What can happen in a femtosecond? One millionth of a billionth (10-15) of a second—it’s a time scale that’s almost impossible to grasp. In 1 femtosecond light travels a distance much less than the thickness of a human hair, even less than the diameter of a bacterium. Give it an entire second, and light travels from the Earth to the Moon. Yet much of the chemistry vital for life punches a femtosecond time clock. The making and breaking of atomic bonds during every reaction in chemistry and biology passes through a transition state, which can be thought of as the moment at which the bond decides if it will break or reform. Movement of the surrounding atomic environment over a few femtoseconds has an equal probability of nudging atoms to form new products or of returning them to their original configuration. Because it is so short-lived, the transition state has always been a mystery to scientists who wanted to understand its role in the action of enzymes, the essential catalysts of life.
In the last decade, using tools that allow us to sketch the shape that atoms adopt in an enzyme’s active site during this minuscule span of time, we have begun to understand how to freeze an enzymatic reaction. By replicating a shape that exists for the femtosecond lifetime of the transition state, and creating a chemical copy of that shape, we can completely halt the action of the enzyme. These synthetic mimics are called “transition-state analogs” and are as powerful as any enzyme inhibitors ever created. Because they are exquisitely specific and required in extremely small doses, transition-state analogs provide an approach that could revolutionize how drugs are developed.
Finding ways to capture the behavior of atoms during an infinitesimally small and difficult-to-observe time span was no easy task.
Although designing transition-state analogs for enzymes is a relatively recent development, the idea has been around for quite some time. Two-time Nobel Prize–winning chemist Linus Pauling was an early proponent of the idea that enzymes recognize and bind tightly to their substrates during the transition state.[1. L. Pauling, “Chemical achievement and hope for the future,” American Scientist, 36:51-58, 1948.] Born in 1901, Pauling studied quantum mechanics with Niels Bohr in Copenhagen and Erwin Schrödinger in Zurich before returning to the States to apply what he had learned to chemistry. He saw the problem as similar to an antibody binding to its antigen and proposed that enzymes were designed to recognize the activated state of reactants, the transition state, with precision. The tight binding would sequester the activated reactants from solution and increase the reaction rate.
In 1972, Richard Wolfenden gave mathematical form to Pauling’s proposal by solving simple equilibrium equations between enzymes and their transition states.[2. R. Wolfenden, “Transition state analog inhibitors and enzyme catalysis,” Ann Rev Biophys Bioeng, 5:271-306, 1976.] Wolfenden’s equations predicted that the conformational changes in the shape of the enzyme’s active site and its substrate at the moment when the transition state forms increased binding strength by a factor of 1010 to 1020 .
This suggested a powerful idea. If scientists could produce chemically stable analogs of actual transition-state reactants, they would bind to the enzyme just as tightly during that brief femtosecond window of the enzymatic reaction and block its action. These proposals were supported by observations that natural-product antibiotics, which have features similar to transition-state analogs, were unusually powerful enzyme inhibitors. But these insightful hypotheses were made before the existence of experimental and computational approaches for observing and predicting the structures of enzymatic transition states. Indeed, finding enzymatic transition-state analogs is challenging, because a typical enzyme-reactant or enzyme transition-state complex often has more than 10,000 atoms whose positions would have to be determined.
Today, those insights have helped us shape the field of transition-state analog chemistry, and are leading to new approaches in drug development. But finding ways to capture the behavior of atoms during an infinitesimally small and difficult-to-observe time span was no easy task.
Catching a molecular moment
As more computational power became available in the 1980s, it was possible to imagine that enzymatic transition-state structure could be solved by a combination of experimental and computational quantum chemistry. To get at the problem, we combined a method often used in physical organic chemistry with computational chemistry that originated from the Manhattan Project.
At the end of the Manhattan Project, Jacob Bigeleisen, one of the program’s alums, released his theories, which were later reduced to computer algorithms and made available to the academic community through the Quantum Chemistry Exchange Program, which provided free access to early computational chemistry. When we started our work using this resource, the lab’s original goal was purely academic: to see if the transition-state structure of AMP nucleosidase, an enzyme involved in purine metabolism in E. coli, would be the same if the reaction was catalyzed by an acid rather than the enzyme. The transition states differed. Although the study did not have great significance biologically, it forced us to develop tools that have become essential to solving the transition-state structure of purine nucleoside phosphorylase (PNP), a known target for T-cell cancers. But figuring out the method took some doing.
Only two features, albeit complex ones, are needed to describe molecular interactions in biology: geometric shape and electrostatic charge. We applied these guiding principles to work out the atomic structures of transition states at the catalytic sites of particular enzymes. The shape of the electron cloud surrounding the atoms, known as the van der Waals surface, predicts how a molecule may occupy space and interact with partner molecules. Electron distribution at the van der Waals surfaces determines whether atomic neighbors will be attracted, like opposite poles of a magnet, or repulsed, like similar poles.
Two approaches are needed to solve these features of enzymatic transition states: the measurement of kinetic isotope effects and computational quantum chemistry. Measuring kinetic isotope effects gives information about both the geometry and electrostatic charge of the transition state. In these experiments we replace the common atoms of nature—hydrogen, carbon and nitrogen—with their heavy-isotope counterparts. For example, deuterium (mass 2) replaces hydrogen (mass 1) in the reactants for the enzyme of interest. Each atomic replacement in the kinetic isotope effect experiment alters the femtosecond bond vibrations of the reactant in the transition state. Thus, by replacing atoms isotope by isotope and then monitoring how those changes affect reaction times, we can collect enough information to deduce the atomic structure of the transition state.[3. V. L. Schramm, “Enzymatic transition states, transition-state analogs, dynamics, thermodynamics, and lifetimes,” Annu Rev Biochem, 80:703-32, 2011.]
Although it was possible to determine kinetic isotope effects as early as the 1960s, it wasn’t until the 1980s that computer-based quantum-chemical approaches became refined enough to begin to interpret the results. Computational quantum chemistry is used to search through thousands of possible transition states to find the one that matches the experimental observations from kinetic isotope studies. This structure is then analyzed using Schrödinger’s equation to obtain the wavefunction, which contains information about both geometry and electrostatic charge, and in fact is the most complete description that can be given of a transition state. This information provides enough of a picture, a virtual blueprint, to allow us to design analogs that mimic its geometry and electrostatic features.
Transition-state analogs are recognized by the enzyme, and the forces that would be applied to bond-breaking in the normal reaction are instead converted into binding energy; those forces are considerable, binding the analogs millions of times more tightly than the normal reactants. The transition-state analog for PNP binds 4,300,000 times tighter to its parent enzyme than does the normal reactant. This means that only tiny amounts of a drug that mimics the normal reactant’s transition-state geometry need be delivered to the target enzyme; and by binding tightly to the catalytic site, such an analog acts to inhibit the enzyme by preventing the normal reactants from binding. Transition-state analogs are now beginning to show results in early-stage trials as therapies for a wide range of unrelated diseases, demonstrating their promise as a better means of hitting a therapeutic target.
A future for transition state drug design?
This new approach to drug design differs from more time-honored methods such as synthesizing a chemical entity that is tailored to fit its target—an approach called structure-based drug design—or chemical-library screening and natural-products chemistry, the random search through millions of chemical compounds in the hopes of chancing on one that inhibits the target enzyme. Each of these methods is universal, that is, they can be used against most targets of interest to the pharmaceutical industry, including receptor molecules, ion channels, and enzymes. Transition-state analysis, by contrast, is limited to enzyme targets. Although a small slice of the pharmaceutical-target pie, enzymes are ubiquitous and essential to life, and thus an area ripe for new drug development.
Even though drug design from enzymatic transition-state analysis is in its infancy, it has already produced a family of potential drugs for an array of biological targets. With many already in clinical trials, we will soon learn the details of their biological impact.
A plethora of applications
Chemical analogs for targeting transition state offer a potent method for inhibiting enzymes involved in disease. Today, transition-state analog inhibitors designed for specific targets are beginning to wend their way into preclinical and clinical trials. These drugs have the potential to produce fewer side effects than other enzyme inhibitors, because their binding is so specific and strong.
We designed an early proof-of-concept molecule, immucillin-H, as one of the first transition-state analogs. The original goal was to design a mimic that would bind to the transition state of bovine purine nucleoside phosphorylase (PNP). Blocking PNP is known to kill rapidly dividing T cells found in T-cell leukemia patients and in patients with autoimmune disorders where the T cells are attacking normal host tissues. Immucillin-H was chemically synthesized for us by Peter Tyler at Industrial Research Ltd. in New Zealand. It proved to be a powerful transition-state analog inhibitor of human PNP, and is now in worldwide clinical trials for several different types of leukemia under the name forodesine.[4. K. Balakrishnan et al., “Phase 2 and pharmacodynamic study of oral forodesine in patients with advanced, fludarabine-treated chronic lymphocytic leukemia,” Blood, 116:886-92, 2010.]
Because the first-generation transition-state analog was difficult to synthesize, and was designed for the bovine enzyme, we created a second-generation PNP inhibitor designed specifically to match the transition state of human PNP. It is called BCX4208 and is in clinical trials for treatment of gout. Gout is caused by an excess of uric acid in the blood, leading to painful and destructive crystal formation in joints. More than 15 million people suffer from gout in North America, Europe, and Japan, and the disease is not easily controlled by current drugs. In humans, uric acid formation requires the action of PNP; hence, blocking PNP with BCX4208 may eventually lead to a unique and effective approach to gout management by using the analog at low levels such that normal T cells are not depleted.[5. S. Bantia et al., “Potent orally bioavailable purine nucleoside phosphorylase inhibitor BCX-4208 induces apoptosis in B- and T-lymphocytes—a novel treatment approach for autoimmune diseases, organ transplantation and hematologic malignancies,” Int Immunopharmacol, 10:784-90, 2010.]
An encouraging feature of these powerful PNP inhibitors is their tight binding to the PNP target. They bind so tightly that after a few hours most of the unbound drug is gone from the blood, while the inhibitor stays bound to the target enzyme for the lifetime of the cell.[6. A. Lewandowicz et al., “Achieving the ultimate physiological goal in transition state analogue inhibitors for purine nucleoside phosphorylase,” J Biol Chem, 278:31465-68, 2003.] This is important because most drugs require careful dosing to maintain sufficient amounts in the blood to allow for constant interaction with the target. This requirement for excess circulating drug exposes all cells and increases the risk that off-target interactions will cause side effects. With specific and long-lasting binding to the target, and no drug in the circulation, side effects would in theory be minimized.
PNP transition-state analogs may also find application in combating malaria. Over the course of evolution, Plasmodium falciparum, the most lethal malaria parasite, lost the ability to make its own purines from amino acids and sugars because those of the host are so readily available, and it now relies on host purines, specifically hypoxanthine, to make its RNA and DNA. Both in humans and in P. falciparum, the only way to make hypoxanthine from purine nucleosides is by the catalytic action of PNPs. The challenging aspect of creating an inhibiting compound for malaria was to find a molecule that mimics the transition-state structures of both human and parasitic PNPs.
To overcome this problem, we compared the transition states of human and P. falciparum PNPs. We discovered that a single transition-state inhibitor called DADMe-immucillin-G was similar to both human and parasite PNP transition states and might act as a powerful inhibitor of both. DADMe-immucillin-G was synthesized by Gary Evans at Industrial Research Ltd., and blocked both the human and parasite enzymes at picomolar concentrations. As P. falciparum can only infect primates, we infected Aotus monkeys with the parasite and then treated them with oral doses of DADMe-immucillin-G twice a day for 7 days. The primates’ blood was cleared of parasites between the 4th and 7th days, and no parasites were detected for up to 9 days post-treatment. Although infections returned after the treatment ended, meaning that not every parasite had been killed, slower parasitic growth was observed. As in other applications of PNP inhibitors, no toxicity was detected. P. falciparum infections in the Aotus test are more virulent than in humans; thus there is hope that similar treatment might be even more effective in humans, but clinical trials are needed to test this hypothesis.[7. M. B. Cassera et al., “Plasmodium falciparum parasites are killed by a transition state analogue of purine nucleoside phosphorylase in a primate animal model,” PLoS One, 6:e26916, 2011.]
Targeting pathways important in rapidly dividing cancer cells is a time-honored approach to designing anticancer agents. But most agents also damage normal cells, causing the well-known side effects of cancer chemotherapy. We designed and synthesized transition-state analogs to disrupt the polyamine pathway, which provides essential counterions necessary for quick DNA strand separation in a rapidly dividing cancer cell. The transition-state analogs block human methylthioadenosine phosphorylase (MTAP), resulting in the cellular accumulation of 5’-methylthioadenosine, a metabolite that inhibits polyamine biosynthesis and halts cancer cell division. MTAP inhibitors have shown remarkable efficacy in blocking the growth of or eradicating human lung cancers and head and neck cancers grown in immune-deficient mice.[8. I. Basu et al., “A transition state analogue of 5’-methylthioadenosine phosphorylase induces apoptosis in head and neck cancers,” J Biol Chem, 282:21477-86, 2007.],[9. I. Basu et al., “Growth and metastases of human lung cancer are inhibited in mouse xenografts by a transition state analogue of 5’-methylthioadenosine phosphorylase,” J Biol Chem, 286:4902-11, 2011.] The MTAP inhibitors are administered orally, can be given once a day, and mice that are fed large quantities of the drug remain healthy as judged by weight, blood chemistry, and tissue histology. These unique properties of the MTAP transition-state analogs in mouse models offer the intriguing possibility of controlling certain human cancers with a once-a-day pill that has few side effects. Although the vast majority of drugs that succeed in the mouse model fail in humans, we are encouraged by the specificity and the low toxicity we observed.
Bacterial antibiotic resistance remains another of the world’s major health challenges. The problem stems from the fact that antibiotics kill all bacteria save those rare individuals that develop resistance. Under continued antibiotic selection pressure, the resistant individuals give rise to a new population. So it has been since the advent of penicillin and with all subsequent bacterial antibiotics. But there are better targets that could control the detrimental aspects of infection without killing bacteria and thereby introducing selection pressure. Children are immunized with the DPT (diphtheria, pertussis, tetanus) vaccine to prevent these bacterial diseases, but the combined vaccine does not necessarily create immune reactions against the bacteria. Rather, the D and T of DPT represent inactivated diphtheria and tetanus toxoids produced by the infecting bacteria. Immunization with toxoids creates antibodies against the toxins, not the bacteria. But it is the toxins that cause damage to human tissue.
Many bacterial toxins are produced under the genetic control of quorum-sensing molecules that detect when the numbers of bacteria grow to a critical threshold—a quorum. Our group hypothesized that blocking the quorum-sensing pathway would cut the telegraph wires and prevent transmission of the “make toxin!” attack message. Without production of pathogenicity factors like toxins, biofilms, and human-cell attachment factors, otherwise harmful bacteria would be disarmed but would continue to live, thereby removing the selection pressure for the development of resistant strains.
We targeted a bacterial enzyme called MTAN that regulates the production of quorum-sensing molecules in gram-negative bacteria such as Pseudomonas, Vibrio, and Escherichia species. Our transition-state analog designed to inhibit bacterial MTAN has led to the synthesis of some of the most powerful inhibitors ever described for enzymes. The best of these bind to MTAN 91 million times tighter than the normal enzyme reactants. At nanomolar concentrations, MTAN inhibitors block quorum-sensing molecules in a virulent strain of Vibrio cholera, the causative agent of cholera, without appearing to cause resistance. In fact, when Vibrio bacteria are grown for 26 generations (a 226 increase in cell population) in the presence of a large excess of MTAN inhibitors, subsequent bacterial generations are equally sensitive to inhibition of the quorum-sensing pathway.[10. J. A. Gutierrez et al., “Transition state analogs of 5’-methylthioadenosine nucleosidase disrupt quorum sensing,” Nat Chem Biol, 5:251-57, 2009.] Now it remains to be seen whether transition-state analogs can translate into new and novel antibiotics. A generation of antibiotics that could prevent disease without causing resistance would indeed be a boon to medical treatment.
Vern L. Schramm is Professor and Chair of Biochemistry and the Ruth Mearns Chair in Biochemistry at Albert Einstein College of Medicine of Yeshiva University, located in the Bronx, New York.