Discussion

The racemization reaction of propargylglycolate and mandelate by the enzyme Mandelate Racemase takes place through at least three mechanisms (mec I, mec II and mec III), that are analogous to those previously found for vinylglycolate[212] as a substrate.

The stabilization of the anionic intermediates seems to be the clue to understand the catalytic role of the enzyme. This stabilization is achieved, on the one hand, by the interaction of the substrate with the active site residues and, on the other hand, by the particular substrate structure. As a consequence, in good agreement with the kinetic experimental data, we have shown that mandelate is racemized faster than vinylglycolate, which, in turn, is isomerized faster than propargylglycolate.

The first step of mec I and mec II consists of a catalytic proton transfer from Lys164 or Glu317, respectively, to a carboxylate oxygen of the substrate through the corresponding hydrogen bond. That proton transfer precedes the abstraction of the $\alpha$-proton (the second step), so avoiding the accumulation of more than one negative charge on the substrate.

The proton transfer along the hydrogen bond between Glu317 and each (S)-enantiomer (mec II) ``catalyzes'' more efficiently the $\alpha$-proton abstraction by Lys166. This fact is in agreement with the experimental proposal by Gerlt and Gassman[238,215] confirmed also by mutagenesis experiments[214], about the importance of Glu317 as a putative general acid catalyst in the Mandelate Racemase pathway and about the role of this residue in reducing the pKa of the $\alpha$-proton.

In contrast, Lys164 has always been described as a charged residue contributing with its $\epsilon$-ammonium group to the positive electrostatic environment of the active site[229]. However, Gerlt and coworkers[228] also indicate that it is likely that Lys164 participates by donating a proton to the substrate carboxyl group but that the confirmation of the proposed binding mode would require additional experimental verification.

Due to the stabilization of the $\alpha$-proton abstraction transition state, we have seen that the highest energy point on the reaction path of mec I and mec II for (S)-propargylglycolate and (S)-mandelate corresponds to different steps of the mechanism: the hybridization change of the substrate and the final proton transfer from the substrate back to Lys164 or Glu317, respectively.

In mec III (the fastest) the stabilization of the intermediate is achieved mainly by the His297 residue. Moreover, it is in this mechanism where the structure of the substrate plays a more decisive role in stabilizing the stationary points found along the path. Without the catalytic proton transfers, more negative charge accumulates on the substrate and its efficiency to delocalize this extra charge is crucial for the enzymatic process. This is the reason why, among the three substrates, mandelate presents the lowest potential energy barrier in mec III.

In addition, we have found minimum energy structures corresponding to formal dianionic intermediates only on the QM/MM potential energy surfaces for vinylglycolate + MR complex (previous paper[212]) and for mandelate + MR complex in this paper. The viability of mec III and the existence of those intermediates (even though they are only slightly stabilized on our semiempirical QM/MM potential energy surfaces with respect to the second-step transition state) are in qualitative agreement with the experimental proposal of a stepwise racemization mechanism via a transiently stable intermediate[214]. The formation of this intermediate is required to explain several experimental isotope effects although its concentration has not yet been determined[217,227].

Finally we must remark that when exploring an enzymatic mechanism with optimization methods some differences from a gas-phase study must be taken into account. An enzyme presents many parallel reaction pathways on a PES which might be very smooth in different regions ^3.10. In this sense, a change between two parallel pathways or two minima is very easy during a bad step of an optimization. Keeping this in mind we must accept the following consequences.

• An accurate optimization method is needed to avoid bad steps which make the system escape from the local valley
• Despite of the last point, we will find small energetic differences between two minima energy structures without any significant geometry differences. The energetic difference is distributed among the high number of degrees of freedom.
• Sometimes the geometric differences between two intermediates will be not relevant to the mechanism.

In conclusion, we need a more accurate optimizer, mainly for the TS structure. So, before any inclusion of the dynamics and temperature effects (chapter 4) an appropriate method to locate TS in big systems will be developed in chapter 3.