Rapid confocal imaging of vesicle-to-sponge phase droplet transition in dilute dispersions of the C10E3 surfactant

The lamellar-to-sponge phase transition of fluorescently labelled large unilamellar vesicles (LUVs) of the non-ionic surfactant triethylene glycol mono n-decyl ether (C10E3) was investigated in situ by confocal laser scanning microscopy (CLSM). Stable dispersions of micrometer-sized C10E3 LUVs were prepared at 20 °C and quickly heated at different temperatures close to the lamellar-to-sponge phase transition temperature. Phase transition of the strongly fluctuating individual vesicles into micrometre-sized sponge phase droplets was observed to occur via manyfold multilamellar morphologies with increasing membrane confinement through inter- and intra- lamellar fusion. The very low bending rigidity and lateral tension of the C10E3 bilayer were supported by quantitative image analysis of a stable fluctuating membrane using both flicker noise spectroscopy and spatial autocorrelation function.

. The low surfactant concentration part of the phase diagram of C 10 E 3 -water is shown in Fig. 1. Around H 0 ≈ 0, bilayer structures are preferred and the binary phase diagram shows a lamellar phase (L α ) and a sponge phase (L 3 ), both of which coexisting with a dilute solution phase (W) at high dilution. In the W + L α two phase region, the lamellar phase can often be fragmented into unilamellar vesicles 15 . The bending rigidity of the C 10 E 3 bilayer was determined to be 5 k B T at room temperature 16 . Vesicles, lamellar phase and sponge phase represent different topologies of the bilayer. Transitions between these states require bilayer fusion or fission, that can be triggered for such non-ionic system by temperature jump experiments. Kinetics of both L α -to-L 3 and L 3 -to-L α phase transitions have been studied by time resolved NMR and small angle neutron scattering 17,18 . Fusion kinetics has also been studied by temperature jumps within the L 3 phase 19 . These studies concluded that membrane fusion requires essentially H 0 < 0 and that the rate of fusion increases with decreasing H 0 . This picture was later confirmed by a vesicle stability study ( Fig. 1) 15 . Extruded unilamellar vesicles, of average radius 25 nm, were found to be kinetically stable for temperatures below the three phase line W + L 3 + L α , while above this temperature the spontaneous fusion occurs with a rate that increases with increasing temperature. The three phase line temperature corresponds approximately to the temperature where H 0 changes sign 20,21 .
This dependence of the fusion kinetics on H 0 is in agreement with previous experimental work on biological lipids and models of membrane fusion, involving a hemifusion stalk intermediate of negative monolayer Scientific RepoRts | (2019) 9:2292 | https://doi.org/10.1038/s41598-019-38620-9 curvature [22][23][24][25] . Studies on lipid systems have shown that fusion kinetics depends on the membrane composition. Lipids having H 0 > 0, like single-chain oleoyl-lysophosphatidylcholine, prevent fusion, while lipids having H 0 < 0, like dioleoyl phosphatidylethanolamine promote membrane fusion. If large enough vesicles can be prepared, they can be studied and characterized with optical microscopy 26,27 . In this paper we further characterize the fusion of vesicles and the topological transitions of C 10 E 3 membranes with temperature. Using rapid laser scanning confocal microscopy, we follow how fluorescently labeled large unilamellar vesicles (LUVs) spontaneously fuse upon a temperature jump (hereafter named T-quenching, or quenching), following the vertical line in Fig. 1, i.e. crossing the horizontal separation line (horizontal three phase line W + L α + L 3 ) between the W + L α and the W + L 3 regions. Final equilibrium structures vary from large multilamellar to droplet-like sponge phase objects depending on the quenching temperature. The sponge phase droplets formed here are analogous to the high genus vesicles studied by Noguchi 28,29 and may also serve as a model for certain cell organelles 30,31 . Furthermore, flicker noise spectroscopy and spatial autocorrelation were applied to analyze the shape fluctuations of an isolated C 10 E 3 bilayer, allowing us to estimate the bilayer bending rigidity κ and the membrane internal tension σ 0 .

Results
Imaging structural transformation. The vesicles are formed in the W + L α two phase region where the vesicles are expected to coexist with C 10 E 3 monomers (Fig. 1) smaller than one, V and R being the actual inner volume of the vesicle and the average radius of its apparent, equivalent sphere respectively, with (ii) a low bending rigidity of the C 10 E 3 bilayer, i.e. κ  k T 5 B 16 , and with (iii) a low enough value of the membrane lateral tension.  When quenched at 28, 30, or 35 °C, LUVs undergo shape and structure transformations that we describe in the following. T-quenching experiments were repeated 3 to 5 times, Figs 2, 3, 4, 5 and 6 retain some of the typical structures observed during these experiments. When tubular vesicles induced by shearing during sample preparation (Methods section) are present in the initial dispersion, they systematically evolve towards LUVs (Fig. 2) prior to any other kind of membrane shape evolution. Such process takes several tens of seconds to a few minutes, depending on the quenching temperature, but it also takes place at 20 °C, over a longer time, i.e. ~30 min. Figure 3 shows a typical transformation under T-quenching, of an initial, at equilibrium LUV, a process that takes ten to twenty seconds. A unilamellar vesicle first grows in size while elongating (Fig. 3b,c), and then curves until closure ( Fig. 3d-f), forming a 'bilamellar' structure. Images in Fig. 3, extracted from Supplementary Movie M3, correspond to optical z-cuts taken at a fixed z-position in the sample. Movies of such vesicle evolution were also acquired while cycling over the z-position, refered to as (z, t) scans, see Supplementary Movie M4 that is a 3D averaging projection of such scan. They strongly suggest that the intermediate, elongated structures similar to those in Fig. 3b-d have a biconcave disk shape, while the final object in Fig. 3f, exhibits average spherical symmetry (Supplementary Movie M4). Insets in Fig. 4b (T-quenching 28 °C) further confirm this assumption. Thus, the most likely scenario for the initial LUV transformation upon T-quenching consists of a growth towards a "red blood-cell"-like structure followed by an invagination, leading to an onion-like object (Fig. 3g). We attribute the process of membrane area growing in Fig. 3b,c to membrane fusion, referred to as 'inter' (-object fusion) in Figs 3, 4 and 5. Indeed, our observations exclude a significant contribution of a membrane-enrichment process via the solubilized fraction of the C 10 E 3 molecules, that would require a simultaneous decrease in size of neighboring LUVs in the sample similarly to the Ostwald ripening process 33 , a phenomenon never observed in our experiments (see Supplementary Movies M4-M8). Fusion events of small LUVs could not be resolved, we could however catch some fusion events of multilamellar objects. Besides, the shape evolutions shown in Figs 3c-e, 4 and 5, corresponding to the 'shape' transition schematically depicted in the figures, is compatible with the expected tendency of the C 10 E 3 bilayers to form saddle shape structures (invagination process) while temperature increases in the sample, to accommodate for the concomitant decrease in H 0 . Objects similar to Fig. 3f were seen to only persist over short times (several seconds), before undergoing further structural transformations, leading to either larger and topologically more complex multilamellar objects, or denser L 3 phase droplets.
When quenched at 28 °C, the initial unilamellar vesicles mainly transform into larger multilamellar L α structures through repeated sequences of growth, anisotropic deformation, fusion, invagination, leading to objects often presenting more than ten apparent membrane layers. Stable structures were obtained after 3 to 8 min, keeping in mind that stability is considered over an experimental time scale of ~30 min. Typical stable multilamellar structures are shown in Fig. 4d,e, being whether onion-like, i.e. symmetric (Fig. 4e), or non symmetric (Fig. 4d), as shown in Supplementary Movies M9-M11. On the contrary, T-quenching at 30 °C (resp. 35 °C) results after a period of 2 to 5 (resp. 0.5 to 3) min in dispersions containing mainly L 3 droplets capable of fusion  Fig. 5l (30 and 35 °C). Beside 'inter'-object fusion and 'shape' transition ( Fig. 3) was an 'intra'-object fusion process clearly evidenced upon T-quenching experiments at 30 and 35 °C, while to a less extent upon T-quenching at 28 °C (Figs 4f and 5l respectively). Schematically we can describe the overall process after a temperature jump as follows. It begins with the fusion of the initial unilamellar vesicles. Such events, when they occur, are rapid and not resolved in our experiments for initial LUVs. However, we clearly see the decrease in the number density of vesicles and the increase of their average size. From these very first moments is the local temperature increase of the sample related to a decrease of H 0 , and hence to a bilayer topology transformation resulting in a gradual decrease of the average Gaussian curvature, 〈K〉, and the Euler characteristic, χ E = 〈K〉A b /2π 34 , of the membrane, where A b is the bilayer membrane area. Bilayers accommodate that temperature increase with the creation of saddle points, that in turn results in the 'shape' transformation depicted above, i.e. in the evolution towards onion-like structures with an increased number of bilayers. As seen in Fig. 3

and in Supplementary
Movies M3-M8, such 'shape' transformation process expands over several seconds. For a low enough temperature increase, and certainly for a low enough value of the rate of temperature increase has the system the time to evolve through the building of such multilamellar objects, where bilayers still exhibit micrometer range values of their radius of curvature. This corresponds mainly to 28 °C T-quenching experiments (Fig. 4b-e). Besides, is the rate of local temperature increase much higher upon T-quenching at 30 °C and 35 °C, concomitant to a stronger and faster decrease of H 0 . As a consequence, bilayers certainly cannot undergo the required rate of sequences of 'inter'-fusion and 'shape'-transformation in order to accommodate with the corresponding high number of saddle points. The system thus rearranges through 'intra'-fusion processes, leading to a denser organisation of the bilayers, with sub-micron range values of their local radius of curvature. This gradual topology transformation and increase of the bilayer volume fraction, φ of the structures, is expected to terminate when φ reaches the value at the sponge phase binodal line to the given temperature. From the phase diagram, we thus expect the steady state φ in the sponge phase droplets to increase with increasing temperature. Our experiments show that a wide variety of structures coexist during T-quenching. As a matter of fact, it can be considered that a competition between inter-vesicle and intra-vesicle fusion is taking place, controlled by the local rate of temperature increase. Thus, that variety of structures found is likely the result of that competition. An example of such variety is given in Fig. 6. One the one side, at 28 °C, amongst major onion-like structures, were found some rare, denser structures that did not fuse at the experimental time scale (Fig. 6a, Supplementary Movie M14), while at 30 °C, amongst major L 3 droplets, were found big, space extended structures, similar to the ones shown in Fig. 6a, but with a strong fusogenic character (Fig. 6b-d, Supplementary Movie M12). We note that the sponge phase droplets often contain one or more unilamellar membrane "domes", as seen in Fig. 5e,j. Certainly, dome precursors are seen in Fig. 6a and Supplementary Movies M14 and M15. While in those, stable intermediary structures (28 °C), multiple bilayers are observed, this is rarely the case in L 3 -phase droplets, i.e. at 30 and 35 °C (Fig. 5e,j), where most often only one dome remains per droplet. Similar domes are also often observed on cubosomes 35 , that could be seen as the crystalline analogs of sponge phase droplets. In the bicontinuous cubic phase the multiply connected bilayer has crystallized in a cubic lattice, while being disordered (liquid) in the case of the sponge phase. In both cases however, the multiply connected bilayer has to meet droplet interface in a particular way to avoid unfavorable bilayer edges. Confined between two solid surfaces, a sponge phase was found to form planar lamellae 36 . Hereby, the sponge phase droplets are pictured as having a unilamellar bilayer envelope, from which the multiply connected sponge structure grows inward as it necks in the envelope. Surprisingly, those domes remain present once L 3 -phase droplets have adhered on the bottom glass substrate (Fig. 5k).
Bending modulus of a single bilayer. Mechanical properties of fluid membranes have been extensively studied over the last three decades starting with the pioneering work by Helfrich 9,[37][38][39] . This field of research was born with the growing interest in phospholipid bilayers, that not only appeared as model systems of cell membranes, but also as interesting tools for investigating soft matter physics. Thanks to their very low solubility, typically in the order of pM to nM, phospholipids build, indeed an almost perfect system, i.e. free standing bilayers in a surfactant-free solution. Conversely, C 10 E 3 has a non negligible monomer solubility, representing 5% of the molecules in our experiments. However, this work demonstrates that C 10 E 3 membranes exhibit long term stability at a constant temperature, i.e. over tens of minutes, though continuously exchanging surfactant molecules with their aqueous environment. We imaged a fluctuating bilayer at a frame rate of 172 frames/sec, that was found to be sufficiently high to enable the resolution of the membrane fluctuations (see Supplementary Movie M16, and Fig. 7). The imaged membrane corresponded to a region of a dome from a stable multilamellar structure at 28 °C, similar to the one shown in Supplementary Movie M15. The observed section of membrane was distant enough from the rest of the structures so that no steric interactions with another membrane was observed during the movie. Images were processed as described in the SI section, in order to extract the membrane profile (x(t), y(t)) with sub-micron resolution. Figure 7 shows two typical snapshots of the Supplementary Movie M16 together with the time averaged image of the movie (grey level insets, average on bottom), as well as the corresponding extracted profiles (rectangular window). As seen in the figure, the average membrane profile is circular, with a radius of curvature R = 9.3 ± 0.1 μm, thus showing that domes are sections of a sphere. (x(t), y(t)) profiles were further converted into circular coordinates, and expressed in terms of r(θ(t), t), the radius of the profile from the center of symmetry of the spherical dome, θ(t) being the angle variable in the cylindrical reference space (in 2D geometry in the images). Finally, we calculated the height function h(θ, t) between average and instantaneous membrane profiles as described in the SI section.
Following methods well described in literature, we applied flicker noise spectroscopy to the space or time variations of the h(θ, t) curves, to extract the bending rigidity κ of the membrane: Table 1 compares κ obtained from the analysis of the average fluctuation spectrum 40 and from the evaluation of the average angular autocorrelation function 41

Discussion
Our study demonstrates the applicability of rapid confocal microscopy and highly fluctuating C 10 E 3 vesicles to investigate complex membrane fusion processes in situ through T-quenching. Those assembly processes are driven by the L α -to-L 3 phase transition of the surfactant, being characterized by the inter and intra fusion of suspended unilamellar vesicles (LUVs), and the formation of either large multilamellar objects or micrometric sponge phase droplets. This was made possible first, by working with dilute samples (0.003 w/w), and secondly, by quenching samples initially in the L α phase at three temperatures close to and above the L α -to-L 3 transition temperature. For the lowest temperature quenching, i.e. 28 °C, we describe how primary LUVs evolve over minutes towards multilamellar objects with increasing number of membranes and local membrane concentration. We also show more rare metastable structures, that appear as intermediates between the lamellar phase L α and the L 3 sponge phase, being characterized by such an increased local surfactant density that no more individual membranes can be seen with our optical technique. In comparison, temperature quenching at 30 °C or 35 °C led to much faster kinetics of membrane transformation. As a consequence, most of the intermediate structures seen at 28 °C were never observed, and LUVs were evolving over a few seconds towards the sponge phase droplets. Finally, we could record an isolated fluctuating part of a membrane at 28 °C, that we analyzed using both flicker noise spectroscopy, and spatial autocorrelation, to estimate the bending modulus κ of the membrane at that temperature. Our measurement of the bending modulus provides values in the range of 2-4 k B T, in good agreement with literature values 16 .

Methods
Vesicle solution and sample preparations. C 10 E 3 (Sigma Aldrich) was mixed with an amphiphilic dye (BODIPY ® 558/568-C12, from Life Technologies) at a dye/C 10 E 3 fraction of 0.3 mol%. This mixture was diluted at a mass concentration 3.33 g/L in 18 MΩ deionized water (Merck Millipore, Germany), corresponding to a molar concentration 11.5 mM, and to a volume fraction φ = 0.315%, considering a density ρ = 0.94 g/cm 3 32 . Once prepared, the dispersion was gently shaked during three minutes. Such solution showed long term stability at 20 °C, as observed under the microscope. The stock solution was used within 2 days. Observation cells were build using the following protocol: a 1 mm thick double face tape, with a 6 mm diameter pinched hole was stuck on one side of a glass coverslip, forming a cuvette, in which 28 μL of the vesicle solution was injected. The well was then closed by sticking on the free tape side a 1 mm thick glass slide, resulting in a watertight observation cell. The preparation was done at room temperature, i.e. at 20 ± 1 °C. It is worth noting that C 10 E 3 vesicles are shear sensitive. The transfer of the 28 μL droplet was done using a pipette tip (Eppendorf, Germany); the sample injection eventually resulted in the shearing of the vesicles into long membrane tubes that relaxe over tens of minutes at 20 °C, and over less than five min during one typical T-quenching experiment, with a strong dependence on both jump temperature and sample preparation conditions. Such evolution is shown in Supplementary Movie M2 and Fig. 2a-c. Confocal microscopy. The confocal micrographs were recorded on a Leica SP5 CLSM (Germany) operated in the inverted mode (D6000I) using a 100×, 1.4 NA, oil-immersion objective. A 543 nm He-Ne laser was used to excite the amphiphilic dye. The samples were monitored in situ at different temperatures, and an environmental system was used to ensure temperature control with an accuracy of 0.2 °C. The microscope was equilibrated over one hour at the quenching temperature before starting the experiments. A typical T-quenching experiment ran as follows: an observation chamber filled with the LUVs was placed on the microscope stage just after it has been sealed, and the real time confocal imaging was started as soon as possible; we consider that image recording started within less than 1 min (typically 30 s) after the observation cell was placed in the microscope. For each T-quenching experiment the first image sequence was acquired at a frame rate ~10 to 30 frames/s, during 1 to 5 min, in a region of the sample located ~50 μm above the coverslip. Then, other image sequences were acquired at different positions in the sample, with the purpose of collecting images on all possible structures present in the dispersion. In particular, reducing the field of view while increasing the frame rate (up to ~200 frames/s) enabled, at high magnification and resolution, to resolve the membrane thermal fluctuations. Confocal images of lamellar and sponge phase structures are shown without modification except global image contrast enhancement and smoothing if necessary.