Previous Research
This is a brief description of my past work; all numbered references are listed at the end.
Quantitation of steric factors in molecular recognition
My work began with a quantitative description of the geometry of any given molecule which was obtained by mapping the radial distance from the centre of mass of the molecule to the molecular surface onto a sphere [1]. Different molecules were compared by rotating the spheres [2]. This method was applied to saxitoxin and tetrodotoxin, both of which bind to the neuronal sodium channel, but possess very different chemical structures. Application of this method made it possible to identify the common structural motif from vastly diverse structures [3]. This method of projection and matching has been incorporated into a similarity program by Oxford Molecular Group (now Accelrys).
Role of electrostatics in molecular recognition
The research that followed demonstrated that it was the electrostatic potential on the apposing surfaces of the ligand and the receptor which determine the electrostatic complementarity between a ligand-receptor pair, and not the partial charges of ligand atoms against those of receptors. This effect was demonstrated by using 34 high-quality datasets of ligand-receptor complexes whose atomic crystal coordinates were available in the Protein Databank. These molecules were treated as rigid molecules, the water molecules approximated as a continuum, and the potential complementarity quantified by regression analysis [4-7].
This potential complementarity principle was applied to optimise the electrostatic interaction between ligands and receptors. A charge assignment method was developed to maximise complementarity by optimising the electrostatic potential on the designed ligand, and this was achieved through assigning appropriate partial charges to atoms of the hypothetical ligand. This method was applied successfully on five receptors, including the bacterial dihydrofolate reductase, a well-known target for antibiotics [8-9].
A fragments database was then created in which each fragment possessed known and consistent geometric and electrostatic properties. This set of fragments were combinatorially generated, were commonly found in known bio-organic molecules, and contained three to eighteen atoms [10]. Their shapes [11] and electric charge distributions [12], when these fragments were constituent parts of bigger molecules, were examined. Regardless of the other parts of the larger molecule, all these fragments retained their respective shapes and charge distributions. These properties make them ideal basic building blocks for designing ligands against a receptor. Novel ligands with predictable properties could be assembled from these fragments against a given receptor site. This molecular fragments database has been incorporated into a commercial design package by Tripos Associates Inc. (St. Louis, Missouri, USA).
Models of hydration in molecular recognition
The above study was generalised to an analysis of ligand-receptor interactions involving flexible rather than rigid molecules, and also including discrete water molecules, all of which were free to move, to give a more realistic representation of these interactions. This entailed the introduction of molecular dynamics simulations. They were first performed on convex and concave surfaces of radii of from 2Å to 12Å, which were models for the ligand and receptor, respectively, to examine hydration effects alone.
The presence of a convex hydrophobic solute primarily influenced water in its first solvation shell; these water molecules possessed fewer hydrogen bonds, and exhibited reduced translational and rotational dynamics [13-15]. In contrast, concave surfaces were associated with more rapid rotation of the neighbouring water molecules [16]. My new analysis methods of molecular dynamics data quantified these changes [17], and showed that they implied a decreased entropy in the first solvation shell water molecules around hydrophobic solutes.
Ligand-receptor interaction : retinol-binding protein/retinol pair
These methods led to a characterisation of the changes in water structure and dynamics as retinol unbound from the serum retinol-binding protein. This entailed a 1-ns simulation of this ligand-receptor pair in water using a novel unbinding method of mutual repulsion, which placed forces of equal magnitude but opposite directions on the ligand and the receptor, and allowed the system to locate an unbinding trajectory [18].
Taken together, these procedures culminated in the first direct demonstration of hydrophobic interaction. This is a water exclusion effect when two hydrophobic solutes bind together, with concomitant increase of entropy. So conversely, when the ligand unbinds, more water molecules will be associated with the ligand and the receptor, and the entropy will decrease. I discovered that as retinol unbound from the protein, more water molecules moved to coat retinol; some of these molecules changed from a bulk-phase configuration to hydrophobic hydration, and the entropy of the system decreased on ligand unbinding. These are the hallmarks of hydrophobic interaction. A rough estimate of the entropy change compares favourably with the experimental value [18]. There was also little movement of the binding site as the ligand dissociated from the protein [19].
I identified the extensive re-arrangement of binding-site water molecules concomitant with ligand unbinding: approximately half of the water molecules inside the binding site exhibited little movement, one-fifth of them moved a long distance, as much as 13 Å, to occupy the space vacated by retinol, while the rest showed a moderate degree of movement [20]. Surprisingly, water molecules outside the binding site did not enter it to take up the space vacated by retinol. Indeed, there was a slight decrease in the number of water molecules inside the binding site as the ligand left the receptor.
Ion channel receptors
I have now extended this work to the cys-loop family of receptors, to which ligand binding alters the function of the ion channel in the centre of the receptor. These receptors include the nicotinic acetylcholine receptor, the GABAA receptor and the 5-HT3 receptor, and are critical for the functioning of excitable cells. However, it was not totally clear how the ligand bound to the receptor. Homology modelling of the extracellular domain of the 5-HT3 receptor and docking simulations of 5-HT were performed, and our theoretical predictions of the binding site and binding mode of 5-HT and granisetron, an antagonist, agreed well with site-directed mutagenesis data [21-22]. Further work has been carried out to locate the unbinding pathway [23], and to define the hydrogen bonds and cation-π interactions between the ligand and the receptor using molecular dynamics simulations [24].
I have also started work on another ligand-gated ion channel, the GABAA receptor, by producing similairity models of this protein [25].
Membrane effects
My recent research has addressed the role of the membrane on the effects of certain ligands, for example, the general anaesthetics. The cell membrane has long been thought to be involved in general anaesthetic action, but it has never been clear how. The effect of general anaesthetics is pressure-dependent; increased ambient pressure reverses their effects. Electron-spin spectroscopy experiments showed that halothane altered the order parameter of a phosphatidylcholine membrane, but this change was reversed when the pressure of the system was raised to 274 atm. However, the molecular mechanism of this pressure-induced reversal was unclear. Using simulations, I show that at 200 atm - 400 atm, halothane tends to aggregate inside the membrane [26]. Previous experiments suggest that halothane acts by binding to a site in the transmembrane domain of the GABAA receptor; this putative binding site is only large enough to accommodate one halothane molecule. Pressure reversal occurs when halothane aggregates, consequently reducing the amount of monomeric halothane available to bind to the binding site. Thus the cell membrane acts as a conduit for general anaesthetics, and its effect on anaesthetic action may be indirect. In drawing these conclusions, I have defined the role of membranes in general anaesthetic action, and solved a pharmacological problem of over 50 years' standing.I am currently seeking to verify these findings by experiment.