Dissection of Amino-Terminal Functional Domains of Murine Coronavirus Nonstructural Protein 3

Virus and cells.

Wild-type (WT), mutant, and revertant MHV strain A59 stocks were propagated in mouse 17 clone 1 (17Cl1) cells; plaque titrations and plaque purifications were carried out in mouse L2 cells. Monolayer cultures of both cell lines were maintained in Dulbecco's minimal essential medium (MEM) containing 10% fetal bovine serum. Spinner cultures of L2 cells, used for transfection of genomic RNA, were kept between densities of 1 × 10⁵ and 2 × 10⁶ cells/ml in Joklik's MEM containing 10% fetal bovine serum.

MHV mutant construction and analysis.

All MHV-A59 mutants in this study were created with an infectious cDNA system generously provided by Ralph Baric; the assembly of full-length cDNA and in vitro transcription of genomic RNA were carried out essentially as described previously (27). Viruses were generated by cotransfection of synthetic gRNA and MHV N mRNA through electroporation of suspension-grown L2 cells (17). Initial viral isolations were performed at 37°C for 1 to 5 days. The wild-type reference virus used throughout this work was previously obtained from isogenic wild-type full-length cDNA (17). To verify constructed viral mutants, RNA isolated from infected cell monolayers with Ultraspec reagent (Biotecx) was reverse transcribed with a random hexanucleotide primer and avian myeloblastosis virus reverse transcriptase (RT; Life Sciences). PCR amplification of cDNA was carried out with the Expand high-fidelity PCR system (Roche). For Sanger sequencing, RT-PCR products were purified with QIAquick spin columns (Qiagen). Whole-genome sequences (excluding 39 and 113 nucleotides [nt] at the 5′ and 3′ ends, respectively) were generated from 18 partly overlapping RT-PCR products spanning the genome. Equimolar amounts of amplicons were pooled, purified with an AMPure XP kit (Beckman Coulter), and quantitated on a Qubit fluorometer (Invitrogen). Tagged libraries of DNA fragments were prepared with a Nextera XT kit (Illumina), and 2- by 250-bp paired-end sequencing was performed on a MiSeq Sequencer (Illumina) using the MiSeq 500-cycle v2 kit. Viral genome sequences were fully assembled using SPAdes version 3.0.0 with the read correction mode active and K-mer sizes of 21, 33, 55, 77, 99, and 127.

To isolate revertants of the N3S mutant, 12 independent plaques obtained at 37°C were each used to start cultures that were grown at 37°C and thereafter serially passaged at 39°C in L2 cells at a low multiplicity of infection. When an increase in growth rate was noted, following 10 to 13 passages, titers of viral supernatants were determined on L2 cells at 39°C, and a single larger plaque was chosen from each independent sample and purified for further analysis.

Genetic constructions were all made via manipulation of pBar-A3 (17), a modified version of MHV cDNA clone A (27) that is here referred to as pA-WT. For pA-N3S, clone A of the N3S mutant, the BsiWI-BstBI fragment of pA-WT (running from the leader through nsp3 Ubl1) was replaced with a fragment that was synthesized by PCR from overlapping oligonucleotides. The replacement fragment removed half of nsp1 and all of nsp2 and created 30 point mutations in 5′ gRNA secondary structures, as described in detail in Results. For pA-SSN3, clone A of the SSN3 mutant, the BamHI-BstBI fragment of pA-WT (running from nsp1 codon 86 through nsp3 Ubl1) was replaced with a synthetic fragment encoding residues 86 to 127 of nsp1, His₇, and 58 residues of foot-and-mouth disease virus (FMDV) VP1-2A (from GenBank accession number X00871). Plasmid pA-SSN3x was constructed identically to pA-SSN3, except that the NPGP consensus autoprocessing sequence was changed to NLAP (28). An epitope-tagged version of pA-SSN3, designated pA-SSN3^HA, was built in two steps. First, a synthetic fragment encoding the nsp3 Ac region, followed by FLAG and hemagglutinin (HA) tags, was inserted between the SalI and SacI sites bounding the deleted Ac region of the previously described pA3Ac2, which had been used to generate the ΔAc2 mutant (17). Then the BstBI-SpeI fragment of this intermediate (running from nsp3 Ubl1 to PLP1) was used to replace the corresponding fragment of pA-SSN3. Four derivative plasmids were obtained from pA-SSN3^HA. One, pA-SSN3^HAmutP, was generated by a two-step PCR that replaced the SacI-SpeI fragment (running from the start to the middle of nsp3 PLP1) to create the PLP1 catalytic mutation C289A. For the initial PLP1 deletion construct, pA-SSN3^HAΔP1, the SacI-BstXI fragment encompassing PLP1 was removed from pA-SSN3^HA, and following the creation of blunt ends, the vector was reclosed. In a second deletion mutant, pA-SSN3^HAΔP2, the SacI-BstXI fragment of pA-SSN3^HA was again removed, and this time it was replaced with a synthetic fragment that moved the deletion boundaries inward, as specified in Results. For the third deletion construct, pA-SSN3^HAΔP3, the SacI-Acc65I fragment encompassing both PLP1 and ADRP was removed from pA-SSN3^HA, and following the creation of blunt ends, the vector was reclosed. Oligonucleotides for PCR and DNA sequencing were obtained from Integrated DNA Technologies. The overall compositions of constructed plasmids were confirmed by restriction analysis, and all ligation junctions and regions resulting from PCR amplification were verified by DNA sequencing.

Viral growth kinetics.

To measure growth kinetics, confluent monolayers of 17Cl1 cells (75 cm²) were inoculated at a multiplicity of 1.0 PFU per cell for 2 h at 37°C, with rocking every 15 min. Following the removal of inocula, monolayers were washed three times with phosphate-buffered saline (PBS), and incubation was continued in fresh medium at 37°C. Sample aliquots of medium were withdrawn at various times from 2 to 48 h postinfection, and infectious titers were subsequently determined in L2 cells.

Northern blotting.

RNA was extracted from infected cell monolayers with TRI-Reagent (Zymo) according to the manufacturer's instructions. Northern blotting of intracellular RNA was performed as previously described in detail (29) by using a PCR-amplified probe corresponding to the 3′-most 539 nucleotides of the N gene and the entire 3′ untranslated region (UTR) of the MHV genome. The probe was labeled with an AlkPhos Direct kit, and blots were visualized using CDP-Star detection reagent (GE Healthcare).

In vitro translation.

Capped mRNAs were produced with T7 RNA polymerase (mMessage mMachine; Ambion) by runoff transcription of BstZ17I-truncated pA-WT or SpeI-truncated pA-N3S, pA-N3SΔ1, pA-N3SΔ2, pA-SSN3, or pA-SSN3x. Synthetic transcripts (0.46 μg) were translated in 17 μl of rabbit reticulocyte lysate (Ambion) labeled with [³⁵S]methionine (MP Biomedicals); protein products were analyzed in 10% polyacrylamide gels by SDS-PAGE, followed by fluorography. Vector pA-N3SΔ1 was derived from pA-N3S by removal of the BsiWI-NcoI fragment and religation of the plasmid following the creation of blunt ends. The same strategy was used to make pA-N3SΔ2 by removal of the ApaI-NcoI fragment.

Western blotting.

Sets of lysates were prepared from 75-cm² confluent monolayers of L2 cells that were mock infected or infected with wild-type and mutant MHV at either a high or a low multiplicity of infection at 37°C. At 8 to 10 h postinfection (for infections begun at a multiplicity of 1 PFU per cell) or 17 to 27 h postinfection (for infections begun at a multiplicity of 0.01 PFU per cell), monolayers were washed twice with PBS and then lysed by the addition of 500 μl of 50 mM Tris-HCl (pH 8.0), 150 mM NaCl, 1.0% Nonidet P-40, 0.7 μg/ml pepstatin, 1.0 μg/ml leupeptin, 1.0 μg/ml aprotinin, and 0.5 mg/ml Pefabloc SC (Roche). Lysed cells were held on ice for 15 to 30 min and then clarified by centrifugation. For immunoprecipitations, lysates were cleared with preimmune antiserum and were then incubated for 1 h at 4°C with anti-nsp3 antibody D3 (30) or VU164 (31), generously provided by Susan Baker (Loyola University Chicago, Maywood, IL) or by Mark Denison (Vanderbilt University, Nashville, TN), respectively. Samples were incubated for an additional 1 h at 4°C with a 75% slurry of nProtein A Sepharose (GE Healthcare) in lysis buffer. Sepharose beads were collected by centrifugation, washed three times with lysis buffer, and used directly for SDS-PAGE sample preparation. Lysate samples or immunoprecipitates were separated by SDS-PAGE through 7.5% or 4-to-15% gradient polyacrylamide gels, with prestained protein markers (NEB) in flanking lanes, and were transferred to polyvinylidene difluoride membranes. Blots were probed with one of the following: anti-HA monoclonal antibody (MAb 12CA5; Roche), anti-nsp1 antibody VU221 (32), or anti-nsp8 antibody VU124 (33); the last two antibodies were also gifts from Mark Denison. Bound primary antibodies were visualized using a chemiluminescent detection system (ECL or West Dura; Pierce).

Construction of a mutant containing nsp3 as the first subunit of the replicase-transcriptase complex.

To begin to dissect the amino-terminal domain structure of nsp3, it was first necessary to disentangle nsp3 from its obligation to process upstream nsp subunits. The PLP1 module of MHV nsp3 carries out the nsp1-nsp2 and nsp2-nsp3 cleavage events of replicase-transcriptase polyprotein maturation (Fig. 1). Our first strategy to obviate these cleavages was to create a mutant, designated nsp3-start (N3S), in which polyprotein translation initiated with nsp3, while most of the preceding coding region was deleted. The major constraint on the construction of the N3S mutant was the need to preserve genomic elements that participate in viral RNA synthesis. The 5′ ends of coronavirus genomes, including the 5′ UTR and part of the nsp1 coding region, contain multiple cis-acting RNA structures that are critical for replication and transcription (34). In MHV and other betacoronaviruses, the genus in which these structures have been most extensively studied, the genomic 5′ end contains a series of at least eight stem-loops (SL) (35–39), as well as a four-way junction formed by a long-range interaction between the regions upstream of SL 5 (SL5) and downstream of SL7 (40) (Fig. 2A). Mutations or deletions in many of these elements have been found to debilitate or kill the virus or to abrogate defective interfering (DI) RNA replication.

We consequently designed the 5′ UTR of the N3S mutant to contain a modified version of the 5′-most 594 nucleotides of the MHV genome, thereby retaining some margin beyond the 467-nt minimal segment previously shown to be sufficient to support DI RNA replication (41, 42). The AUG codon that occurs at nucleotides 592 to 594 was fused to the beginning of the nsp3 coding region (nt 2707 of the wild-type sequence) to serve as the start codon for the replicase-transcriptase polyprotein (Fig. 2B); at the −4 position relative to the AUG codon, nt 588 was mutated to optimize the context for translation initiation (43). To ensure utilization of the newly created start codon, the original start codon for nsp1 and the other four in-frame upstream AUG codons were each knocked out by mutations at 2 or 3 positions. Additionally, six out-of-frame upstream AUG codons were each mutated at 1 or 2 positions. Further point mutations were made to restore any base pairs in RNA secondary structures that would be disrupted by the start codon knockouts. The RNA secondary structure shown in Fig. 2 is the lowest free-energy mfold structure (44) for wild-type MHV nucleotides 79 to 594 (with no imposed constraints to generate the long-range interaction). The mfold-predicted structure for the N3S mutant was verified to be identical to that of the wild type. The only upstream start codon that was not eliminated was that for the small upstream open reading frame (uORF) in SL4 (Fig. 2B). Similarly situated uORFs are found in the majority of coronavirus genomes, and in MHV there exists selective pressure for the uORF to be maintained (45).

Following multiple trials, a single isolate of the N3S mutant was obtained. Initially, the sequence of the 5′ genomic end of this recombinant was determined, confirming that it contained exactly the engineered 30 point mutations and 2,112-nt deletion (Fig. 2B) and, in particular, that the 5′-most AUG codon occurred at nt 592 to 594 (Fig. 3A). To verify the functionality of this relocated start codon, we translated synthetic mRNAs that were identical to the 5′ ends of the wild-type or N3S mutant genomes. These mRNAs were generated by runoff in vitro transcription of truncated pA-WT or pA-N3S, the first of the seven plasmids used for assembly of full-length wild-type or N3S mutant cDNAs, respectively (Fig. 3B). Additionally, we produced two N3S-related mRNAs from plasmids pA-N3SΔ1 and pA-N3SΔ2, which contained deletions in the 5′ UTR of pA-N3S. The mRNAs were translated in a reticulocyte lysate system, and [³⁵S]methionine-labeled protein products were analyzed by SDS-PAGE. For pA-WT mRNA, a product corresponding to the entire nsp1 fused to a small fragment of nsp2 had a size consistent with its expected molecular mass of 40.2 kDa (Fig. 3C). Translation of pA-N3S mRNA yielded a partial nsp3 fragment with an apparent molecular mass of 46 kDa, somewhat larger than its calculated size of 35.9 kDa. However, since a product of identical mobility was obtained with pA-N3SΔ1 and pA-N3SΔ2 mRNAs, the first of which has a 5′ UTR of only 25 nt, this showed that translation of pA-N3S mRNA had started at the expected location. As anticipated for cap-dependent protein synthesis, the efficiency of translation of the N3S-related mRNAs increased as the size of the 5′ UTR decreased. The aberrantly slow migration of the partial nsp3 product was likely due to the highly acidic (Ac) region near the amino terminus of nsp3 (Fig. 1); the same mobility was seen for an almost identical nsp3 fragment that was initiated by a completely different mechanism (see below). A faint, slower-migrating band was observed for translated pA-N3S mRNA (Fig. 3C, open circle), but this same band was present for the 5′-UTR-deleted N3S mRNAs and it therefore could not have resulted from upstream initiation at a noncanonical start codon.

Together, these results demonstrated that the complete elimination of nsp1 expression coupled with deletion of the entirety of nsp2 does not abolish the viability of MHV in tissue culture. This represents the largest truncation of the replicase-transcriptase that has yet been made, although it must be noted that successful individual deletion of nsp2 in both MHV and SARS-CoV has already been well established (31). Combined with our prior work (17), this means that Ubl1, the amino terminus of nsp3, is the first functional domain of the replicase-transcriptase with respect to viral RNA synthesis. Nevertheless, because the N3S mutant had been very difficult to isolate, this raised the possibility that it needed to acquire one or more adaptive mutations in order to be fully viable. The determination of the complete genomic sequence of this mutant revealed that, indeed, it harbored two mutations in the nsp3 coding region. The first of these, E851G, was at the amino-terminal end of PLP2, near the Ubl2 domain. The second mutation, D1791V, fell in the carboxy-terminal Y domain. There were only three additional mutations in the N3S genome outside nsp3. One of these, T220I in nsp15, was at the boundary of a nonessential surface loop on this subunit, and we thus think it is not functionally relevant. Likewise, two mutations in the spike (S) protein, Q99H in the ectodomain and C1280S in the transmembrane domain, could conceivably affect growth but are highly unlikely to signify an interaction between S and nsp3.

The phenotype of the N3S mutant also indicated that it was conditionally impaired. Plaques of N3S were moderately smaller than wild-type plaques at 33°C and 37°C. However, at 39°C, the mutant formed plaques that were tiny compared to those of the wild type (Fig. 4A). To explore the basis for this deficiency, 12 cultures begun from individual plaques of N3S were passaged serially 10 to 13 times at 39°C in L2 cells. Titers of the viral supernatants were then determined at 39°C, and one revertant plaque originating from each culture was purified for analysis. Each chosen revertant formed plaques that were substantially larger than those of the N3S mutant (two examples are shown in Fig. 4A), but none reached the size of wild-type MHV plaques at 39°C. For all 12 independent revertants, we sequenced the 5′ end of the genome, ranging from the end of the leader through the start of the nsp4 gene, as well as the entire N gene and the 3′ UTR. Only one revertant (rev12) had a nucleotide change (U476A) in the 5′ UTR, which occurred in a bulge base of SL A (SLA) (Fig. 2B). Otherwise, no alterations were found in the 5′ or 3′ UTRs of any revertant. This suggested that the changes engineered in the original N3S mutant did not adversely affect the functioning of cis-acting RNA elements at either end of the genome.

In contrast, all revertants exhibited from one to three mutations in the nsp3 coding region, and these were gathered in three main loci (Fig. 4B). Multiple reverting mutations mapped to the amino-terminal Ubl1 and Ac domains. Three of these—A38V, A60T, and D98Y—were positioned within or close to the α2 helix of Ubl1 previously shown to interact with N protein (16). Moreover, A38V was identical to one of the reverting mutations originally seen in the N protein chimeric mutant that led us to discover the Ubl1-N interaction (15). Conversely, though, none of the N3S revertants contained changes in the serine- and arginine-rich region of N protein. An nsp3 mutation found in one revertant, L480I, mapped to the border between the PLP1 and ADRP domains but is not likely to be significant, because most other strains of MHV also have an isoleucine at this position. All other reverting mutations fell in regions already brought to our attention by the two nsp3 mutations in the original N3S mutant (Fig. 4B). One of these, I813T in the Ubl2 domain, was near the E851G mutation of N3S in PLP2. The remainder of the reverting mutations were clustered in the carboxy-terminal Y domain or else in the region adjacent to the first transmembrane domain, which is likely proximal to the Y domain. Notably, most of the revertants contained a further alteration of D1791. This residue, which was changed to valine in the original N3S mutant, was again changed to either alanine or glycine in 8 of the 12 revertants. Collectively, the N3S reverting mutations appear to demarcate the most critical domains of nsp3. We speculate that they point to regions of nsp3 that interact with other proteins or protein domains important for RNA synthesis. However, it is not currently clear why the unique architecture of the N3S mutant genome necessitates its acquisition of these compensatory changes in order to improve nsp3 function.

Initiation of nsp3 by an alternative mechanism.

Owing to our uncertainty about the underlying cause of the mutations in N3S and its revertants, this mutant did not provide an optimal starting point for further dissection of nsp3. We therefore turned to an alternative strategy to circumvent the necessity for the nsp1-nsp2 and nsp2-nsp3 cleavage events, making use of the 2A peptide autoprocessing element from the picornavirus foot-and-mouth disease virus (FMDV). Through an undefined mechanism, the FMDV 2A sequence effectively dictates termination followed by reinitiation of protein synthesis at two adjacent codons in a single ORF (46). Exploiting this device, we constructed a mutant, designated stop-start-NSP3 (SSN3), in which synthesis of the replicase-transcriptase polyprotein began with an amino-terminal segment of nsp1 that was separated from nsp3 by a 2A element (Fig. 5A). The fragment of nsp1 (amino acids 1 to 127) was chosen to preserve the same cis-acting genomic RNA structures as had been retained in the N3S mutant. Autoprocessing of the 2A element would generate a single extra residue, proline, at the amino terminus of nsp3. This was expected to be tolerated, because the first 18 residues of MHV Ubl1 are unstructured and do not contribute to the interaction with N protein (16).

We first implemented this design with the minimal 20-amino-acid FMDV 2A peptide, but no viable virus was obtained. Similarly, inclusion of a homolog of the 2A peptide, from Thosea asigna picornavirus, appeared to be only partially effective, since recovered virus was very impaired. Amino-terminal extensions of the 2A element have been shown to greatly increase stop-start efficiency, consistent with the notion that the stretch of polypeptide occupying the ribosomal exit tunnel causes or contributes to this anomalous translational mechanism (47, 48). Accordingly, the final version of the SSN3 construct incorporated a 20-amino-acid FMDV 2A peptide plus 38 upstream residues from its native polyprotein sequence, preceded by a His₇ epitope tag (Fig. 5A). Three independent isolates of the SSN3 mutant were obtained, all having the same phenotype, and one was chosen for further analysis. The sequence of this virus was determined from the 5′ end of the genome through the PLP1 domain and was confirmed to be exactly as expected, with no additional mutations. Plaques formed by the SSN3 mutant were only slightly smaller than wild-type plaques at 37°C and 39°C (Fig. 5). However, at 33°C, we consistently noted heterogeneity in plaque size, with the sporadic appearance of much smaller plaques. The basis for this growth characteristic at 33°C is unknown, but we have observed it for all constructed mutants that contain large deletions in both nsp1 and nsp2, irrespective of the presence of either the 2A element or the His₇ epitope tag (data not shown).

Contrary to the direct fashion in which the SSN3 and subsequent derivative mutants were acquired, we were unable to obtain a related mutant, SSN3x (Fig. 5A), that was modified by two changes in the FMDV 2A sequence known to abolish autoprocessing (28). In four separate trials, each with multiple independent samples and robust positive controls, transfected SSN3x full-length RNA produced no signs of infection, indicating that the mutated 2A sequence was lethal. Since the only difference between the SSN3 and SSN3x constructs was the absence of a functional stop-start element in the latter, this strongly suggests that a permanent blockage of the amino terminus of nsp3 results in a nonfunctional replicase-transcriptase.

To seek further evidence for the relative efficacies of the SSN3 and SSN3x stop-start elements, we in vitro-translated synthetic mRNAs that were identical to the 5′ ends of these mutant genomes. Translation of pA-SSN3 mRNA gave rise to two discrete products, the smaller of which was exactly the size expected (21.5 kDa) for the nsp1 fragment fused to the FMDV 2A element (Fig. 5C). The larger pA-SSN3 product, corresponding to the processed nsp3 fragment, migrated more slowly than expected for a 39.5-kDa protein. However, its mobility was identical to that of the product obtained by translation of pA-N3SΔ1 mRNA, which, because of its minimal 5′ UTR, had served to produce a size standard for this nsp3 fragment (see Fig. 3C). (The nsp3 fragments encoded by pA-SSN3 and pA-N3SΔ1 differ only at their amino-terminal residues, proline and methionine, respectively.) In contrast to pA-SSN3, translation of pA-SSN3x mRNA yielded just a single, larger protein with an apparent molecular mass of 65 kDa, bigger than the expected 57.4 kDa by the same differential observed for the other nsp3 products. The total absence of this unprocessed product upon translation of pA-SSN3 mRNA indicated that there was complete efficiency of termination by the incorporated 2A element. Moreover, the relative intensities of the bands for the nsp3 and nsp1-2A fragments (allowing for their different methionine contents) showed that reinitiation was also highly or entirely efficient. Conversely, the lack of production of either the nsp3 or nsp1-2A fragment upon translation of pA-SSN3x mRNA demonstrated that the mutations in the SSN3x construct effectively eliminated autoprocessing by its 2A element. Taken together, these results allowed us to conclude that the SSN3 construct enabled normal initiation of nsp1 with equimolar production of mature nsp3, thereby providing a suitable platform for further dissection of nsp3.

The PLP1 and ADRP domains of nsp3 are entirely dispensable.

Having eliminated the need for the two processing events carried out by PLP1, we were thus able to explore whether this module is required for other roles in the replicase-transcriptase complex. Toward this end, the SSN3 mutant was rebuilt with FLAG and HA epitopes inserted in tandem immediately downstream of the Ac region, yielding a construct designated SSN3^HA (Fig. 6A). A PLP1 catalytic knockout, SSN3^HAmutP, was then created by mutating an active site residue that has been shown to be essential for MHV PLP1 activity in vitro (18) and in vivo (19). To minimize the probability of spontaneous reversion of this mutation, C289A, we changed all three bases of the wild-type codon (UGU to GCC). Multiple independent isolates of both the SSN3^HA and SSN3^HAmutP viruses were readily obtained. As no phenotypic differences were noted among them, one isolate of each was chosen for further analysis. Sequence determination from the 5′ end of the genome through most of nsp3 confirmed the presence in each of exactly the engineered composition with no extraneous mutations. Plaques formed by the SSN3^HA and SSN3^HAmutP mutants were identical to each other and to those of the original SSN3 mutant at 33°C, 37°C, and 39°C (examples at 37°C are shown in Fig. 6B). This showed that, as anticipated, the epitope tags were fully tolerated in what is thought to be an unstructured linker between the Ac region and PLP1 (12). Likewise, we had expected the knockout of PLP1 catalytic activity to have little or no effect in the SSN3 background, based on the prior demonstration of the viability of a PLP knockout mutant and its partial rescue by simultaneous small deletions in the nsp1-nsp2 and nsp2-nsp3 cleavage sites (19).

We next designed deletions of PLP1 in the SSN3^HA background (Fig. 6A). Because no structural information is yet available for a betacoronavirus PLP1, other measures were used to infer the limits of this domain. In the first mutant, SSN3^HAΔP1, the extent of the deletion (amino acids 234 to 484) was based on the smallest expressed segments of MHV nsp3 that exhibited protease activity in vitro (49, 50), as well as on the boundaries of the upstream Ac (15) and downstream ADRP (23, 51) domains. In the second mutant, SSN3^HAΔP2 (deletion of amino acids 274 to 460), the boundaries of the first deletion were moved inward, to the edges of the region of PLP1 that is most conserved among the lineage A betacoronaviruses. In addition, the more conservative downstream boundary preserved residue M475, a mutation of which has very far-reaching deleterious effects in a particular temperature-sensitive mutant, Brts31 (52). In a final construct, the SSN3^HAΔP3 mutant (deletion of amino acids 234 to 651), the downstream deletion boundary was extended in the opposite direction, to encompass both PLP1 and ADRP, addressing the possibility that the original deletion juxtaposed the essential Ubl1 domain too close to ADRP. In multiple independent trials, we were unable to isolate SSN3^HAΔP1 or SSN3^HAΔP2 virus. Remarkably, only the mutant containing the largest deletion was viable. Two independent isolates of the SSN3^HAΔP3 virus were recovered. For one of these, we determined the sequence of the entire genome, which revealed exactly the constructed mutations and rearrangements, with no additional mutations. The SSN3^HAΔP3 mutant formed plaques only slightly smaller than those of the corresponding SSN3^HA parent at 33°C, 37°C, and 39°C (plaques at 37°C are shown in Fig. 6B). This demonstrated that MHV can retain a surprising degree of hardiness despite the loss of half of nsp1, all of nsp2, and one-fifth of nsp3; the sum of the deletions in the SSN3^HAΔP3 mutant totals 3,120 nucleotides, nearly one-tenth of the viral genome.

To corroborate the genetic compositions of the SSN3^HA virus and its derivatives, we examined the expression of different subunits of the replicase-transcriptase. Infected cell lysates were immunoprecipitated with an antibody to the amino-terminal 347 amino acids of nsp3 (antibody VU164) (31) and then probed by Western blotting with an antibody to the HA epitope tag that had been inserted at the end of the Ac domain (Fig. 6A). (The FLAG epitope tag planted at the same internal position, adjacent to the HA tag, did not work in Western blots.) For cells infected with the SSN3^HA and SSN3^HAmutP viruses, a protein with an apparent molecular mass of 260 kDa was detected (Fig. 7A). nsp3 has a calculated molecular mass of 222 kDa (or 224 kDa with the epitope tag insertion) and has been described as 210 kDa (30, 31). However, we observed the same band for wild-type, SSN3, SSN3^HA, and SSN3^HAmutP viruses in direct Western blots with VU164 antiserum (data not shown). We therefore conclude that the apparent 260-kDa band is nsp3, not a precursor, and that its lower mobility is probably attributable to differences between the molecular mass standards used by us and by others. Consistent with this, a single band of 235 kDa was observed for the SSN3^HAΔP3 mutant, again larger than the calculated molecular mass of 178 kDa (Fig. 7A). This finding confirms that mature nsp3 of SSN3^HAΔP3 was stably produced and that it harbored a large deletion. All of the same protein species were observed when immunoprecipitations were performed with a different anti-nsp3 antiserum (antibody D3) (30), which had been raised against the amino-terminal 205 amino acids of nsp3 (Fig. 7B). Notably, nsp3 of the SSN3^HAΔP3 virus was only weakly immunoprecipitated by D3 despite containing the complete region of nsp3 recognized by this antibody. This may indicate that the deleted nsp3 of this mutant is conformationally different from wild-type nsp3.

Western blotting with anti-nsp1 antibody revealed a 26-kDa protein for wild-type-infected cell lysates (Fig. 7C), consistent with the expected molecular mass of 27.4 kDa for nsp1. The same antiserum also detected the 21.5-kDa nsp1-2A fragment of the SSN3 mutant and all of its derivatives, even though this fragment contained only half of the nsp1 molecule. (The His₇ epitope tag that had been engineered upstream of the FMDV 2A element [Fig. 5] did not work in Western blots.) This result demonstrated that there was efficient in vivo termination by the 2A element, as we had observed in vitro (Fig. 5C). Additionally, Western blots of infected cell lysates probed with anti-nsp8 antibody detected the same 21.6-kDa species in wild-type- and mutant-infected cell lysates. This showed that, at the peak of infection, nearly equivalent amounts of this downstream replicase subunit were produced, suggesting that reinitiation by the FMDV 2A element was also highly efficient in vivo.

To more completely assess the fitness of the SSN3^HA and SSN3^HAΔP3 mutants, we evaluated their growth and RNA synthesis capabilities relative to those of the wild type. In infections begun at a multiplicity of 1.0 PFU per cell (limited by the titers that could be achieved for stocks of SSN3^HAΔP3), wild-type virus reached peak titers of 7.8 × 10⁷ PFU/ml between 8 and 12 h postinfection (Fig. 8A). In contrast, the SSN3^HA and SSN3^HAΔP3 mutants reached maximal titers that were 1.6 and 2.0 log₁₀ lower than those of the wild type with somewhat delayed kinetics, peaking between 12 and 16 h postinfection. Data from a second, independent growth experiment were nearly identical to those shown in Fig. 8A. We also noticed that 17Cl1 cell monolayers infected with wild-type MHV exhibited extensive syncytia and cytopathic effect and were almost completely detached by 24 h postinfection. However, 17Cl1 monolayers infected with either of the two mutants underwent a period of partial detachment from 16 to 24 h postinfection but had recovered to roughly 80% confluence by 48 h postinfection. This recovery was not observed with L2 cell monolayers. The differential growth kinetics (Fig. 8A) make clear that the major detriment caused by the engineering of the SSN3-derived mutants was due to the loss of nsp1 and nsp2. The gap between the growth curves for wild-type virus and both the SSN3^HA and SSN3^HAΔP3 mutants was roughly equal to the sum of the reductions seen previously for an individual nsp1 carboxy-terminal deletion mutant (53) and an individual nsp2 deletion mutant (31). Comparatively, the removal of the PLP1 and ADRP domains in the SSN3^HAΔP3 mutant had only a minor effect. Northern blotting of intracellular RNA harvested at the peak of infection revealed a corresponding quantitative drop in RNA synthesis by the SSN3^HA and SSN3^HAΔP3 mutants relative to that of the wild type (Fig. 8B). However, both mutants produced the same subgenomic RNA species in the same relative proportions as wild-type MHV. This showed that deletion of the PLP1 and ADRP domains of nsp3 did not qualitatively alter the ability of MHV to execute the characteristic pattern of coronavirus RNA replication and transcription.