Abstract
Lipid compositional asymmetry across the leaflets of the plasma membrane is an ubiquitous feature in eukaryotic cells. How this asymmetry is maintained is thought to be primarily controlled by active transport of lipids between leaflets. This strategy is facilitated by the fact that long-tail phospholipids and sphingolipids diffuse through the lipid bilayer slowly—taking many hours or days. However, a lipid like cholesterol—which is the most abundant lipid in the plasma membrane of animal cells—has been harder to pinpoint in terms of its favored side. In this work we show that, when a saturated lipid is added to a mix of the unsaturated lipid palmitoyl-oleoyl-phosphatidylcholine (POPC) and cholesterol, both cholesterol and the long-tail phospholipids organize asymmetrically across the membrane’s leaflets naturally. In these extruded unilamellar vesicles, most cholesterol as well as the saturated lipid—dipalmitoylphosphatidylcholine or sphingomyelin—segregated to the inner leaflet while POPC preferentially localized in the outer leaflet. This asymmetric arrangement generated a slight phospholipid number imbalance favoring the outer leaflet and thus opposite to where cholesterol and the saturated lipids preferentially partitioned. These results were obtained using magic-angle spinning nuclear magnetic resonance (MAS NMR) in combination with small-angle neutron scattering (SANS) using isotope labeling to differentiate lipid species. We suggest that sidedness in membranes can be driven by thermodynamic processes. In addition, our MAS NMR results show that the lower bound for cholesterol’s flip-flop half-time at 45°C is 10 ms, which is at least two orders of magnitude slower than current MD simulations predict. This result stands in stark contrast to previous work that suggested that cholesterol’s flip-flop half-time at 37°C has an upper bound of 10 ms.
Graphical abstract
Significance
Lipid membrane asymmetry is critical in cellular processes and results from a delicate balance between thermodynamics and active transport—the plasma membrane being a prime example. Understanding the role and consequences of lipid asymmetry in model systems has so far required to use challenging engineering protocols to drive lipid compositional asymmetry in unilamellar vesicles taking advantage of the slow motion of long-tail phospholipids across the lipid bilayer. Here, we demonstrate that equilibrium lipid composition asymmetry naturally exists in low-curvature membranes composed of three representative lipids—a saturated lipid, an unsaturated lipid, and cholesterol. These results will trigger a revision of the existing concepts of the origin of plasma membrane asymmetry as well as in bioengineering applications.
Introduction
The membranes of cells are a complex mixture of lipids and proteins. In the early 1970s the fluid mosaic model described lipid molecules as having a passive role: mostly to solvate its surrounding functional proteins (1). The structural complexity of multilipid multicomponent membranes quickly evolved as it was discovered that the plasma membrane (PM) of mammalian erythrocytes was asymmetric in lipid composition (2). By the early 1990s PM asymmetry, where phosphatidylserine (PS) and phosphatidylethanolamine (PE) are concentrated in the inner leaflet, and where phosphatidylcholine (PC) and sphingomyelin (SM) are concentrated in the outer leaflet, was firmly established (3). More recent studies have shown that almost all major classes of phospholipids are asymmetrically distributed across the PM leaflets (4). In addition to the PM, other cell organelle membranes are thought to be asymmetric, such as the membrane of the Golgi apparatus (5). However, not all cell membranes are asymmetric—the endoplasmic reticulum being a case in point (5). The process by which asymmetry is thought to be established and maintained is through ATP-dependent flip-flop activity of translocase proteins, such as ABC transporters (6). Although energy dependent, the process is less costly due to the slow spontaneous lipid randomization across leaflets by flip-flop—which takes hours or days—making it efficient from the point of view of the life cycle of lipids and cells (3). The physiological importance of this asymmetric lipid arrangement—as well as the transient loss of this asymmetry (7)—is still being uncovered. An example of how lipid asymmetry determines a physiological event is the asymmetry of PS across the PM whose location drives different signaling outcomes: when exposed on the external surface (opposite to where it is normally located) it triggers blood coagulation and macrophage cell disposal (8). A recent review by Doktorova et al. highlights the biophysical and functional consequences of membrane asymmetry (9).
However, the inner/outer leaflet location of cholesterol, which is the single most abundant lipid in the PM (∼40 mol %), remains ambiguous because different experimental approaches have produced different results. On the one hand, cholesterol has been found to be located in the outer facing leaflet of the PM, preferring a lipid environment provided by saturated lipids such as SM (4). Lorent et al. found that this location was likely favored by an imbalance in the phospholipid number between the inner and outer leaflets, which was found to be significant: a 2:1 ratio (4). This phospholipid number difference would likely drive cholesterol to the outer, phospholipid-depleted leaflet (4). And more recently, Doktorova et al. further explored the role of phospholipid composition and number imbalance in simulations, model membranes and cell systems and found that cholesterol is predominately in the saturated-rich but phospholipid-poor leaflet (10). Previous work by Buwaneka et al. supports this finding; using cholesterol sensors they showed that cholesterol is localized in the outer leaflet in multiple mammalian cells (11). The physiological importance of cholesterol asymmetry was highlighted by a lack of cholesterol asymmetry in the PM of cancer cells (12).
On the other hand, cholesterol, as well as fluorescent sterols that closely resemble cholesterol:dehydroergosterol, have instead been reported to be located on the inner cytosolic-facing leaflet (13,14). These studies suggest that this enrichment of the sterols in the inner leaflet of the PM is driven by either positive interactions with PE lipids or by exclusion from the outer leaflet by long-chain SM molecules. Since cholesterol is thought to rapidly diffuse through the lipid bilayer (seconds or less), cholesterol’s chemical equilibrium is sensibly what drives its distribution in this phospholipid/sphingolipid asymmetric environment (15,16).
Because of the complexity of biological membranes, simpler model membrane systems are used to examine interactions between components by controlling with precision their composition and environment. Free-standing membranes, typically vesicles, have been found to sustain manufactured lipid compositional asymmetries (17) using methodologies that have been recently reviewed by Scott et al. (18). Compositional asymmetry of cholesterol had not been achieved with these methods (19,20) until Courtney et al. (13) successfully generated an asymmetric distribution of cholesterol from symmetric vesicles using the β-cyclodextrin approach (18) at a very low temperature (0°C), where the asymmetry lasted at least an hour.
Courtney et al. also demonstrated that a natural cholesterol asymmetric distribution is attained in vesicles (100 nm in diameter) when the outer leaflet was manufactured to be enriched with very long saturated acyl chains lipids (C24 SM). Their work also demonstrated that, in erythrocytes, cholesterol preferentially settles in the inner leaflet driven by the very long saturated acyl chain lipids in the outer leaflet. However, at physiological temperatures (37°C), that asymmetry was lost in both model vesicles and erythrocytes (13).
Although MD simulations show that cholesterol flip-flop rates between leaflet is in the microsecond range (21,22,23), recent MD studies of cholesterol’s distribution in prebuilt asymmetric membranes—where the lipid composition or both lipid composition and number of phospholipids in each leaflet is not the same—show that cholesterol can distribute asymmetrically (10,23). Of these, the most striking cholesterol distribution comes from a phospholipid number-enriched cytosolic-like leaflet having unsaturated lipids against a phospholipid number-depleted but saturated lipid-rich exoplasmic-like leaflet where cholesterol is mostly found (10). This asymmetry revealed a more impermeable membrane due to the tight packing induced by cholesterol in one leaflet, but also highly fluid and prone to forming defects in the phospholipid number-rich but cholesterol-poor leaflet. These predictions were verified in cells where it was observed that the disruption of PM lipid asymmetry produced both higher permeability as well as a relocation of peripheral proteins away from the PM into the cytosol environment (10). Recent theoretical and simulation work by Varma and Deserno (16), showed that an asymmetric distribution of cholesterol between membrane leaflets can result from a chemical potential balance of two opposing effects: differential stress/differential lipid number between leaflets as well as a cholesterol partitioning bias resulting from a preference of cholesterol for a certain membrane composition occurring in one leaflet and not the other (a Flory-Huggins mixing free energy). Although these authors do not treat curvature specifically in their work, curvature can certainly be thought of as a driver for differential stress (24) and therefore the ideas set forth by their work are applicable in vesicles.
Here, we demonstrate, using magic-angle spinning (MAS) NMR and small-angle neutron scattering (SANS), that large unilamellar vesicles (LUVs) having a mixture of saturated and unsaturated lipids and cholesterol, can sustain a naturally asymmetric distribution of cholesterol as well as the other lipid components. We found that cholesterol and the saturated lipid—in this case DPPC or SM—preferentially segregate to the inner leaflet while the unsaturated lipid partitions to the outer leaflet. Furthermore, a slight phospholipid imbalance favoring the unsaturated lipid-rich outer leaflet is detected. Surprisingly, this lipid arrangement is observed at 45°C where the membrane is in the fluid phase and in thermodynamic equilibrium.
In addition, our MAS NMR results on symmetric vesicles made of unsaturated lipids (mostly POPC) and cholesterol, establish that cholesterol’s flip-flop half-time must be greater than 10 ms. This lower bound of 10 ms is two orders of magnitude slower than what all-atom (21,22) and coarse-grained (23) simulations predict. This result stands in stark contrast to previous work by Bruckner et al. who had reported that cholesterol’s flip-flop half-time at 37°C had an upper bound of 10 ms (25).
Materials and methods
Lipids and chemicals
Cholesterol (bovine), palmitoyl oleoyl phosphatidyl-choline (POPC), d31 palmitoyl oleoyl phosphatidyl-choline (dPOPC), palmitoyl oleoyl phosphatidyl-glycerol (POPG), d31 palmitoyl oleoyl phosphatidyl-glycerol (dPOPG), di-palmitoylphosphatidylcholine (DPPC), d62 di-palmitoylphosphatidylcholine (dDPPC), and egg SM were purchased from Avanti Polar Lipids (Alabaster, AL) and used without further purification.
13C40-Di-palmitoylphosphatidylcholine (13C DPPC), 13C-3,4-cholesterol (13C cholesterol), and D2O (99.9% purity) were purchased from Cambridge Isotopes (Cambridge, MA) and used without further purification. 13C-skip-Cholesterol was provided by Prof. Chad M. Rienstra (National NMR Facility at Madison [NMR FAM], at the University of Wisconsin-Madison) (26).
Milli-Q H2O was obtained from a MilliporeSigma (Burlington, MA) ultrapurification water system having an 18.2 MΩ·cm resistivity. Chloroform and methanol (high-performance liquid chromatography [HPLC] grade) were purchased from Fisher Scientific (Waltham, MA). Manganese(II) sulfate monohydrate was purchased from Sigma-Aldrich (Burlington, MA).
Sample preparation
Lipids in powder form were weighed according to the desired lipid ratios (see Table S2) into a glass vial and thoroughly mixed using chloroform or chloroform/methanol mixtures (2:1) and vortexed until the solution was clear. The removal of this chloroform was performed, initially, with a stream of N2 gas using a nitrogen evaporation system (N-EVAP) until a thin in-appearance dry lipid film remained on the vial’s walls. The vials were then placed in vacuum at 60°C overnight to remove any residual organic solvent. Once dry, the samples were hydrated in D2O for MAS NMR measurements, and in mixtures of D2O and H2O for SANS measurements, where the aqueous solvent was slightly off from the mean scattering length density (SLD) of the membranes (Δ = +0.3 Å−2) (see details in Table S1). The lipid concentration targeted was 50 mg/mL.
LUVs with 5 mol % of charged lipids (POPG or dPOPG) were prepared by extrusion (27). The extrusion was performed with a modified Avanti Polar Lipids extruder system that included heating via an external water bath. A syringe pump mechanically manipulated the syringes of the extruder (New Era Pump Systems, Farmingdale, NY). The final extrusion consisted of 41 passes through a polycarbonate filter with a 100 nm average pore size. The extrusion of the vesicles was performed at 50°C, which is well above the melting temperature of the lipid systems studied. The speed of extrusion can be controlled on the syringe pump, which allows extruding high-concentration solutions (see supporting material, section 1 in Krzyzanowski et al. (28) for more details). For MAS NMR measurements the vesicle solution was concentrated to ∼150 mg/mL (∼200 mM) using a tabletop Eppendorf 5424 Centrifuge at a speed of 10k rcf and Amicon Ultra-0.5 Centrifugal filters (pore size 100 kDa NMWCO).
MAS NMR
MAS NMR data were obtained at the NMR FAM, University of Wisconsin-Madison.
Nineteen microliters (∼3 mg) of a concentrated vesicle solution (∼150–200 mM) plus an additional 1 μL of D2O was packed into a 3.2 mm MAS rotor using a specialized rotor packing device. To determine the lipid distribution across the leaflets of the vesicles’ membrane we used the well-established technique of exposing the vesicles to a paramagnetic salt. We chose the paramagnetic salt Mn2+ at a 1:570–760 ion/lipid ratio. This ratio was determined using 1H NMR and testing several Mn2+ concentrations on a POPC vesicles solution to determine at what point the ratio of lipids to Mn2+ deviated from the expected drop in intensity by half, i.e., the full quenching of the outer leaflet choline signal. We used a Mn2+ concentration that was double the lowest we found. In the literature a 1:1200 ratio of Mn2+/lipid was determined to be sufficient to quench the NMR signal from the outer leaflet lipids (29,30). For the MAS NMR measurements with salt, we used 19 μL of the concentrated lipid vesicle solution plus 1 μL of a 5 mM MnSO4 solution in D2O. We validated the lipid/Mn2+ ratio by MAS NMR observing a consistent 50% decrease in the intensity of the phosphate and 13C cholesterol peaks in POPC/cholesterol vesicles (shown in Fig. 1).
Figure 1.
MAS NMR data for ∼150 mM 100 nm diameter POPC vesicle solution having 50 mol % 13C-labeled cholesterol (∼3 mg of lipids). Black: no Mn2+; red: in the presence of Mn2+.; scale bar 100nm. Selected regions of 1D 13C MAS NMR spectra for carbon 3—as shown in the schematic with a red dot—identify ∼50% of cholesterol on the outer leaflet determined from the 50% peak quenching due to Mn2+. In addition, the 31P MAS NMR confirmed the unilamellarity and symmetry of the phosphate composition of the leaflets since 50% of the phosphate signal was quenched with Mn2+ as well. The Mn2+ ion/lipid ratio was ∼1:750.
MAS NMR experiments were performed on a Bruker 600 MHz spectrometer with a 3.2 mm triple-resonance Phoenix MAS probe configured for 1H-31P-13C. The sample spinning rate was maintained at 10 kHz, with a variable temperature set point at 45°C.
In the acquisition of 1D-31P NMR spectra, proton decoupling was applied. The parameters set for these experiments were a 31P 90° pulse length of 5 μs, acquisition time of 50 ms, recycle delay of 3 s, and a total of 1024 scans. Proton decoupling was carried out with an RF field strength of 30 kHz using the SPINAL64 sequence (31).
For the 1D-13C NMR spectra, the Hahn echo method was employed with the following parameters: a 13C 90° pulse of 4.2 μs, acquisition time of 40 ms, echo time of 100 μs, and a total of 2048 scans. Proton decoupling was performed using an RF field strength of 65 kHz with the SPINAL64 sequence (31).
Analysis of the peaks of interest presented in Figs. 1 and 2 as well as Fig. S1, A and B were done by comparing MAS NMR spectra from aliquots of the same sample, one having no salt and one with Mn2+. The scale factor applied to the spectra with salt so that it visually overlapped with the spectra without salt provided the amount of quenching produced by Mn2+, as reported in Table 1.
Figure 2.
Selected regions of 1D 13C MAS NMR spectra for 100 nm vesicles (∼150 mM) having 30 mol % skip-labeled 13C-labeled cholesterol (26) and a 1:4 13C-labeled DPPC/POPC ratio (∼3 mg of lipids). Black: no Mn2+; red: with Mn2+ (∼1:750 ions:lipids). The 1D 13C MAS NMR spectra identify ∼90% of 13C cholesterol (carbon 3 as shown in Fig. 1) in the inner leaflet as well as 65% of 13C DPPC in the inner leaflet, as determined from peak quenching due to Mn2+. The 31P MAS NMR shows the phosphate composition of the leaflets indicating that 55% of the phospholipids are in the outer leaflet as 45% of the original 31P spectra remains. The full 13C spectra is shown in Fig. S1A.
Table 1.
Inner leaflet percent partitioning for each lipid relative to their total in the membrane
POPC + cholesterol | ||||
100 nm LUVs | cholesterol | POPC + POPG | – | – |
MAS NMR | 50 ± 10 | 50 ± 10 | – | – |
SANS | 50 ± 5 | 50 ± 5 | – | – |
POPC + DPPC + cholesterol | ||||
100 nm LUVs | cholesterol | DPPC | POPC + POPG | DPPC + POPC + POPG |
MAS NMR | 90 ± 10 | 65 ± 10 | – | 45 ± 10 |
SANS | 70 ± 5 | 64 ± 5 | 39 ± 5 | 46 ± 10a |
POPC + SM + cholesterol | ||||
– | cholesterol | SM | POPC + POPG | SM + POPC + POPG |
MAS NMR | 76 ± 10 | – | – | 45 ± 10 |
SANS | 76b | 68 ± 5c | 37 ± 5 | 45b |
Calculated using (6) in materials and methods.
Taken from MAS NMR.
Calculated using Eqs. 5 and 6 in materials and methods.
SANS
The measurements were performed on the SANS D22 instrument at the Institut Laue-Langevin (ILL) (Grenoble, France), https://doi.org/10.5291/ILL-DATA.9-13-428. Instrument configuration covered a Q range of 0.002 ≤ Q ≤ 0.5 Å−1. The wavelength used was 6 Å with a wavelength resolution of Δλ/λ = 10%.
Data were collected on 2D detectors and reduced using GRASP (32), the reduction package provided by the ILL. The solution was loaded into 1 mm thick, banjo quartz Hellma cells. Temperature was controlled by water bath heating the entire sample environment stage to 45°C.
SANS analysis was performed with SASview (33). The model used was that of the core-multishell model modified for the correct scattering volume for vesicles. We used five shells corresponding to headgroups, tails, and a small methyl region. A low-Q correlation peak in our scattering curves was produced by charge-induced interactions between vesicles. We modeled this interaction with the Hayter-Penfold rescaled mean spherical approximation structure factor.
Data from samples having the same mean SLD for the membrane were fitted simultaneously. The mean SLD for the membrane was given by:
(1) |
The index corresponds to each lipid in the composition of the membrane. corresponds to the nominal molar amount of the corresponding lipid. The nominal molar fractions are given in Table S2. The corresponds to the scattering length of a given lipid and are obtained from their atomic and isotopic composition. The volume, , for each lipid at 45°C was obtained from known published values (28,34,35). In the second equivalent expression, corresponds to the volume fraction of lipid in the mixture. Indeed, the SLD for any given lipid is:
(2) |
Similarly, the SLD of the solvent is given by:
(3) |
where corresponds to the volume fraction of either D2O or H2O and where and correspond to the SLDs of D2O and H2O, respectively. The SLD for the solvent at 45°C was obtained according to Krzyzanowski et al. (28).
Fitting proceeded by having the bilayer dimensions: headgroup, tails, and methyl region, and the bilayer’s SLDs for the headgroup and methyl region constrained to be the same for samples sharing the same mean SLD. The SLDs for the tail region for each leaflet were free parameters except for the symmetric system for which the inner and outer shells were constrained to have the same SLD. In addition to obtaining scattering curves for the concentrated lipid vesicle solutions, we acquired spectra for diluted vesicle solutions in D2O for some of the cases (shown in Fig. S2). The dilution factor was ∼10. The values of the profile fit parameters shown in Figs. 3, B, D and 4 B are listed in the Table S1.
Figure 3.
SANS spectra for POPC/cholesterol LUVs with and without DPPC (A and C) and corresponding SLD profiles (B and D) (see Tables S1, A and B for fit details). The lines through the data in (A) and (C) correspond to the calculated scattering obtained from the SLD profiles presented in (B) and (D). (A) SANS spectra for cholesterol/dPOPC LUVs without (blue symbols) and with (black symbols) DPPC (1:1 mixture hDPPC/dDPPC), highlighting cholesterol. (C) SANS spectra for cholesterol/hPOPC with dDPPC (red symbols), and cholesterol/hDPPC with dPOPC (black symbols); blue symbols correspond to LUVs with cholesterol and a ∼1:2.5 hPOPC/dPOPC mixture with no DPPC. The fraction of cholesterol in the all cases was ∼30–35 mol %, and for LUVs with DPPC, the DPPC content was ∼17–25 mol % (sample composition details are found in the Table S2).
Figure 4.
(A) SANS spectra for LUVs composed of POPC/cholesterol (blue symbols), DPPC/POPC/cholesterol (black symbols) and SM/POPC/cholesterol (red symbols). The spectra for POPC/cholesterol and DPPC/POPC/cholesterol are also presented in Fig. 3C. Lines through the data are calculated scattering curves obtained from the corresponding SLD profiles presented in (B) and Fig. 3D. (B) SLD profiles obtained from fitting the SANS spectra for membranes with either SM or DPPC; the latter is also presented in Fig. 3D (see Table S1B for fit details). (C) Selected regions of 1D 13C MAS NMR spectra for LUVs having 30 mol % 13C-labeled cholesterol and a 1:2 SM/POPC ratio. Black: with no Mn2+; red, in the presence of Mn2+ (∼1:750 ions:lipids). The MAS NMR spectra identifies ∼76% of 13C cholesterol (carbon 3 as shown in Fig. 1) in the inner leaflet determined from peak quenching due to Mn2+. The 31P MAS NMR spectra shows the phosphate composition of the leaflets indicating that 55% of the phospholipids are in the outer leaflet as 45% of the original 31P spectra remains.
The total fraction of lipids in each leaflet as presented in Table 1 was calculated as follows:
(4) |
where the subindex refers to the lipid being highlighted according to the specific deuteration scheme. The subindex , corresponds to all other lipid species in the membrane. corresponds to the SLD of leaflet —outer or inner—obtained from the fit, as shown in Table S1. corresponds to the total mole fraction of lipid A in leaflet . Indeed, within the error of the SLD profile.
The partitioning of each lipid in the membrane had to satisfy an outer to inner leaflet volume ratio of ∼1:
(5) |
Because exchanging indices 1 for 2 and vice versa between the numerator and denominator does not identify the outer versus inner leaflet, SANS alone cannot identify the inward/outward leaflets.
The only way to resolve this was with complementary information from MAS NMR. Using the fraction of the 31P signal after the addition of Mn2+, we were able to establish the fraction of phospholipids in the inner leaflet, (as reported in Table 1) and that can be used to identify which leaflet corresponds to the outer/inner leaflet as follows:
(6) |
The leaflet j (1 or 2) that satisfies is identified as the inner leaflet.
HPLC-ELSD
Simultaneous determination of total phospholipid and cholesterol content in LUVs by HPLC
To quantify the different lipid classes within the LUVs, a high-performance liquid chromatography-evaporative light scattering detector (HPLC-ELSD) method was performed. Separation was carried out on an Agilent chromatographic system (1260 Infinity II series, Agilent Technologies, Les Ulix Cedex, France) coupled to a SEDEX 90 ELSD detector (Sedex Sedere, Olivet, France). Standards and experimental samples were taken to dryness under an argon (Ar) stream and the residue dissolved in 1 mL, CHCl3/CH3OH (9:1), and injected through a semipreparative diol-modified silica stationary phase (Nucleosil 100-5 OH column (10 × 250 mm), Macherey-Nagel, Hoerdt, France). The column was eluted with a binary gradient system consisting of solvent mixture A: (CHCl3/CH3OH [70:25, v/v] plus 1% NH4OH) and solvent mixture B: (CHCl3/CH3OH/Milli-Q H2O [60:40:5.5, v/v/v] plus 0.5% NH4OH), at a flow rate of 1.0 mL/min throughout the run. A method was programmed such that at time 0 min, the proportion of solvent mixture B was 5%, gradually increased to 100% at 60 min, and stayed at this value until 85 min after which it was again decreased to 5% at 86 min and kept there until the end of the gradient at 96 min. During all the measurements, the column temperature was set to room temperature, the autosampler module at 5°C and the injection volume was set to 50 μL. The ELSD nebulization temperature was set at 60°C and N2 was used as a carrier gas at 3.5 psi inlet pressure. Data were analyzed by the Open Lab chromatographic workstation (Agilent Technologies, Les Ulix Cedex, France).
HPLC method validation
Separation of cholesterol and PC was achieved in approximately 96 min and their identification was based on the retention time of known standards (0.4 μg/μL). A standard mixture composed of cholesterol/PC/PG (1:1:1) with concentrations ranging between 20 and 325 μg each were injected for HPLC analysis. The method employed was validated by assessing the linearity, limit of detection and reproducibility through injecting standard solutions in CHCl3 (of wide-ranging concentrations), that in addition allowed to establish calibration curves and as a reference check. The precision of the concentration determination was tested by repetitive injections of the same standard mixture of a given concentration. The HPLC system was flushed with 100% 2-propanol for 10 min between each run to eliminate any carryover. The ELSD response, expressed as peak areas as a function of the analyzed lipid mass, was quantified against the calibration curves. See Table S2 for the corresponding composition results.
FAME analysis by GC-FID
The fatty acyl chain composition, both hydrogenous and deuterated in the LUVs, was determined by gas chromatography-flame ionization detection (GC-FID) (GC 2010 Plus, Shimadzu, Noisiel, France), post their methanolysis thus allowing for the release of the fatty acids as their corresponding fatty acid methyl esters (FAMEs). To carry out GC-FID measurements, the samples were first derivatized upon adding about 3 mL of methanolic-HCl to screw cap glass vials containing approximately 1 mg of the lipid mixtures each. The vials were then bubbled with argon, vortexed, and sealed tightly with a Teflon-lined cap and incubated at 95°C for 1 h. After cooling down the vials to room temperature (≈10 min), 3 mL of H2O was added and the solution vortexed, following which 3 mL of C6H14 was added and again vortexed vigorously to create an emulsion. Slow centrifugation (≈500 × g for 5 min) at room temperature was carried out to break the emulsion and produce an upper C6H14-rich phase containing the FAMEs. The supernatant phase of ≈2.8 mL was then transferred into a fresh vial, evaporated under a stream of Ar and the resulting dried film was reconstituted in 50 μL C6H14 and transferred into a GC autosampler vial that was then loaded into the GC’s automatic liquid sampler. The GC instrument (GC 2010 Plus, Shimadzu, Noisiel, France) was equipped with a split/splitless injector and an SGE BPX70 70% cyanopropyl polysilphenylene-siloxane column (25 m × 0.22 mm ID and 0.25 μm film thickness). Helium (He) was used as a carrier gas at a flow rate of 1.04 mL/min with a linear velocity of 35 cm/s and a purge flow rate of 1 mL/min. The column was allowed to equilibrate for 3 min at 155°C before injection and then the temperature was ramped up to 180°C at a rate of 2°C/min and then to 220°C at a rate of 4°C/min and finally held at 220°C for 5 min resulting in a 27.5-min total run time. Samples (8 μL) were injected into the column at 250°C using an AOC-20i autoinjector. Detection was through an FID operating at 260°C with 40 mL/min H2, 400 mL/min compressed air and 30 mL/min He make-up flow. Data analyses of the FAMEs obtained from the lipid mixtures were carried out by LabSolutions software (Shimadzu, Noisiel, France), which helped assign and integrate the total ion chromatogram peaks from which the total mole fraction amount of each FAME was obtained. See Table S2 for the corresponding composition results.
Cryoelectron microscopy
Samples were prepared for cryo-TEM imaging by application of 3.5 μL of sample onto plasma-cleaned (Gatan Solarus. Pleasanton, CA) Quantifoil grids (1.2/1.3, copper 200 mesh; EMS, Hatfield, PA) and subsequent plunge freezing in liquid ethane using the Vitrobot Mark IV (Thermo Scientific, Hillsboro, OR). Grids were stored in liquid nitrogen until use. Grids were imaged on either the Thermo Scientific Glacios (200 kV) with the Ceta-D camera or Thermo Scientific Titan Krios 3Gi (300 kV) with the K3 direct detector (Gatan) using EPU software (Thermo Scientific).
Results
Cholesterol is symmetrically distributed in POPC membranes
To directly assess the distribution of cholesterol within the membrane of unilamellar vesicles, we used MAS NMR coupled with paramagnetic agents since cholesterol’s signal does not appear in traditional solution NMR due to the slower molecular motion of cholesterol in the lipid bilayer (36). Paramagnetic ions, while remaining impermeable to the membrane, can broaden or quench the resonance peaks from the chemical groups they interact with due to paramagnetic relaxation (37). This interaction is distance dependent and acts mostly on the headgroup region of the outer leaflet lipids (30,38). We chose the paramagnetic salt Mn2+ at a ∼1:750 ion/lipid ratio to quench the resonances from groups as far as the phosphate region of the lipids’ headgroup in the outer leaflet of the vesicles’ membrane (29,30). Comparing MAS NMR spectra from identical aliquots of the same sample, one having no salt and one with Mn2+, provides enough information to determine the distribution of lipids in the vesicles’ membrane (29,30). Fig. 1 shows the MAS NMR spectra, taken at 45°C, from 100 nm diameter POPC vesicles containing 50 mol % of 13C-labeled cholesterol. The 13C peak shown in the figure corresponds to carbon 3 as depicted in the schematic of cholesterol’s structure in Fig. 1. Carbon 3 is the closest carbon to the OH moiety and therefore closest to the aqueous medium and most susceptible to Mn2+ quenching. Also, in Fig. 1 we show the 31P spectra of the phosphate group of the phospholipids in the vesicles. The 31P peak for the phospholipids and the 13C peak for cholesterol describe a symmetric distribution for both lipids because after adding Mn2+ these peaks drop by ∼50% compared with the no salt case. Indeed, this 50% attenuation confirms the unilamellarity of the LUVs. Cryo-TEM images of the vesicles after the MAS NMR measurements (diluted to ∼10 mM) also revealed the unilamellarity of the vesicles. The distribution of nested vesicles (an example shown in Fig. 1) versus single unilamellar vesicles remained marginal—<10%, as expected (27)—with and without Mn2+.
Furthermore, the permanence of the MAS NMR peaks after the addition of Mn2+ indicates that the lipid residence time in a leaflet must be longer than 10 ms, which is the set timescale of the NMR measurement. Indeed, if this leaflet residence was shorter, i.e., the flip-flop time was faster, lipids from both leaflets would interact with Mn2+ resulting in the complete quenching of their signal. Such behavior has been observed for cholesterol by Bruckner et al. (25) in very small, sonicated vesicles. However, here we find that in extruded LUVs the signal from cholesterol remains at half of its value after Mn2+ addition. Therefore, the takeaway message from these measurements is that cholesterol’s flip-flop between leaflets is in a timescale slower than 10 ms. The reported flip-flop half-time for cholesterol varies by several orders of magnitude at physiological temperatures; from submilliseconds in MD simulations (21,22) to under 10 ms (25), to a few minutes (39), up to hundreds of minutes to hours (40,41). The data presented in Fig. 1 set a solid lower bound for cholesterol’s flip-flop half-time at greater than 10 ms in LUVs at 45°C.
Cholesterol, DPPC, and POPC distribute asymmetrically between membrane leaflets
Next, we measured vesicles (100 nm in diameter) containing skip-labeled 13C cholesterol (26), POPC, and fully labeled 13C DPPC (with 5 mol % POPG (27)). The concentration of cholesterol was 30 mol % and the ratio of DPPC/POPC was ∼1:4. The measurements were taken at 45°C where the membrane is known to be fluid (42). As shown in Figs. 2 and S1 A, we surprisingly found that 90% of the total cholesterol was in the inner leaflet since 90% of the signal persisted after the addition of Mn2+. Although this fraction is high, for membranes with a 30 mol % total cholesterol this results in a 54 mol % concentration of cholesterol in the inner leaflet, which is below the solubility limit of cholesterol in PC membranes (43,44). Furthermore, we found that ∼65% of DPPC was in the inner leaflet since ∼65% of the 13C signal for DPPC’s carbon closest to the phosphate environment remained after the addition of Mn2+. To verify the asymmetric distribution of cholesterol in this system we measured vesicles where only cholesterol was 13C labeled (skip-labeled 13C (26)). As shown in Fig. S1 B, the signal for cholesterol was again quenched by only 10% after the addition of Mn2+. In both cases (with 13C-labeled DPPC and with unlabeled DPPC) the phosphate 31P signal was quenched by ∼55%, indicating that there is a slightly higher fraction of phospholipids in the outer leaflet. A 55% attenuation of the 31P spectra also confirms the unilamellarity of the LUVs with a near symmetric distribution of phospholipids across both leaflets.
To support these results, we turned to SANS, which in recent years has been effective in detecting lipid compositional asymmetry in LUVs (50–100 nm in diameter) (45,46,47,48,49,50). Detecting lipid compositional asymmetry across the leaflets of membranes in vesicles using SANS requires that the membrane be composed of deuterated (d) and hydrogenated (h) lipids and that these acquire an asymmetric distribution between leaflets—generating internal leaflet contrast. As with 13C labeling, deuteration preserves the chemical identity of the lipids and thus are expected to preserve their physical properties albeit with slight shifts; for example, the melting temperature of hydrogenated DPPC (hDPPC) is 41°C while for tail deuterated DPPC (dDPPC) it is 37°C (51).
Thus, we set out to detect the asymmetric distribution of lipids and cholesterol in unilamellar vesicles with SANS having a similar composition as the ones studied by MAS NMR. To do so we chose deuteration schemes that highlighted only one lipid component at a time. To highlight cholesterol, we used a deuteration scheme in which all phospholipids became indistinguishable from each other. Fig. 3 A shows the SANS spectra for vesicles composed of cholesterol (∼35 mol %), deuterated POPC (dPOPC) and 5 mol % of deuterated POPG (dPOPG). These lipids only have their palmitoyl tail deuterated. This figure also shows the SANS spectra for vesicles that have, additionally, a 1:1 mixture of hDPPC and dDPPC, the latter having both palmitoyl tails deuterated. This mixture of deuterated and perdeuterated DPPC has an equivalent SLD to that of dPOPC (see Eqs. 1 and 2 in materials and methods). Thus POPC, POPG, and DPPC became indistinguishable from each other. The ratio of DPPC/POPC was ∼1:3 and the cholesterol fraction was ∼36 mol % (see Table S2 A for composition details). To highlight the bilayer region in the spectra, the solvent was chosen to be slightly off the mean SLD of these membranes (Δ = +0.3Å−2). In both cases (with or without DPPC), the resulting contrast is between cholesterol and its surrounding phospholipid membrane. We observed that the vesicles having POPC and cholesterol show a sharp minimum at Q ∼0.08 Å−1, while the vesicles containing additional DPPC have a clear increase in intensity in this Q region. Fits to the data, shown as continuous curves, were obtained by modeling the vesicle with inner and outer headgroup regions, tail regions for each leaflet, and a contribution from the relatively small methyl group region in the middle of the bilayer. The overall quality of the fits (χ2 = 14.3 and χ2 = 16.3 for the symmetric and asymmetric cases, respectively, as shown in Table S1 A) is satisfactory up to a Q ∼0.2 Å−1. The model captures the differences in cholesterol composition between leaflets in the Q range 0.2 Å−1 ≥ Q ≥ 0.05 Å−1. A discrepancy between the model and the data arises for Q values greater than 0.2 Å−1, possibly reflecting some form of inner structure within each leaflet which is not accounted for in our model. The results in SLD profiles indicate that cholesterol is symmetrically distributed in POPC/cholesterol vesicles and asymmetrically distributed in DPPC/POPC/cholesterol vesicles, as shown in Fig. 3 B. The partitioning of cholesterol deduced from the SLD of each leaflet corresponds to a distribution where ∼70% of the total cholesterol is in one leaflet of the membrane (see Table S1 A for SLD profile value details). This asymmetric distribution of cholesterol is in agreement with what was found by MAS NMR, as detailed in Table 1.
While Fig. 3 A highlights cholesterol asymmetry, Fig. 3 C highlights DPPC asymmetry as well as POPC asymmetry. To highlight these lipids in mixtures of DPPC/POPC/cholesterol, we used their respective deuterated versions: dDPPC/hPOPC/cholesterol to highlight DPPC and hDPPC/dPOPC/cholesterol to highlight POPC (see Table S2 B for composition details). Also in the figure is the spectra for the system without DPPC. The mean SLD of the membranes was the same for the three cases and the solvent’s SLD was chosen to highlight the bilayer region by setting it slightly higher than the membranes’ mean SLD (Δ = +0.3 Å−2).
As shown in Fig. 3 C, the scattered intensity for the cholesterol/POPC system, having no DPPC, drops to the background value of 1 × 10−3 cm−1 for Q values of 0.1 Å−1 and greater. The fit to the data indicates that the system is symmetric in the distribution of all membrane components, as shown in the corresponding SLD profile in Fig. 3 D. The scattered intensity for the system highlighting DPPC (dDPPC), on the other hand, shows an enhanced signal compared with the scattering from the symmetric system in the mid/high Q values. This indicates that deuterated and hydrogenated lipids mostly populate opposing leaflets. The partitioning of DPPC deduced from the SLD of each leaflet (see Table S1 B for SLD profile details) was that 36% is in one leaflet, while 64% is in the opposite leaflet. This quantitative partitioning is similar to what was found by MAS NMR, where 65% of DPPC was found in the inner leaflet. Similarly, the scattering intensity for the system highlighting POPC also shows enhanced scattering compared with the scattering from a symmetric system in the mid/high Q range. Quantifying the partitioning of dPOPC (+dPOPG) using the SLD of each leaflet, we find that there is a 38–62% partitioning between leaflets.
Despite the fact that SANS can clearly detect and quantify lipid asymmetry in membranes, it cannot assign it to a specific leaflet due to the loss of the phase of the scattered amplitude; the scattered signal is unchanged by mirroring (out-to-in) the leaflets’ SLD values. For the systems shown in Fig. 3, the particular lipid location was resolved using the results from MAS NMR, which had identified DPPC and cholesterol in the inner leaflet. The additional requirement that the ratio of the inner to outer leaflet volumes be ∼1 (Eq. 5 in materials and methods) implied that POPC preferentially partitions in the outer leaflet. Indeed, with this arrangement it was possible to verify that the phospholipid number fraction in the outer leaflet was ∼55% as measured by MAS NMR. These results are summarized in Table 1, and most importantly highlight that both MAS NMR and SANS robustly find similar asymmetric partitioning of all three lipids.
Cholesterol, SM, and POPC distribute asymmetrically between membrane leaflets
Because SM is a more representative saturated lipid of the PM (4,5) and known to have a preferential interaction with cholesterol (52,53), we replaced DPPC with SM. Interestingly, as shown in Fig. 4 A, we found that the scattering spectra for cholesterol/hSM/dPOPC is very similar to cholesterol/hDPPC/dPOPC. Quantifying the partition of POPC between leaflets in the SM containing membrane from the SLD profile shown in Fig. 4 B (see Table S1 B for SLD profile details), we found that there is a very small increase in POPC segregation compared with the DPPC system: with 37% in one leaflet and 63% in the opposite leaflet. This finding is consistent with the slight but higher intensity in the mid/high Q found for the SM containing membranes compared with the system with DPPC, as shown in Fig. 4 A. We also performed MAS NMR on LUVs having an equivalent composition to that studied by SANS: SM, 13C cholesterol, and POPC (and 5 mol % POPG). As shown in Fig. 4 C, we found that cholesterol segregated to the inner leaflet with ∼76% of the carbon 3 13C signal remaining after the addition of Mn2+. The phosphate 31P signal indicated that ∼ 55% of the phospholipids are in the outer leaflet, like the membranes containing DPPC. Using the fraction and preference of cholesterol for the inner leaflet revealed by MAS NMR and the partitioning of POPC derived from the SANS analysis and by requiring the outer to inner leaflet volume ratio be ∼1 and by setting the phospholipid number fraction in the outer leaflet be 55%, as was found by MAS NMR, we were able to deduce that SM partitions preferentially to the inner leaflet (68%). This fraction is slightly higher than was found for DPPC. The results are summarized in Table 1, where we highlight how SANS and MAS NMR can be combined to obtain the partitioning of all lipids in the membrane.
Discussion
In the 1970s de Kruijff, Cullis, and Radda (54,55,56), using solution NMR and paramagnetic agents, were able to identify lipid asymmetry in really small (<30 nm in diameter) model vesicles consisting of a mixture of a few lipids. They report lipid asymmetry in saturated and unsaturated lipid mixtures, mixtures having different headgroup compositions, as well as mixtures that included cholesterol (38). However, the location of cholesterol—which was not independently detected—was inferred from the analysis of the phospholipid 31P peaks. These studies suggested that cholesterol (above 30 mol %) preferred the inner leaflet (54,55,57). On the other hand, saturated lipids, when mixed with unsaturated lipids, prefer to localize in the outer leaflet (29). Furthermore, an asymmetric distribution of lipids was found to be strongly dependent on curvature (38,58). The vesicles studied by de Kruijff and co-workers were sonicated and thus really small, the largest ones being less than 30 nm in diameter (38). Indeed these vesicles had a relatively large difference in the number of lipids found in the inner leaflet compared with the outer leaflet (∼1:2) inducing significant lateral stress, as reported by Nakano et al. (59).
By the 1980s the method of extrusion emerged, whereby an aqueous suspension of multilamellar lipid vesicles was sieved through a polymeric membrane having a submicron mesh (60,61,62). The size of the mesh defined the vesicle diameter and larger unilamellar vesicles were created. The extrusion method is now widely used in part because these larger vesicles behave symmetrically between the outer and inner leaflets with regard to lipid partitioning (63) as well as lateral stress, as recently confirmed by Nakano et al. (59), and thus behaving like planar membranes. Furthermore, the thermotropic behavior of lipids in extruded vesicles is conserved; this latter property is significantly altered in sonicated vesicles (59).
The planar properties of these large extruded vesicles were assumed to not produce leaflet lipid asymmetry as had been observed in small vesicles (20). Indeed, these ideas then propelled the effort of manipulating the lipid composition of vesicles by a lipid exchange vehicle (18,45,64) to generate lipid composition asymmetry across the membrane’s leaflets. As shown here, however, equilibrium lipid asymmetry across membrane leaflets is established in LUVs containing a saturated lipid, an unsaturated lipid, and cholesterol. While the asymmetry for saturated lipids in these vesicles is not as pronounced as what is found in PMs (4), for cholesterol the asymmetry between leaflets is substantial and similar to what has been reported in PMs (∼1:9 (11,12)). The role of saturated lipids is highlighted by the fact that, when the membrane only contains cholesterol and an unsaturated lipid (POPC), the distribution of both components is symmetrical, as detected by both MAS NMR and SANS, or only very slightly asymmetric as detected by SAXS (63).
In the context of the lipid distribution found in eukaryotic cells there is evidence for saturated lipids localizing in the outer leaflet (Lorent et al. (4) and Doktorova et al. (10) being the most recent). Cholesterol, however, is controversial. Although there is significant evidence that cholesterol is located in the outer leaflet (9,11,12), Courtney et al. (13), using a cyclodextrin approach, found that cholesterol’s preference is for the inner leaflet of the PM in erythrocytes (albeit at low temperature). Similarly, Solanko et al. (14) found that dehydroergosterol (DHE)—a fluorescent cholesterol analog—locates in the inner leaflet of yeast cells, although self-quenching effects for DHE may be at play as Doktorova et al. (10) have recently pointed out.
Our findings are in general agreement with those by Doktorova et al. (10) whereby cholesterol favors one leaflet—that with saturated lipids—and where the phospholipid content is reduced. There are two things that stand out, however, the first is that we observe this significant asymmetric lipid arrangement across leaflets in equilibrium, in the fluid phase—a process thought to necessitate active lipid transporters, particularly for long phospholipids (65)—and the second is that we do not find that a large phospholipid number difference between leaflets is required for the emergence of a significant asymmetric partitioning of cholesterol between leaflets– which is the current hypothesis for cholesterol asymmetry in the PM (4,9,10).
The fact that compositional asymmetry builds naturally across leaflets for all lipid components in the membrane, including a slight imbalance in the phospholipid number between leaflets, suggests that the thermodynamics of these mixed systems is more nuanced. Indeed, changing the vesicles’ diameter from 200 to 50 nm for the symmetric system as well as the system that highlights DPPC (as in Fig. 3 C) did not impact the lipid distribution, as shown in Fig. S3. In this figure, the scattering curves for the corresponding symmetric as well as the asymmetric vesicles perfectly overlap for both diameters. Disentangling the relative importance of differential stress induced by a difference in phospholipid numbers across leaflets—a key new concept (15,16,66)—and lipid partitioning/affinity, will help to elucidate what drives mixtures of cholesterol and saturated and unsaturated lipids to undergo an asymmetric distribution across leaflets naturally.
For the simple three-component mixtures presented here, DPPC or SM and cholesterol preferentially locate in the inner leaflet, which is reversed from what is observed in the PM. Notwithstanding, the fraction of SM partitioning to the inner leaflet is slightly higher (68 ± 5%) than for DPPC (64 ± 5%), congruent with a higher affinity of SM with cholesterol (53). Additional exploration into different lipid ratios for these ternary mixtures as well as more complex model membranes mimicking the PM is essential to map and uncover whether equilibrium asymmetric configurations exist that can, to some degree, naturally resemble the asymmetry observed in the PM (4,10). From such mappings we will gain insight into the intricacy of biological membrane behavior and the still elusive but relevant enzymatic flippase activity associated with lipid movement across the membrane (8). Indeed, the finding that, in simple ternary systems presented here (DPPC/SM, POPC, and cholesterol), the asymmetry is reversed from what is found in the PM of cells would suggest an even more important role of such enzymatic activity. On the other hand, we pose that exploring more complex lipid compositions is very relevant as these could induce a reversal of lipid arrangements, one that reflects the lipid distribution in the PM. Lipids to include could be polyunsaturated lipids, as well as those having different headgroup types, such as PS and PE.
A final but important point to be highlighted is our finding that the lower bound for cholesterol’s residence time in a membrane leaflet at 45°C is 10 ms. This finding is very different from the results obtained by Bruckner et al. who had reported, instead, that 10 ms was the upper bound for cholesterol’s residence time at 37°C in a membrane leaflet (25). The main difference between our work and theirs is how the vesicles were prepared. Bruckner et al. prepared their vesicles by sonication, which meant they were smaller than 30 nm in diameter and, as Nakano et al. recently showed, they do not preserved lipid order: their calorimetric melting transition peak completely disappears (Fig. S1 in (59)). In fact, we turned to 13C MAS NMR measurements because, on trying to reproduce the results from Bruckner et al., we could not detect any signal from 13C-labeled cholesterol in extruded vesicles from simple solution NMR.
Conclusions
Our results highlight that an equilibrium compositional asymmetry between the leaflets of PM model membranes composed of a saturated lipid, an unsaturated lipid, and cholesterol: DPPC, POPC, and cholesterol, or SM, POPC, and cholesterol, exists in unilamellar vesicles that are the size of small prokaryotic cells (∼100 nm in diameter). We used MAS NMR and SANS, which are techniques ideally suited to detect compositional asymmetry between the leaflets of membranes. These techniques use isotopic labeling, which does not alter the chemical identity of the lipids or the aqueous environment. In these vesicles cholesterol strongly favored the inner leaflet. In addition, we found a slight asymmetry (compared with what is found in the PM) in the distribution of the saturated lipids, which also favored the inner leaflet. The unsaturated lipid POPC, which is found to be essentially symmetrically distributed in the PM (4), had a slight preference for the outer leaflet. A small imbalance in the number of phospholipids between leaflets—a feature not found in single-component lipid vesicles of this size—may play a role in cholesterol’s highly asymmetric distribution. Lipid distribution asymmetries across membrane leaflets identified in the 1970s, were found to be driven by high curvature (vesicles less than 30 nm in diameter). Here, we showed that lipid asymmetry is attained in LUVs and independent of curvature (from 50 up to 200 nm in diameter). Because the model systems studied here, containing a saturated lipid, an unsaturated lipid, and cholesterol, attained a reversed asymmetric configuration to what is found in the PM, further experiments are needed to both completely understand this phenomenon as well as investigate whether more complex lipid compositions can recapture the asymmetric lipid arrangement found in the PMs of cells naturally.
Our findings suggest that the ubiquitous asymmetry observed in the PM of cells may have an important thermodynamic component and not necessarily necessitating the constant intervention of energy-dependent transport proteins to move lipids across the bilayer. Signaling lipids, like PS, may be the few lipids that may need active transport across membranes (8). Having found equilibrium compositional asymmetry in model membranes, they could become model asymmetric lipid platforms to investigate integral membrane proteins (65).
Finally, MAS NMR measurements presented here involving cholesterol showed a peak that was partially quench by Mn2+ and never disappearing into the baseline setting the lower bound for cholesterol flip-flop half-time at 10 ms at physiological temperatures. This result is at least two orders of magnitude longer than what MD simulations predict (21,22,23).
Acknowledgments
The authors thank Prof. Chad M. Rienstra for providing the 13C skip-labeled cholesterol (26). They also thank the Institute Laue Langevin for providing neutron beam time under https://doi.org/10.5291/ILL-DATA.9-13-428. U.P.-S. acknowledges travel support from the ILL to promote scientific collaboration. U.P.-S. thanks the PSCM at the ILL for access to personnel and equipment during sample preparation. U.P.-S. gratefully acknowledges the support from the NSF CAREER award DMR-1753238. U.P.-S. is supported by the T32 Lung Biology and Pathology Training Grant HL007829. This study made use of the National Magnetic Resonance Facility at Madison, which is supported by NIH grant P41GM136463, P41GM103399 (NIGMS). Equipment was purchased with funds from the University of Wisconsin–Madison, the NIH P41GM136463, P41GM103399, S10RR02781, S10RR08438, S10RR023438, S10RR025062, S10RR029220), the NSF (DMB-8415048, OIA-9977486, BIR-9214394), and the USDA.
Author contributions
U.P.-S., L.P., and T.R. designed the research. Y.Z., U.P.-S., L.P., T.R., and K.C.B. performed the research. L.P., T.R., K.C.B., and T.L. contributed new reagents or analytic tools. Y.Z., U.P.-S., L.P., T.R., and K.C.B. analyzed the data. U.P.-S. wrote the original draft. Y.Z., U.P.-S., L.P., T.R., K.C.B., T.L., and Y.L. reviewed and edited the manuscript. All authors read and agreed to the published version of the manuscript.
Declaration of interests
Authors declare no competing interests.
Editor: Daniel Huster.
Footnotes
Supporting material can be found online at https://doi.org/10.1016/j.bpj.2024.10.004.
Supporting material
References
- 1.Singer S.J., Nicolson G.L. The fluid mosaic model of the structure of cell membranes. Science. 1972;175:720–731. doi: 10.1126/science.175.4023.720. [DOI] [PubMed] [Google Scholar]
- 2.Bretscher M.S. Asymmetrical lipid bilayer structure for biological membranes. Nat. New Biol. 1972;236:11–12. doi: 10.1038/newbio236011a0. [DOI] [PubMed] [Google Scholar]
- 3.Devaux P.F. Static and dynamic lipid asymmetry in cell membranes. Biochemistry. 1991;30:1163–1173. doi: 10.1021/bi00219a001. [DOI] [PubMed] [Google Scholar]
- 4.Lorent J.H., Levental K.R., et al. Levental I. Plasma membranes are asymmetric in lipid unsaturation, packing and protein shape. Nat. Chem. Biol. 2020;16:644–652. doi: 10.1038/s41589-020-0529-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.van Meer G., Voelker D.R., Feigenson G.W. Membrane lipids: where they are and how they behave. Nat. Rev. Mol. Cell Biol. 2008;9:112–124. doi: 10.1038/nrm2330. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Borst P., Zelcer N., van Helvoort A. ABC transporters in lipid transport. Biochim. Biophys. Acta. 2000;1486:128–144. doi: 10.1016/s1388-1981(00)00053-6. [DOI] [PubMed] [Google Scholar]
- 7.Doktorova M., Symons J.L., Levental I. Structural and functional consequences of reversible lipid asymmetry in living membranes. Nat. Chem. Biol. 2020;16:1321–1330. doi: 10.1038/s41589-020-00688-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Clarke R.J., Hossain K.R., Cao K. Physiological roles of transverse lipid asymmetry of animal membranes. Biochim. Biophys. Acta Biomembr. 2020;1862 doi: 10.1016/j.bbamem.2020.183382. [DOI] [PubMed] [Google Scholar]
- 9.Doktorova M., Levental I., Heberle F.A. Seeing the Membrane from Both Sides Now: Lipid Asymmetry and Its Strange Consequences. Cold Spring Harbor Perspect. Biol. 2023;15 doi: 10.1101/cshperspect.a041393. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Doktorova M., Symons J.L., et al. Levental I. Cell Membranes Sustain Phospholipid Imbalance Via Cholesterol Asymmetry. bioRxiv. 2023 doi: 10.1101/2023.07.30.551157. Preprint at. [DOI] [Google Scholar]
- 11.Buwaneka P., Ralko A., et al. Cho W. Evaluation of the available cholesterol concentration in the inner leaflet of the plasma membrane of mammalian cells. J. Lipid Res. 2021;62 doi: 10.1016/j.jlr.2021.100084. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Liu S.L., Sheng R., et al. Cho W. Orthogonal lipid sensors identify transbilayer asymmetry of plasma membrane cholesterol. Nat. Chem. Biol. 2017;13:268–274. doi: 10.1038/nchembio.2268. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Courtney K.C., Pezeshkian W., et al. Zha X. C24 Sphingolipids Govern the Transbilayer Asymmetry of Cholesterol and Lateral Organization of Model and Live-Cell Plasma Membranes. Cell Rep. 2018;24:1037–1049. doi: 10.1016/j.celrep.2018.06.104. [DOI] [PubMed] [Google Scholar]
- 14.Solanko L.M., Sullivan D.P., et al. Wüstner D. Ergosterol is mainly located in the cytoplasmic leaflet of the yeast plasma membrane. Traffic. 2018;19:198–214. doi: 10.1111/tra.12545. [DOI] [PubMed] [Google Scholar]
- 15.Hossein A., Deserno M. Spontaneous Curvature, Differential Stress, and Bending Modulus of Asymmetric Lipid Membranes. Biophys. J. 2020;118:624–642. doi: 10.1016/j.bpj.2019.11.3398. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Varma M., Deserno M. Distribution of cholesterol in asymmetric membranes driven by composition and differential stress. Biophys. J. 2022;121:4001–4018. doi: 10.1016/j.bpj.2022.07.032. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Marquardt D., Heberle F.A., et al. Pabst G. 1H NMR Shows Slow Phospholipid Flip-Flop in Gel and Fluid Bilayers. Langmuir. 2017;33:3731–3741. doi: 10.1021/acs.langmuir.6b04485. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Scott H.L., Kennison K.B., et al. Katsaras J. Model Membrane Systems Used to Study Plasma Membrane Lipid Asymmetry. Symmetry. 2021;13 doi: 10.3390/sym13081356. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Doktorova M., Heberle F.A., et al. Marquardt D. Preparation of asymmetric phospholipid vesicles for use as cell membrane models. Nat. Protoc. 2018;13:2086–2101. doi: 10.1038/s41596-018-0033-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.London E. Membrane Structure-Function Insights from Asymmetric Lipid Vesicles. Acc. Chem. Res. 2019;52:2382–2391. doi: 10.1021/acs.accounts.9b00300. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Gu R.-X., Baoukina S., Tieleman D.P. Cholesterol Flip-Flop in Heterogeneous Membranes. J. Chem. Theor. Comput. 2019;15:2064–2070. doi: 10.1021/acs.jctc.8b00933. [DOI] [PubMed] [Google Scholar]
- 22.Baral S., Levental I., Lyman E. Composition dependence of cholesterol flip-flop rates in physiological mixtures. Chem. Phys. Lipids. 2020;232 doi: 10.1016/j.chemphyslip.2020.104967. [DOI] [PubMed] [Google Scholar]
- 23.Aghaaminiha M., Farnoud A.M., Sharma S. Quantitative relationship between cholesterol distribution and ordering of lipids in asymmetric lipid bilayers. Soft Matter. 2021;17:2742–2752. doi: 10.1039/d0sm01709d. [DOI] [PubMed] [Google Scholar]
- 24.Cino E.A., Tieleman D.P. Curvature-based sorting of eight lipid types in asymmetric buckled plasma membrane models. Biophys. J. 2022;121:2060–2068. doi: 10.1016/j.bpj.2022.05.002. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Bruckner R.J., Mansy S.S., et al. Szostak J.W. Flip-flop-induced relaxation of bending energy: implications for membrane remodeling. Biophys. J. 2009;97:3113–3122. doi: 10.1016/j.bpj.2009.09.025. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Della Ripa L.A., Petros Z.A., et al. Rienstra C.M. Solid-State NMR of highly (13)C-enriched cholesterol in lipid bilayers. Methods. 2018;138–139:47–53. doi: 10.1016/j.ymeth.2018.01.008. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Scott H.L., Skinkle A., et al. Heberle F.A. On the Mechanism of Bilayer Separation by Extrusion, or Why Your LUVs Are Not Really Unilamellar. Biophys. J. 2019;117:1381–1386. doi: 10.1016/j.bpj.2019.09.006. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Krzyzanowski N., Porcar L., Perez-Salas U. A Small-Angle Neutron Scattering, Calorimetry and Densitometry Study to Detect Phase Boundaries and Nanoscale Domain Structure in a Binary Lipid Mixture. Membranes. 2023;13:323. doi: 10.3390/membranes13030323. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Michaelson D.M., Horwitz A.F., Klein M.P. Transbilayer asymmetry and surface homogeneity of mixed phospholipids in cosonicated vesicles. Biochemistry. 1973;12:2637–2645. doi: 10.1021/bi00738a014. [DOI] [PubMed] [Google Scholar]
- 30.Bystrov V.F., Dubrovina N.I., et al. Bergelson L.D. NMR differentiation of the internal and external phospholipid membrane surfaces using paramagnetic Mn2+ and Eu3+ ions. Chem. Phys. Lipids. 1971;6:343–350. [Google Scholar]
- 31.Fung B.M., Khitrin A.K., Ermolaev K. An improved broadband decoupling sequence for liquid crystals and solids. J. Magn. Reson. 2000;142:97–101. doi: 10.1006/jmre.1999.1896. [DOI] [PubMed] [Google Scholar]
- 32.Dewhurst C.D. Graphical reduction and analysis small-angle neutron scattering program: GRASP. J. Appl. Crystallogr. 2023;56:1595–1609. doi: 10.1107/S1600576723007379. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Doucet, J. H. M.C., Alina G., et al. Washington A. 2017. SasView. Version 4.1.2. [Google Scholar]
- 34.Greenwood A.I., Tristram-Nagle S., Nagle J.F. Partial molecular volumes of lipids and cholesterol. Chem. Phys. Lipids. 2006;143:1–10. doi: 10.1016/j.chemphyslip.2006.04.002. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Kucerka N., Nieh M.P., Katsaras J. Fluid phase lipid areas and bilayer thicknesses of commonly used phosphatidylcholines as a function of temperature. Biochim. Biophys. Acta. 2011;1808:2761–2771. doi: 10.1016/j.bbamem.2011.07.022. [DOI] [PubMed] [Google Scholar]
- 36.Mouret L., Da Costa G., Bondon A. Sterols associated with small unilamellar vesicles (SUVs): intrinsic mobility role for 1H NMR detection. Magn. Reson. Chem. 2014;52:339–344. doi: 10.1002/mrc.4069. [DOI] [PubMed] [Google Scholar]
- 37.Su Y., Mani R., Hong M. Asymmetric insertion of membrane proteins in lipid bilayers by solid-state NMR paramagnetic relaxation enhancement: a cell-penetrating Peptide example. J. Am. Chem. Soc. 2008;130:8856–8864. doi: 10.1021/ja802383t. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Bergelson L.D., Barsukov L.I. Topological asymmetry of phospholipids in membranes. Science. 1977;197:224–230. doi: 10.1126/science.327544. [DOI] [PubMed] [Google Scholar]
- 39.Leventis R., Silvius J.R. Use of cyclodextrins to monitor transbilayer movement and differential lipid affinities of cholesterol. Biophys. J. 2001;81:2257–2267. doi: 10.1016/S0006-3495(01)75873-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Garg S., Porcar L., et al. Perez-Salas U. Noninvasive neutron scattering measurements reveal slower cholesterol transport in model lipid membranes. Biophys. J. 2011;101:370–377. doi: 10.1016/j.bpj.2011.06.014. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Breidigan J.M., Krzyzanowski N., et al. Perez-Salas U. Influence of the membrane environment on cholesterol transfer. J. Lipid Res. 2017;58:2255–2263. doi: 10.1194/jlr.M077909. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Veatch S.L., Keller S.L. Separation of liquid phases in giant vesicles of ternary mixtures of phospholipids and cholesterol. Biophys. J. 2003;85:3074–3083. doi: 10.1016/S0006-3495(03)74726-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43.Huang J., Buboltz J.T., Feigenson G.W. Maximum solubility of cholesterol in phosphatidylcholine and phosphatidylethanolamine bilayers. Biochim. Biophys. Acta. 1999;1417:89–100. doi: 10.1016/s0005-2736(98)00260-0. [DOI] [PubMed] [Google Scholar]
- 44.Garg S., Castro-Roman F., et al. Perez-Salas U. Cholesterol solubility limit in lipid membranes probed by small angle neutron scattering and MD simulations. Soft Matter. 2014;10:9313–9317. doi: 10.1039/c4sm01219d. [DOI] [PubMed] [Google Scholar]
- 45.Liu Y., Kelley E.G., et al. Perez-Salas U. Creating Asymmetric Phospholipid Vesicles via Exchange With Lipid-Coated Silica Nanoparticles. Langmuir. 2020;36:8865–8873. doi: 10.1021/acs.langmuir.0c01188. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46.Heller W.T. Small-Angle Neutron Scattering Study of a Phosphatidylcholine-Phosphatidylethanolamine Mixture. ACS Omega. 2023;8:33755–33762. doi: 10.1021/acsomega.3c04164. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47.Nguyen M.H.L., DiPasquale M., et al. Marquardt D. Peptide-Induced Lipid Flip-Flop in Asymmetric Liposomes Measured by Small Angle Neutron Scattering. Langmuir. 2019;35:11735–11744. doi: 10.1021/acs.langmuir.9b01625. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 48.Marx L., Frewein M.P.K., et al. Pabst G. Antimicrobial peptide activity in asymmetric bacterial membrane mimics. Faraday Discuss. 2021;232:435–447. doi: 10.1039/d1fd00039j. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 49.Heberle F.A., Marquardt D., et al. Pabst G. Subnanometer Structure of an Asymmetric Model Membrane: Interleaflet Coupling Influences Domain Properties. Langmuir. 2016;32:5195–5200. doi: 10.1021/acs.langmuir.5b04562. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 50.Heberle F.A., Pabst G. Complex biomembrane mimetics on the sub-nanometer scale. Biophys. Rev. 2017;9:353–373. doi: 10.1007/s12551-017-0275-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 51.Bryant G., Taylor M.B., et al. Garvey C.J. Effect of deuteration on the phase behaviour and structure of lamellar phases of phosphatidylcholines - Deuterated lipids as proxies for the physical properties of native bilayers. Colloids Surf. B Biointerfaces. 2019;177:196–203. doi: 10.1016/j.colsurfb.2019.01.040. [DOI] [PubMed] [Google Scholar]
- 52.Barenholz Y., Thompson T.E. Sphingomyelins in bilayers and biological membranes. Biochim. Biophys. Acta. 1980;604:129–158. doi: 10.1016/0005-2736(80)90572-6. [DOI] [PubMed] [Google Scholar]
- 53.van Duyl B.Y., Ganchev D., et al. Killian J.A. Sphingomyelin is much more effective than saturated phosphatidylcholine in excluding unsaturated phosphatidylcholine from domains formed with cholesterol. FEBS Lett. 2003;547:101–106. doi: 10.1016/s0014-5793(03)00678-1. [DOI] [PubMed] [Google Scholar]
- 54.de Kruijff B., Cullis P.R., Radda G.K. Differential scanning calorimetry and 31P NMR studies on sonicated and unsonicated phosphatidylcholine liposomes. Biochim. Biophys. Acta. 1975;406:6–20. doi: 10.1016/0005-2736(75)90038-3. [DOI] [PubMed] [Google Scholar]
- 55.De Kruijff B., Cullis P.R., Radda G.K. Outside-inside distributions and sizes of mixed phosphatidylcholine-cholesterol vesicles. Biochim. Biophys. Acta. 1976;436:729–740. doi: 10.1016/0005-2736(76)90402-8. [DOI] [PubMed] [Google Scholar]
- 56.Berden J.A., Barker R.W., Radda G.K. NMR studies on phospholipid bilayers. Some factors affecting lipid distribution. Biochim. Biophys. Acta. 1975;375:186–208. doi: 10.1016/0005-2736(75)90188-1. [DOI] [PubMed] [Google Scholar]
- 57.Huang C.H., Sipe J.P., et al. Martin R.B. Differential interaction of cholesterol with phosphatidylcholine on the inner and outer surfaces of lipid bilayer vesicles. Proc. Natl. Acad. Sci. USA. 1974;71:359–362. doi: 10.1073/pnas.71.2.359. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 58.Op den Kamp J.A.F. Lipid asymmetry in membranes. Annu. Rev. Biochem. 1979;48:47–71. doi: 10.1146/annurev.bi.48.070179.000403. [DOI] [PubMed] [Google Scholar]
- 59.Nakano M., Nakao H., et al. Ikeda K. Energetic and Structural Insights into Phospholipid Transfer from Membranes with Different Curvatures by Time-Resolved Neutron Scattering. J. Phys. Chem. Lett. 2022;13:6024–6030. doi: 10.1021/acs.jpclett.2c01080. [DOI] [PubMed] [Google Scholar]
- 60.Olson F., Hunt C.A., et al. Papahadjopoulos D. Preparation of liposomes of defined size distribution by extrusion through polycarbonate membranes. Biochim. Biophys. Acta. 1979;557:9–23. doi: 10.1016/0005-2736(79)90085-3. [DOI] [PubMed] [Google Scholar]
- 61.Hope M.J., Bally M.B., et al. Cullis P.R. Production of large unilamellar vesicles by a rapid extrusion procedure: characterization of size distribution, trapped volume and ability to maintain a membrane potential. Biochim. Biophys. Acta. 1985;812:55–65. doi: 10.1016/0005-2736(85)90521-8. [DOI] [PubMed] [Google Scholar]
- 62.Mayer L.D., Hope M.J., Cullis P.R. Vesicles of variable sizes produced by a rapid extrusion procedure. Biochim. Biophys. Acta. 1986;858:161–168. doi: 10.1016/0005-2736(86)90302-0. [DOI] [PubMed] [Google Scholar]
- 63.Kucerka N., Nieh M.P., Katsaras J. Asymmetric distribution of cholesterol in unilamellar vesicles of monounsaturated phospholipids. Langmuir. 2009;25:13522–13527. doi: 10.1021/la9020299. [DOI] [PubMed] [Google Scholar]
- 64.Krompers M., Heerklotz H. A Guide to Your Desired Lipid-Asymmetric Vesicles. Membranes. 2023;13 doi: 10.3390/membranes13030267. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 65.Bogdanov M. The power and challenge of lipid (a)symmetry across the membrane and cell. Emerg. Top. Life Sci. 2023;7:1–6. doi: 10.1042/ETLS20220088. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 66.Foley S.L., Varma M., et al. Deserno M. Elastic and thermodynamic consequences of lipid membrane asymmetry. Emerg. Top. Life Sci. 2023;7:95–110. doi: 10.1042/ETLS20220084. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.