DNA-Origami Line-Actants Control Domain Organization and Fission in Synthetic Membranes

Cells can precisely program the shape and lateral organization of their membranes using protein machinery. Aiming to replicate a comparable degree of control, here we introduce DNA-origami line-actants (DOLAs) as synthetic analogues of membrane-sculpting proteins. DOLAs are designed to selectively accumulate at the line-interface between coexisting domains in phase-separated lipid membranes, modulating the tendency of the domains to coalesce. With experiments and coarse-grained simulations, we demonstrate that DOLAs can reversibly stabilize two-dimensional analogues of Pickering emulsions on synthetic giant liposomes, enabling dynamic programming of membrane lateral organization. The control afforded over membrane structure by DOLAs extends to three-dimensional morphology, as exemplified by a proof-of-concept synthetic pathway leading to vesicle fission. With DOLAs we lay the foundations for mimicking, in synthetic systems, some of the critical membrane-hosted functionalities of biological cells, including signaling, trafficking, sensing, and division.


■ INTRODUCTION
Biological membranes coordinate numerous pathways critical to life, from trafficking to signal transduction and cellular motility, 1 many of which rely on regulating the distribution and interactions of membrane machinery. Among other regulatory principles, cells are believed to exploit membrane phase separation and critical composition fluctuations to dynamically generate local heterogeneity. 2−10 Cell membranes are also known to contain inclusions that can accumulate at the lineinterface between coexisting lipid domains 11−15 and relax the associated line tension. 16 −18 It has been proposed that cells use these line-actants as an additional means to dynamically control membrane phase behavior and the lateral organization of membrane inclusions 19 underpinning, for instance, the function of signaling hubs 20,21 and the generation of endo/ exosomes. 22,23 Artificial cell science aims to construct, from the bottom up, synthetic analogues of biological cells that mimic their sophisticated behaviors 24,25 and are expected to widely impact next-generation diagnostics, therapeutics, and bioprocessing. 26−28 Often constructed from lipid bilayers, 29 syntheticcell membranes have been engineered to replicate an array of biomimetic responses, 30 including mechanotransduction, 31 energy conversion, 32−34 and membrane deformation. 35 In parallel to solutions relying on reconstituted membrane proteins, 32−34 fully synthetic DNA nanodevices, 36 anchored to the bilayers by means of hydrophobic tags, have emerged as a versatile toolkit for biomembrane engineering, 37−39 having enabled the design of biomimetic pathways for transport, 40−43 trafficking, 44,45 cell adhesion, 46,47 tissue formation, 48−50 signal detection, 51 membrane remodeling, 52−55 and surface patterning. 56−59 Despite these advances, the design of pathways enabling systematic control over the local structure and composition of synthetic-cell membranes, to the same degree of what is afforded by extant biological machinery, remains an elusive task.
Here we introduce DNA-origami line-actants (DOLAs) to control the formation, stability, and three-dimensional morphology of lipid domains on synthetic bilayers. Leveraging the modularity of amphiphilic DNA nanotechnology, 56 we designed DOLAs to selectively accumulate at the line-interface between coexisting domains on giant unilamellar vesicles (GUVs). Using a combination of experiments and coarsegrained simulations, we show that, similar to interfacial inclusions in three dimensions, DOLAs modulate the tendency of lipid domains to coalesce, stabilizing two-dimensional analogues of Pickering emulsions. The devices can be deactivated upon exposure to a molecular stimulus, triggering domain coarsening and unlocking the sought dynamic control over lateral membrane organization. Combined with osmotic unbalance, we demonstrate DOLAs can form the basis of a synthetic pathway leading to membrane budding-off and fission, exemplifying three-dimensional morphological control.
Owing to their modular design and robust working principle, we argue that DOLAs could constitute a versatile toolkit for synthetic-cell membrane engineering, allowing us to take full advantage of the rich phenomenology of lipid phase separation and design ever-more-advanced membrane-hosted functionalities.

■ RESULTS AND DISCUSSION
Engineering Line-Actants with DNA Nanostructures. Our design of choice for DNA line-actants, depicted schematically in Figure 1a, exploits the versatility of the Rothemund rectangular origami (RRO) as a "molecular breadboard", with an array of regularly spaced binding sites. 61−64 The rectangular tiles (Supplementary Figure 1 and  Supplementary Table 1) feature two sets of either 6, 12, or 24 single-stranded (ss) DNA overhangs extended from the same face of the origami, as shown in Figure 1b. The sticky ends, with sequence δ1* or δ2*, can respectively bind complementary overhangs on double-stranded (ds) DNA anchoring modules functionalized with two cholesterol moieties (dC) or a single tocopherol (sT). 56 We have recently determined that dC and sT modules display, respectively, thermodynamic preferences for accumulating within liquid-ordered (L o ) and liquiddisordered (L d ) domains of phase-separated membranes. Specifically, transporting a dC module from L d to L o produces a moderately favorable free-energy shift (ΔG p,Ld o dC ) of ∼−0.8 k B T, while sT has a stronger affinity for disordered domains (ΔG p,Ld o sT ≈ 1.9 k B T). 56 The two sets of dC and sT modules were distributed on opposite halves of the tiles, with a spacing of 25 30nm between the two distinct sets of anchors ( Figure  1b). Because is comfortably greater than the estimated width of the boundary separating L o and L d phases (ξ ∼ 8 nm, Supplementary Note I), free-energy minimization is expected to drive the accumulation of the tiles across line-interfaces, accommodating each set of anchors in their respective preferred phase (Figure 1a). We produced three DOLA designs, with either 6×, 12×, or 24× anchors of each type,  Table 2) were extended to include overhang domains δ1* and δ2*, with complementarity to dC (blue) or sT (red) anchor modules, respectively (Supplementary Tables 3 and 4). Note that tiles are expected to have all of their δ1 and δ2 overhangs saturated with sT and dC anchoring modules, owing to the strong affinity between overhangs (see Supplementary Note III). We also expect partially bound states, in which tiles are linked to the membrane through a subset of the available anchors, to be thermodynamically unlikely. (c) Hexagonal grids depict the arrangement of staples across the origami plates, where each hexagon corresponds to the 3′ terminus of a staple. We positioned sets of 6× (light pink), 12× (pink), or 24× (magenta) overhangs targeting the same type of anchorpoints. The two sets of anchors are separated by a distance 30nm (6× and 12×) or expecting an increase in the free-energy gain associated with line-accumulation for a larger number of anchors ( Figure 1c).
DOLAs were further labeled with fluorescent (Alexa488) beacons, located on the face opposite to that hosting the anchors (Supplementary Figure 2 and Supplementary Table  5), thus allowing us to monitor their distribution by means of confocal microscopy. Accumulation of the three DOLA designs at the L d −L o line-interface in phase-separated GUVs (DOPC/DPPC/Chol 2:2:1) is demonstrated in Figure 1d with confocal 3D reconstructions, where the equatorial boundary separating the two hemispherical domains is clearly delineated by the fluorescent plates.
A custom-built image segmentation routine, detailed in Supplementary Note II and Supplementary Figure 4, was applied to confocal data to sample the DOLAs' fluorescenceintensity profile across line-interfaces, as sketched in Figure 2a.
The resulting curves, shown in Figure 2b, display clear peaks at the L d −L o boundary location (x = 0), confirming lineaccumulation. As detailed in Supplementary Note II, fitting of the diffraction-limited fluorescent peak and baseline signals from the surrounding L d and L o regions allowed us to estimate a line-partitioning coefficient K p,int , defined as the ratio between the surface density of origami at the line-interface and the average origami surface density on the entire GUV. Figure 2c summarizes the experimentally determined K p,int for the three tested designs, rendering similar results for the 12× and 24× tiles, with a slightly larger value for the 6× variant. Line-interface adsorption models, outlined in Supplementary Note III (see associated Supplementary Figure 5 and Supplementary Tables 6, 7, and 8), enable the estimation of K p,int for the three DOLA designs. We fitted the theoretical estimates to experimentally determined K p,int values, using the overall surface coverage of the origami on the GUVs, σ, as a fitting parameter. The obtained estimates of 1.5 < σ theory 12× < 9.3% and 2.9 < σ theory 24× < 14.4% for 12× and 24×, respectively, are in good agreement with the nominal experimental surface coverage σ exp ∼ 5% (Supplementary Figure 6). For the 6× design, we estimated σ theory 6× < 3.2%, slightly below the nominal experimental value and reflecting the larger observed K p,int and the smaller absolute fluorescence-intensity values (see Supplementary Note III and Supplementary Figure 7). A smaller than expected σ is likely the result of a decreased membrane affinity, consistent with the smaller number of anchor-points available for the 6× tiles. It should further be noted that RRO tiles are prone to bending in the direction orthogonal to that of the double helices, 65 possibly leading to thermal fluctuations in curvature while tiles are in the bulk. These fluctuations would be suppressed upon confining the tiles to the membrane, which would carry an entropic penalty to tile−membrane binding. It is natural to expect that anchoring overhangs may limit the amplitude of shape fluctuations in the bulk, implying that designs with a higher number of anchoring points, i.e., 24× and 12×, would be less affected by the confinement entropic penalty compared to the (arguably) more flexible 6× design. Possible differences in spontaneous (equilibrium) curvature of the tile designs 66 may also impact membrane−tile affinity.

DOLAs Stabilize Two-Dimensional Pickering Emulsions.
Having demonstrated that DOLAs can programmably accumulate at the boundary between coexisting lipid phases, we sought to verify whether accumulation leads to stabilization of the line-interface, as hypothesized for biological lineactants. 14 To this end, we decorated the phase-separating GUVs with 12× DOLAs at a nominal surface coverage of ∼30%, further promoting line-interface saturation (ϕ ∼ 0.75− 0.87). The GUVs, fluorescently labeled with L d -partitioning TexasRed-DHPE for ease of visualization, were heated to ∼37°C , well above their miscibility transition temperature of T m ∼ 33 ± 1°C, 68 to induce lipid mixing. After incubating at high temperature for ∼5 min, the sample was quenched back to room temperature (T = 25°C), leading to the nucleation of L o or L d domains in a background of the opposite phase.
In nonfunctionalized GUVs, domains rapidly coarsen until most vesicles display two quasi-hemispherical domains, resulting in Janus-like morphologies (Figure 3a, bottom). However, in DOLA-functionalized GUVs, coalescence was arrested, leading to morphologies with a large number of small, stable domains, analogous to two-dimensional Pickering emulsions (Figure 3a, top). To quantify the domain-stabilizing ability of DOLAs, we analyzed epifluorescence micrographs collected ∼3 h after quenching and extracted the fraction F of non-Janus vesicles, namely, those that failed to relax into the two-domain morphology. Results, summarized in Figure 3b, clearly demonstrate how most line-actant-decorated GUVs display Pickering morphologies (Figure 3a, top). In turn, the Janus morphology dominates when GUVs lack either dC or sT anchor modules, no tiles are included, or if the origami are prepared omitting the overhangs to target either dC or sT. The latter controls confirm that domain stabilization emerges solely thanks to line-interface accumulation of the rationally designed DOLAs, rather than as a result of nonspecific effects associated with the individual membrane inclusions.
As further confirmation that domain stabilization is underpinned by interfacial accumulation of the origami, vesicles lacking TexasRed-lipids and decorated with fluorescent DOLAs were subjected to an analogous heating−cooling program and inspected with confocal microscopy. Besides detecting domain stabilization, line-accumulation was confirmed at the boundaries of the stabilized lipid domains, as demonstrated in Figure 3c with a 3D confocal reconstruction.
Coarse-Grained Simulations Capture DOLA Line-Accumulation and Domain Stabilization. To further validate our line-actant engineering strategy, we conducted Monte Carlo (MC) simulations using a coarse-grained representation of the experimental system. As presented in Figure 4a and detailed in Supplementary Note IV, we defined a two-dimensional Ising model on a triangular lattice, which displays phase coexistence at sufficiently low temperature. We interpret the phase rich in spin s = 1 (green) as L d and that rich in To replicate the experimental protocol, we prepared hightemperature systems with uniform spin (and tile) distributions, before quenching to a lower temperature at which demixing Programmable, domain-stabilizing, activity of DOLAs as conveyed by the fraction, F, of GUVs exhibiting more than two domains with respect to that of Janus vesicles. As shown in the graphical legend, besides the complete DOLAs (n = 363 GUVs), we exploited the modularity of the platform to produce various control functionalization schemes expected to lack line-active behavior, including omitting either sT (n = 97 GUVs) or dC (n = 187 GUVs) modules, using tiles that lack overhangs targeting sT (n = 125 GUVs) or dC (n = 165 GUVs), and omitting tiles altogether (n = 160 GUVs). (c) Line-accumulation of DOLAs scaffolds lipid domains and confers them with stability against coalescence, as shown with a 3D view of a reconstructed vesicle after heating above and quenching below the miscibility transition temperature. Interfacial accumulation of fluorescent (Alexa488) 12× DOLAs readily shows stable domains rendered from a representative confocal z-stack using Volume Viewer (FIJI 60 ) with contrast enhancement (see Supplementary Figure 9 for reconstructions without contrast enhancement). All scale bars = 10 μm.
occurs. The quenching temperature was chosen to replicate the experimental line tension of ∼1 pN 67 (see Supplementary Note IV). The system was simulated with Kawasaki dynamics, 69 conserving the number of s = −1 and s = 1 spins, initially set to a ratio of 2/3. The tiles, present at a surface coverage σ = 30%, readily localized at the boundaries of the blue domains emerging after quenching, leading to their stabilization (see Figure 4b and Supplementary Movie 1). Figure 4c compares the time-evolution of systems featuring and lacking 12× tiles. When tiles were present, coarsening was slowed down and small domains persisted at long simulation times (t MC = 2 × 10 7 steps). Conversely, over the same time window, the system lacking the tiles achieved complete coarsening, resulting in a morphology with one domain of each phase, analogous to the Janus GUVs (Supplementary Movies 2 and 3, respectively). Supplementary Figure 10 compares frequency histograms of the number of domains in 15 runs at various t MC , confirming the ability of the tiles to slow down domain coarsening, in line with experimental findings outlined in Figure 3.
Despite its coarse-grained nature, simulations offer insights on the different mechanisms leading to domain coarsening and on how these are influenced by the presence of DOLAs. Generally, we observed coarsening followed two distinct pathways: domain coalescence and Ostwald ripening. 70 Simulations suggest that the presence of tiles slows down the former mechanism, by offering a steric barrier that prevents the interfaces of two distinct domains from coming in sufficient proximity and undergo coalescence. Instead, Ostwald ripening, supported by the exchange of particles (spins) between domains via diffusion through the background phase, occurs regardless of the presence of tiles. This second coarsening process is solely responsible for the decrease in total line length observed in the presence of tiles at very long simulation times (Figure 4d, inset), when no further domain coalescence is observed. In experiments, tile-decorated GUVs display stable 2D Pickering emulsions ∼3 h after quenching, when most nonfunctinonalized vesicles have fully equilibrated (Figure 3b), hinting that domain coalescence may be the primary mechanism leading to coarsening, with Ostwald ripening playing a comparatively minor role. Figure 4d compares the simulated time-evolution of the overall line-interface length (L) for systems featuring each DOLA design with those lacking tiles. At late simulation times, the interface length is higher for 24×, followed by 12× and 6×, with the latter design showing values identical to the tileless control, confirming the expected link between the number of anchors and line-stabilizing ability. See also Supplementary Figure 11 and the associated Supplementary Movies 4 and 5 for representative simulation runs with 24× and 6× DOLAs.
Consistently, as shown in Figure 4e, the simulated tileoccupied fraction of the line (ϕ) converges to values that monotonically increase with the number of anchors, in agreement with the experimental/theoretical findings outlined in Figure 2d. Analogous trends are noted when monitoring the time-dependent fraction of tiles at the line-interface, found to increase with anchor-point number (Supplementary Figure  13). Further insights can be gathered from Figure 4f, where we plot the fraction of tiles at the line as a function of interface length. We note that, as L decreases due to domain coarsening, strongly binding 24× tiles remain more persistently pinned at the interface, while 6× tiles are readily expelled. Finally, in Figure 4g, we explicitly explore the correlation between ϕ and line length. As L shrinks, a plateau in ϕ is approached for 6× tiles able to readily desorb from the line, while a correlation is retained for strongly pinned 24× DOLAs.
Fueling Lipid Domain Reorganization with Dynamic DNA Line-Actants. The ability of DOLAs to regulate domain coarsening can be readily coupled to nanostructure reconfigurability afforded by toeholding reactions, 71−73 thus enabling dynamic control over membrane lateral organization. To demonstrate this functionality, 12× line-actants were modified with a 6-nt toehold domain of sequence α on the 3′-end of cholesterol-targeting overhangs (see Supplementary Table 9). As depicted schematically in Figure 5a, a chemical fuel, in the form of an oligonucleotide with sequence α*δ2, can selectively displace the dC-anchoring modules from the tile through a toeholding reaction (see Supplementary Figure 14 for an agarose gel confirming the sought molecular response). When removing the dC anchors, DOLAs are expected to lose their line-interface affinity and instead partition to L d driven by the remaining sT moieties, negating the ability of the tiles to stabilize lipid domains.
As sketched in Figure 5b (left) and shown in Figure 5c (left) with fluorescence micrographs, GUVs decorated with the responsive 12× tiles form stable 2D emulsions when subjected to the heating−cooling cycle outlined above. Figure 5b,c (center and right) show domain coarsening triggered by exposure to the fuel strand, which translates into a clear decrease in the fraction F of non-Janus GUVs, shown in Figure  5d. The series of snapshots in Figure 5c show that, while some GUVs acquire a Janus morphology after fuel exposure (top), others retain a larger number of domains. Nonetheless, all observed GUVs experience a substantial degree of domain coarsening, as outlined in Figure 5e, where we track changes in the number of domains following fuel addition in individual GUVs. As exemplified in Supplementary Figure 15, domains often bulge after fuel addition, possibly due to a slight osmolarity mismatch coupled with differences in spontaneous curvature between the phases. Bulging may in turn be responsible for enhancing stability against coalescence of the domains after line-actant deactivation, preventing some GUVs from relaxing into the Janus configuration. 74 Responsive DOLAs for Domain Fission in Synthetic Membranes. Reversible domain stabilization with DOLAs can be exploited to synthetically replicate the action of biological membrane-remodeling machinery, tackling a critical bottleneck in synthetic-cell engineering. 75 For instance, phospholipase A 2 has been postulated to exhibit enhanced catalytic activity at line-interfaces between coexisting L d −L o phases, driving domain budding and fission. 76 We propose our line-actants could form the basis of an analogous pathway coordinating fission and, therefore, three-dimensional membrane transformation.
We prepared GUVs in an initial configuration featuring DOLA-stabilized microdomains, as outlined above. We then applied a hyper-osmotic shock by increasing the osmolarity of the outside solution (C O ) to ∼1.28 × C I , where C I is the internal vesicle osmolarity, thus leading to an increase in GUV excess area. 77 While in some cases hyperosmosis led to membrane internalization (Supplementary Figure 16), some domains responded by bulging-out, as depicted in Figure 6a and experimentally observed in Figure 6b. At this stage, some bulged-out domains could be "primed" for complete buddingoff or fission, driven by minimization of line-interface energy and opposed by membrane fluctuation entropy 78 and, possibly, steric repulsion between DOLAs preventing line shrinkage. Deactivation of the line-actants through fuel strand addition is expected to increase line tension and negate any steric hindrance to interface shrinkage, tipping the marginally stable domains into full budding-off and fission (Figure 6c, top). Fission events can be promptly recorded experimentally upon fuel-strand addition, as shown in Supplementary Movie 6 and Figure 6c (bottom). The sequence of epifluorescence images in the latter clearly shows the dark region corresponding to a budding L o -domain shrinking progressively and ultimately disappearing. Vesicle fission is confirmed with bright-field micrographs, where arrows highlight both the "parent" and "offspring" vesicles regaining spherical morphologies. An analogous example is provided in Figure 6d and the associated Supplementary Movie 7, with overlaid confocal and bright-field micrographs.
In further support of our proposed synthetic fission pathway, DOLA-functionalized liposomes with stabilized microdomains were gently added to imaging wells containing hyper-osmotic buffer baths with and without fuel. In the former scenario, confocal microscopy revealed further examples of phaseseparated GUVs found in proximity of a smaller single-phase vesicle (Figure 6e, left, and Supplementary Figure 17), similar to the confirmed fission occurrences shown in Figure 6c and d. In turn, we did not observe such presumed fission events in vesicles immersed in baths lacking fuel (Figure 6e, middle), where GUVs appeared more Janus-like, also in line with our findings on sample heterogeneity summarized in Supplementary Figure 16. Finally, DOLA-lacking GUVs, immersed in fuel-supplemented hyper-osmotic buffer, simply adopted dumbbell shapes thanks to their Janus morphology ( Figure  6e, right), consistent with the expectation that the used C O /C I is insufficient to drive fission of quasi-hemispherical domains. 79

■ CONCLUSIONS AND OUTLOOK
In summary, we presented DNA origami nanostructures� dubbed DOLAs�designed rationally to accumulate at the one-dimensional line-interface between lipid domains, imitating biological line-active membrane inclusions. The synthetic line-actants exploit the programmability and modularity of DNA origami, combined with the natural tendency of cholesterol and tocopherol "anchors" to enrich distinct lipid phases. Through experiments and numerical modeling, we demonstrated that the affinity for line-interfaces can be programmed by tuning the number of hydrophobic anchors, thus outlining a general design principle readily applicable to different origami designs, hydrophobic groups, or membrane compositions. We further demonstrated that our line-actants are able to stabilize small lipid domains against coarsening, giving rise to two-dimensional analogues of Pickering emulsions on the surface of giant liposomes and achieving a previously unattainable degree of control over the lateral organization of synthetic bilayers. The nanodevices can be externally deactivated through a toehold-mediated strand displacement operation, triggering domain coalescence. Coarse-grained Monte Carlo simulations replicate the experimental phenomenology, identifying origami−origami steric repulsion as the primary mechanism leading to domain stabilization and ultimately validating our lineactant engineering strategy. Finally, we combined externally triggered line-actant deactivation with osmotic unbalance to implement a proof-of-concept synthetic fission pathway, whereby offspring liposomes programmably bud-off from parent vesicles, demonstrating the ability of the nanodevices to promote large-scale membrane restructuring similar to biological nanomachines. 23,80,81 The simplicity and modularity of the mechanism underpinning line-action in our devices point at possible design variations that exploit different origami geometries, anchor chemistry, and polymerization approaches to further program the number, size, and morphology of lipid domains, marking a conceptual shift in synthetic-membrane engineering. Responsiveness to a wider range of physical and chemical stimuli can also be achieved by incorporating active elements such as DNAzymes, 82 G-quadruplexes, 83 aptamers, 84 and/or photoactuable moieties (e.g., azobenzene), 85 thus broadening the design space and applicability of synthetic membranerestructuring pathways.
Thanks to their ability to program membrane restructuring and the lateral distribution of membrane inclusions, the DOLAs constitute a valuable toolkit in synthetic-cell science that could underpin elusive and highly sought-after behaviors such as synthetic-cell division, vesicle-based trafficking, and signal transduction pathways, ultimately unlocking disruptive applications to advance therapeutics and diagnostics.
Finally, DOLAs could be used to exert control over the local domain structure of living-cell membranes. One could thus envisage applying the origami line-actant toolkit as the basis of biophysical studies aimed at clarifying the role of microphase separation in cell signaling, by triggering the formation and disruption of "lipid rafts" at will. 20 If strong correlations between disease-related pathways and membrane-domain structure are established, the origami line-actants could carry potential therapeutic value by unlocking a previously unattainable strategy for conditioning cell behavior. ■ ASSOCIATED CONTENT
Experimental methods, supplementary notes, supplementary figures, supplementary table with parameters for model and simulations, DNA sequences of the nanostructures used throughout this work (PDF) Movie S1: Representative simulation trajectory with 12× DNA-origami line-actants following the system after quenching for 2 × 10 7 MC cycles. Box size: 320 × 372 lattice points (AVI) Movie S2: Representative simulation trajectory with 12× DNA-origami line-actants following the system after quenching for 3 × 10 7 MC cycles. Box size: 160 × 172 lattice points (AVI) Movie S3: Representative simulation trajectory lacking DNA-origami line-actants following the system after quenching for 3 × 10 7 MC cycles. Box size: 160 × 172 lattice points (AVI) Movie S4: Representative simulation trajectory with 24× DNA-origami line-actants following the system after quenching for 3 × 10 7 MC cycles. Box size: 160 × 172 lattice points (AVI) Movie S5: Representative simulation trajectory with 6× DNA-origami line-actants following the system after quenching for 3 × 10 7 MC cycles. Box size: 160 × 172 lattice points.(AVI) Movie S6: Evolution of L o -domain fission event in a representative vesicle observed with bright-field and epifluorescence microscopy. Time-stamps refer to the time elapsed after initial acquisition. Fuel was added 2 min before acquisition began. The fluorescent marker is TexasRed-DHPE, which labels the L d phase. Scale bar: 10 μm (AVI) Movie S7: Evolution of L o -domain fission event in a representative vesicle observed with overlaid confocal and bright-field microscopy. Time-stamps refer to the time elapsed after initial acquisition. Fuel was added 90 min before acquisition began. The fluorescent marker is TexasRed-DHPE, which labels the L d phase. Scale bar: 10 μm (AVI)