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

Background: MERS-CoV papain-like protease (PLpro) processes viral polyproteins and has deubiquitinating activity. Results: A crystal structure of MERS-CoV PLpro bound to ubiquitin guided mutagenesis to disrupt PLpro deubiquitinating activity without affecting polyprotein cleavage. Conclusion: The deubiquitinating activity of MERS-CoV PLpro suppresses the induction of interferon-β expression. Significance: Our strategy to selectively disable PLpro deubiquitinating activity enables the study of its specific functions in infection.

The Middle East respiratory syndrome coronavirus (MERS-CoV) 6 was first isolated in June 2012 from a patient in Saudi Arabia who had died from progressive respiratory and renal failure (1). Since then, over 800 cases have been reported, with a case fatality rate surpassing 30% (2). The progression and severity of the symptoms observed in MERS patients resemble the severe acute respiratory syndrome (SARS) observed in patients infected with SARS-CoV, which caused a global pandemic in 2003, resulting in over 8000 cases, with a case fatality rate of ϳ10% (3). Whereas the SARS-CoV outbreak was contained within months, MERS cases continue to occur 2 years after the emergence of MERS-CoV in the human population. Currently, dromedary camels are suspected to be one of the direct reservoirs for the zoonotic transmission of MERS-CoV, although the exact chain of transmission remains to be explored in more detail (4,5).
MERS-CoV and SARS-CoV are enveloped, positive-sense single-stranded RNA (ϩRNA) viruses that belong to the Betacoronavirus genus in the family Coronaviridae of the Nidovirales order (6). The CoV non-structural proteins (nsps), which drive viral genome replication and subgenomic RNA synthesis, are encoded within a large replicase gene that encompasses the 5Ј-proximal three-quarters of the CoV genome. The replicase gene contains two open reading frames, ORF1a and ORF1b. Translation of ORF1a yields polyprotein 1a (pp1a), and Ϫ1 ribosomal frameshifting facilitates translation of ORF1b to yield pp1ab (7). The pp1a and pp1ab precursors are co-and post-translationally processed into functional nsps by multiple ORF1a-encoded protease domains. CoVs employ either one or two papain-like proteases (PL pro s), depending on the virus species, to release nsp1, nsp2, and nsp3 and a chymotrypsin-like protease (3CL pro ) that cleaves all junctions downstream of nsp4 (reviewed in Ref. 8). Comparative sequence analysis of the MERS-CoV genome and proteome allowed for the prediction and annotation of 16 nsps, along with the location of the probable proteolytic cleavage sites (6). The MERS-CoV PL pro domain, which resides in nsp3, has recently been confirmed to recognize and cleave after the sequence LXGG at the nsp122 and nsp223 junctions, as defined previously for other CoV PL pro s, as well as an IXGG sequence, which constitutes the nsp324 cleavage site (9,10).
These recognition sequences within pp1a/pp1ab resemble the C-terminal LRGG motif of ubiquitin (Ub), an 8.5-kDa protein that can be conjugated to lysine residues or the N terminus of target proteins as a form of post-translational modification through the action of the cellular E1/2/3 ligase system (reviewed in Ref. 11). Additional Ub molecules can be linked to any of the 7 lysine residues in Ub itself or to its N terminus to generate polyubiquitin (poly-Ub) chains of various linkage types (11). The best-studied linkages are the ones occurring at Lys 48 of Ub, which results in the targeting of the tagged substrate to the 26 S proteasome for degradation, and at Lys 63 , which generates a scaffold for the recruitment of cellular proteins to activate numerous signaling cascades, including critical antiviral and proinflammatory pathways (11). The C terminus of Ub can be recognized by deubiquitinating enzymes (DUBs), which catalyze the deconjugation of Ub, thus reversing the effects of ubiquitination (12). Interestingly, CoV PL pro s, including those of MERS-and SARS-CoV, have been suggested to act as multifunctional proteases that not only cleave the viral polyproteins at internal LXGG cleavage sites but also remove Ub and the antiviral Ub-like molecule interferon-stimulated gene 15 (ISG15) from cellular proteins, presumably to suppress host antiviral pathways (9,(13)(14)(15)(16)(17)(18)(19).
Activation of antiviral and proinflammatory pathways is a critical first line of defense against virus infections, including those caused by nidoviruses. Viral RNA molecules are recognized by pattern recognition receptors, such as the cytoplasmic RIG-I-like receptors (RLRs) RIG-I and MDA5, which are activated by intracellular viral RNA transcripts bearing 5Ј tri-and diphosphates and double-stranded RNA (dsRNA) replication intermediates, respectively (20,21). Upon their stimulation, RLRs signal through the mitochondrial antiviral signaling protein (MAVS), leading to the formation of a signaling complex at the mitochondrial membrane and ultimately to the activation of transcription factors IRF-3 and NF-B. These transcription factors in turn regulate the expression of antiviral type 1 interferons (IFN), including IFN-␤, which acts through autocrine and paracrine receptor-mediated signaling pathways to induce the transcription of numerous interferon-stimulated genes (ISGs) that will interfere with virus replication as well as proinflammatory cytokines, such as IL-6, IL-8, and TNF-␣. Regulation of the antiviral and proinflammatory pathways is largely Ub-dependent, because multiple factors in the innate immune cascade are ubiquitinated, including RIG-I, which is critical for downstream signaling. Cellular DUBs function to prevent excessive inflammation and immune responses during infection by removal of Ub from innate immune factors (reviewed in Ref. 22).
The DUB activities of MERS-and SARS-CoV PL pro have been implicated in the suppression of host antiviral pathways because these proteases can suppress IFN-␤ induction upon their ectopic expression (9,13,15,16,19,23). Previous work has shown that during infection, SARS-CoV indeed suppresses the host's antiviral responses by preventing the induction of IFN-␤ expression in cell culture (24 -26). Similarly, MERS-CoV infection has been found to elicit a poor type-1 IFN response in cultured monocyte-derived dendritic cells (27) and alveolar epithelial A549 cells (28) as well as ex vivo in bronchial and lung tissue samples (28). Furthermore, delayed induction of proinflammatory cytokines in human airway epithelial cells infected with MERS-CoV has been reported (29).
Although the above observations suggest that MERS-and SARS-CoV actively suppress antiviral responses, such as IFN-␤ production and inflammation, they do not directly implicate the DUB activity of PL pro as being responsible for (part of) this suppression. Due to the dependence of MERS-CoV replication on the ability of PL pro to cleave the nsp1-nsp3 region of the replicase polyproteins, studying the role of PL pro DUB activity, specifically in the suppression of the cellular innate immune response, is difficult because both activities depend on the same enzyme active site. Selective inactivation of only the DUB activity of PL pro would enable the study of how this activity alone affects cellular signaling; however, achieving this requires detailed information on the structural basis of Ub recognition and deconjugation by PL pro . To this end, we determined the crystal structure of MERS-CoV PL pro bound to Ub to elucidate the molecular determinants of Ub recognition. Based on the structure of this complex, mutations were introduced that selectively disrupted Ub recognition by targeting regions of the Ub-binding site on PL pro that were sufficiently distant from the active site of the protease. Using this approach, we were able to remove the DUB activity from PL pro without affecting its ability to cleave the nsp324 cleavage site in trans. This enabled us, for the first time, to demonstrate that the DUB activity of MERS-CoV PL pro can suppress the MAVS-mediated induction of IFN-␤ expression.
Construction of MERS-CoV PL pro Expression Plasmids-A cDNA fragment encoding the PL pro domain (amino acids 1479 -1803 of the MERS-CoV pp1a/pp1ab polyprotein (NCBI ID: JX869059); pp1a/pp1ab amino acid numbering is used throughout the rest of this work) was cloned into bacterial expression vector pASK3 in-frame with N-terminal Ub and a C-terminal His 6 purification tag to produce pASK-MERS-CoV-PL pro .
Using standard methodologies, the sequence encoding amino acids 1480 -1803 of MERS-CoV pp1a/pp1ab was PCRamplified, cloned downstream of the T7 promoter of expression vector pE-SUMO (LifeSensors), and used to transform Escherichia coli BL21 (DE3) GOLD cells (Stratagene) grown under kanamycin selection (35 g/ml). Recombinant expression plasmid (pE-SUMO-PL pro ) was isolated from a single colony, and DNA sequencing confirmed the expected sequence of the PL pro domain and the in-frame fusion of the 5Ј-end to a sequence encoding a His 6 -SUMO purification tag, which facilitated purification of the product by immobilized metal (nickel) affinity chromatography as described below.
To obtain high expression in eukaryotic cells, the sequence of MERS-CoV nsp3-4 (amino acids 854 -3246) flanked by an N-terminal HA tag and a C-terminal V5 tag was optimized based on the human codon usage frequency, and potential splice sites and polyadenylation signals were removed. This sequence was synthesized (Invitrogen) and subsequently cloned into the pCAGGS vector (Addgene) using standard methodologies. The following expression constructs were generated: pCAGGS-HA-nsp3-4-V5 (amino acids 854 -3246), pCAGGS-HA-nsp3C-4-V5 (amino acids 1820 -3246, which does not include the PL pro domain), and pCAGGS-HA-nsp3-Myc (amino acids 854 -2739). The sequence encoding MERS-CoV PL pro (amino acids 1479 -1803) was PCR-amplified using synthetic plasmid DNA as a template and cloned in frame with a C-terminal V5 tag in the pcDNA3.1(Ϫ) vector (Invitrogen). The pASK-MERS-CoV-PL pro and pcDNA3.1-MERS-CoV-PL pro expression constructs served as templates for site-directed mutagenesis using the QuikChange strategy with Pfu DNA polymerase (Agilent). All constructs were verified by sequencing. The sequences of the constructs and primers used in this study are available upon request.
Purification of MERS-CoV PL pro and in Vitro DUB Activity Assay-In vitro DUB activity assays were performed with recombinant MERS-CoV PL pro batch-purified from lysates of E. coli strain C2523. Cells transformed with pASK-MERS-CoV-PL pro were cultured to an A 600 of 0.6 in lysogeny broth at 37°C. Protein expression was then induced with 200 ng/ml anhydrotetracycline for 16 h at 20°C. The cells were pelleted, resuspended in lysis buffer (20 mM HEPES, pH 7.0, 200 mM NaCl, 10% (v/v) glycerol, and 0.1 mg/ml lysozyme), and lysed for 1 h at 4°C, followed by sonication. The lysate was clarified by centri-fugation at 20,000 ϫ g for 20 min at 4°C, and the soluble fraction was applied to Talon resin (GE Healthcare) pre-equilibrated with lysis buffer. After a 2-h rolling incubation at 4°C, the beads were washed four times with wash buffer (20 mM HEPES, pH 7.0, 200 mM NaCl, 10% (v/v) glycerol, and 20 mM imidazole), followed by the elution of the protein with elution buffer (20 mM HEPES, pH 7.0, 200 mM NaCl, 10% (v/v) glycerol, and 250 mM imidazole). Eluted protein was dialyzed against storage buffer (20 mM HEPES, pH 7.0, 100 mM NaCl, 50% (v/v) glycerol, 2 mM dithiothreitol (DTT)) and stored at Ϫ80°C. N-terminal Ub is cleaved from the Ub-PL pro -His 6 fusion protein by the PL pro domain itself during expression. To achieve removal of the Ub from mutated and/or inactive PL pro , E. coli strain C2523 containing pCG1, expressing the ubiquitin-specific processing protease 1 (Ubp1), was used (37).
In vitro DUB activity assays were performed as described by van Kasteren et al. (30). Briefly, the indicated amounts of purified MERS-CoV PL pro wild type or active site mutant (C1592A) were incubated with 2.5 g of either Lys 48 -linked poly-Ub chains or Lys 63 -linked poly-Ub chains (Boston Biochem) in a final volume of 10 l. Isopeptidase T (Boston Biochem) served as a positive control. After a 2-h incubation at 37°C, the reaction was stopped by the addition of 4ϫ Laemmli sample buffer (4ϫ LSB; 500 mM Tris, 4% SDS, 40% glycerol, 0.02% bromphenol blue, 2 mM DTT, pH 6.8). SDS-polyacrylamide gels were stained with Coomassie Brilliant Blue (Sigma-Aldrich) and scanned using a GS-800 calibrated densitometer (Bio-Rad).
Expression and Purification of MERS-CoV PL pro for Crystallization-E. coli BL21(DE3) GOLD cells harboring pE-SUMO-PL pro were grown at 37°C with aeration in 500 ml of lysogeny broth containing kanamycin (35 g/ml) to an A 600 of 0.6 -0.8. Expression of the His 6 -SUMO-PL pro fusion protein was then induced by the addition 1 mM isopropyl ␤-D-1-thiogalactopyranoside for 18 h at 16°C with aeration. Cells were pelleted by centrifugation and stored at Ϫ80°C.
Cell pellets were resuspended in ice-cold lysis buffer (150 mM Tris, pH 8.5, 1 M NaCl, 0.1 mM phenylmethanesulfonyl fluoride (PMSF), 2 mM DTT) and lysed using a French pressure cell (AMINCO). Cell lysate was clarified by centrifugation (17,211 ϫ g at 4°C), and the supernatant containing the His 6 -SUMO-PL pro fusion was applied to a column containing nickelnitrilotriacetic acid affinity resin (Qiagen). The column was washed with 10 column volumes of lysis buffer supplemented with 25 mM imidazole, followed by elution of the fusion protein with lysis buffer containing 250 mM imidazole. The His 6 -SUMO tag was then removed from PL pro by adding His 6 -tagged Ulp1 SUMO protease to the eluted SUMO-PL pro fusion, followed by dialysis of the protein mixture overnight against 2 liters of cleavage buffer (150 mM NaCl, 50 mM Tris, pH 8.0, 1 mM DTT) at 4°C. Tag-free PL pro was separated from His 6 -SUMO and the His 6 -Ulp1 SUMO protease by passing the dialyzed protein mix through a nickel-nitrilotriacetic acid gravity column. The flow-through contained purified PL pro that was subsequently dialyzed against 20 mM Tris, pH 8.5, 150 mM NaCl, 2 mM DTT and further purified by gel filtration using a Superdex 75 (GE Healthcare) gel filtration column.
Covalent Coupling of Ub to PL pro -Ub(1-75)-3-bromopropylamine (Ub-3Br) is a modified form of Ub with a reactive C terminus that forms an irreversible covalent linkage to the active site cysteine of DUBs and was prepared according to Messick et al. (38) and Borodovsky et al. (39). Purified PL pro was incubated with a 2-fold molar excess of Ub-3Br and incubated for 1 h at room temperature with end-over-end mixing. The resulting PL pro ⅐Ub complex was dialyzed into 20 mM Tris, pH 8.5, 150 mM NaCl, 2 mM DTT, and excess Ub-3Br was removed by gel filtration using a Superdex 75 column.
Crystallization of PL pro and PL pro ⅐Ub Complexes-The purified PL pro ⅐Ub complex was concentrated and crystallized at 20°C in two different conditions using the vapor diffusion method: 1) 20% PEG 4000, 0.1 M trisodium citrate, pH 5.4, 20% isopropyl alcohol at 10 mg/ml, which yielded the structure of open PL pro ⅐Ub (see "Results"), and 2) 1.80 M ammonium sulfate (AmSO 4 ) at 20 mg/ml, which yielded the structure of closed PL pro ⅐Ub (see "Results"). Crystals of unliganded PL pro were also grown using the vapor diffusion method in 18% PEG 4000, 0.1 M trisodium citrate, pH 5.6, 16% isopropyl alcohol after concentrating the protein to 12 mg/ml. Immediately prior to crystallization, 1 M DTT was added to the protein to a final concentration of 5 mM, which was found to improve crystallization.
In preparation for x-ray data collection, single crystals of open PL pro ⅐Ub from condition 1 above were briefly swept through a droplet of cryoprotectant composed of 22% PEG 4000, 0.1 M trisodium citrate, pH 5.6, 20% 1,2-propanediol before flash cooling in liquid nitrogen. Similarly, single crystals of closed PL pro ⅐Ub from condition 2 above and unbound PL pro were cryoprotected in 1.85 M AmSO 4 , 15% glycerol and 22% PEG 4000, 0.1 M trisodium citrate, pH 5.6, 10% 1,2-propanediol, respectively, before flash cooling in liquid nitrogen.
Data Collection and Structure Determination-X-ray diffraction data were collected from all crystals at the Zn-K absorption edge at beamline 08B1-1 of the Canadian Light Source and integrated using XDS (40). Integrated data were then scaled using Scala (41). Initial phase estimates for reflections collected from unliganded and Ub-bound PL pro were determined via a single wavelength anomalous dispersion experiment. The position of the zinc anomalous scatterer was identified using HySS (42), and density modification was performed with RESOLVE (43) within the phenix.autosol pipeline (44). Initial models were constructed using phenix.autobuild, and further model building and refinement were carried out using Coot (45) and phenix.refine (46). Crystallographic statistics for all structures are found in Table 1.
Protease Activity Assays in Cell Culture-HEK293T cells, grown to 80% confluence in 12-well plates, were transfected using the calcium phosphate transfection method (47). To determine the DUB activity of MERS-CoV PL pro , plasmids encoding FLAG-tagged Ub (0.25 g), GFP (0.25 g), and MERS-CoV-PL pro -V5 (0.2 g) were co-transfected. A combination of plasmids encoding GFP (0.25 g), HA-nsp3C-4-V5 (0.2 g), and MERS-CoV-PL pro -V5 (0.15 g) were transfected to assess the in trans cleavage activity of MERS-CoV-PL pro . Total amounts of transfected DNA were equalized to 2 g by the addition of empty pcDNA vector. At 18 h post-transfection, cells were lysed in 2ϫ LSB. Proteins were separated in an SDS- polyacrylamide gel and blotted onto Hybond-P (GE Healthcare) using the Trans-blot turbo transfer system (Bio-Rad). Aspecific binding to the membrane was blocked with dried milk powder solution, and after antibody incubation, protein bands were visualized using Pierce ECL 2 Western blotting substrate (Thermo Scientific).
Luciferase-based IFN-␤ Reporter Assay-Using the calcium phosphate method, 80% confluent HEK293T cells in 24-well plates were transfected with 5 ng of plasmid pRL-TK (Promega) encoding Renilla luciferase; IFN-␤-Luc firefly reporter plasmid (25 ng); innate immune response inducer plasmids encoding RIG-I (2CARD) , MAVS, or IRF3 (5D) (25 ng); and the indicated quantities of MERS-CoV PL pro -or MERS-CoV nsp3-encoding expression plasmids. Total amounts of transfected DNA were equalized to 1 g by the addition of empty pcDNA vector. At 16 h post-transfection, cells were lysed in 1ϫ passive lysis buffer (Promega). Firefly and Renilla luciferase activity was measured using the Dual-Luciferase reporter assay system (Promega) on a Mithras LB 940 multimode reader (Berthold Technologies). Experiments were performed in triplicate and independently repeated at least four times. Firefly luciferase activity was normalized to Renilla luciferase, and statistical significance was determined using an unpaired two-tailed Student's t test. Values of Ͻ0.05 were considered statistically significant. 4ϫ LSB was added to the remaining lysates, and these samples were analyzed by Western blotting as described above.

DUB Activity of Recombinant MERS-CoV PL pro
It was recently shown in cell culture experiments that ectopic expression of MERS-CoV PL pro resulted in deconjugation of poly-Ub and ISG15 from cellular targets (9,16). DUB activity of purified recombinant MERS-CoV PL pro was also demonstrated using Ub-7-amino-4-trifluoromethylcoumarin (48) or Ub-7amino-4-methylcoumarin (49) as a substrate. To characterize the direct activity of recombinant MERS-CoV PL pro toward poly-Ub, we purified the enzyme from E. coli and incubated it with either Lys 48 -or Lys 63 -linked poly-Ub chains. Wild-type PL pro degraded both Lys 48 -and Lys 63 -linked chains in a concentration-dependent manner, whereas mutating the active site nucleophile (C1592A) severely reduced the activity of the enzyme toward both Ub linkage types (Fig. 1). No clear preference of the enzyme for cleaving either the Lys 63 or the Lys 48 Ub linkage was observed under the conditions used in this in vitro DUB assay (Fig. 1, compare A and B). This assay clearly demonstrated that the protease domain used throughout this study for ectopic expression and crystallization experiments possesses DUB activity toward Lys 48 -and Lys 63 -linked Ub chains and that this activity does not require other viral or cellular proteins. During the preparation of this manuscript, an article by Báez-Santos et al. (50) was published in which similar results were presented.

Crystal Structures of MERS-CoV PL pro and PL pro ⅐Ub Complexes
MERS-CoV PL pro -The crystal structure of PL pro was determined both on its own and as a covalent complex with Ub (PL pro ⅐Ub). The PL pro domain crystallized in space group P6 3 , and consistent with another recently determined crystal structure of MERS-CoV PL pro (49), we found the protease to adopt a fold consistent with DUBs of the ubiquitin-specific protease (USP) family. The structure includes a C-terminal catalytic domain containing a right-handed fingers, palm, and thumb domain organization as well as an N-terminal Ub-like (Ubl) domain found in many USPs, including that of SARS-CoV (51, 52) ( Fig. 2A). The packing of the palm and thumb domains forms a cleft leading into the active site in a manner consistent with the domain organization prototyped by the Clan CA group of cysteine proteases (53). The Ubl domain packs against the thumb domain composed of helices ␣2-7, which in turn packs against the palm domain composed of strands ␤6, ␤7, and ␤14 -19. Extending from the palm, the fingers domain is composed of strands ␤10, ␤11, ␤13, ␤14, and ␤19 and contains a C 4 zinc ribbon motif (54) coordinating a zinc atom via residues Cys 1672 , Cys 1675 , Cys 1707 , and Cys 1709 in tetrahedral geometry, similar to that of SARS PL pro , transmissible gastroenteritis coronavirus PL1 pro , and cellular USP2 and USP21 (51,(55)(56)(57).
PL pro Covalently Bound to Ub-The MERS-CoV PL pro ⅐Ub complex crystallized in two different space groups (P6 3 and P6 5 DECEMBER 12, 2014 • VOLUME 289 • NUMBER 50 lized in space group P6 3 revealed weak density for the covalently bound Ub molecule. Although the entire bound Ub molecule could be modeled within its binding site on PL pro in this crystal form, high temperature factors for atoms comprising the modeled Ub molecule suggested that it was not rigidly bound to the protease despite being covalently linked to the active site cysteine. Further analysis of the crystal packing revealed that the Ub molecule was fully exposed to solvent and not involved in crystal contacts, which provided a degree of mobility to Ub when bound to PL pro (Fig. 3A). This result  A, the open PL pro ⅐Ub structure crystallized in space group P6 3 , where Ub was found to face the solvent, uninvolved in crystal contacts. B, the closed PL pro ⅐Ub structure crystallized in space group P6 5 22, where Ub no longer faces the solvent, and is involved in crystal contacts. Images were created using PyMOL (70). encouraged us to pursue additional crystallization conditions, which yielded crystals of PL pro ⅐Ub in space group P6 5 22 (Figs. 2B and 3B). The crystal packing in this space group allowed for multiple crystal contacts between the bound Ub monomer and surrounding symmetry mates and resulted in clear, well defined density for the Ub molecule (Fig. 3B). Interestingly, relative to the P6 3 crystal forms of PL pro , the fingers domain in this crystal form was moved toward Ub (Fig. 2C). In light of these movements, the PL pro ⅐Ub structure with the fingers domain positioned away from Ub (space group P6 3 ) will hereafter be referred to as "open" PL pro ⅐Ub, whereas the structure with the fingers domain shifted toward Ub (space group P6 5 22) will be referred to as "closed" PL pro ⅐Ub. An overlay of the different PL pro crystal structures that have been determined reveals that these structures vary in the position of the zinc ribbon motif, further suggesting a high degree of mobility for this region (Fig.  2C). In line with this observation, movement of the fingers domain toward bound Ub was also reported for the SARS-CoV PL pro domain, which displayed a 3.8-Å movement of the zinc atom when comparing the Ub-bound and unbound structures (58). Further comparison of the closed MERS-CoV PL pro ⅐Ub structure with the recently determined SARS-CoV PL pro ⅐Ub structure (58) revealed differences in the relative orientation of the fingers domain of the two proteases. The MERS-CoV PL pro fingers domain was found to be shifted ϳ26°away from the palm domain compared with that of SARS-CoV PL pro , resulting in a slight difference in the Ub binding orientation, with the MERS-CoV PL pro -bound Ub being positioned closer toward helix ␣7 of the palm domain (Fig. 4).

PL pro Active Site Organization and Interaction with the C-terminal RLRGG Motif of Ub
The cleft formed between the palm and thumb domains of PL pro guides the C-terminal 72 RLRGG 76 motif of Ub toward the protease active site, and the interactions between the C-terminal motif of Ub and the active site cleft are depicted in Fig. 5 (A  and B). The PL pro active site is composed of a Cys 1592 -His 1759 -Asp 1774 catalytic triad, which adopts a catalytically competent arrangement in both the unliganded and Ub-bound structures of PL pro (Fig. 5C). The oxyanion hole of the PL pro active site appears to be composed of backbone amides from residues Asn 1590 , Asn 1591 , and Cys 1592 , which appear suitably arranged to stabilize the negative charge that develops on the carbonyl oxygen of the scissile bond during catalysis (Fig. 5C). Interestingly, as noted by Lei et al. (49), the MERS-CoV PL pro active site appears incomplete. In SARS-CoV PL pro , Trp 107 (amino acid numbering according to the structure of Protein Data Bank entry 2FE8) is positioned within the enzyme's active site with the indole nitrogen of its side chain oriented such that it is probably involved in forming part of the oxyanion hole (51). In the case of MERS-CoV PL pro , we and others (48,49) have found the structurally equivalent residue in MERS-CoV PL pro to be Leu 1587 , which would be unable to participate in stabilizing the oxyanion during catalysis. Furthermore, it was recently shown that MERS-CoV PL pro L1587W mutants show greater catalytic efficiency than wild-type PL pro (48,49). Given the effect this residue has on the catalytic rate of PL pro , it will be very interesting to understand how this residue influences MERS-CoV replication kinetics. It has been proposed that the decreased catalytic efficiency may influence maturation of the MERS-CoV polyprotein (48) and could be involved in the recognition of residues downstream of the scissile bond of the polyprotein cleavage sites or in the modulation of PL pro DUB activity.
Interestingly, differences were observed in the position of a loop on PL pro connecting strands ␤15 and ␤16, which is structurally analogous to the blocking loop (BL2) first described in the structure of USP14 (59). This loop is disordered in our unliganded PL pro structure and that previously determined by others (49); however, in both of our PL pro ⅐Ub structures, we found this loop to be fully resolved, supported by the main-chain hydrogen-bonding interactions between Arg 74 of Ub and Gly 1758 of PL pro , as well as a hydrophobic interaction between Val 1757 and Pro 1644 , two PL pro residues present on opposite sides of the active site cleft (Fig. 5A). The side-chain -amino group of the Ub residue Arg 74 is also hydrogen-bonded to the main-chain carbonyl group of PL pro residue Thr 1755 ; however, this interaction is only seen in the open PL pro ⅐Ub structure. The SARS-CoV PL pro domain has also been crystallized both in the presence (51) and absence (58) of Ub, and although the BL2 loop of unbound SARS-CoV PL pro was resolved in two of three monomers of the asymmetric unit, the third showed weak electron density for BL2 and high temperature factors, indicating a high degree of mobility. In addition, in the transmissible gastroenteritis coronavirus USP domain PL1 pro , a structurally analogous BL2 loop was found to be in an open conformation with poorly defined electron density in the absence of substrate (55). It is interesting to note that all three coronavirus USP DUBs crystallized to date (from MERS-CoV, SARS-CoV, and transmissible gastroenteritis coronavirus) demonstrate a significant degree of flexibility within the BL2 loop region in the absence of substrate and that none of the structures determined in their unbound form demonstrate obstruction of the active site via BL2.

Structure-guided Design of PL pro Mutants Defective in DUB Activity
We previously demonstrated that the DUB activity of the papain-like protease 2 (PLP2) from equine arteritis virus (another member of the nidovirus order), which resembles the ovarian tumor (OTU) domain-containing family of DUBs (60), could be selectively removed without affecting its ability to process the equine arteritis virus replicase polyprotein. This allowed us to establish that the DUB activity of PLP2 is directly responsible for suppressing Ub-dependent antiviral pathways during infection of primary host cells (61). Subsequently, Ratia et al. (62) applied a similar strategy to the SARS-CoV PL pro domain in order to partially remove the DUB activity of PL pro while maintaining the nsp2-3-processing function. We now used the crystal structure of the USP-like MERS PL pro ⅐Ub complex to guide the design of mutations targeting the Ub-binding site on PL pro that would completely disrupt Ub binding without affecting the structural integrity of the active site. PL pro residues interacting directly with Ub were replaced with larger, bulkier residues that would prevent Ub binding by altering both shape and surface electrostatics of the Ub-binding site. Individual mutation of eight different PL pro residues (Arg 1649 , Thr 1653 , Ala 1656 , Asn 1673 , Val 1674 , Val 1691 , Val 1706 , and Gln 1708 ) and combinations thereof were generated (Fig. 6, A-D). Importantly, these residues are located at a distance from the PL pro active site, and thus we hypothesized that they would only participate in DUB activity and not polyprotein processing.
Despite significant movement within the fingers domain of PL pro , most interactions between the protease and Ub are consistent between the open and closed Ub-bound complexes. Residue Ile 44 of Ub, which forms part of the hydrophobic patch that is commonly recognized by Ub-binding proteins (63), interacts with the hydrophobic side chain of Val 1691 of PL pro (Fig. 6B). Residues Gln 49 and Glu 51 of Ub form hydrogen-bonding interactions with Thr 1653 that is present on helix ␣7, which runs through the center of PL pro . Two arginine residues, Arg 1649 of PL pro and Arg 72 of Ub (the latter of which forms part of the C-terminal tail of Ub that is bound in the PL pro active site cleft) are oriented such that the guanidinium groups of these residues are arranged in a stacked conformation (Fig. 6C). In addition, due to the inward movement toward Ub of the closed PL pro ⅐Ub fingers domain, a unique hydrogen-bonding interaction between Gln 62 of Ub and Gln 1708 of PL pro and a hydrophobic interaction between Phe 4 of Ub and Val 1706 of PL pro were found to occur in this complex (Fig. 6D). Residue Ala 1656 is positioned near the C terminus of PL pro helix ␣7, and although it is not directly involved in Ub binding, we believed that it was positioned such that the introduction of larger residues (e.g. arginine or phenylalanine) could disrupt Ub recognition, and thus this residue was targeted for mutation (Fig. 6C). Two residues on the solvent-facing region of the PL pro zinc ribbon motif, Asn 1673 and Val 1674 , were also targeted for mutagenesis. Although they do not bind Ub at the S1 binding site (the substrate binding site on PL pro responsible for binding mono(Ub) in our structure; see Ref. 64 for nomenclature), we hypothesized that it may inhibit association with the distal Ub on Lys 63 poly-Ub chains based on a superposition of a Lys 63 -linked di-Ub model onto the PL pro -bound Ub molecule of the closed PL pro ⅐Ub complex structure determined here (not shown). In addition, the crystal structure of USP21 bound to linear di-Ub was recently determined and revealed that the tip of the fingers domain of this DUB acts as an S2 recognition site, binding to the distal Ub of a linear di-Ub molecule (57). Given the structural similarity between Lys 63 di-Ub and linear di-Ub and the clear activity we observed for MERS-CoV PL pro toward Lys 63 , we hypothesized that mutating residues Asn 1673 and Val 1674 near the zinc ribbon may also disrupt Ub processing.

Targeted Mutations within the PL pro ⅐Ub Binding Site Disrupt Ub Processing but Not Proteolytic Cleavage of the nsp324 Site
Using a previously described ectopic expression assay (61), we monitored the effects of amino acid substitutions in PL pro , as described above, on overall levels of Ub-conjugated proteins in HEK293T cells as well as the ability of these PL pro variants to process the MERS-CoV nsp324 polyprotein cleavage site in trans. V5-tagged PL pro (wild type and mutants) was coexpressed with N-terminally HA-tagged and C-terminally V5-tagged MERS-CoV nsp3C-4 excluding the PL pro domain, hereafter referred to as HA-nsp3C-4-V5. We assume that the successful processing of the nsp324 site in HA-nsp3C-4-V5 is indicative of unaltered proteolytic cleavage capability of PL pro , which during infection facilitates the release of nsp1, -2, and -3 from the viral polyproteins. Processing of HA-nsp3C-4-V5 in trans by wild-type PL pro and our panel of mutants was visualized via Western blotting (Fig. 7A). Whereas wild-type PL pro was able to cleave HA-nsp3C-4-V5 substrate in trans, the PL pro active site mutant C1592A was unable to cleave the nsp324 site (Fig. 7A, compare lanes 5 and 6 and lanes 19 and 20). As expected, each of the substitutions in the Ub-binding site of PL pro only minimally affected nsp324 cleavage, with the exception of the A1656R mutant that displayed a clearly reduced ability to cleave HA-nsp3C-4-V5 compared with wildtype PL pro (Fig. 7A, compare lanes 5 and 10). This suggests that Ala 1656 of PL pro may be involved in recognizing and binding sequences in the vicinity of the nsp324 cleavage site. Most double and triple substitutions tested were also slightly less efficient in cleaving HA-nsp3C-4-V5 compared with the wild-type control.
In order to analyze the effect of the mutations on overall DUB activity, PL pro -V5 was co-expressed with FLAG-Ub, and the levels of FLAG-Ub-conjugated cellular proteins were visualized via Western blotting (Fig. 7B). Expression of wild-type PL pro resulted in a strong decrease of the accumulation of FLAG-Ub conjugates, whereas a negligible effect was observed upon expression of active site mutant C1592A (Fig. 7B, compare  lanes 3 and 4 and lanes 16 and 17). Substitutions of residue Val 1691 , positioned on strand ␤12 of PL pro , and Thr 1653 and Ala 1656 , residues located on helix ␣7 (Fig. 6, B and C), displayed the clearest reduction of PL pro DUB activity (Fig. 7B, lanes  5-8). The V1691R mutation had the most pronounced effect, and a PL pro T1653R/V1691R double mutant also displayed severely reduced DUB activity, comparable with that seen for the active site mutant (Fig. 7B, compare lanes 4 and 5 and lanes  17 and 22). Notably, a more conservative substitution at the same position, V1691L, had a much less pronounced effect on DUB activity (Fig. 7B, lane 6). Substitution of Val 1674 with either Ser or Arg impaired DUB activity but to a much lesser extent than substitutions targeting Val 1691 , Thr 1653 , and Ala 1656 (Fig. 7B, compare lanes 5-8, 10, and 11). The N1673R substitution did not negatively affect DUB activity of PL pro at all, whereas the N1673R/V1674S double substitution resulted in slightly greater DUB activity (Fig. 7B, lanes 9 and 20). These results do not support our hypothesis based on modeling that Asn 1673 and Val 1674 might form part of an S2 binding site that recognizes an additional distal Ub within a Lys 63 -linked chain. Further structural studies are needed to validate the role of these residues in binding Ub chains. It should be noted, however, that these mutants may still be able to process Lys 63linked poly-Ub chains by recognizing a single Ub monomer at the end of a poly-Ub substrate, which may explain the ineffectiveness of these mutations in disrupting DUB activity. Mutations at residues Val 1706 and Gln 1708 did not influence DUB activity of PL pro (Fig. 7B, lanes 18 and 19). Given that these residues were only found to interact with Ub in our closed PL pro structure (Fig. 6A), their failure to inhibit DUB activity in this cellular DUB assay is not surprising and indicates that these residues are not essential for Ub recognition. Interestingly and repeatedly observed, the R1649Y mutant was found to have even greater DUB activity than wild-type PL pro (Fig. 7B, compare lanes 3 and 12). This residue was found to interact with residue Arg 72 of Ub, and although this result was unexpected, it is possible that the R1649Y mutant retains the ability to interact with Arg 72 of Ub via a cationinteraction between the aromatic tyrosine inserted into PL pro and the positively charged arginine of Ub. Together, the findings from our mutagenesis study demonstrate that it is possible to selectively decouple the DUB and polyprotein processing activities of MERS-CoV PL pro through structureguided site-directed mutagenesis.

PL pro DUB Activity Suppresses the Innate Immune Response
Conjugation and deconjugation of Ub plays an important role in the regulation of the innate immune response, and not surprisingly, pathogens have evolved mechanisms to subvert these Ub-  1 and 15), and cleavage resulted in the generation of full-length HA-tagged nsp3 and V5-tagged nsp4. Cells were lysed 18 h post-transfection, and expressed proteins were analyzed by Western blotting. Proteolytic cleavage was measured from the generation of N-terminal HA-tagged nsp3C and C-terminal V5-tagged nsp4. B, HEK293T cells were transfected with a combination of plasmids encoding FLAG-Ub, PL pro -V5 (wild-type and mutants), and GFP (as a transfection control). Cells were lysed 18 h posttransfection, and expressed proteins were analyzed by Western blotting to visualize the deconjugation of FLAG-tagged Ub from a wide range of cellular proteins by MERS-CoV PL pro wild-type and mutants. dependent pathways (reviewed in Ref. 22). For arteriviruses, which are distant relatives of CoVs within the nidovirus order, it has been shown that the DUB activity of their PLP2 is involved in antagonizing IFN-␤ activation upon ectopic expression, and for equine arteritis virus, this was confirmed during infection in host cells (61,65). Coronavirus papain-like proteases have been suggested to act as IFN-␤ and NF-B antagonists as well (15,23,66,67). MERS-CoV PL pro is thought to possess these properties based on its capability to inhibit RIG-I-, MDA5-, and MAVS-induced IFN-␤ promoter stimulation and to reduce TNF-␣-induced NF-B reporter gene activity (9,16). We therefore designed luciferase-based reporter gene assays to establish whether the DUB activity of MERS-CoV PL pro alone suffices to antagonize the IFN-␤ pathway. To this end, we first assessed at which level of this innate immune signal transduction pathway MERS-CoV PL pro is most active as a suppressor.
Innate immune signaling was induced in HEK293T cells by expression of one of three signaling factors, RIG-I, MAVS, or IRF3, which stimulate the pathway leading to IFN-␤ production at different levels. Because RIG-I and IRF3 normally need to be activated through post-translational modification (ubiquitination and phosphorylation, respectively), constitutively active variants were used (RIG-I (2CARD) and IRF3 (5D) ), which efficiently induce downstream signaling independent of these activation steps. Cells were co-transfected with plasmids encoding one of these innate immune signaling proteins and wild-type PL pro , the PL pro active site mutant C1592A, or full-length MERS-CoV nsp3 containing the PL pro domain. The inhibitory effect of the PL pro variants on the activation of the IFN-␤ promotor by the different stimuli was measured via co-expression of a firefly luciferase reporter gene under control of the IFN-␤ promoter. Another co-transfected plasmid encoding Renilla luciferase was included as an internal control in order to be able to correct for variability in transfection efficiency. At 16 h posttransfection, luciferase activities were measured, and activation of the IFN-␤ promoter induced by expression of RIG-I (2CARD) , MAVS, or IRF3 (5D) was set at 100% (Fig. 8). In accordance with Mielech et al. (16), we observed that MERS-CoV PL pro significantly reduced the IFN-␤ promoter activation that could be induced by expression of either RIG-I (2CARD) or MAVS. This effect was concentration-dependent, whereas the PL pro active site mutant was unable to block IFN-␤ promoter activation (Fig. 8, A and C). MERS-CoV nsp3 expression also inhibited RIG-I-and MAVS-mediated IFN-␤ promoter induction (Fig. 8,  B and D), and together this suggested that PL pro inhibits innate immune signaling at least downstream of the MAVS adaptor and possibly also in the signaling between RIG-I and MAVS. MERS-CoV PL pro also inhibited activation of the IFN-␤ promoter after stimulation with IRF3 (5D) in a concentration-dependent manner, whereas the C1592A mutant did not reduce IFN-␤ promoter activation (Fig. 8E). However, expression of full-length MERS-CoV nsp3 did not significantly inhibit IFN-␤ promoter activation after stimulation with IRF3 (5D) (Fig. 8F). This suggests that the subcellular localization of the protease, which in the case of full-length nsp3 is membrane-anchored and in the case of the PL pro domain is presumably cytosolic, may be important in determining its substrate specificity. Taken together, our results suggest that MERS-CoV PL pro pri-marily interferes with the IFN-␤ signaling pathway at the level between MAVS and IRF3.
We therefore chose to use MAVS-mediated induction of IFN-␤ promoter activation in subsequent experiments. This also resulted in the strongest inhibition by PL pro , providing a maximum window to assess the effects on IFN-␤ promoter inhibition by the PL pro mutants with specifically inactivated DUB activity. Inhibition of IFN-␤ promoter activation by wildtype and mutant PL pro was determined by calculating the relative luciferase activity (Fig. 9). Expression of wild-type PL pro reduced MAVS-induced IFN-␤ promoter activity to ϳ20% of the control, whereas active site mutant C1592A reduced it by only a few percent compared with the untreated control (Fig. 9). Substitutions T1653R and A1656R resulted in greatly impaired DUB activity (Fig. 7B, lanes 7 and 8), and compared with wildtype PL pro , expression of these mutants resulted in higher IFN-␤ promoter activity, with relative luciferase values of ϳ54 and 58% respectively (Fig. 9). It should, however, be noted that the A1656R mutant was also impaired in cleaving the nsp324 site, and therefore this mutation nonspecifically disrupted the two proteolytic functions of PL pro . Strikingly, each mutant containing the V1691R substitution was completely unable to inhibit IFN-␤ promoter activation, resulting in relative luciferase activity levels similar to those seen with the active site mutant (Fig. 9, lanes 4, 16, and 17). This strongly suggested that the DUB activity of PL pro , which we found to be severely impaired in V1691R mutants (Fig. 7B), is responsible for suppressing MAVS-induced IFN-␤ promoter activity in this assay. The level of reduction in DUB activity corresponded to the degree of inhibition of IFN-␤ promoter activation for all PL pro mutants tested, which strengthens this conclusion. In accordance with its increased DUB activity, mutant R1649Y suppressed MAVS-induced IFN-␤ promoter activity more effectively than wild-type PL pro .
Taken together, our data show that the DUB activity of MERS-CoV PL pro suffices to efficiently suppress MAVS-induced IFN-␤ promoter activation and that this activity can be selectively disabled, without disrupting protease activity toward the nsp324 cleavage site, by targeting the Ub-binding site of the enzyme. This demonstrates for the first time that the DUB activity of MERS-CoV PL pro is specifically responsible for suppressing the innate immune response.

DISCUSSION
Guided by the MERS PL pro ⅐Ub crystal structures, we here describe how the DUB activity of PL pro can be selectively disabled by introducing mutations into the S1 binding pocket of the protease (Fig. 6). Particularly, the substitution of Val 1691 with the bulky and charged arginine residue severely impaired DUB activity in our cell culture-based assays. In addition, our results demonstrate that the majority of the mutations within the S1 Ub-binding site of PL pro that were tested do not affect trans cleavage of the nsp324 junction, with the exception of an A1656R mutant that did disrupt cleavage of the nsp324 site. The latter result indicates that Ala 1656 resides in a region of PL pro that recognizes both Ub and a region of the nsp3C-4 construct that was used to test cleavage efficiency. DECEMBER 12, 2014 • VOLUME 289 • NUMBER 50

JOURNAL OF BIOLOGICAL CHEMISTRY 34677
Our results demonstrate that the DUB activity of MERS-CoV PL pro inhibits IFN-␤ promoter activation when innate immune signaling is induced by co-expression of either RIG-I or MAVS. The fact that suppression of IFN-␤ promoter activation was completely eliminated for several of our mutants (Fig. 9) strongly suggests that the proteolytic activity still present in those mutant enzymes has no additional role in the suppression of this particular branch of the innate immune response (e.g. by directly cleaving RIG-I or MAVS). A number of other CoV papain-like proteases with DUB activity have also been implicated in antagonizing the host innate immune response (15,23,66,67). In agreement with our data, recent studies have demonstrated the ability of MERS-CoV PL pro to inhibit RIG-I-, MDA5-, and MAVS-dependent IFN-␤ promoter activation as well as to down-regulate the level of IFN-␤ mRNA transcripts in MDA5-stimulated cells (16). The current data support the hypothesis that all of these activities solely depend on the deubiquitinating capacities of these coronavirus enzymes. Reports regarding the dependence of MERS-CoV PL pro -mediated IFN-␤ antagonism on the enzyme's protease activity have, however, varied thus far. Mielech et al. (16) recently demonstrated that a MERS-CoV nsp3 fragment containing PL pro but excluding the transmembrane domain can inhibit MAVS-, RIG-I-, and MDA5-dependent IFN-␤ promoter activation, and MDA5 mediated IFN-␤ mRNA transcription only with a functional PL pro active site. Yang et al. (9) on the other hand used a MERS-CoV PL pro expression product extending into the nsp3 transmembrane region to demonstrate that down-regulation of RIG-I-stimulated IFN-␤ promoter activity is seen even with an active site knock-out mutant. Here we show that inhibition of RIG-I-, MAVS-, and IRF3-induced IFN-␤ promoter activity by the MERS-CoV PL pro domain is clearly dependent on a functional active site and that it is specifically the DUB activity of the protease that mediates this inhibition. However, the possibility cannot be ruled out that other parts of nsp3 contain additional innate immune suppressing activities, which may be responsible for the protease-independent effects reported with longer expression products.
Ubiquitination plays an important role in the regulation of pathways involved in detecting and counteracting viral infections, and, not surprisingly, a number of viruses of substantial diversity have been found to deploy DUBs that manipulate these signaling processes by reversing the post-translational modification of cellular proteins by Ub conjugation (19,68). Some of these DUBs, specifically those found in (ϩ)RNA viruses, are also critical for viral replication by catalyzing the proteolytic cleavage of specific sites in viral polyproteins, thus complicating our ability to study the direct effects of the additional DUB activity of these viral proteases. Ultimately, these effects need to be studied in the context of a viral infection; however, a simple inactivation of the protease/DUB would not only fail to prove the specific involvement of the DUB activity, it would also prevent viral replication. The method described here selectively removed the DUB activity of the MERS-CoV PL pro domain while leaving polyprotein processing activity at the nsp324 site unhindered, thus paving the way for the application of these mutations to recombinant MERS-CoV and the direct study of the role of DUB activity during infection.
We were able to show that Lys 48 -and Lys 63 -linked poly-Ub chains are processed in vitro by MERS-CoV PL pro at similar rates, which is in accordance with a recent report by Báez-Santos et al. (50). In contrast, SARS-CoV PL pro rapidly cleaves Lys 48 -linked poly-Ub and displays only moderate activity for Lys 63 linkages in similar assays (62). It has been suggested that SARS-CoV PL pro may recognize Lys 48 -linked di-Ub via its S1 and S2 sites (62), although to date, no crystal structures have been reported of SARS-CoV PL pro in complex with a di-Ub substrate. Similarly, no such structural data have been obtained  nsp3 (B, D, and F). Upon induction of the innate immune response with RIG-I (2CARD) and IRF3 (5D) , cells were transfected with the PL pro (0, 150, 350, or 500 ng) or nsp3 (0, 350, 500, of 1000 ng) constructs. Upon induction with MAVS, cells were transfected with the PL pro (0, 50, 75, 100 or 150 ng) or nsp3 (0, 150, 350 or 500 ng) constructs. At 16 h post-transfection, cells were lysed, and luciferase activity was measured. All experiments were repeated independently at least four times. Significance was evaluated using an unpaired two-tailed Student's t test; p values of Ͻ0.05 were considered significant. Bars, mean; error bars, S.D. Western blotting was used to verify expression of MERS-CoV PL pro and nsp3.  for MERS-CoV PL pro , and thus future structural studies are necessary to determine precisely how MERS-CoV PL pro recognizes poly-Ub substrates and whether the preferences observed in expression systems can be confirmed in situations representative of an infection. In addition to deconjugating Ub, MERS-and SARS-CoV PL pro also recognize the antiviral Ubl molecule ISG15 (16,17). In the absence of a crystal structure of a DUB from the USP family in complex with ISG15, it is difficult to predict which regions of PL pro may be specifically responsible for ISG15 binding. However, it is interesting to note that both the palm and fingers domains of the SARS-CoV PL pro domain (62) and the cellular USP21 (57), respectively, have been implicated in ISG15 recognition, probably through additional interactions between PL pro and the N-terminal Ubl fold of ISG15. Future structural work is necessary to identify the specific determinants of ISG15 recognition by MERS-CoV PL pro . Structureguided mutagenesis of MERS-CoV PL pro to selectively disrupt deISGylation without affecting polyprotein cleavage would further expand our insights into the role of this additional activity in coronaviral immune evasion. The specific removal of DUB and potentially deISGylating activity from viral proteases that suppress the host innate immune response may open new avenues to engineer attenuated viruses for use as modified-live virus vaccines.