Calculating the Effect of Membrane Thickness on the Lifetime of the Gramicidin A Channel: A Landmark

Main Text

In this issue of the Biophysical Journal, Sun et al. (1) report the results of a remarkable molecular dynamics (MD) study about the effect of membrane thickness on the dimerization and dissociation of the gramicidin A (gA) dimer channel. Although MD simulations of biomolecular systems of fantastical complexity are becoming almost commonplace, this work deserves our attention.

Despite the progress in MD simulations, a quantitative characterization of the interactions between membrane proteins and the surrounding lipid bilayer—a fundamental problem in biology and biophysics—has remained excessively challenging. This is especially true for nonspecific effects mediated by the overall bulk mechanical properties of the lipid membrane bilayer, which emerge from a statistical average of a multitude of small structural fluctuations spanning a broad range of length scales and timescales. Calculating the effects of bulk mechanical properties of a membrane on the function of an embedded protein had long been beyond the reach of available resources because it requires very long simulations of very large systems. In the present study, the authors were able to show, by relying on extensive umbrella sampling potential of mean force MD calculations, that the lifetime of the gA dimer channel is directly affected by the thickness of the membrane. The results are in quantitative agreement with classic experimental observations, opening the door to a rational and quantitative understanding of the protein-membrane interactions for the first time.

In a way, it is fitting (and somewhat ironic) that this impressive study involves gA. Since the initial experimental studies going back to the 1970s (2), the channel formed by this small linear pentadecapeptide molecule has played a central role as a prototypical model for fundamental studies of ion permeation and fundamental membrane biophysics.

Early on, it was noted that the lifetime of the gA channel was sensitive to the lipid bilayer thickness (2, 3). Intuitively, this could be rationalized in the light of the head-to-head dimer membrane-spanning structure proposed by Urry (4); the dimer channel is destabilized as the membrane gets thicker and “pulls” on the two monomers. Arguments based on continuum mechanics were used to further support this concept, but it is only now that we are able to fully elucidate the mechanical protein-membrane coupling on the basis of detailed atomic MD simulations.

To put this within proper historical context, it is worth recalling that the first MD simulation of a protein was reported in 1977 (5); a simple globular protein in vacuum without explicit solvent molecules was simulated for less than 5 ps. The first MD simulations of the gA channel, carried out a little later, were based on a model that included a few water molecules but no explicit membrane bilayer (6). In fact, membranes were scary to simulators; for a long time, there were concerns that the simple nonpolarizable force field used for MD simulations would just be unable to account for the balance of hydrophobic and hydrophilic forces sufficiently accurately to maintain the integrity of a lipid bilayer. Moreover, the treatment of long-range electrostatics and pressure were still very rudimentary and known to be inadequate. These concerns were openly discussed in 1992 at an important workshop on membrane simulations at the Centre Européen de Calcul Atomique et Moléculaire in Orsay (France). The idea of being able to compute membrane mechanical features from MD simulations, such as the effect of thickness on the lifetime of the gA channel for example, was even brought up during these discussions, but only as a dream that, one day, might be realized. Simulations of actual atomic models of phospholipid membrane bilayers with explicit water molecules came of age soon after in 1992 (7), followed by the first MD of gA embedded in an explicit membrane in 1994 (8). Since then, there has been a steady progress in simulation algorithms, numerical methods, and computer software and hardware. The all-atom force field, which was used by Sun et al. (1), was first introduced in the 1990s and was refined for more than 20 years (9, 10, 11, 12, 13).

In retrospect, it took almost 50 years to bridge the classic experimental measurements showing that the lifetime of the gA channel depends on membrane thickness with the MD simulations of Sun et al. (1), which account quantitatively for this observation. In its historical context, this landmark study offers a sobering lesson regarding the pace of fundamental research and the perseverance that is required to move forward, but it also gives us the chance to acknowledge the great advances that have been accomplished in computational biophysics during this period.

Editor: D. Peter Tieleman.