Hepatitis B virus (HBV) is an enveloped DNA virus with a spherical capsid (or core). The capsid is constructed from 120 copies of the homodimeric capsid protein arranged with T = 4 icosahedral symmetry. We examined in vitro assembly of purified E. coli expressed HBV capsid protein. After equilibration, concentrations of capsid and dimer were evaluated by size exclusion chromatography. The extent of assembly increased as temperature and ionic strength increased. The concentration dependence of capsid assembly conformed to the equilibrium expression: K(capsid) = [capsid]/[dimer](120). Given the known geometry for HBV capsids and dimers, the per capsid assembly energy was partitioned into energy per subunit-subunit contact. We were able to make three major conclusions. (i) Weak interactions (from -2.9 kcal/mol at 21 degrees C in low salt to -4.4 kcal/mol at 37 degrees C in high salt) at each intersubunit contact result in a globally stable capsid; weak intersubunit interactions may be the basis for the phenomenon of capsid breathing. (ii) HBV assembly is characterized by positive enthalpy and entropy. The reaction is entropy-driven, consistent with the largely hydrophobic contacts found in the crystal structure. (iii) Increasing NaCl concentration increases the magnitude of free energy, enthalpy, and entropy, as if ionic strength were increasing the amount of hydrophobic surface buried by assembly. This last point leads us to suggest that salt acts by inducing a conformational change in the dimer from an assembly-inactive form to an assembly-active form. This model of conformational change linked to assembly is consistent with immunological differences between dimer and capsid.

The capsids of most spherical viruses are icosahedral, an arrangement of multiples of 60 subunits. Though it is a salient point in the life cycle of any virus, the physical chemistry of virus capsid assembly is poorly understood. We have developed general models of capsid assembly that describe the process in terms of a cascade of low order association reactions. The models predict sigmoidal assembly kinetics, where intermediates approach a low steady state concentration for the greater part of the reaction. Features of the overall reaction can be identified on the basis of the concentration dependence of assembly. In simulations, and on the basis of our understanding of the models, we find that nucleus size and the order of subsequent "elongation" reactions are reflected in the concentration dependence of the extent of the reaction and the rate of the fast phase, respectively. The reaction kinetics deduced for our models of virus assembly can be related to the assembly of any "spherical" polymer. Using light scattering and size exclusion chromatography, we observed polymerization of assembly domain dimers of hepatitis B virus (HBV) capsid protein. Empty capsids assemble at a rate that is a function of protein concentration and ionic strength. The kinetics of capsid formation were sigmoidal, where the rate of the fast phase had second-power concentration dependence. The extent of assembly had third-power concentration dependence. Simulations based on the models recapitulated the concentration dependences observed for HBV capsid assembly. These results strongly suggest that in vitro HBV assembly is nucleated by a trimer of dimers and proceeds by the addition of individual dimeric subunits. On the basis of this mechanism, we suggest that HBV capsid assembly could be an important target for antiviral therapeutics.

Hepatitis B virus (HBV) is an enveloped virus with an icosahedral capsid. Its homodimeric capsid protein ("core antigen") assembles into particles of two sizes, one with T = 3 icosahedral symmetry (90 dimers) and the other with T = 4 symmetry (120 dimers). We have investigated this assembly process in vitro, using a variety of purified, bacterially expressed, capsid proteins. All of our constructs lacked the predominantly basic C-terminal 34 amino acids of the full-length capsid protein (183 amino acids) and were further truncated to terminate at specific points between residues 138 and 149. While the smallest construct (138 residues) did not assemble into capsids, those terminating at residue 140, and beyond, assembled into mixtures of T = 3 and T = 4 particles. The two kinds of capsids could be separated on sucrose gradients and did not interconvert upon protracted storage. The proportion of T = 3 capsids, assayed by sucrose gradient fractionation, analytical ultracentrifugation, and cryoelectron microscopy, was found to increase systematically with larger deletions from the C-terminus. The variant terminating at residue 149 formed approximately 5% of T = 3 capsids, while the 140-residue protein produced approximately 85% of this isomorph. For the 147-residue capsid protein, the structures of both capsids were determined to 17 A resolution by three-dimensional reconstruction of cryoelectron micrographs. In these density maps, the boundaries of the constituent dimers can be clearly seen and the quaternary structures of the two capsids compared. The arrangement of dimers around their icosahedral five-fold axes is almost identical, whereas the quasi-six-fold arrangements of dimers are distinctly different.