Obligatory Role of Cholesterol and Apolipoprotein E in the Formation of Large Cholesterol-enriched and Receptor-active High Density Lipoproteins*

The formation of large cholesterol-enriched high density lipoproteins (HDLI/HDL,) from typical HDL3 requires 1ecithin:cholesterol acyltransferase activity, additional cholesterol, and a source of apolipoprotein (apo-) E. The present study explores the role of apo-E in promoting HDLJHDL, formation and in imparting to these lipoprotein particles the ability to interact with the apo-B,E(low density lipoprotein (LDL)) receptor. Incubation of normal canine serum with cholesterol-loaded mouse peritoneal macrophages resulted in the formation of HDL1/HDL, that competed with 12’I-LDL for binding to the apo-B,E(LDL) receptors on cultured human fibroblasts. Cholesterol efflux from macrophages was necessary because incubation of normal canine serum with nonloaded macrophages did not cause HDLl/HDL, formation. However, cholesterol delivery to the serum was not sufficient to result in HDLI/ HDL, formation. Apolipoprotein E had to be available. Incubation of apo-E-depleted canine serum with cho-lesterol-loaded 5774 cells, a macrophage cell line that does not synthesize apo-E, demonstrated that no HDL1/ HDL, formation was detected even in the presence of significant cholesterol efflux. However, addition of ex- ogenous apo-E radioimmunoassay previously Fibroblast Binding Assay-The receptor binding activities of the lipoprotein fractions isolated by Pevikon block electrophoresis were determined by measuring the ability of the fractions to compete with lZ5I-LDL for binding to apo-B,E(LDL) receptors on cultured fibroblasts. Binding assays were performed at 4 “C as previously described (29). fibroblasts (0.9 X lo4) were grown in 16-mm wells in DMEM with 10% fetal calf serum. After 5 days, the medium was replaced with DMEM containing 10% (v/v) lipoprotein-deficient serum, and incubation was continued for 48 h at 37 “C. The cells were then used in the binding assays. The medium used in the binding assays was composed of DMEM containing 25 mM Hepes and 10% lipoprotein-deficient serum. The fibroblasts were incubated at 4 “C in medium containing varying amounts of the lipoprotein fractions and ‘=I-LDL at 2 pg of protein/ml for 2 h. The cells were then washed three times with cold, phosphate-buffered saline containing 2% bo- vine serum albumin. This was followed by two 10-min washes at 4 “C using the same buffer. The cells were then washed once with phos- phate-buffered saline alone and then dissolved with two 0.5-ml ali-quots of 0.1 M NaOH. The radioactivity contained in the 1-ml sample was measured in a y counter. Nonspecific binding was defined as the amount of lZ5I-LDL bound in the presence of 200 pg of unlabeled LDL protein/ml. The amount of radioligand specifically bound in the presence of lipoprotein fractions was expressed as a per cent of maximum specific binding, i.e. the amount of T-LDL specifically bound in the absence of any and triglycerides)

The formation of large cholesterol-enriched high density lipoproteins (HDLI/HDL,) from typical HDL3 requires 1ecithin:cholesterol acyltransferase activity, additional cholesterol, and a source of apolipoprotein (apo-) E. The present study explores the role of apo-E in promoting HDLJHDL, formation and in imparting to these lipoprotein particles the ability to interact with the apo-B,E(low density lipoprotein (LDL)) receptor. Incubation of normal canine serum with cholesterolloaded mouse peritoneal macrophages resulted in the formation of HDL1/HDL, that competed with 12'I-LDL for binding to the apo-B,E(LDL) receptors on cultured human fibroblasts. Cholesterol efflux from macrophages was necessary because incubation of normal canine serum with nonloaded macrophages did not cause HDLl/HDL, formation. However, cholesterol delivery to the serum was not sufficient to result in HDLI/ HDL, formation. Apolipoprotein E had to be available. Incubation of apo-E-depleted canine serum with cholesterol-loaded 5774 cells, a macrophage cell line that does not synthesize apo-E, demonstrated that no HDL1/ HDL, formation was detected even in the presence of significant cholesterol efflux. However, addition of exogenous apo-E to the serum during the incubation with cholesterol-loaded 5744 cells promoted the formation of large receptor-active HDLl/HDL,. The receptor binding activity of these particles produced in vitro correlated with the amount of apo-E incorporated into the HDLJHDL,. Apolipoproteins A-I and C-I11 were ineffective in promoting HDLJHDL, formation; thus, apo-E was unique in allowing HDL1/HDLC formation. These results demonstrate that when lecithinxholesterol acyltransferase activity, cholesterol, and apo-E are present in serum, typical HDL can be transformed in vitro into large cholesterol-rich HDLI/ HDL, that are capable of binding to lipoprotein receptors.
The inverse correlation between plasma high density lipoprotein (HDLl) levels and the development of coronary artery disease (1-3) has sparked considerable interest in the role of * 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 solely to indicate this fact. HDL in cholesterol homeostasis. The postulated antiathero: genic effect of HDL may involve its ability to promote the efflux of cholesterol from peripheral tissues and transport the cholesterol to the liver for excretion (for review see Refs. 4 and 5). In 1968, Glomset (6) proposed the concept of reverse cholesterol transport, but there is a paucity of experimental evidence defining the mechanism whereby HDL could participate in this process. However, it has been shown that HDL can acquire cholesterol from cells (7)(8)(9)(10)(11)(12), that the HDL become enlarged when enriched with cholesterol and cholesteryl esters (which promote an apparent expansion of the cholesteryl ester-rich core), and that during this process the HDL particles acquire apolipoprotein (apo-) E (12). The presence of apo-E could direct the metabolism of these cholesteryl ester-enriched HDL particles. Apolipoprotein E mediates lipoprotein uptake by extrahepatic and hepatic apo-B,E(LDL) receptors and hepatic apo-E receptors (for review see Refs. 5 and 13) and binds with very high affinity to both the apo-B,E(LDL) and the hepatic apo-E receptors; apo-E could therefore efficiently mediate the uptake of these lipoproteins.
The HDL can be fractionated into several subclasses using a variety of techniques. However, for discussion in this paper, it would be useful to consider two metabolically distinct subclasses of HDL, HDL-without apo-E (HDL2 and HDL3) and HDL-with apo-E (HDL1 and HDL,) (see Refs. 12 and 14 for a discussion of this nomenclature). The HDL-with apo-E can be isolated because they bind to heparin-Sepharose affinity columns (14), or in certain species, such as the dog, the HDL-with apo-E can be observed as distinct lipoprotein subclasses by Pevikon block electrophoresis (15). The HDL-with ?PO-E (HDL1/HDLc) are the largest of the HDL (-120-300 A in diameter). They are enriched in cholesteryl ester and possess apo-E (12, 16), usually in association with apo-A-I (sometimes along with apo-A-I1 and the C apolipoproteins). By virtue of the presence of apo-E, this subclass of HDL is capable of interacting with the apo-B,E(LDL) and apo-E receptors. The HDL-without apo-E are the smaller HDL (-70-100 A). They possess apo-A-I as their principal apolipoprotein and are thought to be the precursors of the larger, cholesteryl ester-rich HDL-with apo-E (12).
Several lines of evidence support the concept that the HDLwithout apo-E (HDL2/HDL3) can be converted to the larger, cholesteryl ester-enriched HDL-with apo-E. When canine serum is incubated with an exogenous source of cholesterol (cholesterol-Celite particles or cholesterol-loaded mouse peritoneal macrophages), Gordon et al. (12) have shown that HDL2/HDL3 are reduced in concentration as the HDLJHDL, concentration increases. In addition, in experiments using canine lZ5I-HDL3, it was demonstrated that these particles were converted to HDLJHDL, as they acquired cholesterol (12). In other studies, Schmitz et al. (17) showed that incubating human serum for 24 h at 37 "C resulted in the formation of HDLr-like lipoproteins and a marked reduction in the presence of HDL3. This conversion required lecithin:cholesterol acyltransferase, and the production of cholesteryl esters appeared to drive the reaction toward formation of cholesteryl ester-enriched HDL1. Gordon et al. (12) demonstrated that the formation of HDLl/HDL in. vitro, using cholesterol-Celite as a source of cholesterol, also required the presence of 1ecithin:cholesterol acyltr~sferase. The proposed role of 1ecithin:cholesterol acyltransferase in the conversion of HDL2/HDLs to HDL,/HDL is further supported by studies that have compared the cholesteryl ester fatty acid composition of the HDL lipoproteins in swine and dogs (5, 18). The fatty acid compositions of the cholesteryl esters of the HDL2/HDLs and the HDLJHDL, were very similar, and the predominant cholesteryl fatty acid of both was linoleate. Linoleate is the principal fatty acid that appears in cholesteryl esters formed by the t r~s e s t e r i~c~t i o n reaction of lecithin:cholesterol acyltransferase. In contrast, the cholesteryl ester fatty acid in very low density lipoproteins, intermediate density lipoproteins, and LDL in both swine and dog is oleate, the principal product formed by the intracellular acyl-CoAcholesteroi acyltransferase reaction.
The present study examined the conditions necessary for the formation of HDL,/HDL, in vitro. Two major questions were addressed are HDL1/HDL, that are formed in vitro capable of binding to lipoprotein cell-surface receptors and is apo-E required for the production of HDLl/HDL,? The results demonstrated that apo-E is required for the formation of the large HDLJHDL, and that apo-E from various sources (exogenously added, redistributed from other lipoproteins, or newly synthesized by cultured macropha~es) can participate in this process. Finally, it was found that HDL1/HDL, particles, regardless of their source of apo-E, are able to bind to the apo-B,E(LDL) receptors.

MATERIALS AND METHODS
Experimental Protocol-A schematic representation of the general protocol followed in the present studies is shown in Fig. 1 peritoneal macrophages or cells of the murine macrophage-like cell line 5774.2 were maintained in culture as described below. The cells were plated in tissue culture flasks 24 h prior to cholesterol loading. Macrophages or 5774 cells were loaded with cholesterol by incubating the cells for 24 or 48 h at 37 "C with acetoacetylated LDL added at a concentration of 100 pg of lipoprotein cholesterol/ml in Dulbecco's modified Eagle's medium (DMEM) (GIBCO, Grand Island, NY). In some studies, acetoacetylated LDL containing [3H]cholesterol, prepared as previously described (191, were added. At the end of the incubation period (24-48 h), the acetoacetylated LDL were removed, and the cholesterol-loaded cells (or nonloaded control cells) were washed three times with D M E~ and then incubated with fresh 10% normal canine serum or with apo-E-depleted serum (see below) for 20 h at 37 "C. In several studies, [3SS]methionine (100 pCi/ml) was added in the serum for incorporation into newly synthesized proteins. Selected studies were performed with the addition of M monensin (Calbiochem-Behring) or 10 ng/ml endotoxin (Sigma) to the 10% canine serum.
The cell-free medium was collected at the end of the incubation period, and the d < 1.006 g/ml fraction was removed by ultracentri-  (15). The Pevikon block was divided into 20 l-cm-wide zones, with divisions beginning at the origin and extending through the zones known to contain the most rapidly migrating HDL. The major cholesterol peak marked the position of H D L on the Pevikon block. The fractions between LDL and the HDL, corresponded to the HDLI/ HDL, fractions, which are labeled 1, 2, and 3 in the various figures.
The position of the lipoproteins vary slightly in different Pevikon blocks. Control and experimental samples were separated on the same Pevikon block to ensure appropriate comparison of the different fractions in each experiment, and fractions of equivalent migration distances were compared to each other. The lipoproteins eluted from the Pevikon were characterized by distribution of radiolabeled proteins and Cholesterol, chemical composition, polyacrylamide gradient gel electrophoresis, and receptor binding activity. The cholesterol content of the fractions and unfractionated serum was measured by the enzymatic spectrophotometric assay from Bio-Dynamics (Boehringer Mannheim), and the protein concentration was determined by the method of Lowry et ai. (20). The various procedures are outlined below. Macrophage Cultures-Mouse peritoneal macrophages were isolated from female Swiss Webster mice (Taconic, NY) 4 days after an intraperitoneal injection of 1 ml of thioglycolate medium (21). The peritoneal cells (10 mi) were suspended in DMEM (supplemented with penicillin G (100 units/ml) and streptomycin sulfate (100 mg/ ml)) at a concentration of 4 X lo6 macrophages/ml and allowed to adhere to 75-cm2 culture flasks (Falcon, Oxnard, CA) for 1.5 h at 37 "C. The nonadherent cells were removed by three washes with DMEM, and the adherent cells were incubated overnight in DMEM supplemented with 10% fetal bovine serum before cholesterol loading.  [1,2,6,7-3H]cholesteryl linoleate (New England Nuclear) by a procedure similar to that used for labeling lipoproteins with the fluorescent probe l,l'-dioctadecyl-3,3,3',3'tetramethylindocarbocyanine percholate (24), except that 0.05 mCi of [3H]cholesteryl 1in.oleate in dimethyl sulfoxide was added for each milligram of lipoprotein protein in place of the fluorescent probe. The LDL were then acetoacetylated with diketene as detailed by Weisgraber et al. (19).
Gradient gel electrophoresis was performed as previously described (25,26). Samples were electrophoresed on 2-16% or 4-30% polyacrylamide gradient slab gels for 24 or 40 h, respectively, using a Pharmacia electrophoresis apparatus in a pH 8.35 buffer (90 mM Tris base, 80 mM boric acid, and 3 mM EDTA). The gels were then stained for 48 h with 0.04% oil red 0 in 60% ethanol at 40 "C. Gels were then destained in 5% methanol and 7% acetic acid and scanned with a Beckman DU-8 spectrophotometer at 510 nm.
Preparation of Apo-E-depleted Serum-Normal canine serum was depleted of apo-E by incubation with the anti-canine apo-E monoclonal immunoglobulin 25D2 (27) complexed to Sepbarose 4B. The apo-E monoclonal antibody was prepared by the procedure previously described (27). To remove the apo-E, canine serum (20 ml) was incubated at room temperature for 2 h with 40 mg of anti-apo-E coupled to Sepharose. The unbound fraction (apo-E-depleted serum) was then used in the experiments. The apo-E level was determined by radioimmunoassay as previously reported (28).
Fibroblast Binding Assay-The receptor binding activities of the lipoprotein fractions isolated by Pevikon block electrophoresis were determined by measuring the ability of the fractions to compete with lZ5I-LDL for binding to apo-B,E(LDL) receptors on cultured fibroblasts. Binding assays were performed at 4 "C as previously described ( 2 9 ) . Human fibroblasts (0.9 X lo4) were grown in 16-mm wells in DMEM with 10% fetal calf serum. After 5 days, the medium was replaced with DMEM containing 10% (v/v) lipoprotein-deficient serum, and incubation was continued for 48 h at 37 "C. The cells were then used in the binding assays. The medium used in the binding assays was composed of DMEM containing 25 mM Hepes and 10% lipoprotein-deficient serum. The fibroblasts were incubated at 4 "C in medium containing varying amounts of the lipoprotein fractions and '=I-LDL at 2 pg of protein/ml for 2 h. The cells were then washed three times with cold, phosphate-buffered saline containing 2% bovine serum albumin. This was followed by two 10-min washes at 4 "C using the same buffer. The cells were then washed once with phosphate-buffered saline alone and then dissolved with two 0.5-ml aliquots of 0.1 M NaOH. The radioactivity contained in the 1-ml sample was measured in a y counter. Nonspecific binding was defined as the amount of lZ5I-LDL bound in the presence of 200 pg of unlabeled LDL protein/ml. The amount of radioligand specifically bound in the presence of lipoprotein fractions was expressed as a per cent of maximum specific binding, i.e. the amount of T -L D L specifically bound in the absence of any competitors.
Calculation of Particle Core Radius and Shell Width-The per cent core volume can be estimated from the composition of the core components (cholesteryl ester and triglycerides) of the various lipoproteins by the relationship where TG, CE, FC, PL, and AP are the concentrations of triglycerides, cholesteryl ester, free cholesterol, phospholipid, and apolipoproteins, and Ul, dl, 03, a, and O5 are the partial specific volumes for TG, CE, FC, PL, and AP, respectively. The V,,, is the core volume, and V,,, is the volume for the entire particle. Two assumptions were made in arrivingat the above relationship. First, the core is assumed to contain all of the triglycerides and cholesteryl esters and a fraction of the free cholesterol because free cholesterol has limited solubility in a cholesteryl ester-rich core? Second, the partition coefficient of free cholesterol between the surface and the core is assumed to be 6, which represents the maximum free cholesterol solubility in the core? The values 01 = 1.10, 02 = 1.07, 03 = 0.95, O4 = 0.98; a typical partial specific volume for proteins, = 0.74 (30,31), and the per cent values reported in Table I for the composition of the various fractions may then be used to calculate VcO,/Vtoh,.
Donald Small, personal communication.
The radius of the core can then be calculated by the relationship vco, vat* &td where raM is the radius determined from negative staining electron microscopy (Table I) observations. The shell width may then be described by the relationship -=rt.,@lr,,, = shell width.

Characteristics of H D L Fractions Separated by Pevikon Block Electrophoresis-As
previously demonstrated by Gordon et al. (12), various subclasses of canine HDL can be isolated by Pevikon block electrophoresis. In the present studies, HDL subclasses obtained from the plasma of a cholesterol-fed dog were isolated by Pevikon block electrophoresis and characterized ( Table I). The HDL subclasses from the plasma of a cholesterol-fed dog resemble the fractions obtained after incubation of normal canine serum with a source of cholesterol (to be described below). The d = 1.006-1.087 fraction of the plasma from a cholesterol-fed dog was subjected to Pevikon block electrophoresis as described under "Materials and Methods." The LDL fraction was located 4-7 cm from the origin. One-centimeter fractions extending from the end of the LDL region were obtained, and the lipoproteins were analyzed by SDS-polyacrylamide gel electrophoresis ( Fig. 2 A ) , as well as by polyacrylamide gradient gel electrophoresis (Fig. 2B). The fractions were numbered 1-5, with Fraction 1 located immediately adjacent to the LDL fraction and Fraction 5 located farthest from the origin. Fig. 2B (gradient gel patterns) demonstrates that the size of the HDL decreased as the distance of migration increased. The HDL migrating farthest from the origin (12-16 cm, Fraction 5 ) represented predovinantly HDL3, i,e. HDL equivalent to the small (-70-100 A in diameter), dense particles that lack apo-E. Fractions migrating closer to the origin (7-11 cm) were larger particles (HDLl/HDL,) that floated at a lower density and contained more cholesterol ( Table I) and apo-E ( Fig. 2A). The largest HDL particles, representing Fraction 1, are the HDL,. Fig. 2A illustrates the apolipoprotein content of the various lipoprotein preparations. The largest HDL (Fraction 1) contained the most apo-E. The apo-A-I content increased with decreasing particle size. The chemical composition of each fraction was determined, and the particle sizes were measured from electron micrographs ( Table I). The larger particles contained more free cholesterol, as well as cholesteryl ester, whereas the percentage of protein was lower as compared to that of the smaller particles.

Formation of Receptor-active HDLl and HDL, by Incubation of Normal Canine Serum with
Cholesterol-loaded Macrophages -Thioglycolateelicited mouse peritoneal macrophages were loaded with cholesterol by incubation with human acetoacetylated LDL, as described under "Materials and Methods." After the cells were loaded with cholesterol, normal canine serum (10%) in tissue culture medium was added to promote cholesterol efflux and to serve as a cholesterol acceptor. After 20 h of incubation with the cholesterol-loaded macrophages (or nonloaded, control macrophages), the medium containing the 10% canine serum was removed from the cells. "The size of the particles was determined from negative staining electron microscopy observations. In each fraction, the diameters of 200 free-standing particles that were observed in the micrographs were measured. The mean values and standard deviations are mesented.
I, Calculated as detailed under "Materials and Methods." ND, not detectable.

A.
HMc the d > 1.006 fraction after incubation with cholesterol-loaded or nonloaded macrophages is shown in Fig. 3. As previously demonstrated (12), incubating canine serum with cholesterolloaded macrophages results in a shift of the lipoprotein cholesterol distribution to slower-migrating lipoproteins (HDLwith apo-E, i.e. HDLl and HDL,) (Fig.  3). Macrophages labeled with ["S]methionine secrete newly synthesized ['"SI methionine-labeled proteins, one of which has been identified as apo-E (32, 33). As shown in Fig. 3, the cholesterol-loaded macrophages secreted more ["S]methionine-labeled proteins than did the nonloaded cells, and the majority of the newly synthesized ["S]methionine-labeled proteins migrated in the HDLI/HDLc region of the Pevikon block. Fluorography of SDS-polyacrylamide gels of the d > 1.006 fractions revealed that apo-E accounted for a large portion of the observed augmentation in protein secretion induced by cholesterol loading (Fig. 3, inset). Densitometric scans of these fluorograms revealed that apo-E accounted for 35% of the total ["'SS]methionine-labeled proteins secreted by the cholesterol-loaded cells and only 14% of the proteins secreted by the nonloaded cells. Because the total secretion of ["SS]methionine-labeled proteins by the cholesterol-loaded cells was approximately 2-fold greater than that of the nonloaded cells, and at least 2-fold more of the labeled protein was apo-E, the actual increase in apo-E secretion was at least 4-fold greater in the cholesterol-loaded macrophages. Thus, since a major portion of the secreted proteins was apo-E, the majority of the [:15SS]methionine-labeled proteins would indeed be located in the HDLI/HDLc region (between LDL and HDLJ on the Pevikon block.

Am-A-I--
The lipoproteins in the HDLJHDL, region of the Pevikon block were isolated and tested for receptor binding to cultured human fibroblasts. As shown in Fig. 4 (representing a typical experiment), the HDL, subfraction (equivalent to Fraction 3, Fig. 3) isolated from serum after incubation with cholesterolloaded cells was much more effective than an equivalent fraction obtained from serum incubated with nonloaded cells in competing with '""ILDL for binding to fibroblast apo-  ]methionine-labeled apo-E than was present in the HDL, obtained from serum incubated with nonloaded cells (Fig.  4, inset). Densit,ometric scans of the fluorogram revealed that there was approximately 40-fold more [%]methionine-labeled apo-E in Fraction 3 from serum incubated with cholesterol-loaded macrophages than in the corresponding fraction from serum incubated with nonloaded macrophages. As shown in Fig. 4, Fraction 3 from the cholesterol-loaded macrophages also demonstrated about 40-fold more receptor binding activity.
In addition, it was possible to demonstrate that the HDLJ HDL, subfractions obtained from the serum after incubation with cholesterol-loaded cells possessed variable binding activity that correlated qualitatively with the amount of [?3] methionine-labeled apo-E present. As shown in Fig. 5 , Fractions 1 and 2 (see Fig. 3 for migration characteristics), which represented the largest HDLJHDL,, were more potent in competing with '*'I-LDL for receptor binding to fibroblasts than Fraction 3, which represented the smaller, less cholesterol-rich HDLl (12). The average concentrations of Fractions 1-3 (from seven experiments) that were necessary to inhibit "'I-LDL binding activity by 50% were 0.1, 0.31, and 2.7 pg of cholesterol/ml, respectively. In contrast, equivalent fractions (1-3) obtained from serum incubated with nonloaded cells required concentrations of 30,40, and >lo0 pg of cholesterol/ ml, respectively, to inhibit '*'I-LDL binding activity by 50%. The binding activity of the HDLI/HDL, fractions appeared to correlate with the apo-E content of these particles (Fig. 5 ,  inset). Fractions isolated from the HDL2/HDLn region of the Pevikon block were inactive (data not shown) and lacked the presence of detectable, radiolabeled apo-E.

Effect of Inhibition of Macrophage Apo-E Secretion on the Generation of HDLJHDL, Fractions-The HDLI/HDL, frac-
tions that demonstrated receptor binding activity could have obtained apo-E by transfer from other plasma lipoproteins, as well as from the macrophages (as evidenced by the presence of [:"SS]methionine-labeled apo-E) (34). To determine the relative importance of newly synthesized macrophage apo-E versus transferred apo-E in the formation of the HDLl/HDL,, apo-E secretion by cholesterol-loaded macrophages was inhibited by the addition of monensin or endotoxin to the tissue culture medium. Monensin alters receptor-mediated endocytosis and cellular protein secretion (35, 36). Endotoxin produces a variety of responses in macrophages, including the inhibition of apo-E secretion (37, 38). Addition of monensin M ) or endotoxin (10 pg/ml) to cholesterol-loaded macrophages did not significantly affect cholesterol efflux. The total cholesterol concentration in the serum after incubation with the cholesterol-loaded macrophages in the absence or presence of monensin or endotoxin was significantly elevated compared to the cholesterol concentration in serum incubated with nonloaded macrophages (total cholesterol concentration was increased from 1.84 mg/ml after incubation with nonloaded macrophages to 2.3 mg/ml after incubation with cholesterol-loaded macrophages in the presence of M monensin). Basu 6, W). There was a marked decrease in [:"S]methionine-labeled apo-E in these fractions after the addition of either monensin (Fig. 6,O) or endotoxin (data not shown). The distribution of ["sS]methionine-labeled apo-E in the d = 1.006-1.21 fraction from serum incubated with nonloaded, untreated macrophages (Fig. 6, 0 ) is shown for comparison. In an additional study, it was found that monensin treatment of the cholesterol-loaded cells markedly reduced the secretion of total ['"Ss] methionine-labeled protein in the d = 1.006-1.21 serum fraction by -60%. Furthermore, ['"S]methionine-labeled apo-E secretion was reduced by 86% after monensin treatment as assessed by densitometric scanning of a fluorogram of an SDS-polyacrylamide gel electrophoretic pattern of the d = 1.006-1.21 serum fraction.
The HDL,/HDL, fractions from these studies were tested for receptor binding activity. As shown in Fig. 7 (data from a representative experiment), Fraction 3 from the serum incubated with cholesterol-loaded macrophages treated with monensin or endotoxin demonst.rated potent receptor binding activity, although not as much as an equivalent fraction from cholesterol-loaded macrophages. These results are consistent with data obtained in three independent studies. Therefore, when the secretion of newly synthesized mouse macrophage apo-E was markedly inhibited, cholesterol delivery to canine serum st,ill resulted in the formation of HDLJHDL, fractions that demonstrated appreciable receptor binding activity. As will he shown, this binding activity was mainly due to the transfer of apo-E from other serum lipoproteins to the newly formed HDLJHDL,.
Formation of HDLl/HDL, following Incubation of Serum with J774 Cdls-Alt.hough monensin and endotoxin inhibited i \ the secretion of apo-E by mouse peritoneal macrophages, significant quantities of apo-E were still secreted. To elimimte the possibility of macrophage apo-E production, a mouse macrophage-like cell line (5774) that does not secrete apo-E was used (40). The procedures described above for the analysis Qf HDLl/HDL, formation were used to demonstrate that the 5774 cells released cholesterol and resulted in the generation of HDLl/HDL, particles that could be isolated by Pevikon Mock electrophoresis. The distribution of lipoprotein cholestvrol in the d > 1.006 fractions was essentially identical to that shown in Fig. 2.
The isolated HDLl/HDLc fractions obtained by Pevikon Mock electrophoresis were tested for receptor binding activity. As shown in Fig. 8 (data from a representative experiment), the HDL,/HDL, fraction (Fraction 1) from the serum incubated with cholesterol-loaded 5774 cells showed significant ability to compete with lZ5I-LDL for receptor binding activity. By comparison, an equivalent fraction from the nonloaded d774 cells showed less binding activity. These results, along with those obtained from the use of monensin and endotoxin, demonstrated that receptor-active HDLI/HDL, could be formed in the absence of apo-E secretion by cells, presumably because apo-E was redistributed from other serum lipoproteins, such as very low density lipoproteins, to the HDLJ HDL,, but not in the absence of cholesterol released from intracellular stores. Inhibitiotz of HDL1/HDLc Formation by Depletion of Apo-E from the Serum-The apo-E in canine serum can be depleted by immunoaffinity chromatography using an anti-canine apo-E-Sepharow column. As determined by radioimmunoassay, 6040% of the serum apo-E could be removed by the immunoaffhity column. Normal canine serum or canine serum depleted of apo-E was added to cholesterol-loaded cells at a protein concentration equivalent to 10% normal canine mum. The d > 1.006 lipoproteins obtained after the incubation were subjected to Pevikon block electrophoresis; the fractions were characterized by polyacrylamide gradient gels stained for lipid and scanned by densitometry. As previously  "Defined according to Gordon et al. (12).
Lipid distribution was determined by densitometric scanning of lipid-stained polyacrylamide gradient gels, and the results were calculated as a per cent of the amount present in control serum, i.e. normal canine serum incubated with non-cholesterol-loaded 5774 cells. The results presented are the per cent change from this control; an increase would be expressed as + and a decrease as -. The amount of 3H-labeled cholesterol released into the normal canine serum and the apo-E-depleted canine serum by the 3H-labeled, cholesterolloaded 5774 cells was similar; the cholesterol level in the apo-Edepleted serum was 75% of that observed in the normal serum. described (12), polyacrylamide gradient gel electrophoresis bas revealed the presence of large (-210 A) and small (-150 A) HDL, and HDL, (-250 A) in dog serum. As shown in Table I1 (data from a typical experiment), when apo-E-depleted canine serum was incubated with the cholesterolloaded 5774 cells, there was no significant production of HDL, or HDL,. However, when normal canine serum was incubated with cholesterol-loaded 5774 cells, an increase in the amount of HDLl and HDL, was observed. These results indicated that a source of apo-E (either from macrophages or from serum) was essential for the formation of the large, cholesterol-enriched HDL,/HDL,.
The HDLJHDL, subfractions obtained from normal serum after incubation with cholesterol-loaded 5774 cells were very potent in competing with lZ5I-LDL for binding to the apo-B,E(LDL) receptors of cultured fibroblasts (Fig. 9A, H). However, an equivalent fraction obtained after incubation of apo-E-depleted serum with the cholesterol-loaded 5774 cells displayed little or no receptor binding activity (Fig. 9A, 0). Parallel experiments using mouse peritoneal macrophages (instead of 5774 cells) revealed that the HDLJHDL, subfractions obtained from normal canine serum after incubation with the cholesterol-loaded peritoneal macrophages displayed potent receptor binding activity (Fig. 9B, a). Furthermore, when apo-E-depleted serum was incubated with the cholesterol-loaded peritoneal macrophages, the "HDL,/HDL, subfractions possessed almost equal receptor binding activity (Fig. 9B, 0). Thus, it appears that the newly synthesized apo-E produced by the cholesterol-loaded peritoneal macrophages actively participated in the formation of HDLJHDL, and was responsible for the receptor binding activity.

Effect of Exogenously Added Apolipoproteins on the Formation of ~e c e p t o r -~t i v e
HDL-The results from the experiments described above suggested that apo-E from plasma or apo-E synthesized by macrophages could be used in the formation of cholesterol and apo-E-enriched HDL. To determine whether or not exogenously added apo-E could be incorporated into these particles, cholesterol-loaded 5774 cells were incubated with apo-E-depleted canine serum with or without the addition of human apo-E3 (2.5 pg of pro~in/ml of serum).
The d > 1.006 fractions obtained after the incubation were analyzed by 2-16% polyacrylamide gradient gel electrophoresis to obtain a better resolution of the HDLJHDL, region of the gel. The results from a representative experiment are shown in Fig. 10. In the absence of added human apo-E, there was a paucity of large HDL subclasses (HDL, and HDLJ (Fig. 10, ~~~e  d  line). When human apo-E was added to the canine apo-E-depleted serum, cholesterol delivery to the serum resulted in an increase in the formation of HDL, and HDL, (Fig. 10, solid line).
The receptor binding activity of the Pevikon fractions was assayed (Fig. 11). The results confirmed that in the absence of added apo-E virtually no HDL~/HDL= were formed; therefore, no binding activity was observed (Fig. 11B, El). When apo-E was added, large HDL were formed, and they were highly active in competing with lZ5I-LDL for receptor binding (Fig. 11B, m). The addition of human apo-E to normal serum incubated with cholesterol-loaded 5774 cells resulted in increased binding activity of the HDLJHDL, fraction (Fig.   1lA).
To determine the specificity of the effect of apo-E on the formation of large HDL during cholesterol addition, other apolipoproteins were also tested. Apolipoproteins A-I, , and E2 were added separately to apo-E-depleted canine serum at concentrations of 2.5 Gg/ml, and the serum was incubated with cholesterol-loaded 5774 cells. The distribution of lipoproteins in the d > 1.006 fractions obtained after incubation was analyzed on 2-1676 and 4-30% polyacrylamide gradient gels and quantitated by densitometric scanning of the lipidstained gels. The results are summarized in Table 111. The addition of apo-E3, as demonstrated above, resulted in a substantial increase in HDL, and HDLl formation during cholesterol delivery to the apo-E-depleted canine serum. Apolipoprotein E2(Arglbs "-* Cys) was also effective in inducing the formation of large HDL. The addition of apo-A-I or apo-C-111 did not result in a significant increase in HDL, or HDLI. However, the addition of apo-A-I did result in the f o~a t i o n of HDLz and produced an increase in HDLs. These experiments therefore support the conclusion that apo-E (either  Gordon et al. (12).
Lipid distribution was determined by densitometric analysis of lipid-stained polyacrylamide gradient gels, and the results were calculated as average per cent increase in total area of each lipoprotein subclass over the area of each subclass in apo-E-depleted canine serum incubated with 5774 cells without the addition of human apolipoproteins.
ND, not determined.
apo-E3 or apo-E2) is necessary for the formation of large HDL species during cholesterol delivery to canine serum.

DISCUSSION
Data from tissue culture studies have shown that HDL can participate in the removal of cholesterol from cells (7)(8)(9)(10)(11)(12). This might constitute an initial event in the process referred to as reverse cholesterol transport. According to this hypothesis, cholesterol leaves the cells as free cholesterol, which may be acquired by HDL either through the binding of the HDL particles to the cell surface (41,42) or by diffusion of the cholesterol through the unstirred aqueous layer surrounding the cell and subsequent incorporation into the HDL particle (43). Eventually, the majority of the free cholesterol becomes esterified through the action of 1ecithin:cholesterol acyltransferase (6).
It has been shown that the HDL particles that acquire cholesterol from an exogenous source become cholesterol enriched, increase in size, and acquire apo-E (12). The studies by Gordon et al. (12) have demonstrated that the smallest HDL (HDL3) serve as precursors for the formation of the cholesterol-rich HDLl and HDL,, which acquire apo-E as they become cholesterol enriched. The apo-E that associates with these particles could originate from other plasma lipoproteins (34) or may be newly synthesized and released from macrophages (32) or other peripheral cells (44).
The results of gradient gel electrophoresis, which was used to separate HDL particles on the basis of size, have demonstrated that three distinct subclasses of HDL-with apo-E could be formed when cholesterol was provided to serum either from cholesterol-Celite particles or cholesterol-loaded peritonea! macrophages (12). These included osmall HDLl (140-160 + in diameter), large HDLl (180-200 A), and HDL, (240-270 A). These size increments appear to correlate with the formation of one, two, or three layers of cholesteryl esters within the core of the small HDL1, large HDL1, and HDL,, respectively (12). It has also been demonstrated that lecithin:cholesterol acyltransferase activity was necessary for the enrichment of cholesteryl esters in the formation of the HDLJHDL, (12). The activation of 1ecithin:cholesterol acyltransferase by @-mercaptoethanol enhanced the formation of HDLJHDL,. Conversely, when the 1ecithin:cholesterol acyltransferase was inhibited by N-ethylmaleimide, HDL1/HDLc formation was profoundly depressed. The cholesteryl esters formed in the HDLJHDL, were not due to cholesteryl ester exchange activity since canine serum is noted for its lack of this enzymatic activity (45).
The presence of apo-E on these HDL particles could be expected to profoundly alter the metabolism of the lipoproteins by virtue of the ability of apo-E-containing lipoproteins to interact with both apo-B,E(LDL) and apo-E receptors (13). Theoretically, these lipoproteins could participate in the redistribution of cholesterol among the various cells that possess the apo-B,E(LDL) receptors, or they could be taken up by the liver by apo-B,E(LDL) or apo-E (remnant) receptors, subsequently eliminating cholesterol from the body. Alternatively, the cholesteryl esters may be transferred from the HDL to very low density lipoproteins or LDL, and the cholesterol could then be delivered to the liver by this indirect pathway. Either the direct or indirect delivery of the HDL cholesterol to the liver would be compatible with the reverse cholesterol transport hypothesis (for review see Refs. 4, 5, 13, and 46).
Large HDL enriched in cholesterol, cholesteryl esters, and apo-E are present in the plasma of animals and humans (5,18). The HDL-with apo-E, which resemble the small HDL, formed in vitro, are present in significant concentrations in human cord blood (28); however, they represent a minor constituent of the HDL in adult humans (47). In patients with abetalipoproteinemia, a condition in which there is an absence of very low density lipoproteins and LDL and in which the HDL are the major cholesterol-carrying lipoproteins, the HDL-with apo-E (HDLl) represent a major subclass of the HDL (28, 48). In these patients, it appears that the HDL-with apo-E may be major lipoproteins responsible for transport and redistribution of cholesterol among peripheral cells possessing the apo-B,E(LDL) receptors or hepatic cells possessing both apo-B,E(LDL) and apo-E receptors. Diets high in fat and cholesterol induce an increase in the concentration of HDL-with apo-E in both humans (49) and various animals (4, 5, 18). In certain species, high fat and high cholesterol diets induce marked increases in the large, cholesteryl ester-rich HDL,, in which apo-E becomes the major or exclusive apolipoprotein constituent (4,5,23).
The present study establishes that the apo-E-containing HDLJHDL, that are formed in vitro in normal canine serum possess the ability to bind to the apo-B,E(LDL) receptors of fibroblasts. Furthermore, it has been shown that the tissue or species source of the apo-E does not make a significant difference. In some experiments, the apo-E acquired by the cholesterol-enriched HDLJHDL, was newly synthesized and secreted by mouse peritoneal macrophages. In other experiments using 5774 macrophages, which do not synthesize apo-E, the apo-E acquired by the HDLl/HDLc was either redistributed from other serum lipoproteins or was derived from apo-E added directly to serum previously depleted of apo-E. Likewise, monensin-or endotoxin-treated peritoneal macrophages produced much less (7to &fold) apo-E than untreated macrophages, and in these studies, the HDLJHDL, appeared to acquire the apo-E from other lipoproteins. In all of these cases, the receptor binding activity of the HDLl/HDL, correlated with the presence of apo-E, regardless of its source. These studies also demonstrate an important role for apo-E in the formation of cholesteryl ester-enriched HDLl/HDL,. The HDLl/HDL, could be formed only under conditions in which apo-E was synthesized by macrophages, redistributed by other serum lipoproteins, or added exogenously. In control studies in which apo-A-I or apo-C-I11 was substituted for apo-E, the addition of these apolipoproteins did not result in the formation of the HDLJHDL, (Table 111). Human apo-EZ was nearly as effective as human apo-E3 in eliciting the formation of the cholesterol-enriched HDLJHDE. However, apo-E in the medium is not necessary for cholesterol release from cells. When apo-E-deple~d serum was incubated with cholesterolloaded 5774 cells that did not synthesize apo-E, the cells still released cholesterol; however, virtually no HDL,/HDL, were formed. Under these conditions, the cholesterol released (-75% of the level released in the presence of apo-E-containing serum) was principally associated with HDL~/HDL~. Hence the relative absence of apo-E in the serum did not substantially affect the release of cholesterol by cholesterolloaded cells.
The formation of large HDLJHDL, requires the acquisition of apo-E and cholesterol and the subsequent formation of cholesteryl esters. Previously, Gordon et al. (12) speculated that as the particles increased in size, there was an increase in the cholesteryl ester layers from 1 to 3 within the core of the HDLJHDL,. An analysis of the particle size and chemical composition of various HDL subclasses within the plasma of cholesterol-fed dogs (Table I) suggests that, as the particle increases in size, not only does the core expand but also a concomitant increase in the shell or surface constituents occurs. The increase in surface constituents could be accommodated by the shell if the particles assume a nonspherical configuration. For example, a sphere has the smallest surface area to volume ratio while an ellipsoid of equal volume would have a greater surface to volume ratio and would therefore be able to accommodate larger amounts of surface constituents. However, if a spherical shape is retained, the surface constituent could be accommodated only if there is an increase in shell width (Table I). Of course a combination of shape change and increase in shell width could also allow the accumulation of the larger amounts of surface constituents. Regardless of the exact model, it appears that apo-E is necessary for the formation of an expanded cholesteryl ester core (12,50) and an increase in the shell width or surface area.
In conclusion, these in vitro studies have demonstrated a mechanism whereby cholesteryl ester-enriched HDL can be formed in plasma and participate in reverse cholesterol transport. The acquisition of apo-E by the cholesterol-loaded HDLJHDL, enables the lipoproteins to interact with apo-B,E(LDL) and apo-E receptors. By virtue of the presence of apo-E, these particles can participate in the redistribution of cholesterol to other cells, including those of the liver. Furthermore, it has been established that apo-E is essential for the formation of large HDLl and HDL,. Apolipoprotein E appears to be uniquely capable of eliciting the expansion of the shell and the core of these particles to accommodate free cholesterol and additional layers of cholesteryl ester. Precisely how apo-E functions in the formation of these large HDL particles remains to be determined.