Ubiquitin (Ub) and the Ub-like (Ubl) modifier interferon-stimulated gene 15 (ISG15) participate in the host defence of viral infections. Viruses, including the severe acute respiratory syndrome human coronavirus (SARS hCoV), have co-opted Ub–ISG15 conjugation pathways for their own advantage or have evolved effector proteins to counter pro-inflammatory properties of Ub–ISG15-conjugated host proteins. In the present study, we compare substrate specificities of the papain-like protease (PLpro) from the recently emerged Middle East respiratory syndrome (MERS) hCoV to the related protease from SARS, SARS PLpro. Through biochemical assays, we show that, similar to SARS PLpro, MERS PLpro is both a deubiquitinating (DUB) and a deISGylating enzyme. Further analysis of the intrinsic DUB activity of these viral proteases revealed unique differences between the recognition and cleavage specificities of polyUb chains. First, MERS PLpro shows broad linkage specificity for the cleavage of polyUb chains, whereas SARS PLpro prefers to cleave Lys48-linked polyUb chains. Secondly, MERS PLpro cleaves polyUb chains in a ‘mono-distributive’ manner (one Ub at a time) and SARS PLpro prefers to cleave Lys48-linked polyUb chains by sensing a di-Ub moiety as a minimal recognition element using a ‘di-distributive’ cleavage mechanism. The di-distributive cleavage mechanism for SARS PLpro appears to be uncommon among USP (Ub-specific protease)-family DUBs, as related USP family members from humans do not display such a mechanism. We propose that these intrinsic enzymatic differences between SARS and MERS PLpro will help to identify pro-inflammatory substrates of these viral DUBs and can guide in the design of therapeutics to combat infection by coronaviruses.
The emergence of novel infectious agents of zoonotic origin that can cross species barriers and pose a significant risk to human health is a constant threat today . Over a decade ago, a novel coronavirus caused severe upper respiratory infection in humans and was termed the severe acute respiratory syndrome human coronavirus (SARS hCoV) . SARS caused a global pandemic and had mortality rate of ∼10%  due to its high infectivity, but was eventually contained by healthcare professionals through early detection and patient isolation. However, to date, there is no specific treatment available for SARS or any other type of coronavirus infection, either in the form of small-molecule inhibitors or vaccines [2,4]. Although the last known human-to-human SARS infection chain was broken in 2003, the Centers for Disease Control and Prevention (CDC) has recently classified SARS as a select agent (posing a high risk to human health), owing to the high risk of re-emergence and the lack of available treatment options. Just as SARS began to fade from the public's memory, a new hCoV emerged in 2012, termed Middle East respiratory syndrome (MERS) hCoV, which to date has infected close to a thousand individuals with a ∼30% mortality rate (http://www.cdc.gov/coronavirus/mers/index.html). MERS hCoV, like SARS hCoV, is a β-coronavirus of zoonotic origin, which is also related to coronaviruses found in bats  and recent evidence suggests that MERS hCoV probably utilized dromedary camels as intermediary hosts . MERS hCoV, like SARS hCoV, causes severe respiratory infection, as well as acute renal failure in some cases [3,7]. MERS is currently an ongoing pandemic, with the first U.S. cases reported in May 2014.
Ubiquitin (Ub) is a small protein that post-translationally modifies target substrates through a defined enzymatic cascade involving E1–E2–E3 enzymes [8–10]. Conversely, deubiquitinating enzymes (DUBs) cleave Ub from substrates [11,12]. Ub can be conjugated to itself via any of its seven lysines, as well as via its N-terminus, leading to different functional consequences for the modified substrate . The most common function of ubiquitination is proteasomal degradation of the substrate, mediated by Lys48-linked polyUb chains . Interferon-stimulated gene 15 (ISG15), a ubiquitin-like (Ubl) modifier composed of two Ubl folds in tandem, is conjugated to substrates by an orthogonal enzymatic cascade  and plays a role in tagging newly synthesized proteins during an antiviral response [15,16]. Topologically, ISG15 resembles the open conformation of a linear or a Lys63-linked di-Ub molecule [17,18], whereas Lys48-linked chains are more dynamic and have context-dependent topologies, according to their interacting proteins .
Both Ub and ISG15 are involved in fighting viral infection and several viruses have co-opted Ubl conjugation pathways for their own advantage or have evolved effector proteins to counter the pro-inflammatory properties of Ubl-conjugated host proteins . For coronaviruses, one such mechanism relies on dampening the host immune response via the action of viral isopeptidases . Thus, for SARS, a highly promising drug target is one of its two viral processing proteases, the SARS papain-like processing protease (PLpro) , a multi-functional papain-like cysteine protease . In addition to cleaving the viral polypeptide at defined positions, SARS PLpro can remove Ub and ISG15 from cellular conjugates. These activities are believed to be important for its anti-inflammatory function, although the specific ubiquitin- and/or ISG15-conjugated substrates have yet to be identified. Recently, the equivalent processing protease from the MERS hCoV, termed MERS PLpro, has also been shown to have deubiquitinating and deISGylating activities when expressed in cells [24,25].
In the present study, we undertook a biochemical characterization of MERS PLpro and a comparison of its isopeptidase activity with SARS PLpro. We show that, similar to SARS PLpro, purified, recombinant MERS PLpro is a dual-function isopeptidase that can cleave ISG15 and Ub conjugates. Using a positional scanning substrate library screen, we show that MERS PLpro has broad cleavage site specificity in terms of amino acid sequence, a feature it shares with SARS PLpro. Furthermore, we demonstrate MERS PLpro to be a mono-distributive DUB that recognizes and cleaves one Ub unit within a chain. We also demonstrate MERS PLpro to be a broad specificity DUB with respect to Ub–Ub linkages. In contrast to MERS PLpro, we found that the unit of recognition and cleavage for SARS PLpro is a Lys48-linked di-Ub unit within a chain length of three or greater, making it a unique ‘di-distributive’ DUB. This finding allows us to dispel the recent notion that SARS PLpro has a preference for cleaving ISG15 conjugates, as opposed to Ub conjugates, as SARS PLpro possesses the capacity to efficiently cleave both polyUb chains and ISG15 conjugates at similar rates. Furthermore, we show that this unique di-distributive mechanism of SARS PLpro can result in the accumulation of di-Ub Lys48 units as well as of monoUb-conjugated substrates, which could have functional consequences during coronavirus infection. Our study reveals key mechanistic differences between two related coronavirus proteases that may be used in identifying their cellular substrates and which may aid in inhibitor design strategies. Finally, our study reinforces the need to investigate the comparative advantage that deubiquitination compared with deISGylation might confer on coronaviruses.
Ubiquitin probe labelling assays
Unless otherwise indicated, DUBs (1 μM) were incubated with an excess of activity-based probes [ABPs; Ub–PA (ubiquitin-propargylamide), monoUb–VME (ubiquitin-vinylmethyl ester) or di-UbLys48-ABP, 5 μM] for the indicated times (1 min–1 h) at 37°C in 20 mM Tris, pH 8.0, 150 mM NaCl and 5 mM DTT. Reactions were terminated with loading sample buffer [4× lithium dodecyl sulfate (LDS), Invitrogen], boiled for 5 min at 95°C and analysed by SDS/PAGE and SYPRO-staining. Gels were imaged on Bio-Rad Gel-Doc imagers.
Ubiquitin chain cleavage assays
Unless otherwise indicated, in a reaction volume of 10 μl, 0.5 μg of Ub chains (in 20 mM Tris/HCl, pH 8.0, 150 mM NaCl and 5 mM DTT) were cleaved with 2–1000 nM DUB at 37°C for the indicated times in a time-course assay or by 0.2-fold serial dilution of DUBs (starting at 1 μM) for 1 h at 37°C. Reactions were terminated with 4× loading sample buffer (LDS, Invitrogen), boiled for 5 min at 95°C and analysed by SDS/PAGE and SYPRO-staining. Gels were imaged on Bio-Rad Gel-Doc imagers, quantified using ImageJ (NIH) and cropped where indicated.
Ubiquitin–and ISG15–AMC kinetics
To determine apparent kcat/Km values for SARS and MERS PLpro, Ub–AMC (amino-methyl coumarin) and ISG15–AMC were prepared as 0.5-fold serial dilutions (starting at 20 and 12 μM respectively) in 20 mM Tris/HCl, pH 8.0, 150 mM NaCl and 5 mM DTT. SARS PLpro was used at 10 nM, MERS PLpro was used at 50 nM and the final reaction volume was 20 μl. Substrates and DUBs were pre-incubated at 37°C for 5 min and cleavage of Ubl–AMCs was performed at 37°C using a Spectramax fluorescence plate reader running SoftMax Pro 5 software (Molecular Devices) operated in kinetic mode, in black flat-bottomed 384-well plates (Corning), where free AMC fluorescence was monitored by excitation at 355 nm and emission at 460 nm. Initial linear cleavage rates (Vi) were fitted by the Michaelis–Menten equation using Prism software (GraphPad) based on a free AMC standard curve. Where indicated, experiments were performed in duplicate; error bars indicate S.E.M.
Synthesis of the diverse tetrapeptide-ACC (7-amino-carbamoylmethylcoumarine) Positional Scanning Substrate Combinatorial Library
Synthesis of the PS-SCL (Positional Scanning Substrate Combinatorial Library) for screening of P4, P3 and P2 sub-sites was carried out as described previously [26,27]. After completing the synthesis and analysis steps, each sub-library was dissolved in biochemical grade DMSO at a concentration of 20 mM and stored at −80°C until use.
Fluorogenic substrate synthesis
The synthesis of individual fluorogenic substrates was carried out using classic solid-phase peptide synthesis methods as described previously . Each substrate was purified by HPLC on a Waters M600 solvent-delivery module with a Waters M2489 detector system using a semi-preparative Waters Spherisorb S10ODS2 column. The solvent composition was as follows: phase A [water/0.1% TFA (trifluoroacetic acid)] and phase B (acetonitrile/0.1% TFA). The purity of each substrate was confirmed by analytical HPLC using a Waters Spherisorb S5ODS2 column. Finally, the molecular mass of each substrate was confirmed by MS analysis. All compounds were at least 95% pure. Individual substrates were dissolved at 20 mM in DMSO and stored in −80°C until use.
Assay of the PS-SCL and individual tetrapeptides
Kinetic assays were performed at 37°C in buffer containing 150 mM NaCl, 6 mM DTT and 20 mM Tris/HCl, pH 7.5. Assay conditions for the P2, P3 and P4 libraries were as follows: the reaction volume was 100 μl; the total final substrate mixture concentration was 300 μM; MERS PLpro was assayed at 1 μM. Enzyme was pre-incubated for 30 min at 37°C in buffer, followed by addition into 96-well Corning® plate wells containing each substrate. The reaction was monitored for 60 min in kinetic mode. To calculate initial cleavage rates (Vi), the linear portion of the obtained progress curves was used. Results are presented as an average for at least three experiments. Analysis of the results was based on total relative fluorescence units (RFU) for every sub-library, setting the highest value to 100% and adjusting the other results accordingly. Individual substrates were screened against MERS PLpro at 37°C in buffer containing 150 mM NaCl, 6 mM DTT, 20 mM Tris/HCl, pH 7.5 and 1 M of the Hofmeister salt sodium citrate. Enzyme was pre-incubated for 30 min at 37°C before being added to wells containing each substrate. Assay conditions were as follows: 100 μl reaction volume; final substrate concentrations ranged from 500 to 29.3 μM; and MERS PLpro concentration was 1 μM. The concentration of DMSO in the assay was less than 1% (v/v). Release of the ACC fluorophore was monitored continuously with excitation at 355 nm and emission at 460 nm. Initial velocity rates were fitted using GraphPad Prism 6 to the Michaelis–Menten curve, determining Vmax and Km.
DUB assay in cell lysates
Lysates from human interferon β (IFNβ)-treated HeLa cells (10 μg of total protein per reaction) were incubated with a 0.1-fold serial dilution of 10 μM DUB in 20 μl reaction volumes for 1 h in the presence of 25 mM DTT or with 0.5 μM DUBs for the time-course assay. Reactions were terminated by boiling in SDS loading buffer and the results were analysed by SDS/PAGE and Western blotting. Blots were developed by chemiluminescence.
MERS PLpro is a broad-specificity isopeptidase
To characterize MERS PLpro, we cloned and expressed the putative MERS PLpro (including its Ubl domain) using domain limits based on alignments with the related SARS PLpro (Supplementary Figure S1A). First, we tested MERS PLpro by comparing its activity on the fluorogenic model substrates Ub–and ISG15–AMC. Unlike SARS PLpro, which displays a strong preference for cleaving ISG15–AMC over Ub–AMC (Figure 1A; Supplementary Figures S1B and S1C) [23,29,30], MERS PLpro did not display a significant preference for either, but processed ISG15–and Ub–AMC similarly when determining overall relative kcat/Km values (Figure 1A). Additionally, MERS PLpro was more reactive than SARS PLpro towards a Ub ABP, Ub–PA, which is used to covalently label active DUBs (Figure 1A, inset), these results suggest divergent preferences of SARS and MERS PLpro towards Ub and ISG15 model substrates.
MERS PLpro is a broad-specificity isopeptidase
Next, to determine the cleavage site specificity of MERS PLpro with regard to the primary amino acid sequence of the cleavage site (P4–P1 residues of the substrate, which correspond to Leu73-Arg74-Gly75-Gly76 in Ub and Leu154-Arg155-Gly156-Gly157 in ISG15), we generated a PS-SCL (Supplementary Figure S1D) that has been previously used to assay the specificity of SARS PLpro and other DUBs . The results of the PS-SCL screen demonstrated that MERS PLpro has an absolute requirement for glycine in the P2 position, but can accommodate a wide variety of hydrophobic amino acids in both the P3 and the P4 positions (Figure 1B). In the context of this tetrapeptide screen, the most preferred amino acid in P3 was leucine, followed by arginine, lysine and hydrophobic residues. The most preferred amino acid in P4 was also leucine, followed by mostly hydrophobic residues. The library screen results are in line with the three cleavage sites that MERS PLpro processes within the viral polypeptide (Figure 1C), where the P1 and P2 sites are both glycine, the P3 residues are isoleucine, lysine or valine and the P4 is either a leucine or an isoleucine . Thus, these results indicate that the active site of MERS PLpro appears to be less stringent. When comparing PS-SCL library screen results of other DUBs  from the USP family (which contains both SARS and MERS PLpro), it appears that MERS PLpro exhibits a lesser degree of specificity in P4 than either SARS PLpro or USP5. Its specificity in P3 is broad, yet more stringent than that of SARS PLpro (Supplementary Figure S1E).
PS-SCL screen reveals relaxed cleavage site specificity for MERS PLpro, allowing cleavage of a DUB-resistant ubiquitin mutant
Notably, to date, no isopeptidase has been shown to accommodate proline in P4  and the positional library screen suggested that MERS PLpro may be able to cleave substrates containing a proline in the P4 position (Figure 1B, bottom panel). To confirm the cleavage of proline in P4, we generated unique tetrapeptides with defined sequences and determined the rates of cleavage by MERS PLpro (Figure 1D). The results showed that proline can be accommodated in P4 for both tetrapeptides tested, although with reduced rates and that there is little discrimination between leucine and arginine in the P3 position.
In a recent study, we generated a UbL73P mutant that was resistant to cleavage by different families of human DUBs when conjugated to substrates or as polyUb chains . Intrigued by the ability of MERS PLpro to accommodate proline in P4 (corresponding to UbL73P) we tested whether MERS PLpro can cleave UbL73P. Indeed, we show that in the context of di-Ub chains, MERS PLpro, but not SARS PLpro or USP2CD, could cleave UbL73P chains, although more weakly when compared with wild-type chains (Figure 2E). The ability of MERS PLpro to cleave UbL73P was observed for Lys11-, Lys48- and Lys63-linked di-Ub chains, but not for M1-linked (linear) di-Ub chains, suggesting some level of linkage specificity for MERS PLpro. Thus, we determined the linkage specificity of MERS PLpro by assaying a panel of di-Ub molecules (all eight linkages)  in a time-course assay. Using different concentrations of MERS PLpro, we demonstrate that MERS PLpro is a broad-specificity DUB, displaying cleavage of Lys6-, Lys11-, Lys29-, Lys33-, Lys48- and Lys63-linked chains, but not linear di-Ub chains (Figure 1F, top panel; Supplementary Figure S1F). In contrast, cleavage of di-Ub chains by SARS PLpro, which displays broad linkage specificity in this assay, is not as robust as cleavage by MERS PLpro when comparing the two enzymes at similar enzyme concentrations (Figure 1F, bottom panel). Neither DUB was able to cleave linear di-Ub or Lys27-linked di-Ub, although the latter could be structurally quite compact.
Enhanced activity of SARS PLpro on longer Lys48-linked Ub chains
SARS PLpro recognizes a Lys48-linked tri-Ub as a minimal unit for ubiquitin chain processing
The low activity of SARS PLpro on di-Ub chains compelled us to test its activity on longer Ub chains. Surprisingly, when analysing Lys48-linked tetra-UbL73P chains, both MERS and SARS PLpro were able to cleave the mutant Ub chains (Figure 2A, right panel). The cleavage of UbL73P chains was impaired compared with wild-type chains; however, the slowed kinetics allowed visualization of different cleavage intermediates for MERS and SARS PLpro (compare Figure 2A, left and right panels). The tetra-Ub cleavage pattern for USP2CD and MERS PLpro appeared to resemble a distributive cleavage pattern, as opposed to the SARS PLpro cleavage pattern, which accumulated di-Ub as an intermediate form concomitant with the appearance of monoUb products, during cleavage of UbL73P, suggesting that SARS and MERS PLpro recognize Ub chains in a different manner. Interestingly, a detailed look at the active-site clefts of the crystal structures of USP2CD, SARS and MERS PLpro bound to mono-Ub [32–34] suggests a lack of a defined pocket for P4 in MERS PLpro and SARS PLpro, whereas that pocket is present in USP2CD (Supplementary Figure S1G), potentially explaining the ability of viral DUBs to accommodate proline in P4. Thus, overall, MERS PLpro is a dual-specificity isopeptidase that cleaves both ISG15 and various Ub conjugates at similar rates. In contrast, SARS PLpro appears to be a poorer DUB on di-Ub substrates, but a superior deISGylase as compared with MERS PLpro. Importantly, the cleavage of wild-type tetra-Ub chains revealed that SARS PLpro can cleave tetra-Ub chains more efficiently than MERS PLpro, whereas the opposite is true for the cleavage of di-Ub chains. These differences suggest that MERS and SARS PLpro are mechanistically distinct in their ability to recognize ISG15 and Ub conjugates.
To further investigate the unique property of SARS PLpro in stabilizing di-Ub cleavage products from a poly-Ub chain, we compared the processing rates for SARS and MERS PLpro on tetra-, tri- and di-Ub substrates (Figures 2B and 2C). We show that SARS PLpro (Figures 2B, top panels, and 2C, black lines) exhibits negligible cleavage of Lys48-linked di-Ub chains but readily process tetra- and tri-Ub chains. The patterns of cleavage intermediates from the time-course study indicate that SARS PLpro cleaves a tri-Ub chain into a di-Ub and a mono-Ub product. The cleavage of a tetra-Ub chain by SARS PLpro is more complex, as it can either generate a tri-Ub and a mono-Ub product, which can then be further processed into a di-Ub and a mono-Ub or the tetra-Ub can be cleaved to generate two di-Ub units. Thus, different modes of tetra-Ub cleavage will result in the accumulation of di-Ub (and monoUb) along with the transient appearance of the tri-Ub form (Figure 2B). SARS PLpro therefore appears to preferentially recognize and cleave Lys48-linked chains in a ‘di-distributive’ manner. This di-distributive cleavage mechanism is best demonstrated when analysing cleavage of Lys48-linked penta-Ub chains at lower temperatures (Figure 2D), allowing visualization of immediate cleavage intermediates. In Figure 2D, it is visually striking that the initial cleavage products, arising at highly comparable rates from processing a penta-Ub chain, are tri- and di-Ub intermediates. This result strongly indicates that cleavage of a five-member Ub chain can take place stochastically either between the second and third Ub moiety or between the third and fourth Ub moiety (from either end of the chain), both of which will result in the transient accumulation of tri- and di-Ub cleavage intermediates. The lack of transient accumulation of a single cleavage intermediate, combined with the observation of mono- and tetra-Ub cleavage intermediates suggests a lack of directionality in Ub chain cleavage, strongly suggesting that SARS PLpro is a distributive DUB that processes Ub chains in units of two. Consequently, further cleavage of a di-Ub molecule into monomers appears to be inhibited.
Unlike SARS PLpro, cleavage rates with MERS PLpro (Figure 2B, bottom panels, and Figure 2C, grey lines) were nearly identical against Ub chains of various lengths and the patterns of cleavage intermediates demonstrate that MERS PLpro is a mono-distributive DUB that cleaves single units of Ub (like the majority of DUBs studied thus far). The di-Ub-stabilizing cleavage pattern of SARS PLpro appears to be Lys48-specific, as no di-Ub intermediates accumulate during cleavage of Lys63-linked tetra-Ub chains (Figure 2E). Additionally, Lys63-linked chains are cleaved poorly by SARS PLpro compared with MERS PLpro (Supplementary Figure S2A). Both SARS and MERS PLpro failed to cleave linear (M1-linked) tetra-Ub chains, even at equimolar DUB/substrate concentrations (Supplementary Figure S2B). These results collectively support the idea that MERS PLpro is a broad-specificity mono-distributive DUB, whereas SARS PLpro is di-distributive and is significantly more active on longer Lys48-linked Ub chains.
The comparatively poor cleavage activity of SARS PLpro on di-Ub chains could be due the inability of SARS PLpro to recognize and bind di-Ub chains. To test this, we generated recombinant histidine-tagged catalytic mutant DUBs and used them for pull-down assays by incubating them with different mixtures of Ubl units (Figures 3A and 3B). We show that SARS PLpro preferentially binds to Lys48-linked di-Ub chains over mono-Ub when both were present at equimolar ratio. In contrast, USP2CD preferentially favoured mono-Ub over di-Ub and MERS PLpro bound to both similarly (Figure 3A). Interestingly, SARS PLpro slightly favoured Lys48-linked di-Ub chains over ISG15 when both were present, whereas the opposite was true for MERS PLpro (Figure 3B). Thus, it is not the inability of SARS PLpro to recognize di-Ub in order to cleave it, but perhaps the binding of di-Ub by SARS PLpro is mediated by a region that renders the Ub–Ub internal cleavage site inaccessible for cleavage by the active site of the protease.
SARS PLpro is a di-UbLys48-binding, triUbLys48-cleaving DUB
Recent advances in chemical biology have enabled the generation of Ub ABPs for DUBs that are linkage-specific, where the warhead used to covalently label the active-site cysteine of a DUB is positioned in between the two Ub moieties in a di-Ub probe [35,36]. Using this tool, we tested whether SARS could efficiently bind di-Ub across its active site in the S1-S1′ position (adapting the nomenclature of amino acid positions in protease cleavage sites  to positions occupied by the Ubl protein domains), in order to favour labelling with the ABP; or whether another position is utilized (such as S2-S1), which would prevent reactivity (Figure 3E). Also, as both SARS and MERS PLpro contain ubiquitin-like domain (Ubls) in their N-terminus, we wondered whether the Ubl domain plays a role in the correct positioning of longer Ub chains for SARS PLpro. We show that whereas the Lys48-specific di-Ub ABP (di-UbLys48–ABP), as well as a monoUb–ABP, efficiently labelled MERS PLpro and its ∆Ubl mutant to form a covalently modified DUB (Figure 3C, right side), the di-UbLys48–ABP failed to efficiently label SARS PLpro or its ∆Ubl mutant (Figure 3C, left side). It appears that the lack of di-UbLys48–ABP labelling for SARS PLpro is probably due to selective binding of the di-UbLys48 chain at a position other than the S1-S1′ position. Thus, the active-site cysteine cannot react with the warhead of the ABP. Consistent with Ub–ABP labelling, removal of the Ubl domains of either SARS or MERS PLpro had no effect on their Lys48-linked tetra-Ub cleavage rates or the pattern of intermediates observed (Supplementary Figures S2C and S2D). Thus, the Ubl domains of both MERS and SARS PLpro are dispensable for their Ub chain recognition and processing activities.
To determine whether the different mechanisms for substrate recognition and processing between MERS and SARS PLpro remain observable in the context of multiple cellular substrates, we aimed to create an experimental system that would contain both ISGylated and Lys48-linked polyubiquitinated substrates, where we could simultaneously monitor the appearance of cleavage intermediates of both Ubl proteins. To this end, we treated HeLa cells with IFNβ and the proteasome inhibitor MG132, to induce stabilization of both ISGylated and Lys48-linked polyubiquitinated substrates in cells. Next, the lysates were incubated with recombinant viral DUBs in a time-course assay and the substrate-processing patterns were analysed by Western blotting by probing for total cellular Ub and ISG15 conjugates (Figure 3D). The results indicated that MERS and SARS PLpro are both potent deISGylating enzymes, whereas SARS PLpro appeared to process higher-molecular-mass (HMr) Ub conjugates slightly more efficiently. Importantly, in the context of a mixture of cellular substrates, SARS PLpro initially accumulated di-Ub chains as a cleavage product, whereas MERS PLpro did not (Figure 3D, black box). Assaying IFNβ-/MG132-treated cell lysates with a serial dilution of viral DUBs (Supplementary Figures S2E and S2F) also indicated initial stabilization of di-Ub products for SARS PLpro, overall demonstrating the same mechanistic differences between SARS and MERS PLpro with respect to Ub chain processing in the context of cellular substrates.
In summary, we have demonstrated that SARS PLpro preferentially recognizes di-UbLys48 as a unit (possibly in the S2-S1 pockets) and efficiently cleaves the next Ub in the chain (in S1′), resulting in accumulation of di-UbLys48 products. These results strongly suggest that the minimal preferred substrate for SARS PLpro is a Lys48-linked tri-Ub chain (spanning S2-S1-S1′; Figure 3E). Importantly, the related MERS PLpro lacks this unique recognition and cleavage preference.
Orientation of ubiquitin chain cleavage by SARS PLpro
Recognition of a di-Ub moiety and cleavage of a third Ub in a tri-Ub chain by SARS PLpro is presumed to be via the S2-S1/S1′ sub-sites (in a ‘distal’ position; Figure 3E, top, shown by a black unbroken line), largely based on the presumed similar recognition of ISG15, where the two Ubl domains of ISG15 could only occupy those same S2-S1 sub-sites, restricted by the positioning of its C-terminal glycine that is cleaved. Alternatively, recognition of a tri-Ub chain may occur in an S1/S1′-S2′ binding mode (in a ‘proximal’ position; Figure 3E, top, shown as a grey dotted line), where, recognition of a di-Ub would take place on the other side of the cleavage site. The results presented thus far cannot differentiate between the two modes of orientation within a tri-Ub chain, given the symmetry in the system and the identical size of the cleavage intermediates. Thus, we sought to determine the orientation of the Ub chain cleavage mechanism by SARS PLpro to differentiate between ‘distal’ Ub chain processing (S2-S1/S1′) compared with a ‘proximal’ (S1/S1′-S2′) mechanism, which would place recognition of ISG15 and the di-UbLys48 unit on opposing sides of the SARS active site.
To determine the directionality in the cleavage mechanism, we generated a purified histidine-tagged substrate conjugated with FLAG-tagged Ub (FLAG–Ub) chains (Figure 4A; Supplementary Figure S3A). We chose Cdc34 (cell divison cycle 34, UbE2R1) as a model substrate, whose auto-conjugation results in Lys48-linked Ub chains. When purified histidine–Cdc34 conjugated with FLAG–Ub chains was cleaved by SARS PLpro in a time-course assay, Cdc34–Ub conjugates of various lengths, as well as free Cdc34, appeared as initial cleavage products (Figure 4B, compare lanes 1 and 2). Ultimately, two unique product species were stabilized by SARS PLpro: monoUb-conjugated Cdc34 and free Cdc34 (Figure 4B, lanes 5 and 6; Supplementary Figure S3B). Cleavage of the same poly-Ub-conjugated Cdc34 by MERS PLpro, on the other hand, did not result in the accumulation of monoUb-conjugated Cdc34. In the case of MERS PLpro, even though the initial cleavage intermediates were similar to those of SARS PLpro (various length Cdc34–Ub conjugates), the final product was free fully deconjugated Cdc34 (Figure 4B, right side). These results are consistent with previous experiments illustrating mechanistic differences between SARS and MERS PLpro (shown in Figures 2 and 3), where SARS PLpro stabilized di-UbLys48 cleavage products arising from unanchored Ub chains, whereas MERS PLpro did not. Additionally, these results strongly support a mechanism where recognition of Lys48-linked Ub chains by SARS PLpro in the context of a conjugated substrate take place via the ‘distal' approach (via the S2-S1/S1′ sub-sites; Supplementary Figure S3C) and not by the ‘proximal’ binding mode. In the latter case, the final accumulated cleavage product would have been a di-Ub-conjugated substrate (the two Ub moieties in S1′-S2′, whereas the conjugated substrate would occupy the S3′ sub-site), which was not observed (Figure 4B; Supplementary Figure S3B). Furthermore, the appearance of initial cleavage products of various lengths also supports a distributive cleavage mechanism, where there is no discrimination between the positions at which SARS or MERS PLpro cleaves a Ub chain.
SARS PLpro is a unique ‘distal di-distributive’ DUB
Interestingly, such a ‘distal di-distributive’ cleavage mechanism of Lys48-linked chains appears to be unique to SARS PLpro, as two human DUBs from the same USP family (USP2 and USP21) also do not accumulate monoUb-conjugated Cdc34 as the final cleavage product, but completely deconjugate the substrate (Figures 4C and 4D; Supplementary Figure S3D). Collectively, these results strongly suggest a di-distributive cleavage mechanism for SARS PLpro (where the di-Ub moiety is recognized in the S2-S1 sub-sites), which may lead to accumulation of monoUb-conjugated substrates, depending on the positioning of the cleavage event with respect to the substrate (Figure 4E).
We show in the present study that although the viral polyprotein processing proteases from the MERS and SARS coronaviruses are both efficient DUB and deISGylating enzyme, they utilize different mechanisms for substrate recognition and processing. MERS PLpro is mono-distributive DUB, whereas SARS PLpro prefers recognition of a di-UbLys48 unit, thus it is a di-distributive DUB (Figure 4E; Supplementary Figure S3E; and discussion below).
Previously, MERS PLpro has been demonstrated to cleave Ub and ISG15 conjugates when expressed in cells [24,25]; in the present study, we provide direct in vitro evidence demonstrating that MERS PLpro is a dual-specificity isopeptidase that processes Ub and ISG15 conjugates similarly. Our analysis is consistent with a very recent report on the substrate specificity of MERS PLpro , where although the authors noted slightly higher processing rates for ISG15–AMC, MERS PLpro was also demonstrated to cleave Ub chains one Ub at a time. We additionally demonstrate that MERS PLpro is a broad-specificity DUB, like most human DUBs from the USP family , which cleaves various Ub linkages with similar efficiency, with the exception of linear chains. This lack of specificity for cleaving di-Ub chains is somewhat shared by SARS PLpro, when a di-Ub molecule occupies the S1-S1′ sub-sites. However, given the preference of SARS PLpro to recognize di-UbLys48 units (discussed in detail below), the physiological relevance of the broad linkage specificity of SARS PLpro at the S1-S1′ sub-sites remains to be determined.
It is interesting to note that linear Ub chains are not cleaved by either SARS or MERS PLpro, even though the cleavage of a linear di-Ub is proteolytically the same as cleavage of a viral polypeptide at the processing sites, where the protease is cleaving a peptide bond (as opposed to various isopeptide-bond linkages in Ub–Ub dimers). Furthermore, since cleavage of Lys63-linked Ub chains (which are similar in topology to linear Ub chains ) is well tolerated by MERS PLpro, it would be expected that viral DUBs might also cleave linear Ub chains. Interestingly, some human DUBs of the USP family, such as USP7, USP8 and USP11, also cannot tolerate linear Ub chains, even though they are less specific towards isopeptide linkages . One reasonable explanation is that the methionine sidechain of the glycine-methionine junction of a linear Ub is stearically hindered for SARS and MERS PLpro (and for some human DUBs) to accommodate, whereas an isopeptide bond via a lysine side chain is not. The flexibility of the isopeptide linkage could also help to explain the lack of linkage specificity for Ub–Ub linkages (when occupying S1-S1′-binding sites), suggesting that there could be only minimal contacts with the substrate on the primed side of the cleavage site (S1′ sub-site for Ub chains or the sequence following the viral polypeptide following the cleavage site). On the other hand, in a linear di-Ub, the flexible C-terminus of the distal Ub (in S1) ending in glycine continues immediately into the initiator methionine of a rigid β-strand of the proximal Ub (in S1′), potentially leading to a Ub–Ub conformation which is not suitable for binding in the S1′ pocket of viral PLpros. Further structural work on viral DUBs in complex with longer Ub chains, or with viral polyprotein substrates, that span the active site are needed to shed light on the primed side specificity of viral DUBs.
On the other hand, addressing the issue of unprimed side specificity of viral DUBs, using a PS-SCL we show that, unlike human DUBs, both SARS and MERS PLpro have broad cleavage-site specificities in terms of amino acid preference. Both SARS and MERS PLpro appear to have loosely defined pockets in their active-site clefts, even though available crystal structures of both SARS and MERS PLpro in complex with mono-Ub [33,34] indicate an intricate network of hydrogen bonds stabilizing the C-terminus of Ub in their active-site clefts. Nevertheless, they are both able to accommodate a hitherto DUB-resistant mutation in Ub , suggesting greater flexibility of their active-site clefts than human DUBs. Interestingly, the accommodation of proline in P4 for SARS PLpro is only observable in the context of longer, tetra-Ub chain substrates, suggesting that recognition and processing of poly-Ub chains by SARS PLpro is via a unique mechanism. Whether these differences in viral PLpro active-site specificities may be exploited in inhibitor design strategies has yet to be addressed.
Furthermore, we demonstrate mechanistic differences between SARS and MERS PLpro in how they recognize and process poly-Ub chains (Figure 4E; Supplementary Figure S3E): MERS PLpro is a mono-distributive DUB, which cleaves Ub chains one Ub at a time (like the majority of human DUBs), whereas SARS PLpro is preferentially a di-distributive DUB that recognizes Lys48-linked di-Ub molecules as its minimal recognition unit and a processes the third Ub in a polyUb chain or removes a di-Ub from a substrate. This unique recognition is possibly mediated by an extended binding surface in SARS PLpro, which is not conserved in MERS PLpro. Traditionally, it has been accepted that SARS PLpro is primarily a deISGylating enzyme, mainly based on its preference for ISG15–AMC over Ub–AMC [21,23,30]. We propose that this is an inaccurate description of the activity and function of SARS PLpro, as in the present study show the cleavage of poly-Ub chains (minimally a tri-Ub- or a di-Ub-conjugated substrate) to be highly efficient and comparable to that of ISG15-conjugates. MERS PLpro does not share this mechanism, as it possibly recognizes Ub or ISG15 in the S1 site and cleaves the next Ub in the S1′ site, in a broadly linkage-independent manner, with no or minimal contacts in the S2 site (Figure 4E; Supplementary Figure S3E).
While our paper was in preparation, two recent papers from Mesecar and colleagues [33,38] demonstrated a similar mechanism of bivalent Ubl recognition for SARS PLpro, which is not shared by MERS PLpro. By modelling and mutational analysis, the authors identified a region in SARS PLpro (Phe70-Leu76) that is responsible for both di-UbLys48 and ISG15 recognition in the S2 sub-site and which is not conserved in MERS PLpro. In their model, Ratia et al.  also suggested the S2 interaction to be via the classical Ile44 hydrophobic patch of Ub, for which interaction remains to be demonstrated experimentally. In the future, more precise mutations in S1 or in S2 sub-sites, guided by structural work on ISG15- and di-Ub-bound SARS PLpro complexes, will allow for discrimination between the surfaces involved in Ub compared with ISG15 recognition, which will ultimately pave the way for determining the relative contributions of deubiquitination compared with deISGylation to coronavirus pathogenesis.
Our study fully supports the Ratia et al.  model and further shows that the unique di-distributive cleavage mechanism of SARS PLpro can probably result in the stabilization of not only di-UbLys48 species, but also of monoUb-conjugated substrates that accumulate as remnants following substrate deconjugation by SARS PLpro. These monoUb-conjugated substrates are stabilized when the remnant Ub on the substrate occupies the S1′ sub-site during the first cleavage event by SARS PLpro, leaving a poor substrate for SARS PLpro to recognize in subsequent cleavage events. In the present study, we also show that related USP-family DUBs and MERS PLpro do not share this feature of SARS PLpro to accumulate di-Ub units and monoUb-conjugated substrate cleavage products as they are mono-distributive DUBs. Although it is possible that, in vivo, SARS PLpro may ultimately remove all Ub molecules from its substrates, it is tempting to speculate that transiently stabilizing monoUb-conjugated substrates could have functional consequences during coronavirus infection. However, the exact in vivo substrates of SARS (and MERS) PLpro are currently ill-defined and the biological significance of this mode of di-Ub recognition remains to be elucidated. However, there are examples of human DUBs that appear to function similarly. The Machado–Josephin family cysteine protease DUB ataxin-3 has been shown to be a poor DUB overall, but has been demonstrated to edit longer Lys48- and Lys63-mixed chains by recognizing a Ub chain via its UIM (Ub-interacting motif) domains and cleaving proximal to the recognized elements . Additionally, a recently solved crystal structure of OTUD2 (a human DUB from the OTU (ovarian tumor protease) family) in complex with a Lys11-linked di-Ub substrate has revealed an extended Ub (S2)-binding surface in OTUD2 responsible for its Lys11-linked tri-Ub processing into a mono-Ub and di-Ub unit . The function of these additional Ub-recognizing domains in DUBs and how they contribute to biological pathways driven by each unique DUB has yet to be addressed.
MERS and SARS PLpro are primarily processing proteases that cleave the viral polypeptide at defined positions to release non-structural effector proteins [24,42]. Interestingly, the positional scanning library screen for both SARS  and MERS PLpro (the present study) faithfully recapitulated the nature of the sequences cleaved within the viral polypeptides , although these cleavage sites are not part of any Ubl domains in the viral polypeptide. Whether SARS and MERS PLpro have the capability to cleave Φ-Φ/+-G-G motifs (where Φ is a hydrophobic and + is a positively charged amino acid) in exposed loops of human proteins, pointing to potential additional virulence-related non-Ubl substrates in human cells, has yet to be determined. The contribution of SARS PLpro to the infectivity of SARS hCoV and the contribution of MERS PLpro towards MERS hCoV infectivity is well established, as the catalytic activity of these proteases is required for viral polypeptide processing . Thus, both viral DUBs are targets of intense drug development efforts in attempts to inhibit viral propagation. Recently, promising small-molecule inhibitors have been developed for SARS hCoV based on their ability to inhibit SARS PLpro; however, they do not inhibit MERS PLpro , suggesting that finding a common anti-coronaviral drug based on PLpros could be a difficult task. The in vivo testing of such compounds is also hampered by safety concerns using these highly pathogenic BSL-3 (Biosafety Level 3) viruses, thus the recent development of a proxy experimental system, using transgenic mice harbouring an engineered Sindbis virus (a BSL-2 organism) expressing SARS and MERS PLpros  is an extremely important step.
During viral infection, after viral genome translation and polypeptide expression, the host immune response must be suppressed to ensure timely viral particle assembly to propagate infection. It is accepted that the Ub/ISG15-deconjugating activity of these membrane-bound viral PLpros plays a role in promoting viral replication in the host by deconjugating ISG15- and Ub-conjugated proteins in the infected cell , contributing to down-regulation of innate immune functions . However, the identities of the Ub-/Ubl-conjugated substrates whose deconjugation by viral PLpros are essential for pathogenicity remain ill-defined. Whether they cleave Ub or Ubl conjugates from viral proteins (i.e., to spare them from degradation by the proteasome or remove the ISG15 mark that each newly synthesized protein receives ) or cellular substrates (i.e., to dampen the host immune response by deconjugating signalling molecules [21,25]), or both, within the vicinity of the membrane environment remains unknown, as most studies have been performed using overexpression systems relying on model pro-inflammatory substrates. It is also currently unknown to what extent deISGylation and/or deubiquitination activities contribute to pathogenesis. Thus, identifying viral PLpro mutants that can differentiate between Ub and ISG15 recognition and cleavage are of considerable interest. Interestingly, mutations in PLpros that affect global Ubl deconjugation, but not viral polypeptide processing, have been recently described [33,34], which could aid in the development of attenuated viruses for coronavirus vaccine development. It would also be of interest to determine whether the two PLpro enzymes can be functionally swapped between the two viruses to determine whether coronavirus infection merely requires an active processing protease/isopeptidase or if substrate specificity of that protease is a key factor. With reverse genetics systems for coronaviruses in place  that have been used to uncover the function of SARS proteins in viral pathogenesis , together with the rapid development of the full-length MERS hCoV genome as a bacterial artificial chromosome system , it may be possible to answer these questions in the near future.
Miklós Békés conceived the study, designed the research, carried out experiments (with the exception of the PS-SCL screen in Figure 1, which was carried out by Wioletta Rut and Paulina Kasperkiewicz under the supervision of Marcin Drag; whereas Monique Mulder and Huib Ovaa synthesized the di-UbLys48-ABP probe used in Figure 3C), analysed the data and wrote the paper. Christopher Lima and Tony Huang provided data analysis, experimental guidance and support.
We thank members of the Huang, Lima and Reinberg laboratories for helpful discussion and reagents and Erin Bekes for editorial support before submission.
This work was supported by the National Institute of General Medical Sciences [grant number F32GM100598 (to M.B.)]; the Foundation for Polish Science [FOCUS FNP grant (to M.D)]; the Polish Ministry of Science and Higher Education for the Faculty of Chemistry at Wroclaw University of Technology (to M.D.) the National Institute of General Medical Sciences [grant number GM065872 (to C.D.L); C.D.L. is an Investigator of the Howard Hughes Medical Institute; the National Institutes of Health [grant number GM084244 (to T.T.H.)]; and the American Cancer Society [grant number RSG-12-158-DMC (to T.T.H.)].
interferon-stimulated gene 15
lithium dodecyl sulfate
Middle East respiratory syndrome
Positional Scanning Substrate Combinatorial Library
severe acute respiratory syndrome
vinyl methyl ester