Growth Factor Independence 1b (Gfi1b) Is Important for the Maturation of Erythroid Cells and the Regulation of Embryonic Globin Expression

Growth factor independence 1b (GFI1B) is a DNA binding repressor of transcription with vital functions in hematopoiesis. Gfi1b-null embryos die at midgestation very likely due to defects in erythro- and megakaryopoiesis. To analyze the full functionality of Gfi1b, we used conditionally deficient mice that harbor floxed Gfi1b alleles and inducible (Mx-Cre, Cre-ERT) or erythroid specific (EpoR-Cre) Cre expressing transgenes. In contrast to the germline knockout, EpoR-Cre mediated erythroid specific ablation of Gfi1b allows full gestation, but causes perinatal lethality with very few mice surviving to adulthood. Both the embryonic deletion of Gfi1b by EpoR-Cre and the deletion in adult mice by Mx-Cre or Cre-ERT leads to reduced numbers of erythroid precursors, perturbed and delayed erythroid maturation, anemia and extramedullary erythropoiesis. Global expression analyses showed that the Hba-x, Hbb-bh1 and Hbb-y embryonic globin genes were upregulated in Gfi1b deficient TER119+ fetal liver cells over the gestation period from day 12.5–17.5 p.c. and an increased level of Hbb-bh1 and Hbb-y embryonic globin gene expression was even maintained in adult Gfi1b deficient mice. While the expression of Bcl11a, a regulator of embryonic globin expression was not affected by Gfi1b deficiency, the expression of Gata1 was reduced and the expression of Sox6, also involved in globin switch, was almost entirely lost when Gfi1b was absent. These findings establish Gfi1b as a regulator of embryonic globin expression and embryonic and adult erythroid maturation.


Introduction
The continuous process of hematopoiesis initiating from pluripotent hematopoietic stem cells and giving rise to all hematopoietic lineages compensates for the restricted life span of mature blood cells. Each terminally differentiated blood cell is the result of chronological steps of proliferation and differentiation, which are stringently controlled by underlying lineage specific and ubiquitously expressed transcription factors. The DNA binding repressors of transcription growth factor independence 1b (GFI1B) and its paralogue GFI1 are expressed in a complementary and partially overlapping manner in hematopoietic stem cells and several hematopoietic lineages as well as cells of the sensory and nervous systems [1][2][3]. Although knockout mutants for both proteins in mice resulted in different hematopoietic phenotypes [4][5][6][7][8][9][10][11], GFI1B can functionally replace GFI1 throughout the hematopoietic system, but not in sensory cells such as the inner ear hair cells [2].
Repression of transcription by Gfi1 or Gfi1b fully depends on its N-terminal Snail/Gfi (SNAG) domain, which enables the recruitment of the GFI1/GFI1B cofactors Lysine (K)-specific demethylase 1A (LSD1/KDM1A) and CoREST/Rcor1. Consequently, a knockdown of LSD1 has been shown to cause a phenotype reminiscent of Gfi1b or Gfi1 knockout phenotypes affecting HSCs, granulopoiesis, erythropoiesis and platelet production [36]. The function of the GFI1B/LSD1/CoREST complex in erythroid proliferation and differentiation was intensively studied [37,38]. Interestingly, the GFI1B/LSD1/CoREST complex binds to the Meis1 promoter in erythroid cells, but not in megakaryocytes, despite the fact that it is highly expressed in both cell types, suggesting a functional difference of Gfi1b between the two lineages.
Germline deletion of Gfi1b in mice causes lethality at around day 14.5 of embryonic development, probably due to a combined phenotype of inappropriate erythropoiesis and severe bleeding caused by a failure to produce platelet-generating megakaryocytes [1,6]. However, other not yet discovered mechanisms may also play a role. This early lethality of Gfi1b deficient mice restricted all analyses to either prenatal hematopoiesis or to cell culture systems. The recent generation of conditional Gfi1b knockout mice [9] allowed us to perform a more specific analysis of pre-and postnatal function of Gfi1b in erythropoiesis. We inactivated the Gfi1b gene by crossing conditional Gfi1b knockout mice with EpoR-Cre knock-in mice to delete specifically in the erythroid lineage or by inducibly ablating it in adult mice using Mx-Cre and Rosa-Cre-ERT mouse lines. Our results show that Gfi1b is required for the differentiation from pro-erythroblasts to mature erythrocytes and for the silencing of globin genes during embryonic development and at adult stages.

Ethics Statement
The protocols for the in vivo experiments described here were reviewed and approved by the IRCM Animal Care Committee Figure 1. The erythroid specific knockout of Gfi1b causes perinatal lethality and a delayed maturation of fetal erythroid cells. A: Typical situs of either wild type (wt) or specific knockout of Gfi1b in the erythroid lineage (EpoR-Cre, Gfi1b fl/fl ). While germline deletion of Gfi1b results in anemic, pale embryos with severe hemorrhage, the erythroid lineage specific knockout of Gfi1b results in pale embryos but not in hemorraghe (right). B: Bar graph illustrating the percentage of expected EpoR-Cre transgenic Gfi1b fl/fl embryos found viable at different stages of development as indicated. The erythroid specific ablation of Gfi1b does not reduce the survival rates of mouse embryos until birth, but leads to perinatal lethality with few exceptions (two out of 129 at age 6-weeks). The total numbers of embryos (N) analyzed at each stage are indicated at the bottom. Pregnant females were humanly euthanized according to procedures approved by the Canadian Council on Animal Care (CCAC) at indicated gestational time points and embryos were taken for analysis. New born pups that survived until a few hours after birth were examined as soon as possible for signs of anemia (paleness), weakness and difficulty breathing (endpoints) and those showing such signs were humanly euthanized for analysis following procedures approved by the Canadian Council on Animal Care (CCAC). The two mice that survived to adulthood were monitored semiweekly but never showed any sign of distress and were perfectly healthy. These mice were humanly euthanized after 6 weeks for analysis. C: Flow cytometry of isolated fetal liver cells at 14.5 dpc from wt and erythroid specific Gfi1b-KO embryos using antibodies against TER119 in combination with the developmental markers CD71, cKIT and CD9 as well as GFP for the detection of the GFP-CRE fusion protein. GFP-Cre is expressed early in erythroid development before cells become TER119 + (TER119/GFP-CRE panel). Gfi1b-KO embryos show an accumulation of TER119-low, CD71-low, cKIT-high, CD9-high erythroid precursors indicating a delayed maturation of erythrocytes. FACS plots are representative for at least 6 or more independent samples analyzed. doi:10.1371/journal.pone.0096636.g001  (ACC); protocol numbers are: #2009-12/#2013-04. All animal experiments were conducted according to institutional rules put in place by the IRCM ACC, which follow the regulations and requirements of the Canadian Council on Animal Care (www. ccac.ca).

Mice
The generation of Gfi1b 2GFP knock-in mice and Gfi1b fl/fl conditional knockout mice has been described previously [1], [9]. All mice were housed under specific pathogen-free conditions and institutional animal ethics committees reviewed animal experimentation protocols and certified animal technicians regularly observed the mice in sign of distress. Adult mice were sacrificed by carbon dioxide inhalation whereas newborn pups were euthanized by decapitation following anesthesia by carbon dioxide inhalation as per standard operating procedure approved by the IRCM ACC and the CCAC. All efforts were made to minimize the number of animals used and to reduce their suffering. All mice were backcrossed with C57BL/6 mice for at least 8 generations. No phenotype or differences in number of cells was observed for Gfi1b fl/fl or Gfi1b KO/fl mice. Rosa-Cre-ERT mice were obtained from The Jackson Laboratory (Bar Harbor, Maine, USA), strain B6;129-Gt(Rosa)26Sortm1(cre/ERT)Nat/J Stock Nr. 004847. The generation of EpoR-Cre knock-in mice used for erythroid specific ablation of Gfi1b expression has been described previously [39].

Flow cytometry, cell sorting, microarray analysis and Q-PCR
Hematopoietic cell populations were analyzed by flow cytometry using an LSR (BD Biosciences) and sorted using a MoFlo (Cytomation). Cells were passed through a 23-gauge needle, filtered through a cell strainer and resuspended in PBS (1% FCS, 10 mM EDTA). 1-5 X 10 6 cells were stained with antibodies at a 1:200 concentration for 20 min, washed with PBS and measured or sorted immediately. Antibodies used were ordered from BD-Biosciences (Missisauga, ON, Canada) or Bio-Legend (San Diego, CA, USA).

Erythroid specific ablation of Gfi1b by EpoR-Cre causes accumulation of immature erythroid cells and perinatal lethality
We generated a cell type specific knockout of Gfi1b using EpoR-Cre mice, which express a GFP-Cre recombinase fusion protein in immature erythroid cells ( Figure 1A, Figure S1), which also allows monitoring the expression of the Cre-transgene by measuring green fluorescence. EpoR-Cre Gfi1b fl/fl mice did not show internal bleeding at stage E14.5 ( Figure 1A) or at birth, but appeared pale compared to controls ( Figure 1A, Figure S1A). Most EpoR-Cre Gfi1b fl/fl mice died within minutes after birth, but a few survived to adulthood ( Figure 1B, Figure S1A). EpoR-Cre Gfi1b fl/fl fetal livers (E14.5) appeared similar to wild type controls ( Figure S1C), but showed an accumulation of CD71 2 TER119 2 erythroid precursor cells that are c-Kit + and GFP + (i.e. express the EpoR-Cre transgene) and a decrease of CD71 + TER119 + erythroblasts ( Figure 1C), suggesting that deletion of Gfi1b delays the differentiation of embryonic erythroid cells.

Acute disruption of Gfi1b in adult mice affects erythroid differentiation and causes anemia
To investigate the role of Gfi1b in adult erythroid development, we used Gfi1b fl/fl mice expressing the inducible Mx-Cre transgene   [42] and injected them with pIpC ( Figure 2A) to induce the recombination of the Gfi1b alleles throughout the hematopoietic system. Deletion of the floxed Gfi1b alleles in bone marrow, spleen and in FACS sorted TER119 + cells was substantial but still incomplete ( Figure S2). Nonetheless, Gfi1b knockout mice showed a relative increase of MEP percentage over GMPs and CMPs in bone marrow compared to controls ( Figure 2B, middle panel). Also, the proportion of CD71 + , TER119 + pro-erythroblasts and of more mature CD71 2 , TER119 + erythroblasts was decreased ( Figure 2B, upper panel), indicating a delay in erythroid maturation similar as seen in EpoR-Cre, Gfi1b fl/fl mice. Spleens of Mx-Cre, Gfi1b fl/fl mice were larger in size and showed a significant increase of TER119 + cells, suggesting ongoing extramedullary erythropoiesis ( Figure 2B, lower panel and Figure S3). Peripheral blood analysis revealed a significant decrease in red blood cell count (RBC), hematocrit (HCT) and hemoglobin (Hgb) in Gfi1b deficient mice compared to wild type controls ( Figure 2C). Consequently, numbers of reticulocytes (Retic), immature reticulocytes (IRF-H) and macrocytic RBCs (Macro) were increased in the absence of Gfi1b as well as the mean corpuscular volume (MCV) and red cell size and shape (RDW), while the white blood cell count (WBC) did not change significantly ( Figure 2C), indicating that Gfi1b deficient mice suffer from anemia.
To further confirm these findings, we used mice carrying a tamoxifen inducible Cre recombinase integrated into the Rosa26 locus for the ablation of Gfi1b expression in adult mice. In these mice the recombination of one or both conditional Gfi1b alleles was still incomplete when both Gfi1b alleles were floxed, but was efficient when only one floxed allele was present ( Figure S4). We therefore used mice in which one floxed allele was replaced by a Gfi1b:GFP knock-in allele, which disrupts the Gfi1b coding region and allows to measure Gfi1b mRNA expression by monitoring green fluorescence [1][2][3]. Rosa-Cre-ERT, Gfi1b GFP/fl mice were sacrificed between eight and 15 days after the last of four treatments with tamoxifen ( Figure 3A). Flow cytometric analysis of bone marrow cells from these mice showed strongly reduced percentages of late erythroblasts (TER119 + , CD71 + and TER119 + , CD71 2 cells), a slight increase of pro-erythroblasts (TER119 2 , CD71 + cells) and a marked increase of the percentage of MEPs at the expense of GMPs compared to controls ( Figure 3B). The spleen in Rosa-Cre-ERT, Gfi1b GFP/fl mice was enlarged about twofold (not shown) and showed a significant accumulation of TER119 + cells ( Figure 3C), which is indicative for ongoing extramedullary erythropoiesis as was also seen in Mx-Cre, Gfi1b fl/fl mice. This suggests that Gfi1b deficiency initiates a mechanism to compensate for marrow insufficiency. In addition, Rosa-Cre-ERT, Gfi1b fl/fl mice showed similar alterations in their blood parameters as Gfi1b fl/fl -MxCre animals, namely reduced red blood cell count (RBC), hematocrit (HCT) and hemoglobin (Hgb) and increased numbers of reticulocytes, indicating again an anemic state in the absence of Gfi1b (not shown).
To gain more insight into the effects of tamoxifen on more differentiated erythroblast populations, we analyzed bone marrow of Rosa-Cre-ERT, Gfi1b GFP/fl and Rosa-Cre-ERT, Gfi1b GFP/ + animals from day 2 to day 8 after two tamoxifen injections using gates that divide TER119 + cells into proerythroblasts, basophilic erythroblasts, polychromatophilic erythroblasts and orthochromatophilic erythroblasts. Between day 2 and day 5 after tamoxifen injection, both basophilic and polychromatophilic erythroblasts almost entirely disappeared regardless whether a Gfi1b allele remained intact or not ( Figure S5). Between day 6 and day 8 after tamoxifen injection, basophilic erythroblasts became detectable in the control mice but not or only to a lower extent in the Gf1b deleted animals ( Figure S5). This suggests that Cre-ERT has a detrimental effect on TER119 + cells upon tamoxifen treatment and that these cells require Gfi1b for the differentiation of the TER119 2 precursor population.
To test the long term effect of Gfi1b ablation by activating Cre-ERT, Rosa-Cre-ERT, Gfi1b GFP/+ and Rosa-Cre-ERT, Gfi1b GFP/fl mice were analyzed two or nine months after tamoxifen-induced deletion of Gfi1b. TER119 + cells were gated again into four erythroblast populations. While a decrease of basophilic erythroblasts was still maintained, all other TER119 + cells were present at wt frequencies at both nine months (data not shown) and two months after tamoxifen induced Gfi1b ablation ( Figure S6A, B). In animals with floxed Gfi1b alleles, only TER119 2 cells showed efficient Cre mediated excision two months after tamoxifen induction, whereas the TER119 + erythroblast population only contained floxed or GFP alleles ( Figure S6C). This indicated that these TER119 + erythroblasts very likely emerged from few nondeleted precursors in the TER119 2 population, which supersede those with efficient excision of the Gfi1b allele and develop into TER119 + erythroid cells.

Gfi1b deficient cells shows defects in the regulation of globin gene expression
To gain more insight into the maturation defect caused by Gfi1b deficiency and to avoid any non-specific effects seen with tamoxifen, we performed two independent genome wide expression profiling experiments with FACS-sorted TER119 + bone marrow cells from Gfi1b fl/fl , Mx-Cre mice (see Figure 2B, lower panel for sorting gate) and from TER119 + fetal liver cells from EpoR-Cre Gfi1b fl/fl mice. Most of the significantly regulated protein coding genes were upregulated in Mx-Cre/pIpC induced Gfi1b deficient cells compared to controls, which is in agreement with the role for Gfi1b as a transcriptional repressor ( Figure 4A). Gene set enrichment analysis (GSEA) revealed targets of Gata2 and genes negatively regulated by Stat5 to be most affected by Gfi1b deficiency (Figure 4B, C). Platelet, HSC and AML specific genes were also significantly enriched among Gfi1b effector genes, which were up regulated in Gfi1b deficient cells ( Figure 4B). Analysis of Figure 5. Deregulation of embryonic globin genes in Gfi1b deficient TER119 + cells. A: Scatter plot comparison of gene expression levels (log2 of normalized signal intensities) in TER119 + pIpC induced Mx-Cre, Gfi1b fl/fl bone marrow cells compared to pIpC induced cells from control mice. Dots represent probesets and are jittered for better visualization. Probesets were classified as indicated and probesets for hemoglobin genes and important regulators of hemoglobin gene expression and globin switch were labeled (red dots). Two RNA samples for each genotype were pooled and analyzed on single arrays. B: Scatter plot comparing the changes in gene expression induced by inactivation of Gfi1b in erythroid cells of adult mice (y-axis) as in (A) with genes regulated in fetal liver cells during development of the mouse embryo from 11 dpc (E11) compared to 16 dpc (E16). Raw data for fetal liver development was taken from GEO data series GSE13149 and reanalyzed. Embryonic globin genes (red), megakaryocyte/ platelet specific genes (blue) and integrins known to be targets of Gfi1b in hematopoietic stem cells (yellow) are indicated. Genes that are downregulated in fetal liver cells during mouse development but upregulated in Gfi1b-KO cells are likely to be direct targets of the transcriptional repressor Gfi1b and were subjected to GSEA analysis (green frame and table to the right) and show a high enrichment in megakaryocytic/coagulation related genes and globin genes. doi:10.1371/journal.pone.0096636.g005 factors known to be associated with erythropoiesis showed that the expression of Gata2, Klf2, Bcl11a, Sox6 and the embryonic globin genes Hba-x, Hbb-bh1 and Hbb-y were affected by the deletion of Gfi1b ( Figure 5A). A comparison of genes up-regulated in Gfi1b deficient TER119 + erythroid cells with those down-regulated during embryonic development revealed that expression of the embryonic globin genes Hba-x, Hbb-bh1 and Hbb-ywas strongly affected by Gfi1b ablation ( Figure 5B). In addition, integrins alpha6 and alpha2b/b3 (CD41/61), already described as Gfi1b effectors in HSCs [9] and several other megakaryocyte/platelet specific genes such as Gp1bb (CD42C), Timp3 or Pf4 showed increased expression in Gfi1b deficient cells ( Figure 5B).
The second expression profiling experiment comparing mRNA prepared from TER119 + fetal liver cells from EpoR-Cre, Gfi1b fl/fl mice and wild type littermates isolated at 14.5 dpc also demonstrated that over 60% of protein coding genes whose expression changed more than two-fold were up-regulated in Gfi1b knockout cells, which was again in agreement with a repressor function of Gfi1b ( Figure 6A, pie diagram). Exon analysis confirmed deletion of the targeted Gfi1b exons in the TER119 + fetal liver cells used for the analysis ( Figure S7). GSEA analysis revealed that gene sets related to coagulation and immune response, hypoxia and CBFA2T3 (Eto2) targets were enriched among up-regulated genes, whereas gene sets related to cell cycle regulation, targets of E2F and DNA synthesis were enriched among down-regulated genes ( Figure 6B, C). Similar to the analysis of TER119 + bone marrow cells from Mx-Cre, Gfi1b fl/fl mice, a deregulated expression of genes encoding GATA2, SOX6 and of the Hba-x and Hbb-bh1 embryonic globin genes was observed ( Figure 6D).

Gfi1b is required for the developmental repression of embryonic globin gene expression
To validate a potential regulation of embryonic globin genes by Gfi1b, we FACS-sorted CD71 + , TER119 2/lo (proerythroblast) and CD71 + , TER119 + (late erythroblast) fractions of E15.5 fetal liver cells from wt and EpoR-Cre Gfi1b fl/fl mice for RT-PCR expression analysis. The expression of the genes for Hba-x, Hbb-bh1 and Hbb-y were up-regulated about 10 fold in Gfi1b deficient cells compared to their wt counterparts ( Figure 7A). A developmental expression analysis of the Hba-x, Hbb-y and Hbb-bh1 genes showed significantly higher levels in Gfi1b deficient fetal liver cells than in wild type controls throughout development from stages E12.5 to E16.5 ( Figure 7B). A similarly enhanced expression was found in constitutive Gfi1b deficient (Gfi1b GFP/GFP ) fetal liver cells (stage E 13.5) where the expression of Hbb-bh1, Hbb-y and Hba-x was induced up to over 25 fold over controls ( Figure 7C). The expression of Bcl11a, a target of KLF1, remained similarly regulated during development in Gfi1b knockout fetal liver cells and wt controls ( Figure 8A). In contrast, Sox6 was almost absent in Gfi1b deficient fetal liver cells at all developmental stages analyzed and Gata1 was not induced to wt expression levels from stage E 13.5 onwards ( Figure 8A). Expression of Glycophorin A (GYPA), an erythroid differentiation marker gene was only induced at a very late stage (E 15.5) in Gfi1b deficient cells and never reached wt expression levels ( Figure 8A).
To test whether the deregulation of globin genes is also maintained in adult Gfi1b deficient mice, we used FACS sorted CD71 + , TER119 2 (proerythroblast) or CD71 + , TER119 + (late erythroblast) bone marrow cells from two surviving adult EpoR-Cre Gfi1b fl/fl mice for Q-PCR analysis ( Figure 8B). Only a partial deletion of the floxed Gfi1b allele was detected in TER119 + cells ( Figure 8C). However, despite this partial deletion, expression of beta-like embryonic hemoglobin genes (Hbb-y and Hbb-bh1) was still strongly up-regulated in these cells over 60 to over 100 fold, respectively, compared to wild type controls ( Figure 8B). Hba-x or the adult hemoglobin genes alpha (Hba) and beta (Hbb) or Gata1were only mildly affected by Gfi1b deficiency in these cells ( Figure 8B). When we compared the effect of Gfi1b deficiency on a number of known and suspected regulators of globin gene expression ( Figure S8) between fetal liver and adult TER119 + bone marrow cells from our array data, the only significant overlap turned out to be a strong overexpression of Gata2 and of the fetal globin genes themselves. This suggests an important role for Gata2 in the effect of Gfi1b deficiency on fetal globin gene expression and a more direct involvement of GFI1B in the regulation of fetal globin gene expression.

Discussion
In this study we present evidence that the transcriptional repressor Gfi1b is an important factor for murine embryonic and adult definitive erythropoiesis. It has been described previously that Gfi1b is highly expressed in megakaryocyte and erythrocyte progenitors (MEPs) and to a lower extent throughout erythrocyte maturation [1]. However a complete study of the role of GFI1B in erythroid differentiation throughout development and in adult stages was hampered by the early embryonic lethality of germline Gfi1b knockout mice. We have analyzed three different mouse models, which enabled the deletion of conditional Gfi1b alleles either specifically in erythropoiesis at early developmental stages (EpoR-Cre mediated) or upon treatment with either pIpC (Mx-Cre) or tamoxifen (Rosa-Cre-ERT). The results from analyses of all three models indicate that GFI1B is an essential factor required for erythroid maturation during embryonic development in the fetal liver and in adult stages for the production of mature erythroid cells in the bone marrow. This is supported by the reduced frequencies of TER119 + erythroblasts that were observed in all Gfi1b deficient mice regardless how the ablation was achieved. Analysis of mice 2 and 9 months after a deletion of the conditional Gfi1b allele even suggested that Gfi1b is absolutely essential to maintain erythropoiesis at long term.
We also found that MEPs are present and even increased in percentage in adult Gfi1b deficient mice, which excludes a lack of precursor cells as a the underlying cause for the low frequencies of erythroblasts in the absence of Gfi1b and supports a regulatory role Figure 6. Transcriptome analysis of TER119 + fetal liver cell derived microarray data. A: Pie chart visualization of the proportion of genes up and down regulated more than two-fold in EpoR-Cre induced Gfi1b deficient fetal liver cells at 14.5dpc derived from microarray analysis. Array analysis for each genotype was done in duplicates. B: Gene set enrichment analysis shows enrichment in genes upregulated in Gfi1b deficient fetal liver cells related to hemostasis/coagulation, hypoxia and targets of the transcriptional corepressor Cbfa2t3 (upper enrichment plots and table). C: Cell division related gene sets and targets of the master regulator of replication E2F are enriched in downregulated genes in the Gfi1b knockout. D: Scatter plot comparison of gene expression levels (log2 of normalized signal intensities) in TER119 + cells from EpoR-Cre induced Gfi1b deficient fetal liver cells at 14.5dpc compared to the respective wild type cells. Dots represent probesets and are jittered for better visualization. Probesets were classified as indicated and probesets for hemoglobin genes and important regulators of hemoglobin gene expression and globin switch were labeled. doi:10.1371/journal.pone.0096636.g006 of Gfi1b during erythroid commitment and development. Probably as a result of the erythroid maturation defect, adult Gfi1b deficient mice suffer from anemia as indicated by the low RBC counts, the low hematocrit and hemoglobin levels. In addition, adult Gfi1b deficient mice also show extramedullary erythropoiesis, which may be a consequence of the anemia.
Mice in which Gfi1b ablation was mediated by the EpoR-Cre transgene did not die at midgestation but mainly at birth. This finding points to the possibility that a delayed or inhibited erythroid development is not entirely responsible for the embryonic lethality observed in germline Gfi1b knockout mice. However, it cannot be ruled out that incomplete deletion of the floxed Gfi1b   Sox6) and glycophorin A (Gypa) normalized to Gapdh. Sample sizes were as described in Figure 7B. B: Q-PCR analysis on RNA from CD71 + , TER119 2 (pro-erythroblasts) and TER119 + (late erythroblasts) live bone marrow cells from a surviving EpoR-Cre induced Gfi1b-KO mouse compared to a wild type littermate. All measurements were done in triplicates. C: RT-PCR detection of EpoR-Cre and Gfi1b wt, flox and KO (excised) alleles on total RNA from bone marrow of a surviving mouse with erythroid specific inactivation of Gfi1b by EpoR-Cre. doi:10.1371/journal.pone.0096636.g008 alleles has allowed enough erythrocytes to mature to allow full development past E13. 5-14.5. It remains unclear however, why most EpoR-Cre, Gfi1b fl/fl mice die shortly after birth. Additional studies are necessary to clarify this, but a recently described role of the EpoR in vascular cells and hypoxic stress [43] may have contributed to this lethality. Different from what was observed in fetal liver, ablation of Gfi1b in adult mice, whether erythroid specific or not, did not lead to noticeable accumulation of proerythroblasts (CD71 + , TER119 lo cells) in the bone marrow, but rather to a loss of erythroblast populations (CD71 + , TER119 + cells). It is thus possible that a role of Gfi1b in proerythroblast maturation is different in embryonic and adult development.
Our data also demonstrate that Gfi1b plays an important role in the regulation of the expression of embryonic globin genes. Regardless how Gfi1b ablation was achieved, all animals that lack or are deficient of Gfi1b showed a significant increase in embryonic globin gene expression both in fetal liver cells and in bone marrow derived adult erythroid cells. The expression of embryonic globin genes is dependent on fine-tuning by many transcription factors, co-activators and co-repressors. An important role in this regulation has previously been assigned to two complexes, the NF-Y/Gata2 activator hub and the BCL11a/COUPTFII/ GATA1 repressor hub that are both present in embryonic and adult erythroid cells on repressed and active y-globin regulatory sequences [44]. The embryonic globin gene repressor function of BCL11a is dependent on the presence of another factor, SOX6, to form long-range interactions [45]. Our experiments showed a strong down-regulation of the expression of both Sox6 and Gata1 in Gfi1b deficient fetal liver cells and largely unaffected levels of Bcl11a expression. Since SOX6 and GATA1 are both required for a functional repressor complex that occupies the embryonic beta globin locus at regulatory sequences, these findings provide compelling evidence that the deregulation of these two genes is responsible for impaired repression of the embryonic globin genes in Gfi1b deficient mice. In contrast, TER119 + cells from adult Gfi1b knockout mice showed up-regulation of Gata2 mRNA levels, but almost no change in other known regulators of embryonic globin gene expression. Gata2 overexpression is known to stimulate fetal globin gene expression [46] and moreover, downregulation of Gata1 induces Gata2 expression and results in impaired differentiation of erythroblasts [47], This phenotype is similar to what we observe in our Gfi1b knockouts. The analysis of published ChIPseq data of Gfi1b [48] did not reveal a direct occupation of the embryonic globin genes suggesting that Gfi1b affects the expression of embryonic globin genes likely via a different mechanism in erythroid cells from fetal liver or adult bone marrow. Although our data clearly establish a role of Gfi1b in embryonic globin expression, future studies will have to show whether this occurs through a direct repression of the globin locus by the GFI1B/ LSD1/CoREST repressor complex, or whether Gfi1b acts indirectly on globin expression possibly in a complex with other regulatory factors. A: Flow cytometric analysis of cells from the indicated mice to detect different erythroblast cell populations according to CD71 and TER119 marker expression. Mice were analyzed 2 months after tamoxifen treatment as described in Figure S5. B: Quantification of the frequency of the indicated cell subsets from the mice characterized in (A). Proerythroblast: CD71 + TER119 lo/ 2 , Basophilic erythroblasts: CD71 + , TER119 + , Polychromatophilic erythroblasts: CD71 med , TER119 + , Orthochromatophilic erythroblasts: CD71 lo , TER119 + . C: PCR analysis of DNA from total bone marrow from Rosa-Cre-ERT, Gfi1b GFP/fl or Rosa-Cre-ERT, Gfi1b fl/fl mice to detect the wt, floxed or excised (KO) alleles. (TIF) Figure S7 Box-and-Whisker plot of gene level normalized intensity for Gfi1b in wt and Gfi1b-KO fetal liver cells. The upper plot shows smallest value, first quantile, median, third quantile and largest value of the Gfi1b-gene level normalized intensities of wild type (red) and Gfi1b knockout (blue) TER119 + fetal liver cells analyzed in duplicates on Affymetrix gene-1.0-ST arrays that allow for exon-level analysis. Exons 10481308, 1048130 and 10481310 (including first ATG) are bordered with loxP sites in the conditional Gfi1b-KO and should be deleted by CRE recombination. This is a proof for the deletion of Gfi1b by EpoR-Cre in TER119 + fetal liver cells. The lower plot does show the exon-intron structure and gene-1.0ST array probesets covering the Gfi1b-gene and analyzed here. Both plots were generated using the web-tool ''Gene array analyzer'' (http://gaa. mpi-bn.mpg.de/) [49]. (TIF) Figure S8 Change of expression of globin genes and their regulators induced by Gfi1b deficiency. A: Scatter plot demonstrating the relation of the magnitude of gene expression changes induced by Gfi1b deficiency in fetal liver cells at day 14.5 relative to the probability of a significant change of expression (rawp). Values were taken from array data sets described in Figure S7. Genes visualized are either globin genes or known or suspected regulators of globin gene expression. Labels represent the official gene symbols and dots represent the data for gene level analysis of array data. B: Bar graph representing the magnitude of gene expression changes induced by Gfi1b deficiency in TER119 + bone marrow cells from adult mice as measured on affymetrix MOE430-2 expression arrays. Data are from single array experiments, not allowing for p-value determination. The same genes as in (A) were analyzed. Multiple probesets for single genes were averaged. Gene expression changes are indicated in log-scale. Dotted lines indicate the levels of 1.5-fold or 2-fold changes in gene expression level as indicated. (TIF)