Hidden electrostatic energy contributions define dynamic allosteric communications within p53 during molecular recognition

Abstract

Molecular recognition is fundamental to transcription regulation. As a transcription factor, the tumor suppressor p53 has to recognize either specific DNA sequences or repressor protein partners. However, the molecular mechanism underlying the p53 conformational switch from the DNA-bound to repressor-bound states is not fully characterized. The highly charged nature of these interacting molecules prompted us to explore the nonbonded energy contributions behind molecular recognition of either a DNA or the repressor protein iASPP by p53 DNA binding domain (p53DBD), using molecular dynamics simulation followed by rigorous analyses of energy terms. Our results illuminate the allosteric pathway by which iASPP binding to p53 diminishes binding affinity between p53 and DNA. Even though the p53DBD uses a common framework of residues for recognizing both DNA and iASPP, a comparison of the electrostatics in the two p53DBD complexes revealed significant differences in residue-wise contributions to the electrostatic energy. We found that an electrostatic allosteric communication path exists in the presence of both substrates. It consists of evolutionarily conserved residues, from residue K120 of the binding loop L1 to a distal residue R213 of p53DBD. K120 is near the DNA in the p53DBD-DNA complex, whereas iASPP binding moves it away from its DNA binding position in the p53DBD-iASPP complex. The “energy hubs” (the residues show a higher degree of connectivity with other residues in the electrostatic networks) determined from the electrostatic network analysis established that this conformational change in K120 completely rewires the electrostatic network from K120 to R213, thereby impeding DNA binding. Furthermore, we found shifting populations of hydrogen bonds and salt bridges reduce pairwise electrostatic energies within p53DBD in its DNA-bound state.

Significance

The tumor suppressor p53 is implicated in an array of cellular functions; mutational alteration of p53 plays a critical role in most human cancers, and thus, controlling the stability of p53-ligand interactions could lead to cancer therapies. The connectivity of binding sites to distal allosteric sites is an emerging target of the allosteric drug design field. Understanding the molecular recognition patterns in terms of residue-wise electrostatic energies in complexes between the DNA binding domain of p53 (p53DBD) with either DNA or the iASPP repressor would provide important insight into drug target selection. We show that electrostatic energy contributions can explain the spatial and temporal events of allosteric communications within p53DBD during molecular recognition.

Introduction

Molecular recognition of and by transcription factors, one of the fundamental processes of genomic decoding, is accomplished by selectively stabilizing a distinct conformational state. The “guardian of the genome” p53, being a central hub in the molecular network of transcription in living cells, undergoes conformational transitions during the process of molecular recognition. p53 binds to specific DNA targets (at “response elements” or response element sequences) (1,2) and regulates gene expression for diverse biological functions, e.g., cell cycle arrest, DNA repair, and apoptotic pathways, in response to a myriad of stressors (3, 4, 5, 6). Apart from recognizing specific DNA sequences, p53 also interacts with various protein factors in a context-dependent manner during transcription regulation.

The central DNA binding domain of p53 (p53DBD; amino acids 93−293) is responsible for sequence-specific recognition of DNA (7). p53DBD consists of a β-sandwich scaffold that supports the loop-sheet-helix motif, which interacts with the major groove of the DNA through hydrogen bonding, whereas the L3 loop (residue Arg248) contacts the minor groove as well as the DNA backbone (8). Major groove interactions include Lys120 with G8, Cys277 with C9, and Arg280 to G10. DNA binding residue R280 in the H2 helix (belonging to loop-sheet-helix) and residue K120 of the L1 loop (9) are involved in target selection, and acetylation of the DNA binding residue K120 contributes to the p53 promoter specificity (10, 11, 12, 13, 14, 15). Evidently, charged residues of the p53DBD are primarily responsible for DNA binding.

The crystal structure of p53DBD in complex with the iASPP (inhibitory member of the apoptosis stimulating protein of p53 family protein, which, depending on cellular requirements, silences the DNA binding activity of p53) (14) revealed that p53 also binds the repressor protein iASPP, and the L1 loop of the binding interface is displaced in such a way that the DNA binding residue K120 is no longer accessible to the DNA, although iASPP binding does not block the major DNA binding surface of the p53DBD (14).

Strikingly, residue R213, at a position remote from the active site of p53DBD, is a mutational hot spot (16). The R213Q mutation is found in certain tumor cell lines, and methylation of R213 has also been reported (17). Mutations of R213 affect the interactions of different DNA and proteins with the binding interface of p53, and the mutations prevent the trans activation of the p21 gene (18). Surprisingly, overlaying the backbones of the crystal structures of p53DBD complexes with either DNA or iASPP does not show any substantial conformational changes in p53DBD, except in the L1 loop at the binding site. Thus, modulation of the p53DBD binding affinity by mutation of a distal residue suggests the involvement of “dynamic allostery” (i.e., allostery without showing significant conformational alterations) (17,19).

A recent study found that dynamic allostery in a PDZ3 domain protein does not result from an entropic effect alone, as traditionally believed, but rather that modulation in the internal electrostatic energy profile also contributes significantly to the process (20). DNA and iASPP are highly negatively charged, whereas arginine-rich p53 is positively charged, also suggesting the involvement of electrostatic energy contributions to the molecular recognition process. However, the nonbonded energy contributions to the molecular-recognition-mediated conformational switch in p53 have not been reported thus far.

To illuminate the allosteric regulation in p53-mediated molecular recognition, here we have used molecular dynamics to study the complexes of p53DBD with either DNA or iASPP repressor protein. We aimed to understand the communication path for the allosteric response and the residue-wise nonbonded energy contributions to molecular recognition by the p53DBD in these two complexes. To this end, the molecular dynamics (MD) simulation studies on both the p53DBD-DNA and p53DBD-iASPP complexes were performed, after which exhaustive analysis of the residue-wise nonbonded energy contributions (individually with the DNA or iASPP) and the pairwise interaction energies of the p53DBD were carried out.

Our results showed that the motions are restricted to the loop regions of p53DBD, whereas no significant change happens in the side-chain and backbone dynamics. Remarkably, however, the “energy hubs” (energy hubs are defined as the electrostatically highly connected residues in the electrostatic network, which are potentially important for structural integrity and communication) identified from the electrostatic network analysis revealed an allosteric communication (which was not apparent from the structure or dynamics) between L1 loop residue K120 of the binding interface to the distal residue R213 (in the S6-S7 connecting loop) of p53DBD in both the DNA and iASPP complexes. Notably, the allosteric communication path is constituted with evolutionarily conserved residues (21) implying the biological relevance of the allosteric connection identified here.

K120 on the L1 loop (part of the DNA binding interface) faces DNA in the p53DBD-DNA complex, whereas iASPP binding forces it to move away from the DNA binding site in the p53DBD-iASPP complex. Interestingly, analysis of the residue-wise electrostatic network showed that movement of the single charged residue K120 completely rewires the K120 to R213 electrostatic communication in the p53DBD-iASPP complex, as compared to that of the p53DBD-DNA complex, which apparently makes the DNA interaction less favorable. Furthermore, the results also revealed that perturbations in specific pairwise electrostatic interactions occur because of the distance population shift (distance between two residues connected either by hydrogen bond or by salt bridge) in the two complexes of the p53DBD.

Materials and methods

MD simulation

Starting structures for MD simulations of p53DBD bound to the DNA and iASPP were obtained from Protein Data Bank (PDB), PDB:3KMD; p53 residues 92–291 of chain C were extracted to construct monomer-DNA complex and PDB:6RZ3, p53 residues 92–291, respectively. They were equilibrated for 1 ns before doing the final simulation. The C- and N-termini were capped with N-methyl amide and acetyl groups, respectively. MD simulations for the complexes were performed with GROMACS (GROningen MAchine for Chemical Simulations) 5.1.2 software (22) using the Amber99SB-Ildn force field (23) and TIP3P water model (24). The complex was centered in a cubic box placed at least 1.0 nm from the defined box edge. The protonation states for the titratable residues were determined using the multiconformation continuum electrostatics method (25) and also from PROPKA (26) Web server (https://www.ddl.unimi.it/vegaol/propka.htm) assuming pH 7.4. All the Asp, Glu, Arg, and Lys residues are charged, and His residues are neutral. Force field parameters for the Zn²⁺ ion were obtained from Lu et al. (27). The systems were neutralized by adding 150 mM NaCl using the Na⁺ and Cl⁻ force field parameters described in the Amber99SB-Ildn force field. Substitution of the appropriate number of solvent molecules by Na⁺ and Cl⁻ ions mimics experimental (the condition used for crystallization of the complexes studied here) salt conditions (14). We performed another short 400 ns of MD simulation for the complexes using 225 mM NaCl. The higher salt concentration was used to check whether residue-wise electrostatic contributions of p53 could arise because of low salt concentration. The high-salt simulations confirmed that the electrostatic contributions arise predominantly from the interactions with either of the substrates (DNA or iASPP), not ions in solution. The structures were energy minimized, followed by NVT equilibration and then NPT equilibration. The temperature was controlled through velocity rescaling (28) at 310 K with a time constant of 0.1 ps, and pressure was controlled using a Parrinello-Rahman barostat (29) at 1 bar. The particle mesh Ewald algorithm (30) was applied to calculate long-range electrostatic interactions. The cutoff for short-range electrostatic and van der Waals (vdW) interaction is 1.2 nm. After equilibration, MD simulations of 1.0 μs for each complex were performed and frames were recorded at every 2 ps (31). Stabilization and convergence of simulations were ensured by calculating root mean-square deviations (from the starting equilibrated conformation for the final MD run) throughout the whole trajectory and comparing root mean-square fluctuations (RMSF), order parameters, and contact frequency maps from the first and second half of the simulations (Figs. S2–S5).

Differential contact frequency map and dynamic cross-correlation matrix

The last frame from the final MD simulation each complex was extracted using the tool provided by GROMACS. A contact is identified between any two atoms of two residues if they are less than 4.5 Å apart (32). A differential contact map was generated from the contact matrices for p53DBD-DNA and p53DBD-iASPP, considering contacts that include p53DBD as well as substrates (DNA and iASPP).

Throughout an MD simulation, contacts among residues can form or break. The contact frequency map is therefore defined as

$$C{F}_{ij}^{p53-substrate}=\frac{n_{ij}}{N},$$

where n_ij and N indicate the frame numbers at which the i and j residues are in contact and the total number of frames throughout the trajectory, respectively. CF_ij varies from 0 to 1, where values 0 and 1 indicate the absence and presence of a contact between residues i and j throughout the trajectory, respectively. The contacts are defined here in such a way to include intra-p53DBD contacts as well as inter-p53DBD contacts with substrates (substrate is either the DNA or iASPP) simultaneously.

The differential contact map (ΔCF_ij) is defined as

$$\Delta C{F}_{ij}=C{F}_{ij}^{p53-DNA}-C{F}_{ij}^{p53-iASPP},$$

where ΔCF_ij = 1 and ΔCF_ij = −1 indicate a contact between p53DBD i and j residue pairs present exclusively in p53DBD-DNA and p53DBD-iASPPcomplexes, respectively.

Cross-correlation maps are used to identify the regions that move in phase or out of phase with each other during the simulations (33) for the p53DBD residues in the p53DBD-DNA and p53DBD-iASPP complexes. The elements of the matrix (C_ij) are obtained from their position vector (r) as shown in the equation below:

$$C_{ij}=\frac{⟨\left(r_i-⟨r_i⟩\right)\left(r_j-⟨r_j⟩\right)⟩}{\sqrt{\left(⟨r_i^2⟩-{⟨r_i⟩}^2\right)\left(⟨r_j^2⟩-{⟨r_j⟩}^2\right)}},$$

where i and j correspond to any two atoms, residues, or domains; r_i and r_j are position vectors of i and j; and the angle brackets denote an ensemble average. Interatomic cross-correlation fluctuations between any two pair of p53DBD residues (all atoms are considered) in the p53DBD-DNA and p53DBD-iASPP complexes are calculated by using this expression and are represented graphically by the dynamic cross-correlation matrix (DCCM). The value of C_ij can vary from −1 (completely anticorrelated motion) to +1 (completely correlated motion).

Order parameter S²

Backbone N-H vectors were selected for the p53DBD residues in the p53DBD-DNA and p53DBD-iASPP complexes to calculate S² over the duration of the trajectory, The order parameter represents dynamics of the protein, with the value 1 indicating complete rigidity and decreasing toward 0 for enhanced flexibility.

$$S^2=\frac{1}{2}\left(3{\sum }_{i=1}^3{\sum }_{j=1}^3⟨{⟨{\mu }_i{\mu }_j⟩}^2-1⟩\right),$$

where μ₁, μ₂, and μ₃ are the x, y, and z components of the bond vector scaled to unit magnitude, μ (34). Angular brackets indicate averaging over the snapshots.

Energy perturbation

Throughout the entire trajectory, the residue-wise nonbonded interaction energies between the p53DBD and its two binding partners (DNA and iASPP) are calculated as described:

$$\Delta E_{nonbonded}^{p53-substrate}=\Delta E_{Electrostatic}^{p53-substrate}+\Delta E_{vdW}^{p53-substrate}.$$

The nonbonded interactions $\Delta E_{nonbonded}^{p53-substrate}$) include both electrostatic $\left(\Delta E_{Electrostatic}^{p53-substrate}\right)$ and vdW $\left(\Delta E_{vdW}^{p53-substrate}\right)$ interactions and are modeled using a Coulomb and Lennard-Jones potential function, respectively. For the ith residue, the $\Delta E_{nonbonded,i}^{p53-substrate}$ can be further broken down in terms of contributions from p53DBD, substrate (either DNA or iASPP), and water in the following manner:

$$\Delta E_{nonbonded,i}^{p53-substrate}=\Delta E_{nonbonded,i}^{p53}+\Delta E_{nonbonded,i}^{substrate}+\Delta E_{nonbonded,i}^{Water}$$

$$=\left(⟨E_{nonbonded,i}^{p53-substrate}⟩-⟨E_{nonbonded,i}^{p53-unbound}⟩\right)+⟨E_{nonbonded,i}^{substrate,bound}⟩+\left(⟨E_{nonbonded,i}^{water,bound}⟩-⟨E_{nonbonded,i}^{water,unbound}⟩\right),$$

where $\Delta E_{nonbonded,i}^{p53}$, $\Delta E_{nonbonded,i}^{p53}$, and $\Delta E_{nonbonded,i}^{water}$ represent the change in the average nonbonded interaction energy throughout the whole trajectory between the bound and unbound state due to the interactions between the ith residue and p53DBD, substrate, and water, respectively. A large cutoff of 2 nm is applied for computing the interaction energy with water molecules. In addition, we calculated the electrostatic interaction energies of Zn²⁺ with each residue of p53 in both of the complexes throughout the 1 μs of MD trajectories.

For the nonbonded interaction, no cutoff is used. Every 10 time steps during MD simulation, we stopped the overall translation and rotation to prevent all the kinetic energy from becoming associated with the center of mass motions and overall rotations. The residue-wise vdW and electrostatic (Coulomb) energy contributions to $\Delta E_{nonbonded,i}^{p53-substrate}$ were calculated separately for the p53DBD residues in the complex of p53DBD-DNA and p53DBD-iASPP, but the vdW terms are found to be numerically much smaller than the respective electrostatic ones, which is obvious because of the high negatively charged nature of both the DNA and iASPP. Therefore, we have focused on the electrostatic interactions in calculating the perturbation in pairwise interactions ΔE_i-j:

$$\Delta E_{i-j}=\Delta E_{i-j}^{Electrostatic}+\Delta E_{i-j}^{vdW}\approx \Delta E_{i-j}^{Electrostatic}.$$

The values of dielectric constants for the solvent and protein used in this calculation are 80 and 2, respectively. In the presence of both the DNA and iASPP, pairwise electrostatic perturbations are calculated for the p53DBD residue pairs throughout the whole MD simulation trajectories.

The interaction energy between two residues i and j is the sum of the nonbonded interaction energies already defined in a force field where

$$\Delta E_{i-j}^{Electrostatic}=\mathcal{\zeta }\frac{q_i{q}_j}{{\epsilon }_r{r}_{ij}},\text{where}\mathcal{\zeta }=\frac{1}{4\pi {\epsilon }_o}=138.935$$

$$\Delta E_{i-j}^{vdW}=4{\epsilon }_{ij}\left({\left(\frac{{\sigma }_{ij}}{r_{ij}}\right)}^{12}-{\left(\frac{{\sigma }_{ij}}{r_{ij}}\right)}^6\right),\mathrm{w}\mathrm{h}\mathrm{e}\mathrm{r}\mathrm{e}{\sigma }_{ij}=\frac{1}{2}\left({\sigma }_{ii}+{\sigma }_{jj}\right)\mathrm{a}\mathrm{n}\mathrm{d}{\epsilon }_{ij}={\left({\epsilon }_{ii}{\epsilon }_{jj}\right)}^{\frac{1}{2}}.$$

Construction of energy networks, the definition of energy hubs, and defining shortest paths

Energy networks are constructed by considering the p53DBD residues in the p53DBD-DNA and p53DBD-iASPP complex as nodes. A weighted edge is made between any pair of residues i and j by considering the interaction energy as the weight. Energy hubs are defined as nodes that have a higher degree of connectivity in the network. Note that for better comparison, we have considered the same hubs in both the complexes. The shortest path was calculated based on the procedure described earlier (33). The shortest path represents the bonds that should be strong enough to maintain the path and at the same time be flexible for the transmission of information. Thus, the shortest communication paths are optimal in terms of both strength and stiffness.

Betweenness-centrality is computed using the following equation:

$$BC\left(\nu \right)=\frac{2}{\left(n-1\right)\left(n-2\right)}{\sum }_{i\ne \nu \ne j\ne {\rm E}}\frac{{\sigma }_{ij}\left(\nu \right)}{{\sigma }_{ij}},$$

where BC(ν) is the betweenness-centrality of residue ν, n is the number of residues within the network, σ_ij(ν) is the number of shortest paths between residue i and j that pass through residue ν, and σ_ij is the total number of shortest paths from i to j. Dijkstra’s algorithm (35) is applied to calculate the shortest path between residues i and j.

Software used

The following software packages were used to create all the illustrations: INKSCAPE 0.92 (https://inkscape.org/release/0.92.2/mac-os-x/), UCSF Chimera (35), PyMOL (The PyMOL Molecular Graphics System, Version 2.0; Schrödinger, LLC), GraphPad Prism8 (https://www.graphpad.com/scientific-software/prism/), and GIMP (https://www.gimp.org/downloads/).

Results

Overall structure and dynamics of p53DBD in complex with DNA or iASPP

The crystal structures show that iASPP binding displaces the L1 loop of p53DBD away from the DNA binding interface (14). We intended to find the allosteric conformation changes, if any, in backbone traces of p53DBD in the DNA and iASPP complexes after 1 μs of MD simulation. To this end, the backbone traces of the final conformations of p53DBD complexes with DNA and iASPP resulted from the MD simulations were superimposed, and it was found that the structural changes in p53DBD are confined to loop regions only (e.g., loops L2 and L3 and loops connecting the S4-S3, S6-S7, S7-S8, and S9-S10 β-sheets) (Fig. 1 A).

Figure 1.

(A) Superposition of the final simulated structures of p53DBD in the DNA-bound state (starting structures: PDB: 3KMD, colored in red) and iASPP-bound state (PDB: 6RZ3, colored in blue), extracted from MD simulation. DNA and iASPP are colored in salmon and pale cyan, respectively. (B) Differential contact map between DNA- and iASPP-bound simulated structures. Contacts exclusively present in DNA- and iASPP-bound states are shown in red and blue, respectively, with the common contacts colored gray. Most of the differential contacts appear in the binding regions to the different binding partners. To see the figure in color, go online.

We then focus on the p53DBD side-chain reorganizations in these two complexes. A commonly used method to distinguish variations in side-chain conformations is the residue pairwise contact map (36,37). It has been postulated that, for allosteric involvement, contacts of neighboring residues are more likely coupled compared to the distal ones (38,39), and different contact-based frameworks have been proposed to evaluate the structural basis for the transmission of information (40, 41, 42). The residue pairwise differential contact matrix for the p53DBD residues (Fig. 1 B) is constructed following similar methods based on the final state of the simulated structures of p53DBD-DNA and p53DBD-iASPP complexes to detect differences in the p53DBD side-chain contacts. The contacts found exclusively in either of the complexes, p53DBD-DNA or p53DBD-iASPP, are mapped for p53DBD residues in red and blue, respectively. The majority of pairwise contacts were found to be common (gray regions). However, some contacts are exclusively present in p53DBD in one or the other of the complexes. As expected, these complex-specific contacts occur in the regions of p53DBD where the binding partners interact. Thus, the differential side-chain contact map upon DNA or repressor binding represents subtle structural rearrangements in p53DBD. This implies that structural comparison does not provide any information about p53 molecular recognition.

Evidently, the conformational dynamics play (as stated before, structures do not explain recognition) a crucial role (43) behind the energy modulation (i.e., variation in energies) in p53DBD in the two complexes. Residue-wise RMSFs based on Cα atoms and the order parameter (S²) of N-H vectors have been calculated for the p53DBD residues in the p53DBD-DNA and p53DBD-iASPP complexes with the anticipation that fluctuations would provide deeper insight into the dynamic nature of the p53DBD backbone. In contrast to DNA binding, repressor binding increases the overall fluctuations of the Cα atoms of the p53DBD. In specific regions of p53DBD, such as the loops connecting the S3-S4 (V147-T155), S7-S8 (Y220-C229), and S9-S10 (D259-N263) β-sheets, noticeable differences were observed between the p53DBD-DNA and p53DBD-iASPP complexes. Interestingly, however, loop L1 (F113-C124) of p53DBD, which contains both DNA binding (K120) and iASPP binding (H115) residues, showed smaller Cα atom fluctuations in the repressor-bound state compared to the DNA-bound state (Fig. 2 A), suggesting that L1 loop flexibility also plays a crucial role in iASPP complex formation by p53DBD (14). On the other hand, with the exception of loop L3, which contains DNA binding residues R248 and Zn²⁺ binding residues C238 and C242, N-H bond fluctuations calculated from the order parameter (S²) showed comparable overall backbone dynamics of p53DBD in the two complexes (Fig. 2 A). It is noteworthy that the dynamic regions identified from the S²-values of N-H vectors are in agreement with the NMR-derived identification of flexible regions in p53DBD (44). Evidently, Cα dynamics capture pronounced differences in complex formation for different ligands, whereas only minute changes are seen in N-H fluctuations.

Figure 2.

(A) Comparison of backbone dynamics of p53DBD bound with DNA (red) and iASPP (blue), based on residue-wise RMSFs and order parameters (S²). Dynamic cross correlation (strong: |C_ij| = 1.0–0.7; moderate: |C_ij| = 0.7–0.5; and weak: |C_ij| = 0.3–0.5) between residues of p53DBD is represented in the form of matrix in (B) with DNA-bound state, and (C) with the iASPP-bound state. Areas that show differences in DNA-bound state are indicated by boxes. They show correlation between the DNA and Zn²⁺ binding regions, which are missing in the iASPP-bound state. To see the figure in color, go online.

For determining dynamic allostery in proteins, the DCCM (C_ij) has recently been used to quantify correlated movements among residues (33,45). C_ij-values for correlated and anticorrelated motions range from 1 to −1 with cutoffs for strong (|C_ij| = 1.0–0.7), moderate (|C_ij| = 0.7–0.5), and weak (|C_ij| = 0.3–0.5). We calculated the DCCM for p53DBD in both complexes to uncover any distal correlations in p53DBD. Interestingly, the DNA binding residues in p53DBD (e.g., K120, R248, R273, C277, and R280), as well as Zn²⁺ binding residues (e.g., C238 and C242), tend to have strong to moderate correlations among each other in the presence of DNA, whereas such correlations are completely missing in the complex with iASPP (8) (Fig. 2, B and C).

Furthermore, dynamic correlations between L1 and helix H2 to Zn²⁺ binding loop L2 were seen in the DNA-bound state. The observation implies that Zn²⁺ plays a crucial role in providing a dynamic interplay between DNA binding loop L1 and helix H2 of p53DBD that is suitable for DNA binding. The results also reinforce the previous reports showing removal of the Zn²⁺ ion (i.e., disruption of stable architecture for DNA binding) abolishes DNA binding activity (46). The L1 loop moves from the DNA binding surface upon binding with the iASPP that leads to the disruption of the stable DNA binding architecture of p53, and as a result, no correlated motion was found from the DCCM analysis in the case of the iASPP-bound complex. The result, showing the correlated motions of the abovementioned residues in the DNA-bound state, is significant as compared to the same in the repressor-bound state (47,48) is an important observation revealed in our study.

Although differences in the overall dynamics of p53DBD are observed in the two complexes, it is still not clear how distal residues contribute to molecular recognition. We suspected that the dynamics might be regulated by intrinsic electrostatic contributions, so we further investigated the intraprotein interaction energy network.

Contributions of nonbonded interactions in p53DBD complexes with DNA and iASPP

It is clear that the structure and dynamics of p53DBD, as stated earlier, could not establish the functional contributions of the distal residues. Therefore, we analyzed more fundamental electrostatic and vdW interaction energies. It has been well established that complex formation induces a drastic perturbation in the energy of the binding site, which propagates to the allosteric site via an intraprotein network of contacts (20).

It may be noted here that ions present in solution could have significant contributions to electrostatic energy. Therefore, MD simulations for both the p53DBD-DNA and p53DBD-iASPP complexes were carried out using two different salt conditions, 150 and 225 mM. The residue-wise electrostatic contributions using different salt concentrations were found to be similar (Fig. S5, E and F). This confirms electrostatic contributions solely arise from the interactions between p53 and its substrates (here DNA and iASPP). Further analyses were continued on the simulation with 150 mM salt, the experimental condition used for crystallization of the complexes studied here.

Throughout the MD simulation, we have computed the alterations in the residue-wise nonbonded interaction energies in terms of the electrostatic $\left(\Delta E_{Electrostatic}^{p53-substrate}\right)$ and vdW $\left(\Delta E_{vdW}^{p53-substrate}\right)$ contributions of p53DBD residues in the p53DBD-DNA and p53DBD-iASPP complexes (Figs. 3, A and B and S6). A global decrease in residue-wise electrostatic energy in p53DBD is observed when it is bound to iASPP as compared to its DNA-bound state. Significant differences in the p53DBD residue-wise electrostatic energies $\left(\Delta E_{Electrostatic}^{DNA-iASPP}=\Delta E_{Electrostatic}^{p53-DNA}-\Delta E_{Electrostatic}^{p53-iASPP}\right)$ between p53DBD-DNA and p53DBD-iASPP interactions were found (up to ∼150 kcal/mol), although the contributing residues remain the same for both the complexes (Figs. 3 A and S7 A; Tables S1 and S3). This observation implies that p53DBD uses a common electrostatic framework, regardless of its binding partner, but the spectrum of residue-wise energy contributions is different depending on the binding partners. The interaction energy profile of vdW $\left(\Delta E_{vdW}^{DNA-iASPP}=\Delta E_{vdW}^{p53-DNA}-\Delta E_{vdW}^{p53-iASPP}\right)$, on the other hand, shows a modest range of variations up to ∼10 kcal/mol (Figs. 3 B and S7 B; Tables S2 and S4). Clearly, electrostatic energy dominates in specifying subtle structural rearrangements.

Figure 3.

Residue-wise contributions to the nonbonded energy between p53DBD and DNA (red) or iASPP (blue) from (A) electrostatics (B) vdW interactions. All of the residues of p53DBD that have electrostatic contributions with either iASPP or DNA are shown. The electrostatic contributions to DNA binding are much higher than for iASPP (extended red bars), including both direct binding and allosteric residues. The vdW contributions are much smaller than electrostatic contributions (note scales). (C) Electrostatic energy contributions for both the complexes are mapped onto a merged crystal structure. Red spheres indicate that the dominant electrostatics interaction is seen in the DNA-bound complex, and blue ones represent higher electrostatic contributions correlated with iASPP binding. To see the figure in color, go online.

Our analysis identified the residues that interact with either the DNA or the iASPP strongly and favorably. For DNA-interacting residues K120, R248, R273, and R280 (8), the electrostatic contributions are greater than −100 kcal/mol each (Fig. 3, A and C), whereas the iASPP binding residue H115 (14) is stabilized in the p53DBD-iASPP complex by the vdW energy contribution (Fig. 3, B and C). Furthermore, we found that the Zn²⁺ ion has significant electrostatic interactions with the DNA binding residues Arg248, Arg273, Arg280, and Arg283 of p53DBD in the DNA-bound complex. In contrast, the Zn²⁺ interaction energies decrease upon iASPP binding (Fig. S8).

Next, we sought to find out whether the allosteric coupling in p53DBD complexes can be captured by electrostatic interaction energy or not. Remarkably, in the p53DBD-DNA complex, several noninteracting (distal from DNA) residues, e.g., R213, E258, R110, and E171, displayed large values of $\Delta E_{Electrostatic}^{p53-substrate}$ (Fig. 3 C), indicating allosteric modulation. In addition, in switching from the DNA to iASPP-bound states, some of the p53DBD residues undergo drastic perturbations in nonbonded energies. Distal p53DBD residues such as K101, R110, D148, and D228 showed larger electrostatic contributions in the iASPP-bound state than the DNA-bound state (Fig. 3, A and C).

Interactions of p53DBD residues with water molecules (cutoff distance within 2 nm) were seen in both the complexes, prompting us to consider solvation effects. The results show opposite patterns of solvation energy for p53DBD in the two complexes; the overall distribution is positive for DNA recognition and negative for repressor interaction (Fig. S9). DNA being more negatively charged than iASPP, a more pronounced effect on solvation is seen in the DNA-bound p53DBD complex than in iASPP-bound p53DBD, in which the repressor shields a large amount of the solvent-exposed area of p53DBD. p53DBD-solvent interaction energies compensate to some extent for the p53DBD-DNA electrostatic interaction energies, mainly at the primary DNA binding residues, for example, K120, R248, R273, and R280. The incorporation of solvent molecules into the p53DBD-DNA contact interface was observed throughout the simulation.

Electrostatic energy reveals an allosteric communication path comprising evolutionarily conserved residues

Beyond just identifying the p53DBD residues that show different electrostatic energies in the two different substrate-bound states, we generated a connectivity network by scrutinizing the pairwise electrostatic interactions (ΔE_{i = j} between two intraresidues i and j of p53DBD) that lead to these differences (49). ΔE_{i = j} for residues with |ΔE_{i = j}| > 3 kcal/mol are tabulated in Tables S5 and S6, and Fig. 4, A–L highlight the residues with |ΔE_{i = j}| > 6 kcal/mol. Residues that are electrostatically connected to a large number of other residues and are likely to be involved in the transfer of electrostatic energy from one part of the complex to another are called the energy hubs. Employing the electrostatic betweenness-centrality approach (see Materials and methods), the same energy hubs are defined for p53DBD in DNA- and iASPP-bound complexes, namely K132, E171, R196, R158, E110, and E271 (Fig. 4, A–L). Notably, residues corresponding to the defined energy hubs are evolutionarily conserved, as reported previously (21), in which the conserved residues are aligned for p53-mouse, p53-chicken, p53-fly, and human p63 and p73. The evolutionary conservation of these residues implies their involvement in real biological functions.

Figure 4.

Allosteric connection propagation in p53DBD mediated by energy hubs in (A)–(F) for p53DBD-DNA and (G)–(L) for p53DBD-iASPP complex, with lines indicating pairwise electrostatic interaction energy ΔE_i-j > 6 kcal/mol. Connectivity networks shown in (M) for DNA-bound and (N) for iASPP-bound p53DBD indicate the electrostatic energy linkages between active sites and the allosteric regions. The electrostatic interactions arise because of different binding partners. Residues for DNA- and iASPP-bound p53DBD are shown in red and blue spheres, respectively. Dashed lines indicate lower values of the interaction energy at which the difference between the pairwise interaction energies in the two complexes is more than 6 kcal/mol. Significant changes among pairs like E171-R249, E171-R213, and R110-D148 are clearly noticeable. DNA contacts are indicated by red lines. Note that the most prominent differences between the complexes occur at allosteric sites. To see the figure in color, go online.

Although inter-energy hub connections in p53DBD are mostly through common residues in both DNA- and repressor-bound complexes, some new interacting partner residues also appear that are exclusive in either of the complexes. This implies that electrostatic connections between energy hubs and other partner residues are not static but dynamic in nature (i.e., vary depending on the binding partner) in the two complexes; for example, residue E171 is linked to C238 in the DNA-bound complex, but the same E171 residue linked to N210 in the iASPP complex. Most of the pairwise interaction energies of p53DBD are higher in the iASPP-bound state compared to the DNA-bound one (Tables S5 and S6 for comprehensive pairwise energy), but the residue-wise energy contributions of p53DBD toward DNA are higher than that of the iASPP (Tables S1 and S3). To identify communication networks based on electrostatic energy in p53DBD, we further scrutinized pairwise electrostatic interactions of other p53DBD residues with intra- and inter-energy hubs. Interestingly, for both the complexes, p53DBD undergoes drastic energy perturbations (|ΔE_{i = j}| > 8 kcal/mol) in distal allosteric regions (these can be favorable in one complex and unfavorable in the other) (Fig. 4, M and N). In particular, pairwise electrostatic interaction energies in p53DBD are more favorable in the DNA-bound state for residue pairs R196-D186 and R175-H179, whereas in the iASPP-bound state, pairwise electrostatic interaction energies are more favorable for R110-D148, E171-R213, R175-D184, and E171-R249 pairs. Notably, as verified by the respective crystal structures, these residue pairs mentioned above are connected to each other by either salt bridges or hydrogen bonding.

In both the complexes, residue R213, far from each binding interface, is connected to DNA binding residue R248 via the communication path: R248 => R249 => E171 => R213, where the residues are evolutionarily conserved (Fig. 5, A and B). It is noteworthy that in endocrine cancers, R213 is a frequently mutated amino acid site (16). The H-bond between R213-E171 is retained in iASPP-bound p53DBD, whereas the same is broken in the DNA-bound state (Fig. S10, A and B). The H-bond linking E171-R249 is also disrupted in DNA-bound state. Such conformational alteration would favor the p53-binding protein 1 (53BP1) to interact with the R249 of p53DBD after DNA binding (50). p53, being a transcription factor, has to interact with other partner proteins, with the communication path being utilized to transfer the information from the DNA binding site to the binding sites of other partner proteins (like 53BP1) to carry out p53-mediated cellular processes.

Figure 5.

Representative snapshots showing comparisons of communication paths from the binding sites that transfer an electrostatic signal from the DNA binding sites to the allosteric site mapped on (A) DNA-bound and (B) iASPP-bound p53DBD structures. Hydrogen bonds between E171-R249 and E171-R213 that are broken in the DNA-bound state are shown in the repressor-bound state. Loop L1 conformation mediates a reconfiguration of the dynamic electrostatic path of connecting residue K120 to R213 in (C) DNA-bound and (D) iASPP-bound p53. To see the figure in color, go online.

We further considered residue K120 of the L1 loop of p53DBD, which is involved in both DNA and repressor binding, to find any allosteric communication with distal residue R213, as the pairwise interaction energy between R213-E171 becomes favorable upon iASPP binding because of the movement of the L1 loop from DNA-bound to iASPP-bound state. It is rational to consider residue K120 to decide the connecting path between K120 and R213 because the main role played by the iASPP repressor is to shift the L1 loop away from helix H2. We found that K120 to R213 communication follows a different path in the iASPP-bound complex as a consequence of the conformational alteration of the L1 loop (Fig. 5, C and D). Clearly, this path is actually an extension of the allosteric path previously described (connecting R248-R213) in the p53DBD-DNA complex (Fig. 5 C), but not in the p53DBD-iASPP complex (Fig. 5 D). Notably, although the communication paths connecting K120-R213 are different in the two complexes, the residues involved in both cases are evolutionarily conserved (21).

Pairwise side-chain distance population shifts modulate energy perturbations

So far, significant differences in electrostatic energy contributions of p53DBD residues have been demonstrated in DNA-bound and repressor-bound states. We have measured the pairwise distance distribution of R110-D148, E171-R213, R158-E258, R175-D184, R196-D186, R175-H179, E171-R249, D281-R273, and E285-N288 residue pairs throughout the trajectory to correlate with the underlying dynamics behind this finding (Fig. 6, F–N). It is apparent that most pairs have a population peak around 0.28 nm (hydrogen atoms are not indexed for calculation here), suggesting H-bond- or salt-bridge-mediated contacts within the pairs. The iASPP-bound p53DBD crystal structure shows an H-bond between E171 and R213 that is confirmed by a single population peak without a shoulder (Fig. 6 G). The population peak of this pair practically disappears in the DNA-bound complex (Fig. S10, A and B), in agreement with the DNA-bound crystal structure showing no H-bond between E171 and R213 (7). This correlation may also be implicated in either enhancing conformational entropy in the DNA-bound state, in which pairwise energy decreases by disrupting H-bonds (thus causing more dominant side-chain contribution), or decreasing the conformational entropy in the iASPP-bound state, in which the pairwise energy is increased by strengthening H-bonding.

Figure 6.

Representative snapshots showing the changes in contact patterns between DNA- and iASPP-bound states for (A and B) E171-R213 and (C and D) E171 and R249 residue pairs. (E) The differential contact frequency map of p53DBD residues is calculated from MD simulation trajectories of both complexes. Differential contact regions (a–e) are marked with circles. Regions b, c, and d reflect changes in residue contacts for E171, R213, and R249, respectively. Regions a and e represent contact differences in the two complexes for the L1 loop and H2 helix. (F–N) Comparison of the probability distribution of pairwise distance throughout the trajectory of p53DBD residues bound to both DNA (red) and iASPP (blue). The corresponding pairwise interaction energy is also shown with corresponding color codes. Peaks around 0.28 nm (hydrogen atoms are not included in the calculation) represent the existence of H-bonds or salt bridges. To see the figure in color, go online.

Because p53DBD is flexible in solution, contacts break and form. We have calculated the differential contact frequency map throughout the whole trajectory to comprehend the variations in the contact pattern in p53DBD. Five main regions (marked “a” to “e” in Fig. 6 E) illustrate major differences in the interaction maps for the p53DBD-DNA and p53DBD- iASPP complexes. Regions “b,” “c,” and “d” reflect changes in residue contacts for E171, R213, and R249, respectively, whereas regions “a” and “e” represent contact differences for the L1 loop (involved in both DNA and iASPP binding) and H2 helix (responsible for DNA binding only). DNA binding-mediated rearrangement is accompanied by a change in contact frequency for the case of the R175-H179 pair (Fig. 6 K), which is absent in the iASPP-bound complex.

In conclusion, pairwise energy perturbation, i.e., change in pairwise interaction energy is strongly correlated to the change in distance over time between two residue pairs that are connected through H-bonding or salt bridge.

Discussion

The tumor suppressor transcription factor p53, being at the hub of an extensive transcriptional network (a hub protein), needs to interact with DNA and other binding partners, and changes in either conformation or dynamics occur during the interaction process (44). Here, we have demonstrated that electrostatically driven dynamic allostery, modulated by ligand-binding, plays a pivotal role in p53 molecular recognition. Based on our study, we speculate that while interacting with different other partners, electrostatics would have a significant contribution too for a hub protein such as p53. We have shown that residue-wise electrostatic contributions in the p53DBD are far more stabilizing for DNA binding than for iASPP binding. Especially in the DNA-bound state, pairwise interaction energy decreases because of distance population shift (caused by H-bond or salt bridge breakage), indicating enhanced side-chain flexibility. We suggest changes in connecting partners to energy hubs result in the transmission of information in p53DBD via ligand-dependent dynamics (Video S1). In other words, to acquire a particular conformational state corresponding to a “valley” in the energy landscape, substrate-dependent (e.g., DNA, repressors, etc.) modulation of energy landscape is required for p53DBD. Further, our analysis showed that the Zn²⁺ ion in the p553DBD-DNA complex contributes a significant amount of electrostatic energy to the DNA binding residues present in the L3 loop and H2 helix, whereas its electrostatic contribution decreases in the case of the iASPP-bound complex. It appeared from our result that the presence of the Zn²⁺ ion strengthens the L3 loop and H2 helix architecture with a significant number of electrostatic contributions. In this context, a previous report (46) showed that the removal of the Zn²⁺ ion abolishes site-specific DNA binding activity.

Video S1. Dynamic conformational changes from DNA-bound state to iASPP-bound state

In DNA-bound state H-bonds are broken within E171-R249 and E171-R213 pairs (shown in magenta) allowing p53DBD to interact with cofactors like 53BP1, 53BP2 etc. (where those residues are involved) during the process of transcription. On the Contrary, in the iASPP-bound state formation of H-bond in those residue pairs impairs p53DBD to interact with cofactors. Apparently, motion of the allosteric loop (containing R213) is the key driving factor for the H-bond breakage and formation. Factor-binding loop L1 (shown in brown color) motion indicates the dynamic switch in the recognition process of iASPP from DNA and allosteric path connecting K120 of L1 and R213.

The biological significance of our study lies in the following observations. We have shown that the communication path for an allosteric response connects R213 to DNA binding residue K120, which helps rationalize the previous report that the R213A mutation results in a loss of binding to DNA and transcription cofactor 53BP1 (18). We suggest that either a change in the electrostatic interactions or side-chain rearrangements due to an H-bond population shift are likely responsible for the inactivation of the R213A mutant. Moreover, although the p53DBD has numerous charged residues, the K120 to R213 connecting path appears to be completely rewired by a conformational change of a single charged residue (K120) on the L1 loop that accompanies the conversion from the p53DBD-DNA to p53DBD-iASPP complex. The R213 mutations in p53DBD prevent p53 trans-activation activity in cells, make targeting the upstream p21 gene less effective, and can mitigate p53-mediated G1/S arrest (18). Clearly, the distal residue R213 of the p53DBD has a pivotal role in modulating ligand interaction at the p53 binding site (17) via allosteric control mediated by electrostatics.

The observations from our study may be extrapolated to explain p53 inactivation due to mutations. Our results suggest that a differential pattern of electrostatic interaction energies in p53DBD upon complex formation with different partners accounts for allosteric control that is not apparent from the structural parameters (e.g., contact map), which show no noticeable change. For “hotspot” mutations R248Q or R248W, R249S, R175H, and R282H (including both the “contact” and “structural” mutants), our results suggest that loss of the Arg charge would cause a drastic change in either pairwise electrostatic interactions or electrostatic contributions toward DNA binding. It is noteworthy that the energy hubs’ constituent residues K132, E171, and R196 are in direct contacts with the abovementioned “hotspot” mutants either by H-bond or by a salt bridge (51). On the other hand, energy hubs E158 and E271 are the temperature-sensitive hotspots (52).

Our analysis of the p53DBD down to the residue level identifies electrostatic contributions toward substrates binding and pairwise interaction energies along with vdW and solvation energy, which for p53 have not been highlighted earlier. We suggest a process of molecular-recognition-mediated allosteric communication via a biologically relevant communication path, made up of evolutionarily conserved residues, that connects the binding site to the allosteric site of p53DBD and an inherent energy flow network within p53DBD during that process. This information would be highly significant in the field of target-specific orthosteric, as well as allosteric, drug designing.

Author contributions

J.S. and S.B. conceived the project. S.B. designed and performed research and analyzed data in consultation with J.S. S.B. and J.S. wrote the manuscript.

Editor: Jason Kahn.