**3. Determination of transition state architecture**

mimic geometric and electrostatic features of a transition state (or other intermediates of high energy) are considered as excellent enzyme inhibitors (Fig.1). They bind up to 108

tighter than substrate. Thus, the goal of transition-state analogs design is to create stable chemical structures with van der Waals geometry and molecular electrostatic potential sur‐

Although some reviews on the subject have been published, this concept has not been re‐ viewed in detail [Wolfenden, 1999; Robertson, 2005; Schramm, 2005; Schramm, 2007; Dyba‐ ła-Defratyka et al. 2008; Schramm, 2011]. In this review the current trends, alongside with

The sequencing of the human genome has promised a revolution in medicine. The genome encodes 20,000- 25,000 human genes, and thousands more proteins as a result of alternative gene splicing. Many of these hold the keys to treating disease, especially numerous enzymes of undefined so far physiologic functions [Gonzaga-Jagureui et al., 2012]. Out of 1200 registered drugs over 300 act as enzyme inhibitors. Most of them are simple analogs of substrates of cer‐ tain enzymatic reaction. Analogy to transition state as a mean to obtain effective inhibitors emerged in 1970s [Lienhard, 1973]. Through the 1970s and 1980s, most of the known examples were natural products [Wolfenden, 1976]. The situation has changed in 1990s when synthetic inhibitors became the predominate examples of transition-state inhibitors. In 1995, there were

transition-state analogues for at least 130 enzymes [Radzicka & Wolfenden, 1995].

appropriate case studies in designing of such inhibitors will be presented.

faces as close as possible to those of the transition state.

**Figure 1.** Progress of the enzymatic reaction versus uncatalyzed one.

**2. Choice of the target enzyme**

326 Drug Discovery

times

At present, the most reliable method to determine three-dimensional architecture of transi‐ tion state is through the use of computational methods in conjunction with experimentally measured kinetic isotope effect (KIE).

Isotopic substitution is a useful technique for probing reaction mechanisms. The change of an isotope may affect the reaction rate in a number of ways, providing clues to the pathway of the reaction. The advantage of isotopic substitution is that this is the least disturbing structural change that can be effected in a molecule. Replacement of one isotope of the sub‐ strate by another at vicinity where bonds are being or re-hybridizing typically leads to a change in the rate of the reaction. Thus, kinetic isotope effects measurements compare kcat/KM values between isotope-labeled and natural abundance reactants. This provides in‐ formation about which bonds are broken or formed, and identifies changes in hybridization that occur during the rate limiting step of a reaction. It is reached by conversion of atom-byatom KIE values to a specific static model with fixed bond angles and lengths by computa‐ tional matching to a quantum chemical model of the reaction of interest. Substrate, intermediate and product geometries are located as the global minima. Transition-state structures are located with a single imaginary frequency, characteristic of true potential en‐ ergy saddle points.

Such an analysis was performed recently for human thymidine phosphorylase, an enzyme responsible for thymidine homeostasis, action of which promotes angiogenesis. Thus, inhib‐ itors of this enzyme might be considered as promising anticancer agents. Its transition state was characterized using multiple kinetic isotope effect measurements applying isotopically ( 3 H, 14C and 15N) enriched thymidines, which were synthesized enzymatically [Schwartz et al, 2010]. A transition state constrained to match the intrinsic KIEs was found using density functional theory. In the proposed mechanism (Fig.2), departure of the thymine results in a discrete ribocation intermediate. Thymine likely leaves deprotonated at N1 and undergoes enzyme-catalyzed protonation before the next step. In the following step, the intermediate undergoes nucleophilic attack from an activated water molecule to form the products. The latter step is a reaction rate limiting step as determined by energetics of its transition state. The transition state model predicts that deoxyribose adopts a mild 3′-*endo* conformation during nucleophilic capture (Fig. 2).

ies. Reaction mechanisms may have already been proposed in the literature, and thus pro‐ vide a logical starting point for modeling studies. The three-dimensional structure of the enzyme, preferably with a bound substrate analog, reaction product or inhibitor, is among the most critical sources of information. In practice, this usually means that a high-resolu‐

Transition State Analogues of Enzymatic Reaction as Potential Drugs

http://dx.doi.org/10.5772/52504

329

Molecular mechanics methods are important in simulations of enzymes, even though these methods cannot model chemical reactions. For that molecular dynamics simulations, or combination of molecular mechanics with quantum mechanical methods are commonly used [Senn & Thiel, 2007; Hou & Cui, 2011; Kosugi & Hayashi, 2012; Londsdale et al., 2012]. Enzymes are large molecules consisting of thousands of atoms whereas the active site may comprise only around 100 atoms. Since quantum chemical calculations are nowadays af‐ fordable only for up to a few hundred atoms (depending on the level of accuracy) the sys‐ tem is split into two regions: a small region encapsulating the reaction at the active site is modeled with a quantum mechanical methods, while the rest of the enzyme alongside with

**Figure 3.** Quantum mechanics/molecular mechanics calculation of an enzymatic reaction illustrated by cytochrome

These calculations do not take in the consideration such an important factor as protein dy‐ namic. There is an agreement that fast (at nano- or picosecond scale) protein motions couple directly to transition state formation in enzymatically catalyzed reactions and are an integral part of the reaction coordinate. Slower protein dynamic motions also influence the heights of barriers in enzymatic reactions, however detailed description of these effects require elab‐

tion X-ray crystal structure of a reacting enzyme complex is required.

surrounding water is modeled using molecular mechanics (Fig. 3.)

P450 with bound cyclohexene [Lonsdake et al., 2010].

oration of new computational methods [Saen-oon et al., 2008].

Such studies, although cumbersome and difficult, are being recently more and more popu‐ lar, as demonstrated by representative studies on *Escherichia coli t-*RNA-specific adenosine deaminase [Luo & Schramm, 2008], glucoside hydrolases [Lee et al, 2004], human purine nu‐ cleoside phosphorylase [Murkin et al., 2007], *Trypasnosoma cruzi trans*-sialidase [Pierdomini‐ ci-Sottile et al., 2011], *L-*dopa decarboxylase [Lin & Gao, 2011] or *cis-*prenyltransferase [Hu et al., 2010].

**Figure 2.** Mechanism of human thymidine phosphorylase catalyzed depirymidation of thymidine. The dash line repre‐ sents protonation step.

Computational chemistry provides techniques for the generation and exploration of the multi-dimensional energy surfaces that govern chemical reactivity. Therefore, energy mini‐ ma and saddle points can be located and characterized. The pathways that interconnect them can be determined. Thus, computational methods are increasingly at the forefront of elucidating mechanisms of enzyme-catalyzed reactions, and shedding light on the determi‐ nants of specificity and efficiency of catalysis [Kollman et al., 2002; Parks et al., 2010; Wil‐ liams, 2010; Lonsdake et al., 2012].

At the beginning of a molecular modeling study choice upon the specific catalytic process to model has to be undertaken. This decision may sound simple, but it includes the nontrivial task of exhaustively searching the literature to determine what is already known about the selected enzymatic system, either from experiments or from previous computational stud‐ ies. Reaction mechanisms may have already been proposed in the literature, and thus pro‐ vide a logical starting point for modeling studies. The three-dimensional structure of the enzyme, preferably with a bound substrate analog, reaction product or inhibitor, is among the most critical sources of information. In practice, this usually means that a high-resolu‐ tion X-ray crystal structure of a reacting enzyme complex is required.

The transition state model predicts that deoxyribose adopts a mild 3′-*endo* conformation

Such studies, although cumbersome and difficult, are being recently more and more popu‐ lar, as demonstrated by representative studies on *Escherichia coli t-*RNA-specific adenosine deaminase [Luo & Schramm, 2008], glucoside hydrolases [Lee et al, 2004], human purine nu‐ cleoside phosphorylase [Murkin et al., 2007], *Trypasnosoma cruzi trans*-sialidase [Pierdomini‐ ci-Sottile et al., 2011], *L-*dopa decarboxylase [Lin & Gao, 2011] or *cis-*prenyltransferase [Hu et

**Figure 2.** Mechanism of human thymidine phosphorylase catalyzed depirymidation of thymidine. The dash line repre‐

Computational chemistry provides techniques for the generation and exploration of the multi-dimensional energy surfaces that govern chemical reactivity. Therefore, energy mini‐ ma and saddle points can be located and characterized. The pathways that interconnect them can be determined. Thus, computational methods are increasingly at the forefront of elucidating mechanisms of enzyme-catalyzed reactions, and shedding light on the determi‐ nants of specificity and efficiency of catalysis [Kollman et al., 2002; Parks et al., 2010; Wil‐

At the beginning of a molecular modeling study choice upon the specific catalytic process to model has to be undertaken. This decision may sound simple, but it includes the nontrivial task of exhaustively searching the literature to determine what is already known about the selected enzymatic system, either from experiments or from previous computational stud‐

during nucleophilic capture (Fig. 2).

al., 2010].

328 Drug Discovery

sents protonation step.

liams, 2010; Lonsdake et al., 2012].

Molecular mechanics methods are important in simulations of enzymes, even though these methods cannot model chemical reactions. For that molecular dynamics simulations, or combination of molecular mechanics with quantum mechanical methods are commonly used [Senn & Thiel, 2007; Hou & Cui, 2011; Kosugi & Hayashi, 2012; Londsdale et al., 2012]. Enzymes are large molecules consisting of thousands of atoms whereas the active site may comprise only around 100 atoms. Since quantum chemical calculations are nowadays af‐ fordable only for up to a few hundred atoms (depending on the level of accuracy) the sys‐ tem is split into two regions: a small region encapsulating the reaction at the active site is modeled with a quantum mechanical methods, while the rest of the enzyme alongside with surrounding water is modeled using molecular mechanics (Fig. 3.)

**Figure 3.** Quantum mechanics/molecular mechanics calculation of an enzymatic reaction illustrated by cytochrome P450 with bound cyclohexene [Lonsdake et al., 2010].

These calculations do not take in the consideration such an important factor as protein dy‐ namic. There is an agreement that fast (at nano- or picosecond scale) protein motions couple directly to transition state formation in enzymatically catalyzed reactions and are an integral part of the reaction coordinate. Slower protein dynamic motions also influence the heights of barriers in enzymatic reactions, however detailed description of these effects require elab‐ oration of new computational methods [Saen-oon et al., 2008].
