A Comparative Analysis of Coronavirus Nucleocapsid (N) Proteins Reveals the SADS-CoV N Protein Antagonizes IFN-β Production by Inducing Ubiquitination of RIG-I

Coronaviruses (CoVs) are a known global threat, and most recently the ongoing COVID-19 pandemic has claimed more than 2 million human lives. Delays and interference with IFN responses are closely associated with the severity of disease caused by CoV infection. As the most abundant viral protein in infected cells just after the entry step, the CoV nucleocapsid (N) protein likely plays a key role in IFN interruption. We have conducted a comprehensive comparative analysis and report herein that the N proteins of representative human and animal CoVs from four different genera [swine acute diarrhea syndrome CoV (SADS-CoV), porcine epidemic diarrhea virus (PEDV), severe acute respiratory syndrome CoV (SARS-CoV), SARS-CoV-2, Middle East respiratory syndrome CoV (MERS-CoV), infectious bronchitis virus (IBV) and porcine deltacoronavirus (PDCoV)] suppress IFN responses by multiple strategies. In particular, we found that the N protein of SADS-CoV interacted with RIG-I independent of its RNA binding activity, mediating K27-, K48- and K63-linked ubiquitination of RIG-I and its subsequent proteasome-dependent degradation, thus inhibiting the host IFN response. These data provide insight into the interaction between CoVs and host, and offer new clues for the development of therapies against these important viruses.

SADS-CoV has the genome order typical of CoVs, consisting of seven independent ORFs that encode 16 non-structural proteins, 4 structural proteins and one accessory protein (7,14). Among the structural proteins, the nucleocapsid (N) protein is highly expressed and has functions at multiple steps during viral infection (15). The N protein is highly conserved between different CoVs and is indispensable in the viral life cycle. It forms nucleocapsids with genomic RNA, promotes viral genome replication and subgenomic RNA transcription, and interacts with other viral proteins (such as the membrane [M] protein) to promote virion assembly (15). Additionally, an increasing number of studies suggest that the N protein is involved in viral evasion of the host innate immune response (16)(17)(18)(19)(20)(21)(22).
Innate immunity represents the first line of defense against pathogens and includes the type I interferon (IFN) signaling pathway, which plays an essential role in protection against viral infection (23). The IFN response starts with the recognition of pathogen associated molecular patterns (PAMPs) by pattern recognition receptors (PRRs). As RNA viruses, CoVs produce PAMPs including dsRNA and 5′-ppp RNA intermediates in the cytoplasm during replication, which can then be recognized by PRRs, specifically retinoic acid-inducible gene I (RIG-I)-like receptors (RLR) (24). After recognition and subsequent activation of RIG-I, an adaptor localized on the surface of mitochondria (mitochondrial antiviral signaling protein; MAVS) is activated by interaction between their caspase activation and recruitment domains (CARDs). MAVS then forms prion-like polymers and recruits TBK1 and IKKϵ (and other components) to form a complex that induces phosphorylation of IRF3, which subsequently promotes the production of type I IFN. This IFN production leads to expression of hundreds of IFN-stimulated genes (ISGs) in an autocrine and paracrine manner and keep the host cells in an antiviral state (25). Viruses, on the other hand, have evolved various strategies to antagonize and even benefit from the antiviral pathways of the host cell (26)(27)(28).
To elucidate their roles in IFN suppression during infection, we analyzed the amino acid similarities between N proteins from several representative CoVs of four different genera, and compared targets of each N protein in IFN signaling. More importantly, we studied the mechanism of IFN inhibition by the SADS-CoV N protein by a comparative analysis. Our data show that the PAMP recognition step is a critical target for N protein suppression of IFN signaling. We also demonstrate that SADS-CoV N protein interacts with RIG-I and inducing its ubiquitination, which leads to its proteasome-dependent degradation and, consequently, the inhibition of the host IFN responses. Our data reveals one of the mechanisms by which SADS-CoV suppresses host innate immunity and provides novel clues for treatment and vaccine development against CoV infections.

Luciferase Reporter Gene Assay
293T cells were seeded in 12-well plates at 70% confluence and then cotransfected with 500 ng of indicated N protein expression plasmid, 100 ng IFN-b-Luc and 50 ng RL-TK (Renilla luciferase positive control) per well. 24 h later, cells were lysed by 100 ml lysis buffer (25 mM Tris-HCl, 200 mM NaCl, 10 mM NaF, 1 mM Na 3 VO 4 , 25 mM b-glycerophosphate, 1% NP40 and protease inhibitor cocktail), and 10 ml lysate of each well was subjected to dual luciferase reporter gene assay following the manufacturer's instructions.

Western Blotting
293T cells were transfected with indicated plasmids as mentioned above and harvested after 24 h by addition of lysis buffer, and incubated on ice for 15 min. Cell lysates were centrifuged at 12,000 ×g for 15 min before the supernatants were either subjected to immunoprecipitation (IP) or denatured directly at 100°C for 10 min. Denatured cell lysates were separated by SDS-PAGE and transferred to PVDF membranes using 200 mA for 70 min. For immunoblotting, indicated primary antibodies were used to incubate the membranes for 3 h at RT or overnight at 4°C; HRP-conjugated goat anti-mouse or goat anti-rabbit IgG were used as secondary antibodies. The bands were visualized with chemiluminescent reagent (#1705061, Bio-Rad, Hercules, CA, USA) and were imaged by an ECL imaging system (LI-COR biosciences), and amounts of b-actin, GAPDH or TPB of each sample were used as controls to demonstrate equal loading of protein samples among lanes.

Co-Immunoprecipitation (IP) Assay
293T cells were seeded in 6-well plates and transfected with indicated plasmids as mentioned above and lysed at 36 h posttransfection (hpt) in 300 ml lysis buffer per well. Lysates were centrifuged at 12,000 ×g for 15 min, and the supernatants were subjected to IP, with 40 ml reserved as a whole cell lysate (WCL) control. Briefly, 1 ml of antibody was added to 200 ml cell lysate and incubated on a shaker at 4°C for 2 h, followed by addition of dynabeads protein G (#10004D, Invitrogen, Carlsbad, CA, USA) which had been washed by lysis buffer three times, and then incubated for another 2 h at 4°C. The mixture was centrifuged at 5,000 ×g for 90 s and the beads were collected magnetically and washed in lysis buffer three times. Next, the beads were mixed with loading buffer and the mixture was denatured at 100°C for 10 min before analysis along with the WCL by western blot.

Fractionation of Cytoplasmic and Nuclear Proteins
293T cells were seeded in 6-well plates at 70% confluence and transfected with indicated plasmids as mentioned above, and then harvested at 24 hpt. Cytoplasmic and nuclear proteins were fractionated and extracted by using a nuclear and cytoplasmic protein extraction kit (Thermo Scientific) following the manufacturer's instructions. Extracted proteins were quantified using a BCA protein assay kit (Thermo Scientific, #23225), and an equal amount of sample was used for western blot analysis.

Statistical Analyses
Data were analyzed by GraphPad Prism 6 software and Student's t-tests were performed to determine significance. Data were expressed as mean ± standard deviation (SD) of three independent experiments. P values < 0.05 were considered to be statistically significant.

Sequence Comparison of CoV N Proteins
To analyze their conservation, the SADS-CoV N protein was compared with those of other CoVs using ClustalW, and amino acid sequence similarities were calculated. The full length N protein of SADS-CoV contains 375 amino acids, with the NTD and CTD domains predicted based on sequence alignment with HCoV-NL63 N protein from the genus Alphacoronavirus, which is the only N protein whose structure has been solved so far (48). The putative NTD spans from residue 10 to 144 while the CTD spans from residue 215 to 336 ( Figure 1A), similar to other CoV N proteins in length. To illustrate the evolutionary relationship, we constructed a phylogenetic tree based on N protein amino acid sequences from representative CoVs of each genus (Supplementary Figure S1). As expected, the N protein amino acid sequence identity was consistent with the genus clustering observed for the whole genome ( Figure 1B). Next, we analyzed the amino acid sequence similarities between the full-length, NTD and CTD sequences from the N proteins of various representative CoVs. As shown in Figure 1C, the similarity to SADS-CoV among the alphacoronaviruses was relatively high, 42.5% for PEDV and 41.9% for TGEV. The similarities of N proteins from the other genera ranged from 26.4% for mouse hepatitis virus (MHV) to 16.7% for PDCoV. Notably, there was greater similarity in the NTD and CTD than in the full sequence, suggesting that the intrinsically disordered regions (IDRs) are more divergent. The NTDs were more highly conserved (60.2-29.9%) than the CTDs (43.6-17.6%), although this was not statistically significant (P=0.1342) ( Figure 1C). These data demonstrate that N proteins of CoVs are relative conserved, especially within genus, thus we hypothesized that they might exhibit similar functions with respect to evasion of innate immunity.

CoV N Proteins Inhibit IFN Production by Interfering With Different Stages of the Signaling Pathway
To compare the mechanisms by which CoV N proteins from different genera inhibit IFN production, a dual luciferase reporter gene assay was performed in transfected cells, using a RIG-I activating model virus, SeV, to activate RLR signaling. To ensure that the effects of each N protein from distinct CoVs were comparable, 293T cells were used in all reporter assays. The activity of the IFN-b promoter was inhibited by all the N proteins we tested, including N proteins of alphacoronaviruses TGEV, PEDV, SADS-CoV, betacoronaviruses SARS-CoV, SARS-CoV-2, MERS-CoV, the gammacoronavirus IBV and the deltacoronavirus PDCoV ( Figure 2A). Next, we confirmed the effects of N protein expression on IFN-b promoter activity, this time induced by RIG-IN, a truncated, constitutively active form of RIG-I (49). The N proteins from PEDV, SARS-CoV-2, MERS-CoV and IBV were able to inhibit IFN-b promoter activity in these tests, suggesting that they target downstream of RIG-I or interfere with the interaction between RIG-I and MAVS ( Figure 2B).
Moving downstream, we wanted to test the effect of overexpression of other key signaling components on the observed inhibition by CoV N proteins. For the mitochondrial adapter MAVS, the N protein of TGEV, SADS-CoV, SARS-CoV, MERS-CoV, or PDCoV lost its inhibitive effect ( Figure 2C). Notably, the N proteins of PEDV, SARS-CoV-2 and IBV were all capable of inhibiting the activation of IFN-b pathway by TBK1, which is recruited to form complexes with MAVS that phosphorylate and subsequently activate IRF3. The PEDV N protein did not block IKKϵ induction of IFN-b, although the N proteins of SARS-CoV-2 and IBV did (Figures 2D, E). However, all N proteins in this study lost their inhibitory effect when the pathway was activated by IRF3-5D ( Figure 2F), which is a constitutively active form of IRF3 that is able to enter the nucleus to spontaneously activate IFN. To confirm the expression of N proteins, western blotting was conducted and the expression levels of N proteins have been shown (Figure 2). These data indicate that N proteins of different CoVs use various strategies to inhibit IFN expression, and even N proteins with relatively high similarity might inhibit IFN expression by targeting different stages of the signaling cascade.

The N Protein of SADS-CoV Inhibits IRF3 Phosphorylation and Nuclear Translocation Induced by SeV, But Not by RIG-IN
We wanted to explore in greater detail the mechanism of IFN inhibition utilized by the recently emerging SADS-CoV. Since the previous experiment suggested that the SADS-CoV N protein inhibits IFN-b promoter after SeV infection but not after induction by RIG-IN (Figures 2A, B), we took a closer look at the effect of SADS-CoV N protein on IRF3 activity. Activation of IRF3 via phosphorylation and nuclear translocation is essential for its induction of IFN-b production (50). IRF3 nuclear translocation was observed in cellular fractionation experiments, detecting IRF3 phosphorylation by western blot with S396 phosphorylated IRF3 MAb. As shown in

The N Protein of SADS-CoV Interacts
With RIG-I Independently of Its RNA Binding Activity The above results led us to speculate that the SADS-CoV N protein inhibits IFN signaling by interfering with the function of RIG-I. As a PRR for RNA PAMPs in the cytoplasm, RIG-I binds directly to RNA and subsequently activates downstream signaling (51). To investigate the mechanism by which SADS-CoV N inhibits RIG-I, we first asked whether there was a direct protein-protein interaction between them. We cotransfected 293T cells with flag-tagged fulllength RIG-I and myc-tagged SADS-CoV-N, followed by co-IP at 36 hpt by using anti-flag-tag antibodies. As shown in Figure 4A, SADS-CoV N protein was detected in the IP product, suggesting direct interaction between SADS-CoV N protein and RIG-I. Because N protein has RNA binding activity similar to RIG-I, we investigated whether the interaction is dependent on RNA binding of these two proteins. For this aim, the cell lysates were treated with RNaseA before co-IP was performed. As shown in Figure 4B, the SADS-CoV N protein was detected in the IP product even after RNaseA treatment, demonstrating the interaction with RIG-I was independent of shared RNA binding.

The SADS-CoV N Protein Induces Ubiquitination of RIG-I Leading to Its Degradation
The ubiquitination of RIG-I has been proven to be essential for activation of the RIG-I signaling pathway (52)(53)(54). Previous studies have identified several ubiquitination states of RIG-I that regulate its activity in different ways (52,55,56). To investigate the mechanism by which SADS-CoV impacts RIG-I, we determined whether the SADS-CoV N protein regulates ubiquitination of RIG-I. 293T cells were cotransfected with ubiquitin, RIG-I, and SADS-CoV N or empty vector, with or without poly(I:C) (which is a specific agonist for RLR). Cells were harvested at 36 hpt and the level of ubiquitination of RIG-I was analyzed by using co-IP and western blot. Poly(I:C) significantly induced ubiquitination of RIG-I ( Figure 5A, Lane 3), and, interestingly, SADS-CoV N enhanced the ubiquitination of RIG-I both in the presence and absence of poly(I:C) ( Figure 5A, Lanes 2 and 4). To determine which type of ubiquitination was upregulated by the SADS-CoV N protein, K48-and K63-ubiquitin were used. Whereas wild-type ubiquitination of RIG-I was significantly enhanced by the SADS-CoV N protein ( Figure 5A), K63-ubiquitination was decreased while no significant change in K48-ubiquitination was found ( Figure 5B). Most interestingly, we found significant decreases in the amount of RIG-I either in the presence of K48-ubiquitin or K63-ubiquitin in whole cell lysates, but not the wild-type ubiquitin, and this decreased amount was further reduced by SADS-CoV N coexpression. Since proteasome-dependent degradation is a common outcome of ubiquitination modification (57), we asked whether the reduced RIG-I levels in the whole cell lysate was caused by proteasome-mediated protein degradation. MG132 treatment was performed at 4 hpt to inhibit proteasome activity. The RIG-I level in whole cell lysates completely recovered after MG132 treatment, both in the presence of K48-ubiquitin and K63-ubiquitin ( Figure 5C). In addition, both K48-and K63-ubiquitin modification of RIG-I were enhanced by the SADS-CoV N protein after MG132 treatment, indicating that SADS-CoV N induces ubiquitination of RIG-I leading to its proteasome-dependent degradation.
To further explore which type of RIG-I ubiquitination was being induced by SADS-CoV N, K6-, K11-, K27-, K29-and K33ubiquitin, covering all the lysines present in the amino acid sequence of ubiquitin, were used to perform co-IP assay. Interestingly, K27-ubiquination was significantly increased by SADS-CoV N when cells were treated with MG132, whereas only a slight change in K27-ubiquitination was induced without MG132 treatment ( Figures 5D, E, Lanes 7 and 8). Notably, in K27-ubiquitin co-transfected cells, SADS-CoV N increased the level of RIG-I in whole cell lysate in the presence of MG132, indicating a proteasome-dependent degradation of RIG-I mediated by K27-ubiquitination ( Figure 5E).

DISCUSSION
Viruses have been reported to interfere with innate immunity and delay IFN responses, an ability that creates a suitable replication environment typical for infections by several CoVs (29,41,43,(58)(59)(60)(61)(62)(63). SADS-CoV is a newly emerged porcine CoV with potential for zoonotic transmission (7), and it has been proven to antagonize IFN-b production (58). However, the mechanism used by the SADS-CoV N protein in interfering with host innate immunity remains to be clarified.
The N protein is more highly conserved than other CoV structural proteins such as the spike protein, which may be due to its vital functions and the lower selective pressure placed on it. Thus, we wanted to see whether N protein amino acid similarity was predictive of a functional role in suppression of IFN response (Supplementary Figure S1). For the various N proteins tested, ability to suppress IFN response was not clearly related to the amino acid sequence similarity (Figures 1 and 2). SARS-CoV-2 N inhibited IFN promoter activity induced by RIG-IN, MAVS, TBK1 and IKKϵ, whereas the SARS-CoV N protein did not ( Figures 2B-D), despite an overall amino acid similarity of 91.2%, the highest among N proteins tested in this study (data not shown). This shows the importance of tertiary structure in determining protein function, which is not always evident from the simple amino acid sequence. To date, there is no available complete tertiary structure of the CoV N protein, whereas structures of the NTD and CTD have been reported for various CoV N proteins (48,(64)(65)(66)(67)(68). As described by earlier reports, the core NTD and CTD structures of CoV N proteins are highly conserved (48,(64)(65)(66)(67)(68); however, little evidence has been found so far to link the CoV N protein antagonism of host innate immunity to the tertiary structure of N protein, which is urgently needed to demonstrate the structural basis of the innate immune antagonistic activities of N proteins.
To date, N proteins from several CoVs have been proven to inhibit host IFN response ( Figure 6). The PEDV N protein has been reported to antagonize host IFN responses by interacting with TBK1 directly (16). The SARS-CoV N interacts with TRIM25 and protein activator of protein kinase R (PACT) directly to interfere with activation of RIG-I (17,18). The MERS-CoV N suppresses type I and type III IFN induction by interaction with TRIM25 (18,69). The MHV N protein interacts with PACT to inhibit activation of RIG-I (17). The PDCoV N protein interacts with RIG-I directly and interferes with binding of dsRNA and PACT to RIG-I (21,22). The N protein of SARS-CoV-2, the etiological agent of COVID-19, also has been shown to inhibit IFN induced by SeV, poly(I:C), RIG-I, MAVS, TBK1, and IKKϵ (70). Our current results are consistent with these previous studies (Figures 2 and 6), but go further in suggesting that SADS-CoV N targets the very early steps of the IFN response, and may directly interfere with activation of RIG-I (Figures 2A, B and 3). During preparation of our manuscript, Zhou et al. reported that SADS-CoV N blocked interaction between TBK1 and TRAF3 (71). CoV N proteins may target multiple host protein simultaneously such as SARS-CoV N protein interacting with TRIM25 and PACT (17,18). Since SADS-CoV infection interfere with host IFN signaling upstream of MAVS (58), and since SADS-CoV N appeared not to suppress IFN-b induced by TBK1, IKKe and IRF3-5D, which is indicated by the comparative analysis in Figures 2D-F, we speculate that interference via RIG-I is likely a dominant mode for SADS-CoV N protein to inhibit RLR signaling ( Figure 6). RLRs are cytoplasmic PRRs that recognize RNA PAMPs, and they have been proven to play critical roles in IFN induction during CoV infection (24,72,73). There are several CoV N proteins that have been shown to suppress host IFN response by targeting RIG-I or related proteins, such as those of SARS-CoV, MERS-CoV, MHV and PDCoV (18,(20)(21)(22)69). These studies led us to look closely at RIG-I as a likely target for CoV to circumvent the host IFN response. Consistently, our data demonstrated that the SADS-CoV N protein targets RIG-I to inhibit IFN-b promoter activity ( Figure 2). Our co-IP results indicate that SADS-CoV N protein interacted with RIG-I independently of RNA binding activity ( Figure 4), making it the only alphacoronavirus N protein that interacts with RIG-I directly. Recently, PDCoV, a deltacoronavirus that infects swine and has the potential for interspecies transmission, was shown to bind RIG-I with its N protein (21,22).
Full activation of RIG-I depends on multiple post-translational modifications beyond the simple presence of PAMPs. It is reported that ubiquitination, removal of inhibitory phosphorylation as well as some cofactors, such as PACT, are essential for full activation of RIG-I (52,74,75). Among these, ubiquitination has been reported to be critical for regulation of RIG-I, allowing versatile outcomes in IFN signaling depending on the type of ubiquitin modification (52,55,56,76,77). One common outcome of ubiquitination is proteasomedependent degradation of the target protein. To date, many viruses have been demonstrated to modulate ubiquitination of RIG-I in order to interfere with host immune signaling (78,79). For instance, the E6 oncoprotein of human papillomavirus (HPV) interacts with TRIM25, which is a ubiquitin E3 ligase that triggers K63-linked ubiquitination of RIG-I, thereby inhibiting the activation of downstream signaling (80). Paramyxovirus V proteins interact with RIG-I as well as TRIM25 to antagonize RIG-I-mediated IFN induction (81). Our data demonstrated that the SADS-CoV N protein enhances K27-, K48-and K63-linked ubiquitination of RIG-I ( Figure 5). Previous studies reported that K48-linked ubiquitination of RIG-I leads to degradation while K63-linked ubiquitination leads to activation of RIG-I (53,77). Consistently, K48-linked ubiquitination of RIG-I induced by SADS-CoV N protein led to proteasome-dependent degradation of RIG-I in our study ( Figures 5B, C). Unexpectedly, SADS-CoV N-mediated K63ubiquitination of RIG-I also led to degradation of RIG-I ( Figures 5B,  C). There are multiple E3 ligases that mediate K63-linked ubiquitination of RIG-I identified so far, such as TRIM25, Riplet, MEX3C and TRIM4 (53). These E3 ligases deliver a K63-linked polyubiquitin moiety to various domains of RIG-I, and might lead to divergent physiological outcome. The role of K27-linked ubiquitination of RIG-I is still elusive. Some components of the RLR signaling pathway such as MAVS (82) and IRF3 (83) have been reported to be modified by K27-linked polyubiquitination leading to autophagic degradation, suggesting that this type of ubiquitination plays an important role in host antiviral innate immunity. The K27linked ubiquitination of RIG-I mediated by SADS-CoV N protein did not decrease the level of RIG-I in whole cell lysates ( Figure 5D,  lanes 7 and 8), indicating a mode of action other than degradation by this type of modification on RIG-I activation. More importantly, while the wild type ubiquitination induced by SADS-CoV N protein did not lead to significant reduction of amount of RIG-I in whole cell lysates ( Figure 5), we speculate that K48-and K63-linked ubiquitination might not be the major form of ubiquitination induced by SADS-CoV N protein.
In conclusion, the present study provides evidence, by a comprehensive comparative analysis with representative CoVs from different genera, which the N protein of SADS-CoV interacts with RIG-I independent of its RNA binding activity. The N protein also induces ubiquitination and subsequent proteasome-dependent degradation of RIG-I. These effects may contribute to the ability of SADS-CoV N to inhibit IFN. Our data provide insight into interaction between SADS-CoV and host, and offer new clues for therapeutic treatment and vaccine development against SADS-CoV.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

AUTHOR CONTRIBUTIONS
Y-WH and BW conceived and designed the study and critically revised the manuscript. YL and Q-ZL performed the experiments and conducted data analysis. WL, Y-LY, and RC helped in experimental implementation. YL and Y-WH wrote the manuscript. All authors contributed to the article and approved the submitted version.

ACKNOWLEDGMENTS
We thank the professional editing service NB Revisions for technical preparation of the text prior to submission.