Fluorescence and multiphoton imaging resolve unique structural forms of sterol in membranes of living cells.

Although cholesterol is an essential component of mammalian membranes, resolution of cholesterol organization in membranes and organelles (i.e. lysosomes) of living cells is hampered by the paucity of nondestructive, nonperturbing methods providing real time structural information. Advantage was taken of the fact that the emission maxima of a naturally occurring fluorescent sterol (dehydroergosterol) were resolvable into two structural forms, monomeric (356 and 375 nm) and crystalline (403 and 426 nm). Model membranes (sterol:phospholipid ratios in the physiological range, e.g. 0.5-1.0), subcellular membrane fractions (plasma membranes, lysosomal membranes, microsomes, and mitochondrial membranes), and lipid rafts/caveolae (plasma membrane cholesterol-rich microdomain purified by a nondetergent method) contained primarily monomeric sterol and only small quantities (i.e. 1-5%) of the crystalline form. In contrast, the majority of sterol in isolated lysosomes was crystalline. However, addition of sterol carrier protein-2 in vitro significantly reduced the proportion of crystalline dehydroergosterol in the isolated lysosomes. Multiphoton laser scanning microscopy (MPLSM) of living L-cell fibroblasts cultured with dehydroergosterol for the first time provided real time images showing the presence of monomeric sterol in plasma membranes, as well as other intracellular membrane structures of living cells. Furthermore, MPLSM confirmed that crystalline sterol colocalized in highest amounts with LysoTracker Green, a lysosomal marker dye. Although crystalline sterol was also detected in the cytoplasm, the extralysosomal crystalline sterol did not colocalize with BODIPY FL C(5)-ceramide, a Golgi marker, and crystals were not associated with the cell surface membrane. These noninvasive, nonperturbing methods demonstrated for the first time that multiple structural forms of sterol normally occurred within membranes, membrane microdomains (lipid rafts/caveolae), and intracellular organelles of living cells, both in vitro and visualized in real time by MPLSM.

Cholesterol is essential for optimal membrane transport, receptor-effector coupling, cell recognition, and other eukary-otic cellular processes (reviewed in Ref. 1). Increasing evidence indicates that plasma membrane cholesterol is organized into lateral and transbilayer cholesterol-rich microdomains (reviewed in Ref.
2) such as lipid rafts and caveolae (reviewed in Refs. [2][3][4][5][6]. These domains contain proteins involved in multiple cellular functions including signaling and cholesterol transport. For example, overexpression of caveolin-1 in mice stimulates high density lipoprotein uptake via plasma membrane caveolae and markedly increases plasma high density lipoprotein-cholesterol (8). In contrast, caveolin-1-deficient mice are lean and resistant to diet-induced obesity but show hypertriglyceridemia and exhibit vascular abnormalities (9 -11). There is a strong association of cholesterol abnormalities with cytotoxicity, sickle cell acanthacytosis, Niemann-Pick C disease, Alzheimer's disease, atherosclerosis, diabetes, and obesity (reviewed in Refs. [12][13][14][15][16]. However, almost nothing is known regarding cholesterol organization within cholesterol-rich microdomains. The available evidence suggests that the cholesterol content of plasma membrane microdomains is very high, possibly sufficient for cholesterol phase separation into crystals. Plasma membranes exhibit the highest cholesterol:phospholipid molar ratio (e.g. 0.5-1.0) for any intracellular membrane, and this ratio appears to be closely regulated (reviewed in Ref. 2). Excess membrane cholesterol phase separates into crystalline cholesterol at cholesterol:phospholipid molar ratios higher than 1.0 in model membrane bilayers (reviewed in Ref. 17), plasma membranes (18), and the lysosomal matrix as well as cytosol of macrophage foam cells (19). Cholesterol monohydrate crystals within plasma membranes of smooth muscle cells (18,20,21), lysosomes and cytoplasm of macrophage-derived foam cells (14,19), and atherosclerotic plaques (13) are cytotoxic. Plasma membranes of certain types of cell derivation are composed of cholesterol-rich cytofacial leaflets (reviewed in Ref. 22) and lateral lipid raft/caveolar microdomains (reviewed in Refs. 2, 3, 5, 23, and 24), wherein the cholesterol:phospholipid ratio is expected to be Ͼ Ͼ1. However, if cholesterol phase separates into crystalline cholesterol at such ratios, it is difficult to conceive how these microdomains effectively facilitate cholesterol transport, potocytosis, and signaling (2,3,5,23,24). Therefore, it is important to determine the structural form of cholesterol within the plasma membrane, especially in lipid rafts/caveolae, as well as in other intracellular organelles and membranes. Unfortunately, relatively few noninvasive, nonperturbing techniques exist for real time visualization of cholesterol structures in biological membranes or in living cells (18,25).
The present investigation addresses these issues by taking advantage of two major technological advances. First, the spectral properties of a naturally occurring fluorescent sterol, de-5 ml of 36% (w/v) sucrose (d ϭ 1.139)) in 10 mM Tris, 1 mM EDTA, pH 7.8. After 90 min at 39,000 rpm on an SW40Ti swinging bucket rotor and Beckman XL-90 ultracentrifuge (Beckman Instruments, Fullerton, CA), pure sterol crystals appeared at d Ͻ1.054 (13), whereas plasma membrane-enriched fractions appeared at the 36 -24% (w/v) sucrose interface (43). The relative enrichment of the lipid raft/caveolar membrane fraction was determined by quantitative Western blotting using antisera to caveolin-1 and flotillin-1, basically as described for other membrane markers (39).
Measurement of Force-area Isotherms of Pure and Mixed Monolayers-Pure cholesterol or dehydroergosterol monolayers or mixed monolayers containing varying amounts of sterol together with 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine were compressed on 140 mM NaCl in water at 22°C (under an argon atmosphere in the dark) with a KSV surface barostat (KSV Instruments Ltd., Helsinki, Finland). The barrier speed during compression did not exceed 3.4 Å 2 /molecule/min. Isotherms at a surface pressure of 35 mN/m were recorded using proprietary KSV software (44).
Model Membrane Vesicle Preparation-Small unilamellar vesicles (SUV) and large unilamellar vesicles (LUV) were prepared to contain 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC), dehydroergosterol (DHE), and cholesterol in the following indicated proportions Spectral Properties of DHE in Solvents and Membranes-Absorbance spectra were obtained with a Lambda 2 Dual Wavelength Spectrophotometer (PerkinElmer life Sciences). Steady-state fluorescence excitation and emission spectra were obtained with a PC1 Photon Counting Fluorometer (ISS Instruments, Champaign, IL). Spectra were analyzed with Grams32 (Thermo Galactic, Salem, NH). Light scatter was avoided by use of narrow monochromator slits, low concentrations, and appropriate cut-off filters. Any Raman scattering was subtracted from all excitation and emission spectral data. Artifacts because of inner filter effects were avoided by keeping the absorbance of sample solutions at the excitation wavelength (324 nm) below 0.15. All absorbance measurements were performed in 1 ml-or 3.5-ml quartz cuvettes. Fluorescence measurements of organelles and associated membranes were performed with samples in 2 ml of filtered 10 mM PIPES buffer, pH 7.4, in a quartz cuvette with the temperature regulated to 37 Ϯ 0.3°C through use of a water heating bath (Fisher). Six integrated intensity measurements (335-600 nm) were performed on concentrations of DHE in 10 mM PIPES, pH 7.4, and 100% ethanol, ranging from 2.5 to 10 M.
Aqueous Solutions of DHE and Cholesterol for Micelle Determination-Solutions of 2.5 and 10 M cholesterol or DHE were prepared in 95% EtOH and evaporated onto glassware under dry nitrogen. Each of the deposited sterols was then redissolved in 10 mM PIPES buffer at 37°C under heavy vortexing. Filtration of portions of these solutions was made with a single pass through either an Avanti Mini-Extruder with a 0.1-polycarbonate membrane (Whatman) or a 0.5-filter (Millipore, Bedford, MA). Similarly, DPH in tetrahydrofuran was evaporated onto glassware followed by addition of the unfiltrated and filtrated sterol solutions and intense vortexing. The sterol:DPH ratio was maintained at 250:1. Fluorescence emission spectra were obtained as described above.
Determination of DHE Steady-state Polarization during Exchange Assays-Steady-state fluorescence polarization measurements of DHE in plasma membranes, endoplasmic reticulum, mitochondria, lysosomes, and lysosomal membranes at 37°C were performed as described earlier (38 -40). Residual light scatter (from both donor and acceptor membranes) contribution to polarization data was corrected by converting polarization to anisotropy according to the equation r ϭ 2P/(3 Ϫ P), and subtracting the residual fluorescence anisotropy of both donor and acceptor membranes (i.e. not containing DHE) from all experimental data. Absorbance (324 nm) of sample solutions in 10 mM PIPES buffer, pH 7.4, was kept below 0. 15.
Lysosomal Sterol Transfer-Sterol transfer between isolated lysosomes was determined using a fluorescent sterol (DHE) exchange assay as described previously (39,46). The basis of the assay (release from self-quenching of DHE in the donor lysosome and transfer to acceptor lysosome lacking DHE), validation of DHE as a probe for cholesterol transfer, in-depth descriptions of the assay, proper controls, and justification for the lack of polarization change in the absence of acceptor membranes were provided in the above cited publications. Standard curves for DHE in lysosomal-lysosomal membrane exchanges were determined earlier (39,46). Multiphoton Laser Scanning Microscopy (MPLSM) and Image Analysis-MPLSM of DHE, Nile Red, LysoTracker Green, and BODIPY FL C 5 -ceramide was performed on intact L-cells (L arpt Ϫ tk Ϫ ) cultured on two-well Lab-Tek chambered cover glasses (VWR, Sugarland, TX). Prior to imaging, cells were supplemented for 2 days with the addition to the medium of 20 g/ml DHE, either from a 5 mg/ml stock solution of DHE in anhydrous ethanol or with 10 mM 65:35 large unilamellar vesicles of POPC and DHE (prepared as described above). For colocalization experiments of DHE with Nile Red, the cells cultured with DHE were washed with Puck's buffer and then incubated with 100 -400 nM Nile Red for ϳ30 min. The concentration was titered to different levels in this range depending upon the amount of excitation power used in the multiphoton imaging process described below. For imaging with LysoTracker Green, L-cells were grown to confluency on a chambered cover glass supplemented with 20 g/ml medium of DHE for 2 days and then washed in Puck's buffer (1 mM Na 2 HPO 4 , 0.9 mM H 2 PO 4 , 5.0 mM KCl, 1.8 mM CaCl 2 , 0.6 mM MgSO 4 , 6 mM glucose, 138 mM NaCl, and 10 mM HEPES). Cells were then incubated with 50 nM LysoTracker Green in Puck's buffer at 37°C for 1 h; medium was removed, and cells were washed several times with Puck's buffer. To ascertain crystalline DHE and Golgi colocalization, BODIPY FL C 5 -ceramide was purchased as already made complexes with bovine serum albumin. Murine L-cells with DHE (20 g/ml medium) were grown on chambered cover glasses for 2 days and then washed with a Hanks' buffered saline solution (HBSS) with 10 mM HEPES, pH 7.4, subsequently termed HBSS/ HEPES. After washing, the cells were loaded for 30 min with 5 M BODIPY FL C 5 -ceramide-bovine serum albumin at 4°C in HBSS/ HEPES medium. Afterward, the medium was removed, and the cells were washed with HBSS/HEPES at 4°C. After replacing the medium, the chambered cover glass were then placed back in the incubator for 30 min at 37°C. The cells were once again washed with HBSS/HEPES medium and immediately imaged under multiphoton excitation.
MPLSM fluorescence imaging was performed (33) using an MRC1024 Multiphoton Laser Scanning Microscope controlled by Laser-Sharp software and equipped with an external descanned 3 detector unit (Bio-Rad) (47). Briefly, the excitation source was a femtosecond Coherent Mira 900 mode-locked Ti:Sapphire laser with broadband optics pumped at 12 watts with a Coherent Sabre argon ion laser (Coherent, Palo Alto, CA). The excitation light (900 -930 nm) was delivered to an Axiovert 135 (Zeiss Inc., New York, NY) microscope stage via a modified epiluminescence light path. A Zeiss 63ϫ Plan-Apochromat (1.4 N.A.) or 100ϫ Fluar (1.3 N.A. with higher transmittance between 350 and 400 nm) oil immersion objective was used for all images. Selection of 900 -925 nm as the multiphoton excitation wavelength range allowed simultaneous excitation of crystalline DHE and monomeric DHE (three-photon excitation), with either Nile Red, LysoTracker Green, or BODIPY FL C 5 -ceramide (two-photon excitation). The fluorescence emission of the respective probe molecules was collected by the same objective and passed into the Bio-Rad external detector unit where the emission was appropriately separated into different wavelength regions and collected by three photomultiplier tubes (PMT) as follows: (i) multialkali PMT1 collected emission through a D455:30-nm filter; (ii) multialkali PMT2 collected emission through a HQ575:150-nm filter; (iii) bi-alkali PMT3 collected emission through a D375:50-nm filter. All filters and dichroics used in detection were from Chroma Technology (Brattleboro, VT). Images (Kalman filtered) were analyzed and presented using a combination of software packages including MetaMorph Image Analysis Software (Advanced Scientific Imaging, Meraux, LA), Adobe Photoshop 5.0 (San Jose, CA), CorelDraw 9 (Ottawa, Ontario, Canada), and National Instruments LabView 6i equipped with IMAQ Vision 5.0 (Austin, TX). MPLSM of autofluorescence in L-cells was obtained for each experiment at the same excitation power and wavelength as used for collecting probe fluorescence. The gain and black levels of each photomultiplier tube were optimized to minimize the autofluorescence signal and to maximize the fluorescence signal in the probe-supplemented cells. Any residual autofluorescence signal at high detection sensitivity in the 375:50-nm channel was subtracted.
In order to separate crystalline and monomeric DHE signals for colocalization with other probes and thus minimize the spectral "bleed through" of each form into the other channel, software was written using LabView 6i equipped with IMAQ Vision 5.0 utilizing the pixel color intensity representation of pixel fluorograms. The program, called FLUOROGRAM_IM, provided the user with a way to import an RGB image and form a pixel fluorogram from the red and green components by plotting the red intensities along the vertical axis and the green intensities along the horizontal axis. Once a pixel fluorogram was created, the program allowed the user to collectively select pixels by drawing a region (specifiable in any shape or size) around the pixels belonging to either form of DHE. Pixels belonging to crystalline DHE were chosen based upon the ratio I 475nm :I 375nm Ն 1, whereas the remaining pixels with I 475nm :I 375nm Ͻ1 were deemed monomeric (intensity ratio ϭ 1 along the diagonal of the fluorogram). After the selection was made, the software program turned off those pixels in the original image whose correlating intensities do not fall in the selected region. The new gray scale images corresponding to each form (crystalline, 455 channel; monomeric, 375 channel) is then saved for further colocalization analysis with the second probe (third channel) used in the experiment. Areas of low concentration and/or low sensitivity as well as saturable conditions produce ambiguous correlations but mostly can be overcome by imaging cells at different levels of sensitivities depending upon the focus of the localizable product. Generally, the bulk of the intensity was separable and highly indicative of the structural form of sterol.
After background correction, colocalization coefficients for each pseudo-color (red or green channel) were calculated using Equation 1 (Bio-Rad Tech Note 8) (48), where C red and C green are the colocalization coefficients; I i r and I i g represent the red and green intensities of the ith pixel after background subtraction, and ␦ rg (i) and ␦ gr (i) are Kronecker deltas that represent intensity independent colocalization weights. For every ith pixel, ␦ rg (i) ϭ ␦ gr (i) ϭ 0, if there is only a red intensity or green intensity but not both. ␦ rg (i) ϭ ␦ gr (i) ϭ 1, only if there is both a red and green intensity associated with the ith pixel. Therefore, a coefficient represents the fraction of the total intensity of the filtered fluorescence of the corresponding probe in an image that is colocalizing with the filtered fluorescence of another probe.

Synthesis and Purification of DHE-To determine structural
properties of sterol in membranes of living cells, it was essential that DHE be free of contaminants. A commercially available DHE as well as DHE freshly synthesized and purified herein (see "Experimental Procedures") were examined. Absorbance spectra of the two preparations differed only slightly in the region where DHE maximally absorbs (peak maxima near 311, 324, and 340 nm) (Fig. 1A). However, below 275 nm the absorbance of the two preparations deviated significantly, indicating the presence of impurities. HPLCs of commercially available DHE (10 g/10 l solvent) revealed four peaks (Fig.  1B). However, only peak 3 (representing 83% of total) exhibited absorption characteristics of DHE. Peaks 1 and 2 were sample impurities, whereas peak 4 was due to a small solvent impurity. In contrast, the DHE synthesized and purified as indicated under "Experimental Procedures" was 98% pure as indicated by HPLC (Fig. 1C). When Peak 3 (pure DHE in Fig. 1C) was pooled from a number of HPLC runs, dried under nitrogen, and reinjected on the HPLC column, the resultant DHE was 99.7% pure (Fig. 1D, peak 3).
Absorbance Spectral Properties of DHE in Ethanol and in Aqueous Buffer-Although DHE is monomeric at low concentrations in ethanol, aggregation of DHE occurs in aqueous buffers due to the low critical micellar concentration (20 -30 nM for cholesterol and DHE) of sterols (26 -30). Absorbance spectral properties of DHE in aqueous buffers (10 M in 10 mM PIPES, pH 7.4) differed significantly from those of monomeric DHE (10 M in ethanol). In ethanol DHE displayed maxima at 311, 324, and 340 nm ( Fig. 2A). In contrast, DHE in aqueous buffer (Fig. 2C) exhibited the following differences. (i) The absorption spectrum in aqueous buffer was slightly broader with maxima (314, 329, and 347 nm) that were red-shifted by 3, 5, and 7 nm from those in ethanol. (ii) The absorption peak profile of DHE in aqueous buffer was significantly altered. The ratio of absorption maxima at 329:314 nm was 1.03, signifi-cantly lower than that for absorption maxima at 324:311 nm in ethanol near 1.14. (iii) The molar extinction coefficient of DHE in 10 mM PIPES, pH 7.4 (E M ϳ5500 M Ϫ1 cm Ϫ1 ) was 2.4-fold lower than for DHE in anhydrous ethanol (E M ϳ13,000 M Ϫ1 cm Ϫ1 ), both measured at 324 nm.
Fluorescence Emission Characteristics of DHE in Ethanol and in 10 mM PIPES Buffer, pH 7.4 -The fluorescence emission spectral properties of DHE in ethanol and aqueous buffer differed even more than the absorption spectra. The fluorescence emission of DHE in ethanol ( Fig. 2B) was Stokes shifted ϳ46 nm as compared with the absorption spectrum ( Fig. 2A), essentially the same as that obtained for DHE in other alcohols (1-butanol, 2-propanol, and methanol) (28). The fluorescence emission spectrum of DHE in ethanol (Fig. 2B) was a near mirror image of the absorption spectrum ( Fig. 2A) and exhibited vibrationally resolved fluorescence emission maxima near 354, 370, and 390 nm (Fig. 2B). In contrast, the aqueous fluorescence emission spectrum of DHE was consistent not only with the presence of monomeric DHE spectral features as observed for DHE in ethanol ( Fig. 2B) but also indicated the appearance of a completely new fluorescent entity differing markedly from that in ethanol (Fig. 2B) as follows. (i) DHE emission maxima (356 and 375 nm) were red-shifted as compared with those in ethanol. (ii) New fluorescence emission maxima appeared at 403 and 426 nm (Fig. 2C). (iii) The relative emission intensity of the new emission maxima was severalfold greater than those at 356 and 375 nm (Fig. 2D). This increased intensity was at some expense of the DHE emission at 356 and 375 nm because their intensities were 0.65 that of the emission maxima in ethanol (Fig. 2B). However, the overall integrated intensity of the emission from the DHE in 10 mM PIPES was 2.7 Ϯ 0.2 greater than that in ethanol. (iv) The quantum yield (88) of DHE in 10 mM PIPES, pH 7.4, was increased 6-fold from 0.04 in ethanol (56) to 0.25 Ϯ 0.02, due to the overall increase in integrated emission and decrease in absorption.
A fluorescence emission spectrum of the new DHE fluorescing species, formed in 10 mM PIPES, pH 7.4 ( Fig. 2E), was resolved from that of the monomeric DHE by subtraction of the normalized fluorescence emission spectrum of DHE in ethanol from that in aqueous buffer. The new spectral profile of DHE exhibited fluorescence emission maxima at 403 and 426 nm, with a slight shoulder at 460 nm. The appearance of this new fluorescence could be due to an impurity in the DHE preparation. Although this may potentially be the case in studies where partially pure DHE (e.g. Fig. 1, A and C) was used, in the present investigation the DHE was highly purified (99.7% pure, Fig. 1D). In contrast to the commercially available preparation, the highly purified DHE did not contain significant impurities. These data indicated that the DHE species emitting maximally at 426 nm in PIPES, pH 7.4, was not due to a contaminant in the DHE preparation, but rather it was due to the presence of aggregating DHE in the form of microcrystals or micelles. That the DHE was present primarily as microcrystals, rather than micelles, was resolved as follows.
First, aqueous solutions of DHE or cholesterol were examined under polarized light microscopy using the Zeiss Axiovert Microscope with 63ϫ oil objective. Clearly visible under crosspolarization of DHE solutions were small birefringent crystals. For comparison, aqueous solutions of cholesterol were also made and examined using the same technique. Under these conditions, small birefringent crystals of cholesterol monohydrate were detected, confirming earlier results of others (49). Importantly, the DHE microcrystals were similar in size and shape as that those of cholesterol.
Second, aliquots of the aqueous solutions of DHE (2.5 and 10 M) were filtered through 0.1-m filters, which would allow passage of micelles (but not microcrystals) with sizes on the order of 4 nm. Emission spectra over the region 350 -600 nm of the aqueous solutions of DHE were recorded before and after filtration using an excitation wavelength of 324 nm. Before A, absorbance spectra of DHE (19 g/ml ethanol). Curve 1 represents DHE purchased commercially as indicated under "Experimental Procedures." Curve 2 represents recrystallized DHE that was synthesized and freshly prepared as indicated under "Experimental Procedures." B, HPLC of commercially available DHE (10 g/10 l solvent). Peak 3, the only material absorbing at 325 nm (absorption maximum of DHE) represents 83% of total. Peaks 1 and 2 represent impurities present in the sample, and peak 4 is due to a small solvent impurity. C, HPLC scan of DHE (10 g/10 l solvent) synthesized and recrystallized as indicated under "Experimental Procedures." DHE present in peak 3 represents 98% of total. D, DHE was collected from chromatograms (peak 3 in C), dried under nitrogen, and reinjected (30 g/10 l solvent) on the HPLC column. The final purified DHE (peak 3 in D) represented 99.7% pure DHE. All HPLC eluants were monitored by absorbance at 205 nm as indicated under "Experimental Procedures." filtration, the spectrum of DHE showed the new fluorescing species as shown in Fig. 2D. However, after filtration no spectral emission was observed above the background noise. By increasing the slit sizes on the monochromator, sensitivity was increased but without any detectable DHE emission. In an effort to determine whether there are possible contributions from larger aggregates of micelles, the filtration was repeated using a 0.5-m filter. Once again no DHE emission was detected in the filtrate.
Third, the fluorescence probe DPH was used to detect the presence of micelles. In order to excite DPH and not DHE, the excitation wavelength was changed to 369 nm, and emission spectra were recorded over the range 385-600 nm. DPH was added to the unfiltrated and filtrated solutions of DHE as described under "Experimental Procedures." Initially, a spectrum was obtained on DPH in buffer without DHE to determine any background level of DPH emission. No detectable emission was observed. Next, the spectra of DPH added to the DHE solutions before and after filtration were obtained, also showing no DPH emission. As a control, similar concentrations of cholesterol in aqueous buffer were prepared and filtered similarly to that of the DHE. As was the case with DHE, no DPH emission was detected for cholesterol.
These experiments suggest that the spectral emission of the unfiltered solutions of DHE is derived from microcrystals larger than 0.5 m. This is supported by comparison of spectral characteristics of aqueous dispersions of DHE with those of DHE in the form of crystalline powder (30). Although the purity of the latter preparation of DHE is unknown, the fluorescence emission spectrum and lifetime components of this crystalline powder were remarkably similar to that shown for the new aqueous form (Fig. 2D). In summary, the fluorescence emission spectrum of the new, microcrystalline form of DHE appearing in 10 mM PIPES, pH 7.4, was distinct from that of monomeric DHE appearing exclusively in ethanol.
Stability of Crystalline DHE in Aqueous Media-The stabil- ity of DHE microcrystals in 10% fetal bovine serum-containing media (used to supplement and incorporate DHE into cultured cells over a period of 2-3 days) was determined to show if serum constituents in the cell culture medium could solubilize DHE microcrystals. Spectra of DHE in media at 37°C were collected for different time points over a period of 70 h, followed by calculation of emission peak ratios (426:355 and 426:373 nm) and comparison to those of DHE in ethanol and 10 mM PIPES, pH 7.4. At time t ϭ 0 (Table I) monomeric DHE emission peak ratios at 426:355 and 426:373 were very low, 0.42 Ϯ 0.01 and 0.298 Ϯ 0.004, respectively (n ϭ 6). This was in contrast to those of crystalline DHE in aqueous buffer, which were 10-fold higher, 4.8 Ϯ 0.1 and 3.4 Ϯ 0.2 (n ϭ 6) ( Table I). In media containing 10% fetal bovine serum at time 0, the ratios of the crystalline DHE peak at 426 nm to those of monomeric DHE (355 or 373 nm) were 2.91 Ϯ 0.07 and 2.81 Ϯ 0.05, respectively (n ϭ 6). These ratios were not significantly altered with increasing incubation time over 70 h (Table I). A similar finding has been reported for crystalline cholesterol in the presence of plasma (13).
Spectral Properties of DHE in Model Membranes, Effect of Sterol:Phospholipid Ratio-As indicated in the Introduction, it is thought that sterol phase separates at molar ratios of sterol: phospholipid Ͼ1.0. Yet relatively little is known regarding the exact nature of phase-separated sterol (transbilayer dimers, pure crystalline phase, and superlattices) in membranes. Furthermore, the methods (NMR and x-ray crystallography) used to detect the phase separation are not sensitive enough to detect the presence of low quantities/clusters/microdomains of phase-separated sterol in membranes. To resolve these issues, DHE was incorporated into model membranes (large unilamellar vesicles (LUV)) composed of varying molar ratios of DHE and 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC).
The fluorescence emission spectrum of DHE in LUVs composed of POPC:DHE at 65:35 (i.e. sterol:phospholipid molar ratio of 0.54) exhibited emission maxima near 355, 373, and 394 nm, as well as a small shoulder near 426 nm (Fig. 3A). These maxima were shifted from those observed in ethanol but not quite to the extent of those found in aqueous buffer. The ratios of DHE fluorescence emission peak intensities at 426: 355 and 426:373 were 0.54 and 0.33, respectively, slightly higher than for monomeric DHE in ethanol and suggesting the presence of a small amount of crystalline DHE (Table I). When the emission spectrum of DHE in ethanol was normalized to that of DHE (at 373 nm) in the LUV and subtracted, the difference spectrum (Fig. 3B) exhibited emission maxima characteristic of crystalline DHE in aqueous buffer (Fig. 2E). The spectra before and after subtraction were integrated to give values for the fluorescence emission that were corrected for the increased quantum yield exhibited by the crystalline DHE. These corrected values were used to obtain the approximate % of DHE in a crystalline arrangement which was very small (ϳ0.2%) for LUV composed of POPC:DHE at 65:35 (i.e. DHE: phospholipid molar ratio of 0.54) (Table I). It should be noted that in membranes composed of POPC:DHE at 65:35, the DHE undergoes significant self-quenching by Forster energy transfer (29). The fluorescence emission intensity was 63% lower than expected, based on comparison with that of DHE in membranes composed of POPC:DHE:cholesterol of 65:5:30 (Fig. 3A) and corrected for dilution. However, as expected from Forster TABLE I Relative proportions of crystalline/monomeric DHE in solvents and membranes DHE was 10 M in solvents and model membranes. Where indicated, values represent the mean Ϯ S.E., n ϭ 5-7. LUV refers to large unilamellar vesicle membranes. SUV refers to small unilamellar vesicle membranes. All membrane spectra of DHE emission were obtained in 10 mM PIPES, pH 7.4. Control L-cell fibroblasts were cultured with 10% fetal bovine serum medium containing 20 g/ml DHE, and subcellular organelles and lysosomes were isolated as described under "Experimental Procedures." ND, not determined. c DHE donor lysosomes (isolated from cells cultured with 20 g/ml DHE in the medium as in membranes) in 10 mM PIPES, pH 7.4. At time t ϭ 0 min SCP-2 (1.5 M) was added, and a DHE emission spectrum was obtained. Another DHE emission spectrum was obtained after 270 min of incubation at 37°C with 10-fold excess of acceptor lysosomes (containing no DHE). energy transfer, the shape of the DHE emission spectrum was unaffected. Consequently, the self-quenching did not account for the presence of the small percentage of crystalline DHE in these membranes.
The fluorescence emission spectrum of DHE in LUV composed of POPC:DHE at 40:60 (i.e. sterol:phospholipid molar ratio of 1.5) exhibited very similar emission maxima near 355, 373, and 395 nm, along with a more distinct shoulder near 426 nm (Fig. 3A) as compared with that at 0.54 DHE:phospholipid ratio. At 1.5 DHE:phospholipid molar ratio, the ratios of intensities of the DHE fluorescence emission peaks at 426:355 and 426:373 nm were 0.84 and 0.47, respectively (Table I). These ratios were significantly higher than those of monomeric DHE in ethanol, higher than for DHE in LUV with 0.54 molar ratio of sterol:phospholipid, and thereby suggested the presence of a small amount of crystalline DHE (Table I). Subtraction of the normalized emission spectrum of DHE in ethanol (normalized to that of 1.5 molar ratio DHE:phospholipid LUV at 373 nm) yielded a difference spectrum, which again exhibited emission maxima (Fig. 3B) characteristic of crystalline DHE in aqueous buffer (Fig. 2E). After correction for the higher quantum yield of crystalline DHE, this indicated that increasing the molar ratio of sterol:phospholipid by 3-fold (from 0.54 to 1.5) increased by 15-fold the amount of crystalline sterol to 7.7% (Table I). Further increment in DHE:phospholipid ratio to 2 and 3 increased the crystalline DHE content to 9.5% (Table I). Thus, emission spectral properties of DHE detected the presence of small amounts (1-10%) of crystalline sterol in LUV membranes composed of a broad range of sterol:phospholipid molar ratios (i.e. 0.54 -3.00).
Force-area Isotherms of Monolayers Formed from DHE or Cholesterol or Mixtures of the Two Sterols with POPC-In order to examine whether the DHE separation into crystalline form in membrane lipids reflects that of cholesterol, force-area isotherms of DHE and cholesterol in POPC monolayers were compared (Table II) at a surface compression of 35 mN/m, in the range of that typical of lipid bilayers (44,50). Increasing the mol % of either DHE or cholesterol in the monolayer reduced the mean molecular area (Å 2 ), due to the condensing effect of sterols on phospholipid membranes. At 35 mol % sterol, where no crystalline cholesterol was detected by x-ray crystallography and birefringence (51), the mean molecular area of the monolayer composed of DHE and POPC, 43.6 Å 2 , was not statistically different from that composed of cholesterol and POPC, 43.4 Å 2 (Table II). Even at 70 mol % sterol, where crystalline cholesterol was detected by x-ray crystallography and birefringence (51), the mean molecular area of the monolayer composed of DHE and POPC, 41.0 Å 2 , was not statistically different from that composed of cholesterol and POPC, 39.9 Å 2 (Table II). Finally, there was no statistical difference in mean molecular area occupied by pure DHE (40.40 Å 2 ) versus pure cholesterol (39.15 Å 2 ) (Table II). Because there was no difference in the condensing effects of DHE versus cholesterol in the POPC monolayers, these data are consistent with DHE behaving similarly to cholesterol with regard to formation of crystalline forms in monolayers at high mol % of sterol.  (52), and sterol in the inner leaflet is much more tightly packed than in the outer leaflet (22). To examine if this increase might induce the formation of crystalline phase sterol, SUVs with limiting radii of curvature were prepared as described under "Experimental Procedures." The radii of curvature in the SUV and LUV, measured by photon correlations spectroscopy, were 15 Ϯ 3 and 53 Ϯ 10 nm, respectively. The fluorescence emission spectrum of DHE in SUV (POPC:DHE of 65:35, i.e. sterol:phospholipid molar ratio of 0.54) exhibited emission maxima (Fig. 3C) indistinguishable from those in LUV with the same composition (Fig. 3A). The ratios of intensities of the DHE fluorescence emission peaks in the SUV at 426:355 and 426:373 nm were 0.69 and 0.43, respectively, slightly higher than for monomeric DHE in ethanol, thereby suggesting the presence of low levels of crystalline DHE in SUV (Table I). When the emission spectrum of DHE in ethanol was normalized to that of DHE (at 370 nm) in the SUV and subtracted, the difference spectrum ( Fig.  3D) exhibited emission maxima characteristic of crystalline DHE in aqueous buffer (Fig. 2E).
Because any DHE expelled from the membrane would form microcrystals in the buffer and give rise to spectra resembling Fig. 2E, the small unilamellar vesicles were subjected to a single pass extrusion using a 0.1-m membrane as performed for the aqueous solutions of DHE. The spectra obtained before and after the extrusion were exactly the same, decreasing the likelihood that the spectra (Fig. 2E) originates from extraneous DHE in the buffer. Also, an additional centrifugation run like the one used to sediment debris and multilamellar vesicles (sediments Ͼ85% of the microcrystals) did not produce any variability in the amount of the crystalline portion of the SUV DHE spectrum. Model membrane data from Table I indicated that with decreasing radius of curvature of the membrane the crystalline sterol increased 30-fold from ϳ0.2% in LUV to ϳ6% in SUV.
Spectral Properties of DHE in Plasma Membranes Isolated from L-cell Fibroblasts-The plasma membrane exhibits the highest molar ratio of sterol:phospholipid ratio in the cell, 0.5-1.0 (2). The fact that the inner (cytofacial) leaflet of the plasma membrane contains 80 -90% of plasma membrane sterol but only half of the plasma membrane phospholipid (2, 53 54) indicates molar ratios of sterol:phospholipid in the cytofacial leaflet as high as 1.8. However, it is not known if this sterol in the plasma membrane may be phase-separated into crystal-line sterol. Therefore, L-cells were cultured in the presence of DHE, and the DHE-enriched plasma membranes were isolated as described under "Experimental Procedures." Excitation and emission spectral intensities of DHE were higher in plasma membranes than any other cellular membrane fractions (Fig. 4A). Whereas emission spectra of DHE in plasma membranes (Fig. 4A) largely resembled those of DHE in model membranes (Fig. 3), after subtraction of the monomeric DHE (Fig. 4C), a small amount of crystalline DHE (about 5.4% of total) was detectable in the plasma membrane fraction (Table I). This was confirmed by examination of emission spectral peak ratios, all of which were greater than those of monomeric DHE (Table I). Detection of a small amount of crystalline sterol in these membranes was not due to adherence of DHE crystals to the membranes followed by coisolation on sucrose gradients. On the contrary, plasma membrane vesicles appeared at much higher density (27-30% sucrose) than did DHE crystals (14% sucrose). Furthermore, mixing of crystalline DHE with membranes not containing DHE, followed by reisolation on sucrose gradients, showed that there was no crosscontamination of the purified membranes with crystalline DHE.
Spectral Properties of DHE in Plasma Membrane Microdomains, Lipid Raft-enriched Subfractions from L-cell Fibroblasts-Plasma membrane sterol is laterally distributed into cholesterol-rich (lipid rafts and caveolae) and cholesterol-poor regions. Lipid rafts/caveolae account for about 10% of total plasma membrane sterol but consist of only 1% of plasma membrane surface (reviewed in Ref. 23). Thus, the small amount (i.e. 5.4%) of crystalline sterol in the plasma membrane may be highly enriched in these microdomains. To examine this possibility, DHE containing lipid raft/caveolae-enriched subfractions were isolated from the plasma membrane fraction by a nondetergent technique (see "Experimental Procedures"). The lipid raft/caveolar subfraction was enriched nearly 3-fold in caveolin-1 and flotillin-1 as compared with the parent plasma membrane fraction (Fig. 4B). The emission spectra of DHE in lipid raft/caveolar membrane subfraction (Fig. 4B) were similar to those of plasma membranes (Fig. 4A). DHE exhibited low intensities of crystalline DHE emission peaks at 403 and 426 nm (Fig. 4B), in the range of those observed for plasma membranes (Fig. 4A) and model membranes containing low molar ratios of DHE:phospholipid (Fig. 3A). Close examination of the DHE emission spectra in lipid raft/caveolar membrane subfraction after subtraction of monomeric DHE (Fig.  4C) showed Ͻ0.5% crystalline DHE, considerably less than in the plasma membrane. The very small and almost negligible presence of the crystalline DHE in the lipid raft/caveolar membrane subfraction was confirmed by examination of emission spectral peak ratios, all of which were only slightly greater than those of monomeric DHE (Table I).
Spectral Properties of DHE in Lysosomes-In order to examine the structural form of DHE in the intracellular organelle membrane containing the second highest cellular sterol phospholipid ratio, i.e. lysosomal membrane, it was necessary to first isolate the lysosome organelle. Lysosomes contained high quantities of crystalline sterol as follows. (i) The emission spectrum of DHE in isolated lysosomes exhibited large peaks near 426 and 403 nm (Fig. 4, D and E). (ii) The DHE emission spectral peak ratios at 426:355 and 426:373 were 2.02 Ϯ 0.15 and 1.42 Ϯ 0.06, respectively (n ϭ 3) (Table I). These ratios were much higher than for monomeric DHE in ethanol (Table  I) but were more similar to those of crystalline DHE in buffer or cell culture medium (Table I). (iii) Quantitative evaluation of the emission spectra showed that 75% of the fluorescence intensity was representative of the crystalline form. (iv) Lyso- a Pure cholesterol or dehydroergosterol monolayers, or mixed monolayers containing the indicated amount of sterol together with 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine, were compressed on 140 mM NaCl in water at 22°C (under an argon atmosphere in the dark) with a KSV surface barostat and proprietary KSV software. The barrier speed during compression did not exceed 3.4 Å 2 /molecule per min.
b Each value is average of 3 isotherms measured at 35 mN/m with an average variation Ͻ0.5 Å 2 . somes had very high ratios of sterol:phospholipid, consistent with the separation of DHE into a crystalline phase, crystallization of DHE in the lysosomal matrix, and/or incomplete solubilization/efflux of DHE crystals from the lysosomal matrix into the cell interior. The total DHE:phospholipid molar ratios in the lipids extracted from the lysosomes, determined from the extinction coefficient of DHE and by HPLC analysis, were 1.6 and 1.5, respectively.
Spectral Properties of DHE in Lysosomal Membranes-To determine whether the crystalline DHE detected in the lysosome (see above) was largely confined to the lysosomal matrix or actually a part of the lysosomal membrane, lysosomes were subjected to hypotonic lysis followed by an additional sucrose gradient purification to resolve DHE crystals (matrix-derived) from lysosomal membranes (see "Experimental Procedures"). The lysosomal membranes (density Ͼ30% sucrose) were clearly resolved from lysosomal matrix-derived crystals of DHE (density Ͼ16% sucrose). The molar ratio of total sterol:phospholipid (determined by HPLC of lipids extracted from purified lysosomal membranes) was 0.38, consistent with earlier reports (39). This molar ratio was more than 4-fold lower than that of intact lysosomes, confirming the presence of a large amount of crystalline DHE in the lysosomal matrix.
The emission spectral intensity of DHE in the lysosomal membrane fraction (not shown) was lower than that of DHE in plasma membranes but higher than that of DHE in endoplasmic reticulum and mitochondrial membranes (Fig. 4A). This reflected the relative sterol content of the lysosomal membranes versus plasma membranes, microsomes (endoplasmic reticulum), and mitochondrial membranes. Close examination of the DHE emission spectra in lysosomal membranes showed (after subtraction of monomeric DHE) a clearly resolvable crystalline DHE emission spectrum (Fig. 4C). The crystalline DHE present in the purified lysosomal membrane fraction represented about 6.9% of total sterol. The presence of a small amount of crystalline DHE in the lysosomal membranes was also substantiated by examination of emission spectral peak ratios, both of which were greater than those of monomeric DHE (Table I).
The possibility that the crystalline DHE in the lysosomal membrane was the result of DHE crystals (from the lysosomal matrix) sticking to the lysosomal membrane and cosedimenting at the same density as lysosomal membranes was considered as follows. (i) Crystalline DHE was mixed with isolated lysosomal membranes that did not contain DHE. (ii) The mixture was placed on the same sucrose gradient and the two fractions were clearly separated. Examination of emission spectra of the two fractions revealed that all the crystalline DHE appeared much higher in the gradient (density Ͻ14% sucrose) and was clearly separated from the much denser lysosomal membrane fraction (density Ͼ30% sucrose), which did not contain any DHE. Finally, it is important to note that the small amount of lysosomal membrane DHE found in the crystalline form was in the same range as that observed for DHE in model membranes wherein no crystalline DHE was present during the membrane preparation (Table I).
Spectral Properties of DHE in Microsomal Membrane and Mitochondrial Membrane Fractions Isolated from L-cell Fibroblasts-The ratio of sterol:phospholipid in microsomes (endoplasmic reticulum) and mitochondrial membranes, near 0.2 and Ͻ0.1, respectively, was much lower than that in plasma membranes or lysosomal membranes (reviewed in Refs. 32 and 39). The excitation and emission spectral intensities of DHE in the microsomal and mitochondrial membranes were both lower than in plasma membranes (Fig. 4A) or lysosomal membranes (not shown). The low intensities of DHE emission spectra in these membrane fractions (Fig. 4A) produced, upon subtraction of monomeric DHE, spectra with large signal to noise but indicated only small amounts of crystalline DHE, Ͻ1% of the total. (Fig. 4D), crystalline DHE can be converted to monomeric form both in vitro and in intact cells. Addition of ethanol converted the emission spectrum of DHE in the lysosomes from primarily crystalline to essentially monomeric (Fig. 4E). This was confirmed upon examination of the DHE emission peak ratios at 426:355 and 426:373 nm (Table I).

Disruption of Crystalline DHE in Lysosomes, Potential Role of Sterol Carrier Protein-2-Whereas crystalline DHE clearly accumulated in lysosomes of L-cells supplemented with crystalline DHE in the medium
The spontaneous transfer of DHE from lysosomal donors (enriched in crystalline DHE) to lysosomal acceptors (containing no DHE) was determined as the release from self-quenching and detected as increased DHE polarization (see "Experimental Procedures"). Spontaneous DHE transfer from lysosomes was relatively slow (Fig. 5A, solid circles) with an initial rate of molecular sterol transfer of 0.048 Ϯ 0.020 pmol/ min. However, upon addition of 1.5 M sterol carrier protein-2, the DHE fluorescence polarization and anisotropy increased rapidly, consistent with transfer of DHE from donor lysosomes to acceptor lysosomes (Fig. 5A, open circles). The initial rate of DHE transfer was enhanced nearly 13-fold, from 0.048 Ϯ 0.020 to 0.601 Ϯ 0.096 pmol/min. To determine whether SCP-2-mediated DHE transfer from the lysosomes disrupted the crystalline DHE or simply transferred DHE from the lysosomal membranes, emission spectra of DHE were obtained for lysosomes (isolated from L-cells supplemented with DHE) at the beginning (t ϭ 0 min, spectrum 1, Fig. 5B) and at the end (t ϭ 270 min, spectrum 2, Fig. 5B) of SCP-2-mediated DHE transfer from lysosomal donors (enriched in crystalline DHE) to lysosomal acceptors (containing no DHE). DHE emission spectral peak ratios (426:355 and 426:373 nm) decreased 3-4-fold to 0.54 and 0.36, respectively, after 270 min of SCP-2-mediated sterol exchange (Table I). These ratios resembled those observed for DHE in purified lysosomal membranes, 0.66 and 0.48, respectively (Table I). It should be noted that simply adding SCP-2 to DHE crystals in the intact lysosomes did not alter the DHE emission spectral properties. These data suggest that SCP-2 converted the crystalline DHE to the monomeric spectral form by quickly transferring the DHE from the crystalline DHE in the donor lysosomes to acceptor lysosomes.
In order to confirm further that sterol exchange from DHE crystals was facilitated by SCP-2, the anisotropy of the transfer of sterol between DHE crystals and lysosomal membrane acceptors was monitored. The experiment consisted of three steps involving donors of DHE crystals created by adding 2.5 g of DHE (similar to that seen in lysosomes) to 2 ml of 10 mM PIPES buffer. (i) Anisotropy of donor crystals was monitored for 3 h (the flat curve in Fig. 5C). (ii) A spontaneous transfer was performed between the donor crystals and lysosomal membrane acceptors (middle curve of Fig. 5C). (iii) 1.5 M SCP-2 was included and once again the anisotropy of donor crystal and lysosomal membrane acceptors was monitored for 3 h (the top curve of Fig. 5C). Initial rates were calculated by using a standard curve based upon previously published lysosomal membranes exchange data. Exchanges between crystals and lysosomal membrane acceptors showed that 1.5 M SCP-2 increased the initial rate of molecular sterol transfer nearly 7-fold, from 4.67 Ϯ 0.65 (spontaneous) to 30.73 Ϯ 5.2 pmol/min (with 1.5 M SCP-2). Clearly, anisotropy increased for spontaneous sterol transfer from both lysosomal and pure crystalline DHE donors to acceptor lysosomal membranes, an effect 13and 7-fold larger, respectively, in the presence of SCP-2.
Multiphoton Laser Scanning Microscopy of Crystalline DHE-The above differences in DHE fluorescence emission suggested that it might be possible to visualize simultaneously monomeric and crystalline DHE in living cells in real time through fluorescence microscopy. Unfortunately, the fact that DHE absorbs in the ultraviolet region over the range 275-345 nm (excitation maxima near 311, 324, and 340 nm, Fig. 1A) makes visualization by conventional and confocal fluorescence microscopy difficult because this requires the use of quartz optics, results in significant photobleaching and photodestruction, and yields images with limited resolution and excessive photobleaching over time (32,34). These difficulties are largely avoided by the use of multiphoton excitation with infrared light at 900 -925 nm resulting in simultaneous absorption of three infrared photons equivalent to single photon excitation in the ultraviolet range (i.e. 300 -308 nm). Furthermore, in multiphoton excitation only those DHE molecules within the objective focal volume (about 0.01 cu ) are excited, thereby allowing emission to be measured only from the DHE molecules excited in the focal volume. Through use of scanning optics and external detectors, images similar to those obtained by confocal microscopy are obtained but lacking in the drawbacks of the latter and exhibiting dramatically reduced photobleaching. These features of multiphoton excitation were therefore used to determine whether crystalline and monomeric forms of DHE could be discriminated by multiphoton laser scanning microscopy.
DHE was crystallized on coverslips by evaporation of carrier ethanol solvent under a stream of nitrogen followed by threephoton excitation at 900 nm (Fig. 6). To resolve crystalline and monomeric DHE, fluorescence emission was monitored simultaneously through separate dichroic filters as described under "Experimental Procedures": crystalline DHE, distinctly visible as needle-like structures (up to 50 m long and usually 1-10 m wide), was preferentially detected by monitoring emission with a 455:30-nm dichroic filter (Fig. 6A, green channel); monomeric DHE was preferentially detected with a 375:50-nm dichroic filter (Fig. 6B, blue channel). For the sake of imaging the crystals without saturation, the gain was decreased in the 455:30-nm channel to prevent saturation, whereas the gain in the 375:50-nm channel was maximal. The red channel (Fig. 6C, 575:150-nm dichroic filter) detected neither monomeric nor crystalline DHE and was used to establish whether DHE fluorescence emission spilled over into this wavelength range. Analysis of the merged image (comprised of A-C) revealed primarily green DHE crystals (Fig. 6D). Thus, multiphoton excitation at 900 nm was useful for determining the presence of crystalline DHE by MPLSM utilizing basically the same spectroscopic techniques as for the in vitro studies except that emission was collected simultaneously in the 455:30-and 375: 50-nm channels.
Multiphoton Laser Scanning Microscopy of Crystalline DHE in Lysosomes of Living L-cell Fibroblasts-As indicated above, DHE crystals were stable over many days in 10% fetal bovine cell culture medium incubated at 37°C. Because crystalline DHE was detected in lysosomes isolated by subfractionation of L-cells cultured with DHE crystals, consistent with the observation that L-cells actively phagocytose particles as large as several microns (55), the possibility that these crystals could be visualized in lysosomes by MPLSM was examined. L-cell lysosomes, ranging from 0.4 to 3.5 m in diameter, were readily visualized by staining intact cells with LysoTracker and monitoring LysoTracker emission with a 575:150-nm dichroic filter (Fig. 7, red pixels). Microcrystalline DHE was simultaneously visualized by monitoring emission through a 455:30-nm dichroic filter (blue pixels). Finally, some of the crystalline DHE taken up by the L-cells was solubilized/metabolized as indicated by the presence of monomeric DHE detected by monitoring emission with a 375:50-nm dichroic filter (Fig. 7, green  pixels). The merged image of all three photomultiplier tubes (Fig. 7) revealed the following.

FIG. 5. SCP-2-mediated DHE transfer from intact lysosomes disrupts crystalline DHE in vitro.
A, the transfer of DHE from lysosomal donors (enriched in crystalline DHE) to lysosomal acceptors (contain no DHE) was determined as described under "Experimental Procedures." In the absence of sterol carrier protein-2, the spontaneous transfer of DHE was indicated by slight change in the anisotropy over the time period of the experiment. However, upon addition of 1.5 M SCP-2, the DHE fluorescence anisotropy increased rapidly, consistent with transfer of DHE from donor lysosomes to acceptor lysosomes. B, emission spectra of DHE in lysosomal membranes (isolated from L-cells supplemented with DHE) at the beginning (curve 1, t ϭ 0 min, spectrum 1) and at the end (curve 2, t ϭ 270 min, spectrum 2) of sterol carrier protein-2-mediated DHE transfer from lysosomal donors (enriched in crystalline DHE) to lysosomal acceptors (contain no DHE). C, the bottom curve represents the anisotropy of 1.25 g/ml DHE in 10 mM PIPES buffer over 3 h. With the addition of lysosomal membrane acceptors, a near-linear increase in anisotropy occurs over the same period, indicating spontaneous exchange of DHE with cholesterol. Upon addition of 1.5 M SCP-2, the protein-mediated transfer of DHE as measured by a change in anisotropy follows an exponential rise over time.
Tracker (red)) were present in some lysosomes (Fig. 7, arrow 1). Second, distinct magenta pixilated areas (indicating colocalized crystalline DHE (blue), but not monomeric DHE, and Lyso-Tracker (red)) were present in some lysosomes (Fig. 7, arrow 2). Third, distinct yellow/orange pixilated areas (indicated colocalization of monomeric DHE (green), but little crystalline L-cell fibroblasts were cultured on chambered cover glass for 2-days in medium containing 10% fetal bovine serum and DHE (20 g/ml), washed 3-4 times with Puck's buffer, and incubated with LysoTracker Green as described under "Experimental Procedures." Excitation was at 925 nm. Red represents Lyso-Tracker detected through a dichroic filter (575:150 nm); green represents DHE detected through a dichroic filter (375:50 nm), and blue represents DHE detected through a dichroic filter (455:30 nm). Arrow 1 points to a representative white region where red, green, and blue colocalize (i.e. LysoTracker, monomeric DHE, crystalline DHE); arrow 2 points to a representative magenta region where red and blue colocalize (LysoTracker and crystalline DHE); arrow 3 points to a representative yellow pixilated area where green and red colocalize (monomeric DHE and LysoTracker); arrow 4 points to a representative red region where LysoTracker stains lysosomes containing neither monomeric nor crystalline cholesterol; arrow 5 points to a representative green region showing primarily monomeric DHE; arrow 6 points to a representative cyan region where green and blue colocalize (monomeric DHE and crystalline DHE). DHE, and LysoTracker (red)) were present in some lysosomes (Fig. 7, arrow 3). Fourth, distinct red pixilated areas (indicating LysoTracker staining regions without either monomeric or crystalline cholesterol) showed that L-cell fibroblasts had a substantial population of lysosomes that did not appear to contain significant amounts of either monomeric or crystalline DHE (Fig. 7, arrow 4). Fifth, distinct green (Fig. 7, arrow 5) and cyan (Fig. 7, arrow 6) pixels showed that some monomeric and crystalline/monomeric DHEs, respectively, were not present in the lysosomes.
Resolution of the Two Forms of DHE in Living L-cell Fibroblasts-Due to the complexity of visually distinguishing the different color tones necessary for deciphering the spatial colocalization of crystalline and monomeric forms, a procedure for separating the two forms utilizing the differences in spectral ratios was developed (see "Experimental Procedures").
First, the 455:30-nm channel and the 375:50-nm channel were pseudo-colored red and green, respectively, and merged (Fig. 8A). This was imported into the software program that plotted the fluorogram (Fig. 8B) as described under "Experimental Procedures." This fluorogram showed that there was a strong correlation in the more red (crystalline) pixels, which clearly have a higher ratio over the more green (monomeric) pixels and appeared in somewhat linear fashion along the ordinate axis. The more green pixels appeared in a more complex pattern due to the saturation of monomeric regions of localization, with a population aligned linearly along the horizontal axis and another population saturating along the right vertical axis. Some saturation was allowed in order to detect lower concentrations. As will be shown later, at least some of those saturated monomeric DHE regions in the pixelgram represented monomeric DHE in lipid droplets.
Second, the more red pixels (pixels with I red :I green Ն1) were selected by drawing a line along the diagonal and then enclosing a triangle around the upper left scattering of more red pixels. The resulting gray scale image was created (Fig. 8C) and then merged with the third or 575:150-nm channel (Lyso-Tracker emission, not shown). This merged image (Fig. 8E) showed the crystalline (green) colocalizing with LysoTracker Green (red) displayed as a fluorogram (Fig. 8F) and colocalization coefficients; C green ϭ 0.97 (crystalline DHE) exhibited a strong correlation with the LysoTracker Green, whereas C red ϭ 0.32 (LysoTracker Green) did not. This was corroborated by the image (Fig. 7) that showed many lysosomes colored in bright red and thereby evidencing the fact that they did not have any crystalline DHE.
Third, the more green pixels (pixels with I red :I green Ͻ1) were selected by drawing a triangle around the lower right scattering of more green pixels. This resulting image was created (Fig.  8D) and then merged with the 575:150-nm channel (Lyso-Tracker emission, not shown). The resulting new merged image showed monomeric DHE pseudo-colored green and Lyso-Tracker pseudo-colored red (Fig. 8G). The fluorogram of the merged image (Fig. 8H) indicated that the LysoTracker, C red ϭ 0.54, exhibited a slightly larger amount of colocalization with the monomeric DHE, C green ϭ 0.46. Both coefficients indicated additional separate localization of each probe in other cellular sites.
In summary, MPLSM allowed simultaneous three-photon excitation (crystalline and monomeric DHE) and two-photon excitation (LysoTracker Green) and separate, simultaneous detection of each in living cells in real time. This noninvasive technology took advantage of the enhanced emission properties of crystalline DHE to image multiple structural forms of sterol in living cells both within lysosomes and outside of the lysosomal compartment. It should be noted that the intensities of the crystalline DHE in L-cells were so bright that, to avoid saturating the photomultiplier tubes, it was not possible in these images to visualize simultaneously the DHE enriched in the plasma membrane. However, the latter was clearly imaged under other conditions (see below).
Multiphoton Laser Scanning Microscopy of DHE and BODIPY FL C 5 -ceramide, a Golgi Marker, in L-cell Fibroblasts-To determine whether crystalline DHE entered the Golgi compartment of L-cell fibroblasts, cells were incubated with BODIPY FL C 5 -ceramide (see "Experimental Procedures"). MPLSM at 900 nm excitation resulted in simultaneous three-photon excitation of DHE (monomeric and crystalline) and emission detected as above. Concomitantly, MPLSM at 900 nm resulted in simultaneous two-photon excitation of BODIPY FL C 5 -ceramide whose emission was detected through a dichroic filter (575:150 nm). When the images were treated as described above (not shown), the fluorogram of the Golgi marker versus crystalline DHE (not shown) indicated that only 9% of DHE in Golgi was crystalline. Thus, once the DHE left the lysosomal compartment it did not significantly accumulate in the Golgi in a crystalline form.
Multiphoton Laser Scanning Microscopy of DHE and Nile Red, a Lipid Droplet Marker, in L-cell Fibroblasts-To determine whether DHE was translocated for storage in intracellular lipid droplets of L-cell fibroblasts, the cells, as described under "Experimental Procedures," were cultured on chambered cover glass and incubated with Nile Red, a probe that accumulates primarily in neutral lipid droplets. Tuning the Ti: Sapphire laser to 925 nm resulted in simultaneous threephoton excitation of DHE (monomeric and crystalline) and two-photon excitation of Nile Red. The fluorescence emission of all fluorophores was simultaneously detected as follows: Nile Red with a 575:150-nm dichroic filter; crystalline and monomeric with a 455:35-and a 375:50-nm dichroic filter, respectively. Two levels of sensitivity and objectives were used to collect images as follows: (i) no attenuation of excitation power and a Zeiss 100ϫ Fluar oil immersion lens (high transmission characteristics for 350 -400 nm but decreased flatness of field); (ii) 80% attenuation of excitation power and a Zeiss 63ϫ Plan Apochromat oil immersion lens. To determine the degree of colocalization of the monomeric DHE with Nile Red in L-cell lipid droplets, a region of cells was selected showing little crystalline DHE; sensitivity was increased, and the images were merged (Fig. 9A). The blue/cyan regions showed the presence of very low crystalline DHE (Fig. 9A, arrow 1). Importantly, monomeric DHE was clearly visualized in plasma membranes and other membranes (Fig. 9A, arrow 2). The bright saturated yellow regions (Fig. 9A, arrow 3; upper right corner of fluorogram, Fig. 9B) were lipid droplets wherein monomeric DHE colocalized with the lipid droplet probe, Nile Red. The colocalization coefficient, C red ϭ 0.98 showed that almost all of the Nile Red colocalized with monomeric DHE, whereas C green ϭ 0.42 showed that just under half of the monomeric DHE intensity was colocalized with Nile Red in the lipid droplets (Fig. 9B). This implied that Ͼ50% of monomeric DHE intensity was in membranes, vesicles, and in diffuse distributions in the cytoplasm.
Some of the bright pixilated areas in the merged image were cyan (Fig. 9A, arrow 1), representing crystalline DHE colocalized with monomeric DHE. To obtain a more quantitative estimation of the distinction between crystalline and monomeric DHE with Nile Red, another image was taken at lower laser excitation power and with the 63ϫ objective centered upon cells containing more phagocytosed DHE crystals. Once again the images were merged under the same pseudo-colors as previously (Fig. 9C), where the blue regions represented crystalline DHE (Fig. 9C, arrow 1), green regions showed monomeric DHE (Fig. 9C, arrow 2), and the yellow regions represented monomeric DHE colocalized with Nile Red in lipid droplets (Fig. 9C,   arrow 3). Once again, an image was formed (not shown) composed of DHE alone: crystalline (red) and monomeric (green). Two distinct and well separated populations of pixels were obtained in this pixel fluorogram (Fig. 9D). The high slope, upper pixel population was composed of crystalline DHE, whereas the lower pixel population was composed of monomeric DHE. Each population was encircled, and a corresponding gray scale image was created as described under "Experimental Procedures." Each of these images was merged with the 575:150-nm channel (Nile Red in lipid droplets, not shown). The fluorograms resulting from the merged images of crystalline DHE (green) with Nile Red (red) and monomeric DHE (green) with Nile Red (red) were shown in Fig. 9, E and F, respectively. The coefficients of colocalization with respect to the crystalline form were quite small for both DHE and Nile Red: C red ϭ 0.01, C green ϭ 0.03 (Fig. 9E). However, the monomeric form colocalized quite well with Nile Red in lipid droplets. With this concentration of Nile Red, most of the Nile Red was observed in the bright lipid droplets, confirmed in the image (yellow round regions and no visible red areas), by the fluorogram, and the coefficient, C red ϭ 0.99 (Fig. 9F). The monomeric DHE, on the other hand, clearly resided in other areas illustrated by regions of green. At this excitation level, 74% of the 375:50 nm intensity integrated over the whole image colocalized to some extent with the Nile Red in lipid droplets, as seen by a coefficient of C green ϭ 0.74, whereas the other 25% was in other cellular structures (Fig. 9F). Thus, monomeric DHE was significantly colocalized with lipid droplets as well as other intracellular structures and the plasma membrane. Crystalline DHE was primarily associated with lysosomes.
Multiphoton Laser Scanning Microscopy of L-cell Fibroblasts Cultured with Large Unilamellar Membrane Vesicles Containing DHE-To determine whether L-cells cultured with noncrystalline DHE might exhibit the absence of crystalline DHE and/or a different intracellular DHE distribution, L-cell fibroblasts were cultured for 2 days with medium supplemented with DHE (20 g/ml) in the form of LUV composed of POPC: DHE (65:35) as described under "Experimental Procedures." As FIG. 9. Multiphoton laser scanning microscopy of DHE and Nile Red, a lipid droplet marker, in L-cell fibroblasts. L-cell fibroblasts were cultured on cover glass for 2-days in medium containing 10% fetal bovine serum and DHE (20 g/ml) as described under "Experimental Procedures." The cells were washed 3-4 times with Puck's buffer, followed by incubation with Nile Red as indicated under "Experimental Procedures." Excitation at 925 nm elicited simultaneous three-photon excitation of DHE (monomeric and crystalline) and Nile Red. Nile Red emission (red pixels in A and C) was detected through a 575:150-nm dichroic filter; crystalline DHE (blue pixels in A and C) was detected through a 455:30-nm dichroic filter; monomeric DHE (green pixels in A and C) was detected through a 375:50-nm dichroic filter. A shows the merged image obtained from high power excitation on a region of cells with small amounts crystals (arrow 1), plasma membrane (arrow 2) as well as other intracellular membranes, and lipid droplets (arrow 3). B is the fluorogram resulting from colocalization of the monomeric DHE (green) that was obtained using our software as described under "Experimental Procedures" and Nile Red (red). C is a merged image obtained from slightly lower power excitation of a region of cells showing larger amounts of crystals (arrow 1), organelles with DHE but with no Nile Red (arrow 2), and lipid droplets (arrow 3). D is the fluorogram of the merged images detected through the dichroic filters 455:30 (red) and 375:50 nm (green). The crystalline (upper reddish correlated pixels) and the monomeric forms (lower greenish correlated pixels) are clearly separated across the diagonal. As described under "Experimental Procedures," each clustering of pixels was selected to produce individual gray scale images representative of crystalline and monomeric DHE. E is the fluorogram determined by merging the crystalline portion with Nile Red, and F is the fluorogram obtained from merging the monomeric portion with Nile Red. Clearly, the correlation coefficients confirm that the majority of Nile Red emission from lipid droplets colocalizes with monomeric DHE. shown in Fig. 3, DHE in these LUVs was primarily in monomeric form with Ͻ0.2% in crystalline form. After the cells were washed 3-4 times with Puck's buffer and incubated with Nile Red (see "Experimental Procedures"), all fluorophores were excited at 920 nm by MPLSM. Fluorescence emission of all fluorophores was simultaneously collected by the Zeiss 63ϫ oil immersion objective and directed to three separate external photomultiplier tubes: Nile Red (red pixels) with a 575:150-nm dichroic filter; crystalline DHE (blue pixels), with a 455:35-nm dichroic filter; and monomeric DHE (green pixels) with a 375: 50-nm dichroic filter. To show the intracellular distribution of DHE (in cells labeled with LUV DHE and with Nile Red), the simultaneously acquired images from the three photomultiplier tubes were merged (Fig. 10A). It is clear from this merged image that supplementing the L-cells with DHE in the form of LUV (primarily noncrystalline DHE) resulted in a somewhat different intracellular distribution of DHE as compared with that detected in cells supplemented with crystalline DHE (Fig. 9).
The plasma membrane was prominent in the merged image (Fig. 10A, arrow 1) and contained primarily monomeric DHE (green pixels, detected through a 375:50-nm dichroic filter). An examination of the intracellular regions in the merged image (Fig. 10A) showed that, within the perinuclear region of some cells, a small amount of crystalline DHE (blue/magenta, arrow 2) colocalizing with Nile Red was visible. As seen previously, monomeric DHE colocalized with Nile Red (yellow/orange, arrow 3) in lipid droplets.
Once again, to help separate the forms of DHE, a merged image (not shown) composed of the 375:50-(green) and 455: 30-nm (red) channel was created and imported into the newly developed software program. The resultant DHE fluorogram is shown in Fig. 10B. The pixels having a I red :I green Ͻ1 and I red :I green Ն1 were selected to produce gray scale images. Both of these images were merged with the Nile Red gray scale image. The merged image of monomeric DHE with Nile Red was shown in Fig. 10C with its corresponding fluorogram (Fig.  10D). With a C red ϭ 0.96, most of the Nile Red colocalized with monomeric DHE even though it was not all in lipid droplets in this figure (Fig. 10D). A higher concentration of the Nile Red probe was used in order to delineate other nonpolar lipid regions within the cell. Again, the DHE in monomeric form was distributed among cellular components, but Ͼ50% of the total intensity of monomeric DHE colocalized to some degree with Nile Red staining regions (including lipid droplets) as indicated by C green ϭ 0.64 (Fig. 10D). Whereas the brightest regions were the lipid droplets, the perinuclear regions also appeared to accumulate significant colocalized Nile Red and monomeric cholesterol.
As indicated above, supplementation of L-cells with DHE in liposomal form resulted in the near absence of crystalline DHE, with only a few crystalline pixels detected colocalizing with Nile Red in the perinuclear region (not shown). With a C red ϭ 0.04, almost none of the Nile Red colocalized with crystalline DHE.

DISCUSSION
Because of its influence on membrane fluidity, permeability, cell-cell recognition, transport, receptor-effector coupling, and microdomain (e.g. caveolae and rafts) function, the cholesterol content of cellular membranes must be tightly regulated (reviewed in Refs. 2 and 53). This is especially important in view of the fact that abnormal regulation of membrane cholesterol content and/or distribution impairs membrane function (reviewed in Refs. 2, 32, 53, and 56 -60) and thereby cell survival as indicated by cytotoxicity, sickle cell acanthacytosis, Niemann-Pick C disease, Alzheimer's disease, and atherosclerosis (reviewed in Refs. [12][13][14][15][16]. Despite these findings relatively little is known regarding the organization of membrane cholesterol, especially when cholesterol is present in molar excess over membrane phospholipid. Likewise, almost nothing is known regarding the real time, direct visualization of different structural forms of cholesterol in membranes or cells (reviewed in Refs. 18 and 25). The data presented here address these issues and provide the following new insights.
First, crystalline and monomeric forms of sterol were differentiated spectroscopically. Monomeric DHE exhibited emission maxima near 354, 370, and 394 nm in anhydrous ethanol, whereas those of crystalline DHE were significantly red-shifted to 356, 403, and 426 nm. These emission characteristics of crystalline DHE were not due to the presence of impurities. The onset of the red-shift was concomitant with the formation of DHE microcrystals in aqueous buffers (27,30) at concentrations similar to that where cholesterol has been shown to form microcrystals in aqueous buffers (49,(61)(62)(63). Several techniques were applied to determine whether the change in spectral characteristics from ethanol to water was due to formation of microcrystals or of micelles. Observation by polarizing light microscopy revealed birefringent structures resembling those reported to be cholesterol monohydrate microcrystals (60 -63) and are therefore presumed to be dehydroergosterol monohydrate crystals. Filtration up to 0.5 m eliminated all detectable emission, indicating that the fluorescence is derived from an aggregate with a lower size limit of Ͼ0.5 m, most likely the observed microcrystals. In corroboration, the addition of DPH to M concentrations of DHE in aqueous buffer did not reveal any incorporation into micelles of DHE, regardless of the possible occurrence of micellar agglomeration. These results compared favorably with the same experiments applied to cholesterol at similar concentrations and conditions.
Although determination of the exact photophysical nature of crystalline DHE involves a complex discussion of exciton theory (64), one can consider a simple interaction mechanism with a nearest neighbor in the form of a physical sandwich dimer, because steroid fused rings form ␣-face-to-␣-face dimers as well as ␤-face-to-␤-face dimers in crystals (65). For example, the crystalline absorption spectrum exhibited characteristics such as a bathochromatic shift accompanied by spectral broadening, some increase in intensity of higher order vibrational bands, and overall hypochromism. The crystalline spectral emission exhibited a red shift in the emission maxima (354 to 356 nm and 370 to 375 nm) whose intensities were decreased from those in the monomer, whereas the lower energy transitions (exhibiting maxima at 403 and 426 nm) were increased dramatically in intensity. This was indicative of enhanced Franck-Condon factors and fast relaxation times into the lowest energy level where there must be less nonradiative processes and stronger wave function overlap with higher vibrational levels in the ground state. A larger lifetime component consistent with a smaller nonradiative transfer rate was observed in the crystalline form (30). The possibility that the enhanced emission could also be due to the formation of excited state dimers (excimers) was considered based on the fact that the conjugated triene double bonds in DHE are part of a planar, ring structure. Such structures, when stacked face-to-face, allow the overlap of the electron clouds of the two molecules and thereby facilitate excimeric interaction. Typically, excimer formation is concentrationdependent and fluoresces at longer wavelengths than monomers as well as exhibits higher quantum yield than the monomer. However, excimeric emission is usually broad and structureless because the excimer dissociates due to the strong repulsive force between the molecules in the ground state. As can be seen in Fig. 1D, vibrational structural features were present in the subtracted spectrum. Regardless of whether the DHE formed physical dimers and/or excimers, these differences in the spectral emissive characteristics of the monomeric and crystalline DHE allowed the determination of their relative proportions in solvents, model membranes, biological membranes, and organelles (e.g. lysosomes).
Second, these differences in the spectral emission characteristics of the monomeric and crystalline DHE permitted determination of their relative proportions in model membranes. The structural form of DHE in model membranes was primarily monomeric even at molar ratios of sterol:phospholipid substantially Ͼ1. This observation was surprising in view of previous studies on sterol packing in model membranes. While at low mol % the sterol is uniformly dispersed/solubilized in the membrane, and increasing the sterol:phospholipid molar ratio from 0.25 to 1 results in formation of interdigitated, transbilayer sterol dimers (66). Over the same concentration, DHE self-quenches in model membranes by Forster energy transfer (29,30). The Forster energy transfer distance for this homotransfer between face-to-face oriented DHE molecules is 13.3 Å (30). Because sterol and phosphatidylcholine are 6 (13) and 8 Å (30) thick, respectively, Forster homotransfer of energy is efficient even if the DHE molecules are separated by one phospholipid molecule. In contrast, at sterol:phospholipid molar ratios Ͼ1 an immiscible sterol phase forms in model membranes (reviewed in Refs. 13, 67, and 68). Based on x-ray and NMR data, it had previously been assumed that this immiscible sterol phase is identical to pure crystalline sterol monohydrate (reviewed in Refs. 13, 67, and 68). If massive formation of pure crystalline DHE had occurred in membranes with DHE:phospholipid molar ratios Ͼ1, the concomitant formation of face-toface DHE dimers would have resulted in preponderant emission at longer wavelength. On the contrary, the data presented here with DHE clearly showed that even at sterol:phospholipid molar ratios Ͼ1 there was only a small amount of phaseseparated DHE with properties identical to that in pure DHE crystals. The molecular basis for this apparent contradiction between results obtained by the various techniques (x-ray and NMR versus fluorescence) may be explained as follows.
One possibility is that the different techniques detecting phase-separated immiscible sterol may not necessarily report on the same aspects of sterol structural packing in membranes.
Although the x-ray and NMR techniques largely report on transbilayer thickness/distance between sterol molecules, it is unclear whether these techniques differentiate the multiple potential forms of lateral phase-separated sterol in membranes: pure cholesterol patches, ribbons of sterol, and/or sterol-sterol with different orientations (edge-to-edge, ␣-face-to-␣face, ␤-face-to-␤-face, and ␣-face-to-␤-face), etc. (reviewed in Refs. 65, 69, and 70). In contrast, DHE emission characteristics detect the formation of face-to-face lateral interactions between sterols, which emit with greater quantum yield and at higher wavelength. As shown here over the DHE:phospholipid molar ratio ranging from 0.5 to 3, DHE detected only small amounts (1-10%) of immiscible DHE whose structure (identified by longer wavelength emission maxima) appeared identical to pure crystalline DHE or other face-to-face sterol packings.
An alternative explanation of the different results obtained by the x-ray and NMR of cholesterol versus fluorescence of DHE is that the phase behavior of DHE may differ markedly from that of cholesterol in model membranes. Although it is important to recognize that DHE and cholesterol are not identical, this latter possibility was considered unlikely based on the very similar behavior of these sterols in a variety of model membrane systems (reviewed in Refs. 26, 30, 71, and 72) summarized briefly as follows. (i) In pure and mixed monolayers at pressures similar to bilayer systems, cholesterol and DHE showed similar packing arrangements even at mol % where cholesterol has been shown by x-ray crystallography and birefringence to separate into crystalline structures (51). (ii) The dissociation constants of spontaneous desorption of cholesterol and DHE from membranes were 0.044 and 0.031 mol/mol, respectively (30). In contrast, due to the apparent higher affinity of cyclodextrin for DHE, the cyclodextrin-catalyzed extraction of cholesterol from model membranes was severalfold slower than that of cyclodextrin-catalyzed extraction of DHE (73). (iii) The rate constants of spontaneous transfer of cholesterol and DHE between POPC containing model membranes did not significantly differ (71). (iv) The fractional distribution of cholesterol into kinetically resolvable sterol domains was not significantly different from that of DHE in POPC containing model membranes (71). Third, it was demonstrated that crystalline and monomeric membrane sterol could be differentiated spectroscopically in vitro in subcellular membrane fractions isolated from L-cell fibroblasts. Plasma membranes, lysosomal membranes, microsomes (endoplasmic reticulum), and mitochondrial membranes differ markedly in sterol:phospholipid ratio ranging from near 1.0 in plasma membranes to as low as 0.04 in mitochondria (reviewed in Refs. 2 and 53). When these membrane fractions were isolated from L-cell fibroblasts supplemented with DHE, only a small proportion (1-7%) of crystalline DHE was detected. This was confirmed by MPLSM of L-cells supplemented with DHE in monomeric form (i.e. LUV) and represents the first real time imaging of a fluorescent sterol in the plasma membrane of a living cell as well as the first real time resolution of multiple structural forms of DHE in the plasma membrane of a living cell. Consistent with the findings of only a small amount of crystalline sterol in the plasma membrane, other investigators (18) using x-ray crystallography also showed that plasma membranes isolated from control smooth muscle cells did not contain significant levels of crystalline sterol.
The lack of substantial quantities of crystalline sterol in plasma membranes, the most highly sterol-enriched membrane in the cell, was especially interesting because the transbilayer distribution of sterol in the plasma membrane is asymmetric, i.e. 4-fold higher in the cytofacial leaflet (reviewed in Refs. 22 and 58). Because the L-cell plasma membranes in this study exhibited a sterol:phospholipid molar ratio of 1.2 (38), this would imply that the sterol:phospholipid molar ratio in the cytofacial leaflet was almost 2.0. Nevertheless, neither DHE in isolated plasma membranes in vitro, x-ray crystallography in vitro, nor DHE in real time multiphoton images of living cells detected very much crystalline sterol in plasma membranes. It should be noted that these observations for the structure of DHE in biological membranes were not due to DHE perturbing biological membrane structure or function because of the following: DHE is a natural product found in high percentage in the membranes of eukaryotes such as sponge and yeast; DHE can be supplemented to microorganisms and cultured L-cells to replace nearly 90% of endogenous sterol; DHE has no adverse effect on sterol:phospholipid ratio or phospholipid composition, fatty acid composition, cell growth, or cholesterol-sensitive enzymes; DHE codistributes with endogenous sterol among intracellular membranes; and DHE does not alter the structure of the plasma membrane or the function of sterol-sensitive membrane proteins (reviewed in Refs. 26 and 31).
Fourth, the detection of crystalline DHE in the isolated plasma membrane fraction suggested that this small amount of crystalline sterol might represent microdomains present in small amounts in the plasma membrane. It is known that the lateral distribution of sterol in the plane of the bilayer is not uniform because cholesterol-rich lipid rafts/caveolae have been detected in the cell surface and isolated by several methods (reviewed in Ref. 23). Caveolae represent only 1-2% of the cell surface membrane area, but lipid raft/caveolar membrane-enriched fractions exhibit severalfold higher sterol content than the plasma membrane (reviewed in Ref. 23). If the small amount of crystalline DHE detected in L-cell plasma membranes was due to its presence in lipid raft/caveolar membranes, then an enrichment of severalfold might be predicted therein with DHE:phospholipid ratios near 2-3. However, model membranes with such DHE:phospholipid ratios contained 7.1-9.5% crystalline DHE. Surprisingly, the level of crystalline DHE in isolated lipid raft/caveolar membranes was Ͻ0.3%, 18-and up to 32-fold lower than in the bulk plasma membrane and model membranes, respectively (Table I). These observations, showing that lipid rafts/caveolar membranes contain very little crystalline sterol despite their high sterol content, may explain for the first time why so many processes (signaling, transport, etc.) important to cell viability readily function in these domains. The future challenge will be to determine what features of lipid raft/caveolar proteins and/or lipids prevent crystalline sterol formation.
Fifth, MPLSM for the first time resolved the uptake and intracellular distribution of crystalline sterol in real time in living cells. L-cell fibroblasts ingested large amounts of crystalline DHE that accumulated in lysosomes as evidenced by colocalization with LysoTracker Green, a lysosomal stain, in living cells. The lysosomal localization of substantial amounts of crystalline DHE was confirmed by examination of the spectral properties of DHE in lysosomes isolated from L-cells cultured with crystalline DHE. In addition, some crystalline DHE was also detected outside the lysosomal compartment. However, very little extralysosomal crystalline DHE colocalized with BODIPY FL C 5 -ceramide (Golgi marker) or Nile Red (lipid droplet marker) in living cells. Consistent with this observation, crystalline cholesterol has been detected in macrophage foam cells, both within lysosomes and outside (as crystals surrounded by a membrane) the lysosomal compartment (13,14,19). The phagocytosis of DHE crystals was consistent with the known ability of L-cells to phagocytose particles as large as several microns and by the fact that L-cells endocytose the equivalent of their entire cell surface membrane within about 2 h (55,74), similar to the activity of macrophages (14,19).
Sixth, MPLSM for the first time resolved the uptake and intracellular distribution of monomeric sterol in real time in living cells. The pattern of monomeric sterol distribution depended on the mode of DHE supplementation to the cells. (i) Although most cells incubated with crystalline DHE exhibited high amounts of crystalline form in lysosomes, those cells, which had digested/solubilized the ingested crystals, distributed monomeric DHE throughout the cell. The highest concentrations of monomers were in lipid droplets, followed by perinuclear regions, the plasma membrane, and punctate areas resembling vesicles in the cytoplasm (Fig. 10). (ii) In contrast, when cells were incubated with DHE in monomeric form (i.e. vesicles composed of POPC:DHE, 65:35) almost no crystalline form was detected in the cells. Instead highest concentrations of monomeric DHE were found in the plasma membrane and lipid droplets, followed by lower levels in the perinuclear region. Much less monomeric DHE was present in a punctate vesicular pattern throughout the cytoplasm. Thus, the use of multiphoton laser scanning microscopy together with the spectral differences in monomeric versus crystalline DHE allowed for the first time the noninvasive, real time imaging of multiple forms of sterol in living cells.
In summary, the data presented herein demonstrate that the spectral properties of DHE, together with multiphoton laser scanning microscopy, form a powerful tool to noninvasively resolve and visualize the individual dynamics of multiple structural forms of sterol in real time in living cells. Crystalline DHE was enriched within lysosomes but not plasma membranes or lipid raft/caveolar membrane-enriched subfractions. Interestingly, the presence of sterol carrier protein-2 in vitro significantly enhanced the sterol transfer of crystalline DHE (in donor lysosomes) to acceptor lysosomes (containing no DHE) as evidenced by the change in spectral characteristics of the DHE emission from crystalline to monomeric. These data for the first time showed a potential protective role for sterol carrier protein-2 in mitigating the deleterious effects of crystalline DHE in the cell. Consistent with this possibility, the cellular level of sterol carrier protein-2 was up-regulated as much as 3-fold in macrophage foam cells (7,75) wherein crystalline cholesterol is known to accumulate and become cytotoxic (13,14,19). These noninvasive, nonperturbing methods also demonstrated for the first time that multiple structural forms of sterol normally occurred within cell membranes and could, for the first time, be visualized real time in membranes and intracellular organelles of living cells. These findings are especially important because relatively few noninvasive, nonperturbing techniques exist for real time visualization of cholesterol structures in biological membranes or in living cells (reviewed in Refs. 18 and 25).