Formation and Characterization of Supported Lipid Bilayers Composed of Phosphatidylethanolamine and Phosphatidylglycerol by Vesicle Fusion, a Simple but Relevant Model for Bacterial Membranes.

Supported lipid bilayers (SLBs) are simple and robust biomimics with controlled lipid composition that are widely used as models of both mammalian and bacterial membranes. However, the lipids typically used for SLB formation poorly resemble those of bacterial cell membranes due to the lack of available protocols to form SLBs using mixtures of lipids relevant for bacteria such as phosphatidylethanolamine (PE) and phosphatidylglycerol (PG). Although a few reports have been published recently on the formation of SLBs from Escherichia coli lipid extracts, a detailed understanding of these systems is challenging due to the complexity of the lipid composition in such natural extracts. Here, we present for the first time a simple and reliable protocol optimized to form high-quality SLBs using mixtures of PE and PG at compositions relevant for Gram-negative membranes. We show using neutron reflection and quartz microbalance not only that Ca2+ ions and temperature are key parameters for successful bilayer deposition but also that mass transfer to the surface is a limiting factor. Continuous flow of the lipid suspension is thus crucial for obtaining full SLB coverage. We furthermore characterize the resulting bilayers and report structural parameters, for the first time for PE and PG mixtures, which are in good agreement with those reported earlier for pure POPE vesicles. With this protocol in place, more suitable and reproducible studies can be conducted to understand biomolecular processes occurring at cell membranes, for example, for testing specificities and to unravel the mechanism of interaction of antimicrobial peptides.

Introduction

Supported lipid bilayers (SLBs) are often used as models of biological membranes to either characterize the structural and dynamic properties of the membrane itself or to study biomolecular interactions at the membrane. The great majority of the studies include SLBs formed either from saturated or unsaturated phosphatidylcholine (PC) species.[1−4] To make more appropriate mimics of mammalian membranes, PC is typically mixed with either sterols[5] or charged lipids such as phosphatidylserine,[6] phosphatidylglycerol (PG),[7] and most recently phosphatidylinositols.[8] For bacterial membranes, on the other hand, mixtures of PC with phosphatidylethanolamine (PE)[9] and PG[10,11] have been used mainly due to the belief that it is not possible to form SLBs of high coverage via the vesicle fusion method using PE and PG lipids.[12] In some cases, pure PC membranes were used to determine the mechanism of action of antimicrobial compounds.[13−15] Recently, PE-rich SLBs were formed in mixtures with PC in an attempt to create more relevant model bacterial membranes.[9] As an alternative to vesicle fusion Langmuir–Blodgett/Langmuir–Schaefer deposition has been successfully applied to produce advanced models of bacterial membranes using PC and liposaccharides,[16−18] but this method is significantly more time-consuming and requires high technical skills to succeed. However, the membranes produced by the Langmuir–Blodgett/Langmuir–Schaefer method are asymmetric in nature and represent an accurate model of the outer membrane of Gram-negative bacteria.

The molar ratio of PE to PG in bacterial membranes is 1:6 (17% PG) in the outer and 1:3 (33% PG) in the inner or cytosolic membranes of Gram-negative bacteria, respectively.[19,20] Therefore, typical bilayer models that mimic bacterial membranes contain 20–30% PG mixed with PC.[10]

The lack of success in forming SLBs with PE and PG has been attributed in part to the negatively charged PG head groups and in part to the molecular shape of POPE. It has been argued that lipids with cylindrical shapes and a packing parameter close to 1 such as PC are required to form SLBs.[21] PE has been reported to give hexagonal liquid crystalline structures (explained as a consequence of its small headgroup) even though the PE rich polar extracts of Escherichia coli present a lamellar packing.[20] However, stable SLBs were formed by 1,2-dipalmitoylphosphatidylethanolamine (DPPE) using Langmuir–Blodgett/Langmuir–Schaefer deposition in the past.[22] We have previously shown by a set of complementary surface-sensitive techniques including atomic force microscopy, quartz crystal microbalance with dissipation (QCM-D), and neutron reflection (NR) that SLBs made from polar lipid extracts of E. coli could be formed by vesicle fusion.[23] Deposition was successfully achieved by tip sonication of a lipid suspension in Tris buffer at pH 7.4 supplemented with 2 mM CaCl2 at 50 °C. Natural E. coli lipid extracts include cardiolipin (CL) besides PG and PE, and a detailed structural analysis is therefore complicated due to the unknown and complex lipid composition of the final SLB.[23] For more controlled studies, it is thus favorable to use mixtures of standard commercially available lipids with known molecular properties such as mass and component volumes.

Here, we present the optimization of the formation of SLBs composed of negatively charged 1-palmitoyl-2-oleoyl-sn-glycero-3-phospho-(1′-rac-glycerol) (POPG) and zwitterionic 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine (POPE) on silica (see Figure A for the chemical structures of these lipids). The optimization was performed using both NR and QCM-D, and the structure of POPE–POPG SLBs is reported for the first time. In contrast to QCM-D, NR measures the scattering length density perpendicular to the surface, which relates to the dry surface coverage, the level of hydration, and the structure down to 5 Å resolution. Therefore, NR excels at determining the exact coverage of the SLB upon vesicle fusion. Figure B gives a simulation of the NR signal as a function of the SLB coverage. QCM-D, on the other hand, determines the wet mass of the film and is sensitive to the presence of unfused vesicles on the surface. In this study, the PG content was varied between 10 and 25 mol % to create mimics of the Gram-negative bacteria membranes.

Figure 1

(A) Molecular structures of POPE and POPG lipids. (B) Neutron reflection data simulation for an SLB in D2O with low (red tones) to intermediate (yellow tones) and high (purple tones) surface coverage. (C) Sketch of a lipid bilayer slab model color coded to the scattering length density (SLD) of each of the components in the system.

Results and Discussion

Divalent cations like Mg2+ or Ca2+ are commonly used to neutralize the net negative charge of the glycerol headgroup in POPG while also bridging to the anionic silica surface. This neutralization and bridging leads to vesicle aggregation as measured by dynamic light scattering (DLS) (see Supporting Information (SI) Figure S1). Recently, it was shown that CaCl2 induced an apparent size increase in vesicle suspensions made from E. coli extracts; however, the size increase was mainly due to vesicle aggregation and not due to vesicle fusion, and this vesicle aggregation was reversible upon diluting with Ca2+-free buffer.[24] Here, CaCl2 was used as the fusion promotor, and several concentrations were tested for optimizing the deposition with NR and QCM-D connected in series on the NR beamline. SLB formation via vesicle fusion is facilitated by using lipids in the fluid phase. The melting temperature (Tm) of POPG is −2 °C,[25] and the calorimetric Tm of POPE was reported to be 38 °C[26,27] although Avanti Lipids report 25 °C in their specifications. Thus, if deposition is carried out at room temperature, the lipids may still be in the gel phase rather than in the liquid phase due to the high phase-transition temperature of POPE. It is possible to form SLBs in the gel phase;[28] however, for lipids with a typical nonfusing behavior, it is particularly important to optimize the deposition conditions to get high-quality SLBs.

Here, initially NR was used to investigate the formation of an SLB of 75 mol % POPE and 25 mol % POPG (hereon referred to as 25 mol % PG) at 37 °C, well above the Tm of POPE. NR data were collected at the time-of-flight INTER reflectometer at the ISIS neutron source in the U.K. over the Q-range of 0.08–0.25 Å−1 by using two incident angles (0.7 and 2.3°) at a Q resolution of 3%. Figure gives the NR profile upon continuous addition of 0.1 mg/mL lipid vesicles in 1 or 2 mM CaCl2 at a flow rate of 1 mL/min by the use of a syringe pump. The addition of 20 mL of the lipid solution in 1 mM CaCl2 showed no significant surface adsorption (<10 vol %) as seen from Figure (red curve). To assess the importance of the calcium concentration, an additional 20 mL of lipids in 2 mM CaCl2 was added to the same surface under continuous flow. The elevated CaCl2 concentration led to poor bilayer coverage (yellow curve in Figure ; compare with the simulation in Figure ). A second, independent, experiment was then carried out with continuous flow of 50 mL of the lipid solution in 2 mM CaCl2 to a clean surface, resulting in a surface coverage similar to that obtained when adding only 20 mL of the solution (blue curve in Figure ). This suggests that the deposition of 25 mol % PG in the presence of 2 mM CaCl2 at 37 °C leads to SLBs with unsatisfactory surface coverage.

Figure 2

NR profiles obtained in D2O contrast after continuous flow of a solution of 0.1 mg/mL 25 mol % PG vesicles at 37 °C onto a silicon oxide substrate at 1 mL/min. Red: 20 mL lipid solution in 1 mM CaCl2; yellow: the same experiment as the red curve, but after the addition of another 20 mL of lipids in 2 mM CaCl2; blue: a replicate experiment with 50 mL of lipid solution added in 2 mM CaCl2 to a clean surface.

To assess the effect of temperature, a third experiment was carried out at 50 °C: a temperature that previously resulted in successful SLB deposition of polar lipid extracts from E coli.[23]Figure A gives time-resolved kinetic NR data for the continuous addition of 25 mol % PG vesicles at 50 °C using 3 mM CaCl2. In this case, measurements were carried out at a single incidence angle (2.3°) corresponding to a Q-range of 0.03–0.3 Å–1, where changes in the reflectivity are the most pronounced (see Figure B). The NR data were fitted to a layered SLB model consisting of three slabs (heads–tail–heads: see Figure C), and Figure B (red curve) gives the fitted SLB coverage as a function of time during continuous lipid flow. This figure also includes the data for deposition at 2 mM CaCl2 (blue curve) under the same experimental conditions. The data show that an SLB with full coverage could be obtained using both 2 and 3 mM CaCl2 and that 3 mM CaCl2 gave a slightly faster kinetics of SLB formation. Our data indicate that electrostatics are critical for fusion to take place and it is essential to have enough Ca2+ present to bridge vesicles and to fuse them to the SiO2 surface. However, reproducible deposition of lipids prepared for different experiments (with different lipid and CaCl2 stocks) was only obtained at 3 mM CaCl2. This suggests not only that deposition at 50 °C is necessary to obtain a high bilayer coverage, but also that the critical CaCl2 concentration for fusion is close to or just above 2 mM. Hence, small differences in the lipid film composition or the actual CaCl2 stock concentration caused by inaccuracies in the sample preparation can influence the degree of SLB coverage when using 2 mM CaCl2.

Figure 3

(A) NR data and best fits for the kinetics of deposition of 25 mol % PG with 3 mM CaCl2 at 50 °C. The inset in (A) shows the corresponding SLD profiles for the best fits. In (B), the fitted SLB coverage is given as a function of time for a 10 mol % PG SLB deposited at 50 °C in 3 mM CaCl2 (purple) or 25 mol % PG at 50 °C with 2 mM (blue) or 3 mM CaCl2 (green and red). The model used to fit the SLB is described in detail in the main text regarding the discussion of Figure . The NR kinetic data for the other data sets are shown in the Supporting Information Figures S3–S5.

Figure 4

NR profiles and best fits for a 25 mol % PG bilayer formed at 50 °C (red) and subsequently cooled to 37 °C (yellow) and 25 °C (blue). The bilayer was characterized in D2O and H2O at all three temperatures. The inset shows the SLD profile for the SLB that gives no structural differences between 50 and 37 °C, whereas a 1.0 ± 0.5 Å thickening of the lipid core took place for 25 °C. The same data have been replotted as RQ(4) vs Q in the Supporting Information Figure S9 to highlight the quality of the fits.

The final SLB structure was obtained upon collection of NR data over the full Q-range (0.08–0.25 Å1) in two isotopic contrasts: 100% D2O and 100% H2O. Figure gives the NR data and best fits for a 25 mol % PG SLB formed with 3 mM CaCl2 at 50 °C and after cooling down to relevant experimental temperatures such as 37 and 25 °C. Details on data fitting and the final fitted parameters are given in the methods section and the Supporting Information. Briefly, the data were fitted assuming a symmetric bilayer in which the inner heads, tails, and outer heads were represented as individual layers. In the bilayer model, an additional water layer was added below the SLB, which represents the water trapped in the roughness of the SiO2 substrate. At 50 and 37 °C, the bilayer displayed full coverage (no solvent penetration into the lipid core), thin headgroups (5.5 ± 0.5 Å), a tail thickness of 31 ± 0.5 Å2, and a mean molecular area of 60 Å2. Our data suggest that the lipid core thickens slightly from 31 to 32 Å upon lowering the temperature from 37 to 25 °C, leading to negligible changes in the head group hydration and a slight decrease in the mean molecular area of the lipids (59 Å2). The molecular area of pure POPE bilayers were reported to be ∼61 and 58 Å2 at 50 and 35 °C,[29,30] respectively, in excellent agreement with our results. Importantly, our NR data clearly show that there is no significant loss of lipids upon cooling and that there are no significant changes in the structure of the SLB.

Earlier it was shown that stable DPPE SLBs with a headgroup thickness of 7 Å in the fluid phase could be formed by Langmuir–Blodgett/Langmuir–Schaefer deposition.[22] This headgroup thickness is slightly larger than those found here for POPE–POPG. Fluid POPC SLBs were previously reported to have a headgroup thickness of 8 Å and a core thickness of 31 Å with an mean molecular area (MMA) of 61 Å2 at 25 °C.[1] Slightly thinner cores (30 Å) were reported for fluid POPC–POPG (75:25 mol %) SLBs at 25 °C,[1,7] whereas the headgroup was found to be between 8[7] and 10 Å[1] for these systems. Moreover, the lipid core thicknesses for POPE were reported to be 32 and 31 Å at 50 and 35 °C,[29,30] respectively,[29,30] also in excellent agreement with our results. Molecular dynamics (MD) simulations suggested that the P–P distance in POPE–POPG bilayers was only 0.5 Å thinner than that in POPC bilayers (values that are at the very limit of what NR can detect) and that the PE preferentially interacted with PG rather than with other PE headgroups.[31] Such a preferential interaction between the PE and PG heads might explain why we see a 2 Å thinner headgroup than that reported for pure PE earlier.[22] A different MD simulation study suggested that below the Tm, POPE bilayers contained primarily lipids in a tilted arrangement with a more compact lipid packing compared with POPC, due to the smaller headgroup in POPE molecules. This tilting, however, disappeared as the bilayers became fluid.[32] Therefore, our data seem to be in agreement with the structural parameters predicted by MD simulations and earlier reports for pure PE[22] and PC bilayers.[1] Finally, similar changes in the thickness of the POPE–POPG SLBs due to changes in temperature were reported earlier for POPE and POPG SLBs made by Langmuir–Blodgett deposition.[33]

To investigate the role of the lipid composition, time-resolved NR data were subsequently collected for SLB formation of 10 mol % PG vesicles at 50 °C upon constant flow using a syringe pump at 1 mL/min and a solution containing 0.1 mg/mL lipids in 3 mM CaCl2 (Supporting Information Figure S3). As the negative charge was significantly lower, faster deposition took place and full coverage (95 ± 1 vol %) was obtained within 10 min, as seen in Figure . The experiment was again carried out for one incident angle (2.3°), but this time in event mode, where one continuous reflectivity measurement is performed and, afterward, sliced into sensible time frames to give optimized statistics for fast deposition kinetics (kinetic data for 10 mol % PG can be found in the Supporting Information Figure S5). The structure of the 10% PG membrane was similar to that of the 25% PG membrane.

So far, we demonstrated the successful formation of SLBs made of POPE and POPG deposited via vesicle fusion using NR. Although this technique is not highly sensitive toward diffuse structures such as vesicles, we have observed the effect of vesicle adsorption on the reflectivity profile previously.[28] QCM-D, on the other hand, excels at detecting small adsorbed amounts of such large, water-filled structures[10] as these give rise to a large change in dissipation (softness of the adsorbed layer) and a small frequency change (wet adsorbed mass). Therefore, we used the Q-sense E4 QCM-D system connected in-line with the NR cell and monitored the SLB deposition in parallel on a small SiO2 sensor. Figure and Supporting Information Figure S7 give the QCM-D signals for the deposition of POPE–POPG SLBs using 3 mM CaCl2 for 25 and 10 mol % PG, respectively. Upon vesicle injection, no overshooting in the frequency or dissipation took place, which is the typical fingerprint for successful SLB formation.[34] Instead, a continuous decrease in frequency occurred until steady state was reached at approximately −25 Hz regardless of the composition. For the dissipation, steady state was reached at 1.5 or 0.5 × 10−6 units for 25 and 10 mol %, respectively. Moreover, a very small spreading of the overtones was measured for 25 mol % PG, whereas no spreading at all took place for 10 mol % PG. The QCM-D data altogether suggest that the lipid vesicles fuse upon contact with the SiO2 surface and that there is no critical coverage of bound vesicles needed for the vesicles to break as has been suggested earlier for PC lipids in buffer.[34,35] However, it is clear that vesicle adsorption is the limiting step in the deposition of POPE–POPG SLBs in agreement with the theoretical model and stochastic simulations for the adsorption process of vesicles on surfaces reported earlier by Plunkett et al.[35] Interestingly, this model indicated that a critical vesicle concentration with overshooting in lipid adsorption (vesicles) occurred only under a very narrow set of conditions, and it was therefore more likely to observe saturation kinetics in the adsorption of the SLB without the typical overshooting in frequency and dissipation as observed here for POPE–POPG bilayers.

Figure 5

QCM-D signals for the formation of an SLB with 25 mol % PG in D2O. Events labeled with consecutive numbers mark the following: 1, MQ water with 3 mM CaCl2; 2, 0.1 mg/mL lipids; 3, pump stopped; 4, D2O rinse for 30 min (peak from removal of CaCl2); and 5, pump stopped. While the pump is on (at a high flow rate, 1 mL/min), it creates a pressure, which appears like added mass. When the pump is off, the bilayer frequency stabilizes at −25 Hz.

Moreover, the minimal overtone spreading measured indicates that no significant additional vesicle adsorption takes place on top of the SLB (this is especially true for the 10 mol % PG SLB; Supporting Information Figure S7). For 10 mol % PG, the final frequency value was slightly above −25 Hz, supporting the NR data with almost complete SLB formation (95% surface coverage).

Note that the use of an inline QCM-D implies that the lipid solution was pumped first through the NR cell setup and subsequently through the QCM-D sensor cells. This means that the high flow rate needed for the NR experiment creates a local pressure upon pushing the solution inside the much smaller QCM-D cell. This gives signal alterations in the QCM-D that relax upon turning the pump off. Supporting Information Figure S8 gives the QCM-D data collected at 37 °C and 2 mM CaCl2, which gave insignificant lipid adsorption on the Si crystal in the NR setup (red curve in Figure ), although some adsorption occurred on the QCM-D sensor (Δf = ∼ −15 Hz). This experiment clearly shows that a discontinued flow leads to a plateau in the QCM-D signals and arrested lipid adsorption. Addition of a new lipid solution prepared in 2 mM CaCl2 led to typical QCM-D signals that are considered as full bilayer coverage in the QCM-D, but in the NR cell the bilayer coverage was low. This suggests that the mass transfer to the surface is the limiting factor to determine successful deposition of an SLB of 25 mol % PG and that it is important to keep a constant lipid flow until a complete coverage is reached. Moreover, it is important to properly rinse off the vesicle suspension as soon as the SLB is formed since prolonged exposure of the SLB to the vesicle suspension can lead to deposition of a liquid crystalline phase from the bulk solution resulting in the appearance of a Bragg peak (Supporting Information Figure S6). The phase separation occurs in solution due to vesicle aggregation in the presence of CaCl2. Prolonged vesicle incubation was also shown to lead to irreversible vesicle binding of E. coli extracts in the presence of CaCl2.[24]

Although both NR and QCM-D sensor surfaces are made of SiO2, it should be stressed that the surfaces are not identical. In particular, the roughness of the QCM-D Si sensors is 1 order of magnitude higher than that for polished NR surfaces (4–6 Å RMS). A certain surface roughness may enhance lipid attachment and vesicle fusion and thus make bilayer formation occur in the QCM-D under particular conditions other than those for an NR setup. The flow rate has also proved important for successful bilayer formation and, since the dimensions of the neutron cell and the QCM-D cell are different, the mass transfer to the surface will be higher in the smaller QCM-D cell under the experimental conditions necessary for NR.

We also succeeded in forming both 10 and 25% PG bilayers in the QCM-D using the standard setup (peristaltic pump, dragging mode at 0.1 mL/min). In this case, however, the experiments were notoriously more difficult to reproduce due to the formation of air bubbles at high temperatures (even after degassing) and the deposition of vesicles on the surface due to aggregation and phase separation rather than due to surface related phenomena. We have previously demonstrated the importance of surface directionality to differentiate between surface phenomena and deposition to surfaces due to phase-separation processes and the effect of gravitation.[36] Therefore, using the QCM-D upside down might be advantageous for studies where aggregation occurs in the bulk as well as the use of higher flow rates.

Conclusions

In summary, we have, by vesicle fusion, formed and for the first time structurally characterized SLBs made of POPG and POPE, which can be considered as better mimics of simple bacterial membranes. Indeed, antimicrobial compounds have been reported to differentiate between PE and PC lipids in the past[37] and this is particularly important in discriminating between mammalian cells, Gram-positive bacteria, and Gram-negative bacteria. Our experiments suggest that a PE:PG vesicle concentration of 0.1 mg/mL in 3 mM CaCl2 introduced at continuous flow by a syringe pump at a rate of 1 mL/min at 50 °C are the optimal conditions for bilayer formation. The time needed to achieve full coverage depends on the amount of PG in the vesicle preparation, as long as a critical CaCl2 concentration is used (which is just above 2 mM CaCl2). It is important that the flow is kept constant for as long as it is needed to achieve full coverage (15–20 or 50–55 min for 10 or 25 mol % PG, respectively) and that the lipid suspension is rinsed off as soon as full coverage is obtained to avoid excess lipid vesicle adsorption. From a stability point of view, it is also desirable to work with a high flow rate and low vesicle concentration due to the propensity of vesicle aggregation (and eventually phase separation) in the bulk solution after the addition of CaCl2. It is highly recommended to add the CaCl2 as a 1:1 dilution into freshly sonicated small unilamellar vesicles immediately prior to introduction into the sample cell to limit aggregation and premature vesicle fusion as DLS data indicate that aggregation starts to take place within 5 min of mixing. Finally, the structure of POPE–POPG bilayers is similar to those of pure POPE bilayers (vesicle studies) with a small headgroup (5 Å) and slightly thicker lipid cores as compared with POPC. There seems to be considerably less thermal expansion for POPE–POPG than for PC systems as subjecting the POPE–POPG bilayers to temperatures below the Tm only leads to a 1 Å thickening of the tail core and a 1 Å2 decrease in the mean molecular area.

Materials and Methods

POPE and POPG were bought from Avanti Lipids and used without further purification. MQ H2O was used in all preparations, and D2O (sigma) was provided by the neutron facility. CaCl2 was bought from Sigma Aldrich. Lipids were mixed in chloroform to the appropriate molar ratio using stocks of the individual lipids. The organic solvent was dried under N2 flow and subjected to vacuum for at least 1 h. Films were kept at −20 °C until use. Lipid films were resuspended in MQ H2O (QCM-D) or D2O (NR) to a concentration of 0.2 mg/mL and bath sonicated for 1 h. Just prior to use, the lipid suspension was tip sonicated for 5 min, 5 s on/off at 10% of the maximum power (50 W) using a Fisher Scientific tip sonicator model FB50 and a tip model CL-18. Finally, the vesicle suspension was mixed with equal volumes of CaCl2 having double the desired final concentration.

Neutron Reflection (NR)

Specular NR was used to analyze the structure of the SLB at the solid/liquid interface. NR allows one to obtain structural information perpendicular to the surface in a nondestructive manner. INTER, a time-of-flight reflectometer at the ISIS neutron source (Didcot, U.K.), was used in this study. An incident neutron beam, with a chosen wavelength (λ), is directed on the solid/water interface and partially reflected and refracted depending on the incident angle (θ) and the wavelength (λ). The ratio between the incident and the reflected beams (R) is then measured as a function of the wavevector, which is defined as Q = 4π/λ sin θ. To obtain detailed structural information, the isotopic contrast between the molecules in the sample is important. The scattering length density (SLD) determines the isotopic contrast and highlights different components in the sample. The SLD profile is described by the sum of the coherent scattering length b multiplied by the number of nuclei in a given volume n (SLD = ∑nb). By exchanging hydrogen with deuterium, the SLD of a molecule increases and the isotropic contrast can be manipulated to partially highlight specific parts of a molecule. In this study, the isotropic contrast was achieved by exchanging the bulk solvent with either D2O or H2O. The resolution was set to Δλ/λ = 3% using two incident angles of 0.8 and 2.3°. Homemade solid/liquid flow cells were used and connected to an HPLC pump. The area exposed to the neutron beam was 4.5 × 6.5 cm2. All cell parts made of polyetheretherketone were cleaned by bath sonication in 2% Hellmanex and MQ water. Silicon(111) surfaces were cooked in piranha solution (H2SO4/H2O2 7:3) at 80 °C for 10 min followed by extensive rinsing in MQ water. The data were analyzed using the Motofit software, which uses the Abeles optical matrix method for fitting reflectometry data of thin layers.[38] Three layers were used to model the lipid bilayer: headgroup, tail, headgroup. An additional layer was added between the SiO2 layer and the SLB to represent the solvent trapped in the roughness of the substrate. To keep physically realistic models of lipid bilayers, the lipids were restricted to maintain the same mean molecular area (MMA) of the head and the tailwhere d is the thickness of the layer, ø is the volume fraction, and V is the volume of the headgroup or the tail. Since there are two tails per lipid head group, the area of the tail is multiplied by two. Error estimations associated with each fitted parameter were obtained by varying each parameter individually until a significant change in the χ2 values between the simulated NR data and the experimental data was observed. Theoretical values of the lipid components obtained from MD simulations are given in Supporting Information Table S1.

Quartz Crystal Microbalance with Dissipation (QCM-D)

Measurements were performed using a Q-SENSE E4 system (Q-Sense, Sweden) connected to an HPLC pump through the neutron cells. Silicon oxide sensor crystals, 50 nm (Q-Sense, Sweden), were used. Cleaning protocols were used as specified by Q-Sense. Prior to experiments, the fundamental frequency (5 MHz) and five overtones (3rd, 5th, 7th, 9th, and 11th) were found and recorded in MQ water. The flow rate was set to 1 mL/min, and the temperature was controlled as specified. A baseline was obtained in D2O prior to injection of the vesicle solution. A constant flow of vesicles was maintained until stable signals were obtained before rinsing with excess D2O. The QMC-D measurements were reproducible when running at high flow rate and when flipping the instrument upside down. This is due to less probability of air bubble formation in the QCM-D cell as well as the lack of vesicle deposition due to aggregation and phase separation in the bulk induced by calcium bridging of vesicles (Figure S1 gives DLS data for POPE–POPG vesicles incubated with 1, 2, and 3 mM CaCl2 over time and shows that vesicle aggregation occurs within minutes of mixing the lipids with the calcium solution and stabilizes at around 10 min for 10 mol % PG but takes longer for 25 mol % PG). Moreover, the lipid solution should never be left standing in the cell without flow, but should be rinsed with water right away after successful SLB deposition.