R273C and R273H Are Classic Contact Mutants.

R273C and R273H are two of the five most frequent cancer-associated mutations. They affect Arg-273, which makes major contacts with the phosphate backbone of target DNA. Having previously solved the crystal structure of T-p53C-R273H (16), we have now determined the structure of T-p53C-R273C (at 1.8-Å resolution) to elucidate the structural changes in the cysteine variant. Overall, the structures of T-p53C, T-p53C-R273H, and T-p53C-R273C are virtually identical. A pairwise superposition of the C^α atoms from equivalent chains gives rmsds of <0.16 Å. The R273C mutation simply removed a DNA contact without perturbing the conformation of neighboring residues, such as Phe-134 and Asp-281 (Fig. 2). The preservation of the overall architecture of the DNA-binding surface in the structures of T-p53C-R273C and T-p53C-R273H explains why these mutants exhibit residual DNA-binding activity despite the loss of a crucial DNA contact (ref. 18 and unpublished data).

Open in a separate window

Fig. 2.

Crystal structure of T-p53C-R273C. Stereoview of the mutation site in the structure of T-p53C-R273C [Protein Data Bank (PDB) ID code 2J20, yellow] superimposed on the structure of T-p53C (PDB ID code 1UOL, orange), T-p53C-R273H (PDB ID code 2BIM, gray), and DNA-bound wild type (PDB ID code 1TSR, light gray). One strand of bound DNA in the vicinity of Arg-273 is shown as a gray line, with small spheres indicating the continuation of the DNA backbone.

G245S Induces Small Conformational Changes in the L3 Loop.

The G245S mutation is located in the L3 loop, which binds to the minor groove of DNA response elements via Arg-248 (Fig. 1). The crystal structure of T-p53C-G245S revealed distinct structural changes in the immediate environment of the mutation site. Observed conformational changes in some surface loops (e.g., in the L1 loop and the S7–S8 loop) can be attributed to differences in crystallization conditions and crystal packing for T-p53C and T-p53C-G245S and reflect the intrinsic conformational flexibility of these loop regions (12, 17, 19). In T-p53C-G245S, the hydroxyl group of Ser-245 points toward the zinc ligands Cys-238 and Cys-242 and displaces a structural water molecule that is observed in the wild-type and various mutant structures (Fig. 3A). The main-chain atoms of Ser-245 form the same hydrogen bonds with neighboring residues (Cys-242 and the side chain of Arg-249) as observed for the glycine in the wild type and T-p53C, although the backbone conformation of Ser-245 is slightly different. Both dihedral angles and ψ have changed by ≈20° and fall within a “generously allowed” region of the Ramachandran plot ( = −141°, ψ = −102°). As such, Ser-245 adopts a moderately unfavorable main-chain conformation.

Open in a separate window

Fig. 3.

Structural effects of the cancer mutations G245S and R282W. (A) Stereoview of the zinc-binding region of T-p53C-G245S (PDB ID code 2J1Y, molecule A, yellow) superimposed on the structure of T-p53C (PDB ID code 1UOL, molecule A, gray). The C^α atom of the DNA-contact residue Arg-248 is depicted as a small sphere in the color of the corresponding chain. A structural water molecule that is present in T-p53C but displaced by the side chain of Ser-245 in T-p53C-G245S is shown as a magenta sphere. (B) Stereoview of the loop–sheet–helix motif in the crystal structure of T-p53C-R282W (PDB ID code 2J21, molecule A, yellow) superimposed on T-p53C (PDB ID code 1UOL, molecule A, gray). Stabilizing interactions mediated via Arg-282 in T-p53C are shown as dotted lines. Residues 117–121 of T-p53C-R282W were disordered. The C^α atoms on both sides of the chain break are highlighted by small yellow spheres. (C) C^α trace of a wild-type core domain dimer bound to a DNA half-site (PDB ID code 2AC0, black) (10). The two strands of bound DNA are shown in blue and magenta. T-p53C-G245S, shown in yellow, is superimposed on one of the two monomers. Selected residues at the core–core domain interface are highlighted.

Most interestingly, the mutation appears to trigger a flip of the peptide bond between Met-243 and Gly-244, resulting in a displacement of the corresponding C^α atoms by 0.7 and 1.7 Å, respectively. This peptide flip relative to the conformation found in the structures of T-p53C and DNA-bound wild type was observed in all four molecules of the asymmetric unit. There is also a significant structural response from Pro-177, which follows the movement of Gly-244 to avoid steric clashes and is shifted to a similar extent. The backbone conformation of residues 246–250 of the L3 loop, which includes the DNA-contact residue Arg-248, is only marginally affected, and their C^α displacements are in the 0.5-Å range. Some of the residues with the most significant shifts in their C^α atoms (Met-243, Gly-244, and Pro-177), however, are key residues in the subunit interface of the core domain dimer bound to a DNA half-site (10, 20, 21). (Fig. 3C). Although the G245S mutation has no major direct effect on the DNA-contact residue Arg-248 in the DNA-free form of T-p53C-G245S, it presumably puts conformational strain on the subunit interface upon DNA binding. Such effects on the subunit interface would explain DNA-binding studies on full-length T-p53-G245S, which show that specific binding to the gadd45 response element is ≈15-fold reduced relative to T-p53 at physiological ionic strength (18). In contrast, R249S, the other structural cancer hotspot mutation in the L3 loop, substantially perturbs the L3 loop, induces high flexibility, and favors alternative conformations with much more severe consequences for the overall protein stability and DNA binding (16, 18). Hence, for these two structural hotspot mutations in the L3 loop, there is a direct correlation between the magnitude of loss of DNA binding and the extent of the observed structural perturbation.

Our structures also explain interactions of p53 mutants with other proteins, such as the apoptosis-stimulating protein ASPP2. Under conditions where the mutants are fully folded, several cancer mutants bind to the 53BP2 domain of the ASPP2 protein at wild-type levels, whereas no binding is detected for the mutants T-p53C-R249S and T-p53C-G245S, which have distinct structural changes in the region that forms the central part of the p53–53BP2 interface (22).

The R282W Mutation Affects the Packing of the Loop–Sheet–Helix Motif.

Arg-282 plays a crucial role in maintaining the structural integrity of the loop–sheet–helix motif that binds to the major groove of DNA target sites (Fig. 1). Its side chain is involved in a network of interactions that pack helix H2 against the S2–S2′ β-hairpin and loop L1 (Fig. 3B). The 1.6-Å crystal structure of T-p53C-R282W revealed that the overall fold of the protein is not compromised by the R282W hotspot mutation. The C^α atoms of T-p53C-R282W and T-p53C can be superimposed with a rmsd of 0.26 Å. There are, however, substantial structural perturbations in the loop–sheet–helix motif (Fig. 3B), and several stabilizing interactions are lost, which accounts for the loss of thermodynamic stability of 3 kcal/mol. The region of the L1 loop next to Trp-282 is disordered because of steric hindrance, and no conclusive electron density was observed for several residues within this loop (residues 117–121, including the DNA contact Lys-120). As such, Trp-282 displaces what appears to be the weakest structural link in the loop–sheet–helix motif. The side chain of Trp-282 is solvent-exposed and much less well embedded within the protein structure than the arginine in T-p53C. The destabilization of the loop–sheet–helix motif is also reflected in the B factor profile, which shows an increased mobility for the main-chain atoms at both ends of the chain break in loop L1 and around the mutation site in the C-terminal helix, and this trend is propagated toward the chain end (Fig. 6, which is published as supporting information on the PNAS web site). The overall packing characteristics of the C-terminal helix containing the DNA-contact residues Cys-277 and Arg-280 against the core of the protein, however, are preserved. Hence, at the subphysiological temperatures at which the mutant R282W is largely folded, it is still able to bind gadd45 DNA (14) despite the structural disorder in the L1 loop containing the DNA-contact residue Lys-120. This retention of binding activity is consistent with a recent mutational analysis of the L1 loop (23) and with the observation that this loop is a mutational “cold spot” that is only rarely mutated in human cancer (3), indicating that structural perturbations in the L1 loop are tolerated to a certain degree without abrogating p53 function.

Y220C Creates a Solvent-Accessible Cleft at the Periphery of the β-Sandwich.

Y220C is the most common cancer mutation outside the DNA-binding surface (compare release R10 of the TP53 mutation database at www-p53.iarc.fr) (3). It is located at the far end of the β-sandwich, at the start of the loop connecting β-strands S7 and S8 (Fig. 1) and destabilizes the core domain by 4 kcal/mol (Table 2). The benzene moiety of Tyr-220 forms part of the hydrophobic core of the β-sandwich, whereas the hydroxyl group points toward the solvent. The 1.65-Å crystal structure of T-p53C-Y220C showed that the Y220C mutation creates a solvent-accessible cleft that is filled with water molecules at defined positions but leaves the overall structure of the core domain intact (Fig. 4). The structural changes upon mutation link two rather shallow surface clefts, preexisting in the wild type, to form a long, extended crevice in T-p53C-Y220C, which has its deepest point at the mutation site (Fig. 4 B and C). Cys-220 occupies approximately the position of the equivalent atoms of Tyr-220 in the wild type. The positions of neighboring hydrophobic side chains located in the core of the β-sandwich have not shifted significantly. The mutation, however, results in a loss of hydrophobic interactions and a suboptimal packing of these hydrophobic core residues. The side chain of Leu-145, which was buried in the wild type, for instance, becomes partly solvent accessible in T-p53C-Y220C. The largest structural changes in the immediate environment of the mutation site are found in the S7–S8 loop for Pro-222. Throughout the structure, there is no C^α displacement >0.9 Å.

Open in a separate window

Fig. 4.

Crystal structure of T-p53C-Y220C. (A) Stereoview of the mutation site at the periphery of the β-sandwich in T-p53C-Y220C (PDB ID code 2J1X, molecule A, yellow) superimposed on the structure of T-p53C (PDB ID code 1UOL, molecule A, gray). Several water molecules close to Cys-220 in T-p53C-Y220C that fill the cleft created by the mutation are shown as red spheres. (B) Molecular surface of T-p53C around Tyr-220. (C) Molecular surface of T-p53C-Y220C. The view is the same as in B. The position of the side chain of Tyr-220 in T-p53C is shown as a stick model.

β-Sandwich Mutations and the Molecular Basis of Temperature Sensitivity.

About one-third of the reported cancer mutations in the p53 core domain are located outside the structural elements that form the DNA-binding surface (loops L2 and L3 and the loop–sheet–helix motif). The V143A mutant is of particular interest because of the well documented temperature sensitivity of its binding to many response elements in both yeast and mammalian systems. At body temperature, the mutant is inactive and unfolded, whereas it retains transactivational activity at lower temperatures (24, 25). The structures of T-p53C-V143A and T-p53C-F270L provide the molecular basis for understanding the temperature-sensitive behavior of many β-sandwich mutants. Both mutations created internal cavities in the hydrophobic core of the β-sandwich without collapse of the surrounding structure (Fig. 5). The cavity volume was 47 ų (T-p53C-V143A) and 51 ų (T-p53C-F270L), as defined by the volume that can be occupied by a probe mimicking the size of a water molecule (1.4-Å probe radius). Additional information on these cavities is provided in Table 3, which is published as supporting information on the PNAS web site. In agreement with the hydrophobic nature of the cavities, no ordered buried water molecules were detected in the crystal structure. Although the overall structure of the core domain was perfectly conserved, the creation of void volumes came at a high energetic cost of 3.7 and 4.1 kcal/mol. These observations are consistent with studies on “large-to-small” substitutions in the hydrophobic core of T4-lysozyme and barnase (26–28).

Open in a separate window

Fig. 5.

Crystal structures of T-p53C-V143A and T-p53C-F270L. (A) Stereoview of the structure of T-p53C-V143A (PDB ID code 2J1W, yellow) superimposed on T-p53C (PDB ID code 1UOL, gray). All residues in the hydrophobic core of the β-sandwich within a 4.5-Å radius of the Val-143 side chain in T-p53C are shown. (B) Stereoview of the structure of T-p53C-F270L (PDB ID code 2J1Z, yellow) superimposed on T-p53C (PDB ID code 1UOL, gray). All residues within a 6-Å radius of the Phe-270 side chain in T-p53C are shown.

A recent study has identified a large number of temperature-sensitive p53 mutants (29). Most mutations were clustered in the β-sheet region of the protein, and the substitutions were mainly from large hydrophobic residues to smaller hydrophobic residues. Incidentally, V143A was not detected in this study, whereas mutations at residue 270 were (F270I and F270C). Interestingly, the Y220C mutation has also been reported to cause temperature-sensitive behavior (25). Again, this behavior is in agreement with our crystallographic data, which show that the mutation-induced structural changes are very localized, far away from the DNA-binding surface. A common structural feature of the β-sandwich mutants seemed to be that there were only minor structural disruptions upon mutation, although the effect on the thermodynamic stability of the protein was generally more severe than for the hotspot mutations in the DNA-binding surface. The much more compact and robust structural framework of the β-sandwich compared with the zinc-binding region and the loop–sheet–helix motif renders it generally much less susceptible to mutation-induced structural changes, in particular for large-to-small substitutions. The absence of structural changes in surface regions, especially in the DNA-binding surface, however, is key for functionality. Hence, temperature-sensitive behavior can be expected for all cancer mutations that destabilize the core domain without compromising the surface complementarity that is crucial to the function of p53, not only for binding to specific promoter sequences but also for interactions with regulatory proteins and for correct domain organization in tetrameric full-length p53 (9, 30–33).

Implications for Rescue Strategies.

At first glance, the discussion of the local structural changes induced by highly destabilizing mutations appears to be rather academic when assessing the functional consequences under physiological conditions, because these mutant proteins will most likely be largely unfolded. Our structural observations, however, have profound implications for therapeutic strategies that aim to rescue the function of p53 with small-molecule drugs that stabilize p53 (reviewed in refs. 34 and 35). The mutations N239Y and N268D have been reported to restore transcriptional activity in a subset of cancer mutants, including G245S (36). Double-mutant cycles suggest that they act as global stability suppressors (37). As such, they mimic the effects of a hypothetical generic small-molecule drug. The structure of T-p53C-G245S, however, exhibits distinct structural perturbations despite the presence of these stabilizing suppressor mutations, concomitant with altered functional properties (e.g., impaired binding of 53BP2; see above). For mutants with local structural changes in the DNA-binding surface, simple stabilization by means of a generic small molecule may not be enough to fully restore activity. A prime example is the mutant R249S. Studies on second-site suppressor mutations have shown that full restoration of DNA-binding activity is found only in the presence of a mutation (H168R) that specifically reverses the structural changes induced by the oncogenic mutation (16, 36, 37).

On the basis of our structural studies, β-sandwich mutants, such as V143A and F270L, represent much more promising targets for rescue by generic small-molecule drugs, because, in this case, stabilizing the wild-type conformation of the protein by ligand binding may be sufficient to restore wild-type-like activity under physiological conditions. The Y220C mutant not only has the potential of being rescued by a generic wild-type-binding compound but also has the possibility of being a target for a mutant-selective drug that can bind in the crevice formed by the deletion (Fig. 4C). This crevice region is particularly attractive because it appears to be distant from the functional sites and interfaces of the protein.

Equilibrium Denaturation.

Urea denaturation studies were performed as described in ref. 16 by using the equations described in ref. 14 for data analysis.

Crystallization and Structure Determination.

All crystals were grown at 17°C by using the sitting drop vapor-diffusion technique. Crystals of T-p53C-V143A, T-p53C-Y220C, T-p53C-F270L, and T-p53C-R282W were grown under the conditions described for T-p53C (17); crystals of T-p53C-R273C were grown under the conditions described for T-p53C-R273H (16). Crystals of T-p53C-G245S were grown by mixing 2 μl of protein solution (6 mg/ml) with 0.4 μl of water and 1.6 μl of the reservoir solution (19% polyethylene glycol 3350/0.19 M calcium acetate, pH 7.2). Crystals were flash-frozen in liquid nitrogen by using mother liquor with either 20% polyethylene glycol 200 or 20% glycerol as a cryoprotectant. X-ray data sets were collected at 100 K on beamlines 10.1 and 14.1 at the Synchrotron Radiation Source (Daresbury, U.K.). Data processing was performed by using Mosflm (44) and Scala (45). All crystals except T-p53C-G245S belonged to space group P2₁2₁2₁ with two molecules per asymmetric unit and were isomorphous to those obtained for T-p53C and T-p53C-R273H (16, 17). Structure solution and refinement were performed with CNS (46). After an initial round of rigid-body refinement using the structure of either T-p53C (PDB ID code 1UOL) or T-p53C-R273H (PDB ID code 2BIM) as a starting model, the structures were refined by iterative cycles of refinement with CNS and manual model-building with MAIN (47). Crystals of T-p53C-G245S belonged to space group P2₁ with four molecules per asymmetric unit. The structure was solved by molecular replacement with CNS using T-p53C as a search model. Subsequent refinement was performed as described above for the other mutants. The final models (1.6- to 1.8-Å resolution) had a crystallographic R factor of 18.3–19.8% (R_free = 20.6–22.3%) and excellent stereochemistry as verified with PROCHECK (48). Detailed data collection and refinement statistics are summarized in Table 1.

We thank Caroline Blair for protein purification, Fiona Sait for setting up crystallization trials for T-p53C-G245S, Dr. Frank Boeckler for help generating Fig. 4 B and C; and the staff at the Synchrotron Radiation Source for helpful advice and assistance in data collection. This work was supported by the Agency for Science, Technology, and Research of Singapore (H.C.A.), Cancer Research UK, the Medical Research Council, and European Community FP6 funding.