Regulation of Cholesterologenesis by the Oxysterol Receptor, LXRα*

Cholesterol is required for normal cellular and physiological function, yet dysregulation of cholesterol metabolism is associated with diseases such as atherosclerosis. Cholesterol biosynthesis is regulated by end product negative feedback inhibition where the levels of sterols and oxysterols regulate the expression of cholesterologenic enzymes. Sterol regulatory element-binding protein-2 is responsive to both sterols and oxysterols and has been shown to mediate the transcriptional response of the cholesterologenic enzymes to these lipids. Here, we show that the nuclear hormone receptor for oxysterols, the liver X receptor α (LXRα), regulates cholesterol biosynthesis by directly silencing the expression of two key cholesterologenic enzymes (lanosterol 14α-demethylase (CYP51A1), and squalene synthase (farnesyl diphosphate farnesyl transferase 1)) via novel negative LXR DNA response elements (nLXREs) located in each of these genes. Examination of the CYP51A1 gene revealed that both the SRE and nLXRE are required for normal oxysterol-dependent repression of this gene. Thus, these data suggest that LXRα plays an important role in the regulation of cholesterol biosynthesis.

Cholesterol is essential for maintenance of cellular membrane fluidity and is a precursor for the production of steroid hormones and bile acids. Although cholesterol is required for normal cellular and physiological function, dysregulation of cholesterol metabolism is associated with atherosclerosis and increased risk of heart disease (1). Thus, cholesterol metabolism is under strict biological regulation, and drugs inhibiting cholesterol biosynthesis such as the statins, inhibitors of 3-hydroxy-3-methylglutaryl-coenzyme A reductase (HMGCR) 2 (see Fig. 1a), have been successfully utilized clinically to reduce the risk of atherosclerosis. The biosynthesis of cholesterol has long been known to be regulated by end product feedback inhibition, and this regulation has been attributed to direct regula-tion of the expression of several cholesterol biosynthetic genes by the sterol sensing sterol regulatory element-binding protein-2 (SREBP-2) (2). The SREBPs are responsive to both alterations in cholesterol levels as well as oxidized cholesterol metabolites known as oxysterols (3).
The nuclear hormone receptors LXR␣ (NR1H3) and LXR␤ (NR1H2) have been demonstrated to be physiological receptors for oxysterols (4 -6) and regulate components of the cholesterol metabolic pathway including reverse cholesterol transport and cholesterol elimination (7). No role for the LXRs in direct regulation of cholesterol biosynthesis has been described. A recent study by Wong et al. (8) suggested that 24,25-epoxycholesterol was important for "fine-tuning" acute control of cellular cholesterol biosynthesis. Because this endogenous oxysterol is known to be one of the more potent natural oxysterol ligands for the LXRs, it is possible that some of its activity may be mediated by these receptors. Here, we show that two key cholesterologenic enzyme genes are LXR target genes and that LXR plays a role in the regulation of cholesterol biosynthesis.

EXPERIMENTAL PROCEDURES
Plasmid Construction-The CYP51A1 promoter (Ϫ2217 to ϩ155) was amplified from genomic DNA isolated from HepG2 cells and cloned into pTAL-Luc luciferase report vector (Clontech, CA) by introducing KpnI and BglII into the primers. The proximal portion (Ϫ2217 to Ϫ1749) containing LXRE and the distal portion (Ϫ1749 to ϩ155) containing the SRE of the promoter were amplified from the CYP51A1 promoter vector and cloned into pTAL-Luc vector using the same restriction sites. Three copies of CYP51A1 LXRE, FDFT1 LXRE, and NR1H3 LXRE were cloned into pTAL-Luc through MluI and BglII. LXR␣ and RXR␣ were cloned into pDEST14 (Invitrogen) using Gateway technology (Invitrogen) for EMSA analysis. LXR␣ and LXR␤ were cloned into pcDNA3.1 vector for overexpression analysis. The LXRE mutant CYP51A1 promoter reporter was created using QuikChange II sitedirected mutagenesis kit (Stratagene) according to the manufacturer's instructions.
Cell Culture and Cotransfections-HepG2 cells were maintained and routinely propagated in minimum essential medium supplemented with 10% fetal bovine serum at 37°C under 5% CO 2 . HEK293 cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum at 37°C under 5% CO 2. In experiments where lipids and sterols were depleted, cells were maintained on charcoal treated serum (10% fetal bovine serum) and treated with 7.5 M lovastatin and 100 M mevalonic acid. 24 h prior to transfection, HepG2 or HEK293 cells were plated in 96-well plates at a density of 15 ϫ 10 3 cells/well. Each transfection contained 100 ng of the pTAL-Luc reporter, 50 ng of pGL4.73 reporter, and 50 ng of receptor (as indicated) as described in the figure legend using Lipofectamine TM 2000 (Invitrogen). 16 h post-transfection, the cells were treated with LXR ligands. 24 h post-treatment, the luciferase activity was measured using the Dual-Glo TM luciferase assay system (Promega). The values indicated represent the means Ϯ S.E. from three independently transfected wells. The experiments were repeated at least three times, and representative experiments are shown.
CYP51A1, FDFT1 HMG-CoA, and LXR␣ protein levels were assessed by Western blot using antibodies from Abnova (Taiwan). LXR␣ and control siRNA was obtained from Ambion (Austin, TX). The transfection was performed using siPORT TM NeoFX SiRNA transfection reagents in accordance with the manufacturer's instructions.
Cholesterol Quantification and Biosynthesis Measurement-Cholesterol quantification was performed using a cholesterol/cholesterol ester quantification kit (BioVision). The cholesterol level was normalized to protein content. Cholesterol de novo biosynthesis was measured by TLC. HepG2 cells were treated with siRNA for 45 h, and the media were replaced with growth media supplemented with 1 Ci/ml of 14 C sodium acetate (Amersham Biosciences). After 3 h of incubation at 37°C under 5% CO 2 , total lipid was extracted and dissolved in chloroform:methanol (2:1). TLC was performed in heptane:isopropylether:acetic acid (160 ml/40 ml/2 ml) for 1 h. The TLC plate was scanned using Bioscan AR2000 Imaging Scanner (Bioscan, Washington, D. C.). The number of counts/min was normalized to protein content.
ChIP-on-chip-ChIP-on-chip procedures were performed as previously described (9).
Microarray-Microarray identification of genes altered by knock-down of LXR expression in mice was previously described (10).
HMG-CoA Reductase Enzyme Assay-HMG-CoA reductase activity was measured by spectrophotometric assay as described by Takahashi et al. (12). The reaction buffer consisted of 25 mM potassium phosphate (pH 7.2), 50 mM KCl, 1 mM EDTA, and 5 mM dithiothreitol. Control and siRNA-treated HepG2 cells were cultivated in 6-well plates. After treatment, the cells were harvested in 100 l of reaction buffer/well. The assay was performed in reaction buffer with 0.3 mM NADPH, 0.3 mM R,S-HMG-CoA and 2 l protein in a final volume of 200 l. The kinetics program in the Benchmark Plus TM microplate spectrophotometer was used, and the reaction was monitored every 20 s for up to 10 min at 37°C. One unit of HMG-CoA reductase activity is defined as the amount of the enzyme that caused oxidation of 1 mol of NADPH/min at 37°C.

RESULTS
Identification of Novel LXR␣ Target Genes-To identify novel biochemical pathways regulated by LXR␣, we performed LXR␣ chromatin immunoprecipitation followed by whole genome analysis of the enriched DNA fragments by microarray experiment (ChIP-on-chip) (9). Regions of the genome with significant LXR␣ occupancy (termed promising active regions (PARs)) were screened for putative LXREs using a previously described algorithm (13,14). PARs were then queried against the human genome data base, and 1304 unique genes were identified with a PAR within 1 kb of a gene. We then compared these data with microarray data obtained from the livers of mice treated with a LXR-directed antisense oligonucleotide that significantly reduced LXR␣ expression (10). This allowed us to identify genes that were significantly altered by LXR␣ knockdown that also demonstrated LXR␣ occupancy and contained a defined LXRE. Fifty-seven genes were identified as having met these criteria and included the well characterized LXR target genes ABCA1 and NR1H3, indicating that the method was successful. Upon closer examination of the putative target genes, it was noted that there were examples of genes whose expression was increased in response to a decrease in LXR␣ expression, suggesting that LXR␣ may play a direct repressive role in regulation of these genes. We found this particularly intriguing and novel given that LXR has not been previously characterized as a direct silencer of gene expression.
Interestingly, two of the negatively regulated genes encode key enzymes within the cholesterol biosynthesis pathway. The two genes identified were squalene synthase (FDFT1) and lanosterol 14␣-demethylase (CYP51A1), and their positions within the cholesterologenic pathway are illustrated in Fig. 1a. Both have regions of significant LXR␣ binding (spanning 2-3 kb) as determined by ChIP-on-chip either in the 5Ј region or in the intronic region of the genes (Fig. 1b). To investigate the putative role of LXR␣ in regulation of cholesterologenesis, we decreased LXR␣ expression in the human hepatoma cell line, HepG2, by 37% (LXR␣ protein decreased 56%) using specific siRNA ( Fig. 2a) (9). As illustrated in Fig. 2b, decreasing LXR␣ expression led to a significant increase in expression of both identified genes. CYP51A1 and FDFT1 expression increased by ϳ50 and ϳ42%, respectively. Protein levels of CYP51A1 and FDFT1 were also examined following LXR␣ knock-down and as illustrated in Fig. 2c, CYP51A1 expression increased ϳ70% and FDFT1 increased ϳ42%. These data suggest that the dominant influence of LXR␣ on the expression of these genes is negative because "removal" of a significant amount of LXR␣ results in increased expression of these genes.

Regulation of Cholesterologenic Enzyme Expression by LXR␣-Both FDFT1
and CYP51A1 encode enzymes that are essential for cholesterologenesis. FDFT1 encodes the enzyme squalene synthase that catalyzes the production of squalene from farnesyl phyrophosphate and is the first committed step in sterol synthesis (Fig. 1a). Inhibitors of this enzyme have been shown to be effective inhibitors of cholesterol synthesis (15,16). CYP51A1 encodes lanosterol 14␣-demethylase, which catalyzes removal of the 14␣-methyl group from the cholesterol intermediate lanosterol (17) (Fig. 1a). Inhibition of CYP51A1 activity by ketoconazole in humans results in a 27% decrease in total serum cholesterol and a 41% decrease in serum low density lipoprotein levels, illustrating the importance of this enzyme in the cholesterol biosynthesis pathway (18).
Inhibition of CYP51A1 activity results in accumulation of lanosterol in humans (18), and lanosterol has been suggested to regulate HMGCR proteasomal degradation (19). A more recent study indicated that dihydrolanosterol and not lanosterol is responsible for regulation of HMGCR proteosomal degradation (20). Because CYP51A1 utilizes both dihydrolanosterol and lanosterol as a substrate (17), regulation of CYP51A1 activity has similar effects on the levels of both of these sterols (18), and in fact, it appears that inhibition of CYP51A1 activity has a much larger effect on the level of dihydrolanosterol relative to lanosterol (18). Thus, we hypothesized that LXR␣ may regulate HMGCR expression via modulation of dihydrolanosterol levels by its direct ability to repress CYP51A1 expression. Indeed, we found that depletion of LXR␣ expression resulted in a 3-fold increase in HMGCR expression (Fig. 3a) and a 46% increase in HMGCR activity (Fig. 3b). Thus, LXR␣ regulates the expression of three key enzymes in the cholesterol biosynthesis pathway.
LXR␣ Is Required for Normal Levels of Cholesterologenesis-Based on these data, we predicted that LXR␣ knock-down would result in increased cholesterol content in HepG2 cells. Fig. 4a shows that this is indeed the case. When LXR␣ expression is decreased by 56%, intracellular cholesterol levels  increase ϳ30%. We also examined de novo cholesterologenesis in this model and noted a 44% increase in cholesterol biosynthesis in LXR␣-depleted cells (Fig. 4b). These data are consistent with a model where LXR␣ regulates the expression of CYP51A1 and FDFT1 and the stability of HMGCR, resulting in modulation of cholesterol biosynthesis.
Identification of Negative LXR DNA Response Elements-Because CYP51A1 and FDFT1 were identified based on LXR␣ occupancy in ChIP-on-chip analysis in addition to the requirement for a putative LXRE within the occupancy site, these data indicate that these may be the first observations of "negative" LXREs in the genome. Analysis of the PARs from CYP51A1 and FDFT1 (Fig. 1b) by the LXR␣ ChIP-on-chip study revealed putative LXREs. We performed EMSAs on all putative LXREs and identified two LXREs (one in each gene) that bound to the LXR␣/RXR␣ heterodimer effectively (Fig. 5). The novel LXREs exhibited similarity to other well characterized "positive" LXREs identified in the ABCA1 and NR1H3 genes (Fig. 2a). One is a DR4 element, whereas other is a very unique DR2 element. No LXREs with a DR2 organization have been previously identified. RXR␣ heterodimerization was required for LXR␣ binding to both novel LXREs (Fig. 5b). To assess the functionality of these novel LXREs, we cloned three copies of each upstream of a minimal promoter-luciferase reporter construct and examined their activities in the absence and presence of either synthetic (GW3965 or T0901317 (T1317)) or natural (oxysterol; OHC) LXR ligands in HepG2 cells. The positive LXRE and CYP51A1 and FDFT1 LXREs both affected basal transcription but with opposing effects (Fig. 6a).
The NR1H3 positive LXRE increased transcription by 200%, whereas the two other LXREs decreased basal transcription by 42-56% (Fig. 6a). The addition of ligands had no effect on basal transcription from a reporter lacking a LXRE (data not shown). In cells transfected with the positive LXRE, a further increase in transcription was noted with treatment with either of the two synthetic ligands (5-6-fold induction) (Fig. 6b). Both LXREs derived from the CYP51A1 and FDFT1 showed minimal responsiveness to the synthetic LXR ligands (Fig. 6, c and d), and quite novel was our observation that the nLXRE reporters from each of the both cholesterologenic enzyme genes conferred oxysterol-   dependent repression of transcription to the reporter construct (Fig. 6, c and d). We termed these LXREs "negative" LXREs (nLXREs) based on their ability to confer negative responsiveness to LXR. Both of these nLXREs were respon-sive to oxysterols that are known agonists of LXR such as 24,25-epoyxycholesterol and 25-hydroxycholesterol but were unresponsive to well characterized agonist such as 22(R)-hydroxycholesterol. This unusual ligand selectivity, taken together with their relative lack of responsiveness to synthetic LXR ligands, indicates that these novel nLXREs may confer differential ligand sensitivity to the receptor. We also examined the activity of these three LXRE containing reporter constructs in HEK293 cells. Like HepG2 cells, HEK293 cells also express LXR, and the LXR responsiveness of reporter genes can be enhanced by overexpression of LXR (21). We transfected the reporters into HEK293 cells in the absence or presence of a vector directing the overexpression of either LXR␣ or LXR␤. As illustrated in Fig. 7a, the reporter containing the positive LXREs from the NR1H3 gene displayed enhanced transcription (53%), and overexpression of either LXR resulted in further enhancement of transcription (Fig. 7a). Consistent with our data from the HepG2 cells, transfection of the cells with the reporters containing the nLXREs derived from the CYP51A1 or FDFT1 genes led to decreased basal transcription (44 -50%) that was further suppressed upon overexpression of either LXR␣ or LXR␤. We next examined the ability of LXR ligands to alter the transcriptional activity of LXR␣ transfected into HEK293 cells along with the nLXRE reporters derived from either the CYP51A1 or FDFT1 genes. The results were very similar to those obtained in HepG2 cells in that the identical oxysterols that repressed transcription in HepG2 cells also repressed in HEK293 cells (Fig. 7, b and c). These data are consistent with our observations that the nLXREs mediate LXRdependent repression of transcription.
Both the nLXRE and the SRE Are Required for Oxysterol Regulation of CYP51A1 Expression-Because the CYP51A1 promoter contains a SRE that is also responsive to oxysterols via SREBP, we assessed the LXR␣-derived component of repression by depleting LXR␣ expression with siRNA and examining the 25-OHC responsiveness in HepG2 cells cultured in lipid and oxysterol-depleted conditions. Depletion of LXR␣ resulting in a significant loss of the repressive effects of 25-OHC, suggesting that a significant component of the repressive effects is mediated by this receptor (p Ͻ 0.05) (Fig. 8a). Based on previous publications illustrating that SREBP-1a/2 plays an important role in CYP51A1 expression (22) as well as our data indicating that LXR␣ also is important, we hypothesized that both the nLXRE and the SRE play a role in oxysterol responsiveness of the gene. To investigate this, we constructed a CYP51A1 promoter reporter by cloning ϳ2 kb of the CYP51A1 promoter upstream of a luciferase reporter and as shown in Fig.  3b. When this construct was transfected into HepG2 cells, transcription is repressed ϳ20% relative to the control (Fig. 8b). Treatment with 25-OHC results in further repression (ϳ75% relative to control; Fig. 8b). We manufactured CYP51A1 promoter-reporter constructs in which we deleted the nLXRE or the SRE as illustrated in Fig. 8b. Interestingly, removal of the nLXRE resulted in an increase in transcription (Fig. 8b), suggesting a basal repressive function of this element consistent with what we have seen in heterologous systems such as our previous cotransfection studies (Figs. 6 and 7). 25-OHC treatment represses transcription driven by this CYP51A1 promoter fragment containing only the SRE, but not to the same absolute level as that observed when both the nLXRE and the SRE elements were present (Fig. 8b). Deletion of the SRE results in basal expression similar to control and 25-OHC treatment represses transcription, but the magnitude of repression is clearly less efficacious than when the nLXRE is present. We examined the role of the nLXRE more closely by creating a mutant LXRE CYP51A1 promoter (Fig. 8c) where the LXRE is unable to bind to LXR. We confirmed that the mutant LXRE is unable to bind to LXR␣ in an EMSA (Fig. 8d). Both the wild type promoter and the LXRE mutant promoter were transfected into HepG2 cells, and the ability of 25-OHC to repress transcription was examined. As illustrated in Fig. 8e, mutation of the LXRE within the CYP51A1 promoter results in significant loss (but not entire loss) of the ability of the oxysterol to inhibit transcription. Thus, these data clearly indicate that both the nLXRE and SRE play a role in mediating the oxysterol responsiveness of transcription of this gene and additionally indicate that the nLXRE also has a basal repressor effect on transcription of this gene.

DISCUSSION
These findings indicate that, beyond the role of LXR␣ in the regulation of cholesterol transport and elimination, this nuclear receptor also plays an important role in end product inhibition of cholesterol biosynthesis (Fig. 9). It is interesting to note that the two enzymes whose expression is directly repressed by LXR␣ are in steps of cholesterologenesis where the pathway has committed to sterol synthesis. This is in contrast to SREBP-2, which regulates the expression of enzymes throughout the pathway. Our data suggest that LXR␣ works in concert with SREBP to regulate the expression of CYP51A1 in response to 25-OHC. 24,25-Epoxycholesterol has been recently shown to be an important for fine-tuning acute control of cellular cholesterol homeostatis (8). Because this endogenous oxysterol is one of the more potent natural ligands of LXR, it is possible that at least some of the activity of 24,25-epoxycholesterol described by these investigators may be mediated via the LXR-dependent mechanism we have described.
All of the genes directly regulated by LXR described to date are under positive regulation by LXR (7) similar to the NR1H3 FIGURE 8. Examination of the role of the nLXRE in the CYP51A1 promoter. a, LXR␣ depletion by siRNA treatment results in significant loss of the ability of 25-OHC to repress CYP51A1 expression in HepG2 cells. b, deletion analysis suggests that both the nLXRE and the SRE mediate a component of the ability of 25-OHC to repress transcription of a CYP51A1 promoter driven luciferase reporter in HepG2 cells. c, schematic illustrating the LXRE mutant that was created to eliminate LXR binding to the CYP51A1 promoter reporter construct. d, EMSA demonstrating that the mutant LXRE does not bind to LXR␣. A competition experiment was performed with the wild type LXRE as the labeled probe, and the binding was competed with either cold wild type oligonucleotide (5ϫ, 25ϫ, and 100ϫ molar excess) or mutant oligonucleotide (5ϫ, 25ϫ, and 100ϫ molar excess). e, elimination of LXR binding results in reduced ability of 25-OHC to repress transcription of the CYP51A1 reporter. HepG2 cells were treated with 10 M 25-OHC resulting in ϳ72% repression of transcription of the wild type promoter but only ϳ45% inhibition of transcription from the LXRE mutant CYP51A1 promoter. All of the experiments within this figure utilized lipid depleted medium as described under "Experimental Procedures." Me 2 SO concentrations in all wells were held constant. *, p Ͻ 0.05 versus control as determined by Student's t test.
gene, indicating that the mode of regulation of CYP51A1 and FDFT1 is very unique. Promoter-specific transcriptional repression by unliganded LXR has been reported previously (23), but in this case the addition of an LXR ligand results in activation of transcription not further repression as we have noted for the CYP51A1 LXRE. A recent study indicates that LXR␣ may repress Ucp1 expression via a mechanism that involves competition with other transcription factors within the promoter (24). LXR has also been shown to repress the C-reactive protein gene, but it is not clear whether this is mediated by an LXRE (25). Additionally, LXR has been demonstrated to repress a number of genes involved in the inflammatory response utilizing a mechanism that has been attributed to antagonism of the NF-B pathway (26). Thus, the CYP51A1 and FDFT1 LXREs represent a novel class of LXRE (nLXRE) that confers negative responsiveness to LXR and oxysterols. Additionally, the nLXRE is differentially responsive to distinct classes of LXR ligands (e.g. synthetic versus oxysterols). The mechanism responsible for the differential responsiveness is currently unclear, but it has been suggested that synthetic versus oxysterol binding to LXR results in different conformations of the receptor (27). Although our data suggest that these oxysterol-specific repressive effects are mediated via direct binding of the ligands to LXR because we observe dependence on LXR binding to the promoter, it is possible that other effects of these specific oxysterols may also be playing a role (e.g. oxysterol-dependent post-translational modification of the receptor).
It remains to be determined whether there are additional genes that contain a functional nLXRE. Because LXR is a putative target for the development of anti-atherogenic agents (28), the activity of synthetic LXR ligands on nLXRE containing genes such as CYP51A1 and FDFT1 must be considered. These results also suggest putative therapeutic potential of LXR agonists as inhibitors of cholesterol biosynthesis. Our data support a model for oxysterol regulation of cholesterol biosynthesis in which SREBP-2 increases cholesterologenic enzyme expression in response to decreasing oxysterol levels, whereas LXR␣ responds to increasing oxysterol levels to repress the expression of key cholesterologenic enzymes (Fig. 8). This mode of dual regulation of cholesterologenesis may provide for exceptionally sensitive and stringent control of cholesterol produc-tion where its levels must by delicately balanced to support cellular function but also avoid toxicity.