All vertebrates synthesize cholesterol (1), which is found predominantly in the plasma membranes of cells (2). Cholesterol has at least two classes of important functions. First, it controls metabolic processes and is a specific precursor of D-vitamins, sex hormones, and bile acids (1). However, the amount of cholesterol in animal tissues is much higher than needed in these roles. Another function of cholesterol may be to alter the properties of lipids in membranes (1). Above the lipid melting temperature, the presence of cholesterol orders lipid acyl chains. Below the melting temperature, cholesterol prevents lipids from condensing into solid phases. Furthermore, the addition of cholesterol to lipid mixtures promotes the formation of two coexisting liquid phases, a familiar “liquid crystalline” (L_α) phase and a “liquid-ordered” (L_o) phase (3,4). The liquid-ordered phase has been distinguished from a solid phase by faster translational diffusion of the lipids (5) and greater mobility of the acyl chains (6). The separation of lipids into two immiscible liquid phases is often linked to the formation of “rafts” in cell membranes (7). These raft domains are characterized as rich in cholesterol and saturated lipids, and they have been implicated in a host of important cell functions such as membrane trafficking and signaling (8).

Although it is known that lipids and sterols interact in membranes through hydrogen bonding and van der Waals forces, little is known about how these interactions give rise to immiscible liquid phases in bilayers (7). To directly probe how sterol structure affects the miscibility phase behavior of lipids, we incorporated the sterols in Fig. 1 into giant unilamellar vesicles (GUVs) and utilized fluorescence microscopy to observe micron-scale domains of coexisting phases.

FIGURE 1 Sterol structures drawn to highlight assignment of hydrogen, hydroxy, and methyl groups to either the α-face (bottom) or β-face (top) of the sterol. Distortion of the rings due to double bonds is not shown for the sterols ergosterol and (more ...)

We chose some of the sterols because they are naturally related to cholesterol. For example, lanosterol is a precursor of both cholesterol, which is found primarily in animal tissue, and ergosterol, which is found primarily in yeast and fungi (1). The sterol 25-hydroxycholesterol is an inhibitor of HMG-CoA reductase and in turn prevents the synthesis of cholesterol (1). Cholesterol oxidase catalyzes the conversion of cholesterol to cholestenone (9). Finally, as its name suggests, coprostanol appears in feces and arises due to the conversion of cholesterol by bacteria (10).

In our experiments, we use fluorescence microscopy to directly observe micron-scale domains in membranes containing equimolar ratios of the saturated lipid DPPC (dipalmitoyl phosphatidylcholine), the unsaturated lipid DOPC (dioleoyl phosphatidylcholine), and a sterol. We classify sterols as promoters if vesicles contain immiscible liquid phases as temperature is lowered. Vesicles made with inhibitor sterols instead form coexisting solid and liquid phases as temperature is lowered. We measure the transition temperature (miscibility or melting) of vesicles containing different sterols and determine the area fraction of the more ordered phase. Our findings are consistent with previous work that classified the sterols used as either promoters or inhibitors of ordered membrane domains using fluorescence quenching and detergent resistance (4,11).

GUVs of 1:1:1 DPPC/DOPC/sterol and 0.8 mol % TR-DPPE were prepared by electroformation at 60°C in deionized water (18 MΩ-cm, Barnstead, Dubuque, IA) as described previously (16,17). The resultant vesicles were transferred between glass coverslips for examination by fluorescence microscopy on a temperature controlled microscope stage. The vesicles were free floating and not attached either to the glass slide or to other vesicles. The phase transition in a single vesicle is identified as the temperature at which domains of at least 1 μm appear (or disappear) in the membrane as temperature is lowered (or raised). Temperature was ramped at 0.2°C/s and the sample equilibrated for 30 s before viewing. The lowest temperature probed was 10°C. All transitions are fully reversible. Errors bars reported in transition temperature represent the full range of temperatures over which many vesicles pass through the transition in a single preparation of GUVs. The measured range of transition temperatures is reproducible when vesicles are prepared with the same initial lipid composition and experimental conditions (e.g., growth temperature). The spread in transition temperature between vesicles is due to variation in lipid composition and the location of that composition with respect to the nearest phase boundary. Previously we have noted that there is increased spread in measured transition temperatures near steep regions of the phase diagram, where small changes in lipid composition can lead to large changes in transition temperature (18). The midpoint of this range (black points in Fig. 2) roughly corresponds to the temperature at which 50% of vesicles contain coexisting phases. The spread in transition temperature is unusually large for membranes containing the sterol 25-OH-cholesterol (ΔT = 22°C). This sample contains some vesicles with a transition close to the growth temperature. Previously we have observed a large distribution in transition temperatures for vesicles of 1:1 DOPC/DPPC + 35% cholesterol when vesicles are grown at or below their miscibility transition temperature (19,25).

FIGURE 2 Transition temperatures of vesicles produced from 1:1:1 DOPC/DPPC/sterol with 0.8 mol % TR-DPPE. As temperature is decreased, a transition is observed from one uniform phase to either (middle) two coexisting liquid phases for vesicles containing promoter (more ...)

TABLE 1 Summary of data for vesicles produced from 1:1:1 DOPC/DPPC/sterol with 0.8 mol % TR-DPPE

Our characterization of liquid versus solid phases for the particular case of vesicles containing 25-hydroxycholesterol is based on lipid mobility and is consistent with previous measurements of lipid diffusion. Lipid diffusion constants (D) have been measured in liquid states (both L_o and L_α) by a variety of experimental methods to be ~1–20 μm²/s (5,21–23). Therefore, the time required for a lipid to traverse a micron across the membrane's surface ranges between seconds and hundredths of seconds. In contrast, lipids in S_o states diffuse significantly slower (D ~ 10⁻² − 10⁻⁸ μm²/s) (5,22,24), leading to much longer timescales for μm-scale lipid motion, from hundreds of seconds to hundreds of days. Interestingly, binary membranes of saturated lipids and cholesterol with S_o-L_o coexistence also show slow diffusion (D ~ 1–5 × 10⁻² μm²/s) (5,23), and large-scale phase separation is not detected by fluorescence microscopy (16). For this reason, we are unable to distinguish pure S_o domains from domains containing both S_o and L_o phases in this study, although it is a focus of current work in our laboratory. We have discussed S_o-L_o coexistence and evidence for three phase S_o-L_o-L_α coexistence in greater detail in the context of mixtures containing sphingomyelin lipids and cholesterol (19,25).

To study the influence of different sterols on lipid phase behavior, we produced free-floating GUVs containing equimolar ratios of DOPC, DPPC, and each of the sterols shown in Fig. 1. The lipids DOPC and DPPC were chosen because they are readily available as pure, single components, and we have studied the ternary mixture of DOPC/DPPC/cholesterol extensively (16,20).

FIGURE 3 Fluorescence micrographs of vesicles produced from 1:1:1 DOPC/DPPC/sterol with 0.8 mol % TR-DPPE at 23°C. Vesicles contain (a) 25-hydroxycholesterol, (b) cholesterol, (c) dihydrocholesterol, (d) epicholesterol, (e) ergosterol, (f) no sterol (control), (more ...)

In the absence of cholesterol, vesicles of DOPC and DPPC contain coexisting solid and liquid phases below the melting transition temperature (Fig. 3 f) (16,26). A dark domain is clearly solid rather than liquid if it has a static, noncircular boundary. The dark solid phase is rich in saturated lipid, and the bright liquid phase is rich in unsaturated lipid and the fluorescent dye (16).

Vesicles of DOPC, DPPC, and the various sterols in Fig. 1 exhibit either two coexisting liquid phases or coexisting solid and liquid phases at low temperatures. Above the transition temperatures shown in Fig. 2, vesicles are in one uniform liquid phase. Vesicles containing the sterols 25-hydroxycholesterol, cholesterol, dihydrocholesterol, epicholesterol, or ergosterol produce coexisting liquid phases below a miscibility transition (Fig. 3, a–e). We classify these sterols as promoters because they promote the formation of a second liquid phase as cholesterol does. In contrast, we observe a transition to coexisting solid and liquid phases in vesicles containing the sterols androstenolone, cholestane, cholestenone, or coprostanol (Fig. 3, g–j). We classify these sterols as inhibitors because they do not promote the formation of a second liquid phase. The remaining sterol, lanosterol, does not produce coexisting liquid or coexisting solid and liquid phases over a wide range of temperatures (5°C–60°C) and lipid compositions (DOPC/DPPC/lanosterol with ratios 33/33/33, 35/35/30, 37/38/25, 22/45/33, and 24/49/27). Coexisting liquid domains can be induced in membranes containing lanosterol by exposure to light. Previously we noted that light can induce coexisting liquid domains in vesicles containing DOPC, DPPC, and cholesterol near the miscibility transition boundary (16).

We find no distinct correlation between transition temperatures and the phases produced, solid domains for the inhibitor sterols, and coexisting liquid phases for the promoters (Fig. 2). When a solid phase is formed, transition temperatures are lower than for pure DPPC (41°C), as expected from a colligative melting point depression due to the addition of DOPC and sterol to the membrane. Liquid miscibility transition temperatures are not constrained to be below the melting point of pure DPPC, although they generally fall below 41°C. Vesicles that exhibit liquid miscibility transitions may also have melting transitions to solid phases at lower temperatures, including coexistence of three phases containing a solid with two distinct liquids (19,25). We find no direct evidence for a gel phase in vesicles containing promoter sterols down to 23°C, except perhaps in membranes containing 25-hydroxycholesterol, in which the coalescence of two dark, circular domains at 23°C requires ~1 min (see partially coalesced domains in Fig. 3 a and the Materials and Methods section). It is useful to relate these results to membranes containing cholesterol in which slow domain coalescence is observed in regions of the phase diagram where there is evidence for coexisting S_o, L_o, and L_α phases (S. L. Veatch and S. L. Keller, unpublished results, and (25)).

Based on our previous results with membranes containing cholesterol (16), we conclude that the lipids in the dark domains in Fig. 3 are more ordered and contain more DPPC, whether those domains are solid or liquid. Area fractions of dark phases for vesicles containing each sterol are shown in Fig. 4. For the most part, vesicles containing solid domains with inhibitor sterols have smaller areas of dark phase than liquid domains with promoter sterols. This difference in area fraction is consistent with a smaller molecular area in solid phases than in liquid phases. In addition, solid phases should contain little sterol (27), whereas ordered liquid phases can incorporate significant fractions of sterol (20). By extension, inhibitor sterols are expected to have low solubility in DPPC below its melting temperature. The fraction of solid phase in vesicles with inhibitor sterols in Fig. 4 follows the trend in melting temperature in Fig. 2. This is simply explained by an increase in solid phase as temperature is lowered further below the melting temperature. Similarly, as temperature decreases from 29°C to 23°C, the fraction of dark area in vesicles either remains constant or increases (Fig. 4). This is consistent with our previous observations in vesicles of DOPC/DPPC/cholesterol (20).

FIGURE 4 Area fraction of dark phase of vesicles produced from 1:1:1 DOPC/DPPC/sterol with 0.8 mol % TR-DPPE. In general, vesicles containing liquid domains have larger area fractions of dark domains. As temperature decreases, the fraction of dark area either (more ...)

We would like to emphasize that we measure transition temperatures and dark phase surface fractions at a single point in the ternary phase diagram for each sterol system and do not attempt to map entire miscibility phase boundaries. It is important not to draw too much significance from a single data point. Although higher transition temperatures could be due to an increased interaction energy between the sterol and saturated lipid DPPC, it could also indicate that the phase boundary is shifted to higher sterol concentrations or that <33% sterol partitions into the membrane. Similarly, a wide spread in transition temperatures could be due to the close proximity of a steep region of the phase diagram. A wide spread may also indicate that the sterol produces nonuniform films during preparation or that vesicles of that sterol should be grown at alternate temperatures (19). Clearly, studies over a range of lipid compositions are necessary to make direct comparisons between individual sterols. Our results provide a first step toward those broad studies.

We also investigated the effect of gradually substituting cholestenone for cholesterol in vesicles (Fig. 5). The activity of cholesterol oxidase to produce cholestenone has been used as a measure of lipid phase and stability of cholesterol in the membrane (9,28). We find that the addition of cholestenone has two effects on vesicles of DOPC, DPPC, and cholesterol. First, at low cholestenone concentrations (0–50% of total sterol), the miscibility transition temperature is depressed, although coexisting liquid phases are still observed below the transition. When cholestenone is 50% of total sterol, the liquid phase boundaries fluctuate, implying that line tension is small and that the transition temperature, as well as a critical point, are probably nearby. Second, at higher cholestenone fractions (>50–100% of total sterol), a solid phase is observed. Our results indicate that a gradual conversion of cholesterol to cholestenone by cholesterol oxidase can change the phase state of the membrane. At higher temperatures, two coexisting liquid phases become one uniform liquid when cholesterol converts to cholestenone. At lower temperatures, a solid phase can appear when cholesterol converts to cholestenone.

FIGURE 5 Transition temperatures (top) and vesicle micrographs (bottom) for vesicles of 1:1:(1 − x):x DOPC/DPPC/cholesterol/cholestenone with 0.8 mol % TR-DPPE. Coexisting liquid phases are observed below the miscibility transition temperature in membranes (more ...)

Wenz and Barrantes have suggested that a sterol's ability to promote DPPC-rich phases in membranes may be correlated with higher sterol hydrophobicity. They compiled and calculated values for the log of the partition coefficient (LogP) of sterols between octanol and water (29). We find no clear correlation between this measure of hydrophobicity and either our own measurements of phase behavior and dark area fraction or published values of fluorescence quenching (Table 1).

Bloch lists several attributes required for cholesterol's functional competence in membranes, among which are 1), a flat, rigid fused ring system, 2), the absence of methyl groups on the α-face, 3), an equatorial 3-hydroxy group, 4), bridgehead methyl groups at the C10 and C18 positions, and 5), an unmodified branched, isooctyl chain (1). Using our results above, we evaluate these criteria with respect to lipid miscibility. If part of cholesterol's “functional competence” involves the production of coexisting liquid phases, then we expect all sterols that follow Bloch's criteria to be promoters.

For criterion 1, all of our sterols possess rigid fused rings. However, coprostanol does not fulfill the requirement of flatness. Comparing coprostanol to dihydrocholesterol, the two sterols differ only in the linkage of the first two rings. Coprostanol is curved due to a cis fusion of the A and B rings (31), which likely prevents it from packing closely with neighboring phospholipids and explains why it is an inhibitor. Vesicles containing coprostanol exhibit coexisting solid and liquid phases, whereas vesicles containing dihydrocholesterol exhibit coexisting liquid phases. For criterion 2, lanosterol is our only sterol with methyl groups on the α-face, and it does not consistently produce coexisting liquid domains. Hence, within our set of sterols, criterion 2 is valid. The enhanced planarity of cholesterol with respect to lanosterol may explain its ability to alter the ordering and phase behavior of lipids (32,33). For criterion 3, three of our sterols do not possess equatorial hydroxyl groups at the 3 position: cholestane has a hydrogen, cholestenone has a ketone group, and epicholesterol has an axial hydroxyl group. Cholestane and cholestenone are inhibitors, yet the promoter epicholesterol behaves like cholesterol in its ability to produce coexisting liquid domains in vesicles, in its fraction of dark area, and in its miscibility transition temperature. Given our set of sterols, the criterion for a hydroxyl group is valid, but the criterion for the group to be equatorial is not important for miscibility phase behavior. For criterion 4, only androstenolone is missing the C10 methyl group, and it is an inhibitor sterol. However, androstenolone is not a sufficient test of criterion 4 since it also violates criterion 5. For criterion 5, several sterols do not possess unmodified isooctyl chains, including the promoter sterols of ergosterol and 25-hydroxycholesterol. Hence, the criterion for isooctyl chains is not sufficient to explain miscibility phase behavior. Overall, we conclude that Bloch's requirements for functionally competent sterols are more strict than necessary to predict the formation of liquid phases, particularly for the promoter sterol epicholesterol (without an equatorial hydroxyl group) and the promoter sterols ergosterol and 25-hydroxycholesterol (with modified isooctyl chains).

Barenholz presents less stringent requirements for “membrane active sterols”, which usually predict our observed formation of liquid phases. A membrane active sterol is defined as affecting short- and long-range lipid order within membranes, minimizing volume, and decreasing membrane permeability. Specifically, the sterol should possess 1), a flat, fused ring system, 2), a hydroxyl or other small polar group at position 3, 3), a “cholesterol-like” tail, and 4), a small area per molecule (<40 Ų when assembled at the air/water interface at a surface pressure of 12 mN/m) (30). As an example of applying rule 4, cholestenone is an inhibitor. Although it satisfies rules 1–3, it violates rule 4 since it has an area per molecule of 52 Ų in a monolayer at 12 mN/m (34). The rules above correctly predict the observed formation of liquid phases except in the case of 25-hydroxycholesterol, which has an alkyl tail that terminates in a hydroxyl group that is not “cholesterol-like”. We do not know how 25-hydroxycholesterol orients in the membrane or exactly why its miscibility transition temperature is high with a large spread in values (see Materials and Methods section).

Other studies have replaced the hydroxyl headgroup of cholesterol with a ketone, an acetate, or an ether and found that the compressibility of DPPC/sterol monolayers is lowest for the hydroxy-sterols (35). Similarly, replacing the hydroxyl headgroup of cholesterol with a ketone to produce 6-ketosterol renders the sterol less effective at reducing monolayer lateral elasticity (36). Returning to our own results on hydroxysterols versus ketosterols, we find that cholestenone is an inhibitor. The structural reason that cholestenone is an inhibitor is likely because its ketone headgroup lacks the ability of a hydroxyl group to directly hydrogen bond with the ester linkage of phosphatidylcholine lipid acyl chains (37).

With respect to the alkyl-chain structure of sterols, lipid vesicles are more permeable when they contain sterols with alkyl chains that are longer and more unsaturated than the alkyl chain of cholesterol (38). In a separate experiment, a group of cholesterol analogs with branched and unbranched chains containing 3–10 carbons all exhibited roughly equal area condensation in monolayers of DPPC or 18:0-18:1PC (28). Sterols with unbranched alkyl chains had smaller molecular areas and smaller collapse pressures than those with branched chains (28). Although unbranched cholesterol analogs fit the requirement of a smaller molecular area for a “membrane active sterol” (30), it is unclear if an unbranched tail qualifies as “cholesterol-like”. In any case, it is clear that membrane active sterols require an alkyl chain. Androstenol (5-androsten-3β-ol) is similar to androstenolone, lacks an alkyl chain, has no significant condensing effect on monolayers (28), and has a much smaller effect on monolayer lateral elasticity than cholesterol (36).

Applying this information to our results, it is difficult to find correlations that consistently apply between the membrane phases we observe and previous measurements of membrane order and permeability. We find that membrane phase often, but not always, correlates with the general magnitude of area condensation of sterols with oleoyl-stearoyl PC (34), the solubilization of sterols in sonicated eggPC extracts, the decrease in vesicle permeability (10), or the equilibrium constants for lipid dimer exchange (31).

A similar conclusion applies to monolayers. A variety of cholesterol analogs condense lipid monolayers (DPPC or stearoyl-oleoylPC in this case) (28), but not all produce immiscible liquid phases in lipid monolayers (DPPC or palmitoyl-sphingomyelin in this case) (35,39). As another example, sterol molecular area in monolayers predicts the temperature of the inverted hexagonal phase transition in bilayers of lipids and many sterols, with the striking exception of androstenol (40).

From the experimental side, the effects of lanosterol on lipid membranes are typically less pronounced than of cholesterol and ergosterol. For example, cholesterol reduces vesicle permeability more than lanosterol (42). Similarly, the microviscosity of eggPC/cholesterol membranes is greater than of eggPC/lanosterol membranes as measured by fluorescence microscopy (43). Nuclear magnetic resonance finds that sterols produce conformational ordering of saturated lipid chains of dimyristoylPC and PPetPC in the hierarchy ergosterol > cholesterol > lanosterol. For the mixed saturated and unsaturated chains of palmitoyl-oleoylPC, ordering occurs as cholesterol > ergosterol ~ lanosterol (33,44). For the unsaturated chains of DOPC, ordering occurs as cholesterol > lanosterol (45). Cholesterol and ergosterol also produce more membrane order than lanosterol as measured by a spin probe incorporated in vesicles of bovine brain lipids (46). However, smaller concentrations of lanosterol (15%) than cholesterol (25%) are required to diminish the endothermic melting peak of PPetPC (33).

In summary, we find coexisting liquid phases in bilayer lipid membranes containing those sterols previously characterized as promoting the formation of DPPC-rich phases. Although detailed phase diagrams are required to completely compare the effectiveness of each promoter sterol in producing liquid phases, it seems likely that the following structural features that fulfill requirements of “membrane active sterols” play an important role: 1), flat, fused rings, 2), a hydroxyl or other small polar group at position 3, 3), a “cholesterol-like” tail, and 4), a small area per molecule (30).

M.E.B. was partially supported by a National Science Foundation (DGE-0086280) GK-12 Fellowship. Research in the Keller laboratory is supported by a National Science Foundation Career Award (MCB-0133484), a Research Corporation Innovation Award, and a Cottrell Scholar Award.