Fragment-based drug design is a new approach that has been successfully applied to challenging targets, such as protein-protein interactions. While 3-D protein structures have been used in drug discovery for many years now, fragment-based drug design uses x-ray crystallography or other physical techniques to screen fragment libraries for specific binding to a target protein. Knowledge of exactly how the fragments bind to the protein target allows the hits to be optimized by growing the fragments or by combining and linking different fragments.
Fragment-based drug design has a number of attractive features compared to high-throughput screening. The approach allows for the screening of substantially fewer compounds, usually several hundred to a thousand. It detects fragments that bind specifically but with low affinity (~100 µM to 10 mM). Low-affinity binders are generally difficult to detect by most other methods.
Additionally, there are few, if any, false positives, and nonspecific binding is a nonissue, because the approach only detects compounds that bind specifically in a regular fashion. The small size of the fragments (usually less than 250 Daltons) makes subsequent optimization relatively easier. It is generally easier to build up a small molecule than it is to reduce the size of a large one.
Compounds designed using this method are more likely to be novel compared to screening of larger compounds, because the smaller fragments explore chemical space more effectively. “You can get to new chemistry outside of the standard databases where you start with compounds 300-400 molecular weight,“ says Tom Blundell, Ph.D., Sir William Dunn Professor of Biochemistry at the University of Cambridge and one of the founders of Astex Therapeutics (www.astex-therapeutics.com). “Very quickly you can move to new compounds that are patentable.“
Advances in protein expression, automation of crystallization and x-ray data collection, and complex structure determination have increased throughput, as well as the availability of targets. Originally developed at Abbott Laboratories (www.abbott.com) to increase throughput specifically for fragment-based drug design, the ACTOR robot from Rigaku (www.rigaku.com) automates the handling of cryo-cooled samples in protein crystallography. Although by most standards the throughput is modest, these advances have allowed protein crystallography to contribute early in a drug design project rather than in hindsight.
Fragment-based lead optimization is a high-content approach rather than a high-throughput one. “The goal is to identify better leads. Were looking for maximal potency for minimal mass,“ says Dr. Stephen Burley, CSO of SGX Pharmaceuticals (www.sgxpharma.com). Fragments identified through this approach tend to be high-efficiency binders, that is, a larger percentage of their atoms are directly involved in binding as compared to larger molecules. Also, because the approach is empirical, there are no preconceived notions about what sites are important and what types of fragments will bind. Companies, such as SGX Pharmaceuticals, Plexxikon, Sareum, Takeda San Diego, ActiveSight, and Astex Therapeutics, among others, have been successful in championing these techniques, particularly as measured by their ability to establish collaborations with pharmaceutical companies.
Defining the Process
Fragment-based drug design using x-ray crystallography requires that a protein target that can readily form crystals with or without ligand and its three dimensional structure is available. Thanks to the efforts of structural genomics initiatives, on-going worldwide, the number of available targets is constantly growing.
A number of different physical techniques can be used to analyze the binding of low-affinity fragments to protein targets, including NMR and mass spectroscopy. “Each technique has its own particular advantages and disadvantages,“ says Duncan McRee, Ph.D., president of ActiveSight (www.active-sight.com). “But x-ray has an advantage because it provides knowledge of how the fragments bind in the active site.“ By contrast, NMR has not benefited from the same degree of automation that x-ray crystallography has, and NMR results are more subject to interpretation.
For protein crystallography, ten to a hundred milligrams of purified protein are needed to set up and grow large numbers of protein crystals, at least 100 or so. Each crystal is soaked in a sample of five to 10 compound fragments, mounted, and cryo-cooled for X-ray data collection. High-speed data collection can be accomplished at synchrotron facilities. “Using our proprietary beamline facility at the Argonne National Laboratory, we can typically screen our entire library, about 1,000 fragments, in about 24 to 48 hours,“ comments Dr. Burley. “Access to this beamline allows us to do experiments that we would never have contemplated otherwise.“
Reasonable throughput rates can be achieved at home labs given that access to synchrotron sources is limited. New X-ray generators, such as the Rigaku FR-E Superbright, along with optical focusing systems can produce intense x-ray beams that make collection times reasonably short, particularly when combined with other automation technology. Astex Therapeutics reported collection and processing of diffraction data from 54 crystals of protein tyrosine phosphatase 1B in an 80-hour period using such a system.
If a fragment binds to the protein in the crystal, the complex causes a change in the x-ray diffraction pattern as compared to the native, or unbound, diffraction pattern. Difference maps between the complex and the native data show where the fragment is bound on the protein structure. The exact binding constant cannot be determined by this method as in NMR, but subsequent refinement determines the fragments 3-D structure in the protein binding pocket.
The fragment must bind specifically to the protein molecules in the crystal to be detected. Nonspecific binding, often a major problem in other high-throughput screening, is invisible in the x-ray data because nonspecific binders bind randomly and essentially do not affect the x-ray diffraction signal. Fragments identified using this technology bind too weakly to be considered hits —their binding affinity is usually in the high µM to mM range. There are two basic strategies to create higher affinity compounds from the fragments: extension and linking.
Extension or growing consists of adding functionality to an initial fragment that explores the adjacent areas of the binding pocket to find favorable interactions that increase the binding affinity. This has been used successfully to develop inhibitors to a number of targets, including p38 MAP kinase inhibitor (Astex), DNA gyrase (Hoffman-La Roche), Erm methyl transferase (Abbott), and urokinase (Abbott).
High-affinity compounds can also be created by linking low-affinity fragments together. This requires multiple fragments that bind in close proximity, knowledge of their binding modes, and orientations, and finally, a suitable linker that can maintain the binding interactions of the individual fragments. NMR or mass spectroscopy is used more commonly in conjunction with fragment linking. Variations on this include fragment self-assembly where reactive fragments can self-assemble in the presence of a binding site that brings two fragments into close proximity.
Knowledge of where specific chemical fragments bind to a protein can be used in a broad range of computational techniques. The information can be used in a variety of computational techniques, such as pharmacophore searches of available chemical libraries. The empirical data reduces the combinatorial complexity of the problem dramatically. These searches allow researchers to find or synthesize compounds that hopefully validate the binding hypothesis and extend the compound fragment into other parts of the pocket.
From Hindsight to the Forefront
The dramatic progress in robotics, automation, and computational power in recent years has taken protein crystallography from a tool used in hindsight to the forefront of new drug design. X-ray crystallography screening is used to identify small molecule fragments that bind specifically to a protein target. Knowledge of the binding structure of the complex allows these fragments to be extended or linked with other fragments to create higher affinity compounds. The results also provide empirical data in which to ground computational and combinatorial chemistry experiments.
Fragment-based drug design opens the door to develop inhibitors for targets that have been difficult to address by other methods, such as those involving protein-protein interactions. “What is most exciting right now is how we can use it in multiprotein systems, which dont have deep binding pockets,“ says Dr. Blundell.
Fragment-based drug design is becoming an effective technology that is complementary to the existing approaches to drug design. It is clear that this technique will be used extensively in the future as more groups adopt the technology.