild-type (WT) virus or the NS1-104 virus, the NS segments were reverse transcribed (RT) (+) and amplified by PCR. The obtained NS cDNAs were distinguished by restriction analysis of the EcoRV tag site. The positions of size markers (in kilobases) are indicated on the left. (B) Metabolic labeling of proteins in virus-infected MDCK cells. Cells were either mock infected or infected with the recombinant wild-type (WT) or NS1-104 virus at an MOI of 10 and incubated at 33°C. Cells were metabolically labeled at 7 h p.i. with [³⁵S]methionine for 1 h. Cell extracts were prepared and analyzed by SDS-polyacrylamide gel electrophoresis and autoradiography. The positions of the molecular mass markers (in kilodaltons) and of the NP, NS1, NS1-104, and M1 proteins are indicated on the left and right, respectively.

In an IFN-β reporter gene assay, the NS1-104 virus activated the IFN-β promoter even slightly more than the ΔNS1 virus did (Fig. 7A). This was also reflected by a pronounced activation of purely IRF-3-dependent gene expression (Fig. 7B), and the NS1-104 mutant virus also caused a more than 10-fold higher percentage of nuclear accumulation of EGFP-IRF-3 compared to that caused by the wild-type virus (data not shown). Furthermore, cytokine ELISAs revealed that A549 cells infected with the NS1-104 virus secreted about twice the amounts of IFN-β and IFN-α compared to cells infected with ΔNS1 virus (Fig. 7C and D). Thus, the loss of the C-terminal amino acids strongly diminishes the IFN-antagonistic capacity of the NS1 protein. Notwithstanding the massive IFN-α/β induction, the NS1-104 virus replicated surprisingly well in 6- and 11-day-old embryonated chicken eggs, reaching titers of 1 × 10⁸ and 3 × 10⁷ ffu/ml, respectively, that were close to wild-type virus titers (Fig. 7E). In MDCK cells, the multicyclic replication of the mutant was affected to some extent, as the titers of the NS1-104 virus were 1 to 2 log units lower than those of the wild-type virus (Table 2). Thus, despite the strong IFN induction by the NS1-104 mutant, there was only a relatively mild reduction in virus propagation, showing that the NS1 N-terminal domain plays a critical role in supporting efficient viral replication.

FIG. 7.

The NS1-104 protein supports viral replication and inhibition of PKR but not inhibition of IFN-α/β. (A) MDCK-C3 cells that contain a stable IFN-β promoter luciferase reporter gene were infected with recombinant wild-type (WT), ΔNS1, or NS1-104 virus at an MOI of 1. Luciferase activity was determined 8 h p.i. and is presented as the change in activation or induction (n-fold) compared to the value for luciferase activity in mock-infected cells. The graph shows average values of a typical experiment conducted in duplicate, which has been repeated five times. Error bars indicate the standard deviations. (B) MDCK cells were transfected with IRF-3 reporter plasmid and pRL-TK-luc for normalization. Twenty-four hours posttransfection, cells were mock infected or infected with wild-type, ΔNS1, or NS1-104 virus at an MOI of 1. Luciferase activities were determined 14 h p.i., and promoter induction is presented as the change in activation (n-fold) compared to the luciferase activity in mock-infected cells. The graph shows average values of two experiments conducted in duplicate. Error bars indicate the standard deviations. (C and D) A549 cells were mock infected or infected with wild-type, ΔNS1, or NS1-104 virus at an MOI of 1. Eighteen hours p.i., the supernatants were removed, and IFN-β (C) and IFN-α levels (D) were quantitated by ELISAs. The graphs show the average values of three experiments conducted in duplicate. Error bars indicate the standard deviations. hu, human. (E) Six- and eleven-day-old embryonated chicken eggs were inoculated with 10³ infectious units of wild-type or NS1-104 virus and incubated for 72 h at 33°C, and virus titers were determined. The indicated values represent the averages of at least three independent experiments. Error bars indicate the standard deviations. (F) A549 cells were either mock infected or infected with wild-type, ΔNS1, or NS1-104 virus at an MOI of 1. Cell extracts were prepared 10 h p.i. and were analyzed by immunoblotting with antibodies specific for phospho-PKR (P-PKR), total PKR, and the viral NP protein.

TABLE 2.

Replication in MDCK cells^a

Expt no. (MOI)	Virus titer (ffu/ml)
	WTb	NS1-104
Expt 1 (0.01)	3.8 × 107	1.7 × 106
Expt 2 (0.01)	4.3 × 107	1.5 × 106
Expt 3 (0.01)	7.7 × 107	1.2 × 106
Expt 4 (0.01)	7.1 × 107	3.5 × 105

^aConfluent monolayer cultures were infected at an MOI of 0.01 and incubated for 3 days at 33°C. Virus titers were determined as described in Materials and Methods.

^bWT, wild type.

Our previous analyses had indicated a major impact of PKR in determining viral growth (Fig. 4 and 5 and Table 1). We therefore hypothesized that the truncated NS1-104 virus was able to inhibit the activation of PKR and tested this assumption by immunoblot analysis of lysates from infected A549 cells. Indeed, the truncated NS1 protein inhibited PKR phosphorylation as well as the wild-type protein did, while the total amount of PKR was not affected (Fig. 7F). Taken together, these experiments strongly suggest that the function of the influenza B virus NS1 protein as an antagonist of IFN-α/β induction is not mainly dependent on its dsRNA-binding capacity but rather on its C-terminal domain. Nevertheless, binding of dsRNA by the N-terminal part of the NS1 protein appears to play a pivotal role in efficient viral replication and inhibition of PKR activation.

DISCUSSION

Previously, we have characterized the influenza B virus NS1 protein as a viral factor of 281 amino acids that antagonizes induction of the IFN-α/β system and boosts viral replication (7, 9). By using in vitro binding and reporter gene assays, it was established in previous studies that the B/NS1 protein contains an N-terminal domain (aa 1 to 93), known to confer dimerization and binding to dsRNA (60), and a larger C-terminal part (aa 94 to 281) that can suppress activation of the IFN-β promoter by Sendai virus infection (9). We showed recently that alanine replacements of the highly conserved basic amino acids at positions 47/50, 58/60/64, and 77/78 eliminate dsRNA binding, whereas substitutions at positions 33/38, 52/53/54, and 83/86 had only a partial or no effect (9). However, the specific functions of distinct B/NS1 domains or even amino acids in the complex processes involved in the subversion of antiviral responses had previously not been addressed.

In this report we used reverse genetics to explore in infected cells the role of the dsRNA-binding activity of the influenza B virus NS1 protein. This function is shared by the type A NS1 protein, and the sequestration of dsRNA had been suggested as a mechanistic basis for its IFN-antagonistic function (61). According to this hypothesis, the viral NS1 proteins would prevent recognition of virus-induced dsRNA in infected cells through sensor proteins, such as the recently described cellular RNA helicases RIG-I and/or MDA5 that signal for activation of type I IFN genes (1, 19, 23). However, in disagreement with this model, we found that all three influenza B mutant viruses expressing dsRNA-binding-defective NS1 proteins were as weak activators of the transcription factor IRF-3 and type I IFN genes as mutants that had retained partial or full dsRNA binding. The slight increase in IFN-β secretion by these viral recombinants compared to the wild-type virus and the failure of some NS1 proteins to inhibit nuclear migration of IRF-3 may reflect a partial involvement of dsRNA sequestration in silencing transcriptional activation of the IFN-β gene. However, none of these mutants induced IFN-α gene expression significantly higher than the wild-type virus did. Furthermore, two recent studies also indicated that the NS1 protein expressed by influenza A virus does not necessarily require dsRNA binding to antagonize IFN induction (8, 39). Thus, the original dsRNA sequestration model needs to be reconsidered (see below).

Ectopic expression of either the N-terminal dsRNA-binding domain or the residual C-terminal part of the B/NS1 protein alone (amino acids 1 to 93 and 94 to 281, respectively) was previously shown to inhibit IFN-β promoter induction by Sendai virus in a reporter gene setting (9). However, when expressed in the context of an authentic influenza B virus infection, the N-terminal part was not sufficient for this activity, since the loss of amino acids 105 to 281 was accompanied by a drastic increase in IFN-β secretion. Thus, the inhibition of Sendai virus-mediated IFN induction may have different quantitative or qualitative requirements compared to influenza virus infections, emphasizing the importance of reverse genetic tools in such studies. The observation that the NS1-104 virus induced twice as much IFN-α/β as the ΔNS1 virus did might be caused by the better replication of the former mutant, leading to increased production of activating dsRNAs. In reflection of this effect, we suggest that the previously unexplained strong cytolytic activities of two variants of influenza B/Yamagata/1/73 virus expressing shortened NS1 proteins of 90 and 128 amino acids may also have been caused by high levels of IFN that can trigger cell death (55, 57). Thus, the present study shows that the dsRNA-binding activity contributes relatively little to the inhibition of IFN-α/β induction by the NS1 protein. Instead we propose that the major IFN-antagonistic function of NS1 depends on interference at some stage with the signaling module that is activated after recognition of dsRNA by the cellular RIG-I and/or MDA5 sensor proteins. In this scenario, through its C-terminal domain, NS1 may either affect these RNA helicases directly or block any of the downstream steps, which involves recruitment of the kinase(s) TBK-1 and/or IKKɛ by the adaptor IPS-1, resulting in the activation of IRF-3/IRF-7 (19). In fact, inhibition of RIG-I signaling that is independent of dsRNA binding has recently also been described for other virus families, such as the Borna- and Filoviridae (4, 58). Experiments are under way to define such interactions for the influenza virus NS1 proteins.

Although all recombinant viruses expressing NS1 mutant proteins of 281 amino acids induced only low levels of IFN-β, the three viruses with abolished dsRNA-binding capacity and one mutant with low dsRNA-binding capacity replicated poorly in IFN-competent 11-day-old chicken eggs and MDCK cells. In contrast, the highly interferonogenic NS1-104 virus grew only slightly less than the wild-type virus did. These were unexpected findings, as efficient replication in IFN-competent hosts has in the past been mainly attributed to effective inhibition of IFN production (14). Consequently, we reasoned that replication of the attenuated viruses was due to another dsRNA-mediated antiviral activity, such as PKR. Indeed, the present study provides the first direct evidence that the B/NS1 protein inhibits activation of PKR in vivo, because it was phosphorylated in ΔNS1-infected cells, but not in wild-type virus-infected cells. Analysis of the phosphorylation status of PKR and its target eIF2α after infection with recombinant viruses further revealed a direct correlation of the dsRNA-binding activity of the B/NS1 protein and its capacity to inhibit PKR activation. Further, the N-terminal 104 amino acids of the B/NS1 protein that contain little more than the dsRNA-binding domain proved to be sufficient for this function. Thus, the findings establish an important role for the dsRNA-binding activity of the B/NS1 protein to prevent PKR activation. Whether this involves a general sequestration of any dsRNA species in infected cells or a rather selective shielding of nucleic acids associated with viral ribonucleoprotein remains a subject of further analysis. We would like to add that the same activity may also protect against other dsRNA-dependent antiviral enzymes, such as the 2′-5′ oligoadenylate synthetase that was recently shown to be silenced by type A influenza virus (39). It is also possible that the dsRNA-binding activity of NS1 contributes to another step during influenza B virus replication that is not known at this point. Surprisingly, we detected no phosphorylated eIF2α after ΔNS1 virus infection, although PKR was clearly activated. A possible explanation is that the cellular stress imposed by the simultaneous viral induction of PKR and IFN-α/β triggered activation of a phosphatase complex that dephosphorylates eIF2α, which could involve GADD34 (growth arrest- and DNA damage-inducible protein 34) (20, 42). We also considered an altered interaction between mutant NS1 proteins and its only known cellular binding partner, the IFN-induced ISG15 protein, to explain growth defects of the recombinant viruses. However, this possibility was excluded in coprecipitation assays showing that all three dsRNA-binding-defective NS1 proteins have largely retained their ISG15-binding capacity (J. Schneider and T. Wolff, unpublished results). The truncated NS1-104 protein also bound to ISG15, but it is expected to be inactive in blocking the conjugation of ISG15 to its cellular targets (69), which may contribute to the mild growth phenotype of the corresponding mutant virus.

The blockade of PKR activity is a common theme in many virus families, highlighting that this enzyme is a key factor of the cellular antiviral response (27). Interestingly, viral PKR inhibitors function by quite diverse mechanisms. This includes sequestration of activating dsRNA as known for the vaccinia virus E3L and reovirus σ3 proteins, as well as dsRNA-independent activities, such as serving as a pseudosubstrate for PKR or mediating the dephosphorylation of eIF2α, which has been described for the vaccinia virus K3L and herpes simplex virus γ₁34.5 proteins, respectively (27). It is well established that the inhibition of PKR activation is a major function of the influenza A virus NS1 protein (3). However, inconsistent findings exist concerning the underlying mechanism. One line of evidence argues in favor of a sequestration mode, as the mutational loss of dsRNA-binding activity was accompanied by a failure of the A/NS1 protein to prevent PKR activation both in vitro and in vivo (17, 33). In contrast, another recent report showed that a single recombinant virus expressing a dsRNA-binding-defective A/NS1 protein was a weak PKR activator (30). These authors proposed that the A/NS1 protein prevents the conformational switch involved in PKR activation by a direct binding interaction. However, the existence of such complexes in infected cells remains to be established. Thus, further investigations involving careful side-by-side comparisons of viral mutants will be required to determine whether type A and B influenza viruses whose NS1 proteins have little sequence identity use the same or different strategies to avoid PKR induction.

In summary, our results demonstrate that efficient replication of influenza B virus depends on the dsRNA-binding activity of the NS1 protein, which is important to prevent activation of PKR and possibly other dsRNA-responsive factors. Interestingly, dsRNA shielding does not appear to be sufficient to prevent IFN-α/β induction by influenza B virus. Possible explanations include the existence of different intracellular threshold levels of dsRNA required for the induction of the RNA helicases RIG-I/MDA-5 and PKR or a selectivity of NS1 for dsRNA targets recognized by PKR, but not by RIG-I/MDA-5. The observed lack of activity of the NS1 N-terminal domain towards IFN inhibition in influenza B virus-infected cells underscores the important role of the C-terminal part of the protein in this process, which presumably involves an RNA-independent mechanism. Thus, further analyses of the intracellular RNA and protein targets of the NS1 domains appear to be a promising strategy not only to learn more about the details of the IFN suppressive activities of influenza viruses but also about the factors and mechanisms that drive cellular responses to virus infections in general.