X-Ray Structures of Native HIV-1 Capsid Protein Reveal Conformational Variability

Abstract

The detailed molecular interactions between Human Immunodeficiency Virus type 1 (HIV-1) capsid protein (CA) hexamers have been elusive in the context of a native protein. We report crystal structures describing novel interactions between CA monomers related by 6-fold symmetry within a hexamer (intra-hexamer) and by 3-fold and 2-fold symmetry between neighboring hexamers (inter-hexamer). These structures help elucidate how CA builds a hexagonal lattice, the foundation of the mature capsid. Lattice structure depends on an adaptable hydration layer that modulates interactions among CA molecules. Disruption of this layer by crystal dehydration treatment alters inter-hexamer interfaces and condenses CA packing, highlighting an inherent structural variability. Capsid stability changes imparted by high concentrations of CA-targeting antiviral PF74 can be explained by variations at inter-hexamer interfaces remote to the ligand binding site. Inherent structural plasticity, hydration layer rearrangement, and effector molecule binding may perturb capsid uncoating or assembly and have functional implications for the retroviral life cycle.

The mature capsid of Human Immunodeficiency Virus type 1 (HIV-1) is formed from a single multifunctional capsid protein (CA), which contains highly helical N-terminal (CA_NTD, residues 1-145) and C-terminal (CA_CTD, residues 150-231) domains connected by a flexible inter-domain region (residues 146-149) (1-5). The capsid is made from a lattice of ~250 CA hexamers closed by the insertion of 12 CA pentamers. The hexamer building block comprises an inner ring of six CA_NTDs held together by CA_NTD-CA_NTD and CA_NTD-CA_CTD contacts between adjacent CA molecules. The six CA_CTDs of CA hexamers form an outer “girdle” that is engaged in inter-hexamer interactions (1-5). CA-CA interactions are of critical importance for the structural integrity of the capsid and viral infectivity (1-3, 6-11). Following infection and fusion of the viral and cellular membranes, the capsid undergoes controlled dissasembly (or uncoating), which seems coordinated with reverse transcription (7, 9). Hence, CA is an attractive therapeutic target and several antivirals have been reported to bind at the CA_NTD or CA_CTD (12-17). Among them, PF-3450074 (PF74) has been reported to have a bimodal mechanism of action (18-21): at lower concentrations (nanomolar to ~2 μM) PF74 exerts its antiviral effect by directly competing with binding of nuclear host factors cleavage and polyadenylation specific factor 6 (CPSF6) and nucleoporin 153 (NUP153), whereas at higher concentrations (~10 μM) it blocks the uncoating, assembly, and the reverse transcription steps of the viral life cycle (18-23). Crystal structures of PF74 in complex with the CA_NTD fragment alone (CA_NTD-_PF74) (14) or more recently with crosslinked CA hexamers (CA_XL-PF74) (18, 19), have provided important insights into the structural mechanism by which PF74 binds CA. It was shown that PF74 preferentially binds the hexamer form of CA at the same general site used by CPSF6 and NUP153. However, the structural mechanism by which high concentrations of PF74 affect uncoating and assembly remains incompletely defined. Such effect of CA-targeting drugs would be relevant in a therapeutic context where antivirals are used at concentrations significantly higher than their inhibitory effective concentrations (EC₅₀).

Cryo-electron microscopy (cryo-EM) studies combined with advanced molecular modeling have provided information on the hexagonal lattice of CA and helped build informative models of the mature HIV-1 capsid (2, 3, 24). X-ray and NMR structures of isolated CA_NTD and CA_CTD domains or full-length CA monomers or dimers have provided key details for interactions at the 2-fold interface (1-5, 16, 25, 26). Structures of engineered crosslinked CA hexamers (CA_XL) or pentamers elucidated the interactions between CA subunits in the hexameric and pentameric building blocks (1, 3). Although engineered mutations improved the biophysical aggregation properties of the crosslinked hexamers they also affected key interactions at the 2-fold interfaces. Hence, despite advances, critical details regarding the CA-CA interactions that govern virus uncoating and assembly remain insufficiently described.

Here we describe the elusive crystal structure of native full-length CA in a form that reveals a complete set of inter- and intra-hexamer interactions. Additional structures illustrate how these interactions are altered by changing the hydration layer of the CA lattice or binding of a CA-targeting antiviral, highlighting a structural variability that may affect uncoating, assembly, host factor recognition, and trafficking across the capsid shell.

Native CA (CA) crystals were grown at low protein concentrations (<5 mg/ml), which may have reduced the propensity of native CA for aggregation without having to include any mutations (Fig. 1A). The protein crystallized in space group P6 with a single CA molecule per asymmetric unit (Fig. 1F, table S1). Crystallographic 6-fold symmetry generates CA hexamers, which in turn assemble the hexagonal lattice (Fig. 1B). CA subunits from neighboring hexamers are related by crystallographic 2-fold and 3-fold symmetry (Fig. 1, C and D). The structure of CA is in general agreement with the 9 Å resolution cryo-EM maps of the flattened CA hexagonal lattice (24) (fig. S1A) and tubes (2) (fig. S1B). The overall fold of CA is also in general agreement with the previously published crystal and NMR structures of full-length CA (1, 3-5, 24) and the CA_NTD and CA_CTD fragments (12, 14, 16, 25, 27, 28). The CA_NTD domain starts with a partially ordered β-hairpin, includes seven α-helices and a cyclophilin binding loop (CypA-BL) (Fig. 1A). The CA_NTD is followed by a short flexible hinge linker (1, 3-5) that leads to the CA_CTD, which comprises a 3₁₀-helix, an extended strand (major homology region) and four α-helices (H8-H11) (Fig. 1A). Key interactions between the CA_NTD and CA_CTD domains that are likely to stabilize the capsid particle are intra-hexamer CA_NTD-CA_NTD and CA_NTD-CA_CTD contacts among adjacent CA subunits around the 6-fold axis, inter-hexamer CA_CTD-CA_CTD contacts at the 2- and 3-fold axes (Fig. 1D) (1, 2) and to a lesser extent CA_NTD-CA_CTD intra-subunit contacts (Fig. 1A).

Fig. 1. Crystal structure of native CA.

(A) Secondary structure and ribbon diagram of native CA. The CA_NTD (1-145) comprises β-hairpin (1-13, brown), CypA-BL (85-93, cyan), and seven α-helices: H1 (17-30; yellow), H2 (36-43; black), H3 (49-57; purple), and H4 (63-83), H5 (101-104), H6 (111-119), and H7 (126-145) in gray. The CA_CTD (150-231) comprises 3₁₀-helix (150-152; green), and four α-helices: H8 (161-173; gray), H9 (179-192; orange), H10 (196-205; blue), and H11 (211-217; pink). The CA_NTD and CA_CTD are connected with a flexible linker (146-149). (B) Each crystallographic asymmetric unit comprises a single native CA molecule (shown in surface view representation; pink CA_NTD, red CA_CTD). Application of 6-fold crystallographic symmetry generates the CA hexagonal lattice. (C) The six CA subunits in a hexamer are related by 6-fold crystallographic symmetry (yellow hexagon); CA subunits from neighboring hexamers are related by 2-fold (orange ovals) and 3-fold (pink triangles) crystallographic symmetry, shown at the inter-hexamer interfaces. (D) Orthogonal views of three native CA hexamers colored as in A. The hexamers are stabilized by interactions at the 6-fold (brown β-hairpin, yellow H1, black H2, and purple H3), 2-fold (green 3₁₀ and orange H9), and 3-fold interfaces (blue H10 and pink H11). (E) Inter-hexamer interactions at the 2-fold interface. Interpretable electron density is now observed for all residues at the 2-fold interface of CA (2.4 Å; 2Fo–Fc; σ=1.2) including residues 176-187, which were previously disordered in crosslinked hexamer structures. (F) Summary of X-ray data.

The inter-hexamer interactions at the crystallographic 2-fold interface are clearly defined in the CA structure through interpretable electron density (Fig. 1E). The observed interactions involve multiple residues, including S149, L151, S178, E180, V1181, W184, M185, L189, and Q192, as well as water molecules (Fig. 2A). These contacts are reminiscent of, but different than those in crystal or NMR structures of isolated CA_CTD domains (fig. S2A, table S2) (16, 25). They are also significantly different than those in the original crosslinked hexameric (CA_XL, PDB ID: 3H47) (fig. S2B, table S2) and recent related CA structures that contain W184A and M185A mutations in helix H9 at the dimerization interface (1, 18, 19).

Fig. 2. Inter-hexamer interactions at the 2-fold and 3-fold interfaces.

Stereo views of CA_CTD regions that are related by 2-fold (A and C) or 3-fold symmetry (B and D) from CA (A and B), and dCA (C and D). The 2-fold interface comprises helices H9 and 3₁₀ in A and C. The 3-fold interface comprises helices H10 and H11 in B and D. Ordered water molecules (spheres) at the 2-fold (A) and 3-fold (B) interfaces of CA are modeled in 2.4 Å simulated annealing omit Fo–Fc electron density maps at σ=2.5 (green mesh). B-factors of refined waters were between 30-50 Ų and matched well with those of interacting atoms from S149, E175, Q176, W184, I201 and A204. Helices are shown in cartoon representation; black dashed lines between residues (in sticks) indicate that they are within ~4 Å. No water molecules are present at the 2-fold and 3-fold of dCA (C and D).

Inter-hexamer interactions between pairs of helices H10 and H11 at the 3-fold have been reported in cryo-EM studies enhanced with molecular modeling (2). The higher resolution of the CA structure now reveals the details of these contacts (Fig. 2B, table S3). Previous studies proposed that the inter-hexamer interactions are hydrophobic in nature (2, 25). However, the CA structure highlights an additional hydrophilic component that is based on water-mediated interactions at both the 2-fold and 3-fold interfaces (Fig. 2, A and B). These water molecules engage in hydrogen bond interactions with either the side chains of highly conserved residues (S149, E175, W184 at the 2-fold), or main chain carbonyl groups (Q176 at the 2-fold; I201, A204 at the 3-fold) (Fig. 2, A and B). As a result, the closest distance between subunit residues at the 3-fold interface of CA is ~6 Å compared to up to ~15 Å in CA_XL (as measured between the A204 carbonyl and its symmetry related mates) (fig. S3A and C). Water molecules at the inter-hexamer interface contribute to the stabilization of the capsid. Out of 450 assigned water molecules per CA hexamer in CA (75 per CA monomer), 30 are at the 2-fold and the 3-fold inter-hexamer interfaces (3 waters for each of the six 2-folds and 2 waters for each six 3-folds per hexamer). As there are ~250 CA hexamers per assembled capsid, thousands of water molecules at the inter-hexamer interfaces are likely to significantly contribute to capsid stability and geometry.

To further investigate the role of this hydration layer in CA structure we used a controlled dehydration protocol (29) that alters the water content of native CA crystals. Dehydration of CA crystals (dCA) resulted in shrinkage of the three dimensions of the unit cell by ~6%, ~6%, and ~3%, respectively (Fig. 1F, table S1). Superposition of the CA and dCA structures reveals significant conformational rearrangement in the relative orientation of the CA_NTD and CA_CTD domains, imparted by a hinge-like motion of the rigid bodies using the linker region as a pivot point (fig S4). These changes correlate with significant packing differences at the 2-fold and 3-fold (Fig. 2A vs. 2C, 2B vs. 2D, and fig. S3A vs. S3B). Hence, the hexamers of dCA (fig. S3B) arrange more tightly against each other (the closest distance ~3 Å as measured between residues A204 and L205 at the 3-fold interface) than those of CA_XL (fig. S3C) and CA (fig. S5A) creating up to 22 and 15 new molecular contacts in the 2-fold and 3-fold of dCA (tables S2-S3). Calculations of solvent accessible area reveal that dCA has ~150% and ~200% more buried surface than CA and CA_XL at the 2-fold, and ~400% more buried surface than CA at the 3-fold interface (table S4). The experimentally observed interactions at the 3-fold are between K203–T216, K203–A217 (involving main-chain atoms), and A204–L205, G206–T216, and P207–E213 (involving side-chain atoms) (table S3, Fig. 2D). Sequence alignment (fig. S5) of residues at the 2-fold and 3-fold shows that almost all of these are entirely conserved among HIV-1 subtypes (205, 206, 213, and 216; 207 is highly conserved). Consistent with our data, substitutions of A204 result in non-infectious virions with unstable or abnormal cores (2, 7). In addition, an extensive body of mutagenesis studies also confirms the importance of the observed residues as their substitution may lead to total loss of infectivity, defects in core morphology and stability, and abrogation of viral DNA synthesis (tables S2 and S3) (1, 6-11, 26, 30).

In addition to these inter-hexamer differences we also identify significant variations in the CA_NTD-CA_NTD and CA_NTD-CA_CTD intra-hexamer contacts (table S5). These involve changes in hydrogen bond networks, as well as water-mediated and hydrophobic interactions (Fig. 3A). Extensive variations in the CA_NTD-CA_NTD and CA_NTD-CA_CTD intra-hexamer contacts are also observed in the crystal structure of the Bovine Leukemia Virus (BLV) CA hexamer (Obal et al.), supporting the notion that variability in these interactions is a common feature among multiple retroviruses.

Fig. 3. Changes at the intra-hexamer interfaces.

(A) Stereo view of intra-hexamer CA_NTD-CA_NTD and CA_NTD-CA_CTD inter-subunit interfaces. Two neighboring CA subunits are shown as cartoons outlined for clarity (CA_NTDs in lighter colors than the corresponding CA_CTDs). Sites of varying interactions among CA, dCA, CA_XL, and CA_PF74 are marked in blue/orange for CA_NTD-CA_NTD interfaces or pink/brown for CA_CTD-CA_NTD interfaces. (B) Intra-subunit rearrangement linked to changes at the 2-fold interface. Enlarged stereo view of the boxed region shows changes in the position of H9 helices in neighboring subunits (marked with primed symbols). Least squares superposition (residues 143–174, 192–219) of dCA (cyan CA_NTD, blue CA_CTD) on CA (pink CA_NTD, red CA_CTD). Crystal dehydration results in slight extension of helix H8 (blue arrow), interaction of R143 with main chain E175 carbonyl instead of Q176, and repositioning of helix H9 (black arrow). In dCA, W184' from the H9' helix (light blue for dCA and salmon for CA) forms a hydrogen bond with main chain E175 carbonyl from a neighboring subunit, whereas in CA, W184' interacts with Q176 and side chain of E175 through water-mediated contacts. Moreover, in dCA R143 also interacts with E187' and T188'from the neighboring subunit, thus, becoming a part of the 2-fold interface. Black dashed lines connect residues that interact through hydrogen bonds.

Finally, a significant structural difference between CA and dCA involves key interactions of R143 in CA_NTD with E175 or Q176 in helix H8 of CA_CTD within the same subunit (Fig. 3, A and B). The R143-Q176 interaction in CA is replaced with R143-E175 in dCA. This change repositions helix H9, leads to loss of the W184-bound water, and alters the H9-H9 inter-hexamer interactions at the 2-fold (Fig. 3B). Similar alternative interactions of R143 with CA_CTD residues were also observed in engineered unliganded (PDB ID: 3H4E) or ligand complex structures (PDB ID: 4U0E), crystallized in orthorhombic (1) or hexagonal space groups respectively (18), highlighting its functional importance, which is further supported by virological studies (7, 18).

Collectively, the observed structural changes suggest that this remarkable plasticity may provide opportunities for therapeutic intervention by stabilizing conformations sampled during uncoating and assembly, directing CA to non-productive intermediates, or replacing key waters of the hydration layer at the 2-fold or 3-fold interfaces. Notably, residues at these interfaces are highly conserved among various clades and clinical isolates (7, 9, 31) providing a high genetic fragility that makes CA an appropriate target for antivirals (9).

Among more than twenty CA-targeting compounds, PF74 is the most potent CA-targeting antiviral that blocks HIV-1 by affecting multiple steps of the virus life cycle (12-16). In the structure of CA_NTD-PF74 the ligand occupies a preformed pocket defined by helices H3, H4, H5, and H7 (14). As this binding site is proximal to CA_CTD, it was proposed that PF74 could interact with CA_CTD as well. This prediction was confirmed in recent reports (18, 19), explaining how low concentrations of PF74 prevent binding of CPSF6 and nuclear entry (18-23). However, the crystal structures of crosslinked engineered CA did not fully address how PF74 can affect inter-hexamer interactions that affect capsid stability and reverse transcription. Hence, we solved the crystal structure of PF74 in complex with native HIV-1 CA (CA_PF74) permitting the detailed description of these interactions.

PF74 binds similarly across two CA monomers in the recent crosslinked structures by Price et al. (CA_XL-PF74) (18) and by Bhattacharya et al. (19), as well as in the present CA_PF74 structure (Fig. 4A, and fig. S6A) with the indole moiety bound differently than in CA_NTD-PF74 (Fig. 4B). In addition to the interactions observed in CA_NTD-PF74 (14), PF74 also interacts with helices H8 and H9 of CA_CTD from a neighboring subunit in the CA_XL-PF74 and CA_PF74 (18, 19). In CA_XL-PF74, helix H9 mutations W184A and M185A leave part of the H9-H9 2-fold interface disordered and there are also no interactions observed at the neighboring 3-fold region. These interactions are observed in CA_PF74. We also observe some differences in the intra-hexamer contacts of PF74 with CA_CTD residues at or near dimerization helix H9 (Y169 and K182 in CA_PF74, vs. Q176, S178, and E179 in CA_XL-PF74) (Fig. 4A) (18, 19). Interestingly, PF74 binding results in subtle structural changes at the remote 3-fold and 2-fold regions (Fig. 4C and fig. S6, B and C). These variations are reminiscent of the allosteric changes observed in the complex of CA_CTD with the CA-targeting antiviral peptide CAI that binds near the CA_CTD-CA_CTD interface (PDB ID: 2BUO) (16). Specifically, we observe changes in hydrophobic and water-mediated interactions at the CA_PF74 2-fold that affect the buried surface area in this region (tables S2 and S4). Also, changes of the H10 helices of CA_PF74 (rmsd ~1 Å with CA) lead to their convergence at the 3-fold, with the distance between A204 main chain oxygen atoms changing from 5.5 Å to 3.6 Å. In turn, this change leads to a displacement of the water molecules seen in CA (Fig. 4C and fig. S6C) and a 150% increase in buried surface area at the 3-fold of CA_PF74 (tables S3 and S4). Hence, the CA_PF74 structure provides insights into the mechanism by which high concentrations of PF74 change intra- and inter-hexamer interactions and affect the core stability (tables S2, S3, and S5).

Fig. 4. Effects of PF74 on HIV-1 CA structure.

(A) PF74 binding at the CA_NTD-CA_CTD interfaces of neighboring subunits within a CA hexamer. Top view of CA_PF74 hexamers (side view is shown in fig. S6A). The larger box shows a close-up stereo view of the PF74 binding site in relation to the 3-fold (triangle) and 2-fold (oval) interfaces. CA_NTDs and the corresponding CA_CTDs are colored by the same colors (light and dark, respectively). CA_NTD and a CA_CTD of a neighboring subunit participate in PF74 binding (magenta and cyan ribbons). Helix H9 is omitted for clarity. PF74 is shown modeled in a 2.7 Å simulated annealing omit map (σ=2.5). (B) PF74 binding geometries in CA_PF74 (blue), CA_XL-PF74 (PDB ID: 4U0E; PF74 in grey), and CA_NTD-PF74 (PDB ID: 2XDE; PF74 in magenta) PF74 in 4QNB is almost identical to 4U0E and omitted for clarity. (C) Stereo view of the 3-fold interface of CA_PF74 superposed onto CA (aligned on residues 1-219). Helices H10 of CA_PF74 and CA are in yellow and blue. Water molecules are shown as yellow spheres in CA; no water molecules were present in CA_PF74.

Obal et al. report a structural heterogeneity of CA monomers in the BLV CA hexamers (Obal et al.). Comparison of the related CA, dCA, and CA_PF74 structures also suggests a structural variability and conformational plasticity of CA. The variability in inter-hexamer CA_CTD contacts at the 2-fold and 3-fold in our related structures is likely representative of the heterogeneous arrangement of CA molecules in an asymmetric capsid core. Such structural variability is in agreement with the molecular models of an asymmetric HIV-1 capsid (2, 24).

The nature of CA structural variability goes beyond mere side-chain rearrangement. A key structural determinant important for the stabilization of variable interfaces is likely to be the presence of structured water molecules observed at strategic regions (1), including the 2-fold and 3-fold interfaces (Fig. 2, A and B). This hydration layer could function as an extension of CA structure, contributing to surface complementarity among flexible CA molecules. It is likely to be adaptable and facilitate nearly isoenergetic structural rearrangements that contribute to quasi-equivalent structural variability.

In addition, similar to previous reports (1) we observe changes in the relative orientation of CA_NTD and CA_CTD (~10° rotation in fig. S4) that may cause a tilt among neighboring hexamers and contribute to the surface curvature of an authentic capsid. Different inter-hexamer tilt angles would be anticipated at the narrow and broad ends of the capsid. In addition, intra-hexamer interactions between CA monomers from laterally or longitudinally positioned hexamers are likely to differ in tilt, because of differences in surface curvature in the respective directions. Given the asymmetric nature of capsid, it is likely that no two CA molecules are identical in any one HIV-1 capsid core (2, 24). Hence, CA plasticity allows a wide range of conformations that contribute to the structural variability of the HIV-1 asymmetric core.

CA pliability may provide biological and evolutionary benefits to viruses by allowing interactions with multiple host factors. For HIV-1, these include TRIM5α, CPSF6, MxB, cyclophilin, NUP153, and NUP358 (recently reviewed in (32, 33)). Such diverse interactions afford a functional versatility reminiscent of a Swiss Army knife. Moreover, the large variability in subunit distances at the 3-fold region (ranging from ~3 Å to ~15 Å in fig. S3) may affect permeability to deoxynucleoside triphosphates that can have a footprint diameter of ≤10 Å. This may be relevant to the reverse transcription process, although nucleotide trafficking could also be affected by possible occasional imperfections in the malleable capsid core structure. Extensive structural variability may also become an Achilles heel for the virus by providing multiple opportunities for pocket targeting, destabilization of core, or stabilization of non-productive structural intermediates.

One Sentence Summary.

X-ray structures of native HIV-1 capsid protein reveal interactions important for uncoating, assembly, and CA-targeting by antivirals.