Analysis of Constructed E Gene Mutants of Mouse Hepatitis Virus Confirms a Pivotal Role for E Protein in Coronavirus Assembly

Virus and cells.

Wild-type, mutant, and revertant virus stocks of all MHV strains were propagated in mouse 17 clone 1 (17Cl1) cells. The sources of strains MHV-1 and MHV-3 were described previously (30). It should be noted that our laboratory strain of MHV-3 differs from the Bicêtre strain of MHV-3 (9). MHV-2 and MHV-DVIM were generously provided by Ehud Lavi (University of Pennsylvania School of Medicine) and Kathryn Holmes (University of Colorado Health Sciences Center), respectively. Plaque titer determinations and plaque purifications were carried out with mouse L2 cells, which were also maintained in spinner culture for RNA transfection via electroporation, as described previously (29).

Cloning and sequencing of the E gene of various MHV strains.

A region encompassing the E genes of MHV-1, MHV-2, MHV-3, and MHV-DVIM RNA isolated either from infected cells or from purified virus was amplified by reverse transcription followed by PCR (RT-PCR), with primers corresponding to nucleotides (nt) 147 to 164 of gene 5a and nt 244 to 261 of the M gene. Although these primers were based on the known sequence of MHV-A59, a 710-bp PCR product was successfully obtained for each of the four other strains and was cloned into a pCR TA cloning vector (Invitrogen). A consensus nucleotide sequence was determined from three or four clones for each MHV strain.

Plasmid constructs.

T7 transcription vectors generated for the transfer of mutations into the E gene of MHV were derived from pCFS8 (14), which encodes a pseudo-defective interfering (DI) RNA that includes the complete S gene and all of the MHV-A59 genome downstream of S through the poly(A) tail. As detailed previously (14), in this region pCFS8 contains two differences from the authentic MHV sequence: (i) a coding-silent change in the S gene that eliminates a HindIII site, and (ii) a phenotypically silent 19-nt tag at the start of gene 4 that contains an XbaI site.

Clustered charged-to-alanine mutations 1 to 4 were initially created in pCFS1 (14), a precursor of pCFS8 that contains the segment extending from the MluI site at nt 3262 of the S gene through the KpnI site at nt 394 of the M gene (see Fig. 3). Mutations 1 and 2 (see Fig. 2) were each generated by a single step of PCR mutagenesis followed by replacement of the BstEII-EcoRV fragment of pCFS1 with the BstEII blunt-end fragment of the mutant PCR product to yield pCFS9 and pCFS10, respectively. Mutations 3 and 4 were generated by splicing overlap extension PCR mutagenesis (18), and the BstEII-EcoRV fragment of each mutant PCR product was used to replace the corresponding segment of pCFS1 to yield pCFS11 and pCFS15, respectively. Mutations 1 to 4 were then transferred into the pseudo-DI vector by incorporation of the MluI-EagI fragment of pCFS9 or the XbaI-EcoRV fragment of pCFS10, pCFS11, or pCFS15 in place of the corresponding fragment of pCFS8 to produce pFF45-pFF48, respectively.

Second-site reverting mutations, in conjunction with the original mutations of mutant 2 (Alb154), were also introduced into MHV via transcription vectors derived from pCFS8. For the construction of these, total RNA isolated from cells infected with Alb154RevA, Alb154RevB, Alb154RevD, or Alb154RevI was reverse transcribed under standard conditions (35) with a random primer, p(dN)₆ (Boehringer Mannheim). The 170-bp region from gene 5a through the M gene (see above) of each cDNA was amplified by PCR, restricted with BstEII and EcoRV, and, together with the XbaI-BstEII fragment of pCFS8, incorporated in place of the XbaI-EcoRV fragment of pCFS8 in a three-way ligation to produce plasmids pFF61 to pFF64.

Standard techniques were used for all recombinant DNA manipulations (35). The composition of all constructed plasmids was verified by restriction analysis; all PCR-generated regions and newly created junctions of each plasmid were verified by DNA sequencing by the method of Sanger et al. (36) with modified T7 DNA polymerase (Sequenase; U.S. Biochemical Co.) or by automated sequencing with an Applied Biosystems 373A or 377 DNA sequencer.

Targeted recombination and identification of recombinant mutant viruses.

E gene mutations were transduced into the MHV genome by the targeted recombination method previously used to incorporate mutations into the S (14), M (10), and N (13, 32, 33) genes, gene 4 (14), and the 3′ untranslated region (19, 22). Mutant viruses were obtained as progeny recombinants arising from transfection of synthetic pseudo-DI RNAs from HindIII-truncated pFF45-48 and pFF61-64 into cells infected with the temperature-sensitive and thermolabile N gene mutant, Alb4. Details of RNA synthesis and transfection were as described previously (14, 29).

Of the four clustered charged-to-alanine mutants, only mutant 2 was selected as a recombinant capable of forming large (i.e., wild-type-size) plaques at 39°C, the nonpermissive temperature for Alb4. Three candidate plaques for this mutant were analyzed by RT-PCR with a primer pair flanking the locus of the Alb4 deletion and with a primer pair flanking the gene 4 tag (see Fig. 3). The repair of the Alb4 deletion was ascertained by the size of the PCR product from the former (22). The PCR product from the latter was digested with XbaI to determine the presence of the gene 4 tag. Following sequence verification, a recombinant from this set was designated Alb153.

Since targeted recombination with donor RNA containing the other three mutations did not yield large-plaque mutants at 39°C, a screening approach was used for their identification. Following heat treatment to counterselect Alb4 parental virus (22), surviving plaques obtained at 33°C were combined in multiple pools of 10 plaques and RNA was isolated from a cell monolayer infected with each pool and was analyzed by RT-PCR. Recombinants were detected with primer pairs in which one primer originated within the heterologous 19-nt gene 4 tag or in which one primer originated within the region that is deleted in the Alb4 N gene. Plaques from pools giving a positive PCR signal for one of these markers were subsequently analyzed by RT-PCR.

Individual candidates for mutant 3 (20 plaques) and mutant 4 (40 plaques) were then screened for the gene 4 tag or for repair of the Alb4 deletion with primer pairs flanking or originating within these markers, as above. Positive candidates were then screened by RT-PCR amplification of the E gene and tested for the presence of an FspI site created by the mutations in mutant 3 or mutant 4 (see Fig. 2). One mutant of each type was found in this search, and following sequence verification, they were designated Alb154 (mutant 3) and Alb183 (mutant 4). A similar screening procedure was used to search for mutant 1. In this case, the RT-PCR product derived from the E gene was tested for the loss of the EcoRV site that should be disrupted by one of the mutations in this mutant (see Fig. 2). In multiple trials, no recombinant meeting this criterion was found.

Final confirmation of the composition of each of the three successful recombinants was obtained by sequencing purified genomic RNA to establish the presence of the E gene mutations, the gene 4 tag, and the region that is absent in the Alb4 N gene deletion. MHV recombinants were purified by polyethylene glycol precipitation followed by two cycles of equilibrium centrifugation on preformed potassium tartrate-glycerol gradients, and viral genomic RNA was isolated exactly as described previously (13, 22). For the isolation of total cytoplasmic RNA from virus-infected 17Cl1 cell monolayers, a Nonidet P-40 gentle lysis procedure was used (21). Direct RNA sequencing was carried out by a modification of a dideoxy termination protocol (12, 32).

Revertant analysis.

Independent spontaneous revertants of mutant 3 (Alb154) and mutant 4 (Alb183) were isolated starting from individual plaques obtained at 39°C from passage 2 stocks. Each plaque was serially passaged 10 times (mutant 3) or 11 times (mutant 4) at 37°C at an estimated multiplicity of 0.2 PFU/cell. Following the final passage, revertants were isolated as viruses forming large (wild-type-size) or intermediate-size (between mutant size and wild-type size) plaques at 39°C. Mutations responsible for reversion were determined by sequencing of the E gene and other genes in RNA from cells infected with purified plaques or in purified virion RNA.

Electron microscopy.

Samples of mutant 4 (Alb183) and the isogenic wild-type strain Alb129, freshly passaged at 33°C, were harvested, clarified of cell debris by centrifugation for 10 min at 1,200 × g, and used directly for electron microscopy without further concentration or purification (see Fig. 7 and 8). In a separate experiment (see Table 2), mutant 4 and wild type were freshly passaged at 33 and at 39°C and clarified and 8-ml samples were then concentrated by centrifugation for 2 h at 151,000 × g through 1.5 ml of 50 mM Tris-maleate (pH 6.5)–1 mM EDTA–10% glycerol onto a cushion of 100% glycerol in a Beckman SW41 rotor at 4°C. Viral pellets were resuspended in 1 ml of 50 mM Tris-maleate (pH 6.5)–1 mM EDTA. Drops (40 μl) of the different viral suspensions were placed on dental wax, and previously glow-discharged, Formvar-coated copper grids were placed on the drops for 2 to 4 min. Excess liquid was removed by wicking with filter paper, and the grids were immediately washed by floating on one or two drops of 10 mM ammonium acetate. After the final wash, the grids were briefly floated on 2% sodium phosphotungstate (pH 7.0), the excess stain liquid was removed by wicking, and the samples were viewed in a Zeiss (LEO) 910 transmission electron microscope operating at 80 keV. Micrographs were recorded at a magnification of ×25,000, and the negatives were enlarged photographically.

Nucleotide sequence accession numbers.

The coding regions of the E genes of MHV-1, MHV-2, MHV-3, and MHV-DVIM have been assigned GenBank accession no. AF051146, AF051147, AF051148, and AF051149, respectively.

Comparison of MHV E protein sequences.

As a preliminary step in establishing the importance of different regions and residues of the E protein, we determined the sequences of the E genes of MHV-1, MHV-2, MHV-3, and MHV-DVIM. An alignment of the resulting deduced amino acid sequences with that of our reference strain, MHV-A59, and with those previously determined for MHV-S and MHV-JHM, was performed to obtain a sampling of the range of sequence variation allowed for the E protein (Fig. 1). This comparison presumes, and all existing interstrain recombinational evidence supports the idea, that these E proteins are functionally interchangeable.

As shown in Fig. 1, the MHV E proteins are all quite small, comprising 83 to 88 residues (9.6 to 10.2 kDa), and exhibit a high degree of sequence identity (89 to 98%). Almost one-third of the protein is made up of a relatively long hydrophobic region, a putative membrane anchor spanning residues V10 through I37. Although the topology of the E protein has not been unequivocally established, evidence from a study of the E protein of transmissible gastroenteritis virus (16) suggests that the molecule is in a C-exo N-endo configuration. The membrane anchor is highly conserved among the seven strains of MHV and is flanked by two invariant charged residues (D8 and K38). Four absolutely conserved cysteine residues (C23, C40, C44, and C47), which are potential sites for palmitoylation (47) or interchain disulfides, are found within the membrane anchor or adjacent to it on the carboxy side. The only hydrophilic portion of the molecule occurs near the carboxy terminus. Although the charged residues here are highly conserved, this segment represents the most divergent part of the E protein, most notably at the extreme carboxy terminus, which shows a 5-amino-acid extension (IIQTL) in many of the strains.

Generation of clustered charged-to-alanine mutants of the MHV E protein.

To address whether the E gene is an essential gene, we first attempted, by targeted RNA recombination, to create a null mutant of MHV-A59 in which the E gene was replaced by the gene for the green fluorescent protein. While we could successfully construct multiple recombinants in which gene 4 was replaced by the green fluorescent protein gene (14), we were not able to generate such a deletion in the E gene despite repeated efforts. Although this negative result could not be interpreted unequivocally, it implied that ablation of the E gene would be lethal to the virus.

We thus sought next to construct conditional lethal E gene mutants by using the strategy of clustered charged-to-alanine mutagenesis (2, 3, 44). Owing to the small size and limited hydrophilicity of the E protein, the algorithm of Wertman et al. (44) suggested only four possible sets of mutations, all of them falling in the carboxy-terminal one-third of the molecule (Fig. 2). The four clustered charged-to-alanine mutants were constructed in transcription vectors that also contained a phenotypically silent marker in gene 4 (14) and were then introduced into the MHV genome by targeted recombination. Synthetic mutant donor RNAs were independently transfected into cells infected with the N gene deletion mutant Alb4 (Fig. 3), and released progeny viruses were harvested and analyzed for the desired mutations. Since we were interested in generating temperature-sensitive mutants, plaques of progeny virus were first assayed under nonpermissive conditions (39°C), either directly or with prior heat treatment, to eliminate mutants capable of forming large (i.e., wild-type-size) plaques at this temperature. Only one of the four possible mutants, mutant 2 (designated Alb153), was found to fall into this category and its apparently wild-type phenotype was not characterized further. Although viruses forming large plaques at 39°C were identified among progeny from the other three donor RNAs, these proved to be either revertants of Alb4 or recombinants that had repaired the Alb4 deletion but did not contain the upstream E gene mutations.

Other clustered charged-to-alanine mutants had to be identified by molecular screening, because there was no other clear basis for differentiating them from the Alb4 parent, which is also temperature sensitive. Since we expected that the E mutants would be temperature sensitive but not thermolabile (or not as thermolabile as Alb4 [22]), progeny viruses from targeted recombination were heat treated (22) to enrich for potential recombinants. In retrospect, this may have hindered rather than helped subsequent screening. Pools of candidate recombinants, obtained as progeny plaques at the permissive temperature of 33°C, were assayed by RT-PCR, and individual plaques from pools that gave positive signals were then analyzed further, as described in Materials and Methods. This process yielded a recombinant of mutant 3 (Alb154) and a recombinant of mutant 4 (Alb183). A recombinant representing mutant 1 was never obtained among the numerous candidates screened, suggesting that this mutant, if not lethal, was more impaired than the Alb4 parent. Analysis of mutants 2, 3, and 4 by RT-PCR and by direct sequencing of purified genomic RNA revealed that each had the constructed set of four or five point mutations in the E gene, resulting in the conversion of two or three charged residues to alanines (Fig. 4). The sequence shown in Fig. 4 for Alb183 was obtained with considerable difficulty owing to the small amount of genomic RNA that could be purified from this virus, possibly because of virion fragility (see below). The mutations in this virus were verified in total cellular RNA isolated from Alb183-infected cells and also in revertants of Alb183. As expected, the Alb4 N gene deletion in each of the three recombinants had been repaired, and in addition, the gene 4 tag from the donor RNA was present (data not shown). This distribution of markers was consistent with each recombinant having been generated by a single crossover event upstream of gene 4 (Fig. 3).

At elevated temperatures, the mutant 3 and mutant 4 recombinants, Alb154 and Alb183, each produced significantly smaller plaques than those of wild-type MHV-A59 or Alb129, an isogenic recombinant that contains the gene 4 tag but no other mutations (14). As indicated in Table 1, this partially temperature-sensitive phenotype was evident at both 37 and 39°C. At either of these temperatures, each mutant produced plaques that were less than one-third the diameter of wild-type plaques, although they were not as small as plaques of the N gene deletion mutant Alb4.

Revertant analysis.

The sizes of plaques of mutants 3 and 4 differed sufficiently from those of plaques of the wild type to serve as the basis for the isolation of revertants. Following either 10 or 11 serial passages at the nonpermissive temperature of virus stocks originating from independent plaques of mutant 3 or mutant 4, spontaneous revertants were selected as viruses able to form plaques detectably larger than those of the mutant from which they were derived. In most cases, revertants formed wild-type-size plaques at 39°C; in some cases, the revertant plaque size was intermediate between those of the wild type and the mutant. It should be noted that despite the selective pressure applied over multiple passages, not every stock yielded a revertant.

For mutant 3, nine independent revertants were isolated. Sequencing of the E genes of these revealed that each contained a single second-site mutation in the E protein and that they could be divided into five groups (Fig. 5). Similarly, 14 independent revertants of mutant 4 were isolated and were divided into eight groups (Fig. 5). In most cases, each mutant 4 revertant also contained a single second-site mutation. The two exceptions to this observation, mutant 4 RevC and mutant 4 RevM, each contained an additional mutation that was most probably of no significance. For mutant 4 RevC, the S62I mutation that it had in common with mutant 4 RevG must have been sufficient for reversion; likewise, for mutant 4 RevM, the S62C mutation that it had in common with mutant 4 RevK and mutant 4 RevP must have been sufficient for reversion.

Almost all of the second-site reverting mutations of both mutant 3 and mutant 4 fell at one of three positions close to the original clustered charged-to-alanine mutations in the E gene, and almost all of these (S55A, S62I, S62C, and Y66C) occurred in both revertant sets (Fig. 5). Surprisingly, the two revertants that deviated from this pattern, mutant 3 RevI and mutant 4 RevA, each had a relatively conservative amino acid substitution within the putative membrane anchor (V31A and T27A, respectively). This may suggest either that compensatory structural changes within the membrane domain of E protein can be transmitted to the ectodomain or else that monomer-monomer interactions within the membrane can make up for a loss of interaction between ectodomain moieties.

To ascertain whether the observed second-site mutations were indeed responsible for reversion, we analyzed in more detail a subset of mutant 3 revertants representative of each position at which second-site mutations were found. Initially, we sequenced the entire M gene of each member of this group, mutant 3 RevA, mutant 3 RevB, mutant 3 RevD, and mutant 3 RevI, to check for the possible presence of mutations in this protein, which potentially interacts directly with the E protein (4, 42). No mutations were found in any of the M genes of these revertants. Next, we attempted to introduce each reverting mutation, in conjunction with its cognate set of clustered charged-to-alanine mutations, into the MHV genome. This was carried out by targeted recombination between the appropriate donor RNA construct and the Alb4 N gene deletion mutant. For mutant 3 RevA, mutant 3 RevB, and mutant 3 RevD, recombinants were obtained that contained both the mutant 3 clustered charged-to-alanine mutations and the corresponding second-site mutation. Since all of these produced large plaques at 39°C, this demonstrated that the E gene second-site mutation found in each of these revertants was sufficient to compensate for the original mutations of mutant 3. However, for reasons that we currently do not understand, repeated attempts to generate a recombinant corresponding to mutant 3 RevI were unsuccessful. This result, coupled with the unexpected membrane anchor location of the second-site mutation in mutant 3 RevI, prompted us to sequence the remaining structural genes (S and N) of this virus as well as the packaging signal domain within gene 1b (15). No mutations other than that in the E protein (V31A) were found. Thus, although we cannot rule out the possibility of a compensating mutation in a nonstructural gene, it is likely that the V31A mutation accounted for the reversion of mutant 3 RevI.

Taken together these analyses confirmed that for both mutant 3 and mutant 4 one or more of the clustered charged-to-alanine mutations constructed in the E gene was responsible for the observed plaque size phenotype. Moreover, the results demonstrated that a single second-site amino acid change within the E protein was able to reverse the phenotypic effect of the original mutation(s). No intergenic suppressors of the mutant 3 or mutant 4 mutations have yet been found.

Phenotype of E protein mutants.

Difficulties encountered in the preparation of purified virions of mutant 3 and mutant 4 led us to suspect that there might be defects in their structural integrity. To test the thermolability of these mutants, stocks of mutant 3, mutant 4, and their isogenic wild-type counterpart, Alb129, were grown at 33°C and also at 39°C. Released viruses were then heat treated for various intervals at 40°C and pH 6.5 (22), and survivors were measured by plaque titer determination at 37°C. As shown in Fig. 6A, virions of the two mutants were at least an order of magnitude more thermolabile than were wild-type virions when grown at 33°C. By 24 h of heat treatment, the infectious titers of mutant 3 and mutant 4 had dropped 15- and 78-fold, respectively, more than that of the wild type. In contrast, when grown at 39°C, both mutant 3 and mutant 4 were much less strongly affected by heat treatment and did not differ significantly from the wild type, which showed roughly the same heat sensitivity at 39°C as at 33°C (Fig. 6B). However, the infectious titers of stocks of the two mutants grown at 39°C were markedly lower than those of stocks grown at 33°C. One explanation consistent with these data is that for the two mutants grown at 33°C, most virions were assembled in a flawed manner that was sufficient for their stability at the lower, permissive temperature but led to a loss of viability upon heat treatment. Conversely, for mutant viruses grown at 39°C, only a relatively small fraction of virions assembled successfully, but these had passed some threshold of structural stability and were no more susceptible to heat treatment than were wild-type virions. Further work is required to more precisely elucidate the basis for the thermolability differences observed in Fig. 6. It should be noted that both the plaque size and the frequency of survivors of heat treatment among mutants 3 and 4 grown at 39°C ruled out the possibility that these survivors were revertants.

For mutant 4, the more defective of the two E protein mutants, we next sought further indications of structural impairment. When virions of mutant 4 were examined by electron microscopy, a remarkable difference from virions of the isogenic wild-type control, Alb129, was seen. Negatively stained preparations of wild type had an appearance typical for MHV and other coronaviruses (Fig. 7). With few exceptions, each virion was a roughly spherical particle 80 to 100 nm in diameter, with a more densely staining center and a complete halo of surface spikes projecting approximately 20 nm beyond the periphery of the particle. By contrast, although some virions of mutant 4 conformed to this description, most particles of this E protein mutant deviated dramatically from the typical coronavirus structure, as exemplified in Fig. 8. The majority of mutant 4 virions exhibited a more narrow, tubular appearance. These elongated virions were often pinched at multiple points, producing dumbbell-shaped forms. Even the more spherical particles of mutant 4 frequently contained two or three lobes. Between 80 and 90% of mutant 4 virions, whether grown at 33 or at 39°C, could be classified as having an abnormal morphology, whereas no more than 15% of wild-type virions fell into this category (Table 2). The estimated volumes of mutant 4 virions were, on average, similar to those of wild-type virions but varied much more from the mean value. As can be seen in Fig. 8, despite the abnormality of virion shape, the mutant showed no detectable aberration in its distribution of surface spikes. These observations provided striking evidence for the critical role of the coronavirus E protein in virion morphogenesis, clearly corroborating the results obtained by expression studies that established the requirements for virus-like particle formation (4, 42).