DDX3 Interacts with Influenza A Virus NS1 and NP Proteins and Exerts Antiviral Function through Regulation of Stress Granule Formation

ABSTRACT

DDX3 belongs to the DEAD box RNA helicase family and is a multifunctional protein affecting the life cycle of a variety of viruses. However, its role in influenza virus infection is unknown. In this study, we explored the potential role of DDX3 in influenza virus life cycle and discovered that DDX3 is an antiviral protein. Since many host proteins affect virus life cycle by interacting with certain components of the viral machinery, we first verified whether DDX3 has any viral interaction partners. Immunoprecipitation studies revealed NS1 and NP as direct interaction partners of DDX3. Stress granules (SGs) are known to be antiviral and do form in influenza virus-infected cells expressing defective NS1 protein. Additionally, a recent study showed that DDX3 is an important SG-nucleating factor. We thus explored whether DDX3 plays a role in influenza virus infection through regulation of SGs. Our results showed that SGs were formed in infected cells upon infection with a mutant influenza virus lacking functional NS1 (del NS1) protein, and DDX3 colocalized with NP in SGs. We further determined that the DDX3 helicase domain did not interact with NS1 and NP; however, it was essential for DDX3 localization in virus-induced SGs. Knockdown of DDX3 resulted in impaired SG formation and led to increased virus titers. Taken together, our results identified DDX3 as an antiviral protein with a role in virus-induced SG formation.

IMPORTANCE DDX3 is a multifunctional RNA helicase and has been reported to be involved in regulating various virus life cycles. However, its function during influenza A virus infection remains unknown. In this study, we demonstrated that DDX3 is capable of interacting with influenza virus NS1 and NP proteins; DDX3 and NP colocalize in the del NS1 virus-induced SGs. Furthermore, knockdown of DDX3 impaired SG formation and led to a decreased virus titer. Thus, we provided evidence that DDX3 is an antiviral protein during influenza virus infection and its antiviral activity is through regulation of SG formation. Our findings provide knowledge about the function of DDX3 in the influenza virus life cycle and information for future work on manipulating the SG pathway and its components to fight influenza virus infection.

INTRODUCTION

DDX3 belongs to the DEAD box RNA helicase family and harbors ATPase and RNA helicase activities (1). Like most other DEAD box helicases, DDX3 is a multifunctional protein with functions related to RNA metabolism, RNA export, ribosome biogenesis, cellular signaling, apoptosis, and viral infection (2, 3). DDX3 is known to enhance antiviral innate immunity by interacting with specific proteins of the type I IFN pathway (4). However, many viruses employ viral proteins, such as vaccinia virus (VACV) K7, hepatitis B virus (HBV) polymerase (Pol), and hepatitis C virus (HCV) core protein, to counteract DDX3 function and in turn use DDX3 to enhance its own replication (5–9). In contrast to its antiviral function, DDX3 is required for the replication of several viruses, such as human immunodeficiency virus (HIV), West Nile virus (WNV), Japanese encephalitis virus (JEV), HCV, and norovirus (4). Therefore, existing literature portrays DDX3 both as a host factor required for viral replication and as a component of the antiviral innate immune response.

Stress granules (SGs) are discrete cytoplasmic foci containing untranslated mRNA in nucleoprotein aggregates. They form in eukaryotic cells in response to a variety of environmental stress conditions, including viral infections (10). The first step in the signaling cascade leading to SG assembly is the phosphorylation of eukaryotic translation initiation factor 2α (eIF2α), which can be regulated by any of the four serine/threonine kinases, namely, double-stranded RNA-dependent protein kinase R (PKR), heme-regulated translation inhibitor kinase (HRI), PKR-like endoplasmic reticulum kinase (PERK) and general control nonderepressible 2 (GCN2) (11–13). PKR is activated by heat, UV irradiation, and viral infection (14), HRI is activated in erythroid cells subject to oxidative stress and when levels of free heme are limiting during hemoglobin assembly (15, 16), PERK is activated in response to unfolded protein accumulation in endoplasmic reticulum (17, 18), and GCN2 is activated during amino acid deprivation (19). Phosphorylation of eIF2α reduces the availability of ternary complex eIF2-GTP-tRNA_i^Met, which is required to load the initiator tRNA_i^Met onto the small ribosomal subunit to initiate translation (20). This results in the accumulation of stalled translation preinitiation complexes, containing translationally inactive messenger ribonucleoproteins, which recruit RNA-binding proteins such as T-cell intracellular antigen 1 (TIA-1) and TIA-1-related protein (TIAR), which in turn mediate the formation of SGs (21).

Several RNA helicases, including DDX3, have been shown to localize in SGs (22). A recent study reported that DDX3 localized in SGs induced by a variety of cellular stresses, including sorbitol, arsenite, dithiothreitol (DTT), heat shock treatment, and UV irradiation. Further, DDX3 was found to be an SG-nucleating factor, and DDX3-eIF4E interaction is essential for SG formation (23). Many viruses induce SGs through the activation of PKR kinase and in some cases GCN2 by detection of viral RNA in the cytoplasm (24, 25). Most viruses, including influenza A virus, have mechanisms to inhibit SG formation, implicating the antiviral role of SGs in virus life cycle (10). In case of influenza viruses, NS1 protein is known to inhibit PKR, thereby preventing eIF2α phosphorylation and SG formation (26). Besides NS1, the nucleoprotein (NP) and polymerase acidic protein X (PA-X) have also been shown to aid influenza virus in overcoming stress-induced translation arrest (27). Besides inducing translation arrest, SGs have also been shown to play a role in interferon (IFN) synthesis by sequestering retinoic acid-inducible gene I (RIG-I) and influenza viral RNA (vRNA), thereby serving as a platform for sensing of viral RNA by RIG-I (28). In addition, other antiviral proteins, such as MDA5, LGP2, RNase L, OAS, and PKR, have also been shown to localize in influenza virus-induced SGs (28). These studies underscore the antiviral role that SGs play in influenza virus infection and highlight the involvement of DDX3 in virus life cycle and SG formation. These studies prompted us to explore the role of DDX3 in influenza virus-induced SG formation.

Several host factors involved in RNA metabolism, including DDX3, have been shown to associate with the viral polymerase complex and colocalize with NP (29). However, detailed studies on the viral interaction partners and the function of DDX3 during influenza virus replication were not conducted. Another study, attempting to assess the effect of DDX3 downregulation on influenza virus polymerase activity, could not determine conclusively whether DDX3 regulates this function (30). Thus, we set out to investigate whether DDX3 plays a role during influenza virus infection and the mechanism by which DDX3 is involved in the regulation. We found that DDX3 is an interaction partner with viral NS1 and NP proteins and localizes in virus-induced SGs. NS1 is able to counteract virus-induced SG formation and DDX3 localization in these SGs. We also identified the domains in DDX3 that are critical for interaction with NS1, NP, and SG formation. Moreover, knockdown of DDX3 impaired SG formation and increased virus titers upon infection with a PR8 NS1 deletion virus. Thus, we demonstrated DDX3 as an antiviral protein for influenza virus infection with a prominent role in regulating SG formation, which warrants further study and understanding.

MATERIALS AND METHODS

Cells and viruses.

Madin-Darby canine kidney (MDCK) cells and newborn porcine tracheal epithelial (NPTr) cells were maintained in minimum essential medium (MEM) supplemented with 10% fetal bovine serum (FBS) (Life Technologies). 293T cells were maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% FBS.

Influenza A/Puerto Rico/8/34 (H1N1) (PR8-WT) and A/Sw/SK/18789/02 (H1N1) (SIV/SK-WT) viruses were propagated in 11-day-old embryonated chicken eggs as described previously (31). PR8 virus lacking NS1 protein (del NS1) was kindly provided by A. García-Sastre and was propagated in Vero cells maintained in MEM with 10% FBS. PR8-WT and SIV/SK-WT virus were titrated by plaque assay on MDCK cells, while del NS1 was titrated on Vero cells. PR8 virus carrying mutations R38A and K41A in NS1 was rescued by reverse genetics (32). The virus was propagated by using a MDCK cell line stably expressing the NS1 protein (MDCK-NS1) and titrated by plaque assay on MDCK-NS1 cell line.

Antibodies and reagents.

Rabbit polyclonal NS1 and NP antibodies were generated in our laboratory as previously described (33). The other antibodies were purchased from different sources, as follows: mouse anti-HA antibody and mouse anti-Flag M2 antibody, Sigma-Aldrich; rabbit anti-Flag (DYKDDDDK) tag antibody, Cell Signaling Technology; mouse anti-influenza A virus NP AA5H antibody, AbD Serotec; rabbit polyclonal antibody to DDX3, Abcam; rabbit polyclonal antibody to PABP1, Abcam; rabbit polyclonal antibody to HA tag (chromatin immunoprecipitation [ChIP] grade), Abcam; mouse monoclonal antibody to β-actin, Cell Signaling Technology; goat anti-TIA-1 antibody, Santa Cruz; donkey anti-rabbit IgG secondary antibody with Alexa Fluor 405, Abcam; donkey anti-mouse IgG secondary antibody with Alexa Fluor 488, Life Technologies; donkey anti-goat secondary antibody with Alexa Fluor 633, Life Technologies; IRDye 680RD anti-rabbit antibody, Li-Cor; and IRDye 800CW anti-mouse antibody, Li-Cor. Horse serum used in immunofluorescent staining was purchased from Life Technologies. Stellaris fluorescence in situ hybridization (FISH) probes specific to PR8 M vRNA for FISH assays were purchased from Biosearch Technologies.

Transfection and IP.

To examine DDX3 interaction with viral proteins, 293T cells were seeded at a density of 1 × 10⁶ cells/well in six-well plates. One microgram of each plasmid expressing the protein of interest was transfected using TransIT-LT1 as per the manufacturer's recommendation. For transfection in NPTr cells, Lipofectamine LTX with Plus reagent (Life Technologies) was used as per the manufacturer's recommendation. The plasmids used for transfection include pcDNA-HA-DDX3, pcDNA-SK-NP, pcDNA-PR8-NP, pcDNA-PR8-NS1, and pCMV-3×Flag-PR8-NS1, pCMV-3×Flag-DDX3 (DDX3), pCMV-3×Flag-core helicase DDX3 (DDX3-CH), pCMV-3×Flag-C-term deletion DDX3 (DDX3-del CTD), and pCMV-3×Flag-N-term deletion DDX3 (DDX3-del NTD). The protein of interest was cloned into pCMV-3×Flag and pcDNA-HA plasmids such that the Flag or HA tag was fused to the N-terminal sequence of the expressed fusion protein. For examining protein interactions by immunoprecipitation (IP), the cell lysate was collected in 1 ml Flag lysis buffer (FLB) (50 mM Tris HCl, pH 7.4, with 150 mM NaCl, 1 mM EDTA, and 1% Triton X-100). The cell lysate was then sonicated and cleared of cellular debris by centrifugation. For RNase A treatment, RNase A (Life Technologies) was added to the lysate at a concentration of 10 μg/ml and incubated on ice for 30 min before the immunoprecipitation assay.

For IP, 1.5 μg of mouse anti-HA (Sigma-Aldrich) or mouse monoclonal anti-Flag M2 (Sigma-Aldrich) antibody or mouse anti-influenza A virus NP (AbD Serotec) antibody was added to the cell lysate and incubated with gentle rocking at 4°C for 1 h and 15 min. Then, 35 μl of Dynabeads protein G (Life Technologies) was added to the lysate and incubated for another 1 h and 15 min with gentle rocking at 4°C. The beads were then washed extensively with FLB, and the precipitated proteins were subjected to sodium dodecyl sulfate–10% polyacrylamide gel electrophoresis (SDS-PAGE) and Western blotting with the appropriate antibodies.

Western blotting.

Samples were resolved by SDS-PAGE and transferred onto nitrocellulose membranes (Bio-Rad). Membranes were blocked with 10% skim milk for 30 min and then incubated with a primary antibody diluted in Tris-buffered sodium chloride solution with 0.1% Tween 20 (TBST) at 4°C overnight. The following primary antibodies were used for Western blotting: rabbit polyclonal antibody to NS1 and rabbit polyclonal antibody to NP (in-house generated), rabbit anti-Flag (DYKDDDDK) tag antibody (Cell Signaling Technology), rabbit polyclonal antibody to DDX3 (Abcam), rabbit polyclonal antibody to HA tag (ChIP grade) (Abcam), and mouse monoclonal antibody to β-actin (Cell Signaling Technology). After being washed in TBST, membranes were incubated in TBST containing IRDye 680RD goat anti-rabbit IgG antibody or IRDye 800CW goat anti-mouse IgG antibody. Membranes were washed again in TBST and were scanned using an Odyssey imager (Li-Cor Biosciences).

Immunofluorescent staining.

NPTr cells were grown on glass chamber slides. After experimental treatment, the cells were washed with PBS, fixed with 4% paraformaldehyde (PFA) in PBS for 15 min at room temperature, and then permeabilized with ice-cold methanol for 15 min at room temperature. The cells were washed in PBS, blocked in 5% horse serum in PBS for 45 min, and then incubated with the primary antibody in blocking buffer for 2 h at room temperature or overnight at 4°C (21). The following primary antibodies were used for immunostaining: mouse anti-Flag M2 antibody (Sigma), rabbit anti-Flag (DYKDDDDK) tag antibody (Cell Signaling Technology), mouse anti-influenza A virus NP AA5H antibody (AbD Serotec), rabbit polyclonal antibody to DDX3 (Abcam), rabbit polyclonal antibody to PABP1 (Abcam), and goat anti-TIA-1 antibody (Santa Cruz). The cells were again washed in PBS and incubated with the Alexa Fluor-conjugated secondary antibodies in blocking buffer for 1 h at room temperature. After washing in PBS, the cells were mounted using Prolong Gold antifade reagent (Life Technologies), and the images were captured using Leica TCS SP8 confocal laser microscope.

For quantification of SG-forming cells, approximately 50 cells immunostaining positive for NP were considered from two random microscopy panels. The cells showing punctate TIA-1 staining among the NP-positive cells were counted visually.

Knockdown of DDX3.

NPTr cells were plated at densities of 4 × 10⁴ cells/well in 24-well plates, 3.8 × 10⁴ cells/well in 4-well chamber slides, and 8 × 10⁴ cells/well in 12-well plates. The next day, medium was replaced with Opti-MEM and the small interfering RNA (siRNA)-containing transfection mix was made with Opti-MEM and Lipofectamine 2000 as per the manufacturer's protocol. An siRNA mixture containing four independent siRNAs targeting DDX3 (GS1654) and negative siRNA (SI03650318) were obtained from Qiagen. The transfection mix was added to the cells in Opti-MEM for 5 to 6 h and then was replaced with complete medium and incubated for 48 h before the experimental treatment.

FISH assay.

For the FISH assay, cells were washed with PBS and fixed with 4% paraformaldehyde (PFA) for 10 min at room temperature after virus infection. After being washed with PBS, the cells were permeabilized with 70% ethanol overnight at 4°C. The cells were then incubated in a hybridization solution (10% dextran sulfate, 2 mM vanadyl-ribonucleoside complex, 0.02% RNA-free bovine serum albumin [BSA], 1 mg/ml Escherichia coli tRNA, 2× saline-sodium citrate [SSC; 1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate], and 10% formamide) containing 125 nM vRNA probe (Stellaris FISH probes; Biosearch Technologies) at 28°C overnight in the dark. The cells were then washed again and imaged with a Leica TCS SP8 confocal laser microscope after addition of the mounting solution. The Stellaris FISH probes against the vRNA M segment were conjugated with Quasar 670 dye. The probes were custom designed using the Stellaris probe designer (version 4.1) with the sequence of the M segment as the target. A total of 39 probes were generated, and this mixture of vRNA probes were used in the FISH assay to visualize the M vRNA in the infected cells.

To visualize DDX3 and NP, primary antibodies against the two proteins were added to the hybridization solution along with the vRNA probe during the overnight incubation period described above. Next day, the cells were washed and incubated with respective Alexa Fluor secondary antibodies in hybridization solution for 1 h in dark at room temperature. Cells were washed again and observed with the confocal microscope.

Statistical analysis.

The statistical analysis was performed using GraphPad Prism 6 software. Two-way analysis of variance with the Sidak multiple-comparison test was used to compare the virus titer between cells treated with DDX3 siRNA and those treated with off-target siRNA. P values of <0.05 were considered statistically significant. Data presented are means and standard deviations.

RESULTS

DDX3 interacts with viral NS1 and NP proteins.

DDX3 is known to affect the life cycle of a number of viruses by interacting with their respective viral proteins, such as HIV Tat, HCV core, HBV pol and VACV K7 (3). Therefore, we speculated that DDX3 could affect influenza virus life cycle by interacting with one or several influenza virus proteins during infection. To identify the DDX3 interaction partners, we tested the interaction between HA-tagged DDX3 and viral proteins expressed during infection by IP. 293T cells were first transfected with HA-tagged DDX3 plasmid and then were infected with the influenza virus for 11 to 12 h. The cell lysate was subjected to IP with an antibody against the HA tag. The precipitated proteins were then subjected to Western blotting with antibodies against the HA tag, viral NS1 protein, and NP protein, respectively. Expression of viral proteins is similar in the input of the infected-cell lysates (Fig. 1A, lanes 1 and 3); both NS1 and NP proteins coprecipitated readily with HA-DDX3 (lane 2) but not with HA-vector (lane 4). Neither HA-DDX3 nor the viral proteins were detected in the input and pulldown samples when HA-vector was transfected in uninfected cells (Fig. 1A, lanes 5 and 6). These data demonstrated that viral proteins NS1 and NP interact with DDX3 protein during infection.

FIG 1.

Identification and characterization of viral proteins interacting with DDX3. (A) 293T cells were transfected with HA-DDX3 or HA-vector plasmid. At 36 h posttransfection, cells were infected with SK-WT at a multiplicity of infection (MOI) of 1 or left uninfected. Cell lysates were prepared at 11 to 12 h p.i and subjected to IP with HA antibody. Precipitated proteins were subjected to Western blotting using antibodies against HA tag and NS1 and NP proteins. (B) Flag-DDX3 or Flag-vector plasmid was cotransfected with or without PR8-NS1-expressing plasmid in 293T cells. At 48 h posttransfection, cell lysates were collected and subjected to IP with Flag antibody. Precipitated proteins were subjected to Western blotting using antibodies against the Flag tag and NS1 protein. (C) HA-DDX3 or HA vector plasmid was transfected with or without SIV/SK-NP-expressing plasmid in 293T cells. At 48 h posttransfection, cell lysates were collected and subjected to IP with HA antibody. Precipitated proteins were subjected to Western blotting using antibodies against the HA tag and NP protein. (D) Flag-PR8 NS1 or Flag vector plasmid was cotransfected with HA-DDX3 or HA-vector in 293T cells. At 48 h posttransfection, cell lysates were collected and subjected to IP with Flag antibody and protein G Dynabeads. Precipitated proteins were subjected to Western blotting using antibodies against the Flag tag and the HA tag. (E) HA-DDX3 or HA vector plasmid was transfected into 293T cells and either infected with SIV/SK-WT at an MOI of 1 or left uninfected. At 11 to 12 h p.i., cell lysates were collected and subjected to IP with NP antibody. Precipitated proteins were subjected to Western blotting using antibodies against NP protein and the HA tag. I/P, input; P/D, pulldown.

To examine whether NS1-NP interaction is dependent upon the expression of other viral proteins, we cotransfected 293T cells with plasmids expressing Flag- or HA-tagged DDX3 and NS1 or NP and examined their interaction by precipitating DDX3 from the lysate. To verify NS1-DDX3 interaction, Flag-DDX3 or Flag-vector plasmid was transfected along with or without NS1-expressing plasmid, and the cell lysates were subjected to IP with Flag antibody. The levels of input NS1 protein expression were similar (Fig. 1B, lanes 1 and 3), but NS1 was detected in the precipitated complex only when Flag-DDX3 was expressed, not when Flag-vector was expressed (lane 2 versus 4). To verify NP-DDX3 interaction, HA-DDX3 or HA-vector plasmid was transfected with or without NP-expressing plasmid (Flag beads bind to NP protein nonspecifically; thus, we used the HA tag) and the cell lysates were subjected to IP with HA antibody. The levels of input NP protein expression were similar (Fig. 1C, lanes 1 and 3), but NP was detected in the precipitated complex only when HA-DDX3 was expressed, not when HA-vector was expressed (lane 2 versus 4). These results showed that NS1 and NP protein can interact with DDX3 independently of infection and other viral components.

To further confirm this interaction, we conducted the reciprocal pulldown assay. Flag-NS1 or Flag-vector plasmid was cotransfected with HA-DDX3 or HA-vector plasmid in 293T cells, and the cell lysates were then subjected to IP using the Flag antibody. Flag-NS1 coprecipitated HA-tagged DDX3 (Fig. 1D, lane 2), while HA-DDX3 was not detected in the precipitated complex expressing Flag-vector and HA-DDX3 (lane 4). For DDX3-NP interaction, 293T cells were transfected with HA-DDX3 or HA-vector and then were infected with wild-type (WT) virus. The cells were subjected to IP with NP antibody. HA-DDX3 coprecipitated with NP (Fig. 1E, lane 2) in infected-cell lysate. No HA-DDX3 was detected in the precipitated complex of mock-infected-cell lysate, which does not express NP (Fig. 1E, lane 4). When HA-vector was expressed in infected cells, although NP was precipitated successfully, no HA-DDX3 was detected (Fig. 1E, lane 6). These results demonstrated that NS1 and NP interact with DDX3 during infection and that the interaction is specific.

DDX3-NS1 and DDX3-NP interaction is RNA independent.

NS1, NP, and DDX3 exhibit RNA-binding activity (34–36). Therefore, we were interested in investigating whether DDX3-NS1 and DDX3-NP interactions were mediated by RNA. To this end, plasmids expressing the NS1 protein and HA-DDX3 were cotransfected in 293T cells, and the cell lysates were pretreated with RNase A before being subjected to IP with HA antibody. Efficient degradation of RNA in the sample by RNase A treatment was confirmed by measuring the rRNA ratio (28S/18S) and the RNA integrity number (RIN) (37, 38) using an Agilent 2100 Bioanalyzer (data not shown). While NS1 was not coprecipitated with HA-vector (Fig. 2A, lane 4), NS1 was coprecipitated with HA-DDX3 irrespective of the RNase A treatment (lane 2 versus 5). It is well established that substitution of R38 and K41 to alanine in NS1 disrupts its RNA binding activity (39); we therefore coexpressed NS1 with R38A/K41A mutation (MT NS1) and HA-DDX3 in 293T cells and performed IP with antibody against the HA tag. Surprisingly, MT NS1 was not coprecipitated with HA-DDX3 (Fig. 2B, lane 2). Additionally, we also tested the interaction of MT NS1 expressed during infection with PR8 virus expressing the NS1 protein containing R38A/K41A mutations (MT virus) and HA-DDX3 (Fig. 2C). Similar to the results in cotransfection with plasmids, MT NS1 expressed during infection did not coprecipitate with HA-DDX3 (Fig. 2C, lane 2). These data suggest that even though DDX3-NS1 interaction is RNA independent, the R38/K41 site in NS1 that is important for RNA binding activity by itself is essential for interaction with DDX3.

FIG 2.

RNA dependency of DDX3 interaction with NS1 and NP. (A) PR8-NS1 expressing plasmid was cotransfected with HA-DDX3 or HA-vector plasmid in 293T cells. At 48 h posttransfection, cell lysate was collected, pretreated with RNase A at 10 μg/ml or left untreated, and subjected to IP with HA antibody. (B) HA-DDX3 or HA-vector plasmid was cotransfected with or without plasmid expressing PR8-NS1 R38A/K41A (MT NS1) in 293T cells. At 48 h posttransfection, cell lysate was collected and subjected to IP with HA antibody. (C) 293T cells were transfected with HA-DDX3 or HA-vector plasmid. At 24 h posttransfection, cells were either infected with PR8 virus carrying the R38A/K41A mutation in NS1 (MT virus) at an MOI of 10 or left uninfected. At 24 h p.i., cell lysate was collected and subjected to IP with HA antibody. (A, B, and C) Precipitated proteins were subjected to Western blotting using antibodies against the HA tag and NS1 protein. (D) SIV/SK-NP-expressing plasmid was cotransfected with HA-DDX3 or HA-vector plasmid in 293T cells. At 48 h posttransfection, cell lysates were collected, pretreated with RNase A at 10 μg/ml or left untreated, and subjected to IP with HA antibody. Precipitated proteins were subjected to Western blotting using antibodies against the HA tag and NP protein. I/P, input; P/D, pulldown.

To investigate the RNA dependency of DDX3-NP interaction, cells were cotransfected with plasmids expressing HA-DDX3 and NP, respectively. Cell lysates were treated with RNase A before being subjected to IP with HA antibody. The RNase A treatment efficiency was tested as described before. As shown in Fig. 2D, Western blotting detected NP protein in the precipitated complex in HA-DDX3 expressing cell lysate irrespective of RNase A treatment (lane 2 versus 5), whereas NP was not coprecipitated with HA-vector (lane 4). This suggested that the interaction between DDX3 and NP is independent of RNA.

The C-terminal domain of DDX3 has a predominant role in mediating DDX3 interaction with NS1 and NP.

DDX3 belongs to the DEAD box family of proteins and, like all the helicases in the family, contains a core-helicase domain flanked by a highly variable N-terminal region and a C-terminal region (36, 40). The core-helicase domain in human DDX3 comprises residues 168 to 582, while the flanking N- and C-terminal regions comprise residues 1 to 167 and 583 to 662, respectively (41). Based on sequence alignment with other DEAD box helicases, the core-helicase domain contains the nine conserved motifs involved in ATPase, helicase, and RNA binding activity (42). The N-terminal region contains a conserved leucine-rich nuclear export signal and is critical for CRM1-mediated nuclear export and eIF4E binding (3, 23, 36), while an SR-rich region in the C terminus is associated in an interaction with the nuclear export receptor NXF1/TAP (43). Therefore, we constructed Flag-tagged plasmids expressing different truncated DDX3 mutants and examined their ability to interact with NS1 and NP.

As shown in Fig. 3A, plasmids expressing Flag-tagged full-length DDX3 and three truncation mutants expressing the core helicase domain of DDX3 alone (amino acids [aa] 168 to 582, DDX3-CH), DDX3 lacking the C-terminal domain (aa 1 to 582, DDX3-del CTD), or DDX3 lacking the N-terminal domain (aa 168 to 662, DDX3-del NTD) were constructed. To identify the domains critical for NS1 and NP interaction, 293T cells were cotransfected with one of the Flag-tagged DDX3 truncation mutant plasmids and a plasmid expressing either NS1 or NP. Cell lysates were subjected to IP with Flag antibody (for DDX3-NS1 interaction) or with NP antibody (for DDX3-NP interaction). Full-length DDX3 demonstrated strong interaction with NS1 (Fig. 3B, lanes 2 and 10), whereas the core-helicase domain of DDX3 lost the ability to interact with NS1 (lane 4). Addition of the N-terminal domain slightly restored interaction with NS1 (Fig. 3B, lane 12), whereas the addition of the C-terminal region had a stronger influence on restoring the interaction with NS1 (lane 14). Considering the amount of pulled-down truncated DDX3 in this sample, the effect is more profound. IP of DDX3 truncation and NP showed results similar to those observed in Fig. 3B. While full-length DDX3 was coprecipitated with NP (Fig. 3C, lanes 2 and 12), the core-helicase domain of DDX3 was not (lane 6). While the addition of the N-terminal domain of DDX3 restored interaction slightly (Fig. 3C, lane 14), the addition of the C-terminal region had a stronger impact on restoring the interaction with NP (lane 18). Additionally, none of the DDX3 truncations or full-length DDX3 bound to the beads when expressed alone, demonstrating the specificity of interaction of DDX3 and its truncated proteins with NP (Fig. 3C, lanes 4, 8, 16, and 20). These results suggested that the C-terminal domain (aa 583 to 662) and the N-terminal domain (aa 1 to 167) are essential for DDX3 interaction with its viral partners, although the C-terminal domain has a more prominent role than the N-terminal domain.

FIG 3.

Identification of DDX3 protein domains critical for viral protein interaction. (A) Flag-tagged plasmids expressing different DDX3 truncations were constructed to identify the domains important for NS1 and NP interaction and for SG formation. A schematic of the domains cloned into each plasmid is shown. (B and C) PR8-NP- or PR8-NS1-expressing plasmid was transfected with one of the Flag-tagged plasmids expressing different DDX3 truncations in 293T cells. At 48 h posttransfection, cell lysate was collected and subjected to IP with Flag antibody (B) or NP antibody (C). Precipitated proteins were subjected to Western blotting using antibodies against NS1 protein and the Flag tag (B) or NP protein and the Flag tag (C). I/P, input; P/D, pulldown.

DDX3 localizes in the SGs in response to del NS1 virus infection.

DDX3 has been documented to localize to cytoplasmic SGs and has been reported be an essential SG nucleating factor (23, 43, 44). Several other host proteins, such as NF90, RAP55, and FMRP, known to localize to SGs and involved in SG formation have also been shown to associate with influenza virus NP and NS1 (45–48). Influenza A virus lacking a functional NS1 and influenza A viruses that are impaired in NS1 RNA binding activity have been shown to induce robust SG formation, while the WT virus expressing a fully functional NS1 does not induce SGs throughout the virus life cycle (26, 28). A recent study revealed that viral NP is recruited into SG and that expression of NS1 is able to inhibit formation of RAP55- and NP-associated SGs (45). These studies led us to speculate that DDX3 could localize in influenza virus-induced SGs and might affect the virus life cycle by interacting with NP in SGs. In order to test this hypothesis, we first examined SG formation and DDX3 localization upon virus infection in NPTr cells. Unlike many human cell lines, such as A549 and HeLa cells, which are derived from carcinomatous tissues, NPTr was established following serial culture of primary cells (49). Additionally, the cell line is derived from the trachea, which is the primary site of influenza virus replication during infection (50). Thus, infection of NPTr cells would resemble conditions similar to natural infection, and hence, we used NPTr cells to study SG formation.

We infected NPTr cells with either a recombinant PR8 virus completely lacking the functional NS1 protein (del NS1) (51) or the WT PR8 virus and observed the formation of SGs at different time points by staining the SG-specific marker T-cell-restricted intracellular antigen-1 (TIA-1; red) (21). In addition, we also stained viral NP protein (green) and cellular DDX3 (blue) to identify infected cells and the localization of DDX3. As shown in Fig. 4A, no cytoplasmic punctate staining with TIA-1 was observed in mock-infected cells and at any of the time points tested in WT-virus-infected cells, indicating SGs did not form under these conditions. However, in del NS1 virus-infected cells, SGs started to form at 6 h p.i. and was sustained throughout the time points tested. More strikingly, we found that DDX3 colocalized with TIA-1 in virus-induced SGs at all the time points where SG formation was observed.

FIG 4.

Kinetics of SG formation in PR8-WT- and del NS1-infected cells and vRNA and NP localization in relation to SGs. (A) SG formation was analyzed by immunofluorescent staining of NPTr cells either mock infected or infected with PR8-WT or del NS1 at an MOI of 0.5. The cells were stained at predetermined time points. Virus-infected cells were identified by antibody staining for NP (green), SGs were indicated by staining for TIA-1 (red), and DDX3 localization was observed by staining for DDX3 (blue). (B) SG formation was further confirmed by immunofluorescent staining of NPTr cells that were either mock infected or infected with PR8-WT or del NS1 at an MOI of 0.5. The cells were stained at 11 h p.i. Virus-infected cells were identified by antibody staining for NP (green), SGs were indicated by staining for TIA-1 (red) and PABP1 (blue), both of which are SG markers. (C) Enlarged images of a del NS1-infected cell stained at 11 h p.i. for DDX3 (blue), TIA-1 (SG marker) (red), and NP (green). The arrowheads show NP colocalization with TIA-1 and DDX3, all of which form punctate structures characteristic of SG formation and localization. (D) NPTr cells were infected with del NS1 virus at an MOI of 0.5 or mock uninfected. At 11 h p.i., the cells were subjected to FISH (red) using probes specific for vRNA of M segment and counterstained with antibodies against DDX3 (blue) and NP (green).

In order to confirm the formation of SGs in del NS1 virus-infected cells, we stained the cells for another SG marker, PABP1 (21). Cells were also stained for NP (green) to identify infected cells and TIA-1 (red). As shown in Fig. 4B, no punctate staining with PABP1 (blue) or TIA-1 (red) was observed in mock-infected and WT-virus-infected cells. However, TIA-1 and PABP1 colocalized and formed cytoplasmic punctate staining in del NS1 virus-infected cells. Thus, these results confirm the formation of SGs and identify TIA-1 as an authentic marker for observing SG formation in del NS1-infected cells.

DDX3 and NP colocalize in SGs, but vRNA is not sequestered in virus-induced SGs.

Our IP experiments clearly demonstrate the interaction between viral NP protein and cellular DDX3 during infection (Fig. 1A, C, and E and 2D). Therefore, we were interested in studying where the NP and DDX3 interaction could occur in virus-infected cells. In the immunofluorescent-staining experiment, we noticed that NP also formed some granules in the cytoplasm, which colocalized with DDX3 and TIA-1 staining (Fig. 4C), suggesting that DDX3 interacts with NP in virus-induced SGs. Since NP protein encapsidates influenza viral RNA (vRNA) to form viral ribonucleoprotein (vRNP) complex, which is essential for viral transcription and replication (34), we wanted to understand whether the NP staining in the SGs is a result of vRNP being recruited to virus-induced SGs. We conducted a FISH assay using a probe against the M vRNA segment. FISH analysis clearly showed that vRNA (Fig. 4D, red) did not colocalize with the granular NP (green) in SGs. Note that we always observed granular DDX3 colocalized with TIA-1 in virus-infected cells (Fig. 4A and C). Therefore, in this particular experiment, we used DDX3 (blue) granule formation as a marker for SGs. These results demonstrated that NP and DDX3 colocalized in virus-induced SGs, while vRNA/vRNP is not sequestered in these SGs.

Influenza virus NS1 inhibits virus-induced SG formation and DDX3 localization in SGs.

SGs did not form at any of the time points tested with WT virus infection, but SG formation was readily observed in cells infected with del NS1 virus (Fig. 4A) starting at 6 h p.i. Therefore, we speculated that expression of the NS1 protein might inhibit virus-induced SG formation and DDX3 localization in SGs in WT-virus-infected cells. To test this, we studied SG induction upon del NS1 virus infection in cells expressing WT NS1. We transfected NPTr cells with plasmid expressing Flag-tagged NS1 protein (Flag-NS1) or Flag-vector and then infected cells with del NS1 virus. To study the DDX3 recruitment into SGs, cells were stained with antibodies against DDX3 (blue), TIA-1 (red), and Flag (green) (Fig. 5A). To monitor del NS1 virus-infected cells, the cells were also stained with antibodies against Flag (blue), NP (green), and TIA-1 (red) (Fig. 5B). As we expected, SGs and granular DDX3 were not found in NS1-expressing cells upon del NS1 virus infection (Fig. 5A and B). Additionally, we also examined whether NS1 could inhibit other forms of stress-induced SG formation, such as oxidative stress-induced SG formation via treatment with sodium arsenite (NaAs) (52). For this, we transfected NPTr cells with a plasmid expressing Flag-tagged NS1 protein (Flag-NS1) or Flag-vector and then treated the cells with NaAs for 1 h. Cells were then stained with antibodies against DDX3 (blue), TIA-1 (red) and Flag (green) (Fig. 5C). In contrast to the results observed with del NS1 virus-induced SGs, NS1 expression did not interfere with NaAs-induced SG formation, and NS1 was even recruited to these SGs in a few cells (Fig. 5C).

FIG 5.

Effect of NS1 on virus-induced and NaAs-induced SG formation. NPTr cells were transfected with 500 ng of Flag-NS1 or Flag-vector plasmid (A and B). At 36 h posttransfection, the cells were infected with del NS1 virus at an MOI of 0.5. At 11 h p.i., one set of cells were stained with antibodies against the Flag tag (green), DDX3 (blue), and TIA-1 (SG marker) (red) (A) and the other set of cells were stained with antibodies against the Flag tag (blue), NP (green), and TIA-1 (SG marker) (red) (B). (C) At 46 to 48 h posttransfection, the cells were treated with 0.75 mM NaAs for 1 h and were then stained with antibodies against the Flag tag (green), DDX3 (blue), and TIA-1 (SG marker) (red).

The core-helicase domain of DDX3 alone is sufficient for its localization into del NS1 virus-induced SGs.

To further our understanding of DDX3 function in virus-induced SGs, we considered studying the relative importance of the DDX3 domains in SG localization. NPTr cells were transfected with plasmids encoding either Flag-tagged full-length or truncated versions of DDX3 (depicted in Fig. 3A) and then infected with del NS1 virus. At 11 h p.i., cells were stained with antibodies against Flag (blue), TIA-1 (red), and NP (green). As shown in Fig. 6, full-length DDX3 colocalized with virus-induced SGs and so did the core-helicase domain (DDX3-CH) and the DDX3 with the C-terminal deletion (DDX3-del CTD). However, in the many microscopic fields we examined, none of the cells expressing DDX3 with the N-terminal deletion (DDX3-del NTD) exhibited SG formation upon virus infection.

FIG 6.

Characterization of DDX3 domains in virus-induced SG formation. NPTr cells were transfected with 500 ng of plasmids expressing Flag-tagged DDX3 truncations. At 36 h posttransfection, the cells were infected with del NS1 virus at an MOI of 0.5, and at 11 h p.i., cells were stained with antibodies against the Flag tag (blue), TIA-1(SG marker) (red), and NP (green).

DDX3 downregulation interferes with SG formation and enhances virus replication.

The dominant negative effect of N-terminal-deletion DDX3 on SG formation (Fig. 6) led us to postulate that DDX3 is not just recruited passively into SGs and might have a critical function in the formation of SGs. To confirm our speculation, we treated NPTr cells with siRNA specific to DDX3 (siRNA-DDX3) and observed SG formation in del NS1 virus-infected cells (Fig. 7A). The infected cells were stained with antibodies against DDX3 (blue), TIA-1 (red), and NP (green). Treatment with DDX3 siRNA resulted in a significant reduction of endogenous DDX3 expression compared to cells treated with off-target siRNA (siRNA-OT) (Fig. 7A, DDX3), and interestingly, virus-induced SG formation was also suppressed in cells with DDX3 downregulation (Fig. 7A, TIA-1 and NP). Notably, quantitation of infected cells exhibiting virus-induced SGs showed that the number of SG-forming cells diminished by 60% in DDX3 siRNA-treated cells compared to off-target-siRNA-treated cells. This observation suggested that DDX3 is not just passively recruited to SGs but has a critical function in the formation of virus-induced SGs.

FIG 7.

Effect of DDX3 downregulation on virus-induced SG formation and virus titer. NPTr cells were transfected with siRNA specific to DDX3 or off-target siRNA. (A) At 48 h after siRNA treatment, the cells were infected with del NS1 virus at an MOI of 0.5. At 11 h p.i., the cells were stained for DDX3 (blue), TIA-1 (SG marker) (red), and NP (green). (B) At 48 h after siRNA treatment, the cell lysates were subjected to Western blotting with antibodies against DDX3 and β-actin. The level of DDX3 expression was determined by normalizing the intensity of the DDX3 bands with the corresponding β-actin bands for each siRNA treatment and is represented as a graph below the Western blot images. (C) At 48 h after siRNA treatment, the cells were infected in triplicate with either PR8-WT, del NS1, or SIV/SK-WT virus at an MOI of 0.01. The supernatant was collected every 12 h, and the virus titer at each time point was determined by plaque assay. A growth curve was plotted using the mean titer values at each time point, and the associated standard deviations are displayed as error bars. Two-way analysis of variance was used for significance comparison, with a P value of <0.05 being considered significant. (D) Cytopathic effects on cells infected with del NS1 virus and mock-infected cells were documented under the microscope (D).

From the above data, it is clear that DDX3 has a critical function in virus-induced SG formation, but the effect of DDX3 downregulation and SG formation on virus replication is not known. Thus, we infected DDX3 siRNA- and off-target-siRNA-treated cells with either PR8 WT, del NS1, or SIV/SK-WT virus, a field-isolated strain (53), and examined the virus titer every 12 h. Downregulation of endogenous DDX3 expression was confirmed by subjecting the DDX3 siRNA- and off-target-siRNA-treated cells to Western blotting with antibodies against DDX3 and β-actin (Fig. 7B). Normalization of DDX3 level to the β-actin level in the same sample showed that siRNA-DDX3-treated cells demonstrated a 50% reduction in endogenous DDX3 expression compared to off-target-siRNA-treated cells (Fig. 7B). Moreover, DDX3 siRNA treatment resulted in an increased virus titer with del NS1 virus infection at all the time points tested. The peak difference, with an 11.5-fold increase in virus titer, was observed at 36 h p.i. and was found to be statistically significant (P < 0.001). The increases in virus titers observed at 24 and 48 h p.i. were also statistically significant (P = 0.0032 and 0.0017, respectively). Considering that only 50% of DDX3 expression was knocked down, the effect of DDX3 on inhibiting del NS1 virus replication may be more profound. However, DDX3 downregulation did not result in any statistically significant differences in the virus titers upon PR8-WT virus infection and SIV/SK-WT virus infection at any of the time points tested (Fig. 7C). The degree of cytopathic effect on the infected cells was also more severe in del NS1 virus-infected DDX3 knockdown cells, while DDX3 knockdown alone did not result in cytotoxicity (Fig. 7D).

DISCUSSION

DEAD box proteins are the largest helicase family in eukaryotes and are multifunctional proteins involved in various aspects of RNA metabolism, cell cycle regulation, tumorigenesis, and virus life cycle (2). Several RNA helicases, including DDX1, DDX6, RHAU, eIF4A, and DDX3 have been shown to localize in SGs (22). Among these, DDX3 has been reported to function as an essential component for SG assembly and has been reported to interact with other SG proteins, such as eIF4E and PABP1 (23). DDX3 interaction with eIF4E traps eIF4E and the associated mRNA in a translationally inactive complex, thus inhibiting cap-dependent translation (54). Formation of this inactive complex and the inhibition of translation in turn triggers SG assembly (23). In our study, we have shown that infection with del NS1 virus triggers SG formation and that DDX3 and NP localize in these SGs (Fig. 4A and C). We also observed impairment of virus-induced SG formation and increased replication of del NS1 virus upon DDX3 knockdown (Fig. 7A, B, and C). These results suggested that DDX3 functions as an antiviral protein and that DDX3 colocalization in virus-induced SGs might make a major contribution to this antiviral function. This is in agreement with previous studies which have shown that SG formation negatively affects influenza virus replication (26, 45). The antiviral function of DDX3 could be exerted in several ways. In one study, it was reported that the vRNA along with RIG-I is sequestered into virus-induced SGs and that these SGs serve as a platform for RNA detection by RIG-I (28). Because DDX3 interacts with NP, it is possible that DDX3 sequesters NP bound vRNA into SGs. Even though we did not stain for RIG-I in del NS1-infected cells, our results clearly showed that vRNA does not localize in SG containing NP (Fig. 4D). Therefore, it seems less likely that DDX3 exerts its antiviral role through facilitating vRNA sensing by RLR in SGs. Another possible mechanism of antiviral function would consist of sequestering viral mRNA into SGs through DDX3 interaction with the translation initiation factor eIF4E. During influenza virus infection, eIF4E binds viral mRNA in the cytoplasm after nuclear export and triggers recruitment of other factors like eIF4G and PABP1 to initiate translation (55). As mentioned above, DDX3 inhibits translation by binding eIF4E and locking it in a translationally inactive state in SGs. Therefore, it is conceivable that DDX3 inhibits viral mRNA translation by binding to eIF4E-viral mRNP complex, trapping it in a translationally inactive state and thereby sequestering the eIF4E-viral mRNP in the SGs. However, as the FISH assay requires a mixture of probes for the target RNA molecules, it would be difficult to differentiate cRNA from viral mRNA. Thus, additional studies are needed to confirm viral mRNA localization in SGs and to explore the contribution of DDX3-eIF4E interaction to SG nucleation. The third possible mechanism is sequestration of NP into SGs via DDX3-NP direct interaction, as observed from our IP (Fig. 1A, C, and E) and immunofluorescent staining (Fig. 4A and C) experiments. NP is a highly conserved viral protein with a primary function of encapsidating the viral genome (34). However, NP is a multifunctional protein with additional functions in nuclear and cytoplasmic trafficking of vRNA, viral polymerase activity, and the regulation of transcription and replication during the virus life cycle (56, 57). Therefore, sequestration of NP protein from the appropriate cellular compartments into SGs as observed in our study (Fig. 4C), could have a serious negative effect on viral replication. Taken together, these data suggest that DDX3-NP interaction could sequester NP into virus-induced SGs, isolating it away from the proper cellular compartments. This reduces the availability of NP to carry out its normal cellular functions and has an unfavorable effect on virus replication. Additionally, ectopic NP expression at high levels has been shown to inhibit SG formation in an eIF2α-independent manner through an unknown mechanism (27). Thus, NP sequestration in SGs could serve as a mechanism to counteract its SG-antagonistic function.

DDX3 has been reported to function as an inducer of type I IFN production by interacting with several components of the RIG-I-mediated IFN production pathway (4). More importantly, constitutively expressed DDX3 has been shown to act as a viral RNA sensor during the initial stages of infection by forming a complex with MAVS and thereby triggering IFN-β production. Many known antiviral components, such as OAS, RNase L, RIG-I, and MDA5, and several regulators of RIG-I activation, such as TRIM25, RIPLET, and MEX3C, have been shown to colocalize in SGs (10). A recent study also suggests that SGs might act as a platform for facilitating RIG-I sensing of viral RNAs during influenza virus infection (28). Considering the function of DDX3 in inducing IFN-β production and SG formation, DDX3 might play an important role in facilitating signaling mediated by RIG-I sequestered in SGs.

It has been reported that NaAs treatment induces oxidative stress and SG assembly via activating the eIF2α kinase HRI (52), while del NS1 virus infection activates the PKR kinase. NS1 is able to inhibit PKR activation but not HRI kinase activation (26, 45), which might explain the ability of NS1 to suppress virus-induced SG formation (Fig. 5A and B) and the inability of NS1 to suppress NaAs-induced SG formation (Fig. 5C). Failure of NS1 to suppress SG assembly in NaAs-treated cells was also observed by Khaperskyy et al. (27). These data provide a glimpse into the distinct pathways activated by NaAs treatment and influenza virus for SG assembly.

Studies with DDX3 truncations in inducing SG formation revealed that the core helicase domain of DDX3 alone is sufficient for SG localization during virus infection, while addition of the N-terminal domain does not affect its localization (Fig. 6). Interestingly, we observed that addition of the C-terminal domain though has a negative effect on the formation of SGs itself (Fig. 6). This prompts us to speculate that the N-terminal-deletion-containing DDX3 has a dominant negative effect on SG formation. Moreover, downregulation of DDX3 also weakens virus-induced SG formation (Fig. 7A). These results portray DDX3 as an essential factor for virus-induced SG formation, which is supportive of another study which identified DDX3 as an essential nucleating factor for oxidative stress-induced SG formation (23).

In our study, we observed that WT-virus-infected cells did not form SGs throughout the infection at all the time points tested (Fig. 4A). We identified the NS1 protein as being responsible for the SG-inhibitory function, since SGs did not form in del NS1 virus-infected cells upon NS1 expression (Fig. 5A and B). Previous studies have demonstrated the ability of NS1 to inhibit SG formation (27, 45). The NS1 protein is believed to suppress SG formation through its ability to inhibit the activation of eIF2α kinase PKR. However, formation of SGs can occur independent of eIF2α phosphorylation (eIF2α-P), upon treatment with drugs that interfere with translation initiation and when certain translation initiation factors are depleted via siRNA treatment (58–61). More specifically, SGs are induced by depletion of PABP1 and eIF4E and by preventing eIF4E association with eIF4G (61). Many viruses, such as poliovirus, herpes simplex virus (HSV), and mammalian orthoreovirus, are known to modulate SG formation independent of eIF2α-P (62–64). HSV ICP8 protein binds to G3BP, while poliovirus 3C proteinase cleaves G3BP to counteract eIF2α-independent SG formation (64, 65). Hence, an alternative NS1-mediated mechanism to inhibit SG formation downstream of eIF2α may exist. The NS1 protein of influenza virus is reported to enhance viral mRNA translation by facilitating ribosomal recruitment via interactions with viral mRNA, eIF4G, and PABP1 (66). In our study, we clearly demonstrated the existence of DDX3 and NS1 interaction (Fig. 1A, B, and D). DDX3 interaction with eIF4E prevents eIF4E-eIF4G association, thereby trapping the associated mRNA in a translationally inactive complex, inducing the formation of SGs (23). DDX3 also interacts with PABP1, which is a key component of SGs (23). We speculate that during WT influenza virus infection, besides interacting with DDX3, NS1 also associates with viral mRNA, PABP1, and eIF4F (which comprises eIF4E, eIF4G, and eIF4A). This might prevent DDX3 binding to eIF4E and PABP1. The inability of DDX3 to interact with eIF4E and PABP1 could in turn suppress SG formation, which might be an additional mechanism to explain the resistance of WT-virus-infected cells to SG formation.

In this study, we reported for the first time that DDX3 is an antiviral protein during influenza virus infection. DDX3 localizes in SGs upon virus infection and is critical for virus-induced SG formation. DDX3 is able to interact with the viral proteins NS1 and NP in infected cells; NP but not vRNA is sequestered in SGs. As mentioned in the discussion above, NP recruitment into SGs mediated through DDX3-NP interaction could contribute to DDX3's antiviral function, and NS1 interaction with DDX3 might suppress SG formation in WT-virus-infected cells. Supportive of SG-mediated DDX3 antiviral function, DDX3 downregulation suppresses SG formation and increases virus titer in del NS1-infected cells. Overall, we have discovered a novel antiviral function for DDX3, which is mediated through SG formation during influenza virus infection.