Preferential ATP-binding Cassette Transporter A1-mediated Cholesterol Efflux from Late Endosomes/Lysosomes*

Recently, ATP-binding cassette transporter A1 (ABCA1), the defective molecule in Tangier disease, has been shown to stimulate phospholipid and cholesterol efflux to apolipoprotein A-I (apoA-I); however, little is known concerning the cellular cholesterol pools that act as the source of cholesterol for ABCA1-mediated efflux. We observed a higher level of isotopic and mass cholesterol efflux from mouse peritoneal macrophages labeled with [ 3 H]cholesterol/acetyl low density lipoprotein (where cholesterol accumulates in late endosomes and lysosomes) compared with cells labeled with [ 3 H]choles-terol with 10% fetal bovine serum, suggesting that late endosomes/lysosomes act as a preferential source of cholesterol for ABCA1-mediated efflux. Consistent with this idea, macrophages from Niemann-Pick C1 mice that have an inability to exit cholesterol from late endo-somes/lysosomes showed a profound defect in cholesterol efflux to apoA-I. In contrast, phospholipid efflux to apoA-I was normal in Niemann-Pick C1 macrophages, as was cholesterol efflux following plasma membrane cholesterol labeling. These results suggest that cholesterol deposited in late endosomes/lysosomes preferentially acts as a source of cholesterol for ABCA1-mediated cholesterol efflux.

phages from these animals have a profound defect in apoA-Imediated cholesterol efflux (8 -10), indicating that apolipoprotein-mediated cholesterol efflux is primarily mediated by ABCA1. In contrast, ABCA1 shows only slight interaction with HDL 3 and no interaction with HDL 2 (11). Cellular cholesterol efflux mediated by HDL is thought to involve a "passive" process that may be diffusion-mediated or may involve an interaction of HDL with scavenger receptor B-I (SR-BI) (12,13).
ABCA1 is a full transporter with 12 membrane-spanning domains (5,14). Transfection of ABCA1 in 293 cells reveals a predominant cell surface localization and suggests a direct interaction of ABCA1 with apoA-I (11). The primary activity of ABCA1 appears to be the translocation of phospholipid at the plasma membrane rather than direct interaction with cholesterol (15,16). Phospholipid-apoA-I complexes formed by ABCA1 may promote cholesterol efflux in a secondary fashion perhaps involving distinct areas of the plasma membrane (15,17). The nature of the cellular sites that donate cholesterol to these phospholipid-apoA-I complexes is poorly understood. This may involve specific plasma membrane domains that derive cholesterol from intracellular stores. The nature of intracellular sites that potentially donate cholesterol to the plasma membrane for ABCA1-mediated efflux is also unclear. Niemann-Pick C (I and II) molecules play an essential role in intracellular cholesterol trafficking, particularly in the exit of cholesterol from late endosomes/lysosomes (18 -21). Earlier studies suggested a defect in cholesterol efflux to phospholipid vesicles in NPC1 fibroblasts (22), but the specific role of NPC1 in ABCA1-mediated cholesterol efflux has not been investigated.
The ABCA1 gene is up-regulated by cellular cholesterol loading (23). The mechanism of this effect is increased gene transcription mediated by the oxysterol-activated transcription factor liver X receptor (LXR) acting in a complex with retinoid X receptor (RXR) at a site on the proximal promoter of the ABCA1 gene (24). While studying cholesterol efflux from macrophages that had been treated with the LXR/RXR ligands 22(R)-hydroxycholesterol and 9-cis-retinoic acid to up-regulate ABCA1, we noticed a marked discrepancy between the magnitude of ABCA1 expression and the resulting stimulation of cholesterol efflux, depending on the method of cellular cholesterol labeling. This led to an investigation of the hypothesis that ABCA1 stimulates cholesterol efflux preferentially from a pool of cholesterol found in late endosomes/lysosomes. This hypothesis has been evaluated by comparing cholesterol efflux under different labeling conditions and supported by the demonstration of a profound defect in cholesterol efflux to apoA-I using macrophages from NPC1 mice.

EXPERIMENTAL PROCEDURES
Ribonuclease Protection Assay-Reverse transcription-polymerase chain reaction was used to obtain a fragment of the murine ABCA1 cDNA. Murine ABCA1 and ␤-actin antisense riboprobes were prepared by in vitro transcription using murine ABCA1 ␤-actin cDNA plasmid constructs. The protected hybrid fragments for ABCA1 and ␤-actin were 290 and 160 base pairs, respectively. Ribonuclease protection assay was performed as described (25). In brief, 20 g of total RNA were hybridized with 5 ϫ 10 5 cpm ABCA1 and ␤-actin riboprobes at 48°C overnight in 30 l of a buffer consisting of 40 mM PIPES, pH 6.0, 400 mM NaCl, 1 mM EDTA, and 80% formamide. The hybridization mixture was digested with 20 units of T 2 ribonuclease (Life Technologies, Inc.) at 37°C for 1 h, extracted with phenol/chloroform, precipitated with ethanol, and dissolved in 5 l of RNA loading buffer. The protected RNA hybrid fragments were resolved on a 6% polyacrylamide/urea gel and subjected to autoradiography.
Immunoblot Analysis of ABCA1-For immunoblot analysis of ABCA1, peritoneal macrophages were washed and scraped in PBS and lysed in 10 mM Tris-HCl, pH 7.3, 1 mM MgCl 2 , and 0.5% Nonidet P-40 in the presence of protease inhibitors (0.5 g/ml leupeptin, 1 g/ml aprotinin, 1 g/ml pepstatin A; Roche Molecular Biochemicals). Postnuclear supernatants from cell lysates were prepared by centrifugation at 3000 ϫ g for 10 min at 4°C. Samples containing the indicated amounts of protein were reduced with 2-mercaptoethanol in gel loading buffer, fractionated by 7.5% SDS-polyacrylamide gel electrophoresis, and transferred to 0.22-m nitrocellulose membranes. Immunoblotting was performed using an anti-ABCA1 antiserum (Novus, Littleton, CO) and ECL (Amersham Pharmacia Biotech). The relative intensities of the bands were determined by densitometry (Molecular Dynamics, model 300A).
Isolation and Culture of Mouse Peritoneal Macrophages-Homozygous NPC1 were produced by intercrossing BALB/cNctr-npc1 N /ϩ (BALB-npc1 N /ϩ) mice (stock number 003092; Jackson Laboratory, Bar Harbor, ME). Mouse peritoneal macrophages were isolated from NPC1 and wild type (wt) littermates by peritoneal lavage with PBS 3 days after intraperitoneal injection with 1 ml of 3.85% thioglycollate (Becton Dickinson, Sparks, MD). The isolated cells were plated onto 24-well plates and allowed to adhere by incubation for 4 h at 37°C in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS) (Life Technologies). After removal of nonadherent cells by washing with PBS, the cells were further incubated for 2 days and then used for cholesterol labeling and efflux experiments. Meeting, PA) were added during the labeling to induce ABCA1 expression. After the labeling, the cells were washed with PBS and equilibrated with DMEM, 0.2% BSA for 1 h. Efflux was then performed as described below. For procedure b, the cells were labeled with 1 Ci/ml [ 3 H]cholesterol in 0.5 ml of DMEM supplemented with 10% FBS for 24 h. The cells were then equilibrated overnight in DMEM, 0.2% BSA with or without the LXR/RXR ligands 22(R)-hydroxycholesterol and 9-cis-retinoic acid. After washing, the cells were used for efflux experiments. For procedure c, the cells were first treated with the ligands 22(R)-hydroxycholesterol and 9-cis-retinoic acid in DMEM, 0.2% BSA overnight to induce ABCA1. Then the medium was replaced by 5 mM methyl ␤-cyclodextrin:cholesterol at molar ration 8:1 ([ 3 H]cholesterol, 1 Ci/ml) for 15 min at 37°C. After washing, the cells were used for efflux step.
[ 3 H]Cholesterol Efflux Study-After labeling and equilibration, the cells were incubated by 10 g/ml purified human apoA-I or 15 g/ml human HDL 2 in 0.5 ml of DMEM, 0.2% BSA with or without the LXR/RXR ligands for 4 h. Then medium was collected and centrifuged at 6000 ϫ g for 10 min to remove cell debris and cholesterol crystal, and radioactivity in an aliquot of supernatant was determined by liquid scintillation counting. The cells were finally lysed in 0.5 ml of 0.1 M sodium hydroxide, 0.1% SDS and the radioactivity in an aliquot was determined. Cholesterol efflux was expressed as the percentage of the radioactivity released from the cells into the medium relative to the total radioactivity in cells and media.
Cholesterol Mass Analysis-The cells in 6-well plates were [ 3 H]cholesterol labeled by procedure a as described above. After 4 h of incubation with 10 g/ml apoA-I (see Fig. 2) or 15 g/ml human HDL 2 (see Fig.  6), medium was collected, and the cells were lysed in 0.1 M sodium hydroxide and 0.1% SDS. Lipids were extracted in chloroform:methanol (2:1). The organic residue was dissolved in 0.5% Triton X-100. Cholesterol was determined enzymatically (Wako Chemicals USA, Richmond, VA). Protein was determined by the Lowry method.
[ 3 H]Phospholipid Efflux Study-Macrophages in a 24-well plate were choline labeled for 24 h in 0.5 ml of DMEM, 10% FBS supplemented with 1.0 Ci/ml [ 3 H]choline (PerkinElmer Life Sciences). After overnight equilibration in DMEM, 0.2% BSA with or without the LXR/ RXR ligands treatment, the cells were washed twice in PBS, 0.2% BSA. Efflux was performed by incubation with 10 g/ml apoA-I for 4 h in 0.5 ml DMEM, 0.2% BSA with or without the ligands. Then medium was collected and centrifuged at 6000 ϫ g for 10 min to remove cell debris.
[ 3 H]Phospholipids in an aliquot of supernatant were first extracted with chloroform:methanol (2:1), and then the radioactivity was determined by scintillation counting. The cells were finally lysed in 0.5 ml of 0.1 M sodium hydroxide, 0.1% SDS, and the radioactivity in an aliquot after lipid extraction was determined. The percentage of secreted [ 3 H]phospholipid was calculated by dividing the medium-derived counts by the sum of the total (medium plus cell).
Statistical Analysis-The results are presented as the means Ϯ S.D. The tests for the significant differences between groups were performed by Student's t test.

RESULTS
We recently showed that ABCA1 mRNA is up-regulated by activation of LXR/RXR (24). To determine whether this resulted in an increase in functional ABCA1 protein, we measured cholesterol efflux to apoA-I in mouse peritoneal macrophages cells treated with the LXR/RXR ligands 22(R)hydroxycholesterol and 9-cis-retinoic acid. This treatment resulted in a marked up-regulation of ABCA1 mRNA (not shown) and protein levels (Fig. 1A) and an increase in cholesterol efflux (Fig. 1B) as anticipated (27). Surprisingly, the level of cholesterol efflux was about 2.5-fold higher in activated cells labeled with [ 3 H]cholesterol AcLDL compared with cells labeled with [ 3 H]cholesterol, 10% FBS (Fig. 1B, compare bars 4 and 2), despite comparable levels of ABCA1 expression ( Fig. 1A; note that ABCA1 protein appears as a doublet for unknown reason). AcLDL is internalized by the scavenger receptor A and accumulates primarily in late endosomes and lysosomes (28), whereas the [ 3 H]cholesterol, 10% FBS method appears to preferentially label recycling endosomes and the trans-Golgi network (29). These findings suggested the hypothesis that ABCA1 might preferentially stimulate cholesterol efflux from late endosomes/lysosomes rather than from cellular cholesterol pools labeled by [ 3 H]cholesterol, 10% FBS.
To further explore this idea, cholesterol mass and isotopic efflux to apoA-I were measured in cells labeled with [ 3 H]cholesterol, 10% FBS, with [ 3 H]cholesterol/AcLDL, or with [ 3 H]cholesterol/cyclodextrin. In the latter procedure, the cells are labeled briefly (15 min) with cyclodextrin:cholesterol (8:1, molar ratio), and the radiolabel is thought to reside mostly in the plasma membrane (30,31). These experiments showed greater isotopic and mass efflux of cholesterol in activated cells labeled with [ 3 H]cholesterol/AcLDL, compared with cells labeled in either of the other two ways (Fig. 2), consistent with the idea that the late endosomal/lysosomal cholesterol pool is a preferential source for cholesterol efflux by ABCA1. Activation of LXR/RXR did result in a marked increase in isotopic and mass cholesterol efflux from cells labeled with [ 3 H]cholesterol/ cyclodextrin, suggesting that the plasma membrane cholesterol also contributes significantly to ABCA1-mediated cholesterol efflux. However, for the [ 3 H]cholesterol, 10% FBS labeling method, isotopic efflux was only slightly increased by LXR/RXR activation, whereas cholesterol mass efflux was increased in a similar fashion to [ 3 H]cholesterol/cyclodextrin-labeled cells. This finding could arise if the radiolabel was primarily present in pools of cholesterol inaccessible to ABCA1 (i.e. recycling endosomes) (29), whereas cholesterol mass efflux reflected efflux from the plasma membrane where cholesterol would be unlabeled by this method.
To further explore the hypothesis that late endosomes/lysosomes represent a preferred source of cholesterol for ABCA1mediated cholesterol efflux, we next carried out efflux studies using macrophages from Niemann-Pick C1 mice, which have a defect in trafficking of cholesterol out of late endosomes (32). Cholesterol loading was carried out using [ 3 H]cholesterol/ AcLDL. Compared with macrophages from wild type mice, there was a profound decrease in cholesterol efflux to apoA-I in NPC1 macrophages, especially following induction of ABCA1 (Fig. 3A). Measurements of ABCA1 mRNA and protein revealed similar levels of induction in control and NPC1 macrophages (not shown). In earlier studies, Liscum et al. (22) reported that human NPC1 fibroblasts had a moderate defect in cholesterol efflux to small unilamellar vesicles; this was manifested as a delay in cholesterol efflux that became normal following longer incubation periods. However, a time course study revealed a profound 3-4-fold decrease in cholesterol efflux to apoA-I in NPC1 macrophages that was not ameliorated by prolonged incubation (Fig. 3B). If ABCA1 preferentially stimulates cholesterol efflux from late endosomes/lysosomes, then it might be anticipated that there would be a less pronounced defect in cholesterol efflux in NPC1 cells labeled with [ 3 H]cholesterol, 10% FBS. Accordingly, using this labeling method, basal cholesterol efflux to apoA-I was similar in wild type and NPC1 cells, and efflux was only moderately decreased in NPC1 macrophages compared with wild type macrophages following LXR/RXR activation (Fig. 3C). Following plasma membrane labeling with [ 3 H]cholesterol/cyclodextrin, there were identical levels of cholesterol efflux in NPC1 and wild type cells (Fig. 3D). These experiments suggest that both lysosomal and plasma membrane cholesterol pools serve as a source of cholesterol for ABCA1 and that the lysosomal pool requires the activity of the NPC1 molecule.
ABCA1 is thought to act as a phospholipid flippase at the plasma membrane (15,16). This activity may lead to the formation of phospholipid-apoA-I complexes that secondarily stimulate cholesterol efflux from a distinct region of plasma membrane (15,17). We measured phospholipid efflux to apoA-I in NPC1 and wt macrophages. Phospholipid efflux was stimulated following activation of LXR/RXR, but there was no defect in phospholipid efflux in NPC1 cells (Fig. 4). This indicates that the primary action of ABCA1, i.e. formation of phospholipid-apoA-I complexes, is intact in NPC1 cells.
Cholesterol efflux to HDL 2 was also significantly decreased in NPC1 cells loaded with [ 3 H]cholesterol/AcLDL (Fig. 5) or by the [ 3 H]cholesterol, 10% FBS method (not shown). Because HDL 2 does not interact with ABCA1 (11), this indicates a defect in cholesterol efflux via pathways not mediated by ABCA1. Interestingly, cholesterol efflux via HDL 2 was also induced by LXR/RXR activation (Fig. 5). This suggests the presence of other LXR/RXR target genes in the HDL 2 -mediated efflux pathway. Because apolipoprotein E (apoE) was recently identified as an LXR/RXR target (33), we considered the possibility that increased cholesterol efflux might be due to increased apoE synthesis by mouse peritoneal macrophages. However, LXR/RXR activation similarly increased cholesterol efflux to HDL 2 in macrophages from apoE knock-out mice (not shown), eliminating this possibility. The ability of HDL 2 to stimulate increased cholesterol efflux following LXR/RXR activation was also confirmed by cholesterol mass measurements, which indicated primarily an increase in HDL 2 free cholesterol (Fig. 6). SR-BI neutralizing antibodies (34) did not affect cholesterol efflux mediated by HDL 2 in either basal or LXR/RXR-stimulated conditions (not shown). DISCUSSION Our findings suggest that phospholipid-apoA-I complexes formed by ABCA1 initially stimulate cholesterol efflux from regions of the plasma membrane that preferentially utilize cholesterol deposited by modified LDL in late endosomes/lyso-somes rather than cholesterol deposited at other intracellular sites. The equilibration of cell surface cholesterol with these intracellular sites requires the activity of the NPC1 molecule and could perhaps also involve trafficking of the ABCA1 molecule itself (35). ABCA1 is markedly less efficient in stimulating cholesterol efflux from cells that have been labeled with [ 3 H]cholesterol, 10% FBS, which probably primarily labels recycling endosomes (29). A profound defect in ABCA1-mediated cholesterol efflux in NPC1 mutant macrophages may be an important factor explaining our recent observations showing an increase of atherosclerosis in apoE knock-out/NPC1 mutant mice, compared with apoE knock-out control mice. 2 Massive cholesteryl ester accumulation in TD macrophages indicates that ABCA1 has an essential role in mediating cho- lesterol efflux from these cells. It is likely that cholesterol from a variety of sources, including effete red cells and modified forms of LDL, is taken up by macrophages eventuating in delivery to late endosomes/lysosomes. Our studies are consistent with the idea that apoA-I/ABCA1-mediated cholesterol efflux plays an essential role in removing cholesterol from these intracellular sites. The underlying mechanisms are unclear. Intracellular trafficking of apoA-I has been reported in macrophages, and this process appears to be defective in TD (36). The trafficking of apoA-I has been suggested to have a role in mediating cholesterol efflux (37). Moreover, recent studies using fluorescence confocal microscopy have revealed trafficking of ABCA1 itself between the cell surface and intracellular sites including late endosomes/lysosomes but not recycling endosomes (35). Thus, it is conceivable that apoA-I bound to ABCA1 trafficks directly to late endosomes/lysosomes and somehow mediates cholesterol efflux from these sites. However, it is notable that ABCA1 effectively stimulates efflux of plasma membrane cholesterol, as deduced from a recently developed plasma membrane labeling procedure (30,31) (Fig. 2) and that this process is normal in NPC1 mutant macrophages (Fig. 3D). Moreover, phospholipid efflux to apoA-I is unaffected in NPC1 mutant cells (Fig. 4), suggesting that trafficking of ABCA1 to late endosomes/lysosomes is not required for formation of phospholipid-apoA-I complexes.
An alternative explanation is that phospholipid-apoA-I complexes are formed by ABCA1 at the cell surface and that these complexes then stimulate cholesterol efflux from specialized regions of the plasma membrane that preferentially obtain cholesterol from intracellular sources derived from late endosomes/lysosomes rather than recycling endosomes. We propose a model in which cholesterol trafficks from late endosomes/ lysosomes to the trans-Golgi in a process requiring the activity of NPC1 (38 -40). Once cholesterol has arrived in the Golgi, it may be formed into cholesterol/sphingolipid complexes, which give rise to cell surface cholesterol-enriched microdomains or rafts. These plasma membrane domains could then act as a preferential source of cholesterol for ABCA1-mediated efflux. This could explain why mass and isotopic efflux following AcLDL labeling is even greater than following general plasma membrane labeling with cyclodextrin. This model is not necessarily inconsistent with recent data, suggesting that ABCA1 stimulates cholesterol efflux from non-raft regions of the plasma membrane (41), because there could be several different types of cholesterol-enriched microdomains in the plasma membrane.
Following entry of lipoprotein cholesterol into late endosomes and lysosomes, NPC1 has an essential role in allowing cholesterol to gain access to the ABCA1 efflux pool (Fig. 3). Liscum et al. (22) reported a delay in cholesterol efflux to unilamellar vesicles in fibroblasts from NPC1 patients. However, with time the efflux became normal. In contrast, the apoA-I stimulated cholesterol efflux in NPC1 mutant mouse macrophages was profoundly reduced at all time points (Fig.  3B). Recently, it has been shown that following labeling of NPC1 mutant Chinese hamster ovary cells with 3 H-cholesteryl ester LDL, early time points of cholesterol efflux to cyclodextrin show no or little defect (31,42). However, after the initial appearance at the plasma membrane and subsequent internalization to an intracellular pool, cholesterol shows delayed trafficking back to the plasma membrane and poor activation of acyl-CoA-cholesterol acyltransferase in (ACAT) NPC1 mutant cells. This has led to the proposal that NPC1 acts on an intracellular pool of cholesterol that is derived from the plasma membrane and is in equilibrium with ACAT. Whether NPC1 is acting in late endosomes/lysosomes (43,44) or on another pool of cholesterol (31), our studies suggest that this pool of cholesterol represents an important source of cholesterol for ABCA1stimulated efflux.
Recently, Leventhal et al. (45) have found a defect in basal cholesterol efflux to apoA-I in acid sphingomyelinase-deficient macrophages. These studies suggest that endososomal/lysosomal sphingomyelin accumulation leads to cholesterol sequestration and, thus, defective cholesterol trafficking and efflux. The present findings extend these observations by providing direct evidence that the late endosomal/lysosomal cholesterol pool represents the preferred source of cholesterol for ABCA1mediated efflux. Also, consistent with the present findings, Kojima et al. (46) recently reported that progesterone suppressed apoA-I-mediated cellular lipid release in human fibroblasts. Progesterone has been reported to sequester cholesterol in lysosomes and block cholesterol trafficking to plasma membrane similar to the effects of the NPC1 mutation (47). However, these earlier studies (45,46) did not specifically compare cholesterol efflux from different cellular pools under conditions of high ABCA1 activities (i.e. following LXR/RXR activation and marked up-regulation of the ABCA1 (Fig. 1)) and did not specifically evaluate the effect of the NPC1 molecule in ABCA1mediated cholesterol efflux as in the present study.
Our findings that NPC1 mutant macrophages have a prominent defect in cholesterol efflux from the late endosomal/lysosomal pool but only a moderate decrease in efflux from recycling endosomal pool (Fig. 3, A and C) may explain some earlier in vivo work showing that de novo synthesized cholesterol or cholesterol entering cells through the HDL/SR-BI pathway can be metabolized and excreted normally, whereas LDL-derived cholesterol becomes sequestered in the lysosomal compartment and is metabolically inactive in NPC1 mutant mice (48,49) and in NPC1 patients (50). Labeling of cholesterol by DMEM, 10% FBS might mimic more closely the trafficking of de novo synthesized cholesterol or cholesterol derived from the HDL/SR-BI pathway, whereas labeling by AcLDL is similar to LDL cholesterol trafficking to lysosomes.
The HDL 2 -mediated cholesterol efflux pathway, distinct from that mediated by ABCA1 (12,11), was also defective in NPC1 cells. An intriguing, unexpected observation was the finding that macrophage cholesterol efflux to HDL 2 was increased by treatment with LXR/RXR activators (Fig. 5), suggesting a novel efflux process unrelated to ABCA1, SR-BI, or apoE. The mechanism of HDL 2 -mediated cholesterol efflux appears quite distinct from apoA-I-mediated cholesterol efflux. Thus, apoA-I binds and interacts with ABCA1 to mediate cho- lesterol efflux, whereas HDL 2 is inactive in this regard (11). HDL can stimulate cholesterol efflux by interacting with SR-BI (13), but this pathway does not appear to be very active in mouse peritoneal macrophages, 3 and SR-BI neutralizing antibodies had no effect on the LXR/RXR-induced cholesterol efflux to HDL 2 . These findings suggest that there is a novel molecular target of LXR/RXR activation in the cholesterol efflux pathway to HDL 2 . This may well have physiological importance because differences in overall HDL levels between different subjects, such as male/female differences, are primarily due to different HDL 2 levels, whereas HDL 3 levels are relatively constant in the population (51).