ABSTRACT
The hepatitis B virus (HBV) capsid or core protein (Cp) can self-assemble to form an icosahedral capsid. It is now being pursued as a target for small-molecule antivirals that enhance the rate and extent of its assembly to yield empty and/or aberrant capsids. These small molecules are thus called core protein allosteric modulators (CpAMs). We sought to understand the physical basis of CpAM-resistant mutants and how CpAMs might overcome them. We examined the effects of two closely related CpAMs, HAP12 and HAP13, which differ by a single atom but have drastically different antiviral activities, on the assembly of wild-type Cp and three T109 mutants (T109M, T109I, and T109S) that display a range of resistances. The T109 side chain forms part of the mouth of the CpAM binding pocket. A T109 mutant that has substantial resistance even to a highly active CpAM strongly promotes normal assembly. Conversely, a mutant that weakens assembly is more susceptible to CpAMs. In crystal and cryo-electron microscopy (cryo-EM) structures of T=4 capsids with bound CpAMs, the CpAMs preferentially fit into two of four quasi-equivalent sites. In these static representations of capsid structures, T109 does not interact with the neighboring subunit. However, all-atom molecular dynamics simulations of an intact capsid show that T109 of one of the four classes of CpAM site has a hydrophobic contact with the neighboring subunit at least 40% of the time, providing a physical explanation for the mutation's ability to affect capsid stability, assembly, and sensitivity to CpAMs.
IMPORTANCE The HBV core protein and its assembly into capsids have become important targets for development of core protein allosteric modulators (CpAMs) as antivirals. Naturally occurring T109 mutants have been shown to be resistant to some of these CpAMs. We found that mutation of T109 led to changes in capsid stability and recapitulated resistance to a weak CpAM, but much less so than to a strong CpAM. Examination of HBV capsid structures, determined by cryo-EM and crystallography, could not explain how T109 mutations change capsid stability and resistance. However, by mining data from a microsecond-long all-atom molecular dynamics simulation, we found that the capsid was extraordinarily flexible and that T109 can impede entry to the CpAM binding site. In short, HBV capsids are incredibly dynamic and molecular mobility must be considered in discussions of antiviral mechanisms.
INTRODUCTION
Chronic hepatitis B virus (HBV) infection is a worldwide health issue. More than 240 million patients suffer from chronic HBV infection. Chronic HBV can lead to liver failure, cirrhosis, or hepatocellular carcinoma, causing more than 700,000 deaths annually. Current treatments against chronic HBV suppress the virus but usually do not lead to a sustained virological response (1). New approaches, including core-protein-directed antivirals, are now being investigated for their potential to contribute to a curative treatment (2).
HBV is an enveloped DNA virus with an icosahedral core. The protein shell of the core, the capsid, is a T=4 icosahedral complex with a 36-nm diameter that is assembled from 120 copies of the homodimeric core protein (Cp). Cp is a 183-residue protein. The first 149 residues comprise the alpha-helical assembly domain, which is dimerized by a four-helix bundle, with two helices contributed by each half-dimer (Fig. 1) (3). The remaining 34 residues form a flexible, arginine-rich C-terminal domain that binds nucleic acid and signals nuclear localization and export (4–7). Reverse transcription takes place inside the capsid, in a reaction that depends on the integrity of both the assembly and C-terminal domains (8, 9). Cp is also found in the nucleus, in association with covalently closed circular DNA (cccDNA) (10, 11). The capsid protects the packaged viral genome, provides a metabolic compartment for reverse transcription, has multiple functions in the nucleus, delivers the viral genome to the nucleus for infection or to the endoplasmic reticulum for secretion, and interacts with the viral envelope proteins.
Core protein and CpAMs. (a) Cp homodimer showing assembly domains. The dimer interface is a four-helix bundle that forms spikes on the capsid surface. In this view, the capsid interior is at the bottom of the dimer and the exterior is at the top. The interdimer contact, as well as the HAP binding pocket, is formed by the helix-loop-extended structure on the far right and left of the structure. The HAP is shown as pink spheres, and the T109 residue is shown in cyan. (b) View of the interaction of two dimers from the capsid exterior. The HAP pocket is at the interdimer interface. Subunits are labeled A to D, with A subunits at the 5-fold symmetry vertex. HAPs fill pockets in only the B and C subunits (labeled “1” and “2”). (c) HAP12 and HAP13 differ by a single atom, highlighted in bold.
The multifunctionality of Cp makes it an attractive antiviral target. Core protein allosteric modulators (CpAMs) include the heteroaryldihydropyrimidines (HAPs), phenylpropenamides (PPAs), and sulfamoylbenzamides (SBAs). These compounds have antiviral activity that roughly corresponds to their ability to stimulate capsid assembly (12–14). HAPs, PPAs, and SBAs bind to a pocket formed at the interface between two Cp dimers. These CpAMs enhance assembly kinetics and strengthen the association between dimers. HAPs in particular lead to assembly of aberrant structures. However, all of these CpAMs are believed to inappropriately nucleate Cp assembly, leading to empty and defective particles in lieu of nascent virions.
An understanding of drug resistance to CpAMs remains elusive. Unlike that of many viruses, HBV replication has high fidelity, and mutations in Cp are relatively uncommon (15, 16). In part, this may be because most nucleotides are in more than one gene. The lack of a culture system that supports multiple rounds of high-yield replication is a major barrier to identifying resistance to novel antivirals. Consequently, surveying published sequences is perhaps the most effective approach for identification of resistant mutants. A recent study found that Cp residue T109, which is located near the CpAM pocket, is 97% conserved based on a database of 2,800 published HBV sequences (16), making it one of the most variable residues located near the HAP pocket. T109 may be replaced by several alternative amino acids that differentially modify HBV sensitivity to CpAMs in vitro and in clinical trials (16, 17). However, examination of crystal and cryo-electron microscopy (cryo-EM) structures of HBV capsids does not provide a structural explanation for T109-based resistance. In the present study, we investigated the mechanism by which T109X mutants alter assembly and antiviral activity by testing Cp assembly kinetics, assembly thermodynamics, and sensitivity to CpAMs. We employed all-atom molecular dynamics (MD) simulations of an intact HBV capsid to elucidate T109's involvement in dimer-dimer interactions, and we suggest a structural basis for CpAM resistance upon T109 mutation.
RESULTS
Cp residue T109 is 97% conserved, which makes it one of the most frequently mutated residues in HBV Cp (15). It is one of the few mutated residues found near the CpAM binding pocket. More specifically, T109 is near the mouth of the CpAM pocket facing the capsid interior. Based on crystal and cryo-EM structures, T109 is predicted to have little direct contact with a bound HAP and no interaction with an adjacent dimer (Fig. 1). Nonetheless, T109I, T109M, and T109S mutants were shown to have various levels of HAP resistance (16). To test whether T109X mutants affect the physical chemistry of assembly and whether this may provide a basis for resistance, Cp149-T109X mutants were expressed in Escherichia coli and purified.
An easily accessible experimental quantitation of assembly is the pseudo-critical concentration (PCC) of capsid formation (18). The PCC arises because capsid formation involves an equilibrium between many subunits (120 dimers for HBV) and a complete capsid. In practice, the PCC is the maximum concentration of free dimers at equilibrium; above the PCC almost all dimers are in the form of a capsid. The PCC can readily be used to calculate the pairwise association energy between two dimers (19).
HBV assembly is proportional to ionic strength (20). To work in a convenient range of protein concentrations, we examined capsid assembly in 300 mM NaCl (Fig. 2; Table 1). The measured PCC for Cp149 was 3.7 μM, consistent with the results of previous experiments (20). Mutating T109 to more hydrophobic residues (T109I and T109M mutations) decreased the PCC. Substituting a more hydrophilic amino acid (T109S mutation) led to a higher PCC. To gain a better understanding of the effect of a mutation, it is often useful to consider intermolecular interactions in terms of their association energy, ΔGcontact, which is directly proportional to the log of the PCC (19). It is immediately clear that the change of association energy is small, at about 0.4 kcal/mol between the T109I and T109S mutants; by comparison, a previous study found that changing the hydrophobic surface within the CpAM pocket by mutating a single residue led to a change of up to 2.1 kcal/mol (9). As there are 240 intersubunit contacts holding a capsid together, even these modest changes in energy have a significant effect on capsid stability. Though the sample size is small, it is striking that the per-contact association energy follows the same rank order as sensitivity to the HAP molecule GLS4, i.e., ΔGcontact,Ile > ΔGcontact,Met > ΔGcontact,Thr > ΔGcontact,Ser (Table 1; Fig. 3) (16).
Pseudo-critical concentration (PCC) of capsid assembly. Capsid assembly is steeply concentration dependent, resulting in the PCC: all Cp above the PCC is found as capsid (29). Data are from reaction mixtures in 50 mM HEPES-300 mM NaCl, pH 7.5, at 23°C, for which both capsid (squares) and dimers (circles) were quantified by HPLC.
Assembly correlates with HAP resistance for HBV Cp149 and Cp149-T109X mutants
There is a linear correlation between dimer-dimer association energy and antiviral resistance. Data are from Table 1.
In previous work, it was observed that the antiviral effect of a CpAM could be related to assembly kinetics (21). To compare Cp149-T109X mutants, we used 90° light scattering, measured in a fluorometer, which is sensitive to weight-average molecular weight (i.e., it is particularly sensitive to large complexes) (22). Assembly kinetics are modulated by nucleation and elongation phases. Because assembly of a 120-dimer capsid necessarily has a plethora of intermediates, the status of the intermediates can dramatically affect kinetics. Decreasing the reversibility of the component reactions notably decreases the lag phase of assembly (23, 24). The Cp149-T109I mutant had the fastest assembly, while Cp149-T109S, even though it assembled after 24 h, did not show a measurable signal over a 5-min span (Fig. 4). For Cp149, the lag phase is approximately 20 s. For both Cp149-T109I and Cp149-T109M, the lag phase is much shorter than the 10-s dead time of the measurement, while the Cp149-T109S mutant appears to have a lag phase of >250 s. The substantial differences in kinetics indicate that the T109X mutants affect the rate constants of nucleation and elongation, not just the thermodynamics of assembly.
Assembly kinetics measured by 90° light scattering. Assembly was monitored by 90° light scattering at 320 nm. Assembly was initiated at 30 s in a 300-s measurement, when an equal volume of 50 mM HEPES, 600 mM NaCl, pH 7.4, was added to 10 μM Cp149-T109X in 50 mM HEPES, pH 7.4. These data are also shown compared directly to drugged assemblies in Fig. 5. All samples were incubated at 23°C.
We examined the responses of T109X mutants to an aggressive CpAM, HAP12, and a weak CpAM, HAP13. These two molecules were chosen because they are nearly identical (Fig. 1). HAP12 is also structurally very similar to GLS4, in that both have a morpholine group at position 6 of the dihydropyrimidine that strongly activates antiviral activity; HAP13 has the deactivating, positively charged piperazine group at this position (21). HAP12 is estimated to strengthen the pairwise association energy by −1.9 kcal/mol, compared to −0.6 kcal/mol for HAP13, based on extrapolating the change in binding energy to saturation (21); similarly, HAP12 is estimated to increase the association kinetics 20-fold (1.28 log) more than HAP13 does. The antiviral effects of the two compounds in HBV-expressing HepG2.2.1.5 cells follow a similar trend: the 50% effective concentration (EC50) for HAP12 is 12 nM, and that for HAP13 is 6.1 μM (21). In this work, we observed that Cp149 assembly was accelerated by HAP13 and was even faster with HAP12 (Fig. 5). HAP12 also accelerated the assembly of all three T109 mutants and increased the maximal light scattering, even for Cp149-T109I, which was almost entirely assembled under these conditions. Because light scattering is sensitive to weight-average molecular weight, the high light scattering signal seen with HAP12 implied the presence of complexes that were larger than a capsid. Conversely, though Cp149 and Cp149-T109S showed assembly acceleration due to HAP13, Cp149-T109M and Cp149-T109I showed no discernible effect.
Assembly kinetics of T109X mutants are differentially sensitive to HAP12 and HAP13. Assembly of protein alone (solid black lines) was compared to the effects of HAP12 (dashed red lines) and HAP13 (thick solid blue lines). Assembly was monitored by 90° light scattering at 320 nm. The data collected in the absence of drug are also shown in Fig. 4. Assembly was initiated at 30 s in a 300-s measurement, when an equal volume of 50 mM HEPES, 600 mM NaCl, ±20 μM HAP, pH 7.5, was added to 10 μM Cp149-T109X in 50 mM HEPES, pH 7.5. All samples were incubated at 23°C.
We used electron microscopy (EM) to examine the products of Cp149-T109X assembly reactions performed with a range of concentrations. Note that HAP12 and HAP13 are racemic mixtures and that only half of the molecules will affect HBV (25, 26). To minimize transient kinetic traps in control samples, we chose to work at a lower ionic strength, using 150 mM NaCl. At final concentrations of 10 μM protein and 150 mM NaCl, Cp149, Cp149-T109I, and Cp149-T109M formed normal capsids (Fig. 6). Under these conditions, Cp149-T109S remained unassembled because of its high PCC. At 10 μM HAP12 (a stoichiometry of only one active HAP molecule for every two dimers), all four T109 variants showed a large number of aberrant structures and some seemingly normal capsids. The largest proportion of aberrant capsids were seen with the T109S mutant, which is presumably the most dependent on the HAP to drive assembly. Conversely, Cp149-T109I had a relatively large fraction of capsids. At 40 μM and 80 μM HAP12, only large aberrant structures were observed for all four T109 variants. These larger-than-capsid complexes explain the light scattering increase shown in Fig. 5.
TEM images of Cp149-T109X protein assembled in the presence of 0 to 80 μM HAP12. Assembly of 10 μM dimers was induced by 150 mM NaCl (final concentrations), under which conditions the T109S mutant was below the PCC. At 10 μM HAP12, where half of the molecules are the inactive isomer, all four T109 variants showed a mixture of normal and aberrant capsids. At higher HAP12 concentrations, all observed complexes were inconsistent with an icosahedral capsid. Samples were subjected to negative staining with uranyl acetate.
Biophysical experiments suggested that HAP13 should show a pattern of morphological defects that is different from that with HAP12. For 10 μM HAP13, Cp149 assemblies were mainly capsids, with a few aberrant structures, while Cp149-T109S yielded mainly aberrant particles (Fig. 7). Cp149-T109I and Cp149-T109M mutants yielded only normal capsids. At superstoichiometric concentrations of HAP13, Cp149 and especially Cp149-T109S often formed tubular and conical structures. It is interesting that with Cp149 and Cp149-T109S, larger assembly products were found with HAP13 than with HAP12, suggesting that HAP13 led to fewer nuclei. Cp149-T109M showed progressively more aberrant complexes at 40 μM and 80 μM HAP13, indicating that the CpAM concentration could overcome resistance. Unlike the T109M mutant, Cp149-T109I appeared to be resistant to HAP13-induced assembly defects at all concentrations.
TEM images of Cp149-T109X protein assembled in the presence of 0 to 80 μM HAP13. The same control sample images are also shown in Fig. 6. Assembly of 10 μM dimers was induced by 150 mM NaCl (final concentrations), under which conditions the T109S mutant was below the PCC. At 10 μM HAP13, normal capsids were seen with the T109I and T109M mutants, while the wild-type protein and the T109S mutant were grossly deformed. At higher HAP13 concentrations, the T109I mutant protein retained a normal morphology, whereas the other proteins assembled into increasingly bizarre shapes. Samples were subjected to negative staining with uranyl acetate.
A change in the amount of assembly does not require a morphological change. To quantify HAP12- and HAP13-induced assembly of Cp149-T109X mutants, different concentrations of HAP12 or HAP13 were titrated against 10 μM protein. Since large aberrant structures will clog a size exclusion chromatography (SEC) column, assembly products were removed from equilibrated assembly reaction mixtures by centrifugation at 150,000 × g for 30 min. Supernatants were examined by high-pressure liquid chromatography (HPLC), and free dimers were quantified by peak integration. HAP12 decreased the remaining dimer concentration in a dose-dependent manner for all four T109 variants (Fig. 8). At 40 μM HAP12 (one active HAP per dimer), we observed essentially no free dimers, i.e., approximately 100% assembly. However, surprisingly, at the substoichiometric HAP12 concentration of 10 μM (one active HAP for every two dimers), HAP12 had a greater effect on the amount of remaining dimers for the more resistant Cp149-T109I mutant (dimer concentration decreased ∼80%) than for the less resistant wild type and Cp149-T109S (dimer concentration decreased 44% and 56%, respectively).
Free dimers remaining in solution after assembly of Cp149-T109X with various CpAM concentrations. (Left) All proteins showed steep dose-dependent assembly induced by HAP12. (right) With HAP13, the T109I mutant was highly resistant to changes in assembly, even with 80 μM CpAM. The amounts of dimers were evaluated by HPLC after assembly reaction mixtures were incubated for 24 h and large polymers, including capsids, were removed by centrifugation.
With HAP13, the responses were more diverse (Fig. 8). Cp149-T109I showed little change in free dimer concentration at all HAP13 concentrations. The Cp149-T109M response to HAP13 was dose dependent, but some free dimers persisted even at high HAP13 concentrations. The wild type and the Cp149-T109S mutant were about as responsive to HAP13 as they were to HAP12.
The ability of T109 mutants to alter capsid stability is difficult to explain based on crystal and cryo-EM structures; the two highest-resolution T=4 capsid structures (Protein Data Bank [PDB] entries 1QGT and 3J2V) show T109 bathed in solvent and making no contact with V120 in the neighboring subunit (3, 27). A recent study employing all-atom MD simulations of an intact HBV capsid in explicit solvent (150 mM NaCl) showed that the particle is wildly dynamic (28). These simulations were performed without symmetry constraints, and following relaxation from the crystallographic starting structure, the capsid explored a large ensemble of asymmetric conformations. Asymmetry extended to the local level, such that each of the 120 CpAM-accessible pockets explored unique ensembles of configurations.
We mined the data set generated by capsid simulations to investigate the behavior of T109. A total of 50,000 capsid conformers were recorded over 1 μs of simulations, each with 60 B and 60 C CpAM pockets. This provided 3 million samples for B and for C. The results indicate that T109 can form hydrophobic contacts with V120 and V124 of the neighboring subunit. A hydrophobic contact is defined here as a distance of ≤4.0 Å, center to center, between methyl carbons. This proximity is sufficient to exclude water and allow attractive van der Waals interaction. T109, which is located at the mouth of the CpAM pocket, contacts V120 of the neighboring subunit, with 40% occupancy in C sites (Fig. 9 and 10; see Movie S1 in the supplemental material). Such an interaction is expected to slightly strengthen the capsid association energy; also, by decreasing dissociation during assembly reactions, this interaction will speed up the apparent kinetics (29). Hydrophobic contacts with T109 do not appear to be common in the B site. In both the B and C sites, the methyl-methyl distances between T109 and V120/V124 are frequently ≤5.5 Å (Fig. 9), a distance more than close enough to partially occlude the opening of the pocket. The frequency at which the methyl-methyl distances are ≤5.5 Å may explain the slow kinetics of CpAM binding (30).
MD simulations show that T109 can make hydrophobic contacts with the neighboring subunit. Simulations of an intact capsid, performed without symmetry constraints, produced 50,000 capsid conformers, providing 3 million samples for each of the B and C CpAM pockets. Using a cutoff of 4.0 Å between methyl carbons for hydrophobic interaction, contact between T109 and V120 in the neighboring subunit was present in the C pocket, with 40% occupancy. Thus, contact with the neighboring subunit was present with >20% occupancy over all CpAM binding sites. Extending the cutoff to 5.5 Å suggests that the additional side chain length (1.5 Å) conferred by the Cp149-T109I mutation would allow contact between T109 and V120, with nearly 75% occupancy in C sites and 10% occupancy in B sites, and between T109 and V124, with 27% occupancy in C and B sites.
Hydrophobic contacts between T109 and valines of the neighboring subunit indicate a structural basis for our experimental findings. A representative conformation from simulations based on PDB ID 2G33 shows a 4-Å hydrophobic contact between T109 and V120 in the C pocket blocking entry to the CpAM site. For reference, the T109-V120 distance in PDB structure 2G34, which has a HAP in its C pocket, is 7 Å; based on that structure, a HAP was modeled into the site shown in this figure. Hydrophobic contact between T109 and V120 has not been observed in experimental capsid structures containing HAPs (31, 32, 38), yet it clearly occludes the CpAM binding pocket in MD simulations. With the Cp149-T109I mutant, the longer side chain may more easily bridge the gap to make contact with V120, strengthening the subunit-subunit interaction and blocking CpAM binding. A longer side chain may also allow contact between T109 and V124, directly interfering with HAP binding.
Increasing the interaction cutoff to 5.5 Å suggests that the additional side chain length conferred by the Cp149-T109I mutation (1.5 Å for an additional carbon-carbon bond) results in a hydrophobic contact with V120, with nearly 75% occupancy in C sites and 10% occupancy in B sites, and a hydrophobic contact with V124, with 27% occupancy in C and B sites (Fig. 9). Contact with V124 in particular may serve to block HAP binding, as this residue has been shown to interact directly with HAPs in crystal structures (Fig. 10) (31, 32). These results indicate a structural basis for both the enhanced assembly and increased resistance to HAP observed experimentally for the Cp149-T109I mutant (16).
DISCUSSION
In this paper, we studied the basis of HBV resistance to CpAMs by using a series of related mutants and two closely related CpAMs with markedly different antiviral efficacies. This comparison allows a more careful examination of the physical chemistry of resistance without resorting to designed mutants that fill the CpAM pocket (9, 33). The choice of mutants in this study was based on their location near the HAP site, their relative frequency (though T109 is 97% conserved), and the recent observation that T109 mutants do confer some resistance to HAPs (15, 16). Spontaneous mutations seem to be relatively rare in HBV, particularly in the core protein (15). Mutants in the pool of virus that has not been subject to selective pressure are therefore a likely source of resistance. The two structurally similar HAPs allow us to essentially ignore steric differences between the two small molecules.
The fundamental observation is that a mutant that promotes normal assembly (Fig. 2 and 4) has substantial resistance even to a highly active CpAM (Fig. 3 and 7). Conversely, a mutant that has weaker assembly is more susceptible to CpAMs. A seeming paradox arises with the observation that substoichiometric amounts of the highly active CpAM HAP12 lead to disproportionately more assembly in the strongly resistant T109I mutant than in the weakly resistant T109S mutant. This counterintuitive peculiarity is actually predictable from assembly theory (29, 34, 35). A regulatory step for assembly is nucleation, and the amount of assembly is largely a function of protein-protein interaction. With limiting amounts of a strong CpAM and a protein with relatively strong subunit-subunit association energy, enhanced nucleation will lead to more starting points for assembly; the additional association energy of the bound CpAM will then be leveraged to generate more capsid. In vivo, initiating assembly in the absence of the biological nucleating complex of viral RNA and viral polymerase will promote formation of empty particles, an antiviral mechanism. This scenario is consistent with the observation that antiviral activity of a HAP series correlates with its effect on assembly kinetics (21). Conversely, a protein with weaker association energy would require saturating concentrations of CpAM to stabilize capsid formation, and consequently, fewer particles would form. In an infected cell, the concentrations of Cp are much lower than those in these in vitro experiments, and the nucleic acid-binding C-terminal domain plays a critical role in supporting assembly (36, 37). Thus, a highly active CpAM may be able to achieve an antiviral effect in vivo at concentrations much lower than those shown in vitro, even with a resistant mutant. Furthermore, a mutation that strengthens assembly while blocking binding of a CpAM may give resistance to CpAM but may also have its own antiviral effect that results in formation of empty capsids, as with the V124W mutation (9, 33).
The structural basis for T109X resistance to HAPs is problematic if one looks only at static structures. T109 is at the mouth of the CpAM pocket, not in it. The side chain is largely exposed to the lumen of the capsid. It does not directly contact HAPs either in capsid structures (32, 38) or in the trigonal sheets of Cp seen in crystals of an assembly-deficient mutant (12, 16, 39). However, recent all-atom MD studies of an intact HBV capsid (28) showed that, in the CpAM pocket of the C subunit, T109 can form a hydrophobic contact with V120 of the neighboring subunit (Fig. 10). The equivalent interaction in the B pocket is relatively rare. In both pockets, the dynamic proximity of T109 to V120 and V124 may affect the kinetics of drug binding. It is interesting that the CpAM density was greatest in the C subunit of the HAP1-capsid crystal structure (31) but in the B subunit of the phenyl propenamide-capsid crystal structure (13). However, there was little obvious difference between B and C sites in either of these static structures.
The static representations given by crystal or cryo-EM structures suggest all-or-nothing interactions between protein residues and thus provide limited information regarding the role of a given residue in capsid function. In contrast, the ensemble of conformers produced by MD simulations offers a more comprehensive characterization given the remarkable flexibility of the HBV capsid. For example, a previous study captured subtle morphological changes induced by the chemical interactions of bound HAP molecules (40), and this HAP-driven faceting of capsids has now been observed with both crystallography and cryo-EM (30, 31). Further, local motions sampled during microsecond-long capsid simulations showed high correlations with both crystallographic B factors and the calculated EM resolution for capsid structures (28). Although the simulations analyzed in the present work describe the wild-type capsid, not T109X mutants, the transient interactions they reveal are highly pertinent to assessing the functional role of T109 and the effects of mutations at this site. Altogether, our simulation results support a reasonable and intuitive structural explanation for the experimentally observed behavior of T109 mutants.
The overarching question of this study is how sensitive to resistance the CpAMs will be. The HBV genome is generally slow to mutate, though its response to the reverse transcriptase inhibitor lamivudine is a striking exception (41). The HBV core protein is highly conserved, with very few mutations near the CpAM pocket (15, 16). Critically, there are statistical barriers to CpAM resistance mutations that arise because the core protein forms a large oligomer. CpAMs drive capsid assembly, stimulating nucleation and stabilizing dimer-dimer interactions (38). CpAMs are effective because the core protein is already primed to self-assemble—about 90% of cores are empty (42)—and CpAMs stimulate this reaction. Mutants that overstimulate assembly can lead to empty capsids (9), mimicking the behavior of some CpAMs (43). However, resistant mutants may avoid some of the effects of CpAMs by assembling faster and stronger without the CpAM (Table 1; Fig. 3). Thus, resistance is a double-edged sword. Furthermore, if a resistant mutant arises in a cell already infected with a less resistant Cp, it can result in the formation of chimeric capsids that include the sensitive Cp. We have observed coassembly of HBV Cp for mutants with enhanced and attenuated assembly properties (9, 33, 39, 44). Coassembly means that the infection will remain CpAM sensitive, as the sensitive Cp can nucleate a misassembly. Indeed, in other systems, the sensitivity of chimeric capsids to antivirals, such as CpAMs, has led capsid proteins to be described as a dominant drug target (45).
MATERIALS AND METHODS
Cloning of Cp149 T109X mutants and core protein purification.T109 of the HBV genotype D strain adyw core protein was mutated to isoleucine, methionine, and serine by PCR mutagenesis (Table 2). pET-11c plasmids carrying the Cp149-T109X genes were transformed into E. coli BL21(DE3) and grown in Terrific broth (BD Biosciences) with 50 mg/liter carbenicillin at 37°C for 16 h. Core proteins were purified as previously described (20). Cp was quantified by measuring the absorbance at 280 nm, using an extinction coefficient of 60,900 M−1 cm−1.
Primers used for T109X mutations in Cp149
Assembly kinetics.Light scattering, which is sensitive to the weight-average molecular weight of solute (22), was monitored at 90° by use of a 320-nm excitation and emission wavelength on a Photon Technology International fluorimeter at 23°C for 5 min. Cp149 or Cp149-T109X mutant dimers (10 μM) were mixed with an equal volume of 50 mM HEPES-600 mM NaCl to a final concentration of 50 mM HEPES-300 mM NaCl to initiate assembly. For HAP-treated samples, 20 μM HAP, dissolved in dimethyl sulfoxide (DMSO), was added to 50 mM HEPES-600 mM NaCl prior to mixing with Cp. The final conditions were 50 mM HEPES, 300 mM NaCl, 5 μM Cp149, 10 μM HAP12 or HAP13, and 1% DMSO.
SEC.For pseudo-critical concentration determination experiments, a Superose 6 10/300GL column (GE Healthcare) was used to separate capsid and dimers. Cp dimers (0 to 20 μM) were induced to assemble by use of 300 mM NaCl and incubated for 24 h before injection. For the free dimer quantitation experiment, 10 μM Cp149 or Cp149-T109X dimers were preincubated with 0 to 80 μM HAP12 or HAP13 for at least 20 min, induced to assemble with an equal volume of 300 mM salt, to a final concentration of 150 mM NaCl, and incubated for 24 h at 23°C. Samples were centrifuged at 150,000 × g for 30 min to remove large aggregates, including capsids, before quantification of free dimers by HPLC over Agilent Bio-SEC5 100-Å- and 1,000-Å-pore-size columns in tandem.
TEM.Aliquots of assembly products isolated from either SEC or light scattering experiments were examined by negative-stain transmission electron microscopy (TEM). Five microliters of 10 μM Cp with HAP was applied to a carbon-coated grid and incubated for 30 s. The grid was blotted and stained with 2% uranyl acetate. Samples were imaged with a JEOL-1010 TEM equipped with a 1k*1k Gatan charge-coupled device (CCD) camera.
MD simulations.A complete model of the HBV capsid assembly domain (Cp149) was constructed and employed as the basis for all-atom MD simulations (28). The simulations were performed without symmetry constraints and, after a 100-ns period of equilibration, were continued for 1 μs (28). This gave rise to 60 μs of cumulative sampling for B and C CpAM binding pockets. A total of 50,000 capsid conformers recorded during simulations (frames were saved every 20 ps) provided ensembles of 3 million samples each for B and C binding pockets. VMD (46) was used to characterize protein-protein interactions involving residue T109. Hydrophobic contacts between T109 and valines of neighboring subunits were defined by a center-to-center distance between methyl carbons of ≤4.0 Å. Results are reported as occupancies (percentages of samples in which the hydrophobic contact was present). Trajectory analysis was carried out on the Blue Waters supercomputer at the University of Illinois at Urbana-Champaign.
ACKNOWLEDGMENTS
This work was supported by NIH grant R01-AI118933 and funding from the Indiana Clinical and Translational Sciences Institute (CTSI) to A.Z. and by a fellowship from the University of Delaware to J.A.H. This work is also part of the Blue Waters sustained-petascale computing project, which is supported by the National Science Foundation (awards OCI-0725070 and ACI-1238993) and the state of Illinois. Blue Waters is a joint effort of the University of Illinois at Urbana-Champaign and its National Center for Supercomputing Applications. We acknowledge use of the Blue Waters supercomputer through a Broadening Participation allocation, entitled “Revealing mechanistic insights into the drug-induced disruption of the hepatitis B virus capsid through the computational microscope.”
We acknowledge the Indiana University Electron Microscopy Center for use of its facilities.
A.Z. has an interest in a biotech start-up company.
FOOTNOTES
- Received 20 June 2018.
- Accepted 30 July 2018.
- Accepted manuscript posted online 8 August 2018.
Supplemental material for this article may be found at https://doi.org/10.1128/JVI.01082-18.
REFERENCES
- Copyright © 2018 American Society for Microbiology.