C/EBP Regulates Hepatic Transcription of 11β-Hydroxysteroid Dehydrogenase Type 1

Glucocorticoid action within individual cells is potently modulated by 11β-hydroxysteroid dehydrogenase (11β-HSD), which, by interconverting active and inert glucocorticoids, determines steroid access to receptors. Type 1 11β-HSD (11β-HSD1) is highly expressed in liver where it regenerates glucocorticoids, thus amplifying their action and contributing to induction of glucocorticoid-responsive genes, most of which are also regulated by members of the C/EBP (CAAT/enhancer-binding protein) family of transcription factors. Here we demonstrate that C/EBPα is a potent activator of the 11β-HSD1 gene in hepatoma cells and that mice deficient in C/EBPα have reduced hepatic 11β-HSD1expression. In contrast, C/EBPβ is a relatively weak activator of 11β-HSD1 transcription in hepatoma cells and attenuates C/EBPα induction, and mice that lack C/EBPβ have increased hepatic 11β-HSD1 mRNA. The 11β-HSD1promoter (between −812 and +76) contains 10 C/EBP binding sites, and mutation of the promoter proximal sites decreases the C/EBP inducibility of the promoter. One site encompasses the transcription start, and both C/EBPα and C/EBPβ are present in complexes formed by liver nuclear proteins at this site. The regulation of 11β-HSD1 expression, and hence intracellular glucocorticoid levels, by members of the C/EBP family provides a novel mechanism for cross-talk between the C/EBP family of transcription factors and the glucocorticoid signaling pathway.

Glucocorticoids, synthesized and secreted by the adrenal cortex, play a vital role in maintaining homeostasis, particularly during stress. The control of energy metabolism is central to the maintenance of homeostasis, and glucocorticoids play an important role in regulating glucose availability and utilization. In addition, during inflammation or injury, glucocorti-coids potently suppress the immune response and play a role in the acute phase response in liver (1). The actions of glucocorticoids are principally mediated by the type II or glucocorticoid receptor and, in a limited number of tissues, by the related type I or mineralocorticoid receptor. The interaction between glucocorticoid hormone and receptors is crucially modulated by the glucocorticoid-metabolizing 11␤-hydroxysteroid dehydrogenase (11␤-HSD; EC 1.1.1.146) 1 enzymes (reviewed in Refs. 2 and 3).
11␤-HSD interconverts active glucocorticoid hormones (corticosterone, cortisol) and inert 11-keto metabolites (11-dehydrocorticosterone, cortisone), thus controlling intracellular availability of active glucocorticoids. Two isozymes of 11␤-HSD have been identified; type 1 (11␤-HSD1) and type 2 (11␤-HSD2). 11␤-HSD2 is a high affinity NAD ϩ -dependent enzyme. It functions exclusively as a dehydrogenase (inactivating glucocorticoids) (4,5) and is expressed in placenta and in aldosterone target tissues (e.g. kidney) where it protects otherwise non-selective mineralocorticoid receptors from occupation by glucocorticoids (6 -8). In contrast, 11␤-HSD1 is a lower affinity, NADP(H)-dependent enzyme, which is widely expressed, with highest levels in liver (9). Although 11␤-HSD1 is bi-directional in cell homogenates, in intact cells, including primary rat hepatocytes, the enzyme is predominantly a reductase, regenerating active glucocorticoids from inactive 11-dehydroglucocorticoids (10,11). Mice homozygous for a targeted disruption of the 11␤-HSD1 gene cannot reduce 11-keto glucocorticoids to active 11-hydroxy steroids (12). Hepatic 11␤-HSD1 is the major site of regeneration of active glucocorticoids; in humans, inert cortisone administered orally is rapidly converted to cortisol, predominantly by the liver (13). As well as contributing toward plasma glucocorticoid levels, 11␤-HSD1 increases intracellular glucocorticoid levels, amplifying glucocorticoid action. 11␤-HSD1-deficient mice show impaired activation of gluconeogenic enzymes upon starvation, resulting in lower blood glucose levels than their wild-type littermates, and resist hyperglycemia provoked by obesity or stress (12). In humans, 11␤-HSD inhibition leads to increased insulin sensitivity, presumably by attenuating glucocorticoid antagonism of hepatic insulin action (14). The enzyme thus provides a novel control of hepatic glucose/insulin relationships and represents a potential therapeutic target for the manipulation of hepatic glucose homeostasis and insulin sensitivity. Clearly, the regulation of hepatic 11␤-HSD1 expression is of substantial interest. Although the control of 11␤-HSD1 expression has been widely studied in vivo and in cell culture (reviewed in Ref. 2), the molecular mechanisms governing 11␤-HSD1 transcription have yet to be determined.
The promoter region of the rat 11␤-HSD1 gene has been cloned (15). A single major promoter is active in liver and hippocampus, although kidney utilizes two additional promoters (15). The promoter used in liver lacks a TATA box, but has a CCAAT sequence at Ϫ73 to Ϫ69 (the transcription start site is designated ϩ1) as well as a GCAAT sequence in the inverse orientation between Ϫ63 and Ϫ67. In this study, we have investigated the transcriptional control of the 11␤-HSD1 gene in liver and demonstrate the essential role played by C/EBP␣ in the regulation of the gene encoding this key glucocorticoidmetabolizing enzyme.

EXPERIMENTAL PROCEDURES
Subcloning and Sequence Analysis of the Rat 11␤-HSD1 Promoter A 6.8-kb EcoRI fragment encoding part of the rat 11␤-HSD1 gene was subcloned from A (15) and sequenced using Sequenase II (Amersham Pharmacia Biotech). Putative transcription factor binding sites were identified using computer software available at the UK Medical Research Council Human Genome Mapping Project Resource Center.
HepG2 cells were maintained and transfected as described previ-ously (16). 5 ϫ 10 5 cells seeded per 60-mm dish were transfected using the calcium phosphate procedure with 5 g of test plasmid, 1 g of pCH110 (Amersham Pharmacia Biotech) (as internal control), 1 g of C/EBP expression plasmid (0.5 g of each when added together), made to a total of 10 g with pGEM3 (Promega). In some experiments, pMSV was used in control transfections that did not include C/EBP; results for transfections with pMSV were not significantly different to those in which no "empty" vector was used. 48 h after transfection, luciferase activity was assayed in cell lysates as described previously (16). ␤-Galactosidase activity was assayed using the Tropix Galacto-Light kit (Cambridge Bioscience, Cambridge, UK).

DNase I Footprinting and Electrophoretic Mobility Shift Assays
DNA Fragments and Oligonucleotides-Complementary oligonucleotides were synthesized by OSWEL (Department of Chemistry, University of Southampton, UK) and are shown in Table I. A series of overlapping restriction fragments covering the 11␤-HSD1 promoter between Ϫ812 and ϩ76 were each labeled at the 5Ј-end on the lower strand. In most cases the fragment was first subcloned into pGEM-3 (Promega) to generate suitable restriction sites for end labeling. DNA fragments and double stranded oligonucleotides were end-labeled using [␣-32 P]dATP and the Klenow fragment of DNA polymerase I.
Preparation of Nuclear Extracts-Nuclei were purified from rat liver by centrifugation through buffered sucrose as described (18). Nuclear extracts were prepared by the method of Dignam et al. (19) and stored in liquid N 2 .
Bacterial Expression of Recombinant C/EBP␣-pT5, encoding a 35-kDa fragment of rat C/EBP␣ lacking the N-terminal 60 amino acids (a gift of W.-C. Yeh and S. L. McKnight) was used to express recombinant C/EBP␣ (rC/EBP␣) in Escherichia coli BL21, as described (20). rC/ EBP␣ was partially purified over a DEAE-cellulose column (Amersham Pharmacia Biotech, St. Albans, UK) according to a previous study (20), and fractions containing rC/EBP␣ were identified by EMSA analysis, as described below, using oligonucleotide O E .
Electrophoretic Mobility Shift Assays-EMSA reactions (20 l) contained 5-10 g of rat liver nuclear extract or 0.5 g of bacterial extract containing rC/EBP␣, 3 g of poly(dI-dC) (Amersham Pharmacia Biotech), 4 mM Tris-HCl (pH 7.5), 100 mM KCl, 5 mM MgCl 2 , 1 mM EDTA, 10% glycerol, 0.2% Triton X-100 and were preincubated at 22°C for 15 min prior to addition of 0.1 pmol of 32 P-labeled DNA. After a further 15-min incubation, reactions were electrophoresed on 5% non-denaturing polyacrylamide gels containing 350 mM Tris, 450 mM boric acid, 100 mM EDTA. Where appropriate, a 10-or 100-fold molar excess of competitor DNA was included in the preincubation, prior to addition of 32 P-labeled DNA. For antibody supershift assays, 1 l of C/EBP␣ antiserum or C/EBP␤ antiserum (Santa Cruz Biotechnology Inc., Santa Cruz, CA) or, as a control, COUP transcription factor (COUP TF) antiserum (a gift of M. Parker) were included in the preincubation before addition of 32 P-labeled DNA.
DNase I Protection Analysis-50 g of rat liver nuclear extract or 2 g of bacterial extract containing rC/EBP␣ were preincubated on ice for 15 min in 30 l containing 10% glycerol, 10 mM Tris-HCl (pH 7.5), 2.5 mM MgCl 2 , 1 mM CaCl 2 , 0.1 mM EDTA, 75 mM KCl, 4 mM spermidine, 0.5 mM dithiothreitol, 1 g of poly(dI-dC). Reactions were incubated for 15 min on ice after addition of 32 P-labeled DNA then DNase I added for 1.5 min. Reactions were terminated by addition of EDTA and NaCl to final concentrations of 21 and 400 mM, respectively. DNA was purified using protein removal cartridges (NBL Gene Sciences Ltd., Cramling- CAGGTCCTGTAGCACTGTACC ton, UK), precipitated with ethanol, resuspended, and separated on a 10% polyacrylamide gel containing 7 M urea. Size markers were produced by carrying out Maxam and Gilbert DNA sequencing reactions using the same 32 P-labeled DNA fragment.

Analysis of 11␤-HSD1 mRNA in Livers of C/EBP␣ and C/EBP␤ Knock-out Mice
Mice homozygous for deletions in the genes encoding C/EBP␣ and C/EBP␤ were generated as described in References 21 and 22, respectively. 10 g of total RNA isolated from livers of newborn homozygous C/EBP␣-deficient mice or from the livers of 16-week-old male mice deficient in C/EBP␤ was analyzed by Northern blotting as described previously (23). cDNA probes used for hybridization were: 11␤-HSD1, a mouse cDNA probe encoding nucleotides 158 -618 (24) and 7S, a cDNA encoding mouse 7S RNA (25). Hybridization signals were quantified by phosphorimaging (Fuji FLA-2000).

Statistics
Data were analyzed using analysis of variance. Significance was set at p Ͻ 0.05. Values are mean Ϯ S.E.

Subcloning and Sequence Analysis of the 11␤-HSD1 Gene
Promoter-To identify putative regulatory sequence elements in the rat 11␤-HSD1 gene, we extended the previous limited sequence analysis (15) and determined the complete sequence of a 6.8-kb EcoRI fragment encoding 5287 nucleotides of 5Јflanking DNA and part of the structural gene, including the first two exons. Several microsatellite sequences are present 5Ј of Ϫ2500, including a 170-bp polypurine sequence and a 300-bp (NA) n repeat. At Ϫ462 there is a (CT) 26 repeat, immediately followed by (GT) 19 and 65 nucleotides rich in CCTT repeats. Several putative transcription factor binding sites were noted, including sequences corresponding to glucocorticoid response element consensus half sites and putative binding sites for HNF1, HNF3, and the C/EBP family of transcription factors.
Transfection of HepG2 Cells Identifies C/EBP-responsive Sequences in the 11␤-HSD1 Gene Promoter-Because the region 5Ј to the 11␤-HSD1 gene contains several putative C/EBP binding sites, the involvement of C/EBP in the transcriptional control of the promoter was investigated in HepG2 cells (a human hepatoma cell line) using a series of plasmids fusing 5Ј-flanking DNA and 49 bp of exon 1 of the rat 11␤-HSD1 gene to a luciferase reporter gene. HepG2 cells express low levels of endogenous 11␤-HSD1 (data not shown) and have been extensively used to characterize gene regulation by C/EBP. Basal expression of pr11␤1(Ϫ3618/ϩ49) was low in transfected HepG2 cells, and deletion of 5Ј-flanking DNA to Ϫ812 had no significant effect on luciferase activity (Fig. 1A). Deletion of DNA between Ϫ812 and Ϫ754 resulted in higher basal activity, which was not significantly altered with further deletion to Ϫ88 (Fig. 1A). Co-transfection of C/EBP␣ with pr11␤1(Ϫ3618/ ϩ49) increased luciferase reporter activity approximately 15- to 20-fold (Fig. 1A), and deletion of 11␤-HSD1 5Ј-flanking DNA to Ϫ579 had no significant effect on the magnitude of the C/EBP␣ induction (Fig. 1A). However, deletion of DNA between Ϫ579 and Ϫ322 reduced C/EBP␣ induction (approximately 3-fold), with a further reduction on deletion to Ϫ124 and elimination with deletion to Ϫ88 (Fig. 1A). These results show that a repressor element lies between Ϫ812 and Ϫ754 and that several regions of the 11␤-HSD1 promoter, particularly between Ϫ579 and Ϫ88, contribute to C/EBP␣ inducibility.
Both the DNA fragment encoding Ϫ599 to Ϫ88 and that encoding Ϫ812 to Ϫ599 of the rat 11␤-HSD1 gene were able to confer induction by C/EBP␣ on the SV40 promoter (Fig. 1B), demonstrating that the promoter contained at least two C/EBP␣ responsive regions. Neither fragment, however, conferred the same magnitude of inducibility as seen with pr11␤1(Ϫ812/ϩ49) or pr11␤1(Ϫ599/ϩ49) (Fig. 1B). In addition, the fragment encoding Ϫ812 to Ϫ599 significantly decreased the activity of the SV40 promoter in the absence of C/EBP␣ (Fig. 1B), consistent with the location of a repressor element in the region between Ϫ812 and Ϫ754.
Under normal conditions in rat liver, most of the C/EBPbinding activity is attributable to C/EBP␣ and C/EBP␤ (26). To see if C/EBP␤ is also able to activate the 11␤-HSD1 promoter, HepG2 cells were co-transfected with pMSV-C/EBP␤ and pr11␤1(Ϫ1799/ϩ49). Full-length C/EBP␤ caused a small but significant induction of transcription from the rat 11␤-HSD1 promoter (Fig. 1C) and, when co-transfected with C/EBP␣, resulted in a similar level of pr11␤1(Ϫ1799/ϩ49) activity to that induced by C/EBP␤ alone (Fig. 1C). However, it has been previously shown that removal of the N-terminal 21 amino acids of C/EBP␤ decreases production of LIP, an inhibitor of C/EBP activity produced from the C/EBP␤ mRNA by translation initiating at a downstream ATG (27). When an N-termi-nally truncated C/EBP␤, C/EBP␤(t), was co-transfected into HepG2 cells, it activated transcription from pr11␤1(Ϫ1799/ ϩ49), although to a lesser extent than C/EBP␣ (Fig. 1C). As with C/EBP␤, co-transfection of C/EBP␤(t) and C/EBP␣ resulted in a similar level of pr11␤1(Ϫ1799/ϩ49) activity to that induced by C/EBP␤(t) alone (Fig. 1C).
Footprint Analysis of the 11␤-HSD1 Gene Promoter-DNase I protection analysis was used to identify sites of rat liver nuclear protein interaction with the 11␤-HSD1 promoter between Ϫ812 and ϩ76 using a series of overlapping restriction fragments, each labeled at the 5Ј-end on the non-coding strand. Parallel reactions were carried out with bacterially expressed rC/EBP␣ protein to identify the footprints that may be due to C/EBP-related factors (control bacterial extracts that did not express rC/EBP␣ did not show any binding to 11␤-HSD1 DNA; data not shown). Representative results from the DNase I protection analysis are shown in Fig. 2 and summarized in Fig.  3. Liver nuclear proteins bind to at least 11 sites on the 11␤-HSD1 promoter between Ϫ812 and ϩ76. Furthermore, 10 of these sites can be wholly or partially occupied by C/EBP-related proteins.
In the proximal region between Ϫ88 and ϩ76 there were two clear regions of protection, FP1 and FP2, within the transcribed region and spanning the transcription start, respectively (Fig. 2). The CCAAT sequence at Ϫ73 to Ϫ69 and the GCAAT sequence in the inverse orientation between Ϫ63 and Ϫ67 were weakly protected only with high concentrations of liver nuclear extract (data not shown), suggesting that these sequences, present in a position typical of such regulatory elements, play little or no role in the expression of 11␤-HSD1 in rat liver.
Between Ϫ599 and Ϫ88, six regions, FP3-8, were protected from DNase I digestion by rat liver nuclear extract, all of which were similarly protected by rC/EBP␣ protein (Fig. 2). In some experiments FP5 appeared to extend to Ϫ255, however, in all experiments the region between Ϫ224 and Ϫ244 was clearly protected. Hypersensitive sites induced by liver nuclear extract (but not by rC/EBP␣) were also observed at Ϫ498, Ϫ510, and Ϫ550.
Between Ϫ812 and Ϫ599 two regions, FP9 and FP11, were protected both by liver nuclear extract and by rC/EBP␣ (Fig. 2). A third region, FP10, was protected by liver nuclear extract but weakly, if at all, by rC/EBP␣ (Fig. 2). Although this footprint clearly includes Ϫ768 to Ϫ759, in some experiments protection appeared to extend as far in the 3Ј direction as Ϫ747 (data not shown), possibly as a result of DNA distortion by liver nuclear extract (see below). rC/EBP␣ and liver nuclear proteins produced identical footprints at FP11; however, several differences were observed at FP9. Rat liver nuclear extract, but not rC/ EBP␣, induced a DNase I hypersensitive site at Ϫ694, within the protected region (Fig. 2) and also induced a region of hypersensitivity 3Ј to FP9 (enhanced cleavage at Ϫ663, Ϫ661, and Ϫ655) (Fig. 2), the extent of which varied with different preparations of extract. Furthermore, those preparations of liver nuclear extract that resulted in the appearance of strong hypersensitive sites at Ϫ663 to Ϫ655 apparently protected the adjacent DNA (Ϫ704 to Ϫ664 and approximately Ϫ647 to Ϫ616) from DNase I digestion as well ( Fig. 2 and data not shown). Because the adjacent "protected" regions were only seen with preparations of extract that strongly induced the DNase I hypersensitivity, it is likely that they are due to exclusion of DNase I by distortion of DNA rather than protection from DNase I by direct protein binding.
The Promoter Proximal Footprints, FP1, -2, -3, and -4 Are Required for Full C/EBP␣ Inducibility-The introduction of a mutation into FP2, which dramatically reduced C/EBP binding (data not shown), reduced C/EBP␣ inducibility of pr11␤1(Ϫ196/ ϩ49) (Fig. 4), although it also diminished basal levels of transcription (Fig. 4). Mutation of FP1, within the transcribed region, eliminated C/EBP␣ induction, while having no effect on basal transcription (Fig. 4). Similarly, mutation of either FP3 or FP4 alone reduced the C/EBP␣ effect (Fig. 4). Mutation of any two of the footprinted sites together further diminished or even eliminated induction by C/EBP␣ (Fig. 4), and mutation of FP2, -3, and -4 together both eliminated the induction by C/EBP␣ and decreased basal levels of transcription (Fig. 4). These data confirm the requirement for C/EBP␣ binding in the transcriptional activation of the rat 11␤-HSD1 promoter.
C/EBP Binds to at Least Two of the Footprinted Sites-The similarity between the footprints produced by liver nuclear extract and rC/EBP␣ strongly suggested that the factors present in liver nuclear extract, which bind to the rat 11␤-HSD1 promoter, belong to the C/EBP family. To characterize the liver nuclear protein(s) binding to these sites, EMSA analysis was carried out on FP2, spanning the transcription start, and FP9, which exhibited differences between liver nuclear extract and rC/EBP␣ binding.
Several specific protein-DNA complexes were formed on oligonucleotide O A (encoding FP2) by rat liver nuclear extract (Fig. 5A). Competition assays demonstrated that all the specific complexes contained proteins with DNA-binding specificity related to C/EBP; all were similarly competed by O E and O D (encoding an optimal C/EBP binding site and the P3(I) C/EBP binding site from the PEPCK gene, respectively). Oligonucleotides encoding cAMP-responsive elements or AP-1 binding sites also competed for liver nuclear protein binding to a limited extent (data not shown) consistent with the pattern of specificity previously described for C/EBP (28). Little or no competition was seen when oligonucleotides O B and O C (encompassing the CCAAT box at Ϫ73 and the GCAAT sequence at Ϫ67, respectively, of the 11␤-HSD1 gene) were included in EMSA reactions (Fig. 5A). rC/EBP␣ bound to oligonucleotide O A as a single specific complex of similar electrophoretic mobility to the major complex formed by liver nuclear proteins, and the rC/EBP␣-O A complex showed similar (although not identical) competition by oligonucleotides as seen with liver nuclear proteins (Fig. 5B). Very similar results were obtained with rat liver nuclear extract and rC/EBP␣ binding to oligonucleotide O F , encompassing FP9 (data not shown), with rank order of competition of showing that FP9 also represents a high affinity site for C/EBP-related proteins.
C/EBP␣ Is the Major Isoform of C/EBP Binding in Liver Nuclear Extracts-Addition of antiserum specific for C/EBP␣ to EMSAs demonstrated that the majority of the complexes formed by liver nuclear extract on O A include C/EBP␣ (Fig. 5C). Addition of C/EBP␤ antiserum supershifted a minor proportion of the complexes, however, addition of both antisera together supershifted all but one specific complex (Fig. 5C). Very similar results were obtained when oligonucleotide O F was used in supershift EMSA analysis (data not shown).
Hepatic Expression of 11␤-HSD1 mRNA Is Reduced in C/EBP␣ Knock-out Mice, but Increased in C/EBP␤ Knock-out Mice-To examine the relative importance of C/EBP␣ and C/EBP␤ in hepatic expression of 11␤-HSD1 mRNA in vivo, we carried out Northern analysis on RNA isolated from livers of mice that lack either C/EBP␣ (21) or C/EBP␤ (22). 11␤-HSD1 mRNA was dramatically reduced in livers of mice deficient in C/EBP␣, compared with wild-type littermates (Fig. 6A). In marked contrast, 11␤-HSD1 mRNA was increased 2-fold relative to 7S RNA in the livers of C/EBP␤-deficient mice (Fig. 6, B  and C). 3. Summary of the sites of rat liver nuclear protein binding on the rat 11␤-HSD1 gene between ؊812 and ؉76. Regions footprinted on the non-coding strand by rat liver nuclear extract are indicated by boxes below the sequence. For clarity, the coding (top) strand only is shown. The start of transcription (ϩ1) is indicated by a bent arrow. *, indicates the 5Ј-nucleotide present in the 5Ј-deletion plasmid series. Sequences that match consensus C/EBP binding sites are underlined.

DISCUSSION
11␤-HSD1 is highly expressed in liver where it regenerates active steroids from inert 11-keto forms. Hepatic 11␤-HSD1 controls induction of glucocorticoid-responsive genes (12,29), most, if not all of which are also regulated by members of the C/EBP family of transcription factors (30 -33). Here we show that C/EBP␣ is a potent activator of the 11␤-HSD1 gene in hepatic cells both in vivo and in vitro. In contrast, C/EBP␤, a weak activator of 11␤-HSD1 transcription, acts as a relative inhibitor of C/EBP␣-stimulated 11␤-HSD1 promoter activity in vitro, and mice lacking C/EBP␤ have increased hepatic 11␤-HSD1 mRNA. The 11␤-HSD1 gene has an unusually large number of C/EBP binding sites in the proximal promoter region, including one overlapping the transcriptional start site, suggesting that the regulation by C/EBP may be of particular physiological significance.
The promoter of the 11␤-HSD1 gene, between Ϫ812 and ϩ76, contains 11 binding sites for liver nuclear proteins (FP1-11). In transfected HepG2 cells plasmids containing at least 812 nucleotides, and up to 5 kb of 5Ј-flanking DNA from the rat 11␤-HSD1 gene showed low basal promoter activity, which was dramatically increased by C/EBP␣. Although C/EBP␣ is highly abundant in liver, it is present at very low levels in HepG2 cells (34) but may be sufficient to give rise to the low basal activity of the 11␤-HSD1 promoter seen in these cells. Deletion of 11␤-HSD1 DNA from Ϫ812 to Ϫ754, removing FP10 and FP11, increased basal promoter activity, and fusion of DNA encoding Ϫ812 to Ϫ599 (encompassing FP9, -10, and -11) upstream of the SV40 promoter decreased promoter activity, consistent with the presence of a repressor element between Ϫ812 and Ϫ754. The repressor element may correspond to FP10, the only site in the rat 11␤-HSD1 promoter that was bound by liver nuclear extract but not by rC/EBP␣. The identity of the liver nuclear factor(s) protecting this site is currently unknown.
The 11␤-HSD1 promoter is inducible by C/EBP, with C/EBP␣ having a much greater effect than C/EBP␤ on the activity of pr11␤1(Ϫ1799/ϩ49). 5Ј deletion of 11␤-HSD1 DNA to Ϫ579 had little or no effect on C/EBP␣ inducibility, despite the presence of at least two binding sites (FP9 and FP11) for C/EBP-related proteins in this region. However, the fragment encoding Ϫ812 to Ϫ599 conferred C/EBP␣ inducibility upon the SV40 promoter, demonstrating that FP9, FP11, or both, represent functional C/EBP sites. Furthermore, EMSA analysis using an oligonucleotide corresponding to FP9 demonstrated high affinity binding by rC/EBP␣, at least equivalent to that seen with a consensus C/EBP site. Deletion of 11␤-HSD1 DNA from Ϫ579 to Ϫ322, removing FP5, -6, -7, and -8, reduced the C/EBP␣ induction by 3-fold, with the remainder lost with deletion of the region between Ϫ174 and Ϫ88, removing FP3 and -4. Mutation of either FP3 or FP4 diminished the C/EBP␣ inducibility of constructs containing just the four footprinted regions between Ϫ196 and ϩ49, and mutation of both together had a similar effect on promoter activity to 5Ј deletion of DNA to Ϫ88, confirming the functional importance of FP3 and -4 in mediating the effects of C/EBP␣. The region between Ϫ88 and ϩ49, encoding FP1 and FP2 alone was unable to respond significantly to C/EBP␣. However, it was required for the maximal response of the promoter to C/EBP␣. The magnitude of the C/EBP␣-induction conferred on the SV40 promoter by the fragment encoding Ϫ599 to Ϫ88 (containing FP3-8, all attributable to C/EBP) was considerably less than the induction of the intact 11␤-HSD1 promoter, encoding Ϫ599 to ϩ49. Moreover, the C/EBP␣ inducibility of the proximal 11␤-HSD1 promoter was abolished by mutation of FP1 alone and reduced by mutation of FP2. Together, these data demonstrate that the C/EBP binding sites between Ϫ599 and Ϫ88 represent functional C/EBP sites, which act synergistically with the C/EBP binding sites FP1 and FP2 situated between Ϫ88 and ϩ49.
The positions of FP1 and FP2, within the transcribed region and spanning the transcription start of 11␤-HSD1, respectively, are unusual. As well as reducing the C/EBP␣ inducibility of the proximal promoter, mutation of FP2 also reduced basal transcription by 2-fold. C/EBP may be functioning as an initiator (Inr)-binding protein in binding to FP2 at the transcription start of the 11␤-HSD1 gene. Inr sequences overlap the transcription start site and determine the start of transcription in promoters that lack a TATA box (reviewed in Ref. 35). The 11␤-HSD1 promoter lacks a TATA box; the possibility that C/EBP performs the function of an Inr-binding protein remains to be tested. The C/EBP family of transcription factors contains six members, although only two, C/EBP␣ and C/EBP␤, are expressed at significant levels in liver under basal conditions (36). "Super-shift" EMSA analysis demonstrated that C/EBP␣ was present in the majority of complexes formed on FP2 and FP9, with C/EBP␤ present in a substantial minority, suggesting that these footprints are indeed primarily due to C/EBP␣ and C/EBP␤. In both cases, a minor high mobility complex with C/EBP-related DNA binding specificity was unaffected by addition of C/EBP␣ or C/EBP␤ antisera. This complex is unlikely to contain C/EBP␦ (C/EBP␦ antiserum had no effect on the complex in EMSAs) 2 nor is it likely to contain a degradation product or an alternative translation product of either C/EBP␣ or C/EBP␤ (the antisera recognize epitopes within the leucine zipper/DNA binding domains of the proteins). The identity of the protein present in the high mobility complex thus remains to be determined.
The involvement of members of the C/EBP family in the hepatic regulation of 11␤-HSD1 was confirmed in vivo in mice deficient in either C/EBP␣ (21) or C/EBP␤ (22). In striking contrast to C/EBP␣-deficient mice, which showed a dramatic decrease in hepatic expression of 11␤-HSD1 mRNA, mice that lacked C/EBP␤ showed increased expression of 11␤-HSD1 mRNA in their livers. Although it is possible that the increase in 11␤-HSD1 is secondary to increasing levels of circulating IL-6 in these animals (22), it is more likely that it is due to direct unopposed action of C/EBP␣ on the 11␤-HSD1 promoter. The majority of liver nuclear protein complexes formed on FP2 and FP9 contained homo-or heterodimers of C/EBP␣. However, C/EBP␤ was present in a substantial minority. Similarly, both C/EBP␣ and C/EBP␤ were present in the majority of 2 L. J. S. Williams and K. E. Chapman, unpublished data .
FIG. 5. The liver nuclear proteins that bind to FP2 have C/EBP-related specificity and include C/EBP␣ and C/EBP␤. EM-SAs demonstrating (A) specific binding of rat liver nuclear extract to oligonucleotide O A (encoding FP2), (B) specific binding of rC/EBP␣ to oligonucleotide O A and (C) complexes formed by rat liver nuclear proteins on O A were supershifted by C/EBP␣ or C/EBP␤ antisera. Rat liver nuclear extract or rC/EBP␣ was incubated with 32 P-labeled oligonucleotide in the absence of either competitor DNA or antiserum (lanes 2) or in the presence of a 10-or 100-fold molar excess of competitor oligonucleotide (A, B; lanes [3][4][5][6][7][8][9][10][11][12][13][14] or in the presence of added antisera (C, lanes 3-6) as indicated above the lanes. L indicates the lane contains liver nuclear extract (B, lane 15). Lanes 1 contained no protein extract. Supershifted complexes are indicated by an arrowhead; *, indicates a nonspecific complex.
FIG. 6. Analysis of 11␤-HSD1 mRNA in livers of mice deficient in either C/EBP␣ or C/EBP␤. Autoradiograph of a Northern blot analysis of total RNA isolated from livers of C/EBP␣ (A)-or C/EBP␤ (B)-deficient mice. A, 10 g of total RNA from wild type mice (ϩ/ϩ) or mice homozygous for a targeted deletion of the C/EBP␣ gene (Ϫ/Ϫ); B, 10 g of total RNA from wild type (ϩ/ϩ) mice or mice homozygous for a targeted deletion of the C/EBP␤ gene (Ϫ/Ϫ), hybridized to 32 P-labeled cDNAs encoding mouse 11␤-HSD1 and 7S RNA. C, 11␤-HSD1 mRNA is increased relative to 7S RNA in the livers of mice deficient in C/EBP␤. 11␤-HSD1 mRNA/7S RNA levels are expressed relative to wild type (ϩ/ϩ), nominally set to 100 (*, p Ͻ 0.01; n ϭ 5). complexes formed on the haptoglobin C C/EBP binding site by liver nuclear proteins; these were replaced by complexes containing just C/EBP␣ when livers from C/EBP␤-deficient mice were examined (37). In transiently transfected hepatoma cells, C/EBP␤, itself a relatively weak activator of the 11␤-HSD1 promoter, acted as a dominant negative inhibitor of C/EBP␣. These data are consistent with C/EBP␤ acting as a relative inhibitor of 11␤-HSD1 transcription, at least under basal conditions, in vivo, as in vitro.
C/EBP␣ is a central regulator of energy metabolism (30) and loss of C/EBP␣ in transgenic mice dramatically alters energy metabolism (21). Similarly, C/EBP␤ deficiency is associated with perinatal death due to hypoglycemia (33,38) and, in surviving adults, with immunodeficiency (22). The involvement of both C/EBP␣ and C/EBP␤ (and possibly other members of the C/EBP family) in the hepatic regulation of the 11␤-HSD1 gene suggests that C/EBP indirectly regulates the level of active intracellular glucocorticoids in liver by governing the transcription of hepatic 11␤-HSD1. Glucocorticoids are also important regulators of C/EBP transcription, markedly inducing C/EBP␤ and C/EBP␦ mRNAs (39 -41), and with differing temporal and tissue-specific effects on C/EBP␣ mRNA (42,43). Under basal conditions in liver, the ratio of C/EBP␣ to C/EBP␤ will be high, favoring high levels of 11␤-HSD1 transcription and hence local production of active glucocorticoids. Treatments that decrease the ratio of C/EBP␣/C/EBP␤, for example glucagon (which increases intracellular cAMP) or glucocorticoid administration, would be predicted to decrease 11␤-HSD1 transcription, as found in some studies in vitro (16,44) and in vivo (23,45). In addition, C/EBP␤ is regulated both at the level of translation and post-translationally (reviewed in Refs. 32 and 36). The increased activation of the 11␤-HSD1 promoter in transfected HepG2 cells by the N-terminally truncated form of C/EBP␤ compared with the full-length form suggests that translational and/or post-translational regulation of C/EBP␤ is crucial in determining the ability of C/EBP␤ to transactivate the 11␤-HSD1 gene, and thus, ultimately determining ligand supply to the glucocorticoid receptor.
Mechanisms have been described by which "cross-talk" occurs between glucocorticoid receptor and the immune system, involving direct protein-protein interactions between glucocorticoid receptor and transcriptional regulators of the immune response as well as indirect mechanisms (46 -50). We here describe a novel mechanism by which this cross-talk can occur, both in immune cells and metabolic organs such as liver and adipose tissue. This invokes control of 11␤-HSD1 transcription, and hence steroid ligand availability, by members of the C/EBP family, permitting a complex coordinated control of the networks of genes involved in energy metabolism and the cellular response to stress.