C/EBPα Regulates Hepatic Transcription of Hepcidin, an Antimicrobial Peptide and Regulator of Iron Metabolism

Originally identified as a gene up-regulated by iron overload in mouse liver, the HEPC gene encodes hepcidin, the first mammalian liver-specific antimicrobial peptide and potential key regulator of iron metabolism. Here we demonstrate that during rat liver development, amounts of HEPC transcripts were very low in fetal liver, strongly and transiently increased shortly after birth, and reappeared in adult liver. To gain insight into mechanisms that regulate hepatic expression of hepcidin, 5′-flanking regions of human and mouse HEPC genes were isolated and analyzed by functional and DNA binding assays. Human and mouseHEPC promoter-luciferase reporter vectors exhibited strong basal activity in hepatoma HuH-7 and mouse hepatocytes, respectively, but not in non-hepatic U-2OS cells. We found that CCAAT/enhancer-binding protein α (C/EBPα) and C/EBPβ were respectively very potent and weak activators of both human and mouse promoters. In contrast, co-expression of hepatocyte nuclear factor 4α (HNF4α) failed to induce HEPC promoter activity. By electrophoretic mobility shift assay we demonstrated that one putative C/EBP element found in the human HEPCpromoter (−250/−230) predominantly bound C/EBPα from rat liver nuclear extracts. Hepatic deletion of the C/EBPα gene resulted in reduced expression of HEPC transcripts in mouse liver. In contrast, amounts of HEPC transcripts increased in liver-specific HNF4α-null mice. Decrease of hepcidin mRNA in mice lacking hepatic C/EBPα was accompanied by iron accumulation in periportal hepatocytes. Finally, iron overload led to a significant increase of C/EBPα protein and HEPC transcripts in mouse liver. Taken together, these data demonstrate that C/EBPα is likely to be a key regulator of HEPC gene transcription and provide a novel mechanism for cross-talk between the C/EBP pathway and iron metabolism.

the acquired immune system, characterized by highly specific but delayed onset of response to initial pathogen exposure and the non-targeted innate immunity acting as a first barrier host defense. The ability to avoid microbial infection relies for the major part on constitutive or inducible endogenous synthesis of a battery of microbicidal substances named antimicrobial peptides that behave as natural antibiotics of the organism (1,2). These molecular effectors, universally found in procaryotes, plants, and animals ranging from insects to mammals, exhibit a broad spectra of activities on bacteria and fungi (3)(4)(5)(6). Antimicrobial peptides, typically 20 -40 amino acids in length, are rich in cationic residues and adopt an amphipathic structure.
Recently, we identified a novel mouse gene named HEPC (7) encoding a small 83-aa protein that shares significant homology in its C-terminal region with human peptide hepcidin (also termed LEAP-1) isolated in two parallel studies from human urine (8) and plasma ultrafiltrate (9). Hepcidin exhibits a broad range of activity against Gram-positive and Gram-negative bacteria and fungi (8,9). However, in contrast to described mammalian antimicrobial peptides including ␣and ␤-defensins (4,5), which are expressed mainly in circulating neutrophil granulocytes and in various epithelia including gastrointestinal, urogenital, and respiratory tracts, hepcidin exhibits predominant liver expression. Because in Drosophila most antimicrobial peptides were produced in the fat body, a functional equivalent of the liver (10), hepcidin could represent the first mammalian hepatic homologue of insect cysteine-rich antimicrobial peptides.
Hepcidin was identified as a gene induced in mouse liver by iron excess (7). In addition, expression of hepcidin was enhanced in mouse hepatocytes both in vivo and in vitro after exposure to bacterial lipopolysaccharides (7) and was dramatically induced following bacterial challenge in bass liver (11). Induction of hepcidin in these two unrelated physiopathological situations could indicate that besides its antimicrobial functions this peptide exhibits other functions. Several antimicrobial peptides display functions not directly related to host defense including prevention from oxidative stress (12), antitumor activity (13), and regulation of angiogenesis (14).
Recent reports argue in favor of a role of hepcidin in iron metabolism. Indeed, analysis of upstream stimulatory factor 2 (USF2) 1 knockout mice revealed a development of iron overload in liver, pancreas, and heart. In these mice, it appeared that two very highly related mouse hepcidin genes, HEPC1 and HEPC2, lying in close proximity to the USF2 gene, were transcriptionally inactive, and livers of USF2 Ϫ/Ϫ mice did not contain HEPC transcripts. It has been suggested that the lack of hepcidin expression was responsible for the liver iron-overload phenotype observed in USF2 Ϫ/Ϫ (15,16). Inversely, early liver expression of the HEPC1 gene in transgenic mice resulted in phenotypic traits of iron deficiency (17). These results strongly suggest that hepcidin plays an important role in the regulation of iron homeostasis (15,16).
In this study, to understand molecular mechanisms involved in the regulation of hepcidin in the liver, we have cloned and analyzed human and mouse HEPC promoters and investigated the transcriptional control of the HEPC gene in vitro and in vivo. We demonstrated that liver-enriched nuclear factor C/EBP␣, a critical regulator of several hepatic metabolic processes (18), is likely to play an essential role in the control of HEPC gene expression.

MATERIALS AND METHODS
Animals-For fetal rat liver samples, pregnant female Sprague-Dawley rats were purchased from Charles River (St. Aubin-les-Elbeuf, France). Breeding was done by placing female rats with males of the same strain overnight, and the noon of the next day was considered as 0.5 day postcoitum. On the indicated days of gestation, rats were anesthetized, embryos were removed, livers were minced, and washed briefly with phosphate-buffered saline.
The mice with liver-specific deletion of the C/EBP␣ gene will be reported elsewhere. Briefly, C/EBP␣-floxed allele described earlier (19) was bred with the rat albumin Cre transgene (20). The gene deletion commenced at about 2 weeks of age at the onset of expression of the albumin promoter and completely deleted at age 6 weeks when the promoter was maximally expressed (21). Liver-specific HNF4␣-null mice were generated by Cre-loxP-mediated deletion of exons 4 and 5 of the HNF4␣ gene as previously described (21,22).
For experimental iron overload, 5-week-old BALB/cJ male mice were obtained from CERJ (Le Genet St. Ile, France). Minced rat and mice liver samples were frozen in liquid nitrogen and kept at Ϫ80°C until further processing.
Iron Overload-Carbonyl iron overload was performed as previously described (23). Briefly, a group of five mice was iron-overloaded by 3% carbonyl iron supplemented in the diet (AO3, UAR, France) over 8 months. Control mice had a carbonyl iron-free diet. Under these experimental conditions, liver iron concentration was on average 13.8-fold higher than the control value (158.2 mol of iron/g of dry weight liver versus 11.5 mol, respectively).
Hepatocyte Isolation and Culture-Hepatocytes were isolated by a two-step collagenase perfusion procedure from normal male Sprague-Dawley rats or C57BL6 mice. Hepatocytes were seeded in a mixture of 75% minimum essential medium (MEM) and 25% medium 199, supplemented with 10% FCS and per ml: 100 IU of penicillin, 50 g of streptomycin sulfate, 1 mg of bovine serum albumin, and 5 g of bovine insulin. Four hours later, the medium was renewed with the same medium deprived of FCS and supplemented with 1 ϫ 10 Ϫ6 M hydrocortisone hemisuccinate. It was changed every day thereafter.
RNA Isolation and Northern Blot Analysis-Total RNA was extracted by the guanidine thiocyanate method following sedimentation through a cesium chloride cushion or using the SV total RNA isolation system (Promega, Charbonnières, France). 10 -20 g of RNA was separated by electrophoresis through 1.2% agarose gel in 2.2 M formaldehyde, 20 mM MOPS (pH 7.0), 10 mM sodium acetate, 1 mM EDTA and transferred onto a nylon membrane (Hybond Nϩ, Amersham Biosciences) by capillary blotting. Hybridization was carried out in the presence of 32 P-labeled cDNA probe. The equivalence of RNA loading was assessed by ethidium bromide staining and/or hybridization with a glyceraldehyde-3-phosphate dehydrogenase cDNA probe.
cDNA Probes-Mouse hepcidin cDNA probe was prepared as described previously (7). cDNAs encoding rat and human hepcidin were generated by polymerase chain reaction (PCR) from reverse-transcribed rat liver and human hepatoma HepG2 mRNAs, respectively. Obtained PCR products were cloned in pCRII-TOPO vector (InVitrogen, NV Leek, The Netherlands) and sequenced. The cloned fragments were then released by EcoRI digestion and used as probes for Northern blot analysis. Rat ␣-fetoprotein (AFP) probe was obtained from J. Kruh. The mouse C/EBP␣ and HNF4␣ cDNA probes were generated as described in Refs. 19 and 21, respectively. Mouse ApoC-III probe was amplified from mouse liver cDNAs using gene-specific primers and cloned in pCRII-TOPO vector and sequenced.
Real-time RT-PCR Analysis-Total RNA extraction was carried out from normal human liver and HuH-7, HepG2, and U-2OS cells using the SV total RNA isolation system (Promega), and cDNA first-strand synthesis was performed with the Advantage RT-for-PCR kit (Clontech, Palo Alto, CA). Primers and probes were designed with the assistance of Primer Express Software (PE Biosystems). The following probes were used to detect human HEPC and ␤-actin amplification products: 5Ј-FAM-CTGCGGCTGCTGTCATCGATCAA-TAMRA-3Ј and 5Ј-FAM-AT-GGGACGACATGGAGAAAATCTGGC-TAMRA-3Ј, respectively. The relative expression of HEPC in the different samples was determined by standardization to its expression in U-2OS cells (which exhibits the lowest level of HEPC expression) using a two-step calculation method. First, for each sample, Ct actin value was subtracted from the corresponding Ct HEPC . Then, these normalized values were used to calculate relative level of HEPC expression as follows: HEPC expression ratio (sample/U2OS) ϭ 2 ⌬Ct where ⌬Ct ϭ (Ct U2OS Ϫ Ct sample ). The HEPC expression level in U-2OS cells was used as a reference and was set as 1-fold expression. All quantitative real-time PCR assays were performed in triplicate in 96-well microplates using qPCR™ Core reagent kit (Eurogentec, Seraing, Belgium) and ABI PRISM 7700 Sequence Detection System (PE Biosystems). For each microplate the reactions contained target cDNAs, two standard curves (prepared from human liver), and two negative controls: no-template control (sterile water) and no-amplification control (omitting the RT step). The liver sample was obtained from a patient who underwent hepatic surgery for a liver metastasis of colonic adenocarcinoma. A sample of non-tumoral area of the liver was immediately frozen in liquid nitrogen and stored at Ϫ80°b efore use. Surgical procedures and collection of liver samples were performed under informed consent and in compliance with French law and regulations.
Western Blot Analysis-Liver tissue lysates were prepared as follows: minced liver samples were lysed in a buffer containing 50 mM HEPES, pH 7.5, 150 mM NaCl, 1 mM EDTA, 5 mM EGTA, 0,1% Tween 20, 10% glycerol, 0.1 mM phenylmethylsulfonyl fluoride, and 1ϫ protease inhibitor mixture (Roche Diagnostics, Meylan, France). Protein concentrations were determined by Bio-Rad protein assay. Total cellular proteins (100 g/lane) were resolved by SDS-PAGE, transferred to nitrocellulose membrane (Amersham Biosciences), incubated with blocking buffer (3% bovine serum albumin in Tris-buffered saline). Rabbit polyclonal anti-C/EBP␣ antibody was a gift from G. Darlington (Baylor College of Medicine, Houston). Following incubation with peroxidase-conjugated goat anti-rabbit immunoglobulins (DAKO, Denmark) as second antibodies, peroxidase activity was detected by enhanced chemiluminescence according to the manufacturer's instructions (Amersham Biosciences).
Cloning of Human and Mouse HEPC Promoters and Construction of Reporter Vectors-950-and 624-bp nucleotide fragments of the 5Ј-flanking genomic region of the human HEPC gene were obtained from human genomic DNA (Clontech) by PCR amplification using the following forward primers: 5Ј-GGCTCGAGGTACTCATCGGACTGTAGATG-TTAGC-3Ј and 5Ј-GGCTCGAGCTGTGCTGGGCCATATTACTGCTGT-C-3Ј, respectively, and the reverse primer: 5Ј-GGAAGCTTGTGACAG-TCGCTTTTATGGGGCCTGC-3Ј (the incorporated XhoI and HindIII sites are underlined). PCR fragments were digested with XhoI-HindIII and were inserted into the same sites of the promoter-less luciferase reporter vector pGL3-Basic (Promega). These were designated pHEPCϪ960/Ϫ9 and pHEPCϪ633/Ϫ9. Constructs containing 5Ј-deletions of the human HEPC promoter (pHEPCϪ327/Ϫ9, Ϫ275/Ϫ9, Ϫ223/ Ϫ9, and Ϫ108/Ϫ9) were generated by PCR using pHEPCϪ633/Ϫ9 as template and specific primers containing XhoI and HindIII restriction sites. After digestion with XhoI-HindIII, PCR fragments were cloned in the pGL3-Basic vector. A 789-bp nucleotide fragment of the 5Ј-flanking genomic region of the mouse HEPC1 gene were obtained from mouse genomic DNA (Clontech) by PCR amplification using the forward primers: 5Ј-GCGCTCGAGGAATACATCGTCAAGCCAGAC-3Ј and the re-verse primer: 5Ј-GCGAAGCTTGTGTGGTGGCTGTCTAGGAGC-3Ј (the incorporated XhoI and HindIII sites are underlined). The PCR fragment was digested with XhoI-HindIII and was inserted into the pGL3-Basic vector. The obtained construct was designated pmHEPCϪ783/ϩ5. Similar to human HEPC promoter constructs, 5Ј-deletions of mouse HEPC1 promoter (pmHEPCϪ554/ϩ5, Ϫ338/ϩ5, and Ϫ267/ϩ5) were generated by PCR. Details of constructs are available upon request. All constructs were verified by DNA sequencing. Searches for transcription factor binding sites was performed using MatInspector (24).
Cell Transfection and Luciferase Assay-U-2OS or HuH-7 cells were seeded in 35-mm-diameter dishes and grown to 50 -70% confluency. Cells were transiently transfected with 1.5 g of promoter-less pGL3-Basic vector or positive control pGL3-Promoter vector or constructs containing 5Ј-flanking regions of the HEPC gene and 0.5 g of C/EBP␣, C/EBP␤, HNF4␣ expression plasmids, or empty vector. The pEGFP-C3 vector encoding green fluorescent protein (GFP) used as a co-transfectant (0.5 g per dish) was served as the control for transfection efficiency. Transfections were performed using liposome-based transfection reagent Lipofectin (Invitrogen). Cells were maintained in the presence of transfection reagent in Opti-MEM (Invitrogen) for 16 h and then switched to growth medium. In all cases, the transfection efficiency was at least 10 -15%. Primary mouse hepatocytes were transfected 24 h after seeding using GB12 reagent as previously described (25). The following expression plasmids were used for co-transfection: pCMV-C/EBP␣ and pCMV-NF-IL6 encoding C/EBP␣ and C/EBP␤, respectively (kindly provided by G. Darlington, Baylor College of Medicine, Houston) and a HNF4␣ expression vector pSG5/rHNF4. 42 or 30 h after transfection of cell lines or primary mouse hepatocytes, respectively, cells were lysed in luciferase cell culture lysis reagent (Promega), and cellular extracts were analyzed for luciferase activity by liquid scintillation counting using the Luciferase assay system from Promega. The luciferase activity was expressed as cpm/g of cellular protein.
Double-stranded C/EBP wild type and mutant oligonucleotides were purchased from Geneka Biotechnology (Montréal, Canada). The singlestranded oligonucleotides, which contain the putative C/EBP binding site in the 5Ј-flanking region of the HEPC gene, were from InVitrogen.
Oligonucleotide probes were labeled with [␥-32 P]ATP and T4 polynucleotide kinase (Amersham Biosciences). Binding reactions were performed in 24 l of binding buffer containing 6 g of rat liver nuclear extract and 100 ϫ 10 3 cpm of labeled oligonucleotide probe. The reaction mixture was incubated at 4°C for 20 min and then was loaded onto a 6% nondenaturating polyacrylamide gel. The electrophoresis was run in 0.5ϫ Tris-Boric acid electrophoresis buffer.
For competition experiments, cold competitors were included in the binding mixture before adding the nuclear extract. For supershift experiments, nuclear extracts were preincubated with the antibody to C/EBP␣ or C/EBP␤ (Geneka Biotechnology) for 20 min at 4°C.
Preparation of Liver Tissues and Iron Staining-Livers from 2-month-old liver-specific C/EBP␣-null and control mice were fixed in 10% neutral buffered formalin and embedded in paraffin. Then, sections cut at a thickness of 5 m were mounted on the slides. Iron staining was performed using Accustain iron stain kit (Sigma).
Statistical Analysis-Results were expressed as mean Ϯ S.D. Student's t test was used for estimation of statistical significance (p Ͻ 0.01 was considered as significant).

HEPC Gene Expression Is Regulated during Rat Liver Development and Is
Related to Hepatocyte Phenotype-Expression of the HEPC gene was analyzed during rat liver development from day 14.5 of embryonic age. HEPC transcripts were undetectable in 14.5-and 16.5-day fetal liver, appeared at 18.5 day of embryonic age, was very strongly induced shortly after birth, and then abruptly ceased from the third day of postnatal period. Then, amounts of HEPC mRNA began to accumulate from 28th day after the birth, reaching the highest levels in adult liver (Fig. 1A). In parallel, expression of AFP, a well known liver-specific developmentally regulated gene, was studied. AFP transcripts were highly expressed in fetal and perinatal liver and abruptly shut off after 15 days, in accordance to previous studies (28). These observations, along with our previous analysis of HEPC expression (7), suggested that this gene is liver-specific and its expression is related to the fully differentiated hepatocyte. To confirm this issue, we analyzed levels of HEPC transcripts by Northern blotting in normal hepatocytes and in a set of rat and human liver-derived cell lines. In accordance with our previous data on mice hepatocytes (7), freshly isolated rat hepatocytes contained higher amounts of HEPC transcripts in comparison with 48-h-old cultures (Fig.  1B). Furthermore, HEPC mRNAs could not be detected in rat hepatoma cell lines HTC and FAZA after a 2-week exposure.
The survey of four human hepatoma cell lines (HepG2, Hep3B, HuH-7, and PLC/PRF/5) for hepcidin expression by Northern blotting revealed a hybridization signal in HuH-7 and HepG2 cells following a 1-week exposure time. In contrast, HEPC transcripts from normal human liver were clearly detectable as early as 2 h after autoradiography (data not shown). In order to assess more accurately the relative level of hepcidin expression in adult liver in comparison to HuH-7 and HepG2 hepatoma cells and non-hepatic U-2OS cell line, we performed quantitative real-time RT-PCR assays. Expression of the HEPC gene in adult human liver was found to be 142 times stronger than in HuH-7 hepatoma cells. (Fig. 1C).
Localization of the Human HEPC Gene, and Sequence Analysis of Human and Mouse HEPC Promoters-To understand molecular mechanisms that confer liver specificity and are involved in transcriptional regulation of hepcidin during development and iron overload, we isolated and characterized the 5Ј-flanking region of the human HEPC gene and analyzed its promoter activity. Analysis of genomic sequence data of human clones R30879 and F24108 (GenBank TM accession numbers AD000684 and AC002132, respectively) allowed us to map the HEPC gene to chromosome 19, band q13.1 between the USF2 gene and the MAG gene encoding myelin-associated glycoprotein precursor (Fig. 2). Interestingly, two liver-specific genes LISCH7 and C/EBP␣ are also mapped to 19q13.1. In particular, the LISCH7 gene is located immediately upstream of the USF2 gene.
A 950-bp 5Ј-flanking genomic fragment of the human HEPC gene was cloned in the pGL3-Basic vector, sequenced, and analyzed using bioinformatic tools for the presence of potential transcription factor binding sites. This region contains a sequence with a TATA box homology. Among a number of putative response elements, we found binding sites for the liverenriched transcription factor HNF-4 (Ϫ75 to Ϫ62 and Ϫ602 to Ϫ589 from the predicted translation start site) and for members of C/EBP family (Ϫ249 to Ϫ236, Ϫ298 to Ϫ285 and Ϫ90 to Ϫ77) (Fig. 3). The proximal putative C/EBP binding site TTCT-TGGAAATGA overlaps with the STAT consensus binding site TTCNNNGAA.
Sequence analysis of a 789-bp nucleotide fragment of the 5Ј-flanking genomic region of the mouse HEPC1 gene revealed that, similarly to human HEPC promoter, it contains at least three putative binding sites for the C/EBP family of transcription factors, including the proximal C/EBP binding site that overlaps with the STAT site and a sequence related to the HNF4 binding site (data not shown).
Hepatocyte-specific Promoter Activity of Hepcidin Gene-Two reporter vectors, pHEPC-960/Ϫ9 and pHEPCϪ633/Ϫ9, containing respectively 950-bp and 624-bp nucleotide fragments of the human HEPC 5Ј-flanking genomic region fused to a luciferase reporter gene were tested for transcriptional activity in human hepatoma cell line HuH-7 and non-hepatic osteosarcoma U-2OS cells. Promoter-less pGL3-Basic vector and pGL3-Promoter vector containing SV40 promoter upstream of the luciferase gene were used as negative and positive control, respectively. Transfection of reporter vector containing 950 bp of the HEPC 5Ј-flanking region (pHEPCϪ960/Ϫ9) resulted in high luciferase activity in HuH-7 cells whereas the promoter activity was very low in non-hepatic U-2OS cells. Likewise, the truncated 624-bp HEPC promoter construct exhibited a very low luciferase activity in U-2OS cells; however, in HuH-7 cells it was even more active than the 950-bp construct (Fig. 4A) suggesting that the region between Ϫ960 and Ϫ633 contains negative cis-acting elements.
Similar to the human HEPC promoter, a reporter vector pmHEPCϪ783/ϩ5 containing 789 bp of the mouse HEPC1 gene 5Ј-flanking region exhibited low luciferase activity in U-2OS cells. In contrast, transfection assays with primary mouse hepatocytes showed high reporter activity (Fig. 4B). Furthermore we performed a more detailed analysis of the mouse HEPC1 promoter activity. For this purpose several promoter-luciferase reporter plasmids were constructed and transfected into primary mouse hepatocytes. As shown in Fig. 4C, deletion of the 5Ј-flanking DNA from bp Ϫ783 to bp Ϫ554 that contains a distal putative C/EBP binding site had no significant effect on the activity of the mouse hepcidin promoter. Deletion to bp Ϫ338 resulted in moderate increase of luciferase activity and, finally, further deletion of DNA between Ϫ338 and Ϫ267 containing second putative C/EBP binding site, strongly (ϳ12fold) decreased promoter activity.
Liver-enriched Transcription Factor C/EBP␣ but Not HNF4␣ Can Transactivate Human and Mouse HEPC Gene Promoters in Vitro-Because sequence analysis of 5Ј-flanking region of human and mouse HEPC genes revealed several putative C/EBP and HNF4 binding sites we studied the involvement of C/EBP␣ and ␤, members of C/EBP family of transcription factors, and HNF4␣ on its activity. While co-transfection of C/EBP␣ with the human hepcidin promoter reporter vector pHEPCϪ633/Ϫ9 into U-2OS cells very strongly increased luciferase activity, the C/EBP␤-mediated induction of promoter activity was relatively weak. In contrast, co-expression of HNF4␣ had ϳ2-fold inhibitory effect on luciferase activity (Fig.  5A). In a similar manner, co-transfection of C/EBP␣ or C/EBP␤ with pmHEPCϪ783/ϩ5 reporter vector that contains 5Ј-flanking region of the mouse HEPC1 gene, resulted respectively in relatively strong and weak increase of luciferase activity. HNF4␣ co-expression had no significant effect on the mouse HEPC promoter activity (Fig. 5B).
Finally, to analyze precisely the involvement of C/EBP␣ in the transcriptional control of HEPC promoter, a series of plasmids fusing 5Ј-flanking DNA of the human HEPC gene were constructed and tested for promoter activity in the absence or presence of C/EBP␣ in U-2OS cells. Deletion of the 5Ј-flanking DNA between Ϫ633 and Ϫ327 resulted in an increase on the magnitude of C/EBP␣ induction. Deletion of the DNA region (between Ϫ327 and Ϫ275) that contains the distal putative C/EBP␣ binding site led to the decrease of C/EBP␣ inductibility to levels observed with the pHEPC (Ϫ633/Ϫ9) reporter vector.
Cotransfection of C/EBP␣ with pHEPC (Ϫ223/Ϫ9) reporter vector lacking the second putative C/EBP␣ binding site had an important inhibitory effect on C/EBP␣-mediated modulation of promoter activity. Finally, further deletion to Ϫ108 had no significant effect on the magnitude of the C/EBP induction (Fig. 5C).
C/EBP␣ Binds to the HEPC Promoter-Furthermore, in order to demonstrate that the positive effect of C/EBP␣ on HEPC promoter activity was direct, we performed EMSA using adult rat liver nuclear extracts, together with three 32 P-labeled double-stranded oligonucleotides, HEPC(Ϫ300/Ϫ280) HEPC-(Ϫ250/Ϫ230), and HEPC(Ϫ92/Ϫ72) whose sequences encompass putative C/EBP binding sites in the 5Ј-flanking region of the human HEPC gene. As expected, C/EBP binding activity was detected by EMSA in rat liver nuclear extracts using reference C/EBP oligonucleotide probes. The DNA binding was specific since it was efficiently suppressed by addition of 100fold molar excess of the unlabeled wild-type but not mutant C/EBP probe. Preincubation of nuclear extracts with specific antibodies to C/EBP␣ resulted in partial supershifting of binding complexes (Fig. 6). HEPC(Ϫ300/Ϫ280) probe that contains the distal potential C/EBP binding site showed low levels of binding activity with rat liver nuclear extracts. In contrast, using labeled HEPC(Ϫ250/Ϫ230) probe representing the second C/EBP binding site we found that liver nuclear extracts contained strong binding activity, and the migration profile was similar to that obtained with reference wild-type C/EBP oligonucleotide probe. The DNA binding was efficiently com- FIG. 4. Hepatocyte-specific activity of human and mouse HEPC promoters. A, luciferase reporter vectors pHEPCϪ960/Ϫ9 and pHEPCϪ633/Ϫ9 containing 5Ј-flanking regions of the human HEPC gene were transfected into hepatoma HuH-7 and non-hepatic U-2OS cells. pGL3-Basic and pGL3-Promoter vectors were used as negative and positive controls, respectively. In all cases, transfection efficiency assessed by co-transfected vector pEGFP-C3-encoding GFP was ϳ10 -15%. B, the reporter vector pmHEPC-783/ϩ5 containing 789 bp of the 5Ј-flanking region of the mouse HEPC1 gene was transfected into primary mouse hepatocytes (MH) and U-2OS cells. C, various luciferase reporter plasmids containing the mouse HEPC1 promoter (Ϫ267, Ϫ338, Ϫ554, and Ϫ783/ϩ5 bp from transcription initiation codon) were transfected into primary mouse hepatocytes. Transfection efficiency assessed by expression of GFP in mouse hepatocytes was ϳ15-20%. Luciferase activity for each condition is presented as cpm/g of cellular protein.
Transfections were performed in triplicates, and results are presented as means Ϯ S.D. Similar results were obtained from three independent experiments for the human HEPC promoter activity and from two independent experiments for mouse promoter activity in mouse hepatocytes. *, p Ͻ 0.01 between promoter activity in HuH-7 (or mouse hepatocytes) and U-2OS.

FIG. 5. Analysis of transactivation of the human and mouse HEPC promoters by C/EBP␣, C/EBP␤, and HNF4␣.
Expression vectors encoding C/EBP␣, C/EBP␤, and HNF4␣ or empty vectors were co-transfected with human HEPC promoter-luciferase vector pHEPCϪ633/Ϫ9 (A) or mouse HEPC1 promoter-luciferase plasmid pm-HEPC-783/ϩ5 (B) into U-2OS cells. C, a series of reporter plasmids fusing variable amounts of human HEPC promoter (Ϫ108, Ϫ223, Ϫ275, Ϫ327, and Ϫ633/Ϫ9 bp from transcription initiation codon) to luciferase were co-transfected with or without C/EBP␣ into U-2OS cells. In all cases, transfection efficiency assessed by expression of GFP was ϳ10 -15%. Luciferase activity is presented as cpm/g of cellular protein.
Transfections were performed in triplicates, and results are presented as means Ϯ S.D. * indicates significant modulation of promoter activity by co-transfected expression vectors (p Ͻ 0.01).
peted with 100-fold molar excess of unlabeled reference wildtype C/EBP and homologous HEPC (Ϫ250/Ϫ230) oligonucleotide probes but not with mutant C/EBP probe. Furthermore, a supershift reaction was evidenced when anti-C/EBP␣ but not anti-C/EBP␤ antibodies were added to the binding mixture. Finally, the HEPC (Ϫ92/Ϫ72) probe that contains the overlapping C/EBP and STAT sites bound rat liver nuclear proteins. All binding was efficiently suppressed by addition of the unlabeled homologous oligonucleotide probe. However, competition performed with reference wild-type C/EBP probe resulted in partial blocking of binding activities. In addition, no supershifting was detected by addition of either C/EBP␣ or C/EBP␤ antibodies (Fig. 6). Taken together, the results of transfection experiments and gel shift assays suggest that the HEPC (Ϫ250/ Ϫ230) C/EBP binding site is more important for activation of the human HEPC promoter by C/EBP␣ than two other sites.

C/EBP␣ and HNF4␣ Control Expression of HEPC Transcripts in Mice Livers in an Opposite
Manner-To validate the data obtained from in vitro experiments and to determine whether liver-enriched transcription factors C/EBP␣ and HNF4␣ can control transcription of the HEPC gene in the same manner in vivo, we studied the role of these proteins in the regulation of hepcidin expression in an intact mouse model. Because the standard HNF4␣ and C/EBP␣ gene knockout led to early embryonic lethality or to death shortly after the birth, respectively (29,30) we investigated expression of hepcidin in mice with conditional liver-specific disruption of C/EBP␣ or HNF4␣ genes. Specifically, amounts of HEPC transcripts were determined by Northern blotting in mice in which C/EBP␣ gene was specifically disrupted in the liver by Cre-mediated recombination (C/EBP␣LivKO) and in control mice containing a functional C/EBP␣ gene (C/EBP␣Flox). As shown in Fig. 7A, control mice expressed C/EBP␣ transcripts (lanes 1 and 2) and contained also high amounts of liver HEPC mRNA. In contrast, the hepatic C/EBP␣ gene-deficient mice (lanes 3 and 4) showed decreased amounts of HEPC transcripts in the liver. Specifically, the mouse with undetectable expression of C/EBP␣ gene (lane 3) contained the lowest amount of HEPC transcripts and those expressing some C/EBP␣ mRNA (lane 4) exhibited elevated expression of hepcidin.
To investigate whether the expression of hepcidin was affected in vivo by disruption of the hepatic HNF4␣ gene, RNA from 45-day-old male HNF4␣LivKo mice (21) lacking HNF4␣ in the liver and control animals was analyzed for the presence of HEPC transcripts. While mRNA levels for apolipoprotein C-III (ApoC-III) were drastically decreased in liver-specific HNF4␣-null mice in accordance with previous data (21), liver expression of hepcidin was 4.1-fold higher (p Ͻ 0.01) in these animals (Fig. 7B). Nicolas et al. (15) demonstrated previously that USF2 Ϫ/Ϫ mice had undetectable expression of hepcidin mRNA in the liver and exhibited massive iron overload in the liver. We investigated whether the decreased expression of hepcidin in hepatic C/EBP␣-deficient mice could also result in alteration of iron metabolism. Iron staining was performed on livers of 2-month-old control (C/EBP␣Flox) and liver-specific C/EBP␣-null mice (C/EBP␣LivKO). While no iron accumulation was detected in the group of control mice, 4 of 5 C/EBP␣null mice exhibited the presence of iron deposits, visible as blue staining, mainly in periportal hepatocytes (Fig. 8).

Liver-specific Deletion of the C/EBP␣ Gene Results in Hepatic Iron Overload in Mice-
Iron Overload Induces C/EBP␣ Expression-Iron overload observed in liver-specific C/EBP␣-null mice strongly indicates a possible involvement of C/EBP␣ in the regulation of iron metabolism. The positive effect of C/EBP␣ on the HEPC gene transcription in vitro and in vivo, and previously described up-regulation of hepcidin expression in the liver by iron excess suggested that this effect is mediated by C/EBP␣ and prompted us to investigate expression of this transcription factor during iron overload. In accordance with our previous observations, amounts of HEPC mRNA in liver tissue strongly increased in all animals overloaded by 3% carbonyl iron for 8 months (Fig.  8). Interestingly, these mice also showed a 2.1-fold increase (p Ͻ 0.01) of the C/EBP␣ protein in the liver compared with the control ones (Fig. 9).  1 and 2) and from C/EBP␣LivKO mice (C/EBP␣ fl/fl ϫ albumin-Cre ϩ/Ϫ ) in which the C/EBP␣ gene was specifically disrupted in the liver (lanes 3 and 4). 10 g of RNA were separated by electrophoresis and subjected to Northern blot analysis using mouse hepcidin and C/EBP␣ cDNA probes. B, total RNA was extracted from livers of 45day-old control HNF4␣Flox (HNF4␣ fl/fl ϫ albumin-Cre Ϫ/Ϫ ) mice and from liver-specific HNF4␣-null mice HNF4␣LivKO (HNF4␣ fl/fl ϫ albumin-Cre ϩ/Ϫ ). 10 g of RNA were separated by electrophoresis, transferred to a nylon membrane, and hybridized with the indicated 32 Plabeled cDNA probes.

DISCUSSION
Hepcidin, a small circulating cysteine-rich peptide, exhibits a broad range of antimicrobial and antifungal activities (8,9). However, several lines of evidence including pattern of expression and regulation by iron overload strongly suggest that hepcidin differs from described antimicrobial peptides (7,8). This hypothesis is reinforced by recent data describing the association of iron-overload phenotype with lack of hepcidin expression in USF2 Ϫ/Ϫ mice (15,16). After submission of this manuscript, Nicolas et al. (17) reported development of phenotypical traits of severe iron deficiency anemia in transgenic mice expressing hepcidin under control of the liver-specific transthyretin promoter. This observation reinforced the idea that this peptide plays an essential role in regulating liver metabolism.
Here we confirm and extend previous data on hepatocytespecific expression of the hepcidin gene and the relation of its expression on hepatocyte differentiation phenotype in humans, mice, and rats. In addition to its predominant hepatic expression, the HEPC gene is likely one of the most highly expressed genes in adult liver. Indeed, as assessed by Northern blot analysis of mice, rat, and human adult liver mRNA, hybridization signal was detected after a very short time of exposure to film (as early as 2-4 h). In addition the real-time RT-PCR analysis revealed HEPC gene expression in human adult liver in a range compatible to that of ␤-actin gene supporting the previously reported analysis of hepcidin expression in human tissues (9). Considering a strong liver-specific pattern of hepcidin expression, activation of its expression by iron and implication in the control of iron metabolism, we analyzed the molecular mechanisms involved in hepcidin gene transcription regulation.
A study of HEPC gene expression during rat liver development, which is an in vivo model of hepatocyte differentiation, revealed very low amounts of HEPC transcripts in fetal liver, and its relatively late accumulation from the 28th day of post-natal development. Interestingly, we found a strong and transient induction of HEPC mRNA shortly after birth. After submission of this manuscript, Nicolas et al. (17) published a report describing expression of hepcidin mRNA during liver development in wild type C57BL/6 mice. In these mice it appeared that hepcidin genes were strongly and transiently induced at birth and at first and second days of postnatal development. Thus, at least in rodents, hepcidin follows the similar pattern of developmental regulation. Previously it has been reported that a number of genes, mainly involved in the regulation of hepatic metabolism, are activated near the time of birth (31). The shift from fetal to neonatal life is manifested by dramatic metabolic changes in the liver, including mobilization of stored glycogen and development of hepatic gluconeogenesis in order to maintain glucose homeostasis during the postnatal period. Thus, an increase of hepcidin shortly after birth is probably a part of metabolic patterning of the liver that occurs during the perinatal period and might have an adaptive role to prevent microbial infection and/or to modulate iron metabolism after birth.
Analysis of sequences of 5Ј-flanking regions of human and mouse HEPC genes and measurement of promoter activities provided important elements that help to understand the phenomenon of predominant expression of this gene in the liver and mechanisms of its regulation during various physiopathological situations. Reporter vectors containing nucleotide fragments of 5Ј-flanking genomic region of human and mouse HEPC genes exhibited strong basal promoter activity in HuH-7 hepatoma cells and primary mouse hepatocytes, respectively. However, activity of both human and mouse HEPC promoters was very low in non-hepatic U-2OS cells suggesting that expression of hepcidin is regulated in the liver primarily at transcriptional level. Sequence analysis of the 5Ј-flanking region of the human and mouse HEPC gene identified several binding sites for liver-enriched transcription factors C/EBP and HNF4 known to be involved in the regulation of liver-specific genes (18,32,33). Interestingly, four putative C/EBP binding sites were also described in 5Ј-flanking region of bass hepcidin gene (11). The C/EBP family of transcription factors includes at least six members, although two of them, C/EBP␣ and C/EBP␤, are enriched in the liver. The important functional role of C/EBP␣ in HEPC gene transcription was confirmed in cotransfection experiments. The promoter activity of both the human and mouse HEPC gene was dramatically increased by C/EBP␣ in U-2OS cells. However, C/EBP␤ isoform had a relatively weak capacity to transactivate HEPC promoters. Characterization of three putative C/EBP binding sites in the human HEPC promoter revealed high DNA binding activity of rat liver nuclear extracts to one site (HEPCϪ250/Ϫ280) and the predominant presence of C/EBP␣ in the bound complex. Together with the results of cotransfection of C/EBP␣ expression vector with various human HEPC promoter-luciferase reporter plasmids demonstrated that the HEPCϪ250/Ϫ280 site represents a functional C/EBP site and is likely the most important for induction of the human HEPC promoter by C/EBP␣.
It is interesting to note that despite the similar binding specificities of C/EBP␣ and C/EBP␤ to CAAT sites, they regulate different genes in liver. C/EBP␣ also specifically controls the retinaldehyde dehydrogenase gene 4 (34). In contrast, the Cyp2d5 gene is controlled by C/EBP␤ and not by C/EBP␣ (35). While the mechanisms governing this specificity between C/EBP␣ and C/EBP␤ is presently unclear, it is interesting to note that replacement of the C/EBP␣ gene with the C/EBP␤ does not change the phenotype of liver (36).
Since C/EBP␣ is highly abundant in adult liver (37), it is likely that this transcription factor is mainly involved in main- FIG. 9. Iron overload leads to an increase of C/EBP␣ protein expression in mouse liver. A, liver samples from mice iron-overloaded by 3% carbonyl iron for 8 months and control animals were used for extraction of total RNA and for preparation of tissue lysates. Expression of C/EBP␣ and hepcidin was studied by Western blot and Northern blot analysis, respectively. B, densitometry analysis. The average for the group of control and iron-overloaded mice is shown in histograms; left panel represents data from Western blot analysis of C/EBP␣ and right panel from Northern blot analysis of hepcidin. The lowest densitometry value from the group of control mice was arbitrarily set as 1. *, p Ͻ 0.01. tenance of the hepatic HEPC gene expression in mature mouse liver. Moreover, among different regulatory mechanisms, liverenriched transcription factors that belong to C/EPB family are likely to largely account for developmental changes in liver metabolism after birth. Indeed, C/EBP␣ appears in the liver during the last trimester of fetal development, reaches the highest levels near the time of birth, decreases during the suckling period, and reaccumulates in adult liver (18). Thus, this transcription regulator is likely to be involved in HEPC gene activation in liver both shortly after birth and at the adult stage. In addition, the previous observation of low concentration of C/EBP␣ in cultured hepatoma cells in comparison with adult hepatocytes (38) is consistent with our data on low or undetectable amounts of HEPC transcripts in rat and human liver-derived cell lines.
Finally, the involvement of C/EBP␣ in the hepatic regulation of the HEPC gene was confirmed in vivo in mice lacking C/EBP␣ expression in the liver. While, as expected, 6-week-old control mice contained high levels of HEPC transcripts, hepatic C/EBP␣-null animals exhibited a strong decrease of HEPC gene expression.
In addition to well documented antimicrobial activity, it has been hypothesized that hepcidin acts as a key regulator of iron homeostasis (15,16). Mice with totally silent HEPC genes in USF2 Ϫ/Ϫ mice exhibited strong iron accumulation in the liver and pancreas. It is worthy noting that although homozygote USF2 Ϫ/Ϫ mice with undetectable expression of HEPC gene transcripts had profound abnormalities of iron metabolism, this phenotype was not observed in heterozygote USF2 ϩ/Ϫ mice containing a reduced amount of hepcidin mRNA in the liver (15). We investigated whether a decrease of hepcidin expression in livers of hepatic C/EBP␣-null mice could result in modification of liver iron content. Histological examination of liver section from 2-month-old C/EBP␣LivKO mice revealed weak but obvious iron staining of periportal hepatocytes. Although we cannot fully exclude that C/EBP␣ is involved in the regulation of other genes encoding proteins related to iron metabolism, correlation of reduced levels of hepcidin with iron overload in liver-specific C/EBP␣-null mice supports the proposed role of hepcidin as a regulator of iron homeostasis.
In contrast to C/EBP␣, HNF4␣ either had no significant effect on the HEPC promoter activity (mouse promoter) or moderately decreased its activity (human promoter). In vivo, expression of HEPC transcripts increased in mice with liverspecific disruption of the HNF4␣ gene, suggesting that this transcription factor is directly or indirectly involved in the negative regulation of HEPC gene transcription.
We have shown previously that expression of hepcidin was strongly up-regulated by iron excess using different models including experimental carbonyl iron and iron-dextran overload (7). However, mechanisms of this regulation remained unknown. Here we found that carbonyl iron overload led to a 2-fold increase of C/EBP␣ protein expression in mice livers. Thus, taking into consideration the role of C/EBP␣ in the regulation of HEPC gene promoter activity and on expression of HEPC transcripts in vivo, it is likely that, at least partly, this transcription factor is involved in an iron-mediated increase of hepcidin expression. Considering that C/EBP␣ is also involved in the regulation of a large set of liver-specific genes including key glyconeogenic enzymes and is essential for energy homeostasis (18,30), this nuclear factor could modulate expression of a large number of target genes under condition of iron excess. In this setting, it will be interesting to analyze if iron-induced glucose metabolism abnormalities observed during genetic hemochromatosis (39) are resulted, at least partly, from an increase of C/EBP␣ expression.
In conclusion, we demonstrated that liver expression of hep-cidin is developmentally regulated and is related to hepatocyte functional status. C/EBP␣ transcription factor is likely to play a key role in regulation of HEPC gene transcription. The correlation between expression of C/EBP␣ and hepcidin and iron accumulation in the liver of mice with hepatic C/EBP␣ disruption and, on the other hand, the increase of both C/EBP␣ and hepcidin during iron overload could represent a novel mechanism for cross-talk between the C/EBP signaling pathway and iron metabolism.