Annual Scientific Meeting, Belfast; November 2000.

Physico-chemical models for the action of intravenous anaesthetic agents are few; they include hydrophobic correlations, pressure-reversal studies, as well as their actions on ligand-gated, and voltage-gated channels and other receptors at clinically relevant drug concentrations. We have previously reported models that relate the hypnotic properties of anaesthetic drugs to their three-dimensional molecular properties such as shape [1] and electrostatic potential (a measure of the distribution of charge around the molecule) [2].

For the volatile agents, Sewell and Halsey [3] generated a model based on molecular shape and electrostatic potential that could predict with an accuracy of 97.8% whether a volatile compound was an anaesthetic or not; while a model based on shape alone could accurately predict the potencies of halogenated ether (r² = 0.98) and ethane (r² = 0.94) general anaesthetics [1].

Our preliminary data for intravenous agents demonstrated the importance of electrostatic potential in determining in vivo potency, explaining 87% of the variance in observed potencies [2], with a true predictive power of 83% under conditions of cross-validation. This represented substantial improvement compared with a 'conventional' activity model based on the octanol/ water partition coefficients of the compounds, which explained only 66% of the variance in observed activity (55% under cross validation). A model based on shape similarity alone explained 81% of the variance of observed potencies of the structurally diverse group of intravenous anaesthetics (78% under cross validation).

However none of these results took into account the effects of chirality of the various agents [4, 5], and hence the present study is an extension of our earlier work to develop an improved activity model for the intravenous agents that can address this shortcoming.

We calculated the free plasma drug concentrations data that abolished the response to a noxious stimulus for 11 hypnotic agents (racemic thiopentone, pentobarbitone, thiamylal, and methohexitone; minaxolone, alphaxalone, eltanolone; propofol; R & S ketamine and R etomidate) from published studies in the literature. Where possible, these were taken from studies where no other adjuvant drugs were administered up to the time of surgical incision. Midazolam was excluded from the present analyses as we know of no data describing its use as a sole anaesthetic.

Computer-based models of the agents were then constructed, and low energy conformers identified using a random search procedure (implemented in SYBYL 6.6, Tripos Inc, USA). The resultant conformers were geometry optimised at the semi-empirical level using the PM3 Hamiltonian, and partial charges assigned using the NAO-PC method (VAMP 7, Oxford Molecular Ltd, UK). A reiterative procedure was used to superimpose the conformers of the structurally diverse agents, so as to maximize the similarities in their molecular shapes or electrostatic potentials calculated as the Carbo and Hodgkin similarity indices respectively (ASP 3.2, Oxford Molecular Ltd, UK). The optimal alignments were identified by formulating activity models that correlated hypnotic activity with the evaluated molecular similarity, and retaining the model with the highest correlation.

The bench mark model based on octanol-water partition coefficients (hydrophobic correlation) explained only 62% of the variance in the observed activities of the compounds, failed to predict the potency order for the enantiomers of ketamine and had a predictive power of only 59%.

In contrast, the best model obtained was based on the similarities of all the agents to the electrostatic potentials of S-thiopental and R-ketamine, and explained 89% of the variance in the observed activities of the hypnotic agents. Cross validation (in which 20% of the compounds were excluded at each stage) showed that the model had a predictive power of 84%. The model correctly predicts the potency order for the enantiomers of the chiral agent ketamine; but identifies alphaxalone (the main active component of the steroid Althesin) as a clear outlier, being less potent than would be predicted from its molecular similarity.

The exclusion of the outlier alphaxalone from the molecular similarity analysis resulted in an improved model (r²=0.97), which was based on the similarities of the remaining 10 hypnotic agents to the geometric shapes of S-pentobarbital and RS-methohexitone. Again cross validation revealed a high predictive power of 96%.

In this study, we have used computer modelling applications to assess whether 'molecular fields' (including molecular shape and electrostatic potentials) can offer some commonality between the eleven diverse chemical entities that can induce and maintain anaesthesia.

Our results demonstrate that a single activity model can be formulated for chiral and non-chiral intravenous agents using a molecular similarity approach. Furthermore the procedure identifies an optimal alignment for the structurally-diverse general anaesthetics that could enable the rational design of hypnotic agents.

It is difficult to offer firm reasons for alphaxalone being an outlier. Alphaxalone is the major hypnotic steroid in Althesin, with the alphadolone acetate being present to increase the former's solubility. Our studies to determine the 'anaesthetic' concentration measured plasma alphaxalone concentrations; we do not know the relative in vivo potencies of alphaxalone and alphadolone acetate during clinical anaesthesia. Furthermore, the only available reported estimate for its plasma protein binding is low (40%), and not in keeping with other pregnane steroids.

What does this study offer towards the determination of the site of anaesthetic action? It is important to distinguish between a common molecular basis for the mechanism of general anaesthetic action and the implication of a common site of action. Target-orientated studies suggest that the different classes of intravenous anaesthetic agents probably act at different binding sites on single or multiple classes of receptors [6], while our 'molecular-orientated' approach provides information about the common molecular features of a diverse group of molecules that determine their in vivo anaesthetic potency.

References:

1. Sewell JC, Halsey MJ. Europ J Med Chem 1997, 32: 731-737

2. Sewell JC, Sear JW, Halsey MJ. Br J Anaesth 1999, 82: 470P

3. Sewell JC, Halsey MJ. Toxicol Letters 1998, 100-101; 359-364

4. Tomlin SL, Jenkins A, Lieb WR, Franks NP. Anesthesiology 1998, 88: 708-717

5. Tomlin SL, Jenkins A, Lieb WR, Franks NP. Anesthesiology 1999, 90: 1714-1722

6. Krasowski MD, Harrison NL. Cell Mol Life Sci 1999; 55: 1278-1303