Structural basis of HIV-1 capsid recognition by PF74 and CPSF6

Significance

Events that occur between entry of the HIV-1 capsid into the cytoplasm of the target cell and the delivery of the viral genetic material into the nucleus constitute some of the less well understood processes in the viral life cycle. We demonstrated that PF74, a small-molecule inhibitor of HIV-1, and the host proteins CPSF6 and NUP153 bind to a preformed pocket within the CA protein hexamers that exist within the assembled capsid. Our results suggest that key features of the CA hexameric lattice remain intact upon docking at the nuclear pore. In addition, low molecular weight ligands that better mimic virus–host, protein–protein interactions at the intersubunit interfaces within the assembled viral capsid may offer novel avenues for therapeutic intervention.

Abstract

Upon infection of susceptible cells by HIV-1, the conical capsid formed by ∼250 hexamers and 12 pentamers of the CA protein is delivered to the cytoplasm. The capsid shields the RNA genome and proteins required for reverse transcription. In addition, the surface of the capsid mediates numerous host–virus interactions, which either promote infection or enable viral restriction by innate immune responses. In the intact capsid, there is an intermolecular interface between the N-terminal domain (NTD) of one subunit and the C-terminal domain (CTD) of the adjacent subunit within the same hexameric ring. The NTD–CTD interface is critical for capsid assembly, both as an architectural element of the CA hexamer and pentamer and as a mechanistic element for generating lattice curvature. Here we report biochemical experiments showing that PF-3450074 (PF74), a drug that inhibits HIV-1 infection, as well as host proteins cleavage and polyadenylation specific factor 6 (CPSF6) and nucleoporin 153 kDa (NUP153), bind to the CA hexamer with at least 10-fold higher affinities compared with nonassembled CA or isolated CA domains. The crystal structure of PF74 in complex with the CA hexamer reveals that PF74 binds in a preformed pocket encompassing the NTD–CTD interface, suggesting that the principal inhibitory target of PF74 is the assembled capsid. Likewise, CPSF6 binds in the same pocket. Given that the NTD–CTD interface is a specific molecular signature of assembled hexamers in the capsid, binding of NUP153 at this site suggests that key features of capsid architecture remain intact upon delivery of the preintegration complex to the nucleus.

The HIV-1 CA protein (also known as p24) performs multiple different roles in viral replication (1). CA is now well characterized with respect to its functions in virion assembly and morphogenesis. As the central region of the precursor Gag polyprotein, CA, together with the downstream SP1 element, facilitates virus assembly by forming the bulk of the intermolecular interactions that generate the spherical, immature protein shell of the virion (2). Upon proteolytic maturation and disassembly of the immature Gag lattice, approximately two-thirds of the mature CA proteins in the virion reassemble into the mature capsid, a fullerene cone shell that encapsulates the viral genome and associated enzymes. CA is a highly helical protein composed of two domains: an N-terminal domain (NTD) and a C-terminal domain (CTD), which are connected by a flexible linker. To form the mature capsid, CA assembles into a lattice of hexameric rings composed of an inner ring of six NTDs surrounded by a belt of six CTDs (3). The lattice incorporates 12 CA pentamers to form a closed protein shell. Each CA hexamer and pentamer is formed by intermolecular NTD–NTD and NTD–CTD contacts, and the extended lattice is formed by dimeric CTD–CTD interactions (3–7).

In addition to its structural role in viral assembly and maturation, CA mediates interactions with numerous host proteins upon cell entry. The assembled capsid of the mature viral core shields the other viral components from the cytoplasm, and many studies point to the key role that CA plays in orchestrating an intricate network of virus–host interactions, resulting in either productive infection or in restriction of viral replication by intrinsic cellular defenses. Precisely how these virus–host protein–protein interactions contribute to viral replication or restriction remains a matter of debate for most of the CA-binding host proteins. In many cases, the host factors seem to affect the still poorly understood process of core uncoating or capsid disassembly. However, it is not well established whether these factors act by directly perturbing capsid stability and whether they can recognize specific states of the capsid in the viral life cycle.

The HIV-1 CA protein is now established to be a key determinant of the ability of HIV-1 to infect nondividing cells. In particular, CA is genetically and functionally implicated in nuclear import of the provirus, by mediating interactions with cellular transport factors, including cleavage and polyadenylation specific factor 6 (CPSF6) (8-10), TNPO3 (11-13), NUP538/RanBP2 (14), and nucleoporin 153 kDa (NUP153) (15). A conserved pocket between NTD helices 3 and 4 mediates the interactions between CA and the transport factors CPSF6 and NUP153 (10, 15). This same pocket is also the binding site for the small-molecule inhibitor PF-3450074 (PF74) (16). Indeed, competitive binding of these three ligands to CA has been experimentally demonstrated (10, 15, 17). In the cases of PF74 and CPSF6, binding to the isolated NTD is structurally well characterized (10, 16). Although the molecular details are necessarily different, both are stabilized by packing of a key aromatic ring from the ligand (Phe321 in CPSF6) against the center of the NTD binding pocket. The NTD pocket is not involved in any of the CA–CA interactions in the mature capsid. Rather, the pocket lines the rim of a deep groove between CA subunits that is accessible to solvent from the outer surface of the capsid.

In the intact capsid, the NTD pocket sits atop the NTD–CTD interface, which is an intermolecular interface between the NTD of one subunit and the CTD of the adjacent subunit within the same hexameric ring (3). This interface is critical for capsid assembly both as an architectural element of the CA hexamer and pentamer and as a mechanistic element for generating lattice curvature (3–5). Therefore, we decided to investigate whether ligands that occupy the NTD pocket could also reach toward the NTD–CTD interface. Here we report biochemical experiments showing that PF74, CPSF6, and NUP153 bind to the CA hexamer with at least 10-fold higher affinities compared with nonassembled CA or isolated CA domains. Crystal structures of PF74 and CPSF6 in complex with the hexamer reveal that the common binding pocket is larger than previously thought and actually encompasses part of the NTD–CTD interface. Our aggregate data indicate that CPSF6 and PF74 specifically recognize the NTD–CTD interface, and this may be a general molecular mechanism by which cellular host factors can distinguish the assembled capsid from other functional forms of CA.

Results and Discussion

The PF74 Binding Site Is the NTD–CTD Interface.

To understand more clearly how PF74 interacts with the CA protein, we soaked the inhibitor with previously described P6 crystals of disulfide-stabilized CA hexamers (3, 18) and determined the structure of the bound complex at 2-Å resolution (Fig. 1 and Table S1). We observed very clearly defined unbiased difference density for the bound compound (Fig. 1B, magenta). In our structure, each of the six available binding pockets in the hexamer is occupied by the compound. The six pockets are crystallographically equivalent in this crystal form (Fig. 1A).

Fig. 1.

Structure of PF74 in complex with the HIV-1 CA hexamer. (A) Top view of the hexamer, with each subunit in a different color and the bound PF74 in white. (B) Unbiased mFo-DFc density at 3σ (magenta mesh) clearly defines the bound compound at the NTD–CTD interface. NTD is in cyan, and CTD is in green. (C) Chemical structure of PF74, with the R1, R2, and R3 rings labeled. (D) Superposition of hexamer-bound PF74 (NTD in cyan, compound in white) and NTD-bound PF74 (16) (NTD in dark blue, compound in cyan). (E) Close-up view of the R3 indole bound to its subpocket. Relevant residues are shown as spheres (NTD in blue and CTD in green). The hydrogen bond between the indole N3 amide and NTD Gln63 sidechain is shown in yellow. (F) Superposition of intersubunit NTD–CTD interfaces in the structure with bound PF74 (light shades) and unbound hexamer (dark shades). Key binding residues are shown as sticks and labeled.

PF74 contains three aromatic moieties: R1 and R2 are both phenyl rings, and R3 is a substituted indole ring (16) (Fig. 1 B and C). As illustrated in Fig. 1D, the R1 and R2 moieties interact with the NTD in exactly the same configuration as previously observed in 2XDE (16), the crystal structure of PF74 in complex with the isolated NTD. In the hexamer-bound form, however, the R3 indole is flipped relative to 2XDE (encircled in Fig. 1D). In this configuration, the indole N3 amide is oriented to donate a hydrogen bond (Fig. 1E, yellow) to the sidechain carbonyl of Gln63. The indole ring also reaches out to the neighboring CTD and stacks against the guanidinium group of Arg173 and also packs against the aliphatic sidechain of Leu172 (Fig. 1 D and E). Furthermore, the loop connecting helices 8 and 9 (residues 178–180) contacts the six-membered ring of the R3 indole, and the CTD Glu180 sidechain points toward the NTD to interact with the sidechain amine of Lys70. The configuration of this loop very likely reflects the enhanced flexibility in this region caused by the downstream W184A and M185A mutations in the CTD (3).

An important feature is that the PF74 binding pocket within the hexamer is preformed. Superposition of the PF74-bound NTD–CTD interface with that in the unbound hexamer (3) indicates no structural changes in the hexamer structure (Fig. 1F). The configurations of the Gln63 and Arg173 sidechains are therefore “locked in” before binding, and PF74 essentially replaces bound water molecules (compare Fig. S1 A and B). Because the native NTD–CTD interface seems to be a unique feature of the mature capsid lattice, we take this as an indication that the principal inhibitory target of PF74 is the assembled capsid, and not the unassembled CA subunit.

Biochemical Analysis of the PF74 Interaction with CA.

Our structural analysis indicated that PF74 would show greater affinity for the hexamer compared with the isolated NTD or unassembled forms of the CA protein. To test this prediction, we performed isothermal titration calorimetry experiments (Table S2). We first confirmed published findings that the compound bound to the isolated NTD (Fig. S2A) and full-length, unassembled CA (Fig. S2 B and C) with equivalent dissociation constants (K_d) of ∼4 μM (16). In contrast, and as predicted by our structure, PF74 bound to the soluble, disulfide-stabilized CA hexamer with >10-fold greater affinity (262 nM) (Fig. S2D). We note that this value is more in line with the reported EC₅₀ of ∼500 nM for PF74 (16), again supporting the idea that PF74 specifically targets the assembled capsid.

Given the stacking interaction of the R3 indole with the CTD, we also measured affinities for soluble hexamers containing the R173A and R173K mutations, and indeed found these to be at micromolar levels (Fig. S2 F and G). PF74 did not bind to the isolated CTD (Fig. S2H), consistent with the interpretation that engagement of the indole/R173 stack requires the CTD to be appropriately oriented relative to the NTD. We also found that the binding affinity of PF74 for soluble, disulfide-stabilized N21C/A22C CA pentamers is similar to the hexamer (Fig. S2E). Unfortunately, we were unsuccessful in obtaining a structure of PF74 bound to the CA pentamer, because the pentamer crystals rapidly dissolved upon addition of the compound. One possible interpretation of this observation is that, in the pentamer case, the CTD moves significantly to accommodate the compound within the NTD–CTD interface.

In vitro selection experiments have previously identified NTD mutations (T107N and 5Mut = Q67H, K70R, H87P, T107N, and L111I) that conferred viral resistance to PF74 and its analogs (16). Accordingly, monomeric NTD and hexameric CA constructs containing these mutations all displayed lower affinities for these compounds, compared with the corresponding wild-type constructs (Table S2 and Fig. S3). These results suggest that the mutations confer resistance to PF74 by impairing the compound’s ability to productively bind the capsid.

Effect of PF74 on Capsid Assembly.

Our finding that the PF74 binding site is preformed in the hexamer is consistent with published studies showing that PF74 significantly increases the rate of CA assembly in vitro (16). This also suggested that PF74 would stabilize preassembled CA tubes, which are composed almost exclusively of hexamers. To test this, we used a centrifugation assay using CA-NC tubes. Upon centrifugation through a sucrose cushion, the disassembled proteins remain in the supernatant, whereas intact tubes form a pellet at the bottom of the centrifuge tube. As previously described, centrifugation in the presence of a destabilizing buffer (19) resulted in essentially no CA-NC proteins in the pellet (Fig. 2A, lane 2). In contrast, the presence of increasing amounts of PF74 resulted in correspondingly increasing amounts of CA-NC proteins detected in the pellet (Fig. 2A, lanes 3–6). Thus, PF74 stabilizes preassembled CA hexamers and the hexagonal capsid lattice.

Fig. 2.

Effect of PF74 on CA assemblies. (A) In vitro-assembled CA-NC tubes were incubated with either stabilization buffer (lane 1) or destabilization buffer with increasing amounts of PF74 (lanes 2–6), then subjected to centrifugation through a sucrose cushion. CA-NC proteins were detected by Western immunoblot analysis using anti-p24 antibodies. The CA-NC proteins in the pellet reflect the presence of intact assemblies. Similar results were obtained in three independent experiments, and a representative experiment is shown. (B) Human HeLa cells stably expressing rhesus TRIM5α (TRIM5α_rh) or containing the empty vector LPCX were challenged with similar amounts of HIV-1-GFP in the absence (lanes 1 and 3) or presence of 10 µM PF74 (lanes 2 and 4) for 12 h. Infected cells were used to prepare postnuclear supernatants (INPUT) that were layered onto a 50% sucrose cushion to separate soluble (SOLUBLE) from pelletable (PELLET) HIV-1 capsids. Fractions were analyzed by Western immunoblot analysis using antibodies against HIV-1 CA. The percentages of pelletable capsids relative to the infected control in the presence of DMSO are shown. Similar results were obtained in three independent experiments, and SDs are shown. There is a significant reduction in the TRIM5α signal versus LPCX (P = 0.02).

Perplexingly, incubation of PF74 with isolated HIV-1 cores has been previously shown to instead have a destabilizing effect on the assembled, fullerene capsid (20). An extensive set of electron microscopy experiments examined wild-type and disulfide-stabilized preparations of two-dimensional CA crystals and helical tubes assembled as described in refs. 3, 18, and 21. A dramatic and reproducible phenotypic change in the presence and absence of PF74 could not be discerned. Therefore, as an alternative, we performed the “fate of the capsid” assay in the presence and absence of PF74. In this assay, the relative levels of soluble vs. pelletable capsid proteins in the cytoplasm of infected cells are determined, which reflects the disassembly state of the HIV-1 core (22). As shown in Fig. 2B, the presence of PF74 significantly reduced the amount of pelletable CA proteins during infection (compare lanes 1 and 2). As a control, Fig. 2B, lanes 3 and 4 also illustrate the ability of rhesus TRIM5α to accelerate the uncoating process of HIV-1. We therefore conclude that PF74 stabilizes CA hexamers and hexameric (tubular) assemblies but destabilizes the mature capsid.

We propose that the seemingly contradictory effects of PF74 on CA assembly might be explained by the fact that in the mature fullerene capsid, the assembled CA subunits must adopt multiple different quasi-equivalent conformations to form both hexamers and pentamers and a variably curved capsid lattice. The multiple CA conformations have been shown to arise, in part, from flexion about the NTD–CTD assembly interface (3–7). Because the NTD–CTD interface also forms the binding pocket for PF74, we speculate that the compound severely limits the flexibility of CA subunits, perhaps to a range that is compatible only with the constant curvature found in tubes and not the full range required for the variably curved capsid. However, given that these effects are observed at significantly higher doses of the compound than the affinities we measured here or its reported IC₅₀ (16), the relevant effect of PF74 is more likely to be due to competitive inhibition of binding factors.

The CPSF6 Binding Site Is the NTD–CTD Interface.

CPSF6 residues 313–327 have been previously shown to constitute the minimal binding epitope of CPSF6 for the CA protein, and a structure of this peptide in complex with the isolated NTD has been reported [Protein Data Bank (PDB) ID: 4B4N] (10, 23). To understand how CPSF6 binds to assembled CA, we cocrystallized GB1-tagged CPSF6 peptide with soluble CA hexamers and determined the structure of the bound complex at 2.7-Å resolution (Fig. 3 and Table S1). The complex crystallized in the orthorhombic space group previously described for one of the crystal forms of hexameric CA (PDB ID: 3H4E) (3). The asymmetric unit of the crystal contains two CA hexamers oriented head-to-head, with their NTDs and CypA-binding loops facing each other. The space surrounding the lateral surfaces of hexamer pairs is sufficiently large to accommodate 12 GB1 domain tags in the asymmetric unit, but the tags are not resolved in the electron density, indicating significant disorder in their crystal packing. In contrast, the conformation of the CPSF6 peptide is clearly defined by difference density at each of the 12 crystallographically independent binding sites in the asymmetric unit (Fig. 3A).

Fig. 3.

Structure of CPSF6 in complex with the HIV-1 CA hexamer. (A) Electron density of CPSF6_313–327 (orange mesh) and of the surrounding NTD (green) and CTD (cyan) residues from two adjacent CA monomers that form the binding pocket in mesh representation. CPSF6 is shown as sticks (orange), and the NTD (green) and CTD (cyan) as cartoons. Electron density at better-resolved binding sites suggests the possibility of a hydrogen bond between the sidechain of CTD Lys182 and the carbonyl oxygen atom of CPSF6 Gly318 (dotted black line). The mesh representations for CPSF6 residues on chain p, CTD chain c residues 165–190, and NTD chain d residues 95–109 were set at 0.9, 1.2, and 1.2 σ, respectively. (B) Side view of the CA hexamer with bound CPSF6 peptides (yellow). The CPSF6 binding site is located between the NTD (green) and CTD (cyan) of neighboring CA monomers. The remaining CA subunits in the hexamer are shown as a heat map depicting the degree of conservation based on the analysis of 97 unique HIV-1/HIV-2/other SIV sequences. Red is least conserved, and blue is most conserved. One of the bound CPSF6 peptides is shown as a backbone trace to reveal the conservation of the binding site. (C) A detailed view of the CPSF6_313–327 polypeptide (yellow) in the binding pocket formed by NTD α-helices 3 and 4 (green) and CTD α-helices 8 and 9 (cyan). (D) Weblogo representation of the sequence conservation of capsid CTD residues located in helices 8 and 9 and the linker between them. Ninety-seven independent primate lentiviral sequences were analyzed. The height of a particular residue indicates its degree of conservation. Lys182 is highly conserved, whereas Gln179 is not.

As in the case of PF74, the interaction of the CPSF6 peptide with the NTD in the hexamer context is not significantly altered with respect to 4B4N (10). Specifically, residues 313–327 of the peptide adopt a compact fold that is stabilized by both intramolecular and intermolecular hydrogen bonds, with the phenyl ring of Phe321 projecting downward into a triangular cleft made by NTD helices 3 and 4 and packing against hydrophobic sidechains in both helices (Fig. 3A). The binding pose of the CPSF6 Phe321 sidechain is equivalent to that of the PF74 R2 ring, as previously noted (15) (Fig. S4). Both the amino and carboxyl termini of the peptide point toward what would be the outer surface of the assembled capsid, indicating that the protein segment containing residues 313–327 of CPSF6 most likely protrudes from the folded bulk of the protein to access its binding site on the assembled capsid, as previously suggested (10).

In our crystal structure, the NTD–CTD interface positions the CTD of the adjacent CA subunit in close proximity to the bound CPSF6 peptide (Fig. 3B, with NTD colored in green, CTD in cyan, and CPSF6 in yellow). Helix 8 of the CTD makes up the bottom of the deep groove that accommodates the bound CPSF6 peptide, whereas helix 9 flanks the peptide on the side opposite to the CPSF6–NTD interface (Fig. 3A). The flexible linker between CTD helices 8 and 9 can be resolved in some of the 12 independent binding sites; importantly, the configurations of these resolved loops are similar to those observed in CTD structures lacking the W184A and M185A mutations (4, 24). These resolved loops wall off part of the binding site and are in contact with the peptide (Fig. 3A). In two of these resolved sites, the electron density indicates that the sidechain amine of CTD Lys182 may be forming a hydrogen bond with the backbone carbonyl of CPSF6 Gly318 (Fig. 3C). In the same two sites, the sidechain of CTD GLN179 appears to form hydrogen bonds with the backbone carbonyl of CPSF6 Pro317 and the sidechain of CPSF6 Gln319. Contact with the CTD forms the lowest anchoring point of the bound peptide, with the CPSF6 Pro317 imino ring embedded within a subpocket at the bottom of the binding site. As noted previously, the CPSF6 peptide is rich in prolines and glycines, which introduce kinks and impart backbone flexibility (10). Our new structure indicates that the sharp turn around Pro317-Gly318 of the bound peptide is well accommodated and is likely to be stabilized by the presence of the CTD.

The CPSF6-binding surface of the NTD is conserved across HIV-1 strains (10). We extended this analysis to the CTD residues forming part of the CPSF6-binding groove. Alignment of CTD sequences from the curated sequence databases maintained by the Los Alamos National Laboratory revealed remarkably high conservation of Lys182 among primate immunodeficiency viruses (Fig. 3D and Fig. S5). This observation supports the notion that the CTD-mediated contacts, including Lys182, may contribute to the biological function of the CPSF6-binding interface.

Biochemical Analysis of the CPSF6 Interaction with CA.

As with PF74, we compared the binding affinities of the CPSF6 peptide with the CA hexamer and other forms of CA. CPSF6–CA interactions are relatively weak and quantitative determination of the dissociation constants is challenging. We therefore used two independent methods to measure binding affinities: sedimentation velocity analytical ultracentrifugation (Fig. 4 A–C) and fluorescence polarization spectroscopy (Fig. 4 D–F). The binding studies were performed using fluorescein-labeled CPSF6 peptide (fCPSF6) (SI Materials and Methods). In the analytical ultracentrifugation assay, sedimentation rates of the fCPSF6 were monitored in the presence of increasing concentrations of CA. van Holde–Weischet analysis of the sedimentation data revealed that the free CPSF6 peptide sedimented at ∼0.7 Svedberg units (S) (Fig. 4 B and C, red curves). In the presence of the isolated NTD or soluble CA hexamer, the bound fraction sedimented at ∼1.5S (Fig. 4B) and ∼7S, respectively (Fig. 4C). These values are consistent with the expected hydrodynamic behavior of the CA protein complexes. Fig. 4A plots the fraction of bound peptide (estimated from their boundary fractions) as a function of added binding partner, and the fitted dissociation constants are listed in Table S3. We found that the CPSF6 peptide bound to the hexamer (Fig. 4A, blue) with >10-fold higher affinity compared with the isolated NTD (Fig. 4A, red) (100 μM vs. 1170 μM, respectively). Affinities measured by fluorescence polarization were comparable: 83 μM for the hexamer and 872 μM for the isolated NTD (Fig. 4D). We also used the polarization assay to measure the affinity for full-length CA (K_d = 436 μM) (Fig. 4D, green). Thus, CPSF6 binds to the hexamer more tightly compared with the isolated NTD or unassembled CA, as expected from the structure. To examine the importance of the CTD contacts observed in our crystal structure, we also used fluorescence polarization to test the binding of the CPSF6 peptide to CA hexamers harboring the K182R and K182A mutations. These experiments yielded K_ds of 587 μM for K182R (Fig. 4E, red) and 424 μM for K182A (Fig. 4E, green), which are comparable to the affinity for unassembled CA. Overall, our results indicate that full-affinity binding of CPSF6 to the CA hexamer involves energetically significant contacts with both the NTD and the CTD, and that engagement of the CTD contacts requires the two CA domains to be appropriately oriented relative to each other via the intersubunit NTD–CTD interface within the CA hexamer.

Fig. 4.

Analysis of CPSF6 and NUP153 binding to HIV-1 CA. (A) Binding isotherms of fluorescein-CPSF6 (fCPSF6) peptide to the isolated NTD (red) and the CA hexamer (blue), which were derived from van Holde–Wischet analysis of analytical ultracentrifugation data. (B) Representative van Holde–Wischet model-free analysis of the fCPSF6 sedimentation rates in the presence of increasing concentrations of NTD. The red-orange-yellow-green-blue-purple rainbow color series corresponds to NTD concentrations of 0 μM, 63 μM, 250 μM, 625 μM, 1460 μM, and 1800 μM, respectively. fCPSF6 concentration was 2.2 μM for the free peptide (red) and 12 nM in the rest of the titration series. The data show that the sedimentation coefficient of the free fCPSF6 (∼0.7S) does not depend on the concentration of the peptide. (C) van Holde–Wischet analysis of fCPSF6 binding to the CA hexamer. The red-orange-yellow-green-blue-purple rainbow color series corresponds to CA (monomer) concentrations of 0 μM, 45 μM, 65 μM, 108 μM, 305 μM, and 530 μM, respectively. (D) Fluorescence polarization measurements of fCPSF6 binding to the CA hexamer (blue), wild-type full-length CA (green), and isolated NTD (red). (E) Fluorescence polarization measurements of fCPSF6 binding to the wild-type (blue), K182A (red), and K182R (green) hexamers. (F) Fluorescence polarization measurements of fNUP153 binding to the CA hexamer (red) and NTD (green) is compared with fCPSF6 binding to the hexamer (blue). The fNUP153 FG peptide binds more tightly to the hexamer compared with the isolated NTD.

Implications for Capsid Recognition and the Mechanism of Uncoating.

Although it now seems established that the CA protein remains associated with the newly synthesized viral DNA for a significant period after entry, the exact timing of uncoating, or disassembly of the capsid lattice and its physical separation from the genome, is still being debated. Some studies support early capsid dissociation, whereas others suggest docking of intact cores at the nuclear pore. We propose that a complementary approach to delineate the timing of capsid disassembly is to systematically classify postentry host factors as either capsid binders that can detect molecular signatures of assembled CA, or CA protein binders that recognize free capsid subunits, and then correlate this property with the known cellular localization of these factors. We further propose that the intersubunit NTD–CTD interface is a specific signature of the assembled capsid. The native capsid is held together by highly cooperative, noncovalent interactions between the subunits. Notwithstanding the fact that soluble HIV-1 CA hexamers and pentamers can be obtained (as we have done here) by artificially stabilizing the NTD–NTD interactions with engineered disulfide bonds, the native NTD–CTD interface most likely remains intact only in the presence of the other noncovalent NTD–NTD and CTD–CTD interactions in the assembled CA lattice.

Our results provide a strong argument for classifying CPSF6 as a cellular protein that binds the retroviral capsid before its disassembly. Although the precise role of CPSF6 in postentry events remains to be defined, the importance of this protein in HIV-1 replication is underscored by a recent study demonstrating that there is strong selective pressure in vivo for HIV-1 CA to maintain CPSF6 binding (25). One possibility is that CPSF6 binding is required for evasion of innate immunity because HIV-1 mutants impaired in CPSF6 binding were recently shown to trigger innate immunity responses in macrophages (26). On the other hand, CPSF6 seems to suppress replication of the viral variants that escape CTL responses in individuals carrying the HLA-B27 allele (25). Interestingly, the S41A capsid mutation, which frequently arises in HLA-B27⁺ individuals and is associated with increased viral loads, alleviates the inhibitory effect of CPSF6 on viral replication without disrupting CPSF6 binding (25, 27). Our structure confirms that whereas Ser41 is located in the same deep groove between CA monomers where CPSF6 peptide binds, it is not in direct contact with CPSF6 and is not expected to directly interfere with CPSF6 binding. However, the proximity of Ser41 to the CPSF6-binding site and its location on a neighboring CA monomer further support the notion that CPSF6 acts before disassembly of the capsid.

It is quite significant that at least two nucleoporins, NUP358/RANBP2 and NUP153, have been demonstrated to directly bind CA (14, 15). In particular, the NUP153 FG repeats bind to the same NTD pocket as CPSF6 and PF74, and it is likely that an equivalent FG-derived phenylalanine ring engages the hydrophobic NTD pocket (15). We have also analyzed biochemically the relative CA binding affinities of a fluorescent NUP153 FG peptide (fNUP153; described in SI Materials and Methods). As shown in Fig. 4F, fNUP153 binds to the CA hexamer with lower affinity compared with CPSF6 peptide, although the affinity is likely to be higher for full-length NUP153 because of the highly repetitive nature of the FG motif. However, even the short fNUP153 peptide, which contains a single FG motif, displays much higher affinity to the CA hexamer (Fig. 4F, red) compared with the isolated NTD (Fig. 4F, green). The apparent preference of NUP153 for assembled capsid supports a model wherein the capsid remains intact (or largely intact) immediately before docking at the nuclear pore.

In summary, our study reveals that the small-molecule inhibitor PF74 and the host protein CPSF6 (and most likely NUP153) bind to the HIV-1 capsid by making contacts with two neighboring CA subunits within the assembled capsid. The atomic details of the interface and mutations we describe may inform future studies of the mechanisms by which capsid-interacting host factors contribute to viral uncoating, nuclear import, and viral restriction. Low molecular weight ligands that better mimic virus/host protein–protein interactions at the intersubunit interfaces within the assembled viral capsid may offer novel avenues for therapeutic intervention.

Materials and Methods

Protein Purification.

Plasmids encoding the wild-type CA, NTD, and CTD were kind gifts of W. I. Sundquist, University of Utah, Salt Lake City, UT. CA proteins containing cysteine mutations for disulfide bond stabilization have been previously described (3, 18). All other mutants described in this study were made by Quikchange mutagenesis (Illumina). CA proteins were purified as previously described (3). Stock solutions (10 mg/mL) of soluble, disulfide-stabilized HIV-1 CA hexamers and pentamers were prepared by sequential dialysis as previously described (3).

DNA encoding the CPSF6 peptide used for crystallization, spanning residues 313–327 (PVLFPGQPFGQPPLG) was cloned into the previously described pET30a-based vector developed for expression of proteins fused to the GB1 solubility tag (28). The tagged peptide was expressed in Rosetta2(DE3) Escherichia coli cells (Invitrogen). For purification, cells were resuspended in lysis buffer [50 mM Tris (pH 7.5), 100 mM NaCl, 10 mM β-mercaptoethanol (βME), and 1% Tween 20] and lysed by sonication. The clarified supernatant was applied on Ni-NTA resin, and the bound fraction was purified to ∼95% homogeneity by size exclusion chromatography. The protein yielded a sharp peak at ∼79 mL with a HiLoad 16/60 Superdex 75 column (GE Healthcare).

Structure Determinations of the PF74-Bound and CPSF6-Bound CA Hexamers.

P6 crystals of CA hexamers were obtained in 10–14% PEG 8000, 2% (wt/vol) Tacsimate, and 100 mM Tris (pH 7.6–8.6), as previously described (18). The crystals were soaked in mother liquor containing 25% glycerol and several test dilutions of PF74 (from a 30 mM stock in DMSO) for up to 30 min, then flash-frozen in liquid nitrogen. Crystals of the CPSF-bound hexamers were identified in an automated screen at the X-ray Crystallography Core Laboratory, University of Texas Health Science Center, San Antonio (SI Materials and Methods).

Biochemical and Biophysical Methods.

The centrifugation-based in vitro stability assay of assembled CA-NC tubes was measured as previously described (19). Input and pellet fractions were analyzed by Western immunoblot analysis using antibodies against HIV-1 CA. The fate of the capsid assay was performed as previously described (22, 29) (SI Materials and Methods). Isothermal titration calorimetry (ITC) was performed using an ITC200 microcalorimeter (MicroCal) operating at 28 °C. Sedimentation velocity runs were carried out at a speed of 50,000 rpm and temperature of 20 °C in a Beckman Optima XLA-1 ultracentrifuge equipped with an eight-slot An50-Ti rotor. Fluorescence polarization spectroscopy was performed with a Biotek Synergy 2 plate reader (SI Materials and Methods).

Note Added in Proof.

The reader is referred to a comprehensive study by Price et al. (30) also showing binding of small molecule drugs and host factors to the NTD-CTD interface of disulfide-stabilized CA hexamers, which appeared after submission of our manuscript.