THE JOURNAL OF BIOLOGICAL CHEMISTRY 0 1989 by The American Society for Biochemistry and Molecular Biology, Inc.

Vol. 264, No. 7, Issue of March 5, pp. 3786-3793, 1989 Printed in U. S. A.

Plasma Membranes Contain Half the Phospholipid and 90% of the Cholesterol and Sphingomyelin in Cultured Human Fibroblasts*

(Received for publication, October 11, 1988)

Yvonne LangeSQ, Mark H. SwaisgoodllII , Benita V. RamosS, and Theodore L. Steckn From the $Departments of Pathology and Biochemistry, Rush-Presbyterian-St. Luke’s Medical Center, Chicago, Zllinois 60612 and the llDepartment of Biochemistry and Molecular Biology, the University of Chicago, Chicago, Illinois 60637

The literature suggests that cholesterol and sphin- gomyelin might be essentially confined to plasma mem- branes in mammalian cells; however, this premise has thus far escaped a direct test. We explored the issue in three ways. First, we fractionated whole homogenates of cultured human fibroblasts by equilibrium sucrose density gradient centrifugation. We found that the pro- files of cholesterol and sphingomyelin were indistin- guishable from those of two plasma membrane mark- ers, 5’ nucleotidase and [3H]galactose, which was con- jugated to the surface of intact cells from an exogenous donor by galactosyltransferase.

Second, we determined the relative surface areas of intact cells from their uptake of 1-(4-trimethyl- amino)phenyl-6-phenylhexa-1,3,5-triene, a cationic fluorescent dye which partitions into but does not cross plasma membranes. Relative to human red cell ghosts, the apparent surface area of the fibroblasts was 17,500 pm2/cell while for canine hepatocytes, the value was 11,500 pm2/cell. The relative ratios of cell cholesterol to dye binding (hence, surface area) were quite similar in ghosts, fibroblasts, and liver cells; namely 1.0,1.12, and 0.67, respectively.

Finally, we found that the specific ratios of both cholesterol and sphingomyelin to 5’ nucleotidase were only 10% less in gradient-purified plasma membranes than in whole homogenates. Similar results were ob- tained using an entirely different method of purifica- tion: two-phase aqueous partition. The cholesterol and sphingomyelin in fractions rich in other membranes was closely proportional to their 5‘ nucleotidase con- tent, suggesting that the presence of these lipids re- flected contamination by plasma membrane fragments. The 5’ nucleotidase/phospholipid ratio in the purified plasma membrane fraction was roughly twice that in whole cells. We conclude that the compartment marked by 5’ nucleotidase in cultured human fibroblasts con- tains approximately 90% of the two named lipids and half the cell phospholipid phosphorus.

A central goal in cell biology is to describe the essential

* This work was supported by American Cancer Society Grant BC- 95 (to T. L. S.) and National Institutes of Health Grants HL-28448 and HL-32466 (to Y. L.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “aduertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

§To whom correspondence should be addressed Dept. of Pathol- ogy, Rush-Presbyterian-St. Luke’s Medical Center, Chicago, IL 60612.

IIRecipient of a traineeship from the Cardiovascular Pathophysi- ology and Biochemistry Training Grant HL-7237 at the University of Chicago.

molecular characteristics of the membranes. Early in this endeavor, it was recognized that cholesterol and sphingomye- lin were enriched in the plasma membranes of mammalian cells (Colbeau et al., 1971; Steck and Wallach, 1970). It was also noted at that time that the distribution of these two lipids was covariant among subcellular fractions (Patton, 1970); however, the mechanism underlying this phenomenon was not clarified. Numerous later studies have confirmed these findings (Dawidowicz, 1987a). Furthermore, recent work with sphingomyelinase (Perret et al., 1979; Rawyler et al., 1983) and cholesterol oxidase (Lange and Ramos, 1983) as probes has demonstrated that the bulk of their substrates is attacked at the outer surface of intact cells.

These findings support the hypothesis that cholesterol and sphingomyelin are, for the most part, confined to plasma membranes. While this premise has been challenged for var- ious reasons (Dawidowicz, 1987a; Van Meer, 1987), it has not received a critical test. The present study sought a definitive answer to the question.

EXPERIMENTAL PROCEDURES

Materials-Uridine diphospho[~-6-~H]galactose (ammonium salt, 17.3 Ci/mmol) and [N-methyl-’4C]sphing~myelin (bovine, 58 mCi/ mmol) were purchased from Amersham Corp. Galactosyltransferase and polyethylene glycol 6000 were from Sigma. TMA-DPH’ was from Molecular Probes (Eugene, OR). Dextran 500 was obtained from Pharmacia LKB Biotechnology Inc. (Piscataway, NJ).

Cells-Human foreskin fibroblasts were derived from primary ex- plants and were grown, harvested, and homogenized according to Lange and Muraski (1988). Canine liver cells (mostly hepatocytes) were prepared as described (Bonnevie-Nielsen et al., 1982). Human erythrocytes were obtained from outdated units of normal blood provided by United Blood Services, Inc.

Linear Density Gradients-Linear density gradients of 12 or 38 ml were prepared in 5 mM Napi (pH 7.5) containing 20-43 or 23-54% sucrose (w/w). Approximately 8 X IO6 fibroblasts from two 175-cm2 flasks were homogenized in 1 ml of buffer A (0.25 M sucrose in 5 mM NaPi (pH 7.5)). Homogenates were cleared of unbroken cells and debris by centrifugation in a Sorvall H T 1000 rotor at 500 rpm for 5 min and either loaded onto or mixed into the gradients; the two methods gave equivalent results. Gradients were spun a t 4 “C for a minimum of 16 h a t 28,000 rpm in a Beckman SW 41 rotor. Gradients were fractionated from the bottom of the tube and aliquots taken for refractive index determination. The remainder of each fraction was diluted in buffer A and spun for 1 h a t 55,000 rpm in a Beckman 65 rotor to pellet the membranes. The pellets were resuspended in buffer A for analysis.

Lipid Analysis-Samples were extracted with 5 volumes of chlo- roform:methanol (2:1, v/v). The organic phase was dried under Nz. Cholesterol was converted to cholestenone with cholesterol oxidase and assayed spectrophotometrically (Lange and Ramos, 1983) or by HPLC (Lange and Muraski, 1987). Phospholipid phosphorus was determined using the micro method described by Bartlett (1959).

’ The abbreviations used are: TMA-DPH, 1-(4-trimethylarn- ino)phenyl-6-phenylhexa-1,3,5-triene; HPLC, high pressure liquid chromatography; C/P, moles of cholesterol/mole of phospholipid.

3786

Subcellular Cholesterol and Sphingomyelin 3787

Sphingomyelin Assay-Approximately 0.03 nmol of [14C]sphingo- myelin was added as a recovery standard to aliquots of lipid extracts containing 10-20 nmol of sphingomyelin. The samples plus parallel reference standards were chromatographed on plates of Silica Gel G in ch1oroform:methanol:acetic acikwater (25:15:4:2, v/v) and visual- ized with iodine vapor. Sphingomyelin spots were scraped into glass tubes, and the sphingosine released by acid hydrolysis was quantified with fluorescamine (Naoi et al., 1974). Aliquots of the aqueous phase remaining from the ether extraction of the sphingosine-fluorescamine complex were counted for I’C for recovery correction.

Other Methods-The surface of intact fibroblasts was labeled with [3H]galactose transferred from the exogenous sugar nucleotide using galactosyltransferase (Thilo, 1983). 5’ nucleotidase was measured as described (Lange and Steck, 1985). Protein was assayed by the Bio- Rad procedure (Bradford, 1976). NADH-cytochrome c reductase and galactosyltransferase activities were measured as described (Lange and Muraski, 1988). Cytochrome oxidase was assayed spectrophoto- metrically (Wharton and Tzagaloff, 1967). DNA was determined fluorimetrically using bis-benzimidazole (Cesarone et al., 1979).

RESULTS

We used four analytical criteria to estimate the fraction of human fibroblast cholesterol and sphingomyelin confined to the plasma membrane: (a) the qualitative congruence of the density profile of these lipids with established plasma mem- brane markers; (b ) the extent of copurification of these two lipids with plasma membrane markers on isopycnic sucrose density gradients; ( c ) the extent of copurification of these two lipids with plasma membrane markers on two-phase aqueous partition; and (a’) the correlation of the surface area of various cells with their total cholesterol content. We relied on 5’ nucleotidase as an endogenous marker for plasma membranes because its cell surface disposition has been unequivocally established and, contrary to some cell systems (Widnell et al., 1982), it was completely accessible on the surface of intact cells (as also found by Tulkens et al., 1974).

Isopycnic Fractionation-Cultured human fibroblasts were surface-labeled, homogenized, and spun gently to remove large particles. Supernatants then were spun to equilibrium on sucrose density gradients. The profiles of three relevant con- stituents from a representative experiment are shown in Fig. 1. 5’ nucleotidase and surface-conjugated [3H]galactose de- scribed a broad peak with a modal density of 1.15 g/cm3 (see also Lange and Steck, 1985). The profile of cholesterol par- alleled these plasma membrane markers so faithfully that a common locus seems likely. The copurification of cholesterol and plasma membrane markers resembled earlier studies on rat liver (Thines-Sempoux et al., 1969) and rat embryo fibro- blasts (Tulkens et al., 1974).

The density profile of sphingomyelin was indistinguishable from those of 5’ nucleotidase and cholesterol (Fig. 2A) . Most of the sedimentable protein and approximately half of the phospholipid were found at a higher density than the plasma membrane markers (Fig. 2B). Although markers for mito- chondria, lysosomes, Golgi membranes, nuclei, and smooth endoplasmic reticulum often overlapped those of the plasma membrane, they invariably showed patterns of distribution distinctly different from the four putative plasma membrane constituents followed in Figs. 1 and 2 (not shown, but see Lange and Muraski, 1988).

Co-purification of the Lipids and Plasma Membrane Mark- ers: Isopycnic Fractionation-The sucrose gradient profiles shown in Figs. 1 and 2 strongly suggested that cholesterol and sphingomyelin were closely associated with the same com- partment as 5’ nucleotidase and surface-conjugated [3H]ga- lactose. To make this inference quantitative, we determined the proportions of these constituents in the cell homogenate and in plasma membranes purified by equilibrium sucrose gradient centrifugation (Table I). The peak of 5‘ nucleotidase

1 00

75

50

25

0 10 20 30 40

% SUCROSE FIG. 1. The sucrose density gradient profile of plasma mem-

brane markers. Cultured fibroblasts were harvested, washed, and their surfaces conjugated with [3H]galactose as described under “EX- perimental Procedures.” The labeled cells were then washed, resus- pended, and homogenized. The homogenate was spun for 5 min at 800 X g and the supernatant layered on a linear sucrose gradient which was spun to equilibrium and fractionated. The gradient frac-

buffer and spun for 1 h at 300,000 X g. The pellets were resuspended tions were diluted with approximately 9 volumes of homogenization

in the same buffer for the assay of cholesterol mass (M), 5’ nucleotidase (A-A), and [3H]galactose (o”-O). Values are plot- ted with respect to the sucrose concentration (w/w) in each fraction determined by refractometry.

100

75

50

25

& 100

K 75

50

25

0 15 25 35 45

% SUCROSE

FIG. 2. The sucrose density gradient profile of membrane lipids. This experiment was identical to that shown in Fig. 1 except that the cells were not labeled with [3H]galactose and the gradient was constructed at a higher density. Lipids and protein in the mem- branes of each gradient fraction were determined as described under “Experimental Procedures.” Panel A shows the distribution of 5’ nucleotidase (M), sphingomyelin (A-A), and cholesterol (W). Panel B shows the distribution of protein (W) and phospholipid (A-A).

was enriched 9.3-fold with respect to protein (row 1) and 2- fold with respect to phospholipid (multiply row 2 by row 3). If 5’ nucleotidase uniformly marks only the plasma membrane and if the fraction isolated is pure, then the cell surface bears approximately (1/9.4) or 11% of cell protein and 50% of cell phospholipid. We noticed that as cultures aged past conflu- ence, the cholesterol to phospholipid molar ratio of the ho- mogenate rose from as low as 0.31 to as high as 0.45 mol/mol.

3788 Subcellular Cholesterol and Sphingomyelin

TABLE I Comparison of the specific content of cholesterol and sphingomyelin in the homogenate and the plasma membrane

fraction of human fibroblasts purified by sucrose density gradient centrifugation Unfractionated homogenates and the peak plasma membrane fractions from sucrose density gradients such as

those shown in Figs. 1 and 2 were analyzed for the named constituents as described under “Experimental Procedures.” The averaged specific contents of a number of experiments (given in parentheses) are tabulated f S.D. Units: 5’ nucleotidase (change in absorbance/min); protein (mg); cholesterol, phospholipid, and sphingomyelin (pmol). The quotients in the right column give the means k S.D. of the pairs of values from individual experiments.

Ratio Homogenate Plasma membrane fraction

Plasma membrane/ homogenate

1. 5’ nucleotidase/protein (7) 7.9 1- 1.1 73 f 11 9.3 rl: 1.3 2. Cholesterol/phospholipid (7) 0.38 f 0.04 0.69 f 0.11 1.8 f 0.15 3. 5’ nucleotidase/cholesterol (6) 51 _t 7 56 f. 9 1.1 f 0.14 4.5’ nucleotidase/sphingomyelin (4) 166 f 59 180 f. 78 1.1 f 0.14 5. Cholesterol/spbingomyelin (5) 3.3 _t 1.1 3.3 f. 0.9 1.0 f 0.13

Perhaps growth-arrested cultured cells elaborated plasma membranes differentially.

The ratios of 5’ nucleotidase to cholesterol (row 3) and to sphingomyelin (row 4) were approximately 10% lower in the cell homogenates than in the purified plasma membranes. If the marker enzyme is specific to and uniform in the plasma membranes, we can infer that, within statistical limits, only 10% of each of these lipid components resides outside of the cell surface compartment (presumably, within cytoplasmic membranes). Finally, the cholesterol to sphingomyelin ratio did not change upon plasma membrane purification (row 5), suggesting that they shared a cell-surface disposition to the same extent.

Co-purification of Plasma Membrane Markers: Two-phase Aqueous Partition-Mutual contamination of organelles in isopycnic sucrose density gradients contributes to present uncertainty about the subcellular distribution of cholesterol and sphingomyelin. Therefore, we tested the hypothesis anew with an entirely different approach to subcellular fractiona- tion: the differential partition of organelles between aqueous phases (Albertsson et al., 1982).

We found conditions under which most of the plasma membranes were recovered in the upper phase, most of the cytoplasmic organelles were recovered in the lower phase, and accumulation at the interface was small (Table 11). In these experiments, the major organelles were identified by marker enzymes. In the order of their preference for the upper phase, they were plasma membranes, 5’ nucleotidase; endoplasmic reticulum, cytochrome c reductase; Golgi apparatus, galacto- syltransferase; and mitochondria, cytochrome c oxidase. A similar hierarchy of partition coefficients has been reported for these organelles in other cells (Gierow et al., 1986). We used KI to drive the membranes to the lower phase. (Note in Table I1 that buffer constituents generally affected the chem- ical potential of the two aqueous phases more than the chem- ical potentials of the individual membranes, so that the dis- tributions of the different particles tended to vary in parallel.) The partition equilibrium was exceedingly sensitive to buffer composition but was highly reproducible under any given set of well controlled conditions.

Experiment A in Table I1 also relates the distribution of cholesterol and sphingomyelin to that of the various organelle markers. Clearly, within the limits of the determination, the two lipids partitioned identically with 5’ nucleotidase. Fur- thermore, like 5’ nucleotidase, their specific partition to the upper phase was a t least an order of magnitude greater than any of the cytoplasmic organelle markers. Finally, if the presence of these two lipids in the lower phase and interface is corrected for their content of plasma membranes as gauged by their quotient of 5‘ nucleotidase, the cholesterol and sphin-

gomyelin content of the endoplasmic reticulum, Golgi appa- ratus, and mitochondria manifested in this experiment was indistinguishable from zero.

Table I11 describes the plasma membrane-rich upper phase in more detail. 5‘ nucleotidase was enriched 9.4-fold over the unfractionated homogenate with respect to protein and 2-fold with respect to phospholipid, a purification comparable to that achieved by isopycnic centrifugation (Table I). Choles- terol and sphingomyelin were as enriched in the upper, plasma membrane-rich fraction and as depleted in the lower, organ- elle-rich fraction as was 5’ nucleotidase. As in Table I, there were only small changes in the specific ratios of these three constituents between the homogenate and the purified plasma membrane, again suggesting that the two lipids were essen- tially confined to the 5‘ nucleotidase compartment.

Nuclear Cholesterol-In the experiments described above, we gently centrifuged homogenates to remove intact cells and debris prior to fractionation; however, this also pelleted a significant fraction of the nuclei. Furthermore, the nuclei favored the upper fraction in the two-phase aqueous partition, co-purifying with the plasma membranes. We therefore determined independently the amount of cholesterol in the fibroblasts ascribable to nuclei. For this purpose, crude cell homogenates were analyzed without a low speed spin by equilibrium sucrose density gradient centrifugation. The com- position of the first gradient fraction (density 1.27 g/cm3) and the underlying pellet were compared to the input homogenate (Table IV). The nuclear fractions contained 1-3% of both the 5’ nucleotidase and the cholesterol in the cell as referenced to DNA. The ratio of the abundance of 5’ nucleotidase to cholesterol in the nuclear fractions was not significantly dif- ferent from that in the input, given the substantial error involved in determining the small quantities present. If we correct for plasma membrane contamination in the nuclear fraction using the 5‘ nucleotidase activity, it appears that only 0-1% of cell cholesterol is nuclear.

We note that two early studies reported substantial choles- terol in rat liver nuclear fractions (Keenan et al., 1970; Klei- nig, 1970). Neither study, however, corrected for contamina- tion by plasma membranes.

Correlation of Cell Surface Area with Cholesterol content- The highest cholesterol/phospholipid ratio in the purified plasma membrane fractions of our cultured fibroblasts equaled that of human red cell membranes; i.e. 0.80 mol/mol (Van Deenen and deGier, 1974). Similar values have been reported for other plasma membranes (Demel and DeKruyff, 1976). We reasoned that if human fibroblasts lacked appre- ciable intracellular cholesterol, then the ratio of total cell cholesterol to cell surface area would be close to that of human erythrocytes which lack intracellular membranes altogether.

Subcellular Cholesterol and Sphingomyelin TABLE I1

Distribution of human fibroblast organelles i n aqueous partition: effect of KI Aliquots (0.4 ml) of fibroblast homogenates (containing approximately 3 X lo6 cells in buffer A) were brought

to 2.0 ml (final) of a two-phase system constructed according to Albertsson et al. (1982) in glass Pasteur pipettes which had been sealed at the tip in a flame. The final mixture contained 4.4% (w/v) dextran 500, 3.6% (w/v) polyethylene glycol 6000, 14.6 mM Tris borate (pH 8.0), 0.05 M sucrose, 1 mM Napi, and the stated potassium iodide. The pipettes were thoroughly mixed and then centrifuged for 5 min at 800 X g to facilitate phase separation. The sealed tips of the pipettes were broken off, and the upper phases (-1.3 ml), lower phases (-0.6 ml), and the interfacial layer (-0.2 ml) were collected. Aliquots of the phases were taken for the analysis of marker enzymes, the activities of which were shown to be unaffected by the presence of the polymers. Experiments B and C were a pair, while A was performed separately. In experiment A, the membranes were pelleted at 100,000 X g for 1 h for lipid analvsis.

3789

Experiment KI Phase

mM A 4.87 Upper

Interface

B Lower

5.35 Upper Interface

C Lower

6.33 Upper Interface Lower

Distribution of recovered activity

5’ nucleotidase Galactosyltransferase C~~~~~~~ c~~~~~~ Cholesterol Sphingomyelin

% % 89 43 5 18 6 39

80 22 10 16 10 62 67 13 10 3 23 84

% 41 13 46 33 20 47 19 6

I 6

% % % 21 86 88 31 6 6 48 8 6

0 55 45 0 3 91

TABLE 111 Comparison of the specific content of cholesterol and sphingomyelin i n the homogenate and the plasma membrane

fraction of human fibroblasts purified by aqueous partition Two-phase aqueous partition was performed as described in Table I1 except that the plasma membranes received

a second cycle of purification. That is, the bulk of the upper phase of an initial partition was added to a proportional volume of fresh lower phase obtained from an identical system lacking only homogenate, and the mixture was partitioned again. The membranes in the upper fraction were pelleted at 100,000 X g for 1 h and analyzed together with the input homogenate. The values presented are averages from the number of experiments indicated in parentheses. Units are as in Table I.

Ratio Homogenate Upper phase Upper phase/ homogenate

1. 5’ nucleotidase/protein (3) 5.5 2. Cholesterol/phospholipid (1) 0.30 3. 5’ nucleotidase/cholesterol (2) 68 4. 5’ nucleotidase/sphingomyelin (2) 92 5. Cholesterol/sphingomyelin (2) 2.3

48

71 108

0.59

2.1

9.4 2.0 1.1 1.2 0.90

TABLE IV Comparison of the specific content of cholesterol in the homogenate and nuclear fractions of human fibroblasts

purified by sucrose density gradient centrifugation Fibroblasts were homogenized and an aliquot directly mixed into a gradient of 20-54% sucrose on top of a

cushion of 60% sucrose (all w/w). The tube was spun 16 h at 28,000 rpm in a Beckman SW 41 rotor. The unfractionated homogenate, the densest fraction (Fraction l ) , and the pellet were analyzed for the constituents listed. DNA is expressed in milligrams; the other units are given in Table I.

Ratio Homogenate Fraction I Pellet Fraction I/ Pellet/ho- homogenate mogenate

1. 5’ nucleotidase/DNA 61 1.4 0.73 0.023 0.012 2. Cholesterol/DNA 1.3 0.041 0.012 0.031 0.009 3. 5’ nucleotidase/cholesterol 46 33 60 0.72 1.31

We therefore sought to measure the surface area of fibroblasts. Since the overall lipid composition of various plasma mem-

branes is generally close to that of the human red cell (Daw- idowicz, 1987b; Longmuir, 1987; Rawyler et al., 1983; Renko- nen et al., 1972; Van Deenen and deGier, 1974), we argued that surface-directed membrane-intercalating amphipaths would equilibrate into various plasma membranes with closely similar affinity. In that case, the comparative uptake of the amphipath would reflect the relative plasma membrane com- partment size (i.e. surface area). Plasma membrane specificity can be achieved by applying a membrane-impermeable inter- calator to intact cells. The best probe we identified for this purpose was l-(4-trimethylamino)phenyl-6-phenylhexa-l,3,5-

triene (TMA-DPH; see Bronner et al., 1986, and Kuhry et al., 1983). Being hydrophobic, it vastly prefers membranes to aqueous spaces. Being cationic, it does not traverse the bilayer in experimental times. In addition, the dye is intensely flu- orescent in membranes and not in aqueous media. The con- sequent high sensitivity of the fluorometric assay allowed tracer levels of the dye to be used, thereby avoiding membrane perturbation. We found that the uptake of the dye was com- plete within 1 min at 0 “C, indicating a surface binding reaction rather than internalization. Bound fluorescence was conserved over the assay period, ruling out destruction of the probe.

To test whether the partition coefficient of the probe was

3790 Subcellular Cholesterol and Sphingomyelin TABLE V

Comparison of membrane lipid content and surface area of three cell types In several experiments, approximately 10s/ml sealed or unsealed erythrocyte ghosts and 106/ml cultured human

fibroblasts or canine liver cells were mixed with 5 p~ TMA-DPH in 0.15 M NaCI, 5 mM Napi (pH 8.0) and incubated for 1 min on ice. The cells were pelleted in an HS 4 rotor at 1000 rpm for 3 min at 0 "C. The ghosts were then pelleted in an SS-34 rotor at 15,000 rpm for 15 min at 0 "C. All pellets were resuspended to the original incubation volume and particle numbers in the input and recovered fractions determined. TMA-DPH fluorescence was determined in a Perkin-Elmer MPF-66 fluorimeter. The values for the cell pellets were corrected for contamination by ghosts (usually 4 0 % of the ghosts) estimated from the particle counts. The contribution of light scattering and the loss of fluorescence caused by the light scattering of the particles were corrected for from standard curves constructed using varying cell numbers and fluorescent beads with spectral properties similar to the dye. The corrected fluorescence of the cells was used to calculate their surface area by assuming a direct proportionality (ie. identical partition coefficient) with the ghosts. The chemical composition of the cells was determined on parallel aliquots. The numbers of independent determinations are shown in parentheses. For the three specific ratios (lines 4, 6, and 7), the values are the mean f S.D. of the ratios in the individual experiments.

Constituent Erythrocytes Fibroblasts Liver cells

1. Cholesterol (fmol/cell) 0.32" 47 f 8 (3) 16 f 3.6 (6) 2. Phospholipid (fmol/cell) 0.39" 130 (2) 190 f 41 (6) 3. Protein (pg/cell) 0.30' 210 (1) 720 f 164 (5) 4. Cholesterol/phospholipid (mol/mol) 0.82 0.36 (2) 0.084 f 0.005 (6) 5. Surface area (pm2/cell) X 0.135' 17.5 f 2.7 ( 3 ) 11.5 f 2.6 (5) 6. Cholesterol/surface area (amol/pm2) 2.4 2.7 f 0.2 (3) 1.6 f 0.53 (4) 7. Phospholipid/surface area (amol/um2) 2.9 23.2 (2) 18.8 rt 7.0 (4)

'From Van Deenen and deGier (1974). *From Groner et al. (1986). From Evans and Fung (1972).

the same for lipids of varied composition, we compared the uptake of TMA-DPH into sealed and unsealed human eryth- rocyte ghosts, since the two leaflets of their bilayer are mark- edly different in lipid composition (Etemadi, 1980). The result of such comparisons was that unsealed ghosts took up 2.2 times as much probe as did sealed ghosts. In control experi- ments with the membrane-permeable parent compound, di- phenylhexatriene, we also showed that sealed ghosts took up 90% as much dye as unsealed ghosts. This result suggested that the state of sealing of the ghosts did not affect their affinity for the dye. Thus, TMA-DPH reasonably fulfilled all of our initial criteria.

We determined the relative uptake of TMA-DPH by fibro- blasts compared to ghosts by co-incubating the dye with appropriate mixtures of the two. We first isolated the cells with a low speed spin and then the ghosts with a high speed spin. We removed the buffer from the ghosts to eliminate the unbound probe. (Washing was not feasible because the dye repartitions rapidly.) The pelleted cells and ghosts were then resuspended in identical volumes of the same buffer so that the loss of fluorescence through repartition of the dye into the fresh buffer would be matched for the two fractions. Corrections were made for the contamination of cell pellets by ghosts and for light scattering by the particles. The loss of fluorescence caused by light scattering also was determined and corrected for using mixtures of fluorescent bead standards with both cells and ghosts (see legend to Table V).

Under these conditions, the distribution of fluorescence between ghosts and cells followed ideal partition behavior in the range studied. Assuming the same partition coefficient for the different membrane species, the surface area of the fibro- blasts was determined relative to that of red cells to be 17,500 pm2 (Table V). This value roughly agreed with the observation that the packed cell volume of a fibroblast was approximately 100 times that of a red cell. Table V also shows that the cholesterol per cell surface area (that is, total cholesterol per unit dye bound) was only 12% greater for fibroblasts than for red cells.

To further test the validity of this approach, we performed similar measurements on unfractionated canine liver cells. Their average surface area was found to be 11,500 pm2. Their

cholesterol per surface area was two-thirds that of red cells. (The surface area per liver cell may have been somewhat overestimated because of the tendency of these cells to remain associated as pairs; however, this effect should not have altered the ratio of dye bound per cholesterol.)

Comparison of the data for the fibroblasts and the liver cells offers an additional insight. The liver cells contained far more phospholipid than the fibroblasts; namely, 2.4 times more per cell, 4.5 times more per cholesterol, and 5 times more per surface area ( i e . per unit dye bound). These values show that the binding of dye did not vary with cell phospho- lipid but rather with cell cholesterol. This result cannot be ascribed to selective dye binding to cholesterol, since TMA- DPH is avid for phospholipids (Bronner et al., 1986; Kuhry et al., 1983). Instead, the results suggest that the sterol and the probe are covariant because they are located primarily in the same compartment.

Is the measured surface density of cholesterol compatible with what is known of membrane molecular organization: that is, is there room for all the cell cholesterol in the plasma membrane bilayer? We assumed for the sake of the argument that the bilayer contained only cholesterol and phospholipid in a mole ratio of 0.8 and that the molecular cross-sections of cholesterol, pure membrane phospholipids, and phospholipids condensed by their 1:l association with cholesterol are 0.37 nm', 0.65 nm', and 0.50 nm', respectively (Demel and De- Kruyff, 1976). From these values, it can be calculated that cholesterol should contribute 30% of plasma membrane bi- layer cross-sectional area or 2.5 amol of cholesterol/pm2, in good agreement with the values shown in Table V.

DISCUSSION

Our results suggest that approximately 90% of the choles- terol and sphingomyelin and half of the phospholipid in cultured human fibroblasts reside in the plasma membrane. While these conclusions are consistent with previous findings (see the Introduction and below), they are nevertheless at odds with the prevalent view that the plasma membrane is minor in the great majority of eukaryotic cells (Alberts et al., 1983, p. 321). Reconciliation of these contrasting conclusions can be found in a close consideration of the data.

Subcellular Cholesterol and Sphingomyelin 3791

The case for rat liver is paradigmatic. It has been said that the plasma membrane constitutes 2% of total liver cell mem- branes (Alberts et al., 1983, p. 322). This value presumably refers to protein, since the purification of membrane 5‘ nu- cleotidase may approach 50-fold with respect to total cell protein (Hubbard et al., 1983). (The reciprocal of the purifi- cation factor gives an upper bound for the fractional repre- sentation of the marker compartment.) This value can be used to estimate that about 3.5% of hepatocyte phospholipid resides in the 5’ nucleotidase compartment in rat liver (Amar- Costesec et al., 1974; Colbeau et al., 1971; Von Hoeven and Emmelot, 1972). However, the hepatocyte plasma membrane is physically heterogeneous and, when fragmented by homog- enization, it distributes broadly with respect to sedimentation velocity and isopycnic density (Amar-Costesec et al., 1974; Hubbard et al., 1983; Steck and Wallach, 1970). Since the bile canaliculus is the portion most readily purified, it sometimes has been taken to represent the whole, even though it com- prises only a small fraction thereof (Matsuura et ai., 1984; Weibel et al., 1969). 5’ nucleotidase is highly concentrated in the bile canalicular portion of the plasma membrane com- pared to the basolateral portions (Matsuura et al., 1984) and underrepresents the magnitude of the hepatocyte cell surface in proportion to this enrichment. As a result, a 50-fold puri- fication of this marker with respect to protein does not signify that the plasma membrane contributes 2% of the cell protein but rather that the bile canaliculi may. Conversely, by virtue of their more uniform distribution in the plasma membrane, the lipids will appear to be correspondingly less enriched in the isolate than will selected regional markers. In the same way, cross-contamination will make general plasma mem- brane constituents like lipids appear to be more widely dis- persed in other organelles than regional markers.

In addition, 5’ nucleotidase is not an ideal indicator of the magnitude of the hepatocyte plasma membrane compartment because half of this enzyme is located intracellularly in en- docytic membranes in the rat liver (Stanley et al., 1980). Since the kinetics of internalization of 5’ nucleotidase differ from bulk pinocytosis (Thilo, 1985), the internalized portion, like surface receptors, may reside within the cell in a form more concentrated than in the rest of the plasma membrane.

5‘ nucleotidase is a more straightforward plasma membrane marker in some other cells. We and others (Tulkens et at., 1974) consistently have found that essentially all of the en- zyme activity is accessible in intact fibroblasts; hence, con- fined to the cell surface. In addition, there is no evidence for substantial lateral heterogeneity in its distribution in these plasma membranes. Because exogenous [3H]galactose labeling of the plasma membrane closely paralleled the 5’ nucleotidase and cholesterol distribution, there is no reason to postulate a substantial intracellular pool of the plasma membrane con- stituents with distinctive physical properties.

How much of the unesterified cholesterol in rat liver is in the plasma membranes? Van Meer (1987) has argued that it must be a small fraction because 60% of this sterol was recovered in the microsomes, which he took to represent endoplasmic reticulum. However, the microsomes are a col- lection of small fragments of various membranes, a large portion of which is plasma membrane (Thines-Sempoux et at., 1969). In fact, there is impelling evidence that the choles- terol in rat liver microsomes is derived from the cell surface and not from the endoplasmic reticulum. First, the specific enrichment of cholesterol in the microsomes over the homog- enate was nearly the same as for two cell surface markers, 5’ nucleotidase and alkaline phosphodiesterase I; half of each of these plasma membrane markers was recovered in the micro-

somes (Amar-Costesec et al., 1974). Second, the entire pool of cholesterol in rat liver microsomes was shifted in buoyant density by digitonin while markers for the endoplasmic retic- ulum were not (Thines-Sempoux et al., 1969). Finally, the densest fractions of rough endoplasmic reticulum vesicles (hence, those least likely to be contaminated with buoyant plasma membrane vesicles) have a C/P of 0.05 (Colbeau et al., 1971; Dallman et al., 1969). This should also be the value for the smooth endoplasmic reticulum with which the rough endoplasmic reticulum is thought to be in equilibrium by virtue of lateral diffusion. These values are uncorrected for plasma membrane contamination, the presence of which is suggested by the fact that organelle cholesterol content de- creased with increasing purification. Thus, the case for a major portion of the cholesterol of rat liver in the cytoplasmic organelles is weak.

In contrast, we have found that 80% of intact rat hepatocyte cholesterol is oxidized by cholesterol oxidase (Lange and Ramos, 1983). Assuming that this is the fraction of total cell cholesterol residing in the plasma membrane, we can estimate the fraction ( X ) of total membrane phospholipid in the plasma membrane. Let C1 and PI be the moles of cholesterol and phospholipid in the plasma membrane and CT and PT be the moles of cholesterol and phospholipid in the whole cell. Then,

Substituting C&T = 0.80, CT/PT = 0.16, and Cl/P, = 0.76 (Colbeau et al., 1971) yields X = 0.17. That is, 17% of total membrane phospholipid appears to reside in the plasma mem- brane of these cells. This estimate is similar to the value of 13% calculated by Brotherus and Renkonen (1977) but is 2.5- fold greater than the value of 6.5% suggested by stereology (Bolender et al., 1978). In other cell types, stereology suggests that the surface area of plasma membranes of various cells is three to six times greater than that predicted for a sphere of the same volume; while our measurements (Table V) suggest roughly a 10-fold redundancy. It is difficult to ascertain why stereology would underestimate the surface area, unless the redundancy of the highly convoluted surface is difficult to trace by that technique.

Dawidowicz (1987a) has suggested that 1-2% of total cel- lular lipid mass is in the plasma membrane of eukaryotic cells generally. One source cited (Pagano and Longmuir, 1983) provided such a figure without documentation. The other source cited (Sleight and Pagano, 1983) reported that 3% of the phosphatidylethanolamine in Chinese hamster ovary cells reacted with the extracellular probe, trinitrobenzenesulfonic acid. However, this phospholipid is generally found mostly on the cytoplasmic side of the bilayer, hence, unavailable to the probe. Furthermore, not all of the exposed molecules may react with the reagent and phosphatidylethanolamine may not be as prevalent in the plasma membranes of these cells as in the cell membranes as a whole (see Etemadi, 1980). Thus, the value quoted could substantially underestimate the plasma membrane phospholipid compartment in these cells.

Van Meer (1987) has calculated that no more than 24 and 34% of the cholesterol in BHK cells and MDCK cells, respec- tively, is in the plasma membrane. While these values are plausible, they are derived from data gathered in diverse studies for other purposes. It would be valuable to evaluate directly the distribution of cholesterol in these cell lines. In this regard, our preliminary studies have shown that at least 90% of the cholesterol in intact and fixed MDCK and BHK- 21 cells is susceptible to extracellular cholesterol oxidase.

We had inferred previously from the cholesterol oxidase

3792 Subcellular Cholesterol and Sphingomyelin

susceptibility of intact and fixed cultured human fibroblasts that approximately 90% of their cholesterol is at the cell surface (Lange and Ramos, 1983). (Rapid transbilayer move- ment presumably makes the entire sterol pool accessible to extracellular probes; Lange et al., 1981.) Furthermore, this cholesterol pool can be shifted in buoyant density by digitonin in parallel with 5’ nucleotidase (Lange and Steck, 1985). We have now confirmed the premise by demonstrating that the distribution of cholesterol upon subcellular fractionation closely parallels that of two plasma membrane markers (Figs. 1 and 2; Tables 1-111). In particular, the repeated observation that the cholesterol/5’ nucleotidase ratio was 10% lower in purified plasma membranes than in the homogenate signifies that only about 10% of the sterol is associated with other organelles. (The high cholesterol levels in the plasma mem- brane isolates cannot have been the result of contamination by other organelles, unless they were even more enriched in cholesterol than the plasma membrane itself. However, no such cholesterol-rich intracellular pool has been identified.)

Our data further suggest that sphingomyelin is distributed precisely with cholesterol to the limit of experimental preci- sion. This was also the seminal finding of Patton (1970).

Applying Equation 1 to the results shown in Table I, we calculate that 50% of the cellular phospholipid resides in the plasma membranes of fibroblasts (0.9 x 0.38/0.69 = 0.50). Similar values can be calculated using either 5’ nucleotidase or sphingomyelin in place of cholesterol, so that this result is not contingent on the contested subcellular distribution of the sterol. However, such calculations can overestimate the true value in proportion to the degree of membrane contami- nation of the plasma membrane isolate. We can set an upper bound of 16% for the contamination in the present case (that is, 1 - 0.8/0.69), based on the expectation that the C/P of pure plasma membranes will not exceed 0.8. This provides a minimal estimate of 43% cell phospholipid in the plasma membrane.

Our results are consistent with reliable analytical data in the literature. For example, Tulkens et al. (1974) found the following for cultured rat embryo fibroblasts: (a) no latency of 5’ nucleotidase in intact cells, suggesting that the portion of the cell surface present as inside-out endocytic vesicles was negligible; (b ) 5’ nucleotidase and cholesterol had a similar specific activity ratio among the various subcellular fractions; (c ) the buoyant density of this pair of markers was shifted selectively and coordinately by digitonin. In a study of cul- tured Friend erythroleukemia cells, Rawyler et al. (1983) showed the following: (a) At least 65% of cell cholesterol was in the purified plasma membrane. ( b ) From the C/P in the homogenate (0.307) and in the purified plasma membrane fraction (0.626), it can be calculated that if 65% of cellular cholesterol were in the latter, it would contain 32% of cellular phospholipid. ( c ) The cholesterol/sphingomyelin mole ratio was virtually the same in the plasma membrane (6.1) as in the homogenate (5.5). The corresponding ratios were 5.5, 5.7, and 7.9 in the other subcellular fractions (hence, identical within experimental error). (d ) Sphingomyelinase degraded 56% of its substrate in intact cells. Given that some of the sphingomyelin presumably was hidden in the cytoplasmic leaflet of the plasma membrane and that not all of the exposed fraction may have been digested, the plasma membranes of the erythroleukemia cells could contain as large a fraction of cellular sphingomyelin as we found in fibroblasts.

Our results bear on some interesting issues. (a) If 90% of cellular cholesterol and sphingomyelin reside

in the plasma membranes of at least some cells, their presence in various subcellular fractions probably reflects contamina-

tion by dispersed plasma membrane fragments. This hypoth- esis explains Patton’s (1970) observation of a close correlation of the two lipids in various membrane preparations. It would therefore seem prudent in studies of this kind to analyze the distribution of a constituent like cholesterol or sphingomyelin with reference to independent markers for the relevant com- partments. For example, Brotherus and Renkonen (1977) observed a precise correlation of the succinate dehydrogenase and cardiolipin content of numerous crude subcellular frac- tions and inferred that both were confined to the same com- partment (mitochondria), their broad distribution simply re- flecting cross-contamination. Similarly, Colbeau et al. (1971) inferred from the co-distribution of 5’ nucleotidase that the cholesterol in their inner mitochondrial membrane prepara- tions was attributable to contaminating plasma membranes.

( b ) It is misleading to ascribe the dispersed plasma mem- brane constituents to other organelles which they contami- nate upon subcellular fractionation.

(c) The use of cholesterol and sphingomyelin (and perhaps sterols and sphingolipids generally) as markers for plasma membranes seems justified. They are widely distributed, abundant, and presumably more uniformly distributed at the cell surface than are many proteins.

(d ) If 90% of cellular cholesterol is in the plasma mem- branes of various nonhepatic cells, then the high level of C/P in many whole cell homogenates must reflect an unexpectedly large contribution of the plasma membrane to total cell mem- branes. In particular, the C/P in unfractionated homogenates of a variety of normal cell types was reported to be 0.28-0.42 while in the corresponding purified plasma membranes, the C/P values were 0.52-0.70 (Chakravarthy et al., 1985; Depauw et al., 1985; Koizumi et al., 1981; Rawyler et al., 1983; Record et al., 1982; Renkonen et al., 1972; Robertson and Poznansky, 1985). It could be argued that the C/P values in these plasma membrane preparations may be artifactually lowered by their contamination with intracellular membranes. Nevertheless, even if the true cholesterol content of these plasma mem- branes were as high as is ever found (namely, C/P = 0.8), the plasma membrane would still constitute about 40-50% of total membrane phospholipid according to Equation 1. Since cholesterol adds 30% to the surface area of the plasma mem- brane bilayer (see above), the cell surface could constitute 50- 65% of the total bilayer area in some cells.

The fraction of the total membrane bilayer represented at the cell surface varies widely. If we assume that 80-90% of cell cholesterol is in the plasma membrane, we can calculate from the C/P in the homogenate and the purified fraction that the plasma membrane represents the following percent- ages of the total phospholipid In mammalian red cells, the value is of course 100%; in platelets, 79% (Fauvel et al., 1986); in various cells like the cultured human fibroblast, 50% (see above); in canine cardiac myocytes, 24% (Weglicki et al., 1980); in rat liver hepatocytes, 17%; and in dog liver cells, only about 10% (Table V). Where examined, the distribution of sphingomyelin closely paralleled that of cholesterol.

( e ) The expectation that cholesterol and sphingomyelin should be found throughout the membranes of the cell and not confined to a single locus is sustained by the widely held premise that membrane lipid molecules, despite their extreme insolubility, diffuse between membranes through the aqueous phase with a half-time of a few hours to days (Dawidowicz, 1987b; Phillips et al., 1987). Such a process should cause general mixing of bilayer constituents over their lifetime. However, it appears that membrane lipids are not at diffu- sional equilibrium within the cell (Wattenberg and Silbert, 1983). There is no direct evidence that membrane lipids are

Subcellular Cholesterc

demixed by an energy-dependent process. Conflicting evidence exists as to whether cholesterol and

sphingomyelin are found together because of their mutual chemical affinity (see Yeagle and Young, 1986). However, granting this premise does not explain the codistribution of these two lipids. The cholesterol content of most membranes generally greatly exceeds that of sphingomyelin (see, for ex- ample, Tables I and 111), and cholesterol is the more readily transferred between membranes (Dawidowicz, 198713; Phillips et at., 1987). Similarly, sphingomyelin is highly enriched in the outer leaflet of the red cell membrane (Etemadi, 1980), while cholesterol is almost evenly distributed (Lange and Slayton, 1982). It is difficult to envision a mechanism by which a small amount of sphingomyelin can determine the distribution of so much cholesterol.

Recent studies have suggested that physiological membrane lipids do not diffuse readily through the aqueous phase. In- stead, they appear to be transferred by a complex collisional mechanism (Steck et at., 1988). While fruitful collisions can occur between freely moving particles such as red cells and lipoproteins, such events are presumably restricted within the structured cytoplasm.

Finally, there is growing evidence that newly synthesized cholesterol (Kaplan and Simoni, 1985; Lange and Muraski, 1988) and sphingolipids (Lipsky and Pagano, 1985) are moved vectorially to the cell surface through a membrane rather than an aqueous pathway. Targeted biosynthesis explains how these lipids become specifically concentrated at the cell sur- face.

Acknowledgment-We thank Howard S. Tager for the generous gift of dog liver cells.

REFERENCES

Alberts, B., Bray, D., Lewis, J., Raff, M., Roberts, K., and Watson, J. D. (1983) Molecular Biology of the Cell, pp. 321-322, Garland Publishing, Inc., New York

Albertsson, P. A., Anderson, B., Larsson, C., and Akerlund, H.-E. (1982) Methods Biochem. Anal. 2 8 , 115-150

Amar-Costesec, A., Beaufay, H., Wibo, M., Thines-Sempoux, D., Feytmans, E., Robbi, M., and Berthet, J. (1974) J. Cell Biol. 6 1 , 201-212

Bartlett, G. R. (1959) J. Biol. Chem. 234,466-468 Bolender, R. P., Paumgartner, D., Losa, G., Muellener, D., and

Weibel, E. R. (1978) J. Cell Biol. 77 , 565-583 Bonnevie-Nielsen, V., Polonsky, K. S., Jaspan, J. J., Rubenstein, A.

H., Schwartz, T. W., and Tager, H. S. (1982) Proc. Natl. Acad. Sci. U. S. A. 79 , 2167-2171

Bradford, M. M. (1976) Anal. Biochem. 7 2 , 248-254 Bronner, C., Landry, Y., Fonteneau, P., and Kuhry, J. G. (1986)

Brotherus, J., and Renkonen, 0. (1977) Biochim. Biophys. Acta 486 ,

Cesarone, C. F., Bolognesi, C., and Santi, L. (1979) Anal. Biochem.

Chakravarthy, B. R., Spence, M. W., Clarke, J. T. R., and Cook, H.

Colbeau, A., Nachbaur, J., and Vignais, P. M. (1971) Biochim. Bio-

Dallman, P. R., Dallner, G., Bergstrand, A., and Ernster, L. (1969) J.

Dawidowicz, E. A. (1987a) Annu. Reu. Biochem. 56 , 43-61 Dawidowicz, E. A. (1987b) Curr. Top. Membr. Transp. 2 9 , 175-202 Demel, R. A., and De Kruyff, B. (1976) Biochim. Biophys. Acta 4 5 7 ,

109-132 Depauw, H., De Wolf, M., Van Dessel, G., Hilderson, H. J., Lagrou,

A., and Dierick, W. (1985) Biochim. Biophys. Acta 814 , 57-67 Etemadi, A.-H. (1980) Biochim. Biophys. Acta 604,423-475 Evans, E. A., and Fung, Y.-C. (1972) Microuasc. Res. 4 , 335-347

Biochemistry 25,2149-2154

243-253

100,188-197

W. (1985) Biochim. Biophys. Acta 812,223-233

phys. Acta 249 , 462-492

Cell Biol. 4 1, 357-377

d and Sphingomyelin 3793

Fauvel, J., Chap, H., Rogues, V., Levy-Toledane, S., and Douste-

Gierow, P., Sommarin, M., Larsson, C., and Jergil, B. (1986) Biochem.

Groner, W., Boyett, J., Johnson, A., and Scanttlebury, M. (1986)

Hubbard, A. L., Wall, D. A., and Ma, A. (1983) J. Cell Biol. 9 6 , 217-

Kaplan, M. R., and Simoni, R. D. (1985) J. Cell Biol. 101,446-453 Keenan, T. W., Berezney, R., Funk, L. K., and Crane, F. L. (1970)

Kleinig, H. (1970) J. Cell Biol. 4 6 , 396-402 Biochim. Biophys. Acta 203, 547-554

Koizumi, K., Shimizu, S., Koizumi, K. T., Nishida, K., Sato, C., Ota,

403 K., and Yamanaka, N. (1981) Biochim. Biophys. Acta 649, 393-

Kuhry, J . G., Fonteneau, P., Duportail, G., Maechling, C., and Laus- triat, G. (1983) Cell Biophys. 5 , 129-140

Lanae. Y.. Dolde, J.. and Steck, T. L. (1981) J. Biol. Chem. 2 5 6 ,

Blazy, L. (1986) Biochim. Biophys. Acta 8 5 6 , 155-164

J. 235,685-691

Blood Cells (Berl.) 12, 65-80

229

5321-5323 . .

Lanee. Y.. and Muraski. M. F. (1987) J. Biol. Chem. 262.4433-4436 Lange, Y., and Muraski; M. F. (1988) J. Biol. Chem. 263; 9366-9373 Lange, Y., and Ramos, B. (1983) J. Biol. Chem. 2 5 8 , 15130-15134 Lange, Y., and Slayton, J. M. (1982) J. Lipid. Res. 2 3 , 1121-1127 Lange, Y., and Steck, T. L. (1985) J. Biol. Chem. 260 , 15592-15597 Longmuir, K. J. (1987) Curr. Top. Membr. Transp. 2 9 , 129-174 Lipsky, N. G., and Pagano, R. E. (1985) J. Cell Biol. 100,27-34 Matsuura, S., Eto, S., Kato, K., and Tashiro, Y. (1984) J. Cell Biol.

Naoi, M., Lee, Y. C., and Roseman, S. (1974) Anal. Biochem. 5 8 ,

Pagano, R. E., and Longmuir, K. J . (1983) Trends Biochem. Sci. 8,

Patton, S. (1970) J. Theor. Biol. 2 9 , 489-491 Perret, B., Chap, H. J., and Douste-Blazy, L. (1979) Biochim. Biophys.

Acta 556,434-446 Phillips, M. C., Johnson, W. J., and Rothblat, G. H. (1987) Biochim.

Bwphys. Acta 906,223-276 Rawyler, A. J., Roelofsen, B., Op Den Kamp, J. A. F., and Van

Deenen, L. L. M. (1983) Biochim. Biophys. Acta 7 3 0 , 130-138 Record, M., Bes, J.-C., Chap, H., and Douste-Blazy, L. (1982)

Biochim. Biophys. Acta 688,57-65 Renkonen, O., Gahmberg, C. G., Simons, K., and Kaariainen, L.

(1972) Biochim. Biophys. Acta 255,66-78 Robertson, D. L., and Poznansky, M. J. (1985) Biochem. J. 2 3 2 ,

553-557 Sleight, R. G., and Pagano, R. E. (1983) J. Biol. Chem. 258 , 9050-

9058 Stanley, K. K., Edwards, M. R., and Luzio, J. P. (1980) Biochern. J.

Steck, T. L., and Wallach, D. F. H. (1970) Methods Cancer Res. 5 ,

Steck, T. L., Kezdy, F. J., and Lange, Y. (1988) J. Biol. Chem. 2 6 3 ,

Thilo, L. (1983) Methods Enzymol. 98,415-421 Thilo, L. (1985) Biochim. Biophys. Acta 8 2 2 , 243-266 Thines-Sempoux, D., Amar-Costesec, A,, Beaufay, H., and Berthet,

J. (1969) J. Cell Biol. 43 , 189-192 Tulkens, P., Beaufay, H., and Trouet, A. (1974) J. Cell Biol. 6 3 , 383-

401 Von Hoeven, R. P., and Emmelot, P. (1972) J. Membr. Biol. 9 , 105-

126 Van Deenen, L. L. M., and deGier, J. (1974) The Red Blood Cell

(Surgenor, D. M., ed) pp. 147-211, Academic Press, New York Van Meer, G. (1987) Trends Biochem. Sci. 12, 375-376 Wattenberg, B. W., and Silbert, D. F. (1983) J. Biol. Chem. 258 ,

Weglicki, W. B., Owens, K., Kennett, F. F., Kessner, A,, Harris, L., Wise, R. M., and Vahouny, G. V. (1980) J. Biol. Chem. 255,3605- 3609

Weibel, E. R., Staubli, W., Gnagi, H. R., and Hess, F. A. (1969) J. Cell Biol. 4 2 , 68-91

Wharton, D. C., and Tzagaloff, A. (1967) Methods Enzymol. 10,245- 246

Widnell, C. C., Schneider, Y.-J., Pierre, B., Baudhuin, P., and Trouet, A. (1982) Cell 28, 61-70

Yeagle, P. L., and Young, J. E. (1986) J . Biol. Chem. 261,8175-8181

I

99,166-173

571-577

157-161

186,59-69

93-153

13023-13031

2284-2289