Differential Effect of Plant Lipids on Membrane Organization

Background: The effect of the wide variety of plant lipids on membrane organization is still poorly understood. Results: The local amounts of phytosphingolipids and free and conjugated phytosterols control the relative proportion and spatial distribution of ordered domains. Conclusion: Plant lipids differently modulate, alone or in combination, membrane heterogeneity. Significance: Our results highlight how plant lipid diversity might drive the plasma membrane subcompartmentalization. The high diversity of the plant lipid mixture raises the question of their respective involvement in the definition of membrane organization. This is particularly the case for plant plasma membrane, which is enriched in specific lipids, such as free and conjugated forms of phytosterols and typical phytosphingolipids, such as glycosylinositolphosphoceramides. This question was here addressed extensively by characterizing the order level of membrane from vesicles prepared using various plant lipid mixtures and labeled with an environment-sensitive probe. Fluorescence spectroscopy experiments showed that among major phytosterols, campesterol exhibits a stronger ability than β-sitosterol and stigmasterol to order model membranes. Multispectral confocal microscopy, allowing spatial analysis of membrane organization, demonstrated accordingly the strong ability of campesterol to promote ordered domain formation and to organize their spatial distribution at the membrane surface. Conjugated sterol forms, alone and in synergy with free sterols, exhibit a striking ability to order membrane. Plant sphingolipids, particularly glycosylinositolphosphoceramides, enhanced the sterol-induced ordering effect, emphasizing the formation and increasing the size of sterol-dependent ordered domains. Altogether, our results support a differential involvement of free and conjugated phytosterols in the formation of ordered domains and suggest that the diversity of plant lipids, allowing various local combinations of lipid species, could be a major contributor to membrane organization in particular through the formation of sphingolipid-sterol interacting domains.

The high diversity of the plant lipid mixture raises the question of their respective involvement in the definition of membrane organization. This is particularly the case for plant plasma membrane, which is enriched in specific lipids, such as free and conjugated forms of phytosterols and typical phytosphingolipids, such as glycosylinositolphosphoceramides. This question was here addressed extensively by characterizing the order level of membrane from vesicles prepared using various plant lipid mixtures and labeled with an environment-sensitive probe.
Fluorescence spectroscopy experiments showed that among major phytosterols, campesterol exhibits a stronger ability than ␤-sitosterol and stigmasterol to order model membranes. Multispectral confocal microscopy, allowing spatial analysis of membrane organization, demonstrated accordingly the strong ability of campesterol to promote ordered domain formation and to organize their spatial distribution at the membrane surface. Conjugated sterol forms, alone and in synergy with free sterols, exhibit a striking ability to order membrane. Plant sphingolipids, particularly glycosylinositolphosphoceramides, enhanced the sterol-induced ordering effect, emphasizing the formation and increasing the size of sterol-dependent ordered domains. Altogether, our results support a differential involvement of free and conjugated phytosterols in the formation of ordered domains and suggest that the diversity of plant lipids, allowing various local combinations of lipid species, could be a major contributor to membrane organization in particular through the formation of sphingolipid-sterol interacting domains.
Lateral partitioning of lipid domains was first reported in model systems, using biophysical approaches (1)(2)(3)(4), indicating that the different lipid molecular species are not homoge-neously distributed within lipid bilayers. Fluorescence microscopy studies using giant unilamellar vesicles (GUVs) 2 allowed observation of large domains of different composition, depending on the lipid mixture (5), suggesting that lipid heterogeneity may play a role in the spatial organization of biological membranes (6,7). More recently, the lipid raft model proposed that the phase behavior of different lipid species could account for the lateral heterogeneity observed at the plasma membrane (PM) surface (8). Accordingly, optically resolvable imaging in lipid bilayer membranes involving at least three components reported the coexistence of a liquid-disordered (L d ) phase, formed mainly from unsaturated phospholipids (9 -14), with a liquid-ordered (L o ) phase formed from saturated phospholipids and sphingolipids in the presence of cholesterol (8,15,16). In the latter, a higher degree of conformational order is imposed on the acyl tails of lipids by the rigid ring structure of cholesterol (17), increasing the thickness of the lipid bilayer and lipid packing, although lipids remain laterally mobile. Sterols (18) and sphingolipid/sterol interactions (19 -24) have recently emerged as essential determinants of lipid partitioning within membranes.
Multiple roles for lipids in determining membrane order were thus emphasized and encompassed their tendency to phase separation and organization of biological membranes (25). However, all of these studies were performed on model and animal systems, where the lipid diversity is considerably lower than in plants. Plants contain a huge diversity of lipid molecular species, and the lipid mixture present in their PM is particularly complex. Lipidomic approaches revealed a PM composition of 30 -50% phospholipids, 20 -50% sterols, and 6 -30% sphingolipids, depending on plant species and organs (for a review, see Ref. 26). The structural diversity found in plant lipids mainly arises from the biosynthesis of various combinations into each of these three lipid groups, which could reach to thousands of lipid species.
For instance, higher plant cells contain a vast array of sterols (61 sterols and pentacyclic triterpenes have been identified in maize seedlings) in which ␤-sitosterol, stigmasterol, and campesterol often predominate (27). Phytosterol molecular forms contain typically 28 or 29 carbons and, compared with cholesterol, exhibit an extra alkyl group, either an ethyl or methyl group, on the side chain in the C24 position and an extra carbon-carbon double bond in the sterol nucleus (28). In plant PM, steryl glycosides (SGs) and acyl steryl glycosides (ASGs) are found in addition to free sterols (29 -33). Variations in sterol-conjugated forms originate from differences in the type of phytosterol; the sugar, usually the pyranose form of D-glucose; the configuration of the linkage; the number of sugar groups; and the sugar acylation. Acylation may occur at the C6 of the sugar moiety with fatty acids, palmitic acid (16:0) being most common (34,35). SGs and ASGs, which do not exist in animal membrane, are more effective than free sterols at promoting hexagonal phase formation (36). They constitute up to ϳ30% of the detergent-resistant membrane fraction isolated from plant PM (26), based on their resistance to extraction by Triton X-100 at 4°C (37).
Otherwise, sphingolipidomic approaches identified over 200 sphingolipids in Arabidopsis leaves (38). Glycosylinositolphosphoceramides (GIPCs), which may represent up to 60% of the plant PM, and glycosylceramide (GluCer) are the predominant sphingolipid classes (for a review, see Ref. 39). Although GIPCs belong to one of the earliest classes of plant sphingolipids to be identified (40 -44), few GIPCs have been fully characterized to date because of their high polarity and relatively poor recovery using traditional extraction techniques (38,45,46). GIPCs are formed by the addition of an inositol phosphate to the ceramide followed by several glycosylation of the inositol headgroup (47)(48)(49). In the few plant species characterized to date, it seems that there is a dominant glycan structure with an hexose-glucuronic acid linked to the inositol, although the composition of the other sugars seems to be species-specific (50), polar heads containing up to seven sugars (for a review, see Ref. 51). GIPCs from Arabidopsis and tobacco, analyzed by mass spectrometry, were composed primarily of ceramide with t18:0 and t18:1 and 2-hydroxylated very long chain fatty acids (50,51).
The role of such a wide diversity of plant lipids and the respective influence of the different classes of lipids on membrane organization are still poorly understood. To evaluate the specific involvement of each lipid molecular species in membrane organization, we added sequentially an equimolar amount of different lipids into phospholipid vesicles. To characterize the membrane lateral organization of such vesicles, we used hydroxy-3-(N,N-dimethyl-N-hydroxyethyl)ammoniopropyl-4-(␤-(2-(di-n-butylamino)-6-napthyl)vinyl) pyridinium dibromide (di-4-ANEPPDHQ), an environment-sensitive fluorescent probe that was recently introduced as an alternative to 6-dodecanoyl-2-dimethylaminonaphthalene (laurdan) to assess the level of order of both model and biological membranes (52). Di-4-ANEPPDHQ is inserted into one leaflet of the lipid bilayer with its chromophore aligned to the surrounding tails of the lipid molecules and its headgroup oriented toward lipid polar heads. Its emission spectrum is thus influenced by the membrane packing around the dye molecules in the two lipid phases (53,54) mainly by sensing the reorientation of solvent dipoles, related to the level of water penetration into the lipid bilayer (54). We thus described the individual ability of major phytosterols to modulate the global level of order of a phospholipid bilayer, in synergy with other sterol-conjugated forms and sphingolipids. We further analyzed, using multispectral confocal microscopy on large vesicles, the spatial distribution of the order level of membrane areas at a pixel (i.e. 300 ϫ 300 nm) resolution. This revealed very different abilities of the various phytosterols, alone or in combination, to modulate the proportion of L o phase and the membrane heterogeneity and a particular ability of GIPCs in synergy with conjugated sterols to organize the membrane and promote the formation of large ordered domains. We finally discuss how the diversity of plant lipids might drive the PM subcompartmentalization, allowing the dynamic segregation of membrane components.
Extraction and Purification of Total GIPCs from BY-2 Cells-GIPCs are not commercially available; they were purified according to a method adapted from Carter and Koob (41) to obtain final amounts in the milligram range, described by Buré et al. (55). Briefly, tobacco BY-2 cells (ϳ100 -200 g, fresh weight) were blended with 400 ml of cold 0.1 N aqueous acetic acid in a chilled Waring Blendor at maximum speed for 2 min. The slurry was filtered under vacuum through 16 layers of acidwashed Miracloth on a large funnel. The residue was extracted twice again in the same manner. The aqueous acetic acid filtrate was discarded, and the residue was then re-extracted with hot (70°C) 70% ethanol containing 0.1 N HCl. The slurry was fil-tered hot through Miracloth and washed with acidic 70% ethanol. The residue was re-extracted twice more with acidic 70% ethanol. The combined filtrates were chilled immediately and left at Ϫ20°C overnight. The precipitate was pelleted by centrifugation at 2,000 ϫ g at 4°C for 15 min. The GIPC-containing pellet was washed with cold acetone until washes were colorless and finally with cold diethyl-ether to yield a whitish precipitate. GIPCs were then dissolved in tetrahydrofuran/methanol/water (4:4:1, v/v/v) containing 0.1% formic acid by heating at 60°C, followed by gentle sonication. Glycerolipids were hydrolyzed by methylamine treatment, and GIPCs were re-extracted with hot 70% ethanol. GIPC extracts were further dried and submitted to a butan-1-ol/water (1:1, v/v) phase partition. This process left proteins and cell walls in the lower aqueous phase, whereas more than 97.5% of the GIPCs were recovered in the upper butanolic phase. Finally, the latter phase was dried, and the residue was dissolved in tetrahydrofuran/methanol/water (4:4:1, v/v/v) containing 0.1% formic acid. The amount of GIPCs was determined by quantitative GC-MS analysis measuring the 2-hydroxylated very long chain fatty acids, hallmarks of tobacco GIPCs.
Preparation of GUVs-GUVs were prepared by electroformation (56 -58) to obtain unilamellar vesicles with sizes varying from 15 up to 50 m. A flow chamber (ICP-25 perfusion imaging chamber, Dagan), connected to a function generator (PCGU1000, Velleman) and a temperature controller (TC-10, Dagan) was used for vesicle preparation. Briefly, various lipid solutions (4 l of 1 mg/ml stock solution in chloroform/methanol (9:1, v/v)) were deposited on two microscope slides (18 ϫ 18 mm) coated with electrically conductive indium tin oxide (resistivity 8 -12 ohms). Lipid-coated slides were placed under vacuum for at least 2 h until a thin lipid film was obtained. Cover slides were set up on the flow chamber, and the lipid layer was rehydrated by 200 l of 500 mM sucrose solution preheated to 55°C. A voltage of 3.5 V at 10 Hz and a temperature of 55°C were applied for a minimum of 2 h. After lipid swelling, the chamber was cooled down slowly to 22°C (1 h minimum). The same procedure was used for all lipid mixtures.
Preparation of Large Unilamellar Vesicles (LUVs)-After being prepared as described above, GUVs were extruded and calibrated to 1-m diameter on a hot plate (55°C, corresponding to a temperature above the phase transition temperature of the lipid) by using a miniextruder from Avanti Polar Lipids, Inc. After 15 passes of the GUV solution through membrane, we obtained a homogenous LUV population (59, 60) with a diameter of 1 m. LUVs were kept under a nitrogen atmosphere to minimize oxidation.
Fluorescence Spectroscopy-Probes (stock solution in DMSO) were diluted to 0.1 mol % in 500 mM sucrose solution. Fluorescence measurements were performed on a Fluorolog-3 FL3-211 spectrometer (Jobin-Yvon, Horiba Group) in the T-format with one double-monochromator for excitation and two single-monochromators for emission light. Detection is ensured by two photomultipliers. The light source is a xenon arc lamp. This spectrophotometer is equipped with movable excitation and emission polarizers. All fluorescence signals were recorded with emission and excitation bandwidths of 5 nm and systematically corrected from light scattering of an unlabeled sample. All data acquisitions were done with the Datamax software (Jobin-Yvon/Thermo Galactic Inc.). Experiments were done using a 10-mm special optic glass path cuvette filled with 1 ml of vesicle solution (4 g/ml in 500 mM sucrose solution). All measurements were performed at 22°C in a temperature-controlled chamber using a thermoelectric Peltier junction (Wavelength Electronics Inc.).
Confocal Fluorescence Microscopy-The vesicle suspension was labeled with fluorescent probes (di-4-ANEPPDHQ, DiI-C 18 , or Bdp-cholesterol; stock solution in DMSO; 3 M final) 5 min before microscopy observation and further placed between the slide and coverslip. All experiments were performed at room temperature (22°C). A Sapphire LP (low power) continuous wave blue laser (Coherent) was used as the excitation source (488 nm). A Nikon C1Si/picoquant inverted microscope (Nikon) with spectral camera Nuance CRI was used to observe cells. Images were acquired with a plan apochromat ϫ100.0/ 1.40/0.13 oil objective. A dichroic slide (405-488 nm), which slightly modifies the emission spectrum in red wavelengths, was used to increase the fluorescent signal recovery. High spectral resolution was achieved through the use of a fine ruled diffraction grating, allowing a 5 nm channel resolution. The dispersed light was captured simultaneously at different wavelengths using 32 independent channels of a multianode array of photodetectors. Vesicles were observed by focusing on the maximum area of the vesicle membrane, and acquisition was performed for square surface (51.2-m side lengths) at a 512 pixel resolution, according to the recommended sampling theorem (Nyquist) and resolution limit of the microscope.
Image Analysis-Individual recorded images were transferred to ImageJ software (National Institutes of Health). Red (or green) intensities were then computed by averaging four intensities from 545 to 565 nm (or from 635 to 655 nm). Ratiometric images were designed by dividing the created red images by created green images. To perform spatial analysis, pixels of 288 ϫ 288 nm were analyzed, and central regions of the GUV top surface were cropped in order to avoid edge effects.

Major Phytosterols Differentially Modulate the Global Order
Level of Model Membranes-The ability of each major phytosterol, individually or in combination, to modulate membrane order was investigated using the environment-sensitive probe di-4-ANEPPDHQ. Inserted in the whole bilayer, the dye locates with a slight preference into the L d phase (54,61), where its emission spectrum shows a red shift (62,63). A function corresponding to the ratio of the fluorescence emission signal recorded at 550 and 660 nm, referred to hereafter as RGM (for red/green ratio of the membrane), gives a convenient and quantitative way to measure this emission shift (53, 54, 64 -70). The RGM value allows characterization of the relative proportion of L o and L d phases in a lipid bilayer and is inversely correlated to the relative amount of L o phase.
LUVs, homogeneously sized (1-m diameter) to avoid any differential effect of the curvature on the fluorescence parameter recorded, were prepared using DOPC (33%), DPPC (33%), and sterol (33%): either cholesterol, stigmasterol, ␤-sitosterol, or campesterol. Lipid vesicles were labeled with di-4-ANEP-PDHQ (3 M), and the ability of each phytosterol to modulate membrane order was evaluated through RGM quantification and assessed over a range of temperatures from 5 to 40°C (Fig.  1A). The effect of each sterol on LUV membrane order level was only slightly enhanced by increasing temperature (Fig. 1A), and whatever the temperature, the same profile was observed (Fig.   1A). In detail, a strong ability of cholesterol to decrease the RGM value was observed, from 0.93 of LUV of DOPC/DPPC (1:1) to 0.56 upon cholesterol addition, at 20°C (Fig. 1A), confirming that cholesterol efficiently increases membrane order level, as widely described in the literature (71-78). (error bars) (n ϭ 5 or more independent repetitions). B, illustrative effect of the ability of each phytosterol to increase or decrease membrane order was evaluated at room temperature. RGM measured for LUVs containing DOPC/DPPC/phytosterol (1:1:1 in molar weight) was normalized to the control value without sterol. The ratio was expressed as a percentage. Tobacco mix, a mixture of the major sterols found in the PM of 7-day-old tobacco suspension cells (34% stigmasterol, 33% campesterol, 32% sitosterol). C, illustrative phase diagram for a ternary lipid mixture containing DOPC/DPPC/phytosterol mixture (1:1:1 in molar weight). The phytosterol mixture corresponds to various amounts of campesterol, stigmasterol, and sitosterol. The RGM of LUVs was measured (at least 5 independent experiments), and the mean value was reported according to the color scale.
We checked that LUV production buffer did not change the RGM. When prepared on a pH-controlled buffer, such as PBS, we measured similar RGM (e.g. 0.58 Ϯ 0.02 and 0.57 Ϯ 0.02 for DOPC/DPPC/cholesterol (1:1:1) in sucrose buffer and PBS, respectively), showing that our results are applicable at physiological ionic strength.
Campesterol presented an effect close to the that of cholesterol, by decreasing the RGM to 0.65 at the same temperature (Fig. 1A). ␤-Sitosterol seemed less efficient at changing the membrane order level and decreased the RGM only to 0.79 (Fig.  1A). The same amount of stigmasterol induced a significant increase of RGM value up to 1.10, above the value recorded for LUV of DOPC/DPPC (Fig. 1A), suggesting its surprising potency to decrease membrane order level.
The sterol content of plant cell membranes is made up by a mixture of phytosterols, the properties of which had thus far not been investigated in combination. We therefore analyzed membranes of LUVs containing stigmasterol, ␤-sitosterol, and campesterol in different proportions, with a final concentration of total phytosterols of 33%. We first characterized the mixture corresponding to tobacco suspension cell PM (i.e. equimolar concentration (ϳ10 mol %) of each phytosterol) (69). The RGM of such LUVs labeled with di-4-ANEPPDHQ significantly decreased from 0.93 to 0.74 Ϯ 0.05 (S.D.), compared with control vesicles of DOPC/DPPC (1:1). Interestingly, this decrease of 20% reflects the sum of effects of individual phytosterol (Fig.  1B), suggesting that each sterol could participate, in an additive or subtractive way, in membrane organization. Fig. 1C shows the order level for LUVs of DOPC/DPPC/different mixtures of phytosterols (1:1:1). As campesterol concentration increases in the LUV mixture, RGM continuously decreases right up to its minimum value (0.65) reached at the maximum campesterol amount (33%, when campesterol is the only sterol used), suggesting that its ability to organize membrane depends on its relative concentration (Fig. 1C). The same is true for the two other phytosterols tested, stigmasterol and ␤-sitosterol; they modify the LUV order level, depending on both their own ability to modify membrane organization (Fig. 1A) and their relative concentration (Fig. 1C). Altogether, these results confirm the various abilities of different phytosterols to order lipid bilayers and reveal the individual and independent involvement of the different molecular species of phytosterols in the global order level of model membranes.
Conjugated Sterols Strongly Order the Membrane of LUVs-PM contains SGs and ASGs, which are specific to the plant kingdom and some bacteria. They are 1.6 -3.4-fold enriched in detergent-resistant membrane fractions, suggesting their possible involvement in the formation of membrane ordered domains. When 33% of SGs or ASGs, mostly composed of conjugated ␤-sitosterol, were added to LUVs of DOPC/DPPC (1:1), the order level increased significantly, as indicated by the decrease of RGM (to 0.79 and 0.74, respectively; Fig. 2A). The decrease is similar to the one observed with free ␤-sitosterol (0.79; Fig. 2A), suggesting the same ability for free or conjugated forms of sterol to promote order in the lipid bilayer. When SG and ASG were used together, in an equimolar combination, a significant cooperativity to decreasing RGM was observed ( Fig.  2A). Because the total amount of conjugated sterols incorporated in LUVs was maintained at 33%, whatever the combination of SG and/or ASG, this result suggests a strong synergistic ordering effect of the two different sterol-conjugated forms in the bilayer.
To further study possible interactions between different forms of free or conjugated phytosterols, 10% of an equimolar combination of SG and/or ASG was added to LUVs composed of DOPC/DPPC/free phytosterols (1:1:1). This proportion of conjugated sterols corresponds to the one currently found in plant cell PM (32). Although the extra addition of 10% free ␤-sitosterol did not induce any modification of RGM of LUVs of DOPC/DPPC/free phytosterols (1:1:1; Fig. 2B), supplementation with 10% SG, mostly sterylated ␤-sitosterol, triggered a decrease of RGM to a very low value of 0.48 (Fig. 2B). The same effect was observed when ASG or SG/ASG together were added to LUVs containing free sterols (Fig. 2B). Altogether, these data indicate (i) that free phytosterols, SG and ASG, exhibit, alone, the same ability to increase membrane order; (ii) that a mixture of two different sterol forms is more efficient than free or conjugated sterols used separately; and (iii) that a combination of free and conjugated sterols is the most powerful way to increase membrane order.
Plant Sphingolipids Interact with Free Phytosterols to Increase the Order Level of Model Membranes-Although GIPCs are the predominant class of sphingolipids in plant cell PM (51,79,80), these molecular lipid species are not commercially available. We thus recently used a purification procedure described by Carter and Koob (41) modified by Bure et al. (55) on tobacco cell culture to obtain milligram amounts of GIPCs. The biochemical characterization and purity were controlled by GC-MS and TLC and revealed a 90% purified GIPC fraction, deficient in phytosterols and containing only 10% of glycerolipid contamination (phospho-and galactolipids). When 33% GIPCs were added to a DOPC/DPPC mixture (1:1), no significant change in the RGM of LUVs was observed (Fig. 3A). Glu-Cer is also present in plant PM (38), free long chain bases and ceramides being far less abundant (13). When GluCer was added to LUVs of DOPC/DPPC (1:1), alone or with GIPC, the order level did not change significantly (Fig. 3A), suggesting no particular ability of plant sphingolipids, either separately or in combination, to modify the order level of model membranes containing only phospholipids.
When either GIPC or GluCer were incorporated into LUVs containing DOPC/DPPC/different phytosterols, to a final molar ratio of DOPC/DPPC/sterol/sphingolipids of 0.5:0.5:1:1, the order level did not change significantly (Fig. 3B), except with stigmasterol, for which a slight increase in the presence of GIPC and decrease with GluCer was observed (Fig. 3B). However, an equimolar mix of GluCer and GIPC (33% final sphingolipid concentration) significantly increased the order level of LUV membranes (Fig. 3B). The RGM decrease was observed whatever the phytosterol initially present in the vesicles, but it showed different intensities, with a reduction of 32, 20, and 16% for ␤-sitosterol, campesterol, and stigmasterol, respectively. These data suggest that the combinations of molecular species of sphingolipids and sterols present in plant PM may increase membrane order level and that such a synergistic effect depends on the molecular species of phytosterols.

Membranes of LUVs Containing a Complex Plant Lipid Mixture, Including Free Phytosterols, SG, and ASG, Exhibit Very
High Order Level, with or without Phytosphingolipids-We finally used all of these different plant lipids together to evaluate the influence of the complexity of the lipid mixture on the membrane order level. As observed with pure sterol (Fig. 3B), no significant difference between RGM values of LUVs containing either DOPC/DPPC/free sterol tobacco mix (1:1:1), DOPC/DPPC/free sterol tobacco mix/GIPC (0.5:0.5:1:1), or DOPC/DPPC/free sterol tobacco mix/GluCer (0.5:0.5:1:1) was measured (Fig. 4A). However, the introduction of one conjugated sterol (SG or ASG, 10% final), together with a sphingolipid (GIPC or GluCer, 30% final), into LUVs induces a strong decrease of RGM (Fig. 4A), regardless of the molecular mixture of free phytosterol(s) (Fig. 4). A significant decrease (16 -44%) of RGM up to very low values (between 0.35 and 0.61) is obtained, whatever the initial RGM value (Fig. 4B). These results, together with those presented in Figs. 2 and 3, support the outstanding intrinsic ability of SG and ASG, separately or in combination, but independently of the presence of sphingolipid, to increase the order level of membranes.
Di-4-ANEPPDHQ Imaging Reveals Partitioning of L o and L d Domains at the Surface of GUVs-Spectrofluorimetry experiments described above were notably appropriate to measure the mean global order level of LUVs of many different compositions. To investigate the ability of different plant lipids to influence the spatial organization of the membrane, we next moved to confocal microscopy. GUVs with a diameter in the approximate range of 5-15 m, suitable for optical microscopy, were prepared using electroformation with indium tin oxidecoated microscope slides. This methodology gives a high yield of GUVs, about 10 3 vesicles starting from 2 g of lipids, most of them exhibiting a diameter greater than 10 m. The emission signal of fluorescently labeled GUVs was acquired on a tangent plane corresponding to membrane surfaces of 100 -350 m 2 (e.g. see Fig. 5), allowing a spatial characterization of bilayer organization. The fluorescence intensity of di-4-ANEPPDHQ labeling was captured simultaneously at different wavelengths, and corresponding ratiometric images were designed measuring simultaneously the intensities between 545 and 565 nm and between 635 and 655 nm and then dividing the emission intensity of the two band pass filters by one another. A further pixel-to-pixel analysis of red/green ratio (RGP, for red/green ratio of the pixel) was performed to map the order level of the membrane (Fig. 5A), as achieved previously on the cell membrane (67). When used on a classical animal lipid mixture, DOPC/sphingomyelin (SM)/cholesterol (4.5:4.5:1), this method allows observation of the partitioning of reddish and greenish regions indicative of the coexistence of large L o and L d areas (Fig. 5A). Labeling of the same GUVs with laurdan, another widely used order-sensitive membrane dye, confirms such lateral organization of L o and L d phases (data not shown), according to results previously obtained (81).
We then examined on the same material the partitioning of the DiI-C 18 (Fig. 5B) and Bdp-cholesterol (Fig. 5C), classical markers of the L d (11) and sterol-enriched phases (82). Both fluorophores exhibited a spatial segregation in distinct domains (Fig. 5, B and C). Ratiometric images obtained with di-4-ANEP-PDHQ were simplified to separate the RGPs corresponding to either L o or L d phases (Fig. 5D). To discriminate between L o and L d phase, we considered a threshold limit of 1.2 for RGP because pure DOPC GUVs exist in a unique L d phase at room temperature (11,81) and consistently exhibit more than 95% of their RGP values greater than 1.2. In these conditions, the pattern obtained with di-4-ANEPPDHQ (Fig. 5D) looks very similar to the ones obtained with classical fluorescent probes (Fig.  5, B and C), suggesting a striking correspondence between the L o and the cholesterol-enriched phase on one hand and the L d and DiI-C18 phases on the other hand. The proportions of L o and L d domains in different phase-separated GUVs deduced from RGP values are accordingly very similar to the one indicated by the partitioning of Bdp-cholesterol or DiI-C18 (Fig.  5E). A greater number of pixels with an RGP below 1.2 are counted when increasing cholesterol concentration, together with an increase of the area corresponding to Bdp-cholesterol accumulation (Fig. 5E), suggesting a close correspondence

between ordered domains (as revealed by di-4-ANEPPDHQ) and sterol-rich domains (as observed with Bdp-cholesterol). Free Phytosterols Exhibit Different Abilities to Spatially Organize Model Membranes and Create Ordered Domains-To
investigate the spatial organization of membrane of GUVs composed of different mixtures of plant lipids, we analyzed the di-4-ANEPPDHQ emission spectrum of each pixel (288 ϫ 288 nm) of the surface of these different vesicles (Fig. 6A), acquired as described above. Depending on the GUV composition, we clearly observed distinct organizations of reddish and greenish domains at the membrane surface (Fig. 6A) and, accordingly, different distributions of RGP values (Fig. 6B). For example, analysis of RGP values of DOPC/DPPC/stigmasterol vesicles revealed a continuous and flat distribution with a broad range of RGP values (from 0.5 to 3, in green in Fig. 6B). To precisely describe such RGP distribution, we measured its symmetry compared with a normal distribution using the Kurtosis value (Fig. 6C); data sets with Gaussian distribution exhibit a kurtosis of 0, and a negative kurtosis value reveals a uniform distribution that assigns equal probability to all RGP values. The flat shape exhibited by DOPC/DPPC/stigmasterol vesicles is thus in agreement with its negative kurtosis value (Fig. 6C), showing that no class of RGP value is significantly representative of the order level of these GUVs. This result suggests that, at this scale of observation, stigmasterol is not able to induce a spatial organization of the membrane into domains of a particular order level. On the contrary, GUVs containing ␤-sitosterol and campesterol exhibit a low RGM (Fig. 6C) and, accordingly, an RGP distribution centered on lower values, on 1.1 and 0.6, respectively (Fig. 6, B and C), consistent with their ability to order the membrane of LUVs (Fig. 1). RGP distributions were additionally sharper (Fig. 6B), in agreement with a positive Kur-tosis value (1.8 and 2.4 for ␤-sitosterol and campesterol, respectively; Fig. 6C). Such peaked distributions indicate the predominance of few RGP values, corresponding here to domains with a high order level (visualized in reddish colors in Fig. 6A). Focusing on these L o domains, we measured the representativeness of pixels of GUV membrane with an RGP of Ͻ1.2 (in black in Fig.  6D). The relative proportion of L o phase among total pixels is strictly correlated to ability of each phytosterol to modulate the global order level of the membrane (gray squares in Fig. 6E). Furthermore, a granulometric analysis reveals a parallel between the ordering ability of a given phytosterol and its potency to promote large L o domain formation (stigmasterol Ͻ ␤-sitosterol Ͻ campesterol; Fig. 6F). Altogether, our data propose various abilities of phytosterols to promote the formation and consequently the organization of L o domains at the surface of model membranes.
Interestingly, the RGP distribution of vesicles containing a phytosterol mixture exhibits a flat distribution indicative of a high diversity of RGP values (Fig. 6B), confirmed by the negative Kurtosis value (Fig. 6C). The shape of this RGP distribution corresponds to the mean (both in the height and in the width) of the ones of each phytosterol constituting the mixture (Fig. 6B), in agreement with global RGM values (Figs. 2 and 6C). However, the L o phase representativeness is much lower than the one predictable according to the corresponding RGM (Fig. 6E), and the few L o domains found at the surface of GUVs constituted of tobacco mixture appear clustered (Fig. 6, A and D). GUVs constituted of DOPC/DPPC/phytosterol mixture exhibited thus a low L o domain representativeness, similar to that of the stigmasterol (Fig. 6E), but the formation of domains bigger than 2 m, as in campesterol (Fig. 6F). These results suggest a  FEBRUARY 27, 2015 • VOLUME 290 • NUMBER 9 local cooperative effect of combined phytosterols to promote large L o domain formation.

Involvement of SG and ASG, Saturated Fatty Acids, and GIPC in Spatial Organization of Model Membranes-DOPC/DPPC/
sterol is a widely used mixture, well known to promote phase separation (14) and mimicking the saturated/unsaturated ratio found in plant cell. To evaluate the ability of plant sphingolipids to promote the formation of L o domains that separate from unsaturated PC, we performed experiments using DOPC/phytosterol GUVs as target lipid vesicles (Fig. 7). GIPC supplementation to a final composition of DOPC/phytosterol/GIPC (1:1:1) slightly increases the membrane order level of such GUVs, as revealed by the RGM decrease observed, from 1.46 to 1.31 (Fig. 7B). Furthermore, the addition of GIPCs enables the

Differential Effect of Plant Lipids on Membrane Organization
detection of groups exhibiting a size over 2 m (Fig. 7A) without modification of the relative amount of L o phase (Fig. 7B). GIPC-induced L o domain production exhibited an intensity similar to that induced by saturated fatty acid addition (Fig. 7B), even if the global order level was higher for DOPC/DPPC/phytosterol vesicles (1:1:1) than for DOPC/GIPC/phytosterol vesicles (1:1:1; Fig. 7B). Interestingly, an additive effect of sphingolipids and saturated fatty acid forms of PC was observed in DOPC/DPPC/GIPC/phytosterol mixture vesicles (0.5:0.5:1:1). Indeed, an enhancement of order level was measured, with an RGM decreasing to 1.04 and the relative amount of L o phase rising to 47%, nearly 5% of which appeared as domains with a size over 2 m (Fig. 7B).
Finally, the introduction of SG/ASG (5% final each) in vesicles of DOPC/GIPC/TM (30% final each) induced a shift to lower RGP values (Fig. 7A), consistent with the decrease of RGM (Fig. 7B), indicative of an increase of membrane order level, and with the increase of proportion of large ordered areas (Fig. 7B). When SG/ASG (5% final each) were incorporated into vesicles of DOPC/DPPC/GIPC/TM (15:15:30:30% final), the RGM decreased drastically, in agreement with a strong increase of low RGP values (Fig. 7A), the L o phase representativeness rising up to 82% (Fig. 7B). Although tricky due to the very high proportion of ordered phase, a granulometric approach reveals the formation of very large L o domains, with about 15% of L o phase present within domains above 2 M size. Altogether, our results propose the ability of GIPC and DPPC to increase membrane order level, similarly as observed previously with SM and DPPC (83), and also to promote L o domain aggregation. Moreover, a synergy is observed when GUVs are supplemented with the two lipid species together, and the strongest enhancement is seen when saturated and unsaturated PC, free and conjugated sterols, and sphingolipids are all represented. The diversity of lipid species contained in GUVs surprisingly led to the formation of an L o area that covered almost the entire membrane surface, suggesting that all lipids could be among the main actors involved in the membrane order level heterogeneity.

Free and Conjugated Phytosterols Exhibit Various Abilities
That Could Act in Combination to Order Lipid Bilayer-The first investigations identified a planar ring system, a 3␤-hydroxyl group, and an isooctyl side chain requirement for cholesterol's ability to organize membrane and the ⌬ 5 double bond involvement in cholesterol-phospholipid interaction (84). Major plant sterols all satisfy these conditions and are accordingly able to stabilize lipid bilayer (85), to order phospholipid acyl chain, and consequently to reduce membrane water per-FIGURE 6. Free sterols promote various spatial organization of membrane. All experiments were performed using GUVs composed of DOPC/DPPC/ phytosterol (1:1:1), sterol being either stigmasterol, ␤-sitosterol, campesterol, or tobacco mix (TM), which represents a mixture of major sterols found in tobacco suspension cell PM (34% stigmasterol, 33% campesterol, 32% sitosterol). For all lipid mixtures, the final concentration of total sterols was 33 mol %. A, GUVs composed of different lipid mixtures were labeled with di-4-ANEPPDHQ (3 M), and the fluorescence was recovered by multispectral confocal microscopy in a tangent plane, as described under "Material and Methods." The RGPs were measured for each pixel (288 ϫ 288 nm) constituting the membrane surface area of labeled GUVs and were pseudo-color-coded using the color scale shown at the side, hot colors corresponding to higher order level (low RGP). B, distribution of RGP values of individual pixels of the membrane of GUVs and influence of the different phytosterols. The x axis corresponds to the class of RGP values, and only the maximal value of each class is reported on the graph. The y axis corresponds to the percentage of each class. C, characterization of RGP distribution, reporting the peak and the Kurtosis value (for more details, see "Materials and Methods"). D, central regions (3.5 ϫ 3.5 m) of the GUV top surface were cropped in order to avoid edge effects, and pixels exhibiting a RGP value below a fixed threshold of 1.2 were colored in black. E, the density of black pixels was evaluated and correlated to the global RGM. F, the size of black pixel groups was evaluated for the different GUV mixtures and reported according to their relative abundance. Data shown are mean values (n ϭ 21-37 GUVs). FIGURE 7. Plant lipids promote various spatial organization of membrane. GUVs were produced using different lipid mixtures; the proportion of each lipid is given either in molar ratio or in mol % at the top of A. Tobacco mix (TM) represents a mixture of major sterols found in tobacco suspension cell PM (34% stigmasterol, 33% campesterol, 32% sitosterol). Vesicles used for A and B are identical. A, GUVs composed of different lipid mixtures were labeled with di-4-ANEPPDHQ (3 M), and the fluorescence was recovered by multispectral confocal microscopy in a tangent plane, as described under "Materials and Methods." The RGPs were measured for each pixel (288 ϫ 288 nm) constituting the membrane surface area of labeled GUVs and were pseudo-color-coded using the color scale reported at the side, hot colors corresponding to higher order level (low RGP). B, characterization of model membrane spatial organization. Global order level was measured as mean RGP value. Focusing on pixels exhibiting an RGP value below a fixed threshold of 1.2 (ordered domains), their amount was reported for the different GUV mixtures according to their relative abundance (expressed as the percentage of pixels exhibiting an RGP value below 1.2, relative to the total amount of pixels). The representativeness of large ordered domains (Ͼ2 m) was expressed as the percentage of pixels present in such domains among all pixels exhibiting an RGP below 1.2. FEBRUARY 27, 2015 • VOLUME 290 • NUMBER 9 meability (86), although less efficiently than mammalian sterol (87)(88)(89). Here, we described a strong ordering ability for campesterol, in the same range as the cholesterol one, and a less efficient one for ␤-sitosterol (Fig. 1). Accordingly, campesterol exhibits a high potency to stabilize DPPC bilayer (90) and to tightly pack the phospholipid bilayer (91,92), which has been attributed to its smaller freedom of rotation at the C17-hydrocarbon chain. Stigmasterol was previously found to have a much weaker ordering effect than other sterols on diphenylhexatriene anisotropy (93). In agreement, we described here its lack of ability to order lipid bilayer (Fig. 1) and accordingly to promote L o phase formation (Fig. 6). Measurements of parinaric acid fluorescence polarization have similarly shown that stigmasterol is unable to obliterate phospholipid phase transition and to interact with L o phase, whatever the phosphatidylcholine analog used to produce target lipid vesicles (94). Our data thus confirm the specific capacity of phytosterols to modulate the order level of phospholipid bilayers (stigmasterol Ͻ ␤-sitosterol Ͻ campesterol), by using RGM as a new parameter. Such individual phytosterol ability to order membrane was found to only slightly depend on temperature, as previously demonstrated for DOPC/DPPC/cholesterol vesicles (84). Indeed, below 40°C, the temperature at which the L o and L d domains separated in the presence of 33% cholesterol (84), RGM measurement assesses the global order level without taking into account local and dynamic phase transitions.

Differential Effect of Plant Lipids on Membrane Organization
However, phytosterols exist within plant PM as a mixture of stigmasterol, campesterol, and ␤-sitosterol, relative amounts of which vary among plants, tissues, and growth conditions. For example, the ␤-sitosterol/stigmasterol ratio in the membrane has been known to influence the response of a cell to various biotic (95) or abiotic stresses (e.g. in the resurrection plant Ramonda serbica following dehydration and rehydration (96) and in broccoli root in response to salt stress (97)). Using vesicles containing different phytosterol mixtures, we clearly showed that the relative amounts of each phytosterol finely control the membrane order level (Fig. 1), which may be meaningful to allow plant adaptive responses to changes in environmental conditions. In agreement, using sterol biosynthesis mutants, a higher level of stigmasterol within plant PM has been correlated with altered fluidity and permeability characteristics (e.g. a restricted solute efflux) (98,99). Furthermore, the relative increase of more planar sterols, such as campesterol, reduced the membrane permeability to charged ions (87,100) and might favor ion exclusion. The molar ratio between more and less planar sterols thus plays a fundamental role in allowing the plant to tolerate abiotic stress because the ratio is considered to be an index of membrane permeability and functioning (101). Although the precise mechanism by which phytosterols, in combination, regulate membrane properties is not yet understood, we show here that their diversity is a powerful tool, through the relative amounts of their different molecular species, to modulate and to extend possible values of the order level of the membrane (Fig. 1).
Beside free sterols, SGs and ASGs are ubiquitously present in vascular plants, where the composition of sterols in SG usually reflects the free sterol composition (33). Interestingly, the proportions of SG and ASG in PM differ greatly depending on the plant species and the growth conditions, varying from 8% SG plus ASG in Arabidopsis thaliana leaves to 6% SG and 27% ASG in spring oat (Avena sativa) leaf (102,103). The similar ability of conjugated and free forms of phytosterols alone to order membrane evidenced in this study (Fig. 3) is in agreement with the similar efficiency of steryl glycosides, acylated steryl glycosides, and free sterol mixture isolated from soybean to eliminate the thermal phase transition of dipalmitoyl lecithin (104).
However, the ordering ability of free and conjugated forms is strongly enhanced when these different forms of phytosterols are added in combination (Fig. 3). These data demonstrate an impact of ASG/SG amount on membrane physical properties and suggest that the changes in the ratio of conjugated to free forms of sterol could be a major key to modulate the phase behavior of PM, as observed, for example, during dehydration (105). In agreement with such a model, compensatory changes in free sterol versus SG plus ASG suggest an interconversion of the lipids during cold acclimation in rye seedlings (106) or in A. thaliana (107).
Plant Lipid Diversity Determines Spatial Organization of Membrane Order Level-Lateral segregation of lipids has been proposed as one of the possible actors in the modulation of membrane physical properties, leading to L o and L d phase separation (6,7). The DOPC (T m ϭ Ϫ20°C)/DPPC (T m ϭ 41°C) lipid mixture, which has been widely used in the literature to characterize the physical properties of the membrane model system (11,14,83,108,109), exhibits, at temperatures below 45°C, an L o and L d phase coexistence (83). Performing our experiments at 22°C and minimizing temperature fluctuations, we actually favored a long time equilibrium formation of L o and L d phase separation, allowing us to focus our study on the structuring effect of plant lipid on L o and L d phase partitioning. Using multispectral confocal microscopy on giant model membranes labeled with di-4-ANEPPDHQ dye, previously shown to be relevant to characterize L o and L d phase partitioning (54,62), we thus identified a sterol-dependent ordered domain formation (Fig. 5). Such ordered phase formation has been previously associated with the appearance of sterol-rich domains (10). We observed here a close correlation between L o and L d phases and the sterol-enriched and sterol-deprived areas of the membrane, respectively (Fig. 5), suggesting close correspondence of ordered domains and sterol-rich domains, as proposed in the membrane raft hypothesis (8).
Analysis of spatial organization of the membrane performed in this study demonstrates the differential ability of phytosterols to (i) promote the occurrence of ordered domains within the membrane and (ii) organize them at the membrane surface, both related to their ability to increase the global order level of the bilayer. In agreement with our observation, ␤-sitosterol, as a relevant example, has been proved to induce the production of numerous small domains, increasing the surface of L o -L d transition regions (110). Furthermore, we interestingly propose that combined phytosterols exhibit both an additive/subtractive effect to modulate the global order level of the membrane and a cooperative effect to enhance larger L o domain formation. The lateral distribution of L o domains is also strongly affected by the addition of phytosterol-conjugated forms and/or GIPC (Fig. 7B). We thus observed that free sterol-sph-ingolipid and free sterol-SG/ASG combination enhance L o domain representativeness (Fig. 6E) and enlarge ordered domain size (Fig. 6F), arguing in favor of the "membrane raft" model (8). Indeed, SG/ASG are enriched in detergent-resistant membranes (32,111,112), in good agreement with their ability to promote L o domain formation. Moreover, gangliosides, which are structural analogs of plant GIPCs in the mammalian model, are commonly used as raft markers (10,81,(113)(114)(115). Our data demonstrate that each plant lipid class could partly account for the complex lateral distribution of ordered domains within the lipid bilayer. One must note that the diversity of molecular lipid species and their specific amount are very different in animal and plant membranes. For example, animal gangliosides are present in very low amounts in animal membrane, less than few percent. This is not at all the case for plant GIPCs, which represent the major sphingolipids (13). In addition, cholesterol is the sole sterol found in mammal membranes compared with the high diversity of sterol mixture in plant PM. Such potency to interact and to subtly and spatially organize the membrane gives sense to the huge diversity of plant membrane lipids.
Different interesting mechanisms could account for the change in vesicle order level induced by lipids. We interestingly observed no correlation between the global order level, as measured using vesicle RGM, and the L o domain representativeness at the GUV surface, which varies independently when lipids exhibiting ordering ability in LUV are added. For relevant examples, GUVs containing DOPC/DPPC/sitosterol, although exhibiting similar RGM as GUVs containing DOPC/DPPC/ TM, display a different L o domain representativeness (Fig. 6). Moreover, GIPC supplementation, even if it slightly changes the RGM value (Fig. 7), significantly increases the L o domain amount. These results might support a first mechanism, in which an increase of L o domain proportion explains the RGM changes observed. A second possibility is that the order has gone up inside L o domains. In agreement with this hypothesis, GUVs containing either DOPC/DPPC/TM or DOPC/DPPC/ stigmasterol show similar L o domain representativeness, although the first ones exhibit a significantly lower RGM. Likewise, GUVs containing either DOPC/TM or DOPC/DPPC/TM exhibit similar L o domain amounts (34 and 39%, respectively) even if their respective RGM significantly decreases (from 1.46 to 1.14), showing an increase of order level of L o domains with the DPPC addition. Finally, when SGs and/or ASGs are added, both the amount and the order level of L o domains increase, leading to a very high order level of the vesicle membrane.
Altogether, our data interestingly suggested the coexistence of two different mechanisms controlling the local membrane order level, which one accounting for the RGM modification depending on the extra lipid added.
Sphingolipids, Notably GIPCs, Interact with Sterol to Organize L o Domains-Data acquired in model systems by x-ray diffraction, differential scanning calorimetry, and polarized light microscopy indicated that a SM lamellar gel phase and a liquid crystalline phase containing both SM and PC coexist at room temperature (116). More recently, controversy has arisen because no phase separation for DOPC/bovine brain SM (7:3) occurs at 20°C (77,117). Our data show no ability of GIPC, the major plant sphingolipid (51), to order the phospholipid bilayer (Fig. 3) and argue in favor of no extensive phase separation in the binary sphingolipid/phospholipid system. We, however, showed the ability of GIPCs to increase, in a phytosterol-dependent manner, the global order level of the membrane (Fig. 3), even in the absence of DPPC (Fig. 7), mimicking the SM-cholesterol effect (83). In the mammalian model, lateral partitioning of L o and L d phases is thus explained by a preferential interaction of sphingolipids with cholesterol over phospholipids (22), probably due to better shielding from water by the SM headgroup. Their partitioning into L o phase is favored by the fact that sphingolipids can create a complex network of hydrogen bonds due to the presence of the amide nitrogen, the carbonyl oxygen, and the hydroxyl group positioned in proximity to the water/lipid interface of the bilayer (118). Moreover, the hydrogen bond network between lipids stabilizes a more rigid phase in the bilayer and increases order (119,120). Plant GIPC long chain base profiles are dominated by t18:0 and t18:1 species in widely varying proportions (t representing trihydroxylated, called phytosphingosine; 0 or 1 for no or one insaturation; for a review, see Ref. 48), and one hydroxyl residue is very often present at position 2 of the fatty acid. We observed a strong ability of GIPCs to induce, in a sterol-dependent manner, L o domain formation and, in particular, to increase their size (Fig. 7), suggesting that a strong interaction between phytosterols and GIPCs might lead to a well defined phase separation, in agreement with the hydrogen-bond network model. One may hypothesize that the presence of the three hydroxyl groups at the interface between the polar phase and hydrophobic phase of the bilayer may be of importance for phytosphingolipid/phytosterol interactions.
Animal glycosphingolipid with a monosaccharide headgroup is much bulkier than phospholipid ones and occupies a larger volume, as proposed by theoretical calculations of the minimum energy conformation, which results in spontaneous membrane curvature (25). According to these predictions, ganglioside phase separation in model membrane depends on the surface area occupied by the oligosaccharide chain that is usually directly correlated with the number of sugar residues (121)(122)(123). The major GIPC polar head is composed of N-acetylglucosamine/glucuronic acid/inositol/phosphate, and up to seven sugar moieties can be added, particularly in a tobacco BY-2 cell culture (55), from which GIPCs were isolated in this study. Therefore, the volume occupied by the headgroup of GIPC is far bulkier than phospholipid headgroups, explaining their high propensity for partitioning in L o domains.
In summary, we demonstrated the ability of free and conjugated forms of phytosterols to modulate the model membrane order. Moreover, we propose that, in the PM of plant cells, phytosterols could be key compounds to mediate the formation of lipid ordered domains, through interactions with other plant lipids, such as GIPCs. Our data allow new insights into how the diversity of plant lipids might drive the PM subcompartmentalization, allowing the dynamic segregation of membrane components, in the framework of the membrane raft hypothesis (124 -126). The diversity of plant lipid mixture retrieved in plant cell PM raises the question of the coexistence of lipid domains of various compositions and consequently different FEBRUARY 27, 2015 • VOLUME 290 • NUMBER 9 JOURNAL OF BIOLOGICAL CHEMISTRY 5821 lipid-lipid interactions, which have now to be characterized in vivo. Furthermore, the membrane raft hypothesis expands the field of such investigations, suggesting the possible involvement of protein-lipid, membrane-cytoskeleton, and membrane-cell wall interactions that need to be further investigated.