key: cord-0779026-hvrt5r0n
authors: Graziadei, Andrea; Schildhauer, Fabian; Spahn, Christian; Kraushar, Matthew; Rappsilber, Juri
title: SARS-CoV-2 Nsp1 N-terminal and linker regions as a platform for host translational shutoff
date: 2022-02-10
journal: bioRxiv
DOI: 10.1101/2022.02.10.479924
sha: 3135b114a0960b2063360bfb425dcec1783ada5a
doc_id: 779026
cord_uid: hvrt5r0n

In the early stages of SARS-CoV-2 infection, non-structural protein 1 (Nsp1) inhibits the innate immune response by inserting its C-terminal helices into the mRNA entry channel of the ribosome and promoting mRNA degradation. Nevertheless, the mechanism by which Nsp1 achieves host translational shutoff while allowing for viral protein synthesis remains elusive. We set out to characterize the interactome of full-length Nsp1 and its topology by crosslinking mass spectrometry in order to investigate the role of the N-terminal domain and linker regions in host translational shutoff. We find that these regions are in contact with 40S proteins lining the mRNA entry channel and detect a novel interaction with the G subunit of the eIF3 complex. The crosslink-derived distance restraints allowed us to derive an integrative model of full-length Nsp1 on the 40S subunit, reporting on the dynamic interface between Nsp1, the ribosome and the eIF3 complex. The significance of the Nsp1-eIF3G interaction is supported by further evidence that Nsp1 predominantly binds to 40-43S complexes. Our results point towards a mechanism by which Nsp1 is preferentially recruited to canonical initiation complexes, leading to selective inhibition of host-translating ribosomes and subsequent mRNA degradation.

Coronaviruses (CoVs) are positive-sense single-stranded mRNA viruses capable of infecting a large variety of vertebrate species, causing mild to severe respiratory diseases, including severe acute respiratory syndrome (SARS). SARS-CoV-2 is the causative agent of the highly pathogenic respiratory disease COVID- 19 . It belongs to the beta genus of CoVs that includes SARS-CoV and the Middle East respiratory syndrome (MERS) CoV.

Beta-CoVs extensively remodel cell morphology (1) and gene expression upon infection in a time-dependent manner (2, 3) by a combination of viral proteins, transcripts and selective activation of host cell pathways. The hijacking of protein synthesis is a key phenotype of CoV infection, leading to ribosomes selectively translating viral factors and some host proteins. Shortly after internalization, SARS-CoV employs its non-structural protein 1 (Nsp1) to bind the small ribosomal subunit and inhibit canonical mRNA translation, resulting in cellular mRNA degradation (4, 5) . This occurs via an unknown mechanism involving downstream endonucleolytic cleavage of host mRNAs (5, 6) . The resulting suppression of host protein synthesis and immune response includes the translation-dependent type I interferon pathway (7), a phenotype known as host translational shutoff.

CoV-2 Nsp1, which shares 84% sequence identity with CoV Nsp1 (8) , has been identified as a binder of the 40S ribosome in both biochemical and structural work (9, 10) .

The protein has an N-terminal domain (NTD), a linker region and a small C-terminal domain (CTD) made up of two short alpha helices. Both CoV and CoV-2 Nsp1 can induce cleavage of capped and IRES-containing mRNAs in vitro (4, 11, 12) .

Recent cryo-electron microscopy (cryo-EM) structures have illustrated the mechanism by which the Nsp1 CTD renders the 40S and 80S ribosomes, and several 43S initiation complexes, translationally incompetent (9, 10) . The helices making up the C-terminal domain of Nsp1 plug the mRNA entry channel on the 40S subunit, preventing translation. However, only the C-terminal domain and a short portion of the linker were resolved in the structures, presumably due to flexibility of the rest of the protein. Biochemical work has since then highlighted the role of the N-terminal domain and the linker in viral evasion of translational shutoff (6, 13, 14) . However, the mechanism by which this occurs remains unknown.

Here, we characterize the conformation of full-length Nsp1 on the 40S and 43S complexes by crosslinking mass spectrometry, and probe its network of protein-protein interactions. Crosslinking MS shows that the previously unresolved Nsp1 NTD, while highly flexible, occupies a well-defined volume on the 40S subunit, and that its linker contributes to interactions with the 40S, contextualizing previously described clinical variants. Using the crosslinking MS data, we build an integrative model combining current structural knowledge with spatial restraints for Nsp1. We then characterize the Nsp1-bound proteome by affinity purification mass spectrometry (AP-MS) and ultracentrifugation. Taken together, our results show that Nsp1 preferentially binds to 40S subunits and 43S preinitiation complexes, and that both the linker and NTD regions make extensive interactions with the 40S subunit and eukaryotic initiation factors (eIFs). The data point to a role of the eIF3 subunit eIF3G, which can be unambiguously assigned to be in close proximity to Nsp1 near the mRNA entry site based on crosslinks, clarifying the identity of the protein occupying low-resolution density in previous structures (10) . We hypothesize that these dynamic interactions by the Nsp1 linker and NTD to initiation factors contribute to the preferential targeting of canonical initiation complexes, and that translation on the viral 5' UTR may proceed in an eIF3G-independent manner.

In order to understand the topology of the protein-protein interaction network involved in Nsp1-mediated translational repression, we performed a crosslinking-MS analysis of HEK293T cells overexpressing FLAG-tagged Nsp1. After cell lysate was crosslinked with disuccinimidyl sulfoxide (DSSO), the Nsp1-bound proteome was enriched by affinity purification, digested and analyzed by crosslinking-MS. We identified 1,269 crosslinks involving the same protein (self-links) and 869 heteromeric crosslinks at a 2% residue-pair false discovery rate (FDR), resulting in the identification of 515 protein-protein interactions (PPIs) at a 6% FDR. Identified crosslinks covered the ribosome, signal recognition particle, chaperonin-containing T-complex, as well as eIF1, eIF3, eIF2, and eIF4F complexes. As crosslinks directly report on the proximity of residues, we could map the binding site of Nsp1 on ribosomes in solution (Fig. 1) . This was consistent with the arrangement proposed by cryo-EM structures, though several additional interactions were detected involving the regions of Nsp1 not previously resolved. We detected 42 crosslinks between Nsp1 NTD and linker to 40S ribosomal proteins. These include a previously reported interaction with RS2 (uS5) and RS3 (uS3) (9, 10, 15) , and novel contacts with RS9 (uS4), RS10 (eS10), RS12 (eS12), RS17 (eS17), and RS27A (eS31), which line the cavity near the mRNA entry site.

We used the restraints derived from crosslinking-MS data to derive a model of the interactions of full-length Nsp1 on 40S ribosomal subunits using the integrative modeling platform (IMP) (16) (Fig. S1) . A coarse-grained representation of the Nsp1-bound 40S initiation complex, including all regions not resolved in structures, was used in the derivation of models representing conformations consistent with crosslinking and physical restraints.

In the models, Nsp1 NTD occupies a well-defined volume in the cavity facing the mRNA entry site, interacting with S3 ( Fig. 1C, 1D , S3, S4). Nevertheless, multiple positions of the NTD are consistent with crosslinking restraints (Fig. S3) , indicating the dynamic nature of the Nsp1 NTD on the 40S complex and explaining why it had not been resolved in previous structures. In particular, several crosslinks identify a secondary extended conformation in which the Nsp1 NTD lies in proximity to RS10 and RS12 (Fig. 1C) .

The Nsp1 linker region has multiple crosslinks to RS2, RS3, RS10 and RS12. The localization probability density of the linker in our final model indicates that the linker is largely restricted to the right side of the cavity, localizing below the plane of the Nsp1 NTD.

Indeed, the recent observations that the NTD and linker regions are required for translational suppression (13) are consistent with our findings that the linker is not a passive bystander of Nsp1-ribosome recognition, but rather makes extensive and dynamic contacts with the cavity facing the mRNA entry channel.

We further characterized the nature of the Nsp1-bound proteome by affinity purification mass spectrometry (AP-MS) experiments in HEK293T cells. Quantitative AP-MS analysis identified 271 high-confidence co-purifying proteins (Fig. 2 , supplementary table S3), observing good correlation between experimental replicas (R=0.9-0.95, Figure S6 ). As expected, ribosomal proteins made up a majority of co-purifying proteins.

Gene Ontology (GO) enrichment analysis showed translation initiation, overall translation and several terms associated with ribosome biogenesis among the most highly enriched terms (Fig. S7 ). These findings are consistent with literature showing Nsp1-mediated shutdown of host mRNA-translating ribosomes in the early stages of infection (2, 3, 17) .

However, translation initiation factors were among the most highly enriched proteins in the dataset, surpassing even ribosomal subunits that were identified by structural models as direct Nsp1 binders. In addition to factors identified in cryo-EM studies (LYAR, PAIRB, TSR1) (9, 10), we also detected subunits of the exosome complex and the exon junction complex, possibly indicating the nature of downstream mRNA pathways acting on mRNA-stalled ribosomes, and consistent with reports of nuclear activity of Nsp1 (18) . The E3-ubiquitin ligase ZNF598 also co-purified in the pulldown, indicating an activation of the ribosome quality control pathway on Nsp1-stalled complexes. While XRN2 was found to be enriched in the AP-MS data, XRN1, the key 5'->3' helicase involved in nonsense-mediated RNA decay, was not detected. The CCCH-type antiviral protein 1 (ZCCHV/ZAP), a key factor in viral response, 3-5' mRNA decay and decapping, was also enriched in the Nsp1-bound proteome. The presence of these factors in Nsp1 pulldowns from cells overexpressing Nsp1 is somewhat surprising, but ZAP and the exon junction complex have also been found to modulate infection response in SARS-CoV-2 infected cells (3), though this is thought to occur via mechanisms involving Nsp16 and other viral proteins.

The top hit by fold enrichment in the AP-MS screen is eIF2α (EIF2S1), a key player in the integrated stress response pathway (19) , whose phosphorylation by stress-sensing kinase leads to suppression of translation initiation. Phosphorylation of eIF2α is known to be inhibited by beta-CoV factors such as the MERS 4a protein (20, 21) . Interestingly, the interferon-inducible activator of the EIF2AK2 stress-sensing kinase, PRKRA, is also detected as significantly enriched in the AP-MS screen, along with the relevant type I interferon pathway protein DDX60.

In addition to those initiation factors found in previously described cryo-EM structures (eIF3, eIF2, eIF1, eIF1AX) (9, 10) , other canonical components of the 43S initiation complex were detected, such as eIF1B and the helicase DHX29. The subunits of the eIF4F complex, eIF4B and 4IF4H were also found to be enriched. These proteins are involved in unwinding structures in the 5' UTR region and mediating the attachment of the 43S complex to the mRNA, enabling scanning upon ATP hydrolysis (22) (23) (24) . Intriguingly, the polyA binding protein (PABP), an integral part of the eIF4F/eIF4B-mRNA complex that is loaded onto the 43S initiation complex to form the closed loop 48S complex, is not among the enriched proteins. Thus, Nsp1 seems to be stabilizing late-stage 43S intermediates, as well as to be co-purifying with eIF4F/eIF4B. This interaction may occur on the ribosome itself, or via a previously unreported interaction between Nsp1 and eIF4F/eIF4B outside the ribosome. The eIF2 guanosine exchange complex eIF2B was also detected among the enriched translation initiation factors.

The high abundance of translation initiation factors in our data led us to hypothesize that Nsp1 binding to ribosomes is not purely mediated by the interactions between the C-terminal helices and the mRNA entry site, which are shared between late-stage initiation complexes and elongating 80S ribosomes. We therefore analyzed the distribution of ribosomal complexes in cells overexpressing Nsp1 by sucrose gradient centrifugation. Here, we could observe a difference between the subunit distributions of cells overexpressing EGFP versus Nsp1 (Fig.2) , which reproducibly showed an accumulation amount of free 40S subunit and 43-48S complexes for Nsp1.

Proteomic analysis of the sucrose gradient fractions confirmed that Nsp1 preferentially binds pre-initiation intermediates. This co-fractionation analysis indicated that Nsp1 is mainly found bound to 40S subunits and 43S ribosomal initiation intermediates (Fig.   2D ). Interestingly, only a small fraction of Nsp1 fractionated with 80S ribosomes. Instead, Nsp1 showed coelution behavior with subunits from the eIF3 complex (Fig. S8) to which we had found extensive crosslinks. Moreover, the decrease of 80S ribosomes in the Nsp1-overexpressing cells did not lead to a corresponding 60S increase (Fig. 2C ).

This behavior cannot be explained by an mRNA "steric plug" model of Nsp1-mediated translational repression (9), as it requires a targeted shutdown of specific initiation intermediates over the bulk actively translating ribosome and polysome population.

It is also worth noting that this specific block at the 43S stage would be consistent with the reported role of Nsp1 in sensing the 5'-UTR of the viral mRNA in viral evasion of translational shutdown (13) .

Consistent with the enrichment of eIF3 components in Nsp1 AP-MS, we observed extensive crosslinking between the Nsp1 linker and N-terminal domain regions and the G subunit of the eIF3 complex. With 18 detected crosslinks, eIF3G was detected as the primary interactor for Nsp1 in the sample. Using crosslinks, we could identify that the unidentified RNA-recognition motif (RRM) domain found near the mRNA entry site in cryo-EM structures (10) is the RRM domain of eIF3G. This is also consistent with the identification of this domain in the structure of the 48S complex (23) . The RRM domain is connected by a linker to the N-terminal region of eIF3G, which we could model positioned across the surface of the RNA-entry site ( Figure 3C ).

While most crosslinks are between Nsp1 and eIF3G linker regions, structural visualization of the crosslinks on the eIF3G RRM domain showed that residues crosslinked to Nsp1 are positioned on the solvent-exposed face of the domain, consistent with their interaction occurring within the 43S complex. These multiple interactions are in line with the co-fractionation behavior of the two proteins (Fig. 2D, Fig. S8 ). Taken together, AP-MS and crosslinking-MS provide evidence for a direct or RNA-mediated interaction between Nsp1 and eIF3G in the context of late 43S initiation complexes. This is further bolstered by the detection of an extensive network of crosslinks involving eIF3G and Nsp1 K125, a residue that is required for translational repression and mRNA depletion (12) .

As indicated by previous structural work and by our proteomic analysis of ribosome preparations, Nsp1 can bind multiple ribosomal states all the way up to late-stage 43S initiation complexes. The eIF3G RRM covers up RS3, thus precluding some crosslinks satisfied in the model of Nsp1 on the 40S subunit from being satisfied in the 43S-bound state ( Fig. S4) . Indeed, the contacts formed by Nsp1 on the surface of the 43S preinitiation complex clearly show that a single conformer cannot explain all observed crosslinks, and that multiple conformations of the 43S complex must also be present in solution. In particular, our crosslinks indicate multiple conformations within the eIF3 core relative to the 40S body, and strong dynamics for the eIF3I-eIF3B-eIF3G N-terminal region module relative to the mRNA entry cavity. Fluorescence-based measurements have shown that binding of Nsp1 to the small ribosomal subunit is modulated by initiation factors, with eIF3J competing with Nsp1 for binding to the 40S (25).

Other direct Nsp1 interactions included eIF4B and the translational repressor PAIRB/SERBP1, although both of these PPIs were identified by a single crosslink.

Interestingly, the dataset provided a glimpse on the extensive interactions formed by the highly dynamic PAIRB on the ribosome. As Nsp1 and PAIRB share a binding site in the mRNA entry channel, it is possible that PAIRB is displaced there by the Nsp1 CTD (9), leading to a previously-uncharacterised "unplugged" PAIRB-bound conformation of the 40S subunit.

The crosslinks also identified the binding sites of some previously uncharacterized proteins, including the stress granule protein PRC2C, among the top enriched proteins from the AP-MS analysis, which is found crosslinked to eIF3A.

The extensive network of interactions formed by the Nsp1 NTD and linker regions provide insights into the targeting of initiation intermediates by Nsp1, and are consistent with the proposed role of these regions in host shutoff. These interactions may indeed provide a platform for the recruitment of the unknown downstream mRNA degradation machinery, for which we provide plausible candidates in our AP-MS study.

While the cryo-EM structures provided a clear mechanism for Nsp1-mediated host shutoff via its CTD, it has become clear that the Nsp1 NTD and linker are involved in additional mechanisms that allow for the escape of translational repression on viral transcripts and the targeting of host-translating ribosomes.

Our AP-MS results recapitulate the findings of cryo-EM, clarifying earlier results that did not detect an enrichment in ribosomes in the Nsp1-bound proteome, presumably because of the use of C-terminal tags or the high background of the ribosome (26, 27) .

Proteomic analysis of ribosome populations further confirmed that full-length Nsp1 preferentially binds 40S and 43S complexes, which may be explained by a direct or RNA-mediated interaction between Nsp1 and eIF3G.

The crosslinking-MS data show that full-length Nsp1 is involved in a wide range of protein-protein interactions on the platform provided by the 40S subunit or the 43S preinitiation complex. We obtain structural information on these interactions via crosslinking-MS. Our results show that, despite its flexible position, the Nsp1 NTD interacts with the cavity facing the mRNA entry site on the 40S subunit and with the RRM domain of eIF3G. Moreover, the crosslinks further clarify current structural models of Nsp1-bound ribosomes and initiation complexes by unambiguously placing the eIF3G RRM and providing alternative conformations for the eIF3B-eIF3I module.

The interactions reported in our study provide context for recent observations that the NTD and linker regions are required for translational suppression. Indeed, K125, a residue required for 40S subunit binding in in vitro assays (12) , is found extensively crosslinked to eIF3G and several 40S subunits (Fig. 1A) . In integrative models, this residue is located on the face of the Nsp1 NTD that interacts with RS3 (Fig. 1G) . The model also accounts for the correlation of deletion of the 500-532 genome locus, located in the NTD β-sheet, with lower viral loads and non-severe infection traits (28) . These residues may also be involved in contacts with the eIF3G RRM domain.

The dynamic nature of the Nsp1 linker is recapitulated in our data, as the large localization probability density for this region shows. Interestingly, NMR studies of Nsp1 in isolation (29) have shown that the linker is partially but not completely disordered, which is consistent with our data showing a preference for a localization on the right-side of the ribosomal entry site cavity. Deletions in the linker residues K141-S142 in SARS-CoV Nsp1 impairs the translation shutoff function in CoV Nsp1 (30) , indicating that this function may proceed downstream of ribosome binding. A similar deletion, K141-F143, has been observed in some SARS-CoV-2 patients (31). Our integrative model places these residues near RS3. The importance of the Nsp1 NTD and linker is underscored by their highly conserved nature between SARS-CoV species (10, 32) and within SARS-CoV-2 lineages In addition to its role in translational shutoff, Nsp1 is reported to block mRNA export from the nucleus via an interaction with the NXF1 export factor (18) . While our AP-MS protocol is performed after pelleting cell nuclei, we note that the required region for Nsp1 interaction contains the NXF1 RRM domain, which may be further evidence of the ability of Nsp1 to interact with RRM domains.

Our model is consistent with results that indicate a competition between Nsp1 and eIF3J binding to the ribosome (25) . In eIF3J-bound, closed head conformations (34), the space occupied by the Nsp1 NTD in our main cluster of models clashes with the locked rRNA (Fig. S10) . Indeed, eIF3J is less enriched than the core eIF3 complex in AP-MS, suggesting a preference for inhibiting open head conformations. The marked enrichment of the ABC-type ATPase ABCE1, found in some eIF3J-bound 43S states (34) is consistent with the presence of this factor in Nsp1-bound ribosomes (10).

Our work explains the high conservation of the Nsp1 NTD and linker in coronaviruses. Far from being a simple "steric plug", the NTD and linker are involved in an extensive network of protein-protein interactions. The model presented here also provides a framework for understanding the effects of previously reported deletions (13) in terms of protein conformational dynamics. We speculate that the interactions may partially explain the selectivity of Nsp1 for host-translating ribosomes, since translation initiation on a viral transcript may occur via an eIF3G-independent mechanism. Alternatively, the interaction to eIF3G may sequester this protein off of the ribosome. Nevertheless, it is clear that the Nsp1 NTD and linker provide an interaction platform for efficient translational repression and host shutoff. In-lysate crosslinking and affinity purification. 

Lumos Tribrid mass spectrometer (Thermo Fisher Scientific, Germany) connected to an Ultimate 3000 RSLCnano system (Dionex, Thermo Fisher Scientific, Germany), which were operated under Tune 3.4, SII for Xcalibur 1.6 and Xcalibur 4.4. Fractions from peptide SEC were resuspended in 1.6% ACN 0.1% formic acid and loaded onto an EASY-Spray column of 50 cm length (Thermo Scientific) running at 300 nl/min. Gradient elution was performed using mobile phase A of 0.1% formic acid in water and mobile phase B of 80% acetonitrile, 0.1% formic. For each SEC fraction, we used an optimized elution gradient (from 2-18% mobile phase B to 37.5-46.5% over 90 min, followed by a linear increase to 45-55 and 95% over 2.5 min each). Each fraction was analyzed in duplicates. The settings of the mass spectrometer were as follows: Data-dependent mode with 2.5s-Top-speed setting; MS1 scan in the Orbitrap at 120,000 resolution over 400 to 1500 m/z with 250% normalized automatic gain control (AGC) target; MS2 scan trigger only on precursors with z = 3-7+, AGC target set to "standard", maximum injection time set to "dynamic"; fragmentation by HCD employing a decision tree logic with (35) optimized collision energies; MS2 scan in the Orbitrap at a resolution of 60,000; dynamic exclusion was enabled upon a single observation for 60 s.

Each LC-MS acquisition took 120 min.

Protein identifications in pooled peptide SEC fractions from in-lysate crosslinking were conducted using a Orbitrap Fusion Lumos Tribrid mass spectrometer (Thermo Fisher Scientific, Germany) connected to an Ultimate 3000 RSLCnano system (Dionex, Thermo 

Files from raw data obtained from linear identification of proteins were pre-processed using MaxQuant (1.6.17.) (36) . Default settings with minor changes were used: three allowed missed cleavages; up to four variable modifications per peptide including oxidation on Met, acetylation on protein N-terminal. Carbamidomethylation on Cys was set as fixed. For the search 'matching between runs' feature was enabled with default settings. The full human proteome with 20.371 proteins plus amino acid sequence of Nsp1 was used. For protein quantification two or more peptides using the iBAQ approach were applied.

Raw files from crosslinking-MS acquisitions were converted to mgf-file format using MSconvert and were recalibrated to account for mass shifts during measurement. Additionally the maximum number of variable modifications per peptide was set to 1 and the additional loss masses were defined accounting for its cleavability ("A" 54.01056 Da, "S" 103.99320 Da, "T" 85.98264). Defined crosslink sites for DSSO were allowed for side chains of Lys, Tyr, Ser, Thr and the protein N-terminus. The database was composed of 400 Swiss-Prot annotated entries for Homo sapiens (Human) (taxon identifier 9606) with the highest abundance with the addition of the sequence of Nsp1. Results were filtered prior to FDR to matches having a minimum of three matched fragments per peptide, a delta score of > 10% of the match score and a peptide length of at least five amino acids. Additionally, spectral matches were prefiltered before FDR estimation to only those that had cleaved crosslinker peptide fragments for both peptides. Results were then to an estimated false-discovery rate (FDR) of 2% on residue-pair-level using xiFDR (version 2.1.5.2) (37).

The resulting estimated protein-protein interaction FDR was 6%. Self-and heteromeric-crosslinks were handled separately for FDR estimation.

HEK293T cells for the AP-MS were transfected with pcDNA5_FRT_TO-3xFLAG-3C-Nsp1 or pcDNA3 (control) in triplicates as described above. Each replica of HEK293T cells 

Eluates from affinity purification were mixed with four times ice-cold Acetone (Sigma Aldrich) and NaCl for precipitation of proteins. The mixture was stored at -20°C for 1 h and later 

AP-MS experiments were acquired using a Orbitrap Fusion Lumos Tribrid mass spectrometer (Thermo Fisher Scientific, Germany) connected to an Ultimate 3000

RSLCnano system (Dionex, Thermo Fisher Scientific, Germany), which were operated under Tune 3.4, SII and Xcalibur 4.4. 0.1% (v/v) formic acid and 80% (v/v) acetonitrile, 0.1% (v/v) formic acid served as mobile phases A and B, respectively. Samples were resuspended in 2% acetonitrile, 0.1% formic acid before injection onto an Easy-Spray column (C18, 50 cm, 75 µm ID, 2 µm particle size, 100 Å pore size) operated at 45 °C and running with 300 nl/min flow. Peptides were eluted with the following gradient: 0 to 2% buffer B in 2 min, 2 to 7.5% B in 5 min, 7.5 to 42.5%B in 80 min, 42.5 to 50% in 2.5 min. Then, the column was set to washing conditions within 3.5 min to 95% buffer B and flushed for another 5 min. For the mass spectrometer the following settings were used: MS1 resolution = 120,000; AGC target = 250%; maximum injection time = auto; scan range from 375 to 1500 m/z; RF lens = 30%.

The selection criteria for ions to be fragmented were: intensity threshold > 5.0e3; charge states = 2-6; MIPS = peptide. Dynamic exclusion was enabled for 30 s after a single count and excluded isotopes. For MS2 the settings were: quadrupole isolation window = 0.4 m/z; minimum AGC target = 2.5 × 10e4; maximum injection time = 80 ms; HCD (higher-energy collisional dissociation) fragmentation = 30%. Fragment ion scans were recorded with the iontrap in 'rapid' mode with: mass range = normal. Each LC-MS acquisition took 120 min.

Raw data from mass spectrometry were searched as described above using MaxQuant (1.6.17.). The obtained LFQ values for each replica were normalized according to the sum of absorbance units from peptide size exclusion run. Enrichment analysis was performed in Perseus (1.5.6.0) (38) . Normalized LFQ intensities were log-2 transformed and filtered to proteins detected in all three replicas in either +Nsp1 or control experiment ( Figure S6) .

Missing values were then imputed based on a random selection from a normal distribution, downshifted by 1.8 standard deviations shrunk by a factor of 0.3. Significance was determined by performing a two-tailed t-test and imposing a significance cutoff threshold of FDR< 0.05 and ≥2-fold differential abundance on a log 2 scale. FDR was derived by permutation.

HEK293T cells were transfected with pcDNA5_FRT_TO-3xFLAG-3C-Nsp1 or with pEGFP-C1 in triplicates as described above. Lysis and sucrose gradient ultracentrifugation as previously described (39) 

Proteins from sucrose gradient fractions were acquired in a Q Exactive HF mass spectrometer (Thermo Fisher Scientific, Bremen, Germany) coupled to an Ultimate 3000 RSLC nano system (Dionex, Thermo Fisher Scientific, Sunnyvale, USA), operated under Tune 2.9, SII for Xcalibur 1.4 and Xcalibur 4.1. 0.1% (v/v) formic acid and 80% (v/v) ACN, 0.1% (v/v) formic acid served as mobile phases A and B, respectively. Samples were loaded in 1.6% acetonitrile, 0.1% formic acid on an Easy-Spray column (C18, 50 cm, 75 µm ID, 2 µm particle size, 100 Å pore size) operated at 45 °C and running with 300 nl/min flow. Peptides were eluted with the following gradient: 2 to 4% buffer B in 1 min, 4 to 6% B in 2 min, 6 to 37.5%B in 72 min, 37.5 to 42.5% in 5 min, 42.5% to 50% in 6 min, 50% to 90% in 3 min and hold at 90% for 7.5 min followed by 90 to 2%B in 23.5 min. For the mass spectrometer the following settings were used: MS1 scans resolution 120,000, AGC target 3 × 10 6 , maximum injection time 50 ms, scan range from 350 to 1600 m/z. The ten most abundant precursor ions with z = 2-6, passing the peptide match filter ("preferred") were selected for HCD (higher-energy collisional dissociation) fragmentation employing stepped normalized collision energies (29 ± 2). The quadrupole isolation window was set to 1.6 m/z. Minimum AGC target was 2.50 × 10 4 , maximum injection time was 50 ms. Fragment ion scans were recorded with a resolution of 15,000, AGC target set to 1 × 105, scanning with a fixed first mass of 100 m/z. Dynamic exclusion was enabled for 30 s after a single count and included isotopes. Each LC-MS acquisition took 120 min. Raw data from LC-MS runs were searched as described above using MaxQuant (1.6.17.).

The structure of the human 40S initiation complex bound to Nsp1 (PDB ID 6zlw) (10) was used as the starting point for modeling the interactions of full-length Nsp1 to the 40S ribosome using crosslinking-MS restraints. Full-length Nsp1 was modeled using the structure of the N-terminal domain (PDB ID 7k3n) (40) and the C-terminal domain in the Nsp1-bound 43S initiation complex structure (PDB ID 6zlw).

The 43S preinitiation complex structure (PDB ID 6zp4) (10) was adjusted as follows: the RRM domain assigned to "protein X" in the deposited structure was assigned to the RRM domain of eIF3G, according to crosslinking-MS. The domain was built in MODELLER version 9.23 (41) and fitted in the density for "protein X" in the 43S preinitiation complex. The N-terminal stretch of eIF3G (residues 94-140), known to interact with the WD40 domain of eIF3I, was also built in MODELLER and placed in contact with eIF3I by homology to the structure of the yeast eIF3b-CTD/eIF3i/eIF3g-NTD subcomplex (PDB ID 4u1e) (Erzberger et al. 2014), and in accordance with crosslinking data.

Accessible interaction volume analysis was performed with DisVis (42) using PDB ID 6zlw and 6zp4 as the fixed chain, and the structure of the Nsp1 NTD as the scanning chain, with a 1 Å grid spacing and rotational sampling interval of 12.5°. The allowed Cα-Cα distance for restraints was set between 2 and 30 Å.

Integrative modeling: sampling, scoring and model selection

The building blocks described above (40S ribosome, full-length Nsp1) were used to derive an integrative model based on crosslinking-MS, known structures, homology models and physical restraints with the integrative modeling platform (IMP, version 2.15). The overall modeling workflow is described in fig. S1 . For integrative modeling, all protein regions not present in deposited structures, including the Nsp1 linker and unresolved ribosomal protein regions, were coarse grained as beads. Models were coarse grained as two rigid bodies: one comprising the 40S complex, and the Nsp1 CTD (E148-G180) and another one Crosslink-guided assignment of the unidentified RRM domain in the complex, and modeling of the eIF3I-binding region of eIF3G by homology and crosslinking-MS data. 40S and eIF3G-RRM residues crosslinked to Nsp1 NTD and linker regions are shown as red spheres. D) PAIRB binds Nsp1-stalled ribosomes and 40S subunits View of the face of the 40S subunit with residues crosslinked to PAIRB shown as red spheres. Crosslinks show the large, disordered protein wraps around the 40S subunit, with its plug region displaced by the Nsp1 helices. Crosslinks mapped onto PAIRB-bound 80S ribosome (pdb 6z6n) (46) .

Crosslinks also show interactions between Nsp1 NTD and PAIRB.

Deutsche Forschungsgemeinschaft (DFG) under Germany's Excellence Strategy -EXC

Wellcome Trust Senior Research Fellowship (103139). (JR)

Core funding from the Wellcome Trust (203149). (JR)

Max Planck Institute for Molecular Genetics-International Guest Postdoctoral Fellowship

proteomics: FS Sucrose gradients: MLK Modeling/data analysis: AG Supervision: CS,MLK,JR Writing-initial draft: AG,FS,JR Writing-editing and revision

Integrative Imaging Reveals SARS-CoV-2-Induced Reshaping of Subcellular Morphologies

SARS-CoV-2 uses a multipronged strategy to impede host protein synthesis

Global analysis of protein-RNA interactions in SARS-CoV-2-infected cells reveals key regulators of infection

Severe acute respiratory syndrome coronavirus protein nsp1 is a novel eukaryotic translation inhibitor that represses multiple steps of translation initiation

A two-pronged strategy to suppress host protein synthesis by SARS coronavirus Nsp1 protein

Severe acute respiratory syndrome coronavirus nsp1 facilitates efficient propagation in cells through a specific translational shutoff of host mRNA

Severe acute respiratory syndrome coronavirus nsp1 suppresses host gene expression, including that of type I interferon, in infected cells

Coronavirus Nsp1: Immune response suppression and protein expression inhibition

SARS-CoV-2 Nsp1 binds the ribosomal mRNA channel to inhibit translation

Structural basis for translational shutdown and immune evasion by the Nsp1 protein of SARS-CoV-2

SARS coronavirus nsp1 protein induces template-dependent endonucleolytic cleavage of mRNAs: viral mRNAs are resistant to nsp1-induced RNA cleavage

The N-terminal domain of SARS-CoV-2 nsp1 plays key roles in suppression of cellular gene expression and preservation of viral gene expression

SARS-CoV-2 Nsp1 suppresses host but not viral translation through a bipartite mechanism

The viral protein NSP1 acts as a ribosome gatekeeper for shutting down host translation and fostering SARS-CoV-2 translation

Targeted in situ cross-linking mass spectrometry and integrative modeling reveal the architectures of three proteins from SARS-CoV-2

Putting the pieces together: integrative modeling platform software for structure determination of macromolecular assemblies

Coronavirus biology and replication: implications for SARS-CoV-2

Nsp1 protein of SARS-CoV-2 disrupts the mRNA export machinery to inhibit host gene expression

Signaling Pathways for Translation: Stress, Calcium, and

Inhibition of the integrated stress response by viral proteins that block p-eIF2-eIF2B association

Middle East Respiratory Coronavirus Accessory Protein 4a Inhibits PKR-Mediated Antiviral Stress Responses

The mechanism of eukaryotic translation initiation and principles of its regulation

Structure of a human 48S translational initiation complex

The scanning mechanism of eukaryotic translation initiation

Dynamic competition between SARS-CoV-2 NSP1 and mRNA on the human ribosome inhibits translation initiation

A SARS-CoV-2 protein interaction map reveals targets for drug repurposing

Multilevel proteomics reveals host perturbations by SARS-CoV-2 and SARS-CoV

Genomic monitoring of SARS-CoV-2 uncovers an Nsp1 deletion variant that modulates type I interferon response

1H, 13C, and 15N backbone chemical-shift assignments of SARS-CoV-2 non-structural protein 1 (leader protein)

Identification of residues of SARS-CoV nsp1 that differentially affect inhibition of gene expression and antiviral signaling

Emerging of a SARS-CoV-2 viral strain with a deletion in nsp1

SARS-CoV-2 nsp1: Bioinformatics, Potential Structural and Functional Features, and Implications for Drug/Vaccine Designs

The SARS-CoV-2 RNA-protein interactome in infected human cells

A structural inventory of native ribosomal ABCE1-43S pre-initiation complexes

Correction to Optimized Fragmentation Regime for Diazirine Photo-Cross-Linked Peptides

MaxQuant enables high peptide identification rates, individualized p.p.b.-range mass accuracies and proteome-wide protein quantification

An integrated workflow for crosslinking mass spectrometry

The Perseus computational platform for comprehensive analysis of (prote)omics data

Protein Synthesis in the Developing Neocortex at Near-Atomic Resolution Reveals Ebp1-Mediated Neuronal Proteostasis at the 60S Tunnel Exit

Structural characterization of nonstructural protein 1 from SARS-CoV-2. iScience

Comparative Protein Structure Modeling Using MODELLER

DisVis: quantifying and visualizing accessible interaction space of distance-restrained biomolecular complexes

Assessing Exhaustiveness of Stochastic Sampling for Integrative Modeling of Macromolecular Structures

In-cell architecture of an actively transcribing-translating expressome

jPOSTrepo: an international standard data repository for proteomes

Structure and function of yeast Lso2 and human CCDC124 bound to hibernating ribosomes

We thank Prof. B. Glaunsinger and Dr. A. Mendez (Howard Hughes Medical institute, University of Berkley) for critical reading of the manuscript. We thank Prof. R. Beckmann (Ludwig-Maximilians-Universität München) for kindly providing the Nsp1 construct.

The authors declare that they have no competing interests.

Mass spectrometry raw and processed data for both crosslinking-MS and proteomics