Detergent-free Ultrafast Reconstitution of Membrane Proteins into Lipid Bilayers Using Fusogenic Complementary-charged Proteoliposomes.

Detergents are indispensable for delivery of membrane proteins into 30-100 nm small unilamellar vesicles, while more complex, larger model lipid bilayers are less compatible with detergents. Here we describe a strategy for bypassing this fundamental limitation using fusogenic oppositely charged liposomes bearing a membrane protein of interest. Fusion between such vesicles occurs within 5 min in a low ionic strength buffer. Positively charged fusogenic liposomes can be used as simple shuttle vectors for detergent-free delivery of membrane proteins into biomimetic target lipid bilayers, which are negatively charged. We also show how to reconstitute membrane proteins into fusogenic proteoliposomes with a fast 30-min protocol. Combining these two approaches, we demonstrate a fast assembly of an electron transport chain consisting of two membrane proteins from E. coli, a primary proton pump bo3-oxidase and F1Fo ATP synthase, in membranes of vesicles of various sizes, ranging from 0.1 to >10 microns, as well as ATP production by this chain.


Introduction
Functionalization of artificial lipid bilayers with membrane proteins is a key step in assembly of membrane model systems. The simplest model, proteoliposomes (PL), consists of small  nm diameter) unilamellar vesicles (SUV, also called liposomes), with proteins integrated into their membranes. PL are traditionally formed in two steps 1 . First, preformed SUV are mixed with a membrane protein of interest and a detergent at a concentration above its critical micelle concentration (CMC). Second, the detergent is removed with various dialysis, "bio-beads" or gel filtration techniques, leaving the protein in the membrane. The latter approach is much faster (~30 min 1 ) and is therefore preferable for reconstitution of fragile and sensitive membrane proteins, while the first two approaches are limited by detergent removal speed, which takes many hours and may cause a substantial loss of activity and loss of structural integrity of the proteins. Functionalization of larger vesicles (large unilamellar vesicles, LUV, up to 1 µm diameter) by this approach is more challenging, as vesicle size gets reduced after detergent removal, and it is not possible for giant unilamellar vesicles (GUV, >1 µm), as they are destabilized by detergents (but see Johnson et al. 2 for slow 2Dcrystallization of membrane proteins in large bilayers). Alternative approaches for GUV membrane functionalization 3,4,5 exist but are laborious, time consuming, and still require some detergent at concentrations below CMC. More complex or fragile lipid models (for example, Droplet Hydrogel Bilayers 6 and 3D printable Droplet Interface Bilayer-based artificial tissues 7 ) cannot tolerate detergents. Quickly emerging synthetic biology applications 8,9,10 critically depend on functionalization of such complex membrane structures. Therefore, an easy and robust method allowing fast and gentle delivery of membrane proteins into the target fragile bilayers is highly sought in the field.
An alternative, detergent-free protein delivery method is vesicle fusion, where interacting vesicles' membranes unite into the intact postfusion bilayer, while their intravesicular aqueous contents get mixed, without being released into the external environment. Vesicle fusion is enabled and driven either by conformational rearrangements within complementary fusogenic agents (some proteins 11,12 and peptides 13 or specially modified DNA 14 ) located in the contacting bilayers, or Coulombic interactions between lipid bilayers formed of complementarily charged cationic and anionic lipids 15,16 , or cationic bilayers and negatively charged proteins 17 .
The former approach requires the presence of fusogenic agents in the interacting membranes prior to fusion, is relatively slow (~30 min to reach half-maximum of fusion

Monitoring Vesicle Fusion with Cobalt-Calcein Method
1. Preparation of a gel-filtration gravity column.
1. Soak ~10 g of a gel-filtration resin (e.g. superfine sephadex G-50) in 100 mL of deionized water, and let swell overnight. 2. Pack 3 mL of the resin into a disposable plastic gravity flow column, wash it with ultrapure water, and equilibrate with buffer C (100 mM KCl, 10 mM MOPS, pH 7.4) at room temperature. . It becomes fluorescent again upon addition of EDTA, which having higher affinity for Co 2+ , displaces calcein from the cobalt-calcein complex ( Figure 3A). 2. Add 500 µL of this solution to a dry film of cationic or neutral lipids prepared as described in 1.1. 3. Prepare 100 nm SUV by extrusion as described in 1.2. 4. Pellet extruded SUV at 1,000,000 x g for 20 min in a table-top ultracentrifuge and resuspend in 1 mL of buffer C. Repeat pelleting and resuspension three times; use 0.6 mL of buffer C for the last resuspension. NOTE: These steps remove most of the external cobalt-calcein. 5. Remove the remaining external cobalt-calcein by passing SUV through a disposable gravity-flow column loaded with the gel-filtration resin equilibrated with buffer C as described for step 3.1.2. NOTE: We typically discard the first milliliter volume of the flow-through and collect the second milliliter, which contains external cobaltcalcein free SUV.

Run control experiments as in
Step 2 by mixing, for example, PL and SUV of the same charge (grey trace). NOTE: Proton pumping should be improved by postfusion PL.  (Figure 7B). Run the reaction for ~3 min. NOTE: ATP synthesis reaction proceeds for 5-7 min until depletion of oxygen consumed by bo 3 -oxidase and luciferase. 8. Add ATP reference standard (0.5 nmol ATP) to the reaction mixture twice. Measure ATP produced. 9. Obtain the actual amount of ATP produced in the reaction by dividing the signal from ATP synthesis reaction by the ATP reference standard signal. Adjust this value to the total amount of F 1 F o used in the reaction, and finally express the rate of ATP synthesis in µmol ATP/(mg F 1 F o *min).

Representative Results
The use of fusogenic complementary-charged proteoliposomes for fast detergent-free delivery of membrane proteins into target bilayer membranes includes three steps (Figure 1): A, formation of fusogenic SUV from lipid mixtures with high content of charged lipids; these SUV may optionally carry an intravesicular load; B, conversion of fusogenic SUV into PL using our fast membrane protein reconstitution; C, fusion of fusogenic PL with target bilayers in a low-salt medium, followed by addition of high salt to stop the fusion reaction. In case of large vesicles (D), a preferable strategy is to deliver membrane protein into the anionic target bilayer, which provides a better lipid environment for activity of membrane proteins (discussed further in the text).
The GUV formation protocol with inverted emulsion method is highlighted in detail in Figure 2. We prefer to use hexadecane in the lipid-oil mixture due to its relatively high (18 °C) freezing temperature, which allows easy removal of solidified oil after GUV pelleting. This fast protocol of reconstitution of membrane proteins into fusogenic SUV is illustrated in Figure 4A. Using this protocol, we demonstrate fast reconstitution of primary proton pump bo 3 -oxidase and F 1 F o ATP synthase into PL, and assessment of their specific activity in these membranes. It is important to mention that the yield of protein reconstitution does not depend on the charge of a lipid used 15 and is about 50 -75% 25,12 , and that after reconstitution into PL, the protein can be stored for at least three days, even at room temperature without obvious loss of protein activity. This procedure also provides unidirectional orientation of ATP synthase, where more than 95% of it has its hydrophilic moiety F 1 oriented outwards 15,26 . Fast assembly of a functioning electron transport chain in membranes of large vesicles by means of fusogenic PL is shown in Figure 7. We used F 1 F o SUV + and bo 3 SUV + for fusion with 800 nm LUV -, and demonstrated ATP production by this chain in the postfusion vesicles by sequentially adding Coenzyme Q 1 and DTT to energize membranes, and then adding phosphate to trigger ATP synthesis by F 1 F o .   (A) schematic of the experiment: 100 nm F 1 F o PL + and 100 nm bo 3 -oxidase PL + were fused with LUV -, as described in Figure 6. Fusion was stopped by adding KCl and MOPS to 100 and 50 mM, respectively. The membranes were mixed with ADP-luciferin-luciferase cocktail and energized by addition of DTT and Q 1 , as described in the text. ATP production was initiated by addition of 1 mM phosphate (P i ) and monitored real time with luciferaseluciferin system as described in the text. (B) ATP synthesis by postfusion vesicles (red trace). Control experiments showed no ATP production when PL + were mixed with LUV in high salt (grey), or PL 0 were mixed with LUV -(black). ATP synthesis rate was calculated as explained in the text (steps 8. 8 -8.9). Please click here to view a larger version of this figure.

Discussion
The following few issues need to be considered for success of this experimental approach: Choice of lipid charge for proteoliposomes and target bilayers: Cationic lipids are not found in nature, while anionic lipids are abundant in biological membranes reaching, for example, ~25, 35 and 20% in inner membrane of E. coli, plasma membrane of yeast S. cerevisiae, and inner mitochondrial membranes of many species, respectively 27,28,29 . It would be reasonable to expect that the functionality of membrane proteins in PL + may be affected by the strength of a positive charge of the bilayer, which in turn would depend on a relative content of cationic lipid in the bilayer and external ionic strength. Therefore, it is important to address experimentally to what extent functionality of membrane proteins of interest would depend on the charge of the cationic lipid environment. Here, we show that both F 1 F o ATP synthase and bo 3 -oxidase are sensitive to cationic lipid environment, but we managed to modulate and reverse this effect by first placing the proteins in PL + and delivering them into anionic accepting bilayers, and then increasing the ionic strength of the reaction medium after fusion is finished.
Choice of a particular cationic lipid: Most commercially available cationic lipids are of non-triacylglyceride nature; therefore a potential candidate lipid must be tested for compatibility with membrane proteins of interest. We found previously  30 . In case of suboptimal lipid mixtures or certain conditions, vesicles also demonstrate pronounced liquid content leakage during fusion 31 . Minimal concentrations of charged lipids in fusogenic mixtures enabling true fusion need to be found experimentally for each lipid species, but in general, it is found that membranes with less than 10% charged lipids are rendered non-fusogenic 32 .
While forming fusogenic SUV in presence of multi-valent ions, it is important not to mix oppositely charged components, as this may cause immediate clumping and aggregation of lipids by these ions. For example, while preparing SUV for the cobalt-calcein-EDTA method, it is important to avoid mixing a cationic lipid mixture with EDTA to prevent clogging of the polycarbonate filter by clumped components.
It is also important to mention that the cobalt-calcein-EDTA method, while being very sensitive and convenient for real-time fusion monitoring, may still underestimate the extent of fusion due to 1) self-quenching of fluorescence of free calcein inside postfusion vesicles, which is expected to reach 1 mM, while the self-quenching threshold is reported to be around 20 µM 33 , and 2) surface-bound cobalt-calcein, which remains bound to SUV + even after passing through the gel-filtration resin and colors vesicles to a bright orange color, while SUV 0 have a pale orange color. Also note that release of the bound cobalt-calcein upon addition of detergent Triton X-100 in the presence of EDTA generates much stronger signals for SUV + (Figure 3C, red trace) than SUV 0 (grey trace).

Perspectives:
We expect that the fast approaches described here may greatly facilitate and speed up assembly of complex membranes for the needs of emerging synthetic biology applications. Our membrane protein reconstitution protocol takes only half an hour for reconstitution of fragile large membrane proteins known to be sensitive to lengthy dialysis-based reconstitution techniques, while this fusogenic approach takes only 5 -10 min to deliver such proteins into large lipid bilayers. Here, we demonstrate the advantages of these approaches by manipulating E. coli F 1 F o ATP synthase, which is an example of a fragile protein. It is made of 23 subunits and is known to readily lose its integrity if exposed to suboptimal conditions (for example, heat) during/after solubilization, but being used in these procedures, the protein demonstrates reproducibly high activity in proton pumping.

Disclosures
The authors have nothing to disclose.