Molecular dynamics of the full-length p53 monomer

Abstract

The p53 protein is frequently mutated in a very large proportion of human tumors, where it seems to acquire gain-of-function activity that facilitates tumor onset and progression. A possible mechanism is the ability of mutant p53 proteins to physically interact with other proteins, including members of the same family, namely p63 and p73, inactivating their function. Assuming that this interaction might occurs at the level of the monomer, to investigate the molecular basis for this interaction, here, we sample the structural flexibility of the wild-type p53 monomeric protein. The results show a strong stability up to 850 ns in the DNA binding domain, with major flexibility in the N-terminal transactivations domains (TAD1 and TAD2) as well as in the C-terminal region (tetramerization domain). Several stable hydrogen bonds have been detected between N-terminal or C-terminal and DNA binding domain, and also between N-terminal and C-terminal. Essential dynamics analysis highlights strongly correlated movements involving TAD1 and the proline-rich region in the N-terminal domain, the tetramerization region in the C-terminal domain; Lys120 in the DNA binding region. The herein presented model is a starting point for further investigation of the whole protein tetramer as well as of its mutants.

Introduction

The TRp53 gene, implicated in cancer development as well as in infertility,^1,2 is the most frequently mutated in all human cancers; see http://p53.fr or http://p53.iarc.fr.^3-5

The tumor suppressor protein p53 is composed of three large domains: (1) the intrinsically flexible N-terminal region, which includes 2 transactivation domain (TAD) and a proline-rich region (PRO); (2) the central core region, which contains the DNA binding domain (DBD) region with a coordinated Zn ion between β sheet folds; (3) the C-terminal region, which includes a tetramerization domain (TET) and regulatory C terminus domain (REG; see Fig. 1). The binding to DNA is mediated by recognition of specific sequence motif interacting directly with the DBD.⁶ Conversely, the transcriptional function requires the physical interactions of the 2 TAD with the transcriptional machinery. p53 is in fact a powerful transcription factor, and, consequently, it drives a large number of promoters,⁷ depending on specific activators.^8-11 Possibly the most studied promoter target is p21,^12-14 but new targets emerge as, for example, the connection between IL-7Ra and telomer erosion¹⁵ or the silencing of repeats and noncoding RNA.¹⁶ As a consequence of the transactivation of so many different promoters so far described, p53 is able to activate different biochemical pathways affecting the regulation of life and death of the cell. A recently discovered, relevant set of novel p53 targets include the metabolism of the cell,^17,18 for example the pathways of mevalonate,^19,20 serine,²¹ malate,²² and glutaminolysis.²³ This results in the regulation of cell death²⁴ or cell cycle.^25-27 A relevant metabolic role of p53 is exerted during hypoxia, as oxygen relative pressure reaches particularly lower levels at the center of the tumors, where p53 protein interplays with mTOR²⁸ as well as with the more classic pathways of HIF-1a and VHL.²⁹ A distinct role occurs during re-oxygenation, as during miocardial infartion or trombosis in the central nervous system.³⁰ In both cases p53 status is a relevant factor. Although abnormal protein variants were described several decades ago,^31-34 only recently has a distinct promoter been identified,³⁵ codifying of the full-length p53 or Δ0p53 (promoter 1) or for the Δ33p53 and Δ60p53 (promoter 2), similar to what occurs in other members of the p53 family, namely p73³⁶ and p63.³⁷ This results in significant changes in its functional properties as a transcription factor.^38-43 As a consequence of the p53 isoforms function as a transcription factor and its involvement in DNA damage, it is evident that the status of p53 plays a crucial role in cancer progression^44-52 as well in different physiological² and anticancer responses.⁴² In fact, p53 is frequently mutated in cancer, resulting in novel “gain-of-function”effects. The most frequently mutated region in human cancer is the DBD. Still now, after so many years and publications on the p53–DNA damage connection, new pathways are emerging. Just as an example, p53, with different degrees depending on its isoforms or its polymorphism at codon 72,⁵³ plays a crucial role in single-strand breaks during muscle function,⁵⁴ interacts with PRAP1,⁵⁵ as well as with DNA damage response.^56,57 The regulation of cell death by the p53 protein is quite complex, acting both at the level of autophagy,⁵⁸ lysosomes,⁵⁹ or at the core machinery of programmed cell death.^60-65 In addition to DNA damage response and cell death, p53 plays a crucial role in regulating cellular senescence^66-68 by interacting, for example, with MageA2,⁶⁹ PATZ1,⁷⁰ 4E-BP1,⁷¹ mTOR,^72,73 highlighting the vast complexity of this crucial regulation.

Figure 1. 3D structure model of the monomeric full-length p53 model. (A and B) colors show the different major domains: green, N-terminus, residues 1–100; red, core domain, residues 101–300; blue, C-terminus, residues 301–393. (C and D) colors show the different functional regions: yellow, transactivation domain 1, residues 1–42; orange, transactivation domain 2, residues 43–63; pink, proline-rich domain, residues 63–97; purple, DNA binding domain, residues 102–292; light blue tetramerization domain, residues 323–356; dark blue, C-terminal regulatory domain, residues 363–393. The model refers to the stable conformation between 150–850 ns; see Figure S1.

As the p53 proteins have different isoforms, and it is frequently mutated in cancer,⁷⁴ in order to understand its function with the related transactivation rules and superactivating sequences,⁷⁵ it is crucial to understand its structural interactions and dynamic function.^76-78 p53 is active as a tetramer and its structure bound to DNA has been resolved in the truncated core domain form,⁷⁹ even though only the full-length protein produces the maximum bending and twisting of the consensus DNA RE.⁸⁰ The full-length p53 has been resolved only in its tetrameric form by a combination of NMR, small-angle X-ray scattering, electron microscopy, and FRET,^81-86 showing that in absence of DNA, an open cross-shaped structure is formed, with loosely coupled dimers interacting via the core domain, whereas the structure rigidifies upon DNA binding and becomes more compact.

Because of its structural flexibility, most experimental structural studies have addressed only single domains or portions of them. In particular, the 2 TADs in the N-terminal domain were solved by NMR^87,88 and X-ray diffraction;⁸⁹ the core domain by X-ray diffraction^90-92 and NMR⁹³; the tetramerization region in the C-terminal domain by NMR⁹⁴ and X-ray diffraction.⁹¹ The quaternary structure of the p53 tetramer in complex with DNA has been determined by a combination of SAXS, NMR, and EM.⁹⁵ FRET experiments have been performed on the N-terminal plus DBD form (residues 1–292) and on the full-length in tetrameric form.⁸² However, there are intrinsically disordered regions, functional to the biological role of p53, that were not resolved by any technique.

Full-length p53 in its monomeric form is therefore a particularly elusive protein to be structurally studied with experimental methods. Several in silico studies have been performed, but all focused on single domains or fragments of p53. In particular, the p53 core domain has been the most studied both for its biological role in transcription process^96-100 and in cancer-related mutations.^101-104 MD conformational analyses were also performed for the N-terminal recognition α helix,^105-108 and for C-terminal fragments.^109,110

In order to study the entire monomeric protein behavior and its inter domain interactions, we have built a model for the full-length p53 (residues 1–393). The structure of unresolved fragments (10%) and disordered links between the domains have been predicted and merged with the experimental data. Here, we have accrued out a 850 ns-long molecular dynamics (MD) simulation to validate the stability of the model and predict its structural and dynamic properties.

Results

By means of the root mean square deviation (RMSD) analysis we measure how much the instant conformation, sampled by the MD simulation, differs from the starting structure. Figure S1 shows the RMSD as a function of simulation time for the C-α atoms of the protein. As expected, the N- and C-terminal domains undergo large conformational rearrangements when compared with the core domain, resulting in a constant increase of the RMSD up to 120 ns. After this time, a plateau is reached, and the value oscillates around 10 Å until the end of the simulation. We have therefore excluded the first 150 ns of simulation (red line) and performed all the following analyses on the last 700 ns, i.e., from 150 to 850 ns.

Analysis of the secondary structure content as a function of simulation time (Fig. 2) shows a very good stability of the core domain, mainly characterized by a β sandwich (from β strand S1 to S10) and loop–sheet–helix motifs (composed of loop L1; β strands S2 and S2’; the end of β trand S10 and helix H2, at the C-terminal boundary of the core domain).⁹⁰ Helices H2 and H1, in direct interaction with the Zn ion, show a relatively large variability in their secondary structure content. The intrinsically disordered C- and N-terminal domains are both, as expected, much less structured. In particular, the N-terminal domain does not show any stable secondary structures, although the nascent recognition helix, fully structured upon binding with MDM2, is visible around residues 20–25.¹¹¹ Accordingly, the tetramerization helices are partially folded (region 335–355). Finally, we observe a stable short α helix also in the C-terminal regulatory region (around residue 380–388).

Figure 2. Change in secondary structure of the p53 monomer. Plot of the secondary structure elements calculated by DSSP software as a function of simulation time. Colors (see top of the figure) show the different secondary structures as evolving during the MD simulation at equilibrium, that is between 150–850 ns; see Figure S1. Changes in secondary structure are evident in the entire sequence, especially at the N and C termini; see main text.

The per-residue root mean square fluctuation (RMSF) shows the highest values in the N- and C-terminal domains (Fig. 3). The first TAD and the proline-rich regions contain the most fluctuating residues, while the tetramerization region, in the C-terminal domain, is the third most fluctuating region. In the DBD, Lys120, within loop L1 (residues 113–123) is the most fluctuating residue. Note that this residue directly interacts with the DNA major groove, and it has been proposed to trigger the subtle p53 conformational changes necessary to the specific recognition of different DNA Res.¹¹²

Figure 3. Per-residue root mean square fluctuations. The values for each residue are colored in green, red, and cyan for the N-terminal, core and C-terminal domain, respectively. Larger degrees of fluctuations and secondary structure changes are evident in the functional domains, see Figure 1. The DBD, in red, is the most stable region with the most frequently mutated residues in human cancers located in the more stable areas.

In all the MD simulations, the large concerted structural rearrangements, often linked to the biological function of the protein, are disguised by a great number of small irrelevant fluctuations that can be eliminated performing an essential dynamics (ED) analysis.¹¹³ ED is based on the diagonalization of the covariance matrix built from the atomic fluctuations after the removal of the translational and rotational movement, and it allows the separation of the few 3N directions along which the majority of the protein motion is defined. The RMSF of the C-α atoms along the first 3 eigenvectors (Fig. S2) shows that the widest correlated movements in p53 involve the regions identified as the most fluctuating: i.e., TAD1 and PRO in the N-terminal domain, TET in the C-terminal domain; Lys120 in the DBD. Interestingly, the second and third eigenvectors uncouple the N-terminal domain motions, being dominated by the TAD1 or PRO fluctuations, respectively. Projections of the C-α atoms onto the eigenvectors show that nearly 20% of the whole p53 motion is described by the first eigenvector, and projections along this eigenvector show that this correlated motion is mainly contributed by the N and C terminus (Fig. 4A and B). ED results on p53 monomer allowed us to carry on further analysis to better characterize the main protein motions. The simulation time spent by the protein in a specific region along the ED eigenvectors gives us an estimation of the free energy value of the visited structural basin. The free energy maps of the projection of the trajectory on the essential subspace along eigenvectors 1–2 (Fig. S3A) shows that one major basin is present, spanning from −4 and 3 (eigevector 1) and −2 and 4 (eigenvector 2). The same analysis performed only on the core domain shows a much smaller conformational basin (Fig. S3B), demonstrating that the N- and C-terminal domains are responsible for the widest protein movements .

Figure 4. p53 structure along the first Essential Dynamics eigenvector. (A and B) Projection of the p53 structure along the first eigenvector. The three structures relative to average, minimum and maximum eigenvalues are represented. (C and D) Representative MD snapshots with highlighted the lateral chain of the residues involved in long residence hydrogen bonds between the N-terminal and the core domain; see also the Supplementary Tables. Colors are as in Figure 1A and B.

A number of conformations have been extracted from the minimum free energy region in the middle of the basin of Figure S3A, or along its border at the extreme of the eigenvector subspace, and the electrostatic potential distribution has been calculated. Two representative conformations (identified with the 2 arrows in Fig. S3A) have been chosen and their positive and negative iso-surfaces are shown in Figure 5. Figure 5A, representative of the most stable free energy region, shows that a positive 3-finger surface is formed by 3 lysine residues: 372 and 373 located in the Regulation region and 120 in the DBD, but a contribution originates also from His365, His368, and Lys370. These results indicate how the p53 different domains can cooperate in their interaction with negative partners such as the DNA. Figure 5B, chosen at the extreme along the first eigenvector subspace, shows a negative bulge formed by both TAD1 and TAD2 residues. The insert of Figure 5C highlights how the tip of the bulge is formed by the nascent helix responsible for MDM2 recognition, around residues 20–25. Therefore, wide structural fluctuations, secondary structure variability, and electrostatic fields can provide an efficient mechanism used by p53 in looking for the right molecular partner.

Figure 5. Electrostatic potential surface around p53. (A) Representative structure of the most stable region in the free energy plot (see Fig. S3). (B and C, insert) Representative structure of the basin boundaries in the free energy plot. The red and blue surfaces correspond to electrostatic potentials of −2.5 and +2.5 KT/e, respectively.

Global correlated motions in proteins are naturally driven by local protein–protein interactions, and so the network created by the most stable hydrogen bonds inter- and intra-domains has been analyzed. Despite the described large flexibility of the N-terminal domain, several “anchor points” between the N-terminal and the core domain are present. Three hydrogen bonds, in particular, have residence time longer than 80% of the simulation time (bold in Table S1), thus maintaining a tight interaction between residues 93–96, at the C terminus of the last proline (Pro92) of the PRO region, and 2 loop 1 residues (Thr170 and Val172), plus Arg213, located in the loop between S6 and S7, all in the DBD. Figure 4C shows a representative snapshot for these contacts with the involved lateral chains colored in orange and yellow for the N-terminal and core domain, respectively. Ser95 and Thr170 interact via a main chain hydrogen bond presents for 54% of the simulation time. Other N-terminal residues, contacting the DBD for long simulation times, are located in the TAD2. Gln52, Trp53, and Asp57 form direct hydrogen bonds for more than 50% of the simulation time with His178 (underlined in Table S1 and highlighted in Fig. 4D), close to His179 that binds the Zn ion, and therefore likely playing a role in the maintenance of the metal coordination. Moreover, Gln52 interacts with Arg181, by means of 2 hydrogen bonds, with percentage of simulation time between 35 and 40. The interface between the C-terminal and the core domain is less structured, with no contacts lasting for more than 40% with the exception of the contact between His179 and Leu383 (underlined in Table S2), present for 67% of simulation time. The REG region in the C-terminal domain forms very stable interactions with TAD1 in the N-terminal domain. In detail, Ser6 and His368 forms a direct hydrogen bond for more than 91% of simulation time (bold in Table S3), while Gln5-Thr377 and Ser6-Ser371 contacts are present for 74 and 68% of simulation time, respectively (underlined in Table S3). It is worth noting that helix 1 in the DBD is at the center of a network of interactions that involves both the TADs and the REG regions, acting as a sensor to their structural changes.

Regarding the intra domain interactions, the core domain shows a rich network of very stable hydrogen bonds (87 and 23 hydrogen bonds with residence time greater than 70 and 98%, respectively, see Table S4), in line with its structural role. The high flexibility of the N- and C-terminal domain does not allow for a large number of stable intra domain hydrogen bonds. However 7 and 5 hydrogen bonds are observed in the N- and C-terminal domains, respectively, with residence time greater than 70% (Tables S5 and S6). Note that while the N-terminal intra domain bonds are located all along the domain, 4 out of the 5 stable interactions of the C-terminal domain bind REG residues, the fifth bond being between Arg335 and Glu339 of the TET region.

Discussion

p53 seems evolved from a common ancestor of the p63/p73 proteins, from which a new gene, specifically dedicated to DNA damage response, with less exons, and reduced numbers of residues, has merged.^114-116 Although p53 and p63/p73 diverged during evolution, their relationship is still very strong and relevant. Both in cancer progression, metastasis, and in physiological development, all members of the family interact and regulate each other.^117,118 This is, for example, evident in epidermis,¹¹⁹ as well as in DNA damage response.¹²⁰ The p53 protein is highly conserved in its structure from C. elegans, D. melanogaster to H. sapiens.^121-126 While the DBD domain is highly conserved among vertebrates and invertebrates, the C terminus varies, resulting in a change from dimeric structure to a tetramer in the vertebrates.^127-129 The more ancient members of the family include p73, involved in cancer,¹³⁰ neurodevelopment,^131,132 and aging,¹³³ and p63, involved in epidermal development,^119,134-136 cancer,^137-141 reproduction,¹⁴² and heart development.¹⁴³ Understanding the structural restrain of its structure is pivotal to understand the function of p53^144-146 as well as its potential therapeutic exploitation.^147,148 The regulation of p53 protein half-life is crucial to his function^149,150 and, consequently, for cancer progression.^151-153 This proteosomal degradation is indeed a powerful therapeutic target.^154-158

The full-length p53 tetramer, bound to DNA, acquires different quaternary conformations⁸² where the C-terminal and DBD directly interact; the N-terminal seems to only weakly associate with DBD, and no direct interactions between N-terminal and C-terminal are observed. Additional conformations were detected at the monomeric level. In vitro, there is a realistic possibility for hetero-tetramer formation among the p53/p63/p73 family members.¹⁵⁹ The p53 tetramer seems to bind with the DBD REs with symmetric 10 bp sequences, in 4 classes of structures, consistent with a model where p53 slides along DNA via its C termini while the DBD hopping on/off searches for the responsive Res.⁸¹ The p73 tetramer is somehow similar, but it seems much more sensitive to spacer length.¹⁶⁰ Limited evidence is available on the physical interaction between p53, or mutant p53, with the other member of the family, namely p63. There is, however, a robust evidence of function interaction and immunoprecipatation.^140,161,162 To approach this question, we first created a model for the full-length p53 and performed a full structural and dynamic validation by molecular dynamics.

The in silico results show a core domain with a stable structural role, while intrinsically flexible N- and C-terminal domains explore a large conformational space in search of compatible partners, i.e., other proteins in the case of the N-terminal domain; DNA and pX3 proteins in the case of the C-terminal domain. The tools implemented by these p53 domains to accomplish their biological goal are great flexibility (Fig. 3), correlated movements (Fig. 4A), charged surfaces (Fig. 5), and variable secondary structure (e.g., the MDM2 recognition helix in the N-terminal domain, see Figs. 2 and 5C; the tetramerization helices in the C-terminal domain). In particular, electrostatic potential analysis and per-residue RMSF results (Figs. 5A and 3) are compatible with the proposed model,⁸¹ in which p53 interacts with DNA through its C-terminal domain, while Lys120 acts as a sensor, capable of triggering the conformational transition toward the specific binding of DNA Res.¹¹²

Despite the great conformational space explored by the N- and C-terminal domains as compared with the core domain (demonstrated by the free energy plot of Fig. S3A vs. S3B), they maintain stable inter domain local interactions (Fig. 4C and D; Table S1–3). It is particularly intriguing that the Zn coordinated helix H1 is at the center of these inter-domains interactions both with the N-terminal domain through His178 (Fig. 4D; Table S1) than with the REG region in the C-terminal domain through His179 (Table S2). At the same time H1 through the Zn coordinated cysteine residues is able to sense conformational changes in the DBD structure, both involving the β sandwich or the loop–sheet–helix regions. It is therefore tempting to hypothesize that this helix plays a relevant role in the orchestration of the complex conformational changes needed to accomplish p53 biological functions.

Materials and Methods

The model of the full-length p53 was built using the I-TASSER server service¹⁶³ from the human amino acid sequence. Three template structures were used (PDB id: 3q01,⁹¹ 1ycs,⁹² and 2fej⁹³) for core domain prediction, and the ab initio methodology was used for the unstructured domain fragments (TAD residues 1–60 and C-terminal residues 360–391). The protein structure prediction was made with medium confidence (C-score −2.53; TM-score 0.42 −0.14) in order to consider the intrinsic flexibility of the TADs and REG. We were not interested, in fact, in obtaining the best structure for these regions, but a realistic starting one to be sampled by the MD simulation. Nevertheless, the most important 2 short α helices in the TADs, which are major involved in the “lipophilic” protein–protein interactions (e.g., p53-MDM2¹¹¹ and p53-Bcl-2¹⁶⁴), are present in the final model.

Forcefield parameters for the Zn ion 4-ligand coordination interface were obtained from Lu and co-authors,⁹⁶ who performed a MD study of the core domain. All simulations were performed using Gromacs package v4.6¹⁶⁵ with ff99SB-ILDN force field¹⁶⁶ and periodic boundary conditions. The starting structure has been immersed in a periodic box of TIP3P water model,¹⁶⁷ which extended from 13 Å from the solute and neutralized adding sodium ions. The PME method was used to treat the long-range electrostatics.¹⁶⁸ Bond lengths involving bonds to hydrogens were constrained using LINCS algorithm.¹⁶⁹ A time step of 2 fs was used. The conformational sampling was done at a temperature of 300 K using the V-Rescale algorithm. The equilibration procedure involves: (1) 2 rounds of minimizations (1500 iteration each) and dynamics (25 ps each) of the solvent and sodium ions in the bulk solvent, keeping the solute constrained to its initial position, with decreasing force constants of 500, 100, 300, 50 kcal/(mol Ų); (2) 4 rounds of 2000 steps of minimization of whole system, where the solute restraint was kept as 100, 50, 25, and 5 kcal/(mol Ų); (3) an unrestrained minimization of the whole system. Finally the system was heated to 300 K at constant volume and equilibrated for 300 ps at constant pressure. The production phase was started at this stage and continued up to 850 ns. The conformations were collected in the trajectories at intervals of 2 ps.

The electrostatic potential was calculated with APBS software,¹⁷⁰ while all the other analysis were performed using the Gromacs tools. Default values of g_hbond, the Gromacs tool for hydrogen bond analysis, were used for cut-off distance and angle: i.e., 3.5 Å and 30 °. The plots were generated using VMD¹⁷¹ and POV-Ray (http://www.povray.org/).

Glossary

Abbreviations:

DBD

DNA binding region

ED

essential dynamics

MD

molecular dynamics

TET

tetramerization domain

PRO

proline rich region

RE

responsive element

RMSD

root mean square deviation

RMSF

root mean square fluctuations

SAXS

small-angle X-ray scattering

TAD

transactivation domain

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.

Supplemental Materials

Supplemental materials may be found here: 
www.landesbioscience.com/journals/cc/article/26162