Human embryonic kidney (HEK) 293 cells (ATCC) were maintained in Dulbecco’s modified Eagle’s medium (DMEM, Gibco), supplemented with 10% fetal bovine serum (FBS, HyClone), 1:100 penicillin/streptomycin (Gibco) and 1:100 GlutaMax (Gibco). Human HAP-1 WT or IFN receptor knockout (KO) cells (Horizon Discovery) were maintained in Iscove’s Modified Dulbecco’s Medium (IMDM; Gibco), supplemented with 10% FBS and antibiotics. Cells were grown at 37 °C under 5% CO₂ and split using 0.05% Trypsin-EDTA (Gibco). Cells were seeded prior to all experiments at 2.5–5×10⁴/well in 96-well plates.

Newly-synthesized expression vectors were designed, based on human codon-optimized open reading frame (ORF) sequences of VP35, VP40 and VP24, for the following isolates from each filovirus species or variant (except LLOV, for which an isolate has yet to be recovered), as recorded in GenBank: Ebola virus 2014 Makona and 1976 Yambuku variants (EBOV, Ebola virus/H.sapiens-wt/LBR/2014/Makona-201403007 and Ebola virus/H.sapiens-tc/COD/1976/Yambuku-Mayinga; GenBank accession numbers KP178538.1 and AF086833.2, respectively); Sudan virus (SUDV, Sudan virus/H.sapiens-tc/UGA/2000/Gulu-808892; AY729654.1); Bundibugyo virus (BDBV, Bundibugyo virus/H.sapiens-tc/UGA/2007/Butalya-811250; FJ217161.1); Taï Forest virus (TAFV, Tai Forest virus/H.sapiens-tc/CIV/1994/Pauleoula-CI; FJ217162.1); Reston virus (RESTV, Reston virus/M.fascicularis-tc/USA/1989/Philippines89-Pennsylvania; AF522874.1); Marburg virus (MARV, Lake Victoria marburgvirus strain Uganda 01Uga07; FJ750957.1); and Ravn virus (RAVV, Lake Victoria marburgvirus strain Uganda 02Uga2007; FJ750953.1). In total, 51 wild-type (WT) and FLAG-tagged VP expression vectors were constructed and used in our study. Each codon-optimized construct was generated by Genscript, Inc. (Piscataway, NJ) in pUC57 vector, and then was subcloned using EcoRI-XbaI restriction enzyme sites into pcDNA3.1(+) for expression of WT (untagged) proteins, or subcloned using KpnI-XbaI sites into pcDNA3.1(+)-N-DYK for expression of N-terminally FLAG-tagged proteins. EBOV VP24 from the Makona variant was mutagenized later in the study to produce VP24-W42A/K142A, which has been reported to have less-efficient binding and antagonism of karyopherin α (46). As a negative control, we used an empty pcDNA3.1(+) vector (Invitrogen). As a control of transfection and luciferase assay efficiencies, we also used a dual, constitutively-expressing green fluorescent protein (GFP) and luciferase vector (Cignal, Qiagen).

Unless otherwise noted, antagonist activity experiments in human cells were performed adhering to the same overall protocol, in triplicate, via a standardized panel in 96-well plates (CellBIND, Corning). One plate allowed for simultaneous profiling of up to 29 of the WT or FLAG-tagged VP gene vectors in our library. For such assays, each VP construct was singly transfected at 80 ng/well of plasmid using TransIT-LT1 transfection reagent according to manufacturer’s recommendations (Mirus Bio, LLC) diluted into OptiMEM (Gibco), in conjunction with 10 ng/well of appropriate firefly luciferase-based, innate immune response reporter gene vector, and 10 ng of constitutive Renilla luciferase-based pRL-TK vector (Promega) used as a transfection efficiency control. IFN response reporters used in this study included: IFNβ expression promoter and IRF3 activity reporters (p125-Luc and p55-C1B-Luc, respectively, kind gifts of Takashi Fujita, Kyoto University); NF-κB activity reporter (Stratagene); IRF1 activity and ISRE promoter reporters (Cignal Reporter Assay Kits, Qiagen); and hamster (h)ISG54 promoter reporter (pISG54Lucter, a kind gift from Steven Goodbourn, University of London). Cignal reporters were transfected at 20 ng/well and without addition of pRL-TK, as this was already included within the kit. Except for FLAG-tagged VP24 assays (which used single transfection mix-derived replicates on the same plate that were then harvested for either reporter assays or western blot analysis, described below) and VP24 titration assays (which were conducted across two plates with 20–240 ng/well of VP vector), well positions for each construct were maintained across individual experiments. An equal amount of empty vector (EV) was transfected in dedicated wells for use as negative or positive controls of induction, or included in transfection mixes for VP24 titration assays to normalize amount of total DNA (up to 260 ng/well). Transfected cells were incubated at 37 °C for at least 18 h post-transfection (hpt) prior to induction. Assays shown are representative of multiple independent experiments. In a few experiments, select antagonists that were determined in preliminary data to share redundant activity to similar proteins (EBOV Mayinga VP24 and RAVV VP24) were replaced in our full panel with EBOV VP24-W42A/K142A.

Transfected cells in 96-well VP antagonist panels above were induced the following day via 1) Sendai virus (SeV, Cantell strain, Charles River Laboratories) infection at either high MOI (~22.2), or increasing MOI between ~0.022 and ~22.2, diluted into fresh medium, unless otherwise noted; 2) by dsRNA mimic poly(I:C) at up to 2 μg/mL using Lipofectamine RNAiMax reagent (Thermo Fisher Scientific); 3) by transfection of RIG-I CARD domain vector (60 ng/well, a kind gift of Sonja Best, NIAID) using TransIT-LT1; 4) by universal Type I IFN (U-IFN, 500–1000 U/mL); 5) by tumor necrosis factor (TNF; 50 ng/mL); or 6) by an attenuated recombinant Rift Valley fever virus in which NSs and NSm IFN antagonists have been removed (rRVF-ΔNSs-ΔNSm, “ΔΔrRVF”), used at an MOI of ~2.2 (Bird et al., 2011). Untreated medium was used as an induction control. For SeV and ΔΔrRVF, cells were incubated for 1–2 h to allow for virus adsorption, and then exchanged for 100 μL fresh medium; for HAP-1 KO cells, an equal amount of medium was instead overlaid onto unremoved SeV adsorption medium. For poly(I:C), cells were exchanged with 90 μL fresh medium and then 10uL of transfection mixture diluted in OptiMEM was added to each well; RIG-I CARD transfection mixtures were added similarly to poly(I:C), except 100 μL fresh replacement media was used. For our qRT-PCR assays described below, we instead cotransfected RIG-I CARD vector simultaneously with reporter and antagonist vectors. Apparent cytopathic effects were not observed for any SeV assay in HEK293 cells and were modest with ΔΔrRVF or in HAP-1 cells. Cells were incubated at 4h to 24hpi and harvested using 1X passive lysis buffer (Promega) incubated on a rocker for at least 15 m, and then stored at −20 °C until use, or assayed immediately for relative luciferase activities.

Equal aliquots of induced cell lysates from 96-well plates were transferred to cognate wells in white bottom assay plates (Corning Costar), and firefly and Renilla luciferase activities assessed using the Dual Luciferase Assay Kit (Promega) for reporter gene and transfection control raw counts (relative light units, RLU), respectively, analyzed on a BioTek Synergy 4 microplate reader (Thermo). Relative firefly raw values were normalized to those of Renilla raw values across samples. Normalized values were then compared to EV negative control to obtain fold induction values, and then to EV positive control values to obtain percentage induction, set at 100% (fold change is also shown as “FC” above positive controls for each assay). For most assays, data was stratified into separated, relative activities for each of the three VPs across tested filoviruses.

For western blotting (see below), an anti-FLAG rabbit antibody (Sigma) was used at 1:2000 and 1:10000; and mouse anti-β-actin mAb (Genscript) was used at 1:2000 as an internal loading control. For neutralization assays (see below), a cocktail of IFNα, IFNβ and IFNα1 receptor (IFNAR1, PBL Biosciences) neutralizing antibodies (nAbs) were used at concentrations from 1000 U/mL, 2000 U/mL and 20 μg/mL, respectively, up to 20 μg/mL for each.

For each FLAG-tagged viral protein reporter assay, cells in single transfection mix-derived replicate wells on the same 96-well plate were lysed using up to 100 μL radioimmunoprecipitation assay (RIPA) buffer (Pierce), and then total lysates were centrifuged through QiaShredder columns (Qiagen) at maximum speed. Clarified lysates were transferred to fresh microcentrifuge tubes and stored at −20 °C, or equal aliquot amounts run immediately on a 4–12% NuPAGE Bis-Tris gel (Novex, Life Technologies) to separate proteins by gel electrophoresis. Protein gels were transferred onto nitrocellulose membranes using an iBlot apparatus (Invitrogen) and washed with deionized water and phosphate-buffered saline (PBS, Gibco) supplemented with 0.1% Tween-20 (PBST, Thermo). Blots were blocked for at least 45 m using 5% non-fat dry milk (BD Difco) in PBST, then washed and/or incubated with primary antibodies as indicated above overnight, or for 1h in the case of anti-FLAG used at 1:10000. Membranes were then washed, incubated with horseradish peroxidase (HRP)-conjugated secondary antibodies for 1 h and rewashed using a BenchPro 4100 Card Processing Station (Invitrogen). Finally, membranes were rinsed with water, and protein bands visualized using enhanced chemiluminescence (ECL, Pierce) on a ChemiDoc MP Imager (Bio-Rad).

nAbs were added to reporter- and/or VP-transfected cells as cocktails in the concentrations described above either simultaneously along with SeV diluted into fresh medium, or as a cocktail diluted alone for pre-treatment of cells for at least 4 h; in either case, 50uL of medium was added per well. Following SeV adsorption, 50 μL per well of fresh medium was layered on top of conditioned medium; following pre-treatment of antibody, up to 50 μL per well of media containing SeV, or inducing cytokine at double working concentration, was layered on top.

In a manner analogous to tagged vector-based reporter assays that were compared to western blot analysis, cells in single transfection mix-derived replicate wells on the same 96-well plate were lysed 24hpi, either by a high MOI of SeV (~22.2; induced 1dpt) or by simultaneous cotransfection/induction by RIG-I CARD vector along with antagonist and reporter vectors, for RNA extraction with 100 μL of 1x RNA Lysis/Binding Solution Concentrate (Thermo) followed by magnetic bead purification and TURBO DNase treatment using the MagMAX Total RNA Isolation kit according to the manufacturer’s recommendations (Thermo). Yields of purified RNA samples were verified by NanoDrop spectrophotometer (Thermo) and then interferon gene expression was assessed using human IFNβ and IFNλ gene-specific TaqMan-based primer-probes (Integrated DNA Technologies) as well as a primer-probe for RNase P (RPP30) as an internal control. 25 μL reaction mixes were produced using the SuperScript III One-Step RT-PCR system with Platinum Taq DNA polymerase and ROX dye according to the manufacturer’s recommendations (Thermo). Reactions were run on a Bio-Rad CFX96 Touch Real-Time PCR Detection System. Amplification data was corrected for fluorescence drift and baseline threshold, normalized to internal control, and fold change (“FC”) quantitated by ΔΔCt method in CFX Manager (Bio-Rad). As with reporter assays, data is compared relative to EV positive control induction values, set at 100%. Primer-probe sequences used were: IFNβ (Fwd: 5′-GCAA TTTTCAGTGTCAGAAGCTC-3′, Rev: 5′-TCCTGTCCTTGAGGCA GTATT-3′, Probe: 5′-TGTGGCAATTGAATGGGAGGCTT-3′); IFNλ (Fwd: 5′-GGAGCTAGCGAGCTTCAAGA-3′, Rev: 5′-ACTCCAGTTTTTC AGCTTGAGTG-3′, Probe: 5′-GCCAGGGACGCCTTGGAAGAG-3′); RPP30 (Fwd: 5′-AGATTTGGACCTGCGAGCG-3′, Rev: 5′-GAGCGG CTGTCTCCACAAGT-3′, Probe: 5′-TTCTGACCTGAAGGCTCTGCGC G-3′).

All statistics shown in figures was obtained by conducting a Student’s unpaired t-test for all values in which significance was indeterminate and that remained potentially significant at the highest MOI or protein concentration tested. Significant two-tailed p values were designated as follows: *=p < 0.05; **=p < 0.005; ***=p < 0.0005; ****=p < 0.0001. Standalone asterisks above each bar indicate t-tests that directly compare that antagonist activity value to the positive control induction value in each assay; asterisks above horizontal lines indicate t-tests that directly compare activity values between two antagonists. All statistics were calculated using QuickCalcs software (GraphPad).

To conduct a side-by-side assessment of the potencies of the three known interferon (IFN) response antagonists, VP35, VP40 and VP24, in a single system, we generated a library of pcDNA3.1(+)-based vectors for expression of the three proteins from representative viral isolates for each filovirus species known to cause disease in humans or NHPs. These include reference isolates of EBOV from 2014 and 1976 outbreaks (Makona-201403007 and Yambuku-Mayinga; for clarity, these will be heretofore referred to as “EBOV Makona” and “EBOV Mayinga”), SUDV, RESTV, TAFV, BDBV, and related isolates of MARV and RAVV (both obtained from the same outbreak), as listed in Fig. 1F (Towner et al., 2009). LLOV was not included in this study as no isolate yet exists and there is no evidence of any human or NHP infection or disease. Because specific antibodies for probing many filoviral proteins are not available, vectors containing an N-terminally expressed FLAG tag were also generated, allowing assessment of antagonist expression using a single antibody (Burk et al., 2016). A total of 51 vectors that express WT or cognate tagged viral proteins were designed and used for the bulk of the analyses.

Along with our standardized library, we developed a standardized, 96-well plate reporter assay system for pan-filovirus assessment of IFN antagonist activities. In this format, all WT forms of antagonists in our library were transfected and induced in parallel, cotransfected with one of several luciferase-based reporters to measure responses during either upstream IFN induction (IRF3 and NF-κB activities, IFNβ promoter expression), downstream IFN signaling (ISRE and ISG54 promoter expression) or both (IRF1 activity, shown to play roles in both IFN induction and as well as ISG transactivation), as depicted in Fig. 1A–E (Yarilina and Ivashkiv, 2010; Hu et al., 2008; Kröger et al., 2002). Responses in the presence of filoviral proteins were compared to empty vector (EV) positive and negative controls to establish antagonist activity.

Prior studies with well-studied EBOV VP35 have shown nearlycomplete inhibition of IFNβ induction, while weaker antagonism was found for MARV VP35 (Dunham et al., 2015; Edwards et al., 2016; Albariño et al., 2015; Feagins and Basler, 2015; Basler et al., 2000). Thus, we first measured antagonism of IFNβ expression, predicting a similar pattern of activity for the other VP35 proteins. We found that all ebolavirus VP35 proteins showed equally potent, nearly-complete suppression of IFN reporter activity over a 1000-fold range of Sendai virus (SeV) infection from an MOI of ~0.022 (Fig. 2A). Indeed, very strong antagonism was already evident by 4hpi (Fig. 2B). MARV VP35, as expected, showed diminished IFN inhibition by comparison to ebolaviruses. Interestingly, RAVV VP35 showed approx. 3–5-fold more potent inhibitory activity than MARV VP35 across all SeV MOIs (Fig. 2A), even at the early induction timepoint (Fig. 2B).

We next tested VP35 potencies in cells transfected with activity reporters for IFN-transactivating proteins IRF3 or NF-κB. The VP35 antagonism profiles for both reporters were similar to that for IFNβ, with strong inhibition of SeV-induced IRF3 (Fig. 2C) or NF-κB (Fig. 2D) activity by all ebolavirus proteins, yet weaker activity by MARV VP35, and intermediate activity by RAVV VP35. Conversely, when we induced IFNβ using a surrogate for dsRNA, poly(I:C), a lesspotent, more MDA5-dependent inducer of the IFN pathway than SeV (Li et al., 2005; Gitlin et al., 2006; Kato et al., 2006; Reimer et al., 2008), the relative differences between marburgvirus and ebolavirus VP35 proteins disappeared (Fig. 2E).

We then assayed VP35 antagonistic activities on the downstream IFN signaling pathway, by using either the ISRE (Fig. 2G) or ISG54 promoter reporter (Fig. 2H). We also used our IRF1 dual induction/signaling reporter (Fig. 2F). In all cases, VP35 activity profiles in SeV-induced cells were analogous to those seen for IFNβ induction, with strong ebolavirus VP35, intermediate RAVV VP35 and weaker MARV VP35 antagonism.

Finally, since observed differences in VP35 activities between viruses could be due to inadequate expression of particular proteins in HEK293 cells, we validated these results using FLAG-tagged VP35 constructs transfected in two parallel sets of replicates and harvested for luciferase or western blot assays, respectively. All FLAG-tagged filovirus VP35 proteins were well expressed (Fig. 2J). We also found no impairment of tagged VP35 function compared to either untagged EBOV Makona VP35 directly, or to the cognate WT proteins from previous assays (Fig. 2A and B), as tagged proteins showed near-identical inhibitory effects on IFNβ reporter activity following SeV induction (Fig. 2I). Thus, these assays support that our previous observations for WT VP35 proteins were due to bona fide reporter gene antagonism and not to indirect effects at the level of protein expression.

Taken together, our data indicate that VP35 activity to block the IFN response, as has been observed for EBOV and MARV, are maintained across all filoviruses, but that RAVV VP35 has stronger-than-expected antagonism than previously reported for a marburgvirus.

As MARV VP40 was originally identified to have antagonistic effects on Jak1 activation during downstream IFN signaling, we decided to first analyze pan-filovirus VP40 activities within our panel using the same two IFN signaling reporters, ISRE and ISG54, as well as IRF1 dual activity reporter, described above (Valmas and Basler, 2011). We anticipated that MARV VP40 would prevent signaling upon U-IFN treatment. Indeed, activation of both ISG54 (Fig. 3A) and ISRE (Fig. 3B) reporters were robustly inhibited by MARV VP40 under these conditions. Closely-related RAVV VP40, however, showed half the potency with both reporters after U-IFN induction (Fig. 3A and B). Meanwhile, as previous reports on EBOV VP40 activity during IFN signaling suggested (Ramanan et al., 2011; Basler, 2015; Valmas et al., 2010), every ebolavirus VP40 protein failed to inhibit U-IFN-induced ISG54 (Fig. 3A), and in fact even showed moderate reporter overstimulation, an effect seen in some assays by select proteins lacking inhibitory ability. These effects are likely due to mild background induction of IFN responses by ectopic filoviral proteins, which is observable only for those proteins that are then unable to counteract such responses. Most ebolavirus VP40 proteins also did not antagonize U-IFN-induced ISRE (Fig. 3B), but unexpectedly, we observed a moderate inhibition of ISRE by both SUDV and TAFV VP40 proteins, at levels similar to that of RAVV VP40 (Fig. 3B).

Next, we looked at VP40 IFN signaling antagonism following infection by SeV, a more potent and broader inducer of the IFN pathway (Li et al., 2005; Peters et al., 2008; Zaritsky et al., 2015). In contrast to the strong activity by MARV or RAVV VP40 seen after U-IFN induction of ISG54 and ISRE (Fig. 3A and B), antagonism of these reporters was much weaker following SeV induction (Fig. 3C and D). Further, for ISRE induced by over a 1000-fold range of SeV from an MOI of ~0.022, we observed MOI-dependent effects on marburgvirus VP40 activity (Fig. 3D), with comparatively weaker retention of potencies for RAVV VP40 compared to MARV VP40 at each MOI. Conversely, both SeV-induced reporters showed mostly similar ebolavirus VP40 activity profiles (Fig. 3C and D) to those obtained by U-IFN (Fig. 3A and B). The exception was TAFV VP40, which, interestingly, showed moderate inhibition of SeV-induced ISG54 (Fig. 3C), an activity that was neither observed following U-IFN induction nor shared by SUDV VP40 (Fig. 3A). And unlike the reduced activity found for marburgvirus VP40, antagonism of SeV-induced ISRE by both SUDV and TAFV VP40 was maintained at equivalent levels (Fig. 3D) as compared to U-IFN induction (Fig. 3B), and their activities were mostly unaltered by SeV MOI (Fig. 3D). Further support for the activities of MARV, SUDV and TAFV VP40 proteins on the ISRE reporter could be found when we instead infected cells with an IFN antagonist-deficient recombinant Rift Valley fever virus (ΔΔrRVF). Antagonist activity profiles following induction by ΔΔrRVF (MOI of ~2.2, Supplemental Fig 1A) were highly similar to that obtained with SeV (Fig. 3D), indicating VP40 control of IFN responses are broader effects and are not specific only to SeV infection.

Meanwhile, when we tested IRF1 activity following SeV induction, we saw a VP40 activity profile at a low MOI (~0.022, Fig. 3E) that was analogous to that seen for U-IFN-induced ISRE (Fig. 3B), with more potent inhibitory activity by marburgvirus VP40 than that seen for SeV-induced ISRE and ISG54 (Fig. 3C and D), as well as stronger, MARV VP40-like levels of antagonism by TAFV and SUDV VP40. This effect on IRF1 activity was abrogated at a high MOI (~22.2, Fig. 3E), in which the potency profile returned to being similar to that seen for SeV-induced ISRE (Fig. 3D).

Finally, as we considered previously for VP35, differences in activities by VP40 proteins between viruses could be due to major inefficiencies in protein expression. Thus, we validated our data with FLAG-tagged VP40 constructs transfected in parallel sets of replicates and harvested for either luciferase or western blot assays. Indeed, as we had confirmed for tagged VP35 proteins, FLAG-tagged VP40 proteins were well expressed (Fig. 3I). The tagged proteins gave similar activity profiles for U-IFN-induced ISG54 (Fig. 3F), including strong inhibition by VP40 of marburgviruses, and for SeV-induced ISRE and ISG54 (Fig. 3G and H), including the moderate antagonism by SUDV and/or TAFV VP40, as compared to the cognate untagged proteins from previous assays (Fig. 3A, C and D), thereby corroborating our observations for WT VP40 activities and ruling out indirect protein expression-level effects.

These data suggest that SUDV and TAFV VP40 have a moderate IFN signaling-inhibitory function, an activity previously unreported for ebolaviruses. Meanwhile, VP40 of marburgviruses have filovirus- and inducer-specific potencies whereby MARV VP40, in agreement with prior studies (Valmas and Basler, 2011), is capable of suppressing IFN responses induced by high amounts of SeV or ΔΔrRVF, an activity that is not shared by RAVV VP40.

Following our assessment of downstream IFN signaling, which was known to be inhibited by MARV VP40 (Valmas and Basler, 2011), we next decided to test filoviral VP40 proteins for any potentially unrecognized ability to antagonize the upstream IFN induction pathway. Indeed, as with our ISRE and ISG54 signaling reporters, IFNβ expression over a 1000-fold range of SeV infection from an MOI of ~0.022 (Fig. 4A) or IRF3 activity at high MOI (~22.2, Fig. 4B) were moderately antagonized by VP40 proteins of marburgviruses, but not of ebolaviruses, with the notable exception of TAFV. We also again observed more potent MARV VP40 than RAVV VP40 activity. While IRF3 reporter activity was modestly inhibited by these VP40 proteins, NF-κB activity was not suppressed by any filoviral VP40 (Fig. 4C).

These results suggest that, in addition to its known IFN signaling antagonism, marburgvirus VP40 is also capable of moderate, previously-unreported inhibitory activity during upstream IFN induction; and that TAFV, but not SUDV, VP40 maintains a role in blocking IFN induction, as was shown for both ebolavirus proteins for IFN signaling. Our assays also suggest marburgvirus and TAFV VP40 may inhibit IFNβ expression through IRF3, but not through NF-κB, signaling.

VP24 of EBOV has long been known to block downstream IFN signaling by binding to karyopherin α, which prevents STAT1 interaction and subsequent nuclear translocation (Reid et al., 2006, 2007; Zhang et al., 2012a; Mateo et al., 2010). Given this reported activity, we first decided to compare pan-filovirus VP24 activities within our panel using the ISRE and ISG54 IFN signaling reporters. As shown for ISRE-cotransfected cells induced by U-IFN (Fig. 5A), poly(I:C) (Fig. 5B) or SeV over a 1000-fold range from an MOI of ~0.022 (Fig. 5C), VP24 of ebolaviruses, but not of marburgviruses, strongly antagonized reporter activity for all tested conditions. Interestingly, RESTV VP24 displayed potent antagonism in U-IFN-induced cells, but unlike those of the other ebolaviruses, showed markedly weaker antagonism when cells were induced by poly(I:C) or higher levels of SeV. This impaired antagonism by RESTV VP24 relative to other ebolavirus proteins at increased SeV MOI (~22.2) was also found using our reporter for IRF1 dual activity (Fig. 5D). RESTV VP24 inhibition of ISG54 expression was similarly weak in a VP24-specific titration assay, while all other ebolavirus VP24 proteins had robust, dose-dependent increases in potency (Fig. 5E). So too was reduced RESTV VP24 activity obtained following ISRE induction by ΔΔrRVF (MOI of ~2.2, Suppl Fig. 1B), offering support that this inefficiency was not specific to SeV and poly(I:C) induction. Activity profiles for the other filoviral proteins following ΔΔrRVF induction (Suppl Fig. 1B) were also highly consistent to those seen for SeV-induced assays under similar conditions (Fig. 5B and C), indicative of broader, inherent abilities by ebolavirus VP24 proteins in controlling IFN responses. Among VP24 proteins across all assays, we observed strongest antagonism for SUDV, BDBV or TAFV (Fig. 5A–F; Suppl Fig. 1B).

Further, as an added control for specificity of VP24 activity, we tested in a subset of our assays a mutated form of EBOV VP24 with alanine point mutations at W42 and K142, which have been reported to ablate IFN signaling antagonism due to weakened karyopherin α binding (Mateo et al., 2010). We noted marked differences between EBOV VP24 WT and this signaling antagonism-impaired EBOV VP24-W42A/K142A in U-IFN-induced cells cotransfected with the ISG54 reporter, as anticipated based on the available data (Fig. 5A) (Mateo et al., 2010); however, such reduced function by EBOV VP24-W42A/K142A was minimal in the context of SeV induction, conditions that have not been previously tested (Fig. 5E).

Finally, as we had validated for VP35 and VP40, we ensured that the observed differences in activities by VP24 proteins were not due to deficiencies in protein expression. We transfected FLAG-tagged VP24 constructs with two parallel sets of replicates and harvested cell lysates for luciferase or western blot assays. Like the other antagonists in our library, all tagged filovirus VP24 proteins were well expressed (Fig. 5G). FLAG-tagging had no deleterious effect on VP24 function, as all tagged VP24 proteins of ebolaviruses efficiently blocked SeV-induced ISG54 reporter activity with the exception of tagged RESTV VP24 (Fig. 5F), similarly to what we observed with WT RESTV VP24 (Fig. 5B–E). Meanwhile, tagged marburgvirus VP24 proteins showed no potency, whereas tagged EBOV VP24-W42A/K142A had minimally impaired activity compared to tagged EBOV VP24 WT, both of which had also been seen using the untagged forms of these proteins. These assays again confirmed that the observed effects for WT antagonists, in this case VP24, were due to actual inhibition of reporter activity rather than indirectly due to inefficient protein expression.

Together, our results indicate that VP24 of all ebolaviruses, but not marburgviruses, share robust suppression of IFN signaling except for VP24 of RESTV. Intriguingly, RESTV VP24 appears unable to efficiently antagonize IFN signaling elicited under most tested conditions. Conversely, these results indicate that residues putatively impairing VP24 antagonism of karyopherin α, such as those of EBOV VP24-W42A/K142A, may be less sufficient to unblock IFN signaling in certain inducing conditions than previously appreciated.

Similar to our investigation for VP40, we next decided to assess filoviral VP24 proteins for any previously unrecognized ability to inhibit upstream IFN induction. Unexpectedly, VP24 of all ebolaviruses except RESTV robustly suppressed SeV-induced IFNβ expression in either a SeV MOI-dependent (over a 1000-fold range from an MOI of ~0.022, Fig. 6A) or a VP24 dose-dependent (Fig. 6B) manner. In both experiments, VP24 of RESTV again showed minimal inhibitory activity, similar to that seen with downstream IFN signaling (Fig. 5).

When VP24 proteins were studied using reporters of IRF3 (Fig. 6C) and NF-κB (Fig. 6D) activity following SeV infection, or following direct RIG-I CARD vector-based (Fig. 6E) or poly(I:C)-based (Fig. 6F) induction of IFNβ expression, antagonistic effects were similar, or in the case of NF-κB, appeared to be filovirus species specific. Consistent with our previous data, RESTV VP24 was always weaker compared to other ebolaviruses, even despite a relatively strong inhibition of poly(I:C)-based IFNβ induction, while TAFV and BDBV VP24 proteins were the most potent. Use of FLAG-tagged filoviral VP24 proteins, transfected at equal amounts to their untagged counterparts at well-expressed levels as previously shown (Fig. 5F), also strongly antagonized IFNβ expression after SeV induction (Fig. 6G), once again validating the activities obtained by the WT constructs. Despite more modest differences in this particular assay between tagged ebolavirus VP24 potencies, RESTV VP24 still showed overall weakest activity. Next, in an attempt to determine if the W42A/K142A mutations in EBOV VP24 showed different effects during IFN induction compared to IFN signaling, and thus suggest potentially separable functional regions within VP24 proteins that may be exploited by reverse genetics, we included both untagged and tagged EBOV VP24-W42A/K142A in a subset of induction assays. However, both forms of EBOV VP24-W42A/K142A showed similarly modest differences after SeV induction compared to EBOV VP24 WT (Fig. 6B and G) as we saw with IFN signaling reporters (Fig. 5E and F), making it difficult to assess the importance of putative karyopherin α-interacting residues for induction-specific targeting.

Finally, to further investigate if the apparent property of ebolavirus VP24 to inhibit upstream IFN induction was demonstrable by a different method of measurement, we conducted a side-by-side comparison of reporter assays with qRT-PCR-based IFN gene expression following SeV induction (MOI of ~22.2), analogously to those we conducted when assessing protein expression levels. Using a subset of VP24 proteins along with human TaqMan primer-probes for IFNβ and IFNλ, a Type III interferon gene known to be similarly induced and activate many of the same pathways as IFNβ (Hoffmann et al., 2015), we found that, when compared to the very strong antagonism seen by IFNβ reporter assay (Suppl Fig. 2A), tagged EBOV VP24 displayed a more modest, yet still significant inhibition of IFN gene expression (Suppl Fig. 2B and C), and in both assays, MARV VP24 was incapable of blocking IFN induction. TAFV and RESTV VP24 showed even weaker activity, with little distinction of activities between them (Suppl Fig. 2B and C). We reasoned that this much weaker observable antagonism following SeV induction was likely due to the activation of IFN responses in a substantial proportion of untransfected cells, leading to a high background accumulation of IFN in cell lysates that obscured the actual activity of respective VP24 proteins within transfected cells. To address this problem, we cotransfected RIG-I CARD vector along with VP24 proteins and reporter vector simultaneously, which we hypothesized would greatly increase the chances of viral protein transfection and response induction within the same cell population and thus better resolve VP24 antagonism of IFN expression. As anticipated, under these conditions we achieved highly correlative results between reporter (Suppl Fig. 2D) and qRT-PCR assays (Suppl Fig. 2E and F) for the same IFN gene targets, including potent antagonism by all ebolavirus proteins, but not MARV VP24, and weaker activity by RESTV VP24. Therefore, inhibition of endogenous IFN gene expression by ebolavirus VP24 as observed by qRT-PCR aligns well with activities previously measured in induction gene reporter-based assays and lends support that they were not a consequence of the luciferase reporter approach.

These data taken together suggest marked, noncanonical antagonism by ebolavirus VP24 proteins on upstream induction of both ectopic reporters and endogenous IFN genes, in some cases with variable potencies seen between viruses, but in which RESTV VP24, as was observed for IFN signaling, was consistently the weakest antagonist under all tested conditions.

An open question in understanding the observed antagonistic activities by ebolavirus VP24 proteins on IFN induction is whether there is any indirect effect involving positive feedback (Feagins and Basler, 2015). In such a scenario, VP24 blockade of downstream IFN signaling would prevent IRF7 expression and the subsequent activation of a potential positive feedback loop that can further perpetuate IFNβ expression (Haller et al., 2006; Feagins and Basler, 2015; Sato et al., 1998). To rule out this possibility in our system, we first assessed our induction reporters at very early timepoints (4 h) post-SeV infection, near the limits of assay detection, when IRF7 induction and putative feedback would be unlikely to have a strong impact on results. We found that VP24 of ebolaviruses maintain similarly robust antagonism of both IRF3 activity (Fig. 7A) and IFNβ expression (Fig. 7B) as early as 4hpi. Indeed, VP24 inhibition was equally as strong, or even stronger, than that observed 24hpi as previously shown (Fig. 6A and B, D).

For further support, we directly induced the IFN signaling pathway by U-IFN along with a cocktail of neutralizing antibodies (nAbs) targeting IFNα, IFNβ and the IFNα1 receptor (IFNAR1). We predicted that U-IFN would be unable to upregulate IFNβ, limiting a role for a putative loop, and would be blocked by the cocktail for appropriate ISG54 induction. As we show in Fig. 7C, U-IFN did not induce IFNβ expression regardless of the addition of nAbs; meanwhile, U-IFN strongly induced ISG54 reporter activity, which was completely abolished by the cocktail, thus validating the efficacy of our antibody cocktail and the apparent inability of active IFN signaling pathway components to successfully induce transactivation of the IFNβ promoter. Next, we used the same cocktail to test EBOV VP24 antagonism following SeV induction of IFNβ. We found that increasing amounts of nAbs minimally affected IFNβ reporter activity, and that no loss of EBOV VP24 potency to block IFNβ expression occurred upon nAb treatment (Fig. 7D).

Finally, we employed HAP-1 WT and IFN receptor knockout (IFNAR KO) cell lines to validate our results, as well as provide a cleaner and more direct comparison of VP24 activity in cells in which the IFN signaling pathway is completely abrogated. HAP-1 WT, but not KO, cells exhibited IFN signaling via ISG54 reporter activity as expected (Suppl Fig. 3A), while strong IRF1 reporter activity was not fully ablated in KO cells (Suppl Fig. 3B), likely due to the unmodulated role of IRF1 within the IFN induction pathway. As in HEK293 cells, we saw similarly robust antagonism of IRF1 activity in HAP-1 WT cells by EBOV VP35 and VP24 (Suppl Fig. 3C), and importantly, similar IRF1 inhibitory profiles for filoviral VP35 and VP24 WT proteins in the IFNAR KO (Suppl Fig. 3D and E).

Together, these data suggest that in our system, there is no substantial evidence for an indirect influence on VP24 antagonism by a potential positive feedback loop, and demonstrate that ebolavirus VP24 can indeed inhibit the upstream IFN induction pathway independently from any IFN signaling pathway component, thus appearing to represent a bona fide novel function of these antagonist proteins.

As VP24 of all ebolaviruses except RESTV strongly inhibited IRF3 activity upon SeV induction (Fig. 6C), but showed variable inhibition of NF-κB activity in a species-specific manner, with weak antagonism by VP24 proteins of EBOV, SUDV or RESTV (Fig. 6D), we assessed NF-κB activity following direct TNF induction at early (4 h) or late (24 h) timepoints. NF-κB induction by TNF occurs through a pathway distinct from TLR and RIG-I signaling that may supersede the activation of induction intermediates such as MAVS or TRAF6 (Fig. 1A) (Yarilina and Ivashkiv, 2010; Palladino et al., 2003; Kumar et al., 2013). Under these conditions, we found that all ebolavirus VP24 proteins, except RESTV VP24, were capable of efficiently antagonizing NF-κB activity early after TNF induction (Fig. 8A), but that EBOV and SUDV proteins lost some degree of potency by 24hpi (Fig. 8B), as observed in prior SeV-based assays. Consistent with previous reports, we found that all filovirus VP35 proteins were rendered incapable of inhibiting TNF-induced NF-κB activity (Fig. 8 C), despite their strong antagonism when we previously had induced by SeV (Fig. 2D) (Chang et al., 2009).

These results suggest that VP24 proteins of ebolaviruses, except of RESTV, are also capable of robustly inhibiting early NF-κB signaling when directly induced by TNF, an activity by which VP35 appears to be insufficient. This further suggests possible non-redundant functions of ebolavirus VP35 and VP24, which may together antagonize IFN induction under a variety of alternative and shifting host response conditions.