J. Phys. Chem. B 2009, 113, 8569–8580
8569
Photoreversible Conformational Changes in Membrane Proteins Using Light-Responsive Surfactants Jing Zhang, Shao-Chun Wang, and C. Ted Lee, Jr.* Department of Chemical Engineering and Materials Science, UniVersity of Southern California, Los Angeles, California 90089-1211 ReceiVed: September 4, 2008; ReVised Manuscript ReceiVed: April 8, 2009
Photoreversible control of the conformation of bacteriorhodopsin in the presence of a light-responsive surfactant is demonstrated through combined UV-vis, FT-IR, and 31P NMR spectroscopy and dynamic light scattering (DLS) measurements. The azobenzene-based surfactant photoisomerizes upon 434 nm visible (trans, relatively hydrophobic) and 350 nm UV (cis, relatively hydrophilic) illumination, allowing surfactant micellization to be reversibly controlled. This leads to partitioning of the membrane protein into micelles in the unfolded state under visible light, while UV light leads to solubilization of the protein within purple membrane bilayers in the folded state. A three-stage model of purple membrane-photosurfactant interactions is examined through NMR and DLS measurements. Phototriggered unfolding of bacteriorhodopsin, occurring through RII f RI and reverse β-turn f extended β-strand transitions, requires ∼20 s for completion, while light-induced refolding requires a somewhat longer 80 s as the membrane protein repartitions into the reformed bilayer membrane. Each of these conformational changes can be precisely and reversibly controlled with simple light illumination, providing a novel technique to probe membrane protein folding. Introduction Membrane proteins, which constitute 20-30% of all cellular proteins, are of crucial importance for biological function, serving as mediators between the cell and the surrounding environment, catalyzing the transport of ions and metabolites across the cell membrane, and can further be responsible for a wide range of diseases (e.g., cystic fibrosis, diabetes, and hypertension). Thus, a thorough understanding of membrane protein structure and function is required to allow for the development of new classes of drugs capable of specifically targeting membrane proteins.1 Unfortunately, the fact that membrane proteins simultaneously contain large hydrophobic and hydrophilic regions, which respond to solvent conditions in different and complex ways, has made it exceedingly difficult to examine membrane protein folding pathways as unfolding often leads to irreversible aggregation. Furthermore, the overexpression of membrane proteins in heterologous host organisms (e.g., E. coli)2-4 typically results in unfolded proteins located in inclusion bodies in the cell. Thus, the protein needs to not only be recovered from the cell matrix, but also refolded. A number of techniques for recovery have been developed, including extraction into organic solvents,5 urea,6-10 sodium dodecyl sulfate (SDS) surfactant,11-15 and trifluoroacetic acid,11 or alternatively affinity tags can be used in conjunction with an appropriate affinity column.16-18 Refolding can then achieved by transferring the denatured protein into a “mild” detergent solution (i.e., nondenaturing, as opposed to SDS) that consists of either micelles or vesicles, allowing the native-like conformation of the protein to be reconstituted. The choice of proper extraction and refolding mixtures is by no means trivial, and often must be tailored for each “notoriously individualistic” membrane protein.19 Furthermore, the transfer step mentioned above can bring about unwanted aggregation and other deleterious effects due to the dual hydrophobic and hydrophilic regions * Corresponding author. E-mail:
[email protected].
of the protein. This has led to only a handful of membrane proteins being successfully refolded from the denatured state4,20 and is often a major hurdle in protein characterization. Bacteriorhodopsin (bR), one of the most studied membrane proteins, is the major component of purple membrane (PM) fragments isolated from Halobacterium salinarium, and functions as a light-driven proton pump. PM contains ca. 75 wt % bR and 25 wt % lipids, or equivalently 10 lipid molecules per protein.21,22 Due to this large protein concentration, PM fragments exhibit a highly ordered, paracrystalline lattice23 that, when dispersed in aqueous solutions, exist as stiff, planar sheets with an average diameter of ca. 500 nm.22,23 The high-resolution X-ray crystallographic structure of bR24 reveals seven transmembrane helices connected by extramembranous loops.25 The secondary structure of bR has been studied by FT-IR26 and found to contain about 70% R-helix structure corresponding to the transmembrane helices,27,28 with the remaining 20% β-strands and 10% unordered structures.29 A retinal chromophore, covalently linked within the helix bundle of bR via a protonated Schiff-base linkage to Lys-216 on helix G,25 undergoes photoisomerization from the all-trans to the 13-cis conformation following illumination with 560 nm visible light, representing the first step in the proton-pump photocycle (bR f K state).25,30,31 A proton is then transferred from the Schiff base to Asp-85 (L f M reaction)25,32 and subsequently released into the extracellular region, leading to the formation of the N intermediate.25,32 The Schiff base is then reprotonated from the cytoplasmic side of the membrane via Asp-96 (O state)25 and retinal is isomerized back into the all-trans state. Starting from the unfolded form in SDS micelles, bR can be refolded by rapid exchange of SDS with a milder nonionic detergent system such as DMPC/CHAPSO, DMPC/CHAPS, or DMPC/DHPC (DMPC ) 1,2-dimyristoyl-sn-glycero-3-phosphocholine; CHAPSO ) 3-[(3-cholamidopropyl)dimethylamino]2-hydroxyl-1-propane; CHAPS ) 3-[(3-cholamidopropyl)dimethylammonio]-1-propanesulfonate; DHPC ) dihexanoylphospha-
10.1021/jp807875u CCC: $40.75 2009 American Chemical Society Published on Web 06/01/2009
8570
J. Phys. Chem. B, Vol. 113, No. 25, 2009
Zhang et al.
Figure 1. (a) Miscibility of purple membrane with azoTAB under visible and UV light. Numbers represent the azoTAB concentration (mM). Cloudy mixtures can be detected by the lack of the reflections in the glass test tubes. (b) Optical microscope images of mixtures of PM with azoTAB. [bR] ) 0.18 mg/mL.
SCHEME 1: Photoisomerization of the Azobenzene-Based Surfactant
tidylcholine).33-35 Similarly, even the highly ordered paracrystalline structure of PM sheets could be reformed following removal of the natural lipids with detergents and then reconstitution through exchange of detergent with native lipids isolated from PM.36-38 In the present study, simple light illumination will be used as a novel method of controlling membrane protein folding. Photoreversible control of bR conformation will be achieved through the photoisomerization of a light-sensitive surfactant (see Scheme 1) that can be switched from the “active” to the “passive” state with light. Using this unique property, photoreversible control of soluble protein folding through light-induced molecular binding of the surfactant to the protein has resulted in the formation of a partially folding intermediate of lysozyme39 with enhanced enzymatic activity40 as well as photoreversible preamyloid oligomization in R-chymotrypsin.41 The present study will extend this methodology to membrane proteins, utilizing the photoreversible partitioning of bR between micellar (unfolded) and lipid bilayer (folded) environments that form with the active and passive forms of the photosurfactant, respectively. This concept is similar to the aforementioned rapid exchange of SDS with a milder detergent system, although in the present case the process is entirely light activated, and greatly simplifies the refolding protocol by promoting refolding versus
aggregation as the same surfactant species would be used throughout the refolding process. Several complementary methods will be used to study the effect of the photoresponsive surfactant and light conditions on bR folding, including UV-vis spectroscopy to study absorbance changes at 560 nm of the retinal chromophore, FT-IR spectroscopy to examine secondary structure changes in the protein, and DLS and 31P NMR measurements to examine the formation of PM-azoTAB mixed micelles. Experimental Details Materials. The surfactant 4-ethyl-4′-(trimethylaminobutoxy)azobenzene bromide (azoTAB) shown in Scheme 1 was synthesized as described.42,43 All other chemicals were purchased from Sigma-Aldrich and used as received unless otherwise mentioned. The dipole moment across the nitrogen double bond is ∼0.5 D for the trans isomer compared to ∼3.1 D for the cis form.44 As a result, the trans isomer is significantly more hydrophobic than the cis form, allowing photoinduced changes in a variety of surfactant properties such as critical micelle concentration,42 surface tension,45 and electrical conductivity.42 Purple Membrane Preparation. Purple membrane fragments containing both lipids and the membrane protein bacteriorhodopsin were isolated from Halobacterium salinarium
Conformation of Bacteriorhodopsin (strain S9, a kind gift from Professor Dieter Oesterhelt at the Max Planck Institute of Biochemistry in Martinsried) according to the method of Oesterhelt and Stoeckenius.22 Briefly, bacterial cells were grown in 1 L cultures incubated at 37 °C in the dark in a shaking water bath at 100 rpm for 5-6 days until the absorption at 560 nm was approximately 1.0-1.5.22,46 The purification of PM was then achieved as previously described,22,47 with PM isolated from the medium at 4000 rpm for 40 min. Impurities in PM were removed by washing with DI water at least 15 times at a speed of 24 000 rpm for 45 min (Beckman Coulter Avanti J-25 I Centrifuge, Beckman rotor JA 25.50) until the water no longer exhibited a purple color. The obtained PM wet pellet was stored at -20 °C for future use. The typical yield of the protein was 12-15 mg per liter of culture determined with an Agilent UV-vis spectrophotometer (model 8453) using an extinction coefficient of 54 000 M-1 cm-1 at 560 nm.22 UV-vis Determination of bR Folding. PM suspended in 0.1 M sodium phosphate buffer, pH 6.0, was added directly into vials with the appropriate amount of crystallized azoTAB. The samples were then illuminated for 1 h at 25 °C under gentle stirring with a 200 W mercury arc lamp (Oriel, model no. 6283) equipped with a 400 nm long pass filter (Oriel, model no. 59472), heat absorbing filter (Oriel, model no. 59060), and a fiber-bundle focusing assembly (Oriel, model no. 77557) to isolate the 436 nm mercury line. After the UV-vis spectra were recorded, the samples were then illuminated with the arc lamp equipped with a 320 nm UV filter (Oriel model no. 59980) to isolate the 365 nm mercury line, with surfactant conversion from the trans to the cis isomer assured with absorption measurements. Control experiments to expose the protein to yellow light for 4 h were achieved using a HQ580/20x filter (Chroma) to isolate the 577 nm and 579 nm mercury lines. The relative irradiances at 365, 436, and 577/579 nm are estimated to be 100%, 119%, and 112% from the lamp specifications. Wavelengths of maximum absorption were determined by fitting the spectra to a fifth-order polynomial. The concentration of the protein in each sample was ∼0.18 mg/mL, determined spectroscopically. The phase behavior of each sample was observed visually, while optical microscope images were taken with an Olympus IX71 inverted microscope equipped with 100× oil-immersion objective lens (UPlanFl, N.A. ) 1.3). The dynamics of photoinduced bR conformational changes were studied by in situ UV-vis spectroscopy, using a surfactant concentration of 0.10 mM to ensure rapid surfactant isomerization. The results were obtained over a period of 20 min with spectra collected every 0.5 s. Measured spectra were corrected for scattering effects using a spline routine with a baseline choice of 700-900 nm and 470-480 nm under visible illumination and 700-900 nm and 380-385 nm under UV exposure. FT-IR Measurements. Approximately 0.4 mL of the purified PM in H2O was washed with 5.0 mL of D2O at least three times at the speed of 24 000 rpm for 45 min to remove residual H2O. The sample was then suspended in 0.1 M sodium phosphate D2O buffer (pD ) 6.4, measured with a standard pH electrode and corrected according to pD ) pH + 0.4) at a protein concentration of 7.8 mg/mL detected with UV-vis spectroscopy. Adding the PM suspension into vials containing crystallized azoTAB generated a series of surfactant concentrations. The samples were then gently stirred for 24 h in the dark prior to the FT-IR measurements. Infrared spectra were measured with a Genesis II FT-IR spectrometer (Mattson Instruments). Samples were loaded between a pair of CaF2 windows using a 50 µm Teflon spacer
J. Phys. Chem. B, Vol. 113, No. 25, 2009 8571 with water-jacket circulation used to maintain the temperature at 20 °C. The FT-IR sample chamber was purged with dry air for 12 h prior to and during measurements. The samples were continuously illuminated with either visible or UV light from the mercury arc lamp using the fiber-bundle focusing assembly as previously described.48 For each spectrum, 1000 interferograms were collected with a 2 cm-1 resolution. The visible and UV filter sets were then alternated every 3 h to measure lightinduced changes in protein folding. Difference spectra were obtained by subtracting the spectra measured under visible light from that taken with UV illumination. Photoreversibility was assured through multiple visible T UV light cycles. The protein absorbance was obtained by subtracting the spectrum measured for pure buffer, with the resulting corrected spectra flat in the region between 1900 and 1750 cm-1. The technique of Fourier self-deconvolution (FSD) was applied to the original spectra to resolve the overlapping bands in the amide I region using a band-narrowing factor k ) 2.2 and a full width at half-height of 12.5. NMR Measurements. 31P NMR measurements were performed with a Bruker AMX 500-MHz spectrometer operating at 201.5 MHz and room temperature with continuous 1H decoupling and 33 K data points in the transformed spectra. The data were analyzed using the Nuts program (Acorn NMR Inc.). PM suspended in D2O at a protein concentration of 7.5 mM (determined spectroscopically22) was added directly into vials containing the appropriate amount of crystallized azoTAB to achieve the desired concentrations. The samples were prepared and stirred in the dark for 24 h before the measurement. After the NMR measurements were recorded for trans-azoTAB (dark state), the samples were then illuminated with the arc lamp equipped with a 320 nm UV filter (Oriel model no. 59980) for another 24 h with gentle stirring. Following this conversion of azoTAB to the cis state, the samples were maintained in the dark for no more than 2 h prior to and during measurements, resulting in only minimal thermal conversion of azoTAB back to the trans state (confirmed with UV-vis). Dynamic Light Scattering Measurements. Dynamic light scattering measurements were performed at 25 °C on a Brookhaven model BI-200SM instrument (Brookhaven Instrument Corp.) equipped with a BI-9000AT digital correlator (Brookhaven), a 35 mW HeNe (632.8 nm) laser (Melles Griot, model no. 05-LHP-928), and an BI-APD avalanche photodiode detector (Brookhaven). PM suspended in 0.1 M D2O phosphate buffer, pH 6.0, at a protein concentration of 0.18 mM (determined spectroscopically) was directly added into vials with appropriate amounts of crystallized azoTAB. The samples were then stirred in the dark for 24 h before the measurement. Samples were measured as made without filtration to avoid removal of the large purple membrane fragments, with dust allowed to settle in the instrument until the count rate stabilized. After measurements in the presence of trans-azoTAB (dark state), the samples were then exposed to UV light from the arc lamp for 24 h with gentle stirring. In order to maintain azoTAB in the cis state, UV light from the arc lamp was illuminated onto the sample with the fiber bundle focusing assembly during the entire measurement. The dynamic light scattering data were analyzed with the nonnegative least-squares (NNLS) routine supplied by Brookhaven. Typically, at least five independent scattering runs were conducted for each solution, with the average of all the runs reported. Since the samples could not be filtered, it was ensured that the scattering count rate remained approximately constant over time, while flat baselines for each correlation function were maintained over 1-2 decades in
8572
J. Phys. Chem. B, Vol. 113, No. 25, 2009
Zhang et al.
Figure 2. (a) Normalized correlation function of PM (0.18 mg/mL) mixed with azoTAB (20 mM) in 0.1 M phosphate buffer as a function of light illumination. The presence of azoTAB-PM mixed micelles is evident from the fast relaxation at low values of q2τ. (b) 31P NMR measurements as a function of azoTAB concentration and light illumination. [PM] ) 7.5 mg/mL.
relaxation times. For smaller micellar species, the scattering angle was lowered to 45° in order to observe the full correlation function (i.e., a flat decade at low relaxation times). Hydrodynamic radii, RH, were calculated from the Stokes-Einstein equation RH ) kBT/6ηπD, where kB is Boltzmann’s constant, T is temperature, η is the viscosity of the solvent, and D is the experimentally determined diffusion coefficient. Results and Discussion Phase Behavior of Purple Membrane with azoTAB. The miscibility of purple membrane fragments (PM) with azoTAB is shown in the test tube images in Figure 1a. At azoTAB concentrations below 0.35 mM, transparent suspensions are obtained (evident from the reflections in the glass test tubes), while cloudy samples are observed as the surfactant concentration reached 0.50 mM under both visible and UV light. Under visible light, the mixtures remained cloudy until trans-azoTAB concentrations of 3.0 mM or greater were surpassed, while with UV exposure cis-azoTAB concentrations of 12.0 mM are required for miscibility. This phase behavior phenomena suggests that the classic three-stage model of detergent-lipid membrane interactions may be in operation.49,50 During stage I of this model at low surfactant concentrations (
