Sterol-dependent Transactivation of the ABC1 Promoter by the Liver X Receptor/Retinoid X Receptor*

Tangier disease, a condition characterized by low levels of high density lipoprotein and cholesterol accumulation in macrophages, is caused by mutations in the ATP-binding cassette transporter ABC1. In cultured macrophages, ABC1 mRNA was induced in an additive fashion by 22( R )-hydroxycholesterol and 9- cis -retinoic acid (9CRA), suggesting induction by nuclear hormone receptors of the liver X receptor (LXR) and retinoid X receptor (RXR) family. We cloned the 5 * -end of the human ABC1 transcript from cholesterol-loaded THP1 macrophages. When transfected into RAW macrophages, the upstream promoter was induced 7-fold by 22( R )-hydroxycholesterol, 8-fold by 9CRA, and 37-fold by 9CRA and 22( R )-hydroxycholesterol. Furthermore, promoter activity was increased in a sterol-responsive fashion when cotransfected with LXR a /RXR or LXR b /RXR. Further experiments identified a direct repeat spaced by four nucleotides (from 2 70 to 2 55 base pairs) as a binding site for LXR a /RXR or LXR b /RXR. Mutations in this element abolished the sterol-mediated activation of the promoter. The results show sterol-dependent transactivation of the ABC1 promoter by LXR/RXR and suggest that small molecule agonists of LXR could be useful drugs to reverse foam cell formation and atherogenesis.

Plasma HDL 1 -cholesterol levels are inversely related to the incidence of coronary artery disease (1). Two genetic diseases illustrate this phenomenon, the rare Tangier disease and the more common familial HDL deficiency. Tangier disease is characterized by an extremely low concentration of circulating HDL and the accumulation of cholesteryl esters in tonsils, liver, spleen, and intestinal mucosa, mostly in macrophage foam cells (2). Patients with familial HDL deficiency exhibit a low concentration of HDL particles and an increased risk of coronary artery disease (3). A common explanation for the cardioprotective effect of HDL is the major role it plays in reverse cholesterol transport (4). It is commonly accepted that the efflux of cholesterol from cells is caused by two different pathways: the first is passive and promotes efflux from the cell membrane to HDL and the second is energy-dependent and apolipoproteinmediated (5). The latter was characterized in fibroblasts and macrophages and involves lipid-poor or -free apolipoproteins such as apoA-I, apoA-II, and apo-E (5)(6)(7). This active pathway has been reported to be defective in both Tangier disease and familial HDL deficiency (8 -10). It was recently demonstrated that ABC1 is a key gene in this process (11) and that mutations of ABC1 are the major cause of both Tangier disease and familial HDL deficiency (3,(12)(13)(14)(15)(16)(17). ABC1 (ABCA1) belongs to the large ATP-binding cassette transporter family. These transmembrane proteins transport many diverse substrates across membranes because of their channel-like topology (18,19). The human ABC1 gene was assigned to chromosome 9q31, spanning a minimum of 70 kilobases and containing at least 49 exons (14,16,20). Whereas its expression is ubiquitous, the highest levels of human or murine mRNAs were found in placenta, fetal tissues, liver, lung, and adrenal glands (21,22). The predicted human protein contains 2201 amino acids (220-kDa protein) (21).
The expression of hABC1 is induced by cholesterol loading of human macrophages. Both the protein and the mRNA are up-regulated in the presence of acetylated LDL, and downregulated by cholesterol unloading via HDL3 (21). Whereas the cholesterol-mediated regulation of genes involved in cholesterol uptake or biosynthesis via sterol regulatory element binding protein (SREBP) pathways is well understood (23), much less is known about direct mechanisms of sterol-mediated up-regulation of gene expression. Two families of nuclear receptors are known to be activated by oxysterols and to mediate a positive response by binding to specific DNA elements, the liver X receptor (LXR) and steroidogenic factor 1 (SF1) (24 -27). SF1 acts as a monomer and has been implicated in the regulation of steroidogenic acute regulatory protein gene expression (StAR) activity (26). Recently, two different genes involved in the reverse cholesterol transport pathways, cholesterol 7␣-hydoxylase (24) and cholesterol ester transfer protein (CETP) (25), have been shown to be up-regulated by the heterodimer LXR⅐RXR. This suggests the hypothesis that LXRs might coordinate different steps of reverse cholesterol transport (25). LXR␣ (NR1H3) and LXR␤ (NR1H2) heterodimerize with their partner RXR. The resulting complex up-regulates genes through binding sites typically composed of direct repeats (DR) of the motif AGGTCA, spaced by 4 nucleotides (LXR␣ and LXR␤) or 1 nucleotide (LXR␤) (28 -30). The dimer can be activated by both the ligands of RXR (retinoids) and LXR (oxysterols) separately or together (29,31,32).
Here we report the sequence of the hABC1 promoter and show that this promoter is active in macrophages and that its sterol-mediated activation depends on the binding of LXR/ RXR␣ to a DR4 element. * This work was supported by National Institutes of Health Grants HL54591 and HL56984. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Cloning of the hABC1 Promoter and Introns 1 and 2-The screening of the human RPC.11 BAC clone library was performed (Research Genetics, Inc., Huntsville, AL) with a 68-mer oligonucleotide probe corresponding to nucleotides 11-79 of the published hABC1 sequence. Two BAC clones were recovered that were positive by PCR for exon 1 (BAC553F19) and exon 3 (BAC 522C12). After digestion by PstI, a Southern blot was performed using the 32 P-radiolabeled probes generated by PCR with the previously cited exons. Positive bands were cloned in pBluescript KS(ϩ) (Stratagene, La Jolla, California). A colony hybridization (probes used for Southern blot) (33) allowed us to isolate positive clones for the hABC1 promoter (5 kilobases) and intron 2. Sequencing performed on both strands showed that we also cloned intron 2 from BAC 522C12. The sequences of these introns are contained in the sequence of human genomic clone RP11-1M10, which also contains exons 1, 2, and 3 (see Fig. 2a).
Cell Cultures and Transfection Experiments-The cell lines were purchased from ATCC (Manassas, VA). The murine RAW 264.7, African green monkey CV-1, and human 293 or HepG2 cell lines were maintained in Dulbecco's modified Eagle's medium with 10% fetal bovine serum, 100 units/ml penicillin, and 100 mg/ml streptomycin. THP-1 cells were maintained in RPMI 1640 containing L-glutamine, 10% heatinactivated fetal bovine serum, 100 units/ml penicillin, and 100 g/ml streptomycin supplemented with 0.5 M ␤-mercaptoethanol. Confluent cells were differentiated with 0.2 M phorbol 12-myristate 13-acetate (Sigma) in ethanol over 72 h. Thioglycolate-elicited peritoneal macrophages were isolated from C57 Bl/6 mice as described previously (34). The luciferase activities were measured using the Promega dual luciferase assay system. A reporter plasmid used to analyze the activity of the hABC1 promoter was constructed by subcloning a 1029-bp PCR fragment of the hABC1 promoter (from Ϫ928 to ϩ101 bp) into the pGL3-Luc basic vector (Promega). A shorter promoter (from Ϫ469 to ϩ101 bp) was generated by digestion of this plasmid with SacI. Deletions (see Fig. 7) were performed by enzymatic digestion (ApaI, from Ϫ156 to ϩ101 bp) or PCR. The sequence of the PCR fragments were verified. Where shown, error bars represent S.D. Non-parametric Mann Whitney tests were performed to obtain p values.
Non-parametric Mann Whitney tests were performed to obtain p values.

RESULTS
Increased ABC1 mRNA in Human Macrophages Treated with Sterols and/or Retinoic Acid-To investigate whether the endogenous ABC1 gene can be activated by oxysterols and/or retinoic acid in macrophages, we performed Northern blot analysis of total RNA from human THP-1 macrophages. Fig. 1 shows a significant increase of ABC1 mRNA in cells treated with 22(R)-Hch (2-fold induction, p Ͻ 0.05) or 9CRA (2-fold, p Ͻ 0.05). An additive effect was obtained with combined treatment (4-fold, p Ͻ 0.05 when compared with separate treatments). These responses suggest possible activation of transcription by LXR/RXR (32).
Characterization of the 5Ј Region of the hABC1 Gene-To identify the promoter of the human ABC1 gene, we performed 5Ј-RACE PCR using poly(A) ϩ mRNA from cholesterol-loaded THP-1 macrophages and HepG2 cells (Fig. 2b). In macrophages this revealed a single major transcript (transcript A) consisting of a first exon of 217 bp followed by a second exon of 160 bp, 73% identical to mouse exon 1 (GenBank™/EBI accession number X75926). This exon is then followed by the published human exons 2, 3, and 4 (21).
In HepG2 cells, 5Ј-RACE PCR revealed three different transcripts (Fig. 2b). Transcript B represents a truncated version of exon 2 found in THP-1 cells (only the last 29 bp) followed by the published exons 2, 3, and 4 (21). Transcript C contains one exon of 372 bp upstream of the published exon 2, which is different from the exons found in THP1 cells. Transcript D has the same 5Ј structure as transcript C but lacks the published exon 3.
A BLAST search of the GenBank™/EBI Data Bank (htgs) revealed 100% homology of these exons (Fig. 2a, exons 1-3) with fragments of the human genomic clone RP11-1M10 (working draft sequence, GenBank™/EBI accession number AC012230). A comparison of the sequences from the published exon 2, the 5Ј-RACE PCR product, and RP11-1M10 revealed a C instead of a T at position ϩ15 and a G instead of an A at position ϩ17.
Conceptual translation of the transcripts revealed two new start codons in frame with the previously published ATG located in exon 5 (14) (Fig. 2, a and c). In the case of the transcript characteristic of THP1 cells, a new ATG located in exon 2 resulted in an extra 60 amino acids at the amino terminus. In the case of HepG2 cells, a new start codon at the 3Ј-end of exon 3 may be functional in transcript C and also transcript D, which lacks the previously published start codon. This results in an extra 39-amino acid fragment for transcript C.
A comparison of the putative amino-terminal amino acid sequences of ABC1 (transcripts A, B) with nucleotide data bases revealed strong homology to the amino-terminal sequences of two members of the ABC1 family (57% identity with ABCR and 45% identity with ABC3 (Fig. 1c). This strongly suggests that the amino-terminal sequence of hABC1 is authentic.
Sequence of the hABC1 Promoter-The promoter region upstream of exon 1 was responsive to sterols when transfected into cells (see below), whereas the 2.3-kilobase region upstream of transcript B was not responsive (data not shown). Thus, we focused our attention on the former region. Fig. 3 presents a partial sequence of genomic DNA with a fragment of exon 1 cloned from the human RPCI.11 BAC library. A potential TATA box is present at Ϫ32 bp and an Sp1 site at Ϫ101 bp. An analysis of this sequence revealed several potential transcription factor binding sites.
hABC1 Promoter Is Functional and Sterol-responsive in Macrophages-To investigate the function of the potential hABC1 promoter, we transfected the macrophage-like RAW 264.7 cell line with a promoter-luciferase construct (Fig. 4a). To test for activation by sterols we used 22(R)-Hch, a potent activator of LXR⅐RXR (36) but a poor activator of SF1 (27) and 9CRA, to activate endogenous RXR. Compared with basal conditions, transfected cells treated with 22(R)-Hch or 9CRA exhibited 7-and 8-fold higher promoter activity, respectively (p Ͻ 0.001) (Fig. 4a). When both compounds were added together, there was a synergistic 37-fold induction (p Ͻ 0.001). A similar response was obtained with promoter fragments containing 928 bp (Fig. 4a) or 469 bp (data not shown) of upstream sequence.
Next we compared the response of the ABC1 promoter to different sterols (Fig. 4b). We treated the transfected cells with 25-Hch, which is a good activator of SF1 and a weak activator of LXR (27). Cells were also treated with 7-Kch, which is relatively abundant in human arterial foam cells (37). 25-Hch is a poor inducer of the hABC1 promoter compared with 22(R)-Hch (1.5-fold activation, p Ͻ 0.05). No significant effect of 7-Kch was detected. However, when added in combination with 9CRA, a significant additive effect was detected for 7-Kch (2-fold when compared with 9CRA alone, p Ͻ 0.01). This pattern of sterol responsiveness is consistent with a transcriptional mechanism involving LXR.
LXR␣ and LXR␤ Are Both Expressed in Macrophages-To further investigate the potential role of LXR␣ and LXR␤ in the sterol-mediated up-regulation of hABC1, we verified that the mRNA of these nuclear receptors was present in RAW264.7 FIG. 2. Analysis of hABC1 5 sequence. a, partial gene structure; b, results of 5Ј-RACE PCR; and c, amino-terminal sequence. a, the hABC1 promoter was cloned from the human library RPCI-11, and the structure of the 5Ј-end of the gene was determined by sequence analysis. Positions of exons in RP11-M10 sequence are indicated in the inset. b, 5Ј-RACE PCR was performed on cholesterol-loaded THP-1 macrophages. These cells express exons 1 and 2 (transcript A). A 5Ј-RACE PCR was also conducted on HepG2 cells, which express a truncated version of exon 2 and also exon 3 (transcripts B, C, D). Two possible ATG start codons are present in exons 2 and 3. c, a comparison of the deduced amino-terminal sequence (NH 2 -terminal sequence) of hABC1 with the nucleotide data base (tBlastn) revealed similarities with two members of the ABC1 family, ABCR and ABC3. P, identity; Ⅲ, conserved substitution.  (Fig. 5a). We also analyzed their expression in vivo, using thioglycolate-elicited peritoneal macrophages from mice. Both receptors were detected in macrophages by Northern blot of total RNA, with a much stronger signal for LXR␤. We also selected two cell lines for our transactivation experiments, CV-1 and 293 cells. As shown in Fig. 5, both LXR␣ and LXR␤ could be detected in 293 cells but only LXR␤ in CV-1 cells. Because the quantities of mRNA for LXRs are relatively low in 293 cells and CV-1 cells, we chose to perform transactivation experiments in these cell lines.
LXR/RXR Transactivates the hABC1 Promoter in 293 and CV-1 Cells-To define the involvement of hRXR␣ and/or LXR␣/␤ in the sterol up-regulation of hABC1, we cotransfected 293 cells with the human ABC1 promoter and with these receptors (Fig. 5b). We used the shorter promoter (deletion Ϫ460 bp), which was sterol-responsive in macrophages (data not shown). In 293 cells without cotransfected LXR/hRXR we observed up-regulation of the promoter by 22(R)-Hch (4.5-fold, p Ͻ 0.05) and 9CRA (3.5-fold, p Ͻ 0.05) alone. The combination of 9CRA and 22(R)-Hch resulted in an additive effect (10-fold, p Ͻ 0.05). Basal activity was slightly increased (1.5-fold) when LXR␣/hRXR␣ were cotransfected, as was the sterol response, but there was no additional effect by 9CRA or 9CRA and 22(R)-Hch. However, cotransfection of LXR␤/RXR␣ caused a 5-fold increase in basal expression and a synergistic effect of 9CRA and 22(R)-Hch (19-fold induction compared with ethanol and 3-fold induction compared with 9CRA alone, p Ͻ 0.05).
As a positive control for these experiments we used a construct containing three copies of the LXR⅐hRXR␣ binding site of the CETP promoter (25) (Fig. 5c). Even in the absence of transfected receptors, this construct was highly sterol-and retinoic acid-responsive in 293 cells (14-and 24-fold increase in luciferase activity, respectively, p Ͻ 0.05), and an additive or synergistic effect (58-fold) was obtained with cotreatment. LXR transfection resulted in increased basal activity and increased induction by sterols. However, 9CRA provided no further increase in activity compared with non-transfected cells. These results suggest that endogenous LXRs in 293 cells play a role in the response of both ABC1 and CETP promoters with a further increase in sterol-dependent promoter activity when LXR and RXR are cotransfected.
Deletional Analysis of hABC1 Promoter-To define the region of the hABC1 promoter involved in its sterol-mediated activation we carried out further deletional analysis (Fig. 6). Similar results to Fig. 4a were obtained with fragments of the promoter from Ϫ160 bp or Ϫ100 bp to ϩ101 bp. Deletion of exon sequences (from ϩ3 bp to ϩ101 bp) reduced the basal activity, but the response to 22(R)-Hch and 9CRA was maintained. Notably, deletion of sequences between Ϫ101 and Ϫ36 bp reduced the basal activity and abolished the response to sterol and/or retinoic acid. Interestingly, the region covering Ϫ100 to ϩ36 bp contains an almost perfect DR4 element located on the non-coding strand between Ϫ70 and Ϫ55 bp (Fig. 6, inset).
Mutational Analysis of the DR4 Element-To test the hypothesis that the DR4 element is responsible for sterol activation, we introduced a mutated version of this element in the longer version of the hABC1 promoter (from Ϫ928 to ϩ101 bp). Both half-sites were mutated by changing nucleotides away from the nuclear hormone binding consensus sequence, as shown in Fig. 7. This mutation reduced the basal expression (2.5-fold), but the activity was still readily detectable. Importantly, the mutation abolished the response to 22(R)-Hch alone or in combination with 9CRA.
The DR4 Element Binds LXR/RXR-To determine whether the DR4 element binds LXR/RXR, we used oligonucleotides containing the wild-type DR4 or the mutated version (as in the functional assay, Fig. 7) to perform electrophoretic mobility shift assays using nuclear extracts from 293 cells cotransfected with LXR␤/hRXR␣ (Fig. 8, top) or LXR␣/hRXR␣ (Fig. 8, bottom). When the hABC1 wild-type DR4 element is used alone, a single major shift in activity is detected for both types of nuclear extracts (Fig. 8, lane 1), which disappears with excess of cold competitor indicating specificity (Fig. 8, lane 2). Because the mutated version is unable to compete with the intact oligonucleotide (Fig. 8, lane 3), the integrity of the DR4 itself is necessary for transcription factor binding. In cells cotransfected with LXR␤/RXR, antibodies recognizing the LXR␣ or LXR␤/␣ common region and anti-RXR antibodies markedly reduced binding activity and produced supershifted bands (Fig. 8, top). In cells cotransfected with LXR␣/RXR, antibodies to LXR␣, LXR ␤/␣, and RXR␣ showed similar effects except that the LXR␣ antibody did not produce a supershifted complex (25). Antibodies specific for ROR␣ had no (or a minimal) effect as expected. These results show that LXR␣/RXR␣ or LXR␤/ RXR␣ bind this DR4 element. DISCUSSION In this study we have identified a region of the human ABC1 promoter, which is active in macrophages and is induced by 22(R)-hydroxycholesterol and 9-cis-retinoic acid. Our characterization of the major transcript in cholesterol-loaded THP1 macrophages led to the identification of this promoter and also showed that most of the potential upstream initiation codons (ATG) in the previously published cDNA (21) are unlikely to represent the authentic translation initiation site. LXR␣/RXR␣ or LXR␤/RXR␣ binds to a DR4 element in the hABC1 promoter and mediates its activation by oxysterol and retinoic acid. Thus, LXR and/or RXR agonists could be useful drugs to reverse foam cell formation and atherogenesis.
A unique DR4 element mediates the sterol up-regulation of the hABC1 gene through LXR/RXR, and mutational analysis suggests that this is the only site involved. Unlike previously identified DR4 elements (25,36), the element in the ABC1 promoter is found in an inverse orientation on the non-coding strand (Fig. 6). The almost canonical sequence of this element might explain its high efficiency. Mutations in the DR4 result in a decrease of the basal expression of the promoter, but significant activity is still detectable. A weak activation of the mutated promoter by 9CRA (Fig. 7) suggests the presence of another binding site for RXR. However, the DR4 element we identified is responsible for the entire sterol-mediated activation of the 1-kilobase hABC1 promoter (Fig. 3).
The pattern of activation of the ABC1 promoter by sterols suggests that its expression may be suboptimal in atherosclerotic lesions. Thus, 7-KCh is relatively abundant in oxidized-LDL and in atheroma foam cells (37) and is a poor activator of the hABC1 promoter (Fig. 4b). 27-Hch, also abundant in foam cells, is a relatively poor activator of LXR (32,36). Thus, the accumulation of oxysterols in atherosclerotic lesions probably does not result in optimal activation of hABC1. This suggests that small molecules that are optimal LXR activators might be  Fig. 6) that was identified as a potential binding site for LXR/RXR or a mutated version (mut. DR4) were 32 P-radiolabeled. Competitors (lanes 2 and 3) correspond to cold wild-type or mutant oligonucleotides. After incubation with these oligonucleotides and nuclear extracts from 293 cells transfected with LXR␣/hRXR␣ (B) or LXR␤/hRXR␣ (A), some samples were incubated with polyclonal antibodies for LXR␣/␤, LXR␣, RXR␣, or ROR␣ (negative control) or with a monoclonal antibody targeting LXR␤. Three different experiments were performed with similar results. The arrow indicates the position of the retarded complex. effective drugs at reversing foam cell formation and that they might be useful as a treatment for atherosclerosis. The activation of the hABC1 promoter by 9CRA is increased 2-fold when given with 7-Kch (Fig. 4b). This further suggests that with regard to the induction of ABC1 by oxysterols, an unfavourable foam cell environment could also be switched to a more favorable one by delivery of ligands for RXR.
The inability of LXR␤ to compensate for the lack of LXR␣ in LXR␣ Ϫ/Ϫ mice (24) suggests that these receptors have different targets. In vitro, both LXR␣ and LXR␤ are able to up-regulate hABC1 (Fig. 5, b-d) or CETP (25), but LXR␤ is clearly more effective than LXR␣ in mediating the sterol response of hABC1 (Fig. 5b). This is also consistent with the fact that LXR␤ appears to be more highly expressed than LXR␣ in macrophages (Fig. 5a).
To conclude, we have shown that hABC1 is up-regulated at the transcriptional level by oxysterols and 9CRA acting though LXR/RXR which binds a proximal DR4 element located on the non-coding strand. These results provide strong support for the idea that LXRs act directly to coordinate the activities of molecules mediating reverse cholesterol transport (25). This may lead to a functional coordination of different steps of reverse cholesterol transport. For example, CETP activity results in the remodeling of HDL into small particles and liberates free apoA-I from HDL (38). Free apoA-I appears to be the optimal substrate for ABC1 (39,40). Thus, coordinate induction of CETP and ABC1 by LXR/RXR might act synergistically to enhance cholesterol efflux from macrophage foam cells.