High level cell-free expression and specific labeling of integral membrane proteins


F. Bernhard, Centre for Biomolecular Magnetic Resonance, University of Frankfurt/Main, Institute for Biophysical Chemistry, Marie-Curie-Str. 9, D-60439 Frankfurt/Main, Germany.
Fax: + 49 69 798 29632, Tel.: + 49 69 798 29620,
E-mail: fbern@bpc.uni-frankfurt.de


We demonstrate the high level expression of integral membrane proteins (IMPs) in a cell-free coupled transcription/translation system using a modified Escherichia coli S30 extract preparation and an optimized protocol. The expression of the E. coli small multidrug transporters EmrE and SugE containing four transmembrane segments (TMS), the multidrug transporter TehA with 10 putative TMS, and the cysteine transporter YfiK with six putative TMS, were analysed. All IMPs were produced at high levels yielding up to 2.7 mg of protein per mL of reaction volume. Whilst the vast majority of the synthesized IMPs were precipitated in the reaction mixture, the expression of a fluorescent EmrE-sgGFP fusion construct showed evidence that a small part of the synthesized protein ‘remained soluble and this amount could be significantly increased by the addition of E. coli lipids into the cell-free reaction. Alternatively, the majority of the precipitated IMPs could be solubilized in detergent micelles, and modifications to the solubilization procedures yielded proteins that were almost pure. The folding induced by formation of the proposed α-helical secondary structures of the IMPs after solubilization in various micelles was monitored by CD spectroscopy. Furthermore, the reconstitution of EmrE, SugE and TehA into proteoliposomes was demonstrated by freeze-fracture electron microscopy, and the function of EmrE was additionally analysed by the specific transport of ethidium. The cell-free expression technique allowed efficient amino acid specific labeling of the IMPs with 15N isotopes, and the recording of solution NMR spectra of the solubilized EmrE, SugE and YfiK proteins further indicated a correctly folded conformation of the proteins.




critical micellar concentrations








free induction decay


feeding mixture


green fluorescent protein


heteronuclear single quantum correlation


integral membrane protein




magic angle spinning nuclear magnetic resonance




nondetergent sulfobetaines






reaction mixture


super-glow green fluorescent protein


transmembrane segment




transverse relaxation optimized spectroscopy

Integral membrane proteins (IMPs) account for 20–25% of all open reading frames in fully sequenced genomes, and in bacteria half of all IMPs are estimated to function as transporters. The active efflux of antibiotics caused by multidrug transporter proteins results in the development of clinical resistance to antimicrobial agents and represents an increasing problem in the treatment of bacterial infections. Despite their importance, no high-resolution structure has been determined thus far from any secondary transporter, from either eukaryotic sources or from the bacterial inner membrane. This is due mainly to the tremendous difficulties generally encountered during the preparation of these multispan integral IMPs to the required purity and amounts [1]. Only some 20 IMPs have been overexpressed in Escherichia coli at a level of at least 1 mg·L−1 of culture [2,3]. Problems encountered by using conventional in vivo systems, such as toxicity of the overproduced protein upon insertion into the cytoplasmic membrane, poor growth of overexpressing strains and the proteolytic degradation of the proteins, could easily be eliminated by cell-free expression. Our primary goal was therefore to analyse whether these restrictions could be solved by the production of IMPs in a cell-free expression system. We have analyzed the efficiency of IMP production in a T7 based cell-free approach using an E. coli S30 cell extract in a coupled transcription/translation system [4,5]. During incubation the reaction mixture, containing all enzymes and high molecular mass compounds necessary for gene expression, was dialyzed against a low molecular mass substrate solution providing precursors to extend the protein synthesis for more than 10 h [6,7]. Essential components of the cell-free system such as the bacterial S30 extract preparation, the energy system, the concentrations of precursors and of beneficial additives, have been optimized to yield up to 5 mg of recombinant protein per mL of reaction during a 12 h incubation.

For our expression studies we have chosen secondary transporter proteins from E. coli belonging to the families; small multidrug resistance (EmrE, SugE), TDT (TehA) and RhtB (YfiK) [8,9]. The small multidrug resistance (SMR) transporters are typically 110 amino acids in length and they are supposed to consist of four transmembrane segments (TMS) forming a tightly packed four-helix bundle [8–10]. EmrE is a polyspecific antiporter that exchanges hydrogen ions with aromatic toxic cations [11]. Its molecular transport mechanism, and probably also that of the homologous protein SugE, is an electrogenic drug/proton antiport. EmrE is thought to form homooligomeric complexes and specifically transports aromatic dyes, quaternary amines and tetraphenylphosphonium (TPP+) derivatives [8,11], whilst SugE is presumably only specific for quaternary ammonium compounds [12]. The 36 kDa transporter TehA contains 10 TMS and is responsible for potassium tellurite efflux [13]. Overexpression of TehA further increases the resistance against monovalent cations such as tetraphenylarsonium and ethidium bromide and it decreases the resistance against divalent cations like dequalinium and methyl viologen [13]. A region including TMS 2 to 5, and homologous to proteins of the SMR family, might be primarily responsible for the activity of TehA. YfiK is a 22 kDa transporter with six putative TMS and part of a putative cysteine efflux system [14,15].

Large amounts of pure detergent solubilized IMPs are needed for biochemical characterization or even structural analysis by X-ray crystallography and NMR spectroscopy. This work is the first report of the fast cell-free production of milligram amounts of four different integral transporter proteins, three of which have been amino acid specifically labeled. Whilst a small part of the overproduced proteins could be stabilized post-translationally by the addition of lipids into the cell-free reaction, the precipitated major part of the IMPs could be folded efficiently and solubilized by various detergents. The structural reconstitution of EmrE, SugE, YfiK and TehA was demonstrated by CD spectroscopy, freeze fracturing electron microscopy, NMR spectroscopy and by functional assays.

Experimental procedures

Strains, plasmids, oligonucleotides and DNA techniques

Strains and plasmids used in this study are listed in Table 1. Standard DNA techniques were performed as described elsewhere [17]. The coding sequences for the E. coli EmrE, SugE, TehA and YfiK proteins were amplified by standard PCR using the corresponding oligonucleotide primers from MWG-Biotech (Ebersberg, Germany) (Table 2), Vent polymerase (New England Biolabs, Frankfurt/Main, Germany) and chromosomal DNA from strain C600 as a template. The purified amplified DNA fragments were cloned with the enzymes NdeI and HindIII (New England Biolabs) into the expression vector pET21a(+) resulting in the plasmids pET-emrE, pET-sugE, pET-tehA and pET-yfiK. Expression from these plasmids produced the wild type proteins without any modifications or additional tags.

Table 1. Bacterial strains and plasmids used in this study.
Strains and plasmidsRelevant genotypeReference
  • a 

    E. coli Genetic Stock Center.

BL21 (DE3) StarE. coli B ompT rne131Novagen
C600thr-1 leuB6 thi-1 lacY1 glnV44 rfbD1CGSCa
XL1-BluerecA1 lac[F'Tn10 (Tetr) lacIq lacZM15][16]
A19rna19 gdhA2 his95 relA1 spoT1 metB1CGSCa
pET21a(+)T7 promoter AprNovagen
pQB1-T7-gfpsuper glow gfp, AprQBiogene
pQB1-emrE-gfpemrE NheI in pQB1this study
pET-gfpApr, gfpRoche
pET-emrEemrE NdeI-HindIII in pET21a(+)this study
pET-sugEsugE NdeI-HindIII in pET21a(+)this study
pET-tehAtehA NdeI-HindIII in pET21a(+)this study
pET-yfiKyfiK NdeI-HindIII in pET21a(+)this study
Table 2. Oligonucleotides used in this study.
SugE-upNdcgg cat atg tcc tgg att atc tta gtt att gc
SugE-lowgga aag ctt tta gtg agt gct gag ttt cag acc
EmrE-upNdcgg cat atg aac cct tat att tat ctt ggt ggt gc
EmrE-lowcgg aag ctt tta atg tgg tgt gct tcg tga c
TehA-upcgg cat atg cag agc gat aaa gtg ctc aat ttg
TehA-lowcgg aag ctt tta ttc ttt gtc ctc tgc ttt cat taa aac
YfiK-upcgg cat atg aca ccg acc ctt tta agt gct ttt tgg
YfiK-lowcgg aag ctt tta ata gaa aat gcg tac cgc gca ata gac
EmrE-upNhcgg gct agc aac cct tat att tat ctt ggt gg
EmrE-lowNhcgg gct agc atg tgg tgt gct tcg tga c
SugE-upNhcgg gct agc tcc tgg att atc tta gtt att gc
SugE-lowNhgga gct agc gtg agt gct gag ttt cag acc

In vitro expression of proteins

Bacterial cell-free extracts were prepared from the E. coli strain A19 (E. coli Genetic Stock Center CGSC) in a procedure modified after Zubay [18]. The cells were washed in washing buffer [10 mm Tris-acetate, pH 8.2, 14 mm Mg(OAc)2], with 6 mm 2-mercaptoethanol and 0.6 mm KCl. The lysis buffer was the washing buffer supplemented with 1 mm dithiothreitol and 0.1 mm phenylmethanesulfonyl fluoride. The extract was dialysed in washing buffer supplemented with 0.5 mm dithiothreitol and 0.6 mm KOAc. Endogenous mRNA was removed from the ribosomes by incubation of the extract with 400 mm NaCl at 42 °C for 45 min. Aliquots of the cell-free extract were frozen in liquid nitrogen and stored at −80 °C. The cell-free expression was performed in the continuous exchange mode using a membrane with a cutoff of 15 kDa to separate the reaction mixture (RM) containing ribosomes and all enzymes, from the feeding mixture (FM) providing the low molecular mass precursors. The ratio of RM/FM was 1 : 17 (v/v). Reactions in the analytical scale of 70 µL RM were performed in microdialysers (Spectrum Laboratories Inc., Breda, the Netherlands), and larger dispodialysers (Spectrum Laboratories Inc.) were used for preparative scale reactions with RM volumes of 500 µL to 1 mL. The reactions were incubated at 30 °C in a suitable shaker for 20 h. The protocol for the cell-free reaction mixtures is given in Table 3. Amino acid concentrations were adjusted with regard to the amino acid composition of the overproduced proteins. The least abundant amino acids (present at ≤ 3% in the protein) were added at 1.25 mm, medium abundant (between 3 and ≤ 8%) at 1.8 mm and highly abundant (more than 8%) at 2.5 mm final concentration. Amino acid specific labeling was achieved by replacing the corresponding amino acids by their isotopically labeled derivatives.

Table 3. Protocol for cell-free protein expression. Amino acids were adjusted according to the composition of the expressed protein. RM, reaction mixture; FM, feeding mixture.
ComponentFinal concentration in RMFinal concentration in FM
  • a

    Amersham Biosciences.

  • b

    b Roche Diagnostics.

Tris-acetate, pH 8.23.5 mm3.5 mm
plasmid DNA15 µg·mL
RNasina0.3 U·µL−1
T7-RNA polymerase3 U·µL−1
E. coli tRNAb500 µg·mL
pyruvate kinase40 µg·mL
amino acids0.5–1 mm1–1.5 mm
acetyl phosphate20 mm20 mm
phosphoenol pyruvate20 mm20 mm
ATP1.2 mm1.2 mm
CTP0.8 mm0.8 mm
GTP0.8 mm0.8 mm
UTP0.8 mm0.8 mm
1.4-dithiothreitol2 mm2 mm
folinic acid0.2 mm0.2 mm
complete protease inhibitorb1 tablet per 10 mL1 tablet per 10 mL
Hepes-KOH pH 8.0100 mm100 mm
EDTA2.8 mm2.8 mm
magnesium acetate13 mm13 mm
potassium acetate290 mm290 mm
polyethylenglycol 80002%2%
sodium azide0.05%0.05%

Detergent solubilization of precipitated IMPs

The pellets of cell-free reaction containing the IMPs were suspended in three volumes of washing buffer (15 mm sodium phosphate, pH 6.8, 10 mm dithiothreitol) and centrifuged for 5 min at 5000 g. The washing step was repeated twice. For the reconstitution of proteoliposomes, EmrE was dissolved in one volume of 2%n-dodecyl-β-d-maltoside (DDM) in 15 mm Tris/HCl, pH 6.5, and 2 mm dithiothreitol. The mixture was sonified for 1 min in a water bath and then incubated for 1 h at 75 °C. Non dissolved protein was removed by centrifugation at 20 000 g at 15 °C for 5 min. TehA and SugE were additionally washed in 3%n-octyl-β-glucopyranoside (β-OG) in 15 mm sodium phosphate, pH 6.8, 2 mm dithiothreitol for 1 h at 40 °C. YfiK was first washed in 1%n-nonyl-β-maltoside (NM) in 25 mm sodium phosphate, pH 7.0, 5 mm dithiothreitol for 1 h at 40 °C. Impurities were removed by centrifugation and the pellet was further washed with 1% dodecyl-phosphocholine (DPC) at the previous conditions. Dissolved impurities were removed by centrifugation at 20 000 g for 5 min. The pellets were then dissolved with various concentrations of DDM, DPC, 1-myristoyl-2-hydroxy-sn-glycero-3-[phospho-rac-(1-glycerol)] (MHPG) or SDS if appropriate. β-OG and SDS were from Sigma, DDM, DPC, NM and MHPG were from Avanti Polar Lipids (Alabaster, AL).

Protein analysis

Protein production was analyzed by SDS/PAGE in 17.5% (v/v) Tricine gels [19]. The proteins were silver stained or visualized with Coomassie-Blue (Sigma) as described [17]. Dissolved proteins were quantified according to their specific molar extinction coefficient by measuring the UV absorbance at 280 nm in 6 m guanidine hydrochloride, pH 6.5.

Circular dichroism spectroscopy

Circular dichroism (CD) spectrometry of IMPs dissolved in 15 mm sodium phosphate, pH 6.8, 2 mm dithiothreitol, and containing the appropriate detergents was performed with a Jasco J-810 spectropolarimeter (Jasco Labortechnik, Gross-Umstadt, Germany). Assays were carried out at standard sensitivity with a band width of 3 nm and a response of 1 s. The data pitch was 0.2 nm and the scanning rate 50 nm·min−1. The spectra were recorded from 188 to 260 nm. The presented data are the average of three scans and smoothed by means-movement with a convolution width of 15. The α-helical content of the analyzed proteins was then calculated by the Jasco secondary structure estimation software. In addition, the α-helical content of proteins was calculated according to their primary structure with the predict protein server at http://cubic.bioc.columbia.edu/pp/[20].

Reconstitution of proteoliposomes

The protein concentration of membrane proteins solubilized in 1% DDM was determined by UV measurement at 280 nm in 6 m guanidine hydrochloride, pH 6.5, according to their molar extinction coefficients. Approximately 200 µm of the individual protein samples were used for the reconstitution, and E. coli lipids were added at a molar ratio of protein :lipid of 1 : 500. The solutions were then adjusted to 150 mm NH4Cl and incubated at 40 °C for 30 min. Washed biobeads SM-2 (Bio-Rad), presaturated with E. coli lipids were then added in 10-fold excess to the detergent, and the mixture was incubated overnight at 30 °C on a shaker. The biobeads were exchanged twice. The supernatant was then removed, sonified for 1 min in a water bath sonicator, and assayed immediately or stored in liquid nitrogen.

Freeze-fracture electron microscopy

Droplets of the vesicle suspension were placed between two copper blades used as sample holders and then frozen by plunging into liquid ethane cooled to −180 °C by liquid nitrogen. Freeze-fracturing was performed in a Balzers 400T freeze-fracture apparatus (Balzers, Lichtenstein) with the specimen stage at −160 °C. Platinum/carbon shadowing was at 45° (with respect to the specimen stage) whereas pure carbon was evaporated at 90° onto the sample. After thoroughly cleaning the metal replicas in chromosulfuric acid, they were placed on copper grids and analyzed in an EM208S electron microscope (Philips, Eindhoven, the Netherlands).

Ethidium transport by EmrE proteoliposomes

Transport of ethidium bromide into reconstituted EmrE proteoliposomes was carried out as described [11]. Unilamelar vesicles were prepared by extrusion using 400 nm micropore filters. Fluorescence was measured at excitation and emission wavelengths of 545 and 610 nm, respectively, with a band width of 2.5 nm and a data pitch of 0.1 s. Ten microliters of proteoliposomes (approximately 140 nm EmrE) in 15 mm Tris/HCl, pH 6.5; 2 mm dithiothreitol, 150 mm NH4Cl and 20 µg·mL−1 circular plasmid DNA (pUC18) were suspended in 980 µL of outside buffer (15 mm Tris/HCl, pH 8.5; 2 mm dithiothreitol; 150 mm KCl) and measured immediately. If appropriate, ligands were added at the following final concentrations: tetraphenylphosphonium (TPP; 50 µm), ethidium bromide (2.5 µm) and nigericine (5 µg·mL−1) (Sigma). Green fluorescent protein (GFP) fluorescence was measured at excitation and emission wavelengths of 395 and 509 nm, and at 474 and 509 nm for the red shifted mutant superglow (sgGFP).

NMR spectroscopy

Two dimensional1H,15N correlated spectra of [98%15N]Gly,[98%15N]Ala labeled samples of 0.1 mm EmrE and 0.5 mm SugE in CDCl3/CD3OH/H2O (6 : 6 : 1, v/v/v) with 200 mm ammonium acetate (pH 6.2) and 10 mm dithiothreitol, and of 0.3 mm YfiK in 4% MHPG (v/v) in 25 mm sodium phosphate (pH 7.0) and 5 mm dithiothreitol were obtained with a gradient-sensitivity enhanced [15N,1H]-transverse relaxation optimized spectroscopy (TROSY) pulse sequence [21,22]. The spectra of EmrE (T = 15 °C) and YfiK (T = 30 °C) were recorded on a Bruker DRX600 spectrometer (Bruker BioSpin GmbH, Karlsruhe, Germany) equipped with a 1H{13C,15N} triple-resonance cryoprobe with z-gradient accessory. Acquisition times were adjusted to 140 ms in both dimensions for EmrE. Accumulation of four scans per free induction decay (FID) resulted in a measurement time of 1 h. The spectrum of YfiK resulted from 200 × 768 time-domain data points corresponding to acquisition times of 55 and 53 ms in the 15N and 1H dimensions, respectively. The total recording time was 16 h using 128 scans per FID. The spectrum of SugE was taken at a Bruker DMX500 spectrometer using a xyz-gradient 1H{13C,15N} triple-resonance probe at 15 °C. Acquisition times were 102 ms in both dimensions. Thirty-two transients were recorded for each FID, giving rise to a measurement time of 6 h.


Cell-free expression of integral transporter proteins

The cell-free reaction conditions were first optimized in order to obtain high yields of protein production by titration of each component and by using the expression of green fluorescent protein (GFP) as a monitor. The most critical parameters appeared to be the concentrations of potassium, magnesium and amino acids, and the quality of the prepared S30 extract. The energy regenerating system was most efficient if a combination of phosphoenol pyruvate, acetyl phosphate and pyruvate kinase was used. With the final protocol (Table 3) we received approximately 3 mg of soluble and fluorescent GFP per mL of reaction mixture and almost 80% of the protein was synthesized during the first 7 h of incubation (Fig. 1). Identical reaction conditions were then subsequently used for the expression of the selected IMPs with the only modification being that the amino acid concentrations of the reaction mixtures were specifically adjusted according to the composition of each target protein. The coding sequences of the genes emrE, sugE, tehA and yfiK were amplified from the E. coli genome by PCR and cloned into the expression vector pET21a(+) containing the T7 regulatory sequences. All four proteins were expressed without any modifications and in each case we obtained a high level production in our cell-free system (Fig. 2). In contrast, the conventional in vivo expression using BL21 (DE3) star cells transformed with the same plasmids yielded no expression detectable by SDS/PAGE analysis. The production rate of all four proteins in the cell-free system was estimated to be at least 1 mg IMP per mL of reaction mixture. However, most of the synthesized IMPs precipitated during the cell-free expression remained insoluble. In order to detect whether a small part of the overproduced proteins might stay soluble, we constructed a fusion of emrE to the 5′ end of the gene of the reporter protein sgGFP, resulting in the expression of an EmrE-sgGFP fusion protein. Soluble and correctly folded sgGFP protein can be monitored by its fluorescence at 509 nm and in addition to the more than 1 mg of insoluble fusion protein we could calculate an average of approximately 6 µg of soluble EmrE-sgGFP protein per mL of reaction mixture after standard cell-free expressions.

Figure 1.

Protein production kinetics in the cell-free system. Soluble GFP production in a standard cell-free reaction with a membrane cut-off of 25 kDa and an RM/FM ratio of 1 : 17 was monitored by fluorescence at an emission at 509 nm and after excitation at 395 nm. Data are averages of at least three determinations.

Figure 2.

Cell-free production of membrane proteins. Lanes 1 and 2, in vivo expression. Samples of total cell extracts containing 10 µg of protein were analysed by SDS/PAGE in 17.5% (v/v) tricine gels. Lane 1, total protein of BL21 (DE3) Star × pET21-tehA before induction; lane 2, total protein of BL21 (DE3) Star × pET21-tehA 4 h after induction with 1 mm IPTG. Lanes 3–9, cell-free reactions, samples of 1 µL of the reaction mixtures were analysed. Lane 3, pET21-tehA total protein; lane 4, pET21-tehA soluble protein; lane 5, pET21-tehA pellet; lane 6, pET21-emrE pellet; lane 7, pET21-emrE-GFP pellet; lane 8, pET21-sugE pellet; lane 9, pET21-yfiK pellet. M, marker from top to bottom: 116, 66, 45, 35, 25, 18 and 14 kDa. Arrows indicate the overproduced proteins.

Modification of the cell-free expression system by addition of detergents and lipids

The results obtained with the EmrE-sgGFP fusion gave evidence that a cell-free expression of IMPs in a soluble condition might be feasible and a major reason for the observed precipitation of the vast majority of the IMPs might be the lack of any hydrophobic environment in the cell-free reaction. We therefore analysed whether the addition of detergents or lipids could increase the solubility of overproduced IMPs. As the addition of those substances might impact the general efficiency of the cell-free reaction, we first tested the production of GFP in the presence of various detergents which have been known to support the functional reconstitution of certain IMPs. DDM, DPC, β-OG, Thesit (Avanti Polar Lipids), Triton X-100 and Triton X-114 (Sigma) were added to the reaction mixtures in concentrations starting from the specific critical micellar concentrations (CMC) up to 1.5-fold CMC. With the highest concentrations tested, all detergents showed a negative effect on the GFP expression, and with DPC and β-OG no synthesized GFP was detectable even at the CMC concentrations (Fig. 3). The detergents DDM, Thesit, Triton X-110 and Triton X-114 showed less drastic effects on the GFP expression and even at the highest concentration analysed, only reductions of ≈ 60–80% of that of the control were observed. A slight increase in amount of soluble EmrE-sgGFP expression was only detectable after addition of Triton X-100 at 1.5-fold CMC (Fig. 4). As expected, DPC and β-OG also completely inhibited the EmrE-sgGFP production when at the CMC (data not shown).

Figure 3.

Effect of selected lipids and detergents on the efficiency of cell-free GFP expression. The reactions were incubated for 7 h at 30 °C. The fluorescence of GFP in a standard cell-free reaction corresponding to an average concentration of 2.6 mg·mL−1 was set as 100%. Blank bars, detergents; hatched bars, lipids. Detergent concentrations were 1.5-fold CMC. Lipid concentrations were 4 mg·mL−1. DDM, n-dodecyl-β-d-maltoside; DPC, dodecyl phosphocholine; β-OG, n-octyl-β-glucopyranoside; TX-100, Triton X-100; TX-114, Triton X-114; LPC, l-α-phosphatidylcholine; DMPC, 1,2-dimyristoyl-sn-glycero-3-phosphocholine; POGP, 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine; EL, E. coli lipid mixture.

Figure 4.

Increase of soluble EmrE-sgGFP expression in presence of selected lipids and detergents. The fluorescence was measured at 509 nm. The reactions were incubated for 7 h at 30 °C. The fluorescence of EmrE-sgGFP in a standard cell-free reaction corresponding to an average concentration of 5.8 µg·mL−1 was set as 100%. Blank bars, detergents; hatched bars, lipids. Detergent concentrations were 1.5-fold CMC (TX-110, TX-114, DDM) and twofold CMC (Thesit). Lipid concentrations were 4 mg·mL−1. DDM, n-dodecyl-β-d-maltoside; TX-100, Triton X-100; TX-114, Triton X-114; LPC, l-α-phosphatidylcholine; DMPC, 1,2-dimyristoyl-sn-glycero-3-phosphocholine; POGP, 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine; EL, E. coli lipid mixture.

We next analysed the effect of lipids on the cell-free GFP expression. l-α-phosphatidylcholine (LPC), 1,2-dimyristoyl-sn-glycero-3-phosphocholine (DMPC), 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POGP) and an E. coli lipid mixture were added in increasing concentration only to the RM. POGP resulted in a slight reduction of GFP expression down to approximately 80%, while no negative effects even at the highest analysed concentration of 4 mg lipid per mL RM was noticed with the other three lipids (Fig. 3). The addition of POGP, DMPC and E. coli lipids to the cell-free reaction proved to be beneficial for the soluble expression of EmrE-sgGFP protein. An increase in fluorescent EmrE-sgGFP of up to > threefold could be obtained upon addition of E. coli lipids (Fig. 4), resulting in a concentration of soluble fusion protein of 20 µg·mL.

Detergent solubilization of EmrE, SugE, YfiK and TehA

As the vast majority of the IMPs still remained insoluble we next approached the solubilization of the precipitated proteins using membrane mimicking detergent micelles. First, the solubility of the IMPs in different detergents dissolved in 15 mm sodium phosphate, pH 6.8, and 2 mm dithiothreitol was analysed, and impurities present in the insoluble pellets of the cell-free reactions were removed where possible. The detergents tested for their ability to solubilize the IMPs were β-OG, DDM, DPC, MHPG, NM, nondetergent sulfobetaines (NDSB-195, -201 and -256), SDS, Thesit, Triton X-100 and Triton X-114. The protein pellets containing the overproduced IMPs and other impurities were first washed twice with 15 mm sodium phosphate, pH 6.8, and 10 mm dithiothreitol. EmrE could then be almost quantitatively dissolved in a buffered 2% (v/v) DDM solution. Co-solubilized impurities could be removed easily by heating the solution to 75 °C for 1 h and apparently pure EmrE remained in solution (Fig. 5). The precipitated SugE and TehA proteins could be further purified by washing the pellets first with 3% (v/v) β-OG or with 20% (v/v) NDSBs. These IMPs dissolved only barely in β-OG or NDSB derivatives, and could be harvested by centrifugation, while most impurities remained β-OG or NDSB soluble (Fig. 5). SugE could then be solubilized in 2% (v/v) DPC, 0.1% (v/v) SDS or 1% (v/v) DDM and TehA solubilized best in 3%(v/v) DPC, 1% (v/v) DDM, or 1% (v/v) SDS. YfiK was washed with 1% (v/v) NM and with 1% (v/v) DPC and then solubilized in 3% (v/v) MHPG. For an efficient solubilization, the proteins were incubated on a shaker at 40 °C for 1 h. In addition, the presence of dithiothreitol was important and a higher molecular mass of the proteins observed after SDS/PAGE analysis without reducing agents indicated the formation of disulfide bridges in the protein precipitates (data not shown).

Figure 5.

Purification of cell-free expressed membrane proteins by selective solubilization. Pellets containing the precipitated membrane proteins were dissolved in various detergents in a volume corresponding to the volumes of the original reaction mixtures, and nonsolubilized proteins were removed by centrifugation. 5 µL samples of the soluble fractions were analysed by SDS/PAGE in 17.5% (v/v) tricine gels. Lane 1, EmrE in 2% (v/v) DDM after 1 h at 45 °C; lane 2, EmrE in 2% (v/v) DDM after 1 h at 75 °C; lane 3, SugE in 3% (v/v) β-OG after 2 h at 40 °C; lane 4, SugE in 20% (v/v) NDSB-201 after 2 h at 40 °C; lane 5, SugE in 1% (v/v) DDM after washing with 20% (v/v) NDSB-201; lane 6, TehA in 1% (v/v) DDM after washing with 3% (v/v) β-OG; lane 7, TehA in 1% (v/v) SDS after washing with 3% (v/v) β-OG; lane 8, TehA in 3% (v/v) DPC; lane 9, YfiK in 1% (v/v) DDM after washing with 25% (v/v) NDSB-256. M, marker from top to bottom: 116, 66, 45, 35, 25, 18 and 14 kDa. Arrows indicate the overproduced membrane proteins.

Structural analysis of solublized EmrE, SugE and TehA by CD spectroscopy

The solubilization of precipitated IMPs into detergent micelles might result in the refolding of the proteins. We therefore analysed the formation of secondary structures of the solubilized IMPs. SugE (15 µm) and TehA (10 µm) were measured in 15 mm sodium phosphate buffer, pH 6.8, 2 mm dithiothreitol, and supplemented with DPC, DDM and SDS, respectively. EmrE was measured in 10 mm sodium phosphate, pH 7.4, 2 mm dithiothreitol and with 2% (v/v) DDM. The spectra measured in the various detergent micelles at 25 °C, showing minima at 208 and 222 nm and a large peak of positive ellipticity centered at 193 nm, were characteristic of α-helical proteins (Fig. 6). The analysis of the spectra yielded an estimate of 55 ± 4%α-helical content for EmrE, 72 ± 11% (DPC), 60 ± 11% (SDS) and 84 ± 10% (DDM) for SugE and 78 ± 8% (DDM), 49 ± 3% (DPC) and 40 ± 15% (SDS) for TehA. The predicted α-helical contents, after primary stuctural analysis, were 69% for EmrE, 67% for SugE and 70% for TehA. According to these data, the adoption of the mostly folded conformation of SugE might be favoured upon solubilization with DPC, and with DDM for TehA, respectively.

Figure 6.

CD spectroscopy of solubilized multidrug transporter in detergent micelles. Far-UV spectra were taken at 25 °C in buffered detergent solutions. (A) 24 µm EmrE in 2% (v/v) DDM in 10 mm sodium phosphate, pH 7.4. (B) 15 µm SugE in 15 mm sodium phosphate, pH 6.8, 2 mm dithiothreitol with various detergents. (C) 15 µm TehA in 15 mm sodium phosphate, pH 6.8, 2 mm dithiothreitol with various detergents. SDS, sodium dodecylsulfate; DDM, n-dodecyl-β-d-maltoside; DPC, dodecyl phosphocholine.

Reconstitution of solubilized EmrE, SugE and TehA into proteoliposomes

The precipitated proteins produced by cell-free reactions were solubilized in a 1% (v/v) DDM solution in 15 mm sodium phosphate, pH 6.8, and 2 mm dithiothreitol. Reconstitution into proteoliposomes with E. coli lipids was carried out at a molar protein/lipid ratio of 1 : 500. The insertion of EmrE, SugE and TehA into the lipid membranes was monitored by freeze-fracture electron microscopy (Fig. 7). As would be expected by a functional reconstitution, all three proteins inserted as homogenously dispersed particles into the vesicles. The efficiency of insertion of SugE and EmrE was comparable and an estimated 80% of the vesicles contained inserted proteins. In the case of TehA, the efficiency of proteoliposome generation was less, and ≈ 10% of the vesicles contained proteins.

Figure 7.

Freeze-fracture electron microscopical analysis of reconstituted proteoliposomes. The membrane proteins EmrE (A), SugE (B) and TehA (C) were solubilized in 1% (v/v) DDM and reconstituted in E. coli lipid vesicles (bold arrows). Randomly distributed particles (small arrows) in the fracture faces indicate incorporation of proteins into vesicular membranes. Scale bar = 100 nm.

Ethidium/H+ antiport in reconstituted EmrE proteoliposomes

The functional reconstitution of EmrE into proteoliposomes was tested with an established transport assay using ethidium bromide as a ligand [11]. Intercalation of ethidium into DNA causes an effect on the quantum yield of its fluorescence. Active EmrE protein should therefore generate a significant increase in the fluorescence intensity, by pumping ethidium into the proteoliposomes where it is accumulated in the DNA molecules. Approximately 140 nm EmrE embedded in E. coli lipids were assayed in a total volume of 1 mL. After establishing the baseline, proteoliposomes were added, followed by ethidium bromide after 10 s to a final concentration of 2.5 µm. An immediate large biphasic increase in the fluorescence was monitored (Fig. 8). The first phase of the increase can be attributed to the binding of ethidium to residual DNA in the extraliposomal space [11], while the second phase represents the accumulation of ethidium inside the liposomes due to the transport activity of EmrE. Preincubation of the proteoliposomes with an excess of 50 µm of the high affinity substrate TPP+ completely eliminated the second phase, probably through competition with the ethidium binding site at EmrE. In addition, the collapse of the pH gradient upon addition of nigericine also prevented the accumulation of ethidium in the proteoliposomes, resulting only in the single phase increase of fluorescence after addition of ethidium bromide. The results clearly demonstrate that the ethidium/H+ antiport was responsible for the observed increase in fluorescence, indicating the functional reconstitution of EmrE in E. coli lipids.

Figure 8.

Ethidium transport assay of EmrE proteoliposomes. Transport of ethidium into reconstituted EmrE proteoliposomes in 15 mm Tris/Cl, pH 8.5, 2 mm dithiothreitol, 150 mm KCl was measured by an increase in fluorescence at excitation and emission wavelengths of 545 and 610 nm, respectively. Ten microliters of proteoliposomes (approximately 140 nm EmrE) were added after 30 or 60 s. If appropriate, substances were added at the following final concentrations: TPP (50 µm), ethidium bromide (2.5 µm) and nigericine (5 µg·mL−1). Arrows indicate the time points of addition.

Structural analysis of selectively labeled EmrE, SugE and YfiK by NMR spectroscopy

One advantage of the cell-free expression technique is the rapid and efficient uniform or amino acid specific labeling of the overproduced proteins. Selected amino acids can be replaced by their labeled derivatives and provided in the reaction mixtures. We selected the relatively abundant amino acids glycine and alanine for a specific labeling approach of EmrE, SugE and YfiK and for the generation of samples suitable for NMR spectroscopy. The quality and dispersion of recorded two dimentional 1H,15N correlation spectra could provide information on whether the solubilized IMPs are either aggregated or present in a folded conformation. However, in addition to the size of the proteins, a major problem for the solution NMR analysis of IMPs, is the size of the detergent micelles necessary for the solubilization. We therefore took advantage of the reported high stability of EmrE in the organic solvent mixture CDCl3/CD3OH/H2O (6 : 6 : 1, v/v/v) with 200 mm ammonium acetate, pH 6.2, and 10 mm dithiothreitol [11,23]. The pellets of preparative scale cell-free reactions with a total of 2 mL RM were washed twice with 15 mm sodium phosphate, pH 6.8, and 2 mm dithiothreitol and then suspended in the chloroform mixture in a volume corresponding to one fourth of the volume of the RM. The suspension was incubated on a shaker for 2 h at 40 °C and then centrifuged at 20 000 g for 5 min at 15 °C. The supernatant was then used directly for NMR analysis. Interestingly, the SugE protein shared this stability in the chloroform mixture with its homologue EmrE and could be dissolved by using identical procedures. Both proteins were apparently pure in the chloroform mixture as judged by SDS/PAGE analysis and the impurities obviously remained insoluble during this treatment.

The YfiK protein did not dissolve in the chloroform mixture but it showed good solubility in buffered MHPG solutions. The pellets of six preparative reactions with 0.5 mL RM, each containing the YfiK protein, were combined, washed in 1% (v/v) NM and in 1% (v/v) DPC and dissolved in 2 mL of 1% (v/v) MHPG in 25 mm sodium phosphate, pH 6.0, with 5 mm dithiothreitol. After removal of insoluble protein by centrifugation, the sample was concentrated fourfold and measured by NMR. The final protein concentration of YfiK in the sample was calculated at approximately 6 mg·mL−1, indicating a yield of solubilized labeled YfiK of approximately 1 mg per ml of cell-free RM.

The selectively labeled proteins were subsequently analysed by heteronuclear [15N,1H]-TROSY experiments at 500 or 600 MHz 1H frequency. In the EmrE spectrum, all nine alanine residues and 12 glycine residues are visible and well resolved, spanning an area between 7.5 and 9 p.p.m and indicating a specific folded conformation of the solubilized EmrE protein (Fig. 9A). The spectrum could be nicely aligned with a previously published [15N,1H]-HSQC spectrum of uniformly labeled EmrE, prepared by conventional in vivo expression and labeling in E. coli[23], and all signals of the specifically labeled residues could be assigned accordingly. The dispersion of the amide proton signals also indicated a monomeric conformation of EmrE. The [15N,1H]-TROSY spectrum of the SugE protein also showed a good resolution, and signals of all the 14 alanine and 11 glycine residues were detectable, spanning an area between 7.5 and 8.9 p.p.m, and indicating again a folded conformation of the solubilized protein (Fig. 9B). Despite the size of the 21.3 kDa YfiK protein, the dispersion of its [15N,1H]-TROSY spectrum in MHPG micelles showed a reasonable resolution, and signals of most of the 24 alanine and 13 glycine residues were visible (Fig. 9C).

Figure 9.

[15N,1H]-TROSY spectra of solubilized membrane proteins. The proteins were specifically labeled with [15N]alanine and [15N]glycine by cell-free expression. (A) 0.1 mm EmrE dissolved in CDCl3/CD3OH/H2O (6 : 6 : 1, v/v/v) with 200 mm ammonium acetate (pH 6.2) and 10 mm dithiothreitol. The assignments for the amide proton-nitrogen pairs according to Schwaiger et al.[23] are indicated. The spectrum was taken at 15 °C with a 600 MHz spectrometer. (B) 0.5 mm SugE dissolved in CDCl3/CD3OH/H2O (6 : 6 : 1, v/v/v) with 200 mm ammonium acetate (pH 6.2) and 10 mm dithiothreitol. The spectrum was taken at 15 °C with a 500 MHz spectrometer. (C) YfiK (0.3 mm) solubilized with 4% (v/v) MHPG in 25 mm sodium phosphate (pH 7.0) and 5 mm dithiothreitol. The spectrum was taken at 30 °C with a 600 MHz spectrometer.


We describe a new and versatile approach for the rapid production, purification and reconstitution of large amounts of structurally folded IMPs, and for the generation of amino acid specific labeled samples suitable for NMR spectroscopy. The production of sufficient amounts of protein is the major bottleneck for the structural and functional analysis of membrane proteins in vitro. In addition, if a protein is produced it has to be isolated from complex cellular membranes by time consuming procedures that frequently involve considerable losses. The small multidrug transporter EmrE is one of the few exceptions of IMPs which can also be produced in relatively high amounts by in vivo expression. Yields of up to 1 mg·L−1 after intensive optimizations in E. coli systems have been reported [24] and a hemagglutinin epitope-tagged functional EmrE derivative was expressed in the yeast Saccharomyces cerevisiae at levels of approximately 0.5 mg·L−1[25]. For SugE, TehA and YfiK are no quantitative data available for in vivo expression, and this is the first report of preparative expression of these proteins. We have been able to demonstrate the cell-free production of at least 1 mg·mL−1 of reaction mixture of all of our four target proteins. In the case of SugE and TehA, the production rates were considerably higher. After purification and solubilization into detergent micelles, we could calculate a yield of resolubilizable protein of 1 mg·mL−1 RM for YfiK, 1.5 mg·mL−1 RM for SugE and of 2.7 mg·mL−1 RM for TehA. These calculations did not take into account the amount of proteins which remained insoluble. The obtained production rates of membrane proteins by cell-free expression are therefore comparable to that of other proteins [7,26,27].

The structural reconstitution of EmrE, SugE, YfiK and TehA was monitored by different techniques. EmrE represents one of the best characterized model systems of an integral membrane transporter and its reconstitution is a very well established technique. We included a simple incubation step at 75 °C for the rapid purification of EmrE as it was previously reported that the exposure of EmrE to 80 °C did not affect its transport activity after reconstitution [28]. EmrE is tightly packed without any hydrophilic cytoplasmatic domains [29] and this conformation might cause its somewhat unique solubility and stability in organic solvents [11], and might also favour the observed rapid reconstitution in micelles or liposomes. Homologous proteins of EmrE such as SugE and probably also YfiK and TehA, seem to share these properties and the presented strategy of a cell-free production as precipitate might therefore be advantageous even for this class of IMPs, in order to obtain pure samples of the nonmodified proteins just by using selective resolubilization protocols in suitable detergents. We could demonstrate for the first time that SugE has a high stability in organic solvents comparable to that of EmrE and that it was able to refold into a structural conformation in the identical chloroform mixture. SugE, like EmrE, appears to be monomeric in chloroform as judged by the dispersion of its [15N,1H]-TROSY spectrum. The spectra of both proteins were well resolved, and the [15N,1H]-TROSY spectrum of the cell-free produced and reconstituted EmrE protein is comparable to that of EmrE prepared after in vivo expression [23].

Far-UV CD spectroscopy of solubilized EmrE, SugE and TehA in various detergents revealed spectra typical for predominantly α-helical proteins [30]. EmrE has α-helical estimates of 78% and 80% in chloroform/methanol/water and DMPC, respectively [29,31]. Accordingly, the predicted predominantly α-helical secondary structures of SugE and TehA were in good agreement with the data obtained from CD spectroscopy of the solubilized proteins. The observed differences in α-helicity, in combination with the various detergents, might reflect variations in the protein conformations depending on the type of micelles[32]. An extensive analysis of the effects of different membrane mimetic environments on the conformation of EmrE has recently been published and remarkably, differences in the conformational dynamics, were monitored [33]. The largest amount of α-helical content of EmrE was observed in DDM and the authors assumed that the protein is in a slightly more denatured state in other environments. Their data are in full agreement with our results. Additionally, SugE and TehA also showed the highest α-helicity in DDM.

In MHPG micelles, the YfiK protein showed a reasonable resolution in the [15N]-TROSY spectrum, as would be expected from a protein with a mass in the range ≈ 50–100 kDa. Classical multidrug transporters contain 12 TMS per monomer or functional unit. The EmrE monomer would therefore be three times smaller than this 12 TMS consensus, and it is speculated that functional EmrE might be composed of three subunits [10,34]. It could therefore be possible that the six TMS containing YfiK monomers might reconstitute as oligomers. Considering the estimated micellar size of DPC of ≥ 25 kDa, even as a monomer the analysed molecules would have a minimum size of 47 kDa, which is then in agreement with the observed data.

For the functional analysis of the multidrug transporter EmrE, we could take advantage of a previously established activity assay [11], and the functional reconstitution of the cell-free produced and solubilized protein into proteoliposomes could be clearly demonstrated. The ethidium transport could be specifically competed against the high affinity substrate TPP+ [34], and it was eliminated by affecting the membrane proton gradient with nigericine. Unfortunately, ethidium is not a substrate for SugE and as only nonfluorescent quarternary ammonium compounds have been reported as potential ligands [12], analoguous assays have not be established to date. Ethidium is a potential substrate of TehA but we have not been able to detect any transport activity with proteoliposomes of TehA solubilized either in DPC, DDM or SDS and reconstituted with an E. coli lipid mixture (data not shown). However, the analysis of proteoliposomes by freeze-fracture electron microscopy gave evidence of a structural reconstitution of SugE, EmrE and TehA in E. coli lipid vesicles, and no differences between SugE and EmrE proteoliposomes could be observed. It should also be noted that the function of TehA is not very well analysed yet, and it is not clear so far whether the transport activity requires TehA alone or in a complex with other proteins or cofactors [13].

GFP has been shown to be a sensitive folding indicator for the study of globular and membrane protein overexpression in E. coli[35,36], and it is most likely to become correctly folded as a C-terminal fusion that is not translocated through the membrane into the periplasm. Therefore, at least the C-terminus of the target protein should remain in the cytoplasm. Approximately 70% of all predicted membrane proteins are believed to have this topology. For EmrE, the cytoplasmic localization of the N- and C-terminal ends has been shown [29], and the C-terminal fusion of GFP should therefore not prevent its reconstitution into membranes. In addition, a fully functional chimera between EmrE and GFP was expressed in S. cerevisiae and it conferred resistance against TPP+, acriflavine and ethidium [25]. It can therefore be assumed that the observed fluorescent part of the cell-free produced EmrE-sgGFP fusion also contains a functionally folded EmrE protein. Despite optimized conditions upon addition of E. coli lipids, only an estimate of approximately 1% of the total overproduced protein stayed soluble. While this could already be sufficient for certain analytical assays, higher yields of soluble membrane proteins might be possible by increasing the added amounts of lipids or by providing alternative hydrophobic environments. Dog pancreas microsomes have, for example, been used to produce analytical amounts of completely assembled human T-cell receptor by in vitro expression [37]. The cell-free expression principally offers the opportunity to insert the translated protein directly into the desired membrane of choice. Tedious efforts of delipidation and reinsertion of the overproduced membrane proteins into artificial membranes could therefore be avoided, and the possibility of soluble cell-free membrane protein expression might be considered if the reconstitution of a protein is not possible or if only analytical amounts of protein are needed.

Membrane proteins are difficult to analyse by solution NMR techniques, and the main problems are caused by the sizes of the detergent micelles needed for solubilization. Spectra are frequently very crowded and the low dispersion of signals prevents the effective assignment of residues. A valuable tool to approach this problem is the amino acid specific labeling of membrane proteins by cell-free expression. Whilst the selective labeling of proteins for NMR studies in both individual and commercial cell-free expression systems has already been demonstrated [26,38–41], this report shows the first application of this technique to membrane proteins. The selective labeling of proteins by cell-free expression is highly efficient and advantageous compared with the in vivo labeling. No auxotrophic strains and minimal media are needed, and commonly encountered problems with reduced expression rates are thus eliminated. In addition, due to the lack of any metabolism during cell-free expression, cross-labeling problems usually do not occur. The presented [15N,1H]-TROSY spectra of EmrE, SugE and YfiK nicely demonstrate the highly efficient amino acid selective labeling of membrane proteins without any losses in the protein yields. Together with the fast generation of samples, the selective cell-free labeling of membrane proteins could considerably accelerate the assignment of proteins showing a reasonable resolution. The approach presented here might become especially valuable for solid-state NMR studies. The possibility of producing mg quantites of membrane proteins, with the option of using a range of different isotope labeling schemes, enables structural studies of IMPs reconstituted into lipid membranes. So far, only ligand studies by MAS-NMR have been feasible for these protein families [42], but solid-state NMR studies on some of the presented proteins are already in progress.

Cell-free expression has a high potential to become a valuable tool for the rapid generation of samples suitable for structural analysis [43], and commercially available systems have been developed for the efficient production of proteins on a preparative scale [39,44]. In addition, cell-free expression might also yield more homogenous protein samples more readily suitable for crystallization. The main advantages of the cell-free expression of IMPs were the high level production of insoluble protein and the efficient selective labeling. This is the first report of the solubilization of SugE, YfiK and TehA in micelles and of their reconstitution into membranes. The dissolving of the proteins in suitable detergents obviously resulted in the refolding of the proteins, and renaturation procedures with strong denaturants such as guanidine hydrochloride could be omitted. At least for the family of small multidrug transporters, the cell-free expression technique seems to be a highly appealing way to generate samples suitable for NMR spectroscopy. The production rate of membrane proteins in the cell-free system was not related to the number of transmembrane domains, and the cell-free expression of even larger membrane proteins might therefore be possible. Regardless of this, the cell-free expression could be a suitable tool for the rapid screening of the general likelihood of expression of membrane proteins.


We are grateful to Vladimir Shirokov and Alexander Spirin for valuable advice in establishing the cell-free expression system, Marc Lorch for his help in proteoliposome preparation and Klaus Fendler for helpful discussions. The work was financially supported by DFG (grant GL307-1/3) and BMBF project ProAMP.