Protein Hydration and Stability.

Several hydrophilic side chains that are totally conserved in other vertebrate p53 genes or replaced by polar residues (see Fig. 8, which is published as supporting information on the PNAS web site) had very low solvent accessibility (Table 2). At least five of them (Thr-155, Ser-215, Thr-253, Thr-256, and Ser-269) are temperature-sensitive hot spots (14). Some of these residues had low protection factors in the H/D experiments (e.g., Ser-240, Tyr-163) indicating that their backbone accessibility is higher than their side-chain accessibility. However, most of them had high backbone protection factors (e.g., Cys-238, Thr-253, Thr-256, Tyr-234, and Tyr-236) (see Fig. 9, which is published as supporting information on the PNAS web site) with their corresponding HSQC peaks still present after several days in D₂O.

NMR spectroscopy allows the direct observation of protons and therefore is a very powerful tool for studying protein hydration (15). Assignments of labile protons such as hydroxyl and sulfhydryl are underrepresented, however, in the chemical-shift database (16), because they can only be observed when they are in slow exchange with the solvent (buried). By using triple-labeled (¹⁵N/¹³C/²H) samples, labile resonances not directly bound to ¹⁵N or ¹³C could be readily assigned to hydroxyl or sulfydryl groups in the protein, which could be unambiguously assigned by their NOE interactions with other neighboring protons. We observed and assigned 13 hydroxyl and 2 sulfhydryl groups of low solvent accessibility (Table 2). The cysteines coordinating the Zn²⁺ ion were among the missing sulfhydryl protons: one of them (242) because it is exposed and in fast exchange with the solvent and the other two (176 and 238) because they are likely to be unprotonated ligands. The backbone amides of Cys-176 and 242 exchanged more rapidly than did Cys-238 (protection factor = 7.5 × 10⁵). Interestingly, Cys-238 was buried but not involved in intramolecular H-bonds, whereas Cys-242 formed an interaction with carbonyl of Asn-239. H/D exchange gave additional information about the structure and stability around the metal coordination site.

The presence of a buried unpaired H-bonding group costs between 0.5 and 1.5 kcal·mol⁻¹ (17). Frequently, buried water molecules can act as the H-bonding partners, but at enthalpic and entropic costs (18). Most of the buried hydroxyl or sulfhydryl protons showed strong NOEs (or exchange peaks) with water (Table 2 and Fig. 4), a good indication of the presence of water molecules in their vicinity. Most of the Ser and Thr hydroxyls are close to the surface, and the observed crosspeak is likely to arise from surface waters with residence times in the picosecond to nanosecond time scale (fast exchange with the bulk solvent).

Open in a separate window

Fig. 4.

Detailed view of 2D NOESY spectra showing the NOEs (or exchange) peaks between protected hydroxyl/sulfhydryl protons and the water resonance. Three different spectra are shown at increasing mixing times (200, 400, and 600 ms). The crosspeak between Thr-253 Hγ2 and the water can only be observed at long mixing times, which indicates the absence of structural waters nearby this protected OH.

There were at least four hydroxyls (Thr-155 Hγ1, Tyr-234 Hη, Tyr-236 Hη, and Thr-253 Hγ1) without the water NOE crosspeak. Three of them are critical for the formation of the protein core (Fig. 5). The Thr residues were involved in intramolecular H-bonds: Thr-155 Hγ1··Pro 152 O; and Thr-253 Hγ1··Tyr-236 Oη. The highly conserved Tyr-234 (Fig. 8) is likely to be H-bonded to structural water molecules that have tetrahedral coordination to protein H-bond donors and acceptors and low B-factors in the x-ray structure (e.g., H₂O 2367 in the structure with Protein Data Bank ID code 1TUP). The lack of the intermolecular NOE peak could be explained if this water molecule is in slow to intermediate exchange (<10⁻² s⁻¹) with the bulk solvent. Thus, the overwhelming proportion of solvent molecules will cause a severe line broadening of the structural water resonances that preclude its observation. H/D exchange data supported this hypothesis because all backbone amides neighboring this water are protected, so the water exchange can only occur upon global unfolding.

Open in a separate window

Fig. 5.

Structural detail of the H-bond networks among Thr-253, Tyr-236, and Tyr-234 in the hydrophobic core of p53 DNA-binding domain. These three residues show protected hydroxyl groups (see text) consistent with null accessibility in the structure. (A) Most of the conformers representing the NMR structure show a H-bond between Thr-253 Hγ2 and Tyr-236 Oη. (B) Only little population of conformers displayed a suboptimal H-bond between Tyr-236 Hη and Thr-253 Oγ2. Although protected, there is not an intramolecular H-bond acceptor for Tyr-234 Hη, and the structural analysis suggests the presence of a structural water molecule stabilizing this hydroxyl inside the hydrophobic core of the protein (see text).

Evidence from NMR and crystallographic structures and the lack of NOEs with water supports the absence of a H-bond partner for Tyr-236 Hη. Although the substitution of Tyr-236 by Phe in other p53 genes (Fig. 8) removes a hydroxyl group with one unpaired H-bond from the interior, it also removes the H-bond acceptor of the totally conserved Thr-253 hydroxyl (Fig. 5A). These compensating factors lead to the Tyr-236 → Phe mutant being only marginally more stable (0.27 kcal/mol) than WT (19). However, a double mutation (Tyr/Phe and Thr/Ile) as found in human p53 paralogs p63 and p73 (Fig. 8) could cause significant stabilization. Model building indicated that the p53 structure could accommodate the double mutation without steric clashes. We made the double mutant Tyr-236 → Phe/Thr-253 → Ile and found it to be 1.6 kcal/mol more stable than WT (Fig. 6).

Open in a separate window

Fig. 6.

Urea denaturation curves of WT (open circles) and double-mutant Y236F/T253I at 10°C and pH 7.2.

It is worth summarizing the evidence for the assignment the hydroxyl and sulfydryl protons. Their resonances disappeared after D₂O exchange. Therefore, they are labile protons. They are not ¹⁵N- or ¹³C-bound because they did not show active ¹J coupling with either of these nuclei. The resonances of the hydroxyls in the amide–amide region (Tyr OHs mostly) did not correspond to any ¹⁵N HSQC peak: e.g., the OH of Tyr-205 was at >10 ppm, and the lowest field amide in the HSQC is at 9.5 ppm. We checked that these resonances did not come from other buffer components (i.e., use fully deuterated Tris·HCl) or contaminants in the expression/purification of the sample. The specific assignment of each of these groups was based on the NOEs that each of them had with other NHs in the structure. Because these groups are buried, it is possible to observe several NOEs with NHs (in the perdeuterated sample) and with aromatic or methyls (in the reverse-protonated samples). It is possible to assign without ambiguity nearly all of the OH and SH protons. There was only one that could not be assigned because the lack of NOEs with other protons (Table 2).

Protein Expression and Purification.

Isotopically labeled (¹⁵N,¹³C) samples were expressed in Escherichia coli C41(DE3) cells (20). Cultures were grown in minimal M9 medium supplemented with ampicillin (120 μg/ml), vitamin mix, and the appropriate nitrogen (1 g/liter of ¹⁵NH₄) and carbon (4g/liter of ¹³C-Glucose) sources. After induction at 37°C, cultures were incubated overnight at 25°C and harvested and disrupted by sonication. After centrifugation, p53-containing fusion protein was purified by Ni²⁺ affinity [Ni-NTA (Qiagen, West Sussex, U.K.) or Ni Hi-trap (Amersham Pharmacia)], and the fusion protein was dialyzed and cleaved by thrombin digestion (overnight at room temperature in 150 mM NaCl/25 mM Tris·HCl, pH 7.0/5 mM DTT). The product was diluted twofold with 25 mM Tris·HCl (pH 7.0) plus 5 mM DTT, filtered, and purified by ion-exchange chromatography (Resource 15S, Amersham Pharmacia).

The production of perdeuterated and reverse-protonated samples followed a similar protocol using ²H₂O-based M9 medium supplemented with 0.3–1% rich triple-labeled (¹³C/²H/¹⁵N) medium (BioExpress, Cambridge Isotope Laboratories, Cambridge, MA). Cells were grown directly in ²H₂O without prior adaptation and at 25°C. Reverse-labeled samples were produced by adding the corresponding α-ketoacids (12) or amino acid (Tyr, Phe, Arg, or Met) to the culture 2–3 h before induction.

p53 constructs used in the urea denaturization (WT and double mutant) were expressed and purified according to published protocols (19).

NMR Spectroscopy.

The core domain of p53 aggregates at temperatures of ≈37°C (21). The transverse relaxation times of the amide protons indicated the presence of species of high M_r in fast exchange with the monomers at concentrations >400 μM (pH 7.2; 150 mM ionic strength). Excellent spectra were obtained at 20°C by using 200–400 μM samples. Triple resonance experiments were performed on a triple-labeled (¹³C,¹⁵N,²H) sample to confirm and complete our previous backbone assignments of p53 core domain (K. B. Wong, S.M., and V.F., unpublished data). HNCA, HNCO, HN(CO)CA, CBCA(CO)HN, HN(CO)CC (8 and 12 ms for carbon spin lock), and HNCACB experiments were modified from the standard (22) to include deuterium decoupling. Side-chain assignments were completed with various HCCH-COSY in double (¹³C,¹⁵N) or triple (¹³C,¹⁵N,²H 50%) labeled samples.

Distance constraints were obtained from a collection of NOESY spectra with mixing times ranging from 100 ms (protonated samples) to 200 ms (deuterated and reverse-labeled samples). The following series was used in the order (sample number, labeling regime, and spectra): 1, unlabeled, 2D NOESY water, and D₂O (80 ms); 2, ²H (50%), ¹³C, ¹⁵N, 2D NOESY and various 3D NOESY-HSQC; 3, ²H (100%), ¹³C, ¹⁵N, 2D NOESY (100–600 ms mixing), 3D NOESY-HSQC; 4, as 3 plus reverse ¹³CH₃ Ile,Leu,Val, 2D NOESY, 3D NOESY-¹³C-HSQC, 3D ¹³C-HSQC-NOESY-¹³C-HSQC; 5, as 4 plus reverse ¹²C/¹H Tyr, 2D NOESY, 3D NOESY-¹³C-HSQC; 6, as 3 plus reverse ¹²C/¹H Phe, Met, 2D NOESY (200 ms); and 7, as 3 plus reverse ¹²C/¹H Phe, Arg, 2D NOESY (200 ms). For the perdeuterated sample, several 2D NOESY with 100, 200, 400, and 600 ms were recorded to check the absence of spin-diffusion. In general, all of the NOESY-like experiments were recorded for samples between 300 and 400 μM in the optimal buffer conditions (25 mM d₁₁-Tris·HCl, pH 7.1/150 mM NaCl/10 mM d₁₀-DTT) at 298 K in 600- (with cryoprobe) and 800-MHz spectrometers. For the Tyr reverse-labeled sample, an additional set of 2D NOESY experiments was recorded at various temperatures to analyze the changes in chemical exchange regime of the ring resonances as a function of temperature.

Amide Proton-Exchange Experiments.

¹⁵N-labeled p53 core domain was buffer-exchanged into D₂O buffer (25 mM phosphate, pH 7.2/150 mM NaCl/5 mM DTT) using a NAP-column (Amersham Pharmacia) resulting in a final concentration of ≈280 μM and placed immediately into an NMR tube and into the magnetic field. A series of ¹H-¹⁵N fast HSQC spectra (23) were acquired on a DRX600 spectrometer (Bruker, Karlsruhe, Germany) with a TXI cryoprobe over 4 d with further experiments after 10 and 20 d, respectively. The middle of the acquisition time for each spectrum was taken as the time base for the decay of the signals. The volume integrals for each crosspeak were integrated, and the decays were fitted to a single exponential decay. Protection factors (P) for backbone amide were extracted according to the formula P = k_int/k_ex^obs, where k_int was obtained from elsewhere (24).

Structure Calculation.

Distance constraints were obtained from several 2D and 3D NOESY-type experiments in deuterated, reverse-protonated, and partially deuterated samples. The interproton distance constraints were classified into four categories corresponding to 1.8–2.8, 1.8–3.5, 1.8–4.5, and 1.8–5.5 Å, respectively. Crosspeaks from known secondary structures within the protein were used for calibration purposes. Backbone and ϕ angular constraints were determined from the ¹³C chemical shifts by using the program talos (25). The final ensemble of structures was calculated with the program cns (26). Thirty-six structures out of 50 were accepted where no distance violation was >0.3 Å and no angle violation was >0.5°. Structure calculation statistics are summarized in Table 1. The data are deposited in the Protein Data Bank (PDB ID code 2FEJ).

We thank Drs. Mark Bycroft, Mark Allen, and Andreas Joerger for helpful discussions. This work was supported by Cancer Research UK, the Medical Research Council, and European Community FP6 funding. H.T. was supported by a Boehringer Ingelheim Ph.D. scholarship. This publication reflects the authors’ views and not necessarily those of the European Community. The Community is not liable for any use that may be made of the information.