Biochemical Reconstitution of HIV-1 Assembly and Maturation

Assembly and maturation are essential steps in the replication of orthoretroviruses such as HIV-1 and are proven therapeutic targets. These processes require the coordinated functioning of the viral Gag protein’s multiple biochemical activities. We describe here the development of an experimental system that allows an integrative analysis of how Gag’s multiple functionalities cooperate to generate a retrovirus particle. Our current studies help to illuminate how Gag synergizes the formation of the virus compartment with RNA binding and how these activities are modulated by the small molecule IP6. Further development and use of this system should lead to a more comprehensive understanding of the molecular mechanisms of HIV-1 assembly and maturation and may provide new insights for the development of antiretroviral drugs.

ABSTRACT

The assembly of an orthoretrovirus such as HIV-1 requires the coordinated functioning of multiple biochemical activities of the viral Gag protein. These activities include membrane targeting, lattice formation, packaging of the RNA genome, and recruitment of cellular cofactors that modulate assembly. In most previous studies, these Gag activities have been investigated individually, which provided somewhat limited insight into how they functionally integrate during the assembly process. Here, we report the development of a biochemical reconstitution system that allowed us to investigate how Gag lattice formation, RNA binding, and the assembly cofactor inositol hexakisphosphate (IP6) synergize to generate immature virus particles in vitro. The results identify an important rate-limiting step in assembly and reveal new insights into how RNA and IP6 promote immature Gag lattice formation. The immature virus-like particles can be converted into mature capsid-like particles by the simple addition of viral protease, suggesting that it is possible in principle to fully biochemically reconstitute the sequential processes of HIV-1 assembly and maturation from purified components.

IMPORTANCE Assembly and maturation are essential steps in the replication of orthoretroviruses such as HIV-1 and are proven therapeutic targets. These processes require the coordinated functioning of the viral Gag protein’s multiple biochemical activities. We describe here the development of an experimental system that allows an integrative analysis of how Gag’s multiple functionalities cooperate to generate a retrovirus particle. Our current studies help to illuminate how Gag synergizes the formation of the virus compartment with RNA binding and how these activities are modulated by the small molecule IP6. Further development and use of this system should lead to a more comprehensive understanding of the molecular mechanisms of HIV-1 assembly and maturation and may provide new insights for the development of antiretroviral drugs.

INTRODUCTION

The assembly, architecture, and maturation of orthoretroviruses such as HIV-1 are directed by a virally encoded polyprotein called Gag (recently reviewed in references 1 and 2). Several thousand copies of HIV-1 Gag assemble at the plasma membrane and bud to form an immature virion. Subsequent to budding, Gag is proteolytically cleaved into the mature structural proteins matrix (MA), capsid (CA), and nucleocapsid (NC) that rearrange to form mature infectious particles. The immature HIV-1 Gag protein contains additional peptide domains, including p6, which facilitates virus release from the cell, and two smaller spacer peptides, SP1 and SP2, that connect NC to the CA and p6 domains, respectively.

Studies have shown that the CA-SP1 regions of Gag play an important role in assembly. In particular, a 14-residue segment spanning the CA-SP1 boundary folds into an α-helix (the “junction helix”), which self-associates into a 6-helix bundle that holds together and harbors the bulk of hydrophobic contacts in the immature Gag hexamer (3, 4). Similar 6-helix bundles or functionally equivalent regions are observed in other orthoretroviruses (5–8). Consistent with its key role in HIV-1 Gag lattice formation, mutations in the CA-SP1 boundary abrogate or significantly impair immature HIV-1 assembly, both in vitro and in cells (9–15). Furthermore, inositol hexakisphosphate (IP6), a small molecule that is abundant in mammalian cells, induces immature HIV-1 Gag assembly in vitro by binding just above the 6-helix bundle and stabilizing its folded conformation (16).

In addition to serving as the protein scaffold for virus assembly, Gag plays a critical role in the selective packaging of the viral genomic RNA (gRNA) (reviewed in references 17 and 18). Orthoretrovirus particles contain two copies of the gRNA, which is a full-length, 5′-capped mRNA transcript of the chromosomally integrated provirus. The gRNA directly binds the Gag NC domain, which consists of one or two zinc knuckles that can bind promiscuously to nucleic acids and with specificity to cognate RNA molecules. Current models posit that gRNA packaging is coupled to Gag assembly at an early stage, and it is proposed that retrovirus assembly is initiated by an oligomeric complex containing the gRNA (which may or may not be dimeric at this step) and several molecules of Gag (17–25). This Gag/RNA complex is thought to nucleate virus assembly by forming a “seed” that more and more Gag molecules add to, eventually resulting in the formation of a spherical Gag lattice (Fig. 1A). The architecture of this putative Gag/gRNA seed is unknown but at a minimum must include high-affinity and specific modes of interaction between the gRNA 5′ leader that contains the Ψ packaging signal and bound Gag NC domains. Consistent with the above-described seeding mechanism, previous in vitro studies have also invoked nucleic acid-induced Gag clustering as an important mechanism of initiating immature lattice assembly (18, 26). Indeed, it was previously thought that Gag assembly is completely RNA dependent (or at least nucleic acid dependent). However, we and others have shown that HIV-1 Gag can assemble in vitro in the absence of nucleic acid templates and that, in fact, assembly in the absence of nucleic acid can occur at a much higher efficiency than in its presence (4, 16, 27). While it appears almost certain that gRNA recruitment and Gag assembly are coupled, the mechanistic details of this coupling and how it is modulated by assembly factors like IP6 remain unclear.

FIG 1.

HIV-1 assembly mechanism and assay design. (A) HIV-1 assembly initiates when a few Gag molecules bind to the 5′ leader (Ψ) in the viral genomic RNA. This oligomeric ribonucleoprotein complex nucleates or seeds the assembly of the virus particle. During the propagation phase of assembly, more and more Gag molecules are added until assembly is complete. (B) Experimental protocol to biochemically reconstitute HIV-1 assembly in vitro. Purified Gag^ΔMA protein in bulk solution (50 μM in PI buffer) remains soluble for around 3 h. The assembly of virus-like particles (VLPs) is initiated by the addition of naked RNA or preformed “seed” Gag^ΔMA/RNA complexes. Assembly is monitored during incubation by simultaneously using light scattering and negative-stain EM. (C) Schematic of HIV-1 Gag, with the major domains (MA, CA, NC, and p6) and spacer peptides (SP1 and SP2) labeled. The Gag^ΔMA construct is diagrammed below. (D) SDS-PAGE profiles of purified Gag^ΔMA proteins used in this study. Three micrograms of each protein was loaded into each lane and visualized by Coomassie staining. (E) Secondary-structure model of the HIV-1 dimeric 5′ leader (Ψ) with major structural elements labeled. The core encapsidation signal (Ψ^CES) is shown in blue, which has the TAR, poly(A), and primer binding site (PBS) regions deleted. A GAGA tetraloop used to replace the PBS in Ψ^CES is indicated by red dotted lines. (F) Agarose gel electrophoresis illustrating dimerization of Ψ and Ψ^CES when incubated in PI buffer. M, monomer; D, dimer.

In vitro assembly model systems have been a critical tool of many studies that led to important advances in our understanding of HIV-1 assembly and morphogenesis (26). A recombinant HIV-1 Gag construct called ΔMA-CA-NC (missing residues 16 to 100 from the MA domain and truncated at NC; referred to here as Gag^ΔMA) has been particularly informative because it displays binary assembly behavior in vitro, forming both mature capsid-like particles and immature virus-like particles (VLPs) (28). This binary behavior is interpreted to reflect the conformation of the CA-SP1 junction during assembly: when the junction folds and associates into the 6-helix bundle, Gag^ΔMA forms immature VLPs, and when the junction helix fails to fold, mature particles are produced. In this study, we developed a biochemical reconstitution system composed of purified HIV-1 Gag^ΔMA protein and an HIV-1 gRNA 5′ leader to generate immature VLPs whose assembly kinetics and assembly products can be monitored in real time. Using this system, we examine how the CA-SP1 junction 6-helix bundle, RNA binding, and the assembly factor IP6 are functionally integrated during HIV-1 assembly in vitro. Our results formally support the hypothesis that folding of the CA-SP1 junction 6-helix bundle is a rate-limiting step of HIV-1 Gag assembly and reveal new insights into how gRNA and IP6 promote immature lattice formation.

RESULTS

Assay design and initial optimization.

Our assembly assay was conducted with purified Gag^ΔMA protein (Fig. 1C and D) (50 μM) in bulk solution in “PI buffer” [20 mM Tris (pH 7.5), 140 mM KCl, 10 mM NaCl, 5 mM MgCl₂, 10 mM Tris(2-carboxyethyl)phosphine hydrochloride (TCEP)]. Assembly was initiated by adding test nucleants, such as naked RNA or preformed protein/RNA seed complexes (Fig. 1B). We monitored the progress of VLP assembly in real time by using light scattering and negative-stain electron microscopy (EM). Thus, our assay simultaneously monitored both the kinetics of assembly and the assembly products.

We first tested the effect of the protein purification method on the assembly properties of Gag^ΔMA, as the purification buffers in the original protocol contained the chelator EDTA, which allowed the efficient removal of contaminating nucleic acids but also would have stripped off bound zinc from the NC domain (28). Despite the potential to unfold NC, this treatment was nevertheless shown to yield protein that assembles into immature VLPs, either by overnight dialysis with short oligonucleotides (28) or by incubation with tartrate salts (4, 27). To confirm that Gag^ΔMA purified in the presence of EDTA remains assembly competent, we induced assembly by adding tartrate (375 mM final concentration [4]) to bulk Gag^ΔMA in PI buffer (Fig. 2A, black curve). This resulted in a time-dependent increase in the light scattering signal at 320 nm, which is indicative of the formation of large particles. Negative-stain EM images of aliquots from the same sample revealed that the particles consisted of spherical VLPs and that the numbers of VLPs increased in correspondence with the magnitude of the light scattering signal (Fig. 2B). Thus, the correct assembly products were being produced. However, when we tested various types of RNA, we found that they produced similar light scattering profiles (Fig. 2A, compare colored curves to the black control curve). These results indicated that the assembly behavior of Gag^ΔMA protein purified in the presence of EDTA is relatively insensitive to RNA; the protein was therefore not suited for the purposes of this study.

FIG 2.

Effect of the protein purification method and RNA on Gag^ΔMA assembly in vitro. (A) Light scattering profiles of Gag^ΔMA purified in the presence of 1 mM EDTA (28), as a function of the incubation time after the addition of tartrate (375 mM final concentration) (4), in the absence (black curve) or presence (colored curves) of the indicated RNAs. The gray curve shows the profile of a negative-control sample in which only buffer was added. (B) Negative-stain EM images of assembly products from panel A, sampled at the indicated time points. An example of an immature virus-like particle (VLP) is indicated (“imm.”) in each image. Bars, 100 nm. (C) Light scattering profiles of Gag^ΔMA purified without EDTA (29). (D) Assembly products from panel C, sampled at the indicated time points. Examples of mature capsid-like particles (“mat.”) and immature VLPs (“imm.”) are indicated in each image. Bars, 100 nm. (E and F) Zoomed-out views of the assembly products induced by the addition of the indicated RNAs, sampled at the indicated time points. Colored boxes indicate the regions corresponding to images in panel D. Bars, 500 nm. (G) Assembly reaction mixtures in the presence of a 50-nt ssRNA oligonucleotide (2-h time point of the brown curve in panel C) were fractionated by centrifugation, and the amounts of Gag^ΔMA protein in the pellet and supernatant fractions were visualized by using SDS-PAGE and Coomassie staining. Data are representative of results from experiments performed using two independent protein preparations, with two replicates per preparation.

We then purified Gag^ΔMA using a different protocol that preserves the zinc knuckles (29). In stark contrast to protein purified with EDTA, the addition of 375 mM tartrate did not induce the assembly of Gag^ΔMA, and the light scattering profile remained flat at baseline for more than 2 h (Fig. 2C, black curve). Importantly, however, this protein was now quite sensitive to added RNA. The addition of single-stranded RNA (ssRNA) oligonucleotides of 25 or 50 nucleotides (nt) in length just after the addition of tartrate induced robust assembly, as evidenced by sigmoidal transitions in the light scattering profiles (Fig. 2C). A lag time of ∼2 h was observed for the 25-nt oligonucleotide, versus ∼1 h for the 50-nt oligonucleotide (Fig. 2C, compare red and brown curves). Examination of the assembly products at late time points revealed that the assembled VLPs were predominantly spherical and immature in appearance (Fig. 2D, top). Thus, under these experimental conditions, ssRNA oligonucleotides can induce the assembly of immature Gag^ΔMA VLPs with kinetics dependent on the length of the RNA, provided that EDTA is omitted from the purification procedure and that, by extension, the NC domain is properly folded. We also examined the yield of particles and found that >95% of the bulk Gag^ΔMA proteins are incorporated into pelletable VLPs (Fig. 2G). All experiments described below were therefore performed with protein purified using the protocol described previously by Ganser et al. (29).

HIV-1 Ψ and large RNAs induce two assembly pathways of Gag^ΔMA in vitro.

We next tested the effect of the HIV-1 5′ leader sequence on the assembly of Gag^ΔMA by using a previously described 159-nt RNA construct, called Ψ^CES, which contains the minimal core encapsidation signal of HIV-1 (Fig. 1E) (30, 31). Previous studies have shown that this RNA adopts a well-defined dimer structure in PI buffer (Fig. 1F) and can direct the specific packaging of test RNAs into virus particles that bud from cells (30–32). In stark contrast to the simple RNAs, Ψ^CES induced assembly with essentially no lag phase (Fig. 2C, dark blue curve). Unexpectedly, however, the Ψ^CES curves plateaued at lower absorbance values, and, more importantly, negative-stain EM imaging revealed that the vast majority of particles were in fact not immature VLPs but rather long, narrow tubes (Fig. 2D, bottom, and Fig. 2F), which are the mature capsid-like particles formed by Gag^ΔMA (4, 28). Immature VLPs were observed only at later time points, and these were relatively few in number. We also tested other large RNAs, such as the full dimeric 5′ leader [here termed Ψ, and which contains the primer binding site (PBS) and the TAR and poly(A) hairpins that are missing from Ψ^CES (Fig. 1E)] as well as purified Escherichia coli rRNA. Both of these RNA molecules induced the same Gag^ΔMA assembly behavior as Ψ^CES (Fig. 2C, compare dark blue, cyan, and green curves). Overall, these results demonstrate that under the above-described experimental conditions, two independent assembly pathways occur: one with essentially no lag phase that produces mature capsid-like particles and a second with a longer lag phase that produces immature VLPs. Given these results, we reexamined more closely the assembly products produced by ssRNA. As with the larger RNAs, mature-like assemblies could also be observed at early time points with 50-nt ssRNA (Fig. 2D, top, and Fig. 2E); however, the tubes were much shorter, and the final particle distribution was heavily skewed toward the late-assembling immature VLPs in this case.

The binary assembly behavior of Gag^ΔMA has been described previously and is most parsimoniously explained by the CA-SP1 junction 6-helix bundle being unfolded in the mature tubes and folded in the immature VLPs (28). Indeed, we previously showed that structure-based mutations designed to disrupt the integrity of the 6-helix bundle favored the assembly of Gag^ΔMA into mature capsid-like particles rather than immature VLPs (4). In that previous study, we also observed an all-or-none in vitro assembly behavior of the Gag^ΔMA CA-SP1 mutants, which we speculated might be due to failure at nucleation (4). Here, we tested this notion by preincubating Gag^ΔMA with RNA in a seed mixture, which we surmised might both enhance the yield and accelerate the kinetics of immature VLP assembly once added to Gag^ΔMA in bulk solution. Such an experimental setup was also suggested by observations that in cells, new virions initially appear as oligomeric complexes of the gRNA and several Gag molecules (Fig. 1A) (22, 25). We thus preincubated Gag^ΔMA and the Ψ^CES dimer at a 6:2 molar ratio (1 Gag hexamer per RNA dimer), which was the highest proportion of Gag to RNA tested that formed only oligomeric complexes; the addition of more excess Gag to the RNA began to induce immature lattice assembly. We then seeded the bulk Gag^ΔMA with the Gag^ΔMA/Ψ^CES oligomers. In this experimental format, the resulting light scattering curve still showed an initial, rapid rise with a low plateau but now had an additional, higher transition at ∼2 h (Fig. 3A). Negative-stain EM confirmed that this later transition corresponded to the formation of immature VLPs, which were now much more numerous than with seeding with RNA alone (Fig. 3B). As observed in reaction mixtures seeded with naked RNA, the early transition corresponded to the assembly of mature capsid-like particles.

FIG 3.

Seeding of Gag^ΔMA assembly with preformed Gag^ΔMA/Ψ complexes. (A, C, E, and G) Light scattering profiles of reaction mixtures in which Gag^ΔMA in bulk solution is seeded with the indicated complexes containing Gag^ΔMA/Ψ (6 Gag molecules to 2 RNA molecules or 1 RNA dimer). When IP6 is present, it is used at the indicated molar ratios with Gag^ΔMA. (B, D, F, and H) Assembly products sampled at the indicated time points. Individual examples are indicated and labeled in images that contain both mature capsid-like particles (“mat.”) and immature VLPs (“imm.”). Bars, 100 nm. Experiments in panels A, B, E, and F were performed with three independent protein preparations and three times for each preparation; experiments in panels C, D, G, and H were performed with two preparations and two times each.

Nonspecific interactions between bulk Gag and RNA lead to rapid nucleation of mature capsid-like assemblies.

In comparing the light scattering curves, it was apparent that the principal effect of seeding with Gag^ΔMA/Ψ^CES oligomers was to enhance the yield of immature VLPs; these still assembled with a significant lag time, and so the Gag^ΔMA/Ψ^CES seed had no apparent effect on assembly kinetics compared to naked Ψ^CES. Neither the yield nor the kinetics of the mature assembly pathway was significantly affected (although the capsid-like tubes seemed shorter in the seeded reaction mixtures). We therefore surmised that the mature assemblies were arising from interactions between the bulk Gag proteins and the added RNA. To test this, we used as bulk Gag a mutant construct called Gag^ΔMANC11 that lacks the zinc binding knuckles but retains the first 11 residues of NC. This construct is not capable of high-affinity binding to RNA but should retain the ability to bind weakly and nonspecifically (33). As shown in Fig. 3C, seeding bulk Gag^ΔMANC11 with Gag^ΔMA/Ψ^CES oligomers still induced the robust assembly of immature VLPs. Compared to the experiments using Gag^ΔMA to propagate assembly, the light scattering curve was now more cleanly sigmoidal, indicating that the mature assembly pathway was suppressed (compare black curves in Fig. 3A and C). This was confirmed by negative-stain EM: the vast majority of the VLPs observed had the immature phenotype, and mature capsid-like particles were observed only rarely, if at all (Fig. 3D).

Taken together, the above-described data demonstrate that the mature assembly pathway is an in vitro phenomenon that arises from the ability of the bulk Gag molecules to bind with high affinity to nucleic acids. Once the obscuring mature pathway has been taken into account, Ψ RNA induces Gag^ΔMA assembly with kinetics similar to those of simpler RNAs. This finding underscores the importance of monitoring both assembly kinetics and products with the same sample. Our results also suggest that high-affinity binding of the NC domain to RNA is required only during the early steps of immature assembly and becomes dispensable once the particle has started to form. Although NC binding to gRNA sites outside the Ψ packaging signal very likely contributes to the overall assembly efficiency of authentic virions, it appears that the critical function of NC during assembly is to recruit the genome.

IP6 dramatically accelerates the kinetics of RNA-induced Gag^ΔMA assembly.

The above-described results support the model that an initial oligomer composed of a few Gag protein molecules in complex with gRNA nucleates the assembly of the immature HIV-1 virion. However, the significant lag time indicates the presence of a rate-limiting step in assembly, which we hypothesized to be slow folding of the CA-SP1 junction into the 6-helix bundle that forms the hydrophobic core of the Gag hexamer. Since IP6 binds and stabilizes the 6-helix bundle (16), we predicted that the inclusion of IP6 in the seed mixtures would promote the folding of the 6-helix bundle and, hence, eliminate the lag time of assembly. Indeed, when seeds were made in the presence of stoichiometric levels of IP6 relative to Gag^ΔMA and Ψ^CES, the lag time of assembly was eliminated (Fig. 3E, black curves). This enhancement is due to the combined effects of the RNA and IP6, since omitting Ψ^CES from the seed mixture recovered the sigmoidal assembly curve with an ∼1-h lag (Fig. 3E, red curve). Furthermore, the presence of IP6 virtually eliminated the mature assembly pathway even when wild-type (WT) Gag^ΔMA was used as bulk Gag because only immature VLPs were observed (Fig. 3F). Similar results were obtained when Gag^ΔMANC11 was used as bulk Gag (Fig. 3G and H).

Structural requirements of in vitro Gag assembly.

Our seeding assembly protocol provides a convenient assay to separately analyze the structural and functional requirements of the nucleating and propagating Gag molecules. To do this, we purified 11 Gag^ΔMA proteins (Fig. 1D) that contained alanine substitutions for residues within the protein-protein interfaces that hold together the immature lattice and/or were previously shown to be important for HIV-1 assembly and infectivity (3, 4, 16, 34) (Table 1). As also observed in a previous study (4), all the mutant Gag^ΔMA constructs were biochemically well behaved, and we found no significant variations in expression levels, protein behavior during purification, solubility, or yield. We performed complementary experiments, testing the ability of the mutant to function as a nucleator when preincubated with Ψ^CES and then seeded into bulk WT Gag^ΔMA or as a propagator in bulk solution when seeded with the WT Gag^ΔMA/Ψ^CES complex. Unfortunately, some of the mutants displayed wide variations in their inherent propensity to spontaneously aggregate in assembly buffer (in the absence of seeding); this precluded comparative analysis by light scattering, and thus, we monitored the assembly products only by using negative-stain EM. The results are summarized in Table 1 and Fig. 4, and representative data are shown in Fig. 5 and 6.

TABLE 1.

In vitro assembly phenotypes of Gag^ΔMA mutants^a

Mutation	Location	Phenotype
		Control reaction		Nucleation		Propagation
		No IP6	+IP6	No IP6	+IP6	No IP6	+IP6
None (WT)		i>m	imm.	i>m	imm.	i>m	imm.
E160A/E161A/K162A	NTD	mat.	i<m	i<m	i>m	mat.	imm.
E207A/E208A	NTD	mat.	i>m	i<m	i>m	mat.	i<m
R232A/S234A	NTD	n.a.	imm.	i<m	imm.	mat.	imm.
T242A/Q244A	NTD	n.a.	imm.	mat.	imm.	n.a.	imm.
R275A	NTD	mat.	i<m	mat.	mat.	mat.	mat.
W316A/M317A	CTD	n.a.	n.a.	i>m	imm.	n.a.	n.a.
K290A	CTD	mat.	mat.	i∼m	i∼m	i∼m	i∼m
V353A	β-Turn	mat.	mat.	i>m	imm.	n.a.	i<m
L363A	6HB	mat.	mat.	mat.	i<m	n.a.	i<m
M367A	6HB	mat.	mat.	i<m	i<m	n.a.	i<m
T371A	6HB	i<m	imm.	i<m	i>m	n.a.	imm.

^a6HB, 6-helix bundle; imm., all immature or very few mature; i>m, mostly immature; m>i, mostly mature; mat., all mature or very few immature; n.a., no assembly or only aggregates.

FIG 4.

Assembly behavior of Gag^ΔMA mutants. Residues selected for alanine substitution are mapped onto the immature Gag hexamer structure (3) and colored according to their assembly phenotypes (Table 1) when seeded with naked Ψ (A and B), when used as a nucleator (C and D), and when used as a propagator (E and F). Experiments were performed in the absence (A, C, and E) or presence (B, D, and F) of IP6.

FIG 5.

Alanine-scanning mutagenesis of the CTD and SP1 regions of Gag^ΔMA. Shown are negative-stain EM images that illustrate the assembly products of the indicated mutants. In each panel, the assembly phenotype is indicated at the bottom left corner (imm., essentially all particles observed are immature VLPs; mat., essentially all particles observed are mature tubes; i>m, most assembly products are immature VLPs; i<m, most assembly products are mature tubes; n.a., no assembly products or only aggregates are observed). Individual examples are indicated and labeled in images that contain both mature and immature particles. Bars, 200 nm. (A and B) Assembly phenotypes of the indicated mutants upon the direct addition of Ψ RNA, either without (A) or with (B) IP6. (C and D) Nucleation experiments in which mutant Gag^ΔMA/Ψ complexes without (C) or with (D) IP6 are added to bulk WT Gag^ΔMA. (E and F) Propagation experiments in which WT Gag^ΔMA/Ψ complexes without (E) or with (F) IP6 are added to bulk mutant Gag^ΔMA. Experiments were performed with two independent protein preparations for each mutant and three to four times for each preparation. Boxes are colored according to the coloring in Fig. 4.

FIG 6.

Alanine-scanning mutagenesis of the NTD of Gag^ΔMA. Shown are negative-stain EM images that illustrate the assembly products of the indicated mutants. In each panel, the assembly phenotype is indicated in the bottom left corner, as described in the legend of Fig. 5. Individual examples are indicated and labeled in images that contain both mature and immature particles. Bars, 200 nm. (A and B) Assembly phenotypes of the indicated mutants upon the direct addition of Ψ RNA, either without (A) or with (B) IP6. (C and D) Nucleation experiments in which mutant Gag^ΔMA/Ψ complexes without (C) or with (D) IP6 are added to bulk WT Gag^ΔMA. (E and F) Propagation experiments in which WT Gag^ΔMA/Ψ complexes without (E) or with (F) IP6 are added to bulk mutant Gag^ΔMA. Experiments were performed with two independent protein preparations for each mutant and three to four times for each preparation. Boxes are colored according to the coloring in Fig. 4.

We tested six sets of alanine substitutions in the CA C-terminal domain (CTD) and SP1 regions, which generate the core Gag hexamer and the “minimal” immature lattice (Fig. 4 and Table 1): W316A/M317A, which abrogates the CTD dimerization interface that connects adjacent Gag hexamers; K290A, which is located in the major homology region loop and abrogates the binding of IP6 to the center of the Gag hexamer; V353A, which is located in a β-turn element that connects the top of the junction helix bundle to the bottom of the CTD hexamer; L363A and M367A, which are located in the junction helix itself and would disrupt “knobs-in-holes” packing of the 6-helix bundle; and T371A, which is the first disordered residue that immediately follows the junction helix, as seen in the crystal and cryoEM structures (3, 4, 16, 34). In control assembly experiments, all mutants except T371A were incapable of independently forming immature VLPs when induced with Ψ^CES and/or IP6 (Fig. 4A and B and Fig. 5A and B); T371A formed VLPs but at an appreciably lower yield than the WT control. These results are all consistent with the results of a previous mutagenesis study (4). Thus, the experiments described below serve to delineate whether the underlying defects are at the nucleation or propagation steps of assembly.

In the nucleation experiments (adding mutant Gag^ΔMA/Ψ^CES complexes to bulk WT Gag^ΔMA), the L363A and M367A mutants were the most significantly impaired, consistent with an important role for the 6-helix bundle in initiating assembly (Fig. 4C and Fig. 5C). T371A was also impaired but not to the same extent as L363A or M367A. Somewhat surprisingly, both the K290A and V353A mutants were competent as nucleators, even though Gag^ΔMA harboring either of these mutations is incapable of independent assembly into immature VLPs in vitro (4). Likewise, the W316A/M317A mutant was also competent as a nucleator despite the important role of the CTD-CTD dimer interface in extending the immature Gag lattice (3, 4). These results suggest that the known immature Gag hexamer structure may not precisely describe all the functional features of the nucleating Gag oligomers.

In the propagation experiments (adding WT Gag^ΔMA/Ψ^CES complexes to bulk mutant Gag^ΔMA), the assembly behavior of the mutants correlated more cleanly with the immature Gag lattice structure (Fig. 4E and Fig. 5E). W316A/M317A did not support assembly, consistent with the critical role of the CTD dimer in connecting adjacent Gag hexamers (3, 4). Likewise, the β-turn and junction helix mutants were also impaired in propagating assembly, yielding only aggregates. We expected, however, that the latter mutants should have formed mature particles, because under these conditions, WT Gag^ΔMA propagated both mature and immature particles (Fig. 3B). Since we have established that the competing mature pathway arises from Gag clustering promoted by nonspecific binding of the bulk Gag molecules with the RNA, one interpretation of these results is that the junction helix mutations somehow interfere with these interactions in vitro. However, we refrain from ascribing any biological significance to this observation, as there is no evidence that the assembly of the mature capsid in actual virions would involve fragments of Gag larger than CA.

We also tested the effect of IP6. In both the nucleation (Fig. 5D) and propagation (Fig. 5F) formats, IP6 generally mitigated the deleterious effects of the mutations (compare Fig. 4C to Fig. 4D and likewise Fig. 4E to Fig. 4F), results that are consistent with IP6-induced stabilization of the 6-helix bundle. An exception to this was the K290A mutant, which is impaired in binding IP6 and therefore showed the same phenotypes in its presence and absence. W316A/M317A was also insensitive to IP6 in propagation experiments (Fig. 5F and compare Fig. 4E to Fig. 4F), underscoring the importance of the CTD dimer interface in extending the Gag lattice and also demonstrating that its requirement for assembly is independent of the 6-helix bundle. Among the junction helix mutants (L363A, M367A, and T371A), the recovery of the ability to propagate assembly had a pronounced gradation toward the C terminus, with L363A being the least and T371A being the most responsive to the added IP6 (Fig. 5F and compare Fig. 4E to Fig. 4F).

We tested five sets of alanine substitutions designed to disrupt protein-protein interfaces involving the CA N-terminal domain (NTD), as observed in the cryoEM structure of the immature HIV-1 Gag lattice (3): E160A/E161A/K162A, E207A/E208A, R232A/S234A, T242A/Q244A, and R275A (Fig. 4 and Table 1). Like the CTD-SP1 mutants, the NTD mutant proteins were generally incapable of assembling immature VLPs when induced with Ψ^CES alone, although the correct phenotype was recovered in the presence of IP6 (Fig. 4A and B and Fig. 6A and B); nevertheless, VLP yields for some mutants remained appreciably lower than those for the WT.

In nucleation experiments, Gag^ΔMA/Ψ^CES complexes harboring the T242A/Q244A and R275A mutations were nonfunctional, and only mature particles were observed when these were used to seed bulk WT Gag^ΔMA (Fig. 6C). IP6 completely rescued the T242A/Q244A phenotype, but R275A remained unable to nucleate immature VLPs (Fig. 6D and compare Fig. 4E to Fig. 4F). In propagation experiments, all the mutants were nonfunctional and formed only mature particles (E160A/E161A/K162A, E207A/E208A, R232A/S234A, and R275A) or aggregates (T242A/Q244A) (Fig. 6E). Again, IP6 rescued both the nonassembling and mature-only phenotypes, except for R275A, which propagated only mature particles (Fig. 6F and compare Fig. 4E to Fig. 4F). This result is consistent with an important role for the NTD in extending the immature lattice, as indicated by the cryoEM structure (3). The R275A mutant is quite notable, as it is impaired in both nucleation and propagation.

In vitro maturation.

Our in vitro assembly system provides an attractive vehicle for attempting to recapitulate and analyze the dynamic transitions that occur during maturation. This process has been difficult to study with bulk methods, because in preparations of authentic virions, the particles do not mature synchronously (35, 36). To determine whether our immature VLPs can undergo synchronous maturation, we digested the particles with recombinant HIV-1 protease (PR). As shown in Fig. 7A and B, maturation products accumulated in the expected order and with the CA-SP1 intermediate being the most long-lived. As expected for an acid protease, proteolysis was more efficient at pH 6, at which almost all the Gag^ΔMA proteins were converted into CA and NC after 1 h (compare Fig. 7A and B). Also as expected, cleavage at the CA-SP1 junction was further attenuated by the maturation inhibitor bevirimat (BVM), and this effect was more significant at pH 7.5, as previously shown (37). Interestingly, the addition of IP6 further inhibits proteolysis, synergistic to BVM. This suggests that the two compounds may not compete for the same binding site in the 6-helix bundle.

FIG 7.

In vitro maturation of HIV-1 VLPs. (A) SDS-PAGE gel showing proteolysis products at the indicated time points after the addition of PR to VLPs assembled by seeding with Gag^ΔMA/Ψ/IP6. Proteolysis reactions at pH 7.5 were performed either with or without additional IP6 in the reaction buffer, as indicated. (B) Same as panel A except that the reactions were performed at pH 6.0. (C) Negative-stain EM images of the indicated proteolysis reactions at the zero time point (top row) or after 120 min of incubation (bottom row). Insets show zoomed-in views. Bars, 200 nm. Experiments were performed two times with independent protein preparations.

Protease digestion also resulted in the formation of mature capsid-like particles, although this effect was dependent on pH and on the presence of both IP6 and nucleic acids (Fig. 7C). At pH 7.5, residual undigested immature particles could be observed by negative-stain EM even after 2 h (Fig. 7C, last column). Particles resembling mature cones or tubes were observed only at pH 6 when nucleic acids were present (Fig. 7C, second and third columns). Interestingly, when the starting immature VLPs contained Ψ, 46% of the mature particles (241/527 total particles examined) formed as tubes, some as long as 1 μm; the remainder appeared more capsid-like (cones and short capped cylinders). When the VLPs contained single-stranded RNA, the cone-shaped particles were more prevalent, with tubes accounting for only 5% (28/541). Internal densities were observed in many of the mature particles (Fig. 7C, inset boxed in red), indicating that the RNAs were reencapsidated, although this requires further confirmation. Importantly, the presence of IP6 was necessary for mature particle formation. We believe that this is due to the ability of IP6 to also promote the assembly and stability of the mature lattice (16, 38), which increases the probability that the newly mature CA protein molecules would reassemble rather than diffuse away.

DISCUSSION

Much of the current knowledge on the mechanism of HIV-1 assembly has been derived from reductionist studies of minimal in vitro systems, which allow investigators to focus on each of the various types of interactions and biochemical activities of the Gag and CA proteins that are required for virion formation (26). There is therefore a need to develop experimental approaches that will allow an integrative analysis of how these functionalities cooperate to generate a retrovirus particle, as exemplified by recent studies (39). Previously, it was shown that nucleic acids and IP6 can independently promote HIV-1 Gag assembly in vitro (16, 28, 40, 41), that the CA-SP1 junction 6-helix bundle is a required element of the immature Gag hexamer structure (3, 4, 11, 15, 28, 42), and that the CA CTD dimer interface is critical for extending the Gag lattice (3, 4, 34, 43). Here, we describe the development of an in vitro reconstitution system that we used to analyze how these distinct biochemical activities of Gag are integrated during HIV-1 assembly. A key feature of this system is that we simultaneously monitor both the assembly kinetics and assembly products in real time. Of particular importance, we found that it was critical to first take into account the known propensity of HIV-1 Gag to assemble into both mature and immature particles, as the competing mature assembly pathway (which we treat here as an in vitro artifact) occurs with rapid kinetics and is vastly amplified by large RNA molecules. Our results therefore raise a cautionary note with regard to minimalist studies of Gag/RNA binding, as the interpretation of the resulting data may be misleading unless steps are taken to ensure that the interactions lead to the formation of the appropriate immature assembly products.

We found that RNA can indeed induce the assembly of Gag into immature VLPs but with slow assembly kinetics. IP6-induced assembly in the absence of RNA is likewise slow, and it is the combination of RNA and IP6 that yields the maximal assembly rate. Taken together with data from previous studies, our results support the idea that Gag harbors all the information necessary to nucleate and propagate the assembly of the immature lattice and that RNA and IP6 are more appropriately viewed as modulators of assembly. It is important to note that our current system does not take into account other possible assembly modulators, such as the Gag/membrane and possibly Gag/tRNA interactions mediated by the MA domain (39, 44–51). We expect that a membrane-bound system would have even more enhanced efficiency; however, our current setup is analogous to the case of Mason-Pfizer monkey virus, which assembles complete immature capsids in the cytoplasm prior to membrane targeting and envelopment (52, 53).

It is important to understand the nature of the Gag^ΔMA/Ψ^CES complexes that we used to seed the assemblies and the role of IP6 in facilitating assembly. In assembly systems initiated by nucleation, the addition of the preformed structural nucleus eliminates the lag time of assembly (54). By this criterion, the Gag^ΔMA/Ψ^CES oligomers clearly did not satisfy the definition of a preformed nucleus, and the presence of IP6 in the seed mixture (but not the propagating mixture) was required to eliminate the lag time. This could be interpreted to mean that the nucleus requires a properly folded 6-helix bundle. On the other hand, kinetic models of virus assembly show that the initial lag does not reflect the formation of the nucleus, as commonly thought, but rather reflects the time required to achieve a steady state of assembly intermediates (55). Within this conceptual framework, the effect of IP6 is thus explained as promoting the formation of the intermediates, which in turn accelerates the overall assembly kinetics. Regardless, we emphasize that it is the combination of RNA and IP6 that allows both rapid assembly kinetics and the formation of the correct immature assembly products in our system.

The assembly phenotypes of the various HIV-1 Gag mutants, when used as either a nucleator or a propagator, agree very well with the known structure of the immature HIV-1 Gag lattice (3, 4). These results indicate that many of the structural requirements of the initiating Gag and propagating Gag are similar. Interestingly, certain alanine substitutions conferred different degrees of impairment depending on whether the mutant protein was used to nucleate or propagate assembly. One of these mutations is W316A/M317A, which disrupts the CTD dimerization interface that is critical for extending the immature lattice (3, 4, 34, 43). Whereas the W316A/M317A protein was significantly impaired when used as a propagator, it nonetheless retained significant functionality as a nucleator. Similarly, the V353A substitution, which disrupts a buried β-turn element that precedes the junction helix (4), remained capable of supporting nucleation. In cells, Gag proteins harboring this and other β-turn mutations have been reported to remain diffusely distributed in the cytosol and fail to form plasma membrane-associated Gag patches (56). Our interpretation of these results is that the nucleating Gag oligomer may have distinct structural requirements, or at least different dynamic properties, compared to the propagating Gag molecules. Understanding the details of these differing requirements will require further studies.

Interestingly, the CA NTD mutants highlight the importance of this subdomain in the assembly of immature virions, even though the NTD is dispensable for generating the minimal Gag lattice structure (4, 12). These mutant proteins contained intact CA-SP1 regions and so, for the most part, were responsive to rescue by IP6. An important outlier, however, is R275A, which induced mature-like behavior as both a nucleator and a propagator, in both the presence and absence of IP6. In the immature Gag hexamer, Arg275 is part of a buried hydrophilic surface that connects three different subunits: two NTDs and one CTD (7). In the mature lattice, this residue (Arg143 according to CA numbering) does not appear to significantly contribute to lattice-stabilizing interactions (57–59). These results uncover a previously unappreciated but apparently crucial role for contacts between the NTD and CTD in stabilizing the immature lattice. Although additional experiments are required to confirm this, the apparent all-or-none behavior of the R275A mutant further suggests that the system has been poisoned. Thus, the immature NTD-CTD lattice interactions may be a viable target for assembly inhibitors.

Although we did not address packaging specificity in detail in this study, our data indicate that RNA modulation of Gag assembly in vitro requires an intact NC domain but only during the nucleation steps; the NC zinc knuckles become dispensable once assembly is initiated. We therefore suggest that gRNA packaging specificity is determined at nucleation. Our results do not appear to support a recent model wherein Ψ RNA induces Gag assembly with more rapid kinetics than non-Ψ RNA (60, 61). Rather, we observe that Ψ^CES induced assembly kinetics similar to those of a generic 50-nt ssRNA oligonucleotide, although this fact was initially obscured by the competing pathway that produces mature particles with rapid kinetics. We surmise that packaging specificity is established by a more complicated mechanism that may involve the localized exposure of a critical number of high-affinity NC binding sites (62) (similar in concept if not in detail to murine leukemia virus [63]), the coordinated colocalization of Gag and gRNA molecules at specific sites within infected cells (64–68), and/or interdependence with other aspects of Gag assembly, such as membrane binding (39).

Finally, we show that our immature VLPs can be induced to mature into capsid-like particles by the simple addition of viral protease. Although aspects of this in vitro system remain to be improved, these results demonstrate that it is possible to biochemically reconstitute the sequential processes of HIV-1 assembly and maturation from purified components. Moving forward, the inclusion of membranes would be required to understand the contribution of MA-mediated interactions to assembly initiation and kinetics, and the inclusion of non-Ψ competitor RNA would be required to address packaging specificity. Experiments toward the further development of this assembly system are under way, and we expect that such studies will help to enhance our understanding of how the large number of biochemical activities expressed by the Gag protein functionally integrate and synergize during HIV-1 assembly and maturation.

MATERIALS AND METHODS

Protein expression and purification.

HIV-1 Gag^ΔMA and its mutant versions were purified using a previously reported protocol for CA-NC, which was designed to preserve the zinc knuckles of the NC domain that are critical for high-affinity binding to RNA (29). In brief, proteins were expressed in E. coli BL21(DE3) cells by induction with 1 mM isopropyl-β-d-thiogalactopyranoside (IPTG) for 4 h at 25°C. Bacteria were harvested by centrifugation and resuspended in lysis buffer (25 mM Tris [pH 7.5], 0.5 M NaCl, 1 mM ZnSO₄, 10 mM β-mercaptoethanol) supplemented with 0.3% (wt/vol) deoxycholate and protease inhibitor tablets (Roche). Cells were lysed by incubation with lysozyme and sonication. Polyethyleneimine (pH 8.0) was added to precipitate nucleic acids as described previously (29). The lysate was clarified by centrifugation, and the protein was precipitated by the addition of a 1/3 volume of saturated ammonium sulfate. The precipitated fraction was recovered by centrifugation, resuspended in lysis buffer, and then dialyzed overnight against a solution containing 25 mM Tris (pH 7.5), 50 mM NaCl, 1 mM ZnSO₄, and 10 mM β-mercaptoethanol. The protein was further purified by using ion-exchange chromatography (MonoS 10/100 GL column; GE Healthcare) with a 50 mM to 1 M NaCl gradient. Pooled fractions were dialyzed into a solution containing 25 mM Tris (pH 7.5), 1 mM ZnSO₄, and 10 mM β-mercaptoethanol; concentrated to ∼10 mg/ml by using 10-kDa-cutoff Vivaspin 15R centrifugal concentrators (Millipore); flash-frozen in liquid nitrogen; and stored at −80°C until use. The original ΔMA construct described by Gross et al. included the SP2 spacer (28). The construct used in this study was further truncated at NC and is the same as the one used by Wagner et al. (4).

HIV-1 protease (PR) used for the in vitro maturation experiments was expressed in inclusion bodies and purified based on a previously reported protocol (69). Briefly, PR was expressed in E. coli BL21(DE3) cells by induction with 1 mM IPTG for 4 h at 37°C. Cells were harvested by centrifugation, resuspended in PR buffer (20 mM Tris [pH 8.0], 0.1 M NaCl, 5 mM imidazole, 1 mM β-mercaptoethanol) supplemented with protease inhibitor tablets (Roche), and lysed by incubation with lysozyme and sonication. After centrifugation, the pellet containing inclusion bodies was washed once with 30 ml of PR buffer supplemented with 2 M urea and 1% Triton X-100, washed once with pure water, and then resuspended in 30 ml of PR buffer supplemented with 8 M urea. After overnight incubation at room temperature, the precipitate was removed by centrifugation, and the protein was purified on Ni-nitrilotriacetic acid (NTA) resin (Qiagen). Fractions were pooled and diluted with a solution containing 20 mM Tris-HCl (pH 7.9), 100 mM NaCl, 5 mM imidazole, and 8 M urea to ∼100 μg/ml protein. PR was refolded by stepwise dialysis against a solution containing 20 mM Tris-HCl (pH 7.9), 100 mM NaCl, 1 mM phenylmethylsulfonyl fluoride (PMSF), 10% glycerol, and 1 mM dithiothreitol (DTT), with gradually decreasing urea concentrations. The refolded protein was concentrated by using 5-kDa-cutoff Vivaspin 15R centrifugal concentrators (Millipore) to 0.1 mg/ml and stored at −80°C.

RNA preparation.

RNA oligonucleotides (GU repeats of a 25- or 50-nt total length) were purchased from Integrated DNA Technologies.

Ψ and Ψ^CES RNAs were prepared by in vitro transcription using T7 RNA polymerase as previously described (31). Briefly, small-scale reactions were used to optimize the amounts of MgCl₂, nucleoside triphosphates (NTPs), and the PCR-amplified DNA template to obtain the best RNA yield. Reaction mixtures (15 ml) for both Ψ and Ψ^CES RNAs were incubated at 37°C for 5 h before quenching by the addition of EDTA-urea (250 mM EDTA, 7 M urea [pH 8.0]). The RNA products were then purified using 6% denaturing PAGE gels (National Diagnostics). RNAs were eluted from the gel bands using the Elutrap electroelution system (Whatman). The eluted RNAs were then washed twice with 2 M NaCl and desalted eight times with H₂O using 30-kDa Amicon Ultra-4 centrifugal filter devices (Millipore). Ψ and Ψ^CES were dimerized prior to use by incubation in PI buffer, as described previously (31).

E. coli rRNA was purified from a 2-liter pellet of BL21(DE3) cells grown overnight in 2× YT (yeast-tryptone) medium at 37°C. Cells were suspended in 40 ml of a solution containing 10 mM Tris-HCl (pH 7.5), 10 mM MgCl₂, and 50% (vol/vol) phenol. After incubation at room temperature for 2 h, phases were separated by centrifugation, and the aqueous layer was collected and reextracted with chloroform. Total RNA was ethanol precipitated, resuspended in 20 mM Tris-HCl (pH 7.5)–10 mM MgCl₂, and purified by using ion-exchange chromatography (MonoQ; GE Healthcare) with a 50 mM to 1 M NaCl gradient. Fractions containing rRNA were identified by agarose gel electrophoresis, pooled, and further purified using a Superdex 200 column (GE Healthcare) in a solution containing 10 mM Tris-HCl (pH 7.5), 10 mM MgCl₂, and 100 mM NaCl. Final fractions were combined and stored at −80°C.

Assembly reactions.

The ratios of nucleic acid to protein in the final assembly reaction mixtures were maintained at 10 nucleotides per protein molecule, which means that concentrations of seed complexes were adjusted according to the length of the RNA component. Seed complexes containing Gag and RNA (e.g., Gag^ΔMA/Ψ^CES at a 6:2 molar ratio and Gag^ΔMA/Ψ^CES/IP6 at a 6:2:6 molar ratio) were generated at 37°C for 2 h in PI buffer (20 mM Tris [pH 7.5], 140 mM KCl, 10 mM NaCl, 5 mM MgCl₂, 10 mM TCEP). Assembly was initiated by adding the seed complex to a 50 μM solution of free Gag^ΔMA in PI buffer containing 375 mM tartrate. We found that tartrate was necessary for assembly to proceed within the time window utilized in the experiment; omitting tartrate resulted in partial precipitation after ∼120 min but not the formation of VLPs. Assembly kinetics was monitored by measuring light scattering at 320 nm in a SpectraMax M5 plate reader (Molecular Devices). Aliquots were sampled at various time points (typically 20, 60, and 120 or 180 min) and processed for negative-stain EM.

Electron microscopy.

Samples were prepared for electron microscopy according to two different protocols, as we found that the immature and mature particles had different optimal staining conditions. For samples made up of mostly immature and mixed particle types, 3 μl of the sample was applied to a carbon-coated grid for 3 min, and the grid was then incubated with 2% (wt/vol) uranyl acetate for 3 min. Excess liquid was blotted off by touching the edge of the grid with filter paper. For mature particles, 3 μl of the sample was applied to a carbon-coated grid for 2 min. Subsequently, the grid was washed with 100 mM KCl for 2 min, blotted, stained with 2% (wt/vol) uranyl acetate for 2 min, and blotted again. Images were recorded using a Tecnai F20 or Tecnai Spirit system (FEI), both operating at 120 kV.

In vitro maturation.

For the PR digestion assays, Gag^ΔMA/Ψ^CES/IP6 complexes at a 6:2:6 molar ratio were incubated for 2 h at 37°C and added to 100 μM Gag^ΔMA in PI buffer or MI buffer (20 mM morpholineethanesulfonic acid [MES] [pH 6.0], 140 mM KCl, 10 mM NaCl, 5 mM MgCl₂, 10 mM TCEP) containing 375 mM tartrate. Assembly reaction mixtures were incubated for 3 h at 37°C, after which the IP6 concentration was adjusted to 1 mM. PR was added to the assembly mixture at a 1:50 ratio relative to Gag^ΔMA, and the sample was incubated for 2 h at room temperature. Aliquots were removed at the indicated time points and analyzed by using SDS-PAGE with Coomassie staining and/or negative-stain EM.