This book provides an overview of the biophysical methods applied in drug discovery today, including traditional techniques and newer developments. Perspectives from academia and industry across a spectrum of techniques are brought together in a single volume. Small and biotherapeutic approaches are covered and strengths and limitations of each technique are presented. Case studies illustrate the application of each technique in real applied examples. Finally, the book covers recent developments in areas such as electron microscopy with discussions of their possible impact on future drug discovery.
This is a go-to volume for biophysicists, analytical chemists and medicinal chemists providing a broad overview of techniques of contemporary interest in drug discovery.
Read an Excerpt
Impact and Evolution of Biophysics in Medicinal Chemistry
M. EGGEN AND J. SCHINDLER
Fifty five years ago, Max Perutz and John Kendrew received the 1962 Nobel Prize in Chemistry. Their pioneering work delivered the first 6 Å structure of the protein Myoglobin, and represented the foundation for modern structural biology. This first atomic-level picture heralded the complexity of the protein universe and provoked questions about how to predict the driving forces that allow proteins to rapidly fold into biologically relevant conformations. This was the beginning of biophysics. In 1969, the first enzyme structure of the extracellular nuclease of Staphylococcus Aureus as an enzyme-inhibitor complex was solved by F. Albert Cotton and coworkers.
Thus began significant motivation to devise and evolve computational methods to predict protein folding. "Force fields" in modern computer simulation draw their information primarily from the ~40 000 protein families represented in the Protein Data Bank (PDB), where structures obtained by X-ray crystallography, nuclear magnetic resonance (NMR) spectroscopy and, increasingly, cryo-electron microscopy (cryo-EM) are deposited. Today, X-ray crystallography has been extended further to characterise inherently disordered proteins and protein aggregation.
Analysis of structural information has been applied to study protein equilibria and dynamics. The driving forces of (1) hydrogen bonding, (2) van der Waals interactions, (3) backbone conformational preferences, (4) electrostatics, and (5) hydrophobic interactions leading to the precise folding of proteins produce the minimum energy conformation, as can be seen crystallographically. Beyond the minimum energy conformation determined through x-ray crystallography, additional lower energy minima representing the dynamic movement of proteins can be studied by other biophysical techniques such as protein NMR and proton deuterium exchange mass spectrometry (HDX-MS).
Small molecules must be identified and optimized to productively interact with the protein's disease-relevant conformation or surface. This requires several pieces of information: (1) disease biology, (2) an understanding of the protein form or forms within the cell, and (3) an understanding of the protein surface and the potential interactions and MOA that can occur with the small molecule. Many of the same forces that impact protein conformation influence the low energy conformations and physical properties of drugs and can be assessed through small molecule crystallography, analytical methods and computational minimizations. Ultimately, the small molecule design team, the chemist, computational scientist, and structural biologist are challenged to identify opportunities for optimal molecular recognition from the protein. Optimization requires an understanding of interaction geometries and approximate contributions identified through crystal structures and demonstrated through affinity data. However, molecular interactions behave in a highly "non-additive fashion" and are context-dependent. Solvation of the protein and small molecule, long-range interactions and conformational changes of the protein (see Figure 3, ref. 9) all influence binding energetics.
Design of the small molecule must focus on specific intermolecular reactions based on available structural information. The most important and well defined are hydrogen bond interactions directly with the protein or through structural water. Weaker hydrogen bonds are also sometimes available when an aromatic ring can act as H-bond acceptor. Although H-bonds are among the strongest, these interactions should not be the sole focus of the designer. Orthogonal multipolar interactions of C–F to C=O such as the Bürgi–Dunitz angle, halogen bonds that leverage sigma-hole anisotropy, and cation–Pi interactions are also opportunities in the design of a molecule.
Ultimately, the amount of hydrophobic surface buried through Van der Waals interactions upon ligand binding appears to best correlate with binding affinity. This concept holds for a diverse set of protein–ligand complexes, including protein–protein interactions. All other types of interactions are highly context-dependant and their utility must be tested within a given target. The thermodynamics and kinetics of the target interactions can be evaluated to understand progress and will be discussed in a later section.
Intramolecular interactions within a protein as well as stabilizing intermolecular interactions with an inhibitor were key to delivering the first biophysical technique, X-ray crystallography. The power of interpreting a low-energy atomic-level picture of a molecule bound to an enzyme created a paradigm shift in drug discovery. However, the dynamic nature of proteins requires an understanding of both their minimum energy states and also the status of their movement in other conformations. Further, these states must be measured and interpreted amidst the complex and dynamic processes in the cell. Techniques such as molecular imaging can provide direct measurement of the location and progression of processes. Connecting these techniques to non-invasive measurements of patient disease pathology and response in the evaluation of neurodegeneration, cardiovascular disease and oncology can leverage molecular imaging through PET-CT and MRI scans.
Biophysics is utilized throughout drug discovery strategies from target identification and selection, construct selection, screening, ligand optimization, drug development and ultimately imaging within our patients. The following sections will provide context for the techniques presented in this book.
1.2 Evolution of Biophysics in Medicinal Chemistry
1.2.1 Phenotypic Drug Discovery
The earliest drug identification and approval were driven by small molecule phenotypic drug discovery (PDD) approaches. This strategy involved identifying and optimizing efficacy in animal physiological and behavioural models for human diseases, and ex vivo tissue-based or phenotypic cellular assays. While many drugs used clinically today were discovered agnostic of their target protein, work in this fashion increased the necessary investment in the systematic iteration and synthesis of lead molecules focusing on potency and phenotype. Without specific target and mechanism of action information, agents often displayed poly-pharmacology, which could be good for disease modification or bad for toxicology. Coupled with extended times in early stage discovery, the advent of protein crystallography, and the potential for off-target liabilities that wouldn't be identified until entry into the clinic, drug discovery attempted to minimize these challenges through targeted therapies, that is, target directed drug discovery (TDD) (Figure 1.1).
1.2.2 Targeted and Fragment-based Drug Discovery
With the advent of a deeper understanding of the proteins involved in a disease pathway through cell biology tools such as siRNA, proteomics, knock-outs, and human genetics (mutations found in patient populations), drug discovery efforts have focused more on specific proteins believed to be involved in a disease. This knowledge has led to a paradigm shift in drug discovery that went from dosing a limited set of compounds in animal models to screening 250 000 to 1 million compounds in biochemical assays in a specific protein. Typically, the biochemical assays were performed in formats that looked at competition binding or enzyme activity against recombinant proteins.
Drug discovery targets, in response, moved into more complex protein targets such has epigenetics, protein–protein interactions (PPI), membrane-bound proteins and transporters, and protein aggregation. This resulted in gaps or greater challenges in screening approaches, and protein expression, production and purification. The new target space ultimately affected structural biology approaches. As an example, discrete protein target EZH2 (Section 1.5.2), actually a component of a complex of proteins, requires this complex for activity and stability. In addition, protein complexes could in theory be inhibited by binding in different proteins and read-out as competitive, non-competitive, uncompetitive, mixed or irreversible in nature. In the case of protein–protein interactions, one may simply be interacting with a protein and disrupting a productive, functional interaction with a second protein.
This paradigm shift in targets also resulted in re-thinking how compound libraries were computationally selected for screening. For example, PPI targets required larger molecules that were not present in most traditional libraries based on Lipinki's rules, and when starting points were found in traditional screening, there was very little room to maintain potency while fixing the physical properties. To overcome gaps in compound collections and property challenges, some drug discovery groups implemented a new strategy and proposed to screen compound "fragments" with MW < 300 Da. Fragment-based drug discovery (FBDD) takes advantage of Jencks' description of Gibb's free energy and attributes of the binding energy between enzyme and substrate. By beginning with fragments, a greater chemical space could be sampled with smaller sets of molecules while optimizing biological activity and physical properties in parallel. This introduced a strategic shift in how to approach drug discovery and an opportunity to identify chemical matter without a large compound collection.
The challenge with FBDD is that the initial compound actives have highly efficient per heavy atom (non-hydrogen) target binding but low measured binding constants. Given their low affinity, high concentration biochemical assays (enzyme or competition binding) were used, but did not behave or were still not sensitive enough to pick up fragments in all targets or target classes. Sensitive biophysical methods had to be identified. Teams also needed orthogonal methods to confirm the fragment as a legitimate active, identify a potential MOA, and to provide clues about vectors in order to decorate the molecule for efficient, key interactions within the larger protein active site.
1.2.3 Phenotypic Drug Discovery 2.0
A renewed interest in PDD strategies was influenced by a focus on the complexity of patients and their disease states in oncology, autoimmune disease and neurodegeneration. Coupled with technological developments at the interfaces of disciplines scientists have been able to deliver complex cell culture systems and gene-edited disease relevant mutations for phenotypic screening assays. Advances in mass spectrometry methods such as affinity selection (AS-MS), cellular thermal shift assays (CETSA) and cDNA expression microarray technologies have allowed probing of complex pathways. Improvements in cell-based and model-organism based automated screens, along with the development of new biophysical methodologies have enabled improved screening, faster identification of targets and profiling of the MOA in complex biological samples at the genomic, proteomic and phenotypic levels.
1.2.4 Evolving Compound Collections and Chemical Technologies
Highly elaborate compounds that became part of compound collections were developed during work on well-characterized targets and focused target class research. Re-entry into PDD to identify an alternative molecular MOA and the most relevant but broadened diversity of targets made it clear that there were gaps in the available chemical matter. Scientists have developed different chemical matter strategies and technologies to identify starting points, identify their cellular targets, and select opportunities for enablement of structural information. The lack of chemical starting points for new targets has led to the initiation and expansion of Open Innovation models for accessing novel chemical matter and enabling potential collaborations. The breadth of molecular weight, property space, and molecular topology has expanded to cover FBDD, traditional compound collections and renewed interest in natural products and peptides. For example, given the low affinity of initial fragments, measured biochemical activity typically ranges from low mM to high mM. Chemical technologies such as phage display, DNA-encoded libraries and covalent technologies have further expanded the available tools. These developments, as a whole, elevate the requirements for biophysical methods to cover broad sensitivity ranges and be compatible with diverse physical properties. The new target space coupled with FBDD and the regenerated interest in natural products and peptides have led to an explosion of biophysical approaches to support chemical discovery, and these can be grouped into four areas: (1) biophysical screening, (2) validation of screening hits, (3) chemical optimization, and (4) structural biology. Most significantly, biophysical methods have become a key component in drug discovery from selectively interacting with a protein to demonstrating target engagement in the cell (Figure 1.2).
1.3 Biophysical Screening Approaches
1.3.1 Protein- and Ligand-based NMR
Two biophysical screening approaches that enabled FBDD to be established are the "protein-observed" and the "ligand-observed" nuclear magnetic resonance (NMR) methods. Both NMR approaches are carried out against a small library (15 K) of low molecular weight compounds or fragments. While these two techniques measure the binding of compounds differently, they both bridge the gap in necessary sensitivity by detecting weak ligand binding. Protein-observed NMR and 19F NMR can pick up binding constants in the range of 1 mM to 10 mM, respectively, and are among the most sensitive biophysical techniques.
In protein-observed NMR, relatively small proteins are 15N and 13C double labelled, however other novel approaches observing and labelling tryptophan or other individual residues that may be prevalent in binding sites are also being developed. Protein size becomes the main limitation over 40 kDa and is dependent on protein behaviour and increasing complexity. In protein NMR, the structure is first solved through a suite of experiments including 15N-HSQC and 3D NMR that determine hydrogen bonds, dihedral angles and distance constraints through the nuclear Overhauser effect (NOE), to assign specific residues and connectivity. Once individual residues are assigned, binding of the ligand to the protein is observed through 15N or 1H-amide chemical shift changes observed where ligands bind or indirectly influence movement of the protein. Protein-observed NMR has two additional features. Firstly, the region of the protein that the ligand binds to can be mapped out using a double-labelled (15N, 13C) protein and following the NOEs of the protein in the presence of the compound. When these data are linked to a crystal structure, one can begin to optimize interactions with the minimum energy conformation of the protein while understanding something about protein movement. Linking the fragment hits to structure is a key requirement for FBDD. Given the low affinity of the initial fragments, very little to no cell or biochemical activity may be measurable, therefore linking the hits to structure is key to driving rapid SAR design to improve target affinity.
The second key feature of this screening approach is the ability to differentiate between compounds binding to the same or overlapping region of the protein. Analysis of the individual compound patterns of shift in the protein structure and comparison with ligands of known MOA enables computational modelling and design. If two fragments are binding to adjacent sites of the protein, NOE experiments can demonstrate their proximity and enable compound merging and linker optimization. This approach has delivered rapid improvements in binding affinity, although the precision of vectors required can sometimes lead to synthetic and design challenges. This approach has been used to screen difficult targets such as BACE, Hsp90, MCl-1 and BCl.
Excerpted from "Biophysical Techniques in Drug Discovery"
Copyright © 2018 The Royal Society of Chemistry.
Excerpted by permission of The Royal Society of Chemistry.
All rights reserved. No part of this excerpt may be reproduced or reprinted without permission in writing from the publisher.
Excerpts are provided by Dial-A-Book Inc. solely for the personal use of visitors to this web site.