Glucocorticoid Receptor, C/EBP, HNF3, and Protein Kinase A Coordinately Activate the Glucocorticoid Response Unit of the Carbamoylphosphate Synthetase I Gene

Abstract

A single far-upstream enhancer is sufficient to confer hepatocyte-specific, glucocorticoid- and cyclic AMP-inducible periportal expression to the carbamoylphosphate synthetase I (CPS) gene. To identify the mechanism of hormone-dependent activation, the composition and function of the enhancer have been analyzed. DNase I protection and gel mobility shift assays revealed the presence of a cyclic AMP response element, a glucocorticoid response element (GRE), and several sites for the liver-enriched transcription factor families HNF3 and C/EBP. The in vivo relevance of the transcription factors interacting with the enhancer in the regulation of CPS expression in the liver was assessed by the analysis of knockout mice. A strong reduction of CPS mRNA levels was observed in glucocorticoid receptor- and C/EBPα-deficient mice, whereas the CPS mRNA was normally expressed in C/EBPβ knockout mice and in HNF3α and -γ double-knockout mice. (The role of HNFβ could not be assessed, because the corresponding knockout mice die at embryonic day 10). In hepatoma cells, most of the activity of the enhancer is contained within a 103-bp fragment, which depends for its activity on the simultaneous occupation of the GRE, HNF3, and C/EBP sites, thus meeting the requirement of a glucocorticoid response unit. In fibroblast-like CHO cells, on the other hand, the GRE in the CPS enhancer does not cooperate with the C/EBP and HNF3 elements in transactivation of the CPS promoter. In both hepatoma and CHO cells, stimulation of expression by cyclic AMP depends mainly on the integrity of the glucocorticoid pathway, demonstrating cross talk between this pathway and the cyclic AMP (protein kinase A) pathway.

The hepatocytes in the mammalian liver are morphologically very similar but show a remarkable regional heterogeneity in their enzyme content: hepatocytes surrounding the terminal branches of the (afferent) portal vein have an enzymatic phenotype different from that of the hepatocytes around the smallest branches of the (efferent) hepatic vein. Amino acid degradation, gluconeogenesis, conversion of ammonia into urea, and oxidative phosphorylation are found predominantly in the upstream periportal region, whereas glycolysis, xenobiotic metabolism, and glutamine synthesis occur mainly in the downstream pericentral region (28, 29, 42). Key enzymes of these pathways are therefore expressed in gradients from the portal to the central vein. It is believed that this physical separation of mostly opposite metabolic pathways largely avoids futile cycling (e.g., between glycolysis and gluconeogenesis) and is essential for the homeostatic function of the liver.

To gain a better understanding of the establishment of liver architecture, we seek to establish the molecular basis of these differences in porto-central enzyme gradients. Using transgenic mouse studies, we and others have shown that many of these enzyme gradients are regulated at the transcriptional level (3, 6, 35, 41, 53). A genetic dissection of the DNA elements that confer periportal or pericentral expression upon a reporter gene should lead to insight into the signal transduction cascades that regulate gene expression in hepatocytes according to their position along the porto-central axis.

The carbamoylphosphate synthetase I (CPS) gene has proven to be a useful model for such an approach (6). CPS, the first enzyme of the ornithine cycle, converts ammonia, originating from amino acid degradation, into urea (42, 44). In the liver, CPS expression is confined to the hepatocytes surrounding the portal veins (12, 42). In the adult, CPS gene expression is regulated by intracellular levels of cyclic AMP (cAMP) and by glucocorticoids (33, 34, 45). During development, CPS expression is upregulated perinatally (6), presumably due to the late fetal rise in circulating free glucocorticoids and the neonatal rise in intracellular cAMP (19, 23, 34). CPS shares the periportal expression and hormonal regulation with other genes that encode key enzymes in amino acid breakdown, gluconeogenesis, and urea synthesis (27, 28, 42, 44, 66). In vivo, the hepatic expression of CPS is controlled by a single far-upstream enhancer (6), making the gene an attractive model to identify the minimal sequence that is necessary to bring about liver-specific, hormone-inducible periportal expression.

Here we report the characterization of the structure and function of the 469-bp upstream enhancer. The enhancer is shown to comprise a cAMP response element (CRE), a glucocorticoid response element (GRE), and sites for the liver-enriched transcription factor families C/EBP and HNF3. In liver but not in fibroblast cells, the function of the GRE depends on the presence of the adjacent C/EBP and HNF3 sites; i.e., it functions as a glucocorticoid response unit (GRU). The GRU appears to be sufficient for transcriptional activation by glucocorticoids and cAMP, indicating that cross talk between the protein kinase A (PKA) pathway and the glucocorticoid-dependent pathway occurs at this unit, independently of the CRE. The in vivo relevance of factors that were found to interact with the enhancer in vitro was assessed with mice deficient for these factors.

MATERIALS AND METHODS

Nuclear extract preparations, in vitro footprinting and EMSA.

Nuclear extract preparations, in vitro footprinting with DNase I, and electrophoretic mobility shift assays (EMSA), were performed as described previously (14), except that the extracts from highly purified nuclei were used for both footprinting and EMSA. Purification of rat liver HNF3 was performed as described previously (57). Purified glucocorticoid receptor (GR) DNA-binding domain was a gift from K. R. Yamamoto, and purified C/EBPα was a gift from S. L. McKnight.

The oligonucleotides used for EMSA were as follows (complementary strands are not shown): site P1, GTT TAT TAT ATC AGA TAT CCT GTT; site P2, GTG TCC TGG CAC ATG ACC CGG AT; site P3, TGA CAA GTT GAA AAA ACA AGT TCA TC; CRE of CPS enhancer, GTC CTC AAC GTC ATT CTA AA; CRE of tyrosine aminotransferase (TAT) gene (49), AGC TTC TGC GTC AGC GCC AG; TAT C/EBP site s1 (21), AAG CCC AAG GTT TAC CAA TCT CTG C; TAT C/EBP site s3 (21), CTG AAA GTT TCC CCA TGT CCA ACA; and TAT HNF3/GRE site s4 (21), CTA GAA CAA ACA AGT CCT GCG T.

In situ hybridization and RNA analysis.

In situ hybridization on 4% formaldehyde-fixed livers was performed as described previously (43, 65). Serial sections 7 μm thick were probed for CPS and glutamine synthetase (GS) with the respective ³⁵S-labeled cRNA probes (6, 26). Quantification of the signal obtained with the in situ hybridization was performed as described previously (26).

Total RNAs from newborn GR^+/+ and GR^−/− mouse liver and from HNF3α^−/−, HNF3γ^−/−, and HNF3α^−/−, γ^−/− mouse liver were separated on formaldehyde-containing agarose gels for Northern blot analysis as described previously (59). Filters were hybridized with antisense RNA probes in vitro transcribed from rat CPS cDNA (58) and mouse cytochrome oxidase (8) or with CPS and GAPDH (glyceraldehyde-3-phosphate dehydrogenase) DNA probes (11, 58).

Plasmid construction, cell culture, and transfections.

CMV-HNF3α and -β were gifts from J. E. Darnell, 6RGR (GR expression vector) was a gift from K. R. Yamamoto, and pMSV C/EBPα and -β were gifts from S. L. McKnight. Reporter constructs used are based on pSPluc+ (Promega, Madison, Wis.). The 305-bp XbaI-PvuII bovine growth hormone polyadenylation sequence (from pcDNA3; Invitrogen, San Diego, Calif.) was inserted into the XbaI-EcoRV sites in the polylinker 3′ of the modified luciferase cDNA. The CPS promoter (positions −160 to +138 [7]) was inserted into the HindIII-KpnI sites of the upstream polylinker. Fragments derived from the 469-bp enhancer fragment (7) and mutational substitutions were obtained by PCR. To dissect the enhancer, PCR-mediated mutagenesis was performed with primers to create a BamHI site at position 141 or 298 and a PstI site at position 229 or 400 (Fig. 1C). Subsequently, enhancer fragments were inserted into the PstI-BamHI sites of the polylinker upstream of the promoter fragment. To remove the C/EBP site from the 103-bp GRU fragment (positions 298 to 400), a BamHI site was created at position 339. To inactivate the HNF3 site in the GRU fragment (Fig. 1D), oligonucleotide CATCAGTGTAGGCTTTGACA (mutated bases are in italics) was used. First, the 3′ flank was synthesized by using PCR and primer PstI (position 400). The resulting product and primer BamHI (position 298) were used in a second PCR to synthesize the complete GRU with the mutations. All fragments were verified by sequence analysis.

FIG. 1.

In vitro DNase I protection analysis of the upper strand of the 469-bp enhancer fragment. (A) Footprint analysis with purified factors. Lane 1, chemical cleavage at pyrimidines (Py). Lane 2, control (C): naked DNA treated with DNase I. Lanes 3, 4, and 5, DNA incubated with purified HNF3, purified GR DNA-binding domain (GR), or purified C/EBPα. Lane 6, DNA incubated with 40 μg of liver nuclear extract (NE). Vertical lines indicate the positions of protected regions. Arrows indicate the positions of the typical DNase I-hypersensitive sites obtained with either purified HNF3 or liver nuclear extract (55). The footprints shown were confirmed on the lower strand (not shown). (B) Competition footprint analysis with liver nuclear extract and competition with unlabeled oligonucleotides. Lane 1, control (C): naked DNA treated with DNase I. Lanes 2 to 4, DNA incubated with 40 μg of liver nuclear extract and 30 ng of HNF3, C/EBP, or no competitor, respectively. (C) Sequences in the 469-bp enhancer fragment that are protected by transcription factors in footprint analyses (horizontal lines). Arrowheads show the positions of DNase I-hypersensitive sites in the HNF3 sites. Boxes indicate the positions of the CRE and GRE half-sites. (D) Schematic representation of the structure of the enhancer, based on the results obtained with the footprint analysis and functional assays. Black circles, positions of sites P1, P2, and P3; black boxes, positions of the HNF3 sites; hatched ovals, C/EBP sites; gray ovals, GR-binding site (GRE) and CRE. The grouping of sites into a GRU is shown.

FTO-2B and CHO-K1 cells were grown in DMEM/F12 (Gibco BRL) supplemented with 10% fetal calf serum as described previously (7). Transfection of FTO-2B and CHO-K1 cells and luciferase and chloramphenicol acetyltransferase activity assays were performed as described previously (7). Sixteen hours after transfection, the medium was changed and, when indicated, supplemented with 100 nM dexamethasone and/or 0.25 mM chlorothiophenyl-cAMP (ctp-cAMP) for another 24 h prior to extract preparation.

RESULTS

The enhancer at −6.3 kbp has sites recognized by the liver-enriched factors HNF3 and C/EBP and by the GR.

The 469-bp enhancer, in conjunction with the CPS promoter, provides hepatocyte-specific, hormone-dependent activation of expression of a reporter gene (7). We have used DNase I protection assays to study this enhancer. Both strands were analyzed by using nuclear extracts prepared from rat liver, purified HNF3, recombinant C/EBPα, and GR DNA-binding domain (38). Figure 1A shows eight of the nine protected sites obtained with the liver nuclear extract. Of these, three sites each correspond to protections obtained with purified C/EBPα and with purified HNF3 (Fig. 1A and C). Protected sites obtained with purified HNF3 and the corresponding sites obtained with whole liver nuclear extracts both show the characteristic DNase I hypersensitivity in the middle of the recognition site (21, 57). Incubation of purified GR with the enhancer resulted in two protected areas which were not visible when liver nuclear extracts were used, one around position 280 of the enhancer (GRE I) and one around position 390 (GRE II). Only the site around position 390 corresponds to a site with good similarity to the well-defined recognition site of the GR (positions 382 to 396; AGAGCANNNTGTTCT). The other protected region (GRE I) contains a sequence resembling a half-site and may be occupied by the monomer of the GR-binding domain. Competition footprinting with unlabeled oligonucleotides corresponding to binding sites for HNF3 and C/EBP showed that both oligonucleotides successfully compete with the respective sites in the enhancer (Fig. 1B). This demonstrates that the corresponding binding sites are not occupied by other factors present in the liver extract.

The three footprints formed with the liver extract, P1, P2, and P3, were also formed with HeLa cell nuclear extracts (not shown). The corresponding sites were further analyzed by EMSA with either liver or HeLa cell nuclear extract (Fig. 2). Similar, but not identical, patterns were obtained for each site with the two nuclear extracts. The complexes obtained with each site, however, differed substantially, indicating that these sites were recognized by a different set of proteins. Indeed, each of the sites was unable to compete for the other two, both in EMSA and in footprint assays performed with both nuclear extracts (not shown). Comparison of sites P1, P2, and P3 with the transcription factor database showed some similarities with sites for HNF1 (P1), HNF4 (P2), NF1 (P2), and AP1 (P2). However, none of the P1, P2, and P3 complexes were competed by bona fide binding sites for any of these factors (not shown).

FIG. 2.

Characterization of binding activities to sites P1, P2, and P3 in liver and HeLa cell nuclear extracts. EMSA were performed with oligonucleotides corresponding to sites P1 (lanes 1 to 6), P2 (lanes 7 to 12), and P3 (lanes 13 to 18). The probes were incubated with 2 μg of liver or HeLa cell nuclear extracts (L and H, respectively) and 10 ng of competitor oligonucleotides (P1, P2, and P3, respectively), as indicated on the top.

The CPS CRE and the CRE of the TAT gene are recognized by similar factors.

A putative CRE, which is functional in transient-transfection assays (7), is present at positions 148 to 155. This CRE was only weakly protected in the footprint assay but gave rise to retarded complexes when analyzed with EMSA (Fig. 3). The CPS CRE was analyzed in parallel with the well-characterized CRE of the TAT gene enhancer at −3.6 kbp (49). Both probes gave rise to retarded complexes of similar size, and cross-competition between the probes was observed (Fig. 3). Two minor complexes (B and C) formed with the CPS CRE were not competed by the TAT CRE. These results, together with the functional assay (7), indicate the presence of a bona fide CRE in the CPS enhancer.

FIG. 3.

Characterization of binding activities to the CPS CRE. EMSA were performed with an oligonucleotide corresponding to the CRE of CPS or the CRE of TAT (49) as a probe. The probes were incubated with hepatoma cell nuclear extracts and competitor oligonucleotides as indicated at the top. The positions of the various retarded bands are indicated by arrows. Complex A is formed both with the CPS CRE and TAT CRE probes. Complexes B and C are specific for the CPS CRE probe. NS, nonspecific.

Dissection of the enhancer into active regions in FTO-2B hepatoma cells.

To identify the smallest functionally active region, the enhancer was further dissected, and the resulting fragments were tested by transient transfections in FTO-2B hepatoma cells (Fig. 4). The promoter without enhancer (Fig. 4, construct 1) was not significantly stimulated by any of the hormones added. In the absence of hormones, the enhancer (Fig. 4, construct 2) did not add to the basal activity of the promoter, while addition of cAMP alone was only slightly effective. In contrast, dexamethasone activated the enhancer 7-fold over basal activity, and the addition of cAMP caused an additional stimulation to 17-fold over basal activity. Shortening of the enhancer by removal of site P1 and the HNF3 site 3 (construct 3 in Fig. 4) caused a twofold decrease of the basal activity but did not affect the hormone-induced activity of the enhancer. Cotransfection of an expression vector for GR (Fig. 4, construct 3) revealed that overexpression of GR enhanced basal and hormone-induced activities of this construct by approximately 35%.

FIG. 4.

Dissection of the CPS enhancer into functional regions. Transient-transfection assays of the enhancer and fragments thereof in FTO-2B hepatoma cells were performed. The left part shows the constructs used. The CPS promoter (positions −161 to +138) (construct 1; gray box with arrow at transcription start site) was used in all experiments, without (as a control) or in conjunction with various portions of the enhancer (constructs 2 to 5, detailed in Fig. 1C and D). Mutations were introduced into the GRE (construct 6), the HNF3 site (construct 7), and the C/EBP site (construct 8) of the GRU (detailed in Fig. 1C and D). Where indicated (GR), an expression vector for rat GR was cotransfected. White bars, luciferase activity in the absence of hormones; hatched bars, activity in the presence of 0.25 mM ctp-cAMP; cross-hatched bars, activity in the presence of 0.1 mM dexamethasone; black bars, activity in the presence of both ctp-cAMP and dexamethasone. The effect of ctp-cAMP alone was not tested in transfections with constructs 6 to 8 (ND). The activity of the promoter alone in the absence of hormones was set to 1. Error bars indicate standard errors of the mean.

The smaller, fully active enhancer fragment (Fig. 4, construct 3; Fig. 1C, positions 141 to 400) was divided into a fragment from position 141 to 229 that contains the CRE and a fragment from position 298 to 400 that contains the GRE. The CRE-containing fragment (Fig. 4, construct 4) had a basal activity similar to that of the full-size enhancer and was stimulated slightly by cAMP, about twofold over basal activity by dexamethasone, and sixfold over basal activity by the combination of both agents. This CRE-containing fragment represents approximately 30% of the hormone-induced activity of the full-size CPS enhancer (compare constructs 2, 3, and 4 in Fig. 4).

The hormonal stimulation pattern of the GRE-containing fragment (Fig. 4, construct 5) was very similar to that of the enhancer fragment. Like the shortened enhancer (construct 3), its basal activity was decreased 2-fold but was strongly stimulated upon addition of dexamethasone (20-fold over basal activity) and of the combination of both dexamethasone and cAMP (30-fold over basal activity). The activity of this GRE-containing fragment represents approximately 75% of that of the full-size and shortened enhancers (constructs 2 and 3). Thus, the GRE fragment accounts for most of the hormonal response of the far-upstream CPS enhancer.

The GRE fragment is a functional GRU.

To test whether sites within the 103-bp GRE fragment (Fig. 4, construct 5) each independently add to the activity of the fragment or act cooperatively, the GRE, HNF3, and C/EBP elements were each inactivated (Fig. 4, constructs 6 to 8, respectively). Mutation of one nucleotide in the inverted repeat of the GRE that is necessary for GR binding (68) and activation (52) caused a fivefold decrease in the dexamethasone-stimulated expression (construct 6). Two substitution mutations in the HNF3 site, shown to effectively inactivate this site (57), caused a 10-fold reduction in the response to hormones as well (construct 7). Deletion of the C/EBP site, which does not result in alterations in the distance between other sites, almost completely abolished the induction by hormones (construct 8). Therefore, all three sites need to be intact for the fragment to be active. This region can therefore be referred to as an autonomously active GRU.

The GRE in the CPS enhancer does not cooperate with the C/EBP and HNF3 elements in CHO cells.

The contributions of GR, HNF3, and C/EBP to the activity of the 103-bp GRE fragment were tested in (non-liver-derived) CHO cells. In CHO cells, GR, C/EBPα, C/EBPβ, HNFα, and HNF3β mRNA levels were 2.8-, 1.9-, 1.6-, 0.35-, and 0.25-fold, respectively, the levels found in adult rat liver (not shown). Cotransfection of different amounts (3 ng to 3 μg) of an expression vector encoding rat HNF3α or -β resulted in a dose-dependent, maximally twofold reduction of basal and hormone-induced activities of the CPS promoter (not shown), whereas under the same conditions, overexpression of C/EBPα activated the CPS promoter about twofold (not shown). Coexpression of either HNF3 or C/EBPα had no effect on enhancer activity in the absence or presence of hormones (not shown), whereas coexpression of GR activated the enhancer approximately 7-fold over basal activity in the presence of dexamethasone and 14-fold over basal activity in the presence of dexamethasone and cAMP (Fig. 5, construct 5). To test to what extent the GRE, C/EBP, and HNF3 sites contribute to the enhancer activity of construct 5 in CHO cells, expression of constructs 6 to 8 of Fig. 4 was assayed (Fig. 5, constructs 6 to 8). As in hepatoma cells, mutation of the GRE (Fig. 5, construct 6) caused a fivefold decrease in hormone-induced expression relative to the control (construct 5). However, inactivation of either the HNF3 site (construct 7) or the C/EBP site (construct 8) resulted in a reduction to only 60 to 85% of control activity (construct 5). These results show that CHO cells differ from FTO-2B hepatoma cells in that the HNF3 and C/EBP elements, or the availability of their corresponding transcription factors, do not contribute to the activity of the CPS enhancer.

FIG. 5.

Functional analysis of the GRE fragment of the CPS enhancer in CHO-1 cells. See the legend for Fig. 4 for explanation.

Importance of GR, HNF3α, HNF3γ, C/EBPα, and C/EBPβ in vivo.

To assess the in vivo relevance of transcription factors interacting with the GRU in the regulation of CPS gene expression, we compared CPS mRNA levels in wild-type and mutant neonatal mice with targeted disruptions of the genes encoding some of these transcription factors. Newborn mice are well suited for such analysis, because CPS mRNA levels are induced at birth (11), probably as a result of the activation of the CPS GRU (60).

In livers of neonatal mice deficient for GR (8), CPS mRNA levels were reduced three- to fourfold, as determined by Northern blot analysis (Fig. 6). Livers of neonatal HNF3α and -γ knockout mice (30) were tested similarly (Fig. 7). In livers deficient for either HNF3α or -γ, CPS mRNA concentrations in wild-type and homozygous knockout mice were comparable. HNF3β could not be tested in this way, because its deficiency is lethal at embryonic day 10 to 11 in mice (2, 71). In HNF3α/γ double mutants, which die within the first week after birth, CPS mRNA levels were also normal 2 days after birth.

FIG. 6.

Northern blot analysis of CPS mRNA from livers of GR wild-type and knockout mice. Total RNAs (20 μg) from newborn wild-type (+/+) and knockout (−/−) mice were analyzed with cRNA probes for rat CPS and for mouse cytochrome oxidase (COX) to control for RNA loading. The right panel shows the quantified signal (CPS signal/COX signal). Error bars indicate standard errors of the mean.

FIG. 7.

Northern blot analysis of CPS mRNA from livers of HNF3α-, HNF3γ-, and HNF3α/γ-deficient mice. Total RNAs (20 μg) from neonatal day 2 (ND2), ND8, and adult wild-type (+/+) and knockout (−/−) mice were analyzed with cDNA probes for rat CPS and for GAPDH to control for RNA loading. The right panel shows the quantified CPS signal (CPS signal/GAPDH signal).

CPS mRNA levels in livers of newborn mice homozygous for a deletion in the C/EBPα gene (70) were compared with those in heterozygous and wild-type littermates, using quantitative in situ hybridization with cRNA probes (Fig. 8A and B). Quantification of the optical density shows that CPS mRNA levels are decreased approximately 10-fold in the homozygous C/EBPα-deficient mice (Fig. 8I). The effect is specific for CPS, because the mRNA levels for GS and argininosuccinate synthetase, another ornithine cycle enzyme, were not affected (Fig. 8C and D and data not shown). The zonation of expression of both CPS and GS is perturbed in the C/EBPα knockout mice. Normally, CPS is expressed in the hepatocytes surrounding the terminal portal veins, and GS is expressed in the layer of two or three hepatocytes surrounding the central veins (Fig. 8A and C) (18). In the mutant mice, both genes are expressed outside their respective territories (Fig. 8B and D). In addition, the normal arrangement of the hepatocytes into plates is replaced by rosette-like structures in the C/EBPα knockout mice (16, 70). These results emphasize the importance of C/EBPα for the establishment of mature liver architecture and gene expression patterns. Livers of C/EBPβ-deficient mice were tested similarly. Both CPS (Fig. 8E and F) and GS (Fig. 8G and H) mRNA levels were similar in livers of mice heterozygous for the mutation and in those of knockout mice (Fig. 8J). Also, the zonation patterns of both mRNAs and liver morphology were comparable in heterozygous and knockout mouse livers.

FIG. 8.

Distribution and abundance of CPS RNA in livers of neonatal C/EBPα- and C/EBPβ-deficient mice. (A to D) Serial sections of livers of newborn heterozygous (A and C) and homozygous (B and D) C/EBPα-deficient littermates, hybridized with a ³⁵S-labeled cRNA probe for CPS mRNA (A and B) or GS mRNA (C and D). The livers of wild-type mice are very similar to the livers of heterozygous littermates and are not shown. The morphology in the knockout mice is abnormal, showing tubular rosettes of hepatocytes (B and D). Furthermore, the gradient in the distribution of CPS RNA from the portal to the central vein and the GS RNA localization around the central vein that is seen in the livers of heterozygous mice are disturbed in the livers of knockout mice. (E to H) Serial sections of livers of newborn heterozygous (E and G) and C/EBPβ-deficient (F and H) littermates, hybridized with a ³⁵S-labeled cRNA probe for CPS mRNA (E and F) or GS mRNA (G and H). Bar, 0.2 mm. (I and J) Quantitation of the CPS and GS in situ hybridization signals on sections from two or three littermates per group of C/EBPα-deficient mice (I) and C/EBPβ-deficient mice (J). Optical densities in two independent series of in situ hybridizations were registered with a charge-coupled device camera and quantified. The density of the tissue is subtracted as background from the total density.

DISCUSSION

The 469-bp far-upstream enhancer fragment of the CPS gene is both necessary and sufficient for liver-specific, hormone-dependent periportal expression (6), making the gene an attractive model for defining the mechanism underlying liver-specific, periportally localized, and hormone-activated gene expression. We did not include an analysis of the proximal promoter in this study for several reasons. In this and previous studies, the promoter was shown to lack any cell specificity and hormonal responsiveness in transient transfections (7, 69). Furthermore, the enhancer was also functional when combined with a heterologous promoter (7), suggesting that the enhancer is sufficient for tissue-specific, hormone-dependent activation of expression. In the present study we further characterized the 469-bp far-upstream enhancer.

Footprint analysis with rat liver and HeLa cell nuclear extracts, competing oligonucleotides, and purified factors provided evidence for binding of C/EBP, HNF3, and GR. Furthermore, three sites for different, as-yet-unidentified, non-liver-specific factors were observed. We positively excluded NF1, HNF1, HNF4, Ap1, NFY, Coup-TF, Oct, PEA3, and CACCC box binding factor as candidates.

The GR and liver-enriched factors mediate tissue-specific and hormone-dependent activation of CPS in vitro and in vivo.

A liver-specific, direct response to glucocorticoids and cAMP is thought to be mediated by hormone response units (HRUs) in the regulatory regions of genes. HRUs are loci at which the activity of ubiquitous hormone-responsive factors depends on the nearby binding of other transcription factors. HRUs of genes expressed in liver generally require binding sites for members of one of the liver-enriched transcription factor families, like HNF3 or HNF4 and C/EBP (21, 37, 51, 55, 61, 63). Our data show that the region between positions 320 and 400 of the 469-bp enhancer fragment (Fig. 1) meets the requirement for a GRU. The unit, composed of a GRE, an HNF3 site, a C/EBP site, and a site recognized by an unidentified factor, was shown to be an autonomously functional region within the enhancer. Inactivation of either the C/EBP site, the HNF3 site, or the GRE abolished the activity of the GRU in FTO-2B cells, confirming the involvement of these factors in the activation of the GRU (Fig. 4).

To test whether the cooperative interaction between the occupied GRE, C/EBP, and HNF3 sites constrained the activity of the CPS enhancer in a tissue-specific way, expression of the enhancer in hepatoma cells was compared to that in (non-liver-derived) CHO cells (Fig. 5). In contrast to the case for hepatoma cells, hormone-dependent transactivation of the CPS promoter was affected only by mutation of the GRE in these fibroblast-like cells, whereas deletion of the C/EBP site and mutation of the HNF3 site did not affect the transactivation activity. Furthermore, overexpression of C/EBP (α or β), HNF3 (α or β), or the combination of these factors did not increase reporter gene expression from the intact CPS enhancer, in either the absence or presence of overexpressed GR. These findings show that the GRE, in conjunction with the C/EBP and HNF3 sites, functions as a sensitive HRU only in the context of hepatoma cells and imply the existence of an as-yet-unidentified protein(s) in these cells that confers cooperatively upon the binding properties of GR, HNF3, and C/EBP to the GRU or that mediates their transactivation. The finding that the CPS enhancer can be activated by hormones in CHO cells in the presence of very high levels of GR, that is, by overexpression of GR beyond the endogenous levels in CHO cells (72), shows that CHO cells can be used in complementation assays to identify this accessory factor(s).

To confirm the in vivo role of GR, C/EBP, and HNF3, the CPS mRNA contents in livers of neonatal mice deficient for these factors were determined (Fig. 6 to 8). Hepatic CPS expression is induced at birth in response to high circulating levels of glucocorticoids and glucagon (reviewed in reference 42). Our observations therefore reveal the capacity of the transcription factor-deficient animals to express CPS. As predicted by the in vitro results, both GR^−/− mice and C/EBPα^−/− mice had strongly reduced CPS mRNA levels, indicating that changes in the concentrations of these factors control the activation of CPS gene expression in vivo. A dominant role for C/EBPα in liver-specific gene expression is also indicated by other mouse models. Both albino lethal mice and mice with juvenile visceral steatosis have reduced CPS and C/EBPα levels, whereas other transcription factors are normally expressed (58, 67). Both the levels and positions of CPS and GS gene expression and liver morphology were normal in C/EBPβ-deficient neonates, indicating either that C/EBPβ is not involved in CPS expression or that other members of the C/EBP family can take over its function in liver. In agreement with this conclusion, CPS mRNA can be induced in adult C/EBPβ-deficient mice by glucocorticoids and fasting (not shown). CPS expression is unaffected in HNF3α and -γ knockout mice and even in the HNF3α/γ double-knockout mice, which die in the first postnatal week. Furthermore, we observed that CPS levels were similar in adult, 48-h-fasted, dexamethasone-treated control and HNF3γ-deficient animals (not shown). HNF3α and -γ deficiency may be compensated for by HNF3β, which could not be studied because inactivation of this gene causes lethality at embryonic day 10 (2, 71).

Mechanism of glucocorticoid-dependent CPS gene transcription.

In rat and mouse liver, as well as in cultured hepatocytes and FTO-2B hepatoma cells, CPS gene transcription is stimulated by glucocorticoids and the synthetic analog dexamethasone (31, 34, 44, 48). CPS and other ornithine cycle enzymes are induced by dexamethasone with delayed onset in cultures of primary hepatocytes (48). Protein synthesis inhibitors (cycloheximide) repress the induction of ornithine cycle genes, including the CPS gene, and the phosphoenolpyruvate carboxykinase gene (47, 48). This has been interpreted to mean that de novo protein synthesis is required (48) and that the response is secondary, with the induction of CPS depending on the glucocorticoid-induced synthesis of an activator. However, our present analysis shows that (i) the DNA-binding domain of GR interacts directly with the GRE in the enhancer (Fig. 1A), (ii) a substitutional mutation in the GRE markedly reduces glucocorticoid induction (Fig. 4 and 5), and (iii) GR is essential for dexamethasone-induced activation of the GRE (Fig. 5). These observations are consistent with a direct response to glucocorticoids. Our findings therefore show that at least part of the response of CPS to glucocorticoids is primary and is mediated by the GRE in the enhancer. This conclusion does not rule out the possibility that de novo synthesis of a factor(s) is necessary to realize a full response. The expression of C/EBPα and -β is, for example, rapidly induced by glucocorticoids and cAMP and is involved in the hormone-induced expression of several genes (9, 10, 20, 39, 50). A crucial role for members of the C/EBP family of transcription factors in achieving the full response of the GRU fits with our observation that the C/EBP site within the GRU is essential for the function of the unit and, hence, for the activation by glucocorticoids.

Activation by cAMP and glucocorticoids is synergistic and does not require the CRE: cross talk between the PKA and glucocorticoid pathways at the GRU.

Both in vivo and in vitro, induction of CPS expression by cAMP or glucocorticoids alone is moderate, with the combination of both hormones being necessary to cause a strong synergistic response (34, 44, 45). The cAMP-mediated stimulation of CPS expression is mediated by PKA (32), since downregulation of PKA by overexpression of the regulatory subunit RIα decreases CPS mRNA levels in hepatoma hybrids (5, 25, 58). In accordance with this model, we have previously shown that the integrity of the CRE is necessary for full activity of the CPS enhancer (7). Nevertheless, its activation by cAMP alone is remarkably weak (Fig. 4). Presently, we do not know why the CRE, present in the enhancer, is not sufficient for stimulation by cAMP alone. However, Nitsch et al. (51) have shown that the CRE in the TAT gene operates in synergy with an adjacent HNF4 site. Mutation of the HNF4 site causes a complete loss of induction by cAMP, and replacement of the HNF4 site by various HNF3 sites does not restore the activity. The CRE in the CPS enhancer might be unable to synergize with the adjacent HNF3 site and therefore shows only a weak response to cAMP.

In contrast to the weak stimulation of glucocorticoids or cAMP alone, a strong synergistic response is achieved by the combination of both hormones (34, 44, 45). Several mechanisms can account for the combinatorial action of these two hormones. The classical model is that cAMP and glucocorticoids act via distinct pathways which activate transcription through separate hormone response elements, the CRE (73) and the GRE (37, 63). Our analysis shows, however, that the activation of the CPS enhancer by cAMP depends mainly on the integrity of the glucocorticoid pathway: (i) activation of both the complete enhancer and the GRE fragment by cAMP requires the presence of dexamethasone in FTO-2B cells (Fig. 4) and CHO cells (Fig. 5), and (ii) the presence of a GRE alone is sufficient for activation by both hormones, that is, does not require the simultaneous presence of a CRE (Fig. 4 and 5). Moreover, the finding (5a) that the activation of the CPS GRE fragment by glucocorticoids is fourfold stronger in FTO-2B cells, which have a constitutively high PKA activity (5, 25), than in FTO-2B cells overexpressing the regulatory subunit RIα of PKA (WT-8 cells), which causes repression of PKA activity (25), also demonstrates cross talk between the pathways. Thus, our present analysis clearly suggests that the glucocorticoid and cAMP signal transduction pathways are integrated further upstream in the cascade and converge upon a single DNA-binding region.

Synergism between hormone response elements and liver-specific transcription factors was shown to be necessary for full activity of the TAT enhancer in hepatoma cells (51). Our observation that a dimer of an optimized GRE is insufficient to confer sensitivity to fasting upon a reporter gene, whereas the TAT GRU is (60), indicates that the DNA-binding region that is necessary to integrate the glucagon and glucocorticoid signal transduction pathways requires, in addition to a GRE, binding sites for C/EBP and HNF3. A possible mechanism for the cross talk between the cAMP (PKA) pathway and the glucocorticoid pathway (15, 24, 46, 54, 74) may then be one in which glucocorticoids determine the activation of the enhancer and PKA modulates the degree of this stimulation. A cAMP-dependent increase in the trans-activating potential or stability of the protein-DNA interaction of GR and other factors, such as HNF3 and C/EBP family members, has been reported (15, 36, 54, 56), as well as cAMP-mediated induction of expression of GR or other factors, such as C/EBP family members (39). Moreover, PKA activation in FTO-2B and WT8 hepatoma cells has been shown to both stabilize the binding of GR and HNF3 to the TAT GRU at −2.5 kbp and to enhance the glucocorticoid response via this GRU in WT8 hepatoma cells (15). The glucocorticoid-induced activation of the CPS enhancer fragment that lacks the GRE (Fig. 4, construct 4), on the other hand, may well be indirect and reflect glucocorticoid-induced expression of transcription factors such as C/EBP (9, 20), for which binding sites are present in the CRE-containing fragment.

Based on our results, we hypothesize a crucial role for the GRU within the enhancer in the tissue-specific activation of the CPS gene. In this model, ligand-bound GR will initiate the assembly of factors such as C/EBP and HNF3 at the GRE fragment, by modulation of the chromatin structure and/or direct protein-protein interactions (1, 4, 40). The rate of CPS expression can be adapted by glucagon (cAMP), via its effect on the glucocorticoid-dependent activation complex. This sequence in the activation of a GRU-containing enhancer would nicely explain the 20-year-old observation that exposure to glucocorticoids has to precede that to cAMP to obtain a full-scale biological effect (13, 17, 22, 62, 64). The ability of the GRU to activate hormone-dependent transcription in vivo is currently being investigated with transgenic mice.