Algorithm to catalogue topologies of dynamic lipid hydrogen-bond networks

Graph-Based Analyses of Dynamic Water-Mediated Hydrogen-Bond Networks in Phosphatidylserine: Cholesterol Membranes

“…The topology of a H-bond lipid cluster is defined as the geometric arrangement of the nodes and edges in that cluster. To identify lipid clusters, we used the recently developed DFS algorithm [ 17 ] that analyses the H-bond graph to identify transient lipid clusters and the topologies of these clusters. Briefly, the DFS algorithm treats the nodes and edges of the H-bond graph as, respectively, data points and relationships between these data points.…”

“…Altered distribution of PS has been associated with a number of human diseases [ 13 ], including cancer [ 14 , 15 , 16 ]. The formation of lipid clusters, particularly in the case of PS [ 17 , 18 , 19 ], PG [ 17 ], and phosphatidylinositol 4,5-bisphosphate (PIP2) [ 20 ], might be promoted by the presence of cations. Experiments using, e.g., time-resolved fluorescence spectroscopy [ 18 ], fluorescence correlation spectroscopy [ 20 , 21 ], steady-state probe-partitioning fluorescence resonance energy transfer (SP-FRET) [ 20 ], and molecular dynamics (MD) simulations [ 18 , 19 , 22 ] suggested that the presence of sodium [ 9 , 18 , 21 , 23 , 24 , 25 , 26 ], potassium [ 18 , 22 ], calcium [ 19 , 22 , 25 , 26 , 27 , 28 ], or magnesium ions [ 28 ] may be associated with the clustering of lipids.…”

“…MD studies have shown that lipids can form H-bond networks mediated by direct or water-mediated H-bonds between lipid headgroups [ 5 , 17 ]. The water dynamics at the lipid interface is slower than the bulk [ 1 , 40 , 41 ].…”

“…To identify and characterize dynamic water-mediated clusters of lipid headgroups, we have recently implemented a Depth-First-Search (DFS)-based graph algorithm that uses Connected Component searches to explore the nodes of the H-bond graphs and identify four main types of lipid H-bond clusters [ 17 ]. Clusters are then characterized according to properties such as cluster size, which is given by the number of nodes (lipid headgroups) that constitute the cluster, and cluster length, which is given by the number of H-bonds within the longest H-bond path within the cluster [ 17 ]. Using the DFS-based algorithm we implemented, we dissected the water-mediated H-bond clusters at the interface of hydrated lipid bilayers composed of POPE, POPG, 3:1 POPE:POPG, 5:1 POPE:POPG, POPS, and an Escherichia coli lipid membrane model, all without cholesterol [ 17 ].…”

“…Clusters are then characterized according to properties such as cluster size, which is given by the number of nodes (lipid headgroups) that constitute the cluster, and cluster length, which is given by the number of H-bonds within the longest H-bond path within the cluster [ 17 ]. Using the DFS-based algorithm we implemented, we dissected the water-mediated H-bond clusters at the interface of hydrated lipid bilayers composed of POPE, POPG, 3:1 POPE:POPG, 5:1 POPE:POPG, POPS, and an Escherichia coli lipid membrane model, all without cholesterol [ 17 ]. We found that, regardless of the lipid membrane composition, short linear H-bond arrangements, typically two lipid headgroups that H-bond directly or via one water molecule, were sampled frequently during atomistic MD simulations [ 17 ]; relative to the other lipids studied, POPS could engage more frequently in circular arrangements of three lipid headgroups [ 17 ].…”