Stat5 regulates cellular iron uptake of erythroid cells via IRP-2 and TfR-1

Erythroid cell formation needs to be tightly regulated to maintain proper tissue oxygenation. Although about 10¹¹ red cells are produced every day in humans, total red cell numbers are kept within a narrow range in bone marrow, spleen, and fetal liver. While early erythroid lineage commitment is controlled by numerous transcription factors and their binding partners (like GATA-1, FOG-1, and EKLF),¹  later stage differentiation from erythroblasts to mature erythrocytes is strictly regulated by erythropoietin (Epo).² 

Epo induces dimerization of erythropoietin receptor (EpoR), which results in juxtaposition and activation of the receptor-associated Janus kinase 2 (Jak2). Jak2 subsequently phosphorylates several tyrosine residues in the cytoplasmic tail of EpoR, which in turn acts as docking sites for signaling molecules such as PI3-K³ , MAPK,⁴  PKC,⁵  and PLC-γ.⁶  A central pathway of EpoR signaling is the activation of the transcription factor signal transducer and activator of transcription 5 (Stat5).^7,8  Upon EpoR phosphorylation, Stat5 molecules are tyrosine-phosphorylated by Jak2 and translocate to the nucleus. This leads to activation of Stat5 target genes such as Pim-1, c-myc, Bcl-x_L, D-type cyclins, oncostatin M, and SOCS members that play important roles in differentiation, proliferation, apoptosis, and other processes.^9–15 

The importance of Epo signaling is evidenced by the phenotypes of Epo^−/−, EpoR^−/−, and Jak2^−/− mice, which die in utero at embryonic day (E) 13.5 due to defects in erythropoiesis. Fetal livers of Jak2^−/− animals completely lack BFU-E (burst forming unit–erythroid) and CFU-E (colony forming unit–erythroid) progenitors. In line with this, Epo- and EpoR-deficient embryos also fail to develop mature erythroid cells in vivo.^16–18  Interestingly, mice expressing a truncated form of EpoR (EpoR_H), which solely activates Stat5 and lacks all critical phosphotyrosine sites required to activate other signaling pathways, exhibited no strong erythroid phenotype,^15,19  suggesting Stat5 as a critical component of the EpoR signaling pathway. Unexpectedly, however, mice expressing an EpoR mutant additionally lacking the binding site for Stat5 (EpoR_HM) displayed no overt erythroid phenotype except under stress conditions.^15,19  Moreover, the mice initially reported to be deficient for Stat5a and Stat5b were viable and had a surprisingly mild erythroid phenotype.^7,20,21  Although showing fetal anemia, adult animals only displayed mild defects in recovery from phenylhydrazine-induced erythrolytic stress. This was explained by increased apoptosis in early erythroid progenitors, due to lack of Stat5-induced expression of the antiapoptotic gene Bcl-x_L.²¹  Later, however, these initial Stat5 knockout animals—now referred to as Stat5^ΔN/ΔN—were found to express a N-terminally truncated protein able to bind DNA and to activate a subset of Stat5 target genes.^14,22–25  In contrast to Stat5^ΔN/ΔN, the recently described bona fide Stat5a/b null knockout mice (referred to as Stat5^−/−)⁸  die during gestation or at latest (< 2%) shortly after birth. Complete ablation of Stat5 resulted in severe defects in generation of all lymphoid lineages.^23,24,26  However, an accurate analysis of the erythroid compartment from these mice is still missing.

Maturing erythroid progenitors depend on large amounts of bioavailable iron (humans require 20 mg iron complexed to transferrin per day) to enable efficient heme synthesis. Cellular uptake of iron-loaded transferrin is mediated by transferrin receptor-1 (TfR-1; also called CD71),²⁷  followed by receptor internalization. Accordingly, maturing erythroid cells express high levels of TfR-1. However, as excess iron would lead to oxidative damage, expression of proteins involved in iron uptake, storage, and utilization is tightly controlled. In case of TfR-1, this occurs primarily at the level of TfR-1 mRNA stability²⁸  and to a lesser extent at the transcriptional level.²⁹  At low-iron conditions, trans-acting iron regulatory proteins (IRP1 + 2) bind to conserved hairpin structures (iron-responsive elements [IREs]) in the 3′-untranslated region (UTR) of TfR-1 mRNA, which selectively stabilizes this mRNA and ensures proper TfR-1 cell-surface expression and iron uptake.^30–32  Upon iron excess, IRP-1 is converted to a cytosolic aconitase catalyzing isomerization of citrate to isocitrate,³³  while IRP-2 is degraded by the proteasome.³⁴  Thus, both proteins no longer bind to IREs, resulting in strongly reduced TfR-1 mRNA stability, which leads to reduced TfR-1 cell-surface expression and iron uptake.^35–37  Recently generated knockout mouse models for IRP-1 and IRP-2 suggested IRP-2 as the master regulator of IRE-regulated mRNAs, as ablation of IRP-2 led to a decrease in TfR-1 expression and microcytic anemia, while IRP-1 knockout animals had no overt phenotype.^38,39 

Here we show that Stat5^−/− embryos suffer from severe microcytic anemia, a disease mostly associated with iron deficiency and characterized by small-sized red blood cells. In Stat5-deficient animals this anemia had 2 causes: first, fetal livers were much smaller in knockout embryos due to a strong increase of apoptosis in the erythron. We demonstrate that the antiapoptotic proteins Mcl-1 and Bcl-x_L were largely absent in Stat5^−/− cells, but ectopic expression of Mcl-1 complemented the survival defect of Stat5^−/− erythroid cells. Second and more important, we demonstrate for the first time a direct link between Stat5 and iron metabolism. In the absence of Stat5, IRP-2 expression was strongly decreased, resulting in more than 2-fold lower cell-surface expression of TfR-1 and thus strongly reduced iron uptake in erythroid cells. Together, the high levels of apoptosis and impaired iron uptake caused severe microcytic anemia and probably contributed to the death of Stat5^−/− embryos.

Stat5^+/− mice⁸  were maintained under pathogen-free conditions and bred on a mixed background (C57Bl/6 × Sv129F1) to obtain Stat5^−/− embryos. For the determination of blood indices, heparinized blood was measured in a Vet ABC Blood Counter (Scil Animal Care, Viernheim, Germany). Serum Epo levels were determined using a Quantikine mouse erythropoietin enzyme-linked immunosorbent assay (ELISA) kit (R&D Systems, Minneapolis, MN) according to the manufacturer's protocol, measured on a Victor³ V 1420 multilabel counter (PerkinElmer, Waltham, MA). All animal experiments described in our manuscript were performed in accordance with Austrian and European laws and under approval of the ethical and animal protection committees of the Veterinary University of Vienna.

Fetal livers of E13.5 mouse embryos (Stat5^−/− and wild type [wt]) were resuspended in serum-free medium (StemPro-34; Invitrogen, Carlsbad, CA). In brief, for self-renewal,⁴⁰  cells were seeded into medium containing 2 U/mL human recombinant Epo (Erypo; Cilag AG, Sulzbach, Germany), murine recombinant stem cell factor (SCF; 100 ng/mL; R&D Systems), 10⁻⁶ M dexamethasone (Dex; Sigma-Aldrich, St Louis, MO), and insulin-like growth factor 1 (IGF-1; 40 ng/mL; Promega, Madison, WI). The resulting erythroblast cultures were expanded by daily partial medium changes and addition of fresh factors, keeping cell densities between 2 and 4 × 10⁶ cells/mL. Proliferation kinetics and size distribution of cell populations were monitored daily in an electronic cell counter (Casy-1; Innovatis AG, Reutlingen, Germany).⁴⁰  For retroviral infections, fetal liver erythroblasts were cocultured for 72 hours with retrovirus-producing fibroblast cell lines selected for high virus production.¹⁴  Infection efficiency was 75% to 95% as measured by flow cytometry for green fluorescent protein (GFP) expression. Cytospun cells were fixed in methanol, embedded in Entellan (Sigma-Aldrich), and stained with either Benzidine, May-Gruenwald Giemsa, or Wright-Giemsa solutions according to manufacturer's protocols (all Sigma-Aldrich). Photomicrographs were taken using an Axiovert 10 microscope (Zeiss, Jena, Germany) equipped with a 63× oil-immersion lens (numeric aperture 44-07-61), a Zeiss AxioCam MRc5, and Axiovision LE software, version 4.5. Images were processed with Adobe Photoship CS2 and are presented at ×630 magnification.

Cultured erythroblasts or single-cell suspensions of freshly isolated fetal livers were stained with fluorescence-conjugated antibodies (all from BD Biosciences, San Jose, CA) against Ter-119 (PE-conjugated) and TfR-1 (biotinylated) for in vivo erythroid development analyses. Annexin V (APC-conjugated) staining was performed according to the manufacturer's instructions. For in vivo proliferation assays, pregnant mice were injected with 80 mg BrdU per kilogram of body mass. After 1 hour, embryos were isolated, and fetal liver cells were fixed and stained with anti-BrdU–FITC plus Ter-119–PE following the manufacturer's protocol (BrdU flow kit; BD Biosciences). Samples were analyzed on a FACSCalibur flow cytometer (BD Biosciences). Where indicated, Ter-119⁺ cells were isolated by magnetic cell sorting using Auto-MACS (Miltenyi Biotec, Auburn, CA).

The following antibodies were used for Western blotting: anti–mouse ERK1/2, anti–horse ferritin H, anti–mouse actin (all from Sigma-Aldrich), anti–mouse TfR-1 (BioSource, Camarillo, CA), anti–mouse Bcl-x_L, anti–mouse PCNA, anti–mouse Stat5 (all from BD Transduction Laboratories, Erembodegem, Belgium), anti–mouse Mcl-1 (Abcam, Cambridge, United Kingdom), anti–mouse eIF4E, anti–mouse eIF2-alpha, pSer-eIF2-alpha (all from Cell Signaling Technology, Danvers, MA), anti–rat IRP-1,⁴¹  and anti IRP-2.³⁹ 

Chromatin immunoprecipitation was performed as in LeBaron et al.⁴²  A total of 2 × 10⁷ primary erythroblasts were stimulated with 10 U Epo/mL for 30 minutes. Cells were fixed with 1% formaldehyde for 30 minutes. DNA was sonicated using a Bandelin Sonoplus GM70 sonicator (cycle count, 30%; power, 45%; 6 × 30 seconds; Bandelin Electronic, Berlin, Germany). DNA fragments were recovered using anti–mouse Stat5ab (C20; Santa Cruz Biotechnology, Santa Cruz, CA) or anti–mouse Stat5 (N20; Santa Cruz Biotechnology). Recovered DNA fragments were directly used for polymerase chain reaction (PCR) analysis.

RNA was isolated using the RNeasy Mini Kit (QIAGEN, Valencia, CA). RNA integrity was checked with the Agilent Bioanalyzer (Agilent, Palo Alto, CA). A total of 2.5 μg RNA was reverse transcribed using Superscript II reverse transcriptase (Invitrogen). Real-time PCR was performed on an Eppendorf Master-cycler RealPlex using RealMasterMix (Eppendorf, Hamburg, Germany) and SYBR Green. Results were quantified using the Delta Delta C(T) method.⁴³ 

Statistical analyses were performed using Excel 2004 (Microsoft, Redmond, WA). The Student t test was used to calculate P values (2-tailed). P values of .05 or less are indicated in the figures by one asterisk; P values of .01 or less are indicated in the figures by 2 asterisks. Data are presented as mean values plus or minus standard deviation (SD).

Embryos lacking the entire Stat5 locus (ie, Stat5a/b) die during gestation or at latest perinatally (approximately 99%) with severe defects in diverse hematopoietic lineages.^23,24  The previously demonstrated pivotal role of Stat5 in Epo signaling^19–22  prompted us to analyze the function of Stat5 in erythropoiesis in detail. E13.5 Stat5^−/− embryos and newborn Stat5^−/− animals appeared paler than their wt littermates, particularly in the fetal liver region (Figure 1A). The relative abundance of erythroid cells in Stat5^−/− fetal livers (approximately 80% of all cells) remained unchanged, as determined by staining for the panerythroid marker Ter-119, followed by flow cytometry (Figure 1B left). However, the size of Stat5^−/− fetal livers in E13.5 embryos was visibly reduced and total fetal liver cellularity was decreased by 50% (n = 6), corresponding to a similar reduction in the total number of erythroid cells (Figure 1B right). Since anemia causes elevation of Epo levels⁴⁴  to counteract hypoxia, sera from E16.5 or newborn wt and Stat5^−/− animals were analyzed for Epo levels. These were highly elevated in Stat5^−/− versus wt embryos (3.8 ± 0.6-fold; n = 5); newborn animals showed an even higher elevation (35.2 ± 5.1-fold; n = 5; Figure 1C). This strongly suggested that Stat5-deficient animals suffer from severe anemia. To determine the specific type of anemia, blood from E16.5 wt and Stat5^−/− animals was analyzed (Table 1). Red blood cell counts of E16.5 Stat5^−/− embryos were lowered to 1.0 plus or minus 0.3 × 10⁶/mm³, in contrast to 2.4 plus or minus 0.4 × 10⁶/mm³ in wt embryos. In line with this, hematocrit (Hct) of E16.5 Stat5^−/− embryos was reduced to 9.8% plus or minus 0.7% compared with 31.4% plus or minus 1.8% in wt embryos. Likewise, mean corpuscular volume (MCV), hemoglobin content (Hgb), and mean corpuscular hemoglobin (MCH) of E16.5 Stat5^−/− blood was strongly reduced. These effects also were clearly visible in blood smears, showing hypochromic microcytic erythrocytes (Figure S1A,B, available on the Blood website; see the Supporting Materials link at the top of the online article).

An additional cause for early lethality and high serum Epo levels could have been a lung defect leading to reduction in red cell oxygenation. Analysis of tissue sections from wt and Stat5^−/− newborn animals, however, did not reveal any histologic differences (data not shown). Taken together, Stat5^−/− mice suffer from microcytic anemia.

Hypomorphic Stat5^ΔN/ΔN mice displayed enhanced erythroid cell death, which was attributed to reduced expression of the antiapoptotic protein Bcl-x_L.^20,21  To determine apoptosis in Stat5^−/− mice, freshly isolated fetal liver cells were stained for Ter-119 and apoptosis assessed by annexin V staining. In Stat5^−/− embryos, the frequency of annexin V⁺ cells was enhanced more than 2-fold, regardless of developmental stage (Figure 2A). In line with this, Stat5^−/− fetal liver cells showed a greater than 6-fold reduction of erythroid colony numbers in CFU-E assays regardless of Epo concentrations, supporting the notion that lack of functional Stat5 reduces cell survival (Figure 2B). In contrast, no significant alterations were observed in BFU-E assays (Figure S1C).

To assess potential differences in viability of Stat5-deficient erythroid cells in an adult (bone marrow) versus an embryonic (fetal liver) microenvironment, short-term transplantation experiments were performed. Equal amounts of sorted GFP-expressing wt- or Stat5^−/− proerythroblasts cultured from fetal livers (c-Kit⁺/TfR-1^high/Ter-119⁻ cells¹⁵ ) were injected into lethally irradiated wt recipients (Figure 2C). This setup was chosen to circumvent any influence of the well-known repopulation defect of Stat5-deficient hematopoietic stem cells.^45–47  At 3 days after transplantation, bone marrow cells were harvested and scored for GFP⁺ mature Ter-119⁺ erythroid cells, as previously reported.⁴⁸  In line with the preceding experiments, a 5-fold reduction in abundance of transplanted mature Stat5^−/− versus wt cells was determined in the adult microenvironment (bone marrow; Figure 2D). Moreover, these data indicated a cell-autonomous survival defect of Stat5^−/− erythroid cells.

Consistent with the apoptotic phenotype described, Ter-119⁺ cells from Stat5^−/− fetal livers showed decreased levels of Bcl-x_L as compared with wt cells (Figure 3A left). Nevertheless, Bcl-x_L protein levels in Stat5^−/− cells were still approximately 50% of that of wt. This prompted us to analyze the expression of other antiapoptotic Bcl-2 family members. We focused on Mcl-1 for 3 reasons: (1) Mcl-1 is up-regulated during early erythroid commitment in human cells⁴⁹ ; (2) its bone marrow–specific ablation reduces blood formation⁵⁰ ; and (3) it appears to be regulated by Stat5.^51–53  Indeed, Mcl-1 protein and mRNA expression were drastically reduced in Stat5-deficient cells (Figure 3A right; Figure 3B). To assess whether Mcl-1 is an Epo-inducible Stat5-regulated gene, primary wt and Stat5^−/− erythroblasts were factor-deprived for 2 hours and subsequently restimulated with Epo for 30 minutes. Quantitative PCR (qPCR) revealed a 3.5-fold increase in Mcl-1 mRNA in Epo-stimulated cells, which was abrogated in Stat5-deficient cells (Figure 3C).

To test whether exogenous Mcl-1 provides protection against apoptosis to erythroid cells, primary wt or Stat5^−/− fetal liver cells were transduced with retroviral constructs encoding GFP, Bcl-x_L, or Mcl-1. Erythroblasts were cultivated for 48 hours in the presence or absence of Epo, and apoptosis was determined by flow cytometry. Ectopic expression of either Mcl-1 or Bcl-x_L completely prevented apoptosis of wt as well as Stat5 knockout erythroblasts upon Epo withdrawal (Figure 3D).

The decrease of fetal liver size and cellularity in Stat5^−/− embryos could also have been due to reduced proliferation of erythroid cells, as suggested by the known ability of Stat5 to enhance expression of proliferation-promoting genes such as c-Myc, Cyclin D2 and D3, or oncostatin M.^14,23,54,55  Cell division kinetics of erythroid cells in vivo and in vitro, however, were similar in Stat5^−/− and wt cells (Figure S2; Document S1).

Taken together, Stat5-deficient fetal liver erythroid cells were massively apoptotic. This effect could be attributed to reduction of Bcl-x_L levels together with complete loss of Mcl-1, translating into massive decrease of fetal liver cellularity.

To analyze whether Stat5-deficient mice had a defect in erythroid lineage commitment, wt and Stat5^−/− fetal livers were analyzed for the presence of megakaryocytic-erythroid progenitors^56,57  (MEPs), the first erythroid-committed progenitor detectable by flow cytometry. Interestingly, we observed a 2-fold increase of the MEP compartment in Stat5-deficient fetal livers (Figure S3), suggesting a compensatory attempt to counteract the increased erythroid cell death during definitive erythropoiesis.

To determine whether the anemia in Stat5^−/− embryos was due to a defect in erythroid differentiation, fetal liver cells were analyzed for erythroid markers Ter-119 and TfR-1. This combination allows staging of maturing erythroid cells from immature progenitors (TfR-1^low Ter-119^low) over an intermediate stage (TfR-1^high Ter-119^high) to late orthochromatophilic erythroblasts (TfR-1^neg Ter-119^high; Figure 4A left; gates R1 to R5, increasing maturity⁵⁸ ). Stat5-deficient and wt fetal livers contained cells of all differentiation stages at indistinguishable frequencies (Figure 4A). For detailed morphologic analysis, wt and Stat5^−/− fetal liver cells were sorted according to their cell-surface marker phenotype (R2-R5), spun onto glass slides, and subsequently stained with either May-Grunwald Giemsa or Benzidine-Wright Giemsa (Figure S4). No apparent morphologic differences in maturity between wt and Stat5^−/− cells were observed. Thus, the reduction in fetal liver cellularity (Figure 1B) was probably not due to differentiation arrest at a distinct step of maturation. We did, however, observe a reproducible decrease in TfR-1 cell-surface expression, which prompted us to align the gating strategy accordingly.

Accumulation of hemoglobin is the hallmark of terminal erythropoiesis, requiring an enormous up-regulation of iron intake via increased expression of TfR-1. Quantification of TfR-1 levels in Stat5-deficicient versus wt cells by flow cytometry revealed a greater than 2-fold reduction in knockout cells (Figure 4B). This was confirmed at the mRNA level (Figure 4C) and further corroborated by Western blot analysis of wt, Stat5^+/−, and Stat5^−/− fetal liver cell lysates (Figure 4D,E). A recent report described functional Stat5-binding sites (interferon-gamma–activated sequence [GAS] elements) in the first intron of the TfR-1 gene, using an erythroleukemic cell line expressing a constitutively active Stat5 variant.⁵⁹  To corroborate these data, we decided to analyze DNA binding of endogenous Stat5 to these elements in primary fetal liver erythroblasts after Epo stimulation by chromatin immunoprecipitation (Figure 4F). Indeed, DNA binding of Stat5 to all 3 sites analyzed was confirmed and apparently resulted in an Epo-induced increase of TfR-1 mRNA as quantified by qPCR (Figure 4G). As expected from the well-known inverse relation in expression of TfR-1 to the iron-storage protein ferritin,³⁰  Stat5^−/− cells showed elevated levels of ferritin protein (Figure S5).

As a direct consequence of reduced TfR-1 cell-surface expression, we observed a significant reduction of intracellular iron (approximately 40%) in freshly isolated Stat5^−/− fetal livers as measured by atomic absorption spectrometry (Figure 4H), further supporting the idea of altered iron metabolism in Stat5^−/− cells. Reduced iron availability leads to a drop in heme synthesis,⁶⁰  known to result in activation of heme-regulated inhibitor (HRI).⁶¹  This kinase, via inactivation of translation initiation factor eIF2-alpha, throttles expression of globins to ensure that heme, alpha-globin, and beta-globin are always synthesized at the appropriate ratio of 4:2:2.⁶¹  To test whether this regulatory circuit was disturbed in Stat5^−/− erythroid cells, an abundance of globin mRNAs in lysates from erythroid cells sorted out of wt or Stat5^−/− fetal livers were analyzed by qPCR. Indeed, relative levels of both globin mRNAs were significantly reduced in Stat5^−/− cells; moreover, eIF2-alpha showed the expected increase in phosphorylation (Figure S6). In summary, these data demonstrated that Stat5^−/− erythroid cells were severely iron-deficient.

Stabilization of TfR-1 mRNA by binding of IRP-1 and IRP-2 is considered the predominant mechanism to satisfy the iron demand of proliferating cells.^30,31  A possible activation of IRPs by Epo has been discussed.^62,63  Accordingly, Western blot analyses for IRP-1 and IRP-2 from lysates of wt and Stat5^−/− primary erythroblasts revealed a striking, 5-fold down-regulation of IRP-2 in Stat5 knockout cells (Figure 5A,B right), accompanied by a 2-fold up-regulation in IRP-1 expression. IRP-1 and IRP-2 mRNA levels were similarly changed in Stat5-deficent cells (data not shown). Determination of IRP-2 RNA binding in Stat5^−/− cells using in vitro–transcribed, radioactively labeled IRE probes in electrophoretic mobility shift assays (EMSAs; Figure 5C) showed a similar decrease of IRP-2 activity (Figure 5D). Given the important role of IRP-2 in TfR-1 expression in erythroid cells,^38,39  these data strongly suggested that the decrease of IRP-2 was an additional cause for reduction of TfR-1 cell-surface expression.

To test a possible direct role of Stat5 in regulating IRP-2 expression, we analyzed the IRP-2 promoter in detail. A region 1030 to 1100 bp upstream of the transcriptional start site contained one perfect Stat5 DNA-binding site (TTCN₃GAA)⁶⁵  plus 2 adjacent low-affinity Stat5 response elements with a mismatch in one half-site of the inverted repeat (Figure 6A). Annotated Stat5 sites^24,65  together with the IRP-2 sequence I suggested that the latter fulfilled bioinformatic criteria for perfect Stat5 binding (Figure 6B).

To test whether there was a direct transcriptional induction of IRP-2 by Stat5, the 2 kb fragment of the IRP-2 promoter upstream of the predicted transcription start site, comprising all 3 putative Stat5 response elements (REs), was inserted into a firefly-luciferase reporter gene construct. 293T cells were cotransfected with Stat5a and EpoR cDNAs together with the respective reporter construct. Transfected cells were stimulated with Epo or left untreated. Epo-treated cells displayed a significant increase of luminescence over untreated controls (Figure 6C). Direct Epo-induced expression of endogenous IRP-2 and oncostatin M (a bona fide Stat5 target gene) was demonstrated in murine erythroid leukemia cells: following 3 hours of factor deprivation, cells were restimulated for 1 hour with Epo and mRNA expression levels determined by qPCR. Epo stimulation induced expression of IRP-2 as well as oncostatin M about 3-fold (Figure 6D).

To further substantiate that IRP-2 is a direct target of Stat5, EMSAs were performed. 293T cells were cotransfected with constructs encoding EpoR and murine Stat5a. Transfected cells were Epo-stimulated or left untreated. Extracts were subsequently subjected to EMSAs using radiolabeled oligonucleotides encompassing either the newly identified Stat5-RE I of the IRP-2 promoter or a well-described Stat5-RE probe from the bovine β-casein promoter as positive control. For supershifts, a serum directed against the C-terminus of Stat5 was added to the oligonucleotide/lysate mixture. Stat5-DNA complexes were clearly evident in Epo-stimulated extracts, using both the IRP-2-I or the β-casein probe, as these complexes were readily super-shifted upon addition of anti-Stat5 serum (Figure 6E). Similar results were obtained using Epo-stimulated cells transfected with murine Stat5b (data not shown).

To test whether Stat5 recognizes one of these putative DNA-binding sites in vivo, we finally performed chromatin-immunoprecipitation (ChIP) experiments using 2 different antisera directed against N- or C-terminal epitopes in Stat5. PCR analysis of immunoprecipitated Stat5-DNA complexes from Epo-stimulated primary wt erythroblasts yielded a PCR product representing Stat5-binding sites in the IRP-2 promoter in both specific Stat5 ChIPs, but not in control IgG ChIP experiments (Figure 6F).

Together, these results indicated that Stat5 is directly involved in the control of TfR-1 transcription as well as in the modulation of its mRNA stability by regulating expression of IRP-2.

In this paper cooperating mechanisms underlying the erythroid defect leading to microcytic anemia in Stat5^−/− mice were uncovered, demonstrating a novel direct link between the EpoR-Stat5 axis and regulation of iron metabolism in vivo. First, Stat5^−/− fetal livers showed reduced cellularity due to massively enhanced apoptosis of maturing erythroid cells, apparently caused by defective expression of the antiapoptotic genes Mcl-1 and Bcl-x_L. Second, Stat5^−/− erythroid cells exhibited reduced expression of IRP-2 and TfR-1, resulting in a large decrease of TfR-1 cell surface expression, iron uptake, and globin synthesis. Together, these mechanisms appear to be sufficient to explain the severe anemia of Stat5^−/− animals.

None of the conditional Stat5 knockout models created so far in multiple cell types such as hemangioblasts (Stat5fl/fl Tie2-Cre),⁵⁹  B cells (CD19-Cre),⁶⁶  T cells (CD4-Cre, Lck-Cre),^23,24  hepatocytes (albumin-Cre, albumin-alpha-fetoprotein-Cre),^25,67  pancreatic β cells/hypothalamus (Rip-Cre),⁶⁸  endocrine/exocrine pancreas progenitors (Pdx1-Cre),⁶⁸  or skeletal muscle (Myf5-Cre)⁶⁹  die during fetal development. In contrast, ablation of Stat5 in the entire organism resulted in mortality⁸  during gestation or at latest shortly after birth. Since Epo^−/−, EpoR^−/−, and Jak2^−/− mice all die in utero at E13.5 due to defects in definitive erythropoiesis and given the prominent role of Stat5 in EpoR signaling, it was unexpected that a few Stat5^−/− embryos developed to term.

There are several possible explanations for the discrepancy in phenotypes. First, we detected high levels of pY-Stat1 and pY-Stat3 in Stat5-deficient cultivated primary erythroblasts as well as in lysates from freshly isolated fetal livers but not in wt counterparts (M.A.K., F.G., unpublished data). This is in line with increased pY-Stat1 and pY-Stat3 levels found upon liver-specific Stat5 deletion.^67,70,71  Since Stat3 and Stat5 response elements are similar,⁶⁵  increased activation of Stat3 might partially compensate for loss of Stat5. Second, the anemia of Stat5^−/− embryos led to a compensatory elevation of Epo levels in the serum, which was most pronounced in the few newborn animals. This might contribute to prolonged survival mediated by hyperactivation of Stat5-independent EpoR signaling. Third, Stat5-deficient erythroid cells exhibited elevated levels of phosphorylated eIF2-alpha, indicative for an active “integrated stress response” (ISR), presumably via the kinase HRI.⁷²  In mouse models for the red blood cell disorders erythropoietic protoporphyria and β-thalassemia, ablation of HRI exacerbated the phenotype of these diseases.⁷²  Thus, the modulation of translational efficiency to balance heme and globin production could represent another protective mechanism accounting for the “mild” erythroid phenotype of Stat5^−/− animals. Nevertheless, the ultimate reason for the early death of the animals remains to be determined.

Earlier studies addressing the role of Stat5 (ie, Stat5a and Statb) in erythropoiesis were performed with Stat5^ΔN/ΔN animals. These mice are born, viable, and show only a mild erythroid phenotype.^20,21 Stat5^ΔN/ΔN animals express a N-terminally truncated Stat5, which still activates target genes.²²  Here, we used Stat5^−/− mice lacking the entire Stat5a/b locus.⁸  Animals lacking other components of Epo signaling upstream of Stat5 (Epo, EpoR, or Jak2),²  all die in utero around E13.5 due to a block in definitive erythropoiesis. If Stat5 was the only crucial target of this pathway, full Stat5 knockout animals should show a similarly severe phenotype. Indeed, Stat5-deficient animals display erythroid defects and die at latest after birth. Although Stat5 is essential for differentiation of other hematopoietic lineages like maturation of pre-pro– to pro-B cells,^23,24  or in formation of FoxP3⁺ regulatory T cells,²⁶  the observed block in erythroid maturation in vivo was not complete. The presence of erythroid cells at all developmental stages in Stat5^−/− embryos strongly argued against an absolutely essential function of Stat5 in erythroid development. Nevertheless, there were defects in hemoglobinization of Stat5^−/− erythroid cells, which may have several causes. For instance, it could decrease through direct defects in the erythroid differentiation program and/or through a secondary response to iron deficiency.

The most striking observation in the peripheral blood morphology of Stat5^−/− animals was an apparent microcytic hypochromic anemia. This type of anemia, characterized by decreased mean corpuscular volume and reduced mean cell hemoglobin, is frequently associated with iron deficiency. Thus, it was tempting to investigate the molecular players involved.

The normal adaptive response to low iron is the well-characterized feedback regulation that increases TfR-1 mRNA stability upon binding of IRPs to its 3′-UTR (Figure 7). In Stat5^−/− cells, this response apparently was impeded as delineated from the reduced expression of TfR-1, which in turn was the direct result of decreased IRP-2 protein levels. This mechanistic link was further substantiated by the measured reduction in total intracellular iron in Stat5^−/− fetal livers, finally resulting in decreased globin mRNA expression. It remained unclear, however, whether a connection between Stat5 and IRP-2 expression existed. The promoter region of the IRP-2 gene contains 3 adjacent potential binding sites for Stat5, which indeed turned out to be functional. Moreover, qPCR showed reduced IRP-2 mRNA abundance in the absence of Stat5. Thus, one could envision a chain of events in Stat5-deficient erythroid cells, starting with decreased IRP-2 and TfR-1 expression, resulting in a net decrease of TfR-1 mRNA stability and abundance, followed by diminished TfR-1 surface expression. The consequence would be insufficient iron uptake (even in iron-depleted cells), ultimately leading to decreased heme synthesis, activation of the integrated stress response pathway and reduced globin mRNA translation. Interestingly, no functional compensation for low IRP-2 levels by the highly homologous IRP-1 protein was observed. This is in line with in vivo data from corresponding IRP-1 or IRP-2 knockout animals,^38,39  which indicated that IRP-2 is the predominant regulatory factor modulating TfR-1 mRNA stability. While ablation of IRP-1 produced no overt phenotype, loss of IRP-2 resulted in hypochromic microcytic anemia due to reduced TfR-1 expression, a phenotype reminiscent to the one of Stat5 knockout animals described here. Accordingly, lowering the expression of TfR-1 by 50% led to a similar phenotype, as TfR-1^+/− mice also displayed the same type of anemia.⁷³  It should be mentioned, however, also other conditions are known to result in microcytosis, including ablation of the genuine Stat5 target Pim-1.⁹ 

Involvement of Stats in iron metabolism might even be a more general mechanism. Hepcidin,^74,75  the dominant regulator of dietary iron absorption in enterocytes and iron release from macrophages, is a direct Stat3 target gene^74,75 : upon infection, the inflammatory cytokine IL-6 promotes hepcidin expression via Stat3, trapping iron in macrophages, resulting in decreased plasma iron concentrations. Hepcidin expression is decreased by hypoxia and anemia, directly responding to increased Epo levels.^74–76  Thus, its regulation in the anemia resulting from Stat5 deficiency may be of interest in future studies.

In erythropoiesis, up-regulation of Bcl-x_L was found to be defective in Stat5^ΔN/ΔN erythroid cells.^20,21  Other studies, however, suggested that Bcl-x_L prevents apoptosis only of late-stage erythroblasts,^11,77  but not directly via EpoR.⁷⁷  Upon readdressing this question in mice that are fully devoid of Stat5, we observed a reduction in Bcl-x_L levels of about 50% in fetal liver erythroid cells. Furthermore, Mcl-1 expression in Stat5^−/− erythroid cells was analyzed, based on the finding that this Bcl-2 gene family member could be a Stat5 target gene.^52,53,78  Indeed, Mcl-1 was completely absent in Stat5^−/− fetal liver cells, whereas Epo stimulation of wt primary erythroblasts led to a 3.5-fold increase of Mcl-1 mRNA. Furthermore, reintroduction of Mcl-1 or Bcl-x_L completely prevented apoptosis of wt as well as Stat5 knockout erythroblasts upon Epo withdrawal.

Besides the finding that Mcl-1 is a Stat5-dependent Epo target gene, down-regulation of Mcl-1 in Stat5^−/− erythroid cells could also occur through an additional mechanism, which is induced through iron deficiency in Stat5 knockout animals. Recently, it was suggested that Mcl-1 levels decrease after activation of the phospho-eIF2-alpha–mediated ISR pathway already mentioned.^79–81  eIF2-alpha phosphorylation can be induced by different kinases in response to several stress stimuli,⁸²  including HRI. Iron deficiency activates HRI, which in turn phosphorylates eIF2-alpha on its inhibitory Ser51, resulting in global reduction of mRNA translation,⁶¹  which immediately affects Mcl-1, as it is a highly unstable protein.⁸³  The observed reduced iron levels, together with elevated eIF2-alpha phosphorylation in Stat5^−/− primary erythroblasts, suggested an active ISR in Stat5^−/− cells. Hence, loss of Stat5 could lead to a direct decrease of Mcl-1 mRNA, but alternatively also to a down-regulation of Mcl-1 protein due to iron deficiency–induced ISR. Taken together, the apoptosis in Stat5^−/− fetal livers most probably reflects a composite effect of reduced levels of Bcl-x_L and loss of Mcl-1.

This contribution should help to clarify the long-discussed role of Stat5 in erythropoiesis in vivo: We identify Stat5 as a key factor regulating erythroid iron metabolism in vivo and, additionally, link the antiapoptotic machinery of erythroid cells with their iron uptake system.

We thank L. 6ennighausen for providing Stat5^−/− mice, G. Stengl for FACS sorting, G. Litos for technical assistance, and M. von Lindern for critical reading of the manuscript. We are grateful to R. S. Eisenstein and T. Roault for antibodies against IRP-1 and IRP-2, respectively.

This work was supported by grants WK-001 (to F.G. and M.A.K.) and SFB F028 (to H.B., R.M., and E.W.M.) from the Austrian Research Foundation, Fonds zur Föderung der wissenschaftlichen Forschung (FWF), and the Herzfelder Family Foundation (to E.W.M.).

Contribution: M.A.K., F.G., H.G., M.S., M.A., and B.K. conducted experiments; M.A.K. and E.W.M. designed experiments and interpreted results; H.B. contributed essential reagents and worked on the draft of the manuscript; and M.A.K., R.M., and E.W.M. wrote the paper.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Ernst W. Müllner, Max F. Perutz Laboratories, Department of Medical Biochemistry, Medical University of Vienna, Dr. Bohr-Gasse 9, A-1030 Vienna, Austria; e-mail: ernst.muellner@meduniwien.ac.at.