A single regulatory module of the carbamoylphosphate synthetase I gene executes its hepatic program of expression.

A 469-base pair (bp) upstream regulatory fragment (URF) and the proximal promoter of the carbamoylphosphate synthetase I (CPS) gene were analyzed for their role in the regulation of spatial, developmental, and hormone-induced expression in vivo. The URF is essential and sufficient for hepatocyte-specific expression, periportal localization, perinatal activation and induction by glucocorticoids, and cAMP in transgenic mice. Before birth, the transgene is silent but can be induced by cAMP and glucocorticoids, indicating that these compounds are responsible for the activation of expression at birth. A 102-bp glucocorticoid response unit within the URF, containing binding sites for HNF3, C/EBP, and the glucocorticoid receptor, is the main determinant of the hepatocyte-specific and hormone-controlled activity. Additional sequences are required for a productive interaction between this minimal response unit and the core CPS promoter. These results show that the 469-bp URF, and probably only the 102-bp glucocorticoid response unit, functions as a regulatory module, in that it autonomously executes a correct spatial, developmental and hormonal program of CPS expression in the liver.

In liver, many metabolic pathways are predominantly found in either the upstream, periportal region (e.g. gluconeogenesis) or the downstream, pericentral region (e.g. glycolysis) (1)(2)(3). Key enzymes of these pathways are therefore expressed in gradients along the sinusoids that connect the portal and the central vein. These zonal differences in gene expression largely avoid futile cycling of opposite metabolic processes (such as between glycolysis and gluconeogenesis) but also comprise complementary functions such as the low and high affinity detoxifying functions of the urea cycle and glutamine synthetase, respectively (1,2,4). Because the heterogeneous expression of genes in mammalian liver seems to relate to the function of the organ, much effort has been invested in understanding the mechanisms underlying its regulation. Nevertheless, it remains obscure what causes these differences in porto-central enzyme gradients and how critical the maintenance of these gradients is for preservation of normal liver function. For several hepatic genes it was shown that their localization within the liver is regulated at the transcriptional level (5)(6)(7)(8). For that reason, we decided to embark upon a genetic dissection of the regulatory DNA elements that confer periportal expression upon a reporter gene.
The carbamoylphosphate synthetase (CPS) 1 I gene has proven to be a very convenient model for such an approach (7). The CPS gene is expressed selectively in the periportal hepatocytes of the liver and, in the epithelial cells of the villi of the small intestine (2,9), is induced perinatally (7) and is tightly controlled by glucocorticoids and intracellular cAMP levels (10 -12). CPS shares the periportal expression and hormonal regulation with other genes that encode key enzymes in amino acid catabolism, gluconeogenesis, and urea synthesis (2,10,(13)(14)(15)(16)(17). The 12-kbp upstream DNA fragment of the CPS gene has been shown to contain all information required for the proper control of expression in vivo (7). We previously isolated a 469-bp fragment at 6 kbp upstream of the transcription initiation site which, together with a 0.3-kbp (Ϫ161 to ϩ138) proximal promoter fragment, confers hepatocyte specificity upon a reporter gene (7,18,19). This upstream regulatory fragment (URF) contains multiple recognition sites for HNF3 and C/EBP family members, the glucocorticoid receptor (GR) and several unknown factors. The GR recognition site is present within a 102-bp fragment, together with a C/EBP and a HNF3 site, and forms a functional response unit (20). Here, we present an experimental analysis of the regulatory functions of the 469-bp URF and the 0.3-kbp promoter in transient transfection assays and in transgenic mice. We found that the URF meets the criteria for being a regulatory module (21), which autonomously executes all regulatory functions in liver that cause the typical expression pattern of the endogenous CPS gene. Within this fragment, a 102-bp subfragment comprising the GRU controls hepatocyte-specific, hormone-controlled expression in vitro and, in addition, periportal expression in vivo.

Plasmids
Constructs 1-10 are based on a pBluescript-derived vector containing the promoterless luciferase gene (1.8-kbp HindIII/EcoRI lucϩ fragment of pSPlucϩ (Promega, Madison, WI)) in conjunction with the SV40 small T-antigen intron and polyadenylation signal, optimized for expression (22). Construct 1 contains the fully active genomic rat CPS I gene promoter from Ϫ161 to ϩ138 bp, containing sites I, II, and III (19,23,24), the GAG box, the TATA motif, the transcription initiation site, and 138 bp of the 140-bp 5Ј-untranslated region, upstream of the lucϩ gene (see Fig. 1). Construct 2 contains the 469-bp genomic rat CPS I enhancer fragment (19), in conjunction with the complete CPS promoter. Construct 3 is identical, except that the promoter extends from Ϫ74 to ϩ138, that is, lacks sites I, II, and III. The promoter of construct 4 extends from Ϫ38 to ϩ138, that is, additionally lacks the GAG box (19,25). Constructs 5-7 are comparable with constructs 2-4, respectively, except that the URF is replaced by the 102-bp GRU fragment (20). * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
¶ To whom correspondence should be addressed: Dept. of Anatomy and Embryology, Academic Medical Center, University of Amsterdam, Meibergdreef 15, 1105 AZ, Amsterdam, The Netherlands. Fax: 31-20-6976177; E-mail: w.h.lamers@amc.uva.nl. Construct 8 is identical to construct 4, except that the sequences of the URF upstream of the GRU were removed. Constructs 9 and 10 are comparable with construct 7, except that they contain the minimal heat shock 70 gene promoter and the minimal cytomegalovirus promoter, respectively. The minimal heat shock 70 gene promoter was isolated as a 315-bp SmaI-HindIII fragment from pIND (Invitrogen). The minimal cytomegalovirus promoter was isolated from pCI (Promega) by generating a BamHI site at position Ϫ55 relative to the transcription start site and a NcoI site at ϩ320 including the TATA box, transcription start side, 5Ј-untranslated region and chimeric intron. The various parts of the URF and promoter were generated by polymerase chain reaction. All constructs were checked by sequence analysis. The plasmids were isolated using Jetstar columns (Genomed, Bad Ö ynhausen, Germany) according to the manufacturer's instructions.

Transient Transfections
FTO-2B hepatoma, NIH3T3, and CHO-K1 cells were grown in Dulbecco's modified Eagle's medium/Ham's F-12 medium (Life Technologies, Inc.), supplemented with 10% fetal calf serum as described (19). Transfection of FTO-2B and CHO-K1 cells and luciferase and ␤-galactosidase activity assays were performed as described (19). 16 h after transfection medium was changed and, when indicated, supplemented with 100 nM dexamethasone and/or 0.25 mM chlorothiophenyl cAMP for another 24 h prior to extract preparation.

Generation and Screening of Transgenic Mice
Four constructs were used for the generation of transgenic mice as described (26). The lines resulting from constructs 2-4 were named E, G, and T, respectively. Construct 5 was modified to contain, in addition to the 102-bp GRU and the extended CPS promoter (Ϫ161 to ϩ138), a 0.2-kbp chimeric intron fragment from pCI (Promega), a nuclear targeted lacZ reporter gene, and the 0.3-kbp polyadenylation signal from the bovine growth hormone gene from pcDNA3 (Invitrogen, Carlsbad, CA). The resulting transgenic line was named GRU-E. Transgenic offspring was identified by polymerase chain reaction on tail tip lysates, using primers specific for SV40 sequences (SV40ϩ, CAG GCA TAG AGT GTC TGC and SV40-, CTG GGG ATC CAG ACA TGA) or lacZ sequences (lacZϩ, GCA TCG AGC TGG GTA ATA AGC GTT GGC AAT and lacZϪ, ACT GCA ACA ACG CTG CTT CGG CCT GGT AAT). Copy numbers of constructs integrated into the genome were determined by Southern blot analysis of 10 g of tail tip DNA digested with PstI.

Hormonal Treatment of Animals
Induction of the Hepatic Cyclic AMP Level-To increase hepatocellular cAMP levels in adult mice, animals were starved for 48 h (27). For prenatal cAMP treatment, fetuses of timed pregnant females were exposed by midabdominal incision and injected with 50 mg/kg body weight (1 ϫ 10 Ϫ4 M) chlorothiophenyl cAMP in physiological saline. Control females were laparotomized, and the fetuses were injected with physiological saline alone. After 6 h, the livers of the fetuses were isolated for luciferase activity determination.
Manipulation of Circulating Glucocorticoids-Mice were injected intraperitoneally with dexamethasone phosphate in phosphate-buffered saline (5 mg/kg body weight) 16 h before sacrifice. For prenatal dexamethasone treatment, timed pregnant females were injected with 5 mg/kg body weight (1 ϫ 10 Ϫ5 M) dexamethasone phosphate in phosphate-buffered saline. After 6 h, the livers of the fetuses were isolated for luciferase activity determination. Mice were adrenalectomized via the dorsal approach. Adrenalectomized and control animals had access to water supplemented with 0.7% (w/v) NaCl and 4% (w/v) sucrose and were sacrificed after 7 days.

Animal Care
Animals were housed with a 12-h light and 12-h dark cycle, and permitted ad libitum access to water and standard pellet-type diet. This study was performed in accordance with the Dutch guidelines for the use of experimental animals.

Reporter Enzyme Assay
Organs were isolated and immediately frozen in liquid nitrogen. Homogenates were prepared at 4°C in 100 mM KPO 4 (pH 7.8), 10% glycerol, 0.2% Triton X-100, 1 mM EDTA, and 1 mM dithiothreitol with a cooled Potter-Elvehjem type homogenizer. Total protein content in the homogenate was measured using the bicinchonic acid protein assay reagent (Pierce). Luciferase activity of 5-15 l of supernatant of the homogenate was determined as described (7). The ␤-galactosidase activity of 10 l of supernatant was measured using the Galacto-Light kit (Tropix, Bedford, MA) according to the manufacturer's instructions. The light emission was measured in a Berthold Lumat LB 9501 Luminometer.

In Situ Hybridization and Histochemical Analysis
Freshly isolated tissues and embryos were fixed in 4% formaldehyde and embedded in paraplast as described (6). Serial sections of 7 m thickness were probed for the presence of CPS, luciferase, glutamine synthetase (GS) and phosphoenolpyruvate carboxykinase (PEPCK) mRNAs by in situ hybridization with the respective 35 S-labeled cRNAs as described (7,28). For histochemical analysis, frozen tissue was embedded in Tissue-Tek O.C.T. compound (Sakura Finetek, Torrence, CA) and sectioned. Serial cryostat sections 40 m thick were used to detect ␤-galactosidase activity or glutamine synthetase protein as described (29).

RNA Analysis
RNA isolation and Northern blot analysis were performed as described (30). Glycogen and residual DNA was removed by precipitation in 2 M LiCl (16 h, 4°C). Quantitative Northern blot analysis was performed using a PhosphorImager (Molecular Dynamics, Sunnyvale, CA).

RESULTS
The Interaction between the Upstream Regulatory Fragment and Proximal Promoter-Ten constructs were used to study the function of the URF and proximal promoter with respect to tissue-specific expression and hormonal regulation (Fig. 1). The reference construct (construct 1) only contains the minimized, fully active promoter. Construct 2 contains the 469-bp URF, in conjunction with the minimized, fully active promoter (18, 19). Construct 3 lacks sites I, II, and III. In construct 4, the GAG box has been removed additionally. The constructs were tested in transient transfections to FTO-2B hepatoma cells, in the absence and presence of hormones. Constructs 2-4 were stimulated 4 -6-fold by dexamethasone and 7-9-fold by the combination of cAMP and dexamethasone, whereas construct 1 is not stimulated (Fig. 1). In NIH 3T3 fibroblasts and CHO cells, all constructs do show basal activity and no response to hormones (Ref. 20 and not shown).
When the URF is replaced by the 102-bp GRU within the URF (constructs 5-7), basal activity decreases to approximately one-third of that seen in the comparable constructs containing the entire URF. The GRU suffices to confer strong hormone inducibility upon the minimized, fully active CPS promoter (construct 5 and Ref. 20). Mutation of the C/EBP, HNF3, or GR recognition site within this unit almost completely abolishes the hormonal response, showing the strict requirement of the intact GRU for such a response (20).
When sites I, II, and III were removed from the promoter (construct 6), hormone-inducible luciferase levels decreased 4-fold. When, in addition, the GAG box was removed (construct 7), hormone inducibility decreased another 2-fold. The replacement of the TATA box, transcription start site and 5Ј-untranslated region of the CPS gene with those of the Drosophila hsp70 gene (construct 9) or the cytomegalovirus minimal promoter (construct 10) resulted for cytomegalovirus in relatively strong basal activity in the case of the cytomegalovirus promoter and for both promoters in only a slight increase in inducibility relative to construct 7.
Whereas sites I-III and the GAG box are required for hormonal inducibility in construct 7, they are not in construct 4. One possibility for the loss of activity and decline in hormonal inducibility of constructs 5-7 could be the short distance between the TATA box and the GRU, causing crucial sites to be positioned at opposite sites of the DNA helix (asynchronous phasing). The other possibility could be that the HNF3 site, positioned between the GRU and the TATA box in construct 4, forms a crucial component for this construct to be active. To investigate these possibilities, construct 4 was modified by removing the part of the URF upstream of the GRU, leaving all sequences downstream of the deletion in exactly the same configuration (construct 8). The modification caused a complete loss of inducibility. This result shows that the GRU requires either promoter sequences (sites I-III and GAG box) or URF sequences upstream of the GRU to confer its hormonal response.
In Vivo Assay of the Upstream Regulatory Fragment and Proximal Promoter-The DNA in transient transfection assays is present in episomal particles rather than embedded in the chromatin and therefore not subject to many important components of the transcription regulation machinery (31)(32)(33). Furthermore, topological and developmental aspects of regulation are by definition not fully reconstituted in transfection assays. We therefore tested the constructs 2-4 ( Fig. 1) in transgenic mice (lines E, G and T, respectively). This series should enable us to investigate 1) the regulatory potential of the URF, in conjunction with the core promoter and 2) the role of the additional sites (GAG, I, II, and III) within the immediate upstream region of the promoter in the in vivo transactivation of the gene.
To investigate the organ distribution of reporter gene expression, luciferase activity was assayed in liver, small intestine, kidney, spleen, and heart of transgenic littermates of the offspring of founders (Table I). Of lines selected for further analysis, activity in colon, stomach, lung, and pancreas was deter-mined additionally. Eight independent lines with construct E were tested, of which four showed liver-specific expression, one showed equal expression in liver and heart, and three showed no activity. Lines E4 and E5 were used for further analysis. Nine independent lines with construct G were tested, of which four showed liver-specific expression, one showed ectopic expression in the kidney, and four did not show any expression in the organs tested. One positive founder (G5) was not able to pass on the transgene and was lost. Lines G7 and G8 were used for further analysis. Eight independent lines with construct T were analyzed, of which four showed liver-specific expression, one showed expression in all organs tested, and three did not show any expression in the organs tested. Lines T1 and T2 were used for further analysis. Assuming that the transgenic lines that were investigated are a random sample of all possible integrations, our data show that the presence of the GAG box (construct G) and sites I-III (construct E) enhance expression. A similar conclusion is reached when the lines the highest specific expression (E5, G7, T2) are compared. These data therefore correspond to those obtained in vitro, albeit that the differences in vivo are much larger. In addition to organ distribution, we studied three highly characteristic features of CPS expression: selective periportal expression, hormone inducibility, and developmental appearance in the selected transgenic lines.
Localization of Luciferase mRNA in the Liver-The localiza- tion of luciferase mRNA in the different transgenic mice was analyzed with in situ hybridization and compared with that of endogenous CPS, PEPCK, another periportal enzyme, and GS, a strictly pericentral enzyme (Fig. 2). CPS mRNA is expressed in a gradient declining from the portal to the central vein but is excluded from hepatocytes directly surrounding the central veins. Luciferase mRNA is localized in the periportal region in lines E4 and E5 ( Fig. 2 and not shown) and excluded from the hepatocytes surrounding the central veins. Mosaic expression, a frequently occurring problem in transgenic mice (e.g. Refs. 34 and 35), was more prevalent in E4 than in E5. The pattern of expression is very similar to that of CPS in normal, fed conditions, except that a smaller region around the portal veins expresses luciferase, similar to the pattern of PEPCK. After starvation combined with dexamethasone administration, CPS mRNA levels are induced, and its gradient is inverted, now increasing from the portal toward the central vein (7,36). Nevertheless, its expression remains silent in the GS-positive pericentral cells. In both E4 and E5, the gradient in expression of luciferase mRNA did not respond to the challenge (not shown). G5 also showed clear periportal localization, resembling that of E4 and E5 and that of endogenous CPS and PEPCK (Fig. 2). Both the lines T1 and T2 showed mosaic expression with only a small number of the hepatocytes producing detectable amounts of luciferase mRNA. In starved T1 mice treated with dexamethasone, we were best able to visualize the luciferase mRNA in the liver. In these animals positive cells were always localized around the portal veins (Fig. 2).
Hormonal Inducibility of the Constructs in Adult Transgenic Mice-Adult mice of lines E4 and 5, G7 and 8, and T1 and 2 were tested for hormonal inducibility of luciferase expression. Starved animals were injected intraperitoneally with dexamethasone. All lines tested showed significant inducibility of reporter gene expression, the fold induction ranging from 3 to 8 (Fig. 3A). The response of luciferase expression was in good agreement with the increase in endogenous CPS mRNA (Fig. 3B).
Line E4 was used for further analysis of the hormonal regulation of reporter gene expression. As Fig. 4A shows, administration of dexamethasone alone is not effective in increasing reporter gene activity, whereas starvation causes a significant 3-fold increase compared with control. The combination of both  Ϯ 20 in all organs a Activity in kidney, spleen, or heart. Absence of luciferase activity in colon, stomach, lung, and pancreas was confirmed additionally for lines E4, E5, G5, G7, and G8 and T1 and T2.
b Founder without offspring.
dexamethasone and starvation results in a synergistic stimulation of 8-fold compared with control. The hormonal induction of luciferase activity is in good agreement with the induction of endogenous CPS mRNA expression (Fig. 4B), except that treatment with dexamethasone alone failed to induce luciferase expression, whereas it caused a 2-3-fold stimulation of expression of the endogenous CPS gene. When transiently transfected to primary hepatocytes, reporter activity of the constructs 2-4, used for generating E, G, and T mice, strictly depends on dexamethasone (19). The lack of induction upon administration of dexamethasone to the E mice may therefore be due to the fact that endogenous glucocorticoids are present at almost saturating concentrations for the URF alone. Therefore, a group of homozygous E5 mice was adrenalectomized, which causes a decrease in endogenous glucocorticoid levels (37). Both the luciferase activity (Fig. 4A) and the endogenous CPS mRNA levels (not shown) in these adrenalectomized mice were decreased significantly compared with control (Fig. 4A). These results indicate that indeed the endogenous levels of glucocorticoids under normal conditions are saturating for the URF. However, despite the almost complete depletion of glucocorticoid hormones after adrenalectomy (37), CPS and luciferase activity in control and adrenalectomized mice only declines to approximately 50% of that in control animals. It is presently unknown what mechanism maintains CPS expression in vivo in the absence of glucocorticoids, but it is noteworthy that part of the late fetal and preweaning increases in CPS expression are glucocorticoid-independent (37).
Perinatal Activation and Developmental Regulation-Before birth, CPS mRNA levels are low but detectable and increase strongly around birth (7). This pattern is mimicked by the developmental profile of chloramphenicol acetyltransferase mRNA of transgenic mice in which a 12-kbp upstream region of the CPS gene drives reporter gene expression (7). To study the perinatal regulation of the E, G and T transgenes, livers of embryonic day 18.5 mice and neonatal day 0.5 were assayed (Fig. 5D). In contrast to CPS or chloramphenicol acetyltransferase mRNA, the luciferase activity in livers of mice of line E4 and E5, G7 and T2 is hardly detectable before birth, whereas directly after birth a strong induction can be seen. The induction is strongest in T2 and weakest in E4.
To investigate the contribution of the URF and promoter to the developmental regulation after birth, E4, G7, and T2 were subjected to a more detailed analysis of expression of the reporter gene during this period (Fig. 5, A-C). One day after birth, the luciferase activity and endogenous CPS mRNA level have decreased substantially. Lowest CPS mRNA levels were observed at neonatal day 10. Thereafter, levels increased toward weaning. The upsurge of expression toward weaning was seen in all three lines but stronger in E4 and weaker in T2.
To investigate which mechanism underlies the inactivity of the CPS URF/promoter before birth, fetuses were injected in utero with a saturating dose of cAMP or dexamethasone one day before birth (embryonic day 18 -18.5), followed 6 h later by isolation of the livers. Fig. 6 shows that both hormones induced expression before birth, cAMP being much more potent than dexamethasone. These results show that the URF/promoter is highly competent to activate reporter expression before birth and that mainly cAMP is the limiting component of the activation. The activity of the endogenous CPS gene before birth FIG. 3. The CPS minimal promoter suffices for the 469-bp URF to confer hormone inducibility upon a reporter gene. A, transgenic mice of lines E4, E5, G7, G8, T1, and T2 were starved and treated with dexamethasone (see "Experimental Procedures"). 16 h after dexamethasone injection, the livers were isolated, and the luciferase activity was assayed in whole liver extracts (white bars). Black bars represent the specific luciferase activity of fed, untreated transgenic littermates, which were set to one for each group. Error bars give the S.E. of five to ten animals. *, p Ͻ 0.05, using the Student's t test. B, Northern blot analysis of CPS mRNA levels in livers of E and T mice after hormonal treatment. d/s, starved mice treated with dexamethasone; Ϫ, fed, nontreated mice. rRNA staining profiles used to correct for unequal loading are shown (18 and 28 S).

FIG. 4. Starvation activates reporter gene expression by the 469-bp URF and sensitizes the URF to glucocorticoid treatment.
A, analysis of the luciferase activity in whole liver extracts of control (Ϫ), dexamethasone treated (d), starved (s), and dexamethasonetreated starved (d/s) mice of line E4 (left part) and of control (Ϫ) and adrenalectomized (adr) mice of line E5 (right part). The activity in livers of the control group of mice was set to one. Error bars give the S.E. of four to ten animals. *, p Ͻ 0.05, using the Student's t test. B, analysis of total RNA from livers of control (Ϫ), dexamethasone-treated (d), starved (s), and dexamethasone-treated starved (d/s) mice of line E4 as described for Fig. 3B. apparently requires sequences outside the URF and promoter.
The CPS GRU/Promoter in Vivo-To assay the function of the GRU in vivo, construct 5 was used to generate transgenic mice. Of 12 founders only three expressed the transgene. Of these, only one line displayed the highly characteristic periportal expression (Fig. 7). All three lines showed ectopic expression in most organs (not shown). To further substantiate the role of the GRU/promoter in the localization of expression, the construct was flanked by four copies (two at each flank) of the insulators of the chicken ␤-globin gene (38,39). These sequences purportedly increase the number of transgenic lines correctly expressing the transgene (40). This time we directly screened for activity and localization in the livers of the founders (F0 screen) and found two of ten founders expressing the transgene. Both mice showed the characteristic periportal distribution of expression in the liver, showing that the GRU/ promoter contains the information required for this highly characteristic topographical aspect of CPS gene expression.

DISCUSSION
The Upstream Regulatory Fragment Is a Regulatory Module-For some genes, in particular in Drosophila but also in sea urchin and mice, it has been shown that individual regulatory DNA fragments carry out distinct parts of the overall developmental or spatial regulatory function (21,41,42). Such DNA fragments are referred to as regulatory modules (21). We here report the identification and analysis of an in vivo regulatory module in the CPS gene. In conjunction to the minimal promoter (TATA motif, transcription start site, and 5Ј-untranslated region), the URF shows a striking ability to recapitulate the regulatory control over reporter gene expression that is characteristic for the CPS gene in liver. The URF is therefore an essential module in the CPS gene for its spatial, developmental, and hormonal expression in liver in vivo, without the contribution of other sequences. Fetuses of line G7 were treated 0.5 day before birth in utero with chlorothiophenyl cAMP or dexamethasone (dex). 6 h later the livers were isolated, and luciferase activity was measured. As a control, luciferase activity of livers of newborns (ND0), approximately 2 h after birth, was assayed. Compared with the untreated embryonic day 18.5 group, p Ͻ 0.05 for all groups (Student's t test).
The 12-kbp upstream region of the CPS gene exerts regulatory control over a reporter that is virtually identical to the endogenous gene (7). In contrast, the URF is not able to 1) confer expression in the small intestine (Ref . 7 and Table I), 2) to initiate expression before birth (Fig. 5), and 3) to respond to dexamethasone without prior sensitization by cAMP (Figs. 4 and 6). Thus, the URF controls part of the total expression pattern of the CPS gene, whereas different sequences within the 12-kbp upstream region are required for other characteristics of CPS expression, such as expression in the gut and modulation of the hormonal responsiveness. These functional criteria can serve as parameters to identify these additional sequences.
The GAG Box and Sites I, II, and III Are Not Required for Developmental, Spatial, and Hormonal Regulation of Transgene Expression-We and others showed that, in transient transfection experiments, GAG boxes (25,43,44) significantly add, in a position-dependent manner, to basal activity and fold stimulation by hormones (19,44). The sites I, II, and III, which show similarity with the recognition site for C/EBP family members (24), were shown to be involved in the modulation of activity of the GAG box bound factor(s) (19,25). In contrast, the present study indicates that although the GAG motif is required for strong expression in vivo (Table I, compare lines G  and T), neither the GAG motif nor sites I, II, and III qualitatively contribute to transgene regulation; the E, G and T mice showed a very similar response to hormones (Fig. 3), similar developmental profiles, and similar distribution of expression in the liver. Apparently, the URF suffices to exert all these regulatory functions upon the core promoter (TATA box and 5Ј-untranslated region) in vivo.
Communication of the GRU with a Minimal Promoter Requires Additional Elements-The URF depends for its function on an intact GRU formed by an array of four factor-binding sites. Three of these sites (C/EBP, HNF3, and GR recognition site) are functionally linked in that each is indispensable for the function of the GRU, and, therefore, for the entire URF (20). The GRU itself, if linked to the fully active, extended CPS promoter (Fig. 1, construct 5), or the entire URF if linked to the core promoter (Fig. 1, construct 4) is able to respond hepatocyte-specifically to hormones. Thus, the GRU is an essential component of the URF. The GRU alone, however, cannot mediate hormonal activation of transcription if coupled directly to a core promoter (constructs 7, 9, and 10; construct 10 differs from the other two constructs mainly by its high basal activity); either the remaining upstream part of the URF or the upstream CPS promoter sites (I-III and GAG) are required for the GRU to function. Both the proximal promoter and the URF contain sites for liver-enriched factors such as members of the C/EBP family (20,24), which could be one of the requirements for the additional fragments. The function of the additional sequences could be to stabilize the activation complex formed by the GRU-associated factors, including the ligand-dependent glucocorticoid receptor and accessory factors HNF3 and C/EBP and the transcription initiation complex at the basal promoter (33,45). In this complex the GRU functions as a tissue-restricted and ligand-dependent activity switch.
The URF Requires Sensitization by Cyclic AMP to Respond to Glucocorticoids-The relative impotence of glucocorticoids to activate a GRU in vivo in the absence of elevated levels of cAMP is striking and present throughout life. A vivid example of hormone-dependent activation of the GRU is the neonatal activation of gene expression in the liver. Previous studies demonstrated neonatal activation of reporter gene expression from constructs carrying a GRU (46 -48), or solely a cAMP (49) or glucocorticoid response element (Ref. 49; contrast Ref. 48). However, despite high circulating levels of glucocorticoids, these constructs are almost inactive before birth. By treating fetuses in utero with cAMP, we have now demonstrated that the activation of these constructs only awaited the rise in cAMP levels at birth. Similarly, the GRU hardly responds to glucocorticoids in adult fed animals but does so when cAMP levels are high as a result of fasting. These in vivo and earlier in vitro data (20) showed that the GRU suffices to confer this synergistic interaction between glucocorticoids and cAMP on gene expression. In vitro, treatment with cAMP alone is not sufficient to activate the GRU but requires the presence of dexamethasone, indicating that both signals act in combination on the GRU. Cross-talk between the glucocorticoid and cAMP-dependent pathways (via protein kinase A) has been described. Modulation of DNA binding properties or activity of the glucocorticoid receptor or other accessory factors by protein kinase A activity or direct interaction between the receptor and factors downstream of protein kinase A may account for this cross-talk (20, 50 -54). However, we do not know at present the exact mechanism underlying this interaction.
The GRU Confers Periportal Expression in Hepatocytes-A major goal of this study was to define the minimal sequences necessary for periportal expression in the liver. The GRU coupled to the CPS promoter confers this pattern onto the lacZ reporter gene, thereby providing us with, to our knowledge, the smallest thus far reported regulatory fragment qualitatively responsible for heterogeneous gene expression in liver. However, we needed 22 lines to be able to retrieve this pattern. Of 12 lines without insulators, three showed expression, and only one showed the characteristic pattern in the liver. Of 10 trans- genics with ␤-globin gene-derived insulators, two showed expression, both with the characteristic periportal expression pattern. From these observations several conclusions can be drawn. First, the GRU, although capable of recapitulating the spatial pattern of the entire URF, is sensitive to surrounding sequences in the genome. Apparently, its regulatory capacity is easily overruled by regulatory potential (e.g. enhancers) at the site of integration. Second, the GRU/promoter system does not contain sequences that efficiently activate transcription from the chromatin template. The insulators, reported to increase the number of transgenic lines actually expressing the transgene (40), did not increase the number of expressing lines in our case. In fact, the ectopic expression in the liver was suppressed. This is in line with the proposed function of the insulator (38,39,55). Taken together, the 102-bp GRU is a core component of the URF that contains the information for periportal expression in the liver but depends for its proper function on additional regulatory sequences.
The mechanism by which transcription factors that bind to the GRU selectively induce a periportal activation of gene expression remains to be identified. The mRNAs of the factors that have recognition sites in the CPS GRU, the GR, and members of the C/EBP and HNF3 families of factors do not show clear porto-central gradients in expression in the liver (8, 56 -59). This does, of course, not yet rule out gradients in protein concentration or transactivation activity over the portocentral axis. Furthermore, if the factors responsible for periportal gene expression activate transcription co-operatively, as we have shown to be the case for the CPS GRU in vitro (20), the gradient in expression of their target genes could be much steeper than their own gradients in concentration or activity (e.g. Refs. 8, 60, and 61).