Detergent-mediated incorporation of transmembrane proteins in giant unilamellar vesicles with controlled physiological contents

Abstract

Giant unilamellar vesicles (GUVs) are convenient biomimetic systems of the same size as cells that are increasingly used to quantitatively address biophysical and biochemical processes related to cell functions. However, current approaches to incorporate transmembrane proteins in the membrane of GUVs are limited by the amphiphilic nature or proteins. Here, we report a method to incorporate transmembrane proteins in GUVs, based on concepts developed for detergent-mediated reconstitution in large unilamellar vesicles. Reconstitution is performed either by direct incorporation from proteins purified in detergent micelles or by fusion of purified native vesicles or proteoliposomes in preformed GUVs. Lipid compositions of the membrane and the ionic, protein, or DNA compositions in the internal and external volumes of GUVs can be controlled. Using confocal microscopy and functional assays, we show that proteins are unidirectionally incorporated in the GUVs and keep their functionality. We have successfully tested our method with three types of transmembrane proteins. GUVs containing bacteriorhodopsin, a photoactivable proton pump, can generate large transmembrane pH and potential gradients that are light-switchable and stable for hours. GUVs with FhuA, a bacterial porin, were used to follow the DNA injection by T5 phage upon binding to its transmembrane receptor. GUVs incorporating BmrC/BmrD, a bacterial heterodimeric ATP-binding cassette efflux transporter, were used to demonstrate the protein-dependent translocation of drugs and their interactions with encapsulated DNA. Our method should thus apply to a wide variety of membrane or peripheral proteins for producing more complex biomimetic GUVs.

Membrane proteins are core constituents of the membranes of all cellular compartments; they are involved in major cellular functions such as, for example, cell response to extracellular stimuli, ion and solute transport across the membrane, and cell bioenergetics. They comprise peripheral proteins that transiently bind to the lipid membranes and integral (also called transmembrane) proteins that contain single and several membrane spans and remain in the membranes. Integral proteins are thus ideal targets for drugs. Understanding their role in cell functions or designing drugs requires isolating these proteins to determine their structure and to reincorporate them in a membrane environment to study their functioning.

In this context, giant unilamellar vesicles (GUVs) are micrometer-sized biomimetic membrane systems that are frequently used by biophysicists to study cell functions involving lipid membranes. GUVs are adapted for studying physical properties, such mechanics, adhesion, organization, or dynamics of biomembranes on single vesicles. They are well-suited objects for advanced microscopy studies or micromanipulation (1). In addition, GUVs are promising systems for synthetic biology where one of the ultimate goals is to develop a minimal functional cell integrating biological components in the lumen and in the membrane. Methods to form GUVs with cytosolic or peripheral proteins or nucleic acids in the external or internal volume are now well established (2–5).

In the case of incorporation of transmembrane proteins in GUVs, main methods involved the growth of proteo-GUVs from a quasi-dried film of lipid and proteins by spontaneous swelling or by electroformation (reviewed in ref. 6). This requires that proteins be robust upon drying. Moreover, the protein orientation in the membrane is lost during the formation of GUVs and likely symmetrical. Alternatively, solubilized ion channels in detergent were incorporated after incubation with preformed GUVs (7, 8). However, the final concentrations of detergent were far below the critical micellar concentration (cmc), and thus it is likely that most proteins were aggregated during the reconstitution process. Recently, an elegant strategy to control the orientation of the proteins has been reported and consists of forming the lipid bilayer from two separated monolayers that contain single-span proteins (3). This approach is suitable only for single-span proteins with large extracellular or intracellular domain that can be inserted in a single monolayer, but not for multispan transmembrane proteins.

Thus, there is a need for a method of reconstitution of transmembrane proteins producing suitable GUVs that fulfill several criteria, including no denaturation of proteins, high protein activity and high yield of protein incorporation, unique orientation in the membrane, low membrane passive permeability, and controllable buffer and lipid compositions or lipid asymmetry.

In this context, the major breakthrough for reconstitution in large unilamellar vesicles (LUVs) with diameters in the range of 50–200 nm was the use of detergents. This allowed reconstitution by direct incorporation (9, 10); by detergent removal from a fully solubilized mixture of lipid, protein, and detergent; or by fusion of proteoliposomes or native membranes with pure lipid LUVs (reviewed in ref. 11). The mechanisms of protein insertion during the detergent-mediated reconstitution and the experimental guidelines are now well established. In particular, proteins can be inserted with a unique orientation in a lipid membrane by direct insertion into preformed liposomes destabilized by subsolubilizing concentrations of sugar-based detergents (reviewed in refs. 11–13). Similar concepts have also been used for the unidirectional insertion of proteins in a planar lipid bilayer (14).

Here, we present a method of reconstitution of transmembrane proteins in GUVs based on the concepts of reconstitution by direct insertion into preformed LUVs and on specific properties of detergents in the presence of lipid bilayers. The detergent concentration is maintained above the cmc to keep proteins solubilized during the incorporation process. GUVs that can contain complex, biologically relevant internal solutions are formed in the presence of a subsolubilizing concentration of detergent. Membrane proteins are next unidirectionally inserted in these GUVs before detergent removal. We applied the method to generate GUVs containing bacteriorhodopsin (BR), a photoactivable proton pump that produces a transmembrane-stable and light-switchable electrochemical gradient of protons. We designed a biomimetic system containing different partners to follow the DNA injection by T5 phage upon binding to FhuA, a receptor for ferrichrome-iron in Escherichia coli and its transmembrane receptor. Finally, we extended the method using the fusogenic properties of detergents to incorporate proteins by fusion of purified native membranes or proteoliposomes to GUVs made of lipids mimicking the external leaflet of plasma membranes of eukaryotic cells. After incorporation of Bmc/BmrD, an efflux ATP-binding cassette (ABC) transporter from Bacillus subtilis, we demonstrated the protein-dependent translocation of drugs and their interactions with encapsulated DNA.

Results and Discussion

Principles of Transmembrane Protein Reconstitution in GUVs in the Presence of Detergents.

The method consists of first forming GUVs made of lipids and mild detergents at subsolubilizing concentrations (Fig. 1A). Here, we use electroformation in sucrose solution to ensure formation of larger amounts of GUVs compared with growth in salt buffers (15), but GUVs formed by other methods could in principle be used. Moreover, GUVs can be electroformed with several lipids, including biologically important lipids such as phosphatidylinositol or cardiolipin. The presence of detergent increases the ionic permeability of the GUV membrane and allows for the exchange of internal and external contents after mixing with any buffers. We can thus prepare GUVs with specific lipid compositions that after buffer exchange have different internal ionic contents. This also greatly simplifies working on GUVs in physiological buffers compared with the electroformation of GUVs in the presence of an appropriate buffer. Additionally, this method also preserves GUVs against osmotic shocks during buffer exchange, yet detergent concentration is low enough to maintain large encapsulated molecules such as proteins or DNA in the internal volume. We use a detergent concentration higher than the cmc to keep proteins solubilized during the incorporation step. This is strikingly different from methods in which robust proteins like channels were incorporated well below the cmc of the detergent and a large amount of proteins were thus likely aggregated (7, 8).

Fig. 1.

Detergent-mediated reconstitution of transmembrane proteins in GUVs in physiological buffers. (A) Closed GUVs are grown by electroformation in the presence of subsolubilizing concentrations of mild detergents in sucrose solution (pink background). (B) The presence of detergent allows the fast equilibration of any physiological buffer (light purple background) between internal and external volumes and avoids the denaturation of solubilized transmembrane proteins. (C) Protein incorporation. (C, I) Purified and solubilized transmembrane proteins are unidirectionally reconstituted by direct insertion into GUVs. (C, II) Transmembrane proteins are reconstituted in GUVs after fusion of purified native vesicles in the presence of detergent. (D) Detergent is removed and GUVs are diluted in assay buffer that could be different from the internal buffer (salmon background).

Next, solubilized proteins are either directly incorporated in detergent-destabilized GUVs (Fig. 1 B, I and C, I) or native vesicles are fused to GUVs (Fig. 1 B, II and C, II). Finally, detergent is removed by hydrophobic adsorption onto polystyrene beads, Bio-Beads, or by complexation by cyclodextrins (Fig. 1 D, I and II).

Forming GUVs in the Presence of Detergents and in Physiological Buffers.

GUVs can be formed in the presence of n-dodecyl-β-D-thiomaltopyranoside (DOTM), a sugar-based detergent used for the purification of membrane proteins, at concentration above the cmc. GUVs were formed from a mixed film of lipid and DOTM to allow for a homogeneous distribution of detergent between the inner and outer lipid leaflets (Fig. 2). GUVs of typically 5–50 μm in diameter were formed and were stable for hours in the presence of 100 μM DOTM (i.e., twofold the cmc) (Fig. 2C). They exhibit a size and shape similar to that of GUVs formed in the absence of DOTM (Fig. 2A) or at DOTM concentration at the cmc (Fig. 2B). Higher concentrations of DOTM decrease the amount and the size of GUVs up to the formation of lipid aggregates (Fig. 2D). GUVs were also formed in the presence of two other detergents widely used for membrane protein purification, n-dodecyl-β-D-maltoside (DDM) or octylphenol polyethylene oxide (Triton X-100) at 0.2 mM and 0.25 mM, slightly above the cmc (i.e., 0.17 mM and 0.2 mM, respectively).

Fig. 2.

Formation of GUVs in the presence of detergent and in physiological buffer. (A–D) Formation of EPC/EPA GUVs in the presence of detergent (DOTM) above the cmc. (A) No detergent. (B) Fifty micromolar DOTM, corresponding to the cmc. (C) One hundred micromolar DOTM. (D) Two hundred micromolar DOTM. (E–G) Exchange to buffers. GUVs formed in sucrose and 75 μM DOTM after dilution in a high-salt buffer without (E) or with (F) pyranine (green). DOTM GUVs are permeable to ions but not to large molecules (G) GUVs where DOTM was removed before buffer exchange, aggregate after dilution and in high-salt buffer.

Next, GUVs prepared in DOTM and sucrose solution at 400 mOsm were injected in buffer solutions of higher osmolarity (600 mOsm) with different ionic contents. After injection, GUVs were unaffected, demonstrating that internal and external media were equilibrated (Fig. 2E; Fig. S1 shows experiments with other detergents). However, larger molecules such as pyranine, a soluble pH-sensitive fluorescent probe, remained outside of GUVs after buffer exchange (Fig. 2F). Thus, detergents at concentrations slightly above the cmc increase the ionic membrane permeability without forming or stabilizing pores in the lipid bilayer. In contrast, when detergent is removed by hydrophobic adsorption onto polystyrene beads, a similar injection leads to aggregation or burst of the GUVs, revealing that the membrane has much lower water permeability after detergent removal (Fig. 2G).

Direct Incorporation of Solubilized Membrane Proteins in Preformed GUVs.

For preformed LUVs, unidirectional insertion of transmembrane proteins is dependent on both the type and the concentration of detergent present in the lipid bilayer. Insertion is only mediated by sugar-based detergents, such as n-octyl-β-D-glucoside, DDM, or DOTM. The concentration of detergent is close to saturation of liposomes (i.e., just below the opening of the vesicle). It is established that proteins are inserted in the membrane through their most hydrophobic domains by a transfer mechanism of proteins from the mixed micelles to the lipid membrane not requiring the formation of specific holes in the membrane (ref. 16; reviewed in refs. 11 and 12). Even in the case of highly hydrophobic proteins, short peripheral loops, as for EmrE (17) or for LeuT (18), or five negatively charged amino acids in the carboxylic tail of BR (19), are sufficient to produce a single orientation in the membrane.

We first incorporated BR, a light-induced proton pump, in GUVs destabilized with 75 μM DOTM. GUVs were electroformed in the presence of pyranine (green) and incubated with fluorescently red-labeled BR micelles. After detergent removal, internal pH variations due to proton translocation were measured using pyranine fluorescence (Fig. 3A). GUVs were both red and green, demonstrating that BR was incorporated in all GUVs (Fig. 3B). BR was incorporated in GUVs made with different lipid mixtures including egg phosphatidylcholine (EPC), EPC/egg phosphatidic acid (EPA), and diphytanoyl-phosphatidylcholine (DphPC). It is worth noting that no incorporation was obtained in the absence of DOTM, confirming similarities with the mechanism of direct incorporation in LUVs mediated with sugar-based detergents.

Fig. 3.

Direct incorporation of solubilized transmembrane proteins in GUVs. Reconstitution of red-labeled bacteriorhodospin, a photoactivable proton pump, in DOTM destabilized EPC–EPA GUVs. (A and B) After detergent removal, light-induced proton pumping induces an internal acidification of GUVs. (C) Normalized pyranine fluorescence intensity and variance in the presence (black columns, n = 350) and in the absence (gray columns, n = 440) of valinomycin. (D) Schematic representation of the reconstitution of FhuA, a T5 phage receptor, in red-labeled EPC GUVs encapsulating YOPRO-1, a green fluorescent DNA probe. (E) After binding of T5 phage to FhuA, DNA is injected inside GUVs. (Scale bars: 10 μm.)

Actinic illumination induced a time-dependent fluorescence decrease related to an internal acidification in all GUVs (n = 400) and negligible alkalinization (n = 8). A maximum decrease of fluorescence of 40% was observed corresponding to a pH gradient of 0.8 pH unit (acidic inside) (Fig. 3C). This is fourfold larger than the 10% decrease reported for BR GUVs grown in a medium without buffering from dried BR-lipid film (20). In the later method BRs were likely symmetrically oriented in the membrane of GUVs, thus pumping protons in opposite directions. The high pH gradient found after direct incorporation in GUVs confirmed that BR proteins are collectively pumping protons in the same direction in an inside-out orientation, as also demonstrated after direct incorporation of BR in LUVs (11).

This also showed that the passive permeability of the membrane of GUVs to protons was low. Indeed, it is worth noting that no or a small acidification was observed unless valinomycin, a K⁺ selective carrier, was added. The light-induced proton translocation produced a transmembrane electrical potential (ΔΨ) positive inside responsible for a retroinhibitory effect of proton pumping by BR (21). In the presence of valinomycin, the ΔΨ is overcome by compensatory K⁺ movement and a change in pH can develop due to proton pumping. The large amplitude of the proton gradient results also from the low passive proton permeability of BR-containing GUVs, as shown in an experiment in which GUVs submitted to a 1.5 pH unit jump returned to equilibrium only after 40 min (Fig. S2).

The low proton permeability demonstrates that detergent molecules were fully removed by Bio-Beads or cyclodextrins. The residual DOTM concentration could be calculated from the calibration curves previously reported for LUVs, leading to DOTM concentrations below 10 nM (22).

The amount of incorporated BR was measured using fluorescence of red-labeled BR after calibration of the fluorescence (Materials and Methods and Fig. S3). The BR density was calculated to 1,800 ± 570 BR/μm ² (n = 25), corresponding to a lipid/protein ratio of 2,200 ± 700 mol/mole. This density is high (see ref. 23 for comparison) and is consistent with the high functionality of the BR GUVs.

A more complex reconstituted system was designed with FhuA, an Escherichia coli outer membrane protein for which the functional assay involved several partners. FhuA is a receptor for the ferric siderophore ferrichrome as well as for several bacteriophages. The ability of phage T5 to bind to FhuA reconstituted in LUVs and to release its DNA has been analyzed in detail (for a recent review see ref. 24) but never shown with GUVs. GUVs were grown in the presence of soluble YOPRO-1, a green fluorescent probe with enhanced fluorescence after binding to DNA (Fig. 3D). Internal sucrose solution was exchanged with the buffer required for the function of FhuA and of T5 phage. After incorporation of FhuA and incubation with T5 phage, a large increase of green fluorescence was observed inside GUVs (Fig. 3E and Fig. S4). This shows that the cell-surface exposed loop 8 of FhuA was accessible to T5 phage (25) and that FhuA was inserted through its hydrophobic periplasmic domain in an inside-in orientation.

Reconstitution of Transmembrane Proteins in GUVS by Fusion of Native Membranes or Proteoliposomes.

Several detergents, including Triton X-100, have been reported to induce the fusion of pure lipid LUVs, proteoliposomes, or native membranes when added at subsolubilizing concentrations (26–28). We thus evaluate whether detergents could also induce the fusion of small vesicles containing membrane proteins to GUVs.

Fusion experiments were first performed with two types of native vesicles: inverted inner membrane vesicle (IMV) purified from an E. coli strain overexpressing BmrC/BmrD, a bacterial heterodimeric ABC transporter, and with chromatophores purified from photosynthetic bacteria. In both cases, a nonspecific protein labeling of the native vesicles was achieved with Alexa 488 (green). Native vesicles were 50–100 nm in diameter, unilamellar, and remained with a closed membrane in the presence of detergent as seen by cryoelectron microscopy (Fig. S5).

Unlabeled GUVs were incubated with green-labeled native membranes and increasing concentrations of Triton X-100; fusion was monitored through the appearance of fluorescence at the membrane of the GUVs and the functionality of the reconstituted proteins (see below). Fusion increased with addition of increasing concentrations of Triton X-100, within the range of 100–300 μM, up to concentrations leading to GUV solubilization (Fig. 4 A and B). The concentration of Triton X-100 for fusion was tuned with the concentration of native membranes. Because Triton X-100 molecules partition between lipid membranes of GUVs and native membranes, higher concentrations of Triton X-100 were required with higher concentrations of native membranes. Weak or no fusion occurred in the presence of DOTM or DDM, and no fusion was observed in the absence of detergents.

Fig. 4.

Incorporation of transmembrane proteins by fusion of native membranes with GUVs. Fusion of green-labeled inverted inner vesicles (IMVs) of E. coli containing overexpressed BmrC/BmrD, a multidrug resistance transporter to nonlabeled EPC–EPA GUVs in the absence (A) or the presence (B) of 300 μM Triton X-100. (C) Fusion of green-labeled chromatophores to DOPC/DOPE GUVs containing Tx-red–labeled lipid. (D) Fusion of proteoliposomes containing red-labeled bacteriorhodospin to unlabeled DOPC/DOPE GUVs in the presence of 300 μM Triton X-100. (E and F) Fusion of green E. coli IMVs to red-labeled GUVs containing lipid domains of DOPC/sphingomyelin/cholesterol, in the absence (E) or in the presence (F) of 300 μM Triton X-100. (Scale bars: 10 μm.)

Proteins were incorporated in all GUVs (Fig. 4C). Similar results were found with GUVs prepared with charged EPC/EPA or noncharged mixtures of lipids, DphPC, EPC, or dioleoyl-phosphatidylcholine/dioleoyl-phosphatidylethanolamine (DOPC/DOPE), ruling out an electrostatic mechanism of binding and fusion. Fusion was not related to specific proteins or lipid present in the native membrane because incorporation was obtained with IMVs and chromatophores that have strikingly different protein and lipid compositions and because fusion of pure E. coli lipid LUVs also occurred (Fig. 4C and Fig. S6 B and C).

Moreover, fusion of IMVs and protein incorporation also occurred with GUVs containing cholesterol (chol) and brain sphingomyelin (Sph) and exhibiting lipid domains (Fig. 4 E–F). Because protein labeling on the native vesicles was unspecific, it is not surprising to observe a nondifferential distribution between liquid-ordered (black) and liquid-disordered (red) lipid domains.

We controlled by cryoelectron microscopy that native vesicles remained with a closed and roughly spherical at detergent concentrations used during the fusion process. Thus, it suggests that fusion is mediated by defects generated by detergents in the two adjacent bilayers and thus that the orientation of proteins in GUV membranes is the same as in the native vesicles (29).

We also succeeded in fusing proteoliposomes containing BR or BmrC/BmrD to GUVs (Fig. 5D and Fig. S6A) in the presence of Triton X-100. Again, fusion was dependent on the detergent concentration and no fusion occurred in the absence of detergent. After fusion, the lipid-to-BR ratio was calculated as 1,000 mol/mole, which corresponds to 4,000 ± 2,000 BR/μm².

Fig. 5.

Drug transport by BmrC/BmrD, a bacterial ABC transporter in GUVs. EPC/EPA GUVs are grown encapsulating DNA before incorporation of BmrC/BmrD by fusion of green E. coli IMVs. Addition of ATP induces the translocation of ethidium bromide (EtBr) and its binding to DNA induces an enhancement of its fluorescence. The transport is inhibited by addition of orthovanadate (Vi), a specific inhibitor of BmrC/BmrD. (Scale bars: 10 μm.)

We next demonstrated that this fusion process also preserves protein functionality. We followed the ATP-dependent translocation of drugs in the internal volume of GUVs driven by Bmc/BmrD transporter. GUVs were grown with encapsulated DNA (Fig. S7), and next BmrC/BmrD, overexpressed in native E. coli membranes, was incorporated as above. Addition of ATP and ethidum bromide (EtBr) in the external medium led to an increase of the internal fluorescence due to the binding of EtBr to encapsulated DNA (Fig. 5). The translocation of EtBr was inhibited by orthovanadate (Vi), a specific inhibitor of BmrC/BmrD, or by the absence of ATP.

Conclusion

The current challenge in the design and preparation of GUVs as model membranes either for cell biology-related questions or for synthetic biology perspectives is to increase the complexity of the protein and lipid components while still keeping a biomimetic or efficient distribution of these components between the GUV membrane and the GUV lumen. Current methods only allow either the encapsulation of soluble proteins or nucleic acids in GUVs not containing proteins at the membrane or the nondirectional reconstitution of transmembrane proteins in GUVs containing simple buffers devoid of soluble proteins or nucleic acids. Here we have shown that using detergents it is possible to incorporate transmembrane proteins in GUV membranes made of different lipids, including lipid mixtures representative of the plasma membranes of eukaryotic cells. We have tested our method on different categories of proteins: either constituted of only hydrophobic helices (BR) or of beta-barrel (FhuA) or with a large extramembraneous domain (BmrC/BMCD). Reconstitution was performed either with purified proteins solubilized in detergent micelles or directly with proteoliposomes or purified native membranes. Because our protocol uses the same strategy as for insertion in preformed LUVs, it allows a unidirectional incorporation of the membrane proteins. Internal and external contents were physiological and tunable; in some cases, GUVs contained DNA molecules, but our method should also apply to GUVs encapsulating other protein solutions. Moreover, protein insertion was performed without residual oil present in the lipid membrane or without a drying step, as in previous existing methods. Thus, we anticipate that the method we designed will be accessible to a broader number of transmembrane proteins compared with the existing ones and will open a larger range of applications.

To develop the method, we hypothesized that the concepts of detergent-mediated reconstitution in LUVs should also apply to reconstitution in GUVs. This first report demonstrates that the hypothesis was robust enough. We thus anticipate that further applications could be derived from existing reconstitution experiments in LUVs. For instance, because direct insertion was detergent- and not protein- or lipid-dependent, we expect that, as demonstrated with LUVs, it will be possible to coreconstitute transmembrane proteins that are functionally coupled (30). Moreover, the method would be applicable with other types of membrane proteins purified in detergent, such as peripheral proteins anchored in the lipid bilayer by an aliphatic anchor or single-span proteins. Finally, our experiments of fusion of pure lipid liposomes to GUVs suggests that internal contents of LUVs (e.g., proteins or nucleic acids) could be easily delivered to GUVs, allowing studying their interactions with the “cytosolic” side of a GUV membrane containing the transmembrane proteins. Because our method is versatile and easily applicable to a large range of transmembrane proteins and of biological internal contents, we expect it will provide a significant breakthrough both in synthetic biology and in biophysics.

Materials and Methods

Additional information on materials, protein expression, purification and fluorescent labeling of LUV formation, image acquisition, and analysis is available in SI Materials and Methods.

Formation of GUVs in the Presence of Detergent and Exchange of Buffer.

GUVs grown in the presence of detergents were generated by electroformation following the method reported in ref. 31 with the following modifications. The stock solution of lipids, fluorescent lipids, and DOTM were prepared in CHCl₃. Mixtures were prepared at 1 mg/mL of lipids, 0.5% (wt/wt) Texas Red 1,2-dihexadecanoyl-sn-glycero-3-phosphoethanolamine, triethylammonium salt (Tx-DHPE) or 0.2% (wt/wt) 2-(4,4-difluoro-5,7-dimethyl-4-bora-3a,4a-diaza-s-indacene-3-pentanoyl)-1-hexadecanoyl-sn-glycero-3-phosphocholine (β-Bodipy-FLC5-HPC), and 1.5 mM DOTM. The amount of DOTM was calculated considering the 600-μL volume of the chamber of preparation and 75 μM DOTM (cmc 50 μM), final concentration. A drop of 15 μL of the lipid/detergent mixture was spread on each indium tin oxide-treated glass plate used to form the chamber. After overnight drying under vacuum, the dried film was rehydrated with a 400 mM sucrose solution. GUVs were also formed in the presence of Triton X-100 (cmc 200 μM) and DDM (cmc 170 μM) at 0.25 mM and 0.2 mM final concentrations, respectively (Fig. S1).

For the experiment of exchange of buffers depicted in Fig. 1C, a drop of 20 μL of GUVs prepared in 75 μM DOTM and 400 mM sucrose was mixed with 20 μL of buffer [50 mM Hepes (pH 7), 500 mM KCl, and 75 μM DOTM] with or without 150 μM pyranine. The same experiment was performed with GUV solution from which DOTM was previously removed by addition of Bio-Beads (see below).

In all experiments, for electroformation of GUVs, an AC electric field provided by a pulse generator was applied at room temperature for 3.5 h across the chamber at 1.1 V at 10-Hz frequency. GUVs containing coexisting lipid domains were prepared by electroformation from a dried lipid film of DOPC/Sph/chol 30/40/30 (mol/mol/mol) at 60 °C.

Direct Incorporation of Solubilized Proteins in GUVs.

Bacteriorhodopsin.

GUVs of EPC/EPA 9/1 (mol/mol) or EPC lipid mixtures were prepared in the presence of 75 μM DOTM in a 400 mM sucrose solution supplemented with 150 μM pyranine. A drop of 40 μL of GUVs in 75 μM DOTM was mixed with 20 μL concentrated buffer [200 mM K₂SO₄, 20 mM Pipes (pH 7.5), and 75 μM DOTM]. Then, solubilized BR was added at 0.5–1 μM final concentration and incubated 2–16 h at room temperature before detergent removal.

For confocal imaging, 5 μL of sample was diluted 10–20 times in a solution of 200 mM glucose, 100 mM K₂SO₄, 10 mM Pipes (osmolarity was adjusted to the same value as the sample) (pH 7.8), and 0.01 μM valinomycin, final concentration. Light-induced proton transport by BR was performed illuminating the sample using a 12-V, 100-W halogen lamp with a low wavelength cutoff at 500 nm and a heat filter. It was checked that during illumination no bleaching of pyranine was observed.

Variation of the pH of the internal volume was deduced from the change of intensity of fluorescence of pyranine at 488 nm (21). The calibration curve of the fluorescence intensity of pyranine versus pH is presented in Fig. S2 and shows the linear response of pyranine between pH 6.8 and pH 7.8.

The proton passive permeability of the BR GUVs was evaluated by recording the fluorescence intensity of pyranine after submitting BR GUVs to a 1.5-pH unit gradient. BR GUVs in 125 mM K₂SO₄ and 15 mM Pipes (pH 7.5) were diluted in 125 mM K₂SO₄ and 15 mM Mes (pH 6) (proton equilibration is depicted in Fig. S2).

FhuA.

GUVs made of EPC and 0.5% (wt/wt) Tx-DHPE were grown in 75 μM DOTM in 400 mM sucrose solution supplemented with 5 μM YOPRO-1, a green fluorescent probe of DNA. A drop of 40 μL of GUVs in 75 μM DOTM was mixed with 20 μL of 200 mM NaCl, 20 Pipes (pH 7.5), and 75 μM DOTM. FhuA was added at 0.1–0.25 μM final concentration and incubated 2–16 h without stirring. After detergent removal, 5 μL of sample was diluted 10–20 times in a solution of 200 mM glucose, 10 mM Pipes (pH 7.5), 70 mM NaCl, 1 mM CaCl₂, and 1 mM MgCl₂ (osmolarity was adjusted to the same value as the sample). The function of FhuA was assayed after addition and binding of T5 phage leading to the injection of DNA and its accumulation in the internal volume of GUVs. Because T5 phages can interact with nonincorporated and aggregated FhuA, 0.01 mg/mL DNase was added in the external medium to avoid an increase of the viscosity after ejection of DNA. Confocal images were recorded 30 min after T5 phage addition.

Fusion of IMVs, Chromatophores, and Proteoliposomes with GUVs.

GUVs made of DphPC, DOPC/DOPE (1/1, mol/mol) or DOPC/Sph/chol (33/33/3, mol/mol/mol), were prepared with 75 μM DOTM in 400 mM sucrose solution. A drop of 20 μL GUVs was mixed with 20 μL of buffer at 50 mM Hepes (pH 7), 500 mM KCl, 6 mM MgCl₂, and 100–300 μM Triton X-100 or 50–150 μM DOTM. A drop of 2 μL of IMVs or of 2 μL of chromatophores was added at 1.5 and 1 mg/mL total proteins final concentration, respectively. In the case of fusion of proteoliposomes, 5 μL of proteoliposomes of red-labeled BR or of BmrC/BmrD reconstituted at lipid/protein ratio of 10 (wt/wt) was added at 1 mg/mL final concentration. It is worth noting that the amount of detergent present during the fusion process was far below the solubilization of the native vesicles or of the proteoliposomes. Moreover, GUVs were also not solubilized by the addition of detergent, because detergent partitioned between the populations of membranes, GUVs and natives membranes or proteoliposomes. Fusion was performed by incubation without stirring 2–16 h at room temperature before detergent removal.

Drug Transport by Reconstituted BmrC/BmrD in GUVs.

For the functional assay of drug transport by BmrC/BmrD, DNA was previously encapsulated in GUVs by growing GUVs in sucrose solution supplemented with 70 μg/mL calf thymus DNA (a representative image is depicted in Fig. S5). The solution was changed by mixing 20 μL of GUVs with 20 μL of buffer at 50 mM Hepes (pH 7), 500 mM KCl, 6 mM MgCl₂, and 600 μM Triton X-100. A drop of 2 μL of IMVs was added and incubated without stirring 2–16 h at room temperature. After detergent removal, transport of drugs by BmrC/BmrD was initiated by addition of 5 mM ATP and 25 μM EtBr and was inhibited by addition of 20 mM Vi. For confocal observation 5 μL of the sample was observed after 10–20× dilution in a solution of 100 mM glucose, 290 mM NaCl, 30 mM Hepes, and 3 mM MgCl₂ and osmolarity was adjusted to the same value as the sample.

Detergent Removal.

Detergent was removed either by incubation of the solutions without stirring for 2 h with 5–10 mg of Bio-Beads SM2 (i.e., in large excess to remove the amount of detergents) (11) or by adding methyl-β-cyclodextrin (MβCD) at a ratio of MβCD/detergent of 2/1 (mol/mol) (32).