|Home | About | Journals | Submit | Contact Us | Français|
Mature HIV-1 particles contain conical-shaped capsids that enclose the viral RNA genome and perform essential functions in the virus life cycle. Previous structural analysis of two and three-dimensional arrays provided a molecular model of the capsid protein (CA) hexamer and revealed three interfaces. Here, we present a cryoEM study of a tubular assembly of CA and a high-resolution NMR structure of the CA C-terminal domain (CTD) dimer. In the solution dimer structure, the monomers exhibit different relative orientations compared to previous X-ray structures. The solution structure fits extremely well into the EM density map, suggesting that the dimer interface is retained in the assembled CA. We also identified a novel CTD-CTD interface at the local three-fold axis in the cryoEM map and confirmed its functional importance by mutagenesis. In the tubular assembly, CA intermolecular interfaces vary slightly, accommodating the asymmetry present in tubes. This provides the necessary plasticity to allow for controlled virus capsid dis/assembly.
During retrovirus assembly, the viral protease cleaves the Gag polyprotein into its individual components, which rearrange to form a mature viral protein shell, the capsid, that encloses the RNA genome. The human immunodeficiency virus type 1 (HIV-1) capsid is composed of ~1500 copies of the capsid protein (CA). Among retroviruses, the architecture of the capsid varies, with those of HIV-1 generally forming conical assemblies, while those of the Rous sarcoma virus (RSV) form irregular polyhedra. Such pleiomorphic nature of retroviral capsids is elegantly explained by a fullerene cone model (Ganser et al., 1999; Li et al., 2000) in which the capsid comprises a hexameric lattice containing a variable number of CA pentamers, thus allowing for closure of tubular or conical structures. Studies using cryo-electron microscopy (cryoEM)(Briggs et al., 2004; Ganser-Pornillos et al., 2007; Li et al., 2000; Nermut et al., 1998; Wright et al., 2007) and image reconstruction have provided a general picture of the architecture of immature and mature HIV-1 capsids. HIV-1 CA consists of two domains, a ~150 amino acid N-terminal domain (NTD) and an ~80 amino acid C-terminal domain (CTD), connected by a hinge. A number of atomic structures of individual CA domains have been determined by X-ray crystallography (Gamble et al., 1996; Gamble et al., 1997; Ivanov et al., 2007; Kelly et al., 2006; Momany et al., 1996; Ternois et al., 2005; Worthylake et al., 1999) and by NMR (Tang et al., 2002; Wong et al., 2008). Two different crystal structures of full-length HIV-1 CA (CA-FL) have been reported (Berthet-Colominas et al., 1999; Pornillos et al., 2009).
In solution, HIV-1 CA dimerizes with a Kd of 18μM, and dimerization is dependent on CTD residues Trp184 and Met185 (Gamble et al., 1997). Mutagenesis of these residues interferes with in vitro assembly and abolishes viral infectivity in cell culture (von Schwedler et al., 2003). Experiments using hydrogen-deuterium exchange and chemical cross-linking suggest an intermolecular NTD-CTD interaction in the mature HIV-1 capsid (Lanman et al., 2003). Similar findings have been reported for the RSV capsid (Cardone et al., 2009).
Two pseudo-atomic structures of assembled HIV-1 CA, one in tubes (Li et al., 2000) and the other in 2D crystals (Ganser-Pornillos et al., 2007), have been reported. Both structures were derived from cryoEM reconstructions of assembled CA-FL, and the maps are consistent with a hexameric arrangement of CA. Contacts between the NTDs were reported to stabilize the hexameric assembly, in which three N-terminal helices form an 18-helix bundle at the center, with neighboring hexamers being connected via a dimer interface, which was mapped to helix 9 of the CTD (Gamble et al., 1997). A similar organization was observed in a high-resolution X-ray structure of the murine leukemia virus CA NTD protein (Mortuza et al., 2004). Docking of the individual NTD and CTD domain structures into the HIV-1 CA hexamer revealed novel intermolecular contacts between the NTD and CTD of adjacent subunits (Ganser-Pornillos et al., 2007). A recent HIV-1 CA-FL hexameric X-ray structure further confirmed these intermolecular interactions (Pornillos et al., 2009).
Here, we report a 16Å resolution cryoEM study of a tubular assembled HIV-1 CA in conjunction with a high resolution NMR structure of the CTD dimer. The atomic model of the CTD solution dimer fits extremely well into the EM density map, suggesting that this dimer interface is retained in the assembled capsid. Furthermore, a new interface formed by CTD dimers from three neighboring hexamers was revealed in the current cryoEM map at a pseudo-threefold axis, and mutational data confirmed the importance of the implicated intermolecular contacts.
Tubular assemblies of full-length wild-type HIV-1 CA and several CA mutants, including CA A92E, were prepared and analyzed. Each of the CA-FL variants assembled in vitro into a variety of morphologies, including single- or multi-walled, long helical tubes, cones, and cone-tube hybrids in the presence of high salt (Figure 1A&B), consistent with previous studies (Li et al., 2000). These in vitro assemblies likely contain subunit interactions present in the assembled native HIV-1 capsid, since they exhibited diameters similar to those of native cores (Briggs et al., 2003). CA A92E formed long, well-separated, single-walled, straight tubes (Figure 1B) compared to wild-type CA and other mutants that typically yielded mostly bundled tubes (Figure 1A). For the CA A92E tubes, the helical structures were highly ordered and exhibited layer lines to 9.5Å resolution (Supplementary Materials Figure S1A). Since CA A92E consistently produced more homogeneous and the best quality tubes, we used this variant for three-dimensional (3D) structure determination by cryoEM and image reconstruction.
Analysis of low-dose images of frozen-hydrated tubes resulted in a number of helical families with variable unit cell dimensions and tube diameters ranging from 392 to 480Å. We selected the three highest quality images recorded at different defocus values (Supplementary Materials Figure S1B) from the tubes belonging to the (−13,11) helical family, which displayed well-ordered layer lines (Figure 1C&D). To minimize artifacts from tube distortions that are intrinsic to HIV-1 CA assemblies, we employed real-space 3D reconstruction. The final density map at 16Å resolution (Figure 1E–G), indicated by the Fourier Shell Correlation of 0.5 (Supplementary Materials Figure S1C), was reconstructed by iterative helical real-space reconstruction (IHRSR) (Egelman, 2007) and refined with helical constraints and full contrast transfer function (CTF) correction (Sachse et al., 2007).
The single-walled tube structure exhibits a wall thickness of 80Å, corresponding to the height of full-length CA. The overall tube diameter is 420Å (Figure 1F&G) and matches closely diameters observed for native tubular cores (440±43Å) and conical cores (maximum diameter of 590Å) (Briggs et al., 2003). The dimensions of the surface unit cell are a=98Å, b=102Å, γ =113° (measured at radius=213Å). Consistent with the previous cryoEM map (Li et al., 2000), the NTDs form hexameric rings on the outer surface of the tube (Figure 1E), while CTDs from neighboring hexameric units connect subunits on the inner surface of the tube (Figure 1F&G), mediating the assembly of the overall pseudo-hexagonal surface lattice. The resulting density map displays two-fold symmetry with an overall two-fold phase residual of 19°. The local pseudo six-fold symmetry in the hexamer and pseudo three-fold symmetry between neighboring hexamers is apparent in the non-symmetrized density maps (Figure 2). For the (−13,11) helical tubes, three closest neighboring hexamers that contribute inter-hexamer interactions are derived from three helical arrays, the left-handed 13 start helix (n=−13), the right-handed 11 start helix (n=11), and the left-handed 2 start helix (n=−2) (Figure 2A&B). Interactions between the neighboring hexamers that are mediated by the CTD dimers are not identical along these three helices, resulting in distinct inter-hexamer distances. The 2 start helix exhibits the highest curvature (29°) and shortest inter-hexamer distance (94Å), whereas the 13 start helix shows an 11° rotation and 97Å shift between subunits. The existence of three somewhat different arrangements for the CTD dimers permits variable inter-oligomer interfaces in the native conical structures. We also observed strong densities at the local three-fold axis in the non-symmetrized density map (Figure 2D). These densities are associated with the CTD regions in the inner most layer of the capsid, and three adjacent CTD dimers surround the three-fold axis. This interface has not been described previously. The CTD-CTD interaction at the three-fold interface may play an important role in capsid structure.
In the density map of a tubular assembly, the hexameric arrangement of CA is very similar to that of the flattened 2D CA crystals, with similar surface unit cell dimensions (p6, a=b=92.7Å in the 2D crystals and a=98Å, b=102Å, γ =113° in the tube at the outer diameter). Therefore, it was possible to dock the hexameric model (PDB code: 3DIK, Ganser-Pornillos et al., 2007) into the density map by automated rigid-body fitting using a correlation-based method. In addition, three other docking strategies were employed using (1) the monomeric model, (2) a model generated by preserving the NTD-CTD interface, or (3) the isolated NTD and CTD models from 3DIK. The resulting pseudo-atomic models are very similar, exhibiting only slight variations (Supplementary Materials Figure S2). The separate domain docked model, which exhibited slightly better Fourier Shell Correlation against the density map (Supplementary Materials Figure S1D), fits very well into the density map (Figure 2B–D). The three previously characterized inter-molecular interfaces, NTD-NTD, NTD-CTD, and CTD-CTD (Ganser-Pornillos et al., 2007; Pornillos et al., 2009), exist in the current tubular structure. A slightly asymmetric arrangement for the six NTDs (Figure 3A) is noted, and more pronounced asymmetry for the three CTD dimers is seen (Figure 3B). The arrangement of monomers also deviates somewhat from six-fold symmetry. This results in a shift of the molecules away from the pseudo six-fold axis along the curvature (Figure 3C&D, monomers 3&6, orange), compared to the model derived from hexamer docking (blue). This asymmetry is in clear contrast to the p6 symmetric arrangement seen in the 2D crystals. Indeed, the tubular nature and high curvature (up to 29° rotation between adjacent hexamers) that is correlated with variable inter-subunit distances (94 to 97Å) is most likely responsible for the altered arrangement between CTDs (Figure 3B). Likewise, formation of two different cowpea chlorotic mottle virus capsid assemblies has been attributed to small changes in the packing angle between two CA monomers at the dimer interface (Tang et al., 2006).
To further examine the protein-protein contacts that contribute to capsid assembly, we investigated the solution behavior of the full-length CA protein. Previous NMR studies were hampered by poor spectral quality due to resonance broadening (Bosco and Kern, 2004) caused by monomer-dimer exchange (Rose et al., 1992). For our wild-type CA protein, we compared its 1H,15N HSQC spectrum to those of the individual isolated domains (Figure 4A). Resonances arising from residues in the spectra of the individual domains exhibit very similar chemical shifts to the corresponding ones in the spectrum of CA-FL (Figure 4A), demonstrating that no gross structural differences are present. For example, most, if not all, of the N-terminal resonances in CA-FL overlap with those of the NTD. Only a subset of resonances associated with the CTD, however, are seen in the CA-FL spectrum. Furthermore, resonances only observed in the spectrum of the isolated CTD domain, and not in that of CA-FL, are very weak as a result of severe line broadening. This behavior is associated with milli- and micro-second timescale exchange involving an association/dissociation between the CTDs, either alone or in the full-length CA context, as evidenced by dilution studies (Figure 4B). At 2mM, additional resonances are observed that are not present in the spectrum at low concentration (12.4μM). These resonances arise from residues in the CTD dimer interface, are of low intensity, and are not observed in the spectrum of CA-FL. Both monomer and dimer resonances for Ser146, Thr148, Phe168, and Thr171 in the CTD (residues 144–231) 1H,15N HSQC spectrum at 12.4μM were assigned unambiguously and were used to calculate the dissociation constant (Kd) of the dimer. The resulting Kd value of 9.8±0.6μM is in excellent agreement with that (10±3μM) determined for the CTD construct containing CA residues 146–231 by equilibrium sedimentation (Gamble et al., 1997).
An investigation of the CTD by multi angle light scattering (SEC-MALS, Supplementary Material Figure S3) revealed a small degree of concentration dependence: a 5-fold dilution in protein concentration resulted in a change in the measured weight-averaged molecular mass from 18.9 to 17.9kDa (change in elution time from 33.5 to 33.6 minutes). The measured molecular mass is only slightly less than the expected dimer mass (19.6kDa), confirming that the CTD exists mostly as a dimer at these concentrations. The SEC-MALS data on CA exhibited a more pronounced concentration dependence: 5-fold sample dilution shifted the measured molecular mass from 43.7 to 37.8kDa. Therefore, full-length CA interacts more weakly than the isolated domain, which agrees with the higher Kd value of 18±0.6μM (Gamble et al., 1997). The fact that, in the 1H,15N HSQC spectrum of CA-FL, only resonances of the CTD are broadened, with those of the NTD remaining essentially identical to those in the isolated NTD spectrum, supports the notion that the NTD is flexibly linked to the CTD and tumbles independently in solution.
In summary, the NMR data clearly demonstrate that, in solution, only the CTD is responsible for dimerization, without any significant contribution from an NTD-NTD and/or NTD-CTD interaction.
Since severe line-broadening was present in all spectra of CA-FL, no NMR structure determination was possible. However, the isolated CTD spectra were of sufficient quality to permit us to solve the solution structure by NMR. Nearly complete (>95%) backbone and side chain assignments for the dimeric CTD were obtained using a 2mM uniformly 13C,15N-labeled CTD sample in which CTD exists predominantly as a dimer (~95%). Analysis of 3D-13C/15N-filtered, 13C-edited NOESY data recorded on a 2.8mM mixed labeled/natural abundance sample (~96% dimer), yielded intermolecular NOEs that clearly delineated the dimer interface.
The NMR structure of dimeric CTD was solved using methodology commonly employed in our laboratory, based on intra- and inter-molecular NOE and H-bond distances and dihedral angle, H-N, and Hα-Cα RDC constraints (Supplementary Material Table S1). The structure is well-defined, satisfies all experimental constraints, displays excellent covalent geometry, and exhibits atomic r.m.s. deviations of 0.41±0.08Å and 0.82±0.07Å with respect to the mean coordinate positions for the backbone (N, Cα, C′) and all heavy atoms, respectively. A 30-conformer ensemble is displayed in Figure 5A. Critical side chains involved in the dimer interface are shown in the backbone ribbon structure in Figure 5B. The interface is characterized by extensive hydrophobic interactions between residues at the N-terminus of the CTD, such as Tyr145, Thr148, and Leu151, and residues residing in helix 9, such as Val181, Trp184, Met185 and Val191. Note that the side chain of Tyr145 plays a pivotal role at the dimer interface (Figure 5C) as supported by numerous intermolecular NOEs (Figure 5D). This amino acid resides at the junction between NTD and CTD. In previous X-ray NTD structures, Tyr145 is exposed to solvent and not involved in any interactions. In contrast, all CTD crystal structures solved so far do not contain Tyr145, using sequences from 146 to 231 or 151 to 231. Our current CTD structure clearly illustrates the important role of Tyr145 for the CTD interface. Indeed, Y145 mutations exhibited a modest impairment in particle assembly (Supplementary Materials Figure S4H) and are seriously defective in capsid assembly, as evidenced by profoundly reduced viral capsid stability, infectivity and RT activity (Figure 5E). In addition, the Y145A/F mutants failed to assemble into tubes in vitro and formed particles containing aberrant cores (Supplementary Materials Figure S4). These results demonstrate the critical involvement of Tyr145 in capsid assembly.
Although many crystal structures of dimeric CTD are currently available (Gamble et al., 1997; Ternois et al., 2005; Worthylake et al., 1999), no solution structure of a CTD dimer has been reported to date. Only structures of monomeric units were determined by NMR (Alcaraz et al., 2007; Wong et al., 2008). The available X-ray structures of CTD display considerable variability at their dimerization interfaces, while the structures of the monomeric units are very similar (backbone atomic r.m.s.d. of ~ 0.7Å). The current solution dimer exhibits an interface distinct from those observed in the crystal structures. The backbone (N, Cα, C′) atomic r.m.s.d. between our NMR dimer solution structure and the various dimer crystal structures, 1BAJ (Worthylake et al., 1999), 1A8O (Gamble et al., 1997) and 2BUO (Ternois et al., 2005), are 3.8, 7.6, and 9.5Å, respectively, while the corresponding r.m.s.d. values for the monomeric units are all within 1.7Å (Figure 6A). The solution dimer structure is most different from 2BUO and 1A8O and is more similar to the 1BAJ and 1A43 dimer structures. The differences are predominantly determined by the crossing angles of helices 9 and the shape of their arrangement (Figure 6B). In the NMR solution structure, these two helices form a V-shape with a crossing angle of ~56° with their point of closest approach at their N-termini, opening up towards their C-termini. In the 1BAJ and 1A43 structures, the helices are arranged with a crossing angle of ~43° and 38°, respectively, with very little opening. The 1A8O and 2BUO structures are quite different: helices 9 display crossing angles of ~140° and 121° with no fanning at the C-termini. Note that our NMR structure and the crystal structures were determined using different protein constructs and experimental conditions, including different media (solution vs. crystal lattice), oxidation states (reduced vs. disulfide bonded Cys198 and Cys218), and the presence of ligands. For example, the crystal structure constructs were 2–7 amino acid shorter, missing Tyr145 at the N-termini, compared to the current NMR structure. The active involvement of Tyr145 at the dimer interface in solution (Figure 5B–D) indicates that any shortening of the sequence at the N-terminal end of the CTD constructs might contribute to the structural differences observed between the solution and crystal structures.
A high resolution crystal structure of full-length HIV-1 CA was recently determined in which the NTD hexamer was stabilized by thiol cross-linking and CTD dimerization was disrupted by introducing the W184A/M185A double mutation (Pornillos et al., 2009). In this structure, Tyr145 seems to belong to the NTD, although one must keep in mind that the CTD dimer interface is greatly perturbed by the mutation. In addition, the N-terminal two-thirds of helix 9 exhibits undefined electron density, possibly due to partial unfolding, and such changes in helix 9 have also been noted in an NMR structure of the monomeric W184A/M185A CTD mutant (Wong et al., 2008). In contrast, the present structure of the wild-type CTD dimer (Figure 5A&B) clearly shows a well-defined helical conformation for the entire helix 9.
We also examined how well our solution CTD dimer structure fits to the CA full-length structure from flattened 2D sheets (Ganser-Pornillos et al., 2007). Contrary to the poor fit seen with the crystal structures, a very good fit was noted for the NMR CTD dimer, yielding backbone Cα atomic r.m.s.d. values of 1.7Å for the symmetry related CTD dimers in the 2D crystalline sheets (far right image in Figure 6A&B). The CTD structure in the sheets also exhibits a V-shape for helixes 9 with a slightly larger crossing angle (70°) and a slightly larger separation (Figure 6B), which may be caused by the lower resolution (9Å) of the EM model. The close resemblance between the solution dimer and the EM CTD dimer suggests that the solution interface is retained in the assembled capsid. In addition, the average backbone Cα atomic r.m.s.d. values between the NMR CTD dimer and the domain fitted structure in the tubular capsid assembly were 2.4Å for the D1 dimer and 2.8Å for the average of three dimers (Figure 3B). These values are significantly smaller than the r.m.s.d. values between the NMR structure and any of the X-ray CTD dimer structures (ranging 3.8Å to 9.6Å). We, therefore, rebuilt our tubular EM model using rigid-body docking of monomeric NTD and dimeric solution CTD into the EM map. The NMR CTD dimer docked model (Figure 6C) indicates a non-symmetric trimer arrangement of CTD dimers, compared to the symmetric array seen in the 2D sheets (Figure 6D).
The pseudo-atomic model, reconstructed from CA tubular assembly, allowed the delineation of four inter-molecular interfaces: the NTD-NTD; the NTD-CTD, essentially as described previously (Ganser-Pornillos et al., 2007); the CTD-CTD dimer interface, similar, but slightly different than seen before; and a novel interface located at the local three-fold axis. The first two interfaces are intimately connected to the hexameric ring structure, while the latter two involve adjacent CTDs and are responsible for inter-hexamer interactions and stabilization of the extended lattice.
The novel CTD-CTD interface at the inner surface of the tube was evident in our raw density map (Figure 2D, ,7A),7A), as well as in the reconstructed trimer model (Figure 7B). This additional CTD-CTD interface has not been described before. It involves pairs of helices, H10 and H11, with H10 from one CTD in close proximity to H11 from a neighboring CTD. Although the limited resolution of our density map precludes resolving α-helical elements in the tube structure and many details of the capsid structure will have to await higher resolution studies, some important features of this new CTD-CTD interface are clearly evident. The interacting helical face of H10 consists of residues Lys203, Ala204 and Pro207 while that of H11 comprises Glu213, Thr216, Ala217 and Gln219. Several biochemical and genetic studies have identified functionally important residues that reside in H10 and H11, including Lys203, Glu212, and Gln219. Lys203 and Gln219 mutations destabilize the capsid, causing markedly reduced infectivity (Forshey et al., 2002; Ganser-Pornillos et al., 2004; von Schwedler et al., 2003). Interestingly, close distances, averaged between the three asymmetric pairs, of 6.5Å and 6.0Å for the side chains and ~9Å for the backbone between Lys203 and Gln219 and Thr216 and Pro207, respectively, were observed (Figure 7B). To probe and verify contacts between these residues, double cysteine mutations were introduced at K203/Q219 and P207/T216 for chemical cross-linking. The P207C/T216C mutant exhibited reduced infectivity, whereas the K203C/Q219C mutation resulted in noninfectious particles (Figure 7C), consistent with the importance of K203 and Q219 in maintaining capsid stability (Forshey et al., 2002). Intermolecular cross-linked species corresponding to apparent dimers and trimers of CA accumulated when the P207C/T216C mutant virus particles were treated with oxidizing reagents (Figure 7D), lending further support to the close spatial arrangement at the threefold axis. The cross-linked bands ran slightly more slowly than expected, possibly due to altered mobility after crosslinking (Phillips et al., 2008).
Our structural model also suggested that Glu213 at the trimer interface may play an important role in capsid assembly or disassembly. This was tested by generating E213A and E213Q mutants, both of which exhibited reduced viral infectivity (Figure 7E), without altering mature particle formation (Supplemental Materials S5). Cores isolated form the E213A/Q mutant viruses contained elevated levels of CA and uncoated more slowly in vitro than the wild-type (Figure 7E&F), indicating that the defect involves a hyperstable capsid.
A fullerene cone model of the HIV-1 CA core was proposed based on early EM analysis of native cores and in vitro assemblies (Ganser et al., 1999; Li et al., 2000). Our results from tubular assemblies of HIV-1 CA strongly support this model. Furthermore, on the basis of the packing interactions in our pseudo-atomic model of three adjacent CA hexamers, we propose a mechanism for the curved and asymmetric assembly of CA. To accommodate non-planar assembly, structural alterations from a flat P6 lattice are necessary, and a slight tilt of the NTD along the curved surface is indeed seen in our model, suggesting some variability of the NTD-NTD interface (Figure 3C&D). However, the most pronounced changes involve the CTD (Figure 3C&D, ,6D),6D), either at the CTD dimer interface (Figure 3B), the trimer interface of CTD dimers (Figure 6C&D), or both. Nevertheless, our solution NMR CTD dimer structure is well accommodated in the assembled 2D sheets and tubes. Given that the CTD-CTD dimer interface exhibits extensive hydrophobic contacts, it may be less variable than the NTD-NTD, NTD-CTD (Ganser-Pornillos et al., 2007; Pornillos et al., 2009), and CTD trimer interfaces (this study, Figure 7B), which predominantly involve polar groups. Therefore, the quasi-equivalence and asymmetry in the assembly of cones and tubes most likely involves the hinge between NTD and CTD (Supplementary Materials Figure S2), resulting in variable trimer interfaces. Such variability between NTD and CTD has been suggested previously (Cardone et al., 2009; Pornillos et al., 2009). We propose that the newly characterized trimer interface, involving H10 of one hexamer and H11 from an adjacent hexamer, is likely to play a key role in uncoating, as evidenced by the capsid-destabilizing effects of amino acid substitutions at Lys203 and Gln219 (Forshey et al., 2002; von Schwedler et al., 2003) and the hyperstabilizing effect of E213 mutations (this study). Further structure-guided mutational studies as well as biochemical investigations of capsid assembly and uncoating in vitro and vivo will allow characterization of the molecular interactions at this new interface in more detail.
One important feature of the HIV-1 capsid is its intrinsic asymmetry; therefore, multiple levels of interaction and flexible interfaces are necessary to impart stability to the native core architecture, while also permitting controlled disassembly in the uncoating step of infection. In the current CA tubes, different arrangements between adjacent hexamers are present, mediated through varying relative orientation of the NTD and CTD while likely preserving the CTD-CTD dimer interface. This allows for flexibility and plasticity in the formation of the varied morphologies seen in HIV-1 capsids and those of other retroviruses. The adjustable and distinct inter-molecular interfaces in the tubular assembly described here provide the potential means for the formation of various types of curved and asymmetric retroviral capsids starting from a generic hexameric lattice.
The cDNA encoding gag polyprotein, pr55gag was obtained from the NIH AIDS Research and Reference Reagent Program, Division of AIDS, NIAID, NIH (Erickson-Viitanen et al., 1989). Regions encoding CA-FL (Pro1-Leu231) and CA-CTD (Met144-Leu231) were amplified and subcloned into pET21 (EMD chemicals, Inc. San Diego, CA) using NdeI and XhoI sites. Proteins were expressed in E. coli, Rosetta 2 (DE3), cultured in Luria-Bertani media or modified minimal medium, and induced with 0.4mM IPTG at 23°C for 16h. All proteins were purified over a 5mL Hi-Trap QP column (GE Healthcare, Piscataway, NJ) in 25mM sodium phosphate, pH 7.0, 1mM DTT, and 0.02% sodium azide. CA-FL and CA-CTD were further purified over 5 mL Hi-Trap SP columns (GE Healthcare, Piscataway, NJ) using a 0–1M NaCl gradient in 25mM sodium phosphate, pH 5.8, 1mM DTT, and 0.02% sodium azide. The final purification step for all proteins comprised gel-filtration over Hi-Load Superdex 200 16/60 columns (GE Healthcare) in 25mM sodium phosphate, pH 6.5, 100mM NaCl, 1mM DTT, and 0.02% sodium azide. Proteins were concentrated to 10mg/mL using Amicon concentrators (Millipore, Billerica, MA), flash-frozen with liquid N2 and stored at −80°C. Molecular masses of purified CA proteins were confirmed by LC-TOF mass spectrometry (Bruker Daltonics, Billerica, MA). Uniform 15N and 13C-labeling of CA-CTD was carried out by growth in modified minimal medium using 15NH4Cl and 13C6- glucose as the sole nitrogen and carbon sources, respectively. A mixed 13C/15N-labeled and unlabeled (1:1 ratio) CA-CTD sample was prepared by un/refolding labeled and unlabeled CA-CTD protein as described in Supplementary Materials. The A92E mutant CA protein was expressed in E. coli and purified by ion exchange chromatography as described previously (Yang and Aiken, 2007).
CA-FL proteins (10 mg/ml) were diluted to 2 mg/ml in high salt buffer (50mM Tris pH 8.0, 1M NaCl) and incubated at 37°C for 1hr for tubular assembly. Samples (3–5μl) were applied to glow discharged perforated Quantifoil grids (Quantifoil Micro Tools, Jena, Germany), blotted and plunge-frozen in liquid ethane using a manual gravity plunger. Low dose (10e−/Å2) projection images were collected using a Tecnai Polara microscope at 200 KV cooled at liquid nitrogen temperature. Images were recorded on a Gatan 4K × 4K CCD camera or Kodak SO163 films at a nominal magnification of 59,000 and under focus values ranging from 1.0 to 2.5 μm. For films, the best images were selected and digitized using a Nikon super coolscan 9000 ED scanner (Nikon, Japan) at a resolution of 4000 dpi.
Selected tube images were Fourier transformed and indexed for their helical symmetries. Only tubes within the same helical family (−13,11) were included for further processing and reconstruction. The tube images were binned to a pixel size of 2.15Å and boxed into small segments of 332×332 pixels with an overlap of 90% along the helical axis by EMAN boxer program (Ludtke et al., 1999). An initial density map was calculated using the IHRSR real-space processing package (Egelman, 2007). The structure was further refined iteratively with helical constraints and CTF correction using software developed for high resolution helical refinement (Sachse et al., 2007). A total of 270 segments were included in the final round of refinement. The helical symmetry was imposed only in the final density map. MRC based helical processing software (Yonekura and Toyoshima, 2000, 2007) was used for initial Fourier transformation, helical indexing, and repeat distance calculation. The resulting density maps were visualized with Chimera (Pettersen et al., 2004).
Pseudo-atomic models of a CA hexamer were constructed by docking of the following models (PDB code 3DIK, (Ganser-Pornillos et al., 2007) into tubular EM density map using an automated map fitting feature implemented in Chimera: (1) hexamer; (2) monomer; (3) separate NTD and CTD domains; (4) combined NTD from one molecule with CTD from adjacent molecule to preserve the NTD-CTD interface; (5) separate NTD and NMR CTD dimer. Pseudo-atomic trimer models of hexamers were constructed by combining the models of three closest neighbors.
All NMR spectra for the structure determination of CA-CTD were recorded at 25°C on NMR samples containing 2mM (13C/15N-labeled) or 2.8mM (mixed) CA-CTD in 25mM sodium phosphate buffer, pH 6.5, containing 0.02% azide and 2mM DTT using Bruker AVANCE900, 800, 700, and 600 spectrometers, equipped with 5mm triple-resonance, three-axes gradient probes or z-axis gradient cryoprobes. Backbone and side chain resonance assignments were carried out using HNCACB, HN(CO)CACB, HNCA, HN(CO)CA, HBHA(CO)NH and HCCH-TOCSY experiments (Bax and Grzesiek, 1993; Clore and Gronenborn, 1998) at 700 MHz. Distance constraints were derived from 3D simultaneous 13C- and 15N-edited NOESY (Sattler et al., 1995), 3D 13C-edited NOESY, and 2D NOESY experiments on the 13C/15N-labeled sample and from 3D 13C/15N-filtered, 13C-edited NOESY experiment on the mixed CA-CTD sample (Lee et al., 1994). All NOESY spectra were acquired at 900 MHz using a mixing time of 120ms. Residual HN and HαCα dipolar couplings were measured using in-phase/anti-phase (Ottiger et al., 1998) 2D 1H-15N HSQC and 3D coupled-HN(CO)CA experiments at 600 MHz. Two different alignment media were employed, a C12E5/hexanol mixture (5% w/v, molar ratio 0.96) (Ruckert and Otting, 2000) and a colloidal pf1 phage solution (10mg/mL) (Hansen et al., 1998). The CA-CTD samples for RDC data collection contained 100mM NaCl. Spectra were processed with NMRPipe (Delaglio et al., 1995) and analyzed using SPARKY3 (version 3.115, Goddard and Kneller, 2004). HNCACB and HN(CO)CACB spectra for backbone assignments of CA-NTD were recorded on a 0.94mM sample of 13C/15N-labeled CA-NTD at 900 MHz. 1H,15N HSQC spectra of CA-FL, CA-NTD and CA-CTD at concentrations ranging from 6μM to 2.8mM were collected to assess self-association.
All NOE cross peaks were picked using SPARKY3 (version 3.115, Goddard and Kneller, 2004) and manually inspected/sorted for accuracy. Intermolecular NOE cross-peaks were assigned manually from the 3D filtered/edited NOESY data, while all other NOE cross peaks were initially assigned using the CANDID algorithm in CYANA (Herrmann et al., 2002) in an automated fashion. The chemical shift input table contained >95% resonance assignments for the CA-CTD dimer. The CYANA structure calculations were performed including 206 backbone torsion angle constraints ( and ψ) from TALOS (Cornilescu et al., 1999) and 96 hydrogen bond constraints identified via H-D exchange and NOESY data analysis (1.8–2.3Å for H-O and 2.7–3.3Å for N-O). CYANA generated a total of 4066 NOE distance constraints (which include 198 inter-subunit constraints) and structure models that were used as initial constraints and models to generate the simulated annealing (SA) structures in XPLOR-NIH using a slow cooling protocol (Brunger, 1992; Schwieters et al., 2003). The final SA CA-CTD dimer structures were obtained in an iterative fashion by extensive, manual cross-checking of all distance constraints against the 3D NOESY data sets and the generated structures. Residual dipolar coupling (RDC) constraints, measured in two alignment media (pf1 phage and C12E5/hexanol mixtures, respectively), were added in the final SA calculations. 500 structures were generated and the 30 lowest energy structures were selected and analyzed using PROCHECK-NMR (Laskowski et al., 1996) and PALES (Zweckstetter and Bax, 2000). For the final 30 conformer ensemble, 99.7% of all residues are located in the most favored and additionally allowed regions of the Ramachandran plot. All structure figures were generated with MOLMOL (Koradi et al., 1996) or PyMOL (DeLano Scientific). The atomic coordinates and NMR constraints have been deposited in the RCSB Protein Data Bank under accession code 2KOD.
Mutations were generated by PCR amplification and transferred into the full-length HIV-1 molecular clone R9 (Gallay et al., 1997). Viruses were produced by transfection of 293T cells, and infectivity was quantified by titration on Hela-P4 cells, as described previously (Yang and Aiken, 2007). Capsid stability was determined by measuring CA protein after purification of cores and the kinetics of uncoating were followed by quantitation of CA release from purified cores at 37°C (Aiken, 2009).
For Cys crosslinking studies, viruses were pelleted by ultracentrifugation (100,000×g, 30min) and disulfide bond formation was induced by oxidation (Phillips et al., 2008). Viral particles were incubated with copper phenanthroline (60μM CuSO4, 267μM o-phenanthroline), mixed by vortexing for approximately 5sec, and immediately quenched with 20mM iodoacetamide and 3.7mM Neocuproine (Sigma). Where indicated, samples were withdrawn from the reactions, treated with 2-mercaptoethanol to reduce cysteine bonds, denatured, separated by non-reducing SDS-PAGE and CA was detected by immunoblotting with rabbit anti-CA serum.
We thank Dr. Joanne Yeh for useful discussions and Dr. Teresa Brosenitsch for critical reading of the manuscript. We thank Drs. Koji Yonekura, Edward Egelman and Niko Grigorieff for sharing their image processing software and Dr. Jing Zhou for technical assistance. This work is a contribution from the Pittsburgh Center for HIV Protein Interactions and was supported by the National Institutes of Health (GM082251 and AI076121).