The structure and flexibility of conical HIV-1 capsids determined within intact virions

See allHide authors and affiliations

Science  16 Dec 2016:
Vol. 354, Issue 6318, pp. 1434-1437
DOI: 10.1126/science.aah4972

Structural insights into capsid flexibility

Viral capsids are protein structures that enclose the genetic material of viruses. Previous structural studies of the HIV-1 capsid have relied on recombinant, cross-linked, or mutant capsid proteins. Mattei et al. now report subnanometer-resolution cryo–electron tomography structures of the HIV-1 capsid from intact virions. These structures confirm the hollow cone shape of the capsid and allow for the specific placement of each individual capsid hexamer and pentamer within the lattice structure. The structures also reveal the flexible nature of the capsid, which likely helps it to accommodate interactions with host cell factors.

Science, this issue p. 1434


HIV-1 contains a cone-shaped capsid encasing the viral genome. This capsid is thought to follow fullerene geometry—a curved hexameric lattice of the capsid protein, CA, closed by incorporating 12 CA pentamers. Current models for core structure are based on crystallography of hexameric and cross-linked pentameric CA, electron microscopy of tubular CA arrays, and simulations. Here, we report subnanometer-resolution cryo–electron tomography structures of hexameric and pentameric CA within intact HIV-1 particles. Whereas the hexamer structure is compatible with crystallography studies, the pentamer forms using different interfaces. Determining multiple structures revealed how CA flexes to form the variably curved core shell. We show that HIV-1 CA assembles both aberrant and perfect fullerene cones, supporting models in which conical cores assemble de novo after maturation.

HIV-1 is initially produced as an immature, noninfectious particle with the main structural protein Gag forming a membrane-associated protein layer. Cleavage of Gag separates individual domains, leading to morphological conversion into the mature, infectious virion. The cleaved CA protein, consisting of two α-helical domains connected by a flexible linker, assembles into a conical core encasing the viral genome (14). The core has been suggested to constitute a fullerene cone: a lattice of CA hexamers that flexes to accommodate different local curvatures on the conical surface, and which is closed by insertion of five pentamers toward the narrow end of the cone and seven toward the broad end (5). The different core morphologies seen in a minority of HIV-1 particles and in other retroviruses can be modeled as fullerene structures in which different distributions of the pentamers determine the overall shape (5, 6).

The structure of the mature hexameric assembly of CA has been studied by using recombinant CA assembled into tubular arrays for cryo–electron microscopy (cryo-EM) (7, 8) and by x-ray crystallography of mutant, cross-linked CA (9) or wild-type CA in a hexameric lattice (10). The 10-nm hexamer-hexamer spacing in these arrays matches that detected in cores purified from intact viruses (11). The mature CA hexamer consists of a central ring formed by the N-terminal domains of CA (NTD), and an external ring formed by the C-terminal domains of CA (CTD) that links hexamers together. The x-ray structure of a CA pentamer, stabilized by targeted insertion of cysteine residues, was found to be quasi-equivalent to the hexamer (12). Models for the core have been generated by arranging hexamers and pentamers into cones resembling those seen in vivo and subjecting these arrangements to coarse-grained (13) and all-atom molecular dynamics (MD) simulations (8). Thus, currently available structural information about hexamers, pentamers, and cores derives from the analysis of crystals, in vitro assemblies, and in silico simulations.

Here, we applied cryo–electron tomography (cryo-ET) and subtomogram averaging to determine the structure, arrangement, and flexibility of CA within intact, wild-type HIV-1. Purified and inactivated HIV-1 was vitrified and imaged by using a dose-symmetric tilt scheme (14) (table S1 and Fig. 1A). From reconstructed tomograms of 539 particles, 552 cores were identified and subtomograms were extracted along the surface of the cores. The subtomograms were subjected to a reference-free, iterative alignment and averaging procedure (15). From the aligned, motion-corrected, dose-filtered, and contrast-transfer function–corrected subtomograms, we generated a final, sixfold symmetric structure of the hexameric CA lattice within the virion core at 6.8 Å resolution (Fig. 1, B and C, movie S1, and fig. S1).

Fig. 1 The structure of the CA hexamer.

(A) Computational slice through a representative tomographic reconstruction of HIV-1 virions. Scale bar, 100 nm. Inset shows a computational slice through the top of the core in the boxed virion, indicating that individual hexamers can be resolved. (B) Cryo-ET reconstruction of the mature CA hexameric lattice (gray isosurface) viewed from outside the core. The final structural model of the hexamer is shown, with the NTD in cyan and the CTD in orange. One monomer is highlighted in blue (NTD) and red (CTD). (C) As in (B), viewed perpendicular to the lattice. See also movie S1.

We compared our reconstruction to the available crystal structures of hexameric CA (9, 10, 16) and found the fully hydrated wild-type hexameric crystal structure [Protein Data Bank (PDB) 4XFX] (10) to be the closest representation of the hexameric CA lattice in intact HIV-1. CA monomers taken from PDB 4XFX were flexibly fitted into each of the CA densities in the EM map (Fig. 1, B and C). The fitted structure preserves the tertiary domain structures seen in the crystal, whereas the relative positions of the domains have undergone a slight rotation around the linker connecting NTD and CTD to adapt from the flat crystal lattice to the curved in-virus lattice (fig. S2A). The presence of additional density within the pore formed by the NTDs suggests that the recently proposed binding site for deoxynucleoside triphosphates (17) is occupied in the assembled core (fig. S2, B and C).

To assemble a fullerene cone, the CA lattice must bend to adopt different local curvatures—more tightly curved at the tip than at the base of the cone, and curved in one direction, but not in the other, on the side of the cone. CA has previously been suggested to achieve this by modulating the relative orientation of its two domains (9, 10), by motion around the CTD dimer interface (8), or by variability in the trimeric interface formed by the CTD (18). To determine how CA flexes in the viral cone, we calculated the tilt and the twist between all pairs of neighboring hexamers in our data set (Fig. 2A). We then classified the pairs of hexamers according to the measured angles (fig. S3) and determined the structure of each class. In this way, we determined the structures, at subnanometer resolution, of 18 different conformations with different tilt and twist angles (table S2, fig. S3, and movies S2 and S3). Into each EM map, we rigid-body fitted the NTDs and CTDs that had resulted from the flexible fit into the symmetrized hexamer. We found that tilt and twist are accommodated by movements of the NTD relative to the CTD around the flexible linker (Fig. 2B), as well as by relative rotations around the CTD dimer and trimer interfaces (Fig. 2, C and D). In all cases, the main range of these movements was considerably smaller than previously suggested (8, 10, 18) (fig. S4). We conclude that variable curvature is achieved by small structural movements in CA that are distributed over multiple CA-CA interfaces.

Fig. 2 Structural flexibility in the CA hexameric lattice.

(A) The distribution of tilt (left column) and twist (right column) angles between hexamers within a single core. The tilt/twist angle is indicated by the color of the connecting lines between hexamer positions from blue (low/negative values) to red (high/positive values). Insets show a schematic illustration of tilt and twist angles. (B to D) The flexibility in the CA lattice that accommodates different tilts and twists is illustrated by superimposing the CA domains from structures determined with tilt angles from –1 ± 3° to 29 ± 3°, and twist angles from –12 ± 3° to 12 ± 3°, color-coded as in (A), for (B) the monomer, (C) the CTD dimeric interface, and (D) the trimeric CTD trimeric interface. See also movies S2 and S3.

Forming a fullerene cone requires that the capsid incorporates 12 pentamers or pentameric defects. We analyzed the positions and arrangements of hexamers to identify the position and the orientation of pentamerically coordinated positions in the lattice. We extracted subtomograms at these positions and averaged them directly to generate a low-resolution structure in which a CA pentamer was clearly visible (fig. S5A), demonstrating that pentameric CA is present in the HIV-1 core. After further iterative alignment and averaging, we obtained a structure of the pentamer within HIV-1 at 8.8 Å resolution (Fig. 3, A and B, movie S4, and fig. S1). We placed CA monomers from our final hexameric CA structure into the pentameric EM map and flexibly fit them into the map to produce a structural model for the pentamer (Fig. 3, A to C).

Fig. 3 The structure of the CA pentamer.

(A) Cryo-ET reconstruction of the CA pentamer (gray isosurface) viewed from outside the core. The final structural model of the core pentamer is shown with the NTD in cyan and the CTD in orange. One monomer is highlighted in blue (NTD) and red (CTD). (B) As in (A), viewed perpendicular to the lattice. See also movie S4. (C) Our structural model of the pentamer, for comparison to (D) the crystal structure of the pentamer (PDB 3P05). (E) Comparison of the NTD-CTD interface in the pentamer (cyan/red), and in the pentamer crystal structure (magenta). (F) Two adjacent monomers within our in-virus pentamer, with one shown as a surface view. To illustrate how the NTD-CTD interface is opened up in the pentamer compared to the hexamer, the residues involved in the CPSF6 binding interface are colored in gray. (G) As in (F) for the two adjacent monomers in the cross-linked pentameric crystal structure. (H) As in (F) for two adjacent monomers in our hexamer. See also movie S5.

Notably, the structure differs appreciably from the available crystal structure of in vitro cross-linked pentameric CA (12). The NTDs, organized around the fivefold axis, have undergone a rotation of ~19° relative to the hexameric conformation that excludes helix 3 from the interprotomer interface, resulting in a central 10-helix bundle where helix 1 interacts with helix 2 in the neighboring protomer (Fig. 3, C and D). Despite this difference, as in the crystal structure, the five arginine 18 residues are in close proximity in the center of the bundle, consistent with a proposed role for the charge state of arginine 18 in regulating the hexamer to pentamer transition (12). The relative positions of the CA domains have also changed, leading to a different NTD-CTD interface in the in-virus pentamer than in the cross-linked pentameric CA and the in-virus hexamer (Fig. 3E and movie S5). In the in-virus pentamer, the loop between helices 3 and 4 sits above helix 8 of the neighboring CTD, whereas the C-terminal part of helix 7 and a few residues from the NTD-CTD linker interact with helix 11 of the neighboring CTD. As a result, the pocket that exists in the hexamer between the two CA domains and that is the binding site for cellular host factors is opened out, and the NTD face of the pocket is exposed (Fig. 3, F to H).

By solving the pentameric structure, we also visualized the interactions between the pentamer and the neighboring hexamers. We found that the pentamer-hexamer CA-CTD dimer interface is similar to that in the interhexamer interface, but that to accommodate the higher local curvature, the three helices (helix 10) around the threefold interface move closer together than in any of the hexamer-hexamer interfaces (fig. S4F).

For a subset of 107 cores, we analyzed the overall distribution of hexamers and pentamers identified during subtomogram averaging and applied a further multireference alignment step to minimize the number of false-negatives (Fig. 4 and figs. S6 and S7). The cores were mostly conical, but as expected, other morphologies were present, including tubular, polyhedral, and triangular shapes. Most of the cores showed one or more local regions where the CA lattice was disrupted or absent. These imperfections did not have characteristic shape, size, or location and were observed at the tip, side, and base of the core (Fig. 4 and fig. S7). We did not observe a recurrent seam along the longitudinal axis of the cores (19). The observed imperfections may represent assembly defects, or may result from damage during virus purification. In six cores, we observed complete, hexameric lattices closed by the insertion of 12 pentamers—they were perfect fullerene structures (Fig. 4, movie S6, and fig. S7) containing between 1122 and 1314 copies of CA. As predicted, the shape of the cores varied according to the distribution of pentamers. HIV CA is therefore able to assemble perfect, closed fullerene cones within virus particles. Although it is currently unclear whether imperfect cores can retain infectivity, we consider it reasonable to assume that correct CA assembly results in closed structures.

Fig. 4 The arrangement of hexamers and pentamers in cores.

(A) Computational slices through the tomographic reconstructions of four HIV-1 particles, two with complete conical cores (left columns), and two with incomplete cores (right columns). Other cores are illustrated in fig. S7. (B) As in (A), superimposed with the position and orientation of each aligned hexameric and pentameric unit, revealing the fullerene structure. The CTDs are colored gray; the NTDs of the hexameric units are colored according to the quality of their alignment from red (for low cross-correlation values) to green (for high cross-correlation values); the NTDs of the pentameric units are depicted in blue. (C) Three views of the cores. The number of copies of CA making up each of the illustrated cores is shown. See also movie S6.

A recent report suggests that the core is formed by rewrapping of the immature lattice in a nondiffusional manner, leading to pseudo-fullerene structures with strain defects such as seams (20). Our observations are inconsistent with this model. In most models for core assembly, the immature lattice disassembles upon maturation and the mature lattice assembles de novo from individual CA molecules or small multimers (2125). Our observations are consistent with de novo polymerization of CA in a reversible manner that allows relaxation of the local structure during assembly (13). This would permit formation of perfect fullerene cones with only small, distributed changes in CA hexamer structure as observed here. We found that at sites of high curvature, pentamers are incorporated. These are stabilized by different CA-CA interaction interfaces than those in the hexamer. When perfect, the resulting core is tightly closed, preventing access of cellular components to the genetic material. On the outside of the core, pentamer and hexamer expose different parts of the CA protein surface to the cytoplasm during infection, which would allow them to interact differently and specifically with host cell factors.

Supplementary Materials

Materials and Methods

Figs. S1 to S7

Tables S1 to S3

Movies S1 to S6

References (2639)

References and Notes

  1. Materials and methods are available as supplementary materials on Science Online.
  2. Acknowledgments: We thank S. Dodonova, K. Qu, F. Schur, and W.Wan for technical assistance and discussion. This study was supported by Deutsche Forschungsgemeinschaft grants BR 3635/2-1 (to J.A.G.B.) and KR 906/7-1 (to H.-G.K.). The Briggs laboratory acknowledges financial support from the European Molecular Biology Laboratory and from the Chica und Heinz Schaller Stiftung. This study was technically supported by the European Molecular Biology Laboratory IT services unit. A representative tomogram, the EM structures, and the fitted models have been deposited in the Electron Microscopy Data Bank and the PDB. The database accession numbers are listed in table S3. S.M., H.-G.K., and J.A.G.B. designed the experiments. B.G. and S.M. prepared samples. W.H.J.H. implemented tomography acquisition schemes. S.M. acquired the data. S.M. performed image processing. S.M. and J.A.G.B. analyzed the data. S.M. and J.A.G.B. wrote the manuscript with support from all authors.
View Abstract

Stay Connected to Science

Navigate This Article