Hepatitis B Virus Capsids Have Diverse Structural Responses to Small-Molecule Ligands Bound to the Heteroaryldihydropyrimidine Pocket

ABSTRACT

Though the hepatitis B virus (HBV) core protein is an important participant in many aspects of the viral life cycle, its best-characterized activity is self-assembly into 240-monomer capsids. Small molecules that target core protein (core protein allosteric modulators [CpAMs]) represent a promising antiviral strategy. To better understand the structural basis of the CpAM mechanism, we determined the crystal structure of the HBV capsid in complex with HAP18. HAP18 accelerates assembly, increases protein-protein association more than 100-fold, and induces assembly of nonicosahedral macrostructures. In a preformed capsid, HAP18 is found at quasiequivalent subunit-subunit interfaces. In a detailed comparison to the two other extant CpAM structures, we find that the HAP18-capsid structure presents a paradox. Whereas the two other structures expanded the capsid diameter by up to 10 Å, HAP18 caused only minor changes in quaternary structure and actually decreased the capsid diameter by ∼3 Å. These results indicate that CpAMs do not have a single allosteric effect on capsid structure. We suggest that HBV capsids present an ensemble of states that can be trapped by CpAMs, indicating a more complex basis for antiviral drug design.

IMPORTANCE Hepatitis B virus core protein has multiple roles in the viral life cycle—assembly, compartment for reverse transcription, intracellular trafficking, and nuclear functions—making it an attractive antiviral target. Core protein allosteric modulators (CpAMs) are an experimental class of antivirals that bind core protein. The most recognized CpAM activity is that they accelerate core protein assembly and strengthen interactions between subunits. In this study, we observe that the CpAM-binding pocket has multiple conformations. We compare structures of capsids cocrystallized with different CpAMs and find that they also affect quaternary structure in different ways. These results suggest that the capsid “breathes” and is trapped in different states by the drug and crystallization. Understanding that the capsid is a moving target will aid drug design and improve our understanding of HBV interaction with its environment.

INTRODUCTION

Hepatitis B virus (HBV) causes degenerative liver disease and is the leading cause of liver cancer, with 240 million chronically infected individuals (1, 2). Current antiviral therapies control the progression of the disease but fail to eliminate the virus (3, 4). There is a need for improved therapies to combat chronic infections. One approach is to target virus assembly (5–7).

HBV is an enveloped, double-stranded DNA virus with an icosahedral nucleoprotein core. The HBV capsid is the protein shell of the core. Beyond genome protection, the capsid is involved in intracellular trafficking, interaction with nuclear import machinery, regulation of reverse transcription, signaling, completion of reverse transcription, RNA chaperoning, and envelope acquisition (8). Capsid assembly is central to HBV replication. In vivo HBV capsids assemble around an RNA copy of the viral genome bound to the viral reverse transcriptase to form immature HBV cores (9). These cores mature in the host cytoplasm through the reverse transcription of the pregenomic RNA to a relaxed circular double-stranded DNA (10).

A homodimer is the basic functional unit of the HBV core protein (Cp) (Fig. 1d) (11). Cp has no human homolog, which contributes to its value as an antiviral target. Cp dimers self-assemble to form capsids in a tightly regulated and dynamic process (12–14). Wild-type Cp is 183 residues long and consists of two domains: the N-terminal assembly domain (residues 1 to 149), which forms the contiguous shell, and the C-terminal RNA-binding domain (residues 150 to 183) (15). The assembly domain (Cp149) is capable of self-assembly in vitro and in vivo, forming capsids that are indistinguishable from wild-type capsids (15–17). Comparisons of free and capsid-constrained Cp149 reveal a stable central chassis connected by glycine or proline hinges to subdomains: the spikes (residues 63 to 94), interdimer contact region (residues 111 to the C terminus), and a fulcrum that mechanically connects the other two subdomains (residues 10 to 25) (18).

FIG 1.

Core protein allosteric modulators and the Cp dimer. (a) HAP18; (b) HAP1; (c) AT-130; (d) a Cp dimer with secondary structural elements labeled.

The capsid has a T=4 icosahedral symmetry, whereby the 60 asymmetric units are each comprised of four Cp monomers (designated A, B, C, and D or AB dimers and CD dimers) in quasiequivalent environments that differ slightly from each other (Fig. 2a) (19, 20). The A subunit interfaces with other A subunits to form the 12 5-fold vertices of the capsid. The 20 3-fold vertices contain three CD dimers each. The 30 quasi-6-fold vertices (2-fold symmetry) contain two each of the B, C, and D subunits. Crystal structures, showing multiple conformations of Cp in capsid and dimer forms, have been determined (11, 18, 21–23).

FIG 2.

Crystal structure of HAP18-bound capsid and its comparison to the apo-capsid. (a) A cartoon representation of the quaternary structure of an HBV capsid complexed with HAP18. A capsid asymmetric unit of two dimers is highlighted by surface representation, with the A subunit in blue, B subunit in red, C subunit in green, and D subunit in yellow. The positions of the HAP18 molecules in the asymmetric unit are marked with numbered black circles: 1, B-associated HAP18; 2, C-associated HAP18. (b) Close-up of a CD dimer showing HAP18 (sphere representation) in the C subunit binding pocket, which in the context of a capsid will be capped by the neighboring D subunit. The bottom of this image is on the interior of the capsid; the partially unraveled four-helix bundle forms the spike that punctuates the capsid outer surface. (c) A stereoimage of the CD dimer (red) with the 2Fo-Fc map (blue) at 1.5 σ contouring shows the fit of the model to the electron density.

In vitro HBV capsid assembly depends on a nucleus, believed to be a trimer of dimers, followed by rapid elongation (14). Assembly is entropically driven by burial of the hydrophobic surface (12, 24). Dimers associate with weak contact energy, which allows “thermodynamic editing” to remove incorrectly and weakly associated dimers (25). Aggressive assembly conditions (increased temperature, Cp concentration, and ionic strength) that result in stronger association energies overnucleate the assembly reaction and lead to the formation of kinetically trapped intermediates (12, 25, 26).

Small-molecule core protein allosteric modulators (CpAMs) (Fig. 1) can activate Cp assembly and lead to the formation of a variety of nonuniform kinetically trapped structures (27–30). They act by favoring an assembly-active form of Cp and by stabilizing protein-protein interactions (21, 27, 30–32), implying a competition between functional assembly and misassembly (29, 33). Besides being noninfective, misassembled structures also sequester and deplete functional Cp dimers that would otherwise be involved in virion assembly. These molecules are predominantly hydrophobic and were originally discovered as potent nonnucleoside inhibitors of HBV replication (31). Heteroaryldihydropyrimidines (HAPs) and phenylpropenamides are two families of CpAMs that lead to suppression of HBV infection (34, 35). In vitro studies have established a strong correlation between inhibition of HBV replication and HAP-induced assembly (29). At saturating concentrations, some CpAMs (e.g., HAPs but not phenylpropenamides) induce the formation of aberrant structures (36). Crystal structures have been determined for preassembled HBV capsids bound to a representative HAP molecule, HAP1, and a representative phenylpropenamide, AT-130 (Fig. 1) (21, 23). These molecules bind to the same hydrophobic pocket at a dimer-dimer interface. Both affect the quaternary structure of the capsid, altering the spatial relationship between dimers. These changes were large enough to cause these capsids to crystallize with different unit cell parameters from capsids without bound CpAM (apocapsids). However, while the quaternary changes observed for these structures were in the same mode, i.e., expansion around the 5-fold axes, the extent of these changes were distinctly different; HAP1 induced a much more dramatic quaternary rearrangement than AT-130. Also, there was not a consensus on a preferred binding site; HAP1 bound preferentially to C subunits, while AT-130 bound preferentially to B subunits. AT-130 binding also resulted in significant tertiary structural changes, while HAP1 binding did not (23). These tertiary changes were suggested to be compensatory, to allow the formation of morphologically normal capsids, in contrast to HAP1, where the quaternary changes alone appeared to result in assembly misdirection.

Based on the structure of the HAP1-bound capsid, we developed a series of HAP derivatives that differentially affected assembly (29). One of these compounds, HAP18 (Fig. 1C), increased the stability of protein-protein interactions (considerably more than HAP1), increased the rate of Cp assembly (less than HAP1), and induced the formation of large, regular complexes—believed to be tubes—that were hundreds of micrometers long (29). As HAP18 is larger than other HAPs, its orientation in the HAP site was likely to be readily discerned, even at a modest resolution. Thus, to confirm the HAP binding site, identify the orientation of HAP and its derivatives in the binding site, and understand the effect of CpAM size on HBV capsid structure, we determined the crystal structure the HBV capsid in complex with HAP18 at a 3.9-Å resolution. Comparison to other CpAMs shows a diversity of conformational responses, suggesting that these molecules allow us to observe individuals from an ensemble of structures.

MATERIALS AND METHODS

Sample preparation.

Hepatitis B subtype adyw 3CA Cp150 (a Cp149 variant with the three native cysteines mutated to alanine and a C-terminal cysteine appended) capsid protein dimers were expressed and purified as described in detail previously (37, 38). Briefly, Escherichia coli cells expressing the capsid protein were lysed by sonication and the assembled HBV capsid in the cell lysate was purified by size exclusion chromatography. T=4 and T=3 3CA Cp150 capsids from pooled fractions were isolated by sucrose gradient centrifugation and immediately dialyzed into 10 mM HEPES buffer at pH 7.5 with 150 mM NaCl.

Crystallization.

3CA Cp150 T=4 capsids at a 17.5-mg/ml concentration were mixed with a 1:1 molar concentration (based on the molar concentration of a Cp149 protein dimer) of a racemic mixture (R and S conformers) of HAP18 in dimethyl sulfoxide (DMSO) at a 2:1 molar ratio of HAP18 mixture to 3CA Cp150 dimer, prior to crystallization. The crystallization trials were carried out at room temperature with 4 μl sitting drops. The final crystallization well solutions were composed of 5 to 10% polyethylene glycol 5000 monomethyl ether, 0 to 5% polyethylene glycol 8000 monomethyl ether, 10 to 16% 2,3-butanediol, 100 mM Tris (pH 9.0), 150 mM NaCl, and 300 mM KCl. The crystals took about a week to grow, although diffraction quality crystals were relatively rare. The crystal used for this data set came from a well with 5% polyethylene glycol 5000 monomethyl ether, 5% polyethylene glycol 8000, 14% 2,3-butanediol, 100 mM Tris (pH 9.0), 150 mM NaCl, and 300 mM KCl.

Diffraction data collection.

Selected crystals for data collection were cryoprotected as described previously (21). Cryosoaked crystals were frozen in a stream of gaseous nitrogen at −170°C before they were transported in liquid nitrogen to the Advanced Photon Source (APS) synchrotron. Data were collected on the 14BMC beamline with an exposure time of 30 s, 0.2° oscillations, and a detector distance of 500 mm.

Structure determination and refinement.

Diffraction data were processed using the HKL2000 program (39). Molecular replacement and subsequent refinement were carried out with programs in the Phenix suite (40). A previously determined HBV capsid structure (PDB ID 1QGT) was used as the initial phasing model (11). Initial phases from the 1QGT model were calculated to a resolution of 6 Å. The phases were then extended in eight steps in subsequent refinement cycles to a final resolution of 3.89 Å. This is the same approach as that used with other HBV-CpAM structures to minimize model bias (21, 23).

For refinement of the final molecular model, global noncrystallographic symmetry restraints were used in the refinement with isotropic group B-factor refinement. This allowed averaging of the 60 copies of the icosahedral asymmetric unit comprised of one AB dimer and one CD dimer. Using Ramachandran restraints in the initial cycles of refinement improved the R factors, model fitting, and the overall bonds and angles. Manual model building with refinement in the density was performed with Coot (41). Though a 5% test set of structure factors was reserved for cross-validation, the values of R_work and R_free were highly correlated due to the extensive noncrystallographic averaging and did not provide a valid comparison. The models of the HAP18 R and S stereoisomers were generated using the Prodrg program (42). Geometrical restraints were generated for the molecule using the eLBOW program (43) in Phenix. The HAP18 was modeled into the binding site density after phase extension and visual inspection of the density. The model developed with Phenix.refine resulted in a final R_work of 27.8% and a correlation coefficient of 91.66% (Table 1). PyMOL (The PyMOL Molecular Graphics System, version 1.8; Schrödinger, LLC, New York, NY) was used to generate figures. The Maestro program (Maestro, version 10.1; Schrödinger, LLC, New York, NY) was used to identify interacting residues and generate two-dimensional (2-D) projections. Buried surface areas and binding energies were calculated using the PDBe PISA tool (44). Superpositions and root mean square deviation (RMSD) calculations were done using Phenix tools.

TABLE 1.

Crystallographic data collection and refinement statistics for HAP-18-bound capsid^a

Characteristic (unit)	Value
Data collection
Space group	C2
Unit cell parameters
a, b, c (Å)	529.37, 367.41, 542.00
Angles (°)	α = γ = 90, β = 104.79
Resolution (Å)	35.68–3.89 Å
Rsym(%)	13.3 (44.7)
Average I/σ	12.7 (4.3)
Completeness (%)	90.6 (77.8)
Redundancy	3.1 (2.8)
No. of unique reflections	677,647
Refinement
Noncrystallographic symmetry	60-fold
R factor (%)	27.8
Correlation coefficient (%)	91.66
No. of atoms	4,610
RMSD bond length (Å)	0.004
RMSD bond angle (°)	0.964
Ramachandran outliers (%)	3
Rotamer outliers (%)	11.2
Ramachandran favored (%)	88.1
Mean B factor size (Å2)
Protein	180.9
HAP18	267.1

^aValues in parentheses are for the highest-resolution shell of data (3.98 to 3.89 Å). R_free is not reported because it is highly correlated to R_work due to noncrystallographic symmetry averaging during refinement and is not statistically independent.

Protein structure accession number.

Coordinates and structure factors for HAP18-capsid have been deposited in the Protein Data Bank under accession code 5D7Y.

RESULTS

HAP18 leads to minimal quaternary and tertiary structural changes compared to the apocapsid.

For crystallography with CpAMs, we used a Cp149 variant that crystallizes more readily, 3CA Cp150, in which the three native cysteines were mutated to alanine and a C-terminal cysteine was appended (21). The C-terminal cysteine cross-links and stabilizes capsids. This mutant has been used previously for crystal structure determination in apo, HAP1-bound, and AT-130-bound forms (21, 23). 3CA Cp150 capsids and HAP18 were cocrystallized at a 2:1 HAP18-to-dimer ratio. Because HAP18 is a racemic mixture with only one active enantiomer, this meant that there was one active HAP per dimer (35). Like other CpAM-capsid crystallization experiments, initial crystals formed within 1 week of mixing protein with precipitant; by comparison, capsids without CpAMs take at least 1 month to crystallize (21). The crystal packing and unit cell dimensions for the HAP18-HBV capsids were very similar to those of previously investigated CpAM-HBV crystals and distinct from crystals of capsid without bound CpAM. Indeed, we hypothesized that CpAMs influence packing by distorting the capsid quaternary structure to favor a HAP-capsid unit cell and expected to see the same quaternary changes in the HAP18-bound capsid (21, 23).

When we examined the structure of the capsid, determined by molecular replacement, we were surprised to find that the quaternary structure of the HAP18-bound capsid was unlike that of the HAP1-capsid complex and instead was very similar to apo forms (PDB ID 1QGT and 2G33) (11, 21) (Fig. 2a). Previous HBV-CpAM complexes showed clear and significant deformation of the tertiary and/or quaternary structure, resulting in expansion of the average capsid diameter by as much as 10 Å, particularly around the 5-fold axes, while still retaining icosahedral symmetry (21, 23).

The structure of the HBV-HAP18 capsid was a T=4 arrangement of 120 tetravalent dimers (Fig. 2a). Each dimer is roughly shaped like an upside-down T with the intradimer contact, comprised of a four-helix bundle, forming a protruding spike (Fig. 2b and c). Interdimer contacts were made at the extremities of the dimer base. The surface of the capsid is fenestrated by two classes of pores: those between trimers of dimers at 3-fold and quasi-3-fold groups of trimers of dimers and those at quasi-6-fold vertices.

A superposition of the HAP18-bound capsid structure on the apo structure revealed that the capsid diameter was very similar (Fig. 3a). The HAP18-bound capsid, in fact, seemed to be radially compressed at the 5-folds by about 3 Å. Given the lack of striking global structural changes, in contrast to previous HBV-CpAM structures (Fig. 3b and c), detailed comparisons of the HAP18-capsid structure to the apo structure were then performed. We observed an absence of quaternary structure change at the level of the whole capsid, the two dimers in the icosahedral asymmetric unit, and the individual dimeric asymmetric units. The two-dimer icosahedral asymmetric unit from a whole-capsid-based superposition of the HAP18-bound structure on the asymmetric unit of the apocapsid had an RMSD of only 0.98 Å for α-carbons; we note that the C-terminal 20 amino acids demonstrate a systematic shift (Fig. 4). The relative positions of the Cp monomers were all but indistinguishable from the apocapsid.

FIG 3.

HAP18, HAP1, and AT-130 each lead to distinctly different changes in tertiary and quaternary structures at the level of the capsid. Overlays of CpAM-bound capsids (magenta) on apo-capsid (cyan) when viewed down the 5-fold reveal systematic differences in capsid structure in the exterior (upper panels) and interior (lower panels). (a) There is a negligible size difference between the HAP18-bound capsid and the apo-capsid. Variegation of the coloring on interior and exterior views indicates the close level of agreement. (b) The HAP1-bound capsid has a significantly larger diameter, shown by the excess of magenta on the exterior and cyan on the interior. This difference is most evident at the 5-fold vertices, formed by A subunits. On the exterior view, the cyan patch ringing the 5-fold is the side of the CD spike. (c) The AT-130-bound capsid is also expanded compared to the apo-capsid to an even greater degree than the HAP1 structure. Most of the exterior cyan-colored surface is due to differences in the tertiary structure of the C subunit.

FIG 4.

Tertiary structure changes in the HBV capsid are associated with HAP18 binding. (a) The two-dimer asymmetric unit from the HAP18-bound capsid is overlaid on the asymmetric unit (A, blue; B, red; C, yellow; D, blue) of the apo structure (cyan), based on superposition of complete capsids. (b) The structures are quantitatively compared by plotting displacement of Cαs against residue number. As seen in the molecular comparisons in panels a, c, and d, the major changes occur near the spike tips, notably helix 4a, of the C and D subunits and the C termini of all the subunits. (c) Overlay of the AB dimer, extracted from the all-capsid overlay in panel a, shows little structural difference between these dimers. (d) Overlay of the CD dimer, extracted from the all-capsid overlay in panel a, shows major changes at the spikes, indicated by the brace bracket.

Despite the lack of quaternary structural changes, tertiary structure differences were observed on the exterior surface near the 5-fold vertices, the exterior surface near the quasi-6-fold vertices, the walls of 3-fold pores, and the helical domains that form the quasi-6-fold spikes (Fig. 3a and 4a). There were changes in the Cα positions near the C-terminal subdomains of all four subunits in the capsid asymmetric unit, with most changes mapping to positions after Gly123 in the middle of the C-terminal helix (Fig. 3a). Differences in the exterior and interior surfaces near the 5-fold vertices were a result of C-terminal tertiary changes in the A subunit, leading to a decrease in internal diameter, as measured 5-fold to 5-fold. Structural differences near the quasi-6-fold vertices were a result of corresponding C-terminal conformational changes in the quasiequivalent B and C subunits. Differences in the inner walls of the 3-fold pores resulted from C-terminal structural changes in the D subunit (Fig. 3a). Contacts between subunits are modulated by the C-terminal helix-loop-extended structure; we find that these motifs can move slightly to accommodate a CpAM without altering quaternary structure.

Superposition of the asymmetric units of HAP18-bound and apocapsid structures allowed visualization and analysis of tertiary structural changes (Fig. 4a). A plot of Cα displacement versus residue number, comparing the monomers in the HAP18-capsid to the corresponding subunits in the 1QGT structure, showed specific structural changes particularly in helix 4a (residues 80 to 92) and in the latter half of helix 5 and the following turn (residues 123 to 135) (Fig. 4b). The electron density and corresponding model at the spike tips (residues 75 to 79) were weaker than the rest of the structure, and the model does not provide a reliable indication of movement. The AB dimer from the HAP18-bound structure superposed with the AB dimer from the apo structure had a small mean displacement (RMSD, 0.50 Å), but with systematic shifts in the C-terminal residues after amino acid 120, in helix 5, and in helix 2, which crosses helix 5 (Fig. 4c and d and 1d). Superposition of the CD dimers had a larger mean displacement (RMSD, 1.41 Å), but most of the difference was localized to large conformational shifts at the spikes, specifically shifts in helices 3 (near residue 70) and the junction of helix 4a and 4b (residues 82 to 95), similar to those observed in the AT-130-bound capsid (Fig. 4d) (23). The systematic movement of the spikes, distal to the HAP site, may reflect an allosteric response. A comparison to a second apo structure (PDB ID 2G33) (21) showed similar minimal quaternary structure differences between it and the HAP18-bound capsid.

Electron density establishes the orientation of the HAP molecule in its hydrophobic pocket.

The CpAM-binding pocket is a hydrophobic site buried at the interface of neighboring dimers, previously identified to bind HAP1 and AT-130. As in previous studies, CpAM density was observed in B and C subunits but not A and D (21, 23). The site is comprised of a pocket in one subunit that is capped by residues from the neighboring subunit. For A and D subunits, the pocket is occluded by a neighboring A subunit or a neighboring B subunit, respectively. In contrast, the B pocket is capped by the C subunit within the asymmetric unit (Fig. 5a), and the C subunit is capped by the neighboring D subunit (Fig. 5b). Most of the residues forming the site are from the C-terminal subdomains of the two contributing subunits. The site is bounded by a “C-shaped” 4-helix motif, which includes helices 2, 4b, and 5 from the subunit with the pocket and helix 5 from the capping subunit (Fig. 5a and b). This site is not open to the exterior surface but is open to the capsid interior, particularly in the case of the C-associated pocket (Fig. 5b). The curvature of the pocket imposes a preferred orientation of the HAP18 molecule, and helix 5 of the capping subunit clamps down on the molecule. In the absence of a HAP molecule, there is a naturally occurring gap between the hydrophobic surfaces of the two adjacent subunits. HAP molecules fill this gap to effectively increase buried hydrophobic surface at the protein-protein interaction site, promoting aggressive, error-prone assembly (25).

FIG 5.

HAP density and the HAP-binding site viewed from the capsid interior. The HAP-binding site is delineated by a C-shaped four-helix motif (orange polygon), with one of the helices donated by the capping subunit. (a) The B-associated HAP site, viewed from the capsid interior, shows the R enantiomer of HAP18 binding at the interface of the B (red) and C (dark green) subunits. The HAP pocket is viewed from the interior of the capsid here and in panel b. (b) The C-associated HAP18, also an R enantiomer, is capped by the D subunit (yellow). (c) Viewed from the capsid exterior, electron density (black mesh) is shown for the B-associated HAP18 (magenta) and the C-associated HAP18 (green) at a contour of 1.5 σ. The R enantiomer fits density well. Shape complementarity of the CpAMs for the pockets establishes the orientation of the CpAMs in their binding sites. Each molecular model is shown with its chlorophenyl group (Fig. 1) pointing to the top of the figure.

Previously, we observed that HAP1 bound preferentially to the C subunit while AT-130 bound preferentially to the B subunit, with very weak density in the other available site, though both had weaker densities than observed for HAP18 (21, 23). The HAP18 density was equally strong in B and C subunits (Fig. 5c) and similar in strength to the nearby protein. The asymmetric shape of HAP18 and the relatively high-quality electron density constrained the placement and orientation of HAP18 in its binding sites despite the modest resolution of the structure and the relatively high B factor required for the molecule (Fig. 5c).

Though the HAP18 used in crystallization experiments was a racemic mixture, only the R enantiomer has assembly activity and suppresses HBV replication (30, 35). The R enantiomer was observed in the crystal structure (Fig. 5c). In the ∼5-Å HAP1-bound capsid structure, the enantiomer could not be identified (21); in fact, we were cautious of assigning the HAP to the weak density observed. To examine the basis of selectivity with HAP18, we performed parallel refinements with both the R and S enantiomers. Refinement with the S enantiomer in both binding sites did not yield improvements in the R factor (data not shown). It was observed that the halogenated phenyl moiety in the S enantiomer of HAP18 sterically clashed with the Phe23 side chain of both the B and C subunits. The S enantiomer in the B-subunit-binding site (which is sterically more restrained) also clashed with Trp102 and Tyr132 from the C chain. The R enantiomer improved refinement and showed complementarity with electron density in both sites (Fig. 5c).

HAP18 binds different structural states in quasiequivalent sites.

Quasiequivalence is reflected in the HAP18 sites. The dialkyl moiety of the HAP18 in the C-associated pocket was more exposed to the capsid interior. The buried surface area of the B-associated HAP18 in its binding site was 299.6 Ų. The surface area for the C-associated HAP18 was only slightly smaller, at 274.2 Ų, even though the dialkyl moiety was solvent exposed. In contrast, the substantial change in quaternary structure in the HAP1-bound capsid (21) increased the solvent exposure at the C-associated site (the interfacial area between HAP1 and the site was only 233 Ų).

HAP18 had strong shape complementarity to both B and C sites (Fig. 5). Most of the interactions in the binding sites were based on the association of hydrophobic surfaces between HAP18 and amino acid side chains. However, specific interactions varied between the two sites. In the B-associated site, the B subunit Trp102 indole nitrogen hydrogen bonded to the HAP18 pyrimidine ring. In the C-associated site, the guanidine group of Arg133 of the capping D subunit interacted with both the dihydropyrimidine and pyridine ring of the C-associated HAP18. The C-associated HAP18 also appeared to have a π-stacking interaction with Phe23 from the C subunit. Hydrophobic interactions in the B-associated site were observed to result from residues from both the B and C subunits, while the interactions observed in the C-associated site were primarily from the C subunit.

The two HAP18 molecules in the icosahedral asymmetric unit flank the C subunit C-terminal subdomain (residues following G123). Conformational shifts in the C subunit, compared to the apo structure, created room for the molecules in both sites where HAP18-binding traps a specific capsid conformation (Fig. 6). There were minimal conformational changes in the C terminus of the B subunit, suggestive of lock-and-key binding, to accompany the observed shift in the capping C subunit. For the C-associated HAP18, the C-terminal tail of the C subunit was moved away from the bound HAP18 compared to the apo structure, suggesting induced fit, creating room to accommodate the ligand in the pocket. A modest conformational change was also observed in the main chain of the capping D subunit; the loop at the end of helix 5 bent toward the binding site. In the apo structure, the loop faced away from the pocket. The observed shift allowed Arg133 from the D subunit loop to interact with HAP18. The movement of the C-terminal subdomain of the C subunit correlates with distal structural changes in the helices of the dimer interface, a correlation previously noted in a free-core dimer structure (18).

FIG 6.

The C-terminal helix-turn-extended structure of the C subunit mediates binding of both B- and C-associated HAP-18 molecules. An overlay of the HAP18 structure on the apo structure (gray) reveals structural changes near the HAP binding sites. To bind the B-associated HAP18 (magenta), minimal changes are seen in the residues of the B subunit (red), but there is a shift in the C subunit (dark green) C-terminal loop, after helix 5, to create room in the B-associated HAP pocket. To bind the C-associated HAP18, the loop between helices 4 and 5 of the neighboring D subunit (yellow) shifts by up to 3 Å compared to the apo structure to form contacts with the C-associated HAP18 while the C terminus of the C subunit shifts away to create a binding pocket for the CpAM.

Comparing CpAM-bound capsids.

The same site binds HAP1, AT-130, and HAP18, but each CpAM led to distinctly different crystal structures (Fig. 3). HAP1 and AT-130 both expanded capsids, while HAP18 led to a contraction. In overlays, the differences are most readily visible when looking at the capsid interiors, where the HAP18 structure is a clear outlier (Fig. 7). However, the mechanisms of structural change for each CpAM had striking differences. HAP1 minimally affected the tertiary structure of the subunits but introduced large global quaternary structure changes; the icosahedral asymmetric unit pivoted on a point near the center of the quasi-6-fold so that the A subunit was elevated by about 5 Å and the CD subunit was tilted to offset the dimer interface spike, with respect to the apo structure (Fig. 7, top panels, and 3b). We do note that the low resolution of the HAP1-capsid structure (∼5 Å) may have obscured tertiary structural changes (21). AT-130 introduced both tertiary and quaternary changes in the capsid, including a 4-Å upward rigid body shift in the A and B subunits, similar to, but to a lesser extent than, what was observed in HAP1, and substantial refolding of the CD interface at the spike tip (Fig. 7, bottom panels, and 3c). These observations initially suggested that CpAM bound in the interfacial hydrophobic pocket to induce quaternary changes in the capsid structure and in turn led to the alternate unit cell.

FIG 7.

HAP18, HAP1, and AT-130 each lead to distinctly different changes in tertiary and quaternary structures at the level of the dimer. Though the same HAP-binding sites are filled, the effect on local structure is distinct in each case. (Top panels) Overlay of AB and CD dimers from the HAP1-bound capsid structure (2G34, red) on corresponding subunits from the HAP18-bound capsid structure (yellow), based on the superposition of capsid asymmetric units. There are minimal differences in structure, as HAP1 induces primarily quaternary structural changes affecting dimer-dimer position rather than structure. (Bottom panels) Overlay of AB and CD dimers from the AT-130 phenylpropenamide-bound capsid structure (4G93, magenta) on corresponding subunits from the HAP18-bound capsid structure (yellow). There are large differences in tertiary structure, especially in the CD dimer. The upper half of helix 4 in both the C and D subunits is displaced by about a half turn.

While the HAP18-bound capsids crystallized with the same unit cell dimensions as other CpAM-bound capsids, the quaternary structure of the capsid was very similar to that of an apocapsid. HAP18 interacts with capsid protein almost entirely through atoms that it has in common with HAP1. While HAP1 led to a large shift in quaternary structure compared to the apocapsid, the almost identical binding interactions resulted in subunit movement limited to modest tertiary changes in the C-terminal subdomain of the A, C, and D subunits and the intradimer contacts of the CD spikes. The shift of the C-terminal subdomain is thus the basis of the smaller internal diameter in HAP18 capsids.

A commonality seen in all the CpAM-bound capsid structures is that minimal structure changes were observed in the AB dimer. HAP1 preferentially bound the C pocket and caused quaternary changes. AT-130 preferentially bound the B pocket and yet changed the tertiary structure of the CD dimer as well as the capsid quaternary structure. HAP18 bound both B and C subunits, causing no detectable quaternary changes, but affected the tertiary structure of the CD dimer.

DISCUSSION

We have compared the structures of HBV capsids with and without bound CpAMs. We had hypothesized that CpAMs would expand capsid structure by distorting interdimer interfaces, but this was incorrect. Unlike previous HBV-CpAM structures, the HAP18-capsid has essentially the same quaternary structure as the apocapsid. This is all the more surprising as the protein-CpAM interactions with HAP18 and HAP1, the latter of which led to gross changes in quaternary structure, are the same. The absence of quaternary change in the HAP18-capsid structure is not due to lack of CpAM binding in the HAP site, and based on the strength of the density the sites are filled with high occupancy.

Thus, the expansion of the capsid seen with HAP1 and AT-130 (Fig. 3b and c) was not specifically the effect of filling the CpAM-binding site. The range of HBV capsid protein structures, liganded and apo, icosahedral and nonicosahedral, led us to suggest that in solution Cp can be found in a diverse ensemble of quaternary structures that can be biased by CpAM binding, mutations, and crystal contacts. An analogous observation was made with experimental antiviral ligands that bind to picornavirus capsids, which make small tertiary structure changes to the capsid and stabilize capsids to inhibit nucleic acid release; these ligands appear to act by modifying picornavirus capsid dynamics, entropically stabilizing them (45–49).

CpAM binding, allostery, Cp assembly, and structure.

The CpAM-binding site is believed to act as a regulatory site for Cp dimer conformation that affects subsequent assembly by triangulating subdomains (18). The CpAMs studied thus far bind to the pockets in B and C subunits around the quasi-6-fold vertices. Unlike the HAP1 and AT-130 structures, where either B or C pockets were preferentially filled (21, 23), HAP18 has equal occupancy in both quasiequivalent sites, leading to 120 HAP18 molecules per capsid. Filling the HAP site allows CpAMs to move one end of helix 4b to alter the spikes, which thus allosterically interact with the other half of the dimer. This effect is seen most clearly in the C-associated pocket.

Though HAP18 is the largest CpAM yet examined, it showed the smallest structural shifts. Nonetheless, HAP18 accelerated assembly and stabilized protein-protein interactions, each more than 100-fold (29). Indeed, small differences in the CpAM led to large, unpredictable differences in products of assembly reactions (21, 23, 27, 29, 30, 50). We expected that the piperidinyl dialkyl moiety of HAP18 would result in additional interactions with core protein (compared to HAP1 [Fig. 1a and b]). We unexpectedly found that all HAP18 interactions with Cp were via the core HAP1 structure, with no contributions from the additional moiety. Though HAP18 led to aberrant structures, it had very modest effects on a preformed capsid structure, unlike HAP1; this is consistent with an effect on structural dynamics rather than a static effect.

When present in molar excess during assembly reactions, CpAMs can alter assembly geometry. HAP1 led to large baglike complexes with large patches of hexameric sheet (27). BAY41-4109 led to similar, but usually incomplete, particles (30, 32). AT-130, which has a completely different chemistry, yielded near-normal capsids (33). HAP18 led to sheets of uniform width and irregular length; we suggest that these are flattened tubes of regular diameter (29). Similarly, single mutations that alter the HAP site led to altered assembly and altered solution properties (18, 24, 50).

CpAMs preferentially bind at quasi-6-fold vertices; this forms a basis for hexamer-rich macrostructures generated by stoichiometric excess of HAPs (27, 29). Without the curvature provided by 5-fold vertices, the hexameric HAP-bound Cps form sheetlike structures (27, 29). The differences in macrostructure features likely are based on CpAM-modified quasiequivalence. The propensity for HAP18 to form tubes presumably correlates with its ability to evenly fill four sites in the quasi-6-fold vertex (two B sites and two C sites in the context of a capsid). We suggest that HAP18 introduces a 2-fold polarity in hexamers. During assembly in the presence of excess HAP18, this would be the basis for polymerizing a uniform hexagonal lattice.

In capsids, AB dimers show little change in tertiary structure with bound CpAM but can shift their position with reference to capsid quaternary structure (23). This provides a rationale for assembly of spherical capsids in the presence of substoichiometric concentrations of CpAM (30). In the absence of excess CpAM, 5-fold vertices form at the correct locations for icosahedral geometry. For AT-130, which led exclusively to “spherical” capsids (33, 51), it appeared that tertiary changes in the CD dimer compensated for the CpAM and led to AB quaternary shifts that preserved the icosahedral symmetry (23). We suggest that the choice between aberrant and regular capsid assembly is determined by the structural state of the quasi-6-fold.

Capsid dynamics.

The HBV capsid is plastic. Hydrogen-deuterium (H-D) exchange experiments showed that although Cp149 has a stable fold, the contained peptide amides are very labile (52). Furthermore, binding antibodies to capsids changed H-D exchange rates distal to the epitope site. When Cp149 capsids were treated with trypsin, the first cut was at Arg127 in helix 5 for both free dimer and capsid (53). For Arg127 to fit the protease active site, this helix must transiently partially unfold. Similarly, restraining subunit mobility through a highly conserved intradimer disulfide interfered with capsid assembly (54). Paradoxically, the component cysteines readily oxidize in capsids (54, 55). These results are consistent with a fluctuating ensemble of icosahedral states, which can be modified by bound CpAM.

The allosteric changes in HBV capsid and dimer and the associated ensemble of structural states play pivotal roles in the HBV life cycle. For example, capsids appear to be responsive to changes in nucleic acid conformation, as reverse transcription generates double-stranded DNA from an RNA template (56). Nucleic acid-induced changes to virus structure also modulate interactions with surface protein. Conversely, changes in interdimer contacts affected reverse transcription (24). Mature capsids are less stable than RNA-filled capsids, presumably due to both the tension from the packaged double-stranded DNA and the oxidation of the C61 disulfide (54, 57, 58). The C-terminal domains of full-length Cp become at least transiently exposed as a function of genome maturation and phosphorylation (59–62). Therefore, limiting Cp conformation by blocking allostery could be devastating to the virus life cycle.

Crystallization also suggests that CpAMs affect capsid dynamics. Crystallization of structurally heterogeneous proteins can be facilitated by solutes that favor a more homogeneous form (63). One commonality between the CpAM-bound capsid crystals is that they grow much faster (on the order of days) than apocapsid crystals (on the order of months), which could be explained by a decrease in capsid dynamics with CpAM binding allowing for more-rapid crystal elongation (11, 21, 23, 37). Capsid-CpAM crystals and apo crystals also have differences in the way capsids are packed, i.e., they are not isomorphic, despite the fact that the HAP18-capsid structure is very similar to the native structure. Strikingly, the crystals of CpAM-bound capsids have fewer packing interactions than capsids of the same protein without bound CpAM (21, 23). We now observe that the difference in crystallization is not a function of capsid quaternary structure as previously hypothesized. Therefore, we suggest that the difference is a function of capsid stability and protein dynamics.

Antiviral mechanism and final comments.

HAP18 has a marked antiviral effect but modest quaternary structural effects. In cell culture, HAP18 suppressed HBV replication, though not as effectively as HAP1 and other related HAPs (29). A comparison of various HAPs and phenylpropenamides showed that a reduction in viral load correlates with the CpAM's effects on the kinetics of assembly (29, 33). While HAP18 stabilized the interdimer interactions to a greater extent than HAP1, its effects on the rate of assembly were weaker. HAP18 had a 50% effective concentration (EC₅₀) much closer to that of AT-130, which has a comparable kinetic effect while having a negligible effect on intersubunit association energy (23, 29). The differential kinetic effect could arise from the interactions between CpAM-bound subunits and/or from the ability of the bound CpAM to alter subunit conformations. The differences between binding energy, kinetic effect, and antiviral activity indicate a complex structure-activity relationship. We suggest that HAP18's ability to redirect the structural states of the capsid protein (36) may have a role in its antiviral effect. These results have led to ongoing studies on HBV structure and dynamics.

Recently, Klumpp et al. published a description of the activities of a HAP molecule (NVR-010-001-E2, here HAP-NVR) including a high-resolution crystal structure (1.95 Å). HAP-NVR is a close relative of HAP12 (29). In their study, HAP-NVR had been soaked into crystals of Cp149-Y132A, an assembly-deficient core protein mutant (64). Their structural results are largely consistent with those reported here. In Cp149-Y132A crystals, the protein forms an approximately trigonal sheet with a repeat of three, nearly 2-fold symmetric dimers forming a closed triangle. As with the HAP18-capsid structure, they observed that the HAP core of HAP-NVR interacted with residues of the HAP site but there was little interaction with the HAP-NVR morpholino group, equivalent in position to the piperidinyl moiety in HAP18. The biophysical effects of HAP18 and HAP-NVR also had similarities. HAP18 was found to stabilize dimer-dimer interactions by −1.8 kcal/mol per contact, HAP12 increased stability by −1.9 kcal/mol (29), and HAP-NVR was found to raise the unfolding temperature of Cp149 capsids by about 10°C to 89°C. A critical difference between the two structures arises from the quasiequivalence of a T=4 capsid: there are substantial differences in HAP-protein interactions at quasiequivalent B and C sites, whereas HAP-NVR in Cp149-Y132A appears to be limited to a single environment that is distinct from those in a capsid. The molecular detail provided by the high-resolution structure, therefore, does not explain or predict the multiple binding modes for the same HAP molecule seen in a biological context.

In this study, we have established the structural basis for HAP18 binding and activity and that not all CpAMs share a structural effect. The exposure of the acetylene tails of HAP18 toward the capsid interior would make it an attractive motif for additional modifications. Our findings show that very modest structural changes, possibly transient changes, can still disrupt capsid assembly to significant antiviral effect. We therefore suggest a dynamic as well as a structural basis for HAP antiviral activity. A dynamic capsid is not unique to HBV, as demonstrated in studies with picornaviruses (49, 65) and HIV (66). We therefore propose that small molecules that limit dynamic ensembles can serve as an effective approach to antiviral treatment.

Funding Statement

Some elements of this work were supported by Assembly Biosciences, including a supported research project to Adam Zlotnick. Beam time at the Advanced Photon Source, Argonne National Laboratory, was supported by the U.S. Department of Energy, Basic Energy Sciences, Office of Science, under contract W31-109-Eng-38. This work was supported by grant R01 AI067417 from the NIH to A.Z.