Apolipoprotein A-I Stimulates Secretion of Apolipoprotein E by Foam Cell Macrophages*

Apolipoprotein A-I (apoA-I) overexpression inhibits atherogenesis in mice, and apolipoprotein E (apoE) secreted by foam cell macrophages may exert antiatherogenic effects within the arterial wall. We hypothesized that interaction between apoA-I and apoE contributed to the antiatherogenic properties of apoA-I, and therefore investigated whether apoA-I stimulated secretion of apoE by foam cell macrophages. Cholesterol enrichment of primary murine and human macrophages increased spontaneous apoE secretion 2-fold, as quantified by Western blot and chemiluminescence detection. Human apoA-I caused a further marked increase of apoE secretion from both murine (3.8-fold,p < 0.01) and human (3.2-fold, p = 0.01) foam cells in a time- and concentration- dependent manner, and this increase was confirmed by immunoprecipitation of [35S]methionine-labeled macrophage apoE. The protein synthesis inhibitor cycloheximide, but not the transcription inhibitor actinomycin D, markedly inhibited apoE secretion to apoA-I (73.1 ± 9.8% inhibition at 4 h) and completely suppressed apoE secretion beyond 4 h. Pretreatment of macrophages with Pronase inhibited initial apoA-I-mediated apoE secretion by 70.5 ± 6.5% at 2 h, but by 8 h apoA-I-induced apoE secretion was the same in Pronase-pretreated and non-pretreated cells. Non-apolipoprotein-mediated cholesterol efflux induced by trimethyl-β cyclodextrin did not enhance apoE secretion, whereas phospholipid vesicles inducing the same degree of cholesterol efflux substantially enhanced apoE secretion, and apoA-I and phospholipid vesicles in combination demonstrated additive induction of apoE secretion. We conclude that apoA-I concurrently stimulates apoE secretion and cholesterol efflux from foam cell macrophages and that lipoprotein-derived apoA-I may enhance local secretion and accumulation of apoE in atherosclerotic lesions.

Apolipoprotein A-I (apoA-I, 1 M r 28,000) is the major protein component of HDL, and there is increasing evidence that it contributes to a direct antiatherogenic effect of this lipoprotein. For example, mice transgenic for human apoA-I have decreased susceptibility to atherosclerosis (1,2). Several lines of evidence implicate lipid-poor apoA-I as a particularly important mediator of cholesterol efflux in vivo. There are increased concentrations of lipid-poor apoA-I particles in interstitial fluid and lymphatic fluid (3)(4)(5); pure apoA-I mediates cholesterol efflux in vitro, especially from cholesterol-enriched cells (6 -8); and pre␤-migrating, lipid-poor HDL species are the major initial acceptors of cellular cholesterol in human plasma (9). ApoA-I has also been identified within atherosclerotic lesions by immunohistochemistry (10).
Apolipoprotein E (apoE, M r 34,000) has major roles in the hepatic clearance of triglyceride-rich lipoproteins, and is commonly isolated from VLDL and HDL fractions, and less commonly from LDL fractions, of human plasma (11)(12)(13)(14)(15). ApoA-I is synthesized by the liver and cells of the central nervous system but not by peripheral cells in the arterial wall such as macrophages (16), whereas apoE is also secreted by macrophages, and both synthesis and secretion increase in response to cholesterol accumulation by these cells (17)(18)(19). In vitro, apoE secreted by macrophages is at least in part associated with cell-derived phospholipid and cholesterol (17,20,21). Secretion of apoE by macrophages may thus contribute to spontaneous clearance of cholesterol from macrophages (22), as well as reduce hyperlipidemia and atherosclerosis in apoE knockout (apoE KO) mice transplanted with apoE-secreting macrophages (23,24). ApoE protein and mRNA are abundant within human atherosclerotic lesions, especially in regions rich in monocyte-derived macrophage foam cells (25). In transgenic mice matched for plasma lipoprotein concentrations, secretion of apoE in the artery wall is associated with a reduction in the extent of atherosclerosis (26,27). In apoE KO mice, apoE expression by bone marrow transplanted macrophages alters apoA-I distribution and serum HDL concentration (28), and apoE-deficient macrophages increase atherosclerosis in apoE KO mice indicating local arterial apoE secretion may be antiatherogenic (29). Interestingly, very recent studies have suggested that local macrophage apoE secretion may be proatherogenic (30) or antiatherogenic (31). In addition to inducing local cholesterol efflux (26), apoE secreted by macrophages may have anti-inflammatory and anti-proliferative activity (32,33).
ApoA-I and apoE are both proposed to have important lipid transfer activities in the central nervous system, and apoE is locally secreted during nerve degeneration and repair (34,35). Importantly, Boyles et al. (36) demonstrated that apoA-I, apoE, and monocyte-derived foam cell macrophages were all present during remyelination of damaged nerves. A co-operative transfer of lipid from macrophages by apoA-I and its delivery to the remyelinating nerve via apoE was suggested, but a direct mechanism linking exogenous apoA-I and the secretion of apoE was not demonstrated. Similarly, a possible interaction between exogenous apoA-I and macrophage-secreted apoE has recently been hypothesized to inhibit atherogenesis in the vessel wall in vivo (37).
We here demonstrate that there is a unifying mechanism for the co-localization and interaction of exogenous apoA-I and macrophage-secreted apoE, namely that apoA-I stimulates the secretion of apoE from foam cell macrophages. Primary murine and human macrophages secrete increased quantities of apoE in response to cholesterol loading, but, additionally, there is a further marked increase in secretion of apoE in response to incubation with human apoA-I. The apoE secreted in response to apoA-I is only partially derived from a preformed cell surface pool, most secreted apoE requiring de novo protein synthesis. The direct antiatherogenic effects of apoA-I may thus be augmented by the biological activities of locally secreted apoE.

EXPERIMENTAL PROCEDURES
Reagents-Dulbecco's modified Eagle's medium (DMEM), RPMI 1640, and L-glutamine were supplied by Trace Biosciences, and phosphate-buffered saline (PBS) and penicillin/streptomycin by Sigma Aldrich. White cell buffy coat concentrates (Ͻ24 h ex vivo) and human serum were kindly provided by the New South Wales Red Cross Blood Transfusion Service, Sydney, Australia. All solvents were high performance liquid chromatography (HPLC) grade (Mallinckrodt). Monoclonal mouse anti-human antibodies to human apoE were obtained from Roche Molecular Biochemicals, polyclonal goat antibodies to human apoE were obtained from Fitzgerald, and polyclonal rabbit anti-mouse antibodies to murine apoE were obtained from Biodesign (Kennebunk, ME). Secondary species-specific horseradish peroxidase-linked antibodies, nitrocellulose membranes (0.45 m), and enhanced chemiluminescence reagents and film were obtained from Amersham Scientific (Australia). Phospholipid vesicles (PLV) were prepared from phosphatidylcholine (Sigma P5-388) by sonication as described (38), and hydroxypropyl-␤ cyclodextrin (hp␤CD) and trimethyl-␤ cyclodextrin (tm␤CD) were obtained from Aldrich Chemicals. Pronase, heparinase, cycloheximide, and actinomycin D were supplied by Sigma. [ 35 S]Methionine was supplied by ICN Chemicals.
LDL Preparation-Human LDL (1.02ϽdϽ1.05) and lipoprotein-deficient serum (LPDS, dϾ1.25 g/ml) were isolated from healthy, fasting volunteers in the presence of EDTA, aprotinin, and soybean trypsin inhibitor (all reagents from Sigma) by discontinuous density-gradient ultracentrifugation, dialyzed, filtered, and stored at 4°C under N 2 in EDTA and chloramphenicol as described previously (39).
Preparation of Human Apolipoprotein A-I-Purified human apoA-I was isolated by FPLC, delipidated, lyophilized, and stored at Ϫ20°C prior to reconstitution as described previously (40,41). The purity of each preparation was confirmed on SDS-PAGE by detection of a single band of molecular mass 28 kDa, and confirmed by Western blotting not to cross-react with antibodies to human or murine apoE.
LDL Acetylation-LDL was acetylated (AcLDL), excess reagents removed by dialysis against Chelex-100-treated PBS containing chloramphenicol (0.1 mg/ml), and acetylation assessed using non-denaturing agarose gel electrophoresis on 1% Universal Agarose gels (Ciba-Corning) in Tris-barbitone buffer (pH 8.6) at 90 V for 45 min, as described (8). The LDL band was visualized with Fat Red 7B stain. A relative electrophoretic mobility of Ն3, using native LDL as a reference, was routinely obtained.
Human monocytes (HMDM) were isolated from white cell concentrates using centrifugal elutriation as described (46,47). Purified monocytes (Ͼ95% purity by nonspecific esterase staining) were differentiated by plating at a density of 1.5 ϫ 10 6 cells/22-mm 2 culture dish (Costar) in RPMI 1640 containing antibiotics, glutamine (as above), and 10% (v/v) heat-inactivated whole human serum for 9 days. Following differentiation, the cells were washed and incubated with RPMI 1640 containing 10% LPDS (v/v) and AcLDL (50 g protein/ml) for 4 days to achieve cellular enrichment with FC and CE.
In specified experiments, cells were pretreated with cycloheximide (2.0 g/ml), actinomycin D (1.0 g/ml), heparinase 1 (3 units/ml), and Pronase (10 g/ml) after loading and before exposure to efflux medium as described in figure legends in order to investigate the respective importance of protein synthesis, mRNA transcription, cell surface proteoglycans, and cell surface proteins on apoE secretion. Cycloheximide and actinomycin D exposures were continued during efflux incubations, whereas Pronase and heparinase were removed and cells washed before incubation with efflux medium. The functional activity of heparinase preparations was separately confirmed for each preparation for each experiment by measuring the rate of formation of unsaturated ␤-uronides generated during the degradation of heparin, detected by absorbance at 232 nm (48).
Sterol Efflux and ApoE Secretion from Macrophages-Macrophages that had been incubated with AcLDL (or control medium without AcLDL) for 24 h (MPM) or 96 h (HMDM), were washed in warm PBS and incubated for another 24 h in 1.0 ml of "efflux medium" comprising DMEM (MPM) or RPMI (HMDM) with or without apoA-I (0 -25 g/ml) as described (8,41,44,45). To evaluate physicochemical cholesterol efflux, HMDM were incubated with AcLDL as above and then incubated in efflux medium comprising RPMI with 1.0 mg/ml hp␤CD, 1.0 mg/ml tm␤CD, or 20 -400 g/ml PLV for 24 h. (In comparative experiments, efflux media containing 1.0 mg/ml bovine serum albumin (BSA, Sigma) resulted in cholesterol efflux, apoE secretion, and cell viability identical to that in media not containing BSA, but BSA was typically omitted in later experiments to avoid potential nonspecific binding of antibodies to BSA and potential contamination of media with apoE in BSA preparations.) At the end of this incubation, aliquots of medium were removed from each culture dish, gently mixed with Complete © protease inhibitor (Roche Molecular Biochemicals) and spun in an Eppendorf centrifuge (16,000 ϫ g, 4°C, 10 min) to remove any detached cells. Cell monolayers were washed three times in PBS and then lysed in 0.6 ml of cold PBS containing 2% SDS and Complete © protease inhibitor. Cell and media samples were then analyzed for lipid content and apoE content, and cell lysates were additionally analyzed for cell protein.
Analysis of FC and CE-Aliquots of cell suspensions and of efflux media were placed into separate glass Kimax tubes (Kimble, MA), made up to a final aqueous volume of 1.0 ml with ice-cold PBS containing butylated hydroxytoluene and EDTA, and the total extracted with methanol (2.5 ml) and then hexane (5 ml) as described (39). 4 ml of the hexane layer was evaporated to dryness, redissolved in mobile phase, and analyzed by HPLC as described below (39).
Separation of FC and CE (cholesteryl docosahexanoate, arachidonate, linoleate, oleate, and palmitate) was performed at room temperature on a C-18 column (25 ϫ 0.46 cm length, 5 m pore size, Supelco) by reverse phase HPLC with detection at 210 nm (39,49). Acetonitrile/ isopropanol/water (44/54/2, v/v/v) and acetonitrile/isopropanol (30/70, v/v), respectively, separated FC and CE, commercial standards of which (Sigma) confirmed elution times and 210 nm absorbance. Lipids within cell extracts and media were expressed as nmol/cell culture or nmol/mg cell protein. Where required, percentage of cholesterol efflux was calculated for each culture by expressing the cholesterol released into efflux medium as a percentage of the sum of cholesterol in the efflux medium and total cholesterol in the cell lysate at time of extraction (8).
Quantitation of Apolipoprotein E in Cell Lysates and Efflux Media by Western Blot-Aliquots (40 l) of cell culture medium were mixed with sample buffer (20 l) containing 26.7 mM Tris-HCl (pH 6.8), 2% (w/v) SDS, 133 mM sucrose, 0.01% (w/v) bromphenol blue, 0.67 mM EDTA, reduced with 10 mM dithiothreitol and heated to 100°C for 5 min. Cell lysates (already in SDS) were treated in an identical fashion except for the omission of SDS from the sample buffer.
Samples were applied to a 4% stacking gel and 10% polyacrylamide resolving gel (50), and run under reducing conditions in Tris-glycine buffer for 18 h at 75 V and compared with molecular weight marker protein. Electrophoretic Western blot transfer onto a 0.45-m nitrocellulose membrane was performed at 4°C in Tris-glycine buffer for 1 h at 25-30 V, and subsequent blocking and washing steps were performed using Tris-buffered saline, pH 7.4, containing 0.05% Tween 20 and containing 1-5% skim milk powder. Following transfer, membranes were blocked, washed, and then incubated for 1.5 h with primary antibody to apoE (mouse anti-human monoclonal, 1:1000 dilution, Roche Molecular Biochemicals; or rabbit anti-mouse polyclonal, 1:5000 dilution, Biodesign International), before washing and incubating with 1:5000 dilution of secondary antibody (sheep anti-mouse IgG or donkey anti-rabbit IgG, Roche Molecular Biochemicals) conjugated with horseradish peroxidase. Membranes were incubated with ECL Western blotting detection reagent, exposed to high performance chemiluminescence film (Hyperfilm © , Amersham Pharmacia Biotech), and resulting chemiluminescence signal was quantified using Bio-Rad Gel Doc 1000 Molecular Analyst Software.
For HMDM, a serial dilution of authentic human apoE standard (Biodesign) in sample buffer was run to quantitate secreted apoE (micrograms per culture or per mg of cell protein). The linear response range of chemiluminescence signal to mass of authentic human apoE standard was defined for each Western blot for each experiment. For murine macrophages, data were calculated as arbitrary chemiluminescence units (AU) per cell culture or per mg of cell protein. In preliminary experiments it was confirmed that antibodies to human or murine apoE did not cross-react with apoA-I, and that antibodies to human apoA-I did not cross-react with murine or human apoE (data not shown). Plasma and macrophage samples from apoE KO mice were were confirmed to be apoE-deficient by Western blot.

Metabolic Labeling of Cell Proteins with [ 35 S]Methionine and
Immunoprecipiation-Mature, cholesterol-enriched HMDM and MPM were washed in methionine-free medium (Life Technologies, Inc.) and then incubated in methionine-free medium containing apoA-I and 250 Ci/ml [ 35 S]methionine (1175 mCi/mol, ICN) for 24 h at 37°C. 200 l aliquots of medium were removed at regular intervals as indicated, and HMDM and MPM media, respectively, immunoprecipitated using 1:10,000 dilution of polyclonal goat antibody to human apoE (Fitzgerald) or 1:5,000 dilution of polyclonal rabbit antibody to mouse apoE (Biodesign) and protein A-Sepharose (Amersham Pharmacia Biotech). Immunoprecipitates were subjected to SDS-PAGE prior to quantification of the 34-kDa band of apoE by phosphorimaging (Photostimulated Luminescence, Fujix BAS 1000). Quantitative immunoprecipitation was confirmed in control experiments by analysis of residual supernatants after immunoprecipitation.
Protein Estimation-Protein content of LDL samples and cell lysates was determined in duplicate or triplicate for each cell culture or each lipoprotein sample, and was measured using the bicinchoninic acid (BCA) method (Sigma) with BSA as standard (8,51).
Cell Viability-Cell viability was routinely assessed by light microscopic morphology, by estimation of cell protein and by absence of cell staining with 0.5% (w/v) trypan blue (Trace Biosciences, Australia) after 2 min of incubation. Counting was performed from at least two high powered fields in representative dishes, and viability under various loading conditions was independently and routinely confirmed to be Ն95%. Additionally, in experiments using potential cytotoxins such as actinomycin D, lactate dehydrogenase assay was also used to confirm preserved cell viability compared with control conditions.
Presentation and Statistical Analysis-A minimum of three separate incubations were performed for each condition in each experiment results for each experiment are expressed as the mean Ϯ S.D. of triplicate cultures, and all experiments described are representative of several. Statistical comparisons used Student's t test for unpaired data using Mystat statistical software, and p Ͻ 0.05 was considered significant. Where multiple experiments are pooled, results are expressed as mean Ϯ S.E. of n experiments.

ApoA-I Stimulates ApoE Secretion from Primary Murine
Macrophages-Enrichment of MPM (from QS or C57BL6 mice) with FC and CE by incubation with AcLDL increased apoE secretion into control medium (Figs. 1 and 3), and increased the cellular content of apoE (Figs. 2 and 3), consistent with previous literature. Over multiple independent experiments, expo-sure of lipid-loaded MPM to apoA-I further increased the secretion of apoE to between 2.4-and 5.9-fold (3.8 Ϯ 0.5, mean Ϯ S.E., n ϭ 6 experiments, p ϭ 0.01), of that elicited by lipid enrichment alone, and the difference between apoA-I and control conditions increased progressively over 24 h. ApoA-I also increased the secretion of apoE from non-loaded MPM, but this effect was more modest, ranging from 1.5-to 2-fold and was associated with much less mass efflux of cholesterol than from cholesterol-enriched cells. Stimulation of apoE secretion by apoA-I decreased the cellular pool of apoE in cholesterol-enriched MPM (Figs. 2 and 3) and was concurrent with the depletion of over 50% of total cell cholesterol during cholesterol efflux to apoA-I.
Kinetic studies demonstrated that unlike the rapid displacement of surface-bound apoE achieved by phospholipids within 60 min (52, 53) apoA-I-stimulated apoE secretion extended beyond 8 h, and this kinetic approximately paralleled the secretion of cell cholesterol (Fig. 4). Thus, the stimulation of secretion of apoE was unlikely to simply arise from displacement by apoA-I of preformed apoE resident on the cell surface, but was temporally linked to the process of apoA-I-mediated cholesterol efflux.
ApoA-I Stimulates ApoE Secretion from Primary Human Monocyte-derived Macrophages (HMDM)-Human and murine cells differ in their accumulation, metabolism, and efflux of cholesterol (22, 54 -58), 2 and there are significant differences between between the amino acid composition of human and rodent apoE (59 -61). Consequently, it was important to ensure that apoA-I-mediated apoE secretion was applicable to human macrophage foam cells and human apoE.
As with MPM, cholesterol enrichment caused a clear increase in the secretion of apoE by HMDM, and apoA-I caused a further marked increment in the secretion of apoE while simultaneously promoting cholesterol efflux (Fig. 5). In multiple experiments using monocytes isolated from different donors, the mass of apoE secreted by cholesterol-enriched HMDM under control conditions (i.e. without apoA-I) ranged between 0.5-3.0 g/mg of cell protein/24 h. However, apoA-I further enhanced apoE secretion from cholesterol-enriched cells 2-6fold (3.2 Ϯ 0.6, mean Ϯ S.E., n ϭ 6 experiments, p ϭ 0.01), regardless of the mass of apoE secreted under control conditions. In contrast, when added to HMDM that had not been cholesterol enriched, there was modest (as seen in Fig. 5) and inconsistent (over multiple experiments) stimulation of apoE secretion and cholesterol efflux by apoA-I.
ApoA-I induced a time-and concentration-dependent secretion of apoE from HMDM. As with MPM, total secretion of apoE was maximal at 5 g/ml apoA-I and continued to increase over 24 h (Fig. 6). Unlike MPM, however, there was no detectable depletion in cell-associated apoE after exposure of HMDM to apoA-I (data not shown). ApoA-I always induced less per-centage of cholesterol efflux from cholesterol-enriched HMDM than from cholesterol-enriched MPM (Fig. 3, MPM 53% efflux versus Fig. 5, HMDM 16.8% efflux), and induced negligible efflux from non-loaded HMDM. This is consistent with the known relative resistance of human macrophages to apoA-Imediated cholesterol efflux (56,58,62,63), and may explain the lack of depletion of the cell-associated apoE pool in HMDM after exposure to apoA-I (see "Discussion").
ApoA-I Induces the Secretion of de Novo Synthesized ApoE-In order to establish if apoA-I simply mediated the displacement of preformed, cell surface-associated apoE a series of experiments were undertaken. Cholesterol-enriched HMDM and MPM were metabolically labeled with [ 35 S]methionine during efflux incubation with apoA-I or control medium, and apoE secreted into efflux medium was measured by immunoprecipitation and quantification of 34-kDa bands after SDS-PAGE and phosphorimaging ( Fig. 7 after 24 h of efflux). At all time points up to 24 h, apoA-I achieved greater secretion of [ 35 S]methionine-labeled apoE secretion than control medium, and the amount of apoE secreted to apoA-I increased progressively with the duration of incubation. The protein synthesis inhibitor cycloheximide almost completely inhibited secretion of [ 35 S]methionine-labeled apoE to apoA-I and control media, and, importantly, completely blocked any additional apoE secretion achieved by apoA-I compared with control.
To confirm that protein synthesis was critical for the increment in total apoE in response to apoA-I exposure, and not only relevant to a potentially small pool which was labeled with [ 35 S]methionine, HMDM were incubated with cycloheximide for 1 h prior to efflux (to ensure inhibition of protein synthesis at beginning of incubation with apoA-I), as well as during efflux to apoA-I, and apoE quantified by Western blot (Fig. 8). There was very rapid and very marked inhibition of apoE secretion to apoA-I, which, over five independent experiments using both MPM and HMDM, was always significantly inhibited by 4 h (inhibited by 73.1 Ϯ 9.8% at 4 h), and beyond 4 h there was almost complete inhibition of apoE secretion to apoA-I and control medium (data not shown) without significant impairment of cholesterol efflux. Thus, ongoing de novo apoE synthe- sis contributes very substantially to the pool of apoE secreted to apoA-I. Actinomycin D, a transcription inhibitor, exerted a much more gradual and less significant effect on apoE secretion. In experiments evaluating earlier time points from 2 to 16 h, identical apoA-I-mediated apoE secretion was achieved with and without actinomycin D (data not shown). By 24 h, actinomycin inhibited apoE secretion to apoA-I by only 21.3% (apoA-I alone 361 Ϯ 36.5 ng of apoE/mg of cell protein, apoA-I ϩ actinomycin D 284 Ϯ 20.1 ng of apoE/mg of cell protein). These data indicate that although most apoE released to apoA-I requires active protein synthesis, it is unlikely to be mediated by acute up-regulation of mRNA transcription by apoA-I.
To evaluate the quantitative importance of the cell surface pool of apoE (both total and proteoglycan-bound), to apoA-Imediated secretion, macrophages were pretreated with Pronase or heparinase, respectively (Fig. 9). As our studies (Figs. 2 and 3) had shown that MPM showed most evident depletion of cell apoE after exposure to apoA-I, we hypothesized that MPM were most likely to demonstrate a significant cell surface pool of apoE, which may be displaced by apoA-I. Heparinase treatment had little effect on total cell-associated apoE and did not interfere with apoA-I-mediated apoE secretion in MPM (Fig. 9, A and C) or HMDM (data not shown). In contrast, Pronase pretreatment depleted total cell-associated apoE by 64.0 Ϯ 6.5% compared with control incubation, and reduced secretion of apoE in response to apoA-I (Fig. 9B) over the first 5 h.
However, by 8 h there was identical apoE secretion to apoA-I media from cells that had or had not been exposed to Pronase. ApoA-I-mediated apoE secretion from HMDM was not significantly different with or without Pronase exposure (data not shown).

Differential Effect of ApoE Secretion upon Basal Cholesterol Efflux and ApoA-I-induced
Cholesterol Efflux-Expression of human apoE in murine J774 cells was reported to increase cholesterol efflux to control medium and to HDL (64), suggesting that apoE could play a role in cholesterol efflux induced by other acceptors. We therefore investigated whether the extent of apoA-I-mediated cholesterol efflux was affected by its stimulation of cellular apoE secretion by comparing cholesterol efflux from AcLDL-loaded apoE-secreting C57/BL6 macrophages to that from apoE KO C57/BL6 macrophages. ApoEsecreting MPM spontaneously released more cholesterol to apoA-I-free medium than did apoE KO cells (8.5 Ϯ 0.9% versus 1.3 Ϯ 0.2% cholesterol efflux, respectively; p ϭ 0.005). However, cholesterol efflux to medium containing apoA-I (25 g/ml) from each of the macrophage types was very similar (48.4 Ϯ 3.5% and 53.3 Ϯ 2.0%, respectively; p ϭ NS). Thus, although our data confirmed that constitutive apoE secretion and cholesterol efflux from macrophages may be closely linked (21,22), apoA-I-mediated cholesterol efflux, at optimal concentrations, does not require and is not enhanced by cellular apoE secretion.
Two Different Vehicles of Physicochemical Cholesterol Efflux Exert Differential Effects on ApoE Secretion-Whether apoA-Imediated stimulation of apoE secretion could be explained by its stimulation of cholesterol efflux per se, independent of any apolipoprotein-specific or receptor-specific effect (9, 65) was investigated by comparing it to the effects of efflux to PLV (66) and ␤-cyclodextrins (67) (Fig. 10).
As described previously, at low concentrations (1.0 mg/ml) hydroxypropyl-␤-cyclodextrin did not induce net cholesterol efflux (41,67), whereas trimethyl-␤-cyclodextrin did induce significant cholesterol release (67). At the low concentrations used, ␤-cyclodextrin-mediated cholesterol efflux is not associated with significant net mass release of cell phospholipid (67). Despite the clearly different extent of cholesterol release with the two cyclodextrins, apoE secretion was not enhanced by either agent. In contrast, PLV inducing the same cholesterol efflux as trimethyl-␤-cyclodextrin caused a marked increase in apoE secretion. It was concluded that cholesterol efflux did not in itself necessarily stimulate apoE secretion from macrophages, and that additional properties of the efflux-inducing agent may contribute to such stimulation.
It is well known that that apoA-I induces phospholipid efflux from MPM during cholesterol efflux (7). We have confirmed these observations (68), and very recently, this has been shown to be associated with apoE secretion from THP-1 cells (69). In cholesterol-enriched HMDM, we have recently identified that Ͻ5.0 g of phospholipid is released/ml of apoA-I-containing efflux medium during 24 h of cholesterol efflux. 3 To investigate whether the amount of PL solubilized by apoA-I could completely explain the effect of apoA-I upon apoE secretion from HMDM, we exposed HMDM to 20 g/ml PLV (4-fold the mass of phospholipid removed by apoA-I during cholesterol efflux), apoA-I 10 -25 g/ml, or both (without prior generation of apoA-I-phospholipid discs) and quantified apoE secretion. We found an additive interaction of PLV and apoA-I on apoE secretion ( Fig. 11) with on average a 2.3-fold (p Ͻ 0.01) greater secretion of apoE achieved by the combination of apoA-I and PLV than either alone. This indicates that there are complementary mechanisms by which PLV and apoA-I mediate apoE secretion, and that apoA-I-mediated apoE secretion is unlikely to be completely explained by apoA-I-mediated solubilization of cell phospholipid.

DISCUSSION
To our knowledge, this is among the first mechanistic studies of how one apolipoprotein can stimulate the secretion of another apolipoprotein from macrophages. It provides a mechanism for a direct interaction between lipoprotein-derived apoA-I and the local secretion of apoE in various tissues, in- cluding atherosclerotic tissue in the vessel wall, and in the central and peripheral nervous system. This process is likely to be most important in sites accessible to apoA-I and normally inaccessible to large whole VLDL or HDL, which are the major carriers of apoE in plasma.
ApoA-I is more likely to achieve proximity to foam cells in atherosclerotic lesions than are larger intact lipoproteins. ApoA-I-containing particles isolated from intima are generally smaller than plasma HDL (3) and as lipid-poor apoA-I is abundant at the more dilute lipoprotein concentrations present in interstitial fluid (4,5,70,71), apoA-I-mediated stimulation of apoE secretion is likely to occur in the vessel wall. Our observations for murine macrophages from two strains of mice (QS and C57/Bl6) and human macrophages derived from multiple monocyte donors suggests this is not a process specific to one apoE phenotype or species, but relevant to all apoE-secreting primary macrophages. Stimulation of apoE secretion by apoA-I was much more marked in cholesterol-enriched cells than in control cells, and cellular cholesterol enrichment is known to increase apoE synthesis (18). Thus, this property of apoA-I can be expected to be most relevant in the presence of cholesterolrich foam cell macrophages such as those in atherosclerotic lesions. This may also apply during nerve regeneration, where prior increased synthesis of apoE in cholesterol-enriched foam cells would allow optimal stimulation of apoE secretion during cholesterol efflux mediated by apoA-I.
The ability of apoE to associate with HDL species has been characterized using large HDL particles in whole serum (e.g. Refs. 72 (18). In another, HDL increased apoE secretion and this could be prevented by FIG. 11. Co-operative interaction between PLV and apoA-I enhances apoE secretion from human macrophages. Cholesterolenriched HMDM were effluxed to apoA-I (10 g/ml), PLV (20 g/ml), apoA-I and PLV together, or RPMI alone. ApoE was measured by Western blotting, and data represent mean Ϯ S.D. of triplicate cultures from one experiment representative of three experiments (ratio of apoE secreted to A-IϩPLV compared with that secreted to PLV alone in Fig.  11 was 2.5 Ϯ 0.3, p Ͻ 0.01). Over three independent experiments, apoA-IϩPLV induced 2.3 Ϯ 0.3-fold (mean Ϯ S.E., n ϭ 3, p Ͻ 0.001) greater apoE secretion than PLV alone. tetranitromethane modification of HDL without affecting efflux of cholesterol, implying the two processes are quite separate (74), as previously suggested (75). By determining total mass cholesterol efflux and total apoE secretion, we have demonstrated clear net stimulation of apoE secretion by apoA-I. We have also confirmed our observations using [ 35 S]methioninelabeled apoE, indicating that because apoE turnover is rapid, net mass apoE secretion and labeled apoE secretion generate qualitatively similar data. The correlated kinetics of cholesterol efflux and apoE secretion (Figs. 4 and 6) are consistent with a temporal link between the two processes, but as seen with our data using cycloheximide (Figs. 7 and 8) and cyclodextrin-mediated cholesterol efflux (Fig. 10), this need not be a mechanistic link.
Our findings with cycloheximide, [ 35 S]methionine metabolic labeling, and Pronase indicate preformed cell surface pools and de novo synthesis during efflux may both contribute to apoE secretion induced by apoA-I. Initially, some apoE secreted to apoA-I from MPM is cell surface-derived and/or Pronase-sensitive, but is not bound to the cell surface by a heparinasesensitive proteoglycan. The fact that, with or without Pronase pretreatment, after 8 h identical amounts of apoE are secreted to apoA-I is consistent with a major contribution of de novo apoE synthesis in this process. Most likely, there is ongoing replenishment of a cell surface pool by which apoA-I induces apoE secretion. Importantly, de novo synthesis implies that net secretion of apoE causes net accumulation of apoE in the artery wall and not a simple relocation of preformed pools. However, as Pronase may affect apoA-I binding sites or other plasma membrane proteins, we cannot exclude a Pronase-sensitive apoA-I receptor or binding protein via which apoE secretion is stimulated. It appears that such a Pronase-sensitive apoE pool or apoA-I binding site is either less significant in HMDM or is less amenable to modulation by Pronase than is the case in MPM.
In one previous study, which examined the effects of pure apoA-I on apoE secretion by macrophages, no stimulation of apoE secretion was observed (76). There are two likely explanations for the discrepancy with our data. First, exposure to apoA-I in the earlier study was for only 6 h; second, in the earlier study, the secreted material was ultracentrifuged by density flotation prior to detection of apoE. In both human and murine macrophages, we observed that apoE secretion increased substantially between 8 and 24 h of exposure to apoA-I, implying that sampling at earlier times may markedly underestimate the effect of apoA-I on apoE secretion. Moreover, it has been previously shown that only 20% of cell-secreted apoE is lipid-associated (20). Consequently, prior ultracentrifugation to float secreted apoE, as well as potentially dissociating apoE from any lipoprotein-like particle generated in efflux medium, is very likely to underestimate the extent of apoE secretion. Recent results from Bielicki et al. (69) support the notion that apoA-I can mediate the release of apoE from lipid-rich THP-1 macrophages.
The progressive increase in apoE secretion over 24 h argues against apoA-I acting via simple displacement of apoE adherent to the cell surface as described in response to heparin, heparinase, or phospholipid vesicles in macrophage lines (53) or HepG2 cells (52). Further, displacement is quantitatively less important in some primary macrophages (77). One possible explanation that may unify phospholipid solubilization and our kinetic studies is that apoA-I induces apoE secretion after mobilizing cell phospholipid, which occurs during apoA-I-induced cholesterol efflux from macrophages (6, 7). Cholesterol enrichment increases cell phospholipid synthesis (78), and subsequent phospholipid release from macrophages to apoA-I (7). This may also explain the lack of apoE secretion observed during cholesterol efflux to tm␤CD, which does not induce significant phospholipid release.
Our combined experiments using PLV and apoA-I, at concentrations of PLV at least equal to that of cell phospholipid released to apoA-I during cholesterol efflux, suggest co-operative effects between PLV and apoA-I in inducing apoE secretion. This indicates that the mechanism by which apoA-I and PLV induce apoE secretion are at least to some extent different and complementary, and that apoA-I does not mediate apoE secretion exclusively via solubilization of membrane phospholipid. It also raises the prospect that interactions between apoA-I and phospholipids in the artery wall could enhance the potential for apoE secretion in vivo.
Constitutive apoE secretion from cholesterol-enriched macrophages is associated with release of cellular cholesterol, perhaps to a greater extent with human macrophages (21) than with murine macrophages (54), and our data comparing apoE-KO and control C57BL6 macrophages are consistent with these observations. However, apoE secretion appears to be neither essential or stimulatory for apoA-I-mediated cholesterol efflux from primary murine cells, as distinct from the reported effects in cell lines such as J774 cells (64).
The enhanced secretion of apoE was associated with a clear depletion of cell-associated apoE in MPM but not HMDM, even though apoA-I induced substantial apoE secretion by both cell types. Depletion of cell cholesterol (by HDL or apoA-I) would be expected to decrease apoE synthesis (18,19). Assuming similar intracellular turnover of apoE in HMDM and MPM, the selective depletion of MPM-associated apoE by apoA-I may be explained by the greater depletion of cholesterol from MPM than HMDM and therefore greater continued apoE synthesis in the latter.
As there is not significant re-uptake of secreted apoE by macrophages (79), our observations with actinomycin D and cycloheximide are most consistent with apoA-I-stimulated apoE secretion being a posttranscriptional regulatory effect (74). Glycosylation to the higher M r form of apoE occurs at the trans-Golgi network, therefore regulation of the secretion or degradation of high M r apoE as others (80 -82) have observed is likely to occur in a post-Golgi compartment. We postulate that during apoA-I-mediated cholesterol efflux apoA-I induces local alterations of the plasma-membrane to redirect glycosylated apoE away from normal lysosomal or non-lysosomal degradation (83) and toward surface-connected compartments (82,84), from which it can be secreted into the medium. Whether apoA-I induces a receptor-mediated stimulation of secretion or promotes a physicochemical solubilization of cell surface apoE is the subject of ongoing investigation.