SARS-CoV-2 Nsp5 Demonstrates Two Distinct Mechanisms Targeting RIG-I and MAVS To Evade the Innate Immune Response

ABSTRACT Newly emerged severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) caused a global pandemic with astonishing mortality and morbidity. The high replication and transmission of SARS-CoV-2 are remarkably distinct from those of previous closely related coronaviruses, and the underlying molecular mechanisms remain unclear. The innate immune defense is a physical barrier that restricts viral replication. We report here that the SARS-CoV-2 Nsp5 main protease targets RIG-I and mitochondrial antiviral signaling (MAVS) protein via two distinct mechanisms for inhibition. Specifically, Nsp5 cleaves off the 10 most-N-terminal amino acids from RIG-I and deprives it of the ability to activate MAVS, whereas Nsp5 promotes the ubiquitination and proteosome-mediated degradation of MAVS. As such, Nsp5 potently inhibits interferon (IFN) induction by double-stranded RNA (dsRNA) in an enzyme-dependent manner. A synthetic small-molecule inhibitor blunts the Nsp5-mediated destruction of cellular RIG-I and MAVS and processing of SARS-CoV-2 nonstructural proteins, thus restoring the innate immune response and impeding SARS-CoV-2 replication. This work offers new insight into the immune evasion strategy of SARS-CoV-2 and provides a potential antiviral agent to treat CoV disease 2019 (COVID-19) patients.

To identify viral proteins that inhibit IFN induction, we screened a SARS-CoV-2 expression library with an IFN-b reporter assay induced by MAVS overexpression. MAVS is the adaptor molecule that relays innate immune activation downstream of RIG-I and MDA5. This assay identified several SARS-CoV-2 proteins, including Nsp1, Nsp5, and ORF6, that inhibit IFN induction to various extents (Fig. 1B). The inhibition of IFN induction by Nsp1 was due largely to the high basal induction of IFN by Nsp1 (data not shown), whereas ORF6 was reported to impede nuclear translocation of transcription factors such as IRF3 (36). Thus, we focused on Nsp5, the main protease of SARS-CoV-2, which is important for its replication. Importantly, the enzyme-deficient Nsp5-C145A mutant failed to inhibit IFN induction ( Fig. 1B and C). These results show that Nsp5 inhibits IFN induction in an enzyme-dependent manner.
To characterize the Nsp5-mediated inhibition of the innate immune response, we profiled the expression of several inflammatory genes in response to Sendai virus infection. Real-time PCR analysis indicated that Nsp5, but not the enzyme-deficient Nsp5-C145A mutant, inhibited the expression of antiviral genes, including IFNb, ISG56, ISG15, IL-6, and IL-8, in HCT116 cells (Fig. 1D). The inhibition of gene expression also correlated with reduced cytokine production in the medium of Sendai virus-infected cells (Fig. 1E). Similar inhibition of inflammatory gene expression by Nsp5 was observed in Sendai virus-infected A549 cells ( Fig. S1B and S1C).
The inhibition of the expression of both IRF3-and NF-k B-dependent genes by Nsp5 suggests that it targets upstream components of the RNA-sensing pathways. To determine the point of inhibition, we expressed key components of the RIG-I-IFN pathway, including constitutively active MDA5-N (2CARD), RIG-I-N (2CARD), IRF3-5D, MAVS, and TBK-1, and examined the IFN induction in the presence of Nsp5. As shown by an IFN reporter assay, we found that IFN induction was inhibited by Nsp5 when MDA5-N, RIG-I-N, and MAVS were overexpressed (Fig. S1D). However, IFN induced by the overexpression of TBK-1 and IRF3-5D was not affected by Nsp5. These results suggest that Nsp5 targets upstream molecules of the pathway, specifically MDA5, RIG-I, and/or MAVS. Indeed, transiently expressed RIG-I and MAVS proteins were reduced by Nsp5 as analyzed by immunoblotting ( Fig. 1F and G), while other components of the innate immune pathway, including MDA5, Ik B kinase b (IKKb), IKK« , IKKg (also known as NEMO), TBK-1, and TRAF6, were not affected by Nsp5 (Fig. S1E). Thus, the inhibition of MDA5-mediated IFN induction by Nsp5 is likely due to its effect on MAVS. We also investigated Nsp5 interaction with RIG-I or MAVS by coimmunoprecipitation (Co-IP). As shown in Fig. 1H and I, both RIG-I and MAVS precipitated with Nsp5, supporting their interactions. These results collectively show that SARS-CoV-2 Nsp5 targets RIG-I and MAVS to inhibit the antiviral immune response (Fig. 1J).
Nsp5 cleaves RIG-I at the Q10 residue. To probe the biochemical basis of RIG-I inhibition by Nsp5, we analyzed RIG-I by immunoblotting. We first examined RIG-I downregulation by SARS-CoV-2 using the Caco-2 cell line, which stably expresses RIG-I with a Flag epitope at the N terminus. As shown by immunoblotting analysis, we found that SARS-CoV-2 infection and Nsp5 expression decreased Flag-RIG-I protein in Caco-2 cells and 293T cells ( Fig. 2A and B). Surprisingly, no difference in RIG-I protein was observed in immunoblots analyzed with anti-RIG-I antibody (Fig. 2B). These results suggest that Nsp5 removes a very small N-terminal fragment, such that the apparent molecular weight of the main RIG-I fragment did not significantly change. To test this idea, we used the Flag-RIG-I-N and Flag-RIG-I-DN (deletion of 2CARD) constructs. Indeed, Nsp5  reduced the protein size of Flag-RIG-I-N but not that of Flag-RIG-I-DN ( Fig. 2C and D). To determine whether Nsp5 can function as a bona fide protease for RIG-I, we purified glutathione S-transferase (GST)-RIG-I-N, Nsp5, and the enzyme-deficient Nsp5-C145A mutant from bacteria and performed an in vitro cleavage assay. As shown in Fig. 2E, Nsp5, but not Nsp5-C145A, produced a fragment of approximately the size of GST (27 kDa). This result supports the conclusion that Nsp5 cleaves the very N-terminal end of RIG-I-N, without altering the apparent size of GST.
To identify the site of cleavage within the N terminus of RIG-I, we aligned the 20 Nterminal amino acids of RIG-I proteins of humans and several mammals, including monkey, dog, bat, mouse, and cat. The alignment showed that a putative LQA Nsp5 cleavage sequence is shared only by human and mouse RIG-I proteins (Fig. 2F). Importantly, L and A flanking the putative Q cleavage residue favor the cleavage by Nsp5 as previous studies showed (37)(38)(39). Interestingly, the preferred cleavage site was not shared by RIG-I proteins of bat, cat, dog, and monkey, suggesting that cells of these mammals are likely more resistant to SARS-CoV-2 replication. To determine whether Nsp5 targets Q10 of human RIG-I, we utilized a RIG-I mutant, RIG-I-Q10E, which is the RIG-I deamidated by murine gamma herpesvirus 68 (MHV68) (40). Remarkably, RIG-I-Q10E was resistant to Nsp5-induced downregulation, while wildtype RIG-I (RIG-I-WT) was not (Fig. S2A). In contrast, when RIG-I and RIG-I-Q10E constructs with a V5 epitope at their carboxyl terminus were expressed in the presence of Nsp5, RIG-I and RIG-I-Q10E demonstrated no difference in their apparent molecular weights (Fig. S2B). We also performed an in vitro cleavage assay using GST-RIG-I-N, GST-RIG-I-N-Q10E, Nsp5, and the enzyme-deficient Nsp5-C145A proteins purified from bacteria. As shown in Fig. 2G, GST-RIG-I-N, but not GST-RIG-I-N-Q10E, produced a fragment of approximately the size of GST in the presence of Nsp5. These results imply that there is no additional cleavage site for Nsp5. Thus, Nsp5 targets the Q10 residue of human RIG-I for cleavage.
To probe the biological consequence of Nsp5-mediated cleavage of RIG-I, we reconstituted Rig-i 2/2 mouse embryonic fibroblasts (MEFs) with wild-type RIG-I and RIG-I-Q10E (Fig. 2H). When Nsp5 MEFs were infected with Sendai virus, we found that the expression of antiviral genes, including Ifnb, Isg15, and Isg56, was greatly reduced by Nsp5 in Rig-i 2/2 MEFs reconstituted with wild-type RIG-I (Fig. 2I). In contrast, Nsp5 had no apparent effect on the antiviral gene expression in Rig-i 2/2 MEFs reconstituted with RIG-I-Q10E. These results support the conclusion that RIG-I-Q10E resists Nsp5-mediated cleavage and immune evasion.
Loss-of-function and dominant negative effect of the cleaved RIG-I fragment. Previously, we reported that MHV68 targets the Q10 residue of RIG-I for deamidation to evade cytokine production (40). The fact that SARS-CoV-2 Nsp5 also targets the same residue for cleavage suggests that the 10 N-terminal amino acids are critical for RIG-I function. Indeed, a previous study reported that the loss of the 10 N-terminal amino acids impaired the ability of RIG-I-N to associate with free ubiquitin chains and to activate IRF3 in vitro (41). Thus, we characterized the function of RIG-I-  in activating the innate immune defense. First, we reconstituted Rig-i 2/2 MEFs with wildtype RIG-I and RIG-I-(11-925) via lentiviral infection (Fig. 3A). When antiviral gene expression was examined, we found that Sendai virus infection potently induced the expression of Ifnb, Isg56, and Il-6 in Rig-i 2/2 MEFs reconstituted with wild-type RIG-I but not in those reconstituted with RIG-I-(11-925) (Fig. 3B). This result was also consistent with the phosphorylation and activation of TBK-1 in Rig-i 2/2 MEFs reconstituted with wild-type RIG-I, which was not seen in those reconstituted with RIG-I-(11-925) (Fig. 3C).
To determine whether RIG-I-(11-925) has a dominant negative effect, we performed reporter assays for IFN induction and NF-k B activation upon Sendai virus infection. While  overexpression of wild-type RIG-I elevated IFN induction and NF-k B activation, overexpression of RIG-I-(11-925) potently reduced IFN induction and NF-k B activation in a dose-dependent manner (Fig. 3E). Similarly, real-time PCR analysis further showed that RIG-I-(11-925) inhibited the expression of inflammatory genes, including IFNb and IL-6 ( Fig. S3A). Consistently with reduced inflammatory gene expression, RIG-I-  also inhibited the phosphorylation of IRF3 and TBK-1, signaling events downstream of RIG-I and MAVS, in response to Sendai virus infection (Fig. 3F). Thus, RIG-I-  inhibits innate immune activation induced by dsRNA.
To dissect the mechanism of inhibition, we determined whether RIG-I-(11-925) can bind dsRNA as we previously described (40). Indeed, an RNA gel shift assay showed that RIG-I- , like wild-type RIG-I, retarded dsRNA migration in a dose-dependent manner (Fig. 3G). This result indicates that RIG-I-(11-925) is capable of binding dsRNA, consistent with the notion that the carboxyl-terminal domain of RIG-I is sufficient for dsRNA binding. Finally, we assessed the effect of RIG-I-(11-925) on the heterodimerization between RIG-I and MAVS, induced by Sendai virus infection. Coimmunoprecipitation assay indicated that wild-type RIG-I, but not RIG-I- , pulled down MAVS in Sendai virus-infected cells (Fig. 3H). Moreover, the overexpression of RIG-I-(11-925) inhibited RIG-I-MAVS heterodimerization in a dose-dependent manner in response to Sendai virus infection (Fig. S3B). Taken together, these results collectively demonstrate that RIG-I-(11-925) exerts a dominant negative effect on dsRNA-induced innate immune activation, likely via competing with RIG-I for binding to dsRNA.
Nsp5 promotes MAVS degradation via the ubiquitin-proteosome pathway. MAVS is a mitochondrion-resident protein that relays innate immune activation downstream of RIG-I and MDA5 (20)(21)(22)(23). To validate the supposition that Nsp5 reduces MAVS protein during SARS-CoV-2 infection, we examined MAVS expression in SARS-CoV-2infected Caco-2 cells by immunoblotting and quantitative real-time PCR (qRT-PCR). Interestingly, SARS-CoV-2 infection increased MAVS mRNA, but greatly diminished MAVS protein, at 72 and 96 h postinfection (Fig. 4A). To determine whether Nsp5 targets MAVS for proteolytic cleavage, we inspected the primary amino acid sequence of MAVS for possible consensus Nsp5 cleavage sites. Although several putative sites with low confidence were identified, mutations that change these putative cleavage glutamines to glutamates did not restore MAVS protein in the presence of Nsp5 (data not shown). Additionally, we did not detect any cleaved MAVS fragments by immunoblotting (data not shown). These results suggest that Nsp5 may reduce MAVS protein independently of its protease activity. To test this hypothesis, we measured the half-life of MAVS when translation was inhibited with cycloheximide. As shown in Fig. 4B, Nsp5 reduced the half-life of MAVS to approximately 4 h, whereas its half-life in the presence of Nsp5-C145A or green fluorescent protein (GFP) was calculated to be .8 h. Thus, Nsp5 increases MAVS degradation to reduce MAVS protein expression. In mammalian cells, there are two main protein-degradative pathways, i.e., the proteosome and lysosome. To determine which of these pathways is required for Nsp5-mediated MAVS degradation, we treated cells with MG132, a pan-proteosome inhibitor, and NH 4 Cl, a lysosome inhibitor. Remarkably, MG132 greatly restored MAVS protein in the presence of Nsp5, whereas NH 4 Cl had no apparent effect on MAVS protein (Fig. 4C). These results support the conclusion that Nsp5 induces MAVS degradation via proteosomes.
Proteosomes remove proteins that are tagged with ubiquitin via proteolytic cleavage. Thus, we first determined whether MAVS undergoes ubiquitination. When endogenous MAVS was precipitated and analyzed with antiubiquitin antibody, we found that Nsp5 potently increased the ubiquitination of MAVS but that Nsp5-C145A failed to do so (Fig. 4D). Ubiquitin can be added to target proteins, such as MAVS, via distinct lysine residues of ubiquitin. Two main lysine residues, K48 and K63, are best characterized for  examined MAVS ubiquitination. This assay showed that the ubiquitin-K48R mutant, but not the ubiquitin-K63R mutant, reduced MAVS ubiquitination to background levels ( Fig. 4E), supporting the conclusion that Nsp5 promotes the K48-linked ubiquitination of MAVS. We next determined whether Nsp5 can function as a bona fide E3 ligase, given that Nsp5 is characterized as a cysteine protease. When Nsp5 was purified from bacteria and added to MAVS along with E1 and E2(UbcH5b), a massive ubiquitin signal of higher molecular weight (.72 kDa) was detected in the presence of Nsp5 by immunoblotting (Fig. 4F). In contrast, residual levels of ubiquitin signal were detected in the presence of Nsp5-C145A. Collectively, these results demonstrate that Nsp5 can function as a bona fide E3 ligase to promote MAVS degradation via the ubiquitin-proteosome system.
Nsp5 targets the K136 residue of MAVS to promote protein degradation. Having demonstrated that Nsp5 promotes MAVS ubiquitination, we searched for the ubiquitination site(s) within MAVS. One interesting feature of MAVS is that there are multiple forms of MAVS expressed from the same transcript via internal translation initiation, giving rising to three isoforms of ;70, 50, and 30 kDa, thus referred to as MAVS70, MAVS50, and MAVS30, respectively (42)(43)(44). We explored these isoforms as natural truncations from the N terminus and examined the effect of Nsp5 on their expression. Remarkably, while Nsp5 greatly reduced the expression of the MAVS70 protein, it had no apparent effect on MAVS50, which lacks the first 140 amino acids of MAVS70, compared to the effects of the Nsp5-C145A mutant (Fig. 5A). This result suggests that the N-terminal sequence (amino acids 1 to 140) is important for Nsp5-mediated downregulation of MAVS70. Inspection of the 140 N-terminal amino acids identified three lysine residues, K7, K10, and K136, that can potentially function as ubiquitination receptors. We generated two MAVS mutants containing either a K7, K10R, or K136R mutation. In transfected 293T cells, the MAVS-K7,10R mutant and, to a lesser extent, MAVS-K136R were expressed at higher levels than wild-type MAVS (Fig. 5B). Nsp5 expression detectably reduced the protein levels of MAVS-K7,10R and that of wild-type MAVS. However, Nsp5 had a marginal effect on the protein level of MAVS-K136R, which was comparable to that of Nsp5-C145A. These results suggest that Nsp5 targets the K136 residue of MAVS for ubiquitination and degradation.
To further validate that Nsp5 targets the K136 residue of MAVS for degradation, we determined the half-life of wild-type MAVS and MAVS-K136R in the presence of GFP, Nsp5, and Nsp5-C145A. As with endogenous MAVS, exogenously expressed MAVS had a half-life of more than 8 h. However, Nsp5 reduced MAVS's half-life to approximately 4 h (Fig. 5C). In stark contrast, MAVS-K136R demonstrated a half-life of .8 h in the presence of GFP, Nsp5, or Nsp5-C145A. The MAVS mutant containing all three K!R mutations (triple mutant [TM]) behaved similarly to MAVS-K136R in its half-life (Fig. S5A). When MAVS ubiquitination was examined in stable 293T cells, we found that MAVS-K136R had a background level of ubiquitin signal, while ubiquitination of wild-type MAVS was readily detected under similar conditions (Fig. 5D). These results collectively show that the K136 residue is targeted by Nsp5 for ubiquitination to promote MAVS degradation.
To assess the biological consequence of Nsp5-mediated degradation of MAVS, we reconstituted Mavs 2/2 mouse embryonic fibroblasts (MEFs) with wild-type MAVS and MAVS-K136R (Fig. 5E). We found that the expression of antiviral genes, including Ifnb, Isg15, and Isg56, was greatly reduced by Nsp5 in Mavs 2/2 MEFs reconstituted with wildtype MAVS (Fig. 5F). In contrast, Nsp5 had no apparent effect on the antiviral gene expression in Mavs 2/2 MEFs reconstituted with MAVS-K136R. Similar results were observed in an IFN reporter assay; IFN induction by wild-type MAVS, but not that by MAVS-K136R, was inhibited by Nsp5 (Fig. S4B). These results support the conclusion that MAVS-K136R resists Nsp5-mediated degradation and immune evasion.
RIG-I-Q10E and MAVS-K136R resist destruction and restore innate immune activation during SARS-CoV-2 infection. Considering that SARS-CoV-2 deploys Nsp5 to simultaneously inactivate RIG-I and MAVS, we reasoned that endogenous RIG-I and MAVS are functionally nullified by SARS-CoV-2 during infection. Thus, we ectopically expressed wild-type RIG-I or RIG-I-Q10E in human Caco-2 cells (Fig. 6A) and examined the expression of antiviral genes, including IFNb, ISG15, and ISG56. Upon SARS-CoV-2 infection, we observed clear yet weak induction of these genes in control (EV) Caco-2 cells (Fig. 6B). Ectopic expression of wild-type RIG-I further elevated the expression of these genes by 1-fold for IFNb and ISG56 and by 10-fold for ISG15. Compared to wildtype RIG-I, RIG-I-Q10E further elevated the expression of these antiviral genes by 1fold. These results were reproduced from Rig-i 2/2 MEFs expressing wild-type RIG-I and  RIG-I-Q10E (Fig. S5A and S5B). Very similar results were observed for the expression of wild-type MAVS and MAVS-K136R in both Caco-2 cells (Fig. 6C and D) and Mavs 2/2 MEFs (Fig. S5C and S5D) in response to SARS-CoV-2 infection. Next, we generated Caco-2 cells that express either wild-type RIG-I plus wild-type MAVS or RIG-I-Q10E plus MAVS-K136R and infected them with SARS-CoV-2 (Fig. 6E). Upon SARS-CoV-2 infection, we observed clear reduction in ectopically expressed wild-type RIG-I and MAVS proteins (Fig. 6G). In contrast, RIG-I-Q10E and MAVS-K136R were not significantly reduced by SARS-CoV-2 infection. Wild-type Caco-2 cells were used for SARS-CoV-2 infection, considering that endogenous RIG-I and MAVS can be destructed by Nsp5. When antiviral gene expression was analyzed by real-time PCR, we found that the simultaneous expression of RIG-I-Q10E and MAVS-K136R induced significantly higher levels of transcripts of these antiviral genes than those induced by wild-type RIG-I and MAVS in Caco-2 cells (Fig. 6F) and in Mavs 2/2 MEFs (Fig. S5E and S5F).
To assess the biological significance of the inhibition of RIG-I and MAVS by SARS-CoV-2, we assessed SARS-CoV-2 replication in these cell lines by real-time PCR for viral RNA abundance and plaque assay for viral titer. Real-time PCR analysis showed that the expression of wild-type RIG-I or MAVS reduced viral S RNA by ;75%, while the expression of RIG-I-Q10E or MAVS-K136R reduced it by ;90% (Fig. 6G). However, no additive effect was observed for wild-type RIG-I plus MAVS or for RIG-I-Q10E plus MAVS-K136R. When RNA-dependent RNA polymerase (RdRp) RNA was analyzed, we observed similar results, with the effects of RIG-I-Q10E and MAVS-K136R being much more pronounced than those of their wild-type counterparts. This result was also obtained from MEFs with wild-type RIG-I and MAVS and with RIG-I-Q10E and MAVS-K136R (Fig. S5G to S5I). The reduced viral RNA also translated into lower viral titers by RIG-I-Q10E and MAVS-K136 and to a lesser extent for wild-type RIG-I and/or MAVS (Fig. 6I). Taken together, these results demonstrate that destruction-resistant RIG-I and MAVS more robustly inhibit SARS-CoV-2 replication than their wild-type counterparts.
A small-molecule inhibitor of Nsp5 restores innate immune activation to impede SARS-CoV-2 replication. In addition to antagonizing the innate immune defense, Nsp5 is a cysteine protease that processes most SARS-CoV-2 nonstructural proteins to yield their functional entities, offering a viable target for antiviral inhibition. Thus, we synthesized two small-molecule inhibitors that contain an electrophilic warhead, vinyl sulfone, to covalently target the catalytic residue of cysteine proteases (Fig. 7A). The vinyl sulfone 2CN115, containing a clickable alkyne tag, was shown to covalently engage Nsp5 in an in-gel fluorescence assay (Fig. 7B). A reporter assay showed that 2CN113 restored IFN induction in the presence of Nsp5 when cells were transfected with RIG-I-N (Fig. 6SA); thus, we further characterized 2CN113 as an inhibitor of Nsp5. When 293T cells were treated with 2CN113, RIG-I protein was restored at 2 mM and, to a much lesser extent, at 0.08 mM and 0.4 mM (Fig. S6B). Similarly, 2CN113 also restored MAVS protein in the presence of Nsp5 (Fig. S6C). Furthermore, 2CN113 reduced Nsp5-mediated RIG-I cleavage in a dose-dependent manner when analyzed by an in vitro cleavage assay (Fig. 7C). Considering that Nsp5 is also required for the processing of SARS-CoV-2 nonstructural proteins, we analyzed Nsp8 in SARS-CoV-2-infected cells with 2CN113 treatment. As shown in Fig. 7D, treatment with 2CN113 abolished the production of Nsp8 and an intermediate containing Nsp8 that matches the size of the Nsp7-Nsp8 fusion. When innate immune gene expression was examined in Calu-3 cells infected with SARS-CoV-2, we found that 2CN113 increased the expression of IFNb, ISG15, ISG56, CCL5, and CCL10 in a dose-dependent manner (Fig. 7E). A similar effect of 2CN113 was observed on the expression of ISGs, such as IFNb, ISG15, and ISG56, in SARS-CoV-2-infected Caco-2 cells (Fig. S6D). Conversely, 2CN113 reduced viral RNA transcripts, such as RdRp and Nsp12, in a dose-dependent manner in Calu-3, Caco-2, and Vero cells infected with SARS-CoV-2 ( Fig. 7F; Fig. S6E and S7I). Consistently with this result, 2CN113 also potently reduced infectious virions of SARS-CoV-2 in the medium (Fig. 7G and J and Fig. S6F). Specifically, infectious SARS-CoV-2 in the medium was ;100and 500-fold lower in Calu-3 cells treated with 2CN113 at 1 and 4 mM, respectively, than in the vehicle-treated groups.
To determine whether the inhibition of SARS-CoV-2 replication was due to the cytotoxicity of 2CN113, we assessed the 50% cytotoxic concentration (CC 50 ) or 50% inhibitory concentration (IC 50 ) in Caco-2 cells without and with SARS-CoV-2 infection, respectively. This analysis showed that the CC 50 s of 2CN113 on Calu-3 and Caco-2 cells were approximately ;200-fold greater than their IC 50 s of SARS-CoV-2 replication, indicating that the antiviral activity is unlikely to be due to its cytotoxicity ( Fig. 7H and Fig. S6G). Similarly, the CC 50 of 2CN113 on Vero cells was also ;200-fold higher than the IC 50 of SARS-CoV-2 replication (Fig. 7K). These results collectively show that 2CN113 inhibits SARS-CoV-2 replication via Nsp5.

SARS-CoV-2, the etiological agent for CoV disease 2019 (COVID-19)
, is a positive, single-stranded RNA virus that recently broke into the human population likely via zoonotic transmission (1,16). Given that SARS-CoV-2 replicates within the cytoplasm of infected cells, it is conceivable that dsRNA replication intermediates of SARS-CoV-2 can be recognized by cytosolic RNA sensors, such as RIG-I and MDA5 (45,46). Previous studies have implicated MDA5 in sensing SARS-CoV-2, leaving the function of RIG-I unknown in SARS-CoV-2 infection (47,48). To probe the role of RIG-I in SARS-CoV-2 infection, we discovered that dsRNA derived from SARS-CoV-2-infected NHBE cells potently activates RIG-I. However, SARS-CoV-2 infection induced a weak inflammatory response, as evidenced by IFN-b induction, in NHBE cells. These observations suggest that SARS-CoV-2 subverts dsRNA-induced innate immune activation. Indeed, a screen utilizing the SARS-CoV-2 expression library identified Nsp5 as a potent inhibitor of IFN induction. Biochemical assays show that Nsp5 targets both RIG-I and MAVS, but not MDA5, for destruction. This result may explain the previous implication of MDA5 in sensing coronavirus infection (47,48). Specifically, Nsp5 cleaves RIG-I immediately after the Q10 residue, leaving the nearly full-length RIG-I protein incompetent to interact with and activate downstream MAVS. Nsp5 demonstrates proteolytic activity toward RIG-I in cells and in vitro, adding RIG-I to the growing list of modulators that include NEMO, STAT2, and other coronavirus nonstructural proteins (49)(50)(51)(52). Furthermore, the larger cleaved RIG-I fragment binds dsRNA as competently as wild-type RIG-I, thus exerting a dominant negative effect on the dsRNA-induced immune response. Although N proteins of coronaviruses are proposed to sequester RNA (53), the Nsp5mediated cleavage of RIG-I offers a conceivably more active means of subverting RIG-I

SARS-CoV-2 Nsp5 Targets RIG-I and MAVS
® and likely impacts signaling events downstream of other RNA sensors, such as protein kinase R (PKR) (54). Intriguingly, our previous study showed that murine gamma herpesvirus 68 induces the deamidation of residue Q10 to evade the RIG-I-mediated innate immune response (40). In an independent study, Zhijian Chen's group also demonstrated that the 10 N-terminal amino acids are necessary for RIG-I-mediated IRF3 activation (41). These observations highlight an important element in RIG-I that is targeted by two viral pathogens via distinct mechanisms, thus implying convergent evolution.
In addition to targeting RIG-I, Nsp5 targets MAVS for degradation via the ubiquitinproteasome pathway. Mechanistically, Nsp5 functions as an E3 ligase that adds ubiquitin to MAVS through K48 linkage. Mutational analysis supports the conclusion that K136 of MAVS serves as the receptor for ubiquitin and that the MAVS-K136R mutant is resistant to Nsp5-mediated degradation. Although Nsp5 is defined as a cysteine protease, the ability of its protease domain to catalyze another type of enzymatic reaction is provocative but not unprecedented. For example, the protease domain of Nsp3, a papain-like protease, is known to perform deubiquitination (55,56), the reverse reaction of ubiquitination that is catalyzed by Nsp5. It would be interesting to determine how these two viral proteases impact each other in terms of protein ubiquitination status and associated biological processes. Our work describes a new function of Nsp5 in catalyzing protein ubiquitination and degradation, thereby terminating RIG-I-mediated immune signaling. The fact that Nsp5 targets both RIG-I and MAVS, two key signaling molecules in the innate immune detection of cytosolic dsRNA, suggests the importance of this pathway in restricting SARS-CoV-2 replication. Presumably, MAVS degradation induced by Nsp5 likely dampens the MDA5-mediated immune activation that was reported to be responsible for SARS-CoV-2 infection. Indeed, ectopic expression of Nsp5-resistant RIG-I-Q10E and MAVS-K136R partly restored antiviral gene expression, thus reducing SARS-CoV-2 replication. The restoration of antiviral gene expression was likely limited due to the inhibition of downstream steps of the RIG-I-MAVS pathway by other viral proteins. We noted that Nsp5 was previously reported to block IRF3 nuclear translocation and IFN induction (57). However, Nsp5 demonstrates no significant effect on IFN induction by overexpressing TBK-1 or constitutively active IRF3-5D. The apparent discrepancy is not clear and remains to be determined.
In addition to processing host factors such as RIG-I and MAVS, Nsp5 processes several nonstructural proteins of SARS-CoV-2, thus playing essential roles in key processes of viral infection, such as genome replication (58). Here, we report the development of a small-molecule inhibitor that demonstrates the ability to block the Nsp5-mediated destruction of RIG-I and MAVS. Consequently, this small-molecule inhibitor elevated antiviral gene expression and reduced SARS-CoV-2 replication. The antiviral activity of the Nsp5 inhibitor likely also stems from its blockage of the proteolytic cleavage of the majority of SARS-CoV-2 nonstructural proteins, including Nsp4 to Nsp16, in addition to its restoration of innate immune activation. The CC 50 s of the small-molecule inhibitor on Caco-2 and Vero-E6 cells were approximately 200-fold greater than its IC 50 of SARS-CoV-2 replication in those cells, suggesting that its antiviral activity is not due to nonspecific cytotoxicity. These findings provide a proof of concept for targeting Nsp5 to impede SARS-CoV-2 infection and treat COVID-19. Caco-2 cells were cultured in minimum essential media (MEM) supplemented with 10% FBS and antibiotics. All cell lines were maintained at 37°C in a humidified atmosphere of 5% CO 2 . Viruses. Sendai virus (SeV) was purchased from Charles River. SARS-CoV-2 was propagated in Vero E6 cells. Vero E6 cells were seeded at 1.5 Â 10 6 cells per T25 flask for 12 h. Cells were washed once with FBS-free DMEM before viral infection. After being washed, cells were infected with SARS-CoV-2 at a multiplicity of infection (MOI) of 0.001 in FBS-free DMEM and monitored daily. The supernatant was harvested when 20% of cells demonstrated a cytopathic effect (around 72 h postinfection). The supernatant was centrifuged at 3,000 rpm for 5 min and stored at 280°C, after which the viral titer in the supernatant was determined by a plaque assay using a Vero-E6 monolayer. SARS-CoV-2-related viral propagation, viral infection, and vital titration were performed in a biosafety level 3 (BSL-3) facility at USC.
qRT-PCR. Quantitative real-time PCR (qRT-PCR) was performed as previously described (59). Cells were harvested at various time points after viral infection or transfected with the indicated constructs. Total RNA was extracted using TRIzol reagent (Invitrogen). cDNA was synthesized from 1 mg of total RNA using reverse transcriptase (Invitrogen) according to the manufacturer's instructions. cDNA was analyzed by qRT-PCR using the qPCR-BIO SyGreen Blue mix Lo-ROX (Genesee Scientific). The gene expression level was calculated by the 2 -DDCt method (where Ct is the threshold cycle) using b-actin or GAPDH as an internal control. The qRT-PCR primers are listed in Table 1.
Dual-luciferase reporter assay. HEK293T cells were cotransfected with the indicated expression plasmids and a reporter plasmid cocktail containing 50 ng of luciferase reporter plasmids (ISRE-luc, IFN-b-luc, or NF-k B) and 5 ng of the TK-Renilla luciferase reporter (control vector) by calcium phosphate precipitation. At 24 h posttransfection, whole-cell lysates (WCLs) were prepared and used for a dual-luciferase assay according to the manufacturer's instructions (Promega). The activities of firefly luciferase and Renilla luciferase were determined by a microplate reader (FLUOstar Omega).
Protein expression and purification. HEK293T cells were transfected with plasmids containing Flag-tagged RIG-I and RIG-I-(11-925) for 36 h. WCLs were prepared with Triton X-100 buffer (20 mM Tris, pH 7.5, 150 mM NaCl, 1 mM EDTA, 20 mM b-glycerophosphate, 10% glycerol) supplemented with a protease inhibitor cocktail (Roche). WCLs were sonicated, rotated at 4°C for 20 min, and centrifuged at 12,000 rpm and 4°C for 15 min. The supernatant was filtered and precleared with Sepharose 4B agarose beads (Thermo) at 4°C for 1 h. The precleared samples were incubated with anti-Flag M2 agarose beads at 4°C for 4 h. Agarose beads were washed extensively with lysis buffer. Flag-tagged proteins were eluted with 0.2 mg/ml 3ÂFlag peptide.
Ubiquitination assays. Cells were lysed in 1% NP-40 buffer (100 ml) supplemented with 1% SDS and denatured at 95°C for 5 min. The cell lysates (100 ml) were diluted with 1 ml lysis buffer, and a portion of cell lysates (200 ml) was saved for immunoblot analysis to detect the expression of target proteins. The rest of the cell lysates (800 ml) were precipitated (Denature-IP) with either anti-Flag beads or anti-MAVS (0.2 to 0.5 mg), followed by protein A/G (20 ml). Precipitated proteins were washed three times and subjected to immunoblotting analysis.
For in vitro ubiquitination assay, immunoprecipitated Nsp5-Step, Nsp-C145A-Step, and Flag-MAVS were purified from lysates of E. coli and HEK293T cells, individually. Immunoprecipitates were incubated with 100 ng of E1 (Boston Biochem), 400 ng of E2 (UbcH5b; Boston Biochem), and 2 mg of ubiquitin (Boston Biochem) in reaction buffer (50 mM Tris [pH 7.4], 2 mM MgCl 2 , 4 mM ATP; Boston Biochem) and 1 mM dithiothreitol at 30°C for 2 h, and the reaction was terminated by addition of 4Â the sample buffer.
In vitro cleavage assay. Samples containing recombinant Nsp5 and Nsp5-C145A and its substrate, GST-RIG-I-2CARD, were incubated in Nsp5 cleavage assay buffer {20 mM Tris, pH 7.4, 15 mM NaCl, 0.5 mM TCEP [Tris(2-carboxyethyl)phosphine hydrochloride]} at 37°C for 2 h. Reactions were stopped by the addition of 2Â SDS sample buffer and subsequent boiling at 95°C for 10 min, and samples were used for SDS-PAGE and immunoblotting analysis.
ELISA. Commercial cytokine enzyme-linked immunosorbent assay (ELISA) kits used in this study were human IFN-b (PBL Assay Science) and human RANTES (R&D Systems). Cytokine levels in the media of cultured cells were determined according to the manufacturer's instructions. Absorbance was determined with FLUOstar Omega (BMG Labtech).
Plaque assay. For SARS-CoV-2 viral titration (plaque assay), Vero E6 cells seeded in 12-well plates (;100% cell density) were washed with FBS-free medium and infected with 10-fold-diluted medium containing virion particles. At 1 h postinfection, medium was removed, and overlay medium containing 1% low-melting point agarose and FBS-free 1Â DMEM was added. At 72 h postinfection, cells were fixed with 4% paraformaldehyde (PFA) overnight and stained with 0.2% crystal violet. Plaques were counted on a light box.
Syntheses of 2CN113 and 2CN115. 2CN113 and 2CN115 containing the vinyl sulfone-reactive group were synthesized based on a previously published method, with minor modifications, as illustrated in Fig. S7 (64, 65). Briefly, 9-fluorenylmethoxy carbonyl (Fmoc)-Gln (Trt)-OH was subjected to Weinreb ketone synthesis using N,O-dimethyl hydroxylamine hydrochloride to yield the Weinreb amide, called intermediate 1. The newly obtained Weinreb amide 1 was then reduced to an aldehyde-containing intermediate 2 with an excess of LiAlH 4 . Intermediate 3, containing an electrophilic vinyl sulfone group, was formed when reacting aldehyde 2 with diethyl methylsulfonyl methyl phosphonate in the presence of NaH. The final multistep reaction, including Fmoc deprotection and amide coupling with Z-Leu-OH or alkyne-tagged Z-Leu-OH followed by trityl removal, successfully yielded 2CN113 and 2CN115. All of the reagents and solvents were purchased commercially and used without further purification, unless otherwise stated. All anhydrous reactions were carried out under a nitrogen atmosphere. Reactions were monitored by thin-layer chromatography (TLC) on glass TLC plates with silica gel coated with fluorescent indicator F254. UV light and TLC stains, including ninhydrin and 2,4-dinitrophenylhydrazine, were used as visualizing agents. Individual intermediates and final compounds were purified by flash column chromatography using an automated Teledyne CombiFlash system. 1 H and 13 C nuclear magnetic resonance (NMR) spectra were obtained on a Varian Mercury 400, 400MR, VNMRS-500, or VNMRS-600 spectrometer at ambient temperature.
In-gel fluorescence imaging. HEK293T cells transiently expressing Nsp5-Strep were treated with 2CN115 at the indicated doses at 37°C for 24 h. Following probe treatment, cells were harvested and lysed in NP-40 buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1% NP-40, 5 mM EDTA) supplemented with protease inhibitor (5 mg/ml; Roche). WCLs were centrifuged, and the protein-containing lysate was normalized to 1 mg/ml. Strep-tagged Nsp5 was then enriched by incubation with StrepTactin agarose for 20 min at 4°C. The beads were washed and subjected to on-resin CuAAC conjugation to a fluorescent rhodamine dye under normal click chemistry conditions {50 mM 6-carboxytetramethylrhodamine (TAMRA)-azide, 1 mM Tris(2-carboxyethyl)phosphine (TCEP), 200 mM Tris[(1-benzyl-1 H-1,2,3-triazol-4-yl) methyl]amine (TBTA-HCl), and 1 mM CuSO 4 } for 1 h at room temperature in the dark. Upon completion of the click reaction, beads were washed and bound proteins were eluted by boiling the solutions at 95°C for 10 min. Click-conjugated, enriched proteins were fractionated by SDS-PAGE before visualization at a 532-nm excitation and 610-nm emission on a Typhoon 9400 variable-mode imager (GE Healthcare). After fluorescence scanning, gels were silver stained by standard procedures (Pierce) to track Nsp5 expression levels.
Antiviral and cytotoxicity assays for 2CN113. For the antiviral assay, a clinical isolate of SARS-CoV-2 was propagated in Vero E6 cells, and the viral titer was determined as previously described. All infection experiments were performed at BSL-3. Preseeded Calu-3 and Caco-2 cells (5 Â 10 5 cells per well) were pretreated with the indicated concentrations of Nsp5 inhibitors for 2 h, and SARS-CoV-2 was subsequently added (MOI of 0.1) to allow infection for 2 h. Then, the virus-drug mixture was removed, and cells were further cultured with fresh drug-containing medium. Drugs were added every 24 h after viral infection until the end of the experiments. At 24 h, 48 h, and 72 h postinfection, virus-containing medium was collected and the viral titer in the medium was determined by plaque assay.
For Vero cells, preseeded Vero cells (1 Â 10 5 cells per well) were pretreated with the indicated concentrations of Nsp5 inhibitors for 2 h, and SARS-CoV-2 was subsequently added (MOI of 0.001) to allow infection for 2 h. Then, the virus-drug mixture was removed, and cells were further cultured with fresh drug-containing medium. At 24 h postinfection, virus-containing medium was collected, and the viral titer in the medium was determined by plaque assay.
For cytotoxicity assays, Calu-3, Caco-2, and Vero E6 cells were suspended in growth medium in 96well plates. The next day, appropriate concentrations of Nsp5 inhibitors were added to the medium. After 24 h, the cell viability was measured by cell proliferation kit II (XTT) in accordance with the manufacturer's instructions.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.