Gramicidin structure and disposition in highly curved membranes

Journal of

Structural Biology Journal of Structural Biology 150 (2005) 23–40 www.elsevier.com/locate/yjsbi

Gramicidin structure and disposition in highly curved membranes W. Liu c, M. Caffrey a,b,c,d,* a

College of Science, University of Limerick, Limerick, Ireland Biochemistry Program, The Ohio State University, Columbus, OH 43210, USA c Biophysics Program, The Ohio State University, Columbus, OH 43210, USA Department of Chemistry, The Ohio State University, Columbus, OH 43210, USA

b

d

Received 12 September 2004, and in revised form 17 December 2004 Available online 2 February 2005

Abstract With a view to deciphering aspects of the mechanism of membrane protein crystallization in lipidic mesophases (in meso crystallization), an examination of the structure and disposition of the pore-forming peptide, gramicidin, in the lipidic cubic phase was undertaken. At its simplest, the cubic phase consists of lipid and water in the form of a molecular Ôsponge.Õ The lipid exists as a continuous, highly curved bilayer that divides the aqueous component into two interpenetrating but non-contacting channels. In this study, we show that gramicidin reconstitutes into the lipid bilayer of the cubic phase and that it adopts the channel, or helical dimer, conformation therein. Fluorescence quenching with brominated lipid was used to establish the bilayer location of the peptide. Electronic absorption and emission spectroscopies corroborated this finding. Peptide conformation in the cubic phase membrane was determined by circular dichroism. The identity and microstructure of the mesophases, and their capacity to accommodate gramicidin and other system components (sodium dodecyl sulfate, trifluoroethanol), was established by small-angle X-ray diffraction. Beyond a limiting concentration, gramicidin destabilized the cubic phase in favor of the inverted hexagonal phase. While gramicidin remained bilayer bound as membrane thickness changed, its conformation responded to the degree of bilayer mismatch with the hydrophobic surface of the peptide. These findings support the hypothesis that reconstitution into the lipid bilayer is an integral part of the in meso crystallization process as applied to membrane proteins. They also suggest ways for improving the process of membrane protein crystallogenesis. Ó 2005 Elsevier Inc. All rights reserved. Keywords: Bilayer thickness mismatch; Circular dichroism; Crystallization; Cubic phase; Fluorescence quenching; X-ray diffraction

1. Introduction A tried and true means for understanding how a protein functions is by knowing its three-dimensional structure in atomic detail. This applies to soluble, to structural, and to membrane proteins alike. In the case of the latter, macromolecular crystallography is still the method of choice for structure determination. Its use, however, requires the availability of diffraction-quality crystals. These have proven difficult to procure, and it is the dearth of crystals that, in part, accounts for the relatively few *

Corresponding author. Fax: +1 614 292 1532. E-mail address: martin.caff[email protected] (M. Caffrey).

1047-8477/$ - see front matter Ó 2005 Elsevier Inc. All rights reserved. doi:10.1016/j.jsb.2004.12.007

membrane proteins whose structure is known to high resolution. Given that close to a third of the genome codes for membrane proteins and that over two-thirds of the drugs on the market today target membrane proteins, the need for detailed structures is immediate. Several approaches are available for crystallizing membrane proteins (Hunte and Michel, 2003). A relatively recent addition to the arsenal is a method that makes use of lipidic liquid crystals or mesophases (Landau and Rosenbusch, 1996). This is referred to here as the Ôin mesoÕ method. A mechanism for how in meso crystallization works has been proposed (Caffrey, 2000). It posits that the purified and detergent-solubilized membrane protein reconstitutes into the lipid bilayer of the

24

W. Liu, M. Caffrey / Journal of Structural Biology 150 (2005) 23–40

hosting mesophase, which usually, but not always (Misquitta and Caffrey, 2003), is of the cubic phase type. Crystallants or precipitants, of which there is a wide variety (McPherson, 1999), serve in part to alter the curvature of the membrane and to effect a phase separation of sorts where one of the phases takes the form of a protein crystal. Passage of the reconstituted protein from the bulk cubic phase to the surface of the crystal is facilitated by a lamellar type arrangement in the immediate vicinity of the crystal that acts as a conduit or portal (see Fig. 3 in Caffrey, 2000). Evidence in support of this aspect of the hypothesis has been obtained recently in the course of attempts to crystallize the acetylcholine receptor–bungarotoxin complex (Paas et al., 2003). With a view to exploiting this and other methods for crystallizing proteins in a rational way, it is imperative that the mechanism of the underlying process be understood. This paper is an attempt to do just that by testing one aspect of the proposed mechanism for in meso crystallization. Specifically, we are interested in determining if reconstitution is an integral part of the process in going from purified protein to protein crystal by way of a lipidic mesophase. By reconstitution is meant that the protein, freed of its native membrane and typically solubilized in detergent, becomes incorporated into the lipid bilayer of the hosting mesophase. The assumption is that the protein adopts a conformation and an orientation similar to that in the native membrane with its hydrophobic ÔwaistbandÕ suitably solvated by the apolar interior of the lipid membrane. To date, several membrane proteins have been successfully crystallized by the in meso method and used for high-resolution structure determination. These include bacteriorhodopsin (Luecke et al., 1998), halorhodopsin (Kolbe et al., 2000), sensory rhodopsin (Luecke et al., 2001), the sensory rhodopsin/transducer complex (Gordeliy et al., 2002), and the photosynthetic reaction center from Rhodobacter sphaeroides (Katona et al., 2003). Other proteins have produced crystals or microcrystals that diffract poorly or do not diffract at all. These include the acetylcholine–bungarotoxin complex (Paas et al., 2003), the reaction center from Blastochloris (formerly Rhodopseudomonas) viridis (Chiu et al., 2000), and the light harvesting 2 complex (Caffrey, 2003; Chiu et al., 2000). Recently, the method has been used to grow diffraction-quality crystals of a b-barrel protein, the bacterial outer membrane Vitamin B12 transporter, BtuB (Misquitta et al., 2004a). Ideally, we would like to investigate the proposed mechanism of in meso crystallization using one of the aforementioned proteins. The reality of doing so is fraught with difficulties, however. These have to do with the fact that without exception, the proteins whose structure has been determined to date by the in meso method are complexes that include one or several cofactors, pigments in particular. Mechanism testing relies heavily on

spectroscopy. Unfortunately, interpretation of such data is non-trivial and complicated by the electronic coupling that exists between the components of these proteins. Therefore, it was decided to perform the first round of hypothesis testing with the relatively simple pentadecapeptide, gramicidin (Fig. 1), which is currently in in meso and other crystallization trials (unpublished data). Gramicidin is an antibiotic produced by Bacillus brevis (Wallace, 1998). It acts by creating pores in membranes, rendering them leaky to small cations and incapable of supporting life-sustaining transmembranal gradients. There are several advantages to using gramicidin in the current study. The most important of these include the fact that it is a well-characterized peptide that has several aromatic residues and is devoid of cofactors. Furthermore, the peptide is very stable (Greathouse et al., 1994) and easy to handle. In this study therefore we set out to answer the question, Does gramicidin reconstitute into the highly curved bilayer of the cubic phase (schematic shown in Fig. 2) under conditions used for in meso crystallization? Advantage was taken of the tryptophan content of gramicidin which was used as a spectroscopic probe to investigate the disposition of the peptide within the cubic phase. To this end, electronic absorption and fluorescence emission measurements were combined with quenching studies, circular dichroism spectroscopy, and small-angle X-ray diffraction. The data support the view that reconstitution does occur. They also show that the peptide is sensitive to the identity of the lipid that constitutes the cubic phase reflecting hydrophobic mismatch between the bilayer and the channel form of gramicidin.

2. Materials and methods 2.1. Materials Monoolein (9.9 MAG,1 lots M239-JA22-N and M239MA26-N, 356 g/mol), monopalmitolein (9.7 MAG,2 lot 1

The monoacylglycerol (MAG) chain shorthand notation used here is based on the N.T system introduced previously (Misquitta et al., 2004b). In this system, N refers to the Neck of the chain extending from and including the carbonyl carbon to the first carbon of the olefinic bond. T refers to the Tail of the chain extending from and including the second olefinic carbon to the carbon of the methyl terminus. The sum N + T is the total number of carbon atoms in the chain. Thus, 9.9 MAG is monoolein which has an acyl chain 18 carbon atoms long and a double bond between carbons 9 and 10. 2 Abbreviations used: 9.7 MAG, monopalmitolein; 9.9 MAG, monoolein; 11.9 MAG, monoeicosenoin; A, absorbance; bromo-MAG, 2, 3-dihydroxypropyl (7Z)-9,10-dibromooctadecanoate; CD, circular dichroism; DOPC, 1,2-dioleoyl-sn-glycero-3-phosphocholine; F, uncorrected fluorescence; Fc, fluorescence corrected for the inner filter effect; FQ, fluorescence quenching; I, inner filter correction factor; SDS, sodium dodecyl sulfate; SPP, 25 mM sodium potassium phosphate buffer, pH 5.60; TFE, trifluoroethanol; TLC, thin layer chromatography.

W. Liu, M. Caffrey / Journal of Structural Biology 150 (2005) 23–40

25

Fig. 1. Gramicidin models. (A,B). Space-filling model of the channel or helical dimer form of gramicidin A based on PDB entry 1JNO. Trypotophans in the two monomers are shown in purple and cyan, and are identified by their position in the chain. Dimensions are based on measurements of the molecular model performed using Protein Explorer. (A) View from within the lipid bilayer along the dimeric twofold axis. (B) View perpendicular to the bilayer plane where the pore through the dimer can be seen. Non-tryptophan hydrogen, carbon, nitrogen and oxygen atoms are shown in white, grey, blue, and red, respectively. Cartoon representations of gramicidin displaying the polypeptide backbone fold in the helical dimer (channel or pore) conformation (C) and in a double helical conformation (D). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this paper.)

M219-JY22-N, 328 g/mol), and monoeicosenoin (11.9 MAG, lot M274-N26-M, 384 g/mol) were purchased from Nu Chek Prep (Elysian, MN). 2,3-Dihydroxypropyl (7Z)-9,10-dibromooctadecanoate (bromo-MAG, lots 180BR-10, 180BR-11, and 180BR-13, 516 g/mol) and 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC, lot 181PC-138, 786.12 g/mol) were obtained from Avanti Polar Lipids (Alabaster, AL). Trifluoroethanol (TFE, lot 023K3647, 100 g/mol), L-tryptophan (lot 119H0344, 204.2 g/mol), and gramicidin D (lots 121K1236 and 023K3647, 1,880 g/mol) were from Sigma (St. Louis, MO). Gramicidin D is a natural mixture of gramicidin A (80%), gramicidin B (6%), and gramicidin C (14%). The amino acid sequence of gramicidin A is: formylNH-L-Val-Gly-L-Ala-D-Leu-L-Ala-D-Val-L-Val-D-Val11 L-Trp-D-Leu-L-Trp -D-Leu-L-Trp-D-Leu-L-Trp-CO-NHCH2-CH2-OH. In gramicidin B and C, Trp at position 11 is replaced by L-Phe and L-Tyr, respectively (Townsley et al., 2001). Sodium dodecyl sulfate (SDS, lot 100146, 288 g/mol) was purchased from Boehringer–Mannheim (Indianapolis, IN). Ethyl alcohol (200 proof, analytical

grade) was purchased from Quantum Chemical (Tuscola, IL) and was used in co-dissolution studies as described below. Sodium potassium phosphate was from Fluka (Buchs, SG, Switzerland). For the thin layer chromatography work, solvents of analytical grade from Fisher Scientific (Pittsburgh, PA) were used. Quartz cuvettes (0.01, 0.1, 1, and 3 mm path length) were from Hellma International (Plainview, NY). Syringes were from Fisher Scientific (Hamilton 81030, 100 lL gas-tight). Water, with a resistivity of >18 MX cm, was purified by using a MilliQ Water System (Millipore, Bedford, MA) consisting of a carbon filter cartridge, two ion exchange filter cartridges, an organic removal cartridge, and a final filter (Sterile Millipore, millipak 40, lot F2PN84024). 2.2. Methods 2.2.1. Thin layer chromatography Thin layer chromatography (TLC) of fresh (unused) 9.7 MAG, 9.9 MAG, 11.9 MAG, and bromo-MAG was used to determine lipid purity. For this purpose,

26

W. Liu, M. Caffrey / Journal of Structural Biology 150 (2005) 23–40

Pn3m or cubic-Ia3d phase. The actual compositions of the samples used in this study are as follows: 40%(w/ w) SPP for 9.9 MAG and 11.9 MAG (Briggs, 1994; Qiu and Caffrey, 2000), 48%(w/w) SPP for 9.7 MAG (Qiu, 1998), and 35%(w/w) SPP for bromo-MAG (unpublished data). To verify that the mesophase sought was actually obtained, aliquots of the above samples were sealed in 1 mm quartz X-ray capillaries and used in diffraction measurements, as described below.

Fig. 2. Cartoon representation of the cubic-Pn3m phase with the channel form of gramicidin reconstituted into its lipid bilayer and dissolved as a mixed micelle in the aqueous channel. Dimensions are based on the cubic phase formed by 9.9 MAG at 39%(w/w) water and 20 °C. Gramicidin is shown as a space-filling model taken from Fig. 1A. Lipids and detergents are shown as lollipop figures with the pop part representing the polar headgroup and the stick representing the acyl or alkyl chain. The shaded beige and yellow ovals represent, respectively, the glycerol headgroups of the monoacylglycerols and the sulfate headgroups of sodium dodecyl sulfate. The continuous blue zones are the aqueous channels that permeate the cubic phase. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this paper.)

1, 5, 50, and 100 lg samples of lipid dissolved in chloroform were run on Adsorbosil Plus plates (silica gel H, 16385, lot 146794, Alltech, Deerfield, IL) using three different solvent systems: chloroform/acetone (96/4, v/v), chloroform/acetone/methanol/acetic acid (73.5/25/1/ 0.5 by vol.), and hexane/ethyl acetate/acetone (73.5/ 1.5/25 by vol.). The plates were pre-run twice in chloroform/methanol (10/1, v/v). Spots were visualized by spraying with 4.2 M sulfuric acid followed by charring on a hot plate (Type 2600, Thermolyne, Dubuque, IA) at 250 °C. Estimated purity of the lipid was in excess of 99%. 2.2.2. Sample preparation 2.2.2.1. Cubic phase samples. All samples were prepared at room temperature (20–24 °C). For the bulk of the studies, 25 mM sodium potassium phosphate buffer, pH 5.60 (SPP), was used as the dispersing lyotrope. Lipids were removed from the freezer at
70 °C and thawed at room temperature (bromo-MAG, 9.7 MAG), 40 °C (9.9 MAG) or 50 °C (11.9 MAG). Molten lipid was transferred to the syringe mixing device and the cubic phase was prepared by mechanical homogenization following published procedures (Cheng et al., 1998; Qiu and Caffrey, 1998). To obtain optically clear samples for spectroscopic measurements, sample composition was adjusted to ensure formation of a pure cubic-

2.2.2.2. Reconstitution of gramicidin into the cubic phase. Direct mixing. Stock solutions of gramicidin in molten MAGs were prepared as follows. The lipid (9.7 MAG, 9.9 MAG, and 11.9 MAG) was first melted and solid gramicidin was added to a known final concentration. A true solution was obtained following sonication at full power in a bath sonicator (Model G112SP1T, Power Laboratory Supplies, Hicksville, NY) at room temperature for 10 min. Samples were stored under argon at
20 °C. To prepare the cubic phase, stock solutions were diluted with additional molten MAG (9.7 MAG, 9.9 MAG, 11.9 MAG, and bromo-MAG) to the required absolute and relative concentration of peptide and/or quenching lipid. The molten mix was used subsequently in cubic phase sample preparation, as described above, with SPP buffer as the lyotrope. Co-dissolution mixing. Gramicidin and lipid (typically 100 mg/mL) were co-dissolved in organic solvent (ethanol, methanol, and chloroform) in appropriate molar ratios. The samples were dried in a stream of inert (nitrogen or argon) gas and were further dried under high vacuum (VLP120, Savant Instruments, Holbrook, NY) at 1.5 mTorr and 20 °C for at least 48 h. The dried material was subsequently used in cubic phase preparation as above with SPP buffer as the lyotrope. Detergent mixing. A TFE solution containing 75 mg gramicidin/mL was prepared. To this was added a micellar dispersion of 0.02 or 0.25 M SDS in SPP buffer. The gramicidin/SDS mix was sonicated to clarity in a bath sonicator at full power for 20 min at room temperature. The composition of the final solution was 0.75 mg gramicidin/mL, 1%(v/v) TFE, and 0.02 or 0.25 M SDS. It was used to prepare the cubic phase by mechanical mixing, as described above. Vesicle reconstitution. Vesicles were prepared by initially co-dissolving the gramicidin and DOPC (1:10 mole ratio) in ethanol at a concentration of 10 mg peptide/ mL. Solvent was removed in a stream of dry nitrogen followed by overnight drying under high vacuum (1.5 mTorr, Savant) at room temperature. The dry gramicidin/lipid mix was dispersed in SPP buffer to a final lipid concentration of 2.5 mM by vortex mixing in a nitrogen atmosphere. Bath sonication at full power and at room temperature for 30 min as above produced a dispersion of vesicles. This was centrifuged for 10 min

W. Liu, M. Caffrey / Journal of Structural Biology 150 (2005) 23–40

27

at 14 000g (Model MR 14.11, fixed angle rotor, Jouan, Winchester, VA) at room temperature and the optically clear supernatant was used as the source of sonicated vesicles. The gramicidin concentration in the supernatant was determined by measuring its absorbance at 280 nm and using a molar extinction coefficient of 20 840 M
1 cm
1 (experimentally determined in-house). 2.2.3. Spectroscopic measurements 2.2.3.1. UV–visible absorption. Spectra were recorded with a Uvikon XL dual beam spectrophotometer (Research Instruments Intl., San Diego, CA). Data were collected from 750 to 250 nm in 1 nm steps at 100 nm/ min with air as the reference. The absorption spectrum of a lipid/SPP buffer dispersion or SPP buffer alone in a quartz cuvette (Hellma) of suitable path length recorded against air was subtracted from sample spectra as appropriate. Loading cells with the highly viscous cubic phase was achieved in different ways depending on the cuvette used. In all cases, a flat tipped 22-gauge needle (point style 3, inner diameter, 0.41 mm; outer diameter, 0.72 mm) was used to transfer the homogenous sample from the microsyringe mixing device to the cell. In the case of the 3 mm path length cuvette, the loaded cell was capped with Parafilm and then spun at 14 000g as above for 10 min at room temperature. The centrifugation step was not required when 1 mm path length cells were used. By carefully dispensing into the cuvette, a continuous column of cubic phase could be procured. The 0.1 and 0.01 mm path length cuvettes consist of a front and a back window. To load, a small volume of cubic phase was placed on one of the windows in the region to be interrogated by the light beam and immediately a sandwich was made by covering it with the second window. These were held together in a springloaded cuvette adapter. The UV absorbance of the lipid was considerable (Fig. 3). This problem of ÔbackgroundÕ absorbance is exacerbated in the current application by virtue of the fact that most of the cubic phase samples used are close to 60%(w/w) or 1–2 M lipid. Thus, a small amount of a highly absorbing contaminant in the lipid will contribute significantly to the background signal. Effort was made to find commercial and homemade 9.9 MAG with lower absorbance, to no avail. By TLC, the lipid is 99% pure. The precise origin of the absorbance signal is not known. Thus, to quantify the spectral properties of additives such as tryptophan or gramicidin, all spectra were recorded against an air reference. The analyte spectrum was obtained subsequently by subtracting the lipid/buffer background from the analyte/lipid/buffer sample spectrum. 2.2.3.2. Fluorescence. Emission spectra were recorded using an SFM3 fluorimeter (Bio-Logic Science Instru-

Fig. 3. Absorption (dashed line) and corrected fluorescence emission spectra (solid line) of the cubic-Pn3m phase of hydrated 9.9 MAG. Samples were prepared at 40%(w/w) SPP buffer as described under Section 2.2. The absorption and fluorescence spectra were recorded in quartz cuvettes with a path length of 3 mm at room temperature. The absorption spectrum was recorded with air in the reference beam. The excitation wavelength used for the fluorescence spectrum was 305 nm. Peaks in the fluorescence spectrum are labeled. Other conditions and details are as described under Section 2.2.

ments, Claix, France) that included a 150-W Hg/Xe lamp (L2482, Hamamatsu, Japan) and Jobin Yvon monochromators (MM-200 driving unit, BH10 UV monochromators). One millimeter slits were placed at the entrances to and exits from the excitation and emission monochromators providing an 8 nm band pass. Light intensity was measured with a photomultiplier tube (Bio-Logic PMT 200) with an integration time of 100 ms. Excitation spectra were recorded with an exciting wavelength of 305 nm. The latter was chosen to selectively excite tryptophan in the gramicidin with minimal contribution from phenylalanine and tyrosine and to minimize photobleaching. Since 305 nm is on the red edge of the tryptophan absorption peak, possible inner filter effects (see below) are lessened. Additionally, this wavelength is close to a peak in the Hg/Xe lamp emission spectrum at 302 nm which provides for a stronger signal. Generally, emission spectra were recorded over a limited range from 340 to 320 nm (1 nm steps, 100 nm/min) to minimize bleaching. When sample absorption exceeds 0.1, fluorescence intensity no longer rises linearly with concentration. A correction for this so-called inner filter effect (I) has been applied to all relevant data following established procedures (Lakowicz, 1983). Thus, I ¼ 10½Aðkex ÞþAðkem Þ=2 ; where A (kex) and A (kem) are the absorption of the sample at the excitation and emission wavelengths, respectively. The corrected fluorescence (Fc) is calculated as:

28

W. Liu, M. Caffrey / Journal of Structural Biology 150 (2005) 23–40

Fc ¼ IF; where F is the recorded fluorescence value. Just like the problem encountered with lipid UV absorbance, we also had to contend with background fluorescence from the lipid. This is illustrated in Figs. 3 and 4. All of the MAGs used in the study, including the bromo-MAG, had background fluorescence. It was corrected for by subtracting spectra recorded using reference samples which included everything except analyte. 2.2.3.3. Circular dichroism. The conformation of gramicidin in solution and reconstituted into various lipid dispersions was determined by CD (Wallace, 1998). Spectra were recorded using an AVIV Circular Dichroism Spectrometer (Model 202, Protein Solutions, Lakewood, NJ) in the range from 250 to 190 nm in 1 nm steps with an equilibration time of 2 s per step. The bulk of the solution work was done using 1 mm path length quartz cuvettes. Because the MAG samples have high background absorbances in the UV region (see above), the CD spectra were noisy when collected using 1 mm cells. Accordingly, the bulk of the cubic phase data was acquired using either 0.01 or 0.1 mm path length cuvettes which were loaded as described above. CD data on sonicated DOPC vesicles containing gramicidin were made using 1 mm path length cells. All measurements were made after a 1 h equilibration at 25 °C.

For the bulk of the spectroscopic measurements described above, samples were prepared and analyzed in triplicate. 2.2.4. X-ray diffraction Phase characterization was carried out using low- and wide-angle X-ray diffraction. The samples used were transferred to 1 mm quartz capillary tubes (Hampton Research, Laguna Niguel, CA) and flame-sealed using a propane/oxygen torch (Smith Equipment, Watertown, SD). A bead of 5-min epoxy (Devcon, Danvers, MA) was applied to protect and ensure the integrity of the flame seal. Samples were stored at room temperature for at least 1 day and then incubated at 20 °C for at least 4 h before being used in diffraction measurements. Measurements were performed using a rotating anode X-ray generator (Rigaku RU-300 operating at 45 kV and 250 mA) producing Ni-filtered Cu Ka radia˚ ) as described (Cherezov tion (wavelength k = 1.5418 A et al., 2002a). Sample-to-detector distance (typically 340 mm) was measured using a silver behenate standard ˚ ; Blanton et al., 1995). Samples were contin(d001, 58.4 A uously translated at a rate of 2 mm/min back and forth along a 2 mm section of the sample to average the contributions to total scattering from different parts of the sample and to minimize possible radiation damage effects (Cherezov et al., 2002b). The temperature inside the sample holder (Zhu and Caffrey, 1993) was regulated by two thermoelectric Peltier effect elements controlled by a computer feedback system. Measurements were performed at 20.0 ± 0.05 °C. A typical exposure time was 30 min. Diffraction pattern registration on high-resolution image plates and subsequent analysis have been described (Cherezov et al., 2002a).

3. Results 3.1. Direct reconstitution

Fig. 4. Corrected fluorescence emission spectra of tryptophan in SPP buffer and in the cubic phase of hydrated 9.9 MAG. Reference spectra for the buffer and the cubic phase without tryptophan are included. The concentration of tryptophan in the buffer alone and in the buffer used to prepare the cubic phase was 1.70 and 4.25 mM, respectively. The cubic phase sample consists of 40%(w/w) buffer and 60%(w/w) lipid. Other conditions are as described in the legend to Fig. 3 and under Section 2.2. Buffer (solid line); 9.9 MAG cubic phase (dashed line); Trp in buffer (dotted line); and Trp in 9.9 MAG cubic phase (dash dotted line).

Gramicidin is a hydrophobic peptide that is soluble in apolar organic solvents. It is likely therefore to be soluble in molten monoolein (9.9 MAG), the lipid upon which the in meso method for membrane protein crystallization is based (Landau and Rosenbusch, 1996). If this is so, then combining the solution of gramicidin in molten monoolein with aqueous buffer in the correct ratio at room temperature is likely to give rise to the cubic phase with the gramicidin reconstituted into its lipid bilayer. For purposes of discussion, a cartoon representation of gramicidin reconstituted in the cubic phase membrane is shown in Fig. 2. In essence, once the cubic phase with its distinct lipidic and aqueous compartments forms, the gramicidin really has nowhere else to go but to partition or incorporate into the lipid bilayer. In what follows, we provide experimental proof that this occurs.

W. Liu, M. Caffrey / Journal of Structural Biology 150 (2005) 23–40

To initiate the project, the solubility of gramicidin in molten 9.9 MAG was tested by visual inspection, as described under Section 2.2. An optically transparent and colorless solution was obtained up to concentrations of 5 mg gramicidin/mL of molten monoolein at room temperature. The absorption, fluorescence (Fig. 5), and CD spectra (Fig. 6A) of such a solution have been recorded. Comparison spectra for gramicidin in an assortment of organic solvents are shown in Fig. 7. The spectral features of gramicidin in molten 9.9 MAG are similar to those of the peptide in methanol and in ethanol. The lack of scatter, as is evident in the absorption spectra in the 400–750 nm range (data not shown), is consistent with the material forming a true solution in molten lipid under these conditions. The gramicidin/molten 9.9 MAG solution was used to prepare the cubic phase following a standard mixing protocol (Section 2.2). The mesophase produced had all the characteristics of the cubic phase in that it was optically isotropic, transparent and extremely viscous. Verification that the phase so formed is of the cubic-Pn3m type was obtained by using small-angle X-ray diffraction (Fig. 8). The latter shows that the cubic phase is retained up to 1 mol gramicidin/20 mol 9.9 MAG and that its lattice parameter is sensitive to gramicidin loading. Beyond this level of added gramicidin, the inverted hexagonal (HII) phase forms (Fig. 8). Care was taken to perform all subsequent measurements at gramicidin concentrations where the cubic phase is stable. Since gramicidin is such an apolar molecule, our working hypothesis was that following the above in meso protocol the peptide would undergo reconstitution into the bilayer of the cubic phase. Its spectral properties, as shown in Fig. 7, support this view. Of particular note is the fact that the fluorescence of the tryptophan in

Fig. 5. Absorption (dashed line) and corrected fluorescence emission spectra (solid line) of gramicidin D in molten 9.9 MAG. Samples were prepared at 0.37 mg gramicidin/mL lipid. All other conditions are as described in the legend to Fig. 3 and under Section 2.2.

29

gramicidin is decidedly blue shifted (by 12 nm) and of enhanced intensity compared to tryptophan in aqueous solution (Fig. 9). Both properties are consistent with the peptide being associated with the apolar interior of the bilayer. The hypothesis was further tested by means of intrinsic tryptophan fluorescence quenching. To this end, the reconstitution experiment was repeated using molten lipid that contained varying ratios of 9.9 MAG and its dibrominated, and thus quenching (OÕKeeffe et al., 2000), analog referred to as bromo-MAG. Prior to the quenching experiments, it was established using smallangle X-ray diffraction that the bromo-MAG formed the cubic phase and that it is completely miscible with 9.9 MAG in the liquid crystalline state. The data (not shown) demonstrate that the cubic-Pn3m phase prevails over the entire range of concentrations from 100% 9.9 MAG to 100% bromo-MAG. For the quenching study, tryptophan fluorescence was recorded as a function of bromo-MAG/9.9 MAG mole ratio in the cubic phase. The data, presented in Fig. 10, show that gramicidin fluorescence is quenched almost fully (95%) in a sample containing 90 mol% bromo-MAG. The quenching profile, as well as the degree of quenching, is consistent with the hypothesis that the peptide is membrane bound. 3.2. Conformation of gramicidin in the cubic phase membrane Gramicidin functions to create pores in membranes through which water and small cations can passage. The helical dimer is considered to be the active species that creates such transient holes (Figs. 1A–C). An alternative, double helical conformation (Fig. 1D) can be adopted in a more polar environment (Chen and Wallace, 1997). The two conformations are distinguished by their characteristic electronic CD spectra in the UV region. Accordingly, CD spectroscopy was used to probe the conformation adopted by gramicidin in the cubic phase membrane of 9.9 MAG. Reference spectra were recorded for gramicidin in methanol (Fig. 6B), in ethanol (Fig. 6C), in molten 9.9 MAG (Fig. 6A), and in small unilamellar dioleoyl phosphatidylcholine (DOPC) vesicles (Fig. 6J) where one or other of the two conformations are stabilized (Killian et al., 1988). For example, when reconstituted into the bilayer of DOPC vesicles, gramicidin adopts the helical dimer conformation. In contrast, in methanol, ethanol, and molten 9.9 MAG the spectral properties reveal the double helix conformation. However, a given system can have different relative amounts of the two conformations (Wallace et al., 1981). The relative amounts present can be gauged by measuring the ratio of the CD signal intensity associated with the positive peak at 220 nm to that of the negative peak at 230 nm, as discussed below.

30

W. Liu, M. Caffrey / Journal of Structural Biology 150 (2005) 23–40

Fig. 6. Circular dichroic properties of gramicidin D in molten 9.9 MAG (A), in methanol (B), in ethanol (C), in the cubic phases of 9.7 MAG (D), 9.9 MAG (E), and 11.9 MAG (F) with (G,H) and without SDS (I), and in small unilamellar vesicles of DOPC (J). Spectra were recorded at 25 °C in cuvettes with the following path lengths: 0.01 mm (B), 0.1 mm (A,D–I), and 1 mm (C,J). Spectra (D)–(J) are consistent with the helical dimer (channel) conformation for the peptide while spectra (A)–(C) suggest the double helical form. The methods used to reconstitute gramicidin into the cubic phase are as follows: direct (D–F), co-dissolution (I), and detergent mixing methods (G, H). SDS was used at a concentration of either 0.02 M (G) or 0.25 M (H). The prevailing cubic phase was of the Pn3m (D–G, and I) or Ia3d type (H). The negative peak (at 230 nm) to the positive peak (at 220 nm) intensity ratio is
1.54,
4.30,
10.43, and
1.89 in spectra (D), (E), (F), and (I), respectively. See text for details.

W. Liu, M. Caffrey / Journal of Structural Biology 150 (2005) 23–40

Fig. 7. Absorption (A) and corrected fluorescence emission spectra (B) of gramicidin D in different organic solvents and dispersions. Solvents include methanol (red), ethanol (green), TFE (brown dashed line), and molten 9.9 MAG (blue) where the peptide concentration was 0.3 mg/ mL. Dispersions include SDS micelles (cyan) and the cubic-Pn3m phase of 9.9 MAG at 40%(w/w) SPP buffer (magenta). The SDS concentration was 0.25 M. All other conditions are as described in the legend to Fig. 3 and under Section 2.2. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this paper.)

The CD spectrum of gramicidin reconstituted into the cubic-Pn3m phase of hydrated 9.9 MAG is presented in Fig. 6E. It exhibits dichroism characteristics of the helical dimer suggesting that its conformation in this setting is akin to that in the membrane of DOPC vesicles (Fig. 6J). Sonicated phosphatidylcholine vesi˚ (Huang, cles have a limiting diameter of about 300 A 1969). The curvature of the lipid bilayer in such vesicles derives from an internal aqueous spheroidal volume defined by this dimension. In contrast, the cubic phase aqueous channel diameter is about one-sixth of this and thus, is much more highly curved (Fig. 2). Nonetheless, the conformation of gramicidin in these two contrasting bilayer environments is similar as judged by their respective CD signatures. This suggests that conformation is not sensitive to the degree of bilayer curvature in the range studied. However, it should be noted that while vesicles have little curvature by comparison, the cubic phase expresses simultaneously curvatures that are both positive and negative and of equal magnitude in each leaflet of the bilayer (Chung and Caffrey, 1994).

31

Fig. 8. Sensitivity of the cubic-Pn3m phase of hydrated 9.9 MAG to gramicidin D at 20 °C. Phase identity and phase microstructure (lattice parameter) were determined using small-angle X-ray diffraction as described under Section 2.2. Gramicidin was incorporated into the lipidic mesophase by the co-dissolution mixing method. The concentration of gramicidin is expressed as mole% corresponding to 100(mole gramicidin)/[(mole gramicidin) + (mole 9.9 MAG)]. At the highest concentration of gramicidin used the initial phase observed was of the cubic-Pn3m phase. This transformed to the HII phase over a period of time at room temperature and at an elapsed time of 2 days is completely HII.

Fig. 9. Comparison of the absorption and corrected fluorescence emission spectra of gramicidin D in the cubic-Pn3m phase of hydrated 9.9 MAG (solid lines) and of L-tryptophan in buffer (dashed lines). The concentration of tryptophan used (640 lM) is four times that of gramicidin. All spectra have been corrected for background from glass, buffer, and lipid. All other conditions are as described in the legend to Fig. 3 and under Section 2.2.

3.3. Bilayer thickness There is considerable interest in bilayer thickness effects as possible modulators of membrane composition and function (Cantor, 1999). Some of the earliest studies of this effect in model systems were performed with

32

W. Liu, M. Caffrey / Journal of Structural Biology 150 (2005) 23–40

Fig. 10. Fluorescence quenching of gramicidin D in the cubic-Pn3m phase of hydrated 9.9 MAG. Bromo-MAG is the quenching lipid and its concentration is expressed as mole% corresponding to 100 (mole bromo-MAG)/[(mole bromo-MAG) + (mole 9.9 MAG)]. Gramicidin was combined into the lipidic mesophase by the direct mixing method. Fluorescence data have been corrected for background fluorescence from SPP buffer and lipid and for the inner filter effect, and have been normalized to the quencher-free (Fc, 0) value. Data are shown as the average of at least three replicate sample preparation and fluorescence measurements along with error bars. The insert shows a comparison of the quenching data for all three lipids collected under identical conditions.

gramicidin where ion movement across the membrane was monitored as a function of bilayer thickness (Veatch et al., 1975). These were extended to include major transporting proteins, such as the calcium-ATPase from sarcoplasmic reticulum (Caffrey and Feigenson, 1981). Subsequent studies with gramicidin have shown that its conformation in bilayered vesicles is indeed sensitive to membrane thickness (Greathouse et al., 1994; Wallace et al., 1981). To shed light on the in meso reconstitution hypothesis, we investigated if the conformation of gramicidin would likewise be sensitive to the thickness of the bilayer that constitutes the cubic phase. To achieve this, CD spectra of the peptide reconstituted into cubic phases formed by MAGs of different chain lengths were recorded. For this purpose, the 9.7, 9.9, and 11.9 MAGs were used. Separate X-ray diffraction measurements demonstrated that all form the cubic-Pn3m phase under conditions of measurement at room temperature (see below). The CD data show that the conformation of the peptide is not profoundly sensitive to the identity of the MAG chain used in the range studied (9.7 MAG, Fig. 6D; 9.9 MAG, Fig. 6E; and 11.9 MAG, Fig. 6F). In all cases, spectra are consistent with a helical dimer conformation. However, the ratio of the negative to the positive dichroism peaks changed in a systematic way in going from the long to the short-chained

MAG. This finding is consistent with an enhanced helical dimer character in the shorter lipid. This is expected given the better match between the bilayer thickness of the short-chained 9.7 MAG and the cylindrical length of the channel form of the peptide, as discussed below. To investigate if the fluorescence quenching profile, which should shed light on the lipidic environment of the gramicidin, was different in MAGs with different chain lengths, quenching data were collected with 9.7, 9.9, and 11.9 MAG at increasing mole% of bromoMAG (Fig. 10). In all cases, quenching occurs in a way that is entirely consistent with a bilayer location for the peptide. The degree of quenching at the highest mole% bromo-MAG used is the same for all three MAGs, as expected (data not shown). However, the details of the quenching dependence on bromo-MAG concentration are different for the different lipids. The difference is pronounced at the lower quenching lipid concentrations where the individuality of the MAGs is most apparent (Fig. 10). Thus, quenching is particularly effective in this region for the longest chain lipid and least effective for the shortest chain MAG. 9.9 MAG, with its intermediate chain length, shows a quenching behavior that is bracketed by that of the other two. This disparate quenching behavior reflects the way in which the gramicidin molecule interacts with the cubic phase bilayers created by the different lipids. One additional point worthy of note in regard to the quenching data presented in Fig. 10 is the simple fact that they are different for the different MAGs used. This indicates that gramicidin is sensitive to lipid chain characteristics which again supports the hypothesis that the peptide resides in the bilayer of the cubic phase. 3.4. Detergent reconstitution The overall goal of this study was to investigate the reconstitution of gramicidin into the cubic phase with the peptide serving as a model for membrane proteins undergoing in meso crystallization. The results presented thus far were obtained with gramicidin reconstituted directly without the use of detergents or solvents to effect dissolution. However, in the case of membrane proteins, such direct reconstitution is simply not practical. Most membrane proteins are introduced to the in meso system as mixed micelles dispersed in detergent. For the current gramicidin work to be relevant to the membrane protein field, it is important that we examine the disposition of the peptide in the cubic phase beginning with it dispersed in a detergent micellar solution. Gramicidin can be solubilized in SDS micelles when accompanied by a small amount of TFE. In fact, a recent NMR structure of the peptide was determined in such a solution (Townsley et al., 2001). The effects of SDS and TFE on cubic phase behavior of the hosting

W. Liu, M. Caffrey / Journal of Structural Biology 150 (2005) 23–40

lipid were investigated by X-ray diffraction studies. The results show that at the usual concentration of detergent used, SDS triggers a transition from the cubic-Pn3m to the cubic-Ia3d phase (Table 1). Even higher concentrations induce lamellar liquid crystal (La) phase formation. TFE has an effect similar to that of SDS (Table 2) in that increasing concentrations lead successively to the cubic-Ia3d and the La phases. However, at the levels used typically (1%(v/v) in SPP buffer, see Section 2.2), TFE does not affect phase behavior. While not ideal given the phase space group change, CD and fluorescence quenching measurements were per-

Table 1 Effect of SDS and of sample lipid/water content on the phase properties of hydrated monoolein at 20 °C Sample composition

SDS concentrationa (M)

Phase identity

Lattice parameter ˚) (A

60

0 0.02 0.03 0.05 0.10 0.20 0.25 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00

Pn3m Pn3m Ia3d Ia3d Ia3d Ia3d Ia3d Ia3d Ia3d La La La La La La

97 97 136 146 161 167 167 169 167 54 56 51 53 54 57

50

50

0 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00

Pn3m Ia3d Ia3d Ia3d La La La La La La La

103 136 192 186 75 64 64 61 57 57 56

60

40

0 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00

Pn3m Im3m Ia3d Ia3d La La La La La La La

102 188 238 216 72 69 67 65 63 63 58

Aqueous phase (%(w/w))

Monoolein (%(w/w))

40

Samples were prepared with 40, 50, and 60%(w/w) lipid and 60, 50, and 40%(w/w) water or detergent solution, respectively. a The detergent concentration in the aqueous SPP buffer used to prepare the lipidic dispersion.

33

Table 2 Effect of TFE on the phase properties of hydrated monoolein at 20 °C Phase identity

Lattice ˚) parameter (A

0 1 2

Pn3m Pn3m Pn3m

101 97 99

5

Ia3d

158

10 15

Ia3d + La Ia3d + La

156 + 47 183 + 48

20

La

48

TFE concentration in aqueous phasea (%(v/v))

All samples were prepared with 40%(w/w) SPP buffer and 60%(w/w) lipid. a The TFE concentration in the aqueous SPP buffer used to prepare the lipidic dispersion.

formed on gramicidin that had been reconstituted into the cubic-Ia3d phase using the standard, more concentrated detergent solution to begin with. The data are shown in Figs. 6H and 11. The fluorescence quenching curve has a profile similar to that observed using the direct reconstitution method (Fig. 11). The measurement was repeated at a lower concentration of SDS wherein the gramicidin is still micellarized initially but the cubic phase remains in the Pn3m space group. The data in Fig. 11 show that the general quenching profile and the extent of quenching in pure bromo-MAG are insensitive to SDS concentration in the range used and to space group type of the cubic phase. These results allow us to conclude that the quenching behavior, and all that

Fig. 11. Fluorescence quenching of gramicidin D in the cubic phase of hydrated 9.9 MAG as affected by the detergent SDS. The concentration of SDS used was 0.02 and 0.25 M which result in the cubic-Pn3m and cubic-Ia3d phase types, respectively. The data at 0 M SDS are taken from Fig. 10. All other conditions are as described in the legend to Fig. 10 and under Section 2.2.

34

W. Liu, M. Caffrey / Journal of Structural Biology 150 (2005) 23–40

it reflects in terms of lipid–peptide interaction, are not influenced dramatically by the initial state of dispersion of the peptide. Therefore, the bilayer location of gramicidin is insensitive to detergent micelles as a vehicle for solubilizing the peptide. The CD (and fluorescence, data not shown) spectra recorded with gramicidin reconstituted into 9.9 MAG at low (Fig. 6G) and high SDS concentrations (Fig. 6H) are virtually identical. This indicates that the conformation of the peptide is insensitive to a cubic phase environment that is of the Pn3m or the Ia3d type. In both cases, the helical dimer conformation for the peptide is confirmed. This is exactly what was found with gramicidin reconstituted directly into the cubic phase without the involvement of detergent (Fig. 6E). However, the CD spectra of gramicidin reconstituted directly or by means of detergent are qualitatively different (Figs. 6E and G). The most significant disparity is in the vicinity of the band at 230 nm. Nonetheless, both have a strong positive peak at 220 nm that is characteristic of the channel conformation. Thus, the overall conformation in the cubic phase membrane is insensitive to the form of the gramicidin used to effect reconstitution. 3.5. Mixed micelles Although the data presented thus far are consistent with a bilayer location for gramicidin in the cubic phase, it is possible that the peptide carried into the cubic phase dispersed in a surfactant solution remains micellarized. The view is that the gramicidin/detergent micelle might reside in the aqueous channels of the cubic phase and, accordingly, that it is not reconstituted into the lipid bilayer (Fig. 2). To effect the observed quenching, it would require that the bromo-MAG partition into the mixed micelle and quench the fluorescence of tryptophan in the gramicidin molecule. To explore the feasibility of this most unlikely possibility the following issues were addressed. First, the capacity of the gramicidin/SDS mixed micelle to incorporate MAG was examined. To this end, increasing amounts of 9.9 MAG were added to the gramicidin/SDS micellar dispersion and incorporation was judged by maintenance of optical clarity (data not shown). Saturation was achieved when 33 mol% (expressed as 9.9 MAG/(9.9 MAG + SDS)) of the SDS was replaced by 9.9 MAG. Beyond that limit, the solution turned cloudy indicating that the system no longer consisted of small micelles. Thus, the carrying capacity of the SDS micelle was put at 33 mol% 9.9 MAG. Separate CD measurements performed on the gramicidin/SDS/9.9 MAG system demonstrated that the conformation of the peptide was not affected by the presence of 33 mol% lipid additive in the micelle (data not shown).

The next task was to replace the 9.9 MAG in the gramicidin/SDS micelle with bromo-MAG and to monitor the fluorescence properties of the peptide. The data in Fig. 12 show convincingly that quenching occurred. Indeed, the degree of quenching is more pronounced than that seen in the cubic phase of 9.9 MAG at low bromo-MAG concentration. These results indicate that the tryptophans in gramicidin supported in a mixed micelle composed of SDS and 33 mol% MAG are indeed accessible to and quenched by the bromines on the chain of the lipid. The significance of this result will be discussed. 3.6. Co-dissolution in organic solvent A standard procedure for preparing mixtures of hydrated lipid and apolar additives involves initially dissolving both in an organic solvent. This ensures homogenous mixing. The solvent is subsequently removed by flushing with an inert gas following by an extended period under high vacuum. In the interests of completeness, we set about using this method also to prepare gramicidin/9.9 MAG samples for use in spectroscopic studies, as above. The expectation was that the results obtained using samples prepared from organic solvent would be identical to those of the direct method. However, we encountered a problem that prevented us from making the desired fluorescence quenching measurements. What we found was that upon subjecting

Fig. 12. Fluorescence quenching of gramicidin D in MAG containing SDS micelles. The quencher is bromo-MAG and it replaces mole for mole 9.9 MAG in the micelle. Thus, sample composition remains fixed at 0.25 M SDS and 33% (m/m) MAG as 9.9 MAG and/or bromoMAG. Excitation and emission wavelengths are 305 and 329 nm, respectively. For comparison part of the quenching curve of gramicidin in the cubic-Ia3d phase of 9.9 MAG at 0.25 M SDS from Fig. 11 is included. All other conditions are as described in the legends to Figs. 10 and 11 and under Section 2.2.

W. Liu, M. Caffrey / Journal of Structural Biology 150 (2005) 23–40

the 9.9 MAG to high vacuum drying, the UV absorption, and fluorescence in the vicinity of 340 nm, increased dramatically. No change in chemical composition of the lipid was detected by TLC and the phase behavior of the lipid was not affected as judged by X-ray diffraction. The problem was not lessened by using fresh vacuum pump oil or additional traps in the vacuum line, or by performing the drying in the dark. The same effect was observed with 9.7 MAG but not with 11.9 MAG. The source of the enhanced absorption and fluorescence remains a mystery but we bring it to the attention of the Reader to alert them to the problem. The absorption rise was not sufficiently severe, however, to prevent us making useful CD measurements on samples prepared by co-dissolution in organic solution. The results presented in Fig. 6I show that the spectrum is that of the helical dimer, and that it is very similar to the spectrum observed for gramicidin incorporated into the cubic phase by the direct method (Fig. 6E). Our conclusion therefore is that the conformation of the peptide reconstituted into the cubic phase bilayer is the same regardless of whether or not organic solvent is used to effect homogeneity as a first step in sample preparation.

4. Discussion 4.1. Evidence for reconstitution The primary purpose of this study was to test the hypothesis that gramicidin, acting as a model membrane protein, becomes reconstituted into the bilayer of the cubic phase under conditions that prevail during in meso crystallization. Gramicidin was chosen for this purpose because it is a small apolar polypeptide that is easily handled, well characterized, and robust. Usefully, it contains several tryptophan residues which have proved to be invaluable spectroscopic intrinsic tags with which to explore the chemical environment experienced by the peptide. What makes gramicidin particularly useful with regard to testing the hypothesis is that it can be incorporated into the cubic phase by at least two quite disparate methods. The first is the direct method, where the peptide is dissolved in 9.9 MAG that is subsequently used in cubic phase preparation. In this case, it is presumed that reconstitution occurs as a result of spontaneous formation of the cubic phase. The cubic phase itself consists of two interpenetrating aqueous channels that never contact one another being separated, as they are by a highly curved, continuous lipidic septum that is a bilayer. As such, gramicidin has two possible destinations within the cubic phase into which it can partition. One is the bilayer, the other the aqueous channels. Given that gramicidin is hydrophobic and essentially water insoluble, it is energetically favorable for it to sequester in the lipid bilayer compartment.

35

The results reported herein present several pieces of evidence in support of this. First, the fluorescence properties of the gramicidin molecule directly reconstituted suggest an apolar environment. Thus, the yield and wavelength of maximum intensity of the fluorescence from the tryptophans in gramicidin are increased and blue shifted, respectively, compared to tryptophan in aqueous solution (Fig. 9). Second, the fluorescence quenching study demonstrates convincingly that the gramicidin is in an apolar environment within the cubic phase. Quenching was carried out using a MAG bearing two bromines on carbons 9 and 10 of an 18-carbon fatty acid tethered to a glycerol backbone. Thus, the bromines are expected to be located to the interior of the lipid bilayer that permeates the cubic phase. Separate diffraction studies with similarly brominated PCs in the lamellar phase support this view of where the bromines are positioned within the membrane (Wiener and White, 1991). Fluorescence quenching was observed to increase approximately exponentially as the fraction of bromoMAG in the system increased (Fig. 10). This is consistent with a bilayer location for the peptide. The third piece of evidence derives from the fact that the quenching behavior proved sensitive to the identity of the non-quenching MAG acyl chains (Fig. 10). Since the hydrocarbon chains are confined to the bilayer interior, it is some property or properties of the bilayer itself that changes with the different MAGs. This is sensed by gramicidin presumably only when it is associated with that same lipid bilayer. In so doing, it responds differently to the quenching effect of the brominated lipid which has a distinct character imprinted on it by the different MAGs. One of the properties that changes with MAG identity is bilayer thickness. This, in turn, defines the relative positions of the apolar/polar interface across the membrane. A final piece of evidence for a bilayer location hinges on the logic that gramicidin is so apolar it is unfavorable for it to reside anywhere else within the confines of the mesophase. What we found was that the cubic phase can accommodate gramicidin up to a point. But beyond that limit, it triggers a transformation from the cubic to the HII phase (Fig. 8). This presumably reflects a change in the energetics associated with mismatch between the peptide and the lipid/water interface, which comes about as a result of a gramicidin that is bilayer bound. The second method used to incorporate gramicidin into the cubic phase involved first dissolving the peptide in a detergent micellar solution. The rationale behind choosing this alternative method was that it mimics the conditions used when proteins are detergent-solubilized from native membranes in preparation for use in in meso crystallization trials. In contrast, the direct method of incorporation is unlikely to work with such complex macromolecular systems. The data obtained for detergent-solubilized gramicidin are essentially the same as those collected using the direct method. The same can

36

W. Liu, M. Caffrey / Journal of Structural Biology 150 (2005) 23–40

be said for mesophase identity and phase microstructure, as determined by X-ray diffraction. Together these data indicate that gramicidin has a bilayer disposition regardless of its route or mechanism of incorporation into the cubic phase. They further support the view that the procedures used to combine gramicidin with the lipidic mesophase leads to a reconstitution of the peptide into a bilayer environment. This implies that the same reconstitution phenomenon happens with membrane proteins as a consequence of preparing them for crystallization by the in meso method. The data therefore support one of the central hypotheses for how membrane protein crystallization takes place by the in meso methods. The obvious next test is to demonstrate it directly with a bona fide integral membrane protein. The current work with gramicidin is a step in that direction. The data presented above in support of a bilayer disposition for gramicidin incorporated by the direct method are convincing. However, if the only form that the peptide can be introduced to the lipidic system takes the form of a mixed micelle along with detergent, the argument could still be made that bilayer reconstitution did not occur. Rather, that the peptide ends up residing in a detergent micelle that is accommodated in the aqueous channel of the cubic phase (Fig. 2). Given that the ˚ , such an arrangement channel diameter is of order 50 A is not unreasonable geometrically. However, the conjunction is untenable thermodynamically. For such an arrangement to exist would require detergent molecules to segregate away from the lipidic bilayer of the mesophase, and to selectively envelop and solubilize the peptide. A preferential partitioning of the gramicidin into the micelle and away from the lipid bilayer would also be required. By the same token, the lipid must confine itself to the membrane of the cubic phase and avoid becoming part of the micelle. None of these are likely to happen. The relative amounts of the components in the system are as follows: gramicidin/SDS/TFE/lipid/ water, 1:50:250:10 000:100 000 by mole in the sample prepared with 0.02 M SDS. Thus, the lipid (bilayer) dominates and will act as a potent sink into which detergent and gramicidin partition. This same reasoning extends to membrane proteins and the likelihood of them failing to reconstitute during preparation for in meso crystallization is very low. 4.2. Bilayer thickness effects The fluorescence quenching measurements made with the 9.7, 9.9, and 11.9 MAGs were done at or close to the full hydration limit for the cubic-Pn3m phase at room temperature. It is at this boundary in the corresponding lipid/water phase diagrams that mensuration can be done, in conjunction with the small-angle scattering data, to decipher the microstructure of the cubic phase (Qiu and Caffrey, 1998). Thus, the thicknesses of the bilayers per-

meating the cubic phase in the case of the 9.7, 9.9, and ˚ , respectively. The 11.9 MAGs are 33.2, 35.6, and 37.4 A corresponding water channel diameters are 56.6, 46.4, ˚ , respectively. and 45.5 A The three-dimensional structure of the channel form of gramicidin has been determined by NMR. In one study, oriented bilayers of hydrated dimyristoyl-PC acted to host the peptide (PDB code: 1MAG; Ketchem et al., 1993). In the other, SDS micelles were used to effect solubilization (Townsley et al., 2001). The two structures are very similar (Townsley et al., 2001). The ˚ long dimer measured end-to-end is approximately 26 A ˚ and has an outer diameter of about 16 A (Fig. 1). The paired b6.3 helices are stabilized by extensive intra- and intermolecular hydrogen bonding through backbone donors and acceptors. The tryptophans at the ethanolamine-capped C-terminus are suitably positioned at the lipid/water interfaces of the membrane. The CD measurements reported herein are consistent with a helical dimer conformation for the peptide in the cubic phase formed by all three MAGs (Figs. 6D–F). The fluorescence quenching profiles obtained using bromo-MAG in combination with the three different MAGs are distinct. The difference between them is pronounced in the low bromo-MAG concentration range where the character of the individual MAGs comes through (Fig. 10). This makes sense since it is here that the concentration of the non-quenching lipid is highest. Thus, we find that the degree of quenching in this region increases with growing chain length. Since bilayer thickness increases with chain length it is reasonable to assume that the differential quenching behavior is related to the span of the membrane and to the peptideÕs response to it. There are several possible interpretations of these quenching data. One that we favor places the C-terminal tryptophans of the dimer in closest contact with the bromines, where it experiences greatest quenching, in the thicker 11.9 MAG membrane. As the bilayer thins progressively in going to 9.9 MAG and to 9.7 MAG, the tryptophan–bromine separation increases which in turn gives rise to an attenuated quenching. The underlying assumptions here are that the glycerol headgroup of the bromo-MAG is anchored at the lipid/water interface and that the position of the bromines in the membrane changes with the chain length of the hosting lipid. Further, the gross as well as the detailed conformation of the gramicidin is expected to remain relatively insensitive to MAG chain identity. The above explanation is, of course, speculative. A more complete understanding of what is actually happening as chain character is adjusted is needed. This will benefit from a determination of the electron density distribution across the lipid bilayer, and thus the position of the bromines in that profile, as well as a detailed structure of gramicidin in the cubic phase bilayer. Both are under investigation in the authorÕs laboratory.

W. Liu, M. Caffrey / Journal of Structural Biology 150 (2005) 23–40

The model that emerges from this study suggests that a water-soluble quencher might affect gramicidin fluorescence in a bilayer thickness or lipid chain length-dependent manner. Thus, the shorter the chain the more accessible should be the tryptophans to quenching for a fixed gramicidin conformation. The hypothesis was tested with the water-soluble quencher, acrylamide, on gramicidin reconstituted into the cubic phase formed by 9.7, 9.9, and 11.9 MAG. Small-angle X-ray diffraction was used to determine the amount of acrylamide that could be added to effect maximal quenching without altering phase behavior. The limit was observed at 0.15 M acrylamide in the aqueous medium beyond which the cubic-Pn3m phase transformed to the Ia3d phase (Table 3). The fluorescence quenching curves are presented in Fig. 13. For all three lipids, quenching to the extent of 20–30% was observed at the highest acrylamide concentration used. This shows that the tryptophans in gramicidin reconstituted in the bilayer of the cubic phase are indeed accessible to quenching by a small, water-soluble molecule. The error in the data suggests that the difference in quenching behavior between the three lipids is not significant. However, the average values (based on three separate measurements) do show the expected trend with a more pronounced quenching in the shorter-chained lipid. 4.3. Gramicidin carrying capacity of cubic phase is finite Part of the evidence in support of the conclusion that gramicidin adopts a bilayer location in the cubic phase derives from its effect on the phase behavior of the hydrated 9.9 MAG system. At low levels, the cubic phase can accommodate the peptide with little or no effect on phase identity or phase microstructure (Fig. 8). However, at higher concentrations of 10 mol% gramicidin it destabilizes the cubic-Pn3m phase and induces HII phase formation. The following is offered as an explanation for how and why the transformation happens. As noted, the Table 3 Effect of acrylamide on the phase properties of hydrated monoolein at 20 °C Acrylamide concentrationa (M)

Phase identity

Lattice ˚) parameter (A

0 0.18 0.35

Pn3m Pn3m Pn3m

101 102 105

0.70 1.41

Ia3d Ia3d

160 159

2.11 2.82

Ia3d + La Ia3d + La

171 + 48 218 + 50

All samples were prepared with 40%(w/w) SPP buffer and 60%(w/w) lipid. a The acrylamide concentration in the aqueous SPP buffer used to prepare the lipidic dispersion.

37

Fig. 13. Accessibility of gramicidin D reconstituted into the cubicPn3m phase of three MAGs (indicated) to the water-soluble quencher, acrylamide, as judged by fluorescence quenching. All other conditions are as described in the legend to Fig. 10 and under Section 2.2. Lines are drawn to guide the eye. Error bars are based on triplicate sample preparation and fluorescence quenching measurements.

˚ long. The gramicidin helical dimer is a cylinder 26 A thickness of the cubic phase membrane formed by 9.9 ˚ , and it is in this that the gramicidin reMAG is 36 A sides at low peptide concentration. It is apparent therefore that an energetically expensive mismatch exists between the hydrophobic surface of the dimer and the apolar interior of the bilayer. The hydrocarbon chains of the HII phase are more disordered than those of the lamellar and cubic phases (Lewis et al., 1989). Thus, the shortest dimension from one hydrocarbon/water interface to another measured through the apolar region of the membrane will be less in the HII phase compared to that of the cubic phase. As gramicidin concentration rises, the energetic cost of mismatch becomes sufficiently large as to favor a global transition to a phase where mismatch is less severe. This then accounts for the peptide-induced phase transformation. The above rationalization leads to the following hypothesis: because mismatch is greater in 11.9 MAG compared to 9.9 MAG, a lower concentration of gramicidin should be needed to trigger the cubic-to-HII phase transition in the case of the former compared to latter MAG. In turn, 9.9 MAG should require less peptide than 9.7 MAG to the same end. Separate small-angle X-ray diffraction measurements support the hypothesis. Thus, the cubicPn3m-to-HII transition occurred at a gramicidin/lipid mole ratio of 1:30 (3 mol%) in the case of 11.9 MAG (Fig. 14F). The corresponding ratio for 9.9 MAG was 1:10 (10 mol%; Fig. 14D). No such transition was observed in the case of the short-chained 9.7 MAG at a mole ratio of 1:10 (10 mol%; Fig. 14B). The higher

38

W. Liu, M. Caffrey / Journal of Structural Biology 150 (2005) 23–40

Fig. 14. Phase behavior (A, C, and E) and phase microstructure characteristics (B, D, and F) of hydrated 9.7 MAG (A,B), 9.9 MAG (C,D), and 11.9 MAG (E,F) as affected by gramicidin D. All data are based on small-angle X-ray diffraction measurements. Time-resolved diffraction measurements in the 20–100 °C range were used for data in panels (A), (C), and (E) while static measurements at 20 °C were used for data in panels (B), (D), and (F). Phases identified include the cubic-Pn3m, HII and the fluid isotropic (FI). Data shown in panel (D) are from Fig. 8. Samples were prepared with 40, 40, and 48%(w/w) SPP buffer for the 11.9, 9.9, and 9.7 MAGs, respectively. Phase boundaries in (A), (C), and (E) are drawn to guide the eye.

concentration of peptide presumably needed to effect the phase change in 9.7 MAG was simply not accessible experimentally. These findings are entirely consistent with the model presented where the HII phase is stabilized to a greater extent when the mismatch between the peptide and its hosting lipid bilayer is exaggerated. We have confirmed this in separate time-resolved Xray diffraction measurements where the cubic-to-HII phase transition temperature was found to be 68 and 26 °C, respectively, for the 9.9, and 11.9 MAG systems at a 1:50 mole ratio of gramicidin and lipid (2 mol%; Figs. 14C and E). As expected, no such transition was observed with the short-chained 9.7 MAG (Fig. 14A).

5. Conclusions The purpose of this study was to test a central tenet of the hypothesis for how membrane proteins crystallize by the in meso method. The hypothesis posits that the protein undergoes an initial reconstitution into the highly curved lipid bilayer upon which the cubic phase is based. Testing the proposal with integral membrane proteins available to us currently is fraught with difficulties. Thus, gramicidin, a short polypeptide that forms helical transbilayer pores in membranes, was used in their stead. Spectroscopic (UV–visible absorption, fluorescence, fluorescence quenching, and circular dichro-

W. Liu, M. Caffrey / Journal of Structural Biology 150 (2005) 23–40

ism) and diffraction measurements demonstrated convincingly that gramicidin reconstitutes into the apolar compartment of the cubic phase bilayer. Spectroscopic signatures and accessibility to quenching moieties known to be confined to the bilayer interior or to be water-soluble were consistent with a peptide molecule oriented in what is referred to as the helical dimer or pore-forming conformation. Additional evidence for the bilayer location came from the differential sensitivity of gramicidinÕs fluorescence quenching behavior to bilayer thickness as dictated by the chain length of the reconstituting lipids. It was also supported by gramicidinÕs demonstrated ability to stabilize the HII phase and to trigger a cubic-to-HII phase transition in a bilayer thickness-dependent manner. The working hypothesis for in meso crystallization invokes a lamellar intermediate acting as a conduit between the bulk cubic phase and the face of the growing crystal. Thus, attempts to crystallize gramicidin in meso will eventually lead to a phase separation of sorts where the peptide concentration rises locally. However, we have shown in the current study that a high enough concentration of gramicidin in the cubic phase of certain MAG systems eventually leads to HII phase formation (Fig. 14). Thus, the rising local concentration of gramicidin as crystallization is approached may well work against crystal formation and growth by destabilizing the planar lamellar portal in favor of the more highly curved HII phase. We know that certain MAGs do not normally form the HII phase. 9.7 MAG is an example (Fig. 14). Further, the propensity for HII phase formation can be countered by the inclusion of additives that induce curvature of an opposite sense to that of the HII phase. Many of the detergents, such as SDS, or small molecules, such as TFE, will do just that as evidenced in the current study (Tables 1 and 2). On-going in meso crystallization trials involving gramicidin are benefiting from these insights and are invoking a number of strategies some of which include the use of the shorter chain MAGs and detergents with a view to facilitating crystal growth.

6. Data deposition Relevant data reported in this paper have been deposited in the Lipid Data Bank (http://www.lipidat. chemistry.ohio-state.edu/GramicidinData.htm).

Acknowledgments We sincerely thank V. Cherezov, J. Clogston, C. Law, Y. Misquitta, and K. Riedl for their many and varied contributions to this work and Dr. W. Shaw, Avanti Polar Lipids, Inc., for the gift of bromo-MAG. Supported in part by the National Institutes of Health (GM61070),

39

the National Science Foundation (DIR9016683 and DBI9981990), and Science Foundation Ireland (02-IN1B266).

References Blanton, T.N., Huang, T.C., Toraya, H., Hubbard, C.R., Robie, S.B., Louer, D., Gobel, H.E., Will, G., Gilles, R., Raftery, T., 1995. JCPDS—International Centre for Diffraction Data round robin study of silver behenate. A possible low-angle X-ray diffraction calibration standard. Powder Diffr. 10, 91–95. Briggs, J., 1994. The phase behavior of hydrated monoacylglycerols and the design of an X-ray compatible scanning calorimeter. Ph.D. dissertation. The Ohio State University. Caffrey, M., Feigenson, G.W., 1981. Fluorescence quenching of model membranes. 3. Relationship between calcium adenosinetriphosphatase enzyme activity and the affinity of the protein for phosphatidylcholines with different acyl chain characteristics. Biochemistry 20, 1949–1961. Caffrey, M., 2000. A lipidÕs eye view of membrane protein crystallization in mesophases. Curr. Opin. Struct. Biol. 10, 486–497. Caffrey, M., 2003. Membrane protein crystallization. J. Struct. Biol. 142, 108–132. Cantor, R.S., 1999. Lipid composition and the lateral pressure profile in bilayers. Biophys. J. 76, 2625–2639. Chen, Y., Wallace, B.A., 1997. Solvent effect on the conformation and far UV CD spectra of gramicidin. Biopolymers 42, 771–781. Cheng, A., Hummel, B., Qiu, H., Caffrey, M., 1998. A simple mechanical mixer for small viscous samples. Chem. Phys. Lipids 95, 11–21. Cherezov, V., Qiu, H., Pector, V., Vandenbranden, M., Ruysschaert, J.M., Caffrey, M., 2002a. Biophysical and transfection studies of the diC14-amidine/DNA complex. Biophys. J. 82, 3105–3117. Cherezov, V., Riedl, K.M., Caffrey, M., 2002b. Too hot to handle? Synchrotron X-ray damage of lipid membranes and mesophases. J. Synchrotron Radiat. 9, 333–341. Chiu, M.L., Nollert, P., Loewen, M.C., Belrhali, H., Pebay-Peyroula, E., Rosenbusch, J.P., Landau, E.M., 2000. Crystallization in cubo: General applicability to membrane proteins. Acta Crystallogr. Sect. D—Biol. Crystallogr. 56, 781–784. Chung, H., Caffrey, M., 1994. The neutral area surface of the cubic mesophase: Location and properties. Biophys. J. 66, 377–381. Gordeliy, V.I., Labahn, J., Moukhametzianov, R., Efremov, R., Granzin, J., Schlesinger, R., Buldt, G., Savopol, T., Scheidig, A.J., Klare, J.P., Engelhard, M., 2002. Molecular basis of transmembrane signalling by sensory rhodopsin II–transducer complex. Nature 419, 484–487. Greathouse, D.V., Hinton, J.F., Kim, K.S., Koeppe II, R.E., 1994. Gramicidin A/short-chain phospholipid dispersions: Chain length dependence of gramicidin conformation and lipid organization. Biochemistry 33, 4291–4299. Huang, C., 1969. Phosphatidylcholine vesicles. Formation and physical characteristics. Biochemistry 8, 344–352. Hunte, C., Michel, H., 2003. Membrane protein crystallization. In: Hunte, C., von Jagow, G., Scha¨gger, H. (Eds.), Membrane Protein Purification and Crystallization 2/e: A Practical Guide. Elsevier Science, USA, pp. 143–160. Katona, G., Andreasson, U., Landau, E.M., Andreasson, L.E., Neutze, R., 2003. Lipidic cubic phase crystal structure of the photosynthetic reaction centre from Rhodobacter sphaeroides at 2.35 angstrom resolution. J. Mol. Biol. 331, 681–692. Ketchem, R.R., Hu, W., Cross, T.A., 1993. High-resolution conformation of gramicidin A in a lipid bilayer by solid-state NMR. Science 261, 1457–1460. Killian, A.J., Prasad, K.U., Hains, D., Urry, D.W., 1988. The membrane as an environment of minimal interconversion. A

40

W. Liu, M. Caffrey / Journal of Structural Biology 150 (2005) 23–40

circular dichroism study on the solvent dependence of the conformational behavior of gramicidin in diacylphosphatidylcholine model membranes. Biochemistry 27, 4848–4855. Kolbe, M., Besir, H., Essen, L.O., Oesterhelt, D., 2000. Structure of the light driven chloride pump halorhodopsin at 1.8 angstrom resolution. Science 288, 1390–1396. Lakowicz, J.R., 1983. Instrumentation of fluorescence spectroscopy. In: Lakowicz, J.R. (Ed.), Principles of Fluorescence Spectroscopy. Plenum Press, New York, NY, pp. 19–49. Landau, E.M., Rosenbusch, J.P., 1996. Lipidic cubic phases: A novel concept for the crystallization of membrane proteins. Proc. Natl. Acad. Sci. USA 93, 14532–14535. Lewis, R.N., Mannock, D.A., McElhaney, R.N., Turner, D.C., Gruner, S.M., 1989. Effect of fatty acyl chain length and structure on the lamellar gel to liquid-crystalline and lamellar to reversed hexagonal phase transitions of aqueous phosphatidylethanolamine dispersions. Biochemistry 28, 541–548. Luecke, H., Richter, H.T., Lanyi, J.K., 1998. Proton transfer pathways in bacteriorhodopsin at 2.3 angstrom resolution. Science 280, 1934–1937. Luecke, H., Schobert, B., Lanyi, J.K., Spudich, E.N., Spudich, J.L., 2001. Crystal structure of sensory rhodopsin II at 2.4 angstroms: Insights into color tuning and transducer interaction. Science 293, 1499–1503. McPherson, A., 1999. Strategies and special approaches in growing crystals. In: McPherson, A. (Ed.), Crystallization of Biological Macromolecules. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY, pp. 271–329. Misquitta, Y., Caffrey, M., 2003. Detergents destabilize the cubic phase of monoolein: Implications for membrane protein crystallization. Biophys. J. 85, 3084–3096. Misquitta, L., Misquitta, Y., Cherezov, V., Slattery, O., Mohan, J.M., Hart, D.J., Zhalnina, M., Cramer, W.A., Caffrey, M., 2004a. Membrane protein crystallization in lipidic mesophases with tailored bilayers. Structure 12, 2113–2124. Misquitta, Y., Cherezov, V., Havas, F., Patterson, S., Mohan, J.M., Wells, A.J., Hart, D.J., Caffrey, M., 2004b. Rational design of lipid for membrane protein crystallization. J. Struct. Biol. 148, 169–175.

OÕKeeffe, A.H., East, M.J., Lee, A.G., 2000. Selectivity in lipid binding to the bacterial outer membrane protein ompF. Biophys. J. 79, 2066–2074. Paas, Y., Cartaud, J., Recouvreur, M., Grailhe, R., Dufresne, V., Pebay-Peyroula, E., Landau, E.M., Changeux, J., 2003. Electron microscopic evidence for nucleation and growth of 3D acetylcholine receptor microcrystals in structured lipid–detergent matrices. Proc. Natl. Acad. Sci. USA 100, 11309–11314. Qiu, H., 1998. The mesophase behavior of the monoacylglycerol/water system: Application in drug delivery. Ph.D. dissertation. The Ohio State University. Qiu, H., Caffrey, M., 1998. Lyotropic and thermotropic phase behavior of hydrated monoacylglycerols: Structure characterization of monovaccenin. J. Phys. Chem. B 102, 4819–4829. Qiu, H., Caffrey, M., 2000. Phase properties of the monoolein/water system: Metastability and equilibrium aspects. Biomaterials 21, 223–234. Townsley, L.E., Tucker, W.A., Sham, S., Hinton, J.F., 2001. Structures of gramicidins A, B, and C incorporated into sodium dodecyl sulfate micelles. Biochemistry 40, 11676–11686. Veatch, W.R., Mathies, R., Eisenberg, M., Stryer, L., 1975. Simultaneous fluorescence and conductance studies of planar bilayer membranes containing a highly active and fluorescent analog of gramicidin A. J. Mol. Biol. 99, 75–92. Wallace, B.A., 1998. Recent advances in the high resolution structures of bacterial channels: Gramicidin A. J. Struct. Biol. 121, 123–141. Wallace, B.A., Veatch, W.R., Blout, E.R., 1981. Conformation of gramicidin A in phospholipid vesicles: Circular dichroism studies of effects of ion binding, chemical modification, and lipid structure. Biochemistry 20, 5754–5760. Wiener, M.C., White, S.H., 1991. Transbilayer distribution of bromine in fluid bilayers containing a specifically brominated analog of dioleoylphosphatidylcholine. Biochemistry 30, 6997– 7008. Zhu, T., Caffrey, M., 1993. Thermodynamic, thermomechanical and structural properties of a hydrated asymmetric phosphatidylcholine. Biophys. J. 65, 939–954.