Ligand Strain and Entropic Effects on the Binding of Macrocyclic and Linear Inhibitors: Molecular Modeling of Penicillopepsin Complexes

Combination of Coevolutionary Information and Supervised Learning Enables Generation of Cyclic Peptide Inhibitors with Enhanced Potency from a Small Data Set

Computational generation of cyclic peptide inhibitors using machine learning models requires large size training data sets often difficult to generate experimentally. Here we demonstrated that sequential combination of Random Forest Regression with the pseudolikelihood maximization Direct Coupling Analysis method and Monte Carlo simulation can effectively enhance the design pipeline of cyclic peptide inhibitors of a tumor-associated protease even for small experimental data sets. Further in vitro studies showed that such in silico-evolved cyclic peptides are more potent than the best peptide inhibitors previously developed to this target. Crystal structure of the cyclic peptides in complex with the protease resembled those of protein complexes, with large interaction surfaces, constrained peptide backbones, and multiple inter- and intramolecular interactions, leading to good binding affinity and selectivity.

A novel genetically-encoded bicyclic peptide inhibitor of human urokinase-type plasminogen activator with better cross-reactivity toward the murine orthologue

The inhibition of human urokinase-type plasminogen activator (huPA), a serine protease that plays an important role in pericellular proteolysis, is a promising strategy to decrease the invasive and metastatic activity of tumour cells. However, the generation of selective small molecule huPA inhibitors has proven to be challenging due to the high structural similarity of huPA to other paralogue serine proteases. Efforts to generate more specific therapies have led to the development of cyclic peptide-based inhibitors with much higher selectivity against huPA. While this latter property is desired, the sparing of the orthologue murine poses difficulties for the testing of the inhibitor in preclinical mouse model. In this work, we have applied a Darwinian evolution-based approach to identify phage-encoded bicyclic peptide inhibitors of huPA with better cross-reactivity towards murine uPA (muPA). The best selected bicyclic peptide (UK132) inhibited huPA and muPA with Ki values of 0.33 and 12.58 µM, respectively. The inhibition appears to be specific for uPA, as UK132 only weakly inhibits a panel of structurally similar serine proteases. Removal or substitution of the second loop with one not evolved in vitro led to monocyclic and bicyclic peptide analogues with lower potency than UK132. Moreover, swapping of 1,3,5-tris-(bromomethyl)-benzene with different small molecules not used in the phage selection, resulted in an 80-fold reduction of potency, revealing the important structural role of the branched cyclization linker. Further substitution of an arginine in UK132 to a lysine resulted in a bicyclic peptide UK140 with enhanced inhibitory potency against both huPA (Ki = 0.20 µM) and murine orthologue (Ki = 2.79 µM). By combining good specificity, nanomolar affinity and a low molecular mass, the bicyclic peptide inhibitor developed in this work may provide a novel human and murine cross-reactive lead for the development of a potent and selective anti-metastatic therapy.

Reliable In Silico Ranking of Engineered Therapeutic TCR Binding Affinities with MMPB/GBSA

Accurate and efficient in silico ranking of protein–protein binding affinities is useful for protein design with applications in biological therapeutics. One popular approach to rank binding affinities is to apply the molecular mechanics Poisson–Boltzmann/generalized Born surface area (MMPB/GBSA) method to molecular dynamics (MD) trajectories. Here, we identify protocols that enable the reliable evaluation of T-cell receptor (TCR) variants binding to their target, peptide-human leukocyte antigens (pHLAs). We suggest different protocols for variant sets with a few (≤4) or many mutations, with entropy corrections important for the latter. We demonstrate how potential outliers could be identified in advance and that just 5–10 replicas of short (4 ns) MD simulations may be sufficient for the reproducible and accurate ranking of TCR variants. The protocols developed here can be applied toward in silico screening during the optimization of therapeutic TCRs, potentially reducing both the cost and time taken for biologic development.

Energy-entropy method using multiscale cell correlation to calculate binding free energies in the SAMPL8 host-guest challenge

Free energy drives a wide range of molecular processes such as solvation, binding, chemical reactions and conformational change. Given the central importance of binding, a wide range of methods exist to calculate it, whether based on scoring functions, machine-learning, classical or electronic structure methods, alchemy, or explicit evaluation of energy and entropy. Here we present a new energy-entropy (EE) method to calculate the host-guest binding free energy directly from molecular dynamics (MD) simulation. Entropy is evaluated using Multiscale Cell Correlation (MCC) which uses force and torque covariance and contacts at two different length scales. The method is tested on a series of seven host-guest complexes in the SAMPL8 (Statistical Assessment of the Modeling of Proteins and Ligands) "Drugs of Abuse" Blind Challenge. The EE-MCC binding free energies are found to agree with experiment with an average error of 0.9 kcal mol-1. MCC makes clear the origin of the entropy changes, showing that the large loss of positional, orientational, and to a lesser extent conformational entropy of each binding guest is compensated for by a gain in orientational entropy of water released to bulk, combined with smaller decreases in vibrational entropy of the host, guest and contacting water.

Macrocycle Cell Permeability Measured by Solvation Free Energies in Polar and Apolar Environments

The relation of surface polarity and conformational preferences is decisive for cell permeability and thus bioavailability of macrocyclic drugs. Here, we employ grid inhomogeneous solvation theory (GIST) to calculate solvation free energies for a series of six macrocycles in water and chloroform as a measure of passive membrane permeability. We perform accelerated molecular dynamics simulations to capture a diverse structural ensemble in water and chloroform, allowing for a direct profiling of solvent-dependent conformational preferences. Subsequent GIST calculations facilitate a quantitative measure of solvent preference in the form of a transfer free energy, calculated from the ensemble-averaged solvation free energies in water and chloroform. Hence, the proposed method considers how the conformational diversity of macrocycles in polar and apolar solvents translates into transfer free energies. Following this strategy, we find a striking correlation of 0.92 between experimentally determined cell permeabilities and calculated transfer free energies. For the studied model systems, we find that the transfer free energy exceeds the purely water-based solvation free energies as a reliable estimate of cell permeability and that conformational sampling is imperative for a physically meaningful model. We thus recommend this purely physics-based approach as a computational tool to assess cell permeabilities of macrocyclic drug candidates.

Peptidic Macrocycles - Conformational Sampling and Thermodynamic Characterization

Macrocycles are of considerable interest as highly specific drug candidates, yet they challenge standard conformer generators with their large number of rotatable bonds and conformational restrictions. Here, we present a molecular dynamics-based routine that bypasses current limitations in conformational sampling and extensively profiles the free energy landscape of peptidic macrocycles in solution. We perform accelerated molecular dynamics simulations to capture a diverse conformational ensemble. By applying an energetic cutoff, followed by geometric clustering, we demonstrate the striking robustness and efficiency of the approach in identifying highly populated conformational states of cyclic peptides. The resulting structural and thermodynamic information is benchmarked against interproton distances from NMR experiments and conformational states identified by X-ray crystallography. Using three different model systems of varying size and flexibility, we show that the method reliably reproduces experimentally determined structural ensembles and is capable of identifying key conformational states that include the bioactive conformation. Thus, the described approach is a robust method to generate conformations of peptidic macrocycles and holds promise for structure-based drug design.