Water under the BAR

Abstract

Many cellular processes require the generation of highly curved regions of cell membranes by interfacial membrane proteins. A number of such proteins are now known, and several mechanisms of curvature generation have been suggested, but so far a quantitative understanding of the importance of the various potential mechanisms remains elusive. Following previous theoretical work, we consider the electrostatic attraction that underlies the scaffold mechanism of membrane bending in the context of the N-BAR domain of amphiphysin. Analysis of atomistic molecular dynamics simulations reveals considerable water between the membrane and the positively charged concave face of the BAR, even when it is tightly bound to highly curved membranes. This results in significant screening of electrostatic interactions, suggesting that electrostatic attraction is not the main driving force behind curvature sensing, supporting recent experimental work. These results also emphasize the need for care when building coarse-grained models of protein-membrane interactions. These results are emphasized by simulations of oligomerized amphiphysin N-BARs at the atomistic and coarse-grained level. In the coarse-grained simulations, we find a strong dependence of the induced curvature on the dielectric screening.

Figure 1

The configuration of the oligomer simulation. (A) Central simulation cell from the top and a periodic image on either side. (B) Close-up of the area highlighted by the red box in panel A, showing lateral contacts between neighboring rows of N-BARs. (Dotted lines) Favorable electrostatic interactions between atoms that are within 5 Å of each other. The view is from the membrane side and a piece of one helix of the scaffold is removed for clarity. (C) Final snapshot after 120 ns, which curved to a radius of 58 nm.

Figure 2

(A) Number of water molecules in between the concave face of the BAR and membrane for three independent simulations of amphiphysin N-BAR domains bound to the membrane by embedded N-terminal helices (dashed lines). The running averages (solid lines) with the standard error computed by dividing the trajectory into 10 equal length blocks (indicated by error bar). (B, side view and C, bottom view) Occupancy of water molecules under the arch of the BAR, averaged over the NBR1 trajectory from 30 ns onward (tight binding). The 30% occupancy isosurface (translucent blue shading) shows regions where water molecules are found at least 30% of the time. Surface of the N-BAR (light shading); conserved positive residues are under the arch (red). All molecular renderings were made with VMD (35).

Figure 3

The number of water molecules in between the concave face of the BAR and membrane for the three proteins in the oligomer simulation. The color of the time traces matches the color of the proteins in Fig. 1; the meaning of the dashed and solid lines and the error bars are explained in Fig. 2's legend.

Figure 4

(A and B) The effective dielectric constant for all six datasets is computed by fitting a potential of the form q₁q₂/ɛɛ₀r (red line) to the long-range part of the approximate potential of mean force (PMF) between negatively charged oxygens (with charge q₁) of the phosphatidylserine headgroups and positively charged hydrogens (with charge q₂) of Lys and Arg side chains under the concave face of the BAR. Panel (A) corresponds to the single N-BAR data in Fig. 2. Panel (B) corresponds to the oligomer data in Fig. 3.The free parameter is the dielectric constant, ɛ; ɛ₀ is the permittivity of the vacuum. The PMF is computed by Boltzmann inversion of the radial distribution function, averaged over the conserved positive charges under the arch of the protein. The fit is performed over the entire range of the potential, following Hess et al. (32). For comparison, the shape of the bare Coulomb (ɛ = 1) potential is shown (green).

Figure 5

(Upper-left panel) Top view of the starting configuration of the six-amphiphysin N-BAR SBCG simulation (counterions omitted for clarity). (Upper-right panel) Side view of the starting configuration of the SBCG simulations, showing counterions. (Lower-left panel) Final configuration of the SBCG simulation with the relative dielectric set to 1. (Lower-right panel) Final configuration of the SBCG simulation with the relative dielectric set to 14.

Figure 6

Position of the membrane midplane (symbols) averaged over the last 1000 configurations of each of the three SBCG simulations. Also shown (solid lines) are the fits of each to the arc of a circle, with the radius of the fit written below each data set.