Asymmetric Bilayers by Hemifusion: Method and Leaflet Behaviors

Abstract

We describe a new method to prepare asymmetric giant unilamellar vesicles (aGUVs) via hemifusion. Hemifusion of giant unilamellar vesicles and a supported lipid bilayer, triggered by calcium, promotes the lipid exchange of the fused outer leaflets mediated by lipid diffusion. We used different fluorescent dyes to monitor the inner and the outer leaflets of the unsupported aGUVs. We confirmed that almost all newly exchanged lipids in the aGUVs are found in the outer leaflet of these asymmetric vesicles. In addition, we test the stability of the aGUVs formed by hemifusion in preserving their contents during the procedure. For aGUVs prepared from the hemifusion of giant unilamellar vesicles composed of 1,2-distearoyl-sn-glycero-3-phosphocholine/1,2-dioleoyl-sn-glycero-3-phosphocholine/cholesterol = 0.39/0.39/0.22 and a supported lipid bilayer of 1,2-dioleoyl-sn-glycero-3-phosphocholine/cholesterol = 0.8/0.2, we observed the exchanged lipids to alter the bilayer properties. To access the physical and chemical properties of the asymmetric bilayer, we monitored the dye partition coefficients of individual leaflets and the generalized polarization of the fluorescence probe 6-dodecanoyl-2-[ N-methyl-N-(carboxymethyl)amino] naphthalene, a sensor for the lipid packing/order of its surroundings. For a high percentage of lipid exchange (>70%), the dye partition indicates induced-disordered and induced-ordered domains. The induced domains have distinct lipid packing/order compared to the symmetric liquid-disordered and liquid-ordered domains.

Significance

This work reports an innovative method to prepare asymmetric giant unilamellar vesicles. This asymmetry is a key property of cell plasma membranes, but it is largely unexplored. Asymmetric domains have different physical chemical properties compared to symmetric domains.

Introduction

The complexity of the plasma membrane (PM) extends beyond the total composition of thousands of lipid species to include the compositional differences between the two leaflets (1, 2). Cell PM inner and outer leaflets are composed of different lipid mixtures that have distinct lateral organizations. The PM exoplasmic leaflet is composed of a highly nonrandom mixture of lipids, which in model membranes can be described as separated liquid-disordered (Ld) and liquid-ordered (Lo) phases. In contrast, the lipid composition of the PM inner leaflet lacks sufficient lipids with high melting points and is likely a single phase (3). Communication between the extra- and intracellular environments must be transduced through the membrane, facilitated by coupling between the inner and the outer leaflets. In this study, we examine the interleaflet interactions of asymmetric giant unilamellar vesicles (aGUVs).

To prepare asymmetric bilayers is challenging (4, 5, 6). An inverted emulsion method can be used to prepare asymmetric vesicles (7, 8), but one concern is that organic solvent is trapped in the bilayer. Lipid exchange also can be mediated by cyclodextrin (CD), a cyclic oligosaccharide that promotes the exchange of outer leaflet lipids of symmetric vesicles with a pool of donor lipid vesicles (9). CD promotes lipid exchange by fitting around cholesterol (chol) or individual lipid acyl chains, thereby solubilizing these in the aqueous medium. A concern in use of CD is lingering contamination of CD in the asymmetric vesicles. Another issue is that an ensemble of vesicles is produced with a wide variation in levels of asymmetry that could mask the actual properties of completely asymmetric bilayers.

We report making aGUVs by hemifusion between a supported lipid bilayer (SLB) and a giant unilamellar vesicle (GUV). Our method is completely free of intermediate steps that involve organic solvents or CD and is therefore free of any exogenous contaminant except for trace fractions of fluorescent dyes. Lipid exchange is inferred from the percentage of fluorescent probes that were exchanged. Hemifusion can be induced by divalent cations (10, 11, 12), osmotic and pH changes (13), peptides and proteins that induce fusion (12, 14), or with an apparatus to control the contact force (15, 16). Here, we establish experimental conditions for Ca²⁺-induced hemifusion between symmetric GUVs and an SLB. In brief, 1) symmetric GUVs and the SLB were prepared with different fluorescent dyes, 2) an aGUV is formed when symmetric GUVs and the SLB hemifuse and exchange lipids between their outer leaflets, then 3) aGUVs are detached from the SLB (see Fig. 1 A). We establish controls to confirm that the newly exchanged lipids in the aGUVs, here named “guest lipids,” are located in the aGUV outer leaflet. We observed stable, nonleaking aGUVs as a result of hemifusion. Lastly, we studied the influence of fluid outer leaflet opposed to a phase-separated inner leaflet. The outer leaflet displays induced domains coupled to genuine phase separation in the inner leaflet, as found by London and colleagues (9). These induced domains have different properties compared to the phase-separated domains in a symmetric GUV.

Figure 1.

Experimental procedure and analyses. (A) A sketch of the hemifusion between a supported lipid bilayer (SLB) and a symmetric GUV is shown. GUVs and SLB were prepared with different fluorescent labels represented by different colored lipids. After hemifusion, aGUVs are detached from the SLB. (B) A comparison of symmetric and aGUVs is shown. Examples of symmetric GUVs are shown as follows (from left to right): control GUV 1 + red DiD, control GUV 2 + green TFPC, and example of an aGUV; the image shows the red, green, and merged channels of the microscope for one aGUV. Line scans are displayed in white, indicating θ = 0. (C) Intensity profiles of the line scans shown on (B), where the intensity peak on the vesicle radius R is measured. The left panel shows a comparison between the intensity of DiD for symmetric and aGUV; the right panel shows a comparison between the intensity of TFPC for symmetric and aGUV. Lipids and TFPC of the SLB replace the lipids and DiD of the GUV. (D) The intensity measured on the GUV or aGUV radius as a function of θ is shown. To see this figure in color, go online.

Materials and Methods

Chemicals

DOPC (1,2-dioleoyl-sn-glycero-3-phosphocholine), 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine, and DSPC (1,2-distearoyl-sn-glycero-3-phosphocholine) were from Avanti Polar Lipids (Alabaster, AL), and chol was from Nu-Chek Prep (Elysian, MN). HEPES, KCl (potassium chloride), CaCl₂ (calcium chloride), and EDTA were from Sigma-Aldrich (St. Louis, MO). Thin-layer chromatography of ∼20 μg lipid samples confirmed the purity of lipids as >99%. Phospholipid concentration was determined to <1% error using inorganic phosphate assay (17). Fluorescent dyes: TopFluor PC, a lipid analog dye here named TFPC (1-palmitoyl-2-(dipyrrometheneboron difluoride) undecanoyl-sn-glycero-3-phosphocholine), and NBD-DOPE (NBD-PE), a lipid analog dye (1,2-dioleoyl-sn-glycero-3-phosphoethanolamine-N-(7-nitro-2-1,3-benzoxadiazol-4-yl)), shown in the plots as NBD-PE, were from Avanti Polar Lipids; DiD (1,1′-dioctadecyl-3,3,3′,3′- tetramethylindocarbocyanine, 4-chlorobenzenesulfonate salt) from Thermo Fisher Scientific (Molecular Probes) (Waltham, MA) and C-Laurdan (6-dodecanoyl-2-[ N-methyl-N-(carboxymethyl)amino] naphthalene) were from SFC (Cheongju-si, Chungcheongbuk-do, Korea). Dyes were prepared as stock solutions in chloroform. Thin-layer chromatography confirmed >99% purity of the dye stocks.

Preparing SUVs for lipid deposition of the supported bilayers

We prepared 500 μL of liposomes with total lipid concentration of 2.5 mM having a small fraction of fluorescent dye. Dye/lipid ratios used in this work were TFPC (1/1500), DiD (1/2500), NBD-PE (1/500), and C-Laurdan (1/100). Sonicated unilamellar vesicle (SUVs) were prepared by sonication for ∼20 min using a circular bath sonicator (Sonblaster; Narda Ultrasonics Corporation, Mineola, NY). Liposomes were prepared in buffer A: HEPES 25 mM, KCl 35 mM (pH 7.43), 100 mOsm/kg. The osmolality of buffers and solutions used in this work was measured using osmometer model 5004 (Precision Systems, Natick, MA).

Preparing SLBs

SLBs were prepared in Lab-Tek II chambered cover glass dishes (155382; Thermo Fisher Scientific). Each dish had four chambers of 1 × 2 cm. The cover slide of each chamber was prewashed with a freshly prepared dilute solution of KOH (Sigma-Aldrich) ∼1 M in ethanol. The chamber was then rinsed thoroughly with Milli-Q water (Thermo Fisher Scientific). After air drying, the chamber was further dried with a flow of N₂. Then the chamber was plasma cleaned for 2 min. After the dish was removed from the plasma cleaner/sterilizer PDC-3 G (Harrick Plasma, Ithaca, NY), 500 μL of the SUV dispersion mixed with 1 M NaCl at 1:1 by volume was dispensed on the cover slides of each chamber. Chambers were kept at 4°C for 2 h. The dish, having a total volume of 2 mL per chamber, was immersed in 2 L of Milli-Q water. SLBs were gently rinsed with water using a 20-mL syringe to remove lipid aggregates loosely attached on the SLB surface. The Milli-Q water in the chamber was then replaced by a buffer A + 1 mM CaCl₂.

GUV preparation

GUVs were prepared using electroformation (18), with modifications. For details, see Supporting Materials and Methods.

GUVs and SLB hemifusion

Using a pipettor (Gilson, Middleton, WI) with a large orifice tip (VWR International, Radnor, PA), we collected 15–20 μL of symmetric GUVs and dispensed them inside the SLB chamber filled with 1.0 mL of buffer A + 1 mM CaCl₂. Aliquots of 3–5 μL of GUVs were placed over the SLB. GUVs settled and hemifused with the SLB. Then, the calcium concentration was increased to 3–7 mM by adding 5–10 μL of buffer B: HEPES 25 mM, KCl 25 mM, CaCl₂ 10 mM (pH 5), 103 mOsm/kg. After 20–30 min, ∼200–500 μL of buffer C was added: HEPES 25 mM, KCl 30 mM, EDTA 5 mM (pH 7.40), 103 mOsm/kg. EDTA was used to chelate Ca²⁺ to prevent further fusion. The salt concentration was adjusted to match the osmolality. To harvest hemifused aGUVs, a 1000-μL pipettor with a large orifice tip was used, pipetting ∼200 μL gently up and down with the pipette tip ∼0.2 cm from the chamber bottom. This procedure gently shears the GUVs off the supported bilayer. The released a GUVs can be placed in a different chamber or slide for study.

Measuring the calcium concentration

To measure the Ca²⁺ concentration of all buffers and samples, the Ca²⁺-sensitive fluorescent dye Rhod-5N (Invitrogen, Carlsbad, CA) was used. Details and results are shown in the Fig. S3.

Quenching assays using sodium dithionite

Cuvette quenching assay using large unilamellar vesicles

Liposomes composed only of DOPC were prepared using rapid solvent exchange (19) at a final lipid concentration of 0.5 mM in buffer A. Vesicles were prepared with a single dye using the dye/lipid ratios described above. After rapid solvent exchange, the lipid dispersion was extruded 31 times using a miniature extruder (Avanti Polar Lipids), with a polycarbonate membrane of pore size 100 nm. Samples were diluted in a 1 × 1 cm cuvette containing 1 mL of buffer A to a final lipid concentration of 50 μM. Using a fluorimeter (Hitachi model 7000, Tokyo, Japan) we monitored the fluorescence of TFPC, DiD and NBD-PE, and light scattering at λ = 400 nm as a function of time. After adding dithionite to a final concentration of 5 μM, we observed the quenching kinetics (Fig. 4 A). After ∼1 h, 10 μL Triton X-100 (Sigma-Aldrich, St Louis, MO, USA) at 10% w/v was added to solubilize the large unilamellar vesicles (LUVs), leading to the total quenching of the fluorescent probes because of the accessibility of the quenching agent. The stock solution of 1 M dithionite was always freshly prepared in Tris buffer (25 mM (pH 10)) as described (5, 20, 21).

Figure 4.

Quenching assay of LUVs using sodium dithionite. (A) Normalized fluorescence (arb. u.) as a function of time for 100 nm LUVs prepared with a single dye is shown: NBD-PE (solid square), DiD (solid circle), or TFPC (open diamond). (B) Normalized light scattering of LUVs simultaneously recorded as a function of time is shown. Arrow 1 indicates the time at which dithionite was added. Arrow 2 indicates the addition of Triton X-100 (10% w/v). Dithionite efficiently quenches NBD-PE and DiD. To see this figure in color, go online.

GUV fluorescence quenching assay

GUVs were prepared as described above and labeled with a single dye, TFPC or DiD or NBD-PE. In the chambered cover glass dish, ∼30 μL of GUVs was diluted with 1 mL of buffer A. For the fluorescence quenching assay, we added 3 μL of 1 M dithionite to the solution of symmetric GUVs. Then we compared the dye intensities of these GUVs in the presence and absence of the quencher (see Fig. S10). GUVs were imaged 15–40 min after the addition of dithionite. This procedure was repeated four times, once for each of the four chambers of the dish. Dithionite was used to quench the outer leaflet fluorescence of aGUVs. These aGUVs were prepared with two different dyes to label the inner and the outer leaflet.

Imaging using a confocal microscope

Imaging was performed with a confocal microscope, Nikon Eclipse C2+ (Nikon Instruments, Melville, NY). For comparison of intensities of symmetric controls and aGUVs, we used the same microscope setup (see Supporting Materials and Methods).

The fluorescence emission of some dyes displays a polarization artifact (22, 23). We corrected this effect by adding a quarter wavelength plate (Thermo Fisher Scientific) in the excitation path. For details, see Fig. S2.

Two-photon microscope spectral imagining

Spectral imaging was performed on a Zeiss 880 inverted confocal microscope (Oberkochen, Germany), equipped with spectrally resolved detection with gallium arsenide phosphide sensitivity, and Spectra-Physics InSight laser (Santa Clara, CA) with 1040 nm and tunable 680–1300 nm multiphoton lines. For the C-Laurdan excitation, we used two-photon excitation at 780 nm. The λ detection was collected in a range of 415–610 nm through 23 channels. In addition, a separate channel was used to monitor the emission of DiD at 650–670 nm.

Analyses

Intensity analyses

The next steps summarize analyses based on the comparison of intensities of symmetric and aGUVs.

• 1)A group of symmetric GUV controls was prepared with a single dye (e.g., DiD shown as GUV 1 in Fig. 1 B or TFPC shown as GUV 2 in Fig. 1 B).
• 2)Intensity profiles were measured in both symmetric GUVs as a control and aGUVs. Fig. 1 C shows a comparison of the intensity profiles obtained for symmetric GUVs and an aGUV.
• 3)Intensities at the peak of 360 intensity profiles were collected and averaged (details below). The percentage of lipid exchange is as follows:

$$\text{lipid exchange }(\%)=\frac{\langle {\text{I}}_{aGUVs}\rangle }{\langle {\text{I}}_{GUVs}/2\rangle }\times 100\%,$$ (1)

where $\langle {\text{I}}_{aGUVs}\rangle$ and $\langle {\text{I}}_{GUVs}\rangle$ are the average intensities measured in asymmetric and symmetric GUVs, respectively. The factor 1/2 yields the intensity expected on one leaflet of the control symmetric GUV. The lipid exchange is inferred from the percentage of fluorescent probes that were exchanged. This percentage is independently calculated for both dyes, and a correlated plot shows their agreement (Figs. 3, 5, and 7).

Figure 3.

Percentage of lipids exchanged after hemifusion. aGUVs prepared from the hemifusion of GUVs labeled with DiD and an SLB labeled with TFPC are shown. (A) Bar plots represent the intensity levels observed in symmetric (solid bar). Dots on the side of solid bars represent the measurements of different symmetric GUVs for which the average intensity was calculated from 360 line scans (see text). Open bars show different percentage of exchanged lipids (right axis) calculated by the comparison of intensities in symmetric and asymmetric vesicles. Errors were calculated from the standard deviation of 360 intensity values for each dye. (B) The percentage of exchanged lipids was independently calculated for DiD and TFPC, [Ca²⁺] = 4–7 mM. Line of slope 1 shows the agreement of the calculated percentage of lipid exchange for DiD and TFPC. Errors were calculated by propagating the uncertainty obtained in the fluorescence signal. To see this figure in color, go online.

Figure 5.

Fluorescence quenching of an aGUV outer leaflet confirms hemifusion. (A) aGUVs prepared from the hemifusion of GUVs labeled with DiD and an SLB labeled with NBD-PE are shown. The percentage of exchanged lipids was independently calculated for DiD and NBD-PE (see text; [Ca²⁺] = 5–8 mM). Errors were calculated by propagating the uncertainty obtained in the fluorescence signal. (B) After the addition of dithionite, the fluorescence of DiD and NBD-PE in the outer leaflet is quenched. The normalized fluorescence of DiD is reduced to half of the value of the unquenched case (left) (n = 21). For a sample of 21 aGUVs, only 5 aGUVs display fluorescence signals of NBD-PE after quenching, with an averaged normalized intensity equivalent to 4% of the value of the unquenched case. Errors were calculated from the standard deviation of 360 intensity values for each dye. The exchanged lipids that come from the SLB are mainly found on the aGUV outer leaflet. To see this figure in color, go online.

Figure 7.

Changes in physical chemical properties of aGUVs. (A) Phase diagram of DSPC/DOPC/chol. The lipid composition of the symmetric GUVs and the inner leaflet of aGUVs are described by the star (middle point of tieline A). aGUVs prepared from the hemifusion of GUVs composed of DSPC/DOPC/chol = 0.39/0.39/0.22 + DiD and an SLB composed of DOPC/chol = 0.8/0.2 + TFPC are shown. For aGUVs, the lipid composition of the outer leaflet follows the trajectory described by the arrow as the percentage of lipid exchange increases. (B) The aGUVs result from lipid exchange between GUVs that hemifused with the SLB. The percentage of exchanged lipids was independently calculated for DiD and TFPC (see text; [Ca²⁺] = 5–8 mM). (C) The apparent partition coefficient, Kp^app, of TFPC as a function of percentage of exchanged lipids is shown. TFPC labels the outer leaflet. (D) The partition coefficient of DiD as a function of exchanged lipids after quenching by dithionite is shown. The Kp of DiD refers to the inner leaflet. Error bars shown in (B)–(D) were calculated by propagating the uncertainty obtained in the fluorescence signal. To see this figure in color, go online.

Measuring intensities on GUV and aGUVs

To measure $\langle {\text{I}}_{GUVs}\rangle$ and $\langle {\text{I}}_{aGUVs}\rangle$, we wrote a script in ImageJ to draw line scans in the images at the GUV equatorial plane. We drew 360 line scans per GUV (i.e., one line scan for every 1° (see Video S1)). Fig. 1 D shows intensities of TFPC and DiD measured in an aGUV that correspond to ∼50% of the values observed in the symmetric GUV controls, leading to a percentage of lipid exchange of ∼100%, by Eq. 1.

Calculation of average intensities, ${\text{I}}_{GUVs}$ and ${\text{I}}_{aGUVs}$ is slightly different for uniform and phase-separated GUVs or aGUVs. For uniform GUVs or aGUVs, we averaged the 360 peak intensities. For symmetric GUVs, we repeated this procedure for n = 15–25 GUVs.

For both phase-separated GUVs and aGUVs, the intensity of Lo and Ld phases are treated separately: we averaged the peak intensities from the Ld phase $\langle I_{Ld}\rangle$ and separately averaged the peak intensities from the Lo phase $\langle I_{Lo}\rangle$. Then, we calculated an average total intensity as follows:

$$\overline{I}={\chi }_{Ld}\langle I_{Ld}\rangle +{\chi }_{Lo}\langle I_{Lo}\rangle ,$$ (2)

where χ_Ld and χ_Lo are the fractions of Ld and Lo phases, and $\langle I_{Ld}\rangle$ and $\langle I_{Lo}\rangle$ are the average intensities measured from line scans of the Ld and the Lo phases. Here, we compared symmetric and aGUVs with similar $\langle {\chi }_{Ld}\rangle$ and $\langle {\chi }_{Lo}\rangle$, and dye fluorescence intensities at the domain phase boundary are not considered in these analysis (see Supporting Materials and Methods).

Measuring the dye partition coefficient in phase-separated GUVs

The dye partition coefficient is the ratio of dye concentration in Ld and Lo phases (Eq. 3) as follows:

$$Kp=\frac{\raisebox{1ex}{$\langle I_{Ld}\rangle \times S_{Ld}$}\!\left/ \!\raisebox{-1ex}{${\varphi }_{Ld}$}\right.}{\raisebox{1ex}{$\langle I_{Lo}\rangle \times S_{Lo}$}\!\left/ \!\raisebox{-1ex}{${\varphi }_{Lo}$}\right.},$$ (3)

where ϕ is the phase-specific fluorescence quantum yield, and S represents the surface area of each phase. The quantum yields were measured from emission spectra at the end points of the tieline for TFPC in DSPC/DOPC/chol (24). If the area occupied by Ld and Lo phases were directly proportional to the mole fractions, then the term S_Ld/S_Lo would cancel because our GUVs had χ_Ld ≈ χ_Lo ≈ 0.5. However, the area per lipid for Ld is ∼30% greater than that for Lo (25, 26, 27). Because the dyes studied in this work exhibit different brightness in the Ld compared to Lo phase, we corrected the Kp for the phase-specific fluorescence quantum yield. Here, Kp > 1 favors the Ld phase. A comparison of symmetric and aGUVs with different Kp values could introduce an error of up to 4% in the calculation of percentage of exchange for the case of aGUVs analyzed here. This error was included in the calculations of the percentage of exchange. Details are in the Supporting Materials and Methods.

Measuring GP of C-Laurdan

Analyses of C-Laurdan generalized polarization (GP) were performed as in previous reports (28), with modifications. We analyzed only the intensities in the GUV equator. We used ImageJ code described above to collect intensity profiles of 360 line scans. We collected the intensities at the GUV equator radius I (θ) for 23 wavelengths from λ = 415–610 nm, as described above. Then, we obtained the intensities as a function of λ for every θ. The spectra for Ld and Lo phases were treated separately, fitted using a Gaussian function for each phase. The GP is then calculated according to Eq. 4 as follows:

$$GP=\frac{I_{440}-I_{490}}{I_{440}+I_{490}},$$ (4)

where I₄₄₀ and I₄₉₀ are the intensities obtained from the fittings at λ = 440 and 490 nm and correspond to the λ of the intensity peak of pure Ld and Lo phases (28), respectively. Finally, we calculated a histogram of GP values per GUV or aGUV, shown in Fig. 8 F. We calculated the mean and the error for the mean, which is the half width of each distribution corresponding to Ld and Lo phases, as discussed and shown in Table 1.

Figure 8.

C-Laurdan reports on lipid packing in symmetric and aGUVs. (A) Spectral images of symmetric and (B) aGUVs using C-Laurdan are shown. (C) A color map of GP values of (left) a symmetric GUV and (right) an aGUV is shown. (D) A comparison between blue-shifted spectra, indicating ordered domains, for symmetric GUVs and aGUVs is shown. (E) A comparison between red-shifted spectra, indicating disordered domains, for symmetric and aGUVs is shown. (F) Frequency counting of GP values measured on the GUV and aGUV is shown. aGUVs, with the outer leaflet enriched in DOPC, display considerable difference in the lipid packing/order compared to the symmetric GUV. The percentage of lipid exchange for this aGUV is ∼85 ± 3% (B and C). Scale bars, 10 μm (A and B). ΔGP = <GP>(ordered domains) − <GP>(disordered domains). To see this figure in color, go online.

Table 1.

GP of C-Laurdan Measured in Symmetric and aGUVs

	<GP> Ld	<GP> Lo	ΔGP	% Exchange
DOPC Symmetric	−0.23 ± 0.10	single Ld phase		0
	−0.25 ± 0.05	single Ld phase		0
DOPC/chol Symmetric	−0.18 ± 0.10	single Ld phase		0
	−0.17 ± 0.05	single Ld phase		0
DSPC/DOPC/chol Symmetric	−0.15 ± 0.08	0.32 ± 0.10	0.47	0
	−0.13 ± 0.07	0.32 ± 0.08	0.45	0
	−0.11 ± 0.05	0.37 ± 0.08	0.48	0
aGUVs	−0.22 ± 0.08	0.13 ± 0.08	0.35	85 ± 3
	−0.23 ± 0.07	0.15 ± 0.08	0.38	82 ± 3
	−0.20 ± 0.07	0.19 ± 0.06	0.39	78 ± 4
	−0.18 ± 0.07	0.20 ± 0.06	0.38	77 ± 4
	−0.16 ± 0.1	0.23 ± 0.09	0.39	75 ± 3

Symmetric GUVs composed of DOPC or DOPC/chol = 0.8/0.2 have a single Ld phase, whereas phase-separated GUVs composed of DSPC/DOPC/chol = 0.39/0.39/0.22 have coexisting Ld + Lo phases. aGUVs have an inner lipid composition DSPC/DOPC/chol = 0.39/0.39/0.22 and an outer leaflet replaced by DOPC/chol at different percentages of exchanged lipids. Each row of the table represents the analyses, as described in Fig. 8, of individual vesicles.

Results

We report a new, method to prepare aGUVs using hemifusion. Hemifusion is the fusion of two proximal lipid layers, the outer leaflets of two independent bilayers, in which these fused layers freely exchange their lipids. The complete fusion of the two bilayers, here called full fusion, has been studied in processes such as virus assembly (14). Hemifusion is commonly seen as an early step of full fusion (29, 30). Here, we make use of lipid exchange by hemifusion and establish experimental conditions to minimize the full fusion of vesicles.

Fig. 2 shows a simple case of lipid exchange between an initially symmetric GUV and an SLB. The symmetric GUVs were prepared with the red DiD, and the SLB was prepared with the green TFPC. After hemifusion, the new aGUV now contains the fluorescence of the dye that was originally and exclusively found on the SLB. Moreover, the SLB, immediately after detaching an aGUV, shows a patch of the red fluorescent dye that originally labeled the symmetric GUV. In the course of a few minutes, this patch spreads and becomes diluted on the supported bilayer. TFPC is a two-chain phospholipid that reports the behavior of the lipids that are initially in the SLB and migrates along with the other SLB lipids into the outer leaflet of the new aGUV. Equilibration of the outer leaflet is attained by the diffusion of lipid and dyes. Notice that the SLB is an abundant reservoir of lipids, with the number of lipids on the SLB (dimensions 1 × 2 cm) hundreds of times larger than the number of lipids in a few hundred GUVs that are in contact with the SLB.

Figure 2.

Microscopy images reveal the hemifusion process. (A) Shown is a hemifusion between a supported lipid bilayer (SLB) of DOPC + TFPC (green) and a GUV of DSPC/DOPC/chol + DiD (red), forming an aGUV that displays both dyes. (B) The SLB immediately after detaching the aGUV is shown. The following are shown from left to right: a three-dimensional view of the microscope red channel, which reveals DiD fluorescence; a three-dimensional view of the microscope green channel, which reveals TFPC fluorescence and red and green merge channels. To see this figure in color, go online.

Later in this section, we discuss the properties of phase-separated GUVs as shown in Fig.2. We divide the Results in two main parts 1): Developing a New Method for Asymmetric Vesicles, and 2) Properties of aGUVs with Coexisting Liquid Phases. First, we describe experiments with uniform and aGUVs. These experiments and controls provide the means for a clear understanding of the protocol and analyses and also highlight the results that show this method to be reliable and robust. We then present and discuss results for asymmetric and phase-separated GUVs.

Developing a new method for asymmetric vesicles

Asymmetric labeled GUVs

Percentage of lipid exchanged

For a simple sample to develop this new method, we prepared aGUVs from the hemifusion of a symmetric GUV composed of DOPC labeled with DiD and a supported DOPC bilayer labeled with TFPC. To investigate the percentage of lipid exchanged in the hemifusion, we compared the fluorescence intensity measured for aGUVs to the intensity observed in symmetric GUV controls, as described in Materials and Methods. Fig. 3 A shows the intensities measured on a set of symmetric GUVs labeled with DiD or TFPC (filled bars). The procedure yields an ensemble of aGUVs with different intensities and therefore different fractions of exchanged lipids. By comparing the intensities of symmetric and aGUVs, we calculated the percentage of exchange. Fig. 3 A shows examples of aGUVs with different percentages of exchanged lipids. Here, 200% represents intensity from both leaflets, and 100% refers to just the outer leaflet. In these analyses, the percentage of exchange is independently calculated for DiD and TFPC, and the agreement between them is shown in Fig. 3 B. Important for this method, the results shown in Fig. 3 B enable us to distinguish and discard both the case of full fusion that leads to the absence of one dye and the case of multilayered vesicles (see Supporting Materials and Methods). Especially useful, we are able to distinguish and study aGUVs having a high percentage of asymmetry: we are not forced to study an ensemble of vesicles with a wide range of exchange percentage.

Outer leaflet fluorescence quenching

Next, we tested if the exchanged lipids that come from the SLB and replace the lipids from the outer leaflet of the symmetric GUV were indeed mainly found on the aGUV outer leaflet. We used the fluorescence quencher dithionite to quench the fluorescence of the aGUV outer leaflet and then measured the fraction of exchanged lipids found in the outer or in the inner leaflet. Dithionite is an efficient quencher of NBD-PE fluorescent probes, reducing the nitro group to an amine (20).

First, we studied the efficiency of dithionite to quench the fluorescent dyes used in this work. For these experiments, spectrofluorometric studies of a suspension of LUVs in a cuvette is more convenient than study of GUVs with a microscope. Fig. 4 A shows a fluorescence time course of LUVs composed of DOPC and labeled with NBD-PE or DiD or TFPC. We monitored the fluorescence signal in the absence and the presence of dithionite as a function of time. In this cuvette assay, we used unilamellar vesicles to investigate the temporal efficiency for dithionite to quench the entire outer leaflet.

We simultaneously measured the light scattering of these unilamellar vesicles (Fig. 4 B) to monitor vesicle integrity. The first 5 min of this time course shows the fluorescence and light scattering of LUVs in the absence of dithionite (Fig. 4, A and B, respectively). Dithionite promotes quick quenching of the LUV outer leaflet when labeled with NBD-PE or DiD. On the other hand, the addition of the same concentration of dithionite in the suspension of LUVs labeled with TFPC leads to only slight quenching of this dye even after 1 h. The quenching of NBD-PE and DiD on the outer leaflet was complete 5–7 min after dithionite addition, with the fluorescence intensity of these LUVs decreasing to half the initial value. After the quenching of the outer leaflet, much slower quenching is observed, suggesting slow penetration of dithionite across the membrane, as previously reported (20). After 45–50 min, the detergent Triton X-100 was added. Triton X-100 disrupts the vesicles, resulting in total quenching of NBD-PE and DiD dyes. This detergent test also indicates that the concentration of dithionite used is enough to quench not only the outer leaflet but also the entire bilayer. Using the same experimental conditions as with NBD-PE and DiD, dithionite does not efficiently quench TFPC, even after disrupting the vesicles. Fig. 4 A shows only a small percent loss of TFPC fluorescence after detergent addition. The addition of dithionite does not disrupt the vesicles by itself; vesicles are destroyed only after the addition of detergent, as shown in the light scattering results in Fig. 4 B.

We also used dithionite to quench the fluorescence of the outer leaflet of symmetric GUVs, finding similar results as with cuvette assays. For GUVs labeled with NBD-PE and DiD, quenching of the entire outer leaflet occurs. For GUVs labeled with TFPC, fluorescence decreased by ∼7–15% compared to the initial value (see Fig. S10).

Because of the great efficiency of dithionite to quench NBD-PE, but not TFPC, we labeled the SLB with NBD-PE. Here, aGUVs were prepared via hemifusion between symmetric GUVs composed of DOPC labeled with DiD and supported DOPC bilayers with NBD-PE. Fig. 5 A shows a range of the percentage of lipids exchanged, which was ∼50–99% using [Ca²⁺] ≈ 6 ± 2 mM. We used dithionite to quench the aGUV outer leaflet and measured the intensities of DiD and NBD-PE after quenching. Because dithionite also quenches DiD, we observed that the DiD fluorescence signal on the aGUV decreases to half of the signal observed on the symmetric GUVs, meaning that the remaining DiD on the aGUV outer leaflet was quenched by dithionite (Fig. 5 B, left plot). After the addition of dithionite to quench the aGUV outer leaflet, we image and analyzed 21 aGUVs, as displayed in the box plot of Fig. 5 B (left panel). For 16 aGUVs of this set of 21, we do not detect NBD-PE fluorescence signal above the background level, indicating that for these aGUVs, all NBD-labeled phospholipids were quenched by dithionite. For this set of 21 aGUVs, we observed only five aGUVs that display some trace of the NBD-PE fluorescence; the average NBD-PE signal on these aGUVs was 4% of the signal measured on the symmetric GUVs or ∼8% with respect to one leaflet. Because dithionite quenches the NBD-PE probes on the aGUV outer leaflet, the remaining detected signal from NBD-PE is presumably from the inner leaflet. This result suggests that the great majority of guest lipids have exchanged with the lipids that were previously in the outer leaflet; but a small percent of lipids could relocate to the inner leaflet, perhaps during hemifusion or detaching of the aGUVs from the SLB.

Fraction of hemifused aGUVs

We counted the fractions of nonleaky aGUVs after hemifusion. We also counted the fraction of GUVs that did not interact with the SLB and the fraction of GUVs that fully fused with the SLB. We trapped the fluorescent water-soluble dye carboxyfluorescein (CF) inside symmetric GUVs (Fig. 6 A). These GUVs had no other dye in the bilayer. We prepared an SLB composed of DOPC labeled with DiD. Unlike the other experiments, here, DiD is the guest dye. We found that the aGUV that hemifused with the SLB displays DiD fluorescence and retains the CF inside, as shown in Fig. 6 B. For GUVs that did not interact with the SLB, CF remains inside the GUV and there is no evidence of DiD in these vesicles, as shown in Fig. 6 A. In addition, for GUVs that fully fused with the SLB, we observed a GUV labeled with DiD but with no dye inside or a very weak fluorescent signal of CF inside (Fig. 6 C). In these experiments, we did not measure the percentage of lipid exchanged and considered as an aGUV vesicles that display any percentage of DiD fluorescence and retain CF similarly to symmetric GUVs.

Figure 6.

Fraction of aGUVs created and aGUV content release. (A) A symmetric GUV prepared with trapped carboxyfluorescein (CF) dye without membrane dye is shown. (B) An aGUV prepared by hemifusion between symmetric GUV as displayed in (A) and SLB labeled with red DiD is shown. (C) Fully fused GUV that has released the trapped dye is shown. (D) The following are shown (from left to right): the fraction of symmetric GUVs that did not interact with the SLB, aGUVs with preserved contents, aGUVs with slight leakage of the trapped dye, and fully fused GUVs obtained after the hemifusion procedure (n = 114). To see this figure in color, go online.

We also observed a fraction of aGUVs labeled with DiD but with an intensity of CF inside the GUV slightly lower, suggesting a little leakage (Fig. 6 D). We found an average intensity of CF inside the symmetric GUV and the aGUV without leakage equal to I_CF ≈ 2500 ± 300 arbitrary unit (arb. u.). For the aGUVs that we counted as slightly leaking CF, the average intensity was I_CF ≈ 2000 ± 200 (arb. u.). Fig. 6 D shows the fraction of symmetric GUVs, nonleaky aGUVs, slightly leaky aGUVs, and fully fused GUVs or aGUVs with total leakage. We counted 114 vesicles in these experiments.

Properties of aGUVs with coexisting liquid phases

Percentage of exchanged lipids and partition coefficient

Next, we studied aGUVs formed from hemifusion between phased-separated GUVs of DSPC/DOPC/chol = 0.39/0.39/0.22 + DiD and an SLB composed of DOPC/chol = 0.8/0.2 + TFPC. We chose to have the same chol fraction for each leaflet of the aGUV to minimize chol flip-flop that could cause changes in lipid properties. For phase-separated GUVs we also monitored the distribution of dyes into disordered and ordered domains. Fig. S6 shows micrograph examples of symmetric and aGUVs and the intensity analysis.

Tracking the percentage of lipid exchange enables estimating the lipid composition of the outer leaflet. For example, 30% of exchange leads to 0.7 × (DSPC/DOPC/chol = 0.39/0.39/0.22) + 0.3 × (DOPC/chol = 0.8/0.2), revealing the new outer leaflet to be DSPC/DOPC/chol ≈ 0.27/0.51/0.22. Fig. 7 A shows the lipid composition of the symmetric GUVs (exchange (%) = 0), which also represents the lipid composition of the inner leaflet as described by the star in the phase diagram. As the outer leaflet is replaced by DOPC/chol = 0.8/0.2, the composition of the outer leaflet changes follows the compositional trajectory described by the arrow in Fig. 7 A. The percentage of lipid exchange for aGUV is plotted in Fig. 7 B. The described trajectory in Fig. 7 A does not follow a thermodynamic tieline. On the contrary, the lipid compositions of the coexisting phases change on the pointed trajectory (Fig.7 A, arrow) where the tielines become shorter as the percentage of lipid exchange increases. In a ternary phase diagram of symmetric vesicles, tielines become shorter as they approach the critical point (31). For shorter tielines, the lipid composition of the Ld and the Lo phase are more alike until they match each other at the critical point. The partition coefficient of a fluorescent dye naturally decreases as the lipid compositions of the Ld and Lo phases become more similar. Because any total composition along a tieline can be described as a linear combination of the tieline end points, the Lo phase described by “tieline N” is enriched with DOPC compared to the Lo phase described by “tieline A” (Fig. 7 A). Thus, even for symmetric GUVs, the changes in the lipid composition for shorter tielines or for the trajectory represented by the arrow in Fig. 7 A would lead to decreased values of the dye partition coefficient.

Our results show that for symmetric GUVs, the partition coefficients of TFPC and DiD are Kp^TFPC = 13 ± 1 (n = 25) and Kp^DiD = 12 ± 1 (n = 18) (Fig. 7, C and D). As shown by the quenching experiments using dithionite (Fig. 5 B), the great majority of the exchanged lipids and dyes are located in the GUV outer leaflet. Therefore, TFPC only monitors the outer leaflet. For percentages of lipid exchange >72%, as indicated by the arrow on Fig. 7 A, the lipid composition on the outer leaflet is expected to form a single Ld phase (Fig. 7 A, phase diagram). However, TFPC reports a nonuniform dye distribution that is aligned with the inner leaflet, suggesting that the inner leaflet is inducing domains on the outer leaflet. At least in terms of lipid composition, these domains are different from the genuine phase separation Ld + Lo of symmetric membranes, so we refer to this Kp as an apparent Kp^app. We cannot distinguish the changes in the phase fractions on the outer leaflet because the inner leaflet phase separation induces domains in the outer leaflet.

Moreover, to investigate the Kp of DiD in the inner leaflet, we also needed dithionite to quench any DiD that remained in the outer leaflet, important for lipid exchange <100% (Fig. 7 C). Fig. S11 shows the difference between DiD Kp before and after outer leaflet quenching by dithionite. Fig. 7, B and C show the Kp of DiD and apparent Kp of TFPC for symmetric and aGUVs as a function of the percentage of exchanged lipids.

The dye partition coefficient measured for aGUVs decreases as the percentage of lipids exchanged with the SLB increases. For TFPC in the outer leaflet, apparent Kp^TFPC = 3.6 ± 1.4 for 98 ± 8% of exchanged lipids. For lipid exchange of 73–98%, apparent Kp^TFPC = 5 ± 1. For DiD exclusively in the inner leaflet, after fluorescence quenching of the outer leaflet, Kp^DiD = 6 ± 2 for 97 ± 9% of lipids exchanged on the outer leaflet. In the range 72–97% exchange, Kp^DiD = 7 ± 2. These results suggest that the DOPC/chol guest lipids from the SLB that replaced the GUV outer leaflet are changing the lipid composition and consequently changing the physical and chemical properties of the membrane, as reflected in lower Kp values, as discussed below.

GP of C-Laurdan in GUVs and aGUVs

To investigate changes in the lipid packing/order of aGUVs compared to the symmetric GUVs, we used the fluorescent probe C-Laurdan, measuring its GP (Eq. 4; Fig. 8). C-Laurdan exhibits a large excited-state dipole moment similar to Laurdan (32, 33, 34), which enables investigating the relative levels of lipid packing by observing the shift of emission spectra when the dye is located in different polarity environments. C-Laurdan exhibits a blue-shifted emission spectrum in the Lo phase and a red-shifted spectrum in the Ld phase (35). Here, we measured and compared the spectral image of C-Laurdan in symmetric and aGUVs. C-Laurdan is reported to cross the membrane (36), and this dye is likely located in both leaflets of the GUVs and aGUVs.

The aGUVs were prepared from hemifusion between symmetric GUVs composed of DSPC/DOPC/chol = 0.39/0.39/0.22 without fluorescence dyes and an SLB composed of DOPC/chol = 0.8/0.2 + DiD + C-Laurdan. In these experiments, the percentage of lipid exchange was calculated based on the intensity of a single dye, DiD on the GUV outer leaflet. Other blue or green dyes overlap with the C-Laurdan spectrum and cannot be used in these experiments. Here, we focus our attention on aGUVs with a high percent of lipid exchange >70%. Fig. 8, A and B show examples of the spectral imaging of symmetric and aGUVs, respectively. Fig. 8 C shows a color map that illustrates the GP values calculated for the symmetric GUV and the aGUV.

Fig. 8 D shows the spectral comparison of the blue-shifted ordered regions of the GUV and the aGUV, whereas Fig. 8 E shows the spectral comparison of the red-shifted disordered regions of the GUV and the aGUV. These spectra depict the points shown in Fig. 8 C. The dots represent the intensity measured as a function of wavelength, and we fit the data using a Gaussian fitting. We obtained the coefficient of determination R² > 0.97 for these fittings.

In Fig. 8 F, we plot the frequency of the GP values measured on GUVs and aGUVs, as previously reported for symmetric vesicles (37). From these distributions of GP values, we calculated an average GP for Ld and Lo phases for symmetric GUVs, and we observed how the GP values change for aGUVs (Table 1). Briefly, the <GP> corresponds to the mean value of the distribution, and the error corresponds to half of the distribution width.

For the symmetric GUVs of DSPC/DOPC/chol = 0.39/0.39/0.22 shown in Fig. 8, <GP_Ld> = −0.15 ± 0.08 and <GP_Lo> = 0.32 ± 0.1. For aGUVs, the inner leaflet has a composition similar to that of the initially symmetric GUVs, DSPC/DOPC/chol = 0.39/0.39/0.22. The lipid composition of the aGUV outer leaflet can be estimated using the percentage of exchanged lipids obtained from the dye intensities. The replacement of 85% of outer leaflet lipids leads to DSPC/DOPC/chol = 0.06/0.74/0.2 (see Fig. 7 A). For the aGUV in Fig.8, <GP_d> = −0.22 ± 0.08 and <GP_o> = 0.13 ± 0.08 for the disordered and the ordered regions, respectively. In this case, the differences between disordered and ordered regions for symmetric and aGUVs are ΔGP_GUVs = 0.47 and ΔGP_aGUVs = 0.35 (Table 1). The new outer leaflet of DOPC/chol ≈ 0.8/0.2 changes the lipid packing/order of the asymmetric bilayer in which the inner Ld phase + the newly exchanged outer leaflet opposed to this Ld phase display a slight decrease of the lipid packing/order, a difference of 0.07 in GP values. The combination of inner Lo phase + the newly exchanged outer leaflet of DOPC/chol opposed to this Lo phase shows a considerable decrease of the lipid packing/order, a difference of 0.19 in GP values. Moreover, the difference with respect to the lipid packing between the disordered and the ordered regions of the asymmetric bilayer ΔGP_aGUVs decreases compared to the difference between symmetric Ld and Lo phases ΔGP_GUVs. Thus, the decrease of the dye Kp is explained by disordered and ordered regions becoming more alike in lipid packing/order, as discussed below.

Other examples of GP values calculated for symmetric and aGUVs are shown in Table 1. Each row on Table 1 represents a different GUV or aGUV. For symmetric GUVs with a single Ld phase composed of DOPC or DOPC/chol = 0.8/0.2, the addition of a small fraction of chol slightly increases the lipid packing/order of the bilayer. We found GP = −0.23 and −0.18 for DOPC and DOPC/chol = 0.8/0.2, respectively. For symmetrically phase-separated GUVs, <GP_Ld>^GUVs is slightly more ordered because of its small fraction of DSPC, on average <GP_Ld>^GUVs ≈ −0.13, whereas the Lo phase has <GP_Lo>^GUVs ≈ 0.34. For aGUVs, the GP values depend on the percentage of lipid exchange, a larger percentage of asymmetry corresponding to reduced lipid packing in the bilayers, and a large effect is observed on the Lo phase, as described above.

Discussion

We developed a new method to make asymmetric bilayers by hemifusion. Lipid exchange by hemifusion is driven toward equilibrium by lipid diffusion between SLB and GUV, which is advantageous in comparison with equilibrium driven by an additional system component of molecules that mediate lipid exchange through an aqueous environment. Moreover, the individual study of aGUV having known percentage of exchanged lipids rather than a distribution of aGUVs having a range of exchange percentages, is a more interpretable way to characterize changes induced by a newly exchanged outer leaflet. In the development of this method, we first tested outer leaflet replacement and the stability of aGUVs. Then, we observed the phase behavior of an aGUV with phase-separated inner leaflet and an outer leaflet having a fluid composition.

To confirm that the guest dye and lipids replace the lipids on the GUV’s outer leaflet, we used dithionite to quench the aGUV outer leaflet (Fig. 5). NBD-PE from the SLB and, by inference, guest lipids were found on the aGUVs after hemifusion (lipid exchange ∼50–99%) (Fig. 5 A) Because NBD-PE and DiD in the outer leaflet are efficiently quenched by dithionite (Fig. 4 A), we could confirm that the guest dye and lipids are almost exclusively found in the aGUV outer leaflet. However, ∼25% of the aGUVs analyzed in these experiments exhibit very weak but detectible fluorescence of NBD-PE, ∼4% of the unquenched fluorescence of NBD-PE (Fig. 5 B). The outer leaflet is completely quenched by dithionite because the normalized intensity of DiD shows a signal that represents only the inner leaflet half of the bilayer (Fig. 5 B). We conclude that this weak signal of NBD-PE is from the inner leaflet: a small fraction of the guest lipids from the SLB, for only a small fraction of aGUVs, somehow migrated to the inner leaflet. Perhaps during hemifusion or more likely when aGUVs are sheared off the SLB, transient pores formed, and perhaps guest lipids and dyes flipped into the inner leaflet. This result is also consistent with a small fraction of aGUVs having a slight leakiness (Fig. 6 B), as discussed below.

We obtained a large fraction of aGUVs with preserved aqueous contents but also a small fraction of aGUVs with slight leakage, ∼15% of the vesicles analyzed. This fraction is similar to the fraction of aGUVs having small contamination of guest dye/lipids in the inner leaflet, detected by the quenching assay experiments on aGUVs (Fig. 5 B). Phospholipid flip-flop occurs only after days for intact model membranes (6, 38). In our experiments, phospholipid flip-flop is minimized because the aGUV preparation and data collection are performed in less than 5 h. Perhaps a transient pore facilitated the flipping of guest dye/lipids from the outer to the inner leaflet and caused slight leakage of trapped dye inside these aGUVs. Thus, aGUVs prepared from hemifusion have a low probability of guest lipid flipping into the inner leaflet, as suggested by the results in Figs. 5 and 6. For the data discussed in this article, the small fraction of guest lipids, ∼4%, found in the inner leaflet does not affect our results. Moreover, for some cases, fully fused GUVs exhibit complete absence of the dye originally on the GUV because when both leaflets fuse with the SLB, fluorescent dyes that originally labeled the GUV quickly diffuse into the SLB. Therefore, fully fused GUVs are easy to recognize without the need to measure trapped dye release. GUVs that did not interact with the SLB are also recognized because they lack the dye originally in the SLB. As a comparison, the experimental procedure that involves lipid exchange mediated by CD is a more invasive method in which lipids are constantly pulled out from the vesicles by free CD (5). This procedure also leads to minor contaminations of donor lipids in the inner leaflet of acceptor vesicles, as previously reported (6, 39).

We investigated the behavior of aGUVs with coexisting liquid phases. Symmetric vesicles prepared with the lipid composition DSPC/DOPC/chol = 0.39/0.39/0.22 exhibit coexistence of Ld and Lo phases, whereas symmetric vesicles composed of DOPC/chol = 0.8/0.2 do not phase separate. By replacing the outer leaflet of DSPC/DOPC/chol vesicles by DOPC/chol, we observed that the inner phase-separated leaflet induces regions of distinct lipid packing/order in the outer leaflet. We observed these induced domains to be in registration with Ld and Lo phase domains of the inner leaflet, with the less packed outer leaflet domain opposed to the inner leaflet Ld phase and the more tightly packed outer leaflet domains opposed to the inner leaflet Lo phase. Detecting domains induced in the outer leaflet requires a high percentage of exchanged lipids so that residual genuine phase separation does not confound the measurements.

Because we measure the percentage of exchanged lipids, we can easily estimate the lipid composition on the outer leaflet, as described above. The exchange of 30% of lipids on the outer leaflet would create DSPC/DOPC/chol ≈ 0.27/0.51/0.22. In symmetric vesicles, this lipid composition would phase separate into Ld and Lo domains (Fig. 7 A, phase diagram). Therefore, when a low percentage of exchange occurs, the outer leaflet will naturally form genuine Ld and Lo phases, although with an increased fraction of DOPC compared to the inner leaflet, DSPC/DOPC/chol = 0.39/0.39/0.22. In part, residual phase separation explains the similarity of the dye partition coefficients between symmetric and aGUVs that have a low percentage of exchanged lipids because both leaflets naturally form phase-separated domains, although with slightly different lipid compositions.

In the simple case in which 100% of its lipids are exchanged, the outer leaflet becomes the Ld phase of the SLB. However, for less than 100% exchange, an Ld phase can incorporate up to a fraction of ∼0.11 of DSPC and a fraction of 0.22 of chol without any formation of an Lo phase ((24, 25, 40); see Fig. 7 A) We can also estimate what percentage of exchanged lipids is required so that the outer leaflet forms a single Ld phase even when that phase contains residual fractions of DSPC and chol. For 72% lipid exchange, the estimated lipid composition of the outer leaflet is DSPC/DOPC/chol = 0.11/0.69/0.21, which in symmetric vesicles, forms a single Ld phase (24, 25, 40) (Fig. 7 A, phase diagram). Note that the phase diagram of symmetric and asymmetric vesicles might differ. However, more importantly for the examples mentioned above (30 and 72% of lipid exchange), the existence of domains in the outer leaflet has a different nature: for the lipid mixtures studied in this work and for percentages of exchanged lipids greater than 72%, the domains in the outer leaflet are induced by the inner leaflet phase separation. As shown by the Kp measurements, these domains have different properties compared to the symmetric ones and to those in which there is still registration of genuine phases (i.e., low percentage of exchanged lipids).

The partition coefficient of TFPC in the outer leaflet (Fig. 7 B) reflects the changes in the lipid packing and compositional changes in the outer leaflet. As the phase-separated outer leaflet is replaced by DOPC/chol = 0.8/0.2, the fraction of DOPC increases, whereas the fraction of DSPC decreases as more lipids are exchanged, and the fraction of chol is almost constant (Fig. 7 A). The increased fraction of DOPC must decrease the lipid packing of this leaflet, resulting in a lower dye partition coefficient because the composition of the outer leaflet tends toward miscibility and a uniform Ld phase (Fig. 7 A, arrow). For 98% of exchanged lipids, the lipid composition of the outer leaflet is expect to form a uniform Ld phase for which in a uniform phase, the dye Kp is defined to be 1. For this percentage of exchange, we measured apparent Kp^{TFPC (aGUV)} = 3.6 ± 1.4, suggesting that these domains are different from domains of symmetric phase separation, and they are induced by the inner leaflet. Again, it should be noted that the changes in the lipid composition do not follow a tieline, and these changes lead to the replacement of DSPC by DOPC on both Ld and Lo phases. As hemifusion proceeds, a disordered lipid is enriched in the outer leaflet; this entire leaflet becomes less ordered, which is also confirmed by C-Laurdan experiments.

The difference in the dye partition coefficients between symmetric and aGUVs is smaller for the DiD that labels the inner leaflet (Fig. 7 C). Kp^{DiD (GUV)} = 12 ± 1 and Kp^{DiD (aGUV)} = 6 ± 2 with ∼97% lipid exchange. The decrease of DiD Kp as a function of percentage of exchanged lipids (Fig. 7 B) shows that the new outer leaflet is either inducing slight changes in the inner leaflet, affecting the lipid packing/order, or else inducing a redistribution of high- and low-melting lipids and chol between the Ld and the Lo phases. The latter might be associated with changes of the phase boundaries of the two-phase region (Ld + Lo), possibly indicating shorter tielines of the two-phase region in asymmetric phase diagrams in comparison with symmetric phase diagrams.

In our investigation of lipid packing/order, comparison between symmetric and aGUVs shows differences in the GP values calculated from the spectral image of C-Laurdan (Fig. 8). C-Laurdan can cross the bilayer (36). Therefore, we can only compare an average lipid packing of the inner and outer leaflets for symmetric and asymmetric bilayers. As the partition coefficient studies show, for a high percentage of exchanged lipids, the newly exchanged outer leaflet not only has different lipid/domain properties but also influences the properties of the inner leaflet, and the GP analyses report the combined effects of both layers.

Comparing the GP values between symmetric and aGUVs, the Ld outer leaflet DOPC/chol of aGUVs is responsible for decreasing the lipid packing of the asymmetric bilayer compared to the symmetric GUV that is phase separated in both leaflets. The ordered region of aGUVs, <GP_o>^(aGUV) = 0.13 ± 0.08, shows significant difference with respect to lipid packing/order compared to symmetric GUVs, <GP_Lo>^(GUV) = 0.32 ± 0.1 (Fig. 8 D). The disordered regions of symmetric <GP_Ld>^(GUV) = −0.15 ± 0.08 and aGUVs <GP_d> = −0.22 ± 0.08 show a smaller difference (Fig. 8 E). For DOPC and DOPC/chol = 0.8/0.2, <GP>^DOPC = −0.23 ± 0.1 and <GP>^DOPC/chol = −0.18 ± 0.1. The average GP obtained for disordered regions of aGUVs is consistent with the values of fluid phases like pure DOPC. We do not know whether any compositional separation of DOPC from chol has occurred, but the value <GP_d> = −0.22 ± 0.08 would suggest disordered domains depleted in chol.

TFPC is uniformly distributed in the SLB, but not in aGUVs. Apparently, the inner leaflet induces a disordered domain across from the inner leaflet Ld phase and an ordered domain across from the inner leaflet Lo phase. Indeed, TFPC in the outer leaflet is enriched ∼4-fold in the domain opposed to the Ld phase and depleted from the domain opposed to the Lo phase, suggesting an increase of packing/order for ordered regions on the outer leaflet (Fig. 7 B). The behavior of DiD Kp as a function of percentage of exchanged lipids suggests that the outer leaflet changes the lipid packing/order of the inner leaflet, decreasing the difference in packing/order between the inner leaflet Ld and Lo phases (i.e., making Ld and Lo phase more alike). Because the packing/order of the disordered domains does not increase, the inner Lo lipid packing/order must become considerably less packed/ordered. Ordered, tightly packed domains were induced in the outer leaflet where the fluorescence of TFPC clearly shows dark domains depleted of fluorescent labels. In contrast, the inner Lo phase across from these order-induced domains must become less packed to explain the decrease of the DiD Kp. Interestingly, not only the order of the inner Lo phase is transmitted across the bilayer, inducing an ordered domain on the outer leaflet, but also this Lo phase is affected by the more disordered outer leaflet enriched in DOPC. The overall lipid packing of the bilayer that includes the inner Lo phase plus the induced-ordered domains is less packed/ordered than symmetric Lo domains. Although the experiments with C-Laurdan do not distinguish packing/order of the individual leaflets, the results are consistent with the dye Kp experiments in which the difference in packing/order between disordered and ordered domains decreases in aGUVs, leading to the observed changes in the dye partition coefficients.

In agreement with other research groups (25, 27), we found coupling between bilayer leaflets. Induced-disordered and induced-ordered domains opposed to the genuine inner Ld and Lo phases suggest significant coupling between leaflets. However, the meaning of leaflet coupling is not well defined, and there is not general agreement about how to measure leaflet coupling. In a sense, different researchers describe leaflet coupling for a particular system without an agreed definition of what “coupling” means. Beyond phase-separated domains in the two leaflets being in register, coupling can mean that one phase influences the melting temperature (4, 41) or the diffusion coefficient of the apposed leaflet (42), the order parameter in the apposed leaflet is changed (41, 42), or that the extent of interdigitation has changed (42). Nonetheless, theoretical/phenomenological models have been proposed to estimate coupling strengths (23, 24, 25). Experimental studies using Förster resonance energy transfer to monitor the thermal stability of Ld and Lo domains found that the longer the acyl chain of the high-melting lipid, the larger the influence on the coupling dominance (27). Such results suggest that interleaflet coupling can be indirectly mapped by monitoring the thermal behavior of asymmetric vesicles and imply that there is a “dominant leaflet” controlling the domain behavior (27). We also find that the leaflet that contains the high-melting lipid DSPC is the dominant leaflet. This phase-separated leaflet induces domains on the other leaflet that would otherwise be expected to be uniform. chol is also suggested to play an important role on the coupling between leaflets (43), either partially entering the bilayer midplane or by flip-flop-induced redistribution between leaflets (44, 45). Coupling between the two leaflets may have relevance to signal transduction in cells.

Conclusions

We propose a new method to prepare aGUVs using hemifusion. Our method yields aGUVs that approach 100% asymmetry without residual organic solvent or CD. Moreover, our method enables the study of highly asymmetric individual GUVs rather than a broad distribution of asymmetries. In the development of our method, we show that the exchanged lipids are mainly found on the outer leaflet of aGUVs using a quenching assay. In addition, our procedure mostly yields hemifused GUVs with preserved vesicle content. In our investigations of the behavior of aGUVs having coexisting liquid phases, we evaluate aGUV physical and chemical properties for measured levels of asymmetry. The inner leaflet of these aGUVS has a lipid composition that phase separates into Ld and Lo domains, whereas the outer leaflet mixture of DOPC/chol = 0.8/0.2, does not phase separate in symmetric vesicles. The nature of domains in aGUVs could be divided in two categories: registration of phases through the bilayer and the induction of disordered and ordered domains. These categories can be differentiated by the percentage of exchanged lipids and the membrane properties. aGUVs with enriched disordered lipid in the outer leaflet exhibit different lipid packing/order compared to the symmetric phase-separated vesicles.

Author Contributions

T.A.E. performed the experiments, analyzed the data, and wrote the manuscript. G.W.F. and T.A.E. designed the research.

Editor: Georg Pabst.