Interaction of cytokine- and glucocorticoid-response elements of acute-phase plasma protein genes. Importance of glucocorticoid receptor level and cell type for regulation of the elements from rat alpha 1-acid glycoprotein and beta-fibrinogen genes.

Dexamethasone increased the transscriptional activity of several acute-phase plasma protein genes in cytokine-treated HepG2 cells, suggesting the presence of functional glucocorticoid receptors (GR). The level of GR was, however, insufficient for stimulation of transiently transfected gene constructs containing glucocorticoid-response elements (GRE). By complementation of HepG2 cells with a GR expression vector, a cell system was generated that allowed analysis of the interaction between GRE and cytokine-response elements of the rat alpha 1-acid glycoprotein gene and identification of the principal regulatory elements of the rat beta-fibrinogen gene. Although the expression of plasmid-derived GR mRNA was reduced by dexamethasone treatment, the concentration of GR was sufficient for full, short-term stimulation of GRE-containing vectors. By comparing the pattern of regulation of the cloned GRES in HepG2 and mouse L-cells, an equivalent, cell-type independent dexamethasone response was monitored for the alpha 1-acid glycoprotein element but a response limited to HepG2 cells was found for the beta-fibrinogen element. The data indicate that, although substantial differences exist in the organization and composition of the regulatory elements of the two genes, the overall function is, nevertheless, remarkably similar.

Dexamethasone increased the transcriptional activity of several acute-phase plasma protein genes in cytokine-treated HepG2 cells, suggesting the presence of functional glucocorticoid receptors (GR). The level of GR was, however, insufficient for stimulation of transiently transfected gene constructs containing glucocorticoid-response elements (GRE). By complementation of HepG2 cells with a GR expression vector, a cell system was generated that allowed analysis of the interaction between GRE and cytokine-response elements of the rat al-acid glycoprotein gene and identification of the principal regulatory elements of the rat B-fibrinogen gene. Although the expression of plasmidderived GR mRNA was reduced by dexamethasone treatment, the concentration of GR was sufficient for full, short-term stimulation of GRE-containing vectors. By comparing the pattern of regulation of the cloned GRES in HepG2 and mouse L-cells, an equivalent, cell-type independent dexamethasone response was monitored for the al-acid glycoprotein element but a response limited to HepG2 cells was found for the ,t?fibrinogen element.
The data indicate that, although substantial differences exist in the organization and composition of the regulatory elements of the two genes, the overall function is, nevertheless, remarkably similar.
Systemic injury causes a coordinate increase in the production of a set of plasma proteins, the acute-phase reactants, in the liver of adult mammals (1, 2). Interleukin-6 (IL-6)' and interleukin-1 (IL-l) have been demonstrated to be principle mediators of this response in rodent and human liver cells (3). Maximal expression of several acute-phase plasma protein genes requires, however, the additional presence of glucocorticoids (4 transcriptional activity of the acute-phase plasma protein genes is assumed to be controlled by hormone-specific, cisacting elements (3, 10). Evidence for this is provided by the example of the rat AGP gene. A GRE is located at position -120 to -64 relative to the transcription start site and functions equally well in hepatic cells (11,12) and mouse fibroblasts (13,14). The regulation with IL-l and IL-6 is mediated by the distal regulatory element located at position -5300 to -5150 (12, 15). When the distal regulatory element is recombined with the GRE, a regulatory pattern is achieved which is similar to that found for the chromosomal rat AGP gene (16). Studies on the human /3 fibrinogen gene elements have also suggested separable GRE and IL-6-RE which, however, are inversely organized relative to AGP elements (17). In the rat haptoglobin gene, the elements responsive to IL-6 and dexamethasone could not be physically separated (18). Analysis of other acute-phase plasma protein genes such as human haptoglobin (19), and rat angiotensinogen (24), has led to the identification of the potential cytokine response elements, but no functional GRE have yet been reported. The functional characterization of the acute-phase gene elements demands a suitable cell test system. Most frequently used are the human hepatoma HepG2 (12, 16) and Hep3B cells (19, 21) because of their high efficiency of DNA uptake, their similarity to normal liver cells in hormone responsiveness, and they contain liver-specific transcription factors. Few applications involved stable transfection into rat hepatoma (6,11,24) or into nonhepatic cells, such as L-cells (13,14) or HeLa cells (23). The results from all these transfection studies have been interpreted to reflect how the corresponding endogenous gene is regulated in normal liver cells. In those instances in which the regulatory properties could not be reproduced in transfected hepatoma cells, the failure had been ascribed to the abnormal phenotype intrinsic to the hepatic cells in culture (25). Although transgenic mice have yielded some information about the functionality of cloned gene sequences (15,25,26), detailed analysis of regulatory elements could only conveniently be done in tissue culture systems (22,27).
Once the shortcomings of the given cell culture systems have been recognized, either the deficiencies can be corrected, or separate, but complementary, systems can be employed for gaining information on generally applicable regulatory mechanisms. In this paper we describe how HepG2 cells have to be modified in order to use them for studying the function of GREs and the interaction of the GREs with the cytokineresponse elements of acute reactant genes. The usefulness of this cell system is demonstrated by the identification of the principle regulatory elements of the rat P-fibrinogen gene.
Gene Constructs-An expression vector for human GR, pRShGRa (31), was provided by Drs. R. M. Evans and V. Giguere, Salk Institute, La Jolla, CA, and the vector for rat GR, pRSVGR (32), by Dr. K. R. Yamamoto, University of California, San Francisco, CA. Plasmid pAGP (90)-CAT and pAGP(140)-CAT contain the rat AGP gene promoter from -64 to +21 (without GRE) and -120 to +21 (with GRE), respectively, upstream of the CAT gene in pSVOCAT (14). Additional AGP GREs, comprising the region from -120 to -42, were inserted in the forward orientation into the NdeI site of pAGP(140)-CAT. The plasmid pAGP(3 X DRE)-140CAT contains 3 tandem copies of the IL-l/IL-6 response element in inverted orientation in the NdeI site of pAGP(140)-CAT (12,16). CAT gene expression vectors containing the following promoter regions of the rat fibrinogen genes (33) were generously provided by Dr. G. R. Crabtree, Stanford University (name of clone in parentheses): afibrinogen, -611 to +32 (4aStu); P-fibrinogen, -5200 to +7 (PSH), -2200 to +7 @XX), -504 to +7 (SD), -118 to +7 (027), and -88 to +7 (827.1); r-fibrinogen, -790 to +36 (~3). Subfragments of the pfibrinogen gene promoter, as indicated in Fig. 7, were subcloned into the polylinker region of pCT, which contains the adenovirus major late promoter linked to the CAT gene (provided by Dr. D. Grayson, Georgetown Medical School). Plasmid PIE-MUP contains the gene of the mouse MUP gene under control of the immediate early promoter of the human cytomegalovirus (12) and was used as an internal standard for transfection efficiency and normalization of CAT activity data (16).
Z'ransfection-Plasmid DNA (20 rg/3 x lo6 cells) was transfected as calcium phosphate precipitates (34) into HepG2 cells or as a DEAE-dextran complex (35) into H-35 and L-cells. After a recovery period of 16 h, the cells were released by trypsin when multiple hormone treatments were planned and divided into separate cultures. Hormone treatments were started 24 h later. Standard conditions were: 1 pM dexamethasone, 250 units/ml IL-6, or 500 units/ml ILla.
Assays-CAT activity in heat-treated cell extracts was quantitated as described (36), and either normalized to the amount of cell protein or, in the cases when co-transfected with PIE-MUP, to the amount of MUP produced by the same cells (expressed as percent conversion of chloramphenicol to acetylated products per h and ng of MUP) (12, 16). The amount of plasma proteins secreted by the cell cultures were quantitated by rocket immunoelectrophoresis using equal medium aliquots.
Total cell RNA were prepared (37) and, when necessary, chromatographed on oligodeoxythymidine cellulose mini columns. Equal aliquots of RNA were analyzed for specific mRNA by Northern blot hybridization to 32P-labeled cDNA encoding human and rat GR; mouse MUP (38); or human haptoglobin (39), AGP (40), and CQantichymotrypsin (41). Transcription run-on was measured in HepG2 cell nuclei which had been prepared according to Almendral et al. (42). The run-on reaction and isolation of transcripts were as outlined by Lamers et al. (43). Equal aliquots of 32P-labeled transcripts (5 X lo7 cpm) were hybridized for 3 days to nitrocellulose strips carrying slot-blotted denatured cDNAs encoding human haptoglobin, AGP, cul-antichymotrypsin, triose-phosphate isomerase (44), c-myc and histone H-4 (45), and pBR322.
GR binding activity in cell extracts was determined with [3H] triamcinolone acetonide by following the procedure of Miesfeld et al. (32).

RESULTS
Dexamethsone Response of HepG2 Cells-Treatment of HepG2 cells with combinations of IL-l, IL-6, and dexamethasone for 12 h led to increased transcription rates of the genes for haptoglobin, AGP, and al-antichymotrypsin ( Fig. 1). The Confluent monolayers of HepG2 cells in lo-cm dishes were treated for 12 h with serum-free medium containing the indicated factors. Nuclei were prepared and used for run-on reactions. Labeled RNA (5 X lo7 cpm each) were hybridized to slot-blotted cDNA encoding human haptoglobin (HP), AGP, a,-antichymotrypsin (ACH), triose-phosphate isomerase (Z'PI), histone H4 (ZZ4), and pBR322 (pBR). The autoradiogram after a 2-day exposure is shown. Cytoplasmic RNA were isolated from the postnuclear supernatants of the cell lysate and 15-rg aliquots were analyzed by Northern blot hybridization for the indicated mRNAs. The autoradiograms after a 20-h exposure are shown. Dex, dexamethasone. HepG2 cells were treated for the indicated length of time with either IL-6 and dexamethasone (Der) or IL-l, IL-6, and dexamethasone. Transcription rates were determined by a nuclear run-on reaction as described in the legend to Fig. 1. The change in cytoplasmic mRNA was quantitated by dot-blot hybridization to serially diluted RNA. regulation of these representative acute-phase reactant genes is cytokine-specific: haptoglobin and cul-antichymotrypsin are stimulated by IL-6, and AGP by the combination of IL-l and IL-6. Dexamethasone alone was ineffective but synergistically enhanced the cytokine response of all three genes. The transcription stimulatory action of the more potent combinations of IL-6 and dexamethasone, or IL-l, IL-6, and dexamethasone was detectable to a minor extent after 1 h, but reached a maximum within 6 h, at which level it was maintained for at least an additional 18 h (Fig. 2). The concentration of the mRNAs for these acute-phase reactants changed qualitatively similar to the transcription rates ( Figs. 1 and 2). The magnitude of these changes, however, did not always coincide. Most notable was that the increase of mRNA for haptoglobin and AGP in cells treated with cytokines plus dexamethasone exceeded the increase in transcription rates by 2-to &fold. These differences might be due to post-transcriptional processes which affect mRNA processing and/or stability.
In all instances, the change in the amounts of protein secreted paralleled the change in the mRNA concentration (compare Figs. 1 and 3A). The quantitation of various acutephase plasma proteins confirmed the prominent synergistic action of dexamethasone with the cytokines (Fig. 3A; Ref. 5). Among all the tested plasma proteins, only a-fetoprotein and albumin showed a positive response to treatment with dexamethasone alone. In these cases, the increases ranged from 2fold for a-fetoprotein to 20% for albumin. The effect of dexamethasone on transcription rates, mRNA accumulation, and protein production appeared to be mediated by GR because the same results were obtained with triamcinolone Autoradiograms were exposed for 14 days (GR), or 24 h (MUP, HP, and AGP).
acetonide or RU28362 but not with aldosterone. The cell response was half-maximal at 2 x 1O-9 M dexamethasone. From these results, we concluded that HepG2 cells must possess functional GRs and that these were involved in controlling the expression of the acute-phase protein genes.
Dexamethasone Regulation of Transiently Transfected Plasmids Requires Enhanced GR Concentration-Although HepG2 cells seemed to have the necessary trans-acting factors for achieving a dexamethasone response, transiently introduced plasmid DNA containing the regulatory elements, including a GRE, of an acute-phase plasma protein gene such as the rat AGP, yielded an extremely low stimulation by dexamethasone alone (Fig. 4, left). As found for the endogenous genes, the transfected plasmid was, however, responsive to dexamethasone when applied in combination with cytokines.
The apparently deficient GR function in the absence of cytokines could be equally well restored by transfection with the expression vector for human ( . concentration of functional GR in the transfected cell population, no significant alteration in dexamethasone regulation of endogenous plasma protein production was observed (Fig.  3). This is to be expected considering that only a minor fraction of the cell population has probably taken up and expressed the GR vector (46). However, the correct function of the introduced GR was demonstrated by the dexamethasone-stimulated expression of co-transfected GRE-containing CAT vectors (Fig. 4, right).
Unexpectedly the plasmid-derived GR mRNA was reduced to lo-20% of the normal level following 24 h of dexamethasone treatment (Fig. 3B), even though the activity of the Rous sarcoma virus promoter present in the GR expression vector has not been found to be influenced by the steroid (47). The reduction seemed to be rather specific to the GR expression vector, since the nonrelated, hormone-independent plasmid PIE-MUP was unaffected (Fig. 3B). Moreover, the negative effect was more prominent on the plasmid-derived GR mRNA than on the endogenous GR mRNA (Fig. 5), suggesting that besides transcription control (48, 49), post-transcriptional events might contribute to the observed mRNA levels. Since dexamethasone treatment lowers GR mRNA, a corresponding reduction of appropriate GR proteins and regulatory potential has to be expected for cells which have been treated with dexamethasone longer than 24 h (49-53). The standard co-transfection protocol using 5 yg of pRSVGR in a total of 20 pg of plasmid DNA yielded nevertheless a GR level (Fig. 5) that allowed a seemingly unrestricted stimulation of plasmids which was proportional to the number of GREs (Fig. 6). The expression of all reported constructs in the absence of dexamethasone was less than 0.01% conversion per h and ng of MUP ( Fig. 6; data for constructs with O-2 GREs are not shown). Control cells, not complemented with GR, indicated that just multimerization of GREs was not sufficient to produce a significantly enhanced dexamethasone response. Because elevated GR levels per se did not influence cytokine regulation of transfected plasmids (Fig. 4), GR-complemented HepG2 cells appear to be a most suitable assay system to uncover and to characterize the interaction between cytokine-and glucocorticoid-response "uGm" Ra+ I -GR elements of acute-phase plasma protein genes which demand a hepatic cell environment for full activity.
Identification of the Regulatory Elements in Rat P-Fibrinogen-The expression of the genes for the three rat fibrinogen chains is known to be dependent on liver-specific factors (54, 55) and the transcription is coordinately stimulated by dexamethasone and IL-6 (9). Indeed, 400-800 bp of 5'-flanking region of the 3 genes contains the information for dexamethasone and IL-6 responsiveness and for mediating in the case of (Y-and P-fibrinogen, an additive, and in the case of yfibrinogen, a synergistic action of the two hormones (Fig. 7, constructs 1, 2, and 6). A notable difference between the 3 fibrinogen gene regions was found in the basal level expression: comparably sized 5'-flanking regions of the (Y-and fifibrinogen genes were approximately 20-and lo-fold, respectively, more active than that of the y-fibrinogen gene. These varying basal level activities might contribute to the different magnitudes of hormone stimulation observed (Fig. 7). Evidence that separate hormone-specific cis-acting elements are involved in the overall regulation of fibrinogen genes was obtained by a more detailed analysis of the fifibrinogen gene (Fig. 7, constructs 3-12). Constructs with progressive 5' deletions revealed that all regulatory properties were confined to the 349-bp promoter region (construct 6). A principal GRE activity, which was dependent on functional GR (Fig. 8B) has been localized to the 157-bp fragment at position -349 to -193 (construct 9). The IL-6 regulation was mediated by the 3' adjacent 150-bp fragment (-192 to -43) (construct 10). The relatively high basal level activity of the latter construct was mainly due to the HNF-1 binding site at -100 (54, 55). The only significant IL-6 regulatory activity within the 150-bp fragment was recovered in the 35-bp se- CAT plasmid constructs (14 pg each) containing the indicated 5'-flanking regions of the rat fibrinogen gene were transfected together with 1 pg of PIE-MUP and 5 pg of pRSVGR into HepG2 cells. Subcultures were treated with the indicated factors. The specific CAT activities shown represent mean values of two to four independent determinations. Dex, dexamethasone.
quence: ATGGGTAAACAAGGCTTGCTGGGAAGATGTT-GCTC (-168 to -134), the very region originally suspected by Fowlkes et al. (33) to be important for acute-phase regulation (construct 11). As found with other hormone response elements, duplication of the 35-bp IL-6-RE increased the magnitude of IL-6 response (construct 12).
The results suggested that separable GRE and IL-6-RE were located within the P-fibrinogen gene promoter region. When these elements were recombined upstream of a heterologous promoter, a additive interaction was obtained (Fig.  8B) that was qualitatively identical to the intact gene regions (Fig. 8A). As apparent from Fig. 8B, correct functionality was also recovered with the inverse order of the IL-6-RE and GRE sequences.
Effect of Dexamethusone on IL-6-RE and GRE Function Is Specific to Hepatic Cells-Although the 35-bp IL-6-RE (construct 12) mediated a predominant IL-6 response in HepG2 cells (Fig. 8B), a minor but reproducible stimulation by dexamethasone was also noted which was similar in magnitude to that of the 150-bp region (construct 10, Fig. 7). A more prominent influence of dexamethasone on the IL-6 regulation through this IL-6-RE was demonstrated in H-35 rat hepatoma cells (Fig. 9A). The CAT expression in these cells, transiently transfected with construct 12, was increased 3-fold with dexamethasone, 2-fold with IL-6, and lo-fold with the combination of the two. This significant dexamethasone action suggested either that the steroid positively interferred with the IL-6 signal transduction pathways, or that the 35-bp sequence functioned unexpectedly as a GRE, even though it is devoid of a nucleotide sequence similar to the consensus GR binding sequence (56, 57). To assess whether the latter possibility applied, we tested the IL-6-RE in a separate cell system which has abundant GR activity on its own and mediates dexamethasone regulation independently of IL-6 co-stimulation. We chose mouse L-cells, since these cells have proven to properly regulate transiently introduced vectors which contain GRE from a variety of genes, including the rat AGP gene ( Fig. 9B;  13 sone treatment (Fig. 9B). More surprising, however, was that the expression of the GRE-containing construct 9 similarly expressed in L-cells, but was not stimulated by dexametha-failed to be regulated by dexamethasone in L-cells (Fig. 9B). Plasmid construct 12 (Fig. 7) was transfected into H-35 cells (A) and pCT constructs containing one copy of the rat AGP GRE (-120 to -64), P-fibrinogen GRE (construct 9, Fig. 7), and IL-6-RE (construct 12) into L-cells. The cells were treated as indicated. The CAT activities in equivalent aliquots of cell extracts are shown. The numerical values for percent conversion of chloramphenicol to products are noted above the thin layer patterns. Dex, dexamethasone.
We concluded from these observations that dexamethasone regulation of both GRE and IL-6-RE of the P-fibrinogen gene requires not only functional GR (Fig. 8) but is also dependent upon either additional trans-acting elements which are present only in HepG2 and H-35 cells, or is inhibited by components present only in L-cells. A future task will be to identify the nature of these contributing factors and to establish to what extent the overall steroid hormone response of each gene involves a direct action of the GR.

DISCUSSION
The ability of HepG2 cells to respond to dexamethasone by increasing transcription rates of acute-phase plasma protein genes ( Fig. 1) has been taken as evidence for the action of functional GR (56,57). The level of GR expression (Fig. 5) appears, however, to be too low for efficient stimulation of GRE-containing episomal plasmid DNA by dexamethasone alone (Figs. 4, 6, and 8). Even multimerization of GREs, which has been noted to generate a cooperatively acting response unit (58-60), could not much improve the dexamethasone stimulation (Fig. 6). The concentration of GR was, nevertheless, sufficient to exert a synergistic action in combination with the cytokine response elements of either the rat AGP or P-fibrinogen gene (Figs. 4 and 8). Although we have noted before that various passages of HepGP cells had a relatively low dexamethasone response compared to other cells (such as L-cells (13,14) or HTC cells (ll)), other clonal lines of HepG2 cells appear to exist which support the dexamethasone regulation of the human P-fibrinogen gene element without complementation (17).
Complementation of HepG2 cells with GR expression vector established conditions which supported efficient, dexamethasone regulation of plasmid DNA (Figs. 4,6,and 8;Ref. 18). The fact that GR expression of the endogenous and especially of the transgene is reduced by dexamethasone, be it by changing transcriptional activity of the GR genes (48, 49) or by changing post-transcriptional processing or stability of GR mRNA, one would predict an attenuating effect on the regulatory potential in long term steroid-treated cells. However, when dexamethasone treatment was limited to 24 h, the reduction of GR expression did not seem critical, since no significant restriction in dexamethasone response was yet apparent (Fig. 6). Most of the endogenous human acute-phase plasma protein genes in normal HepGP cells were not responsive to dexamethasone alone (Figs. 1 and 3). The transient transfection of GR expression vector seemed to involve only a minor fraction of the cell population and, therefore, failed to indicate whether the enhanced GR level would cause a stimulation of these genes by dexamethasone. Stable GRtransformed HepG2 cell lines should yield the necessary information; unfortunately, repeated attempts to select such cell lines were unsuccessful for unknown reasons.
GR-complemented HepG2 cells allowed the identification of the GRE-containing region of the P-fibrinogen gene (Fig.  7). The location of that sequence to the region -349 to -193 differed substantially from the GRE of the human gene, which has been found in the fragment spanning from -2900 to -1505 (17). The rat /?-fibrinogen GRE in conjunction with a heterologous promoter was inactive in L-cells (Fig. 9B), illustrating that a search for a functional GRE by using L-cells would have been fruitless if we would have followed the approach proven successful for the rat AGP gene (14). With the precedence of "liver-specific" activity, a reassessment will be necessary for those functional GRE analyses which were limited to nonhepatic cells.
At the present, we can only speculate about the potential molecular cause for our findings. The function of the GRE fragment is either inhibited in L-cells by a specific component which is absent in hepatoma cells, or is dependent on a transcriptional factor that is limited to hepatic cells. The latter possibility deserves primary attention in view of the repeatedly documented fact that full GRE function is achieved by juxtaposing a GR-binding site to a generic transcriptioncontrolling sequence (59). Ongoing analysis of the GRE sequence should yield the identity of the components determining the GRE activity in HepG2 cells. An additional question to be answered is, where does the GR bind, since the @fibrinogen fragment does not contain a sequence conforming with the extended GR-binding site TGTACANNNTGTTCT (57). Although the P-fibrinogen GRE seems to function similarly to the AGP GRE (Figs. 4,7, and 8), they differ not only in their relative gene position and sequence but also in their requirements for trans-acting components (Fig. 9). This is in line with the observation that evolution of functional GREs has utilized modular arrangements of various combinations of cis-acting sequences (61).
The apparent IL-6-RE of the P-fibrinogen is located close to the promoter which is comparable to the organization of other IL-g-regulated rat and human genes (18)(19)(20)(21)(22)(23)27). The IL-6-RE contains the nucleotide sequence (-152) TGCTGGGAA (-144) that has been proposed to be involved in transcriptional activation of a variety of genes by 21,62). The 35-bp IL-6-RE is essentially ILB-specific in HepG2 cells. However, in H-35 cells the same sequence responds prominently to IL-6 and dexamethasone (Fig. 9A), suggesting the possibility of overlapping signal transduction pathways in these rat hepatic cells. A possibility to consider is that dexamethasone acts indirectly by affecting the activity of those components involved in IL-6 regulation. It has been suggested by Bauer et al. (63) that IL-6 receptor expression can be modulated by acute phase in primary cultures of hepatocytes. However, no significant change in concentration of IL-6 receptor mRNA was found in H-35 cells following treatment with dexamethasone and IL-6 (data not shown). Regardless of the mode of regulation, it is obvious from our observation that just the responsiveness to dexamethasone of a cloned DNA piece in an expression vector is not an accurate defining criterion for a GRE. A comparison to established GRE sequences will demand identification of a GR-binding site (64, 65, and discussion therein).
In summary, this study underscores the necessity that functional analysis of regulatory elements from acute-phase