Protein Kinase A Site-specific Phosphorylation Regulates ATP-binding Cassette A1 (ABCA1)-mediated Phospholipid Efflux*

ATP-binding cassette A1 (ABCA1) is a key mediator of cholesterol and phospholipid efflux to apolipoprotein particles. We show that ABCA1 is a constitutively phosphorylated protein in both RAW macrophages and in a human embryonic kidney cell line expressing ABCA1. Furthermore, we demonstrate that phosphorylation of ABCA1 is mediated by protein kinase A (PKA) or a PKA-like kinasein vivo. Through site-directed mutagenesis studies of consensus PKA phosphorylation sites and in vitro PKA kinase assays, we show that Ser-1042 and Ser-2054, located in the nucleotide binding domains of ABCA1, are major phosphorylation sites for PKA. ApoA-I-dependent phospholipid efflux was decreased significantly by mutation of Ser-2054 alone and Ser-1042/Ser-2054 but was not significantly impaired with Ser-1042 alone. The mechanism by which ABCA1 phosphorylation affected ApoA-I-dependent phospholipid efflux did not involve either alterations in ApoA-I binding or changes in ABCA1 protein stability. These studies demonstrate a novel serine (Ser-2054) on the ABCA1 protein crucial for PKA phosphorylation and for regulation of ABCA1 transporter activity.

Tangier disease is caused by mutations in the ATP-binding cassette transporter, ABCA1 1 (1)(2)(3)(4)(5)(6). Expression of ABCA1 either by cAMP or LXR/RXR stimulation or by introducing ABCA1 cDNA has been shown to increase both cholesterol and phospholipid efflux to ApoA-I (7)(8)(9)(10)(11)(12). Lipid-poor ApoA-I has been shown to be important in promoting the efflux of both cholesterol and phospholipid from cells (13,14). Indirect evidence suggests that phospholipids may be the primary substrate of ABCA1 and that cholesterol is transported through spontaneous diffusion-mediated process after sufficient phospholipid has complexed with lipid-poor ApoA-I to create a good cholesterol sink (8,14). Therefore, the complex of ApoA-I and phospholipid is a much better acceptor of free cholesterol than ApoA-I itself (8,15).
There is evidence that phosphorylation may play an important role in ABCA1 function. cAMP has been shown to be important for induction of ABCA1 at a transcriptional level leading to an increase in protein synthesis (7)(8)(9)(10). Furthermore, the cAMP-dependent protein kinase A (PKA) is a key enzyme for numerous regulatory processes in almost all cell types, and ABCA1 possesses a number of basic consensus sites for phosphorylation by PKA (16). Moreover, a number of ABC transporters, including CFTR (17,18) and P-glycoprotein (17)(18)(19)(20)(21), are phosphorylated by protein kinase A or protein kinase C (17)(18)(19)(20)(21). Phosphorylation of CFTR by PKA is important for activation of chloride ion conductance (22,23). More recently, phosphorylation of P-glycoprotein by PKA and protein kinase C has been shown to modulate swelling-activated chloride current (24).
A number of kinases have been implicated in cholesterol efflux including mitogen-activated protein kinase and protein kinase C (25)(26)(27)(28). Protein kinase C inhibitors substantially reduce apolipoprotein-mediated cholesterol efflux in macrophages (26) whereas protein kinase C agonists such as 1,2dioctanoylglycerol and phorbol myristate acetate increase cholesterol efflux in fibroblasts (29). Cholesterol loading of cultured skin fibroblasts increases protein kinase C activity compared with untreated cells (25,27,28). Taken together, these data suggest that phosphorylation may have an important role in cholesterol efflux with ABCA1 being a potential substrate for kinases.
In this report, we provide evidence that ABCA1 is indeed phosphorylated by PKA or a PKA-like kinase in vivo, and we identify two serines phosphorylated by PKA in vitro that may potentially regulate ABCA1 function. Phosphorylation of Ser-2054 within the nucleotide binding domain 2 was found to be particularly important for ApoA-I-dependent phospholipid efflux activity mediated by ABCA1. Our data suggest that, like other ABC transporters, phosphorylation by PKA is important for modulation of ABCA1 function.

EXPERIMENTAL PROCEDURES
Cell Culture-Human embryonic kidney cells 293T, Flip-In 293 cells (Invitrogen), HeLa cells, and RAW mouse macrophages were maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum, 2 mM L-glutamine, and antibiotics.
Plasmids-Full-length human ABCA1 cDNA was isolated by reverse transcriptase PCR of mRNA obtained from cultured human skin fibroblast cells and cloned into pcDNA3.1 (Invitrogen). PCR-based mutagenesis was performed on the ABCA1 gene to create the S1042A and S2054A mutants as described previously (30). The mutagenesis primers used were as follows: 1042F, 5Ј-CAGAGAAAGCTAGCTGTGGCCTTG; 1042R, 5Ј-CAAGGCCACAGCTAGCTTTCTCTG; 2054F, 5Ј-ACAAACG-CAAGCTCGCTACAGCCATGGCT; 2054R, 5Ј-AGCCATGGCTGTAGC-GAGCTTGCGTTTGT. The S1042A and S2054A mutations in ABCA1 were completely sequence-confirmed. The S1042A mutation was isolated by XhoI/XbaI restriction digestion and moved into the S2054A clone, thereby generating a S1042A/S2054A double mutant that was also sequence-confirmed. All three of these mutants were then cloned into the pcDNA5/FRT vector (Invitrogen) and verified by complete sequencing of the insert. Enhanced green fluorescent protein (EGFP) was fused in-frame to the C terminus of ABCA1 to generate GFP-ABCA1. The EGFP-ABCA1 construct was generated in the plasmid pcDNA3.1 (Invitrogen) by overlapping PCR extension. Briefly, EGFP from pEGFP-C1 (Clontech) was amplified with the following PCR primers: F, 5Ј-GAAAGCTATGTAATGGTGAGCAAG; R, 5Ј-GCTCTAGAT-TACTTGTACAGCTC whereas ABCA1 was amplified with the following primers: F, 5Ј-CACCACAGGCATGGATCCCAAAG; and R, 5Ј-CTTGCT-CACCATTACATAGCTTTC under standard PCR conditions. The products were pooled, and PCR was performed with the two outermost primers. This ABCA1/GFP hybrid was cloned into the pcDNA3.1 ABCA1 vector. PCR-based mutagenesis was performed on the ABCA1 gene to create the S1042A and S2054A mutants as described previously (31). Mouse ABCA1 construct AG is a chimera consisting of a C-terminal fusion of EGFP and the 2261-amino acid full-length mouse ABCA1 transporter and has been described previously (32). The MM mutant contains a lysine 3 methionine substitution in the first and second NBD of mouse ABCA1 and has been shown to be functionally inactive (32,33).
Transfections-Human embryonic kidney cells, 293 Flip-in cells, or HeLa cells were transfected using Fugene-6 (Roche Molecular Biochemicals) as described by the manufacturer.
Generation of ABCA1 Stable Cell Lines-293 Flip-In cells were cultured in DMEM, 10% fetal calf serum, 1% glutamine, 1% penicillin/ streptomycin (Invitrogen) and 100 g/ml Zeocin (Invitrogen). An ABCA1 monoclonal stable cell line was generated by co-transfecting pcDNA5/FRT-ABCA1 and pOG44 (Invitrogen) using Fugene-6 (Roche Molecular Biochemicals) in growth medium without zeocin according to manufacturer recommendations. Twenty-four h after transfection, hygromycin B (Invitrogen) was added to the medium to a final concentration of 50 g per ml, and the media were changed every 3-4 days until hygromycin-resistant colonies were clearly evident. Individual colonies were picked and evaluated for ABCA1 expression by Western blot analysis. At least two independent monoclonal lines were used for all experiments, and a control hygromycin-resistant cell line was generated by co-transfecting the pOG44 with the empty pcDNA5/FRT vector. Polyclonal cell lines were also prepared by pooling at least 500 individual colonies from 293 Flip-in cells co-transfected with pOG44 and pcDNA5/FRT-ABCA1, pcDNA5/FRT-ABCA1 S1042A, pcDNA5/ FRT-ABCA1 S2054A, or pcDNA5/FTR-ABCA1 S1042A/S2054A cDNAs and selected for hygromycin resistance as described above.
Expression and Purification of Recombinant Proteins-GST-ABCA1 fusion proteins were induced in E. coli BL-21 DE3 (Stratagene) and purified as follows. 100 ml of LB containing 100 g per ml ampicillin was inoculated with 3 ml of overnight seed culture and grown at 37°C until the A 600 was between 0.6 to 0.8. Fusion protein expression was induced with 1 mM isopropyl-1-thio-␤-D-galactopyranoside (Sigma) for 1 h at room temperature. Cells were centrifuged at 4500 rpm for 15 min at 4°C and washed in STE buffer (10 mM Tris, pH 8.0, 150 mM NaCl, 1 mM EDTA). GST-ABCA1 fusion proteins were purified from the cell pellets by solubilization with 1.5% sarkosyl detergent followed by glutathione-Sepharose chromatography as described previously (34).
ABCA1 Antibodies-Polyclonal antibodies to ABCA1 were raised against a peptide directed against amino acid (aa) residues 2236 to 2259 of human ABCA1 and subsequently purified using an anti-peptide column as described previously (31). A monoclonal antibody to ABCA1 (AC10) that recognizes an epitope between aa 1873 and 2261 was also prepared as described (31).
Metabolic Labeling, Pulse-Chase, and Immunoprecipitation-RAW macrophages were equilibrated in phosphate-free or methionine-free DMEM in 1% delipidated bovine serum albumin for 2 h before the addition of [ 32 P]orthophosphate (0.7 mCi per ml) or [ 35 S]methionine (0.5 mCi per ml), respectively, in the presence of 4 g per ml 22-R-hydroxycholesterol and 10 M 9-cis-retinoic acid (Sigma). Cells were incubated at 37°C for 12 to 16 h. Cells were lysed by the addition of lysis buffer (20 mM HEPES, pH 7.5, 120 mM sodium chloride, 1% Triton X-100, 5 mM EDTA, 30 mM sodium fluoride, 40 mM ␤-glycerophosphate, 10 mM sodium pyrophosphate, 2 mM sodium orthovanadate supplemented with protease inhibitor mixture (Calbiochem). After 30 min on ice, insoluble material was removed by centrifugation at 14,000 ϫ g for 15 min at 4°C. Pre-cleared lysates were incubated with a rabbit anti-ABCA1 polyclonal antibody for 2 to 4 h at 4°C. Following the addition of 50 l of 50% (w/v) suspension of protein G-Sepharose (Amersham Biosciences) the tubes were rocked for an additional 40 min. Beads were washed four times with lysis buffer, SDS-sample buffer was added, and samples were incubated at 37°C for 10 min. Samples were run on 6% SDS-PAGE gels, dried, and autoradiographed. For detection of PKA phosphorylation of ABCA1 by an antibody specific for PKA-phosphorylated serines, immunoprecipitation using an anti-ABCA1 polyclonal antibody was carried out as described above on unlabeled 293 Flip-in monoclonal cell lines constitutively expressing ABCA1. In some experiments, 0.3 mM 8-Br-cAMP (Sigma) was added to the cells for 16 h at 37°C prior to harvesting. Proteins were transferred to nitrocellulose and probed with a phospho-PKA substrate antibody (Cell Signaling Technology, Inc.) to detect phosphorylation by PKA. Pulse-chase labeling experiments were performed on polyclonal 293 Flip-in cells constitutively expressing wild-type ABCA1, S1042A, or S2054A. Cells were starved in methionine-deficient media containing 5% dialyzed fetal bovine serum for 1 h and pulsed in the same medium containing 0.5 mCi per ml [ 35 S]methionine and [ 35 S]cysteine (Ͼ1000 Ci per mmol; Amersham Biosciences) for 5 min at 37°C. The chase was performed at 2-h time intervals up to 12 h using complete DMEM supplemented with 5% fetal bovine serum and 2 mM cold methionine and cysteine. Cells were solubilized in RIPA buffer, and ABCA1 was immunoprecipitated using anti-myc monoclonal antibody 9E10 (Roche Molecular Biochemicals) and protein G-Sepharose as described above. Immunoprecipitates were eluted in Laemmli sample buffer and analyzed by electrophoresis. The radioactivity associated with ABCA1 was quantitated by densitometry using a Molecular Imager FX system (Bio-Rad).
In Vitro Kinase Assays-About 1 g of purified GST-ABCA1 fusion protein immobilized on glutathione-Sepharose was suspended in 20 l of kinase buffer (20 mM MOPS, pH 7.2, 2 mM EDTA, 10 mM magnesium chloride, 1 mM dithiothreitol, 0.1% Triton X-100) and incubated with 5 units of protein kinase A catalytic subunit (Sigma), 10 Ci of [␥Ϫ 32 P]ATP, and 300 M cold ATP at 37°C for 30 min. The kinase reactions were terminated by the addition of 4ϫ Laemmli sample buffer. Proteins were resolved using 12% SDS-PAGE gel followed by autoradiography of the dried gel. Stoichiometric analysis was performed by excising GST-ABCA1 bands from the gel and counting the radioactivity in a liquid scintillation counter. For immunocomplex kinase assays, ABCA1 was immunoprecipitated as described above and immobilized on protein G-Sepharose. Beads were washed four times with lysis buffer, and kinase reactions were performed as described above.
ApoA-I Binding and Fluorescence-activated Cell Sorter Analysis-Recombinant ApoA-I (Calbiochem) was coupled to the Cy5 fluorochrome with the Fluorolink TM Cy5 monofunctional dye 5-pack (Amersham Biosciences) as described previously (32,33). HeLa cells were transfected with GFP-ABCA1 wild-type, S1042A, S2054A, or S1042A/S2054A as described above. Sixty h post-transfection, cells were bound to Cy5-ApoA-I, and fluorescence-activated cell sorter analysis was performed as described previously (32). Briefly, ApoA-I was conjugated to the fluorochrome with the Fluorolink TM Cy5 monofunctional dye 5-pack (PA25001 Amersham Biosciences). Cy5-ApoA-I was diluted to 100 g/ml in binding buffer (10 mM HEPES, pH 7.4, 1.8 mM CaCl 2 , 1 mM MgCl 2 , 5 mM KCl, 150 mM NaCl), and aggregates were removed by ultracentrifugation for 30 min at 100,000 ϫ g. Binding was performed in the presence of 10 g/ml of Cy5-ApoA-I for 1 h at 4°C on 5 ϫ 10 5 HeLa cells detached by mild trypsinization (0.005% in phosphate-buffered saline). At the end of incubation, cells were washed rapidly prior to fixation with 1% paraformaldehyde. Flow cytometric recordings were performed on a FACScalibur (BD Biosciences) and analyzed by Flowjow software (Tree Star Inc., San Carlos, CA). Transiently transfected cells with the individual ABCA1 cDNAs were gated manually for ABCA1 expression as reflected by EGFP relative fluorescence intensity. Binding data were calculated from the mean of Cy5-ApoA-I relative fluorescence intensity on the selected cell populations. The induced binding was calculated as the point to point difference between the EGFPpositive and -negative cells and was expressed as a percent of the binding induced by the wild-type ABCA1 construct. Statistical analysis was carried out using a paired Student's t test.
Phospholipid Efflux Assay-Efflux of phospholipid was performed as described previously (8,15,32) with the following modifications. For each individual experiment, polyclonal cells expressing the pcDNA5/FRT wild-type ABCA1, S1042A, S2054A, or S1042A/S2054A mutants were plated in triplicate at a density of 80,000 per well in 24 well dishes and labeled with 10 Ci per ml [ 3 H]choline (Amersham Biosciences) in 2 ml of DMEM with 1% fetal bovine serum for 24 h. Twenty-four h after labeling, cells were first washed with equilibration medium (DMEM containing 0.2% fatty acid-free bovine serum albumin) and then equilibrated in the same medium for 1 h. The media were then replaced with 0.5 ml of DMEM in the presence or absence of 10 g per ml ApoA-I and incubated at 37°C for 4 h, which was established to be within the linear range of the assay. Media was collected by spinning cells down at 14,000 ϫ g for 5 min followed by transfer of 0.4 ml into a glass test tube. Cell pellets were lysed in 250 l of 0.2% SDS by rocking for 10 min at room temperature and transferred to a second glass test tube. Remaining cellular debris was collected by washing the wells with 0.25 ml of water and pooling this with the 0.2% SDS lysates. Lipids were extracted by a 30-s vortex with a 1.5-ml chloroform:methanol mixture (1:2, v/v) followed by incubation for 1 h at room temperature. After incubation, 0.5 ml of water and 0.5 ml of chloroform was added, vortexed for 30 s, and centrifuged for 15 min at 3000 rpm at room temperature. The top phase was transferred to an Eppendorf tube, dried under vacuum, and resuspended in 0.2 ml of methanol by vortexing. Samples were then mixed with scintillation fluid and quantitated by liquid scintillation. Percent efflux was calculated by determining the ratio of counts released into the medium over the total counts. All phospholipid efflux data from independent experiments were normalized by densitometric analysis of ABCA1 protein by Western blot analysis. One-way ANOVA with a Neuman-Keuls post-test (Graphpad Prism) was used to detect significant differences among samples.
For fractionation of phospholipids by TLC, cells were labeled with 10 Ci per ml [ 3 H]choline and treated with ApoA-I and processed as described above. The chloroform/methanol extracts were dried under vacuum and suspended in 35 l of chloroform. Samples were then loaded onto Whatman LK5D silica plates and developed in a TLC chamber containing a chloroform:methanol:water mixture (65:25:4) (v/v). Labeled choline-containing phospholipids, identified by their co-migration with pure standards, were visualized by autoradiography and then quantitated by scanning densitometry.

RESULTS
ABCA1 Is a Phosphoprotein-Because protein phosphorylation has been shown to regulate the activity of ABC transporters including CFTR (17,18) and multidrug resistance 1 (MDR1) (19 -21), we determined whether ABCA1 is phosphorylated in vivo. To address this question RAW cells were treated for 16 h with 22-R-hydroxycholesterol and 9-cis-retinoic acid (LXR/RXR ligands) to induce ABCA1 expression and labeled with [ 35 S]methionine or [ 32 P]orthophosphate. Cells were then lysed and immunoprecipitated with either control serum or an anti-ABCA1-specific polyclonal antibody. As evident from Fig. 1A, a 35 S-labeled 220-kDa band was immunoprecipitated by the ABCA1 antibody but not by pre-immune serum (Fig. 1A,  lanes 1 and 2). Additionally, a 220-kDa 32 P-labeled protein was also immunoprecipitated by the anti-ABCA1 antibody but not by pre-immune serum, suggesting that ABCA1 is phosphorylated in RAW cells induced with LXR/RXR ligands (Fig. 1A,  lanes 3 and 4). To provide further evidence for ABCA1 phosphorylation in vivo, we 32 P-labeled a monoclonal 293 stable cell line constitutively expressing ABCA1. Fig. 1B shows that the antibody to ABCA1 recognizes a 220-kDa phosphorylated protein that was not immunoprecipitated by pre-immune serum (lanes 1 and 2). To our knowledge, this represents the first report that ABCA1 is phosphorylated in vivo.
PKA Phosphorylates ABCA1 in Vivo-The discovery that ABCA1 is phosphorylated in vivo led us to investigate the identity of the kinase responsible for the phosphorylation. Because PKA had been shown previously to phosphorylate ABCA1 in vitro (35), we wanted to determine whether PKA is responsible for phosphorylation of ABCA1 in vivo. We compared V8 proteolytic partial digestion profiles for both in vivo 32 P-labeled ABCA1 and ABCA1 phosphorylated in vitro by PKA. Fig. 2A shows that in vivo 32 P-labeled ABCA1 shares at least five common phosphoprotein fragments with that of PKAphosphorylated ABCA1 (compare lanes 2 and 3), suggesting that PKA is the kinase responsible for phosphorylating ABCA1 in both preparations. To demonstrate further that ABCA1 is phosphorylated in vivo by PKA, we used a phospho-PKA substrate antibody to detect phosphorylated ABCA1. This antibody is specific for peptides containing phospho-Thr/Ser with Arg at the Ϫ3 or Ϫ2 position and does not recognize non-PKAphosphorylated substrates. A monoclonal 293 Flip-in cell line constitutively expressing ABCA1 was used in these experiments to avoid the complications of increased transcription of the ABCA1 gene associated with stimulation of RAW macrophages with cAMP. The phospho-PKA substrate antibody recognized a 220-kDa protein immunoprecipitated by anti-ABCA1  lanes 1 and 3) or with rabbit anti-ABCA1 affinity-purified polyclonal antibody (lanes 2 and 4). Immunoprecipitates were resolved by SDS-PAGE, and the gel was dried and autoradiographed. B, ABCA1 is phosphorylated in a 293 Flip-in monoclonal cell line constitutively expressing ABCA1. Cells were labeled with [ 32 P]orthophosphate and ABCA1 was immunoprecipitated with either pre-immune serum (lane 1) or anti-ABCA1 polyclonal antibody (lane 2). Proteins were resolved by SDS-PAGE followed by autoradiography. polyclonal antibody but not by pre-immune serum in ABCA1 monoclonal cell line (Fig. 2B, left panel, compare lanes 1 and 2). In addition, the phosphorylation status of ABCA1 was not further enhanced by 8-Br-cAMP, a known activator of PKA (Fig. 2C, left panel, compare lanes 2 and 3). Taken together, the data suggest that ABCA1 is constitutively phosphorylated by PKA.
Identification of PKA Phosphorylation Sites on ABCA1-Upon examination of the ABCA1 amino acid sequence, we found 19 potential PKA phosphorylation sites based on PKA consensus sequences determined previously (16). Among the 19 potential phosphorylation sites, Ser-2054 contained a high probability consensus motif R(R/K)X(S/T) for PKA phosphorylation (Fig. 3). Interestingly, this PKA potential phosphorylation site within the Ser-2054 RKLS sequence is highly con-served between species, including human, mouse, chicken, and bovine ABCA1 and the Caenorhabditis elegans Ced-7 protein (Fig. 3), suggesting that these residues may be of functional significance. To test whether Ser-2054 is phosphorylated by PKA, we constructed GST-ABCA1 fusion proteins that contain this residue, as well as the corresponding alanine substitution mutant proteins. In vitro kinase assays were performed using the purified catalytic subunit of PKA and the various GST-ABCA1 fusion proteins as substrates. As shown in Fig. 4, PKA strongly phosphorylated GST-ABCA1 aa 1873 to 2261 (lane 2) but not GST alone (lane 1) or GST-ABCA1 1371 to 1650 (data not shown). Stoichiometric analysis revealed that GST-ABCA1 aa 1873 to 2261 incorporated 0.90 mol of phosphate per mol of protein after in vitro phosphorylation by PKA. The S2054A mutant reduced PKA phosphorylation of GST-ABCA1 aa 1873 to 2261 by 95% as determined by liquid scintillation counting of the excised radioactive bands (Fig. 4, left panel, compare lanes  2 and 3). This strongly suggests that Ser-2054 is a major determinant for in vitro PKA phosphorylation between aa 1873 and 2261.
The phosphorylation of Ser-2054 by PKA prompted us to examine other potential sites for this kinase. ClustalW alignment of amino acids of ABCA1 NBD1 and NBD2 regions showed 39% identity, suggesting that these domains may have arisen by duplication (Fig. 5). The most notable feature was the conservation of the RKLS sequence between NBD1 and NBD2 (Fig. 5). This suggests that Ser-1042 within the RKLS sequence of NBD1 may be another potential PKA phosphorylation site. GST-ABCA1 aa 1010 to 1171, which contains the Ser-1042 site, was then tested as a potential substrate for PKA. PKA was found to strongly phosphorylate this fusion protein with a stoichiometry of 0.70 mol of phosphate per mol of protein ( Fig.  6 left panel, lane 2). Moreover, GST-ABCA1 S1042A mutant reduced PKA phosphorylation by 80% (Fig. 6, left panel, lane 3), confirming that Ser-1042 is another PKA phosphorylation site on ABCA1.
Phosphorylation of ABCA1 Influences ApoA-I-dependent Phospholipid Efflux-We next wanted to evaluate the role of PKA phosphorylation on a known ABCA1 function, phospholipid efflux. As shown in Fig. 7A, ApoA-I-dependent phospholipid efflux was found to increase linearly (r 2 ϭ 0.93) over 10 h in a polyclonal Flip-in 293 cell line constitutively expressing wild-type ABCA1. Phosphatidylcholine was the only detectable choline-containing phospholipid whose efflux was stimulated by ApoA-I after 6 h of treatment (100% phosphatidylcholine, 0% sphingomyelin; see Fig. 7B). A 10-fold increase in phosphatidylcholine was observed in the efflux medium upon stimulation of cells with ApoA-I (Fig. 7B). Sphingomyelin or any phospholipid breakdown products were not detectable in the medium after 6 h with ApoA-I. However, sphingomyelin was observed after cells were stimulated for 24 h with ApoA-I, albeit at much lower levels than phosphatidylcholine (90% of phosphatidylcholine, 10% sphingomyelin) (data not shown). Polyclonal Flip-in 293 cells constitutively expressing wild-type ABCA1, S1042A, S2054A, or S1042A/S2054A mutants were then examined for their ability to efflux phospholipid to ApoA-I (8,13,14). Although the steady state levels of the phosphorylation mutants were comparable with that of wild-type ABCA1 (Fig 7C), all phospholipid efflux data were normalized according to ABCA1 protein expression. Overall, the total phospholipid efflux in wild-type ABCA1 cells was stimulated 4-fold or more after 4 h of treatment with ApoA-I (Fig. 7D). Compared with wild-type ABCA1, the S2054A mutant showed a 40% decrease (p Ͻ 0.01) in ApoA-I-dependent phospholipid efflux whereas a 50% reduction (p Ͻ 0.01) was observed for the S1042A/S2054A mutant (Fig. 7D). No significant difference in phospholipid efflux was observed between the S1042A mutant alone and wild-type ABCA1 cells or between the S2054A and S1042A/S2054A mutant cells (Fig. 7D). Comparison among the different mutants showed that of the two individual mutations, S2054A clearly had the greatest effect on phospholipid efflux. Additionally, efflux in cells expressing the combined mutant S1042A/S2054A was not significantly reduced compared with cells expressing S2054A alone, suggesting that Ser-2054 is the critical phosphorylated residue for phospholipid efflux (Fig.  7D). Taken together, the data suggest that PKA phosphorylation of ABCA1 Ser-2054 is important for ABCA1 physiological function and that defects in this process result in a decrease in ApoA-I-dependent phospholipid efflux.
Phosphorylation Does Not Affect ApoA-I Binding or ABCA1 Protein Stability-We next wanted to determine the mechanism by which PKA phosphorylation of ABCA1 Ser-2054 affects ApoA-I-dependent phospholipid efflux. Binding of ApoA-I has been shown previously to be dependent on ABCA1 expression (32,33). To determine whether ABCA1 phosphorylation affects ApoA-I binding, HeLa cells transiently expressing GFP-tagged wild-type ABCA1, S1042A, S2054A, or S1042A/S2054A were analyzed for Cy5-ApoA-I binding by dual-channel flow cytometric analysis. Whereas a non-functional mutant mouse ABCA1 defined previously (33) showed loss of ApoA-I binding compared with its wild-type counterpart ABCA1 (Fig. 8, p Ͻ 0.001; compare AG and MM), no significant difference in ApoA-I binding was observed between cells expressing wildtype human ABCA1 or any of the phosphorylation mutants (Fig. 8, compare WT, S1042A, S2054A, and S1042A/S2054A). This suggests that loss of PKA phosphorylation at either serine residue does not affect the ability of ApoA-I to bind to ABCA1expressing cells. Finally, as phosphorylation has been documented to affect protein stability (36 -38), pulse-chase experiments were performed on polyclonal Flp-in 293 human embryonic kidney cells constitutively expressing wild-type ABCA1 or the phosphorylation mutants. The decay constant for wild-type ABCA1 (about 3 h) was comparable with that found for either S1042A or S2054A alone (Table I). These results suggest that loss of phosphorylation at either site does not play a role in ABCA1 protein stability. Therefore, the data indicate that the reduction in ApoA-I-dependent phospholipid efflux in the S2054A mutant cells is not because of defects in ApoA-I binding or because of decreased ABCA1 protein stability. DISCUSSION Our results show that PKA is an important kinase that regulates ABCA1 functional activity. The ability of ABCA1 to mediate ApoA-I-dependent phospholipid efflux was modulated by PKA site-specific phosphorylation. That ABCA1 is phosphorylated in vivo is supported by immunoprecipitation experiments whereby 32 P-labeled ABCA1 was isolated from either RAW cells or from a stable cell line constitutively expressing ABCA1. Furthermore, direct evidence that PKA is the kinase responsible for in vivo phosphorylation of ABCA1 was provided by the immunoreactivity of the phospho-PKA substrate antibody with cellular ABCA1 and by the similarity in the V8 digestion profile of in vivo phosphorylated ABCA1 and ABCA1 phosphorylated in vitro by PKA. Through in vitro PKA kinase experiments using GST-ABCA1 fusion proteins as substrates, we have identified two PKA phosphorylation sites within ABCA1 (Ser-1042 and Ser-2054), both of which are located within the nucleotide binding domains 1 and 2, respectively. Ser-1042 and Ser-2054 are major PKA phosphoacceptors, because serine to alanine mutations significantly reduce PKA phosphorylation of ABCA1 in vitro. Alignments between the NBD1 and NBD2 show that both serines are highly conserved in ABCA1 across species (see Figs. 3 and 5) and between ABCA1 to ABCA8, indicating that these residues may have a functional importance (39).
The functional consequence of PKA phosphorylation of ABCA1 was examined by analyzing the effects of the phosphorylation mutants on phospholipid efflux. We focused on phospholipid efflux, because there is accumulating evidence that ABCA1 may be a phospholipid translocase and may not directly transport cholesterol (8,14,15). This is supported by experiments where ABCA1 has been shown to bind photoactive phospholipid but not photocholesterol and where ABCA1-conditioned medium in the absence of ABCA1 expression enhances cholesterol efflux (8,15). Our results show that PKA phosphorylation of ABCA1 Ser-2054 is important for maintaining normal phospholipid efflux function. The major choline-containing phospholipid released by wild-type ABCA1 to ApoA-I in the medium was phosphatidylcholine, in agreement with previous findings (15,40). How loss of phosphorylation at Ser-2054 results in decreased phospholipid efflux activity is not clear. Phosphorylation has been shown to modify the function of proteins in a number of different ways including increasing or decreasing biological activity, by stabilizing proteins or marking them for degradation, by facilitating or inhibiting movement between subcellular compartments, or by initiating or disrupting protein-protein interactions (41). In this study we have attempted to address the mechanism by which ABCA1 phosphorylation might influence ApoA-I-dependent phospholipid efflux. We have found that phosphorylation of ABCA1 Ser-2054 does not affect ApoA-I binding or ABCA1 protein half-life. Furthermore, confocal microscopic studies using GFPtagged ABCA1 constructs have revealed that subcellular localization is not altered by ABCA1 Ser-2054 phosphorylation (data not shown). Thus, it is likely that ABCA1 phosphorylation may affect ApoA-I-dependent phospholipid efflux by either altering the conformation of the protein to a more active state or by affecting the interaction between ABCA1 and its partner proteins. Interestingly, mutation at Ser-2054 has a greater effect on phospholipid efflux than at Ser-1042, suggesting that Ser-2054 is the more critical serine for PKA phosphorylation. Therefore, phosphorylation of Ser-2054 may be important for efflux of phospholipids to lipid-poor ApoA-I acceptor with a concomitant increase in cholesterol transfer.
Our results show that ABCA1 is constitutively phosphorylated by PKA. No increase in phosphorylation was observed upon treatment of cells with 8-bromo-cAMP to stimulate PKA. Consistent with the constitutive phosphorylation of ABCA1 is the finding that 8-bromo-cAMP is not required for ApoA-I-dependent phospholipid efflux mediated by the ABCA1-expressing polyclonal stable 293 HEK cell lines. This is in contrast to RAW or J774 mouse macrophage cells where endogenous ABCA1 expression is low, and there is a requirement for 8-bromo-cAMP to up-regulate ABCA1 mRNA and subsequently protein synthesis before cells are cholesterol or phospholipid efflux-competent (8 -10). The lack of increased phosphorylation in response to cAMP suggests that Ser-2054 may be saturated with phosphate and that further activation of PKA does not enhance phosphorylation at this residue.
Our studies indicate that in vivo phosphorylation of ABCA1 by PKA moderates its normal physiological function. These results are consistent with a previous study demonstrating that anion flux mediated by ABCA1 expressed in Xenopus oocytes can be stimulated by cAMP or inhibited by the PKA inhibitor, H89 (35). PKA has been shown to phosphorylate other ABC transporters, including CFTR (17, 18) (reviewed in Ref. 42) and the ATP-binding cassette drug transporter Pglycoprotein (43). Phosphorylation of CFTR has been shown to occur primarily in the regulatory domain and in the first nucleotide binding domain of CFTR (17,44,45). Studies of CFTR phosphorylation variants have shown that serines within the R domain contribute strongly to CFTR chloride channel activation (for review see Ref. 42). The mechanism by which PKA phosphorylation of CFTR leads to increased channel activity is not well understood. Although we have not identified individuals with Tangier Disease or familial hypoalphalipoproteinemia that have mutations at either of the two PKA phosphorylation sites (2,46), there are naturally occurring ABCA1 mutations located close to Ser-1042 and Ser-2054 (Table II). The phosphorylated Ser-1042 and Ser-2054 residues are near the ABCA1 mutations A1046D and R2081W, both of which have resulted in a clinical phenotype in humans with low high density lipoprotein cholesterol, low plasma ApoA-I, and low total cholesterol values (47)(48)(49). This raises the possibility that these ABCA1 mutations may potentially affect the ability of PKA to bind and phosphorylate ABCA1.
To summarize, we have identified PKA as a kinase that can phosphorylate ABCA1 at two serine residues. Loss of phosphorylation at ABCA1 Ser-2054 results in a decrease in ApoA-Idependent phospholipid efflux but does not affect degradation of the ABCA1 protein. Our studies also demonstrate that although these PKA phosphoresidues are highly conserved in NBD1 and NBD2, Ser-2054 is more important than Ser-1042 for influencing ABCA1-mediated phospholipid efflux. Our find- FIG. 7. Effect of phosphorylation on ApoA-I-dependent phospholipid efflux. A, phospholipid efflux on polyclonal cells constitutively expressing wild-type ABCA1 was determined as a function of time following the addition of ApoA-I. Data represent the mean Ϯ S.D. for quadruplicate determinations of percentage of total [ 3 H]choline counts (counts in the medium plus those in the monolayer) that were present in the medium after subtracting counts released non-specifically into the medium in the absence of ApoA-I. Correlation coefficient r 2 was determined by linear regression analysis. B, labeled cholinecontaining phospholipids transferred from wild-type ABCA1-expressing polyclonal cells to extracellular medium were extracted with chloroform and methanol and fractionated by TLC, together with lipid standards as described under "Experimental Procedures." Cells were treated with ApoA-I for 6 h prior to harvest of medium. After autoradiography, the level of each phospholipid was expressed in densitometric units. No other phospholipid other than phosphatidylcholine was detectable in the efflux medium after stimulation of cells with ApoA-I for 6 h. Results represent -fold increase in extracellular phosphatidylcholine relative to that observed in the absence of ApoA-I (normalized to 1). PC, phosphatidylcholine. Results are representative of three independent experiments. C, 293 Flip-in cells constitutively expressing wild-type ABCA1, S1042A, S2054A, or S1042A/S2054A proteins were lysed, and 200 g of protein was analyzed by SDS-PAGE followed by Western blotting with an anti-ABCA1 monoclonal antibody (AC10). glyceraldehyde-3-phosphate dehydrogenase expression is also shown to indicate equal protein loading. D, polyclonal cells constitutively expressing wild-type ABCA1, S1042A, S2054A, or S1042A/S2054A were examined for total phospholipid efflux for 4 h in the presence or absence of ApoA-I as described under "Experimental Procedures." Results represent mean Ϯ S.E. for 9 -15 independent experiments. Phospholipid efflux data were normalized for ABCA1 protein levels quantitated by densitometry as described under "Experimental Procedures." Asterisks indicate p Ͻ 0.01 compared with wild-type (WT) protein as determined by one-way ANOVA with a Neuman-Keuls post-test.

TABLE I
Half-lives of ABCA1 proteins Flip-in 293 cells constitutively expressing wild-type or phosphorylation mutants were pulsed for 5 min and then chased at different times as described under "Experimental Procedures." Cell lysates were then immunoprecipitated with anti-ABCA1 monoclonal antibody, and radiolabeled bands were resolved by SDS-PAGE and quantitated using a phosphorimager. Half-life values derived from first order kinetic degradation plots from three independent experiments are shown. ings show that kinases such as PKA can modulate ABCA1 function, and identification of these signaling molecules may provide leads to novel approaches for regulation of ABCA1 activity.