Analysis of Mason-Pfizer Monkey Virus Gag Domains Required for Capsid Assembly in Bacteria: Role of the N-Terminal Proline Residue of CA in Directing Particle Shape

Abstract

Mason-Pfizer monkey virus (M-PMV) preassembles immature capsids in the cytoplasm prior to transporting them to the plasma membrane. Expression of the M-PMV Gag precursor in bacteria results in the assembly of capsids indistinguishable from those assembled in mammalian cells. We have used this system to investigate the structural requirements for the assembly of Gag precursors into procapsids. A series of C- and N-terminal deletion mutants progressively lacking each of the mature Gag domains (matrix protein [MA]-pp24/16-p12-capsid protein [CA]-nucleocapsid protein [NC]-p4) were constructed and expressed in bacteria. The results demonstrate that both the CA and the NC domains are necessary for the assembly of macromolecular arrays (sheets) but that amino acid residues at the N terminus of CA define the assembly of spherical capsids. The role of these N-terminal domains is not based on a specific amino acid sequence, since both MA-CA-NC and p12-CA-NC polyproteins efficiently assemble into capsids. Residues N terminal of CA appear to prevent a conformational change in which the N-terminal proline plays a key role, since the expression of a CA-NC protein lacking this proline results in the assembly of spherical capsids in place of the sheets assembled by the CA-NC protein.

In the infected cell, the assembly of immature retrovirus capsids occurs by one of two morphogenetically different pathways. In type C retroviruses (e.g., Rous sarcoma virus [RSV]) and primate lentiviruses, the capsid is assembled as it buds from the plasma membrane, while in type B and D viruses (e.g., mouse mammary tumor virus and Mason-Pfizer monkey virus [M-PMV], respectively), immature capsids are preassembled within the cytoplasm prior to transport to the plasma membrane. The immature retroviral capsids formed from the Gag polyprotein precursors are spherical and measure approximately 80 to 110 nm in diameter in thin-section electron micrographs. During or immediately after budding, the virus-encoded protease is activated and cleaves the Gag polyprotein precursor into at least three structural proteins: matrix protein (MA), capsid protein (CA), and nucleocapsid protein (NC). MA remains associated with the inner leaflet of the membrane, while CA forms the surface shell of an electron-dense “core” that encloses a nucleoprotein complex of NC and viral RNA.

We have previously shown that the Gag polyprotein of M-PMV can assemble into procapsids in eukaryotic cells (mammalian, insect, and yeast), in prokaryotic cells, and in an in vitro cell-free system (22, 25, 28, 29, 31, 33; T. Ruml, unpublished data). The gag gene encodes the three domains, MA (p10), CA (p27), and NC (p14), found in all replication-competent retroviruses, as well as additional proteins, pp24/16 (hereafter designated PP), p12, and p4; these are arranged in the following order: NH₂-MA (p10)-PP-p12-CA (p27)-NC (p14)-p4-COOH.

Using both in vivo and in vitro studies, several groups have focused on the domains of human immunodeficiency virus (HIV) type 1 (HIV-1) and RSV necessary for the process of assembly. In vivo studies with both RSV and HIV have shown that the bulk of MA and the N terminus of CA are dispensable for capsid assembly and budding as long as a membrane-targeting domain is located at the N terminus and the NC domain is intact (13, 35, 37). In general, the NC domain or an equivalent region C terminal of CA that can mediate the association of Gag molecules appears to be critical for the assembly of capsids with a density resembling that of wild-type virus (2, 23, 36, 38, 39, 40). Similar conclusions were drawn from in vitro binding experiments with HIV type 1 Gag deletion mutants (3). In simian immunodeficiency virus, MA alone can mediate the formation of virus-like particles, but these have an abnormally low density (15).

In vitro it was found that the HIV-1 capsid protein could assemble into tubular structures with a diameter ranging from 32 to 55 nm at high protein and salt concentrations (10, 17, 19, 34). In contrast, the in vitro assembly of CA-NC occurred at a 20-fold-lower concentration of protein and in low salt but required the addition of RNA (17). This finding is consistent with the results observed with RSV and HIV-1 CA-NC molecules by Campbell and Vogt (5). Surprisingly, the addition of as few as four or five amino acids to the N terminus of HIV-1 CA-NC resulted in a switch from the assembly of tubular structures to the assembly of spherical capsid-like structures (18, 34). However, these capsids were either significantly smaller in diameter (55 nm) than virions (110 nm) (34) or heterogeneous in size and shape (18).

A nearly full-length RSV Gag polyprotein was shown to assemble into spherical particles both in Escherichia coli and in vitro (6), but similar molecules lacking the p10 domain assembled into cylinders. Therefore, it is believed that for RSV it is the p10 domain that is involved in defining the assembly of spherical particles both in E. coli and in vitro. Campbell and Rein (4) have also reported the in vitro assembly of spherical particles from HIV-1 Gag lacking just the p6 domain, although in general, these were smaller (25 to 30 nm) than those normally assembled at the plasma membrane.

Ganser et al. (11) reported the in vitro assembly of an HIV-1 CA-NC fusion protein in the absence of RNA. Both conical and cylindrical structures were observed, but only under conditions of very high ionic strength. They hypothesized that RNA promotes CA-NC assembly by concentrating CA-NC and/or by neutralizing charges and that the polynucleotide chain itself may be dispensable in this process. Recently, a conformational switch controlling tubular or spherical particles assembled in vitro has been suggested for an HIV-1 CA-NC fusion protein retaining the amino terminus of MA (19). The SP1 spacer peptide between CA and NC appeared to control this process, and its presence was required for the formation of spherical structures.

To examine the requirements for and contributions to the process of capsid assembly of the individual domains within the M-PMV Gag polyprotein, we have constructed a series of truncated gag genes encoding both C- and N-terminal deletions for analysis with the bacterial expression and assembly system that we have described earlier (22, 29). Here, we report that in this bacterial overexpression system, the NC domain was essential for the assembly of macromolecular structures. In contrast, none of the domains N terminal of CA was necessary for assembly, although constructs with amino acid extensions at the N terminus of CA programmed spherical particle formation rather than sheets. Remarkably, this requirement for N-terminal residues was shown to be a function of the N-terminal proline of CA, since CA-NC molecules lacking the proline or in which alanine was substituted for the proline could efficiently assemble spherical capsids in E. coli.

MATERIALS AND METHODS

Expression plasmids.

All plasmids are based on the parental construct p10GAG, which is the plasmid encoding the entire M-PMV gag gene in the pGEMEX-2 bacterial expression vector and in phagemid pSIT (1, 22). This construct was created by deletion of a 3.2-kbp fragment by partial digestion with ApaI from pG10MNX as described previously (22). All cloning steps were carried out by established techniques that are described elsewhere (32). The cloning strategies and details of the PCR primers can be obtained upon request from the authors. None of the constructs resulted in any mutation within the Gag-derived products except for MA-CA-NC, where a KpnI site was created between sequences encoding MA and CA. The multiple clones were characterized by restriction analysis. All the newly created mutations as well as the 5′- and 3′-terminal regions of the inserts were verified by DNA sequencing. The sizes of all of the expressed proteins were confirmed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis and corresponded to those predicted from their sequences (data not shown). None of the expressed proteins was unstable, even those found in the soluble fraction of the E. coli lysate (e.g., MA and MA-PP). Protein sequence analyses showed that the N-terminal methionine is removed in bacterial cells and therefore that the amino-terminal amino acid in the CA-NC construct in which the N-terminal proline was deleted and replaced by the initiating methionine [pro(−)CA-NC] is valine.

Bacterial expression.

Luria-Bertani medium containing ampicillin (final concentration, 100 μg/ml) was inoculated with E. coli BL21(DE3) cells carrying the appropriate construct to an optical density at 590 nm of approximately 0.1. When the cells reached an optical density at 590 nm of 0.8, expression was induced by the addition of isopropyl-β-d-thiogalactopyranoside (IPTG) to a final concentration of 0.4 mM. The cells were harvested 4 h postinduction by centrifugation at 5,000 × g for 10 min in a Beckman JA-18 rotor. The cells were lysed with lysozyme in Tris-HCl (pH 8.0, 50 mM) buffer containing 1 mM EDTA and 100 mM NaCl. The cells were sonicated four times for 25 s each time and then incubated at 4°C for 30 min in the presence of sodium deoxycholate (final concentration, 0.1%). The cell lysate was centrifuged at 10,000 × g for 10 min in a Beckman JA-20 rotor. The pellet was washed three times in a buffer containing 0.5% Triton X-100 and 5 mM EDTA and finally washed with phosphate-buffered saline.

Purification of pro(−)CA-NC capsids for negative staining.

Bacterial cells were resuspended in lysis buffer A (50 mM Tris-HCl [pH 8.0], 0.5 M NaCl, 1 mM EDTA, 0.5% mercaptoethanol, 1 μM ZnCl₂), disrupted with lysozyme, sonicated, and incubated with sodium deoxycholate (final concentration, 0.1%). After centrifugation at 10,000 × g (Beckman JA-18 rotor), proteins were precipitated from the supernatant by the addition of ammonium sulfate to 20% saturation, resuspended in lysis buffer A, and dialyzed against a buffer containing 100 mM Tris (pH 8.0), 10% glycerol, 0.1% mercaptoethanol, and 1 M NaCl. The material was pelleted by centrifugation in a microcentrifuge, resuspended in the same buffer, and used for electron microscopy.

Electron microscopy.

Bacterial pellets of the induced or uninduced samples (see above) were fixed in 2.5% glutaraldehyde in phosphate-buffered saline and processed for conventional embedding as described previously (20). For gold immunolabeling, pellets of cells were fixed in 3% paraformaldehyde for 2 days, embedded in agarose, and cut into small lumps. These were infiltrated with 2.3 M sucrose overnight, rapidly frozen in liquid ethane, transferred to methanol-uranyl acetate, and freeze substituted as described previously (16). A CS-Auto low-temperature unit (Reichert-Jung, Vienna, Austria) was used for Lowicryl HM20 embedding and UV polymerization. Thin sections were cut with Ultracut E (Reichert-Jung) and poststained for 6 to 8 min on alcohol-uranyl acetate. Polyclonal goat antibody against the p27^gag domain and rabbit anti-goat antibody–5-nm colloidal gold were used for gold immunolabeling of thin sections. Preimmune goat serum served as a control.

Negative staining of material from inclusion bodies (mainly capsids) was carried out with 4% silicotungstate at pH 8.2. Measurements were made from negatives at a magnification of ×35,000 or ×45,000 using Global Lab Image software (Data Translation, Wokingham, United Kingdom). A Philips CM12 electron microscope operated at 80 or 100 kV was used throughout this work.

RESULTS

We have reported previously that the M-PMV Gag polyprotein assembles in E. coli and in vitro into capsid-like structures that are similar to those assembled in mammalian cells (22). To investigate the localization of assembly domains required for M-PMV capsid formation in this prokaryotic overexpression system, we generated a series of truncated gag genes that encoded both C- and N-terminal deletions (Table 1). Each construct was expressed in E. coli BL21 cells; 4 h after induction, the formation of capsid-like structures was assessed by electron microscopy of the cells. The term “capsid” is used here for a spherical shell formed by Gag-derived proteins. This definition includes spherical structures of the same size assembled from CA-containing proteins. All of the structures assembled from the proteins engineered in this study can be considered to be immature capsids or procapsids, as they do not undergo the process of maturation, i.e., proteolytic cleavage of the precursors and rearrangement of the mature proteins to form an electron-dense core structure.

TABLE 1.

Constructs used in this study

Construct	Composition	Morphology	Inclusion bodies
Gag	MA-PP-p12-CA-NC-p4	Spherical	Yes
GagΔp4	MA-PP-p12-CA-NC	Spherical	Yes
GagΔNC-p4	MA-PP-p12-CA	None	Yes
MA-PP-p12	MA-PP-p12	None	Yes
MA-PP	MA-PP	None	No
MA	MA	None	No
GagΔMA	PP-p12-CA-NC-p4	Spherical	Yes
GagΔMA-PP	p12-CA-NC-p4	Spherical	Yes
p12-CA-NC	p12-CA-NC	Spherical	Yes
CA-NC-p4	CA-NC-p4	Sheets	No
CA-NC	CA-NC	Sheets	No
CA	CA	None	Yes
pro(−)CA	(P300Δ)CA	Polymers?	Yes
MA-CA-NC	MA-CA-NC	Spherical	Yes
pro(−)CA-NC	(P300Δ)CA-NC	Spherical	Yes
(P300A)CA-NC	(P300A)CA-NC	Spherical	Yes
KDIF-CA-NC	KDIF-CA-NC	Spherical	Yes
F-CA-NC	F-CA-NC	Spherical	Yes

C-terminal deletions demonstrate the importance of NC in the formation of higher-order structures.

In order to examine the roles of the different domains C terminal of MA, we constructed a panel of truncated gag genes in which each domain (p4, p14 [NC], p27 [CA], p12, and PP) was progressively removed. Each construct was expressed following induction in E. coli BL21. Control cells showed no evidence of inclusion bodies or capsid-like structures (data not shown), but as we have shown previously, the expression of full-length M-PMV Gag resulted in the formation of inclusion bodies which contained assembled procapsids (Fig. 1A). Deletion of the C-terminal protein p4 (GagΔp4) did not have any detectable effect on the capability of the polyprotein to assemble. The majority of particles assembled in E. coli from the GagΔp4 precursor were morphologically indistinguishable from those assembled from the wild-type Gag polyprotein (compare Fig. 2A and B). However, a small portion of particles showed a higher level of irregularity. The 36-amino-acid p4 protein contains eight prolines and resembles proline-rich proteins of other retroviruses (e.g., p6 in HIV-1) that play a role late in the budding process and facilitate the “pinching off” and release of the virus particle from the host cell membrane (21).

FIG. 1.

Electron micrographs showing thin sections of E. coli cells expressing M-PMV Gag and its N-terminal deletion forms. (A) Gag. (B) GagΔMA. (C) GagΔMA-PP. (D) CA-NC-p4. Bar, 100 nm. In panel D, the white arrow indicates a single sheet of CA-NC-p4 and the black arrow indicates the ends of bilamellar structures, in which single sheets splay out parallel to the membrane.

FIG. 2.

Electron micrographs showing thin sections of the capsids released from E. coli expressing M-PMV Gag and its deletion forms. (A) Gag. (B) GagΔp4. (C) GagΔMA. (D) GagΔMA-PP. (E) MA-CA-NC. Bar, 100 nm. Insets show individual capsids at magnifications of ×2 relative to the panels.

In contrast to the results obtained with GagΔp4, deletion of both NC and p4 (GagΔNC-p4) completely abrogated the formation of capsid-like structures in E. coli. The truncated polyprotein continued to accumulate in the cytoplasm in the form of inclusion bodies, but no evidence of macromolecular structures was observed (data not shown). Thus, the NC domain of the M-PMV Gag precursor protein appears to play a key role in the assembly of capsids.

Some of the other truncated forms of the Gag polyprotein (MA-PP and MA) did not form any defined, visible structures within the bacterial cells and were present as soluble proteins. In contrast, MA-PP-p12 (and GagΔNC-p4) formed large intracytoplasmic inclusion bodies comprised only of amorphous material (data not shown). This result suggests that the p12 domain might be responsible in part for the accumulation of the precursors in insoluble inclusions, consistent with its recently described internal scaffolding role, which appears to increase the efficiency of Gag-Gag interactions (30).

N-terminal deletions highlight an important role for sequences N terminal of CA.

As with the study of the C-terminal domains, we designed a panel of truncated Gag precursors in which the following domains were progressively removed: MA, PP, p12, and p27 (CA). To create a form of Gag with a deletion of MA (GagΔMA), the C-terminal methionine codon of MA was retained at the beginning of the phosphoprotein domain. Surprisingly, the lack of the N-terminal MA domain had no effect on the capability of Gag precursors to assemble into capsid-like structures in E. coli (Fig. 1B). This result is in contrast to the results of studies of mammalian cells, where sequences in MA located at the surface of the molecule are essential for the intracytoplasmic targeting of precursors to the assembly site and for capsid-membrane interactions (7, 8, 28). However, the efficiency of expression of GagΔMA was significantly lower than that of wild-type Gag; therefore, fewer capsid-like structures in the cell were observed (Fig. 1B). Interestingly, the GagΔMA capsids appeared to be less embedded than wild-type Gag in distinct inclusion bodies and showed clearly delineated margins (compare Fig. 1A and B), suggesting that the MA domain might contribute to the intercapsid material observed in inclusions with wild-type Gag.

A further truncation of N-terminal sequences that deleted both the MA and the PP domains of Gag (GagΔMA-PP) resulted in the loss of almost half of the molecular mass of the M-PMV Gag polyprotein but had little effect on capsid formation (Fig. 1C and 2D). Large accumulations of spherical and irregular capsids could be seen in inclusions of cells expressing the protein, although again the amorphous intercapsid material seen with wild-type Gag was missing in these aggregations. The lack of this material might result in some distortion of the capsid structures as they accumulate in the bacteria, since released capsids appeared to be primarily spherical (Fig. 2D).

In contrast to the first two truncations, the additional deletion of the p12-encoding sequence resulted in a dramatic change in the morphology of the assembled macromolecular structures in the E. coli cytoplasm. Instead of spherical capsids, long bilamellar structures could be seen spanning the width of the E. coli cells (Fig. 1D). While these structures initially appeared to be tubes, closer inspection revealed that they were most probably sheets of assembled CA-NC-p4 molecules that had associated into paired structures. This conclusion comes from the lack of any evidence of circular elements that would be expected from cross-sectional cuts through tubes, the occasional observation of a single sheet of CA-NC-p4, and the appearance of the ends of the bilamellar structures, in which single sheets splayed out parallel to the membrane. Thus, for M-PMV, deletion of sequences upstream of CA-NC-p4 results in the assembly within the cytoplasm of planar sheets rather than spherical capsids or tubes. The same structures were observed following the expression of CA-NC, in which p4 was additionally deleted from the C terminus.

Expression of the CA protein alone did not result in the formation of any electron microscopically observable structures in the inclusion bodies of induced E. coli (see Fig. 3E).

FIG. 3.

Electron micrographs of thin sections of E. coli showing the effect of N-terminal modifications of M-PMV CA-NC fusion proteins on the morphology of assembled structures. (A) p12-CA-NC. (B) MA-CA-NC. (C) CA-NC. (D) pro(−)CA-NC. (E) CA. (F) pro(−)CA. Bar, 100 nm. The arrow in panel E indicates the absence of structures in the inclusion bodies of induced E. coli. The arrow in panel F indicates inclusion bodies with the appearance of skeins of wool within which the pro(−)CA molecule formed balls of thread-like structures.

N- and C-terminal deletions result in the assembly of capsids with thinner shells.

A preliminary analysis of the electron micrographs of the capsids assembled from the different truncation mutants in the bacterial cells suggested that while the spherical capsids had similar diameters, the thickness of the shell varied proportionally with the molecular mass of the precursor. To investigate this finding further, intact capsids were released from washed inclusions as described in Materials and Methods and then embedded and sectioned to determine the mean relative thickness of the capsid wall. Figure 2A to D show the capsids released from wild-type Gag, GagΔp4, GagΔMA, and GagΔMA-PP. As might be expected, truncations of MA and MA-PP from the N terminus resulted in progressively thinner shells (and a corresponding decrease in the overall diameter of the capsid). Surprisingly, the C-terminal truncation of p4 also affected both the thickness of the shell (15 ± 2 nm versus 19 ± 2 nm) and its diameter (87 ± 7 nm versus 97 ± 7 nm), suggesting that, despite its C-terminal location, it might play a role in defining the packing interactions of the precursors within the spherical shell.

A nonspecific N-terminal extension of CA-NC is sufficient to program the formation of spherical capsids.

Previous work on HIV-1 capsid assembly in vitro had shown that the addition of as few as four or five amino acids to the N terminus of CA could convert the assembly of CA-NC from tubes to spheres (18, 34). We therefore investigated whether specific sequences extending from the N terminus of CA were required for the shift in morphogenesis observed with M-PMV by substituting sequences unrelated to p12. For this purpose, an MA-CA-NC-encoding expression vector was constructed in which MA sequences precisely replaced those of p12. Induction of chimeric Gag expression resulted in the efficient formation of spherical capsids (Fig. 2E and 3B) similar in diameter and wall thickness to those assembled by p12-CA-NC (Fig. 3A) (82 ± 11 nm versus 83 ± 10 nm and 12 ± 1 nm versus 12 ± 1 nm, respectively). As might be predicted, the shells of both of these structures were significantly thinner that those of structures assembled from wild-type Gag (19 ± 2 nm), and these structures were correspondingly smaller in diameter (compare to the value for Gag, 97 ± 7 nm) (Fig. 2A). Thus, merely extending the N terminus of CA with nonhomologous amino acids is sufficient to induce the assembly of spherical capsids.

The N-terminal proline of CA plays a critical role in defining the structure assembled by CA-NC.

The work of Gitti et al. (14) suggested that following cleavage of the N terminus of CA from its adjacent sequences, HIV-1 CA undergoes a conformational change that allows the assembly of cylindrical versus spherical structures. Evidence was presented to suggest that this mechanism might be shared by other mammalian retroviruses. We have therefore explored whether the N-terminal proline of M-PMV CA plays a similar role in defining the spatial arrangement of CA-NC molecules during assembly and whether deletion of the proline could prevent the conformational switch. Four different CA-NC constructs were prepared in which the N-terminal proline was (i) deleted and replaced by the initiating methionine [pro(−)], (ii) replaced by alanine, or preceded by one (iii) or preceded by four (iv) N-terminally adjacent amino acids of p12. The posttranslational removal of the initiating methionine in the bacterial products was confirmed by amino acid sequencing (data not shown). Remarkably, pro(−)CA-NC molecules assembled in a dramatically different fashion from CA-NC molecules (Fig. 3C), forming spheres and cylinders (Fig. 3D) (diameter of spheres, 88 ± 7 nm) similar to those seen with the GagΔp4 precursor (diameter of spheres, 87 ± 7 nm). Each of the other CA-NC constructs, in which CA was extended by one or four amino acids or in which alanine was substituted for the N-terminal proline, showed the same phenotype (data not shown). The morphology of negatively stained pro(−)CA-NC spherical particles released from the bacterial cells is shown in Fig. 4. Thus, we conclude that the N-terminal proline of CA plays a crucial role in determining the nature of the interactions between the domains of the Gag precursor that are important for particle assembly.

FIG. 4.

Electron micrograph of negatively stained pro(−)CA-NC. The capsids were purified from E. coli as described in Materials and Methods. Bar, 100 nm.

Further evidence for the above conclusions comes from the expression of CA and pro(−)CA molecules in E. coli in the absence of NC. The expression of CA alone resulted in the formation of inclusion bodies that lacked any evidence of higher-order structures (Fig. 3E). In contrast, the expression of the pro(−)CA molecule yielded inclusion bodies with the appearance of skeins of wool within which the pro(−)CA molecule formed balls of thread-like structures (Fig. 3F). Thus, merely removing the proline from the N terminus of CA allowed it to undergo stable interactions with other CA molecules to yield what appeared to be polymeric structures. The identity of the material observed following the expression of pro(−)CA was confirmed by immunogold labeling of thin sections using a polyclonal goat antibody against p27^gag (data not shown).

DISCUSSION

We showed previously that M-PMV Gag assembles into capsid-like structures in bacteria (22, 29). In the present study, we have focused on the domains critical for the assembly of organized structures and have located a domain that determines the formation of spherical particles.

We show here that almost half of the Gag polyprotein precursor is dispensable for the assembly of immature capsid-like structures. Neither the simultaneous deletion of the PP, p12, and p4 domains nor the deletion of MA together with PP and p4 abrogated the assembly of spherical structures. In contrast, CA-NC alone had a high propensity to form sheets of protein within the bacterial cytoplasm. These observations indicate that there is no specific structural requirement for the domains upstream of CA for spherical particle assembly in bacteria. The bacterial high-level expression system is distinct from virus-infected cells, where MA directs the transport of Gag-related polyproteins to an intracytoplasmic assembly site and assembled capsids to the plasma membrane (26, 27, 28). The data presented here are in agreement with the finding that M-PMV MA is not required for the capsid assembly process per se but provides a targeting signal (26) that allows polyproteins to achieve a critical intracytoplasmic concentration required for assembly. Such a signal is clearly dispensable in bacteria, where expression yields sufficiently high cytoplasmic levels of proteins for efficient assembly. Similarly, the p12 domain, which is essential for efficient capsid assembly in virus-infected cells and in the in vitro translation-assembly system (30, 33), is not necessary for the efficient assembly of spherical capsids in the bacterial high-level expression system.

We hypothesize that the major role for sequences immediately N terminal of CA in the assembly process is to constrain the N-terminal proline residue of CA so that it cannot undergo a conformational rearrangement that determines whether CA assembles into sheets rather than spheres or cylinders. This hypothesis is supported by results obtained with both a deletion of the N-terminal proline and a substitution mutation in which the N-terminal proline of CA was replaced by alanine. Both of these changes resulted in the formation of spherical CA-NC particles. It has been shown previously that the free amino terminus of HIV-1 CA folds into a β-hairpin/helix structure that is probably stabilized by the formation of a salt bridge between the N-terminal proline and a conserved aspartate residue at position 51 (14). A similarly located aspartate residue is highly conserved in the CA sequence of retroviruses (34). The fact that in M-PMV an aspartate residue is located at position 50 of CA predicts a similar structural arrangement in this protein. We show here that this putative conformational change is not directed by any specific structural motif located upstream of CA and that this process is controlled solely by the availability of an N-terminal proline residue. These findings are consistent with the data of Gross et al. (18), who showed that the spherical shape of the HIV capsid is determined by an N-terminal extension of the CA domain and that the formation of the core shell requires the liberation of the CA N terminus. In summary, it can be concluded that the conformation of CA within the context of flanking sequences in the Gag precursor defines the spherical form of the capsid.

Using electron microscopy, we have not observed large changes in particle size with any of the deletion mutants reported here. However, small decreases in capsid diameter correlated with reductions in the size of the Gag precursor protein. Our comparison of the deletion mutants with wild-type Gag shows a reduction in capsid wall thickness, which was particularly significant for structures assembled from proteins with large deletions, such as GagΔMA-PP and MA-CA-NC. Garnier et al. (12, 13) demonstrated a role for HIV-1 protein p6 in controlling capsid size following the expression of Gag in COS-1 cells, although Gross et al. (19) recently observed properly sized particles assembled in vitro from an HIV-1 p6 deletion construct (ΔMA-CA-NC). Similarly, deletion of the corresponding C-terminal domain in M-PMV Gag, p4, resulted in only a small decrease in the diameter of the capsids.

The data presented here also demonstrate that the NC domain plays a critically important role in the process of M-PMV capsid assembly in bacteria, consistent with the data on the in vivo assembly of HIV-1 (9). However, the in vitro assembly of small or heterogeneously shaped and sized spherical particles from an HIV Gag fragment lacking NC also has been described (18, 34). The data elucidating the role of RNA in the process of capsid assembly have been controversial. The RNA binding capability of NC suggested a role for RNA in interprotein interactions between NC molecules either by concentrating the protein molecules or by serving as scaffolding for the formation of highly organized tubular structures (5, 11, 17). However, efficient assembly of CA-NC into conical cores in the absence of RNA was achieved in vitro by promoting the interactions with high salt concentrations (11). Direct interactions between NC molecules also were demonstrated by cross-linking studies (24). While we previously described the in vitro assembly of isolated M-PMV Gag into capsid-like structures without the addition of purified RNA (22), the data presented here show that the NC domain is normally required for the assembly of organized protein sheets by CA-NC. In contrast to the in vitro assembly of HIV-1 CA (10, 17), no structures were formed in E. coli by M-PMV CA alone. Thus, the fusion of CA to NC has a dramatic effect on the potential to assemble in E. coli, resulting in what appear to be unilamellar and bilamellar sheets. Preliminary experiments indicate that the function of NC in stimulating the formation of sheet-like structures can be provided by a short, 12-amino-acid C-terminal extension of CA that includes six amino acids from NC and six histidines (data not shown). Experiments are currently under way to determine whether this short sequence functions through binding RNA or through some other mechanism.

The studies described here show the utility of a bacterial assembly system to determine, under relative physiological conditions, the ability of Gag components to assemble into macromolecular structures and to define elements necessary for the capsid assembly process. This system, which is independent of eukaryotic intracellular transport mechanisms and the factors of the natural host cell, has allowed us to define domains critical for M-PMV capsid assembly and to locate unequivocally the amino acid residue responsible for the switch from spherical particles to planar structures.