Gfi-1B controls human erythroid and megakaryocytic differentiation by regulating TGF-β signaling at the bipotent erythro-megakaryocytic progenitor stage

Growth factor independence-1 (Gfi-1) and -1B (Gfi-1B) are homologous transcriptional repressors.¹  They share highly conserved Snail/Gfi1 (Snag) domain at their N-termini and 6 zinc finger domains at their C-termini.²  They bind to the same consensus sequence TAAATCAC(A/T)GCA,^2-4  and are both important regulators of hematopoiesis,^5-7  even though they are differentially expressed in the various hematopoietic cell populations. Gfi-1 and Gfi-1B expression are mutually exclusive.⁸  Whereas Gfi-1 is expressed in cells of the immune system, early B lymphocytes and T lymphocytes,^2,9  monocytes, and granulocytes,^6,10,11  Gfi-1B is rather found in erythroid and megakaryocytic cells.^12,13  Using Gfi-1B:green fluorescent protein (GFP) knock-in mice,¹⁴  it was highlighted that its expression is highly dynamic during erythropoiesis. We^12,15  and others¹³  have followed Gfi-1B expression in an ex vivo model of human erythropoiesis. Gfi-1B expression was higher as the number of erythroblasts increased during the culture of human CD34⁺ progenitors with erythropoietin (EPO) and remained sustained at the terminal differentiation stages. Recently, chromatin regulatory proteins (lysine-specific demethylase 1 [LSD1], co-RE1-silencing transcription factor [CoREST], and histone deacetylase) have been suggested to mediate transcriptional repression of Gfi-1B and its target genes.¹⁶ 

Inactivation of the GFI1B gene leads to embryonic lethality at day 15, as the result of arrest of erythroid maturation and failure to produce definitive enucleated erythrocytes.⁷  Examination of the blood from GFI1B^−/− embryos revealed an abnormal primitive erythropoiesis, and flow cytometry of E12.5 fetal liver cells showed an increase in the c-kit⁺, ter119⁺ cell population. It was concluded from this study that GFI1B deficiency induces a developmental arrest of erythroid progenitors in the fetal liver at the burst-forming unit erythroid (BFU-E) stage or earlier.

Although erythropoietin (EPO) is the main erythropoiesis-regulating growth factor, other cytokines play an important role in this developmental process. Among them is transforming growth factor-β (TGF-β), which is recognized by hematopoietic cells through 3 cell-surface receptors, namely, TGF-β type I, II, and III receptors (TβRI, TβRII, and TβRIII), all displaying different affinities for their ligand.¹⁷  In contrast to TβRI and TβRII, TβRIII is a membrane-anchored proteoglycan that does not possess an intrinsic kinase activity but enhances TβRI- and TβRII-mediated signaling by stabilizing their complex with TGF-β,¹⁸  regulating TGF-β receptor complex formation¹⁹  and internalization,^20,21  and/or induces Smad-independent signaling.^22-24  Its importance is emphasized by the fact that TGFBR3 knockout mice lead to embryonic lethality at day 16 resulting from liver and heart developmental defects, but their phenotype in terms of erythroid development remains unexplored.^25,26  However, it has been shown that, in immature adult hematopoietic progenitor cells, phospho-Smad2/3 associates with either Smad4 to inhibit progenitor proliferation or with TIF1-γ to promote erythroid differentiation.²⁷ 

In this report, we explore the function of Gfi-1B in primary human and immature adult progenitor cells. Using shRNA-mediated silencing, we show that Gfi-1B controls the development of erythroid cells and megakaryocytes at the bipotent erythro-megakaryocytic progenitor (MEP) stage. We further identified the TGFBR3 gene that encodes TβRIII, as a target for Gfi-1B in the MEP cell population and showed that Gfi-1B specifically binds to and represses the activity of the TGFBR3 gene promoter. Consistent with these findings, Gfi-1B depletion alters TGF-β signaling by compromising the formation of the phospho-Smad2/TIF1-γ complex that is required for erythroid differentiation. These data propose that Gfi-1B controls erythropoiesis and megakaryopoiesis at least in part by regulating TGFBR3 expression in MEP.

Human HEK293 and K562 cell lines were expanded in Dulbecco minimum essential medium containing 10% fetal calf serum, 2mM l-glutamine, and penicillin/streptomycin.

Human umbilical cord blood samples were collected from normal full-term deliveries, after informed consent of the mothers, according to the approved institutional guidelines of Assistance Publique-Hôpitaux de Paris. After a Ficoll-metrizoate gradient separation (Biochrom), CD34⁺ cells were enriched by an immunomagnetic selection (Miltenyi Biotec, purity ≥ 95%) and cultured for 4 to 5 days in Stem Span H3000 medium (StemCell Technologies) supplemented with 50 ng/mL human stem cell factor (SCF), 100 ng/mL Flt3-ligand (FL), 60 ng/mL interleukin-3 (IL-3), and 20 ng/mL thrombopoietin (TPO; Promocell Bioscience). For erythroid differentiation, CD34⁺ cells were grown in Stem Span H3000 medium containing 25 ng/mL SCF, 10 ng/mL IL-3, 10 ng/mL IL-6 (Promocell Bioscience), and 2 IU/mL EPO (a gift from Dr Brandt, Roche Diagnostics). For megakaryocytic differentiation, cells were cultured in medium containing 5 ng/mL SCF and 10 ng/mL TPO. Myeloid and megakaryocytic colony assays were performed as described.¹⁵ 

For cell-cycle analysis, sorted CD34⁺ GFP⁺ cells were fixed in 1% paraformaldehyde, permeabilized with 70% ethanol for 1 hour at −20°C, and then labeled with 1 mg/mL propidium iodide in the presence of 20 μL of RNAse at 10 mg/mL for 30 minutes at room temperature. Cell cycle–oriented histogram analysis was performed using FlowJo software. Phosphatidylserine exposure was measured by annexin V binding. Cells were incubated with 1 μg of phycoerythrin (PE)-coupled annexin V and 2 μL per 10⁴ cells viability dye 7-amino-actinomycin D (7-AAD) for 15 minutes at 37°C in 50mM N-2-hydroxyethylpiperazine-N′-2-ethanesulfonic acid (pH 7.4), 140mM NaCl, 1mM CaCl₂ buffer. The proportion of the cells in different quadrants was determined.

Three lentiviral vectors containing Gfi-1B shRNA were used (supplemental data, available on the Blood website; see the Supplemental Materials link at the top of the online article). The specific TβRIII shRNA was from Sigma-Aldrich (pLK0.1 TGFBR3 RNAi), and the shRNA control with a random sequence was generously given by Goardon et al.²⁸  Recombinant lentiviruses were produced as previously described.¹² 

Protein extracts were separated on sodium dodecyl sulfate-polyacrylamide gels, transferred to nitrocellulose membrane (Schleicher & Schuell), and immunoblotted with the appropriate antibody. The primary antibodies were specific of human Gfi-1, TβRI, TβRII, Smad2/3 and Smad4 (Santa Cruz Biotechnology), TβRIII (R&D Systems), High-Mobility Group Box protein HMGB2 (BD Biosciences PharMingen), phospho-Smad2 (Cell Signaling), p21^cip/waf1 (Oncogene Research Products), β-actin (Sigma-Aldrich), TIF1-γ,²⁷  and Gfi-1B, an antibody that was prepared in our laboratory.¹⁵  The peroxidase-conjugated secondary antibodies were specific of goat (Southern Biotechnology), mouse, or rabbit immunoglobulins (IgGs; Cell Signaling Technology).

Cell immunostaining was performed with PE-conjugated antihuman CD34 antibody, allophycocyanin-conjugated anti-CD36, and PE-conjugated antiglycophorin A (GPA) antibodies (Caltag Laboratories). Dead cells were excluded by propidium iodide staining. Appropriate isotype-matched controls were used to determine the background staining level. Cells were analyzed using FACSCalibur analyzer (Becton Dickinson Immunocytometry Systems).

Common myeloid progenitor (CMP), granulo-macrophagic progenitor (GMP), and MEP populations were sorted on FACSAria cell sorter (BD Biosciences) after PE.Cy5.5-coupled anti-CD34 (Caltag), biotinylated anti-CD38 (BioLegend) revealed by Pacific orange-conjugated streptavidin, PE-coupled anti-CD123 (BioLegend), and allophycocyanin-coupled anti-CD45RA (Caltag) labeling.

Extraction and reverse transcription (RT) of total RNA were performed as described.¹⁵  cDNA was amplified by quantitative polymerase chain reaction (PCR) using human TβRI, II, III, Gfi-1B, and glyceraldehyde 3-phosphate dehydrogenase (GAPDH) specific primers (supplemental data) according to the described thermal cycling program.¹⁵  The fold-change expression of each gene is represented by the ratio between the copy number of tested cDNA and the GAPDH cDNA calculated with the curve obtained with serial dilution of control cDNA.

Affymetrix analyses were performed with purified total RNA from shControl- and shGfi-1B-infected CD34⁺ cells and MEP. Gene expression profiling of each population was determined by hybridization on Affymetrix microarrays (Gene Chip Human Genome U133 Plus 2.0 array). The microarray dataset was deposited at ArrayExpress (http://www.ebi.ac.uk/arrayexpress; array accession no. E-MEXP-2560).

HEK293 cells were transiently transfected using Lipofectamine 2000 (Invitrogen) as described¹⁵  using plasmids containing the reporter luciferase gene under proximal TGFBR3 promoters²⁹  with or without Gfi-1B expression plasmid. pRL-TK (Renilla luciferase, Promega) and pGL2-luciferase were used as an internal control to correct transfection efficiency.

Oligonucleotide pull-down assays were performed as described¹⁵  with increasing amounts of wild-type or mutated double-strand biotin-labeled oligonucleotide corresponding to the TGFBR3 proximal promoter (supplemental data).

For immunoprecipitation assays, nuclear extracts from 10⁷ cells were precleared by addition of 2 μg of normal IgG for 2 hours. After incubation with 50 μL of protein G beads (GE Healthcare) and centrifugation, 2 μg of human Smad2/3 antibody or control IgG was added on the supernatant. The following day, 50 μL of protein G beads was added and protein complexes were analyzed by immunoblotting.

K562 cells were fixed, lysed, and sonicated as described.¹⁵  Immunoprecipitations were performed following the Upstate protocol (www.upstate.com) using a human Gfi-1B antibody (from our laboratory) and control rabbit IgG (Santa Cruz Biotechnology). PCR was performed on immunoprecipitated DNA as described¹⁵  using specific primers of TGFBR3 proximal promoter and a control region located upstream of the promoter (supplemental data).

Inactivation of GFI1B in mice leads to embryonic lethality, impeding the use of these knockout mice to study the role of Gfi-1B on adult hematopoiesis. We thus used shRNA-containing lentiviral vectors to down-regulate Gfi-1B expression in human primary immature progenitors. Amplified infected CD34⁺ cells were plated in erythroid (E0-E7) or megakaryocytic (T0-T10) liquid culture conditions and analyzed for the expression of Gfi-1, Gfi-1B, and p21 various days after infection. We observed that Gfi-1B was efficiently knocked down at all differentiation stages (from E0 to E6 and from T0 to T4, the result of one shRNA is shown), validating our experimental system to study the role of Gfi-1B in human adult erythropoiesis (Figure 1E). Interestingly, contrary to what was observed in control cells, Gfi-1 expression remained elevated during erythroid and megakaryocytic development of shGfi-1B–transduced cells, whereas Gfi-1B expression decreased in agreement with a recent report showing that Gfi-1B represses Gfi-1 transcription in erythroid and megakaryocytic cells.⁸  Furthermore, it is interesting to note that p21 expression was up-regulated in the absence of Gfi-1B. We thus evaluate the role of Gfi-1B in CD34⁺ cell proliferation by infecting them twice with lentiviruses carrying an shControl or shGfi-1B. Forty-eight hours after the second infection, CD34⁺ and GFP⁺ cells were sorted, cultured in the same amplification culture medium, and tested for their ability to grow in the same previous culture medium (D5-D9) (Figure 1A). Gfi-1B knockdown decreased the proliferation of immature CD34⁺ progenitor cells (Figure 1B). No apoptosis was observed in the cultures (Figure 1C). The proportion of the cells in G₀/G₁ phases of the cell cycle was more important at day 9 in Gfi-1B knockdown cells compared with control cells (Figure 1D). Equivalent results were obtained with 2 other Gfi-1B shRNA sequences (data not shown). Thus, in the absence of Gfi-1B, immature progenitors have a reduced cell growth.

To determine whether Gfi-1B is indispensable for terminal erythroid and megakaryocytic differentiation, CD34 cells were plated in the presence of EPO, IL-3, IL-6, and SCF (E0-E7) to induce erythroid differentiation, or in the presence of TPO and SCF (T0-T10) to induce megakaryocytic differentiation. We observed by cytofluorometry that the proportion of CD36⁺GPA⁺ cells, which increased on EPO stimulation in control transduced cells (from 1.8% to 57.4%), did not show such an important increase in shGfi-1B–transduced cells (from 1.2% to 24.2%) (Figure 2A). In agreement with this result, cytology performed on May-Grunwald-Giemsa–stained cytospins showed that, 7 days on induction of differentiation, the percentage of immature erythroblasts (ER; pro- and basophilic; 64.1 ± 15 vs 13.0 ± 7.1, mean ± SD, n = 4) as well as the percentage of differentiated erythroblasts (polychromatophilic and acidophilic Er; 22.0% ± 15.5% vs 1.5% ± 0.7%, mean ± SD, n = 4) were decreased in Gfi-1B–depleted cells (Figure 2B). Similarly, after 10 days of culture in the presence of TPO, shGfi-1B cells gave rise to decreased amounts of megakaryocytes than shControl cells (72% vs 14%; Figure 2D). Accordingly, shGfi-1B–transduced megakaryocytes did not display proplatelet-like protrusions compared with shControl-transduced cells (Figure 2C). Equivalent results were obtained with the 2 other Gfi-1B shRNA (supplemental Table 1). Interestingly, although Gfi-1B depletion led to an increase of Gfi-1 expression in erythroid and megakaryocytic cells (Figure 1E), such increase was not sufficient to drive normal erythropoiesis and megakaryopoiesis. We conclude that Gfi-1B affects proliferation in immature progenitor cells and is specifically required for terminal human erythroid and megakaryocytic differentiation.

We next studied the effect of Gfi-1B knockdown on the different progenitor compartments to investigate whether the requirement of Gfi-1B for terminal erythroid and megakaryocytic differentiation results from its function in immature progenitors. For this, 2 to 3 days on transduction, the CD34⁺GFP⁺ cell population was cultured in a semisolid medium to allow progenitor development. A drastic decrease in the BFU-E number was observed when plating shGfi-1B-transduced cells in methylcellulose compared with shControl-transduced cells (23.5 ± 8 vs 122.0 ± 45, n = 8; Figure 3A). Similarly, shGfi-1B cells plated in semisolid medium with megakaryocytic cytokines gave rise to very few colony-forming units-megakaryocyte (CFU-MK) containing CD41⁺ cells, compared with control cells (10.1 ± 2 CFU-MK vs 39.3 ± 4, n = 6; Figure 3B). Importantly, no difference was observed in the number of granulo-macrophagic colonies generated from the Gfi-1B knocked down population, progenitors that are known to not rely on Gfi-1B for their differentiation (Figure 3A). Importantly, the same decrease in erythroid and megakaryocytic progenitors was observed with the 2 other Gfi-1B shRNA (supplemental Table 1). Hence, Gfi-1B knockdown leads to a decrease in the amount of erythroid and megakaryocytic colonies derived from BFU-E and CFU-MK.

Importantly, such results do not formally demonstrate whether Gfi-1B is indeed required at the progenitor stage because a blockade in late differentiation would prevent colony formation and, thus, BFU-E and CFU-MK colony detection. To circumvent this problem, we thus evaluated by flow cytometry the size of the 3 progenitor compartments: CMP (CD123^lowCD45RA⁻), GMP (CD123^lowCD45RA⁺), and MEP (CD123⁻CD45RA⁻).³⁰  We observed that, whereas the proportion of CMP slightly increased in the absence of Gfi-1B (41.4% vs 42.8%), the percentage of GMP increased (7.2% vs 14.8%) and the percentage of MEP decreased (30.7% vs 18.7%; Figure 3C). Although these populations were not cell homogeneous, this result can be interpreted in 2 ways: Gfi-1B deficiency (1) favors the commitment of CMP toward the granulo-macrophagic lineage or (2) impedes erythroid development, implying that the increase in the percentage of GMP is the result of the decrease in the percentage of MEP. To distinguish between these 2 possibilities, we determined the number of progenitors present in the sorted shControl- or shGfi-1B–infected CMP, GMP, and MEP cell populations (Figure 3D-E). Both shControl and shGfi-1B GMP populations gave rise to similar numbers of G-, GM-, and M-CFC colonies (Figure 3D). In contrast, decreased BFU-E and CFU-MK were generated from both shGfi-1B CMP and MEP populations, highlighting the differences in the erythroid and megakaryocytic potential of Gfi-1B–sufficient and –deficient cells (Figure 3D-E). Importantly, introduction of a mouse Gfi-1B in Gfi-1B knockdown CD34⁺ cells restored the cellular phenotype at the progenitor and differentiated stages (supplemental Figure 2). We conclude that Gfi-1B is not involved in the commitment of CMP toward erythro-megakaryocytic or granulo-macrophagic development but controls the size and differentiation of the MEP population.

In agreement with this result, we found that Gfi-1B was highly expressed in the MEP population, whereas it was expressed at low level in CMP and not expressed in GMP (Figure 3F). Gfi-1B was barely detected in the CD34⁺ cell population that is heterogeneous and contains the 3 CMP, GMP, and MEP cell populations. In contrast, Gfi-1 was poorly expressed in MEP and highly expressed in CMP and GMP, confirming results obtained by others in mice.¹⁴  Thus, among progenitors, MEP express the highest amount of Gfi-1B and their number is altered in the absence of Gfi-1B, strongly suggesting that Gfi-1B plays a key role in the maintenance and development of these progenitor cells.

Having shown that the transcription repressor Gfi-1B regulates the size and differentiation of the MEP compartment, we next aimed at identifying its target genes in this cell population. We used microarrays (Affymetrix 20.1) to compare the transcriptome of MEP or CD34⁺ cells infected with lentiviruses carrying either the shGfi-1B or the shControl. We found that only 146 genes were up-regulated in shGfi-1B CD34⁺ cells compared with shControl CD34⁺ cells. In contrast, 1852 genes are up-regulated in shGfi-1B MEP compared with their shControl counterpart, in agreement with their high Gfi-1B expression level and the highest homogeneity of the cell population (Figure 4A). Among these genes, 60 were found to be up-regulated in both Gfi-1B–depleted CD34⁺ and MEP populations by comparison of Affymetrix data. Interestingly, this group of 60 genes included several genes encoding proteins involved in TGF-β signaling (Figure 4B). In particular, the expression of the type III TGF-β receptor (TβRIII or betaglycan) was up-regulated in CD34⁺ and MEP populations lacking Gfi-1B, suggesting that Gfi-1B represses the transcription of this gene in immature progenitors.

To verify these Affymetrix data, we prepared mRNA from shControl- and shGfi-1B–transduced MEP and CD34⁺ cells and analyzed, by quantitative RT-PCR, the presence of the 3 TGF-β receptor transcripts. We reproducibly observed that Gfi-1B knockdown led to a strong up-regulation in the levels of TβRIII mRNA in both CD34⁺ cells (4.6- ± 0.9-fold increase) and MEP (2.6- ± 0.4-fold increase), whereas the levels of TβRI and TβRII transcripts were barely modified at the mRNA (Figure 4C) and protein levels (supplemental Figure 1). We conclude that Gfi-1B negatively regulates the transcription of the TGFBR3 gene in human primary erythroid and megakaryocytic progenitors.

To investigate whether Gfi-1B–dependent transcriptional repression of TGFBR3 is direct or not, we assessed the ability of Gfi-1B to bind to and to regulate the activity of the TGFBR3 promoter. To do so, we switched to K562 cells, which represent a valuable cell line model for immature MEPs³¹  and which respond to TGF-β by concomitant induction of hemoglobin and reduction in cell proliferation.³²  Accordingly, as observed in CD34⁺ cells and MEP population, TβRIII expression was increased at both the mRNA and protein levels in shGfi-1B–transduced K562 cells (Figure 5A-B), whereas the protein expression of TβRI and TβRII remained unchanged (Figure 5B). Two important regulatory regions have been described in the TGFBR3 promoter²⁹ : (1) the proximal region and (2) the distal region, located at 25 kb and 45 kb from the translation initiation codon, respectively (Figure 5C). To investigate the regulation of the TGFBR3 promoter activity by Gfi-1B in the regulation of TGFBR3 transcription, we transfected the wild-type distal or the wild-type proximal promoter luciferase constructs into HEK293 cells together with or without a Gfi-1B expression plasmid. Gfi-1B expression reduced the activity of the proximal TGFBR3 promoter by 45% (Figure 5D) but had no effect on the activity of the distal one (data not shown). Sequence analysis identified 4 motifs AATC/GATT in the proximal promoter sequence (−525, −474, −134, and −55). To define the role of the different Gfi-1B putative binding sites in the proximal region of the TGFBR3 promoter, we transfected plasmids in which −165/+60 or −75/+60 region was cloned upstream of the luciferase gene.²⁹  Gfi-1B induced a significant reduction of luciferase activity when luciferase gene expression was driven by the −559/+60 and −165/+60 TGFBR3 promoters but not when using the −75/+60 promoter. These results indicate that the −134 Gfi-1B binding site, which is very well conserved through species (data not shown), plays an important role in the transcriptional repression of the TGFBR3 promoter by Gfi-1B. We conclude that Gfi-1B can repress the activity of the TGFBR3 promoter by acting on its proximal regulatory element.

We next verified whether Gfi-1B binds to the proximal region of the TGFBR3 promoter in K562 cells by performing DNA affinity precipitation experiments. We thus used biotinylated oligonucleotides representing the wild-type or mutated region of the TGFBR3 promoter that enclosed the −134 binding site (Figure 5E). Strikingly, we found that Gfi-1B binds in a dose-dependent manner to the wild-type oligonucleotide corresponding to the proximal TGFBR3 promoter sequence, whereas no binding was detected when using the oligonucleotide mutated at the Gfi-1B consensus binding site (Figure 5E). This result and the fact that 2 different groups^3,4  characterized by the random oligonucleotide binding-selection strategy the Gfi-1B binding site as being AATC (or GATT) in 100% of the cases strongly suggest that GATT is a nondegenerated Gfi-1B consensus-binding site. Because Gfi-1 is not expressed in K562, we did not find Gfi-1 binding to these oligonucleotides. We conclude that, in vitro, Gfi-1B binds to the proximal TGFBR3 promoter.

To verify these results in vivo, we performed chromatin immunoprecipitation (ChIP) experiments in shControl- or shGfi-1B–transduced K562 cells. Chromatin was immunoprecipitated using antibodies directed against Gfi-1B or control IgG. Primers amplifying a specific region of the TGFBR3 promoter (−0.15 kb) or a region located upstream of the promoter (−2.3 kb) were used. The result showed that Gfi-1B binds to the TGFBR3 proximal promoter (Figure 5F). This interaction was specific as shown by a significant binding decrease in Gfi-1B knocked down cells (n = 3, P < .001). No enrichment was found on the region located upstream the promoter (−2.3 kb). Hence, the TGFBR3 promoter is a target for Gfi-1B–mediated transcriptional repression in bipotent MEP cells.

To establish whether increase TGFBRIII expression is responsible for the suppressed erythroid and megakaryocytic differentiation in Gfi-1B knockdown cells, we performed rescue experiments using lentiviral vectors carrying an shRNA against TβRIII in Gfi-1B–depleted CD34 cells. CD34⁺ cells infected with lentivirus containing a Gfi-1B shRNA (including a GFP sequence) alone or together with TβRIII shRNA (including a puromycin sequence). Infected cells were selected with puromycin, and GFP⁺ cells were sorted by fluorescence-activated cell sorter (FACS; Figure 6A). The efficiency of the TβRIII shRNA was shown on K562 cells, which expressed abundant endogenous TβRIII (Figure 6B). Whereas Gfi-1B–depleted cells gave rise to few BFU-E (68.0 ± 4) or CFU-MK (11.0 ± 3) per 1000 cells plated, cells with both Gfi-1B and TβRIII knockdown gave rise to the same numbers of progenitors as shControl-infected cells (179 ± 9 BFU-E and 36.5 ± 2.5 CFU-MK for the control and 169 ± 3 BFU-E and 28.5 ± 0.5 CFU-MK for cells depleted for both Gfi-1B and TβRIII). Interestingly, depletion of TβRIII alone did not affect erythroid and megakaryocytic progenitor development (Figure 6C-D). GPA expression analyses showed that, in the absence of both Gfi-1B and TβRIII, erythroid maturation developed normally (Figure 6E). This result was confirmed by benzidine staining: 5 days after induction of erythroid differentiation, the same percentage of benzidine-positive cells was observed in control and in shGfi-1B⁺ shTβRIII-transduced cells (data not shown). Thus, knockdown of TβRIII in Gfi-1B–depleted cells rescues the erythroid and megakaryocytic differentiation, indicating that Gfi-1B drives erythroid and megakaryocytic development by repressing TGFBR3 transcription.

Our results show that, in the absence of Gfi-1B, TGFBR3 transcription is up-regulated in immature human MEPs. We thus hypothesized that this increase in TβRIII expression leads to altered TGF-β signaling. To address this question, we analyzed the activation of the canonical TGF-β pathway in shControl- or shGfi-1B–transduced K562 cells.²⁷  Whereas K562 cells had elevated basal levels of phosphorylated Smad2, probably resulting from Brc-Abl expression, which is known to constitutively activate many signaling pathways, Gfi-1B knockdown led to a significant increase in the phosphorylation of Smad2 in immature K562 cells, which paralleled the increase in TβRIII expression (Figure 7A). Hence, TGF-β signaling is altered in MEP cells that lack Gfi-1B.

So far, we have shown that Gfi-1B knockdown led to both altered TGF-β signaling and impaired erythroid and megakaryocytic differentiation. We thus hypothesized that Gfi-1B–mediated regulation of TGF-β signaling might contribute to its function in erythroid development. Indeed, TGF-β signaling was described as triggering erythroid differentiation from immature progenitors through the formation of a protein complex between phosphorylated Smad2/3 and TIF1-γ. To address this question, we investigated whether the formation of the complex between phosphorylated Smad2/3 and TIF1-γ was modified in K562 cells depleted for Gfi-1B by performing coimmunoprecipitation experiments. Strikingly, we found that, whereas phosphorylated Smad2 associated to both Smad4 and TIF1-γ in shControl cells, complexes between phosphorylated Smad2 and TIF1-γ did not form in Gfi-1B knockdown cells (Figure 7B). Thus, depletion of Gfi-1B specifically prevents the formation of the Smad2/TIF1-γ complex, which is known to be required for erythroid differentiation²⁷  but does not affect the association between Smad2 and Smad4, which is responsible for the inhibition of cell proliferation. Consistent with this result, the proliferation of immature CD34⁺ progenitors was indeed inhibited in the absence of Gfi-1B (Figure 1B).

Altogether, our data suggest that Gfi-1B controls erythroid and megakaryocytic development at the bipotent MEP stage, at least in part, by regulating TGF-β signaling.

We here show that Gfi-1B, a zinc-finger transcription factor, is essential at early stages of human adult hematopoietic differentiation and more specifically at the bipotent MEP stage. Gfi-1B is highly expressed in the MEP cell population; and, in its absence, a reduction in proliferation and an inhibition of differentiation of immature progenitors are concomitantly observed.

Interestingly, whereas erythroid progenitors from fetal liver of GFI-1B^−/− mice form large dispersed colonies of immature progenitors,⁷  shGfi-1B MEPs do not proliferate or differentiate in semisolid medium in response to a cocktail of growth factors. These results suggest that Gfi-1B target genes may be different in mouse embryonic and human adult stem cells. The existence of potential differences between human and mouse Gfi-1B functions is further supported by the fact that Gfi-1 expression, which is maintained in Gfi-1B–depleted CD34 cells, fails to restore the Gfi-1B deficiency. This result is indeed in sharp contrast with data showing that GFI1/GFI1B knock-in mice display normal hematopoiesis.³³  However, the development of the inner ear hair cell is altered in these mice, suggesting that Gfi-1 and Gfi-1B may have some different functions. In particular, one Gfi-1 function that is not shared by Gfi-1B is to enhance STAT3-mediated transactivation by binding and sequestrating the STAT3 inhibitor PIAS3.³⁴  Furthermore, we¹⁵  and others¹⁶  have shown that LSD1/CoREST mediated the repression induced by Gfi-1 and Gfi-1B. We found that, contrary to Gfi-1B, Gfi-1 did not recruit LSD1/CoREST to the TGFBR3 promoter (data not shown). Thus, it seems probable that Gfi-1 cannot replace Gfi-1B in its specific function in human immature hematopoietic progenitor cells.

Comparative transcriptomic analyses of Gfi-1B knocked down CD34⁺ or MEP show that, among the 60 genes that are up-regulated in both CD34⁺ cells and MEP depleted for Gfi-1B, several genes encode proteins involved in TGF-β signaling. In particular, the expression of one of its receptors, the coreceptor TβRIII, was strongly increased both in CD34⁺ cells and in MEPs depleted for Gfi-1B, suggesting that Gfi-1B modulates TGFBR3 gene expression in immature MEPs. In accordance with the up-regulation of TGFBR3 transcription in the absence of Gfi-1B, we found that Gfi-1B binds to the proximal part of the TGFBR3 promoter and represses its activity. This result, together with the rescue of the erythroid and megakaryocytic maturation observed after the TβRIII knockdown in Gfi-1B–depleted CD34⁺ cells, indicates that TβRIII repression is responsible for Gfi-1B–mediated regulation of erythropoiesis and megakaryopoiesis.

TGF-β superfamily members have indeed been described as regulating erythropoiesis mainly through Smad pathways.^35,36  Here we show that Smad signaling pathway is altered in the absence of Gfi-1B in erythroid cells. Indeed, the knockdown of Gfi-1B increases Smad2 phosphorylation. Surprisingly, TGF-β stimulation does not increase the phosphorylation of Smad2 in K562 cells. However, many signaling pathways are constitutively activated in K562 cells because of the presence of Bcr-Abl; this result suggests that TGF-β signaling pathway is also activated in K562 cells. The increase in Smad2 phosphorylation that follows the increase of TβRIII expression in the absence of Gfi-1B may be the result of the enhanced activation of TβRI, which is responsible for Smad2 phosphorylation. A striking observation we made is that, whereas both Smad2/Smad4 and Smad2/TIF1-γ complexes form in the presence of Gfi-1B, formation of Smad2/TIF1-γ is completely impaired in Gfi-1B–depleted erythroid cells. This result provides a simple explanation for our data showing that Gfi-1B knocked down cells exhibit (1) an inhibition in progenitor proliferation, which is known to result from Smad2/Smad4 association and (2) an impairment in erythroid development, which was described as requiring Smad2/TIF1-γ association.²⁷  In the same line, knockdown of TF1γ in immature hematopoietic leads to altered erythroid differentiation,²⁷  and mutations in the moonshine gene that encodes the zebrafish ortholog of mammalian TIF1-γ were shown to specifically disrupt both embryonic and adult hematopoiesis, resulting in severe red blood cell aplasia.³⁷ 

Why is the formation of Smad2/TIF1-γ complexes compromised in cells lacking Gfi-1B? It is tempting to speculate that increased TβRIII expression in the absence of Gfi-1B is in part responsible for the lack of Smad2/TIF1-γ association. Enhanced Smad2 phosphorylation resulting from TβRIII overexpression may specifically impair its interaction with TIF1-γ, whereas its association to Smad4 would remain unaffected. Alternatively, TβRIII overexpression might modify an additional component of the pathway that would regulate Smad2/TIFγ interaction. Gfi-1B, through regulation of TβRIII expression, may also modify TGF-β signaling through alternative non-Smad pathways. Indeed, TβRIII has been demonstrated to regulate the MAPK p38,²³  the small GTPase Cdc42,²²  and the nuclear factor-κB (NF-κB) pathways²⁴  through its interaction with β-arrestin 2.²²  Accordingly, among the genes up-regulated in cells deficient for Gfi-1B, we found Egr-1³⁸  and Rho-B,³⁹  also known to participate to this non-Smad TGF-β signaling. In addition, TβRIII also binds inhibin⁴⁰  and bone morphogenetic protein⁴¹  to regulate both activin and bone morphogenetic protein-mediated signaling.^41-43  In any case, the current data suggest that the expression balance between TGF-β receptors might be a key parameter in controlling cellular responses to TGF-β signaling during hematopoiesis. The nature of the factors that regulate TβRIII expression in addition to Gfi-1B and the TβRIII function responsible for the role defined here is currently under investigation.

In conclusion, we here propose that Gfi-1B represses TβRIII in MEPs and thereby controls the ability of these cells to differentiate in response to TGF-β. This study provides the first evidence for a crosstalk between a transcriptional repressor and a cytokine signaling pathway during erythropoiesis and megakaryopoiesis.

The authors thank Dr J. Massagué (New York, NY) for the TIF1-γ antibody, Dr E. Soler (Rotterdam, The Netherlands) for Gfi-1B shRNA, Dr A.M. Lennon-Duménil (Paris, France) for critical reading of the manuscript, Dr L. Bénit (Paris, France) for analyzing DNA sequences, Dr E. Lauret (Paris, France) for helpful discussion, F. Letourneur and N. Cagnard (Paris, France) from “Séquençage et Génomique” and informatics platforms for hybridization on Affymetrix microarrays and analysis of the results, and B. Chanaud, L. Stouvenel, and M. De Sousa (Paris, France) from Flow Cytometry platform of Cochin Institute.

This work was supported by Inserm, France and Fondation de France (grant 20007002071). B.L. was supported by a fellowship from Association pour la Recherche contre le Cancer.

Contribution: V.R.-H. and B.L. conceived, designed, and performed the research, analyzed and interpreted data, and contributed to writing the manuscript; V.B. analyzed and interpreted cytology and approved the final manuscript; G.C.B. provided plasmid constructs and approved the final manuscript; F.H. performed FACS analysis and cell sorting purification, provided critical ideas, and approved the final manuscript; and D.D. conceived and performed the research, analyzed and interpreted data, coordinated the research, and wrote the manuscript.

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

Correspondence: Dominique Duménil, Institut Cochin, Batiment G. Roussy, 27 rue du Faubourg Saint-Jacques, 75014 Paris, France; e-mail: dominique.dumenil@inserm.fr.