ARTICLE pubs.acs.org/Langmuir
Composition and Asymmetry in Supported Membranes Formed by Vesicle Fusion Hanna P. Wacklin*,† Institut Laue Langevin, 6 rue Jules Horowitz BP 156, 38042 Grenoble, France, and Australian Nuclear Science and Technology Organisation, New Illawarra Road, Lucas Heights NSW 2234, Australia
bS Supporting Information ABSTRACT: The structure and formation of supported membranes at silica surfaces by vesicle fusion was investigated by neutron reflectivity and quartz crystal microbalance (QCM-D) measurements. The structure of equimolar phospholipid mixtures of DLPCDPPC, DMPCDPPC, and DOPCDPPC depends intricately on the vesicle deposition conditions. The supported bilayer membranes exhibit varying degrees of compositional asymmetry between the monolayer leaflets, which can be modified by the deposition temperature as well as the salt concentration of the vesicle solution. The total lipid composition of the supported bilayers differs from the composition of the vesicles in solution, and the monolayer proximal to the silica surface is always enriched in DPPC compared to the distal monolayer. The results, which show unambiguougsly that some exchange and rearrangement of lipids occur during vesicle deposition, can be rationalized by considering the effects of salt screening and temperature on the rates of lipid exchange, rearrangement, and vesicle adsorption, but there is also an intricate dependence on the lipidlipid interactions. Thus, although both symmetric and asymmetric supported bilayers can be prepared from vesicles, the optimal conditions are sensitive to the lipid composition of the system.
1. INTRODUCTION Supported lipid membranes1 are widely used as models of biological cell membranes and as biomimetic interfaces with potential applications in biosensing and biomedicine. Vesicle fusion to surfaces2 is, because of its simplicity, one of the most popular methods of forming supported bilayers and has been successfully applied to chemically different surfaces under conditions ranging from pure water at room temperature to buffers containing several ionic species up to 500 mM concentration and elevated temperatures. The process of supported bilayer formation has been studied in great detail in order to understand the mechanism by which vesicles transform into a continuous membrane. In general, vesicles are thought to adsorb to a surface intact before fusing into a continuous bilayer. Quartz crystal microbalance measurements with dissipation (QCM-D)3 and dual polarization interferometry (DPI)4 have indicated that a continuous bilayer is formed only once a critical density of adsorbed vesicles has been reached but that this behavior can be suppressed in the absence of salts at low ionic strength,5 when the bilayer is formed directly. However, partially or fully ruptured vesicles on the surface have been found to promote the rupture of more vesicles,6 and there is also compelling evidence that the edge of the forming bilayer acts as the primary fusion site for adsorbing vesicles.7 The results of the vesicle fusion process are variable, and the structural symmetry of supported membranes, as represented by domain coupling between the bilayer leaflets, also depends intricately r 2011 American Chemical Society
on the membrane preparation method,8 vesicle fusion conditions,9 and physical properties of the surface10 and can change on the timescales of typical experiments by lipid flip-flop across the bilayer.11 In the majority of investigations, the possible effect of the solid surface on the lipid composition of the supported membranes has been completely ignored. It has always been assumed that the membranes are transferred onto a solid substrate as they are in vesicles (i.e., that no exchange of lipids between the vesicles and the supported bilayer or rearrangement of lipids occurs). We recently discovered using specular neutron reflection that binary mixtures of zwitterionic phosphochatidylcholines DOPC and d62-DPPC can spontaneously form supported bilayers that are asymmetric in composition and also differ in total composition from the vesicles, indicating that some lipid rearrangement takes place during or after vesicle fusion. Similar asymmetric lipid arrangements on solid surfaces have also been observed by AFM9 or indirectly derived from protein affinity data.12 The aim of the current work was to determine how and to what extent the vesicle deposition conditions that are known to affect the membraneformation pathway influence the composition and structure of the supported bilayers. At the same time, the purpose was to
Received: February 22, 2011 Revised: May 2, 2011 Published: May 25, 2011 7698
dx.doi.org/10.1021/la200683e | Langmuir 2011, 27, 7698–7707
Langmuir elucidate the effect of lipid spontaneous curvature and miscibility on the supported bilayer structure. In our previous study, it was found that saturated phosphatidyl choline d62-DPPC was always preferentially located on the proximal side of the membrane facing the silica surface compared to unsaturated lipid DOPC whereas the solution facing the distal side had a composition closer to that of the vesicles. The degree of asymmetry also varied with the vesicle composition. The results suggested that they might be linked to the bulk phase diagram of the mixture, which has a broad area of immiscibility between the gel and fluid phases below the main-phase transition temperature of DPPC (42 °C). However, the cis unsaturated chains of DOPC give it a small negative spontaneous curvature (c0 = 0.1 nm1),13 which could mean that it is preferentially located in the inner leaflet of small, highly curved vesicles and that this may be the origin of the asymmetry in the supported bilayer. In addition, the total composition of the supported membranes differed from that of the vesicles in a manner that suggested that DPPC has a higher affinity for the silica surface than DOPC. In this article, we show that compositional asymmetry is a general feature of phosphatidylcholine (PC) mixtures in supported bilayers and that it can be either enhanced or eliminated by modifying the vesicle deposition conditions. We investigated three different equimolar mixtures in order to elucidate the role of spontaneous lipid curvature and lipidlipid interactions on the membrane structure: DOPC (21 °C)d62-DPPC (39 °C), DLPC(1 °C)d62-DPPC, and DMPC (22 °C)d62-DPPC, where the main gel- to fluid-phase transition temperature for each lipid is given in parentheses. The motivation for this was that DLPCDPPC mixtures exhibit the same kind of gel- to fluidphase immiscibility as DOPCDPPC but DLPC has no preference for interfacial curvature whereas DMPC and DPPC show no such immiscibility and DMPC has no preference for interfacial curvature but there is still a difference between the mainphase transition temperatures of the individual lipids. All of these mixtures form a homogeneous fluid phase at all compositions if the temperature is raised above the main-phase transition temperature of DPPC. By investigating the effect of temperature and salt screening by NaCl on these mixtures, we also wished to understand what role electrostatic interactions, domain formation, and lipid mobility play in the resulting structure. We will show that it is not possible to attribute the asymmetry exclusively to either lipid spontaneous curvature or phase segregation, which rules out the vesicle origin of asymmetry. Instead, we propose that asymmetry is a result of lipid rearrangement at the surface and exchange with vesicles.
ARTICLE
2. MATERIALS AND METHODS
Figure 1. (a) Sensitivity of neutron reflectivity to membrane composition at the siliconD2O interface and simulated reflectivity profiles for a 50 Å membrane. (b) Sensitivity to membrane asymmetry. The pink and blue circles show an experimental data set recorded in D2O and CmSi (38% D2O, contrast matched to silicon), and the simulated data sets show the variation in reflectivity if 10 or 20 mol % deuterated DPPC is added to either the proximal or distal side of the membrane. The black line shows the change in reflectivity observed upon adding 5 mol % on each side of the membrane, or 10 mol % in total. The scattering length density profiles corresponding to the simulated reflectivities are shown in the inset.
2.1. Neutron Reflection. Specular neutron reflectivity measures the neutron scattering length density profile F(z) of a supported membrane, which can be decomposed to quantify the lipid composition by using selective deuteration,14 thus allowing the direct detection of compositional membrane asymmetry. The intensity of reflected neutrons is measured as a function of the scattering vector Q = (4π sin θ)/λ, where θ is the angle of incidence and λ is the neutron wavelength. Figure 1a shows the simulated reflectivity and scattering length density profiles of a lipid bilayer at the siliconD2O interface depending on the percentage of deuteration of the lipids by fitting such optical matrix simulations of reflectivity to experimental data measured in several isotopic solution contrasts, where it is possible to determine the fraction of a perdeuterated lipid in a mixture to within (3 mol %.
The magnitude of compositional asymmetry can be determined unambiguously by neutron reflection provided that the difference in the leaflet composition is at least 10 v/v% as illustrated in Figure 1b, which shows a set of experimental data and simulated reflectivity curves for a DOPCd62-DPPC bilayer in D2O and in 38% D2O demonstrating that it is also possible to distinguish whether the fraction of d62-DPPC is higher on the distal (solution facing) or proximal (facing the silicon support) side of the membrane and that both of these changes are clearly distinguishable from a difference in the total d62-DPPC content of the membrane. Thus, although specular reflectivity cannot detect the membrane lateral structure, it has the unique ability to determine the average lipid composition of 7699
dx.doi.org/10.1021/la200683e |Langmuir 2011, 27, 7698–7707
Langmuir each leaflet without the use of chemical markers. The structural resolution of the measurements is in principle limited by the incoherent scattering background from the aqueous solvent (typically on the order of 106), which determines the accessible q range. To compensate for the restricted q range, measurement of the same membrane in several water contrasts (e.g., H2O, D2O, and their mixtures) allows the thickness and composition information to be effectively decoupled from each other. In general, the resulting structural resolution is of the order of (3 Å in total membrane thickness and (10% in total lipid volume fraction, as has been described previously.14 2.2. Materials and Experimental Procedures. All lipids were purchased from Avanti Polar Lipids, and 99.9% D2O was provided by the ILL or purchased from Sigma-Aldrich. Small unilamellar vesicles (SUV) of 3550 nm average diameter were formed by mixing the lipids in a 1:1 mol ratio in chloroform before drying in a stream of nitrogen and dispersing by vortex mixing and pulsed probe sonication for 10 min at 55 °C. The lipid solutions (0.5 mg mL1) were characterized by dynamic light scattering (ALV CGS-3, ALV GmbH multiangle instrument, Langen) to verify that the vesicle size distributions were all similar (data not shown). The lipid solutions were formed in pure water or D2O or in 100 mM NaCl in pure water or D2O without modifying the pH or using buffering agents. All solutions were maintained for up to 2 h at either 25 or 55 °C depending on the desired deposition temperature until introduction into the neutron reflectivity cell. The solid supports for neutron reflection were silicon single crystals cut along the (111) plane, purchased from Crystran Ltd. U.K. The surfaces were cleaned for 15 min in 1:4:5 H2O2/H2SO4/H2O at 8085 °C, followed by UV ozonolysis in a BioForce Pro UV cleaner (from BioForce NanoSciences). This treatment leaves a natural oxide layer of 1014 Å thickness and ∼5 Å roughness. The surfaces for QCM-D measurements (silica-coated 5 MHz quartz crystal sensors manufactured by Qsense Ltd.) were cleaned by immersion in 2% Hellmanex II detergent solution for 5 min, followed by extensive rinsing and UV/ ozone cleaning. This treatment was found to clean efficiently the silicacoated sensor surfaces without any observable damage or changes in the quality of the QCM-D data. Between experiments, the sensor crystals and flow system were cleaned in situ by rinsing with absolute ethanol (5 min at 100200 μL min1), followed by 2% Hellmanex solution (at 100 μL min1 for 1020 min or until a flat trace was obtained) and extensive rinsing with water (2030 min at 100200 μL min1) until flat, stable frequency and dissipation traces were obtained; otherwise, the crystal was removed from the flow chamber and cleaned ex situ as described above. Neutron reflectivity data were measured on the D17/FIGARO reflectometers at ILL, Grenoble using flow cells built in-house. Timeof-flight measurements were performed using neutron wavelengths of 220 Å on D17 or 230 Å on FIGARO and two angles of incidence covering a momentum-transfer range of 0.0050.3 Å1. With the relatively short measurement times of 0.51 h, reflectivity from these interfaces could be detected down to 106 or q ≈ 0.2 Å1. The silicon substrates were characterized in D2O and H2O or CmSi (a mixture of 62% H2O and 38% D2O contrast matched to the silicon substrate, with F = 2.07 106 Å2) to determine the silicon oxide layer thickness and interfacial roughness. Prior to the introduction of lipid solutions, the sample chamber was equilibrated to either 25 or 55 °C as required. The lipid solutions were incubated over the silica surfaces for 30 min before rinsing with water or NaCl solution at the same temperature, followed by rinsing out any NaCl solution by pure D2O or H2O. Whenever NaCl was used, the cell was rinsed with the solution before and after vesicle deposition to avoid osmotic shock, but before measurement of the bilayer structure, the NaCl solution was rinsed off with copious quantities of water. In QCM-D experiments performed using a Qsense E4 instrument, the vesicles were introduced at a flow rate of 100 μL min1
ARTICLE
for 9 min, the flow was stopped for at least 20 min or until the trace reached a plateau, and the cell was rinsed at 100 μL min1, first with the same solution in which the vesicles were prepared (100 mM NaCl or water) followed by pure water. 2.3. Data Analysis. Neutron reflectivity profiles were analyzed by model fitting using the Motofit program15 keeping the average molecular volumes for the lipid head and chains fixed, as described previously.16 This ensures that the average area occupied by the lipid chains and headgroups is the same in each monolayer but does not impose any other constraints on the fitting. The phospholipid composition was estimated from the fitted scattering length density of the hydrocarbon chain region, from which the average area per molecule was calculated. The procedure of fitting data from binary lipid mixtures was described in full previously.16 Our models consisted of four layers to describe the inner and outer headgroup regions (inner referring to the silicon side) and the inner and outer chain regions.
3. RESULTS AND DISCUSSION Vesicles were deposited either at 55 °C, which is above the miscibility temperature of all of the mixtures investigated, or at 25 °C, at which temperature 1:1 DLPCDPPC and 1:1 DOPCDPPC have coexisting gel and fluid phases, and for the non-phase-separating 1:1 DMPCDPPC, this temperature is between the main-phase transition temperatures of the two individual lipids. The effect of salt screening by 100150 mM NaCl was determined at both temperatures. Before vesicle deposition in situ, neutron reflectivity data was first recorded from each clean silicon (111) surface in order to use the structure of the native SiO2 layer as a reference in analyzing the lipid bilayer structures, which is necessary in order to resolve the internal structure of the lipid bilayers. The full neutron reflectivity data, fits, scattering length density profiles, and tabulated structural data are available in the Supporting Information. The neutron reflectivity data for DOPCDPPC at 55 °C in the absence of NaCl has been published previously,16 and the results are reproduced here for comparison only. 3.1. General Structure and Quality of Bilayers. Table 1 lists the bilayer thicknesses, lipid volume fractions, and areas per molecule in the bilayers formed under different conditions. Although there are small variations in the lipid surface coverage as indicated by the volume fraction, the majority of the membranes fall within 85 ( 5 v/v% lipid. (This value represents an average over the entire macroscopically large surfaces of 80 mm 50 mm.) The areas per molecule shown in Table 1 correspond to the average areas available per lipid molecule calculated from the bilayer thickness and lipid volume fraction. Although this calculated area does not directly represent the area per molecule but rather the surface coverage of the bilayer, the values correlate quite well with the range of literature values published for these lipids in their liquid-crystalline bulk lamellar phases.17 The thickness of the bilayers does not vary by more than 6 Å between deposition conditions for any of the mixtures and also corresponds closely to the published values for the lipids in the bulk phase. The roughness of the silica surfaces was on average 5 ( 2 Å, and the roughness of the supported bilayers was 3 ( 1 Å, with no significant differences between samples. 3.2. Supported Membrane Composition. Table 1 shows the total lipid composition of the supported membranes represented as the fraction of the second lipid xDXPC in each bilayer, where DXPC = DLPC, DMPC, or DOPC, with the error in the values corresponding to the uncertainty in the scattering length density of the lipid chains ((0.2 106 Å2, which translates to (3% in 7700
dx.doi.org/10.1021/la200683e |Langmuir 2011, 27, 7698–7707
Langmuir
ARTICLE
Table 1. Quality of Lipid Bilayersa DLPC
thickness/ Å
Amol/Å2
lipid vol %
xDLPC
25 °C
42 ( 3
65 ( 19
85 ( 5
0.61 ( 0.03
25 °C NaCl
47 ( 3
55 ( 11
89 ( 5
0.51 ( 0.03
55 °C
47 ( 3
59 ( 12
92 ( 5
0.52 ( 0.03
55 °C NaCl
41 ( 3
63 ( 11
88 ( 5
0.55 ( 0.03
DMPC
thickness/ Å
Amol/ Å 2
lipid vol %
xDMPC
25 °C
49 ( 3
68 ( 11
80 ( 5
0.46 ( 0.03
25 °C NaCl
46 ( 3
65 ( 11
82 ( 5
0.56 ( 0.03
55 °C 55 °C NaCl
43 ( 3 44 ( 3
74 ( 15 6 ( 11
80 ( 5 82 ( 5
0.46 ( 0.03 0.51 ( 0.03
DOPC
thickness/ Å
Amol/ Å 2
lipid vol %
xDOPC
25 °C
46 ( 3
64 ( 13
92 ( 5
0.58 ( 0.03
25 °C NaCl
46 ( 3
62 ( 17
95 ( 5
0.53 ( 0.03
55 °C
46 ( 3
83 ( 13
70 ( 5
0.26 ( 0.03
55 °C NaCl
46 ( 3
68 ( 9
70 ( 5
0.50 ( 0.03
a
Bilayer thickness, surface area avaible per molecule, total lipid volume fraction, and mole fractions of DLPC, DMPC, and DOPC in DLPCDPPC, DMPCDPPC, and DOPCDPPC bilayers measured by neutron reflection under different conditions.
the mol fraction). Although in many cases xDXPC falls within the errors of the expected value of 0.5, in several more cases there are measurable differences up to the most dramatic case of DOPCDPPC at 55 °C in which xDOPC is only 0.26. Screening by NaCl brings the membrane composition closer to the expected value compared to adsorption in the absence of salt at both temperatures. Interestingly, in the absence of salt, DLPCDPPC membranes are enriched in DLPC relative to the vesicles, whereas for DMPCDPPC they contain less than 50 mol % DMPC. Decreasing temperature has the effect of increasing the DLPC enrichment for DLPCDPPC but reverses the enhancement for DOPCDPPC to favor DOPC and has very little effect on DMPCDPPC. The differences in membrane composition show no clear correlation with the lipid chain length or phase segregation of the mixture. 3.3. Membrane Asymmetry. Figure 2 shows the monolayer compositions as the mole fractions of DLPC, DMPC, and DOPC, and in Figure 3, the bilayer asymmetry is represented as the ratio of xDPPC in the proximal to distal leaflet. The asymmetry is always in the same direction (i.e., there is more DPPC in the proximal leaflet), but the ratio varies between 1.05 (xDPPC is 0.02 higher in distal leaflet) and 1.64 (xDPPC is 0.36 higher in distal leaflet). At 25 °C, the asymmetry decreases in the order DLPC > DMPC > DOPC. In the absence of NaCl, increasing temperature reverses the trend in asymmetry, which increases in the order DLPC < DMPC < DOPC. As a result, the xDPPC ratio is reduced for DLPC and DMPC at 55 °C but not for DOPC. The presence of salt reduces the asymmetry at both temperatures for all lipids, but at 55 °C, there is no clear trend between the lipids. Only DMPC forms fully symmetric membranes at 55 °C in the presence of NaCl, although the asymmetry of DOPCDPPC at 25 °C in NaCl also falls within the error limits (represented by the error bars in Figure 3) of the measurement, which is based on the detection sensitivity of 3 mol % in the lipid composition of each monolayer. At 55 °C, all of the mixtures are in the homogeneous fluid phase in vesicles, but at 25 °C, both DLPCDPPC and
Figure 2. Composition of bilayers prepared under different conditions showing the mole fraction of DXPC in the proximal and distal bilayer leaflets for (a) DLPCDPPC mixtures, (b) DMPCDPPC mixtures, and (c) DOPCDPPC mixtures The error in determining each monolayer mole fraction is 3 mol %, as represented by the error bars.
DOPCDPPC are in the two-phase region of the bulk phase diagram and the vesicles have pre-existing domains. The observed temperature trends are not consistent with lipid immiscibility being the driving force behind establishing the asymmetry because this should enhance the asymmetry at 25 °C, and this is not observed. If a preferential interaction of the silica surface with 7701
dx.doi.org/10.1021/la200683e |Langmuir 2011, 27, 7698–7707
Langmuir
ARTICLE
Figure 3. Bilayer asymmetry between the proximal and distal membrane leaflets represented as the ratio xDPPCproximal/xDPPCdistal.
gel-phase lipids compared to fluid-phase lipids were the cause of the asymmetry, then this should also enhance the asymmetry at 25 °C. If the membranes were transferred from the vesicles to the surface exactly as they are, then this would imply that the vesicles must be asymmetric and the formation of the domains after cooling the supported membrane leads to the observed differences in asymmetry. Although it is possible that the surface affects the nucleation of domains, the differences in composition between the supported membranes and the vesicles do not support the “transfer as is” scenario and show that either surfacevesicle and/or surfacelipid interactions lead to the modification of the supported bilayer composition. The composition and asymmetry differences among the three lipid mixtures also imply that the specific lipidlipid and/or lipidsurface interactions lead to the compositional variation among them. For DOPC, it is possible that the small preference for curvature could lead to it being preferentially located in the inner, negatively curved leaflet of the small unilamellar vesicles. However, if the curvature preference were a factor in establishing the asymmetry, then it would not be exhibited by the DMPCand DLPC-containing mixtures as observed because the lipid packing parameter (V/aL, where V is the chain volume, a is the area per headgroup and L is the chain length) remains approximately constant at 0.75 for saturated diacylphophatidylcholines above a chain length of 12 carbons (= DLPC)18 and there is therefore no driving force for curvature-induced asymmetry. On the basis of the results from neutron reflection, we can conclude that the structure of supported membranes is not a straightforward function of the lipid properties or lipidlipid interactions but is modulated by the deposition temperature and salt screening in a manner that implies a role for electrostatic interactions between the vesicles and the surface and a role for membrane fluidity. 3.4. Vesicle Deposition Pathway and Kinetics. To understand how temperature and salt change the membrane-formation process from vesicles, we used dissipative quartz crystal microbalance measurements (QCM-D)19 to determine the pathway of vesicle fusion.20 Figures 47 show the frequency and dissipation changes (seventh overtone) of a 5 MHz quartz crystal monitored during vesicle adsorption. As a first approximation, the magnitude of the frequency decrease is proportional to the mass per
Figure 4. QCM-D frequency and dissipation data (seventh overtone) recorded during vesicle adsorption at 25 °C for (a) DLPCDPPC, (b) DMPCDPPC, and (c) DOPCDPPC.
unit area adsorbed on the crystal, whereas the magnitude of the disspation change reflects the viscoelasticity of the film, which is increased by increasing water content. 7702
dx.doi.org/10.1021/la200683e |Langmuir 2011, 27, 7698–7707
Langmuir
Figure 5. QCM-D frequency and dissipation data (seventh overtone) recorded during vesicle adsorption at 50 °C for (a) DLPCDPPC, (b) DMPCDPPC, and (c) DOPCDPPC.
In the absence of NaCl, all of the mixtures show a small dissipation peak of (0.51.5) 106 during vesicle adsorption, which is accompanied by a frequency shift of 2530 Hz that is
ARTICLE
Figure 6. QCM-D frequency and dissipation data (seventh overtone) recorded during vesicle adsorption at 25 °C in 100 mM NaCl for (a) DLPCDPPC, (b) DMPCDPPC, and (c) DOPCDPPC.
typical for a single lipid bilayer with full surface coverage. The dissipation peak corresponds to the rupture of the adsorbing vesicles,21 which occurs on the same timescale as the bilayer formation because no deep minimum in the frequency is seen that would correspond to the weight of intact vesicles exceeding 7703
dx.doi.org/10.1021/la200683e |Langmuir 2011, 27, 7698–7707
Langmuir
Figure 7. QCM-D frequency and dissipation data (seventh overtone) recorded during vesicle adsorption at 50 °C in 100 mM NaCl for (a) DLPCDPPC, (b) DMPCDPPC, and (c) DOPCDPPC.
the mass of a bilayer. The small residual dissipation of
