13014
J. Phys. Chem. B 2005, 109, 13014-13023
Electrostatic Contributions to Indole-Lipid Interactions Holly C. Gaede, Wai-Ming Yau, and Klaus Gawrisch* Laboratory of Membrane Biochemistry and Biophysics, National Institute on Alcohol Abuse and Alcoholism, National Institutes of Health, 5625 Fishers Lane, Room 3N07, Bethesda, Maryland 20892-9410 ReceiVed: March 3, 2005; In Final Form: May 4, 2005
The role of electrostatic forces in indole-lipid interactions was studied by 1H and 2H NMR in ether- and ester-linked phospholipid bilayers with incorporated indole. Indole-ring-current-induced 1H NMR chemical shifts of lipid resonances in bilayers of 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine, 1,2-dioleoyl-snglycero-3-phosphocholine, 1,2-di-O-octadecenyl-sn-glycero-3-phosphocholine, and 1,2-di-O-octadecenyl-snglycero-3-phosphomethanol show a bimodal indole distribution, with indole residing at the upper hydrocarbon chain/glycerol region of the lipid and near the choline group, when present. 2H NMR of indole-d7-incorporated lipid bilayers reveals that the former site is occupied by about two-thirds of the indole, which adopts a distinct preferred orientation with respect to the bilayer normal. The results suggest that the upper hydrocarbon chain/ glycerol location is dictated by many factors, including interactions with the electric charges and dipoles, van der Waals interactions, entropic contributions, and hydrogen bonding. Indole diffusion rates are higher in lipids with ester bonds and lower in choline-containing lipids, suggesting that interactions between indole and carbonyl groups are of minor importance for lipid-indole association and that cation-π interactions with choline drive the second indole location. Nuclear Overhauser effect spectroscopy cross-relaxation rates suggest a 30-ns lifetime for indole-lipid associations. These results may have important implications for sidedness and structural transitions in tryptophan-rich membrane proteins.
Introduction In membrane-spanning proteins, the aromatic amino acids tryptophan and tyrosine are found preferentially at the interfacial region of the lipid bilayer. In particular, tryptophan is concentrated on the periplasmic side of membrane proteins,1 and interfacial tryptophan has been implicated in protein activity,2-4 selectivity,5,6 folding, and assembly.7 Studies of small peptides and tryptophan analogues suggest that distinct interfacial interactions play a role in the enhanced stability of this protein region.8-15 Tryptophan has an indole side chain that is capable of a variety of interactions in the complex amphipathic environment provided by the lipid interfacial region. As an aromatic molecule, indole possesses an electric quadrupole moment, which may be viewed as two opposing dipoles emerging from the plane of the aromatic ring.16 This moment creates a region of partial negative charge above and below the plane of the aromatic ring, allowing the ring to bind cations. Cation-π interactions may occur between aromatic systems and monovalent cations as well as the quarternary ammonium and primary amino groups present in lipids. Indeed, it has been suggested that this cation-π interaction is a dominant factor in influencing aromatic amino acid behavior.17 Along similar lines, the aromatic ring may act as a hydrogenbond acceptor.18 In a lipid matrix, the primary candidates for a hydrogen-bond partner with the aromatic ring are the hydroxyl groups of the lipid-water interface and interfacial water. Additionally, the amino group of the indole is capable of forming hydrogen bonds with the lipid phosphate and ester oxygens in the interfacial region. However, an NMR study of indole analogues found that hydrocarbon penetration was not * Author to whom correspondence should be addressed. E-mail:
[email protected].
10.1021/jp0511000
increased by a decrease in hydrogen-bonding potential, strongly suggesting that hydrogen bonding of the amino group is not primarily responsible for an interfacial location.14 Furthermore, 2H NMR of ester and ether lipids with indole has suggested that hydrogen bonding with lipid carbonyls is not likely to play a major role in tryptophan’s interfacial preference.8 Entropic factors may also play a role in the location of tryptophan, as an entropic penalty must be paid to insert the rigid ring into the tumultuous hydrocarbon region. In contrast, locating the ring system near the relatively rigid glycerol region may result in an energetically favorable increase of van der Waals interactions, with little entropic cost. Electrostatic interactions are likely candidates for directing the interfacial preference of tryptophan. The charge distribution for an ester phosphatidylcholine lipid is as follows: a zone with weak positive charge located around the carbonyl, a region of negative charge around the phosphate group, and a positive region near the positively charged choline.19 The charge distribution of water shows the opposite trend of the lipid; that is, there is a negative charge density near choline, a positive charge near the phosphates, and very weak negative charge density near the carbonyls.19 The lipid interface is also a rich source of electrical dipoles, with the phosphatidylcholine (PC) headgroup possessing a dipole moment of 18.5-25 D, oriented nearly parallel to the membrane surface.20 Additional dipoles on the order of 2.0 D arise from the carbonyls. In PCs, the carbonyl on the sn1 chain is oriented preferentially in the plane of the bilayer, while the carbonyl group of the sn2 chain is pointed toward the water phase.21 Last but not least, water molecules contribute a dipole moment of 1.8 D each, with the hydrogen atoms of the first water layer pointed preferentially toward the membrane. This arrangement of dipoles results in a positive electrical potential in the membrane interior.22
This article not subject to U.S. Copyright. Published 2005 by the American Chemical Society Published on Web 06/09/2005
Indole-Lipid Interactions Ether-linked lipids represent an attractive model for probing the importance of electrostatic interactions. Removal of the ester carbonyls eliminates both potential hydrogen-bonding sites, a region of weak positive charge, and two electric dipoles. Despite the fact that ether lipids have a dipole potential across the interface that is about 100 mV smaller than that of ester lipids, their bilayer properties are similar.22 Additional removal of the choline headgroup in the case of a phosphomethanol ether lipid allows for investigation of the importance of choline’s positive charge and of the headgroup dipole. To assess the role of electric dipole interactions between tryptophan and lipid in dictating the location and orientation of tryptophan within the lipid bilayer, we undertook a series of solid-state NMR studies with indole incorporated into bilayers formed from both ester and ether lipids. The order parameters of 2H-C bonds of indole were measured by 2H NMR. The data were analyzed in terms of indole orientation and wobble in the bilayer-associated state. Indole location in the bilayer was determined by 1H magic angle spinning (MAS) ring-current-induced chemical shift measurements performed on indole incorporated into ester and ether lipid matrixes. Since the aromatic ring current generates field lines that oppose the external magnetic field in the center of the ring but reinforce the external magnetic field at the edges of the ring, either upfield or downfield shifts may be observed, depending on the location of the indole relative to the magnetic field. These measurements report the distance and relative orientation of the indole aromatic ring to lipid protons. Twodimensional 1H MAS nuclear Overhauser effect spectroscopy (NOESY) experiments of the same indole-lipid systems were also used to assess the indole location within the lipid matrixes. In addition to spatial proximity, NOESY cross-relaxation rates are sensitive to lipid and indole dynamics. Finally, pulsed field gradient (PFG) MAS experiments were used to determine lateral diffusion of both indole and lipid. Comparison of indole and lipid diffusion rates provided a measure of the strength of indole-lipid associations. When combined with a quantitative assessment of indole-lipid crossrelaxation rates, the lifetime of indole-lipid associations may be estimated. Materials and Methods Sample Preparation. 1-Palmitoyl-2-oleoyl-sn-glycero-3phosphocholine ((16:0-18:1)-PC), 1,2-dioleoyl-sn-glycero-3phosphocholine ((18:1-18:1)-PC), 1,2-di-O-octadecenyl-snglycero-3-phosphocholine ((18:1-18:1)-O-PC), and 1,2-di-Ooctadecenyl-sn-glycerol-3-phosphomethanol ((18:1-18:1)-OPM), and 1-palmitoyl-d31-2-oleoyl-sn-glycero-3-phosphocholine ((16:0-18:1)-PC-d31) were synthesized by Avanti Polar Lipids (Alabaster, AL). D2O (99.9% D) was purchased from Cambridge Isotope Labs (Andover, MA). Indole (purity 99% or higher) was purchased from Aldrich (Milwaukee, WI). Indole-d7 was purchased from C/D/N Isotopes (Pointe-Claire, Quebec, Canada). Specifically deuterated indoles were prepared by reaction with Raney nickel in a series of deuterated solvents.23 2H NMR Samples: Indole and Lipid. A stock solution of indole-d7 (1.7754 × 10-4 mol/mL) was prepared in CH3OD. An appropriate amount of the stock solution (70-100 µL) was added to 40 mg of dry lipid to prepare samples of approximately 5-10 lipid molecules per indole molecule, allowing for evaporative losses. The samples were then dried under a stream of nitrogen while spinning to generate a large surface area for efficient removal of the solvent. The samples were then hydrated with 40.0 µL of deuterium-depleted H2O and homogenized by
J. Phys. Chem. B, Vol. 109, No. 26, 2005 13015 vortexing and freeze-thawing. A small amount of each sample (∼5 mg) was transferred to 4-mm NMR rotors equipped with 11-µL spherical inserts. The remaining sample was transferred to 5-mm glass tubes with ground glass joints. Specifically Labeled Indole and Lipid Samples. The 16:0-18:1-PC (30.0 mg) and specifically deuterated indoles (0.5-1.0 mg) were dissolved in methanol to result in an indole/ 16:0-18:1-PC molar ratio of 0.1-0.2. The methanol was removed with a stream of argon, and then the lipid film was redissolved in cyclohexane. The cyclohexane was removed by freeze-drying in a vacuum. The duration of the freeze-drying procedure was adjusted to remove the solvent while maintaining losses of indole that resulted in the desired lipid/indole ratio. The powder was dispersed in 35.0 mL of deuterium-depleted water, and the resulting liposomes were pelleted by centrifugation. (16:0-18:1-PC)-d31-Containing Samples. Deuterium-depleted water (8 µL) was added to 16:0-18:1-PC-d31 (5 mg) and vortex-mixed for homogeneity. In the case of (18:118:1)-PC, (18:1-18:1)-O-PC, and (18:1-18:1)-O-PM lipids, approximately 2 mg of (16:0-18:1)-PC-d31 was added to 20 mg of the lipid in a small sample tube and dissolved in 50 µL of methanol. The mixture was vortexed to homogenize. The methanol was removed while the sample was spun under a flow of dry N2 gas. The residual solvent was removed under vacuum. The sample was hydrated with 20 µL of deuterium-depleted water and vortex-mixed to homogenize. All samples were transferred by centrifugation into 5-mm-diameter glass sample tubes with ground glass stoppers. 1H NMR Samples. A stock solution of indole (1.449 × 10-4 mol/mL) was prepared in methanol. An appropriate amount of the stock solution (20-50 µL) was added to 5 mg of dry lipid to prepare samples of approximately 2:1 lipid/indole molar ratios, allowing for evaporative losses. The samples were then dried under a stream of nitrogen while spinning to promote the homogeneous distribution of indole. The samples were then hydrated with 5 µL of D2O and homogenized by vortexing. The samples were transferred by centrifugation to 4-mm NMR rotors equipped with 11-µL spherical inserts. A 16:0-18:1-PC liposome sample containing indole was prepared by taking 7.0 mg of a hydrated lipid indole mixture and diluting it in 1.00 mL of D2O. The resulting dispersion was extruded 15 times through 0.1-µm-pore-diameter polycarbonate filters (Osmonics). NMR Measurements. 2H NMR. Solid-state 2H NMR spectra were recorded on a Bruker DMX 300 NMR spectrometer (Billerica, MA) equipped with a high-power probe with a 5-mm solenoidal sample coil. The experiments were performed at 46.1 MHz using a quadrupolar echo sequence24 with a 2.2-µs 90° excitation pulse, a 50-ms delay between pulses, and a relaxation delay of 200 ms. A spectral width of 200 kHz was used with 16 000 data points for each scan, and 50 Hz of line broadening was applied before Fourier transformation. Typically, between 250 000 and 1 million scans were accumulated for deuterated indole incorporated into lamellar phase lipid at 10 °C, and between 100 000 and 500 000 scans were accumulated for samples containing (16:0-18:1)-PC-d31. The powder spectra were dePaked using the algorithm of Sternin et al.25 The order parameters SCD for the individual C-2H bonds were directly calculated from the experimental quadrupolar splittings according to the following equation
∆νq ) 3/4(e2qQ/h)SCD
(1)
where SCD ) 1/2(3 cos2 θ - 1), where θ is the angle between the bilayer normal and the C-D vector. The static quadrupolar
13016 J. Phys. Chem. B, Vol. 109, No. 26, 2005 coupling constant (e2qQ/h) of 180 kHz for aromatic ring deuterons was used for indole-containing samples and 167 kHz for acyl chain deuterons in (16:0-18:1)-PC-d31 samples. Initial assignments of indole quadrupole splittings in (16:018:1)-PC were conducted using specifically deuterated indoles.23 In those experiments, it was established that quadrupole splittings, including the sign, may be assigned in a unique fashion even without relying on specific deuteration. The assigned quadrupolar splittings were used to examine the time-averaged orientation of indole in the membrane relative to the bilayer normal. It was assumed that the indole performs rapid rotation about one axis defined by two angles and that this axis wobbles. The latter motion translates into a reduction of all order parameters by the same factor (Figure 3). The following assumptions were made; field gradient tensors are axially symmetric, and the axis of symmetry of the field gradient tensor is along the C-2H bond axis. The energyminimized coordinates of the indole ring were obtained by GAMESS (general atomic and molecular electronic structure system) connected to CHARMM (chemistry at Harvard macromolecular mechanics, 25n2) in a combined quantum mechanical and molecular mechanical method. 1H NMR. Solution-state 1H MAS NMR spectra were obtained of a 16:0-18:1-PC/indole liposome sample using a 500.13 MHz Bruker (Billerica, MA) DMX500 wide-bore spectrometer operating XWINNMR (version 3.1). The experiments were conducted using a 5-mm-diameter TXI triple-gradient highresolution probe with 32 scans and a pulse delay of 5 s. The terminal methyl was calibrated to 0.885 ppm, and the indole and the lipid headgroup regions were carefully baselinecorrected and integrated. 1H MAS NMR. 1H MAS NMR experiments were performed at 500.13 MHz on a Bruker (Billerica, MA) DMX500 widebore spectrometer operating XWINNMR (version 3.1). The experiments were conducted using 4-mm double-gas-bearing, triple-resonance Bruker MAS probes, one of which was equipped with z-axis gradients, operating at a spinning frequency of 10 kHz and a temperature of 10 °C, unless otherwise specified. One-dimensional spectra were acquired with one scan at a spectral width of 5000 Hz. The chemical shift scale was calibrated by setting the resonance of the terminal methyl groups of the hydrocarbon chains to 0.885 ppm. Because of the indoleinduced ring current effects on chemical shifts, it was necessary to reassign lipid and indole resonances. Spectral assignments are based on 2D-MAS correlation spectroscopy (COSY) and HETCOR experiments. NOESY. Two-dimensional MAS NOESY experiments were acquired in a phase-sensitive mode with 256 t1 increments and 16 scans per t1 increment. To determine cross-relaxation rates quantitatively, integral intensities of diagonal peaks were recorded at a mixing time, tm, close to zero (1 ms) and of cross and diagonal peaks at a longer mixing time (300 ms). Peak intensity is related to the relaxation rate matrix Rij according to the matrix equation Aij(tm) ) exp(-Rijtm)Aij(0), where Aij(0) is the diagonal peak intensity at zero mixing time. After the intensity matrixes Aij(300ms) and Aij(0) were determined experimentally using the integration routine in XWINNMR (Bruker Biospin, Billerica, MA), the elements of the relaxation rate matrix Rij were easily calculated by matrix algebra using Mathcad.26 Diffusion. Diffusion measurements were conducted at sixteen different values of gradient strengths varying from 0.01 to 0.37 T/m with a stimulated echo sequence using sine-shaped bipolar gradient pulses of 5 ms in duration. A longitudinal eddy current
Gaede et al.
Figure 1. Structures of (16:0-18:1)-PC, (18:1-18:1)-PC, (18:1-18: 1)-O-PC, (18:1-18:1)-O-PM, and indole. The letters and numbers refer to resonance assignments of the NMR spectra given in Figures 2 and 4.
delay of 5 ms was used, and diffusion times were varied from 50 to 500 ms. At every gradient strength, 16 scans were acquired with a recycle delay of 4 s. Mathcad was used to fit the signal intensities of the indole C7/C4 resonance, the lipid choline resonance, hydrocarbon chain methylene resonance, and chain methyl resonance to the equation that relates signal intensity to the diffusion constant for powder samples
ln
()
I 2 2 ) - kD + (kD)2 I0 3 45
(2)
where D is the diffusion constant and k is a factor whose exact nature depends on the pulse sequence and on instrumental settings.27 For the stimulated echo sequence, k ) 4γ2g2δ2(∆ T/2 - δ/8), where γ is the gyromagnetic ratio of protons, g is the gradient strength, δ is the gradient pulse length, and T is the time between the gradient pulses sandwiching the 180° pulses. Results and Discussion The sample preparation procedures result in homogeneously distributed indole in multilamellar liposomes in the liquid crystalline phase, as monitored by 1H MAS spectra of lipids and indole that confirmed the existence of single populations of lipid and indole on the time scale of the NMR experiment as well as by solid-state 31P and 2H NMR on selected samples that showed one lamellar lipid phase/one set of indole quadrupole splittings only. The induced chemical shift changes (see below) of the lipid resonances observed in indole-containing 16:0-18:1-PC liposomes reveal that the indole-lipid/water log P is 2.26, which compares well to a measured indole-octanol/ water log P of 2.33.28 With this partitioning, the fraction of a percent of indole present in the water phase of the 50 wt % MLV samples may be safely neglected, as was done in the analysis. The indole and lipid structures are shown in Figure 1. 2H NMR Order Parameters. The solid-state 2H quadrupolar splittings of deuterated indole incorporated into lamellar lipids were directly measured from the dePaked 2H NMR spectra shown in Figure 2. The assignment of each particular quadrupolar splitting to an indole C-2H bond was straightforward and based upon the corresponding solution NMR signal assignments and the calculated degrees of deuteration at specific sites. The characteristics of the quadrupolar splittings for indole are summarized as follows; the C3-2H bond has a very small quadrupolar splitting value and is apparently close to the magic angle; the C4-2H and C7-2H signals are nearly superimposed
Indole-Lipid Interactions
J. Phys. Chem. B, Vol. 109, No. 26, 2005 13017
Figure 3. Plot of the inverse of the square of the difference between observed and calculated quadrupolar splittings for indole-d7 incorporated into the lamellar phase of (16:0-18:1)-PC. The inset shows the coordinate system. The z-axis is perpendicular to the indole ring system, bisecting the two rings. The x-axis bisects the plane of the benzene ring, and the y-axis is orthogonal to the x- and z-axes. θ is the angle between the z-axis and the bilayer normal, and R defines the angle between the x-axis and the projection of the bilayer normal in the x-y plane. Figure 2. (A) Powder and (B) dePaked 2H NMR spectra of indole-d7 incorporated into lamellar (16:0-18:1)-PC, (18:1-18:1)-PC, (18:118:1)-O-PC, and (18:1-18:1)-O-PM. The x-axis of the dePaked spectra has been calibrated according to spectra of bilayers oriented with their normal parallel to the magnetic field (0° spectra).
parallel to the bilayer normal. These maxima occur at R values of 128° and 308°. In the former orientation, the nitrogen is oriented toward the water phase, and in the other, the molecule is rotated by 180° and the nitrogen faces the hydrocarbon region. The four maxima at θ values of 45° and 135° correspond to a specific tilt of the indole ring to the bilayer normal. However, these maxima are very narrow. Even minor deviations from this specific orientation would result in major disagreements with experimentally observed quadrupolar splittings, suggesting that their occurrence is not likely. The calculated order parameters, Smol, are shown in Table 1. To compare Smol to the measured order parameters, the overall motion of the indole must be considered by inclusion of the factor Swobb. The values of Swobb are similar for the PC lipids with values of 0.25, 0.21, and 0.20 for (16:0-18:1)-PC, (18:1-18:1)-PC, and (18:1-18:1)-O-PC, respectively. (18:1-18:1)-O-PM has a considerably higher order with a value of 0.30 for Swobb. The values of R for indole-d7 are 128°, 124°, and 124°, in (18:1-18:1)-PC, (18:1-18:1)-O-PC, and (18:1-18:1)-O-PM, respectively. This constancy in R indicates that the orientation of indole in the lipid matrix is essentially unchanged by the
and have similar quadrupolar coupling tensors; the absolute quadrupolar splitting value of the C2-2H bond is smaller than that of the C5-2H bond, indicating a smaller angle with the bilayer normal for C5-2H than that for C2-2H. These assignments already reveal some information about the indole orientation in the lipid bilayer, and the fairly low 2H NMR order parameters of all C-2H bonds of the indole indicate low orientational constraint with respect to the rotational axis. The quadrupolar splittings and order parameters for indole in each lipid are shown in Table 1. The contour plot in Figure 3 shows the inverse of the square of the sum of the difference between the calculated and observed order parameters for θ angles from 0° to 180° and R angles from 0° to 360° for indole in (16:0-18:1)-PC. Six possible maxima are apparent. The two broad maxima at a θ value of 90° correspond to an indole orientation with its plane on average
TABLE 1: Comparison of Quadrupolar Splittings and Order Parameters of Indole Incorporated into Different Lipid Matrixesa (16:0-18:1)-PC C2-2H C3-2H C4-2H C5-2H C6-2H C7-2H a
(18:1-18:1)-PC
∆νq
SCD
Smolb
9.7 0.54 16.9 26.9 12.3 16.1
0.072 0.004 0.12 0.20 0.091 0.12
0.23 -0.087 0.46 0.76 -0.47 0.45
(18:1-18:1)-O-PC
(18:1-18:1)-O-PM
∆νq
SCD
Smol
∆νq
SCD
Smol
∆νq
SCD
Smol
7.0 0.77 13.7 22.1 12.8 13.6
0.052 0.006 0.10 0.16 0.095 0.10
0.227 -0.087 0.46 0.76 -0.47 0.45
4.9 1.0 15.8 18.4 10.8 14.9
0.036 0.007 0.12 0.14 0.08 0.11
0.072 0.059 0.61 0.63 -0.50 0.60
5.9
