How well do empirical molecular mechanics force fields model the cholesterol condensing effect?

Membrane properties are determined in part by lipid composition, and cholesterol plays a large role in determining these properties. Cellular membranes show a diverse range of cholesterol compositions, the effects of which include alterations to cellular biomechanics, lipid raft formation, membrane fusion, signaling pathways, metabolism, pharmaceutical therapeutic efficacy, and disease onset. In addition, cholesterol plays an important role in non-cellular membranes, with its concentration in the skin lipid matrix being implicated in several skin diseases. In phospholipid membranes, cholesterol increases the tail ordering of neighboring lipids, decreasing the membrane lateral area and increasing the thickness. This reduction in the lateral area, known as the cholesterol condensing effect, results from cholesterol-lipid mixtures deviating from ideal mixing. Capturing the cholesterol condensing effect is crucial for molecular dynamics simulations as it directly affects the accuracy of predicted membrane properties, which are essential for understanding membrane function. We present a comparative analysis of cholesterol models across several popular force fields: CHARMM36, Slipids, Lipid17, GROMOS 53A6L, GROMOS-CKP, MARTINI 2, MARTINI 3, and ELBA. The simulations of 1,2-dimyristoyl-sn-glycero-3-phosphocholine (DMPC) and 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC) membranes with varying cholesterol concentrations were conducted to calculate the partial-molecular areas of cholesterol and other condensing parameters, which are compared to the experimental data for validation. While all tested force fields predict small negative deviations from ideal mixing in cholesterol-DOPC membranes, only all-atom force fields capture the larger deviations expected in DMPC membranes. United-atom and coarse-grained models under-predict this effect, condensing fewer neighboring lipids by smaller magnitudes, resulting in too small deviations from ideal mixing. These results suggest that all-atom force fields, particularly CHARMM36 or Slipids, should be used for accurate simulations of cholesterol-containing membranes.