Partitioning of Benzoic Acid into 1,2-Dimyristoyl-sn-glycero-3-phosphocholine and Blood–Brain Barrier Mimetic Bilayers

Abstract

Using an all-atom explicit water model and replica exchange umbrella sampling simulations, we investigated the molecular mechanisms of benzoic acid partitioning into two model lipid bilayers. The first was formed of 1,2-dimyristoyl-sn-glycero-3-phosphocholine (DMPC) lipids, whereas the second was composed of an equimolar mixture of DMPC, 1,2dimyristoyl-sn-glycero-3-phosphoethanolamine, palmitoylsphingomyelin, and cholesterol to constitute a blood−brain barrier (BBB) mimetic bilayer. Comparative analysis of benzoic acid partitioning into the two bilayers has revealed qualitative similarities. Partitioning into the DMPC and BBB bilayers is thermodynamically favorable although insertion into the former lowers the free energy of benzoic acid by approximately an additional 1 kcal mol −1 . The partitioning energetics for the two bilayers is also largely similar based on the balance of benzoic acid interactions with apolar fatty acid tails, polar lipid headgroups, and water. In both bilayers, benzoic acid retains a considerable number of residual water molecules until reaching the bilayer midplane where it experiences nearly complete dehydration. Upon insertion into the bilayers, benzoic acid undergoes several rotations primarily determined by the interactions with the lipid headgroups. Nonetheless, in addition to the depth of the free energy minimum, the BBB bilayer differs from the DMPC counterpart by a much deeper location of the free energy minimum and the appearance of a high free energy barrier and positioning of benzoic acid near the midplane. Furthermore, DMPC and BBB bilayers exhibit different structural responses to benzoic acid insertion. Taken together, the BBB mimetic bilayer is preferable for an accurate description of benzoic acid partitioning.

Partitioning of Aβ Peptide Fragments into Blood–Brain Barrier Mimetic Bilayer

“…Using REUS simulations, we studied equilibrium partitioning of two peptides into the BBB–mimetic bilayer at 310 K. In the process, the peptides adopt a manifold of states with different orientations, interactions with the bilayer and water, and levels of bilayer disorder. To concisely analyze peptide partitioning, we followed our previous study and distinguished five layers along the bilayer normal, which include water ( z w < ζ), interface ( z b < ζ < z w ), headgroup ( z h < ζ < z b ), hydrophobic ( z c < ζ < z h ), and core (ζ < z c ) layers. In the Supporting Information (Figure S6), we explain the selection of layer boundaries.…”

“…REUS facilitates equilibrium sampling by simulating all replicas in parallel and attempting their periodic swaps between biasing windows. The REUS exchange formalism was described previously. , Replica exchanges result in a random walk of a replica over different biasing windows m and thus across the bilayer normal. Because REUS reduces trapping of a system in local free energy minima, it facilitates conformational sampling for molecule partitioning into a lipid bilayer.…”

“…As a result, REUS provides more accurate free energies for solute partitioning into lipid bilayers compared to umbrella sampling at about equal computational costs . Following our previous REUS study, replica exchanges were attempted every 2 ps by selecting replicas adjacent along z . For the KLVF and VVIA systems, the average acceptance rates for replica exchanges were 53 and 59%.…”

“…Replica-exchange umbrella sampling (REUS) implemented in NAMD was employed to enhance sampling of peptides partitioning into the bilayer. 30 To prepare R = 71 unique replicas distributed over M(=R) biasing windows, we followed the protocol described previously, 31 which is summarized in the Supporting Information. The center of mass of a tetrapeptide ζ in a window m (0 ≤ m ≤ M − 1) was restrained to z m by applying the harmonic potential W m (ζ) = 1/2k(ζ − z m ) 2 , where k = 1.5 kcal mol −1 Å −2 .…”

“…32 Following our previous REUS study, replica exchanges were attempted every 2 ps by selecting replicas adjacent along z. 31 For the KLVF and VVIA systems, the average acceptance rates for replica exchanges were 53 and 59%. Each replica in each trajectory was sampled for 160 ns, totaling τ = 22.72 μs per system.…”