Skip to main page content
Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2014 Jun 5;510(7503):172-175.
doi: 10.1038/nature13419.

Membrane Proteins Bind Lipids Selectively to Modulate Their Structure and Function

Free PMC article

Membrane Proteins Bind Lipids Selectively to Modulate Their Structure and Function

Arthur Laganowsky et al. Nature. .
Free PMC article


Previous studies have established that the folding, structure and function of membrane proteins are influenced by their lipid environments and that lipids can bind to specific sites, for example, in potassium channels. Fundamental questions remain however regarding the extent of membrane protein selectivity towards lipids. Here we report a mass spectrometry approach designed to determine the selectivity of lipid binding to membrane protein complexes. We investigate the mechanosensitive channel of large conductance (MscL) from Mycobacterium tuberculosis and aquaporin Z (AqpZ) and the ammonia channel (AmtB) from Escherichia coli, using ion mobility mass spectrometry (IM-MS), which reports gas-phase collision cross-sections. We demonstrate that folded conformations of membrane protein complexes can exist in the gas phase. By resolving lipid-bound states, we then rank bound lipids on the basis of their ability to resist gas phase unfolding and thereby stabilize membrane protein structure. Lipids bind non-selectively and with high avidity to MscL, all imparting comparable stability; however, the highest-ranking lipid is phosphatidylinositol phosphate, in line with its proposed functional role in mechanosensation. AqpZ is also stabilized by many lipids, with cardiolipin imparting the most significant resistance to unfolding. Subsequently, through functional assays we show that cardiolipin modulates AqpZ function. Similar experiments identify AmtB as being highly selective for phosphatidylglycerol, prompting us to obtain an X-ray structure in this lipid membrane-like environment. The 2.3 Å resolution structure, when compared with others obtained without lipid bound, reveals distinct conformational changes that re-position AmtB residues to interact with the lipid bilayer. Our results demonstrate that resistance to unfolding correlates with specific lipid-binding events, enabling a distinction to be made between lipids that merely bind from those that modulate membrane protein structure and/or function. We anticipate that these findings will be important not only for defining the selectivity of membrane proteins towards lipids, but also for understanding the role of lipids in modulating protein function or drug binding.


Extended Data Figure 1
Extended Data Figure 1. Maintaining intact native membrane protein complexes in the mass spectrometer
a-b, Gas-phase unfolding plots (left, 5 V steps) and mass spectra (right) of detergent stripped AqpZ and AmtB ions (charge inset) from different non-ionic detergent solutions; C8E4 (pink), NG (green), OGNG (purple) and DDM (orange). Membrane protein complexes from NG, OGNG and DDM possessed CCS values substantially larger than those calculated from crystal structures. c, Complete removal of C8E4 at low collision voltages reveals CCS values consistent with those calculated from their respective crystal structures. d, Reported are measured masses, standard deviations, and empirical T0 values used for direct CCS calculation of membrane proteins studied.
Extended Data Figure 2
Extended Data Figure 2. Phospholipid abbreviations and optimization of phospholipid binding experiments
a, Phospholipid abbreviations and headgroup structures. b, Protein to phospholipid to detergent ratios (P:L:D) for each membrane protein, number of resolvable phospholipids bound with their masses. Conditions were optimized empirically to maintain nanospray, sufficient mass spectral quality, and phospholipid binding. Masses for one lipid bound to the protein complex were measured using MassLynx V4.1 software (Waters).
Extended Data Figure 3
Extended Data Figure 3. Phospholipid binding to MscL reveals an insignificant impact on protein gas-phase stabilization independent of lipid alkyl chain length
a, Representative mass and ion mobility spectra of MscL bound to various phospholipids species (inset top right). b, Representative mass and ion mobility spectra of MscL bound to phospholipids of varying alkyl chain length. c, Cumulative stabilization of MscL is seen for one (blue) to two (red) bound lipid molecules. Reported are the mean ± SEM (N = 3).
Extended Data Figure 4
Extended Data Figure 4. Summary of statistics for molecular dynamics (MD) simulations followed by CCS filtering
a, Representative filtering procedure of AqpZ in a PC bilayer from the MD (top). For each frame, lipids are extracted within 6 Å of the protein (bottom left) then filtered by the experimental CCS of AqpZ bound to a single PC molecule (bottom right), see Methods. b, Summary of statistics. c, Ratios of phospholipid bound to apo CCS values for membrane protein complexes. Reported are the mean and standard deviation (N = 3). d, MD filtered by CCS resulted in similar patches of PC molecules across different time points within the simulation indicating the systems were equilibrated.
Extended Data Figure 5
Extended Data Figure 5. MD simulations filtered by CCS reveals probable one (1x) and two (2x) phospholipid binding sites
a, For each candidate structure, protein and 1x or 2x PC molecule(s), the ratio of their calculated CCS values was determined (CCS Ratio). This procedure generated a large number of candidate structures (grey bars) that were filtered using our CCS measurements (cyan line). The structures in grey that intercept this curve are essentially the ones selected as our most probable ensemble. b, The intersection between the simulated lipid complexes and the experimental data is then projected onto the surface of the protein to identify the most probable binding sites. Probable lipid locations for MscL resembled an annular belt, with no specific patches of lipids likely stemming from the relatively cylindrical geometry of this complex. By contrast for AqpZ and AmtB the most probable location of the lipid molecules were localized to the interfacial regions between protein subunits, as well as other probable locations on individual monomers. c, X-ray derived PG (blue spheres with white tails) located at the subunit interfaces agrees with the predicted PC binding site.
Extended Data Figure 6
Extended Data Figure 6. Modeling and quantification of gas-phase unfolding pathways
a, Representative ion mobility mass spectra are collected over a range of collision voltages. b, Ion arrival time is converted to CCS before generating unfolding plots. c, Model fitting process from 2D to 3D data (See Methods). d, Both a Synapt2 modified with a linear drift cell (DT-IMS) and the commercially available travelling wave SynaptG2 (TW-IMS) produce qualitatively similar unfolding data. e, A contour plot representing the variance of CCS of two gas-phase unfolding species as a function of ion mobility drift cell potential. f, Stacked plots of arrival time distributions for two gas-phase unfolding species as a function of drift cell potential. The lifetime of unfolding protein complexes in the drift tube ranges from 4 to 15 ms depending on the drift cell potential. No additional unfolding post activation occurs implying that the unfolding mechanism is not consistent with an irreversible unfolding model. Such a mechanism would predict time dependence on the population of unfolded species. By contrast the unfolding mechanism is well described by the reversible unfolding mechanism (see Methods).
Extended Data Figure 7
Extended Data Figure 7. AqpZ, AmtB and AmtBN72A/N79A bound to various phospholipid species
a, Representative mass and ion mobility spectra of AqpZ bound to phospholipids. b, Representative mass and ion mobility spectra of AmtB bound to phospholipids and AmtBN72A/N79A bound to PG.
Extended Data Figure 8
Extended Data Figure 8. Summary of water permeability assays and analysis of lipid extracts
a, HPTLC analysis of total polar lipid extract from wild-type E. coli (EPL), cardiolipin-deficient strain (BKT22), or BKT22 cells expressing ClsC and YmdB (BKT22-YC) to restore cardiolipin. Lipids were quantified by densiotometry. b, Reported are the rate constants (kwat) and standard error of replicates (N = 5) for empty liposomes (−) and AqpZ proteoliposomes (+) reconstituted in differing E. coli lipid compositions.
Extended Data Figure 9
Extended Data Figure 9. Structural analysis of AmtB bound to PG
a, Crystal packing with Six AmtB (multicolored) and eight PG (orange) molecules located in the asymmetric unit cell and symmetry related molecules shown in grey and light orange, respectively. b, Fo-Fc and 2Fo-Fc electron density maps after refinement without lipid and near-lipid water molecule (if present) contoured at 2.0 and 1.0 sigma, respectively. c, Comparison of AmtB bound to PG (green chain, this work) aligned with the AmtB structure (maroon chain, pdb 1U7G). d, Structure overlay of AmtB-GlnK complex bound to octylglucoside (purple chain, pdb 2NS1) aligned with AmtB bound to PG reveals a distinct conformational change. The lipid-water interface (grey plane) was determined from coordinates of phosphate atoms from bound PG.
Figure 1
Figure 1. The mechanosensitive channel of large conductance (MscL) resists unfolding in the presence of lipids
a, MscL in electrospray droplets, within a lipid-detergent micelle, undergoes desolvation and activation. b, Mass spectrum reveals multiple MscL phosphatidylinositol phosphate (PI) bound states. Ion mobility (top) of MscL(PI)0-5 with a trend line for native state (dotted). c, Plot of collision voltage against CCS for +12 ions of apo (i) and MscL(PI)4 (ii). Experimental and modeled unfolding plots with collision voltages at which transitions occur and CCS values (horizontal and vertical arrows respectively). R2 values are provided. d, Stabilization calculated from parameters defined by fitting MscL (+12) with lipids. Reported are average and s.e.m. (N=3) in units of electron volts (eV).
Figure 2
Figure 2. AqpZ is stabilized indiscriminately by lipids with the exception of cardiolipin, a lipid that stabilizes the channel significantly and modulates its function
a, Mass spectra reveal AqpZ(POPC)0-5 resolved at 60 V with IM-MS arrival times in agreement with the CCS calculated from the crystal structure. b, Stabilization of AqpZ (+13) bound to lipids, calculated from unfolding parameters. One-way ANOVA (N = 3): **P < 0.01. c, Water permeability assay for AqpZ reconstituted in total polar lipid extracts from E. coli (EPL) (orange), or a cardiolipin-deficient strain (BKT22) (green), or BKT22 cells expressing ClsC and YmdB to restore cardiolipin (BKT22-YC) (cyan) are compared with empty EPL liposomes (pink) (Supplementary Figure 7). Rate constants of water transport (kwat) and standard error of replicates (N = 5).
Figure 3
Figure 3. Lipid binding to AmtB results in a range of stabilizing effects
a, Ion mobility mass spectra of AmtB(PG)0-4 molecules at 175V with IM-MS unfolding data as described in Fig. 1b. Arrival times for the various charge states reveal non-native, trimeric structures. b, Stabilization of AmtB (+15) bound to different lipids and AmtBN72A/N79A (+15) a mutant form designed to disrupt the specific PG-lipid binding site. Gas-phase unfolding reveals that AmtB is weakly stabilized by the majority of lipids but CDL and PG confer significant stabilization. AmtBN72A/N79A shows a reduction in stabilization by PG compared with the wild-type protein. Results are shown as in Fig. 2b with *P < 0.05.
Figure 4
Figure 4. Crystal structure of AmtB bound to PG
a, AmtB, in cartoon representation, is viewed from the periplasm and perpendicular to the membrane plane. Resolved PG molecules (spheres) are oriented with headgroups on the periplasmic side, in the outer leaflet of the inner membrane. b, PG molecules at subunit interface are shown with residue labels and hydrophobic residues in contact with lipid tails (stick representation). Hydrogen bonds (dashed lines) are formed between phospho headgroup and K84, and a water bridge (aqua sphere). c, Conformational change of (yellow arrows) residues 70-81, induced by binding PG. Superposition of AmtB structure without lipid (PDB 1U7G, purple) bound to PG (green).

Similar articles

See all similar articles

Cited by 187 articles

See all "Cited by" articles


    1. Singer SJ, Nicolson GL. The fluid mosaic model of the structure of cell membranes. Science. 1972;175:720–731. - PubMed
    1. Cantor RS. The influence of membrane lateral pressures on simple geometric models of protein conformational equilibria. Chem Phys Lipids. 1999;101:45–56. - PubMed
    1. Hunte C, Richers S. Lipids and membrane protein structures. Curr Opin Struct Biol. 2008;18:406–411. doi:10.1016/ - PubMed
    1. Lee AG. Biological membranes: the importance of molecular detail. Trends Biochem Sci. 2011;36:493–500. doi:10.1016/j.tibs.2011.06.007. - PubMed
    1. Sanders CR, Mittendorf KF. Tolerance to changes in membrane lipid composition as a selected trait of membrane proteins. Biochemistry. 2011;50:7858–7867. doi:10.1021/bi2011527. - PMC - PubMed

Publication types

MeSH terms

Associated data