Accurate quantification of inter-domain partition coefficients in GUVs exhibiting lipid phase coexistence

Abstract

Giant unilamellar vesicles (GUVs) with phase coexistence allow for the recovery of inter-domain partition coefficients (K_p) of fluorescent molecules through comparison of fluorescence intensities in each phase. This method has been extensively used to gather qualitative information regarding the preference of both lipid analogues and other fluorescent molecules for insertion into ordered lipid membrane phases, which is often used as a predictor for incorporation in ordered plasma membrane domains. Methods aiming to recover quantitative information on partition properties from GUV imaging fail to correct for brightness and area per lipid differences, skewing recovered values. Additionally, photoselection effects occurring in the presence of linearly polarized excitation are generally neglected in these calculations. Here, we describe a new methodology for correction of fluorescence imaging data obtained from GUVs to recover accurate partition coefficients, which accounts for changes in the probe's quantum yield, area per lipid differences and photoselection effects. This general methodology is used to quantify liquid ordered/liquid disordered lipid phase partition coefficients of several commonly used fluorescent analogues of phospholipids. Importantly, several sources of error in spectroscopic measurements of K_p, such as incorporation of a fluorescent probe in domain boundaries, clustering in water or in the membrane, or formation of three coexisting phases, are either irrelevant for quantification of inter-domain K_p's through the method presented here or are readily recognized through imaging with GUVs.

---

Introduction

The equilibrium thermodynamic constant defining enrichment of a molecule in a given phase at the expense of its concentration in another lipid phase is the partition coefficient (K_p). While K_p's are normally used to describe the distribution of a molecule between two homogeneous phases, this concept can also be employed to characterize the membrane affinity of a molecule through a membrane/water K_p¹ or membrane inter-domain partition, such as between a liquid-ordered (l_o) and a liquid-disordered phase (l_d).² The partition coefficient between l_o and l_d lipid phases gained significant relevance upon the discovery of cholesterol and sphingomyelin-enriched liquid ordered domains in the plasma membrane.^3,4 These regions are thought to allow for the compartmentalization of different lipids and proteins, contributing to spatially defined activation of specific cellular functions, such as intracellular lipid traffic and cell signalling.⁵

The inter-domain K_p of a given fluorescent molecule can be recovered through different approaches: (i) independent measurements of the membrane/water K_p for each lipid phase⁶ and; (ii) measurements of exchange between two populations of lipid vesicles,⁷ each in one of the membrane phases; (iii) measurements between two phases that coexist in the same lipid bilayer.⁸ While all of these methods have been successfully applied to the quantification of inter-domain K_p's, isolated measurements of the membrane/water K_p for each lipid phase requires considerably more effort than determining a single K_p, and is challenging for molecules with very high membrane/water K_p. On the other hand, measurements of exchange between two populations of lipid vesicles, each in one of the membrane phases is significantly prone to artefacts, as lipid mixing can take place, and the method is also impractical for molecules with a high membrane/water K_p.⁸

For molecules with high membrane/water K_p, measurements between two phases that coexist in the same lipid bilayer are advantageous. In these measurements, the two lipid phases are expected to be under equilibrium. This is critical, since the partition coefficient is an equilibrium thermodynamic quantity.⁸ This method has been extensively applied to the quantification of the partition of molecules between gel and fluid as well as between l_o and l_d lipid phases. Its application has focused on large unilamellar or multilamellar vesicles through fluorescence spectroscopy,^9–11 and supported membranes¹² or giant unilamellar vesicles (GUV's) through fluorescence microscopy.^13,14

The determination of partition coefficients through fluorescence microscopy is particularly attractive, since it does not rely on non-linear fits of the relationship between fluorescence parameters and fraction of each phase to recover partition coefficients, and is carried out in an individual system (supported membrane or GUV). This eliminates the possibility of erroneous interpretation of data obtained from an ensemble measurement with heterogeneous populations of vesicles. Additionally, in lipid membranes with phase coexistence, molecules in some cases show preferential incorporation into domain interfaces,¹³ where lipid packing defects are much higher. Although thermodynamically the domain interfaces are not a formal phase, they are for practical reasons a third “pseudo”-phase, which competes for the incorporation of the fluorescent molecule. In standard fluorescence spectroscopy measurements, partition to this pseudo-phase would remain undetected, while this phenomenon is readily identified through imaging methods.

Supported membranes have the disadvantage of substrate interaction (which can significantly modify the properties of the interacting bilayer). In fact, the distance of a membrane adsorbed onto a hydrophilic support to its surface ranges between 0.5 and 2 nm,¹⁵ which leads to nonnative and non-specific interactions between membrane components with the support.^15,16 Although these interactions are relatively weak, they have been shown to have a significant impact on diffusion properties of membrane components^17,18 On the other hand, GUVs present very limited interaction with the substrate. Among other advantages relative to smaller vesicles and supported membranes, the membrane curvature of GUVs is more closely related to the membrane curvature of the plasma membrane of eukaryotic cells. Importantly, given their dimensions (>1 μm), they offer the possibility of direct observation through phase contrast and fluorescence microscopy of optically resolvable details in membrane morphology and organization.¹⁹ Fluorescence microscopy studies with GUVs have been particularly useful to the characterization of phase coexistence within membranes, namely regarding lipid phase diagrams,^20–22 membrane domain properties,^23–28 morphology,^29,30 and selective lipid phase enrichment in specific molecules.^23,31–34

For these reasons, observation of the partition of fluorescent molecules through fluorescence imaging of GUVs has gained considerable popularity in the recent past. The main limitation of this method is related to its qualitative nature,³⁵ as attempts to recover quantitative K_p's ignored either the relative differences in quantum yields, average molecular areas in each phase or failed to account for the photoselection effect when linearly polarized light was used for excitation. Here, we show how to accurately weight and account for all these factors, in order to retrieve accurate inter-domain partition coefficients of fluorescent molecules through fluorescence imaging of GUVs exhibiting phase coexistence.

Experimental

Materials

1-Palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC), N-palmitoyl-D-erythro-sphingosylphosphorylcholine (PSM), 1,2-dioleoyl-sn-glycero-3-phosphoethanolamine-N-(cap biotinyl) (DOPE-Cap-biotin), 1,2-dioleoyl-sn-glycero-3-phosphoethanolamine-N-(lissamine rhodamine B sulfonyl) (Rho-DOPE), 1,2-dipalmitoyl-sn-glycero-3-phosphoethanolamine-N-(7-nitro-2-1,3-benzoxa-diazol-4-yl) (NBD-DPPE) and 1,2-dioleoyl-sn-glycero-3-phosphoethanolamine-N-(7-nitro-2-1,3-benzoxadiazol-4-yl) (NBD-DOPE) were obtained from Avanti Polar Lipids (Alabaster, AL). Avidin from egg white and cholesterol (Chol) were from Sigma Chemical Co. (St. Louis, MO). All organic solvents were UVASOL grade from Merck Millipore (Darmstadt, Germany). 2-(4,4-Difluoro-5-methyl-4-bora-3a,4a-diaza-s-indacene-3-dodecanoyl)-1-hexadecanoyl-sn-glycero-3-phosphocholine (Bodipy-PC) was obtained from Thermo Fisher Scientific (Waltham, MA, USA).

Liposome preparation

Lipid stock solutions were prepared in chloroform. Quantification of lipid and lipid analogue stock concentration was carried out as described elsewhere.¹³ For preparation of multilamellar vesicles (MLVs), lipid mixtures composed of the adequate amount of lipids were prepared in chloroform to a final concentration of 0.25 mM. The solvent was slowly evaporated under a nitrogen flux and the resulting lipid films were left in vacuum overnight to ensure the complete removal of the organic solvent. Subsequently, the lipids were re-suspended in a 10 mM HEPES buffer (pH 7.4), containing 150 mM NaCl. Freeze–thaw cycles (liquid nitrogen/water bath at 60 °C) were then performed to re-equilibrate and homogenize the samples.

GUVs were obtained by electroformation in Pt wires as previously described.¹³ Briefly, lipid mixtures were prepared in chloroform (from lipid stock solutions) in a probe/lipid ratio of 1/500 for Rho-DOPE and Bodipy-PC, and 1/200 for NBD-DOPE and NBD-DPPE. DOPE-Cap-biotin was included at a biotinylated lipid/total lipid ratio of 1/10⁶ for vesicle immobilization. In this concentration of biotinylated lipids, minimal perturbation of the membrane due to surface adhesion occurs. In addition, since measurements are always carried out in the equatorial slice of the GUV, no artefacts associated to surface adhesion are expected.¹³ After removal of the solvent, electroformation was performed at 58 °C during 75 min, in 1 mL of a 200 mM sucrose solution preheated at the same temperature. After formation, GUVs were transferred to a μ-Slide from Ibidi (Munich, Germany) coated with avidin, and a 200 mM glucose solution was also added to the wells in order to increase GUV deposition and immobilization rates.

Steady-state fluorescence spectroscopy

Fluorescence measurements were carried out with a SLM-Aminco 8100 Series 2 spectrofluorimeter (Rochester, NY) with double excitation and emission monochromators (MC-400), in right-angle geometry. The light source was a 450 W Xe arc lamp and the reference a rhodamine B quantum counter solution. Quartz cuvettes (0.5 × 0.5 cm) from Hellma Analytics were used. Temperature was controlled to 25 °C. When necessary, fluorescence intensities were corrected for inner filter effects as described elsewhere.³⁶ The fluorescence intensity of liposomes without fluorescent lipid analogues was always subtracted to the measured fluorescence signals.

Confocal microscopy

Fluorescence imaging was performed on a Leica TCS SP5 (Leica Microsystems CMS GmbH, Mannheim, Germany) inverted confocal microscope (DMI6000). A 63× amplification apochromatic water immersion objective with NA = 1.2 (Zeiss, Jena, Germany) was used for all measurements, and an argon laser was used for excitation. GUVs were immobilized in avidin coated Ibidi (Munich, Germany) slides through DOPE-Cap-biotin, which was included at a biotinylated lipid/total lipid ratio of 1/10⁶. After GUV immobilization, images were acquired at a line frequency of 100 Hz. The excitation and emission wavelengths (λ_ex/λ_em) were as follows: 514/530–650 nm for Rho-DOPE, 458/480–530 nm for NBD-DOPE and NBD-DPPE, and 488/500–550 nm for Bodipy-PC. Each GUV was measured at the equatorial plane. All data analysis, as described here, was carried out in a Matlab environment (Mathworks, Natick, MA).

Results and discussion

We set-out to obtain the of different fluorescent lipid analogues. We chose the commonly used marker of l_d phases, Rho-DOPE, and both saturated and unsaturated phospholipids labeled with the NBD fluorophore (NBD-DPPE and NBD-DOPE) (Fig. 1). All these lipids are labeled at the headgroup, and as a result of the different dimensions of the fluorophores and of acyl-chain unsaturation, they are expected to have significantly different affinities for the l_o and l_d lipid phases. In fact, rhodamine labeled lipids are reported to always exhibit preference for liquid disordered phases, independently of the lipid acyl-chains.^11,29 On the other hand, the partition of NBD analogues shows considerably greater dependency on acyl-chain structure. NBD-labeled lipids with unsaturated chains are expected to show preference for incorporation into the l_d phase,³² while several studies with NBD-labeled lipids with saturated chains show contradictory results, with preference for either l_d or l_o.^{11,23,32,37–40} For comparison, the l_o/l_d inter-domain partition of a fluorescent lipid analogue where the fluorophore is attached to the acyl-chain (Bodipy-PC) was also studied (Fig. 1). Experiments were carried out at relative concentrations of phospholipid fluorescent analogue between 0.2 and 0.5% to prevent changes in lipid phase properties.

Fig. 1 Structure of the phospholipid fluorescent analogues.

The ternary lipid mixture POPC/PSM/cholesterol is the simplest system to model lipid raft-like composition, with cholesterol, and high and low melting temperature lipids (PSM and POPC, respectively). The phase diagram for this lipid mixture has been characterized at 23 °C, and some tie-lines that give the composition and fraction of each phase within the l_o/l_d phase coexistence region have been obtained.⁴¹ In this work, vesicles were prepared for a POPC : PSM : Chol composition of 45.1 : 29.9 : 25.0 (mol : mol : mol) within this tie-line, so that lipid domains have a known composition defined by the extremes of the tie-line (71.6 : 23.3 : 5.0 for 100% l_d and 25.4 : 34.8 : 39.8 for 100% l_o).⁴¹ However, the procedure carried out here (with the exception of corrections for average lipid area) is also valid in general for lipid mixtures with no available phase diagram or tie-line information. In any case, the composition of GUVs formed from electroformation is somewhat heterogeneous, and the domain composition for individual vesicles is expected to vary around the values dictated by the tie-line. This heterogeneity has been shown to be partially due to the demixing of lipids in the dry film state before electroformation.⁴²

Confocal microscopy measurements were carried and equatorial sections of the GUVs loaded with each of the phospholipid analogues are shown on Fig. 2. Almost no fluorescence is observed for Rho-DOPE within the liquid ordered phase as expected, given the already reported high affinity of this lipid for more disordered membranes and low quantum yield when in liquid ordered membranes. Phase assignment for GUVs loaded with NBD and Bodipy-labeled lipids was carried out through analysis of photoselection in each domain, as photoselection is always considerably higher in the l_o phase as shown below.

Fig. 2 Confocal fluorescence microscopy images of the equatorial section of GUVs prepared from the POPC : PSM : Chol mixture 45.1 : 29.9 : 25.0 (mol : mol : mol) and loaded with each of the studied fluorescent lipid analogues at a molar ratio of 0.2 mol% for Rho-DOPE and Bodipy-PC and 0.5 mol% for NBD-DPPE and NBD-DOPE. Bars: 5 μm.

The average ratios of fluorescence intensities obtained from confocal fluorescence microscopy measurements of phospholipid analogues between each lipid phase (I^l_o/I^l_d) are shown on Table 1. These are directly calculated from the absolute fluorescence values in each GUV. In order to recover quantitative information relative to the l_o/l_d inter-domain partition of these fluorescent lipid analogues from GUV imaging data, it is necessary to identify all factors influencing fluorescence intensities in each lipid phase apart from lipid analogue concentration.

Table 1 Partition coefficients obtained by confocal imaging of GUVs for each fluorescent phospholipid analogue, and ratio of lipid phase-dependent parameters required for calculation of

	〈I^lo/I^ld〉^a	〈I^lo/I^ld〉^P^b	Φ^ld/Φ^lo^c	σ^lo/σ^ld^f	^g
a Mean ratio of fluorescence intensity values in each lipid phase (lo – liquid ordered, ld – liquid disordered) before correction for photoselection of 5–10 individual GUVs.b Mean ratio of fluorescence intensity values after correction for photoselection.c Spectroscopically defined ratio of fluorescence quantum yields of phospholipid analogues in liquid ordered and liquid disordered phases.d Obtained from ref. 34.e Obtained from the integration of steady-state fluorescence emission spectra in lipid vesicles composed of POPC : PSM : Chol 71.6 : 23.3 : 5.0 (mol/mol/mol) (ld) and 25.4 : 34.8 : 39.8 (mol/mol/mol) (lo).f Ratio of the surface density of phospholipid analogues in each phase.g Partition coefficients were calculated according to eqn (5), using a ratio of average lipid molecular areas of Ālo/Āld = 0.74, as discussed in the text.
Rho-DOPE	0.09 ± 0.05	0.057 ± 0.005	3.3^d	0.19 ± 0.02	0.14 ± 0.01
NBD-DPPE	3.7 ± 1.9	1.7 ± 0.3	0.9^d	1.5 ± 0.3	1.1 ± 0.2
NBD-DOPE	0.40 ± 0.09	0.40 ± 0.06	1.0^e	0.40 ± 0.06	0.29 ± 0.04
Bodipy-PC	0.5 ± 0.2	0.36 ± 0.03	1.5^e	0.53 ± 0.04	0.40 ± 0.04

---

The fluorescence intensity value measured in a position r at the lipid phase i in the surface of a GUV through confocal or two-photon excitation microscopy can be defined as:

I_F(r) = LDBⁱk_rⁱC_FⁱS_D (1)

in which L is the excitation intensity, D is the detection efficiency and Bⁱ is the brightness of the fluorophore in the corresponding phase i, defined as the product of the extinction coefficient εⁱ and quantum yield Φⁱ. k_rⁱ is a factor quantifying the extent of photoselection observed in position r and for a particular phase, and C_Fⁱ is the bi-dimensional density of the fluorophore in the membrane surface for phase i. Bi-dimensional concentrations are considered here since the height of the bilayer is at least 50 times smaller than the diffraction-limited resolution of confocal microscopy, and for simplicity, the z-dimension normal to the membrane can be ignored in the calculation of focal volume concentrations. S_D is the membrane surface within the detection volume which can vary depending on the position within the GUV (higher in the equatorial region and lower in the poles). Eqn (1) assumes that no change in fluorescence emission spectra shape of the probe takes place between the two lipid phases. This is true for both Rho-DOPE and Bodipy-PC, while NBD-labeled phospholipids exhibited only a minor 2 nm blue-shift upon incorporation in liquid ordered membranes (results not shown), compared to the spectra in liquid disordered phases.

Within a GUV, S_D is constant in the equatorial plane, and the detection efficiency and the excitation intensity are constant for the different positions within the vesicle. In this way, for a GUV with two different lipid phases within the equatorial plane, the ratio of fluorescence intensities in each lipid phase i in two different positions r_A and r_B (on the equatorial plane) becomes:

where N_Fⁱ is the number of fluorophores in phase i and A_i is the membrane surface area occupied by molecules in phase i. From the definition of mole partition coefficient (K^M_p),¹ the number of molecules of fluorophore (N_Fⁱ) in each phase i is related to the ratio of molecules of each phase (N_i):

However, from microscopy of GUVs only information from fluorescence intensity per area is obtained. So, to retrieve partition coefficients from the ratio of fluorescence values obtained from microscopy, we must first convert N_i to areas, using average lipid molecular areas (Ā_i) characteristic of each lipid phase:

Combining eqn (2) and (4), we have:

Alternatively, if no information is available concerning average lipid molecular areas, the ratio of the density of fluorophore in each phase (σ_F¹/σ_F²) can be used:

In this way, to quantify inter-domain partition of fluorescent molecules from confocal microscopy data obtained from GUVs with phase coexistence, we must first correct the experimental data (〈I^l_o/I^l_d〉) for the photoselection effect, and for brightness and molecular areas differences in each lipid phase. Strategies to carry out all these corrections are discussed below.

Photoselection effect

Since molecules within membranes are neither randomly oriented nor rotating freely, the absorption transition moment [small mu, Greek, vector]_abs of fluorescent membrane probes exhibits a preferential orientation. In fact, it is expectable that specific fluorophore orientations are fully inaccessible due to both structural and energetic barriers, while others are highly probable. Within GUVs, and since these are generally a near perfect sphere, the average orientation of [small mu, Greek, vector]_abs changes depending on the fluorophore's position within the vesicle (Fig. 3). Lasers used for confocal microscopy are often linearly polarized, and molecules with their absorption transition moments aligned parallel to the electric vector E⃑ of the polarized excitation are more likely to become excited (photoselected).

Fig. 3 Spherical coordinates are shown for a given GUV. γ is the polar angle and φ is the azimuthal angle. X₁ is the optical axis of the microscope. X₂ and X₃ define the plane of the equatorial region of the GUV. Excitation light (E⃑) is linearly polarized in the X₂ direction. In case the transition moment ([small mu, Greek, vector]) of the fluorophore within the membrane is oriented parallel to the bilayer normal N⃑ (yellow) the probability of excitation in the equatorial region is proportional to cos² φ (top and bottom insets). On the other hand, if [small mu, Greek, vector] is perpendicular to N⃑ (orange), the probability of excitation in the equatorial region is proportional to cos²(φ + 90).

The probability of absorption of a single photon is proportional to cos² θ, where θ is the angle between [small mu, Greek, vector]_abs and E⃑.⁴³ In a spherical coordinate system within the GUV, the probability of fluorophore excitation will also vary widely for each position within the vesicle, defined by the polar angle γ and the azimuthal φ. In the equatorial plane, in case the polarization of excitation light is parallel to the axis X₂ in Fig. 3, the angle between the membrane normal and the excitation light polarization is φ. If the average [small mu, Greek, vector]_abs of a fluorophore is parallel to the membrane normal (N⃑), the probability of excitation within the equatorial section of the GUV is proportional to cos² φ (Fig. 3). On the other hand, if [small mu, Greek, vector]_abs is perpendicular to N⃑, then the function describing the probability of excitation is shifted by 90°. For two-photon excitation, since two photons are simultaneously absorbed, the probability of excitation becomes proportional to cos⁴ θ and the photoselection effect is considerably higher.

Photoselection also varies considerably for different lipid phases. In fact, more ordered lipid phases hinder fluorophore rotation to a greater extent. As a result, the orientation of the fluorophore population within each position of the GUV becomes more homogeneous, and the amplitude of differences in excitation probability is greater. This dependence of excitation probability on photoselection creates a problem for quantitative measurements of fluorescence intensity within GUVs when using linearly polarized excitation, as an angular dependence of the intensity response of the fluorophore occurs. Several solutions are available to cope with this problem. One possibility is the use of a quarter-waveplate to obtain circularly polarized light. In this case, as long as measurements are restricted to the equatorial region of the GUV, the polarization of excitation light is always parallel to [small mu, Greek, vector]_abs, independently of the position within the equatorial plane of the vesicle, and no photoselection is observed.⁴⁴

Researchers aiming to extract quantitative information on the partition of fluorescent molecules between coexisting phases in a GUV have also chosen to measure fluorescence intensities in the equatorial region at membrane phases on opposite sides of a GUV (φ and φ + 180°), since for an identical molecule at these two angles, photoselection is expected to be the same.³⁴ This approximation is not valid if the extent of photoselection varies significantly between the two lipid phases, which is commonly observed for liquid ordered and liquid disordered phases as we show here (Fig. 4). This option also limits the use of data to the regions of the membrane where two different domains are found at precisely opposed sides at the equatorial plane, dramatically reducing the statistical quality of the measurements.

Fig. 4 Angular (φ) dependence of fluorescence intensities (I_φ) for the different fluorescent phospholipid analogues in both l_d (a) and l_o (b) lipid phases. Lipid composition of GUVs was (a) 71.6 : 23.3 : 5.0 (mol/mol/mol) and (b) 25.4 : 34.8 : 39.8 (mol/mol/mol) POPC : PSM : Chol. Fluorescence confocal images are the equatorial confocal slices of GUVs loaded with each of the lipid analogues. Intensity profiles (c) are the angular (φ) dependences of fluorescence intensities for each GUV (blue – l_d; green – l_o). Thicker lines are the empirical fits of equation 7 to the data. Lipid compositions representative of each lipid phase were selected from a previously published phase diagram for this lipid mixture.⁴¹ Bars: 5 μm.

A more general solution to this problem is the calibration of photoselection through measurements of the studied fluorescent molecule within GUVs of pure lipid phase. In case phase diagram and tie-line information is available for the lipid mixture under study, the composition of these homogeneous GUVs can be set to mimic the composition of domains in GUVs with phase coexistence. Husen and coworkers⁴⁵ showed that a linear combination of a number of spherical harmonics can be fitted to fluorescence intensities in a stack of confocal images obtained at different planes of a GUV, and that these empirical functions could then be used to correct for the photoselection effect in GUVs exhibiting phase coexistence. However, when measuring GUVs only in the equatorial region, simple sinusoidal functions (eqn (7)) can be used to characterize the general angular (φ) dependence of fluorescence intensities (I_φ):

I_φ = A₁ cos²(φ + β) + A₂ (7)

where A₁ is the amplitude of the function, φ is the angle between excitation light polarization and membrane normal, β is the phase shift and A₂ is the intensity observed at maximum photoselection. Angular emission patterns of GUVs loaded with all the fluorescent lipid analogues for each lipid phase are shown on Fig. 4. Empirical fits of eqn (7) to the experimental data are shown as well.

Differences in the extent of photoselection for each fluorescent lipid analogue are clearly identifiable. Rho-DOPE shows the higher extent of photoselection in the l_d phase (Fig. 4). Rhodamine in Rho-labeled phosphatidylethanolamine (PE) lipids has been show to have its transition moment highly oriented 90° relative to the membrane normal,⁴⁶ while the orientation of the transition moment of NBD within the headgroup of NBD-labeled lipids has been shown to have a relatively broad distribution.⁴⁷

As expected, photoselection is more significant in the liquid ordered phase as seen for NBD-DOPE, NBD-DPPE and Bodipy-PC (Fig. 4). For Rho-DOPE, low levels of photoselection were identified in the l_o phase. These results are in agreement with the previous observation of Rho-DOPE segregation into highly probe-enriched membrane nanodomains or clusters within condensed l_o phases.⁴⁸ This segregation is a consequence of the extremely poor solubility of Rho-DOPE in the l_o phase as the large fluorophore in the headgroup and the unsaturated chains inhibit the normal packing of this lipid in highly ordered membranes. Results from Fig. 4 clearly show that the segregated Rho-DOPE within the l_o phase is less oriented than the molecule in the l_d phase.

For correction of I^l_o/I^l_d values for photoselection, the ratio of photoselection parameters k_rA²/k_rB¹ in eqn (5) can be described by:

where, r_A and r_B are the two different positions (φ) within the equatorial region of the GUV in lipid phase 2 and 1, respectively. I_ref(r_x)ⁱ is the fluorescence intensity at position φ = x and in phase i for the homogeneous GUVs used for calibration. Mean ratios of fluorescence intensity of phospholipid analogues in each lipid phase after correction for photoselection 〈I^l_o/I^l_d〉^P are shown on Table 1. Values of I^l_o/I^l_d before this correction present very large errors due to the different photoselection observed for domains in different phases and vesicle positions. An immediate consequence of this correction is a considerable reduction of the measurement error (up to 1 order of magnitude).

The method described here can also be applied to giant plasma membrane vesicles (GPMVs) or to GUVs in the absence of phase diagram information. In that case, since no information is available concerning domain lipid compositions, no calibration with reference GUVs can be performed. Nevertheless, in case the domains occupy at least a full quarter of the GUV/GPMV, empirical photoselection corrections based on the fluorescence of a single domain are still possible.

Brightness

The membrane environment where the fluorophores of these lipid analogues are located is considerably different for ordered and disordered membrane phases. In the liquid ordered phase, lipid dynamics is reduced and water penetration in the membrane decreases accordingly.⁴⁹ Both changes in membrane viscosity and hydration can modify radiative and non-radiative rates of relaxation, influencing the quantum yield of the fluorophore. For these reasons, the fluorescence intensity values obtained for each lipid phase are not equally proportional to fluorophore concentration, and for the calculation of a K_p from a GUV exhibiting phase coexistence, the relative difference in molecular brightness Bⁱ of lipid analogues in each phase i must be estimated. An efficient strategy to recover relative molecular brightness values is to carry out fluorescence correlation spectroscopy (FCS) measurements.³⁴ FCS measurements also allows for the recovery of relative brightness values in GPMVs or in GUVs prepared from lipid mixtures with no phase diagram information. However, FCS setups are not accessible in most laboratories, and other options are available to recover relative brightness values for the lipid analogue in each coexisting lipid phase. Since changes in extinction coefficients (εⁱ) are expected to be small or inexistent, we can consider as an approximation that the relative brightness value in a given lipid phase equals the relative quantum yield value for that phase. In that case, B^l_d/B^l_o = Φ^l_d/Φ^l_o, and relative quantum yields are easily obtained through integration of steady-state fluorescence emission spectra or fluorescence decays in vesicles prepared from lipid compositions representative of each lipid phase. In case tie-line information is available for the phase diagram of the lipid mixture, pure lipid phase vesicles can be prepared from the tie-line extremes. Steady-state measurements have the advantage of accounting for possible static quenching effects if the lipid analogue is found to cluster in one of the lipid phases (as Rho-DOPE in ordered membranes). Static quenching would result in an overestimation of relative quantum yield in both FCS and fluorescence decay measurements.

In addition, in case two different fluorophores are being tracked at the same time in a GUV, very low concentrations (<0.2 mol%) should be used to make sure that no significant Förster resonance energy transfer (FRET) takes place as this would decrease the quantum yield of the putative donor species.

Values of the ratio Φ^l_d/Φ^l_o for each of the fluorescent phospholipid analogues are also included in Table 1. For NBD-DOPE and Bodipy-PC, fluorescence emission spectra were measured in liposomes with compositions defined by the tie-line extremes, and integrated to recover ratios. For NBD-DPPE and Rho-DOPE, previously published values for the same compositions were used.³⁹ The much higher quantum yield of Rho-DOPE in liquid disordered phase is the result of self-quenching of the lipid analogue in the liquid ordered phase due to clustering. While differences in brightness of the NBD-labeled lipids are moderate (<12%), results for Rho-DOPE and Bodipy-PC show considerably higher brightness of these molecules in liquid disordered membranes. For these molecules, neglecting to correct for changes in quantum yield of molecules in the different lipid phases would lead to dramatic errors in the calculation of partition coefficients.

In case no thermodynamic information (phase diagrams and tie-lines) is available for the lipid mixture being studied, eqn (6) can be used to quantify ratios of density (σ^l_o/σ^l_d) of the fluorescent molecules in each lipid phase. Results for the phospholipid analogues studied here are shown on Table 1.

Molecular lipid areas

Molecular areas within different lipid phases are also known to be significantly different, even for l_o/l_d phase coexistence. In case there is phase diagram and tie-line information available for the lipid mixture under study, it is possible to recover more accurate partition coefficients by accounting for this factor. In the specific case of the lipid mixture used in this study, the average area per lipid is considerably different for POPC, Chol, and PSM, and focal volumes in each phase correspond to different numbers of molecules. Average area per lipid values can be estimated on the basis of structural information available from X-ray scattering, surface pressure-mean molecular area isotherms or molecular dynamics simulation. However, the presence of cholesterol is known to induce condensation of lipid phases⁵⁰ and this should also be taken into account. The average area per lipid of POPC and Chol were estimated to be 62 and 26 Ų, respectively.^51,52 The average area per PSM molecule was estimated to be 43 Ų on the basis of literature results for the pure lipid⁵³ and for the impact of cholesterol on the condensation of similar sphingomyelin lipids.⁵⁴ For the compositions on the extremes of the tie-line chosen in this study, the ratio of average area per lipid in each phase (Ā_lo/Ā_ld) was estimated to be 0.74.

Partition coefficients

Partition coefficient values obtained from GUVs presenting phase coexistence were calculated from eqn (5) using all the corrections discussed above, and are presented on Table 1. The recovered partition coefficient for Rho-DOPE is in general agreement with partition coefficient values obtained through fluorescence spectroscopy .³⁷ DOPE-labeled with NBD is also shown to have considerable affinity for l_d phases . This is in agreement with the qualitative preference identified in several studies for NBD-DOPE in l_o/l_d GUVs.^32,38,55,56

For NBD-DPPE, the partition coefficient approached unity . The saturation/insaturation of the acyl-chains of headgroup NBD-labeled phospholipids is, as expected, crucial in dictating preference for disordered or moderate preference for ordered membranes. The importance of carrying out the corrections described here is highlighted by the fact that for NBD-DPPE, although the fluorescence intensities of this lipid are higher in the l_o phase (Fig. 2), the molar concentration of this analogue is slightly lower in l_o than in l_d domains. The higher fluorescence intensities observed for NBD-DPPE in l_o domains are then the combined result of: (i) a higher quantum yield in the environment provided by ordered domains; and (ii) a lower average area per lipid in the l_o phase. This is likely to explain some of the disagreement in literature regarding the partition preference of this analogue as judged by fluorescence microscopy of GUVs.^{23,32,38,40,56} Additionally, the partition preferences of NBD-DPPE have been shown to vary for different giant plasma membrane vesicles (GPMVs) within the same sample,⁵⁶ possibly reflecting the strong dependence of partition coefficient of this molecule on the composition of liquid ordered and liquid disordered phases. This is also likely to explain the larger error associated to the measurement of the of NBD-DPPE relatively to values obtained for the other phospholipid fluorescent analogues (Table 1).

Despite the fact that the PC-Bodipy analogue studied here presents a fully saturated C16 acyl-chain at sn-1 position, the incorporation of the Bodipy fluorophore at the sn-2 acyl-chain is shown to promote significant affinity for the disordered lipid phase , also in agreement with previous observations of preference for insertion in l_d phases for this lipid.^32,38

While in spectroscopic methodologies precise thermodynamic information must be considered to recover partition coefficients (distribution of domain compositions over a tie-line between two coexisting phases),³⁵ the determination of inter-domain K_p through microscopy methods allows for the direct examination of the distribution of the fluorescent molecule of interest. Since the exact position of different phase boundaries and tie-lines for several lipid compositions is controversial and dependent on the methods used for its detection, the ability to measure partition coefficients in a quantitative manner without the pre-condition of knowing the exact position of a tie-line within lipid phase coexisting regions is highly valuable. Also, the visual information accessible through fluorescence microscopy offers additional advantages. The preferential partition of the labelled molecule to domain boundaries (such as observed for NBD-DPPE in gel/fluid lipid phase coexistance¹³), the aggregation of the fluorescent molecule in solution or into large membrane clusters, as well as the formation of a third phase coexistence system (such as a l_d, l_o and gel lipid phase coexistence), would all lead to an error in the estimation of K_p through fluorescence spectroscopy methods, while being irrelevant or readily recognized through imaging of GUVs or GPMVs.

Conclusions

Our data clearly demonstrate the potential of fluorescence imaging of GUVs for the determination of accurate inter-domain partition coefficients. The approach presented here is readily applicable to two-photon microscopy imaging of GUVs as well. With the exception of average lipid area corrections, all other corrections are possible in GPMVs or GUVs prepared from lipid mixtures in the absence of phase diagram and tie-line information. This is a key advantage of this method, since phase diagrams and tie-lines particularly are challenging to define with precision. Importantly, several sources of error in spectroscopic measurements, such as incorporation of fluorescent probe in domain boundaries, clustering in water or in the membrane, or formation of three coexisting phases, are either readily recognized or irrelevant for quantification of inter-domain partition coefficients through the method presented here.

Acknowledgements

MJS and SNP are recipients of fellowships from FCT (SFRH/BD/80575/2011 and SFRH/BPD/92409/2013 respectively). Authors acknowledge funding by FCT project references FAPESP/20107/2014, RECI/CTM-POL/0342/2012 and UID/NAN/50024/2013.

Notes and references

1. N. C. Santos, M. Prieto and M. A. R. B. Castanho, Biochim. Biophys. Acta, Biomembr., 2003, 1612, 123–135.
2. M. A. Gardam, J. J. Itovitch and J. R. Silvius, Biochemistry, 1989, 28, 884–893.
3. K. Simons and E. Ikonen, Nature, 1997, 387, 569–572.
4. A. Rietveld and K. Simons, Biochim. Biophys. Acta, 1998, 1376, 467–479.
5. K. Simons and J. L. Sampaio, Cold Spring Harbor Perspect. Biol., 2011, 3, a004697.
6. J. Mello-Vieira, T. Sousa, A. Coutinho, A. Fedorov, S. D. Lucas, R. Moreira, R. E. Castro, C. M. P. Rodrigues, M. Prieto and F. Fernandes, Biochim. Biophys. Acta, 2013, 1828, 2152–2163.
7. S.-L. Niu and B. J. Litman, Biophys. J., 2002, 83, 3408–3415.
8. W. L. C. Vaz and E. Melo, J. Fluoresc., 2001, 11, 255–271.
9. M. J. Sarmento, A. Coutinho, A. Fedorov, M. Prieto and F. Fernandes, Biochim. Biophys. Acta, 2014, 1838, 822–830.
10. L. Silva, R. F. M. De Almeida, A. Fedorov, A. P. A. Matos and M. Prieto, Mol. Membr. Biol., 2006, 23, 137–148.
11. L. M. Loura, A. Fedorov and M. Prieto, Biophys. J., 2001, 80, 776–788.
12. C. Dietrich, Z. N. Volovyk, M. Levi, N. L. Thompson and K. Jacobson, Proc. Natl. Acad. Sci. U. S. A., 2001, 98, 10642–10647.
13. M. J. Sarmento, M. Prieto and F. Fernandes, Biochim. Biophys. Acta, 2012, 1818, 2605–2615.
14. T. Baumgart, G. Hunt, E. R. Farkas, W. W. Webb and W. Gerald, Biochim. Biophys. Acta, Biomembr., 2007, 1768, 2182–2194.
15. D. J. Muller, Biophys. J., 2006, 91, 3133–3134.
16. M. Tanaka and E. Sackmann, Nature, 2005, 437, 656–663.
17. M. Przybylo, J. Sýkora, J. Humpolíčová, A. Benda, A. Zan and M. Hof, Langmuir, 2006, 22, 9096–9099.
18. R. Macháň and M. Hof, Biochim. Biophys. Acta, Biomembr., 2010, 1798, 1377–1391.
19. L. A. Bagatolli, Biochim. Biophys. Acta, 2006, 1758, 1541–1556.
20. G. W. Feigenson and J. T. Buboltz, Biophys. J., 2001, 80, 2775–2788.
21. S. L. Veatch and S. L. Keller, Biophys. J., 2003, 85, 3074–3083.
22. M. Fidorra, A. Garcia, J. H. Ipsen, S. Härtel and L. A. Bagatolli, Biochim. Biophys. Acta, Biomembr., 2009, 1788, 2142–2149.
23. C. Dietrich, L. a. Bagatolli, Z. N. Volovyk, N. L. Thompson, M. Levi, K. Jacobson and E. Gratton, Biophys. J., 2001, 80, 1417–1428.
24. N. Kahya, D. Scherfeld, K. Bacia, B. Poolman and P. Schwille, J. Biol. Chem., 2003, 278, 28109–28115.
25. S. N. Pinto, F. Fernandes, A. Fedorov, A. H. Futerman, L. C. Silva and M. Prieto, Biochim. Biophys. Acta, 2013, 1828, 2099–2110.
26. H.-J. Kaiser, D. Lingwood, I. Levental, J. L. Sampaio, L. Kalvodova, L. Rajendran and K. Simons, Proc. Natl. Acad. Sci. U. S. A., 2009, 106, 16645–16650.
27. S. L. Veatch and S. L. Keller, Phys. Rev. Lett., 2002, 89, 268101.
28. J. Korlach, P. Schwille, W. W. Webb and G. W. Feigenson, Proc. Natl. Acad. Sci. U. S. A., 1999, 96, 8461–8466.
29. T. Baumgart, S. T. Hess and W. W. Webb, Nature, 2003, 425, 821–824.
30. L. Bagatolli, Chem. Phys. Lipids, 2003, 122, 137–145.
31. A. T. Hammond, F. A. Heberle, T. Baumgart, D. Holowka, B. Baird and G. W. Feigenson, Proc. Natl. Acad. Sci. U. S. A., 2005, 102, 6320–6325.
32. J. Juhasz, J. H. Davis and F. J. Sharom, Biochem. J., 2010, 430, 415–423.
33. K. Bacia, D. Scherfeld, N. Kahya and P. Schwille, Biophys. J., 2004, 87, 1034–1043.
34. E. Sezgin, I. Levental, M. Grzybek, G. Schwarzmann, V. Mueller, A. Honigmann, V. N. Belov, C. Eggeling, Ü. Coskun, K. Simons and P. Schwille, Biochim. Biophys. Acta, Biomembr., 2012, 1818, 1777–1784.
35. J. R. Silvius, Biochim. Biophys. Acta, Mol. Cell Res., 2005, 1746, 193–202.
36. J. R. Lakowicz, Principles of Fluorescence Spectroscopy, Springer, New York, 2006.
37. L. C. Silva, R. F. M. de Almeida, B. M. Castro, A. Fedorov and M. Prieto, Biophys. J., 2007, 92, 502–516.
38. T. Baumgart, G. Hunt, E. R. Farkas, W. W. Webb and G. W. Feigenson, Biochim. Biophys. Acta, Biomembr., 2007, 1768, 2182–2194.
39. R. F. M. de Almeida, L. M. S. Loura, A. Fedorov, M. Prieto, M. S. Loura and R. F. M. De Almeida, J. Mol. Biol., 2005, 346, 1109–1120.
40. Q. Lin and E. London, Biophys. J., 2015, 108, 2212–2222.
41. R. F. M. de Almeida, A. Fedorov and M. Prieto, Biophys. J., 2003, 85, 2406–2416.
42. E. Baykal-Caglar, E. Hassan-Zadeh, B. Saremi and J. Huang, Biochim. Biophys. Acta, 2012, 1818, 2598–2604.
43. J. Lakowicz, Principles of Fluorescence Spectroscopy, Kluwer Academics, New York, 1999.
44. L. A. Bagatolli and E. Gratton, Biophys. J., 2000, 79, 434–447.
45. P. Husen, M. Fidorra, S. Härtel, L. A. Bagatolli and J. H. Ipsen, Eur. Biophys. J., 2012, 41, 161–175.
46. L. A. Bagatolli and E. Gratton, Biophys. J., 2000, 78, 290–305.
47. H. Filipe, L. Santos, J. Prates-Ramalho, M. J. Moreno and L. Loura, Phys. Chem. Chem. Phys., 2015, 17, 20066–20079.
48. B. M. Castro, R. F. M. De Almeida, A. Fedorov and M. Prieto, Chem. Phys. Lipids, 2012, 165, 311–319.
49. G. M'Baye, Y. Mély, G. Duportail and A. S. Klymchenko, Biophys. J., 2008, 95, 1217–1225.
50. H. M. McConnell and A. Radhakrishnan, Biochim. Biophys. Acta, Biomembr., 2003, 1610, 159–173.
51. S. A. Pandit, S.-W. Chiu, E. Jakobsson, A. Grama and H. L. Scott, Biophys. J., 2007, 92, 920–927.
52. O. Edholm and J. F. Nagle, Biophys. J., 2005, 89, 1827–1832.
53. X. M. Li, J. M. Smaby, M. M. Momsen, H. L. Brockman and R. E. Brown, Biophys. J., 2000, 78, 1921–1931.
54. L. Grönberg, Z. S. Ruan, R. Bittman and J. P. Slotte, Biochemistry, 1991, 30, 10746–10754.
55. J. M. Crane and L. K. Tamm, Biophys. J., 2004, 86, 2965–2979.
56. P. Sengupta, A. Hammond, D. Holowka and B. Baird, Biochim. Biophys. Acta, 2008, 1778, 20–32.

---

Footnote

† Current address of M. J. Sarmento: Istituto Italiano di Tecnologia, Genova, Italy.

---