Efflux of sphingomyelin, cholesterol, and phosphatidylcholine by ABCG1.

Cholesterol and phospholipids are essential to the body, but an excess of cholesterol or lipids is toxic and a risk factor for arteriosclerosis. ABCG1, one of the half-type ABC proteins, is thought to be involved in cholesterol homeostasis. To explore the role of ABCG1 in cholesterol homeostasis, we examined its subcellular localization and function. ABCG1 and ABCG1-K120M, a WalkerA lysine mutant, were localized to the plasma membrane in HEK293 cells stably expressing ABCG1 and formed a homodimer. A stable transformant expressing ABCG1 exhibited efflux of cholesterol and choline phospholipids in the presence of BSA, and the cholesterol efflux was enhanced by the presence of HDL, whereas cells expressing ABCG1-K120M did not, suggesting that ATP binding and/or hydrolysis is required for the efflux. Mass and TLC analyses revealed that ABCG1 and ABCA1 secrete several species of sphingomyelin (SM) and phosphatidylcholine (PC), and SMs were preferentially secreted by ABCG1, whereas PCs were preferentially secreted by ABCA1. These results suggest that ABCA1 and ABCG1 mediate the lipid efflux in different mechanisms, in which different species of phospholipids are secreted, and function coordinately in the removal of cholesterol and phospholipids from peripheral cells.

Cholesterol is important to the body as a component of cellular membranes and a precursor of steroid hormones. However, excess cholesterol is a risk factor for atherosclerosis. Thus, cholesterol levels are strictly regulated by synthesis and circulation in the body. Many ABC proteins are reported to function in lipid homeostasis (1). For example, ABCG5 and ABCG8 mediate the efflux of cholesterol and sitosterol from intestine and hepatocytes into intestinal lumen and bile duct (2,3). ABCB4 (multidrug resistance 3) is a phosphatidylcholine (PC) flippase and functions in the secretion of PC into bile duct from hepatocytes (4). ABCA1 mediates the efflux of cholesterol and phospholipids from macrophages to form HDL (5).
ABCG1 is a half-type ABC protein, with a nucleotide binding fold (NBF) in its N-terminal half and a transmembrane region in its C-terminal half. Human ABCG1 cDNA was cloned as a gene homologous to white in Drosophila (6), a transporter of eye pigments, and many N-terminal variant forms of ABCG1 have been reported (7,8). Other members of the ABCG subfamily form a dimer to function. For example, ABCG2 forms a homodimer to participate in multidrug resistance in breast cancer cells (9), whereas ABCG5 and ABCG8 form a heterodimer to function in the efflux of sterol from cells in the small intestine and in hepatocytes (10,11). Cserepes et al. (12) showed that a WalkerA lysine mutant of ABCG4 inhibited the ATPase activity of ABCG1 in a dominant-negative manner and suggested that ABCG1 can heterodimerize with ABCG4, which is most homologous to ABCG1 among the ABCG subfamily. Cross-linking experiments suggested that ABCG1 forms a homodimer (13). However, it has not been clearly shown that ABCG1 functions as a homodimer.
It is thought that ABCG1 is involved in transporting cholesterol, because it is induced upon the loading of macrophages with cholesterol via a pathway of nuclear hormone receptors, liver X receptor (LXR) and retinoid X receptor (RXR) (7,14,15). The ABCG1 gene is expressed in lung, brain, spleen, and macrophages. In liver, ABCG1 is expressed mainly in Kupffer cells (16). Expression of ABCG1 was increased in macrophages from patients with Tangier disease compared with control macrophages (17). Kennedy et al. (18) reported that disruption of ABCG1 in mice on a high-fat, high-cholesterol diet showed accumulation of both neutral lipids and phospholipids in hepatocytes and macrophages, whereas overexpression of ABCG1 protected murine tissues from lipid accumulation. Endogenous ABCG1 is reported to localize to the perinuclear region and, in some cases, is distributed in the plasma membrane of macrophage-derived foam cells (14,17). It was reported that ABCG1 mediates the efflux of cholesterol from cells to HDL-2 or HDL-3 but not to lipidpoor apolipoprotein A-I (apoA-I) (19) and that ABCG1 redistributes cholesterol to cell surface domains accessible for removal by HDL (13). Consistent with these reports, inhibition of ABCG1 protein expression resulted in reduced HDL-3-dependent efflux of cholesterol and phospholipids in macrophages (14). These findings suggest that ABCG1 is involved in lipid efflux in peripheral cells, like ABCA1. However, the mechanism of efflux by ABCG1 is not clear.
In this study, we investigated the subcellular localization and function of ABCG1. We demonstrated that ABCG1 localized to the plasma membrane in a HEK293 stable cell line and that ABCG1 mediates the efflux of cholesterol and phospholipids [preferentially sphingomyelin (SM)] from cells.

Materials
Rabbit polyclonal anti-ABCG1 antibody, goat polyclonal anti-ABCG1 antibody, mouse monoclonal anti-myc antibody, goat polyclonal anti-myc antibody, and mouse monoclonal anti-FLAG antibody were purchased from Santa Cruz Biotechnology. Rabbit polyclonal anti-ABCG8 antibody and mouse monoclonal anti-PentaHis antibody were obtained from Novus Biologicals and Qiagen, respectively. Rabbit polyclonal anti-FLAG and anti-BSA antibodies were purchased from Sigma. Mouse monoclonal anti-ABCA1 antibody was prepared as described previously (20). ABCG5 and ABCG8 cDNA were cloned from a human liver cDNA library. Dithiobis (succinimidylpropionate) (DSP), dimethyl 3,39dithiobispropionimidate-HCl (DTBP), and sulfo-N-hydroxysuccinimidobiotin (sulfo-NHS-biotin) were purchased from Pierce. Marathon cDNA libraries from fetal liver or placenta were obtained from Clontech. 8-Azido-[a-32 P]ATP was purchased from Affinity Labeling Technologies. Other chemicals were purchased from Sigma, Amersham Biosciences, Wako Pure Chemical Industries, and Nacalai Tesque.
Cloning of ABCG1 cDNA ABCG1 cDNA was cloned from a human fetal liver and placental cDNA library. The cDNA sequence cloned in this study was the same as that reported by Kennedy et al. (7) (GenBank accession number NM_207630) except for its N terminus, which corresponded to a variant reported by Chen et al. (6) (GenBank accession number X91249). A WalkerA lysine mutant ABCG1 (ABCG1-K120M) was prepared with the QuikChange II Site-Directed Mutagenesis Kit (Stratagene) as described by the manufacturer. The cloned cDNA was inserted into the Not I site of pcDNA3.1(1) (Invitrogen) to make an expression vector for pcDNA3.1(1)/ABCG1 and pcDNA3.1(1)/ABCG1-K120M, or into the NotI-XbaI site of pcDNA3.1(1)A-mycHis (Invitrogen) or pcDNA3.1(1)A-FLAG [prepared by inserting FLAG tag sequences instead of myc and His tag sequences of pcDNA3.1(1)-A-mycHis] to make an expression vector for pcDNA3.1(1)A-mycHis/ABCG1 or pcDNA3.1(1)A-FLAG/ABCG1.

Cell culture
HEK293 cells were grown in DMEM supplemented with 10% (v/v) FBS in 5% CO 2 at 37jC. THP-1 cells were grown in RPMI 1640 medium supplemented with 10% (v/v) FBS in 5% CO 2 at 37jC. The differentiation of THP-1 monocytes into macrophages was induced with 0.2 mg/ml phorbol 12-myristrate 13acetate (Wako Pure Chemical) for 4 days. The differentiated cells were cultured in RPMI 1640 medium and 0.2% BSA for 24 h, and ABCG1 expression was induced for 24 h by adding TO901317 (Cayman).

Glycosylation of ABCG1
Digestion with endoglycosidase H and peptide N-glycosidase F (New England Biolabs, Beverly, MA) was done as described by the manufacturer. In brief, 20 mg of membrane protein from cells was treated with 500 units of endoglycosidase H or 0.3 units of peptide N-glycosidase F for 1 h at 37jC. The deglycosylated proteins were electrophoresed on a 10% SDS-polyacrylamide gel and immunodetected using rabbit polyclonal anti-ABCG1 antibody.

Immunostaining and fluorescence microscopy
Cells were cultured on glass cover slips, fixed with 4% paraformaldehyde in PBS 1 (phosphate-buffered saline containing 0.1 g/l CaCl 2 and MgCl 2 6H 2 O), and permeabilized with 0.4% Triton X-100 in PBS 1 for 5 min. To diminish the nonspecific binding of antibodies, the cells were incubated in 10% goat serum in PBS 1 . Cells were incubated for 1 h with rabbit polyclonal anti-ABCG1 antibody diluted 1:500 in PBS 1 containing 10% goat serum and then incubated with fluorescent Alexa 488conjugated anti-rabbit IgG (Molecular Probes) for 1 h. Cells were directly viewed with a 633 Plan-Neofluar oil-immersion objective using a Zeiss confocal microscope (LSM5 Pascal).

Biotinylation of cell surface proteins
Cells were washed with ice-cold PBS 1 and incubated with 0.5 mg/ml sulfo-NHS-biotin solubilized in PBS 1 for 30 min on ice in the dark. Cells were washed with PBS 1 to remove unbound sulfo-NHS-biotin and lysed in RIPA buffer [20 mM Tris-Cl (pH 7.5), 1% Triton X-100, 0.1% SDS, and 1% sodium deoxycholate] containing protease inhibitors [100 mg/ml (p-amidinophenyl)methanesulfonyl fluoride, 2 mg/ml leupeptin, and 2 mg/ml aprotinin). ImmunoPure Immobilized Monomeric Avidin Gel (Pierce) was added to the cell lysate to precipitate the biotinylated proteins. The biotinylated proteins were electrophoresed on a 10% SDS-polyacrylamide gel and immunodetected.

Cross-linking
Cells were washed with cold PBS and incubated with 250 mM DSP or DTBP at room temperature for 30 min. The cross-linking reaction was terminated by the addition of Tris-Cl buffer (pH 7.5) to 20 mM, and cells were incubated at room temperature for 15 min. Cells were washed with PBS and lysed in Nonidet P-40 lysis buffer. Samples were denatured in SDS sample buffer with or without DTT, electrophoresed on a 7% SDS-polyacrylamide gel, and immunodetected.

Photoaffinity labeling
Membranes (20 mg of proteins) from HEK293 cells, prepared as described previously (21), were incubated with 50 mM 8-azido-[a-32 P]ATP in 3 ml of TEM buffer [40 mM Tris-Cl (pH 7.5), 0.1 mM EGTA, and 1 mM MgCl 2 ] containing 2 mM ouabain for 10 min on ice. Proteins were ultraviolet light-irradiated for 3 min (at 254 nm, 5.5 mW/cm 2 ) on ice. To remove free 8-azido-[a-32 P]ATP, ice-cold TEM buffer was added to the sample and the supernatant was removed after centrifugation (15,000 g, 5 min, 2jC). Pellets were resuspended in 100 ml of RIPA buffer containing 100 mg/ml (p-amidinophenyl)methanesulfonyl fluoride, and membrane proteins were solubilized for 30 min at 4jC. After centrifugation for 15 min at 15,000 g, the lysates were incubated with antibodies and immunoprecipitated with protein G-Sepharose 4B Fast Flow (Sigma). The immunoprecipitated proteins were washed with RIPA buffer, and samples were electrophoresed on a 10% SDS-polyacrylamide gel and autoradiographed.

Fractional lipid release assay
Cells were subcultured on 12-well plates at a density of 5.0 3 10 5 cells. After incubation for 24 h, the cells were labeled with 2 mCi/ml [ 3 H]cholesterol or [ 3 H]choline for 24 h in DMEM containing 10% FBS. The cells were washed with fresh medium and incubated with DMEM containing 0.02% BSA in the absence or presence of 10 mg/ml apoA-I (Calbiochem) or 20 mg/ml HDL (Calbiochem) for 4 h. For cholesterol efflux, the medium was collected and the cells were lysed with 0.1 N NaOH and 0.1% SDS. For phospholipid efflux, phospholipids were extracted from the medium with chloroform-methanol (2:1) or from the cells with hexane-isopropanol (3:2). PC and SM were separated by TLC on silica gel 60 plates (Merck) developed in chloroformmethanol-acetic acid-water (60:30:10:5). The radioactivity was counted by liquid scintillation counting.

Cellular lipid release assay
Cells were subcultured on six-well plates at a density of 1.0 3 10 6 cells. After incubation for 24 h, the cells were washed with fresh medium and incubated with DMEM containing 0.02% BSA in the absence or presence of 10 mg/ml apoA-I. The lipid content in the medium was determined after 24 or 48 h of incubation as described previously (22).

Mass spectrometric analysis
Cells were subcultured on six-well plates at a density of 1.0 3 10 6 cells. After incubation for 24 h, the cells were washed with fresh medium and incubated with DMEM containing 0.02% BSA in the absence or presence of 10 mg/ml apoA-I for 48 h. The medium was centrifuged for 15 min at 7,000 g to remove cell debris twice. The lipids were extracted from 12 ml of medium by the method of Bligh and Dyer (23) after the addition of di-14:1 PC as an internal standard. The content of choline phospholipids in the medium was analyzed by MS. Mass spectrometric analyses were performed with a triple quadrupole instrument model Quattro micro (Micromass, Manchester, UK) equipped with an electrospray source as described previously (24). The samples were provided by the UltiMate high-performance liquid chromatography system (LC Packings, San Francisco, CA) into the electrospray interface at a flow rate of 4 ml/min in a solvent system of acetonitrile-methanol-water (2:3:1) containing 0.1% ammonium formate (pH 6.4). The mass spectrometer was operated in the positive and negative scan modes. The flow rate of the nitrogen drying gas was 12 l/min at 80jC. The capillary and cone voltages were set at 3.7 kV and 30 V, respectively, argon at 3-4 3 10 4 Torr was used as the collision gas, and a collision energy of 30-40 V was used to obtain fragment ions for precursor ions. The relationship between peak height and amount of SM was examined using bovine brain SM (Avanti).

Statistical analysis
Values are presented as means 6 SD. Statistical significance was determined by Student's t-test. A value of P , 0.05 was considered statistically significant.

Localization of ABCG1 in plasma membrane
ABCG1 was reported to be distributed mainly in the perinuclear region and only partly in the plasma mem-brane in macrophages (14,17). However, immunostaining with anti-ABCG1 antibody suggested that the exogenously expressed ABCG1 was distributed mainly in the plasma membrane of HEK293 cells (Fig. 2B, E). No signal was detected in host HEK293 cells ( Fig. 2A). ABCG1-K120M was also detected mainly in the plasma membrane (Fig. 2C, F), suggesting that the WalkerA lysine mutation did not affect the subcellular localization of ABCG1.
To confirm the cell surface expression of ABCG1, membrane proteins were biotinylated by sulfo-NHS-biotin and precipitated by avidin agarose (Fig. 3, upper panel). We also examined ABCA1 and ABCG8 as a positive and a negative control, respectively. ABCA1, which is localized mainly in the plasma membrane (20), was precipitated by avidin agarose after biotinylation (lane 2) but not without biotinylation (lane 1). On the other hand, ABCG8, which is distributed to the endoplasmic reticulum when expressed alone without ABCG5 (10), was not precipitated by avidin agarose (lane 4). ABCG1 and ABCG1-K120M were precipitated by avidin agarose after biotinylation (lanes 8, 10) but not without biotinylation (lanes 7,9). These results indicate that ABCG1 localizes to the plasma membrane.

Homodimerization of ABCG1
Half-type ABC proteins function as homodimers or heterodimers. ABCG2 localizes to the plasma membrane (25,26) and functions as a homodimer (9), whereas ABCG5 and ABCG8 localize to the plasma membrane and function as heterodimers (10,11). To examine whether ABCG1 forms a homodimer, ABCG1-myc and ABCG1-FLAG, in which each tag sequence was fused to the C terminus, were coexpressed, and immunoprecipitation was done with antibodies against tag sequences (Fig. 4, upper  panel). When ABCG1-FLAG was immunoprecipitated with an anti-FLAG antibody, ABCG1-myc was coprecipitated (Fig. 4A, lane 4), and vice versa (Fig. 4B, lane 4). These results suggest that ABCG1 molecules interact with each other to form a homodimer or a homooligomer.
ABCG2 was reported to form a homodimer via a thiol bond (9) and to form homooligomers (27). The oligomeric features of ABCG1 were analyzed by chemical crosslinking with DSP and DTBP (Fig. 5). DSP and DTBP are cross-linking reagents with arm lengths of 12 and 11.9 Å , respectively, and both can be cleaved when reduced with thiols. When ABCG1 in living cells was cross-linked by DSP (lane 8) or DTBP (lane 14), ABCG1 with a molecular size of z130 kDa was predominantly detected, which corresponds to the size of dimeric ABCG1, and was detected as a monomer when applied to SDS-PAGE after DTT treatment (lanes 11, 17). ABCG1 with higher molecular sizes was faintly detected. The electrophoretic mobility of ABCG1, without chemical cross-linking treatment on SDS-PAGE, was not altered when the sample was denatured in the absence of DTT (lane 2). ABCG1-K120M showed the same features of dimeric formation (lanes 9, 15) as the wild type. These results suggest that ABCG1 exists predominantly as a dimer without forming a thiol bond and that the WalkerA lysine mutation does not affect the dimeric formation of ABCG1.

Binding of ATP by ABCG1
ABC proteins bind nucleotides at their NBFs and transport substrates using the energy of ATP hydrolysis. The binding of ATP by ABCG1 was analyzed using a photoaffinity labeling technique (Fig. 6). ABCB1 (multidrug resistance 1) was specifically photoaffinity-labeled with 8-azido-ATP as reported previously (lane 6) (28). ABCG1 was photoaffinity-labeled with 25 mM 8-azido-[a-

Efflux of cholesterol and phospholipids by ABCG1
Because ABCG1 mRNA expression is induced by ligands for RXR/LXR as is ABCA1 mRNA expression (7,14,15), ABCG1 may be involved in cholesterol and phospholipid homeostasis in the cell. To examine the possibility that ABCG1 is involved in lipid efflux from cells as is ABCA1, . HEK/ABCA1 mediated slightly the higher efflux of cholesterol (7.8 6 1.5%) than host HEK293 cells in the presence of HDL, but the difference was not significant. As reported previously (13,19), the presence of HDL enhanced the cholesterol efflux from HEK/ABCG1 cells (11 6 1.0%) compared with host HEK293 and HEK/ ABCA1 cells. Noteworthy, HEK/ABCG1 mediated significantly higher efflux of cholesterol compared with the host HEK293 cells in the absence (3.6 6 0.78%) or presence   These results suggest that intact NBF is essential for the efflux of cholesterol and phospholipids.
Because labeling of intracellular cholesterol pools varies with methods used to deliver the labeled cholesterol or its precursors (29,30), we further examined the efflux of cellular total cholesterol and choline phospholipids using colorimetric enzyme assays (31) (Fig. 8), by which we successfully measured apoA-I-dependent lipid efflux by ABCA1 and ABCA7 (32)(33)(34). The medium of host HEK293 cells contained minimal amounts of cholesterol and phospholipids. As reported previously (20) The lipid contents were markedly reduced in the medium of HEK/ABCG1-K120M, suggesting that the binding and/or hydrolysis of ATP is required for the efflux of cholesterol and choline phospholipids by ABCG1. The intracellular contents of free cholesterol, total cholesterol, and phospholipids of HEK/ABCG1 were 15 6 0.73, 16 6 0.58, and 61 6 0.50 mg/well, respectively, and not significantly different from those of HEK293 host cells. Therefore, 5.5% and 1.5% of cellular total cholesterol and phospholipids were   [13][14][15][16][17][18]. Cell lysates (20 mg of protein) were treated without (lanes 1-3, 7-9, 13-15) or with DTT (lanes 4-6, 10-12, [16][17][18]. Samples were separated by 7% polyacrylamide gel electrophoresis and detected with rabbit polyclonal anti-ABCG1 antibody. secreted from HEK/ABCG1 in 48 h. In these experiments, expression levels of the wild-type and mutant ABCG1 were similar, and addition of apoA-I did not affect the expression (data not shown). Cell viabilities of HEK293 and HEK/ ABCG1 cells [examined by 2,3-Bis(2-methoxy-4-nitro-5sulfophenyl)-2H-tetrazolium-5-carboxanilide inner salt assay] were similar at 48 h and .98% (data not shown). To confirm that these results did not reflect a clonal bias, we examined another stable cell line of HEK/ABCG1 and found that they mediated the efflux of cholesterol and phospholipids in a similar manner (see supplementary Fig. II).

BSA-dependent efflux of lipids by ABCG1
ABCG1 mediated the efflux of cholesterol and phospholipids even without apoA-I being added to the medium. Because the medium was supplemented with 0.02% BSA and BSA is believed to serve as a cholesterol acceptor in sperm capacitation (35), we examined the possibility that BSA works as an acceptor of cholesterol and phospholipids transported by ABCG1 (Fig. 9A, B). The cholesterol and phospholipid efflux by ABCG1 increased in a BSA concentration-dependent manner up to 0.02%. Surprisingly, a substantial efflux of cholesterol (0.61 6 0.17 mg/well) and phospholipids (0.40 6 0.023 mg/well) was observed even without BSA being added to the medium. Next, we examined whether serum proteins, including albumin, still remained in the medium after cells were washed with fresh medium: we found that significant amounts (z0.0002%) of serum albumin from FBS remained, as judged from Western blotting (see supplementary Fig. III). This suggests that serum albumin may function as an acceptor for cholesterol and phospholipids transported by ABCG1.

Efflux of SM by ABCG1
HEK/ABCA1 and HEK/ABCG1 mediated the efflux of similar amounts of lipids with similar ratios between cholesterol and phospholipids in the presence of apoA-I, as shown in Fig. 8. The difference between the functions of ABCA1 and ABCG1 was that ABCG1 mediated the efflux even in the absence of apoA-I. We speculated that the mechanism of lipid efflux mediated by ABCG1 was different from that by ABCA1 and that phospholipid species secreted by ABCA1 and ABCG1 might be different. It has been reported that ABCA1 mediates the efflux of phospholipid, mostly PC (36). We compared the species of choline phospholipids transported by ABCA1 in the presence of apoA-I and by ABCG1 in the presence of BSA by analyzing molecular species of phospholipids in the medium with mass spectrometry (Fig. 10).
The media of HEK/ABCG1 cells in the presence of BSA (Fig. 10B) and HEK/ABCA1 cells in the presence of apoA-I (Fig. 10C) contained higher amounts of various choline phospholipids (SM 16:0-18:1, PC 16:0-16:1, PC 16:0-18:1, PC 18:0-18:2, and SM 24:1-18:1) than that of the host HEK293 cells in the presence of BSA (Fig. 10A). The major differences between HEK/ABCG1 and HEK/ABCA1 are the peak heights of SM 16:0-18:1 and SM 24:1-18:1. In the medium of HEK/ABCG1 cells, the peak height of SM 16:0-18:1 was much higher than that of PC 16:0-16:1, whereas it was lower than that of PC 16:0-16:1 in the medium of HEK/ABCA1 cells. The peak height of SM 24:1-18:1 was also higher than that of PC 18:0-18:2 in the medium of HEK/ABCG1 cells, whereas it was lower than that of PC 18:0-18:2 in the medium of HEK/ABCA1 cells. Although ion peaks from a triple quadupole mass spectrometer do not allow for direct comparison between phospholipid species, the relationship between the amount of bovine brain SM and the peak height of SM 16:0-18:1 was linear from 25 to 900 mg/ml, as shown in Fig. 10C (inset). Furthermore, the relationship of the peak height of SM 16:0 was linear from 1 to 100 pmol/ml compared with 1 pmol/ ml PC 16:0-16:0 (see supplementary Fig. IV). Therefore, a higher content of SM 16:0-18:1 and SM 24:1-18:1 in the medium of HEK/ABCG1 cells than in the medium of HEK/ABCA1 cells would be dependable. The medium of HEK/ABCG1 cells also included significant amounts of lysoPC 14:0, lysoPC 16:0, and lysoPC 18:1 compared with that of host HEK293 cells (data not shown), but it is not clear whether these lysoPCs were secreted from cells or converted from PC in the medium. These results suggest that ABCG1 preferentially secretes SM compared with PC, whereas ABCA1 preferentially secretes PC compared with SM.
We further examined whether ABCG1 mediated the efflux of SM (Fig. 11). The relative radioactivity of SM against PC secreted from HEK/ABCG1 cells in the presence of BSA (1.3 6 0.42), apoA-I (1.1 6 0.30), or HDL (1.3 6 0.30) was significantly higher than that from HEK/ ABCA1 in the presence of apoA-I (0.14 6 0.045) or HDL (0.16 6 0.012). This shows that ABCG1 mediates the efflux of SM preferentially, whereas ABCA1 mediates the efflux of PC preferentially.

DISCUSSION
In this study, we established HEK293 cells stably expressing human ABCG1 and showed that ABCG1 localizes to the plasma membrane and mediates the efflux of cholesterol and phospholipids (preferentially SM). ABCG1 has been reported to be expressed as several splicing variants in the N-terminal region and within NBF. ABCG1 cDNA, cloned from a fetal liver cDNA library in this study, has the same N-terminal sequence as that originally reported by Chen et al. (6). This type of variant was identified in   placenta and a macrophage cDNA library (12). The exogenously expressed ABCG1 in HEK293 cells showed a similar mobility on SDS-PAGE to the endogenous ABCG1 in THP-1 cells (Fig. 1), suggesting that the variant used in this study could be a major form in macrophages.
N-linked oligosaccharides are involved in the quality control of proteins by the calnexin and/or calreticulin cycle (37), and the expression of several ABC proteins has been reported to be regulated by the endoplasmic reticulum lectin (38). The isolated ABCG1 cDNA does not contain any putative N-linked glycosylation sites, as reported previously in human (6) and mouse (25,26) ABCG1, unlike that reported by Chen et al. (6). The experiment with N-glycosidases clearly showed that the endogenous ABCG1 in THP-1 cells is not modified by N-linked oligosaccharides (Fig. 1). Therefore, the major form of ABCG1 in vivo may not contain an N-linked glycosylation site and may function without N-linked oligosaccharides. It was reported that functional NBD is necessary for the trafficking of ABCG1 to the plasma membrane, because the G121A mutation within the WalkerA motif of ABCG1 abolished the cell surface expression (13). However, ABCG1-K120M localized to the plasma membrane (Fig. 2), suggesting that ATP binding and/or hydrolysis is not essential for the surface expression.
Recently, it was suggested that ABCG2 forms a multimer (27). However, cross-linking experiments show that most of the ABCG1 works as a homodimer of 60-65 kDa in an unglycosylated form, unlike ABCG2. Western blot analysis detected several bands (120-150 kDa) after cross-linking (Fig. 5). Vaughan and Oram (13) also detected similar bands by cross-linking experiment with disuccinimidyl suberate. Because we could not detect any endogenous ABCG subfamily protein (ABCG1, ABCG2, ABCG4, ABCG5, or ABCG8) in HEK293 cells by Western blot (data not shown), it is unlikely that ABCG1 forms heterodimers with endogenous ABCG subfamily proteins. ABCG1 may interact with other endogenous proteins. It was reported that ABCG1 is expressed as a protein of z95 kDa in perinuclear structures within cholesterol-laden macrophages and is present in foamy macrophages within atherosclerotic plaques (17). It was also reported that ABCG1 is expressed on the cell surface and in intracellular compartments of cholesterol-laden macrophages as a 110 kDa protein (14). It is not clear whether the type of splicing variant and/or N-glycosyl modification accounts for the differences in the subcellular localization and molecular size of ABCG1 expressed in cholesterol-laden macrophages.
HEK/ABCG1 cells secrete cholesterol and choline phospholipids as efficiently as HEK/ABCA1 cells in the presence of apoA-I (Fig. 8). Because the lipid efflux by ABCG1 is impaired by the amino acid substitution of lysine in NBF, as reported for ABCA1 (39), the efflux is dependent on the function of ABCG1 operated by ATP binding and/or hydrolysis. Notably, ABCG1 mediates cholesterol and phospholipid efflux even in the absence of apoA-I, whereas lipid efflux by ABCA1 is dependent on the presence of apoA-I. Lipid efflux mediated by ABCG1 was apparently dependent on the presence of BSA (Fig. 9), suggesting that serum albumin can function as an acceptor for cholesterol and phospholipids transported by ABCG1. In the presence of 0.002% BSA, ABCG1 mediated the efflux of 0.94 mg/ml cholesterol, which suggests that the molar ratio of cholesterol and BSA is 7:1 in the medium. Phospholipids in the medium may enhance the binding of cholesterol to BSA. Alternatively, some proteins secreted from HEK293 cells may function as acceptors as well. However, even if that is the case, these molecules cannot serve as an acceptor for lipids secreted by ABCA1, suggesting that ABCG1 has broader acceptor specificity than ABCA1 has.
Recently, Cignarella et al. (40) reported cholesterol efflux from monocyte-derived macrophages in the absence of acceptors correlated with the induction of ABCG1 expression. It was also reported that ABCG1 mediates the efflux of cholesterol and phospholipids from cells to HDL-2 or HDL-3 (19). Because the expression level of ABCG1 in THP-1 cells induced by the LXR ligand was comparable to that in HEK/ABCG1 cells, direct cholesterol efflux by ABCG1 from macrophages could be important within atherosclerotic lesions, in which the amounts of lipid acceptors are limited. Furthermore, in the presence of proper acceptors of secreted cholesterol in plasma, such as HDL-2 and HDL-3, direct efflux by ABCG1 would play an important function in cholesterol and phospholipid secretion.
The mass analysis revealed that ABCG1 secretes SM as well as PC into medium from HEK/ABCG1 cells. The main constituent of the plasma membrane is PC, and it was reported that, when incubated with fibroblasts, lipidfree apoA-I produces lipoproteins containing 69% PC and 18% SM (36). Rough total peak heights of PCs in Fig. 10C can be calculated to be several fold that of SMs. The major differences between phospholipids of HEK/ABCG1 and HEK/ABCA1 cells are the peak heights of SM 16:0-18:1 and SM 24:1-18:1. Although ion peaks from a triple quadrupole mass spectrometer do not always allow for quantitation and ion intensities of the signals are to be corrected (41), there was a linear relationship between the amount of SM and the peak height. Furthermore, the ratio of SM to PC effluxed by ABCG1 was eight times higher than that effluxed by ABCA1 (Fig. 11). From these results, it is assumed that ABCG1 preferentially secretes SM into the medium from HEK/ABCG1 cells. SM is synthesized in the lumen of the Golgi apparatus (42) and moved to the outer leaflet of the plasma membrane by vesicular membrane transport (43). SM has high affinity for cholesterol and tends to form a complex with cholesterol in the outer leaflet of the plasma membrane. ABCG1 may secrete cholesterol together with SM into the medium from the outer leaflet of the plasma membrane. It is known that the LolCDE system catalyzes the release of lipoproteins anchored to the outer leaflet of the inner membrane through the N-terminal fatty acyl chains in Escherichia coli (44).
In this study, we showed that the amounts of cholesterol and phospholipids in the medium of HEK/ABCG1 cells after 48 h of incubation in the presence or absence of apoA-I are similar to those of HEK/ABCA1 cells in the presence of apoA-I, as shown in Fig. 8, whereas the efficiency of cholesterol and phospholipid efflux at 4 h was not so high, as shown in Fig. 7. In a previous study, cells expressing ABCG1 showed an efflux of cholesterol to HDL-2 and HDL-3 but not to lipid-poor apoA-I (13,19). The cause of this discrepancy is not clear, but one possible explanation is that the amounts of free cholesterol and choline phospholipids in the medium were measured by colorimetric enzyme assays in Fig. 8, whereas cells were labeled with [ 3 H]cholesterol or [ 3 H]choline and radioactivity was measured in Fig. 7 and in previous studies. Because it was reported that the labeling of intracellular cholesterol pools varies with the methods used to deliver the labeled cholesterol or its precursors (29), ABCG1 and ABCA1 might mediate the efflux of cholesterol from different intracellular pools. Another possibility may be that cells stably expressing ABCG1, at levels comparable to cells derived from human macrophages, were used in this study, whereas cells transiently expressing ABCG1 were used in the previous study. Although further studies are necessary to understand the precise mechanism of lipid efflux by ABCG1, to our knowledge, this is the first report of a protein that mediates SM efflux.
In summary, we have demonstrated that ABCG1 localizes in the plasma membrane and mediates the efflux of cholesterol and phospholipids, especially SM. ABCA1 and ABCG1 expression is coordinately induced in macrophages in the LXR pathway, and ABCA1 mediates the efflux of cholesterol and phospholipids to apolipoproteins to form preb-HDL. In macrophages, ABCG1 may be involved in the removal of excess lipids by mediating the efflux of cholesterol and phospholipids, especially SM.