RTN3 inhibits RIGI-I-mediated antiviral responses by impairing TRIM25-mediated K63-linked polyubiquitination

Upon viral RNA recognition, the RIG-I signalosome continuously generates IFNs and cytokines, leading to neutrophil recruitment and inflammation. Thus, attenuation of excessive immune and inflammatory responses is crucial to restore immune homeostasis and prevent unwarranted damage, yet few proresolving mediators have been identified. In the present study, we demonstrated that RTN3 is strongly upregulated during RNA viral infection and acts as an inflammation-resolving regulator. Increased RTN3 aggregates on the endoplasmic reticulum and interacts with both TRIM25 and RIG-I, subsequently impairing K63-linked polyubiquitination and resulting in both IRF3 and NF-κB inhibition. Rtn3 overexpression in mice causes an obvious inflammation resolving phenomenon when challenged with VSV, Rtn3-overexpressing mice display significantly decreased neutrophil numbers and inflammatory cell infiltration, which is accompanied by reduced tissue edema in the liver and thinner alveolar interstitium. Taken together, our findings identify RTN3 as a conserved proresolving mediator of immune and inflammatory responses and provide insights into the negative feedback that maintains immune and inflammatory homeostasis. Summary RTN3 is upregulated upon RNA viral infection due to inflammation and ER stress, in turn suppresses antiviral responses by impairing TRIM25-mediated RIG-I K63-linked polyubiquitination and consequently decreases neutrophil populations and inflammatory infiltration, representing a novel mechanism of negative inflammatory resolution.

The innate immune responses can be triggered by pathogen-associated molecular patterns (PAMPs) and 57 then serve as the first line of defense against invading pathogens 1 . Upon viral infection, pattern recognition 58 receptors (PRRs) detect viral RNA, DNA and other viral products and subsequently mediate the activation of 59 downstream signaling pathways 2 3 . Both RIG-I and MDA5, two crucial members of the RIG-I-like receptor 60 (RLR) family, sense cytoplasmic viral RNA and activate antiviral responses. Both RIG-I and MDA5 share the 61 same domain architecture, comprising two N-terminal caspase activation recruitment domains (CARDs), a 62 central DEAD box helicase/ATPase domain and a C-terminal regulatory domain (CTD) 3 . MDA5 has been 63 reported to recognize longer dsRNA molecules (>2 kb), while RIG-I prefers shorter (< 1-2 kb) or 5' 64 triphosphate-containing dsRNA 4 5 . Unlike MDA5, which is conformationally unaltered, upon recognition of 65 dsRNA by the CTD domain, RIG-I unfolds and releases its CARDs from the inward folding state formed by its 66 helicase domain, thereby transforming into an activated state 6,7 . The CARD domain then recruits MAVS and 67 activates the downstream signaling cascade to induce the expression of type I interferons, IFN stimulated 68 genes (ISGs), inflammatory cytokines and/or chemokines 7 . 69 Considering that RIG-I-mediated innate antiviral responses are important, drastic and rapid upon viral RNA 70 recognition, multiple mechanisms have evolved to precisely regulate RIG-I-mediated antiviral signaling to 71 maintain the balance between immunity and tolerance and prevent severe or even fatal unnecessary damage, 72 which may be caused by excessive immune and inflammatory responses 8 . Some modifications of RIG-I 73 reduce its stability. For example, LRRC25 targets RIG-I for ISG15-associated autophagy degradation 9 , while 74 the E3 ubiquitin ligases RNF125, STUB1, c-Cbl and CHIP induce the K48-linked polyubiquitination of RIG-I to 75 promote its proteasome-dependent degradation 10 11 . Alternatively, the posttranslational modification of RIG-I 76 upon K63-linked polyubiquitination has been well shown to regulate RIG-I activation. TRIM25 and Riplet 77 induce this modification in RIG-I 12 13 , while the deubiquitination enzyme USP3 targets the RIG-I CARDs to 78 cleave its K63-linked polyubiquitin 14 . In addition, CKII and PKCα/β phosphorylate the RIG-I CTD and CARDs 79 to negatively regulate its activation 15 16 . 80

81
Although the acute immune response and inflammation defend against viral infection and eliminate 82 virus-induced damage, excessive, unrestricted responses can lead to excessive tissue injury and even organ 83 failure. Several diseases have been shown to be related to chronic inflammation or autoimmunity, such as 84 asthma, arthritis, periodontal disease and neurodegenerative disorders 17 18 19 20 . Inflammation resolution has 85 been shown to be an essential active process 21 , and proresolving mediators regulate inflammation resolution 86 through multiple strategies, including the inhibition of signal transduction, the regulation of cytokines and 87 chemokines, the cessation of leukocyte infiltration and the clearance of apoptotic cells 19  VSV-eGFP or poly(I:C) treatment towards inflammatory induction in stimulated cells ( Fig. 1 B, 1 C, bottom). 114 Interestingly, we observed that TNF-α could upregulate RTN3 expression at both the protein (Fig. 1 D) and 115 mRNA levels ( Fig. 1 E). These results suggest that RTN3 levels are increased during RNA viral infection. 116 Since RTN3 is conserved and ubiquitously expressed in various tissues, we further assessed the pattern of 117 RTN3 upregulation in multiple cell lines, including A549 and THP-1 cells. As expected, the protein ( Fig. S1 A, 118 S1 B) and mRNA (Fig. S1 C) levels of RTN3 were both significantly increased by poly (I:C) stimulation or  119 VSV-eGFP infection in A549 cells, with a similar phenomenon observed in THP-1 cells (Fig S1 D, S1 E, S1 F). 120 Notably, confocal microscopy analysis of HeLa cells transfected with mCherry-tagged RTN3 followed by 121 challenge with poly(I:C) showed that RTN3 was localized on the endoplasmic reticulum (ER), while under 122 poly(I:C) stimulation, RTN3 proteins converged on the ER and formed many aggregated bodies of varying 123 sizes ( Fig. S1 G). Taken together, these data suggest that RTN3 expression is induced by RNA viral infection 124 and its observed upregulation and self-aggregation indicate that RTN3 may be involved in regulating innate 125 immune and inflammatory responses. 126 127

RTN3 overexpression suppresses antiviral immune responses 128
To investigate the function of RTN3 in regulating innate immune responses, we performed luciferase 129 assays using ISRE-luc, IFNβ-Luc and NF-κB-Luc reporters and observed that RTN3 markedly inhibited all 130 activities induced by RIG-I overexpression (Fig. 2 A). In addition, poly(I:C)-or Sendai virus (SeV)-stimulated 131 ISRE-luc activities were both restrained by RTN3 overexpression in a dose-dependent manner (Fig. 2 B). 132 However, RTN3 overexpression had a slightly negative effect on MDA5-induced ISRE-luc activities (Fig. S2 A) 133 and barely inhibited TLR3-induced ISRE-luc activities (Fig. S2 B). These data reveal that RTN3 primarily 134 suppresses RIG-I-mediated antiviral immune responses. 135 To further demonstrate the attenuation of antiviral activities by RTN3, HEK293T cells were transfected with 136 an RTN3-encoding plasmid and subsequently infected with VSV-eGFP. Fluorescence microscopy and flow 137 cytometry results showed that RTN3 overexpression rendered the cells highly susceptible to viral infection 138 compared to the empty vector-transfected cells, which displayed a less sensitive phenotype (Fig. 2 C, 2 D). A 139 consistent result was also observed in long-term puromycin-selected, stable HA-tagged RTN3 vs. HA-tagged 140 empty vector-expressing THP-1 cells (THP-1 HA-RTN3 vs. THP-1 HA-Ev ) ( Fig. S2 D), which excluded the 141 possibility that ectopic RTN3 expression may transiently overload the endoplasmic reticulum (ER) and impair 142 the translation of antiviral proteins. 143 To further assess the role of RTN3 in attenuating antiviral activities, we transfected HEK293T cells with the 144 minimum amount of TBK1-, NF-κB p65-, and IRF3-encoding plasmids as primers together with RTN3 or 145 empty vector and followed by infection with VSV-eGFP. The phosphorylation of IKKα/β, TBK1, p65 and IRF3 146 was dramatically decreased in RTN3-overexpressing cells compared to that observed in the control groups 147 Next, we investigated the potential mechanism for the RTN3-mediated inhibition of antiviral responses. We 163 noted that RTN3 may function as a tripartite motif-containing protein 25 (TRIM25)-binding protein through 164 mass spectrometry analysis 28 , and TRIM25 is a crucial E3 ligase for the K63-linked polyubiquitination 165 modification of RIG-I to promote its activation upon RNA viral infection 12 . To elucidate the relationship 166 between RTN3 and TRIM25 or RIG-I, HEK293T cells were transfected with GFP-tagged RTN3 and 167 Flag-tagged TRIM25 followed by stimulation with or without VSV and then subjected to 168 coimmunoprecipitation and immunoblot analysis. The results showed that RTN3 interacted with TRIM25 169 under basal conditions and was slightly strengthened upon viral infection ( Fig. 3 A, Fig. S3 A). Considering 170 that the RTN3 overexpression leads to its aggregation in the ER and may cause nonspecific interactions 171 (such as structural enfolding), we used GFP-TRIM25 to pull down endogenous RTN3 and confirmed its 172 interaction with TRIM25 and that this interaction was strengthened by VSV infection (Fig. 3 B). Since TRIM25 173 directly targets RIG-I, we speculated that RTN3 may also interact with RIG-I. Therefore, we performed 174 coimmunoprecipitation assays and observed an interaction between TRIM25 and RIG-I ( Fig. 3 C) as well as 175 a similar VSV-promoting pattern as that detected for RTN3 interacting with TRIM25 ( Fig. S3 B). Furthermore, 176 confocal microscopy results showed the colocalization of RTN3 with RIG-I or TRIM25 on the extended ER 177 framework ( Fig. 3 D), indicating that RTN3 aggregation may provide a scaffold for TRIM25 oligomerization or 178 the RIG-I signalosome complex. Collectively, these results suggested that RTN3 interacts with both RIG-I and 179 TRIM25 and that this interaction could be enhanced by RNA viral infection. 180 181

RTN3 impairs the K63-linked polyubiquitination of RIG-I upon RNA viral infection 182
We next assessed how RTN3 suppresses RIG-I-mediated antiviral responses through its interaction with 183 TRIM25. Interestingly, the protein levels of RIG-I and TRIM25 were not decreased by RTN3 overexpression 184 in HEK293T cells (Fig. S3 C, S3 D), excluding the possibility that RTN3 may target RIG-I or TRIM25 for their 185 degradation. We also assessed whether the interaction between RIG-I and TRIM25 was inhibited or 186 promoted by RTN3 overexpression. To this end, HEK293T cells were cotransfected with Flag-RIG-I-and 187 GFP-TRIM25-expressing constructs or GFP-tagged empty vector (GFP-Ev) together with increasing 188 amounts of RTN3-expressing plasmid, which was followed by VSV infection. Unfortunately, the interaction 189 between RIG-I and TRIM25 was completely unchanged (Fig 3 F

RTN3 inhibition of antiviral responses is TRIM25-dependent 204
To elucidate whether TRIM25 promotes the ability of RTN3 to inhibit RIG-I-mediated immune responses, 205 we further investigated whether TRIM25 deficiency abolishes the inhibition of RIG-I-mediated antiviral 206 immune responses. To this end, endogenous TRIM25 expression was depleted in HEK293T cells by shRNA 207 (HEK293T-shTRIM25), after which the phosphorylation of IKKα/β, TBK1 and p65 was detected in both 208 HEK293T-shTRIM25 and HEK293T-shCtrl cells. Compared to the HEK293T-shCtrl cells, TRIM25 depletion 209 counteracted the inhibition of IKKα/β, TBK1 and p65 phosphorylation induced by RTN3 overexpression when 210 challenged with VSV-eGFP (Fig. 4 A). Similarly, RTN3 overexpression in HEK293T-shTRIM25 cells only 211 slightly impaired RIG-I K63-linked polyubiquitination upon VSV infection (Fig. 4 B). Taken together, these data 212 suggest that TRIM25 has a crucial role in the RTN3-mediated suppression exerted of RIG-I activation. 213 Subsequently, we evaluated which domain of RTN3 is responsible for its inhibitory activity towards 214 RIG-I-mediated antiviral responses. RTN3 localizes to the ER membrane and comprises a noncytoplasmic 215 domain (NC) that inserts into the ER lumen, three transmembrane domains (TM) and two cytoplasmic 216 domains. A series of RTN3 truncations were constructed from the C terminus based on its domain structure 217 (Fig. S4 A) and were then assessed for their ability to suppress RIG-I-mediated immune activation through 218 ISRE-luc and NF-κB-luc reporter assays. The TM1-3-truncated construct (T4 mutant) was unable to inhibit 219 RIG-I (N)-induced ISRE-luc activity, while other truncated mutants maintained similar inhibitory activities as 220 full-length RTN3 (Fig. S4 B, top). The same phenomenon was also observed in NF-κB-luc reporter assays 221 (Fig. S4 B, bottom). These results suggest that the first TM domain is required for the inhibitory activity of 222 RTN3. Furthermore, ubiquitination analysis demonstrated that K63-linked polyubiquitination was impaired by 223 the RTN3 mutants T1, T2 and T3 at levels similar to wild-type RTN3, whereas T4 mutants lacked this activity 224 (Fig. S4 C). Collectively, our data suggest that RTN3 inhibits RIG-I-mediated innate immune antiviral 225 responses in a TRIM25-dependent manner and that transmembrane domain 1 (aa 61-92) of RTN3 is 226 indispensable for this activity. 227 228

RTN3 overexpression suppresses antiviral immune responses in mice 229
We next assessed whether RTN3 is upregulated and inhibits RIG-I-mediated innate immune antiviral 230 responses in vivo during viral infection by treating C57BL/6 mice with PBS, poly(I:C) or VSV-eGFP separately 231 via intravenous (I.V.) injection. Real-time PCR results showed that the mRNA levels of Rtn3 in liver extracts 232 were significantly higher in the VSV-and poly(I:C)-treated mice than in those of the PBS control mice (Fig. 5  233 A), and Rtn3 protein levels were also markedly upregulated (Fig. 5 B). These results confirm that RTN3 234 expression is significantly induced by RNA viral infection in mice, at least in the liver. 235 To further investigate whether RTN3 upregulation inhibits antiviral responses in vivo, mice were 236 hydrodynamically injected with a plasmid encoding Rtn3 (HA-Rtn3) or the empty vector (HA-Ev) and then 237 inoculated with VSV-eGFP via intraperitoneal (I.P.) injection to infect mice for another 12 h (Fig. 5 C). At 48 h, 238 as expected, appropriate levels of plasmid-mediated RTN3 overexpression were detected in the livers of 239 mice (Fig. 5 D). After the mice were sacrificed and then bled, the serum samples were incubated with 240 precoated beads, and flow cytometry results showed that the levels of key inflammatory factors, including 241 Mcp-1, IL-12p70, IFN-γ and TNF-α were consistently decreased in the Rtn3-overexpressing mice upon viral 242 infection, especially IFN-γ and TNF-α, both of which were markedly downregulated (Fig. 5 E). We next 243 analyzed the immune cell population within the liver and PBMCs through flow cytometry analysis. Cells were 244 grouped as indicated (Fig. S5 A, S5 B), and upon VSV-eGFP infection, the population of CD11b + Gr1 + 245 neutrophils (Fig 5 F, 5 G) dramatically decreased in the Rtn3-overexpressing mice, while other immune cell 246 populations, including CD11 + Ly6C hi F480 + macrophages, Ly6C hi F4/80 lo monocytes, CD45 + CD3 + T cells and 247 CD45 + CD3 -CD11b + CD11c + dendritic cells, showed no significant change (Fig. S5 C). Taken together, these 248 results indicate that upon RNA viral infection, RTN3 is upregulated and subsequently suppresses antiviral 249 immune and inflammatory responses, decreasing the population of neutrophils in mice. 250 251

RTN3 overexpression suppresses inflammation in mice 252
Based on the observed inhibitory activity of RTN3 in mouse experiments, we speculated that RTN3 may 253 promote the resolution of inflammation in vivo. To assess this possibility, the liver and lung tissue sections of 254 mice were stained with hematoxylin and eosin (H&E), and microcopy analysis showed that upon VSV 255 infection, RTN3 overexpression (indicated by RTN3 IHC) alleviated inflammatory cell infiltration and tissue 256 edema in the liver (Fig. 6 A top and middle, 6 B, 6 C), which was accompanied by a thinning of the alveolar 257 interstitium in the lung (Fig. 6 A bottom, 6 D). These results indicated that RTN3 upregulation is induced by 258 RNA viral infection to attenuate inflammation and protect against tissue injury caused by excessive 259 inflammatory responses. Further experiments evaluating VSV-eGFP infection in the liver revealed that the 260 8 VSV-infected cell population was increased in RTN3-overexpressing mice (Fig. S6 A, S6 B In assessments of whether RIG-I is the predominant target through which RTN3 inhibits antiviral responses, 279 our results demonstrated that RTN3 separately interacts with RIG-I and TRIM25 and that their interaction is 280 augmented upon VSV stimulation. Furthermore, we showed that RTN3 overexpression impairs the 281 K63-linked polyubiquitination of RIG-I in a TRIM25-dependent manner but does not interfere with RIG-I 282 stability or interaction. Notably, the K63-linked polyubiquitination of TRIM25 was promoted while that of RIG-I 283 was inhibited by RTN3 overexpression that led to RTN3 self-aggregation and the formation of a scaffold-like 284 structure, indicating that RTN3 may interfere with the oligomerization of TRIM25 and its E3 ubiquitin ligase 285 activity to promote RIG-I polyubiquitination. Furthermore, mapping analysis of truncated RTN3 constructs 286 showed that RTN3 transmembrane domain 1 (TM1) is crucial for impairing RIG-I K63-linked 287 polyubiquitination and inhibiting immune responses, indicating that a proper topological structure of RTN3 on 288 the ER membrane is essential for its activity. Thus, the underlying mechanism of how RTN3 interferes with 289 this process is worth further investigation. 290

291
Consistent with the in vitro results, in vivo experiments in mice showed that RTN3 is upregulated upon VSV 292 infection or poly(I:C) stimulation. Importantly, our results provide evidence that RTN3 overexpression 293 decreases the production of inflammatory factors and the neutrophil population in the liver and blood during 294 RNA viral infection. The native neutrophil population may be primarily affected by the downregulation of 295 cytokines and chemokines, and the circulating neutrophil population is affected secondarily. Because 296 neutrophils have the capacity to damage tissue 30 31 , neutrophil inhibition is crucial in the inflammatory 297 response 22 32 . The suppression of chemokine CXCL8 (IL-8; Fig. S2) levels by RTN3 indicates that it may be 298 a potential mediator of inflammation resolution by indirectly restricting neutrophil recruitment. Although the 299 decreased macrophage populations in Rtn3-ovrerexpressing mice were not significant, we did observe a 300 decreasing trend for these populations (Fig S5 C). The histological results demonstrated that RTN3 301 overexpression attenuated the infiltration of lymphocytes and greatly relieved tissue inflammation and edema. 302 Taken together, these results provide evidence that RTN3 functions to maintain tissue homeostasis and 303 initiate the resolution of inflammation by inhibiting RIG-I-mediated antiviral responses and even terminating 304 acute inflammation. 305 306 Therefore, we propose a working model that describes how RTN3 suppresses RIG-I-mediated antiviral 307 innate immune responses (Fig. S7). Upon RNA viral infection, the RIG-I signalosome undergoes 308 TRIM25-mediated activation and induces the production of IFNs, cytokines and chemokines, leading to acute 309 immune responses and inflammation. Subsequently, RTN3 is upregulated in an ER stress-and/or 310 inflammation-dependent manner, which is directly associated with acute viral infection. Aggregated RTN3 on 311 the endoplasmic reticulum interacts with TRIM25 and RIG-I, inhibiting the K63-linked polyubiquitination of 312 RIG-I via the E3 ligase activity TRIM25 to block the RIG-I signaling cascade and attenuate innate immune 313 responses. As a result, the production of IFNs, cytokines and chemokines is attenuated, and the resolution of 314 host inflammation is simultaneously initiated. 315 316 In summary, the results of our present study demonstrate that a conserved reticulon protein, RTN3, which 317 is ubiquitously expressed in various tissues and organs, is upregulated upon RNA viral infection. In turn, 318 upregulated RTN3 inhibits RIG-I signalosome activation by impairing the TRIM25-RIG-I axis. Our findings 319 demonstrate a novel negative feedback mechanism for acute immune responses and inflammation resolution 320 and the potential reconstitution of tissue homeostasis during viral infection. Plasmids and transfection 367 The plasmids used in the present study were constructed as follows. RTN3 and TRIM25 were obtained 368 from the HEK293T cDNA library and cloned into the eukaryotic expression vector pcDNA3.1 in-frame with an 369 HA or Flag tag as well as the eukaryotic expression vectors pEGFP-C2 and pEBG. The HA-tagged RTN3 370 fragment was also subcloned into the pMSCV-PURO retrovirus expression vector, and the RTN3 fragment 371 was truncated based on its structure and subcloned into pcDNA3.1 to generate the HA-tagged mutants T1, 372 T2 Generation of gene modified-cell lines 381 HEK293T cells were seeded into 6-well plates at a density of 0.5 × 10 6 cells/ml and incubated overnight. 382 Then, 2.5 ml/well of a TRIM25 shRNA-encoding lentivirus was added to the plate, which was then placed in 383 an incubator. Twenty-four hours post inoculation, the cells were infected with a second batch of virus for 384 another 24 h. The infected cells were then reseeded into 6-well plates with fresh RPMI 1640 medium 385 supplemented with 10% FBS and cultivated for 36 h. Puromycin (2-4 μ g/ml, Thermo Fisher, A113803) was 386 used for TRIM25-knockdown HEK293T cell selection, and immunoblot analysis was performed to determine 387 the knockdown efficiency. For TRIM25 knockdown-HEK293T cells generation, the following TRIM25 shRNA 388 sequences were used: (1), 5'-CCGGAACAGTTAGTGGATTTA-3'; (2), 5'-GAACTGAACCACAAGCTGATA-3'. 389 Both with IP buffer (Beyotime, P0013), after which whole cell extracts were collected and incubated with anti-GFP 424 beads or Glutathione Sepharose 4B at 4°C for 1 h or overnight. Then, the beads were washed 4-5 times with 425 IP buffer, and the immunoprecipitants were eluted with 2× SDS loading buffer and then resolved by 426 SDS-PAGE. Subsequently, the proteins were transferred to PVDF membranes (Millipore, ISEQ00010) and 427 further incubated with the indicated primary and secondary antibodies. The images were visualized using an 428 Odyssey Sa system (LI-COR