Electrostatic Origin of the Catalytic Power of Enzymes and the Role of Preorganized Active Sites*

Enzymatic reactions are involved in most biological processes. Thus, there is a major practical and fundamental interest in finding out what makes enzymes so efficient. Many crucial pieces of this puzzle were provided by biochemical and structural studies (1). Yet, as will be shown below, the actual reason for the catalytic power of enzymes is not widely understood. It is clearly not explained by the statement that " the enzyme binds the transition state stronger than the ground state " because the real question is how the differential binding can be accomplished. Similarly, it is not true that " evolution can use any factor to accelerate reactions. " This review uses energy considerations and the results of computational studies to clarify open questions about enzyme catalysis. Examples of the enormous catalytic power of enzymes have been eloquently compiled by Wolfenden (2). As is clear from this compilation and from kinetic considerations (e.g. Ref. 1), many enzymes evolved by optimizing k cat /K m. The energetics associated with the interplay between k cat and K m is considered in Fig. 1. As is obvious from this figure, enzymes can increase their rate by providing strong binding to the ground state (reducing K m) and reducing ⌬g cat ‡ (increasing k cat). As will be emphasized below, the effect due to K m is quite clear and does not present a major puzzle. The real challenge is to find out what was used by evolution in increasing k cat. The fact that k cat increases upon reduction of ⌬g cat ‡ was basically stated by Pauling (3). However, the way by which ⌬g cat ‡ is reduced was not addressed (except in a suggestion of a strain effect, which is now known to be rather small). Thus, the key question of how this reduction of ⌬g cat ‡ is accomplished remained entirely unresolved. The studies of Jencks and co-workers (4) provided a major insight by emphasizing the effect of binding energy and by trying to find out how an enzyme can reduce ⌬g cat ‡. In the absence of other reasonable explanations (see below) it was concluded that ⌬g cat ‡ is reduced by ground state (GS) 1 destabilization, which was attributed to entropic, strain, and desolvation effects (4). These effects were invoked by many workers in the field (see below) and were fully consistent with the options available to organic chemists …

Examples of the enormous catalytic power of enzymes have been eloquently compiled by Wolfenden (2). As is clear from this compilation and from kinetic considerations (e.g. Ref. 1), many enzymes evolved by optimizing k cat /K m . The energetics associated with the interplay between k cat and K m is considered in Fig. 1. As is obvious from this figure, enzymes can increase their rate by providing strong binding to the ground state (reducing K m ) and reducing ⌬g cat ‡ (increasing k cat ). As will be emphasized below, the effect due to K m is quite clear and does not present a major puzzle. The real challenge is to find out what was used by evolution in increasing k cat . The fact that k cat increases upon reduction of ⌬g cat ‡ was basically stated by Pauling (3). However, the way by which ⌬g cat ‡ is reduced was not addressed (except in a suggestion of a strain effect, which is now known to be rather small). Thus, the key question of how this reduction of ⌬g cat ‡ is accomplished remained entirely unresolved.
The studies of Jencks and co-workers (4) provided a major insight by emphasizing the effect of binding energy and by trying to find out how an enzyme can reduce ⌬g cat ‡ . In the absence of other reasonable explanations (see below) it was concluded that ⌬g cat ‡ is reduced by ground state (GS) 1 destabilization, which was attributed to entropic, strain, and desolvation effects (4). These effects were invoked by many workers in the field (see below) and were fully consistent with the options available to organic chemists in catalyzing reactions in solution. Yet, more recent energy considerations have indicated that such effects are not likely to give large contributions to enzyme catalysis. The problems with GS destabilization mechanisms are illustrated by Fig. 1; increasing the GS energy without changing the TS energy will leave ⌬g ‡ unchanged. In other words, such changes will alter k cat and K m but leave k cat /K m constant (except in the diffusion control limit when k cat is larger than k Ϫ1 , which also supports our conclusion about GS stabilization 2 ). Thus, an enzyme that was optimized under the evolutionary pressure of increasing k cat /K m would not gain much from changing its GS energy without changing the TS energy. To verify the above point, we introduce here a new framework for analyzing mutation experiments that should allow the reader to determine what was really done by evolution. To do so we classify mutation experiments into the three classes described in Fig. 2. If the processes of evolution lead to ground state destabilization, then going backward in evolution by mutations should lead to ground state stabilization, while leaving the TS energy unchanged. This will involve mutations that can be referred to as "GS mutations." If an enzyme has evolved to stabilize its TS, then we will have mutations that can be referred to as "TS mutations." If binding energy is used to stabilize the GS and TS by the same amount we will have mutations that can be referred to as "distant binding (DB) mutations" because such mutations are expected to operate by binding the substrate in regions that are far from its reactive part. An examination of mutations that reduce the activity of enzymes in a substantial way (5-10) has not found GS mutants but only TS and DB mutants. Obviously, our conclusions are not final, and the reader is clearly encouraged to use any mutation experiment.
In view of the above points, it seems to us that GS destabilization mechanisms cannot provide a general way of reducing ⌬g cat ‡ . Thus, one must look for ways by which enzymes stabilize the TS more than the GS. Finding such ways is far from trivial and requires one to address the total effect of the enzyme, while comparing it to the proper reference state in solution. Available experimental results do not yet tell us in a unique way how the total effect of the enzyme is distributed and what are the exact contributions of different free energy factors (e.g. strain, electrostatic, entropy, etc.). In particular, mutation experiments can help enormously by telling us what is the effect of different residues but cannot determine the origin of the overall catalysis. For example, mutating a group that forms a hydrogen bond (HB) to a negatively charged transition state in an active site, which is otherwise completely non-polar, will lead to a very large reduction in k cat . Some people upon being informed about this mutational effect will conclude that the native enzyme is a very good catalyst because its HB is so effective. However, a charge is very unstable in a non-polar environment, and a charged TS that is stabilized by a single HB will still leave us with an enzyme that destabilizes its TS relative to water (11).

Using Energy Considerations and Computer Simulations to Examine Catalytic Proposals
Many reasonable proposals have been put forward to account for the catalytic power of enzymes. However, well defined energy considerations and computer modeling studies are now indicating that most of these proposals cannot account for large catalytic effects. This does not imply that such proposals have not been reasonable but rather that it is difficult to validate them without qualitative and sometimes quantitative energy considerations.
To illustrate the effectiveness of energy considerations and computer modeling we consider below some of the main catalytic proposals and refer the reader to a more detailed analysis in Ref. 12.
The Desolvation Hypothesis-One of the most popular proposals is that enzymes work by providing a non-polar (or sometimes gas phase-like) environment that destabilizes highly charged ground states (13)(14)(15)(16)(17). As shown in Ref. 18 these proposals involve improper thermodynamic cycles and do not use a proper reference state. This amounts to ignoring the desolvation energy associated with taking the GS from water to a hypothetical non-polar enzyme site. With a proper reference state, one finds (18) that a polar TS is less stable in non-polar sites than in water and that the GS destabilization does not help in increasing k cat /K m . In fact, many desolvation models (e.g. Refs. 13, 16, and 17) involve ionized residues * This minireview will be reprinted in the 1998 Minireview Compendium, which will be available in December, 1998. This is the third article of three in the "Enzyme Catalytic Power Minireview Series." This work was supported by National Institutes of Health Grant GM 24492. 1 The abbreviations used are: GS, ground state; TS, transition state; DB, distant binding; HB, hydrogen bond; LBHB, low barrier hydrogen bond; EVB, empirical valence bond. 2 Evolving to the case when k Ϫ1 ϳ k cat cannot be accomplished by GS destabilization but by combining GS stabilization and TS stabilization. that will simply be unionized in non-polar sites. Moreover, in any specific case when the structure of the active site is known (e.g. Ref. 19), one finds by current electrostatic models a very polar (rather than non-polar) active site environment near the chemically active part of the substrate.
Binding Entropy-This idea (4,20) implies that the reduction of ⌬g cat ‡ is because of a restriction of motion of the reacting fragments of the substrate in the enzyme active site. This proposal involves ground state destabilization and is probably not so useful for enzymes that evolved by optimizing k cat /K m (see "The Real Problem is the Origin of the Reduction of ⌬g cat ‡ "). The interpretation of experiments in model compounds that were brought to support this proposal (4,20) is now questioned by computational studies (12,21), and it is also argued to be somewhat irrelevant (12). It is still possible that entropic effects contribute a few kcal/mol to the reduction of ⌬g cat, ‡ , but a quantitative judgment of the importance of entropic effects will have to involve accurate simulations of this effect.
Orbital Steering-This proposal (22) postulated a very narrow dependence of the transition state energy on the angle of approach of the reacting molecules. The corresponding rate acceleration is in conflict with the following. First, as was pointed out eloquently by Bruice and co-workers (23) and confirmed by subsequent calculations (12,25), this proposal requires an unreasonably large force constant. Second, and in some respects more importantly, the proposal has not been properly defined, because it did not consider the corresponding effect of the reference reaction in water (12). With a proper reference state, one finds that the proposed catalytic effect disappears unless the enzyme can constrain the reacting molecules in the ground state to the same narrow angular range as the corresponding range in the transition state of the solution reaction (12). Such a situation is extremely unlikely in flexible enzymes and has little to do with orbitals because we have the same reacting orbitals in the protein and in solution. In this way we simply have another proposal of ground state destabilization by orientational entropy. Finally, a recent work (24) that supported the orbital steering model can be interpreted differently. 3 Ionic Low Barrier Hydrogen Bonds (LBHBs) versus Regular Ionic Hydrogen Bond-This proposal (26 -28) involves three points: (a) ionic HBs that include a negatively charged proton acceptor contribute in a major way to transition state stabilization; (b) the catalytic effect of HBs is associated with their preorganization (28); and (c) hydrogen bonds between the enzyme and charged transition states of substrates involve a much larger covalent character (X Ϫ␦ ⅐⅐⅐H⅐⅐⅐Y Ϫ␦ ) than the corresponding hydrogen bonds in solution; otherwise we have a regular (X Ϫ H-Y) electrostatic HB. The first two parts of the proposal are correct but were introduced long ago as a major part of the electrostatic effect of enzymes (12,29,30). Our analysis indicates that the new assumption of a larger covalent character for HBs in enzymes than in solution results in a reduction of the catalytic effect of these HBs (11,31). The reason is that LBHBs can be effective only in a non-polar environment, but such an environment results in HBs that give less TS stabilization than regular electrostatic HBs. That is, LBHBs can provide very large stabilization to charged TSs in a non-polar enzyme surrounded by non-polar solvent rather than water. However, immersing this enzyme in water will create a very poor catalyst because of the above mentioned desolvation problem. In a polar environment, one finds (11) that an LBHB provides less TS stabilization than a regular HB. In this respect, it is crucial to realize that the alternative to an ionic LBHB is an ionic HB, which acts through electrostatic effects. Thus, in a complete contrast to some assertions (32) we never suggested that the alternative to an ionic LBHB is a neutral (X H-Y) HB. All discussions of strong HBs in enzymes involved ionic TSs, and it was never suggested that regular catalytic HBs are weak (33) but demonstrated that HBs are strong (34).
Distinguishing between LBHBs and HBs is a question of interpretation of experiments rather than experiments alone, and current experiments cannot be interpreted unambiguously (11). A case in point is a recent NMR study (35) that suggested a very different interpretation from that of previous studies (26). Energy considerations can help in reaching a more unique analysis of the nature of ionic HBs in enzyme (31). For example, in the case of serine proteases, it was shown (11,31) that the pK a of the catalytic His-57 and Asp-102 must be very different to stabilize the transition state (relative to water), and thus we cannot have an LBHB in this system. The same arguments can be used to show that a new analysis (33,36) of a negatively charged TS analog of serine proteases proves that the LBHB hypothesis is very problematic. That is, in this case, the pK a of His-57 is raised to about 12, and this was attributed to bindinginduced ground state strain, which has been assumed to be consistent with the LBHB hypothesis (33). However, the presence of a negatively charged TS analog stabilizes the protonated form of His-57 much more than it destabilized the more distant and negatively charged Asp-102. Thus, the TS analog increases rather than decreases the pK a difference in the Asp-His pair, which is, of course, consistent with the experimentally observed pK a . This trend contradicts the LBHB proposal, which requires a matching pK a . Furthermore, the proposed strain effect is inconsistent with mutational studies that moved the COO Ϫ group from residue 102 to residue 214 and still obtained significant catalysis (37).
It is useful to realize that recent theoretical studies (32,38), which were discussed in the first minireview in this series (36), in support of the LBHB hypothesis reflect an improper analysis of an irrelevant system (i.e. no protein is studied) and apparently fail to address the relevant question (i.e. confuse ionic HB with LBHB). That is, the theoretical study (38) that is taken as a support of the LBHB hypothesis (36) attempted to model an enzyme by placing two water molecules around an HB in the gas phase. The resulting ionic HB was found to be extremely strong relative to the improper reference of the dissociated HB in the gas phase. Unfortunately, this strong HB has a very high energy (relative to its energy in water), because it involves a charge in a non-polar environment. Thus, we have here another illustration of the anticatalytic nature 3

FIG. 2. A schematic illustration of the effect of different classes of mutations.
As argued in the text, it is very hard to find GS mutations that lead to a large change in k cat /K m . This indicates that evolution has not used GS destabilization in refining the efficiency of most enzymes. However, the reader is encouraged to use this diagram to find exception to our current observation.

FIG. 1. A schematic description of the energetics of enzymatic reactions.
A is the preexponential factor in the rate expression and ⌬G b is the bind free energy. The figure demonstrates how a reduction of ⌬G b , by a ground state destabilization effect, changes ⌬g cat ‡ but has no effect on ⌬g ‡ or the corresponding k cat /K m . Note that K m ϭ (k Ϫ1 ϩ k cat )/k 1 , but in most cases k Ϫ1 Ͼ Ͼ k cat so that K m ϳ k Ϫ1 /k 1 ϭ K s (where K s is the dissociation constant).
of LBHBs. Similarly, the computational studies of an LBHB in polar solvent (32) that were brought in support of the LBHB proposal (36) involved a comparison of an ionic LBHB and a neutral HB. This has little to do with the LBHB issue which involves the comparison of an ionic LBHB and an ionic HB (see above). Furthermore, Ref. 32 grossly overestimates the actual strength (39) of ionic HBs in polar solvents.
Dynamic Effects-This proposal (40 -42) implies that the enzyme might induce special fluctuations that do not obey the Eyring's absolute rate theory, leading to a transmission factor that is much smaller than unity (note that the transmission factor contains all dynamic effects (see Ref. 43) and that factors such as activation energies that can be obtained by Monte Carlo simulations, rather than only by Molecular Dynamics simulations, are not dynamic effects). However, our simulation studies established quite early that the reactive fluctuations are similar in enzymes and solutions (43) and that the transmission factor is similar in both cases (12,34). The same conclusion was recently reached by others (44). Nevertheless, it was implied in Ref. 44 that the nonequilibrium solvation in enzymes is fundamentally different from the corresponding effects in solution. However, this conclusion was reached without attempting to calculate the dynamics in a solvent cage or to evaluate the activation free energy and its non-equilibrium contributions in the enzyme and in solution. In fact, early studies that compared the dynamics of enzymatic reactions to the corresponding reaction in solution found that both systems have similar solvent fluctuations and similar solute fluctuations (12,43). The only difference is that the amplitude of the fluctuations is smaller in the more preorganized environment. Interestingly, a recent simulation study 4 established that non-equilibrium solvation effects are of similar magnitude in enzymes and in the corresponding solution reaction.
Tunneling-This proposal (45,46) suggests that enzymes can accelerate their reactions by exploiting nuclear tunneling effects. Well chosen experimental studies have indicated that the degree of tunneling changes when the rate constant changes upon mutations (46). However, it was not yet demonstrated that the change in rate constants is because of tunneling (because the large effect of mutations did not change upon isotopic substitution). It is quite possible, for example, that the mutations change the reorganization energy, and this changes the tunneling rather than the other way around. Simulation studies (e.g. Ref. 47) have shown that enzymatic reactions may involve significant tunneling effects, but the corresponding catalytic effects are not expected to be large because similar tunneling corrections occur in the enzyme and the reference solution reaction.
In summary, non-negligible contributions may still be provided by entropy, tunneling, and perhaps some LBHB character, although these are not supported by current simulation studies. Large contributions are expected from electrostatic effects (see below).

Electrostatic Effects Provide the Most Important
Contributions to Enzyme Catalysis As indicated above it is frequently simpler to show what does not lead to catalysis than to establish what does. Even the interpretation of mutation experiments is far from being unique. For example, the large observed effect of mutating Asp-32 in subtilisin can be interpreted as an electrostatic effect (12,48), an LBHB effect (26,27), and an entropic effect (49). It seems to us that only a quantitative, structurally based analysis can help in providing a more unique interpretation. The reason is quite simple; we are dealing with a very complex system with many free energy contributions, and a proper analysis requires a quantitative model that can handle such complexity and predict rather than assume the relevant contributions. Such an analysis can be done by computer simulations.
Computer-aided methods for correlating the structures of enzymes have been developed (12), and some of these approaches, which are referred to as hybrid quantum mechanics/molecular mechanics (QM/ MM) methods (50), are becoming very popular (e.g. Refs. 51 and 52). Yet at present, the only approach that reaches the level of quantitative analysis is the empirical valence bond (EVB) method (12,53). This approach is quantitative because it does not attempt to calculate the energetics of bond-making/bond-breaking processes by an ab initio first principle approach (which requires at present too much computer power) but focuses only on the change in such energetics upon moving from the reference solvent cage to the enzyme active site. This is done by parameterizing the EVB Hamiltonian on ab initio studies and on experimental information about chemical reactions in solution and then keeping the parameters completely unchanged when exploring the free energy of moving from solution to enzymes. Thus, we can determine quantitatively free energy contributions to catalysis without the use of any free parameters. The EVB model has been recently adopted by other research groups who found its features useful for studying reactions in solutions and enzymes (44, 54 -56). All EVB studies (e.g. Refs. 12, 34, 48, 53, and 57) and even more qualitative QM/MM studies (e.g. Refs. 51 and 52), studies that took the protein plus solvent into account, have indicated repeatedly that enzyme catalysis is due mainly to electrostatic effects. It is important to note here that using computational approaches does not always guarantee relevant results. As is now becoming clear in the computational community, high level ab initio calculations that do not include the enzyme cannot tell us much about the role of the enzyme.
Some readers might assume that the importance of electrostatic effects in enzyme catalysis is a rather trivial idea that has been established in the early days of the field. However, although electrostatic effects were proposed quite early (58), they were found to be inconsistent with all early experimental studies. That is, experiments with model compounds in solutions show very small electrostatic effects (20,59). Similarly, changes of ionic strength that were used to probe electrostatic effects did not produce major changes in enzyme activity. Thus, until the emergence of genetic engineering, there was no major experimental evidence for large electrostatic contributions to catalytic effects of enzymes. Theoretical studies, on the other hand, pointed to such effects repeatedly (12,30,50). The difficulties in early experimental verification were traced to the fact that the interior of enzymes cannot be probed by external perturbation and that experiments with model compounds in solution have not produced large electrostatic effects because of large dielectric effects.

Electrostatic Stabilization Is Due to Small
Reorganization Energies The proposal that enzymes work by electrostatic stabilization mechanisms has to overcome one major fundamental problem. That is, it is not obvious how an enzyme active site can provide more electrostatic stabilization than water. More specifically, computer simulation studies (34) indicated that the actual electrostatic interaction between the enzyme and the TS of its substrate, ⌬G Q , is similar to that between water and the corresponding TS. If the interactions are similar, one would expect no catalytic effect. The solution to this fundamental problem has been provided in an early work (30). This work pointed out that in polar solvents about half of the energy gained from charge-dipole interaction is spent on changing the dipole-dipole interaction, ⌬G , so that the free energy of solvation of the transition state is given by (12) the following equation.
In proteins, however, the active site dipoles associated with polar groups, internal water molecules, and ionized residues are already partially oriented toward the transition state charge center. Thus, ⌬G is smaller than in water, and less free energy is spent on creating the oriented dipoles of the protein transition state (Fig. 3). The free energy term ⌬G is basically the so-called "reorganization energy" (12,60) for the process of forming the transition state charges. For example, in water we have to break water-water interactions to form good hydrogen bonds to the TS. In the enzyme, on the other hand, the hydrogen bonds are already partially oriented toward the transition state charges (34). The prediction that the folding energy is used to preorient the enzyme dipoles is supported by the finding that mutations that increase ⌬g cat ‡ also increase the folding energy (61).
The idea that enzymes use preorganized dipoles to catalyze their reactions should not be confused with the role of the Marcus reorganization energy in enzymatic reactions. That is, the activation energy for chemical reactions can be written as (12, 62) follows, where ⌬G 0 is the free energy of moving from the reactant to the product (or an intermediate) and is the so-called "solvent reorganization energy," which mainly reflects the changes in the solventsolvent interaction during the reaction. H 12 describes the mixing between the reactant and product state. is reduced, we state now that ⌬G 0 and/or are reduced. However, we are still left without explaining how these free energies are reduced. Furthermore, insightful but incorrect suggestions that is reduced by having non-polar active sites (65) does not resolve this problem; although non-polar environment does reduce , it would lead to an increase of ⌬G 0 and to anticatalytic effects (see discussion in Ref. 66). What is missing in this proposal is the idea that enzymes reduce both and ⌬G 0 by preorganized polar (rather than non-polar) environment as established by simulation studies (12,53,66). The origin of this reduction is directly connected to the above mentioned preorganized polar environment. Such an environment stabilizes charged intermediates by not investing in ⌬G , and as a result of having preorganized dipoles it also reduces .
If enzymes really use their preoriented environment to stabilize the transition state, then we understand why it was so difficult to elucidate the origin of enzyme catalysis. In this case, the catalytic energy is not stored in the enzyme-substrate interaction but in the enzyme itself. Thus, for example, the reduction of ⌬g cat ‡ is not associated with the entropy loss upon assembly of the substrate fragments but with the free energy invested in fixing the environment.
In view of the length of our arguments, we will summarize their main points as follows. (i) Enzymes attain a large k cat by providing more stabilization to the charges of transition states than the corresponding stabilization in water. (ii) It is not true that evolution can increase k cat by using any free energy factor, because many factors do not provide a physical way of doing so. (iii) Although binding of the "non-chemical part" of the substrate can help in stabilizing the TS, it cannot help in providing a large reduction to ⌬g cat ‡ . Such a reduction requires an active site that is able to stabilize the reacting (chemical) part of the TS more than the corresponding solvent cage does. (iv) The reduction of ⌬g cat ‡ is accomplished by electrostatic stabilization, which is due to a preorganized polar environment (30) .  FIG. 3. Illustrating the catalytic origin of enzyme catalysis. In water (A) we have to pay for orienting the solvent, and this reduces the solvation free energy. In enzymes (B) we already have preorientation dipoles so we do not have to pay so much for the increase in dipole-dipole repulsion (⌬G ).