|Home | About | Journals | Submit | Contact Us | Français|
Understanding self-assembly of icosahedral virus capsids is critical to developing assembly-directed antiviral approaches and will also contribute to the development of self-assembling nanostructures. One approach to controlling assembly would be through the use of assembly inhibitors. Here we use Cp149, the assembly domain of the hepatitis B virus capsid protein, together with an assembly-defective mutant, Cp149-Y132A, to examine the limits of the efficacy assembly inhibitors. By itself, Cp149-Y132A will not form capsids. However, Cp-Y132A will co-assemble with wildtype protein based on light scattering and size exclusion chromatography. The resulting capsids appear indistinguishable from normal capsids. However, co-assembled capsids are more fragile, with disassembly observed by chromatography under mildly destabilizing conditions. The relative persistence of capsids assembled under conditions where association energy is weak compared to the fragility of those where association is strong suggests a mechanism of “thermodynamic editing” that allows replacement of defective proteins in a weakly associated complex. There is fine line between weak assembly, where assembly-defective protein is edited from a growing capsid, and relatively strong assembly, where assembly defective subunits may dramatically compromise virus stability. Thus, attempts to control virus self-assembly (with small molecules or defective proteins) must take into account the competing process of thermodynamic editing.
Viruses have a long history as pathogens (see (1) and references therein) and a much shorter history as tools for the development of nanostructures (2-8). Biologically, virus assembly requires high fidelity and rapid kinetics. Virus assembly is a paradigm for nanostructure self-assembly. Controlling virus assembly therefore has value for development of new classes of antiviral agents and for developing viruses as platforms for nanotechnology.
The object of this study, hepatitis B virus (HBV) is an unusually serious public health problem as a major contributing factor to cirrhosis and hepatocellular carcinoma that chronically infects approximately 400 million individuals (9). HBV is an enveloped DNA virus with an icosahedral core. The protein shell of HBV's core, the capsid, is a self-assembling complex of 120 core protein homodimers (10, 11). In vivo, the nucleoprotein core is critical to virus lifecycle (9, 12-15). In vitro, empty capsids can be recapitulated by the 149 residue core protein assembly domain (Cp149) (16). The interdimer contacts that form a capsid are compact, though not tightly packed, and dominated by hydrophobic residues (17, 18). Small molecules that occupy gaps at this interface are able to strengthen the interdimer association energy and divert assembly from its normal path (19, 20). HBV assembly has been characterized in detail with the aid of model-based analyses specific for icosahedral assembly (21, 22). There is a well-defined nucleation step, formation of a trimer of homodimers, which is critical for avoiding kinetic traps (23). The individual interdimer contacts are weak, though globally they result in a stable structure; the pairwise association energy strengthens from −3.1 to −4.1 kcal/mol as temperature and ionic strength increase (24). Even under conditions where they are nominally unstable, HBV capsids can persist due to hysteresis to dissociation (25). These characteristics – nucleated assembly, weak interaction energy, and hysteresis – may be general to assembly of all virus-like particles (7, 23, 26).
In general, virus capsid assembly can be disrupted by inappropriately enhancing (19, 27), misdirecting/distorting (28, 29), or inhibiting association (29-32). We hypothesize that there are two modes of action for assembly inhibitors. They can weaken or altogether prevent bound subunit from assembling; such an inhibitor will be required in great molar excess as an antiviral. Alternatively, an inhibitor that poisons further assembly by its association with a growing capsid can have a more far-reaching effect. In this study we develop an assembly deficient form of the HBV core protein as a model of an assembly-inhibited protein and investigate its effect on in vitro capsid assembly.
Cp149 was expressed in E coli from a pET11-based plasmid, pCp149, and purified using the detailed protocol described previously (33).
The Cp149-Y132A mutant was generated from pCp149 using a Quik-Change kit (Strategene). E. coli with the mutant plasmid were grown in LB broth with 50 μg/ml carbenicillin at 32°C without induction. Cells suspended in the same lysis buffer used for the Cp149 purification (33) were lysed by sonication. Cell debris was removed by centrifugation for 50’ at 20,000g. Protein was precipitated from the supernatant with 20% ammonium sulfate for 1 hr followed by centrifugation for 50’ at 20,000g. This pellet was resuspended in 20 ml 100 mM Tris HCl, 100 mM NaCl pH 9.0 2 mM DTT, which was loaded onto a 5×70 cm column packed with Sephacryl S300 and equilibrated with the resuspension buffer. Fractions were evaluated by SDS-PAGE and absorbance; Cp149-Y132A typically eluted at about 800 ml, consistent with a dimer rather than a capsid state. Pooled fractions were then adsorbed to a 5ml Hi-Trap Q column, which was then washed with 5 column volumes of 50 mM Tris HCl pH 9.0, 80 mM NaCl, 2 mM DTT (buffer QA). Samples were eluted with a gradient of 0-30% buffer QB (buffer QA with 2M NaCl). Cp149-Y132A eluted at about 0.2 M NaCl. Fractions were pooled based on SDS-PAGE and absorbance and then dialyzed into a storage buffer of 10mM Tris HCL pH 9.0, 2 mM DTT.
Observations of kinetics by light scattering were performed as previously described (23). Briefly, scattering was observed with a Spex Flouromax 2 fluorometer set for 320 nm for both excitation and emission. Samples were in a black masked microcuvette with a 0.3 cm path length (Hellma). Initially, scattering was observed for a protein sample at twice the final concentration in buffer containing no NaCl, then, after 10 s, buffer with twice the final NaCl concentration was manually added and mixed. Some samples were stored at a constant temperature and the light scattering read after 24 hours. Light scattering is reported in arbitrary units.
SEC was performed using a 21 ml Superose-6 column equilibrated with 0.1 M HEPES pH 7.5 and either 0.15 of 1 M NaCl. The column was mounted on an AKTA-FPLC recording absorbance at 280 nm. Peaks were quantified after manual baseline correction using the supplied Unicorn software (GE Health Sciences). Long-term kinetics were observed using a 21 ml Superose-6 column equilibrated with 50 mM HEPES pH 7.5 and 1 M NaCl. The column was mounted on a Shimadzu-HPLC equipped with an autoinjector to facilitate the long time course. Peaks at 280 nm were quantified after manual baseline correction using the supplied LCSolutions software (Shimadzu).
Samples were adsorbed to glow-discharged carbon over paralodian copper grids (EM sciences). Samples were stained with 1-2% uranyl acetate and visualized with an Hitachi H7600 transmission electron microscope equipped with an AMT 2K × 2K CCD camera.
To model a capsid protein complexed with an assembly inhibitor, we designed an assembly-deficient mutant. Interdimer contacts bury 1100 to 1400 Å2 at the four quasi-equivalent interfaces (17, 18). In each case tyrosine 132 contributes about 10% of buried surface (17, 18). The phenolic side chain fits into a hydrophobic pocket formed by P20, D22, F23, F122, W125, A137, P138, I139, and L140A of a neighboring subunit. Similar contacts are made independent of quasi-equivalence. Mutation of Y132 to alanine is expected to decrease buried surface and also leave a destabilizing pocket (Fig. 1).
The effect of the Y132A mutation was apparent during purification. Unlike Cp149, which is isolated from E coli as a capsid, Cp149-Y132A was dimeric based on size exclusion chromatography (SEC). The usual disassembly-reassembly purification of Cp149 (33) was necessarily abandoned and replaced with anion exchange chromatography (see methods). The Y132A mutant retains the characteristic blue-shifted fluorescence spectrum of Cp149 and the helix-rich circular dichroism spectrum (data not shown) (25), and as demonstrated in this paper, is capable of participating in assembly, all of which indicate that the mutant is properly folded. Using light scattering, we found that purified Cp149-Y132A showed no sign of assembly at concentrations of 200 μM dimer in 1 M NaCl, conditions where wildtype Cp149 which has a pseudo-critical concentration of < 0.5 μM dimer (23).
We reasoned that Cp149-Y132A retained sufficient molecular complementarity that it could participate in assembly with wildtype Cp149. Initially we used light scattering to examine the effect of varying association energy on co-assembly (Fig. 2). Cp149 assembly shows a strong correlation between ionic strength and the association constant (24). Co-assembling 10μM Cp149 with 20 μM Cp149-Y132A we observed that there was additional light scattering compared to Cp149 alone. The kinetics of co-assembly showed some peculiarities that were most evident in high salt. In 1 M NaCl, Cp149 alone assembles very rapidly, demonstrated by a very rapid increase in light scattering, faster than the dead time of manual mixing, followed by a near-horizontal plateau (Fig. 2A). By comparison, the co-assembly reaction shows a rapid rise followed by a continuing slow increase.
The light scatter enhancement was indeed correlated with ionic strength and, by extension, association constant (Fig. 2B). Proportionally, the enhancement was always approximately two fold. However, it is the raw light scattering that correlates with the amount of Cp149 HBV assembly (23). Thus, these data suggest a substantial increase in the amount of assembly in the time course of this experiment.
As Cp149-Y132A co-assembles with Cp149 but does not assemble on its own in HEPES and NaCl, we investigated whether there was a minimum proportion of Cp149 needed to nucleate assembly (Fig. 3). For these experiments we induced assembly of 1 μM Cp149 at 21°C by addition of 1 M NaCl and observed the amount of light scattering at 24 hours. For 1 μM Cp149 this should yield approximately 50% assembly, i.e. the pseudo-critical concentration is approximately 0.5μM. More importantly for this experiment, the high NaCl was expected to lead to maximal participation of Cp149-Y132A. Control reactions, where more Cp149 was added to the basal 1 μM and then induced to assemble, showed the expected linear increase in scattering consistent with assembly of all protein in excess of the pseudo-critical concentration (21, 24). Addition of Cp149-Y132A increased the light scattering by approximately twofold, after accounting for the scattering of free dimer, in good agreement with the results shown in figure 2. Higher proportions of mutant did not generate any further excess of light scattering, suggesting that a large proportion of the capsid was wildtype Cp149. We also noted that high concentrations of Cp149-Y132A did not decrease light scattering, suggesting that the mutant did not poison assembly.
To further investigate the effects of Cp149-Y132A on assembly, we examined the long-term kinetics of assembly over a 24-hour period. In capsid assembly reactions, after the initial burst of assembly, a steady state of intermediates is generated and the nucleation rate is equal to the rate of capsid formation (22). For formation of the trimer of dimer nucleus of HBV (23):
Where knuc is the microscopic nucleation rate constant in M−1s−1, s is a statistical factor, Kassoc (M−1) is the association constant for pairwise interaction between dimers, and [Cp] is the concentration of free dimer. A co-assembly reaction of 11.7 μM Cp149-Y132A with 5 μM wildtype protein at 0.6M NaCl was monitored by SEC over a 24 hour equilibration period and compared to a control experiment with 5 μM wildtype Cp149 (Fig. 4). As the reaction approached equilibrium, the co-assembly reaction yielded roughly 20% more capsid. The slopes of the steady-state portion of the reaction, d[capsid]/dt, between 5 and 20 hours, is approximately identical between the co-assembly and the wildtype-only assembly. However, the co-assembly takes substantially longer to reach this steady state.
We examined assembly products by size exclusion chromatography (SEC) and electron microscopy (EM). Initially, SEC was performed using a column equilibrated in 1 M NaCl to stabilize co-assembled capsids (Fig. 5A). Samples with 10 μM Cp149 and varying concentrations of mutant and NaCl were incubated overnight before separation of capsid and dimer by SEC. As the concentration of Cp149-Y132A was increased, we observed small but significant (p < 0.02) increases in the yield of capsid for 0.3, 0.6, and 1.0 M NaCl reactions, though not the great excess suggested by light scattering. The minimal increase in assembly led us to question whether the mutant was actually contributing to capsid formation or if the excess light scattering represented some capsid associated aggregation.
To test this possibility we examined co-assembly reactions using SEC columns equilibrated in a lower ionic strength buffer, 0.15 M NaCl) (Fig. 6). Normal Cp149 capsids persist in mildly destabilizing environments of low salt and low concentration due to hysteresis to dissociation (25). In these experiments we used similar reactions to those shown in figure 5 except that reactions were incubated at 37°C where association is greater. Co-assembly reactions induced by 0.15 M and 0.3 M NaCl showed no significant change in the yield of capsid related to Cp149-Y132A, essentially the same result found when the column was equilibrated with 1 M NaCl buffer. However, co-assembly reactions induced by high salt, 0.6 M and 1 M NaCl showed dramatically different results. When these high NaCl reactions were injected onto a SEC column equilibrated with low NaCl, we observed a dose-dependent decrease in the yield of assembled particles. By comparison, recovery of capsids from high salt assembly reactions using pure Cp149 is nearly quantitative. The difference between results for low and high ionic strength SEC indicates the participation of the Y132A mutant in assembly and its sensitivity to association energy.
To define the morphology of the products from co-assembly reactions, samples equilibrated for 24 hours were visualized by negative stain electron microscopy (Fig. 5b). At all NaCl concentrations co-assembled particles appeared essentially the same as those with wildtype protein alone. Micrographs of co-assembly reactions are almost indistinguishable from those of Cp149 reactions except for the background attributable to high concentrations of free protein, in this case Cp149-Y132A. These micrographs are for samples assembled at 21°C, but the same result was observed in reactions incubated at 37°C (data not shown).
Based on structure, we developed a point mutation within the HBV core protein, Cp149-Y132A, as a tool for examining the process of assembly inhibition. We have shown that this mutant is deficient in assembly but is capable of co-assembling with wildtype Cp149. The extent of co-assembly appears to be increased by high ionic strength, which increases the protein-protein association energy of HBV core protein (24). The critical observation to confirm co-assembly was that high salt reactions generated capsids that were unusually fragile as demonstrated by their dissociation during SEC under conditions where the pseudo-critical concentration of Cp149 is about 5μM but Cp149 capsids persist due to hysteresis (25).
These results can be interpreted in terms of a broad framework of nucleated assembly based on a network of weak interactions (Fig. 7), typical for most viruses (34, 35). Based on structure, it is reasonable to posit that Cp149-Y132A has substantially weaker association energy. Cp149-Y132A appears to be incapable of nucleating assembly. In high salt, which strengthens association energy, it may competitively inhibit nucleation by transiently binding to nucleation competent Cp149. This is consistent with the kinetic effect most obvious in high salt in figures 2A and and4.4. With both strand and weak association energy, the mutant participates in assembly. But with weak association energy, mutant subunits (analogous to defective subunits or inhibitor-bound subunits) are thermodynamically edited from the growing capsid.
We examined the effect of Cp149-Y132A on nucleation rate in a long-term kinetics experiment. Based on model studies, assembly reactions rapidly achieve a steady state of intermediates where the rate of capsid formation is equal to the rate of nucleation, thus the rate of capsid production is direct read out of nucleation rate (22). Quantitative interpretation of co-assembly is complicated by the presence of two species of subunit with different association energies, rate constants, and concentrations that vary in ill-defined ways. However, the data suggest limiting models of the role of Cp149-Y132A. In Figure 4 it is apparent that the rate of capsid formation is essentially unchanged by the presence of Cp149-Y132A, although yield is increased. One explanation for the unchanged reaction rate is that the mutant does not participate in nucleation, the rate-limiting step of capsid assembly. In this scenario, the rate of nucleus formation is determined solely by the wildtype Cp149 and Cp149-Y132A participates only in “elongation” of intermediates, hence the greater yield of capsid. It is the weaker association energy of the mutant that leads to the slower establishment of the steady state.
An alternative explanation is that Cp149-Y132A participates in assembly so poorly that it competitively inhibits nucleation but is incorporated during elongation. Nuclei and pre-nuclear intermediates formed with the weakly associating mutant would have a greater tendency to dissociate rather than elongate. Thus, wildtype protein would be tied up unproductively. The resulting slowdown in nucleation would be partially compensated by the high concentration of mutant. The expected effect would be the apparent lag in reaching steady state, as observed.
The behavior of capsids is straightforward. SEC with the column equilibrated in low salt shows distinct differences in the stability of co-assembly products formed at low versus high NaCl. At lower NaCl concentrations, weakly associated proteins have a chance to be thermodynamically edited from growing and completed capsids, i.e. defective proteins will be preferentially excluded in a stochastic equilibrium. Thus at lower NaCl, there is no change in the yield of capsid. At higher NaCl, with stronger association energy, Cp149-Y132A is retained in capsids resulting in unstable capsids that dissociate in low NaCl SEC. Apparently even a small amount of Cp149-Y132A was sufficient to compromise capsid stability. It appears that only a limited amount of Y132A can be incorporated into capsids (Fig. 3), either because it doesn't bind or is edited away. Two possible explanation arise for this result: (i) a patch of defective protein is too unstable to be retained, or (ii) Y132A promotes a quaternary structure change, weakening binding sites some distance from the defect, that limits association to the wildtype.
The balance between capsid stability and lability is critical to biological function (e.g. (36)). A number of mutants accumulate in chronic HBV that appear to subtly enhance (37-40) and in a compensatory manner attenuate assembly (41-43). However, the effect of Y132A in Cp149 is unsubtle. Even in the full length core protein, where assembly is supported by interaction with nucleic acid, evidence suggests that the Y132A mutation prevents assembly in bacterial (13) and in mammalian expression systems (44). Evidence suggests that the Y132A mutation prevents assembly of full-length core protein (in bacterial (13) and mammalian expression systems (44)), even though the capsid is also stabilized by interaction with packaged nucleic acid.
Our results have implications for development of antiviral approaches that target assembly (29) and for control of self-assembling systems in nanotechnology (3). For an antiviral assembly inhibitor that prevents a bound subunit from assembling or where weak interactions readily allow editing of inhibitor-bound subunits, there will need to be greater than stoichiometric concentrations of inhibitor in a cell. In short, preventing assembly altogether is likely to be an unsuccessful strategy. On the other hand, an inhibitor that participates in assembly to poison the reaction or to generate fragile capsids can be effective at substoichiometric ratios. In low NaCl Cp149-Y132A does not stay included in capsids and thus mimics the effects of a protein with a tightly bound inhibitor that altogether prevents assembly. In high NaCl, Cp149-Y132A participates in assembly to generate fragile capsids, but does not poison assembly in a crystallographic sense by blocking further growth. These diverse activities may be ideal for distinct critical roles in regulating formation of self-assembling nanostructures. It is easy to picture roles for labile inhibitors to prevent inopportune assembly or to act as protecting groups to block assembly. Identification of small molecules and mutants that regulate assembly remains an ongoing area of research.
We are indebted to Drs. Steven Stahl and Paul Wingfield for their preliminary studies with assembly deficient Cp149 mutants and to Jennifer M. Johnson for comments and technical assistance. Electron microscopy was performed at the Oklahoma Medical Research Foundation imaging core facility. This work was supported by NIH grants to AZ and by a Mary Horton-American Cancer Society Fellowship to CRB.