Characterization of protease activity of PLpro and Nsp3

SARS-CoV-2 PLpro, PLpro^CoV-2, is a constituent domain of the 217 kDa multidomain membrane protein, Nsp3 (Fig 1A). PLpro is a multifunctional enzyme with three distinct protease activities that catalyse the cleavage of (i) K48-linked polyubiquitin, (ii) ISG15 and (iii) viral polypeptides. Processing of the viral polypeptide involves cleavage between Nsp1-Nsp2, Nsp2-Nsp3 and Nsp3-Nsp4 (Fig 1A) [19]. While recent studies have characterised the substrate specificity and cleavage activity of PLpro^CoV-2 towards polyUb and ISG15 [7, 9, 20, 21], we wondered if elements outside the minimal PLpro domain might play a role in substrate recognition and cleavage. Nsp3 is a multidomain protein containing a transmembrane segment that mediates its localization to membrane vesicles [22, 23]. In addition to the protease domain, Nsp3 contains an N-terminal Ubl (ubiquitin-like domain) followed by an ADRP (ADP-ribose phosphatase, SUD (SARS-unique domain) domain, another Ubl domain that is part of PLpro. C-terminal of the protease is a NAB (nucleic acid binding) domain followed by a TM (transmembrane segment) and a ZnF (zinc finger) domain at the very C-terminus [23, 24]. Despite being a large multidomain protein, we obtained milligram amounts of pure and stable recombinant Nsp3 truncations when we either removed or replaced the transmembrane segment and the acidic stretch at the N-terminus with a linker (Fig 1B). As all the Nsp3 variants were equally active (S1B Fig), Nsp3 spanning residues 179–1329 (henceforth referred to as Nsp3^core) was chosen for further characterization.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Nsp3^core is a more active protease compared to PLpro.

A) Schematic of cleavage of Nsp1-3 by PLpro from the viral polypeptide and the domain structure of Nsp3: Ubl (ubiquitin-like domain), ADRP (ADP-ribose phosphatase, SUD (SARS-unique domain), PLpro (papain-like protease), NAB (nucleic acid binding domain), TM (Transmembrane domain), ZnF (zinc finger motif). B) Coomassie-stained gel of Nsp3 constructs used in this study with accompanying domain schematics (colour coding as in 1A; double slash indicates deletion). C) DUB assay directly comparing cleavage of Ub3 (upper) and ISG15 (lower) by PLpro and Nsp3^core. D) Assay directly comparing cleavage of Nsp1-2Δ by PLpro and Nsp3^core. E) Protease assay comparing the cleavage of Nsp1-2 FL by PLpro and Nsp3^core. F) DUB assay showing the polyubiquitin linkage preference of Nsp3^core. G) Cleavage assay comparing cleavage of Nsp1-2Δ and Nsp1-2Δ G180A mutant by Nsp3^core. H) UV-traces of analytical gel filtration analyses (Superdex 200 3.2/300) of Nsp3^core overlaid onto molecular weight standards. I) Mass photometry analysis measuring the molecular mass of Nsp3^core. Data shown are representative of two independent experiments.

https://doi.org/10.1371/journal.pone.0253364.g001

To understand how PLpro can cleave such structurally distinct substrates and to characterize these three activities further relative to Nsp3, we first generated K48-linked Ub3, pro-ISG15 and, to monitor the cleavage of the viral polypeptide, we expressed and purified full length Nsp1 fused to the first 19 amino acids of Nsp2 (Nsp1-2Δ). We compared the activities of PLpro and Nsp3^core for each of the substrates in qualitative gel-based assays. As previously observed for PLpro^CoV-1, PLpro^CoV-2 is selective at cleaving K48-linked chains (S1A Fig) and shows a similar mode of cleavage by cleaving triubiquitin into diubiquitin and monoubiquitin as final products [7, 9, 18]. In our assay, nearly all triubiquitin is cleaved by Nsp3^core (50 nM) after 15 minutes whereas a significant portion is still to be cleaved by PLpro (Fig 1C, upper). Similarly, Nsp3^core (20 nM) cleaves most of the pro-ISG15 by 15 minutes while only ~50% is cleaved by PLpro (Fig 1C, lower). Intriguingly, Nsp3^core (1 μM) is able to cleave the Nsp1-2Δ substrate, whilst PLpro is not (Fig 1D). To confirm this observation, we incubated increasing concentrations of PLpro with Nsp1-2Δ, which remained resistant to cleavage (S1C Fig). We therefore conclude that elements outside the minimal PLpro domain are responsible for recognising the viral polypeptide for cleavage. The Nsp1-2Δ substrate tested so far only contains 19 amino acids from Nsp2. To test if a more physiological form of the viral polypeptide could be cleaved by PLpro, we expressed and purified full-length Nsp1-Nsp2 fusions. We again find that PLpro is unable to cleave Nsp1-Nsp2 FL substrate, whereas Nsp3^core readily cleaves it (Fig 1E).

These qualitative assays suggest that Nsp3 is a more active enzyme. To confirm these observations, we used a MALDI TOF based DUB assay (Ritorto et al., 2014), which shows that PLpro efficiently cleaves K48 trimers with a K_m of 29.39 μM, a V_max of 0.66 mM*min^-1 and a k_cat of 22 min^-1 (S1G Fig). In contrast, Nsp3 cleaves K48-linked triubiquitin with a K_m of 17.79 μM, a V_max of 0.62 mM*min^-1 and a k_cat of 20.6 min^-1. The comparatively lower K_m of Nsp3 potentially explains the higher cleavage activity observed. Even though Nsp3 is more active than PLpro, it still retains its specificity for cleaving K48-linked polyubiquitin and does not cleave any of the other polyubiquitin linkages tested (Fig 1F). The C-termini of all PLpro substrates contain a -GG motif, with PLpro cleaving after the last Gly [20] (S1D Fig). Indeed, mutating the C-terminal G180 of Nsp1 renders it resistant to cleavage by Nsp3 as does mutation of the catalytic C856 of Nsp3 to Ala (Fig 1G, S1E Fig), confirming that the extended Nsp3 maintains substrate specificity and does not contain additional protease catalytic sites. We then generated propargylated Nsp1 (Nsp1^Prg), a reactive probe which would covalently modify the catalytic Cys of PLpro and Nsp3 [25]. Indeed, while Nsp3^core was efficiently modified by Nsp1^Prg, barely any conversion was observed for PLpro (S1F Fig).

Since Nsp3^core efficiently cleaves the viral polypeptide fusion Nsp1-Nsp2, which minimal PLpro does not cleave, and Nsp1^core is readily modified by Nsp1^Prg leads us to speculate that regions outside of the PLpro domain may mediate additional interactions with Nsp1 and Nsp2 to facilitate cleavage. As Nsp3 is a multidomain protein, it is difficult to predict which of these may play a role in binding to the viral polypeptide. The SUD domain of SARS-CoV-2 shares 75% sequence identity with that of SARS-CoV-1 SUD and consists of various macrodomains (Mac1, Mac2, Mac3 and DPUP) which bind to nucleotides and to proteins such as the E3 ligase RCHY1 and polyA binding protein 2 [23, 24, 26–29]. In addition to these roles, it is possible that the SUD domain may mediate additional interactions with Nsp1 and Nsp2 to promote viral polypeptide cleavage.

Despite several attempts, we were unable to crystallise longer variants of Nsp3 on its own or in complex with a substrate. Intriguingly, Nsp3 elutes at an apparent molecular weight of ~520kDa on a size-exclusion chromatography column raising the possibility that it could be tetramer (Fig 1J). However, mass photometry analysis reveals Nsp3 to be a monomer (Fig 1K), suggesting that the behaviour on gel filtration could be attributed to an extended conformation or coiled-coil content in Nsp3. Obtaining a structure of Nsp3 in complex with Nsp1-Nsp2 will reveal why it is more active compared to PLpro and the interactions mediated with Nsp1 and Nsp2.

Interactome of PLpro and Nsp3

Since we observe differences in protease activity between Nsp3^core and PLpro, we next wanted to compare the interactomes of PLpro and Nsp3. To date most efforts to comprehensively define the interactomes of SARS-CoV-2 non-structural proteins have failed to identify the interactome of Nsp3 [30]. When we expressed full length Nsp3 in HEK 293 cells, we observed very low expression levels that made it challenging to get enough material for affinity enrichment/mass spectrometry (AE/MS) analysis. Hence, we made a truncation where we deleted the transmembrane segments of Nsp3 (Nsp3ΔTM). The lung epithelial cell line A549 was transfected with either Strep-tagged Nsp3ΔTM or PLpro and cells were lysed following IFN-α stimulation for 36 hours.

Quantitative mass spectrometry analysis revealed that Nsp3 ΔTM and PLpro had a largely overlapping interactome (Fig 2). Interestingly, many of the interactors enriched for both PLpro and NSP3 DTM are interferon stimulated and proteins involved in antiviral signalling. With its preference for cleaving ISG15, we observed ISG15 as a significant interactor that was also identified in the PLpro^CoV-2 interactome [9]. Other prominent overlapping interactors include the IFITs (Interferon-induced protein with tetratricopeptide repeats), IFIT1 and IFIT2, antiviral proteins that function as sensors of viral single-stranded RNAs to inhibit expression of viral messenger RNAs [31]. Of note, IFIT1 (also known as p56 or ISG56) has previously been recognised as a target of ISGylation [32]. Nsp3 ΔTM also interacts with the antiviral proteins SAMHD1, which binds to nucleotides and depletes cellular dNTPs, and 1’-5’-oligoadenylate synthase 3 (OAS3), another interferon-induced gene that inhibits viral replication by degrading viral RNA [33]. In addition, both PLpro and Nsp3 ΔTM interact with the multifunctional ATP-dependent RNA helicase DDX3X, which promotes IRF3 activation leading to the production of the type I interferons, IFN-α and IFN-β [34].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Comparison of Nsp3 and PLpro interactome.

Volcano plot showing significant interactors identified by pull down on Nsp3 DTM or PLpro from transiently transfected cells followed by quantitative mass spectrometry. Interactors enriched in PLpro pull downs are highlighted in blue and those enriched in Nsp3 pull downs are highlighted in red circles.

https://doi.org/10.1371/journal.pone.0253364.g002

Our data also identifies interactors linked to translation (EIF5A2, EIF2S1, ICE1, EIF2S3L) and the actin cytoskeleton (Cofilin-1, tubulins, KIF9, MYL4, PDLIM5), which was also observed in a recent study that used proximity labelling [35]. While most interactors are largely overlapping between Nsp3 ΔTM and PLpro, we identify the RNA binding protein Fragile X mental retardation syndrome-related protein 1 (FXR1) to be enriched only in Nsp3 ΔTM pull downs. Interestingly, Nsp3 from Sindbis virus interacts with FXR1 to bind viral RNAs for the assembly of viral replication complexes [36].

Despite having several additional domains, no additional unique interactors were identified for Nsp3 ΔTM. We attribute the lack of additional interactions to Nsp3 ΔTM to the following possibilities: (i) the Nsp3 used in our experiments lacks the transmembrane segment and hence does not localize to the native subcellular compartment where the relevant cellular interactors of Nsp3 may be present; (ii) the majority of the interactions with Nsp3, i.e., the Ubl1 domain, the SUD domain and macrodomain are with RNA or ADP-ribose, and (iii) Nsp3 has been suggested to function as a scaffold mediating binding to other Nsps to form replicative organelles explaining the lack of interactions with human proteins. Indeed, several interactions of Nsp3 with other viral Nsps have been identified by Y2H, but the functional implications of these interactions to the viral life cycle are unclear [22, 37, 38].

Identification and characterization of inhibitors to PLpro

We employed the previously developed MALDI-TOF DUB assay [39] to evaluate the ability of 1971 FDA approved drugs to inhibit PLpro from SARS-CoV-2. Importantly the bioactivity and safety of these compounds has been tested in clinical trials allowing rapid repurposing of any hit compounds into therapies for the treatment of COVID-19. Compared to other fluorescent-based approaches, the MALDI-TOF-based assay has the advantage of using K48 ubiquitin trimer, a physiological PLpro substrate, for high-throughput screening (HTS) [40]. Further, the final products generated upon cleavage of triubiquitin are diubiquitin and monoubiquitin, which makes it compatible for analysis using the MALDI DUB assay. The reaction is terminated by the addition of 2% TFA and ¹⁵N ubiquitin. To circumvent the spot-to-spot variability typical for MALDI-TOF based detection, the mono-ubiquitin signal is normalized to the ¹⁵N internal standard signal (Fig 3A). The FDA-approved drug library was screened in duplicate at a concentration of 10 μM. The Z’ value is commonly employed to determine the robustness of HTS, with Z’ values greater than 0.5 indicating a robust assay [41]. The MALDI-TOF DUB assay resulted in an average Z’ value of 0.77 ± 0.12 (Fig 3B). Compounds returning percentage inhibition values >50% were identified as hits and 30 compounds met this criterion (1.5% hit rate) (Fig 3B). The hit compounds were cherry picked and retested in a 10-point dose–response to determine IC₅₀ values against both PLpro and Nsp3. We used GRL-0617, a non-covalent inhibitor of both PLpro^CoV-1 [42], and PLpro^CoV-2 [9, 20, 43] as a control. GRL-0617 had been previously shown to inhibit PLpro^CoV-1 with an IC₅₀ of 0.6 μM [42]. We found that GRL-0617 inhibits SARS-CoV-2 PLpro with an IC₅₀ of 2.21 μM, which concurs nicely with the IC₅₀ of ~2 μM determined by recent studies [20, 43]. GRL-0617 inhibits Nsp3^core with an IC₅₀ of 3.39 μM (Fig 3F). The MALDI-TOF-based results were further validated by orthogonal, gel-based assays employing K48-linked triUb and pro-ISG15 as substrate (Fig 3D and 3E; S2A–S2C Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Identification and characterization of PLpro inhibitors.

A) Schematic workflow of the MALDI-TOF DUB assay. K48 trimer is incubated with PLpro for 30 minutes at room temperature. The reaction is stopped by adding 2% TFA and ¹⁵N ubiquitin (as internal standard). Enzymatic activity and inhibition were assessed by MALDI-TOF MS detection of ubiquitin and ¹⁵N ubiquitin signals. B) FDA approved compound library screening by MALDI-TOF, Z’Prime score and HTS data distribution. C) IC50 calculation of compounds returning >50% inhibition values in the HTS: thioguanine, nordihydroguaiaretic acid, Disulfiram, Auranofin and Tideglusib. IC50s were calculated both against PLpro and Nsp3^core using the MALDI-TOF DUB Assay. D-E) Orthogonal, gel-based assays testing PLpro and Nsp3 protease activities against K48 trimer and pro-ISG15 as substrates. Nordihydroguaiaretic acid (NDA), Disulfiram (DSF), Auranofin (AUR) and Tideglusib (TID) were tested at the indicated concentrations against PLpro and Nsp3^core. Data shown are representative of two independent experiments. F) IC50 calculation of GRL0617 against PLpro and Nsp3^core. G) Selectivity assessment of GRL-0617 and HTS positive hits against a panel of 42 human DUBs and the viral ovarian tumor (vOTU). Compounds were tested at the final concentrations of 10 and 1 μM.

https://doi.org/10.1371/journal.pone.0253364.g003

Based on these analyses, five compounds were confirmed for their ability to inhibit PLpro and Nsp3 in vitro (confirmation rate = 16.6%): Thioguanine, Nordihydroguaiaretic acid, Disulfiram, Auranofin and Tideglusib (Fig 3C). Thioguanine inhibits PLpro and Nsp3 with IC₅₀ of 4.54 and 8.67 μM respectively, in the MALDI-TOF-based assay. Intriguingly, despite this reasonable IC₅₀, we could not observe inhibition of PLpro by Thioguanine in the orthogonal gel-based assay Interestingly, thioguanine was previously found to inhibit PLpro^SARS-CoV-1 [44] and SARS-CoV-2 replication [45].

In our assays, we identify nordihydroguaiaretic acid (NDA) to be a potent inhibitor of PLpro and Nsp3, with IC₅₀ of 1.06 and 1.62 μM in the MALDI-TOF based assay. NDA is a compound extracted from plants, used in traditional medicine for the treatment of various diseases, and found to inhibit numerous cellular targets [46]. Its ability to inhibit a range of cysteine-based enzymes is linked to the redox properties of its two catechol rings, resulting in the formation of cysteine adducts [46]. Despite its beneficial effects, the main drawback to the use of this compound is its potential renal and liver toxicity over prolonged use.

Disulfiram is a thiocarbamate drug (commercially available as Antabuse®) that has received FDA approval for the treatment of alcoholism and has been in use for more than 60 years [47]. Disulfiram has been previously identified by in silico screening and fluorescence resonance energy transfer assay to inhibit the main SARS-CoV-2 Cys-protease, Mpro [48, 49]. Disulfiram also inhibits PLpro^CoV2 and Nsp3 with similar potencies, 0.69 and 0.66 μM in the MALDI-TOF-based assay, and between 1 and 2.5 μM when assessed by gel-based orthogonal assay (Fig 3D and 3E). Tideglusib, a Glycogen Synthase Kinase 3 (GSK3) inhibitor, has also been found to inhibit the main SARS-CoV-2 Cys-protease Mpro [48]. We found that Tideglusib strongly inhibits both PLpro and Nsp3 with IC50 of 0.21 and 1.48 μM, respectively in the MALDI-TOF based assay. Similar potency of Tidegusib was also validated in the gel-based assay (Fig 3D and 3E). Auranofin is a gold-containing compound employed for the treatment of rheumatoid arthritis [50], known to suppress inflammation and stimulating cell-mediated immunity. Of note, Auranofin has been recently shown to inhibit replication of SARS-CoV-2 in human cells [51]. Interestingly, we found that Auranofin strongly inhibits PLpro and Nsp3 with potency in the 1 μM range in the MALDI-TOF based assay with a similar inhibition pattern observed in the gel-based assay (Fig 3D and 3E).

The human genome encodes about 100 deubiquitinating enzymes, most of which are cysteine proteases [52]. Since PLpro has architectural features in common with the USP DUBs, we assessed the selectivity of GRL-0617 and the five HTS positive hits by testing them at the final concentrations of 10 and 1 μM against a DUBs panel including 42 human DUBs and the viral ovarian tumor (vOTU) from Crimean Congo Haemorrhagic Virus. Unsurprisingly, we found that Disulfiram, Thioguanine, NDA, Tideglusib and Auranofin inhibited–at different extents–several other human DUBs (Fig 3G), while GRL-0617 showed greater specificity with less off-target inhibition in the panel of DUBs tested (Fig 3G). Interestingly, GRL-0617 was also able to inhibit cleavage of the viral polypeptide (S2D Fig) The presence of highly reactive chemical groups able to covalently modify reactive cysteines might explain the poor specificity of these small molecules. Despite the broad range of activity of these compounds, their relative safety has been extensively evaluated in FDA pre-clinical and clinical trials.

We next tested the efficacy of these compounds to inhibit PLpro in cell-based assays measuring activation of interferon expression using an IFN-β luciferase reporter or the ability to inhibit replication of SARS-CoV-2 in CaCo cells [9]. However, none of the compounds other than GRL-0617 showed any effects of inhibiting PLpro in cells (data not shown). These results contrast with what was shown for some of these inhibitors. For instance, the use of Auranofin and thioguanine has been shown to limit SARS-CoV-2 replication [45, 51] but the underlying mechanism of action is not entirely understood. Our data demonstrate that Auranofin and thioguanine inhibit the deubiquitinating activity of the PLpro protease in vitro. The low potency of the identified hits at limiting viral replication suggests that repurposing may not be an effective strategy to identify inhibitors suitable for clinical development and a systematic approach is required to develop specific and effective inhibitors of PLpro/Nsp3. Our results, showing that extended versions of Nsp3 can cleave viral polypeptides, which the minimal PLpro is not as efficient at suggest that using Nsp3 to screen for inhibitors could lead to the identification of allosteric inhibitors that may inhibit all three activities of Nsp3. Despite not being very potent on their own, combining drugs targeting essential SARS-CoV-2 proteases (PLpro and/or main protease) with other essential viral enzymes such as RNA-dependent polymerase may represent a therapeutic strategy.

Development of nanobodies that inhibit PLpro

Since all the identified inhibitors identified show off-target effects and inhibit other human DUBs, we wanted to devise other ways to selectively inhibit PLpro as a strategy to investigate the roles of Nsp3 in the viral life cycle, in interfering with host immune signalling and to examine if inhibition of PLpro can effectively block viral replication. Previously, ubiquitin variants (Ubv) have been developed to selectively inhibit DUBs by binding to the S1 pocket [53, 54]. Hence, we aimed to engineer a nanobody (a single-domain antibody) that binds to the S1 site with high affinity to inhibit PLpro. The S1 site is important for both deubiquitinating and deISGylating activities of PLpro [17]. We therefore used a recently developed yeast surface nanobody display library to screen and identify nanobodies that specifically bind to the S1 site of PLpro [55]. To steer nanobody selection towards the S1 site, we had to discard nanobodies that recognized other regions of PLpro. Hence, we introduced a negative selection step using PLpro^CoV-2 carrying mutations in the S1 site. As these mutations were designed based on the available structure of PLpro^SARS in complex with ubiquitin (PDB 4MM3, [8]) (S3A Fig), we first tested if these mutations also affected PLpro^CoV-2. Indeed, compared to WT, the mutant PLpro^CoV-2 was unreactive to propargylated Ub and was unable to cleave K48-linked triUb (S3B and S3C Fig). Following multiple rounds of negative and positive selection using magnetic-activated cell sorting (MACS), as outlined in Fig 4A, single yeast clones were tested for their binding to PLpro (Fig 4B). We identified several nanobodies that bound to WT PLpro^CoV-2 but not the S1 site mutant, suggesting that these nanobodies specifically bind to the S1 site (NbSL-17, -18, 19). Sequencing these nanobodies revealed differences in the CDR1, CDR2 and CDR3 loops of the different nanobodies.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Development of nanobodies that inhibit Nsp3^core.

A) Schematic workflow of yeast-surface display selection to identify site-specific nanobodies as competitive Nsp3^core inhibitors. B) Flow cytometry analysis of yeast displaying anti-Nsp3 nanobodies NbSL17, NbSL18 or NbSL19, respectively. Yeast were stained with anti-HA-Alexa488 against HA-tagged Nb and Streptavidin-Alexa647 against biotinylated antigens. Top: Yeast incubated with PLpro wild-type, Bottom: Yeast incubated with PLpro S1*-mutant. Percentages indicate fraction of yeast positive for both fluorescent labels. C) UV-traces and coomassie-stained SDS-PAGE of analytical gel filtration analyses of PLpro-Nb complexes. PLpro alone (black), PLpro+NbSL17 (orange) and PLpro+NbSL18 (blue). D) Isothermal calorimetry titration measuring binding of PLpro to NbSL18. E) Silver stained gels of DUB assays of PLpro (left) and Nsp3^core (right) with K48-Ub3 (top) and proISG15 (middle) and Nsp1-2 FL (bottom) as substrates in the presence of increasing concentrations of NbSL18. Silver-stained SDS-PAGE gels shown are representative of two independent experiments. Also see S3 Fig.

https://doi.org/10.1371/journal.pone.0253364.g004

To further characterize these nanobodies, they were recombinantly expressed and purified from bacteria. Of the three individual nanobody sequences we identified, soluble expression was obtained for only two nanobodies (NbSL17 and NbSL18). We first performed analytical gel filtration, which identified NbSL18 to form a complex with PLpro^CoV-2 (Fig 4C). Using isothermal titration calorimetry (ITC), we find the binding affinity of NbSL18 for PLpro^CoV-2 to be ~500 nM (Fig 4D). We next evaluated the ability of NbSL18 to inhibit PLpro. With increasing concentration of NbSL18, the cleavage of K48-linked triUb by PLpro is inhibited with ~50% inhibition at ~500nM-1000nM NbSL18, which is consistent with the binding affinity of NbSL18 for PLpro (Fig 4E). In contrast, higher concentrations of NbSL18 are required to inhibit the cleavage of ISG15. Recent crystal structures of ISG15 in complex with PLpro^CoV-2 show that the C-terminal UBL of ISG15 uses a slightly different mode of interaction compared to ubiquitin [7, 9], potentially explaining the differences in the inhibition of deubiquitylating and deISGylating activities by NbSL18.

Understanding how the viral polypeptide is recognised by Nsp3 and building on this to inhibit release of mature Nsp1/2/3, could represent a legitimate strategy to suppress viral replication. Nsp1, for instance, plays a key role in the infection process by binding to host 40S ribosomal subunits and blocking the mRNA channel with its C-terminal domain, thereby suppressing translation of host proteins and sequestering ribosomes for translation of viral proteins. Preventing Nsp1 maturation from the viral polypeptide could therefore be vital in hindering viral replication [56, 57]. We therefore tested if NbSL18 could also inhibit cleavage of Nsp1-Nsp2 FL fusions. Our data reveals that NbSL18 can indeed inhibit Nsp1-2 FL cleavage by Nsp3 suggesting that the viral peptide also occupies the same S1 site as ubiquitin (Fig 4H).

Determining a crystal structure of NbSL18 in complex with Nsp3 or PLpro would shed light into the mode of interaction and allow us to engineer mutations to not only increase the binding affinity but also selectively inhibit deubiquitylating, deISGylating or viral polypeptide processing activities. Such nanobodies will be useful to dissect the relative contributions of the different activities of PLpro [58]. However, despite several attempts we were unable to obtain well-diffracting crystals. Nevertheless, the development of these highly specific nanobodies that target PLpro represent valuable tools that could be expressed in SARS-CoV-2 infected cells to directly inhibit PLpro and thereby dissect the contributions of PLpro to viral replication and infection as shown using ubiquitin variants [59]. Further the nanobodies can be used as sensors to monitor subcellular localization of Nsp3 at different stages of viral infection. Should the nanobodies exert significant antiviral effects, they could be used as protein-based inhibitors to protect against SARS-CoV-2 and other related viruses. In summary, our work builds on extensive work characterising the PLpro domain and reveals differences in protease activities between the minimal PLpro domain of Nsp3 and longer versions of Nsp3 and highlights the need to better understand the mechanism of Nsp3. Further, recently published studies suggest that present efforts to develop inhibitors are focussed solely on PLpro [60]. Our findings advocate that it would be prudent to develop inhibitors towards Nsp3 and not just PLpro and minimally to evaluate the efficacy of compounds identified for PLpro on the ability of full length Nsp3 to cleave viral polypeptides.

Supplemental information

Supplemental information includes S1–S3 Figs and S1 and S2 Tables.

cDNA constructs

Human ISG15 cDNA was amplified by RT-PCR from universal human reference RNA (Agilent 740000–41), adding BamHI and NotI restriction sites at the 5’ and 3’ ends respectively. The digested PCR product was cloned into pGEX6P1 cut with the same enzymes to make pGEX6P1 ISG15 (DU8371).

Ubiquitin cDNA was amplified from existing clone DU8724 (pET15b 6His 3C Ubiquitin), adding restriction sites NdeI and BamHI at the 5’ and 3’ ends, respectively. The digested PCR product was cloned into pET24a cut with the same enzymes to make pET24a Ubiquitin (DU20027).

The cDNAs coding for SARS CoV-2 (MN908947.3) Nsp3 residues 743–1072 (PL-pro) and SARS CoV-2 Nsp3 residues 743–1072 (PL-pro) E912R E948R M953E Y1013K were synthesised by NBS Biologicals (www.nbsbio.co.uk) with 5’ BamHI and 3’ NotI sites. Sequences were codon optimised for bacterial expression. These were subcloned into receptor vector pETDuet 6His TEV GST 3C AVi (DU68478) digested with BamHI and NotI to make pETDuet 6His TEV GST 3C AVi SARS CoV-2 Nsp3 PL-pro (DU67765) and pETDuet 6His TEV GST 3C AVi SARS CoV-2 Nsp3 PL-pro E912R E948R M953E Y1013K (DU67766).

pETDUET 6HIS TEV GST 3C SARS-CoV-2 Nsp3 743–1072 (PL-pro) (DU67811) was made by removing the avi tag sequence from DU67765 by PCR mutagenesis.

The cDNA coding for SARS CoV-2 Nsp1 1–180 Nsp2 1–19 8His was synthesised by GeneArt (Thermo Fisher Scientific) adding NcoI and NotI sites at the 5’ and 3’ ends, respectively. This sequence was subcloned into pET28a to make pET28a SARS-CoV-2 Nsp1 1–180 Nsp2 1–19 8His (DU67799).

pTXB1 Halo 3C SARS-CoV-2 (2019-nCoV) Nsp1 1-179-Intein CBD (DU67780 was made using NEBuilder (New England Biolabs), amplifying the vector from existing clone DU28033 (pTXB1-HALO-Mxe Intein-CBD) and the Nsp1 1–179 insert from existing clone DU66413 (pGEX6P1 2019-nCoV Nsp1).

pET15D Twinstrep 3C Nsp3 N179-N1329 6His (DU67831) was made using NEBuilder, amplifying the vector from existing clone DU67810 (pET15D Twinstrep 3C 6His) and the Nsp3 179–1329 insert sequence from Addgene plasmid 141257 pDONR207 SARS-CoV-2 Nsp3.

pET15D Twinstrep 3C Nsp3 1–109 Linker 179–1329 6His (DU67844) was made using NEBuilder, amplifying the vector from DU67831 and the Nsp3 1–109 with linker sequence from Addgene plasmid 141257 pDONR207 SARS-CoV-2 Nsp3.

pET15D Twinstrep 3C Nsp3 N179-N1329 Linker 1584–1944 6His (DU67847) was made using NEBuilder, amplifying the vector from pET15D Twinstrep 3C Nsp3 179–1329 6His (DU67831) and the Linker + Nsp3 1584–1944 sequence from Addgene plasmid 141257 pDONR207 SARS-CoV-2 Nsp3.

pET15D Twinstrep 3C Nsp3 A1-N1329 Linker 1584–1944 6His (DU67881) was made using NEBuilder, amplifying the vector from existing clone pET15D Twinstrep 3C-6His (DU67810) and the Nsp3 A1-N1329 Linker 1584–1944 insert from existing clone pcDNA5D FRT/TO Twinstrep 3C Nsp3 A1-N1329 Linker 1584–1944 6His IRES HA Nsp1 M1-G180 (DU67868).

pET15D Twinstrep 3C SARS-CoV-2 Nsp3 N179-N1329 6His C856A (DU67898) was made by adding the C856A mutation to clone DU67831 using PCR mutagenesis.

pET15D 8His GST 3C Nsp1-Nsp2 (DU70079) was made by reamplifying the Nsp1-Nsp2 sequence from existing clone pcDNA5 FRT/TO Sars CoV-2 Nsp1-Nsp2-Nsp3 A1-G1944-GFP (DU70082) (generated by NEBuilder), directly digesting with BamHI and NotI and cloning into existing plasmid pET15D 8His GST (DU70054) cut with the same enzymes.

pcDNA5D FRT/TO mCherry (DU41458) was made by amplifying the mCherry sequence plus mcs and cloning into the BspTI and NotI sites of pcDNA5D FRT/TO (DU41459), a derivative of pcDNA5 FRT/TO (Thermo Fisher).

pcDNA5D FRT/TO Twinstrep 3C Nsp3 A1-N1329 Linker 1584–1944 6His (DU67879) was made by reamplifying Twinstrep 3C Nsp3 A1-N1329 Linker 1584–1944 6His from pcDNA5D FRT TO Twinstrep 3C Nsp3 A1-N1329 Linker 1584–1944 6His IRES HA Nsp1 M1-G180 (DU67868) (made by NEBuilder) and cloning into pcDNA5D FRT/TO (DU41459) BglII/NotI into BamHI/NotI.

pcDNA5D FRT/TO Twinstrep 3C SARS CoV-2 Nsp3 743–1072 (PL-pro) (DU67916) was made by removing the HA and replacing with Twinstrep 3C using Q5 site-directed mutagenesis (New England Biolabs E0552S) in pcDNA5D FRT/TO HA SARS-CoV-2 Nsp3 743–1072 (PL-pro) (DU67767). pcDNA5D FRT/TO HA SARS-CoV-2 Nsp3 743–1072 (PL-pro) (DU67767) was made by cloning the synthesised SARS-CoV-2 Nsp3 743–1072 (PL-pro) sequence into pcDNA5D FRT/TO HA (DU41456) using BamHI/NotI.

All PCR reactions were carried out using KOD Hot Start DNA polymerase (Novagen).

All RT-PCRs were carried out using Takara’s PrimeScript High Fidelity RT-PCR kit (R022A).

DNA sequencing was performed by MRC PPU DNA Sequencing and Services, School of Life Sciences, University of Dundee (www.dnaseq.co.uk).

See S1 Table for summary of constructs used in this study. All constructs described here can be obtained from https://mrcppu-covid.bio/

Protein expression and purification

All recombinant proteins were expressed in E. coli BL21-CodonPlus (DE3) cells. Bacterial cultures were grown in 2x TY media supplemented with 100 μg/ml ampicillin to an OD₆₀₀ of 0.6–0.8 at 37˚C in shaking incubators. Protein expression was induced by the addition of 300 μM IPTG followed by overnight incubation at 18˚C. Cells were harvested by centrifugation at 3500 g for 30 mins and pellets were suspended in lysis buffer (50 mM Tris-HCl pH 7.5, 300 mM NaCl, 10% glycerol, 0.075% v/v BME, 1 mM benzamidine, 1 mM AEBSF and a complete protease inhibitor cocktail tablet (Roche). Resuspended cells were then lysed by sonication and clarified by centrifugation at 35,000g for 45 min at 4˚C.

Qualitative protease assays

PLpro and Nsp3^core were diluted in buffer A (50 mM Tris-HCl pH 7.5, 50 mM NaCl, 10 mM DTT) and incubated at RT for 10 minutes to fully reduce the catalytic cysteine. For most protease cleavage assays, 3 μM substrate was incubated with varying concentrations of PLpro and Nsp3^core. For assays comparing the relative activities of each protease against a panel of substrates and assessing the linkage specificity, 50 nM of Plpro/Nsp3 was used. For assays comparing the relative activities of PLpro and Nsp3 with Nsp1-2 FL, Ub3, proISG15 and Nsp1-2Δ the respective enzyme concentrations were 50 nM (Nsp1-2 FL, Ub3), 20 nM (proISG15) and 1 μΜ (Nsp1-2Δ). For the assay comparing the cleavage of Nsp1-2(Δ) G180A relative to WT by Nsp3^core and the assay comparing cleavage of Nsp1-2Δ by all Nsp3 constructs, 5 μM protease was incubated with 10 μM substrate. For the assay in which increasing concentrations of PLpro were incubated with Nsp1-2Δ, 10 μM substrate was used. All assays were carried out at 30˚C and quenched at indicated time points by adding LDS loading dye. The samples were separated on 4–12% SDS-PAGE gel (Life Technology) and silver stained using Pierce Silver stain kit (Thermo Fisher).

Preparation of Nsp3-Nsp1^prg complex

HALO-Nsp1 thioester was generated following intein cleavage as described previously for ubiquitin [62]. The HALO-Nsp1-thioester was buffer exchanged into HEPES buffer (pH 8.0) and reacted with 250 mM propargylamine (Sigma) for 3 hours at RT. The reactive probe was then separated from the excess propargylamine by size-exclusion chromatography (Superdex 200 16/60, Cytiva) in 1x PBS. To test reactivity of the probe (S1B), 5 μM Nsp3^core/PLpro was incubated with 10 μM of HALO-Nsp1^prg at 30˚C for 1 hr. The reaction was quenched with 4x LDS dye, resolved on a 4–12% SDS-PAGE gel (Life Technology) and stained with InstantBlue (Expedeon).

Small molecule/nanobody inhibition assays

PLpro and Nsp3^core were diluted in buffer I (50 mM Tris-HCl pH 7.5, 50 mM NaCl, 0.01% (w/v) BSA). The DUB was pre-incubated with the indicated concentration of small molecule inhibitor/nanobody for 30 min at RT. A final concentration of 50 nM DUB was incubated with 3 μΜ substrate at 30˚C for 1 hour. The assay was quenched at indicated time points by adding LDS loading dye. The samples were separated on 4–12% SDS-PAGE gel (Life Technology) and silver stained using Pierce Silver stain kit (Thermo Fisher).

Nanobody selection

ITC measurement

ITC measurements were performed on a MicroCal PEAQ-ITC instrument (Malvern) at 25°C. All proteins were dialyzed overnight at 4°C into 50 mM Tris pH 7.5, 150 mM NaCl and 0.2 mM TCEP prior to ITC analysis. NbSL18 at 148 μM was loaded in the syringe and titrated in 16 2.5 μl injections to 14.4 μM PLpro in the cell. ITC data was fitted with a one-sided binding model (MicroCal Analysis Software, Malvern) to calculate binding constants.

Transfections, lysis and co-immunoprecipitations

A549 human alveolar epithelial cells were transfected with empty vector in triplicate or with vectors expressing SARS CoV-2 Nsp3ΔTM or PLpro fused to a 2xStrep tag at the N-terminus (4 replicates each). 12hr later, cells were stimulated with 50 ng/ml IFN-a (Cell Signalling Technologies) for 36hr. Cells were lysed in ice-cold IP buffer (50 mM Tris-HCl, pH7.5, 150 mM NaCl, 0.5% NP-40, 1 mM AEBSF) and clarified by centrifugation. Viral proteins were affinity purified from 4 mg lysates by tumbling with 40 μl Strep-Tactin Sepharose (IBA Life Sciences) for 90 min at 4°C. Beads were washed two times with IP buffer containing 500 mM NaCl and 0.5% NP-40 and twice more with IP buffer without detergent. 90% of the final wash was transferred to a fresh LoBind tube (Eppendorf) for processing for LC-MS/MS analysis. The remaining 10% was reserved to confirm viral protein expression and pulldown by Western blot for Nsp3. Efficient IFN-a stimulation was confirmed by blotting lysates for ISG15.

On-bead digestion and TMT labeling

The beads from HA-Nsp3, HA-PLpro and control immunoprecipitations were resuspended in 100 μl of 100 mM Tris-Cl (pH 8.0). The proteins were reduced by adding final 5mM DTT and incubated on a Thermomixer with a gentle agitation at 56°C for 20 minutes. Following, the samples were brought to room temperature and alkylated by adding final 20 mM Iodoacetamide and incubated in dark at room temperature on a Thermomixer with a gentle agitation for 30 minutes. Samples were subjected to on-bead tryptic digestion by adding a digestion buffer (2 M Urea + 500 ng of sequencing grade trypsin per sample) and incubated at 30°C temperature on a Thermomixer for about 16hrs. Post digestion, samples were centrifuged (At room temperature at 2000 g for 3 minutes) and the supernatant was collected into a new 1.5ml collection tubes and acidified by adding final 1% TFA. Samples were then subjected SDB-RP peptide clean up using in-house prepared SDB-RP stage-tips. Briefly, two 16guage SDB-RP disks were packed into a 250 ul of pipette tips. The stage-tips were activated by adding 150ul of loading buffer (1%TFA in Isopropanol vol/vol) and centrifuged at 2,500 g at RT for 10 minutes. Subsequently, the acidified peptide digest was loaded and centrifuged at 2,500 g at RT for 10 minutes. The flow-through was reapplied and centrifuged. The stage-tips were washed once with loading buffer and washed again with wash buffer (3% ACN in 0.3% TFA vol/vol). Finally, samples were eluted by adding 60 ul elution buffer I (50% ACN in 1.25% NH4OH vol/vol) and eluted again by adding 60ul of elution buffer II (80% ACN in 1.25& NH4OH). The eluates were immediately snap frozen and vacuum dried in a speed vac.

Mass spectrometry data analysis and Bioinformatics analysis

The MaxQuant software suite [65] version 1.6.10.0 was used for database search with the following parameter. Reporter ion MS3 type: 9plex TMT at a reporter ion mass tolerance of 0.003 Da. The TMT isotopic reporter ion correction factors were manually entered as provided by the manufacturer. The following group specific parameters were used, A built-in Andromeda search engine was used by specifying Trypsin/P as a specific enzyme by allowing 2 missed cleavages, minimum length of 7 amino acids, Oxidation of (M), Acetyl (Protein-N-terminal), Deamidation N and Q were selected as variable modifications. Carbamidomethylation Cys was selected as fixed modification. First search tolerance of 20 ppm and main search tolerance of 4.5 ppm were selected. Global Parameters: Uniprot Human protein database appended with SARS-COV-2 NSP3 with twin strep-tag (release 2017–02; 42,101 sequences) was used for the database search and 1% peptide and protein level FDR was applied. For protein quantification, min ratio count was set at 2 for accurate reporter ion quantification. The MaxQuant output protein group text files were processed using Perseus software suite [65], version 1.6.10.45 was used. The data was filtered for any proteins that are flagged as common contaminants and reverse hits. The TMT reporter ion intensities were log2 transformed and subsequently the TMT reporter tags of PLpro, NSP3 and control conditions were categorized to perform statistical analysis. Two sample Welch’s T-test was performed by applying 5% permutation-based FDR to identify the differentially enriched and significant protein groups between HA-PLPro vs control and HA-NSP3 vs control groups. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE [66] partner repository with the dataset identifier PXD022904.

MALDI TOF assays

MTP Anchor Chip 1536 BC target plates (# 8280787) and MTP target frame (# 8074115) were purchased from Bruker Daltonics (Billerica, MA). 2’,5’- Dihydroxyacetophenone matrix (DHAP matrix) was purchased from Tokio Chemical Industries (Product Number: D1955). LC-MS grade solvents and reagents were purchased from different commercial vendors: Acetonitrile (Merck-Millipore—Cat. No: 1.00030.2500), Ammonium Citrate dibasic (Fluka–Cat. No: 09833), DMSO (Sigma-Aldrich—Cat. No: D8418), Trifluoroacetic Acid–TFA (Thermo Scientific—Cat. Number 85183). Assay were performed in 384 well microplate, PS, small volume, hibase, white (Greiner bio-one–Cat. Number 784904) and covered using Silverseal Sealer Aluminium Foil (Greiner Bio-One—Cat. Number 676090). To prepare DHAP matrix solution, 25 mg diammonium hydrogen citrate were dissolved in 1 ml of milliQ water. 7.6 mg of DHAP matrix were dissolved in a solution of 375 ul of LC-MS grade Ethanol and 125 uL of diammonium hydrogen citrate solution with the aid of vigorous vortexing.

Enzymatic reactions and MALDI-TOF High-Throughput Screening (HTS) were set up in assay as previously described [40]. Briefly, FDA approved compound library was delivered into 384 well plated plates using non-contact acoustic delivering. Two columns were reserved for positive (DMSO only) and negative controls (no enzyme). PLpro and K48 ubiquitin trimer were diluted in buffer containing 40 mM Tris (pH 7.5), and 0.01% BSA at a final concentration of 1.2 ng/μl and 0.2 mg/ml, respectively. 3 μl of PLpro solution was aliquoted into 384 well plates using FluidX XRD system equipped with 16 channel tubing cartridges. Plates were centrifuged for 1 minute at 1000 x g, covered using aluminium sealing foil to prevent evaporation and incubate at room temperature for 30 minutes. Reaction was started by adding 3 μl of K48 ubiquitin trimer on the entire assay plate and incubated at room temperature for 30 minutes. The reaction was stopped by adding 3 μl TFA—final concentration of 2% (v/v) supplemented with 4 μM ¹⁵N labelled ubiquitin. A TTP Labtech Mosquito® HTS equipped with 5 positions plate deck was employed to mix sample and DHAP matrix and to spot the mixture on the MALDI target. Four 384-well plates were combined into one Anchor Chip 1536 BC MALDI-TOF target plates. Spotted and dried MALDI target was subjected to automatic MALDI-TOF MS analysis. Mass spectra were acquired with a RapifleX MALDI-TOF/TOF instrument from Bruker Daltonics equipped with Compass software for FlexSeries 2.0. Mass spectra were acquired in the mass range of m/z 8000–9200, to include ubiquitin (8565.76 kDa) and the ¹⁵N ubiquitin (8669.47 kDa) signals. 4000 laser shots per sample spot were accumulated in positive ionization mode using a 500 ns pulsed ion extraction with a 10 kHz laser frequency, a digitizer setting of 5.00 GS/s, and a single Smartbeam parameter at a 800 μm scan range. The acquired spectra were processed with a centroid peak detection set to a signal-to-noise ratio (S/N) of 5 and Gaussian smoothing (0.5 m/z; 10 cycle). The signal of ¹⁵N ubiquitin was added to the mass control list and used as internal calibrant. MALDI-TOF data, processed with flexAnalysis were exported as a comma delimited (.csv) file. Data sets were further processed using Excel or GraphPad Prism (v7.03; GraphPad Software, La Jolla, CA). PLpro activity was tracked using the measured area (signal) of the enzymatic product (ubiquitin) and the corresponding labelled internal standard (¹⁵N ubiquitin). All data points were reported as ratio of ubiquitin signal over the internal standard signal. Generated sample data were expressed as percentage of control (PoC) according to the following equation:

Were b being the average of the negative controls and c the average of the positive controls.

To assess the robustness and quality of the hit discovery campaign a key metric, the Z’ parameter [41] was employed as for following equation:

Were σ_p, σ_n being positive and negative control standard deviations and μ_p, μ_n the difference of the averages of positive and negative controls. GraphPad Prism was utilized for graphical representation of the data.

FDA approved compound Set: Apex bio DiscoveryProbe™ FDA-approved Drug Library (Catalog No. L1021) is a unique collection of 1971 small molecule, FDA approved drugs with known bioavailability and safety data in humans.