Identification of a glucocorticoid response element contributing to the constitutive expression of the rat liver alpha 1-inhibitor III gene.

alpha 1-Inhibitor III (alpha 1 I3), a broad range plasma proteinase inhibitor, is synthesized with striking tissue specificity in rat livers. The gene is expressed strongly in periportal hepatocytes of healthy adults and less abundantly in regions near the centrilobular vein. This expression pattern is suggestive of a concentration gradient of a blood-borne hormone that enters through the portal vein and diffuses across the lobe toward the centrilobular vein. The alpha 1 I3 gene was known to be regulated both by glucocorticoids and interleukin 6, and therefore the hypothesis was tested that the normal constitutive expression of this gene depended on glucocorticoids. alpha 1 I3 mRNA levels in the livers of hypophysectomized rats with low endogenous glucocorticoid levels were only about 20% of those in control rat livers. Injection of exogenous glucocorticoids reconstituted hepatic alpha 1 I3 mRNA levels up to 64% of their original values in a dose-dependent manner. Similarly, treatment of FAZA rat hepatoma cells with the synthetic glucocorticoid dexamethasone induced alpha 1 I3 mRNA levels in a dose-dependent manner. Taken together these data suggested that glucocorticoids are required for the constitutive high level expression of this gene in normal adult rat livers. A series of 5' deletion constructs and linker scanning mutants of the promoter upstream region were produced and transfected into FAZA cells. A functional glucocorticoid response element was mapped between -168 and -151 base pairs 5' of the transcription start site. This element conforms with an inverted consensus glucocorticoid response element (GRE) but differs in two positions essential for protein DNA interaction between the GRE and the glucocorticoid receptor (GR). The induction of alpha 1 I3 gene promoter region constructs by dexamethasone was abolished by the receptor antagonist RU486, indicating that the GR participated in the activation of the alpha 1 I3 gene. In DNase I footprinting experiments with nuclear protein extracts from untreated and dexamethasone-treated FAZA cells, similar extents of alpha 1 I3 promoter upstream sequences were protected, indicating that proteins capable of binding in the glucocorticoid response-mediating element (GME) region were present before and after arrival of the hormonal signal. However, a purified recombinant fragment of the GR which contained essentially only its DNA binding domain was unable to bind at the GME although it interacted strongly with a consensus GRE sequence.


Identification of a Glucocorticoid Response Element Contributing to the Constitutive Expression of the Rat Liver a,-Inhibitor I11 Gene*
(Received for publication, December 24, 1990) Lawrence J. Abraham al-Inhibitor I11 (a113), a broad range plasma proteinase inhibitor, is synthesized with striking tissue specificity in rat livers. The gene is expressed strongly in periportal hepatocytes of healthy adults and less abundantly in regions near the centrilobular vein. This expression pattern is suggestive of a concentration gradient of a blood-borne hormone that enters through the portal vein and diffuses across the lobe toward the centrilobular vein. The a113 gene was known to be regulated both by glucocorticoids and interleukin 6, and therefore the hypothesis was tested that the normal constitutive expression of this gene depended on glucocorticoids. al13 mRNA levels in the livers of hypophysectomized rats with low endogenous glucocorticoid levels were only about 20% of those in control rat livers. Injection of exogenous glucocorticoids reconstituted hepatic al13 mRNA levels up to 64% of their original values in a dose-dependent manner. Similarly, treatment of FAZA rat hepatoma cells with the synthetic glucocorticoid dexamethasone induced u113 mRNA levels in a dose-dependent manner. Taken together these data suggested that glucocorticoids are required for the constitutive high level expression of this gene in normal adult rat livers. A series of 5' deletion constructs and linker scanning mutants of the promoter upstream region were produced and transfected into FAZA cells. A functional glucocorticoid response element was mapped between -168 and -151 base pairs 5' of the transcription start site. This element conforms with an inverted consensus glucocorticoid response element (GRE) but differs in two positions essential for protein DNA interaction between the GRE and the glucocorticoid receptor (GR). The induction of a113 gene promoter region constructs by dexamethasone was abolished by the receptor antagonist RU486, indicating that the GR participated in the activation of the al13 gene. In DNase I footprinting experiments with nuclear protein extracts from un-treated and dexamethasone-treated FAZA cells, similar extents of a113 promoter upstream sequences were protected, indicating that proteins capable of binding in the glucocorticoid response-mediating element (GME) region were present before and after arrival of the hormonal signal. However, a purified recombinant fragment of the GR which contained essentially only its DNA binding domain was unable to bind at the GME although it interacted strongly with a consensus GRE sequence.
al-inhibitor I11 (al13),' like many of the acute phase proteins, is a broad range proteinase inhibitor. It is the second most abundant plasma globulin in rats with a plasma concentration in the range of 6-10 mg/ml. a113 belongs to the amacroglobulin family of thiolester proteins which also includes al-macroglobulin, a,-macroglobulin, and the complement components C3, C4 and C5 (1)(2)(3). Unlike a,-macroglobulin, which is strongly induced, and al-macroglobulin, which is affected only very little by acute inflammations, a113 is strongly down-regulated during an acute phase inflammatory response (4): during the first 2 days after an inflammation al13 plasma concentrations fall to 1-2 mg/ml. It is the most strongly down-regulated "negative acute phase protein" known to date in rat plasma. In previous studies a113 cDNA clones were isolated and characterized and used to show that al13 mRNA levels decreased during an inflammation to an extent similar to that seen for al13 plasma levels (5-7). The transcription rate of the al13 gene was decreased 12.7-fold at 6 h after an experimentally induced inflammation with a concomitant 16-fold decrease in nuclear aJ3 precursor RNA concentrations (8).
Recently we have analyzed the inflammatory signals that are responsible for the acute phase negative regulation of the al13 gene in rat hepatoma cell lines (9). In that study we have established that interleukin 6 and glucocorticoids can regulate the al13 gene both in a positive and in a negative direction, depending on the precise concentrations of both hormones. We have also identified recently the transcriptional control elements in the 5' region of the al13 gene which are responsible for the highly tissue-specific expression seen in rat livers and have begun to characterize the hepatic nuclear factors that bind to these elements (10).
There is evidence for lobular heterogeneity of hepatocytes, showing systematic differences in the expression of plasma proteins depending on their topographic location in the liver The abbreviations used are: ul13, ul-inhibitor 111; GR, glucocorticoid receptor; GRE, glucocorticoid response element; GME, glucocorticoid response-mediating element; bp, base pairs; AGP, acid glycoprotein; LAP, liver activator protein; HBS, Hepes-buffered saline; EGTA, [ethylenebis(oxyethylenenitrilo)]tetraacetic acid. lobule (11,12). For example, in normal adult rat livers a higher frequency of expression of a113 mRNA and protein was observed in hepatocytes of the periportal and mediolobular areas than in the perivenous zones (11). One possible interpretation of these data was that various hormones entering the liver through the portal vein could create a hormone concentration gradient across the lobe which is reflected in the gradient of a113 RNA and protein concentrations. We knew that glucocorticoids are present in the systemic circulation of normal healthy adult rats in the absence of an inflammatory state in concentrations of approximately IO-' M (13) and that expression of a113 in cultured hepatic cells was sensitive to low levels of glucocorticoids (less than 10"" M (9)). Therefore, we wanted to test the hypothesis that glucocorticoids may be the principal hormones responsible for the maintenance of the normal constitutive expression levels of al13 in adult rat liver tissue.
Here we demonstrate that the al13 gene was regulated by glucocorticoids at the transcriptional level through a response element that was a variant of the consensus glucocorticoid response element (GRE) found in many other genes regulated by glucocorticoids (14)(15)(16). This element differed from a conventional GRE in that it was not sufficient to bind a truncated recombinant glucocorticoid receptor (GR) fragment that contained essentially only the DNA binding domain (17). The element found in the al13 gene partially overlapped with other elements important for the liver-specific expression of this gene which were mapped previously (10). Therefore, binding of the GR at this variant element may require additional contacts with other DNA-binding proteins attached to neighboring sites. The results also suggest that glucocorticoids are indeed required for the constitutive expression of the al13 gene in normal healthy adult rat livers.

MATERIALS AND METHODS
Treatment of Animals and Preparation of RNA-Normal and hypophysectomized male Fisher 344 rats, weighing 125-150 g, were obtained from Charles River Laboratories. After surgery, animals were allowed to recover for a t least 6 weeks. Only rats that showed no increase in their body weight and the absence of testicular development were used. When required, hypophysectomized animals were injected intraperitoneally with the indicated concentrations of dexamethasone in 0.5 ml of phosphate-buffered saline. Control animals were injected with 0.5 ml of phosphate-buffered saline. After the appropriate time interval rats were killed and the liver removed for RNA preparation. Three animals were used per data point.
RNA Preparation, Dot-blot Hybridization, and Primer Extension-Total RNA was prepared from both normal and hypophysectomized rat livers using a guanidinium hydrochloride procedure as described previously (18). RNA was isolat,ed from cultured cell lines using a guanidlnium thiocyanate lysis procedure followed by ultracentrifugation through a cesium chloride step gradient (19). The concentration of the RNA was determined by measuring the absorption a t 260 and 280 nm. Both the integrity and the relative concentrations of the RNA samples were monitored by electrophoresis in formaldehydeagarose gels and ethidium bromide staining.
0113 mRNA levels were quantitated by dot-blot hybridization. RNA was fixed to nylon (Genescreen) membranes by UV irradiation. Filters were then hybridized with cDNA probes for rat al13 that had been :',P labeled (5). In addition, duplicate series of dots were probed with a :',P-labeled ribosomal cDNA probe to monitor for constant RNA loading (20). RNA was quantitated by liquid scintillation counting of individual dots. specific primer, CTTCCATTTTACCAACAGTACCGGAATGCC, Luciferase mRNA was identified by primer extension using a that hybridized with nucleotides a t position +63 to +92 on the luciferase coding sequence (21). The specific 30-mer was end labeled using ['"PIATP and T4 polynucleotide kinase. The primer was hybridized at 50 "C using standard conditions, and cDNA was synthesized using avian myeloblastosis virus reverse transcriptase (Pharmacia LKB Biotechnology Inc. ). The luciferase-specific cDNA was detected after electrophoresis through a denaturing polyacrylamide gel and autoradiography. FAZA Hepatoma Cell Culture and Transfection-The rat hepatoma cell line, FAZA (22), and stably transfected derivatives were grown in a 1:1 mixture of Dulbecco's modified Eagle's medium and Ham's F-12 nutrient solution (GIBCO). Media were supplemented with 10% fetal bovine serum 100 units/ml penicillin and streptomycin. Cells were cultured in 100-mm dishes in an atmosphere of 5% CO, at 37 "C. For dexamethasone (Upjohn) and RU486 (Roussel Uclaf) treatment, cells were grown to 90% confluence (approximately 5 X 10" cells/100mm Petri dish) and supplemented with hormone.
Transfection was carried out by a modified calcium phosphate precipitation technique (23). Plasmid DNA to be used for transfection was isolated using an alkaline lysis procedure (24) and purified by two successive cesium chloride equilibrium centrifugation steps. FAZA cells were plated to be 60% confluent after overnight incubation. Prior to transfection the cells were washed three times with Hepes-buffered saline (HBS; 10 mM Hepes, pH 7.3,6.7 mM potassium chloride, 142 mM sodium chloride). The DNA precipitate was formed by dropwise addition of 250 mM calcium chloride to an HBS solution containing 7.5 mM Na2HP0, and 20 pg/ml test plasmid DNA plus 1 pg/ml pRSVCAT (25). After incubation in fresh medium for 1 h the cells were incubated for 6 h with the precipitate. The cells were then shocked with 25% glycerol (in HBS) for 2 min, washed three times with HBS, and incubated in fresh medium for 40 h. Transient transfectants were then assayed for luciferase activity. Stable transfectants were split 1:l and grown in the presence of 400 pg/ml '2418 (GIBCO) for 3 weeks with a change of medium every 5 days. Pools of approximately 500 stable G418-resistant clones were mixed, replated, and passaged for 2 weeks.
Luciferase assays were performed as described (21,27). After transfection, cell extracts were prepared as follows. After washing three times with HBS, the cells were scraped off the dish into 100 mM potassium phosphate, p H 7.2, 1 mM dithiothreitol, washed, and subjected to three freeze-thaw cycles. After lysis and removal of the cell debris by centrifugation, constant amounts of protein (120 pg) were assayed for luciferase activity using a Monolight 2001 luminometer (Analytical Luminescence Laboratories, San Diego, CA). The relative efficiency of transfection in each experiment was monitored by assaying for the production of constant amounts of chloramphenicol acetyltransferase (26) generated by the cotransfected pRSVCAT construct.
a1Z3 Luciferase Fusion Genes-A nested set of 5' deletions containing al13 promoter fragments extending from -2214 bp to + 58, which had been introduced into the transcription vector, pl9LUC (27), was available from a previous study (10). This construction produced fusion genes consisting of the 5'-noncoding sequences of the rvl13 gene linked to the coding region of the firefly luciferase gene.
Nuclear Extract Preparation-FAZA nuclear extracts were prepared from tissue culture cells that had reached 90% confluence. The cells were detached from the dishes using trypsin EDTA, and extracts were prepared according to the method of Shapiro et al. (28) and dialyzed against 20 mM Hepes, pH 7.9, 20% glycerol (v/v), 0.2 mM EDTA, 2 mM EGTA, 2 mM dithiothreitol, 60 mM potassium chloride. After dialysis, the extracts were stored in liquid Nr until needed.
DNase Z Footprinting and Gel Retardation Assays-DNase I footprint reactions (29) were carried out with unfractionated nuclear extracts. The -225 to -81 fragment from the aI13 promoter region was isolated, end labeled with the Klenow fragment of DNA polymerase I, and then recleaved with either Ssp1 or EcoRI followed by gel purification. The glucocorticoid response element containing pGTCO plasmid2 was prepared similarly after cleavage with Hind111 and EcoRI. The labeled probe fragments (1 ng) were then incubated in 20-pl reaction volumes containing 4 pg of poly(d1-dC) and either 50 pg of nuclear extract or 21 pg of T7X556, a purified glucocorticoid receptor fragment containing the DNA binding domain (17). Control reactions contained 40 pg of bovine serum albumin (molecular biology grade, Boehringer Mannheim) in 20 mM Hepes, pH 7.9,20% glycerol, 0.2 mM EDTA, 2 mM EGTA, 2 mM dithiothreitol, 60 mM potassium chloride. After a 10-min incubation on ice and 40 min a t 22 "C, 20 p1 of 20 mM magnesium chloride, 5 mM calcium chloride was added followed by an appropriate dilution of DNase I. After 120 s a t room temperature the reaction was stopped by the addition of 40 p1 of stop buffer (1% sodium dodecyl sulfate, 20 mM EDTA, sodium chloride, 25 pg/ml tRNA). The probe DNA was recovered by phenol/chloroform extraction and ethanol precipitation and analyzed on denaturing

RESULTS
Dexamethasone Increases a113 mRNA Concentration in Hypophysectomized Rats and in FAZA Hepatoma Cells-Results from previous studies on the acute phase regulation of the al13 gene in rat liver and rat hepatoma cell lines suggested that glucocorticoids may be required for the normal expression of al13 (9,11). To investigate whether glucocorticoids are required for normal aJ3 expression in rats al13 mRNA levels in the livers of hypophysectomized and normal rats were compared. In hypophysectomized rats the levels of al13 mRNA were approximately 20% of those determined in normal animals (Fig. 1A). To determine whether decreased a113 expression was specifically caused by lowered glucocortocoid levels in the hypophysectomized animals, the effect of administration of dexamethasone was studied. Hypophysectomized rats were treated with increasing doses of dexamethasone for varying lengths of time, and the al13 mRNA levels were compared with untreated hypophysectomized control and normal animals. Hepatic mRNA levels in hypophysectomized rats treated with increasing concentrations of dexamethasone (40 pg/kg-8 mg/kg) were increased in a dose-responsive manner (Fig. 1A). For instance, 6 h after administration of 8 mg/ kg dexamethasone the mRNA level was increased from 21 to 40% of normal levels. The a,I3 mRNA levels 12 and 24 h after administration of 8 mg/kg dexamethasone were 57 and 64%, respectively, of normal levels. Overall, aJ3 expression levels in rats with very low levels of glucocorticoids caused by hypophysectomy were decreased and could be restored at least in a major part by intraperitoneal injection of glucocortocoids.
The effects of glucocorticoids on al13 expression in vivo were paralleled in vitro in the rat hepatoma cell line, FAZA. FAZA cells were treated with increasing concentrations of dexamethasone, and 12 h later RNA was prepared. a113 mRNA levels were increased in a dose-responsive manner (Fig. lB), in a similar way as in dexamethasone-treated hypophysectomized rats. The "normal" expression levels in rats are equivalent to the levels observed in glucocorticoidtreated hepatoma cells.
The Promoter Upstream Region of the a113 Gene Mediates Glucocorticoid Induction-To establish that the glucocorticoid regulation of the al13 gene occurred at the level of transcription, transfection studies using 5' promoter constructs were initiated. Approximately 2.2 kilobase pairs of the al13 gene 5"flanking sequence were available (8). A restriction map of this al13 gene promoter region was generated. Subsequently, a region extending from -2,214 to +59 (relative to the transcriptional initiation site in liver) was linked to the coding region of the firefly luciferase reporter gene ( Fig. 2A). This construct, called A113 (-2,214), was used to cotransfect FAZA hepatoma cells. Stable clones expressing luciferase were isolated after G418 selection. Stable transfectants expressing luciferase from the SV40 minimal promoter (33) in the plasmid pSV232LUC were also isolated. A mixed culture of stable A113 (-2,214) transfectants, referred to as F-1, representing approximately 500 clones was used to test the response of (ul13 gene promoter region to different concentrations of glucocorticoid (Fig. 2B). In the absence of dexamethasone, F-1 cells expressed luciferase at a basal level (2,000 light units). Progressively higher concentrations of dexamethasone caused an increase in promoter activity from 1.6-fold (3,200 light units) using 1 X lo-" M dexamethasone to 9-fold (18,300 light units) with 1 X M dexamethasone. In contrast the activity of the SV40 minimal promoter in the transformed line, F-232 was not increased significantly after dexamethasone treatment.
5' Deletion Analysis of the Glucocorticoid-responsive Regions Upstream of the a113 Gene-To map the 5' extent of the region responsible for the glucocorticoid-regulated expression of the aI13 gene a series of deletion constructs with inserts extending from +58 bp to between -48 and -2,214 bp upstream of the transcription start site was used. These constructs were linked to the firefly luciferase gene, and the luciferase activity was measured after transient transfection into the rat hepatoma FAZA cell line. Constructs containing at least 225 bp upstream of the transcription initiation site, A113 (-2,214) and A113 (-225), contained sufficient information to mediate a 25-fold induction of luciferase activity after treatment with 1 X M dexamethasone for 24 h (Fig.  3A). The A113 (-186) construct gave a similar but slightly lower induction. In contrast the shortest constructs, A113 (-150) and A113 (-118) were not induced above that seen with the pSV232LUC control plasmid (4-fold). Thus, sequences located between -186 and -150 were necessary for at least a 6-fold glucocorticoid induction of the al13 gene.
Quantitation of the luciferase mRNA produced after transfection of constructs A113 (-2,214), A113 (-225), and A113 (-118) followed by dexamethasone treatment, by primer extension using a luciferase-specific oligonucleotide (Fig. 3B), showed that the transcriptional start site used in FAZA cells was within 4 bp of that used in rat liver aJ3 and that differences between the luciferase mRNA levels in the A113 (-2,214) and A113 (-225) constructs after dexamethasone treatment were similar to those seen with the luciferase activity measurements, relative to A113 (-118).
Delineation of the Glucocorticoid-responsive Element-A series of Sal1 linker scan mutations extending across the region that had been found to be essential for basal level expression (10) was available from a previous study (10). In the present study, this region was found to be involved also in mediating the induction by dexamethasone. The linker mutant constructs were transfected into FAZA cells, and luciferase activities were measured either with or without dexamethasone treatment (Fig. 4). Mutations LM5, LM6, LM7, and LM9 abolished the glucocorticoid responsiveness of the al13 promoter construct. Previous results (10) indicated that the mutations LM5, LM6, and LM9 completely abolished basal level transcriptional activity, and the present study showed that these same mutations also abolished glucocorticoid-inducible expression. However, the LM7 mutation did not affect basal level transcription (10) but completely abolished inducibility by dexamethasone. Thus, a GRE was contained within or overlapped the -156 to -151 region covered by the LM7 mutation and may extend further to the 5' end of the LM5 mutation at -168. Inspection of the -168 to -151 region revealed partial sequence homology to an inverted copy of the consensus GRE responsible for the transcriptional activation of many genes by glucocorticoids (14)(15)(16)(34)(35)(36) and functioning as a binding site for the GR. An ideal GRE, as defined by saturation mutagenesis (36), has the palindromic sequence quence matches the GME of the a113 gene as follows. 4 . . .

i r l 1 3 G M E 5' C C C T G G C A C A T T T C G T G C A A 3' -168 3' G G G A C C C T G T A A A G C A C G T T 5'
l l l l l l l l l I I I I reverse 3'

Y C N T G T N N N A C A N G R 5'
. . . . .

GRE
However, the GME varies in two nucleotides of one half site which are both important for receptor DNA contact. Therefore, the GME of the a113 gene deviates significantly from a consensus GRE. This sequence variation implies a potentially different way of protein DNA contacts between the GR and a consensus GRE and the GME of the al13 gene, respectively. Stimulation of Luciferase Expression Is Mediated by the Glucocorticoid Receptor-To determine whether the effect of dexamethasone on the transcriptional activation of the al13 promoter was mediated through the glucocorticoid receptor, the effect of the receptor antagonist RU486 was tested (37,38). FAZA cells were transfected with A113 (-225) followed by treatment with 10" M dexamethasone for 24 h. The transcriptional activity of the al13 luciferase fusion increased approximately 10-fold after treatment with this concentration of dexamethasone (data not shown). However, simultaneous treatment with both M dexamethasone and M RU486 led to a low transcriptional activity that was similar to that measured with either RU486 alone or control transfected cells. Thus, RU486 blocked completely the induction by dexamethasone, suggesting that the effect of dexamethasone on aJ3 promoter activity was mediated through the GR.
DNase I Footprint Analysis of the a113 Glucocorticoid Induction-mediating Element (GME)-To investigate whether transcription factors bind to the region that had been mapped as being important for the glucocorticoid induction of the al13 gene, DNase I footprint analysis of the area surrounding the -156 to -151 region of the promoter was carried out using nuclear extracts prepared from either uninduced or 5-h dexamethasone-induced FAZA cells (Fig. 5A and data not shown). A region protected from DNase I digestion which coincided with the -156 to -151 region mapped as containing the GME was found. The protected region included the GME region, indicating that a complex of nuclear factors bound to these sequences as described previously (10). Both uninduced and induced extracts gave a similar region of protection from DNase I digestion ( Fig. 5A and data not shown), suggesting that similar or identical GME binding activities may be present in both uninduced and induced extracts.
These results were confirmed and extended by gel retardation assay using a double-stranded oligonucleotide B3, 5' TCCCCTGGCATTTCGTGCAAC 3', which included the GME sequence (Fig. 5B). When nuclear extracts from uninduced FAZA cells were used in the assay, two predominant retarded complexes were seen. When nuclear extracts from 5or 12-h dexamethasone-treated FAZA cells were used two retarded complexes were seen which appeared to be identical to those obtained with uninduced extracts. The retarded complexes could be competed specifically by unlabeled 3B. There was no apparent change in the GME binding activity when FAZA cells were treated with dexamethasone. If the GR bound to this region only after hormonal treatment then one might have expected to find a corresponding change in the gel retardation pattern, even if the binding was weak.
Purified Glucocorticoid Receptor Does Not Bind to the GME of the a113 Gene-DNase I footprint analysis was carried out on the aJ3 GME region to determine whether the glucocorticoid receptor was able to bind. A 150-amino acid glucocorticoid receptor fragment, T7X556, was used in footprint reactions. This fragment contained the DNA binding domain and protected the same GRE sequence as the full-length receptor (17). When the al13 promoter region was used in footprint reactions this receptor fragment was unable to bind to the sequence delineated by the LM7 mutation (-151 to -156) or to sequences extending through the LM5 and LM6 region (-157 to -168). Only a very minor change in the footprint profile was seen around -151 and also around -163 on the noncoding strand but not on the coding strand. The consensus GRE contained in the pGTCO plasmid2 was very strongly protected from DNase I digestion by this receptor even when %O of the concentration of T7X556 was used as compared with the a113 binding reactions (Fig. 6).

DISCUSSION
Two essential results were obtained in this study. ( a ) Glucocorticoids are required for the constitutive high level expression of the al13 gene in healthy adult rat livers. ( b ) The glucocorticoid response element of the al13 gene is a variant of the consensus GRE, differing from it in qualitatively important aspects.
Glucocorticoids participate in the transcriptional activation of a number of hormone-inducible genes (14)(15)(16). T o our knowledge, however, there are no well documented cases in which glucocorticoids have been shown to be an essential prerequisite for the constitutive, high level expression of a gene in its normal producer cell type. The a113 gene is an example of such a situation. This gene differs strikingly from the majority of other plasma protein genes expressed in the liver which have been studied so far. Other genes, specifying abundant plasma proteins that are constitutively expressed, such as the transferrin, complement C3, C4, or albumin genes, are controlled by a set of transcription factors (HNF1, 39-41; C/EBP; DBP-LAP; 42-45 and others) which are themselves constitutively and preferentially expressed in fully differentiated hepatocytes. In contrast to those genes, the al13 gene is representative of a rare category of strongly expressed liverspecific genes which, in addition to those constitutive factors, require inducible factors such as the GR for their transcription. Our results showed that in rats hypophysectomy greatly decreased the production of al13 mRNA in the liver and that 'me Element of the a113 Gene 18273 administration of glucocorticoids served to restore a major fraction of the normal a113 mRNA expression levels.
In the current study, FAZA hepatoma cells treated with dexamethasone were shown to behave similarly to hepatocytes in vivo; both the endogenous al13 gene and transfected al13 promoter luciferase constructs were induced by dexamethasone in a dose-responsive manner. The results confirmed that FAZA hepatoma cells provided a good model system to study the glucocorticoid-mediated transcriptional regulation of the al13 gene. In this model system, resting hepatoma cells were considered to be equivalent to hepatocytes in hypophysectomized rat liver. Glucocorticoid-treated (induced) hepatoma cells, in contrast, were equivalent to hepatocytes in normal rats with normal levels of circulating glucocorticoids. Using a transient transfection assay, a glucocorticoid-responsive element was located between positions -186 and -150 bp upstream from the al13 transcription initiation site. Linker scan analysis delineated the GRE to the sequence CCCTGGCA-CATTTCGTGC (-167 to -150). Comparison of this sequence with the glucocorticoid response element found in many steroid-responsive genes (14)(15)(16)(34)(35)(36) showed that this element is an inverted consensus GRE element differing in a few nucleotides from the consensus GRE (8). The consensus GRE has been defined in two different ways in the literature: once by comparing the functional elements found in many different genes (14)(15)(16), and, more recently, be systematic saturation mutagenesis in all 15 nucleotide positions of the GRE and comparing relative binding intensities for the GR (36). The "ideal" GRE, defined by this latter approach as the best GR protein binding sequence, showed good agreement with the consensus GRE as defined by the first approach (14-16). The GRE consists of two half-sites in a palindromic arrangement. One half-site is thought to be occupied first and to facilitate binding of a second molecule of the GR at the second halfsite by cooperative interaction (36). Until recently it was thought that the strong half-site must be a hexanucleotide with only one unspecified position (N) separating the TGT triplet from the CY doublet. However, it was shown recently (36) that a half-site can also be a heptanucleotide with two unspecified nucleotides separating the triplet and the doublet: T G T N1 N2 C Y. This result suggests that within each halfsite there are two separable subelements that are possibly contact points for separate subdomains of the GR with the DNA. I t is of interest to note that in the GME of the aI13 gene, one half-site is perfectly conserved whereas in the second half-site two nucleotides differ as shown in the alignment given under "Results." Both of these variant positions are nucleotides involved in protein DNA contact between a consensus GRE and the GR. It is therefore conceivable that the second half-site of the GME becomes a very weak binding site or no binding site a t all and that it could even destabilize or prevent binding of a first copy of the GR at the conserved strong half-site. The variant sequence of the GME suggests the plausible prediction that the GME of the a113 may be a much weaker binding site for the GR than a consensus GRE. In view of the variant sequence of the GME we were prompted to ask whether the glucocorticoid effect involved the GR and whether the receptor bound directly to the GME of the al13 gene. The glucocorticoid receptor antagonist RU486 was able to block the effect of dexamethasone completely, indicating that the al13 promoter activity was mediated through the GR. However, when the purified recombinant GR fragment was used in footprint reactions with the al13 GME, no significant binding was observed. There are a number of possible explanations for this observation.
First, the GR may be able to bind weakly at the GME of the al13 gene, but to initiate transcription successfully it may need to be held in place by interaction with other accessory factors. This interaction possibly requires domains of the GR other than its DNA binding domain. In our experiments with the truncated recombinant GR fragment no binding may have been observed because ( a ) the intrinsic binding to the GME was too weak because of the variant DNA sequence of the site; and ( b ) because the necessary factors were not present. The glucocorticoid receptor is known to undergo cooperative GRE binding with a variety of transcription factors such as nuclear factor 1, Spl, OTF-1, CCAAT box binding factors and the estrogen receptor, to activate an adjacent promoter (46)(47)(48)58). Chang and co-workers (59) have shown that the transcription factor AGP/EBP, a member of the C/EBP family and probably the murine equivalent of rat interleukin 6-DBP/liver activator protein (44,45), binds at a sequence overlapping the GRE of the murine alAGP gene. The fact that the GME of the al13 gene is highly conserved with the functionally active glucocorticoid response regions of the rat and murine alAGP (8,(49)(50)59) and a2"-g1obulin genes (8,51) makes the involvement of accessory proteins plausible. In fact, for the control of the rat alAGP gene by glucocorticoids, Klein et al. (52,53) have demonstrated a requirement for accessory proteins binding at neighboring sites, some of which are labile and need to be resynthesized continuously. The alAGP GRE shares a higher degree of sequence homology with the GME of the rat a113 gene than with the consensus GRE sequence (8). However, unlike alAGP, the al13 GME appears unable to bind recombinant GR in footprint reactions. It is possible that because of the constitutive induction of al13 by glucocorticoids and its down-regulation during the acute phase, the GR requires accessory factors to bind to the GME since cycloheximide completely blocked both the uninduced and the dexamethasone-induced expression of the al13 gene. " A second possible explanation of our data is that the GR may not bind directly to the GME DNA sequence but rather indirectly to other proteins that are in turn attached to the DNA. A similar mechanism was proposed for the regulation of the prolactin gene by steroids. Regulation of this gene involves the estrogen and glucocorticoid receptors interacting with other transcription factors to suppress prolactin gene transcription (54). Similarly, the glucocorticoid receptor is involved in the transcriptional repression of the collagenase gene not by direct binding to the DNA but by interaction with the transcription factor jun in the fluid phase and thus preventing it from activating the gene through an AP1 site (55-57).
A third possible explanation is that the glucocorticoid effect on the al13 gene is indirect and does not involve binding of the GR at the GME, either by direct protein-DNA interaction or by protein-protein interaction. Instead, glucocorticoids could activate an intermediate gene, and the intermediate gene product could exert its control over the al13 gene through the GME site without interaction of the GR with this site.
Among the possibilities, we believe the most likely is that the GR binds at the GME to the DNA and requires auxiliary proteins bound at neighboring sites to hold it in place and to interact productively with the components of the transcriptional machinery to augment the rate of transcription initiation. Previously we have mapped four tissue specific cis-acting elements to sequences surrounding the GME in the al13 gene (10). We have established that nuclear factors bind to these elements in both uninduced and dexamethasone-induced FAZA nuclear extracts. Considering the close proximity of L. Abraham, unpublished data.
mse Element of the a113 Gene the GME and these previously characterized tissue specific cis-acting elements it is likely that there is a close association between the various transcription factors that bind to these sequences.