Genetic Determinants of Rous Sarcoma Virus Particle Size

Abstract

The Gag proteins of retroviruses are the only viral products required for the release of membrane-enclosed particles by budding from the host cell. Particles released when these proteins are expressed alone are identical to authentic virions in their rates of budding, proteolytic processing, and core morphology, as well as density and size. We have previously mapped three very small, modular regions of the Rous sarcoma virus (RSV) Gag protein that are necessary for budding. These assembly domains constitute only 20% of RSV Gag, and alterations within them block or severely impair particle formation. Regions outside of these domains can be deleted without any effect on the density of the particles that are released. However, since density and size are independent parameters for retroviral particles, we employed rate-zonal gradients and electron microscopy in an exhaustive study of mutants lacking the various dispensable segments of Gag to determine which regions would be required to constrain or define the particle dimensions. The only sequence found to be absolutely critical for determining particle size was that of the initial capsid cleavage product, CA-SP, which contains all of the CA sequence plus the spacer peptides located between CA and NC. Some regions of CA-SP appear to be more important than others. In particular, the major homology region does not contribute to defining particle size. Further evidence for interactions among CA-SP domains was obtained from genetic complementation experiments using mutant ΔNC, which lacks the RNA interaction domains in the NC sequence but retains a complete CA-SP sequence. This mutant produces low-density particles heterogeneous in size. It was rescued into particles of normal size and density, but only when the complementing Gag molecules contained the complete CA-SP sequence. We conclude that CA-SP functions during budding in a manner that is independent of the other assembly domains.

Particle assembly of retroviruses is conducted by the Gag polyprotein, which is able to direct budding at the plasma membrane in the absence of the pol and env gene products or of the viral RNA genome (40). Gag proteins are synthesized on free ribosomes in a cytosolic compartment and are then transported to the plasma membrane. At this point, roughly 2,000 Gag molecules interact to drive the budding process. During or shortly after budding, the viral protease (PR) cleaves the Gag molecules into their mature protein products (23), the matrix (MA), capsid (CA), and nucleocapsid (NC); however, proteolytic activity is not required for particle production. For Rous sarcoma virus (RSV), the focus of this report, several additional cleavage products—p2a, p2b, p10, spacer peptides (SP), and PR itself—are liberated upon proteolytic cleavage (see Fig. 1) (2, 21, 29, 31). Particles released when Gag is expressed alone are identical to authentic virions in their rates of budding, proteolytic processing, and core morphology, as well as density and size.

FIG. 1.

Alterations of the RSV Gag protein. The wild-type (WT) RSV Gag protein (Myr0) is shown at the top. The vertical lines represent viral PR cleavage sites, which separate the mature Gag products (MA, p2a, p2b, p10, CA, SP, NC, and PR) as indicated. The horizontal bar above the Gag protein denotes CA-SP (CA plus the spacer peptides). The shaded region within CA marks the MHR. Numbers refer to amino acid residues. Thick, solid bars underneath the Gag protein indicate assembly domains required for budding (M, L, and I). Illustrated below Myr0 are the deletion and substitution mutants utilized in this study. The open box at the amino terminus of some of the constructs denotes the M domain of the Src oncoprotein (Myr1). The shaded region in H32R.ΔMB and Δp2b.ip6 indicates substitution of the RSV Gag M and L domains for the HIV equivalents, respectively. The vertical line in Myr1.L171I indicates a substitution of leucine for isoleucine. The column to the left lists the names of the mutants. The column to the right summarizes the size distribution of the mutants as follows: U, particles that are uniform in size; Sm, particles that are uniform in size but smaller in diameter; Het, particles that are heterogeneous in size.

The mechanism by which Gag directs budding is unknown, but significant advances have been made in dissecting this complex event. Studies with RSV and human immunodeficiency virus (HIV) Gag have revealed three small assembly domains, M, L, and I (Fig. 1), that are required for particle formation. The M domain provides a membrane-binding function and is located at the amino terminus in both the RSV and HIV Gag proteins (1, 28, 42). The L domain maps to a proline-rich sequence within RSV p2b and the C-terminal p6 sequence in HIV Gag. It is believed to operate very late in budding, at the membrane-bud separation step (16, 18, 27, 39). The interaction (I) domain, which is present in two copies within NC, mediates productive interactions between Gag proteins to produce particles of the proper buoyant density (1, 38, 43). Further evidence for the important interactions mediated by I domains is the observation that previously characterized large-deletion mutant RSV Gag proteins lacking these domains cannot be rescued into particles when coexpressed with wild-type Gag. However, mutant Gag proteins that retain a functional copy of I but lack either the M or L domain can still be rescued (39, 42).

While mutations within the assembly domains block or severely cripple particle formation, regions that lie outside these domains can be removed with no effect on the release of particles of normal density. Moreover, although the primary amino acid sequences and the order of M, L, and I differ between RSV and HIV, these functions are fully exchangeable between the two Gag proteins (1, 27, 28).

While much progress has been made in understanding how Gag functions during budding, there is little evidence to suggest how it defines the size of a particle. There is some indication in the literature that the CA sequence is important. In the case of RSV, it has been found that large deletions that remove half of CA and most of the p10 sequence result in particles that are very heterogeneous in size (38), but those studies are difficult to interpret because of the severity of the deletions. Other studies have suggested that regions surrounding the CA sequence are critical for particle size and morphology. For instance, it has recently been demonstrated that the addition of CA to the NC sequences of RSV and HIV enables the resulting CA-NC protein, in the presence of RNA, to self-assemble in vitro into tubular structures (3). In the case of RSV, addition of further N-terminal sequences, for example, p10, p2b, p2a, and part of MA, results in the assembly of spherical particles with dimensions similar to those of authentic virions (4). Along with these findings, linker insertions and deletions made within the CA domain of HIV Gag have been found to alter the size of the particles (5, 10, 32). Taken together, these results indicate that additional functions within Gag are needed to constrain the size of the emerging particle but they do not define the exact regions involved.

To precisely map the size-determining regions of RSV, we have employed rate-zonal gradients and electron microscopy (EM) to systematically study the effect of deletions throughout the Gag protein. The only region we found to be critical for determining particle size is the CA-SP sequence, which is the initial capsid species released from Gag and which contains all of the CA sequence plus the spacer peptides. Some regions of this capsid intermediate appear to be more important than others. Thus, while the CA-SP sequence of RSV Gag is dispensable for budding, it is critical for producing particles of uniform and proper size.

MATERIALS AND METHODS

Previously described Gag constructs.

The RSV gag gene was obtained from pATV-8, an infectious molecular clone of the RSV Prague C genome (33). All of the gag alleles (see Fig. 1) were expressed in simian (COS-1) cells by using the simian virus 40-based mammalian expression vector we have described previously (42). Many of these alleles have been previously reported: pSV.Myr0 (wild type) (42), pSV.Myr1 (42), pSV.Myr1.3h (37), pSV.ΔMA1 (42), pSV.H32R.ΔMB (28), pSV.Myr1.ΔMA6 (25), pSV.Myr1.ΔMA7 (25), pSV.Myr1.ΔMA8 (25), pSV.Myr1.ΔMA9 (25), pSV.Myr0.Δp2a (39), pSV.Myr0.Δp2b.ip6 (27), pSV.Myr1.R-3K (39), pSV.Myr1.R-3A (39), pSV.Myr1.R-3J (38, 39), pSV.Myr1.DM1 (38, 39), pSV.Myr1.Es-Bg (8), pSV.Myr1.L171I (8), pSV.Myr1.LOC1 (8), pSV.Myr1.LOC2 (8), pSV.Myr0.ΔSP3 (30), pSV.Myr1.ΔSP9 (7, 30), pSV.Myr0.ΔSP12 (30), pSV.Myr1.Bg-Xm (8), pSV.Myr1.LON1 (8), pSV.Myr1.Sm-Bs (39), and pSV.Myr1.ΔNC (43). In some cases, the activity of the retroviral protease was eliminated in these constructs by substituting the aspartic acid in the active site with either isoleucine (D37I) (6, 34, 41) or serine (D37S) (6), changes which have no effect on particle release, density, or size (see below). All plasmids were propagated in Escherichia coli DH-1 cells and selected by using medium containing 100 μg of ampicillin per ml.

Construction of additional Gag mutants.

Mutant pSV.Myr1.ΔMA6E was constructed by digesting pSV.Myr1.ΔMA6 with MluI and SpeI and discarding the small fragment. The sticky ends were then treated with Klenow fragment to realign the gag reading frame, and the plasmid was religated. This deletion encodes a protein which contains one foreign residue (Asp) between amino acids 11 and 99 of Gag.

Deletion mutants ΔQM1, Δp10.31, Δp10.52, LOC3, LOC4, LOC5, LOC6, LOC7, LOC8, LOC9, and LOC10 were constructed by oligonucleotide-directed mutagenesis using uracylated, single-stranded MGAG DNA in a manner similar to that described previously (42). The sequences of the mutagenic oligonucleotides are indicated here along with diagnostic restriction endonuclease sites (with the sequence of the sites underlined and the enzymes in parentheses): ΔQM1, 5′-CCTCCTCCTCCTTATGCGGCCGCGGAACAGTCAAGG-3′ (NotI); Δp10.31, 5′-GGGAGAGCAGCAGGGCAGGGTCAGGGAGGAGC-3′; Δp10.52, 5′-AGTGGTTTGTATCCTACTAGTCCCGTGGTGGCCATG-3′ (SpeI); LOC3, 5′-CTGTAGTGATTAAGACTAGTTTGATCACAAGACTG-3′ (SpeI); LOC4, 5′-ACGGTCAGGACCAAGACTAGTGCGCTTATGTCCTCC-3′ (SpeI); LOC5, 5′-CCGCATGACGTCACTAGTTATGCCTTATGGATGG-3′ (SpeI); LOC6, 5′-TGGGGAGTCCAACTCACTAGTCACCCAGCGAACGG-3′ (SpeI); LOC7, 5′-GGGGGGAACGGACTAGTGTGGGCAACCCACAG-3′ (SpeI); LOC8, 5′-GCCGCATTATTAAGAACTAGTCAGGCGTTTAGAGA-3′ (SpeI); LOC9, 5′-GTTGAGGGGTCAGATACTAGTTGCTTTAGGCAGAAGT-3′ (SpeI); LOC10, 5′-ACAGCACCCTCCACTAGTCTAGACAGGCAGAAG-3′ (SpeI, XbaI). As a result of including restriction endonuclease sites to these oligonucleotides, extra amino acid residues were introduced at the site of the deletion: ΔQM1 contains two foreign amino acids (Ala-Ala), as do Δp10.52, LOC4, LOC6, LOC8, and LOC9 (Thr-Ser), but LOC3, LOC5, LOC7, and LOC10 each contain one foreign amino acid (Ser). All mutations were confirmed by DNA sequencing. The gag alleles were then transferred from the replicative-form DNA into the pSV.Myr0 plasmid by exchanging the XhoI-BlpI fragments (for mutants Δp10.31 and Δp10.52), XhoI-BglII fragments (for mutants LOC3 to LOC8), and BlpI-EcoRI fragments (for mutants LOC9 and LOC10). ΔQM1 was transferred into plasmid pSV.Myr1.D37S by an XhoI-BlpI exchange. Recombinants were screened by restriction endonuclease mapping, and two independent clones from each mutagenesis experiment were characterized to confirm that no unwanted mutations were found elsewhere in the gag gene.

Transfection of cells.

COS-1 cells were grown in Dulbecco’s modified Eagle medium (Gibco BRL) supplemented with 3% fetal bovine serum and 7% bovine calf serum (Hyclone, Inc.). RSV-infected turkey embryo fibroblasts were propagated in supplemented F10 medium as previously described (20). COS-1 cells in 60-mm-diameter plates were transfected by the DEAE-dextran-chloroquine method as described previously (41). Before transfection, the plasmid DNAs were digested with XbaI and ligated at a concentration of 25 μg/ml. This step removes the bacterial plasmid sequence and joins the 3′ end of the gag gene with the simian virus 40 late polyadenylation signal for high-level expression (41). Plasmid pSV.Myr0.LOC10 was prepared for transfection by digestion with BssHII and ClaI, treatment with Klenow fragment, and religation of the plasmid. This step, which has the same outcome as XbaI digestion, was exploited since an extra XbaI site had been created by the LOC10 deletion. Typically, 1 μg of DNA was applied to each monolayer. For cotransfections, the cells received 0.5 μg of DNA each.

Metabolic labeling and sucrose gradient analysis.

At 48 h after transfection, COS-1 cells were starved for 0.5 h in methionine-free, serum-free Dulbecco’s medium and then labeled in 0.6 ml of labeling medium supplemented with l-[³⁵S]methionine (50 μCi, >1,000 Ci/mmol) for 8 h. After the labeling period, the medium from each plate was collected and transferred to a microcentrifuge tube, and cellular debris was removed by centrifugation at 15,000 × g for 1 min. For the rate-zonal gradients, labeled particles containing a Gag protein of wild-type density and size was mixed with the samples to provide an internal control. In some cases, infectious RSV was included as the internal control. This virus was grown in turkey embryo fibroblasts and labeled with l-[³⁵S]methionine as described above. The mixture was then layered onto 11.5-ml, 10 to 30% sucrose and centrifuged at 83,500 × g (26,000 rpm) at 4°C for 30 min in a Beckman SW41Ti rotor. Fractions (0.6 ml) were collected through the bottom of each tube and mixed with lysis buffer containing protease inhibitors. The Gag proteins in each fraction were immunoprecipitated with a rabbit antiserum against RSV (reactive with MA, CA, NC, and PR), electrophoresed in a sodium dodecyl sulfate–12% polyacrylamide gel (SDS-PAGE), and detected by fluorography as previously described (37, 38, 41). The resulting films were quantitated by laser densitometry. All gradients were repeated at least once to confirm the results.

Virus-like particle isolation and analysis by EM.

At 48 h posttransfection, medium containing virus-like particles was collected and prepared for EM as previously described (34). Briefly, cellular debris was removed from the medium by low-speed centrifugation at 11,950 × g for 10 min in a Sorvall SS-34 rotor. The particles contained in the supernatant were then pelleted through a 15% sucrose–STE (100 mM NaCl, 10 mM Tris [pH 7.5], 1 mM EDTA) cushion at 108,760 × g for 90 min in a Beckman 50.2 Ti rotor. The particle-containing pellets were softened in STE for 1 h at 4°C before resuspension by pipetting and storage at 4°C. Particles were negatively stained with 2% uranyl acetate.

For thin sections, transfected cells were fixed for 2 h in 0.1 M sodium cacodylate (pH 7.4)–3% glutaraldehyde, washed in 0.1 M sodium cacodylate (pH 7.4), and then postfixed in 1% OsO₄ in the same buffer for 2 h at 4°C. The cells were then rinsed in 0.1 M sodium maleate (pH 5.2), stained with 1% uranyl acetate in 0.1 M sodium maleate (pH 6.0) for 1 h in the dark, rinsed again with sodium maleate, and serially dehydrated with ethanol. The cells were lifted from the plates with propylene oxide and pelleted by centrifugation for 2 min at 15,000 × g in a microcentrifuge before embedding in Spurr embedding medium. Thin sections were counterstained with 2% uranyl acetate and lead citrate.

RESULTS

To ascertain whether particle size determinants could be mapped to a particular region of RSV Gag, we studied the effects of deletions throughout this protein (Fig. 1). To this end, we employed a transient mammalian cell expression system in which RSV-like particles are efficiently produced (41, 42). In this system, the wild-type Gag protein (designated Myr0 to indicate the lack of myristate at the N terminus; 41) drives the release of particles that are identical to authentic virions in terms of their rates of budding, core morphology, size, density, and proteolytic processing of the mature cleavage products (1, 2, 7, 37, 41, 42).

All of the mutants in this study, with just two exceptions, produce particles of normal density (1.16 to 1.18 g/ml) (9). One exception is mutant ΔNC (Fig. 1), which lacks both I domains and therefore produces particles that are dramatically lower in density (1.14 to 1.15 g/ml). The other exception is a group of mutants in which the RSV M domain is completely replaced with smaller membrane-binding domains (Fig. 1, ΔMA1, Myr1.ΔMA6E, and H32RΔMB). The density shift in this case is minor, however, and the particles band at a density (1.15 to 1.16 g/ml) that overlaps the normal range for wild-type retroviral particles.

Control experiments.

Particles produced in the transient expression system were analyzed for size by using rate-zonal sedimentation gradients. For this, culture supernatants containing radiolabeled particles were harvested and cellular debris was removed. A radiolabeled Gag protein of wild-type size was always added to the supernatants to provide an internal control, and the particle mixtures were layered onto 11.5-ml, 10 to 30% sucrose gradients and centrifuged for 0.5 h at 83,500 × g. After centrifugation, the gradients were fractionated and Gag proteins were immunoprecipitated and separated by SDS-PAGE. The resulting X-ray films were then subjected to scanning densitometry to determine the position and amount of Gag protein in the gradients. Under these sedimentation conditions, particles migrate according to their relative sizes. When interpreting the rate-zonal gradients, it is important to note that the position of the peak fraction relative to the internal control and the distribution of the particles in the gradient (e.g., heterogeneous versus uniform) are more important than the heights of the peaks. Rate-zonal gradients provide an important advantage over EM analysis, in that they reveal large differences in particle size and provide information on the total population of particles released from the cell. That is, Gag proteins will be detected whether they are present in particles of recognizable morphology or not.

To determine whether Gag proteins produced in our transient expression system are of a size similar to authentic, wild-type virus, infectious RSV produced in turkey embryo fibroblasts were run in a gradient along with Gag-only particles produced from COS-1 cells. To distinguish the proteins in the transiently produced particles from those of the authentic virus, a Gag mutant was employed that lacks protease activity (Myr0.D37S) and therefore releases only uncleaved Gag precursors. Following centrifugation, the two types of particles were found in the same fractions (Fig. 2A), and the distribution of particle sizes was uniform and homogeneous in each case. Thus, proteolytic maturation of the Gag protein does not influence the size of the particles. This was confirmed in an experiment in which the protease-positive and protease-negative Gag-only particles were both produced by transient expression (Fig. 2B). The addition of a foreign membrane-binding domain to the N terminus of Gag (i.e., the Src membrane-binding domain, which is present in many of our constructs) also did not affect particle size (Fig. 2C). Moreover, this substitution, combined with nearly complete deletion of the protease (mutant 3h, which lacks the last 117 amino acids of Gag), also had no effect on size (Fig. 2D), as previously reported (38). From these control experiments, it appeared that the extremities of Gag do not control particle size.

FIG. 2.

Control experiments. COS-1 cells transfected with the indicated gag derivatives or RSV-infected turkey embryo fibroblasts were labeled with [³⁵S]methionine for 8 h. After the labeling period, the medium from each plate was collected and mixed with labeled control particles. The mixture was then layered onto 10 to 30% sucrose and centrifuged at 83,500 × g at 4°C for 0.5 h. Fractions were collected through the bottom of each tube, immunoprecipitated with a polyclonal rabbit antiserum against RSV, electrophoresed in an SDS–12% polyacrylamide gel, and detected by fluorography. The autoradiogram was then quantitated by laser densitometry. Arrows indicate the direction of sedimentation.

Further evidence of this was obtained by examining the Src chimera (Myr1) and protease deletion mutant (3h) by thin-section EM. Both produced budding structures and released particles typical of C-type retrovirus morphogenesis (Fig. 3). Due to the presence of an active protease, Myr1 was processed to produce electron-dense cores, as expected (Fig. 3A). Mutant 3h particles lack the PR domain and therefore retained the concentric ring structure typical of immature particles (Fig. 3B). Three concentric rings were observed. The outer ring was associated with the lipid envelope. Ten nanometers further toward the center was another, lighter-staining ring. The innermost, darkly staining ring had a diameter of about 40 nm and was located about 10 nm central to the middle ring. Consistent with the rate-zonal gradient data, the particles produced by both Myr1 and 3h were homogeneous in size and shape.

FIG. 3.

Thin-section EM of cells and virus-like particles. At 48 h posttransfection, cells were examined by thin-section EM as described in Materials and Methods. (A) Myr1 particles were homogeneous in size with condensed cores. (B) 3h particles had immature morphology with three rings. (C) Es-Bg particles had noncondensed cores. (D) R-3J particles were heterogeneous in size with condensed cores. Most cores were not in the plane of section or were acentric (arrows). (E) R-3J budding particles; initial bud (left), late bud (center), and released mature particle (right). (F) R-3J.D37S particles were heterogeneous in size with immature morphology. The peripheral material often had gaps. Only two rings were visible. (G) R-3J, flattened patches of protein accumulated on the surface of cells. (H) R-3J.D37S particles budding between patches of accumulated protein. Bars, 100 nm.

The Gag derivatives also were examined by negative staining. Most of the particles did not allow the stain to penetrate the lipid envelope and therefore only provided information on the overall shape and size of the particles. In some instances, the stain did enter the particle to reveal the internal structure (Fig. 4). Central cores produced by proteolytic maturation were evident in Myr1 (Fig. 4A). Penetration of the stain into the center of the particles which had an inactive PR (D37I in Fig. 4B) or lacked a PR (3h in Fig. 4C) suggested that the center of the particle was hollow, as expected, with the protein located at the periphery. Striations similar to those reported for immature HIV (19, 26, 36) were also clearly visible in D37I and 3h (Fig. 4B and C, respectively).

FIG. 4.

Negative-stain EM of virus-like particles. At 48 h posttransfection, virus-like particles were collected by centrifugation and negatively stained with 2% uranyl acetate as described in Materials and Methods. A, Myr1; B, D37I; C, 3h; D, Es-Bg; E, R-3J.D37S; F, DM1. The arrow in C indicates the striated pattern in D37I and 3h. Bars, 100 nm.

The diameters of individual negatively stained particles were determined from photographic negatives (Table 1). The average diameters of Myr1, D37I, and 3h were essentially the same and identical to previously reported measurements for other retroviruses (9). However, it is interesting that even these homogeneous particles are not identical in size but display some variability in particle diameter. This has been observed for other retroviruses, too. For instance, it has been reported that authentic HIV particles vary in diameter between 90 and 160 nm (32) or 95 and 175 nm (10), which is remarkably consistent with the values obtained for Myr1, D37I, and 3h (Table 1). Because all of these Gag derivatives make normal-size particles, they could be used as controls in subsequent experiments to map the genetic determinants of particle size.

TABLE 1.

Analysis of size distribution of virus-like particles by negative staining^a

Gag allele	No. of particles measured	Avg diam (nm) of particles (SD)	Low (nm)b	High (nm)c
Myr1	26	103 (19)	70	150
D37I	29	108 (23)	90	180
3h	24	98 (10)	90	120
R-3J	46	125 (39)	90	240
R-3J.D37S	24	146 (56)	90	270
DM1	16	216 (78)	120	370

^aThe diameters of negatively stained particles were measured from photographic negatives. 

^bLow, diameter of the smallest particle measured. 

^cHigh, diameter of the largest particle measured. 

Replacement of the first half of MA with smaller M domains.

We began our systematic analysis of RSV Gag with mutants that lack sequences within the first half of MA. We have previously reported that small deletions within the first 85 residues of MA, which constitute the M domain, are defective for budding; however, budding is restored when the membrane-binding domain from Src is placed at the amino terminus (1, 42). In mutants ΔMA1 and ΔMA6E, the complete M domain (contained in segments of 84 and 98 residues, respectively) has been replaced with the small Src membrane-binding domain (Fig. 1). When analyzed for particle size, both mutants produced a uniform population of particles that were slightly smaller than the internal control (Fig. 5A and B). The shift of the peak to a position two fractions higher in the gradient was quite reproducible (data not shown). It may be that removal of the bulky 85-amino-acid M domain of RSV allows the membrane to be pulled closer to the core, thereby reducing slightly the diameter of the particle; alternatively, the lower overall mass of the particles might result in the slight shift (see Discussion). This phenotype was not limited to the Src chimeras but was found in all chimeras in which the RSV M domain had been replaced with a smaller M domain, including mutant H32RΔMB (Fig. 5C), which has the 32-residue-long M domain of HIV Gag in place of the first 99 residues of RSV Gag (28), and FynΔMB, in which the first 99 amino acids of MA are replaced with the membrane-binding domain of the Fyn oncoprotein (data not shown). The precise explanation of this minor shift to a higher position in the gradient remains to be determined (see Discussion); however, we conclude from these results that the M domain does not contribute greatly to particle size.

FIG. 5.

Deletions that lead to smaller particles. Particle sizes were analyzed as described in the legend to Fig. 2.

Deletions within the second half of MA, p2, and p10.

The next set of four deletions span the second half of MA, which is dispensable for particle assembly and infectivity in avian cells (25). These mutants (ΔMA6, ΔMA7, ΔMA8, and ΔMA9) collectively lack the residues from 87 in MA to 161 within p2a (Fig. 1). For the most part, these deletions had no effect on particle size (Fig. 6A to D). In the case of ΔMA6, the particles were slightly smaller than the internal control (Fig. 6A) but the density was identical to that of the wild type (data not shown). Moving further down the Gag protein, we found that when all of p2a was deleted, particles of uniform size were released as well (Fig. 6E). It was not possible to analyze a p2b deletion, since this cleavage product contains the proline-rich L domain and its removal blocks particle release (27, 39). However, it was possible to test a chimera, Δp2b.ip6 (Fig. 1), in which the p2b domain of RSV Gag has been replaced with the L domain from p6 of HIV Gag (27). The foreign amino acid sequence had no effect on particle size (Fig. 6F).

FIG. 6.

Alterations in the second half of MA, p2a, p2b, and p10. Particle sizes were analyzed as described in the legend to Fig. 2.

Next, we analyzed mutants that lack various amounts of the p10 sequence. ΔQM1 lacks the first third of p10, while Δp10.31 and Δp10.52 have internal deletions (Fig. 1). ΔQM1 released particles as efficiently as the wild type, but Δp10.31 and Δp10.52 exhibited reduced levels (data not shown). The lower yield of particles from these two p10 mutations was consistent with previously published deletions within p10 (11). The reduction in budding for these latter two mutants is probably due to a conformational problem because large mutants with most of p10 and a large amount of CA deleted (such as R-3A and R-3J [Fig. 1]) release particles at wild-type levels (39). Nevertheless, particles produced by all three of our p10 mutants were homogeneous and uniform in size (Fig. 6G to I). Δp10.31 and Δp10.52 appeared to sediment slightly more slowly than the control particles, similar to the M domain substitutions (Fig. 5); however, unlike ΔMA1, ΔMA6, and H32RΔMB, the p10 deletion mutants possessed wild-type density (data not shown). It may be that large deletions within p10 decrease the distance between the membrane-binding domain and the core, resulting in smaller particles (see Discussion), but this does not explain why ΔQM1 is not shifted to the same extent.

Collectively, the results shown so far indicate that deletions within MA, p2, and p10 (i.e., the first third of RSV Gag) have no effect on uniform particle release. Our next step was to determine whether deletions within the CA sequence would alter particle size.

Large internal deletions within Gag.

To analyze what impact CA deletions have on particle size, we initially made use of three large internal deletion mutants (R-3K, R-3A, and R-3J in Fig. 1) which lack various amounts of p10 and CA. R-3K, R-3A, and R-3J have been previously shown to produce dense particles at the same efficiency as wild-type Gag (39). However, the particles released by these mutants were found to be extremely heterogeneous in size, with material spread throughout the gradient (Fig. 7A to C). Mutant DM1, which combines the R-3J and 3h deletions (38, 39), produced a heterogeneous profile of particles as well (Fig. 7D). Because p10 deletion mutants are not heterogeneous in particle size, we hypothesized that the defects of these large deletion mutants would map to the CA sequence (see below).

FIG. 7.

Large deletions spanning the p10-CA junction. Particle sizes were analyzed as described in the legend to Fig. 2.

To corroborate the rate-zonal gradient data and demonstrate that the heterogeneous profile of particles was not a result of aggregation, we analyzed three of these large deletion mutants by thin-section and negative-stain EM. R-3J and R-3J.D37S differ in having or not having an active protease, respectively. Both produced heterogeneously sized particles ranging in size from normal to extremely large, disrupted particles (Fig. 3D and F, respectively). For the larger particles produced by R-3J, the cores were either not present within the plane of section, aberrant, or off center (Fig. 3D). It was not clear whether the core was located at a fixed distance from the lipid envelope, which in normal-size particles would place it at the center, or whether the off-center cores represent a random distribution of free-floating cores. When the PR was inactivated (R-3J.D37S), the viral protein remained associated with the lipid envelope in discontinuous patches of electron-dense material and no cores were evident (Fig. 3F). A horseshoe-shaped patch of dense material was commonly observed, as if the circumference of the particles was not completely enclosed with protein. Similar results have been reported for HIV when small deletions were made in the N-terminal half of CA (10). Normal budding structures were present for both R-3J and R-3J.D37S (Fig. 3E), but large accumulations of protein underneath the plasma membrane, with little if any curvature, were the predominant structures observed in the cells (Fig. 3G and H, respectively).

Negative-stain EM analysis of R-3J.D37S (Fig. 4E) and DM-1 (Fig. 4F) confirmed that these particles are heterogeneous in size. Penetration of the stain into the center of the particles with an inactive PR (R-3J.D37S in Fig. 4E) or lacking the PR sequence (DM1 in Fig. 4F) suggested that the center of the particle was hollow, as expected, with the protein located at the periphery. When the size distributions of R-3J, R-3J.D37S, and DM-1 were quantitated from negative-stain images, the particles were found to have a very heterogeneous profile, as predicted by the sedimentation analysis (Table 1). The EM results may underrepresent the number of larger particles because numerous disrupted particles were observed, but only spherical particles were counted. In particular, the R-3J and R-3J.D37S Gag proteins differed only in PR activity, but R-3J produced smaller particles on average. Presumably, the largest mature particles from R-3J were less stable during purification or negative staining than the same-size immature particles from R-3J.D37S. DM-1 produced an even wider range of particle sizes as measured by EM, but such differences could not be detected in gradients.

Small deletions within CA.

Having found that large deletions that extend into CA result in heterogeneously sized particles, we decided to see what effect much smaller mutations solely within CA would have. It was possible that certain regions within CA would be critical for determining particle size, with others being dispensable. Mutants LOC3 through LOC8 lack 10- to 11-amino-acid segments between the beginning of CA and the major homology region (MHR; Fig. 1). All of these mutants produced particles that were heterogeneous in size (Fig. 8A to F). Some produced more of a broad peak which overlapped the control particles (LOC3 to 5), while others produced material that was decidedly larger than the control (LOC6 to 8).

FIG. 8.

Small deletions within CA. Particle sizes were analyzed as described in the legend to Fig. 2.

A larger deletion mutant, Es-Bg, which lacks the MHR along with some flanking sequences, produced particles with a broad peak that overlapped control particles (Fig. 8G). Thin-section EM (Fig. 3C) and negative-stain EM (Fig. 4D) of this mutant revealed particles that were heterogeneous in size. However, both EM and rate-zonal gradient data analysis demonstrated that Es-Bg particles are not as heterogeneous in size as some of the other CA deletions. The internal morphology of Es-Bg was interesting. Instead of a central, collapsed core, material seemed to be evenly distributed throughout the volume of Es-Bg particles (Fig. 3C and 4D) even though an active protease is present (data not shown; see reference 8). The thin sectioning and negative staining suggested that Es-Bg has a defect in core assembly.

To look more closely at the MHR, we made use of mutant L171I (8), in which a conserved Leu residue within the MHR is replaced with Ile. L171I has no effect on particle release, but when this point mutation is built back into the viral genome, the resulting viruses are noninfectious in avian cells. When analyzed for size, homogeneous particles were observed (Fig. 8H). Thus, although the MHR region may be critical for proper maturation of the viral core (8), it does not play an important role in defining particle size.

Four additional deletions within the last quarter of the CA sequence also produced heterogeneously sized particles (Fig. 8I to L). Mutants LOC1 and LOC2 appeared as broad peaks that overlapped the internal control, whereas LOC9 and LOC10 particles were more heterogeneous. While the smaller deletions within CA produced particles with various degrees of heterogeneity, all of the CA mutations analyzed (with the exception of L171I) had some effect on particle size. Thus, it appears that CA provides a very critical determinant of particle size.

Spacer peptide deletions.

The CA sequence is initially released from Gag with a small (12-residue) spacer peptide at its C terminus following cleavage between the peptide and the NC sequences (Fig. 1) (7). This form of CA, previously referred to as CA1 (7), is referred to here as CA-SP for clarity. Over the course of several hours after particle release, cleavages within the spacer peptide result in the appearance of two new products that actually run more slowly in SDS-PAGE (2, 7). Recent studies (30) have demonstrated that in mature virus, the CA protein exists as fully mature CA (formerly named CA2) and a form of CA that retains three residues of the spacer, CA-S (formerly referred to as CA3). When precise deletions of these spacer peptides were made (mutants ΔSP3, ΔSP9, and ΔSP12; Fig. 1) and the peptides were separately expressed in avian cells, virions were efficiently assembled, but none of the mutants were infectious (7, 30). Sedimentation analysis revealed that all three mutants were heterogeneous in size (Fig. 9A to C). Thus, it appears that the CA and SP sequences in Gag provide a very critical size determinant.

FIG. 9.

Spacer peptide deletions. Particle sizes were analyzed as described in the legend to Fig. 2.

Deletions within NC.

Another mutant, Bg-Xm (Fig. 1), contains a deletion which removes that portion of the CA sequence downstream of the MHR and the first third of NC, effectively removing the spacer peptides and the surrounding sequence. This deletion produced heterogeneously sized particles (Fig. 9D), which we attributed to deletion of the C-terminal region of CA-SP. However, it was also possible that the extension of the deletion into NC contributed to the defect in particle size. This was explored by using NC deletion mutants.

LON1 and Sm-Bs, which lack sequences within NC and retain one copy of the I domain (Fig. 1), produced particles that were uniform and homogeneous in size (Fig. 10A and B). The slightly smaller size relative to the internal control of these uniformly sized particles might be a consequence of the reduced mass resulting from these large deletions (see Discussion). In contrast, ΔNC, with almost all of NC deleted, produced particles that were heterogeneous in size (Fig. 10C). The latter result was not surprising, since ΔNC lacks both copies of the I domain and releases particles with low density (43). If the I domains are not present, proper interactions cannot take place among the Gag proteins and heterogeneous particles with low density are produced. Thus, along with having an intact CA-SP domain, the Gag protein must have at least one copy of the I domain to produce particles that are uniform and homogeneous in size.

FIG. 10.

Deletions within NC. Particle sizes were analyzed as described in the legend to Fig. 2.

Complementation rescue mediated by CA-SP.

If CA-SP controls the size of RSV, then it most probably does so through self (i.e., CA-SP–CA-SP) interactions. The properties of mutant ΔNC provided an opportunity to test this idea. This is the only mutant we have found that produces heterogeneously sized particles even though it retains the complete CA-SP sequence. We hypothesized that the CA-SP sequence is properly folded in this mutant but the absence of I domains results in local concentrations of Gag that are too low to permit self interactions (i.e., CA-SP interactions themselves are too weak to create high-density particles).

To test the ability of ΔNC to participate in Gag interactions, it was coexpressed with Gag molecules that have all the assembly domains (M, L, and I) and either a mutant or a complete CA-SP sequence (illustrated in Fig. 11, top row). When CA deletion mutant R-3J or Es-Bg was used, no interactions were observed, as shown by the continued appearance of mutant ΔNC in particles of lower density (Fig. 11, left column). Similar results were obtained with mutant R-3K (data not shown). In contrast, when ΔNC was coexpressed with mutant 3h (which lacks protease but retains CA-SP), it was found in particles normal in both density and size (Fig. 11, right column). The simplest interpretation of this result is that the CA-SP sequence of ΔNC is indeed properly folded and provides a means for the mutant Gag protein to be rescued into normal particles.

FIG. 11.

Complementation rescue mediated by CA-SP. (Top row) To test the ability of ΔNC to participate in Gag interactions, it was coexpressed with Gag molecules that have all the assembly domains (M, L, and I) and either a mutant or a complete CA-SP sequence. The white boxes indicate the CA sequence. COS-1 cells were cotransfected with ΔNC and the rescuing gag allele. At 48 h after transfection, cells were labeled with [³⁵S]methionine for 8 h. After the labeling period, the medium from each plate was collected. The particles were then layered onto 10 to 50% sucrose and centrifuged to equilibrium at 83,500 × g at 4°C for 16 h. Fractions were collected through the bottom of each tube, immunoprecipitated with a polyclonal rabbit antiserum against RSV, electrophoresed in an SDS–12% polyacrylamide gel, and detected by fluorography. The autoradiogram was then quantitated by laser densitometry. Left column, density gradient of cotransfected R-3J and ΔNC, as well as Es-Bg and ΔNC, particles. Right column, density and rate-zonal gradients of cotransfected 3h and ΔNC particles. Particle sizes were analyzed as described in the legend to Fig. 2. Arrows indicate the direction of sedimentation.

DISCUSSION

In this study, we have shown that the size determinants within the RSV Gag protein map to just one region, CA-SP, the segment of Gag consisting of CA plus the spacer peptides located between CA and NC. CA-SP is the first capsid protein species to be released from Gag. It previously was referred to as CA1 (7, 30), but we have adopted the new name to better clarify its structure, which is analogous to that of the initial cleavage product of HIV, variously called p25, CA-p2, or CA-SP1. While the sequence comprising CA-SP lies outside of the assembly domains (M, L, and I) and is completely dispensable for budding, this study demonstrates that it plays a fundamental role in constraining the size of the emerging particle. We use the term “particle” in a broad sense, operationally defined as a Gag protein released into the medium in a particulate form with a density similar to that of wild-type virions. However, in those cases where we have looked, objects with the appearance of true virus-like particles have always been seen (i.e., particles are membrane enclosed, have electron-dense cores, etc.).

The major method we employed to analyze particle size, rate-zonal sedimentation in sucrose gradients, has both advantages and limitations that bear on the interpretation of the data presented. The advantages include the display of the entire population of particles and the standardization provided by internal markers. Among the limitations are the relative lack of sensitivity of sedimentation rate to small changes in size. For particles different in sizes but invariant in density, this rate is directly proportional to the mass and inversely proportional to the frictional coefficient. For example, for a sphere of uniform density, a doubling in mass increases the diameter by the 1/3 power and thus the frictional coefficient (which is proportional to the cross section) by the 2/3 power. As a consequence, the sedimentation rate will increase only by a factor of 2^1/3, or about 26%. This value corresponds to only a few fractions of the collected gradient. Seen in this light, even small changes in the peak positions of the deletion mutants imply significant mass differences, with particles near the bottom of the gradients having masses many times larger than that of wild-type virions. For deletion mutants showing the most heterogeneous sedimentation profiles, the budded particles may not be spherical, as suggested by some of the EM results, and thus may have increased frictional coefficients and even larger sizes than expected of spherical particles with this sedimentation rate. On the other hand, a sedimentation profile coinciding with that of wild-type virions does not necessarily imply a completely homogeneous distribution of particles. Recent cryo-EM measurements of murine leukemia virus (44) and baculovirus-expressed HIV virus-like particles (12) show that, unlike icosahedral viruses, these retroviruses are not entirely uniform in size. It remains to be seen to what degree those results can be extrapolated to avian retroviruses. However, as reported here (Table 1), homogeneous RSV Gag virus-like particles do appear to exhibit variability in particle diameter.

It should be noted that analysis of one of our mutants, ΔSP9, using rate-zonal gradients (Fig. 9B) is at odds with our previously published EM data which showed that this mutant re-leases particles identical to those of the wild type (7). While ΔSP9 can produce particles similar in morphology and size to wild-type virions, a certain subset of the population is more heterogeneous and may have been overlooked by EM. The inability to recognize abnormally sized particles by EM is a limitation of that method of measuring particle size.

Model of RSV particle assembly.

Our understanding of how particle assembly occurs during budding is illustrated in Fig. 12. The M, L, and I domains of the RSV Gag protein provide the minimal budding machinery (Fig. 12A). The M domain directs the Gag molecules to the plasma membrane; there, neighboring Gag molecules interact through their I domains via RNA (1, 38, 43). These strong interactions lead to tight packing of the Gag molecules which, in turn, allows the CA-SP sequences to establish contact with one other. The adjacent CA-SP interactions give proper curvature to the bud as it emerges from the surface of the cell. L is believed to function late in budding to separate the viral particle from the cell surface. The result of this process is a viral particle of uniform shape, density, and size.

FIG. 12.

Model of RSV particle assembly. Assembly domains M (black ovals), L (black triangles), and I (black circles) are indicated. The squiggly line represents the CA-SP sequence. The horizontal black line denotes the cell membrane. (A) Wild-type Gag proteins interact to form a budding particle of uniform size and density. (B) Gag mutants that harbor a CA deletion produce heterogeneous particles of normal density. (C) Gag mutant that lack functional I domains (open circles) but retain a properly folded CA-SP domain produce heterogeneous particles of low density.

Deletions within CA-SP do not affect the rate of release of high-density particles, since all three assembly domains are intact (Fig. 12B). However, CA-SP deletions introduce disorder into the size determinant, and when these crippled Gag molecules accumulate at the plasma membrane, the CA-SP sequence is no longer able to constrain the size of the growing particle. This results in the accumulation of large, electron-dense patches of Gag protein underneath the plasma membrane (Fig. 3G and H). These patches rapidly pinch off the cell surface, producing heterogeneous particles of altered morphology that range in size from normal to extremely large.

Interactions provided by CA-SP alone are insufficient both to constrain particle size and to provide the tight packing of Gag protein needed for high density (Fig. 12C). Gag molecules that lack the I domains (as indicated by the open ovals) but retain a properly folded CA-SP domain are targeted to the plasma membrane; however, the wild-type CA-SP domains have difficulty interacting with one another, and this results in the release of heterogeneous particles that are light in density. Complementation rescue experiments demonstrated that such an I domain mutant can be rescued into dense particles of uniform size when coexpressed with a Gag protein containing an intact CA-SP sequence. Thus, it appears that while CA-SP is the major size determinant, it is dependent upon NC to bring the neighboring CA-SP domains into proper juxtaposition for interaction. Together, CA and NC provide the core interactions around which particle assembly occurs, as demonstrated by the self-assembly of in vitro-expressed RSV and HIV CA-NC in the presence of RNA (3). Thus, as previously suggested (8), it appears that CA and NC function as a unit although they are proteolytically cleaved during maturation. CA-SP may organize the viral protein within the particle as NC binds viral RNA. We predict that the assembly functions associated with CA-SP and NC can be replaced with capsid proteins of nonretroviral origin that are capable of self-assembly in vivo, and such experiments are in progress.

Effects on core morphology.

While extensive deletions outside of CA-SP had no effects on particle size, every deletion we analyzed within CA-SP yielded particles that were abnormal in size. Some mutations produced very large and heterogeneous particles (R-3J), whereas others were not as varied (Es-Bg). Es-Bg was particularly interesting because it lacks the MHR and some of the flanking sequence. The MHR is the most highly conserved region within Gag proteins and is involved in the maturation of the virion after budding (8, 24, 35). RSV MHR mutants exhibit defects in core stability and display blocks to infectivity upon entry into the host cell (8). Interestingly, EM analysis of Es-Bg revealed no darkly staining central core but, instead, electron-dense material seemed to be evenly distributed throughout the particle, consistent with a defect in core assembly. This is in stark contrast to R-3J, in which the first half of CA is deleted yet it still forms electron-dense cores. These data suggest that the first half of CA is situated on the outside, while the second half (including the MHR) lies toward the center of the core. This interpretation is compatible with models of the HIV capsid structure (13, 14, 17).

Smaller mutants.

While the CA-SP sequence provides the major size determinant, replacement of the large membrane-binding domain of RSV Gag with smaller ones invariably resulted in slightly smaller particles (Fig. 5). Because the RSV membrane-binding domain is in tight contact with the viral membrane, substitution of a physically smaller M domain may take up less space along the membrane and therefore result in a smaller particle. It also appears that MA may contribute to a small degree to the density of the particle. The idea of MA-MA interactions is consistent with previous reports that I domain mutants can still make particles, although these are low in density (1, 38, 43).

Two large deletions within p10 (Δp10.31 and Δp10.52) appeared to produce smaller particles as well (Fig. 6H and I, respectively); however, the density of these particles was not altered. In this instance, p10 may act as a spacer region between the membrane-binding domain and the core (CA and NC). Deletion of this spacer would effectively bring the core closer to the membrane-binding domain and thus produce a smaller particle without altering its density. We predict that insertion of large polypeptides into the second half of MA or within p10 may result larger particles.

Although the membrane-binding domain mutants and p10 deletions produce particles that sediment more slowly in a rate-zonal gradient than the internal control, it could be argued that the lower overall mass of the particles contributed to the slight shift. While this is a plausible explanation for the difference in migration of these mutants, the protease deletion mutant 3h, in which 17% of Gag is deleted, is perfectly normal in size and diameter (Fig. 2D, 3B, and 4C; Table 1). In contrast, the p10 deletion mutants and the mutants with smaller membrane-binding domains sediment slightly more slowly, even though they have much smaller deletions.

Size determinants in other retroviruses.

Although a systematic search for size determinants has not been done for any other retrovirus, there are reports in the literature that are consistent with our findings. Linker insertion mutations and deletions within HIV CA have been shown to produce particles heterogeneous in size with diameters of 75 to 315 nm (32). These estimates are similar to the values we have reported for the RSV CA deletions in Table 1 (R-3J, R-3J.D37S, and DM1). Moreover, previously published data on the spacer peptide between CA and NC of HIV demonstrated that cells expressing the protein containing a precise deletion of this region released heterogeneous particles that were noninfectious (22). However, there is reason to believe that our findings with RSV may not apply to all retroviruses. In particular, work in our laboratory suggests that the p6 sequence at the end of HIV Gag is very important (15). Clearly, methodical searches for size determinants in other Gag proteins are warranted.

In summary, we do not know much about the nature of the interactions that the CA-SP domain provides, but it is clearly important for constraining particle size during budding. Our results suggest that studies of the mature capsid protein may not reveal the relevant structures needed during assembly and budding since the spacer peptides are absent. Thus, a greater emphasis on the structural properties of CA-SP is warranted.