Peripheral membrane proteins (PMPs) bind temporarily to the surface of biological membranes. They also exist in a soluble form and their tertiary structure is often known. Yet, their membrane-bound form and their interfacial-binding site with membrane lipids remain difficult to observe directly. Their binding and unbinding mechanism, the conformational changes of the PMPs and their influence on the membrane structure are notoriously challenging to study experimentally. Molecular dynamics simulations are particularly useful to fill some knowledge-gaps and provide hypothesis that can be experimentally challenged to further our understanding of PMP-membrane recognition. Because of the timescales of PMP-membrane binding events and the computational costs associated with molecular dynamics simulations, membrane models at different levels of resolution are used and often combined in multiscale simulation strategies. We here review membrane models belonging to three classes: atomistic, coarse-grained and implicit. Differences between models are rooted in the underlying theories and the reference data they are parameterized against. The choice of membrane model should therefore not only be guided by its computational efficiency. The range of applications of each model is discussed and illustrated using examples from the literature.

Spider venom GDPD-like phospholipases D (SicTox) have been identified to be one of the major toxins in recluse spider venom. They are divided into two major clades: the α clade and the β clade. Most α clade toxins present high activity against lipids with choline head groups such as sphingomyelin, while activities in β clade toxins vary and include preference for substrates containing ethanolamine headgroups (Sicarius terrosus, St_βIB1). A structural comparison of available structures of phospholipases D (PLDs) reveals a conserved aromatic cage in the α clade. To test the potential influence of the aromatic cage on membrane-lipid specificity we performed molecular-dynamics (MD) simulations of the binding of several PLDs onto lipid bilayers containing choline headgroups; two SicTox from the α clade, Loxosceles intermedia αIA1 (Li_αIA) and Loxosceles laeta αIII1 (Ll_αIII1), and one from the β clade, St_βIB1. The simulation results reveal that the aromatic cage captures a choline-headgroup and suggest that the cage plays a major role in lipid specificity. We also simulated an engineered St_βIB1, where we introduced the aromatic cage, and this led to binding with choline-containing lipids. Moreover, a multiple sequence alignment revealed the conservation of the aromatic cage among the α clade PLDs. Here, we confirmed that the i-face of α and β clade PLDs is involved in their binding to choline and ethanolamine-containing bilayers, respectively. Furthermore, our results suggest a major role in choline lipid recognition of the aromatic cage of the α clade PLDs. The MD simulation results are supported by in vitro liposome binding assay experiments.

Peripheral membrane proteins (PMPs) bind temporarily to cellular membranes and play important roles in signaling, lipid metabolism, and membrane trafficking. Obtaining accurate membrane-PMP affinities using experimental techniques is more challenging than for protein–ligand affinities in an aqueous solution. At the theoretical level, calculation of the standard protein–membrane binding free energy using molecular dynamics simulations remains a daunting challenge owing to the size of the biological objects at play, the slow lipid diffusion, and the large variation in configurational entropy that accompanies the binding process. To overcome these challenges, we used a computational framework relying on a series of potential-of-mean-force (PMF) calculations including a set of geometrical restraints on collective variables. This methodology allowed us to determine the standard binding free energy of a PMP to a phospholipid bilayer using an all-atom force field. Bacillus thuringiensis phosphatidylinositol-specific phospholipase C (BtPI-PLC) was chosen due to its importance as a virulence factor and owing to the host of experimental affinity data available. We computed a standard binding free energy of −8.2 ± 1.4 kcal/mol in reasonable agreement with the reported experimental values (−6.6 ± 0.2 kcal/mol). In light of the 2.3-μs separation PMF calculation, we investigated the mechanism whereby BtPI-PLC disengages from interactions with the lipid bilayer during separation. We describe how a short amphipathic helix engages in transitory interactions to ease the passage of its hydrophobes through the interfacial region upon desorption from the bilayer.

Peripheral membrane proteins (PMPs) bind temporarily to cellular membranes and play important roles in signalling, lipid metabolism and membrane trafficking. Obtaining accurate membrane-PMP affinities using experimental techniques is more challenging than for protein-ligand affinities in aqueous solution. At the theoretical level, calculation of standard protein-membrane binding free energy using molecular dynamics simulations remains a daunting challenge owing to the size of the biological objects at play, the slow lipid diffusion and the large variation in configurational entropy that accompanies the binding process. To overcome these challenges, we used a computational framework relying on a series of potential-of-mean-force (PMF) calculations including a set of geometrical restraints on collective variables. This methodology allowed us to determine the standard binding free energy of a PMP to a phospholipid bilayer using an all-atom force field. Bacillus thuringiensis phosphatidylinositol-specific phospholipase C (BtPI-PLC) was chosen due to its importance as a virulence factor and owing to the host of experimental affinity data available. We computed a standard binding free energy of -8.2±1.4 kcal/mol in reasonable agreement with the reported experimental values (-6.6±0.2 kcal/mol). In light of the 2.3-μs separation PMF calculation, we investigated the mechanism whereby BtPI-PLC disengages from interactions with the lipid bilayer during separation. We describe how a short amphipathic helix engages in transitory interactions to ease the passage of its hydrophobes through the interfacial region upon desorption from the bilayer.

Phospholipase D (PLD) enzymes are one of the major toxins in the recluse spider (genus Loxosceles) venom. Loxosceles PLD enzymes are classified in either the a clade or the b clade, and some correlation exists between a or b clade membership and high or low catalytic activity against sphingomyelin (SM), respectively. Lajoie et al., recently showed that a b clade enzyme from Sicarius terrosus (St_bID1) had a strong preference for substrates containing ethanolamine headgroups, and suggested that this substrate specificity originated from the interactions between the PLD interfacial face and lipids. To further understand the substrate specificity of these enzymes, we performed simulations of two PLDs from the a clade, Loxosceles intermedia aIA1 (Li_aIA1) and Loxosceles laeta aIII1 (Ll_aIII1), and one from the b clade, St_bIB1i. Simulations were performed in the presence of two types of lipid bilayers, cholinecontaining and ethanolamine-containing bilayer. We observed that the two a clade PLDs bound to bilayers with choline-containing lipids (PC or SM) using the catalytic loop. By contrast, St_ bIB1i, did not bind to those bilayers. Analyses of the trajectories also reveal the importance, for the two a clade PLDs, of three aromatic residues located on the catalytic loop and forming a p-cage interacting with a choline group. A multiple sequence alignment of 18 PLDs of both clades reveal that the aromatic cage is conserved in a clade PLDs and absent from at least one major subgroup of the b clade enzymes. We suggest that this cage controls the affinity for choline headgroups, which is in agreement with earlier reports of PC lipid headgroups interacting with aromatic amino acids.