Factors governing loss and rescue of DNA binding upon single and double mutations in the p53 core domain

Abstract

The mutation of R273→H in the p53 core domain (p53-CD) is one of the most common mutations found in human cancers. Although the 273H p53-CD retains the wild-type conformation and stability, it lacks sequence-specific DNA binding, a transactivation function and growth suppression. However, mutating T284→R in the 273H p53-CD restores the DNA binding affinity, and transactivation and tumour suppressor functions. Since X-ray/NMR structures of DNA-free or DNA-bound mutant p53-CD molecules are unavailable, the factors governing the loss and rescue of sequence-specific DNA binding in the 273H and 273H+284R p53-CD, respectively, are unclear. Hence, we have carried out molecular dynamics (MD) simulations of the wild-type, single mutant and double mutant p53-CD, free and DNA bound, in the presence of explicit water molecules. Based on the MD structures, the DNA-binding free energy of each p53 molecule has been computed and decomposed into component energies and contributions from the interface residues. The wild-type and mutant p53-CD MD structures were found to be consistent with the antibody-binding, X-ray and NMR data. The predicted DNA binding affinity and specificity of both mutant p53-CDs were also in accord with experimental data. The non-detectable DNA binding of the 273H p53-CD is due mainly to the disruption of a hydrogen-bonding network involving R273, D281 and R280, leading to a loss of major groove binding by R280 and K120. The restoration of DNA binding affinity and specificity of the 273H+284R p53-CD is due mainly to the introduction of another DNA-binding site at position 284, leading to a recovery of major groove binding by R280 and K120. The important role of water molecules and the DNA major groove conformation as well as implications for structure-based linker rescue of the 273H p53-CD DNA-binding affinity are discussed.

INTRODUCTION

Mutations of the tumour supressor protein, p53, are associated with ∼50% of human cancers (1) and 95% of lung cancers (2). They are also associated with resistance to therapy (3). The wild-type p53 protein is a key regulator in several cellular processes, including cell cycle control, DNA repair, programmed cell death, differentiation, senescence, angiogenesis and genome stability (4). Following DNA damage, or other genotoxic stress such as hypoxia, oncogene activation and viral infection (5,6), p53 promotes cell growth arrest and/or cell death, thus protecting the genome from accumulating excess mutations. The human p53 protein comprises of 393 amino acids that can be grouped into four functional domains/regions (4). The first 42 amino acids compose a transcriptional transactivation region. Residues 102–292 make up the sequence-specific DNA-binding core/central domain (p53-CD), for which there is a 2.2 Å X-ray structure of a 21 bp DNA duplex and three core domains; two of the core domains bind DNA, while the third does not (7). Residues 323–356 compose the tetramerisation domain for which there are NMR structures (8,9) and an X-ray structure (10). The last 30 residues form a basic regulatory region that can also bind to DNA or RNA, as well as act as a negative regulator of p53 sequence-specific DNA binding (11).

Most of the p53 mutations found in human cancers map to the core domain (12) and ∼40% of them occur at only six hot-spots (R175, G245, R248, R249, R273 and R282) (1). The p53-CD·DNA X-ray structure shows the core domain containing a sandwich of a four-stranded and a five-stranded antiparallel β sheet, and a loop–sheet–helix motif consisting of loop L1 (residues 113–123), a three-stranded β sheet (124–135 and 271–274) and helix H2 (278–286) (7). Residues from the loop–sheet–helix motif bind to the DNA major groove, and one of the two loops stabilised by Zn²⁺ (loop L3, residues 237–250) binds to the minor groove. In particular, two of the hot-spot residues, R248 and R273, whose side chains are within 4 Å of phosphate oxygens in the minor and major DNA groove, respectively, make critical contributions to DNA binding. The other four hot-spot residues stabilise the p53 DNA-binding surface but do not contact DNA directly. Based on the p53-CD·DNA X-ray structure and biochemical data, Cho et al. (7) classified the tumour-derived p53 mutants as conformational if they involve residues important for structural integrity but do not directly contact DNA (e.g. R175, G245, R249 and R282), or contact if they involve residues that directly contact DNA (e.g. R248 and R273). Failure of p53 conformational mutants to bind DNA in a sequence-specific manner has been attributed to structural defects, ranging from small structural shifts to local/global unfolding of the core domain. Failure of p53 contact mutants to act as a sequence-specific transactivator has been attributed to loss of critical DNA contacts.

There is considerable interest in the contact mutants because mutations at R248 and R273 account for almost one-fifth of all p53 mutations (1). Here, we have chosen to study the mutation of R273→H in the p53-CD since this is one of the most common mutations found in human cancer (7). The 273H p53 mutant retains the wild-type conformation (13) (i.e. PAb 1620+, PAb 240–, Hsp 70–) and stability (2) and is estimated to be 98% folded at physiological temperature, 37°C (14). However, the mutant 273H p53-CD lacks sequence-specific DNA binding (14) and sequence-specific transactivation function and growth suppression (15). In contrast, the mutation of T284→R significantly enhances not only the DNA binding of the wild-type protein with the last 30 residues deleted (referred to as p53Δ364–393), but also the corresponding 273H mutant (16). It can also rescue the DNA-binding affinity and specificity as well as transactivation and tumour suppressor functions of the 273H mutant (16). Since the DNA-free and DNA-bound X-ray/NMR structures of the single mutant and double mutant p53-CD have not been reported, the structural and energetic bases for the loss of DNA binding by the 273H p53-CD and the rescue of sequence-specific DNA binding by mutating T284→R in the single mutant are not clear.

Here, our goals are 2-fold. The first is to identify the factors governing the loss of DNA binding upon mutating R273 to a positively charged His in the p53-CD. The second is to elucidate the factors governing the rescue of sequence-specific DNA binding upon introducing a positively charged Arg at 284 in the mutant 273H p53-CD (see above). First, the three-dimensional (3d) solution models of structures of the wild-type, single mutant and double mutant p53-CD, free and DNA bound, were obtained using molecular dynamics (MD) simulations with explicit water molecules. Based on the MD structures, the free energies of wild-type and mutant p53 molecules binding to DNA were computed and decomposed into component energies and contributions from individual residues. To assess the validity of the theoretical results obtained, the findings were verified by their consistency with available experimental data. The energetic and structural analyses help to elucidate the mechanism of p53-CD reactivation, which in turn would aid the search for suitable ligands that may restore the lost interactions with DNA.

MATERIALS AND METHODS

MD simulations

The core domains of wild-type p53, mutant 273H and 273H+284R p53, free and bound to a 15 bp DNA, were subjected to MD simulation using the CHARMM (17) program at a mean temperature of 300 K. The simulations of p53-CD, 273H p53-CD and 273H+284R p53-CD included a total of 24802, 24358 and 25061 atoms, while the DNA-bound simulations included 58389, 58416 and 58417 atoms.

Starting structures. The initial 3d structure for the wild-type p53-CD·DNA simulation was taken from chain B of PDB entry 1TSR, in which the core domain (residues 96–289) is bound to a consensus DNA binding site, GGGCA AGTCT (7). The DNA strands were truncated to provide a three-base overhang of the binding region, yielding a 15 bp DNA duplex, 5′-³AATTGGGCAAGTCTA¹⁷-3′. Initial structures for the mutant p53-CD·DNA simulations were prepared by a side chain rotamer library method (18,19), as described in our previous work (20). Starting structures for the DNA-free simulations were extracted from the averaged MD structure of each complex.

Protonation states. Each simulation was carried out at physiological pH (pH 7) with the following protonation state assignments for the ionisable residues. The three Cys residues that are coordinated to zinc and all Asp, Glu residues and C-terminal groups were deprotonated. The other Cys and all Arg, Lys residues and N-terminal groups were protonated. The His that is coordinated to zinc was deprotonated as well as H214 because of its close proximity to R174. The other His residues were protonated since they are exposed to solvent or in close proximity to carboxylate side chains. In particular, the 273H mutant side chain is close to the negatively charged D281 carboxylate and Thy 11′ phosphate. The protonation state of 273H was verified by calculations (data not shown) showing that of all the His residues in p53, the 273H side chain nitrogens have the most negative electrostatic potential, indicating the highest probability of being protonated. With the inclusion of 20 counterions, the net charge of the solvated Zn²⁺-bound p53-DNA-ion system is –3.

Forcefield and boundary conditions. The simulations employed the all-hydrogen CHARMM forcefield (21) and the TIP3P model of water (22). The vdW interactions were truncated at 15 Å by a potential-shifting function, while the electrostatic interactions were switched from 11 Å to zero at 15 Å by an atom-based force-switching function. The non-bonded interaction list was updated using a heuristic test at a 16-Å cutoff. A spherical boundary potential was applied to the surface water molecules to account for interactions with the implicit bulk solvent (23).

Simulation protocol. The simulation procedure used is described in detail elsewhere (Noskov,S.Y., Wright,J.D. and Lim,C., in preparation.). The fully solvated protein–DNA complex was equilibrated for 60 ps with heavy constraints on only the terminal DNA residues to prevent the DNA strands from unwinding (24). These constraints were reduced and each simulation was continued for 250 ps, during which coordinates were saved every 100 fs.

In the X-ray structure, the root mean square deviation (RMSD) of Cα atoms in the p53-CD·DNA from those in the DNA-free p53-CD is only 0.75 Å, indicating that DNA binding does not alter the protein structure significantly (7). Consequently, the solvated p53 protein was extracted from the average MD structure of each complex and subjected to 250 ps of MD simulation, assuming the same residue protonation state, forcefield and boundary conditions as for the bound-state simulations.

Free energy decomposition

Relative binding free energy. This was based on a thermodynamic cycle (25,26).

ΔG_gas

[p53]_gas + [DNA]_gas→[p53·DNA]_gas

↑–G_solv(p53)↑–G_solv(DNA)↓G_solv(p53·DNA) (Scheme 1)[p53]_sln + [DNA]_sln→ [p53·DNA]_sln

ΔG°

The standard free energy change (ΔG°) for the binding of the p53-CD to DNA in solution according to Scheme 1 is given by:

ΔG° = ΔG_gas + G_solv(p53·DNA) – G_solv(p53) – G_solv(DNA) 1

where ΔG_gas is the standard free energy change per mole for the non-covalent association of the p53-CD and DNA in the gas phase at 300 K, and –G_solv corresponds to the work of transferring the molecule in its solution conformation to the same conformation in the gas phase at 300 K.

The focus in this work is not ΔG° but ΔΔG°, the free energy difference between the wild-type and mutant p53-CD binding to DNA in solution, which is given by:

ΔΔG° = (ΔG_gas^mut – ΔG_gas^wt) + [G_solv(p53^mut·DNA) – G_solv(p53^wt·DNA)] – [G_solv(p53^mut) – G_solv(p53^wt)] 2

The relative free energy (equation 2) was based on ‘snapshot’ configurations taken every 5 ps from the last 100 ps of the MD trajectories (27,28). All water molecules were removed from the MD trajectories to avoid boundary problems for the different sized systems as the DNA-bound and free proteins had the same number of solute atoms but different numbers of solvent atoms (29). The calculations of the terms in equation 2 are outlined below, and further details can be found in our previous work (26).

Gas-phase binding free energy difference, ΔΔG_gas. The free energy difference between wild-type and mutant p53-CD binding to DNA in the gas phase was estimated by:

ΔΔG_gas ∼ ΔΔE_gas^elec + ΔΔE_gas^vdW = (ΔE_gas^elec,mut – ΔE_gas^elec,wt) + (ΔE_gas^vdW,mut – ΔE_gas^vdW,wt) 3

Equation 3 assumes similar changes in conformation and vibrational entropy between wild-type and mutant p53-CDs binding to DNA. Based on the MD structures of wild-type and mutant p53-CDs (see above), the changes in vdW and electrostatic energies upon binding were calculated using a dielectric constant of one without cut-offs. The energies were computed using the CHARMM program (17) and forcefield (30).

Solvation free energy, G_solv. The solvation free energy, G_solv, can be expressed as a sum of three terms:

G_solv = G_solv^cav + G_solv^vdW + G_solv^elec 4

The first two terms in equation 4 were approximated by a linear function of the solvent accessible surface area (SASA) (31):

G_solv^cav + G_solv^vdW = (γ ^cav +γ^vdW) × SASA 5

γ ^vdW was set to –38.8 cal/mol/Ų (32), while γ ^cav was set to 46.0 cal/mol/Ų (33). The absolute SASA of the free or DNA-bound molecule was computed using the GEPOL program (34) and a solvent probe radius of 1.4 Å. The electrostatic contribution to the solvation free energy (G_solv^elec) was calculated using the finite-difference Poisson–Boltzmann method (35), as described in detail elsewhere (Noskov,S.Y., Wright,J.D. and Lim,C., in preparation).

RESULTS

For each simulation, the RMSD of the protein backbone atoms from the initial coordinates during the simulation (data not shown) shows that it initially rises and then plateaus, fluctuating around means between 1.2 and 1.4 Å during the last 150 ps of the simulation. Hence, the earlier part of each trajectory was included in the equilibration phase, while the final 150 ps were employed in generating an average structure, and in analyses such as computing the average distance between a light/heavy donor and heavy acceptor (see below). The p53 backbone RMSDs are similar to those found in other protein simulations (36–39).

Comparison of MD and X-ray p53 structures

To establish the reliability of the simulation results, hydrogen bonds (defined by a heavy donor–heavy acceptor distance ≤3.5 Å, light donor–heavy acceptor distance ≤2.5 Å and a deviation of less than ±60° from linearity) made by interface residues (defined by gas-phase interaction energy changes >1 kcal/mol upon binding to DNA) in the p53-CD·DNA simulation were compared to those in the X-ray structure. The hot-spot residues, R175, G245, R249 and R282, which do not contact DNA directly but are believed to play key structural roles in p53, were also used to verify the integrity of the simulation structures of the free and DNA-bound p53-CD. Among the 26 interface residues (see Tables 1 and 2), K120, S241, R273, A276 and R283 are within hydrogen-bonding distance of the major groove phosphate groups, K120, C277 and R280 are within hydrogen-bonding distance of the major groove bases in the X-ray structure, while the R248 side chain has ‘close’ contacts (defined by a heavy donor–heavy acceptor distance >3.5 Å but ≤4 Å) to the Thy 12′ and Thy 14 phosphate oxygens in the minor groove.

Table 1. Changes in DNA-binding free energy components of the mutant 273H p53-CD relative to wild type^a.

^aValues >5.5 kcal/mol are highlighted in larger font.

^bFree energy contributions from all interface residues.

Table 2. Changes in DNA-binding free energy components of the 273H+284R p53-CD relative to the single 273H mutant.

^aValues >5.5 kcal/mol are highlighted in larger font.

^bFree energy contributions from all interface residues.

Protein–DNA interactions. Among the residues that interact with the DNA in the X-ray structure, R280 makes the most number (six) of contacts (not necessarily hydrogen bonds) within 4 Å of base atoms in the major groove, followed by K120 with five contacts to base or phosphate atoms. R248 has three contacts ≤4 Å of a phosphate or sugar oxygen in the minor groove. S241, R273 and A276 each have two contacts ≤4 Å of the phosphate oxygen, while R283 has one. The C277 side chain is within hydrogen-bonding distance of the Cyt 9′ base N⁴. All the ≤4 Å DNA contacts found in the X-ray structure are preserved in the simulation except the following.

The ‘close’ contact between R283 N^η2 and Gua 7 O^1P (3.53 Å in the X-ray structure) is extended to 4.68 ± 0.41 Å in the simulation. However, the R283 side chain has a high average X-ray B-factor (69 Ų), indicating a flexible side chain that may not be able to maintain a stable salt bridge with the Gua 7 phosphate oxygen. In the major groove the R273 N^η2 is <3.20 Å to two Thy 11′ phosphate oxygens in the X-ray structure, but is 3.95 ± 0.40 Å to Thy 11′ O^5′ and 3.91 ± 0.37 Å to Gua 10′ O^1P in the simulation, while the R280 N^η1…Gua 9 O⁶ and R280 N^η2…Gua 10′ O⁶ ‘close’ contacts in the X-ray structure are attenuated by an intervening water molecule (Fig. 1). New hydrogen bonds are formed between a R280 amino proton and the D281 carboxylate oxygen (Fig. 1) as well as between the K120 backbone oxygen and the G279 and R280 backbone amide protons in the simulation (not shown in Fig. 1). In the minor groove R248 N^η1 is closer to Thy 12′ O^3′ in the simulation (3.41 ± 0.37 Å) than in the X-ray structure (3.83 Å) but it does not form a hydrogen bond (R248 H^η1…Thy 12′ O^3′ = 2.88 ± 0.58 Å), while it is in close contact with Thy 11′ O^2P (3.87 ± 0.55 Å) instead of Thy 14 O^3′ and O^4′ in the X-ray structure (3.83 Å).

Figure 1.

Hydrogen-bonding network in the average wild-type p53-CD·DNA structure in the vicinity of the major and minor grooves. Only certain key residues (discussed in text) and their hydrogen bonding partners (including specific DNA base/phosphate groups and water molecules) are shown for the sake of clarity. However, a water molecule hydrogen bonding to the R249 carbonyl oxygen could not be depicted. The loop–sheet–helix motif is in magenta while the L3 loop is in yellow. Oxygen, phosphate, hydrogen and nitrogen atoms are coloured red, orange, green and blue, respectively.

Protein–water–DNA interactions. Three water molecules form bridges between the protein and the DNA in the p53-CD·DNA simulation (Fig. 1). The first is a water that bridges the R273 side chain and the Gua 10′ phosphate. The second water bridge is between the R280 side chain and the Gua 9 base oxygen (see above). The third water bridge, which is also seen in the crystal structure, links the R248 side chain with the Thy 12′ sugar oxygen. Although the oxygen of this water is within hydrogen-bonding distance to the Gua 13 N (3.35 ± 0.34 Å), the water hydrogen to Gua 13 N distance (2.66 ± 0.44 Å) indicates a non-ideal hydrogen bond.

Protein–protein interactions. All the hydrogen bonds made by the four hot-spot ‘structural’ residues, R175, G245, R249 and R282, in the free and DNA-bound p53-CD X-ray structure are preserved in the simulation except the following. The R175 N^η1…D184 O^δ2 salt bridge in the p53-CD X-ray structure is absent from the DNA-bound crystal structure and in the DNA-free simulation. In the p53-CD·DNA X-ray structure, the R175 N^η1 is 2.59 Å to S183 O^γ and 3.41 Å to P191 O, but in the simulation the hydrogen bond to S183 O^γ is maintained (3.10 ± 0.26 Å), but that to P191 O is unstable (4.56 ± 1.13 Å). In the p53-CD·DNA crystal structure the R282 amino nitrogens (N^η) hydrogen bond to Y126 O and T118 O^γ, while its imino nitrogen (N^ɛ) makes bifurcated hydrogen bonds to the E286 and S127 side chain oxygens. However, in the simulation the R282 N^η hydrogen bonds to T118 O^γ but not to Y126 O, while its N^ɛ breaks and forms hydrogen bonds with the E286 carboxylate oxygens (average R282 N^ɛ…E286 O^ɛ distance = 3.75 ± 0.96 Å).

Dynamical fluctuations. The B-factors (data not shown) averaged over the backbone atoms, N, Cα and C(=O), in the p53-CD·DNA simulation generally reproduce the qualitative features of the X-ray main-chain B-factors with the largest discrepancy found in two loops; the L2 loop (residues 164–194) and the loop connecting β-strands S7 and S8 (residues 220–229). Apart from these two loops, the computed average B-factors are generally lower than the X-ray values, as found and rationalised in previous studies (36,39–41). The RMSD of the DNA backbone atoms (P, O^5′, C^5′, C^4′, C^3′ and O^3′) in the wild-type simulation from the corresponding X-ray structure (data not shown) shows that the major groove region remained relatively stable with an average RMSD of 1.4 Å for base pairs 6–10, while the minor groove was more flexible with an average RMSD of 1.9 Å for base pairs 12–14. The largest deviation was found for base pair 11 and 11′ (RMSD = 2.55 Å).

Changes in component free energies upon mutation

In the following, we will first present the free energy decomposition results for the 273H p53-CD single mutant, then the 273H+284R p53-CD double mutant. The contributions to ΔΔG° (see Material and Methods) from the protein–DNA interface residues were decomposed according to the electrostatic, vdW and cavity components:

6

The summation in equation 6 is over all interface residues listed in Tables 1 and 2. The net electrostatic and vdW contributions were further broken down into protein–solvent (denoted by subscript ‘solv’) and protein–protein or protein–DNA interactions (collectively known as solute–solute interactions, denoted by subscript ‘gas’); i.e.

7a

7b

The individual terms in equations 6 and 7 for the 273H and 273H+284R p53 mutants are summarised in Tables 1 and 2, respectively.

In the following, a single Δ denotes the free energy difference between the DNA-bound and free wild-type or mutant p53-CD. A double ΔΔ represents the DNA-binding free energy of the single mutant 273H p53 mutant relative to the wild-type protein, whereas ΔΔ′ represents the binding free energy of the double mutant 273H+284R p53 relative to the single mutant. Interface residues with and changes in the corresponding residue–protein+DNA and residue–solvent interaction energies upon mutation were identified. The energy changes were correlated with changes in the respective residue–protein, residue–DNA and residue–solvent hydrogen-bonding interactions as well as solvent accessibility seen in the simulations. In interpreting the results, we emphasise the trend (sign) of the large free energy changes upon each mutation rather than their absolute values.

R273→H. When R273 is mutated to His, the p53-CD mutant is predicted to bind to DNA far less well than the wild-type protein (total ΔΔG° = 21.0 kcal/mol). Not surprisingly, the sum of the binding free energy contributions from all interface residues in the mutant is much less favourable than that in the wild-type protein Table 1). This is because the net solute–solute electrostatic and vdW contributions of the interface residues with DNA binding in the mutant are unfavourable compared with those in wild-type p53 and Table 1). The latter term is roughly cancelled by favourable solute–solvent electrostatic contributions Table 1). The remaining solute–solvent vdW and solvent–solvent cavity contributions of the interface residues to DNA binding in the mutant and wild-type protein are similar (Table 1).

The key residues contributing to the net loss of DNA binding across the protein–DNA interface are K120, R280 and D281 in the major groove region and R249 in the minor groove region (ΔΔG_i ≥ 5.5 kcal/mol for i = K120, R249, R280 and D281; Table 1). These four residues as well as the mutant 273H residue have less favourable electrostatic interactions with the protein and/or DNA in the p53 mutant compared with wild-type p53 (positive ΔΔG_i,gas^elec for i = K120, R249, 273H, R280 and D281; Table 1). The positive ΔΔG_gas^elec for the mutant 273H, R280 and D281 is due to the disruption of a hydrogen-bonding network involving the R273, D281 and R280 side chains, the Thy 11′ phosphate and the Gua 10′ base upon mutating R273 to His in the wild-type p53-CD (compare Figs 2 and 1). This has three manifestations. First, the observed loss of protein–protein and protein–DNA interactions in the major groove of the mutant complex is partially compensated by a gain in protein–solvent interactions: the D281 side chain gains three water hydrogen bonds, while the D281 and K120 backbone amide and the T284 side chain oxygen each gain one solvent hydrogen bond (Fig. 2). Second, the loss of protein–protein/DNA hydrogen bonds but gain in protein–solvent interactions in the major groove correlates with an increase in the backbone RMSDs of the major groove base pairs 6–10 and B-factors in the mutant complex simulation relative to the wild-type one (Fig. 3, dotted curves). Third, the behaviour of the RMSD as a function of base pair number in the mutant complex simulation differs from that in the wild-type X-ray structure; e.g., base pair 7 has the highest average RMSD in the mutant simulation but the lowest one in the wild-type X-ray structure (Fig. 3A). This finding, in conjunction with the high major groove backbone RMSDs in the mutant simulation, indicates a change in the major groove conformation (but not its width) in the 273H p53 complex.

Figure 2.

Hydrogen-bonding network in the average 273H p53-CD·DNA structure in the vicinity of the major and minor grooves. See also the legend to Figure 1. The T284 H^N…R280 O and D281 H^N…O (water) hydrogen bonds are not shown for the sake of clarity.

Figure 3.

(A) The RMSD of the DNA backbone atoms (averaged over base pair) from the wild-type simulation structure as a function of base pair number for wild-type crystal structure (solid curve), 273H p53 (dotted curve) and 273H+284R p53 (dashed curve). (B) The B-factors averaged over the protein backbone atoms as a function of residue number in the simulations of DNA bound to wild-type p53 (solid curve), 273H p53 (dotted curve) and 273H+284R p53 (dashed curve).

The distortion of the major groove conformation (in particular, base pairs 7–9) in the mutant complex is likely to cause K120 to lose three hydrogen bonds to the DNA; one from its mainchain nitrogen to the phosphate oxygen of Gua 7, and two from its side chain nitrogens to the base atoms of Gua 8 and Gua 9 (compare Figs 2 and 1), hence ΔΔG_K120,gas^elec is positive. The positive ΔΔG_gas^elec for R249 is due to the loss of salt bridges between the R249 and E171 side chains upon mutating R273 to His (compare Figs 2 and 1).

Although the mutation of R273 to His results in the loss of direct charge–charge interactions with the DNA phosphate group and D281 carboxylate side chain (Fig. 2), the mutant 273H contributes more favourably than the wild-type Arg to DNA binding (negative ΔΔG_273H; Table 1). This is due to the large, negative ΔΔG_273H,solv^elec (–18.2 kcal/mol), which more than compensates for the positive ΔΔG_273H,gas^elec (3.6 kcal/mol; Table 1). The negative ΔΔG_273H,solv^elec is because the mutant His has a smaller desolvation cost upon DNA binding compared with the wild-type Arg (151.4 versus 169.6 kcal/mol; Table 3).

Table 3. Electrostatic contributions to the DNA-binding free energy of wild-type and mutant p53-CD.

^aNumbers in the top row are for the free state, those in the row below are for the DNA-bound state.

^b273H p53-CD. ^c273H+284R p53-CD.

In addition to the mutant residue, the net binding free energy contributions of V122, R248 and R283 in the p53 mutant are more favourable than those in wild-type p53 (negative ΔΔG_i for i = V122, R248 and R283; Table 1). Unlike 273H, these three residues have more favourable electrostatic interactions with the protein and/or DNA in the mutant compared with the wild-type protein (negative ΔΔG_i,gas^elec for i = V122, R248 and R283; Table 1). The negative ΔΔG_gas^elec for V122, R248 and R283 is mainly due to hydrogen bonds formed in the mutant DNA simulation that are absent from the X-ray and MD structures of the wild-type complex. The interactions exclusive to the 273H p53 complex correspond to V122 N…K120 O (3.16 ± 0.20 Å), V122 O…Q136 N^ɛ2 (3.03 ± 0.32 Å) and R283 O…N288 N (3.03 ± 0.18 Å) hydrogen bonds, as well as salt bridges between the R248 guanidinium side chain and the Thy 12′ base oxygen (Fig. 2). The negative ΔΔG_gas^elec for R248 and R283 are partially offset by positive ΔΔG_solv^elec contributions (Table 1) probably due to the hydrogen bonds formed by these two residues, which reduce their solvent accessibility (Table 4), and hence their solvation free energies in the DNA-bound state become more unfavourable (Table 3).

Table 4. %SASA in wild-type and mutant p53-CD, free and DNA bound^a.

^aPercentage ratio of the water-accessible surface area of the side chain X in the protein to the accessible surface area of X in the tripeptide -Gly-X-Gly- using the MolMol program (53) where <20% indicates buried, between 20 and 50% indicates partially buried and >50% indicates solvent exposed (54).

^bNumbers are based on the X-ray structure (7).

^cNumbers are based on the average MD structures.

^d273H p53-CD.

^e273H+284R p53-CD.

R273→H+T284→R. When T284 is mutated to Arg in the 273H p53-CD, the DNA binding ability of the double mutant becomes comparable to the wild-type p53-CD (total ΔΔG° = –2.8 kcal/mol), and the sum of the binding free energy contributions from all interface residues in the double mutant is similar to that in the wild-type protein . The improved DNA binding affinity of the 273H+284R p53-CD relative to the 273H p53-CD is mainly because the net solute–solute electrostatic contribution of the interface residues to DNA binding in the double mutant is more favourable than that in the single mutant although it is partially cancelled by an unfavourable solute–solvent electrostatic free energy (Table 2). The net non-electrostatic solvation free energy contribution of the interface residues to DNA binding in the double mutant is similar to that in the single mutant as the positive solute–solvent vdW free energy is compensated by a negative solvent–solvent cavity term (Table 2).

The net free energy contribution of the 284R mutant residue to the improved DNA binding affinity of the double mutant is not significant (ΔΔ′G_284R, = –1.5 kcal/mol; Table 2). Although the 284R side chain forms a salt bridge with the Thy 12′ phosphate oxygen (Fig. 4) resulting in a large negative ΔΔ′G_284R,gas^elec (–192.8 kcal/mol; Table 2), the latter is offset by an even larger but positive ΔΔ′G_284R,solv^elec (195.7 kcal/mol; Table 2) due to the greater desolvation penalty of the positively charged 284R upon binding DNA compared with neutral T284 in the single mutant (207.6 versus 12.0 kcal/mol; Table 3). The net vdW contribution of the 284R mutant residue to the improved DNA binding is negligible (ΔΔ′G_284R^vdW = –0.1 kcal/mol; Table 2). In contrast, the solvent–solvent cavity ΔΔ′G_284R,solv^cav term is favourable (–4.2 kcal/mol; Table 2) due to the large SASA change at position 284 upon DNA binding (ΔSASA = –70 Ų for 284R compared with –1.8 Ų for T284 in the single mutant using the average MD structures).

Figure 4.

Hydrogen-bonding network in the average 273H+284R p53-CD·DNA structure in the vicinity of the major and minor grooves. See also the legend to Figure 1. As for the wild-type and 273H p53-CD·DNA simulations, a 284R H^N…R280 O hydrogen bond is not shown for the sake of clarity.

Instead of the mutant 284R, the residues that contribute to the loss of DNA binding upon mutating R273 to His in the wild-type p53-CD [namely K120, R249, C277 and R280 (see above and Table 1)] are now responsible for the gain in DNA binding upon mutating T284 to Arg in the 273H p53-CD (ΔΔ′G_i negative and │ΔΔ′G_i│ >6 kcal/mol for i = K120, R249, C277 and R280; Table 2). The negative ΔΔ′G_i values for these four residues stem mainly from the negative ΔΔ′G_i,gas^elec (Table 2), which is due to the restoration of hydrogen bonds formed by these four side chains in the wild-type protein (compare Figs 4 and 1). In the 273H+284R p53-CD·DNA simulation, the K120 and R280 side chains hydrogen bond to Gua 8, Gua 9 and Gua 10′ base atoms in the major groove (Fig. 4), the R249 and E171 side chains form salt bridges (average distance = 3.08 ± 0.39 Å) in the minor groove, and C277 S^γ is close (4.00 ± 0.46 Å) to Cyt 9′ N⁴ (not shown in Fig. 4). These interactions are lost in the single mutant complex simulation, but are present in the wild-type complex simulation and/or X-ray structure. However, the negative ΔΔ′G_i,gas^elec values for K120, R249 and R280 are partially offset by positive ΔΔ′G_solv^elec (Table 2) because these three positively charged residues are better solvated in the DNA-free double mutant than in the unbound single mutant, hence their desolvation penalties in the double mutant are greater than those in the single mutant (Table 3).

The creation of a new hydrogen bond to the DNA phosphate by mutating T284→R in 273H p53 and the restoration of native hydrogen bonds to the major groove bases by K120, C277 and R280 has three other manifestations. First, the observed gain in protein–DNA and protein–protein interactions in the major groove of the 273H+284R p53 complex is at the expense of protein–solvent interactions. Relative to the single mutant complex, 273H, C277 and D281, respectively, lose 2, 1 and 1 water hydrogen bonds (see also %SASA in Table 4). Second, the recovery of protein–DNA/protein–protein hydrogen bonds and loss of protein–solvent interactions in the major groove correlates with a decrease in the backbone RMSDs of the major groove base pairs 6–10 and B-factors in the double mutant complex simulation compared with the single mutant one (compare dashed and dotted curves in Fig. 3). Third, the behaviour of the RMSD versus base pair number in the double mutant complex simulation is similar to that in the wild-type X-ray structure for the major groove base pairs (compare dashed and solid curves in Fig. 3A), suggesting similar major groove conformation in the double mutant and in the wild-type complex.

The residues that contribute favourably to DNA binding in the 273H p53-CD relative to the wild-type protein (namely R248, 273H and R283) now contribute unfavourably to DNA binding in the double mutant compared with the single mutant. The positive ΔΔ′G _R248 and ΔΔ′G_273H stems mainly from the positive ΔΔ′G_i,gas^elec (Table 2). In the case of R248, this is probably because upon mutating T284 to Arg in the 273H p53-CD a direct hydrogen bond from the R248 side chain to the Thy 12′ base oxygen in the single mutant complex (Fig. 2) is screened by a water in the double mutant complex (Fig. 4), which shields the electrostatic interactions. Although the 273H side chain forms a salt bridge with the D281 side chain in the double mutant (Fig. 4) but not in the single mutant (Fig. 2), its interaction with the DNA phosphate backbone is attenuated (273H N^ɛ…Thy 11′ O^P = 5.15 ± 0.86 Å compared with 4.14 ± 0.56 Å in the single mutant). Unlike R248 and 273H, the positive ΔΔ′G_R283 stems from a large, positive ΔΔ′G_283,solv^elec (16.2 kcal/mol; Table 2), which more than offsets the negative ΔΔ′G_R283,gas^elec term (–13.1 kcal/mol; Table 2).

D281 contributes unfavourably to DNA binding upon mutating R273 to His in the wild-type protein (ΔΔG_D281 = 6.9 kcal/mol; Table 1) as well as upon mutating T284 to Arg in the single mutant (ΔΔ′G_D281 = 11.1 kcal/mol; Table 2). The positive ΔΔ′G_D281 term is dominated by a positive ΔΔ′G_D281,gas^elec (8.7 kcal/mol; Table 2) due probably to the greater solvent accessibility of the major groove in the single mutant compared with the double mutant. D281 hydrogen bonds to four and two water molecules in the single mutant and double p53 mutant complexes, respectively. The extra water molecules in the single mutant helps to shield the negatively charged D281 side chain from the negatively charged phosphate backbone, thus reducing the repulsive electrostatic interactions in the single mutant relative to the double mutant complex.

DISCUSSION

Comparison with available experimental data

Wild-type structure. A comparison of the hydrogen bonds in the simulation and X-ray structures of the free and DNA-bound p53-CD shows that most of the hydrogen bonds observed in  the crystal structure are preserved in the simulation (see  Results). Although minor differences between the p53-CD·DNA X-ray and simulation structures are observed for the side chain interactions of R248, R273, R280, R282 and R283, they do not seem to be inconsistent with the relatively high average side chain X-ray B-factors (∼30 Ų for R273 and R280, 46 Ų for R248 and R282, and 69 Ų for R283). Thus, the free and DNA-bound p53-CD simulations have preserved the key protein–protein (and protein–DNA) contacts as well as the qualitative features of the X-ray B-factors.

Mutant structures. Although X-ray/NMR structures of the single mutant and double mutant p53-CD, free or DNA bound, are not available for comparison with the respective simulation structures, we can employ the fact that 273H p53 is recognised by the monoclonal antibody PAb 1620, but not by the monoclonal antibody PAb 240 to verify the mutant structure. PAb 1620 is specific for the wild-type conformation and particularly recognises R209 and N210 (42), while PAb 240 is specific for the ‘denatured’ conformation and recognises residues 212–217 on the S7 β-strand (43). In the average 273H p53 structure, the average %SASA for R209 and N210 (46%; Table 4) indicates that they are solvent exposed, and can thus be recognised by PAb 1620, while the average %SASA for residues 212–217 (22%; Table 4) indicates that they are relatively buried, and are thus inaccessible to PAb 240. In analogy to the 273H p53 mutant, the average %SASA values for residues 209–210 (58%) and 212–217 (13%) in the average 273H+284R structure are similar to those in the respective wild-type structure (58 and 16%), indicating a wild-type conformation for the double mutant (Table 4).

A NMR study of the free 273H p53-CD has shown that structural changes in the mutant protein are found mainly in the loop–sheet–helix motif and the L3 loop, in particular residues D281, R282, E285 and E286 (44). This is consistent with the change in the hydrogen-bonding network involving these four residues in the simulation of the 273H p53-CD relative to the wild-type protein. Upon mutating R273 to His in the DNA-free p53-CD, the D281 carboxylate oxygens lose salt bridges to the 273H and R280 side chains, whereas the R282 and E286 side chains form salt bridges while the E285 carboxylate group forms new hydrogen bonds with K132. Thus, the simulation of the free 273H p53-CD seems to be consistent with both the antibody-binding and NMR data.

DNA binding affinity. Sequence-specific DNA binding to activate transcription of genes containing p53-binding DNA sites appears to be essential for normal p53 function (45,46). Such p53-binding DNA sites contain a repeat of two consensus pentamer sequences: 5′-Pu-Pu-Pu-C-(A/T) (T/A)-G-Py-Py-Py-3′, where Pu represents a purine (Gua or Ade) and Py denotes a pyrimidine (Cyt or Thy) (47). Clearly, the binding affinity (ΔG°) of wild-type p53 for DNA depends not only on the DNA sequence, but also on the p53 molecule itself (i.e. if it is full length, minus the last 30 amino acids, or only the core domain) as well as on the temperature (46). For the consensus site studied here, ⁷GGGCA^{11 12}AGTCT¹⁶, DNA binding specificity stems from hydrogen bonding interactions between the side chains of K120, C277 and R280 and the base oxygen or nitrogen of Gua 8, Gua 9, Cyt 9′ and Gua 10′ in the major groove, while non-specific binding affinity is due to salt bridges between K120, S241, R273, A276 and R283 and the DNA phosphate. Note that K120 and C277 could also hydrogen bond to proton-acceptor base atoms of Ade 8 and Thy 9′, respectively, consistent with the variability of the base sequence at these two positions of the consensus site (7).

The calculations here predict that the DNA binding affinity of the mutant 273H p53-CD is much weaker than that of its wild-type counterpart (positive ΔΔG°) and that it may not be detectable experimentally, as indicated by the relatively large magnitude of ΔΔG° (21 kcal/mol). Furthermore, they predict that most of the lost interactions are between the K120, C277 and R280 side chains and the DNA major groove bases, in particular the invariant Gua 10′, suggesting a loss of DNA binding specificity in the single mutant. These results are consistent with experiments showing that at 25°C the proteolytically excised core domain of wild-type, but not 273H p53, can bind specifically to a synthetic 46 bp DNA containing a site, 5′-TGGCA AGCCT-3′, similar to the consensus DNA site used here (except at the underlined positions) (15). They are also consistent with experiments showing that the p53Δ364–393 and the p53-CD (at 20°C) bind to the gadd45 oligonucleotide containing the consensus site, 5′-GAACA TGTCT-3′, but the respective proteins with R273 mutated to His do not (14,16).

The calculations also predict that the DNA binding affinity and specificity of the 273H+284R p53-CD is similar to that of the wild-type and is enhanced relative to the single mutant (see Results). This is consistent with experiments showing that the 273H+284R p53Δ364–393 binds to the gadd45 oligonucleotide, exhibited wild-type transcriptional activity and suppressed proliferation of tumour cells.

Factors affecting loss of DNA binding upon mutating R273→H in p53

In previous work (7), the 273H mutant was believed not to bind significantly to DNA due to the loss of the hydrogen bond between the wild-type R273 side chain and the Thy 11′ phosphate upon mutation. In contrast, the calculations show that although the mutant H273 does not hydrogen bond to the Thy 11′ phosphate as anticipated, it actually contributes to DNA binding more favourably than R273 in wild-type p53. This is because the mutant His has a smaller desolvation cost upon DNA binding compared with the wild-type Arg (ΔG_273H,solv^elec < ΔG_R273,solv^elec; Table 3). The calculations suggest that the loss of the salt bridge from the wild-type R273 to the D281 side chain (rather than to the Thy 11′ phosphate) probably triggers the loss of DNA binding by the mutant 273H p53-CD (see below).

The lower DNA binding affinity of the mutant 273H p53-CD stems from the disruption of a hydrogen-bonding network involving R273, D281 and R280 in wild-type p53, leading to a loss of major groove binding by R280 and K120, which is partly compensated by a gain in minor groove binding by R248. When R273 is mutated to His, the shorter imidazolium side chain loses a close contact to the Thy 11′ phosphate, and also a salt bridge to the D281 side chain, which in turn loses hydrogen bonds to the R280 guanidinium group (compare Figs 2 and 1). Consequently, R280 is not optimally oriented to bind into the major groove and loses its native hydrogen bonds with the invariant Gua 10′ base. The loss of these protein–protein and protein–DNA hydrogen bonds in the 273H mutant complex is accompanied by (i) an overall increase in the flexibility of the mutant p53 (dotted curves in Fig. 3), (ii) an increase in the solvent accessibility of the major groove (Table 4) and (iii) a change in the major groove conformation (compare dotted and solid curves in Fig. 3A), resulting in a loss of DNA contacts by K120 and a gain in DNA contacts by R248 at the expense of its neighbour R249, which loses salt bridges with the E171 side chain (Fig. 2). The net effect is a loss of favourable protein–protein and protein–DNA electrostatic positive) and vdW positive) interactions as the protein–DNA interface becomes less well packed in the 273H p53 complex (Table 1).

Thus, the calculations suggest a central role for D281 in helping to orient R273 and R280 for hydrogen bonding to the DNA, consistent with the observation that D281 is strictly conserved across 27 species that p53 has been sequenced for (48). Mutation of D281 to Asn or the charge-preserving Glu in p53 has been found in tumour cells, implying that the mutant cannot bind to DNA specifically and function as a tumour suppressor gene. Furthermore, mutations of D281 to Tyr or Gly show a complete loss of DNA binding (49,50).

Factors governing rescue of DNA binding by mutating T284→R in 273H p53

In previous work (16), the mutant 273H+284R p53-CD was believed to restore DNA binding owing to the introduction of a positively charged residue in the vicinity of the DNA-binding interface that could establish novel protein–DNA contacts. As anticipated, the calculations show that the mutant 284R side chain forms a new hydrogen bond to the Thy 12′ phosphate in the double mutant complex simulation. However, the favourable electrostatic interactions between the 284R side chain and the Thy 12′ phosphate are offset by the higher cost of desolvating the positively charged 284R upon binding DNA compared with the neutral wild-type T284 (ΔG_284R,solv^elec >> ΔG_T284,solv^elec; Table 3). Consequently, the mutant 284R residue does not contribute as favourably as other residues (see below) to the improved DNA binding affinity of the double mutant relative to the single mutant (Table 2).

The introduction of another DNA-binding site at position 284 leads to a recovery of major groove binding by R280 and K120, and hence to a ‘wild-type’ DNA binding affinity and specificity of the double mutant. When T284 is mutated to Arg in the 273H p53-CD, the positively charged guanidinium side chain establishes a novel DNA contact that has the apparent effect of squeezing some water molecules out of the DNA binding surface (compare Figs 4 and 2; see also Table 4) and restoring the wild-type major groove conformation (compare dashed and solid curves in Fig. 3A). This enables R280 and K120 to regain native major groove contacts (compare Figs 4 and 1), which are essential for sequence-specific binding affinity. The net effect is a gain in favourable electrostatic and cavity free energies and negative; Table 2) in the double mutant compared with the single mutant.

Thus, the calculations suggest that the 284R mutant does not simply play a role by acting in lieu of R280 to bind to the DNA major groove in the 273H p53-CD. Rather, it plays a key role in helping R280 and K120 to regain their major groove contacts by helping to restore the ‘native’ conformation of the protein–DNA interface. Furthermore, since R280 and K120 contributes significantly to the improved DNA binding affinity of the double mutant relative to the single mutant (Table 2), a triple R273H+T284R+R280X p53-CD mutant would likely bind DNA with reduced affinity relative to the double mutant or wild-type p53-CD. This is consistent with the observation that the charge-preserving R280K p53 mutant is known not to bind to DNA (51), and that mutations of R280 to Lys, Ser, Thr, Ile, Gly and Pro in p53 are present in tumour cells (1).

Implications for rescue using small linker molecules

The mutant structures generated in this study can be used with molecular docking methods [such as those described in our previous work (52)] to screen databases of small molecule compounds to identify potential ligands that may restore the binding affinity of mutant p53-CD towards DNA. The energetic and structural analyses suggest that it may be possible to rescue the DNA binding affinity of the 273H mutant using an appropriate small ligand in the vicinity of the 284 site instead of the mutation site at position 273. The ligand should provide a short bridge to DNA in order to bring the DNA closer to the protein and force the removal of solvent water molecules from the DNA-binding interface. This may allow K120 and R280 to re-establish their native major groove contacts. The suitability of such a ligand could be verified using a combination of MD simulations and free energy decomposition calculations as presented in this work.