TNF-Mediated Inhibition of Classical Swine Fever Virus Replication Is IRF1-, NF-κB- and JAK/STAT Signaling-Dependent

The sera from pigs infected with virulent classical swine fever virus (CSFV) contain substantial amounts of tumor necrosis factor (TNF), a prototype proinflammatory cytokine with pleiotropic activities. TNF limits the replication of CSFV in cell culture. In order to investigate the signaling involved in the antiviral activity of TNF, we employed small-molecule inhibitors to interfere specifically with JAK/STAT and NF-κB signaling pathways in near-to-primary endothelial PEDSV.15 cells. In addition, we knocked out selected factors of the interferon (IFN) induction and signaling pathways using CRISPR/Cas9. We found that the anti-CSFV effect of TNF was sensitive to JAK/STAT inhibitors, suggesting that TNF induces IFN signaling. Accordingly, we observed that the antiviral effect of TNF was dependent on intact type I IFN signaling as PEDSV.15 cells with the disrupted type I IFN receptor lost their capacity to limit the replication of CSFV after TNF treatment. Consequently, we examined whether TNF activates the type I IFN induction pathway. With genetically modified PEDSV.15 cells deficient in functional interferon regulatory factor 1 or 3 (IRF1 or IRF3), we observed that the anti-CSFV activity exhibited by TNF was dependent on IRF1, whereas IRF3 was dispensable. This was distinct from the lipopolysaccharide (LPS)-driven antiviral effect that relied on both IRF1 and IRF3. In agreement with the requirement of IRF1 to induce TNF- and LPS-mediated antiviral effects, intact IRF1 was also essential for TNF- and LPS-mediated induction of IFN-β mRNA, while the activation of NF-κB was not dependent on IRF1. Nevertheless, NF-κB activation was essential for the TNF-mediated antiviral effect. Finally, we observed that CSFV failed to counteract the TNF-mediated induction of the IFN-β mRNA in PEDSV.15 cells, suggesting that CSFV does not interfere with IRF1-dependent signaling. In summary, we report that the proinflammatory cytokine TNF limits the replication of CSFV in PEDSV.15 cells by specific induction of an IRF1-dependent antiviral type I IFN response.


Introduction
The first line of protection of host cells from invading viruses is mediated by the innate immune system. By sensing unique pathogen-associated molecular patterns, conserved cellular pattern recognition receptors initiate multiple intracellular signaling cascades involving interferon regulatory factors (IRF) that culminate in the transcriptional activation and secretion of type I interferons (IFN-α and IFN-β) and type III IFN (IFN-λ) [1]. Specific interactions of IFNs with cellular type I (IFNAR1) and type III IFN receptors subsequently activate Janus kinase (JAK)-and signal transducer and activator of transcription (STAT)dependent signaling in neighboring cells. Consequently, the IFN-mediated JAK/STAT signaling leads to the expression of a multitude of IFN-stimulated genes that synergistically orchestrate cellular antiviral defense [2].
Classical swine fever virus (CSFV) causes a highly contagious hemorrhagic fever in pigs [3]. CSFV is a non-cytopathogenic pestivirus of the Flaviviridae family, and the enveloped virion harbors a single-stranded positive-sense RNA genome [4]. Like most viruses, CSFV is highly susceptible to the antiviral actions mediated by type I and type III IFN and has evolved potent strategies to interfere with the cellular antiviral defense [3,[5][6][7]. In most cells except plasmacytoid dendritic cells (pDC), CSFV interferes with type I IFN induction by means of the viral N pro protein which interacts with IRF3 and induces its proteasomal degradation [7,8]. IRF3 is a key transcription factor of the type I IFN induction cascade triggered by DNA and RNA viruses that is targeted by many viral and bacterial pathogens [9]. Despite N pro -mediated IRF3 degradation, CSFV induces potent IFN-α and proinflammatory host responses in vivo involving pDC, conventional DC and monocytic cells (reviewed in [7]). Among the proinflammatory cytokines, tumor necrosis factor (TNF) represents a key cytokine promoting pleiotropic cellular effects, such as apoptosis, proliferation, survival or differentiation [10]. TNF activates nuclear factor κB (NF-κB) and mitogen-activated protein kinase signaling pathways [11]. Notably, pigs infected with virulent CSFV induce high levels of TNF [12][13][14], and TNF was reported to inhibit the replication of CSFV in porcine cells [15,16]. The antiviral effect of TNF was reduced in p65-silenced PK-15 cells indicating that TNF inhibits CSFV replication via the NF-κB signaling pathway [16]. Interestingly, porcine reproductive and respiratory syndrome virus (PRRSV) infection leads to TNF secretion that in turn inhibits the proliferation of a subsequent CSFV C-strain infection, which may explain CSFV vaccination failures caused by PRRSV infection in the field [15].
Studies conducted with primary macrophages and murine microvascular endothelial cells revealed that TNF induces IRF1-dependent IFN-β responses [17][18][19]. Like IRF3, IRF1 does also bind and activate the IFN-β promoter [20]. Furthermore, IRF1 is critical for the TNF-driven type I IFN response in rheumatoid fibroblast-like synoviocytes [21]. Interestingly, CSFV infection or dsRNA stimulation of PK-15 cells upregulate IRF1 mRNA [22,23]. Overexpression of IRF1 in PK-15 cells triggers antiviral responses against different porcine viruses, although IRF1 is dispensable for IFN-β induction by RNA viruses [23]. Finally, a recent study showed that CSFV N pro antagonizes IRF1-mediated type III IFN production by downregulating IRF1 expression and inhibiting its nuclear translocation in a porcine intestinal epithelial cell line [24]. Altogether, the data described above show that antiviral TNF signaling involves NF-κB and IRF1 and that the anti-CSFV activity of TNF relies on type I IFN responses in an IRF1-and/or IRF3-dependent manner, but the formal proof for a direct link of these signaling elements in the context of CSFV is still missing. In order to explore this in more details, we aimed at deciphering the cellular signaling pathways exhibited by TNF-driven anti-CSFV responses using pharmacological and genetic targeting of selected cellular signaling factors. For this, we used the immortalized near-to-primary porcine aortic endothelial cell line PEDSV.15 [25] that we found to be highly sensitive to the antiviral action triggered by physiological levels of TNF, including porcine (pTNF) and murine TNF (mTNF), as opposed to the common porcine kidney cell lines PK-15 and SK-6. For quantitative virological readouts, we employed a firefly luciferase-expressing CSFV (CSFV-luc). With inhibitory drugs, CRISPR/Cas9-mediated gene knockout and anti-TNF antibodies, we demonstrate that TNF limits the replication of CSFV by activating JAK/STAT signaling in an IRF1-, NF-κB and IFNAR1-dependent way, independently of IRF3.

Viruses
The bicistronic CSFV-luc was derived from a full-length cDNA construct obtained by replacing the N pro -C gene cassette in the pA187-1 cDNA backbone [27] with the corresponding N pro -Luc-IRES-C gene cassette from the bicistronic pA187-N pro -Luc-IRES-C-delE rns replicon construct [28] using standard PCR-mediated cloning. The CSFV-luc and the virulent CSFV vEy-37 [29] were rescued from cDNA as described elsewhere [30] and propagated in PEDSV.15 cells. Viral titers were determined by endpoint dilution in PEDSV.15 cells and expressed as 50% tissue culture infectious dose (TCID 50 )/mL. CSFV E2 was detected in infected cell monolayers by immunoperoxidase staining with the HC/TC-26 monoclonal anti-E2 hybridoma supernatant [31] as described elsewhere [32].

Antiviral TNF Assay and JAK/STAT Compound Library Screening
The PEDSV.15 cells seeded in 96-well plates (3 × 10 4 cells/100 µL/well) were treated with small-molecule compounds of the JAK/STAT Compound Library (Targetmol, Wellesley Hills, MA, USA, cat. No. L3700) at two concentrations (0.5 µM and 5 µM) for approximately one hour prior to stimulation with either LPS (100 ng/mL), pTNF (5 ng/mL) or the medium. After a stimulation period of six hours, the cells were infected with CSFV-luc at a multiplicity of infection (MOI) of 0.1 TCID 50 /cell, and after 22 h of cultivation, the cell extracts were assayed for firefly luciferase activity (Firefly Luciferase Assay Kit 2.0, Biotium, Fremont, CA, USA) using a Centro LB 960 luminometer (Berthold Technologies, Bad Wildbad, Germany). Average relative luminescence units (RLU) with standard deviations from triplicate values were calculated. The data obtained from cytotoxic or antiviral compounds were eliminated from the analysis (RLU below 50% of non-stimulated infected cultures).

Target
Target Sequence (gRNA) 1 The PAM sequences are underlined.  Table 3. Oligonucleotides for the amplification of edited genomic regions.

Western Blot Analyses
The cells were lysed with a denaturing lysis buffer composed of 62.5 mM Tris HCl (pH 6.8), 2% sodium dodecyl sulfate (SDS), 10% glycerol and 0.05% bromophenol blue. The proteins were separated using 4-12% gradient SDS-polyacrylamide gel electrophoresis under nonreducing conditions (ExpressPlus, GenScript, Piscataway, NJ, USA) and analyzed by means of Western blotting using PVDF transfer membranes (Immobilon-FL, Merck Millipore, Burlington, MA, USA) and an Odyssey Infrared Imaging System (LI-COR Biosciences, Bad Homburg, Germany). Porcine IRF3 and viral N pro proteins were detected using the rabbit anti-IRF3 and anti-N pro sera as described previously [8,37]. Using the mouse monoclonal Anti-β-Actin Antibody C4 (Santa Cruz Biotechnology, Dallas, TX, USA), β-actin was detected as the loading control.

TNF Inhibits CSFV Replication in Porcine PEDSV.15 Cells and MDM, but Not in the PK-15 and SK-6 Cell Lines
CSFV-infected pigs show elevated serum TNF, and TNF was shown to inhibit replication in cell lines [15,16]. In order to characterize the antiviral activity of TNF against CSFV more extensively, we quantified the effect of TNF of different origin on the replication of CSFV expressing a firefly luciferase reporter (CSFV-luc) in primary porcine cells versus permanent cell lines ( Figure 1). The near-to-primary endothelial cell line PEDSV.15 [25] responded to mTNF with a significant reduction of CSFV-mediated luciferase activity 20 h after infection, which was not observed in the PK-15 and SK-6 cells, two permanent porcine cell lines used commonly to propagate CSFV (Figure 1a). The PEDSV.15 cells stimulated for six hours with increasing concentrations of mTNF, from 0.4 ng/mL to 10 ng/mL, displayed a dose-dependent reduction of CSFV replication, as determined by CSFV-mediated luciferase activity ( Figure 1b) and by titration of infectious viruses from cell culture supernatants (Figure 1c). Notably, the TNF treatment did not affect the viability of PEDSV.15 cells at 20 h or three days post-treatment (Figure 1d). Time-of-addition experiments revealed the highest mTNF-mediated inhibition of CSFV infection after six hours of treatment (Figure 1e). Prolonged overnight TNF treatment of the PEDSV.15 cells did not result in an enhanced antiviral state (Figure 1f). TNF pre-stimulation of MDM did also interfere with CSFV (Figure 1g), although not as strongly as in the PEDSV.15 cells (Figure 1b), without affecting the viability of the cells (Figure 1h). This suggests differences in TNF responsiveness of MDM versus PEDSV.15 cells. In order to examine the specificity of mTNF and pTNF for triggering the antiviral effects observed, we employed adalimumab (Humira, Abbott Laboratories), a neutralizing human anti-hTNF monoclonal antibody [38]. TNF treatment in presence of increasing concentrations of adalimumab reduced specifically the antiviral effect of TNF but not of lipopolysaccharide (LPS), known to induce a TLR4dependent antiviral type I IFN response (Figure 1i). This was observed with both mTNF and pTNF. Notably, mTNF neutralization blocked the antiviral TNF activity completely and specifically. Altogether, these data demonstrate that TNF inhibits CSFV replication in primary porcine cells but not in PK-15 and SK-6.

The Anti-CSFV Activity of TNF Involves JAK/STAT Signaling
The JAK/STAT pathway is the key element of the signaling cascade engaged in response to type I IFN [39]. In order to explore whether this pathway is also involved in the antiviral action of TNF, we targeted JAK/STAT signaling with small-molecule inhibitors. Strikingly, the antiviral effects of pTNF and mTNF were sensitive to the JAK inhibitor ruxolitinib (Figure 2a Figure S1c,d for drug concentrations of 0.5 µM or 5 µM, respectively. JAK/STAT signaling is triggered typically by type I IFNs. Therefore, we assessed whether TNF induces the IFN-β mRNA in PEDSV.15 cells. As expected, elevated IFN-β mRNA levels were detected four hours after pTNF stimulation (Figure 2e). The pTNF-mediated upregulation of the IFN-β mRNA was independent of the JAK inhibitor ruxolitinib, suggesting that pTNF elicits a direct induction of the IFN-β promoter. Despite several attempts, we failed to detect bioactive type I interferon in cell culture supernatants of LPS-and pTNFstimulated PEDSV.15 cells using a firefly luciferase-based MX-promoter assay or a sensitive VSV-luc-based assay. By applying a transient firefly luciferase reporter gene assay for NF-κB-dependent promoter activity (NF-κB-RE), we observed JAK/STAT-independent activation of NF-κB with pTNF-and LPS-stimulated PEDSV.15 cells (Figure 2f). As expected, IFN-β stimulation did not induce the NF-κB response element. TNF-mediated NF-κB activation was sensitive to TPCA-1, a selective small-molecule inhibitor of IκB kinase 2 known to inhibit NF-κB nuclear localization. Ruxolitinib, on the contrary, did not inhibit the pTNFand LPS-mediated NF-κB-dependent promoter activation, indicating that this activation was JAK/STAT-independent. The discrepancy between the JAK/STAT-dependent antiviral activity and the JAK/STAT-independent activation of an NF-κB-dependent promoter suggests that pTNF-mediated induction of NF-κB-dependent pathways is not sufficient to trigger the antiviral effect. In conclusion, we observed that in porcine PEDSV.15 cells, pTNF stimulates a JAK/STAT-specific antiviral response, induces IFN-β mRNA and activates JAK/STAT-independent NF-κB signaling.

The Anti-CSFV Activity of TNF Requires the Type I IFN Receptor, While IRF3 Is Dispensable
In order to explore the roles of the type I IFN receptor and of IRF3 in antiviral IFN-β, LPS and pTNF signaling, we generated IFNAR1-and IRF3-knockout (KO) PEDSV.15 cell lines (IFNAR1-KO and IRF3-KO, respectively) by introducing small genetic deletions within early exons using CRISPR/Cas9 gene editing (Figure 3). We determined the respective genotypes after editing and clonal expansion using PCR combined with Sanger DNA sequencing.  We identified two IFNAR1-KO clones #5 and #23 carrying deletions within the IFNAR1 loci ( Figure 3a) consisting of heterozygous open reading frame disruptions. One allele from each clone encodes an mRNA with an internal deletion after the first 72 codons leading to a frameshift mutation. The other alleles have an in-frame deletion leading to a 39 amino acid (aa) deletion after aa position 72. This resulted in functional disruption of IFNAR1, since the two clones (#5 and #23) lost the capacity to establish antiviral states upon IFN-β, LPS and pTNF stimulation, contrary to the parent PEDSV.15 cells (Figure 3b). One PEDSV.15 clone with two intact wild-type (WT) IFNAR1 loci called IFNAR1-WT#4 served as the Cas9-exposed negative control and responded to all three stimuli similar to the parent PEDSV.15 cells (Figure 3b). Collectively, these data confirm that IFNAR1 is necessary for the antiviral activity triggered by IFN-β and LPS and demonstrate the requirement of the type I IFN receptor for the antiviral signaling induced by pTNF.
For IRF3, we identified three knockout PEDSV.15 clones with identical out-of-frame homozygous deletions of 190 nucleotides within the IRF3 open reading frame on the deduced mRNA level, leading to a frameshift mutation after the first 38 codons. The IRF3-KO clones #4 and #16 (Figure 3c) served for functional analyses (Figure 3d). During isolation of IRF3-KO clones, we did not obtain any unedited Cas9-exposed negative control. However, since we performed the stimulations in parallel with the IRNAR1-KO clones, the unedited IFNAR1-WT#4 cells (Figure 3b) served as the Cas9-exposed negative control for the IRF3-KO cells. As expected, IRF3-KO cells maintained their capacity to respond to IFNAR1-dependent IFN-β stimulation (Figure 3d). Importantly, the disruption of IRF3 resulted in a fundamental difference between LPS-and pTNF-triggered antiviral innate immune responses. While the LPS-mediated antiviral state was mostly abolished in IRF3-KO cells, IRF3 was completely dispensable for the pTNF-mediated anti-CSFV activity ( Figure 3d). This highlights the mechanistic differences in the initiation of innate immune responses between pTNF-and LPS-triggered signaling.

The Anti-CSFV Activities of LPS and TNF Are IRF1-Dependent
Besides IRF3, IRF1 can also trigger type I IFN induction [20]. Therefore, we generated functional IRF1-KO PEDSV.15 cells using CRISPR/Cas9 to test whether pTNF-mediated antiviral responses require IRF1. We obtained two IRF1-KO clones-IRF1-KO#2 and IRF1-KO#12-that carried homozygous genomic deletions without disruption   Two independent knockout clones (#2 and #12) and the Cas9-exposed negative control (IRF1-WT#1) with intact IRF1 were stimulated with pTNF, LPS, IFN-β or the medium for seven hours followed by infection with CSFV-luc at a MOI of 0.1 TCID 50 /cell for 22 h before the cell lysates were processed for firefly luciferase measurement. (c,d) The effect of two hours of stimulations with LPS (c) or pTNF (d) on the expression of IFN-β mRNA normalized to 18S ribosomal RNA was assessed in the PEDSV.15 and IRF1-KO#2 cells. The data in (b) represent the means and the standard deviations of six independent experimental replicas and significant differences compared with the medium (p < 0.05) were calculated with one-way ANOVA and post hoc tests (p-value indicated; ns, nonsignificant). The data in (c,d) represent the means and the standard deviations of three independent experimental replicas. Significant differences compared with the medium (p < 0.05) were calculated with the unpaired, two-tailed Student's t-test (the p-values are indicated; ns, nonsignificant).

The Anti-CSFV Activity of TNF Is NF-κB-Dependent, but NF-κB Can Function Independently of IRF1
The induction of type I IFN depends on NF-κB, as opposed to the downstream IFN-β signaling [40]. Therefore, we tested the role of NF-κB in mTNF-and pTNF-mediated anti-CSFV activity using the NF-κB inhibitor TPCA-1 and included LPS, p(I:C) and IFN-β as control stimulations (Figure 5a). TPCA-1 prevented the antiviral actions of mTNF, pTNF, LPS and p(I:C), but not of IFN-β. These results confirm that NF-κB signaling is required for the induction of antiviral activity; however, it is dispensable for downstream IFN-β signaling. Accordingly, the JAK inhibitor ruxolitinib efficiently blocked mTNF, pTNF, LPS, p(I:C) and IFN-β-driven antiviral effects as expected. In parallel, we measured the activation of NF-κB after stimulation of the PEDSV.15 cells (Figure 5b). Treatment with mTNF-, pTNF-, LPS-and p(I:C), but not with IFN-β, resulted in NF-κB activation, which was JAK/STAT-independent (see also Figure 2f). In contrast, mTNF-, pTNF-, LPS-and p(I:C)-mediated NF-κB activation was sensitive to TPCA-1, demonstrating the specificity of the drug. As demonstrated previously, IRF1-KO cells were unable to induce antiviral actions triggered by TNF and LPS (Figures 4b and 5c), which was also the case for the p(I:C) trigger (Figure 5c). Interestingly, despite impaired induction of an antiviral state in IRF1-KO cells, we noted intact mTNF-, pTNF-, LPS-and p(I:C)-mediated NF-κB responses, implying that IRF1 is not involved in the activation of NF-κB-dependent signaling (Figure 5d). In summary, we observed that TNF-, LPS-and p(I:C)-mediated activation of NF-κB is required for the establishment of antiviral activity against CSFV, but that NF-κB-dependent signals can function independently of IRF1.  cells (a, c) were infected with CSFV-luc seven hours after stimulation. Firefly luciferase (a,c) or firefly and Renilla dual-luciferase activities (b,d) were measured 24 h after infection or six hours after stimulation, respectively. The NF-κB-dependent promoter activity is plotted as fold induction compared to the medium. The data represent the means and the standard deviations of at least three independent experimental replicas. Significant differences compared with the medium (p < 0.05) were calculated with the unpaired, two-tailed Student's t-test (the p-values are indicated; ns, nonsignificant).

CSFV Infection Does Not Interfere with TNF-and LPS-Mediated IFN-β mRNA Induction in PEDSV.15 Cells
CSFV antagonizes the induction of type I IFN by means of N pro through IRF3 targeting [8]. In addition, a recent report showed that N pro inhibits the expression and nuclear translocation of IRF1, thereby suppressing the production of type III IFN [24]. Therefore, we hypothesized that CSFV may antagonize the TNF-induced IRF1-dependent and the LPS-induced IRF1-/IRF3-dependent IFN-β mRNA induction in PEDSV.15 cells. In order to address this, we infected the PEDSV.15, IRF1-KO#2 and IRF3-KO#4 cells with the virulent CSFV strain vEy-37 for 3 days prior to stimulation with pTNF, LPS or p(I:C) and measured the induction of IFN-β mRNA in comparison with the stimulated mock-infected cells (Figure 6a,b). Mock-and CSFV-infected PEDSV.15 cells had comparable IFN-β mRNA levels after pTNF or LPS stimulation (Figure 6a,b). This was different with p(I:C), where pre-infected PEDSV.15 cells had significantly lower levels of IFN-β mRNA than mockinfected cells. Similarly, the p(I:C)-mediated induction of IFN-β mRNA was sensitive to CSFV in IRF1-KO#2 cells (Figure 6a). As expected (see Figure 4c,d), the IRF1-KO#2 cells did not respond with IFN-β mRNA to pTNF or LPS stimulation (Figure 6a). With IRF3-KO#4 cells (Figure 6b), CSFV pre-infection had no significant effect on the pTNF, LPS and p(I:C)-mediated induction of IFN-β mRNA, suggesting that CSFV is unable to interfere with IRF1-dependent antiviral signaling. At the time of LPS or pTNF treatment, all the infected cells were positive for the virus antigen, as shown by immunostaining of the E2 protein (Supplementary Figure S2). Unfortunately, we were unable to detect endogenous IRF1 protein by Western blot analysis. However, as expected from our previous studies with PK-15 cells [8], the IRF3 protein was degraded in the CSFV-infected PEDSV.15 and IRF1-KO#2 cells (Figure 6c), which is consistent with reduced or absent IFN-β mRNA induction upon p(I:C) stimulation (Figure 6a,b). Notably, IRF3 was not detected in IRF3-KO#4 cells, confirming successful genome editing leading to IRF3 protein knockout. Collectively, these findings indicate that CSFV lacks countermeasures to interfere with IRF1-dependent TNFand LPS-mediated type I IFN induction in PEDSV.15 cells.

Discussion
The induction of high levels of proinflammatory cytokines including TNF is a hallmark of severe and hemorrhagic CSF following infection with highly pathogenic CSFV [7]. Several independent studies report secretion of TNF peaking at 100-500 pg/mL serum 4-5 days after infection of pigs with CSFV [12][13][14]. CSFV-infected alveolar macrophages can also secrete up to 1 ng/mL TNF at 16 h after infection [41]. TNF was reported to inhibit CSFV replication in porcine PK-15 cells [16] and may be at the origin of the vaccination failure with live-attenuated CSFV in PRRSV-infected pigs [15]. Therefore, this study aimed at dissecting the intracellular signaling cascade required for the anti-CSFV activity of TNF.
First, we observed that different cell types responded differently to TNF. While TNF induced an antiviral state in the endothelial cell line PEDSV.15 and in the porcine MDM, TNF did not inhibit CSFV replication in the SK-6 and PK-15 cells (Figure 1a). This was unexpected given the TNF-mediated inhibition of CSFV replication in PK-15 cells reported earlier [16]. Differences in the steady-state levels of rate-limiting factors such as IRF1 may explain this discrepancy. A different degree of dedifferentiation in general may be a reason for the difference between the PEDSV.15 and MDM cultures and the permanent cell lines used commonly for the propagation of CSFV. The PEDSV.15 are immortalized porcine aortic endothelial cells that maintained most morphological and functional properties of primary endothelial cells and were therefore proposed to serve as a prototypical alternative to normal endothelial cells [25]. The MDM were primary cells prepared from porcine blood (see materials and methods). These results emphasize the importance of cell typedependent differences in cellular responses to infection. Further investigation may include selected approaches such as comparison of the IRF1 levels and activity, as well as highthroughput differential transcriptomic and proteomic analyses.
Next, we dissected the TNF-IRF1-IFN-β signaling axis in the context of a CSFV infection of PEDSV.15 cells. More specifically, we showed that TNF, similarly to LPS, induces the expression of IFN-β transcripts by activating IRF1 and NF-κB independently. Thereby, TNF triggers type I IFN receptor-dependent JAK/STAT signaling leading to decreased CSFV replication. Synergistic antiviral effects between TNF and IFNs are known to enhance antiviral responses (reviewed in [18]). Although we cannot rule out such cooperative effects with TNF stimulations, we were able to efficiently blunt the antiviral effects of mTNF and pTNF with an anti-TNF neutralizing antibody (Figure 1i), which confirms the specific activity exhibited by the TNF formulations. The antiviral effects mediated by TNF and LPS were similarly sensitive to individual compounds of a JAK/STAT inhibitor library (Figure 2c,d). Importantly, the downstream antiviral effects of TNF were independent of the type III IFN antiviral pathway in the PEDSV.15 cells since there was a strict IFNAR1 requirement (Figure 3b). This may be different in other cell types such as intestinal epithelial cells that induce type III IFNs in an IRF1-dependent manner [24]. In the PEDSV.15 cells, we demonstrated that the TNF-triggered induction of IFN-β transcripts and the resulting antiviral effect rely on intact IRF1 (Figure 4b,d), whereas IRF3 was dispensable (Figure 3d). IRF1 is also required for adequate LPS-mediated induction of IFN-β mRNA and subsequent antiviral activity (Figure 4b,c). However, in contrast to TNF that solely depends on IRF1, LPS also requires functional IRF3 to establish its antiviral effect. The latter is consistent with the well-established MyD88-independent LPS/TLR4 signal transduction pathway (reviewed in [42]). Previous studies conducted with primary macrophages, murine microvascular endothelial cells and rheumatoid fibroblast-like synoviocytes described TNF-mediated IRF1-dependent type I IFN responses [17,19,21], but this study shows for the first time inhibition of CSFV replication through this axis.
Besides IRF1, NF-κB is also required for the TNF-mediated antiviral effect on CSFV ( Figure 5). Li et al. showed that TNF interferes with CSFV replication via the NF-κB signaling pathway as the antiviral TNF effect was lost in p65-silenced PK-15 cells [16]. Our observation that the antiviral effect of TNF is strictly IRF1-and IFNAR1-dependent (Figures 3 and 4) and that NF-κB activation is independent of functional IRF1 ( Figure 5) demonstrates clearly that the induction of the type I IFN pathway is required besides NF-κB activation for interference with CSFV replication.
Interestingly, CSFV did not interfere with TNF-or LPS-triggered IFN-β mRNA induction ( Figure 6) that both depend on IRF1 (Figure 4c,d). However, as expected from our previous studies with PK-15 cells [8], CSFV prevented p(I:C)-mediated IFN-β mRNA induction in the PEDSV.15 and IRF1-KO#2 cells, which was consistent with the absence of IRF3 when N pro was expressed ( Figure 6). These results are not surprising when considering that CSFV targets specifically IRF3 for proteasomal degradation by means of N pro [8] while the TNF-induced anti-CSFV activity we observed in the PEDSV.15 cells was completely independent of IRF3 ( Figure 3d). However, the lack of interference of CSFV with the IRF1-dependent pathway described here is in apparent contradiction with recent data showing that N pro of CSFV inhibits IRF1 expression and nuclear translocation in porcine intestinal epithelial IPEC-J2 cells, thereby suppressing type III IFN production [24]. This latter finding may, however, be a specific feature of IFN-λ producing cells. The fact that TNF exhibits substantial anti-CSFV activity in certain cell types and that TNF is secreted in response to CSFV infection in pigs may imply that CSFV evolved yet unidentified means of antagonizing antiviral signaling triggered by TNF in vivo. This is matter of ongoing and future studies.

Conclusions
Several reports suggest altogether that the antiviral activity of TNF involves NF-κBand IRF1-dependent signaling and type I IFN responses. In order to test this formally for CSFV, we targeted NF-κB, IRF1, IRF3 and IFNAR1-dependent JAK/STAT signaling pharmacologically or by CRISPR/Cas9-mediated gene knockout. The anti-CSFV activity of porcine and murine TNF was inhibited by antibody-mediated TNF neutralization, NF-κB and JAK/STAT inhibitors and was abrogated completely in the IRF1 and IFNAR gene knockout cells but not in the IRF3 gene knockout cells. IRF1 gene knockout prevented TNFand LPS-mediated IFN-β mRNA induction. Interestingly, CSFV did not counteract TNF-or LPS-mediated IFN-β mRNA induction. This is consistent with CSFV targeting IRF3 for proteasomal degradation [8] but is in apparent contradiction with CSFV-mediated inhibition of IRF1-dependent signaling reported recently [24]. The latter may be restricted to specific cell types that induce type III IFN. Whether CSFV does selectively inhibit the antiviral activity of TNF through IRF1 targeting in mucosal cells still needs to be explored. Nevertheless, the findings of this study contribute to a better understanding of the CSF immunopathogenesis and of the virus-host interaction of CSFV. More generally, this knowledge is valuable for the development of antiviral and immunoprophylactic interventions.