Identification of Sterol-independent Regulatory Elements in the Human ATP-binding Cassette Transporter A1 Promoter

The ATP-binding cassette transporter A1 (ABCA1) shows a differentiation-, cAMP-, and sterol-dependent up-regulation in human monocytes. As part of an ongoing study, we investigated the proximal promoter regions that are highly conserved between the human and murine ABCA1 genes. Using reporter gene assays, we show here that a TATA box 24 bp upstream of the transcription initiation site is essential for promoter activity in RAW 264.7 and HepG2 cells, whereas further enhancement of transcriptional activity is mediated by the −175 bp promoter region. Gel shift assays revealed in vitro binding of Sp1 to a −91 GnC motif as well as binding of Sp1 and Sp3 to a −157 GnC promoter region. In co-transfection experiments using Drosophila S2 cells, we demonstrate that Sp3 competes with Sp1 for binding to the −157 GnC motif and acts as a repressor. On the other hand, overexpression of Sp1 increased ABCA1 mRNA expression in HeLa cells and enhanced cellular cholesterol and phospholipid efflux in RAW 246.7 macrophages. We also show here that the conserved E-box at position −140 binds upstream stimulatory factors 1 and 2 and hepatic nuclear factor 1α and that mutagenesis of the E-box enhanced constitutive ABCA1 expression in RAW 264.7 cells, implying a role for this element in silencing ABCA1 expression. Besides the functional importance for basal gene expression, we have identified that the core promoter region (−175 to +224) is also responsible for the induction of ABCA1 by the cytokine oncostatin M, resulting in a rapid increase in ABCA1 mRNA levels in HepG2 cells. Interestingly, this oncostatin M-induced expression is not dependent on the currently known sequence motifs in the ABCA1 promoter. In conclusion, a functional complex of cis-elements within the proximal human ABCA1 promoter associated with the transcription factors Sp1/3, upstream stimulatory factors 1 and 2, and hepatic nuclear factor 1α has been characterized, which allows a subtle tissue-specific regulation of ABCA1 gene expression.

The ATP-binding cassette transporter A1 (ABCA1) 1 was recently identified as a key regulator of cellular cholesterol efflux (1)(2)(3). Mutations of the ABCA1 gene are the causative defect in genetic HDL deficiency syndromes, and affected subjects have a defect in cellular cholesterol removal, which results in the almost complete absence of plasma HDL cholesterol (4). Cells lacking functional ABCA1 are characterized by structural and functional abnormalities including impaired raft/caveolar processing and Golgi-dependent lipid export processes due to impaired vesicular budding and excess lipid storage in the trans-Golgi network (5).
Expression of ABCA1 gene transcription is up-regulated in human monocytes during phagocytic differentiation, and its expression is further increased by loading with modified lipoproteins (6) or cAMP treatment (7,8). In addition, peroxisome-proliferator-activated receptor (PPAR) agonists and interferon-␥ modulate ABCA1 expression (9,10). Following earlier papers describing the ABCA1 promoter sequence (11,12) (for an update see Refs. 13 and 14), recent reports have mainly focused on the interaction between LXR/RXR and a corresponding DR4 element in the proximal ABCA1 promoter region (15,16). Interestingly, an alternative promoter in the first intron has been recently identified and shown to mediate liver-specific LXR/RXR-dependent ABCA1 expression in two different transgenic mouse models (17)(18)(19). LXR/RXR-responsive elements trigger retinoic acid and oxysterol-dependent activation of the ABCA1 promoter and thereby confer the observed induction of ABCA1 during lipid loading of macrophages. The most likely endogenous ligand for LXR␣ and LXR␤ is 27-hydroxycholesterol, since CYP27-deficient cells are not able to up-regulate ABCA1 in response to sterols and since overexpression of CYP27 activates LXR/RXR (20). Also, the earlier described LXR ligands 20(S)-hydroxycholesterol, 22(R)hydroxycholesterol, and 24(S),25-epoxycholesterol are not present in cholesterol-loaded macrophages, rendering them unlikely to be natural ligands of LXR (20).
Only little functional information is available on how constitutive and tissue-specific expression of ABCA1 is maintained. In a recent work from our group, the transcriptional repressor ZNF202 has been identified as a critical modulator of ABCA1 and ATP-binding cassette transporter G1 gene expression (21).
Following these studies and on the basis of a comparison between human and murine ABCA1 proximal promoter regions, we investigated in detail the functional properties of two conserved GnC motifs that bind the zinc finger proteins Sp1 and Sp3 within the core promoter sequence. We show that both sites are of functional relevance for constitutive ABCA1 gene expression in macrophage and liver cells. Furthermore, an E-box between the two GnC motifs acts as a silencer and binds the basic helix-loop-helix proteins upstream stimulatory factor 1 (USF1) and USF2 and the homeodomain-type protein hepatic nuclear factor 1␣ (HNF-1␣) in in vitro binding assays. Our data also for the first time describe the induction of the ABCA1 promoter by oncostatin M and localize the responsive element within the Ϫ175/ϩ229 bp core promoter region.

EXPERIMENTAL PROCEDURES
Reporter and Expression Plasmid Constructs-Primers for the amplification of the putative ABCA1 promoter sequence were based on a genomic sequence provided by the Whitehead Institute/MIT Center for Genome Research (Cambridge, MA) (accession number AC012230). The human BAC clone RP11-1M10, which contains the genomic ABCA1 5Ј region, was provided by the Resource Center of the German Human Genome Project (RZPD) and served as a template for the amplification of the promoter sequence with the High Fidelity PCR system (Roche Diagnostics).
Reporter constructs of the ABCA1 promoter sequence were cloned by ligation of PCR fragments into the BglII and NheI restriction sites of the pGL3-basic vector. Six inserts of increasing length were obtained by PCR. A promoterless pGL3-basic vector served as negative control, whereas the pGL3-control vector, which contains the SV40 early promoter, was used as positive control. In all experiments, cells were cotransfected with 1 g of the pSV ␤-galactosidase vector to normalize differences in transfection efficiency. Cells were harvested 24 h after transfection and lysed in reporter lysis buffer (Promega). Luciferase assay reagent containing luciferyl-CoA was added after centrifugation. Luciferase activity was determined in a LUMAT LB9501 (Berthold). The ␤-galactosidase enzyme assay (Promega) was used for the determination of ␤-galactosidase activity. Each experiment was repeated three times with two distinct plasmid preparations, and measurements were done in triplicate. The impact of transcription factor binding to consensus motifs within the core promoter region was assessed using promoter reporter constructs containing mutated binding sites or truncated constructs. Mutated inserts of the Ϫ157 GC-box and Ϫ140 E-box sites were generated using extended versions of the forward primer used for the amplification of the Ϫ175/ϩ224 reporter construct. Reporter constructs with mutated Ϫ91 GnC, Ϫ62 DR4, or Ϫ24 TATA motifs were generated following a two-step cloning strategy. Two PCR fragments were generated that overlap at the DNA motif of interest, replacing it by either a SacI (Ϫ91 GnC) or an EcoRI restriction site. In the case of the DR4 element, mutations introduced were identical to those reported by Costet et al. (15) using a SpeI restriction site in its vicinity. Overlapping fragments were ligated after digestion with the corresponding restriction enzyme and subsequently cloned into pGL3 basic vectors. The pPacSp1 and pPacSp3 expression vectors have been kindly provided by L. Lania (Naples, Italy).
Primer Extension Assay-Reverse primers A1-PX1R 5Ј-AAAACA-GAACCGGGGAAAAA-3Ј, A1-PX2R 5Ј-GAGAACCGGCTCTGTTGGT-3Ј, complementary to sequences at position ϩ151 and ϩ112, respectively, of the ABCA1 cDNA obtained by RACE-PCR were used for primer extension analysis. Oligonucleotides were end-labeled with [␥-32 P]ATP by T4 polynucleotide kinase (Invitrogen) and subsequently purified with Sephadex G25 columns (Roche Diagnostics). Fibroblast total RNA (10 g) was annealed with an excess of end-labeled primers at 58°C for 1 h according to the instructions of the manufacturer using the Promega primer extension kit. Reverse transcription reaction was performed for 30 min at 42°C with avian myeloblastosis virus reverse transcriptase. Samples were analyzed on a 8% polyacrylamide gel containing 8 M urea and audioradiographed with Eastman Kodak Co. BioMax MR films at Ϫ80°C.
RNase Protection Assay-DNA fragments were obtained by PCR (High Fidelity PCR System; Roche Diagnostics) using A1-RP1 5Ј-CGT-GCTTTCTGCTGAGTGAC-3Ј as forward primer in combination with A1-PX1R and A1-PX2R (identical to those used for primer extension assays) as reverse primers within exon 1 of the ABCA1 gene. The resulting fragments were cloned into pCRII TOPO dual promoter vector for the generation of riboprobes by in vitro transcription and sequenced. T7 RNA polymerase-mediated in vitro transcription of the linearized construct was performed in the presence of [␣-32 P]CTP using the Ribo-probe® Combination System-SP6/T7 (Promega). Radiolabeled transcripts were treated with DNase I and purified with Sephadex G50 columns (Roche Diagnostics). Fibroblast total RNA (10 g) was hybridized overnight at 50°C with the riboprobes. Finally, the samples were digested by RNase A and RNase T1 using the RNase protection kit from Roche Diagnostics according to the recommendations of the manufacturer. The samples were analyzed on an 8% polyacrylamide gel containing 8 M urea and audioradiographed with Kodak BioMax MR films at Ϫ80°C. Further, the plasmid containing the insert generated with primers A1-RP1 and A1-PX1R was sequenced with [␣-35 S]ATP and primer A1-PX1R using the T7 Sequenase kit (Amersham Biosciences) according to the recommendations of the manufacturer and served as sequencing ladder.
Cell Lines, Transfections, and Oncostatin M (OM) Stimulation-All cell lines were obtained from the American Type Culture Collection (ATCC) (Manassas, VA). RAW 264.7 macrophages, HepG2, Chinese hamster ovary, and HeLa cells were cultured in Dulbecco's modified Eagle's medium (Biowhittaker) supplemented with 10% heat-inactivated fetal calf serum (Sigma). Cells were incubated in a 5% CO 2 atmosphere at 37°C. Drosophila S2 cells were grown at 23°C in Schneider medium (Invitrogen) supplemented with 10% heat-inactivated fetal calf serum and 1% glutamine. RAW 264.7 macrophages, HepG2, Chinese hamster ovary, and HeLa were transiently transfected co-transfected with pPacSp1 and pPacSp3 expression vectors and different ABCA1 promoter constructs using Fugene® reagent (Roche Diagnostics). 1,2-Dioleoyl-3-trimethylammonium propane transfection reagent (Roche Diagnostics) was used for the transfection of Drosophila S2 cells. All transfection protocols were carried out according to the recommendations of the manufacturers. For oncostatin M stimulation experiments, 24 h after transfection with ABCA1 promoter constructs increasing amounts (0, 0.1, 1, and 10 nM) of OM were added to the cells for different periods of time (2,4,8, and 24 h) before luciferase activity was determined.
Electrophoretic Mobility Shift Assays-An equivalent of 40,000 cpm of double-stranded oligonucleotide probe containing the desired promoter sequence was incubated either with 10 g of nuclear extract from RAW cells or 25 g of nuclear extract from Chinese hamster ovary cells transiently transfected with pPacSp1 and pPacSp3 as described previously (22) in a buffer containing 50 mM HEPES/HCl, pH 7.9, 6 mM MgCl 2 , 50 mM dithiothreitol, 100 g/ml bovine serum albumin, 0.01% Nonidet P-40, and 2 g of poly(dI-dC) (Amersham Biosciences) at room temperature for 20 min. Supershift analysis was carried out using 1 l of antisera (Santa Cruz Biotechnology, Inc., Santa Cruz, CA) against the indicated transcription factors. In competition experiments, nuclear extracts were preincubated with a 100-fold molar excess of competitor for 10 min prior to the addition of the radiolabeled probe. In addition, nuclear extracts were incubated with oligonucleotides containing mutated GC-box and E-box sites. A detailed description of the oligonucleotides used in EMSA analysis is shown in Fig. 5. DNA-protein complexes were resolved on a native 8% polyacrylamide gel and audioradiographed with Kodak BioMax MR films at Ϫ80°C.
Determination of ABCA1 mRNA Levels-ABCA1 mRNA expression was analyzed by a quantitative two-step RT-PCR assay established in our laboratory using the Light Cycler technology (23). Total RNA from HeLa cells was isolated with Trizol reagent (Sigma). RNA concentration, purity, and integrity were assessed using the Agilent 2100 bioanalyzer and the RNA 6000 LabChip R kit (Agilent Technologies). Firststrand cDNA synthesis was performed using avian myeloblastosis virus reverse transcriptase (Roche Diagnostics) plus 10ϫ reaction buffer, RNase inhibitor, deoxynucleotide mix (1 mM), random primer p(dN) 6 , MgCl 2 , and 1 g of total RNA. The following program was used: 25°C for 10 min, 42°C for 60 min, and 95°C for 5 min. The FastStart DNA Master Hybridization Probes Kit (Roche Diagnostics) was used with a primer pair specific for the 205-bp ABCA1 fragment comprising base pairs 1327-1532 of the coding sequence and two hybridization probes in order to detect PCR product accumulation. For quantification, a stand-ard curve was generated with external homologous RNA after cloning of this fragment into the pCRII TOPO dual promoter vector and in vitro transcription of the linearized construct by T7 RNA polymerase. Results were controlled for the expression of the housekeeping gene (porphobilinogene deaminase). Primers and hybridization probes for porphobilinogene deaminase were obtained from TIB Molbiol.
Efflux Experiments-RAW 264.7 cells were transiently transfected with increasing amounts of pPacSp1 or mock plasmid. Efflux assays were performed with minor modifications as recently described (5). RAW 264.7 cells were transiently transfected and cultured for 24 h. Subsequently, cells were radiolabeled with 1.5 Ci/ml [ 14 C]cholesterol and 10 Ci/ml [ 3 H]choline and loaded with 40 g/ml enzymatically modified LDL, which was prepared as described elsewhere (24). Cells were incubated for 24 h in six-well plates containing Dulbecco's modified Eagle's medium supplemented with 5% lipoprotein-deficient serum and 10 M 20(S)-OH-cholesterol, 10 M 9-cis-retinoic acid (Sigma). Cells were washed and chased for 17 h with either 100 g/ml HDL 3 protein or 10 g/ml purified ApoA-I (Sigma) or medium alone. Lipids were extracted as previously described (25). Radioactivity was determined by liquid scintillation counting. Lipid efflux is expressed as the ratio of counts in medium to total counts. Specific efflux was calculated by subtraction of efflux rates in the presence of ApoA-I or HDL 3 from those in the absence of lipid acceptors.

RESULTS AND DISCUSSION
Mapping of the ABCA1 Transcription Initiation Site-In order to precisely map the position of the 5Ј-end of exon 1, an RNase protection assay using genomic/cDNA hybrid clones encompassing 138 bp upstream of the sequence obtained by 5Ј-RACE-PCR (data not shown) and either 151 or 112 bp of exon 1 was employed. As depicted in Fig. 1A, a single protected 161or 122-bp fragment appeared, which was appropriate in size and consistent with the results obtained by 5Ј-RACE-PCR. This finding was further confirmed by primer extension analysis employing the same reverse primer (A1-PX1R) within exon 1 used for RNase protection assays and RNA obtained from human fibroblasts. As shown in Fig. 1B, a major transcript can be detected at a size corresponding to that obtained by an RNase protection assay. The exact position of the transcription initiation site, which also defines the ABCA1 promoter region, is indicated in Fig. 2.
Identification of the Promoter Sequence Required for Basal ABCA1 Expression-The genomic region upstream of the transcription initiation site of ABCA1 contains several putative elements that might be responsible for transcriptional regulation (Fig. 2). In addition to the recently described DR4 element mediating oxysterol-dependent activation of the promoter (15,16) and the GnT motif binding the transcriptional repressor ZNF202 (21), a canonical TATA box is located at position Ϫ24. Further upstream, two GnC motifs (position Ϫ91 (GGGGCGGGG) and position Ϫ157 (GGGGCGGGCCC)) and an E-box (position Ϫ140, CACGTG) show high homology to the corresponding region within the murine ABCA1 promoter. To investigate the location of the basal promoter apparatus required for the expression of ABCA1 in macrophages and liver cells, we performed transient transfection experiments with reporter constructs containing exon 1 and increasing portions of the ABCA1 promoter (Fig. 3). Vectors were transfected into RAW 264.7 and HepG2 cells. In both cell lines, the full-length promoter sequence resulted in a more than 100-fold induction of luciferase activity as compared with the promoterless construct. The construct containing only exon 1 (positions ϩ12/ ϩ224) displayed no significant promoter activity. In contrast, a 24-fold increase was observed by extending the 5Ј-end up to position Ϫ79, which includes the TATA box. Promoter activity increased to 40 -70-fold, employing the Ϫ175/ϩ224 construct that includes an additional 187 bp upstream of the transcription initiation site. Taken together, these data implicate that the minimal sequence required for ABCA1 promoter activity in RAW 246.7 and HepG2 cells is located in the region Ϫ175/ϩ224. The ABCA1 Core Promoter Is Regulated by a TATA box, Two GnC Sites, and an E-box-Due to the high homology to the murine ABCA1 promoter sequence, we investigated in detail the impact of the Ϫ24 TATA box, the Ϫ91 and Ϫ157 GnC motifs, and the Ϫ140 E-box. As shown in Fig. 4, the exchange of the TATA box by an EcoRI site almost completely abolished promoter activity of the Ϫ79/ϩ224 region, strongly suggesting a pivotal role of this element for basal promoter induction. Mutations introduced in the Ϫ91 GnC motif within the Ϫ175/ ϩ224 reporter construct diminished promoter activity more than 50% to a level comparable with the wild type Ϫ79/ϩ224 region. Considering the particular location of the Sp1 binding site relative to the transcription initiation site (Fig. 2) and other reports on the interaction of Sp1 and TATA-binding protein or TAF110 (26), Sp1 may play a critical role in the formation or recruitment of the transcription initiation complex at the core promoter. Thus, within the Ϫ175/ϩ224 core promoter region, the TATA box and the Ϫ91 GnC motif are pivotal elements for the basal transcriptional activation of ABCA1. Interestingly, mutations of the Ϫ157 GnC motif and the Ϫ140 E-box both (Fig. 4) increased promoter activity up to 2-fold, suggesting that these regions function as transcriptional repressors.
Transcription Factors Binding to the Proximal Ϫ175 bp ABCA1 Promoter-To identify nuclear proteins interacting with the GnC motifs and the E-box, EMSAs were performed using two probes spanning from Ϫ174 to Ϫ125 and from Ϫ114 to Ϫ65, respectively (Fig. 5A). The interaction of fragment Ϫ174/Ϫ125, comprising an Sp1 site and an E-box, with RAW nuclear extracts resulted in four specific protein-DNA complexes (Fig. 5B, lane 1). The specificity of binding was shown by competition with unlabeled oligonucleotide, which eliminates these complexes completely (lane 7). In order to identify the protein factors that form these complexes, we used EMSA supershift experiments with antibodies specific for transcription factors that are known to be able to bind to GC-and E-boxes. No supershifted complexes were observed with antibodies against AP-2, C/EBP␣, Arnt1, Max1, SREBP1, SREBP2, E2F1, c-Myc, RXR␣, LXR␣, HNF-3␣, or HNF-4␣ (data not shown). In contrast, antibodies specific for Sp1 and Sp3 resulted in supershifted complexes (Fig. 5, lanes 2 and 3). The slowest migrating complex was partially shifted by either antibody, suggesting that this band represents a doublet of unresolved Sp1 and Sp3 bands. The fastest migrating DNA-protein complex was completely shifted by anti-Sp3 antibody, thus showing binding of Sp3 alone. These results indicate that Sp1 and Sp3 are not co-bound on the same DNA molecule, since anti-Sp3 antibody did not affect the formation of the Sp1specific complexes and anti-Sp1 antibody did not supershift the Sp3-specific complexes. This suggests that Sp1 and Sp3 compete for the binding to the region Ϫ174/Ϫ125 of the ABC1 promoter. Using antibodies specific for HNF-1␣, USF1, and USF2 (lanes 4 -6), we could also show the abolition of the intermediate band. However, only USF1 and USF2 antisera produce a supershift, whereas anti-HNF-1 antibody seems to inhibit the formation of the complex rather than produce a supershift. Since both USF1 and USF2 antibodies fully displaced the complex (lanes 5 and 6), we conclude that USF1/ USF2 heterodimers are largely predominant, whereas the amount of USF1 and USF2 homodimers is very low, and they are only visible after long exposure of the autoradiograph. In order to map the binding area in greater detail and to show whether Sp1, Sp3, and USF1/2 bind to the Sp1 site and E-box, respectively, mutations were created with nucleotide substitutions as shown in Fig. 5A (upper line). All specific complexes were displaced when using cold wild type competitor (Ϫ174/ Ϫ125) (lane 7). However, mutant M1 retains only the ability to displace the E-box complex (lane 8), whereas M2 competes for binding to the GnC motif (lane 9). The double mutant M1/2 allows no competition for nuclear protein binding. These data suggest that the specific binding of Sp1, Sp3, and USF1/2 to region Ϫ175/Ϫ125 critically involves the GnC motif at position Ϫ157 and the E-box at position Ϫ140. To determine whether the proximal promoter sequence, including the Sp1 site at position Ϫ91 is bound by nuclear factors, we performed additional EMSAs using a probe spanning from Ϫ114 to Ϫ65 (Fig.  5A, lanes 11-15). Incubation with nuclear extracts from RAW cells resulted in the formation of three specific complexes (lane 11). According to the consensus binding sites for Sp1 and C/EBP␣, we performed supershift analysis using antibodies specific for C/EBP␣ or Sp3. This did not result in a supershifted complex or complex reduction (data not shown and lane 13), whereas the addition of anti-Sp1 polyclonal serum diminished the lowest migrating band and produced a supershifted complex. To confirm the binding of Sp1 to its consensus site, we created the mutated fragment M3 (Fig. 5A, lower line) and used it for competition experiments together with the wild type (Ϫ114/Ϫ65) probe in EMSAs. In contrast to the cold wild type oligonucleotide, which has the ability to displace all specific complexes (lane 14), mutant M3 does not compete for binding of Sp1 (lane 15). These data show that the GnC motif in the proximal promoter fragment (Ϫ114/Ϫ65) interacts with Sp1 but not with Sp3.
Sp1-mediated ABCA1 Promoter Activation Can Be Repressed by Sp3 via the Ϫ157 GnC Motif-We further elucidated the role of Sp1 and Sp3 on ABCA1 transcription by employing Drosophila S2 cells that lack endogenous Sp factors. Using the (Ϫ175/ ϩ224) reporter vector, we observed a dose-dependent increase of promoter activity after transfection with increasing amounts of pPacSp1 (Fig. 6A, left panel), while no significant effect appeared with pPacSp3 alone (middle panel). Since there was evidence from our EMSA analysis (Fig. 5) and luciferase assays (Fig. 4) that Sp3 competes with Sp1 for DNA binding and represses promoter activity, we co-transfected Drosophila S2 cells with increasing amounts of pPacSp3 in the presence of constant pPacSp1 levels. Among these conditions, Sp3 was able to dose-dependently down-regulate promoter activity toward basal levels (Fig. 6A, right panel). As a next step, we investigated the role of both GnC motifs by co-transfecting Drosophila S2 cells with the (Ϫ175/ϩ224) reporter vector containing either a mutated Ϫ91 or a Ϫ157 GnC motif or with the truncated (Ϫ79/ϩ224) vector lacking both sites (Fig. 6B). Cells were transfected either with pPacSp1 (Fig. 6B, crossed bars) or an excess of pPacSp3 in the presence of pPacSp1 (gray bars) or mock (black bars). While no significant Sp1/3 effect was seen, using the Ϫ79/ϩ224 vector mutation of the Ϫ91 GnC motif altered neither the induction by Sp1 nor the Sp3-mediated repression. In contrast, mutation of the Ϫ157 GnC motif abolished Sp3-mediated repression, whereas induction by Sp1 was not influenced. These data implicate that (i) both GnC motifs are capable of inducing ABCA1 expression and (ii) the Ϫ157 GnC box mediates repression within the ABCA1 core promoter. In accordance with the results from electrophoretic mobility shift assays Sp1 and Sp3 seem to compete for DNA binding. Whereas Sp1 acts a potent activator, Sp3 binds with high affinity to the Ϫ157 GnC box without the ability to promote transcription. Thus, the Sp3-mediated repression seems to be the result of occupying the Sp1 binding domain, which has been reported as a common mechanism of how members of the Sp family modulate transcriptional regulation (27).
Cellular Lipid Efflux Rates Are Enhanced by Sp1-mediated Increase of ABCA1 mRNA Expression-In order to examine whether Sp1-mediated transactivation is functionally relevant, we first investigated ABCA1 mRNA expression in HeLa cells transiently transfected with increasing amounts of either pPacSp1. We employed a real time RT-PCR method based on the LightCycler technology using external standards in order to quantitatively and specifically determine ABCA1 mRNA levels. HeLa cells characteristically exhibit a low constitutive ABCA1 expression, which is in the range of 0.3 pg/g total RNA and about 200-fold less than in human macrophages (data not shown). As shown in Fig. 7A, we observed a dose-dependent up to 2.5-fold induction of ABCA1 mRNA levels. To confirm that these transcriptional mechanisms are of functional relevance for cellular lipid transport, we performed efflux experiments with RAW 264.7 cells transiently transfected with Sp1. Cells were radiolabeled with 1.5 Ci/ml [ 14 C]cholesterol and 10 Ci/ml [ 3 H]choline and loaded with 40 g/ml enzymatically modified LDL. Efflux was induced with either 100 g/ml HDL 3 protein or 10 g/ml purified apoA-I or medium alone. Since specific cellular lipid efflux rates in RAW 264.7 cells highly depend on the presence of retinoic acids, we performed these experiments among continuous stimulation with 9-retinoic acid and 20-OH cholesterol. Under these conditions, we observed a significant 2-fold enhancement of apoA-I-and HDL 3 -mediated specific cellular cholesterol and phospholipid efflux (Fig. 7B), which implies that the Sp1-dependent transcriptional mechanism is relevant in vivo. Also, Sp1 plays a critical role in the regulated expression of proteins and enzymes involved in lipoprotein metabolism during myeloid differentiation (22,28) and therefore may also contribute to the differentiation-dependent regulation of ABCA1 observed in human monocytes and macrophages (6). There is evidence that Sp1 mediates not only activity but also specificity and inducibility during differentiation (29,30). How Sp1 mediates this specificity is not clear, but in one case Sp1 binds to its myeloid target in vivo in myeloid cells only (29). Although Sp1 is ubiquitous, which fits to the broad tissue expression pattern of ABCA1 (6), it is preferentially expressed in hematopoietic cells (31). Sp1 is also capable of mediating responses to retinoic acid, thyroid hor- FIG. 4. E-box and GnC motifs modulate ABCA1 core promoter activity in RAW 264.7 cells. 1 ϫ 10 6 RAW 264.7 cells were transfected with 1 g of pSV␤-gal and 2 g of Ϫ175 ABCA1 pGL3 basic vector containing the wild type promoter sequence (Ϫ175/ϩ224) or either mutated or truncated sequences as indicated in the corresponding vector scheme and described under "Experimental Procedures." Transfected cells were cultured for 24 h before they were assayed. A representative experiment is shown, which was independently repeated three times. Luciferase activity was normalized for ␤-galactosidase activity and protein concentrations. Results are expressed as -fold value of promoterless pGL3-basic vector and given as mean Ϯ S.D. of triplicate measurements. mone, and retinoblastoma protein, all of which may influence myeloid differentiation.
Impact of GnC Motifs on Oxysterol-mediated ABCA1 Induction-To analyze whether the recently reported LXR/RXR-mediated induction of ABCA1 (15,16), which is transmitted by a nuclear hormone receptor pathway, also involves Sp1-mediated transcription, we used different mutated core promoter constructs lacking either the DR4 (LXR/RXR) element in different combinations with mutated up-or downstream GnC motifs (Fig. 8). A significant residual transcriptional induction of ABCA1 in the absence of the DR4 element, which was abolished by mutating the Ϫ91 GnC motif, provided evidence that in addition to the effect on macrophage differentiation, the oxysterol-dependent pathway activating ABCA1 transcription depends also on functional GnC motifs. Nevertheless, putative binding sites for other transcription factors involved in myeloid differentiation such as Evi-1 (32) or MZF-1 are present in the ABCA1 promoter region and may also be of importance.
Sterol-independent Regulation of the ABCA1 Promoter via OM-Following the results described above and based on recent reports proposing a critical role for macrophage-derived OM in linking the immune system with hepatic and macrophage lipid metabolism (33), we tested whether the cytokine oncostatin M has an impact on the regulation of the ABCA1 promoter in HepG2 cells. As shown in Fig. 9A, the transcriptional activity of ABCA1 increases steadily with time and in a dose-dependent manner following OM activation. Maximum induction was observed after 24 h with 10 nM OM (6 -7-fold), and a significant increase in promoter activity could already be detected after 4 h of 10 nM OM incubation (4-fold) (Fig. 9A). To further narrow down the upstream region involved in OMinduced ABCA1 activation, we performed the same type of promoter assays with various 5Ј deletion constructs (Fig. 9B). A significant induction by OM could be detected with all constructs down to Ϫ175 bp. Further shortening of the promoter to position ϩ12 resulted in a complete loss of basal activity as well as the stimulatory capacity of OM. This implies that the OMinducible region lies within the first Ϫ175 bp of the proximal ABCA1 promoter. To examine whether the above described GnC motifs and the E-box can mediate or repress this transcriptional induction, respectively, mutated luciferase constructs were analyzed in reporter gene assays (data not shown). As a result, neither mutagenesis of both GC-boxes at positions Ϫ157 and Ϫ91 nor exchange of the Ϫ140 E-box influences the OM responsiveness of the ABCA1 regulatory region. In conclusion, our data for the first time demonstrate a strong, sterolindependent induction of the ABCA1 gene in human HepG2 cells by the pleiotropic cytokine oncostatin M, which involves a Results are given as -fold induction as compared with cells co-transfected with 2 g of Ϫ175 ABCA1 pGL3 and 3 g of empty expression vector. B, 3 ϫ 10 6 S2 cells were transfected with 1 g of pSV␤-gal and 2 g of Ϫ175 ABCA1 pGL3 basic vector containing the wild type promoter sequence (Ϫ175/ϩ224) or either mutated GC boxes as described under "Experimental Procedures" or the Ϫ79 ABCA1 pGL3 basic vector. Mutated sites are indicated by an X in the corresponding vector scheme. Cells were co-transfected either with empty expression vector (black bars), 1 g of pPacSp1 (crossed bars), or 1 g of pPacSp1 and 2 g of pPacSp3 expression vector (gray bars) and incubated for 24 h. Results are given as -fold induction as compared with cells co-transfected with 2 g of Ϫ175 ABCA1 pGL3 and 3 g of empty expression vector. A representative experiment is shown, which was independently repeated three times. Luciferase activity was normalized for ␤-galactosidase activity and protein concentrations. Results are expressed as -fold value of Ϫ175 wild type ABCA1 pGL3-basic vector and given as mean Ϯ S.D. of triplicate measurements.
currently unknown sequence element located in the Ϫ175/ ϩ224 bp core promoter. In this respect, it is of special interest that the LDL receptor promoter is one of the transcriptional targets of OM in primary hepatocytes and HepG2 cells. This induction also occurs independent of intracellular cholesterol levels (33-36) and involves a C/EBP binding site and a cyclic AMP-responsive element (36). Since the proximal ABCA1 promoter region contains neither a C/EBP site nor a cyclic AMPresponsive element consensus sequence, we speculate that these factors are not involved in ABCA1 gene induction. Another report (37) describing OM-stimulated transcription of the human ␣2(I) collagen gene shows that the OM inducibility of the ␣2(I) collagen promoter is located within the core promoter and binds the transcription factors Sp1 and Sp3. Three copies of the 12-bp Sp element confer OM responsiveness to the heterologous thymidine kinase promoter, providing evidence that Sp1/Sp3 binding sites can mediate OM induction. However, although the basal promoter activity in RAW 246.7 and HepG2 cells is critically dependent on Sp1/Sp3 elements in the ABCA1 promoter, these binding sites do not confer OM induction. A likely explanation for this finding may be differences in the specific sequence composition surrounding the GnC motifs and the resulting combination of transcriptional modulators, which are quite different in the ␣2(I) collagen promoter.
Although not involved in OM induction, the Ϫ140 E-box (CACGTG) positioned between both GC-boxes in the core promoter region negatively regulates ABCA1 expression via basic helix-loop-helix transcription factors. We have identified binding of USF1, USF2, and HNF-1␣ by screening transcription factors that are known to bind to E-boxes. USF1 and USF2 are the major basic helix-loop-helix transcription factors in liver nuclear extracts (38). USF1/2 forms heterodimers on the ABCA1 promoter (Fig. 5), since both elements supershifted using either antibodies against USF1 or USF2. In vivo, USF1/2 FIG. 7. Sp1-mediated increase of ABCA1 mRNA expression causes increased cellular lipid efflux rates. A, 1 ϫ 10 7 HeLa cells were transfected with increasing amounts of pPacSp1 and inverse portions of empty expression vector (mock). Total RNA was isolated after 48 h and reverse transcribed. ABCA1 mRNA expression was quantified by LightCycler technology using an external ABCA1-specific standard as described under "Experimental Procedures." Results were normalized for porphobilinogen deaminase expression (PBGD) and shown as ABCA1/porphobilinogene deaminase ratio. B, 2 ϫ 10 5 RAW 264.7 cells were transiently transfected with 2 g of pPacSp1 (black bars) or empty expression vector (mock, gray bars). Cells were radiolabeled 24 h after transfection as described under "Experimental Procedures." Cells were incubated in six-well plates containing Dulbecco's modified Eagle's medium supplemented with 5% lipoprotein-deficient serum and 10 M 20(S)-OH-cholesterol, 10 M 9-cis-retinoic acid (Sigma) and loaded with 40 g/ml enzymatically modified LDL for 24 h. Subsequently, cells were washed and chased for 17 h either with 100 g/ml HDL 3 protein or 10 g/ml purified apoA-I or medium only. Results are given as mean Ϯ S.D. of specific efflux from measurements of four wells. *, p Ͻ 0.05 comparing the specific efflux of pPacSp1 and mock-transfected cells by t test for independent samples.
FIG. 8. The DR4-dependent oxysterol-mediated induction of ABCA1 is modulated by the ؊157 and ؊91 GC-box in HepG2 cells. HepG2 cells were transfected with 1 g of pSV␤-gal and 2 g of Ϫ175 ABCA1 pGL3 basic vector containing the wild type promoter sequence (Ϫ175/ϩ224) or either mutated GC box as described under "Experimental Procedures." Mutated sites are indicated by an X in the corresponding vector scheme. 4 h after transfection, cells were stimulated with 10 M 20-OH(S)-cholesterol and 10 M 9-cis-retinoic acid (9CRA) dissolved in 0.2% (v/v) ethanol. Cells were harvested after 24 h. A representative experiment is shown, which was independently repeated three times. Luciferase activity was normalized for ␤-galactosidase activity and protein concentrations. Results are expressed as -fold values of unstimulated reporter gene vectors and given as mean Ϯ S.D. of triplicate measurements.
FIG. 9. The ABCA1 core promoter contains an oncostatin Mresponsive element. A, HepG2 cells were transfected with 2 g of the Ϫ919/ϩ224 construct and 1 g of pSV␤-gal plasmid and cultured for 24 h. Thereafter, the cells were incubated with different concentrations (0.1, 1, and 10 nM) of oncostatin M for various time points (2,4,8, and 24 h) before measuring luciferase activity. A representative experiment is shown, which was independently repeated three times. Luciferase activity was normalized for ␤-galactosidase activity and protein concentrations. Results are expressed as -fold value of unstimulated construct and given as mean Ϯ S.D. of triplicate measurements. B, HepG2 cells were transfected with 2 g of the indicated promoter constructs and 1 g of pSV␤-gal plasmid and cultured for 24 h. Thereafter, the cells were incubated with 10 nM oncostatin M for 4 h before measuring luciferase activity. A representative experiment is shown, which was independently repeated three times. Luciferase activity was normalized for ␤-galactosidase activity and protein concentrations. Results are expressed as -fold value of promoterless pGL3-basic vector and given as mean Ϯ S.D. of triplicate measurements. heterodimers represent over 66% of the USF binding activity, whereas USF1 and USF2 homodimers represent less than 10%, which strongly suggests a preferential association of heterodimers in vivo (39). Interestingly, USF proteins, while being ubiquitously expressed, are involved in the expression of several tissue-specific or developmentally regulated genes (40). The HNF-4-regulated factor HNF-1␣ has been described as a homeodomain-type transcription factor, which plays an essential role during liver organogenesis by transactivating a large number of hepatic genes (41). Maturity onset diabetes of the young, type III (MODY3) is caused by mutation in the HNF-1␣ gene (42). HNF-1␣ is further involved in the transcription of several apolipoprotein genes and lipid transfer proteins, such as apoB (43), the microsomal triglyceride transfer protein (44), apo(a) (45), and apoIV (46). Interestingly, in healthy Canadian Oji-Cree subjects, the HNF-1␣ G319S genotype variant was significantly associated with higher plasma concentrations of high density lipoprotein cholesterol and apoA-I (47). Since apoA-I and apoA-IV play a pivotal role in the formation of HDL particles and regulation of HDL pool size, similarities can be expected in the transcriptional regulation of these proteins and ABCA1-mediated cholesterol efflux.