Identification of an Inverted CCAAT Box Motif in the Fatty-acid Synthase Gene as an Essential Element for Mediation of Transcriptional Regulation by cAMP*

The antagonistic effect of cAMP on the insulin-in- duced expression of fatty acid synthase (FAS) in liver could be mimicked in vitro using H4IIE hepatoma cells, both by measuring the response of the endogenous FAS gene and by assaying expression of transfected reporter genes containing promoter elements of the FAS gene. 5 (cid:42) -Deletion analysis and replacement mutagenesis revealed that an essential element required for cAMP an- tagonism of the insulin effect is an inverted CCAAT box located between nucleotides (cid:50) 99 and (cid:50) 92. DNase I footprinting and gel shift analysis revealed that this region can bind a protein present in nuclei of liver and spleen, organs that express high and undetectable levels of FAS, respectively. This protein is not a CCAAT/enhancer-binding protein, C/EBP. Thus, the FAS gene appears unusual in that the sequence element required for transcriptional regulation by cAMP is neither a cAMP re- sponse element (CRE) nor a binding site for AP-1, AP-2, or C/EBP. These results suggest that essential to the regulation of FAS transcription by cAMP is the interaction of an inverted CCAAT box motif with a constitu- tively produced trans -acting factor that either itself undergoes modification in response to cAMP or associates with a protein that is produced or modified by cAMP exposure. the DNA binding and dimerization domains of hu- man CREB, residues 254–327) were purchased from Santa Cruz Bio-technology, Inc. The bacterial expression vector, pET-15b-C/EBP (cid:97) was obtained from Dr. Per Flodby at the Karolinska Institute, Huddinge, Sweden and used to produce extracts of Escherichia coli containing soluble C/EBP (cid:97) . Cell Cultures— Rat H4IIE hepatoma cells, purchased from the Tissue (cid:50) 30 FAS-TK Plasmid— Plasmid (cid:50) 124/ (cid:49) 68 was digested with Hin dIII and Dra I, and the Hin dIII site was blunt- ended by treatment with the Klenow fragment of DNA polymerase I. This fragment was then ligated to pTK-CAT which had been digested with Xba I and similarly treated with the Klenow enzyme to give plasmid (cid:50) 124/ (cid:50) 30 FAS-TK. The orientation of the FAS insert was deter- mined by DNA nucleotide sequencing.

The antagonistic effect of cAMP on the insulin-induced expression of fatty acid synthase (FAS) in liver could be mimicked in vitro using H4IIE hepatoma cells, both by measuring the response of the endogenous FAS gene and by assaying expression of transfected reporter genes containing promoter elements of the FAS gene. 5-Deletion analysis and replacement mutagenesis revealed that an essential element required for cAMP antagonism of the insulin effect is an inverted CCAAT box located between nucleotides ؊99 and ؊92. DNase I footprinting and gel shift analysis revealed that this region can bind a protein present in nuclei of liver and spleen, organs that express high and undetectable levels of FAS, respectively. This protein is not a CCAAT/enhancerbinding protein, C/EBP. Thus, the FAS gene appears unusual in that the sequence element required for transcriptional regulation by cAMP is neither a cAMP response element (CRE) nor a binding site for AP-1, AP-2, or C/EBP. These results suggest that essential to the regulation of FAS transcription by cAMP is the interaction of an inverted CCAAT box motif with a constitutively produced trans-acting factor that either itself undergoes modification in response to cAMP or associates with a protein that is produced or modified by cAMP exposure.
It has long been recognized that glycogen metabolism, gluconeogenesis, glycolysis, fatty acid oxidation, and synthesis are all coordinately regulated in the liver by short-term regulatory mechanisms involving primarily allosteric and covalent modulation of key enzymes (1). Critical to this short-term regulation are the opposing actions of insulin and glucagon. Glucagon, via its intracellular messenger, cAMP, activates protein kinases involved in the phosphorylation of key enzymes and transforms the liver from a glycogenic, glycolytic, lipogenic tissue to a glycogenolytic, gluconeogenic, and fatty acid-oxidizing tissue. Initially, when the food supply is stopped, serum insulin concentration falls and glucagon concentration rises, triggering this short-term regulatory mechanism. On refeeding, insulin concentration rises, glucagon falls, and these changes are reversed, and the liver reverts to its role as a glycogenic, glycolytic, lipogenic tissue. It seems likely that antagonism between insulin and glucagon may also serve in long-term regulation of these pathways by controlling transcription of key genes (2).
For example, on prolonged fasting, cAMP may eventually initiate the down-regulation of transcription of lipogenic enzymes. Indeed, it has been shown that injection of dibutyryl cAMP will prevent the activation of FAS 1 transcription in the liver that normally accompanies refeeding (3).
All of the long-term changes in lipogenesis that occur in response to dietary, hormonal, and developmental cues appear to be accompanied by changes in the rate of transcription of the FAS gene in a tissue-specific manner. Elucidation of the sequence of the entire rat FAS gene, including 6.1 kilobases of the 5Ј-flanking region (4,5), has made possible detailed exploration of the mechanisms regulating transcription of this gene, and, recently, sequences required for mediating the effect of insulin have been identified (6,7). The objective of this study was to identify sequences in the FAS gene that are essential for mediation of the cAMP effect on transcription. Because cAMP has been reported to antagonize the insulin effect on transcription of the FAS gene in hepatocytes, we identified and utilized a hepatoma cell line that appears to be good model system for this physiological process. Our strategy was to utilize 5Ј-deletion mutagenesis and replacement mutagenesis of chimeric FAS-reporter genes to narrow down the location of the sequences required for cAMP antagonism of insulin action and to look for trans-acting nuclear factors that might bind to the putative response element. The results of this study are presented in the following report.
Cell Cultures-Rat H4IIE hepatoma cells, purchased from the Tissue * This work was supported by Grant DK 42636 from the National Institutes of Health. 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.
Culture Facility at the University of California, San Francisco, were cultured in minimum essential medium containing Earle's balanced salt solution and non-essential amino acids supplemented with 10% fetal bovine serum and 10% bovine calf serum. Cells were grown at 37°C under an atmosphere consisting of 5% CO 2 , 95% air. Prior to hormonal treatment, cells at 60 -70% confluency were maintained in serum-free medium for 24 h. Cells were exposed to insulin (20 nM) and cAMP (1 mM) plus theophylline (1 mM), usually for 48 h, then washed with PBS, harvested by scraping the plates with a rubber policeman, and lysed by three cycles of freezing and thawing. Cell debris was removed by centrifugation, and the supernatant was retained for analysis.
Transient Transfections-H4IIE cells were grown on 60-mm 2 dishes to 60 -70% confluency and transfected with 10 g of plasmid DNA per plate by the calcium phosphate precipitation method (8). To permit normalization for variable transfection efficiency, 2-5 g of a ␤-galactosidase expression plasmid, pCH110 (Pharmacia LKB, Uppsala, Sweden) was co-transfected together with the DNA of interest. The calcium phosphate/DNA co-precipitate was added to the cells, and transfection was allowed to proceed for 20 h. In some early experiments, 100 M chloroquine was included in the medium in an attempt to increase transfection efficiency; in these experiments, exposure of the cells was reduced to 6 h (9). At the end of the transfection period, the medium was removed and the cells were shocked for 3 min with 20% dimethyl sulfoxide in PBS. Cells were then washed three times with PBS and were fed serum-containing medium.
For hormonal treatments, the transfected cells were maintained in the serum-supplemented media for 16 h, then switched to media containing 0.5% serum for at least 12 h prior to treatment with 20 nM insulin and/or 1 mM cAMP for 48 h. Cells were harvested, resuspended in 0.2 ml of 0.25 M Tris HCl buffer (pH 7.8), lysed by freeze/thawing, and soluble extracts were prepared for ␤-galactosidase assay. The remainder of the supernatant was heated at 65°C for 10 min to destroy endogenous esterases. Heat-denatured proteins were removed by centrifugation, and portions of the soluble extract were assayed for CAT activity.
Enzyme Assays-FAS protein concentration was determined by an immunoassay procedure using a rabbit anti-rat (FAS) IgG preparation as the primary antibody (10) and authentic rat FAS as standard; all assays were performed in duplicate using at least two different concentrations of cytosol. CAT activity was measured in a radiochemical enzyme assay. Reaction mixtures contained 0.25 M Tris-HCl buffer (pH 7.8), 12 M [ 14 C]chloramphenicol and 1 mM acetyl-CoA in a final volume of 100 l. The reaction mixture was incubated at 37°C for 1-2 h and then partitioned with 500 l of ethyl acetate; solvent was removed from the organic phase in vacuo, and the reaction products were analyzed by thin layer chromatography. Product formation was quantitated either by direct phosphorimaging (Molecular Dynamics PhosphorImager SF hardware with ImageQuant software) or by densitometric scanning of autoradiograms. Activity of ␤-galactosidase was determined spectrophotometrically (11).
Preparation of Nuclear Extracts-Nuclear extract from H4IIE hepatoma cells was prepared essentially as described by Dignam et al. (12) except that the protease inhibitors leupeptin (1 g/ml) and pepstatin (1 g/ml) were used in all buffers in addition to phenylmethanesulfonyl fluoride (1 mM). Male rats of the Long-Evans strain were obtained from Simonsen Inc. (Gilroy, CA) and either fasted for 48 h or fasted and refed a fat-free, high carbohydrate diet ("Fat Free Test Diet," ICN Biochemicals) for 48 h. Nuclear extracts were also prepared from livers of fasted, fasted-refed rats and from spleens of fasted-refed rats as described by Gorski et al. (13). Protein concentration was determined using a Bio-Rad Protein Assay Kit (Bio-Rad).
DNase I Footprinting Assays-Various amounts of nuclear extract, or bovine serum albumin, were preincubated with 20 -60 g/ml doublestranded poly(dI-dC) at 0°C for 20 min in 20 mM Hepes (pH 7.9), 60 mM KCl, 1 mM dithiothreitol, 0.5 mM EDTA, and 5% glycerol contained in a final volume of 50 l. The end-labeled probes were then added, and the mixture was incubated for an additional 40 min at 0°C. An equal volume of a solution containing 10 mM MgCl 2 and 10 mM CaCl 2 was added followed by 0.2 unit of DNase I (Promega). The DNA was digested for 3 min at 20°C, and then the reactions were stopped by the addition of 300 l of a mixture consisting of 300 mM NaCl, 25 mM EDTA, 1% sodium dodecyl sulfate, and 40 g/ml E. coli tRNA. Proteins were removed by extraction with phenol:chloroform (1:1, v/v), and the DNA was then precipitated with ethanol and analyzed by electrophoresis on 5% polyacrylamide gels containing 7 M urea. Gels were dried and exposed to x-ray film.
Construction of Plasmids Containing a Series of Chimeric FAS-Reporter Genes-Plasmids containing various FAS-CAT fusion genes were generated by engineering a series of nested deletions from the FAS portion of the DNA using exonuclease III and mung bean nuclease (14). Two different plasmids served as substrates for the deletion constructions. The first plasmid carried FAS sequences from Ϫ1604 to ϩ656 (where ϩ1 is the start of transcription) cloned as a PstI-BamHI fragment into the PstI-XbaI sites of the pCAT-Basic plasmid (Promega). The second plasmid contained FAS sequences from Ϫ450 to ϩ68 cloned in pCAT-Basic as a Sau3AI fragment. The resulting deletion constructs were sized by agarose gel electrophoresis, and the end points of the deletions were determined by nucleotide sequencing of the fusion junctions. Subsequently, the plasmids ending at ϩ656 (generated from the first series of deletions) were subcloned such that nucleotide ϩ68 of the FAS sequence was positioned at the FAS-CAT junction of all chimeric genes.
Construction of Plasmid Ϫ249/ϩ68 ⌬Ϫ75/Ϫ35 FAS-CAT--A plasmid carrying FAS sequences from Ϫ249 to ϩ1272 cloned in pCAT-Basic (Promega) served as the substrate for the deletion construction as follows. The plasmid was digested with HindIII (cleaves the vector just upstream of Ϫ249) and KpnI (cleaves at FAS ϩ400), the resulting ϳ670-bp fragment was isolated and subsequently digested with AluI (cleavage site at FAS Ϫ75). At the same time, another plasmid containing FAS sequences Ϫ150 to ϩ68 in the same vector was digested with SacII (cleaves at FAS Ϫ35). The site was made blunt-ended by treatment with mung bean nuclease, and the plasmid was then cut with HindIII. The large fragment (ϳ4600 base pairs) was ligated to the AluI digestion products of the first plasmid resulting in plasmid Ϫ249/ϩ68 ⌬ Ϫ75/Ϫ35 .

Construction of the FAS-CAT Fusion Plasmid Carrying a Mutation in the Inverted CCAAT Box-Nucleotides
Ϫ99 to Ϫ92 (CATTGGCC) were changed to GTCCAAGG in the FAS-CAT fusion plasmids by in vitro site-directed mutagenesis (15) using the primer 5Ј-CGCAGCC-CCGACGCTGTCCAAGGTGGGCGGCGCAGCC-3Ј.
Construction of Ϫ124/Ϫ30 FAS-TK Plasmid-Plasmid Ϫ124/ϩ68 was digested with HindIII and DraI, and the HindIII site was bluntended by treatment with the Klenow fragment of DNA polymerase I. This fragment was then ligated to pTK-CAT which had been digested with XbaI and similarly treated with the Klenow enzyme to give plasmid Ϫ124/Ϫ30 FAS-TK. The orientation of the FAS insert was determined by DNA nucleotide sequencing.

Characterization of a Hormonally Sensitive Hepatoma Cell
Line as a Hepatocyte Model for Studying Transcriptional Regulation of the FAS Gene-Our first priority was to identify a cell line that would exhibit the modulated transcription of the endogenous FAS gene in response to insulin and cAMP that is typical of normal hepatocytes. We chose a rat hepatoma cell line, H4IIE, that previously has been used effectively for studying the regulation of phosphoenolpyruvate carboxykinase transcription by cAMP and insulin (16,17). These cells normally are cultured in the presence of 20% serum, and, under these conditions, the introduction of either insulin or cAMP to the medium had little affect on the content of endogenous FAS (Table I). However, when the cells were preconditioned to growth in a medium containing only 0.5% serum, addition of 20 nM insulin resulted in a substantial increase in the amount of FAS protein. Thus, the muted insulin effect observed in the presence of 20% serum likely resulted from the masking of the response by an unidentified agent(s) present in the serum. The maximum response to insulin, amounting to a 2.6-fold increase in FAS protein, was observed after 48 h. This response time is cAMP Regulation of FAS Gene Expression similar to that observed in vivo in the livers of rats that are fed a high carbohydrate diet following a period of fasting (18).
Cyclic AMP had no effect on the amount of endogenous FAS present in cells grown in medium containing either 20% or 0.5% serum. Nevertheless, in the presence of cAMP, insulin failed to induce an increase in the endogenous FAS content of H4IIE cells (Table I). Again, this result is similar to that observed in vivo, since administration of dibutyryl cAMP has been shown to block the insulin-dependent induction of FAS expression in the liver that is normally seen following fasting and refeeding (3).
The FAS content of H4IIE cells (ϳ0.7% of the cytosolic protein, by weight) cultured in the presence of insulin (Table I) is considerably less than that of normal liver (ϳ6% of the cytosolic protein, by weight) derived from fasted rats that have been refed a high carbohydrate diet (19). Nevertheless, the observation that H4IIE cells modulate their FAS content in response to insulin and cAMP in a manner similar to that exhibited by normal hepatocytes encouraged us to adopt this cell line as a model system for identifying hormone response elements within the FAS gene.
Preliminary Localization of the cAMP Response Element by 5Ј Deletion Analysis-Our objective was to identify sequences in the 5Ј-flanking region of the rat FAS that confer cAMP responsiveness to transcription of the gene. To this end, we engineered a series of plasmids each containing a promoterless CAT gene placed under the transcriptional control of different upstream sequences from the FAS gene (see "Experimental Procedures"). All of these constructs have a common 3Ј end at ϩ68 (where ϩ1 is the transcriptional start site for FAS), but they carry sequences of different lengths, up to 1604 nucleotides, derived from the 5Ј-flanking region of the FAS gene. These plasmids were transfected into H4IIE cells, and their ability to confer hormonally modulated CAT expression was assayed in the presence of 0.5% serum. Expression of CAT was amplified from 2.6-to 4.1-fold in response to insulin with all of the chimeric genes except that containing only 62 nucleotides of upstream sequence from the FAS gene (Fig. 1); this observation was consistent with the finding by Moustaïd et al. (6) that the insulin response element is located between nucleotides Ϫ68 and Ϫ52.
The addition of cAMP alone to H4IIE cells transfected with various chimeric reporter genes and cultured in the presence of 0.5% serum had little effect on CAT expression. However, in the presence of cAMP, the ability of insulin to stimulate CAT expression was compromised. This muting of the insulin effect by cAMP was evident with all deletion constructs that exhibited an insulin effect (Fig. 1), suggesting that the sequence element responsible for sensitivity to cAMP is located between nucleotides Ϫ149 and ϩ68 of the FAS gene. Similar results were obtained using two different cAMP analogs, 8-bromoadenosine 3Ј,5Ј-cyclic monophosphate and 8-(4-chlorophenylthio)adenosine 3Ј,5Ј-cyclic monophosphate.
Evaluation of Candidate cAMP Response Elements within the Ϫ149 and ϩ68 Region of the Gene-The consensus cAMPresponsive element (CRE) sequence, a well conserved 8-base pair palindrome TGACGTCA (20) is absent from the FAS promoter region. The promoter region does not contain the consensus sequence for AP-1 (21), another well characterized factor that has been implicated as a mediator of the cAMP effect on transcription of some genes. Within the proximal promoter region of the FAS gene, at position Ϫ257/Ϫ250 is a stretch of eight nucleotides that perfectly matches the consensus AP-2 binding site CC(G/C)C(A/G)GGC (22). This factor is believed to be involved in the cAMP-modulated transcription of a number of genes, including acetyl-CoA carboxylase (23). However, since this element is absent from the Ϫ249/ϩ68 and Ϫ149/ϩ68 FAS-CAT chimeric reporter genes that exhibit sensitivity to cAMP (Fig. 1) and is not footprinted by nuclear extracts from H4IIE cells (details not shown), it is unlikely to be required for tran-  cAMP Regulation of FAS Gene Expression scriptional regulation by cAMP. Therefore, we focused our attention on the the sequences downstream of nt Ϫ149 using DNase I footprinting to identify sequence elements capable of binding nuclear proteins. Two regions were found to be protected from DNase I digestion by nuclear extracts from livers of fasted and fasted-refed rats, one between nt Ϫ74/Ϫ52, the other between Ϫ100/Ϫ90 (Fig. 2). These two footprinted regions were then separately mutated in the context of the Ϫ249/ ϩ68 FAS-CAT chimeric gene, and the effect on hormone responsiveness was examined, again using H4IIE cells. When the Ϫ74/Ϫ52 footprinted region was removed from the Ϫ249/ϩ68 FAS-CAT chimeric gene (in the construct Ϫ249/ ϩ68 ⌬Ϫ75/Ϫ35 FAS-CAT) insulin responsiveness of CAT expression was lost, and, thus, the response to cAMP could not be evaluated (details not shown). This finding was consistent with results of an earlier study that identified nt Ϫ68/Ϫ52 as essen-tial for mediating insulin responsiveness of the FAS gene (6).
The second footprinted region, which contains an inverted CCAAT box motif at nt Ϫ99/Ϫ92, was evaluated by replacement mutagenesis in the context of the Ϫ249/ϩ68 FAS-CAT construct. Although mutation of the inverted CCAAT box had no effect on the ability of the chimeric gene to direct the insulinsensitive expression of CAT, responsiveness to cAMP was completely abolished (Fig. 3). The results of this experiment indicated that the insulin and cAMP effects on transcription of the FAS gene are mediated by different regions of the promoter and strongly implicated the inverted CCAAT box in the regulation by cAMP. To verify that the inverted CCAAT box motif specifically was responsible for the nuclear protein binding observed in the DNase I protection assay, we employed the electrophoretic mobility shift assay using a radiolabeled probe corresponding to nucleotides Ϫ106 to Ϫ85 (Fig. 4). Liver nuclear extracts from fasted and fasted/refed rats and spleen nuclear extracts from fasted/refed rats retarded the mobility of the probe to the same extent (slow migrating band, lanes 2, 8, and 5, respectively). Formation of these radiolabeled DNA-protein complexes was abolished in the presence of 100-fold molar excess of an unlabeled oligonucleotide corresponding exactly to nucleotides Ϫ106/Ϫ85 (lanes 4, 10, and 7), but not when the sequence between Ϫ99/Ϫ92 of the unlabeled probe was mutated (lanes 3, 9, and 6). When a radiolabeled Ϫ106/Ϫ85 probe containing the same mutated sequence between Ϫ92/Ϫ99 was used in the electrophoretic mobility shift assay in the presence of nuclear extracts from either liver or spleen, DNA-protein interactions were not observed (lanes 12-14). Faster migrating radiolabeled DNA-protein complexes were also formed with nuclear extracts from liver and spleen. However, formation of these radiolabeled complexes was reduced in the presence of unlabeled oligonucleotides containing either the intact or the mutated inverted CCAAT box sequence (lanes 2 and 3, 6 and 7, 9 and 10), and these complexes were formed from the radiolabeled mutated oligonucleotide probe (lanes 12-14). It would appear therefore that formation of these fast migrating complexes is not specifically dependent on the presence of the inverted CCAAT box sequence. However, from the properties of the slow migrating DNA-protein complex, one could conclude unequivocally that rat liver and spleen contain nuclear protein(s) that specifically interact with the inverted CCAAT box located at nt Ϫ99/Ϫ92 of FAS promoter.
The possibility that the trans-acting factor binding to the inverted CCAAT box might be a member of the well character-FIG. 2. DNase I footprinting analysis of the FAS promoter region in the presence of nuclear extracts from rat liver. An endlabeled DNA fragment comprising FAS promoter sequences from nucleotides Ϫ190 to Ϫ23 was digested with DNase I in the presence or absence of a liver nuclear extract from either fasted or fasted-refed rats, and the products were fractionated by electrophoresis on denaturing polyacrylamide gels. The regions of the DNA probe protected from digestion are indicated by a box in the margin. Lane 1, G ϩ A reaction; lanes 2 and 5, bovine serum albumin included, no nuclear extract; lane 3, 50 g of liver nuclear proteins from fasted rats; lane 4, 50 g of liver nuclear proteins from fasted-refed rats. cAMP Regulation of FAS Gene Expression ized family of C/EBP transcription factors was evaluated and discounted because of the following evidence. Firstly, neither the electrophoretic mobility nor the amount of the DNA-protein complex formed between liver nuclear extract and a radiolabeled oligonucleotide probe representing the inverted CCAAT box was affected in the presence of either authentic anti-C/ EBP␣ antibodies or an unlabeled oligonucleotide representing the consensus binding site for C/EBP␣ (Fig. 5, lanes 1-4). Secondly, formation of the DNA-protein complex between liver nuclear extract and a radiolabeled oligonucleotide probe representing the inverted CCAAT box was prevented when the extract was preheated for 10 min at 65°C, conditions under which the C/EBP proteins are stable. (Fig. 5, lanes 2 and 7). Thirdly, authentic C/EBP␣, expressed as a recombinant protein in E. coli, did not bind to a probe containing the FASinverted CCAAT box (Fig. 5, lane 8), although it did bind to an oligonucleotide containing the consensus binding site for C/EBP␣; the electrophoretic mobility of this complex was supershifted by anti-C/EBP␣ antibodies (Fig. 5, lanes 10 and 11), and association of the radiolabeled probe with the complex was effectively competed out with unlabeled C/EBP consensus probe (Fig. 5, lanes 10 and 12). Fourthly, the amount of DNAprotein complex formed between recombinant C/EBP␣ and an oligonucleotide containing the consensus binding site for C/EBP was unaffected by the presence of a large excess of unlabeled probe containing the FAS-inverted CCAAT box (Fig.  5, compare lanes 10 and 13). The results of these experiments indicate that it is highly unlikely that the cAMP effect on transcription of the FAS gene is mediated by a member of the C/EBP family of transcription factors.
The inverted CCAAT box motif is quite dissimilar from the consensus cAMP-responsive element (CRE), TGACCTCA, and an unlabeled oligonucleotide containing the CRE consensus sequence does not compete with the labeled inverted CCAAT box probe for binding to the liver nuclear protein (Fig. 5, lanes  2 and 5); neither does the labeled inverted CCAAT box probe bind to purified CREB protein (details not shown). Based on this evidence, CREB protein does not appear to be involved in mediating the cAMP effect.
Evaluation of Effectiveness of Putative Hormone Response Elements in the Context of a Heterologous Promoter-Finally, to determine whether cAMP antagonism of the insulin response in the context of the FAS promoter could be transferred to a heterologous promoter, we constructed a chimeric CAT reporter gene containing nucleotides Ϫ124 to Ϫ30 of the FAS gene positioned in front of the thymidine kinase promoter. Expression of chloramphenicol acetyltransferase in H4IIE cells transfected with this construct was stimulated 2-fold by insulin, but no antagonism by cAMP was observed (details not shown). This result indicates that, although the sequence located between nt Ϫ68 and Ϫ52 of the FAS gene can confer insulin responsiveness to a heterologous promoter, the inverted CCAAT box located at nt Ϫ99 to Ϫ92 is itself insufficient to confer cAMP responsiveness out of its normal promoter context. FIG. 4. Demonstration that nuclear proteins bind to the the inverted CCAAT box region of the FAS gene. Electrophoretic mobility shift assays were performed using nuclear proteins from livers (L) or spleens (S) of fasted (F) or fasted-refed (R) rats. The end-labeled DNA probes comprising FAS promoter sequences from nucleotides Ϫ106 to Ϫ85 were either identical with the wild type sequence (Wt) or contained mutations (mu) in the region Ϫ92 to Ϫ99 (see Fig. 3). Some incubations also included 100-fold molar excess of unlabeled competitive oligonucleotides, representing either the wild type (Wt) sequence Ϫ106 to Ϫ85 or the mutated (mu) sequence.

FIG. 5. C/EBP␣ is not the protein involved in mediating the cAMP effect via the inverted CCAAT box.
Two radiolabeled oligonucleotide probes were employed, one containing the FAS inverted CCAAT box, FAS nt Ϫ106 to Ϫ85 (lanes 1-8), the other containing the consensus binding site for C/EBP (lanes 9 -13). Protein extracts were prepared either from liver nuclei of fasted or fasted/refed rats or from E. coli cells that had been transfected with a C/EBP␣ expression vector. Where indicated, a polyclonal antiserum raised against C/EBP␣ was included for "supershifting." In some incubations, unlabeled oligonucleotides representing either the consensus C/EBP binding site, the FAS inverted CCAAT box, or the cAMP response element, CRE, were included at 100-fold molar excess as potential competitors for the ra-

DISCUSSION
In response to insulin and cAMP, the hepatoma cell line H4IIE modulates its endogenous FAS concentration in a manner qualitatively similar to that exhibited by normal liver cells. Thus, the elevation of endogenous FAS protein concentration induced by exposure to insulin can be blocked by simultaneous exposure to cAMP. Using these cells as a model system, we transfected chimeric reporter genes containing various 5Јflanking sequences from the FAS gene and demonstrated that expression of the reporter gene could be enhanced by insulin and that this response could be antagonized by cAMP. In the course of this study, using 5Ј-deletion analysis of the FASreporter gene, we initially localized the insulin response element between nt Ϫ124 and Ϫ30 (details not shown) and subsequently identified within this region a sequence element between Ϫ74/Ϫ52 that was protected from DNase I action in the presence of H4IIE nuclear extract. Deletion of the footprinted region from the chimeric FAS-reporter gene resulted in the loss of insulin responsiveness indicating that this region of the promoter is required for mediating the insulin response. While these studies were in progress, Moustaïd et al. (6) independently reported that the insulin response element of the FAS gene is located between nucleotides Ϫ68 and Ϫ52. Thus, the results of our studies are in complete agreement with their findings.
By examining the effect of deletion and replacement mutagenesis on the ability of cAMP to antagonize insulin action, we were able to identify a sequence element between nt Ϫ99/ Ϫ92 that is essential for cAMP action. This region was also found to be essential for the binding of a nuclear factor present in the livers and spleens of fasted and fasted-refed rats. The electrophoretic mobilities of the DNA-protein complexes formed with these various nuclear extracts were indistinguishable. Since spleen does not express detectable levels of FAS (4) and expression in the liver is elevated on feeding but drastically lowered on fasting, there is clearly no correlation between the presence in a particular tissue of a nuclear protein binding to nt Ϫ99/Ϫ92 and the level of FAS expression. More likely then, this protein is a constitutive transcription factor that either itself undergoes covalent modification as a result of exposure of the cells to cAMP or associates with a protein that is either expressed or modified following cAMP treatment. Mutational disruption of the inverted CCAAT box also decreases activity of the FAS promoter approximately 2-fold, in the absence of insulin or cAMP (Fig. 3), indicating that the constitutive transcription factor binding to the inverted CCAAT box plays a role in supporting basal, as well as cAMP-modulated, transcription.
When placed in front of a heterologous promoter, nt Ϫ124 to Ϫ30 of the FAS gene confer responsiveness to insulin but not to cAMP. This observation indicates that the context of the FAS gene is important in mediating the cAMP effect and raises the possibility that a protein binding to another sequence element within nt Ϫ124 and ϩ68 may also be required for mediating the cAMP effect, perhaps through interaction with the protein binding to the inverted CCAAT box.
The FAS appears to be one of a small group of genes transcriptionally regulated by cAMP that bear cAMP-responsive sequences distinct from the more common CREB, AP-1, and AP-2 binding sites. Other genes that fall into this category include the mouse renin gene (24), the human myelin basic protein gene (25), the bovine steroid hydroxylase (P-450 c21 ) gene (26), the rat c-fos gene (27), the porcine G-protein ␣ i-2 subunit gene (28), and the human tryptophan hydroxylase gene (29). In all of these other examples, cAMP is an inducer of transcription, whereas in the case of the hepatic FAS it is a negative regulator. In several of these genes, CCAAT boxbinding proteins have been implicated as mediators of the cAMP-regulated transcription, notably the c-fos gene which is regulated by the phosphorylated form of C/EBP␣ (27), the G-protein ␣ i-2 subunit which is regulated by an unidentified CCAAT box-binding protein and the human tryptophan hydroxylase gene which, like the FAS gene, has an inverted CCAAT box that is essential for mediating the cAMP effect on transcription. In the case of the FAS, it appears highly unlikely that the protein binding to the inverted CCAAT box is a member of the C/EBP family of transcriptional factors, based on its heat lability and immunochemical properties. Clearly, the purification and characterization of the transcription factor(s) binding to the inverted CCAAT box motif would provide further insight into the mechanism by which cAMP regulates transcription of the FAS gene.