Introduction of phospholipids to cultured cells with cyclodextrin.

Previous studies indicate that methyl-β-cyclodextrin (meβ-CD) can greatly enhance translocation of long-chain phospholipids from vesicles to cells in culture, which is very useful when studying, e.g., phospholipid metabolism and trafficking. However, the parameters affecting the transfer have not been systematically studied. Therefore, we studied the relevant parameters including meβ-CD and vesicle concentration, incubation time, phospholipid structure, and cell type. Because meβ-CD can extract cholesterol and other lipids from cells, thereby potentially altering cell growth or viability, these issues were studied as well. The results show that efficient incorporation of phospholipid species with hydrophobicity similar to that of natural species can be obtained without significantly compromising cell growth or viability. Cellular content of phosphatidyl-serine, -ethanolamine, and -choline could be increased dramatically, i.e., 400, 125, and 25%, respectively. Depletion of cellular cholesterol could be prevented or alleviated by inclusion of the proper amount of cholesterol in the donor vesicles. In summary, meβ-CD mediates efficient transfer of long-chain (phospho) lipids from vesicles to cells without significantly compromising their growth or viability. This lays a basis for detailed studies of phospholipid metabolism and trafficking as well as enables extensive manipulation of cellular phospholipid composition, which is particularly useful when investigating mechanisms underlying phospholipid homeostasis.


Assessment of cell viability
The cells were incubated in DMEM containing donor vesicles and me ␤ -CD for 1 h on 12-well plates, washed three times with DMEM, and incubated in 500 µl of DMEM for 2 h. The cells were then washed once with DMEM and incubated in 500 µl DMEM containing 10% of Alamar Blue for 3 h at 37°C. Then 200 µl of the medium was moved to a 96-well plate and fl uorescence intensity was measured using a Cary Eclipse fl uorescence spectrophotometer (Varian) with the excitation and emission wavelengths set to 530 and 590 nm, respectively. Control samples included: i ) cells incubated in DMEM, ii ) cells killed with 70% ETOH, iii ) cells killed with 5 mM CuSO4, and iv ) DMEM. To determine the number of cells remaining on the dishes, the cells were detached with Trypsin-EDTA and counted with a hemocytometer.
To assess cell intactness based on Trypan Blue exclusion, the cells were detached with Trypsin-EDTA and centrifuged at 500 g for 5 min. Trypan Blue (0.4%) in PBS was then added and the stained cells were counted using a hemocytometer.

Quantifi cation of DNA
Cells were scraped into 0.25 M sucrose, pelleted by centrifugation for 10 min at 1500 g and lysed by incubating for 5 min at 100°C in a buffer consisting of 0.1% SDS, 10 mM Tris-HCl, 1 mM EDTA; pH 8.0. DNA was quantifi ed using the Quant-iT dsDNA Broad-Range Assay Kit (Molecular Probes) according to manufacturer's instructions.

Other assays
Published methods were used to quantify cholesterol ( 22 ) and protein ( 23 ).
We next studied the effect of donor vesicle concentration ( Fig. 1B ). BHK cells were incubated for 1 h in the presence of 8 mM me ␤ -CD and varying concentrations (0.75-3 mM) of donor vesicles of the composition indicated above. Transfer of each PS species to cells increased approximately linearly with donor vesicle concentration and at the donor total lipid concentration of 3 mM, the amount of 14:1/14:1-, 16:1/16:1-, and 18:1/18:1-PS transferred to cells corresponded to 32 ± 3, 37 ± 2, and 10 ± 1% (n = 6) of endogenous PS, respectively. Together, the Invitrogen; cholesterol and other unlabeled lipids from Avanti Polar Lipids (Alabaster, AL), D 9 -choline, D 3 15 N-serine and D 4 -ethanolamine from CIL (Andover, MA); and methyl-␤ -cyclodextrin (me ␤ -CD), phospholipase D ( Streptomyces sp .), hydroxylamine, methyl arachidonyl fl uorophosphonate (MAFP), and Trypan Blue from Sigma. All solvents (HPLC-grade) were obtained from Merck. The phosphatidylethanolamine (PE) and phosphatidylserine (PS) species with a deuterium-labeled head group were synthesized from corresponding phosphatidylcholine (PC) species and D 4 -ethanolamine or D 3 15 N-serine, respectively, using phospholipase D-mediated transphosphatidylation as described previously for unlabeled lipids ( 14 ), except that the reaction volume was reduced 5-fold. The D 9 -PC species were synthesized from corresponding PEs by methylation with D 3 -methyl iodide ( 15 ). The lipids were purifi ed by normal-phase HPLC ( 16 ). Their purity was confi rmed by mass spectrometry and their concentrations were determined using phosphate analysis ( 17 ).

Introduction of exogenous phospholipid to cells
Cells grown to ‫ف‬ 80% confl uency on 6 cm dishes were washed once with DMEM. small unilamellar vesicles, and me ␤ -CD in 2 ml of DMEM were added and the cells were incubated at 37 ° C. After washing three times with DMEM, the cells were chased in DMEM for 0-24 h, washed with PBS, scraped into 0.25M sucrose, and moved to silanized screw-cap tubes. The lipids were then extracted ( 19 ). A cocktail of internal standards was added at the one-phase stage of extraction. After evaporation of the solvents, the lipids were dissolved in methanol/chloroform (2:1, v/v) and stored at Ϫ 20 ° C. A trace amount (0.1 mol%) of 22:0/22:0 PC, which is too hydrophobic to be transferred by me ␤ -CD, was included in the donor vesicles to estimate their adherence to cells. Only negligible amounts of 22:0/22:0-PC were detected in cells after the incubation, thus indicating that adhered donor vesicles do not signifi cantly contribute to cell-associated exogenous lipid.

Mass spectrometry and data analysis
After addition of aqueous NH 3 (4%), the sample was infused at 6 l/min to the ESI source of a Micromass Quattro Micro triplequadrupole mass spectrometer (Waters) operated as previously ( 20 ). The heavy isotope-labeled D 9 -PC, D 4 -PE, and D 3 15 N-PS species and unlabeled PC, PE, PS, SM, and phosphatidylinositol species were selectively detected using head group-specifi c precursor or neutral-loss scanning ( 3,5 ). The species were identifi ed and quantifi ed using the LIMSA software ( 21 ). neous and me ␤ -CD-mediated transfer, is very sensitive to vesicles size ( 24 ). Also the composition of the donor vesicles, which infl uences lipid packing, can have a signifi cant effect on the rate of transfer (see below).

Effect of the polar head group
The effect of the phospholipid polar head group on transfer of exogenous phospholipids was studied by incubating BHK cells for 1 h with donor vesicles (0.5 mM) containing cholesterol, POPC, and labeled 14:0/14:0-PC, -PE, or -PS (10:9:1 mol/mol) in the presence of 4 or 8 mM me ␤ -CD ( Fig. 2 ). Effi cient transfer of all three lipids was observed, albeit incorporation of PE was signifi cantly lower than that of PS or PC. However, no consistent head group dependency was observed when incorporation of di-14:1, di-14:0-, and di-18:1-PE or -PS species to different cell types was investigated (see below). Tests with four different molecular species of PC, PE, or PS (i.e., 14:1/14:1, 14:0/16:0, 16:0/18:1, and 18:0/20:4) in HeLa cells indicated that me ␤ -CDmediated transfer of phospholipids does not depend significantly on the structure of the head group, but rather on the hydrophobicity (i.e., length and unsaturation) of the acyl chains (data not shown; however, see below). These data are consistent with previous studies indicating that differences in cyclodextrin-mediated transfer of PLs with a different head group but identical acyl chains can be explained by their different propensities to effl ux from the bilayer ( 10 ).

Effect of donor vesicle composition
We next studied the effect of donor vesicle composition on me ␤ -CD-mediated transfer of phospholipids by incubating BHK cells with vesicles composed of 50 mol% cholesterol and varying proportions of POPC and deuteriated di-14:1-PC, -PE, or -PS. Transfer of these lipids to cells increased in proportion to their concentration in the donor vesicles, at least up to 25 mol% (data not shown; however, see Fig. 4 ). Vesicles with higher content of these lipids were not studied, because extensive reduction of the POPC content would compromise the incorporation of high amounts of cholesterol in the donor vesicles ( 25 ) and thus probably lead to depletion of cellular cholesterol. exogenous PS species thus comprised ‫ف‬ 79% of total PS in the cells.
We next investigated the effect of incubation time on transfer of PS to cells using donor vesicles (1.5 mM) of the same composition as above and 8 mM me ␤ -CD ( Fig. 1C ). Again, the transfer of each of the three PS species to cells increased nearly linearly with time, and at 2 h, the amount of exogenous 14:1/14:1-, 16:1/16:1-, and 18:1/18:1-PS transferred to cells was 28 ± 6, 43 ± 6, and 14 ± 1% (n = 6) relative to endogenous PS, respectively. Together the exogenous PS species thus comprised ‫ف‬ 85% of total PS in the cells.
Some variability in transfer of phospholipids to cells under seemingly identical conditions was observed (cf. Fig.  1A-C ). The reasons for this are unclear, but could relate to minor unavoidable variations in the average size of the donor vesicles between experiments. It is known that lipid effl ux from a bilayer, which directly correlates with sponta-

Loading of cells with exogenous phospholipids
When studying, e.g., phospholipid homeostasis, it would be highly useful if one could signifi cantly increase the concentration of a particular lipid species or class in cells. Thus, to determine how much exogenous phospholipid can be loaded to cells, we incubated BHK or HeLa cells for up to 5 h with 0.5 mM of donor vesicles containing cholesterol, POPC, and 14:1/14:1-PE, -PS, or -PC (2:1:1 mol/mol) and 2 mM me ␤ -CD. Depending on the species, the transfer increased up to 2-5 h and then leveled off ( Fig. 4 ). The amounts of 14:1/14:1-PS, -PE, and -PC transferred to BHK cells corresponded to 440%, 125%, and 25% of the corresponding endogenous class, respectively, and 14.5%, 10.5%, and 12.5% of the total endogenous phospholipid, respectively. Consistent with these data, the phospholipid to protein ratio (nmol/mg) of the cells increased from 128 ± 7 to 157 ± 11 (n = 8), i.e., 23%. In an analogous experiment, exogenous PC taken up by HeLa cells comprised nearly 60% of total cellular PC or ‫ف‬ 30% of total phospholipid (data not shown). These results show that the relative amount of a phospholipid class can be greatly increased using me ␤ -CDmediated transfer of phospholipids from donor vesicles.

Effects on cellular content of other lipids
Me ␤ -CD is known to extract cholesterol effi ciently from cells ( 26,27 ). In principle, this could be prevented by in-We also studied whether replacement of POPC by di-18:0-, di-20:1-, or di-22:1-PC, which are more hydrophobic and thus less effi ciently transferred by the cyclodextrin than POPC ( 5 ), would prevent the transfer of the matrix phospholipid to cells. Although this was indeed the case, we also found less effi cient transfer of di-14:1-PC, -PE, or -PS from such donors (data not shown), possibly because of tighter lateral packing of the bilayer containing these phospholipids.

Effect of cell line
To compare incorporation of exogenous phospholipids to different cell lines, BHK, CHO, HeLa, or HF cells were incubated for 1 h with 1.5 mM donor vesicles containing cholesterol, POPC, and labeled 14:0/14:0-, 14:1/14:1-, and 18:1/18:1-PE or -PS (30:27:1:1:1 mol/ mol) in the presence of 8 mM me ␤ -CD. As expected, transfer of both PE and PS decreased markedly with increasing acyl chain length (hydrophobicity) with each cell line. The different PE species transferred similarly to CHO, HeLa, and HF cells, but less effi ciently to BHK cells ( Fig. 3A ). On the other hand, incorporation of the 14:0/14:0-and 14:1/14:1-PS species to CHO and HF was signifi cantly more effi cient than to BHK and HeLa cells ( Fig. 3B ). promised by the incubation based on Alamar Blue oxidation ( Fig. 6 ) or Trypan Blue exclusion (data not shown). However, some detachment of treated cells compared with control cells was observed.
Cell growth was assessed by determining the protein and DNA contents of the cells after incubation with donor vesicles and 8 mM me ␤ -CD for 1 h and after a 24 h chase in DMEM with or without 10% FBS ( Fig. 7 ). Both the protein and DNA contents of the treated cells were essentially identical to those of control cells after the introduction period as well as after the 24 h chase, independent of whether serum was present or not.

Decarboxylation of exogenous PS
As an implementation of the method, we studied how the acyl chains of a PS molecule affect its translocation from the plasma membrane to mitochondria. The kinetics of this process can be conveniently determined by monitoring decarboxylation of PS to PE because i ) PS decarboxylase is located exclusively in the mitochondria in mammalian cells and ii ) transfer to mitochondria rather than decarboxylation therein is rate limiting ( 11,29 ). Accordingly, we incubated BHK cells with vesicles containing a heavy isotope-labeled PS species and then determined the kinetics of formation of labeled PE with mass spectrometry. As shown in Fig. 8 , the rate of decarboxylation of 14:1/14:1-PS was very rapid and over 70% of it was converted to PE during the 1 h introduction period, thus showing that this PS species translocates to mitochondria very rapidly. The rate of decarboxylation of the other PS species with longer acyl chains was signifi cantly slower, thus indicating their slower transfer to mitochondria. The rate of PS transfer to mitochondria was inversely proportional to the molecular hydrophobicity ( Fig. 8 ), which has important implications regarding the mechanisms of PS translocation, as well as the maintenance of a high PS content in the inner leafl et of the plasma membrane (see Discussion).

DISCUSSION
It is desirable to effi ciently introduce labeled or unlabeled (phospho)lipids to cells in order to study their metabolism, traffi cking, and homeostasis. However, the very low solubility of natural (hydrophobic) phospholipids in corporating a proper concentration of cholesterol in the donor vesicles ( 28 ). Accordingly, we incubated BHK cells with donor vesicles containing different amounts of cholesterol in presence of 4 or 8 mM me ␤ -CD for 1 h. After washing, the cholesterol and phospholipid contents of the cells were analyzed. As shown in Fig. 5A , ‫ف‬ 75% cellular cholesterol was depleted when the donor vesicles were devoid of cholesterol and the concentration of me ␤ -CD was 4 mM. Addition of cholesterol to donor vesicles reduced the depletion and, at 50 mol%, no change in the cellular cholesterol content was observed, probably because the effl ux of cholesterol from cells equaled its fl ux in the opposite direction. The results obtained with 8 mM me ␤ -CD were generally similar, but even inclusion of 70 mol% of cholesterol in the donor vesicle preparation did not fully prevent depletion of cellular cholesterol ( Fig. 5B ). Notably, however, modest depletion of cellular cholesterol observed has only a minor effect on cell viability and growth (see below).
To study how the incubation of cells with donor vesicles and me ␤ -CD modifi es the cellular phospholipid composition, BHK cells were incubated for 1 h with 0.5 mM donor vesicles (labeled PS or PE/POPC/cholesterol, 1:9:10) and 8 mM me ␤ -CD. Mass spectrometric analysis showed that the POPC content of the cells increased ‫ف‬ 50%, corresponding to a 4% increase in total cellular PC. Also, an ‫ف‬ 30% decrease of total cellular SM was seen, but the relative abundances of the different SM species were not altered. After a chase in DMEM, the cellular content of POPC, most likely originating from the donor vesicles, as well as the total PC were practically normalized in 7 h, whereas the SM content returned to normal after 24 h (data not shown). No signifi cant changes in the total cellular content of PE, PS, or PI were observed (data not shown). During the chase, a decrease in the abundance of polyunsaturated PL species, such as 18:0/20:4 and 18:1/20:4 PE, was observed, but this could be fully prevented by inclusion of FBS in the medium (data not shown).

Effects on cell growth and viability
We next tested whether incubation with vesicles and me ␤ -CD compromises cell viability or growth. Cell viability was assessed based on Alamar Blue oxidation or Trypan Blue exclusion after 1 h incubation with donor vesicles at 0-10 mM me ␤ -CD. Cell viability was not signifi cantly com- decreased at higher me ␤ -CD concentrations. This is not unexpected, because the affi nity of phospholipids for cyclodextrin (relative to that for the bilayer) has been shown to increase with decreasing hydrophobicity ( 9 ). Thus, if the concentration of me ␤ -CD is too high, it binds much of aqueous media presents a major obstacle here. Although phospholipid exchange/transfer proteins seem ideal tools to circumvent this obstacle, they do not work effi ciently with intact cells and, consequently, impractical amounts of protein are needed ( 8,30 ). Fusion of lipid vesicles containing a virus receptor with the plasma membrane has been used to introduce lipids to infected cells ( 31 ) but has complications and has thus been rarely used. Lysophospholipids are readily taken up by cells and can be acylated to intact phospholipids therein ( 32 ), but this approach does not allow introduction of a specifi c molecular species to cells because the reacylation process is not selective regarding the acyl chain added.
Previous studies have shown that certain cyclodextrins can be used to transfer hydrophobic lipids to cells ( 5,10,11 ), but the parameters affecting transfer effi ciency and the possible deleterious effects on cells have not been systematically studied. In the present study, we studied these issues in detail and the key result was that with most phospholipids the amount transferred to cells increased systematically with increasing i ) donor vesicle concentration, ii ) time, and iii ) me ␤ -CD concentration. However, the transfer of the less hydrophobic species, such as Di-14:1-PS,   modeled at very different rates ( 5 ). Such studies were made feasible by the use of me ␤ -CD-mediated transfer of exogenous heavy isotope-labeled phospholipids to cells. Notably, the traditional approach relying on soluble phospholipid precursors (such as serine, ethanolamine, or choline) cannot provide equally detailed information on remodeling pathways and kinetics due to initial labeling of a multitude of molecular species ( 5 ). This would also be the problem if lysolipids are used as precursors, as the reacylation process is not specifi c for the acyl chain to be added. Exogenous phospholipids containing truncated acyl chains, such as 6:0 or 8:0 ( 6, 7, 39 ) do not suffer from this problem but have other potential problems. First, unlike natural phospholipid species, they translocate very rapidly between organelles and may thus be metabolized differently from their natural counterparts. Second, they could be rapidly degraded by homeostatic phospholipases, which tend to recognize and degrade lipids with unnatural acyl chains ( 5 ). Yet due to their very short chains, they may not mix properly with existing lipids, thus potentially compromising structure and function of cellular membranes.
Whereas it is well established that cells maintain the lipid compositions of their organelle membranes within close limits ( 40 ), it is not understood how such an accurate regulation is accomplished ( 41 ). A useful approach to study this is to perturb cellular lipid composition by loading the cells with exogenous phospholipids using me ␤ -CDmediated transport. The alternative approach based on overexpression of rate-limiting synthetic enzymes allows less effi cient loading due to rapid degradation of the phospholipid produced in excess ( 42,43 ). This approach is also less specifi c because the synthesis of numerous phospholipid molecular species and their precursors are simultaneously upregulated. The use of lysolipid precursors suffers from the same problem.

Transport of PS to mitochondria
We have previously shown that transfer of pyrene-labeled PS species from the plasma membrane to mitochondria is inversely proportional to molecular hydrophobicity, which in the case of phospholipids mainly depends on acyl chain length and unsaturation and less on the polar head group structure ( 11 ). Because the biological relevance of data obtained with lipids containing a bulky reporter group can be questioned, here we introduced deuteriumlabeled PS species to BHK cells and then monitored by mass spectrometry the conversion of these PS species to PE. We found that the rate of decarboxylation and thus the rate of transfer, of these PS species was inversely proportional to their molecular hydrophobicity ( Fig. 8 ). This is consistent with a model suggesting that spontaneous lipid transfer, in which effl ux from the donor membrane is the rate limiting step ( 44 ), is the main mechanism of PS transfer from the plasma membrane to mitochondria. However, transfer of phospholipids by noncarrier lipid transfer proteins such as non-specifi c lipid transfer protein (nsLTP) is also inversely proportional to lipid hydrophobicity ( 45 ) and must thus be considered as an alternative or parallel mechanism. However, adequate molecular the phospholipid rather than transfers it to cells. Accordingly, the optimal cyclodextrin concentration depends on the hydrophobicity of the lipid to be transferred. With the most hydrophobic species, the optimum is necessarily a compromise between transfer effi ciency and cell viability.
It is well established that incubation of cells with me ␤ -CD or similar CDs can deplete a major fraction of their free cholesterol ( 33 ), which can severely compromise their viability. Here, we showed that such depletion can be markedly alleviated or even fully prevented by including ‫ف‬ 50 mol% in the donor vesicles. However, the mole fraction of cholesterol needed in the donor vesicles to prevent depletion of cellular cholesterol may depend on several parameters including donor vesicle composition and concentration (unpublished observations). In principle, phospholipid/cyclodextrin complexes could be prepared and used to introduce exogenous phospholipid to cells. However, full complexation of phospholipids requires such a high concentration (>100 mM) of cyclodextrin ( 9 ) that most of the cellular cholesterol and a signifi cant fraction of phospholipids would be most likely depleted, thus killing the cells.
Beside cholesterol, me ␤ -CD has also been shown to extract other lipids from cells ( 12,34 ). In the present study, depletion of up to 30% of total cellular SM was observed. However, analogously to cholesterol, depletion of cellular SM should be preventable by including proper amounts of SM in the donor vesicles. Beside the lipid to be transferred, we also found a small (4%) increase in cellular PC, probably due to transfer of POPC from the donor vesicles. However, the cellular contents of both PC and SM were essentially normalized during a 24 h chase.
Previously, exposure of HaCaT cells to me ␤ -CD was shown to cause i ) a leakage of lactate dehydrogenase, ii ) an increase in caspase-3/7 activity, iii ) a decrease of ROS concentration, and iv ) an increase in interleukin 6 and 8 ( 35 ). In addition, inhibitory effects, such as inhibition of cell growth and cell cycle arrest, have been reported for various cell types ( 36,37 ). Such effects might be due to disturbed lipid composition (e.g., cholesterol depletion), or due to interaction of me ␤ -CD with other cellular components. Typically, the inhibitory effects were observed after an exposure to me ␤ -CD for several hours. In our experiments, the time of exposure was relatively short (typically р 1 h) and cholesterol depletion was prevented, which may explain why the viability and growth of the cells were not signifi cantly affected.

Applications
The present method should be very useful when studying various aspects of phospholipid metabolism. For instance, many phospholipids are remodeled after their synthesis de novo, i.e., one or both of their original acyl chains are replaced by others ( 38 ). This process involves A-type phospholipases as well as acyl transferases, but the identities of the enzymes involved are not clear. Recent data show that the A-type phospholipases involved in phospholipid remodeling are highly selective as indicated by, for example, acyl positional isomers of PS or PE being re-hydrophobicity of PS is required to maintain a high concentration of this phospholipid in the inner leafl et of the plasma membrane ( 11,46 ).
In conclusion, incubation with phospholipid vesicles and me ␤ -CD allows effi cient introduction of intact longchain phospholipids to cultured cells without signifi cantly compromising their growth or viability and should thus be very useful when studying intracellular phospholipid metabolism and traffi cking.