Assembly Properties of the Human Immunodeficiency Virus Type 1 CA Protein

During retroviral maturation, the CA protein oligomerizes toform a closed capsid that surrounds the viral genome. We havepreviously identified a series of deleterious surface mutationswithin human immunodeficiency virus type 1 (HIV-1) CA that alterinfectivity, replication, and assembly in vivo. For this study,27 recombinant CA proteins harboring 34 different mutationswere tested for the ability to assemble into helical cylindersin vitro. These cylinders are composed of CA hexamers and arestructural models for the mature viral capsid. Mutations thatdiminished CA assembly clustered within helices 1 and 2 in theN-terminal domain of CA and within the crystallographicallydefined dimer interface in the CA C-terminal domain. These mutationsdemonstrate the importance of these regions for CA cylinderproduction and, by analogy, mature capsid assembly. One CA mutant(R18A) assembled into cylinders, cones, and spheres. We suggestthat these capsid shapes occur because the R18A mutation altersthe frequency at which pentamers are incorporated into the hexagonallattice. The fact that a single CA protein can simultaneouslyform all three known retroviral capsid morphologies supportsthe idea that these structures are organized on similar latticesand differ only in the distribution of 12 pentamers that allowthem to close. In further support of this model, we demonstratethat the considerable morphological variation seen for conicalHIV-1 capsids can be recapitulated in idealized capsid modelsby altering the distribution of pentamers.

INTRODUCTION

Top

Introduction

Retroviruses initially assemble as noninfectious spherical particlesthat are organized by the viral structural protein Gag (17,23). Upon budding, these immature virions rearrange dramaticallyto form mature infectious viral particles (31). Viral maturationis triggered by the proteolytic cleavage of Gag into three newproteins, namely, MA, CA, and NC. As the virus matures, MA remainsassociated with the plasma membrane, NC associates with thedimeric RNA genome, and CA condenses to form a closed capsidshell that surrounds the NC-RNA complex.

Although the immature capsids of all retroviruses are spherical,the shapes of mature retroviral capsids vary significantly betweengenera (see Fig. 1) (51). Specifically, the capsids of retrovirusesfrom most genera are roughly spherical, whereas the capsidsof beta- and type D retroviruses are cylindrical (e.g., Mason-Pfizermonkey virus [M-PMV]) and the capsids of lentiviruses are conical(e.g., human immunodeficiency virus [HIV]) (4, 55). Despitethese large variations in capsid morphology, however, the tertiarystructures of all retroviral CA proteins are similar (9), suggestingthat the different capsid morphologies may be created by theutilization of common underlying principles (21, 35).

View larger version (67K):
[in this window]
[in a new window]
FIG. 1. Variation in the mature capsid morphologies of retroviruses from different genera. EM images and idealized computer models of the mature capsids of a gammaretrovirus (Moloney murine leukemia virus [M-MuLV]), a betaretrovirus (M-PMV), and a lentivirus (HIV-1) are shown. Note that in all cases, the structures are assembled from hexagonal lattices and are allowed to close by the introduction of 12 pentameric defects (red).

Retroviral CA proteins contain two structural domains, the N-terminalassembly domain (NTD) and the C-terminal dimerization domain(CTD). The NTD is a tapered domain composed of seven ${alpha}$ -helicesand an N-terminal ß-hairpin (5, 9, 10, 12, 22, 26,28, 29, 40, 47). Biochemical and genetic analyses have demonstratedthat this domain is especially important for mature capsid assemblyand for subsequent early steps in the replication cycle (3,6, 13, 43, 52, 53). Nevertheless, several NTD mutants have beenidentified that are also defective in immature capsid assembly,suggesting a role for the domain in this process as well (36,43, 53).

The CTD is a roughly spherical domain composed of four ${alpha}$ -helicesand a 3₁₀ helix (5, 9, 19, 26, 28, 29, 56). Unlike the NTD,most deleterious mutations in the CTD impair the ability ofGag to assemble into immature particles (13, 43, 52, 53). TheHIV type 1 (HIV-1) CA CTD dimerizes in solution and is alsocrystallized as a dimer (19, 44, 56). Mutational analyses haverevealed that this crystallographic CTD dimer interface is importantfor infectivity and assembly (18, 52, 53, 56). In spite of theirstructural similarities, however, not all retroviral CA proteinsform stable dimers in solution. One example is the equine infectiousanemia virus (EIAV) CA protein, which is monomeric in solutionbut has been crystallized as a dimer. Interestingly, the twoCTDs in the EIAV CA dimer associate via the same surface asthe HIV CA CTD dimer, but with different orientations (26).

At high salt concentrations, HIV-1 CA assembles into hollowcylinders (16, 21, 25, 35). Cryo-electron microscopic reconstructionshave revealed that the cylinders are composed of helical arraysof CA hexamers (35). Similar CA hexamers were also observedin Fourier-filtered images of cylindrical cores isolated fromHIV-1 virions, indicating that CA forms similar assemblies invitro and in vivo (8). Moreover, electron diffraction patternsfrom cylindrical and conical viral capsids share common reflections,indicating that both cones and cylinders utilize the same extendedCA lattice, and therefore, that the cylinders formed in vitroare reasonable models for mature capsid assembly (8). The unusualshape of the CA NTD allowed it to be fit into the reconstructedelectron density, resulting in a pseudoatomic structural modelfor the CA lattice (35). In this model, the hexameric ringsare formed by six NTDs and are linked to adjacent rings viaCTD dimers. Helices 1 and 2 of each NTD appear to form a 12-helixbundle that holds the hexamer together. However, this structuremust still be considered a working model owing to the moderateresolution of the reconstruction ( $~$ 20 Å). Indeed, two othergroups have recently proposed variations on this basic model(32, 38).

Based upon the hexameric lattice observed in the CA cylinderreconstructions, together with geometric analyses of syntheticHIV-1 cores, we have proposed that retroviral capsids are organizedon hexameric lattices and that the incorporation of 12 CA pentamersallows each lattice to close. The relative positioning of thepentamers in the hexameric lattice defines the overall shapeof the mature capsid (21) (Fig. 1). In support of this model,CA proteins from Rous sarcoma virus (39) and Moloney murineleukemia virus (20) have also been shown to form hexameric latticesthat are similar to those formed by the HIV-1 CA protein.

The ambiguities inherent in fitting crystal structures to moderate-resolutiondensity maps make it essential to test putative CA-CA interfacesby using complementary approaches. Members of our laboratoryrecently reported an extensive alanine-scanning mutational analysisof HIV-1 CA functioning in the context of a cultured virus (53).That study defined four surfaces that are important for HIV-1assembly, infectivity, and replication. However, the complexityof viral replication made it difficult to determine preciselyhow and when these different surfaces function. For this study,we used a simple CA assembly assay to define which residuesare important for the formation of the mature viral capsid structure.Specifically, purified mutant CA proteins that harbored thedetrimental mutations defined in previous genetic assays (53)were tested for cylinder formation. We also performed computermodeling studies of retroviral capsids to explain the considerablesize and shape variation observed for authentic HIV-1 capsids.

MATERIALS AND METHODS

Top

Materials and Methods

Plasmid construction. HIV-1_NL4-3 CA mutants previously tested in viral replicationassays (53) were transferred from the proviral R9 ${Delta}$ ApaI plasmidto the pET11a expression vector by PCR amplification and standardcloning techniques. The PCR primers were designed to introducean NdeI site and a start codon at the 5' end and a BamHI siteand a stop codon at the 3' end of the CA gene. The four mutantspresented in this paper that were not previously built intoproviral constructs (M39A, R18A, N21A, and ${Delta}$ 87-97) were madeby using PCR-based megaprimer mutagenesis (42). The completeopen reading frame of every CA mutant was confirmed by DNA sequencing.

Protein expression and purification. All CA proteins were expressed and purified as previously describedfor the wild-type protein (52). Purified CA proteins were dialyzedovernight against 20 mM Tris (pH 8.0)-5 mM ß-mercaptoethanol.After dialysis, the mutant proteins were concentrated to $~$ 30mg/ml, divided into aliquots, flash frozen in liquid N₂, andstored at -70°C prior to analysis. The expected mass ofeach mutant protein was confirmed by electrospray mass spectrometry(except for E75A/E76A, which precipitated when dialyzed intowater).

Assembly and EM analyses of recombinant CA proteins. Frozen aliquots of each CA protein were thawed and requantitatedprior to use. CA assembly reaction mixtures were made in 20-µlvolumes (16 h at 4°C; 50 mM Tris [pH 8.0], 1 M NaCl, 5 mMß-mercaptoethanol). For electron microscopy (EM) analyses,carbon-coated grids were placed on 10-µl aliquots of eachassembly reaction for 90 s, washed with 3 to 4 drops of 0.1M KCl, stained with 3 or 4 drops of 4% uranyl acetate, and airdried. Samples were analyzed on Hitachi 7100 and Philips Tecnai12 transmission electron microscopes at magnifications betweenx2,500 and x50,000. Assembly reactions were analyzed at threedifferent protein concentrations (15, 10, and 5 mg/ml) for eachCA mutant. Each reaction was repeated at least three times withproteins from at least two separate purifications.

Isolation and EM analysis of HIV-1 cores. Mature virus cores were isolated as previously described (30).In brief, the virus was produced by transfection of 293T cellswith the R9 proviral HIV-1 expression construct. After 48 h,the supernatant was harvested, filtered through a 0.45-µm-pore-sizefilter, and concentrated through a 20% sucrose cushion. Subsequently,the viral envelope was stripped and the resulting viral coreswere purified by sedimenting the concentrated viral sample througha 30 to 70% sucrose cushion overlaid with layers of a 1% TritonX-100 solution in 15% sucrose and a 7.5% sucrose barrier cushion.After sedimentation, fractions containing viral cores were combinedand concentrated. For EM analyses, viral cores were appliedto glow-discharged carbon-coated grids, briefly washed withbuffer, and stained with 1% uranyl acetate. Samples were examinedin a Philips CM12 transmission electron microscope.

Molecular modeling. Models for the different retroviral capsid shapes were constructedmanually from hexameric and pentameric rings composed of sp²hybridized carbon atoms (21) by using Chem3D software (CambridgeSoftCorporation). Different shapes were explored by inserting pentamersat different positions in the hexagonal lattice. After crudemodels had been constructed manually, the bond lengths and geometriesof the hexameric and pentameric rings were regularized by energyminimization programs within the Chem3D software. The minimizationprotocol was demonstrated in histograms, which showed narrowranges of bond lengths in the structure.

RESULTS AND DISCUSSION

Top

Results and Discussion

Protein purification, solubility, and assembly. An ensemble of 27 mutant HIV-1_NL4-3 CA proteins was tested forin vitro assembly (Table 1). The ensemble included 18 of the21 mutations that reduced viral infectivity at least 25-foldin a previous mutagenesis study (53). The remaining three mutationswere D51A and M185A, whose in vitro assembly was described previously(52), and P224A, which resides in an unstructured C-terminalregion of CA that can be deleted without blocking assembly invitro (data not shown). The ensemble also included five controlmutations that were distributed throughout the protein and thatreduced infectivity <25-fold. We also included one mutant( ${Delta}$ 87-97), which was created by replacing CA residues 87 to 97with two glycine residues, to test the importance of the cyclophilinA (CypA) binding loop for CA assembly. Twenty-six of the 27mutant CAs were soluble in Escherichia coli and were purifiedto high levels of homogeneity (Fig. 2B). An exception was CAmutant E75A/E76A, which was less soluble than the other proteinsand therefore did not purify as well (Fig. 2B).

View this table:
[in this window]
[in a new window]
TABLE 1. List of CA mutants and their assembly phenotypes in vitro and in vivo

View larger version (39K):
[in this window]
[in a new window]
FIG. 2. Mutant CA proteins. (A) Locations of the 35 mutations employed in this study mapped onto the structure of the monomeric HIV-1 CA protein. The amino-terminal ß-hairpin and helices 1 to 11 are red, orange, yellow, green, blue, indigo, magenta, violet, dark red, light purple, cyan, and green, respectively. Gray balls show positions of the Cß atoms. (B) Sodium dodecyl sulfate-polyacrylamide gel electrophoresis analysis of 26 of 27 purified mutant CA proteins. Note the following: (i) the E75A/E76A mutant was less soluble than the WT and was therefore more difficult to purify, (ii) the E28A/E29A mutant was slightly less soluble than the WT but could still be purified to homogeneity, and (iii) the K70A mutant was made late in the study and was therefore not included in this panel. MW, molecular size marker.

Assembly competence was assayed by incubating each CA proteinin a high salt concentration at three different protein concentrations(15, 10, and 5 mg/ml) and by analyzing cylinder formation intransmission electron micrographs (TEM) of negatively stainedsamples. An analysis of the assembly characteristics of thedifferent CA proteins revealed five distinct phenotypes, whichwe termed normal, enhanced, reduced, nonassembling, and altered(Table 1). Mutant CA proteins that assembled with efficienciessimilar to that of the wild type (WT) at all protein concentrationstested were defined as normal (++; 16 of 27), while mutantswhich failed to form cylinders, even at 15 mg/ml, were definedas nonassembling (-; 3 of 27). Mutant proteins with an intermediatephenotype (i.e., cylinder formation was observed at 15 mg/mlbut was severely attenuated at the two lower concentrations)were classified as reduced (+; 4 of 27), and mutants that producedlonger cylinders at higher efficiencies than WT CA were definedas enhanced (2 of 27). Mutant proteins that did not form longcylinders but instead assembled efficiently into spheres, cones,and short capped cylinders were classified as altered (2 of27) (Table 1). Examples of all five of these phenotypes areprovided in Fig. 3.

View larger version (91K):
[in this window]
[in a new window]
FIG. 3. Negatively stained EM images illustrating the five different assembly phenotypes observed for HIV-1 CA proteins (WT, wild-type CA; reduced, A22D CA; nonassembling, A42D CA; altered, R18A/N21A CA). The images in the top panels show assembly reactions performed at a protein concentration of 15 mg/ml, whereas the images in the lower panels show assembly reactions performed at a protein concentration of 5 mg/ml. Note that lower protein concentrations increased the lengths of cylinders for all CA proteins tested.

Mutations in the NTD of CA. Twenty proteins with mutations in the NTD of CA were initiallyanalyzed. Although most of these mutations adversely affectedHIV-1 replication, only four significantly reduced or blockedcylinder assembly in vitro (Fig. 4A and Table 1). Notably, aseries of mutations in helices 5, 6, and 7 that diminished HIV-1replication in culture did not affect CA cylinder formationin vitro. These results are consistent with the idea that thisregion of CA is important for Gag assembly and particle productionbut does not play a direct role in the assembly of the matureviral capsid (53). In contrast, all four of the CA NTD mutationsin our initial screen that exhibited detrimental effects onCA assembly in vitro mapped to helix 1 or 2. Helix 2 harboredthe only two nonassembling mutations in the NTD (M39A and A42D),and helix 1 harbored two mutations that reduced, but did noteliminate, CA assembly (A22D and E28A/E29A).

View larger version (22K):
[in this window]
[in a new window]
FIG. 4. Locations and phenotypes of the different CA mutants described in this study. (A) Structure of the HIV-1 CA protein. (B) Structural model of the hexameric rings formed by the CA NTD. (C) Crystallographic model of the dimer formed by the CA CTD. Color code: green, WT assembly phenotype; red, nonassembling phenotype; yellow, reduced phenotype; purple, altered phenotype; blue, enhanced phenotype. The E75A/E76A mutant protein (denoted by x) was less soluble than the WT.

Helix 1 also harbored one mutant with an altered CA assemblyphenotype (R18A/N21A). This mutant CA protein assembled veryefficiently, but instead of forming long cylinders it producedlarge numbers of small spheres, of $~$ 30 to 50 nm in diameter,as well as cones, spirals, and short capped cylinders (Fig.3). To test which of the two residue changes in the double mutantwere responsible for this phenotype, we also created the R18Aand N21A mutations separately. CA proteins harboring only theR18A mutation were altered in assembly, demonstrating that thisresidue was responsible for the phenotype (data not shown).

Outside of those in helices 1 and 2, the only other mutationin the CA NTD that affected cylinder formation was the deletionmutation ( ${Delta}$ 87-97), which enhanced cylinder production. This mutationremoves the extended loop located between N-terminal helices4 and 5 which acts to package cyclophilin A, a cellular prolylisomerase, into the virion (36, 48).

Implications for pseudoatomic models of the CA hexamer. There is general agreement that the hexameric rings observedin reconstructed CA assemblies are formed by six copies of theCA NTD (35, 38). It was previously proposed that the CA NTDcan be fit into the reconstructed density from helical CA cylinderssuch that the six NTD molecules associate via interactions betweenhelices 1 and 2 at the interior of the hexamer (Fig. 4B) (35).The results of our mutational analyses agree well with thismodel, as all of the detrimental CA NTD assembly mutations observedin our studies mapped to helix 1 or 2. Helices 1 and 2 are alsoprotected from deuterium exchange upon CA cylinder assembly,which again suggests that they likely make important contactsin the assembled CA hexamer (32). Note, however, that not allmutations in helices 1 and 2 reduced CA cylinder assembly. Inparticular, the N21A (helix 1) and P38A (helix 2) mutant proteins,which are both adjacent in the primary sequence to detrimentalmutations, assembled with normal efficiencies. This result maydemonstrate the specificity of the CA assembly interface(s),and it emphasizes the importance of utilizing extensive mutagenesisscreens when attempting to define protein-protein interactionsurfaces.

Two groups have recently proposed variations on the CA hexamermodel shown in Fig. 4B, and it is therefore of interest to considerthese alternative models in light of our mutational analyses.In one alternative model for the CA hexamer, the NTDs are rotated $~$ 180° about axes that are parallel to the six-fold axis ofthe hexamer (38). As a result, helices 1, 2, and 7 from eachCA NTD subunit face outward, and the cyclophilin A binding loopsface toward the center of the ring (where they would presumablymake intersubunit contacts that would help stabilize the hexamer).Several of our observations appear to be inconsistent with thismodel. Firstly, we have now shown that CA can form helical cylinderseven when the entire cyclophilin binding loop is deleted (Fig.2). Secondly, it is not obvious how this model can be reconciledwith the importance of helices 1 and 2 for CA assembly. Onepossibility, suggested by Mayo et al., is that helix 2 mighthave an interhexamer interaction with helix 7 from a subunitof a different hexamer (38). To support this idea, they notedthat an intermolecular four-helix bundle formed between helices2 and 7 of adjacent CA subunits was observed in the crystallattice of EIAV CA (26). However, in the crystallographic EIAVCA interaction, the two copies of helix 7 pack in an antiparallelorientation, whereas in the interaction proposed for HIV CAthey would pack in a parallel orientation (which would thereforerequire a very different interaction). In any event, mutationsin helix 7 generally do not block capsid assembly in vivo (53)or CA assembly in vitro (this work), and CA helix 7 is not protectedfrom deuterium exchange in helical assembles of CA (32).

A second recent study has proposed that the NTD makes intermolecularcontacts with the CTD of an adjacent molecule within the assembledlattice of CA hexamers. This proposal is based upon data fromchemical cross-linking and deuterium exchange protection (32)and is consistent with assembly inhibitor and second-site suppressorstudies (7, 33, 46). Specifically, Lanman et al. have shownthat upon cylinder formation, the C-terminal end of helix 3and the N-terminal end of helix 4 are protected from solventexchange, and that Lys-70 (helix 4, in the NTD) of one CA subunitcan be efficiently cross-linked to Lys-182 (helix 9, in theCTD) in a second subunit (32). In contrast, none of our originalmutations in this region adversely affected cylinder formation(T54A/N57A [helix 3], Q63A/Q67A [helix 4], and E75A/E76A [helix4]). Our initial experiments did not test the effects of theK70A mutation because this mutation did not reduce infectivitybelow our cutoff threshold (21-fold versus the threshold of25-fold). To help with our testing of the validity of the NTD-CTDinterface, however, we subsequently prepared the CA K70A mutant(53) and found that it exhibited an intermediate assembly phenotype.Specifically, the mutation reduced CA assembly in the cylinderformation assay, but short cylinders were still formed at CAconcentrations of 15 mg/ml (data not shown). Similarly, theK70A mutation reduced, but did not eliminate, conical capsidformation in vivo (53). Thus, the phenotype of the K70A mutationis consistent with the involvement of this residue in a protein-proteininterface, although mutations in this region of CA were generallynot as detrimental as mutations elsewhere in the protein.

Implications for retroviral capsid morphology. As described above, we have proposed that all retroviral capsidsmay be organized on hexameric lattices, with the positions ofpentameric "defects" defining the overall capsid morphology.The assembly phenotype of R18A CA is consistent with this unifyingmodel for capsid assembly, since this single protein can simultaneouslyform all of the known retroviral capsid structures (spheres,cones, and capped cylinders). Moreover, many of the structuresformed by R18A CA resemble authentic retroviral capsids in bothshape and size (Fig. 2).

One possible explanation for the phenotype of CA R18A is thatthe different capsid structures arise from an increased frequencyof pentamer incorporation into the assembling hexameric lattice.Precedents for this idea are provided by a number of mutanticosahedral viruses that exhibit similar phenotypes (11, 14,27, 37, 50). For example, point mutations in the structuralprotein of the T=3 icosahedral ilarvirus cause the random insertionof pentamers into the expanding lattice, and the resulting capsidsexhibit highly variable diameters, similar to the spheres producedby the R18A mutant protein (50). Similarly, point mutationsin the VP1 protein of polyomavirus cause the assembly of spirals(34, 45), which are also seen in R18A CA assembly reactions(not shown). Spirals are created when pentamers are improperlyincorporated into an assembling hexagonal lattice, preventingclosure of the structure (14, 27, 37). Finally, a single pointmutant in the coat protein of alfalfa mosaic virus can resultin a complete shift from spheres to cylinders in in vitro assemblyreactions (11).

Importance of the cyclophilin A binding loop. Previous studies have demonstrated that the addition of recombinantcyclophilin A to assembly reactions increases cylinder length(i.e., gives an enhanced phenotype) (24). Moreover, two pointmutations in the cyclophilin A binding loop, G94D and A92E,which were originally discovered as cyclophilin-independentviral escape mutations (1, 57), can also enhance CA cylinderassembly in vitro (35). Taken together, these observations areconsistent with the idea that the cyclophilin A binding loopactually inhibits CA cylinder assembly, suggesting that at leastin vitro, cyclophilin A may increase the efficiency of CA assemblyby suppressing off-pathway CA aggregation caused by the large,hydrophobic cyclophilin A binding loop (24). Off-pathway aggregationcan alternatively be suppressed by mutations that increase loophydrophilicity and charge (e.g., G94D and A92E) or by deletionof the entire loop. Interestingly, HIV-1 CA is the only retroviralCA protein known to use an exposed loop to bind cyclophilinA, and this loop is larger in HIV-1 than in other retroviruses(9). Conceivably, the selective advantage(s) of binding cyclophilinA (49) comes at the cost of reduced CA assembly efficiency.

Mutations in the CTD of CA. The assembly phenotypes of six different CA proteins harboringmutations in the CTD were also analyzed. All of these mutationsreduced the infectivity of the cultured virus, but only threeof the six mutant proteins (W184A, K203A, and K158A) affectedcylinder assembly in vitro (Fig. 4A).

Two of the mutations that diminished cylinder assembly are locatedwithin the crystallographic CTD dimer interface (W184A and K203A).The W184A mutation caused a nonassembling phenotype and is locatedin the hydrophobic core of the crystallographic CA dimer (Fig.4C). This phenotype is similar to that of another hydrophobicdimer interface mutation (M185A) which also blocks cylinderformation in vitro (52). The K203A mutation was less severebut still clearly had a reduced assembly phenotype. K203 makesa salt bridge with D152 across the dimer interface. These resultsconfirm the importance of the crystallographic dimer interfacefor cylinder assembly and support the idea that the CA CTD servesto make dimeric linkages between adjacent CA hexamers.

A second surface located at the "bottom" of the CTD is alsoimportant for particle production and viral infectivity (46).However, the two mutations within this surface that were tested(K158A and D197A) did not diminish CA cylinder assembly, consistentwith the idea that this surface functions during immature particleformation. Indeed, the K158A mutation actually appeared to enhanceCA cylinder assembly. However, unlike the long discrete cylindersproduced by the ${Delta}$ 87-97 CA protein, the long cylinders producedby K158A aggregated into very large clumps (Fig. 3). We canprovide no obvious structural explanation for this phenotype,but we speculate that the replacement of a charged lysine surfaceresidue with an alanine may create a more hydrophobic surfacethat contributes to CA aggregation.

Molecular modeling. EM analyses of thin-sectioned mature virions and isolated HIVcapsids have revealed significant variations in capsid sizeand shape (2, 8, 30, 31, 54). In an attempt to understand whethera fullerene cone model can account for the observed variationin authentic viral cores, we used a molecular modeling programto build three different idealized models for the HIV cone.

From Euler's theorem, only five conical hexagonal lattices areallowed, which correspond to cones with mean angles of 19.2°,38.9°, 60°, 83.6°, and 112.9° (for reviews,see references 15 and 41). Structures corresponding to the narrowestof the allowed cone angles predominate in both synthetic andactual HIV-1 capsids (2, 8, 21, 54). Therefore, the three differentidealized models for HIV-1 capsids were built with conical hexagonallattices with 19.2° cone angles. The cones were closed byincorporating seven pentamers into the wide end and five pentamersinto the narrow end, and the precise positions of the pentamerswere varied between the different models (21). The models werebuilt with 1,000 to 1,500 CA molecules/capsid (assuming eachhexamer represents six CA subunits and each pentamer representsfive CA subunits), which was judged to be a reasonable approximationfor the viral core, as recent reports have suggested that 4,000to 5,000 copies of Gag are present in each virion (V. Vogt,personal communication), and multiple conical capsids are oftenseen in single virions (8).

Six different views of these three idealized cones are shownin Fig. 5, and these images illustrate how the apparent shapeof the capsid can be altered by changing the pentamer placementsand viewing orientations. This is because (i) the inherent asymmetryof the lattice means that cones have different shapes when viewedfrom different orientations and (ii) introducing pentamers intodifferent positions in the hexameric lattice causes one coneto differ from another. As shown in Fig. 5, EM images of bonafide HIV-1 capsids also often show significant variations incapsid shape (e.g., rounded caps versus pointed caps with abruptedges and angles, differences in apparent cone angles, etc.)(54). Similar variations were also apparent in our computermodels when they were viewed from different angles or constructedwith different pentamer distributions. For example, clusteringof the pentameric defects within the lattice increases the localdeclination and causes the capsid to appear more pointed. Thisprinciple is nicely illustrated by the model in Fig. 5B, inwhich an acute cap angle (marked by an arrow) arose becausetwo pentamers were placed in adjacent positions. Conversely,more dispersion of the pentameric defects resulted in roundercaps. Similarly, the observed cone morphology and apparent coneangle can also vary dramatically depending on the orientationfrom which a single model is viewed. This can be seen by comparingthe models in Fig. 5D and E, which show two different viewsof the same cone. Overall, we conclude that the size and shapevariations seen for authentic HIV-1 capsids can be mimickedin idealized cones, providing strong support for the idea thatlentiviral capsids are assembled on the principles of fullerenecones (21).

View larger version (78K):
[in this window]
[in a new window]
FIG. 5. Comparisons of bona fide HIV-1 capsids with idealized models. Upper panels show negatively stained EM images of HIV-1 cores purified from detergent-treated virions. Lower panels show computer-generated images of idealized conical capsids assembled from mixtures of hexamers and pentamers. The location of a single pentamer in the cone shown in panel B is highlighted by an arrow to illustrate how pentameric defects give rise to sites of declination in the hexagonal lattice. The six images represent different views of three different computer-generated conical capsids, each of which has five pentamers capping the narrow end and seven pentamers capping the wide end.

Conclusions. Our studies have defined two CA surfaces that are importantfor the in vitro assembly of CA cylinders and, by extension,the mature HIV-1 capsid. Specifically, we have shown that helices1 and 2 in the NTD and the crystallographic CTD dimer interfaceare required for efficient CA assembly. Our data also allowfor the possibility that there is a third CA-CA interface betweenthe NTD and the CTD (32), as a single mutation (K70A) in helix4 also diminished CA assembly. These data are generally in goodagreement with deuterium exchange mapping experiments, whichimplicate these same surfaces in homo-oligomeric interactionsin the CA cylinders (32). Mutations in these interfaces alsoreduced viral infectivity, supporting the idea that in vitroCA cylinder assembly accurately mimics some (though clearlynot all) aspects of mature capsid assembly (53).

We have also demonstrated that a single point mutation, R18A,allows the HIV-1 CA protein to assemble into cylinders, spheres,and cones in the same reaction mixture. As these structuresmatch the three known retroviral capsid morphologies, this observationprovides support for the idea that all retroviral capsids canbe generated by using similar underlying lattice structures.

Finally, we have used molecular modeling experiments to demonstratethat computer models built with hexamer-pentamer lattices canmimic the considerable shape variation observed for actual HIV-1capsids, which emphasizes the fact that although these objectsare assembled on hexagonal nets using repeating protein-proteininterfaces, they are nevertheless fundamentally asymmetric structures.

ACKNOWLEDGMENTS

We thank Alex Kotov and Paula Flicker for HIV-1 core preparationand imaging, Donovan Anderson for preparing and testing assemblyof the CA K70A mutant protein, and Nancy Chandler and Kurt Albertinefor electron microscopy support in Utah.

This work was funded by an NIH grant to W.I.S. and an NIH predoctoraltraining grant fellowship to B.K.G.-P.

FOOTNOTES

* Corresponding author. Mailing address: Department of Biochemistry, University of Utah, 20 N 1900 E Rm. 211, Salt Lake City, UT 84132. Phone: (801) 585-5402. Fax: (801) 581-7959. E-mail: wes{at}biochem.utah.edu.

Present address: Department of Cell Biology, The Scripps ResearchInstitute, La Jolla, CA 92037.

REFERENCES

Top