Cholesterol Precursors Stabilize Ordinary and Ceramide-rich Ordered Lipid Domains (Lipid Rafts) to Different Degrees

Genetic disorders of cholesterol biosynthesis result in accumulation of cholesterol precursors and cause severe disease. We examined whether cholesterol precursors alter the stability and properties of ordered lipid domains (rafts). Tempo quenching of a raft-binding fluorophore was used to measure raft stability in vesicles containing sterol, dioleoylphosphatidylcholine, and one of the following ordered domain-forming lipids/lipid mixtures: dipalmitoylphosphatidylcholine (DPPC), sphingomyelin (SM), a SM/cerebroside mixture or a SM/ceramide (cer) mixture. Relative to cholesterol, early cholesterol precursors containing an 8-9 double bond (lanosterol, dihydrolanosterol, zymosterol, and zymostenol) only weakly stabilized raft formation by SM or DPPC. Desmosterol, a late precursor containing the same 5-6 double bond as cholesterol, but with an additional 24-25 double bond, also stabilized domain formation weakly. In contrast, two late precursors containing 7-8 double bonds (lathosterol and 7-dehydrocholesterol) were better raft stabilizers than cholesterol. For vesicles containing SM/cerebroside and SM/cer mixtures the effect of precursor upon raft stability was small, although the relative effects of different precursors were the same. Using both detergent resistance and a novel assay involving fluorescence quenching induced by certain sterols we found cholesterol precursors were displaced from cer-rich rafts, and could displace cer from rafts. Precursor displacement by cer was inversely correlated to precursor raft-stabilizing abilities, whereas precursor displacement of cer was greatest for the most highly raft-stabilizing precursors. These observations support the hypothesis that sterols and cer compete for raft-association (Megha, and London, E. (2004) J. Biol. Chem. 279, 9997-10004). The results of this study have important implications for how precursors might alter raft structure and function in cells, and for the Bloch hypothesis, which postulates that sterol properties are gradually optimized for function along the biosynthetic pathway.

Genetic disorders of cholesterol biosynthesis result in accumulation of cholesterol precursors and cause severe disease. We examined whether cholesterol precursors alter the stability and properties of ordered lipid domains (rafts). Tempo quenching of a raft-binding fluorophore was used to measure raft stability in vesicles containing sterol, dioleoylphosphatidylcholine, and one of the following ordered domain-forming lipids/lipid mixtures: dipalmitoylphosphatidylcholine (DPPC), sphingomyelin (SM), a SM/cerebroside mixture or a SM/ceramide (cer) mixture. Relative to cholesterol, early cholesterol precursors containing an 8 -9 double bond (lanosterol, dihydrolanosterol, zymosterol, and zymostenol) only weakly stabilized raft formation by SM or DPPC. Desmosterol, a late precursor containing the same 5-6 double bond as cholesterol, but with an additional 24 -25 double bond, also stabilized domain formation weakly. In contrast, two late precursors containing 7-8 double bonds (lathosterol and 7-dehydrocholesterol) were better raft stabilizers than cholesterol. For vesicles containing SM/cerebroside and SM/cer mixtures the effect of precursor upon raft stability was small, although the relative effects of different precursors were the same. Using both detergent resistance and a novel assay involving fluorescence quenching induced by certain sterols we found cholesterol precursors were displaced from cer-rich rafts, and could displace cer from rafts. Precursor displacement by cer was inversely correlated to precursor raft-stabilizing abilities, whereas precursor displacement of cer was greatest for the most highly raft-stabilizing precursors. These observations support the hypothesis that sterols and cer compete for raft-association ( De novo cholesterol biosynthesis proceeds through various intermediates. Conversion of lanosterol to cholesterol involves a series of demethylation, double bond isomerization, dehydro-genation, and reduction steps (Fig. 1). In genetic disorders of cholesterol biosynthesis mutations in genes coding for enzymes of the biosynthetic pathway lead to a loss or reduction in enzymatic function, and an accumulation of sterol precursors. Mutations in the 7-dehydrocholesterol (7-DHC) 2 reductase gene, which catalyzes the conversion of 7-DHC to cholesterol, result in 7-DHC accumulation and Smith-Lemli-Optiz syndrome (SLOS) (1). In desmosterolosis, a mutation in 24-dehydrocholesterol reductase gene results in accumulation of desmosterol (2). Lathosterolosis is caused by a mutation in the lathosterol 5-desaturase gene, which results in the accumulation of lathosterol (3). In addition, mutations in 3␤-hydroxysteroid ⌬8 sterol isomerase, terminating synthesis at zymosterol, have been implicated in CHILD syndrome and the disorder CDPX2 (4). All of these genetic disorders manifest themselves as broad malformation diseases with overlapping symptoms that include features such as microcephaly, skeletal abnormalities, and both overall growth and mental retardation (5,6).
Participation in the formation and function of ordered lipid domains (lipid rafts) is believed to be an important function of sterols in eukaryotic membranes. Lipid rafts are usually defined as sphingolipid and sterol-rich domains that exist in the liquidordered phase (L o ). In eukaryotic cell membranes they are thought to co-exist with liquid disordered (L d ) domains that are rich in lipids with unsaturated acyl chains. The L o phase is an intermediate state with tightly packed lipids, like the solid-like gel phase, and a high lipid lateral diffusion rate, like that found in the L d phase (7,8). Rafts have been proposed to be important for many cellular processes (9 -12).
Recent studies have revealed that ceramide(cer)-rich rafts can also form. We found that cer is a strong promoter of lipid raft formation (13), and cer has a high affinity for rafts (14,15).
Cer-rich rafts appear to form large platforms when large amounts of cer are generated in plasma membranes in vivo by SMase action (16). Cer-rich rafts have been reported to be involved in the initiation of certain types of apoptosis (17)(18)(19), in bacterial infection by Neisseria gonorrhoeae (20) and Pseudo-monas aeruginosa (21), and in rhinovirus infection (22). Recently, we found that cholesterol is displaced from cer-rich rafts (14). Displacement of cholesterol has now been confirmed by other groups, both in model membranes and cells (23,24).
Because of the importance of cholesterol biosynthesis disorders and lipid rafts we examined the raft-stabilizing properties of combinations of a wide variety of cholesterol precursors and sphingolipids. A novel fluorescence assay (25) was employed to assess raft formation and stability in model membranes. This assay monitors the quenching of fluorescent probes bound to lipid bilayers by tempo, a small nitroxide-carrying quencher, which binds strongly to bilayers that are in the L d state (26). We find that there is a difference in the raft-stabilizing properties of different precursors, but stabilization does not improve gradually along the biosynthetic pathway. Instead, there is a distinct difference between raft stabilization by the early precursors and certain later precursors. We also show that substitution of precursors for cholesterol affects the sterol content of cer-rich rafts. These results have important implications for how sterol composition could affect raft structure and function in cells.
Sterol purity was confirmed by comparison of sterol melting temperatures to those reported by the manufacturers, and by TLC on HP-TLC plates (Merck & Co). Zymosterol and some desmosterol preparations contained impurities and were repurified by TLC. Approximately 1 mg of sterol was dissolved in ethanol, applied to an HP-TLC plate, and then chromatographed using a sequential solvent system. Solvent chambers were equilibrated with solvents for at least 2 h before chromatography. The first solvent (50:38:3:2 (v:v), chloroform/methanol/acetic acid/water) was allowed to migrate halfway up the plate. The plate was then dried, introduced into a second chamber containing the solvent system 1:1 hexane/ethyl acetate (v:v), and chromatographed until the solvent migrated to near the top of the plate. The plate was then dried and sprayed with 5% (w/v) cupric acetate dissolved in 8% (v/v) phosphoric acid in water. To detect sterol, plates were charred at 180°C for 5 min. To purify sterol, the region of the plate containing the pure sterol (identified from a fragment of the plate in which sterol was charred) was scraped off, dissolved in 1:1 (v:v) chloroform/ methanol, centrifuged to remove the silica particles, filtered on 0.22nitrocellulose to remove residual silica, and then dried under nitrogen flow. Purity was confirmed as described above. Commercial desmosterol preparations with no impurities gave experimental results similar to those obtained with purified desmosterol.
Vesicle Preparation-Ethanol dilution small unilamellar vesicles (SUV) and multilamellar vesicles (MLV) were prepared as described previously (14). Vesicles containing the desired lipid mixtures were dispersed at 70°C in 1 ml of phosphate-buffered saline (10 mM sodium phosphate, 150 mM NaCl, pH 7) at a final lipid concentration of 50 M (for SUV) and 500 M (for MLV). Fluorescence measurements were made after samples cooled to room temperature. For fluorescence measurements, vesicles also contained 0.1-0.2 mol % LcTMADPH or 0.5 mol % DPH. Background samples lacking fluorescent probe were also prepared.
Fluorescence Measurements-A scaled-up stock sample of SUV, prepared as described above, was divided into four 1-ml aliquots. Two samples containing quencher (F samples) were prepared by adding tempo to a concentration of 2 mM (from a 353 mM stock solution of tempo dissolved in ethanol). Two samples lacking quencher (F o samples) were prepared by adding a volume of pure ethanol equal to that added to the F samples. Both F and F o samples were then incubated for 10 min at room temperature, after which fluorescence was measured. LcTMADPH or DPH fluorescence was measured at an excitation wavelength of 357 nm and emission wavelength 429 nm. Fluorescence was measured on a Fluorolog 3 spectrofluorimeter using quartz semimicro cuvettes (excitation path length: 10 mm, emission: 4 mm). Unless otherwise noted, slit bandwidths were 3.2-nm excitation and 6.3-nm emission. Fluorescence intensity was measured as a function of increasing temperature as described previously (14). The ratio of fluorescence intensity in the presence of quencher to that in its absence (F/F o ) was calculated after subtraction of background values. Fluorescence anisotropy measurements were performed on SUV prepared as described previously, but using vesicles containing 0.5 mol % DPH or LcTMADPH (14). Anisotropy measurements were made at room temperature using a SPEX automated Glan-Thompson polarizer accessory with slit bandwidths set at 4.2 nm (excitation) and 8.4 nm (emission).
Peptide Quenching by Sterol-SUV containing 50 M total lipid and 2 mol % LW peptide were prepared by ethanol dilution as described above. Peptide fluorescence was measured at 280-nm excitation and 340-nm emission wavelengths. Excitation and emission bandwidths were both 6.3 nm. Fluorescence from background samples lacking peptide was subtracted from that of peptide-containing samples to determine peptide fluorescence. Quenching of peptide fluorescence by sterol was assessed by calculation of F/F o , where F samples contained 7-DHC, and F o samples contained cholesterol in place of 7-DHC.
Detergent Insolubility Assay-To MLV samples (500 nmol of total lipid dissolved in 950 ml of phosphate-buffered saline, pH 7) 50 l of 10% (w/v) Triton X-100/phosphate-buffered saline, pH 7, was added. The samples were incubated for 2 h at 23°C, and then detergent-insoluble lipid was isolated by centrifugation as described previously (14). The detergent-insoluble pellet, equivalent to the detergent-resistant membrane (DRM) fraction, was then resuspended in 100 l of water, transferred to a glass vial, and dried under high vacuum for 45 min. The dried pellet was dissolved in 50 l of 1:1 (v:v) chloroform/methanol. A 5-l aliquot of this was then applied to a HP-TLC plate. Separate spots containing various amounts of the lipids used (0.25-4 g) were also applied to the plate as standards. The sequential two-solvent system described above was used to separate lipids. Lipids were detected by charring as described above, and after the plate cooled, it was scanned as an image (CanonScan N124OU). Spot intensity was analyzed using the NIH/Scion Image program. Unknown lipid and sterol amounts were estimated by comparison to the standards, which generally exhibited a non-linear, but monotonic, increase in staining intensity as the amount of lipid applied was increased.
The percent of each lipid species in the insoluble fraction was computed from the formula: (amount of lipid species in the pellet in nmol/total amount of lipid in pellet in nmol) ϫ 100%.
The total amount of lipid (including sterol) in the insoluble fraction was calculated in nmol, and total % insoluble lipid was computed from the formula: (total amount of insoluble lipid in nmol/total amount of lipid in initial sample (500 nmol)) ϫ 100%.
The amount of insoluble cer was determined by measurement of radioactivity in the pelleted lipid fraction as previously described (14). Samples contained 0.25 Ci of [ 3 H]C16:0 cer.

How Raft Stability Is Assayed by
Fluorescence Quenching-A fluorescence quenching assay was used to determine whether ordered domains/rafts were present in vesicles composed of lipid mixtures containing cholesterol precursors. We developed an assay in which the quencher tempo was used (25) instead of the nitroxide-labeled lipids we used previously (13,25,27,28). Tempo is water soluble, but binds strongly to disordered domains (26). The fluorescent probe used was LcTMADPH, a derivative of DPH. LcTMADPH has a very high affinity for rafts, and is resistant to displacement from rafts by cer (14). To examine the thermal stability of rafts, two sets of vesicles containing LcTMADPH were prepared: vesicles that contain tempo (F samples) and vesicles that lack tempo (F o samples). In vesicles containing co-existing ordered and disordered domains LcTMADPH and tempo segregate, and F/F o is high. As temperature increases the rafts melt, and the bilayer becomes progressively more homogenous. This ultimately abolishes segregation of LcTMADPH and tempo, and F/F o decreases. For samples exhibiting a sigmoidal dependence of quenching upon temperature, melting temperature (T m ) can be defined: the more raft stabilizing the lipid mixture used, the higher the observed T m (13,27,29). T m values for rafts measured using tempo quenching were similar to those in analogous mixtures containing nitroxide-labeled lipids (13,27).
Cholesterol Precursors Stabilize Raft Formation to Varying Degrees-The effect of cholesterol precursor structure upon the thermal stability of rafts was assayed using quenching. The sterols studied are shown in the schematic of the cholesterol biosynthetic pathway (Fig. 1). The two missing precursors were not available. Notice that there are two "vertical" branches to the pathway, which only differ in terms of whether the 24 -25 double bond in the sterol "tail" is unsaturated (left branch) or reduced (right branch). In addition to sterols, the samples contained 1:1 mol:mol mixtures of a lipid that tend to form disor-   Table 1. T m values in the presence of precursors and cholesterol are higher than in the absence of sterol. This shows that all precursors increase the thermal stability of rafts, i.e. they promote raft formation at temperatures where rafts are not stable in the absence of sterol. Cholesterol precursors formed during earlier steps in the cholesterol biosynthetic pathway, lanosterol, dihydrolanosterol, zymosterol, and zymostenol, stabilize ordered domain formation to a significantly lesser degree than cholesterol.
Two precursors that occur late in the branch of the biosynthetic pathway lacking a 24 -25 double bond, lathosterol and 7-DHC, stabilize ordered domain formation to a greater degree than cholesterol. In contrast, desmosterol, which is the last precursor in the branch in which the 24 -25 double bond is present, and has a ring structure identical to that of cholesterol, stabilizes rafts to a much lesser degree than cholesterol. This is somewhat surprising, because T m values show that the 24 -25 double bond does not have a strongly destabilizing effect in the case of lanosterol and zymosterol.
Cholesterol Precursors Stabilize Raft Formation to Varying Degrees: Effect of Sphingolipid Structure-Because rafts in cells should contain sphingolipid, whether sterol structure would affect the stability of rafts was also studied in vesicles containing various sphingolipids. First, the stability of rafts was measured in mixtures of mammalian brain SM/DOPC/sterol 3:3:2 (mol: mol). Fig. 2B and Table 1 show that the relative ability of different precursors to stabilize ordered domain formation in these mixtures is very similar to that observed in analogous mixtures containing DPPC. Experiments were then performed in mixtures containing mixed mammalian cerebrosides in place of almost one-third of the SM. This is of interest because glycosphingolipids can form a considerable fraction of total cellular sphingolipid under some conditions (30). Our previous studies have shown that the presence of cerebrosides can lessen the effect of sterol on raft stability (13). In agreement with this observation, the thermal stability of rafts in 2.04:0.96:3:2 SM/cerebrosides/DOPC/sterol (mol:mol) is only modestly affected by the presence of sterol or precursor structure (Fig.  2C). The former is illustrated most clearly by a comparison of the difference between T m values in samples without sterol and those containing the most raft-stabilizing precursors. For the DPPC-containing samples this difference in T m values is 19°C, and for SM-containing samples the difference is 17°C, but for the samples containing SM plus cerebroside the difference is only 10°C. This behavior is consistent with the possibility that the cerebroside-rich rafts contain only low levels of sterol. If the rafts lack sterol, it could reduce the influence of sterol on their stability. Nevertheless, the overall pattern observed is similar to that in samples lacking cerebrosides, in that all sterols stabilize rafts to some degree, and the most stabilizing sterols are 7-DHC and lathosterol. One interesting difference in cerebroside-containing samples is the relatively strong stabilization of rafts by zymosterol.
Finally, the effect of sterol precursor structure was examined in mixtures of 2.04:0.96:3:2 SM/C16:0 cer/DOPC/sterol (i.e. 12 mol % cer). These mixtures were of interest because of reports that cer-rich rafts can be physiologically important (12). (Samples with 12 mol % cer were studied because at higher cer concentrations sterol displacement might be complete (14), and thus obscure differences between different cholesterol precursors.) As shown in Fig. 2D and Table 1, sterol presence and structure has only small effects on the stability of rafts in cercontaining samples (the difference between the T m of samples

TABLE 1 Relationship between sterol precursor structure and raft melting temperature (T m ) in model membranes containing various sphingolipids or dipalmitoylphosphatidylcholine
The data are derived from the samples used in the quenching experiments in Fig. 2. Samples contained mixtures of sphingolipid (or DPPC), DOPC and sterol in the proportions given in Figure 2. T m was defined as the point of maximum slope in the sigmoidal melting curves (see Fig. 2 without sterol and those containing the most raft-stabilizing sterol is 4°C). This behavior is consistent with the possibility that the cer-rich rafts also contain only low levels of sterol. Fig.  2D shows early cholesterol precursors actually slightly destabilize ordered domain formation in cer-containing vesicles relative to those without sterol. In contrast, some stabilization of rafts in cer-containing samples is observed with lathosterol and 7-DHC, consistent with the greater ability of these sterols to stabilize rafts in other lipid mixtures.
The observation that cer-rich rafts are destabilized when vesicles contained early precursors, but cerebroside-rich rafts are not, is consistent with a model in which cer, but not cerebrosides, are displaced from rafts by sterol (see below and "Discussion").
Sterol Precursors Are Displaced from Ceramide-rich Rafts to Different Degrees-In a previous study we found that moderateto-high cer levels displace cholesterol from lipid rafts (14). If this displacement is important for cer-rich raft function, and if cholesterol precursors differ from cholesterol in terms of displacement, it is conceivable that some aspects of cholesterol biosynthesis disorders arise from the derangement of cer-rich rafts. To test the hypothesis that precursors differ from cholesterol in terms of displacement, cer-induced displacement of cholesterol precursors and cholesterol from rafts was com-pared. It was necessary to use a detergent-insolubility assay for these experiments. This assay relies on the fact that rafts are insoluble in Triton X-100 (TX-100) and can be separated by centrifugation from disordered domains, which are soluble in TX-100. Unlike the situation in cells, TX-100 insolubility in model membrane vesicles can be studied at room temperature, and the amount of insoluble lipid reflects only pre-existing rafts under our experimental conditions (27,28).
The amount of TX-100 insoluble lipid in MLV composed of 3:3:2 SM/DOPC/sterol was compared with that in vesicles composed of 1.56:1.44:3:2 SM/C16:0 cer/DOPC/sterol (i.e.18 mol % cer). The detergent-resistant lipid was pelleted by centrifugation, chromatographed using high performance (HP)-TLC, and visualized by charring. A sample experiment, shown in Fig. 3, shows that the DRM fraction has a considerable amount of sterol in the absence of cer, and that in the presence of cer there is sterol displacement that is dependent on the type of sterol used. The order of displacement by cer in Fig. 3 is lanosterol Ͼ cholesterol Ͼ 7-DHC.
Quantitative data were obtained by comparing the intensity of the TLC spots to a standard curve (see "Experimental Procedures"). The results are summarized in Table 2. The % of the DRM composed of a specific lipid equals the amount of that specific lipid in the DRM divided by the total amount of lipid in the DRM. The % values in the K p ϭ 1 row give the initial fraction of each lipid in the vesicles. This value would be equal that in the DRM if a lipid partitioned equally between detergent-soluble and detergent-resistant fractions (i.e. had a partition coefficient, K p , equal to 1). Table 2 shows that samples lacking cer, (Ϫ) cer samples, sterol and SM tend to be enriched in the DRM fraction, while DOPC is depleted. This is shown by the fact that % of DRM values for sterol and SM are generally higher than predicted for K p ϭ 1 while those for DOPC are lower than predicted by K p ϭ 1. Lathosterol seems to be an exception to the enrichment of sterol in DRM, but this may be an artifact of selective solubilization (see below).
In samples in which almost half of the SM is replaced with cer, (ϩ) cer samples, both sphingolipids (SM and cer) are enriched in the DRM fraction, but sterol is generally not. Thus, sterol is displaced from the DRM fraction when cer is present. The ratio of % of DRM values in the presence of cer to that in the  The amount of lipid in these samples was below the detection limit of the standard curve (Ͻ1.5%). AUGUST 4, 2006 • VOLUME 281 • NUMBER 31

Cholesterol Precursors and Rafts
absence of cer show that the most strongly raft-stabilizing sterols, lathosterol, and 7-DHC, are the most resistant to displacement by cer (% of DRM that is sterol decreases by about onethird), whereas weakly raft-stabilizing sterols like lanosterol and zymostenol are almost completely displaced by cer (% of DRM that is sterol drops by 80 -90%). Intermediate levels of displacement are observed for cholesterol. Intermediate levels of displacement are also observed for desmosterol, but this may be an artifact of low DRM recovery (see below). It should be noted that the yield of total DRM lipid was lower in the presence of cer than in its absence, despite the fact that the rafts were more stable in the presence of cer. This could reflect a true difference in the amount of rafts present in the model membranes in the presence and absence of cer. However, we have previously shown that sterol enhances the insolubility of rafts in Triton X-100 (27), so a more likely explanation is that the cer-rich rafts are not as detergent-resistant as those in the absence of cer because they are not as sterol-rich. Consistent with this explanation, Table 2 shows that the yield of total lipid in cer-rich rafts is highest in the presence of lathosterol and 7-DHC, the sterols that were not markedly displaced by cer. Partial solubilization of rafts may also be occurring in the absence of cer, although to a much lesser degree.
Interestingly, a low yield of Triton X-100 insoluble cer-rich DRM is not a universal phenomenon. In samples containing DPPC in place of SM, the yield of DRM in the presence of cer is higher than in its absence, as judged by %OD (see below). Fully quantitative TLC experiments in DPPC-containing samples were precluded by the overlap of DPPC and DOPC spots, which prevented determining the amount of DPPC and DOPC in DRM by our detection method. Nevertheless, when sterol association with DRM in the presence and absence of ceramide was measured in samples in which DPPC was used in place of SM results qualitatively similar to those in the SM-containing samples were observed (data not shown).
Fluorescence Anisotropy Assay of the Effect of Sterol Structure upon Displacement from Rafts by Ceramide-Partial solubility of rafts in detergent could distort DRM composition relative to that of the rafts from which they were derived. Therefore, assays that would give information about the sterol dependence of the composition and/or physical properties of rafts without use of detergent were also used. We have previously shown that the fluorescent probe DPH is displaced from rafts by cer (14). The dependence of DPH displacement upon the identity of sterol in cer-containing vesicles was examined in order to see if there was a correlation between sterol type and the ability of cer to displace hydrophobic molecules from rafts. DPH fluorescence anisotropy was measured to determine when it was located in rafts. DPH anisotropy is a measure of DPH motion; it is high in ordered domains and low in disordered domains (27). Fig. 4 (filled triangles) shows DPH anisotropy in vesicles composed of SM, DOPC and sterol. These values are intermediate between the values observed in vesicles that are 100% in an ordered state (about 0.3, see Ref. 27) and 100% in a disordered state (Fig. 4, open circles and squares). This is expected because the SM/DOPC/sterol samples have co-existing ordered and disordered domains, and DPH partitions roughly equally between these domains (28). Anisotropy is slightly higher in the samples containing lathosterol or 7-DHC. This seems to reflect, at least partly, higher DPH anisotropy in the disordered state domains, as shown by the higher anisotropy values observed in bilayers composed of DOPC and these sterols relative to the mixtures of DOPC with the other cholesterol precursors (Fig. 4,  open triangles and open squares). In contrast, controls with LcTMADPH, which locates almost exclusively in the rafts in such samples, shows that anisotropy in rafts is not affected significantly by sterol type (Fig. 4) or cer (14). Another contribution to the high anisotropy values in the samples containing lathosterol and 7-DHC may be that they contain a higher fraction of raft domains, so that a larger fraction of DPH molecules locates in the rafts. Fig. 4 also shows that substitution of a little under half of the SM with cer results in a decrease in DPH anisotropy (from filled triangle values to filled circle values). This decrease most likely reflects displacement of DPH from rafts. Previous studies rule out the alternative possibility of decreased anisotropy because of an increase in the fluidity of cer-rich rafts relative to those without cer (14). Another alternative that is unlikely is that the amount of rafts significantly decreases in the presence of cer. Interestingly, the decrease in anisotropy upon introduction of cer approaches values for complete displacement to a lesser degree in the presence of sterols that strongly stabilize rafts and resist displacement from rafts by cer (lathosterol and 7-DHC). The behavior of DPH in these samples is consistent with the hypothesis that these sterols decrease association of cer with rafts (see below). Lower cer concentrations in these rafts could allow them to accommodate more DPH. Overall, the insolubility and anisotropy results indicate that cer-induced displacement of hydrophobic molecules with small polar groups (sterol) or no polar groups (DPH) from rafts is less effective in the presence of cholesterol precursors that strongly stabilize rafts.

Fluorescence Quenching Confirms the Strong Affinity of 7-DHC for Rafts and Its Resistance to Displacement by Ceramide-We previously demonstrated that detergent insolubility
can faithfully report the composition of lipid rafts in model membrane vesicles (28). To confirm this in the context of cerinduced displacement of sterols, we devised a novel quenching assay. This method exploits our discovery that 7-DHC is a good quencher of Trp fluorescence. 7-DHC has an absorbance spectrum that overlaps Trp emission so that quenching is likely to involve fluorescence resonance energy transfer (not shown).
The partition of 7-DHC between sphingolipid-rich ordered domains and DOPC-rich disordered domains, and how its partition was affected by cer, was determined by first constructing a standard curve for the relationship between the amount of 7-DHC-induced quenching of the Trp fluorescence of LW peptide, a transmembrane helix-forming polypeptide (31), and sterol concentration in DOPC containing vesicles. As shown in Fig. 5A, quenching of Trp fluorescence monotonically increases as sterol concentration is increased.
Quenching measurements were then made in samples containing co-existing sphingolipid-rich rafts and DOPC-rich disordered domains. The level of quenching of Trp fluorescence can be used to estimate the amount of sterol in the disordered domains of such samples because previous studies have shown that the LW peptide is basically restricted to the disordered domains (31). In vesicles containing 7-DHC but no cer, the concentration of 7-DHC in the disordered domains is about 3.2 mol% (Fig. 5B). This value is much lower than the overall concentration of 7-DHC in the sample (25 mol %), and thus the concentration of 7-DHC in the rafts must be much higher than 25 mol %. Since the partition coefficient is defined as the ratio of the 7-DHC concentration in rafts to that in disordered domains, this means that 7-DHC preferentially associates with rafts, in agreement with the results shown in Table 2. In the presence of cer, the concentration of 7-DHC in the disordered domains increases significantly (to 19 mol %). However, this is still less than the overall sterol concentration, indicating that 7-DHC still partitions somewhat preferentially into cer-rich rafts relative to disordered domains.
It should be noted that when using this quenching technique it is necessary to know the fraction of the bilayer in the form of ordered domains in order to calculate the exact concentrations of sterol in such domains. If we very crudely estimate that half of the lipid bilayer in the samples containing 7-DHC forms rafts based on the fact that the bilayers contain equal amounts of sphingolipid and DOPC, and the fact that the 30 -33 % value for the amount of insoluble lipid in these samples (Table 2) is likely to be only a lower estimate to the amount of ordered domains present, then the 7-DHC concentration in the rafts would be 47 mol % in the absence of cer and 31 mol % in the presence of cer. These values are very close to those derived from the analysis of the composition of DRM fractions in Table 2.
Correlation between Sterol Stabilization of Rafts and the Ability of Sterols to Displace Ceramide from Rafts-If displacement involves a competition process, there should not only be an inverse correlation between the raft-stabilizing ability of a sterol and its ability to be displaced from rafts by cer, as found above, but also a positive correlation between the raft-stabilizing ability of a sterol and its ability to displace cer from rafts.
To test this hypothesis, the effect of cholesterol precursors upon partitioning of [ 3 H]cer into DRM was measured using MLV containing co-existing ordered and disordered domains (Fig. 6). To avoid the low degree of insolubility encountered in the experiments using SM and cer, lipid mixtures were used that contained DPPC in place of SM. In Fig. 6, % OD (open bars), which measures the amount of light scattering from the detergent insoluble lipid, is a measure of the amount of DRM.
(We have previously shown that when using MLV, % OD is roughly proportional to the concentration of insoluble lipid, but tends to give higher values of insoluble lipid than direct  measurements of lipid amount (14,27,32). This can also be seen in the data shown in Table 2.) Notice that % OD values are in general higher in the DPPCϩcer-containing samples than in analogous SMϩcer-containing samples ( Table 2). Nevertheless, as in the case of the SMϩcer-containing samples (Table 2), the amount of DRM as measured by % OD is highest in DPPCϩcer-containing samples that contain the most strongly raft-stabilizing sterols, 7-DHC and lathosterol (Fig. 6, open  bars).
The ratio of % [ 3 H]cer radioactivity in DRM (% R pellet) to % OD measures the concentration of cer in rafts. Fig. 6 shows that sterols reduce cer concentration in rafts, as shown by the fact that the % R pellet/% OD ratio decreases in the presence of sterols (shaded bars). This decrease is smaller for the precursors that only weakly stabilize raft formation, and larger for precursors that strongly stabilize raft formation (lathosterol and 7-DHC). This agrees qualitatively with the data giving the concentration of cer in cer-rich rafts in Table 2. Nevertheless, the reduction of cer concentration in DRM is relatively modest for most sterols, with the maximal decrease, given by the decrease in % R pellet/% OD in the presence of sterol relative to that in the absence of sterol, being about 2-fold in the most extreme cases, even though overall sterol concentration in the samples is about twice that of cer. In fact, at least much of the decrease in cer concentration in DRM is simply due to the increase in the amount of DRM when sterol becomes DRM-incorporated. On the other hand, Table 2 shows that in cer-containing samples there is a lower cer/SM ratio in DRM containing lathosterol and 7-DHC than in DRM containing more weakly raft stabilizing sterols. This suggests lathosterol and 7-DHC occupy sites in ordered domains that would otherwise accommodate cer.
This weak displacement of cer is consistent with our previous observations that cer resists displacement by cholesterol to a greater degree than cholesterol can resist displacement by cer (14). Combining data in Fig. 6 and Table 2 suggests that the most strongly raft-stabilizing sterols (lathosterol and 7-DHC) might decrease cer concentration in rafts to about the same extent that their concentration in rafts is decreased by cer.

DISCUSSION
Sterol Structure and Stabilization of Rafts-Although some precursors of the cholesterol biosynthetic pathway have been examined for their ability to substitute for cholesterol, this is the first report to compare raft participation for a nearly complete set of cholesterol precursors in combination with a variety of sphingolipids. Several observations in this study have important implications. First, all of the precursors in the early part of the pathway are weakly raft-stabilizing. For precursors on the branch in which the alkyl tail is fully saturated there is a rather abrupt transition between weakly and strongly raft-stabilizing sterols associated with the isomerization of the 8 -9 double bond to a 7-8 double bond. Interestingly, inspection of the membrane sterols of fungi, algae, plants and animals shows that although organisms have sterols with various structures, and sterols with a 5-6 double bond (like cholesterol) and/or 7-8 double bonds are common, sterols with 8 -9 double bonds are rare (33). This suggests the ability of sterols to support the for-mation of ordered states within membranes is an important part of their biological function.
The alkyl tail 24 -25 double bond also has significant effect on raft stability. The observation that desmosterol is significantly less raft-stabilizing than cholesterol agrees with the conclusion of the very recent study of Vainio et al. (34). However, the presence of the 24 -25 double bond is not always destabilizing. For lanosterol and zymosterol, the presence of the 24 -25 double results in the formation of sterols that are not significantly less raft-stabilizing than the analogous sterols lacking the 24 -25 double bond. This shows that the effects of sterol tail and ring structure upon raft-stabilizing ability are not always additive. This contrasts with our observations of additive effects for multiple modifications within the sterol rings (32).
It should be noted that the sterol tail is likely to interact with, or indirectly influence, lipids in the trans leaflet, and in natural membranes the lipids of the inner leaflet are very different than those of the sphingolipid-rich outer leaflet. This contrasts with model membranes, which have a basically symmetric lipid composition. Thus, the effects of tail structure upon raft stability in cells may be different than in model membranes.
Sterols had smaller effects upon raft stability in samples that contained sphingolipid mixtures with a significant fraction of cer or cerebroside than in samples containing SM as the sole sphingolipid, or when DPPC was used in place of sphingolipid, although the relative dependence upon sterol structure was very similar in all of cases. The lesser effects of sterols in the former mixtures can be explained by a lesser capacity of rafts containing cerebrosides and cer to incorporate sterol, as we proposed previously (14,27), and is confirmed for cer-containing samples by the sterol displacement experiments.
It should be noted that we cannot rule out the presence coexisting cerebroside-rich and SM-rich ordered domains in these samples. However, the fact that the melting event detected in vesicles with both SM and cerebroside was at a higher temperature than that observed with SM and no cerebroside suggests that a considerable fraction of the cerebroside dissolved in the SM-rich domains. In addition, previous studies show up to 15mol% cerebroside (our samples had 12 mol % cerebroside) can dissolve in disordered domains before cerebroside-rich ordered domains begin to form (35). This suggests that any cerebroside-dominated domains that co-exist with the SM-rich domains would only be a small component in our samples.
It should be noted that in experiments using DPPC-or SMcontaining samples we found absolute T m values measured using LcTMADPH were 2-5°C higher than those obtained when DPH was used as the fluorescent probe (13,32). This is an expected result due to the fact that LcTMADPH has a very strong affinity for rafts (13), and so will partition significantly into rafts even when they form only a small fraction of the bilayer. In contrast, ordinary DPH partitions nearly equally between ordered and disordered domains (28,36,37). However, due to its displacement from rafts by cer, DPH could not be used to detect cer-rich rafts. Use of LcTMADPH allowed comparison of ordinary and cer-rich rafts.
It should also be noted that using tempo in place of nitroxidelabeled lipids could complicate quenching behavior. In partic-ular, if tempo partitioned favorably into the boundaries between fluid and disordered domains, the level of quenching at the lower temperatures (at which ordered and disordered domains co-exist) might be larger than expected. The fact that both the change in quenching as a function of temperature, and the melting temperatures, we detected were similar to those previously detected in analogous mixtures using a nitroxidelabeled lipid as quencher (13,27,32) suggests any partition of tempo into the boundary between ordered and disordered domains is not sufficient to greatly affect the observed melting temperatures.
Competition Between Sphingolipids and Cholesterol Precursors for Association with Rafts-We found previously that cer has the ability to greatly stabilize lipid rafts (13,14), and that cer-rich rafts exclude cholesterol (14). We now show that the ability of sterols to associate with cer-rich rafts is related to their raft-stabilizing abilities. The more strongly a sterol stabilizes rafts, the harder it is to displace from rafts. In addition, the more strongly a sterol stabilizes rafts, the greater its ability to displace cer from rafts. These differences may have important functional implications (see below).
The displacement of cer by sterol explains the decrease in raft T m values when certain sterols were included in cer-containing samples. Interestingly, a sterol-induced decrease in T m was not observed in the cerebroside-containing samples even though they also exhibited T m values that were not very sterol-sensitive. We interpret this result in terms of the umbrella model for the interaction of sterol and analogous molecules with lipids (14,38). The amount of small headgroup lipids, such as sterol and cer that can be accommodated in a domain is limited: too much would result in exposure of non-polar lipid to water. Thus, sterol and cer compete for ordered domain association. Cerebrosides have large headgroups, and so are not subject to competition in this fashion. Instead, they may pack too tightly in ordered domains, even when mixed with SM, to accommodate a large amount of sterol.
It should be noted that differences in the affinity of ceramide and cholesterol for SM-rich ordered domains probably involve more than just umbrella effects. Differences between the strength of van der Waals and/or H-bonding interactions, or differences in the ability of sterol and ceramide to relieve steric clashes between SM headgroups could also be involved.
Comparison to Previous Studies of Cholesterol Precursor Properties-Several previous studies have revealed that cholesterol precursors differ significantly from cholesterol in terms of their physical properties. We reported that lanosterol did not stabilize rafts in mixtures of DPPC and nitroxide-labeled lipids, and subsequently several groups confirmed this and other differences between lanosterol and cholesterol using techniques such as NMR (39 -41) and fluorescence microscopy (42,43). Unlike many other precursors, lanosterol cannot substitute for cholesterol in mutant CHO cells (44), although selection of lanosterol-containing mammalian cell lines is possible (45).
The effect of desmosterol on membrane structure and function is not as clear. Its mixtures with 1-palmitoyl 2-oleoyl phosphatidylcholine have physical properties similar to those of cholesterol mixed with this phospholipid (46). On the other hand, in agreement with our data, the very recent study of Vainio et al. (47) indicates desmosterol does not stabilize ordered domains as well as cholesterol, and substitution of desmosterol for cholesterol in cells disrupts at least one signal transduction pathway. Additional in vivo data on desmosterol are not definitive. Desmosterol can fully substitute for cholesterol in cells (44) and mice (48), although it leads to serious disease in humans (2). It is possible that with the aid of compensatory changes in membrane composition desmosterol can substitute for cholesterol in almost all biological processes. There is no doubt its behavior is more cholesterol-like than that of lanosterol.
We also previously found that lathosterol and 7-DHC stabilize rafts better than cholesterol (13,32), and Nyholm et al. (49) predicted that lathosterol would stabilize rafts based in its behavior in binary mixtures with sphingolipid. Data on lathosterol from cells also agree well with these results. Cellular lathosterol was found to concentrate in cell-derived DRM at least as well as cholesterol and association of newly synthesized lathosterol with cellular DRM was actually more rapid than that of cholesterol (50). Recently, the ability of 7-DHC to support DRM in mouse brain tissue was also observed (51).
Implications for the Bloch Hypothesis-Konrad Bloch speculated that the cholesterol biosynthetic pathway recapitulated cholesterol evolution (52). According to this hypothesis, cholesterol precursors should have properties that gradually support cellular function better as they progress along the pathway toward cholesterol. Clearly, if raft formation were the only critical function of cholesterol, the data in this report would suggest this hypothesis is not correct. Instead of a gradual and progressive improvement in raft-stabilization properties, the change in raft-stabilizing properties is abrupt, and cholesterol is not the most raft-stabilizing sterol.
However, this is not conclusive. It is possible that cholesterol is the most raft-stabilizing sterol in the complex lipid and protein environment in cell membranes. In this regard, it must be noted that fungal, plant and animal sphingolipids differ significantly (33). Further complicating this analysis is the possibility that cholesterol intermediates might have co-evolved with sphingolipids. There might have been a gradual increase in raftstabilizing ability of various sterol-sphingolipid combinations during evolution, but if the appropriate sphingolipids were only present in extinct organisms, the intermediate raft-stabilizing abilities of the precursor sterols could not now be detected.
Another possibility is that raft formation is not the sole (or even main) function of cholesterol in cell membranes. Cholesterol might be optimized in terms of its total function. For example, suppose that the ability of a sterol to associate with rafts in a regulated manner is important, e.g. assume the ability to be displaced from rafts by cer is a key sterol function. In this case, although 7-DHC and lathosterol are better raft-stabilizers than cholesterol, because they are displaced from cer-rich rafts to a lesser degree than is cholesterol they would not have optimal function. Cholesterol might have an optimal balance of raft-stabilizing properties and ability to be displaced from cerrich rafts.
On the other hand, it is possible that the Bloch hypothesis is incorrect. Intermediate cholesterol precursors might have evolved from lanosterol for reasons that have nothing to do with their function as bulk membrane lipids. For example, a precursor might have first evolved because of selection for hormone-like properties (53), or as an intermediate in the synthesis of some other molecule with a non-structural role, and then later was co-opted as a bulk membrane sterol because of raftstabilizing abilities. Co-option of a structure that has been selected for one function in order to carry out a second function is a common evolutionary theme.
Implications for Diseases of Cholesterol Biosynthesis-In SLOS and lathosterolosis there is an accumulation of cholesterol precursors that are even more raft-stabilizing than cholesterol, while in desmosterolosis and CHILD syndrome there is an accumulation of precursors that are less raft-stabilizing than cholesterol (5). Thus, in these diseases the amount or physical properties of rafts could be affected. This might account for some symptoms of these diseases. For example, precursors altering composition of cer-rich rafts could affect processes such as apoptosis and bacterial and viral pathogenesis, in which cer-rich rafts have been proposed to function (12).
We previously suggested that alterations of rafts due to an accumulation of 7-DHC might impact SLOS (13). It is believed that many SLOS symptoms are a result of modification of signaling in the hedgehog pathway, and genetic mutations in hedgehog signaling proteins result in disorder(s) with phenotype(s) similar to those of SLOS (54). One possible connection between blocked cholesterol biosynthesis and hedgehog function might involve alterations in the covalent modification of hedgehog by sterol, which is essential for the signaling of the hedgehog molecule (55). However, Cooper et al. found that hedgehog autoprocessing and modification by cholesterol precursors proceeded normally in cells when cholesterol biosynthesis was genetically altered to end at 7-DHC or lathosterol, despite altered hedgehog signaling response (54). They suggested that smoothened (Smo), a membrane protein downstream from hedgehog in the signal transduction pathway, is affected by sterol modification via membrane properties altered by the presence of the precursor sterols. Such alterations could involve the 7-DHC-and lathosterol-induced changes in raft formation and properties as noted in this and previous reports (13,32). Alternately, changes in overall physical properties of the membrane, or changes in specific sterol-protein interactions could be involved.