Dual Role of Insulin in Transcriptional Regulation of the Acute Phase Reactant Ceruloplasmin*

Insulin is a potent negative regulator of the response of hepatic cells to pro-inflammatory cytokines, particularly, interleukin (IL)-6. The action of insulin is target-selective because it inhibits transcription of most but not all acute phase genes. We here show that ceruloplasmin (Cp), an acute phase reactant with important functions in iron homeostasis, is subject to a unique dual regulation by insulin. IL-6 increased Cp mRNA expression in HepG2 cells by ∼5-fold. Simultaneous treatment with insulin reduced this stimulation by half. Surprisingly, insulin by itself caused a 2–4-fold induction in Cp mRNA expression. The mechanism of induction by insulin was studied by transfecting into HepG2 cells chimeric constructs of the Cp 5′-flanking region driving luciferase. The activity of a 4800-bp segment of the Cp 5′-flanking region was increased 3-fold by insulin. Deletion and mutation analyses showed the requirement for a single hypoxia-responsive element in a 96-bp segment ∼3600 bp upstream of the initiation site. The domains required for the two activities of insulin were distinct: The distal, hypoxia-responsive element-containing site was sufficient for Cp transactivation by insulin; in contrast, an 848-bp region adjacent to the initiation site was sufficient for IL-6 transactivation of Cp and for the inhibitory activity of insulin. The role of hypoxia-inducible factor-1 in the induction of Cp by insulin was shown by electrophoretic mobility shift assays and by the absence of insulin-stimulated Cp promoter activation in mouse Hepa c4 cells deficient in hypoxia-inducible factor-1 activity. Taken together these results show that insulin functions as a bidirectional, condition-dependent regulator of hepatic cell Cp expression. The unique regulation of Cp may reflect its dual roles in inflammation and iron homeostasis.

Insulin is a potent negative regulator of the response of hepatic cells to pro-inflammatory cytokines, particularly, interleukin (IL)-6. The action of insulin is targetselective because it inhibits transcription of most but not all acute phase genes. We here show that ceruloplasmin (Cp), an acute phase reactant with important functions in iron homeostasis, is subject to a unique dual regulation by insulin. IL-6 increased Cp mRNA expression in HepG2 cells by ϳ5-fold. Simultaneous treatment with insulin reduced this stimulation by half. Surprisingly, insulin by itself caused a 2-4-fold induction in Cp mRNA expression. The mechanism of induction by insulin was studied by transfecting into HepG2 cells chimeric constructs of the Cp 5-flanking region driving luciferase. The activity of a 4800-bp segment of the Cp 5-flanking region was increased 3-fold by insulin. Deletion and mutation analyses showed the requirement for a single hypoxia-responsive element in a 96-bp segment ϳ3600 bp upstream of the initiation site. The domains required for the two activities of insulin were distinct: The distal, hypoxia-responsive element-containing site was sufficient for Cp transactivation by insulin; in contrast, an 848-bp region adjacent to the initiation site was sufficient for IL-6 transactivation of Cp and for the inhibitory activity of insulin. The role of hypoxia-inducible factor-1 in the induction of Cp by insulin was shown by electrophoretic mobility shift assays and by the absence of insulin-stimulated Cp promoter activation in mouse Hepa c4 cells deficient in hypoxia-inducible factor-1 activity. Taken together these results show that insulin functions as a bidirectional, condition-dependent regulator of hepatic cell Cp expression. The unique regulation of Cp may reflect its dual roles in inflammation and iron homeostasis.
Vertebrates respond to tissue damage and other inflammatory stimuli by implementing a coordinated series of processes known as the acute phase reaction. A characteristic of this response, and the basis of common clinical tests, is the increase in the plasma concentration of a limited group of proteins, e.g. C-reactive protein. The major function of the acute phase response is almost certainly to protect the host against injury, trauma, infection, and other inflammatory events (1,2). The response is initiated by surveying leukocytes that respond to these events by secretion of inflammatory mediators. These mediators have been broadly classified into two groups: (i) cytokines, such as interleukin (IL) 1 -6, IL-1, interferon-␥, and tumor necrosis factor-␣, which are the primary regulators of acute phase gene expression; and (ii) insulin (and other growth factors) and glucocorticoids, which function as modulators of cytokine action (1,2). Among the pro-inflammatory mediators, IL-6 is considered to be the major physiological regulator of acute phase gene expression (2,3). The liver is a principal target of inflammatory mediators, and the specific actions of these mediators with respect to the pattern of acute phase gene expression have been defined in detail in hepatocarcinoma cell lines (2). Insulin is a highly effective negative regulator of the cytokine-stimulated acute phase response in vitro (4 -8) and in vivo (9,10). Insulin inhibits IL-6-mediated induction of haptoglobin, thiostatin, complement C3, and C-reactive protein in hepatocytes and HepG2 cells (6,11). The extent of the inhibition depends on the target protein, and insulin does not alter IL-6-induced expression of several acute phase proteins including ␣ 1 -antichymotrypsin and ␣ 1 -acid glycoprotein (6,11).
Ceruloplasmin (Cp) is a 132-kDa, copper-containing acute phase protein of mainly hepatic origin. Its plasma concentration during inflammation in humans increases by ϳ50 -100%, much less than the 100 -1000-fold increases for C-reactive protein and serum amyloid A (12). The function of Cp during inflammation is unclear. Several groups have reported an antioxidant activity of Cp that protects lipids and DNA against free radical-mediated injury (13). In contrast, we and others have shown that Cp copper can cause oxidative modification of lipoproteins (14 -16). In addition to its participation in free radical reactions, Cp is an important regulator of iron homeostasis. It is the major ferroxidase in plasma, catalyzing the conversion of Fe 2ϩ to Fe 3ϩ for binding by apo-transferrin (17).
The key role of Cp in iron homeostasis is supported by recent reports of iron overload in patients with hereditary Cp deficiency (18), and in mice with targeted disruption of the Cp gene (19). These findings, together with early organ culture and animal studies (20,21), suggest that Cp is required for efficient release of iron from cells and tissues. In contrast, Cp has been shown to mediate inward iron flux as well in some cell culture systems. Cp increases iron uptake by iron-deficient cells of hepatic and erythroid origin (22,23), and by glioblastoma cells (24,25); the Cp homologue fet3p has a similar role in high affinity iron uptake in Saccharomyces cerevisiae (26,27).
The spectrum of agents that increases plasma Cp concentration in vivo, and hepatic cell synthesis of Cp in vitro, is consistent with its involvement in both inflammation as well as iron homeostasis. Plasma levels of Cp increase in animal models of inflammation, including after injection of tumor necrosis factor-␣, turpentine, IL-6, and endotoxin (28 -30). Cp secretion is increased in human hepatoma cells exposed to IL-6 (31, 32), but the underlying molecular mechanism has not been investigated in detail; a recent report suggests that regulation may be post-transcriptional (33). We have shown recently that iron depletion increases hepatic cell transcription of Cp via activation of hypoxia-inducible factor (HIF)-1, which binds to a hypoxia-responsive element (HRE) in the 5Ј-flanking region of Cp (22,34). HIF-1 is a heterodimer containing HIF-1␣ and HIF-1␤/aryl hydrocarbon receptor nuclear translocator (ARNT) subunits, two basic helix-loop-helix/Per-ARNT-Sim proteins (35). Upon cell activation, HIF-1␣ is induced and it translocates to the nucleus where it forms dimers with HIF-1␤; these dimers bind to HREs in multiple genes to induce transcription. HIF-1 is activated by iron deficiency, hypoxia, and CoCl 2 , and recent evidence shows activation of HIF-1 by insulin and insulin-like growth factor (36,37). Thus, HIF-1 activation may provide a common mechanism for induction of multiple genes by hypoxia and insulin, e.g. erythropoietin, vascular endothelial growth factor, and the glucose transporter Glut1 (36,38).
The known transcriptional responses of Cp present a potential paradox regarding the effect of insulin on Cp gene expression. As a negative regulator of the acute phase response, insulin would be expected to diminish transcriptional activation of Cp, particularly by IL-6. In contrast, as a HIF-1-responsive protein, insulin may increase Cp expression. To investigate this paradox, we have studied the effect of insulin (and IL-6) on the transcriptional response of Cp in HepG2 cells. Like multiple other acute phase proteins, the IL-6-mediated induction of Cp is inhibited by insulin. However, our results indicate that Cp is unique among acute phase proteins in that transcription is induced by insulin, in the absence of IL-6, through HIF-1 activation. This unusual dual response of Cp to insulin may be the result of its unique metabolic positioning at the intersection of inflammation and iron metabolism.

EXPERIMENTAL PROCEDURES
Reagents-Human Cp was purchased from Calbiochem and bovine insulin from Invitrogen. Rabbit polyclonal anti-human Cp IgG and peroxidase-conjugated anti-rabbit IgG were obtained from Accurate Chemical (Westbury, NY) and Roche Molecular Biochemicals, respectively. Other reagents were from Sigma.
Cell Lines and Culture Conditions-Human hepatocarcinoma HepG2 cells were obtained from American Type Culture Collection and cultured in Dulbecco's modified Eagle's medium (Sigma) supplemented with 10% heat-inactivated fetal bovine serum, 100 units/ml penicillin, 100 mg/ml streptomycin, and 2 mM L-glutamine (Invitrogen). Mouse hepatoma cell lines Hepa-1c1c7 and Hepa c4 were kind gifts of Oliver Hankinson and were cultured in modified Eagle's medium with the same supplements used for HepG2 cells. Cells at 50 -60% confluence were used in all experiments. Cells were maintained in a humidified atmosphere containing 5% CO 2 at 37°C. For experiments in which cells were exposed to hypoxia, oxygen tension in the chamber (Billups-Rothenberg, San Diego, CA) was set at either 1% O 2 for hypoxia or 20% O 2 for normoxia.
Immunoblot Analysis of Cp-Conditioned medium from serum-deprived HepG2 cells was subjected to 7% SDS-PAGE using Protogel (National Diagnostics, Atlanta, GA) and transferred by a semi-dry method to an Immobilon-P membrane (Millipore, Bedford, MA). The membrane was incubated with anti-human Cp IgG (1:10,000) as primary antibody, and then with peroxidase-conjugated secondary antibody (1:5000). Cp was detected by chemiluminescence using ECL (Amersham Biosciences). The blot was subsequently incubated with Coomassie Blue dye to verify uniform loading of all samples.
RNA Blot Analysis of Cp and Other Transcripts-RNA was isolated from HepG2 cells using TRIzol reagent (Invitrogen). Total RNA (20 g) was denatured in formamide/formaldehyde, electrophoresed through a 1% agarose gel containing 6% formaldehyde, and then blotted onto nylon membranes (Schleicher & Schuell). After cross-linking by ultraviolet irradiation (Stratalinker, Stratagene), the filters were hybridized to a 646-base pair (bp) BstXI/BamHI restriction fragment (positions 984 -1629 of the open reading frame) of a human Cp cDNA labeled by random priming with [␣-32 P]dCTP. For Northern blot analyses of other genes, blots were hybridized with radiolabeled probes consisting of a 1.5-kb EcoRI fragment of human C-reactive protein cDNA (pCRP1, ATCC 59496), a 1.6-kb PstI fragment of human transferrin cDNA (TfR27A, ATCC 53106), a 525-bp fragment of human serum albumin generated by reverse transcription-PCR, and 0.65-bp EcoRI fragment of VEGF (provided by Bela Anand-Apte).
Construction of Vectors Containing Cp Promoter and Enhancer Segments-Cp promoter/enhancer constructs, engineered to contain SacI and XhoI restriction sites, were made by PCR amplification using Pfu polymerase (Stratagene), primers containing these restriction sites, and a 4774-bp fragment of the 5Ј-flanking region of the human Cp gene as template (pGEM-Cp). pGEM-Cp was obtained by a PCR-based screen of a human genomic library in the bacterial artificial chromosome vector pBeloBAC11 (Research Genetics, Huntsville, AL) as described (34). A proximal construct was made from Ϫ848 to Ϫ1 (the nucleotide upstream of the translation-initiation site, which is here defined as ϩ1) and ligated into SacI and XhoI sites of pGL3basic vector (Promega). For the long constructs inserted into pGL3basic, the PCR products from two separate amplification reactions were ligated to form a single construct. In brief, a proximal construct was made from Ϫ2389 (just upstream of an EcoRI site) to Ϫ1. Several distal constructs were PCR-amplified from 5Ј-termini at Ϫ4774, Ϫ3639, and Ϫ3576 to the 3Ј-terminus at Ϫ2325. The proximal and distal products were ligated at the EcoRI site and then into the 5Ј-SacI and 3Ј-XhoI sites upstream of luciferase in pGL3basic. An upstream Cp enhancer construct was made by PCR amplification of pGEM-Cp between Ϫ3639 and Ϫ3544 and was ligated into the SacI and XhoI sites of the pGL3prom vector (Promega) upstream of the SV40 promoter and luciferase. Site-directed mutagenesis of the Cp HRE in this construct (from CGT to AAA) was done by the megaprimer method (39). All constructs were verified by sequencing.
Transient Transfection of Cells and Reporter Gene Assays-To measure transcriptional efficiency of chimeric Cp promoter/enhancer constructs, HepG2 cells at ϳ50% confluence in six-well plates were transiently transfected for 16 h with a reporter plasmid (2 g) using Lipofectin (Invitrogen). To monitor transfection efficiency, a reporter gene construct (0.25 g) containing ␤-galactosidase behind an SV40 promoter was co-transfected. Transfected cells were allowed to recover for 6 h in Dulbecco's modified Eagle's medium with 10% fetal bovine serum and then incubated with insulin or subjected to 1% O 2 in serumfree medium for 18 h. Luciferase (Promega) and ␤-galactosidase (Tropix, Bedford, MA) activities in cell extracts were determined by chemiluminescence.
Preparation of Nuclear Extracts-Nuclear extracts were prepared from HepG2, Hepa-1c1c7, and Hepa c4 cells as described (40). Briefly, 1 ϫ 10 8 cells were washed with ice-cold phosphate-buffered saline and then with a solution containing 10 mM Tris-HCl, pH 7.8, 1.5 mM MgCl 2 , and 10 mM KCl, supplemented with a protease inhibitor mixture containing 0.5 mM dithiothreitol, 0.4 mM phenylmethylsulfonyl fluoride, and 2 g/ml each of leupeptin, pepstatin, and aprotinin (Sigma). After incubation on ice for 10 min, the cells were lysed by 10 strokes with a Dounce homogenizer and the nuclei were pelleted. The pellet was resuspended in a solution containing 420 mM KCl, 20 mM Tris-HCl, pH 7.8, 1.5 mM MgCl 2 , and 20% glycerol, supplemented with the protease mixture described above, and incubated at 4°C with gentle agitation. The nuclear extract was centrifuged at 10,000 ϫ g for 10 min, and the supernatant was dialyzed twice against a solution of 20 mM Tris-HCl, pH 7.8, 100 mM KCl, 0.2 mM EDTA, and 20% glycerol. Protein concentration was determined using the Bio-Rad reagent with bovine serum albumin as standard.
Electrophoretic Mobility Shift Assay (EMSA)-Sequences of the sense strands of the oligonucleotide probes used for EMSA were as follows: 5Ј-TCT GTA CGT GAC CAC ACT CAC CTC-3Ј (Cp HRE), 5Ј-TCT GTA AAA GAC CAC ACT CAC CTC-3Ј (mutated Cp HRE), and 5Ј-GCC CTA CGT GCT GTC TCA CAC AGC-3Ј (erythropoietin HRE). The sense and antisense strands were annealed, gel-purified, and endlabeled with [␥-32 P]ATP (PerkinElmer Life Sciences) using T4-polynucleotide kinase (Promega). Unincorporated nucleotide was removed by gel filtration using G-25 Sephadex columns (Quick Spin TM TE, Roche Molecular Biochemicals). To measure DNA-protein interaction, 1 ϫ 10 5 cpm of oligonucleotide probe was incubated with 5 g of nuclear extract and 0.5 g of sonicated, denatured salmon sperm DNA (Invitrogen) in 10 mM Tris-HCl, pH 7.8, 50 mM NaCl, 1 mM MgCl 2 , 1 mM EDTA, 5 mM dithiothreitol, and 5% glycerol, for 20 min at 4°C in a total volume of 20 l. The reaction mixture was subjected to electrophoresis (200 V in 0.3ϫ Tris-buffered EDTA solution at 4°C) using 5% nondenaturing polyacrylamide gels. Dried gels were subjected to autoradiography up to 24 h. For competition experiments, a 10 -300-fold molar excess of unlabeled, annealed oligonucleotide was pre-mixed with radiolabeled probe before addition to the binding reaction. For gel supershift analysis, 1 l of rabbit monoclonal antibody against HIF-1␣ or rabbit polyclonal antibody against ARNT/HIF-1␤ (both from Novus Biologicals, Littleton, CO) was added after the initial 20-min incubation, and the solution was further incubated for 30 min at 4°C before electrophoresis.
Statistical Analysis-All experiments have been performed at least three times with similar results, and representative experiments are shown. Densitometric results are normalized with respect to internal controls and are expressed relative to the results in untreated control wells. Results from reporter experiments are expressed as mean values Ϯ standard error of the mean (n ϭ 3 replicate wells).

Induction of Cp Expression by Insulin-We first determined
whether insulin blocked the IL-6-mediated induction of Cp gene expression. Northern blot analysis of HepG2 cells showed that IL-6 increased the steady state level of both major Cp mRNA transcripts by more than 4-fold (Fig. 1A). The induction by IL-6 was substantially reduced by approximately half by insulin, to a level that was still 2-fold higher than untreated control cells. Surprisingly, Cp mRNA expression was increased by ϳ2-fold by insulin itself. A dose-response experiment was done to determine whether the inhibitory activity was observed at physiological concentrations of insulin (between 0.1 and 5 nM). Marked inhibition of IL-6-mediated Cp expression was observed at 0.3 nM insulin and maximal inhibition at ϳ1 nM (Fig. 1B). The average serum insulin concentration in fasting adults is ϳ0.1 nM, but insulin concentrations from 0.3 to 0.9 nM are transiently reached after meals (in normal adults and obese adults with impaired glucose tolerance, respectively), and lev-  Fig. 1A. B, HepG2 cells were incubated with insulin at concentrations up to 100 nM for 18 h. Aliquots of conditioned medium were subjected to 7% SDS-PAGE followed by immunoblot analysis using polyclonal rabbit anti-human Cp IgG (top panel). The Coomassie Blue-stained gel was used as a loading control (middle panel). Cp secretion was quantitated densitometrically, normalized by the major, ϳ180-kDa band, and expressed compared with untreated control cells (lower panel). C, to determine specificity of the effect of insulin on Cp, the expression of several acute phase and other genes was determined. HepG2 cells were treated with insulin (100 nM) for 18 h and total RNA isolated. The expression of Cp, C-reactive protein (CRP), albumin, transferrin, and VEGF mRNA was determined by Northern blot analysis using genespecific radiolabeled cDNA probes. Ultraviolet visualization of the 28 S ribosomal RNA subunit served as a control. els of 1 nM and higher are seen after insulin administration and in patients with insulinomas (41,42). The concentration of IL-6 used in this experiment, 2.5 ng/ml, is typical of the serum concentration in patients with acute inflammatory syndromes including bacteremia or meningitis (43). Thus, insulin has a dual activity with respect to Cp expression; it is a negative regulator of Cp under inflammatory conditions characterized by elevated IL-6, but it is a positive regulator under noninflammatory conditions. Northern blot analysis of HepG2 treated with increasing amounts of insulin showed that half-maximal stimulation of steady-state level of Cp mRNA occurred at ϳ1 nM insulin, and a maximal 4-fold increase was observed at 10 nM. (Fig. 2A). In multiple repetition of this experiment, the maximal stimulation of Cp mRNA expression by insulin was between 2-and 4-fold compared with untreated controls. Analysis of secreted Cp, by immunoblot analysis of conditioned medium, showed that the increase in synthesis and secretion of the protein nearly paralleled the induction of the transcript (Fig. 2B). Half-maximal induction was observed between 0.3 and 1 nM, and maximal stimulation was seen at 3-10 nM insulin. Cp was secreted entirely in the intact, 132-kDa form required for several of its biological activities (44,45). The specificity of the induction of Cp was shown by Northern analysis of other secreted hepatic proteins. Insulin did not alter HepG2 cell mRNA expression of C-reactive protein, a positive acute phase protein, or albumin and transferrin, two negative acute phase proteins (Fig. 2C). In a positive control experiment, we measured the induction of VEGF gene expression as described previously for insulin (36) and insulin-like growth factor-1 (37); comparable inductions of VEGF and Cp mRNA were observed (Fig. 2C). A time-course experiment showed only marginal stimulation of Cp mRNA expression after 8 h, and essentially maximal expression by 16 h (Fig. 3A). A time course of protein secretion, as expected, showed delayed stimulation of Cp by insulin (Fig.  3B). Very little stimulation was seen after 16 h, but substantial stimulation by 24 h.
Transcriptional Activation of Cp by Insulin: Requirement for a HRE in the Cp 5Ј-Flanking Region-We have previously reported that the Cp gene contains a functional HRE activated by both iron deficiency and hypoxia (34). This finding, coupled with recent reports that insulin induces multiple genes through activation of HIF-1 (36 -38), led us to examine the possible role of HIF-1 in insulin-inducible Cp expression. We first compared the magnitude of the inductions in response to insulin and hypoxia. Exposure of cells to hypoxic conditions, i.e. 1% O 2 , induced an ϳ10-fold increase in Cp mRNA compared with normoxic control cells (range was between 6-and 19-fold in multiple experiments), an induction substantially greater than that by a maximal dose of insulin (2.6-fold in this experiment) (Fig. 4A). Exposure of cells to both hypoxia and insulin did not increase Cp expression beyond that induced by hypoxia alone, indicating the absence of a synergistic effect. Because the induction of HRE-containing genes by hypoxia requires new protein synthesis, we tested whether protein synthesis was also required for Cp induction by insulin. Cycloheximide completely blocked the induction of Cp, indicating a common dependence on new protein synthesis for both agonists (Fig. 4B).  To investigate the molecular mechanism of insulin-stimulated Cp gene expression, we prepared chimeric constructs of the Cp gene 5Ј-flanking regulatory region ligated with luciferase as a heterologous reporter. The human Cp 5Ј-flanking region contains six consensus HIF-1 binding sites, i.e. (G/ A)CGTG (35), arranged in three adjacent pairs between positions Ϫ4535 and Ϫ2447 (Fig. 5A). A series of promoter/enhancer fragments was constructed with progressively larger 5Ј-deletions, and the fragments were ligated upstream of luciferase in the promoterless reporter gene pGL3-basic. Subconfluent HepG2 cells were transiently transfected with these constructs and then incubated for 18 h with insulin or under hypoxic conditions. Transactivation of the entire 4774-bp construct (Cp Ϫ4774,Ϫ1 -Luc) was increased ϳ3-fold by insulin (Fig.  5A). Hypoxia also transactivated this construct; the -fold stimulation was approximately twice that for insulin, consistent with their relative increases in Cp mRNA expression. Insulin and hypoxia similarly transactivated a chimeric construct (Cp Ϫ3639,Ϫ1 -Luc) lacking the upstream-most pair of consensus HIF-1 binding sites. However, activation by insulin and hypoxia was completely abrogated in a promoter/enhancer construct lacking the three distal consensus HIF-1 binding sites (Cp Ϫ3576,Ϫ1 -Luc). Basal activity was similar for all Cp promoter/enhancer constructs. These results suggest that a single HRE located between positions Ϫ3639 and Ϫ3577 is required for Cp transactivation by insulin and hypoxia.
To test whether this single HRE was sufficient for Cp transactivation by insulin, a mutagenesis strategy was used. A 96-bp segment containing this HRE was subcloned upstream of the SV40 promoter driving luciferase in the pGL3prom vector (Cp Ϫ3639,Ϫ3544 -SV40-Luc). After transient transfection into HepG2 cells, insulin stimulated reporter gene expression by almost 3-fold (Fig. 5B). As before, the stimulation by hypoxia was approximately twice that by insulin. To show that the HRE in this enhancer region was responsible for the observed activation, the core 5Ј-ACGTG-3Ј sequence was mutated to 5Ј-AAAAG-3Ј. This construct was completely inactive with respect to transactivation by insulin or by hypoxia (Fig. 5B). These results suggest that a single HRE present in the Cp enhancer is necessary and sufficient for transactivation by insulin.
Relationship between Cp Promoter/Enhancer Sites Required for Stimulatory and Inhibitory Activities of Insulin-We investigated the relationship between the cis-acting site at which insulin transactivates Cp and the site required for the inhibition of IL-6-mediated Cp transactivation. HepG2 cells were transfected with the full-length promoter/enhancer construct (Cp Ϫ4774,Ϫ1 -Luc) and then were treated with IL-6, insulin, or both, and luciferase activity was measured. IL-6 induced reporter gene expression by almost 5-fold. Insulin reduced IL-6mediated induction to the level seen with insulin alone, between 2-and 3-fold (Fig. 6A). Thus, both the positive and negative regulatory activities of insulin are reconstituted using the full-length Cp promoter/reporter construct. To determine whether the two activities of insulin utilize the same site, fragments of the 5Ј-flanking region were tested. The vector containing the 96-bp segment containing the active Cp HRE (Cp Ϫ3639,Ϫ3544 -SV40-Luc) was transfected into HepG2 cells and treated with insulin and IL-6. As expected, insulin induced luciferase expression by ϳ2.5-fold; however, there was essentially no induction by IL-6 (Fig. 6B). The inability of insulin to inhibit the induction by IL-6 suggested that the inhibitory site was not in the region containing the HRE. Cells were then transfected with a construct containing the presumptive basal Cp promoter (Cp Ϫ848,Ϫ1 -Luc), previously shown to drive constitutive, but not hypoxia-stimulated, reporter gene expression (34). As expected, insulin did not increase luciferase expres-sion, but IL-6 increased expression by ϳ5-fold (Fig. 6C). The stimulatory activity of IL-6 was completely abrogated by insulin, indicating the presence of the site of negative regulation in this proximal construct. Together these results show that distinct cis-acting sites in the Cp gene 5Ј-flanking region are required for the stimulatory and inhibitory activities of insulin.
Role of Hypoxia-inducible Factor-1 in Transcriptional Activation of Cp by Insulin-To identify the transcription factor(s) involved in insulin-stimulated induction of Cp, EMSAs were done using a radiolabeled 24-bp Cp HRE probe. HepG2 cells were treated with insulin or hypoxia, and nuclear extracts were incubated with the Cp HRE probe. Treatment with insulin led to the formation of a single radiolabeled complex (Fig. 7A). The mobility of the complex induced by insulin was the same as that induced by hypoxia, and the same as that bound to an erythropoietin HRE probe, suggesting that a similar HIF-1containing complex may be induced by both hypoxia and insulin. Competition experiments were done to show specificity of binding of the complex to the Cp HRE probe. The binding of radiolabeled Cp HRE probe was partially competed by a 10-fold molar excess of unlabeled Cp HRE and almost completely competed by a 100-fold molar excess (Fig. 7B). Binding was similarly blocked by an excess of unlabeled erythropoietin HRE probe, whereas a mutated Cp HRE probe showed only partial competition even at 300-fold molar excess.
Gel supershift studies were done to confirm the presence of HIF-1 in the insulin-induced complex that binds the Cp HRE. Rabbit monoclonal anti-HIF-1␣ shifted the complex formed in insulin-treated cells (Fig. 8A). Furthermore, a polyclonal antibody against ARNT/HIF-1␤ (and a mixture of both antibodies) completely blocked the formation of the complex. To further confirm the requirement for HIF-1 in the formation of insulinstimulated, Cp HRE-binding complexes, we took advantage of the Hepa c4 mouse hepatoma cell line, which is deficient in ARNT/HIF-1␤ (46). These cells, derived from the parental Hepa-1c1c7 cell line, have been used to show the requirement for HIF-1 in gene transactivation by hypoxia (40), and more recently by insulin (36). Both cell lines were transfected with the chimeric reporter gene Cp Ϫ3639,Ϫ3544 -SV40-Luc. Treatment of the wild-type Hepa-1c1c7 cells with insulin increased luciferase expression by almost 2-fold, whereas the stimulation by hypoxia was ϳ3-fold (Fig. 8B). Neither treatment increased Cp enhancer activity in the HIF-1-␤-deficient cell line. The mouse cell lines were also used to establish HIF-1 binding by electrophoretic mobility shift assay. Hepa-1c1c7 and c4 cells were treated with insulin or with CoCl 2 , and the binding of the HIF-1 complex to the Cp HRE probe was determined; we previously showed that CoCl 2 , like hypoxia, activated HIF-1 and Cp transcription in these cells (34). Inducible binding was observed in nuclear extracts of wild-type Hepa-1c1c7 cells treated with either agonist, whereas essentially no HIF-1 binding was seen in the ARNT/HIF-1␤-deficient Hepa c4 cells (Fig.  8C). Together, these studies clearly demonstrate the requirement for HIF-1 in activation of Cp transcription by insulin. DISCUSSION We have found that the effect of insulin on hepatic cell transcription of Cp depends on the inflammatory state of the cell. In the absence of a pro-inflammatory stimulus, i.e. treatment with IL-6, insulin enhances Cp gene expression and protein production in HepG2 cells. This induction requires activation of HIF-1 and binding to an HRE in the distal 5Ј-flanking region of Cp. In contrast, in the presence of IL-6, insulin markedly represses Cp gene expression in these cells. Thus, insulin functions as a bidirectional, condition-dependent regulator of hepatic cell Cp expression.
Negative Regulation of Cp Transcription by Insulin-We have confirmed previous reports of induction of Cp in hepatoma cells by IL-6 (31, 32). Our finding that IL-6 increases the steady-state level of Cp mRNA and activates a chimeric Cp promoter-Luc reporter shows that regulation is at the level of transcription. Although these results are consistent with the mechanism of IL-6-mediated induction of other acute phase genes, transcriptional regulation of Cp by IL-6 has not been previously shown. We cannot account for the difference between our results and an earlier report that Cp induction by IL-6 is primarily post-transcriptional, but a variance in culture conditions is a possible explanation (33). Elucidation of the mechanism of transcriptional activation of acute phase genes by IL-6 is an area of intense investigation. An important role for C/EBP␤ in IL-6-stimulated transactivation of multiple acute phase genes has been shown by up-regulation of C/EBP␤ expression, and by its binding to specific sequences in the 5Ј-flanking regions of these genes (47)(48)(49). More recently, the IL-6-activated "acute phase response factor" has been shown to be STAT3, which binds to palindromic, TT(N) 5 AA motifs (termed acute phase response elements) in the promoters of multiple (but not all (Ref. 50)) acute phase genes (51)(52)(53)(54). The essential role of STAT3 in the acute phase response in vivo has been shown in mice with an inducible STAT3 gene deletion (55). The molecular mechanism of Cp transactivation by IL-6 has not been investigated in detail. Binding of C/EBP␤ to the rat Cp promoter has been shown, but agonist-induced responses have not been examined (56). Likewise, the role of STAT3 in basal or agonist-induced Cp transcription has not been reported; however, the presence of three near-consensus STAT3-binding sites in the proximal Cp promoter between nucleotide positions Ϫ848 and Ϫ1 merits further analysis.
Our results show that insulin suppresses the IL-6-mediated increase in Cp mRNA level to nearly the same level as that seen with insulin alone. The experiments with a chimeric reporter gene (Cp Ϫ848,Ϫ1 -Luc) confirm that IL-6 transcriptional activity is completely blocked by insulin. Thus, Cp joins a family of insulin-inhibited acute phase genes including haptoglobin, thiostatin, complement C3, and C-reactive protein (6,11). The negative regulation by insulin may be mediated by at least three pathways. Insulin reduces IL-6-induced STAT-3 activity as measured by gene transcription, mRNA accumulation, protein concentration, and DNA binding activity (7). Insulin also reduces the expression of the IL-6 receptor ␣-subunit and IL-6 receptor binding (7). Finally, insulin suppresses transactivation by C/EBP␤ (8). The presence of an active C/EBP␤-binding site in the rat Cp promoter (56) and the consensus STAT3-binding sites suggest that either or both of these pathways may be involved. Our results indicate that the direction and magnitude of the response of the Cp promoter to insulin is condition-dependent; under normal physiological conditions, insulin increases Cp transcription, whereas, during the systemic acute phase response or localized inflammation, insulin reduces Cp transcription. We are not aware of other acute phase reactants that are transcriptionally activated by insulin; thus, the response of Cp may be unique. The dual response of Cp to insulin may reflect its pleiotropic nature, with FIG. 7. Specific binding of an insulin-stimulated transcription factor complex to the Cp HRE. A, EMSAs were done to determine Cp HRE-binding complexes. HepG2 cells were treated for 8 h with insulin (Ins., 100 nM) or with 1% O 2 (Hpx.). Nuclear extracts were mixed with a 32 P-labeled, double-stranded 24mer probe containing either the Cp HRE or the erythropoietin (Epo) HRE. Probebound complexes were resolved by 5% nondenaturing PAGE and visualized by autoradiography. Cont., control. The positions of the putative HIF-1 and constitutive (Const.) complexes are indicated by arrows. B, competitor binding to show specificity of binding of a complex to the Cp HRE. HepG2 cells were treated with insulin (100 nM) for 16 h and nuclear extracts prepared. Radiolabeled probe was pre-mixed with unlabeled, annealed, 24mer oligonucleotide competitor at 10-, 100-, or 300-fold molar excess (increasing concentration is indicated by upward slope of triangle) before addition to the nuclear extracts. The competitor probes were wild-type (wt) Cp HRE, mutant (mut) Cp HRE containing a CGT-to-AAA mutation (underlined), and erythropoietin (Epo) HRE. The extracts were mixed with a 32 P-labeled probe containing the Cp HRE, and probe-bound complexes identified as in A.
possibly distinct functions in inflammation and iron homeostasis. Suppressive activities of insulin have been reported in other enzyme systems, e.g. it inhibits transcriptional activation of fructose-2,6-bisphosphatase by dexamethasone in rat hepatoma FTO-2B cells (57). Insulin by itself causes transcriptional activation of this enzyme in hepatic FAO-1 (58). Thus, the function of insulin may be bidirectional in this system as well; however, a direct demonstration using a single type of cell has not been demonstrated.
Activation of Cp Transcription by Insulin-We observed a 2-5-fold insulin-stimulated increase in Cp transcription, in Cp mRNA level, and in the rate of Cp secretion in HepG2 cells. We also found that the stimulation of Cp gene expression by a maximal concentration of insulin was much lower than that seen upon treatment with 1% O 2 . Our observations are comparable with a previous report of 2-and 4-fold induction of aldolase A transcription in HepG2 cells by insulin and hypoxia, respectively (36). The differential expression may reflect the 2-fold higher level of HIF-1␣ in cells treated with hypoxia compared with cells treated with insulin (59); however, these results were obtained in embryonic kidney 293 cells, and cell type-specific differences in HIF-1␣ activation by insulin have been observed (36). The mechanism(s) of HIF-1␣ induction by hypoxia and by insulin is not clearly understood. Hypoxia increases the amount of HIF-1␣ by stabilization of the protein (60). Under normoxic conditions, rapid, ubiquitin-dependent degradation of HIF-1␣ occurs, whereas hypoxic stress inhibits this process (61). There is evidence that insulin utilizes a similar mechanism for stabilization of HIF-1␣ (36). Recent reports suggest that HIF-1␣ activation by both hypoxia and insulin utilize the phosphatidylinositol 3-kinase pathway (62,63).
Role of Insulin in Regulating Cp Levels in Vivo-Our results may help to clarify contradictory studies on Cp expression in diabetic patients and in animal models of diabetes. Elevated plasma Cp levels have been reported in non-insulin-dependent diabetes mellitus (NIDDM, type II diabetes) (64 -66), but normal Cp levels have been reported as well (67,68). The reports on insulin-dependent diabetes mellitus (IDDM, type I diabetes) are similarly inconclusive, with one report indicating decreased plasma Cp (69), whereas others indicate increased Cp (70). The lack of a consistent response of serum Cp to diabetes has been ascribed to variability in the acute phase response (64), diabetic vascular complications (71), oxidative stress (70), extent of hyperglycemia (66), serum nitric oxide (72), and to methodological differences (70).
Our in vitro results suggest that the circulating level of insulin may be a critical regulator of hepatic Cp synthesis, and consequently, plasma Cp level. In this case, part of the variation in Cp levels in NIDDM patients may be a result of the large range in plasma insulin values from normal to very high (e.g. in early phases of the disease and in diabetes related to obesity), and possibly of the oscillatory nature of insulin levels in NIDDM. IDDM patients require insulin for survival, and thus the method and extent of diabetic control may influence Cp status. Animal models of diabetes provide additional insights. In both alloxan-and streptozotocin-induced diabetic rats, the decrease in insulin level is accompanied by substantial reductions in plasma Cp (73)(74)(75), and treatment of diabetic rats with insulin restores plasma Cp to normal levels (75). Similarly, treatment of rats with streptozotocin decreases Cp synthesis by the isolated perfused liver, but is restored to normal after treatment of the rats with insulin (76). Taken together, these data suggest a possible regulatory activity of insulin on hepatic Cp synthesis and Cp plasma concentration in vivo. Our results suggest that the relationship between plasma Cp and diabetes is likely to be a complex function including dependences on both insulin levels and the extent of inflammation.