Effect of linker on the binding free energy of stapled p53/HDM2 complex

Inactivation of the tumor suppressor p53 resulting from the binding with a negative regulator HDM2 is among the predominant defects in human cancers. p53-mimicking peptides whose conformational and proteolytic stability is enhanced by an all-hydrocarbon staple are being recognized as promising anticancer agents for disrupting the p53–HDM2 binding and reactivating p53. Herein, we conduct a computational modeling and thermodynamic characterization of stapled p53/HDM2 complex via molecular docking, simulations, and binding free energy analysis. The binding thermodynamics analysis is done based on the end-point calculation of the effective binding energy—a sum of the direct peptide–protein interaction energy and the dehydration penalty—and on its decomposition into contributions from specific groups constituting the complex. This allows us to investigate how individual amino acids in the stapled p53 and HDM2 contribute to the binding affinity. We find that not only the epitope residues (F19, W23 and L26), but also the hydrocarbon linker of the stapled p53 impart significant contributions. Our computational approach will be useful in designing new stapled peptides in which the staple location is also optimized to improve the binding affinity.


Introduction
The p53 protein is a transcription factor regulating cell cycle and apoptosis in response to DNA damages and cellular stresses [1]. It plays a critical role for maintaining the genome integrity and preventing the development of tumor [2]. The p53 levels in normal cells are controlled by negative regulator proteins such as HDM2 that neutralize the p53 transactivation activity via a direct binding interaction [3]. The loss of p53 activity resulting from the binding with overexpressed HDM2 is among the predominant defects in human cancers [4]. The reactivation of p53 by perturbing the p53-HDM2 binding has therefore been regarded as a promising approach for suppressing tumor growth in cancer cells [5][6][7].
The p53-HDM2 interaction occurs primarily between the short α-helical segment of p53 and the hydrophobic pocket of the HDM2 surface. In particular, three residues (F19, W23 and L26) within this segment form critical contacts with HDM2 [8]. p53-mimicking peptides incorporating these epitope residues will hence serve as potential anticancer agents that reactivate p53 by driving it out from the interaction with HDM2. In this context, there is recently a growing interest in stapled peptides in which an α-helical conformation appropriate to binding with HDM2 is rigidified through an introduction of an all-hydrocarbon linkage connecting helix forming residues [9][10][11][12][13]. Furthermore, it has been demonstrated that the stapling also enhances the proteolytic stability and promotes cell permeability, which are crucial for in vivo therapeutic activity [14]. Systematic mutational analysis, including the optimization of the staple locations, are being carried out experimentally and computationally in search of higher binding affinity and more improved potency and specificity [15,16]. Herein, we propose a computational method that is useful for designing new stapled p53-mimicking peptides (to be simply referred to as stapled p53 peptides from here on). We start from a peptide sequence and conduct a template-based modeling using an experimental structure. The stapled p53 peptide so constructed is subjected to molecular dynamics simulations to explore a representative structure in an aqueous solution and its conformational stability. The simulated stapled p53 structure is then used for a molecular docking onto the HDM2 surface. Starting from the docked complex structure, we perform molecular dynamics simulations. For the simulated complex configurations, we finally carry out thermodynamic analysis. This is done based on the end-point calculation of the effective binding free energy Δf [17]. It comprises the direct peptide-protein interaction energy (ΔE u ) and the solvation free energy contribution (ΔG solv ), Δf = ΔE u + ΔG solv . The quantity Δf is connected to the binding free energy (ΔG bind ) via ΔG bind = Δf-T(ΔS config + ΔS ext ) in which ΔS config and ΔS ext are the configurational and external entropies, respectively [18,19]. Since these entropy terms are typically negative, the favorable contributions to the binding affinity arise mainly from Δf. The formation of peptide-protein contacts, such as hydrogen bonds and van der Waals contacts, leads to favorable changes in the direct interaction energy (ΔE u < 0). However, the formation of these peptide-protein contacts involves the dehydration penalty (ΔG solv > 0). Therefore, in arguing the net contribution to the binding affinity, it is essential to analyze Δf that simultaneously takes into account both ΔE u and ΔG solv . Importantly, Δf can be decomposed into contributions from specific groups constituting the complex [20,21]. Thereby, our method allows us not only to investigate the relevance of individual amino acids, but also to quantify the contribution from the hydrocarbon linker to the binding affinity.

Modeling stapled p53/HDM2 complex
We investigated a stapled p53 peptide, referred to as sMTide-02 in Ref. [14], whose sequence is Ac-TSFXEYWALLX-NH2 (X: linker positions): two additional peptides stapled at shifted positions, Ac-TXFAEYWAXLS-NH2 (to be referred to as sMTide-02b) and Ac-XSFAEYWXLLS-NH2 (sMTide-02c) were also studied to analyze the dependence of the structural stability and binding affinity on the linker location (see Fig 1A). Each peptide structure was constructed using an experimental structure for a stapled p53 (PDB ID 3V3B) [15] as a template. The hydrocarbon linker was added with GaussView [22]. A 100 ns molecular dynamics simulation was carried out for each stapled peptide starting from the constructed structure (see below for details on the simulation). A representative peptide conformation for each system was selected based on the k-means clustering with a radius of 4.0 Å (shown in Fig  1A), which was then used in the docking onto the HDM2 surface. We employed the X-ray structure (PDB ID 1YCR) [8] for HDM2 ( Fig 1B). The docking was performed using Auto-Dock Vina [23]. Only the side chains were permitted to rate in the docking simulation. 1,000 complex structures were generated from the docking carried out for each system, and we chose the most stable structure. Molecular dynamics simulations for the stapled p53/HDM2 complexes were then carried out starting from the respective docked structures.

Molecular dynamics simulations
Simulations for the free stapled p53 peptides and the stapled p53/HDM2 complexes were carried out using AMBER16 [24]. Each system was solvated by waters and counter ions. The ff99SB-ILDN [25,26] was adopted for the peptide/protein and for water the TIP3P model [27] was used. The linker partial charges were determined from the restrained electrostatic potential (RESP) method [28] after carrying out an HF/6-31G � quantum mechanical calculation with Gaussian 09 [29]. We employed the general Amber force field [30] for the other parameters. After the standard minimization and equilibration steps, 100 ns NPT-ensemble simulations were performed at 300 K and 1 bar. A single and 10 independent runs were conducted for the free peptide and the complex, respectively.

Thermodynamic calculations
The effective binding free energy, Δf = ΔE u + ΔG solv , was computed using the simulated complex structures saved with a 1 ns interval. The direct peptide-protein interaction energy (ΔE u ) can be obtained easily from the force field parameters. The solvation term (ΔG solv ) comprises the ones for the complex and its components, ΔG solv = G solv; complex −(G solv; stapled p53 + G solv; HDM2 ), and was computed using the 3D-RISM theory [31,32].
Within the classical force field, the direct interaction energy (ΔE u ) is expressed as a sum of atomic contributions. For the solvation free energy (G solv ), we have recently developed an exact atomic decomposition method [20,21]. Thereby, the effective binding free energy (Δf) can be partitioned into contributions from constituent atoms. By an appropriate grouping of these atomic terms, individual residue and linker contributions to Δf can be obtained.

Results and discussion
We first performed a 100 ns free-peptide simulation for each of the stapled p53 peptides (sMTide-02, sMTide-02b and sMTide-02c) to examine its conformational stability in an aqueous environment. The free sMTide-02 was quite stable (the Cα root-mean-squared deviation (RMSD) to the initial structure remained <1.0 Å) during the simulation. Its overall helical structure was also maintained (Fig 2A) with an average helical content of 55%. It is well known that short helical segments taken out from globular proteins cannot keep up its secondary structure when isolated [33]. Our simulation thus confirms the significantly enhanced conformational stability of a short helical peptide brought about by the staple linker. On the other hand, the structures of the other stapled peptides (sMTide-02b and sMTide-02c) exhibited somewhat larger deviations: the Cα RMSDs to the respective initial structures increased up to 1.6 Å and 2.4 Å and the average helical contents dropped to 35% ( Fig 2B) and 39% (Fig 2C), respectively. Thus, the stability and helicity of the stapled p53 peptides depend on the linker positions.
Finally, we conducted the end-point calculations of the effective binding free energy (Δf) based on the simulated complex structures. As we stated above, Δf provides the major favorable contribution to the binding affinity. Therefore, its decomposition into specific group contributions enables us to identify critical residues. The effective binding free energies for HDM2 with sMTide-02, sMTide-02b and sMTide-02c were computed to be -18.7 ± 0.5 kcal/mol, -13.2 ± 1.0 kcal/mol, and -7.0 ± 0.8 kcal/mol, respectively (average ± standard error estimated from the respective 10 independent runs). This trend is in accord with the experimental observation for the related SAH-p53 (stabilized alpha-helix of p53) peptides [9]: The linker positions in sMTide-02, sMTide-02b and sMTide-02c corresponds to those in SAH-p53-8, SAH-p53-2 and SHA-p53-3 peptides, respectively, whose binding affinities for HDM2 decrease in this order. The decomposition of the effective binding energy into constituent amino acids is shown in Fig 4. We find that the epitope residues (F3, W7 and L10; colored magenta) of the stapled peptides and the hydrophobic residues located at the HDM2 binding surface (colored yellow) are in fact the principal contributors to the binding affinity. Interestingly, we observe that the hydrocarbon linker of sMTide-02 (colored orange) also provides a significant contribution. In this regard, we emphasize the more relevance of analyzing Δf than just examining the direct peptide-protein interaction energy (ΔE u ). Indeed, as can be inferred from S1 Fig  that further partitions Δf into ΔE u and ΔG solv terms, the contributions from the epitope residues and the hydrocarbon linker to ΔE u are comparable to those from the other residues in the stapled peptide, and hence, their significance cannot be elucidated solely in terms of ΔE u . Only after taking into account the dehydration penalty embodied in ΔG solv , the special role of the epitope residues and the hydrocarbon linker becomes evident. This demonstrates the essential importance of analyzing Δf in identifying the residues critical to binding. The significant contributions from the hydrocarbon linker is also observable in sMTide-02b and sMTide-02c, albeit to a lesser extent. (The partitioning of Δf into ΔE u and ΔG solv for sMTide-02b and sMTide-02c is presented in S2 and S3 Figs, respectively.) Correspondingly, the contributions from the epitope residues of sMTide-02b and sMTide-02c are somewhat smaller than those of sMTide-02. This indicates that, in designing stapled peptides, the location of the linker should be optimized also from the standpoint of binding affinity.

Conclusions
We present a computational method for modeling and characterizing stapled peptides and illustrate its application to a stapled p53/HDM2 complex. Putative initial structures for the free stapled p53 peptides and their complexes with HDM2 are generated by a template-based modeling and docking tools, which are subsequently validated via molecular dynamics simulations in an aqueous environment. Thermodynamic characterization of the stapled p53/HDM2 complex is done by decomposing the effective binding free energy into specific constituent groups. This method allows one to identify hot spot residues critical to binding. In fact, we identify the epitope residues of the stapled p53 and the hydrophobic residues of the HDM2 surface to be the principal contributors to the binding affinity. We also find that the The epitope residues of the stapled p53 peptides are colored magenta and the hydrocarbon linker is colored orange; hydrophobic residues located at the binding interface of HDM2 are colored yellow; and residues present in the wild-type p53/HDM2 binding interface are indicated by the red stars.
https://doi.org/10.1371/journal.pone.0232613.g004 hydrocarbon linker of the stapled p53 provides a significant contribution. Thus, the linker plays an important role not only in stabilizing the helical peptide appropriate to binding, but also in determining the binding thermodynamics. Our method will be useful in designing new stapled peptides in which the staple location can be optimized from the thermodynamic viewpoint.