Lipid-protein interactions mapped by Markov state modeling and cryo-electron microscopy.

(A) GLIC structure (gray) embedded in a lipid bilayer (blue) with interacting lipids highlighted (yellow). (B) Markov state models were used to cluster simulations conducted under resting (R) or activating (A) conditions into five states, including closed (orange outline) and open (blue outline), with experimental structures plotted in white. (C) The duration of lipid interactions were measured in individual trajectories. (D) From clustered trajectory frames, lipid-occupied densities were obtained for closed (orange) and open (blue) states. (E) Further simulations were performed to test the role of residues involved in extensive lipid interactions. (F) View of a manually built cryo-EM model of closed GLIC (gray) with computationally derived lipid occupancies in orange (semi-transparent). (G) Cryo-EM reconstruction of closed GLIC (gray), with partly defined non-protein densities (orange). (H) Equivalent view of a manually built cryo-EM model as in F, with newly built lipids shown as sticks (yellow, heteroatom; phosphorus, orange; oxygen, red; nitrogen, blue).

Cryo-EM data collection and model refinement statistics

Cryo-EM and simulations reveal lipid interaction sites.

(A) Cryo-EM reconstruction of closed GLIC with distance to the interacting lipids colored according to the bar below. (B) Mean duration time of lipid-residue interactions, scaled (white–red) according the the color bar below, projected onto the closed-state GLIC structure (PDB ID: 4NPQ). For both panels, a single subunit is displayed in cartoon. (C) Secondary structure schematic for GLIC (top), mean duration time of each lipid-residue interaction (middle), and number of lipids interacting with each residue during simulations (bottom).

Protein and lipid determinants of state-independent inner-leaflet binding.

(A) Computationally-derived densities from the open (blue) and closed (orange) states generally agree. (B) Modeled lower leaflet lipids (yellow, stick, heteroatom coloring) from cryo-EM densities (brown mesh). (C) Zoom-in view of the lower leaflet buried lipid interaction site at the M1-M3 subunit interface. Lipids particularly interact with residues T274 (purple) and W217 (magenta). (D) The 100 simulation frames that best fit the computational densities could be clustered into two distinct binding poses at the T274 binding site. When the tail with a double-bond was in the pocket (left) lipid heads are directed out from the channel, while it samples a larger variety of poses when the lipid tail without a double-bond is in the pocket (right), with possibility of the head to enter into a crevice on the bottom of the channel, approaching the pore. The M4 helix is not shown for clarity in the top panels. (E) Number of contacts made by specific POPC lipid atoms with residue T274 in simulations at both resting and activating conditions (red indicate high numbers and white low). For the tail with a double-bond the interactions are concentrated around the double-bond region while for the tail without a double-bond interactions are interspersed along the tail. (F) Mutation of residue W217 lining this pocket, reveals shortened interactions for an alanine mutation (magenta).

State-dependent protein-lipid interactions in the upper leaflet.

(A) Zoom view of the outer transmembrane domain of the closed GLIC cryo-EM structure (left, gray), showing non-protein cryo-EM densities (brown mesh) overlaid with built lipids (yellow). For comparison, the same region is shown of an open state x-ray structure (PDB ID: 6HZW) [16] with built lipid and detergent in equivalent positions (right). Conformational changes of the M2-M3 loop, including residue P250, in channel opening are highlighted. Lipids are colored by heteroatom (phosphorus, orange; oxygen, red; nitrogen, blue). (B) Same region as in (A) showing densities occupied by lipids (≥40%) in simulation frames clustered as closed (left, orange) or open (right, blue) states (top). Open-state differences include an enhanced density on the right-hand side of the intersubunit cleft, and a more deeply penetrating density in the intrasubunit cleft. Overlaid densities represent simulations conducted under resting (dark shades) or activating (light shades) conditions, which were largely superimposable within each state. (C) Radial distribution of the lipid atoms closest to the upper pore, showing closer association of lipids in the open (light blue, dark blue) versus closed (orange, red) states. (D) Same region and cryo-EM densities (mesh) as in (A) overlaid with lipids from a few simulation frames where the lipids had the highest correlation with the closed state computational occupancies in (B). (E) Snapshots from simulation frames that had the highest correlation with the occupancies (activating conditions) in (B). The open states are characterized by an additional lipid at the intersubunit site, interacting with the oxygen of P250. (F) The minimum distance between P250 and the closest lipid projected onto the free energy landscapes obtained from Markov state modeling. P250-lipid interactions are possible in the open state, but not in the closed state.

Lipid sites identified by cryo-EM and simulations.

(A) In the closed state, five lipid interaction sites were independently identified through both cryo-EM and simulations. In the outer leaflet, two intrasubunit sites are situated at the principal (P) and complementary (C) subunits, with the lipid at the principal subunit site interacting with S196 of the pre-M1 loop. Particularly, no lipid site could be found at the subunit interface. In the inner leaflet, a buried lipid pose was found at the subunit interface, determined by interactions with T274 and W217 of the M3 and M1 helices, respectively. (B) In the open state, six lipid interaction sites were identified from simulations, out of which three sites were known from previous structures and one presumably occupied by a detergent molecule [16] (thin blue stripes). In the outer leaflet, an additional lipid site could be found at the subunit interface in the open state, interacting with P250 on the M2-M3 loop. Additionally, the site at the principal intrasubunit site is slighty closer to the M4 helix compared to the closed state, and the lipid site at the principal intrasubunit site has an increased likelihood of tail penetration into an allosteric pocket compared to the closed state. Lipid sites in the inner leaflet are at positions equivalent to those in the closed state.

Cryo-EM data and processing pipeline.

(A) Representative micrograph from a dataset collected on a Titan Krios, showing detergent-solubilized GLIC particles (top. Representative 2D class averages at 0.82 Å/px in a 256 × 256 pixel box and a 180 Å mask bottom. (B) Overview of the cryo-EM processing pipeline for a merged GLIC data. (C) FSC curves for unmasked (red) and masked (blue) map.

W217 and T274 mutants mean duration times compared to wild type.

Mean duration time (µs) of lipid interaction time for GLIC wild type (black, top row), T274A mutant (purple, middle row) and T274W mutant (light purple, bottom row).

State-dependent lipid interactions and conformational changes in the upper transmembrane domain.

(A) Radial distributions of nearest lipid atom from the upper pore, with a presumed pre-desensitized state displaying the highest probability of lipid tail penetration (pink), followed by open states (dark blue, light blue) and closed states (red, orange). (B) Radial distance to the -2’ gate projected onto the tIC space at activating conditions. The presumed pre-desensitized state is marked in pink. (C) Lipid densities derived from simulations with ≥40% occupancies, representing open (blue) and closed (orange) states. (D) Lipid densities from closed state simulation frames with ≥40% occupancy (orange) and ≥30% (white). Built cryo-EM lipids are shown in yellow. (E) Minimum residue-lipid distance projected onto the activating condition free energy landscape for residues Q193, S196 and P250. (F) The normalized number of residue-lipid atom contacts displayed for each of the residues in (E), with red areas highlighting atoms with close interactions to the residue in (E). (G) Representative GLIC open and closed state simulation frames with the highest correlations between lipid positions and the computational occupancies (Figure 4B).