Phase Diagram Determination for Phospholipid/Sterol Membranes Using Deuterium NMR

Transcription

1 Phase Diagram Determination for Phospholipid/Sterol Membranes Using Deuterium NMR YA-WEI HSUEH, 1 MARTIN ZUCKERMANN, 2 JENIFER THEWALT 2,3 1 Department of Physics, National Central University, Chung-li 320, Taiwan 2 Department of Physics, Simon Fraser University, Burnaby, British Columbia, Canada, V5A 1S6 3 Department of Molecular Biology and Biochemistry, Simon Fraser University, Burnaby, British Columbia, Canada, V5A 1S6 ABSTRACT: Deuterium NMR is used to investigate the phase behavior of membranes containing ergosterol and to measure the temperature ergosterol mole fraction phase diagram. In particular, our approach to analyzing the NMR spectra has allowed us to definitively locate and characterize the region where two liquid crystalline phases coexist. The general methodology for phase boundary determination in phospholipid/sterol membranes is presented Wiley Periodicals, Inc. Concepts Magn Reson Part A 26A: 35 46, 2005 KEY WORDS: phase diagram; phosphatidylcholine; ergosterol; 2 H NMR INTRODUCTION Received 3 January 2005; revised 21 February 2005; accepted 21 February 2005 Correspondence to: Jenifer Thewalt; address: jthewalt@ sfu.ca Concepts in Magnetic Resonance Part A, Vol. 26A(1) (2005) Published online in Wiley InterScience ( com). DOI /cmr.a Wiley Periodicals, Inc. Membranes Not Homogeneous Eukaryotic cell plasma membranes are composed of an impressive variety of lipids and proteins. The physical organization of the lipids has been a subject of interest ever since the idea of the lipid bilayer, where the lipids hydrophobic chains are sequestered from the aqueous surroundings in the center of a quasi two-dimensional sheet, was proposed (1). Refinement of this original lipid bilayer picture to include proteins (2) led to the fluid mosaic model of the cell membrane (3). The fluid mosaic model has lipids and proteins mixed at random hence, mosaic and undergoing rotational and translational diffusion within the liquid crystalline membrane hence, fluid. Recent observations indicate that biological membranes are far from random. The inner (cytoplasmic) and outer leaflets have distinct lipid and protein 35

2 36 HSUEH, ZUCKERMANN, AND THEWALT compositions. Furthermore, specialized regions of the plasma membrane known as rafts are thought to be important for many cell functions (4). The idea that these rafts exist in a different phase, the cholesterolrich liquid-ordered (lo) phase, compared with the rest of the membrane has been widely proposed in recent years (5). Model Membranes Why Is It Important to Understand Lipid/Lipid Interactions? Because cell membranes are composed of such a wide array of different molecules, to study the fundamental behavior of the major lipid species it is common to employ model membranes having defined lipid composition. Such simplified systems, consisting of at most a few different types of lipids, allow the basic membrane thermodynamics to be measured. If there are no more than three different lipid species, a phase diagram can be constructed. Such a phase diagram contains in principle a complete description of the system. Ternary lipid phase diagrams are often measured at a single temperature, as a function of composition, whereas binary lipid phase diagrams are typically measured as functions of temperature and composition. Ternary and binary lipid membranes containing a sterol (usually cholesterol) as one of the components have been studied extensively, and their phase behavior is thought to be relevant to the study of rafts in cell membranes (6). Phases Commonly Displayed by Phospholipid Membranes Containing Sterols The bilayer phases that are represented in lipid/sterol phase diagrams were described by Vist and Davis (7) for dipalmitoylphosphatidylcholine (DPPC)/cholesterol and the notation we use to characterize them is due to Ipsen et al. (8). The gel phase of phospholipid bilayers having a low cholesterol content is referred to as the solid-ordered (so) phase, the liquid crystalline phase of these lipid bilayers is referred to as the liquid-disordered (ld) phase and the beta phase of Vist and Davis (7), which is found at higher cholesterol concentrations, is referred to as the liquid-ordered (lo) phase. The terms solid and liquid are used to characterize the nature of the phase, whereas the words ordered and disordered indicate the conformational nature of the lipid acyl chains. Thus, the so phase is made up of lipids having very limited motional freedom and straight, essentially all-trans, hydrocarbon chains. The ld phase consists of lipids with kinking, conformationally explorative chains undergoing translational and rotational diffusion. The lo phase, then, contains straight-chained lipids that are free to diffuse. A major difference between ld and lo phase membranes is that the hydrophobic thickness of the lo phase is much larger than that of the ld phase. Windows on the Lipid World: What Can We Learn About Phases and Transitions? These phases are observed in many ways. The existence of and consequences of transforming between these phases has been studied by such techniques as differential scanning calorimetry (DSC), infrared (IR) spectroscopy, x-ray diffraction, atomic force microscopy (AFM), fluorescence microscopy, and single particle tracking. This is not the place for an exhaustive comparison of techniques but briefly, DSC detects heat absorption or emission and thus is sensitive to the existence of phase transitions in membranes. The frequency of molecular vibrations detected by such methods as IR spectroscopy varies with phase state. Small-angle x- ray diffraction monitors bilayer spacing and hydrophobic thickness, whereas wide-angle x-ray diffraction is useful for determining chain packing and the extent of lipid motion. AFM is a relatively recent technique that can sensitively determine membrane thickness. Fluorescence microscopy has been used extensively in lipid vesicles to look at micron scale phases distinguishable by fluorescent probe partitioning characteristics. Single particle tracking can yield information about the translational diffusion characteristics of the labeled membrane constituent. 2 H NMR: Advantages and Disadvantages Regarding Membrane Phase Determination The chief disadvantage that all spectroscopy must overcome is a psychological one: human beings are predisposed to believe what they see. Thus pictures exert great power compared with arcane spectral patterns. Our aim in this article is to demystify one extremely useful technique for exploring phase transitions in membranes, 2 H NMR. Note that the use of 2 H NMR to explore phase transitions relies on a robust understanding of 2 H NMR of membranes as provided by pioneering researchers in the 1970s and 1980s (9 11). Another disadvantage of 2 HNMRis that to do experiments in a reasonable length of time, tens of mg of deuterium-labeled lipid must be used per sample, which can be expensive. In contrast to techniques such as infrared spectroscopy, 2 HNMR monitors only the deuterated molecule. The advan-

3 MEMBRANE PHASE DIAGRAMS 37 tages, of course, are many. No bulky and potentially phase-boundary distorting probes are needed with this nondestructive technique. Spectroscopic information is exceedingly rich and unambiguous. Because every spin is observed, coexisting phases are represented in a quantitative way by their spectra. Membrane structure on small scales, from m bond disorder along the acyl chains to 10 7 m liquid crystalline coexisting domains, is observed. Recently, it has become possible to determine using 2 H NMR exclusively all the major phase diagram boundaries for model membranes containing sterols (12). The model membrane system we use to illustrate the technique of phase diagram determination by 2 H NMR in this article is DPPC-d31/ergosterol. DPPC is a well-studied phosphatidylcholine with two 16-carbon saturated acyl chains. When one chain is perdeuterated to form DPPC-d31, the hydrated lipid has a sharp so to ld phase transition temperature of 40 C. Ergosterol is similar to cholesterol in structure and is the sterol that is found in the plasma membranes of yeast. SAMPLE PREPARATION DPPC-d31/erg multilamellar dispersions were prepared for erg concentrations of 0, 5, 10, 13, 16, 20, 25, 27.5, 30, 35, and 42 mol%. DPPC-d31 was obtained from Avanti Polar Lipids Inc. (Alabaster, AL). Erg was obtained from Sigma-Aldrich Canada Ltd. (Oakville, Ontario, Canada). DPPC-d31 and erg were mixed in the appropriate quantities, dissolved in benzene/methanol, 4:1 (v/v), and then freeze-dried. Samples were hydrated using a ph 7.4 buffer prepared in deuterium-depleted water containing 50 mm HEPES, 150 mm NaCl, and 4mM EDTA. Hydration was performed by freeze-thaw-vortex cycling five times between liquid nitrogen temperature and 50 C. Deuterium-depleted water was from Sigma-Aldrich Canada Ltd. 2 H NMR SPECTROSCOPY 2 H NMR experiments were performed on a locally built spectrometer at 46.8 MHz using the quadrupolar echo technique (13). The typical spectrum resulted from 10,000 to 15,000 repetitions of the two-pulse sequence with 90 pulse lengths of 3.95 s, interpulse spacing of 40 s, and dwell time 2 s. The delay between acquisitions was 300 ms and data were collected in quadrature with Cyclops 8-cycle phase cycling. The sample was heated from 15 C to 60 C. At Figure 1 Partial phase diagram of the DPPC/erg membrane., midpoint of the transition from M 1 (T) curves (Fig. 5);, onset or end of transition in M 1 (T) curves;, obtained by inspection of the depaked spectra as a function of erg concentration (Fig. 12); Œ, obtained by inspection of the depaked spectra as a function of temperature (Fig. 14); E, obtained from M 1 (erg) curves (Fig. 13);, obtained from spectral subtraction;, the onset of transition in M 1 (T) curves for membranes having erg concentrations of 25%, 27.5%, or 30%. each temperature, the sample was allowed to equilibrate for 20 minutes prior to a measurement. The spectra were depaked using the procedure described by Lafleur et al. (14). The spin-spin relaxation time, T 2e, was measured by varying the interpulse spacing from 40 to 100 s and taking the initial slope of the echo peak signal versus echo time. PHASE DIAGRAM DETERMINATION The partial phase diagram for DPPC-d31/erg constructed from the 2 H NMR spectra is shown in Fig. 1. We have identified three two-phase regions and a three-phase line in the DPPC/erg partial phase diagram. There is a so ld phase coexistence region at about 40 C between 0 mol% and 9 mol% erg. There is a so lo phase coexistence region below 39.5 C between 9 mol% erg and mol% erg and a ld lo phase coexistence region above 39.5 C between mol% erg and mol% erg. The upper bound of the ld lo phase coexistence region lies in the neighborhood of 53 C. The boundary separating the so lo and ld lo regions is a so ld lo

4 38 HSUEH, ZUCKERMANN, AND THEWALT so/so ld/ld Region As each phase state displays a distinct spectrum, the phase state of a given membrane is easy to identify. Phase changes can be observed by examining the spectra as a function of temperature or composition. As shown in Fig. 3(A), the spectra of pure DPPC-d31 obtained below 39 C are characteristic of a pure so phase membrane. The spectra obtained above 41 C display pure ld phases. The spectrum obtained at 40 C is composed of both so and ld components. It is obvious that pure DPPC-d31 MLDs undergo a gel- (so)-to-liquid disordered (ld) phase transition near 40 C (T m ) as the temperature is raised. To trace the so component in a so ld spectrum, we usually examine the tail of the spectrum. The tails of pure so and ld spectra extend to 60 khz and 30 khz, where the spectra disappear into the baseline respectively (see Fig. 2). Therefore, the signal between 30 and 60 khz (or between 30 and 60 khz) would be an indication of the so component. The spectra obtained above 41 C in Fig. 3(A) show no signal but noise in khz, suggesting that there is no so component in these spectra. Figure 2 Characteristic spectra of the pure (A) so, (B) ld, and (C) lo phases. three-phase coexistence line at 39.5 C. We discuss in the following text the methods of determining these regions and phase boundaries using 2 HNMR. As described in the introduction, membrane lipids in aqueous environments display a variety of phases, such as the liquid crystalline ld and lo, and the gel so. The 2 H NMR spectrum of each phase state displays distinct features. These features are associated with the lipid molecular motion and molecular packing arrangements that characterize these phases. Figure 2(B) (and Fig. 2(C)) shows the typical liquid crystalline spectrum obtained from membranes consisting of randomly oriented lipid bilayers. The spectrum is a superposition of Pake doublets indicating that the membrane is in the liquid crystalline phase where the acyl chain undergoes rapid, axially symmetric reorientation about the bilayer normal. Because of the more conformationally disordered lipid chain, the average spectral width of the ld phase spectrum (Fig. 2(B)) is nearly half that of the lo phase (Fig. 2(C)). Figure 2(A) displays the spectrum of the so phase. Deuterons in so membranes do not undergo axially symmetric motion on the NMR time scale because the acyl chains are closely packed. First Moment M 1 Phase change can also be detected by examining the first moment as a function of temperature or composition. The first moment, M 1, calculated from the 2 H NMR spectrum is defined as M 1 1 x I A x Figure 3 (A) 2 H NMR spectra and (B) the first moment M 1 of pure DPPC-d31 as a function of temperature.

5 MEMBRANE PHASE DIAGRAMS 39 Figure 4 2 H NMR spectra of DPPC-d31/erg as a function of erg concentration at T 33 C. where is the frequency shift from the central (Larmor) frequency, I( ) is the spectral intensity, and A x x I( ). M 1 measures the average spectral width. Each phase state has a distinct 2 H NMR spectrum and thus a distinct M 1 value. A significant change of M 1 is usually observed as membranes go from one phase to another. In Fig. 3(B), M 1 of pure DPPC-d31 decreases significantly from 39 C to 41 C, where membranes undergo transitions from so phase to ld phase, consistent with the spectrum change in Fig. 3(A). The midpoint of the transition (40 C) is defined as T m. In 95:5 DPPC-d31/erg MLDs, the so to ld phase transition occurs at a lower temperature, where a spectrum composed of so and ld components is observed (12). Unlike pure DPPC-d31, temperature dependences of spectrum and M 1 display a broader two-phase (so ld) region. The temperatures where the onset and the end of transitions occur are the temperatures where the so/so ld and so ld/ld phase changes, respectively. The open squares in Fig. 1 denote these two temperatures. The spectra as a function of erg concentration for DPPC-d31/erg at 33 C are shown in Fig. 4, indicating that below T m DPPC-d31/erg undergoes a broad so to lo phase transition as erg concentration increases. The 0 mol% erg and 5 mol% erg samples display so phase spectra. For the 10 mol% erg sample, a small lo component appears in addition to the so spectrum, as indicated by the emergence of peaks near 15 khz and 20 khz. The coexistence of so and lo phases indicates that erg induces lo domains in the DPPCd31 bilayers. The proportion of lo component increases as the erg concentration increases (see spectra at 13 mol%, 16 mol%, and 20 mol%). Concurrently, the proportion of so component decreases. The 25 mol% and 27.5 mol% samples seem to display pure lo spectra. However, whether the so component has totally disappeared in these two spectra is not clear. Examining the tail of the spectrum as we did for the so ld spectrum would be difficult in this case, as the tails of so and lo spectra both extend to 60 khz (see Fig. 2). The spectra of 30 mol%, 35 mol%, and 42 mol% should be in the pure lo phase. Therefore, Fig. 4 implies a so/so lo phase change and a so lo/lo phase change occurring between 5 and 10 mol% and between 20 and 30 mol% respectively. Further analysis will be discussed to narrow down where the phase changes. Figure 5 shows the spectra as a function of erg concentration for DPPC-d31/erg at 45 C. It suggests that above T m DPPC-d31/erg undergoes a broad ld to lo phase transition with increasing erg concentration. The 0 mol% erg, 5 mol% erg, and 10 mol% erg samples display pure ld phase spectra. As erg concentration increases, the average spectral width increases and the individual peaks broaden. At 30 mol% erg, the individual peaks become sharper and remain sharp at higher erg concentration. For erg concentrations between 13 mol% and 25 mol%, the broadening of the individual doublets in the spectrum implies an ld lo phase coexistence. The determination of the exact ld/ld lo and ld lo/lo boundaries are discussed later. The spectra of 84:16 DPPC-d31/erg as a function of temperature are shown in Fig. 6. The spectra of 26 C, 30 C, 34 C, and 38 C show that there is so lo phase coexistence in the DPPC/erg membrane. At 26 C, the spectrum consists mostly of the so component with a very small amount of the lo component as indicated by the emergence of Pake doublet peaks within the central part of the spectrum ( 20 khz). These Pake doublet peaks become more prominent as temperature increases and the so component is reduced. At 38 C, the lo component dominates the spectrum. At 42 C, 46 C, and 50 C, ld lo phase coexistence is observed. This is found by examining the individual peaks of the depaked spectra. These peaks are broadened, indicating the coexistence of ld

6 40 HSUEH, ZUCKERMANN, AND THEWALT Figure 5 2 H NMR spectra of DPPC-d31/erg as a function of erg concentration at T 45 C. Figure 6 2 H NMR spectra as a function of temperature for 84:16 DPPC-d31/erg. and lo components (see discussion of Fig. 14 below). At 55 C and 60 C, a pure liquid crystalline phase is observed, where the individual peaks of the depaked spectra become sharp. To explore the phase change across all erg concentrations and temperatures, we have plotted M 1 as a function of temperature for all DPPC-d31/erg samples shown in Fig. 7. Pure DPPC-d31 undergoes a pretransition at 31 C from the L phase to the P phase and a main transition at 40 C from the P phase to the L phase as the temperature is raised. Note that the L phase and the P phase are so phases, and the L phase is synonymous with the ld phase. The M 1 (T) curve plunges less dramatically at T m as more erg is added: below T m, M 1 (T) decreases with increasing erg concentration, whereas above T m, M 1 (T) increases with increasing erg concentration. The so to ld transition for DPPC-d31/erg membranes containing erg concentrations between 10 mol% and 20 mol% erg occurs at a constant temperature, T m C, implying the existence of a three-phase (so ld lo) line in the phase diagram. The solid circles in Fig. 1, determined from the midpoint of the so to ld transition, define the three-phase line. For DPPC-d31/erg membranes containing erg concentrations of 25 mol% and above, the so to ld phase transition is absent. The Figure 7 The temperature dependence of M 1 for DPPCd31/erg at various erg concentrations., 0%;, 5%;, 10%; E, 13%;, 16%; ƒ, 20%;, 25%;, 27.5%; Œ, 30%;, 35%;, 42%.

7 MEMBRANE PHASE DIAGRAMS 41 transitions in M 1 (T) of 25mol% erg 30mol% erg membranes correspond to phase changes from a so lo phase coexistence to a pure lo state. The temperatures at which the onsets of transitions occur are obtained and shown as open triangles in Fig. 1. Spectral Subtraction We now discuss the determination of the phase boundaries below the three-phase line. The boundaries of the so lo coexistence region are determined by the spectral subtraction method (7, 12). Within the two-phase region, the 2 H NMR spectrum S isasuperposition of the weighted so and lo spectra: S x A, T f A S f x f, T 1 f A S s x s, T [1] S x B, T f B S f x f, T 1 f B S s x s, T [2] where x A and x B are the ergosterol concentrations of two DPPC-d31/erg MLDs, x s and x f are the ergosterol concentrations at the so and lo phase boundary, f A and f B are the fractions of lo phospholipid in the two samples, and S s and S f are the end-point spectra characteristic of so- and lo-phase domains, respectively, at the so and lo phase boundaries. Given two areanormalized 2 H NMR spectra S( x A, T) and S( x B, T), the so (or lo) spectrum can be obtained by subtracting a fraction K (or K ) of one spectrum from the other, i.e. S s x s, T S x A, T KS x B, T [3] S f x f, T S x B, T K S x A, T [4] where K is the ratio of lo phase phospholipid fractions in the two samples, f A /f B, and K is the ratio of so phase phospholipid fractions in the two samples (1 f B )/(1 f A ). Using the K and K values determined from the spectral subtraction, the phase boundaries x s and x f can then be calculated: x s 1 x B x A K 1 x A x B 1 x B K 1 x A x f 1 x A x B K 1 x B x A 1 x A K 1 x B [5] [6] We now demonstrate spectral subtraction using 13 mol% erg and 20 mol% erg samples at 33 C. The spectra of 13 mol% erg and 20 mol% erg in Fig. 4 are chosen to be S(x A,T) and S(x B,T), respectively. The end-point spectra S s (x s,t) and S f (x f,t) in Eq. [3] and Figure 8 The difference spectra obtained from spectral subtraction using K 0.37, 0.4, 0.43, and 0.46 at T 33 C. The arrows indicate the edge features mentioned in the text. Eq. [4] can be obtained using K 0.31 and K 0.49, respectively. At the best K value (i.e., 0.31), the lo components from the first and second terms in Eq. [3] cancel out, such that the difference spectrum contains only the so component. If K is smaller than the best value, the lo component do not disappear completely. The features associated with the lo spectrum, i.e. the methyl peaks near 15 khz and 20 khz and the vertical plateau edges, will be found in the difference spectrum. If K is larger than the best value, a negative lo component is introduced to the subtracted spectrum, causing an edge feature visible around 28 khz as shown in Fig. 8. The criterion for determining the best K value is to make K large enough such that the methyl peaks near 15 khz and 20 khz and the vertical plateau edges disappear from the difference spectrum, but not too large such that the edge feature around 28 khz does not appear. It is difficult to determine precisely the best K value by eye examining the presence/absence of these features, and this uncertainty would be the error in K. x s is calculated to be 0.08 by plugging the actual erg concentrations of the 13 mol% and 20 mol% samples (i.e., x A mol% and x B mol%, respectively), and K 0.31 into Eq. [5]. The normalized so endpoint spectrum (see Fig. 9(A)) obtained at K 0.31 is compared with the spectrum measured near the

8 42 HSUEH, ZUCKERMANN, AND THEWALT Figure 9 (A) The so end-point spectrum obtained from spectral subtraction at T 33 C, normalized in area. (B) The difference between two normalized spectrum: the so end-point spectrum in (A) and the spectrum of the 5 mol% sample in Fig. 4. component do not disappear completely. The contribution of a so component to the area around the plateau edge ( 27 khz) can be considered as adding a negative-slope line to that area, which makes the intensity near the plateau edge of the difference spectrum lower than that of the pure lo spectrum as shown in Fig. 10(A). If K is larger than the best value, a negative so component is introduced to the difference spectrum. Its contribution to the area around 27 khz can be considered as adding a positive-slope line to that area, which causes the intensity near the plateau edge of the difference spectrum higher than that of the pure lo spectrum as shown in Fig. 10(C). The reason of choosing the area near the plateau edge to examine is because the presence of the so component is relatively prominent in this area than the rest of the spectrum. By plugging the actual erg concentrations of the 13 mol% and 20 mol% samples, and K 0.44 into Eq. [6], x f is calculated to be The lo end-point spectrum (see Fig. 10(B)), obtained at K 0.49, is compared with the spectrum measured near the so lo/lo boundary, i.e. the 30 mol% erg sample in Fig. 4. The difference between the two so/so lo boundary, i.e. that of the 5 mol% erg sample in Fig. 4. The difference between the two spectra, shown in Fig 9(B), indicates that these two spectra are not too different. At the best K value (i.e., 0.49), the so components from the first and second terms in Eq. [4] cancel out, such that the difference spectrum contains only the lo component. To determine the best K value, the methyl peaks close to the plateau edge ( 27 khz) of the difference spectrum is compared with that of a measured pure lo spectrum obtained from the 35 mol% sample. In principle, the end-point spectrum obtained from spectral subtraction is expected to be identical to the measured spectrum obtained at the phase boundary. Because the pure lo spectrum changes little with increasing erg concentration, the spectrum obtained at the phase boundary (somewhere between 20 mol% and 30 mol%, as discussed in Fig. 4) and that obtained at 35 mol % should be very much alike. Figure 10 shows the difference spectra obtained from various K values (the black lines), and each overlaps with a pure lo spectrum of the 35 mol% sample (the gray lines). All spectra in Fig. 10, except Fig. 10(D), are normalized. Figure 10(B) shows that at the best K value, the black and gray lines match well. If K is smaller than the best value, the so Figure 10 Comparison of the difference spectra obtained from spectral subtraction using (A) K 0.39, (B) K 0.49, and (C) K 0.59, with the pure lo spectrum of 65:35 DPPC-d31/erg. Each difference spectrum (black line) overlaps with the pure lo spectrum (gray line). All spectra are normalized; (D) The difference between the normalized lo end-point spectrum in (B) and the normalized spectrum of 70:30 DPPC-d31/erg in Fig. 4. T 33 C.

9 MEMBRANE PHASE DIAGRAMS 43 normalized spectra shown in Fig 10(D) indicates that these two spectra are nearly identical. Together with the discussion of Fig. 9 where the so end-point spectrum is similar to the spectrum measured near the phase boundary, the spectral subtraction method is valid. In addition to 13 mol% erg and 20 mol% samples, we have also performed spectral subtraction using 13 mol% erg and 16 mol% erg as well as 16 mol% erg and 20 mol% erg. All three sets of data give consistent x s and x f values within error. The final x s (or x f )is obtained by taking the average of all three x s (or x f ). Note that although the 25 mol% erg sample is also in the two-phase region, it is too close to the so lo/lo boundary. The spectra are distorted near the phase boundary, giving incorrect x s and x f values. Therefore, spectral subtraction involving 25 mol% erg sample is not valid. In addition, two samples having very different erg concentrations, e.g. 13 mol% erg and 20 mol% samples, will give a difference spectrum of a better signal-to-noise ratio. Figure 11 T 2e as a function of temperature for various membranes., 95:5 DPPC-d31/erg;, 70:30 DPPC-d31/ erg; Œ, 65:35 DPPC-d31/erg. T 2e Correction to Spectral Subtraction The spectral subtraction method is valid as long as certain assumptions hold. The spectra of the two phases, S s and S f, have to be sufficiently different that one can easily distinguish and carry out the subtraction procedure. The exchange of labeled lipid between two kinds of domains must be slow on the NMR time scale so that it can be neglected. In addition, the domain must be sufficiently large so that the signal from the lipid on the boundary of the domains is negligible. Also, this method assumes that both phases have the same relaxation time T 2e, which is not true in this case. Figure 11 displays the T 2e values measured near the so lo phase boundaries. The T 2e values of the pure so phase obtained from 95:5 DPPCd31/erg are compared with those of the pure lo phase obtained from 65:35 DPPC-d31/erg or 70:30 DPPCd31/erg above 33 C. We found that the lo phase has a T 2e one to six times larger than the so phase T 2e (depending on the temperature), as shown in Fig. 11, hence the so component decays with time faster than the lo component (the signal e t/t 2e )). Thus, at any given quadrupolar echo time 2, the 2 H NMR spectrum will contain a smaller so component S s than is representative of the sample. f A and f B in Eq. [1] and Eq. [2], which should be denoted as f A (t 2 ) and f B (t 2 ), do not reflect the actual fraction of fluid phospholipid in the samples due to this T 2e effect. Thus the K and K (which should be denoted as K(t 2 ) and K (t 2 ), respectively) determined from the spectral subtraction will not be correct, leading to a deviation of x s and x f from the actual values (16). To eliminate this T 2e effect, corrected f A and f B values (i.e., f A (t 0) and f B (t 0)) are calculated by extrapolating the height of the respective echo signal back to t 0 using the measured T 2e for a given temperature, and then the corrected K and K values (i.e., K(t 0) and K (t 0)) can be derived and expressed in terms of the experimentally determined K, K (i.e., K(t 2 ), K (t 2 )), and T 2e s. The new K and K are plugged into Eq. [5] and Eq. [6] to obtain the corrected x s and x f. The x s and x f values with and without the T 2e correction are listed in Table 1. Note, the lo phase T 2e s used to calculate the corrected values below 33 C are from 65:35 DPPC-d31/erg and that at 35 C is from 70:30 DPPC-d31/erg. The x s and x f values obtained at 27 C and 29 C remain unchanged after the correction. At higher temperatures ( 31 C), where the T 2e discrepancy in so and lo phases is more prominent, the corrections cause the x s and x f values Table 1 Comparison of the x s and x f Values and Those with T 2e Corrections No Correction With T 2e Correction T ( C) x s x f x s x f

10 44 HSUEH, ZUCKERMANN, AND THEWALT Figure 12 The depaked spectra of DPPC-d31/erg as a function of erg concentration at T 45 C. Because the spectra are symmetrical, we show only the high-frequency half of each. ld/ld lo phase boundary lies between 10 mol% and 13 mol% erg and the ld lo/lo phase boundary between 27.5 and 30 mol% erg. We have plotted M 1 as a function of erg concentration at various temperatures above T m in Fig. 13. Because in the liquid crystalline phase M 1 is proportional to the average order parameter, each curve in Fig. 13 displays the progression of the average order parameter with increasing erg concentration at a given temperature. The M 1 (erg) curves, except at 57 C, exhibit two distinct erg concentration-dependent behaviors: below 27.5 mol% erg the average order parameter increases rapidly as erg concentration increases, whereas above 27.5 mol% erg the average order parameter increases slowly or levels off as erg concentration increases. As we know from the NMR spectra at these temperatures, the 75:25 DPPC-d31/ erg has coexisting ld lo phases while the 70:30 DPPC-d31/erg is in the lo phase. The average order parameter of membranes containing coexisting lo and ld phases is expected to be most sensitive to increasing erg concentration as a proportion of ld phase will be converted to the much more ordered lo phase upon erg addition. For lo phase membranes the effect of added erg will be more modest. Therefore, the changes of slope in the M 1 (erg) curve suggests a ld lo/lo boundary near 27.5 mol% erg. To determine this boundary, we drew a line through the points for 20 and 25 mol% erg (or 25 and 27.5 mol%) and fitted to increase 1 2 mol%. However, taking the error into account, all corrected x s and x f agree with the corresponding uncorrected values. The corrected x s and x f values are plotted in Fig.1 (solid squares). The ld lo Phase Boundaries To further investigate the ld lo phase coexistence region, the depaked spectra are examined. The depaked spectrum provides better resolution in observing the variations in spectral width and sharpness of individual peaks. Figure 12 shows the depaked spectra as a function of erg concentration at 45 C. At low erg concentrations ( 10 mol%), the spectral lines are sharp. They broaden from 13 mol% erg to 27.5 mol% erg (and the spectral width also increases rapidly for this concentration range) and are sharp again at erg concentrations 30 mol%. Lipids diffusing between ld and lo domains with a rate faster than the NMR time scale will yield an averaged spectrum containing broadened individual peaks. Therefore the spectral subtraction method cannot be applied to this region. On the other hand, Fig. 12 indicates that the Figure 13 M 1 as a function of erg concentration for DPPC-d31/erg., 41 C;, 42 C;, 43 C; V, 44 C; Œ, 45 C; E, 47 C;, 48 C;, 49 C, 50 C;, 53 C;, 57 C. The lines are guides to the eye.

11 MEMBRANE PHASE DIAGRAMS 45 another line to the points between mol% erg (or mol%). The open circles in Fig. 1, denoting the ld lo/lo boundary, are obtained from the intercepts of the two lines at various temperatures. In principle there should be a similar change in slope at low erg concentrations, reflecting crossing the ld/ld lo phase boundary. However, these slope changes are less obvious than those at the ld lo/lo boundary. Thus the ld/ld lo phase boundary has instead been determined by direct examination of the depaked spectra shown in Fig. 12. As discussed in Fig. 12, the ld/ld lo phase boundary at 45 C lies between 10 and 13 mol% erg, we therefore define the phase boundary to be midway between 10 and 13 mol% erg, i.e mol% erg, and the error to be the distance to 11.5 mol%. The ld/ld lo phase boundary, indicated by the open diamonds in Fig. 1, is obtained. The T 41 C curve in Fig. 13 levels off at an erg concentration of 27.5 mol%, implying that at 41 C the DPPC chain has reached its maximum sterolenhanced order. As noted previously, the T 57 C curve behaves differently from the others; in particular, the value of M 1 increases steadily with increasing erg concentration all the way to 42 mol% erg, implying no phase boundary crossing. Another approach to determining the ld lo/lo boundary is to examine the temperature-dependent variation in width of individual peaks of the depaked spectrum. Figure 14 shows the depaked spectra of 25 mol% erg from 37 C, where the membrane is in the lo phase, to 60 C, where the membrane is in the ld phase. There is a slow decrease in the quadrupolar splittings from T 37 C to 41 C. The individual peaks in the spectrum remain sharp. Above 41 C, the individual peaks broaden significantly and the quadrupolar splittings decrease faster as a function of temperature, indicating the onset of ld lo phase coexistence. The broad individual peaks persist at higher temperatures until at 53 C the individual peaks become narrow again, implying that the membrane no longer displays ld lo phase coexistence. The individual peaks remain narrow at higher temperatures and the rate of decrease in the quadrupolar splittings as a function of temperature slows down. Thus a second phase boundary (from ld lo to a single liquid crystalline phase) occurs around 53 C. Similar analysis of the spectra of 72.5:27.5 DPPC-d31/erg yielded phase boundaries for the ld lo coexistence region at 43 C and 50 C. These results, indicated as solid triangles in Fig. 1, agree well with the open circles obtained from the analysis of M 1 (erg) curves. Figure 14 The depaked spectra of 75:25 DPPC-d31/erg as a function of temperature. CONCLUSION We have explained how a phospholipids/sterol partial phase diagram is obtainable exclusively from 2 H NMR spectra. The model membrane we used to illustrate this, DPPC/ergosterol, exhibits both so lo and ld lo coexistence regions with a clear three-phase line separating them. Our observation of ld lo phase coexistence provides evidence that two liquid crystalline phases can coexist even in model membranes containing no proteins. Thus rafts in cell membranes may be strongly influenced by lipid/lipid interactions. The phase diagram is obtained by analyzing the NMR spectra in various ways. The boundaries of the so lo coexistence region are obtained using the spectral subtraction method. In contrast to earlier NMR phase diagram determinations, we have been able to definitively locate and characterize the ld lo coexistence region. The boundaries of the ld lo coexistence region are obtained by examining the depaked spectra as a function of temperature and erg concentration, as well as by analyzing M 1 (erg) curves. Note that these methods require a reasonable number of samples/ spectra to determine the lo ld phase boundaries accurately. Fewer samples/spectra will result in larger

12 46 HSUEH, ZUCKERMANN, AND THEWALT error bars. We believe that these observations may be generalized to other sterol-containing model membranes. ACKNOWLEDGMENTS This work was supported by research grants from the Natural Sciences and Engineering Research Council of Canada. Ya-Wei Hsueh received support from National Science Council of Taiwan. 13. Davis JH, Jeffrey KR, Bloom M, Valic MI, Higgs TP Quadrupolar echo deuteron magnetic resonance spectroscopy in ordered hydrocarbon chains. Chem Phys Lett 42: Lafleur M, Fine B, Sternin E, Cullis PR, Bloom M Smoothed orientational order profile of lipid bilayers by 2 H-nuclear magnetic resonance. Biophys J 56: Thewalt JL, Bloom M Phosphatidylcholine: cholesterol phase diagrams. Biophys J 63: Thewalt JL, Hanert CE, Linseisen FM, Farrall AJ, Bloom M Lipid-sterol interactions and the physical properties of membranes. Acta Pharm 42:9 23. REFERENCES 1. Gorter E, Grendel F On bimolecular layers of lipoids on the chromocytes of the blood. J Exp Med 41: Danielli JF, Davson H A contribution to the theory of permeability of thin films. J Cell Comp Physiol 5: Singer SJ, Nicolson GL The fluid mosaic model of the structure of cell membranes. Science 175: Simons K, Ikonen E Functional rafts in cell membranes. Nature 387: Brown DA, London E Functions of lipid rafts in biological membranes. Ann Rev Cell Dev Biol 14: Silvius JR Role of cholesterol in lipid raft formation: lessons from lipid model systems. Biochim Biophys Acta 1610: Vist M, Davis JH Phase equilibria of cholesterol/ dipalmitoyl-phosphatidylcholine mixtures: 2 H nuclear magnetic resonance and differential scanning calorimetry. Biochemistry 29: Ipsen JH, Karlstrom G, Mouritsen OG, Wennerstrom H, Zuckermann MJ Phase-equilibria in the phosphatidylcholine-cholesterol system. Biochim Biophys Acta 905: Seelig J Deuterium magnetic resonance: theory and application to lipid membranes. Q Rev Biophys 10: Stockton GW, Smith ICP Deuterium nuclear magnetic resonance study of condensing effect of cholesterol on egg phosphatidylcholine bilayer membranes. 1. Perdeuterated fatty-acid probes. Chem Phys Lipids 17: Davis JH The description of membrane lipid conformation, order and dynamics by 2 H-NMR. Biochim Biophys Acta 737: Hsueh Y-W, Gilbert K, Trandum C, Zuckermann M, Thewalt J The effect of ergosterol on dipalmitoylphosphatidylcholine bilayers: a deuterium NMR and calorimetric study. Biophys J (in press). BIOGRAPHIES Ya-Wei Hsueh is currently a professor of physics at National Central University in Taiwan. She received her Ph.D. in physics at University of Toronto, Canada, in Her doctoral work focused on the study of hightemperature superconductors by NMR. She did her postdoctoral work at Simon Fraser University (Canada), where she studied membrane structure/phase behavior by NMR. Her current research interests are domain formation and mechanical properties in model membranes by NMR and AFM. Martin Zuckermann earned his D.Phil. in Theoretical Physics at Oxford University, England, in After postdoctoral training at the University of Chicago with Jim Phillips and Leo Falicov, he joined the Physics Department of the University of Virginia as an assistant professor in 1965 and the Physics Department of Imperial College, London, as lecturer in In 1969, he was appointed associate professor in the Physics Department of McGill University where he remained until 1999, when he retired from the post of Macdonald Professor of Physics. He is currently adjunct professor of physics at Simon Fraser University, but he retains the position of Macdonald Emeritus Professor of Physics at McGill University. He is a Fellow of the Royal Society of Canada and an International Member of the Royal Danish Society of Letters and Science. Jenifer Thewalt earned her doctorate in chemistry at Simon Fraser University (SFU) in After postdoctoral training in physics at the University of British Columbia with Myer Bloom, in 1995 she was jointly appointed as an assistant professor in the Physics Department and the Institute of Molecular Biology and Biochemistry (MBB) at SFU. She currently holds a joint appointment as associate professor in the Departments of MBB and Physics at SFU, as well as an adjunct professorship in medicine at the University of British Columbia.