EmrE and TBsmr were purified into n-dodecyl-β-d-maltoside (DDM), as His-tagged protein following overexpression in E. coli according to previous methods 7. Protein yields were ~1 mg for EmrE and 2.5 mg of TBsmr per litre of culture at > 95% purity.

Reconstitution and assay conditions were optimised using EmrE and E. coli lipid vesicles. Transport activity was quantified using pH driven transport of radiolabelled, ¹⁴C, MV a method that has been extensively applied to EmrE 8. Previous reconstitution procedures 28 were optimised to maintain a pH gradient across the liposome bilayer during the transport measurement. EmrE was purified into DDM, but was then exchanged into octyl-β-D-glucopyranoside (OG) prior to reconstitution into lipid vesicles. Although EmrE is less stable in OG, this detergent is easier to remove from the bilayer than DDM (following reconstitution into lipid vesicles) as OG has a higher critical micelle concentration (CMC) than DDM (0.53% or 18 mM for OG as opposed to 0.009% or 0.17 mM for DDM). Thus this initial exchange of the protein to OG reduces proton leakage through the lipid bilayer enabling a proton gradient to be maintained which can drive MV translocation over minutes. The amount of OG remaining after reconstitution into E. coli lipid vesicles was 0.3 mM (well below the CMC of 18 mM and leaving a lipid:OG ratio 430:1). If EmrE was reconstituted directly from DDM into lipid vesicles (i.e. omitting the initial transfer to OG), 0.09% (1.8 mM) of DDM remained in E. coli protein-containing lipid vesicles. This is above the CMC of 0.009% and resulted in leaky lipid vesicles that could not maintain a pH gradient. The OG reconstitution procedure used here also involved only partial pre-saturation of the lipid vesicles by OG to help maintain bilayer integrity, but which also meant that <50% of the protein reconstituted. For example, 50 μg of EmrE was initially mixed with E. coli lipid vesicles but only 21 μg were found to be associated with the vesicles after reconstitution. Since SMR proteins are more unstable in OG compared to DDM, the proteins were immediately reconstituted into lipid vesicles following detergent exchange into OG.

EmrE was reconstituted, via OG, into E. coli lipid vesicles that were initially prepared by sonication or extrustion to give differing sizes (50 and 200 nm in diameter). As shown in figure 1 increased assay reproducibility is observed when vesicles were formed by extrusion rather than sonication. In addition 50 nm diameter lipid vesicles were chosen over 200 nm lipid vesicles, since it was possible to determine both a linear regime of transport to determine an "initial rate of transport" as well as saturation of transport (by loss of pH gradient or substrate accumulation). The initial rate of MV transport by EmrE in 50 nm E. coli lipid vesicles was ~250 nmol. min^-1. mg^-1.

In summary, reconstitution of OG solubilised SMR proteins in OG pre-saturated lipid vesicles, gave tight lipid vesicles with low proton leakage and enabled linear MV transport to be observed in lipid vesicles over several minutes and thus initial rates to be determined. Radiolabelled MV transport provides a quantitative measure of EmrE transport activity. Activities over specified range of lipid conditions were quantified for EmrE and TBsmr by MV transport, as MV is a substrate for both proteins.

The native membranes of EmrE and TBsmr are of similar composition with PE lipids being the majority constituent closely followed by an anionic lipid. The most abundant anionic lipids in EmrE native membranes are PG lipids opposed to PI lipids found in native TBsmr membranes. The initial rate of MV transport at low MV concentrations proved a reliable method of comparing the lipid dependences of the two protein activities. MV uptake has been previously used to determine EmrE function and initial rate values ranging between 160 to 30,000 nmol.min^-1.mg^-1 have been reported for EmrE reconstituted into E. coli vesicles782627. The diverse values are attributed to variations in reconstitution methodologies, solubilisation conditions and lipid concentrations. Our initial rate values (of 250 nmol.min^-1.mg^-1 for EmrE in E. coli) are in agreement with the lower end of this range of values. In vivo experiments have been utilised to identify SMR ligands 78 but limited kinetic characterisation has been undertaken. Here, we have investigated how varying lipid compositions, which mimic aspects of the in vivo environments, affect SMR transport in vitro.

The incorporation of PE into PC bilayers increases the initial rate of MV transport for both EmrE and TBsmr up to 0.4 PE mole fraction (see figures 2c and 3b). TBsmr is less active in DOPC lipid vesicles than EmrE and shows a smaller, 4-fold increase in transport rate with PE, compared to 20-fold for EmrE. Both proteins also showed rate increases when DOPG was added to DOPC; about 25-fold for EmrE and 5-fold for TBsmr. TBsmr showed a larger, 20-fold, increase in rate when its native PI lipid headgroup was added instead of DOPG (figure 3c). These results show that a fluid DOPC lipid bilayer structure imparts a degree of transport activity to SMR proteins. However, transport is suboptimal in DOPC and regardless of the lipid added to DOPC, an increase in transport rate is observed when the DOPC mole fractions falls below 0.6 (see figures 2c,d and 3b,c). This implies a threshold level of PC of ~0.6 mole fraction in binary lipid mixtures, above which transport is hindered. An increase in rate occurs upon addition of PE or an anionic lipid to PC. A faster transport rate still is seen in the absence of PC, but with a combination of PE and the anionic lipid. These are the dominant lipid types in the native membranes: PG/PE for EmrE and PI/PE for TBsmr, which give ~70-fold and 25-fold increases in initial rate over that in DOPC (which is not present in either native membrane). In this study, synthetic lipids with defined chain compositions were used rather than the mixtures of differing chain length and saturation present in native lipid compositions, thus these varying natural chains could also further enhance transport activity. The data for EmrE in PC/PG (figure 4) show that the increase is due to PG increasing the maximum transport rate, V_max. The negative charge of PG or PI is likely to play a role in attracting the positively charged MV substrate. The larger increase in transport observed with PI over PG for TBsmr, presumably reflects a preferential interaction with its native inositol group.

Membrane proteins are heavily influenced by their surrounding lipid environment. Specific lipid interactions have been identified with lipids either being seen tightly bound in X-ray structures or a certain lipid being essential for function 33. Alternatively, generic bilayer properties can significantly influence protein function 23. Key properties include a mis-match between the hydrophobic length of the protein and lipid bilayer as well as the elastic properties of the bilayer, which include curvature energy and lateral pressure. Thus, alamethicin channel formation and function has been directly linked to bilayer curvature in PC/PE bilayers 2122, folding is also dependent on this curvature and lateral pressure in PC/PE systems 20. Here we have shown that PE favours EmrE or TBsmr transport activity, in an anionic lipid background (PG for EmrE or PI for TBsmr). Although, the greatest influence on transport is the anionic lipid.

Lipids have also previously been shown to affect transport function but it is only recently that systematic studies have been undertaken to investigate their effect on integral membrane proteins. E. coli lactose permease is known to be affected by lipids 34, with initial rates of lactose showing similar dependences on PE as reported here for SMR proteins. The lactose rate increased upon addition of 0.5 mole fraction PE to PC. The effect of PG differed to that reported here, as 0.5 mole fraction PG did not increase the lactose rate, unlike the large 20-fold increase seen here for EmrE. PE has also been reported to affect the biogenesis, folding and topology of lactose permease and other transporters 353637. Other multidrug transporters of the major facilitator superfamily (MFS) including GabP, PheP and LmrP 383940 have shown lipid dependencies for function. In the case of LmrP the replacement of PE for PC in the bilayer lead to significant alterations in both structure and function. Further studies on LmrP using PE-methylated moieties suggest that direct interactions between the PE headgroup and certain LmrP amino acids are directly involved in pH sensing required for substrate transport 41. There are also some indications that multidrug transporters from other protein families are also dependent on lipid composition. In the case of the ATP binding cassette multidrug transporter, p-glycoprotein, drug binding has also been shown to be influenced by lipid environment 42.

EmrE-His and TBsmr-His were cloned into pT7-7, transformed into E. coli TA15/pGP1-3 and overexpression (induced by heat shock) and purification were performed as previously described 35. A 6 L volume of cells was disrupted by passage through a French ™pressure cell at 1000 psi (6900 kPa). Isolated membranes were solubilised at 4°C for at least 2 hours in 50 ml of solubilising buffer (20 mM Tris, pH 7.5, 100 mM NaCl, 10 mM 2-mercaptoethanol, and 1% (w/v) DDM). Insoluble material was removed by centrifugation at 35 000 g and 4°C. NaCl and imidazole were added to final concentrations of 350 mM and 15 mM respectively, and DDM solubilised protein was consequently incubated with 1.5 ml of Ni-NTA agarose (QIAGEN) for 1.5 h at 4°C. Non-specifically bound protein was removed by washing with 30 column volumes of 20 mM Tris pH 7.5, 400 mM NaCl, 15 mM imidazole, 0.1% DDM and 5 mM 2-mercaptoethanol, followed by 15 column volumes of the above buffer with the exception of 20 mM NaCl. Proteins were eluted with 5 column volumes of elution buffer (20 mM Tris, pH 7.5, 25 mM NaCl, 200 mM imidazole, 0.1% DDM and 5 mM 2-mercaptoethanol). Eluted protein was concentrated in an Amicon Ultra 50 000 MWCO centrifugal concentrator (Millipore) to approximately 1–2 mM as determined by A₂₈₀, snap frozen in liquid N₂ and stored at -80°C until required. Protein purity was determined by SDS-PAGE and Western blot analysis. His-tagged protein was used throughout. EmrE purified into DDM was assayed for activity by TPP binding, by radiolabelled TPP binding curves as well as isothermal titration calorimetry. The latter gave a dissociation constant, K_d of 36 nM in 0.08% DDM with a binding stoichiometry of 0.48 TPP per EmrE, i.e. binding to dimer EmrE, with > 90% of the protein being active.

Lipid vesicles were prepared with 50 nm, 100 nm or 200 nm diameter by extrusion as previously 4344. Vesicles were also prepared by sonication.

Transport assays were performed as previously described 8 using radiolabelled ¹⁴C MV²⁺ (Sigma-Aldrich) (referred to as MV throughout). Protein was first exchanged from DDM into OG 28: DDM-protein was incubated with 1.5 ml of Ni-NTA beads (pre-washed 1% OG and 20 mM Tris pH 7.5) for 1.5 h at 4°C; loaded onto a Pharmacia XK-16/20 glass column; washed 4-times with buffer containing 150 mM NaCl, 1% OG, 15 mM mercaptoethanol and 15 mM Tris pH 7.5 and then mixed with elution buffer (the previous buffer with 200 mM imidazole and 15 mM mercaptoethanol) and incubated at room temperature for 15 min. Eluted protein was concentrated by centrifugation at 4000 g using Amicon Ultra 50 000 MWCO centrifugal concentrator (Millipore). Protein in OG was immediately reconstituted into lipid vesicles (at a starting protein:lipid mole ratio of ~2900:1). Lipids were rehydrated (to 50 mg.ml^-1 in 150 mM NaCl, 15 mM Tris, pH 7.5) and sonicated or extruded to 50 nm, 100 nm or 200 nm as stated in the text. The vesicles were mixed with OG to a final concentration of 0.65% w/v, or a 25:1 molar ratio of lipid to OG. 75 μg of OG solubilised EmrE-His was added to 375 μl of these OG-lipid vesicles and OG was removed by a ~70 fold dilution into 25 ml of 190 mM NH₄Cl, 15 mM Tris pH 7.5 followed by mixing at room temperature for 20 min. The protein-containing vesicles were centrifuged at 250 000 g for 60 min, at 25°C and the pellet resuspended in 100 μl of reconstitution buffer (190 mM NH₄Cl, 15 mM Tris pH 7.5) and stored at -80°C. Protein-containing vesicles were thawed, resealed by a 20 s sonication step and transport was initiated by dilution of 3 μl of protein-containing vesicles into 200 μl of assay buffer (140 mM KCl, 5 mM MgCl₂ and 10 mM Tricine) containing 42 μM ¹⁴C MV²⁺, at pH 8 for EmrE and pH 9 for TBsmr. At given times ranging between 0 and 20 min the reaction was stopped by dilution with 2 ml of ice cold assay buffer and protein-containing lipid vesicles were collected by filtration through Scheicher and Schull (0.2 μm) (Millipore, Watford, UK) filters. (Although the procedure commenced with extruded unilamellar vesicles, their size will have altered during reconstitution and the transport assay. We found, as previously reported 8, that 0.2 μm filters worked well and better than smaller pore sizes). Filters were placed into scintillation vials with 10 ml of Emulsifier-safe scintillation fluid (PerkinElmer) and radioactivity was measured in a TRI-CARB 2100TR Liquid Scintillation Counter. Experiments were carried out in triplicates.

The amount of OG and DDM remaining after reconstitution was determined by a phenol sulphuric acid carbohydrate assay 45. For this assay, 50 μl of the protein-containing vesicles were added to 250 μl of 5% (w/v) phenol and 600 μl of concentrated sulphuric acid. This mixtures was incubated for 5 min and the absorbance was measured at 460 nm.

Maintenance of pH was monitored by incorporation of the pH sensitive dye carboxyfluorescein (Molecular Probes) inside protein-containing lipid vesicles. Protein-containing vesicles were formed as described above in the presence of 10 μM carboxyfluoroscein. 3 μl of protein-containing vesicles were diluted into 200 μl of assay buffer and thoroughly mixed by pipetting. Changes in carboxyfluoroscein fluorescence were monitored over 15 min, at 25°C by exciting the sample at 475 nm and measuring emission at 520 nm.

Initial transport rates of MV were determined from the gradient of the linear part of the MV accumulation versus time (e.g. figure 1a): over ~the first 6 mins for EmrE or 16 mins for TBsmr. Rates (apart from those for K_m and V_max determinations) were determined for 42 μM MV. All fitting was carried out using MicroCal Origin 6.0 software. Initial rates were calculated per mg of protein associated with lipid vesicles after reconstitution. Note that identical protein samples were used for each lipid series; protein aliquots were taken from the same protein sample exchanged from DDM into OG for all DOPC/DOPE measurements, or all from a separate protein sample for all DOPC/DOPG or DOPG/DOPE measurements.

The amount of protein was quantified on samples after the transport assay, on the same sample, from colorimetric development of Western blots 46, by densitometry using Quantity One^® software (BioRad). The actual amount of reconstituted protein was calculated by comparing band densities between protein-containing lipid vesicle samples, normalised to an aliquot of the same protein preparation which had been previously quantified (each Western always contained a band from a protein sample that had already been quantified on another gel, to enable normalising between gels). Protein-containing vesicle samples were run on 10% SDS denaturing gels and transferred to polyvinyl difluoride membranes prior to incubation with anti c Myc alkaline phosphatase (1:5000) (Sigma Aldrich) (the protein constructs contained a Myc epitope). Alkaline phosphate substrate solution, BCIP-NBT (Sigma Aldrich), was applied to the membrane as per manufacturer's protocol for protein visualisation and quantification. The BCIP-NBT detection system is based on the hydrolysis of 5-bromo-4-chloro-3 indolyl phosphate (BCIP) and the reduction of p-nitrobluetetrazolium chloride (NBT) producing a purple product, which can be colorimetrcally quantified. This Western blot method due to its high sensititivity (0.5 ng of substrate) 4748proved effective for quantifying the relatively small amounts of EmrE reconstituted by the method here, which was optimised to maintain a pH gradient rather than to optimise EmrE concentrations. The amount of reconstituted protein was also assessed by a modified Lowry method as well as following solvent extraction of protein and lipid. For the latter EmrE samples in lipid were solubilised using a 43:43:14 chloroform:methanol:sample (v/v) ratio and absorbance spectrum measured between 250 and 310 nm. EmrE standards of known concentration in DDM were used for calibration and lipid samples in the absence of EmrE were used to subtract any non-protein absorbance in the region from 250–310 nm.

In order to determine K_m and V_max MV transport assays were performed with different concentrations of ¹⁴C MV at three different lipid mole fractions for each lipid series: DOPC/DOPE, DOPC/DOPG and DOPC/DOPG. MV concentrations were between 5 μM and 2 mM MV (spiked with 5% ¹⁴C MV²⁺). The extent of non-specific binding of MV was determined in control experiments in protein-free lipid vesicles (for all lipid compositions). This non-specific binding was corrected for in the binding curves shown. At 42 μM MV this background binding was significantly lower than the specific binding and transport. 42 μM MV corresponds to: ~250:1 mole ratio MV to protein and ~50:1 lipid: MV; 600 μM MV to: ~4000:1 MV to protein, and 3:1 lipid: MV.