The dependence of lipid asymmetry upon phosphatidylcholine acyl chain structure.

Lipid asymmetry, the difference in inner and outer leaflet lipid composition, is an important feature of biomembranes. By utilizing our recently developed MβCD-catalyzed exchange method, the effect of lipid acyl chain structure upon the ability to form asymmetric membranes was investigated. Using this approach, SM was efficiently introduced into the outer leaflet of vesicles containing various phosphatidylcholines (PC), but whether the resulting vesicles were asymmetric (SM outside/PC inside) depended upon PC acyl chain structure. Vesicles exhibited asymmetry using PC with two monounsaturated chains of >14 carbons; PC with one saturated and one unsaturated chain; and PC with phytanoyl chains. Vesicles were most weakly asymmetric using PC with two 14 carbon monounsaturated chains or with two polyunsaturated chains. To define the origin of this behavior, transverse diffusion (flip-flop) of lipids in vesicles containing various PCs was compared. A correlation between asymmetry and transverse diffusion was observed, with slower transverse diffusion in vesicles containing PCs that supported lipid asymmetry. Thus, asymmetric vesicles can be prepared using a wide range of acyl chain structures, but fast transverse diffusion destroys lipid asymmetry. These properties may constrain acyl chain structure in asymmetric natural membranes to avoid short or overly polyunsaturated acyl chains.

stability of membranes, and membrane shape ( 4-10 ). Therefore, biologically realistic asymmetric vesicles could be widely useful in model membrane studies. However, artifi cial membranes produced in the laboratory usually have a symmetric distribution of lipids in the inner and outer leafl ets of the bilayer. Asymmetric artifi cial membranes have been diffi cult to produce, although some important progress has been made (11)(12)(13)(14)(15)(16). We recently developed a robust method in which methyl-␤ -cyclodextrin (M ␤ CD)-induced lipid exchange is used to prepare model membrane vesicles with lipid asymmetry, and we have applied the method to produce asymmetric small, large, and giant unilamellar vesicles (17)(18)(19).
In the present study, we extended the range of this method by using it to construct asymmetric vesicles containing SM in the outer leafl et and various species of PCs in the inner leafl et. PC acyl chain structure was systematically varied, and how this variation affects the ability to form asymmetric membranes was assayed. We found that whether the resulting vesicles were asymmetric [SM outside/ PC inside (SMo/PCi)] depended upon PC acyl chain structure. Despite this limitation, asymmetric vesicles with a wide range of acyl chain structures could be produced. This is an important advance that should allow new applications of asymmetric vesicles. These studies also allowed analysis of the effect of lipid type upon membrane structure and behavior. We investigated the origin of the acyl chain dependence of asymmetry by measuring the transverse diffusion rate of the various PCs used. There was a correlation between asymmetry and transverse diffusion. Asymmetric vesicles could be prepared only when PCs with slow transverse diffusion were used. It was found that the presence of two polyunsaturated acyl chains or overly short acyl chains prevented the maintenance of asymmetry. This Abstract Lipid asymmetry, the difference in inner and outer leafl et lipid composition, is an important feature of biomembranes. By utilizing our recently developed M ␤ CDcatalyzed exchange method, the effect of lipid acyl chain structure upon the ability to form asymmetric membranes was investigated. Using this approach, SM was effi ciently introduced into the outer leafl et of vesicles containing various phosphatidylcholines (PC), but whether the resulting vesicles were asymmetric (SM outside/PC inside) depended upon PC acyl chain structure. Vesicles exhibited asymmetry using PC with two monounsaturated chains of >14 carbons; PC with one saturated and one unsaturated chain; and PC with phytanoyl chains. Vesicles were most weakly asymmetric using PC with two 14 carbon monounsaturated chains or with two polyunsaturated chains. To defi ne the origin of this behavior, transverse diffusion (fl ip-fl op) of lipids in vesicles containing various PCs was compared. A correlation between asymmetry and transverse diffusion was observed, with slower transverse diffusion in vesicles containing PCs that supported lipid asymmetry. Thus, asymmetric vesicles can be prepared using a wide range of acyl chain structures, but fast transverse diffusion destroys lipid asymmetry. These properties may constrain acyl chain structure in asymmetric natural membranes to avoid short or overly polyunsaturated acyl chains. The eukaryotic plasma membrane exhibits asymmetry with respect to its lipid distribution across the bilayer; the majority of sphingomyelin (SM) and phosphatidylcholine (PC) is in the outer leafl et, and the aminophospholipids phosphatidylethanolamine (PE) and phosphatidylserine (PS) are in the cytoplasmic leafl et (1)(2)(3). It is known that the asymmetric distribution of lipids in the bilayer affects various biological properties, such as membrane permeability, membrane potential, surface charge, the mechanical mixture and vortexed at 55°C for 30 min. After cooling for 5 min, samples were centrifuged at 49,000 g for 5 min using an air-driven microultracentrifuge (Beckman Airfuge). The supernatant was applied onto a Sepharose CL2B column (dimensions, 25 cm length and 1 cm diameter). We found that Sepharose CL2B gave better resolution of small and large vesicles than did the Sepharose CL4B we used previously. Fractions of 1 ml were collected, with SUVs mainly eluting into fractions [16][17][18][19]. These fractions were combined for further analysis. Generally the lipid concentration in the peak SUV fractions was ‫ف‬ 150 M as measured by the intensity of lipid bands on HP-TLC relative to standards with known amounts of lipid (see below). In cases in which lipid concentration was not explicitly measured, a 150 M lipid concentration was assumed unless otherwise noted.

Analysis of lipid composition by HP-TLC
Lipids were extracted with chloroform/methanol (1:1 v/v) from ordinary and exchange vesicles as described ( 18 ). The extracted lipids or pure lipid standards were applied to HP-TLC (Silica Gel 60) plates and chromatographed in solvent as described ( 17 ). The lipids in the vesicles were quantifi ed by measuring TLC spot intensity versus a standard curve for each lipid generated on the same TLC plate as described ( 17 ).

Fluorescence measurements
Fluorescence was measured on a SPEX FluoroLog 3 spectrofl uorometer (Jobin-Yvon, Edison, NJ) using quartz semimicro cuvettes (excitation path length, 10 mm; emission path length, 4 mm) according to described protocols ( 17 ). DPH fl uorescence was measured at an excitation wavelength of 358 nm and emission wavelength of 427 nm. TMADPH fl uorescence was measured at an excitation wavelength of 364 nm and emission wavelength of 426 nm. C 6 -NBD-PC fl uorescence was measured at an excitation wavelength of 465 nm and emission wavelength of 534 nm. The slit bandwidths for fl uorescence measurements were generally set to 4.2 nm (2 mm physical size) for excitation and to 4.2 nm (2 mm physical size) for emission. The backgrounds were less than 1% of fl uorescence-containing samples and were not subtracted from the sample signals. An exception to this was for measurements of TMADPH fl uorescence anisotropy (see below).

Steady-state fl uorescence anisotropy measurements
DPH or TMADPH dissolved in ethanol (stock solutions of 20 µM) were added to preformed vesicles to a concentration of 0.1 mol% of total lipid. Anisotropy measurements were made using a SPEX automated Glan-Thompson polarizer accessory according to previously described protocols ( 17 ). For DPH, the bandwidths of slit for anisotropy were set to 4.2 nm (2 mm physical size) for both excitation and emission. For TMADPH, the bandwidths of slit for anisotropy were set to 4.2 nm (2 mm physical size) for excitation and to 8.4 nm (4 mm physical size) for emission. For TMADPH, the anisotropy was calculated after subtraction of fl uorescence intensity in background samples lacking fl uorophore. The anisotropy was determined using the following equation: A = [(( I vv × I hh )/( I vh × I hv )) -1]/[(( I vv × I hh )/( I vh × I hv )) + 2], where A represents anisotropy, I represents the fl uorescence emission intensity, with the vertical ( v ) and horizontal ( h ) orientation of the excitation and emission polarization fi lters ( 20 ).

Vesicle scrambling experiments
The scrambling process, in which vesicles are destroyed and then reformed as symmetric vesicles, was carried out similarly to previously described protocols ( 17 ). Fractions 16-19 from CL2B column were combined and then divided into four tubes (1 ml each). Two of the samples were subjected to scrambling, and two fi nding may explain why such lipids are not generally abundant in biological membranes.
Concentrations of lipids were measured by dry weight. Concentrations of DPH and TMADPH were determined by absorbance as described previously ( 17 ). Sepharose CL-4B and Sepharose CL-2B were purchased from Amersham Biosciences (Piscataway, NJ). High-performance thin-layer chromatography (HP-TLC) plates (Silica Gel 60) were purchased from VWR International (Batavia, IL). The digital thermometer was purchased from Fisher Scientifi c (Pittsburgh, PA).

Formation of asymmetric SUV
Asymmetric small unilamellar vesicles (SUV) were formed using a protocol adapted from that described in Cheng et al. ( 17 ). First, donor multilamellar vesicles (MLV), the population from which bSM would be exchanged with the outer leafl et of the SUVs, were prepared. Donor MLVs composed of 16 mM bSM were formed by drying the lipid in a fi lm under nitrogen followed by high vacuum for at least 1 h, followed by hydration at 70°C with 500 µl of PBS (1.8 mM KH 2 PO 4 , 10 mM Na 2 HPO 4 , 137 mM NaCl, and 2.7 mM KCl, pH 7.4). This sample was vortexed in a multitube vortexer (VWR International) at 55°C for 15 min and then subjected to centrifugation at 11,000 g for 5 min at room temperature to remove any small vesicles present prior to lipid exchange. The supernatant was discarded, the pellet was resuspended with 500 l PBS, and then 95 l of 390 mM M ␤ CD dissolved in PBS was added. This mixture was vortexed in the multitube vortexer at 55°C for 2 h.
Next, an acceptor solution, which contains sonicated PC SUV, was prepared. We formed 600 l of 8 mM PC MLVs by drying the lipid in a fi lm and rehydrating it with 70°C in PBS just like the SM MLVs described above. This mixture was also vortexed in a multitube vortexer (VWR International) at 55°C for 15 min and then sonicated in a bath sonicator (Special Ultrasonic Cleaner Model G1112SP1, Laboratory Supplies Co., Hicksville, NY) at room temperature for at least 15 min until the solution became nearly transparent, and then it was diluted to 4 mM with PBS.
Third, this 500 l of 4 mM sonicated SUVs containing unsaturated glycerophospholipid were added to the SM MLV-M ␤ CD

Dynamic light scattering measurements
The measurement of SUV size was carried out at 23°C using Protein Solutions DynaPro 99 dynamic light -scattering instrument (Wyatt Technology, Santa Barbara, CA) ( 18 ). Fractions 16-19 from the Sepharose CL2B column were subjected to the dynamic lightscattering experiment 5 min and 1 day after eluting from the column. Data were analyzed with the software Dynamics version 5.25.44 (Protein Solutions, Inc. Charlottesville, VA).

Preparation of asymmetric (exchange) SUVs and their SM content in the inner and outer leafl ets
We used M ␤ CD-induced lipid exchange methods to produce asymmetric SUVs with SM in the outer leafl et and PC in the inner leafl et. In this protocol, the lipid to be introduced into the outer leafl et is in donor MLV, whereas the lipid to be located in the inner leafl et is in acceptor SUV. One change made from our previous protocol was to fractionate vesicles using Sepharose CL2B, which gave better resolution of vesicles by size than the Sepharose CL4B used previously ( 17 ). Using this approach, the properties of acceptor SUV having PCs with various acyl chains were investigated to determine whether the capability to form asymmetric vesicles is affected by lipid acyl chain structure. The PCs tested had monounsaturated acyl chains of different lengths (di14:1PC, di16:1PC, di18:1PC, di20:1PC, and di22:1PC); diphytanoyl chains, which have four methyl groups per chain (diphyPC); one saturated and one polyunsaturated chain (16:0-18:2PC and 16:0-20:4PC); and two polyunsaturated chains (di18:2PC, di18:3PC, and di20:4PC). HP-TLC analysis of the acceptor vesicles after exchange (i.e., exchange vesicles) revealed that they contained both SM and PC. Based on HP-TLC, the SM content in exchange vesicles was in the range 50-70% ( Table 1 ). Because the theoretical value for complete exchange of the entire were used without scrambling. The samples to be scrambled were dried by a nitrogen stream, dissolved in 20 l of ethanol, and then dispersed at 70°C in 980 l of distilled water (which should reconstitute the PBS as well as the lipid vesicles). After cooling to room temperature, DPH or TMADPH, which was dissolved in ethanol, was added as described above.

Measurement of lipid transverse diffusion (fl ip-fl op)
The transverse diffusion of C 6 -NBD-PC was measured using the dithionite method ( 21 ). First, 1 ml of 100 M SUVs were prepared by ethanol dilution, and then 1 mol% C 6 -NBD-PC (from a stock solution dissolved in ethanol) was added to the preformed vesicles, which were then incubated overnight at room temperature. Alternately, large unilamellar vesicles (LUV ) were prepared by lipid extrusion technique using a Mini-Extruder set (Avanti Polar Lipids, Alabaster, AL). A sample containing 1 ml of MLVs (1 mM lipid) were disrupted by fi ve freeze-thaw cycles and then forced 15 times through a polycarbonate fi lter with a pore size of 100 nm. The sample was then diluted to 100 µM lipid with PBS. Then C 6 -NBD-PC was added (to 1 mol% of lipid). Next, initial NBD fl uorescence emission ( F 0 ) was measured. Then, for both SUV and LUV, 10 l of 1.0 M sodium dithionite freshly prepared in 1.0 M Tris pH 10.0 was added, and the fl uorescence emission intensity ( F P ) was recorded versus time for up to 1 h. The fraction of C 6 -NBD-PC protected from external dithionite due to location in the inner leafl et is given by the ratio F P / F 0. The fraction of C 6 -NBD-PC that had fl ipped to the inner leafl et was estimated from the F P / F 0 value at time = 0, determined by extrapolation from fl uorescence values between 10 and 60 min after dithionite addition, assuming that the leakage of dithionite into the vesicles followed a fi rst-order reaction. All calculations were performed by using the software Slide-Write Plus (Advanced Graphics Software, Inc., Encinitas, CA).

Prevention of lipid oxidation in polyunsaturated lipids
The possibility of lipid oxidation that might occur was determined by the increase in absorption at 233 nm. The polyunsaturated PCs that we used, such as 16:0-18:2PC, 16:0-20:4PC, di18:2PC, di18:3PC, and di20:4PC, are highly oxidation-prone especially during the process of sonication. Free radical autoxidation of polyunsaturated lipids is accompanied at an early stage by diene conjugation, which results in signifi cant absorbance at 233 nm ( 22 ). To minimize oxidation during vesicle preparation, prior to sonication, the test tubes containing polyunsaturated PC MLVs were tightly sealed with tefl on tape, after a nitrogen gas stream was directed down into the test tube. For all of our lipids, we estimated at most a small percentage of conjugated diene was present, indicating little oxidation. We also carried out the all experiments immediately after the formation of the exchange vesicles, so that all experiments on the exchange vesicles were completed within 1 h of their preparation .

Percentage SM in exchange vesicles calculation
The percentage of SM in exchange vesicles was calculated from HP-TLC and, for the outer leafl et, TMADPH anisotropy. TLC data gives total percentage of SM in the bilayer (both leaflets), and percentage of SM derived from the anisotropy of TMADPH incorporated into the outer leafl et, when compared with a standard curve (supplemental To determine the percentage of outer leafl et lipids that were SM after exchange, we examined the steady-state fl uorescence anisotropy of TMADPH. This probe is localized to the outer leafl et of the bilayer when it is added to the preformed vesicles due to its charged quaternary amino group. First, a standard curve of anisotropy versus SM content was generated in ordinary, symmetric vesicles for each of the PCs used. Anisotropy increases with SM content because at room temperature SM tends to form the solid-like gel state, in which TMADPH is highly ordered and has high anisotropy, whereas by themselves, the PCs used are in the liquid disordered (Ld) state, in which TMADPH exhibits low anisotropy. Next, the anisotropy in the exchange vesicles (supplemental Table I ( Table 1 and Fig. 1 ).
The percentage of inner leafl et lipids that were SM in the exchange vesicles was then calculated from the total SM in the vesicles ( Fig. 1 ). The SM content in the inner leafl et of exchange SUVs was found to be highly dependent on the type of PC present. The inner leafl et SM values were close to zero in many cases (di16:1PC, di18:1PC, di20:1PC, di22:1PC, diphyPC, and 16:0-18:2PC), indicating that SM in those vesicles was largely restricted to the outer leafl et (i.e., they were SMo/PCi vesicles). In contrast, SM was less asymmetrically distributed (i.e., located to a substantial extent in both inner and outer leafl ets of the bilayer) for PC with short acyl chains (di14:1PC) or highly polyunsaturated acyl chains (16:0-20:4PC, di18:2PC, di18:3PC, and di20:4PC). In fact, in the case of lipids with two polyunsaturated chains, the inner and outer leafl et SM concentrations were almost the same, indicative of no asymmetry.

Assessment of asymmetry in exchange vesicles compared with scrambled vesicles with the same lipid composition
It is important to prove that exchange vesicles are or are not asymmetric with additional methods. When exchange vesicles are asymmetric, they have different properties than the corresponding symmetric vesicles ( 17,18 ). The presence or absence of such differences was used to confi rm the presence or absence of asymmetry. To prepare ordinary symmetric vesicles with the same lipid compositions as those in the exchange vesicles, a protocol was used in which the exchange vesicles were dissolved in organic solvent and the lipids were scrambled by reforming vesicles from them to yield ordinary symmetric vesicles (i.e., scrambled vesicles).
The fi rst assay used to compare physical properties of exchange and scrambled vesicles was outer leafl et membrane order measured by TMADPH anisotropy (supplemental Table I and Fig. 2 ). As noted above, in asymmetric vesicles containing SM in their outer leafl et, the outer leaflet membrane order is especially high at 23°C because the SM is primarily in the solid-like gel state. For the lipid compositions studied, outer leafl et order is lower when vesicles  Table 1 and for the inner leafl et calculated from data in Table 1  are symmetric due to the introduction of a low melting temperature (Tm) PC into the outer leafl et. As shown in Fig. 2 , relative to exchange vesicles, there was no decrease in outer leafl et order for scrambled vesicles containing di18:3PC and di20:4PC. This indicates that the exchange vesicles prepared from these lipids are not asymmetric, in agreement with the estimates above showing that there are vesicles could appear to be as ordered as asymmetric vesicles. However, in our studies, relatively small changes in order suggest incomplete asymmetry in the exchange vesicles, as it is in agreement with conclusions based upon SM content in the inner and outer leafl ets.
The second assay used to evaluate the difference between the properties of exchange and scrambled vesicles was the thermal stability of gel-state domains in the vesicles. The bilayer of SM-containing vesicles undergoes a readily detected phase transition from the ordered gel phase to the liquid-disordered phase as temperature is increased. The Tm can be defi ned as the midpoint of this phase transition ( 23,24 ). It was previously found that Tm values are higher in asymmetric SMo/PCi vesicles than symmetric vesicles of the same composition ( 17 ). This refl ects the fact that the nearly pure SM outer leafl ets in asymmetric vesicles melt at a temperature similar to that of SM in pure SM vesicles, whereas in symmetric vesicles, the mixture of SM with low-Tm PC in each leafl et results in a Tm value lower than that of pure SM ( 17,18 ). Table 2 shows Tm values of exchange vesicles both before and after the scrambling process, and the difference between Tm in exchange vesicles and scrambled vesicles prepared from them is shown in Fig. 3 . The results were virtually identical to those obtained from measurements of outer leafl et order in terms of the relationship between PC acyl chain structure and asymmetry at 23°C.

Relationship between PC acyl chain structure and lipid transverse diffusion
By defi nition, the ability to maintain membrane asymmetry, in which the inner and outer leafl ets have different lipid compositions, requires, in the absence of active lipid transport catalyzed by a protein, that the movement of lipids between inner and outer leafl ets is slow. Therefore, we determined whether the different extents of asymmetry observed could be explained by differences in lipid transverse diffusion (i.e., fl ip-fl op). The transverse diffusion assay used was based on the ability of sodium dithionite to reduce C 6 -NBD-PC into a nonfl uorescent derivative. C 6 -NBD-PC was incorporated into the outer leafl et of preformed vesicles, similar fractions of SM in the inner and outer leafl ets of these vesicles. In addition, there was only a very small decrease in order for scrambled vesicles versus exchange vesicles containing di14:1PC, suggesting vesicles with this lipid have very little asymmetry. For the other PCs studied, Fig. 2 shows there was a signifi cant decrease in order in scrambled vesicles relative to exchange vesicles, indicative of a signifi cant degree of lipid asymmetry in the exchange vesicles, which is in agreement with the conclusions based upon inner and outer leafl et SM content in these vesicles.
However, for vesicles containing 16:0-18:2PC, 16:0-20:4PC, and di18:2PC, the decrease in outer leafl et order in scrambled vesicles relative to exchange vesicles was not as large as for the other lipids. By itself, changes in overall membrane order are not conclusive, because if SM is very laterally immiscible with PC in symmetric vesicles, the scrambled

Before Scrambling
After Scrambling time for transverse diffusion because these conditions showed the most sensitivity to PC acyl chain structure (see supplemental materials for details). Fig. 4 shows the degree of protection of C 6 -NBD-PC fl uorescence in symmetric SUVs containing PCs with different acyl chain structures. Varying the PC structure gave an extent of transverse diffusion that closely paralleled the effect of PC structure on asymmetry in exchange vesicles. Transverse diffusion was low for vesicles containing any of the PCs that were able support full asymmetry, whereas transverse diffusion was high with vesicles containing di14:1PC, di18:3PC, and di20:4PC, which do not support asymmetry. The behavior of vesicles containing di18:2PC was surprising in that they seemed to allow only slow C 6 -NBD-PC transverse diffusion even though the di18:2PC cannot support the formation of highly asymmetric vesicles. Possible explanations of this are considered in the Discussion.
Alternate explanations of the acyl chain dependence of transverse diffusion were tested. One possible factor could be vesicle size/curvature. However, exchange vesicles are isolated from the same, fi xed fractions of a size exclusion chromatography column, so there should be no dependence of exchange vesicle size upon PC acyl chain structure. Nevertheless, transverse diffusion in LUV was compared with that in SUV to investigate the possibility that curvature strongly impacts lipid transverse diffusion (supplemental Fig. VII). The pattern in which lipids supported less or more C 6 -NBD-PC transverse diffusion was very similar in SUV and LUV, indicating that the effect of bilayer curvature is not great.
Another possibility is that lipid randomization during spontaneous vesicle fusion could cause loss of asymmetry. However, using dynamic light scattering, no vesicle fusion was detected over 24 h for di18:3PC or di14:1PC, which failed to form asymmetric vesicles (data not shown).

Asymmetric lipid distribution is associated with a difference between order in the inner and outer leafl et for SM outside/PC inside vesicles
Previous studies using asymmetric vesicles have shown that when the outer leafl et of a vesicle is SM in the gel state (100% ordered) and the inner leafl et is composed of DOPC or POPC, the inner leafl et remains largely in the Ld state ( 17,18 ). This means that the physical states of the inner and outer leafl ets are not strongly coupled to each other ( 17,18 ). To see how general this behavior is, we examined the degree of coupling between the inner and outer leafl et physical states for exchange vesicles with a wider variety of PC acyl chains. To do this, the degree of order in the inner leafl et was calculated from the fl uorescence anisotropy of DPH (which measures order in both leafl ets) and externally added TMADPH (which measures outer leafl et order) ( 18 ), as shown in Fig. 5 .
Supplemental Tables I and II contain the raw anisotropy  data for TMADPH and DPH, and supplemental Table III shows the calculated percentage order in the outer and inner leafl ets. Fig. 5 illustrates the difference between the level of order in the inner and outer leafl ets. In most cases, the percentage order in the outer leafl et was higher than that in the inner leafl et, as expected for asymmetric vesicles and translocation of C 6 -NBD-PC into the inner leaflet was detected by the amount of NBD fl uorescence that was protected from reduction by dithionite added to the external solution. If the extent of transverse diffusion is slow, there should have been very little protection, but if transverse diffusion is fast, there should have been signifi cant protection from dithionite (supplemental Figs. III and IV).
To identify the best conditions in which to carry out this assay, the effects of varying time over which transverse diffusion could occur and varying lipid composition were measured (supplemental Figs. V and VI). Transverse diffusion was both time and composition dependent, and ultimately, mixtures (1:1 mol:mol) of the PCs of interest with DOPC or SM were used together with a 24 h incubation DPH, and was also observed previously ( 17,18 ).] We cannot make any statement about coupling in these vesicles.

The acyl chain dependence of asymmetric vesicle formation is explained by lipid transverse diffusion (fl ip-fl op)
This article shows that asymmetric vesicles can be prepared from lipids having a wide range of acyl chain structures. This ability should be useful in future studies of the effect of acyl chain structure upon the properties and function of asymmetric bilayers. However, the ability to form asymmetric bilayers was not a universal property of lipids. A question that arises from this is why asymmetric vesicles cannot be formed specifi cally from lipids with two short acyl chains or two polyunsaturated acyl chains. Because maintenance of asymmetry requires, by defi nition, that lipids do not equilibrate between inner and outer leafl ets by transverse diffusion across the bilayer, the logical implication is that rapid transverse diffusion for lipids with two short or polyunsaturated acyl chains explains their inability to form asymmetric vesicles. This was confi rmed by the measurements of the extent of transverse diffusion for a probe lipid in vesicles with various lipid compositions. The lipid mixtures that allowed most rapid transverse diffusion of the probe lipid were the same ones in which lipid asymmetry was not present, which means that the lipid compositions were such that they increased the rate of transverse diffusion for both labeled and unlabeled lipids.
In the case of almost all of the vesicles that were fully or mostly asymmetric (inside leafl et composed of di16:1PC, di18:1PC, di20:1PC, di22:1PC, 16:0-18:2PC, and 16:0-20:4PC), the difference was in the range 50-100%. Because the outer leafl et was almost fully ordered ( ‫ف‬ 90%) in these vesicles (supplemental Table III), this means that the inner leafl et was largely disordered (supplemental Table  III) and thus that the physical states of the inner and outer leafl ets are largely uncoupled. (The experimental error in anisotropy values makes it diffi cult to make a precise statement concerning low levels of coupling.) The behavior of diphytanoyl PC was surprising because it is in vesicles that are fully asymmetric, but there did not appear to be any difference in the order of the inner and outer leafl et. This means either that the inner and outer leafl et physical states are coupled in SMo/diphyPCi vesicles or that DPH does not partition well into the inner leafl et composed of diphytanoyl PC. We suspect the latter possibility is more likely because the methyl groups of diphytanoyl PC would interfere with van der Waals interactions between lipid acyl chains and the fl at DPH molecule.
In the cases of di14:1PC and lipids with two polyunsaturated chains, the percentage ordered in the outer leafl et was at most modestly higher than in the inner leafl et. This was expected, because it refl ects the lack of asymmetry in these vesicles. In the absence of asymmetry, the percentage ordered should be almost equal in the inner and outer leafl ets. [The observation that the calculated value of order is higher for the inner leafl et than the outer leafl et in some of these cases is an artifact of fact that TMADPH does not partition into ordered domains as well as does creases. This could increase the spaces in which a molecule crossing the bilayer might dissolve. Another factor arises from the fact that the barrier to transverse diffusion is very likely to be the movement of the polar headgroup of a lipid across the hydrophobic core of the bilayer. In the case of shorter acyl chains, the width of this hydrophobic barrier is small, which should allow more rapid transverse diffusion. In the case of lipids with polyunsaturated acyl chains, the lesser hydrophobicity of the bilayer core, due to the presence of multiple carbon-carbon double bonds ( 26 ), may be a property that allows faster transverse diffusion. In the case of polyunsaturated lipids, the issue of acyl chain oxidation must also be considered. This is important because it has been shown that oxidized lipids exhibit fast transverse diffusion across bilayers ( 27 ). Indeed, when a preparation of polyunsaturated lipid that had been stored for an extended period of time was used, we noted very fast transverse diffusion, suggesting oxidation could have been involved (data not shown). For this reason, we took precautions against oxidation of the lipids with polyunsaturated acyl chains, and no or very little acyl chain oxidation was detected (see Methods). Nevertheless, we cannot rule out a small degree of oxidation.
We also observed that mixtures of lipids with slow and fast transverse diffusion tended to show an intermediate extent of transverse diffusion, although generally they did not show a linear relationship between lipid composition and extent of transverse diffusion (supplemental Fig. VI). This suggests that transverse diffusion during the lipid exchange process might follow a complex course of events. At the outset of the lipid exchange process, the acceptor vesicles are composed only of the intended inner leafl et lipid, so SM molecules would be expected to fl ip across the bilayer with a rate refl ecting that of the PC present. However, toward the end of the SM exchange process, the membrane composition will be very different, and it is likely that transverse diffusion rates will have changed. This may explain why the extent of transverse diffusion observed using symmetric vesicles did not always exactly parallel the loss of asymmetry observed in the exchange vesicles. Another factor could be a difference between the transverse diffusion of the labeled C 6 -NBD-PC molecules and that of the unlabeled lipids.
It should be noted that previous studies of the acyl chain dependence of lipid transverse diffusion (in symmetric vesicles) are consistent with our results. In the study of Armstrong et al., it was found that increased unsaturation increased transverse diffusion rates for a headgroup labeled NBD-lipid ( 25 ). The study of Volinsky et al. found that oxidized PC transverse diffusion was much more rapid than that of unoxidized PC ( 27 ).
Fast lipid transverse diffusion also helps explain the tendency to have the highest extent of SM exchange in cases in which the lipid composition did not result in asymmetric vesicle formation. In vesicles that have low transverse diffusion, once the outer leafl et is nearly 100% SM, there can be no further increase in SM content. However, if transverse diffusion is fast, then SM introduced into the outer leafl et could fl ip into the inner leafl et, allowing exchange to continue, so that both leafl ets would end up with a high SM concentration.
A fundamental question that arises from these observations is why lipids with certain acyl chains impart rapid transverse diffusion to all of the lipids in a membrane. Armstrong et al. proposed that lipid packing was the important parameter determining transverse diffusion rates ( 25 ). Packing should become looser as the number of double bonds in-