Identification of protein-protein interactions of SARS-CoV-2 using mammalian two-hybrid system assays
To analyse the genome-wide protein-protein interactions of SARS-CoV-2, We cloned all the coding sequences of Nsps and ORFs into the pM and pVP16 vectors separately (Table 1). To achieve a better expression, we separated the coding sequence of Nsp3 into 3 parts, Nsp3.1, Nsp3.2, and Nsp3.3, which was designed according to the known functional domains of Nsp3 (Figure S1). S was separated at the cleavage site of furin into S1 and S2. The sensitivity and efficiency of the mammalian two-hybrid system were confirmed by detecting the interaction of p53 and SV40T, which could lead to more than 130 times of increase in the relative expression level of reporter genes compared with negative controls (Figure 1A).
We screened 784 interaction combinations between all the coding sequences of Nsps or ORFs of SARS-CoV-2, and the assays on each combination were repeated at least 3 times (figure 1B). 48 positive interaction combinations were screened out and analyzed using Cytoscape software to visualize molecular interaction networks (figure 1B and 1C). Some interactions were confirmed using co-immunoprecipitation (figure 2). Besides 5 self-interactions, 5 out of 48 interactions between different viral proteins, Nsp1/Nsp3, Nsp3/N, Nsp3/Nsp12, Nsp10/Nsp14, and Nsp10/Nsp16, were examined to be positive bidirectionally and 79.2% interactions could only be detected in one direction, indicating that the fusion domains may influence the interactions, which happened in our previous studies15. We identified 14 novel interactions, including Nsp1/Nsp7, Nsp1/M, Nsp1/Nsp3.2, Nsp1/ORF7a, Nsp3.1/Nsp10, Nsp3.1/ORF3, Nsp3.2/Nsp10, Nsp3.3/Nsp3.1, Nsp12/M, E/Nsp4, ORF6/Nsp4, ORF6/Nsp10, ORF6/M, and Nsp3.1/N.
Unlike the Nsp1 of SARS-CoV, which barely interacts with other viral proteins, Nsp1 of SARS-CoV-2 has the most interaction partners, 10 Nsps and 7 accessory proteins, among all the viral proteins, and likely functions as the hub of the group of viral proteins (figure 1B and 1C). As the first viral protein being expressed after viral infection, the function of Nsp1 as a hub likely was an advantage for virus to organize the assembly of replicase complex and replication of viral genomic RNAs. Besides its function in suppression of protein expression and defense against host innate immune response16, the vital role of Nsp1 in the viral replication is supported by the clinical observation that a deletion in the C-terminus of Nsp1 attenuated the viral pathogenicity17. Nsp3 is one of the most complex proteins and has multiple functional domains18. The well-known function of Nsp3 is to process viral polyprotein at cleavage sites of Nsp1/2, Nsp2/3 and Nsp3/4. In our interaction screening, Nsp3 was found to interact with Nsp10, Nsp12, Nsp13, and Nsp14, indicating its possible roles in the replication/transcription of viral RNAs besides the process of viral polyprotein. Nsp5 is responsible for the cleavage at the sites separating Nsp4 to Nsp16, spanning 1a and 1b regions. Similar to SARS-CoV, Nsp5 of SARS-CoV-2 interacts with Nsp12, indicating that Nsp12 could be the sites in 1b for Nsp5 to grasp its substrate15. In agreement with the previous findings that Nsp8 and Nsp12 form the core RNA polymerase complex, Nsp8 of SARS-CoV-2 interacts with Nsp1219, 20. As an RNA binding protein, Nsp8 may facilitate the substrate recognition of Nsp12. We identified 10 interactions Nsp10 involved in, and among them, the interactions of Nsp10/Nsp14 and Nsp10/Nsp16 were also uncovered in our previous work for SARS-CoV15, indicating a conservative mechanism of the genus Betacoronavirus for regulation of the activity of methyltransferase of Nsp14 and Nsp16. We also identified many interactions between Nsps and accessory proteins, indicating that the possible roles of accessory proteins in the replication/transcription of viral RNAs. 5 self-interactions, Nsp3, Nsp5, Nsp15, and N reported here were also found in SARS-CoV, except for ORF8, which is one of the most distinct ORFs between SARS-CoV and SARS-CoV-221, 22.
Nsp3 interacts with N protein
N protein is one of four well known structural proteins identified in the viral particles5. It forms a long helical nucleocapsid structure which viral RNA was packed inside. This ribonucleoprotein (RNP) complex protects the viral RNA from the attack of host nucleases and recognition of host nucleotide sensors triggering the immune response23. This structure could also play an essential role in the replication/transcription of viral RNA15. However, the molecular details about the role of N in replication were obscure. In this screening, the interaction between Nsp3.1 and N was among the strongest ones (figure 1A) and was confirmed by co-immunoprecipitation (figure 2G). Since Nsp3 is the component of the viral replication and transcription complex (RTC), this interaction suggested the N could regulate the replication of viral RNA through the association with Nsp3.
N protein interacts with Nsp3 through its NTD domain
N protein has three major defined domains, N-terminal domain (NTD), serine-arginine-rich (SR) domain, and C-terminal domain (CTD) (figure 3A)24, 25. To determine N protein’s key domains interacting with Nsp3.1, we examined the interactions between various domains of N and Nsp3.1 using co-immunoprecipitation. NTD retains the capability to interact with Nsp3.1, while this capability of N protein is largely lost in CTD (figure 3B and 3C). We also examined the locations of N and Nsp3.1 proteins in the cells. The immunostaining results showed that similar to the wild-type (wt) N, NTD colocalized with Nsp3.1 in the 293T cells, while CTD lost the colocalization with Nsp3.1 (figure 3D, 3E, and 3F). We could not detect the expression of the coding sequence of SR, which happens typically in the expression of proteins with a molecular weight of less than 15 kDa. In our laboratory practice, we increase the size of protein by fusing our target protein with a tag protein, like EGFP, which has a relatively independent structure and unlikely interferes with the function of target protein. As predicted, we detected the decent expression of EGFP-tagged SR. However, Nsp3.1-HA could not be detected in the pull-down samples of EGFP-tagged SR, similar as that of EGFP, while, in contrast, the decent level of Nsp3.1-HA was detected in that of EGFP-tagged N (figure 3G and 3H). Moreover, NTD seems to interact with Nsp3.1 in a stronger manner than N protein, indicating that the other domains of N protein could negatively impact the interaction with Nsp3.1 (figure 3C).
Nsp3 interacts with N through its acidic domain
Thus far, limited knowledge about the functions and structures of Nsp3.1 was available. Based on analysis of the protein sequence, we found a special domain rich in negatively charged amino acids in the N-terminus of Nsp3.1 (figure 4A). As a nucleic acid-binding protein, N protein has a 10.1 of pI and is composed of many positively charged amino acids, which facilitate its interaction with nucleic acids with negative charges and likely also with acidic proteins (figure S2). Accordingly, we first examined the interaction between N protein and the aa 1-235 of N-terminus, which possessed the most acidic amino acids in Nsp3.1. Despite the loss of nearly two-thirds of Nsp3.1, aa 1-235 retained the capability to interact with N protein which was even stronger than that of the full length of Nsp3.1 (figure 4B and 4C). To further narrow down the region that interacts with N protein, we removed aa 1-102, which has the most acidic amino acids, from Nsp3.1 and examined its interaction with N protein. As predicted, the deletion of aa 1-102 largely abolished Nsp3.1/N interaction (figure 4G and 4H). By observing the colocalizations between N and truncated Nsp3.1, we confirmed that only aa 1-235 of Nsp3.1 plays an indispensable role in the interaction between Nsp3.1 and N (figure 4D, 4E, and 4F).
Nsp3 and N formed a protein complex in vitro
Next, we sought to confirm N could directly interact with Nsp3. Since aa 1-102 of Nsp3.1 was indispensable for N-Nsp3.1 interaction, we used aa 2-111 of Nsp3.1 instead of Nsp3.1 to check the interaction in vitro. We co-expressed aa 2-419 of N and aa 2-111 of Nsp3.1 proteins in the bacteria, and the two proteins were purified together. Gel filtration analysis showed aa 2-419 of N and aa 2-111 of Nsp3.1 migrated together, and their molar ratio in the protein complex is close to 1. Through the conversion of elution volume to the approximate molecular weight, at least two aa 2-419 of N proteins could be in the fraction of complex peak, indicating that the interaction sites between Nsp3 and N should be different from the sites for the formation of the oligomer of N and that of Nsp3 (figure 5A), and the recognition of replicase complex on N through Nsp3 should not disturb the structure of N oligomers. Consistent with our IP results, Partial (figure 4B) or complete (figure 4C) deletion of CTD retained the interaction with Nsp3.1.
Sequence analysis indicated that 43 basic amino acids at N-terminus of N were likely dispensable for the interaction if the N-Nsp3.1 interaction depended on the attraction between the acidic and basic domains. Indeed, gel filtration analysis confirmed that the deletion of 43 aa did not impact the interactions and the molar ratio of the interaction (figure 5E and 5F).
The interaction between Nsp3 and N protein played an essential role in the replication and transcription of viral genomic RNAs
As the direct interacting protein with viral genomic RNAs, N protein composed the nucleocapsid structure wrapping viral RNA and thus could be involved in viral replication and transcription5, 15. Nsp3 is processed from the viral polyprotein, which composed RTC and thus it could also join in the viral replication and transcription. Although both could play important roles in viral replication and transcription, whether the association of RTC and N of SARS-CoV-2 was essential for viral replication and transcription was not defined.
We investigated whether inhibition of the interaction between Nsp3 and N could influence the replication and transcription of viral genome. Firstly, we determined whether the interaction could be inhibited. To this end, we constructed the N or Nsp3.1 fused with Nuclear Localization Sequence (NLS). Next, we investigated whether NLS-N or NLS-Nsp3.1 could disturb the interaction between BD-N and AD-Nsp3.1 in the nucleus. Indeed, both NLS-N or NLS-Nsp3.1 could inhibit the interaction between BD-N and AD-Nsp3.1 in a dose-dependent manner (figure 6B). We also confirmed the aa 1-235 of Nsp3.1 could compete with Nsp3.1 in the interaction with N protein using co-immunoprecipitation (figure 6C and 6D).
Since the limited availability of biosafety level 3 (BSL3) laboratory, we utilized the viral replicon instead of live SARS-CoV-2 as a model to study the impact of inhibition of the interaction between Nsp3.1 and N on the viral replication (figure S3). The replicon of SARS-CoV-2 (nCoV-replicon) constructed by our lab expressed the S gene-deleted full-length RNA of viral genome. The deletion of S gene abolished the generation of viruses which raised concerns of biosafety. We replaced the coding sequence of S gene with the reporter gene firefly luciferase, and the replicon with firefly luciferase is named as Rep-Luci. The activity of firefly luciferase could reflect the level of replication and transcription of the replicon (manuscript in submission).
To investigate the role of viral proteins on viral replication and transcription, we expressed plasmids expressing viral proteins, Rep-Luci and RL-TK plasmids in 293T cells and measured the relative luciferase activities (figure 6A). Most of viral proteins promoted the activity of Rep-Luci, but Nsp3.1 inhibited the activity, indicating its potential as an inhibitor for viral replication. We further examined the inhibitory effect of aa 1-80, aa 1-101, or aa 1-235 of Nsp3 on the activity of Rep-Luci (figure 6E). All of the truncated Nsp3.1 proteins, as well as full-length Nsp3.1 protein, inhibited the replication and transcription of replicon, and the inhibitory effect of aa 1-235 of Nsp3 was in a dose-dependent manner (figure 6F). Except that the inhibitory effects of aa 1-80 and aa 1-101 of Nsp3.1 protein were comparable, the inhibitory effects increased in the order of aa 1-101, aa 1-235, and the full-length of Nsp3.1. Interestingly, the coverage of the region of Nsp3.1 of these truncated mutants was also increased in this order. The co-immunoprecipitation results showed that the interaction between aa 1-235 and N was much stronger than that between wt Nsp3.1 and N, indicating that Nsp3.1 could inhibit the function of Nsp3 independent of its interaction with N. As the truncated form of natural Nsp3 protein, Nsp3.1 could exhibit a dominant-negative effect through inactivation of the other function of wt Nsp3 protein.