Exportin-1-Dependent Nuclear Export of DEAD-box Helicase DDX3X is Central to its Role in Antiviral Immunity

DEAD-box helicase 3, X-linked (DDX3X) regulates the retinoic acid-inducible gene I (RIG-I)-like receptor (RLR)-mediated antiviral response, but can also be a host factor contributing to the replication of viruses of significance to human health, such as human immunodeficiency virus type 1 (HIV-1). These roles are mediated in part through its ability to actively shuttle between the nucleus and the cytoplasm to modulate gene expression, although the trafficking mechanisms, and impact thereof on immune signaling and viral infection, are incompletely defined. We confirm that DDX3X nuclear export is mediated by the nuclear transporter exportin-1/CRM1, dependent on an N-terminal, leucine-rich nuclear export signal (NES) and the monomeric guanine nucleotide binding protein Ran in activated GTP-bound form. Transcriptome profiling and ELISA show that exportin-1-dependent export of DDX3X to the cytoplasm strongly impacts IFN-β production and the upregulation of immune genes in response to infection. That this is key to DDX3X’s antiviral role was indicated by enhanced infection by human parainfluenza virus-3 (hPIV-3)/elevated virus production when the DDX3X NES was inactivated. Our results highlight a link between nucleocytoplasmic distribution of DDX3X and its role in antiviral immunity, with strong relevance to hPIV-3, as well as other viruses such as HIV-1.


Introduction
DEAD-box helicase 3, X-linked (DDX3X) is a conserved ATP-dependent RNA helicase with various roles in RNA metabolism/gene expression, facilitated by localization in the cytoplasm or the nucleus. DDX3X is crucial in regulating innate antiviral immune responses initiated by the retinoic-acid-inducible gene I (RIG-I)-like receptors (RLRs) [1]. RLRs recognize cytoplasmic RNA derived from viruses such as hepatitis C (HCV), influenza A, human immunodeficiency virus type 1 (HIV-1) [2], and parainfluenza virus type 3 (hPIV-3) [3], a major cause of bronchiolitis, bronchitis, and pneumonia in children, the elderly, and immunocompromised, and a cause of significant mortality in hematopoietic stem cell transplant recipients [4,5]. Despite being an important respiratory pathogen

DNA Transfections
Plasmid transfections were performed using FuGENE HD (Promega, Madison, WI, USA) according to the manufacturer's instructions.

Co-Immunoprecipitation and Immunoblotting
HEK-293T cells were transfected to express mCherry or mCherry fusion proteins, then the mCherry positive populations were FACS-sorted and expanded. 1 × 10 7 cells were scraped into microfuge tubes, then 200 µL ice-cold co-IP buffer (20 mM Tris-Cl pH 7.4, 150 mM NaCl, 0.1% v/v IPEGAL) supplemented with 10 µg mL −1 RNaseA (Sigma-Aldrich, St. Louis, MO, USA) and cOmplete Ultra EDTA-free protease inhibitor cocktail tablets was added. Cells were briefly sonicated and clarified by centrifugation, then 100 µL Protein G-coupled magnetic resin (Thermo Fisher Scientific, Waltham, MA, USA) pre-bound to mCherry antibody was added to each supernatant. Protein-antibody complexation proceeded with end-over agitation for 30 min at 4 • C, then the resin was washed once with tris-buffered saline (TBS), transferred to clean tubes, and 50 µL 2× Laemmli sample buffer was added before incubating for 10 min at 95 • C. Samples were centrifuged at 16,100× g for 1 min and immediately subjected to SDS-PAGE (sodium dodecyl sulfate polyacrylamide gel electrophoresis) on 10% polyacrylamide gels. Proteins were transferred to PVDF (polyvinylidene difluoride) and probed using specific antibodies diluted in 5% w/v skim milk + 0.1% v/v Tween-20.

NanoString RNA Profiling
Whole-cell lysates of freshly sorted A549 cells were prepared 24 h postinfection by thoroughly washing and resuspending the cells in 50 µL CL buffer (10 mM Tris-Cl pH 7.4, 150 mM NaCl, 0.25% (v/v) IPEGAL). Cells were homogenized and RNA hybridization reactions were performed using the 770-plex Human PanCancer Immune Profiling CodeSet (NanoString Technologies, Seattle, WA, USA) with 5 µL clarified supernatant, corresponding to approximately 4000 cells in accordance with the manufacturer's instructions. The nCounter ® SPRINT system (NanoString Technologies, Seattle, WA, USA) was used to quantify captured reporter probes. Average linkage Pearson correlation heatmaps on optimally ordered data were generated using MeV software. Principal component analysis was performed using XLStat. Experimentally-validated human ISGs were interrogated using the Interferome database [24].

Enzyme-Linked Immunosorbent Assay
ELISA titrations were performed in triplicate using the sandwich method employed by the LumiKine hIFN-β kit (Invivogen, San Diego, CA, USA) in accordance with the manufacturer's instructions. Measurements were performed on a ClarioStar plate reader (BMG Labtech, Ortenberg, Germany) equipped with a liquid injector using 30 flashes per well.

Live and Indirect Immunofluorescence Microscopy
Fixed and live cell imaging was performed using a Nikon C1 inverted confocal laser scanning microscope (Monash Micro Imaging, Monash University), equipped with a CO 2 -and temperature-controlled live imaging chamber and stage, a 100× NA 1.4 oil-immersion objective, and running NIS Elements (Nikon, Tokyo, Japan) for image acquisition. Specimens were optically sliced through the maximum dimension of the nucleus using a pinhole diameter of 1.0 AU. Images were analyzed blind using Fiji/ImageJ (NIH). Pixel intensities (fluorescence) in the middle of the nucleus and cytosol were determined by sampling equally sized representative regions of interest (ROIs), free of inclusions and oversaturated pixels, as performed previously [25]. Background was calculated by defining a ROI in each image lacking cells or specific staining, and measuring the pixel intensity of an area equivalent to that used for cell sampling. This was subtracted from the nuclear and cytosolic pixel intensity values, thereby enabling the nuclear/cytoplasmic (Fn/c) ratio to be calculated. Similarly, DDX3X-exportin-1 colocalization around the nuclear membrane was determined by measuring pixel intensity (fluorescence) along a representative line bisecting the nucleocytosolic boundary as indicated.

Expression and Assembly of Exportin-1-Ran-GTP
GST-Exportin-1 and GST-Ran(Q69L) were expressed separately in E. coli BL21(DE3) cells at 16 • C following induction at OD 600nm = 0.6 with 0.5 mM IPTG. 18 h post-induction, bacteria were harvested by centrifugation and resuspended in PBS supplemented with 5 mM DTT and cOmplete protease inhibitor tablets (Sigma-Aldrich, St. Louis, MO, USA). Proteins were extracted by sonication and applied to sepharose G4B resin (GE Healthcare, Chicago, IL, USA) for GST-affinity purification, then washed and the GST-free proteins eluted by incubation with PreScission protease. Exportin-1 and Ran(Q69L) were then each applied to a Superdex 200 16/60 gel filtration column (GE Healthcare, Chicago, IL, USA) equilibrated in GF1 buffer (20 mM Tris-Cl pH 7.5, 100 mM NaCl, 5 mM MgOAc, 2 mM DTT). For production of Ran-GTP, 1 mM GTP was added to Ran(Q69L) and the complex was purified on a Superdex 200 16/60 column. The formation of the Ran-GTP complex was confirmed by absorbance at 260 nm. For binding studies using exportin-1-Ran-GTP, the complex was pre-formed by incubating equimolar amounts of exportin-1 and Ran-GTP in GF1 Buffer at 20 • C for 30 min.

Circular Dichroism
Protein circular dichroism spectra were measured in 20 mM Tris-Cl pH 8.0, 500 mM NaCl, 10% glycerol, and 0.5 mM TCEP using a J-815 circular dichroism (CD) spectrometer (Jasco, Easton, MD, USA). Spectra were recorded at 0.2 mg mL −1 between 190-250 nm in a 1 mm quartz cuvette at 20 • C. Mean ellipticity values per residue (θ) were calculated as θ = (3300 × m × ∆A)/(lcn), where l is the path length (0.1 cm), n is the number of residues, m is the molecular mass (Da), and c is the protein concentration (mg mL −1 ).

RNA-Dependent ATP Hydrolysis Assays
RNA-dependent ATP hydrolysis activity was measured using the Biomol ® Green phosphate detection kit (Enzo Life Sciences, Farmingdale, NY, USA). 200 nM RIG-I∆CARDS, DDX3X, or variants thereof were diluted in ATPase assay buffer (20 mM Tris-Cl pH 7.5, 1.5 mM DTT, 1.5 mM MgCl 2 ), 10 µM poly(I:C) (Invivogen, San Diego, CA, USA), and 20 nmol ATP (Sigma-Aldrich, St. Louis, MO, USA), then incubated for 25 min at 37 • C. Phosphate standards were serially diluted from 2 µM to 0.031 µM using 1× ATPase reaction buffer and added to the control wells. Reactions were performed in pentaplicate in a final volume of 100 µL in 96-microwell assay plates (Corning, Corning, NY, USA). Reagents were diluted using diethylpyrocarbonate (DEPC)-treated water. Following incubation, 100 µL Biomol Green reagent was added to the control and sample wells to stop the reactions. Sample absorbance was measured by absorbance at 620 nm using a ClarioStar plate reader (BMG Labtech, Ortenberg, Germany) using 30 flashes per well.

Analytical Ultracentrifugation
Sedimentation velocity experiments on wild-type DDX3X and NES mutants alone and in complex with exportin-1-Ran-GTP were performed in an Optima analytical ultracentrifuge (AUC; Beckman Coulter, Brea, CA, USA) at 20 • C. Proteins were incubated individually or together at 20 • C for 30 min prior to centrifugation in GF1 buffer (20 mM Tris-Cl pH 7.5, 100 mM NaCl, 5 mM MgOAc, 2 mM DTT). 380 µL of sample and 400 µL of reference solution (GF1 buffer) were loaded into a conventional double sector quartz cell and mounted in an An-50 Ti rotor (Beckman Coulter, Brea, CA, USA). Samples were centrifuged at 40,000 rpm and the data was collected continuously at 280 nm. Solvent density (1.041 g mL −1 at 20 • C) and viscosity (1.0149 cp at 20 • C), as well as estimates of the partial specific volume (DDX3X: 0.7215 mL g −1 , exportin-1-Ran-GTP: 0.7450 mL g −1 at 20 • C), were computed using SEDNTERP [26]. Sedimentation velocity data were fitted to a continuous size [c(s)] distribution model using SEDFIT [27].

Quantification and Statistical Analysis
Statistical parameters are reported in the figures and figure legends. Statistical analysis was performed using GraphPad Prism software. For nuclear/cytosolic fluorescence ratio measurements (Figures 1B,F, 2B and 4B,D), n represents the number of cells measured per sample and is represented as mean ± SEM, as previously [23,28,29]. Significance was calculated using Student's t-test (two-tailed) or one-way ANOVA with Tukey's, Dunnett's, or Holm-Sidak multiple comparisons post hoc analysis as indicated. For ATP hydrolysis assays ( Figure 3C), n represents the number of experimental replicates and is represented as mean ± SD. Significance was calculated using the Student's t-test with Holm-Sidak multiple comparisons post hoc analysis. For plaque assays ( Figure 5A,C) and ELISA ( Figure 5B,D), n represents the number of biological replicates and is represented as mean ± SD. Significance was calculated using one-way ANOVA with Dunnett's or Tukey's multiple comparisons post hoc analysis as indicated. For NanoString RNA profiling ( Figure 6A-C, Table S1), low (<10) count data was discarded, then the remaining data was background corrected by subtracting the maximum value of the available negative control probes and normalized to the geometric mean of 10 stable housekeeping genes across all samples, as described previously [30].

The DDX3X N-Terminus Mediates Its Nuclear Export
To investigate the exportin-1-dependent nuclear export of DDX3X, we generated plasmids for mammalian expression of HA-fused full-length DDX3X, the first 168 residues (HA-DDX3X(1-168)), or DDX3X lacking the first 168 residues (HA-DDX3X(169-662)). Subcellular localization of these proteins was analyzed in HEK-293T cells by confocal laser scanning microscopy and quantitative image analysis (qCLSM). As a control, we used a GFP fusion of the well-characterized HIV-1 Rev NES (GFP-HIV-RevNES), which is exported via exportin-1 [31]. As another control, we used a GFP fusion of the Simian virus 40 T-antigen (Tag) NLS (GFP-TagNLS), which undergoes nuclear import dependent on importin α/β1 [32]. As expected, GFP-HIV-RevNES and GFP-TagNLS localized to the cytoplasm and nucleus, respectively ( Figure 1A,B). Surprisingly, despite being only 19.5 kDa in size, and small enough in principal to passively diffuse across the nuclear pore, HA-DDX3X(1-168) was predominantly cytosolic, as per the full-length protein ( Figure 1A,B). In contrast, DDX3X lacking this region, HA-DDX3X(169-662), was localized strongly within the nucleus ( Figure 1A,B), supporting the idea that the N-terminus specifically mediates nuclear export of DDX3X [20]. Additionally, because the 55.9 kDa HA-DDX3X(169-662) truncation has very limited ability to passively diffuse through the nuclear pore, its strong nuclear accumulation is likely due to active nuclear import. Thus, DDX3X residues 1-168 and 169-662 appear to harbor, respectively, at least one NES or nuclear localization signal (NLS), and these interact specifically with one or more subcellular trafficking receptors to facilitate nucleocytoplasmic shuttling of full-length DDX3X.

DDX3X Harbors an Exportin-1 Recognized NES in the N-Terminus
To confirm the NES (hereafter termed NESα) is functional in mediating DDX3X nuclear export, and for further use as a tool to explore DDX3X function, we substituted the four hydrophobic residues to alanine (L12A/F16A/L19A/L21A, termed qmNESα), which are required for transport of other exportin-1 cargos [33]. We compared subcellular distribution of the qmNESα variant with wild-type DDX3X expressed as mCherry fusion proteins in HEK-293T cells by live-cell qCLSM. The qmNESα DDX3X variant showed significantly (p ≤ 0.0001) impaired nuclear export compared to wild-type, with almost 9-fold higher levels of nuclear accumulation ( Figure 1E,F). LMB treatment did not further increase the nuclear fluorescence signal, in stark contrast to wild-type which showed significantly (p ≤ 0.001) increased nuclear accumulation. As expected, these results confirmed the finding by Brennan et al. [20] that the DDX3X N-terminal NESα is functional in exportin-1-dependent nuclear export ( Figure 1E,F). Consistent with this result, we successfully captured the transient receptor-cargo interaction between endogenous exportin-1 and mCherry-DDX3X, but not mCherry-DDX3X(qmNESα) by co-immunoprecipitation ( Figure 1G). Collectively, these data confirm that exportin-1-mediated nuclear export of DDX3X is dependent on the N-terminal NESα of DDX3X.

DDX3X's C-Terminal Tail Is Dispensable for Nuclear Export
DDX3X residues 260-517, comprising a truncated portion of the helicase core (residues 211-575), were previously proposed to bind exportin-1 without dependence on a modular NESor Ran-GTP [17,21,22]. Additionally, DDX3X C-terminal residues 536-662 have been reported to mediate nuclear export by nuclear RNA export factor 1 (NXF1/TAP) [18]. To test these possibilities, we generated mCherry-fused DDX3X lacking the NXF1-binding region but harboring the wild-type NESα, termed DDX3X(1-535). Using live-cell qCLSM, we found the subcellular distribution of this protein was identical to full-length (Figure 2), indicating the C-terminal tail is dispensable for DDX3X's subcellular trafficking. Next, we introduced our qmNESα mutations into this truncated construct, termed DDX3X(1-535)(qmNESα), to test the contribution of any NES-independent binding of exportin-1. As expected, this protein was localized in an identical manner to full-length DDX3X(qmNESα) (Figure 2). Collectively, these data suggest DDX3X's bulk nuclear export occurs via exportin-1, and that this is mediated by the N-terminal NESα-sequence.

DDX3X's C-Terminal Tail is Dispensable for Nuclear Export
DDX3X residues 260-517, comprising a truncated portion of the helicase core (residues 211-575), were previously proposed to bind exportin-1 without dependence on a modular NES-or Ran-GTP [17,21,22]. Additionally, DDX3X C-terminal residues 536-662 have been reported to mediate nuclear export by nuclear RNA export factor 1 (NXF1/TAP) [18]. To test these possibilities, we generated mCherry-fused DDX3X lacking the NXF1-binding region but harboring the wild-type NESα, termed DDX3X(1-535). Using live-cell qCLSM, we found the subcellular distribution of this protein was identical to full-length (Figure 2), indicating the C-terminal tail is dispensable for DDX3X's subcellular trafficking. Next, we introduced our qmNESα mutations into this truncated construct, termed DDX3X(1-535)(qmNESα), to test the contribution of any NES-independent binding of exportin-1. As expected, this protein was localized in an identical manner to full-length DDX3X(qmNESα) (Figure 2). Collectively, these data suggest DDX3X's bulk nuclear export occurs via exportin-1, and that this is mediated by the N-terminal NESα-sequence.

DDX3X Nuclear Accumulation
nctional significance of exportin-1-dependent nuclear export of DDX3X in in the context of invasive RNA. To determine whether the subcellular nges in correlation with immune stimulation, we challenged HeLa cells by thetic double-stranded RNA analog poly(I:C) and then examined the f endogenous DDX3X by qCLSM. Strikingly, poly(I:C) caused rapid rom the cytosol to the nucleus, with significantly (p ≤ 0.0001) increased (~2n observed 6 h post-stimulation, with levels of nuclear protein remaining (Figure 4a-b). To test whether the same effects were induced by an RNA used hPIV-3, the most virulent hPIV subtype for respiratory illness [34],   For (D-I) residuals from the c(s) distribution best fit plotted as a function of radial distance from the axis of rotation are displayed above. The presence or absence of larger-sedimenting species corresponding to complex formation is indicated by black arrows. See also Table 1.

Invasive RNA Triggers DDX3X Nuclear Accumulation
We next probed the functional significance of exportin-1-dependent nuclear export of DDX3X in innate immune signaling in the context of invasive RNA. To determine whether the subcellular distribution of DDX3X changes in correlation with immune stimulation, we challenged HeLa cells by transfection with the synthetic double-stranded RNA analog poly(I:C) and then examined the subcellular distribution of endogenous DDX3X by qCLSM. Strikingly, poly(I:C) caused rapid redistribution of DDX3X from the cytosol to the nucleus, with significantly (p ≤ 0.0001) increased (~2-fold) nuclear accumulation observed 6 h post-stimulation, with levels of nuclear protein remaining constant for at least 24 h ( Figure 4A,B). To test whether the same effects were induced by an RNA virus infection model, we used hPIV-3, the most virulent hPIV subtype for respiratory illness [34], and A549 human alveolar epithelial cells. Indeed, hPIV-3 infection significantly (p ≤ 0.0001) increased (~2-fold) the nuclear localization of ectopically expressed DDX3X ( Figure 4C,D). Notably the magnitude of DDX3X relocalization between poly(I:C) stimulation and virus infection was identical (~2-fold), suggesting a specific response to invasive RNA. In addition, hPIV-3 infection induced accumulation of DDX3X into cytosolic inclusions in some cells, possibly p-bodies or stress granules typically associated with translational regulation of cellular or viral RNA. HeLa cells showed identical results (data not shown). These results imply that nuclear redistribution of DDX3X may be a general, acute-phase cellular response to viral challenge, arising as a specific cellular response to invasive RNA.

Overexpression of Wild-Type But Not Nuclear Export Defective DDX3X Can Protect Against hPIV-3 Infection
To dissect the role of exportin-1-mediated nuclear export of DDX3X in regulating immune signaling events in the nucleus and cytosol, we infected A549 cells expressing either mCherry-DDX3X, mCherry-DDX3X(qmNESα), or mCherry alone with hPIV-3, and then measured viral replicative fitness using plaque assays. Strikingly, cells overexpressing mCherry-DDX3X were significantly (p ≤ 0.01) more resistant to infection than those expressing mCherry alone, with almost a 10-fold reduction in infectious virus production as measured by plaque assay ( Figure 5A). This is consistent with the idea that DDX3X plays an important antiviral role. In stark contrast, cells overexpressing mCherry-DDX3X(qmNESα) were substantially more susceptible to infection, with 200-fold higher levels of virus production (p ≤ 0.01) than those expressing wild-type DDX3X, strongly indicating that DDX3X's ability to undergo nuclear export through the exportin-1-recognized NESα is key to its antiviral activity. Parallel monitoring of production of IFN-β by ELISA in response to infection indicated that DDX3X(qmNESα)-expressing cells secreted significantly (p ≤ 0.001) more IFN-β (~2-fold) than those expressing wild-type DDX3X ( Figure 5B). These results suggest that the increased hPIV-3 titer observed in nuclear export defective DDX3X-expressing cells is not due to a general defect in IFN-β production during hPIV-3 infection, and that increased IFN-β production is insufficient to inhibit hPIV-3 replication. These results strongly imply that DDX3X's antiviral role in hPIV-3 infection is dependent on its nuclear export/nuclear trafficking ability.

Exportin-1 Is Important to hPIV-3 Replication
Even though RNA viruses such as paramyxoviruses replicate entirely in the host cytosol, inhibition of exportin-1 by LMB has been reported to inhibit virus production in the case of Hendra virus [23], RSV [35], and Venezuelan equine encephalitis virus [36], suggesting their replication is facilitated by exportin-1-dependent nuclear export. Since hPIV-3 also replicates entirely in the cytoplasm, we tested the importance of exportin-1 mediated nuclear export by treating hPIV-3-infected cells with LMB, again monitoring virus production and IFN-β as above. Controlling for limited cytotoxic effects, we found a dose-dependent reduction in both hPIV-3 titer and IFN-β secretion with increasing LMB concentration ( Figure 5C,D), again consistent with the importance of exportin-1-dependent nuclear export of host/viral factors being central to hPIV-3 virus production fitness, as opposed to IFN-β levels.

DDX3X's Nuclear Trafficking Potentiates Immune Gene Induction
Since IFN-β production in response to hPIV-3 infection did not appear to be impaired by inactivation of DDX3X nuclear export, we hypothesized that altered expression of antiviral genes besides IFNB1 might be responsible for the effects on infection observed in Figure 4A. To address this directly, we profiled host gene transcription using the NanoString nCounter ® SPRINT system. We transfected A549 cells to express mCherry-fused DDX3X, mCherry-fused DDX3X(qmNESα), or mCherry alone, then sorted the mCherry-expressing populations and assayed mRNA transcript levels 24 h post-hPIV-3 or mock infection. Transcript levels were monitored using the PanCancer Immune Profiling RNA probe library. After internal normalization and discarding low-count data, we measured transcription across a total of 730 human genes relevant to immunity and cancer (Table S1). The vast majority of genes were downregulated in uninfected cells expressing mCherry-DDX3X compared to mCherry alone ( Figure 6A), suggesting DDX3X may act as a 'brake' on immune genes at steady-state. Consistent with this idea, many of the genes were upregulated in uninfected cells expressing mCherry-DDX3X(qmNESα), implying that DDX3X's ability to traffic between the nucleus and cytoplasm, dependent on its exportin-1 recognized N-terminal NESα, is central to this function. As expected, viral infection resulted in strong activation of many of these genes in cells expressing wild-type, but not in cells expressing nuclear export defective, DDX3X. This is reflected in the distant clustering of the wild-type DDX3X samples between steady-state and infection, as opposed to the much closer clustering of the DDX3X(qmNESα) samples in the absence or presence of infection ( Figure 6A. See also principal component analysis in Figure 6B). The data show that the anti-hPIV-3 inflammatory response in lung tissue is overwhelmingly characterized by the induction of IFN-β and ISGs including proinflammatory cytokines and chemoattractants for neutrophils (e.g., CXCL1, CXCL2, CXCL3, IL1A, IL6, IL8, PTGS2, and SAA1) and T-cells (e.g., CCL5, CCL20, CXCL10, CXCL11, IL6, and IL8), as well as innate immune signaling proteins (e.g., MX1, IFI27, IFIT1, IFIT2, IRF7, ISG15, ISG20, STAT1, and TLR8) and inducers of apoptosis (e.g., IL1B, IFI27, and IFIT2) ( Figure 6C).
Comparison of the mRNA levels for cells ectopically expressing DDX3X with or without a functional NES revealed clear differences in the subsets of ISGs expressed. In resting cells, ectopic expression of DDX3X(qmNESα) resulted in increased mRNA levels of 523 genes ( Figure 6D). Only 173 of these showed similar effects upon overexpression of wild-type DDX3X. There were an additional set of 49 genes, distinct from those impacted by DDX3X(qmNESα), showing elevated levels upon overexpression of wild-type DDX3X ( Figure 6D). Upon hPIV-3 infection, wild-type DDX3X-expressing cells distinctly upregulated 192 genes, whereas only 85 were distinctly upregulated in DDX3X(qmNESα)-expressing cells ( Figure 6D). Overall, these data highlight that nucleocytoplasmic trafficking of DDX3X is critically important in regulating gene induction during viral infection, with elevated nuclear expression of DDX3X impacting the resting state transcriptome as well as that in response to viral infection.

Nuclear DDX3X Contributes to IFNB1 Transcription and Influences ISG Subset Induction
Chromatin-immunoprecipitation experiments indicate DDX3X can associate with the IFNB1 promoter [10]. Our observation that expression of nuclear-localizing DDX3X(qmNESα) led to elevated levels of IFN-β secretion in response to hPIV-3 infection compared to wild-type DDX3X ( Figure 5B) correlated nicely with the fact that a large number (217) of the genes upregulated upon overexpression of DDX3X(qmNESα) were ISGs, including IFNB1 itself. For the latter, hPIV-3-infected A549 cells expressing either DDX3X or DDX3X(qmNESα) showed enhanced IFNB1 transcription versus mCherry alone (normalized induction of 0.924 and 0.942 versus 0.870, respectively), with DDX3X(qmNESα) showing the greatest overall induction (Table S1). Consistent with the idea that DDX3X's nucleocytosolic distribution modulates its role as a brake on immune induction at rest, uninfected cells overexpressing DDX3X showed lower IFNB1 transcription than cells overexpressing mCherry only, whereas cells overexpressing DDX3X(qmNESα) once again showed enhanced IFNB1 transcription (normalized induction of -0.979, -0.909, and -0.847, respectively) ( Table S1). The nuclear trafficking of DDX3X thus appears to modulate IFNB1 gene transcription, modulated by exportin-1 binding to the DDX3X N-terminal NESα in a Ran-GTP-dependent manner.
To further validate the above results, we examined protein expression levels of a subset of the above genes in addition to IFNB1, representing a broad range of cellular pathways and antiviral defenses. Changes in transcriptional activity of CASP3, RIPK2, LAMP2, and TBK1 ( Figure 6E) were reflected in corresponding changes in expression of encoded proteins as determined by immunoblot ( Figure 6F), giving confidence that our overall dataset for IFN-1/ISG induction/expression is robust.

Discussion
DDX3X is a key host cellular factor in the RLR signaling cascade and is implicated in the replication strategy of a large and growing list of evolutionarily divergent pathogens of significance to human health, including hepatitis B virus [37], hepatitis C virus [38], influenza A virus [39], Japanese encephalitis virus [40], West Nile virus [41], dengue virus [42], and HIV-1 [17]. Understanding the link between DDX3X subcellular localization and the host-and pathogen-directed roles of DDX3X are central to unlocking novel strategies to target DDX3X activity in infection.
Consistent with a NES-dependent interaction, we and others [8,[17][18][19][20] have shown that DDX3X's nuclear export is inhibited by the exportin-1-specific inhibitor LMB, which blocks cargo protein binding and trafficking by covalently modifying the NES-binding interface of exportin-1 [15,16]. Correspondingly, our qCLSM, co-immunoprecipitation, and analytical ultracentrifugation sedimentation velocity data confirms DDX3X harboring a nonfunctional NES is incapable of binding exportin-1 even at supraphysiological concentrations, and reciprocally, Ran-GTP is strictly required for DDX3X binding to exportin-1. Previously, DDX3X's nuclear export was attributed to a unique exportin-1-dependent mechanism requiring DDX3X helicase domain residues 260-517, but neither a recognized exportin-1 NES nor the Ran-GTP gradient [17]. However, consistent with Brennan et al. [20] our results do not support this finding. Collectively, we confirm the mechanism of DDX3X's exportin-1-dependent nuclear export is typical of other receptor-cargo interactions and aligns with that of An3, the Xenopus laevis orthologue of DDX3X, which shares 87% sequence identity overall and an identical NES within the N-terminus of DDX3X [43]. Notably, the key hydrophobic residues of the DDX3X/An3 NES are also conserved down to the Saccharomyces cerevisiae orthologue Ded1p, which also undergoes exportin-1-mediated nuclear export in a NES-and Ran-GTP-dependent manner [44].
Although exportin-1 is an exporter of DDX3X we observe residual DDX3X in the cytosol following inactivation of the NES or LMB treatment. One explanation may be that DDX3X utilizes other nuclear export pathways in addition to the exportin-1 pathway. The C-terminal region of DDX3X was previously reported to mediate binding and nuclear export by NXF1 [18]. We did not observe any contribution of the DDX3X C-terminal region to its nucleocytosolic distribution in this study, but cannot formally exclude the possibility that other nuclear export receptors may bind and traffic DDX3X in certain circumstances. We propose it is equally plausible that the nuclear import of DDX3X is weak at steady-state, but is then enhanced during specific events or stages of the cell cycle [20]. Importantly, the nuclear import mechanism of DDX3X remains unknown and warrants further study. Previous studies [20], as well as our own, have only implicated regions involved in the nuclear import of DDX3X.
Our study suggests invasive RNA is a trigger for authentic nuclear accumulation of DDX3X, which supports IFN-β induction and secretion. Nearly all proinflammatory genes strongly activated during hPIV-3 infection were positively associated with expression of wild-type DDX3X, while IFN-β itself and a particular subset of IFN-I signaling and effector genes were more strongly expressed when DDX3X accumulated more strongly in the nucleus. This reveals that regulated trafficking of DDX3X between the nucleus and cytosol is crucial for controlling IFN-β levels, at least in response to hPIV-3 infection, as well as supporting transcription of a particular subset of IFN-I signaling and effector genes in order to amplify the IFN-I response.
While we do not exclude the possibility that endogenous DDX3X expression levels may play a role, we propose the following model of DDX3X trafficking-dependent immune regulation based on our observations. In the resting state, DDX3X acts as a 'brake' on immune gene induction to prevent unnecessary immune activation. However, upon exposure to invasive RNA during acute-phase virus infection, DDX3X supports cytosolic signaling events leading to IFN-I expression, redistributing to the nucleus to help drive transcription contributing to IFN-I mediated immunity and T-cell recruitment/activation. Notably, the nuclear export of DDX3X via exportin-1 is critical for maximal gene induction, and thereby presumably results in a more effective innate and adaptive immune response to infection, and as demonstrated in our hPIV-3 infectious model.
Our results indicate that DDX3X plays a hitherto unrecognized antiviral role in hPIV-3 replication that is contingent upon its export into the cytosol, and seemingly independent of its role in IFN-β induction. Consistent with this finding, IFN-α and type III IFN (IL29A, IL-28A and/or IL28B), as opposed to IFN-β, are reported to have anti-hPIV-3 action [45,46], and type III-IFN receptor deficiencies increase susceptibility to hPIV-3 infection [47]. Despite nuclear export-deficient DDX3X being permissive to hPIV-3 replication, LMB treatment, which blocks the nuclear export of all exportin-1 cargos, including DDX3X, suppressed hPIV-3 replication. This suggests that the nuclear export of unknown host and/or hPIV-3 viral proteins plays a pivotal role in hPIV-3 replication, and that compounds such as LMB specifically targeting exportin-1 in this context may be effective against hPIV-3, as reported for other viruses [23,36,48].
DDX3X subcellular localization is central to its function in antiviral immunity and hence paramount to the infectivity of microbes that exploit DDX3X as an essential host cofactor. For example, HIV-1 Rev requires nuclear DDX3X to export HIV-1 transcripts to the cytoplasm [17], whilst cytoplasmic DDX3X is required for RSV M2 translation [49]. This suggests that host-orientated agents that alter the nuclear import/export of DDX3X are likely effective antiviral agents. Indeed, inhibitors of exportin-1, such as those developed by Karyopharm ® Therapeutics [50], that inhibit nuclear export of all cargoes recognized by exportin-1 bearing a NES can be efficacious broad-spectrum antivirals (e.g., against RSV and influenza infection). Accordingly, LMB treatment inhibited hPIV-3 replication in the current study, and Hendra virus [23], RSV [35], and Venezuelan equine encephalitis virus [36] in previous studies, suggesting their replication requires exportin-1-dependent nuclear export. However, there are currently no cargo-specific nuclear export inhibitors, which are critically important in reducing the cytotoxic effects of global inhibition of exportin-1. We anticipate our work exploring the exportin-1 mediated nuclear export of DDX3X and understanding its functional relevance in directing antiviral immune signaling outcomes will support the pursuit of DDX3X-specific nuclear export inhibitors that will have implications for viruses of significance to human health such as HIV-1 and RSV.