Structure of the Human Papillomavirus E7 Oncoprotein and Its Mechanism for Inactivation of the Retinoblastoma Tumor Suppressor*

The E7 oncoprotein from human Papillomavirus (HPV) mediates cell transformation in part by binding to the human pRb tumor suppressor protein and E2F transcription factors, resulting in the dissociation of pRb from E2F transcription factors and the premature cell progression into the S-phase of the cell cycle. This activity is mediated by the LXCXE motif and the CR3 zinc binding domain of the E7 protein. In this study we report the x-ray crystal structure of the CR3 region of HPV E7 and a structure-based mutational analysis to investigate its mode of pRb and E2F binding and E2F displacement from pRb. The structure reveals a novel zinc-bound E7-CR3 obligate homodimer that contains two surface patches of sequence conservation. Mutation of residues within these patches reveals that one patch is required for pRb binding, whereas the other is required for E2F binding. We also show that both E7-mediated interactions are required to disrupt pRb·E2F complexes. Based on these studies we present a mechanistic model for how E7 displaces E2F from pRb. Because the CR3 region of HPV E7 has no detectable homology to other human proteins, the structure-function studies presented here provide an avenue for developing small molecule compounds that inhibit HPV-E7-mediated cell transformation.

E7 proteins resemble adenovirus E1A and SV40 large T antigen both in primary sequences and in transactivation and transformation properties. Based on amino acid sequence homology within E7 proteins, they can be separated into three conserved regions denoted in an analogous fashion to adenovirus E1A as CR1, CR2, and CR3 (12). The CR2 and CR3 regions of HPV E7 share sequence homology with the corresponding regions of adenovirus E1A and SV40 large T antigen, including a strictly conserved LXCXE motif that mediates high affinity binding to pRb (13,14). The CR3 region of E7 contains two CXXC motifs that are separated by 29 or 30 residues, forming a novel zinc binding domain (3), which is also present in two copies in the primary sequence of E6 (15). The HPV E7 CR3 region has been shown to mediate protein dimerization (15)(16)(17) and to mediate direct interaction with several E7-interacting proteins. In particular, the E7 CR3 region contacts the C-terminal region of pRb (18), and full-length E7 proteins are at least 100-fold more potent in pRb binding than E7 CR1/CR2 constructs. The E7 CR3 region also mediates inactivation of the cyclin-dependent kinase inhibitors p27 and p21 (19,20) and several transcription factors that apparently contribute to HPV-mediated oncogenesis, including the TATA box-binding protein (TBP), a component of the NURD histone deacetylase complex Mi2␤ (21,22), the acetyltransferases p300/CBP, p300/CBPassociated factor (P/CAF) (23), and the transcription factor E2F (24).
A major cellular target for HPV-E7 and other viral oncoproteins in cell transformation is the pRb tumor suppressor. pRb represents a family of closely related proteins, including p107 and p130 that bind to E2F transcription factors and block their transcriptional activation function (25)(26)(27). Phosphorylation of pRb by cyclin/cyclin-dependent kinase results in the dissociation of pRb⅐E2F complexes and the transcriptional activation of E2F-regulated S-phase genes (28). pRb binding by viral oncoproteins SV40 large T antigen, adenovirus E1A, and HPV E7 results in the misregulated release of E2F transcription factors and the activation of S-phase genes. pRb contains pocket domain A and B and a C-terminal domain harboring target sites for cell cycle-dependent posttranslational modifications, including phosphorylation and acetylation (29). Although the A-B pocket of pRb contains a high affinity binding site for the E2F transactivation domain, the C-terminal pRb domain also contains a lower affinity binding site for other E2F regions, and the site of pRb interaction has been mapped roughly to a region that overlaps the HPV E7 CR3 binding site (18,30). The crystal structures of the A-B pocket of pRb in complex with peptides derived from either the high affinity E7 LXCXE peptide or a E2F 18-residue activation domain peptide show that, whereas the E2F peptide binds to the cleft formed by the A/B interface, the E7 peptide binds to the B domain nearly 30 Å away. These findings are consistent with in vivo and in vitro studies showing that LXCXE peptides cannot displace E2F from pRb (31, 32) and iso-thermal titration calorimetry studies showing that the two peptides bind to pRb independently (33,34). Moreover, these studies are consistent with observations that the E7 CR3 region cooperates with the CR2 region for E2F displacement from pRb (18) and subsequent cell transformation, although the mechanism for this is not clear.
Despite the large number of biological activities that map to the CR3 region of HPV-E7 as well as its related region within HPV-E6, this domain has previously resisted detailed structural and mechanistic analysis. In this study we now report on the high resolution x-ray crystal structure of the CR3 region of E7 from genotype 1A of HPV and use structure-based mutagenesis to dissect how the protein disrupts pRb⅐E2F complexes to mediate cell transformation.

MATERIALS AND METHODS
Protein Expression and Purification for Crystallization-DNA fragments encoding residues 44 -93 of native type 1A HPV E7 (HPV 1A E7-(44 -93)) protein preceded by DNA encoding an MK sequence was cloned into the pRSET A vector for protein overexpression in Escherichia coli cells BL21 (DE3) (Invitrogen). Cells were grown in LB medium supplemented with 100 g/ml ampicillin at 37°C, and when the A 595 nm reached 0.5, 1 mM isopropyl 1-thio-␤-D-galactopyranoside and 100 M zinc acetate were added to the media, and the cells were grown for 3 additional hours before they were harvested by centrifugation at 4000 ϫ g for 20 min to isolate the cell pellet that was stored at Ϫ70°C before protein purification. Frozen cell pellets were suspended in 20 mM Tris-HCl, pH 7.5, 100 mM sodium chloride, 10 M zinc acetate, and 10 mM dithiothreitol and lysed by sonication. The cell lysates were centrifuged at 40,000 ϫ g for 30 min, and the insoluble fraction containing the HPV 1AE7-(44 -93) was solubilized in 20 mM Ches, pH 10.15, 100 mM sodium chloride, 50 M zinc acetate, 10 mM dithiothreitol, and 6 M guanidine-HCl denaturant. This HPV 1AE7-(44 -93) denaturant solution was dialyzed against the same buffer without guanidinium-HCl overnight followed by a 4-h dialysis against a buffer containing 20 mM Tris-HCl, pH 7.5, 50 mM sodium chloride, 10 M zinc acetate, and 10 mM dithiothreitol. The solution containing the refolded HPV 1AE7-(44 -93) protein was centrifuged to remove precipitates formed during the refolding process, and the supernatant was then applied to a pre-equilibrated ion exchange Q-Sepharose column (Amersham Biosciences). The flow-through from the Q-Sepharose column, harboring the HPV 1AE7-(44 -93) protein, was collected, concentrated (Millipore), and loaded onto a Superdex 200 gel filtration column (Amersham Biosciences). Peak fractions containing HPV 1AE7-(44 -93) were judged to be greater than 95% pure by SDS-PAGE analysis with a typical yield of 100 mg of recombinant protein from 12 liters of growth culture. For the preparation of selenium-derivatized protein, base substitution encoding a single L76M mutation was introduced into the above construct by mutagenesis (QuikChange mutagenesis kit, Stratagene), and protein overexpression was carried out in the methionine auxotrophic E. coli cells B834 (DE3) (Novagen). B834 (DE3) cells were grown in selenomethionine-containing minimal media supplemented with the other 19 amino acids (35). Selenomethionine-derivatized protein was purified essentially as described for the wild-type construct with similar yield.
Crystallization, Data Collection, Structure Determination, and Refinement-Crystals of HPV 1AE7-(44 -93) were obtained by mixing 2 l of 6 mg/ml protein in a buffer containing 100 mM Ches, pH 10.15, 100 mM NaCl, and 10 mM dithiothreitol with an equal volume of reservoir solution containing 100 mM sodium citrate, pH 6.5, 200 mM MgCl 2 , and 2.2 M sodium chloride and equilibrating over reservoir solution at 20°C. Both native and selenium-derivatized crystals grew to a typical size of 300 ϫ 80 ϫ 80 m 3 after 2 days. The crystals were cryoprotected by transferring them to a reservoir solution supplemented with 50% saturated sucrose solution. Only the selenomethionine-derivatized crystals produced useful diffraction after cryoprotection and, therefore, were the only crystals used for data collection, structure determination, and refinement. Complete three-wavelength (peak, inflection, and remote) Zn-MAD (50 -2.3Å) and Se-MAD (50 -3.5 Å) data were collected from the selenium-derivatized crystals on beamline 19ID at the Advanced Photon Source (Argonne National Laboratories) using a 3 ϫ 3 mosaic CCD detector at 100 K. The data were processed and scaled using HKL2000 suite (HKL Research). The space group was determined to be P2 1 2 1 2 with two molecules in one asymmetric unit.
The structure of HPV 1AE7-(44 -93) was solved by Zn-MAD. The two zinc positions per 2 protein subunits in the asymmetric unit dimer were located and refined with the program SOLVE (36), yielding an initial figure of merit of 0.60. The initial experimental map was of excellent quality, facilitating the building of a partial model automatically with the program RESOLVE (37,38). A partial model refinement with the program CNS (39) employed simulated annealing, and torsion angle dynamic refinement protocols allowed for the placement of additional residues not present in the initial model. Subsequent refinement with CNS and iterative manual adjustments of the model with the program O (40) followed by translation, liberation, and screw-rotation refinements in the program REFMAC5 (41) resulted in a final model with excellent crystallographic and stereochemical parameters (see Table 1) and including residues 44 -93 of the protein and 88 water molecules. The final model was checked for errors against a composite omit map and analyzed by Procheck (42), revealing no residues outside of the allowed regions of the Ramachandran plot.
GST pull-down assays were carried out by mixing 30 g of GST-E7 fusion protein in a batch with 20 l of glutathione-Sepharose 4B beads (Amersham Biosciences) that were pre-equilibrated with the binding buffer 1ϫ PBS, pH 7.4, 100 mM sodium chloride, 5 M ␤-mercaptoethanol, 0.1% TWEEN 20. Equivalent molar amounts of either pRb or E2F1 protein were then added to the GST-E7 equilibrated beads in a total volume of 350 l, and the protein/bead mixture was further equilibrated at 4°C for 1 h accompanied by gentle mixing. Protein not bound to the glutathione beads was removed by centrifugation at 500 ϫ g for 3 min, and the beads were washed 2 times with 1 ml of binding buffer. The beads were resuspended in 30 l of SDS-loading dye and boiled for 5 min. Samples were analyzed on 12% PAGE Duramide gels (Cambrex) followed by either Coomassie Blue staining for pRb proteins or for E2F1 protein using Western analysis with anti-E2F1 antibody (Genetex) for E2F1-(243-437).
E2F Displacement Assays-10 g of His-pRb-(376 -928) was premixed with equivalent molar amounts of non-tagged E2F1-(243-437) and added to 30 l of pre-equilibrated nickel nitrilotriacetic acid beads in a total volume of 200 l of binding buffer containing 1ϫ PBS, pH 7.4, 100 mM sodium chloride, 5 M ␤-mercaptoethanol, 0.1% Tween 20 for further equilibration at 4°C for 30 min. Increasing molar amounts of GST-E7-E/D-CR3 (10-, 20-, 30-, and 50-fold excess) or untagged E7-CR2/CR3 (1-, 2-, and 5-fold excess) relative to His-pRB input was added to the beads with bound His-pRB-E2F1 complex to a final volume of 350 l for each displacement reaction. The reaction mixtures were mixed at 4°C for an additional 1 h. Protein not bound to the glutathione beads was removed by centrifugation at 500 ϫ g for 3 min, and the beads were washed 2 times with 1 ml of binding buffer before analysis of the protein-bound beads on SDS-PAGE as described above.

RESULTS
Overall Structure of HPV E7-CR3-The recombinant HPV 1AE7 CR3 region (residues 44 -93; Fig. 1A) was prepared from bacteria and crystallized in space group P2 1 2 1 2. The crystal structure was determined using MAD from the two bound zinc atoms per asymmetric unit and refined to a resolution of 1.6 Å, with excellent crystallographic statistics and protein geometry ( Table 1). The E7 CR3 domain assembles as a roughly globular and obligate dimer of approximate dimensions 30 Å ϫ 20 Å ϫ 25 Å (Fig. 1B). Each subunit of the dimer contains a two-stranded antiparallel ␤-sheet formed by residues 44 -52 (␤1) and 58 -65 (␤2) followed by a sharp U-turn leading to helix ␣1 (residues 67-79) that sits on one side of the sheet with its axis roughly parallel to the sheet. From the ␣1 helix, a 90-degree bend leads to ␤3 (residues 81-83) followed by an extended strand that cuts across the ␤1-␤2 sheet and leads to a final short ␣2 helix (residues 89 -91). The ␣2 helix sits on the opposite side of the ␤1-␤2 sheet relative to the ␣1 helix. A structural zinc ion is coordinated by four cysteines residues; two from the hairpin loop between ␤1 and ␤2 strands, one from the loop connecting the ␣1 and ␣2 helices, and one from the ␣2 helix (Fig. 1B). The dimer is formed by the face-to-face dyad-related packing of the two subunit ␤1-␤2-␣1 faces and ␤-sheet interactions between ␤2 and ␤3 strands of opposing subunits. FIGURE 1. Sequence alignment and overall structure of E7. A, sequence alignment of a selected set of HPV E7 proteins. Strictly conserved residues are shaded black, and conservative substitutions are shaded in gray. The CR1 and CR2 regions of HPV E7 (not included in this structure) are indicated above the alignment, and the secondary structural elements of the CR3 region (included in this study) are indicated above the alignment. Also indicated above the sequence alignment of the CR3 region are residues in 1AE7 that make hydrogen bond (open circle) and van der Waals (closed circles) dimer interactions and residues implicated by the mutational studies reported here to make interactions with pRb (solid triangles), E2F1 (solid diamonds), or residues that do not make detectable interactions with these proteins (solid stars). B, a schematic structure of the HPV 1AE7 dimer is shown in two orthogonal views, with the two protomers of the dimer colored in blue and green, and the bound zinc atoms and their cysteine ligands colored in purple (ball) and yellow (stick), respectively.
Superposition of either the E7-CR3 monomer or dimer structure against a structure database using the Dali server (www.ebi.ac.uk/dali/ index.html) identifies no structures with Z-scores over 2.8 (less than 2.0 is considered dissimilar). Analysis of the structures with Z scores between 2.0 and 2.8 reveals that none of them bind zinc or form homodimers and the structural homology is primarily restricted to the ␤1-␤2-␣1 region of the monomer. Based on this, we conclude that the structure of the E7-CR3 described here represents a novel zinc binding fold. A sequence alignment of E7 proteins shows a high degree of sequence conservation within the CR3 region with greater than 40% identity and 60% homology between proteins, suggesting that all E7-CR3 regions adopt the same dimeric structure (Figs. 1, A and B).
The E7-CR3 Dimer Interface-The E7-CR3 dimer interface appears to be maintained in large part by a hydrophobic core that buries a total solvent-accessible area of 2058 Å 2 ( Fig. 2A). A sequence alignment of the CR3 domains from E7 proteins shows that, of the residues that stabilize the core, five residues are strictly conserved, and three show high conservation. Residues that contribute to this dimer-induced core include dyad-related interactions from Val-48 from ␤1, Val-59, Leu-61, and Val-63 from ␤2, Leu-73, Met-76, and Leu-77 from ␣1, and Leu-81 and Ile-83 from ␤3. The carbonyl oxygens of Thr-62 and Leu-64 from one protomer make main chain hydrogen bonds to the amide groups of Asn-82 and Val-84 from the other protomer, and the amide groups of Leu-64 and Asp-66 from one protomer make hydrogen bonds to the carbonyl oxygens of Ser-80 and Asn-82 from the other protomer. An additional hydrogen bond is formed between the side chain of Gln-72 and the side chain of Met-76 from the other protomer. Finally, ␤3 from one protomer makes anti-parallel ␤-strand interactions with the end of ␤2 of the other subunit. Taken together, it would appear that the E7-CR3 domain would not be able to maintain this tertiary structure in monomeric form or in the absence of zinc coordination and is, therefore, likely to be an obligate and zinc-dependent dimer. This is consistent with solution studies showing that HPV-E7 undergoes a monomerdimer equilibrium with a dissociation constant of about 1 M (16).
Surface Features of E7-CR3-In an effort to identify regions of the E7 CR3 domain that may mediate interaction with pRb and other E7-CR3interacting proteins, we mapped HPV E7 conserved residues onto the surface of the HPV 1AE7-CR3 dimeric structure (Fig. 3A). This analysis reveals two conserved surface patches that are implicated as contact sites.
One surface patch maps to the outer edges of the two ␣1 helices of the dimer. Residues defining these edges, Ile-70, Arg-71, Glu-74, Glu-75, and Leu-78 of the ␣1 helix show a high degree of conservation among E7 proteins from different HPV genotypes. A second conserved patch maps to the ␤1 strands within a pronounced groove at the dimer interface (Fig. 3A). Specifically, residues Arg-60 and Leu-61 in this region are highly conserved among the E7 proteins. Interestingly, an electrostatic surface map of the HPV 1AE7 CR3 domain reveals that the regions around the ␣1 helix (patch 1) and ␤1 strand (patch 2) harbor the most electronegative and electropositive surfaces of the protein, respectively where T is a test data set of 9.8% of the total reflections randomly chosen and set aside before refinement).
( Fig. 3B). Based on the E7 sequence alignment, the structural features are likely to be conserved among the E7 proteins. This observation suggests that if the ␣1 and ␤1 regions of the CR3 domain of E7 do indeed mediate protein contact, this contact likely involves an electrostatic component.
Mutational Analysis of E7-CR3 to Probe pRb Interactions-Previous studies have shown that the E7 CR3 region mediates interactions to both pRb and E2F (18,24). To confirm that the CR3 region used for structure determination in this study is sufficient for pRb binding, we prepared GST-tagged 1AE7-(44 -93) (called GST-1AE7-CR3) for use in a GST pull-down assay with pRb, employing a pRb construct containing the A and B domains as well as region C (hereafter, pRb ABC , residues 376 -928 of pRb). As shown in Fig. 4A, GST-1AE7-CR3 specifically pulls down pRb, whereas the GST control fails to do so. In light of the structure of HPV 1AE7-CR3 and the noted conservation of solvent-exposed residues in the ␣1 helix (patch 1) and ␤1 strand (patch 2), we first tested the hypothesis that these patches might mediate the interaction with pRb. For this analysis we prepared a series of site-directed mutations of HPV-E7-conserved residues within the ␣1 and ␤1 strand within genotypes 1A and 16 of GST-tagged HPV E7 proteins and compared their affinity for pRb using a GST pull-down assay. The GST-E7 fusion proteins used for these studies did not include the high affinity LXCXEmediated protein binding motif within CR2 so that LXCXE-protein binding would not mask CR3-mediated protein binding, as the presence of the CR2 within the HPV protein increases its pRb affinity by about 300-fold (Supplemental Fig. 2). In addition, the E/D-rich region of E7 (located between the CR2 and CR3 regions of E7) was included (called GST-E7-E/D-CR3) because this region had previously been implicated to make weak pRb interactions (43) and because we anticipated that it may serve as a flexible linker between GST and the minimal CR3 domain of E7. From these experiments, wild-type and mutant E7 proteins from genotypes 1A and 16 showed similar patterns for pRb binding (Fig. 4B). Specifically, although most mutations of residues within patch 1 and patch 2 of E7 did not affect pRb binding, the E74R/E75R double mutant within patch 1 of HPV 1AE7 and the corresponding E80R/D81R double mutant of 16E7 reduced pRb binding by ϳ5and 6-fold, respectively. This experiment suggests that the electronegative region within patch 1 of the E7 proteins mediates pRb interaction. Interestingly one of the patch 2 mutants R66E of 16E7 showed abnormal electrophoretic properties and enhanced pRb binding affinity (Fig. 4B).
Given that our studies implicate that an acidic region of E7-CR3 mediates pRb binding and that a previous study mapped CR3 binding to residues within 803-841 of the C domain of pRb (33), a region harboring several strictly conserved lysine and arginine residues, we addressed whether basic residues in this region of pRb may mediate interaction to patch 1 of E7-CR3. Three single pRb mutants K810E, K814E, K824E, a triple mutant K810E/K814E/K824E, and a quintuple mutant K810E/ K814E/K824E/R828E/R830E were made and subject to GST pull-down assays with native GST-1AE7-E/D-CR3 and GST-16E7-E/D-CR3. Although none of the single lysine mutants within the C domain of pRb had significant effects on E7 CR3 binding, the pRb mutants containing multiple lysine substitutions resulted in a 3-5-fold reduction in E7 CR3 binding (Fig. 4C). This was true for both the 1A and 16 genotypes of HPV E7. Together these studies suggest that there is an electrostatic contribution to the interaction between the E7 CR3 domain and the C region of pRb.   Fig. 1B. The surface is color-coded according to amino acid conservation among the HPV E7 proteins (Fig. 1A) by the Consurf Server (consurf.tau.ac.il), where dark and light green represents highly conserved and partially conserved residues, respectively. Residues that were mutated for analysis in this study are highlighted. B, the same molecular surface representations are as in A but color-coded according to electrostatic potential as calculated by the program GRASP.
The finding that there is an electrostatic component to the interaction between an acidic patch on E7-CR3 and a basic region within the C-region of pRb suggested that mutations that reversed the charge in the interacting surfaces of both the E7 and pRb proteins might support binding between the two proteins. To test this hypothesis, GST pulldown experiments were carried out to compare the binding of the E74R/E75R 1AE7 and the E80R/D81R 16E7 CR3 domain mutants to native pRb as well as several charge reversal mutants within the C domain of pRb. These experiments revealed that the E7 mutants bound pRb less well than wild-type E7 (Fig. 4D), and as hypothesized, several single lysine to glutamate mutations within the C domain of pRb (K810E and K814E) restored pRB binding to the mutated E7 proteins from genotypes 1A and 16 to near wild-type levels (Fig. 4C). As a control, a single lysine mutant, K844E, outside the E7 binding region of pRb, failed to restore binding between pRb and the E7 mutant (Fig. 4D). Overall, the E7 mutants interacted less efficiently with the pRb C domain mutants containing multiple basic to acidic substitutions, implicating that the electrostatic contribution to the interaction between the pRb C domain and the E7-CR3 domain is not the sole determinant for binding between these protein domains.
Although our studies reveal an interaction between the E7 CR3 domain and the C domain of pRb in the context of a pRb ABC construct, similar pull-down studies between the E7 CR3 domain and the isolated C domain of pRb (residues 803-841) in the form of an N-terminal MBP fusion protein failed to show an interaction between the two proteins (data not shown). This result demonstrates that the A-B pocket domain FIGURE 5. Disruption of the pRb⅐E2Fcomplex by HPV E7 proteins. A and B, the ability of wild-type (WT) 1AE7-E/D-CR3 and 16E7-E/D-CR3 in the form of GST fusion proteins were analyzed for their ability to disrupt the pRb-E2F complex. Equimolar amounts of His-tagged pRb ABC -(376 -928) and E2F-(237-423) were pre-mixed and incubated with increasing amounts of the respective E7 protein. E2F that was bound to pRb was isolated by capture on nickel beads and analyzed by anti-E2F antibody Western blotting. C and D, the experiments were carried out as described in A and B, except that the ability of the PATCH 1 double mutants, 1AE7-E/D-CR3 E74R/E75R and 16E7-E/D-CR3 E80R/D81R, were used. E and F, the experiments were carried out as described in A and B, except that the pRb ABC K810E/K814E/K824E mutant was used. G, the experiments were carried out as described in A and B, except that the pRb ABC K810E and 16E7-E/D-CR3 E80R/D81R mutants were used.  A, GST pulldown assay using GST, GST-1AE7-CR3 fusion protein, and pRb ABC -(376 -928). Equal molar amounts of these three proteins were used. B, GST pull-down studies using wildtype (WT) and mutants of GST-1AE7-E/D-CR3 and 16E7-E/D-CR3 fusion proteins and pRb ABC -(376 -928). The HPV E7 mutants are designated as PATCH 1 and PATCH 2 as illustrated in Fig. 3A. Pulled down complexes were Coomassie-stained to show equal input of GST-E7 fusion proteins and the respective amounts of bound pRb ABC -(376 -928). Results shown were the representative one of three duplicates. Gel bands were quantified using the program Image J from NIH (rsb.info.nih.gov/ij). C, GST pull-down experiments in which wild-type E7-E/D-CR3 proteins were used to pull down both wild-type and mutant pRb ABC -(376 -928) proteins. D, the HPV 1AE7-E/D-CR3 E74R/E75R and HPV 16E7-E/D-CR3 E74R/E75R double mutants that showed significant loss of pRb binding capacity were subject to pull-down assays with wild-type and mutant pRb ABC -(376 -928) proteins, some of which restore HPV-E7/pRb binding. of pRb must also cooperate with the pRb C region for E7 CR3 binding. Moreover, the observation that pRb binds to two relatively divergent genotypes of E7, type 1A and type 16 (Supplemental Fig. 1), with similar sensitivities to mutation further suggests that the mode of pRb binding to E7-CR3 is conserved among E7 proteins from different HPV genotypes.
E7 CR3-mediated disruption of pRb⅐E2F Complexes-A major transforming activity of HPV correlates with the ability of the E7 protein to disrupt pRb⅐E2F complexes. To directly determine whether the CR3 region of E7 contributes to the disruption of pRb⅐E2F complexes, we prepared a 1:1 complex of bacterially expressed His-tagged pRb large pocket (His-pRb ABC ) and a bacterially expressed E2F fragment harboring residues 237-423 that was previously shown to be sufficient for pRb binding (33). To this pRB⅐E2F complex, we titrated in increasing amounts of GST-E7-E/D-CR3 domain from either genotypes HPV 1A or HPV 16. The amount to E2F retained in complex with pRb was then analyzed by Western analysis . Figs. 5, A and B, show that the E7 CR3 proteins from both HPV genotypes 1A and 16 disrupted pRb⅐E2F complexes in a dose-dependent manner, although these recombinant proteins were much less potent for pRb⅐E2F displacement than the LXCXEcontaining E7 and E1A protein constructs (Supplemental Fig. 3), confirming a requirement for the LXCXE motif for full displacement. To address whether the pRb⅐E2F displacement activity of HPV E7 was dependent on the interaction between patch 1 of the E7 CR3 domain and the C-domain of pRb, we carried out the similar experiments using either the E74R/E75R 1AE7 or K810E/K814E/K824E pRb mutants. Figs. 5, C and E, show that neither of these mutants supports the E7-mediated displacement of pRb⅐E2F complexes. Similar results were observed for the E80R/D81R 16E7 mutant (Figs. 5, D and F). This data demonstrate that the interaction between patch 1 of the E7 CR3 domain and the basic region within the C domain of pRb contributes to the disruption of pRb⅐E2F complexes.
Because the K810E pRb mutant restored the binding of pRb to the E7 patch 1 charge reversal mutants (1AE7 E74R/E75R and 16E7 E80R/ D81R), we asked whether these E7 mutants might promote E2F displacement from the K810E pRb mutant. As can be seen in Fig. 5G, the E7 charge reversal mutants are unable to disrupt preformed pRb⅐E2F complexes, suggesting that E7-CR3 binding to the basic region of the pRb C domain is not sufficient for disruption of pRb⅐E2F complexes.
Because E2F was previously reported to bind directly to the CR3 region of HPV16 E7 (24), we asked whether this interaction might be mediated through patch 1 and/or patch2 of E7-CR3 and whether this interaction contributed to E7-CR3-mediated disruption of pRb⅐E2F complexes. For these studies we carried out pull-down studies with E2F-(237-423) using wild-type, patch 1, and patch 2 mutants of GST-E7-E/D/CR3. The wild-type E7 protein efficiently pulls down E2F, as expected (Fig. 6A). However, patch 2 mutants, harboring R60E and L61Q substitutions (residues that are strictly conserved across most HPV genotypes), show a reduction in E2F binding by about 5-fold for each mutant. To address whether the interaction between patch 2 of E7-E/D-CR3 and E2F is important for E7-mediated disruption of pRb⅐E2F complexes, we assayed the ability of the patch 2 E7-CR3 mutants (R60E for 1AE7 and R66E for 16E7) to disrupt the complex. We found that the arginine to glutamate mutants of E7 patch 2 were unable to disrupt pRb⅐E2F complexes in either the context of 1AE7 or 16E7 (Figs. 6, B and C). Mutation of the conserved leucine residue within patch 2 (Leu-61 in 1AE7 and Leu-67 in 16E7) has been previously shown to disrupt other E7-mediated protein contacts, including loss of binding to P/CAF and loss of histone deacetylase interaction (22,23), whereas the pRb binding capacity of this E7 mutant remains unchanged. Although mutation of this residue also disrupts E2F binding (Fig. 6A) and the E2F displacement activity of E7 (data not shown), it may also destabilize the structure of the E7-CR3 dimer, since this residue sits at the interface of the homodimer. This observation may explain the deleterious nature of this mutation in E7. Together, these studies demonstrate that patch 2 of E7-CR3 mediates E2F interaction, and this interaction is required for disruption of pRb⅐E2F complexes.

DISCUSSION
Crystallographic studies of the pRb pocket domain bound to either a high affinity LXCXE motif within the CR2 region of HPV E7 (34) or a high affinity transactivation peptide of E2F (33,44) have revealed that, whereas the E2F peptide binds to a groove between the A and B domains of pRb, the E7 peptide binds to the B domain nearly 30 Å away from the E2F peptide. This is consistent with in vitro data showing that the CR2 region of E7 is not sufficient to mediate E2F displacement and that the CR3 region is also required for this activity. Moreover, in vitro studies also reveal that a region N-terminal to the E2F transactivation domain, harboring a marked box region, contains a low affinity binding site for pRb and that this binding is destabilized by the binding of a CR3-containing E7 construct (33). Putting these previous studies together with our findings that the E7 CR3 region harbors binding sites for both the C domain of pRb and the marked box region of E2F and contributes to the disruption of pRb⅐E2F complexes through these interactions, we are able to propose the following model for E7-mediated dissociation of pRb⅐E2F complexes (Fig. 7). First, E2F binds to pRb through at least two sets of interactions; that is, an interaction between the E2F transactivation domain and the pRb-A/B pocket and an interaction between the E2F-marked box with the pRb-A/B pocket and C-domain. We also propose that the interaction between the E2F marked box and the pRb C-domain involves the basic region of the pRb C domain. HPV E7 then mediates disruption of the pRb⅐E2F complex by first forming a stable complex with pRb via an interaction between the LXCXE-containing CR2 region of HPV E7 and the B domain of the pRb pocket region. This high affinity pRb/HPV E7 interaction then presents patch 1 of the HPV E7 CR3 domain to make low affinity contacts to the basic region within the C-domain or pRb and patch 2 of the CR3 domain of HPV E7 to make contacts to the C-terminal region marked box region of E2F. We propose that the HPV-E7 CR3 mediated destabilization of the pRb⅐E2F marked box interaction is sufficient to drive E2F displacement from pRb. Although we do not currently know the mechanism for destabilization of the interaction between the pRB-A/B pocket and the E2F transactivation domain, it may occur through some allosteric mechanism (Fig. 7A). Additionally, it is well established that cyclin-dependent kinase/cyclin hyperphosphorylation of the pRb C-domain leads to the displacement of E2F and, specifically, to the displacement of the transactivation domain of E2F (28,33,45). In light of this, it is also possible that displacement of the E2F marked boxed from the pRb C domain makes the C domain more accessible to cyclin-dependent kinase/cyclin-mediated phosphorylation in vivo, which in turn leads to the displacement of the E2F transactivation domain from the pRb pocket region (Fig. 7B). Consistent with this hypothesis, the adenovirus E1A has also been shown to modulate the phosphorylation levels of the pRb pocket proteins (46). Interestingly, the cyclin-dependent kinase/cyclin phosphorylation sites within the C domain of pRb are located within the basic region of the pRb C-domain that mediates HPV E7-CR3 binding.
In this study we have determined the high resolution crystal structure of the CR3 domain of the HPV E7 protein and have employed structure-based mutagenesis to address the molecular basis for how this domain of HPV E7 perturbs the function of the human pRb tumor suppressor protein. In particular, we have identified two conserved surface patches on the E7 CR3 domain that mediates its pRb inhibitory function. High risk forms of HPV are causative agents for cervical cancer and are, therefore, important therapeutic viral targets. In light of the fact that the HPV E7 protein plays a key role in cell transformation and our observation that the E7 CR3 domain contains a novel protein fold with no structural homologues in humans, we propose that the CR3 domain of E7 domain may be an excellent candidate for targeted inactivation by small molecule compounds. In particular, the development of molecules that bind patch 1 and/or patch 2 of the E7 CR3 domain may disrupt the ability of E7 to perturb pRb function. In addition, the CR3 domain of E7 has been implicated in mediating inactivation of several other human proteins associated with human cancer including histone deacetylases (22) and the p300/CBP acetyltransferases (47,48). Therefore, small molecule compounds that specifically bind to the HPV-E7 CR3 domain might disrupt other cell-transforming activities of this virus. Indeed, a recent report has demonstrated that the integrity of the CR3 domain of HPV-E7 is essential for the life cycle of HPV (49). The structure-based development of small molecule HPV-E7 CR3-inactivating compounds may, therefore, be a therapeutically beneficial avenue of investigation.