Glucosomes: Glycosylated Vesicle‐in‐Vesicle Aggregates in Water from pH‐Responsive Microbial Glycolipid

Abstract Vesicle‐in‐vesicle self‐assembled containers, or vesosomes, are promising alternatives to liposomes because of their possible hierarchical encapsulation and high stability. We report herein the first example of sugar‐based vesicles‐in‐vesicles, which we baptize glucosomes. These were prepared by using a natural microbial glycolipid (branched C22 sophorolipid) extracted from the culture medium of the yeast Pseudohyphozyma bogoriensis. Glucosomes spontaneously formed in water between pH 6 and pH 4 at room temperature, without the requirement of any additive. By means of pH‐resolved in situ small angle X‐ray scattering, we provided direct evidence for the vesicle‐formation mechanism. Statistical treatment of the vesicle radii distribution measured by cryo‐tansmission electron microscopy by using a derived form of the Helfrich bending free‐energy expression provided an order of magnitude for the effective bending constant (the sum of the curvature and the saddle‐splay moduli) of the lipid membrane to K=(0.4±0.1) k B T. This value is in agreement with the bending constant measured for hydrocarbon‐based vesicles membranes.


Introduction
Liposomes are well-knowns upramolecular systems classically obtained from phospholipid mixtures. Sized between 20 nm and severalm icrons and obtained throughas eries of wellknown methods (e.g. thin-film hydration, solventi njection, and reverse-phase evaporation), they are largely studied for their ability to load ag iven cargo,m olecule, or nanoparticle for therapeutic purposes and for their interesti na rtificial cell design and analytical science, just to cite some. [1,2] Unfortunately,c lassical liposomes have one major drawback that concerns their stability throughout the delivery process and leach-ing of the cargo due to destabilization or degradation of the bilayer membrane by the action of enzymes, such as phospholipase A2. [3] To overcome such problems and to make more stable liposomes, severaln ew technologies have been studied, and these include, among others, the development of vesosomes( liposomes-in-liposomeso rm ulticompartmentl iposomes), [1,[3][4][5] polymersomes (polymer-based capsules), [6] capsosomes (liposome-containingp olyelectrolyte capsules), [7] proteinosomes (multicompartment protein-polymer conjugates), [8] and av ariety of stimuli-responsive vesicles [1] based on lipids, polymers, and their mixtures.
Vesosomes are an interesting class of vesicles used in drug delivery, [9] and they refer to am ulticompartmentl iposomeo btained by using an umber of encapsulations trategies. In the first one, spiral folded cochleate cylinders composed of negatively charged dioleyl phospholipidsi nt he presence of calcium ions are mixed with preformed liposomes;u pon removal of Ca 2 + by using ethylenediaminetetracetic acid (EDTA), the cylinders unroll and encapsulate the smallerl iposomes. [4,10] In the second one, [11] interdigitated lipid bilayers( ILBs) are formed by adding3m ethanolt oas mall unilamellar vesicle solution composed of dipalmitoylphosphatidylcholine (DPPC) below their meltingt emperature (T m ). The ILBs are then used as precursor membranes to encapsulate ap reformed liposomal preparation after heating the solution above T m .I nt he third, [12] multistep approach, cholesterol is used as ab inder to "glue" liposomes together and is eventually encapsulated by ah ydrated1 ,2-dimyristoyl-sn-glycero-3-phosphocholine (DMPC)c holesterol mixture. In the fourth method, [13] externalc ompounds such as salts or carbohydrates are used to induce membrane fluctuations in giant vesicles;t his results in wobbling of ap art of the membrane inward and eventually leads to detachmentw ithin Vesicle-in-vesicle self-assembled containers, or vesosomes,a re promisinga lternatives to liposomes because of their possible hierarchical encapsulation and high stability.W er eport herein the first example of sugar-based vesicles-in-vesicles, which we baptizeg lucosomes.T hese were prepared by using an atural microbial glycolipid (branched C22 sophorolipid) extracted from the culture mediumo ft he yeast Pseudohyphozyma bogoriensis. Glucosomess pontaneously formed in water between pH 6a nd pH 4a tr oom temperature, without the requirement of any additive. By means of pH-resolved in situ small angle X-ray scattering, we provided direct evidence for the vesicle-formation mechanism. Statistical treatment of the vesicle radii distributionm easured by cryo-tansmission electron microscopy by using ad erived form of the Helfrich bending free-energy expression provided an order of magnitude for the effective bending constant (the sum of the curvature and the saddlesplay moduli)o ft he lipid membrane to K = (0.4 AE 0.1) k B T. This value is in agreement with the bending constant measured for hydrocarbon-based vesiclesm embranes. the vesicle itselfa fter enclosing externalw ater and forminga n inner secondary vesicle. In the fifth method, [14,15] engineered phospholipid-based liposomes with complementary surfaceactive groups are used to strengthen the interliposomal interaction upon fusion and the possible formation of vesosomes. More recently,m icrofluidics were used to prepare vesosomes composed of an asolectin/cholesterol mixture. [16] Althoughv esosomesh ave broad application potentiala nd despite the fact that various preformed liposome formulations can be encapsulated, the process is still not easy to handle and has severald rawbacks. The mechanism of formationo f the externalm embranei sl imited to af ew phospholipidt ypes generally containing saturated lipid chains, such as DPPC,b ecause of their ability to form interdigitated bilayers. In most cases, it is even necessary to use additives, such as EDTA, ethanol, or cholesterol, or to heat above the T m of the lipid (typical T m values are highert han 45 8C, av alue which can be too high for some applications). Furthermore, the compositions of the externalm embrane and the internal lipids are generally not the same, af act that increases the complexity of the overall formulation, and preformed liposomes, obtained by traditional methods, [2] must generally be prepared beforev esosomef ormation. Last, but not least, all technologiesy ield vesosomes with ab road size distribution varyingf rom several hundred of nanometers to severalm icrons, whereas small vesosomes below 200 nm in size, interesting for delivery applications, can only be obtained after extrusion.
In this work, we report fort he first time the formation of glycosylated vesosomes, which we call glucosomes, obtained in water at room temperature andn ear neutralp H( pH 6), by using ap H-sensitive microbialg lycolipid obtained from the yeast Pseudohyphozyma bogoriensis. Stimuli-responsive vesicles are of large scientific and medical interestd ue to the possibility to use one, or ac ombination of, external stimuli (temperature, pH, magnetic fields, redoxp otential, light) to trigger the releaseo facargo in biomedical applications. [1,17] Although the literaturei sr ich in specifically pH-responsive phospholipid and fatty-acid vesicles, [18][19][20][21][22] this work illustrates the first example of as timuli-triggered vesicle-in-vesicle system. P. bogoriensis produces ab ranchedb olaform lipid composed of a b-d-glucose b(1,2) carbohydrate (sophorose) head group covalently linked at the C13 position of behenic acid (C22:0) throughaglycosidic bond and af ree COOH group at the C1 positionofbehenic acid itself ( Figure 1). Reports on thesecompoundsd ate back to 1968 when this novel type of sophorolipid was isolated and structurally characterized. [23] In contrastt o the well-known sophorolipids produced by yeasts of the Starmerella clade, these compounds display ab ranched structure insteado fl inear structure and do not occur in the lactonic form (i.e. intra-esterification of the free COOH to the sugar unit). Despite their intriguing structure and hereto-related properties, applicationo fP. bogoriensis sophorolipids is largely neglected, mainly due to the low production titers of 0.5 to 1.0 gL À1 compared to the ones mentioned above for which yields of several hundreds of grams per liter can be obtained. [24] Herein, we report for the first time on the self-assem-bly behavior of branched sophorolipidsa nd discuss their potential in biomedical applications.
Using ac ombination of cryo-transmission electron microscopy (TEM) and pH-resolved in situ small-angle X-ray scattering (SAXS),w es how that this compound behavesa satypical pHsensitivem icrobial glycolipid and is ablet of ormg lycosylated vesiclesb ys imply changing the pH in the 6 < pH < 4r ange at room temperature. However,d ifferentf rom other related microbial glycolipids [25,26] that form single-or multiwall vesicles, we observe the formation of glucosomes. If glucosomes and vesosomes are structurally similar,o ne shouldn ote the following:1 )none of the classical compounds employed for vesosomes( e.g. phospholipids, cholesterol) are used, and glucosomesa re only composed of b-d-glucose b(1,2) (sophorose)containing lipids;2)the compositions of the externaland internal membrane vesicles is expected to be the same;3 )wep roduce av esicle-in-vesicle system by using as imple pH-change approachi nstead of any of the known complex methods described in the literature. The glycolipidss hown here certainly contributet oe nlarging the wide complexity of glycolipid selfassembly behaviori nw ater; [27][28][29][30][31][32] in addition, given the importance of carbohydrates in glycobiology and medicine, [33] this new class of stimuli-responsive multicompartment glycosylated vesicles could pave the way to the development of complex delivery systems.

Materials and Methods
Synthesis of Sophorolipid aSL-C22:0 13 The chemical structure of the microbial glycolipid used in this work is shown in Figure 1. The structure represents the sophorose derivative of 13-hydroxydocosanoic acid (C22 lipid moiety corresponding to behenic acid), in which the glycosidic bond occurs at the C13 position of the fatty acid. We will refer to this compound as aSL-C22:0 13 ,f or which as tands for acidic, SL stands for sophorolipid, C22 indicates the length of the fatty acid, 0i st he number of unsaturations, and 13 is the position of the glycosidic bond. This compound was obtained through microbial production by the yeast Rhodotorula bogoriensis MUCL 11 796 recently renamed Pseudohyphozyma bogoriensis according to ap ublished procedure. [23] The yeast was grown at 25 8Ci nm edium containing 50 gL À1 glucose, 4gL À1 yeast extract, 0.2 gL À1 MgSO 4 ·7H 2 O, and 1gL À1 KH 2 PO 4 .I nocula in shake flasks were agitated at 200 rpm, and 3-5% inoculation volume was used for production in a1 50 Lf ermenter (Sartorius). The pH was controlled at 3.5 by NaOH addition, and the stirring and airflow rates were set at 500-800 rpm and 0.5-1 vvm, respectively,t ok eep the partial pressure of oxygen (pO 2 )a t4 0% during the entire stationary phase. Discontinuous feeding of glucose was applied to keep the concentration above 30 gL À1 ,a nd 20 gL À1 rapeseed oil was added upon inoculation. The total cultivation time was 133 h. Cells and remaining oil were removed by microfiltration, and SL was extracted with EtOAc (2 ). The SL precipitated upon partial evaporation of EtOAc and was recovered by filtration. The sophorolipids obtained by microbial fermentation contained two, one, or no acetylation(s) at the C6 carbon atom of one or the other glucose moiety.T os tandardize the SL mixture, the acetyl groups were removed by alkaline hydrolysis:the SLs were dispersed in reverse osmosis water,and the solution was brought to pH 12.5 with sodium hydroxide. The pH was monitored and kept constant during hydrolysis. The reaction was followed up by using HPLC with an evaporative light-scattering detector (ELSD) ( Figure S1 in the Supporting Information) and was stopped by acidification (pH 4.5). The final aSL-C22:0 13 product was extracted with ethyl acetate and was subsequently freeze dried.

Sample Preparation
The aSL-C22:0 13 glycolipid was dissolved in Milli-Q-grade water at room temperature to give ac oncentration of 5mgmL À1 .T he pH was increased up to about 11 by adding 5 m NaOH (10-15 mL), and it was eventually decreased by adding microliter amounts of 0.1-1 m HCl until the solution became turbid in the vicinity of pH 6. This procedure was the same as that used for the pH-dependent study of related sophorolipids [34] and glycolipids, [25] which thus makes the results of this work comparable to those of other microbial glycolipids of as imilar chemical nature. This process was adapted for analysis in situ by means of small-angle X-ray scattering (SAXS), whereas aliquots were extracted at ag iven pH and were analyzed by using cryogenic transmission electron microscopy.

TransmissionElectronMicroscopy Experiments under Cryogenic Conditions (cryo-TEM)
These experiments were performed with aF EI Te cnai 120 twin microscope operating at 120 kV and equipped with aG atan Orius CCD numeric camera. The sample holder was aG atan Cryoholder (Gatan 626DH, Gatan). DigitalMicrograph software was used for image acquisition. Cryofixation was done with ah omemade cryofixation device. The solutions were deposited on aglow-discharged holey carbon coated TEM copper grid (Quantifoil R2/2, Germany). Excess solution was removed, and the grid was immediately plunged into liquid ethane at À180 8Cb efore it was transferred into liquid nitrogen. All grids were kept at liquid-nitrogen temperature throughout all experiments. Images were handled by using ImageJ software. [35] In Situ Small-Angle X-ray Scattering (SAXS) Data were acquired with the high-brilliance ID02 beamline (E = 12.46 keV,s ample-to-detector distance = 1m)a tt he ESRF synchrotron (Grenoble, France). In situ experiments employed af lowthrough polycarbonate 2mmc apillary connected to the samplecontaining solution at pH 11.6 through ap eristaltic pump. The pH was controlled in situ by using ac lassical KCl pH meter directly in the experimental hutch, which constantly monitored the pH. The pH changes were obtained by using a0 .1 m HCl solution introduced by am otor-controlled press syringe. Error bars on the experiments were calculated on the basis of the estimated number of photons detected (accounting for the gain and quantum efficiency of the CCD and phosphor layer), assuming Poisson statistics. The noise of the detector was accounted for by comparison with dark current. Data were acquired by using aC CD camera and were integrated azimuthally to obtain at ypical I(q)( I is the intensity as af unction of the magnitude q of the scattering vector) spectrum. Contribution of the solvent (water at pH 11.6) and capillary were measured prior to the experiment and were duly subtracted during the data treatment. Data were corrected for the transmission of the direct beam and were scaled to be in absolute scale. The q-range calibration was made by using as ilver behenate standard sample (dref = 58.38 ).

Fit of SAXS Data:MicelleM odel
The SAXS analysis of the data presented here was performed according to Ref. [26].Acore-shell ellipsoid of revolution form factor model was chosen to fit the data in the region above q > 1nm À1 and in the basic pH region, whereas ac ore-shell bicelle form factor was employed to describe the lipid bilayer membrane. All models used were developed in the SasView 3.0.0 software (Core-ShellEllipsoidXT and CoreShellBicelle). [36] The general equation of the scattering intensity [I(q)] is given by Equation (1): in which scale is the volume fraction, V is the volume of the scatterer, 1 is the scattering length density (SLD) and is equivalent to the electron density of the object, 1 solv is the SLD of the solvent, P(q)i st he form factor of the object, bkg is ac onstant accounting for the background level, and S(q)i st he structure factor.F or the purpose of the present work, we assumed aunitary value of S(q)i n the analyzed range of q values (q > % 1nm À1 ). The analytical expressions of P(q)f or ac ore-shell ellipsoid of revolution and Core-ShellBicelle models are provided in Ref.
[37].I ns ummary,t he ellipsoid model, employed in the 12 < pH < 6r egion, is characterized by the equatorial core radius and shell thickness, the core and shell aspect ratios, and the SLDs of the core, shell, and solvent. In the fitting process, the volume fraction (0.5 wt %), core SLD (7.9 10 À4 nm À2 ), and solvent SLD (9.4 10 À4 nm À2 )a re fixed. The procedure to estimate the core and solvent SLD was described elsewhere. [38] The core SLD was estimated on ad ry basis of behenic acid. If water penetration could occur, [38] we assumed this to be negligible to keep the number of independent parameters as low as possible. To control the fit, we also used as eries of additional assumptions:t he shell SLD was between 10.0 and 12.0 10 À4 nm À2 , which are reasonable values for hydrated sophorose; [38] the equatorial shell thickness was kept below 1.5 nm to be consistent with the values found in previous studies on analogous sophorose-con-  [25,26,38] The core-shell bicelle model was used to analyze the 6 < pH < 4 region. To adapt this model to the analysis of the vesicles bilayers, we used large values of the bicelle radius (fixed), R = 100 nm, and we fixed the rim size to zero, which thus made it de facto al ipid core-shell bilayer model. As before, the volume fraction (0.5 wt %), core SLD (7.9 10 À4 nm À2 ), and solvent SLD (9.4 10 À4 nm À2 )w ere fixed. The face thickness, the face SLD, and the length of the bilayer core were optimized on the basis of previous studies on similar compounds. [25,26] The fit was controlled by assuming, just as above, that the shell SLD was contained in the range of hydrated sophorose, whereas the shell thickness was also below about 1.5 nm. Finally,t he evolution of the micelle-to-vesicle ratio with pH was evaluated by al inear combination of the two model functions [Eq. (1)] over the entire pH range to fit the data, as shown in Equation (2): in which x,b etween 0a nd 1, identifies the relative proportion of morphology 1, described by the model function I(q) 1 ,and morphology 2, described by the model function I(q) 2 .T he general expression for I(q) 1 and I(q) 2 is given in Equation (1), and all assumptions made above are valid. The main difference lies in the expression for the form factor, P(q): P(q) 1 identifies the core-shell ellipsoid model, whereas P(q) 2 identifies the core-shell bilayer model. Given that the use of Equation (2) involves the fitting process over al arge number of independent variables ( % 20), ap rocess introducing many artifacts, we only used it to evaluate x,w hich corresponds to the micelle-to-vesicle ratio (x!1s tands for micelles only and x!0s tands for bilayers only). All other parameters were kept fixed at their optimum values for both micelle and bilayer,a sg iven above.

LC-MSAnalysis
Analysis of aSL-C22:0 13 was done with aT hermo LCQ Deca by the use of RP18 solid phase (150 2mm, Phenomenex Luna) and solvents 1) methanol with 0.1 %f ormic acid and 2) water with 0.1 % formic acid as gradients. LC-MS analysis was done with aShimadzu LC-10-AD HPLC system (Shimadzu Europe GmbH, Germany) connected to aq uadrupole mass spectrometer (Waters, Milford, MA). Molecules were identified by their native molecular masses after ESI (electrospray ionization) without collision. The results were consistent with previous reports. [23,39] 2. Results Figure 2s hows the entires et of in situ SAXS data recorded for aSL-C22:0 13 between pH 12 and 4. At basic pH, two signals can be identified:b elow q = 0.2 nm À1 the strong scattering intensity indicates the presenceo fl arge objects, whereas above q = 0.2 nm À1 the typical signature of micelles can be identified. Upon lowering the pH, at ransition occurs at about pH 6, at which as trong increase in the scattered intensity starts at q % 0.9 nm À1 .F inally,t wo diffraction peaks at q = 2.11a nd 4.25 nm À1 ,w hich are indicative of the presence of al ongrange orderedl amellar or onion phase, appear below pH 4. In the 6 < pH < 4r ange, the SAXS signal is characteristico falipid bilayer given strong similarities with previousS AXS data collected on microbial glucolipids. [25,26] The nature of the bilayer was investigated by using cryo-TEMa sacomplementary technique.C ryo-TEM analysis of aSL-C22:0 13 in water was recorded in the vicinity of pH 5t oo bserve the nature of the bilayer structures. Figure 3s hows the massive presence of vesicles throughout the sample holder, and the estimated membrane cross section is (4.0 AE 0.5) nm. The SAXS datac ould then be fitted by using ac ore-shell ellipsoid form factor in the micellar regiona bove pH 6a nd ab ilayer form factor in the 4 < pH < 6 range until precipitation and formation of al amellar phase below pH 4; the results from the fit are shown in Figure 2. The micelle-to-vesicle ratio throughout the pH jump shows atransition pH at about 6.1 AE1.0. The typical morphological character- istics of the micelles showapractically constant hydrophobic core radius and hydrophilic shell thickness throughout their stabilityr ange;t he radius is between 0.90 and0 .95 nm, and the thickness is between 1.00 and 1.05 nm. The hydrophobic core radius is comparable in size to that classically estimated for "classical" acidic sophorolipids and glycolipids; [26] the shell size is comparable to the size of sophorose and to the experimentally fitted value obtained for acidic C18:1 sophorolipid micelles below pH 6( % 1.2 nm) but is twice as large as the core radius of acidic C18:1 sophorolipidm icelles above pH 7 ( % 0.5 nm). [26] The morphological features of thea SL-C22:0 13 micelles seem to be in line with that classically found for microbial glycolipids, whereas possible differences in the hydration layer can explain minor variations (fractionso fananometer) in the hydrophilic shell thickness. After the micelle-to-vesicle transition pH, the core-shell bilayer model indicates af ace thickness of 0.90 nm and al ength of 0.85 nm at about pH 5. According to these data, the total vesicle cross section (2 thickness + length) is estimated to be 2.65 nm, av alue that is comparable to that obtained by cryo-TEM, despite a3 0% discrepancy.A tt he moment, it is unclear whether the bilayer cross-section is overestimated by cryo-TEM or underestimated by the fitting process of the SAXS data, as both techniques are subject to experimental error.I na ll cases, both techniques indicate that the bilayer size is constituted by no more than two aSL-C22:0 13 molecules that are most likely interdigitated if one estimates the size of aSL-C22:0 13 to be less than 3nmb yu sing the classical Ta nford formula and the length of the hydrocarbon region to be only half of aC 22 chain. Figure 2s hows as cheme of the morphological evolu-tion of the self-assembleds tructures formed by aSL-C22:0 13 against pH.
Finally,t he SLD of the shell, which is equivalent to the electron density in the hydrophilic shell region composed of sophorose and water,c an be estimatedf rom the SAXS data in the micellar stabilityd omain and its value is constant (11.0 10 À4 nm À2 ), whereas it slightly increases if vesicles are formed, which is probably indicativeo fadehydration process, as already observed for the formation of liposomes from glucolipids. [26] Analysis of the SAXS data reveals that, as found for saturatedm icrobial glycolipids, the micelles are stable objects until pH 6, as their size and SLD stay practically unchanged.A lthoughi tw as found that the mechanism of formation of vesicles from pH-responsive glucolipids followed ac lassical micelle-cylinder-disk-vesicle mechanism, described for other lipid systems, [40] it seems that aSL-C22:0 13 -based vesicles are formed through ad ifferent route that does not include the micelle-to-disk transition. In fact, although giant micellesa nd large disks could be clearly detected for the glucolipids both by cryo-TEMa nd SAXS, no hint of their stable presence could be observed in the aSL-C22:0 13 system. Considering the sharp, as opposed to smooth, transitions in terms of size and SLD observed between the micellar and vesicle regions, it is likely that the micelles act as reservoirs of matter for the formation of vesicles.S imilar arguments were used to describe the micellesto-bilayer formation in microbial glucolipids. [26] These assumptions are corroborated by additional cryo-TEM data recorded at pH 9.4 (Figure 4), in the micellar regime, and at pH 5.9, at the micelle-to-vesicle transition. The images show the presence of largem icellar aggregates,a si ndicated by arrows labeled with 1i nF igure 4a,b.L arge (arrowsl abeled with 2, Figure4a, b) and small (arrows labeled with 3, Figure 4a,b)v esicles are also observed at basic pH in close proximity of the micellar cloud. Very similar images are obtained at pH 5.9, and in no case could micellar fusion into cylinders and disks be observed, as seen for other microbialg lucolipids. [26] Furthermore, the coexistence of (large amounts of) micelles and (few) vesicles at basic pH is neither as urprise nor uncommon. In fact, the corresponding SAXS data have as trong low-q scattering signal, which indicates the presence of spurious  amountso fl arge objects, af act that is compatible with the vesicles observed by cryo-TEM. Similar large objects, including rolled sheets and vesicles,w ere described for glycolipidsu nder basic conditions. [26] Interestingly,i nt he transition pH region, cryo-TEMs hows data that is very similar to those presented in Figure 4, that is, an increasing amount of micelles coexisting with micellar aggregates.N of ilamentous micelles or disks could be observed, which thus corroborates the hypothesis of the reservoir model.
Ac loser look at Figure 3t aken on the aSL-C22:0 13 system at pH 4.9 shows the presence of large vesicles of several hundred nanometers with the embedding of smaller single and multiwalled vesicles ranging from about 25 to 200 nm. Similar systems, called vesosomes,w ere first reported by the group of Zasadzinski. [10,11] By analogy,w ea ddress the system presented here as glucosomes, referring to as ystem of vesicles-in-vesicles only composed of b-d-glucose b(1,2) (sophorose)-containing lipids. Sophorolipids are ac lass of microbial glycolipids that commonlyr efers to C 18 compounds in their lactonic and acidic forms. It was reported earlier that the sodium saltso fa cidic sophorolipids formed vesicles in water, [41] but thesed ata could not be verified, as micelles and nanoscale platelets were reported for the same compound. [42] Dhasaiyan et al. reported the formation of vesicles by using linolenic acid sophorose lipids, [43] but scanning electron microscopy (SEM) wase mployed as the analytical technique,a nd it was not reported whether the vesicles were single-walled or if they contained other vesicles.H ere, we give direct proof that as ophorosecontaining lipid massively forms vesicles, as corroborated by SAXS, and that the vesicles form am ulticompartments ystem. Nonetheless, despite the structurala nalogy,g lucosomes are phospholipid-free and are obtained in as ingle step in water at neutralp H, af act that makes ab ig difference with classical vesosomes. [1,10,11] The latter are generally prepared by embedding ap reformed liposomal solution in micron-scalegiant vesicles, either obtainedf rom cochleate cylinders in the presence of Ca 2 + and EDTA [10] or from heating lamellar bilayers above the T m of the lipid in the presence of ethanol. [11] The use of cholesterol as av esicle binder [12] and the employment of functional vesicles [14,15] were also reported.
The glucosomesd escribed in this work strongly differ from these systemsi nt erms of both the compositiona nd formation mechanism. According to the cryo-TEM and in situ SAXS data, af ew vesicles coexist with al argem ajority of micelles in the basic-neutral pH region ( Figure 4);h owever,p H-resolved in situ SAXS shows an abruptm icelle-to-bilayer transition, which suggestsareservoir-to-vesicle-to-lamellar [26] model rather than am icelle-to-cylinder-to-disks-to-vesicle-to-lamellar mechanism, as reportedf or other glucolipids( Figure 6). [25,26] Althoughs till unclear,u ponv esicle formation below pH 6, the glucosome formation mechanismp robably goes through membrane wobbling episodes due to spontaneousf luctuations of the surfaceb ilayer,a sr eported foro ther systems. [13] The cryo-TEM images in our possession seem to confirm this. The arrows in Figure 5a-e show points at specific wide undulations of the bilayer having ap eriod in the order of tens, or even hundreds, of nanometers, and they can be relatedt o invagination (negative curvature) ande xvagination( positive curvature) processes occurring at the membrane surfacea nd are probably at the origin of the glucosome phenomenon. One must also note the enhanced roughness at the vesicle surface (Figure 3b,d and Figure 5a-e) compared with the smooth vesicle bilayer observed at basic pH ( Figure 4, arrow 2) and where the oscillation period is rather in the order of nanometers. These observations indicate strongf lexibility of the membrane at room temperature, af act that is also reflected by the large polydispersity in terms of vesicle radius. According to the meltingt emperature of aSL-C22:0 13 , T m = 58-59 8C, [23] at room temperature one expects solid-like behavior of this compound within the membrane, which should then then be stiffer than that observed above. However,t he comparison between the actual T m and the membrane properties should only be indicative of this class of molecules, as demonstrated by the self-assembly behavior of acidic subterminal C18:1cis sophorolipids, which can form stable crystalline ribbonsa tr oom temperature in their pure form, [44] even thought he T m of C18:1cis (oleic acid) is well-below room temperature. In contrast, the membranesf ormed by C18:1cis glucolipids are more flexible than those formed by the corresponding C18:0 glucolipids. [25] The qualitative observations concerning the strong flexibility of the aSL-C22:0 13 membrane surfacei sdiscussed below. Figure 5f reports the radius-size distribution of the glucosome population ( % 350 individual vesicles measured), which varies from af ew nanometers up to 300 nm. Even though the histogrami sl imited to 100 nm for readability purposes, about 5% of the populationh as radii in the 100 to 300 nm range. The data can be fitted by using al ognormal distribution (dotted curve)h aving am ean of m = 12.2 AE 0.6 nm and as tandard deviationo fs = 0.4 AE 0.1 nm. Interestingly,i ti sp ossible to exploit the radius-size distribution obtained by cryo-TEMe xperiments to determinet he effective bending constant of the membrane, K, [45,46] which is defined as [Eq. (3)]: in which k and " k represent the curvature and the saddle-splay moduli, respectively,a nd they identify the bending elasticity of the bilayer: k is relatedtothe size and " k is related to the topology.T he molar or number fraction, C N ,o fv esicles of aggregation number N and general radius R is related to K through [Eq. (4)]: in which k B is the Boltzmannc onstant, T is the temperature, and C M is the molar or number fraction of vesicles of aggregation number M and radius r 0 ,defined as follows [Eq. (5)]: and R 0 is defined as the minimum energy,s pontaneous, radius of the vesicles. Equation (4) is derived from the Helfrich expression of the bending free energy [47] by using several approximations, including the assumption that the vesicles are in am etastable state, and if not at equilibrium, they form spontaneously;f urthermore, ideal mixing of the vesicles should be verified. The Helfrich expression itself is based on the so-called harmonic approximation, for which the bilayer thickness (3-4 nm) and the Debye length of ionic surfactants are small relative to the principal radii of curvature of the membrane, R 1 and R 2 (> 30 nm). [46] If k and " k cannotb ed etermined directly,t he value of K is still useful to characterize the stiffness of am embrane relative to that of other systems: for K % k B T,t he membrane is plastic and thermalu ndulations generate ab road size distribution, whereas for K @ k B T,t he membrane is more rigidt han the thermal fluctuationsa nd the size distribution is sharper.E quation (4) can be directly applied to fit the size distribution of the glucosome aSL-C22:0 13 system in Figure 5f (continuous curve), and one finds r 0 = (12.9 AE 2.0) nm and K = (0.4 AE 0.1) k B T. The value of K below k B T is in agreement with that expected by ab road size distribution of vesicles, and it can be comparedt o soft membranes composed of hydrocarbon-based amphiphiles, such as am ixture of cetyltrimethylammoniumb romide (CTAB) and sodium octyl sulfonate( SOS) (0.2 < K < 0.7 k B T according to the weightr atio) or cetyltrimethylammonium tosylate (CTAT) and sodium dodecylbenzene sulfonate (SDBS). [40,41] On the contrary,s tiffer membranes containing am ixture of hydrocarbons and perfluorinated surfactants can achieve K values in the order of 10 k B T. Despite the nice agreement between our data and the literature, one mustb ec autious to consider the absolute value of K = (0.4 AE 0.1) k B T as granted. Upon repeating the cryo-TEMe xperiment,w es ystematically found the same type of vesicle-in-vesicle system,afact that gives credit to the data. However, our system is pH dependent, and even if we perform the cryofixation severalh ours after fixing the pH, the system may still not be at true equilibrium. The membrane cross-section (2 thickness + length) according to the SAXS data presentedi nF igure 2c is about 2.7 nm, av alue that is fivet imes lower than the mean value for r 0 .F rom Figure 5, it is clear that Equation ( 4) does not properly describe the radius distribution, which is better described by the log-normal distribution. For this reason,t he actual value of K may be even smaller,a nd one should probably take the value of K = (0.4 AE 0.1) k B T as the upper limit. Finally,t he system under study forms vesicles under acidic pH conditions, and consequently,t he sophorolipids are close to neutral, and in this sense, the Debye length condition should be fulfilled. Figure 6summarizes the process of glucosome pH formation for the aSL-C22:0 13 -branched sophorolipid molecule. We must highlight the fact that stimuli-responsive vesosomes have so far received little, if no, attention at all. [13]

Conclusions
In this work, we provided evidencet hat ab ranched form of C 22 sophorolipids, aSL-C22:0 13 ,p roduced by Pseudohyphozyma bogoriensis formed av esicle-in-vesicle colloidal dispersion, which we called glucosomes by analogy with vesosomes,a lthought hey were entirely composed of ag lycosylatedc ompound.G lucosomes were prepared in water in the 6 < pH < 4 range at room temperature without the need to add external compounds, as was classically done for the preparation of phospholipid-based vesosomes. We employed pH-resolved in situ small-angle X-ray scattering (SAXS) by using synchrotron radiation to follow the self-assembly mechanism of aSL-C22:0 13 from basic to acidic pH. The starting system at pH % 12 was mainlyc omposed of micelles coexisting with as mall percent- Figure 6. Schemeo ft he pH-driven formation of glucosomesb yu sing aSL-C22:0 13 branched sophorolipid. Accordingt ot he in situ SAXS data,t he system is composed of amajority of micellesand am inorityo fvesicles from basict on eutralpH. In the transition pH region,a ta round pH 6, micelles seem to act as reservoirs of matter for glucosome formation.Belowp H4, al amellar phase forms.
ChemistryOpen 2017, 6,526 -533 www.chemistryopen.org 2017 The Authors. Publishedb yWiley-VCH Verlag GmbH &Co. KGaA, Weinheim age (< 5%)o fl arge vesicles. This mixture was very stable down to pH % 6, below which glucosomes were formed. The SAXS data were confirmed by cryo-transmission electron microscopy (TEM) data recorded at three pH values. Glucosomes were supposed to form from surface fluctuations of the glycolipids bilayer,w hich we believed to be interdigitated. Theb ilayer thickness was estimated to be between (4.0 AE 0.5) nm, according to cryo-TEM measurements, and 2.65 nm, according to modeling the SAXS data at about pH 5; moreover,t he length of aa SL-C22:0 13 molecule (between COOH andC 13, at which branching takes place, and including sophorose) was estimated to be below 3nm. Employmento fafree-energy-minimizing mass-action model in conjunction with the Helfrichb ending free-energy expression allowed estimation of the order of magnitude of the effective bending constant (which is the sum of the curvature and the saddle-splay moduli)o ft he lipid membrane to K = (0.4 AE 0.1) k B T. This value well describes the broad radii distribution, and it is coherent with similar hydrocarbonbased vesicle membranes found in the literature (0.1 < K < 1 k B T), but its absolute value should be taken as an upper limit for the aSL-C22:0 13 -branched sophorolipid glucosome system.