Use of Metadynamics in the Design of isoDGR-Based αvβ3 Antagonists To Fine-Tune the Conformational Ensemble

Abstract

The integrin family of cell-adhesion receptors regulates cellular functions crucial to the initiation, progression, and metastasis of solid tumors. In particular, integrin avb3 plays a key role in endothelial cell survival and migration during tumor angiogenesis. It is therefore gaining increasing importance as a drug target in antiangiogenic cancer therapy. The sequence Arg-Gly-Asp (RGD), which is contained in natural avb3 interactors, such as vitronectin, fibronectin, fibrinogen, osteopontin, and tenascin, is by far the most prominent ligand to promote specific cell adhesion through stimulation. This sequence is therefore attractive as a lead for the development of different integrin antagonists. Recent biochemical studies showed that deamidation of the NGR sequence gives rise to isoDGR, a new avb3-binding motif. This sequence constitutes a novel class of peptidic integrin ligands and paves the way to drug-design studies with a focus on the synthesis and characterization of a new generation of isoDGR-based macrocycles. For the design of low-molecular-mass isoDGR-containing molecules, an accurate determination of their biologically active conformation is a prerequisite. The presence of the b bond induces high flexibility in isoDGR-containing macrocycles and thus augments the range of accessible interconverting conformations. However, the identification of relevant conformations that might affect binding affinity is challenging for standard spectroscopic and diffraction techniques. Atomistic simulations, such as molecular dynamics (MD), replica-exchange molecular dynamics (REMD), and Monte Carlo (MC) simulations, can complement experimental data. However, they often fail to generate reliable equilibrium conformations because of the rugged and complex nature of the free-energy surface (FES) that is accessible to the system. As a consequence, computational sampling is often relegated to some local, unrealistic minima, which compromise subsequent docking studies. As computational drug design becomes increasingly reliant on virtual screening and on high-throughput 3D modeling, the need for fast and accurate computational methods for sampling of the ensemble of energetically accessible conformations is warranted. In this context, several techniques, including the local-elevation method, taboo search, the Wang–Landau method, adaptive force bias, conformational flooding, umbrella sampling, weighted histogram techniques, transitionstate theory, and path sampling, have been developed to address the sampling problem, through either reconstruction of the free energy or the direct acceleration of events that might happen on a long timescale (“rare events”). Related to these methods, metadynamics (MetaD) has emerged as a powerful coarse-grained non-Markovian molecular-dynamics approach for the acceleration of rare events and the efficient and rapid computation of multidimensional free-energy surfaces as a function of a restricted number of degrees of freedom, named collective variables (CVs). If the CVs are appropriately chosen for the system under investigation, MetaD directly provides a good estimate of the free energy of the system projected into the CVs (see the Supporting Information for details). Notably, the free energy is not immediately deducible by other sampling methods, such as umbrella sampling, in which the free-energy profile is not obtained directly from the simulations and requires an additional computational step, such as the weighted histogram analysis method (WHAM). In this study, we developed a protocol based on the combination of MetaD and docking simulations to analyze the conformations and the avb3-binding properties of isoDGR-containing cyclopeptides and to predict the conformational effects of chemical modifications and discriminate binders from nonbinders in silico. To investigate the conformational equilibrium of RGD-, DGR-, and isoDGRcontaining cyclopeptides (cyclization mode involving cysteine side chains) and to exhaustively explore their FESs, we performed well-temperedMetaD simulations, for which we chose Gly f and y angles as CVs (Figure 1; for simulation details, see the Supporting Information). As it lacks a side chain, Gly has large conformational freedom around its backbone dihedral angles. Therefore, Gly can explore a considerably larger area in the Ramachandran energy diagram than any other amino acid, and occupies five [*] Dr. A. Spitaleri, Dr. M. Ghitti, Dr. S. Mari, Dr. G. Musco Dulbecco Telethon Institute, Biomolecular NMR Laboratory c/o Center of Genomic and Bioinformatics S. Raffaele Scientific Institute via Olgettina 58, 20132 Milan (Italy) Fax: (+39)02-2643-4153 E-mail: giovanna.musco@hsr.it