Reconciling ASPP-p53 binding mode discrepancies through an ensemble binding framework that bridges crystallography and NMR data

Abstract

ASPP2 and iASPP bind to p53 through their conserved ANK-SH3 domains to respectively promote and inhibit p53-dependent cell apoptosis. While crystallography has indicated that these two proteins employ distinct surfaces of their ANK-SH3 domains to bind to p53, solution NMR data has suggested similar surfaces. In this study, we employed multi-scale molecular dynamics (MD) simulations combined with free energy calculations to reconcile the discrepancy in the binding modes. We demonstrated that the binding mode based solely on a single crystal structure does not enable iASPP's RT loop to engage with p53's C-terminal linker-a verified interaction. Instead, an ensemble of simulated iASPP-p53 complexes facilitates this interaction. We showed that the ensemble-average inter-protein contacting residues and NMR-detected interfacial residues qualitatively overlap on ASPP proteins, and the ensemble-average binding free energies better match experimental KD values compared to single crystallgarphy-determined binding mode. For iASPP, the sampled ensemble complexes can be grouped into two classes, resembling the binding modes determined by crystallography and solution NMR. We thus propose that crystal packing shifts the equilibrium of binding modes towards the crystallography-determined one. Lastly, we showed that the ensemble binding complexes are sensitive to p53's intrinsically disordered regions (IDRs), attesting to experimental observations that these IDRs contribute to biological functions. Our results provide a dynamic and ensemble perspective for scrutinizing these important cancer-related protein-protein interactions (PPIs).

Fig 1. Structural background of p53-ASPP PPI.

(A) Domain organizations of p53 and ASPP proteins. p53’s DBD and ASPP’s ANK-SH3 are folded domains whereas the remaining sequences are intrinsically disordered. (B) Crystal structures of ASPP2-p53 complex (PDB 1YCS [5]) and iASPP-p53 complex (PDB 6RZ3 [6]). Solution NMR studies [4] also suggest a iASPP-p53 binding mode that is different from PDB 6RZ3. (C) In this study, we combined all-atom and coarse-grained (CG) MDs plus free energy calculations to explore the binding mechanisms between three p53 constructs (p53_DBD p53_P-DBD and p53_P-DBD-L) and the ANK-SH3 domains of ASPP2/iASPP.

Fig 2. Examining the interactions between p53’s C-terminal linker and iASPP’s RT loop.

(A) Samplings of p53’s C-terminal linker (yellow) and iASPP’s RT loop (red) from 3 ×1 μs all-atom MDs starting from the PDB 6RZ3 binding mode. (B) Starting from PDB 1YCS binding mode. For each case, MD trajectories were concatenated and were aligned on p53’s DBD against the shown respective reference structure. The X and Y Cartesian coordinates of the protein backbone atoms were used for the projection. ASPP and p53 are colored green and cyan, respectively. The gray lines depict the average structures of protein backbone.

Fig 3. Martini CGMD simulated p53-ASPP binding complexes.

(A) Comparison of NMR-detected residue chemical shift changes and Martini CGMD sampled residue contact frequency on the ANK-SH3 domains of ASPP. Martini contact frequency was calculated by first summing all contacts involving each ASPP residue and divided by the total number of frames, followed by normalization to the largest value. ASPP residues that are involved in interacting with p53 as determined by crystallography were marked by differently-colored asterisks: blue from PDB 1YCS (ASPP2-p53_DBD), red from PDB 6RZ3 (iASPP-p53_DBD), and gray from the NMR-suggested 1YCS-like binding mode for iASPP. (B) Mapping the data shown in panel A onto the ANK-SH3 domain structures. (C) Martini CGMD sampled p53_P-DBD-L-iASPP complexes. The normalized atoms densities of p53 in the 3D space around ASPP, and the projections of samplings onto the XY plane were shown to the left and right, respectively. The p53’s C-terminal linker (yellow) is interacting with iASPP’s RT loop (red).

Fig 4. Energetic characterizations of Martini CGMD simulated binding complexes.

(A) PMF curves of the dissociation process between p53_DBD and ASPPs, starting from the crystal-structure binding modes. The PMF errors were estimated by the Monte Carlo bootstrapping procedure of the WHAM program. PMF-derived binding free energy (ΔG) was calculated by defining the bound state at RC = 1.5 Å and the unbound state at RC = 14.5 Å. The experimentally measured ΔGs were shown as red dashed lines. (B) Martini CGMD simulated binding between p53_DBD and ASPPs. The trajectories are projected onto the distance (COM of p53_DBD to COM of ANK-SH3 domain), and the RMSD (with respect to crystal structures) axes. The color scale means frequency of binding. The red shaded areas highlight the binding poses that are close to crystal structures, and those resemble exactly to crystal structures were superimposed onto crystal structures. (C) Overall atom density distribution of p53_DBD (cyan) around ASPP (green) in the 3D space. (D) Comparison of experimentally measured ΔGs, calculated ΔGs based on crystal structures, and ΔGs calculated from multiple representative Martini complexes (detailed PMF curves are shown in S7 Fig). t-tests were performed to assess the statistical significance.

Fig 5. Effects of p53’s IDRs on p53-ASPP binding.

(A) First contact formation times of ASPPs binding to p53_DBD and p53_P-DBD-L. For p53_P-DBD-L, besides the normal definition of first contact as “any ASPP residue to any p53 residue”, we also counted the first contact defined as “any ASPP residue to residues only from p53 DBD domain” (hatched bars). For ASPP2-p53_P-DBD-L binding, an additional set of 50 × 4 μs Martini CGMD was performed in which p53’s IDRs were rigidified by constraints (green bar, using the normal definition of first contact). (B) Average time-dependent number of contacts between ASPP and different p53 domains of p53_P-DBD-L calculated from 50 × 4 μs Martini CGMD simulations. The contacts based on p53_DBD-ASPP simulations were also shown as gray bars. (C) Projections of Martini CGMD sampled ASPP-p53 complexes onto 2D plane using the sketch-map dimentionality reduction method. We defined “prime” and “secondary” binding modes to refer to the most highly-sampled states, and the less sampled but still have considerable population states (if present), respectively. Representative complexes were shown right to the sketch-map projection.

Fig 6. Assessing the refolding of Martini CGMD sampled ASPP-p53 complexes.

(A) Refolding is gauged by a RMSD metric, namely, starting from the moment when ASPP and p53 first contact, the subsequent complexes’ RMSDs with respect to the first contact complex were calculated. (B-C) Five Martini CGMD trajectories were randomly selected to assess the refolding for ASPP2 and iASPP in the binding of p53_DBD, respectively.