Inducible Fli-1 gene deletion in adult mice modifies several myeloid lineage commitment decisions and accelerates proliferation arrest and terminal erythrocytic differentiation

Mature blood cells in adults are permanently regenerated from a limited pool of pluripotent hematopoietic stem cells (HSCs) localized in the bone marrow. This process occurs through the hierarchical generation of intermediate progenitors with more and more restricted differentiation and proliferation potential that can be prospectively purified. According to current models, myeloid lineages are derived from a common myeloid progenitor (CMP) generating 2 distinct bipotent progenitors, the megakaryocytic-erythrocytic progenitors (MEPs) and the granulocytic-monocytic progenitors (GMP).¹  MEPs generate in turn erythrocytic and megakaryocytic monopotent progenitors, whereas GMP generate granulocytic and monocytic monopotent progenitors. Increasing data also indicate an alternative pathway generating MEPs directly from HSCs.^2-4 

All this process is controlled by permanent crosstalk between extracellular signals and intracellular regulatory networks of transcription factors. Although having no instructive role, erythropoietin (EPO) is the major cytokine controlling erythropoiesis through the stimulation of proliferation and survival of committed erythrocytic progenitors expressing the specific receptor EPOR.⁵  Thrombopoietin (TPO) is the major cytokine controlling megakaryopoiesis through the stimulation of proliferation and differentiation of committed megakaryocytic progenitors expressing the specific receptor TPOR/MPL. Contrasting with the strict dependence of mature red cells production on EPOR signaling, significant platelets production can occur without TPO signaling.⁶  For example, stromal cell–derived factor-1 is known to stimulate platelets production independently of TPO.⁷  Furthermore, Notch signaling stimulates megakaryopoiesis directly from HSCs even in the absence of TPO.⁸  More recently, in vitro cultures have shown that TPO induces BMP4 production and autocrine stimulation of megakaryopoiesis initiated from human CD34⁺ progenitors.⁹ 

The specific combination of transcription factors allowing erythrocytic or megakaryocytic lineages commitment remains poorly understood.^10,11  Indeed, both lineages are dependent on many common transcription factors including Gata1, Tal1, Nfe2, Gfi1b, Zbp89, as well as cofactors like Fog1, Lmo2, or Ldb1 acting in large multiprotein complexes such as Fog1/Gata1/Zbp89¹²  or Gata1/Tal1/Lmo2/Ldb1.¹³  No differential requirement of these factors for either lineage has been reported to date. In contrast, low levels of c-Myb favor megakaryocytic at the expense of erythrocytic lineage,¹⁴  whereas the same is true for Stat5.¹⁵  Aml1/Runx1 cooperates with Gata1 to activate some megakaryocytic promoters,¹⁶  and conditional inactivation of Runx1 gene in adult inhibits megakaryocytic maturation but has no effect on erythrocytic differentiation.¹⁷  Eklf and Fli-1 are also known binding partners of Gata1 that competitively activate either erythrocytic or megakaryocytic promoters, respectively.¹⁸  Fli-1–mediated activation of megakaryocytic promoters is further dependent on the cooperation with Fog1 and/or Aml1.¹⁹  We²⁰  and others^21,22  recently demonstrated that Eklf promotes the commitment of bipotent progenitors toward erythrocytic at the expense of megakaryocytic differentiation. Reciprocally, concordant results from 2 different mouse models of Fli-1 gene disruption clearly established that the thrombocytopenia observed in Fli-1^−/− embryos is at least partially due to a maturation defect of fetal liver megakaryocytes.^23-25  Further studies have shown that this contribution of Fli-1 to late stages of megakaryocytic differentiation is shared by at least 2 other ETS family transcription factors, Gapbα²⁶  and Erg.²⁷  However, the real contribution of Fli-1 in the commitment decision toward megakaryocytic versus erythrocytic lineage remains rather controversial. For example, somewhat contradictorily to normal numbers of colony-forming unit megakaryocyte (CFU-MK) progenitors in Fli-1^−/− fetal liver,²³ Fli-1^−/− embryonic stem (ES) cells failed to generate megakaryocytic colonies in vitro, and the in vitro generation of megakaryocytic cells from Fli-1^−/− AGM (aorta-gonad-mesonephros) was almost completely suppressed.²⁸  Furthermore, Fli-1^−/− ES cells failed to contribute to bone marrow megakaryocytes in reconstituted chimeric mice, whereas Fli-1^+/− ES cells displayed a massively increased contribution to erythrocytic cells compared with wild-type ES cells.²⁵  Taken together, these data raise the intriguing possibility that megakaryocytic commitment might be more stringently dependent on Fli-1 in adult and during early development than during fetal life.

To address the role of Fli-1 in adult hematopoiesis, we have generated new transgenic mice allowing inducible Fli-1 gene disruption. We found that Fli-1 gene disruption in adult mice induced a mild thrombocytopenia associated with a maturation defect of bone marrow megakaryocytes as previously observed in fetal liver. Further investigations revealed that Fli-1 gene deletion modified commitment decisions at several hierarchical levels of myelopoiesis characterized by a specific increase of MEPs that are prone to preferential and accelerated erythrocytic differentiation at the expense of megakaryocytic differentiation and by an unaltered number of GMP that are prone to preferential monocytic at the expense of granulocytic differentiation.

Fli-1 gene exon 9 and its upstream and downstream flanking sequences were amplified from ENS ES cell²⁹  DNA using primers indicated in supplemental Table 1A (available on the Blood Web site; see the Supplemental Materials link at the top of the online article) and inserted into BamH1/Nhe1, SacII/Not1, and Xho1/Kpn1 sites of the KSloxPfrtNeoBS (3′-5′) vector,³⁰  respectively. The resulting targeting construct contained 2 loxP sites flanking Fli-1 exon 9 coding sequence, 1 Frt-flanked neomycine resistance cassette and 2 homology arms (supplemental Figure 1A). ENS ES cells were transfected with linearized targeting vector by electroporation, and homologous recombinant clones were screened among G418-resistant clones by Southern blot analysis as described in supplemental Figure 1B. The neomycine-resistance cassette gene was excised with Flp recombinase by transfection with the pCAGGS-Flpe vector³¹  and selection with puromycin for 72 hours. Correctly targeted clones obtained after neomycine-resistance cassette excision were injected into C57BL/6J blastocysts and implanted into recipient females. Germ line transmission of the Fli-1^fl allele was obtained for 1 of 4 chimeric mice generated with ES recombinant clone 79.8 of genotype Fli-1⁺/Fli-1^fl. Constitutively deleted allele Fli-1^Δ (supplemental Figure 1A) has been generated in vivo by crossing mouse females of genotype Fli-1⁺/Fli-1^fl with Sycp1Cre transgenic males expressing Cre recombinase during male meiosis.³²  Mice harboring either Fli-1^Δ or Fli-1^fl alleles were then crossed with Mx1-Cre transgenic mice³³  expressing Cre recombinase under the control of the interferon inducible Mx1 gene promoter to generate couples of mice with genotypes Fli-1^fl/Fli-1^Δ or Fli-1^fl/Fli-1^fl harboring or not the Mx1-Cre transgene. Between 8 and 12 weeks after birth, couples of Cre⁺ and Cre⁻ mice of either Fli-1^fl/Fli-1^Δ or Fli-1^fl/Fli-1^fl genotypes were injected intraperitoneally 3 times on alternate days with 300 μg of polyinosinic:polycytidic acid (poly I:C; Sigma-Aldrich) dissolved in 0.9% NaCl/20 g body mass. Injected mice were killed for analyses between 8 and 10 days after the first injection. Mice were bred and maintained under specific pathogen-free conditions at the ALECS-SFP animal facilities of the Faculté de Médecine Lyon-Est (Université Claude Bernard, Lyon1, France) and experimentations were performed according to procedures approved by the local animal care and experimentation authorities (Ministère Délégué de la Recherche et des Nouvelles Technologies, agreement no. 4936; Direction des Services Vétérinaires, agreement nos. 69266317 and 7462).

Peripheral blood was obtained by cardiac puncture and mixed with EDTA (ethylenediaminetetraacetic acid). Red cell, white cell, and platelet counts were determined manually using Mallassez counting chambers after appropriate dilutions in phosphate-buffered saline (PBS), Hayem blue, or BD Unopette (BD Biosciences) solutions, respectively. Bone marrow cells (BMC) were flushed from femurs and tibiae, pooled, washed in PBS containing 0.5% fetal calf serum (FCS) and 2mM EDTA, and made in single-cell suspensions by repeated pipetting and filtering through a 100-μm nylon mesh. Single-cell suspensions from spleen and thymus were prepared using the same procedure after initial organs mincing using scissors and clamps.

Relative proportions of mature cells were determined using a FACSCalibur or FACSCanto II flow cytometer (BD Biosciences) after single or double labeling with the following monoclonal conjugated antibodies: anti-CD71–fluorescein isothiocyanate (FITC) and anti-Ter119–phycoerythrin (PE) for erythrocytic lineage, anti-CD41–FITC for megakaryocytic lineage, anti–Gr-1–PE, and anti-CD11b–FITC for granulocytic/monocytic lineages, anti-CD45R(B220)–PE, and anti-IgM–FITC for B-lymphoid lineage, anti-CD4–PE and anti-CD8a–FITC for T-lymphoid lineages, and anti-NK1.1–PE for natural killer (NK) lineage. All these antibodies were purchased from Caltag Laboratories except for CD41 (Acris Antibodies).

For quantification and sorting of CMP, MEP, and GMP progenitors, BMC suspensions were treated with red blood cells ACK lysing buffer (Lonza) before filtering through a 70- or 100-μm cell strainers (BD Biosciences). BMC suspensions were then labeled with a lineage cocktail of biotinylated antibodies (Lineage cell depletion kit; Miltenyi Biotec) supplemented with biotinylated anti–Sca-1 (BD Pharmingen), anti-CD3 (BD Pharmingen), anti-ILR7 (eBiosciences), and anti-CD19 (Serotec). After magnetic depletion of lineage-positive cells using MS columns (Miltenyi Biotec), Lin⁻Sca⁻ cells were numbered and further labeled with streptavidin-PE-Cy7 (BD Pharmingen) for the detection of residual biotinylated stained cells and with anti–c-Kit–APC (BD Pharmingen), anti-CD34–FITC (e-Biosciences), and anti–Fcγ-RII/III–PE (BD Pharmingen) antibodies. MEPs, CMPs, and GMPs were then identified as previously described³⁴  as the Lin⁻Sca⁻Kit⁺ Fcγ-RII/III^low/CD34^low, Fcγ-RII/III^low/CD34^high, and Fcγ-RII/III^high/CD34^high subpopulations, respectively, and sorted using FACSAria or FACSVantage cell sorters and DIVA software (BD Biosciences). Dead cells were excluded by 7-aminoactinomycin D staining (Sigma-Aldrich).

Hematopoietic progenitors in unfractionated BMC suspensions were evaluated by colony assays scored at day 7 using MethoCult^RSF^BIT M3236 methylcellulose semisolid medium (StemCell Technologies) supplemented with 20% FCS (StemCell Technologies), mIL3 (10 ng/mL) and huEPO (3.3 U/mL). Hematopoietic progenitors in sorted MEP, GMP, and CMP populations were evaluated by colony assays scored at day 7 using MethoCult M3234 methylcellulose semisolid medium (StemCell Technologies) supplemented with 30% FCS, murine interleukin 3 (mIL3; 10 ng/mL), murine stem cell factor (mSCF; 50 ng/mL), mFlt-3l (5 ng/mL), murine granulocyte-macrophage colony-stimulating factor (mGM-CSF; 5 ng/mL), huIL11 (50 ng/mL), huEPO (2 U/mL), and mTPO (50 ng/mL). Megakaryocytic progenitors (CFU-Mk) in unfractionated BMC suspensions were evaluated by colony assays using MegaCult-C collagen semisolid medium (StemCell Technologies) supplemented with mIL3 (10 ng/mL) and mTPO (50 ng/mL) and by numbering acethylcholinesterase-positive colonies at day 7. All cytokines were purchased from PeproTech except huEPO (kindly provided by F. Nicolini).

MEPs were seeded at 2000 cells/mL in Iscove modified Dulbecco liquid medium supplemented with either 10% FCS and myeloid cytokines exactly as previously described³⁵  or with 15% BIT 9500 serum substitute (StemCell Technologies) and only SCF, IL3, IL11, TPO, and IL6.⁹  After 5 days of culture, viable cells excluding trypan blue were numbered, and the relative proportions and maturation of erythrocytic and megakaryocytic cells were determined by fluorescence-activated cell sorting (FACS) after double-labeling with labeled anti-CD71 and anti-Ter119, anti-CD61 and anti-CD41, or anti-CD41 and anti-CD42b antibodies. For single-cell progeny analyses, single MEPs were plated in 96-well plates using a FACSVantage flow cytometer coupled with single-cell depositor in 60 μL of the same liquid medium with serum. Each individual well was evaluated by light microscopy after 6 to 7 days of culture to determine the proportion of wells containing at least 1 identifiable living cell as well as the relative proportions of erythrocytic colonies (containing a large number of small cells), megakaryocytic colonies (containing a small number of large cells), or mixed colonies as illustrated in supplemental Figure 3.

Mice genotyping was performed by polymerase chain reaction (PCR) analysis using crude lysates of tail biopsies using primers indicated in supplemental Table 1B. Hematopoietic cells DNA used to control deletion efficiencies were prepared either with QIAampDNA Mini kit (QIAGEN; for > 50 000 cells) or with NucleoSpin Tissue XS kit (Macherey-Nagel; < 50 000 cells). Deletion efficiencies were determined either by densitometry of PCR products obtained with primers KO14 and KO8 (supplemental Table 1B) or by quantitative (q)PCR using Fli-1 exon 9 and 129K primers located in α globin locus used as standard reference (supplemental Table 1D). Quantitative PCRs were performed using Light-cycler 480 SybR-Green-Master-Roche kit on a Mx3000 PTM PCR instrument and analyzed using Mx Pro software (Stratagene). Fli-1 genotyping of isolated colonies was performed using primers indicated in supplemental Table 1C.

Bone marrow large mature megakaryocytes were purified by 2 rounds of sedimentation on discontinuous 0%-3% BSA gradient as previously described³⁶  (4 mL 3% BSA/PBS in a 15-mL Falcon tube overlaid with 4 mL 1.5% BSA/PBS and then 4 mL BMC suspension in PBS; sedimentation 40 minutes at 1g). Single mature megakaryocytes were then pipeted individually under visual microscope control and transferred in Eppendorf tubes. Large mature megakaryocytes from MEP liquid cultures were collected directly without enrichment after appropriate dilution of the cells suspension. Reverse transcription (RT) reactions were performed in the same tubes in 10 μL using 3′ Actin and Fli-1 primers (supplemental Table 1E), followed by PCR and then nested PCR on 1/50th of the first PCR.

Cell cytospins were examined under light microscope after either May-Grünwald-Giemsa or acetylycholinesterase staining.

Western blots analyses were performed on total cell lysates as previously described¹⁸  using the following antibodies: rabbit polyclonal anti Fli-1 (sc-356; Santa Cruz Biotechnology) and anti-Actin (sc-8432; Santa Cruz Biotechnology).

A conditionally disrupted allele of Fli-1 (Fli-1^fl) has been generated by inserting 2 loxP sequences flanking the Fli-1 gene exon IX encoding the whole ETS DNA binding and C-terminal domains of the Fli-1 protein (supplemental Figure 1A). A constitutively disrupted allele (Fli-1^Δ) has been also generated in vivo by crossing transgenic mice harboring the conditional Fli1^fl allele with Sycp1Cre mice expressing Cre recombinase in primary spermatocytes³²  (supplemental Figure 1A-C). All subsequent analyses were performed by comparing couples of mice of the same Fli-1 genotype (Fli1^fl/Fli1^fl or Fli1^Δ/Fli1^fl) carrying (Cre⁺) or not carrying (Cre⁻) the Mx1-Cre gene encoding interferon inducible Cre recombinase³³  and submitted to the same polyinosinic:polycytidylic acid (pIpC) injection protocol. Fli-1 gene deletion efficiencies detected in BMCs from injected Cre⁺Fli1^fl/Fli1^fl or Fli1^Δ/Fli1^fl mice ranged from 60% up to 90% and were associated with proportional decrease in Fli-1 protein levels (supplemental Figure 2). Deletion efficiencies and Fli-1 protein decreases in spleen and thymus were slightly and strongly lower than in bone marrow, respectively (supplemental Figure 2).

A first series of experiments performed on couples of injected Fli1^fl/Fli1^fl Cre⁺ and Cre⁻ mice revealed that injected Cre⁺ mice were characterized by a significant 33% mean decrease in circulating platelets number (630 ± 361 vs 950 ± 382 10⁹/L; P < .05, n = 9) with no significant variation of circulating red and white cell numbers. Concomitantly, analyses of BMC cytospins revealed a marked decrease in the mature/immature ratio of megakaryocytes in injected Cre⁺ mice indicating a maturation defect (Figure 1A-B). In addition to this maturation defect, injected Cre⁺ mice were characterized by a significant 2-fold decrease in CD41⁺ cells (Table 1A) and by a 10-fold decrease in bone marrow CFU-MK (Table 1B). Similar decrease in circulating platelets and bone marrow CFU-MK numbers were observed by comparing couples of injected Fli1^fl/Fli1^Δ Cre⁺ and Cre⁻ mice, and a compilation of the results obtained for all Cre⁺ and Cre⁻ couples of mice analyzed so far is presented in Figure 1C and D, respectively. Intriguingly enough, although the decrease in platelets and CFU-MK numbers were both proportional to the extent of Fli-1 gene deletion, the decrease in CFU-MK appeared much more drastic than the decrease in platelets numbers. Altogether, these data indicated that loss of Fli-1 in adult mice induced a mild thrombocytopenia associated with a marked reduction of megakaryocytic progenitors and a maturation defect of megakaryocytes present in bone marrow.

Further analyses of BMCs failed to detect any significant variation in the proportion of cells expressing B or T lymphoid lineage markers between injected Cre⁺ and Cre⁻ mice (Table 1A). However, in addition to the decrease in megakaryocytic cells, these analyses revealed a 3-fold increase in natural killer NK1.1⁺ cells, a 1.5-fold increase in CD71⁺Ter119⁺ erythrocytic cells and a 2-fold decrease in Gr1⁺⁺/Mac1⁺ granulocytic cells (Table 1A). In agreement with the increase in erythrocytic cells, the number of BFU-E (burst-forming unit erythroid) and CFU-E (colony-forming unit erythroid) erythrocytic progenitors increased by 2.2- and 1.7-fold, respectively (Table 1B). The opposite changes in the numbers of BFU-E and CFU-MK raised the intriguing possibility that a commitment switch occurred at the level of their common bipotent progenitor MEPs. Furthermore, in agreement with the decrease in granulocytic cells, the number of granulocytic progenitors CFU-G (colony-forming unit granulocytic) significantly decreased by 1.7-fold (Table 1B), whereas the number of monocytic CFU-M (colony-forming unit monocytic) progenitors symmetrically increased by 1.4-fold, although this increase was not statistically significant (Table 1B). Likewise, these symmetrical changes in the relative numbers of CFU-G and CFU-M progenitors also suggested a commitment switch occurring at the level of their common bipotent progenitor GMPs. To address these possibilities, we thus decided to analyze the progeny of purified MEP and GMP bipotent progenitors as well as of their parent multipotent progenitor CMPs.

Bipotent MEPs, GMPs, and multipotent CMPs were purified from the bone marrow of injected Fli1^Δ/Fli1^fl Cre⁺ and Cre⁻ mice according to published protocol based on their differential expression of CD16/32 and CD34.³⁴  Compared with injected Cre⁻ mice, injected Cre⁺ mice displayed a significant increase in MEP number (Figure 2A) reaching a more than 3-fold increase for the highest Fli-1 gene deletion efficiencies obtained (Figure 2B). Concomitantly, the numbers of CMPs and GMPs did not change significantly (Figure 2A), although both of them slightly increased by less than 2-fold for the highest Fli-1 gene deletion efficiencies (Figure 2C-D). The clonogenic progenies of these different classes of Cre⁻ or Cre⁺ multipotent progenitors were then compared in the presence of a complete cocktail of cytokines allowing the development of all myeloid progenitors (Figure 3). As expected, Cre⁻ CMPs generated all types of single lineage myeloid colonies as well as mixed granulo-monocytic and a few percent of mixed erythro-megakaryocytic colonies (Figure 3A left). Cre⁺ CMP generated the same number of colonies with an increased proportion of monocytic (M) and erythrocytic (E), a decreased proportion of granulocytic (G) but no identifiable megakaryocytic (Mk) or mixed erythro-megakaryocytic (E-Mk) colonies (Figure 3A right). Cre⁻ and Cre⁺ GMPs generated the same number of only G, M, or GM colonies but also with an increase of the M to G colonies ratio for Cre⁺ compared with Cre⁻ GMPs (Figure 3B). This increase in the M/G ratio occurring without any change of the total number of colonies strongly suggests that Fli-1 gene deletion slightly favors the commitment of bipotent GMPs toward monocytic at the expense of granulocytic lineage. Most strikingly, whereas Cre⁻ MEPs generated approximately one-third of megakaryocytic or mixed erythro-megakaryocytic colonies, Cre⁺ MEPs generated the same total number of colonies, but belonging almost exclusively to the erythrocytic lineage (Figure 3C). PCR analyses performed on single colonies revealed that whereas Fli-1 gene deletion could be evidenced on most erythrocytic colonies derived from Cre⁺ MEPs (Figure 3D lanes 5-7,13,14), all the few residual megakaryocytic colonies that could be analyzed were found to be undeleted (Figure 3D lanes 4,12, and data not shown). Taken together, these results thus indicated that loss of Fli-1 led to the generation of an increased number of MEPs that became virtually unable to generate mature megakaryocytic colonies in vitro. We also analyzed the progeny of single purified Cre⁻ or Cre⁺ MEPs after 7 days in liquid culture containing a full cocktail of myeloid cytokines (Figure 3E). No significant difference was found in the proportions of single Cre⁺ and Cre⁻ MEPs generating viable progeny (51% vs 56%) with Cre⁺ MEPs generating a significant increase in the proportion of erythrocytic colonies (78% vs 53%) and a compensating decrease of mixed (4% vs 10%) and pure megakaryocytic colonies (18% vs 36%; Figure 3E). Importantly, these single-cell liquid cultures showed that all classes of pure megakaryocytic colonies containing either 1, from 2 to 6, or more than 6 megakaryocytes were decreased to the same extent in Cre⁺ MEP compared with Cre⁻ MEPs (Figure 3E). Altogether, these results suggested that Fli-1 gene deletion is acting mainly by switching the commitment of MEPs toward the erythrocytic lineage rather than by lowering the proliferation or viability of their committed megakaryocytic progeny. Intriguingly however, we noticed a marked difference in the decrease of megakaryocytic and mixed colonies between single-cell liquid and methocell cultures that remained hardly explained by the slight difference in Fli-1 gene deletion efficiencies displayed by mice used in the 2 experiments (46% ± 5% vs 66% ± 11%, respectively).

The morphology of erythroid colonies generated by purified MEPs in methocell was rather heterogeneous ranging from large typical BFU-E–like colonies containing a high number of thick erythrocytic cell clusters (Figure 4A left, “large” colonies) to colonies containing a small number of tiny cell clusters looking more fully differentiated (Figure 4A left, “small” colonies). Interestingly, Fli-1 gene deletion increased the relative proportion of smaller colonies (Figure 4A right). Furthermore, cell cytospins of pooled colonies revealed an increased proportion of more differentiated cells characterized by smaller size and acidophilic cytoplasm after Fli-1 gene deletion (Figure 4B). These 2 observations thus suggested that Fli-1 gene deletion accelerated the erythrocytic differentiation of MEPs at the expense of their proliferation. To address this question more directly, we compared the proliferation rate and differentiation kinetics of purified MEPs in liquid cultures containing serum and a complete cocktail of myeloid cytokines. In agreement with previous report,³⁵  undeleted MEPs expanded up to 100-fold after 7 days in these conditions. In the same conditions, deleted MEPs expanded by only 58-fold (Figure 5A). Interestingly, deleted MEPs generated a significantly higher proportion of Ter119-positive cells (Figure 5B) among which a higher proportion already displayed reduced CD71 expression (Figure 5C) as well as reduced size (Figure 5D) indicating an increased proportion of more differentiated cells. These data thus indicated that, in addition to promote the commitment of MEPs toward the erythrocytic lineage, Fli-1 gene deletion concomitantly reduced their proliferation and accelerated their erythrocytic differentiation.

As expected from the results of semisolid colony assays (Figure 3), liquid bulk cultures of Cre⁺ MEPs displayed a drastic decrease in the number of large mature megakaryocytes (Figure 6A) proportionally to the extent of Fli-1 gene deletion (Figure 6B). Single-cell RT-PCR analyses revealed that most of these few remaining megakaryocytes (9/11) still expressed Fli-1 mRNA, indicating that most of them were derived from undeleted MEP (Figure 6C lanes 11-21). Likewise, we found a drastic reduction in the number of large acetylcholinesterase-positive megakaryocytes present in the bone marrow of injected Cre⁺ compared with Cre⁻ mice (Figure 6D). Furthermore, single-cell RT-PCR analyses revealed that most of these few remaining large mature megakaryocytes (4/5) still expressed Fli-1 mRNA and thus escaped Fli-1 deletion (Figure 6C lanes 22-27). Contrasting with this drastic effect on megakaryocytes maturation, the number of bone marrow CD41-positive cells were only slightly reduced (Table 1), and their polyploidy was only slightly affected (40% decrease of > 8n cells; supplemental Figure 3). Moreover, sorted double-positive bone marrow CD41⁺CD61⁺ cells displayed similar Fli-1 deletion efficiencies compared with the unsorted BMC populations, thus confirming that most of them were produced in the absence of Fli-1 (60% deletion in sorted cells CD41⁺CD61⁺ cells compared with 50% or 40% in unsorted BMCs in 2 different experiments, respectively; data not shown). In agreement to these in vivo data, Cre⁺ MEP displayed a similar 50% decrease in the production of CD41⁺ cells in both liquid (Figure 7B) and semisolid cultures (Figure 7C). The expression level of CD41 was clearly higher than the expression level in the initial MEPs (Figure 7A), thus confirming true megakaryocytic commitment. In addition, the decrease in the production of CD41⁺ cells by Cre⁺ MEPs in both liquid (Figure 7B) and semisolid cultures (Figure 7C) gradually increased with the expression level of CD41 (Figure 7D) and preferentially affected the most mature CD41⁺CD42b⁺ cells (supplemental Figure 5). Taken together, these data indicate that Fli-1 becomes increasingly required during the progression of megakaryocytic maturation ending with a critical requirement for the increase in cell size, CD41 and CD42b expression levels, accompanying and immediately following polyploidization.³⁷ 

In the present study, we used new transgenic mice allowing inducible Fli-1 gene deletion to determine the consequences of Fli-1 loss on adult myelopoiesis. Our results showed that Fli-1 loss modified multiple lineage commitment decisions as well as the proliferation/differentiation balance in the erythrocytic lineage.

At the highest hierarchical level analyzed, Fli-1 loss induced a specific increase in the number of bipotent MEPs. Importantly, this MEP increase did not occur at the expense of the GMP number. Furthermore, this MEP increase occurred without concomitant increase in the overall erythrocytic and megakaryocytic progeny of CMPs that are supposed to generate MEPs and GMPs. These data thus indicate that the increase in MEPs induced by Fli-1 loss did not originate from a commitment switch occurring at the CMP level, but probably rather through the alternative pathway initiated directly at the level of HSCs.^3,4,38,39  Similar increase in MEPs has also been observed following inducible deletion of the gene encoding the other ETS transcription factor Spi-1/Pu.1⁴⁰  in adult mice, and this increase is currently at least partly explained by direct interaction and functional cross-antagonism between Spi-1/Pu.1 and Gata-1.⁴¹  Although Fli-1 is also a binding partner of Gata-1, available data indicate that Fli-1 cooperates rather than inhibits Gata-1 activity, thus suggesting that Fli-1 most probably repress MEP production through a different mechanism. Interestingly enough, a recent study led strong support for a positive contribution of Notch signaling in the direct generation of MEPs by HSCs.⁸  Furthermore, activation of Notch signaling in HSCs has also been reported to stimulate the direct production of NK cells.⁴²  In that context, it is worth noting that our present study confirmed that Fli-1 loss also induced a significant increase in NK cells as already observed by others in chimeric Fli-1^−/− mice.²⁵  An exciting possibility suggested by our study, that could simultaneously explain both the increase in MEPs and NK cells induced by Fli-1 loss, would be therefore that Fli-1, as already reported for Ikaros,⁴³  might be a negative regulator of Notch signaling in HSCs.

In agreement with previous studies with chimeric Fli-1^−/− mice,²⁵  the present study confirmed that Fli-1 loss is also associated with decreased granulopoiesis in adult and further clarified the origin of this decrease. We found indeed that this decreased granulopoiesis was associated with a specific decrease of granulocytic progenitors and symetrical slight increase of monocytic progenitors. Furthermore, while the numbers and the overall clonogenicity of CMPs and GMPs were not affected by Fli-1 loss, both of them generated an increased number of monocytic colonies compensating the decrease in granulocytic colonies. Altogether, these new observations strongly suggest that Fli-1 loss favors the commitment of multipotent and bipotent progenitors toward the monocytic lineage at the expense of the granulocytic lineage. Interestingly, while G-CSF is known to instruct the commitment of bipotent GMPs toward granulocytic differentiation,⁴⁴  other studies have shown that G-CSF stimulation stabilizes Fli-1 protein.⁴⁵  Fli-1 can thus be added to the growing list of transcription factors that, like Irf8, Spi-1/Pu.1, Mafb, Cebpϵ, and Gfi1, contribute differentially to the granulocytic versus monocytic commitment of bipotent GMP in response to growth factors.^2,46 

The most drastic effect of Fli-1 loss was observed at the level of erythrocytic and megakaryocytic lineages. First, we found that Fli-1 loss drastically impaired late stages of megakaryocytic differentiation. This first point was evidenced by a drastic reduction in the number of large mature megakaryocytes present in bone marrow, a drastic reduction of their in vitro production either by total BMCs (CFU-MK) or by purified MEPs, as well as by our finding that the large majority of these few remaining mature megakaryocytes turned out to be undeleted. Furthermore, we found that Fli-1 loss reduced approximately 2-fold megakaryocytic commitment at the benefit of erythrocytic commitment. This second point was evidenced by the increased number of bone marrow BFU-E and most importantly by our finding that MEPs generated the same number of colonies with an increased proportion of erythrocytic versus megakaryocytic cells. Taken together, our results established that, in agreement with its expression profile, Fli-1 slightly favors megakaryocytic commitment at the expense of erythrocytic commitment and becomes increasingly required during late stages of megakaryocytes maturation. Interesting challenges for future studies will be to determine whether this differential requirement of Fli-1 simply reflects stoichiometric variations between different ETS members with completely redundant functions along the megakaryocytic differentiation process or is due to real nonredundant functions among the ETS family in the regulation of specific target genes and/or differential regulation of common target gene as previously suggested by others.²⁶ 

Lastly, by comparing the proliferation and erythrocytic differentiation of purified Fli-1–deleted and –undeleted MEPs, our study is the first to demonstrate the critical role of Fli-1 in the control of the proliferation/differentiation balance during normal erythropoiesis. Interestingly, this dual property of Fli-1 observed in normal erythropoiesis is quite similar to its role evidenced in erythroleukemic cells.^47,48  Moreover, Spi-1/Pu.1 has also been reported to play a similar role in erythroleukemic cells as well as during fetal erythropoiesis,^49,50  although its role during normal adult erythropoiesis has not been clearly addressed. An interesting perspective raised by all these data will be to know if this common property of Fli-1 and Spi-1/Pu.1 relies on the deregulation of common targets genes such as genes involved in ribosome biogenesis that we recently identified in erythroleukemic cells.⁴⁸ 

In conclusion, the present study demonstrates that Fli-1 controls the commitment decisions of several multipotent progenitors during adult hematopoiesis as well as proliferation/differentiation balance during normal erythropoiesis. We believe that our new transgenic mice allowing inducible Fli-1 gene deletion in adult mice should be a useful tool to further investigate the downstream molecular events involved in these new functions of Fli-1.

We are very grateful to W. Reith, D. Aubert, K. Rajewsky, A. de La Coste, M. Rassouzaldegan, and F. Nicolini for providing us with plasmids for the cloning of homologous recombination arms, ES cells clone ENS, Mx1-Cre mice, Sycp1-Cre mice, and recombinant EPO, respectively. We thank also A. Gy and S. Girard for their help in the initial establishment of recombinant ES cells, C. Pondarre for her help in blood cells analyses during the initial characterization of our transgenic mice phenotype, A. Dorier for his help with short-term mice housing during pIpC injections, and W. Vainchenker for helpful discussions. Highly expert assistance of C. Bella and S. Dussurgey at the IFR128 cell sorting platform is also greatly acknowledged.

This work has been supported by grants from the CNRS and Université Lyon 1 and by specific grants from the Ligue contre le Cancer (Labellisation 2005-2007, 2010-2012, and Comité du Rhône). F.M. is a member of Inserm. J.S., B.G., C.G., M.W.G., and J.M.V. are members of the CNRS.

Contribution: J.S., M.W.-G., and C.G. performed most of the experiments, interpreted the data, and assisted with the writing of the paper; B.G. performed cytological analyses of bone marrow megakaryocytes and assisted with initial FACS analyses and writing of the paper; J.-M.V. performed ES cell injections for the generation of founder mice and managed most of the mice crosses; and J.S. and F.M. designed the experiments, interpreted the data, and wrote the paper.

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

Correspondence: François Morlé, CNRS UMR5534, Centre de Génétique Moléculaire et Cellulaire, 16 rue Dubois, Université Lyon 1, Villeurbanne, F-69622, France; e-mail: francois.morle@univ-lyon1.fr.