Ineffective erythropoiesis in myelodysplastic syndromes: correlation with Fas expression but not with lack of erythropoietin receptor signal transduction


  • Michaëla Fontenay-Roupie,

    1. Département d'Hématologie, AP-HP, INSERM U363, Université René Descartes, Hôpital Cochin, Paris, France,
    2. Institut Cochin de Génétique Moléculaire (ICGM), INSERM U363, Université René Descartes, Hôpital Cochin, Paris, France
    Search for more papers by this author
  • Didier Bouscary,

    1. Département d'Hématologie, AP-HP, INSERM U363, Université René Descartes, Hôpital Cochin, Paris, France,
    2. Institut Cochin de Génétique Moléculaire (ICGM), INSERM U363, Université René Descartes, Hôpital Cochin, Paris, France
    Search for more papers by this author
  • Martine Guesnu,

    1. Département d'Hématologie, AP-HP, INSERM U363, Université René Descartes, Hôpital Cochin, Paris, France,
    Search for more papers by this author
  • Françoise Picard,

    1. Département d'Hématologie, AP-HP, INSERM U363, Université René Descartes, Hôpital Cochin, Paris, France,
    Search for more papers by this author
  • Josiane Melle,

    1. Département d'Hématologie, AP-HP, INSERM U363, Université René Descartes, Hôpital Cochin, Paris, France,
    Search for more papers by this author
  • Catherine Lacombe,

    1. Département d'Hématologie, AP-HP, INSERM U363, Université René Descartes, Hôpital Cochin, Paris, France,
    2. Institut Cochin de Génétique Moléculaire (ICGM), INSERM U363, Université René Descartes, Hôpital Cochin, Paris, France
    Search for more papers by this author
  • Sylvie Gisselbrecht,

    1. Institut Cochin de Génétique Moléculaire (ICGM), INSERM U363, Université René Descartes, Hôpital Cochin, Paris, France
    Search for more papers by this author
  • Patrick Mayeux,

    1. Institut Cochin de Génétique Moléculaire (ICGM), INSERM U363, Université René Descartes, Hôpital Cochin, Paris, France
    Search for more papers by this author
  • François Dreyfus

    1. Département d'Hématologie, AP-HP, INSERM U363, Université René Descartes, Hôpital Cochin, Paris, France,
    2. Institut Cochin de Génétique Moléculaire (ICGM), INSERM U363, Université René Descartes, Hôpital Cochin, Paris, France
    Search for more papers by this author

Dr M. Fontenay-Roupie Laboratoire d'Hématologie, Hôpital Cochin, 27 rue du Faubourg Saint-Jacques, F75679 Paris Cedex 14, France.


Ineffective erythropoiesis in myelodysplasia is characterized by a defect in erythroid progenitor growth and by abnormal erythroid differentiation. Increased apoptosis of erythroid, granulocytic and megakaryocytic lineages is thought to account for cytopenias. Erythropoietin (Epo)-induced BFU-E and CFU-E growth was studied in 25 myelodysplastic syndrome (MDS) marrow specimens and found to be drastically diminished. To investigate the functionality of Epo-R in MDS marrow, we focused on Epo-induced STAT5 activation. Epo was able to stimulate STAT5 DNA binding activity in all normal and 12/24 MDS marrows tested, with no correlation between the level of STAT5 activation and the development of erythroid colonies in response to Epo. In contrast, impaired proliferation of erythroid progenitors was related to an increased expression of the transmembrane mediator of apoptotic cell death Fas/CD95 on the glycophorin A+ subpopulation. Therefore we conclude that the stimulation of pro-apoptotic signals rather than the defect of anti-apoptotic pathways resulting from Epo-stimulated Jak2-STAT5 pathway, predominantly accounts for ineffective erythropoiesis in myelodysplasia.

Erythropoietin (Epo) is the growth factor required for normal erythroid cell proliferation, survival and differentiation (Krantz, 1991). Epo is dispensable for burst-forming unit-erythroid (BFU-E) generation and BFU-E differentiation to colony-forming units-erythroid (CFU-E) (Wu et al, 1995). However, CFU-E are absolutely dependent on Epo and/or Epo receptor (Epo-R) for their differentiation (Wu et al, 1997), and Epo deprivation leads to apoptosis (Koury & Bondurant, 1990). Epo binds to the Epo-R, a member of the cytokine receptor superfamily with no intrinsic tyrosine kinase activity. Ligand binding is followed by Epo-R dimerization and activation, within minutes, of the Epo-associated intracellular Janus kinase Jak2 (Witthuhn et al, 1993). This results in tyrosine phosphorylation of the Epo-R (Dusanter-Fourt et al, 1992) and of several intracellular substrates. The signal transducer and activator of transcription (STAT)5 binds to the activated Epo-R, becomes phosphorylated in a Jak2-dependent manner (Damen et al, 1995; Gobert et al, 1996) and translocates to the nucleus. Whether Epo-induced STAT5 activation is a positive signal for either proliferation/survival or differentiation of the erythroid progenitors is still a matter of debate.

The myelodysplastic syndromes (MDS) are a group of clonal preleukaemic diseases characterized by impaired haemopoiesis. Erythropoiesis is severely compromised leading to anaemia and transfusion requirement. The mechanisms of ineffective erythropoiesis may be produced by a defect of erythroid progenitor growth, abnormal erythroid differentiation and increased apoptosis of erythroid progenitors. We and several other groups have recently confirmed the increase of apoptotic cells in the bone marrow of MDS (Raza et al, 1995; Bouscary et al, 1997; Parker & Mufti, 1998). Tumour necrosis factor (TNF)-α and interferon (IFN)γ are thought to participate in this apoptotic process (Kitagawa et al, 1997). Additionally, we previously reported that the transmembrane receptor Fas (CD95/Apo1), a member of the TNF-α receptor family able to transduce a cell death signal after binding of its ligand, was highly expressed on MDS-deriving bone marrow cells bearing the CD34, CD33 or glycophorin A (GPA) antigens (Bouscary et al, 1997). In vitro culture studies in MDS have established that erythroid progenitors and mainly BFU-E have an impaired capability to expand in response to Epo (Backs et al, 1992; Brada et al, 1996). Epo binding to MDS bone marrow mononuclear cells (BMMNC) is normal and no structural abnormalities of the Epo-R have been described (Backs et al, 1996). Therefore, a defect in Epo-induced signal transduction could account for impaired erythropoiesis. To address this question, we investigated the erythropoiesis of a group of 25 cases of MDS. Impairment of erythroid progenitor growth was found to be statistically correlated to an elevated Fas expression on the GPA-positive subpopulation but did not correlate with the level of STAT5 activation following Epo addition.


Patient material

Bone marrow (BM) cells were collected, after informed consent, from 25 patients with MDS classified according to the criteria of the French–American–British (FAB) group as refractory anaemia (RA, n = 4), refractory anaemia with ring sideroblasts (RARS, n = 9), refractory anaemia with excess of blasts (RAEB) and RAEB in transformation (RAEBt) (n = 12). Normal BM cells were obtained from 11 donors, after informed consent. BM mononuclear cells (BMMNC) were isolated on Ficoll-Hypaque and immediately used for further experiments.

Antibodies and inhibitors

For supershift experiments we used either an anti-STAT5 monoclonal antibody (MoAb) directed to the N-terminal part of human STAT5 (Upstate Biotechnologies, Lake Placid, U.S.A.) or a mixture (1:1) of two polyclonal antibodies recognizing the C-terminus of human STAT5A and STAT5B. These latter antibodies were produced by immunizing rabbits with peptides corresponding to the 12 C-terminal amino-acids (AGLFTSARSSLS) of STAT5A or the eight C-terminal amino-acids (QWIPHAQS) of STAT5B coupled to KLH (Verdier et al, 1998).

The proteasome inhibitor, N-acetylleucylleucylnorleucinal (LLnL) and the calpain inhibitor N-acetylleucylleucylnormethioninal (LLnM) used as control were from SIGMA (L'Isle-d'Abeau, France).

Erythroid colony cultures

BMMNC were seeded in duplicate at 105 cells/ml in 0.8% methylcellulose medium containing 10% fetal calf serum, 0.9% bovine serum albumin (BSA), 1.7 mm glutamine, and 0.9 × 10−4 mβ-mercaptoethanol for 14 d at 37°C under humidified atmosphere (5% CO2/95% air). BFU-E and CFU-E growth were assayed in culture supplemented with 4 IU/ml Epo (Boehringer Mannheim, Germany) and 100 IU/ml interleukin (IL)-3 (a generous gift from S. Shimosaka, Kirin Brewery Company Ltd, Tokyo, Japan). CFU-E and BFU-E-derived colonies were numbered at days 7 and 14, respectively.

STAT5 activation

After 2 h of incubation in Iscove's medium containing 0.4% BSA and 2 μg/ml holotransferrin, BMMNC (at a concentration of 1 × 106/ml) were stimulated with 10 IU/ml Epo for either 10 min or 60 min at 37°C. In some cases cells were preincubated with 50 μm LLnL or LLnM for 15 min at 37°C before stimulation with 10 IU/ml Epo for 10 or 60 min according to Verdier et al (1998). The reaction was stopped by the addition of ice-cold phosphate-buffered saline (PBS), pH 7.40, and nuclear extracts were prepared as previously described (Gobert et al, 1996). Briefly, cells were collected by centrifugation and lysed in buffer A (HEPES 20 mm, pH 7.8, KCl 10 mm, Na3VO4 1 mm, EDTA 1 mm, Nonidet P-40 0.2%, glycerol 10%, phenylmethylsulphonylfluoride, leupeptin, aprotinin, pepstatin as protease inhibitors and dithiotreitol 1 mm) for 5 min on ice. Nuclear fractions were collected by centrifugation at 10 000 g for 5 min and extracted for 30 min in lysis buffer B (NaCl 350 mm, glycerol 20%, HEPES 20 mm, pH 7.8, KCl 10 mm, Na3VO4 1 mm, EDTA 1 mm and the same protease inhibitors). Nuclear extracts were clarified by centrifugation at 10 000 g for 8 min and frozen at −80°C until used. Electrophoretic mobility shift assay (EMSA) was performed as followed: 4 μl of nuclear extract and 1 μl of γ[32P]ATP (specific activity: 111 TBq/mmol)-labelled synthetic oligonucleotide probe derived from the β-casein gene promoter (5′-AGATTTCTAGGAATTCAATCC-3′) were added to 15 μl of mix containing 100 μg/ml of poly(dI-dC) competitor. Samples were incubated for 30 min at 4°C in binding buffer (HEPES 20 mm pH 7.80, EDTA 0.2 mm, NaCl 100 mm, KCl 10 mm, MgCl2 10 mm, spermidin 8 mm, dithiotreitol 4 mm, bovine serum albumin 200 μg/ml, glycerol 5% and Ficoll 8%). Oligonucleotide-protein complexes were separated on a 5% polyacrylamide gel in 0.25× Tris-Borate/EDTA buffer. For supershift assays, specific antibodies were added 1 h before incubation with the oligonucleotide probe. UT-7 cells and FDCP-1 cell line transfected with the Epo-R were used as controls (Gobert et al, 1996; Verdier et al, 1998).

Cell-cycle analysis

Cell cycle was determined at days 0, 1 and 2 of a short-term liquid culture. 105 BMMNC/ml were seeded in serum-free IMDM supplemented with 0.4% BSA and 2 μg/ml holotransferrin. Cell cycle was analysed by staining 105 permeabilized cells with propidium iodide using the DNA Coulter preparation kit (Beckman Coulter, Miami, Florida). Results were analysed by a cell cycle dedicated software and expressed as percentages of S- and sub-G1-phase cells, the latter corresponding to necrotic and apoptotic cells.

Analysis of apoptosis

Apoptosis was detected on BM smears using the TUNEL technique (terminal deoxynucleotidyl transferase [TdT]-mediated dUTP Nick End Labelling) with the In Situ Cell Death Detection Kit POD (Boehringer Mannheim, Germany). After thawing for 1 h, the smears were fixed in a paraformaldehyde solution (4% in PBS) for 30 min at room temperature. Endogenous peroxidase was inhibited with 0.3% H2O2 in methanol for 30 min at room temperature. Slides were rinsed twice in PBS and incubated in permeabilization solution (0.1% Triton X100 in 0.1% sodium citrate) for 2 min on ice. The Tunel reaction mixture (TdT, nucleotides containing fluorescein-labelled dUTP in reaction buffer) was added on samples. Smears protected with coverslips were incubated for 1 h at 37°C. In negative controls, TdT was ommitted. The preparations were then incubated with anti-fluorescein antibody conjugated with horseradish peroxydase for 30 min at 37°C. Finally, samples were stained for 5 min using 3.3′-diaminobenzidine tetrahydrochloride diluted in 0.05 m TRIS buffer, pH 7.8, with 0.05% H2O2, shaded from light. Slides were rinsed three times in PBS and stained with Gill's haematoxylin. Samples were analysed by light microscope and cells in which staining was confined to the nuclear periphery were considered negative. The percentage of these cells in controls (n = 8) and MDS patients ranged from 1% to 4%.

Fas expression

CD34- and GPA-positive BM subpopulations were quantified by specific labelling of BMMNC using flow cytometry. Fas expression was studied on these subpopulations by double staining with a FITC-labelled anti-Fas MoAb (CD95, Clone UB2, Immunotech, Marseille, France) and a PE-labelled MoAbs directed to either CD34 or GPA (Immunotech). Results were expressed as relative fluorescence intensity index (RFI), defined as the ratio of the fluorescence intensity of the Fas antigen to an isotypic-matched unrelated control antibody (Beckman Coulter, Miami, Florida). Mean RFI was defined for normal CD34+ or GPA+ BM cells. The signal intensity for Fas on cells from MDS patients was considered to be increased for a RFI above the mean +2 SD of normal controls (n = 8). All cell samples were analysed on a XL cytometer (Beckman Coulter, Miami, Florida).

Statistical analysis

The Student t-test was used to compare quantitative data expressed as means ±standard deviation (SD).


Progenitor colony formation

Erythroid progenitor colony formation was studied in the 25 cases of myelodysplasia described in 1Table I. BFU-E-derived colonies were obtained in 11/25 (44%) of MDS cases (Table II). Mean numbers of BFU-E were significantly diminished in MDS (13 ± 24/105 BMMNC) as compared to normal bone marrows (318 ± 212/105 BMMNC; n = 11; P < 0.005). CFU-E-derived colonies were obtained in 13/24 (54%) of MDS cases and were also markedly decreased in MDS (70 ± 127/105 BMMNC) as compared to normal marrows (612 ± 215/105 BMMNC; P < 0.05) (Table II). This indicates that erythroid colony formation in the presence of Epo was severely compromised.

Table 1. Table I. Patient characteristics. Abbreviations: N, normal; C, complex.Thumbnail image of
Table 2. Table II. Erythroid growth, STAT5 activation upon Epo stimulation and Fas expression on GPA+ cells. * Expressed as number of colonies/105 BMMNC.† Fas expression < 1.8 was referred as (−).Thumbnail image of

STAT5 activation

In an attempt to investigate an impairment of the Epo signal transduction in MDS, STAT5 activation was analysed in Epo-stimulated cells. BMMNC from normal and MDS marrows were stimulated in vitro by Epo (10 IU/ml) after cells had been serum-deprived for 2 h. Nuclear extracts were used for the detection of STAT5 DNA-binding activity. No case of constitutive activation of STAT5 was detectable. In 12/24 tested cases Epo-induced STAT5 DNA binding activity as assessed by the presence of protein–oligonucleotide complexes (2, 1Table II; Fig 1). Stem cell factor (SCF) addition was unable to rescue STAT5 DNA-binding activity in samples in which no STAT5 activation was detected following Epo stimulation (data not shown).

Figure 1.

. STAT5 DNA binding activity in normal and MDS marrows. Electrophoretic mobility shift assay (EMSA) was performed on nuclear extracts from BMMNC stimulated (+) or not (−) with 10 IU/ml Epo for 10 min at 37°C. Epo-stimulated UT-7 and FDCP-1 cell lines were used as positive controls.

In all normal and MDS marrows tested, except in one case of RARS, STAT5 was present as its faster migrating form (Fig 1). This form was found to comigrate with a C-terminal truncated form of STAT5 detectable in the FDCP-1 cell line (Gobert et al, 1996). C-terminal truncated STAT5 was characterized by its ability to supershift when incubated with anti-STAT5 antibodies directed against the N-terminus of the molecule (N) and by its inability to do so in the presence of anti-human STAT5A and B antibodies directed against the C-terminus of the molecule (A + B). By pre-incubating UT-7 nuclear extracts with anti-human STAT5 antibodies, STAT5-oligonucleotide complexes could be partially or completely supershifted with N or A + B antibodies. In most cases no supershift was observed from normal or MDS marrows using A + B antibodies, whereas it could be detected with N antibodies (Fig 2).

Figure 2.

. Supershift assays. Nuclear extracts were incubated with anti-human STAT5 antibodies directed either to the N-terminus (N) or to the C-terminus of STAT5A and STAT5B (A+B) prior to incubation with the labelled β-casein probe. In MDS, STAT5-oligonucleotide complexes were supershifted by N antibodies. UT-7 cell line was used as control for N or (A + B)-induced supershift (full-length forms) and FDCP-1 cell line was used as control for N-induced supershift (C-terminal truncated forms).

Kinetics experiments in the UT-7 cell line have demonstrated that STAT5 DNA binding activity was maximal at 10 min and decreased after 60 min of Epo stimulation (Verdier et al, 1998). In normal and most MDS marrows, the signal increased from 0 to 60 min of Epo stimulation. However, in some cases of MDS the signal was transient, i.e. maximal after 10 min and returned to basal levels after 60 min of Epo stimulation (Fig 3).

Figure 3.

. Kinetics of Epo-induced STAT5 activation in normal and MDS marrows. BMMNC were incubated with 10 IU/ml Epo for 0, 10 or 60 min at 37°C before preparing nuclear extracts. STAT5 activation was revealed by EMSA, as described.

Proteasome inhibitors have been described as effective in sustaining Epo-induced signal transduction (Verdier et al, 1998). As shown in Fig 44A, the proteasome inhibitor LLnL was able to maintain a high level of STAT5 DNA binding activity in UT-7 cells after 60 min of Epo stimulation. In normal marrow, LLnL had no significant effect after 10 min of Epo stimulation, whereas it drastically reinforced the signal after 60 min of Epo stimulation. In an attempt to determine whether the Epo-induced Jak2-STAT5 pathway was functional, we preincubated MDS-deriving mononuclear cells with 50 μm LLnL before stimulation with Epo. Examples of the effects of proteasome inhibitor on MDS-deriving mononuclear cells are shown in Fig 44B. After 10 min of Epo stimulation LLnL was either ineffective in modifying STAT5 DNA binding activity in 4/6 cases (Fig 4B: RA and RAEB) or able to rescue STAT5 activation in two non-responsive cases (Fig 4B: RAEBt). After 60 min of Epo stimulation, STAT5–oligonucleotide complexes were more detectable in the presence of LLnL than in the absence of LLnL in 4/6 cases and remained undetectable in 2/6 cases (Fig 4B). LLnM, used as control, was unable to modify STAT5 activation induced by 10 or 60 min of Epo stimulation (Fig 4B). No signal was seen in LLnL- or LLnM-incubated cells when Epo was omitted. These results suggest that the inhibition of the proteasome-dependent proteolytic pathway could restore a prolonged activation of STAT5 in response to Epo.

Figure 4.

. Effects of proteasome inhibitors on Epo-induced STAT5 activation. Cytokine-deprived UT-7 or serum-deprived bone marrow mononuclear cells were preincubated with 50 μm LLnL or 50 μm LLnM for 15 min at 37°C before 10 IU/ml Epo stimulation for 10 or 60 min at 37°C. The reaction was stopped by the addition of ice-cold PBS pH 7.40 and the samples were processed for nuclear extraction. (A) UT-7 cell line and normal bone marrow (N BM). (B) MDS marrows (one RA, one RAEBt).

We did not find any correlation between the state of Epo-induced STAT5 activation and the BFU-E/CFU-E characteristics of growth. 8/12 (67%) patients without activation of STAT5 and 5/12 (42%) patients with STAT5 activation had no or poor erythropoiesis (i.e. no BFU-E and few CFU-E) (Table II). STAT5 pathway could be functional without effective in vitro erythropoiesis and, conversely, erythroid progenitor growth could also be present without early activation of STAT5 in response to Epo.


The sensitivity of MDS BMMNC to apoptosis was first evaluated by cell-cycle analysis. We found that cytokine deprivation for 1 d induced a 4-fold increase of apoptotic/necrotic cells (sub-G1) in RA and RAEB/t versus a 2-fold increase of sub-G1 cells in normal marrows. After 2 d of deprivation the percentage of sub-G1 cells in RA and RAEB/t remained higher than in normal samples (Fig 5A). At the same time the percentage of S-phase cells was decreased by 50% in RA and RAEB/t versus 10% in normal marrows (Fig 5B). In contrast, RARS-deriving cells retained higher proliferative capabilities than normal marrows. A 3-fold increase in the percentage of sub-G1 cells was observed in RARS, whereas the percentage of S-phase cells remained stable during the 2 d of culture (5Figs 5A and 5B).

Figure 5.

. Cell-cycle analysis by flow cytometry. BMMNC were cultured in cytokine-deprived conditions for 2 d. (A) Percentage of sub-G1 cells. (B) Percentage of S-phase cells. Normal bone marrows (n = 11) closed circles, RA (n = 4) open squares, RARS (n = 9) open circles, RAEB/t (n = 12) closed triangles.

Spontaneous apoptosis was also quantified using the TUNEL technique on MDS and normal bone marrow smears. Sample analysis was performed on 100 cells by two different persons. In normal marrows (n = 8), the mean percentage of TdT+ cells was 15%. Similar experimental conditions showed that 12/23 MDS patients had geqslant R: gt-or-equal, slanted30% TdT+ cells corresponding to an increase in apoptosis (Fig 6A). Most TdT+ cells belonged to the myeloid lineage (myelocytes, promyelocytes and neutrophils) in RA and RAEB/t. Among the nine cases of RARS studied, five exhibited >30% TdT+ cells. However, preparations from the RARS subgroup were re-examined with more attention given to the differentiated erythroid cells. We found a massive apoptosis in 80% and 90% erythroblasts in two cases (nos. 6 and 13), whereas the others had a range of TdT+ erythroblasts between 0 and 20%.

Figure 6.

. Apoptosis of BMMNC and Fas expression on the CD34+ population. (A) Apoptosis was quantified on bone marrow smears using the TUNEL technique as described in Materials and Methods and reported as percentages of TdT+ cells. The normal bone marrow mean value was 15%. (B) Fas expression on the CD34+ subpopulation was studied by flow cytometry and expressed as a ratio of fluorescence intensity (RFI) to an isotypic control.

Fas expression on CD34+ and GPA+ subpopulations

We have previously shown that the expression of the transmembrane receptor Fas was increased in MDS, whatever the FAB subtype, as compared to normal marrow (Bouscary et al, 1997). In the present work, Fas was quantified as a ratio of fluorescence intensity (RFI) to a control antibody in the CD34+ or the GPA+ subpopulations. In normal marrows (n = 8), mean RFI value (±SD) for Fas was 0.7 ± 0.3 and 0.9 ± 0.4 on CD34+ and GPA+ subpopulations, respectively. Fas expression was considered to be increased for RFI values geqslant R: gt-or-equal, slanted mean +2 SD, i.e. for geqslant R: gt-or-equal, slanted1.3 on CD34+ cells and geqslant R: gt-or-equal, slanted1.8 on GPA+ cells.

Elevated Fas expression was found on CD34+ BM cells from 13/25 MDS patients (3/4 RA, 5/9 RARS and 5/12 AREB/t) and was well correlated with the increase in apoptosis detected using the TUNEL technique (Fig 6B). RFI values of Fas expression on GPA+ cells were also increased in 11/25 MDS (4/4 RA, 3/9 RARS and 4/12 RAEB/t). 9/11 (82%) of patients with versus 5/14 (36%) of patients without increased Fas expression on GPA+ cells had no or poor erythropoiesis (Table II). Comparing the BFU-E formation to the expression of Fas on GPA+ cells (Fig 7) indicated that the mean number (±SEM) of BFU-E was significantly higher in patients without (22.2 ± 7.9/105 BMMNC) than in patients with increased Fas expression (2.1 ± 1.6/105 BMMNC) on GPA+ cells (P < 0.05).

Figure 7.

. BFU-E growth according to Fas expression on GPA+ cells. Means of BFU-E numbers (±SEM) were determined in the Fas(−) and Fas(+) subgroups. Student t-test was used to compare mean values: * P < 0.05.


In vitro erythropoiesis in response to Epo is either completely abolished or severely reduced in MDS, except in the RARS subtype. Few data concerning Epo-R expression in MDS marrows are available due to the lack of analytical tools. However, one study of [125I]Epo binding to BMMNC suggested that the percentage of Epo-R expressing cells was equivalent in normal and MDS marrows (Backs et al, 1996). Additionally, no evidence has been provided for deleterious mutations of the EPO receptor gene in acquired or familial myelodysplasias (Backs et al, 1996; Mittelman et al, 1996). Therefore studies were conducted to investigate Epo-R signal transduction pathways. Because detection of transcription factor activation upon cytokine stimulation required few cells, an adapted method for biochemical analysis of the corresponding signal transduction pathway was used. Here, we report on a group of patients with various degrees of dyserythropoiesis in which STAT5 activity upon Epo stimulation was studied. We found that a majority (67%) of patients with a defect of STAT5 activation had poor erythropoiesis, but in vitro BFU-E/CFU-E growth could exist despite the absence of detectable levels of Epo-induced STAT5 activation and, conversely, Epo-induced STAT5 activation was not always sufficient for erythroid development. A recently published work reported on the link between STAT5 pathway defect and altered in vitro erythroid progenitor growth (Hoefsloot et al, 1997). The discrepancy between the two studies may be explained by patient distribution. We have conducted our work as a prospective study whatever the degree of anaemia and the results of erythroid progenitor growth. Nine cases of RARS versus one case, and no case of chronic myelomonocytic leukaemia versus six cases in Hoefsloot's study were included. Most of the RARS exhibited erythroid colony formation and 5/9 RARS had positive Epo-induced STAT5 activation.

In the present work the question of Epo-R expression and activation was also addressed. However, from as much as 30 × 106 normal or MDS-deriving non-separated BMMNC we were unable to detect Epo-R tyrosine phosphorylation after Epo stimulation. In addition, Western blotting experiments with anti-Epo-R antibodies performed either on total cells lysates or after Epo-R immunoprecipitation also remained negative (data not shown). This was probably due to the low level of Epo-R expression and to the dilution of Epo-R expressing cells among the majority of Epo-R-negative cells. This explanation remains accurate in the RARS subgroup which exhibited a majority of terminally differentiated erythroblasts in which Epo-R was not expressed. Therefore we could not exclude that both Epo-R expression and STAT5 activation remained below the threshold of detection in some cases.

Proteasome, a multicatalytic protease complex, is involved in ubiquitin-dependent proteolysis. Recent papers have shown that receptors for platelet-derived growth factor β-receptor (Mori et al, 1995) and Met tyrosine kinase receptor (Jeffers et al, 1997) could be degraded through a proteasome-dependent pathway after ligand binding. Additionally, proteasome inhibitors such as lactacystin or LLnL were shown to prolong the signal of STAT activation in response to IL-2 (Yu & Burakoff, 1997), IL-3 (Callus & Mathey-Prevot, 1998) and Epo (Verdier et al, 1998). LLnL allowed a sustained activation of STAT5 and rescued a signal of STAT5 activation in patients unresponsive to Epo alone. This could be due to the inhibition of receptor–ligand degradation and to the persistence of activated receptors at the cell surface. Therefore we could not exclude that a faint and/or transient signal could be abortive and ineffective in mediating the transcription of target genes.

The implication of the Jak2-STAT5 pathway in propagating proliferative signals of erythroid or multipotent cell lines has been reported (Damen et al, 1995; Chrétien et al, 1996), but the role of STAT5 activation remains still controversial. Depending on the erythroid cell line used, blocking STAT5 activation by antisense strategy or Epo-R mutations led to enhancement or suppression of erythroid differentiation (Iwatsuki et al, 1997). Additionally, retroviral infection of Epo-R−/− mouse fetal liver cells with a mutant of Epo-R lacking intracytoplasmic tyrosine residues (except Y479) demonstrated that STAT5 is dispensable for both the proliferation and differentiation of erythroid progenitors, thereby suggesting the predominant involvement of PI3-kinase and MAPK pathways (Klingmüller et al, 1997). In addition, Epo responses were found to be unaffected in STAT5A and STAT5B double knock-out mice (Teglund et al, 1998).

In early normal haemopoietic cells, the highly homologous STAT5A and STAT5B are almost exclusively present as a C-terminal truncated form named STAT5β (Meyer et al, 1998). Here, we found that both normal and MDS marrow always (except in one case) expressed the faster migrating form of STAT5 which was also detectable in the FDCP-1 cell line used as control (Gobert et al, 1996). Supershift experiments confirmed that this form corresponded to STAT5β. An unidentified nucleus-associated serine protease has been suggested to be responsible for in vivo STAT5 truncation in haemopoietic progenitors (Azam et al, 1997; Meyer et al, 1998). But STAT5 cleavage could also occur in vitro during nuclear extraction and we must consider the possibility that either an endogenous protease from Epo-R expressing cells or an exogenous protease from neighbouring bone marrow cells could cleave STAT5. The functional significance of full-length and truncated forms of STAT5 is still questionable.

Although we demonstrated that Epo-R could be functional in MDS, our data emphasized the role of apoptosis in the defect of erythropoiesis. It is known that Fas expression is induced by IFNγ on normal erythroid cells (Maciejewski et al, 1995) and that Fas ligand is normally present in human purified BFU-E and CFU-E (Maciejewski et al, 1995; Dai et al, 1998). The functionality of Fas in these cells has been suggested by Dai et al (1998) who showed that IFNγ-induced apoptosis of BFU-E was mediated by the Fas–Fas ligand interaction. Of particular interest is the clear relationship that we found between the increased Fas expression on GPA+ cells and the decrease of in vitro erythroid progenitor growth. We and others have previously reported that the Fas antigen expressed in MDS bone marrow is functional in terms of BFU-E and CFU-E growth inhibition and apoptosis induction (Maciejewski et al, 1995; Bouscary et al, 1997; Dai et al, 1998). TUNEL-labelled smears from RARS examined regarding the presence of TdT+ erythroblastic cells exhibited percentages ranging from 0 to 90%. The two cases with the highest percentages (80% and 90%) were both responsive to Epo in terms of STAT5 activation. This may argue for the predominant role of apoptosis in the pathogenesis of anaemia in the early stages of MDS. We have previously found that secondary AML and RAEBt were mostly Fas negative and appeared to escape the control of apoptotic signals. Here both Fas expression and erythroid progenitor growth were low in five cases of RAEBt (nos. 19, 22, 23, 24 and 25). This confirms the minor role of Fas-dependent apoptosis in the dyserythropoiesis of advanced stages of MDS. In addition, there is an increased susceptibility of immature progenitors to the apoptotic process (Rajapaksa et al, 1996; Parker et al, 1997). In our series, Fas was found heavily expressed on CD34+ cells in more than a half of patients. Fas ligand is also overexpressed and functional on erythroid and myeloid MDS cells (Gersuk et al, 1998; Gupta et al, 1999). Therefore Fas-mediated apoptosis could occur in MDS early progenitors, as already suggested (Bouscary et al, 1997; Okada et al, 1997; Lepelley et al, 1998).

Interestingly, increased expression of Fas was combined with a defect of STAT5 activation in six cases of MDS (nos. 1, 4, 8, 16, 18 and 25) with altered erythropoiesis, suggesting that ineffective anti-apoptotic signals may contribute to emphasize pro-apoptotic events. A cross-talk between survival and apoptotic pathways does exist, as erythropoietin is known to induce expression of anti-apoptotic molecules such as Bcl-XL and Bcl-2 to ensure erythroid progenitor survival (Silva et al, 1996; Gregoli & Bondurant, 1997). In addition, Bcl-XL is probably a STAT5 target gene (Dumon et al, 1999). The effect of Epo on apoptosis effectors and anti-apoptotic mediators has to be determined in MDS erythroid progenitors. Finally, the use of rHuEpo in the MDS treatment remains questionable: although both rHuEpo and SCF could stimulate erythroid progenitor proliferation, no evidence has been found that rHuEpo alone could surpass Fas-mediated pro-apoptotic signals.


This work was supported by a grant of the Fondation de France ‘Fondation contre la Leucémie’ (no. 96003069) and of the Association pour le Don de Moelle Osseuse ‘Moelle, Partage et Vie’.