Specific Contribution of the Erythropoietin Gene 3′ Enhancer to Hepatic Erythropoiesis after Late Embryonic Stages

Abstract

Erythropoietin (Epo) is secreted from the liver and kidney, where Epo production is strictly regulated at the transcriptional level in a hypoxia- and/or anemia-inducible manner. Here, we examined the in vivo function of the enhancer located 3′ to the Epo gene (EpoE-3′). Reporter transgenic-mouse analyses revealed that the EpoE-3′ enhancer is necessary and sufficient for the liver-specific and hypoxia-responsive expression of the gene after embryonic day 14.5 (E14.5). However, the enhancer is dispensable for Epo gene expression in the kidney and early-stage embryonic liver. Genetic removal of EpoE-3′ from the endogenous Epo gene resulted in mice with severe anemia at late embryonic and neonatal stages due to defects in hepatic erythropoiesis, but early hepatic and splenic erythropoiesis was not affected. The mutant mice recover from the anemia in the juvenile period when major Epo production switches from the liver to the kidney. These results demonstrate that EpoE-3′ is necessary for late hepatic erythropoiesis by specifically supporting paracrine production of Epo in the liver. In contrast, Epo production in the kidney utilizes distinct regulatory machinery and supports erythropoiesis in the bone marrow and spleen in adult animals.

INTRODUCTION

During mammalian development, erythropoiesis moves from the yolk sac to the fetal liver and, finally, to the bone marrow and spleen (8, 26). As yolk sac erythropoiesis is unique and easily distinguishable from fetal-liver and adult erythropoiesis (27), it is often referred to as primitive erythropoiesis, while the latter is termed definitive erythropoiesis. Primitive and definitive erythroid cells show distinct dependence on erythropoietin (Epo), a growth factor that specifically sustains growth and survival of cells of the erythroid lineage (12). Although the number of primitive erythrocytes is partially decreased in Epo gene knockout mice, they differentiate and support primitive erythropoiesis so that the Epo-deficient embryos live and develop to embryonic day 13.5 (E13.5) (48). In contrast, definitive erythroid progenitors cannot differentiate without Epo (48). In the embryo and fetus, the liver is the major Epo-producing tissue, but in adult animals, Epo is produced mainly in the kidney (3, 13, 49). Therefore, renal failure often causes Epo-dependent anemia (12). It has been shown that erythroid homeostasis in adult animals is strictly maintained through fine-tuning of Epo production (16).

Transcription of the Epo gene is the major regulatory point for the maintenance of erythroid homeostasis (38). Epo gene transcription is regulated in a tissue-specific manner and also in response to tissue hypoxia, which is mainly caused by anemia (5, 16, 39). The molecular mechanism involved in hypoxia-inducible Epo gene expression has served as a prototype for study of oxygen-dependent transcriptional regulation. The discovery of two human hepatoma cell lines, Hep3B and HepG2, which retain a unique ability to express Epo in response to hypoxia (9), markedly facilitated this study. Utilizing these two cell lines, a hypoxia-responsive element (HRE) located in the 3′ flanking region of the Epo gene (35) and a hypoxia-inducible factor (HIF) were identified (46). Since then, many hypoxia-responsive genes have been found to be regulated by the HIF family of transcription factors acting through HREs (17, 33).

Whereas many important findings were accumulated by extensive analyses of Epo gene transcriptional regulation using these cell lines and conventional transgenic-mouse systems (5, 19, 34, 36), these studies have a limitation, especially with regard to tissue-specific Epo gene expression. It is now important to define the mechanisms operating in vivo to regulate Epo gene activity. With this in mind, we utilized a bacterial artificial chromosome (BAC)-based transgenic-mouse system (24, 39). We found that a single nucleotide substitution in the promoter GATA box resulted in unexpected ectopic Epo gene expression in several epithelial cell lineages, which suggests a novel aspect of Epo gene regulation. The study also revealed that the Epo gene regulatory mechanisms clarified in vivo share many fundamental similarities to, but also demonstrate substantial differences from, those reported in vitro.

The Epo gene enhancer located 3′ to the gene (EpoE-3′) has been reported to contain two important cis-acting elements, HRE and a direct repeat (DR2) motif for nuclear receptor binding (2, 5). Expression of a 4-kb fragment of the human Epo gene containing all the exons and EpoE-3′ was investigated in a transgenic-mouse system and revealed hypoxia-inducible transgene expression in the liver, as well as constitutive and ectopic expression in the other tissues (34, 36). While these reports suggested that EpoE-3′ might be the hypoxia-responsive Epo gene enhancer in the liver, it was difficult to detail the physiological and regulatory functions of EpoE-3′ due to polycythemia caused by the transgene overexpression (36). Although critical contributions of this enhancer to the hypoxia response of the Epo gene have been intensively studied (2, 29, 35), many unsolved questions still remain. For instance, it is not known how widely this enhancer is utilized or how it interacts with other tissue-specific elements within the Epo gene.

Therefore, in this study we addressed the function of the EpoE-3′ enhancer by adopting a BAC-based transgenic-mouse assay. We created a BAC-based Epo-GFP reporter transgene and multiple mouse lines harboring the transgene. Analyses of the reporter mice revealed that EpoE-3′ is essential for green fluorescent protein (GFP) expression in hepatocytes after E14.5 but that EpoE-3′ is not required for GFP expression in the kidney and early-stage hepatocytes. In terms of Epo gene regulation, hepatocytes are developmentally divided into two types of cells, the earlier Hep-E (E9.5 to E14.5) and later Hep-L (E14.5 and after). To examine EpoE-3′ function in a physiological context, we also prepared mice with a germ line deletion of EpoE-3′. Surprisingly, homozygous EpoE-3′ deletion mice are born alive and with normal adult hematopoiesis and fertility, although the mice show a marked decrease in Epo production in Hep-L and severe neonatal anemia, indicating that EpoE-3′ is a Hep-L-specific regulatory element for Epo gene expression and is transiently essential for late embryonic erythropoiesis. In the homozygous EpoE-3′ deletion mice, hepatic erythropoiesis was significantly enfeebled, and transient anemia was provoked from the late embryonic to neonatal stages, even though renal Epo expression levels were preserved or rather higher than those of wild-type mice. Our results demonstrate that paracrine-secreted Epo supports erythropoiesis in the liver, while endocrine-secreted Epo from the kidney fully supports erythropoiesis in the bone marrow and spleen in adult stage mice.

MATERIALS AND METHODS

GFP reporter transgenic mice.

The 180-kb BAC transgene (wt-Epo-GFP) containing a GFP reporter and the regulatory region of the Epo gene was described previously (24, 39). In this study, an m3-Epo-GFP transgene with a mutation (underlined) in the 3′ HRE (ACGTG to AAAAG) was constructed by using a standard procedure (46). To construct the BS-HSP-GFP transgene, the BS region (see below) was ligated to a 0.8-kb mouse heat shock protein 68 promoter (15) and enhanced GFP (EGFP) cDNA (BD Biosciences). Transgenes were injected into fertilized eggs from BDF1 mice (Clea, Japan), and transgenic-mouse lines were established. The GFP transgene in the mouse genome was detected by PCR as described previously (41). All mice were treated according to the regulations of the standards for human care and use of laboratory animals of the University of Tsukuba.

Targeted deletion of the BS region from the mouse Epo gene.

To delete the EpoE-3′ enhancer region of the Epo gene, a targeting vector was constructed by internal ribosome entry site (IRES)-GFP (BD Biosciences), MC1-Neo (22), and MC1-DT3 cassettes (42); 129SV mouse-derived genomic fragments (11); and loxP sequences (22). Two loxP sequences were integrated on both sides of the BS region. Through homologous recombination and positive-negative selection, 22 mutant embryonic stem (ES) cell lines (Epo^ƒn/w genotype, where f is floxed BS region, n is Neo, and w is wild type) were obtained, and 2 independent germ line chimeric mice were established from these ES cell lines. The Neo cassette was deleted by injection of a Cre-expressing plasmid into fertilized eggs from wild-type female and Epo^ƒn/ƒn male mice (1). After establishment of mice with Neo deleted (Epo^ƒ/w), the BS region was excised by mating them with Ayu1-Cre transgenic mice that express Cre recombinase in their germ line, as described previously (22). PCR was used for mouse genotyping. All primer sequences used in this study are available upon request. Southern blot analysis of XbaI-digested genomic DNA was also performed for genotyping using a ³²P-labeled probe.

RT-PCR and RNA blot analyses.

RNA samples were isolated from tissues using Isogen (Nippon-Gene). For quantitative reverse transcription (RT)-PCR analysis, total RNA was reverse transcribed with Superscript II (Promega), and Epo mRNA levels were measured with an ABI Prism 7700 (Perkin-Elmer) (24). GAPDH (glyceraldehyde-3-phosphate dehydrogenase) was used as an internal standard (41). To detect the BAC transgene expression, cDNA samples were put through PCR (see Fig. 1A); primer sequences are available on request. For RNA blot analysis, 10 μg of total RNA was used for electrophoresis and hybridization with a ³²P-labeled probe.

Fig. 1.

Transgenic reporter assay of the EpoE-3′ region. (A) Structures of GFP reporter transgenes. The m3-Epo-GFP transgene construct has a mutation in the 3′ HRE (black dots) of the wt-Epo-GFP transgene containing 180 kb (60 kb upstream and 120 kb downstream) of the Epo gene flanking region. The BS-HSP-GFP transgene consists of the BS region, the human HSP68 gene promoter, and a GFP cassette. pA, polyadenylation signal sequence. The open and shaded boxes indicate untranslated and translated regions, respectively. The numbers of transgenic lines positive for GFP expression in embryonic hepatocytes at E11.5 and E15.5 are shown. The numbers of transgenic lines expressing GFP in hepatocytes and REP cells of adult (6 to 8 weeks of age) mice under anemic conditions are also shown. The expression profiles of the wt-Epo-GFP transgene were similar in 3 transgene-expressing lines, while 1 line did not express GFP in any tissues. Similarly, the expression profiles of the GFP reporter were identical in 3 m3-Epo-GFP transgene-expressing lines, but 2 lines did not express GFP. GFP expression was observed in 3 BS-HSP-GFP transgenic-mouse lines out of 6 transgene-positive mouse lines. The expression profiles of GFP in these 3 lines were close to each other. Tg, transgenic lines analyzed. (B) Anti-GFP immunohistochemistry of the adult liver under anemic conditions. GFP expression (brown) was detected in hepatocytes around the central vein (#) of wt-Epo-GFP (line wt-A) and BS-HSP-GFP (line BS1) transgenic mice but was not detected in m3-Epo-GFP (line m3-1) transgenic mice. Scale bar, 200 μm. (C) RT-PCR analysis of transgene expression in the liver in early-stage embryos (E11.5; Hep-E) and neonates (P0; Hep-L). The expression of wt-Epo-GFP (wt) and m3-Epo-GFP (m3) transgenes was detected by specific primers for the GFP transgene shown in panel A. HPRT was used as an internal control. −, not present.

Histological analyses.

Tissues were fixed in 4% paraformaldehyde for 1 h and embedded in OCT compound (Sakura Fine Technical) in liquid nitrogen. Sections (8-μm thickness) were incubated with rabbit anti-GFP polyclonal antibody (40) or anti-β-globin antibody (23) at 4°C overnight. After treatment with hydrogen peroxide, the sections were incubated with a horseradish peroxidase-conjugated anti-rabbit IgG secondary antibody (Biosource) at room temperature for 2 h. Color detection was performed using diaminobenzidine as a chromogen (brown staining). Hematoxylin was used for counterstaining. A terminal deoxynucleotidyltransferase-mediated dUTP-biotin nick end labeling (TUNEL) assay was performed using an in situ apoptosis detection kit (Takara) (40).

Induction of anemia and measurement of plasma Epo concentration.

To expose mice to hypoxic conditions by bleeding anemia, 6- to 8-week-old mice were bled by taking 0.4 ml of blood from the retro-orbital plexus at 48, 36, 24, 12, and 6 h before analysis (40). Blood samples from neonate mice were obtained from carotid arteries. The plasma Epo concentration was measured using a photometric enzyme-linked immunosorbent assay (ELISA) kit (Roche).

Chromatin immunoprecipitation (ChIP) assay.

The fresh, minced liver (10 mg) was cross-linked by using formaldehyde and sonicated for 180 s on ice with a 2-mm sonicator tip (Branson). The chromatin complex was purified and precipitated with anti-acetylated histone H3 and H4 (AcH3 and AcH4) antibodies included in the ChIP assay kit (Upstate Biotechnology). Quantitative PCR was performed with primers (available on request) to detect the precipitated promoter sequence of the Epo gene with an ABI Prism 7300 (Perkin-Elmer) with SYBR green reagent (Nippon-gene). The promoter region of the α-catenin gene was used as an internal positive control (44).

Analyses of hematopoietic cells.

Single-cell suspensions from embryonic livers and spleens were cultured in methylcellulose medium (StemCell Technologies) supplemented with 100 ng/ml stem cell factor (SCF; R&D Systems) and 4 U/ml Epo (Chugai Pharmaceutical) for 3 days (erythroid CFU [CFU-E]) or 7 days (erythroid burst-forming unit [BFU-E]), and benzidine-positive erythroid colonies were counted (40). For flow cytometry analysis, cells were stained with phycoerythrin (PE)-conjugated anti-Ter119 and allophycocyanin (APC)-conjugated streptavidin after incubation with biotin-conjugated anti-CD71. Cells incubated with antibodies were stained with propidium iodide to gate out dead cells and analyzed by FACSCalibur with the CellQuest program (Becton Dickinson). All antibodies were purchased from BD Biosciences (41). An aliquot of the collected blood sample was used for determining the hematopoietic indices (hematocrit [Hct], red blood cell number [RBC], hemoglobin concentration [Hb], and platelet number [Plt]) using an automatic counter (Nihon-Kouden) (40).

RESULTS

EpoE-3′ is essential for gene expression in the liver but not in the kidney.

The enhancer located 3′ to the mouse and human Epo genes (EpoE-3′) is the first cis-acting regulatory element described that acts in a hypoxia-responsive manner (35). We previously reported that a GFP transgene knocked in to the mouse Epo locus within a BAC (wt-Epo-GFP) containing a 180-kb genomic fragment faithfully recapitulates endogenous Epo gene expression in vivo (Fig. 1A) (24, 39), indicating that the 180-kb genomic fragment in the BAC contains information essential for Epo gene expression. In this study, using the same BAC-based reporter transgenic system, we examined the functional contributions of EpoE-3′ to Epo gene expression in vivo.

We constructed a mutant transgene (m3-Epo-GFP) that has 3 nucleotide substitutions in the HRE within EpoE-3′ (Fig. 1A). This substitution (underlined) (ACGTG to AAAAG) has been shown to abrogate the binding of HIF transcription factors (46). Three out of 5 mutant transgenic-mouse lines expressed the GFP reporter in kidney REP (renal Epo-pro- ducing) cells in response to hypoxia (Fig. 1A). Closer immunohistochemical analysis of the GFP expression revealed that the GFP reporter was indeed expressed in the REP cells specifically under anemic conditions, as is the case for the wt-Epo-GFP transgenic mice (data not shown) (24). In contrast, we could not detect expression of the m3-Epo-GFP transgene in adult liver, even if mice were rendered anemic by bleeding to stimulate their Epo production (Fig. 1B, middle). This was in stark contrast to the wt-Epo-GFP transgenic mice, in which expression of GFP was detected in the hepatocytes surrounding the central vein when the mice were exposed to severe anemia concurrently with m3-Epo-GFP mice (Fig. 1B). These results suggest that EpoE-3′ is a liver-specific inducible enhancer and is dispensable for Epo gene expression in the kidney.

We examined the GFP expression profile of the m3-Epo-GFP mutant transgene during embryonic development. The expression of GFP from the wt-Epo-GFP transgene begins at E9.5 in the liver rudiment and continues in hepatocytes throughout the embryonic stage (24). Since GFP expression in the fetal liver could be detected more sensitively by RT-PCR than by GFP immunohistochemistry, we examined m3-Epo-GFP expression by means of RT-PCR using a primer set shown in Fig. 1A. In the E11.5 liver, GFP expression was detected in all 3 lines of m3-Epo-GFP mice (Fig. 1C; two lines are shown). When we examined GFP expression in the postnatal day 0 (P0) liver, expression was detected only in wt-Epo-GFP mice, but not in m3-Epo-GFP transgenic mice. In the m3-Epo-GFP transgenic mice, GFP expression was detected until E13.5 but was completely diminished by E15.5 (data not shown). These results indicate that hepatocyte development is divided into two stages in terms of Epo gene regulation. Therefore, in this study, we refer to the earlier-stage (E9.5 to E14.5) and later-stage (E14.5 and after) hepatocytes as Hep-E and Hep-L, respectively (Fig. 1C). Taken together, these BAC-based transgenic reporter analyses demonstrate that EpoE-3′ activity is required for inducible expression of the Epo gene in Hep-L and adult hepatocytes but that the activity is dispensable for Hep-E and REP cells.

EpoE-3′ is sufficient to support inducible expression of the Epo gene in late hepatocytes.

We now asked if EpoE-3 is sufficient to support inducible expression of Epo in the liver. For this purpose, we subcloned a 0.5-kb mouse genomic fragment containing the EpoE-3′ enhancer sequence, which is highly conserved throughout the mammalian genome (37). We named this genomic fragment the BS region, as it is located between BglII and SphI restriction enzyme sites (Fig. 1A). The BS region was ligated to both the mouse HSP68 (heat shock protein 68) gene promoter (15) and a GFP cassette to make the construct BS-HSP-GFP (Fig. 1A), in which GFP is expressed under the regulatory influence of the BS region. We generated six transgenic-mouse lines carrying this genomic fragment, three of which expressed GFP in adult hepatocytes in an anemia-inducible manner, as was the case for the wt-Epo-GFP transgene (Fig. 1A and B). This demonstrates that the BS region containing EpoE-3′ is sufficient for inducible expression of the Epo gene in the liver.

However, all three lines of mice also expressed the transgene in tissues and cells, such as placenta and heart, which do not normally express the endogenous Epo gene or the wt-Epo-GFP transgene (data not shown). Unlike wt-Epo-GFP and m3-Epo-GFP, expression of the BS-HSP-GFP transgene was not detectable in hepatocytes at E11.5 but was detectable by E15.5 (data not shown). These results indicate that, while the BS region is sufficient for inducible expression of the Epo gene in Hep-L cells, a further regulatory region(s) is required for correct timing of transgene expression and suppression of ectopic expression (data not shown), problems which are considered to be caused by the HSP68 gene promoter (15). Taken together, these transgenic-reporter analyses revealed that the BS region is necessary and sufficient for Epo gene expression in Hep-L cells in vivo. In contrast, EpoE-3′ in the BS region is dispensable for Epo gene expression in the other Epo-producing cells, i.e., Hep-E and REP kidney cells.

Deletion of EpoE-3′ from the mouse Epo gene.

We addressed the physiological roles of the BS region by adopting a Cre-loxP-based homologous-recombination strategy (Fig. 2 A). For this purpose, we prepared a homologous-recombination knockin vector harboring an IRES-GFP gene cassette in the 3′ untranslated region of the Epo gene and a neomycin resistance gene (Neo) cassette 3′ to the Epo gene (Fig. 2A). This vector was introduced into ES cells, and several homologous knockin lines of mice bearing the vector (Epo^fn) were established from homologously recombined ES cell clones. These lines of mice were crossed sequentially with Ayu1-Cre mice that express Cre recombinase ubiquitously (1). We obtained an allele containing a germ line Neo cassette deletion (Epo^f) and another allele containing the Neo cassette plus a BS region deletion (Epo^δ) through these crossings. We determined the genotype of these mice by PCR (data not shown) and Southern blotting (data not shown).

Fig. 2.

Establishment of mouse lines lacking the BS region of the Epo gene. (A) Deletion strategy for the BS region, including 3′ HRE (black dots). By homologous recombination in ES cells using the indicated vector DNA, loxP sequences, IRES-GFP, and Neo cassettes were inserted near the BS region of the mouse Epo gene. Floxed (ƒ) and deleted (δ) Epo-BS alleles were obtained by Cre-mediated recombination. Five exons of the Epo gene are represented by open (untranslated region) and shaded (translated region) boxes. Arrowheads, XbaI recognition sites; pA, polyadenylation signal sequence; DT3, gene cassette for expression of diphtheria toxin. The expected sizes of XbaI-digested genomic fragments hybridized with probe (red line) are shown. (B) BS-null (δ/δ) and wild-type (w/w) littermates 1 day after birth (P1). Note that the BS-null neonate is pale compared with its wild-type littermate. (C) Hct and Hb concentrations in mice with each genotype at 8 weeks of age. The data are shown as means with standard deviations calculated for more than 5 animals in each genotype.

Homozygous deletion of the BS region (Epo^δ/δ) did not cause lethality in adult mutant mice. The Epo^δ/δ mice were born alive and conforming to a Mendelian frequency, as was the case for Epo^f/f mice (data not shown). Since both the Epo gene knockout mouse and the EpoR gene knockout mice were reported to suffer from lethal anemia (48), these results demonstrate that the BS region of the Epo gene is dispensable for mouse development and survival to adulthood.

Meanwhile, we noticed that Epo^δ/δ newborn mice were severely pale compared with littermates with other genotypes (Fig. 2B). However, they grew normally and recovered from severe anemia in the juvenile period (see below). Hct and Hb contents in the adult Epo^δ/δ mice were comparable to those in the wild-type (Epo^wt/wt) and Epo^f/f mice (Fig. 2C and data not shown). These results indicate that the BS region deletion gives rise to the loss of Epo gene expression in embryonic stages, resulting in neonatal anemia.

Loss of EpoE-3′ causes anemia in the neonatal mouse.

We examined how BS-null (Epo^δ/δ) neonates recover from anemia by monitoring the Hct and plasma Epo concentrations. In the wild-type mouse, the Hct was approximately 30% at birth and gradually increased, along with the growth of the mice (Fig. 3 A). Since the heterozygous mutant mice (Epo^δ/w) showed erythroid parameters similar to those of wild-type mice throughout their lives (data not shown), we compared the Hct scores of heterozygous Epo^δ/w pups derived from the crossing of Epo^δ/w and Epo^δ/δ parents.

Fig. 3.

Deletion of the BS region causes a severe defect in adult hepatic Epo production. (A) Change of Hct and Epo concentrations in the peripheral blood of Epo^δ/w and Epo^δ/δ littermates. Hct and Epo concentrations were measured 0, 3, 7, and 14 days after birth. n = 3 for each point. The data are represented as means with standard deviations. *, P < 0.01 compared with wild-type mice. (B) Epo concentration in the peripheral blood. The Epo concentration was measured by ELISA before (Normal) and after (Anemia; Hct < 15) induction of bleeding anemia in each genotype of 8-week old mice (n = 5 or 6). The data are represented as means with standard deviations. (C) Relative Epo mRNA levels in the kidney and liver. Epo mRNA levels were examined in the kidney and liver under normal and anemic conditions by quantitative RT-PCR. The Epo mRNA level was normalized to the GAPDH expression level. The data are represented as means with standard deviations (n = 5). The arrows indicate undetectable expression levels. (D) RNA-blotting analysis of Epo mRNA in the kidneys of anemic mice. Epo mRNA was examined using the probe shown in Fig. 2A. Epo-GFP mRNA from the modified alleles (Epo^ƒ and Epo^δ) was longer than the endogenous Epo mRNA (2.4 × 10³ and 1.6 × 10³ nucleotides, respectively) due to linking to the IRES-GFP cassette. 18S rRNA was used as a loading control.

The Hct scores of the homozygous (Epo^δ/δ) and heterozygous (Epo^δ/w) BS mutant neonates were 17.0% ± 1.8% and 27.1% ± 1.9%, respectively (Fig. 3A). The Hct increased along with the growth of both types of pups, but the Hct scores of Epo^δ/δ mice caught up with those of the Epo^δ/w mice by P14 (Fig. 3A). Thus, the BS deletion affects mouse erythropoiesis in the first 2 weeks after birth.

In contrast, the plasma Epo concentration of BS-null (Epo^δ/δ) mice was comparable to that of the heterozygous mice even from the neonatal stage (Fig. 3A), indicating that the loss of the BS region did not affect the circulating plasma Epo concentration in neonatal stages. These observations suggest that the enfeebled erythropoiesis of the BS-null neonates is not directly related to the plasma Epo concentration, but rather, correlates with the lack of Epo production in the fetal/neonatal liver.

Epo production in the adult liver is dependent on EpoE-3′.

Under normal physiological conditions, deletion of the BS region did not affect the plasma Epo concentration in adult (Fig. 3B, gray bars) or neonatal (Fig. 3A) mice. The plasma Epo concentration measured by ELISA was markedly raised in adult mice displaying a lowered Hct level of 15% induced by bleeding anemia. All genotypes, including BS-null mice, were similarly affected (Fig. 3B, solid bars). The BS-null mice recovered from the anemia on a time course similar to that of the heterozygous mutant and wild-type mice (data not shown), strongly supporting our contention that the BS region is dispensable for stress erythropoiesis.

We also examined expression levels of Epo mRNA in the adult liver and kidney under normal and anemic conditions by quantitative RT-PCR analyses. In the liver, Epo mRNA expression was undetectable under normal conditions but was markedly induced by bleeding anemia in the heterozygous mutant and wild-type mice (Fig. 3C). In contrast, Epo mRNA was undetectable in the BS-null mouse livers even under anemic conditions, indicating that the BS region is critical for Epo gene expression in adult hepatocytes. In the kidney, we did not find any detectable changes in Epo mRNA expression in all mouse genotypes (Fig. 3C).

In Fig. 3C, the Epo mRNA expression levels are shown after normalization with GAPDH mRNA levels, and this allowed us to compare actual expression levels of Epo mRNA in both kidney and liver. The results indicate that the Epo mRNA level in the kidney (204 ± 48 in wild-type mice) is significantly higher than that in the liver (7.3 ± 1.7) in an anemic situation. As sizes of mutant Epo mRNAs (Epo-GFP mRNA) from Epo^ƒ and Epo^δ alleles are larger than that from the wild-type allele due to insertion of the IRES-GFP cassette (Fig. 2A), Epo mRNA expression levels from the mutant alleles (Epo^ƒ and Epo^δ) can be compared with that from the normal allele (Epo^w) in heterozygous individuals (Fig. 3D). The result indicates that Epo gene expression in the kidney was not affected by the deletion of the BS region.

We also examined GFP expression from Epo^ƒ and Epo^δ alleles by using the IRES function. Showing very good agreement with the mRNA analysis, GFP fluorescence was detected in REP cells of anemic BS-null mutant (Epo^δ/δ) mice (data not shown). Taken together, these results indicate that the BS region is essential for hepatic Epo expression but not for renal Epo production in the adult mouse. Plasma Epo is mainly produced and supplied by the kidney, and the contribution of hepatic Epo production is very low. Although BS-null mice do not produce Epo in the liver, their erythropoiesis and plasma Epo concentration are within the normal range, even in a stress erythropoiesis situation.

The BS region is essential for histone acetylation of the Epo promoter in liver.

Hepatocytes express Epo in response to anemic or hypoxic stimuli. As the histone acetylation status reflects characteristics of chromatin competence for transcriptional activation (21), we took advantage of mice with the cis element deleted to examine a possible contribution of the BS region to Epo gene regulation via chromatin organization around the gene by using ChIP assay of acetylated histones.

To clarify whether the BS region is required for the histone modification needed for Epo gene expression, chromatin complexes were immunoprecipitated from the livers of BS-null and wild-type mice under anemic or normal conditions by anti-acetylated histone antibodies. Under both normal and anemic conditions, the Epo promoter region was recovered in the chromatin complex immunoprecipitated with anti-AcH3 and -AcH4 antibodies from wild-type liver (Fig. 4). We used the α-catenin gene promoter as an internal positive control (44). In contrast, in the BS-null liver, histones in the Epo promoter region were not acetylated even under anemic conditions (Fig. 4). This result indicates that the chromatin configuration surrounding the Epo promoter is actively poised for Epo gene expression in the liver and that the BS region is essential for this. Hepatocytes seem to prepare for immediate Epo gene induction through constitutive histone acetylation and chromatin opening. Since we could not detect histone acetylation in the Epo allele with BS deleted, lack of promoter competence may be one of the reasons why BS-null hepatocytes lose their ability to express Epo.

Fig. 4.

Deletion of the BS region results in defects of histone acetylation in the promoter region. The necessity of the BS region for the acetylation of histones around the Epo gene promoter was examined by ChIP assay. Cross-linked chromatins from the wild-type (wt/wt) and BS-null (δ/δ) mouse livers were prepared under normal and anemic conditions. Quantitative PCR was performed to detect the Epo gene promoter region from material precipitated with anti-AcH3 or -AcH4 antibody. The promoter region of the α-catenin gene was detected as an internal positive control, and normal rabbit IgG was used as a negative control for immunoprecipitation. The data are represented as means with standard deviations in one representative experiment.

Hep-L stage-specific defect of Epo expression in BS-null liver.

In our analyses of BS-null embryos, we noticed that the anemic phenotype was evident in the BS-null embryos after E15.5 (Fig. 5 A, right), but before E13.5, BS-null embryos were not anemic (Fig. 5A, left). The embryonic anemia continued after birth, so that pale neonates were seen (Fig. 2B). Since these observations further support the notion that there is multilayer regulation of Epo gene expression during embryonic development, we investigated the expression levels of the Epo gene in the liver and kidney, the two major Epo-producing tissues, during embryonic and neonatal stages. In the liver, Epo expression was first detected at E9.5 (data not shown) and continued up to 2 weeks after birth in wild-type and heterozygous mutant mice (Fig. 5B). The levels of Epo expression varied during mouse development, peaking at P3 in the normal mouse (Fig. 5B). Importantly, BS-null embryos appeared not to express Epo in the liver after E14.5 (Fig. 5B).

Fig. 5.

Loss of Epo expression in the late-stage embryonic liver in BS-null mice. (A) Epo^wt/wt and Epo^δ/δ embryos at E13.5 (left) and E15.5 (right). Note the anemic appearance only in the E15.5 BS-null embryos. (B) Relative Epo mRNA levels in the liver and kidney were measured by quantitative RT-PCR analysis at the indicated developmental stages. GAPDH expression levels were used as internal controls. The data are represented as means with standard deviations. The arrows indicate undetectable expression levels. Epo expression levels in E12.5 and E13.5 kidneys were not determined (nd) because they were not developed. n = 3 at each point. (C) GFP expression was detected in the peritubular interstitial cells of Epo^δ/w and Epo^δ/δ kidneys at P0 (arrowheads). The number of GFP-positive cells in the Epo^δ/δ kidney is higher than that in the Epo^δ/w kidney. Scale bar, 200 μm.

In the kidney, Epo expression was first detected at E17.5, and afterward, the levels of Epo mRNA were much higher in the kidney than in the liver (Fig. 5B). We found that the Epo mRNA level was much higher in the BS-null mouse kidney than in the wild-type and heterozygote (Epo^δ/w) mouse kidneys at E18.5 and P6 (Fig. 5B), suggesting that renal Epo production was upregulated to compensate for the loss of hepatic Epo production in the BS-null mouse. Indeed, the plasma Epo levels were comparable between the BS-null mutant and wild-type mice after birth (Fig. 3A). These data show very good agreement with GFP reporter expression analysis, in which GFP-positive cell numbers increased approximately 4-fold in the kidney of the BS-null mouse compared with the heterozygote kidney (Fig. 5C).

It is interesting that, whereas both Epo- and EpoR-deficient embryos suffer from severe anemia and die by E13.5 (48), the BS-null embryos are not anemic up to this stage or during the Hep-E stage. In our histological and flow cytometric analyses of E13.5 BS-null embryos, we did not notice any abnormalities in fetal liver and yolk sac erythropoiesis (data not shown). Taken together, these results demonstrate that the loss of EpoE-3′ function provokes anemia that starts around E15.5. Whereas renal Epo expression is upregulated in a compensatory manner, erythropoiesis in the liver is not fully activated by the increase of Epo production in the kidney.

Loss of hepatic Epo production impairs hepatic erythropoiesis.

To dissect impairment of erythropoiesis in the BS-null embryonic liver, we performed erythroid colony-forming assays and flow cytometric analysis of the liver and spleen at E17.5. We found that there were comparable (liver) or slightly increased (spleen) numbers of erythroid progenitors (BFU-E and CFU-E) in BS-null embryos compared with heterozygote embryos (Fig. 6 A). This is consistent with the previous report that hematopoietic progenitors are formed normally in Epo-deficient embryos (48). We then examined erythroblasts by flow cytometry and found a marked decrease in mature erythroblasts in the BS-null liver (CD71⁻ Ter119⁺ fraction in Fig. 6B). In contrast,the numbers of splenic mature erythroblasts were virtually identical in BS-null and heterozygous embryos (Fig. 6B). We did not notice any abnormalities in the other hematopoietic lineages of BS-null embryos throughout these analyses (data not shown). These results indicate that the Epo-EpoR signaling pathway is essential for the terminal maturation of erythroblasts at the late embryonic Hep-L stage (such as E17.5) and nicely complement the initial observation reported in the Epo gene knockout mouse at E13.5 (48).

Fig. 6.

BS deletion causes arrest of erythroid cell terminal maturation in the liver but not in the spleen. (A) The erythroid colony formation assay showed the numbers of CFU-E and BFU-E erythroid progenitors in the livers and spleens of Epo^δ/w and Epo^δ/δ embryos at E17.5. The data are represented as means with standard deviations in a triplicate experiment with a representative embryo among 4 embryos investigated. (B) CD71 and Ter119 expression of hematopoietic cells in the livers and spleens of Epo^δ/w and Epo^δ/δ embryos. The percentages of cells in each quadrangle are shown. Note that the mature erythroblast fraction (CD71⁻ Ter119⁺) of the BS-null liver was severely decreased. (C) Schematic description of the relationship between the definitive erythropoietic tissues and Epo-producing tissues during mouse development. The hepatocytes and REP cells produce Epo for definitive erythropoiesis. Hepatocytes are divided into Hep-E and Hep-L in terms of their Epo gene regulation. Hep-E cells start to produce Epo at E9.5 and switch to Hep-L cells around E14.5. The BS region is required for hypoxia-inducible Epo gene expression in Hep-L cells throughout life (*). BS deletion results in transient anemia between E15.5 and P14. Fetal hepatic erythropoiesis requires the paracrine-secreted Epo from hepatocytes. The endocrine-secreted Epo from kidney REP cells does not support hepatic erythropoiesis, but the endocrine action of Epo from REP cells is required for erythropoiesis in the spleen and bone marrow. Most Epo is produced by the kidney after the major erythropoietic tissues switch from the liver to the bone marrow and spleen around P14, and the loss of hepatic Epo production does not affect adult erythropoiesis.

Since the fetal liver is the major erythropoietic tissue in the embryo, severe anemia in the BS-null embryos must be caused by a defect in hepatic erythropoiesis that is not compensated for by erythropoiesis in the spleen and bone marrow. Based on these observations, we propose that erythropoiesis in the liver specifically requires Epo production in hepatocytes and that this paracrine step cannot be replaced by the increase of circulating Epo. On the contrary, kidney-derived Epo is responsible for erythropoiesis in the spleen and bone marrow (a schematic model is shown in Fig. 6C). Thus, paracrine-secreted Epo stimulates erythropoiesis in the fetal and neonatal liver, while endocrine-secreted Epo supports erythropoiesis in adult hematopoietic tissues (Fig. 6C).

The BS-null mouse shows liver-specific defects in erythropoiesis.

In the analyses focused on the difference between hepatic and splenic erythropoiesis, we found that neonatal livers of BS-null mice were slightly small and pale compared with Epo^δ/w mouse livers (data not shown). In contrast, there were only marginal changes in the size and color of Epo^δ/δ and Epo^δ/w neonatal spleens (data not shown). At E17.5, BS-null fetal liver harbored fewer β-globin-expressing erythroid cells than did the heterozygote and wild-type embryos (Fig. 7 A). At the same time, the number of apoptotic cells increased in the BS-null liver compared with the heterozygote Epo^δ/w liver (Fig. 7B).

Fig. 7.

Erythroid cells are decreased in the livers of BS-null embryos and neonates. (A and B) Detection of erythroblasts positive for anti-β-globin staining (A) and TUNEL-positive apoptotic cells (B) in the livers of Epo^δ/w and Epo^δ/δ embryos at E17.5. Positive signals are detected as brown staining. Scale bars, 50 μm. (C and D) Immunohistochemistry with anti-β-globin antibody was performed on liver (C) and spleen (D) sections from Epo^δ/w or Epo^δ/δ neonates at P7. (C) Erythropoietic cell clusters stained brown with anti-β-globin antibody (red circles) were observed in the Epo^δ/w liver, whereas there were no erythropoietic clusters in the BS-null liver. The arrowheads indicate the enucleated circulating red cells in the vessels. (D) The BS-null spleen was normal compared with the Epo^δ/w spleen in the size and number of β-globin-expressing erythroblasts. Scale bars, 50 μm (C) and 100 μm (D).

Whereas there were many erythroid clusters in the livers of wild-type mice, no such clusters were found in the BS-null liver at P7 (Fig. 7C). On the other hand, histological and flow cytometric analyses revealed that the splenic erythropoiesis of the BS-null mouse was normal throughout the mouse life (Fig. 7D and data not shown). In conclusion, these results suggest that paracrine-secreted Epo stimulates anti-apoptotic signals in Hep-L stage erythroblasts and that endocrine-secreted Epo is required for the maintenance of erythroid homeostasis in adult mouse stages.

DISCUSSION

In this study, we examined the in vivo function of an Epo gene 3′ enhancer, EpoE-3′, which is the first known hypoxia-responsive enhancer (35). The BAC-based GFP reporter transgenic-mouse analysis clearly demonstrated the cell type- and developmental-stage-specific activity of the enhancer, in addition to its stress response ability. The physiological significance of the enhancer was also examined by adopting a gene-targeting strategy. In homozygous EpoE-3′ deletion (Epo^δ/δ) mice, hepatocytes in the Hep-L stage lost their ability to produce Epo, and the mutant mice suffered from anemia during the late embryonic to neonatal stages. The results further demonstrate a specific contribution of hepatic Epo to the Hep-L stage of erythropoiesis through paracrine secretion.

Using a GFP reporter transgenic-mouse system, we identified two major Epo-producing cells in adult mice, i.e., REP cells of the kidney and hepatocytes (24). GFP reporter expression was undetectable in other tissues or cell types, and even in these two types of Epo-producing cells, GFP fluorescence was detected only under anemic or hypoxic conditions. Since reporter transgenes were constructed by utilizing a 180-kb BAC clone encompassing the Epo gene, these cell-type-specific and hypoxia (or anemia)-inducible cis-regulatory elements must reside within the 180-kb region covered by the BAC. We also found that Epo gene expression in Hep-L stage hepatocytes is controlled specifically by EpoE-3′, which is located in the BS region and contains a canonical HRE.

We verified the EpoE-3′ activity by exploiting two distinct approaches. One is the BS-HSP-GFP transgenic-mouse analysis. In this reporter mouse analysis, we found that the BS region, and hence the EpoE-3′ enhancer, is active in Hep-L stage hepatocytes when driven either by the endogenous Epo or the HSP68 promoter. The other approach is cis element targeting of the BS region by homologous recombination. We unexpectedly found that the mice homozygously lacking the BS region, BS-null (Epo^δ/δ) mice, are born alive. However, we also found that the mice lose Epo gene expression specifically in the Hep-L stage. As this markedly affects hepatic erythropoiesis, the BS-null mice suffered from severe anemia during the neonatal period. At the start of splenic and bone marrow erythropoiesis, the mice recovered from anemia. These results in combination demonstrate that EpoE-3′ contributes to erythropoiesis in the Hep-L stage liver.

There are one HRE motif and one DR2 motif within the BS region (2, 5, 39). DR2 is recognized by nuclear receptors, such as HNF4α, and retinoic acid receptors (2, 20). We surmise that HRE and HIFs are the major functional players in EpoE-3′ activity. In fact, a 3-nucleotide mutation of HRE caused complete loss of Epo gene expression in the liver. In this regard, it is interesting that an HRE in the VEGF gene is reported to be critical for both neural-cell-specific and hypoxia-inducible expression of the gene (25). This observation suggests that two distinct activities, i.e., tissue specificity and inducibility, may reside in a single HRE motif. Alternatively, there remains a possibility that HNF4α plays an important role in determining cell type specificity or inducibility in collaboration with HIFs, as it has been suggested that DR2 regulates tissue-specific Epo gene expression in collaboration with the HRE (7). Indeed, we found that a point mutation in the DR2 motif caused the impairment of reporter transgene expression similar to HRE mutation (N. Suzuki and M. Yamamoto, unpublished data). These results suggest that there may be cooperativity between the tissue-specific and stress-inducible enhancer elements in EpoE-3′. However, erythropoiesis in the late embryonic stage was found to be sustained within the normal range in both HNF4α-deficient and RXRα-deficient embryos (20, 28), implying that both HNF4α and RXRα may be dispensable for Epo gene expression in Hep-L. This issue is now under investigation utilizing the system developed in this study.

In contrast to the situation in Hep-L, both HNF4α deficiency and RXRα deficiency in midgestation (E11.5 to E12.5) embryos cause anemia and reduction of Epo production in hepatocytes (20, 28). These observations imply the necessity of these transcription factors for Epo gene expression in Hep-E. Since the BS-null mice do not show abnormalities in early hepatic Epo expression in the Hep-E stage, we envisage that the DR2 motif localized somewhere in the 180-kb genomic region, but not in the BS region, may be recognized by RXRα and HNF4α and regulates Epo gene expression in Hep-E. An alternative interpretation of the data is that the loss of Epo production is caused by maturation defects in HNF4α-null and RXRα-null hepatocytes (10, 18), suggesting that the proper maturation of early hepatocytes in Hep-E may be essential for initiation of Epo production in the E9.5 liver. Detailed assessments are required to elucidate the important contributions of these two factors to Epo gene expression during the Hep-E stage.

Recently, several reports demonstrated that HIF-2α is more critical than HIF-1α for Epo gene expression in Hep3B cells derived from human hepatocellular carcinoma (47) and in the adult mouse liver (30). Additionally, double-null mutation of genes for the proline hydroxylase domain proteins 1 and 3 (PHD1/3) caused accumulation of HIF-2α and overproduction of Epo in the adult mouse liver (43). Since hepatic erythropoiesis is also induced in this mutant mouse line, PHD1/3 are key factors for Epo gene expression and erythropoiesis in Hep-L through the regulation of HIF-2α stability. These observations support the notion that HIF-2α regulates hepatic Epo production in hypoxic Hep-L through binding to the HRE in the BS region.

We tested the importance of the GATA box in the Epo promoter for hepatic Epo gene expression using the BAC-based transgenic-mouse system, and the results showed that the GATA box is not important for Epo gene expression in both the Hep-E and Hep-L stages (reference 24 and data not shown). Consistent with this observation, our in vivo studies have shown that EpoE-3′ is the only enhancer that is necessary and sufficient for Epo gene expression in Hep-L stage hepatocytes.

We found constitutive acetylation of histones in the Epo promoter region of the wild-type mouse liver by ChIP assay. This is consistent with the previous observation that RNA polymerase II resides in the Epo promoter of Hep3B cells, even when the cells are under normoxic conditions and not expressing the gene (4). The constitutive histone acetylation and chromatin opening may represent the molecular basis for the immediate response of the Epo gene to hypoxia. In contrast, however, Wang et al. reported that histones in the Epo promoter region of Hep3B cells are acetylated in a hypoxia-inducible manner (45). The reason for this discrepancy is unknown.

In the ChIP assay utilizing BS-null liver cells, we found that histones in the Epo promoter region are not acetylated in BS-null cells under unstressed normoxic conditions, suggesting that EpoE-3′ is necessary for the promoter histone acetylation. One plausible explanation for this observation is that constitutive histone acetylation may be initialized in a cell-type-specific manner and epigenetically maintained during the unstressed period. This might be achieved through a transcription factor bridge between the BS and promoter regions, which ensures competence to express the Epo gene in the hepatocytes. However, in BS-null Hep-L cells, this initialization is abrogated, and the Epo gene is no longer competent in inducible expression. Indeed, CBP and p300 coactivators harboring histone acetyltransferase activity interact with HIF-α/β subunits (6, 14, 31), suggesting that the histones in the Epo promoter region are acetylated by the transcription factor bridge composed of a HIF-CBP/p300 complex through promoter-enhancer contact.

The embryonic liver is the tissue in which definitive erythropoiesis operates, and Epo expression is detectable in the liver at E9.5 before the explosion of erythroid cells (32). Hep-E produces Epo up to E14.5, and the numbers of erythroid cells are increased in the liver. Actually, the Epo gene knockout mice are lethal around E13.5 (48). After E14.5, the Epo-producing hepatocytes transit from Hep-E to Hep-L, and both the proportion and number of erythroid cells in the liver begin to decrease (32). At the same time, the splenic erythroid progenitors develop and the character and ultrastructure of hepatocytes alter as they take on the role of metabolic regulation (32, 50, 51). We suggest that the function of the liver changes from a hematopoietic tissue to a metabolic tissue around E14.5, when the Epo gene regulatory system also transits from Hep-E to Hep-L. The microenvironment of Hep-E seems to be suitable for erythropoiesis, but that of Hep-L may not be so favorable to erythroid cell growth. Consequently, fetal liver erythropoiesis is shut down when the hepatocytes stop producing Epo around P14 (Fig. 6C). Thus, hepatic Epo production is specifically required for terminal erythropoiesis during the embryonic to neonatal stages.

In conclusion, this study has revealed the relationship between erythropoietic and Epo-producing tissues, both of which change during development. One of the most important points in this relationship is that hepatic erythropoiesis in embryonic and neonatal periods requires paracrine action of Epo from the hepatocytes for the growth and maturation of the erythroblast. In contrast, the loss of hepatic Epo production does not affect erythropoiesis in the spleen and bone marrow, demonstrating that the Epo secreted from the kidney acts in an endocrine manner on erythroid cells in the bone marrow and spleen via the circulation.