Structure and activity of lipid bilayer within a membrane-protein transporter

Lipid Bilayer Ordering in Native Cell-Membrane Particles of AcrB.

We prepared native cell-membrane nanoparticles of E. coli AcrB using membrane-active SMA polymers. The nanoparticles were purified by single-step Ni-affinity chromatography, applied directly to grids, and vitrified for single-particle cryo-EM analysis. A 3D reconstruction with C1 symmetry achieved a final density map of 3.2-Å resolution (Fig. 1 A and B and SI Appendix, Fig. S1). We initially tried to reconstruct the 3D EM map in C3 symmetry; however, that density map was fragmented, especially so in the TM region, and we could not see lipid-bilayer structure in the central cavity. The C1 reconstruction was fitted by an asymmetric AcrB trimer (Fig. 1C and SI Appendix, Fig. S2), where each protomer exists in a distinct state (L for loose, binding-ready; T for tight, substrate-bound; and O for open, substrate release) as in the asymmetric crystal structures (28, 29, 32), but here in this cryo-EM structure with differences from corresponding subunits in the crystal structures [r.m.s.d. on Cα positions of 1.9 Å (L), 1.2 Å (T), and 1.0 Å (O) vs. PDB ID code 4U8Y (32)].

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

EM density of AcrB in a native cell-membrane nanoparticle. (A and B) Surfaces of EM density features. Gray-colored surfaces show density features covered by the protein model for the AcrB trimer. Yellow-colored surfaces show remaining density features, presumed to be from lipids. The density within the central cavity between AcrB TM domains is interpreted as a patch of lipid bilayer. (A) Side view, as seen from within the membrane. (B) Bottom view, as seen from the cytoplasm. (C) Ribbon diagram of the AcrB trimer (L-state chain A: cyan; T-state chain B: orange; O-state chain C: gray) with superimposed EM density in the central cavity (yellow). (D) Enlarged view of boxed region of EM density for the native cell-membrane lipid bilayer.

There is a distinct lipid belt around the TM region and a patch of ordered lipid bilayer in the lipid cavity of the AcrB trimer (Fig. 1). The TM density covered by the protein model has a diameter of ∼9 nm, whereas the diameter including the lipid belt is ∼12 nm. We resolved 24 lipid molecules in the central cavity patch, and an overall total of 31 complete lipid molecules plus an additional 11 individual alkyl chains that could be from other lipid molecules. We found no evidence of ordered SMA molecules.

The patch of lipid bilayer in the central cavity is shaped like a triangular two-layer cake (Fig. 2 A–D) with each side facing the TM domain of a particular subunit (Fig. 2E). The overall shapes and orientations of the leaflet on the periplasmic side (outer leaflet) and of the cytosol-facing leaflet (inner leaflet) are distinctly different (Fig. 2 B–D). Twelve lipid molecules, built as the predominant bacterial lipid phosphatidylethanolamine (PE) (33), could be fitted into the EM density of each leaflet (Fig. 2 C and D). The thickness of the lipid bilayer is ∼31 Å, as calculated from the averaged Z coordinates for phosphorus atoms of the modeled lipid molecules.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Features of the lipid bilayer from the central cavity. (A) Side view of the EM density of lipid-bilayer structure (blue). The patch of the lipid bilayer looks like a triangular two-layer cake. Lipid molecules fitted at the apices of each triangle are drawn in red. (B) Lipid bilayer as viewed from the periplasmic space (top view). The outer leaflet (orange) is rotated ∼10° clockwise relative to the inner leaflet (blue). (C and D) EM density with superimposed lipid models for the outer leaflet (C) and inner leaflet (D). Here, both EM maps are colored blue, but distinctive triangular shapes relate C to orange and D to blue in B. The lipid model is in stick representation with colored phosphorus (orange), carbon (gray), and nitrogen (blue). (E) Stereoview of the central cavity region of AcrB viewed as in B. Protein ribbons are colored as in Fig. 1C. The phosphoryl heads from lipids 1, 5, and 9 are at the cytoplasmic surface of the inner leaflet of the bilayer, and those from lipids 13, 17, and 21 are at the periplasmic surface of the outer leaflet. Triangles that connect phosphorous atoms in these two sets of apical lipids are drawn to emphasize differences in size, shape, and orientation for the two leaflets.

A striking feature of the lipid bilayer is that the outer triangle of lipids is rotated relative to the inner triangle by ∼10° clockwise, as viewed from the periplasm (Fig. 2 B–D). Whereas the inner leaflet has lipid molecules packed quite tightly, there are some spaces between lipids within the outer leaflet, but these are occupied by protein side chains (Lipids Interact Extensively with AcrB Protein, both in the Central Cavity and in the Belt Outside). Thus, in terms of lipids alone, the outer leaflet is larger than the inner one (Fig. 2 C and D). A particular lipid molecule is at each triangular apex, and the distances between the phosphorus atoms at these apices are, respectively, 26.8, 25.4, and 22.4 Å for the inner leaflet and 28.5, 29.9, and 33.4 Å for the outer leaflet (Fig. 2E). Thereby, these triangles have areas of 263.5 and 399.7 Ų, respectively. The phospholipids of the inner leaflet are disposed in a regular pattern with alkyl tails mostly straight; outer-leaflet phospholipids are also ordered, meaning that they are defined by distinct density features, but their alkyl tails are mostly curved (Figs. 2 C and D and 3 A–C).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Structure of lipid bilayer and protein–lipid interactions. (A) Atomic model of the lipid bilayer in stick representation with coloring of phosphorus (orange), oxygen (red), nitrogen (blue), and carbon (gray). (B) Atomic model of the outer leaflet showing 12 lipid molecules (sticks) and protein residues that protrude from each subunit into the lipid array. Protein residues are colored as in Fig. 1C: subunits A (cyan), B (orange), and C (gray). (C) Atomic model of 12 lipid molecules in the inner leaflet. (D) Hydrophobic interactions in the central cavity between the lipid bilayer (yellow density surfaces) and AcrB subunit A. A385, F386, and F458 interact with the outer leaflet, and M1, F4, F11, and M447 do so with the inner leaflet. Subunits B and C make similar but distinct interactions at other faces of the triangular bilayer patch. (E) Interactions of residue R8 from AcrB subunit B (T state) with lipid 9 from the inner leaflet. Hydrogen bonding between the guanidyl and phosphoryl groups is indicated. (F) Interactions of residue H338 with lipid A from the lipid belt surrounding TM domains of the AcrB trimer.

Lipids Interact Extensively with AcrB Protein, both in the Central Cavity and in the Belt Outside.

We found several specific hydrophobic protein–lipid interactions in the central cavity. A385, F386, and F458 from each protomer all protrude into the outer leaflet (Fig. 3B), occupying spaces left from disrupted packing of the lipid tails (Fig. 2C). As well, M1, F4, F11, and M447 from each protomer interact with the inner leaflet (Fig. 3D and SI Appendix, Fig. S3). In addition, certain lipid head groups interact with the protein through hydrogen bonding. In particular, the guanidyl group of R8 on subunit B (T state) hydrogen bonds (3.1 Å) with the phosphoryl group of lipid 9 in the inner leaflet (Fig. 3E and SI Appendix, Fig. S4E), and the corresponding R8 on subunit C (O state) is at possible hydrogen-bonding distance (3.4 Å) to the phosphoryl group of lipid 5 (SI Appendix, Fig. S4F); however, R8 of subunit A (L state) is too far (closest distance of 4.3 Å) from the phosphoryl group of corresponding lipid 1 (SI Appendix, Fig. S4D). In the outer leaflet, the phosphoryl group from lipid 13 hydrogen bonds (3.1 Å) with the backbone N of G460 from subunit C (SI Appendix, Fig. S4I); however, corresponding lipids 17 and 21 are too distant from their corresponding F459–G460–G461 segments for hydrogen bonding (SI Appendix, Fig. S4 G and H). The distinctions in lipid interactions with the AcrB trimer reflect the intrinsic asymmetry of AcrB and its associated lipid bilayer.

The lipid belt surrounding the TM region of AcrB is generally much less ordered than the lipid-filled central cavity, but some lipid molecules do interact directly with outward-facing TM helices. For example, the phosphoryl group of lipid A hydrogen bonds (3.3 Å) to N_δ of H338 from subunit A (Fig. 3F and SI Appendix, Fig. S5A). Lipid B is similarly disposed in relation to H338 from subunit B; however, this interaction seems slightly too long for hydrogen bonding (SI Appendix, Fig. S4B). Lipid C occupies an analogous site in relation to subunit C, but here H338 is remote from the head group (SI Appendix, Fig. S4C). Besides those lipid molecules nearby to H338, several other lipid-belt lipid molecules also make hydrophobic interactions with outer helices from TM domains (SI Appendix, Fig. S5).

Intimate lipid–protein interactions were also observed in 2D crystals of aquaporin-0 (34), although in this case the system was reconstituted from 1,2-Dimyristoyl-sn-glycero-3-phosphocholine lipids, whereas our AcrB particles were extracted directly from natural membranes.

Lipid-Bilayer Function in Harmonizing Peristaltic Conformational Changes Through AcrB Trimers.

Whereas lipid tails in the outer layer are irregularly curved and loosely packed, those in the inner layer are relatively straight and quite close-packed (Fig. 3A). In cross-section, midway through the inner leaflet, the EM density shows a hexagonal pattern (Fig. 4A) that is remarkably similar to the pattern in a PE crystal structure (2), predicted 40 y ago to reflect natural membranes; however, the real lipid cell-membrane structure is much more complex than that of artificial lipid structure. We numbered the 12 PE lipid molecules in the inner leaflet, 1–12, and we identify associated alkyl chains in the grid with these numbers (Fig. 4B). Protein also contributes to the hexagonal pattern; for example, densities for M1 from each of the subunits and F4 from subunit A (L state) are seen in Fig. 4C, and M1 is actually cut through in the cross-section. For 10 of the 12 PE molecules, the head-group density is well defined and the two pair of unconnected lipid tails are adjacent; thus, all 24 lipid tails are assigned to head groups of specific lipid molecules as drawn in Fig. 4D.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

Hexagonal pattern within the lipid bilayer. (A and B) EM density drawn as a solid yellow surface and sliced through the inner leaflet. Sliced surfaces in A are colored red. Lines of a hexagonal grid are drawn in yellow. Hydrocarbon tails in B are numbered by lipid identifiers 1–12. Density for protein residue M1 also occupies a grid position in this cut. (C) Stereoview of the EM density of the inner leaflet of the lipid bilayer. (D) Hexagonal grid as in A. The red dot represents the phosphoryl head, green is for a tail on glycerol position C2, purple is for a tail on glycerol position C3, and yellow is for unassigned positions. (E) Proposed hexagonal pattern for the inner leaflet in a C3-symmetric AcrB trimer. The red triangle marks the threefold axis position. Blue arrow-directed lines specify translations and rotations in lipid positions that can shift this symmetric pattern into the asymmetric pattern as that of D. (F) Correspondence of lipid numberings in the asymmetric lipid array (D) with those in the symmetric lipid array (E).

Natural lipid bilayers are fluid and they can adapt, albeit with certain resistance, to conformational changes in associated proteins. AcrB itself takes on multiple states. The three subunits in asymmetric AcrB structures (28, 29, 32) are distinct, both from one another and also from the conformation in symmetric AcrB (25). This distinction has implications for the lipid bilayer that we expect will occupy the central cavity for all of these states when in a natural membrane. The lipid structure must accommodate all of these protein states. In Fig. 4E, we present a lipid array reorganized into threefold symmetry from the one observed in our EM structure, and we include paths of transformation that can move from the lipid array back into the asymmetric array of Fig. 4D. The proposed transport mechanism for AcrB (32) has each protomer moving successively and in coordination from state L through T into O and then back to L (L → T → O → L). As these protein movements occur, the lipid structure must also move to accommodate.

The net result for the working system gives the appearance of 120° rotations at each step; in fact, however, subunits stay in place while undergoing the succession of conformational changes, which in turn must be accompanied by shifts in the lipid structure, as envisioned in SI Appendix, Fig. S7. From examination of the extensive protein–lipid interactions in AcrB (Fig. 3D and SI Appendix, Figs. S3 and S4), we propose that the lipid bilayer in the central cavity serves to harmonize conformational changes in the peristaltic mechanism of drug extrusion by AcrB. Through defined protein contacts, the lipid bilayer senses the conformational changes that occur in each TM domain and then transduces effects of these changes through the lipid bilayer to neighboring protomers in a viscous interplay between cavity lipids and the AcrB trimer. This process happens reciprocally, such as to synchronize movement of client drugs through the pseudorotatory AcrB trimer.

Lipid Bilayer in the AcrB D407A Trimer Mutant.

The TM domain of AcrB contains conserved amino acid residues proposed to be important for a proton relay mechanism in the drug/proton antiport activity. Crystal structures of AcrB with mutated proton-relay residues, D407A, D408A, K940A, and T978A, all showed a dramatic collapse of TM domains toward the central cavity (35). Most noticeably, the distances between F386 positions dropped from 17.6 Å in wild-type AcrB to 6.7 Å for mutant AcrB D407A for the detergent-solubilized protein in crystal structures (Fig. 5E). Similar TM shifts occurred for the other mutant proteins.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5.

Lipid bilayer in AcrB D407A trimer prevents large distance movement of F386. (A) EM density drawn as solid yellow surface and sliced through the inner leaflet (red cross-sections). The hydrocarbon tails show a hexagonal pattern as in Fig. 4A. (B) Hydrophobic interactions in the central cavity between the lipid bilayer (yellow density surface) and AcrB subunit A. A385, F386, and F458 interact with the outer leaflet, and M1, F4, and M447 interact with the inner leaflet. Subunits B and C interact similarly but distinctly at the other faces of the triangular bilayer patch. (C) EM density and conformation of A407, D408, K940, N941, and T978 within AcrB D407A mutant. K940 and T978 form a hydrogen bond with a distance of 3.3 Å; K940 and N941 form a hydrogen bond with a distance of 2.5 Å. (D) EM density and conformation of D407A, D408, K940, N-941, and T978 within wild-type AcrB. K940 forms a hydrogen bond with D407 with a distance of 3.5 Å; K940 also forms a hydrogen bond with T978 with a distance of 3.4 Å. (E) Stick model of AcrB D407A for F386 residues superimposed on space-filing wild-type AcrB residues. (F) F386 residues from the crystal structure of detergent-extracted D407A AcrB (magenta) compared with those from the wild-type crystal structure (cyan). F386–F386 distances shift from 17.6 Å down to 6.7 Å as shown. For reference, the stick model for F386 residues in the cryo-EM model of wild-type AcrB (white) is also copied here from E.

To test importance of the lipid bilayer for the structure and activity of AcrB, in light of the remarkably large changes reported previously, we also determined the cryo-EM structure for SMA-extracted AcrB D407A trimer at 3.0-Å resolution. As for wild-type AcrB, the AcrB D407A structure has a similar lipid-bilayer patch located in its central cavity. As for wild-type AcrB, lipid hydrocarbon tails in this inner leaflet are also hexagonally arranged (Fig. 5A), and we observed similar hydrophobic interactions between AcrB and the lipid bilayer (Fig. 5B). In comparison with the structure of the wild-type AcrB trimer, the amino acid residues that interact with the lipid bilayer are relatively unmoved (Figs. 3D and 5B). Consistent with this observation, the r.m.s.d. value is less than 1 Å in a comparison between the wild-type AcrB and AcrB D407 using all Cα atoms.

The role of the conserved residues, and of the lipid bilayer, in the proton relay mechanism for AcrB needs further study. We do observe changes in the D407A mutant structure, and also in its lipid bilayer, compared with our wild-type AcrB structure (Fig. 5 C and D); however, these changes are subtle compared with the dramatic shifts seen for proton relay mutants in the detergent-extracted situation (Fig. 5 E and F). The likely factor leading to collapse of TM domains in crystal structures of the mutants is the absence of the supporting lipid bilayer because of the use of detergents. The lipid bilayer, as preserved in the central cavity after SMA extraction, provides a restraining structural support for the TM domains (Fig. 5B). The tight packing of lipid molecules in the inner leaflet also suggests that the central cavity is not part of the drug-transport pathway.

Prospects for Native Cell Membrane Nanoparticles.

Detergents have been essential for advances in membrane-protein structural biology, but they also have limitations because the lipid bilayers that detergent solubilization destroys may be crucial for membrane-protein function and stability. The best niche for a membrane protein is in its native cell-membrane environment. The system that we have been developing for detergent-free membrane-protein solubilization into native cell-membrane nanoparticles appears to preserve much of the natural lipid environment. From this cryo-EM analysis of AcrB as solubilized directly with SMA copolymer, we could build a total of 31 lipid molecules and 11 additional hydrocarbon chains that likely derive from lipids. Most remarkably, the central cavity between AcrB TM domains sustains a 24-lipid patch of mostly well-ordered bilayer structure. Regularity in the hexagonal pattern of the inner leaflet is similar to that in a PE crystal structure (2), and this regularity contrasts with highly irregular packing in the outer leaflet. Protein side chains interact with both leaflets and participate in the hexagonal pattern. Lipid ordering in the protein confines of the AcrB central cavity may be a special situation, but we also see well-ordered lipid structure in the surrounding lipid belt, even with our relatively unsophisticated analysis of the map.

A system such as ours for preparing native cell-membrane nanoparticles has certain advantages. First, and most importantly, a protein can be extracted with its native local membrane structure largely intact. Whereas apolipodisc (36), bicelle (37), or saposin (38) alternatives need to include detergents at some stage, here we could remain truly detergent-free. Second, the native cell-membrane nanoparticle system might catch membrane-protein complexes that are labile in detergents, as was demonstrated for a plant metabolon (39). Lastly, this detergent-free system is well suited for single-particle cryo-EM analysis, providing evenly distributed particles. Our current native cell-membrane nanoparticle system still has shortcomings. Not all tested membrane proteins performed well. However, the system has much scope for improvement, and we expect a very positive impact on membrane-protein research.