Human ABCA1 BAC transgenic mice show increased high density lipoprotein cholesterol and ApoAI-dependent efflux stimulated by an internal promoter containing liver X receptor response elements in intron 1.

By using BAC transgenic mice, we have shown that increased human ABCA1 protein expression results in a significant increase in cholesterol efflux in different tissues and marked elevation in high density lipoprotein (HDL)-cholesterol levels associated with increases in apoAI and apoAII. Three novel ABCA1 transcripts containing three different transcription initiation sites that utilize sequences in intron 1 have been identified. In BAC transgenic mice there is an increased expression of ABCA1 protein, but the distribution of the ABCA1 product in different cells remains similar to wild type mice. An internal promoter in human intron 1 containing liver X response elements is functional in vivo and directly contributes to regulation of the human ABCA1 gene in multiple tissues and to raised HDL cholesterol, apoAI, and apoAII levels. A highly significant relationship between raised protein levels, increased efflux, and level of HDL elevation is evident. These data provide proof of the principle that increased human ABCA1 efflux activity is associated with an increase in HDL levels in vivo.

A significant step in the elucidation of mechanisms of reverse cholesterol transport resulted from the identification of mutations in ABCA1 underlying Tangier disease, as well as familial hypoalphalipoproteinemia associated with reduced efflux (1)(2)(3)(4)(5). These and further investigations and characterizations of the biochemical phenotype of heterozygotes for ABCA1 deficiency (6) have demonstrated that lipidation of the nascent apoAI-rich HDL 1 particle is a rate-limiting step in the mainte-nance and regulation of HDL cholesterol (HDL-C) levels in humans. The ABCA1 gene is also rate-limiting for cholesterol efflux and HDL-C levels in different species, including mouse (7,8) and chicken (9), demonstrating conservation of this pathway in cholesterol metabolism over at least 400 million years.
Studies of heterozygotes for ABCA1 deficiency have also demonstrated a very strong relationship between levels of cellular cholesterol efflux and HDL-C levels in plasma, with ϳ82% of the variation in HDL-C levels in these patients being accounted for by the decrease in cellular cholesterol efflux. This clearly has demonstrated in these patients that ABCA1 is the major but not the only contributor to cellular cholesterol efflux in humans (6).
There have been recent significant additional advances with regard to understanding regulation of ABCA1 expression. A direct mechanism of sterol-mediated up-regulation of gene expression of ABCA1 has been shown to be due to transactivation of the ABCA1 promoter by LXR and RXR (10 -12), two members of the nuclear receptor superfamily. This sterol-mediated activation has been shown to be dependent on the binding of RXR/LXR heterodimers to a DR4 element in the promoter of the ABCA1 gene. Transcriptional sequences representing LXR response elements (or LXREs) are composed of direct repeats of the motif AGGTCA separated by four nucleotides, and this element has been shown to be activated by both ligands of RXR (rexinoids) and LXR (oxysterols) separately and together (11). These data (10 -12) have clearly shown that the LXRE in the promoter influences ABCA1 regulation in vitro. However, thus far this has been the only LXRE described in the ABCA1 gene, and there has been no in vivo validation of the sterol responsiveness of the human ABCA1 protein.
Levels of mRNA may be poor predictors of protein expression (12), as mRNA levels can vary almost 20 times and still yield the same level of gene product. Alternatively, the same level of expression of an mRNA can result in vastly different levels of a protein (13,14). Therefore, even though in vitro studies have shown an increase of ABCA1 mRNA on oxysterol stimulation (10,11), it is most important to determine whether there is an increase in ABCA1 protein associated with raised ABCA1 mRNA expression. Furthermore, whereas decreases in cellular cholesterol efflux secondary to either antisense in vitro inhibition of the gene (1) or in vivo mutations (1)(2)(3)(4)(5) are associated with decreased efflux and decreased HDL-C levels, it is currently unknown whether overexpression of ABCA1 in vivo is associated with increased HDL-C levels and an increase in tissue-specific cholesterol efflux.
The use of transgenic technologies using BACs offers important advantages for generating mice expressing human ABCA1. The inclusion of endogenous regulatory elements within the transgene allows for assessment of normal temporal, tissue-, and cell-specific expression of human ABCA1. Furthermore, inclusion of selected endogenous promoter sequences allows for dissection of the contribution of different sequences to the normal regulation of the ABCA1 gene. Such information is not possible using cDNA transgenic approaches that often result in poorly expressed genes that are not physiologically regulated.
Here we demonstrate, both in vitro and in vivo, that the ABCA1 gene has an internal promoter containing LXREs in intron 1. Activation of this functional internal promoter in human intron 1 by oxysterols in vivo directly contributes to an increase in human-specific mRNA in tissue and leads to increased protein expression. These experiments have led to the identification of three novel ABCA1 transcripts with different transcription initiation sites that utilize sequences in intron 1. In addition, increased human ABCA1 expression results in a remarkable and significant increase in cholesterol efflux and HDL-C levels. These studies provide important proof of the principle for therapeutic strategies directed toward the activation of ABCA1 expression and activity.

EXPERIMENTAL PROCEDURES
Transient Transfection Assay-Cells were transfected for 3 h by lipofection using ExGen 500 (Euromedex) in Opti-MEM I. Medium was then replaced with Dulbecco's modified Eagle's medium (DMEM) containing 0.2% fetal calf serum, and cells were incubated for 48 h. Cell extracts were prepared and assayed for luciferase activity as described (15). Twenty four hours before transfection, HepG2, HUH7, CaCo2, COS-1, and RK13 cells were plated in 24-well plates in DMEM supplemented with 10% fetal calf serum at 5 ϫ 10 4 cells/well. Transfection mixtures contained 100 ng of tkpGl3 reporter vector or pGl3 containing an 8-kb fragment from ABCA1 intron 1 (pGl3-8kb). Transfection mixtures contained 50 ng of reporter plasmid (tkpGl3) containing multiple copies of the putative LXREs and 25 ng of LXR␣ and RXR expression plasmids, in the presence of the internal control ␤-galactosidase expression vector. After transfection, cells were treated for 48 h with 1 M 22(R)-hydroxycholesterol (Sigma).
Gel Mobility Shift Assay-LXR␣ and RXR were transcribed and translated in vitro using pCDNA3-LXR␣ and pSG5-mRXR␣ as templates and the TNT-coupled transcription/translation system (Promega). Gel mobility shift assays (20 l) contained 10 mM Tris (pH 8), 40 mM KCl, 0.1% Nonidet P-40, 6% glycerol, 1 mM dithiothreitol, 0.2 g of poly(dI-dC), 1 g of herring sperm DNA, and 2.5 l each of in vitro synthesized LXR␣ and RXR proteins. The total amount of reticulocyte lysate was maintained constant in each reaction (5 l) through the addition of unprogrammed lysate. After a 10-min incubation on ice, 1 ng of 32 P-labeled oligonucleotide was added, and the incubation was continued for an additional 10 min. DNA-protein complexes were resolved on a 6% polyacrylamide gel in 0.5ϫ TBE. Gels were dried and subjected to autoradiography at Ϫ80°C.
Multicopy Cloning-250 picomoles of each oligonucleotide to which half-sites for BamHI and BglII restriction enzymes had been added were phosphorylated using T4 polynucleotide kinase (Roche Molecular Biochemicals), incubated for 5 min at 95°C and then 10 min at 65°C, and cooled to room temperature. Multimeric copies were then generated using T4 ligase, cloned in TKpGl3 vector, and verified by sequencing.
Generation of BAC Transgenic Mice-BACs containing the ABCA1 gene were identified by screening high density BAC grid filters from a human BAC library. Four BACs containing ABCA1 were sequenced as described previously (1, 6). Version 1.7 of ClustalW with modifications was used for multiple sequence alignments with Boxshade for graphical enhancement. The 5Ј end of BAC 269 is at position Ϫ13491 in intron 1 (i.e. 13491 nucleotides from the 5Ј end of exon 2). This BAC was chosen for further purification as it alone contained intron 1 sequence without the human ABCA1 promoter, allowing us to test for functionality of the putative intronic regulatory elements. The BACs were purified for injection using the Qiagen Maxi Prep kit, followed by cesium chloride purification (16) and dialysis overnight. BACs were quantified using agarose gel electrophoresis, and sets of 300 C57BL/6xCBA eggs were injected with 30 ng of the purified BAC DNA. Founders were genotyped with DNA extracted from tail pieces, followed by subsequent PCR amplification of exon 2, exon 26, and exon 49 of the ABCA1 gene.
Feeding of High Cholesterol Diets-BAC mice and control littermates were provided free access to water and a high fat/high cholesterol or a control chow diet for 7 days. The diets were purchased from Harlan Teklad with the high fat/high cholesterol diet (TD 90221) containing 15.75% fat, 1.25% cholesterol, and 0.5% sodium cholate. This diet has been shown previously to result in up-regulation of ABCA1 mRNA levels in mouse liver, assessed at 7 days after feeding (17). The control diet contained 0.5% sodium cholate (TD 99057).
Quantification of Human and Mouse ABCA1 Transcripts-RNA from mouse liver and peritoneal macrophages were isolated using Trizol (Life Technologies, Inc.), and 3 g of total RNA was reverse-transcribed using Superscript II (Life Technologies, Inc.). Human-and mousespecific primers were used along with 18 S primers (Ambion Inc.) to quantitate transcript abundance. Human-specific primers are as follows: Ex3F, CAAACATGTCAGCTGTTACTGGAAG, and Ex4R, GAGC-CTCCCCAGGAGTCG. Mouse-specific primers are as follows: Ex5F, CATTAAGGACATGCACAAGGTCC, and Ex6R, CAGAAAATCCTGCA-GCTTCAATTT. The standard cycling conditions of denaturation at 94°C for 30 s, annealing at 60°C for 45 s, and extension at 72°C for 1 min were used. PCR products were separated on 2% agarose gels, and images were captured using Bio-Rad multianalyst software using a geldoc system. Bands were quantified using NIH Image version 1.6. All values are ratios with the corresponding 18 S bands.
Quantitative Real Time PCR for Human and Mouse ABCA1 Levels-The human ABCA1 primers, mouse ABCA1 primers, and their Taqman probes were designed using Primer Express software (Applied Biosystems, Foster City, CA). The TaqMan probe contains a reporter dye at the 5Ј end and a quencher dye at the 3Ј end. The sequences of the primers and the probes are as follows: human ABC1 forward primer, 5Ј-CCTGACCGGGTTGTTCCC-3Ј, and human ABC1 reverse primer, 5Ј-TTCTGCCGGATGGTGCTC-3Ј; human ABC1 TaqMan probe, 5Ј-AC-ATCCTGGGAAAAGACATTCGCTCTGA-3Ј; mouse ABC1 forward primer, 5Ј-TCCGAGCGAATGTCCTTC-3Ј, and mouse ABC1 reverse primer, 5Ј-GCGCTCAACTTTTACGAAGGC-3Ј; and mouse ABC1 Taqman probe, 5Ј-CCCAACTTCTGGCACGGCCTACATC-3Ј. The RT-PCR was carried out on ABI P RISM 7700 in a final volume of 50 l, containing 40 ng of total RNA, 200 M primers, and 600 M probe in 1ϫ TaqMan One-step RT-PCR Master mix (PE Biosystems, CA), according to the manufacturer's instruction. The primers and probe for 18 S or rodent GAPDH were used as the internal controls for human ABC1 and mouse ABC1, respectively. The reverse transcription reaction was run at 48°C for 30 min. After activation of the AmpliTaq Gold at 95°C for 10 min, the PCR was carried out for 40 cycles (denaturation at 95°C for 15 s and annealing and extension at 60°C for 1 min). Data quantification and analysis were performed according to the manufacturer's protocol (PE Biosystems). Values were calculated relative to the level of the control. Each sample was assayed in triplicate during two independent experiments.
Detection of Alternate Transcripts Involving Intron 1 Sequence Arising from Three Different Transcription Start Sites-In order to identify the transcript generated in the BAC mice lacking exon 1, ABCA1 intron 1 sequence was searched by ProScan for putative transcription sites, and several likely sites were determined. Primers were synthesized, and CLONTECH marathon-ready mouse and human liver cDNAs were used to amplify putative transcripts using the predicted transcript primers and an ABCA1 exon 3 reverse primer, following the manufacturer's instructions. Positive transcripts were confirmed using nested PCR and sequenced. In addition, RNA was isolated from BAC transgenic and control littermate tissue using Trizol (Life Technologies, Inc.), and 5Ј-rapid amplification of cDNA ends was performed using primers described previously (18) and following the manufacturer's instructions (CLONTECH). All products were TA-cloned (Invitrogen) and sequenced using an ABI 3100 automated DNA sequencer. Primers used for transcript amplification are as follows: exon 1bF, GTTGTCATCTTTGAA-CAAACTG; exon 1cF, GAGAAGGGAACTCACATTGCTTTG; exon 1dF, CACGGTAGAACTTTCTACTGTG; and Ex3R, CATTCATGTTGTTCA-TAGGGTG. Standard cycling conditions were used for PCR amplification of all three transcripts.
Western Blot Analysis of the Distribution of ABCA1-BAC transgenic mice and control littermates were sacrificed by CO 2 inhalation, and various tissues were isolated and placed in 500 l of low salt lysis buffer containing complete protease inhibitor tablets (Roche Molecular Biochemicals) on ice. The tissues were homogenized and sonicated. The resulting homogenate was centrifuged at 14,000 rpm for 10 min at 4°C, and the supernatant was aliquoted into tubes. Protein levels were quantified using the Lowry assay. 100 -150 g of protein was separated on 7.5% polyacrylamide gels and was transferred to polyvinylidene difluoride membranes (Millipore). Membranes were probed with ABC1PEP4 polyclonal rabbit antibody (directed against residues 2236 -2259 in ABCA1) 2 or monoclonal anti-glyceraldehyde phosphate dehydrogenase (Chemicon) as a control. The membrane was dipped in ECL (Amersham Pharmacia Biotech) and exposed to X-Omat blue film (Eastman Kodak Co.). Protein levels were quantitated using NIH image software. ABC1 Immunocytochemistry-To compare further the in vivo cellular expression pattern of ABCA1 protein in both ABCA1 BAC transgenic mice and their wild type littermates, immunocytochemistry was performed on a variety of fixed tissues using the polyclonal ABC1PEP4 antibody described above. Transgenic and wild type mice were deeply anesthetized with pentobarbital, injected intraperitoneally with 100 units of heparin in sterile water, and then transcardially perfused with cold 3% paraformaldehyde in 0.1 M phosphate-buffered saline (PBS). The brain and liver were removed from each mouse and post-fixed for 24 -48 h in the same fixative. For each organ 30 -50-m sections were cut on a vibrating microtome (Vibratome). Sections were collected in sterile PBS at 4°C, rinsed in 0.1 M PBS with 0.3% Tween 20, and incubated in blocking solution (0.1% PBS with 0.3% Tween 20, 3% whole goat serum, and 5% bovine serum albumin) for 2 h at room temperature.
Sections of liver were incubated for 48 h with ABC1PEP4. Brains from each mouse were processed for combined immunocytochemistry with a neuron-specific (NeuN, Chemicon) antibody and ABC1PEP4. Sections were sequentially placed into primary antisera against ABCA1 (diluted 1:2500 in block solution) and NeuN (dilution 1:50 in block solution) for 48 h at 4°C. Following incubation with the primary antibody, sections were washed several times in blocking solution and incubated in secondary antibody for 48 h at 4°C. Secondary antibodies (Molecular probes) were used as follows: goat anti-mouse Alexa 488 with NeuN at a dilution of 1:200, and goat anti-rabbit CY-3 with ABCA1 primary at a dilution of 1:200.
Following further washes with 0.1 M PBS the sections were drymounted on gelatin-coated slides, dehydrated by serial ethanol washes, and permanently mounted with Fluoromount (Gurr). Sections were analyzed using an upright fluorescence microscope (Zeiss), and digital images were captured on a CCD camera (Princeton Instrument Inc.). Combined and NeuN/ABCA1-stained sections were processed into double immunofluorescence figures using Northern exposure image program.
Measurement of Plasma Lipid and Apoprotein Levels-Mice were either bled by saphenous vein withdrawal or by cardiac puncture, and the collected blood was added to tubes containing 5 l of 0.1 M EDTA. For the measurement of HDL-C, the plasma was mixed 1:1 with 20% PEG20, vortexed, incubated at room temperature for 10 min, and spun at maximum speed for 5 min at room temperature (19). 20 l of the resultant supernatant was added to 96-well maxisorp plates (Millipore), and 200 l of Infinity cholesterol reagent (Sigma) was added to the wells. The plates were quantified in an enzyme-linked immunosorbent assay reader at 492 nm. For the measurement of total cholesterol, 5 l of plasma was added to the same plates; 200 l of Infinity cholesterol reagent was added, and the plate was quantified as above. Triglycerides were measured by adding 10 l of the plasma to a 96-well plate, followed by the addition of 100 l of solutions from a triglyceride kit (Roche Molecular Biochemicals). FPLC separation of plasma lipoproteins was performed using two Superose TM 6 (Amersham Pharmacia Biotech) columns in series as described previously (19). Equal volumes of plasma (40 l) from mice (n ϭ 8) in each group were pooled for the analysis. The cholesterol and triglyceride content in each 0.5-ml fraction was assessed using commercially available enzymatic kits (Roche Molecular Biochemicals). Apoproteins were measured as described previously (20,21).

Establishment of Primary Fibroblast and Macrophage
Cultures-For the isolation of macrophages, mice were injected intraperitoneally with 2 ml of 3% thioglycollate and 3 days later were sacrificed by CO 2 inhalation. 5 ml of DMEM containing 10% fetal bovine serum, L-glutamine, and penicillin/streptomycin (all from Life Technologies, Inc.) was injected into the body cavity. The mouse was gently massaged, and the media were withdrawn and placed in tubes on ice. The cell pellet was resuspended in 1 ml of the above media and plated at a density of 5 ϫ 10 5 cells/ml in a volume of 300 l. The cells were incubated in a humidified atmosphere of 5% CO 2 at 37°C until used. Fibroblasts were isolated by dissecting the femurs of the mice and triturating the above media through the bone to remove all bone marrow. The media were added to tubes on ice and spun at 1200 rpm for 5 min at 4°C, and the pellet was resuspended in 10 ml of media. The cells were plated on 10-cm tissue culture plates (Corning Glass) and left in a humidified atmosphere with 5% CO 2 at 37°C.
Measurement of Efflux in Fibroblast and Macrophage Cells-24 h post-plating of the macrophage cells, [ 3 H]cholesterol (2 Ci/ml) (PerkinElmer Life Sciences) and 50 g/ml acylated low density lipoprotein (Intracel) were preincubated at 37°C for 30 min. Media were made containing DMEM, 1% fetal bovine serum, penicillin/streptomycin, Lglutamine, 1 M acyl-CoA:cholesterol O-acyltransferase inhibitor (CI-976, a kind gift from Dr. Minghan Wang) and the preincubated 50 g/ml acylated low density lipoprotein and [ 3 H]cholesterol. The media in the 24-well plates were replaced with 300 l of this cholesterol-containing media, and the plates were incubated for 24 h. The labeled media were then replaced with 0.2% defatted bovine serum albumin (Sigma) or 10% delipidated serum (Sigma) containing media for about 24 h. The media were again replaced with 300 l of DMEM, penicillin/streptomycin, L-glutamine, either with or without 20 g/ml apoAI (Calbiochem), and the treatment compounds 9-cis-retinoic acid (Sigma) and 22(R)-hydroxycholesterol (Steraloids). 24 h later, the media were withdrawn and centrifuged at maximum speed for 5 min at room temperature. 100 l of the supernatant was added to scintillation vials, and radioactivity was quantitated. 200 l of 0.1 N NaOH was added to each well containing the cells and incubated for 20 min at room temperature. 100 l of this lysate was added to scintillation vials and quantified. Efflux was calculated as the total counts in the medium divided by the sum of the count in the medium plus the cell lysate.
Statistical Analyses-All statistical analyses were performed using one-way analysis of variance followed by the Newman-Keuls post-test, except for the protein quantification and the analyses of the statistical significance between the means and S.D. of the data provided in the footnotes of Table IV. The statistical analyses of these two data sets were performed using unpaired t tests.

Intron 1 Contains a Functional
Promoter-To investigate whether the internal intron 1 fragment could drive transcription of a reporter gene, we transfected an 8-kb fragment of intron 1 upstream of exon 2 into different cell types, including several hepatic (HepG2 and HuH7), intestinal (CaCo 2 ), and renal (RK13) cell lines. Indeed, in these cell lines, a significant activation of the reporter gene was observed as compared with transfection of the empty pGl3 vector alone ( Fig. 2A).
In order to detect the presence of regulatory elements, we scanned the human ABCA1 intron 1 from position Ϫ1 to Ϫ24,156 and discovered several putative regulatory elements. Among these, we discovered several possible LXREs containing imperfect direct repeats of the nuclear receptor half-site AG-GTCA separated by four nucleotides (DR4) ( Fig. 1 and Table I) (22). LXREs are regulated by oxysterols (23,24) and are important transcription control points in cholesterol metabolism (25). An LXRE in exon 1 of ABCA1 had been shown previously to be active in the regulation of the gene in vitro (10 -12). All the putative LXREs are contained within the 8-kb fragment.
To investigate whether these DR4 elements were indeed able to bind LXR-RXR heterodimers, gel retardation assays were performed (Fig. 2B). As shown before when LXR␣ and RXR proteins were incubated with the labeled CYP7-LXRE oligonucleotide in vitro, a complex was observed in the presence of the LXR-RXR heterodimer (24). An excess of unlabeled CYP7 competed efficiently for binding to the probe, but no competition for binding was observed with a DR-2 oligonucleotide. As a control, the unlabeled LXRE described previously in ABCA1 exon 1 (ϩ4-LXRE) (10, 11) also competed efficiently for binding. The signal was then competed with increasing quantities of each unlabeled putative LXRE. As shown in Fig. 2B, all three constructs competed for binding of the CYP7A probe in a dosedependent manner, although with seemingly different efficiencies.
To determine whether these potential LXREs also possessed functional relevance, multiple copies of these oligonucleotides cloned in front of a luciferase reporter gene were assayed by cotransfection with the expression plasmids for LXR␣ and RXR in COS-1 cells (Fig. 3). As described previously, three copies of the consensus LXRE (3XLXRE), 5 copies of the CYP7 LXRE (5XCYP7-LXRE), 2 copies of the ϩ4 LXRE in ABCA1 exon 1 (2Xϩ4-LXRE) showed a strong activation in the presence of cotransfected LXR and RXR plasmids (10,24). The 3 copies of the putative 4686 LXRE and 7656 LXRE of ABCA1 intron 1 showed a 2-and 6-fold induction, respectively. In contrast, two copies of the 7174 LXRE showed a weaker activation by the LXR-RXR heterodimer.
Detection of Alternate Transcripts in Intron 1-In order to determine if the LXREs in intron 1 that we identified by bioinformatic and in vitro methods are functional in vivo, we generated mouse lines transgenic for the ABCA1 gene in which the promoter and exon 1 regions had been deleted. BAC 269 was used to aid in the identification of novel elements differing from the previously known active LXR element (10) in the promoter. Protein expressed from this BAC would be predicted to be full-length as the translation initiation site is in exon 2. We obtained several BAC founder lines, and all analyses were performed on two individual founder lines. These two founder lines had different copy numbers of human BACs. Southern blot analysis revealed founder XA with 3-4 BACs and founder XB with 1-2 BACs (data not shown). In order to elucidate the transcript generated by the BAC transgenic mice lacking exon 1 of the ABCA1 gene, we performed RT-PCR and 5Ј-rapid amplification of cDNA ends. By utilizing these approaches followed by sequencing, we identified three novel transcripts containing published exon 2 and exon 3 sequences and an additional sequence from intron 1 (Fig. 4A). These transcripts were seen at different levels in hepatic tissue. For example, the transcript with exon 1b was expressed at the highest level in livers from chow-fed BAC transgenics compared with wild type mice fed the same diet. Furthermore, all three transcripts occur in wild type mice and humans (Fig. 4B).
Human-specific ABCA1 mRNA Is Up-regulated in BAC Transgenic Mice in Response to Cholesterol Feeding-In order to determine if the human-specific transcript is present in BAC transgenic mice, and further up-regulated upon stimulation when mice are fed a high cholesterol-containing atherogenic diet, we performed human-and mouse-specific quantitative RT-PCR and quantitative real time PCR on mouse liver RNA, normalized to 18 S RNA (Fig. 5, A and C). Human ABCA1specific primers indicated the presence of human ABCA1 transcript only in the BAC transgenic lines. Quantitation of mRNA levels by real time PCR resulted in a 1.4 -2.3-fold induction of the human ABCA1 transcript when the BAC mice were fed an atherogenic diet (Fig. 5C). In the same samples, the endogenous mouse transcript is also increased in both wild type and BAC mice by 1.2-1.8-fold (Fig. 5C). In the conventional RT-PCR assay, the human ABCA1 band showed an increase of ϳ38% when the BAC mice were fed an atherogenic diet (Fig.  5B). In the same assay, the endogenous mouse transcript in both the wild type and BAC mice were increased by ϳ17% in response to feeding of the atherogenic diet (Fig. 5B). There was no difference observed in the amount of the endogenous mouse transcript in BAC mice when compared with wild type mice under chow or atherogenic diet conditions in both assays. Similar increases in both the human and the endogenous mouse ABCA1 transcripts was observed in peritoneal macrophage cells isolated from mice on atherogenic diets (data not shown). This clearly shows that the human ABCA1 gene contained within the BAC is regulated in response to dietary cholesterol in vivo. This induction is most likely due to the oxysterol-dependent activation of LXR␣ or LXR␤ from the LXREs present within the first intron of the ABCA1 gene.
Human ABCA1 Protein Is Increased in BAC Transgenic Mice-In order to determine if the ABCA1 protein is expressed in the absence of its upstream promoter and exon 1 sequences, we first performed Western blot analysis of several different tissues in the mice. We observed that there was indeed an increase in ABCA1 protein levels in the liver, small intestine, testis, stomach, and brain compared with nontransgenic mice that was distinguishable by our anti-ABCA1 antibody (Fig.  6A). When the mice were fed an atherogenic diet, of the various tissues tested, the levels of ABCA1 protein were further induced in the liver (Fig. 6C), giving us our first indication that ABCA1 expression levels could be up-regulated by elements separate from those found in its promoter and exon 1 regions, and that LXREs identified in intron 1 are likely to be functional in vivo. This alternative promoter is also active in macrophages and fibroblasts as determined by the increase in ABCA1 protein in these tissues in BAC transgenic mice and their response to oxysterol stimulation (Fig. 6B).
Protein expression in tissues does not necessarily give any indication of the cellular distribution of a protein and can lead to misinterpretation of the expression level of a protein in a particular cell. We performed further immunohistochemical analysis on liver and brain (Fig. 7, A-M) of wild type littermates and BAC transgenic mice. We observed qualitative increases in ABCA1 expression in the liver in the transgenic mice (Fig. 7, B and D) compared with wild type littermates (Fig. 7, A  and C). There was no observable alteration in the subcellular

FIG. 1. Localization of LXRE elements in the ABCA1 5 region.
A schematic diagram of the putative LXRE elements discovered in intron 1 is shown. ABCA1 genomic organization at the 5Ј end is shown above, and the ABCA1 BAC 269 is shown below. The BAC contains 13.5 kb of intron 1 sequence followed by the rest of the gene, with the ATG occurring in exon 2. Novel putative LXR elements were identified at positions Ϫ7656 bp, Ϫ7174 bp, and Ϫ4686 bp in the ABCA1 genomic DNA and are also contained within the BAC.

ABCA1 BAC Mice Show Increased HDL-C and Cholesterol Efflux
distribution of ABCA1. For example, in the cortex, ABCA1 is predominantly located in the nucleus of neurons in both transgenic (Fig. 7, K-M) and wild type mice (Fig. 7, H-J). This is the first indication of ABCA1 protein expression in different tissues including brain. There was virtually no staining observed in the primary antibody-omitted control (Fig. 7E).
ABCA1 BAC Transgenic Mice Show Increased HDL-C Apoprotein Levels-We next determined if the increase in ABCA1 protein in the BAC mice resulted in an increase in its activity by measuring the plasma lipid levels in these mice. A significant increase in HDL-C levels in the ABCA1 BAC transgenic mice compared with control littermates was seen both on chow and atherogenic diet (n ϭ 4) (p ϭ 0.005 and 0.007, respectively) (Table II and Fig. 8A). These data show that the alternate promoter in intron 1 is important and sufficiently functional to result in increased expression of ABCA1 protein and increased HDL-C levels. Furthermore, in both the BAC transgenic mice and wild type littermates, the HDL-C levels increased significantly upon feeding of a cholesterol-rich diet (n ϭ 4, p Ͻ 0.001 and p ϭ 0.002, respectively) (Table II), consistent with upregulation of the ABCA1 protein. The level of up-regulation of HDL-C in the BAC mice on atherogenic compared with chow diet was higher than the level of HDL-C increase in the wild type littermates on atherogenic versus chow diet, providing additional proof that the human ABCA1 transcript is indeed up-regulated upon stimulation through the LXR pathway.
Apoproteins AI and AII were also significantly increased in the BAC transgenic compared with wild type mice on a chow diet (n ϭ 16, p Ͻ 0.05 and p Ͻ 0.0001, respectively) (Table III).
To assess for qualitative differences in lipoprotein particles  2. Intron 1 of ABCA1 has promoter activity. A, HepG2 (liver), HuH7 (hepatoma), CaCo 2 (intestinal) and RK13 (kidney) cell lines were cotransfected with empty pGl3 vector or pGl3 containing an 8-kb fragment upstream of exon 2 of ABCA1 intron 1 (pGl3-8kb). Cells were then incubated for 48 h. Luciferase activity was determined and plotted as fold activation relative to empty pGl3-transfected cells. B, gel mobility shift assays are shown in which LXR␣ and RXR were incubated as indicated with the radiolabeled probe corresponding to CYP7-LXRE. Binding of the LXR␣-RXR heterodimer was tested by competition, by adding 5-, 25-, or 50-fold molar excess of each unlabeled oligonucleotide corresponding to the putative LXREs shown in Fig. 1 and Table I.  between the human ABCA1 BAC transgenics and their littermate controls, FPLC analysis was performed (Fig. 8B).
HDL-C levels, as indicated by the total area of the HDL peak (fractions 30 -38), were increased in the transgenic mice, compared with the non-transgenic controls. The size distribution of the HDL particles appears slightly different, as the peak appears in fraction 34 in the wild type and fraction 35 in transgenic mice, indicating an increase in slightly smaller HDL particles, with increased expression of ABCA1. This is in keeping with the role of ABCA1 in the initial lipidation of apoAI and not its subsequent enlargement. Remnant lipoproteins and low density lipoprotein cholesterol (fractions 12-20, 24 -28, respectively) levels were not readily different between transgenics and controls. HDL-C levels were further increased on feeding of the atherogenic diet (Fig. 8B) with peaks occurring in the same fraction. Thus, the increased HDL-C concentration in ABCA1 BAC transgenic mice likely reflects an increased number of HDL particles and not the presence of larger HDL particles.
ABCA1 BAC Transgenic Mice Show Increased Efflux-A defect in cholesterol and phospholipid removal mediated by apolipoproteins has been observed previously in ABCA1-defective Tangier disease fibroblasts (26), and ABCA1 has been shown to mediate cholesterol efflux to apoAI or HDL from cells (4,27). In order to determine if there was an increase in efflux of cholesterol in the mice expressing high levels of ABCA1, we established primary peritoneal macrophage and fibroblast cultures from these mice. We observed increased efflux of [ 3 H] cholesterol to apoAI from both peritoneal macrophage (Fig. 9A) (Table IV) and fibroblast (Fig. 9B) (Table IV) cultures obtained from the transgenic mice when compared with wild type littermates. These efflux levels were further significantly increased when the mice were fed the atherogenic diet (Fig. 9, A and B). We observed that there was no increase in efflux when apoAI was omitted as the efflux acceptor (data not shown). In addition, we observed that stimulation of cultures with 9-cis-retinoic acid and 22(R)-hydroxycholesterol, which are specific activators of the LXR/RXR pathway, also significantly upregulated the efflux levels from both the macrophage (Fig. 9A) and fibroblast cells (Fig. 9B and Table IV). BAC mice on atherogenic versus chow diets showed higher levels of up-regulation of efflux when compared with wild type littermates on atherogenic and chow diets, indicating a larger induction of efflux, and a larger response to LXR/RXR activation in the presence of the human ABCA1 gene, especially in fibroblasts where the FIG. 4. Characterization of the alternative splice variants containing ABCA1 intron 1. A, schematic diagram indicating the location of the splice variants discovered in the BAC transgenic mice and also confirmed in wild type mice and humans. Exon 1b is 120 bp in length and contains a TA-rich region ϳ2.5 kb upstream. Exon 1c is 136 bp in length, and exon 1d 178 bp in length. A CAAT box is located upstream of exon 1b, and CAAT and TATA boxes are found immediately upstream of exon 1c. B, identification of the alternative transcript involving exon 1b, 1c, and 1d in wild type mouse and human liver tissue. Duplex RT-PCR was performed on mouse RNA with primers generating an exon 1b transcript fragment of ϳ360 bp and a fragment corresponding to the previously characterized transcript of ϳ250 bp. Both transcripts were found in wild type and transgenic mice, with the alternative transcript being highly up-regulated in the BAC transgenic mice. Different levels of each transcript were seen in liver from humans compared with wild type mice. RT-PCR generated fragments of ϳ380 and 450 bp for exon 1c and exon 1d transcripts, respectively. These transcripts were found to be present in both mouse and human liver RNA.

ABCA1 BAC Mice Show Increased HDL-C and Cholesterol Efflux
difference in up-regulation of efflux between stimulated BACs and stimulated wild type mice was 73%. DISCUSSION Here we show that increasing human ABCA1 protein expression results in a significant increase in HDL-C, apoAI, and apoAII levels in vivo. No major change in the distribution of HDL particles is seen, suggesting that this increase in ABCA1 protein predominantly results in an increase in the number of HDL particles. We have previously shown, based on families with low HDL-C, a strong correlation between the reduction in cholesterol efflux and the decrease in plasma HDL-C in these patients (6). Here we demonstrate that increasing efflux is associated with a proportionate and predictable increase in HDL-C and raised apoAI and apoAII levels. The relationship between the increase in efflux and increase in HDL-C appears to be linear, with a correlation coefficient (r 2 ) of 0.87 (p ϭ 0.007) showing that raised efflux levels are associated directly with a proportionate increase in HDL-C in vivo. Furthermore, the rate of efflux was almost completely correlated with the level of ABCA1 protein (r 2 ϭ 0.98, p ϭ 0.001), showing that any approach that results in an increase in net functional ABCA1 protein levels in the cell could be expected to have a proportionate increase in cholesterol efflux. Moreover, the establishment of BAC transgenic mice containing sequence from all of the introns, including intron 1 without promoter sequence, has allowed for the identification of three novel transcripts initiating in intron 1 and demonstrated that an internal promoter containing LXREs in intron 1 contributes to the normal regulation of ABCA1 and its responsiveness to oxysterol stimulation.
It could be argued that the increase in ABCA1 protein and HDL levels in this study is due to in vivo regulation of the endogenous mouse ABCA1 protein alone. Numerous findings FIG. 5. Regulation of human and mouse ABCA1 transcripts. A, quantitative human and mouse ABCA1-specific RT-PCR was performed on wild type and BAC mouse tissue, both on chow and an atherogenic diet. An 18 S primer control was included with each sample. The PCRs were separated on agarose gels and quantified using NIH image. The 1st, 4th, 7th, and 10th lanes were amplified using mouse-specific ABCA1 primers; the 2nd, 5th, 8th, and 11th lanes were amplified using human-specific primers; and the 3rd, 6th, 9th, and 12th lanes were amplified using 18 S-specific primers. Mouse and 18 S-specific transcripts were amplified in all the four mice. Human-specific transcripts were only amplified in the BAC transgenic mice. B, the transcripts were quantitated using NIH image, and the ratios between the transcripts and the corresponding 18 S bands were obtained. The human transcript was up-regulated by 38% (p Ͻ 0.001) in the BAC mice fed atherogenic diet compared with those fed chow diet. The endogenous mouse transcript was up-regulated by 17% (p Ͻ 0.001) in both the wild type and BAC mice fed an atherogenic diet when compared with those fed a chow diet. There was no significant difference observed in endogenous mouse transcript levels when comparing wild type and BAC mice both on chow diet or when comparing wild type and BAC mice both on atherogenic diet. C, quantitative real time PCR was performed on RNA isolated from the liver of wild type and BAC transgenic mice, both on an atherogenic diet and a control chow diet. Two sets of mice were used in this analysis (solid bars representing one set, and the mottled bars representing the next set), each analyzed in two separate experiments in triplicate. As observed with the conventional RT-PCR method, both the human and the endogenous mouse transcript showed induction in response to the atherogenic diet. The human ABCA1 gene was upregulated by 1.4 -2.3-fold in response to the high cholesterol diet, and endogenous mouse ABCA1 transcript showed an up-regulation of 1.2-1.8-fold in response to the same diet.
FIG. 6. Expression of ABCA1 protein in BAC transgenic mouse tissue. A, Western blot analysis of various mouse tissues from BAC transgenic mice and control littermates on chow diet. Tissue from two different founder lines were analyzed and showed similar results. ABCA1 was detected in liver, brain, small intestine, testis, lung, and stomach, with highest level of up-regulation in the BAC mouse liver when compared with control. B, increase of ABCA1 protein levels in liver in response to ad libitum feeding of an atherogenic diet for 7 days. There was a graded increase in ABCA1 protein levels, with liver from wild type chow fed animals showing the lowest levels and transgenic animals fed the atherogenic diet showing the highest levels. All Western blots were probed with an anti-GAPDH antibody (Sigma) to ensure equal protein loading levels, and the corresponding GAPDH lanes are shown below the Western blots. C, Western blot analysis was also performed on cultured macrophage and fibroblast cells that were used for efflux assays. ABCA1 protein was detected in both peritoneal macrophage and fibroblast cells, with the transgenic animals showing higher levels of protein that the control littermates. The protein levels in these tissues were also increased in response to feeding of an atherogenic diet.
argue against this. First, the base line and increase in ABCA1 protein in the BAC transgenic mice was significantly greater than seen in the control littermate mice (p ϭ 0.049, n ϭ 3). This could only reasonably be ascribed to the effects of the human ABCA1 protein. In addition, quantitative PCR using mouseand human-specific primers clearly has shown an increase after feeding (38%) of human ABCA1 mRNA that was more than 2ϫ greater than the increase in endogenous mouse mRNA in the same experiments. This provides formal proof that transcription of the human ABCA1 with only intron 1 shows cholesterol-responsive regulation in vivo.
The human ABCA1 gene consists of 50 exons spanning 149-kb genomic DNA (28). Translation begins in exon 2, and transcription had been shown previously to be initiated at a 303-bp exon located 24,459 bp upstream of exon 2 (18). Here we have shown three other transcription initiation sites utilizing FIG. 7. Distribution of ABC1 protein in wild type and ABC1 BAC transgenic mice. Localization of ABC1 protein was determined by immunocytochemical analysis using an ABC1-specific polyclonal antibody (ABCPEP4) (red) in liver and brain tissues. Endogenous levels of ABC1 are identified in sections from wild type liver (A, low power, and C, high power). Elevated ABC1 levels are seen in liver tissues from BAC transgenics (B, low power, and D, high power). The tissue distribution of ABC1 is similar in sections from cerebral cortex from both wild type and BAC transgenic brains (F and G).   (1)(2)(3)(4)(5)(6). However, in our own studies, in some patients in whom mutations have been mapped to this particular gene, no DNA sequence variation in the coding region or splice donor/acceptor sites has been detected that could account for the phenotype observed. The approach to assessing the mutations had been to look at each splice site and exon, as well as the regular promoter in an effort to identify potential DNA variants that could account for the disruption of protein function (1,6). The failure to detect mutations in some of these patients, together with the finding of the importance of these alternate transcripts in the regulation of the ABCA1 gene, may explain how expression could be compromised in some patients with defects in efflux that map to this gene but in whom no mutations have yet been described. One example of mutations disrupting the ratios of alternative protein isoforms implicated as the cause of abnormal phenotype is that affecting urogenital development in Denys-Drash syndrome (29). Further analysis and comparison of the sequence of the different ABCA1 transcripts may help to identify missing mutations and confirm the functional significance of these sequences. It is apparent that different transcription start sites using an alternate promoter involving sequence in intron 1 can be used to enhance the information contained with the ABCA1 gene. Alternate splicing of nuclear pre-mRNA is a general mechanism for controlling gene expression leading to various RNA isoforms from a single primary transcript (30 -32). What is unusual here is that the splicing event involves intronic sequence, which in contrast to alternate splicing of exonic sequence has only been described infrequently (33)(34)(35). The specific capacities of these sequences in intron 1 for protein FIG. 8. Analysis of plasma lipid levels in transgenic mice. A, BAC transgenic mice show a 65% increase in HDL-C levels compared with control littermates on a chow diet. Furthermore, HDL-C levels in both BAC transgenic and wild type mice were increased by Ͼ100% in response to feeding of the atherogenic diet, with the BAC transgenic mice having close to 2ϫ the HDL-C seen in the nontransgenic littermates. B, whereas quantitative changes are apparent, there are no major qualitative changes in HDL-C in transgenic versus nontransgenic mice, either on chow (A) or atherogenic diets (B).  9. Analysis of efflux levels in primary macrophage and fibroblast cells. A, primary macrophage cultures were established from mice as described. Efflux was measured 24 h after the addition of apoAI and after stimulation by 9-cis-retinoic acid and 22(R)-hydroxycholesterol (n ϭ 4, see Table IV). We observed a significant increase (46%) in efflux levels of BAC transgenic macrophages when compared with macrophages from wild type (wt) littermates on the same chow diet (p Ͻ 0.001). Both sets of animals showed an increase in efflux upon stimulation of the cultures with 9-cis-retinoic acid and 22(R)-hydroxycholesterol. BAC transgenic mice on the atherogenic diet showed an increase in efflux when compared with BAC and Wt mice on the control chow diet (p Ͻ 0.0001 and p ϭ 0.0004 respectively). This efflux rate was further increased when the cultures were stimulated with 9-cis-retinoic acid and 22(R)-hydroxycholesterol (p Ͻ 0.01). B, efflux was also performed on fibroblast cultures established from BAC transgenic and control mice (n ϭ 4, see Table IV). An increase in efflux was seen in BAC transgenic mice when compared with Wt littermates on chow diet (p ϭ 0.03). This level is further increased in both transgenic and Wt mice by 55.8% in response to stimulation by 9-cis-retinoic acid and 22(R)-hydroxycholesterol. BAC transgenic mice fed the atherogenic diet showed a significant (Ͼ100%, p Ͻ 0.0001) increase in efflux levels when compared with chow fed transgenic animals. These levels were mildly increased (by 11.2%) when the cultures were stimulated with 9-cisretinoic acid and 22(R)-hydroxycholesterol.
interactions and the importance of the contribution of these specific sequences in modulating cellular responses to physiological signals, such as oxysterol stimulation, when compared with the promoter LXREs, remains to be determined. However, it is clear that alternate transcript decisions in regard to intron 1 sequence are influenced by specific factors that may vary in different cell types, suggesting this event is of primary importance.
These newly discovered alternate transcripts are not seen equally in all tissues and, therefore, may provide further insights into the complex tissue-specific regulation of this gene, with certain transcripts likely to play a more major role in certain tissues. The presence of these alternate transcripts is also seen in endogenous mouse tissues, but there appears to be species-specific regulation of ABCA1, with these transcripts not being detected in all tissues at the same level as they are seen in humans.
Species-specific regulation of other genes involved in HDL metabolism has been reported. Fibrates, as an example, decrease the transcription of the ApoAI gene in rats, whereas in humans this clearly results in activation of ApoAI gene expression (20,21). The availability of human ABCA1 transgenic mice further allows the investigation of the role of other transcription factors influencing the responsiveness of the intron 1 promoter to oxysterol stimulation. The breeding of these mice to others where various transcription factors are no longer present will help to determine their role in influencing the responsiveness of this promoter to oxysterol stimulation.
We have shown that intron 1 of the ABCA1 gene contains an internal promoter that is sufficient to drive ABCA1 protein expression and can regulate responsiveness to LXR/RXR stimulation in vitro and in vivo. These LXREs, which are more than 15 kb away from the previously identified promoter, clearly identify the importance of intragenic sequences for the regulation of ABCA1. The LXREs we identified appear functional in vivo, resulting in significantly raised HDL-C levels and increased ABCA1 protein expression, particularly in liver, brain, small intestine, macrophages, and fibroblasts. Our experiments also demonstrate cross-species functional complementarity with murine LXR-␣, -␤, and RXR-␣ sufficient to transactivate the human ABCA1 gene.
Cavalier et al. (36) have recently described a similar line of human BAC transgenic mice lacking exon 1 of the ABCA1 gene, although its effects on plasma lipid levels and cholesterol efflux are not reported. Interestingly, they only describe one transcript in these mice, which is equivalent to our exon 1c tran-script. They also describe a line of transgenic mice containing a full-length BAC but were unable to demonstrate differences in cholesterol efflux and plasma lipid levels in these mice.
The mice described by Cavalier et al. (36) were created on the FVB background. Our mice are C57BL/6xCBA/J hybrids. Strain differences in HDL and its metabolism and response to a high fat diet are well documented (37)(38)(39)(40)(41)(42)(43)(44)(45)(46)(47). As we are unaware of any studies comparing lipid metabolism in FVB mice to other strains, a direct comparison cannot be made. However, there are numerous ways these strain differences might affect HDL-C levels in transgenics created in them. For example, strain differences may contribute to factors such as apoAI acceptor levels which may, in turn, influence the ability of increased ABCA1 to increase plasma HDL-C concentrations. Similarly, factors influencing cholesterol removal from HDL or the turnover of various HDL subspecies may also influence the ability to observe ABCA1-mediated increases in HDL-C.
The availability of mice described in this paper will now allow us to address the question as to how effectively these animals can resist experimental atherosclerosis. Since the first description of the cellular defect in Tangier disease, where decreased HDL levels appeared to be associated with a decrease in cholesterol efflux (48,49), the question as to whether increasing efflux would result in an increase in HDL-C levels and decreased atherosclerosis has been present. This challenge assumed greater importance with the discovery of the ABCA1 gene as the gene mutated in Tangier disease (1)(2)(3)(4)(5)(6). The discovery and demonstration that increasing cholesterol efflux can indeed be associated with an increase in HDL-C levels provide additional support for the development of therapeutics influencing ABCA1 protein expression.