THEJOURNAL OF BIOLOGICAL CHEMISTRY 8 1985 by The American Sociew of Biological Chemists, Inc.

VoL 260, No. 29, Issue of December 15, pp. 15592-15597 1985 Printed in lk7.A.

Cholesterol-rich Intracellular Membranes: A Precursor to the Plasma Membrane* (Received for publication, June 10,1985)

Yvonne LangeSB and Theodore L. Steckll From the $Departments of Pathology and Biochemistry, Rush Medical College, Chicago, Illinois 60612 and the nDepartment of Biochemistry and Molecular Biology, University of Chicago, Chicago, Illinois 60637

The disposition of newly synthesized sterols in cultured human fibroblasts has been examined in this study. We beganby demonstrating that cholesterol mass and exogenously added [sH]cholesterolboth are markers for the plasma membrane, perhaps better than 5'-nucleotidase. Cells were incubated with radioactive acetate to label their endogenous sterols biosynthetically, treated with cholesterol oxidase to convert plasma membrane cholesterol to cholestenone, and then homogenized and spun to equilibrium on sucrose gradients. The density gradient profiles of the various organelles were monitored using these markers: plasma membrane, radioactive cholestenone; smooth endoplasmic reticulum, 3hydroxy-3-methylglutaryl-CoAreductase (HMG-CoA reductase); and Golgi apparatus, galactosyltransferase. The buoyant density profiles of radioactive intracellular cholesterol and lanosterol both had a peak at 1.12 g/cm3, similar to Ei'-nucleotidase and galactosyltransferase but not to HMG-CoA reductase. This result suggests that cholesterol biosynthesis is not taken to completion in the endoplasmic reticulum. Digitonin treatment shifted the profiles of both plasma membrane andintracellular cholesterol to higher densities. Pretreatment of intact cells with cholesterol oxidase abolishedthe digitonin shift of plasma membranes but not the intracellular cholesterol, indicating that these two membrane pools are not entirely physicallyassociated.Because intracellular cholestero1,wasshifted more than anyof the organelle markers, it must reside in a separate membrane. Since digitonin selectively shifts thedensity of membranesrich in cholesterol, we infer thatnewly synthesized cholesterol accumulates in such membranesprior to its delivery to the plasma membrane. Taken together, these results suggest that cholesterol may be concentrated for delivery to the plasma membrane by being synthesized from a sterol precursor such as lanosterol in a discrete but undefined intracellular membrane.

The mechanisms by which newly synthesized proteins are brought to the plasma membrane are understood in some detail (1, 2). Incontrast, very littleis known about the

* This work wassupported by National Institutes of Health Grants HL-28448 and HL-32466 ( t o Y. L.) andGrant BC-95 from the American Cancer Society (to T. L. S.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solelyto indicate this fact. f To whom correspondence should be addressed Department of Pathology, Rush-Presbyterian&. Luke's Medical Center, 1753 West Congress Parkway, Chicago, IL 60612.

introduction of lipids into cell-surface membranes.Eighty to 95% of unesterified cellular cholesterol is associated with the plasma membranein various cells, leaving only a few per cent to all the internal membranes and lipid droplets (3). The transfer of cholesterol from its siteof synthesis to theplasma membrane in cultured cells was reported to have a half-time of 10 min (4)and 60 min (5) at 37 "C in two studies; there was no detectable flow of cholesterol from the cell surface to internal membranes (5). The mechanism by which the vectorial transfer of newly synthesized cholesterolto theplasma membrane occursis not known. However, there is evidence that specific membrane vesicles are involved (4,5).Because the cell-surfacemembrane is rich in cholesterol, it is reasonable to suppose that membrane vesicles which convey newlysynthesized cholesterol to the plasma membranewould also have a high cholesterol content. To examine this possibility, we probed the intracellular membranes with digitonin which is known to selectively intercalate into and increase the buoyant density of membranes containing cholesterol by virtue of its stoichiometric interaction with this sterol (6). Our results provide evidence for special membranes within the cell which bear newly synthesized sterol and are rich in cholesterol. EXPERIMENTALPROCEDURES

Materials-[3H]Acetic acid (sodium salt, 90 mCi/mmol), ["C] acetic acid (sodium salt, 56 mCi/mmol), uridine diphosphate [4,5-3H] galactose (40.3 Ci/mmol), ~~-3-hydroxy-3-methy1[3-'~C]glutaryl coenzyme A (57 mCi/mmol), [5-3H]mevalonicacid lactone (24 Ci/ mmol), and [7-3H]cholesterol (34.6 Ci/mmol) were purchased from New England Nuclear. Cholesterol (>99% pure) was obtained from NuCheck Prep (Elysian, MN) and lanosterol and cholest-4-en-3-one (cholestenone) from Steraloids Inc. (Wilton, NH). Cholesterol oxidase (EC 1.1.3.6; Breuibacterium sp.) was used as obtained from Beckman Instruments. HPLC' grade solvents were obtained from Fisher. Cell Culture-Human foreskin fibroblasts derived from primary explants were grown as described previously (5). Forty-eight hours prior to initiating experiments where HMG-CoA reductase was induced, the medium was replaced with one depleted of lipoproteins (5). Labeling Sterols (5)"Radiolabeled acetate (0.1-1.0 mCi) wasadded in ethanol ( 4 % ) t o the medium and the incubation continued at 37 " C for various times. The medium was then removed, the monolayer rinsed, and the cells detached from the flask by trypsin treatment. The cells were washed twice in phosphate-buffered saline a t 0 "C prior to cholesterol oxidase treatment. Treatment of Cells with Cholesterol Oxidase(7)-Cells were washed once in 40 volumes of 310 mM sucrose, 0.5 mM NaPi (pH 7.5) and resuspended in approximately 10 volumes of this buffer. 10 W/ml cholesterol oxidase was added, and thecells were incubated a t 37 "C for 10 min. The reaction was stopped by chilling on ice. The abbreviations used are: HPLC, high performance liquid chromatography; HMG-CoA reductase, 3-hydroxy-3-methylglutaryl-CoA reductase.

15592

15593

Cholesterol in Intracellular Membranes Digitonin Treatment of Cell Homogenates-Cells were homogenized in 0.25 M sucrose, 5 mM Napi (pH 7.5) on ice by 25-35 strokes of a glass-glass coaxial pestle homogenizer. Homogenates were spun for 5 min at 800 X g to remove unbroken cells and nuclei. Ten pl of a solution of digitonin in ethanol was added per ml of cell homogenate and incubated for 10 min on ice. (The amount of digitonin added was approximately the molar equivalent of the anticipated unesterified cholesterol in thepreparation, allowing for the conversion of plasma membrane cholesterol to cholestenone by cholesterol oxidase. Some of the variation in our digitonin density shift results may be due to fluctuations in theamount of digitonin actually bound to thesubcellular membrane fractions, a parameter we could not measure or control.) A parallel control contained 10 p1 of ethanol/ml of homogenate. In experiments in which the distributions of membranes treated and untreated with digitonin were compared on a single gradient, it was necessary to remove free digitonin prior to mixing the two homogenates. For this purpose, 80-100 pl of red cells prewashed in sucrose buffer was added to the digitonin-treated homogenate. The suspension was incubated for 5 min on ice, mixed with the untreated homogenate, spun for5 min at 800 X g to sediment the red cells, and the supernatant taken for density gradient analysis. Density Gradients-Homogenates were layered on 13-ml linear sucrose gradients of density 1.09-1.25 g/cm3 prepared in 5 mM Napi (pH 7.4). 0.1 mM EDTA and 10 mM dithiothreitol were added to gradients in which HMG-CoA reductase activities were measured. The gradients were spun in a Beckman SW 40 rotor for a minimum of lo8 X ,g min a t 3 “C. Gradients were collected in equal volume fractions (1.2-1.5 ml) from the bottom of the tube. Aliquots of each fraction were taken for refractive index determination and enzyme assays, and the remainder was extracted with chloroform/methanol (21) for lipid analysis. Lipid Analysis-Samples were extracted with 5 volumes of chloroform/methanol (2:1, v/v). The organic phase wasremoved and dried under Nz. The radioactivity in sterols was determined by thin layer chromatography as described (5) or by HPLC on a Beckman Ultrasphere ODS reversed-phase column with acetonitrile/isopropano1 (9010, v/v) as themobile phase, using a Beckman system detecting at 254 nm. Sterols were identified by their co-migration with standards on HPLC and intwo different thin layer chromatography systems (chloroform/methanol, 1002, v/v, and petroleum ether/ethyl ether/acetic acid, 90101, v/v). Enzyme Assays-Galactosyltransferase was .assayed as the ovalbumin-stimulated incorporation of [3H]galactose into trichloroacetic acid precipitates (8).For the assay of membranes labeled from [3H] acetate, the final precipitates were dispersed in 50 pl of 1 N NaOH, and 0.5 ml of ice-cold isopropyl alcohol was added. After 15 min on ice, the samples were spun, the supernatants containingthe extracted lipids discarded, and the pellets washed in 8% trichloroacetic acid prior to redissolving in NaOH for the determination of transferred [3H]galactose. HMG-CoA reductase was measured as described (9). Kyro EOB used in theassay was the generous gift of Dr. D. H. Hughes of Procter and Gamble. 5’-Nucleotidase was assayed by a modification of the method of Avruch and Wallach (10). Aliquots of cell homogenates (10-50 pg of protein) were incubated for 20-40 min at 37 “C in a final volume of 0.75 ml containing 50 mM glycine (pH 9.0), 0.4 mM MgClZ,and 0.16 mM 5’-AMP. To stop thereaction, the tubes were chilled on ice, and ZnSO, and Ba(OH)* were added in succession to 37.5 mM (final) to precipitate unreacted substrate. After 10 min on ice, the tubes were spun. The supernatant was made to 70 mM in ZnS04 andBa(OH)z, and the precipitate was removed by centribation. Theabsorbance of the adenosine in the supernatantwas measured at 260 nm. RESULTS

Effect of Digitonin and Cholesterol Oxidase on the Density Profile of Plasma Membranes-We demonstrated previously that 95% of the unesterified cholesterol in cultured human fibroblasts resides in the plasma membrane, so that the remaining 5% must be distributed among all of the internal membranes and the lipid droplets (3). This great enrichment of plasma membranes in cholesterol suggests that it could serve as a reliable marker for this membrane. We therefore compared the distribution of cholesterol withthat of a widely used marker, 5’-nucleotidase, in cellhomogenatesspun to

w

2

4w

1.6

0.8

i; 0.8

0.4

I-

v)

w

I

0

0

a 3.0

6.0

2 2.0

4.0

c)

X

E, 0

2 .o

1.0

FRACTION

FIG. 1. The effect of digitonin and cholesterol oxidase on the buoyant density profile of plasma membranes.Panels A-C, cells were removed from flasks by trypsin treatment, washed, and resuspended in 0.15 M NaCl, 5 mM NaPi (pH 8.0). [3H]Cholesterol was added and the cells incubated for 5 min on ice. The cells were pelleted by centrifugation, washed in 0.31 M sucrose, 0.5 mM Napi (pH 7 3 , resuspended in this buffer, and homogenized. After spinning for 5 min at 800 X g, the supernatant was divided into two 1-ml portions. 150 pg of digitonin in 10 pl of ethanol was added to 1aliquot and the same volume of ethanol to the other. The mixtures were incubated for 10 min on ice and layered on two linear sucrose gradients which werespun to equilibrium, fractionated, and analyzed as described under “Experimental Procedures.” To facilitate analysis, cholesterol was converted to cholestenone prior to HPLC. Cholestenone mass (Panel A), [3H]cholesterol (panel B ) , and 5’-nucleotidase activity (panel C) were measured in each fraction of the gradients for control (0)and digitonin-treated (0)cell homogenates. Density is expressed in g/cm3. Panels D-F, the experiment was the same as that described above except that the cells were treated with cholesterol oxidase (10 IU/ml) for 10 min at 37 “C in 0.31 M sucrose, 0.5 mM Napi (pH 7.5) prior to homogenization. Cholestenone mass (panel D), [3H]cholestenone (panel E ) , and 5’-nucleotidase activity (panel F)were measured in each fraction of the sucrose gradient for control (0)and digitonin-treated cells (0).

equilibrium on sucrose gradients. In addition, intact were cells prelabeledwithexogenous[3H]cholesterol,since wehave found that this label does not equilibrate with the intracellular pool over many hours (5). As shown in Fig. 1 (panels A X ) , the distribution of [3H]cholesterol, cholesterol mass, and5’nucleotidase coincided exceptthat 5”nucleotidase was more broadly distributed than cholesterol. Labeled cell homogenates were also treated with digitonin prior to centrifugation. All threemarkerswere. shifted by digitonin to a similar higher density (Fig. 1, panels A-C). The increase in buoyant density induced in the plasma membranes by digitonin (from 1.12 to 1.19 g/cm3) is comparable to that seen in red cell ghosts:butgreater than that reported in

* Y. Lange and T. L. Steck, unpufilished observation.

15594

Cholesterol in Intracellular Membranes

similar digitonin experiments using rat liver membranes (6). We examined the effect of cholesterol oxidase treatment of intact cells on thedistribution of plasma membrane in sucrose gradients. The cholesterol oxidase treatment caused the peaks of cholesterol mass and [3H]cholesterolto split intotwo peaks of density 1.09 and 1.13 g/cm3 (Fig.1,panels D and E ) . (This reproducible finding may be caused by a change in the pattern of fragmentation of plasma membranes in cholesterol oxidasetreated cells during homogenization or by a density effect of cholestenone itself.) The low density peak did not appear in the 5”nucleotidase profile (Fig. 1,panel F).This finding could indicate the latency of plasma membrane 5”nucleotidase in these fractions, perhaps due to sequestration of the enzyme within sealed inside-out vesicles. In anycase, the coincidence of cholestenone mass and the [3H]cholestenone marker with 5’-nucleotidase in such experiments verifies the plasma membrane localization of these constituents, since the cholesterol oxidase only could have oxidized sterols at the surface of the 1.2intact cells. (All plasma membrane cholesterol is a substrate for the enzyme because the sterol equilibrates across the 7 bilayer within seconds; see Ref. 11.) x 0.8Next we examined the effect of digitonin on the density E profile of plasma membranes from cells which had been Q treated with cholesterol oxidase. In these experiments, intact u [3H]cholesterol-labeled fibroblasts were exposed to choles0.4terol oxidase prior to homogenization and digitonin treat5 U ment. The enzyme converted greater than 95% of cell cholesterol to cholestenone and inhibited the digitonin-induced I density shift of all three markers (Fig. 1, panels D-F). (We occasionally observed a small shift to higher density of one plasma membrane marker or another under these conditions, e.g. panels E and F. It may be that the small amount of cholesterol remaining in the oxidized plasma membrane acPpn;aqg counts for this effect. We also have observed’ that cholesterol 4 6 8 IO oxidase-treated red cell ghosts bind small amounts of digiFRACTION tonin, suggesting that membrane cholestenone may form FIG. 2. The buoyant density profile of newly synthesized weak complexes with this probe.) In any case, the observation sterols. Confluent monolayers of fibroblasts were incubated for 48 h that cholesterol oxidase pretreatment suppressed the effect of at 37 “C in medium containing 5% lipoprotein-deficient serum, and digitonin on plasma membrane density provided a powerful then 0.07 mCi of [14C]acetatewas added to the flask. The cells were means of examining the effect of digitonin on intracellular incubated for 2 h at 37 “C, the medium was removed, the monolayer rinsed, and fresh medium containing 1 mCi [3H]acetate added for a cholesterol (see below). Distribution of Newly Synthesized Sterol on Sucrose Gra- further 0.5 h incubation at 37 “C. The labeled medium was removed, the cells were rinsed and dissociated by trypsin treatment. The dients-To trace the pathway of newly synthesized sterols and cells were treated with cholesterol oxidase, homogenized, spun to through the cell, we labeled cultures with [?“C]acetatefollowed equilibrium on a sucrose gradient, and assayed as described under after adefined interval by [3H]acetate. In addition to revealing “Experimental Procedures.” The distribution of radioactivity in lathe temporal distribution of sterols within the cell, the use of nosterol ,).( cholesterol (O), and cholestenone (A)was measured. two radiolabels facilitated the identification of precursors in Panel A shows the distribution of 3H (“young” label) and panel B the pathway of cholesterol biosynthesis. At early times of shows the distribution of 14C (“old” label). The profiles of ratios of 3H to lac(young to old label) in each sterol are shown in panel C. biosynthesis, we noted a major radioactive component which Fraction 1 is the bottom of the gradient. The density of fraction 8 was identified as lanosterol. We also used cholesterol oxidase was 1.12 g/cm3. to discriminateintracellular cholesterol againsta massive background of plasma membrane cholesterol (5). At the time of assay, most of the “young,” second-added In the experiment shown in Fig. 2, suitably prepared monradiolabel, 3H, was in lanosterol,-(Fig. 2, panel A ) , whereas olayers were incubated for2 h with [14C]acetateat 37 “C. The radiolabel was removed and [3H]acetate added for a further proportionately more of the first added label, I4C, had been 0.5 h incubation at 37 “C. The cells were removed from the converted to cholesterol and transferred to theplasma memflask, treated withcholesterol oxidase, homogenized, and spun brane (Fig. 2, panel B). Consistent with itsidentityasa to equilibrium on a sucrose density gradient. The profiles of precursor in the biosynthesis of cholesterol from acetate, radiolabel in lanosterol, cholesterol, and cholestenone were lanosterol was more enriched in the second-added radiolabel determined. Since cholesterol oxidase converted all of plasma 3H than was cholesterol (Fig. 2, panel C).Similarly, cholestenmembrane cholesterol to cholestenone whiie leaving intracel- one had a lower, 3H/14Cratio than did intracellular choleslular cholesterol pools untouched, the radiolabel in cholesten- terol. There was also significant variation in the isotope ratio one is attributable to that cholesterol which had reached the (i.e. age) of lanosterol across the gradient. The intracellular plasma membrane at thetime of enzyme treatment (3). Con- cholesterol was only slightly younger than the cholestenone versely, the radiolabel in cholesterol resistant to cholesterol (plasma membrane cholesterol) and much older than lanosterol. These data suggest that thelanosterol pool is converted oxidase marks the intracellular pool.

a

Cholesterol in Intracellular Membranes relatively quickly into intracellular cholesterol while the latter is transferred to theplasma membrane more slowly. This view is consistent with our previous finding that cholesterol moves to the cell surface in a first-order fashion (Le. from a single pool in whichold and young intracellular cholesterol are completely mixed) with a half-time of approximately 1 h at 37 "C (5). A striking feature of the data in Fig. 2 is the low buoyant density (i.e. 1.12 g/cm3) not only of intracellular cholesterol (previously noted in Ref. 5) but also its precursor, lanosterol. This finding was unexpected in view of the belief widely held since the early studies of Chesterton (12) that cholesterol biosynthesis is completed inthe endoplasmic reticulum. Therefore, we compared the equilibrium density distribution of newly synthesized sterol with that of HMG-CoA reductase and galactosyltransferase on sucrose gradients (Fig. 3). HMGCoA reductase, an early and rate-determining enzyme in the pathway of cholesterol biosynthesis, is known to be ,located in the smooth endoplasmic reticulum (13) and was found primarily near the bottom of the gradient at 1.18 g/cm3 (Fig. 3, panel A). In contrast, galactosyltransferase, a Golgi membrane marker (14), was recovered mostly at the top of the gradient at 1.12 g/cm3 (Fig. 3,panel A). Newly synthesized of intracellular cholesterol and lanosterol also were at the top the gradient (Fig. 3,panel B), clearly demonstrating that they were not associated with the bulk of the smooth endoplasmic reticulum.

A

e

15595

Effect of Digitonin on the Density of Newly Synthesized Sterols-The coincidence of the peaks of newly synthesized sterols with the plasma membrane and galactosyltransferase activity in Fig. 3 was observed repeatedly. (The co-migration of newly synthesized cholesterol with plasma membrane has been noted previously (54.) That these components are found at the same density couldbe fortuitous, but alternatively could indicate the physical association of newly synthesized sterol with either the cell surface or Golgi membranes. To test these hypotheses, we examined the effect of digitonin onthe density distribution of newly synthesized sterols (Fig. 4). In these experiments, cells in one flask were labeled with [3H]acetate, while a duplicate monolayer was labeled with [14C]acetate.The cells were then treatedwith cholesterol oxidase and homogenized separately. Digitonin was added to the 3H-labeled homogenate which then was mixed with the l4C-labeled homogenate, layered on a sucrose gradient, and spun to equilibrium. The profiles of each radiolabel in cholesterol and cholestenone were determined. As in Fig. 1, panels D-F, the pretreatment of cells with cholesterol oxidase abolished the effect of digitonin on the buoyant density of the plasma ,membrane (Fig. 4, panel A). Nonetheless, digitonin induced a striking shift in the distribution of newly synthesized intracellular cholesterolto higher density (panel B ) . The extent to which digitonin shifted newly synthesized cholesterol varied among experiments and was not always as great as inFig. 4.We infer that the co-migration of newlysynthesized cholesteroland plasma membrane markers on sucrosegradients was not due to a physical association between the coresponding cell components unless digitonin

Y

W v)

4

I-

o

3

n

0.8 I m 2 F

W

U

n

h

0

0

a0

v

In

0

0

0.41

0 v-

v-

X

X

I

n

0' 4

U

FRACTION FIG. 3. The buoyant densityprofile of marker enzymesand newly synthesized sterols. Confluent monolayers of fibroblasts were incubated for 47 h a t 37 "C in medium containing 5% lipoprotein-deficient serum, and then0.75 mCi I3H]acetate was added to the flask. The cells were incubated for 1h a t 37 "C,the medium removed, and thecells rinsed and dissociated from the flask with trypsin. The cells were treated with cholesterol oxidase, homogenized, and spun to equilibrium as described under "Experimental Procedures." The gradient was analysed for enzyme activity and lipid radioactivity. Fraction 1 is the bottom of the gradient. Panel A shows the distribution of HMG-CoA reductase (0)and galactosyltransferase (0)activities expressed in arbitrary units of lo9 X cpm. Panel B shows the distribution of [3H]lanosterol (O),cholestenone (A),and cholesterol (0).

0

I

I

I

I

2 4 6 8 1 0 FRACTION FIG. 4. The effect of digitonin on the buoyantdensity profile of newly synthesized cholesterol. Duplicate flasks of cells were labeled with ["CC] or [3H]acetate for 1h as described in the legend to Fig. 2. The labeled cells were treated with cholesterol oxidase and homogenized. The 3H-labeledhomogenate was treated with digitonin, mixed with the I4C-labeled homogenate, and the buoyant density analyzed as described under "Experimental Procedures." Gradient fractions were extracted for the determination of 3H ( 0 plus digitonin) and 14C ( 0 ; minus digitonin) in plasma membrane (cholestenone; panel A ) and intracellular cholesterol (panel B ) .

15596

Cholesterol in Intracellular Membranes

1.12 g/cm3 (Fig. 3). Therefore, these nascent sterols are not in the same membranes asthe earliest, rate-determining enzyme in thesterol biosynthetic pathway. Since HMG-CoA reductase appears to be confined to the smooth endoplasmic reticulum (13),we infer that newly synthesized lanosterol and cholesterol either weremoved from that site to the more buoyant membranes or were in fact synthesized in the more buoyant membranes from sterol intermediates whichwere transported from the endoplasmic reticulum.Certain soluble cytoplasmic proteins are known to stimulate sterol biosynthesis in microsomal membranes(20-22) and may carry lipids (22). Although this hypothesis was not favored in previous studies of cholesterol biosynthesisin liver homogenates(23), our data might signifythat such proteins catalyze intermembrane transfer of intermediates during sterol biosynthesis. FRACTION Since microsomesare typically mixtures of small vesicles from FIG. 5. The effect of digitonin on the buoyant density profile a variety of cell membranes, including endoplasmic reticulum, of galactosyltransferase. The experimentwas as described in the plasma membrane, and Golgi membranes (241, the stimulawas measlegend to Fig. 1, panels D-F. Galactosyltransferase activity uredforcontrol (0) and digitonin-treatedhomogenates (0) and tion of sterol biosynthesis in microsomal homogenates by soluble cytoplasmic proteins could reflect the rate-limiting expressed in arbitrary unitsof cpm X W 3 . transfer of intermediates between different membranes as happened to disassociate them by some unknown mechanism. well as intramembrane events. The fact that very little of the Newly synthesized sterols often were found to have the nascent lanosterol or cholesterol coincided with the dense same buoyant density profile as galactosyltransferase. How- peak of HMG-CoA reductase activity (Fig. 3) suggests that ever, the effect of digitonin on galactosyltransferase was al- either their movement out of endoplasmic reticulum is very ways slight compared to thesterols (Fig. 5).A small digitonin- rapid or that they are made elsewhere. The fact that these induced increase in thedensity of Golgi membranes has also newly madesterols show a discrete, essentiallyunimodal peak been reported for liver homogenates (6). The finding that in buoyant density (Fig. 2) suggests that they are either intracellular cholesterol is much more sensitive to digitonin selectively transferred to or are synthesized at a single subthan is galactosyltransferase suggests that they are not con- cellular site. The age of the lanosterol ( i e .the ratio of labels incorporated stituents of the same membrane. However, in that neither galactosyltransferase nor cholesterol may be uniformly rep- from acetate added at two times) was not uniformly distribresented throughout the stacks of the Golgi apparatus (15, uted with buoyant density (Fig. 2). It is possible that the3H/ E ) , the present data do not rule out the possibility that newly '*C ratio was reduced in the fractions of lowest lanosterol synthesized cholesterol travels through specific regionsof the content simply becauseof a background of contamination by older sterol species. Alternatively,variation in theage profile Golgi apparatus. might signify multiple pools of lanosterol which are not in DISCUSSION rapid diffusional equilibrium and which turn over at different rates, even though all of the lanosterol appears to chase into It became apparent in thecourse of this study that cholesterol may be a preferred plasma membrane marker, at least cholesterol. In any case, there is no evidence that lanosterol forsomecells such as fibroblasts, for several reasons. (i) equilibrating at the high denisty of HMG-CoA reductase is Enzyme markers such as 5'-nucleotidase are found in intra- younger than therest, asmight be the case if it were synthecellular as well as surface membranes (17, 18). This may be sized in thesmooth endoplasmic reticulumand subsequently why the 5'-nucleotidase profile is broader than that of cho- moved to more buoyant membranes. It is our working hypothlesterol in Fig. 1. (ii) In contrast, cholesterol is 95%, and esis, therefore, that thepools of lanosterol, intracellular choexogenously added radiocholesterolis 9&99% exposedat the lesterol, and cell-surface cholesterolmay each be homogeneextracellular surface (3). (iii) Even greater stringency is af- ous in age ( i e . turnover rate), the dispersion in their buoyant forded by treatment of intact cells with cholesterol oxidase densities notwithstanding. It is believed that digitonin selectively increases the buoywhich converts the plasma membrane cholesterol to cholestenone, which thus becomes a marker specific for the cell ant density of plasma membranes becauseof their high consurface. Furthermore, intracellular cholesterol is not a sub- centration of cholesterol (cf. Ref. 6). We corroborated this strate for cholesterol oxidase evenin cell homogenates?Pre- premise by demonstrating that oxidation of cholesterol to sumably, this is because the enzyme does not attack choles- cholestenone with cholesterol oxidase abolishesthe increase terol in membranes witha cholesterol content below a critical in density induced in plasma membranes by digitonin (Figs. level (7,19). (iv) Cholesterol is ubiquitous among mammalian 1 and 4). Similar results were obtained in erythrocyte ghost plasma membranes.Even where it is not normally present or, membranes? Because intracellular but not cell-surface choas in liver, where there is significant cholesterol within the lesterol is shifted in density by digitonin following cholesterol cell, both exogenous radiocholesterol and cholestenone gen- oxidase treatment of labeled cells (Fig.41, we infer both that erated by cholesterol oxidasetreatment may be used as uni- the intracellular cholesterol isnot associated withthe plasma versal plasma membrane-specific labels. (v) The determina- membranes and that itresides in cholesterol-richmembranes. tion of cholestenone is as simple and sensitive as thatof 5'- By the criteria of native density and digitonin density shift, nucleotidase; the exogenous radiocholesterol labelis simpler we surmise that theintracellular cholesterol is not in typical endoplasmic reticulum, Golgi, or any other major organelle and more sensitive to assay than most enzymes. A principal finding of this study is that thepeak buoyant membrane characterized to date, although it could residein a these prior to density of HMG-CoA reductase was 1.18 g/cm3, while newly specialized membranein continuity with one of synthesized 'lanosterol and cholesterol both equilibrated at homogenization. That is, the buoyant membranes bearing

Cholesterol in Intracellular Membranes intracellular lanosterol and cholesterol could derive from a special region of the endoplasmic reticulum or Golgi apparatus. If there is a specific cytoplasmic membrane bearing newly synthesized cholesterol, it must be minor. It can contain no more than 5% of the cellular cholesterol mass (3) and, given the possible accumulation of cholesterol within lysosomes and cholesterol ester droplets (25) as well as other intracellular membranes, it could carry much less. Since the plasma membrane represents about a third of this cell’s phospholipid,2 the intracellular cholesterol-rich membrane probably contains less than 1%of the cellular membrane mass. Our data suggest that a pathway may exist in which cholesterol biosynthesis and movement to the plasma membrane are coordinated, While HMG-CoA reductase is confined to the smooth endoplasmic reticulum (13), its product, mevalonate, is water-soluble. Squalene is presumably the first sterol precursor which wouldaccumulate appreciablyin membranes. Since we have found that squalene epoxidase has the same buoyant density profile as HMG-CoA reductase in rat liver homogenates? we infer that squalene epoxide and perhaps other early membrane-bound sterol precursors arise in the smooth endoplasmic reticulum. We suggest that, ata step not yet determined, one or more later sterol intermediates are brought to specialized buoyant membranes, perhaps via a carrier protein (22) or through the flow and fusion of membrane vesicles. There, cholesterol synthesis is completed. Because sterol biosynthesis is unidirectional, cholesterol translocation is vectorial. If these precursor membranes are (or become upon homogenization)small vesicles,they might fractionate with microsomes and thus have been ascribed,to the endoplasmic reticulum in earlier work (12). The last step in the process presumably is fusion of the cholesterol-rich precursor membranes with the plasma membrane. It may thus not be a coincidence that thecholesterol-bearingintracellular membranes often have the same buoyant density asthe plasma membrane (e.g. Figs. 2 and 3); they may be similar physically. These special membranes could also be related to those bearing plasma membrane precursor phospholipids and proteins. There must be a mechanism by which cholesterol is concentrated to itshigh value in the plasma membrane (approximately 0.9 mol/mol phospholipid’). The plasma membrane bilayer probably grows by the accretion of precursor vesicles, and these must have a mature cholesterol content; otherwise, “pumping” of plasma membrane cholesterol or phospholipid molecules against their concentration gradients must be postulated. The combined biosynthetic/topogenetic pathway described above can provide such a mechanism. The biosynthesis of late sterol intermediates in a discrete membrane compartment would of itself drive their local concentration in the bilayer to a high level, and lipid need not be pumped. Since thereis no detectable “backflow” of sterols in the cytoplasm of these cells (5,26), the mature level of cholesterol

15597

could readily be achieved in such a precursor vesicle just prior to its incorporation into the plasma membrane. The mechanism of accumulation of cholesterol at the cell surface thus may be analogous to other topogenetic processes:the covalent modification of a mobile precursor of a membrane-associated molecule immobilizesand thusconcentrates it at itsdestination. An example isthe placement of acidhydrolases in lysosomes (27). Acknowledgments-We wish to thank H. J. G. Matthies, B. V. Ramos, and M. F. Muraski for expert technical assistance and G. Byrd for typing the manuscript.

REFERENCES 1. Farquhar, M. G., and Palade, G. E. (1981) J. Cell Bwl. 91, 7751035 2. Rothman, J. (1981) Science 213,1212-1219 3. Lange, Y.,and Ramos, B.V. (1983) J. Biol. Chem. 258, 1513015134 4. DeGrella, R. F., and Simoni, R. D. (1982) J. Bid. Chem. 257, 14256-14262 5. Lange, Y.,and Matthies, H. J. G. (1984) J. Biol. Chem. 259, 14624-14630 6. Wibo, M., Thines-Sempoux, D., Amar-Costesec, A., Beaufay, H., and Godelaine, D. (1981) J. Cell Bwl. 89,456-474 7. Lange, Y.,Matthies, H., and Steck, T. L. (1984) Biochim. Biophys. Acta 769,551-562 8. Harrison, E. H., and Bowers, W. E. (1983) J. Bwl. Chem. 258, 7134-7140 9. Brown, M. S., Dana, S. E., and Goldstein, J. L. (1973) Proc. Nutl. Acad. Sci. U. S. A. 70,2162-2166 10. Avruch, J., and Wallach, D. F. H. (1971) Biochim. Biophys. Acta 233,334-347 11. Lange, Y.,Dolde, J., and Steck, T. L. (1981) J. Biol. Chem. 256, 5321-5323 12. Chesterton,’C. J. (1968) J. Biol. Chem. 243, 1147-1151 13. Orci, L., Brown, M. S., Goldstein, J. S., Garcia-Sequra, L.M., and Anderson, R. G. W. (1984) Cell 36,835-845 14. Hino, Y.,Asano, A., Sato, R., and Shimizu, S. (1978) J. Biochem. 83,909-923 15. Hino, Y., Asano, A., and Sato, R. (1978) J. Biochem. 83,935942 16. Orci, L., Montesano, R., Meda, P., Malaisse-Lagae, F., Brown, D., Perelet, A., and Vassalli, P. (1981) Proc. Nutl. A d . Sci. U. S. A. 78, 293-297 17. Widnell, C. C., Schneider, Y. J., Pierre, B., Baudhuin, P., and Trouet, A. (1982) Cell 28, 61-70 18. Tartakoff, A. M. (1980) Int. Rev. Exp. Puthol. 22,227-251 19. Pal, R., Barenholz, Y., and Wagner, R. R. (1980) J. Biol. Chem. 255,5802-5806 20. Yamamoto, S., and Bloch, K. (1970) J. Biol. Chem. 245, 16701674 21. Srikantaiah, M. V., Hansbury, E., Loughran, E. D., and Scallen, T. J. (1976) J. Biol. Chem. 251, 5496-5504 22. Dempsey, M. E. (1984) Curr. Top. Cell. Regul. 2 4 , 6 3 4 6 23. Friedlander, E. J., Caras, I. W., Lin, L. F. H., and Bloch, K. (1980) J. Bid. Chem. 255,8042-8045 24. Thines-Sempoux, D., Amar-Costesec, A., Beaufay, H., and Berthet, J. (1969) J. Cell Biol. 43, 189-192 25. McGookey, D.J., and Anderson, R.G. W. (1983) J. Cell BWZ. 97, 1156-1168 26. Wattenberg, B. W., and Silbert, D. F.(1983) J. Bwl. Chem. 258, 2284-2289 27. Sly, W. S., and Fischer, H. D. (1982) J. Cell Biochem. 18, 67-85