Cryo-electron microscopy of tubular arrays of HIV-1 Gag resolves structures essential for immature virus assembly

Significance

HIV-1 undergoes a two-step assembly process. First, an immature noninfectious particle is assembled, which leaves the infected cell. Second, the structural protein, Gag, is cleaved in the virus by the viral protease, and this leads to formation of the infectious virus. The immature virus particle therefore represents the key intermediate in HIV-1 assembly. There is currently no high-resolution information available on the arrangement of Gag within immature HIV-1. We have assembled part of HIV-1 Gag in vitro to form immature virus-like tubular protein arrays, and have solved a subnanometer-resolution structure of these arrays by using cryo-EM and tomography. This structure reveals interactions of the C-terminal capsid domain of Gag that are critical for HIV-1 assembly.

Abstract

The assembly of HIV-1 is mediated by oligomerization of the major structural polyprotein, Gag, into a hexameric protein lattice at the plasma membrane of the infected cell. This leads to budding and release of progeny immature virus particles. Subsequent proteolytic cleavage of Gag triggers rearrangement of the particles to form mature infectious virions. Obtaining a structural model of the assembled lattice of Gag within immature virus particles is necessary to understand the interactions that mediate assembly of HIV-1 particles in the infected cell, and to describe the substrate that is subsequently cleaved by the viral protease. An 8-Å resolution structure of an immature virus-like tubular array assembled from a Gag-derived protein of the related retrovirus Mason–Pfizer monkey virus (M-PMV) has previously been reported, and a model for the arrangement of the HIV-1 capsid (CA) domains has been generated based on homology to this structure. Here we have assembled tubular arrays of a HIV-1 Gag-derived protein with an immature-like arrangement of the C-terminal CA domains and have solved their structure by using hybrid cryo-EM and tomography analysis. The structure reveals the arrangement of the C-terminal domain of CA within an immature-like HIV-1 Gag lattice, and provides, to our knowledge, the first high-resolution view of the region immediately downstream of CA, which is essential for assembly, and is significantly different from the respective region in M-PMV. Our results reveal a hollow column of density for this region in HIV-1 that is compatible with the presence of a six-helix bundle at this position.

The major structural component of HIV-1 is the 55-kDa Gag polyprotein. Gag proteins from all retroviruses contain an N-terminal membrane-binding matrix (MA) domain; two central domains that together comprise capsid (CA); and an RNA-binding nucleocapsid (NC) domain located toward the C terminus of Gag. Between, and downstream of these domains are regions that differ between retroviruses. In HIV, a spacer peptide, SP1, is found between CA and NC, whereas downstream of NC are a second spacer peptide, SP2, and the p6 domain (1).

Assembly of an infectious HIV-1 particle proceeds in two stages (2). In the first stage, Gag oligomerizes into a curved lattice on the cytoplasmic face of the plasma membrane, leading to outward protrusion of the membrane and subsequent membrane scission to release an immature virus particle. Within the immature virus particle, Gag is arranged in a radial manner with the N-terminal MA domain on the inner surface of the membrane, and the C terminus of Gag pointing toward the center of the particle. In the second stage of assembly, the viral protease becomes activated and cleaves Gag at five positions, releasing the component domains, leading to a structural rearrangement within the virus particle. MA remains bound to the viral membrane, NC and the associated viral genome condense in the center of the particle, and CA assembles to form a conical core around the viral genome. This proteolytic maturation is required for the particle to become infectious and is a target of antiretroviral drugs (3). The immature virus particle therefore represents a key intermediate in HIV-1 assembly. Obtaining detailed structural information on the arrangement of Gag within immature HIV-1 is essential to reveal the protein interfaces that mediate assembly of the immature virus, and to reveal the structure of the substrate that must be cleaved by the viral protease to induce maturation to the infectious form.

There has been an extensive effort to apply structural biology methods to understand the structures and interactions that mediate assembly of HIV-1, and the structural changes associated with maturation (4, 5). High-resolution structural data exist for all the folded domains of HIV-1 Gag in their isolated forms. Recently, a series of breakthrough studies have revolutionized our understanding of how the CA domains self-assemble to form the mature capsid core of the virus (6–8), and a pseudoatomic model for the assembled mature core is now available (9). In contrast, much less structural information is available on the immature virus. A structure of the HIV-1 Gag lattice within immature-like particles assembled in vitro from a Gag-derived protein has been resolved to a resolution of ∼17 Å by using subtomogram averaging (10). At this resolution, the approximate positions of the protein domains within the lattice can be defined, but their orientation is not clearly determined. We recently solved the structure of immature-like tubular assemblies of a Gag-derived protein (ΔProCANC) from another retrovirus, Mason–Pfizer monkey virus (M-PMV), to a resolution of ∼8 Å (11). At this resolution, the arrangement of the CA domains could be clearly determined, revealing the regions of CA that interact within the assembled immature M-PMV Gag lattice. Based on the expected structural homology between the HIV-1 and M-PMV CA domains, a model for the arrangement of the CA domains in immature HIV-1 was generated and compared with the CA arrangement in mature HIV-1 cores. Strikingly, the interfaces that stabilize the immature and the mature CA lattices are almost completely distinct in this model.

The SP1 peptide plays a crucial role in HIV-1 assembly and maturation (2). A stretch of ∼25 residues downstream of the folded C-terminal domain of CA comprising the C-terminal residues of CA and SP1 is known to be essential for immature HIV-1 assembly (12, 13). This stretch encompasses the CA-SP1 proteolytic cleavage site, mutation of which abolishes formation of mature cores and infectious viruses (14). The bevirimat class of HIV-1 assembly inhibitors blocks HIV-1 replication by specifically targeting processing at this site (15). The structure of this important region of Gag has not yet been resolved with high resolution, however. Early mutational studies and molecular modeling led to the proposal that this region forms an extended α-helix (13), and this was supported by subsequent NMR studies of the isolated sequence (16). The presence of thick columns of density linking the CA and NC layers in low-resolution EM structures is consistent with an extended six-helix bundle forming in this region (10, 13, 17). M-PMV, which contains no spacer peptide between CA and NC, diverges structurally from HIV-1 in this region (18), and, instead of the thick density columns, six splayed densities were observed in the M-PMV ΔProCANC tubes (11).

The available structural data therefore do not answer several critical questions: Most importantly, (i) does the arrangement of the CA domains in immature HIV correspond to the model generated based on structural homology from the M-PMV–derived tube structure; and (ii) what is the structure of the region immediately downstream of CA that is critical for immature assembly and the regulation of maturation?

The key step in obtaining detailed structural data for the immature M-PMV lattice was to generate immature-like tubular arrays of a Gag construct. The helical symmetry of the M-PMV ΔProCANC tubes allowed application of a hybrid cryo-electron tomography (ET) and cryo-EM image processing approach for structure determination (11). HIV-1 Gag derivatives that assemble tubular arrays adopt a mature-like arrangement of CA (19, 20), and have indeed been important tools in determining the structure of the mature CA core (9, 21). In contrast, HIV-1 Gag derivatives known to assemble into immature-like arrays form approximately spherical structures (22, 23) that are unsuitable for application of high-resolution helical reconstruction methods.

Here we identify an HIV Gag-derived protein competent to assemble partially immature-like tubular arrays, and determine their structure to a resolution of 9.4 Å. The structure reveals the immature-virus-like arrangement of the C-terminal domain of HIV CA (CA-CTD) and the CA-SP1 linker region.

Results and Discussion

Immature-Like Tubular Arrays.

We previously assembled immature-like tubular arrays of M-PMV Gag from a protein comprising CA, NC, and the intervening residues, but lacking the N-terminal proline residue of CA (ΔProCANC) (11). Similar HIV-1 Gag-derived proteins assemble narrow tubular arrays (Fig. 1A) (19, 20, 24), which have a mature-like arrangement of CA. Prior studies had indicated that certain HIV-1 Gag-derived proteins can assemble tubular or spherical particles depending on the conditions (22) and that certain mutations in the CA-SP1 region converted the assembly from spherical to tubular or aberrant sheet-like structures (12, 25). Furthermore, mutations in the region targeted by the capsid assembly inhibitor CAI do not alter assembly of immature particles in vitro or in tissue culture, whereas they completely disrupt assembly of the mature-like lattice (26). We therefore purified CANC variants and other Gag-derived proteins with these mutations, and tested them for assembly of wide tubular particles. We found that CANC proteins with the mutation Y169L or Y169S yielded tubes with a wider diameter upon negative stain EM, whereas all other constructs tested did not lead to this phenotype or abolished in vitro assembly. Other constructs tested included Y169S/L mutations within constructs beginning after the N-terminal proline residue of CA or carrying mutations in predicted N-terminal interfaces, constructs with deletion or partial deletion of the N-terminal β-hairpin sequences, and hybrid constructs made by fusing the N-terminal domain of CA (CA-NTD) from M-PMV with the CA-CTD and the downstream sequence of Gag from HIV-1. Accordingly, CANC variants Y169L and Y169S were selected for further study.

Fig. 1.

Arrangement of the Gag lattice is immature-like in the HIV-1 CANC Y169L tubes. (A) WT HIV-1 CANC protein was assembled along with DNA. On a slice through a tomogram of the assembled tubes, thin mature-like tubes were observed. (B) A single amino acid mutation Y169S/L converts thin mature-like tubes to thick immature-like tubes (Fig. S1). (C) Comparison of radial density profiles of CANC Y169L tubes with immature HIV-1 virus particles and mature-like WT CANC tubes. (D) A schematic representation of the HIV-1 Gag polyprotein aligned with the radial density profile shown in C (green). Positions of the NC, SP1, CA-CTD, CA-NTD, and MA domains along with the viral membrane are indicated.

We assembled CANC Y169L or Y169S in vitro in the presence of nucleic acid and imaged the assembly products by cryo-ET. Both proteins assembled into tubular arrays with an overall average diameter of 744 ± 23 Å (Fig. 1B and Fig. S1) and were indistinguishable in cryo-ET images. Radial density profiles were calculated for CANC Y169L tubular assemblies from the cryo-ET data. The results were compared with those obtained for WT CANC tubes (diameter, 300–400 Å), and for immature HIV-1 particles produced from transfected cells in the presence of a protease inhibitor (Fig. 1 C and D). CANC Y169L tubes showed a similar radial density profile as immature HIV-1: both displayed three separate peaks corresponding (from the inside) to the ribonucleoprotein complex (NC layer), the CA-CTD, and the CA-NTD, whereas immature HIV-1 exhibited two additional peaks corresponding to the MA-bound lipid bilayer. These observations suggest that CANC Y169L and CANC Y169S tubes assemble immature-like protein lattices.

CA residue Y169 is located in a tight interface formed between the CA-NTD and CA-CTD within the assembled mature HIV-1 core (7, 27). It strongly coevolves with its interaction partners within this interface (28). Mutation of Y169 abolishes proper mature core formation and thereby infectivity, but is fully compatible with immature assembly in vitro and with release of virus particles with regular immature morphology (26). It seems likely that, for Y169L and Y169S mutants, the mature lattice is destabilized by the mutation, making assembly of an immature-like arrangement of CA the most energetically favorable state.

Solving the Structure of the Tubular Arrays.

To obtain a 3D structure of the tubes, we applied the same hybrid tomography/helical reconstruction methodology that we had previously used to determine the structure of immature-like M-PMV ΔProCANC tubular arrays (11) (Fig. 2, Table S1, and SI Materials and Methods). Data from both mutants was pooled and two independent reconstructions were generated and found to be equivalent to a resolution of 9.4 Å (Fig. S2A). They were averaged together to generate the final structure (Fig. 2 C–F).

Fig. 2.

Structure of the HIV-1 CANC Y169S/L tubes at 9.4-Å resolution. (A) Structure of a single CANC Y169L tube solved by using real-space helical reconstruction. One hexamer of Gag is colored red to illustrate the arrangement of the lattice. Isosurface threshold is 1.5σ away from the mean. (B) The same structure as in A rotated 90° and magnified to show the different layers in the lattice (annotated). (C) On averaging the asymmetric unit extracted from 41 different tubes, a 9.4-Å resolution structure was obtained (Fig. S2 and Movie S1). Isosurface threshold is 1.0σ away from the mean in C–E. One pseudohexameric asymmetric unit of Gag CA has been colored red. The dotted line shows the position of the clipping plane used to produce E and F. (D) The same reconstruction as C, in an orthogonal orientation. (E) Six copies of HIV-1 CA-NTD molecules (blue ribbon) were flexibly fitted into the outermost layer of the pseudohexameric asymmetric unit. (F) Six corresponding HIV-1 CA-CTD molecules (orange ribbon) were similarly fitted into the inner layer of the hexamer (SI Materials and Methods and Fig. S3). Isosurface threshold is 1.5σ away from the mean.

As previously described for in vitro assembled HIV Gag spheres (10), the tubes show three layers of density: outermost is the CA-NTD layer, below this the CA-CTD layer, and innermost is the NC layer (Fig. 2B). The CA-CTD and NC layers are joined by column-like densities. Although the NC layer appears disordered, the two CA layers are structured (Fig. 2 B and C). The resolution of the final reconstruction varied through the structure from an average resolution of 10.0 Å in CA-NTD, 9.4 Å in CA-CTD, and 10.8 Å at SP1 (Fig. S2B), likely reflecting differing degrees of order at different radii. In both CA layers, the electron density contains three independent copies of the monomer (Fig. S3), which are not related to one another by the twofold symmetry or the helical symmetry implicit in the reconstruction protocol.

We wished to compare the resulting density with high-resolution crystal structures of the individual protein domains. Crystal structures are available for the NTD and CTD of CA (5). To ensure that the mutations at position 169 of CA do not cause substantial changes in the structure of the CA-CTD, we expressed, purified, and crystallized both Y169L and Y169S CA-CTD according to established protocols (SI Materials and Methods), and solved their structures by X-ray crystallography (Table S2). We found them to be similar to the previously resolved structures of the HIV CA-CTD Y169A protein (26) (rmsd, 0.50 Å and 0.74 Å for CA-CTD Y169S and Y169L proteins, respectively).

We carried out rigid body fitting of the Y169S CA-CTD X-ray structure into the central density layer of the cryo-EM structure. We also carried out rigid body fitting of an available high-resolution atomic structure of the CA-NTD [Protein Data Bank (PDB) ID code 2JPR] (29) into the outer density layer of the cryo-EM structure and manually joined the domains. We then applied established molecular dynamics-based flexible fitting (MDFF) (30) methods to the joined structure to generate an improved model of the CA layer (Figs. 2 E and F and 3, Fig. S3, and Movie S1). For the CA-NTD and CA-CTD domains, a good rigid body fit was observed into each of the three independent copies of the monomer, and only minimal movements of helices took place during flexible fitting (Fig. S4). This indicates that the structures of the domains within the tubes are similar to those resolved in the input atomic structures.

Fig. 3.

Arrangement of CA domains in HIV-1 CANC Y169S/L tubes. (A) The arrangement of the CA lattice on the HIV-1 CANC Y169S/L tubes viewed from outside the surface of the tube. The CA-NTD domains are colored blue and the CA-CTD domains are colored orange. A gray hexagon is superimposed on the image to show the pseudohexameric asymmetric unit on the surface of the tubes. (B) Arrangement of the CA-NTD lattice. (C) Arrangement of the CA-CTD lattice. Fig. S5 A and B show CA-NTD and CA-CTD symmetry axes. (D) The CA lattice viewed from the side of the tube. A gap is observed between the CA-NTD and CA-CTD layers. (E) Additional densities are observed in the final reconstruction, outside the fitted first seven α-helices of CA, which are consistent with the folded N-terminal β-hairpin. Isosurface threshold is 1.0σ away from the mean. (F) The CA-CTD and CA-NTD layers are connected by the 7–8 linker (arrowheads), which is flexible and forms three different arrangements (Fig. S6). Isosurface threshold is 0.75σ away from the mean.

Finally, to allow direct comparison of the protein arrangements in the HIV-1 CANC Y169S/L tubes with those previously described for M-PMV ΔProCANC tubes, we also revisited the M-PMV data and applied the same MDFF-based flexible fitting to generate a revised structural model of the M-PMV–derived tubes (Fig. S3F).

Arrangement of the CA Domains.

Unexpectedly, the CA-NTDs in the HIV-1 CANC Y169S/L tubes are not assembled to form a hexameric lattice, as had been observed in M-PMV ΔProCANC tubes and in previous studies on immature HIV-1 Gag assemblies (10, 17), but instead pack together to form a lattice that is “p2” in crystallographic terms, containing four twofold positions (Figs. 2E and 3B and Fig. S5 A and C). Although a lattice with this symmetry can form a tubular array, it is difficult to adapt it to form spherical structures as would be necessary in the virus. Within the lattice, clear additional densities outside the positions of the fitted first seven CA α-helices are observed, corresponding to the N-terminal β-hairpin of the CA-NTD (Fig. 3E). This feature cannot be present in the immature virus, because the β-hairpin only forms upon proteolytic cleavage of the N terminus of CA from the preceding MA domain. The β-hairpin appears to mediate interactions between CA-NTDs in the HIV-1 CANC Y169S/L tubes. Taken together, these observations suggest that the arrangement of the CA-NTD within the HIV-1 CANC Y169S/L tubes is unlikely to represent the arrangement of the CA-NTD within the immature virus and that it is most probably specific to the in vitro assembly of this mutant protein. We have therefore not interpreted it further. However, we note the remarkable capacity of the CA-NTD to oligomerize in at least three different ways to form curved lattices—first as seen in the mature HIV-1 capsid, second as seen in the M-PMV ΔProCANC tubes, and third as seen in the HIV-1 CANC Y169S/L tubes described here.

The CA-CTD monomers are well resolved (Figs. 2F and 3C and Fig. S2 E and F). They form an approximately hexameric lattice, “p6” in crystallographic terms (Fig. 3C and Fig. S5 B and D). The lattice spacing derived from the helical parameters was 72.5 Å at the CA-CTD:SP1 interface, corresponding to 81 Å at the CA-NTD. This is in good agreement with the interhexameric distance of ∼80 Å at the CA-NTD measured in the immature HIV-1 Gag lattice by previous cryo-EM studies (17, 31), and significantly smaller than the ∼95 Å interhexameric spacing observed in the mature lattice (6). In the structure, there is a space between the CA-NTD and CA-CTD domains (Fig. 3D). The space between the domains is larger than that previously observed in the M-PMV ΔProCANC tubes. No direct protein–protein contacts were observed between the two domains, but the polypeptide chain linking the domains (the 7–8 linker) is resolved (Fig. 3F and Fig. S3 C and D). Because of the mismatch between the p2 symmetry of the CA-NTD and p6 symmetry of the CA-CTD lattices, there are three different relative positions of the CA-CTD and CA-NTD. All these positions differ from the relative positions of the CA-CTD and CA-NTD in the M-PMV ΔProCANC tubes and in the mature HIV-1 CA lattice (Fig. S6). To link the domains, there are therefore three different orientations of the 7–8 linker (Figs. S3 C and D and S6). Together, these data indicate that the arrangement of the CA-CTD within the immature-like lattice can be formed independent of the arrangement of the CA-NTD, and that the linker between them is highly flexible.

The CA-CTD layer is made of compact α-helical domains (Fig. 2F and Fig. S2 E and F). The overall arrangement of the CA-CTD monomers resembles that seen in the M-PMV ΔProCANC tubes (11) and in one of the previously derived models of the immature HIV-1 CA-CTD lattice (10). The CA-CTD monomers in the HIV-1 CANC Y169S/L tubes show a kink in helix 9 of CA, a feature also observed in the X-ray atomic structure, but not observed in the M-PMV ΔProCANC tubes. The dimer interface and the overall arrangement of the HIV-1 CA-CTD dimers around the hexameric axis are similar to that obtained for M-PMV (Fig. S3 E and F). In HIV-1 and M-PMV tubes, CA-CTD dimerization is mediated by a hydrophobic interface between helix 9 in each monomer (Figs. S3 E and F and S7). In HIV-1 CA, residues W184 and M185 are close to this interface, whereas, in M-PMV CA, residue Y180 (in the sequence VDYV) is close to this interface. The relative orientation of the CA-CTD monomers within a CA-CTD:CA-CTD dimer is similar to that in the M-PMV ΔProCANC tubes (only 1° difference in the crossing angle of the two helix 9s), and to that in the crystal structure of CA Y169A (4° difference in crossing angles). In contrast, the relative orientation differs strongly from “mature-like” dimeric forms such as mature-like CA arrays (138° difference) or HIV-1 CA crystals (118° difference; Fig. S7). Despite the highly similar structures of the dimers, the sizes of the hexamers formed by these CA-CTD dimers are different: 72.5 Å for HIV-1 and 86 Å for M-PMV (measured at the CA–SP1 interface). To accommodate the smaller hexamer size, adjacent CA-CTDs around the sixfold axis are closer together in the HIV-1 CANC Y169S/L tubes (18.9 Å center to center) than in the M-PMV ΔProCANC tubes (22.6 Å; Fig. S3 E and F). The interface between adjacent CA-CTDs around the sixfold is correspondingly more extensive in the HIV-1 CANC Y169S/L tubes (including CA residues R154-P157, E212, and M215-T216) than in the M-PMV ΔProCANC tubes [including CA residues K149, P152 (in the sequence VKQGPD), R211, and S214 (in the sequence IRLCSD)]. The major homology region, conserved across retroviruses, covers the base of helix 8, and the adjacent part of the 7–8 linker. This region includes residues likely to function in positioning the 7–8 linker and stabilizing the CA-CTD fold, but also the residues R154–P157 that contribute to the interactions around the hexamer.

Arrangement of the CA-SP1 Region.

Below the CA-CTD layer are columns of ordered density that connect the CA-CTD to the NC along the sixfold symmetry axes of the lattice (Fig. 4 A and B and Fig. S2G). They are formed by six extended structures surrounding a central cavity. The arrangement is approximately sixfold symmetric, but deviates from this symmetry toward the NC. These columns of density are in the position corresponding to the C-terminal residues of CA and N-terminal part of SP1, a region of the protein that is essential for assembly. These residues can adopt a helical conformation in solution (16), but there is currently no high-resolution structural information available for this region in an oligomeric state. The resolution of our cryo-EM structure is not high enough to permit ab initio modeling of this region, but the density we observe is consistent with the presence of the predicted six α-helices arranged around a central hole (Fig. S2G). To test whether the density has the dimensions expected for six helices packed together, we compared it with the X-ray structure of the six-helix bundle from cucumber mosaic virus coat protein (32). The dimensions of the helical bundles are almost identical (Fig. 4A). Together, these observations provide strong support for the presence of a hollow six-helix bundle in this region.

Fig. 4.

Arrangement of the region immediately downstream of CA. (A) View of a hexamer of HIV Gag from inside the tube. The region downstream of helix 11 of CA forms columns of density that connect the CA-CTD to the NC along the pseudo–six-fold symmetry axes of the lattice (a slice through the density is presented in Fig. S2G). This arrangement is consistent with a six-helix bundle model. The six-helix bundle found in cucumber mosaic virus (green) has been placed in the density for comparison. (B) A view from the side, orthogonal to A. Six extended densities are observed linking the CA and NC layers. Isosurface threshold is 1.0σ away from the mean. (C) A view of the previously solved M-PMV ΔProCANC tubes in the same orientation as A. No density consistent with a six-helix bundle is seen. (D) A view of the M-PMV ΔProCANC tubes from the side. Isosurface threshold is 1.5σ away from the mean.

The hollow columns of density seen in the HIV-1 CANC Y169S/L tubes is in contrast to the six splayed densities that point away from the sixfold symmetry axis in the M-PMV ΔProCANC tubes (Fig. 4 C and D). This contrast indicates substantial structural divergence in this region in the two viruses. The point at which the two electron densities strongly diverge corresponds approximately to the position of HIV-1 CA residue L231 (the last residue of CA) based on a provisional model of the region downstream of CA (SI Materials and Methods and Fig. S8 B and D). The observed structural divergence is consistent, with structural differences observed in lower-resolution structures of spherical in vitro assembled HIV-1 and MPMV immature virus-like particles (18) and also with sequence divergence: the region downstream of CA in HIV-1 shows a strong hydrophobic helical moment compatible with the presence of a helical bundle (13), whereas no strong helical-hydrophobic moment is seen in this region of the M-PMV sequence (18, 33). Nevertheless, it is not obvious how to reconcile the observed structural divergence of this region with the apparent functional equivalence: mutagenesis data indicates that M-PMV contains a “spacer-like” sequence that, as in HIV-1, functions as a key determinant of assembly (34). Higher-resolution structural analysis will be required to understand whether the different structures of this sequence in the two viruses could conceal related mechanisms for stabilizing the assembling Gag lattice.

Implications for Assembly and Its Inhibition.

It has long been known from structural studies of isolated domains that CA-CTD and CA-NTD can fold independently of one another. Here we found that the CA-CTD can assemble an immature-like lattice independent of the arrangement of the CA-NTD. This “mismatched” lattice is accommodated by the flexibility of the linker between the domains. These observations raise the question: what is the role of the CA-NTD during assembly in vivo? Indeed, previous observations have shown that proteins consisting only of the CA-CTD and the adjacent SP1 peptide, replacing upstream regions with a myristoylation site and downstream regions with a leucine zipper, are competent to assemble VLPs in cells, albeit at lower efficiency, despite the absence of CA-NTD (35). Conceivably, the CA-NTD may modulate assembly, while being more important for its structural role within the mature capsid, or its role in the early stages of infection, mediated via interactions with cellular factors such as CypA, CPSF-6, or nuclear porins. Consistent with this suggestion, although assembly inhibitors with binding sites in the CA-NTD in some cases lead to a reduction in virus production, binding of other molecules to the CA-NTD instead inhibit virus maturation (36) or act during early infection (37).

The independence of the two CA domains may also be functionally important in maturation. Cleavage upstream or downstream of CA is not sufficient to induce disassembly of a preassembled immature Gag lattice, but cleavage on both sides of CA is required (38). The data presented here indicate that cleavage between CA and MA could even permit β-hairpin formation—and potentially rearrangement of the NTD—without destabilizing the immature CA-CTD arrangement. Low-resolution structural studies of inhibited viruses suggest that the bevirimat class of inhibitors, which inhibit proteolytic cleavage at the CA–SP1 boundary, stabilize an immature conformation of the lattice (39). We speculate that bevirimat could also arrest the virus in such a hybrid maturation state, in which partial maturation of the CA-NTD has already occurred.

Essentially all stages of the viral lifecycle have been proposed as potential targets for antiretroviral drugs (15), including assembly and cleavage of the immature virus. In the absence of structural data on the immature Gag lattice, the binding sites of assembly inhibitors relative to the protein–protein interfaces remain unknown. Indeed for inhibitors such as bevirimat, the binding site exists only in the assembled immature lattice. For the CA-CTD, the structure presented here provides the best structural model to date for the arrangement of the CA-CTD and downstream regions in immature HIV-1, and, together with the M-PMV structure presented previously (11), provides a framework for understanding how inhibitor-binding sites relate to the positions of interfaces important in virus assembly. Mutagenesis analysis has shown the CA-CTD regions to be most critical in mediating virus assembly, and the interfaces they form during assembly include the most conserved residues across retroviruses. The observed similarity in CA-CTD:CA-CTD interactions between M-PMV and HIV makes it reasonable to assume that these interactions will be conserved across retroviruses.

Materials and Methods

Protein Purification and in Vitro Assembly.

HIV-1 CA-CTD and CANC proteins were expressed and purified from Escherichia coli as described previously (22, 26). Assembly of purified proteins into tubes was induced as described previously (22). Immature HIV-1 particles (for radial-density profiles) were produced by transfection of 293T cells with the HIV-1 proviral plasmid pNLC4-3 in the presence of 5µM amprenavir followed by purification as described previously (40). (SI Materials and Methods).

X-Ray Structure Determination of Y169S and Y169L CA-CTD Proteins.

Crystallization, data collection, structure refinement, and model building of CA-CTD Y169S and CA-CTD Y169L proteins were performed as described previously (26) (SI Materials and Methods). PDB accession codes are 4COC for Y169L and 4COP for Y169S.

Cryo-EM and Image Processing.

Cryo-EM sample preparation and data collection was performed as described previously (11). The helical parameters of each tube were extracted using cryo-ET and subtomogram averaging (10, 11, 41). Real-space helical reconstruction was conducted by using extracted helical parameters (42). The pseudohexameric asymmetric units from all reconstructed tubes were averaged to obtain a final reconstruction of the HIV-1 CANC Y169S/L tubes at 9.4 Å (SI Materials and Methods). The structure and fit are deposited under EMDB accession number EMD-2638, and PDB accession code 4D1K.

Atomic Structure Fitting and Analysis.

Atomic structures of retroviral CA domains were fitted into the cryo-EM densities by using MDFF (30). Analysis of final obtained PDB files was done by using the UCSF Chimera package (43). (SI Materials and Methods).