key: cord-0712186-bv42mbnp
authors: Piñón, Josefina D.; Mayreddy, Ravi R.; Turner, Julie D.; Khan, Farah S.; Bonilla, Pedro J.; Weiss, Susan R.
title: Efficient Autoproteolytic Processing of the MHV-A59 3C-like Proteinase from the Flanking Hydrophobic Domains Requires Membranes
date: 1997-04-14
journal: Virology
DOI: 10.1006/viro.1997.8503
sha: ae26337e6fe2ba4e94f90b8884854b18eabb6dfe
doc_id: 712186
cord_uid: bv42mbnp

Abstract The replicase gene of the coronavirus MHV-A59 encodes a serine-like proteinase similar to the 3C proteinases of picornaviruses. This proteinase domain is flanked on both sides by hydrophobic, potentially membrane-spanning, regions. Cell-free expression of a plasmid encoding only the 3C-like proteinase (3CLpro) resulted in the synthesis of a 29-kDa protein that was specifically recognized by an antibody directed against the carboxy-terminal region of the proteinase. A protein of identical mobility was detected in MHV-A59-infected cell lysates.In vitroexpression of a plasmid encoding the 3CLpro and portions of the two flanking hydrophobic regions resulted in inefficient processing of the 29-kDa protein. However, the efficiency of this processing event was enhanced by the addition of canine pancreatic microsomes to the translation reaction, or removal of one of the flanking hydrophobic domains. Proteolysis was inhibited in the presence ofN-ethylmaleimide (NEM) or by mutagenesis of the catalytic cysteine residue of the proteinase, indicating that the 3CLpro is responsible for its autoproteolytic cleavage from the flanking domains. Microsomal membranes were unable to enhance thetransprocessing of a precursor containing the inactive proteinase domain and both hydrophobic regions by a recombinant 3CLpro expressed fromEscherichia coli.Membrane association assays demonstrated that the 29-kDa 3CLpro was present in the soluble fraction of the reticulocyte lysates, while polypeptides containing the hydrophobic domains associated with the membrane pellets. With the help of a viral epitope tag, we identified a 22-kDa membrane-associated polypeptide as the proteolytic product containing the amino-terminal hydrophobic domain.

along with helicase and zinc finger motifs. ORF1a encodes three predicted proteinase domains. Two are cys-The murine coronavirus, mouse hepatitis virus strain teine proteinases (PLP-1 and PLP-2) that are distantly A59 (MHV-A59), has a positive-stranded RNA genome related to the cellular proteinase papain, and the third is that is approximately 31 kb long (Bonilla et al., 1994) . a poliovirus 3C-like proteinase (3CLpro). A growth-factor-Viral replication begins with the translation of the genolike region is also predicted to be encoded at the 3 end mic RNA upon entry into permissive cells. This results of ORF 1a. Several potential proteinase cleavage sites in the production of the viral RNA-dependent RNA polyhave been identified throughout both ORFs, suggesting merase that is believed to be encoded in gene 1 at the that the polyprotein products of gene 1 are processed by 5 end of the viral genome (Bonilla et al., 1994; Lee et these viral-encoded proteinases to generate individual al., 1991) . Approximately 70% of the coding potential of polypeptide products (Lee et al., 1991) . It has been shown the viral genome is contained within the two overlapping that the inhibition of proteolytic activity results in the rapid open reading frames (ORF 1a and ORF 1b) of this first shutoff of MHV RNA synthesis (Kim et al., 1995) . The locus (Bonilla et al., 1994; Bredenbeek et al., 1990) . Toproteolytic processing of the gene 1 polyprotein, theregether, translation of ORFs 1a and 1b via a frameshift fore, is essential for MHV replication. mechanism could result in a polyprotein of greater than

The 3CLpro domain is present in all coronavirus ge-800 kDa. Several functional domains are predicted to nomes studied to date (Bonilla et al., 1994;  Boursnell et reside within these two ORFs (Bonilla et al., 1994; Herold al., 1987; Eleouet et al., 1995; Herold et al., 1993; Lee et et al., 1993; Lee et al., 1991) . The predicted RNA-depenal., 1991) . The 3CLpro spans a 303-amino acid region, dent RNA polymerase motif (SDD) is encoded in ORF 1b between Ser3334 and Glu3636, and is flanked by two hydrophobic, and potentially membrane-spanning, domains (Bonilla et al., 1994; Lee et al., 1991; growth-factor-like regions at their junctions with other domain (Bonilla et al., 1994; Lee et al., 1991) . Preimmune serum was also collected from rabbits. These antibodies protein domains (Fig. 1) . Several of these predicted sites have recently been demonstrated to be cleaved by the were partially purified by passage through a protein A-Sepharose CL-45 column. The sera were diluted in Tris-3CLpro of several coronaviruses (Grotzinger et al., 1996; Lu et al., 1995; Tibbles et al., 1996; Ziebuhr et al., 1995) . saline buffer (50 mM Tris, pH 8.6, 150 mM NaCl, 0.02% sodium azide) and applied to the column. Unbound pro-These observations suggest an important role for the 3CLpro in the maturation of these domains into functional teins were removed by washing the column with 50 mM Tris-HCl buffer, pH 7.2. Antibodies were eluted with 0.1 proteins.

We have investigated the processing activity of the M citric acid (pH 3.0), neutralized with 1 M Tris, and dialyzed in 11 phosphate-buffered saline, pH 7.2 (20 mM MHV-A59 3CLpro using an in vitro transcription-translation system. In our studies, we made use of a pET21-sodium phosphate, 150 mM NaCl). Protein concentration was determined by UV spectroscopy. based plasmid that encodes most of the amino-terminal hydrophobic domain (HD1), the entire 3CLpro domain,

The antipeptide antibody 12CA5 is directed against the epitope YPYDVPDYA derived from the influenza virus and approximately one-half of the carboxy-terminal hydrophobic domain (HD2). A 29-kDa protein, representing hemaglutinin (HA) protein (Kolodziej and Young, 1991) . The murine ascites fluid containing this antiserum the 3CLpro, can be efficiently released from the flanking hydrophobic sequences, but only if the in vitro reactions (BAbCO) was provided by Dr. Michael Malim (Philadelphia, PA). are supplemented with canine microsomal membranes. This 29-kDa proteinase can also be identified in lysates from MHV-A59-infected cells. The requirement for mem-Plasmids branes is not surprising, given the hydrophobic nature of the sequences flanking the proteinase domain (see The plasmids used in this study are illustrated in Figs. 1C and 5A. The cDNA 917-919b (Pachuk et al., 1989) Fig. 5A ). The IBV 3C-like proteinase has recently been reported to require microsomal membranes for pro-had been previously cloned into pACYC5.3 to yield the plasmid pACYC 5.3-917/919. An SpeI-SacI fragment cessing activity in vitro (Tibbles et al., 1996) . In contrast, recent studies on the in vitro activity of the MHV-A59 from pACYC 5.3-917/919 was cloned into the NheI and SacI sites of pET21a (Novagen), resulting in pET21-3CLpro did not report a requirement for membranes (Lu et al., 1996 . We show here that the membrane HD1.3C.HD2. This plasmid encodes ORF 1a amino acids from Ser3149 to Leu3783. requirement is related to the presence of the two flanking hydrophobic domains. We have also expressed the MHV-

The plasmid pET21-3C was constructed by polymerase chain reaction (PCR) amplification of the predicted 3CLpro A59 3CLpro in E. coli and have used the partially purified recombinant fusion proteinase in in vitro trans cleavage domain using pET21-HD1.3C.HD2 as the template and the primers F3CP (5-GGAATTCCATATGTCTGGTATAGTG-assays using radiolabeled precursors generated by transcription-translation as substrates for the proteinase.

AAGATG-3) and R3CP (5-GAATTCCTCGAGTTACTGTAG-CTTGACACCAGCTAG) which introduces the restriction Unlike the autocatalytic cis processing observed during transcription-translation, posttranslational trans cleav-sites for NdeI and AvaI at the 5 and 3 ends of the amplified product, respectively, as denoted by the underlined letters age of precursors containing an inactive 3CLpro, as well as portions of both hydrophobic domains, was not depen-in the primer sequences. The PCR fragment was digested with NdeI and AvaI and cloned into the corresponding sites dent on membranes.

in the pET21a vector. The resulting plasmid encodes an additional methionine residue immediately upstream of the MATERIALS AND METHODS 3CLpro coding region.

The plasmid pET21-HD1.3C.HD2 was digested with the restriction enzyme HindIII. The larger DNA fragment DBT cells were maintained in Dulbecco's modified eaof approximately 7312 nucleotides was religated regle medium supplemented with 10% fetal calf serum sulting in the plasmid pET21-HD1.3C.Hind3. This plasmid (DMEM/10% FCS). The MHV-A59 isolate used in this contains a 93 amino acid truncation in HD2 compared study was originally provided by Dr. Julian Leibowitz (Colto the parental construct pET21-HD1.3C.HD2. lege Station, TX).

A region of MHV-A59 gene 1 from nucleotides 9912 to 10661 was PCR amplified from pET21-HD1.3C.HD2 using Antisera the primers FSP 9912-9930 (5-GGAATTCCATATGGCC-AAAATTGGTACCGAGGTT-3) and RSP 10661-10644 Rabbit antiserum UP313 and preimmune serum were raised by Cocalico (Reamstown, PA). Antiserum UP313

(5-AACATATCCTACAGAACC-3), digested with NdeI and BamHI, and cloned into the corresponding sites of is directed against the peptide 3618-EDELTPSDVYQQ-LAGVKLQ-3636, corresponding to the carboxy-terminal-pET21-3C resulting in the plasmid pET21-NX.3C. An NdeI-BamHI fragment from pET21-HD1.3C.HD2 was most 19-amino acid residues of the predicted 3CLpro subsequently subcloned into the NdeI and BamHI sites A region containing the C3478A mutation in the 3CLpro coding region was PCR amplified from pET21-of pET21-NX.3C to yield pET21-HD1.3C. Similarly, the region from nucleotides 10599 to 11595 of MHV-A59 ORF HD1.3C.HD2 C3478A using the primers FSP 10599 -10622 and R3CP. Following digestion with BamHI and 1a was PCR amplified from pET21-HD1.3C.HD2 using the primers FSP 10599-10622 (5-CGTAGTAGCCAT-AvaI, the amplified fragment was cloned into the corresponding sites in pET21-NX.3C to yield pET21-NX.3C ACCATAAAGGGC-3) and RMP 11595-11562 (5-CCT-CTTCCTCGAGATTGGCTCCAAAATACCACA-3). Follow-C3478A. ing digestion with BamHI and AvaI, the PCR fragment Cell infection, cell lysis, and immunoblot analysis was cloned into the BamHI and AvaI sites of pET21-3C. That fragment was later replaced by a BamHI-AvaI DBT cells were grown to confluency and infected with fragment from pET21-HD1.3C.HD2, resulting in the plas-MHV-A59 at a multiplicity of infection of 10 PFU/cell. A mid pET21-3C.HD2.

parallel culture was mock infected. Syncytia formation, The coding sequences representing the MHV-A59 observed 8-9 hr postinfection, was used as a visual 3CLpro domain were PCR amplified from pET21indicator of viral protein expression. At 9 hr postinfection HD1.3C.HD2 using the primers FSP 3C (5-GCGthe cells were collected and lysed in buffer containing CCGGAATTCTCTGGTATAGTGAAGATGGTGTCG-3 ) 10 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1% NP-40, and and RSP 3C (5-GCTCTAGAGCTTACTGTAGCTTGACAthe proteinase inhibitors phenylmethylsulfonyl fluoride CCAGC-3 ). The PCR amplified fragment was then di-(PMSF) and aprotinin (100 and 20 mg/ml, respectively) gested with EcoRI and XbaI (denoted by the underlined for 10 min on ice. The intact nuclei were pelleted and sequences in the primers) and cloned into the correthe clarified extracts stored at 080Њ. Equal amounts of sponding sites of the pMALC2 vector (New England protein were analyzed by sodium dodecyl sulfate-poly-Biolabs), resulting in the plasmid pMAL-3C.wt, which acrylamide gel electrophoresis (SDS-PAGE). The sepaencodes the 3CLpro fused to and downstream of the rated proteins were transferred to polyvinylidene fluoride maltose binding protein (MBP). The plasmid pMAL-3C.mut, (PVDF) membranes (Millipore). The membranes were inencoding the 3CLpro containing the inactivating mutation cubated in Tris-buffered saline (10 mM Tris-HCl, pH 8.0; C3478A, was constructed in a similar manner using the 150 mM NaCl) containing 0.1% Tween 20 and 5% nonfat construct pET21-HD1.3C.HD2 C3478A (see below) as the milk (TBST/milk). Following this blocking step the memtemplate for PCR amplification.

branes were incubated with either antiserum UP313 or the preimmune serum as specified for each experiment.

The membranes were then washed three times in 11 TBST solution without milk and then incubated with horse The vector pET21a-HindIII (0) was constructed by diradish peroxidase-conjugated anti-rabbit IgG (Cappel) for gesting the plasmid pET21a with the restriction enzyme another hour. Antigenically active polypeptides were vis-HindIII, filling in the recessed ends using the Klenow ualized using phosphatase coupled antibody (ECL kit, fragment of DNA Polymerase I, and blunt-end ligation of Amersham) and exposed to X-ray film. the filled-in DNA ends. A BglII-SalI fragment from pET21-HD1.3C.HD2 was cloned into the corresponding sites

In vitro transcription and translation in pET21a-HindIII (0) resulting in the construct pET21-HD1.3C.HD2-HindIII (0).

Expression of the plasmid DNAs was carried out using the TnT rabbit reticulocyte lysate coupled tran-A BamHI-HindIII fragment from cDNA 917 (Pachuk et al., 1989) was cloned into M13mp18 utilizing the BamHI scription -translation system (Promega) as previously described . Where indicated, 5 Eq and HindIII sites in the replicative form (RF) of the DNA to give rise to the recombinant plasmid M13mp18B 2 H. of canine pancreatic microsomal preparations (Promega) was also added to the translation reaction. The A 345-bp fragment containing a mutation in the catalytic cysteine was PCR amplified using the primers FMP C3478A cysteine proteinase inhibitor N-ethylmaleimide (NEM) (0.025 mg/ml) was added to the reaction where indi-(5-ATGTGGATCCGCCGGTTCTGTG-3) and RSP 10980-10959 (5-TGGCCTGTCATAGAAGCAAGCGC-3). This cated to test its effect on processing. The incorporation of [ 35 S]methionine into acid precipitable counts was fragment was purified and digested with the restriction enzymes BamHI (denoted by the underlined sequence) and used as an indicator of protein synthesis. Equivalent amounts of acid precipitable counts were directly ScaI and cloned back into the corresponding sites in M13mp18B 2 H to yield M13mp18B 2 H C3478A. A cysteine to analyzed by SDS -PAGE or immunoprecipitated with UP313 prior to electrophoresis. Radioimmunoprecipi-alanine mutation (TGC to GCC) was confirmed by sequencing of the RF DNA. A BamHI-HindIII fragment from tations (RIP) were carried out as described previously Denison et al., 1991) . M13mp18B 2 H C3478A was cloned into the corresponding sites in pET21-HD1.3C.HD2-HindIII (0). The resulting con-Time course analysis of processing was carried out by both pulse -labeling and pulse -chase assays. struct was named pET21-HD1.3C.HD2 C3478A. Pulse -label reactions were performed as described (New England Biolabs). Briefly, E. coli DH5a cells transformed with either the plasmid pMal-3C.wt or pMal-above in reaction volumes of 50 ml in the absence and presence of microsomal membranes. At specific time 3C.mut were induced with isopropyl-thio-b-D-galactoside (IPTG) at a final concentration of 0.3 mM for 3-4 hr at points between 0 and 300 min, 5-ml aliquots were removed, added directly to 21 Laemli buffer, and ana-30Њ. Cells were harvested and lysed by sonication. Cell debris were pelleted by centrifugation at 9000 g for 30 lyzed by SDS -PAGE. When pulse -chase assays were performed, the translation reactions were terminated min. The crude lysates were loaded onto an amylose column equilibrated with column buffer (20 mM Tris-by the addition of cyclohexamide (0.5 mg/50-ml reaction), excess cold methionine (2 mM ), and RNase A HCl, pH 7.4, 200 mM NaCl, 1 mM EDTA, and 1 mM DTT) at a flow rate of 1 ml/min. The column was washed with (0.16 mg/ml) (Bonilla et al., 1997) .

To determine membrane binding, plasmid DNAs were 12 column volumes of column buffer and the MBP-3CLpro fusion protein was eluted with column buffer con-transcribed and translated in the presence of microsomes. Following translation, membranes were pelleted taining 10 mM maltose. The concentration of the fusion protein was estimated on a 10% SDS-polyacrylamide gel by centrifugation at 12,000 g for 15 min at 4Њ (Echeverri and Dasgupta, 1995) and then resuspended in 11 low against known concentrations of bovine serum albumin (BSA) protein. The recombinant proteinase was stored at salt RIP buffer (10 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1% NP-40, 1% sodium deoxycholate, 1% SDS). Volumes 080Њ in column buffer supplemented with 20% glycerol. of soluble and membrane fractions containing equivalent Posttranslation proteolytic assays counts per minute were directly analyzed by SDS-PAGE. The amount of proteins analyzed in this manner were Radiolabeled substrates were generated using the not representative of the actual ratio of proteins in the TnT rabbit reticulocyte system as described above. Lysoluble versus the pellet fraction, with the proteins in the sate volumes containing equivalent counts per minute pellet fraction being slightly overrepresented. We estiwere incubated with approximately 3-4 mg of the recommated the percentage of total proteins partitioned into binant proteinase or an equivalent volume of column each fraction by calculating the amount of acid precipitabuffer/20% glycerol for 12-16 hr at 30Њ. The processed ble counts in each fraction as a percentage of the total products were analyzed by SDS-PAGE. acid precipitable counts of both fractions.

Overexpression and partial purification of the recombinant 3CLpro

The HA epitope derived from the influenza virus hemaglutinin (HA) protein was introduced into pET21-HD1.3C by Expression and purification of the recombinant 3CLpro were carried out according to manufacturer's instructions PCR amplification of pET21-HD1.3C.HD2 using the primers . The p29 protein is therefore specifically expressed from the 3CLpro encoding plasmid, TACAGAACC-3). FSP HA-MHV9648 contains a 40-nucleotide overhang encoding the restriction site NdeI, as denoted pET21-3C. Since the UP313 antibody recognizes the carboxy-terminal end of the 3CLpro, p29 most likely repre-by the underlined sequences, as well as the HA epitope, denoted by the bold-faced sequences. The PCR amplified sents the full-length 3CLpro translation product.

To determine if p29 represents the mature 3CLpro, we fragment was digested with NdeI and BamHI and cloned back into pET21-HD1.3C resulting in pET21-HA.HD1.3C. looked for the expression of a similar protein in vivo.

Lysates from MHV-A59-infected cells were separated by Membrane association assays were performed using this plasmid. Equivalent counts per minute from soluble and SDS -PAGE, transferred to a PVDF membrane, and analyzed by Western blotting with UP313 (Fig. 2B) . A 29-membrane fractions were immunoprecipitated with the antibody 12CA5 as previously described  kDa protein was recognized specifically by the UP313 antibody in lysates of infected cells (lane 4), but not in Denison et al., 1991) . Immunoprecipitated proteins were then analyzed by SDS-PAGE. uninfected cells (lane 3). Preimmune serum was unable to detect p29 in infected or uninfected cell lysates (lanes 1 and 2). These in vivo results provide further support RESULTS that the 29-kDa protein detected in in vitro transcription-Identification of the 3CLpro in vitro and in infected translation experiments represents the mature 3CLpro. cells

Demonstration of 3CLpro processing in vitro A plasmid encoding the core 3CLpro of MHV-A59, pET21-3C, was in vitro-transcribed and translated in the To study the processing activity of the 3CLpro in vitro, we constructed the plasmid pET21-HD1.3C.HD2 ( . This plasmid encodes a precursor protein with a predicted molecular weight of approximately 72 kDa that analyzed by SDS-PAGE. In vitro expression of pET21-3C resulted in the production of a protein with an apparent contains the 3CLpro as well as the two QS sites at its junctions with the flanking hydrophobic domains. Pro-migration of 29 kDa (predicted molecular weight, 33 kDa) that is immunoprecipitated by UP313 (Fig. 2A, lane 3) .

cessing at these two sites would result in the release of the 3CLpro from the flanking hydrophobic sequences. This 29-kDa protein (p29) is neither present in translation 3CLpro when the translation reaction was supplemented with microsomes (lane 2). These results suggest a requirement for microsomal membranes for the processing of the MHV-A59 3CLpro from the flanking hydrophobic regions.

The catalytic residues of the 3CLpro have been previously identified . Amino acid residues His3374 and Cys3478 were shown to be essential for proteinase activity. Using site-directed mutagenesis a mutant of pET21-HD1.3C.HD2 was created in which the catalytic residue Cys3478 was replaced by alanine (C3478A). This C3478A mutant was utilized in order to determine whether or not the observed processing of p29 was dependent on the proteinase activity of the 3CLpro itself. The cysteine proteinase inhibitor, NEM, was also used for this purpose. In vitro transcriptiontranslation of the wild-type pET21-HD1.3C.HD2 in the presence of the inhibitor resulted in a decreased production of the 29-kDa proteinase product (Fig. 3, lane 3) . In translation reactions with the C3478A mutant, p29 production was abolished (lane 4). These results demonstrate that the 29-kDa 3CLpro results from its own autocatalytic processing from the flanking domains.

Translation of pET21-HD1.3C.HD2 in the absence of after 90 min of translation. In contrast, p29 was detected by 45 min and its levels increased steadily in translation reactions performed in the presence of microsomes. This The 3CLpro could then be identified based on its molecular weight and its ability to be immunoprecipitated by the suggests that while microsomal membranes are not absolutely required for the in vitro processing of p29, their antibody UP313. In vitro transcription-translation reactions of the pET21-HD1.3C.HD2 plasmid by itself, how-presence in the translation reaction can greatly enhance the processing activity of the 3CLpro. ever, did not result in appreciable processing of the 29-kDa 3CLpro (Fig. 3, lane 1) . Only a precursor, with an Pulse-chase experiments performed either in the absence or presence of microsomal membranes gave simi-apparent molecular weight of 50 kDa, was observed. Due to the hydrophobic nature of the sequences flanking the lar results. The 29-kDa proteinase could always be detected at much shorter times in those reactions carried 3CLpro, we decided to perform our in vitro reactions in the presence of canine pancreatic microsomal mem-out in the presence of microsomes (data not shown). We also studied the effect of adding microsomal mem-branes. The [ 35 S]methionine-labeled translation products generated from pET21-HD1.3C.HD2 in the presence of branes only after the pulse-labeling period. In vitro transcription and translation of pET21-HD1.3C.HD2 was car-microsomal membranes were immunoprecipitated with UP313. Analysis of the immunoprecipitated products by ried out in the absence of microsomal membranes. After a 90-min pulse period, the translation reaction was electrophoresis revealed the production of the 29-kDa This plasmid encodes the amino-terminal hydrophobic domain, HD1, and 3CLpro, like pET21-HD1.3C.HD2, but only 52 residues of the carboxy-terminal hydrophobic domain, HD2. In the construct pET21-HD1.3C, the entire HD2 was deleted whereas in pET21-3C.HD2, all of HD1 from the parental construct was removed.

The results obtained from in vitro expression of these constructs are shown in Fig. 5B . In all cases the fulllength translation products detected migrated faster than expected from their predicted molecular weights. pET21-HD1.3C.Hind3, pET21-HD1.3C, and pET21-3C.HD2 gave rise to primary translation products with apparent molecular weights of 46, 43, and 37 kDa, respectively (predicted molecular weights 62, 55, and 51 kDa). The parental construct, pET21-HD1.3C.HD2, is shown in lanes 1 and 2 for comparison. As expected, this construct demonstrated a requirement for microsomal membranes in order for the 29-kDa 3CLpro to be efficiently released from the precursor protein. The construct, pET21-HD1.3C.Hind3, con- hydrophobic domain, while both the pET21-HD1.3C.HD2 and pET21-HD1.3C.Hind3 constructs encode parts of both hydrophobic domains. It seems therefore, that the stopped by the addition of cyclohexamide, RNase, and requirement for membranes is directly related to having excess cold methionine. The reaction was then divided hydrophobic sequences flanking both ends of the prointo two equal volumes and microsomal membranes teinase. were added to one half of the reaction. Aliquots were removed at the times indicated during the 180-min chase Trans proteolytic activity of the E. coli-generated and analyzed directly by SDS-PAGE. In this experiment, recombinant 3CLpro we were able to detect a low level of p29 at the end of the 90-min pulse period (Fig. 4C) . This low level of p29

The enhancement of proteolytic processing in the remained constant throughout the entire chase period presence of microsomal membranes could be due to the and did not increase even when the reactions were supefficient presentation of the cleavage sites to the 3CLpro plemented with microsomes posttranslationally. This by the membranes. Alternatively, the membranes may suggests that in order to exert its enhancing effect on contribute to the proper folding of the proteinase into an the proteolytic processing of the 3CLpro the microsomal active conformation. To distinguish between these two membranes have to be present in the reaction cotranslapossibilities, we conducted in vitro trans cleavage assays tionally.

as previously described using substrates and enzymes both generated in a coupled transcription-Determinants of membrane requirement translation system. In these experiments we used radiolabeled translation reactions of the plasmid pET21-The role of membranes in the efficient processing of the 3CLpro was further investigated using a series of HD1.3C.HD2 C3478A, encoding an inactivating mutation in the 3CLpro domain, as substrate. Unlabeled transla-truncated constructs which encode shorter portions of the surrounding hydrophobic regions. These constructs tion reactions of pET21-3C were used as the source of enzyme. However, we were unable to demonstrate trans are schematically represented in Fig. 5A and were all derived from the parental construct pET21-HD1.3C.HD2. cleavage using this system. A recombinant 3CLpro was subsequently tested in trans cleavage assays using ra-The plasmid pET21-HD1.3C.Hind3 encodes a polyprotein carrying a 93-amino acid truncation from the carboxy-diolabeled substrates generated in the absence or presence of microsomal membranes. The 3CLpro domain terminal end compared to that of the parental construct. 1 and 2, pET21-HD1.3C.HD2; lanes 3 and 4, pET21-HD1.3C.Hind3; lanes 5 and 6, pET21-HD1.3C; lanes 7 and 8, pET21-3C.HD2 . The recombinant 3CLpro, we did not detect any processing from this substrate. The precursor protein was then used as a substrate in a posttranslation proteolytic assay with the recombinant 3CLpro. In the absence of microsomal membranes, the precursor generated from pET21-HD1.3C.HD2 C3478A was processed by the recombinant proteinase, albeit inefficiently (lane 9). The addition of microsomal membranes during translation did not enhance the ability of this precursor to be processed by the recombinant 3CLpro (lane 10). This suggests that membranes are not involved in the presentation of the QS cleavage sites to the exogenously added recombinant proteinase. It is clear, however, that the NX.3C C3478A precursor is much more efficiently cleaved by the recombinant proteinase. This suggests that the presence of the hydrophobic regions in the HD1.3C.HD2 precursor contributes to its inefficiency as a substrate for trans cleavage and that this inefficiency is not overcome by the addition of membranes.

We also used the wild-type precursor HD1.3C.HD2, carrying an active 3CLpro, as a substrate for the recombinant proteinase. This precursor, by itself, is able to undergo autoproteolytic cleavage giving rise to the 29-kDa proteinase even in the absence of microsomes (lane 3), during the 16-hr incubation. This processing event is TnT system. Substrate volumes containing equivalent counts per miproduction (lanes 5 and 6), suggesting that the mechanute were incubated with 3-4 mg of the recombinant 3CLpro or with nism by which the 29-kDa proteinase is released from an equivalent volume of column buffer/20% glycerol (0// enzyme) for the flanking domains involves an autoproteolytic cis 12-16 hr at 30Њ. Lanes 1-2, substrate generated from pET21-NX.3C C3478A; lanes 3-6, substrate generated from pET21-HD1.3C.HD2 concleavage. taining an active 3CLpro; lanes 7-10, substrate generated from pET21-HD1.3C.HD2 C3478A.

The hydrophobic nature of the sequences surrounding (Ser3334-Gln3636) was overexpressed in, and partially purified from, E. coli as part of a fusion protein with the the 3CLpro, coupled with the prediction that these domains may be membrane spanning, prompted us to in-maltose binding protein (MBP). Figure 6 shows the ability of the recombinant MBP-3CLpro fusion proteinase to effi-vestigate the ability of these hydrophobic sequences to interact with the microsomal membranes. We conducted ciently cleave an ORF1a protein substrate generated from the plasmid pET21-NX.3C C3478A (lanes 1 and 2) . membrane association assays, wherein the microsomal membranes and any attached proteins were separated This substrate, which encodes the carboxy-terminal half of HD1 and the entire 3CLpro domain, inactivated by the from the rest of the lysates by a brief centrifugation period at 4Њ. Following this treatment, the 29-kDa 3CLpro from C3478A mutation, was cleaved in trans by the recombinant proteinase resulting in the production of the 29-kDa translation of pET21-HD1.3C.HD2 was present in the soluble fraction, with residual amounts associated with the 3CLpro. To investigate the effect of microsomal membranes on the presentation of the cleavage sites to the pellet (Fig. 7, lanes 1 and 2) . We were unable to detect p29 in the soluble or pellet fractions of translations car-enzyme supplied in trans, we translated the plasmid pET21-HD1.3C.HD2 C3478A in the absence (lane 7) and ried out using the C3478A mutant (lanes 3 and 4) , again confirming that this observed processing is dependent presence (lane 8) of microsomes. In the absence of the 0// signs above each lane indicate the absence or presence of microsomal membranes in the translation reaction. The molecular weight of the protein markers are indicated on the left. Shown on the right are the full-length translation products generated by each construct and the 29-kDa 3CLpro. In addition to a full-length translation product of 43 kDa and the 29-kDa 3CLpro, translation of pET21-HD1.3C also yields another major translation product of 35 kDa. p35 is most likely a product of internal initiation since its synthesis is not inhibited by NEM and it contains the carboxy-terminal end of the 3CLpro as evidenced by its immunoprecipitation by UP313 (data not shown). mediately upstream of HD1 in the construct pET21-HD1.3C. The resulting plasmid, pET21-HA.HD1.3C, was used in membrane association assays and compared to pET21-HD1.3C. In translation reactions using both plasmids the 3CLpro was located almost exclusively in the soluble fraction of the reactions (Fig. 8, lanes 1 and 3) . In the pellet fraction, both p22 and p15 were present in each case (lanes 2 and 4) . Immunoprecipitation of the soluble and pellet fractions with the antibody 12CA5 prior to SDS-PAGE analysis revealed the full-length precursor (43 kDa) in the soluble fraction and the 22-kDa protein, as well as the full-length precursor, in the pellet fraction of the pET21-HA.HD1.3C translation reaction (lanes 7 and 8). No proteins were detected upon immunoprecipitation of the soluble and pellet fractions of the pET21-HD1.3C translation reaction with the 12CA5 antibody (lanes 5 and p15 protein may simply represent a degradation product of p22. These observations represent the first demonstration of a physical association between HD1 and mem-on the activity of the 3CLpro. In addition, we were able to detect three other polypeptides with molecular weights of branes for a coronavirus in vitro. approximately 37, 22, and 15 kDa that were present in the pellet fraction of translations with the wild-type construct.

DISCUSSION These bands were not present in the pellet fraction of translations using the C3478A mutant, suggesting that A common strategy adopted by positive-stranded RNA viruses is the expression of a large polyprotein precursor, these are products of processing by the 3CLpro. That these bands are present almost exclusively in the pellet followed by the processing of this precursor polyprotein into individual, mature viral proteins. Responsible for fraction implies that they are associated with the membranes and may therefore represent the fully processed these processing events are several viral-encoded proteinases whose significance in viral genomic expression hydrophobic domains or partially processed products containing either hydrophobic domain. In particular, p37 and replication have been well documented (Dougherty and Semler, 1993) . For the murine coronavirus, two such has an electrophoretic migration identical to the precursor protein encoded by the plasmid pET21-3C.HD2. This viral proteinase activities have been identified. Both the PLP-1 and 3CLpro, whose proteolytic activities have band, therefore, likely represents a partially processed product containing the 3CLpro and the carboxy-terminal been demonstrated in vitro (Baker et al., 1990 (Baker et al., , 1993 Lu et al., 1995) , are encoded as part hydrophobic domain.

We have also performed membrane association of the large polyprotein precursor that results from the translation of the viral RNA genome. Flanked by hy-assays using our truncated constructs (Fig. 7) . In all cases, centrifugation of the lysates following in vitro drophobic domains, the serine-like proteinase, 3CLpro, is believed to be the principal proteinase responsible for translation led to the distribution of the 3CLpro almost exclusively in the soluble fraction of the reactions (lanes the processing of the large precursor polyprotein into individual, functional replicase proteins. As many as 11 5, 7, and 9). Interestingly, both the p22 and p15 proteins were also detected in the pellet fractions of transcrip-potential 3CLpro cleavage sites have been predicted flanking the potentially important domains encoded in tion-translation reactions using pET21-HD1.3C.Hind3 (lane 6) and pET21-HD1.3C (lane 8), but not pET21-gene 1 (Gorbalenya et al., 1989; Lee et al., 1991) . These include cleavage sites that flank the putative polymerase, 3C.HD2 (lane 10). This observation suggests that these two proteins are related to the first hydrophobic domain helicase, and growth factor-like regions as well as cleavage sites surrounding the 3CLpro. Processing at several (HD1) rather than the second (HD2).

In order to identify HD1 from among the processed of these sites has already been demonstrated in several coronaviruses (Grotzinger et al., 1996; Lu et al., 1995 ; products, we placed the influenza virus HA epitope im- 1 and 3) . This protein, however, is not immunoprecipitated by 12CA5 (lanes 5 and 7) , further supporting the suggestion that this protein is a product of internal initiation (see legend to Fig. 5B ). Tibbles et al., 1996; Ziebuhr et al., 1995) . For MHV-A59, 3C-like proteinase, however, this requirement for membranes is not absolute for the MHV-A59 3CLpro since the QS dipeptide flanking the N-terminus of the proteinase has been demonstrated to be processed by the extended incubations of the in vitro translation of pET21-HD1.3C.HD2 in the absence of microsomal membranes 3CLpro in vitro .

In this study, we identified the 3CLpro using an anti-produced low levels of p29 processing. However, their presence in the reaction cotranslationally greatly en-body directed against the last 19 amino acids at the carboxy-terminus of the proteinase domain. The 3CLpro

hanced the efficiency of p29 processing (Figs. 4A and 4B). This requirement for membranes suggests an inter-was identified from in vitro translations of several plasmids encoding the proteinase as part of a precursor action between the precursor protein and membranes in order to achieve the proper protein conformation neces-polyprotein, and also from lysates of MHV-A59-infected cells. In both cases, the 3CLpro migrated on SDS-poly-sary for efficient processing at the flanking QS cleavage sites resulting in the release of the 29-kDa proteinase. acrylamide gels as a 29-kDa protein. This is similar in size to, and is presumably the same protein as the pre-Moreover, the inability of microsomal membranes to enhance proteolytic activity posttranslationally suggests viously reported p27 3CLpro . Because the epitope recognized by the UP313 antibody is at the that this interaction occurs cotranslationally and may involve the insertion of the hydrophobic domains into the extreme end of the proteinase domain adjacent to the predicted QS cleavage site at the junction between the membranes. The membrane requirement is directly related to the 3CLpro and HD2, we are confident that this p29 protein represents the full-length 3CLpro.

presence of hydrophobic sequences flanking both termini of the 3CLpro. The translation reactions of pET21-We also demonstrate in this report a requirement for microsomal membranes for the efficient processing of HD1.3C and pET21-3C.HD2, constructs encoding only one of the flanking hydrophobic domains, produced simi-the 3CLpro from the flanking hydrophobic sequences. This membrane requirement has not been previously re-lar levels of p29 independent of the presence of microsomal membranes. In pET21-HD1.3C.Hind3, the addition ported for the MHV-A59 3CLpro. However, the 3C-like proteinase of IBV was reported to show an absolute re-of the first 52 amino acid residues of HD2 to the carboxyterminal end of pET21-HD1.3C resulted in the loss of the quirement for microsomes for its processing activity in vitro (Tibbles et al., 1996) . We report here that only when ability to process p29 in the absence of microsomes. These results are consistent with those reported pre-the translation of the pET21-HD1.3C.HD2 plasmid was carried out in the presence of microsomal membranes viously for the MHV-A59 3CLpro , wherein the construct used, pGpro, contained a KpnI-HindIII frag-was the processed p29 readily detectable. Unlike the IBV ment of MHV-A59 ORF 1a (see Fig. 5A ) that encoded respect to each other for autocatalytic cis processing by the 3CLpro to take place. the 3CLpro and portions of the surrounding sequences. Compared to our pET21-HD1.3C.Hind3 construct, pGpro

One challenging aspect of this study has been the proper identification of the precursor polyproteins as lacks the amino-terminal 87 amino acids representing the hydrophobic residues of HD1. The only hydrophobic well as all processed products, partial or complete. It has been our experience that these precursors migrate sequences encoded in this plasmid flank the carboxyterminal end of the 3CLpro and, based on our observa-on SDS -polyacrylamide gels with electrophoretic mobilities faster than that predicted based on their pri-tions, would not be expected to require microsomes for autoproteolysis. Not surprisingly, pGpro did not require mary amino acid sequence. This apparent conflict between the predicted and observed molecular weights the presence of microsomes for in vitro activity (Lu et al., 1996 . We speculate that when both termini of the of the precursor and processed proteins was previously reported in investigations of both the HCV 229E 3CLpro are flanked by hydrophobic sequences, these domains may aggregate together in the absence of mem- (Ziebuhr et al., 1995) and MHV-A59 3C-like proteinases . The observed molecular weight of branes in order to shield themselves from the aqueous environment. The amino acid sequences representing the 3CLpro itself, 29 kDa, is smaller than its predicted molecular weight of 33 kDa. Furthermore, each of the the 3CLpro that lie in the middle of this precursor inevitably become misfolded, resulting in an inefficient protein-3CLpro-encoding plasmids used in our study resulted in primary translation products with electrophoretic ase. Since precursor proteins with only one hydrophobic domain do not require membranes for processing by mobilities faster than expected. It is therefore difficult to positively identify all precursor and processed prod-the 3CLpro, this suggests that in such precursors the hydrophobic region alone may become misfolded. The ucts based solely on their apparent molecular weights without the benefit of having a panel of antibodies di-rest of the protein, however, may still be able to achieve a conformation that permits processing by the 3CLpro.

rected against these regions under study. We were able to positively identify the 3CLpro using a antipep-Hence, in such precursors, membranes are not necessary for proteinase activity.

tide antibody directed at the carboxy-terminal end of the proteinase. However, such antibodies are not, at This speculation is further supported by the observation that a recombinant 3CLpro supplied in trans is present, available for either of the hydrophobic domains. Instead, we placed an epitope (HA-) tag imme-able to efficiently process a precursor protein generated from the pET21-NX.3C C3478A plasmid which diately upstream of the HD1 in the construct pET21-HA.HD1.3C and used the 12CA5 antibody directed does not encode any hydrophobic sequences. However, the HD1.3C.HD2 C3478A precursor, containing against this epitope to unambiguously identify the precursor protein encoded by this plasmid. This precursor both hydrophobic domains, serves as a very inefficient substrate for the recombinant proteinase. Generating migrated with an apparent molecular weight of 43 kDa, faster than its predicted molecular weight of 55 kDa. the HD1.3C.HD2 C3478A precursor protein in the presence of microsomal membranes, however, does not Using the HA-tag the processing product corresponding to the amino-terminal hydrophobic domain was increase its efficiency as a substrate for the recombinant 3CLpro, demonstrating that membranes have no also identified in association with membranes upon expression of this plasmid in vitro. This is the first in effect on trans processing by the recombinant 3CLpro. Furthermore, the addition of the recombinant protein-vitro demonstration of a physical interaction between this amino-terminal hydrophobic domain and mem-ase to the HD1.3C.HD2 substrate in which the proteinase had not been inactivated did not increase the lev-branes. However, in such an analysis the nature of this interaction, whether peripheral or integral, cannot be els of p29 production from these precursors. These results, taken together, strongly suggests that the en-discerned. We estimated that approximately 37% of the total translated proteins becomes associated with hanced processing of the 3CLpro observed in the presence of membranes upon coupled transcription -trans-membranes (data not shown). This number is variable and seems to be independent of the hydrophobic con-lation of wild-type precursors, containing both substrate and active enzyme, cannot be attributed to an tent of the precursor protein, that is to say, increasing hydrophobic content of the precursor does not corre-increase in trans cleavage, but rather is due to an enhancement of cis processing by the 3CLpro. It is late with a greater partitioning into the membranes. Instead, this calculation may reflect the capacity of the also clear that the microsomes are not required as a cofactor for enzyme activity since the processing of microsomes to take in proteins or the fact that the microsomes may be a limiting reagent in these in vitro the NX.3C C3478A precursor protein takes place efficiently in the absence of microsomes. These data sup-reactions. Preliminary studies show that the hydrophobic domains and hydrophobic-containing precursors port our contention that membrane binding plays a role in keeping the entire HD1.3C.HD2 precursor, both remain in the membranes following alkali extraction (Mostov et al., 1981) to the same extent as the coro-substrate and enzyme, in a proper conformation with navirus S protein under the same conditions (data not

Identification of the catalytic sites of a papain-like cysteine proteinase of murine coronavirus

Intracellular localization of polypeptides encoded in mouse retic mobilities, it would not be surprising if HD2 migrated hepatitis virus open reading frame 1a

Characteristics of the poliovipected. rus replication complex

Structural and 3CLpro suggests a role for these domains not only in functional characterization of the poliovirus replication complex. J. the proper folding of the precursor protein, but also in Virol

Mouse hepatitis the localization of this region of the gene 1 polyprotein virus strain A59 RNA polymerase gene ORF 1a: heterogeneity among to a target organelle during coronavirus infection

Charactionarily related to the coronavirus-like superfamily (den terization of the leader papain-like proteinase of MHV-A59: identifica-Boon et al., 1991), it has recently been demonstrated that tion of a new in vitro cleavage site

Characterization of the proteolytic products of ORF 1b generated by the EAV a second cleavage site and demonstration of activity in trans by the nsp4 proteinase are localized to a membranous compartpapain-like proteinase of the murine coronavirus MHV-A59

The nsp4 proteinase is a chymotrypsin-like

Completion of the sequence of the genome of the coronavirus avian infectious bronchitis virus

The primary structure and exside of the 3C-like nsp4 proteinase domain (Snijder et pression of the second open reading frame of the polymerase gene of the coronavirus MHV-A59; a highly conserved polymerase is exal., 1994). Based on the accumulating evidence in EAV, it pressed by an efficient ribosomal frameshifting mechanism. Nucleic has been suggested that the replicase cleavage products Acids Res

Flavivirus chored in the membrane through these hydrophobic regenome organization, expression, and replication

The replication complexes

Equine arteritis virus is not a togavirus found to be membrane associated (Bienz et al., 1994, but belongs to the coronaviruslike superfamily

Thus, it is plausible that in MHV

Identification of polypeptides encoded in open reading frame 1b of the putative polymerase gene of the murine coronavirus replicase gene are targeted to certain membranous ormouse hepatitis virus A59

support of this idea, we have demonstrated

Expression of virus-encoded that the PLP-1 is localized to the Golgi apparatus during proteinases: functional and structural similarities with cellular en-MHV infection

Amino terminal regions of poliovirus 2C protein mediate membrane binding

Complete sequence (20 kb) of the polyprotein-encoding ACKNOWLEDGMENTS gene 1 of transmissible gastroenteritis virus

Alphavirus RNA The authors thank Xinhe Jiang and Xiurong Wang for their excellent replicase is located on the cytoplasmic surface of endosomes and lysosomes

Coronavirus genome: prediction of putative functional domains in the non-structural polyprotein by comparative amino acid supported in part by Merck and Training Grant AI-07325. P.J.B. was sequence analysis

Molecular cloning of the gene encoding the putative gene 1 of the human coronavirus HCV 229E

tide sequence of the human coronavirus 229E RNA polymerase locus

Proteolytic processing of the replicase ORF1a protein of equine arteritis virus

Coronavirus protein processing and RNA synthesis is inhibited by Virol

and Gorbathe cysteine proteinase inhibitor E64d

The arterivirus nsp4 protease is the prototype of a novel group of chymotrypsin-like enzymes, the 3C-like serine surveillance

The complete sequence (22 kilobases) of murine coronavirus gene 1 encoding the putative acterization in vitro of an autocatalytic processing activity associated with the predicted 3C-like proteinase domain of the coronavirus avian proteases and RNA polymerase

Intracellular and in vitro-transinfectious bronchitis virus

Processing of the equine arteritis virus replicase ORF1b protein: identification of cleavage products containing the

Identification and characterization of a serine-like proteinase of the murine coronavirus MHV-A59. putative viral polymerase and helicase domains

tional membrane integration of calcium pump protein without signal sequence cleavage