key: cord-0712610-4cgrjnjo
authors: Lei, Jian; Hilgenfeld, Rolf
title: Structural and mutational analysis of the interaction between the Middle-East respiratory syndrome coronavirus (MERS-CoV) papain-like protease and human ubiquitin
date: 2016-05-30
journal: Virol Sin
DOI: 10.1007/s12250-016-3742-4
sha: 5dd0f03bb34af58c469693cf5f3e311a80d804d4
doc_id: 712610
cord_uid: 4cgrjnjo

The papain-like protease (PL(pro)) of Middle-East respiratory syndrome coronavirus (MERS-CoV) has proteolytic, deubiquitinating, and deISGylating activities. The latter two are involved in the suppression of the antiviral innate immune response of the host cell. To contribute to an understanding of this process, we present here the X-ray crystal structure of a complex between MERS-CoV PL(pro) and human ubiquitin (Ub) that is devoid of any covalent linkage between the two proteins. Five regions of the PL(pro) bind to two areas of the Ub. The C-terminal five residues of Ub, RLRGG, are similar to the P5–P1 residues of the polyprotein substrates of the PL(pro) and are responsible for the major part of the interaction between the two macromolecules. Through sitedirected mutagenesis, we demonstrate that conserved Asp165 and non-conserved Asp164 are important for the catalytic activities of MERS-CoV PL(pro). The enzyme appears not to be optimized for catalytic efficiency; thus, replacement of Phe269 by Tyr leads to increased peptidolytic and deubiquitinating activities. Ubiquitin binding by MERS-CoV PL(pro) involves remarkable differences compared to the corresponding complex with SARS-CoV PL(pro). The structure and the mutational study help understand common and unique features of the deubiquitinating activity of MERS-CoV PL(pro). ELECTRONIC SUPPLEMENTARY MATERIAL: Supplementary material is available for this article at 10.1007/s12250-016-3742-4 and is accessible for authorized users.

To date, six coronaviruses infecting humans have been characterized. Infections with human coronaviruses (HCoVs) 229E (Hamre & Procknow, 1966) , OC43 (McIntosh et al., 1967) , NL63 (van der Hoek et al., 2004) , and HKU1 (Woo et al., 2005) cause relatively mild symptoms in most cases, whereas severe acute res-piratory syndrome coronavirus (SARS-CoV; Ksiazek et al., 2003; Kuiken et al., 2003; Peiris et al., 2003) and Middle-East respiratory syndrome coronavirus (MERS-CoV; Zaki et al., 2012) are connected with severe respiratory-tract infection and, in particular in case of MERS-CoV, acute renal failure (Eckerle et al., 2013) , leading to high case-fatality rates of ~10 and 35%, respectively. In spite of 13 years of research on SARS-CoV (Hilgenfeld & Peiris, 2013) , no approved drugs or vaccines are available for the treatment or prevention of coronavirus infection (Wang et al., 2016) . This is mainly due to the fact that although these emerging viruses have devastating effects on those infected, the absolute numbers of cases (~8000 for SARS, 1733 so far for MERS; (http://www.who.int)) imply that the de-velopment of specific antivirals is very likely not commercially viable. On the other hand, the global risk posed by MERS-CoV must not be underestimated. Since its discovery in September 2012, the number of MERS cases reported has been rising steadily, with some intermittent peaks connected to hospital outbreaks in Saudi Arabia (Assiri et al., 2013) . Man-to-man transmission of MERS-CoV has also been impressively demonstrated by the recent outbreak of MERS in South Korea, which was traced back to a single traveller from the Arab peninsula (Butler, 2015) . Therefore, it is imperative that academic laboratories help increase the preparedness against a possible MERS-CoV pandemic by characterizing antiviral drug targets and by identifying lead compounds interfering with them. In order to successfully infect humans, a virus has to meet at least two conditions: 1), it should maintain a sufficiently correct replication of its genetic material; 2), it should inhibit the host antiviral response. The papainlike protease (PL pro ) of MERS-CoV (or SARS-CoV) is involved in both of these tasks (Yang et al., 2013; Barretto et al, 2005) . The PL pro is a domain located in the middle part of the largest non-structural protein, Nsp3, of MERS-CoV (or SARS-CoV) . It is responsible for releasing Nsp1, Nsp2, and Nsp3 from the polyproteins 1a (pp1a) and 1ab (pp1ab), an essential step of replication (Harcourt et al., 2004) . Like its SARS-CoV counterpart, the MERS-CoV PL pro also has deubiquitinating (DUB) and deISGylating activities in vivo as well as in vitro (Yang et al., 2013; Mielech et al., 2014; Lei et al., 2014; Baez-Santos et al., 2014b) . K48-and K63-polyubiquitin poly (Ub) and ISG15 (interferon-stimulated gene 15)conjugated targets are usually involved in host innate immune regulation (Liu et al., 2005; Maringer & Fernandez-Sesma, 2014) . The PL pro has the ability to digest K48-and K63-linked polyUb chains and to remove ISG15 from ISG15-linked proteins (Baez-Santos et al., 2014b) , thereby interrupting the signalling pathways leading to the innate immune response. Thus, the PL pro can block the activation of IFN regulatory factor 3 (IRF3) (Yang et al., 2013) and subsequently the production of interferon β (IFNβ) (Mielech et al., 2014) . Interestingly, MERS-CoV PL pro shows a similar cleavage rate for K48-and K63-linked polyUb chains, while the SARS-CoV enzyme prefers K48-over K63-linked chains (Baez-Santos et al., 2014b) . The former enzyme degrades a polyUb chain by removing mono-Ubs, whereas the latter cleaves di-Ub units off the polyUb chain (Bekes et al., 2015) .

We have reported the first crystal structure of the MERS-CoV PL pro (Lei et al., 2014) . Later, two other groups also described the structure of this enzyme (Bailey-Elkin et al., 2014; Lee et al., 2015) . The structure of PL pro can be divided into two parts: a ubiquitin-like (Ubl) domain and a catalytic domain with thumb, palm, and fingers subdomains. The overall fold of MERS-CoV PL pro is not only similar to that of SARS-CoV PL pro , but also to that of several human ubiquitinspecific proteases (USPs) (Hu et al., 2005) . In 2014, the X-ray structure of the complex of SARS-CoV PL pro with ubiquitin has been reported (Chou et al., 2014; Ratia et al., 2014) . Several key residues (such as Glu168 or Tyr265) of SARS-CoV PL pro that are important for ubiquitin recognition (Chou et al., 2014; Ratia et al., 2014) , are not conserved in MERS-CoV PL pro . Bailey-Elkin et al. (2014) described the structure of an artificially linked, covalent complex between ubiquitin and MERS-CoV PL pro . Here, we present the crystal structure of a noncovalent complex between the two proteins and a mutational study of the interactions involved. For these studies, we used the Cys111Ser active-site variant of MERS-CoV PL pro .

The PL pro of MERS-CoV (strain 2c EMC/2012; Gen-Bank no. AFV09327.1) contains 320 residues, from Gln1482 to Asp1801 of pp1a/1ab. For simplification, Gln1482 was renumbered into Gln1 here. The DNA plasmid coding for MERS-CoV PL pro was produced earlier (Lei et al., 2014) .

The MERS-CoV PL pro C111S, D164E, D164A, D165E, D165A, and F269Y variants were produced using the same strategy that we described before (Lei et al., 2014) . All primers for these variants are listed in Supplemental  Table S1 . All DNA plasmids coding for the altered PL pro were verified by sequencing.

Genes coding for wild-type (WT) MERS-CoV PL pro and for its variants were expressed and the corresponding proteins were purified according to our previous description (Lei et al., 2014) .

performed according to the procedure described for MERS-CoV PL pro (Lei et al., 2014) .

Purified MERS-CoV PL pro (WT) and a variant that had the active-site Cys111 replaced by Ser (PL pro (C111S)) were both concentrated to ~24 mg/mL in 20 mmol/L Tris-HCl, 150 mmol/L NaCl, pH 8.8, 10 mmol/L β-mercaptoethanol (BME). Human ubiquitin (BostonBiochem) was dissolved to 6 mg/mL in 20 mmol/L Tris-HCl, 150 mmol/L NaCl, pH 8.8. 500 μL PL pro (WT) or PL pro (C111S) were mixed with 500 μL ubiquitin (~1:1 molar ratio) at 4°C

overnight. (C111S)-Ub were observed under condition No. 9 of MD1-01, whereas no crystal of PL pro (WT)-Ub was obtained. Optimized crystals of PL pro (C111S)-Ub were subsequently obtained within two days under the condition: 22% w/v PEG 4000, 15% v/v 2-propanol, 0.1 mol/L tri-sodium citrate pH 4.8, and 10% glycerol. 2 μL of protein and 2.5 μL of reservoir were mixed to equilibrate against 500 μL reservoir.

Crystals were placed in a nitrogen-gas stream (100 K). A 3.16-Å dataset was collected at wavelength 0.91841 Å at beamline 14.2 of BESSY, Berlin (Mueller et al., 2012) . The program XDS (Kabsch, 2010) was used to process the diffraction data. The space group was found to be P6 3 , with unit-cell parameters a = b =138.14 Å, c = 57.59 Å, γ = 120°. Diffraction data statistics are shown in Table 1 .

The structure of the MERS-CoV PL pro (C111S)-Ub complex was solved by molecular replacement using MOLREP (Vagin & Teplyakov, 2010) . The program selected the MERS-CoV PL pro (Protein Data Bank (PDB) entry 4P16, Lei et al., 2014) as the first search model. Human ubiquitin (PDB entry: 1UBQ, Vijay- Kumar et al., 1987) was used as the second search model. The model of the complex was inspected and rebuilt using Coot (Emsley et al., 2010) , and refined using phenix.refine (Headd et al., 2012) . The refinement statistics are shown in Table 1 Kinetic assays of purified PL pro s All enzymatic assays were performed using a 96-well microtiter plate and the reaction buffer 20 mmol/L Tris-HCl, 150 mmol/L NaCl, pH 7.9, 2 mmol/L dithiothreitol (DTT). The fluorogenic substrates Cbz-Arg-Leu-Arg-Gly-Gly-7-amino-4-methylcoumarin (Z-RLRGG-AMC) (Bachem), Z-LRGG-AMC (BostonBiochem), and ubiquitin-AMC (Ub-AMC) (BostonBiochem) were used. The fluorescence of free AMC with different concentrations (5 nmol/L-2.5 μmol/L) in reaction buffer was measured to generate a calibration curve, in order to convert the change of fluorescence intensity per unit of time, Δ(AFU) /s, into the amount of hydrolyzed substrate in μmol/L/s. The enzymatic cleavage reactions were run with an Flx800 fluorescence spectrophotometer (BioTek), to measure the increased fluorescence signal (λ ex : 360 nm; λ em : 460 nm) resulting from AMC release. Reactions were initiated by addition of the proteases to the reaction system. The peptide-hydrolysis kinetic assays were performed with the following conditions: 1 μmol/L MERS-CoV PL pro variant (D164E, D164A, D165E, F269Y), or 10 μmol/L D165A, or 0.1 μmol/L SARS-CoV PL pro , with different concentrations (10, 20, 40, 80, 100 μmol/L) of Z-RLRGG-AMC or Z-LRGG-AMC in a final volume of 100 μL at 25 °C. The kinetic curves for the proteases and their variants with the substrates Z-RLRGG-AMC or Z-LRGG-AMC were linear and the initial velocities also increased linearly with substrate concentration. No saturation was observed. Therefore, the data were fitted to the equation v/[E] tot. = k app [S] , where k app approximates k cat /K M , as described previously (Barretto et al., 2005; Wojdyla et al., 2010) .

The deubiquitinating kinetic assays were performed under the following conditions: 0.1 μmol/L MERS-CoV PL pro wild-type or its variants D164E, D164A, D165E, F269Y, or 0.5 μmol/L D165A, or 0.025 μmol/L SARS-CoV PL pro were incubated with increasing concentrations (1, 2, 4, 8 μmol/L) of Ub-AMC in a final volume of 50 μL, at 25 °C. Although PL pro actually cleaves the isopeptide bond between the carboxyl group of the C-terminal Gly in Ub and the ε-amino group of Lys in ubiquitinated targets in vivo, we used the hydrolysis of Ub-AMC here to test the deubiquitinating activity in vitro. The kinetic curves of proteases and variants with the substrate Ub-AMC were hyperbolic and the initial velocities were not linear over the concentration of substrate. However, saturation was still not observed within a reasonable time. Only when the ratio of protein to substrate was 1:1 or larger, were we able to achieve saturation within a limited time (data not shown). As the initial velocities did not increase in a strictly linear fashion with substrate concentration, application of the equation v/[E] tot. = k app [S] to mimic k cat /K M would lead to large standard errors. We were however able to fit the data to the Michaelis-Menten equation using the GraphPad Prism program (GraphPad Software), even though saturation could not be reached (Supplemental Figure 1 ). All assays were performed in duplicates.

The substrate-binding site of MERS-CoV PL pro features significant differences from those of the corresponding SARS-CoV enzyme and human ubiquitin-specific proteases (USPs, such as, USP14) (Hu et al., 2005; Chou et al., 2014; Ratia et al., 2014) . It is therefore of interest to determine the crystal structure of the complex between MERS-CoV PL pro and its substrate, human ubiquitin. Hence, we crystallized the ubiquitin (Ub) complex of a MERS-CoV PL pro variant that had the active-site Cys111 replaced by serine (C111S) and determined the structure at 3.16 Å ( Figure 1A ).

There is one PL pro (C111S)-Ub complex per asymmetric unit. Using the PDBePISA server (Krissinel & Henrick, 2007) , the total interface region of MERS-CoV PL pro (C111S)-Ub was determined as 813 Å 2 , close to the 915 Å 2 interface of SARS-CoV PL pro -Ubal (ubiquitin aldehyde; PDB entry: 4MM3, Ratia et al., 2014) and the 999 Å 2 of SARS-CoV PL pro (C112S)-Ub (PDB: 4M0W, Chou et al., 2014) , but less than the 1503 Å 2 observed for USP14-Ubal (PDB: 2AYO, Hu et al., 2005) . The overall structure of MERS-CoV PL pro (C111S) is very similar to that of the substrate-free PL pro (Lei et al., 2014) , with a root-mean-square deviation (RMSD) of 0.91 Å for corresponding Cα atoms between these two structures. Nevertheless, two differences are immediately visible: 1), the zinc-finger motif has moved and closed in onto the Ub, compared to the free PL pro . The zinc ion position has shifted by about 4 Å and the largest deviation between the two structures is ~6 Å for Cys228 of the zinc-finger region ( Figure 1B) ; 2) the mobile loop 271 GIETAVG 277 , also named "BL2 loop", is defined by clear main-chain electron density ( Figure 1C ). This loop is disordered in substrate-free PL pro (Lei et al., 2014; Bailey-Elkin et al., 2014) . Compared to MERS-CoV PL pro , the BL2 loop 267 GNYQCG 272 is shorter by one residue in SARS-CoV PL pro .

At the time when we deposited in the PDB the coordinates for the crystal structure of the non-covalent complex MERS-CoV (C111S)-Ub (PDB: 4WUR), Bailey-Elkin et al. (2014) described two crystal structures for a covalent complex MERS-CoV PL pro -Ub, in which an alkyl bromide group introduced at the C-terminus of Ub had formed a thioether with the active-site Cys111 of MERS-CoV PL pro . These structures were in space groups P6 5 22 and P6 3 and were named "closed" and "open" PL pro -Ub complexes (PDB: 4RF0 and 4RF1), respectively (Bailey-Elkin et al., 2014) . Even though the zinc-finger motif of PL pro in the former complex (space group P6 5 22) is closer to the Ub than in the P6 3 complex, both the "closed" and "open" complex show almost the same interactions between PL pro and Ub (Bailey-Elkin et al., 2014). The differences may be caused by crystal packing. Our non-covalent complex is similar to the "open" form of the covalent PL pro -Ub complex (RMSD = 0.58 Å for the PL pro and 0.85 Å for Ub, based on all Cα atoms).

All parts of the PL pro except for the ubiquitin-like (Ubl) domain interact with Ub. Most interactions in-volve five surface regions of the PL pro and two regions of Ub ( Figure 2A ). These five regions of the PL pro are labelled by Roman numbers: I, Leu106-Tyr112; II, Ala162-Arg168; III, Cys208-Val210; IV, Gly248-Pro250; V, Phe269-Tyr279. I and II are situated in the thumb domain; III is in the fingers domain; and IV and V are in the palm domain ( Figures 1A and  2A ). In the following and in the figure labels, residues of Ub are indicated in italics to distinguish them from residues of PL pro . The two interacting regions of Ub are: A, Arg42-Gln49; B, Arg72-Gly76 (Figure 2A ). Region A of Ub consists of β3, β4, and the loop between β3 and Structure of MERS-CoV PL pro -ubiquitin complex β4 ( Figure 2B ). Regions II and III of PL pro interact with region A of Ub. Region B comprises the five C-terminal residues, RLRGG, and is in contact with regions I, II, IV, and V of PL pro ( Figure 2C ). The C-terminal RLRGG motif contributes the majority of the interactions with the PL pro ; the buried surface between Ub region B and PL pro is 477 Å 2 (out of a total of 813 Å 2 ). In order to make these interactions clear, we describe here in some detail the contacts between the PL pro and regions A and B of Ub.

Region A (Arg42-Gln49) of Ub inserts into the space between the thumb and fingers domains of PL pro (Figures 2A-B) . Residues Ile44, Ala46, and Gly47 engage in hydrophobic interactions with Tyr209 and Val210 of region III of PL pro ( Figure 2B ). The hydrophobic patch (Ile44, Ala46, and Gly47) is a common interaction region utilized by Ub-binding proteins (Dikic et al., 2009) . In particular, the interaction of Ile44 with Val210 of PL pro is important for the deubiquitinase (DUB) but not for the protease activity. The variant V210R shows dramatically reduced DUB activity, as demonstrated by Bailey-Elkin et al. (2014) (according to the numbering scheme of these authors, V210 is V1691).

A salt-bridge exists between region A of Ub and the PL pro ( Figure 2B ), namely between the side-chains of Arg42 and Asp165 (region II in PL pro ). In addition, a hydrogen bond is formed between the main-chain O atom of Gly47 and the main-chain amide of Val210 (region III). In the SARS-CoV PL pro (C112S)-Ub complex (Chou et al., 2014) , Arg42 forms a salt-bridge with the negatively charged Glu168, a residue which is replaced by the positively charged Arg168 in MERS-CoV PL pro . Consequently, the same salt-bridge cannot be formed in the MERS-CoV PL pro (C111S) -Ub complex. Instead, MERS-CoV PL pro has Asp165 interacting with Arg42. In our deubiquitinating (DUB) kinetic assays, the D165A variant shows a dramatically reduced DUB activity; the k cat /K M is about 78-fold decreased compared to that of the wild-type ( Table 2 ). The K M value of the D165A variant is about 4-fold higher than that of the wild-type MERS-CoV PL pro , suggesting that this amino-acid replacement reduces the Ub binding affinity. Meanwhile, the D165E amino-acid replacement shows a catalytic efficiency comparable to wild-type towards Ub-7-amino-4methylcoumarin (Ub-AMC) ( Table 2 ). However, we noticed that the K M value of D165E for the DUB activity is about 2-fold larger than for the WT enzyme. Although Glu165 can mimic Asp165 here, we propose that the longer side-chain of Glu may fit less perfectly compared to Asp. These results indicate that the salt-bridge between Asp165 and Arg42 could be important for the PL pro 's DUB activity, in addition to the interaction between Asp165 and the P4-amide group (see below).

Region B of Ub comprises the five C-terminal residues, RLRGG. These five residues bind to the narrow activesite channel between the thumb and palm domains of the PL pro ( Figure 2C ). They mainly interact with regions I, II, IV, and V of the protease. Residues RLRGG are compatible with the PL pro cleavage motif, (R/K) (L/I) XGG (P5-P1), in the MERS-CoV polyproteins; their interactions with the PL pro are discussed here in terms of subsites S1 to S5. S1 and S2 subsites. Pro163 (region II of PL pro ) and the side-chains of Asn109 and Tyr112 (located in region I) form a space-restricted S1 site to accommodate Gly76 (P1). The carbonyl oxygen atom of Gly277 (region V) accepts a hydrogen bond from the amide of Gly76 (Figure 3) . The side-chains of Tyr112 (region I) and Phe269, Val276 as well as Tyr279 (region V), and Gly277 form a restricted space for Gly75 (P2). Two hydrogen bonds link the main-chain at Asp164 of the PL pro and Gly75 (Figure 3) . The S1 and S2 sites of the protease are too small to accommodate any other residue but glycine.

S3 subsite. The main-chain O atom of Arg74 (P3) accepts a hydrogen bond from the Gly277 amide. In the complexes SARS-CoV PL pro (C112S)-Ub or SARS-CoV PL pro -Ubal, the main chain at Arg74 forms two hydrogen bonds (Chou et al., 2014; Ratia et al., 2014) , namely with the amide of Gly272 and with the hydroxyl group of Tyr265. The former hydrogen bond is conserved in the MERS-CoV PL pro (C111S)-Ub complex, but the latter is not. Tyr265 of SARS-CoV is replaced by Phe269 in MERS-CoV, which lacks the ability to form a hydrogen bond with the main-chain amide of Arg74. In agreement with this difference, the DUB activity of SARS-CoV PL pro is 2.5-fold higher than that of MERS-CoV PL pro in our kinetic assay; furthermore, the MERS-CoV PL pro F269Y amino-acid replacement leads to enhancements by about 1.5-, 1.4-, and 1.9-fold of the hydrolytic activities towards carbobenzoxy-Arg-Leu-Arg-Gly-Gly-7amino-4-methylcoumarin (Z-RLRGG-AMC), Z-LRGG-AMC, and Ub-AMC, respectively ( Table 2) .

The side-chain of Arg74 is exposed to the solvent in our MERS-CoV PL pro (C111S)-Ub complex. This sidechain shows remarkable variability in the interactions it makes in the different complexes. In the "open" but not in the "closed" covalent MERS-CoV PL pro -Ub complex (Bailey-Elkin et al., 2014) , it donates a hydrogen bond to the main-chain carbonyl oxygen of Thr1755 of the BL2 loop (corresponding to Thr274 in our numbering scheme). In SARS-CoV PL pro -Ubal (Ratia et al., 2014) but not in the SARS-CoV PL pro (C112S)-Ub complex (Chou et al., 2014) , the side-chain of Arg74 forms a hydrogen bond with the main-chain carbonyl oxygen of Gln270. Instead, Arg74 is involved in a relatively weak salt-bridge with Glu162 in the SARS-CoV PL pro (C112S)-Ub complex (Chou et al., 2014) . None of these interactions exist in our MERS-CoV PL pro -Ub complex.

S4 subsite. The main-chain amide of Leu73 (P4) donates a hydrogen bond to the side-chain of Asp165 (region II) (Figure 3 ). This hydrogen bond is conserved in all the complexes compared here. In our peptide-cleavage assay, the D165A variant shows about 16-fold and 10-fold lower activities towards substrates Z-RLRGG-AMC and Z-LRGG-AMC, respectively, indicating that Asp165 is not only important for interacting with Arg42 of Ub (see above). The side-chain of Leu73 is embedded in a hydrophobic pocket which is formed by the Cβ atom of Asp165, the side-chain of Pro250 (region IV), Phe269, as well as the Cβ and Cγ atoms of Glu273 (region V). Asp165 and Pro250 are conserved in SARS-CoV PL pro (Asp165 and Pro249). Phe269 is replaced by Tyr265, and Glu273, situated in the BL2 loop, is replaced by Tyr269. However, the side-chains of all these non-conserved residues possess the ability to provide a hydrophobic environment to accommodate Leu73. S5 subsite. The side-chain of Arg72 (P5) is located between the PL pro thumb domain and region A of ubiquitin. It forms a salt-bridge with the side-chain of Asp164 in MERS-CoV PL pro . In addition, the guanidinum group of Arg72 may be involved in a π-π interaction with that of Arg168 (Figure 3 ). These interactions have also been described for the covalent complex of MERS-CoV PL pro with Ub (Bailey-Elkin et al., 2014) . Arg72 is not subject to strict space limitations; accordingly, this residue displays different binding patterns in the two SARS-CoV residues are shown in purple in the ball-and-stick style, and they are labeled in red. For clarity, the 2F o -F c electron density (blue; 1.0 σ) of the RLRGG main chain is shown. PL pro residues are displayed in green in the balland-stick style, and they are labeled in black. Hydrogen bonds are displayed as black dashed lines, and the salt-bridge between R72 and D164 is depicted as two red dashed lines.

Structure of MERS-CoV PL pro -ubiquitin complex PL pro -Ub complexes. In the covalent SARS-CoV PL pro -Ubal complex (Ratia et al., 2014) , Arg72 forms a saltbridge with Glu168. In the non-covalent SARS-CoV PL pro (C112S)-Ub complex (Chou et al., 2014) , Arg72 is exposed to the solvent and does not interact with Glu168 (Arg42 instead forms a salt-bridge with Glu168, as mentioned above). In our kinetic assay, the D164A variant of MERS-CoV PL pro displays a ~4.5-fold and a ~3.5-fold reduced activity, respectively, for Z-RLRGG-AMC and Ub-AMC (Table 2) . For Z-LRGG-AMC, the activity is decreased just a little (by about ~1.5-fold), because there is no P5-Arg in this substrate (Table 2 ). These data demonstrate that Asp164 is important for the interaction with Arg72.

In summary, the main-chain heteroatoms of P5-P1 form a hydrogen-bonding network with PL pro . The binding characteristics of P1, P2, and P4 are conserved in all MERS-CoV and SARS-CoV PL pro -Ub complexes. However, P3-Arg and P5-Arg assume binding patterns that differ between the various MERS-CoV and SARS-CoV PL pro -Ub complexes.

Viral proteins are likely to possess multiple functions, as exemplified by non-structural protein 1 (NS1) of influenza A viruses (Hale et al., 2008) , the nucleocapsid (N) protein of coronaviruses (Chang et al., 2014) , or the Nsp14 exonuclease-guanyl-7-methyltransferase of coronaviruses (Chen et al., 2009) . Exhibiting DUB and proteolytic activities, the papain-like proteases of coronaviruses are no exception here. Although the overall folds are conserved, the enzyme activity and substrate-binding modes of CoV PL pro s differ in detail. Therefore, no coronavirus PL pro can be considered a general model for all its homologues (Baez-Santos et al., 2014b) .

We have previously noticed that the oxyanion hole of the MERS-CoV PL pro appears to be deficient (Lei et al., 2014 ; also see the discussion below). Similarly, the recognition of ubiquitin by the enzyme appears to be suboptimal. Thus, the main-chain amide of the P3-Arg residue has no hydrogen-bonding partner on the MERS-CoV PL pro , because the near-by side-chain of Phe269 is incapable of accepting an H-bond. The corresponding residue is Tyr265 in the SARS-CoV PL pro , which is perfectly positioned to accept the hydrogen bond from the P3-amide. Indeed, when we replaced Phe269 by Tyr in MERS-CoV PL pro , the DUB activity of the enzyme increased by a factor of almost 2 and the peptidolytic activity by ~1.5. The evolution of viral enzymes is not neces-sarily driven by optimization of catalytic efficiency. This is particularly true for viral proteases that have to ensure the availability of non-structural proteins in the correct temporal order when they cleave them out of the viral polyproteins; in fact, too rapid a polyprotein processing might be counterproductive. On the other hand, a more efficient DUB activity should help the virus in counteracting the innate immune response of the host cell. As the binding of the P5-P1 residues of ubiquitin and of the polyprotein cleavage sites obviously influences both the DUB and proteolytic activities of the PL pro , the suboptimal catalytic efficiency that we observe may be a consequence of a compromise between the requirements of the two activities.

With regard to the oxyanion hole, we previously proposed that the backbone amide of Asn109 (located in a β-turn connecting β7 and α4, Figure 4 ) may contribute to the stabilization of the oxyanion intermediate in PL pro catalysis, along with the main-chain amide of the active site Cys111, although this may require a slight rearrangement of this β-turn (Lei et al., 2014) . In our non-covalent MERS-CoV PL pro -Ub complex, we do not see any rearrangement of the β-turn. More or less in agreement with our suggestion, Bailey-Elkin et al. (2014) proposed on the basis of their covalent MERS-CoV PL pro -Ub complex structure that the main-chain amides of Asn1590, Asn1591, and Cys1592 (corresponding to Asn109, Asn110, and Cys111 in our numbering scheme) form the oxyanion hole. On the other hand, Lee et al. (2015) argued that the side-chain of Asn109 could contribute to the oxyanion hole. They found that the N109A replacement completely abolished enzyme activity. As we reported earlier (Lei et al., 2014) , the side-chain amide of Asn109 makes a strong hydrogen bond with the conserved Gly161 (Figure 4 ). Any reorientation of the Asn109 side-chain towards the oxyanion would require a disruption of this strong interaction; this is not very likely. In conclusion, lacking a side-chain in the proper spatial orientation and capable of donating a hydrogen bond to the oxyanion (such as Trp107 in SARS-CoV PL pro ), the oxyanion hole of MERS-CoV PL pro seems to be deficient.

Apart from the less than optimum binding of the P3 residue, there are other differences in the way MERS-CoV PL pro and SARS-CoV PL pro recognize human ubiquitin. The formation of a salt-bridge between Arg72, the P5 side-chain of Ub, and Asp164 is unique for MERS-CoV PL pro . The same interaction exists in the two covalent MERS-CoV PL pro -Ub complexes (Bailey-Elkin et al., 2014) . This binding mode is very different from the Glu168 -Arg72 salt-bridge in the SARS-CoV PL pro -Ubal complex (Ratia et al., 2014) . Glu168 is conserved in HCoV-NL63 PL2 pro and replaced by Asp in the HCoV-229E, HCoV-OC43, and HCoV-HKU1 PL2 pro s (for sequence alignments, see Barretto et al., 2005; Baez-Santos et al., 2014b) , so the same type of interaction is likely to be realized in the Ub complexes of these enzymes. However, the corresponding residue in MERS-CoV PL pro is Arg168; hence, Asp164 is used instead to bind Arg72. Asp164 is in fact unique in MERS-CoV. It is replaced by Gly in SARS-CoV PL pro and the PL2 pro s of HCoV NL63 and HCoV 229E, by Ala in HCoV-OC43 PL2 pro , and by Ser in HCoV-HKU1 PL2 pro . Our kinetic results for the D164A replacement (see Results) emphasize the importance of the unique Asp164 residue.

In our PL pro (C111S)-Ub complex, the Ubl domain of PL pro shows no interaction with ubiquitin. The relative orientation of the Ubl domain in the substrate-bound PL pro is the same as in substrate-free PL pro . The Ubl domain of MERS-CoV PL pro is not required for the IFN antagonism activities (Baez-Santos et al., 2014b), but it is required in case of SARS-CoV PL pro (Frieman et al., 2009) . Recently, Mielech et al. (2015) reported that the Ubl domain of mouse hepatitis virus (MHV) is an important modulator of PL pro stability and viral pathogenesis. Although the Ubl domain shows variable effects in different CoVs, the high degree of conservation of the domain throughout the family suggests that it may play a common biological role. One possible function is that the Ubl might be involved in protein-protein interactions. Ubiquitin-like domains are known to function as binding modules in such interactions. For example, the kinase Raf contains a Ubl domain for interaction with human Ras (Fetics et al., 2015) . Also, human ubiquitin-specific protease (USP) 7 includes five Ubl (1-5) domains, of which the second is bound by the Herpes simplex virus-1 immediate-early protein ICP0 to counteract the intrinsic antiviral response of the host cell (Pfoh et al., 2015) .

The BL2 loop of SARS-CoV PL pro is important for binding inhibitors (Baez-Santos et al., 2014a; Lee et al., 2015) . This loop is variable in different CoV PL pro s. A potent inhibitor of the SARS-CoV PL pro , N-[(3-fluorophenyl) methyl]-1-[(1R)-1-naphthalen-1-ylethyl] piperidine-4-carboxamide (compound 3k, IC 50 = 0.15 ± 0.01 μmol/L) was found to have no effect on the MERS-CoV enzyme (Baez-Santos et al., 2014b) . Tyr269 and Gln270 of SARS-CoV PL pro , which are important for binding this inhibitor (Baez-Santos et al., 2014a) , are replaced by Glu273 and Ala275 in the MERS-CoV protease. It seems that this structural difference in the BL2 loop has a remarkable impact on the effectiveness of the inhibitor. The structure of the MERS-CoV PL pro (C111S)-Ub complex presented here will facilitate virtual screening of chemical libraries for specific anti -MERS-CoV PL pro inhibitors (Hilgenfeld, 2014) .

The di-Ub site of PL pro SARS-CoV and MERS-CoV PL pro s can digest K48-and K63-linked (poly)ubiquitin chains and remove ISG15 from proteins covalently linked to it (Ratia et al., 2014 , Baez-Santos et al., 2014b . SARS-CoV PL pro prefers binding of K48-Ub 2 and ISG15 over mono-ubiquitin (Ratia et al., 2014) . All evidence suggests that on the PL pro , at least two ubiquitin-binding sites (or a binding site for diubiquitin-like molecules such as ISG15) exist. Ratia et al. (2014) proposed hypothetic models for complexes of SARS-CoV PL pro with K48-Ub 2 and ISG15. These authors identified two major hydrophobic binding sites on the PL pro for the first Ub (Ub1) and the second Ub (Ub2). The binding site for Ub1 comprises Met209, Pro248, and Pro249. This hydrophobic patch is conserved in MERS-CoV PL pro , although the residues are not exactly the same. Met209 of SARS-CoV PL pro is replaced by Val210 (region III) in MERS-CoV. Pro248 is replaced by Thr249, whereas Pro249 is conserved (Pro250; region V) ( Figure 5 ). The hypothetic second binding site for Ub2 (also named "ridge" region, Ratia et al., 2014) is located to the first α helix (α2) in the thumb domain, including residues Phe70, His74, and Leu76 of SARS-CoV PL pro . According to a structural alignment of the SARS-CoV and MERS-CoV PL pro s, Phe70, His74, and Leu76 are changed to Lys69, Gly73, and Val75, respectively ( Figure 5 ). In SARS-CoV PL pro , the F70S and F70A replacements lost the affinity to K48-Ub 2 and ISG15 in vitro (Ratia et al., 2014) ; therefore, the presence of Lys69 in MERS-CoV PL pro instead of Phe70 strongly suggests that this enzyme should bind Ub2 in a different way. Bailey-Elkin et al. (2014) predicted Asn1673 and Val1674 of the fingers subdomain (corresponding to Asn192 and Val193 in our PL pro ) as the distal Ub site of K63 di-Ub, but they found that their DUB activity data do not support their prediction for the Ub2 binding site. The structure of the MERS-CoV PL pro -Ub complex reported here reveals the exact Ub1 binding site on the PL pro , but the Ub2 binding site should be identified by crystallizing the PL pro in complex with poly-or di-ubiquitin.

In summary, the crystal structure of the MERS-CoV PL pro -Ub complex provides valuable information that helps understand the multiple functions of coronavirus papain-like proteases. Mutational studies additionally highlight features of the MERS-CoV PL pro . The different substrate-binding patterns should be kept in mind when designing inhibitors for PL pro s of CoVs, even though the overall structures of these enzymes are conserved. Furthermore, it would be helpful to obtain the structure of PL pro in complex with di-Ub or ISG15 in the future. ing sites are depicted as purple dots. The N and C termini of PL pro are marked by underlined letters. All residues related to the two Ub binding sites are labeled. The catalytic triad Cys111-His278-Asp293 is indicated by yellow, blue, and red spheres.

KSA MERS-CoV Investigation Team. 2013. Hospital outbreak of Middle East respiratory syndrome coronavirus

Xray structural and biological evaluation of a series of potent and highly selective inhibitors of human coronavirus papain-like proteases

Catalytic function and substrate specificity of the PLpro domain of nsp3 from the Middle East Respiratory Syndrome Coronavirus (MERS-CoV)

Crystal structure of the Middle East respiratory syndrome coronavirus (MERS-CoV) papain-like protease bound to ubiquitin facilitates targeted disruption of deubiquitinating activity to demonstrate its role in innate immune suppression

The papain-like protease of severe acute respiratory syndrome coronavirus has deubiquitinating activity

SARS hCoV papain-like protease is a unique Lys 48 linkage-specific di-distributive deubiquitinating enzyme

South Korean MERS outbreak spotlights lack of research

The SARS coronavirus nucleocapsid protein-forms and functions

MolProbity: all-atom structure validation for macromolecular crystallography

Functional screen reveals SARS coronavirus nonstructural protein nsp14 as a novel cap N7 methyltransferase

Structural basis for catalysis and ubiquitin recognition by the severe acute respiratory syndrome coronavirus papain-like protease

Ubiquitin-binding domains-from structures to functions

Identification of a novel coronavirus in patients with severe acute respiratory syndrome

In-vitro renal epithelial cell infection reveals a viral kidney tropism as a potential mechanism for acute renal failure during Middle East Respiratory Syndrome (MERS) coronavirus infection

Features and development of Coot

Allosteric effects of the oncogenic RasQ61L mutant on Raf-RBD

Severe acute respiratory syndrome coronavirus papain-like protease ubiquitin-like domain and catalytic domain regulate antagonism of IRF3 and NF-κB signaling

The multifunctional NS1 protein of influenza A viruses

A new virus isolated from the human respiratory tract

Identification of severe acute respiratory syndrome coronavirus replicase products and characterization of papain-like protease activity

Use of knowledge-based restraints in phenix.refine to improve macromolecular refinement at low resolution

From SARS to MERS: Crystallographic studies on coronaviral proteases enable antiviral drug design

From SARS to MERS: 10 years of research on highly pathogenic human coronaviruses

Structure and mechanisms of the proteasome-associated deubiquitinating enzyme USP14

Inference of macromolecular assemblies from crystalline state

A novel coronavirus associated with severe acute respiratory syndrome

Newly discovered coronavirus as the primary cause of severe acute respiratory syndrome

Inhibitor recognition specificity of MERS-CoV papain-like protease may differ from that of SARS-CoV

Crystal structure of the papain-like protease of MERScoronavirus reveals unusual, potentially druggable active-site features

Immunity by ubiquitylation: a reversible process of modification

Message in a bottle: lessons learned from antagonism of STING signalling during RNA virus infection

Growth in suckling-mouse brain of "IBV-like" viruses from patients with upper respiratory tract disease

MERS-CoV papain-like protease has delSGylating and deubiquitinating activities

Murine coronavirus ubiquitin-like domain is important for papain-like protease stability and viral pathogenesis

Facilities for macromolecular crystallography at the Helmholtz-Zentrum Berlin

Coronavirus as a possible cause of severe acute respiratory syndrome

Crystal structure of USP7 ubiquitin-like domains with an ICP0 peptide reveals a novel mechanism used by viral and cellular proteins to target USP7

Structural basis for the ubiquitin-linkage specificity and deISGylating activity of SARS-CoV papain-like protease

Identification of a new human coronavirus

Structure of ubiquitin refined at 1.8 Å resolution

Recent progress in the discovery of inhibitors targeting coronavirus proteases

On the use of the merging R factor as a quality indicator for X-ray data

Papain-like protease 1 from transmissible gastroenteritis virus: crystal structure and enzymatic activity toward viral and cellular substrates

Characterization and complete genome sequence of a novel coronavirus, coronavirus HKU1, from patients with pneumonia

Proteolytic processing, deubiquitinase and interferon antagonist activities of Middle East respiratory syndrome coronavirus papain-like protease

Isolation of a novel coronavirus from a man with pneumonia in Saudi Arabia

Technical assistance by Susanne Zoske is gratefully acknowledged. We thank Stefan Anemüller for discussion. We also acknowledge access to beamline BL14.2 operated by the Helmholtz-Zentrum Berlin at the BESSY II electron storage ring (Berlin-Adlershof, Germany). This work was supported by the European Commission through its "SILVER" project (contract no. HEALTH-F3-2010-260644) and by the German Center for Infection Research (DZIF). RH acknowledges support by the DFG Cluster of Excellence "Inflammation at Interfaces" (EXC 306)

This article does not contain any studies with human or animal subjects. Both authors declare no competing interest.

JL and RH designed all experiments. JL performed the experiments. JL and RH analyzed all data. JL and RH wrote the manuscript.Supplementary figure and table are available on the websites of Virologica Sinica: www.virosin.org; link. springer.com/ journal/12250.