Hypoxia-inducible Factor-1 Deficiency Results in Dysregulated Erythropoiesis Signaling and Iron Homeostasis in Mouse Development*

Hypoxia-inducible factor-1 (HIF-1) regulates the transcription of genes whose products play critical roles in energy metabolism, erythropoiesis, angiogenesis, and cell survival. Limited information is available concerning its function in mammalian hematopoiesis. Previous studies have demonstrated that homozygosity for a targeted null mutation in the Hif1α gene, which encodes the hypoxia-responsive α subunit of HIF-1, causes cardiac, vascular, and neural malformations resulting in lethality by embryonic day 10.5 (E10.5). This study revealed reduced myeloid multilineage and committed erythroid progenitors in HIF-1α-deficient embryos, as well as decreased hemoglobin content in erythroid colonies from HIF-1α-deficient yolk sacs at E9.5. Dysregulation of erythropoietin (Epo) signaling was evident from a significant decrease in mRNA levels of Epo receptor (EpoR) in Hif1α-/- yolk sac as well as Epo and EpoR mRNA in Hif1α-/- embryos. The erythropoietic defects in HIF-1α-deficient erythroid colonies could not be corrected by cytokines, such as vascular endothelial growth factor and Epo, but were ameliorated by Fe-SIH, a compound delivering iron into cells independently of iron transport proteins. Consistent with profound defects in iron homeostasis, Hif1α-/- yolk sac and/or embryos demonstrated aberrant mRNA levels of hepcidin, Fpn1, Irp1, and frascati. We conclude that dysregulated expression of genes encoding Epo, EpoR, and iron regulatory proteins contributes to defective erythropoiesis in Hif1α-/- yolk sacs. These results identify a novel role for HIF-1 in the regulation of iron homeostasis and reveal unexpected regulatory differences in Epo/EpoR signaling in yolk sac and embryonic erythropoiesis.

Hypoxia-inducible factor-1 (HIF-1) regulates the transcription of genes whose products play critical roles in energy metabolism, erythropoiesis, angiogenesis, and cell survival. Limited information is available concerning its function in mammalian hematopoiesis. Previous studies have demonstrated that homozygosity for a targeted null mutation in the Hif1␣ gene, which encodes the hypoxia-responsive ␣ subunit of HIF-1, causes cardiac, vascular, and neural malformations resulting in lethality by embryonic day 10.5 (E10.5). This study revealed reduced myeloid multilineage and committed erythroid progenitors in HIF-1␣-deficient embryos, as well as decreased hemoglobin content in erythroid colonies from HIF-1␣-deficient yolk sacs at E9.5. Dysregulation of erythropoietin (Epo) signaling was evident from a significant decrease in mRNA levels of Epo receptor (EpoR) in Hif1␣ ؊/؊ yolk sac as well as Epo and EpoR mRNA in Hif1␣ ؊/؊ embryos. The erythropoietic defects in HIF-1␣-deficient erythroid colonies could not be corrected by cytokines, such as vascular endothelial growth factor and Epo, but were ameliorated by Fe-SIH, a compound delivering iron into cells independently of iron transport proteins. Consistent with profound defects in iron homeostasis, Hif1␣ ؊/؊ yolk sac and/or embryos demonstrated aberrant mRNA levels of hepcidin, Fpn1, Irp1, and frascati. We conclude that dysregulated expression of genes encoding Epo, EpoR, and iron regulatory proteins contributes to defective erythropoiesis in Hif1␣ ؊/؊ yolk sacs. These results identify a novel role for HIF-1 in the regulation of iron homeostasis and reveal unexpected regulatory differences in Epo/EpoR signaling in yolk sac and embryonic erythropoiesis.
HIF-1␣-deficient embryos manifested disorganized vascularization of the yolk sac, although fully formed vessels containing red blood cells were present (7). Furthermore, nucleated red blood cells were detected in the dorsal aorta of the embryo proper at E9.75 (15). These data indicate that the absence of HIF-1␣ does not completely abrogate erythropoiesis.
In this study, we demonstrate that HIF-1␣-deficient embryos have reduced numbers of erythroid progenitors and impaired terminal erythroid differentiation in the yolk sac, which differ from the defects described in HIF-1␤-deficient embryonic stem cells. We show that these defects associate with dysregulated expression of genes encoding Epo, EpoR, VEGFR1, and iron regulatory proteins. The results establish a pivotal role for HIF-1 in the regulation of yolk sac erythropoiesis and iron metabolism in mouse development.

EXPERIMENTAL PROCEDURES
Mice and Genotyping-Hif1␣ ϩ/Ϫ mice were previously generated by gene targeting and have been maintained on a mixed background (C57BL/6 ϫ 129 genetic background) by brothersister mating as described (5). DNA isolated from embryos or neonatal tail biopsies was used for genotyping by multiplex PCR. The PCR was performed using HotstarTaq master mix kit (Qiagen Inc., Chatsworth, CA) in a 12-l reaction mixture (3.5 mM MgCl 2 , 200 mM of each deoxynucleotide triphosphate, 2.4 pM of each primer, 100 ng of genomic DNA, and 0.6 units of HotStarTaq DNA polymerase). A PCR condition included an initial heat activation step at 95°C for 15 min, followed by 30 cycles of 30 s denaturation at 95°C, 30 s annealing at 58°C, 60 s extension at 72°C, and a final 10-min extension at 72°C. The multiplex PCR was used with the following primers: HIF-1␣37, 5Ј-TTT CCA GTA CTG CCC CAA; HIF-1␣REV, 5Ј-GCA AAC AAG CAA ATC ACC AAG G; PGK Pro-REV, 5Ј-GGG GCT GCT AAA GCG CAT GC, which generated 380-and 600-bp bands for the null and wild type alleles, respectively.
Yolk Sac Isolation-Hif1␣ ϩ/Ϫ mice, aged 8 weeks or older, were used for timed mating. Noon of the day when the vaginal plug was detected was considered E0.5. The pregnant females were sacrificed at E9.5. Following embryo dissection, the yolk sac and embryo were washed separately three times in PBS and then digested in 0.25% collagenase in PBS with 20% FBS (Stemcell Technologies, Vancouver, British Columbia, Canada) at 37°C for 1 h. The cells from the yolk sac were either directly used for in vitro hematopoietic culture assays or suspended in Tri-reagent solution (Molecular Research Center, Cincinnati, OH) for RNA extraction. RNA, protein and genomic DNA were extracted in Tri-reagent from digested embryos.
Quantitative Real Time RT-PCR-The mRNA levels of selected genes in the yolk sac and embryo were measured by one-or two-step real time RT-PCR on an ABI Prism 7000 sequence detection system (Applied Biosystems Inc., Foster City, CA). The primers and TaqMan MGB probes for each gene were either designed by using Primer Express software version 2.0 (Applied Biosystems), or obtained as a 20ϫ Assays-on-Demand Gene Expression Assay mix commercially from Applied Biosystems. The sequences of primers/probes and commercial assay identifications are described in Table 1. The one-step real time RT-PCR was performed in a 20-l reaction mixture using TaqMan one-step RT-PCR master mix reagents kit (Applied Biosystems), 900 nM of each primer, and 100 nM of the TaqMan probe. We used 5-500 ng of RNA/reaction giving the linear range of response for selected genes, and 100-fold diluted RNA for 18 S. The universal temperature cycling consisted of 30 min for reverse transcription at 48°C, denaturation and polymerase activation at 95°C for 10 min, followed by 45 cycles of denaturation at 92°C for 15 s and annealing/extension/plate reading at 60°C for 1 min. For two-step, initial total RNA (5-500 ng) was used for first strand synthesis using reverse transcriptase (Invitrogen) and random hexamer. The first strand synthesis followed the manufacturer's protocol. One l of cDNA was used in a 20-l reaction mixture using the TaqMan Universal PCR master mix, No AmpErase UNG reagent kit (Applied Biosystem), 900 nM of each primer, and 100 nM of the TaqMan probe using FAM and VIC fluorescence dyes (Applied Biosystems). The universal temperature cycling consisted of polymerase activation at 95°C for 10 min, followed by 45 cycles of denaturation at 92°C for 15 s and annealing/extension/plate reading at 60°C for 1 min. VEGFR1 and HIF2␣ were analyzed using Cyber Green dye. All samples were analyzed in duplicate wells. RT-PCR in the absence of reverse transcriptase was performed for each sample to rule out genomic DNA contamination. The relative quantitative expression of each gene in each sample was normalized to 18 S rRNA level. The cycle threshold (C T ) value for each of eight selected mRNAs and 18 S rRNA was determined, and RNA levels were calculated as ⌬C T ϭ C T of target mRNA Ϫ C T of 18 S. A lower value of ⌬C T indicates a higher selected gene expression.
Immunohistochemistry-Dissected E9.5 embryos along with the attached yolk sac were fixed overnight in 4% paraformalde-hyde/PBS, dehydrated in a graded series of ethanol, embedded in paraffin, and sectioned at 5 m. Sections from wild type and HIF-1␣ Ϫ/Ϫ embryos were mounted side by side on the same slide to allow for comparison of expression levels. Immunohistochemistry was performed as described (19). Briefly, sections were treated with 3% H 2 O 2 in methanol for Fpn1 antigen retrieval. For Dmt1 and TfR detection, sections were boiled in 0.01 M citric acid, pH 6.0. Blocking was achieved with goat (Fpn1 and Dmt1) or horse serum (TfR). Primary antibody incubations were performed overnight at 4°C using rabbit anti-Fpn1 at 1:200 (20), rabbit anti-Dmt1 at 1:100 (Alpha Diagnostic International), and mouse anti-TfR at 1:1000 dilution (Invitrogen). Secondary antibodies were peroxidase-conjugated with the Vectastain Elite ABC kit (Vector Laboratories), followed by signal detection with Vector NovaRED substrate (Vector Laboratories).
Western Blot Analysis-E9.5 embryos and yolk sacs were dissected from timed pregnant females, and a small portion of the embryo tail was recovered for genotyping. Embryos and yolk sacs were processed individually for Western blot analysis as described (21). Briefly, embryos and yolk sacs were homogenized and lysed in RIPA buffer plus Complete, EDTA-free protease inhibitor (Roche Applied Science). Extract supernatant was collected and protein was quantified using the Bio-Rad DC protein assay kit (Bio-Rad). Seven and five g of total protein from embryo and yolk sac, respectively, were mixed with an equal volume sample buffer with ␤-mercaptoethanol, boiled, separated on 8% SDS-PAGE, and transferred to polyvinylidene difluoride membrane. Blocking was achieved by incubation in Tris-buffered saline containing 5% milk and 0.1% Tween 20. Membranes were incubated overnight at 4°C using the following primary antibodies: mouse anti-TfR at 1:2000 (Invitrogen) and goat anti-actin at 1:5000 dilution (Santa Cruz Biotechnology). After washing, membranes were incubated with horseradish peroxidase-conjugated secondary antibodies (1:10,000 dilution), and the signal was developed using ECL reagent (Santa Cruz Biotechnology).
Statistical Analysis-Student's t test and analysis of variance were used for statistical analysis. Significant differences were

Hif1␣-deficient Yolk Sacs Contain Blood
Islands-Hif1␣-deficient embryos die between E10 and E10.5 (5,7,15). We dissected yolk sacs at E9.5 for analysis of primitive erythroid progenitors or for analyses of definitive erythroid and CFU-Mix progenitors, respectively. At E9.5, somite counts ranged from 11 to 17 for Hif-1␣ Ϫ/Ϫ embryos to 21-25 in wild type littermates. This indicated a moderate developmental delay. Furthermore, it is noteworthy that the experiments described below assess intrinsic properties of erythropoietic cells in yolk sac, which are less likely to be affected by a developmental delay of the embryo.
Erythropoietic Defects in Hif1␣ Ϫ/Ϫ Yolk Sac-To directly examine the role of HIF-1 in erythropoietic development, we studied the growth of yolk sac erythroid progenitors in vitro. First, we analyzed definitive erythropoiesis in the yolk sacs of Hif1␣ Ϫ/Ϫ and wild type mouse embryos dissected from the uterus at E9.5. Yolk sac erythroid progenitor cells were analyzed by in vitro hematopoietic colony assays using various conditions.
Using methylcellulose media with PWM-SCCM (see condition 1 under "Experimental Procedures"), the numbers of erythroid and CFU-Mix colonies derived from the mutant yolk sac were significantly reduced ( Table 2). We considered two possible explanations for these results as follows: 1) the culture condition for progenitors from Hif1␣ Ϫ/Ϫ yolk sacs could not support colony growth or 2) erythropoiesis in Hif1␣ Ϫ/Ϫ yolk sac was severely impaired. To test these hypotheses, we cultured the yolk sac cells under condition 2, which favors BFU-E and CFU-Mix. Under this condition, the number of CFU-Mix from Hif1␣ Ϫ/Ϫ yolk sacs was improved, but not the number of BFU-E and CFU-E (Table 2). However, the size (cellularity) of the Hif1␣ Ϫ/Ϫ CFU-Mix colonies were 2-5 times smaller than the colonies derived from the wild type embryos (Fig. 2, C and D), and erythroid cells (BFU-E colonies and erythroid component of CFU-Mix) were not fully hemoglobinized (Fig. 2, A-D). These results indicate impaired erythropoiesis in Hif1␣ Ϫ/Ϫ yolk sacs.
We then hypothesized that multipotential progenitors might be present in the Hif1␣ Ϫ/Ϫ yolk sacs in comparable numbers with the wild type, but they may require higher doses of cytokines/growth factors to stimulate survival and/or proliferation. Because Tpo was shown to promote mixed lineage colonyforming cell growth in yolk sac cultures (22), we cultured the yolk sac cells in the presence of high serum levels and Tpo (condition 3). Again, the numbers of BFU-E and CFU-E derived from Hif1␣ Ϫ/Ϫ yolk sacs were lower than wild type. However, the numbers of CFU-Mix from wild type and Hif1␣ Ϫ/Ϫ yolk sacs were comparable (Table 2). Furthermore, the hemoglobinization of erythroid cells was slightly improved (Fig. 2, E  and F). These data confirmed the presence of multipotential progenitors in the Hif1␣ Ϫ/Ϫ yolk sacs and indicated that HIF-1␣ is essential for erythroid progenitor survival/ growth/differentiation and for colony-forming capacity of multilineage myeloid progenitors. Analysis of primitive erythroid progenitors in the Hif1␣ Ϫ/Ϫ and WT yolk sacs (E9.5) by using condition 4 revealed a significantly lower number and impaired hemoglobinization of Hif1␣ Ϫ/Ϫ erythroid colonies (Fig. 3,  A and B).  Serum and Fe-SIH but Not VEGF Improve Hemoglobinization of Hif1␣ Ϫ/Ϫ Erythroid Colonies-Serum factors have been shown to support BFU-E in vitro (23). Increasing FBS supplementation up to 30% resulted in partial rescue of hemoglobinization of Hif1␣ Ϫ/Ϫ CFU-Mix (Fig. 2F) in the presence or absence of Tpo in the media (data not shown). We hypothesized that this effect may be promoted by serum factors involved either in iron delivery (e.g. transferrin) or in survival/ differentiation. High dose cytokine supplementation (150 ng/ml SCF and 3 units/ml Epo) to the standard culture media (condition 3) did not rescue the defect in terminal differentiation of Hif1␣ Ϫ/Ϫ erythroid cells (data not shown). Similarly, the addition of 5 ng/ml VEGF to the standard culture media (condition 2) failed to improve either hemoglobinization or plating efficiency of erythroid colonies (data not shown), although it did improve the non-erythroid component of the CFU-Mix colonies (Fig. 2, G and H) and also increased plating efficiency of the non-erythroid CFU-GM colonies (data not shown).
Because supplementation with a high concentration of growth factors failed to improve erythropoiesis, we cultured the yolk sacs cells in the presence of Fe-SIH (17), a compound delivering iron into cells independently of the transferrin receptor/Dmt1 system. As shown in Fig. 3A, the addition of 100 M Fe-SIH to condition 4 significantly increased the number of erythroid colonies derived from the wild type yolk sacs (211%, p ϭ 0.027), even though there was no appreciable increase in the number of erythroid colonies from Hif1␣ Ϫ/Ϫ . However, there was significant rescue of the erythroid differentiation defect in Hif1␣ Ϫ/Ϫ as the size of erythroid colonies and the degree of hemoglobinization were markedly improved (Fig. 3B).
Differential Expression of Epo/EpoR and VEGFR1 Expression in Hif1␣ Ϫ/Ϫ Embryos and Yolk Sacs-To understand the molecular mechanisms of the erythropoietic defects in Hif1␣ Ϫ/Ϫ yolk sacs, we measured the mRNA levels of selected gene products related to erythropoiesis and hypoxia signaling by quantitative real time RT-PCR. As shown in Fig. 4, Epo, EpoR, and VEGFR1 mRNA levels were significantly reduced (3.6-, 2.1-, and 2.9-fold, respectively) in Hif1␣ Ϫ/Ϫ embryos, compared with stage-matched wild type controls. In addition, EpoR, but not Epo and VEGFR1 mRNA levels were significantly reduced (2.8-fold) in Hif1a Ϫ/Ϫ yolk sacs. In addition, the levels of mRNAs encoding erythroid-specific 5-aminolevulinate synthase and embryonic ␤-like globin (mHbY) were significantly lower in Hif1␣ Ϫ/Ϫ as compared with wild type embryos (11.8and 7.9-fold, respectively), but in the yolk sac there was no appreciable expression difference between genotypes. In addition, Hif2␣ mRNA levels were not significantly different in the genotypes studied (Fig. 4).
Altered mRNA Expression of Iron Metabolism Genes in Hif1␣ Ϫ/Ϫ Embryos and Yolk Sacs-The partial rescue of the erythroid differentiation defect in Hif1␣ Ϫ/Ϫ yolk sacs by supplementation with Fe-SIH (Fig. 3B) suggested defects in iron metabolism as a contributing mechanism. To evaluate this hypothesis, we measured the level of mRNAs encoding proteins involved in iron metabolism by real time RT-PCR. As shown in Fig. 5A, the quantitative mRNA analyses revealed that TfR and frascati mRNA levels were significantly lower (10.3-and 2-fold, respectively) in Hif1␣ Ϫ/Ϫ embryos compared with wild type embryos, whereas in the yolk sac no significant difference in mRNA expression between genotypes was detected. In contrast, increased levels of mRNAs encoding ferroportin (Fpn1) and iron regulatory protein 1 (IRP1) (1.5-and 1.6-fold, respectively) were observed in Hif1␣ Ϫ/Ϫ yolk sacs, and hepcidin mRNA expression was markedly up-regulated in both Hif1␣ Ϫ/Ϫ yolk sacs and embryo (4.3-and 5.4-fold, respectively). The expression of other genes involved in iron metabolism such as Dmt1 and iron regulatory protein 2 (IRP2) showed no statistically significant difference between genotypes.
Abnormal Expression of Iron Metabolism Proteins in Hif1␣ Ϫ/Ϫ Yolk Sacs and Embryos-At E9.5, expression of Fpn1, Dmt1 and TfR was detected in yolk sacs, whereas expression in embryonic tissue was significantly lower and generally ranged near detection limits (Fig. 5B and data not  shown). Fpn1 was expressed in visceral endoderm, yolk sac mesoderm, and in endothelial cells (Fig. 5B). Compared with wild type, Fpn1 expression was consistently increased in mesodermal and endothelial cells of Hif1a Ϫ/Ϫ yolk sac. All cell types in the yolk sac, including nucleated primitive hematopoietic cells, demonstrated Dmt1 expression (Fig.   FIGURE 3. Effect of Fe-SIH on erythroid colonies. A, erythroid colonies were counted at day 8 of culture. The number of erythroid colonies derived from Hif1␣ Ϫ/Ϫ yolk sacs was significantly lower than one from wild type yolk sac. By adding Fe-SIH, the number of erythroid colonies was significantly increased in wild type yolk sac, although one from Hif1␣ Ϫ/Ϫ yolk sac shows only a modest increase. B, panels a and b, erythroid colonies from wild type yolk sac treated with PBS or Fe-SIH, respectively. Panels c and d, erythroid colonies from Hif1␣ Ϫ/Ϫ yolk sac treated with PBS or Fe-SIH, respectively. Compared with wild type control (panel a), the erythroid colonies from Hif1␣ Ϫ/Ϫ yolk sac were smaller and poorly hemoglobinized (panel c). The addition of Fe-SIH improved the size and hemoglobinization of mutant erythroid colonies (panel d), approaching the aspect of wild type colonies (panel b). Original magnification, ϫ50. * represents p Ͻ 0.05. 5B). Dmt1 expression levels were indistinguishable between wild type and Hif1␣-deficient yolk sac.
Although TfR expression was readily apparent in wild type visceral endoderm, nucleated primitive hematopoietic cells as well as mesodermal and endothelial cells demonstrated markedly lower TfR protein levels (Fig. 5B). Strikingly, TfR expression was significantly decreased in visceral endodermal cells of Hif1␣ Ϫ/Ϫ yolk sacs compared with wild type. Furthermore, consistent with the immunohistochemistry, Western blot analysis identified significantly decreased TfR protein levels in Hif1␣ Ϫ/Ϫ yolk sacs as well as in the embryo proper (Fig. 5C).

DISCUSSION
Maintenance of oxygen homeostasis is critical for multiple biological processes, including mammalian embryonic development. The appropriate response to hypoxia is required for embryonic metabolism and orderly cardiovascular development (1) and possibly the establishment of hematopoiesis (9,10). During murine embryogenesis, expression of the VEGF receptor Flk1 in yolk sac cells is essential for initiation of hematopoiesis and blood island formation (24). Furthermore, targeted disruption of VEGF revealed a dose-dependent requirement during mouse embryogenesis, because heterozygosity for the VEGF null allele caused embryonic lethality at midgestation because of impaired angiogenesis and blood island formation (25,26). Based on transcriptional regulation of VEGF by HIF-1 (27,28) and studies in HIF-1␤-deficient mice, it was proposed that hypoxia controls the formation and function of hemangioblasts as well as multilineage embryonic hematopoiesis (9). In addition, several serum factors, such as IGF-1, support hematopoiesis in vitro (23). Hif1␤ Ϫ/Ϫ embryos exhibited substantial defects in blood cells and vessel development (6). Ramirez-Bergeron et al. (10) demonstrated proper formation of mesoderm but defects in subsequent differentiation into hemangioblasts, the progenitors of endothelial and hematopoietic lineages in Hif1␤ Ϫ/Ϫ embryos. Transcript levels of mesodermal and hemangioblast markers, such as Brachyury, BMP4, and FLK1, were decreased or even absent in Hif1␤-deficient embryoid bodies following embryonic stem cell differentiation (10). These data suggested that HIF-1␤ influences the kinetics of hemangioblast formation. However, these changes are not necessarily an effect of hypoxia, because HIF-1␤ is constitutively expressed, whereas HIF-1␣ levels increase in response to decreased oxygen tension.
In this study, we studied yolk sac erythropoiesis in HIF-1␣deficient mouse embryos. Hematopoietic colony assay under conditions optimized for the growth and differentiation of multipotential hematopoietic progenitors revealed that Hif1␣ Ϫ/Ϫ yolk sacs contain these progenitors, but they display defects in their ability to form colonies in vitro. We have tested various conditions previously reported to enhance CFU-Mix in vitro colony forming ability (22,23). However, although these conditions partially enhanced the colony number for Hif1␣ Ϫ/Ϫ CFU-Mix, the in vitro colony-forming capacity of these progenitors was still decreased in Hif1␣ Ϫ/Ϫ samples. Thus, our results showed that HIF-1␣ is not essential for the formation of multipotential hematopoietic progenitors, but it plays an important role in expansion of committed erythroid progenitors and in terminal differentiation of erythroid cells. The ability of erythroid colony formation (BFU-E and CFU-E) in Hif1␣ Ϫ/Ϫ yolk sacs was significantly impaired and did not improve under various cytokine conditions. The in vitro proliferative defect appeared to be due to a partial block in expansion and terminal differentiation of the erythroid component of Hif1␣ Ϫ/Ϫ CFU-Mix colonies. Interestingly, the presence of multilineage myeloid progenitors in the Hif1␣ Ϫ/Ϫ yolk sacs contrasted with the previously reported data from HIF-1␤-deficient in vitro differentiated embryoid bodies (9). Whether this unexpected finding relates to potential experimental differences awaits further investigation.
The availability of VEGF, the HIF-1-controlled angiogenic growth factor, is critical for development of bipotential (endo- thelial/hematopoietic) hemangioblasts (29) and, hence, for the generation of downstream hematopoietic progenitors (30). VEGF/VEGFR represent known target genes of HIF-1 (15,28). Interestingly, a previous report (15) described increased VEGF mRNA levels in Hif1␣ Ϫ/Ϫ embryos compared with wild type. Our data provide additional evidence that HIF-1 regulates VEGFR expression. This was evident from decreased VEGFR1 mRNA levels in the embryo and a trend toward decreased levels in the yolk sac (Fig. 4). These results implicate impaired VEGF signaling in defective erythropoiesis in Hif1␣ Ϫ/Ϫ yolk sac. In strong support of this notion, VEGF treatment did not correct the defects in Hif1␣ Ϫ/Ϫ yolk sac erythroid progenitors cultured in vitro.
Yolk sac erythropoiesis in Hif1␣ Ϫ/Ϫ embryos was diminished but not abrogated. Erythropoietic defects caused by Hif1␣ deficiency were evident from the pallor of the embryos and the significant decrease in the number of yolk sac-derived erythroid colonies (CFU-Es and BFU-Es) with small size and poor hemoglobinization. Thus, HIF-1␣ deficiency caused hematopoietic defects downstream of multipotential myeloid progenitors, which were most pronounced in the committed erythroid progenitors. The decreased number and poor hemoglobinization of erythroid colonies were observed in all culture conditions regardless of whether they favored primitive erythroid progenitors, definitive erythroid progenitors, or CFU-Mix progenitors. Conceivably, the moderate developmental delay in Hif1␣ Ϫ/Ϫ embryos could contribute to the impaired erythropoiesis in Hif1␣ Ϫ/Ϫ yolk sacs. However, the defects in colony forming capacity and hemoglobinization during in vitro culture appear to reflect intrinsic properties of Hif1␣ Ϫ/Ϫ erythroid progenitors, which are less likely to be affected by the moderate developmental delay of the embryo.
High levels of Epo, as well as of VEGF and other cytokines, failed to rescue the defect in terminal differentiation of Hif1␣ Ϫ/Ϫ erythroid cells. HIF-1 directly regulates expression of the Epo gene (2), and a recent report showed that EpoR is also regulated by HIF-1 in vascular endothelial cells (32). This study showed that HIF-1 regulates EpoR expression in both the embryo and the yolk sac. Collectively, our results indicate that HIF-1␣ promotes yolk sac erythropoiesis by regulating at least two major hypoxia response pathways (Epo/EpoR and VEGF/VEGFR).
Interestingly, our results revealed that HIF-1 regulated Epo expression in the embryo but not in the yolk sac. These findings confirmed previous reports on the differential regulation of Epo and EpoR in yolk sac (33,34). Conceivably, compensatory up- regulation of HIF-2␣ in the yolk sac could mask the effects of Hif1␣ on Epo regulation. In support of this notion, lack of Hif2␣, a hypoxia-responsive gene related to Hif1␣, resulted in anemia and Epo deficiency in mice (31). However, normal Hif2␣ mRNA levels in the Hif1␣ Ϫ/Ϫ yolk sac and embryos render compensatory activity by this Hif1␣ homolog unlikely. Therefore, our data suggest differential regulation of Epo expression in yolk sac and embryo by HIF-1.
Previous reports suggested a role of HIF-1 in the regulation of genes involved in iron metabolism (35)(36)(37)(38). In particular, there is evidence for direct regulation of TfR by HIF-1. This study reveals that loss of function of HIF-1␣ caused a significant decrease in both TfR mRNA and protein levels. Consequently, HIF-1␣-deficient embryos are predicted to be iron-deficient because of decreased TfR-mediated iron transport in the yolk sac. Indeed, consistent with iron-limited erythropoiesis, Fe-SIH significantly improved the size of Hif1␣ Ϫ/Ϫ yolk sac erythroid colonies and their level of hemoglobinization, but not their number (Fig. 3). Interestingly, Fe-SIH also increased the size and hemoglobin content of erythroid colonies derived from wild type and Hif1␣ ϩ/Ϫ yolk sac. These results suggest that iron influx to erythroid progenitor cells might also represent a limiting factor in wild type yolk sac erythropoiesis as proposed earlier for adult erythroid cells (17).
Several iron metabolism genes, including hepcidin, IRP1, Fpn1 and frascati were differentially expressed in Hif1␣ Ϫ/Ϫ yolk sac and/or embryos. This raises the question as to whether these changes in gene expression reflect direct regulation by HIF-1 or, alternatively, compensatory mechanisms in response to primary defects in iron homeostasis caused by low levels of TfR expression. To the best of our knowledge, with the exception of TfR, there is no evidence to support a model of direct transcriptional regulation of these genes by HIF-1. Thus, we favor the possibility that these genes are differentially expressed in response to underlying defects in iron metabolism, as summarized in a model in Fig. 6.
Little is known about the regulation of embryonic iron homeostasis. Interestingly, the gene response patterns in Hif1␣ Ϫ/Ϫ yolk sac and embryo differ from predictions based on postnatal iron homeostasis. For example, the internal ribosome entry site in the 5Ј-untranslated region of Fpn1 mRNA should inhibit Fpn1 mRNA translation under low intracellular iron conditions and result in decreased Fpn1 protein levels (39 -41). However, a recent study showed that ϳ50% of the FPN1 transcripts in human erythroid cells lack the 5Ј internal ribosome entry site (42). This would render Fpn1 regulation partially independent of intracellular iron levels and, thus, might explain the moderately increased Fpn1 protein levels in the Hif1␣ Ϫ/Ϫ yolk sac. Conversely, high levels of hepcidin are consistent with high intracellular iron balance (43), and hepcidin binding to Fpn1 should result in low levels of Fpn1 expression because of internalization and lysosomal degradation (44). However, Hif1␣ Ϫ/Ϫ yolk sacs revealed increased hepcidin levels in the context of increased Fpn1 expression. Thus, to the best of our knowledge, this study provides the first in vivo evidence for regulatory differences in the hepcidin/Fpn1 homeostatic loop between prenatal and postnatal tissues.
High levels of hepcidin expression could contribute to the defects in erythropoiesis in Hif1␣ Ϫ/Ϫ yolk sac because hepcidin was shown to inhibit erythroid colony formation at low concentration of Epo in vitro (45). Interestingly, the erythropoietic defects described in this study are not due to a block in the developmental switch from embryonic to fetal/ adult erythropoiesis because HbB1 fetal/adult globin expression was detected in Hif1␣ Ϫ/Ϫ embryos and yolk sacs, whereas embryonic globin HbY mRNA levels decreased. This contrasts with a previous study (46) using embryoid bodies, which proposed a hypoxia-regulated developmental switch in globin expression.
HIF-1 plays a significant role in adult erythropoiesis, as demonstrated by the congenital up-regulation of HIF-1␣ in Chuvash polycythemia. This autosomal recessive disorder is caused by homozygosity for the VHL R200W mutation in the von Hippel-Lindau gene (14), and derives, at least in part, from an increase in serum Epo levels. The erythroid progenitors also displayed increased sensitivity to Epo under in vitro conditions, implicating erythropoiesis promoting factor(s) yet to be identified (47). In conjunction with our previous studies on Chuvash polycythemia (14), the differential response of erythropoietic cells of Hif1␣ ϩ/ϩ and Hif1␣ Ϫ/Ϫ yolk sacs to Fe-SIH-mediated cellular iron delivery implicates an unknown HIF-1␣-responsive, iron-containing factor in the promotion of erythropoiesis.
In conclusion, deficiency for HIF-1␣ differentially affects Epo/EpoR signaling during the embryo and yolk sac development. This study also revealed that HIF-1␣ plays a critical role in stimulating the survival, proliferation, and differentiation of erythroid progenitors in yolk sac. Finally, HIF-1 regulates TfR expression in the yolk sac and embryo proper, thereby providing iron for erythropoiesis during early embryonic development.