JOURNAL OF VIROLOGY, Oct. 2008, p. 10262–10270 0022-538X/08/$08.00⫹0 doi:10.1128/JVI.01360-08 Copyright © 2008, American Society for Microbiology. All Rights Reserved.
Vol. 82, No. 20
Small-Molecule Effectors of Hepatitis B Virus Capsid Assembly Give Insight into Virus Life Cycle䌤 Christina Bourne,1† Sejin Lee,2 Bollu Venkataiah,2 Angela Lee,1 Brent Korba,3 M. G. Finn,2* and Adam Zlotnick1* Department of Biochemistry and Molecular Biology, University of Oklahoma Health Sciences Center, 975 NE 10th Street, Oklahoma City, Oklahoma 73126-09011; Department of Chemistry and The Skaggs Institute for Chemical Biology, The Scripps Research Institute, 10550 N. Torrey Pines Rd., La Jolla, California 920372; Department of Microbiology and Immunology, Georgetown University Medical Center, 3900 Reservoir Rd. NW, Washington, DC 200073 Received 29 June 2008/Accepted 30 July 2008
The relationship between the physical chemistry and biology of self-assembly is poorly understood, but it will be critical to quantitatively understand infection and for the design of antivirals that target virus genesis. Here we take advantage of heteroaryldihydropyrimidines (HAPs), which affect hepatitis B virus (HBV) assembly, to gain insight and correlate in vitro assembly with HBV replication in culture. Based on a low-resolution crystal structure of a capsid-HAP complex, a closely related series of HAPs were designed and synthesized. These differentially strengthen the association between neighboring capsid proteins, alter the kinetics of assembly, and give rise to aberrant structures incompatible with a functional capsid. The chemical nature of the HAP variants correlated well with the structure of the HAP binding pocket. The thermodynamics and kinetics of in vitro assembly had strong and predictable effects on product morphology. However, only the kinetics of in vitro assembly had a strong correlation with inhibition of HBV replication in HepG2.2.15 cells; there was at best a weak correlation between assembly thermodynamics and replication. The correlation between assembly kinetics and virus suppression implies a competition between successful assembly and misassembly, small molecule induced or otherwise. This is a predictive and testable model for the mechanism of action of assembly effectors. Hepatitis B virus (HBV), which chronically infects approximately 400 million people and contributes to 1 million deaths per year, is an enveloped virus with an icosahedral core. The core is assembled in the cytoplasm from core (capsid) protein, viral pregenomic RNA, viral reverse transcriptase, and a few host proteins. The core plays indispensable roles in viral DNA synthesis from the pregenome and intracellular trafficking (10). The predominant antiviral strategy against HBV is to attack the reverse transcriptase. Unsurprisingly, HBV reverse transcriptase inhibitors lead to drug-resistant mutants in HBV (7, 24, 30) and also human immunodeficiency virus (14). Resistance can have broader consequences because of the extensive gene overlap in HBV, since some reverse transcriptase mutations lead to surface protein that is insensitive to antibodies generated by the HBV vaccine (25). An alternative therapeutic strategy would be targeting capsid assembly (15, 33). Heteraryldihydropyrimidines (HAPs), first identified by scientists at Bayer AG as having anti-HBV activity in a cell culture-based screen, act in a capsid protein-specific manner (8, 11, 19–22).
* Corresponding author. Present address for Adam Zlotnick: Indiana University, Department of Biology, Simon Hall MSB, Room 220D, 212 S. Hawthorne Drive, Bloomington, IN 47405-7003. Phone: (812) 856-1925. Fax: (812) 856-5710. E-mail:
[email protected]. Mailing address for M. G. Finn: The Scripps Research Institute, Department of Chemistry, 10550 N. Torrey Pines Rd., La Jolla, CA 92037. Phone: (858) 784-8845. Fax: (858) 784-8850. E-mail: mgfinn @scripps.edu. † Present address: 250 McElroy Hall, Department of Veterinary Pathobiology, Oklahoma State University, Stillwater, OK 74078-2007. 䌤 Published ahead of print on 6 August 2008.
The mechanism of HAP activity has been studied in vitro using the 149-residue assembly domain of the HBV capsid protein, Cp149, which is always found as a homodimer (11, 21, 22). Assembly of empty capsids is nucleated by slow formation of a trimer of dimers followed by rapid addition of subsequent dimers (34). We previously reported that HAP 1 (Fig. 1), a variant of the original HAP structure identified by the Bayer laboratory (8, 20), increases the rate of assembly by stabilizing an assembly-active form of Cp149; it also stabilizes sixfold arrangements of capsid protein and destabilizes fivefold associations, with the net effect of disfavoring the formation of spherical (icosahedral) particles in favor of tube- and sheetlike structures at high concentrations of HAP 1 (21). We recently solved the structure of an HBV capsid-HAP cocrystal (2G34) (3). The T⫽4 capsid in this structure, comprised of 240 copies of capsid protein arranged as 60 AB and CD homodimers (Fig. 2A), is slightly distorted compared to “apo” structures (2, 3, 6, 28) due to flattening of the quasisixfolds and expansion of the capsid along the fivefold with a concomitant deformation of interprotein contacts. The binding site of HAP 1 in the cocrystal is a hydrophobic groove in the capsid protein located at the interface between a C⬘ subunit (from a C⬘D⬘ dimer [Fig. 2]) and its neighboring D subunit (from a CD dimer). Only the C site is filled to high occupancy, though quasiequivalent sites are also in A, B, and D subunits. The C⬘D interface is predicted to be the weakest in the HBV capsid structure based on calculations of buried surface area and solvation parameters (3, 5, 16). This structural and thermodynamic information led us to suggest that HAP 1 (and its analogues) acts as a molecular wedge
10262
VOL. 82, 2008
TARGETING VIRUS ASSEMBLY
10263
FIG. 1. HAP derivatives used in this study; see Table 1 for the structures of HAPs 3 to 18. Dihydropyrimidine substituent positions are numbered in HAP 1.
to fill a gap and slightly distort the geometries and energetics of selected protein-protein contacts. What makes a good assembly-directed antiviral agent? To examine this question, we needed a series of related compounds to allow “apples-to-apples” comparison. On the basis of a low-resolution X-ray crystal structure with HAP 1 bound (3), we developed a series of HAPs. These were evaluated in vitro for their effects on the thermodynamics (i.e., the extent of assembly) and kinetics of assembly. Selected compounds were then tested for antiviral effect in cultured cells, allowing correlation between systems that are strictly in vitro and mimic HBV in vivo. Thus, in this paper we develop an outline for evaluating small molecules that target virus assembly. MATERIALS AND METHODS Modeling HAP into electron density. Coordinates for HAP 1 were calculated using the software program Sketcher within the CCP4 suite and energy mini-
mized using the PRODRG2 server using the GROMOS87 force field (http: //davapc1.bioch.dundee.ac.uk/programs/prodrg/). This structure was positioned in the capsid-HAP 1 cocrystal (PDB 2G34) (3) density by eye. Sample preparation. Dimeric Cp149 was expressed and purified as previously described (33, 35). Prior to use, frozen aliquots were dialyzed against 0.1 M HEPES (pH 7.5). The concentration was determined using an ε280 of 60,900 M⫺1 cm⫺1 per dimer. HAP molecules were synthesized based on published procedures (19, 20) and characterized by 1H-NMR, 13C-NMR, FT-IR, and ESMS. The compound was intermediate in the synthesis of compounds 5 to 18. Compounds were stored at ⫺20°C as stock solutions of 10 to 25 mM in dimethyl sulfoxide (DMSO), depending on solubility. High-throughput fluorescence quenching assay. HAP compounds at 10 M were preincubated with 5 M C150Bo for 20 min prior to inducing assembly. Assay conditions were chosen to produce approximately 20% assembly in the absence of an assembly effector (Table 1, entry 1). Assessment of Cp149 assembly. Light-scattering experiments were carried out at 37°C as previously described (21, 22). Protein solutions were incubated with the appropriate small molecule for 20 min at 37°C prior to inducing assembly. Final sample concentrations were 2.5 M Cp149 dimer and 150 mM NaCl with ⱕ2% DMSO. Results from three independent experiments were averaged. The
FIG. 2. HBV capsid and the HAP binding site. (A) A schematic of one facet of the icosahedral HBV capsid (chain A is red, chain B is blue, chain C is yellow, and chain D is purple) showing the location of HAP 1 (cyan). (B) HAP 1 binds in a hydrophobic groove at the dimer-dimer interface. (C) X-ray density (Fo ⫺ Fc, red mesh, 60-fold averaged, B-sharpened, and contoured at 10 ) favors one orientation (cyan) for HAP 1 in this site but does not exclude the (magenta) alternative. (D) Substituents on pyrimidyl position 6 (the methyl of HAP 1 directly above Leu140) are freely accessible to the capsid interior in either orientation. Amino acids abutting the HAP molecule are shown as CPK representations.
10264
J. VIROL.
BOURNE ET AL. TABLE 1. Effects of selected HAP compounds on HBV assembly KD (M)c
⌬⌬Gcontact (kcal/mol)d
C150Bo quencha
Cp149 assemblyb
None
17 ⫾ 5
35 ⫾ 6
1
82 ⫾ 4
73 ⫾ 11
⫹
⫺1.14 ⫾ 0.17
3
48 ⫾ 27
4
77 ⫾ 10
63 ⫾ 2
15
⫺0.78 ⫾ 0.05
5
65 ⫾ 20
6
68 ⫾ 22
7
46 ⫾ 3
8
41 ⫾ 4
9
87 ⫾ 1
10
26 ⫾ 15
11
75 ⫾ 15
82 ⫾ 11
⫹
⫺1.64 ⫾ 0.23
⫺2.50 ⫾ 0.08
Large pleiomorphic
12
71 ⫾ 18
85 ⫾ 10
⫹
⫺1.92 ⫾ 0.07
⫺3.69 ⫾ 0.05
Broken pleiomorphic
13
66 ⫾ 2
48 ⫾ 12
19
⫺0.69 ⫾ 0.11
⫺2.41 ⫾ 0.03
Tubes
14
63 ⫾ 8
80 ⫾ 2
⫹
⫺1.55 ⫾ 0.07
⫺2.47 ⫾ 0.12
Tubes/pleiomorphic
15
18 ⫾ 4
Compound
0
Kinetic indexe
EM observationf
NA
35-nm spheres
⫺3.20 ⫾ 0.01
Large pleiomorphic
Small pleiomorphic
⫺1.80 ⫾ 0.11
60 ⫾ 9
15
⫺0.99 ⫾ 0.11
0
Broken pleiomorphic
0
Continued on facing page
VOL. 82, 2008
TARGETING VIRUS ASSEMBLY
10265
TABLE 1—Continued C150Bo quencha
Compound
Cp149 assemblyb
KD (M)c
⌬⌬Gcontact (kcal/mol)d
Kinetic indexe
16
13 ⫾ 10
NA
17
77 ⫾ 5
⫺2.60 ⫾ 0.04
18
75 ⫾ 10
83 ⫾ 8
⫹
⫺1.76 ⫾ 0.19
⫺2.63 ⫾ 0.08
EM observationf
Large tubes
Percent assembly of the Bo-DIPY-modified C150 mutant under conditions that foster 17% ⫾ 5% assembly after 24 h in the absence of HAP derivative. Percentage of Cp149 assembled (10 M total Cp149, 150 mM NaCl, 100 mM HEPES, pH 7.5, 37°C, ⫾10M HAP) after 24 h, determined by SEC. Binding constant of compound binding to Cp149 dimer, estimated from a hyperbolic fit of titration curves (see Fig. 3). A “⫹” symbol indicates binding was too tight to be estimated by this method (KD ⬍⬍ 3 M). d Difference in the pairwise association energy between Cp149 dimers in the presence and absence of added HAP derivative. Without the small molecule, the association energy is ⫺3.87 kcal/mol. e Semiquantitative comparison of assembly kinetics for Cp149 in the presence of HAP derivative, measured by light scattering (2.5 M Cp149, 150 mM NaCl, 50 mM HEPES, pH 7.5; 37°C) (Fig. 5). A value of 0 indicates that no change in light scattering was visible in the first 30 min after assembly was induced. NA, no assembly under these conditions, even after a 24-h incubation. f Dominant form of Cp149 in reactions of 10 M Cp149 and 30 M HAP compound (37°C, 150 mM NaCl, 24 h), observed by negative-stain electron microscopy (EM). Broken pleiomorphic structures have many exposed edges. Large pleiomorphic structures appear as large irregular envelopes (see Fig. 6). a b c
90° light-scattering signal cannot be rigorously analyzed because of the heterogeneity of assembly products. Thus, we adopted a kinetic index based on the steepest slope of the normalized light scattering in the first 30 min after initiating assembly: kinetic index ⫽ ⫺log (slope/[HAP]), where light scattering is in arbitrary units and the HAP concentration is micromolar. Size exclusion chromatography was carried out as previously described (21, 22) utilizing a 22-ml Superpose 6 column equilibrated with 0.1 M HEPES–0.15 M NaCl (pH 7.5). Prior to inducing assembly, samples were incubated with the small molecule for 30 min at 37°C. Assembly was initiated by addition of NaCl to 150 mM, resulting in a final protein concentration of 10 M with 2% DMSO. Samples were given at least 24 h at 37°C to equilibrate. Aliquots for transmission electron microscopy were adsorbed to glow discharged grids and stained with 2% uranyl acetate. Activity in cultured cells. Assays for antiviral activity were conducted as previously described (12, 13). In general, secreted virus was assayed by testing culture medium for HBV DNA, intracellular virus was assayed by quantitative Southern blotting for the DNA replicative intermediate, and cell viability was tested by uptake of neutral red (13). These assays are outlined elsewhere (http: //niaid-aacf.org/protocols/HBV.htm) and are described briefly here as follows. Confluent cultures of HepG2.2.15 cells were maintained on 96-well flat-bottomed tissue culture plates; confluence in this culture system is required for active, high levels of HBV replication equivalent to that observed in chronically infected individuals. Cultures were treated with nine consecutive daily doses of the test compounds. HBV DNA levels were assessed by quantitative blot hybridization 24 h after the last treatment. Cytotoxicity was assessed by uptake of neutral red dye 24 h following the last treatment. Activity against lamivudineresistant (12, 13) and adefovir dipivoxil-resistant (1) HBV mutants was tested in a 5-day assay using a transient-transfection method as previously described (12). Antiviral activity against the drug-resistant panel was determined by quantitative Southern blot hybridization of intracellular HBV DNA replication intermediates.
RESULTS Modeling HAP 1 into low-resolution electron density. HAP 1 was previously cocrystallized with preformed T⫽4 capsids of Cp149 (3). In an (Fo ⫺ Fc) difference map, where the Fc calculation was based on protein and solvent, a single mass of density observed at the interface between the C and D subunits
was attributed to HAP 1 (Fig. 2). However, assigning the orientation of HAP 1 requires some conjecture, since the 5-Å resolution map provides a well-defined molecular envelope for protein but no details of most side chains or small molecules (3). The chiral, crescent shape of HAP 1 allows it to fit into its assigned density in two antiparallel orientations. One of the two orientations was judged superior because of a better fit of the halogenated phenyl ring attached to C-4 (the chiral center; see Fig. 1 for atom numbering) and of the methyl ester at C-5 of the dihydropyrimidine ring (Fig. 2C). In the opposite orientation, the halogenated phenyl ring extends out of the observed density and makes some collisions with protein density. However, including HAP 1 in the refinement of the HBV structure did not lead to structurally or statistically significant changes. While we cannot be certain of the details of the proposed model using the X-ray diffraction data alone, the density does lead to an important testable hypothesis: regardless of which of the two orientations is used, position 6 of the dihydropyrimidine ring (see HAP 1 in Table 1) faces a tunnel between the C and D subunits that opens to the interior of the capsid and thus should tolerate or even favor binding of compounds with substituents at that position. Synthesis of HAP derivatives. Derivatives of HAP 1 were synthesized by the convergent method reported by Bayer investigators (19, 20) (data not shown). For reasons of synthetic convenience and observed activity, we focused on amine substitutions of the chloromethyl intermediate HAP 2 (Fig. 1) for the preparation of most of the new structures. In addition to these variations, structural analogues with changes at each of the other available positions around the dihydropyrimidine core were also prepared. We found some tolerance of binding to the nature of the ester group at C-5, little tolerance for changes in the 2-chloro-4-fluoroaryl moiety at C-4, insensitivity
10266
BOURNE ET AL.
to the position of the pyridyl nitrogen atom at C-2, and great sensitivity to the installation of larger aryl groups at C-2. Each of these findings is consistent with previous observations reported in the patent literature from the Bayer investigators (19, 20). Based on the length of the appendage at position 6 (in an extended conformation), derivatives with variations of R1 can be grouped into three loosely defined classes as shown in Table 1: “short” compounds bear R1 substituents extending up to 4.5 Å from the original methyl group, “medium” extends from 4.5 to 8.5 Å, and “long” extends beyond 8.5 Å. All of the medium and long groups discussed here bear an amine nitrogen  to the dihydropyrimidine ring. High-throughput assay of HAP derivatives. The HAP compounds shown in Table 1 were initially tested using a fluorescence assay developed for high-throughput screening of assembly effectors (22, 35). A modified HBV capsid protein, C150Bo, which carries a C-terminal Bo-DIPY fluorophore, is brightly fluorescent as a dimer. When C150Bo assembles, the fluorophores from several subunits cluster together and selfquench, resulting in a maximum 95% loss in fluorescence intensity. HAP compounds were preincubated with a twofold molar excess of C150Bo for 20 min prior to inducing assembly. Assay conditions were chosen to produce approximately 20% assembly at equilibrium in the absence of added compound. Most HAP derivatives bearing changes at position 6 enhanced assembly, as did the original molecule HAP 1 (R1 ⫽ methyl [Table 1, entry 2]). Indeed, in the fluorescence quenching assay, HAP 1 was among the most potent compounds tested. The range and pattern of responses correlated in interesting ways with features of size and polarity in the R1 substituent. HAP 3, bearing an isopropyl group instead of methyl, was less potent in promoting dimer association, whereas a phenyl substituent in HAP 4 was highly active. The installation of a propargylamine group in the  position to C-6 was well tolerated in HAP 5, which prompted us to explore a variety of amine derivatives. Of 10 such “medium”-length structures, HAPs 5 to 14, only HAP 7 and HAP 8, bearing adamantyl groups, and HAP 10, with a quaternary ammonium center, were significantly diminished in potency relative to the others. Such differences illuminate the nature of the HAP binding site, which apparently disfavors a permanent positive charge at the  position (compare HAP 10 to its des-methyl analogue HAP 9, the latter being the most effective assembly enhancer in the fluorescence assay). This suggests that the protein environment around the  position is highly hydrophobic, consistent with structural predictions, such that secondary and tertiary aliphatic amines are unprotonated when bound. The opposite trend was observed in the “long” derivatives and the most remote substituents of the “medium” derivatives. Thus, the morpholinyl and piperizinyl compounds HAP 12 and HAP 13 were quite effective, and the propargyl-substituted analogue HAP 14 was only marginally less so. However, substitution at the distal nitrogen with large hydrophobic groups, such as cyclohexylmethyl or p-bromobenzyl of HAP 15 and HAP 16, gave very poor assembly enhancement activity, while the use of relatively polar m-pyridylmethylpiperazine in HAP 17 or 4-bis (ethylnyl)carbinol piperidine in HAP 18 was much better. These observations are consistent with the prediction that
J. VIROL.
FIG. 3. Examples of Cp149 assembly reactions (10 M protein, 150 mM NaCl, 37°C, 24 h) titrated with HAP derivatives. The percent dimer was determined by SEC. HAP 18 (squares) is an example of tight binding to protein and a strong effect on protein-protein association energy. HAP 7 (triangles) binds relatively weakly to Cp149 but elicits a strong protein-protein interaction. HAP 13 (circles) binds relatively poorly to protein and has a weak effect on protein-protein association. The number designation on each curve refers to the HAP used.
“short” and to a lesser extent “medium” substituents remain in the hydrophobic environment of the HAP pocket. In contrast, the ends of the “long” substituents are predicted to emerge into the solvent space of the capsid interior and would therefore favor more polar groups. Many other factors can also contribute, of course, including specific protein contacts, differences in solvation of the unbound compounds, and conformational changes brought about by substitution at dihydropyrimidine position 6. HAPs enhance assembly thermodynamics. To gain deeper insights into potential structure-activity relationships, several compounds (weighted toward the more effective end of the series) were investigated in greater detail. These experiments were performed with Cp149, which has a wild-type adyw strain sequence and somewhat weaker assembly activity than the nonphysiological C150Bo (data not shown) (35). Assembly was assayed by size exclusion chromatography (SEC). In the absence of HAP, after 24 h at 37°C, 10 M Cp149 dimer assembly reactions reached an equilibrium of 6.5 ⫾ 0.6 M free dimer and 3.5 ⫾ 0.6 M capsid (in terms of assembled dimer), with undetectable concentrations of intermediates (concentrations in terms of total protein) (Table 1). The dimer concentration under these conditions defines a pseudocritical value that can be directly related to the average association energy of an individual protein-protein contact (5, 31, 32, 34). The selected HAPs induced a dose-dependent decrease in the free dimer concentration (Fig. 3). The SEC peak for the assembly product increased in size and, in most cases, shifted toward the void volume, indicating the formation of large noncapsid polymers (data not shown). The equilibrium dissociation constant (KD,HAP) values for association of HAPs 4, 7, and 13 with Cp149 were 15 to 20 M, estimated by fitting the SEC titration data to hyperbolic binding curves (Table 1; Fig. 3). For the other derivatives, binding was too tight compared to the protein concentrations required for the assembly assay, preventing calculation of a discrete KD value (Table 1; also data not shown). For each of the HAP derivatives chosen for in-depth study, at 10 M HAP, at least 15% of the protein (of 10 M) existed as a free dimer. Under these conditions, we calculated the
VOL. 82, 2008
TARGETING VIRUS ASSEMBLY
FIG. 4. Graphical representation of data for selected HAP compounds (see Table 1), comparing the extent of capsid protein association using two independent methods. Compounds are identified together with the morphology of the dominant polymer (red) and a quantitative description of assembly kinetics (kinetic index; blue).
effect of each compound on the pairwise Cp149-Cp149 association energy. For these large polymers, it was more appropriate to treat the free dimer concentration as a critical concentration (Kcrit). Kcrit is operationally defined as the constant concentration of free dimer in equilibrium with assembled dimers. This Kcrit value allows an estimation of the average association energy per contact that takes into account the multivalence and symmetry of the dimer interactions: ⬍⌬Gcontact⬎ ⫽ [⫺RT In(Kcrit/2)]/2 ⬍⌬Gcontact⬎ ⫽ (1 ⫺ bound)⌬Gcontact,no HAP ⫹ bound⌬Gcontact, ⫹HAP The value ⬍⌬Gcontact⬎ is a weighted average of HAP-associated contacts (bound interacting with ⌬Gcontact,⫹HAP) and the drug-free fraction (1 ⫺ bound), interacting with the ⌬Gcontact,no HAP value of ⫺3.87 kcal/mol (Table 1). The fraction bound is calculated assuming tight binding of drug or from KD,HAP, as appropriate. For this calculation we assume that only one isomer of the racemic mixture of each compound is active and that each dimer has two HAP binding sites (8). A reasonable correlation was observed between results from the fluorescence assay and the SEC assay and other assembly properties, in spite of the differences in the protein construct and assembly conditions (Fig. 4). The most effective compounds—HAP 11, HAP 12, and HAP 18—strengthened the dimer-dimer association energy by a remarkable 1.6 to 1.9 kcal/mol (Table 1). This corresponds to a 500-fold change in the (pseudo)critical concentration from about 7 M for capsids formed in the absence of HAP to 14 nM for the noncapsid polymer formed in the presence of excess HAP 12. The two HAPs promoting the weakest protein association energies, HAP 4 and HAP 13, both led to the appearance of protein oligomers in SEC that were smaller than the capsid. In the case of HAP 4, polydisperse intermediates were observed at low compound concentrations while large noncapsid polymers predominated at higher concentrations. For HAP 13, the observed decrease in dimer was accompanied by the appearance of an intermediate with an estimated molecular mass of about 100 kDa under the same conditions that led to the tube-like polymers observed in micrographs (see below). Since both HAP 4
10267
FIG. 5. Kinetics of Cp149 assembly (2.5 M, 50 mM HEPES, pH 7.5; 37°C) in the presence of the indicated HAP compounds, analyzed by observing light scattering. Assembly was induced by the addition of NaCl to a 150 mM concentration. In the absence of a HAP, 2.5 M Cp149 does not assemble under these conditions. Traces are an average of three experiments. Different concentrations of HAP compounds were used to keep the experiments on the same scale. KI is the kinetic index (Table 1), as defined in Materials and Methods.
and HAP 13 led to large structures observable in micrographs, it is likely that SEC causes shearing of fragile complexes. We evaluated the effect of HAPs on the speed of Cp149 assembly by light scattering under conditions in which Cp149 alone showed no assembly (Fig. 5). Light scattering is a function of the size and shape of the solute, which for HAP reactions is heterogeneous (Fig. 6). Most of the HAP derivatives showed a strong, approximately linear dependence of the rate of change in light scattering on the HAP concentration (see Fig. 2 in the work of Stray et al. [21]), which suggests weak binding of these compounds to the Cp149 dimer. Thus, we established a kinetic index as the negative log of the maximum observed rate of light-scattering change divided by the effector concentration (see Materials and Methods). As expected for multistep reactions (9), there is a correlation between the kinetic index and the effect of a given HAP on association energy, although the trend is not entirely monotonic. Examination of Cp149-HAP mixtures by transmission electron microscopy revealed different structural outcomes for the HAP-driven assembly products (Fig. 6), with general correlation to the observed association energies and assembly kinetics (Table 1; Fig. 4). Heterogeneous but structurally regular assembly products (capsids and small tubes; HAP 4, HAP 14, and no added HAP) were formed with compounds that led to relatively few nuclei (slower kinetics) and allowed thermodynamic editing of misincorporated subunits (i.e., weak ⌬Gcontact). Native viruses depend on these weak interactions to form regular structures (31, 32). When assembly was both fast and strong, irregular particles were formed with numerous exposed edges, the best example being provided by HAP 12. HAP 7 appears to be an exception, since it makes many broken structures, though in its presence assembly is both very slow and moderately weak. We attribute this apparent contradiction to the bulky nature of its adamantyl substituent, which may exert an atypically great distortion of subunit-subunit contact geometry. Compound HAP 18 is also unique in its effect, producing massive 100-nm-diameter tubes with moderately fast kinetics and strong dimer-dimer association. Its dialkynylcarbinol is unusual among the substituents installed thus far
10268
BOURNE ET AL.
J. VIROL.
FIG. 6. Negative-stain electron micrographs of the assembly products of HBV Cp149 (10 M) induced by 30 M HAP derivative and fixed after 24 h at 37°C. The micrographs show assembly induced by HAP 13 (A), HAP 11 (B), HAP 12 (C), HAP 4 (D), HAP 14 (E), HAP 7 (F) HAP 18, (G), or HAP 18 (H) at a lower magnification. Samples were preincubated with the HAP derivative for 30 to 60 min prior to inducing assembly by the addition of NaCl to a 150 mM concentration. Typical 35-nm-diameter HBV capsids are visible in the presence of HAP 11, HAP 4, and HAP 14. A 100-nm scale bar in panel A applies to panels A to G. The panel H scale bar represents 500 nm.
and may play a unique role in inducing a distinct intersubunit geometry. It therefore appears that different aspects of HAP behavior can contribute to the overall structural outcome. HAPs in cultured cells. To determine how in vitro assembly correlated with antiviral activity, selected compounds were assayed in HepG2.2.15 cells, which constitutively express HBV. We observed a strong correlation between assembly kinetics and the concentration of the compound required to reduce the cytoplasmic concentration of HBV DNA, the replicative intermediate, by 50% (EC50) (Fig. 7). In contrast, a much weaker correlation was seen between the EC50 and the effect of HAPs on association energy. The same trends were observed for the effect of HAPs on the concentration of secreted virions (Table 2; also data not shown). Thus, the HAPs with fastest kinetics, HAP 12 and HAP 1, had the strongest effect on assembly, even though they had very different effects on ⌬⌬G. Conversely, the HAP compound giving rise to the slowest assembly process, 7,
had no detectable effect on HBV production in spite of the fact that its effect on association energy was nearly the same as that of compound HAP 1. Compounds HAP 1 and HAP 12, the two molecules with the highest degree of efficacy in HepG2.2.15 cells, were examined for activity against a panel of drug-resistant virus strains in culture (Table 3). The compounds were equally effective in reducing the replication of strains resistant to the reverse transcriptase inhibitors lamivudine and adefovir dipivoxil, as well as wild-type virus. This broad-spectrum activity is expected given the difference in the proposed mechanism of action for the HAP compounds: the assembly of capsid should be independent of reverse transcriptase activity or mutations in the transcriptase machinery that do not affect the physical interactions of reverse transcriptase with the coat protein. DISCUSSION The studies described here provide new insights into the mechanism by which HAP compounds act in terms of the binding site and the critical features correlating physical chemistry and biological activity. Guided by a low-resolution X-ray crystal structure, we changed substituents at position 6 of the dihydropyrimidine ring to give derivatives that led to different
TABLE 2. Suppression of HBV in HepG2.2.15 by HAPs EC50 (M) for suppression ofa: HAP
FIG. 7. Correlation of HAP-induced assembly kinetics with suppression of HBV in HepG2.2.15 cells. EC50s of HAP compounds were determined by Southern blotting. Data are from Tables 1 and 2. Standard deviations are shown for kinetic index, ⌬⌬G (kcal/mol), and EC50. The error bars appear very small in part because of the log scale of the graph.
1 4 7 11 12 13 18 a
Secreted HBV
Cytoplasmic DNA
0.13 ⫾ 0.02 1.90 ⫾ 0.30 ⬎⬎10 0.23 ⫾ 0.02 0.012 ⫾ 0.002 6.10 ⫾ 0.50 1.10 ⫾ 0.20
0.36 ⫾ 0.05 9.10 ⫾ 0.05 ND 2.00 ⫾ 0.30 0.05 ⫾ 0.01 16.0 ⫾ 2.3 1.70 ⫾ 0.20
Results are means ⫾ standard deviations. ND, not determined.
VOL. 82, 2008
TARGETING VIRUS ASSEMBLY
10269
TABLE 3. Activities of compounds HAP 1 and HAP 12 against drug-resistant strains of HBV EC50 (M) for suppression of secreted virus Mutant
HAP or inhibitor Wild type
HAP 1 HAP 12 Lamivudine Adefovir dipivoxil
0.24 ⫾ 0.03 0.07 ⫾ 0.01 0.2 ⫾ 0.1 1.5 ⫾ 0.3
M204V
M204I
L180M
L180M M204V
N236T
0.27 ⫾ 0.04 0.08 ⫾ 0.01 ⬎100 1.8 ⫾ 0.3
0.31 ⫾ 0.04 0.09 ⫾ 0.01 ⬎100 1.3 ⫾ 0.3
0.29 ⫾ 0.04 0.07 ⫾ 0.01 5.2 ⫾ 0.6 1.9 ⫾ 0.3
0.33 ⫾ 0.04 0.08 ⫾ 0.01 ⬎100 1.3 ⫾ 0.3
0.30 ⫾ 0.04 0.08 ⫾ 0.01 0.2 ⫾ 0.1 8.1 ⫾ 1.0
association energies, kinetics, and structures of the resulting in vitro assembly products. The results are consistent with a model in which position 6 faces a gap on the interior of the capsid at the edge of the contact between two HBV subunits. The gap is surrounded by hydrophobic amino acid residues, favoring nonpolar or unprotonated amine groups near the dihydropyrimidine core and polar groups farther away to interact with the solvent. The rate of assembly in vitro, most likely the nucleation of assembly (23, 34, 36), correlates strongly with antiviral activity. In comparison, it is interesting how both kinetic and capsid-stabilizing effects are critical to the “WIN” compounds which inhibit uncoating of picornaviruses (18, 26, 27). It has been reported that treatment of HBV with HAP compounds in vivo results in proteasomal clearance of the capsid protein(8); in vitro, HAPs accelerate and can misdirect assembly (21). Normal capsid assembly is characterized by a slow nucleation rate and weak pairwise association energies between capsid protein dimers. Both factors are necessary to ensure the maximum yield of capsid with minimal kinetic traps, which would be a waste of protein (from the virus’s perspective) (31, 34). When a nucleus does form, the much faster elongation reactions and multivalent interactions between capsid proteins quickly drive assembly to completion, and nascent capsids are unstable until complete (32, 34). Because individual interactions are weak, errors in assembly are thermodynamically edited to yield the most stable particles. When HAP compounds bind to the protein interface and increase the intersubunit association energy and/or nucleation rate, as well as distorting the geometry of interactions, thermodynamic editing is prevented by stronger association and by entrapping of misincorporated subunits with subsequent additions. The natural assembly pathway is thereby altered to the detriment of normal particles. The data presented here provide the first quantitative demonstration of small-molecule-driven changes in this type of protein-protein interaction. Furthermore, based on our observations, we suggest that it is the rate of initiating assembly incorrectly, or correctly, that is critical for determining the ultimate fate of a nascent virus. The series HAPs 11 to 13 comprises a particularly good example of how small differences in small-molecule structure can propagate to give dramatic effects on protein self-assembly. These compounds differ only at position 4 of the piperidinylmethyl group attached at C-6: CH2 (HAP 11, ⌬Gcontact ⫽ ⫺5.51 kcal/mol, moderate assembly speed), O (HAP 12, ⫺5.79 kcal/mol, very fast assembly), and NH (HAP 13, ⫺4.56 kcal/ mol, slow assembly). We find no evidence of significant differences in conformation of the HAP molecules themselves in
their nuclear magnetic resonance spectra and predicted structures by molecular mechanics calculations. It therefore seems likely that this remote position of the HAP structure interacts in crucial ways with the Cp149 protein as it assembles; indeed, we predict it to lie near the boundary between the intersubunit hydrophobic binding pocket and the solvent-exposed interior surface of the capsid. The ability of HAP compounds to modulate both thermodynamic and kinetic aspects of protein association is important to their in vitro function. For example, HAP 12 strengthens the pairwise protein-protein association energy by almost 2 kcal/ mol. Stronger association energy and faster kinetics can lead to many nucleating centers, where bound subunits are likely to stay bound and get trapped by subsequent binding events, eliminating thermodynamic editing. Consequently, HAP 12 leads to an array of irregular noncapsid polymers. Consistent with this scenario, HAP 13, with slower kinetics and a weaker association energy, gives rise to more-uniform structures which are rod-like rather than being capsids because of the induced change in the angle of subunit-subunit orientation that is apparently common to all effective HAP compounds. Understanding the effect of HAPs on assembly in the test tube provides new insights into assembly in an infected cell. Making the assumption that differences in intracellular availability are small between different HAP compounds, it is kinetics, not the degree of structural aberration nor the change in ⌬⌬G, that correlates best with biological activity. Capsid assembly in the HBV life cycle is believed to be nucleated by a complex of viral mRNA and reverse transcriptase (10, 17). In most polymerization reactions, once a reaction has been nucleated, subsequent reactions are very fast (9, 29). Our results imply that this is the case with HBV as well and that the trick to inhibiting virus replication is to nucleate misassembly faster than virion assembly. That at least some elements of in vivo assembly are regulated by kinetics rather than thermodynamics may be the reason that small molecules, such as HAPs, exerting modest forces can have profound effects on a highly evolved self-assembling system such as the HBV particle. We believe that assembly-directed small molecules may well represent a general antiviral strategy (15, 33) in addition to being an alternative when resistance occurs to agents that operate by alternative mechanisms (4). ACKNOWLEDGMENTS This work was supported by the NIH (grant AI067417) and an American Cancer Society Mary Horton Postdoctoral Fellowship (PF05-237-01-GMC) to C.B. B.K. is supported by NIAID contract NO1AI-30046.
10270
BOURNE ET AL.
We are indebted to the NIAID Antiviral Screening Program for testing the HAP compounds in cultured cells. REFERENCES 1. Angus, P., R. Vaughan, S. Xiong, H. Yang, W. Delaney, C. Gibbs, C. Brosgart, D. Colledge, R. Edwards, A. Ayres, A. Bartholomeusz, and S. Locarnini. 2003. Resistance to adefovir dipivoxil therapy associated with the selection of a novel mutation in the HBV polymerase. Gastroenterology 125:292–297. 2. Bottcher, B., S. A. Wynne, and R. A. Crowther. 1997. Determination of the fold of the core protein of hepatitis B virus by electron cryomicroscopy. Nature 386:88–91. 3. Bourne, C., M. G. Finn, and A. Zlotnick. 2006. Global structural changes in hepatitis B capsids induced by the assembly effector HAP1. J. Virol. 80: 11055–11061. 4. Brazil, M. 2005. HIV-1. Viral assembly inhibitors on the horizon. Nat. Rev. Drug Discov. 4:716–717. 5. Ceres, P., and A. Zlotnick. 2002. Weak protein-protein interactions are sufficient to drive assembly of hepatitis B virus capsids. Biochemistry 41: 11525–11531. 6. Conway, J. F., N. Cheng, A. Zlotnick, P. T. Wingfield, S. J. Stahl, and A. C. Steven. 1997. Visualization of a 4-helix bundle in the hepatitis B virus capsid by cryo-electron microscopy. Nature 386:91–94. 7. Deres, K., and H. Rubsamen-Waigmann. 1999. Development of resistance and perspectives for future therapies against hepatitis B infections: lessons to be learned from HIV. Infection 27(Suppl. 2):S45–S51. 8. Deres, K., C. H. Schroder, A. Paessens, S. Goldmann, H. J. Hacker, O. Weber, T. Kramer, U. Niewohner, U. Pleiss, J. Stoltefuss, E. Graef, D. Koletzki, R. N. Masantschek, A. Reimann, R. Jaeger, R. Gross, B. Beckermann, K. H. Schlemmer, D. Haebich, and H. Rubsamen-Waigmann. 2003. Inhibition of hepatitis B virus replication by drug-induced depletion of nucleocapsids. Science 299:893–896. 9. Endres, D., and A. Zlotnick. 2002. Model-based analysis of assembly kinetics for virus capsids or other spherical polymers. Biophys. J. 83:1217–1230. 10. Ganem, D., and R. J. Schneider. 2001. Hepadnaviridae: the viruses and their replication, 4th ed. Lippincott Williams & Wilkins, Philadelphia, PA. 11. Hacker, H. J., K. Deres, M. Mildenberger, and C. H. Schroder. 2003. Antivirals interacting with hepatitis B virus core protein and core mutations may misdirect capsid assembly in a similar fashion. Biochem. Pharmacol. 66: 2273–2279. 12. Iyer, R. P., Y. Jin, A. Roland, J. D. Morrey, S. Mounir, and B. Korba. 2004. Phosphorothioate di- and trinucleotides as a novel class of anti-hepatitis B virus agents. Antimicrob. Agents Chemother. 48:2199–2205. 13. Korba, B. E., and J. L. Gerin. 1992. Use of a standardized cell culture assay to assess activities of nucleoside analogs against hepatitis B virus replication. Antivir. Res. 19:55–70. 14. Lewis-Hall, F. C. 2007. Re: Important information regarding BARACLUDE (entecavir) in patients co-infected with HIV and HBV. Food and Drug Administration MedWatch Program. Food and Drug Administration, Washington, DC. http://www.fda.gov/MedWatch/safety/2007/Baraclude_DHCP _02-2007.pdf. 15. Prevelige, P. E. J. 1998. Inhibiting virus-capsid assembly by altering the polymerisation pathway. Trends Biotechnol. 16:61–65. 16. Reddy, V. S., P. Natarajan, B. Okerberg, K. Li, K. V. Damodaran, R. T. Morton, C. L. Brooks III, and J. E. Johnson. 2001. Virus Particle Explorer (VIPER), a website for virus capsid structures and their computational analyses. J. Virol. 75:11943–11947. 17. Seeger, C., and W. S. Mason. 2000. Hepatitis B virus biology. Microbiol. Mol. Biol. Rev. 64:51–68.
J. VIROL. 18. Smith, T. J., M. J. Kremer, M. Luo, G. Vriend, E. Arnold, G. Kamer, M. G. Rossmann, M. A. McKinlay, G. D. Diana, and M. J. Otto. 1986. The site of attachment in human rhinovirus 14 for antiviral agents that inhibit uncoating. Science 233:1286–1293. 19. Stoltefuss, J., S. Goldmann, T. Kra ¨mer, K.-H. Schlemmer, U. Niewo ¨hner, A. Paessens, S. Lottmann, K. Deres, and O. Weber. October 1999. New dihydropyrimidine derivatives and their corresponding mesomers useful as antiviral agents. WO patent 19999954326. 20. Stolting, J., J. Stoltefuss, S. Goldmann, T. Kramer, K.-H. Schlemmer, U. Niewohner, A. Paessens, E. Graef, S. Lottmann, K. Deres, and O. Weber. May 2000. Preparation of 4-aryl-2-pyridinyl-1,4-dihydropyrimidine-5-carboxylates for treatment of hepatitis B infection. WO patent; 2000 058302. 21. Stray, S. J., C. R. Bourne, S. Punna, W. G. Lewis, M. G. Finn, and A. Zlotnick. 2005. A heteroaryldihydropyrimidine activates and can misdirect hepatitis B virus capsid assembly. Proc. Natl. Acad. Sci. USA 102:8138–8143. 22. Stray, S. J., J. M. Johnson, B. G. Kopek, and A. Zlotnick. 2006. An in vitro fluorescence screen to identify antivirals that disrupt hepatitis B virus capsid assembly. Nat. Biotechnol. 24:358–362. 23. Stray, S. J., and A. Zlotnick. 2006. BAY 41-4109 has multiple effects on hepatitis B virus capsid assembly. J. Mol. Recognit. 19:542–548. 24. Tennant, B. C., B. H. Baldwin, L. A. Graham, M. A. Ascenzi, W. E. Hornbuckle, P. H. Rowland, I. A. Tochkov, A. E. Yeager, H. N. Erb, J. M. Colacino, C. Lopez, J. A. Engelhardt, R. R. Bowsher, F. C. Richardson, W. Lewis, P. J. Cote, B. E. Korba, and J. L. Gerin. 1998. Antiviral activity and toxicity of fialuridine in the woodchuck model of hepatitis B virus infection. Hepatology 28:179–191. 25. Torresi, J., L. Earnest-Silveira, G. Deliyannis, K. Edgtton, H. Zhuang, S. A. Locarnini, J. Fyfe, T. Sozzi, and D. C. Jackson. 2002. Reduced antigenicity of the hepatitis B virus HBsAg protein arising as a consequence of sequence changes in the overlapping polymerase gene that are selected by lamivudine therapy. Virology 293:305–313. 26. Tsang, S. K., P. Danthi, M. Chow, and J. M. Hogle. 2000. Stabilization of poliovirus by capsid-binding antiviral drugs is due to entropic effects. J. Mol. Biol. 296:335–340. 27. Tsang, S. K., B. M. McDermott, V. R. Racaniello, and J. M. Hogle. 2001. Kinetic analysis of the effect of poliovirus receptor on viral uncoating: the receptor as a catalyst. J. Virol. 75:4984–4989. 28. Wynne, S. A., R. A. Crowther, and A. G. Leslie. 1999. The crystal structure of the human hepatitis B virus capsid. Mol. Cell 3:771–780. 29. Zandi, R., P. van der Schoot, D. Reguera, W. Kegel, and H. Reiss. 2006. Classical nucleation theory of virus capsids. Biophys. J. 90:1939–1948. 30. Zhang, W., H. J. Hacker, M. Tokus, T. Bock, and C. H. Schroder. 2003. Patterns of circulating hepatitis B virus serum nucleic acids during lamivudine therapy. J. Med. Virol. 71:24–30. 31. Zlotnick, A. 2003. Are weak protein-protein interactions the general rule in capsid assembly? Virology 315:269–274. 32. Zlotnick, A. 1994. To build a virus capsid. An equilibrium model of the self assembly of polyhedral protein complexes. J. Mol. Biol. 241:59–67. 33. Zlotnick, A., P. Ceres, S. Singh, and J. M. Johnson. 2002. A small molecule inhibits and misdirects assembly of hepatitis B virus capsids. J. Virol. 76: 4848–4854. 34. Zlotnick, A., J. M. Johnson, P. W. Wingfield, S. J. Stahl, and D. Endres. 1999. A theoretical model successfully identifies features of hepatitis B virus capsid assembly. Biochemistry 38:14644–14652. 35. Zlotnick, A., A. Lee, C. R. Bourne, J. M. Johnson, P. L. Domanico, and S. J. Stray. 2007. In vitro screening for molecules that affect virus capsid assembly (and other protein association reactions). Nat. Protoc. 2:490–498. 36. Zlotnick, A., and S. J. Stray. 2003. How does your virus grow? Understanding and interfering with virus assembly. Trends Biotechnol. 21:536–542.
