Hepatic Erythropoietin Gene Regulation by GATA-4*

Erythropoietin production switches from fetal liver to adult kidney during development. GATA transcription factors 2 and 3 could be involved in modulating this switch, because they were shown to negatively regulate erythropoietin gene transcription through a promoter proximal GATA site. Herein, we analyzed the role of several GATA factors in the regulation of the erythropoietin gene in human liver and in hepatoma cells. Although GATA-3 expression in hepatocytes increases during human development, erythropoietin mRNA accumulation is unaltered in mutant mice lacking GATA-3. We found that GATA-2, -3, -4, and -6 are all expressed in human hepatocytes and that GATA-4 exhibits the most prominent Epo promoter binding activity in vitro and in vivo. Inhibition of GATA-4 expression by RNA interference leads to a dramatic reduction in Epo gene transcription in Hep3B cells. Moreover, GATA-4 expression is high and limited to hepatocytes in the fetal liver, whereas GATA-4 expression in the adult liver is low and restricted to epithelial cells surrounding the biliary ducts. Thus, GATA-4 is critical for transcription of the Epo gene in hepatocytes and may contribute to the switch in the site of Epo gene expression from the fetal liver to the adult kidney.

The erythropoietin (Epo) 1 gene is expressed in a developmental stage and tissue-specific manner in various organs, including the kidney, liver, bone marrow, CNS, intestine, testis, and uterus (1)(2)(3). An interesting aspect is the developmental switch of the primary Epo production site from the fetal liver to the adult kidney. This switch is characterized by species-specific differences in the time of the onset (4 -11).
Conserved cis-acting elements both 5Ј and 3Ј of the Epo gene are important for its regulation. Three sites in the Epo gene 3Ј enhancer mediate Epo gene expression in response to hypoxia; a binding site for hypoxia-inducible factor-1 (HIF-1), a CACA element, and a direct repeat DR-2 site, which interacts with hepatic nuclear factor-4 (12)(13)(14). Mice transgenic for Epo gene constructs harboring different 5Ј-and 3Ј-flanking regions have revealed tissue-specific regulatory elements in the Epo promoter (15)(16)(17)(18). The combined data indicate that Epo gene transcription in the liver is mediated predominantly by cis-regulatory DNA elements in the 5Ј-flanking region, including the minimal promoter. Moreover, Epo transcription is negatively regulated by a subset of GATA transcription factors (GATA-1, -2, and -3), which bind to a GATA site within the minimal promoter (19 -22).
GATA transcription factors belong to the family of zinc finger DNA-binding proteins and play critical roles in cell growth and differentiation (23,24). The six members of the GATA family have been subdivided into two groups based on sequence homology and expression pattern (25). Whereas GATA-1, -2, and -3 were originally characterized as a subgroup essential for the regulation of hematopoiesis, a broader range of functions became obvious after examining the phenotypes of mice with targeted deletions of these factors (26 -29). GATA-1 is expressed in hematopoietic cells and in Sertoli cells of the murine testis (30). GATA-2 is expressed in various tissues, including liver, kidney, and the CNS (31)(32)(33)(34). GATA-3 is also expressed in a variety of non-hematopoietic cells during murine development, most abundantly in the developing central and peripheral nervous system, otic and optic vesicles, liver, kidney, adrenal gland, thymus, and endothelial cells (32, 34 -37).
The second subgroup of GATA factors (GATA-4, -5, and -6) are predominantly expressed in the heart, gut, and extraembryonic endoderm (38 -41). Down-regulation of GATA-4 and -6 in cardiomyocytes results in a reduced expression of several genes, indicating that GATA-4 and -6 act as tissue-specific transcriptional activators (42). GATA-4 and -6 are expressed in the liver during murine embryogenesis (43). During gestation, GATA-6 is clearly expressed in hepatocytes, whereas expression of GATA-4 is detectable late in gestation only in endothelial/epithelial cells surrounding the hepatic vessels. A recent study demonstrated that GATA-4 cooperates with HNF3 (Fox3A) to stimulate albumin gene transcription in liver progenitor cells (44). Based on these observations, we hypothesized that developmental changes in GATA-4 and -6 gene expression occur during human development and that these proteins are involved in the tissue-and developmental stagespecific regulation of Epo gene expression.
In this study, we analyzed the role of several GATA factors in the liver-specific regulation of the Epo gene. The summarized data indicate that GATA-4 is important for hepatic Epo gene transcription during a particular period of development.

EXPERIMENTAL PROCEDURES
Human Tissue Specimens-Tissue biopsy specimens from the liver, and in single cases also from kidneys and the CNS, were collected from 24 individuals at post mortem examination at the Institute of Pathology after written parental consent was obtained. Approval for this study was given by the Institutional Review Board. The study group included 11 fetuses (gestational age (GA) Ͻ 24 weeks post conception (pc)) and six preterm or term neonates with normal organ development, which were spontaneously or electively delivered due to perinatal complications (premature rupture of membranes, amnion infection, or placental/ umbilical complications). To broaden the spectrum of developmental times examined, we included five neonates with congenital malformations who revealed normal medullary and extramedullary hematopoiesis as well as normal development of the liver and kidneys. The diagnoses of these patients were malformations of the CNS (n ϭ 3), vitium cordis, or congenital asplenia. We also included two infants with sudden infant death syndrome. In the 24 individuals, gestational GA ranged from 8 to 67 weeks pc. The body weight ranged from 118 to 5627 g (median 656 g) and the crown-rump length from 18.1 to 64.5 cm (median 32.0 cm); these data were not evaluated in subjects with an age below 16 weeks pc. Technical procedures to collect the tissue biopsy specimens from liver (both right and left lobe) were performed as previously described (4). In addition to the patients' samples, we obtained RNA (liver, kidney, spinal cord, and CNS) from a commercially available panel, which included a liver sample from a 27-year-old Asian male (Clontech Laboratories Inc., Palo Alto, CA).
Immunohistochemistry-6-m-thick slices of formalin-fixed, paraffin-embedded tissue specimens were baked at 80°C for 2 h, deparaffinized, and rehydrated. Endogenous peroxidase was blocked with 3% H 2 O 2 for 10 min. Immunohistochemical staining for GATA-4 was performed as described by Laitinen et al. (50), using the Vector Elite ABC kit (Vector Laboratories). The primary antibody (anti GATA-4, C-20, 1:50 dilution, Santa Cruz Biotechnology Inc.) was a polyclonal antibody raised against a peptide derived from the carboxyl terminus of the human GATA-4. Negative controls were prepared for each sample by following the complete staining procedure in the absence of the primary antibody.
Electrophoretic Mobility Shift Assay-Whole cell extracts from Hep3B cells grown under standard conditions were prepared for EMSA as previously described (51). 100 fmol of an end-labeled 30-bp doublestranded oligonucleotide corresponding to the GATA-binding site at Ϫ30 of the Epo transcription start site was incubated with 5 g of Hep3B extract in 1ϫ binding buffer (13,20), 1 g of bovine serum albumin, and 0.5 g of poly(dI-dC). In competition experiments, a 300-fold excess of cold wild-type or mutant competitor was added, and the reaction was incubated for 20 min at 30°C. After addition of the radiolabeled probe, the binding reaction was allowed to continue for an additional 45 min. Where indicated, reactions were postincubated for 20 min with the following antibodies (all from Santa Cruz Biotechnology): GATA-2 (CG2-96), GATA-3 (HG3-31), GATA-4 (C-20), and GATA-6 (C-20). Immediately after incubation, the samples were loaded onto a 4.5% polyacrylamide gel in 0.5ϫ TBE buffer (45 mM Tris base, 45 mM Boric Acid, 1 mM EDTA, pH 8.0). Electrophoresis was performed at 34 mA for ϳ2.5 h at 4°C. The gel was dried, and complex formation was visualized by autoradiography. The following double-stranded oligonucleotide was used in the experiments (only the non-coding strand is shown): GATA Ϫ30 oligonucleotide Epo promoter 5Ј-CATGCAGATAA-CAGCCCCGACCC CCGGCCA-3Ј.
Western Blot Analysis-Protein extracts from Hep3B or HepG2 cells were prepared and analyzed by Western blot experiments as described by Leach et al. (51,53), using specific antibodies for the GATA transcription factors as listed above as well as antibodies specific for HIF-1␣ (mouse monoclonal, BD Transduction Laboratories) and ␣-tubulin (mouse monoclonal, TU-02, Santa Cruz Biotechnology).
RNA Interference-The plasmid pSilencer TM 1.0-U6 (Ambion, Inc.) was used to express GATA-4 interfering RNAs in Hep3B cells. The RNA-expressing oligonucleotides were ligated downstream of a human U6 promoter as described previously (54). Two different 19-bp oligonucleotides corresponding to sequences immediately downstream of AA dinucleotides in the 5Ј region of the GATA-4 coding sequence (Gen-Bank TM , accession number L34357) were ligated into pSilencer. Sequences were compared with the human genome data base to exclude significant homology to other genes. Two DNA oligonucleotides were generated for each construct. In the forward oligonucleotide, the 19nucleotide sense siRNA sequence was linked to the reverse complementary antisense strand by a 9-nucleotide spacer. Five Ts were added to the 3Ј-end of the oligonucleotides, and 4-nucleotide overhangs for EcoRI and ApaI restriction sites were added to the 5Ј and 3Ј ends of the complementary sequence, respectively. The oligonucleotides have the following sequences: Construct 1: GATA-4/1 siRNA 852-870 sense, 5Ј-T-CTCGATATGTTTGACGACTTCAAGAGAGTCGTCAAACATATCGAG-ATTTTTT-3Ј; antisense, 5Ј-AATTAAAAAATCTCGATATGTTTGACG-ACTCTCTTGAAGTCGTCAAACATATCGAGAGGCC-3Ј. Construct 2: GATA-4/2 siRNA 897-915 sense, 5Ј-CTGTGGGGCTATGTCCACCTTCA-AGAGAGGTGGACATAGCCCCACAGTTTTTT-3Ј; antisense, 5Ј-AATTAAAAAACTGTGGGGCTATGTCCACCTCTCTTGAAGGTGG-ACATAGCCCCACAGGGCC-3Ј.
The forward and reverse oligonucleotides were annealed and ligated into the EcoRI and ApaI sites in pSilencer. Hep3B cells were freshly passaged overnight at a concentration of 5 ϫ 10 4 /ml to 1 ϫ 10 5 /ml in 6-well plates under standard conditions. 15 min prior to transfections, medium was changed to quiescent medium (no fetal bovine serum). Transfections were carried out using FuGENE 6 transfection reagent (Roche Applied Science) according to the manufacturer's instructions. Highest efficiency of transfection was reached using 6 l of FuGENE transfection reagent and 1 g of DNA per 9.4 cm 2 well (2-ml medium). 4 h after transfection, 2% fetal bovine serum and 1% antibiotics (penicillin/streptomycin) were added. The reduction of GATA-4 expression by siRNA was analyzed by semiquantitative RT-PCR and Western blotting analysis. Experiments were repeated at least twice and performed for each set in duplicate or triplicate.

Role of GATA-3 in Epo Gene
Regulation during Development of the Liver-Previous studies have shown that GATA-3 negatively regulates Epo gene transcription through a promoter proximal GATA-site in vitro (23). To investigate whether GATA-3 is involved in regulating the switch of Epo gene transcription from the fetal liver to the adult kidney, we examined GATA-3 expression levels in human liver specimens taken at different developmental stages. Immunohistochemical analysis revealed that GATA-3 is expressed in early hematopoietic progenitor cells and in hepatocytes (not shown). In contrast, GATA-1 is expressed only in hematopoietic cells and thus represents an indicator of hematopoietic activity in the fetal liver as previously shown (47). In the semiquantitative RT-PCR analysis of GATA-3 expression, we have grouped the samples according to well defined stages of hematopoietic activity in the liver. As shown in Fig. 1 (A and B), GATA-1 expression declines in liver specimens after mid-gestation, while expression of GATA-3 increases. The reciprocal expression pattern of GATA-3 and Epo during development in the liver (Fig. 1A and Ref. 4), led us to directly examine the role of GATA-3 in regulating the switch from liver to renal Epo gene expression in fetal mice deficient for GATA-3 (Fig. 1, C and D). Hepatic Epo expression decreased significantly between days E14.5 pc and E18.5 pc, indicating that the switch from liver to renal Epo synthesis in mice occurs before birth. More importantly, Epo gene expression is not affected by the absence of GATA-3.
Expression Pattern and Epo Promoter Binding Activity of GATA Transcription Factors in Human Hepatoma (Hep3B) Cells-The above observations and unpublished data showing that expression of GATA-2 in human hepatocytes does not significantly change during development, 2 led us to investigate the expression pattern of all six GATA transcription factors in Hep3B cells. Both RT-PCR ( Fig. 2A) and Western blot analysis (data not shown) indicate that Hep3B cells express GATA-2, -3, -4, and -6, but not GATA-1 and GATA-5.
To examine whether GATA-2, -3, -4, and -6 directly interact with GATA binding sites in the Epo promoter, we performed EMSA with protein extracts prepared from Hep3B cells (Fig.  2). There are two GATA binding sites in the Epo upstream promoter region, one at Ϫ30 and the other at Ϫ267, relative to the transcription initiation site (55). Previous in vitro binding analysis of the Epo promoter region focused on the GATA binding site at Ϫ30 and was performed in cells in which GATA-1, -2, or -3 were overexpressed (20,21). Incubation of the oligonucleotide containing the Ϫ30 GATA site with whole cell extracts prepared from Hep3B cells leads to the formation of several protein-DNA complexes (Fig. 2B). While all of these complexes diminished significantly in intensity after preincubation with excess wild-type competitor oligonucleotides, three distinct protein-DNA complexes remained after preincubation with a mutant oligonucleotide (indicated by a single arrowhead, a double arrowhead, and a filled circle) implying that formation of these protein-DNA complexes require a GATA site. We used antibodies specific for GATA-2 and -3 to test whether these proteins bind to the Ϫ30 GATA site. Mutant competitor was added to the reactions to reduce background. Postincubation with GATA-2 or GATA-3 antibodies did not affect the formation the protein-DNA complexes (Fig. 2B, lanes  5 and 6). One GATA-specific complex (single filled circle) is reduced by postincubation with antibodies against GATA-4, and formation of another protein-DNA complex (single arrowhead) is inhibited by postincubation with GATA-4-and GATA-6-specific antibodies.
To address the question which of the GATA factors interact with the Epo gene in hepatocytes in vivo, we performed ChIP analysis in Hep3B cells (Fig. 3). In these experiments, antibodies specific for GATA-2, -3, -4, and -6, RNA polymerase II, the co-activator CBP, and the transcription factors Sp1 and NF-B (p65 subunit) were included. The NF-B antibody was not expected to precipitate the Epo promoter under normal conditions and thus served as a negative control in this experiment. There are several potential Sp1 binding sites in the Epo promoter, and SP1 is known to interact with GATA factors (56). The positions of GATA and NF-B binding sites relative to the Epo transcription start site are shown in Fig. 3A. As a control we analyzed the binding of all these factors to the human GAPDH gene. The data indicate that all of the GATA factors, polymerase II, and CBP can be cross-linked to the Epo promoter region in vivo (Fig. 3C). GATA-4 appears to exhibit greater binding activity than the other GATA factors. As expected, NF-B did not bind to the Epo gene under normal conditions. In contrast to the Epo gene promoter, only polymerase II cross-linked to the control GAPDH gene, indicating that the interactions of GATA-2, -3, -4, -6, and CBP are specific for the Epo promoter (Fig. 3C).
Consequence of GATA-4 RNA Interference on Epo Gene Expression in Hep3B Cells-We used RNA interference to reduce expression of GATA-4 in Hep3B cells and then analyzed the consequence on Epo gene expression by RT-PCR (Fig. 4). We generated two constructs expressing two independent GATA-4-interfering RNAs (GATA-4/1 and GATA-4/2). Hep3B cells were transiently transfected with the two constructs, and, 48 h after transfection, RNA and proteins were harvested and analyzed by RT-PCR or Western blot experiments. The results, shown in Fig. 4, demonstrate that GATA-4 expression was significantly reduced by RNA interference using one of the constructs (GATA-4/1) in transfected compared with the control cells (Fig. 4B). More importantly, Epo mRNA was not detectable in the transfected cells but was expressed in the control cells. The results in Fig. 4A demonstrate that GATA-4 protein levels were also significantly reduced in the transfected cells. The fact that GAPDH was expressed at similar levels in the transfected and non-transfected cells demonstrates that the reduction of GATA-4 levels by RNA interference was specific. The second construct (GATA-4/2) did not lead to significant inhibition of GATA-4 expression (Fig. 4B). It thus represents a control showing that transfection with the plasmid does not cause unspecific alterations in gene expression in these cells. We also analyzed the expression of other GATA factors in transfected and untransfected Hep3B cells (Fig. 4C). The results show that reducing the levels of GATA-4 does not cause drastic alterations in the expression of GATA-2, -3, or -6 in these cells.
Expression of GATA-4 and -6 in Fetal and Adult Human Tissues-RT-PCR for GATA-4 and -6 in human tissues (liver, kidney, brain, and spinal cord) demonstrated that GATA-4 is highly expressed in the fetal liver (Fig. 5A). Expression of GATA-4 in the fetal or adult kidney and CNS, two other organs expressing the Epo gene, is extremely low or absent.
We next analyzed the expression of GATA-4 in human fetal and adult liver specimens by immunohistochemistry (Fig. 5, B  and C). GATA-4 is expressed in different cell types during development. Early in development, GATA-4 is expressed in hepatocytes and endothelial cells. Some of these cells show morphological characteristics of oval cells/hepatic progenitor cells, which are known to express the Epo gene (Fig. 5B). In the adult liver GATA-4 staining is restricted to epithelial cells of the biliary duct (Fig. 5C).

FIG. 1. Role of GATA-3 in Epo gene regulation during liver development.
A, RT-PCR analysis of GATA-1, -3, and GAPDH expression in human liver tissue specimens during gestation. The [␣-32 P]dCTP-labeled PCR products were electrophoresed in 6% non-denaturing polyacrylamide gel. B, semiquantitative RT-PCR analysis of GATA-3 expression (per GAPDH) in the human liver specimens. For statistical analysis, the samples were grouped according to major developmental stages of hepatic hematopoiesis: GA Ͻ 20 weeks pc: primary hepatic hematopoiesis; GA Ն 20 to Ͻ30 weeks pc: peak of hematopoietic activity in the liver; GA Ն 30 to Ͻ40 weeks pc: decrease of hepatic hematopoiesis; GA Ն 40 weeks pc: no relevant contribution of the liver to hematopoiesis. C, RT-PCR products of murine GAPDH, Epo, and GATA-3 mRNA were [␣-32 P]dCTP-labeled and electrophoresed in 6% non-denaturing polyacrylamide gel. Samples were taken from breedings of mice with heterozygous deletions for GATA-3 at days E14.5 and E18.5 of gestation. The genotype of the animals is indicated as follows: homozygous for GATA-3 deletion: Ϫ/Ϫ; heterozygous for GATA-3 deletion: Ϫ/ϩ; wild-type: ϩ/ϩ. D, semiquantitative RT-PCR of Epo mRNA expression per GAPDH mRNA. Data are shown as mean Ϯ S.D.

Effects of Hypoxia and Inflammation on Expression of GATA-4 in HepG2
Cells-Because expression of Epo is activated by hypoxia and repressed by inflammation, we were interested to analyze whether these conditions would also alter the expression of GATA-4 in hepatic cells. In these experiments we used HepG2 instead of Hep3B cells. HepG2, like Hep3B, is a hepatoma cell line, which also represents an established model for analyzing Epo gene regulation. We were thus interested in examining expression of GATA-4 in these cells as well. As shown in Fig. 6, GATA-4 is expressed in HepG2 cells, and in contrast to HIF1-␣, its expression is not altered by hypoxia. There is also no significant change in GATA-4 expression in response to well characterized mediators of inflammation.

DISCUSSION
Expression of Epo is controlled at the transcriptional level by cis-acting DNA elements located within the 5Ј promoter and the Epo 3Ј enhancer (1). One key aspect in the tissue-specific regulation of the Epo gene is the switch of the Epo production site from the fetal liver to the mature kidney. Data from organ ablation experiments in sheep strongly suggest that this process is transcriptionally regulated (57). Recent observations that GATA-2 and GATA-3 bind to the Epo promoter and repress Epo gene expression in transfection studies suggest that these two regulatory proteins could be involved in controlling Epo gene expression in the liver and perhaps also in the kidneys or other Epo-producing organs.
It is shown here that hepatic expression of GATA-3 increases during development (Fig. 1). In addition, Imagawa et al. (20,21) showed that GATA-3 acts as a negative regulator of Epo gene transcription in cell culture experiments. Thus GATA-3 might be involved in the switch of the primary Epo production site from the fetal liver to the adult kidney. However, our data obtained from GATA-3-deficient mice provide evidence that GATA-3 is not critical for the developmental regulation of Epo gene expression in the murine liver. These results can be interpreted in different ways. It is possible that the Epo gene is regulated by different GATA factors in mice versus human, which could also explain species-specific differences in the onset and kinetics of the switch (4,8,10). Alternatively, it is conceivable that an as yet unknown activity compensates for the loss of GATA-3.
To examine whether other GATA factors participate in the regulation of Epo gene expression during liver development, we analyzed the expression of all known GATA factors in Hep3B cells. Hep3B and HepG2 are hepatoma cell lines with similarity to fetal hepatocytes and are generally accepted as good models for hepatic Epo gene regulation (48). RT-PCR analysis and Western blot experiments reveal that GATA-2, -3, -4, and -6 are expressed in Hep3B cells. In addition, the ChIP experiments demonstrate that all of these proteins interact with the Epo gene in vivo. These data are partially consistent with previous results from Imagawa et al. (20,21), who demonstrated binding of GATA-2 and -3 to the Ϫ30 GATA site in the Epo promoter using protein extracts recovered from cells in which GATA-2 or -3 were overexpressed. However, both EMSA and ChIP analyses suggest that GATA-4 directly interacts with the GATA sites in the Epo promoter region and exhibits the most prominent Epo promoter binding activity among the GATA factors. The stronger signal consistently observed for GATA-4 in the ChIP assay could indicate that most of the cells have GATA-4 bound at the promoter, whereas GATA-2, -3, and -6 only bind to the Epo promoter in a sub-population of cells. This interpretation, however, should be treated with caution, because it is possible that the epitopes for GATA-2, -3, and -6 may not be as accessible in the ChIP assay as the one for GATA-4. A, expression analysis of GATA factors in Hep3B cells. RT-PCR analysis was carried out as described under "Experimental Procedures" using primers specific for GATA-1, -2, -3, -4, -5, and -6. The PCR products were electrophoresed in 2% agarose gels and stained with ethidium bromide. B, EMSA was carried out as described under "Experimental Procedures" using an oligonucleotide harboring the GATA binding-site located 30 bp upstream of the Epo transcription initiation site (Ϫ30, lanes [1][2][3][4][5][6][7][8]. Protein extracts were preincubated with cold competitor DNA or postincubated with specific antibodies as indicated. Protein-DNA complexes were resolved on 4.5% polyacrylamide gels. The results from the GATA-4 RNA interference experiments strongly suggest that GATA-4 plays a critical role in Epo gene regulation in Hep3B cells. This activity can not be compensated by other GATA factors present in Hep3B cells, including GATA-6. A recent study by Cirillo et al. (44) demonstrates that GATA-4 cooperates with hepatic nuclear factor-3 to modify the chromatin structure over the albumin gene enhancer. Other studies also implicate GATA factors as critical components of chromatin modification mechanisms leading to transcriptional activation (58). The ChIP analysis presented here provides additional information on the transcriptional regulation of the Epo gene: the interaction of RNA polymerase II with the Epo promoter indicates that the chromatin is open and accessible to transcription factors under conditions of normoxia. In this respect, it is interesting to note that CBP, a co-activator with histone acetyl transferase activity (59, 60), can be cross-linked to the Epo promoter region. CBP is known to associate with components of the protein-DNA complex at the Epo 3Ј enhancer (61).
The functional data together with the expression analysis of GATA-4 in the fetal and adult human liver suggests that GATA-4 participates in the switch of the primary Epo production site from fetal liver to adult kidney. The expression of GATA-4 in hepatocytes parallels that of Epo and is restricted to early developmental stages. At later stages, GATA-4 is detectable only in epithelial cells surrounding the biliary ducts. These results are consistent with the expression pattern of GATA-4 in the murine liver (62). The summary data thus strongly suggest that GATA-4 is critical for Epo gene expression in the fetal liver and that the absence of GATA-4 expression in adult hepatocytes contributes to the repression of Epo gene transcription.
GATA-4 does not appear to be involved in hypoxia-mediated up-regulation or inflammation-mediated down-regulation of Epo gene expression (Fig. 6), processes that are mediated by HIF-1␣ and NFB, respectively (22,61). Instead, we propose that GATA-4 binds to the Epo promoter in a tissue-specific manner and interacts with chromatin-modifying activities, like CBP, to render the Epo gene promoter accessible to transcription factors and RNA polymerase II, very much like the documented role of GATA-4 in opening chromatin over the albumin enhancer (44). Numerous studies have implicated GATA factors in the modification of chromatin structure. For example GATA-1, known to interact with CBP (63), appears to be required for the acetylation of nucleosomes over the ␤-globin gene locus (64). In contrast, GATA-2 was shown to interact with histone deacetylases thereby repressing transcription of genes (65). We thus suggest that GATA-4 binds to the Epo promoter in fetal hepatocytes and recruits chromatin-modifying activities ultimately leading to expression of the gene. At the adult stage, the lack of GATA-4 expression, perhaps in combination with the interaction of GATA-2 or-3 with the Epo promoter, may lead to the formation of a repressive chromatin structure and inactivation of Epo gene transcription. Clearly, the function of GATA-4 is tissue-specific, and other activities, probably other GATA factors, will mediate activation of the Epo gene in tissues in which GATA-4 is not expressed, e.g. kidney.