DDX50 is a viral restriction factor that enhances TRIF-dependent IRF3 activation

The transcription factors IRF3 and NF-κB are crucial in innate immune signalling in response to many viral and bacterial pathogens. However, mechanisms leading to their activation remain incompletely understood. Canonical RLR signalling and detection of viral RNA is dependent upon the receptors RIG-I, MDA5 and TLR3. Alternatively, the DExD-Box RNA helicases DDX1-DDX21-DHX36 activate IRF3/NF-κB in a TRIF-dependent manner independent of RIG-I, MDA5 or TLR3. Here we describe DDX50, which shares 55.6% amino acid identity with DDX21, as a component of the dsRNA sensing machinery and signalling pathway. Deletion of DDX50 in mouse and human cells impaired activation of the IFNβ promoter, IRF3-dependent endogenous gene expression and cytokine/chemokine production in response to cytoplasmic dsRNA (polyIC transfection), and infection by RNA and DNA viruses. Mechanistically, DDX50 co-immunoprecipitated with TRIF and DDX1, promoting complex formation upon stimulation. Furthermore, whilst MAVs/TBK1 induced signalling is intact in Ddx50 KO cells, TRIF-dependent signalling was impaired suggesting DDX50 drives TRIF-dependent Ifnβ transcription. Importantly, loss of DDX50 resulted in increased replication and dissemination of vaccinia virus, herpes simplex virus and Zika virus highlighting its important role as a viral restriction factor. Author summary The detection of viral RNA or DNA by host RNA or DNA sensors and the subsequent antiviral immune response are crucial for the outcome of infection and host survival in response to a multitude of viral pathogens. Detection of viral RNA or DNA culminates in the upregulation of inflammatory cytokines, chemokines and pathogen restriction factors that augment the host innate immune response, restrict viral replication and clear infection. The canonical RNA sensor RIG-I is a member of the large family of DExD/H-box helicases, however the biological role of many DExD/H-box helicases remain unknown. In this report, we describe the DExD-Box helicase DDX50 as a new component of the RNA sensing machinery. In response to DNA and RNA virus infection, DDX50 functions to enhance activation of the transcription factor IRF3, which enhances antiviral signalling. The biological importance of DDX50 is illustrated by its ability to restrict the establishment of viral infection and to diminish the yields of vaccinia virus, herpes simplex virus and Zika virus. These findings increase knowledge of the poorly characterised host protein DDX50 and add another factor to the intricate network of proteins involved in regulating antiviral signalling in response to infection.


Introduction 45
Interferon regulatory factor 3 (IRF3)-dependent signalling leading to type I interferon (IFN) 46 expression is crucial for pathogen clearance and host survival in response to infection by many viral and bacterial pathogens (1,2). IRF3 signalling is tightly regulated and is triggered 48 to infected WT cells (Fig. 3A-B). This correlated with decreased secretion of CXCL10 and IL-150 6 as determined by ELISA (Fig. 3C-D). Although significant, the effect of Ddx50 KO on 151 Cxcl10 expression by RT-qPCR following infection with both viruses was less pronounced 152 ( Fig. 3A-B). Overall, this highlights the importance of DDX50 in innate immune signalling 153 during viral infection. 154

Loss of Ddx50 does not alter IL-1α or TNFα-mediated NF-κB activation 155
Deletion of Ddx50 impaired the induction of NF-κB/IRF3-co-transcribed genes in response to 156 dsRNA transfection, ssRNA virus infection (Fig. 1) and dsDNA virus infection (Fig. 2) and 157 previously was reported to modulate MAP kinase signalling (22). Therefore, alternative 158 pathways were tested to determine if the observed defect was specific to RLR signalling. WT 159 or KO MEFs were treated with IL-1α or TNFα and activation of the NF-κB promoter or 160 expression of Il-6 and Nfkbia were measured by Luciferase reporter gene assay or RT-161 qPCR, respectively. No differences in NF-κB promoter activity or NF-κB-dependent gene 162 expression were observed ( Fig. 4A-C or without prior pathway stimulation was used to assess the subcellular localisation of 178 DDX50 in response to cytoplasmic dsRNA. LaminA/C and α-tubulin served as nuclear and 179 cytoplasmic fraction controls, respectively. As described, under resting conditions, the 180 majority of DDX50 was in the nuclear fraction (21) (Fig. 5C). However, DDX50 accumulated 181 in the cytoplasm 1 h post-stimulation ( Fig. 5C). At 2 h post-stimulation the level of DDX50 in 182 the cytoplasm returned to basal levels (Fig. 5C). To support this finding the assay was 183 repeated and the localisation of HA-tagged DDX50 was analysed by immunofluorescence. 184 Under resting conditions DDX50 was restricted to the nucleolus, with weak nuclear staining. 185 In agreement with the biochemical fractionation assay, accumulation of DDX50 in distinct 186 cytoplasmic puncta was observed 1 h post infection with SeV (Fig. 5D). 187 The nucleocytoplasmic shuttling of DDX50 upon stimulation led us to investigate at which 188 stage in the activation of the IRF3/NF-κB pathway DDX50 might function. The IRF3/NF-κB 189 pathway can be activated by transfection and overexpression of key proteins acting at specific stages of the pathway. Therefore, to map in more detail where DDX50 acts, 191 plasmids encoding TBK1, MAVS or TRIF were co-transfected into WT or KO MEFs along 192 with pLuc-IFNβ and pTK-RL. Activation of the pathway was measured by Firefly and Renilla 193 Luciferase activation as before. No differences in fold activation were observed between the 194 WT and KO cells upon expression of TBK1 or MAVS (Fig. 5E). However, activation was 195 significantly impaired in the KO cell line upon expression of TRIF, mapping DDX50 upstream 196 to or independent of MAVS, but at or downstream of TRIF activation. Notably, DDX50 197 shares 55.6 % amino acid identity with DDX21, which is essential for TRIF recruitment to 198 MAVS via complex formation with DDX1 and DHX36 in response to cytoplasmic dsRNA 199 (15). 200

DDX50 augments TRIF recruitment to activate signal transduction 201
An essential TRIF-binding domain of DDX21 has been mapped to residues 467-487 within 202 the RNA helicase C domain (15). Strikingly, this motif shares 86% amino acid identity with 203 DDX50 (Fig. 6A). This level of homology was specific for DDX50 and not due to the helicase 204 C domain consensus sequence, because it was not detected within other DExD-box family 205 members such as DHX36 (Fig. 6A). Due to the high level of identity between DDX21 and 206 DDX50 we investigated whether DDX50 can co-immunoprecipitate the DDX1-DDX21-207 DHX36-TRIF complex. To this end, co-immunoprecipitation assays were performed using 208 extracts of MEF cell lines that stably expressed DDX50-HA and that were transfected with 209 TRIF-cTAP or GFP-Flag. Following stimulation, DDX50-HA specifically co-210 immunoprecipitated TRIF-cTAP (Fig. 6B). This was confirmed by reciprocal 211 immunoprecipitation in HeLa cells, where hDDX50-HA specifically co-immunoprecipitated 212 with TRIF-cTAP (Fig. 6C). Due to the quality of available anti-TRIF antibodies, co-213 immunoprecipitation of endogenous TRIF could not be tested. However, endogenous DDX1 214 did co-immunoprecipitate with DDX50-HA, albeit at low levels ( Fig. 6D). This led to the 215 hypothesis that DDX50 may form a cytoplasmic RNA sensing complex with TRIF, to activate 216 TRIF-dependent NF-κB and IRF3 activation. To test this hypothesis, the ability of TRIF to form a complex with DDX1 in WT or DDX50 KO cells was investigated. Interestingly, in the 218 absence of DDX50 the co-immunoprecipitation of endogenous DDX1 with TRIF was 219 diminished, indicating DDX50 may facilitate optimal TRIF recruitment to the RIG-I/MDA5 220 independent RNA sensing complex (Fig. 6E). 221

DDX50 is a viral restriction factor 222
IRF3 is a crucial viral restriction factor that controls the transcriptional upregulation of Monolayers of WT or KO MEFs/HEK293Ts were infected at MOI 0.0001 or 0.0003 with A5-242 GFP-VACV or at MOI 0.01 with HSV-1 strain 17 (S17) encoding GFP fused to Vp26 (Vp26-243 approximate 6-and 3.5-fold increase in the yield of VACV at 24 and 48 h p.i., respectively 245 ( Fig. 8A-B). This difference was not restricted to VACV, and loss of DDX50 resulted in an 246 increase in yield of HSV-1 following low MOI (Fig. 8C). In line with higher viral titres, 247 synthesis of the VACV specific late gene product D8 was enhanced in KO MEFs (Fig. 8F). 248 Notably, the number of plaques formed by VACV was increased on the KO MEFs and 249 HEK293Ts compared to control cells ( colleagues did not report on DDX50. However, this was performed in mouse dendritic cells 296 and the expression of DDX50 varies from cell type to cell type (human atlas data). It may be 297 that DDX50 plays a more significant role in non-haematopoietic cells or that it was below 298 detection in the initial screen. Alternatively, the high sequence identity between DDX50 and 299 DDX21 may have masked the role of DDX50 following knockdown of DDX21. In the DDX1-300 DHX36-DDX21 complex, DDX21 acts as a scaffold to recruit TRIF upon DDX1 agonist 301 binding. Therefore, we hypothesise that DDX50 may act in a similar fashion. Notably, whilst 302 DDX21 and DDX50 have non-redundant roles, DDX50 is essential for DDX21 helicase 303 activity in vitro (21). So, it is possible that DDX50 may function to support DDX21 or even 304 DDX1 (the RNA binding protein) activity in this complex, which may explain why DDX50 is suggest that it may inhibit DENV replication (28). Following knockdown, the authors reported a reduction in IFNβ promoter activity and therefore hypothesised that DDX50 may regulate 322 type I IFN production during DENV infection (29). This is consistent with our findings that the 323 ZIKV titre is increased in the absence of DDX50, and together provides evidence that 324 DDX50 is a viral restriction factor in response to multiple RNA and DNA viruses. Therefore, 325 DDX50 as a restriction factor may extend beyond the viruses tested in this study and act 326 broadly to detect viral RNA and restrict viral replication through activation of IRF3-dependent 327 gene transcription. 328 Although DDX50 was required for optimal signal transduction its absence did not abolish 329 signalling in response to viral infection or stimulation. Given that DDX50 binds TRIF, a 330 protein that is non-essential for RIG-I/MDA5 signalling, and that DDX1 acts independent of 331 canonical RIG-I signalling, we propose that DDX50 acts in concert with other receptors for 332 optimal antiviral signalling and restriction. Whilst this study identifies a role for DDX50 in 333 in activating host defences is illustrated by the fact that VACV, despite being a dsDNA virus, 344 encodes a dsRNA binding protein called E3 (38), that contributes to virulence (39). It is 345 important to note that TRIF is also an essential component of the STING pathway (31). 346 Therefore, the level to which DDX50 restricts DNA viruses in an RNA-sensing dependent 347 manner, or whether it can further influence TRIF signalling in the cGAS-STING pathway, warrants future investigation. Furthermore, the importance of DDX50 in RNA sensing and its 349 contribution in antiviral immunity requires validation in vivo. Unfortunately, to date there are 350 no KO mice or models available, however with the recent success in generating Ddx21 KO 351 mice, it may soon be a plausible avenue for investigation. 352 In conclusion, the DExD-Box RNA helicase DDX50 is identified as a crucial component in 353 the host cell RNA sensing machinery, acting to facilitate IRF3 activation and inhibit viral 354 dissemination. It is proposed that DDX50 may act through the recruitment of TRIF to the 355 DDX1 RNA sensing complex.

Acknowledgements 357
We would like to thank Dr. B.J. Ferguson, University of Cambridge and Dr. A. Shenoy, 358 Imperial College London, for their helpful feedback and advice and Dr. Trevor Sweeney, 359 University of Cambridge for providing ZIKV for this project. pCW57-GFP-2A-MCS was a gift 360 from Adam Karpf (Addgene plasmid #71783). 361

CRISPR-cas9 generation of knockout cell lines 381
Guide RNA design and synthesis, and pX459 plasmid construction was performed following 382 the Zhang lab protocol (45). Specific guide RNAs are described in Table S1. To generate 383 KOs, MEFs were transfected with pX459 plasmids using LT1 following the manufacturer's extraction, cDNA synthesis and RT-qPCR were carried out as described previously using 434 first strand synthesis (Invitrogen) (47). qPCR was performed using the primers indicated in 435   Table S2.