The CA-SP sequence controls the size of RSV particles in cooperation with the I domains in the NC sequence (13
). This prompted us to test the possibility that foreign capsid proteins can substitute for CA-SP-NC. We initially chose the major SV40 capsid protein, VP1, for several reasons. First, it contains a quantity of amino acids similar to that of CA-SP-NC (338 for CA-SP-NC versus 361 for VP1). Second, when expressed in the absence of all the other SV40 gene products, it self-assembles into icosahedral capsids within the nucleus of the cell, and when calcium ionophores are present, it also assembles in the cytoplasm (17
), where retroviral budding takes place. Third, the particles produced by VP1 are similar in size to the spherical particles produced in vitro by ΔMA-ΔPR, a slightly longer form of CA-SP-NC that also contains the N-terminal p2 and p10 peptides of RSV Gag (40 nm for VP1 versus 50 nm for ΔMA-ΔPR) (7
We began by constructing two Gag-VP1 chimeras (Fig. ). Myr1.VP1 has VP1 in place of the last few residues of p10 and most of CA but retains the last 25% of CA as well as SP, NC, and PR. In Myr1.VP1t, the VP1 sequence is linked to the same N-terminal portion of Gag, but the C-terminal Gag sequence is absent. Both chimeras utilize MSrc
, which is functionally equivalent to the M domain of RSV Gag (36
), and both contain the L domain of RSV, which is required for the virus-cell separation step late in budding. However, Myr1.VP1t lacks both of the I domains required for the production of particles of normal density (2
Release of Gag-VP1 chimeras into the medium.
The Gag-VP1 chimeras were expressed in COS-1 cells by an expression system that has been used in many previous studies of RSV Gag. Expression of the wild-type Gag protein (referred to as Myr0; Fig. ) results in the release of virus-like particles into the growth medium (Fig. A, left panels), and these have been shown to be identical to authentic RSV (and the Src chimera Myr1) in terms of their rate of release, core morphology, size, density, and proteolytic maturation (2
). When the viral PR is inactivated by a single amino acid substitution (Myr0.D37I), budding continues at the normal rate, but Gag cleavage products are not released into the medium (Fig. A, right panels).
FIG. 2 Release of Gag-VP1 chimeras within virus-like particles. COS-1 cells were transfected with the indicated DNAs; 48 h later, the cells were labeled with [35S]methionine for 2.5 h. (A) Three clones of Myr1.VP1 (with PR) and three clones of (more ...)
Three clones of Myr1.VP1 and three clones of Myr1.VP1t were tested in the COS-1 cell system, and all were found to be released into the medium. For Myr1.VP1, which contains the viral PR, several proteolytic cleavage products were identified (Fig. A, left panels); two of these correspond to known Gag products, namely, PR and the p23 form of MA. The migration of p23 was slightly slower than that observed for wild-type Gag (Myr0), presumably because of the addition of myristate, which results when MSrc is present. The amounts of PR and p23 released into the medium were the same as for the wild-type control (Myr0), and the Myr1.VP1 chimera was found to be released with normal efficiency. As expected, the three closely migrating species of CA normally seen with the wild type (Myr0) were not produced by this chimera. Instead, a 48-kDa product was observed; this product was slightly smaller than that expected (53 kDa) for a chimeric CA containing most of p10, all of VP1, and the C-terminal segment of CA (Fig. A, left panels). The reason for this small discrepancy is unknown. DNA sequence analysis did not reveal any errors in the construct, and a precursor protein of the expected mass for the uncleaved Gag-VP1 species (97 kDa) was observed in the cell lysates. If the faster migration is the result of cleavage by the viral PR, then the portion removed must be small and no more than the p10 fragment (53 residues) or the CA fragment (62 residues) linked to the N and C termini of VP1, respectively.
For Myr1.VP1t, which lacks the viral PR, only one major Gag-related product was found in the medium (Fig. A, right panels); the budding efficiency of this chimera was found to be somewhat diminished (about 50%) relative to that of the control. Surprisingly, the apparent mass of the major product (74 kDa) was larger than expected (67 kDa) and very close to that of the uncleaved Gag protein (Myr0.D37I; 76 kDa); however, a faint signal of the proper size was observed in the lysates along with the larger species. This size discrepancy is consistent with the addition of ubiquitin (76 amino acids; 8.5 kDa), which has been shown to be involved in the retroviral budding mechanism (21
). Moreover, a ladder of higher-molecular-weight species was observed in both the lysates and the medium samples for this chimera, a result which would be expected if multiple ubiquitin molecules were added.
Further proof that Myr1.VP1 and Myr1.VP1t indeed carry the VP1 sequence was obtained by immunoprecipitating the chimeric proteins with antisera against VP1 (Fig. A, right panel) and SV40 (data not shown). Evidence that the released proteins were in a particulate form was obtained by sedimentation analysis (see below).
FIG. 3 Requirement of the M and L domains for the release of VP1 chimeras. Viral proteins from cell lysates and growth medium were analyzed as described in the legend to Fig. . The positions of the full-length Gag protein (Pr76) and its detectable (more ...) Mechanism of release for Gag-VP1 chimeras.
It has been reported that SV40 particles can be released from polarized cells by a mechanism that (like budding) does not involve cell lysis (8
). Therefore, it was of interest to examine the pathway utilized by the Gag-VP1 chimeras. Several approaches were used, and all of the results argue strongly that the retroviral budding pathway is utilized by the chimeras.
(i) Trypsin resistance.
If the chimeras travel the retroviral budding pathway to reach the growth medium, then they should be surrounded by a lipid bilayer, which would protect them from digestion with exogenously added PRs. Analyses of the chimeras revealed that both are released in a trypsin-resistant form (Fig. B), including the MA and PR species produced by Myr1.VP1 (data not shown). (The partial sensitivity of Myr1.VP1t in this particular experiment was not reproducible.). Trypsin resistance was lost when detergent was added, consistent with the removal of protective membranes. These characteristics are identical to those of wild-type Gag (Myr0; Fig. B) (14
) but are unlike those of Gag proteins with a signal peptide (SPG; Fig. B) (14
), which travel the secretory pathway and are released into the medium in a trypsin-sensitive, soluble form. These results suggest that the chimeras are membrane enclosed.
(ii) Necessity of the M and L domains for Gag-VP1 release.
Retrovirus budding requires an M domain for plasma membrane targeting and an L domain for particle release. If the VP1 chimeras travel the budding pathway, then the removal of either domain should abolish their release into the medium. To inactivate MSrc
, we used a previously described mutant, Myr1(−), which fails to bud (Fig. A) because the site of myristylation has been destroyed (4
). This change was introduced into both of the Gag-VP1 chimeras to create Myr1(−).VP1 and Myr1(−).VP1t (Fig. ), and the resulting myristate-lacking forms were found to be incapable of particle release despite good levels of expression in the cells (Fig. A); hence, the M domain is required.
To examine the importance of the L domain, we deleted this region from the Myr1.VP1 construct to make Myr1.VP1.ΔL (Fig. ). This deletion does not remove any of the other regions of Gag essential for budding and is similar to that of a mutant named T10C, which can be rescued into virus particles by complementation using budding-competent Gag proteins (4
) and which becomes budding competent on its own when the L domain of human immunodeficiency virus type 1 or equine infectious anemia virus is fused to its C terminus (20
). We found that the removal of the L domain from Myr1.VP1 resulted in a block of budding (Fig. B), indicating that the release of the chimera requires the L domain as well as the M domain.
(iii) Subcellular localization.
VP1 contains nuclear targeting information (12
) which enables it to be efficiently transported to the nucleus, where it is rapidly assembled into virus-like particles (17
). If the M and L domains of Gag efficiently direct VP1 into the budding pathway, then the chimeras should not be found in the nucleus. Consistent with this notion, cell fractionation and immunofluorescence experiments using various sera (antisera to RSV, SV40, and VP1) failed to reveal the presence of the chimeras in the nucleus, although T antigen (which is constitutively expressed in COS-1 cells) was readily detectable in the nucleus in both assays (data not shown). Moreover, the chimeras remained cytoplasmic even when the M domain was inactivated by eliminating the site of myristylation (data not shown). It is possible that the nuclear targeting information of VP1, which resides near the N terminus (18
), is masked by fusion to the p10 sequence. In any case, it is clear that the attached Gag sequences are dominant over the nuclear targeting signals of VP1, enabling this capsid protein to bud from the plasma membrane of the cell.
VP1 can substitute for retroviral I domains.
The I domains within the RSV Gag protein (Fig. ) provide the major regions of interaction and enable the tight packing of molecules needed for the efficient release of particles of normal density (2
). When the I domains are absent, fewer particles are produced and those that are released have a lower density, as measured in isopycnic sucrose gradients (e.g., mutant ΔNC; Fig. A). Deletions elsewhere in Gag have little or no effect on particle density (13
). Because Myr1.VP1 contains the RSV I domains, we expected the particles produced by that chimera to have the same density as control particles, and this was found to be the case (Fig. B). However, it was unclear what to expect for Myr1.VP1t. This chimera buds with an efficiency that is reduced only twofold relative to that of RSV Gag (Fig. A; compare with Myr0.D37I), even though it lacks both of the RSV I domains and all of the CA-SP sequence (Fig. ). Analysis of Myr1.VP1t particles in density gradients revealed that they have a density that is indistinguishable from that of particles produced by wild-type Gag (Fig. C), a result that strongly suggests that interactions provided by VP1 can substitute for those of the I domains. This hypothesis is further supported by a deletion mutant that lacks the C-terminal 109 amino acids of VP1 (Myr1.VP1ΔABt; Fig. ) and that was found to have reduced particle density (Fig. D). To what extent the assembly of Myr1.VP1t (or Myr1.VP1) bears resemblance to that of authentic SV40 will require further investigation, but the following experiment shows that there is some similarity.
FIG. 4 Particle density of the Gag-VP1 chimeras. COS-1 cells transfected with the indicated Gag derivatives were labeled with [35S]methionine for 8 h. The medium from each plate was collected and mixed with labeled control particles of normal (more ...) Gag-VP1 chimeras respond to calcium ionophores.
To ascertain whether VP1 can substitute for the size determinant of RSV, we used rate-zonal sedimentation analysis. Retroviral particles move quickly through the gradient as a well-defined band, and these were used as an internal control within each gradient (Fig. ). As previously reported (13
), mutants of RSV that have deletions within the CA-SP sequence make particles that are very large and heterogeneous (e.g., mutant Myr1.R3J; Fig. A, left panel). Similar heterogeneity was also observed with both of the Gag-VP1 chimeras (Fig. B and C, left panels). However, it is well documented that the assembly of VP1 requires calcium (11
) and that particles are not found in the cytoplasm (where retroviral budding is initiated) unless ionophores are used to increase the levels of calcium in this compartment (17
). Therefore, we repeated the sedimentation analysis using particles obtained from cells that were treated with the calcium ionophore A23187
. Although the CA-SP mutant did not respond to A23187
(Fig. A, right panel), both Gag-VP1 chimeras did and, as a result, the particles became much more uniform in size (Fig. B and C, right panels). Moreover, the chimeric particles were found to sediment more slowly than the internal control particles; the peaks were shifted two (Myr1.VP1) and four (Myr1.VP1t) fractions higher in the gradient, indicative of a considerable change in particle size. The smaller size of Myr1.VP1t particles probably was due to the absence of the mass contributed by the long C-terminal Gag sequence and not to the presence of the short foreign peptide (Fig. ) because an identical shift in sedimentation rate was seen when this peptide was removed (data not shown). In contrast, the particles from Myr1.VP1ΔABt did not appear to respond significantly to increased levels of calcium (Fig. D). The inability of this chimera to produce uniformly sized particles could be explained in part by its limited ability to produce particles of the proper density (Fig. D). Collectively, these results suggest that VP1 has some assembly capabilities when inserted into the budding pathway.
FIG. 5 Calcium ionophores change the size of Gag-VP1 particles. COS-1 cells expressing the indicated Gag derivatives were labeled with [35S]methionine in the presence or absence of the calcium ionophore A23187. The medium from each plate was (more ...) Gag-STMV chimeras.
Having found that VP1 can substitute for the assembly functions contained within CA-SP-NC, we next constructed Gag-STMV chimeras to see whether our findings could be extended to a capsid protein from another nonenveloped virus. STMV encapsidates a single-stranded RNA genome and requires coinfection with tobacco mosaic virus for propagation (1
). This plant virus is much smaller (only 17 nm) than SV40 and RSV, and its coat protein is half the size of CA-SP-NC (159 versus 338 amino acids). It does not require calcium for assembly, as would be predicted for a virus that replicates in the cytoplasm.
The STMV capsid sequence was inserted into Gag in a fashion similar to that used for the Gag-VP1 chimeras (Fig. ). Myr1.STMV contains all of the domains required for budding (M, L, and I), whereas Myr1.STMV.t lacks the I domains and all of the C-terminal sequences of Gag as a result of a stop codon immediately after the STMV sequence. These two chimeras lack some or all of the RSV size determinant, respectively.
The expression of the Gag-STMV chimeras in COS-1 cells revealed that they are released into the growth medium (Fig. ) in particulate form (see below). As expected, Myr1.STMV was proteolytically processed, resulting in the release of MA p23, PR, and a chimeric CA protein of the predicted size (30 kDa) into the medium. Several species that migrated more slowly than the CA-STMV fusion protein were also detected, and these most likely were processing intermediates, as were the species that migrated more slowly than the CA bands in the control (Myr0). Because of the incomplete processing of the precursor, it was more difficult to estimate the efficiency of budding by comparing the amounts of p23 and PR relative to those in the wild-type control, but it is clear that the STMV sequence does not have a severe impact on budding. This was also found to be true for Myr1.STMV.t, which lacks the viral PR and which was released as a single protein species of the expected mass (43 kDa). The relatively weaker radioactive signal obtained with this chimera is in large part due to the reduced numbers of methionines available for labeling (21 for the Myr0 control versus 8 for Myr1.STMV.t).
FIG. 6 Release of Gag-STMV chimeras into the growth medium. COS-1 cells were transfected with the indicated DNAs and analyzed as described in the legend to Fig. . The positions of the full-length Gag protein (Pr76) and its detectable cleavage (more ...)
To ascertain whether the STMV capsid has any assembly capabilities when linked to Gag, we analyzed the particles in sucrose gradients. If this foreign capsid had no activity, then particles of very large and heterogeneous size would be expected in rate-zonal gradients, much like those of CA-SP mutants (Fig. A, left panel). This is not what we found. Both of the Gag-STMV chimeras produced particles that were smaller than those of the wild-type RSV control (Fig. A and B), and the distributions of these particles in the gradient were much tighter than that of a typical CA-SP mutant (Fig. A, left panel). Because wild-type STMV particles are very small, this phenotype is not surprising. Indeed, the particles produced by Myr1.STMV.t were even smaller than those of Myr1.STMV, perhaps due to the reduced mass following the removal of the C-terminal Gag sequence in the former chimera.
FIG. 7 Analysis of particle density and size of the Gag-STMV chimeras. Particles released from COS-1 cells expressing the indicated Gag-STMV chimeras were mixed with control retroviral particles and analyzed for size (A and B) or density (C and D) in sucrose (more ...)
The Gag-STMV particles were also analyzed in isopycnic gradients. Myr1.STMV contains the density-determining region (I domains) of RSV Gag and, as expected, this chimera was found to have a density very similar to that of the wild-type control (Fig. C). However, removal of the C-terminal Gag sequence resulted in a shift to a lower density (Myr1.STMV.t; Fig. D). The reason for this shift is unclear. If the STMV sequence within this chimera was unable to interact with itself, then particles of very large and heterogeneous size would be expected, as is the case for low-density particles produced by I domain mutants (13
). Because such particles were not observed (Fig. B), we favor the idea that the reduced mass of Myr1.STMV.t particles accounts for their lower density. In any case, the data indicate that the STMV coat protein can promote interactions among Gag proteins when inserted into the retroviral budding pathway.