Multiscale simulations combined with experiments identify the protein motifs responsible for membrane binding.

ABHD5 selectively binds to droplet-embedded vesicles (DEVs). Schematic of a DEV consisting of a phospholipid bilayer (gray) enclosing a LD monolayer (yellow) containing TAG. The white arrow indicates the bilayer, and the black arrow indicates the monolayer (A). Brightfield image of a DEV showing the phase-separated bilayer (white arrow) and droplet monolayer (black arrow) (B). Fluorescence image of ABHD5-CFP bound to the DEV, showing strong enrichment at the TAG-containing monolayer (black arrow) and absence from the bilayer (white arrow) (C). Our computational workflow incorporates validating and equilibrating the AlphaFold protein structure (D), converting this structure to a CG model (E), simulating the protein conformation changes and membrane remodeling with CGMD (F), and back-conversion and equilibration of the structures with GaMD (G).

ABHD5 interacts with membranes via two discrete regions.

(A-B) Percentage of contact time between each ABHD5 residue and individual lipid species (POPC, DOPE, SAPI, and TAG) from CGMD simulations on ER (A) and LD (B) membranes. Two key membrane-contacting regions are highlighted: the N-terminal region (amino acids 1-43) and a membrane-anchoring domain within the insertion segment (amino acids 180-230). A zoomed view of the interaction pattern in these regions is provided in Figure S1. This data represents the average of contact time from three independent CGMD simulations, with the corresponding error shown in Fig. S2. (C) LD binding induces protection against H/D exchange in the ABHD5 membrane-contacting regions. Horizontal lines indicate peptic peptides from ABHD5 subjected to H/D exchange for 10 s at pH 7.3 and 22°C, in the absence or presence of LDs. The y-axis shows differential deuterium uptake (free ABHD5 minus LD-bound). Fifteen peptides (red lines) spanning residues 8-44 and 190-280 exhibit significant protection upon LD binding. Peptides from other regions (blue lines) show no significant change in deuterium uptake.

Atomistic simulations reveal insertion of ABHD5 into LD and ER membranes.

(A, D) Representative side views from GaMD simulations show ABHD5 inserted into ER (A) and LD (D) membranes. The membrane surface is indicated by a straight (ER) or curved (LD) line. Membrane-interacting regions, N-terminal (residues 1-43, blue) and the domain within the insertion segment (residues 180-230, magenta), are highlighted. (B, E) Close views of the N-terminal interactions with ER (B) and LD (E) membranes. Charged residues are colored blue, polar residues green, and nonpolar residues gray. (C, F) Detailed snapshots of the domain within the insertion segment engaging the ER (C) or LD (F) membrane. Membrane embedding occurs through hydrophobic and electrostatic contacts that differ across organelle-specific lipid environments.

Confocal fluorescence imaging of live cells and model LDs validate the importance of hydrophobic residues for ABHD5 binding to membranes.

(A) WT ABHD5 was targeted to LD and ER membranes, whereas (B) the triple-mutant W199A, F210A, F222A ABHD5 was not. (C) LDs were visualized by their incorporation of Bodipy. (D) The color merge image shows the membrane and LD-bound WT ABHD5 colocalized with the LDs while the mutant ABHD5 is cytosolic and within the nucleus. (E) 3D reconstruction image of artificial LD caps shows the concentration of the ABHD5 variants (yellow) on the membrane (red). (F) Mutation of W181A or the triple mutant decreased binding rate of ABHD5 during the first 5 mins to the aLD surface. (G) After 24 hours of equilibration, all three mutants demonstrate significantly lower equilibrium binding to the aLD than the WT ABHD5. (A-D) The scale bars represent 20 µm.

Membrane binding induces conformational changes of the ABHD5 lid helix.

(A-C) Representative snapshots from GaMD simulations showing ABHD5 in solution (A), bound to ER membrane (B), and bound to LD membrane (C). The membrane-anchoring domain within the insertion segment (residues 180-230) is shown in magenta; the lid helix (residues 198-207) is highlighted in pink. Key residues within the pseudosubstrate pocket (F86, W181, Y272, Y330) are shown in orange spheres. (D) Distance between the center of mass of the ABHD5 core and the lid helix during GaMD simulations, illustrating lid displacement upon membrane binding. (E) Per-residue Cα RMSF values averaged across three independent simulations for each state (solution, ER-bound, LD-bound). The N-terminal region (residues 1-43) and the membrane-anchoring domain within the insertion segment (residues 180-230) exhibit state-dependent differences in flexibility.

Spatial distribution of lipid species surrounding ABHD5 on ER and LD membranes.

Lipid density maps from CGMD simulations showing the average local enrichment of four lipid species, POPC (red), DOPE (green), SAPI (blue), and TAG (purple), around ABHD5 on ER (A) and LD (B) membranes. Darker colors indicate higher local lipid density. The dashed white outlines represent the projected footprint of ABHD5 on the membrane surface.

ABHD5 induces local membrane curvature and lipid redistribution on LDs.

(A) Time-averaged height map of the LD membrane surface over the final 200 ns of GaMD simulation with ABHD5 bound (dashed outline). A local elevation of phospholipids is observed beneath the protein. White pixels indicate absence of phospholipid coverage. (B) Radially averaged phospholipid height as a function of distance from the ABHD5 center for three independent LD simulations. (C) Time-averaged height map of TAG molecules in the same simulation, showing a pronounced TAG-rich nanodomain beneath ABHD5. (D) Radially averaged TAG height from the same simulation set. Error bars in (B) and (D) represent standard deviations across the 200 ns trajectory window for each replica. (E-F) Side views of ABHD5 bound to ER (E) and LD (F) membranes from representative simulation snapshots. The membrane-anchoring domain (residues 180-230) is highlighted in magenta, and the lid helix (residues 198-207) in pink. POPC, DOPE, and SAPI phosphates are shown as orange spheres; TAG molecules forming the nanodomain are shown as a gray surface/lines.