Coarse-Grained Simulations of Phosphorylation Regulation of p53 Autoinhibition

Abstract

Intrinsically disordered proteins (IDPs) are key components of cellular signaling and regulatory networks. They frequently remain dynamic even in complexes and thus rely on potentially subtle shifts in the disordered conformation ensemble for function. Understanding the molecular basis of these fascinating mechanisms of IDP function and regulation requires detailed characterization of dynamic ensembles in various biologically relevant states. Here, we study the phosphorylation dependence of the dynamic interaction between the N-terminal transactivation domain (NTAD) and DNA-binding domain (DBD) of tumor suppressor p53, which plays a key role in autoinhibition and regulation of p53 activation or termination during various stages of stress response. By extending the hybrid-resolution (HyRes) coarse-grained (CG) protein force field to model phosphorylated sidechains, we show that HyRes simulations accurately recapitulate the effects of phosphorylation on the p53 NTAD/DBD interactions. The simulated ensembles show that phosphorylation of Thr55 as well as Ser46 enhances dynamic NTAD/DBD interactions and further induces conformational shifts that promote trans interactions between two p53 dimers to drive dissociation from DNA. These CG simulations thus provide a strong molecular basis in support of previous experimental studies suggesting the central role of dynamic interactions of disordered domains and phosphorylation in function of p53. Success of the current study also suggests that HyRes provides an efficient and viable tool for studying dynamic interactions and post-translational modifications in IDPs function and regulation.

Introduction

Intrinsically disordered proteins (IDPs) play crucial roles in cellular regulation and signaling^1–3 and are frequently associated with various human diseases⁴. Unlike structured proteins, IDPs lack a stable tertiary structure under physiological conditions, primarily due to their low sequence complexity that is generally enriched with charged and hydrophilic residues^1–3. Unbound IDPs exist as dynamic and presumably heterogeneous ensemble of disordered conformations with various degrees of secondary structure propensities. Upon specific binding, IDPs often undergo binding-induced folding transitions^5–8. It has also been recognized that IDPs can remain unstructured even in specific complexes and functional assemblies^9–14. The dynamic nature of the disordered ensembles in the unbound and often bound states arguably allows IDPs to respond sensitively to various cellular stimuli, such as the binding of ligands and cofactors, changes in cellular environments, and post-translational modifications^15–18. Multiple cellular signals can be naturally integrated through cooperative responses of the entire dynamic ensemble^19–21. To understand these fascinating mechanisms of IDP function and regulation, we need to integrate experiment and molecular modeling to resolve the molecular detail of dynamic ensembles of IDPs under various biologically relevant states.

The tumor suppressor protein p53 is considered one of the most important proteins in cancers and plays a central role in cellular stress responses^22,23. The full-length p53 protein is composed of several distinct domains as shown in Figure 1A, including an intrinsically disordered N-terminal transactivation domain (NTAD) and proline-rich domain (PRD), a structured DNA-binding domain (DBD), a short disordered region containing the nuclear localization signal (NLS), a folded tetramerization domain (TET), and a disordered C-terminal regulatory domain (CTD)²⁴. While DBD directly binds with specific genes for transcription activation activities, disordered domains interact with numerous proteins from various signaling pathways that regulate the p53 level and activity^25–28. The p53 NTAD, in particular, is one of the most studied IDPs with nontrivial transient local and long-range structural features^29–33 (Figure 1B). It is characterized by two activation motifs, AD1 and AD2, which have substantial residual helical propensities and can fold into short helices in various complexes^34–37. The disordered nature of NTAD is considered an essential property for p53’s role as a major hub protein^5,8. How the disordered ensemble responds to multi-site phosphorylation and cancer-associated mutations arguably governs the preference of p53 interactions with various cofactors that ultimately control its stability and activity^27,28,38,39.

Figure 1. Sequence and structure of p53.

A) Function domains of p53. Residue S121 in DBD is the paramagnetic spin labelling site in PRE experiments⁴². B) Sequence of p53 NTAD. Negatively charged residues are colored in blue. C) The structure of p53 tetramer, which is a dimer of dimers (colored in gray and cyan). For the monomer on the top left, the PRD and NTAD are colored in green and red, respectively. The negatively charged NTAD residues and positively charged DBD residues are highlighted using blue and red beads, respectively. The two key phosphorylation sites studied here (Ser46 and Thr55) are also labeled in panels B and C.

Besides mediating numerous intermolecular interactions, negatively charged NTAD also interacts dynamically with the positively charged DNA-binding interface of DBD and competes with DNA for binding to DBD^40,41. This intramolecular interaction is regulated by phosphorylation and contributes to the control of p53 activation or termination during various stages of stress response⁴². NMR experiments show that phosphorylation of Thr55 in AD2 enhances the binding of NTAD to DBD and disrupts cooperative binding to p53-specific DNA targets, especially those with weaker p53 binding affinities⁴². The autoinhibition is further enhanced by phosphorylation of Ser46, leading to biphasic behavior in the concentration dependence of DNA binding as assessed by fluorescence anisotropy⁴². Dephosphorylation of Thr55 by protein phosphatase 2A releases the autoinhibition and activates the transcription during damage of DNA^43–45. At late stages of DNA damage, Thr55 would be re-phosphorylated by TAF1 kinase to drive DNA disassociation and terminate the stress responses⁴⁶. While these experiments have firmly established the role of Thr55 phosphorylation as a master switch to activate or terminate p53-mediated transcription activities, much remains to be known about how phosphorylation regulates the dynamic NTAD/DBD interaction at the molecular level.

Paramagnetic relaxation enhancement (PRE) measurements suggest that pThr55 and pSer46 mainly enhance the interactions of the AD2 of NTAD with DBD⁴². However, these experiments do not resolve NTAD/DBD interactions within the same monomer or between NTAD of one monomer with DBD of another within the dimer or tetramer of p53. It has been speculated that the ability to form interactions within the same dimer (cis) across the dimer-dimer interface (trans) may provide the molecular basis of regulation by Thr55 (and Ser46) phosphorylation⁴² (Figure 1C). Further resolving this mechanism requires molecular dynamics (MD) simulations to generate the disordered ensembles of p53 and examine how phosphorylation may shift the conformational equilibrium^47–51. Such a task is computationally too demanding for atomistic MD simulations given the large system size and complexity of conformational space. Instead, coarse-grained (CG) models that dramatically reduce the system size and extend the accessible length and timescales are desired⁵².

We have previously developed a hybrid-resolution (HyRes) force field specifically designed to describe disordered conformational ensembles of IDPs^53–55. In HyRes, the protein backbone is represented atomistically whereas side chains are represented with single or multiple CG beads (Figure 2A). The covalent parameters and side chain CG bead size were parameterized based on atomistic simulations to maximally reproduce the covalent geometry and flexibility, which can be important in accurate describing packing at the protein-protein interface. The effective energy function of HyRes also includes backbone hydrogen-bonding, Debye-Hückel electrostatics, and van de Waals (vdW) interactions, parameterized based on to reproduce pair-wise backbone and side chain interactions as well as the helical propensities and chain dimension of small model IDPs. The latest HyRes model, implemented on GPU, can provide a semi-quantitative description of secondary structures as well as long-range interactions of a set of IDPs⁵⁵. It has been successfully applied to study the dynamic interactions of several proteins including phase separation with success^14,54–56. Therefore, HyRes may provide an effective balance between accuracy and computational efficiency to elucidate the molecular basis of how phosphorylation regulates the dynamic interactions between p53 NTAD and DBD.

Figure 2. HyRes modeling of phosphorylated serine and threonine.

A) HyRes representation of all natural or canonical residues, which includes atomistic backbone and intermediate-resolution side chains. B) Mapping of phosphorylated Ser and Thr residues. The blue bead represents the phosphate group (bead CC). C) Distributions of the CB-CC bond length and D) CA-CB-CC bond angle in HyRes (blue traces) in comparison to those derived from atomistic simulations (red).

In this work, we have simulated the wild-type and phosphorylated p53 tetramers using the HyRes force field to study the molecular mechanism of the p53 autoinhibition. The generated conformational ensembles were validated by comparing to the available NMR PRE experimental data⁴². This was followed by detailed conformational analysis to examine how various local and long-range conformational properties of NTAD as well as the dynamic NTAD/DBD interaction depended on phosphorylation of Thr55 and Ser46. The results provide a detailed description of how pThr55 and pSer46 enhance AD2-DBD interactions and promote the trans mode of NTAD/DBD interactions within the tetramer for autoinhibition. This study shows that HyRes can capture complex IDP dynamic interactions and provides an important tool for molecular simulations of disordered proteins and understanding their roles in function and diseases.

Materials and Methods

Parameterization for Phosphorylated Sidechains

To model the effects pThr55 and pSer46 on p53, we first developed an HyRes model for phosphorylated serine and threonine, by including an additional CC bead for the phosphate group (Figure 2B). The CC bead carries −1 net charge. The parametrization of bond, angle and vdW parameters followed the HyRes convention^53,54 (also summarized above). We first performed 100-ns atomistic simulations of serine and threonine dipeptides in CHARMM22⁵⁷ with GBSW implicit solvent⁵⁸. The equilibrium pseudo bond lengths (CA-CB and CB-CC) and bond angle (CA-CB-CC) were set to the averaged values in the all-atom simulations (Table 1). The bond and angle force constants k_b and k_θ were set using the standard deviations derived from GBSW trajectories, such that the CG model would roughly reproduce the flexibility of these sidechains in atomistic simulations. The vdW radii of CC beads were set to reproduce the volume of the corresponding group of atoms from atomistic representation, whereas ε was taken form that of the glutamate acid side chain bead (Table 2). For the CB bead of pSer, ε_i and vdW radius were taken from those of alanine; for pThr CB, the vdW parameters were taken as the average of alanine and valine CB values. The distributions of the bonds and angles of the phosphorylated serine and threonine dipeptides calculated from 20 ns trajectories using the above HyRes model are shown in Figure 2C and 2D, in comparison to those derived from GBSW simulations. The results show that HyRes is able to reproduce the covalent geometry and flexibility for both pSer and pThr.

Table 1.

Bonded parameters of phosphorylated residues, k_b/k_θ in kcal/mol, b₀ in Å, and θ_o in °.

Bond/Angle	k b /k θ	b0/θo	Bond/Angle	k b /k θ	b0/θo
pSer CA-CB	262.9	1.61	pSer N-CA-CB	20.0	110.3
pSer CB-CC	200.0	2.53	pSer CA-CB-CC	20.0	132.8
pThr CA-CB	200.0	2.03	pThr C-CA-CB	20.0	101.2
pThr CB-CC	200.0	3.20	pThr N-CA-CB	20.0	98.4
pSer C-CA-CB	20.0	112.3	pThr CA-CB-CC	20.0	72.6

Table 2.

vdW parameters of pSer and pThr side chains. The CC bead has −1.0 net charge.

Bead type	εi (kcal/mol)	vdW radii (Å)
pSer CB	−0.15	2.12
pThr CB	−0.22	2.44
pSer and pThr CC	−0.07	2.25

MD Simulation Protocols

To construct the initial structural model of p53 tetramer, we first mapped the X-ray structure of the p53 DBD tetramer (PDB ID: 3KMD⁵⁹) onto HyRes representation using the “at2hyres” script (https://github.com/mdlab-um/HyRes_GPU). A disordered conformation of NTAD and PRD (residues 1–94) was generated using CHARMM^60,61 and then attached to p53 DBD using MODELLER⁶². The C-terminal domains after DBD were not included in the current study. The tetramer structure was then used to initiate a high-temperature simulation at 800 K to randomize the N-terminal domain structures. The p53 DBD C⍺ atoms were harmonically restrained with a force constant of 1.0 kcal/(mol Ų). In addition, pulling forces were imposed on the N-termini during high-temperature simulations to generate random conformations with various degrees of separation between NTAD from DBD. The final set of 8 initial conformations of the tetramer are shown in Figure S1. Note that using this diverse set of initial structures for independent simulations minimizes the impacts of initial structures and ensure comprehensive sampling of dynamic NTAD/DBD interactions. All production simulations were performed using CHARMM/OpenMM^60,63,64 on GPU. Langevin thermostat was used with a friction coefficient of 0.2 ps⁻¹ and reference temperature of 300 K. Lengths of all hydrogen-containing bonds were constrained by the SHAKE⁶⁵ algorithm, and the MD integration time step was 4 fs. Non-bonded interactions were smoothly switched off from 1.6 to 1.8 nm. For each construct (wild type, pThr55, and pSer46 pThr55), eight replicates were run from different initial conformations (Figure S1) for 400 ns with snapshots saved every 20 ps. Atomistic ensembles of NTAD alone generated using a99SB-disp were taken from the previous work³³.

Data Analysis

The generated trajectories were analyzed using in-house scripts utilizing MDAnalysis⁶⁶ and MDTraj⁶⁷ libraries. In all the analyses the last 300ns of the simulations are used, unless otherwise specified. The helicity probability was calculated using the standard Dictionary of Secondary Structure of Proteins (DSSP) protocol⁶⁸. For the tetramer, the average helicity was calculated across all frames for the four NTAD monomers in each replica run. Subsequently, the mean and standard deviation were computed across all eight independent runs.

To validate the dynamic ensembles from HyRes simulations, theoretical PRE values were calculated between all NTAD and PRD residues and the paramagnetic labeling site (S121 in DBD⁴²; see Figure 1A),

$$\frac{I_{ox}}{I_{red}}=\frac{R_2\text{exp}\left(-R_2^{sp}t\right)}{R_2+R_2^{sp}},$$ (1)

$$R_2^{sp}=k⟨r^{-6}⟩\left(4{\tau }_c+\frac{3{\tau }_c}{1+{\omega }_H^2{\tau }_c^2}\right),$$ (2)

where r is the distance between C⍺ atoms of the paramagnetic spin-labeled residue and other residues in NTAD and PRD. The other parameters are k = 1.23 × 10⁻³² cm⁶/s², ω_H = 600 MHz, τ_C = 3.3 ns, R₂ = 16 s⁻¹, and t = 9.8 ms, taken from our previous work³³. Note that for each snapshot of the tetramer, a total of 16 distances needs to be calculated between four monomers and four spin-labeled S121 on DBDs. The ensemble average $⟨r^{-6}⟩$ in Eq. 2 includes all distances.

The principal component analysis (PCA) was performed utilizing the scikit-learn⁶⁹ library using only C⍺ atoms of NTADs. The first two principal components were first determined using conformations pooled from all three sets of simulations (wild type, pThr55, and pSer46 pThr55). The same two principal components were then used to derive the conformational distributions of the three different constructs. All molecular visualization was prepared using VMD⁷⁰.

Results and Discussion

Phosphorylation effects on p53 NTAD conformational equilibrium

To assess the ability of the coarse-grained force field to capture relevant IDP dynamics, we first compare the HyRes conformational properties with atomistic ensembles generated using a99SB-disp, one of the best explicit solvent atomistic protein force fields for IDPs^33,71. As summarized in Figure 3, the radius of gyration (R_g) distributions for NTAD alone are highly consistent between HyRes and a99SB-disp (red vs blue traces), with much smaller errors indicating superior convergence with HyRes. The average R_g for NTAD is 25.8 Å in HyRes, compared well to that of 26.2 Å in a99SB-disp. HyRes also describes the helical propensities of NTAD alone well (Figure 3C), supporting that HyRes is well equipped to describe the dynamic ensemble of p53 NTAD. Note that a99SB-disp slightly over-estimates the helicity of p53 NTAD compared to the experimental results^33,54, whereas HyRes interestingly better reproduce experimental distributions. Within the tetramer context, NTAD forms dynamic interactions with DBD, mainly driven by nonspecific electrostatic interactions^42,54. This interaction has modest impacts on the chain dimension (Figure 3A, green vs blue traces), but appears to significantly reduce the helicity in the AD1 region (Figure 3C). Interestingly, phosphorylation of Thr55 and Ser46 has minimal impact on either the local and global structures of NTAD (Figure 3B and 3D). This is not necessarily surprising considering the transient and dynamic nature of NTAD/DBD interactions. There is only a very modest decrease in the residual helicity in the AD2 region, which is the primary region affects by phosphorylation indicated by PRE experiments⁴² as well as the current HyRes simulations (see below). Chemical shift analysis of the pThr55 and p53 construct also suggest no substantial perturbation except near the phosphorylation site, supporting a lack of significant changes in NTAD residual helicity⁴².

Figure 3. Residue helicity profiles and R_g distributions of p53 NTAD in various constructs.

A) Probability distributions of R_g of NTAD calculated using HyRes (blue; NTAD alone) vs a99SB-disp (red; NTAD alone) vs HyRes (green; NTAD in tetramer). B) Probability distributions of R_g of NTAD in p53 tetramers without and with phosphorylation of Thr55 and Ser46. C) Residue helicity profiles calculated using a99SB-disp (NTAD alone) and HyRes (NTAD alone or in tetramer). D) Residue helicity profiles of NTAD in tetramers without (green: wild type) or with phosphorylation (blue: pThr55; red: pSer46pThr55) calculated using HyRes.

HyRes recapitulates the effects of phosphorylation on dynamic NTAD/DBD interactions

To determine if HyRes correctly captures long-range intramolecular interactions of p53 tetramer, we calculated the theoretical PRE profiles of the NTAD and PRD residues induced by a paramagnetic spin label on residue S121 on DBD⁴². PRE experiment can detect long-range transient interactions between the paramagnetic spin label and the rest of the protein up to ~ 35 Å⁷², and is thus uniquely suitable for critical validation of the simulated ensemble. We have previously shown that HyRes is able to reproduce the experimental PRE profiles measured on NTAD alone, at a level that is comparable to a99SB-disp⁵⁴. In Figure 4 and Figure S2, we compared the PRE profile calculated from HyRes simulation of the p53 tetramer (red trace) with experimental values (red bars). There is a general agreement between the two profiles, even though there is clearly an overestimation of the contact between N-terminal segment (residues 1–25) in HyRes generated ensemble. Note that there are 6 Asp/Glue residues in this region, which is comparable to a total of 7 Asp/Glu’s within residues 40–60. As such, the N-terminus of NTAD could be expected to be similarly attractive to DBD, leading to comparable PRE effects around residues 1–25 and 40–60. The experimental observation that the N-terminal segment (especially residues 1–12) is rarely in the proximity of DBD suggests that NTAD may be stiffer than HyRes predicts, such that the entropy cost of bringing the N-terminus of NTAD to DBD should have been much greater. This may reflect a limitation of HyRes, and likely CG protein models in general.

Figure 4. Calculated and measured PRE profiles of p53 NTAD and PRD region with and without phosphorylation.

The experimental profiles are shown in red (wild type) and blue (pThr55) bars. Theoretical profiles were calculated sets of 8 independent HyRes simulations (see Methods). The error bars are not shown for clarity; see Figure S2 for error bars associated with each calculated PRE profile.

We then evaluate if the simulated ensembles could recapitulate the effects of pThr55 and pSer46. The results show that HyRes correctly predicts pThr55 enhances the NTAD-DBD interaction mainly through the AD2 region, leading to significant increases in PRE effects (Figure 4, blue vs red traces compared to blue vs red bars). Interestingly, the simulated ensemble reveals that increased AD2 interactions with DBD is achieved at the expense of reduced AD1 interactions with DBD, leading to weaker PRE effects in the N-terminal region of NTAD. This is exactly what was observed experimentally (blue bars in Figure 4). The simulation thus supports the notions that AD1 and AD2 compete with each other for dynamic binding to DBD and that pThr55 shifts the balance towards AD2 binding⁴². Additional phosphorylation of Ser46 further strengthens the effects of pThr55, leading to stronger PRE effects in the AD region and further reduced PREs in the N-terminus (Figure 4, green trace). This is consistent with functional and physical studies of the double phosphorylated construct⁴². Analysis of the probabilities of direct contacts between DBD and NTAD/PRD domains show consistent trends with those observed in PRE profiles. In the AD1, the contact frequencies are higher for the wild-type and the lowest for the pSer46pThr55 construct (Figures S3A). However, AD2 exhibits higher DBD contact probabilities in pThr55 and pSer46pThr55 tetramer, especially for the phosphorylated residues and neighboring charges. Contacts involving the PRD only show minor increases with pSer46 or pThr55. For the DBD, residue contact frequencies remain similar in all three p53 constructs (Figures S3B). The ability of HyRes to recapitulate these nontrivial PRE effects of pThr55 and pSer46 suggests that the simulated ensembles are likely realistic and biologically relevant, allowing further analysis of the molecular basis of phosphorylation regulation of p53 autoinhibition.

Phosphorylation promotes trans-dimer NTAD-DBD contacts to enhance autoinhibition

PCA was performed to further analyze the conformational distributions of the dynamic NTAD-DBD interactions within each of the three p53 tetramer constructs. For this, we first derived two common principal components (PCs) using snapshots collected from all three sets of HyRes simulations (see Methods). The results show that, for the wild-type p53 tetramer, the conformational space is predominantly occupied by cis-like states (Figure 5A and Figure S4). That is, in the absence of phosphorylation, the NTAD and DBD interact primarily within the same dimer (cis). Whereas in the pThr55 p53 tetramer, there is a shift in the conformational equilibrium towards more states where NTADs from one dimer interacts DBDs from the other dimer (trans) (Figure 5B and Figure S5). Among the 7 major basins of pThr55 p53 tetramer, only one is cis-like and the rest is either mixed (both cis and trans interactions present) or trans-like. The shift towards trans-like states is even more prominent in the pSer46/pThr55 p53 tetramer, where the population of trans-like and mixed states continue to increase (Figure 5C and Figure S6). Taken together, the simulated ensembles provide a direct support at the molecular level for the proposal that phosphorylation of Thr55 and Ser46 may promote the trans-like NTAD-DBD interaction within the p53 tetramer to further strengthen autoinhibition⁴².

Figure 5. PCA analysis of dynamic NTAD-DBD interactions in various p53 tetramer constructs.

(A) Wild-type p53 tetramer, where all major states are cis-like. B) pThr55 p53 tetramer, where new states with trans NTAD-DBD interactions start emerging. For the 7 major basins shown, states 1, 3, 4 are trans-like, states 2, 6, and 7 are mixed, and only state 5 is cis-like. (C) pSer46pThr55 p53 tetramer, where there is a further increase in populations of trans-like states. With the 9 major basins, states 1, 3, 5, 7, and 9 are trans-like, states 4, 6, and 8 are mixed, and states 2 and 6 are cis-like. For the representative snapshots, DBDs of two dimers are shown in yellow and gray. Only one of the NTADs from the gray dimer is shown in red cartoon for clarity. Locations of (p)Thr55 and (p)Ser46 are marked using green spheres and paramagnetic spin labelling site S121with yellow beads. Contours are drawn at every kT. See Figure S4–6 for representative snapshots for all major basins of all three p53 constructs.

Conclusion

In this study, we have extended the HyRes protein force field to model phosphorylated side chains and demonstrated the capability of the force field to generate biologically relevant dynamic conformational ensembles of the p53 tetramer. The simulated ensembles correctly recapitulate nontrivial effects of pThr55 and pSer46 on local structures and transient long-range interactions between NTAD and DBD, providing new insights into how phosphorylation modulates the disordered conformational equilibrium in p53 autoinhibition. Specifically, pThr55 and pSer46 do not only enhance the association of the AD2 region with the DNA binding interface of DBD, through nonspecific electrostatic interactions, but further promote the trans interaction mode where NTADs from one dimer preferentially engage with DBDs from the other dimer of the p53 tetramer. The conformational shift towards the trans mode further stabilizes the DNA-unbound state of p53 tetramer and strengthens autoinhibition. The current work also supports the potential of HyRes as a powerful tool for studying the conformational landscapes and dynamic interactions of IDPs in physiological and pathological pathways.

Supplementary Material

Supplementary figures showing the initial conformations, error bars of calculated PRE profiles, contact probabilities, and illustration of NTAD-DBD interactions of all major states or various p53 tetramer constructs. The online version contains supplementary material available at https://doi.org/xxxx