Crystal structure of the E2 transactivation domain of human papillomavirus type 11 bound to a protein interaction inhibitor.

Interaction between the E2 protein and E1 helicase of human papillomaviruses (HPVs) is essential for the initiation of viral DNA replication. We recently described a series of small molecules that bind to the N-terminal transactivation domain (TAD) of HPV type 11 E2 and inhibits its interaction with E1 in vitro and in cellular assays. Here we report the crystal structures of both the HPV11 TAD and of a complex between this domain and an inhibitor, at 2.5- and 2.4-A resolution, respectively. The HPV11 TAD structure is very similar to that of the analogous domain of HPV16. Inhibitor binding caused no significant alteration of the protein backbone, but movements of several amino acid side chains at the binding site, in particular those of Tyr-19, His-32, Leu-94, and Glu-100, resulted in the formation of a deep hydrophobic pocket that accommodates the indandione moiety of the inhibitor. Mutational analysis provides functional evidence for specific interactions between Tyr-19 and E1 and between His-32 and the inhibitor. A second inhibitor molecule is also present at the binding pocket. Although evidence is presented that this second molecule makes only weak interactions with the protein and is likely an artifact of crystallization, its presence defines additional regions of the binding pocket that could be exploited to design more potent inhibitors.

Interaction between the E2 protein and E1 helicase of human papillomaviruses (HPVs) is essential for the initiation of viral DNA replication. We recently described a series of small molecules that bind to the N-terminal transactivation domain (TAD) of HPV type 11 E2 and inhibits its interaction with E1 in vitro and in cellular assays. Here we report the crystal structures of both the HPV11 TAD and of a complex between this domain and an inhibitor, at 2.5-and 2.4-Å resolution, respectively. The HPV11 TAD structure is very similar to that of the analogous domain of HPV16. Inhibitor binding caused no significant alteration of the protein backbone, but movements of several amino acid side chains at the binding site, in particular those of Tyr-19, His-32, Leu-94, and Glu-100, resulted in the formation of a deep hydrophobic pocket that accommodates the indandione moiety of the inhibitor. Mutational analysis provides functional evidence for specific interactions between Tyr-19 and E1 and between His-32 and the inhibitor. A second inhibitor molecule is also present at the binding pocket. Although evidence is presented that this second molecule makes only weak interactions with the protein and is likely an artifact of crystallization, its presence defines additional regions of the binding pocket that could be exploited to design more potent inhibitors.
Human papillomaviruses (HPVs) 1 are the etiological agents of malignant and benign lesions of the differentiating squamous or mucosal epithelium, notably of cervical cancer. Approximately 25 HPV types replicate in mucosal tissues of the anogenital tract. HPV16, -18, and -31 are the most prevalent "high-risk" types found in pre-cancerous or malignant lesions of the cervix. HPV6 and -11 are the most common "low-risk" types, which cause benign genital warts (condyloma acuminata), a less serious condition but one of the most common sexually transmitted diseases (1). Currently, no specific antivirals are available for the treatment of HPV infections.
The small circular double-stranded DNA genome of papillomavirus is actively maintained as a multicopy episome in the nucleus of infected epithelial cells. This process is dependent on replication of the viral genome by the viral E1 and E2 proteins, in conjunction with the host DNA replication machinery. E2 is a sequence-specific DNA-binding protein that has a number of functions in the viral lifecycle. In addition to its role in the initiation of viral DNA replication, E2 is involved in regulating the transcription of viral genes (2)(3)(4)(5)(6)(7), and in the segregation of the viral genome during cell division (8,9). As a replication initiation factor, E2 binds with high affinity to specific sites located within the viral origin (ori) to help recruit it to the E1 helicase (10 -13). Formation of a ternary complex between E1, E2, and the origin relies not only on the interaction of E1 and E2 with specific DNA sequences at the origin but is also critically dependent on a direct interaction between these two proteins (14 -18).
The 40-kDa E2 protein exists in solution as a dimer and consists of an N-terminal transactivation domain (TAD) of ϳ200 residues, a C-terminal DNA binding/dimerization domain of ϳ100 residues, and a connecting "hinge" region of ϳ70 amino acids (19). The N-terminal TAD is the region of E2 that binds to E1 (14,15,18).
Crystal structures have been reported for the TAD of HPV16 (amino acids 1-201) (20) and for a truncated TAD of HPV18 (amino acids 66 -215) lacking two N-terminal ␣-helices (21). The HPV16 and -18 TADs share ϳ45% sequence identity (see Fig. 1) and have a very similar L-shaped structure composed of a three-helix bundle, partly missing in the structure of the truncated HPV18 TAD, followed by an antiparallel ␤-sheet.
We recently described a series of small molecules that antagonize the E1-E2 protein-protein interaction by binding reversibly to the E2 TAD (22,23). This class of inhibitors, termed "indandiones" because they feature an indandione system spirofused onto a substituted tetrahydrofuran ring, blocks assembly of the HPV11 E1-E2-ori ternary complex at submicromolar concentrations in vitro and abrogates viral DNA replication in a transient transfection assay. To our knowledge, these are the only small molecule inhibitors of HPV DNA replication with cellular activity.
Here, we report the three-dimensional structure of the HPV11 TAD and that of a TAD-indandione inhibitor complex. The HPV11 TAD structure is very similar to that of the analogous domains of HPV16 and -18. Inhibitor binding does not significantly affect the protein backbone but results in the * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
formation of a deep pocket that makes several well defined interactions with the inhibitor. Residues that have previously been implicated in binding E1 or the indandione inhibitors, including Gln-12 (24), Glu-39 (25)(26)(27), and Glu-100 (23) are in or near this pocket.

EXPERIMENTAL PROCEDURES
Proteins-Wild type (WT) HPV11 E2 TAD with an N-terminal polyhistidine tag (His-TAD, sequence MGHHHHHH followed by E2 amino acids 2-201) was purified from bacteria as described (23). To incorporate selenomethionine, the protein was expressed in Escherichia coli strain B834 (28), a methionine auxotroph, grown in DL30 medium (29) lacking methionine and supplemented with 2 g/ml biotin and thiamin and 50 g/ml seleno-D,L-methionine (Se-Met). Purification was performed as described for the WT protein, except that buffers were sparged with helium and all purification steps were completed within 1 day. Addition of a 4-lysine tail to the C terminus of His-TAD was performed by PCR using a primer encoding four lysines, followed by ligation of the PCR product into the plasmid pET15b (Novagen). The encoded His-Lys-TAD protein was purified as described for His-TAD. Electrospray mass spectrometry was used to verify the molecular weights of these three proteins and to determine that they lacked the N-terminal methionine. Complete incorporation of Se-Met was also verified by amino acid analysis. Purified proteins were Ͼ95% pure by SDS-PAGE. Protein concentrations were determined by absorbance at 280 nm using a molar extinction coefficient of 49,060 M Ϫ1 cm Ϫ1 .
Site-directed mutagenesis of HPV11, HPV6, and HPV31 E2 or of His-Lys-TAD was performed using the QuikChange kit (Stratagene) according to instructions supplied by the manufacturer. E2 proteins were synthesized in vitro by coupled transcription/translation using the TNT reticulocyte lysate system (Promega).
Inhibitors-Inhibitors were synthesized by the same procedures as described for other compounds in the same series (22). Single enantiomers were isolated by chiral chromatography. Radiolabeled inhibitor 3 was synthesized from an enantiomerically pure Boc-protected amine precursor. The acetate moiety, tritiated at the methyl group, was incorporated using [ 3 H]N-acetoxyphthalimide (American Radiolabeled Chemicals Inc.). Specific activity of the product was 9.2 Ci/mmol.
Size-exclusion Chromatography-Size-exclusion chromatography was performed using the SMART System (Pharmacia), on a Superdex-75 (PC 3.2/30) column in buffer composed of 25 mM Tris-HCl (pH 8.0), 500 mM NaCl, 5 mM TCEP, 0.1 mM EDTA at a flow rate of 40 l/m. (All buffers were adjusted at room temperature to the stated pH.) Analytical Ultracentrifugation-Sedimentation velocity and equilibrium experiments were performed at 20°C using an XL-A analytical ultracentrifuge (Beckman-Spinco) with an An-60 Ti rotor, in 1.2-cm charcoal-filled Epon centerpieces (double sector for velocity and sixsector for equilibrium) and quartz windows. TAD proteins were exchanged into a buffer composed of 20 mM Tris (pH 8.0), 500 mM NaCl, 2 mM TCEP, 0.1 mM EDTA. Velocity data were collected at 60,000 rpm at 0.005-cm intervals, with 5 replicate readings per point. Protein concentrations were ϳ16 M. Scans were spaced as closely as the absorbance optics would allow. Equilibrium data were collected successively at 26,000, 30,000, and 38,000 rpm using TAD concentrations of 2-8 M. 5 replicate readings were taken at 0.003-cm intervals. Scans were taken every hour, and at each speed the system was judged to have reached equilibrium when no difference was observed in successive scans, as determined by subtraction using the Origin software (Microcal). Buffer density (1.01922 g/ml) and protein partial specific volumes at 20°C (0.7292 and 0.7311 cm 3 /g for HPV11 His-TAD and HPV11 His-Lys-TAD, respectively) were calculated using the program SEDNTERP (30). Velocity data were analyzed using the program SVEDBERG 6.38 (31), and equilibrium data were analyzed using Win-Nonlin 1.06 (32). For equilibrium experiments, data were truncated at an absorbance of ϳ1.0, which corresponds to a TAD concentration of 17 M. Very similar molecular weight estimates were obtained if more data, up to absorbance 1.2, was retained, but variances were higher and residuals were nonrandom.
Isothermal Titration Calorimetry-Titration of inhibitor 1 into a solution of HPV11 His-Lys-TAD was performed using a VP-isothermal titration calorimetry instrument (MicroCal Inc.) in a buffer composed of 20 mM Tris (pH 7.5), 100 mM NaCl, 0.5 mM TCEP, 0.1 mM EDTA, and 1% (v/v) Me 2 SO, at 20°C, and was analyzed as described previously (23). Data from a blank run, using a solution of compound 1 injected into buffer without His-Lys-TAD, were subtracted from the inhibitorprotein run prior to fitting the data to a one-site binding model.
Crystallization and Data Collection-Final crystals of apo-HPV11 His-TAD containing either Met or Se-Met were obtained at 4°C using micro-seeding in a hanging drop vapor diffusion plate. The precipitant solution was composed of 0.1 M sodium succinate (pH 5.0), 18% PEG 5000 monomethyl ether, and 0.2 M (NH 4 ) 2 SO 4 . Diffraction data on the Se-Met E2 TAD were collected on an ADSC Q4 CCD at the National Synchrotron Light Source beamline X4a (Brookhaven National Laboratory, NY). Four datasets were collected from a single crystal, cooled to 100 K, at four different wavelengths near the selenium absorption edge (0.9790, 0.9794, 0.9743, and 0.9879 Å). The first dataset was selected for structure refinement; data collection statistics for this dataset are listed in Table I.
The complex of His-Lys-TAD with inhibitor 1 was formed by adding 1 l of a 60 mM solution of 1 (BILH 434) in Me 2 SO to 74 l of His-Lys-TAD at 10 mg/ml in the purification buffer, and incubating the mixture on ice for 2-3 h. Crystals were then obtained by vapor diffusion, using a precipitant solution composed of 0.1 M sodium citrate (pH 5.5) and 35% methyl-2,4-pentanediol. Diffraction data were collected on the National Synchrotron Light Source beamline X25 (Brookhaven National Laboratory, NY). Data from a single crystal cooled at 100 K were recorded on a Brandeis B4 detector (Brandeis University) mounted on a -axis goniometer (Enraf-Nonius, The Netherlands). Data collection statistics are listed in Table I. Structure Determination and Refinement-All data were processed with the HKL software Denzo and Scalepack (Table I). Initial phasing of the Se-Met HPV11 His-TAD Multiwavelength Anomalous Dispersion data was done using the software SHARP (33) with selenium sites located by SHELX (34). Model building into initial electron density maps was carried out using the software O (Alwyn Jones, Uppsala University, Sweden) and model refinement was done with the software CNX (Accelrys). At a later stage, model building was also helped by molecular replacement phasing (CNX) using the published HPV16 coordinates (Protein Data Bank code 1DTO (20)). 10% of the data was used as a test set for the R free calculation. The final model includes residues 3-121 and 128 -195 (Table I), but 15 residues were modeled as alanines because of inadequate electron density for their side chains (Lys-6, Asp-9, Glu-17, Gln-48, Glu-66, Lys-107, Lys-120, Val-128, Lys-173, Ser-180, Thr-181, Asn-182, His-183, . The structure of the His-Lys-TAD⅐1 complex was solved by molecular replacement using the refined HPV11 His-TAD model. This final model includes residues 2-196 (all side chains were modeled as the correct residue), two molecules of inhibitor 1, 27 water molecules, one molecule of Me 2 SO, and one partially defined buffer molecule (statistics in Table  I). The refined atomic coordinates and structure factors have been deposited in the Protein Data Bank (accession codes 1R6K and 1R6N for the TAD and TAD:inhibitor 1 structures, respectively).
Computational Studies-The atoms making up the binding pocket were identified by applying the Site Finder application within the MOE software package (Chemical Computing Group Inc., Montreal, Canada) to the crystal structure of the complex after both inhibitors, as well as solvent atoms, were removed. The probe radii were 1.4 and 1.8 Å. The isolated donor acceptor radius was 3.0 Å and the connection distance was 2.0 Å.
Inhibitor-protein binding affinities were evaluated using the empirical scoring functions PMF (35), PLP (36), and LUDI (37). Each inhibitor-protein interaction was evaluated after removal of the other inhibitor as well as all solvent atoms (except one water molecule between His-29 and His-32 that appears to be hydrogen-bonded to both residues). For these calculations, hydrogen atoms were added to all residues and molecules.
E1-E2-Ori Complex Formation Assay-This assay was performed using in vitro translated E2 proteins as described previously (23). Briefly, binding reactions (80 l) consisted of HPV11 E1-containing insect cell nuclear extract (titrated to give ϳ70% of maximal activity with purified WT HPV11 E2), 2 l of in vitro translated E2, and 0.4 ng of ori probe generated by PCR using [ 33 P]dCTP. Negative controls contained E2 protein but no E1 extract. Signals were detected by capturing E1-E2-ori complexes using anti-E1-coated scintillation proximity beads.
Circular Dichroism (CD) Spectroscopy-Far UV CD spectra were acquired between 260 and 187 nm using a J-715 spectropolarimeter (Jasco) in a 0.05-cm cell at room temperature (approximately 23°C). Proteins were diluted to 5 M in a buffer composed of 20 mM sodium phosphate (pH 7.6), 50 mM Na 2 SO 4 , and 0.5 mM TCEP. A total of 5 accumulations were taken for each sample at a speed of 20 nm/m, a resolution of 0.2 nm, a response of 2 s, and a bandwidth of 1.0 nm.
Differential Scanning Calorimetry-Melting temperatures of TAD proteins were determined using a VP-DSC (MicroCal Inc.). TAD proteins were exchanged into a buffer composed of 50 mM Tris (pH 7.6), 150 mM NaCl, and 0.5 mM TCEP, adjusted to 5 M, filtered, and degassed. Temperature was scanned between 10 and 65°C at 1.5°min Ϫ1 . Unfolding under these conditions was irreversible for all TAD proteins, but relative denaturation temperatures were estimated, after buffer subtraction, by fitting C p versus temperature curves to a modified two-state model (38), using the Origin software supplied with the instrument. It is acceptable to use the equations of equilibrium thermodynamics to analyze unfolding events that are in practice irreversible, because irreversibility typically results from secondary, slower events, such as aggregation (39).
Binding of Tritiated 3 to TAD Proteins-Titrations were performed in 96-well nickel chelate Flashplates (PerkinElmer Life Sciences) using 50 nM His-Lys-TAD protein and 4 -500 nM tritiated inhibitor 3, in a buffer composed of 25 mM MOPS (pH 7.0), 100 mM NaCl, 0.5 mM TCEP, 0.0025% (v/v) Tween 20, and 2% Me 2 SO. After a 3-h incubation to allow for equilibration of interactions between the His-Lys-TAD and both the nickel plate and inhibitor 3, signals from bound 3 were detected by scintillation counting using a TopCount NXT (PerkinElmer Life Sciences). Blank titrations of compound 3 into wells lacking His-Lys-TAD were performed in parallel and signals from nonspecific binding of the compound was subtracted prior to data analysis. Signals ranged from 30 to 4000 cpm for negative controls and 120 to 7000 for WT TAD (prior to blank subtraction). Binding data were analyzed by non-linear regression (GraFit 3.0, Erithicus Software Ltd.) using the quadratic equation, where Max denotes the counts/min observed at saturation.
ATPase Assays-ATPase activity of purified recombinant E1 (40) was measured using a scintillation proximity assay as described previously (40,41). The concentration of HPV11 E1 was 3 nM and ATP and magnesium acetate were used at 20 and 500 M, respectively. After subtraction of background measured in reactions lacking E1, inhibition data were fit to a three-parameter logistic using SAS (release 6.12, SAS Institute Inc., Cary, NC). TAD mutants Y19A and E39A inhibited only weakly, such that their IC 50 and I max parameters could only be estimated.

RESULTS
Expression, Purification, and Characterization of the HPV11 E2 TAD-The HPV11 E2 TAD (amino acids 1-201, Fig. 1) was expressed in E. coli as an N-terminal polyhistidine-tagged protein. The purified protein eluted in a sharp peak when analyzed by analytical size-exclusion chromatography ( Fig. 2A) and comparison of its elution time to those of molecular mass standards gave an estimated mass of 24.7 kDa, close to that expected for the monomer based on its sequence (23.8 kDa).
The HPV11 TAD also appeared to be monomeric by analytical ultracentrifugation. In an equilibrium sedimentation experiment, no significant trends in apparent mass were observed at increasing rotor speed or loading concentration (Table II). Such a trend would be observed if the protein formed a mixture of monomers and dimers under experimental conditions (42). Data from a centrifugation velocity experiment were best fit to an estimated average molecular mass of 22.5 kDa for a single species, based on individual parameters of S ϭ 1.960 Ϯ 0.003 Svedberg and D ϭ 8.26 Ϯ 0.03 Frick. Collectively, these results indicate that the HPV11 E2 TAD is fully monomeric in solution over the concentration range studied, up to ϳ20 M.
An even more soluble version of the HPV11 TAD containing 4 lysines at its C terminus was also purified and was also found to be monomeric by size-exclusion chromatography and analytical ultracentrifugation (data not shown). Binding of inhibitor 1 (Fig. 3) to this Lys-tailed TAD was analyzed by isothermal titration calorimetry. Formation of a 1:1 complex was observed with a K d value of 40 nM (Fig. 2C).
Crystal Structure of the Apo-HPV11 E2 TAD-The structure of the HPV11 TAD was determined from Multiwavelength Anomalous Dispersion diffraction data obtained using crystals of the protein labeled with Se-Met (see "Experimental Procedures"). Overall the structure of the HPV11 TAD shown in Fig.  4 is very similar to that previously reported for the analogous domain of HPV16. The all-C␣ superposition of the HPV11 and -16 structures (187 residues) yielded an r.m.s. deviation of 1.54 Å, and all secondary structural elements are conserved. The major differences are a slight twist (Ͻ1°) in the relative orientation of the ␣-helical and ␤-sheet subdomains and to variations in the position of loops, in particular those comprised of residues 140 -143 and 160 -165, which are involved in crystal packing in the HPV16 and HPV11 structures, respectively. The N-terminal (amino acids 3-94) and C-terminal (amino acids 95-195) subdomains can be individually superimposed with r.m.s. deviations of 0.85 and 1.14 Å, respectively. Interestingly, the HPV11 protein crystal lattice was distinct from that reported for HPV16, such that a different monomer-monomer interface was observed in the HPV11 crystal structure.
Crystal Structure of an HPV11 E2 TAD-Inhibitor Complex-We obtained crystals of the more soluble Lys-tailed TAD in complex with compound 1, albeit using different conditions than those used to crystallize the apo-His-TAD. The space group of these crystals was also different from any of the previous TAD structures (Table I). The conformation of the HPV11 TAD within the complex (Fig. 5A) is very similar to that observed for the apoprotein; all backbone C␣ could be super-imposed with an r.m.s. deviation of 0.88 Å (187 residues). The two inhibitor 1 molecules are both in the free carboxylate form (as drawn in Fig. 3), one of two distinct forms that this class of inhibitors adopts in solution (43). The presence in the crystal  35 kDa, 1.88 ml). The apparent TAD molecular weight was estimated from a plot of log (M r ) versus elution volume for these three molecular weight standards. B, equilibrium sedimentation data for HPV11 His-TAD. The data shown (absorbance at 280 nm versus radius (r 2 /2, cm 2 )) were obtained at 30,000 rpm, 20°C. The initial TAD concentration was ϳ4 M. The line corresponds to the fit obtained from data at all runs in Table II  structure of the free carboxylate form confirms a prediction that this form is the one with inhibitory activity (43).
Because we had demonstrated that the TAD and inhibitor 1 form a 1:1 complex in solution, we were surprised to find two molecules of inhibitor bound per protein monomer (Fig. 5A). We believe that the molecule referred to below as inhibitor A is the one that binds tightly in biological assays (see below). In con-trast, inhibitor B appears to interact only weakly with a secondary binding pocket, formed not only by the TAD protein alone, but also in part by one surface of inhibitor A, as well as by residues from an adjacent protein monomer in the crystal lattice. In fact, inhibitor B has approximately equal surface area in contact with each of the two TAD molecules, as well as with inhibitor A (data not shown). The close contact between the two inhibitor molecules may distort the observed interaction between the TAD and the dichlorophenyl ring of inhibitor A (see below). The TAD-TAD interface that incorporates inhibitor B is distinct from that in the apo structure, and in fact does not result in formation of a dimer, but rather an infinite array (along the 41 screw axis of the crystal), which is unlikely to exist in solution except at very high protein and inhibitor concentrations.
To further demonstrate that the region of the TAD near inhibitor A constitutes a better binding pocket than does the one near B, a computational analysis of the protein surface was performed using the Site Finder application within the MOE software package (see "Experimental Procedures"). This program uses an ␣ shape-based algorithm (44) to identify concave regions on the protein surface. Furthermore, it evaluates whether these cavities have an appropriate composition of hydrophobic and hydrophilic atoms, as hydrophilic cavities are likely to be highly solvated and thus interact only weakly with organic molecules in aqueous solution. The TAD surface was evaluated after removal of the two inhibitors, and residues that comprised the best small molecule-binding pocket are highlighted in Fig. 5B. Inhibitor A makes a number of close contacts with these highlighted residues, as described below. Inhibitor B is located on the periphery of this pocket and does not make any well defined hydrogen bonding or Van der Waals interactions. The dichlorophenyl ring of inhibitor B is ϳ4 Å from the nearest pocket atoms; other portions of the molecule are further from the protein surface (indandione moiety) or make closest contact with non-pocket atoms (thiadiazoylphenyl moiety).
In a separate computational study, binding energies for each inhibitor molecule to the TAD (in the absence of the other) were predicted using three different empirical protein-binding scoring functions (Table III). As expected, results significantly favored binding of inhibitor A relative to B, with predicted binding energies 2-fold higher for inhibitor A. Taken together, these different methods suggest that inhibitor A is solely responsible  3. Inhibitors used in this study. Inhibitor 1 (BILH 434), used in calorimetry and crystallization studies, has an IC 50 of 180 nM in the E1-E2-ori complex formation assay. Inhibitor 2 was used to obtain the results in Table IV. Inhibitor 3, carrying a tritiated methyl group (*), was used to obtain the results in Fig. 6A and Table V. The unlabeled analogue of this compound has an IC 50 of 100 nM in the E1-E2-ori complex formation assay and a K d for TAD binding of 40 nM, as determined by isothermal titration calorimetry. All three inhibitors are single enantiomers with the absolute configuration as shown. for inhibition at the low concentrations used in biological assays and that the presence of inhibitor B might be an artifact of crystallization. This suggestion is further supported by the mutational analysis of the inhibitor binding pocket, presented below. Although an artifact, it is possible that the fortuitous presence of this second molecule aided in crystallization of the complex.
Inhibitor A lies within 4.5 Å of 12 residues (marked by asterisks in Fig. 1), and the backbone superposition for these in the apo and complex structures gives an r.m.s. deviation of only 0.28 Å. However, binding of this inhibitor induces significant movement of several amino acid side chains. The indandione moiety of the inhibitor is buried in a pocket that is not present in the apo structure, but results from rotations of the side chains of Tyr-19 (1, 85°; 2, 40°) and His-32 (2, 90°) (Fig. 5C). In the apo structure, the aromatic rings of His-32 and Trp-33 form a T-shaped edge-to-face structure, with the hydroxyl group of Tyr-19 close enough to form a hydrogen bond with the ⑀-nitrogen of His-32. In the complex structure, this interaction is lost and the rings of His-32 and Trp-33 become parallel to each other. Two other side chains, Leu-94 and Glu-100, also adjust their conformations to accommodate the inhibitor (Fig.  5C). Together, these changes in side chain conformation result in the formation of the observed binding pocket (Fig. 6).
The new positions of Tyr-19 and His-32 not only open up the pocket, but also allow the indandione moiety of the inhibitor to become a middle layer of a three-tier -stacking interaction between His-32 and Trp-33. Additional Van der Waals contacts from the Leu-94 side chain result in a highly complementary shape to the deepest part of the pocket. Two other important interactions are made with the backbone nitrogen atoms of Tyr-99 and Glu-100; the distance (2.73 and 2.77 Å) and angle between these atoms and the inhibitor carboxylate oxygens suggests the formation of strong hydrogen bonds (Fig. 5C).
Remaining contacts of inhibitor A with the pocket are less well defined hydrophobic interactions. The dichlorophenyl ring lies in a wider but shallower portion of the pocket without making obvious specific interactions. Although it has been found in previous studies (22) that both chlorine atoms contribute significantly to binding, neither atom is very close to the protein; the closest one being 4 Å from the Tyr-19 side chain. The observed position of the dichlorophenyl ring may be due in part to distortions induced by crystal packing or the presence of the second inhibitor molecule, as discussed above. The thiadiazolylphenyl group lies along the protein surface outside of the pocket. Notably, the phenyl-thiadiazole bond is close to the side chain of Met-101. This finding is consistent with recent photoaffinity labeling experiments 2 in which the carbon atom para to the amide bond of an analogue was found to be covalently linked to the terminal carbon of Met-101 (see Fig. 6).
Mutational Analysis of the Inhibitor Binding Pocket-Several residues in or near the inhibitor binding pocket were mutated to examine the roles of specific side chains in binding the indandione, as well as gain insight into the possible overlap between the E1-and inhibitor-binding sites. Most residues were mutated to Ala, but a few residues were replaced by amino acids found in other HPV types: His-32 was changed to Tyr as found in HPV16 E2, whereas Ile-30, Met-31, and Ile-36 were changed to Val, Leu, and Met, respectively, as found in HPV6 E2. These last three residues are the only ones in the vicinity of inhibitor A that are different between HPV11 and -6 E2. Also included in this study was the amino acid substitution I73A that is located on the opposite face of the protein and was previously shown to specifically affect the transactivation activity of E2 rather than its replication or E1 binding activity (25). Each mutant protein was produced by in vitro translation and tested in our previously described E1-E2-ori complex formation assay using radiolabeled origin DNA (Table IV). The ability of each protein to interact with E1 was scored qualitatively, because we did not control for variations in the level of protein produced in the in vitro transcription-translation reaction and, in any case, could not assess the proper folding or stability of the proteins. Most mutant proteins retained at least some ability to productively interact with E1 at the origin.
We also determined the sensitivity of these mutant proteins to inhibitor 2 (Fig. 3), an analogue of compound 1 that we have described previously (23). A number of substitutions had little if any effect on interaction with E1 or on inhibitor binding. These included V64A and K68A, which interact with inhibitor molecule B in the inhibitor-TAD complex, thereby providing further evidence that binding of this second molecule is not relevant to inhibition.
Substitutions I30V/M31L and I36M, which make the HPV11 TAD more similar to that of HPV6 E2, also had little effect on inhibition. These results suggest that none of these three residues are responsible for the 30-fold weaker affinity of HPV6 E2 for compound 2 (23). In further support that these residues do not play a major role in the type selectivity of the indandione inhibitors, we found that the corresponding mutations (V30I/ L31M and M36I) in HPV6 E2 also had little effect on inhibition (data not shown).
Other substitutions (L15A, Y19A, E39A, and S98A) produced inactive or weakly active proteins. In principle this could indicate either that these side chains are directly involved in interaction with E1, or that the mutant proteins are not properly folded. The low signals obtained with the Y19A and E39A E2 proteins in the E1-E2-ori complex formation assay were insufficient to assess inhibition.
Substitution of Gln-12, His-32, Leu-43, Ser-65, and Ile-73 had a more deleterious effect on compound binding (IC 50 val-  (35) and PLP (36), a lower (more negative) score indicates better affinity, whereas for LUDI (37) a higher score predicts better affinity. For each program, scores are proportional to predicted binding energies. ues) than on interaction with E1, suggesting that these residues may have a specific role in binding the inhibitor. However, three of these substitutions are at the periphery of the inhibitor-binding site (Gln-12, Leu-43, and Ser-65), and Ile-73 is on the opposite face of the protein. A possible rationale for these observations is discussed below. In contrast, His-32 is in direct contact with the biologically relevant inhibitor A. The presence of a tyrosine instead of a histidine at position 32 of the E2 proteins of HPV16, -18, and -31, (Fig. 1) may contribute to the insensitivity of these proteins to the indandione class of inhibitors. Consistent with this proposal, we found in other studies 3 that an HPV31 E2 carrying the double amino acid substitution D31M/Y32H was significantly more sensitive to compound 2 (IC 50 ϭ 6 M) than its WT counterpart (IC 50 ϭ 70 M), when tested in an assay format similar to that used to generate the results reported in Table IV. Finally, a particularly interesting result, which we have noted previously (23), was the improved affinity of E100A E2 for this class of inhibitors.
Biophysical and Functional Characterization of Purified Mutant E2 TADs-Because the roles of Tyr-19 and Glu-39 in the binding of E1 or inhibitors could not be assessed using the E1-E2-ori complex formation assay, we produced the corresponding mutant TAD proteins in E. coli to carry out additional assays and biophysical studies. We also expressed the L15A TAD, for which a moderately reduced affinity for E1 and inhibitor 2 were observed, as well as the I73A and E100A proteins, to better understand the significant effects of these two mutations on inhibitor binding. Analysis of all these proteins by circular dichroism indicated no significant change in secondary structure compared with the WT (data not shown). The thermal stability of each mutant protein was determined by differential scanning calorimetry and found to be comparable with that of the WT TAD, with the L15A protein being reproducibly the least stable (Table V).
Next, we measured the affinity of each TAD protein for compound 3 (Fig. 3) using an equilibrium binding assay. The K d values for binding of 3 to WT and Y19A TADs were similar ( Fig. 7A and Table V). Because the side chain of Tyr-19 undergoes a substantial change in position upon inhibitor binding, the lack of effect of the Y19A mutation on the affinity for inhibitor 3 may actually be due to a combination of two offsetting effects: a loss of Tyr-19-inhibitor interactions partially compensated for by a reduction in the conformational change (entropic cost) necessary to form the inhibitor binding pocket. Further mutagenesis of this residue would be necessary to address this possibility. The K d value for 3 binding to E100A was roughly 10-fold lower, but could not be precisely determined in part because the value of K d is well below the protein concentration used in the experiment (5 versus 50 nM). I73A, L15A, and especially E39A had significantly reduced affinities for the compound. Where a direct comparison can be made, results in Table V are in qualitative agreement with those from the E1-E2-ori complex formation assay shown in Table IV.
We assessed the relative affinity of each of these TAD proteins for E1 by taking advantage of our earlier finding that the ATPase activity of E1 is inhibited by addition of the transactivation domain. TAD binding significantly increases the apparent K m of E1 for ATP. However, we observed that TAD-saturated E1 retained some ATPase activity. We also reported a corresponding weakening of E2 binding in the presence of high ATP concentrations, which might play a role in the dissociation of E2 and assembly of E1 hexamers at the origin (40).
Although not a direct binding assay, monitoring ATPase activity in the presence of E2 (or the TAD) provides a DNAindependent, solution-phase method of assessing the E1-E2 interaction. Titration of WT TAD into E1 ATPase assays inhibited activity to a maximum of 86% (I max ϭ 86%), with an IC 50 of 4 nM (Fig. 7B and Table V). Similar results were obtained for I73A and E100A confirming that these mutations do not affect E1 binding substantially. It should be noted, however, that the IC 50 values reported for I73A, E100A, and WT TAD are close to the E1 concentration used in these assays (3 nM), and hence that these IC 50 values are likely to be an underestimate of the affinity of these proteins for E1.
The remaining three mutant E2 proteins (L15A, Y19A, and E39A) differed from their WT counterpart in two ways. First, the maximal degree of E1-ATPase inhibition obtained at saturating concentrations of TAD was lower for these mutants than for WT E2. This was more prominent for Y19A and E39A E2, 3 P. W. White, R. Coulombe, and M. J. Massariol, unpublished data.

TABLE V
Effect of selected amino acid substitutions on properties of E2 TAD proteins Melting temperatures were determined by differential scanning calorimetry. Values shown are averages of two determinations that differed by less than 0.2°C. Dissociation constants for compound 3 were determined using a tight-binding equation (see data in Fig. 7A). Parameters for inhibition of E1 ATPase activity were determined using a three-parameter logistic (WT, L15A, I73A, and E100A) or, when not possible, by estimation (Y19A and E39A), see data in Fig. 7B which achieved a maximum inhibition of only 32 and 38%, respectively. Although we do not completely understand why these mutant E2 TADs achieve a lower maximum level of inhibition it suggests that the geometry of the TAD-E1 complexes that do form for these two proteins is such that the complexes retain greater residual ATPase activity. The second way in which these three mutant proteins differed from WT E2 is in their lower affinity for E1, as measured by the IC 50 values required to reach their respective maximal level of ATPase inhibition. By this measure, L15A was found to reduce only slightly, if at all, the affinity of E2 for E1, whereas Y19A and E39A had a larger effect. Qualitatively, these results are in good agreement with those obtained in the E1-E2-ori complex formation assay (Table IV) in that the E1-E2 interaction is little affected by I73A and E100A, moderately affected by L15A, but more significantly reduced by Y19A and, as anticipated, by E39A. Notably, all mutant proteins retained some ability to interact with E1.

Structure of HPV11 E2 TAD
The tertiary structure of the HPV11 E2 TAD is nearly identical to that of the analogous domain from HPV16 (20). In particular, the orientations of the two subdomains are similar for both TADs, suggesting that the relative position of these subdomains is a fundamental characteristic of this -fold, and that the overall structure is fairly rigid even though it is far from globular. This idea is further supported by the fact that neither the binding of the indandione inhibitor nor the differing crystal packing arrangements of the three structures significantly altered the observed tertiary structure.
Based on the structure of the HPV16 E2 TAD, Antson et al. (20) proposed the existence of an additional E2 dimerization interface within the TAD, comprised of the second and third ␣-helices, and notably involving a cluster of conserved residues (Arg-37, Ala-69, Ile-73, Gln-76, Leu-77, Glu-80, and Thr-81). Although the area of this interface is relatively large (ϳ2000 Å 2 ), it contains a number of water-mediated interactions and relatively few hydrophobic residue contacts, in contrast to most protein interfaces that are typically less hydrophilic than solvent-exposed surfaces (45,46). Our observation that quite distinct monomer-monomer interactions are found in both of our HPV11 TAD structures, together with the results from chromatography and ultracentrifugation experiments that the HPV11 TAD is fully monomeric at concentrations up to ϳ20 M calls into question the generality of the TAD dimerization interface and thus the likelihood that it plays a fundamental role in the function of E2, at least in the case of HPV11.

Nature of the Inhibitor Binding Site
The inhibitor binding pocket (Fig. 7) is not present in the E2 apo structure but rather appears to be induced upon inhibitor binding. This pocket is comprised of a large "shallow end" approximately defined by residues Gln-12, Glu-39, Lys-68, and Gln-71 and a smaller "deep end" defined by His-32, Trp-33, and Leu-94. The indandione moiety of inhibitor A is complementary to the deep end, and the resulting Van der Waals interactions, together with the hydrogen bonds between the carboxylate and backbone amides, account for the strongly exothermic binding of compounds in this series ( Fig. 2C

Conclusions Based on Studies with Mutant E2 Proteins
The Side Chain of Glu-100 May Impede Inhibitor Binding Mutation of Glu-100 to Ala did not affect substantially the interaction of E2 with E1, but specifically enhanced inhibitor binding as we reported previously (23). The structure of the inhibitor-TAD complex provides an explanation for this gain in affinity, namely that the entropic cost associated with the conformational change in the Glu-100 side chain necessary for compound binding (Fig. 5C) is likely reduced in the case of Ala-100.

Amino Acids Outside the Pocket Can Reduce Inhibitor Binding
A number of substitutions (Q12A, L15A, L43A, S65A, and I73A) clearly affected binding of the inhibitor, although they changed residues that do not directly contact the inhibitor in the structure. Further structural information would be required to fully understand the effects of these five substitutions, but all of these residues are located on helices in the three-helix bundle that defines the N-terminal subdomain of the TAD. The residues that form the deep pocket that binds tightly to the indandione moiety are also located within this three-helix bundle, and given the close fit between the inhibitor and this pocket it is possible that these substitutions subtly affect the packing of the three helices in a way that slightly alters the binding pocket. Perhaps similar subtle conformational changes explain why the indandione class of inhibitors binds to HPV6 E2 10 -30-fold less tightly than to HPV11 E2, because the few amino acid differences in or near the binding pocket did not account for this type of differential activity. The effects of these non-binding site residue differences are conceptually similar to some HIV protease resistance mutations that occur well outside of the active site (47). (solid squares). A, dissociation constants for inhibitor 3 were determined by titrating the inhibitor (4 -500 nM) into 96-well Ni-chelate Flashplates containing 50 nM TAD. Lines were determined by nonlinear regression using a tight-binding equation and are based on triplicate measurements. B, E1 binding was evaluated by the ability of the different proteins to inhibit HPV11 E1 ATPase activity. ATPase activity was measured by a scintillation proximity assay as described previously (40,41) using 3 nM E1 and 20 M ATP. Inhibition was measured for each TAD protein at concentrations from 2 to 1000 nM and points are based on averages of three separate experiments. Lines connecting the points are for ease of visualization only.

Importance of His-32 for Inhibitor Binding
We found that the specificity for inhibitor binding to low-risk rather than high-risk HPV types is in part because of a residue forming the pocket, His-32, which is a tyrosine in high risk types (Fig. 1). Substitution of His-32 for tyrosine in HPV11 E2 abrogated inhibitor binding, although it did not significantly affect interaction with E1. Based on our structure of the inhibitor-TAD complex, this substitution is predicted to prevent formation of the deep end of the inhibitor binding pocket. Furthermore, the converse experiment of inserting His-32 into HPV31 E2 generated a mutant protein that could interact with compound 2, albeit not to the same extent as HPV11 or HPV6 E2.

A Role for Tyrosine 19 in Binding to E1
Tyr-19 is highly conserved among HPV types although it is an Ile in bovine papillomaviruses. Substitution of this residue by Ala resulted in an HPV11 E2 protein that was still capable of binding the inhibitor and whose melting temperature and CD spectrum were very similar to that of wild type E2, indicating that this mutation does not affect the overall structure of the TAD. However, the Y19A mutant protein was defective in binding to E1 both in an E1-E2-ori complex formation assay and in an E1-ATPase inhibition assay. Thus, this mutation appears to specifically affect E1 binding. The proximity on the TAD surface of Tyr-19 to Glu-39, the only other residue specifically implicated in binding to E1, suggests that this surface of E2 might form part of the E1-E2 protein-protein interface. If so, our results would imply that the interaction of Tyr-19 with E1 is energetically important for formation of this interface and that the indandione class of inhibitors antagonizes E1 binding by competing directly for the same or an overlapping binding surface on E2. In fact, the structure of an HPV18 E1-E2 TAD complex has recently been determined, and Tyr-19 does appear to make an important interaction with E1. 4 Analysis of the two crystal structures presented here has provided a detailed understanding of the mechanism by which the indandione inhibitors bind to E2, to disrupt the E1-E2 protein-protein interaction. Comparison of these two structures revealed the induced fit mechanism by which the inhibitor binding pocket is formed and further illustrates that it is very difficult to accurately identify binding pockets or model protein-ligand interactions using only the structure of an unliganded protein (48). The binding pocket includes both a deep, narrow cavity filled by the indandione moiety, as well as a wider shallow region that makes only a few contacts with the inhibitor. The finding that a second inhibitor molecule can bind to this shallow region of the pocket raises the possibility that molecules could be found that would retain specific interactions with the deep cavity but gain additional binding energy by interaction with the larger shallow region. The characterization of a well defined small molecule binding pocket on the E2 TAD represents an important step toward the design of HPV antivirals.