Contribution of a nuclear factor 1 binding site to the glucocorticoid regulation of the cytosolic aspartate aminotransferase gene promoter.

Two regions of the cAspAT gene promoter mediate the glucocorticoid regulation of this gene in the Fao hepatoma cell line. The proximal region was localized by deletion studies and stable transfections in the Fao cells to the sequence −553/−398. This region includes the glucocorticoid-responsive element (GRE) A sequence, which consists of two overlapping GREs and which can mediate the glucocorticoid regulation of a heterologous promoter. DNase I footprinting studies have shown that a site 80 base pairs upstream of the GRE A was protected by liver and brain nuclear extracts (site P8). The binding was displaced by an excess of an oligonucleotide containing a typical NF1 binding site and by NF1-specific antibodies. In electrophoretic mobility shift assay using the P8 oligonucleotide as a probe, several complexes were formed. Most complexes were common to liver and brain but were less abundant when testis extracts were used. At least one complex was specific to the liver. All complexes, with the exception of two, were competed for by the NF1 oligonucleotide. Furthermore, the sequence of the P8 site showed a 7/9-base pair homology with a typical NF1 site. A mutation of the P8 site, which prevents the binding of NF1-like proteins to it, considerably decreases the regulation of the cAspAT promoter fragment by glucocorticoids. Surprisingly, the basal activity of the mutant promoter was increased 2-fold. Thus, the regulation of the cAspAT gene promoter is mediated by a regulatory unit comprising the GRE A and a NF1 binding site.

The regulation of gene expression by glucocorticoid hormones is mediated by an intracellular receptor which displays both DNA binding and transactivation properties (1). The receptor interacts usually as a dimer on partially palindromic sites called glucocorticoid-responsive elements (GRE). 1 However, in some gene promoters, GREs are not sufficient to convey hormonal regulation. In these cases, other binding sites for transcription factors, usually located in the vicinity of a GRE, are also required, and constitute, with the GRE, a so-called glucocorticoid-responsive unit, or hormone response domain (2). Following initial observations on the tryptophan oxygenase gene promoter (3) and the mouse mammary tumor virus long terminal repeat promoters (4), several studies have established that glucocorticoid regulation of artificial promoters was strongly enhanced when a binding site for one of several transcription factors was present in the vicinity of a GRE (5,6). These sites were either the CACC sequence, the nuclear factor 1 site, the octamer sequence, the Sp1 element, or a duplication of the GRE itself. The analysis of natural promoters showed that several sites are usually required for efficient regulation. In addition to the mouse mammary tumor virus and the tryptophan oxygenase promoters, the glucocorticoid-responsive units of the tyrosine aminotransferase (7) and of the phosphoenolpyruvate carboxykinase promoters (8) have been extensively analyzed. In the latter case, even two closely located GREs are not sufficient to confer an efficient regulation which requires two additional sites for accessory factors. Furthermore, the phosphoenolpyruvate carboxykinase glucocorticoid responsive unit cooperates with proximal promoter elements to mediate an efficient regulation (9). Clearly, the mechanism of gene regulation by glucocorticoids is complex, and additional studies on different gene promoters are required to allow a better understanding of this biological process.
We are studying the hormonal regulation of the cytosolic aspartate aminotransferase gene promoter. This gene is different from other models for glucocorticoid hormone action in that it is a housekeeping gene characterized by a widespread basal expression and a tissue-specific glucocorticoid regulation (10). The cAspAT gene basal promoter binds transcription factors of the C/EBP family and the NF 1 family (11). The binding pattern differs according to the tissue, which presumably provides the possibility for tissue-specific regulation. Two regions of the cAspAT gene promoter contribute to the glucocorticoid regulation: a distal one (Ϫ1.983/Ϫ1.718 kilobases) which also carries the negative regulation by insulin, and a proximal one (Ϫ553/ Ϫ318 base pairs) (12). The latter one contains an unusual glucocorticoid-responsive sequence consisting of two overlapping GREs that can bind two receptor dimers in a highly cooperative manner (13). Because of this unusual structure and because of the complexity of the cAspAT gene promoter, it was of interest to search for other factors that can cooperate with the glucocorticoid receptor. Here we show that proteins of the NF 1 family bind 80 base pairs upstream of the GREA and that this binding is necessary for an efficient glucocorticoid activation of the cAspAT gene promoter.

Plasmids
The construction of plasmids p(Ϫ553,Ϫ26) CAT and p(Ϫ398,Ϫ26) CAT was described by Garlatti et al. (11). The rat glucocorticoid-receptor expression vector RSV-GR (17) was a generous gift of Dr. Keith Yamamoto (San Francisco). The mutation of the P 8 site was performed by the double polymerase chain reaction method as described previously by Aggerbeck et al. (12). The sequence of the mutated plasmid was verified.

Transfection Experiments
Stable Transfections-One day prior to the transfection, Fao cells (1.5 ϫ 10 6 cells/10-cm dish) were seeded into the usual culture medium containing fetal calf serum (18). Ten milliliters of fresh medium with serum were added to the cells 2-3 h before the transfection. The CAT plasmid (10 g) and the pSV2neo plasmid (2 g) were introduced into the cells by the calcium phosphate co-precipitation technique followed by a glycerol shock (12). Ten ml of fresh medium with serum were added after the glycerol shock. Two days later, the cells were split 1:5 and allowed to grow for 24 h prior to the addition of the neomycin analog G418 (Life Technologies, Inc.: 250 -500 g/ml, depending on the batch). The medium was changed every 3 days. Two to three weeks later, the surviving cells were harvested and pooled for CAT assay. When needed, the cells were treated for 24 h with the relevant hormones.
Transient Transfection-HepG2 cells were transfected as described by Garlatti et al. (11) with some modifications. Five micrograms of the reporter plasmid were co-transfected with varying amounts of a glucocorticoid receptor expression vector and with 1 g of pSV 2 luc (Promega) to correct the variability of transfection efficiency. Luciferase was assayed using a kit from Promega according to the manufacturer's instructions. Data were expressed as the ratio of CAT activity over luciferase activity. Dexamethasone (0.1 M) was added 24 h later when required. After a 24-h treatment, cells were homogenized for CAT assay.

Chloramphenicol Acetyltransferase Assay
CAT activity was determined using the two-phase assay developed by Neumann et al. (19). Briefly, cell homogenates were prepared in 280 l of Tris-HCl, 250 mM, pH 7.8, EDTA, 5 mM, as described previously. Cell extracts (80 l) were treated at 65°C for 5 min, in order to inactivate the endogenous acylases, then centrifuged at 12,000 rpm for 15 min. For each reaction, 20 l of the cell extract supernatant were added to 40 l of a buffer solution in order to have the following final concentrations: 250 mM Tris-HCl, pH 7.5, 5 mM EDTA, 1 mM chloramphenicol, 30 M acetyl-CoA to which 0.5 Ci of 3 H-labeled acetyl-CoA were added (DuPont NEN, NET-290 L). The reaction was carried out for 30 min at 37°C. The solution was then transferred to a minivial and layered with 4 ml of Econofluor (DuPont NEN, NEF 969). After mixing vigorously, the two phases were allowed to separate for at least 15 min, and radioactivity was counted in a scintillation counter. Under these conditions, the product of the reaction, acetylated chloramphenicol, but not the unreacted acetyl-CoA, is allowed to diffuse into the Econofluor phase. In these experiments, blanks were obtained by assaying CAT activity in cells that have undergone the same treatment in the absence of a CAT plasmid. Proteins were assayed according to Bradford (20).

Preparation of Nuclear Extracts
Male Wistar rats, weighing 200 -250 g, were used. Extracts from rat organs (liver, testis, and brain) were prepared as described by Gorski et al. (21).

DNase I Footprinting
Two probes from the cAspAT promoter (nucleotides Ϫ682 to Ϫ226) were end-labeled using the Klenow fragment of DNA polymerase I. The standard reaction was performed according to Vaulont et al. (22) with some modifications. The binding reaction was performed in a final volume of 25 l, containing 50 mM NaCl, 50 mM KCl, 0.1 mM EDTA, 5 mM MgCl 2 , 2 mM dithiothreitol, 4 mM spermidine, 15% glycerol, 100 g/ml bovine serum albumin, 10 mM HEPES, pH 8. Two hundred fifty ng of poly(dI-dC) (Pharmacia) were used as carrier.
The nuclear proteins, 15-80 g, were preincubated 15 min on ice with the competitor oligonucleotides. Then, about 1 ng of labeled probe (20,000 cpm) was added, and the incubation was continued for 15 min on ice. After adjusting the concentration of CaCl 2 to 2.5 mM and incubating for 1 min at 20°C, DNase I was added, and the digestion was carried out at 20°C for 1 min. Subsequent handling of the DNA was performed as described (22).

Gel Retardation Assays
Probes were oligonucleotides labeled using the Klenow fragment of DNA polymerase I. Protein-DNA binding was performed under the same conditions as those described for the footprinting experiments except that 1 g of poly(dI-dC) was used as carrier and that 1.5-7 g of nuclear extracts were added to the probe. After 15 min on ice, the samples were directly loaded onto a 6% polyacrylamide gel in 0.5 ϫ TBE. The gel (0.2 ϫ 16 cm) was pre-electrophoresed at 100 V for 1 h in the cold room (4°C). Electrophoresis was performed at 260 V for 90 min in 0.5 ϫ TBE buffer. For supershift experiments, IgG were purified from either preimmune serum or anti-NF 1 immune serum, using a HiTrap protein A column according to the manufacturer's instructions (Pharmacia). The preimmune or specific anti-NF 1 IgG (2 g) were preincubated with rat liver nuclear extracts (2 g) for 90 min at room temperature before addition of the probe.

RESULTS
In a first series of experiments, we have mapped the proximal region mediating the activation of the cAspAT gene promoter by glucocorticoids. Deletion fragments of the cAspAT gene promoter were subcloned upstream of the CAT gene and the resulting plasmids were stably transfected into the Fao cells. We have previously shown that glucocorticoids stimulated promoter activity approximately 3-fold in cells transfected with the construct (Ϫ682,Ϫ26) (12). Similar experiments were performed with plasmids carrying promoter fragments that were further deleted at the 5Ј end. The results depicted in Fig. 1 show that the fragment Ϫ553/Ϫ398 is required for glucocorticoid activation of the promoter. The glucocorticoid effect is even more striking in the presence of cAMP, which is known to potentiate the induction of cAspAT by glucocorticoids (12). This fragment includes a glucocorticoid-responsive sequence (called GRE A) that we have previously shown to be functional when inserted into a heterologous promoter (13). The data suggest that the half-palindromic GRE present at Ϫ380 is not sufficient to confer glucocorticoid regulation. We then asked whether other putative transcription factor binding sites in the vicinity of the GRE A could contribute to the glucocorticoid induction of the promoter activity.
DNase I footprinting analysis was carried out to identify the binding sites involved in the glucocorticoid regulation. A DNA fragment spanning the region Ϫ682/Ϫ226 was labeled at either strand, and allowed to bind to liver nuclear proteins. In the experiment depicted in Fig. 2, four sites were protected from DNase I digestion. These sites were called P 6 , P 7 , P 8 , and P 9 . Two more proximal sites P 5 and P 4 are outside the portion of the gel that is shown here. Site P 4 had been described in detail during the study of the promoter fragment carrying the basal activity (11). The closest site to the GRE A is P 8 which is included in the promoter fragment Ϫ553/Ϫ398. Thus, we have focused this study on the P 8 site.
As shown in Table I, the DNA sequence corresponding to the P 8 site displays an identity of 7/9 bases with the consensus NF 1 binding sequence (23). Furthermore the P 8 sequence is partially palindromic and the distance between the two half-sites is identical to that of a typical NF 1 -binding site; addition of an excess of the NF 1 oligonucleotides prevented the protection observed at the P 8 site (Fig. 2).
A similar experiment was conducted using a fragment la-beled on the noncoding strand at the distal end (Fig. 3). In this case, the P 8 and P 9 sites can be clearly detected when liver extracts were used. Addition of a P 8 oligonucleotide (which covers the protected region) prevented protection at the P 8 site, but not at the P 9 site. A similar result was obtained with an NF 1 oligonucleotide confirming the data of Fig. 2. Other oligonucleotides including Sp 1 or CACC sites were not able to compete efficiently. These data suggest that a NF 1 -like protein binds at the P 8 site. When brain nuclear extracts were used instead of liver extracts, similar results were obtained for the P 8 site, with a similar pattern of competition. However, there was no protection at the P 9 site suggesting that binding at this site may be tissue-specific. In fact, when testis extracts were used, the P 9 site was also protected, while the P 8 site was only partially protected (not shown).
An electrophoretic mobility shift assay was used to further characterize the binding at the P 8 site. Fig. 4 shows the data obtained with liver, brain, and testis extracts using as a probe either the P 8 oligonucleotide (Fig. 4A) or an oligonucleotide containing a typical NF 1 site (Fig. 4B) (origin of replication of the adenovirus). Both probes gave very similar patterns of retarded complexes. A strong binding was observed with liver and brain extracts, but not with testis extracts, in agreement with the footprinting experiments. Several complexes were

TGCG TGTCA GCCCG
formed in each case, an expected observation since there are several NF 1 or NF 1 -like proteins in nuclear extracts. Complexes a, b, and c are the most abundant and display the slowest migration in the gel. These complexes were displaced by an excess of P 8 or NF 1 oligonucleotides. However, the NF 1 oligonucleotide was more efficient suggesting that it could bind NF 1 or NF 1 -like proteins with a higher affinity than P 8 . Note that one of the half-sites in the P 8 sequence is not optimal for NF 1 binding. Complexes e and f appear to be specific for the P 8 probe (Fig. 4, A and B). As opposed to other complexes, they are poorly competed for by the NF 1 oligonucleotide. The nature of these complexes is not known. However, it is unlikely that they play a significant role as they can only be seen in EMSA, not in DNase I protection experiments (absence of a footprint on the P 8 site in the presence of the NF 1 oligonucleotide). Furthermore, these complexes are clearly detected only when a high amount of nuclear extracts (7 g) is added to the probe; this suggests that the abundance of the corresponding proteins is poor or that their affinity for the site is low. One interesting observation is that, although most complexes formed by liver and brain nuclear extracts are similar in migration and abundance, complex c appears to be more abundant in binding reactions with the liver extracts. These complexes were further analyzed using a supershift assay with polyclonal antibodies directed against the C-terminal half of the NF 1 protein (Fig. 4C). The anti-NF 1 IgG, but not preimmune IgG, displaced complexes b, c, and d, and to a lesser extent complex a. The other complexes were barely detected in these experiments. The results confirm that proteins of the NF 1 family of transcription factors are the major components of the complexes formed using the P8 probe.
In order to assess the role of the P 8 site in the glucocorticoid induction of the cAspAT gene promoter, a mutation was introduced in the putative NF 1 site. An oligonucleotide containing the mutated sequence (called P 8 mut) was unable to compete for the complexes formed by the P 8 probe and liver nuclear proteins (Fig. 5A). This suggests that the mutation has inactivated the NF 1 site. This was further confirmed by gel shift assays using the P 8 mut oligonucleotide as a probe (Fig. 5B). Under these conditions, there was a dramatic decrease in the intensity of the shifted bands, suggesting that the mutation did inactivate the NF 1 site, and did not generate a novel site with high affinity to liver nuclear proteins.
The P 8 mut mutation was then introduced into the sequence Ϫ553/Ϫ26 and the promoter activity of this mutated fragment was compared to that of the wild-type fragment (Table II). Plasmids containing the wild-type and mutated fragments cloned upstream of the CAT gene were co-transfected into the HepG 2 cells with an expression vector coding for the glucocorticoid receptor. The experiment depicted in Table II shows that the mutated fragment displays a higher basal activity than the wild-type fragment. Interestingly, the 3-fold glucocorticoid activation of the wild-type promoter was significantly decreased when the P 8 site was mutated. In order to assess the specificity of this phenomenon, we evaluated the effect of C/EBP␤, a well known activator of the cAspAT gene promoter (11). A C/EBP␤ expression vector was co-transfected with the wild-type or with the mutated cAspAT gene promoter. Both promoters were activated to a similar extent under these conditions (data not shown). Thus, the mutation of the P 8 site specifically reduces the effect of the activated glucocorticoid receptor. This suggests that proteins binding to this site contribute to the glucocorticoid regulation. DISCUSSION In this study, we have further characterized the complex regulation of the cAspAT gene promoter by glucocorticoids. This regulation is tissue-specific and is modulated both negatively and positively by insulin and cAMP, respectively (12). Two promoter regions are required for the glucocorticoid effect. We have focused this study on the proximal region which includes a functional site called GRE A. The GRE A displays a FIG. 4. Electrophoretic mobility shift assay. A, labeled oligonucleotide covering the P 8 site was incubated with 7 g of liver, brain, or testis nuclear extract and the reaction was subjected to EMSA. In the indicated lanes, competing oligonucleotides (50 ng) were added to the binding reactions. Specific complexes are labeled a to f to the left. B, similar experiment using a labeled oligonucleotide containing the NF 1 binding site of the adenovirus major late promoter. C, rat liver nuclear extracts (2 g) were preincubated for 90 min at room temperature in the absence (0), or in the presence of preimmune (IgG) or anti-NF 1 IgG (Ab NF 1 ) before addition of the P 8 probe. Supershifted complexes are indicated by a star to the right. unique structure composed of two overlapping imperfect GREs; it binds two dimers of the glucocorticoid receptor in a highly cooperative manner (13).
The finding that overlapping GREs could constitute a functional structure has raised several questions that remain unanswered. In fact, much is known about units composed of adjacent GREs which are found in several gene promoters that are regulated by glucocorticoids. In the case of tyrosine aminotransferase, the functional regulatory unit is composed of two GREs which act synergistically (24), but also comprises other binding sites for transcription factors such as HNF 3 and ets (25). The glucocorticoid-responsive unit of the phosphoenolpyruvate carboxykinase gene is also composed of two imperfect GREs and of two binding sites for accessory factors (8). Conversely, in some experiments using artificial promoters, two adjacent GREs coupled to a minimal promoter were found to be sufficient to confer a potent regulation by glucocorticoids (5). Thus, depending on the promoter structure, the multimerization of GREs in an adjacent manner may, or may not, be sufficient to constitute a potent regulatory unit. Examination of the structure of several promoters suggests that additional transcription factors are often required.
Since the arrangement of the GREs as overlapping elements in the cAspAT GRE A site is different from that of other sites, it was of interest to ask whether this site was part of a larger regulatory unit. We have shown here that this is indeed the case since a NF 1 binding site is critical for glucocorticoid regulation of the cAspAT gene promoter. The fact that the accessory factor in this case is NF 1 is interesting because this factor is implicated in one of the best studied examples of gene regulation by steroid hormones, namely the mouse mammary tumor virus long terminal repeat promoter. In the latter case, the integrity of the NF 1 site is required for efficient glucocor-ticoid regulation of the promoter (26,27). Lately, it has been suggested that the chromatin structure is critical to understand the mechanisms by which the glucocorticoid receptor activates transcription (28 -30). Indeed, one function of the GR is to bind to, and either disrupt or at least alter, the nucleosome structure at the mouse mammary tumor virus promoter, thus allowing NF 1 to bind to DNA and activate transcription (31). In initial studies (4), only the binding of NF 1 was detected in vivo after hormonal treatment, but other studies have detected the binding of both GR and NF 1 (32). In the case of the cAspAT gene promoter, the inactivation of the NF 1 site results not only in a decrease in the glucocorticoid effect but also in an increase in the basal activity of the promoter. Thus, the NF 1 site exerts a negative effect on promoter activity in the absence of hormone. It is possible that one function of the activated glucocorticoid receptor would be to relieve this negative effect. However, we have no direct evidence that the effect of the NF 1 site on promoter activity and its effect on the glucocorticoid regulation are linked.
Nuclear factor 1 has been shown to bind to several promoters and to collaborate with tissue-specific and regulated transcription factors. One interesting example is that of the eH site of the albumin enhancer (33). In this case, an NF 1 site and a HNF 3 site are in close apposition, but NF 1 can inhibit transcriptional activation by HNF 3 ␣. In most cases, NF 1 is a potent transcriptional activator (23,34,35). Thus, depending on the structure of the regulatory unit, NF 1 can have either positive or negative effects. The mechanism by which NF 1 exerts its negative effect on the cAspAT gene promoter, and actually allows glucocorticoid stimulation, is yet to be elucidated. It is noteworthy that the NF 1 site is not in the immediate vicinity of the GRE A but is 80 base pairs upstream. However, this distance is optimal to bring proteins binding at those sites close together if a nucleosome is positioned on this region of the promoter.
Could NF 1 contribute to the tissue specificity of the glucocorticoid regulation of the cAspAT gene? Clearly NF 1 binding activity is ubiquitous. However, in addition to the initially discovered NF 1 , several NF 1 -like proteins bind to the same site (36 -38). Some of these proteins are liver-specific as evidenced in EMSA by the presence of DNA-protein complexes specifically in this tissue (39). One possible model is that only the liver-specific NF 1 -like proteins could collaborate with the glucocorticoid receptor in the context of the cAspAT gene promoter.
The presence of different forms of NF 1 -like proteins and mRNAs has been shown in several studies. In one study, a NF 1 site was shown to be critical to the activity of the adipocyte- FIG. 5. Effect of the mutation of the P 8 site on binding to nuclear proteins. The EMSA were conducted under the same conditions as in Fig. 4A. A, competition with various oligonucleotides: a 100-fold excess of oligonucleotides containing either the NF 1 site, the P 8 site, or the mutated P 8 site was added during the binding reaction. B, EMSA using 2 g of liver nuclear extracts with either the P 8 or the P 8 mut probe.

TABLE II
Effect of the P 8 site mutation on basal and regulated activities of the cAspAT gene promoter Plasmids containing the cAspAT gene promoter upstream of the CAT gene were transiently transfected into the HepG2 cells with either 500 or 1000 ng of an expression vector for the glucocorticoid receptor. Cells were treated or not with 0.1 M dexamethasone. Plasmids were either the wild type p(553,Ϫ26) CAT or the mutant p(Ϫ553,Ϫ26) mutCAT. The mutation is described in Table I. The data represent the CAT/ luciferase activities and correspond to the average Ϯ S.E. of four independant determinations. Statistical analysis was conducted using the Mann-Whitney U test. specific enhancer of the P 2 gene (35). In a different example, the NF 1 site of the collagen gene was shown to be implicated in the transforming growth factor ␤ regulation of this gene as well as in its modulation by acetaldehyde (40). All these studies as well as ours demonstrate the contribution of NF 1 or NF 1 -like proteins to various regulatory regions in gene promoters and enhancers.