Conformational Memories and the Exploration of Biologically Relevant
Conformations of Anandamide

Judy Barnett-Norris, Frank Guarnieri and Patricia Reggio

 Introduction
 Methods
 Results
 Conclusions

Introduction

The endogenous cannabinoid ligand arachidonylethanolamide (anandamide) is an inherently flexible ligand. While the presence of unsaturation does constrain the structure somewhat, there are still 16 torsion angles that can vary, making hundreds of low energy conformations possible. The presence of so many variable torsion angles makes a simple conformational analysis using the molecular mechanics dihedral driver approach, for example, unwieldy. An alternative approach which is used with such flexible ligands is constrained Molecular Dynamics in which a hypothesized pharmacophore alignment is maintained during the simulation, thus resulting in an exploration of a limited part of the conformational space of the ligand.

We have recently employed a new method for the conformational analysis of highly flexible ligands, the Conformational Memories technique [F. Guarnieri and H. Weinstein J. Am. Chem. Soc. 118, 5580-5589, 1996.]. This method permits complete sampling of conformational space. It not only yields the energy of each conformation, but also the probability that the ligand will adopt each particular conformation relative to all the other conformations accessible in an equilibrated thermodynamic ensemble. PGB₂-EA is a prostaglandin ethanolamide synthesized and tested for CB1 affinity by Dr. Allyn Howlett some three years ago [J.C. Pinto et al Mol. Pharmacol. 45, 516-522, 1994]. PGB₂-EA failed to alter [³H]CP-55,940 binding to CB1 at concentrations up to 100mM. We applied the Conformational Memories technique to PGB₂-EA to explore if conformational differences between anandamide and PGB₂-EA could explain why one bound to CB1 and the other did not bind to CB1.

Methods

The Conformational Memories technique employs multiple Monte Carlo annealing (MC/SA) random walks using the MM3 force field and the Born/surface area (GB/SA) continuum solvation model (chloroform in this case) as implemented in the Macromodel molecular modeling package. The molecular simulation technique of Conformational Memories is a two stage process consisting of an Exploratory Phase and a Biased Sampling Phase.

A. Exploratory Phase

In the Exploratory Phase, repeated runs of Monte Carlo simulated annealing (MC/SA) are carried out in order to map the entire conformational space of the flexible ligand. The simulated annealing of the Exploratory Phase uses 19 temperatures from 2070K to 310K. We used many different starting structures and did multiple runs to check the consistency of our results. The anandamide run used as an example here was started from an anandamide structure which had been overlayed with the key pharmacophoric elements of D ⁹-THC and then minimized before submission to the Conformational Memories program. The collective histories of all random walks are transformed into mean field dihedral distribution functions called "conformational memories" (e.g. sixteen conformational memories for anandamide and fourteen for PGB₂-EA; one for each torsion angle).

The figure provides an example of the graphical output from the Exploratory phase. On the y axes are dihedral angle values from -170 to 180 degrees. On the X axes are the 19 temperatures ranging from 2070K to 310K. The population percentage is plotted in the Z direction. The plots yield the identification of structural motifs. For example, the graph for dihedral 3, exhibits a classic three state distribution seen for rotation about a bond that connects two sp³ hybridized carbons: trans, gauche plus, and gauche minus. For Dihedral 8, which is for rotation about a bond that connects the sp³ hybridized methylene carbon to an sp² hybridized carbon, we see two peaks corresponding to classic skew angles (118° ) seen in the crystal structure of arachidonic acid, for example.

B. Biased Sampling Phase

The biased sampling phase uses a temperature slice from the memories. We used the 19th temperature, 310K. The MC random walk in the 2nd phase, therefore samples only from the populated areas in the specified slice (310K for us). We sampled 100 structures from a 500,000 step MC random walk. Each resulting batch of 100 structures was analyzed with a program (X-Cluster) that inputs the series of 100 conformations and computes the root-mean-square (RMS) difference between all possible pairs of conformations to form clusters of conformational families.

Results

A. Anandamide

When we analyzed the conformational families of anandamide three large families were found. The first family was an extended conformation (this was the major cluster), the second was a hairpin or U shape and the third was a helical shape. The figure shows all of the members of the predominant cluster, the extended shape.

B. PGB₂ - EA

For PGB₂-EA, we found two large clusters. In the first cluster, the predominant one, there were U or "rockette" type conformers. A smaller percentage of structures were in an extended or L shape. The figure shows all of the members of the "rockette" family. A comparison of the shape of the predominant anandamide family with that of the predominant PGB₂-EA family, reveals their shapes to be quite different.

C. Receptor Docking : Model CB1 Receptor

Next, we explored how each major cluster of both anandamide and PGB₂-EA would fit inside our CB1 receptor model. In our model, classical cannabinoids, non-classical cannabinoids and anandamide are hypothesized to have two key interaction sites: K3.28 and V6.43 / I 6.46. Each conformer was docked using K3.28 as an interaction site for the carbonyl oxygen. (The carbonyl oxygen rather than the hydroxyl is hypothesized to be an important interaction site for the ligand because ligands which have the hydroxyl group replaced by a methyl group have been shown to have higher CB1 affinity indicating that the hydroxyl is not essential for affinity. [J.C. Pinto et al. Mol. Pharmacol. 46, 516-522, 1994].) The hydrophobic binding pocket (groove) formed by V6.43 and I 6.46 was used as a second interaction site, one for the pentyl tail of each conformer.

For anandamide, we found that the predominant, extended shape worked well inside the receptor. The other anandamide shapes: the U or haripin, and the helical shape did not fit inside the receptor due to profound steric clashes and/or an inability to reach both key interaction sites.

For PGB₂-EA, neither shape would dock into the CB1 receptor model.

Conclusions

The results of the Conformational Memories Calculation have identified the major conformational families of anandamide and PGB₂-EA to be an extended shape and a U or rockette shape, respectively. Our results point to the extended shape of anandamide as the conformation that anandamide prefers in CHCl₃. The preferred conformation of PGB₂-EA is quite different even though 2 dimensional drawings make the two ligands look very much alike. Furthermore, we have found that only the extended form of anandamide fits inside our CB1 receptor model.