Different cellular traffic of LDL-cholesterol and acetylated LDL-cholesterol leads to distinct reverse cholesterol transport pathways.

Endocytosis of LDL and modified LDL represents regulated and unregulated cholesterol delivery to macrophages. To elucidate the mechanisms of cellular cholesterol transport and egress under both conditions, various primary macrophages were labeled and loaded with cholesterol or cholesteryl ester from LDL or acetylated low density lipoprotein (AcLDL), and the cellular cholesterol traffic pathways were examined. Confocal microscopy using fluorescently labeled 3,3′-dioctyldecyloxacarbocyanine perchlorate-labeled LDL and 1,1′-dioctyldecyl-3,3,3′,3′-tetramethylindodicarbocyanine perchlorate-labeled AcLDL demonstrated their discrete traffic pathways and accumulation in distinct endosomes. ABCA1-mediated cholesterol efflux to apolipoprotein A-I (apoA-I) was much greater for AcLDL-loaded macrophages compared with LDL. Treatment with the liver X receptor ligand 22-OH increased efflux to apoA-I in AcLDL-loaded but not LDL-loaded cells. In contrast, at a level equivalent to AcLDL, LDL-derived cholesterol was preferentially effluxed to HDL, in keeping with increased ABCG1. In vivo studies of reverse cholesterol transport (RCT) from cholesterol-labeled macrophages injected intraperitoneally demonstrated that LDL-derived cholesterol was more efficiently transported to the liver and secreted into bile than AcLDL-derived cholesterol. This indicates a greater efficiency of HDL than lipid-poor apoA-I in interstitial fluid in controlling in vivo RCT. These assays, taken together, emphasize the importance of mediators of diffusional cholesterol efflux in RCT.

The development of atherosclerosis is initiated by the formation of macrophage-derived foam cells (1). As dedicated scavenger and sentinel cells, macrophages actively take up and process apoptotic and necrotic cells (2) as well as excess plasma and tissue LDL and modified LDL (1), which under pathological conditions lead to the accumulation of large amounts of cholesterol. To maintain cellular cholesterol homeostasis, the macrophage distributes and transports the excess cholesterol into specific cellular compartments and converts it into nontoxic cholesteryl ester for storage (1,3). The cell can also export excess cholesterol to appropriate extracellular acceptors by transfer mechanisms through cholesterol gradients that involve mostly HDL, mediated via scavenger receptor class B type I (SR-BI) (4) and the ABCG1/G4 transporter (5,6) or the apolipoprotein A-I (apoA-I)-mediated pathway that operates through ABCA1 (7)(8)(9)(10). In addition to the canonical receptor pathway for regulated LDL uptake by the LDL receptor (LDLr) (11,12), the macrophage can take up LDL via receptor-independent pathways, such as macropinocytosis, which, in a dosedependent manner, can lead to macrophage foam cell formation under certain conditions in vitro (13)(14)(15).
On the other hand, unregulated internalization via a number of scavenger receptors and other unknown receptors (16) accounts for up to 95% of uptake of modified lipoproteins. Receptor-mediated uptake of LDL and acetylated low density lipoprotein (AcLDL) is mostly via clathrin-coated pits (17). Whereas LDL is delivered to centrally located vesicles, bVLDL or AcLDL is observed in peripherally distributed vesicles, where its catabolism is slower (18,19). Comparison of the uptake of AcLDL and oxidized low density lipoprotein (OxLDL) also showed that the two ligands traffic to different endosomes and accumulate in distinct lysosomal compartments (20), an observation compatible with the reported morphological and functional heterogeneity of the endocytic compartment in macrophages (21)(22)(23)(24). However, the cellular uptake of OxLDL is complex and mediated by a variety of receptors that include macrophage receptor with collagen domain, scavenger receptor with C-like lectin, scavenger receptor type B-CD36, and LOX1 as well as scavenger receptor type A-I/II (16). In addition, OxLDL contains both oxidized proteins and lipids, which have multiple effects, including impaired degradation (25) and enhanced inflammatory stimulation (26).
To further explore how the differential internalization mechanisms of native and modified lipoproteins affect cholesterol transport and homeostasis in macrophages and avoid the complications of multiple effects of OxLDL, we used the modified lipoprotein AcLDL in comparison with native LDL. Here, we demonstrate that LDL and AcLDL traffic to discrete endosomal compartments and that cholesterol derived from each of these lipoproteins enters different cellular pools and is effluxed in vitro through largely distinct pathways: LDL-derived cholesterol effluxes preferentially via the HDL-mediated pathway dependent in part on ABCG1, whereas AcLDL-derived cholesterol effluxes preferentially via the more specific lipid-poor apoA-I/ABCA1-dependent pathway. Interestingly, in vivo reverse cholesterol transport (RCT) to the liver and bile is shown to be more significant for the LDL-derived cholesterol, reflecting the importance of the available mediators of diffusional cholesterol efflux in RCT.

Animals
ABCA1 2/2 mice were a kind gift from Dr. Edward M. Rubin (Department of Energy Joint Genome Institute, Berkeley, CA). SRA 2/2 mice were transferred from Dr. T. Kodama (University of Tokyo). C57Bl6, NPC1 2/2 , SR-BI 2/2 , LDLr 2/2 , and caveolin-1 2/2 mice were purchased from Jackson Laboratories and maintained and bred at the animal facility of the Ottawa Heart Institute. All protocols were approved by the University of Ottawa Animal Care Committee.

Lipid efflux
Unless indicated in the legends, labeling conditions were as follows. Macrophages were washed three times with plain DMEM and then labeled with LDL or AcLDL ( (32). Efflux to apoA-I (50 mg in 2 mg/ml BSA medium) was monitored for 3-5 h. Efflux to BSA (2 mg/ml BSA of DMEM) was allowed to proceed for 16 h. Efflux to mb-CD (10 mM in 2 mg/ml BSA of DMEM) was carried out for 15 min at 37jC or 4jC.

Lipid analysis
Cellular lipids were extracted (33) and separated by TLC using hexane-diethylether-acetic acid (105:45:1.5) as running solvent on Sil-G TLC plates (EMD Chemicals, Darmstadt, Germany). Lipid bands were detected by exposure to iodine vapors and scraped off the TLC plate, and radioactivity was measured with a scintillation counter. For total cholesterol determination, cells were washed with cold PBS, and cholesterol was extracted by isopropanol and measured by colorimetric assay (Wako Chemicals, Richmond, VA).

RCT
Macrophages from ABCA1 2/2 , ABCA1 2/1 , or ABCA1 1/1 mice were labeled with cholesterol delivered by LDL or AcLDL for 24 h. Cells are removed with 5 mM EDTA PBS and injected intraperitoneally into C57BL6 mice (34). Gallbladders, livers, and feces were harvested 24 h later. Tissues and feces were treated with 0.5 N NaOH, and lipid radioactivity was counted.

Statistics
Student's t-test was applied to evaluate significant differences.

RESULTS
Macrophages labeled and loaded with AcLDL-derived cholesterol preferentially efflux through ABCA1, and macrophages labeled and loaded with LDL-derived cholesterol preferentially efflux by a diffusional pathway BMDMs were labeled and loaded with [ 3 H]cholesterol delivered by either AcLDL or LDL. AcLDL-loaded macrophages effluxed cholesterol to apoA-I at a rate nearly 7-fold higher than that of the macrophages loaded with LDL (Fig. 1A). Furthermore, apoA-I-mediated cholesterol efflux from the cells loaded with AcLDL was totally dependent on ABCA1 expression, whereas cells loaded with LDL displayed z25% residual apoA-I-specific cholesterol efflux in ABCA1 2/2 macrophages (Fig. 1A). In contrast, macrophages loaded with LDL effluxed cholesterol to HDL and BSA at a significantly greater rate than those loaded with AcLDL ( Fig. 1B, C). This difference was not related to the preferential plasma membrane cholesterol labeling by LDL cholesterol, because mb-CD extraction at 4jC, which measures the proportion of cholesterol present at the cell surface, showed equivalent plasma mem- Fig. 1. Macrophages labeled with cholesterol derived from acetylated low density lipoprotein (AcLDL) or LDL efflux differentially to different acceptors. Fetal liver-derived macrophages or bone marrow-derived macrophages (BMDMs) from ABCA1 wild-type (1/1), heterozygote (1/2), or homozygote (2/2) knockout mice were labeled with cholesterol (5 mCi/ml) delivered by LDL or AcLDL (50 mg protein/ml) in 1% FBS in DMEM for 24 h, followed by overnight equilibration in DMEM with 2 mg/ml BSA. A, B: Efflux of cholesterol expressed as percentage of total cell radioactivity was measured over 5 h in the presence or absence of 50 mg of apolipoprotein A-I (apoA-I) (A) or HDL (B). C: Efflux to 2 mg/ml BSA in DMEM was measured over 16 h. D: Cholesterol efflux to 10 mM methyl-b-cyclodextrin (mb-CD) was determined by incubation for 15 min at 4jC or 37jC. E: Efflux from peritoneal macrophages [resident (Res.PM) and elicited (Ind.PM)], J774 cells, fetal liver-derived macrophages (FLDMs), and BMDMs. F: Specific apoA-Imediated cholesterol efflux from J774 cells measured by immunoprecipitation of eff lux medium with anti-apoA-I antibody. Each data point is the mean and SD of triplicate determinations and representative of two or more independent experiments. *P , 0.001.
Cholesterol traffic in macrophages brane labeling by LDL (9.9%) and AcLDL (10.2%) (Fig. 1D). However, short-term cholesterol efflux to mb-CD at 37jC, which measures cholesterol present at the plasma membrane and the recycling compartment (36), was significantly greater from cells loaded with LDL (44%) compared with AcLDL (22%), regardless of ABCA1 expression level (Fig. 1D). Therefore cholesterol derived from LDL preferentially loaded the recycling compartment, a conclusion supported by the greater efflux to HDL and to mb-CD at 37jC. This difference was not related to the use of specific differentiated murine macrophages, because the same efflux specificity existed in all tested murine macrophages (Fig. 1E). The same specificity was also observed for efflux to immunoprecipitated apoA-I (Fig. 1F).
To further document the differential trafficking of LDLand AcLDL-derived cholesterol and its independence of the labeling efficiency of cell surface compartments, macrophages were labeled to an equivalent level with [ 3 H]cholesteryl oleate and preincorporated into LDL or AcLDL to the same specific activity ( Fig. 2A). Under these conditions, cholesterol efflux to apoA-I from macrophages loaded with AcLDL remained significantly greater (P , 0.004) than that from cells loaded with LDL by z3-fold (Fig. 2B). When cells were equally loaded with cholesteryl ester-labeled lipoproteins, the proportion of accumulated ACAT-generated cholesteryl esters was z10-fold greater with AcLDL compared with LDL ( Fig. 2C). Similar to [ 3 H]cholesterol labeling of lipoproteins, cholesterol derived from [ 3 H]cholesteryl oleate was equally accessible to mb-CD extraction at 4jC whether delivered by LDL or AcLDL (Fig. 2D). However, mb-CD extraction at 37jC was much higher for LDL-loaded macrophages, similar to the previous results (Fig. 1D) and confirming a preferential labeling by LDL of the recycling compartment.
It is well known that AcLDL is a potent ACAT activator (37,38), and our own data showed that treatment with the same amount of lipoprotein (either LDL or AcLDL) in macrophages causes significantly more cholesterol esterification by labeling with [ 3 H]oleate in AcLDL-treated cells (data not shown). To understand whether the increased apoA-I-mediated efflux was attributable simply to increased cholesteryl ester formation, because cholesterol efflux to apoA-I is closely related to the level of cellular cholesteryl ester (see supplementary Fig. II) and in keeping with previous studies (39), macrophages were loaded for 5 or 24 h with AcLDL or LDL to achieve similar levels of radioactive labeling (Fig. 2E) and cholesteryl ester formation (Fig. 2G). Under these conditions, a clear difference in the amount of apoA-I-mediated cholesterol efflux was evident (4-fold increase for AcLDL-loaded cells) (Fig. 2F). Increased efflux to apoA-I might reflect preferential cholesterol mobilization from the ACAT-accessible pool; however, ACAT inhibition during the labeling period increased apoA-I-mediated cholesterol efflux from the macrophages labeled with AcLDL but not LDL (Fig. 2H). Increased efflux to apoA-I might reflect preferential cholesterol mobilization from the ACAT-accessible pool (cholesteryl ester droplets). However, administration of the ACAT inhibitor Sandoz 58-035 during the labeling period completely suppressed the cholesterol reesterification cycle (data not shown) and increased apoA-I-mediated cholesterol efflux from the macrophages labeled with AcLDL but not LDL (Fig. 2H). These results demonstrate that the observed differences of LDL-and AcLDL-loaded macrophages are not simply attributable to increased loading or preferential targeting to an ACAT-accessible pool; rather, they further demonstrate that lipoproteinderived cholesterol, which internalizes through distinct mechanisms (see below), remains in two functionally distinct pools.
Level of cholesterol loading with LDL or AcLDL does not affect efflux specificity Next, we showed that the specificity of the pathways is independent of the net uptake of lipoprotein cholesterol. Increasing LDL concentration resulted in a linear and nonsaturating increase in uptake, which at 250 mg/ml doubled total cellular cholesterol and increased cholesteryl ester level to z5% (Fig. 3A, B), indicating that a non-receptor-independent mechanism contributed to macrophage LDL uptake, as reported by others (13,14). In contrast, increasing AcLDL concentration up to 50 mg/ml rapidly increased cellular cholesterol, which plateaued at a 3-fold increase, and was accompanied by an increase in cholesteryl ester to z40% (Fig. 3C, D) as well as a greater upregulation of ABCA1 (Fig. 3E). Under conditions that achieve equivalent cholesterol loading with LDL and AcLDL, we measured the total cellular cholesterol level before the addition of apoA-I and then carried out an efflux assay (Fig. 3G, H). Thus, AcLDL at 6.25 mg/ml and LDL at 150 mg/ml loaded the cells to the same level, 30.2 and 29 mg of total cholesterol mass, respectively (SD , 15%), but AcLDL elicited twice as much ABCA1-mediated efflux. The same difference in efflux was maintained with 12.5 mg/ml AcLDL and 250 mg/ml LDL, which increased cellular cholesterol to 38.7 and 35.6 mg/mg cellular protein, respectively.
Because ABCG1 mediates efflux to HDL (5, 40) and LDL-derived cholesterol is effluxed to HDL at a higher rate than AcLDL-derived cholesterol (Fig. 1B), we also measured ABCG1 protein level in LDL-and AcLDL-treated cells (Fig. 3F). Clearly, ABCG1 protein expression was increased in LDL-treated cells, but not as much as in AcLDLtreated cells, suggesting that the robust cholesterol efflux to HDL in LDL-treated cells was not related directly only to ABCG1 expression but also to a specific targeting of cholesterol, and possibly of the transporter, to the recycling compartment. Alternatively, this discrepancy may reflect the contribution of another, yet uncharacterized, mediator of the diffusional pathway (41).

Specificity of LDL and AcLDL cholesterol internalization
To determine the specificity of LDL cholesterol internalization via the LDLr and AcLDL via SRA in macrophages, we first labeled BMDMs from LDLr 2/2 and wild-type mice with [ 3 H]cholesteryl oleate-labeled LDL. Both BMDMs were pretreated with 3% lipoprotein deficient serum to upregulate LDLr by depleting cellular cho-lesterol. Under this condition, z70% of LDL cholesterol internalization could be inhibited by the addition of cold LDL, indicating its dependence on LDLr (Fig. 4A), whereas the fraction of LDL cholesterol uptake, which could not be inhibited by cold LDL, may be mediated by macropinocytosis (13)(14)(15) (Fig. 4B), indicating that the majority of LDL cholesterol uptake  6-, and 16-fold, respectively. F: ABCG1 protein was measured by Western blot under conditions that achieved equal cholesterol loading, AcLDL (12.5 mg/ml) and LDL (250 mg/ml) loaded cells, and ABCG1 induction was 2.8and 1.9-fold, respectively. G, H: Conditions that achieve similar cholesterol loading were selected for an efflux assay. Equivalent total cholesterol was obtained by loading with 150 and 250 mg LDL/ml, accumulating 30 and 36 mg/well cholesterol mass, respectively (G), and by loading with 6.25 and 12.5 mg AcLDL/ml, accumulating 29 and 38 mg/well cholesterol mass, respectively (H). The cells, thus labeled and cholesterol-loaded, were equilibrated with 2 mg BSA/ml in lipid-free medium overnight, and efflux to apoA-I was started with the addition of apoA-I (25 mg/well) for 5 h. Mean and SD values are as described for Fig. 1. in vivo was dependent on macropinocytosis or another uncharacterized mechanism. Next, we tested the specificity of AcLDL cholesterol internalization via SRA. The uptake of AcLDL was decreased in SRA 2/2 macrophages compared with C57BL6 macrophages (Fig. 4C), but these cells still accumulated a large amount of cholesterol from AcLDL, suggesting a contribution of other scavenger receptors in normal and SRA 2/2 macrophages. The uptake of AcLDL cholesterol by C57BL6 macrophages could not be competed by the addition of cold LDL (Fig. 4C), consistent with the absence of overlap between the two types of receptors in normal macrophages, as reported previously (42). Uptake of AcLDL could be almost completely inhibited (93%) by the addition of a 50-fold excess of cold AcLDL in C57BL6 macrophages (Fig. 4D) and was independent of the expression of LDLr, indicating that AcLDL uptake was receptordependent (Fig. 3C, D). Thus, the AcLDL cholesterol uptake is receptor-dependent, but LDL cholesterol could be taken up by either receptor-mediated or non-receptormediated pathways.
Importantly, the patterns of cholesterol efflux to apoA-I were not altered when comparing macrophages of different genetic backgrounds (i.e., wild-type, SRA 2/2 , or LDLr 2/2 ) that were labeled with [ 3 H]cholesterol or [ 3 H]cholesteryl oleate incorporated in either LDL or AcLDL (Fig. 4E, F). Similarly, the specificity of efflux to HDL or mb-CD was not altered by LDLr or SRA deficiencies (data not shown) but reflected the ligand (i.e., LDL vs. AcLDL) properties.

LDL-and AcLDL-related cholesterol traffic pathways are differentially regulated
Caveolin-1, which has been shown to transport cholesterol between intracellular compartments/endoplasmic reticulum and the plasma membrane (43,44), had no effect on apoA-I-mediated cholesterol efflux from macrophages labeled with LDL or AcLDL cholesterol (data not Fig. 4. Specificity of LDL and AcLDL internalization via the LDL receptor (LDLr) and SRA in macrophages. BMDMs from control or LDLr 2/2 (A-E) or SRA 2/2 (F) mice were labeled with [ 3 H]cholesteryl oleate (5 mCi/ml) delivered with LDL or AcLDL (50 mg/ml) in 5% LPDS for 24 h. Before labeling, all of the cells (except those for B) were pretreated with 3% LPDS for 24 h to upregulate LDLr by depriving cellular cholesterol. After labeling, cells were washed twice with DMEM. The cells were then lysed by the addition of 0.5 N NaOH and incubation at room temperature overnight. Cell count was measured. Cholesterol efflux to apoA-I is presented in E, F.

Cholesterol traffic in macrophages
shown). On the other hand, deficiency in Niemann-Pick type C1 (NPC1), which controls cholesterol traffic from late endosomes (45,46), led to a 44% reduction in cholesterol efflux from LDL-labeled macrophages and to a 72% reduction from AcLDL-labeled cells (Fig. 5A). A similar difference in cholesterol traffic from late endosomes was also observed upon treatment with progesterone, which blocks transport from late endosomes to the plasma membrane (36,47,48). Progesterone abrogated cholesterol efflux to apoA-I in LDL-labeled cells, whereas a small but significant (16%) cholesterol efflux remained in AcLDL-labeled cells compared with control cells (Fig. 5B). Because liver X receptor ligands, such as hydroxycholesterol, are potent ABCA1 inducers, we treated the cholesterol-loaded cells with 22-hydroxycholesterol overnight during the equilibration time (before incubation with apoA-I) in an attempt to bypass the differential induction of ABCA1 by the two lipoproteins. As expected, 22-hydroxycholesterol significantly increased ABCA1 protein expression (data not shown) in both LDL-and AcLDL-labeled cells and significantly increased cholesterol efflux to apoA-I in AcLDL-labeled cells, but cholesterol efflux to apoA-I in LDL-labeled cells was not altered (Fig. 5C). These results further indicate that trafficking of LDL-derived and AcLDL-derived cholesterol is differentially regulated.

Imaging of LDL-and AcLDL-related traffic
Dual fluorescent label experiments were performed in which DiO-labeled LDL and DiD-labeled AcLDL were jointly incubated with macrophages. The fluorescently labeled lipoproteins trafficked to distinct endosomes, which only partly overlapped at early time points but remained separated at later time points (over a period from 20 min to 2 h) (Fig. 6). We followed the fluorescence for up to 24 h (the conditions that were used for loading) and saw no colocalization at any point (data not shown). These compartments were Lysotracker-positive (data not shown), suggesting that the DiO-labeled LDL and DiD-labeled AcLDL were trafficked to distinct late endosomes/lysosomes, in agreement with the observations of others (19,20). To specifically follow cholesterol movement, we also labeled LDL and AcLDL with dansyl-cholestanol, a fluorescent analog of cholesterol. Dansyl-cholestanol-labeled LDL or AcLDL was separately added to wild-type or ABCA1-knockout macrophages under conditions that would achieve equal loading or labeling of [ 3 H]cholesterol (24 h of incubation) and examined by fluorescence microscopy. Dansyl-cholestanol delivered by LDL showed diffuse fluorescence throughout the cell with some punctate structures. On the other hand, dansyl-cholestanol delivered by AcLDL was present in multiple bright punctate structures (see supplementary Fig. IB). In the absence of ABCA1, both LDL-and AcLDL-delivered dansyl-cholestanol-loaded macrophages showed increased total cellular fluorescence (data not shown). The fluorescence patterns observed here suggest that, over time, AcLDL-delivered dansyl-cholestanol tends to accumulate in the late endosomes, whereas LDLdelivered dansyl-cholestanol is transferred to other cellular membrane compartments.

LDL-and AcLDL-derived cholesterol returns to the liver at different rates
To test whether the separate metabolic pathways that were shown to exist in vitro for LDL and AcLDL cholesterol could also be documented in vivo, an RCT experiment (34) was carried out. Equal cell numbers of BMDMs from normal C57BL6 mice, labeled to the same extent with [ 3 H]cholesterol delivered either by LDL or AcLDL, were injected intraperitoneally into normal C57BL6 mice. After 24 h, the animals were killed (after a 5 h fast) and the cholesterol radioactivity transported from the intraperitoneal site of injection to the liver, gallbladder, or feces was measured. At 24 h after injection of green fluorescent protein-labeled macrophages, we estimated that the majority of injected macrophages remained in the peritoneal cavity. Histological examination of liver sections failed to detect any fluorescent macrophages (data not shown), suggesting that the injected macrophages them- selves did not migrate to the liver, in agreement with the results of others (34), and thus further demonstrating that the radioactivity measured in the liver represents cholesterol transported to the liver. Agarose gel electrophoresis of the concentrated peritoneal fluid demonstrated the presence of both a-HDLand preb-HDL-migrating apoA-I (Fig. 7A, B, lane 3), with a large preponderance of preb-HDL, which can mediate cholesterol efflux in an ABCA1-dependent manner. The return of cholesterol from LDL-labeled macrophages to the liver was significantly greater than that from AcLDL-labeled macrophages by z50% (Fig. 7D). Similar results were obtained for RCT to the bile, which was 35% higher for LDL-derived cholesterol (Fig. 7C), and to feces (results not shown). As established above, the in vitro efflux of cholesterol from macrophages follows distinct pathways that are faster for AcLDL-derived cholesterol; however, in vivo efflux and RCT are dependent on the concentration of HDL and apoA-I present in the peritoneal fluid. Therefore, these results reflect the respective in vivo contributions of efflux mediated by apoA-I and ABCA1 versus HDL and ABCG1 or SR-BI or other pathways and the peritoneal fluid concentration of their acceptors.

DISCUSSION
We have demonstrated that the intracellular transport pathways for AcLDL-and LDL-derived cholesterol are different and that cholesterol derived from these lipoproteins segregates into two mostly distinct cellular pools. AcLDL-delivered cholesterol is preferentially transported Fig. 6. LDL-and AcLDL-related cargoes traffic into different endosomes. Fluorescent hydrophobic lipid markers were incorporated into LDL and AcLDL to follow the intracellular trafficking of the lipoproteins. 3,3 ¶-Dioctyldecyloxacarbocyanine perchlorate-labeled LDL and 1,1 ¶-dioctyldecyl-3,3,3 ¶,3 ¶-tetramethylindodicarbocyanine perchlorate-labeled AcLDL were incubated with the macrophages for 5 min at 37jC, then washed and followed by fluorescence microscopy for 2 h. LDL (green) and AcLDL (red) were trafficked to distinct late endosome compartments. The results are representative of four independent experiments. Bars 5 1 mm.
Cholesterol traffic in macrophages into late endosomes and lysosomes and further converted to cholesteryl ester in endoplasmic reticulum-associated lipid droplets, whose pools are readily accessible to efflux to apoA-I by an ABCA1-dependent pathway, in agreement with earlier reports (7,49). On the other hand, the pool of LDL-derived cholesterol is preferentially transported to a recycling compartment, where it is apparently more accessible to efflux mediated by HDL, BSA, and mb-CD.
These specific efflux phenotypes for AcLDL-and LDLderived cholesterol were consistently observed in all murine macrophages tested (Fig. 1E). In addition, cholesterol from other types of modified or pathological lipoproteins (OxLDL, bVLDL) was also shown to be readily accessible to apoA-I (see supplementary  (50), and the accumulated cholesterol can be mobilized primarily by ABCA1-mediated efflux. The dichotomy of cholesterol traffic is dictated by the specificity of entry (different receptors, adaptors, and receptorindependent uptake, such as macropinocytosis), transport, accumulation in different endosomes, and mobilization for efflux. The specificity of entry starts with the retention of AcLDL in the cell periphery in protrusions, such as microvilli, in contrast with the rapid delivery of LDL to late endosomes (19). The initial association/retention of AcLDL at the cell surface appears to take place in large cell surface structures different from the classical clathrincoated pit pathway (19,51), a site of initial retention also shared by bVLDL (18,52).
The binding of the lipoproteins to the receptors elicits the recruitment of specific sets of adaptor proteins, such as autosomal recessive hypercholesterolemia-protein for LDLr and Dab1 for LDLr-related protein (53)(54)(55), but no adaptor proteins have been identified for SRA. Presumably, distinct adaptors and their cognate receptors control the specific targeting of AcLDL and LDL into distinct lysosomes with little overlap (Fig. 6). The segregation of the ligands is not only morphological but also functionally distinct and results in the formation of two differently accessible cholesterol pools. The mobilization of cholesterol from these pools proceeds through different pathways with partial overlap: one operates through apoA-I and ABCA1 and the other through diffusional efflux across cholesterol gradients, including those mediated by ABCG1 (6,40) and by SR-BI (56), which both deliver cholesterol to HDL rather than to lipid-poor apoA-I. The regulation of diffusional cholesterol efflux through cholesterol gradients from cellular membranes to lipoprotein acceptors is apparently dependent on the transport from various cellular pools in both the plasma membrane and intracellular membranes that differ in their accessibility to exogenous acceptors. Cells have been shown to have slow and fast cholesterol pools in terms of mb-CD accessibility (57,58). Here, we have shown that LDLderived cholesterol is more rapidly removed by mb-CD compared with AcLDL-derived cholesterol, indicating that AcLDL preferentially labels a slow pool in terms of accessibility for diffusional efflux. This slow pool may be in late endosomes, given its preferential mobilization by ABCA1. In contrast, LDL, like HDL, preferentially labels cholesterol pools with rapid access to diffusional efflux (50), which includes the recycling compartment shown here (Fig. 1D) and in a recent report (36).
Our observation that LDL-derived cholesterol stored in macrophages can return to the liver of wild-type mice and be secreted through the bile more efficiently than AcLDL-derived cholesterol indicates that diffusional efflux through cholesterol gradients, including ABCG1-and SR-BI-mediated efflux, is an efficient pathway. Although the net amount of LDL cholesterol taken up by macrophages is limited by the downregulation of the LDLr, the efficient labeling and delivery with lipoproteins and phagocytosis of apoptotic cells requires an efflux mechanism, which we know is less dependent on apoA-I (see supplementary Fig. II). Therefore, the greater efficiency of in vivo RCT from LDL-loaded versus AcLDL-loaded macrophages in the control mice implies that HDL and other lipoproteins mediating diffusional efflux are more effective in RCT than lipid-poor apoA-I/preb-HDL, although the cavity fluid level of preb-HDL is apparently higher than that of a-HDL (Fig. 7B). The other possibility is that the preb-HDL becomes a-HDL via lipid efflux in an ABCA1-dependent manner, which then further quickly removes lipids from the macrophage foam cells. These interstitial HDLs become effective ligands for hepatic ABCG1, which is upregulated by both LDL and AcLDL loading of macrophages (Fig. 3F). It was recently shown to play a major role in the control of tissue lipid levels, and although its deficiency does not affect HDL levels, it has been suggested to contribute to in vivo RCT via HDL (40).
Although the downregulation of SR-BI in AcLDL-loaded macrophages (59) (data not shown) would argue against a major role for this transporter in efflux and in RCT at the step of efflux, we observed slight increases in SR-BI protein in LDL-loaded macrophages (data not shown), as reported by others (60). In vitro, macrophages loaded with AcLDL (i.e., foam cells) are exquisitely dependent on ABCA1-mediated, apoA-I-dependent efflux, in agreement with reports in the literature (7,61,62). However, in vivo RCT from AcLDL-loaded macrophages is clearly less effective than that from LDL-loaded macrophages (Fig. 7). We also recently showed that RCT from abca1 2/2 macrophages is decreased by z50%, indicating the existence in vivo of a significant efflux and RCT independent of ABCA1 activity (M. D. Wang et al., unpublished data). Therefore, we must conclude that other transporters and efflux pathways operate in vivo for RCT. A recent and timely report from Rothblat and colleagues (41) highlights the heterogeneity of diffusional efflux pathways and the importance of uninhibitable or background efflux, independent of ABCA1 and SR-BI. Such pathways may contribute to RCT from LDL-labeled macrophages. Peripheral lymph and interstitial fluid have been shown to contain all plasma lipoproteins (63)(64)(65)(66)(67), albeit at reduced concentrations and with significantly increased free cholesterol compared with plasma counterparts. This increased free cholesterol in interstitial lipoproteins may reflect the reduced lymph LCAT activity (63,66) and the active cholesterol efflux to interstitial lipoproteins. More studies of these lipoproteins and their relevance to RCT are clearly needed. Because it is a bidirectional process, diffusional efflux has never been recognized as a significant contributor to RCT. The in vivo evidence presented here that links in vitro diffusional efflux to irreversible transport to the liver and bile demonstrates its physiological importance.
Finally, in mitigating against atherosclerosis, it is critical that appropriate acceptors (i.e., HDL or lipid-poor apoA-I) be present to promote the regression of foam cells. Macrophages that have been loaded with modified lipoproteins are critically dependent on ABCA1/apoA-I-mediated cholesterol efflux for the mobilization of loaded cholesterol (i.e., in vitro lipid-poor apoA-I can efficiently promote the regression of foam cells). Treatments such as apoA-I Milano or apoA-I mimetic peptide infusion (68)(69)(70) appear to promote foam cell regression without increasing plasma HDL levels, although it is unclear whether regression is achieved by efflux and/or other antiatherogenic functions of apoA-I. In conclusion, the existence of dual pathways for macrophage cholesterol transport implies that effective intervention against atherosclerosis may require LDL cholesterol-lowering therapy with statins in combination with specific agonists to increase the expression and function of both ABCA1 and ABCG1 (40,71) as well as SR-BI (72,73).