Antagonism of STAT3 signalling by Ebola virus

Many viruses target signal transducers and activators of transcription (STAT) 1 and 2 to antagonise antiviral interferon signalling, but targeting of signalling by other STATs/cytokines, including STAT3/interleukin 6 that regulate processes important to Ebola virus (EBOV) haemorrhagic fever, is poorly defined. We report that EBOV potently inhibits STAT3 responses to interleukin-6 family cytokines, and that this is mediated by the interferon-antagonist VP24. Mechanistic analysis indicates that VP24 effects a unique strategy combining distinct karyopherin-dependent and karyopherin-independent mechanisms to antagonise STAT3-STAT1 heterodimers and STAT3 homodimers, respectively. This appears to reflect distinct mechanisms of nuclear trafficking of the STAT3 complexes, revealed for the first time by our analysis of VP24 function. These findings are consistent with major roles for global inhibition of STAT3 signalling in EBOV infection, and provide new insights into the molecular mechanisms of STAT3 nuclear trafficking, significant to pathogen-host interactions, cell physiology and pathologies such as cancer.

Here, we aimed to examine the effect of EBOV on STAT3 responses, showing for the first time that EBOV VP24 antagonises STAT3 using a combination of mechanisms analogous to and distinct from that used for STAT1, to inhibit both STAT3 homodimers and heterodimers. We further reveal that the STAT3 complexes use distinct mechanisms for nuclear accumulation, apparently necessitating VP24's multipronged strategy.

EBOV inhibits STAT3 responses
Despite likely roles in EBOV infection for dysregulation of cytokines/STATs other than IFN/ STAT1/2, antagonism of other STATs by EBOV remains unresolved. To determine whether EBOV affects STAT3, we infected COS7 cells with EBOV before treatment with OSM, commonly used to analyse IL-6 family cytokine/STAT3 responses [18,20,26], and analysis of STAT3 localisation by immunofluorescence staining and confocal laser scanning microscopy (CLSM; Fig 1A). In mock-infected cells, STAT3 was diffusely localised between the nucleus and cytoplasm of resting cells, with nuclear accumulation clearly observed following OSM treatment, as expected. In EBOV-infected cells, however, OSM-dependent STAT3 nuclear accumulation was inhibited, with quantitative image analysis confirming a significant decrease in nucleocytoplasmic localisation in EBOV-infected compared with mock-infected cells (Fig 1A and 1B). To exclude possible impact by virus-induced type-I IFN, we confirmed that EBOV also antagonises STAT3 responses in Vero cells, which do not produce IFN (Fig 1A and  1B). Notably, in infected cells, we observed accumulation of STAT3 into large, distinct cytoplasmic regions (zoom images, Fig 1A). Co-staining for EBOV nucleoprotein (NP) using different monoclonal (Fig 1C) or polyclonal (S1 Fig) antibodies indicated that these regions correspond to cytoplasmic viral replication/inclusion bodies [30]. This concentration/colocalization of STAT3 with cytoplasmic viral inclusions is, to our knowledge, the first such observation for any virus, and suggests that STAT3 is accumulated into these virus-induced structures.

Stimulation of cells with OSM before infection inhibits EBOV replication
The finding that EBOV inhibits STAT3 responses suggests that STAT3-dependent cytokine signalling is likely to have antiviral functions toward EBOV, analogous to reports for hepatitis C virus [31], herpes simplex virus-1 [32], coxsackievirus B3 [33], influenza virus and vaccinia virus [34]. To examine this, we assessed the effect of OSM stimulation of Vero cells on EBOV replication by treatment of cells with OSM 24 h pre-infection or 24 h post-infection. Analysis using RT-qPCR for EBOV NP [35] indicated that OSM treatment pre-infection inhibits replication (Fig 2A). This suggests that OSM/STAT3 signalling can induce an antiviral state which impedes subsequent viral infection, resulting in a lag in replication. No comparable effect was detected following stimulation of cells after establishment of infection (Fig 2B), consistent with antagonism of antiviral signalling by EBOV, resulting in a less potent effect of OSM on replication.
The above data clearly indicated that VP24 has autonomous activity in antagonising STAT3 responses, independent of other viral components. However, our observation that STAT3 appears to accumulate into NP-enriched inclusion bodies in EBOV-infected cells (Fig 1), combined with the observation that VP24 and NP interact [39], suggested that NP/inclusions might enhance or otherwise regulate VP24-mediated antagonism. We thus assessed STAT3 responses to OSM in cells expressing VP24 with and without NP (S3 Fig). In OSM-treated cells expressing NP and GFP (control), STAT3 retained clear capacity to accumulate into the nucleus, and this did not differ significantly from cells expressing GFP alone. In cells expressing VP24, STAT3 responses were significantly inhibited to a similar extent in cells expressing or not expressing NP. These data confirm a principal role for VP24 in antagonism of STAT3, and indicate that NP does not significantly contribute to (or impede) this function. Thus, -infection with or without OSM (10 ng/ml, 15 min) before fixation, immunofluorescent staining for STAT3 (green), and analysis by CLSM. DAPI (blue) was used to localise nuclei. Representative images are shown. Arrowheads indicate accumulation of STAT3 in cytoplasmic regions; indicated regions in merged images are expanded in panels below (Zoom). (B) Images such as those shown in A were analysed to calculate the nuclear to cytoplasmic fluorescence ratio (Fn/c) for STAT3 (mean ± SEM, n � 70 cells for each condition). Statistical analysis (Student's t-test) was performed using GraphPad Prism software; ���� , p < 0.0001; No add., no addition. (C) E6 Vero cells were infected and treated as in A, before fixation, immunofluorescent staining for STAT3 (green) and EBOV NP (using monoclonal anti-NP, red), and analysis by CLSM. DAPI (blue) was used to localise nuclei. Representative images are shown. Arrowheads indicate colocalization of STAT3 and NP in discrete cytoplasmic regions/inclusions. Scale bars, 30 μm.
https://doi.org/10.1371/journal.ppat.1009636.g001 VP24 is sufficient to effect STAT3 antagonism in the absence of NP or viral infection. However, given the apparent accumulation of STAT3 into large cytoplasmic viral inclusions in infected cells (Fig 1), we cannot rule out that this contributes to efficient STAT3 antagonism by VP24 during infection, potentially involving other viral components. Our observations in this respect are analogous to reports of sequestration of STAT1, STAT2 and other antiviral signalling molecules (e.g. TANK-binding kinase 1, IFN regulatory factor 3) in cytoplasmic inclusions formed by other negative-sense RNA viruses [16,17,40,41]. However, given the diverse functions attributed to STAT3, including a number of reports that STAT3 can have pro-viral rather than antiviral functions depending on the specific virus, cell type and other parameters [42], it is possible that recruitment of STAT3 to inclusion bodies plays distinct roles in EBOV infection. These possibilities will form the basis of future research.
To confirm the functional outcome of VP24-mediated inhibition of STAT3 nuclear accumulation, we next analysed OSM-dependent signalling using a luciferase reporter gene assay [18,43], in which the luciferase gene is under the control of a STAT3-dependent promoter (m67). This indicated that VP24 effects significant suppression (c. 70% reduction) of OSM/ STAT3 signalling in HEK293T and U3A cells (Fig 4D; upper panel); RT-qPCR analysis confirmed that VP24 can inhibit OSM-induced expression of the STAT3-dependent socs3 gene (S4 Fig). Mumps virus V-protein (MUV-V, used as a positive control in our assays) induces STAT3 degradation to suppress IL-6 signalling [16]. We confirmed that MUV-V inhibits STAT3 responses and that this correlates with reduced levels of STAT3 expression in cell lysates. Since no similar effect was observed on STAT3 expression in VP24-expressing cells ( Fig 4D; lower panel), it appeared that VP24 uses a different antagonistic mechanism.
Intriguingly, however, co-immunoprecipitation assays in U3A cells indicated that VP24 does not affect Kα1-pY-STAT3 interaction (Fig 5A and 5B), despite clear impact on STAT3 responses in these cells (Fig 4). It has been suggested that karyopherin interactions of STAT homo-and heterodimers might differ [24,25]. Our data support this idea, providing evidence that the association of Kα1 with STAT3-STAT1 heterodimers differs from its association with STAT3 homodimers, as VP24 binding to Kα1 competes with the former, but not the latter interaction. The competitive nature of binding of VP24 and STAT1 to Kα1 results from the binding of both proteins at a non-classical STAT1/VP24-binding site. Our data thus suggest that STAT3-STAT1 heterodimers bind at, or proximal to, this non-classical site and so can be displaced by VP24 (Fig 5A and 5B). Since STAT1 homodimers and STAT1-STAT2 heterodimers also bind to this site [10], this might represent a common interface for STAT1-containing complexes, such that competitive binding by VP24 would be likely to occur for heterodimers activated by other cytokines/mediators (e.g. STAT4-STAT1 heterodimers). However, it would appear that STAT3 homodimers bind in a different fashion, possibly at a distinct site such as in a conventional cargo binding region, resulting in a lack of competition since VP24 binds elsewhere. Identification of the site in Kα1 that mediates binding to STAT3 homodimers, and other molecular details such as whether the binding requires additional host factors, will form the focus of future research. Nevertheless, since the data from U3A cells indicate that STAT3 homodimers bind to Kα1 via a site not competitively bound by VP24, it appears that an alternative mechanism is required to antagonise signalling by these complexes.

VP24 does not inhibit STAT3 binding to DNA
Reports supporting constitutive nuclear trafficking of STAT3 suggest that STAT3 accumulates in the nucleus in response to cytokine activation due to intra-nuclear interactions/sequestration, such as through induced DNA binding [25]. We therefore considered that VP24 may inhibit STAT3 nuclear accumulation in U3A cells by inhibiting the capacity of STAT3 to bind (red) and CLSM (A) to determine the Fn/c for STAT3 (B; mean ± SEM, n � 36 cells for each condition; results are from a single assay representative of two independent assays). Filled and unfilled arrowheads indicate cells with or without, respectively, detectable expression of the transfected protein. Scale bars, 30 μm. MUV V, Mumps virus V protein. (C) Lysates of COS7 and U3A cells were analysed by immunoblotting (IB) for STAT1 and STAT3. (D) upper panel: HEK293T or U3A cells co-transfected with m67-LUC and pRL-TK plasmids, and plasmids to express the indicated proteins, were treated 16 h posttransfection with or without OSM (10 ng/ml, 8 h) before determination of relative luciferase activity (mean ± SEM; n = 3 independent assays); lower panel: cell lysates used in a representative assay were analysed by IB using antibodies against the indicated proteins. Statistical analysis used Student's t-test; ��  to DNA, similar to the antagonistic mechanism of RABV P-protein for STAT1, where the Pprotein binds proximal to or within the STAT1 DNA binding domain [44,45]. To assess DNA binding by STAT3, we performed electrophoretic mobility shift assay (EMSA) analysis of cell lysates using the m67 probe (Fig 6), which is a high affinity variant of the sis-inducible element

PLOS PATHOGENS
Antagonism of STAT3 signalling by Ebola virus from the c-fos gene, commonly used to analyse STAT3-DNA binding [46][47][48]. OSM induced clear DNA binding of both endogenous and over-expressed STAT3 in the absence and presence of VP24, with VP24 having no evident inhibitory effect. Thus, the principal mechanism of antagonism does not appear to involve a direct hindrance of STAT3-DNA interaction.

VP24 interacts with STAT3, independently of VP24-karyopherin binding
While there is substantial evidence that VP24 mediates antagonism of STAT1 by competitive binding to karyopherins [12][13][14][15], recombinant purified VP24 and STAT1 proteins were reported to interact in vitro [49], suggesting that direct VP24-STAT1 binding may also contribute to antagonism. However, immunoprecipitation of VP24 expressed in mammalian cells and analysis of co-precipitated proteins by immunoblotting (IB) or mass spectrometry did not detect a VP24-STAT1 interaction [15,50,51], so there is currently a lack of clear evidence that this is significant to STAT1 antagonist function. Nevertheless, since many IFNantagonists inhibit STATs through direct or indirect physical interaction [5], we tested whether VP24 can bind to STAT3. Endogenous and transfected STAT3 co-precipitated with VP24 from U3A cells (Fig 7A and 7B), and reciprocal immunoprecipitation via STAT3 confirmed an association (S6 Fig). Thus, VP24 can interact with STAT3 independently of STAT1, consistent with data for antagonism of OSM/STAT3 signalling (Fig 4). We also confirmed coprecipitation of STAT3 with VP24 from HEK293T and COS7 cells (Figs 7C and S10). While the association of VP24 and STAT3 is clearly specific compared with controls, the amount of STAT3 detected in immunoprecipitates of VP24 appears relatively low, potentially reflecting a transient interaction and/or poor retention of the complex during cell lysis and immunoprecipitation. Since immunoprecipitation from cells does not differentiate direct and indirect interactions, it is also possible that VP24-STAT3 interaction is mediated via other cellular components, which may result in some dissociation during lysis/immunoprecipitation. pY-STAT3 co-precipitated with VP24 exclusively from OSM-treated cells as expected since pY-STAT3 is induced by OSM. However, IB for total STAT3 (detecting both non-phosphorylated and phosphorylated forms), indicated comparable co-precipitation with VP24 from cells treated with or without OSM (Figs 7 and S10). Thus, VP24 appears to interact with both the phosphorylated and non-phosphorylated STAT3. To confirm this, we examined the association of VP24 with STAT3-R609Q/Y705F, which contains mutations at the pY site (Y705F) and SH2 domain (R609Q), preventing cytokine-induced phosphorylation and dimerisation [52]. FLAG-tagged STAT3-R609Q/Y705F and FLAG-tagged wild-type (WT) STAT3 co- To further investigate the antagonistic mechanism, we analysed a karyopherin-binding deficient VP24 protein (VP24 MUT), wherein mutations of key residues at the VP24:karyopherin interface (L201A/E203A/P204A/D205A/S207A) strongly impair karyopherin binding and STAT1/IFN antagonism [15]. The effect of the mutations in inhibiting STAT1-antagonist function was confirmed using a STAT1/2-IFN-dependent luciferase reporter assay (using pIS-RE-LUC plasmid), which indicated an almost nine-fold increase in luciferase activity in IFNα-treated HEK293T cells expressing mutated protein compared with WT protein (Fig 8A; left  panel). Analysis using the STAT3/OSM-dependent signalling assay (using m67-LUC plasmid) in U3A cells showed no significant impact of the mutations on VP24 inhibitory activity ( Fig  8A; middle panel), indicating that specific antagonism of STAT3 by VP24 is independent of karyopherin-binding activity, consistent with the lack of an effect of VP24 on Kα1-STAT3 association in U3A cells. Assays of STAT3/OSM-dependent signalling in HEK293T cells, however, indicated some dependence on karyopherin-binding, probably reflecting a contribution to signalling by STAT3-STAT1 heterodimers (Fig 8A; right panel), consistent with the capacity of VP24 to compete with STAT1 and STAT3 for Kα1 binding in these cells. Thus, it appears that, in contrast to STAT1 (and STAT3-STAT1 heterodimers), antagonism of signalling by STAT3 homodimers is independent of VP24-karyopherin binding. Consistent with this, the mutations had no evident effect on VP24-STAT3 interaction in U3A cells (Fig 8B). Taken together, these data indicate that the nuclear trafficking mechanisms of STAT1 and STAT3 are distinct, and, accordingly, antagonism by VP24 uses different mechanisms, likely including competition with STAT1-containing complexes for karyopherin binding, as well as physical interaction with STAT3, which effects cytoplasmic localisation. This is perhaps consistent with reports that STAT3 nuclear import can be mediated by multiple karyopherins including non-NPI-1 karyopherins, such as Kα4 [24,27,28], which may have necessitated the development in VP24 of a distinct, karyopherin-independent strategy to antagonise STAT3.
Taken together, our data indicate that VP24 can inhibit STAT3 signalling by two distinct mechanisms, consistent with important roles for global shutdown of STAT3 in EBOV infection. Since OSM stimulation of cells can impair viral replication, it appears that STAT3 antagonism forms part of the viral strategy to counteract antiviral processes in the infected cell. However, given the established roles for STAT3 in regulating the broader immune response and other diverse cellular and physiological processes, it is possible that STAT3 antagonism is also related to pathogenic processes such as the dysregulation of inflammation, coagulation and mucosal wound healing, observed during EBOV infection [6,[21][22][23]53,54]. Recent reports indicate that STAT3 antagonism by MUV is associated with neurovirulence in vivo [55], and suppression of IL-6 signalling by influenza A virus early in infection contributes to a cytokine storm implicated in disease severity [56]. Interestingly, although the IFN-antagonist VP40 of the filovirus Marburg virus does not specifically target STATs, it inhibits upstream kinases resulting in inhibition of activation of both STAT1 and STAT3 [57]. Together these data indicate that potent suppression of STAT3 responses by filoviruses may contribute to excessive inflammatory responses associated with severe haemorrhagic fever. The apparent importance of STAT3 targeting to filoviruses, and previous reports of roles in infection by paramyxoviruses and rhabdoviruses, also indicates that specific and direct antagonism of STAT3 is important to diverse pathogens in the order Mononegavirales [16][17][18][19][20]. These data suggest that virus-STAT3 interactions could provide new directions for the development of antivirals for diverse pathogens.
Beyond the implications for viral infection, the study also provides, to our knowledge, the first clear indication of distinct nuclear import strategies for STAT3 homodimers and heterodimers, potentially accounting for the contrasting trafficking models previously proposed [24][25][26][27][28]. Together with our recent finding that lyssaviruses differentially target STAT3 dimers to modulate cytokine-induced transcription [20], this supports the idea that these complexes have distinct roles in signalling by STAT3, a pleiotropic molecule important to processes including cancer, development and immunity.

Plasmids and cell culture
Constructs to express EBOV-VP24 and MUV-V fused to GFP were generated by PCR amplification from pCAGGS-FLAG-VP24 (kindly provided by Christopher Basler, Georgia State University) and MUV V-FLAG (a gift from Curt Horvath [16], Addgene plasmid #44908), and cloning into the pEGFP-C1 vector C-terminal to GFP (Clontech). Constructs to express mCherry-or FLAG-tagged STAT3-WT were kind gifts from Marie Bogoyevitch (University of Melbourne), and the constructs to express FLAG-tagged Kα1 and Kα4 were a kind gift from Christopher Basler (Georgia State University). Primers/constructs to express FLAG-tagged STAT3-R609Q/Y705F were a kind gift from Daniel Gough (Hudson Institute of Medical Research). Other constructs have been described elsewhere [18,43,58]. U3A (a kind gift from George Stark, Lerner Research Institute, Cleveland Clinic), COS7, E6 Vero, HeLa and HEK293T cells were maintained in DMEM supplemented with 10% FCS and GlutaMAX (Life Technologies), 5% CO 2 , 37˚C. Transfections used Lipofectamine 2000 (Invitrogen), Lipofectamine 3000 (Invitrogen), or FuGene HD (Promega), according to the manufacturer's instructions.

Virus infection
All work with infectious virus was conducted at Physical Containment Level 4 (PC4) at the Australian Centre for Disease Preparedness (ACDP, formerly AAHL). EBOV infections used Mayinga 1976 isolate, which was originally received from NIH Rocky Mountain Laboratories and passaged three times in Vero cells at ACDP after receipt.

CLSM
For analysis of STAT3 localisation, cells growing on coverslips transfected with plasmids or infected with EBOV (MOI = 10) were incubated in serum-free-(SF)-DMEM for 1 h and treated without or with 10 ng/ml recombinant human OSM (BioVision) for 15 min (analysis of fixed/immunostained cells) or 30 min (analysis of STAT3-mCherry in living cells) before fixation using 3.7% formaldehyde (10 min, room temperature (RT) for transfected cells) or 4% paraformaldehyde (48 h, 4˚C for infected cells), followed by 90% methanol (5 min, RT) and immunostaining. Antibodies used for were: anti-STAT3 (Santa Cruz, sc-482; or Cell Signaling Technology, 9139), monoclonal (Absolute Antibody, Ab00692-23.0) or polyclonal (rabbit clone #691, final bleed 1410069) anti-EBOV NP, and anti-mouse or anti-rabbit Alexa Fluor 488, 568 or 647 secondary antibodies (ThermoFisher Scientific). Imaging used a Leica SP5 or Nikon C1 inverted confocal microscope with 63 X objective. For live cell analysis, cells were imaged in phenol-free DMEM using a heated chamber. Digitised confocal images were processed using Fiji software (NIH). To quantify nucleocytoplasmic localisation, the ratio of nuclear to cytoplasmic fluorescence, corrected for background fluorescence (Fn/c), was calculated for individual cells expressing transfected protein [18,20,43]; mean Fn/c was calculated for n � 24 cells for each condition in each assay.

Analysis of the effect of OSM on EBOV replication
E6 Vero cells were treated without or with 10 ng/ml or 20 ng/ml OSM for 24 h before removal of media and infection with EBOV (MOI = 1, 1 h). The inoculum was then replaced with media without or with OSM to continue the treatment for a further 24 h, 48 h or 72 h postinfection before collection of supernatant. Alternatively, E6 Vero cells were infected with EBOV (MOI = 1, 1 h) before replacement of inoculum with media for 24 h. Cells were then treated without or with OSM (10 ng/ml or 20 ng/ml) before collection of supernatant at 24 h, 48 h and 72 h post-treatment. For quantification of virus in samples, RNA was extracted using the MagMAX-96 Viral RNA Isolation Kit (Applied Biosystems) utilizing the Kingfisher Flex (ThermoFisher Scientific) before RT-qPCR (AgPath-ID One-step reverse transcription-PCR kit, Applied Biosystems) targeting the NP gene [35]. Copy numbers were calculated using a standard curve. Assays were performed with biological triplicates.

Luciferase reporter gene assays
Cells were co-transfected with m67-LUC or pISRE-LUC (in which Firefly luciferase expression is under the control of a STAT3 or STAT1/2-dependent promoter, respectively) and pRL-TK (transfection control, from which Renilla luciferase is constitutively expressed), as previously described [18,59], together with protein expression constructs. Cells were treated 16 h (OSM) or 8 h (IFN-α) post-transfection with or without OSM (10 ng/ml for 8 h) or IFN-α (1,000 U/ml for 16 hours) before lysis using Passive Lysis Buffer (Promega). Firefly and Renilla luciferase activity was then determined in a dual luciferase assay using a BMG CLARIOstar plate reader, as previously described [18,59]. Briefly, the ratio of Firefly to Renilla luminescence was determined for each condition, and then calculated relative to that determined for GFP-N-proteinexpressing cells treated with OSM (relative luciferase activity). Data from � 3 independent assays were combined, where each assay result is the mean of three replicate samples.

Statistical analysis
Unpaired two-tailed Student's t-test was performed using Prism software (version 7, GraphPad).