ABCG1 Redistributes Cell Cholesterol to Domains Removable by High Density Lipoprotein but Not by Lipid-depleted Apolipoproteins*

ATP binding cassette transporter G1 (ABCG1) mediates the transport of cholesterol from cells to high density lipoprotein (HDL) but not to lipid-depleted apolipoprotein A-I. Here we show that human ABCG1 overexpressed in baby hamster kidney cells in the absence of lipoproteins traffics to the plasma membrane and redistributes membrane cholesterol to cell-surface domains accessible to treatment with the enzyme cholesterol oxidase. Cholesterol removed by HDL was largely derived from these domains in ABCG1 transfectants but not in cells lacking ABCG1. Overexpression of ABCG1 also increased cholesterol esterification, which was decreased by the addition of HDL, suggesting that a proportion of the cell-surface cholesterol not removed by HDL is transported to the intracellular esterifying enzyme acyl-CoA:cholesterol acyltransferase. A 638-amino acid ABCG1, which lacked the 40 N-terminal amino acids of the predicted full-length protein, was fully functional and of a similar size to ABCG1 expressed by cholesterol-loaded human monocyte-derived macrophages. Mutating an essential glycine residue in the Walker A motif abolished ABCG1-dependent cholesterol efflux and esterification and prevented localization of ABCG1 to the cell surface, indicating that the ATP binding domain in ABCG1 is essential for both lipid transport activity and protein trafficking. These studies show that ABCG1 redistributes cholesterol to cell-surface domains where it becomes accessible for removal by HDL, consistent with a direct role of ABCG1 in cellular cholesterol transport.

Many epidemiological studies have found that plasma levels of HDL 1 are inversely related to the onset of cardiovascular disease (1)(2)(3)(4)(5), and low levels of HDL are the most common lipid abnormality seen in humans (6,7). Conversely, high levels of HDL have been recognized as a negative risk factor in the development of atherosclerosis (6). One of the most important atheroprotective roles HDL plays is in reverse cholesterol transport, where excess cholesterol that builds up in peripheral cells is transported to the liver for excretion in the bile (8). The main protein component of HDL is apolipoprotein A-I (apoA-I).
ApoA-I promotes efflux of excess cholesterol and phospholipids from cells through its direct interaction with the ATP-binding cassette (ABC) transporter A1 (ABCA1) (9 -12).
A fully functional ABC transporter has two membrane spanning domains and two nucleotide binding domains, and the hydrolysis of ATP in the nucleotide binding domains provides energy for the transport of substrates across membranes. In the artery wall, ABCA1 is thought to play an important role in the removal of lipids from cholesterol-loaded macrophages of the developing coronary plaque (13). ABCA1 expression is elevated by cholesterol loading of macrophages (9), and this is because of the activation of the liver X receptor and retinoid X receptor (LXR/RXR) heterodimer (14), which binds to the promoter of ABCA1 and activates ABCA1 transcription (15,16). Macrophage cholesterol can be converted to oxysterols that are ligands for LXR (17,18).
Genes other than ABCA1 involved in lipid metabolism are induced by activated LXR/RXR heterodimers, including those encoding the half-transporters ABCG1, ABCG5, and ABCG8 (14,19). ABCG5 and ABCG8 form an obligate heterodimer (20) that promotes phytosterol and cholesterol transport across intestinal enterocytes. Mutations in either protein can lead to sitosterolemia (21). ABCG1 appears to form homodimers that are highly expressed in macrophages (22) and facilitate the efflux of cholesterol from cells to HDL (23,24). The peroxisomal proliferator-activated receptor-␥ (PPAR␥) activators also increase ABCG1 expression in macrophages and stimulate cholesterol efflux to HDL independent of their induction of LXR (25). The targeted disruption of ABCG1 in mice on a high fat and high cholesterol diet causes the massive accumulation of both neutral lipids and phospholipids in hepatocytes and macrophages of multiple tissues, despite no changes in plasma lipids (26). Conversely, tissues in ABCG1 transgenic mice are protected from dietary fat-induced lipid accumulation (26).
In the present study, we characterized the cholesterol transport function of ABCG1. By using an inducible expression system, we demonstrate that human ABCG1 can be highly expressed in baby hamster kidney (BHK) cells, where it redistributes cholesterol to cell-surface domains that are accessible for removal by HDL but not by lipid-free apoA-I. ABCG1 traffics to the plasma membrane, probably as a homodimer, and this process requires an active nucleotide binding domain. Furthermore, a 638-amino acid form of ABCG1 that is missing the 40 N-terminal amino acids of the predicted full-length protein has full lipid transport activity and may be the predominant form of ABCG1 expressed in human macrophages. Overexpression of both a putative full-length and a 19-amino acid N-terminal truncated form of a close homolog of ABCG1, ABCG4, also increased cholesterol efflux to HDL. Thus, the N termini of the predicted full-length forms of both ABCG1 and ABCG4 are not essential for lipid transport activity.

MATERIALS AND METHODS
Cultured Cells-All cell culture incubations were performed at 37°C in a humidified 5% CO 2 incubator. BHK cells were obtained from the ATCC (Manassas, VA). BHK cells expressing human ABCG1 were generated using the mifepristone-inducible GeneSwitch system (Invitrogen). An N-terminal FLAG-tagged human ABCG1 cDNA coding for a 638-amino acid protein (GenBank TM accession number NP_004906.1) was a generous gift from David Wade (CV Therapeutics). Cells were transfected initially with the pSwitch plasmid using FuGENE 6 (Roche Applied Science). Clonal lines were isolated and then transfected with pGene/V5-His/lacZ and assayed for ␤-galactosidase activity with the ␤-galactosidase assay kit (Invitrogen). Clonal pSwitch lines that gave the highest mifepristone-induced ␤-galactosidase activity were then transfected with linearized pGene/V5-HisA containing the FLAGtagged 638-amino acid human ABCG1. An N-terminal FLAG-tagged cDNA clone containing the reported full-length 678 amino acids (based on GenBank TM accession number NP_004906.3) was constructed by ligation of cDNA coding the 40 extra upstream amino acids into the original clone. The cDNA was obtained using the SuperScript Firststrand Synthesis System for RT-PCR (Invitrogen) and subsequent PCR. Total RNA for the RT-PCR was obtained from THP-1 macrophages treated with acetylated LDL and the LXR agonist 22-hydroxycholesterol (27). Site-directed mutagenesis of an amino acid residue common to all ABC transporter nucleotide binding domains (NCBI conserved domains data base, cd00267.1) (28) within the Walker A motif (29) of ABCG1 (glycine121 to alanine, GenBank TM accession number NP_004906) was carried out using the QuikChange XL site-directed mutagenesis kit (Stratagene, CA). A human IMAGE clone (5763981; ATCC, Manassas, VA) coding for ABCG4 was used to amplify the predicted full-length 646-amino acid open reading frame, which was then ligated into pGene/V5 HisA in-frame with the C-terminal V5-His tag. In a similar manner, a 627-amino acid open reading frame, which lacked the first 19 amino acids of the predicted full-length protein, was also constructed. Mock-transfected BHK cells were derived from the same pSwitch clonal line transfected with linearized pGene/V5-HisA. Cells were grown and maintained in DMEM containing 10% fetal bovine serum until experimental treatments. Unless indicated otherwise, ABCG1 and ABCG4 were induced by incubating cells for 18 -20 h in DMEM with 1 mg/ml fatty acid-free bovine serum albumin and 10 nM mifepristone.
HDL and Apolipoprotein A-I-HDL was prepared by sequential ultracentrifugation in the density range 1.125-1.21 g/ml and was depleted of apolipoprotein E (apoE) and apoB by heparin-agarose chromatography, and apoA-I was purified from HDL and subsequently delipidated (30).
To measure lipid efflux, cells were incubated with DMEM/bovine serum albumin with or without 10 g/ml apoA-I or 25 g/ml HDL for 2-6 h at 37°C and chilled on ice, and the medium was collected and centrifuged to remove detached cells. For cholesterol efflux, the medium was counted for 3 H, and the cells were assayed for free and esterified [ 3 H]cholesterol after isolation by TLC (31).
To measure cell-surface cholesterol redistribution, cells were washed once with PBS and then incubated with 1 unit/ml cholesterol oxidase (Calbiochem) in DMEM at 37°C for 10 min (32). Cells were then washed twice with PBS; cellular lipids were extracted, and [ 3 H]cholesterol and [ 3 H]cholestenone were measured after isolation by TLC.
Immunoblotting-Control, ABCG1-, and ABCG4-transfected BHK cells were solubilized in 50 mM Tris buffer containing 1% SDS, 0.1 M mercaptoethanol, and 0.5 mM EDTA, and proteins were resolved by 8% PAGE. Proteins were transblotted onto nitrocellulose, and ABCG1 was identified with an anti-FLAG antibody conjugated to horseradish peroxidase (Sigma) or an anti ABCG1 antibody (Santa Cruz Biotechnology) using an enhanced chemiluminescence detection assay. ABCG4 was identified using an anti-V5 antibody conjugated to horseradish peroxidase (Invitrogen).
Membrane Labeling of ABCG1-For selective labeling of plasma membrane ABCG1, cells were incubated for 30 min at 4°C with PBS containing 1 mg/ml sulfo-N-hydroxysuccinimide-biotin to biotinylate cell-surface proteins (9,33). Cells were washed and dislodged from the dish at 4°C in Tris-buffered saline. To isolate ABCG1, cell proteins were solubilized in buffer containing 1% Triton X-100 plus protease inhibitors, and the extract was incubated overnight at 4°C with a 1:1000 dilution of anti-FLAG antiserum (Stratagene, CA) raised against the FLAG peptide. The antibody-antigen complex was isolated by protein G-coated magnetic beads (Dynal) and electrophoresed in SDS using an 8% polyacrylamide gel. Each gel lane received immunoprecipitated protein corresponding to equal amounts of cells. To measure biotinylated ABCG1, proteins were transferred to nitrocellulose and identified with a streptavidin-horseradish peroxidase ECL assay (Pierce).
Cross-linking of ABCG1-To assay if ABCG1 exists as a homodimer at the plasma membrane, BHK cells expressing ABCG1 were washed twice with ice-cold PBS and incubated for 30 min at 0°C with PBS containing 1 mg/ml disuccinimidyl suberate (DSS), a homobifunctional, noncleavable, amine reactive cross-linking agent. The cells were then washed twice with cold PBS containing 20 mM glycine (to quench cross-linker) and twice with cold PBS. The existence of ABCG1 homodimers was assayed using SDS-PAGE of solubilized cellular proteins probed with the anti-FLAG antibody.
Statistics-Data were analyzed by paired Student's t test to determine significance. Each experiment shown is representative of at least three similar experiments. The apparent K d value for HDL-mediated cholesterol efflux specifically to ABCG1 was calculated using the onesite saturation model of nonlinear regression.

Induction of ABCG1 Alters the Distribution of Cellular
Cholesterol-To examine the effects of ABCG1 expression on cellular cholesterol homeostasis, BHK cells were stably transfected with an N-terminal FLAG-tagged human ABCG1 cDNA, which was inducible by treatment with mifepristone. As a control, cells were also transfected with the same vector lacking the ABCG1 cDNA. This inducible system allows for time-dependent expression of high levels of ABCG1 and minimizes the potentially cytotoxic effects of the long term overexpression of ABCG1 in cell membranes.
The mifepristone-treated mock-transfected cells had no detectable FLAG tag, but the ABCG1-transfected cells expressed a FLAG-tagged protein of ϳ60 kDa (Fig. 1A). A commercially available ABCG1 antibody also detected the FLAG-tagged protein but did not detect ABCG1 in mock-transfected cells (results not shown). As has been demonstrated previously (23,24), the increase in ABCG1 expression led to an increase in HDLmediated cholesterol efflux from cells but no significant change in the efflux of cholesterol to lipid free apoA-I (Fig. 1B). Because the population of ABCG1-transfected BHK cells was mixed, a clonal line was isolated that had a higher level of ABCG1 expression. This line also had a significantly higher level of HDL-mediated cholesterol efflux compared with the mixed population (ϳ280%) and was used for further studies.
We tested the possibility that ABCG1 expression alone could alter cellular cholesterol homeostasis by radiolabeling cellular cholesterol, treating cells for 18 h with increasing concentrations of mifepristone, and measuring labeled cholesterol distribution between cellular pools. The relative distribution of cellsurface free cholesterol was determined by treating cells for 10 min at 37°C with the enzyme cholesterol oxidase and assaying for cellular cholestenone. Mock-transfected cells had low levels of oxidizable labeled cholesterol (ϳ4%) and esterified cholesterol (ϳ5%), which did not change upon the addition of mifepristone (data not shown). Furthermore, the HDL-mediated efflux of labeled cholesterol from these cells did not change with increasing concentrations of mifepristone. In contrast, mifepristone treatment increased the fraction of oxidase-accessible labeled cholesterol in ABCG1 transfectants from 4% to over 10% and increased the fraction of labeled cholesterol that was esterified from 5 to 20% (Fig. 2). When these cells were subsequently incubated for 4 h with HDL, cholesterol efflux increased from 3 to 15% with increasing mifepristone treatment ( Fig. 2). ABCG1 is thus altering the distribution of membrane cholesterol, and this is occurring independent of cellular lipoprotein interactions.
We examined the effects of increased HDL levels on radiolabeled cholesterol efflux and cellular cholesterol pool sizes in cells lacking or expressing ABCG1. With both mock-and ABCG1-transfected cells, increasing HDL levels up to 400 g of protein/ml progressively increased cholesterol efflux, but this was greater for the ABCG1-expressing cells at all concentrations (Fig. 3A). The difference in cholesterol efflux between the two cell types revealed a single saturable component. By using the one-site saturation model of nonlinear regression, the apparent K d value for the reaction was 31 g/ml HDL protein with a standard deviation of 5.7 g/ml and an R 2 value of 0.98 (Fig. 3A, inset). This ABCG1-dependent cholesterol efflux was largely derived from a cellular pool that was accessible to cholesterol oxidase (Fig. 3B). ABCG1-independent cholesterol efflux was almost exclusively from the nonoxidizable, unesterified cholesterol pool (Fig. 3B). This suggests that inducing ABCG1 alters the distribution of cholesterol within the plasma membrane, making it more accessible to both cholesterol oxidase and HDL.
Inducing ABCG1 increased the fraction of cholesterol that was esterified (Fig. 2), suggesting that ABCG1 also promotes flux of cholesterol into the substrate pool for acyl-CoA:cholesterol acyltransferase (ACAT), which resides in the endoplasmic reticulum. In the above experiment, HDL had no effect on this fractional esterification because it was added after these esters were formed. To determine whether HDL affects this flux, we radiolabeled plasma membrane cholesterol for 1 h, chased cells without or with HDL for 4 h, and measured the labeled free and esterified cholesterol content. After the chase without HDL, the labeled free cholesterol content of mock cells was higher than that of ABCG1 transfectants (Fig. 4A). Only a small fraction of the radiolabeled cholesterol became esterified in the mock cells (ϳ2%) after the short term labeling and chase incubations, but the ABCG1 transfectants had a significantly higher content of labeled cholesteryl esters (ϳ7%, Fig. 4B). These results are consistent with ABCG1 promoting the flux of cholesterol into the ACAT substrate pool. Addition of HDL reduced both the free and esterified cholesterol content of the cells (Fig. 4, A and  B). In mock cells, the reduction in labeled free cholesterol was ϳ15%, and for esterified cholesterol the reduction was 19%. The reduction in the ABCG1 transfectant was greater; ϳ24% for free cholesterol and ϳ30% for esterified cholesterol. These data imply that a substantial amount of the increased cholesteryl esters in ABCG1-expressing cells is derived from the free cholesterol pool that is removable by HDL.
Because there is a marked increase in cholesteryl ester content in ABCG1 transfectants, it is possible that this alteration in the cellular cholesterol pool is necessary for ABCG1 function.
To determine whether the esterification of cholesterol was necessary for the increased HDL-mediated cholesterol efflux in ABCG1 transfectants, the ACAT inhibitor compound 58035 (3-(decyldimethylsilyl)-N-[2-(4-methylphenyl)-1-phenylethyl] propanamide) was used to prevent the esterification of cholesterol. The cholesteryl ester content of both mock and ABCG1 transfectants was significantly reduced with 58035 ( Fig. 4C), but this did not affect HDL-mediated cholesterol efflux, implying that the esterification of cholesterol was not necessary for ABCG1 cholesterol efflux activity.
ABCG1 Is Expressed at the Cell Membrane and Requires an Active Nucleotide Binding Domain-Results showing that induction of ABCG1 increases oxidizable cholesterol at the cell surface support the possibility that this transporter is localized to the plasma membrane. To test this, we biotinylated cellsurface proteins, isolated ABCG1 by immunoprecipitation and SDS-PAGE, and identified biotinylated ABCG1 with a streptavidin assay. ABCG1-but not mock-transfected cells expressed a FLAG-tagged protein at the cell surface of the size expected for ABCG1 (Fig. 5A). These data indicate that ABCG1 traffics to the plasma membrane where it may form oxidizable, HDLaccessible cholesterol pools.
We used a cross-linking reagent, DSS, to determine whether FLAG-tagged ABCG1 expressed at the plasma membrane is dimerized. Results showed that cross-linking formed an anti-FLAG antibody immunoreactive protein of approximately twice the size of monomeric FLAG-tagged ABCG1 (Fig. 5B). Because DSS does not enter the cell and has a linker of only 11 Å, it is likely that the immunoreactive protein is an ABCG1 homodimer expressed at the cell surface. We also observed, however, a cross-linked immunoreactive band with a molecular mass consistent with a trimer (ϳ180 kDa). Thus, we cannot exclude the possibility that ABCG1 is forming a complex with proteins other than itself. Because covalent cross-linking has a relatively low efficiency, these studies do not provide an estimate of the relative fraction of ABCG1 that exists as multimers.
To determine whether the presence of a functional NBD is necessary for ABCG1 activity, we performed site-directed mutagenesis to substitute alanine for an essential glycine residue within the Walker A motif (29). The mutated protein was expressed at similar levels when compared with the wild type protein (Fig. 5C). Both wild type and mutant ABCG1 appeared as a single band on reduced and nonreduced SDS-polyacrylamide gels, suggesting that disulfide bridging does not promote the dimerization of ABCG1 required for activity. The G121A mutation almost completely abolished the ability of ABCG1 to redistribute cholesterol to oxidase-and ACAT-accessible pools and to promote cholesterol efflux to HDL (Fig. 5D). Cell-surface biotinylation studies showed that the NBD mutant did not appear on the cell surface (Fig. 5E). This might be due to the incorrect folding of the mutant protein or the failure of the protein to dimerize, which could be essential for protein trafficking. However, it would appear that a functional NBD in ABCG1 is necessary for both lipid transport activities and protein trafficking to the plasma membrane.
The 638-Amino Acid ABCG1 Protein Has Similar HDL-mediated Efflux Capabilities When Compared with the Predicted 678-Amino Acid Full-length Protein-The first published human ABCG1 amino acid sequence contained an open reading frame of 638 amino acids (GenBank TM accession number NP_004906.1 (34)). This FLAG-tagged protein was used in the above studies. More recent studies have suggested that the full-length protein is 678 amino acids (GenBank TM accession number NP_004906.3), although many other isoforms of the protein have also been proposed (27,35,36). To determine whether the 678-amino acid protein also had HDL-mediated efflux function, the predicted coding region for the upstream 40 amino acids was amplified from mRNA isolated from acetylated LDL-loaded THP-1 human macrophages (27). This was ligated appropriately into the 638-amino acid ABCG1, retaining the N-terminal FLAG tag. When expressed in BHK cells, this protein was expressed at a similar level when compared with the original transfected protein and appeared larger in size (Fig. 6A). Induction of both the 638-and 678-amino acid proteins increased the fraction of cholesterol that was accessible to cholesterol oxidase on the cell surface and to intracellular ACAT, and this was associated with an increase in cholesterol efflux to HDL (Fig. 6B). To analyze further the ability of these ABCG1 proteins to mediate cholesterol efflux to HDL, cells were incubated with or without HDL for 4 h, and changes in the cholesterol content of the cells were measured. There was no significant decrease in cholesterol content during the HDL incubation in the mocktransfected cells or in the ABCG1 G121A transfectants (Fig.  6C). However, in both the 638-and 678-amino acid ABCG1 transfectants, there was a significant decrease in cellular cholesterol content (ϳ9 and 14%, respectively). The previous studies on ABCG1, showing an increased efflux of cholesterol to HDL, did not address the actual size of the mature protein (24,26). Although the larger form may be modestly better at promoting cholesterol esterification (Fig. 6B), the 638-and 678amino acid proteins did not differ significantly in mediating cholesterol efflux, suggesting that the N-terminal domain is not critical for this function.
We examined the size of the naturally occurring human ABCG1 protein by immunoblot analysis. The 638-and 678amino acid ABCG1s were compared with the protein expressed in acetylated LDL-loaded, monocyte-derived human macrophages. Based on gel migration, the 638-amino acid protein (which has only an additional 11 amino acids due to the FLAG tag) appeared to be closest in size to the human macrophage form of ABCG1 (Fig. 6D). This is also in agreement with the size of the ABCG1 protein expressed in both human umbilical vein endothelial cells and human aortic cells (37). We sometimes observed a faint immunoreactive band in the 678-amino acid ABCG1 of a similar size to that of the 638-amino acid ABCG1 Fig. 6D). This could either be nonspecific antibody interaction or a processed form of the protein that is similar in size to the 638-amino acid ABCG1.
To examine further the importance of the N terminus in ABCG-mediated cholesterol transport, we compared the amino acid sequences of ABCG1 and ABCG4. Like ABCG1, ABCG4 is expressed and regulated by oxysterols in monocyte-derived human macrophages and has been shown to mediate cholesterol efflux to HDL particles (24). ABCG1 and ABCG4 share a high degree of homology (84%) and identity (72%), and the original published putative 627-amino acid ABCG4 sequence (GenBank TM accession number NP_071452.1) is 19 amino acids shorter than the current putative sequence (GenBank TM accession number NP_071452.2). Despite the high overall homology between ABCG1 and ABCG4, the amino acids upstream of the 638amino acid ABCG1 and the 627-amino acid ABCG4 share very little homology, and hence we examined if the 19 extra amino acids in ABCG4 were required for function. As with expression of the two forms of ABCG1, both the 627-and 646-residue forms of ABCG4 are expressed at similar levels and are of an appropriate size when transfected into baby hamster kidney cells (Fig. 7A). The two ABCG4 isoforms expressed in this cell system were also able to mediate an increase in radiolabeled cholesterol efflux to HDL when compared with mock cells, as well as causing an increase in radiolabeled cholesterol esterification and an increase in oxidase-accessible labeled cholesterol (Fig. 7B). Thus, as with ABCG1, the N terminus of ABCG4 does not appear to be critical for its cholesterol efflux activity.

DISCUSSION
Previous studies have shown that overexpression of ABCG1 causes an increase in HDL-mediated cholesterol efflux from cells (23,24), but the mechanism by which this occurs was not described. Here we show that ABCG1 traffics to the plasma membrane, and its overexpression redistributes cholesterol to a cell-surface pool that is accessible to enzymatic oxidation and to removal by HDL particles. Mutation of an essential amino acid within the Walker A motif abolished the redistribution of cholesterol as well as ABCG1 trafficking to the plasma membrane, indicating that a functional NBD is essential for ABCG1 activity. Our studies also show that the first 40 amino acids of the predicted full-length ABCG1 protein are not essential for its activity. Furthermore, the truncated form of the protein is similar in size to ABCG1 expressed in human monocyte-derived macrophages.
Overexpression of ABCG1 in BHK cells increased radiolabeled cholesterol translocation to sites accessible to exogenously added cholesterol oxidase. Addition of HDL, but not apoA-I, promoted cholesterol efflux from these sites, indicating that they were the source of cholesterol removed by HDL. Radiolabeled cholesterol efflux was also promoted by HDL in BHK cells not expressing ABCG1. However, only the ABCG1 overexpressing cells showed a decrease in total cholesterol mass following HDL-mediated cholesterol efflux. It is thus likely that the apparent efflux of radiolabeled cholesterol in mock cells is largely due to the exchange of radiolabeled cellular cholesterol with nonlabeled HDL cholesterol. The ABCG1specific efflux of radiolabeled cholesterol to HDL was shown to have a single saturable component with a K d of ϳ30 g/ml HDL protein. These findings fit the hypothesis that ABCG1 expression is directly associated with the efflux of cellular cholesterol to HDL.
The current study does not address the physical properties of the cholesterol oxidase-accessible membrane domains in ABCG1-expresssing cells. Previous studies (32) have shown that ABCA1 also redistributes membrane cholesterol to cholesterol oxidase-accessible pools and that this cholesterol is selectively removed by apoA-I. However, apoA-I is unable to remove the oxidase-accessible cholesterol from ABCG1-expressing BHK cells, suggesting that the cholesterol is in a membrane domain different from those formed in ABCA1-expressing cells. ABCG1 could generate these cell-surface domains by several mechanisms. First, ABCG1 could redistribute cholesterol between different domains of the plasma membrane. Such a mechanism has been described for the ABC transporter Pglycoprotein, which utilizes ATP to translocate cholesterol as well as phospholipids from the cytosolic leaflet to the exoplasmic leaflet of the plasma membrane (38,39). Second, ABCG1 could be involved in the formation of lipid vesicles that traffic between intracellular membranes and the plasma membrane, thereby facilitating the delivery of cholesterol to the cell surface for removal by HDL. Consistent with this concept are studies showing that ABCA1-containing vesicles rapidly recycle between the plasma membrane and endosomes (40). Third, rather than redistribute cholesterol, ABCG1 could increase the thermodynamic activity of cholesterol and its rate of projection from the membrane.
Induction of ABCG1 also increased the flux of cholesterol into the ACAT substrate pool, causing an increase in the cholesteryl ester content of the cells. This increase was reduced in the presence of HDL. This suggests that a fraction of the ABCG1-transported cholesterol not removed by HDL translocates from the cell surface to intracellular sites, where it is esterified. Similar results were observed for cholesterol pools generated by ABCA1 (32). It is unclear if the inward flux of cholesterol is the result of recycling of the transporters or if it occurs by an independent pathway. In either case, these results raise the possibility that these ABC transporters play a role in the previously described inward flux of plasma membrane cholesterol to ACAT (41,42).
Results showing an absence of the ABCG1 NBD mutant on the cell surface suggest that a functional NBD domain is required for trafficking of ABCG1. This is not the case for many of the loss-of-function mutations in ABCA1 (43). However, none of the ABCA1 mutations examined were in either of the two NBDs. Work in our laboratory has shown that a similar substitution mutation within the Walker A motif of ABCA1 NBD1 (A937V) abolishes its lipid transport and apolipoprotein binding activities without impairing trafficking of the protein to the plasma membrane. 2 It is possible that the mutation in the NBD of ABCG1 interferes with its dimerization, which in turn prevents its movement to the plasma membrane. In support of this idea are studies showing that mutations in either ABCG5 or ABCG8, which form functional heterodimers, impair their dimerization and trafficking to the plasma membrane (20,44). Likewise, it is possible that any mutation in ABCG1 that causes a loss in activity would also prevent its dimerization. To date, no mutation in ABCG1 that affects lipid transport but not dimerization has been found.
Several studies have identified a large number of ABCG1 mRNA transcripts with alternate amino acid sequences (27,36). Based on different deduced methionine start sites, ABCG1 was predicted to encode either a 638- (34) or a 678-amino acid protein (36). Here we show that these two gene products function similarly when overexpressed in BHK cells. We also found similar cholesterol efflux activities of a putative full-length and N-terminal truncated form of ABCG4, a close structural and functional homolog of ABCG1. These findings imply that the N-terminal amino acids of these ABCG proteins are not critical for their lipid transport function, dimerization, or intracellular trafficking. It is noteworthy that the apparent mass on SDSpolyacrylamide gels of the human monocyte-derived macrophage ABCG1 (ϳ60 kDa) is similar to the 638-amino acid protein expressed in this study and is in agreement with the findings from two other studies, where it was estimated that ABCG1 had a mass of ϳ65 kDa (20,37). This suggests that the originally reported 638-amino acid form of ABCG1 may be the common native gene product. This form of ABCG1, however, has a calculated mass of 71 kDa, consistent with post-transcriptional modifications that increase its mobility in SDS-PAGE. Identification of the true N-terminal sequences of the native forms of ABCG1 will require direct sequencing of purified protein or development of domain-specific antibodies.
Our study confirms that both ABCG1 and ABCG4 mediate cholesterol efflux to HDL when expressed in the same cell type (24). ABCG4 shares the highest amino acid identity (72%) with ABCG1 among the ABCG family members. Because they have overlapping tissue expression in the brain (23,45,46) and are both expressed in macrophages (22,47), it is possible the two transporters can exist as heterodimers. In support of this idea are results showing that co-expression of a loss-of-function ABCG4 mutant with ABCG1 causes a loss of the ATPase activity of the ABCG1 (48). This would imply that both of the transporters must have active NBDs to form a functional heterodimer, as is the case for ABCG5 and ABCG8. Results showing increased cholesterol efflux with hyperexpression of either transporter in the absence of the other indicate that heterodimer formation is not obligatory for this function. Although we cannot exclude interactions with other proteins, cross-linking studies show that ABCG1 forms a multimer on the cell surface of the size predicted for a homodimer. It is possible, however, that some cells express ABCG1/ABCG4 heterodimers that have different transport functions or activities than the homodimers.
Targeted disruption of ABCG1 in mice leads to massive accumulation of neutral lipids in macrophages (26) as well as in the liver, clearly demonstrating the importance of ABCG1 in tissue lipid homeostasis. Disruption of the ABCA1 gene is also associated with accumulation of cholesterol in tissues (49), indicating that both of these transporters play a role in clearing excess cholesterol from cells in vivo. Both ABCA1 and ABCG1 appear to function as translocators of cholesterol to plasma membrane domains accessible for removal by lipid-depleted apolipoproteins and lipidated lipoproteins, respectively. In macrophages, both ABCA1 and ABCG1 are induced by oxysterols through the LXR nuclear receptor system, which is further evidence for a common role of these transporters in exporting cholesterol from macrophages. There does appear to be some dissociation in the regulation of these transporters, however, as it was shown that activation of PPAR␥ induced expression of ABCG1 in an LXR-independent manner (25), whereas PPAR␥-mediated induction of ABCA1 occurs through subsequent activation of LXR (50). Moreover, disruption of the ABCG1 gene in mice leads to tissue accumulation of triglycerides and phospholipids as well as cholesterol, whereas a lack of ABCA1 selectively increases cellular cholesterol. Additional studies are needed to define the precise role of ABCG1 in modulating cellular lipids.
In summary, the current study shows that inducing ABCG1 redistributes cholesterol to cell-surface domains, where it becomes available for removal by lipidated lipoproteins. These lipid domains are different from those formed by ABCA1, which are only accessible to lipid-free apolipoproteins. These studies show that ABCA1 and ABCG1 can act in concert to transport excess cellular cholesterol from tissue macrophages into the