Sterol Regulation of 3-Hydroxy-3-Methylglutaryl-coenzyme A Synthase Gene through a Direct Interaction Between Sterol Regulatory Element Binding Protein and the Trimeric CCAAT-binding Factor/Nuclear Factor Y*

3-Hydroxy-3-methylglutaryl-coenzyme A (HMG-CoA) synthase, a key regulatory enzyme in the pathway for endogenous cholesterol synthesis, is a target for negative feedback regulation by cholesterol. The promoter for HMG-CoA synthase contains two binding sites for the sterol regulatory element-binding proteins (SREBPs). When cellular sterol levels are low, the SREBPs are released from the endoplasmic reticulum membrane, allowing them to translocate to the nucleus and activate SREBP target genes. In all SREBP-regulated promoters studied to date, additional co-regulatory transcription factors are required. In the HMG-CoA synthase promoter there are several potential co-regulatory transcription factor binding sites, including an inverted CCAAT box. A similar element has been shown to function with SREBP to mediate sterol regulation of another gene involved in cholesterol metabolism, farnesyl diphosphate synthase. Here, we show that CCAAT binding factor/nuclear factor Y (CBF/NF-Y) binding to the CCAAT box is required for sterol-regulated transcription of HMG-CoA synthase. The SREBP sites and the inverted CCAAT box are normally separated by 17 base pairs, and we show that increasing this distance results in a decrease in the level of transcriptional regulation by sterols. Furthermore, we provide evidence that there is a direct interaction between CBF/NF-Y and the basic helix-loop-helix-zipper region of SREBP. Interestingly, this interaction does not occur efficiently with any of the isolated subunits and appears to require all three nonidentical CBF/NF-Y subunits in a preassembled complex. Since CBF/NF-Y only binds to DNA when all three subunits are in a complex, this would prevent SREBP from forming nonproductive associations with the individual subunits.

3-Hydroxy-3-methylglutaryl-coenzyme A (HMG-CoA) 1 syn-thase is an early enzyme in the cholesterol biosynthetic pathway and is a target for negative feedback regulation by cholesterol (1,2). When cellular sterol levels are low, the transcription of the HMG-CoA synthase gene is turned on, and when cellular sterol levels rise, transcription is turned off. This negative feedback regulation is mediated in part by the sterolregulatory element-binding proteins (SREBPs). The SREBPs are synthesized as 125-kDa membrane-spanning proteins. When sterol levels fall, these proteins are released from the endoplasmic reticulum membrane by a two step proteolytic process that results in the release of the mature ϳ68-kDa amino-terminal portion (3). This mature protein translocates to the nucleus and activates genes involved in both cholesterol and fatty acid metabolism through binding to sterol-regulatory elements (SREs) present in the promoters of target genes (4 -10).
By themselves, SREBPs are very inefficient transcriptional activators and therefore require additional co-regulatory transcription factors to achieve high level activation. In the low density lipoprotein receptor, acetyl CoA carboxylase, and fatty acid synthase promoters, the co-regulatory factor is Sp1 (5,7,11,12). In the case of farnesyl diphosphate (FPP) synthase, squalene synthase, and HMG-CoA synthase, the co-regulatory factor is the CCAAT binding factor (CBF), also called nuclear factor Y (NF-Y) (6,10,13).
CBF/NF-Y is a ubiquitously expressed, trimeric transcriptional activator that binds to CCAAT motifs in a number of eukaryotic promoters (14). In most cases, it functions together with more specific transcriptional activators to achieve high levels of transcriptional activation in specific tissues or under certain conditions. In the albumin promoter, for example, CBF/ NF-Y functions together with liver specific members of the CCAAT/enhancer-binding protein (C/EBP) family to achieve a high level of liver specific expression (15).
The purpose of the current studies was to further define the role of CBF/NF-Y in the sterol regulated transcription of the HMG-CoA synthase gene. The sterol regulatory region of the HMG-CoA synthase promoter is contained within a 100-bp segment of the promoter, from -324 to Ϫ225 relative to the mRNA start site. This region contains an inverted CCAAT motif and two SRE sites, as shown in Fig. 1A. The CCAAT motif and the SRE sites are separated from each other by 17 bp in the wild-type promoter. Here we show that CBF/NF-Y binds to this inverted CCAAT motif located at position Ϫ295 and that this site is required for sterol-regulated transcription. We also show that inserting additional nucleotides between these two sites resulted in a loss of sterol-regulated transcription, suggesting that a physical interaction occurs between SREBP and NF-Y on the promoter that is required for regulation. We provide evidence that the bHLH region of SREBP interacts with * This work was supported in part by Grants from the National Institutes of Health (HL48044) and the American Heart Association (93001330 and California Affiliate Grant 93-275). 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.
‡ Supported by National Institutes of Health Training Grant GM07311.
CBF/NF-Y in solution in the absence of DNA. Interestingly, this interaction does not occur with any of the individual CBF/ NF-Y subunits, but requires all three subunits to be present at the same time.

MATERIALS AND METHODS
Cells and Media-CV-1 cells were obtained from Dr. K. Cho (University of California, Irvine). HepG2 cells were obtained from the ATCC. HeLa nuclear extracts were purchased from Endotronics (Minneapolis, MN). All cell culture materials were purchased from Life Technologies Inc. Lipoprotein-deficient serum was prepared by ultracentrifugation of newborn bovine serum as described previously (16). Cholesterol and 25-hydroxycholesterol were obtained from Steraloids Inc., and stock solutions were dissolved in absolute ethanol.
Plasmids-pGL2 basic was purchased from Promega and was used as the source of the luciferase reporter gene in all constructs. Standard techniques were used in all cloning procedures (17). pSynWT contains nucleotides Ϫ488 to ϩ38 relative to the mRNA start site of the hamster HMG-CoA synthase promoter inserted upstream of the luciferase reporter gene in pGL2 basic (18). To construct the plasmid pSynSRE, polymerase chain reaction amplification of the sequence corresponding to Ϫ324 to Ϫ225 of the HMG-CoA synthase promoter was performed using oligonucleotides containing SstI and NheI restriction sites attached to the 5Ј and 3Ј synthase oligonucleotides, respectively. The resulting fragment was inserted upstream of the minimal HMG-CoA synthase TATA box (Ϫ28 to ϩ39) of the plasmid "TATA only" described previously (11). The insertion mutations, pSynSREϩ5 and pSyn-SREϩ10 were constructed using the Altered Sites mutagenesis kit (Promega) according to the manufacturer's protocol. The oligonucleotides used were complementary to the 5Ј and 3Ј sequences relative to the insertion site at Ϫ290 and contained the following insertion sequences: 5Ј-CATGG-3Ј, 5Ј-CATGGTGGAC-3Ј, for the ϩ5 and ϩ10 mutants, respectively. All mutants were confirmed by sequencing. pSyn⌬NF-Y was constructed in the same way using an oligonucleotide containing the scramble mutation at the site of the inverted CCAAT box (5Ј-ATTGGC-3Ј 3 5Ј-AGATCT-3Ј) beginning at position Ϫ295 relative to the mRNA start site.
The plasmid pNF-YA29, which encodes a dominant-negative version of the A subunit of NF-Y (19) expressed from the simian virus 40 early promoter, was a kind gift of Dr. Robert Mantovani from the Universita Degli Studi Di Milano, Italy. pGex2T (Pharmacia) was used to construct all glutathione S-transferase (GST) fusion genes for protein expression. The pCiteCBF-A, pCiteCBF-B, and pCiteCBF-C constructs for use in in vitro transcription/translation reactions were a kind gift of Dr. Sankar Maity (Dept. of Molecular Genetics, University of Texas M. D. Anderson Cancer Center, Houston, TX). pCMV 2 ␤-galactosidase contains the cytomegalovirus early promoter linked to the Escherichia coli ␤-galactosidase gene (11).
Cell Culture and Transient Transfection Assays-CV-1 cells were cultured at 37°C and 7% CO 2 in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum (normal) and were plated at 125,000 cells/60-mm dish on day 0. On day 1, cells were refed normal media and transfected by the calcium phosphate co-precipitation method as described (11). Precipitates contained 20 g each of the luciferase reporter test plasmid and the nonsterol-regulated pCMV 2 ␤-galactosidase plasmid. Where indicated, precipitates also contained the expression plasmid for pNF-YA29. Salmon sperm DNA was added to bring the total DNA to 60 g/2 ml. An equal volume (0.5 ml) was added to each of four dishes, for a total of 15 g DNA/60-mm dish. 12-16 h later, the cells were washed three times with phosphate-buffered saline, refed either induced (Dulbecco's modified Eagle's medium ϩ 10% lipoprotein-depleted serum (LPDS)) or suppressed medium (same as induced medium but also containing 10 g/ml cholesterol and 1 g/ml 25-hydroxycholesterol) and returned to the incubator. On day 3, cells were harvested by scraping in phosphate-buffered saline, and duplicate dishes were pooled prior to preparation of soluble protein extracts by 3 freeze-thaw cycles.
Enzyme Assays-The luciferase activities were measured in an analytical luminescense monolight model 2010 luminometer with a luciferin reagent from Promega. ␤-galactosidase assays were performed by a standard colorimetric procedure with 2-nitrophenyl-␤-D-galactopyranoside as substrate (17). The normalized luciferase activity was obtained by dividing the luciferase activity in relative light units by the ␤-galactosidase activity (A 420 /h).
DNase I Footprinting-Recombinant 6XHis-tagged SREBP-1a (amino acids 1-490) protein was expressed in E. coli BL21DE3(pLysS) and purified via nickel chelation chromatography as described previously (11). The concentration and purity of the protein was determined by SDS-PAGE analysis performed with marker proteins followed by staining with Coomassie Blue. A double-stranded probe, end-labeled on the top strand with [␥-32 P]ATP and T4 polynucleotide kinase (Pharmacia Biotech Inc.) in the linker region attached to Ϫ324 of the HMG-CoA synthase promoter was prepared and incubated with or without SREBP 1a protein and treated with DNase I as described previously (11).
Gel Mobility Shift Assay-A 45-bp double-stranded oligonucleotide probe corresponding to bases Ϫ324 to Ϫ285 of the wild-type HMG-CoA synthase promoter was end-labeled with [␥-32 P]ATP and T4 polynucleotide kinase (Pharmacia). The mutant probe (Syn⌬NF-Y) is exactly the same except that it contains the CCAAT box scramble mutation, 5Ј-ATTGGC-3Ј 3 5Ј-AGATCT-3Ј beginning at position Ϫ295. The labeled probes were used in an electrophoretic mobility shift assay as described previously (11). HepG2 nuclear extracts were prepared by the method of Attardi and Tjian (20), and 2 g total nuclear protein were used in each electrophoretic mobility shift assay. Where indicated, either 1 g of antiserum directed against the A subunit of NF-Y antibody (Rockland Inc.) or 1 g of anti-Sp1 antibody (Santa Cruz Biotechnology, Inc.) was incubated with the HepG2 nuclear extract for 60 min on ice prior to incubation with the labeled probes.
In Vitro Translation of CBF-pCiteCBF-A, pCiteCBF-B, and pCiteCBF-C plasmids (21) were purified by the cesium chloride gradient method (17). Using the TNT coupled transcription/translation kit (Promega), rabbit reticulocyte lysates were programmed with 1 g of either pCiteCBF-A, pCiteCBF-B, or pCiteCBF-C in the presence of 35 S-translabel (ICN) according to the manufacturer's protocol. For the co-translation of CBF-A and CBF-C, 0.75 g of each plasmid was used, and for the translation of all three CBF subunits together, 0.5 g of each plasmid was used. The synthesis of full-length proteins was verified by autoradiography following SDS-PAGE.
Expression and Preparation of GST Fusion Proteins-GST expression constructs were made by polymerase chain reaction amplification with Pfu polymerase (Stratagene, Inc.) of the coding region for amino acids 1-490 of SREBP 1a, the corresponding region of SREBP-1c, or amino acids 1-481 of SREBP-2, and inserting the resulting fragments in frame with the GST-coding region of pGex2T. Truncated SREBP 1a constructs encoding amino acids 1-320 or 320 -420 were generated in the same way with appropriate oligonucleotide primers. The fusion proteins were expressed in E. coli strain BL21DE3. 500-ml cultures were grown to A 600 of 0.6 to 0.8, induced by adding isopropyl-1-thio-␤-D-galactopyranoside to final concentration of 1 mM, and grown for an additional 3 h at 37°C. Cells were then harvested by ultracentrifugation. The cell pellets were resuspended in NETN (100 mM NaCl, 20 mM Tris, pH 8, 1 mM EDTA, 0.5% Nonidet P-40) and lysed by sonication, and the cell debris was removed by ultracentrifugation. Glycerol was added to the supernatant to a final concentration of 10%, and extracts were stored in aliquots at Ϫ70°C. The expression of the fusion proteins was verified by SDS-PAGE analysis followed by Coomassie Blue staining, as well as by ECL Western blot (Pierce) analysis using anti-GST antibody and horseradish peroxidase-conjugated anti-rabbit IgG (Sigma).
Protein-Protein Interaction Assay-500 l of the indicated crude GST fusion protein extract and 1 ml of buffer Z'0.1 (50 mM Hepes, pH 7.5, 2 mM MgCl 2 , 100 mM KCl, 10% (v/v) glycerol, 0.1% (v/v) Nonidet P-40) ϩ 0.5% nonfat dry milk and 5 mM dithiothreitol were added to 25 l of glutathione-agarose beads, followed by an incubation on a rotating wheel at 4°C for 2 h. The bound GST fusion proteins were pelleted in a microcentrifuge, and nonspecifically bound proteins were removed by washing the pellets 3 times with 1 ml of buffer. Then, 25 l of the indicated in vitro translation reaction or 150 g of total protein from HeLa nuclear extract were added to the bound proteins along with 1 ml of buffer, and samples were incubated at 4°C for 2 h on a rotating wheel. The reactions were again washed 3 times with 1 ml of buffer, and the specifically bound proteins were pelleted, resuspended in sample buffer, and analyzed by SDS-PAGE followed by either autoradiography in the case of the in vitro translated proteins, or by ECL Western blot analysis using anti-NF-YA antibody in the case of the HeLa nuclear extract.

RESULTS
In an effort to gain further insight into how SREBPs function together with unique co-regulators in the different SREBPregulated promoters we have been studying how the SREBPs function together with CBF/NF-Y to activate the promoter for HMG-CoA synthase. A schematic representation of the promoter for hamster HMG-CoA synthase is shown in Fig. 1A. When the synthase promoter was fused to the luciferase re-porter gene and transfected into CV-1 cells, the promoter was seven times more active in cells cultured in the absence of sterols compared with cells cultured in the presence of sterols.
To simplify further analysis of regulatory elements involved in this process, we made a smaller version of the synthase promoter, pSynSRE, which lacks bases from Ϫ488 to Ϫ325 and from Ϫ224 through Ϫ29 as shown in Fig. 1. When this truncated version was fused to the luciferase reporter gene and analyzed, it was subject to efficient sterol regulation as well.
The HMG-CoA synthase promoter region in pSynSRE contains two consensus binding sites for the SREBP proteins, as well as binding sites for the ATF/AP-1 family of transcription factors and an inverted CCAAT box that is a consensus binding site for the CBF/NF-Y protein. The two SRE elements were previously shown to be required for sterol regulation (2). To directly determine if they both bind SREBP proteins, we performed a DNase I footprinting experiment with a probe from the synthase promoter and purified recombinant SREBP-1a protein (Fig. 1B). When increasing amounts of SREBP-1a were added to the synthase probe, both predicted sites were protected from DNase I cleavage consistent with both sites binding SREBP protein.
FIG. 1. The HMG-CoA synthase promoter. A schematic of the HMG-CoA synthase promoter from position Ϫ488 to ϩ38 relative to the mRNA start site (SynWT) or from Ϫ324 to Ϫ225 linked to Ϫ28 to ϩ38 (SynSRE) is shown in panel A. The relative positions of the SRE I 5Ј and 3Ј which bind to SREBPs, the inverted CCAAT box which binds NF-Y, the CRE/AP-1 site which binds members of the CREB/ATF family of transcription factors as well as AP-1 proteins, and the GC-rich sequences which bind Sp1 are indicated. The fold regulation indicated to the right is the promoter-reporter activity observed in CV-1 cells transiently transfected with 5 g of the corresponding reporter plasmid and grown in the absence (induced) or the presence (suppressed) of cholesterol (fold regulation ϭ induced/suppressed). A double-stranded 32 P-end-labeled probe corresponding to Ϫ324 to Ϫ225 of the HMG-CoA synthase promoter was used in a DNase I footprint analysis (B), as described under "Materials and Methods" in the presence of increasing amounts of recombinant SREBP-1a protein (amino acids 1-490) (lanes 2-9). Lanes 2-9 correspond to 180, 360, 720, and 900 ng, and 1.3, 1.8, 2.7, and 4.5 g, of SREBP-1a protein, respectively). The position of the actual "footprints" are shown on the left, and the position of G residues in the sequence are indicated on the right.
To determine if the CCAAT sequence might function as a co-regulatory site for sterol regulation by the SREBPs, we made a mutant derivative of pSynSRE, which simultaneously alters all bases of the inverted CCAAT box beginning at position Ϫ295. The mutant construct was designated pSyn⌬NF-Y. When it was transfected into CV-1 cells and compared directly to the wild-type pSynSRE plasmid, it was completely defective for sterol regulation (Fig. 2, A and B, compare lanes 1 and 6).
Since NF-Y is one protein that has been shown to bind to CCAAT elements, we transfected a dominant-negative expression construct for the A subunit of NF-Y, pNF-YA29, along with the wild-type pSynSRE plasmid. When the cells were cultured in medium containing lipoprotein-depleted serum, a dose-dependent decrease in transcriptional activity was observed (Fig.  2, A and B, lanes 1-5). Taken together, these data show that the inverted CCAAT box at Ϫ295 is required for the sterolregulated transcription of the HMG-CoA synthase promoter, and consistent with an earlier report (13), support a role for CBF/NF-Y in sterol regulation of the promoter for HMG-CoA synthase.
Several transcription factors have been shown to bind to CCAAT box sequences, including CBF/NF-Y, C/EBP, and NF-1 (14). To determine if CBF/NF-Y could specifically bind to the HMG-CoA synthase CCAAT box sequence, we performed gel shift experiments with HepG2 nuclear extract and 32 P-endlabeled probes encompassing this region of the synthase promoter corresponding to either the wild-type (SynSRE) or CCAAT box scramble mutant (Syn⌬NF-Y). There were two complexes that were produced only with the wild-type probe (Fig. 3, compare lanes 2 and 6), indicating that they contain proteins specific for the CCAAT box sequence. The migration of the upper of these two complexes was further retarded in the presence of an antibody specific for the A subunit of NF-Y (lane 3) but not in the presence of an antibody that is specific for Sp1 (lane 4). Thus, this upper CCAAT box-specific complex is due to the binding of NF-Y.
The lower CCAAT box-specific complex observed with the wild-type probe has not been identified and probably represents another CCAAT box-binding protein. Several shifted bands were observed with both the wild-type probe and the mutated probe (NS), indicating that these complexes are notspecific for the CCAAT box sequence. These results extend the observations of an earlier report (13), and taken together with the results of Fig. 2, further support a role for NF-Y in sterol regulation of the HMG-CoA synthase promoter.
The binding of CBF/NF-Y to a site within the FPP synthase promoter enhances the binding of SREBP-1a at an adjacent SRE element (22). This stimulatory effect and sterol regulation were both lost when 4 bp were inserted between the CCAAT box and the SRE element in FPP synthase. In the wild-type HMG-CoA synthase promoter, the inverted CCAAT box and the SRE I 5Ј are separated by 17 bp. To determine if the relative spacing of these two sites was important for sterol regulation of the HMG-CoA synthase promoter, we inserted 5 or 10 bp between the synthase SRE and the inverted CCAAT element as diagrammed in Fig. 4C. These insertion mutants were then transfected into CV-1 cells and assayed for sterol regulation as described above. The activities from the ϩ5 and ϩ10 promoters under induced conditions were significantly reduced, indicating that the relative spacing between the SRE and CCAAT elements is important for sterol regulation of the synthase promoter. These results support the notion that the CCAAT box and SRE sites must function together to achieve optimal sterol regulation.
The experiments presented so far suggest there may be a functional interaction between the CBF/NF-Y protein, which binds to the inverted CCAAT site, and SREBP, which binds to the SRE element. To test this hypothesis, we constructed GST-SREBP fusion proteins for use in in vitro protein-protein interaction assays (Fig. 5). For these experiments we fused the entire mature forms of SREBP-1a, -1c, or -2 to the glutathione S-transferase coding sequence, and as a control we used a vector that expressed the GST portion without any fused sequence.
Plasmids encoding each of the GST proteins were induced in E. coli, and the crude extracts were bound to glutathioneagarose beads and then incubated either alone or in the presence of nuclear extract protein from HeLa cells. Material that FIG. 2. NF-Y is involved in the sterol-mediated regulation of the HMG-CoA synthase promoter. CV-1 cells were transiently transfected with luciferase reporter constructs pSynSRE or pSynSRE containing a mutation that scrambles the inverted CCAAT box sequence beginning at position Ϫ295 from 5Ј-ATTGGC-3Ј to 5Ј-AGATCT-3Ј (pSyn⌬NF-Y). After transfection, the cells were cultured in medium containing 10% lipoprotein-depleted serum in the absence (ind) or presence (sup) of 10 g/ml cholesterol and 1 g/ml 25-hydroxycholesterol. Where indicated (lanes 2-5), increasing amounts (100 ng, 300 ng, 1 g, or 3 g) of pNF-YA29, a dominant-negative expression construct for the A subunit of NF-Y (19), were co-transfected with pSynSRE. For all transfections, a CMV 2 -␤-galactosidase construct was co-transfected as a control for transfection efficiency. The average of the luciferase reporter activity under induced (ind) and suppressed (sup) conditions for three individual experiments is shown in panel A. The luciferase activity is plotted as "normalized" luciferase activity, which was obtained by dividing the relative luciferase units (RLU) by the ␤-galactosidase activity (A 420 /h) to obtain the normalized values. The data from panel A is represented as "fold regulation" by taking the normalized luciferase activity under induced growth conditions divided by that under suppressed growth conditions (B). bound specifically to the beads was retained after extensive washing and then analyzed by SDS-PAGE followed by Western blot analysis using an antibody specific for the A subunit of NF-Y. A single, Ϸ42-kDa immunoreactive band corresponding to NF-YA was specifically detected in the HeLa nuclear extract (lane 1). An identical band was present in the sample eluted from the GST-SREBP-1a (lanes 3 and 4) GST-SREBP-1c (lanes 5 and 6) and GST-SREBP-2 (lanes 7 and 8) reactions. In contrast, NF-YA was not present in the sample incubated with only the GST protein (lanes 2 and 3). These data are consistent with a direct interaction between the SREBP proteins and NF-Y.
CBF/NF-Y is a trimeric protein complex consisting of A, B, and C subunits, which are simultaneously required for DNA binding (21). The experiment shown in Fig. 5 indicates that SREBPs interact with at least the A subunit of NF-Y (B subunit of CBF). However, since CBF/NF-Y is a trimeric complex, the results from Fig. 5 are also consistent with SREBP interacting only with the preassembled trimeric complex, perhaps via the B or C subunit, and thereby indirectly interacting with the A subunit. To address this question, we transcribed/translated the subunits of CBF/NF-Y in vitro in the presence of 35 S-labeled amino acids either individually or in several combinations and used the crude translation mixes in GST-interaction assays with the SREBPs similar to the experiments described in Fig. 5.
When GST-SREBP-1a-bound agarose beads were incubated with the translated reactions for either CBF-A or CBF-C, there was no detectable interaction (Fig. 6A, lanes 5 and 7) There was a low level of CBF-B retained, but this was also observed for the GST control resin (Fig. 6C). In contrast, when all three CBF subunits were co-translated and incubated with the GST-SREBP-1a resin, all were strongly retained (Fig. 6A, lane 9). When the GST-SREBP-1c or GST-SREBP-2 bound material was analyzed, similar results were obtained (Fig. 6A, lanes 11  and 13). This indicates that SREBP and CBF/NF-Y interact in solution in the absence of DNA, and all three subunits must be present for an efficient interaction to occur.
To identify which regions of the SREBPs interact with CBF/ NF-Y, we created mutant derivatives of SREBP-1a. In one version aa 1-320 were fused to GST, and in the other version, amino acids 320 -420 were fused to GST. As depicted at the bottom of Fig. 6B, the first 320 amino acids of SREBP-1a contain both its acidic activation domain and a serine-threonine rich region, whereas amino acids 320 -420 includes the  4. Insertions between the SRE I 5 and the inverted CCAAT box of the HMG-CoA synthase promoter compromise the fold regulation in response to sterols. Insertion mutations of 5 or 10 base pairs between the SRE I 5Ј and the inverted CCAAT box of SynSRE were created as described under "Materials and Methods," and analyzed as described in the legend to Fig. 1. A, the induced (ind) and suppressed (sup) relative luciferase activities are shown for the wild-type and mutant promoter-reporter constructs. The value for the relative luciferase activity under induced conditions for pSynSRE was set to 100%, and all values were normalized relative to this value. B, the data from panel A is represented as fold regulation by taking the normalized luciferase activity under induced growth conditions divided by that under suppressed growth conditions. C, the sequence of the wild-type HMG-CoA synthase promoter from Ϫ306 to Ϫ266 is shown with the location and sequence of the insertion mutations indicated. The inverted CCAAT box which binds NF-Y and the SRE I 5Ј which binds SREBPs are also noted.
These fusion proteins were individually expressed in E. coli and used in the GST-protein interaction assay as described above. When incubated with all three co-translated CBF/NF-Y subunits, the truncated SREBP-1a including the bHLH-Zip domain interacted very strongly with CBF/NF-Y, as evidenced by the presence of three 35 S-labeled bands corresponding to CBF-A, -B, and -C (Fig. 6B, lanes 6 and 7), whereas the other truncated version, which contained the first 320 amino acids of SREBP-1a, did not show any significant retention of the labeled CBF/NF-Y subunits (Fig. 6B, lanes 4 and 5). Thus, the heteromeric CBF/NF-Y protein interacts with the bHLH region of the SREBPs.

DISCUSSION
The region of the HMG-CoA synthase promoter necessary for sterol-regulated transcription includes two SRE I elements and an inverted CCAAT box (Fig. 1A). The two SRE I elements were previously shown to be simultaneously required for sterol-regulated transcription (2), and we have shown that mature SREBP-1a binds to both the 5Ј and the 3Ј SRE I sites (Fig. 1B). SREBP-2 also binds to both the 5Ј and 3Ј SRE I sites (data not shown).
Although both SRE I sites are required, by themselves they are not sufficient for sterol regulation of the HMG-CoA synthase promoter (2,11,12). This is similar to the situation for all other SREBP-regulated promoters that have been carefully studied to date; additional co-regulatory factors are required for efficient transcriptional activation (5, 7, 9 -11, 22). In HMG-CoA synthase at least one of the co-regulatory sites is an inverted CCAAT element, and we have shown that this site is necessary for sterol-regulated transcription (Fig. 2).
The promoter for FPP synthase, another gene encoding a cholesterol synthetic enzyme, also contains a CCAAT box in addition to an SRE element. In FPP synthase, this CCAAT box is required for sterol-regulated transcription of this gene as well (22). Several transcription factors bind to CCAAT box sequences (14). Here we have shown that the CBF, also termed NF-Y, binds to the CCAAT box at Ϫ295 in the HMG-CoA synthase promoter (Fig. 3). This is consistent with earlier studies where overexpression of a dominant-negative version of one CBF/NF-Y subunit decreased sterol regulation of both FPP synthase and HMG-CoA synthase (13).
To extend these earlier studies we made a mutation that specifically altered the CCAAT element in the HMG-CoA synthase promoter, and we demonstrated that this mutation eliminated the binding of CBF/NF-Y (Fig. 3) and abolished regulation by sterols (Fig. 2). Additionally, we have shown that when increasing amounts of a vector expressing the dominant-negative NF-YA was transfected along with the HMG-CoA synthase promoter a dose-dependent decrease in promoter activity was observed (Fig. 2).
CBF/NF-Y is a ubiquitously expressed transcription factor that is composed of three nonidentical subunits all of which must be present in a complex for specific DNA recognition. CBF/NF-Y binding to CCAAT boxes present in several eukaryotic promoters is important in their transcriptional regulation (23)(24)(25)(26). In the case of the albumin promoter, CBF/NF-Y cooperates with a member of the C/EBP transcription factor family to activate a high level of albumin expression specifically in the liver (15). In this case, the positive synergism between C/EBP and NF-Y does not appear to occur at the level of DNA binding but rather at a subsequent step in transcriptional activation. In contrast, in the FPP synthase promoter NF-Y appeared to stimulate the binding of SREBP to an adjacent SRE (22). These results are consistent with a physical interaction between CBF/ NF-Y and SREBP.
To study this further in the HMG-CoA synthase promoter, first we inserted additional bases between the 5Ј SRE I and the inverted CCAAT box to separate the two elements and possibly decrease the potential for an interaction between proteins bound at the two sites. As shown in Fig. 4, the insertion of 5 or 10 bp did result in a loss of regulated transcription, which is also consistent with an interaction occurring between SREBP and CBF/NF-Y bound at the two sites. Because one turn of the DNA helix is roughly 10 bp, this experiment also provides evidence that any interaction that occurs is not strictly dependent on helical phasing of the DNA.
Furthermore, we have shown by in vitro protein-protein interaction assays, that GST-SREBP fusion proteins interact specifically with CBF/NF-Y (Fig. 5). Therefore, in the case of the HMG-CoA synthase promoter, it appears that there is a physical interaction between SREBP and CBF/NF-Y that is required for maximal transcriptional activation. These results do not exclude the possibility, however, that CBF/NF-Y is also important at a subsequent step in transcription, as was seen in the case of the albumin promoter mentioned above.
The three subunits of CBF/NF-Y assemble in an ordered fashion prior to binding DNA (21). First CBF-A and CBF-C, each of which contains a histone-fold motif (27), form a dimer that is then recognized by CBF-B. This heterotrimer is then competent to recognize DNA. The Western blot procedure used to show SREBP interacted with CBF/NF-Y present in the crude HeLa extract was performed with an antibody specific for the A subunit of NF-Y (analogous to the B subunit of CBF). From this experiment, it was impossible to determine if SREBP interacted specifically with the A subunit of NF-Y or indirectly via interaction with one or more of the subunits of the complex.
To distinguish between these possibilities, protein-protein interaction assays with the GST-SREBP fusion proteins and individual CBF/NF-Y subunits or various combinations were performed. The results demonstrated that GST-SREBP-1a did FIG. 5. CBF/NF-Y interacts with SREBP-1a, -1c, and -2 in solution. GST-SREBP fusion proteins were expressed in E. coli, and the resulting bacterial extracts were used in protein-protein interaction assays as described under "Materials and Methods." The fusion proteins (GST only, lanes 2 and 3; GST-SREBP-1a, lanes 4 and 5; GST-SREBP-1c, lanes 6 and 7; GST-SREBP-2, lanes 8 and 9) were incubated with (lanes 3, 5, 7, and 9) or without (lanes 2, 4, 6, and 8) 150 g total of nuclear protein from HeLa cells. After binding to glutathione-agarose beads and extensive washing, aliquots were analyzed by SDS-PAGE followed by ECL Western blot analysis using anti-NF-YA antibody. Lane 1 corresponds to 12 g of HeLa nuclear extract loaded directly as a control for the presence of NF-YA. not interact efficiently with any of the individual subunits or with the CBF A and C dimer. However, when all three subunits were co-expressed, there was an efficient interaction between agarose bound SREBP and CBF/NF-Y (Fig. 6).
Taken together, the data indicate that SREBP must interact with a motif in the CBF/NF-Y molecule that is recognizable only when all three subunits are assembled as a trimer. Since CBF/NF-Y only binds DNA in its trimeric state, this would prevent SREBP from forming nontranscriptionally active complexes with the individual CBF/NF-Y subunits. This type of combinatorial interaction between transcription factors is unusual, and there are relatively few examples of similar observations.
In two related reports, the MADS box containing serum response factor (SRF) and MEF2 proteins were shown to interact only with a mixed heterodimer composed of one subunit of either myoD or myogenin, both of which are muscle specific bHLH proteins and the ubiquitous E12 protein (28,29). In these two cases, the SRF or MEF2 protein only bound the bHLH heterodimer and did not appreciably interact with either bHLH partner protein in isolation.
By using truncated versions of SREBP-1a fused to GST we showed that the bHLH zipper participates in the interaction with CBF/NF-Y. This region is the most highly conserved domain between SREBP-1 and SREBP-2 because they share identity at 76% of the positions in this domain (30), so it is not surprising that both SREBP-1c and SREBP-2 interacted with CBF/NF-Y (Fig. 6). Since SREBP and CBF/NF-Y are both required for sterol regulation of HMG-CoA synthase, the physical interaction between them is probably fundamental to their ability to function together to activate transcription of the HMG-CoA synthase promoter.
To date, Sp1 and CBF/NF-Y are the only ubiquitous transcription factors that have been directly implicated in gene activation along with the SREBPs. In the low density lipoprotein receptor system, SREBP enhanced the binding of Sp1 to the promoter (11), whereas in FPP synthase, CBF/NF-Y enhanced the binding of SREBP to the promoter (22). Even though DNA recognition may occur by different mechanisms, in each case SREBP and the co-regulator were both required for transcriptional activation in response to low sterols. Thus, the co-regulatory factors Sp1 and CBF/NF-Y may function similarly at a subsequent step in transcriptional activation perhaps by contacting a common component of the basal transcription machinery.
Interestingly, both Sp1 and subunits A and C of NF-Y (subunits B and C of CBF) have glutamine-rich regions reminiscent of activation domains present in other transcription factors (31). These regions are important for the transcriptional activities of both proteins (27,32,33), and in the case of Sp1, the glutamine activation domain contacts the TATA box-binding protein-associated factor TAF110 (34). Further studies will be necessary to determine exactly how these transcription factors work together with SREBPs to activate transcription.  1, 2, and 3, respectively. GST-SREBP-1a was bound to glutathioneagarose beads, then incubated with mock-translated extract (lane 4) or with aliquots of in vitro translated CBF subunits (lanes 5-9) alone or in combinations, and followed by extensive washing before analysis. The CBF subunit(s) present in the translation reactions corresponding to each lane are indicated (ϩ). The same procedure was followed for GST-SREBP-1c (lanes 10 and 11) or GST-SREBP-2 (lanes 12 and 13) as indicated. B, 2 l of the 50-l crude transcription/translation reaction for CBF-A, CBF-B, or CBF-C, were loaded in lanes 1, 2, and 3, respectively. Protein-protein interaction assays for GST fusion proteins of the truncated forms of SREBP-1a containing either amino acids (aa) 1-320 (lanes 4 and 5) or amino acids 320 -420 (lanes 6 and 7) were performed as well. Reactions included mock extract (Ϫ) or extract from reactions containing all three CBF subunits as indicated (ϩ). A diagram of the mature transcriptionally active SREBP-1a protein is shown at the bottom with the location and amino acid positions of the acidic activation and bHLH-leucine zipper (bHLH-Zip) DNA binding domains indicated. C, the simple GST protein was used as a control in proteinprotein interaction assays as described above and under "Materials and Methods." The translated CBF subunit(s) used in each reaction are indicated (ϩ).