The Interface between Hepatitis B Virus Capsid Proteins Affects Self-Assembly, Pregenomic RNA Packaging, and Reverse Transcription

ABSTRACT

Hepatitis B virus (HBV) capsid proteins (Cps) assemble around the pregenomic RNA (pgRNA) and viral reverse transcriptase (P). pgRNA is then reverse transcribed to double-stranded DNA (dsDNA) within the capsid. The Cp assembly domain, which forms the shell of the capsid, regulates assembly kinetics and capsid stability. The Cp, via its nucleic acid-binding C-terminal domain, also affects nucleic acid organization. We hypothesize that the structure of the capsid may also have a direct effect on nucleic acid processing. Using structure-guided design, we made a series of mutations at the interface between Cp subunits that change capsid assembly kinetics and thermodynamics in a predictable manner. Assembly in cell culture mirrored in vitro activity. However, all of these mutations led to defects in pgRNA packaging. The amount of first-strand DNA synthesized was roughly proportional to the amount of RNA packaged. However, the synthesis of second-strand DNA, which requires two template switches, was not supported by any of the substitutions. These data demonstrate that the HBV capsid is far more than an inert container, as mutations in the assembly domain, distant from packaged nucleic acid, affect reverse transcription. We suggest that capsid molecular motion plays a role in regulating genome replication.

IMPORTANCE The hepatitis B virus (HBV) capsid plays a central role in the virus life cycle and has been studied as a potential antiviral target. The capsid protein (Cp) packages the viral pregenomic RNA (pgRNA) and polymerase to form the HBV core. The role of the capsid in subsequent nucleic acid metabolism is unknown. Here, guided by the structure of the capsid with bound antiviral molecules, we designed Cp mutants that enhanced or attenuated the assembly of purified Cp in vitro. In cell culture, assembly of mutants was consistent with their in vitro biophysical properties. However, all of these mutations inhibited HBV replication. Specifically, changing the biophysical chemistry of Cp caused defects in pgRNA packaging and synthesis of the second strand of DNA. These results suggest that the HBV Cp assembly domain potentially regulates reverse transcription, extending the activities of the capsid protein beyond its presumed role as an inert compartment.

INTRODUCTION

Hepatitis B virus (HBV) chronically infects more than 240 million people worldwide and contributes to 780,000 deaths per year (1). Currently there is no reliable cure for chronic hepatitis (2, 3). HBV is an enveloped double-stranded DNA (dsDNA) virus with an icosahedral core. HBV replicates its genome through an RNA intermediate. During replication, the capsid protein (Cp; also called the core protein) assembles around pregenomic RNA (pgRNA) and reverse transcriptase (the P protein) to form RNA-filled cores (4–8). Unlike HIV, reverse transcription of the HBV genome is a prerequisite for the progeny viruses to leave the host cell (5, 9).

HBV reverse transcription is a complex reaction involving three template switches. P protein is incorporated into capsids during assembly by interacting with the packaging signal, a stem-loop structure at the 5′ end of pgRNA. Next, P acts as the primer and polymerase to initiate minus-strand DNA synthesis using the stem-loop as a template. After synthesizing 4 nucleotides, P switches the template to a point near the 3′ end of pgRNA and then completes the synthesis of minus-strand DNA, digesting the pgRNA template during DNA elongation via its RNase H activity. Upon polymerization of minus-strand DNA to the 5′ end of the pgRNA template, the remaining short segment of the 5′ end of pgRNA and P pirouette to an acceptor site near the 5′ end of the minus-strand DNA for the second template switch. The pgRNA segment is the primer for plus-strand DNA synthesis. After synthesis of a few hundred nucleotides, using the RNA segment as a primer, the plus strand reaches the 5′ end of its template and switches the template for the third and final time to the 3′ end of the minus-strand DNA, which serves as the template for the remainder of plus-strand DNA synthesis. The result of this molecular gymnastics is a relaxed circular DNA (rcDNA) with P covalently linked to the 5′ end of minus-strand DNA (4, 10). Capsids with rcDNA are secreted as virions or deliver their genomes to the nucleus to amplify or maintain the viral covalently closed circular DNA (cccDNA) pool, which maintains the chronic infection (11, 12). cis-Acting elements on the templates facilitate the three template switches (13–18). However, whether the capsid participates in reverse transcription is murky at best.

The HBV capsid is in T=4 or T=3 icosahedral symmetry, and T=4 capsids account for a major population in the natural HBV pool (19–23). The basic building block of HBV capsid is Cp, which is 183 amino acids (aa) in length and exists as a homodimer in solution (24). The N-terminal 149 aa are the assembly domain (Cp149). The C-terminal 150 to 183 aa are enriched in arginine and have nucleic acid-binding properties (25). Also in the C-terminal domain, differential phosphorylation of serine and threonine sites has been shown to regulate DNA synthesis (26–30). However, the role of the assembly domain of Cp on reverse transcription has not been investigated.

HBV capsid assembly has extensively been studied in vitro using Cp149 (31–33). Weak protein-protein interactions at the interdimer interface drive capsid assembly and maintain capsid stability. The HBV Cp is a dynamic molecule that can adopt multiple conformations and undergo transient unfolding/refolding transitions (34–36). We have hypothesized that allosteric changes regulate its assembly (37). Mutations at the intradimer interface affect assembly biophysics and correlate with changes in virus secretion (38, 39). Small antiviral molecules, called core protein allosteric modulators (CpAMs), can target the interdimer interface and modulate assembly. For example, heteroaryldihydropyrimidines (HAPs) and phenylpropenamides bind a hydrophobic pocket at the interdimer interface and enhance assembly kinetics and thermodynamics (40–44). The phenylpropenamides notably affect the Cp tertiary structure (44). Remarkably, a hydrophobic substitution at the HAP binding pocket of Cp demonstrated a correlation between enhancing assembly and inhibiting virus replication (45). These studies showed that the interdimer interface plays a critical role in regulating capsid assembly and replication.

Here we hypothesize that biophysical properties of the capsid affect nucleic acid packaging and metabolism. An implication of this hypothesis, on the basis of thermodynamic linkage (46), is that, conversely, the nucleic acid can affect the capsid. Reverse transcription requires three template switches, likely to have a high energy barrier. The resulting dsDNA also imposes strain on the capsid. To test our hypothesis, we engineered capsid biophysical properties by filling the HAP binding pocket with different hydrophobic residues, which changed Cp assembly kinetics and association energies. These substitutions led to complex changes to HBV replication in cell culture, altering capsid formation, pgRNA packaging, and reverse transcription. These data provide evidence of a connection between capsid biophysical chemistry and the metabolism of packaged nucleic acid. In effect, the HBV core—the capsid, polymerase, and nucleic acid—acts as a finely tuned molecular machine.

MATERIALS AND METHODS

Cloning of Cp149 mutants and protein purification.

The HBV Cp149-pET11c (the SwissProt database accession number for full-length Cp is P03147) construct was mutated to Cp149-V124F, Cp149-V124L, and Cp149-V124A with a QuikChange mutagenesis kit (Stratagene). Cp149-V124X dimers were expressed in Escherichia coli BL21(DE3) in Terrific Broth or Superior Broth (Athena Enzyme System) with 50 μg/ml carbenicillin at 37°C overnight. Cp149 wild-type (Cp149-WT) and -V124W dimers were purified as previously described (45). Cp149-V124A, -V124L, and -V124F dimers were purified by the same process as Cp149-WT with a few modifications. Cp149-V124F dimers were assembled at 50 mM NaCl and fell apart on the column. Initially, we did not know that high NaCl concentrations would help stabilize the V124F capsids on the column, so we purified the V124F assembly products by centrifugation at 150,000 × g for 30 min at 4°C. Such centrifugation pellets capsids and large intermediates, but not dimers (47). We were able to purify assembly-active Cp149-V124F in this way. Cp149-V124L reassembly was induced with 250 mM NaCl, and assembly products were separated on a column equilibrated with 50 mM HEPES, pH 7.5, 500 mM NaCl to maintain capsid stability. Cp149-V124A reassembly was induced in a step increase of ionic strength from 62.5 mM NaCl, 125 mM NaCl, 250 mM NaCl, and 500 mM NaCl to a final 1 M NaCl. In this way, we achieved efficient Cp149-V124A assembly at high ionic strength while avoiding kinetic traps. The elution buffer for the V124A capsid purification was 50 mM HEPES, pH 7.5, 1 M NaCl.

The extinction coefficient of the Cp149-V124W dimer at 280 nm was 70,025 M⁻¹ cm⁻¹ (45). The extinction coefficient used for the Cp149-V124A, -WT, -V124L, and -V124F dimers was 60,900 M⁻¹ cm⁻¹ at 280 nm. The standard assembly buffer was 50 mM HEPES, pH 7.5, at 23°C with varied NaCl concentrations. The protein stock was treated with 1% to 5% β-mercaptoethanol for 20 min before assembly.

Stopped-flow spectroscopy.

The assembly kinetics of Cp149-V124X were monitored with a stopped-flow spectrometer (SF-300X; KinTek). Ninety-degree light scattering was observed with excitation and emission at 400 nm and 23°C for 200 s after induction of assembly. Only two syringes were used: syringe A contained 2× protein samples; syringe B contained 1 M NaCl in 50 mM HEPES, pH 7.5. Equal volumes of samples from syringes A and B were mixed to induce Cp149-V124X assembly at 500 mM NaCl. Varied protein concentrations were chosen to show the kinetic differences between different mutants. Each reaction was repeated three times independently, and the averaged results are shown.

SEC and calculation of thermodynamic parameters.

Size-exclusion chromatography (SEC) and calculation of thermodynamic parameters were done as previously described (45). Cp149-V124X mutants (where X is V, L, or F) were induced to assemble at 100 mM NaCl to equilibration for 72 h or longer at 23°C. The Cp149-V124A association energy is very weak, and it was induced to assemble at 700 mM NaCl. Capsid and dimer fractions were separated by use of a 21-ml Superose 6 column (GE Life Sciences) mounted to a Shimadzu high-pressure liquid chromatography (HPLC) system. The pairwise association energy (ΔG_contact) and the apparent dissociation constant (K_D,apparent) were calculated as previously described (31, 32). Briefly, based on 240 intersubunit contacts in a T=4 capsid, K_capsid = [capsid]/[dimer]¹²⁰, K_D,apparent = K_capsid^(−1/119), K_capsid = Π_j s_i K_contact²⁴⁰, and ΔG_contact = −RT ln K_contact. K_capsid is the equilibrium constant for capsid assembly; [capsid] and [dimer] are the capsid and dimer concentrations at equilibrium, respectively; and Π_j s_i is a statistical term that describes the degeneracy of the association on the basis of capsid geometry. R is the universal gas constant and T is the temperature in Kelvin. The buried hydrophobic area of residue 124 was taken from reference 48.

HAP12 titration of Cp149-V124X assembly.

HAP12 (the most active HAP molecule) titration of Cp149-V124X assembly was done as previously described (45). Cp149-V124X (10 μM) was incubated with HAP12 (2.5 to 60 μM) at 23°C for 20 min and then assembled in the presence of 50 mM NaCl to reach equilibrium for 24 to 72 h. Capsid, abnormal structures, and dimer fractions were separated and quantified by SEC. For Cp149-V124A and -V124L, assembly products were separated with two Bio SEC-5 HPLC columns in series, the first with a pore size of 1,000 Å and the second with a pore size of 300 Å (Agilent). All the other mutant assembly products were separated with a Superose 6 column, as described in reference 45.

HAP12 absorbance and light scattering were subtracted in the calculation of Cp149-V124A, -WT, and -V124F aberrant structure concentrations. The HAP12 absorbance in the Cp149-V124W peak was not detectable. Only the light scattering of capsids was subtracted in the calculation of the Cp149-V124W concentration. For Cp149-V124L, we observed a decreased total area at very high HAP12 concentrations (50 to 60 μM), as well as HAP12 absorbance in the aberrant structure peak. The decreased total area at very high HAP12 concentrations was probably due to large aggregations of the V124L assembly products that were removed by the filter coupled to the Bio SEC-5 columns. The Cp149-V124L free dimer percentage was calculated by dividing the dimer area by the total protein area in the absence of HAP12.

Samples of Cp149-V124X assembly products under different conditions were prepared and checked by transmission electron microscopy as previously described (45).

Cell culture and transfections.

All plasmids expressing HBV used in the cell culture experiments were derived from the plasmids described previously (45). These plasmids expressed pgRNA under the control of the cytomegalovirus (CMV) immediate early promoter. Plasmid P^Y63F C^X expressed pgRNA, P^Y63F, Cp183-V124X (where X is A, the WT amino acid, L, F, or W), and HBx protein but did not express envelope proteins. The Y63F mutation eliminated P's priming function, thus preventing reverse transcription, but the mutant with the Y63F mutation was functional for pgRNA packaging. These plasmids allowed us to analyze capsid formation and pgRNA packaging for Cp183-V124X. Plasmid P^WT C^X expressed pgRNA, P, Cp183-V124X (where X is A, the WT amino acid, L, F, or W) and the HBx protein but did not express envelope proteins. These plasmids allowed us to analyze the synthesis of Cp183-V124X DNA. Green fluorescent protein (GFP) was expressed from a separate plasmid to monitor transfection efficiencies. The cell line Huh7 was cultured and transfected as previously described (45). Ten micrograms of plasmid P^WT C^X or P^Y63F C^X and 250 ng of the GFP expression plasmid were used in each transfection.

Velocity sedimentation and Western blot analysis of HBV capsid assembly.

Cytoplasmic lysates were used to detect Cp and GFP by Western blot analysis, as previously described (45).

Cytoplasmic lysate from P^Y63F C^X transfection was analyzed with velocity sedimentation and Western blotting, as described previously (26). Briefly, a fraction of lysate was loaded onto a step sucrose gradient (15%, 30%, 45%, and 60% [wt/vol]), and the gradient was centrifuged in an SW60 rotor for 2.5 h at 52,000 rpm and 10°C. Twelve fractions, each of which was 365 μl, were carefully separated from top to bottom. Ten microliters of each fraction was run on an SDS-polyacrylamide gel, followed by detection of Cp as described previously (33).

Analysis of pgRNA packaging in velocity sedimentation fractions.

A portion of the velocity sedimentation fractions was electrophoresed through a 0.8% agarose gel, followed by transfer to a polyvinylidene difluoride membrane for fluorescence applications (PVDF-FL membrane; Millipore) and a Hybond-N membrane (GE Life Sciences). The two membranes were placed front to back, and this method gave reliable, quantitative signals for both Cp and pgRNA detection. Native capsid was detected by immunostaining of the PVDF-FL membrane. Packaged pgRNA was detected using ³²P end-labeled oligonucleotide probes specific to pgRNA.

Southern blot analysis of encapsidated DNA.

Nucleic acid from capsids was isolated from cytoplasmic lysates of P^WT C^X transfections and detected as previously described (45).

RESULTS

Design of substitutions at the HBV Cp interdimer interface to modulate Cp biophysics.

HBV assembly is driven by hydrophobic interactions at the interdimer interface, where about 75% of the buried surface area is hydrophobic (31, 41). CpAMs bind a hydrophobic pocket at the interface and strengthen protein-protein interactions by filling that gap (41, 44). Previously, by replacing V124 at the HAP binding pocket with tryptophan, we designed a Cp mutant with enhanced assembly kinetics and association energy (45). Cp-V124W recapitulated the effect of CpAMs both in vitro and in cell culture. To further investigate the Cp interdimer interface and vary its biophysical properties, we designed a series of V124X substitutions (where X is A, L, or F) (Fig. 1). We predicted that these substitutions would alter assembly properties in proportion to their change on the buried hydrophobic surface area at the interdimer interface.

FIG 1.

HBV Cp structure and substitution design at the interdimer interface. (A) HBV Cp149 dimer structure with HAP1 (cyan spheres) bound (Protein Data Bank accession code 2G34). (B) Residue V124 (yellow spheres) contacts HAP1 (cyan spheres) at the Cp interdimer interface viewed from the interior of the capsid. (C) Hydrophobic residues used as replacements at residue V124 and the structure of HAP12. Me, methyl. Panels A and B are adapted from Tan et al. (45).

The assembly domain of Cp (Cp149) was used in all the biophysical studies on capsid assembly in vitro. The full-length Cp (Cp183) was used to characterize in vivo assembly, pgRNA packaging, and reverse transcription. The dimeric wild-type (WT) Cp is referred to as Cp149-WT or Cp183-WT. The designed dimeric Cp mutants are referred to as Cp149-V124X or Cp183-V124X (where X can be A, L, F, or W). Some of the Cp149-V124W in vitro assembly data (association energy and HAP12 titration) are adapted from a previous study performed under the same conditions (45).

The Cp association energy correlates with the size of the hydrophobic substitution.

Capsid assembly is an entropy-driven process, suggesting that it is particularly sensitive to changes in the buried hydrophobic surface area (49). We determined the association energy from equilibrated assembly reactions by quantifying the assembly products by SEC (Fig. 2). Cp149-V124W, -V124F, -V124L, and -WT assembly was quantified at 100 mM NaCl at 23°C. For Cp149-V124A, assembly was studied at 700 mM NaCl, because its association energy is too weak to be determined at low ionic strength.

FIG 2.

Thermodynamic study of Cp149-V124X. (A) The equilibrium concentrations of capsid and dimer were determined by size exclusion chromatography. Two representative chromatographs show the elution of 10 μM Cp149-WT and Cp149-V124F assembled at 100 mM NaCl. mAU, milli-absorbance units. (B) Based on the capsid-dimer equilibrium, we extracted the average pairwise association energy between dimers. There is a linear correlation between the association energy and the hydrophobic surface area at residue 124. The Cp149-WT, -V124L, -V124F, and -V124W association energies were measured at 100 mM NaCl, while the Cp149-V124A association energy was measured at 700 mM NaCl and extrapolated to that at 100 mM NaCl.

Due to the steep concentration dependence of the 120-dimer capsid assembly, at equilibrium the concentration of the Cp dimer remains approximately constant. K_D,apparent, the apparent dissociation constant, is approximately equal to this observable pseudocritical concentration (Table 1) (31, 50). The pairwise association energy (ΔG_contact) for Cp149-V124X was calculated from K_D,apparent with the assumption that the 240 interdimer contacts in a T=4 capsid are equivalent. For Cp149-WT and the Cp149-V124W, Cp149-V124F, and Cp149-V124L mutants, K_D,apparent was readily measured at 100 mM NaCl, allowing direct comparison, but this was not so for Cp149-V124A. However, based on the ionic strength dependence of the association energy, we estimated the ΔG_contact of Cp149-V124A at 100 mM NaCl by extrapolation (31, 51). The calculated K_D,apparent for Cp149-V124A was 1,750 μM at 100 mM NaCl, which is almost 60 mg/ml, a concentration that is experimentally inaccessible.

TABLE 1.

Pairwise association energy and T number of HBV Cp149-V124X capsids^a

Cp149	ΔGcontactb (kcal/mol)	KD,apparentc (μM)	% T=4d
V124A	∼−1.66	∼1,750	83 ± 3
WT	−2.74 ± 0.04	43.28 ± 4.97	78 ± 6
V124L	−3.10 ± 0.02	11.86 ± 0.51	ND
V124F	−3.46 ± 0.02	0.46 ± 0.02	55 ± 5
V124W	−3.84 ± 0.14	0.99 ± 0.42	44 ± 10

^aValues are means ± standard deviations.

^bΔG_contact was measured for all Cp149-V124X capsids at 100 mM NaCl and 23°C except Cp149-V124A, for which assembly was measured at 700 mM NaCl. The experimental measurement of ΔG_contact for Cp149-V124A was −2.92 ± 0.08 kcal/mol, and that for K_D,apparent was ∼26 μM. ΔG_contact was extrapolated to the value obtained with 100 mM NaCl (see the text) on the basis of the dependence of contact energy on ionic strength (31).

^cThe experimental measurement of K_D,apparent for Cp149-V124A was ∼26 μM at 700 mM NaCl and 23°C.

^dThe percentage of T=4 capsids was based on counting well-defined particles from negatively stained micrographs of Cp149-V124X capsids (Fig. 3, top). Cp149-V124X capsids were assembled at ionic strengths to highlight the capsid formation of each mutant identified at the top of Fig. 3. ND, not determined. Cp149-V124L assembly produced very few normal capsids, and the T number cannot be quantified statistically.

ΔG_contact varied linearly with the buried hydrophobic surface area, which is the predicted correlation (Fig. 2). The extrapolated value for Cp149-V124A fell very close to this line. Excluding Cp149-V124A, the observed slope was −14.6 cal/mol/Ų, which is in excellent agreement with estimates of hydrophobic interactions at protein-protein interfaces in the literature (52, 53). We note that the fraction of T=4 and T=3 capsids affects the calculated value of ΔG_contact by no more than 0.01 kcal/mol, which was below our ability to differentiate.

Morphology of the assembly products.

To determine if the substitutions at the Cp interdimer interface affected capsid assembly morphology, we examined assembly products by transmission electron microscopy. In the absence of HAP12, we observed capsid-like particles for all Cp149-V124X mutants under conditions favoring capsid formation, although they required different NaCl conditions (Fig. 3). Notably, although Cp149-V124L assembly produced some capsids, noncapsid polymers were prevalent even at mild ionic strength. These noncapsid polymers eluted in the void volume on SEC. The V124W and V124F mutants were enriched for T=3 particles (56% and 45%, respectively), while the proportions of T=3 particles in the V124A mutant (17% T=3) were similar to those in the wild type (22% T=3) (Table 1).

FIG 3.

Electron micrographs of Cp149-V124X assembly products. (Top) Under conditions favoring capsid formation, Cp149-V124X assembled into capsid-like particles. Notably, Cp149-V124L mainly yielded irregular noncapsid polymers. Cp149-V124W was assembled at 50 mM NaCl, Cp149-V124F and -V124L were assembled at 100 mM NaCl, Cp149-WT was assembled at 500 mM NaCl, and Cp149-V124A was assembled at 1 M NaCl. (Middle) At 500 mM NaCl, Cp mutants with strong association energies and fast kinetics (V124W, V124F, and V124L) formed complexes with trapped assembly defects. Cp149-V124W had some partial or irregular capsids. Cp149-V124F and -V124L mainly yielded noncapsid polymers. (Bottom) Except for Cp149-V124W, 20 μM HAP12 (with assembly induced at 50 mM NaCl) resulted in the formation of large aberrant Cp149-V124X structures. The scale bar applies to all micrographs.

We further tested the ability of V124X substitutions (V124W, V124F, and V124L) to affect morphology by examining assembly at higher ionic strength, 500 mM NaCl. At higher ionic strength, the stronger association energy of the assembly-enhancing mutants inhibits the dissociation of incorrectly assembled subunits, trapping intermediates and leading to aberrant assembly (50). At 500 mM NaCl, we observed that both Cp149-V124F and -V124L formed noncapsid polymers. However, the defects observed for Cp149-V124F and -V124L must in part be due to the geometry that they impose at the interdimer interface, as we observed only capsid-sized particles for Cp149-V124W under the same conditions (Fig. 3).

Assembly kinetics correlates with the size of the hydrophobic substitution.

Kinetics reveals the ability of Cp to overcome the energy barrier to assembly. We compared the assembly kinetics of Cp149-V124X by 90° light scattering with a stopped-flow spectrometer. The time-dependent light scattering signal indicated the formation of large structures: capsids or kinetically trapped intermediates (Fig. 3). We found that at 500 mM NaCl, we were able to measure the assembly kinetics of all Cp149-V124X mutants, albeit at varied protein concentrations (Fig. 4). Cp149-V124W and -V124F assembled very aggressively, and we observed a well-defined lag phase of assembly only at low protein concentrations (1.5 to 3.5 μM) (Fig. 4A). Cp149-V124W assembled faster than Cp149-V124F at all concentrations. Cp149-V124L barely assembled at 3.5 μM, though assembly was easily observed at 7.5 μM (Fig. 4B). Cp149-WT assembled more slowly than Cp149-V124L at 7.5 μM. Cp149-V124A did not assemble at 10 μM; observation of assembly kinetics required protein concentrations up to 40 μM.

FIG 4.

Assembly kinetics of Cp149-V124X at 500 mM NaCl observed by stopped-flow light scattering. (A) Assembly kinetics of Cp149-V124W and -V124F at 1.5 to 3.5 μM protein concentrations. (B) Assembly kinetics of Cp149-WT, -V124L, and -V124A at 3.5 to 40 μM protein concentrations.

Substitutions at the interdimer interface affect the hydrodynamic radius of Cp dimers.

Assembly kinetics for capsids, a multistep cascade of reactions, is expected to be sensitive to not only the stability of intermediates (Table 1) but also the ability of Cps to participate in assembly, which is regulated by Cp allostery. The conformations of Cp mutants are reflected on SEC due to differences in their hydrodynamic (Stokes) radii. The assembly-enhanced dimer, Cp149-V124W, eluted faster than Cp149-WT on SEC, indicating that it had a more compact conformation than the WT (45). Cp149-V124F eluted more slowly than the WT but faster than the V124W mutant, indicating that Cp149-V124F is relatively more compact than the WT but less compact than the V124W mutant (Fig. 5). In contrast, the assembly-inactive dimer, Cp149-Y132A, eluted earlier than the WT (37). Altogether, these SEC results suggest a link between a compact dimer conformation and faster assembly kinetics.

FIG 5.

Cp149-V124X dimer elution positions on a Superose 6 column. Though all mutants had nearly the same mass, Cp149-Y132A eluted the fastest, indicating that it has the largest Stokes radius. Cp149-V124W eluted the slowest, indicating that it has the most compact conformation. Cp149-V124F eluted between the WT and the V124W mutant. The elution position correlates with assembly kinetics and is consistent with allosteric changes in the Cp149 average conformation and, thus, the allosteric regulation of HBV assembly.

Substitutions at the interdimer interface change the size of the hydrophobic pocket.

Since the V124X substitutions were designed to modify the HAP pocket at the interdimer interface, we predicted that Cp149-V124X would have different sensitivities to HAPs. For these studies, we examined Cp149-V124X sensitivity to HAP12, a highly active CpAM requiring low NaCl concentrations (42). We titrated Cp149-V124X with HAP12, allowed assembly at 50 mM NaCl and 23°C to reach equilibrium, and quantified the extent of assembly by measuring the percentage of free dimer by SEC (Fig. 6). Cp149-V124W was relatively insensitive to HAP12 treatment. At 60 μM HAP12, assembly of the V124W mutant was increased by only ∼15%. Cp149-V124F assembly was increased by ∼70% at 60 μM HAP12. Cp149-WT, -V124L, and -V124A did not assemble at 50 mM NaCl without HAP12. Nonetheless, at 60 μM HAP12 all three achieved about 95% assembly. At lower HAP12 concentrations (ca. 10 μM HAP12) a rank order of HAP12 resistance showed a correlation to the size of the V124X side chain: V124W mutant > V124F mutant > V124L mutant ≥ WT > V124A mutant.

FIG 6.

HAP12 titration on Cp149-V124X assembly. Cp149-V124X (10 μM) was induced to assemble at 50 mM NaCl and different HAP12 concentrations. The dimer fraction at equilibrium was used to indicate the assembly extent. The sensitivity to HAP12 correlated with the size of the V124X residue, with Cp149-V124W being nearly insensitive to the small molecule. Data for WT and V124W are from Tan et al. (45) and are included for comparison.

HAPs can induce Cp assembly into complexes dominated by hexagonal repeats (40, 54, 55). However, Cp149-V124W assembled into capsid-like particles in the presence of HAP12 (Fig. 3), which was consistent with its high insensitivity to HAP treatment (45). In contrast, HAP12 induced the formation of large aberrant structures for Cp149-V124A, -WT, -V124L, and -V124F assembly at a 1:2 dimer-to-HAP12 ratio (Fig. 3).

Substitutions at the Cp interdimer interface affect virus assembly in predictable ways in cell culture on the basis of their biophysical chemistry.

In vitro characterization of Cp149-V124X assembly led to a comprehensive understanding of the biophysical chemistry of HBV capsid assembly. However, in cells assembly is a more complex reaction that occurs in a complex cytoplasmic environment and is coordinated with pgRNA packaging. To determine if the assembly of capsids in a cell would parallel the biophysical chemistry of in vitro capsid assembly, we examined Cp183-V124X assembly in Huh7 cells. We transfected Huh7 cells with plasmids carrying an HBV genome that expressed pgRNA, a derivative of P (P^Y63F) that is active for pgRNA packaging but unable to synthesize DNA, HBx protein, and Cp183-V124X. The use of P^Y63F (to observe capsid formation and pgRNA packaging) avoided the potential complication of heterogeneity of the nucleic acid content due to the conversion of pgRNA to DNA (10). By Western blotting, we found that cytoplasmic Cp183-V124X accumulated to a similar level for all of the mutants (Fig. 7A).

FIG 7.

V124X substitutions affect capsid formation, pgRNA packaging, and reverse transcription in Huh7 cells. (A) A Western blot shows that Cp183-V124X mutants accumulated to similar levels in the cytoplasm. Relative sample volumes are indicated below the image. (B) Velocity sedimentation profiles of Cp183-V124X assembly obtained by Western blotting show that capsid formation correlates with in vitro assembly. Cp183-WT capsids sedimented at fractions 7 to 9, with dimers sedimenting at fractions 1 to 3. Cp183-V124W and -V124A had capsid sedimentation profiles similar to the WT profile, though Cp183-V124W had less free dimer and Cp183-V124A had more free dimer. Cp183-V124W capsid sedimentation was shifted to a slightly lower density. Cp183-V124F and -V124L had a large proportion of capsid assembly products sedimenting at intermediate fractions. Lane U, a sample of unfractionated cell lysate. (C) pgRNA was detected in capsid fractions 7 to 9 by Northern blotting. Cp183-V124W and -V124A capsids packaged much less pgRNA than the WT. Cp183-V124W is more defective in pgRNA packaging than Cp183-V124A. The amount of pgRNA packaged by Cp183-V124F and -V124L was too low to be detected. (D) A Southern blot shows that Cp183-V124X differentially affected ssDNA and rcDNA production.

To determine the ability of Cp183-V124X to support the formation of cores in cell cultures, we subjected cytoplasmic capsids to velocity sedimentation centrifugation, fractionated each gradient into 12 fractions, and detected Cp in each fraction by Western blotting. Cp183-WT capsids sedimented characteristically in fractions 7 to 9, with the WT dimers appearing in fractions 1 to 3 (Fig. 7B). Cp183-V124W and -V124A exhibited similar sedimentation profiles, except that there was less free dimer for Cp183-V124W and more free dimer for Cp183-V124A. Notably, the V124W cores sedimented slightly more slowly than the WT cores (shifted to a lower density). The Cp183-V124F and -V124L sedimentation profiles showed a large amount of intermediate fractions that sedimented in between the capsid and the dimer. For Cp183-V124L, the intermediate fractions were dominant compared to the prevalence of the capsids.

Substitutions at the Cp interdimer interface lead to a defect in pgRNA packaging.

For Cp183-V124X, capsid formation in cells was consistent with the biophysical chemistry of Cp assembly in vitro. Because the assembly-accelerating phenylpropenamides lead to reductions in pgRNA packaging (43, 56), we predicted that an assembly-accelerating Cp with no other defect in capsid morphology, e.g., Cp183-V124W, would similarly encapsidate less pgRNA. Conversely, Cps with slower assembly kinetics, e.g., Cp183-V124A, would have little or no effect on pgRNA packaging due to the lack of kinetic competition between the packaging and assembly of empty capsids. We examined the packaged pgRNA in V124W, V124A, and WT capsid fractions resolved on sucrose gradients. To measure pgRNA encapsidation, each capsid fraction was electrophoresed on a native agarose gel and the pgRNA content was measured by Northern blotting.

We detected pgRNA in Cp183-WT fractions 7 to 9, indicating that the capsids in these fractions contained pgRNA (Fig. 7C). In contrast, though protein levels were similar, the pgRNA levels in the Cp183-V124W and -V124A capsids were much lower than those in the WT capsid, indicating that they packaged less pgRNA. Cp183-V124F and -V124L had severe defects in capsid formation, making the reliability of measurement of packaged pgRNA difficult; the pgRNA signal was at or below the detection limit.

Substitutions at the Cp interdimer interface cause pleiotropic defects in reverse transcription.

We then examined the effect of the V124X substitutions on reverse transcription. Huh7 cells were transfected with a plasmid carrying an HBV genome expressing pgRNA, a WT P protein, the HBx protein, and Cp183-V124X. The codon for V124 of Cp does not overlap any known genes or cis-acting elements required for successful genome replication (13–18). Southern blotting of cytoplasmic, encapsidated Cp183-WT DNA showed the characteristic pattern of full-length rcDNA, double-stranded linear DNA (dlDNA), partially double-stranded replication intermediates, and a large amount of single-stranded DNA (ssDNA). All Cp183-V124X capsids accumulated less ssDNA than the WT and no detectable rcDNA (Fig. 7D). When 10-fold more sample for the Cp183-V124X variants was loaded and robust levels of ssDNA were seen, no rcDNA was detected. The presence or absence of micrococcal nuclease in sample preparation did not affect the yield of DNA (26), indicating that purified capsids were intact, but we cannot exclude the possibility that dsDNA destabilized mutant capsids, leading to their destruction prior to isolation. The relative accumulation of ssDNA was WT > V124A mutant ≫ V124F mutant ≈ V124W mutant ≫ V124L mutant. Thus, the V124X substitutions had pleiotropic effects on DNA synthesis.

DISCUSSION

The overarching goal of this study was to correlate the in vitro properties of capsid assembly and stability with the in vivo activities of capsid assembly, RNA packaging, and reverse transcription. In this discussion, we first analyze the in vitro assembly data, which are based on a highly simplified system. Then, we analyze the effect of the V124X substitutions on cellular core protein activities. We observed striking parallels, except with respect to transcription of the plus-sense DNA strand, suggesting that the capsid itself plays a role in catalyzing elements of the reverse transcription process.

In vitro, the biophysical chemistry of the HBV Cp interdimer interface is critical for capsid assembly. Using structure-guided design, we have found that substitutions at the HAP binding site within the Cp interdimer interface modulate in vitro assembly in predictable ways that correlate with the change in the buried hydrophobic surface area. More interestingly, substitutions that enhanced or attenuated assembly in vitro also affected capsid formation, pgRNA packaging, and reverse transcription in cell culture. Our results lead to a model in which the interdimer interface of HBV capsids plays a pleiotropic role during virus replication.

Our substitutions had the predicted effects on capsid assembly, e.g., their effects on ΔG_contact and HAP12 binding. The observed −14.6 cal/mol of contact energy per square angstrom is surprisingly strong, given the poor complementarity at the interdimer interface, which has numerous gaps, including the hydrophobic pocket for CpAM binding. Our observed free energy is only slightly lower than the typical 20 cal/mol/Ų of contact energy for protein-protein interactions between highly complementary surfaces, such as receptor-ligand interactions (53). The sensitivity of Cp149-V124X to HAP12 correlates well with the size of the hydrophobic substitution, confirming that the V124X residues fill the pocket to different extents.

Some substitutions affected in vitro assembly geometry. Cp149-V124L preferentially formed noncapsid polymers, even at low ionic strength, which should minimize kinetic traps (32, 33). At high ionic strength, Cp149-V124F formed noncapsid polymers similar to those seen with Cp149-V124L. These assembly defects were probably due to an altered geometry at the interdimer interface competing with the energy minimum of a complete capsid, where all dimers are involved in the maximum number of interdimer interactions. Cp149-V124W, the most aggressive Cp, did not produce any noncapsid polymers under any conditions that we tested. However, we did observe some partial capsids at high ionic strength, which probably were metastable intermediates with insufficient dimer available to complete them.

Our substitutions had an unexpected effect on the Stokes radius of Cp149-V124X that correlated with assembly kinetics. A smaller Stokes radius correlated with faster kinetics, e.g., for Cp149-V124W and -V124F. This is also consistent with the findings of studies on an assembly-inactive Cp, Cp149-Y132A, which has a larger Stokes radius and thus an extended structure.

In cell culture, HBV capsid assembly was concordant with the in vitro biophysical properties of Cp. Considering the crowded environment and the affinity of the C-terminal domain for nucleic acid, assembly could be predicted to be more aggressive in cells (6). Though the protein levels of the Cp183-V124X mutants were similar to each other in culture, they led to different levels of accumulation of capsids. The V124W substitution, which had a stronger contact energy in vitro, resulted in the production of more capsids and fewer dimers in cells, while the V124A substitution, which led to a weaker contact energy in vitro, shifted the assembly equilibrium to fewer capsids in cells. The V124F and V124L mutations led to a large amount of intermediate fractions, suggesting incomplete capsids or noncapsid protein complexes.

Compared to the pgRNA packaging obtained with WT Cp183, the V124W and V124A substitutions caused defects in pgRNA packaging. Cp183-V124F and -V124L had severe defects in capsid accumulation, making reliable measurement of packaged pgRNA difficult. Cp183-V124W capsids sedimented more slowly than the WT capsid, possibly indicating the presence of empty capsids, partial capsids, or a larger fraction of T=3 capsids. The formation of empty capsids has been observed with phenylpropenamide treatment, which speeds up assembly dramatically and inhibits pgRNA packaging (43, 56); empty particles are also observed with the WT Cp (57). These results suggest that there is a kinetic competition between the assembly of empty capsids and the process of pgRNA packaging: increasing or decreasing assembly kinetics will disrupt this process. This concept is consistent with the hypothesis that a pgRNA-P complex is required for packaging of pgRNA (58); i.e., such a complex is a nucleating factor. On the basis of the behavior of V124A, we suggest that failure to complete assembly of a capsid in a timely manner is equally disruptive.

The amount of ssDNA synthesized in the Cp183-V124W and -V124A capsids was approximately proportional to the amount of pgRNA packaged. Hence, minus-strand DNA synthesis was not affected by the V124 substitutions. However, after accounting for the lower level of ssDNA synthesized, all substitutions had an independent, additional defect in the synthesis of full-length rcDNA. This defect of plus-strand DNA synthesis suggests that the interdimer interface plays an additional and unexpected role in plus-strand DNA synthesis. Recent structural studies suggest that the capsid holds pgRNA in a defined scaffold (10, 59) and that HBV Cp can act as an RNA chaperone (59, 60). Thus, we hypothesize that the conformational mobility at the interdimer interface is necessary for plus-strand DNA elongation and/or plus-strand template switching.

A defect arising from the V124 mutations that is specific to plus-strand synthesis came as a surprise. In vitro and in cell culture, we observed that assembly led to morphologically normal capsids. In cell culture, while RNA packaging was modestly attenuated, first-strand DNA synthesis was not. The mutation itself is buried at a protein-protein interface of the assembly domain, distant from the packaged nucleic acid. As the structures of the V124 mutant capsids are similar to the structure of the wild type, we propose that the mutations that affect reverse transcription do so in part by changing the dynamic properties of the HBV capsid. However, HBV capsids are known to be dynamic. Capsid proteins in a capsid can transiently unfold and refold (35). During assembly, Cp is proposed to undergo conformational changes to become assembly competent (31, 34, 37). C-terminal domains can move and structurally modify packaged RNA in response to phosphorylation (59). The dynamic properties of Cps are also implicated throughout the virus life cycle. A recent review compiled several examples where protein dynamics had an allosteric function (61).

The connection between capsid biophysical chemistry in vitro and reverse transcription in cells suggests that the encapsidated RNA/DNA could act synergistically with capsid dynamics. Communication between the nucleic acid content and the capsid conformation, nucleoprotein allostery, could contribute to the maturation signal that distinguishes capsids containing rcDNA (mature genomes) from capsids containing immature genomes (57, 62). It has been observed that mature capsids can have increased protease and nuclease sensitivity, suggesting a dynamic change of capsids during reverse transcription (63). A naturally occurring mutation of Cp residue 97 (Phe or Ile) to leucine, a mutation located at the intradimer interface, enhances assembly and increases secretion of immature ssDNA capsids, also implying a linkage (46) of capsid biophysics to the secretion/maturation signal (38, 39).

In summary, our findings indicate that the interdimer interface of the subunits of the HBV capsid has pleiotropic effects on capsid assembly, pgRNA packaging, and reverse transcription. Our findings suggest that the capsid is an essential part of a carefully tuned replication machine that works together with the cis-acting elements on the HBV genome to facilitate DNA synthesis. Changing the biophysical chemistry of capsids affects several stages of genome replication. In the current study, substitutions at the HAP binding pocket of the Cp interface are destructive to HBV replication. In effect, the HBV core is an integrated nucleic acid-processing machine whose activities correlate with capsid biophysics.