Drugs by Design

Some 15 years or so ago, before combinatorial chemistry stole its limelight, structure-based drug design got lots of attention. Its now lower profile notwithstanding, structure-based design is being used effectively in drug discovery, and progress continues to be made in both the methods and their applications, Charles H. Reynolds, group leader of computer-aided drug discovery at Johnson & Johnson Pharmaceutical Research & Development (J&JPRD), Spring House, Pa., said last month.

Reynolds spoke in Philadelphia at the ACS ProSpectives meeting Advances in Structure-Based Drug Discovery, which he coorganized with chemistry professor Kenneth (Kennie) M. Merz Jr. of the University of Florida, Gainesville, and professor of chemistry and biochemistry Dagmar Ringe of Brandeis University, Waltham, Mass.

In structure-based drug design, the three-dimensional structures of bioactive agents and their targets are the basis of drug discovery. In the past few years, dozens of drugs discovered this way at least in part have reached the market or late-stage clinical trials, and research in the area continues to grow, Reynolds noted. Those approved drugs include the AIDS medications Crixivan and Viracept, the flu drug Tamiflu, the leukemia therapy Gleevec, and the cancer agent Tarceva. Success breeds success. Structure-based design continues to fill drug development pipelines and to play a key role in developing leads and improving drug properties.

Nuclear magnetic resonance spectroscopy is a key source of 3-D information. At Utrecht University, in the Netherlands, researchers are using NMR data to help develop a lantibiotic-based alternative to vancomycin, the last-resort treatment of antibiotic-resistant infections. Lantibiotics are antibiotics containing the amino acid lanthionine. They kill bacteria by binding to lipid II, an essential component of bacterial cell wall biosynthesis, thereby causing pores to form in bacterial membranes.

An NMR structure obtained last year by Eefjan Breukink, Robert Kaptein, and coworkers revealed that the lantibiotic nisin binds in an unprecedented manner to a pyrophosphate on lipid II. Vancomycin also binds lipid II, but bacteria can modify its binding site, and the change gives rise to bacterial resistance to the drug. On the other hand, no bacteria have yet been shown to develop resistance by modifying the lipid II pyrophosphate to which nisin binds. The researchers hope to generate nisin analogs for clinical testing in collaboration with an industrial partner.

At Abbott Laboratories, Abbott Park, Ill., meanwhile, associate research investigator Chaohong Sun and colleagues recently used NMR to discover agents that target X-linked inhibitor of apoptosis protein (XIAP), an anticancer target. Using NMR structures they obtained of XIAP with and without Smac, a protein ligand, they designed nonpeptide Smac mimics. Among them, one binds XIAP nearly two orders of magnitude more strongly than Smac does.

Another key source of 3-D information is X-ray crystallography. For example, researchers at Rib-X Pharmaceuticals, New Haven, Conn., use X-ray crystallography to develop antibiotics that target the ribosome. From solving more than 115 antibiotic-bound ribosome X-ray structures, they have identified promising targets in the ribosome's 50S subunit.

To find compounds likely to interact with those sites and to imbue those compounds with favorable drug properties, the scientists use modeling programs such as Analog and QikProp, developed by Yale University chemistry professor William L. Jorgensen and coworkers and customized by Rib-X researchers. Using this strategy, the Rib-X group has identified several antibiotics from fewer than 700 synthesized compounds and is taking two into human clinical trials, according to Erin M. Duffy, Rib-X senior director of structure-based drug discovery.

At Rutgers University, Piscataway, N.J., professor of chemistry and chemical biology Edward (Eddy) Arnold and coworkers have determined X-ray structures of human immunodeficiency virus (HIV) reverse transcriptase in complexes with various antiviral drugs. On the basis of these structures, they discovered diarylpyrimidines that inhibit reverse transcriptase. These are now being developed as potential AIDS treatments by researchers at several J&JPRD and Janssen Pharmaceutica sites and at Tibotec, in Mechelen, Belgium, in collaboration with Arnold's group. A Phase II clinical trial of one of these, TMC278 (rilpivirine), is currently recruiting patients, and a Phase III trial of another, TMC125 (etravirine), is starting.

In 1985, Arnold also was part of a team that obtained the first structure of a virus that infects animals, the cold-causing human rhinovirus. On the basis of that structure, he, Rutgers research professor Gail Ferstandig Arnold (his wife), and their coworkers generated libraries of modified human rhinoviruses that display immune-response-inducing substances, including one from HIV's gp41 surface glycoprotein. Such substances might serve as constituents of an AIDS vaccine. The researchers are now determining structures of modified rhinoviruses in complex with an anti-HIV antibody as a basis for structure-based vaccine design.

Using X-ray crystallography in combination with other techniques often enhances its usefulness. For example, senior principal scientist Zhaoning Zhu and colleagues at Schering-Plough, Kenilworth, N.J., used X-ray crystallography in combination with SAR (structure-activity relationship) analysis to develop novel hydroxamates as potent inhibitors of tumor necrosis factor- converting enzyme (TACE). One of these, Sch-709156, inhibits TACE in a number of animal models. TACE inhibitors are potential anti-inflammatory agents.

At Merck, West Point, Pa., Joseph P. Vacca, executive director of medicinal chemistry, and coworkers used molecular modeling and X-ray crystallography to identify potent and selective inhibitors of human -secretase-1 (BACE-1) as potential Alzheimer's treatments. They haven't yet identified a specific candidate for development, but we're starting to get close to something we can move forward with, Vacca said.

Also using molecular modeling and X-ray crystallography, distinguished research fellow and vascular research team leader Bruce E. Maryanoff and coworkers at J&JPRD recently developed a potent, selective oral chymase inhibitor that entered clinical trials for asthma and dermatitis. Maryanoff noted that it had previously been notoriously difficult to identify orally active, potent, and selective inhibitors of chymase.

Despite the harvest from X-ray crystallography, the technique still needs improvement. For instance, it's generally difficult to obtain X-ray structures of membrane proteins, which are hard to isolate and crystallize, said biochemistry professor Sir Tom L. Blundell of the University of Cambridge. With about 25% of genes coding for membrane proteins, better techniques for their structural analysis ought to be developed, he added. Novel approaches such as small-volume crystallization and automated crystal imaging will eventually facilitate X-ray crystallography of membrane proteins, he predicted.

In some studies, molecular modeling is the main focus. Assistant professor of pharmacy Lakshmi P. Kotra and coworkers at the University of Toronto recently used homology modeling and molecular dynamics to refine an electron microscopy structure of the insulin receptor with an attached insulin molecule. They then analyzed interactions between insulin and its receptor to design insulin mimics. Insulin currently has to be administered to diabetics by injection, and Kotra and coworkers hope to identify oral agents.

And at Merck, molecular modeling helped develop potent inhibitors of glycogen phosphorylase, an enzyme in glycogen metabolism regulation and a diabetes target. Merck senior research scientist Qiaolin Deng designed and synthesized naphthyl diacid compounds with better access to the enzyme's binding site than that of a lead compound, and the designed versions showed three- to 14-fold better potency.

An important trend in structure-based drug design has been a growing interest in fragment-based approaches. These are based on screening of fragments, which are small molecules bearing one or more functional group motifs. Although in general the fragments bind only weakly to biomolecular targets, they can be combined and elaborated into useful lead compounds.

Fragment-based approaches are a viable alternative and are complementary to traditional lead identification by screening of corporate compound libraries, said Deborah A. Loughney, director of computer-assisted drug design at Bristol-Myers Squibb, Princeton, N.J. In the traditional approach, lead compounds are generally optimized by adding additional groups, paring back a molecule to get to a better starting point, or both. The fragment-based approach instead builds up by starting with less complex molecules and then enhancing potency by increasing complexity as needed.

An early fragment-based approach is SAR by NMR, developed in 1996 by senior research fellow Stephen W. Fesik and coworkers at Abbott. In this technique, NMR is used to obtain structure and affinity data on fragments that bind to proteins. Fragments are then combined and modified to yield higher affinity ligands and drug leads.

Recently, professor of pharmaceutical sciences James A. Wells of the University of California, San Francisco, and coworkers at Sunesis Pharmaceuticals, South San Francisco, developed a related approach using mass spectrometry. Called extended tethering, the technique takes advantage of the inherent affinity of an extender for part of a binding site on a protein. The extender recruits fragments with affinity for an adjacent part of the site, and these fragments can be identified by mass spectrometry. Extender-fragment conjugates are then tested for activity and elaborated as needed. Wells and the Sunesis group recently used extended tethering to identify a new inhibitor for caspase-3, a target for a range of diseases.

At Locus Pharmaceuticals, Blue Bell, Pa., vice president of technology and informatics Jeffrey S. Wiseman and coworkers recently redesigned a p38 kinase inhibitor by using a computational approach to rapidly predict fragment binding energies. Preclinical toxicological studies of the redesigned agents, which have better selectivity than their predecessor, are scheduled to begin next month, with clinical trials for arthritis possibly starting next year. The computational method allowed us to compute binding potency accurately and about 10,000 times faster than any alternative method, Wiseman noted.

And at Astex Therapeutics, Cambridge, England, researchers use a combination of fragment screening, computational chemistry, and structural biology for fragment-based drug discovery. A compound called AT7519, discovered by this approach, went from synthesis to clinical trial approval in just 14 months and is currently in a Phase I clinical trial for treatment of refractory solid tumors.

The Astex group is also among those using structure-based techniques to improve drug ADME/Tox (absorption, distribution, metabolism, excretion, and toxicological) properties. Cytochrome P450 enzymes metabolize most drugs in the body and limit their useful lifetimes of action; P450 interactions can also cause side effects and failures in drug development. In 2001, an Astex group solved the structure of CYP 2C9, the first of an enzyme of this class. A year later, Astex scientists also obtained the first structure of CYP 3A4, the P450 that metabolizes more drugs than any other. The researchers hope to use such structures to rationally design drugs that work longer and are less toxic, said Marcel L. Verdonk, head of virtual screening technologies at Astex.

Similarly, at J&JPRD, researchers are using structure-based techniques to reduce toxic effects that lead to a condition known as long QT syndrome. A form of cardiac toxicity, long QT syndrome disrupts the normal activation and inactivation of electrical currents that control heart ventricles, leading to arrythmia. More than 60 drugs are known or suspected to cause long QT syndrome. Most cases have been traced to blockage of a potassium ion channel encoded by hERG (human ether-a-go-go-related gene).

J&JPRD senior scientist Brett A. Tounge and coworkers are using models of the potassium channel to mitigate the toxicity leading to long QT syndrome. They have developed a model for hERG channel-ligand interactions to guide the synthesis of compounds that don't tend to block the channel and therefore won't cause long QT syndrome.

Overall, there's been pretty steady growth in structure-based drug design in the past few years, Reynolds said. The number of protein structures that have been determined is on a steep climb, so there are a lot more targets available to do modeling around. Computational capabilities have been improving rapidly, too, he adds. More and more people are realizing that structure-based drug design allows you to do discovery in a focused, hypothesis-driven way, where you can probably make many fewer compounds and improve your prospects for success.