g-Cyclodextrins Greatly Enhance Translocation of Hydrophobic Fluorescent Phospholipids from Vesicles to Cells in Culture IMPORTANCE OF MOLECULAR HYDROPHOBICITY IN PHOSPHOLIPID TRAFFICKING STUDIES*

Short-chain, fluorescent derivatives are commonly used to investigate intracellular phospholipid trafficking. However, their use can yield misleading results because they, unlike the native species, can rapidly distribute between organelles due to their low hydrophobicity. On the other hand, hydrophobic derivatives are very difficult to introduce to cells and thus have hardly been used. Here we show that carboxyethylated g-cyclodextrin (CE-g-CD) greatly enhances transfer of a variety of hydrophobic fluorescent phospholipid derivatives from vesicles to cultured cells. Several lines of evidence indicate that CE-g-CD enhances transfer of lipid molecules by increasing their effective concentration in the aqueous phase, rather than by inducing membrane fusion or hemifusion. Incubation with CE-g-CD and donor lipid vesicles does not extract cholesterol or phospholipids from the cells or compromise plasma membrane intactness or long term cell viability. Using CE-g-CD-mediated transfer, we introduced hydrophobic pyrene-labeled phosphatidylserine to the plasma membrane of fibroblast cells and followed their distribution with time. In contrast to what has been previously observed for other, less hydrophobic species, transport of this lipid to the Golgi apparatus or mitochondria was not detected. Rather, much of this fluorescent PS remained in the plasma membrane or was incorporated to various endocytotic compartments. These findings indicate that the native, typically hydrophobic phosphatidylserine molecules efflux only very slowly via the cytoplasm to intracellular organelles. This helps to explain how cells can maintain a very high concentration of phosphatidylserine in the inner leaflet of their plasma membrane. Furthermore, the present results underline the importance of using hydrophobic analogues when studying intracellular trafficking of many phospholipid classes.

In eukaryotic cells, most phospholipids are synthesized in the endoplasmic reticulum but are abundant in all organelle membranes (1). Accordingly, efficient transport of those phospholipids from the endoplasmic reticulum to the other organelles must take place. There are several mechanisms that could accomplish this transfer, i.e. vesicle traffic, spontaneous or protein-assisted diffusion via the cytoplasm, or diffusion via membrane fusion (or hemifusion) sites (reviewed in Refs. 2 and 3). However, the involvement and relative contributions of these various transport mechanisms have not been established. It is also unknown how intracellular trafficking of phospholipids is controlled so that the different membranes can maintain their distinct lipid compositions. A major reason for the lack of this crucial information is largely a technical one: there are no simple and efficient methods to trace movements and distributions of phospholipids inside the cell. For instance, immunolocalization, which has played the key role in resolving the transport mechanism and sorting of proteins, cannot be applied, as (i) phospholipids are generally poor antigens, (ii) most phospholipids are abundant in all membranes and thus there is no adequate concentration gradient, and (iii) it is not possible to obtain stoichiometric complex formation due to the large size of the antibody molecule.
To circumvent the tracing problem, short-chain fluorescent (NBD 6 ) lipids were introduced to this field by Pagano and co-workers (reviewed in Refs. 4 and 5). Although such derivatives have proven to be very useful to study sphingolipid trafficking (6,7), they have been less successful when glycerophospholipid transport is being investigated. One reason for this is that short chain NBD 6 -phospholipids are rapidly degraded (half-life is 1 h or less) in mammalian cells at the physiological temperature (8 -10). The altered conformation of such lipids (11,12) may render them targets for some cellular phospholipases. However, an even more serious limitation is that the short chain lipids, unlike the natural ones, can distribute spontaneously between all accessible membrane surfaces due to their hydrophilicity (8). Thus, their use is very problematic when one studies intracellular trafficking of lipids having access to the cytoplasmic leaflet of organelle membranes, which is the case with most glycerophospholipids. Accordingly, fluorescent derivatives with hydrophobicity similar to that of endogenous phospholipids need to be employed to obtain relevant information on intracellular trafficking of the latter. Unfortunately, such derivatives are difficult to introduce into cells just because of their hydrophobicity. Although several methods allowing introduction of hydrophobic fluorescent lipids to cells have been reported (13), none of them has gained significant popularity, for various reasons.
Recently, ␤-cyclodextrin derivatives were shown to mediate efficient transfer of cholesterol between lipid vesicles and cells in culture (14,15). This led us to examine whether some cyclodextrins could be used to introduce hydrophobic fluorescent phospholipid derivatives into cultured cells. Primarily pyrenelabeled phospholipids were employed in these studies because several systematically constructed sets were available (16,17), which helps to draw conclusions regarding the mechanism of lipid transfer (18). The pyrenyl derivatives are also attractive because they, in all respects studied so far, behave similarly to natural lipids (16,17,19) and should allow one to map microscopically lipid concentration gradients in cellular membranes (20). Additional studies were carried out with Bodipy-labeled derivatives, because they offer some advantages over the pyrene derivatives.

EXPERIMENTAL PROCEDURES
Lipids and Other Reagents-All unlabeled lipids were obtained from Avanti Polar Lipids (Alabaster, AL). Bodipy 12 -PC, 1 -PE, and -SM, as well as NBD 12 PC, were supplied by Molecular Probes Europe (Leiden, Netherlands). Bodipy-PC and -PE were repurified shortly before use by HPLC on silica gel and Bodipy-SM on a reverse phase column. Typically, such repurification greatly reduced labeling of cells in the absence of cyclodextrin by removing free fatty acid and/or other impurities. The pyrenylacyl glycerophospholipids were synthesized as described previously (21,22). Pyrenyldecanoyl sphingomyelin was synthesized using the method of Via et al. (23) and purified by HPLC on a silica column using a chloroform/methanol gradient and then further purified on an octadecylsilica column. [ 3 H]Cholesterol oleate was synthesized as described previously (24). All labeled lipids were at least 99.5% pure as determined by HPLC. The cyclodextrins were obtained from Cyclolab (Budapest, Hungary). The cell culture media were obtained from Life Technologies, Inc., and all other chemicals were from Sigma.
Assays for Cyclodextrin-mediated Lipid Transfer and Binding-The previously described assays (16) were employed to study the transfer and binding of pyrenyl lipids. The donor vesicles consisted of a pyrenyl lipid, POPC, and TNP-PE (0.2:45:5 nmol), whereas the acceptor vesicles consisted of POPC and POPA (480:20 nmol). Transfer of the pyrenyl lipid molecules from quenched donor vesicles to acceptor vesicles results in de-quenching of pyrene fluorescence, which is recorded. The initial slope of the progress curve is used as the measure of transfer rate. To measure binding, donor vesicles were titrated with a cyclodextrin solution, and the pyrene fluorescence intensity was recorded. Fluorescence measurements were carried out using either a Hitachi F-4000 or PTI QuantaMaster spectrofluorometer equipped with a thermostatted cuvette holder. The excitation and emission wavelengths were 345 and 395 nm for pyrene and 480 and 520 nm for Bodipy, respectively. All measurements were carried out at 25°C.
Cell Culture-Normal Human fibroblasts (GM08333) and BHK-21 cells were grown as before (18,25). For microscopy, the cells were plated on round 32-mm coverslips placed in home-built aluminum-Teflon chambers. Before use, the coverslips had been cleaned by treating them for 1 h at 50°C with 1 M NaOH and 1 M HCl each, followed by washing with water and ethanol. After 1-2 h of incubation at 37°C, the inoculum was removed, and the attached cells were washed, covered with normal growth medium, and placed in the incubator. Labeling and imaging were carried out the following day. This cultivation protocol considerably reduced the amount of cell debris on the coverslips.
Incubation of cells with cyclodextrins and donor vesicles-Cell monolayers were washed twice with CO 2 -independent minimal medium (I-MEM), twice with phosphate-buffered saline and then incubated at room temperature or 37°C in I-MEM containing the donor vesicles with or without cyclodextrin. The donor vesicle compositions are specified under "Results." The CE-␥-CD stock solution was prepared in I-MEM and adjusted to pH 7.4 to avoid acidification of the incubation medium. This adjustment somewhat reduced the efficiency of labeling but was necessary to avoid harmful effects to cells, particularly at higher CE-␥-CD concentrations. After the incubation, the cells were washed three times with I-MEM and then subjected to imaging or lipid extraction.
For the latter, the cells were scraped into phosphate-buffered saline, washed twice by centrifugation, and then extracted according to Ref. 26. The amount of cell-associated fluorescent lipid was determined by measuring the pyrene (excimer) or Bodipy fluorescence intensity of the extract. Extracts of nonincubated cells were used as blanks. Alternatively, the amount of cell-associated fluorescent lipid was determined by HPLC analysis using on-line fluorescence detection (27). For quantification, an internal standard (a Bodipy or pyrene lipid, as appropriate) was added to the cell pellet before the extraction. To determine vesicle binding to cells, [ 3 H]cholesterol oleate (50,000 cpm/dish) was included in the donor vesicles.
Fluorescence Imaging-After washing, the cells on coverslips were covered with I-MEM containing glucose, glucose-oxidase, and catalase to deplete oxygen from the medium (28). Oxygen depletion virtually eliminates photobleaching of the pyrene chromophore but has no detectable effect on fluorophore distribution or cell morphology (25). Imaging was carried out on a Zeiss Axiovert 10 microscope equipped with a cooled CCD camera as detailed (25). For pyrene lipids, a 340 nm (BP, 11 nm) excitation filter, a 480 nm (BP, 80 nm) emission filter, and a 375 nm dichroic mirror were used. The images were corrected for background fluorescence and scatter by subtracting a background image obtained with a 360 nm (BP, 5 nm) excitation filter and multiplied with an empirically determined factor. As will be shown elsewhere, 2 pyrene is not significantly excited with the 360 nm filter, whereas the background fluorescence/scatter is excited similarly with the 360 and 340 nm filters. A particularly useful feature of this type of background correction is that it properly accounts for the typically uneven distribution of cellular autofluorescence. With the Bodipy lipids 480 nm (BP, 30 nm) and 535 nm (BP, 40 nm) excitation and emission filters were used, respectively.
Other Methods-Lactate dehydrogenase was determined essentially as described (29). Concentrations of unlabeled phospholipids were determined by a phosphate assay (30) and protein concentrations by a fluorescamine assay (31).
CE-␥-CD-mediated Transfer Is Strongly Dependent on Pyrene Lipid Hydrophobicity-To obtain information on the mechanism of cyclodextrin-mediated phospholipid transfer, the effect of the length of the pyrene-labeled acyl chain (6 -14 carbons) on the rate of CE-␥-CD mediated transfer was investigated. As displayed in Fig. 2, the rate decreases strongly (exponentially) with increasing length of the pyrene-labeled chain. Analogous results were obtained for lipids that contained two pyrene-labeled chains of identical length (n ϭ 6 -14; data not shown, but see below). Such strong chain length dependence of the rate of fluorescence increase (probe dilution) clearly indicates that CE-␥-CD enhances monomeric transport of the labeled lipid molecules, rather than causes fusion or hemifusion between the donor and acceptor vesicles. If either of the latter mechanisms were to dominate, the rate of fluorescence increase would be independent of the chain length of the pyrene phospholipid, which clearly is not the case. In addition, the fact that hardly any increase of fluorescence was observed for the pyrene PC with longest chain (see Fig. 2) excludes the possibility that transfer of the quencher molecules, if occurring, would significantly contribute to the observed enhancement of pyrene fluorescence in this assay.
To study the mechanism of CE-␥-CD mediated pyrene phospholipid transfer, quenched donor vesicles containing Pyr 10 PC were titrated with CE-␥-CD, and the pyrene emission spectrum was recorded. The spectral intensity increased strongly upon addition of CE-␥-CD, and the increase was proportional to the amount of CE-␥-CD added (data not shown). Parallel results were obtained for the other Pyr n PCs and DiPyr n PCs tested. These results are analogous to those obtained previously for phospholipid carrier proteins (16,17) and therefore strongly suggest that CE-␥-CD forms of soluble complexes with pyrene phospholipids in the aqueous phase, thereby enhancing their translocation between vesicles (see under "Discussion").
Effect of the Lipid Head Group and Backbone Structure on the Rate of CE-␥-CD-mediated Transport-We also studied the effect of the head group on the rate of cyclodextrin-enhanced transfer of pyrene-labeled phospholipids (Fig. 3). Among glycerophospholipids, transfer of phosphatidylglycerol, PE, PS, and PC is similarly enhanced, whereas that of phosphatidylinositol is enhanced about twice as much by 5 mM CE-␥-CD. This more rapid transfer of phosphatidylinositol is probably due to the considerable polarity of the head group. Supporting this, transfer of diglyceride, which has a very small polar moiety, is much less efficient than that of the phospholipids. Notably, Pyr 10 SM is transported almost 3 times more rapidly than Pyr 10 PC. The additional hydroxyl group and double bond in the sphingosine moiety obviously makes SM somewhat less hydrophobic than PC with similar acyl chains. We conclude that the overall molecular hydrophobicity, rather than some structural details, determines the rate of CE-␥-CD-mediated transport of the pyrenyl lipids.

CE-␥-CD Also Greatly Enhances Transfer of Hydrophobic Fluorescent Lipids from Vesicles to Cultured Cells-Having
shown that CE-␥-CD mediates efficient transport of fluorescent phospholipids between phospholipid vesicles in vitro, we next studied whether these compounds also catalyze transport of fluorescent phospholipids from vesicles to cells in culture. To this end, vesicles containing a pyrene-or Bodipy-labeled phospholipid were incubated with cell monolayers at room temperature or 37°C, and after washing, the amount of cell-associated fluorescent lipid was determined using either fluorescence imaging or spectroscopy.
For imaging studies, we employed dipyrenyl phospholipids, i.e. species containing two pyrene-labeled chains, because such species display strong excimer fluorescence (peak at 480 nm) due to frequent interpyrene collisions (32). This excimer fluorescence is more readily visualized with our imaging system than the monomer fluorescence (peaks at 378 and 395 nm) that the monopyrenyl lipids mainly emit. Furthermore, the wide separation of the excitation (345 nm) and emission peaks avoids excitation/emission crossover, as well as allowing efficient correction for background fluorescence as described under "Experimental Procedures." Yet another important advantage of the dipyrenyl derivatives is that the (possible) presence of fatty acid and/or lysolipid impurities in the donor vesicles does not interfere with interpretation of the images, because these compounds do not display detectable excimer fluorescence under the conditions used. Presence of even minor amounts (Ͻ1%) of fluorescent fatty acid or lysolipid could lead to artifactual results due to their very rapid spontaneous transfer (33) as compared with the intact phospholipid. Such selective imaging of the intact phospholipid is obviously not possible with the monopyrenyl or Bodipy derivatives. Fig. 4 shows human fibroblast cells incubated for 15 min at 37°C with donor vesicles containing DiPyr 10 PC in the presence or absence of 30 mM CE-␥-CD. Clearly visible excimer fluorescence, mainly on the plasma membrane, was observed in the presence of CE-␥-CD (Fig. 4a), whereas practically none was detectable in its absence (Fig. 4b). Parallel results were obtained for DiPyr 10 PS (Fig. 4, c and d). Very similar labeling efficiencies were obtained when the incubation was carried at 25°C for 30 min, whereas less efficient or marginal labeling was observed when the incubation was carried out for 30 min at 18 or 10°C, respectively (data not shown). CE-␥-CD also strongly enhanced transfer of the Bodipy-derivatives from the donor vesicles to cells, as shown for Bodipy-SM in Fig. 4, e and f. However, significant labeling of cells by Bodipy-SM (and other Bodipy derivatives) occurred even in the absence of cyclodextrin, apparently because of the relatively low hydrophobicity of the Bodipy derivatives (see below). This is not clearly shown by Fig. 4f because of the lower concentration of Bodipy 12 -SM donor vesicles used (see legend to Fig. 4).
We then carried out analogous labeling experiments with other DiPyr n PC species in order to establish the limits of the method, as well as to confirm that the enhanced labeling in the presence of cyclodextrin is due to monomeric transfer of the fluorescent lipid rather than due to fusion or adherence/endocytosis of the donor vesicles. The results are summarized in Fig. 5. As expected, most efficient CE-␥-CD-mediated labeling was obtained with DiPyr 6 PC (Fig. 5a). The efficiency decreased systematically with increasing length of the acyl chains, and with DiPyr 14 PC, virtually no labeling was observed (Fig. 5i). For DiPyr 6 PC and DiPyr 8 PC, significant labeling was also ob-  (i and j). In a, c, e, g, and i, CE-␥-CD (15 mM) was present, whereas it was omitted in b, d, f, h, and j. The camera settings and image adjustments were identical in each case. The brightly fluorescent structures often seen to attach to cells or matrix, in both the presence and the absence of cyclodextrin (white arrows in a and e), consist of cell debris. Presumably, debris may become strongly labeled either because the Pyr n PC monomers more readily integrate to the less tightly packed (due to degradation) membranes and/or because donor vesicles readily fuse with the membrane remnants of the debris. served in the absence of CE-␥-CD (Fig. 5, b and d), but not for the long chain derivatives (Fig. 5, f, h, and j). Notably, the fact that practically no labeling with the most hydrophobic pyrene lipids was observed in the presence of CE-␥-CD provides strong evidence that CE-␥-CD enhances monomeric transfer of the pyrene derivatives, rather than inducing vesicle-cell fusion, adherence, or endocytotic uptake of vesicles. These data also show that DiPyr 10 PC is the most hydrophobic dipyrenyl derivative for which adequate levels of labeling were achieved under the present labeling conditions.
Although the dipyrenyl lipids used here were of high purity, the fraction of the long-chain dipyrenyl lipids, such as DiPyr 10 PC, transferred from the donor vesicles to cells is rather small (see below). Therefore, it was considered necessary to exclude the possibility that some, or even all, of the cell-associated excimer fluorescence would derive from pyrene fatty acid/lysolipid impurities rather than from intact DiPyr 10 PC. To accomplish this, human fibroblast cells were incubated with donor vesicles containing up to 1 nmol of Pyr 10fatty acid (which corresponds to 5% fatty acid impurity in DiPyr 10 PC, i.e. far more than could be present; see under "Experimental Procedures") in the presence of 15 mM CE-␥-CD. No pyrene excimer fluorescence could be detected in the cells (data not shown), thus demonstrating that the pyrene fluorescence observed in cells incubated with DiPyr 10 PC and CE-␥-CD derives from intact lipids rather than from possible impurities.
To confirm that CE-␥-CD indeed mediates monomeric transfer of the fluorescent lipids rather than causing fusion or association of the donor vesicles with the cells, BHK cell monolayers were incubated with donor vesicles containing a fluorescent lipid and [ 3 H]cholesterol oleate, a nontransferable liposomal marker, in the presence of 0 -30 mM CE-␥-CD for 30 min at room temperature. Fig. 6a shows results for Bodipy-SM. The amount of cell-associated Bodipy-SM increased strongly with the CE-␥-CD concentration up to 15 mM CE-␥-CD. At this concentration, about 5% of the liposomal Bodipy-SM has become cell-associated. In contrast, only about 0.3% of the [ 3 H]cholesterol ester, the liposomal marker, was associated with the cells independent of the CE-␥-CD concentration used. Parallel results were obtained for DiPyr 6 PC except that somewhat less fluorescent PC was transferred to the cells (Fig. 6b). These data are consistent with the proposition that CE-␥-CD enhances monomeric transport of the fluorescent phospholipids from donor vesicles to cells. We also carried out analogous experiments with DiPyr 10 PC but could not obtain conclusive data, probably because the fraction of this (much more hydrophobic) lipid (see below) transferred to cells is quite small and comparable to the fraction of vesicles adhering to cells or cellassociated debris. Nevertheless, the imaging data in Fig. 5 clearly indicate that CE-␥-CD indeed mediates monomeric transfer of DiPyr 10 PC as well.
Spontaneous Transfer of Long Chain Pyrenyl Lipids Is Similar to That of Typical Natural Species-To compare the spontaneous transfer rate of the fluorescent analogues with that of natural phospholipid species, we used a simple reverse-phase HPLC assay. Previous studies have shown that there is a close correlation between the rate of spontaneous transfer of a lipid and its retention time on a reverse-phase column (34). Although such a correlation is strictly valid only for members of a homologue series, reasonable estimates should be obtained using this method. Accordingly, we determined the retention times for different fluorescent PCs as well as for some common natural species and calculated the relative rates of spontaneous transfer. The results are shown in Table I. Notably, the predicted rates for Pyr 10 PC and DiPyr 10 PC are similar to that of 16:0/18:1 and significantly slower than that of two other common natural species, 16:0/20:4-PC and 18:0/20:4-PC. The rate predicted for DiPyr 8 PC is higher than that for 18:2/18: 2-PC but lower than that for 18:3/18:3-PC. In contrast, DiPyr 4 PC, DiPyr 6 PC, Bodipy 12 -PC, and NBD 12 -PC are predicted to transfer far more rapidly than any of these natural species.
To estimate the validity of these predictions, we determined the rate of Pyr 10 PC directly using the assay shown in Fig. 1. The half-time of transfer was found to be 74 h at 25°C as determined from a progress curve recorded over 48 h. This agrees well with the value of 36 h obtained previously for Pyr 10 PC at 37°C (35). In comparison, a half time of 48 h has been obtained for 16:0/18:1-PC at 37°C (36). Thus, the rate of spontaneous transfer of Pyr 10 PC (and probably that of DiPyr 10 PC as well) is indeed similar to that of 16:0/18:1-PC, as was predicted based on the retention time data (Table I).
Incubation with CE-␥-CD Does Not Compromise Plasma Membrane Intactness or Cell Viability-To study whether the labeling protocol perturbs the integrity of the plasma mem- To investigate the effect of the labeling procedure on longterm cell viability, sparse (approximately 15% confluent) BHK cell monolayers were subjected to labeling for 30 min at 25°C and then incubated for several days in the normal growth medium. In cyclodextrin-treated cultures, the amount of cell protein increased from the prelabeling value of 0.4 Ϯ 0.1 mg to 3.5 Ϯ 0.3 mg and in control cultures to 3.4 Ϯ 0.4 mg (n ϭ 6). Hence, the labeling procedure does not seem to affect the long term viability of the cells either.
CE-␥-CD Does Not Seem to Extract Phospholipids or Cholesterol from the Cells-To study whether incubation of with CE-␥-CD and donor vesicles alters the phospholipid composition of the cells, mass spectroscopic measurements (37,38) of the medium and the cells were carried out. Because it is likely that phospholipids in the outer leaflet of the plasma membrane would be preferentially extracted, the analyses were focused on PC, SM, and PE, which together represent nearly 100% of the outer leaflet phospholipids in BHK cells (39). Unfortunately, the phospholipid composition of the labeling medium could not be determined because of some compounds present interfered with the mass spectroscopic analysis. Analysis of the cells did not reveal significant differences between those labeled and controls, except that the former contained somewhat more 16: 0/18:1-PC. However, the increase of 16:0/18:1-PC was quite variable, and further experiments (data not shown) indicated that this extra 16:0/18:1-PC is not truly incorporated to cell membranes, but probably derives from donor vesicles adhering to the cells or cell debris. Notably, the analysis of the total cellular lipid composition do not exclude the possibility that some changes in the lipid composition of the plasma membrane outer monolayer in fact occurs, because this compartment contains only a fairly small fraction (13%) (39) of the total cellular phospholipid.
Previous studies have shown that ␥-cyclodextrin does not extract phospholipids efficiently from erythrocytes, whereas the ␣-derivative does (40). The larger hydrophobic cavity of ␥-cyclodextrins apparently favors complex formation with fluorescent lipids having bulky groups in the acyl chain(s) but disfavors complexation of phospholipids with natural acyl chains.
Certain cyclodextrins extract cholesterol efficiently from cells (14,40), which may compromise cell viability or alter cellular metabolism (41,42). To study whether such cholesterol depletion also occurs under present conditions, we incubated BHK cells with DiPyr 10 PC containing donor vesicles in the presence or absence of 30 mM CE-␥-CD for 30 min at 25°C and then determined the cholesterol content of the cells. The cells incubated with CE-␥-CD contained 22 Ϯ 3.4 nmol of cholesterol/dish (n ϭ 3), whereas the controls contained 24 Ϯ 5. nmol/ dish (n ϭ 4). Hence, the labeling procedure does not cause significant depletion of cellular cholesterol. This, together with the lack of significant extraction of phospholipids, probably explains why the intactness of the cell membrane is maintained under the labeling condition employed.
Fluorescent Lipids Are Deposited Initially to the Outer Leaflet of the Plasma Membrane-To confirm that the CE-␥-CDmediated cell labeling is indeed due to (initial) incorporation of the fluorescent lipid monomers to the outer leaflet of the cells, we labeled BHK cells with DiPyr 8 PC and -PS as above and then followed the distribution of these lipids during a chase at 37°C using fluorescence imaging. DiPyr 8 -derivatives were used here because they can distribute more readily among accessible membranes than the much more hydrophobic DiPyr 10 -derivatives (see below). Fig. 7a shows BHK cells labeled with DiPyr 8 PC and then chased for 5 h at 37°C. As compared with situation before the chase (Fig. 5c), the plasma membrane fluorescence was strongly diminished, whereas there was a marked increase in the labeling of intracellular structures, including punctate pericellular and more diffuse perinuclear structures, which may represent secondary endosomes and the endosytic recycling compartment, respectively (43). On the other hand, no obvious labeling of the mitochondria or the nuclear membrane was apparent. Parallel results were obtained with DiPyr 6 PC (see below) and with DiPyr 4 PC previously (25). These findings indicate that dipyrenyl PCs are initially introduced as monomers to the outer leaflet of the plasma membrane and later incorporate to endosomes and related organelles. If fusion of the vesicles with cell membrane had occurred, labeling of the mitochondria and the nuclear membrane would be expected (see below and cf. Ref. 4).
In contrast, when BHK cells were labeled with DiPyr 8 PS for 15 min at 37°C and then chased for 5 h at 37°C, prominent labeling of mitochondria was observed (Fig. 7b). Labeling of these organelles is in agreement with previous data obtained for NBD 6 PS (44), NBD 12 PS (45), and DiPyr 4 PS (25) and can be explained as follows. The PS molecules are initially introduced to the outer leaflet of the plasma membrane, but they move rapidly to the inner leaflet with the assistance of the aminophospholipid translocase (46). From the inner leaflet, the fluorescent PS molecules move further, possibly via spontaneous diffusion (see below), to various organelles, including mitochondria, where they can be decarboxylated to PE (45,47). We conclude that the intracellular distribution modes of DiPyr 8 PC and -PS are compatible with the proposition that CE-␥-CD mediates monomeric incorporation of the fluorescent lipids initially to the outer leaflet of the plasma membrane, rather than inducing fusion or endocytosis of the donor vesicles.
Hydrophobic Dipyrenyl PS Is Not Transported Efficiently from the Plasma Membrane to Mitochondria-Because DiPyr 8 PS and other fluorescent derivatives used previously (see above) are much less hydrophobic (cf. Table I) than typical native PS species (mostly 16:0/18:1 in BHK cells) (18), it was of interest to determine whether more hydrophobic dipyrenyl PS species, also, would be transported from plasma membrane to mitochondria. DiPyr 10 PS, the predicted spontaneous transfer rate of which is more than 10 times slower than that of DiPyr 8 PS (Table I), was introduced to BHK cells, and the cells were chased in normal medium for up to 23 h. The results are summarized in Fig. 8. After a 30-min chase, the plasma membrane was the most prominently labeled structure, as expected (Fig. 8a), but some labeling of intracellular structures was also obvious. After 3 h (Fig. 8b), the juxtanuclear region, possibly the endosomal recycling compartment (cf. Ref. 43) was prominently labeled, along with the plasma membrane. At 5 h (Fig.  8c), probe distribution remained similar, except that juxtanuclear labeling was somewhat diminished, and punctate fluorescence, presumably representing secondary endosomes, was observed. After 23 h of chase (Fig. 8d), the overall fluorescence intensity was somewhat diminished, and in most cells, the punctate structures were more prominent than before. Notably, in none of these images is there any indication that DiPyr 10 PS would be present in mitochondria. This is in striking contrast with what was observed here for DiPyr 8 PS (Fig.  7b) and previously for DiPyr 4 PS and NBD 12 PS (25,45), thus strongly suggesting that native plasma membrane PS, probably consisting mainly of hydrophobic species (see under "Discussion"), moves only very slowly, if at all, to these organelles. DISCUSSION

CE-␥-CD Mediates Efficient Intermembrane Transfer of Hydrophobic Fluorescent Phospholipids-
The present study demonstrates that CE-␥-CD greatly accelerates translocation of hydrophobic pyrene-labeled phospholipids between vesicles, as well as from vesicles to cells (Fig. 4). Also, the transfer of Bodipy 12 -and NBD 12 -phospholipids was remarkably enhanced by CE-␥-CD, thus demonstrating that we are not dealing with a phenomenon unique for the pyrene derivatives. Several lines of evidence (see under "Results") indicate that CE-␥-CD enhances monomeric transfer of phospholipid molecules, rather than inducing membrane fusion or hemifusion. It is not clear from the present experiments what the precise mechanism of this monomeric transfer is. One obvious possibility is that CE-␥-CD forms soluble complexes with the lipid molecule once it has fully or partially effluxed from the donor membrane surface and then carries the complexed lipid to the acceptor membrane. Several CE-␥-CD molecules in turn could act as carriers (a "relay" mechanism). The alternative, perhaps less probable mechanism is that CE-␥-CD increases the concentration of free phospholipid monomers in the aqueous phase by forming transient complexes close to the donor surface. Previous studies have established that the rate of (spontaneous) transfer of phospholipids between membranes is proportional to the free monomer concentration (48 -50).
Intracellular Distribution of Hydrophobic Pyrene PS Indicates Slow Efflux of Native Plasma Membrane PS Species to Intracellular Organelles-PS makes up as much as one-third of the phospholipids in the inner leaflet of the mammalian cell plasma membrane (39). A high concentration of PS may be required to support various crucial processes, such as fusion of exosome with the plasma membrane or the protein kinase C reaction. It is totally unclear how the cell can maintain so high a concentration of PS in the inner leaflet of the plasma membrane. One possibility is that the plasma membrane PS consists predominantly of saturated, hydrophobic species, as indicated by the study of Keenan and Morre (51). Enrichment of hydrophobic species in the plasma membrane could result from selective depletion of the hydrophilic species due their facile transport to mitochondria and decarboxylation therein, after the synthesis in the endoplasmic reticulum (18). In addition, hydrophobic PS species could have a higher affinity for lipid "rafts" that are targeted to the plasma membrane (52). Once they have reached the plasma membrane, they would be prevented from "leaking" to other organelles due to their very slow spontaneous diffusion (see below). On the other hand, the hydrophobic plasma membrane PS molecules are likely to incorporate into endosytic vesicles and subsequently in other endosomal compartments. Our present results strongly support this model by showing that the hydrophobic fluorescent derivative, DiPyr 10 PS, is present in the plasma membrane and in various endocytic compartments but apparently not in the Golgi apparatus or mitochondria (Fig. 8). However, further studies are needed to determine whether DiPyr 10 PS correctly reports on intracellular trafficking of natural plasma membrane PS species.
Previously, NBD 12 PS has been shown to move quite rapidly from the plasma membrane to the Golgi apparatus and mitochondria (45). Because the authors considered this fluorescent species as being "relatively nonexchangeable," and because inhibition of vesicular transport had no effect on labeling of the Golgi apparatus and mitochondria, they proposed that the transport of NBD 12 PS from the plasma membrane to these organelles is probably mediated by a lipid transfer protein. As shown by Table I, NBD 12 PS is far less hydrophobic than typical native species and can probably therefore move from the plasma membrane to mitochondria by spontaneous diffusion via the cytoplasm. Therefore, it is not necessary to invoke an involvement of a transfer protein.
These results discussed above underline the importance of considering molecular hydrophobicity when studying intracellular trafficking of PS and other lipids having access to the cytoplasmic leaflet of organelle membranes. Hydrophobicity is likely to be a relevant parameter also in other contexts, as indicated by differential sorting of short and long chain fluorescent lipid derivatives within the endosomal compartments (43).