Elsevier

Chemical Physics Letters

Volume 587, 5 November 2013, Pages 68-74
Chemical Physics Letters

Molecular insights into the stabilization of protein–protein interactions with small molecule: The FKBP12–rapamycin–FRB case study

https://doi.org/10.1016/j.cplett.2013.09.042Get rights and content

Highlights

  • Protein–protein interactions are potential target for drug discovery.

  • Protein–protein interactions stabilization is an emerging strategy for drug design.

  • Stabilization of the FKBP12–rapamycin–FRB complex by the small molecule rapamycin is described.

Abstract

Targetting protein–protein interactions is a challenging task in drug discovery process. Despite the challenges, several studies provided evidences for the development of small molecules modulating protein–protein interactions. Here we consider a typical case of protein–protein interaction stabilization: the complex between FKBP12 and FRB with rapamycin. We have analyzed the stability of the complex and characterized its interactions at the atomic level by performing free energy calculations and computational alanine scanning. It is shown that rapamycin stabilizes the complex by acting as a bridge between the two proteins; and the complex is stable only in the presence of rapamycin.

Introduction

Protein–protein interactions (PPIs) play a key role in several biological processes including cell growth, DNA replication, transcriptional activation, translation, immune response and transmembrane signal transduction [1]. It is known that many human diseases are the result of abnormal PPI, either through the loss of an essential interaction or through the formation of a protein complex at an inappropriate time or location [2]. Therefore, modulating PPIs is a viable therapeutic strategy to fight many different diseases. Targeting PPI for drug discovery is a challenging task and is still in its infancy. The challenges are related to the structural features of protein–protein interfaces and the nature of their interactions [3]. Protein–protein interfaces are generally large (∼1500–3000 Å2) and flat [4], [5], lacking well defined pockets, suitable for hosting drug like molecules [6]. Despite these challenges, several studies have provided evidence for development of small molecules that modulate PPIs which has opened new opportunities in the field of drug discovery [7]. In principle, a protein–protein interaction can be exploited as a drug target either by reducing its strength, thus impairing protein complex formation, or by increasing its strength to stabilize the complex. It depends on the nature of the process mediated by the PPI and on the therapeutic aim whichever is the suitable strategy to be adopted in each case. So far, most efforts have been devoted to the design and development of PPIs inhibitors [8], [9], but recently, also PPIs stabilizing agents have gained an increasing attention [10]. Both stabilization and destabilization can be accomplished either with an allosteric or a direct mechanism. In allosteric mechanism of PPI stabilization, stabilizer can bind to a single protein partner, increasing the mutual binding affinity of the protein partners by inducing conformational modifications. In direct stabilization of PPI, the stabilizer binds to both of the protein partners, which also increases the mutual binding affinity [10]. In this Letter, we have studied a typical case of direct mechanism of PPI stabilization, the complex between FK506-binding protein (FKBP12) and FKBP12–rapamycin-binding domain of FRAP (FRB) with the small molecule stabilizer rapamycin (Figure 1). FRAP is FKBP–rapamycin associated protein [11]. The main objective of this study is to characterize the interaction network responsible for the binding of rapamycin with FKBP12 and FRB, and to get the deeper understanding of rapamycin stabilization of the aforementioned protein–protein complex. Rapamycin is an antifungal agent produced by Streptomyces hygroscopicus which was isolated from Easter Island soil sample [12]. Studies have shown that rapamycin has antitumor and immunosuppressive activity [13], [14]. All these activities have been studied and it has been observed that they share a common mechanism [15]. Rapamycin binds to specific family of immunophilins (cytosolic binding proteins), known as FK506 binding proteins (FKBPs). The most relevant protein of this family for the immunosuppressive effects of rapamycin is FKBP12. FKBP12 is a 12 kDa protein and functions as cis/trans peptidylpropyl isomerases [16]. Rapamycin binds to FKBP12 and inhibits its isomers activity, but this drug–immunophilin complex is insufficient to mediate immunosuppressive effect of rapamycin. Rapamycin becomes biologically active only when FKBP12–rapamycin complex binds with a small 11 kDa hydrophobic binding domain (FRB domain) present on the 289 kDa protein FRAP. This FRB domain is the specific intracellular target of FKBP12–rapamycin complex [17]. The inhibition of this protein blocks signal transduction pathways preventing cell cycle progression from G1 to S phase in various cell types, allowing rapamycin to exploit its action [15].

The X-ray crystallographic structure of the ternary complex of FKBP12–rapamycin–FRB has been solved at 2.7 Å resolution by Choi et al. in 1996 [11]. Ternary complexes of rapamycin derivatives were then crystallized and their structures were solved at 1.85 and 2.2 Å resolution by Liang et al. in 1999 [18]. The structure of the complex is shown in Figure 1. FKBP12 protein consists of a β sheet made of five antiparallel β strands. A short α helix is also present. Rapamycin binds in a hydrophobic pocket formed between the α helix and β sheet [11]. FRB domain of FRAP is composed of a bundle of four α helices with rapamycin binding to a hydrophobic pocket formed by helices α1 and α4 [11]. Chemical structure of rapamycin is shown in the Figure 1 (lower panel). It is evident that rapamycin interacts with both receptor proteins. Moreover, experimental evidences are available that the FKBP12 protein is unable to bind with FRB in the absence of rapamycin [19], [20]. To get the deeper insight in the interaction network responsible for the binding of rapamycin with FKBP12 and FRB, we have performed molecular dynamics (MD) simulations and free energy calculations on this ternary system. In particular, we have calculated binding free energies between the components of the FKBP12–rapamycin–FRB ternary system and performed computational alanine scanning (CAS) to evaluate the contribution of each of the amino acids at the protein–protein and protein–rapamycin interface to the binding energy.

Different computational methods have been developed to estimate free energy differences. Free energy perturbation [21] and thermodynamic integration [22] allow to obtain accurate estimates, but are computationally time-consuming and are not suitable for the study of large macromolecules. More computationally affordable yet reliable methods have been developed to tackle this difficult task. These methods, referred to as MM–PBSA or MM–GBSA [23], [24] (Molecular Mechanics–Poisson–Boltzmann (Generalized-Born) Surface Area), combine the molecular mechanical energies in gas phase, the Poisson–Boltzmann or the Generalized Born approach to evaluate solvation energy, and an empirical function to take into account the contribution of protein surface exposure to solvent. They perform an a posteriori evaluation of binding free energy on snapshots extracted from a molecular dynamics trajectory to evaluate the total binding free energy between two proteins forming a complex. The MM–GBSA approach uses a thermodynamic cycle to calculate the binding free energy between the two protein subunits A and B in solution. In principle, three molecular dynamics simulations should be run to calculate binding free energy, one for the complex and one for each of the monomers in order to sample their conformations. Actually, if the conformations of the isolated monomers do not differ too much from their conformations in the complex, a single MD trajectory of the complex can be computed, and conformations of the monomers can be extracted from the same trajectory. This approach, when the structural modification upon binding are not extensive, is accurate and has the advantage to be less computationally expensive and provides results with a lower statistical uncertainty [25]. Crystallographic studies of rapamycin derivatives in binary complex with FKBP12 and in ternary complex with both FKBP12 and FRB have confirmed the little difference in proteins and rapamycin structure between the two cases [18].

The calculation of the binding free energy allows to perform a CAS analysis of the protein–protein interface, in order to highlight the residues that contributes significantly to protein–protein binding, known as hot-spots. The calculation is performed by evaluating the binding free energy between the protein subunits making up the complex (ΔGwild) and the binding energy upon mutation of each interfacial amino acid into an alanine (ΔGmut,X), so, for each residue, a ΔΔG value can be obtained asΔΔGX=ΔGmut,X-ΔGwild

Also in this case, in principle two MD trajectories should be computed, one for the wild type and other one for the mutated complex. This would require to perform a different MD simulation for each of the point mutations to be sampled in the complex, leading to a substantial number of simulations for a large surface. It is possible to make the hypothesis that a single point mutation does not significantly affect the structure of a protein or of a protein–protein complex. This is a strong approximation, because at least locally the protein structure may be modified by the substitution of an amino acid, but in general it is rather accurate. It is therefore possible to use the so-called single-trajectory approach to perform the computational alanine scanning. According to this procedure, a single molecular dynamics simulation is performed on the protein complex, then a set of snapshots is extracted from the trajectory for the ΔGwild evaluation. Subsequently on the same snapshots point mutations are introduced one at a time and the ΔGmut,X is calculated for each of the amino acids at the protein–protein interface, and ΔΔGX is easily obtained [26]. The single trajectory approach proved to be rather accurate, but it tends to overestimate ΔΔGX for charged residues. This problem can be limited by associating a dielectric constant greater than 1 (usually between 2 and 4) to the protein, mimicking the dielectric properties of protein interiors [27]. Finally, it is worth noting that, while the molecular mechanics energy term can be easily obtained from the results of a molecular dynamics simulation, the entropic term is often difficult to achieve. It can thus be approximated to ΔSvib, i.e. the contribution due to the internal vibrations, whose calculation is nonetheless usually time-consuming and can be affected by large statistical uncertainty [28]. The relative contribution of the change in conformational entropy to the ΔΔG is considered to be negligible for the mutational studies, since it is supposed to cancel up when calculating it in the native and in the mutated complex [26].

CAS has been widely used to study protein–protein and protein–peptide [29], [30], [31] as well as protein–ligand interactions [32], [33], allowing to obtain quantitative data on the relative contribution to binding of residues forming either the protein–protein or the protein–ligand interface. This computational technique is by far less expensive and time consuming than experimental alanine scanning approach, and has been proven to be very accurate when compared with available experimental data [34].

Section snippets

Materials and methods

The initial structure of the ternary complex for our simulation was retrieved from Protein Data Bank (PDB id 1FAP) determined at 2.7 Å resolution [11]. It corresponds to a structure of the complex of human FKBP12 and FRB domain of human FRAP with immunosuppressant drug rapamycin.

Molecular dynamics simulation was performed with AMBER 11 package [36] using explicit solvent and periodic boundary conditions. Two parameter sets were used – Amber99SB force field [37] for the two proteins and

Results and discussion

Rapamycin acts as a stabilizing agent that allows the association of FKBP12 and FRB. From SASA calculations, we have evaluated the protein surface buried upon complex formation, which amounts to 1422 Å2. Approximately 50% of the buried surface is occupied by rapamycin, which is placed between the two proteins.

In order to understand the interaction network at the detailed atomic level, binding free energy calculations and CAS studies were performed. To analyze the interface of the

Acknowledgments

We acknowledge the CINECA Award N. HP10BRT60A, 2013 and the LISA 2012 grant for the availability of high performance computing resources. We also gratefully acknowledge Ministero dell’Università e della Ricerca (PRIN project 2010NRREPL: Synthesis and Biomedical Applications of Tumor-Targeting Peptidomimetics) and Fondazione Banca del Monte di Lombardia for financial support. We thank to Dr. Stefano Rendine for useful discussions.

References (47)

  • D.P. Ryan et al.

    Curr. Opin. Struct. Biol.

    (2005)
  • L.L. Conte et al.

    J. Mol. Biol.

    (1999)
  • M. Arkin

    Curr. Opin. Chem. Biol.

    (2005)
  • M.R. Arkin et al.

    Curr. Opin. Chem. Biol.

    (2009)
  • B.D. Kahan

    Transplant. Proc.

    (2004)
  • S.N. Sehgal

    Clin. Biochem.

    (1998)
  • T. Li et al.

    J. Mol. Graphics Modell.

    (2008)
  • S. Pieraccini et al.

    Chem. Phys. Lett.

    (2011)
  • W. Humphrey et al.

    Graphics

    (1996)
  • J.P. Ryckaert et al.

    J. Comput. Phys.

    (1977)
  • O. Futer et al.

    J. Biol. Chem.

    (1995)
  • G. Zinzalla et al.

    Future Med. Chem.

    (2009)
  • O. Keskin et al.

    Chem. Rev.

    (2008)
  • S. Jones et al.

    Proc. Natl. Acad. Sci. USA

    (1996)
  • M.R. Arkin et al.

    Nat. Rev.

    (2004)
  • A. Mullard

    Nat. Rev. Drug Discovery

    (2012)
  • P. Thiel et al.

    Angew. Chem. Int. Ed.

    (2012)
  • J. Choi et al.

    Science

    (1996)
  • S.N. Sehgal et al.

    J. Antibiot. (Tokyo)

    (1975)
  • S.N. Sehgal

    Transplant. Proc.

    (2003)
  • J.J. Siekierka et al.

    Nature

    (1989)
  • J. Chen et al.

    Proc. Natl. Acad. Sci. USA

    (1995)
  • J. Liang et al.

    Acta. Crystallogr. D

    (1999)
  • Cited by (6)

    View full text