Application of Pulsed Field Gradient NMR with High Gradient Strength

(14), pp 7365–7370. DOI: 10.1021/la8002355. Publication Date (Web): June 14, 2008. Copyright © 2008 American Chemical Society. * Corresponding ...
0 downloads 0 Views 372KB Size
Download PDF
Langmuir 2008, 24, 7365-7370

7365

Application of Pulsed Field Gradient NMR with High Gradient Strength for Studies of Self-Diffusion in Lipid Membranes on the Nanoscale Konstantin Ulrich,† Monica Sanders,‡ Farida Grinberg,† Petrik Galvosas,† and Sergey Vasenkov*,‡ Fakulta¨t fu¨r Physik und Geowissenschaften, UniVersita¨t Leipzig, Leipzig, Germany 04103, and Department of Chemical Engineering, UniVersity of Florida, GainesVille, Florida 32611 ReceiVed January 23, 2008 This work demonstrates the feasibility of noninvasive studies of lipid self-diffusion in model lipid membranes on the nanoscale using proton pulsed field gradient (PFG) NMR spectroscopy with high (up to 35 T/m) gradient amplitudes. Application of high gradients affords for the use of sufficiently small diffusion times under the conditions when the width of the gradient pulses is much smaller than the diffusion time. As a result, PFG NMR studies of partially restricted or anomalous diffusion in lipid bilayers become possible over length scales as small as 100 nm. This length scale is important because it is comparable to the size of membrane domains, or lipid rafts, which are believed to exist in biomembranes. In this work, high-gradient PFG NMR has been applied to study lipid self-diffusion in three-component planar-supported multibilayers (1,2-dioleoyl-sn-glycerol-3-phosphocholine/sphingomyelin/cholesterol). The degree of lipid orientation in the bilayers was determined with 31P NMR. A special insert was designed to mechanically align the multibilayer stack at the magic angle with respect to the direction of the constant magnetic field to address the detrimental effects of proton dipole-dipole interactions on the NMR signal. This insert is an alternative to the conventional method of magic angle orientation of lipid membranes, the goniometer probe, which is not compatible with commercial high-gradient coils because of the lack of space in the magnet bore. Macroscopic orientation of the multibilayer stacks using the insert was confirmed with 1H NMR spectroscopic studies and the comparison of results obtained from identical experiments using a goniometer probe for orientation. Diffusion studies were carried out at three different constant magnetic field strengths (B0) over a range of temperatures and diffusion times. The measured diffusivities were found to be in agreement with the data obtained previously by techniques that are limited to much larger length scales of diffusion observation than high-gradient PFG NMR.

Introduction Lateral organization of biological membranes has been attracting an increasing amount of attention in the scientific community since the introduction of the “fluid mosaic model” in the early 1970s. This model describes the cell membrane as a neutral two-dimensional solvent that does not influence membrane protein function.1 Over the past decades, interest in membrane research was further stimulated by the proposal of the lipid raft hypothesis.2 This hypothesis postulates the existence of small functional domains in membranes of eukaryotic cells. Such domains known as lipid rafts are characterized by an enhanced concentration of cholesterol as well as other lipids and proteins. It has been suggested that lipid rafts serve as platforms that play a role in many cellular functions such as signal transduction.3 It is widely hypothesized that rafts are also significant for the survival of cancer cells and for the ability of viruses to penetrate cell membranes.4 Changes of the membrane composition leading to modifications of the raft properties have been explored as an attractive new strategy to fight cancer and viral diseases.5 Success of such strategies requires detailed knowledge of the structural and dynamic properties of rafts. A number of different experimental techniques have been employed to detect and characterize membrane domains in vivo * Corresponding author. E-mail: [email protected]. † Universita¨t Leipzig. ‡ University of Florida.

(1) Singer, S. J.; Nicolson, G. L. Science 1972, 175, 720–730. (2) Simons, K.; Ikonen, E. Nature 1997, 387, 569–572. (3) Simons, K.; Toomre, D. Nat. ReV. Mol. Cell. Biol. 2000, 1, 31–39. (4) Fantini, J.; et al. Expert ReV. Mol. Med. 2002, 20, 1–21. (5) Zaremberg, V.; McMaster, C. R. FASEB J. 2005, 19, A291-A291.

as well as in different types of model membranes, such as lipid bilayers and monolayers. Studies of membrane domains have been carried out for the most part using microscopy techniques, which are based on the detection of fluorescently labeled molecules. The most well-known techniques of this type include fluorescence resonance energy transfer (FRET), fluorescence recovery after photobleaching (FRAP), and fluorescence correlation spectroscopy (FCS).6–9 Although these techniques have proved to be very useful, their applicability for studies of localization and dynamics of relatively small biomolecules, such as lipids, is limited due to an observation that fluorescent labels can affect the physical properties of such molecules in membranes.10 Other methods such as single particle tracking and atomic force microscopy have also been used to study membrane inhomogeneity.11,12 Lastly, nuclear magnetic resonance (NMR) techniques, such as magic angle spinning (MAS) NMR and pulsed field gradient (PFG) NMR, have been used both to detect membrane domains and study dynamics of molecules in these domains.13–16 The main hindrance on the way to unrestricted experimental studies of membrane domains is related to the fact (6) Silvius, J.; Nabi, I. R. Mol. Membr. Biol. 2006, 23, 5–16. (7) de Almeida, R. F. M.; et al. J. Mol. Biol. 2005, 346, 1109–1120. (8) Kahya, N.; et al. J. Biol. Chem. 2003, 78, 28109–28115. (9) Kahya, N.; Schwille, P. Mol. Membr. Biol. 2006, 23, 29–39. (10) Shaw, J. E.; et al. Biophys. J. 2006, 90, 2170–2178. (11) Yuan, C.; et al. Biophys. J. 2002, 82, 2526–2535. (12) Dietrich, C.; et al. Biophys. J. 2001, 80, 1417–28. (13) Lindblom, G.; Oradd, G. J. Dispersion. Sci. Technol. 2007, 28, 55–61. (14) Polozov, I. V.; Gawrisch, K. Biophys. J. 2006, 90, 2051–2061. (15) Oradd, G.; Westerman, P. W.; Lindblom, G. Biophys. J. 2005, 89, 315– 320. (16) Scheidt, H. A.; Huster, D.; Gawrisch, K. Biophys. J. 2005, 89, 2504– 2512.

10.1021/la8002355 CCC: $40.75  2008 American Chemical Society Published on Web 06/14/2008

7366 Langmuir, Vol. 24, No. 14, 2008

that the majority of techniques used until now to study raft-like domains do not have sufficiently high spatial and/or temporal resolution. Recent evidence suggests that lipid rafts in cell membranes can exist as molecular clusters with sizes on the order of 100 nm or smaller in diameter.17–19 It is also possible that rafts are quite unstable.17,20 Clearly, experimental studies of membrane lipid organization and dynamics have to be carried out on length and time scales that are sufficiently small for raft detection to advance the current state of knowledge of membrane rafts. Due to the difficulty of examining and interpreting the intricacies of protein-lipid organization in live cells, multicomponent lipid bilayers have been used as model systems to study structural inhomogeneity of membranes. Lipid bilayers consisting of mixtures of unsaturated and saturated phospholipids, sphingolipids, and cholesterol are believed to be sufficiently complex to capture many important properties of cell membranes. The lipid-raft hypothesis suggests a close link21,22 between lipid rafts and liquid-ordered (lo) phase23 that can be easily observed in cholesterol- and sphingolipid-rich model membranes.12,24–26 Liquid-ordered domains in model membranes are usually quite large (>1 µm) in comparison to those in biomembranes. However, most recently also raft-like domains with the domain sizes in the range of 100 nm were observed in model membranes by a number of different techniques, including FRET,6,7,27 AFM,11,28–30 and solid-state NMR.14,31,32 In some cases such raft-like domains were found to be stable even under conditions when large lo domains disappear,27 i.e., at physiological temperatures and for lipid compositions that approximate outer monolayer of plasma membranes. Raft-like domains in model membranes have shown an ability to grow or agglomerate into larger domains upon temperature or composition change7,14,27,31 or due to crosslinking.33 This is in good agreement with the properties of biomembranes where lipid rafts are believed to be able to agglomerate upon cell stimulation into larger and more stable rafts.34–37 All of these data emphasize the importance of studies of model membranes. In this work we report an application of a new experimental option, i.e., PFG NMR with high (up to 35 T/m) magnetic field gradients, to study lateral diffusion in domain-forming lipid bilayers. Application of such gradients allows the use of sufficiently short diffusion times under the conditions of the narrow-pulse approximation,38,39 which requires that the gradient duration is small in comparison to the diffusion time. As a result, (17) McMullen, T. P. W.; Lewis, R. N. A. H.; McElhaney, R. N. Curr. Opin. Colloid Interface Sci. 2004, 8, 459–468. (18) Anderson, R. G. W.; Jacobson, K. Science 2002, 296, 1821–1825. (19) London, E. Biochim. Biophys. Acta 2005, 1746, 203–220. (20) Kusumi, A.; et al. Semin. Immunol. 2005, 17, 3–21. (21) Reitveld, A.; Simons, K. Biochim. Biophys. Acta 1998, 1376, 467–479. (22) Brown, D. A.; London, E. Annu. ReV. Cell DeV. Biol. 1998, 14, 111–136. (23) Sankaram, M. B.; Thompson, T. E. Biochemistry 1990, 29, 10670–10675. (24) Bagatolli, L. A.; Gratton, E. Biophys. J. 2000, 78, 290–305. (25) Korlach, J.; et al. Proc. Natl. Acad. Sci. U.S.A. 1999, 96, 8461–66. (26) Dufrene, Y. F.; et al. Langmuir 1997, 13, 4779–4784. (27) Silvius, J. R. Biophys. J. 2003, 85, 1034–1045. (28) Tokumasu, F.; et al. Biophys. J. 2003, 84, 2609–2618. (29) Giocondi, M.-C.; Le Grimellec, C. Biophys. J. 2004, 86, 2218–2230. (30) Johnston, L. J. Langmuir 2007, 23, 5886–5895. (31) Veatch, S. L.; et al. Biophys. J. 2004, 86, 2910–2922. (32) Holland, G. P.; McIntyre, S. K.; Alam, T. M. Biophys. J. 2006, 90, 4248– 4260. (33) Hammond, A. T.; et al. Proc. Natl. Acad. Sci. U.S.A. 2005, 102, 6320– 6325. (34) Janes, P. W.; Ley, S. C.; Magee, A. I. J. Cell Biol. 1999, 147, 447–61. (35) Dykstra, M.; et al. Annu. ReV. Immunol. 2003, 21, 457–81. (36) Orr, G.; et al. Biophys. J. 2005, 89, 1362–1373. (37) Kusumi, A.; et al. Semin. Immunol. 2005, 17, 3–21. (38) Mitra, P.; Halperin, B. J. Magn. Reson. A. 1995, 113, 94–101. (39) Malmborg, C.; Topgaard, D.; Soderman, O. J. Magn. Reson. 2004, 169, 85–91.

Ulrich et al.

monitoring anomalous or partially restricted diffusion in lipid membranes on the length scale comparable with sizes of rafts in cell membranes (i.e., around 100 nm) and on the time scale around 10 ms and larger becomes possible. This new PFG NMR option was validated by measuring effective diffusivities of lipids in three-component lipid membranes, which were studied previously by various techniques.8,40–42 The measured diffusivities were found to be in a good agreement with the corresponding experimental data reported previously for displacements larger than the smallest displacements studied in this work.

Experimental Section Oriented Multibilayer Stacks. PFG NMR studies were performed with multibilayer stacks supported on thin glass plates. The bilayer stacks were prepared from ternary lipid mixtures of 1,2-dioleoylsn-glycero-3-phosphocholine (DOPC, >99%), chicken egg yolk sphingomyelin (SM, >95%), and cholesterol (CHOL, >99%) in a molar ratio of 1:1:1. DOPC was purchased from Avanti Polar Lipid, Inc. (Alabaster, AL), while SM and CHOL were supplied by SigmaAldrich (St. Louis, MO). Sample preparation methods for bilayers reconstituted on glass planar supports have been considered tedious43 and very sensitive to solvent choice and hydration techniques.13 Two protocols, which are adapted from the current literature,44,45 are used for the preparation of model lipid membranes in this work. In both techniques, a lipid film is deposited on thin glass plates, which is followed by hydration, causing self-orientation of the lipids into multibilayer stacks. The lipids are individually dissolved in solvents and combined using the appropriate volume ratios to create a final ternary lipid solution. This solution is deposited onto clean glass plates (5 × 5 mm2 or 6 × 7 mm2 with a plate thickness of ∼0.1 mm) obtained from Marienfeldt/Menzel (Lauda-Ko¨nigshofen/ Braunschweig, Germany), and the solvent is evaporated off overnight under high vacuum at ∼313 K. Of the two methods employed here, the differences lie mainly in the solvents used to dissolve the lipids and in the way of introducing water into the system, i.e., either directly as liquid water or from the gas phase. The two techniques will be discussed separately from this point on. It was observed in this work that the PFG NMR data obtained with samples prepared using these two techniques were not influenced by the preparation method within the experimental uncertainty. In the first method, a solution of methanol:1-propanol (1:4 v:v) is used as the solvent. The ternary lipid solution was deposited on the glass plates to a sample mass of 8-9 µg/mm2. After solvent evaporation, from 30 to 40 plates were stacked on top of each other and transferred to a hydration chamber filled with 2H2O saturated with K2SO4. The salt and 2H2O were obtained from Sigma-Aldrich (St. Louis, MO) and Spectra Stable Isotopes (Columbia, MD), respectively. In this method, the hydration takes place only from the gas phase (i.e., indirectly) under the condition of a relative humidity of around 97%. The salt solution is used to keep the humidity slightly below 100% to ensure that condensation of water on the glass plates with lipids does not occur. This prevents direct contact between the lipid films on glass plates and liquid 2H2O. The stack was left in this environment for 7 days at 313 K for equilibration. Once the stack is set up inside the NMR tube (see next section for details), pure 2H O is added to bring the sample to maximum (i.e., 100%) hydration 2 and to keep it at this hydration. The second method uses a solution of chloroform:methanol (2:1 v:v) as the solvent. The ternary lipid solution was prepared as discussed above using three individual lipid stock solutions. A sublimable solid, i.e. naphthalene obtained from Cambridge Isotope Laboratories (Andover, MA), was then introduced into the lipid mixture in a naphthalene:lipid molar ratio of 1:1. The resulting (40) Filippov, A.; Oradd, G.; Lindblom, G. Biophys. J. 2004, 86, 891–896. (41) Ora¨dd, G.; Lindblom, G. Spectrosc.: Int. J. 2005, 19, 191–198. (42) Shahedi, V.; Oradd, G.; Lindblom, G. Biophys. J. 2006, 91, 2501–2507. (43) Gaede, H. C.; et al. Langmuir 2004, 20, 7711–7719. (44) Oradd, G.; Lindblom, G. Magn. Reson. Chem. 2004, 42, 123–131. (45) Hallock, K. J.; et al. Biophys. J. 2002, 82, 2499–2503.

PFG NMR of Lipid Membranes on the Nanoscale

Langmuir, Vol. 24, No. 14, 2008 7367 and Spectroscopy (AMRIS) facility of University of Florida in Gainesville, FL. A Diff60 probe with GREAT60 amplifier was used to generate high gradients for diffusion measurements using the 750 MHz spectrometer. To measure diffusivity, the PFG NMR signal A(g) obtained using a standard stimulated echo sequence or 13interval sequence with bipolar gradients46 was measured as a function of the magnitude of applied field gradient (g). In the measurements with the 750 MHz spectrometer, the amplitudes of the CH3 and/or CH2 lines of frequency-domain spectra of lipids as well the area under these lines were used as a measured signal to obtain the PFG NMR attenuation curves. At the same time, in the measurements using 400 or 125 MHz spectrometers the attenuation of the amplitude of the signal of the time-domain spectra were recorded. Both approaches can be used successfully for diffusion measurements. The advantage of the former approach is related to the possibility of measuring separately diffusivities associated with different NMR lines, which can originate from different types of molecules in the studied samples. The diffusivities D were obtained from either a one-exponential or two-exponential fit of the PFG NMR attenuation curves using

Figure 1. Schematic of a PFG NMR magic angle insert. Orientation of the sample at the magic angle, R ) 54.7° respective to B0, reduces the unwanted effects of T2 NMR relaxation due to dipole-dipole interactions. This insert fits inside a standard 10 mm glass NMR tube.

solution was deposited onto the glass plates to a sample mass of 8.3 µg/mm2. After solvent evaporation, the plates were left unstacked. They were hydrated from the gas phase in the presence of liquid 2H O saturated with K SO for 1-2 days inside of the hydration 2 2 4 chamber. After the initial hydration step, pure liquid 2H2O was directly added to the lipid bilayers using a syringe to achieve a final molar ratio of ∼28 mol of water/mol of lipid. The direct hydration by pure 2H O is expected to bring the sample to a maximum hydration of 2 100%. The 35-40 plates were stacked and wrapped in Parafilm. The stack was then held at 277 K for an additional 1-2 days. NMR Measurements. Once the stacks were fully prepared and hydrated, they were transferred to an NMR tube with an external diameter of 10 mm [for studies using 750 MHz (17.6 T) PFG NMR spectrometer] or 7.5 mm [for studies using 400 MHz (9.4 T) and 125 MHz (2.9 T) PFG NMR spectrometers]. 1H PFG NMR measurements of lipids in lipid membranes are complicated by rather short transversal (T2) NMR relaxation times and the related large line width in a frequency domain. Short T2 times and the related line broadening are mainly caused by strong proton dipole-dipole interactions in lipids that are not completely averaged by molecular motion. In order to minimize the effects of dipole-dipole interactions on T2 times and the line width, the stack of glass plates with bilayers was oriented in the NMR spectrometers in such a way that the angle (R) between the dipole moment of lipids and the direction of the applied magnetic field B0 is very close to the so-called “magic” angle (54.7°). In this case, the scaling factor (3 cos2 R - 1) in the Hamiltonian term describing the dipole-dipole interaction approaches a zero value. To orient the bilayer stacks at the magic angle, a special insert was developed for use with commercially available high-gradient diffusion probes. Figure 1 shows a schematic presentation of the insert. The insert is made of plexiglass and fits into the NMR tubes used in this study. The stacks were placed in the insert and inside the tube. Before sealing the NMR tube with a lipid sample, a few drops of excess 2H2O are added on the tube walls above or below the sample to ensure that the relative humidity in the gas phase surrounding the bilayers is 100%. The measurements of self-diffusion in model lipid bilayers were carried out on three different spectrometers: two homemade PFG NMR spectrometers (FEGRIS NT and FEGRIS FT) operating at a 1H resonance frequency of 400 and 125 MHz, respectively, and a wide-bore Bruker PFG NMR spectrometer operating at a 1H resonance frequency of 750 MHz. The former two spectrometers are located at Universita¨t Leipzig in Leipzig, Germany, while the latter one is located at the Advanced Magnetic Resonance Imaging

Ψ≡

A(g) ) A(g)0)

n)1 or 2

∑

pi exp(-q2teffDim)

(1)

i)1

where q ) γδg for the stimulated echo sequence and q ) 2γδg for the 13-interval sequence. γ is the gyromagnetic ratio, δ denotes the duration of the pulsed field gradient, teff is the effective diffusion time,46 and pi is the fraction of molecules diffusing with a diffusivity Dim. Values of δ and teff were kept constant for measurement of each attenuation curve. Acquisition of each proton spectrum using the PFG NMR sequences required up to 240 scans. The resulting ratios of signal-to-noise or to the background signal at small gradient strengths were around 300 for the measurements at 750 MHz and in the range of 50 for the measurements at 400 and 125 MHz. Effective diffusion times and temperatures were varied in the range of 9.6-59.4 ms and 291-310 K, respectively. The diffusion probes used are suitable for PFG NMR measurements with gradient amplitudes up to 35 T/m, but a maximum of only 28 T/m was needed for most of our experiments. Typically, the signal at 16 gradient strengths was recorded to obtain each PFG NMR attenuation curve. The diffusivities determined by fitting the attenuation curves using equation (1) correspond to the diffusion along the direction that coincides with the direction of the gradients and that of B0 (Figure 1). This direction was different from that of lateral diffusion. The angle between these two directions is β ) 90° - 54.7° ) 35.3° (where 54.7° is the magic angle). Noting that any displacement along the direction of the measurements rm corresponds to the displacement r ) rm/cos β along the direction of lateral diffusion, the diffusivity along the lateral direction (D R 〈r2〉) was calculated by multiplying the measured diffusivities [Dm R 〈(rm)2〉] by a factor (1/cos β)2 ) 1.5.40 In this calculation, we have assumed that the displacements of lipids in the direction perpendicular to the bilayer surface are negligibly small in comparison to the corresponding displacements along the bilayer surface. Orientation of lipids in the bilayer samples and macroscopic orientation of the multibilayer stacks in the samples were probed using 31P and 1H NMR without applying pulsed field gradients. For these measurements, a solid-state NMR spectrometer operating at a 1H resonance frequency of 300 MHz was used. The stacks were placed directly (i.e., without the insert) into a NMR glass tube and a goniometer probe was used to adjust the angle R between the normal to the surface of the bilayer supports and the direction of B0. The measurements were performed using either the free induction decay or Hahn echo NMR sequences.

Results and Discussion Orientation of the lipids in multibilayer stacks prepared for diffusion measurements was characterized using 31P NMR. Figure (46) Cotts, R. M.; et al. J. Magn. Reson. 1989, 83, 252–266.

7368 Langmuir, Vol. 24, No. 14, 2008

Ulrich et al.

Figure 2. Proton-decoupled 31P NMR spectra of a multibilayer stack for different angles between the bilayer normal and the direction of B0. The spectra were obtained using the Hahn echo sequence at T ) 298 K. The 31 P chemical shift was referenced externally to 85% (w/v) H3PO4.

2 shows examples of proton-decoupled 31P NMR frequencydomain spectra, which were recorded for different values of R, the angle of orientation of a multibilayer stack with respect to the direction of the constant magnetic field B0 (Figure 1). The angle was changed using the goniometer probe. Each of the measured spectra consists of a single line with a half-width around 6 ppm. The spectra in this figure collectively demonstrate that the majority of lipids in the sample are arranged in well-oriented bilayers due to the following considerations: (i) the chemical shift of the lines depends on the value of R, as expected for macroscopically oriented lipid bilayers, and (ii) the line halfwidths, which are mostly determined by the chemical shift anisotropy, are sufficiently small, leading to the conclusion that there is insignificant distribution of orientations of individual lipids with respect to the bilayer normal. The signal from randomly oriented lipids in the sample is expected to be independent of R and have a half-width that is several times larger than that of the lines in Figure 2. Within the experimental uncertainty, no such signal was observed. On the basis of the data in Figure 2, we estimate that more than 97% of lipids in our samples are located in well-oriented bilayers. In addition to the orientation of individual lipids, the macroscopic orientation of multibilayer stacks in the insert (Figure 1) with respect to the direction of B0 was investigated. This was accomplished by comparing the 1H NMR frequency-domain spectra of multibilayer which was oriented as close as possible to the magic angle using a goniometer probe (Figure 3A) with those recorded using the magic angle insert (Figure 3B). 1H NMR spectra in Figure 3 were obtained using the standard Hahn echo sequence with different separations (τ) between π/2 and π radio frequency pulses. Under the measurement conditions of this sequence, the signal amplitude is reduced due to T2 NMR relaxation by a factor exp(-2τ/T2). Hence, dependence of the signal amplitude on τ allows the evaluation of the characteristic values of T2, which are mostly determined by the degree of deviation from the magic angle orientation. The spectra in Figure 3 show the partially overlapping lines of the CH2 groups of the hydrocarbon chains and of the CH3 choline groups of the headgroup at around 1 and 3 ppm, respectively. These lines can be assigned to both DOPC and SM. A line at around 4.7 ppm is assigned to a small amount of normal water. It is seen that the spectra in Figure 3A and their dependence on τ essentially coincide with those in Figure 3B. The data in Figure 3A were recorded using a goniometer probe, which allows obtaining the magic angle orientation with the precision around 0.3°. On the basis of comparison of these data with those in Figure 3B, which were

Figure 3. 1H NMR spectra of multibilayer stacks measured by the Hahn echo sequence for different values of pulse separation, τ, at T ) 298 K. The measurements were performed with the stacks oriented at the magic angle. This orientation was achieved using a goniometer probe (A) or the specially designed insert shown in Figure 1B.

obtained using the specially designed magic angle insert, we can conclude that the insert allows the achievement of magic angle orientation with a similar precision. This result demonstrates that it is feasible to orient multibilayer stacks at the magic angle with high precision using a commercial PFG NMR probe and the insert instead of using a goniometer probe. It would be technically very difficult (if not impossible) to build a functional probe of the latter type capable of generating gradients in the range of 35 T/m because of the space limitation in the magnet bore. In addition to orienting samples at the magic angle15,40,44 the following two options have been previously used to increase T2 and to improve spectral resolution: (i) PFG NMR was combined with a magic angle spinning (MAS) technique,14,16,47,48 viz., fast sample rotation along the direction given by the magic angle, and (ii) PFG NMR has been combined with multiple-pulse homonuclear decoupling.49,50 The MAS approach has proven to be especially powerful method allowing discrimination between different membrane components. However, at this time it is not technically compatible with the application of gradients much higher than 1 T/m. Similarly, an approach based on homonuclear decoupling is not compatible with the application of high gradients because measurement artifacts are likely to occur if the gradients necessary for diffusion measurements are superimposed on the decoupling pulse sequence.49,50 Therefore, the approach based on orienting the multibilayer stacks at the magic angle without MAS or the use of decoupling sequences has been chosen for the present study. Figure 4 shows examples of 1H PFG NMR attenuation curves measured by the PFG NMR stimulated echo sequence for different effective diffusion times and at different temperatures. The (47) Pampel, A.; et al. J. Am. Chem. Soc. 2004, 126, 9534–9535. (48) Polozov, I. V.; Gawrisch, K. Biophys. J. 2004, 87, 1741–1751. (49) Dvinskikh, S. V.; Furo, I. J. Magn. Reson. 2001, 148, 73–77. (50) Dvinskikh, S. V.; Furo, I. J. Magn. Reson. 2000, 144, 142–149.

PFG NMR of Lipid Membranes on the Nanoscale

Langmuir, Vol. 24, No. 14, 2008 7369

Figure 5. Lateral diffusivity of lipids in multibilayer stacks plotted as a function of inverse temperature. Diffusivities are obtained from twoexponential fit (filled symbols) and monoexponential fit (empty symbols) of PFG NMR attenuation curves measured at a proton resonance frequency of 125, 400, or 750 MHz. Also shown for comparison are diffusion coefficients (*) recorded by fluorescence correlation spectroscopy (FCS) in lipid membranes with the same composition (from ref 8). The solid line shows the best fit of the diffusivities measured at T > Tm using the Arrhenius law {D ) D0 exp[-(Ea/RT)]} with Ea ) 91 kJ/mol.

Figure 4. Examples of diffusion attenuation curves obtained by the PFG NMR stimulated echo sequence under the following conditions: (A) the 1 H resonance frequency is equal to 750 MHz and δ ) 1.89 ms and (B) the 1H resonance frequency is equal to 400 MHz and δ ) 1.35 ms. The attenuation curves in A and B were obtained by measuring the intensity of the lipid lines in the frequency domain spectra and by measuring the amplitude of the time-domain spectra, respectively (see Experimental Section for more details).

effective diffusion time and sample temperature were kept constant during the measurement of each curve. The data in parts A and B of Figure 4 were obtained using the NMR spectrometers operating at a proton resonance frequency of 750 and 400 MHz, respectively. For each attenuation curve, the initial, i.e. smallest gradient strength, was chosen in such a way that it was sufficiently small to ensure that there is no attenuation of the lipid signal and, at the same time, sufficiently high for suppression of any significant contribution of H2O to the measured signal. Such water suppression is possible because water diffusivity in the lateral direction is around 2 orders of magnitude larger than that of lipids. As a result, even an application of relatively small gradients results in vanishing of the water signal under our experimental conditions. Suppression of the water signal is especially important for the measurements using the amplitude of the time-domain signal (Figure 4B) because contributions of water and lipids to this signal cannot be readily distinguished on the basis of chemical shift difference. It is seen that the data in Figure 4A,B, which were obtained at different B0 strengths and for different types of data processing, i.e., processing of either frequency-domain or time-domain spectra, show qualitatively and quantitatively similar behavior. In addition, no significant changes in the attenuation curves were observed when the 13interval sequence with bipolar gradients was used instead of the stimulated echo sequence to carry out diffusion measurements. In the measurements with the former sequence, all essential parameters such as the total duration of the gradient pulses, effective diffusion time, and temperature were kept to be the same or similar as in the measurements with the latter sequence.

In the presentation of Figure 4A,B, the attenuation curves obtained for different diffusion times are expected to coincide if the diffusivity(ies) of the probed lipids remain the same with increasing or decreasing diffusion time and the corresponding values of the mean square displacement. The data in this figure indicate that there is essentially no dependence of the attenuation curves and of the corresponding diffusion behavior on diffusion time. Such results were obtained for all measured temperatures. Fitting attenuation curves by eq 1 has shown that for temperatures lower than 298 K, the two-exponential fit (1 with n ) 2) produces better results than the one-exponential fit (1 with n ) 1). At the same time, for temperatures higher than 298 K, it has been sufficient to use single-exponential curves to fit the experimental data satisfactorily (Figure 4). In the latter case, the two-exponential fit either produced two diffusivities, which were identical within the uncertainty of the fitting procedure, or it gave an extremely small value (several percent) for one of the fractions, which was comparable to the fitting error. Single-exponential behavior corresponds to a single diffusivity for all lipids contributing to the measured signal. For the studied samples this signal mostly resulted from DOPC and SM. The biexponential behavior indicates the existence of two ensembles of lipids (i.e., DOPC and SM) with different diffusivity values. The figure shows results for different temperatures [i.e., above and below the transition temperature for the liquid-disordered/ liquid-ordered phase coexistence (Tm)] and different values of effective diffusion time, teff. The solid lines show the best fit results using eq 1 with n ) 1 (high temperatures) or 2 (low temperatures). The dashed lines show the monoexponential curves corresponding to the initial decay of the attenuation curves at low temperatures. The range of the signal attenuation in Figure 4B was restricted by a low signal-to-noise ratio of the measured signal. Figure 5 shows an Arrhenius-type plot of the temperature dependence of diffusivities, which result from fitting the attenuation curves for different values of the constant magnetic field B0. It is seen that there is essentially no influence of the B0 strength on the diffusion data. This observation is important in view of the possibility that our PFG NMR diffusion data could be distorted due to differences between the magnetic susceptibility

7370 Langmuir, Vol. 24, No. 14, 2008

of the glass plates, bilayer stacks, insert, and the gas phase in the PFG NMR samples. With the application of a strong external magnetic field B0, the susceptibility differences in the PFG NMR samples can yield additional magnetic field gradients, which are expected to be stronger for larger B0 values. The existence of large susceptibility-induced gradients would inevitably interfere with our PFG NMR measurements. However, the results in Figure 5 show no evidence for such susceptibility effects because the diffusion data remain essentially the same for different B0 values used in this work. Hence, we can conclude that under our experimental conditions the disturbing influence of the susceptibility effects on the diffusion data is negligibly small. This conclusion is supported by the observation that both the 13interval sequence, which is designed to reduce influence of the susceptibility effects on the PFG NMR data, and the standard stimulated echo sequence yield essentially the same results. The PFG NMR observation of a single ensemble of lipids characterized by a single diffusivity at sufficiently high temperatures (T > Tm, where Tm ) 298 K corresponds to the temperature shown by the vertical dashed line in Figure 5) and of two ensembles of lipids with different diffusivities at low temperatures (T < Tm) can be attributed to the microscopic homogeneity of the lipid bilayers at the former temperatures and to the formation of large liquid-ordered (lo) domains that are surrounded by the liquid-disordered (ld) phase at the latter temperatures. For bilayers with the same lipid composition as studied in this work, the formation of giant (viz., micrometersized) lo domains was previously visualized by fluorescence microscopy when the bilayer temperature was decreased beyond Tm.8,12 On the basis of these results, the smallest diffusivity of the two diffusion coefficients obtained by PFG NMR for each T < Tm can be assigned to the diffusion of lipids inside liquidordered domains. These domains are characterized by a tighter packing and a lower diffusivity of lipids in comparison with the liquid-disordered phase. Consequently, the larger diffusivity can be assigned to diffusion of lipids in the liquid-disordered phase. For any particular phase or domain type, the diffusivities of different lipid species, i.e., SM, CHOL, and DOPC, in multicomponent bilayers can be expected to be essentially the same or similar on the basis of the results of previous experimental studies.15,16 This is in agreement with the observation of the monoexponential attenuation curves at T > Tm in the present work (Figures 4 and 5). Scatter in the diffusivity values at temperatures T < Tm when two-exponential fit of the attenuation curves becomes necessary is mostly associated with signal-tonoise limitations of measurements at the two lower proton resonance frequencies. These limitations are especially pronounced at large signal attenuations corresponding to diffusion in the liquid-ordered phase. In comparison, the signal-to-noise ratio of the 1H PFG NMR measurements at 750 MHz was appreciably better. It is important to note that no systematic changes in the measured diffusivities were observed with increasing proton resonance frequency. The results in Figure 5 were obtained with several different bilayer samples prepared in two different countries. Hence, under these circumstances, the agreement between the data in the figure can be considered to be satisfactory. Under our measurement conditions, the char-

Ulrich et al.

acteristic sizes of liqid-ordered domains are expected to be in the range of 1 µm, i.e., much larger than the root-mean-square displacements of lipids [〈r2(t)〉 ) 4Dteff] for all effective diffusion times used. Hence, the possible influence of diffusion restrictions by the domain boundaries and of molecular exchange between different domains on the measured diffusivities can be ruled out. This conclusion is in agreement with our observation that the diffusivities measured at each temperature do not depend on diffusion time. The values of diffusivities obtained by PFG NMR with high gradient strength in this work are in good agreement with those obtained using FCS8 and PFG NMR with conventional gradient strengths40 in lipid membranes with the same or similar composition. This is demonstrated by the comparison of our PFG NMR data and the previously reported FCS results in Figure 5. It is important to note that in comparison with the previous studies, PFG NMR with high gradient strength allows us to record diffusivities under the conditions of the narrow pulse approximation on a length scale as small as 100 nm, i.e. a much smaller length scale than in the previous measurements. In future studies, this new technique will be used to investigate transport and the related structural properties of small, i.e., nanosized, membrane domains, which are believed to be of high biological relevance.18,51,52

Conclusions In summary, we have demonstrated that it is feasible to adapt the PFG NMR technique with high gradient strengths for studies of lateral diffusion in lipid membranes on the nanoscale. The experimental approach used in this work is novel because it allows for noninvasive studies of anomalous or partially restricted diffusion of lipids over much smaller length scales than previously reported. Such studies become possible because an application of high-amplitude gradients in PFG NMR measurements allows the use of small diffusion times. As a result, diffusion data can be obtained for the length scale of lipid displacements which is comparable to the suggested size of lipid rafts in cell membranes (on the order of 100 nm). Due to the advantage of observing diffusion over smaller length scales, study of model lipid systems in which nanosized domains are observed is now possible. In the case that domain size is comparable to the length scale of diffusion measurements, we should now be able to observe the influence of domain boundaries on lipid diffusion. Acknowledgment. Financial support by Deutsche Forschungsgemeinschaft in the framework of the International Research Training Group “Diffusion in Porous Materials” (K.U.) and the start-up fund of University of Florida (S.V.) are gratefully acknowledged. NMR data were obtained at the Advanced Magnetic Resonance Imaging and Spectroscopy (AMRIS) facility in the McKnight Brain Institute of the University of Florida. We thank Prof. J. Ka¨rger, Prof. G. Lindblom, Dr. K. Gawrisch, Dr. Carsten Selle, and Dr. D. Plant for numerous discussions. LA8002355 (51) Sengupta, P.; Baird, B.; Holowka, D. Semin. Cell DeV. Biol. 2007, 18, 583–590. (52) Lommerse, P. H. M.; Spaink, H. P.; Schmidt., T. Biochim. Biophys. Acta 2004, 1664, 119–131.