Molecular Dynamics Study of Realistic Membrane Ion Channels


Principal Investigator

Toby Allen

Department of Chemistry,

The Faculties

The mechanisms behind ion transport across vmembrane channels appear to be governed by two
main factors: intrapore water structure and protein wall charge group configurations. Molecular Dynamics (MD) studies exist which examine the effect of confining boundaries on pure water orientations. However, in the narrow transmembrane segments of channels one must not only consider the effects of protein walls on water structure, but also on the hydration of the ion attempting to pass that segment. The dielectric behaviour of the water in this situation is not well understood and requires considerable investigation.
Many studies of ion channels use continuum methods where a mean dielectric constant and mean ion diffusion constants are assumed (usually the artificially high bulk values). It is commonly accepted that these assumptions are not realistic, but that within their techniques they provide the only possible option. MD on the other hand requires no such assumptions since the solvent water is treated explicitly.


w50 - VPP, PC



What are the results to date and the future of the work?

In our studies we have performed MD simulations of an ion and water within realistic ion channel geometries (cylindrical and catenary shaped protein walls). We have examined the effects of the channel walls on hydration and have used these results to explain the energy barriers experienced by the sodium and potassium ions. In continuum electrostatics calculations (where the water in the channels is treated as a continuum) we have measured the potential energy barriers of ions entering the same channels for varying water dielectric constants. These barriers consist only of the image repulsion component of the potential because the barrier due to hydration losses can not emerge without treating the water in the simulation explicitly. By comparing these barriers to those found in the MD simulations we were able to estimate the hydration barrier component. This hydration barrier has been used in Brownian Dynamics (BD) to reproduce physiological channel conductances.

Our hydration and energy barrier calculations have therefore been very successful. Another feature of our calculations which has never before been attempted, and which has proved quite successful, is the diffusion calculations. Water and ion diffusion have been determined and explained in terms of the water structure within the channels. Ion diffusion constants require large computational time because of the fact that there is only one atom of that species and

- Appendix A



more statistical data is required. Our results showed that the ion diffusion is around 15% of bulk for the narrowest of channels (3 Angstrom radius) and gradually approaches the bulk diffusion as the channel radius increases. In future work these diffusion constants should be employed in BD simulations replacing the constant bulk diffusion constants used.

Our simple channel model has been used to reproduce some findings of existing calculations which employ complex channel wall configurations of polar sidegroups. This is very important as it says that a simple structureless hydrophobic boundary of appropriate dimensions can capture most of the salient features of the real biological ion channel.

We intend to further this argument in future by continuing to reproduce properties of the ion channels using MD and BD with our simple models.

What computational techniques are used?

All MD simulations are carried out with the CHARMM v25b2 package on the SGI Power Challenge (as well as our SGI O2 R5000). The standard Verlet algorithm with bonds constrained by the SHAKE procedure is used. We use simple molecular models and make use of the most efficient list updating and cutoffs so as to improve computational efficiency. Despite this we note that because of the nature of MD simulations where the solvent is treated explicitly, the system sizes are large and even moderately small simulation times (100200ps) require 1224 hours per run. Since we experiment with numerous channels and ions we require a large amount of CPU time to complete a single study.

BD simulations employ our findings from MD so as to produce channel conductances which compare to experiment. These are performed on the VPP with our own highly vectorized package. The BD algorithms have undergone considerable upgrades in the last year to improve vectorization and overall speed. This is the same program used by S-H.Chung (Department of Chemistry) in his studies of permeation of ions through membrane channels.

Appendix A -