Directed Supramolecular Organization of N-BAR Proteins through Regulation of H0 Membrane Immersion Depth

Abstract

Many membrane remodeling events rely on the ability of curvature-generating N-BAR membrane proteins to organize into distinctive supramolecular configurations. Experiments have revealed a conformational switch in N-BAR proteins resulting in vesicular or tubular membrane shapes, with shallow membrane immersion of the H0 amphipathic helices of N-BAR proteins on vesicles but deep H0 immersion on tubes. We develop here a minimal elastic model of the local thinning of the lipid bilayer resulting from H0 immersion. Our model predicts that the observed conformational switch in N-BAR proteins produces a corresponding switch in the bilayer-mediated N-BAR interactions due to the H0 helices. In agreement with experiments, we find that bilayer-mediated H0 interactions oppose N-BAR multimerization for the shallow H0 membrane immersion depths measured on vesicles, but promote self-assembly of supramolecular N-BAR chains for the increased H0 membrane immersion depths measured on tubes. Finally, we consider the possibility that bilayer-mediated H0 interactions might contribute to the concerted structural reorganization of N-BAR proteins suggested by experiments. Our results indicate that the membrane immersion depth of amphipathic protein helices may provide a general molecular control parameter for membrane organization.

Figure 1

Illustration of membrane curvature generation by N-BAR proteins. (A) Molecular structure of the N-BAR protein endophilin (PDB ID code 2C08) viewed perpendicular (upper left panel) and parallel (upper right panel) to the plane of the membrane. Endophilin employs the scaffolding (lower left panel) and wedging (lower right panel) mechanisms for membrane curvature generation through its BAR domain and through its N-terminal H0 helices and insert region, respectively. The molecular structure of amphiphysin is highly homologous to that of endophilin but lacks the insert region^,,. (B) A conformational switch controls generation of isotropic (vesicular) and anisotropic (tubular) membrane curvatures by N-BAR proteins^–, with shallow wedging and little scaffolding on vesicles (left panel) but deep wedging and pronounced scaffolding on tubes (right panel). In the membrane cross sections, green shading indicates membrane regions in which deep membrane immersion of amphipathic helices^, induces compression of the lipid bilayer hydrophobic thickness^–.

Figure 2

Schematic of bilayer-mediated interactions between protein wedges. (A) Amphipathic helices (viewed parallel to the helix axes and indicated in blue) deform the shape h⁺ and hydrophobic thickness u⁺ +a of the upper lipid bilayer leaflet, and may indirectly perturb h⁻ and u⁻ in the lower leaflet via the coupling between upper and lower leaflets. We set here u⁻ = 0 (see main text). For simplicity, we assume in this schematic that the H0 helices are in the face-on orientation, and that the H0-induced membrane deformations are only a function of the spatial coordinate r measured perpendicular to the helix axes. Equation (3) predicts that, for large enough helix immersion depths, two protein wedges can be attracted to each other by bilayer-mediated wedge interactions (black arrows). (B) Our elastic model of bilayer-mediated wedge interactions predicts that, for large enough helix immersion depths, the distance d separating the axes of two neighboring amphipathic helices can take an optimal value set by the key lipid and protein properties captured by equation (3).

Figure 3

Bilayer-mediated interactions between H0 helices. Interaction potential G_int between two parallel H0 helices of neighboring N-BAR proteins versus the immersion depth of the two helices, U, and the distance separating the two H0 axes, d, for τ = 0. We obtained G_int from the exact analytic solution minimizing the energy cost of H0-induced lipid bilayer deformations in equation (3) (see the Methods section), which neglects boundary effects due to the H0 tips. G_int < 0 corresponds to favorable bilayer-mediated interactions between H0 helices, and G_int > 0 corresponds to unfavorable interactions. EPR experiments suggest U ≈ 0 nm and U ≈ −0.9 nm for vesicle- and tube-bound N-BAR proteins^, (see schematics in insets).

Figure 4

Multimerization of N-BAR proteins through bilayer-mediated interactions. Energetically favorable N-BAR arrangements obtained, for U ≈ −0.9 nm^, and τ = 0, from simulated annealing Monte Carlo simulations using (A) the N-BAR pair potential implied by the H0-induced leaflet thickness deformations in equation (3) with h⁺ = 0 and (B) a total N-BAR pair potential that is the sum of the potential used in (A) and a BAR pair potential based on previous calculations (see the Methods section). The left insets in (A) and (B) show the random initial N-BAR configurations employed for (A) and (B). We used the N-BAR shape shown in the right inset in (A), which we modeled after the N-BAR structures described previously, with hardcore steric constraints (see Supplementary Information Sec. S6). We employed periodic boundary conditions, with the N-BAR proteins in (A) forming a single chain. Scale bars, 10 nm.

Figure 5

Effect of membrane tension on bilayer-mediated interactions between H0 helices. Interaction potential G_int between two parallel H0 helices of neighboring N-BAR proteins obtained from equation (3) as in Fig. 3 versus the distance separating the two H0 axes, d, for the deep (U ≈ −0.9 nm) and shallow (U ≈ 0 nm) immersion states of the H0 helices of N-BAR proteins^, at the indicated values of the membrane tension τ.

Figure 6

Concerted structural reorganization of N-BAR proteins through bilayer-mediated interactions between H0 helices. Probability of a pair of N-BAR proteins to be collectively in the observed conformational state with deep immersion of the H0 helices^,, P_d, obtained from equation (8) versus concentration of membrane-bound N-BAR proteins, c, and the energy difference between the observed N-BAR conformational states with deep and shallow immersion of the H0 helices^, in the absence of any interactions between N-BAR proteins, ε_d. We analytically calculated the H0-induced N-BAR interactions following similar steps as for Fig. 3 (see the Methods section) and set τ = 0. The dashed black curve corresponds to P_d = 1/2 and, for reference, the solid white lines delineate the approximate range of N-BAR concentrations found to tubulate large vesicles in vitro.