Simulations long enough to...put you to sleep

For something as widely used for as long as general anesthetics (GAs), one would think that their molecular mechanism of action would have been fairly understood. Far from it.

From Linus Pauling's theory of gases like xenon acting at high concentrations by forming clathrates to more recent theories of GA action on lipids and now on proteins, tantalizing clues have emerged, but speculation remains rife.

In a recent Acc. Chem. Res. review, a group of researchers explains some recent studies on GA action. Now there's a field that to me seems primed for computational studies. This is for two reasons.

Firstly, experimental information on GAs is hard to come by. Consider their chemical features; halogenated, apolar small molecules lacking polar hydrogen bonding and other interactions, binding to their targets with low affinity (it's interesting that halogenation seems to be a key criterion for GA action). In addition most GAs do not bind to a highly specific active site but instead influence protein action indirectly. Such features make any kind of NMR or x-ray structure determination an enormous challenge.

Secondly, molecular dynamics simulations (MDS) have come of age. With recent programs augmented by tremendous gains in hardware and software, microsecond to millisecond simulations have gradually become a reality. This particular field seems to provide a classic and worthy challenge for MDS, since GAs seem to interact indirectly and subtly with proteins by influencing their local and global dynamics rather than binding to well-defined active pockets. Such dynamic perturbations would fail to be captured during the pico to nanosecond timescales typically sampled by MDS. For instance, the most prevalent belief for GAs right now is that they interact with ligand-gated ion channels like the GABA and NMDA receptor and with potassium ion channels. One hypothesis for the mode of action of halothane is that it binds to the open conformation of a potassium ion channel. The channel stays open for milliseconds, thus thwarting experimental study. However, a millisecond transition provides a robust and respectable challenge for long time-scale MD simulations.

At the same time, caveats abound in the field. For instance it's easy to infer that a GA molecule binds to a certain site and obstructs the motion of a tyrosine residue, thus providing support to fluorescence quenching and other studies. But the results of such studies as well as the all-important site-directed mutagenesis studies are notoriously hard to interpret; indirect influences on protein motion may be construed as direct binding to particular sites. Plus, it seems to me that one can read too much into the mere, rather obvious observation that a molecule binding to a protein site inhibits the motion of some residues; whether that observation translates into a realistic phenomenon may be much harder to glean.

So yes, it seems that GA action provides a fertile field for computer simulation. Long MD simulations generally seem to me to be a solution looking for problems; after all most interesting molecular interactions in the body take place on the order of micro to milliseconds. There is a huge number of important problems waiting to be tackled with such tools. However, interpretations of the results will always have to be guided by the sure hand of experiment, with the always important caveat that when it comes to interpretation, one computational study and one experiment can have several offspring.

Vemparala, S., Domene, C., & Klein, M. (2010). Computational Studies on the Interactions of Inhalational Anesthetics with Proteins Accounts of Chemical Research, 43 (1), 103-110 DOI: 10.1021/ar900149j