Chapter 1. Introduction.

This research attempted to determine the conformational behavior of several building blocks employed in our group for the design of enzymatic inhibitors. To this effect, this dissertation covers three main projects:

1. A computational study of the energetics of the hydrolysis of acetylcarnitine, accounting for the effects of changes in the predominant conformation due to solvation.
2. Effect of substitution on the conformational behavior of 4,4-dimethylmorpholinium rings, studied by computational techniques.

   

   Figure 1.1. 4,4-dimethylmorpholinium ring.[fig.morph.gif, 0.44k]

3. Development of parameters for AMBER* force field, and modeling of the conformational behavior of 4,4-dimethyl-2-oxo-1,3,6-dioxazaphosphacinium rings.

   

   Figure 1.2. 2,4,4,-trimethyl-2-oxo-1,3,6-dioxazaphosphacinium ring.[fig.phosp.gif, 0.591KB]

1.1. Purpose of study.

Our group designs and synthesizes transition-structure analogs (TSA) as inhibitors for carnitine acetyl-, octanoyl- and palmitoyltransferases (CAT, COT and, CPT-I and CPT-II). Normally the designer of an inhibitor tries to imitate the geometry of the transition structure [Wolfenden, 1969] for the reaction catalyzed by the enzyme. Ideally, knowing the shape of the active site in an enzyme would enable us to design strong inhibitors. Presently, we only know the aminoacid sequences for CAT, COT, CPT-I and CPT-II, but their tertiary structures remain unknown. Until the resolution of the tertiary structures, only indirect methods can yield information about the shape of the active site. Knowledge of the preferred conformations adopted by TSAs, together with their measured inhibitory potency, can shed some light on the shape of the active site of these enzymes.

Inhibitors for CAT, COT and CPT can help us understand the role of each carnitine acyltransferase in the cell. Such understanding will help us in the design of potentially therapeutic compounds for diseases like diabetes and myocardial ischemia.

As the costs of experimentation and waste disposal continue to increase, it is imperative to find more efficient ways to choose targets for synthesis, especially those compounds with a greater chance of success as inhibitors. Computer modeling allows us to examine many compounds before choosing a target. The question then becomes: are there any modeling methods refined to the point they can yield reliable results for charged species? In particular, modeling of species with charges such as morpholinium, and zwitterions as that of carnitine in solution is difficult. [Note.] We need, however, such modeling to understand the conformational preferences of the reactants --acetylcarnitine, carnitine-- and the enzymic reaction intermediate to achieve a more rational design of inhibitors. As several of our inhibitors include phosphonate substructures, we require methods able to deal with them. Such methods are not widely available; modeling of phosphonates has certainly lagged behind that of more popular phosphorus species, such as phosphates.

For these purposes we want to assess the ability of current state-of-the-art computational methods to model the compounds of our concern. We put particular emphasis on the inclusion of solvent effects, so we want to see the abilities of methods such as COSMO to reproduce properties such as dipole moments and heats of formation in solution. As we need to consider hundreds of conformations and configurations in crowded, charged species, optimization of computer resources is essential. We therefore need to evaluate the applicability of ab initio, semiempirical and molecular mechanics methods so that we spend a minimum of resources without undermining the quality of the predictions.

Among the problems we face are: the high free energy of hydrolysis of acetylcarnitine, which qualifies it as a high-energy molecule. This high energy has remained unexplained for about three decades. [Friedman, 1955] The conformational dynamics of six- and eight-membered heterocycles are still imperfectly understood. [Eliel, 1994] All these pieces of information can help us understand the mode of action of the inhibitors already available to us.

We can take two approaches to deal with modeling:

1. As a tool oriented to engineering of results, meaning that achieving experimental accuracy is the most important matter, even though the methods may not have full theoretical justification.
2. As a research tool, where the theoretical basis of the method can provide insight on the mechanisms underlying the phenomenon of interest, even if the approximation to experiment is only qualitative.

In general, molecular modelers use a combination of the two approaches, and it would be difficult to find examples of ``pure'' techniques, (techniques that adhere strictly to only one approach). Regardless of the ``purity'' of the technique employed, if we can model experimental results, we can gain confidence on the predictions generated for systems that have not been tested yet. How closely should the modeling results approach the experimental results to render the modeling as reliable? This is a somewhat subjective matter, as there are no hard guidelines to make such a decision.

[F.SOLVATION]

Cramer and Truhlar [Cramer, 1995] have reviewed the performance of several solvation models and critically analyzed their implementations. One of their conclusions is that few software packages take into account both the contributions of charge and of multipolar moments, such as dipole, quadrupole, octapole and hexadecapole moments.