Fusion Domains Guide the Oriented Insertion of Light-Driven Proton Pumps into Liposomes

Abstract

One major objective of synthetic biology is the bottom-up assembly of minimalistic nanocells consisting of lipid or polymer vesicles as architectural scaffolds and of membrane and soluble proteins as functional elements. However, there is no reliable method to orient membrane proteins reconstituted into vesicles. Here, we introduce a simple approach to orient the insertion of the light-driven proton pump proteorhodopsin (PR) into liposomes. To this end, we engineered red or green fluorescent proteins to the N- or C-terminus of PR, respectively. The fluorescent proteins optically identified the PR constructs and guided the insertion of PR into liposomes with the unoccupied terminal end facing inward. Using the PR constructs, we generated proton gradients across the vesicle membrane along predefined directions such as are required to power (bio)chemical processes in nanocells. Our approach may be adapted to direct the insertion of other membrane proteins into vesicles.

Main Text

The assembly of molecular systems is a key element in synthetic biology and aims for the engineering of novel devices with functionalities not found in nature. Hereto, minimalistic nanocells built from lipid or polymer vesicles with integrated membrane proteins experience particular interest since they provide a broad spectrum of potential applications in biotechnology, cell biology, and medicine (1, 2). Despite varying functionalities of nanocells, the underlying engineering principles remain the same. A membrane defines the boundary between interior and exterior and forms an impermeable barrier for hydrophilic molecules. This separation of intra- and extravesicular space creates a nano-environment, which can be utilized for internal reactions (3, 4). Regulated transport of molecules to supply and energize these reactions can be achieved by embedding specific transporting and energy-converting proteins in the membrane (5, 6, 7). Utilizing membrane proteins as functional building blocks for such synthetic systems requires the ability to control their orientation in the membrane. In the living cell, the orientation of membrane proteins is determined during insertion and folding, assisted by chaperones, insertases, and translocases (8, 9). This orientation of membrane proteins is not preserved throughout purification and reconstitution into synthetic membranes (10, 11), wherein membrane proteins adopt either inward- or/and outward-facing orientations. However, to our knowledge, no suitable and easily applicable method to control this task exists. Here, we introduce an approach for the directed reconstitution of membrane proteins into synthetic vesicles.

In analogy to cellular membranes, a proton-motive force can provide synthetic systems with energy to power (bio)chemical processes. In this context, light-driven proton pumps are of particular interest (6, 7, 12). By generating a proton gradient upon illumination, such membrane proteins add an element of external control, thus allowing downstream energy-dependent processes such as uptake and release of solutes to be triggered (13). Proteorhodopsin (PR), a light-driven proton pump from proteobacteria, can be employed as an energy-converting module (14, 15). PR translocates protons from the C-terminal cytoplasmic space toward the N-terminal extracellular space upon illumination (16). Thus, in bacteria expressing PR, illumination induces proton outflux and generates a proton gradient across the cellular membrane (17). Here, we wanted to engineer PR as an energy-supplying module to convert light energy into proton gradients of predefined directionality across vesicular membranes. Thus far, the production of PR-containing proteoliposomes relies on protocols in which PR is solubilized, purified, and reconstituted. Thereby, the two possible orientations of PR cannot be controlled to direct the light-driven proton gradient across vesicular membranes (Fig. 1 A). We hence thought to establish a rationale to direct PR insertion into vesicles.

Figure 1.

Orientations and proton-pumping directions of PR reconstituted into liposomes. (A) During detergent-mediated reconstitution, solubilized wt PR (red) can insert into liposomes in two opposing directions. (B) Fusion of the C-terminus of PR to a soluble GFP-domain prevents the terminus from inserting. PR thus inserting from the N-terminal side transports protons into the vesicle. (C) Fusion of the N-terminus of PR to a soluble mCherry domain prevents the terminus from inserting. PR inserting from the C-terminal side transports protons out of the vesicle.

Insertion of membrane proteins with a preferred directionality due to steric effects has previously been reported for proteins that natively comprise bulky domains (18). It has been observed also that membrane proteins tend to insert with their most hydrophobic domain facing the liposomal lumen (19). To control the proton pumping of PR, we wanted to exploit these effects and expected that the hydrophilic properties of soluble protein domains fused to PR would only allow the unoccupied terminal end to traverse the hydrophobic core of the membrane of preformed liposomes (Fig. 1, B and C). Based on this consideration we engineered two different PR constructs to guide their oriented insertion: PR with a C-terminal green fluorescent protein (PR-GFP) and PR with an N-terminal red fluorescent protein (mCherry-PR). The use of green and red fluorescent proteins as soluble fusion domains allowed optical identification of the constructs. Furthermore, we designed the PR constructs so that they would not interfere with the native orientation of PR in the inner membrane of Escherichia coli, which we used for overexpression. PR-GFP was designed to carry GFP in the cytoplasm by adding GFP to the C-terminal end of PR connected by a flexible polypeptide linker (Supporting Material, Sequence S1). To locate mCherry in the periplasm of E. coli, mCherry-PR was designed by replacing the native N-terminal signal sequence of PR with a signal sequence of the periplasmic Skp protein, the mCherry sequence, and a short polypeptide linker (Supporting Material, Sequence S2).

Variations of both PR constructs were expressed in E. coli and colonies were selected for high-level overexpression based on their color and fluorescence intensity (Fig. S1). To assess the proton-pumping activity of the PR constructs, we evaluated their activity in E. coli using a photoactivity assay (Fig. 2 A) (15). Illumination of bacteria overexpressing the PR constructs caused a pH drop in the unbuffered solution, thus indicating that both PR constructs translocated protons from the cytosol to the extracellular solution (Fig. 2 B). Control E. coli not expressing PR did not show such an effect. These results showed that PRs from both constructs inserted in the same direction in the bacterial membrane (Fig. 2 C).

Figure 2.

Photoactivity measurements of E. coli. (A) Experimental setup. The pH of the sample is recorded with a micro pH electrode while being stirred, temperature controlled, and protected from external light. An internal light source activates proton pumping. (B) Measurements of E. coli overexpressing either mCherry-PR (red lines) or PR-GFP (green lines). Gray lines are control measurements performed with E. coli containing an empty pET21a(+) plasmid. Yellow areas indicate periods of illumination and light gray areas indicate dark periods. (C) Orientation of the two PR constructs in the E. coli inner membrane. Both constructs translocate protons to the periplasm.

Once the expression was optimized for both PR constructs, they were purified (Materials and Methods). The yield of purified PR-GFP reached ∼2.5 mg protein per gram of cell pellet (wet weight), whereas that of purified mCherry-PR reached ∼0.5 mg per gram of cell pellet. This variation in expression level explained the differing magnitudes of the pH change detected in the photoactivity measurements of E. coli (Fig. 2). The purity of the overexpressed PR constructs was evaluated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) (Fig. 3 A). Wild-type (wt) PR migrated at ∼22 kDa, PR-GFP at ∼39 kDa, and mCherry-PR at ∼41 kDa. Absorption spectra of both purified PR constructs were recorded (Fig. 3, B and C) and compared to that of purified wt PR, GFP, and mCherry. The spectra of both PR constructs appeared as a convolution of the absorption spectra of the individual proteins from which they were engineered, with coinciding peak positions. The presence of the GFP and mCherry absorption peaks of the PR constructs demonstrated that detergent treatment during purification did not affect the soluble proteins.

Figure 3.

Analytical description of PR constructs. (A) Gradient (4–12%) SDS-PAGE gel of wt PR in lane 1, PR-GFP in lane 2, and mCherry-PR in lane 3. (B) Absorption spectra of wt PR (red), GFP (green), and PR-GFP (black). (C) Absorption spectra of wt PR (red), mCherry (purple), and mCherry-PR (black).

Next, we reconstituted both PR constructs into preformed liposomes under identical conditions (Materials and Methods). Cryo-electron microscopy (cryo-EM) showed mostly unilamellar proteoliposomes with average diameters of ∼100 nm (Fig. 4 A). To assess the orientation of insertion and the functionality of the reconstituted PR constructs, we evaluated their activity in proteoliposomes using the photoactivity assay (Fig. 4 B). The assay monitored the extravesicular change in pH over time, similar to the above measurements performed with E. coli overexpressing the constructs. Each sample underwent multiple light-dark cycles to periodically activate and inactivate proton pumping of the reconstituted PR. Upon illumination, proteoliposomes containing PR-GFP increased the extravesicular pH, which recovered in the dark phase. In contrast, illumination of proteoliposomes containing mCherry-PR decreased the extravesicular pH, which recovered in the dark phase. Proteoliposomes reconstituted with either of the two PR constructs showed reproducible behavior over several light-dark cycles. Control measurements performed with proteoliposomes prepared under identical conditions with wt PR did not show significant pH changes. The photoactivity assay thus demonstrated that proteoliposomes containing the respective PR constructs translocated protons across the vesicular membrane in opposing directions. To evaluate the ratio of PRs in the predefined orientation, we followed an established proteolytic digestion assay with the nonspecific serine protease proteinase K (Fig. 4 C) (20, 21). The fluorescent moieties of the PR constructs are only accessible to proteinase K in the extravesicular space, whereas the lipid bilayer acts as a diffusion barrier and prevents inward-facing moieties from digestion by proteinase K (22, 23). The analysis of the fluorescence before and after digestion on an SDS-PAGE gel revealed that a vast majority of proteins were inserted in the predefined orientation. We can thus conclude that the PR constructs directed the insertion of PR into liposomes.

Figure 4.

Analysis of proteoliposomes. (A) Cryo-EM of PR-GFP. Scale bars, 100 nm. (B) Photoactivity measurements of proteoliposomes reconstituted with either mCherry-PR or PR-GFP. Yellow areas indicate periods of illumination and light gray areas indicate dark periods. Green lines show the proton-translocation activity of proteoliposomes containing PR-GFP and red lines show the activity of proteoliposomes containing mCherry-PR. Gray lines show controls of wt PR proteoliposomes. Each line represents an individually prepared sample. (C) Fluorescence analysis of an SDS-PAGE gel of PR-GFP proteoliposomes before (lane 1) and after proteinase treatment (lane 2) and mCherry-PR proteoliposomes before (lane 3) and after proteinase treatment (lane 4).

Previous studies controlled the directionality of proton-pumping either by altering the lipid composition of the vesicles (21) or by chemically deactivating one orientation of PR reconstituted into vesicles (15). However, such efforts are limited to the usage of specific lipids or, in the latter example, to one particular pumping direction.

Here, we introduced an approach to control the directionality of membrane protein insertion by exploiting the repulsive properties of membranes against hydrophilic proteins. Thereby, the position of a soluble protein fused to a membrane protein determined how it inserted into preformed vesicles. This control of directed membrane protein insertion represents a crucial step toward engineering nanocellular systems of higher complexity. We believe that our approach of fusion-protein-controlled directed insertion of membrane proteins can be readily adapted for many other membrane proteins, such as are needed to functionally equip nanocells and to drive their (bio)chemical reactions. En route toward engineering of multi-modular systems it is conceivable to expand our method to the fusion of two or more membrane proteins for their directed insertion into synthetic vesicles and nanocells. The use of such constructs would not only provide control over the absolute and relative orientation of the membrane proteins but also over their stoichiometry.

Materials and methods

Cloning

PR variants were engineered based on green-light-absorbing PR (GenBank: AY601905.1) from plasmid pZUDF-rbs-PR-GGS-3C10H (15). Plasmid pNR03, harboring PR fused with GFP (PR-GFP) was assembled by adding the gene coding for sf-GFP to the C-terminal end of PR, separated by a flexible linker sequence. The resulting sequence (Supporting Material, Sequence S1) was subcloned into a pET-21a(+) plasmid between the NdeI and HindIII restriction sites. Plasmid pNR09, harboring PR with mCherry linked to its N-terminal end (PR-mCherry) was assembled by replacing the coding sequence for PR’s native signal sequence with the signal sequence of the periplasmic protein Skp from E. coli MG1655 followed by the gene coding for mCherry and a short linker. The resulting sequence (Supporting Material, Sequence S2) was subcloned into a pET-21a(+) plasmid between the NdeI and NotI restriction sites. Both PR constructs contain a His₆-tag from the pET21a(+) backbone on the C-terminal end.

Overexpression of PR-constructs

pNR03 and pNR09 were transformed into E. coli Lemo21(DE3) cells. Colonies were selected for high-level expression from small-scale test expressions based on their color intensity and the fluorescence intensity of GFP or mCherry, respectively. Four liters of Luria-Bertani liquid cultures (100 μg mL⁻¹ ampicillin and 36 μg mL⁻¹ chloramphenicol) were inoculated 1:100 from overnight cultures. For PR-GFP expression, cells were grown under vigorous shaking at 30°C to OD₆₀₀ ∼0.5, 5 μM all-trans-retinal was added, and expression was induced with addition of 0.1 mM isopropyl-β-D-thiogalactopyranoside. For mCherry-PR expression, media were supplemented with 150 μM L-rhamnose. Cells were grown under vigorous shaking at 30°C to OD₆₀₀ ∼ 0.5, 5 μM all-trans-retinal was added, and expression was induced with addition of 0.4 mM isopropyl-β-D-thiogalactopyranoside. After induction, cells were incubated for 4 h at 30°C. The cells were harvested by centrifugation (5000 × g for 12 min at 4°C), resuspended in lysis buffer (20 mM Tris-HCl (pH 7.4), 100 mM NaCl, and 1 mM Tris(2-carboxyethyl)phosphine (TCEP)), and stored at −20°C until further use.

Photoactivity measurements of E. coli overexpressing PR constructs

Photoactivity measurements were performed as described (15). One hundred milliliters of Luria-Bertani liquid cultures containing overexpressing E. coli were prepared as described above. Cells were washed twice with 10 mL 150 mM NaCl (pH 7.4) followed by centrifugation (3200 × g for 10 min at 4°C). Immediately before the photoactivity measurements, another washing step was performed and the concentration of the cells was adjusted to OD₆₀₀ = 40. Eight hundred microliters of the sample was used to measure the light-driven proton translocation activity of the PR constructs (Fig. 1 B). The activity was monitored by recording the pH in the unbuffered extracellular solution using a micro pH-electrode with an integrated temperature sensor (InLab Micro Pro, Mettler Toledo, Columbus, OH). During the measurement, the sample was stirred and the temperature was kept constant at 18°C using a cooling water bath (setup shown in Fig. 1 A). The sample was illuminated by a 2 W, warm white (3000 K) LED lamp (JANSÖ; IKEA, Delft, the Netherlands) for 8 min during four consecutive light-dark cycles. After each period of illumination, the sample was kept in the dark for 8 min to recover. To prevent background illumination, the whole setup was guarded from light. The pH and the temperature were recorded at intervals of 30 s. The pH drift was corrected by subtracting a piecewise linear function from the raw data (Fig. S2). The slope of the function was defined for each peak by two sequential starting points of illumination cycles, as described before (15).

Purification of PR constructs

Cell pellets were thawed, DNaseI from bovine pancreas (Roche Diagnostics, Basel, Switzerland), lysozyme from hen egg white (Fluka Analytical, Honeywell, Mexico City, Mexico), and cOmplete EDTA-free protease inhibitor (Roche Diagnostics, Indianapolis, IN) were added. Cells were lysed by sonication with a Branson digital sonifier with a total pulse time of 25 min (25% amplitude). Unbroken cells were removed by centrifugation (3200 × g for 10 min at 4°C). The membrane fraction was collected by centrifugation (75,000 × g for 1 h at 4°C), resuspended in lysis buffer, and homogenized. The centrifugation step was repeated and the pellet was resuspended and homogenized in 6 mL membrane storage buffer (20 mM Tris-HCl (pH 7.4), 100 mM NaCl, and 10% (v/v) glycerol). The sample was stored at −80°C in aliquots of 1 mL until further use.

For purification, one aliquot was thawed and solubilized in 10 mL solubilization buffer (20 mM Tris-HCl (pH 7.4), 300 mM NaCl, 10% glycerol, 20 mM imidazole, 1 mM TCEP, and 3% (w/v) n-Octyl-β-D-glucoside (OG; Anatrace, Maumee, OH) (pH 7.4)) overnight on a roller shaker at 4°C. Unsolubilized material was removed by centrifugation (75,000 × g for 20 min at 4°C). One milliliter of Ni-NTA agarose resin (Protino, Macherey-Nagel, Dueren, Germany) was added to the solubilized membrane together with 5 mL binding buffer (20 mM Tris-HCl (pH 7.4), 300 mM NaCl, 10% glycerol, 30 mM imidazole, 1 mM TCEP, and 3% (w/v) OG (pH 7.4)) and incubated with rolling for 3 h at 4°C. The sample was washed twice with 10 mL wash buffer (20 mM Tris-HCl (pH 7.4), 300 mM NaCl, 10% glycerol, 10 mM imidazole, 1 mM TCEP, and 1% (w/v) OG (pH 7.4)). The protein was eluted by stepwise addition of elution buffer (20 mM Tris-HCl (pH 7.4), 150 mM NaCl, 10% glycerol, 400 mM imidazole, 1 mM TCEP, and 1% (w/v) OG (pH 7.4)). Elution fractions with strong red color intensity were pooled. Protein concentrations were determined by measuring their absorption at 280 nm (NanoDrop 2000c; Thermo Scientific, Waltham, MA). Molecular weight and extinction coefficients were predicted for each protein, based on the amino acid sequence, using the Expasy ProtParam tool (http://www.expasy.ch/tools/protparam.html). For PR-GFP, a molecular weight of 58.8 kDa and extinction coefficient of 96,510 M⁻¹ cm⁻¹ were predicted, and for mCherry-PR a molecular weight of 57.4 kDa and extinction coefficient of 115,865 M⁻¹ cm⁻¹. Purified protein was stored at 4°C in the dark.

Reconstitution of PR constructs into liposomes

Ten milligrams 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC; Avanti Polar Lipids, Alabaster, AL) in 1 mL chloroform was dried in a 25 mL round-bottom Duran glass flask (Lenz, Dayton OH) under a nitrogen stream. Residual chloroform was removed by applying vacuum overnight at room temperature. The lipid film was hydrated with 2 mL hydration buffer (20 mM KP_i (pH 7.2), 100 mM KCl, and 1 mM TCEP), and liposomes were generated by shaking at 700 rpm for 1 h. Liposomes were destabilized by addition of 0.75% (w/v) OG under shaking for 1 h. The liposomes were extruded (LipX Liposome Extruder, Tools and Technologies, Porto Alegre, Brazil) through a pore diameter of 200 nm by 19 passes. 0.5 mM CaCl₂ was added to the liposomes, followed by shaking for 30 min at 700 rpm. The protein was added to the preformed liposomes and the concentration was adjusted to a final lipid/protein ratio of 11.5 (w/w), which results in ∼80 proteins per proteoliposome, assuming a liposome diameter of 100 nm and a molar lipid/protein ratio of ∼1000. The amount of detergent added with the protein increased the total OG concentration to 0.8%. The sample was shaken for 3 h at 700 rpm. The sample was transferred to a dialysis tube (14 kDa molecular weight cutoff; Visking dialysis tubing, Medicell Membranes, London, United Kingdom) and dialyzed against 2 L of dialysis buffer (20 mM Tris-HCl (pH 7.5) and 0.5 mM CaCl₂) overnight at room temperature.

Cryo-EM

A 3 μL drop of PR-GFP proteoliposomes was deposited on a Lacey carbon grid (Cu 300 mesh; Agar Scientific, Stansted, United Kingdom). The grid was blotted on both sides for 4 s in a Vitrobot (FEI, Hillsboro, OR) at 100% humidity and 4°C, and frozen rapidly by plunging into liquid ethane. Images were collected at liquid nitrogen temperature on a Tecnai F20 electron microscope (FEI) operated at 200 kV equipped with an FEI Falcon 3 direct electron detector at a magnification of 50,000× (step size of 2.08 Å/pixel at the specimen level) and at a defocus value of −5.0 μm (overview in Fig. 4 A) and −2.5 μm (Fig. 4 A, inset). The exposure time was 2 s, resulting in a total electron dose of ∼29 e⁻/Ų.

Photoactivity measurements of proteoliposomes

Proteoliposomes were washed twice with 800 μL 150 mM NaCl (pH 7.4), followed by centrifugation (200,000 × g for 20 min at 4°C), as described (15). Another washing step was performed immediately before measurement. The photoactivity of the proteoliposomes was monitored in a total volume of 800 μL. Measurements were performed as described for the bacteria, but for longer light-dark periods of 15 min each.

Limited proteolysis

Proteinase K (800 U/mL; New England Biolabs, Ipswich, MA) was used to digest exposed moieties of reconstituted PR-constructs. Limited proteolysis was performed as described (21). Proteinase K was added to proteoliposomes to a final concentration of 2.5 mg/mL. Samples were incubated at 37°C for 2 h. Subsequently, phenylmethanesulfonyl fluoride (0.2 M in ethanol; AppliChem, Darmstadt, Germany) was added to a final concentration of 10 mM and the samples were cooled on ice for 30 min to inhibit protease activity. The samples were loaded on a 4–12% SDS-PAGE gel and analyzed with a Gel Doc XR+ Imager and the Image Lab 4.1 software (Bio-Rad, Hercules, CA) using ultraviolet transillumination and an exposure time of 20 s.

Author Contributions

All authors designed and discussed the experimental approach. N.R. and J.T. cloned, expressed, and purified the PR constructs. N.R., J.T., S.H., and D.F. reconstituted the PR constructs. N.R. and S.H. characterized the light-driven proton-pumping activity. D.K., S.H., and D.F. recorded electron microscopy images. All authors discussed and wrote the manuscript.

Editor: Lukas Tamm.