Mutations in the E1 Hydrophobic Domain of Rubella Virus Impair Virus Infectivity but Not Virus Assembly

Abstract

Rubella virus (RV) virions contain three structural proteins, a capsid protein that interacts with viral genomic RNA to form a nucleocapsid and two membrane glycoproteins, E2 and E1. We found that substitution of either an aspartic acid residue at Gly93 (G93D) or a glycine residue at Pro104 (P104G) in the internal hydrophobic domain of E1 affected virus infectivity but not virus assembly. Viruses carrying G93D and P104G mutations had impaired infectivity, reduced 1,000-fold and 10-fold, respectively. A revertant was isolated from the G93D mutant. Sequencing analysis showed that the substituted aspartic acid residue in G93D mutant had reverted to the original glycine residue, suggesting the involvement of Gly93 in membrane fusion during viral entry.

Rubella virus (RV), a small enveloped positive-strand RNA virus, is the sole member of the genus Rubivirus in the family Togaviridae (3). The RV virion contains a nucleocapsid consisting of a 40S genomic RNA molecule and a single species of capsid protein (33 kDa) (17). The nucleocapsid is enveloped within a host-derived lipid bilayer containing two viral glycoproteins, E1 (58 kDa) and E2 (42 to 46 kDa) (16). The structural protein genes are expressed from a 24S subgenomic RNA and are translated in the order NH₂-C-E2-E1-COOH (18). RV matures by budding at the plasma membrane of infected cells (1) and enters uninfected cells by a membrane fusion process in the endosome (6). Both processes are directed by E2-E1 heterodimers (4, 27). Formation of an E2-E1 heterodimer is required for transport of E1 out of the endoplasmic reticulum lumen to the Golgi and plasma membrane (4).

Semliki Forest virus (SFV), a well-characterized alphavirus, uses an acid-triggered membrane fusion to infect host cells. Its fusion with cellular membranes is mediated by the viral spike proteins, heterotrimers of two transmembrane subunits, E2 and E1, and a peripheral protein, E3 (8, 25). Mutagenesis of an SFV spike protein cDNA indicates that the internal hydrophobic domain of E1 is closely involved in membrane fusion (12). Like SFV, RV infects cells via a low-pH-dependent pathway (6, 9, 23). Although little is known about the fusion process of RV, available evidence suggests that RV E1 plays a dominant role in membrane fusion (6, 27).

In previous studies, we constructed and characterized mutations within the putative E1 fusion domain, 29 amino acid residues, extending from amino acids 81 to 109 (Fig. 1). Mutants generated in this domain were based on the rationale that substitution of Cys82 by serine (C82S) would affect E2-E1 interaction, substitution of a charged aspartic acid at Gly93 (G93D) would inhibit the ability of E1 to induce membrane fusion, and substitution of glycine for Pro104 (P104G) would disrupt the amphipathic and helix of fusion peptide. Using a cDNA clone coding for E2 and E1 structural proteins, we demonstrated that C82S mutation or deletion of this hydrophobic domain (dt) of E1 (Fig. 1) resulted in disruption of E1 conformation that ultimately impaired E1-E2 heterodimer formation and cell surface expression of both E1 and E2 (27). However, both G93D and P104G mutations were found to block neither E1-E2 heterodimer formation nor the transport of E1 and E2 to the cell surface. No syncytium formation was detected in cells expressing C82S, dt, and G93D mutations, whereas the wild-type (wt) virus and P104G mutant exhibited fusogenic activity, with 60% and 20 to 40%, respectively, of cells fused at pH 4.8 (27). These studies were performed on RV E2 or E1 structural protein cloned into a plasmid, and the effects of these mutations on virus assembly and infectivity could not be determined.

FIG. 1.

Schematic diagram of RV cDNA constructs and the mutations introduced in the hydrophobic region of RV E1. E1, E2, and C genes are indicated at the top. Signal peptides are indicated by open boxes, and the transmembrane regions are shown by solid boxes. The E1 hydrophobic domain is shown. Beneath the arrows are the single amino acid changes and the deletion mutation. 24S, the cDNA containing all three structural protein genes of RV inserted in pNUT vector (19) under the control of an inducible mouse metallothionein promoter (mMT-1); E2E1, the cDNA containing the E1 and E2 genes in pNUT vector. Restriction endonuclease sites: E, EcoRI; H, HindIII; B, BamHI; N, NcoI; S, SphI; and Bs, BstEII. nt, nucleotides.

In the present study, we incorporated G93D and P104G mutations into an infectious RV cDNA clone and examined their effects on virus infectivity. The effects of mutation on virus assembly were studied using a system in which RV spike proteins can be assembled into RV-like particles in the absence of virus replication (5, 20). We found that virus assembly was not affected by G93D or P104G mutation, but the mutants were 10-fold (P104G virus) and over 1,000-fold (G93D virus) less infective than the wt virus. Passage experiments were carried out to examine the reversion of mutant viruses harvested from BHK cells transfected with RNA transcripts from mutants. In revertants from G93D mutant virus, the substituted aspartic acid residue was found to have changed to the original glycine residue. No revertant was observed with P104G mutant virus.

Effects of mutations on virus assembly.

In SFV, mutations in the putative fusion peptide of E1 glycoprotein confer a strong and heat-reversible budding effect (2). The assembly of mutant spike proteins into mature virions is severely impaired, and a cleaved soluble fragment of E1 is released into the medium (2). In our previous studies, we showed that in the fusion defective G93D mutant, the E2-E1 heterodimers at the cell surface are unstable, resulting in the release of cleaved E2 and E1 into the medium (27). It is of interest to examine whether the assembly of virus particles is also affected by the G93D mutation. Since G93D and P104G mutations reside within the putative fusion peptide, we chose to study virus assembly in the absence of virus replication, using the capacity of RV structural proteins to form virus-like particles (VLPs) (20). In this system, coordinated expression of capsid, E2, and E1 proteins in stable transformed BHK cells (BHK-24S) results in their assembly into VLPs in the absence of genomic RNA.

To isolate a stable BHK-24S cell line expressing the G93D or P104G mutation, cDNA encoding the C protein was inserted into plasmid pNUT-E2E1 (G93D) or pNUT-E2E1 (P104G) (27) by replacing the BstEII fragment with the BstEII fragment from pNUT-24S (Fig. 1). BHKtk⁻ cells (24) were transfected with resultant plasmids by using Lipofectin (Gibco/BRL), and transformed cell lines were isolated as described previously (20). Cell lines were named BHK-24S (wt), BHK-24S (G93D), and BHK-24S (P104G). To determine the effects of G93D and P104G mutations on VLP formation, pulse-chase experiments were carried out. Transformed BHK cells were preincubated with growth medium containing 40 μM ZnSO₄ for 4 h to induce the expression of RV structural proteins from the metallothionein promoter (20). Cells were labeled with [³⁵S]methionine for 60 min and chased with 1 mM unlabeled methionine for 0, 2, 4, or 8 h. Labeled RV proteins from cellular lysates and culture media were immunoprecipitated with human anti-RV serum. Half of each sample was treated with endo-β-N-acetylglucosaminidase H (endo H) to monitor the processing of N-linked glycan (21). In wt and both mutants, intracellular RV specific proteins the sizes of E1 (58 kDa), C (33 kDa), and E2 (39 kDa) were detected, migrating above the capsid protein band; chase with unlabeled methionine resulted in conversion of E2 to a higher-molecular-mass species (42 to 45 kDa) of glycoprotein, while the sizes of E1 and C remained unchanged (Fig. 2A).

FIG. 2.

Assembly of RV structural proteins into VLPs in BHK cells. Transformed BHK cells were induced with ZnSO₄ (40 μM) for 4 h prior to labeling. Induced BHK cells were labeled with [³⁵S]methionine and chased with unlabeled methionine for 2, 4, and 8 h. Labeled RV proteins from cellular lysates (A) and culture media (B) were immunoprecipitated with human anti-RV serum, followed by SDS-PAGE. Lanes: B, BHK cells; Wt, BHK-24S (wt); G93D, BHK-24S (G93D); P104G, BHK-24S (P104G). Positions of E1, E2, C, and apparent molecular weight markers (in kilodaltons) are shown.

The release of RV structural proteins into the medium was found to increase proportionally with the duration of chase in wt and mutants (Fig. 2B). E2 was found to resist endo H digestion, while E1 was partially endo H resistant (data not shown), suggesting that both E2 and E1 had traversed through the Golgi complex and to the cell surface (27). To confirm that RV structural proteins in the medium were VLPs that assembled into subviral particles prior to their release from the cells, media from wt and mutants were subjected to ultracentrifugation (350,000 × g for 20 min) in the presence or absence of 1% nonionic detergent Nonidet P-40 (NP-40). Pellets were resuspended in phosphate-buffered saline and subjected to sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE). As expected, C, E2, and E1 were detected in the pellets in the absence of NP-40, but not in its presence (data not shown), indicating that the viral proteins were secreted as particles sedimenting in a gravitational field. These results indicate that G93D and P104G mutations did not affect the intracellular assembly of mutant spike proteins into VLPs.

Since E1 is the hemagglutinin (HA) protein of RV, we were interested to determine whether G93D and P104G mutations would affect the HA activity. Transformed BHK cells were grown in Dulbecco modified Eagle medium (DMEM) containing 3% fetal calf serum (FCS) that had been treated with kaolin to remove nonspecific inhibitors. After 18 h of induction with ZnSO₄, medium samples were collected and equal amounts of medium from wt, G93D, and P104G mutants were subjected to the conventional polyethylene glycol precipitation procedure for isolation of RV particles (11). The pelleted VLPs were suspended in phosphate-buffered saline. The HA assay was performed as described by Liebhaber (13), and the HA titer was expressed as the endpoint of serial dilutions of VLPs at which erythrocyte aggregation was observed. The VLPs from wt and both mutants all displayed HA activity of 32. Thus, G93D and P104G mutations did not interfere with the expression of HA activity of E1.

Effect of mutations on virus infectivity.

In our previous studies (27), substitution of Gly93 residue by an aspartic acid in the E1 hydrophobic region blocked syncytium formation without affecting the transport of spike proteins to the cell surface. We speculate that this mutation will be lethal in a virion, resulting in the assembly of a nonfusogenic and noninfectious virion. However, it has been shown in Moloney murine leukemia virus that fusion accompanying viral entry and fusion responsible for syncytium formation are independent processes in NIH3T3/DTras cells (26). Therefore, the failure to induce syncytium formation does not necessarily correspond to the complete lack of fusion activity in G93D mutant. To examine whether the G93D mutant could function in the viral envelope to bring about cell infection, we introduced the G93D and P104G mutations separately into an RV infectious full-length cDNA clone, pBR/M33 (28). Initially the plasmid, pBR/M33 (XbaI/HindIII) containing part of the RV nonstructural protein genes and the entire structural protein genes (nucleotides 4952 to 9762) was digested with restriction enzymes SphI and BamHI, and the SphI/BamHI fragment containing G93D or P104G mutation (Fig. 1) was recloned into plasmid pBR/M33 (XbaI/HindIII) (minus the original SphI/BamHI fragment). The resulting recombinant plasmids containing G93D and P104G mutations were digested with enzymes XbaI and HindIII, and the XbaI/HindIII fragments containing G93D and P104G mutations were isolated and inserted into pBR/M33 that had its original XbaI/HindIII fragment deleted. DNA sequencing and restriction analysis were performed to confirm the introduced mutations in the recombinant plasmids (pBR/M33/G93D and pBR/M33/P104G).

Plasmids pBR/M33/G93D and pBR/M33/P104G were linearized at the unique HindIII site, and the linearized DNAs were used as templates for synthesis of RNA transcripts. BHK cells were transfected with RNA by electroporation (14). After transfection, the culture medium was harvested daily and replaced with fresh medium. The released virus in the harvested medium was quantitated by plaque assay. The synthesis and assembly of RV structural proteins into virus particles were monitored by pulse-chase experiments. We found that protein species corresponding to RV E1 (58 kDa), E2 (37 to 45 kDa), and C (33 kDa) were detected intracellularly for the wt and both mutants (Fig. 3A). At day 3 posttransfection, the levels of intracellularly expressed protein in both mutants were significantly reduced (Fig. 3A, day 3). In the culture medium, the wt virions were steadily released into the medium at days 1, 2, and 3 posttransfection, whereas in the mutants, particularly G93D, a significant reduction in the released virus was observed at day 3 posttransfection (Fig. 3B). The reduction in virus release in the mutants could be attributed to the low capacity of mutant viruses to initiate infection in the second round of virus replication (due to the poor fusogenic properties of the mutant proteins) and not to their impaired virus assembly. If this were the case, it would be expected that the virus titers of G93D and P104G mutants in the culture media would be lower than that of the wt.

FIG. 3.

Synthesis of RV structural proteins in BHK cells transfected with full-length RNA transcripts containing G93D and P104G mutations by electroporation. Monolayers of BHK cells (35 mm) were transfected with 2 μg of RNA transcript by electroporation. Culture medium was harvested daily and replaced with fresh medium. Transfected BHK cells were labeled with [³⁵S]methionine for 3 h and chased with unlabeled methionine for 18 h. Labeled RV proteins from cellular lysates (A) and media (B) were immunoprecipitated with human anti-RV serum. wt, wild-type M33 full-length infectious clone; P104G, full-length infectious clone containing P104G mutation; G93D, full-length infectious clone containing P104G mutation. Positions of E1, E2, C, and apparent molecular weight markers (in kilodaltons) are shown.

To determine the infectivity of G93D and P104G viruses, yields of virus in harvested culture media were examined by plaque assay (28). Virus titers for G93D and P104G mutants were significantly lower than for the wt (Table 1). That of G93D virus was 0.005% of the wt value at day 1 posttransfection and increased to 0.6% of the wt value at day 3 posttransfection. In the P104G mutant, the titer was about 10% of the wt value on all 3 days posttransfection.

TABLE 1.

Effects of mutations on virus infectivity^a

Day posttransfection	Virus titer (PFU/ml)			PFU/particle ratiob
	wt	P104G	G93D	wt	P104G	G93D
1	3.0 × 106	4.5 × 105	1.5 × 102	1.7 × 103	1.6 × 102	3.0 × 10−2
2	5.0 × 106	7.0 × 105	1.5 × 103	1.6 × 103	1.4 × 102	2.1 × 10−1
3	1.2 × 107	1.0 × 106	7.0 × 104	2.0 × 103	1.8 × 102	1.6 × 101

^aVero cells infected by a serially diluted virus stock, harvested from the culture media of transfected BHK cells, were overlaid with 0.5% agarose in DMEM containing 5% FCS, incubated at 35°C for 6 days, and stained with 5% neutral red diluted in DMEM containing 5% FCS. Values are the average of three separate experiments. 

^bCalculated by using the area (in square pixels) of the radiolabel in E1 protein from virus particles (Fig. 3B). Image analysis was performed on a PC computer using the Scion Image program for Windows (beta 3b), the PC version of the public domain NIH Image program (developed at the U.S. National Institutes of Health and available on the Internet at http://rsb.info.nih.gov/nih-image/). 

To compare the defects in particle infectivity between wt and mutant viruses, we determined the ratio of virus titer to virus particles for wt and mutant viruses. The amount of virus particles in the culture medium was quantitated by measuring the intensity of radiolabel in E1 (Fig. 3B), using the Scion Image program for Windows (beta 3b; the PC version of the public domain NIH Image program). As shown in Table 1, the PFU/particle ratios clearly show that both G93D and P104G viruses are defective in particle infectivity, and the particle infectivity of wt and P104G viruses remained fairly constant on all 3 days posttransfection. Taken together, our data suggest that both G93D and P104G mutations affect virus infectivity but not virus assembly.

Analysis of G93D revertants.

We found that P104G virus formed small plaques (0.5 mm in diameter) on Vero cells, whereas the plaques of G93D virus were similar to those of wt (3 mm). It is likely that G93D virus formed tiny plaques (due to defects in infectivity), and the observed large-plaque morphology in plaque assay was due to the occurrence of revertants after 6 days of incubation. RV replicates slowly and forms microfocal plaques which are visible after 6 to 8 days of incubation.

The sharp increase in G93D virus titer at 3 days posttransfection as well as the large-plaque phenotype of G93D virus suggest a selection for revertants of G93D mutation. As a low level of virus was produced in BHK cells transfected with G93D RNA by electroporation, progeny were passaged in Vero cells to isolate more viable revertants. Vero cells were infected with wt, P104G, or G93D virus (harvested at day 3 posttransfection) at a multiplicity of 0.1 PFU/cell, and virus replication was monitored by pulse-chase experiments. As shown in Fig. 4A, at days 1 and 2 postinfection, no virus protein was detected in the wt or mutants. However, low levels of virus were detected at day 3 posttransfection in wt and G93D mutant viruses. At day 5 postinfection, similar high levels were released into the medium (Fig. 4B). The amount of released G93D virus was comparable to that of the wt. The released P104G virus was about 20% of the wt value, indicating the emergence of revertant virus in the G93D mutant. To determine the nature of G93D virus reversion, culture medium harvested from day 5 posttransfection was inoculated into a fresh monolayer of Vero cells and incubated for 3 days. After two more passages, viral RNA was isolated from virus particles and used for cDNA synthesis and subsequent PCR amplification. The sequence covering the E1 hydrophobic region was determined in three cDNA clones. In all three sequenced cDNA clones, the substituted aspartic acid residue (93D) had reverted to the original glycine residue. At present we cannot rule out the possibility of other second-site mutations in G93D revertants. No revertant was found in P104G virus stock. This result indicates that the Gly93 residue is critical in membrane fusion and that the increase in virus titers of G93D mutant at days 2 and 3 posttransfection was due to the occurrence of G93D revertants.

FIG. 4.

Synthesis of RV structural proteins in Vero cells infected with wt, P104G, and G93D viruses. Monolayers of Vero cells (35 mm) were infected with wt, P104G, or G93D virus harvested from BHK cells at day 3 posttransfection at a multiplicity of 0.1 PFU/cell. Infected cells were labeled with [³⁵S]methionine for 3 h and chased with unlabeled methionine for 18 h. Labeled RV proteins from cellular lysates and media were immunoprecipitated with human anti-RV serum. (A) Lysates from infected Vero cells at day 1, 2, and 3 postinfection. (B) Labeled RV structural proteins from lysates and media at day 5 postinfection. wt, wild-type M33 virus; P104G, P104G virus; G93D, G93D virus. Positions of E1, E2, C, and apparent molecular weight markers (in kilodaltons) are shown.

Fusion activities of the mutant viruses.

Since a low level of infectious G93D virus was produced in BHK cells transfected with full-length RNA transcript carrying the G93D mutation, it is possible that the inability of G93D/E2E1 mutant protein to induce syncytium formation (27) does not reflect a complete lack of fusion activity. Therefore, we examined the pH dependence of virus-induced polykaryon formation using mutant viruses harvested from BHK cells transfected with infectious RNA transcripts. Vero cells were infected with wt, G93D, or P104G virus, harvested at day 3 posttransfection at 1 PFU/cell, and incubated for 40 or 64 h at 37°C. The infected cells were treated with fusion medium (pH 4.8) for 20 min at 37°C, washed with growth medium (pH 7.0), and incubated with the growth medium at 37°C for an additional 4 h. The polykaryons formed were viewed under a phase-contrast microscope. At 40 h postinfection, cultures of cells expressing the wt and P104G mutant had 60 and 10%, respectively, of fused cells, but no polykaryons were observed with the G93D mutant (Fig. 5A to C). At 64 h postinfection, cell lysis occurred in wt-infected cells, and about 15 and 10% of fused cells were observed in the cell cultures of P104G and G93D mutants, respectively (Fig. 5D). The syncytium formation observed in G93D virus-infected cells could be attributed to the presence of G93D revertants in virus stock. To investigate this possibility, we analyzed the pH-dependent syncytium formation of BHK cells expressing VLPs carrying the G93D mutation (BHK-24S/G93D). Transformed BHK cells were incubated with growth medium containing 40 μM zinc sulfate for 16 h to induce expression of RV structural proteins (20). Induced BHK cells were treated with fusion medium and incubated with growth medium as described above. The percentages of fused cells observed were 50 to 60, 20, and 5% for BHK-24S (wt), BHK-24S (P104G), and BHK-24S (G93D), respectively (data not shown). It is interesting that a low level of polykaryons was detected in BHK-24S (G93D) cells. The likely explanation for not detecting polykaryon formation in E2E1 (G93D) mutant in the previous studies (27) could be the instability of E2-E1 heterodimers in the absence of capsid protein, as soluble E2 and E1 proteins were observed in the medium after a longer time of chase (27). These results indicate that G93D virus possesses a very limited fusogenic activity that correlates well with its low virus infectivity.

FIG. 5.

Syncytium formation in Vero cells infected with wt, P104G, and G93D viruses. Infected cells were exposed to fusion medium at pH 4.8 for 20 min, incubated in regular growth medium for 4 h, fixed, and photographed. Vero cells were infected with wt virus (A), P104G virus (B), or G93D virus (C) at 40 h postinfection or with G93D virus at 64 h postinfection (D).

The initial assay used to characterize the mutants described in this study relied on the capacity of the spike proteins to induce cell-cell fusion in transformed BHK cell lines expressing mutant E2 and E1 structural proteins (27). To examine whether mutant G93D proteins that do not induce syncytium formation can function in viral entry, we incorporated the G93D mutation into virus particles and examined whether the resulting virions were infectious. Similarly, P104G mutant protein exhibiting limited fusogenic activity was examined for its effect on virus infectivity. Alteration of the Pro104 residue to glycine reduced virus infectivity by 90%, whereas substitution of an aspartic acid residue for glycine resulted in a more dramatic drop in virus infectivity (Table 1). The low virus infectivity observed for G93D and P104G mutant viruses is due to the inability of the mutant G93D and P104G viruses to induce fusion of the viral envelope with an intracellular membrane to initiate infection in virus replication, not to defects in virus assembly or the low affinity of mutant protein to bind Vero cells. We found that the assembly of VLPs was not affected by G93D and P104G mutations (Fig. 2), and in binding assays using ³⁵S-labeled wt and mutant viruses, no difference in binding efficiency between the wt and mutant viruses was observed (data not shown).

The endocytic entry mechanism predicts that viruses requiring a low pH for fusion will be sensitive to inhibition by agents that neutralize endosomal pH. We have examined the entry properties of RV by treatment with NH₄Cl, a weak base widely used in studies of the role of low vacuolar pH, and found that RV replication was inhibited by 90% in the presence of 20 mM NH₄Cl prior to addition of virus to Vero cells and throughout virus entry and postentry (data not shown). Since NH₄Cl inhibits acidification of endosomes (15), this result suggests that RV infects cells by the endosomal pathway.

In SFV, mutations in the E1 putative fusion peptide were shown to shift the pH threshold of cell-cell fusion (G91 to A), or block cell-cell fusion completely (G91 to D), when spike proteins were transiently expressed (12). Use of the SFV infectious clone revealed that both mutations conferred a virus assembly defect that was partially reversible at 28°C (2). G91A virus had limited secondary infection and an acid-shifted fusion threshold, whereas G91D was defective and inactive in both cell-cell and virus-liposome fusion assay (7). This differs from our finding in RV, in which the assembly of RV structural proteins into VLPs was not affected by G93D and P104G mutations. It is of interest that in both SFV-G91D and RV-G93D viruses, substitution of a charged aspartic acid for glycine blocks infectivity and cell-cell fusion activity.

Many enveloped viruses, such as influenza virus and SFV, undergo fusion within a cellular endocytic vesicle. Acidification of the endosome is thought to induce a conformational change of a viral envelope glycoprotein, which mediates the fusion event. The requirement for vesicular acidification for viral entry generally assumes that a similar environment is required for virus-mediated cell fusion or formation of syncytia. However, the inability to induce syncytium formation does not necessarily correspond to complete lack of fusion activity, as shown in Moloney murine leukemia virus (26). This finding raises the possibility that different epitopes on the viral envelope glycoprotein may be involved in these two fusion events. In RV, the effect of NH₄Cl on virus entry and the good correlation between block in fusion activity and virus infectivity seems to indicate that the internal hydrophobic domain of RV E1 is involved both in cell fusion after virus entry and in syncytium formation.

In this study, we have shown that VLP assembly was not affected by G93D and P104G mutations in RV E1. VLPs carrying either G93D or P104G mutation resembles the wt in HA activity. G93D virus is barely fusogenic and infectious with a high frequency of reversion to wt. Although the studies reported here do not permit us to definitely conclude that the E1 internal hydrophobic domain is the fusion peptide of RV, our evidence strongly suggests the direct involvement of this domain in fusion events.