The activity of the saponin ginsenoside Rh2 is enhanced by the interaction with membrane sphingomyelin but depressed by cholesterol

The membrane activity of some saponins, such as digitonin or alpha-hederin, is usually attributed to their interaction with membrane cholesterol (Chol). This contrasts with our recent publication showing that Chol, contrary to sphingomyelin (SM), can delay the cytotoxicity of the saponin ginsenoside Rh2, challenging the usual view that most saponins mediate their membrane effects through interaction with Chol. The aim of the present study was to elucidate the respective importance of Chol and SM as compared to phosphatidylcholine (PC) species in the membrane-related effects of Rh2. On simple lipid monolayers, Rh2 interacted more favorably with eggSM and DOPC than with Chol and eggPC. Using Large Unilamellar Vesicles (LUVs) of binary or ternary lipid compositions, we showed that Rh2 increased vesicle size, decreased membrane fluidity and induced membrane permeability with the following preference: eggSM:eggPC > eggSM:eggPC:Chol > eggPC:Chol. On Giant Unilamellar Vesicles (GUVs), we evidenced that Rh2 generated positive curvatures in eggSM-containing GUVs and small buds followed by intra-luminal vesicles in eggSM-free GUVs. Altogether, our data indicate that eggSM promotes and accelerates membrane-related effects induced by Rh2 whereas Chol slows down and depresses these effects. This study reconsiders the theory that Chol is the only responsible for the activity of saponins.

In this study, we focus on the ginsenoside Rh2 (structure presented in Fig. 1A), found in Panax ginseng and responsible to some extent for its medicinal properties and health benefits 15 . This well-known steroid saponin has been reported to disrupt L o domains leading to apoptosis, via either the death receptor FAS oligomerization in human cervical cancer Hela cells 16 or inactivation of the serine/threonine kinase Akt in human epidermoid carcinoma A431 and breast cancer MBA-MB-231 cells 16,17 . In addition, our recent publication showed that the cytotoxicity of Rh2 is accelerated by membrane Chol depletion but delayed by SM depletion in human monocytic leukemia U937 cells 18 . These observations led to the hypothesis that, contrary to membrane SM, Chol could delay the activity of ginsenoside Rh2. The aim of the present study was thus to obtain detailed knowledge about the membrane-related effects of Rh2 and to elucidate the respective importance of Chol and SM.
To these aims, we first evaluated the interaction of Rh2 with membrane in silico and Langmuir experiments on simple lipid monolayers composed of egg sphingomyelin (eSM), egg phosphatidylcholine (ePC),

Results
Rh2 interacts more favorably with pSM and DOPC than Chol as revealed by molecular modeling. Firstly, molecular modeling was used to investigate the structure and molecular interactions of Rh2 with rigid 3D structures of palmitoylsphingomyelin (pSM), Chol and DOPC. The IMPALA method was used to predict the potential insertion of Rh2 into an implicit lipid bilayer 19 . The apolar aglycone moiety of Rh2 was deeply embedded into the bilayer as indicated by its most energetically stable position at ±12 Å (Fig. 1B). Interaction energies between the saponin and the aqueous medium outside of the membrane were highly positive, thus less favorable, suggesting that the most favorable state of Rh2 was within the bilayer (Fig. 1C). In addition, high energies were predicted at the bilayer center, suggesting that Rh2 would not be able to cross the bilayer. A docking method called Hypermatrix 20 was used to calculate the energies of interaction between the central molecule, Rh2, and the surrounding lipids. These molecular assemblies showed more favorable interactions between Rh2 and pSM or DOPC in comparison to Chol molecules (Fig. 1D,E). Calculations predicted that the interactions were mainly apolar in nature (grey bars in Fig. 1D).
Rh2 binds to lipid bilayer vesicles with the following preference: DOPC > eSM > epC. Isothermal Titration Calorimetry (ITC) experiments were then applied to thermodynamically quantify the interaction of Rh2 with liposomes containing eSM, ePC or DOPC. Notice however that formation of liposomes with Chol was not achieved. LUVs were injected in steps into the sample cell containing Rh2 (Rh2/lipid ratio of 0.02) at 25 °C and thermodynamic parameters were determined ( Table 1). The free energy of binding values (ΔG) were negative and relatively similar for the three compositions examined, indicating a favorable and spontaneous interaction. The entropy values (ΔS) were all positive, clearly indicating that the binding of Rh2 with the three types of vesicles was entropically favored but with differences depending on the lipid (eSM > DOPC > ePC). Regarding binding constants, the affinities of Rh2 with lipid bilayers occurred with the following preference: DOPC > eSM > ePC. This could be linked with the difference between DOPC and eSM or ePC liposomes in their interactions with Rh2 as revealed by the enthalpy values (∆H). For instance, the interaction of Rh2 with eSM and ePC was endothermic (∆H > 0) with higher positive enthalpic changes of eSM compared to ePC whereas Rh2 binding with DOPC was found to be exothermic (∆H < 0).

Rh2 adsorbs to a higher extent with eSM and DOPC than with Chol and ePC monolayers.
To explore Rh2:Chol interaction in comparison with that of Rh2 with eSM, ePC and DOPC, we pursued our study on lipid monolayers made of these lipids by measuring adsorption kinetics via the Langmuir trough method. Lipids were spread at the water-air interface to obtain an initial surface pressure of approximately 30 ± 2 mN/m. Rh2 was then injected into the buffer subphase at the final concentration of 10 μM and its adsorption to the different films was determined by measuring the surface pressure (Π) variations as a function of time with a constant film area. The injection of Rh2 beneath the eSM, Chol, ePC or DOPC monolayers gave rise to a rapid increase of Π over time ( Fig. 2A). The ΔΠ upon the adsorption of Rh2 to lipid monolayers was observed with the following preference: eSM > DOPC > ePC and Chol monolayers (Fig. 2B) in agreement with results from Hypermatrix.
Rh2 increases eSM packing and transition temperature. Since Rh2 was able to highly adsorb to eSM monolayers, we then determined the effect of Rh2 on the eSM hydration and main transition temperature by evaluating the generalized polarization (GPex) of Laurdan upon increasing temperatures (Fig. 3) 21,22 . A transition temperature (T m ) around 41 °C was observed for eSM, i.e. close to the values reported in the literature using differential scanning calorimetry 23,24 . A Rh2/lipid ratio of 1 increased the value of Tm to 45 °C. In addition, whatever the temperature, the GPex was higher in presence of Rh2 suggesting that the ginsenoside decreased hydration of eSM. Thus, we suggest that upon addition of Rh2, the packing order would be reinforced giving rise to lower lipid lateral diffusion coefficients. These results are in agreement with the impairment of eSM:water hydrogen-bonds and packing induced by Rh2 as suggested by ITC. www.nature.com/scientificreports www.nature.com/scientificreports/ Rh2 adsorption on mixed lipid monolayers is enhanced in the presence of eSM and the absence of Chol. We next investigated the interaction of Rh2 with binary or ternary lipid systems. We first determined if and how Rh2 could adsorb to lipid monolayers made of either eSM:ePC (1:1), ePC:Chol (1:1) or eSM:ePC:Chol (1:1:1). Rh2 increased the Π within 500 s in eSM:ePC monolayer but only after 1400 s in monolayers containing Chol (Fig. 4A). A plateau was observed in all conditions at 2600 s but a higher ΔΠ was measured for eSM:ePC as compared to eSM:ePC:Chol and ePC:Chol monolayers (Fig. 4B). These data indicated that Rh2 adsorbed on mixed lipid monolayers with the following preference: eSM:ePC > eSM:ePC:Chol and ePC:Chol membrane.
Rh2 increases the size of LUVs containing eSM but no Chol at a low Rh2/lipid ratio. Having shown the favorable adsorption of Rh2 with mixed lipid monolayers in the presence of eSM and the absence of Chol, we then turned our attention to mixed lipid bilayers as a more relevant model system. We first assessed the size of LUVs by dynamic light scattering (Fig. 5). At a Rh2/lipid ratio of 0.4, the size of eSM:ePC liposomes increased from ~140 to ~190 nm, while vesicles containing Chol remained unchanged. However, at a Rh2/lipid ratio of 2, the size of vesicles increased regardless of the liposome composition with an average vesicle diameter of ~225, ~210 and ~190 nm for eSM:ePC, eSM:ePC:Chol and ePC:Chol, respectively. These results indicated that Rh2 increased more favorably the LUV size in the presence of eSM and the absence of Chol in agreement with the preferential binding of Rh2 with monolayers containing eSM but no Chol (Fig. 4).
Rh2 increases membrane rigidity in the presence of eSM and the absence of Chol at a low Rh2/ lipid ratio. To next determine if Rh2 could affect membrane biophysical properties to a differential extent  www.nature.com/scientificreports www.nature.com/scientificreports/ based on membrane composition, we started by evaluating membrane fluidity by the Laurdan generalized polarization (GPex; Fig. 6A) and the DPH fluorescence anisotropy (r; Fig. 6B) on the three liposome compositions. Laurdan locates at the level of the lipid glycerol backbone and allows to detect variations in membrane phase properties through its sensitivity to the environment polarity 22 while the DPH probe is in the hydrophobic core of the membrane and is sensitive to changes in lipid acyl chain physical properties 25 . The increase of these parameters indicates a higher rigidity in the lipid core region. In the absence of Rh2, membranes composed of eSM:ePC, ePC:Chol and eSM:ePC:Chol were characterized as the more fluid to the more rigid. After 5 min of incubation with Rh2 at a 0.4 molar ratio, GPex and anisotropy values were increased in Chol-free liposomes (eSM:ePC) but not in Chol-containing liposomes (ePC:Chol and eSM:ePC:Chol). At a Rh2/lipid molar ratio of 1, Rh2 significantly increased the eSM:ePC:Chol membrane rigidity whereas ePC:Chol exhibited no significant rigidity increase as compared to the control condition. A Rh2/lipid molar ratio of 2 was needed to increase the rigidity of the ePC:Chol system. Thus, it clearly appears that after addition of a low Rh2/lipid ratio, membrane rigidity only increased in the presence of eSM and the absence of Chol.
Rh2 induces membrane fusion to a faster and higher extent in the absence of Chol. We next evaluated the ability of Rh2 to induce membrane fusion by using the self-quenching probe R 18 (octadecyl rhodamine B chloride). Fusion of R 18 -labeled liposomes was monitored via the increase of fluorescence over time,  www.nature.com/scientificreports www.nature.com/scientificreports/ which results from the relief of R 18 self-quenching due to the decrease in its surface density. At a Rh2/lipid ratio of 0.4 which increased the membrane rigidity in eSM:ePC system (Fig. 6), no R 18 dequenching was observed (Fig. 7, ratio of 0.4). In contrast, at higher Rh2/lipid ratios, a plateau was reached after 25 s for all liposome compositions ( Fig. 7, ratios of 1, 2 and 4). Importantly, Rh2 provoked faster and higher increase of R 18 dequenching over time in Chol-free liposomes (eSM:ePC) as compared to Chol-containing liposomes (eSM:ePC:Chol and ePC:Chol). These results suggested that Rh2-induced membrane fusion to a faster and higher extent in the absence of Chol.

Rh2 accelerates liposome permeability in the presence of eSM and the absence of Chol.
To investigate if the insertion of Rh2 into the membrane could induce membrane permeability, the release of calcein from LUVs was measured. Membrane permeabilization is evaluated by measuring the increase of fluorescence emission of LUVs entrapping self-quenched calcein in case of release across the permeabilized membrane. At a Rh2/lipid molar ratio of 4 for which Rh2 was able to induce membrane fusion (Fig. 7), there was no calcein release (Fig. 8, ratio of 4). At a ratio of 10, the fluorescence profile of the three types of liposomes can be discriminated. For instance, after 50 s of incubation with Rh2 (at which a plateau was obtained), 50% of calcein was released from eSM:ePC, ~25% from eSM:ePC:Chol while no release was measured from ePC:Chol liposomes (Fig. 8, ratio of 10). Treatment with a Rh2/lipid ratio of 20 induced a time-dependent increase of fluorescence (Fig. 8, ratio of  20). These results showed that Rh2 accelerated membrane permeability in the presence of eSM and the absence of Chol.
Rh2 induces positive membrane curvature in the presence of eSM and small buds followed by intra-luminal vesicles in its absence. We then took benefit from the resemblance of GUVs with cell membranes in terms of size and membrane curvature to visualize if Rh2 could change the GUV morphology. To this aim, GUVs were labeled with 1,2-dioleoyl-sn-glycero-3-phosphoethanolamine-N-(lissamine rhodamine B sulfonyl) (Rhod-PE, red channel) and N-(7-Nitrobenz-2-Oxa-1,3-Diazol-4-yl)-1,2-Dihexadecanoyl-sn-Glycero-3-Phosphoethanolamine (NBD-PE, green channel) to probe the liquid-disordered (L d ) and liquid-ordered (L o ) phases, respectively 26 . The time course of morphology changes was studied by incubating GUV suspensions with a fixed Rh2/lipid ratio of 3, for which membrane fusion but not permeability occurred (see above results). Control GUVs exhibited a spherical shape with no detectable morphology changes for the whole duration of the experiment (data not shown). In control eSM:ePC GUVs, solid-ordered phases (S o ) were observed (dark areas, white arrows at Fig. 9A, 0 min), as expected 27,28 . Following the same GUV upon addition of Rh2, the progressive formation of positive membrane curvature was accompanied by the progressive vesicle budding after 1 and 2 min www.nature.com/scientificreports www.nature.com/scientificreports/ (Fig. 9A, 1 and 2 min). Then, we were able to evidence membrane fusion and fission from 5 to 8 min (Fig. 9A, 5 to 8 min). In ePC:Chol GUVs, no phase separation was observed in the absence of Rh2 (Fig. 9B, 0 min). After 1 min, Rh2 promoted the formation of small buds (grey arrows) around the membrane that remained connected to the parent GUV, consistent with the induction of membrane curvature changes (Fig. 9B, 1 min). After 9 min, the small buds progressively disappeared but inward budding of GUV membranes appeared (white arrowheads at Fig. 9B, 9 min), leading to intra-luminal vesicles (ILVs) after 10 min (grey arrowheads at Fig. 9B, 10 min). In eSM:ePC:Chol, lipid phase separation was observed, as expected 23,28 (Fig. 9C, 0 min). Over the course of Rh2 treatment, positive membrane curvature occurred and L d domain preferentially localized into these highly curved domains (Fig. 9C, 1 and 10 min). These results suggested that the insertion of Rh2 into the membrane promoted positive membrane curvature in the presence of eSM and the formation of small buds followed by ILV formation in the absence of eSM.

Rh2-induced morphology changes are not directly associated with membrane permeabilization.
To determine the potential link between GUV morphology changes and Rh2-induced membrane permeability,  www.nature.com/scientificreports www.nature.com/scientificreports/ GUVs were placed in IBIDI chambers filled with glucose and calcein and were observed by confocal microscopy. Control GUV membranes containing sucrose were impermeable to free calcein present in the solution, showing a difference of fluorescence intensity between outside and inside of the GUVs (values noted in the merge panels;  www.nature.com/scientificreports www.nature.com/scientificreports/ Fig. 10 A-C; line 1). This difference was maintained for the whole duration of the incubation with DMSO (data not shown). At a Rh2/lipid ratio of 3 and in the presence of eSM, the membrane budding induced by Rh2 reported in Fig. 9A,C was clearly visible but did not allow the entry of calcein into the GUVs (Fig. 10A,C; lines 2 and 3). In ePC:Chol GUVs, the formation of calcein-free small buds mediated by outward budding of GUV membranes was observed already after 1 min (Fig. 10B; line 2) while the formation of ILVs was noticed after 10 min (Fig. 10B; line  3). The difference of fluorescence intensity between outside the GUVs and inside those ILVs was 80 ± 36 which was lower than the difference between outside and inside control GUVs. This indicated that the calcein contained in the solution was encapsulated inside small vesicles during their formation, consistent with the formation of calcein-containing ILVs through inward indentation of the limiting GUV membrane 29 . Thus, Rh2-induced morphology changes were not directly associated with membrane permeability. A Rh2/lipid ratio of 6 was needed to induce wrinkled borders and entry of calcein through the membrane of the three GUV compositions, leading to www.nature.com/scientificreports www.nature.com/scientificreports/ the suppression of difference of fluorescence intensities between outside and inside of GUVs (Fig. 10A-C; line 4). The latter results are compatible with the creation of membrane pores by Rh2 whatever the GUV composition.
In contrast to Rh2, digitonin modulates membrane biophysical properties in a Chol-dependent manner. Altogether, the data reported above indicated that Rh2 modified to a higher extent biophysical properties of LUVs in the presence of SM and the absence of Chol. Since those results challenged the usual view that some saponins mediate their membrane effects through interaction with Chol, the same series of experiments were carried out with digitonin, a steroid saponin known to require Chol for its membrane-related effects 11 . Using the same three liposome compositions, we evaluated the effects of digitonin on vesicle size (dynamic light scattering, Fig. 11A), membrane fluidity (Laurdan, Fig. 11B), fusion (R 18 dequenching, Fig. 11C) and permeability (calcein release, Fig. 11D). The results clearly showed that treatment with digitonin increased vesicle size, membrane fluidity and permeability only in Chol-containing systems without any effect on membrane fusion even at a Rh2/lipid ratio of 13 (data not shown). Interestingly, ePC:Chol vesicles were more susceptible to digitonin as compared to eSM:ePC:Chol, suggesting an influence of Chol concentration. Consistently, digitonin rapidly disrupted the GUV membrane in the presence of Chol but not in its absence (Fig. 11E) even after long time (2 hours, data not shown). Altogether, these results supported the essential role of Chol in the membrane-related effects induced by digitonin and corroborated our observations that Rh2 interacted with the membrane in a differential way.

Discussion
The membrane-related effects of saponins are often considered as Chol-dependent. Our recent observations with the saponin ginsenoside Rh2 let us to reconsider this purpose and to propose that, in contrast to eSM, membrane Chol could decrease the activity of Rh2 18 . We here tested this hypothesis by determining Rh2:membrane interaction by in silico experiments as well as in vitro studies on lipid monolayers and bilayers while modulating the lipid composition, i.e. the presence or absence of Chol or eSM. In silico studies clearly revealed higher relative interaction energy between Rh2 and pSM or DOPC in comparison to Chol, which could result from higher hydrophobic and van der Waals forces but also higher hydrogen bond energies. Accordingly, studies on lipid monolayers showed that Rh2 adsorbed to a higher extent to eSM or DOPC in comparison to Chol and ePC monolayers.
As revealed by ITC, interactions between Rh2 and both eSM and DOPC are favorable with negative values for variation of free energy. They differ however since interaction with eSM was endothermic and mainly driven by entropy whereas interaction with DOPC was slightly exothermic with lower values for variations of entropy. Regarding the interaction of Rh2 with ePC and DOPC, the binding affinity between Rh2 and ePC was lower than with DOPC vesicles which could indicate that variations in the fatty acyl chain lengths, the number of unsaturation found in ePC composition (containing around 30% DOPC) and the resulting biophysical membrane properties could reduce membrane binding of Rh2.
To be closer to physiological condition, we then turned to mixed lipid bilayers. Even if biological membranes are more complex than artificial lipid bilayers (due to the presence of membrane and cytoskeletal proteins but also of membrane asymmetry), our data are consistent with those previously obtained in human cancer cell lines (A549, THP-1, and U937) which highlighted that membrane Chol depletion accelerates Rh2-induced apoptosis 18 . It can be deduced that Rh2 did not specifically interact with Chol which could instead delay and/or decrease the Rh2-induced membrane-related effects. Several hypotheses can be proposed to explain such a delayed/decreased effect. First, due to space constriction, Rh2 could have to at first remove Chol from the membrane, a mechanism reported for digitonin which extracts Chol and forms a complex with the latter 30 . However, three lines of evidence reported here do not support this possibility. Firstly, digitonin increased the size and induced membrane lysis only in Chol-containing liposomes, as previously observed by Sudji et al. 11 , whereas the action of Rh2 was instead delayed and depressed by the presence of membrane Chol. Secondly, Rh2 was able to modulate membrane biophysical properties even without Chol or eSM. Thirdly, based on GPex value of Laurdan, we observed that digitonin increased membrane fluidity whereas Rh2 decreased it, suggesting that Rh2 could preferentially stay into the membrane. To reinforce this idea, hydroxypropyl-β-cyclodextrin (HPβCD), also known to capture Chol from the membrane, has been shown to decrease the surface pressure (Π) of Chol monolayers 31 , in contrast to Rh2. Moreover, like digitonin, alpha-hederin increases membrane fluidity and membrane permeability due to the formation of saponin:Chol complexes. The differential effect between Rh2 and alpha-hederin could be related to the less profound insertion into the membrane hydrophobic core of Rh2 than alpha-hederin owing to two sugar chains, thereby increasing the surface of interaction with Chol and preventing polar interactions between Chol and phospholipids 32 . This highlights the importance of saponin nature since other saponins than Rh2, including chikusetsusaponin III or oleanolic acid glycosylated at both C-3 and C-28, induce Chol-independent membrane disruption 33,34 .
A second hypothesis for explaining the protective role of Chol is through its effect on membrane rigidity, decreasing the possibility for Rh2 to adsorb and interact with the membrane. In agreement with this hypothesis, eSM:ePC liposomes were more fluid than liposomes containing Chol and were also the most sensitive to Rh2. However, even if ePC:Chol and eSM:ePC:Chol liposomes were classified from the more fluid to the more rigid, ePC:Chol vesicles were less susceptible to the action of Rh2 as compared to eSM:ePC:Chol vesicles.
Third, the delayed/depressed effect of Chol could also be explained through its favorable interaction with SM in lipid domains and its similarity with Rh2 structure. Chol could compete with Rh2, thereby depressing Rh2:SM interaction and related effects. Such interaction between Rh2 and lipid clusters is in agreement with previous studies showing the disruption of lipid domains by Rh2 16,17 . Rh2 insertion could in turn increase membrane rigidity and line tension, resulting into the formation of positive membrane curvature to finally reduce the line tension 35 . Accordingly, using coarse-grained molecular dynamics simulations, Lin and Wang have shown that the interaction of dioscin saponin with membrane Chol leads to the destabilization of the L o -phase and the formation of positive membrane curvature which may eventually result in the hemolysis of red blood cells 36 .
The favorable interaction of Rh2 with eSM monolayer is reminiscent to the specific binding to membrane SM of pore-forming proteins actinoporins (e.g. equinatoxin and sticholysin) 37,38 . Hence, binding of the toxins and their permeabilization ability only occur in model systems exhibiting phase coexistence (L o and L d ) 39 . Based on our data related to GUV phase separation and LUV membrane permeability, we rule out a similar mode of action for Rh2. For instance, whatever the presence and the type of GUV phase boundaries, Rh2 induced alterations of GUV morphology and membrane permeability.
In conclusion, our work challenged the usual view that most saponins mediate their membrane-related effects through an interaction with Chol. Here the activity of Rh2 is enhanced by the interaction with membrane SM but depressed by Chol. Further studies are required to determine the structure-activity relationship between different saponins and eventually predict the mechanisms of their interactions with membrane. In silico methods. IMPALA and Hypermatrix methods have been previously described in details 40 .

Materials
Adsorption kinetics. The adsorption kinetics of Rh2 into lipid monolayer was measured by following changes in the surface pressure. Experiments were assayed in an automated Langmuir trough (KSV Mini-trough KSV instruments Ltd., Helsinki, Finland-width = 7.5 cm, length = 20 cm) with two hydrophophilic Delrin mobile barriers, a platinum Wilhelmy plate and a temperature probe. The system was enclosed in a Plexiglas box, and the temperature was maintained at 25 ± 1 °C. Lipid mixed solutions were dissolved at a concentration of 1 mM in CHCl 3 and were added at the air-water interface with a micro-syringe (Hamilton, USA) until surface pressure (∏) reached 30 ± 2 mN/m. After reaching the equilibrium, ginsenoside Rh2 was carefully injected into the 80 ml subphase (buffer 10 mM Tris-HCl, 159 mM NaCl pH 7.4) beneath the lipid monolayer. The subphase is stirred with magnetic stirrer using specialized injection supports. The variation of the surface pressure was recorded in function of time immediately after the injection of Rh2 until a plateau was observed.
preparation and analysis of Large Unilamellar Vesicles. LUVs composed of eSM:ePC (1:1), eSM:eP-C:Chol (1:1:1) or ePC:Chol (1:1) (molar ratio) were prepared by the extrusion from Multilamellar Lamellar Vesicles (MLVs) 41 . In brief, lipids were diluted CHCl 3 at the desired molar ratio. The organic solvent was then removed by evaporation in rotary evaporator (Buchi Rotavapor R-200, Switzerland). The residual lipid films after drying under vacuum overnight were hydrated with buffer solution (10 mM Tris-HCl, 135 mM NaCl pH 7.4). This suspension was submitted to five cycles of freezing/thawing. The latter were extruded through a polycarbonate Nuclepore Track Etch membrane filter (100 nm pore size filter, 19 times) using a mini-extruder system (Avanti Polar Lipids) to form LUVs. The quality control of LUVs was determined by the polydispersity index (PDI) (PDI < 0.2) and particle size distribution (z-average ± 150 nm) via dynamic light scattering (see below). Total phospholipids were determined by phosphorus assay and diluted to the desired final concentration in the buffer solution 42 .
Isothermal titration Calorimetry. Isothermal Titration Calorimetry techniques were assessed using a VP-ITC (MicroCal Northampton, MA, USA) to provide a complete thermodynamic description of the interaction between Rh2 and LUVs made with eSM, ePC and DOPC. All solutions were prepared with the same buffer constituted of 10 mM Tris-HCl, 159 mM NaCl pH 7.4. Prior each analysis, all solutions were degassed using sonicator bath. The cell (volume 1.4045 ml) for the ITC measurement was filled with Rh2 solution at a concentration of 40 μM and stirred at a speed of 305 rpm. The reference cell is filled with milliQ water Titration was carried out at 25 °C using 300 μl syringe filled with the LUV suspension at 2 mM and series of injections of 10 μl were made with a delay of 360 s between each succession injection to allow steady state to be attained. Data were processed by software ORIGIN 7 (Originlab, Northampton, MA, USA) using the cumulative model as described in Razafindralambo et al. 43   where I vv , and I vh are the fluorescence intensities with the excitation and emission polarization filters in vertical (v) and horizontal (h) orientations, respectively. The G factor is an inherent factor to the spectrophotometer used. The fluorescence polarization was measured at λ ex and λ em of 365 nm and 425 nm, respectively.
Membrane permeability of Large Unilamellar Vesicles. The leakage from LUVs of entrapped calcein at self-quenching concentration was followed to determine the effect of Rh2 on membrane permeability 45 . Briefly, the dried lipid film was hydrated with a solution of 10 mM Tris-HCl, 135 mM NaCl pH 7.4 containing purified calcein (73 mM). LUVs were then eluted through a Sephadex-G50 column to remove the non-encapsulated calcein. Fluorescence intensities were continuously recorded during 200 s at 25 °C after the addition of increasing concentrations of Rh2 in the liposome suspension. The λ ex and λ em were 472 and 512 nm, respectively. The percentage of calcein release from liposomes was calculated using the equation (3): t c trl t ot ctrl where F t represents the fluorescence intensity at the time t following the treatment with Rh2. F ctrl corresponds to the baseline fluorescence of liposomes at the same time and F tot , maximum fluorescence is the fluorescence measured after the addition of 0.1% Triton X-100 to induce maximum release.

Membrane fusion of Large Unilamellar Vesicles.
Lipid mixing assay is based on self-quenching of R 18 (octadecyl rhodamine B chloride) in LUVs 46 . The R 18 hydrophic probe was incorporated at 5.7% mol into a lipid mixture before its evaporation. R 18 labeled liposomes were mixed at a ratio of 1:4 with unlabeled liposomes adjusted to the same concentration. The changes of fluorescence intensity were measured during 200 s at 25 °C, using λ ex and λ em at 560 nm and 590 nm, respectively. The percentage of liposomal fusion was calculated using the equation (3).
preparation and analysis of Giant Unilamellar Vesicles. GUVs were prepared using the electroformation method 47 . To investigate phase separation, mixtures of eSM:ePC (1:1), eSM:ePC:Chol (1:1:1) or ePC:Chol (1:1) combined with 0.2% mol Rhod-PE and 0.5% mol NBD-PE were prepared. A small volume of lipid solution was spread and dried inside the chamber of 1 mm thick silicon gasket placed on the surface of an ITO coated glass lamella. The electroformation chamber was filled with 0.2 M sucrose and overlaid with a second ITO-coated slide. The GUVs were formed by exposure to a sinusoidal alternating current of 10 Hz and 1 V for 2 h at 60 °C. The volume containing GUVs were collected from the slide and lipid concentration was determined based on a calibration curve of NBD-PE fluorescence intensity. GUVs were diluted in 0.2 M glucose and placed in a μ-Slide 8-well chamber from IBIDI (Martinsried, Germany) and observed in the red channel for Rhod-PE (λ ex , 561 nm and λ em , 617 nm) and in the green channel for NDB-PE (λ ex , 488 nm and λ em , 530 nm) at room temperature. To investigative the permeabilizing effect of Rh2, same GUV compositions combined with 0.2% mol Rhod-PE were diluted in 0.2 M glucose supplemented with 20 μM calcein and the difference of fluorescence intensity between outside and inside vesicle was measured. GUVs were visualized either with a Zeiss wide-field fluorescence microscope (Observer.Z1) using a plan-Apochromat 40x/1.4 oil Ph3 objective or a Axio Observer 1 inverted microscope equipped with a model CSU-X1 Yokogawa spinning disk and Plan-Apochromat 40x/1.40 oil DIC. Images were analyzed with the Carl Zeiss AxioVision 4.8.2 software.
Fluorescence spectroscopy measurement. All fluorescence measurements were carried out with a LS55 luminescence spectrometer from Perkin Elmer (Waltham, MA, USA). The spectrophotometer was equipped with a temperature-controlled sample holder being set at 25 °C.