Membrane Curvature: How BAR Domains Bend Bilayers Dispatch

There have been a number of reports that adding certain proteins to phospholipid bilayer membranes results in their tubulation or vesiculation, but it is not clear how these proteins work. One hypothesis is that hydrophobic groups on a protein insert into one leaflet of a bilayer, increasing either the spontaneous curvature or the lateral area of this monolayer [1,2]. An alternative hypothesis is that when certain membrane-binding proteins, such as the vesicle coat proteins clathrin or COPI and COPII, polymerize into a coat or a cage, the oligomerization bends the membrane into an endocytic vesicle [2,3]. Peter et al. [4] recently reported the X-ray crystallographic structure of the BAR domain of Drosophila amphiphysin and proposed a third mechanism for protein-induced membrane curvature: electrostatic attraction between the protein and lipids plasters the membrane to the concave surface of the BAR domain, apparently in the absence of any substantial hydrophobic insertion into the membrane. Elegant work from the De Camilli [5] laboratory previously established that addition of amphiphysin forms tubules from larger liposomes in vitro, and that overexpression of amphiphysin 2 in cells leads to internal membrane tubulation [6]. The newly recognized BAR (Bin/Amphiphysin/Rvs) domain is present in a number of proteins — amphiphysins, endophilin, arfaptins, nadrins, beta-centaurins and oligophrenins — some of which are critical for the recycling of synaptic vesicles and T-tubule formation in muscle [4]. The structure of the BAR domain, an elongated ‘banana-shaped’ dimer, has sufficient rigidity, curvature, and charge to immediately suggest its function. Figure 1A, taken from Peter et al. [4], shows the structure of the BAR domain. The authors describe each monomer as “a coiled-coil of three long kinked αhelices, forming a six-helix bundle around the dimer interface” and note “the curvature of the dimer is partly due to the way the monomers intersect and partly due to the kinks in helices 2 and 3”. The hydrophobic residues are largely placed at the interface between the monomers, raising the question: how does BAR bind to bilayers? The authors propose the answer is simple electrostatics. Figure 1B illustrates the electrostatic profile predicted by application of the Poisson–Boltzmann equation to a model of the BAR domain: the blue areas indicate regions of positive electrostatic potential. Note the flexible loop between helices 2 and 3 found at the extreme ends of the dimer is strongly basic, and that the concave surface of the dimer has several patches of positively charged residues. Thus Peter et al. [4] suggest that “this is the surface that interacts with phospholipid membranes”; it would fit a curved membrane with an outer radius of 11 nm. Many studies have shown that clusters of basic residues on proteins can help anchor proteins to the acidic lipids on the cytoplasmic leaflet of the plasma membrane, which typically contains 15–30% monovalent acidic phospholipids, mainly phosphatidylserine, and about 1% multivalent phosphatidyinositol bisphosphate (PIP2). These negatively charged lipids produce a negative electrostatic potential that attracts counterions, such as K+, from the cytoplasm. As first recognized by Helmholtz in the 19th century, this produces a ‘diffuse double layer’ or ion atmosphere which extends a few Debye lengths — a few nanometers under physiological conditions — away from the surface [7,8]. The negative surface potential, which is about –30 mV for a membrane with 20% phosphatidylserine [8], also attracts clusters of basic residues on proteins. This effect is illustrated by the carboxy-terminal basic cluster on K-Ras, consisting of seven contiguous lysine residues, or the amino-terminal basic cluster on Src, with net charge +5, which help anchor these proteins to the plasma membrane by simple electrostatic attraction to the acidic lipids. These basic clusters do not, however, provide quite enough electrostatic energy to anchor the protein tightly to the membrane; K-Ras4B requires an adjacent, covalently attached farnesyl group and Src an adjacent, covalently attached myristoyl group, which act as hydrophobic anchors in concert with the electrostatic interactions [9]. The electrostatic binding energies of these basic clusters can be predicted theoretically by applying the non-linear Poisson–Boltzmann equation to atomic level models of bilayers and proteins/peptides [9]. As predicted theoretically and shown experimentally by a number of different laboratories using simple model peptides and phospholipid vesicles, the binding energy increases linearly with the mole fraction — surface density — of acidic lipids in the bilayer and the number of basic residues in the cluster. An increase in salt concentration screens the charges and decreases the binding. The proposal by Peter et al. [4] that BAR domains use simple electrostatics to bind membranes thus has both biological precedents — for example, Src, KRas4B, MARCKS, HIV-Gag and AKAP79 — and good theoretical and experimental support from model Current Biology, Vol. 14, R250–R252, March 23, 2004, ©2004 Elsevier Ltd. All rights reserved. DOI 10.1016/j.cub.2004.02.060