Antiapoptotic Role of Growth Factors in the Myelodysplastic Syndromes: Concordance Between In vitro and In vivo Observations

The severe anemia and transfusion dependency of patients with low-risk MDS can be alleviated by treatment with erythropoietin and granulocyte colony-stimulating factor (G-CSF). G-CSF alone has no proven effect on the anemia of MDS patients; however, a synergistic effect of erythropoietin and G-CSF is seen in the MDS subcategories, refractory anemia (RA), RA with ringed sideroblasts (RARS), and RA with excess of blasts, and is most pronounced in RARS patients (1–4). Overall, the response rate to growth factor treatment is ∼40% to 50%, with some patients achieving normalization of hemoglobin levels for many years (2, 5, 6). Very little is known about the effects of proerythroid growth factor treatment at the cellular level. Increased apoptosis of bone marrow precursors is thought to contribute to the ineffective hematopoiesis in these individuals. We have previously shown that bone marrow biopsies from MDS patients responding to erythropoietin + G-CSF show a reduced number of apoptotic precursors compared with the pretreatment biopsies (1). However, whether erythropoietin ± G-CSF act by promoting growth of normal bone marrow cells or through repression of excessive apoptosis of abnormal cells remains to be determined.

Mitochondrial signaling plays a pivotal role in the apoptosis of erythroid precursors in low-risk MDS. Hence, we have documented constitutive translocation of cytochrome c from the mitochondrial intermembrane space into the cytosol, with subsequent activation of downstream caspases, in early and intermediate erythroblasts obtained from RA and RARS patients (7). Moreover, the addition of G-CSF to erythroblast cultures from RARS and RA patients significantly decreased cytochrome c release and subsequent apoptosis of erythroid precursor cells (5). The latter results may explain, in part, the beneficial effects of growth factor administration in MDS patients. However, further studies are needed to understand the effects of growth factors on in vivo erythropoiesis and to determine whether G-CSF, alone or in combination with erythropoietin, promotes growth of normal cells as opposed to cells with cytogenetic abnormalities.

About half of all cases with MDS harbor clonal chromosomal aberrations, the most common of which are deletion of 5q, monosomy 7, and trisomy 8 (8, 9). These cytogenetic abnormalities have frequently been used to define the malignant clone in MDS, but recent data suggest that fluorescence in situ hybridization–negative cells may also contribute to the MDS clone, at least in trisomy 8 (10–12). The deletion of 5q is usually found in 95% to 99% of bone marrow stem cells in MDS patients with this aberration (13, 14), and it is difficult to harvest karyotypically normal stem cells after high-dose chemotherapy thus suggesting that the abnormal cells are relatively insensitive to chemotherapy (11). However, it has been reported that patients responding to growth factor treatment with erythropoietin and G-CSF also show a reduced percentage of karyotypically abnormal cells in the bone marrow (11). Similarly, Rigolin et al. (15) showed that MDS patients with cytogenetically abnormal bone marrow cells who respond to erythropoietin treatment also show a reduced percentage of abnormal cells in the bone marrow. Therefore, it is conceivable that growth factor administration may promote the proliferation of erythroid precursors with normal karyotype in the bone marrow of MDS patients.

In the present study, we aimed to elucidate the cellular effects of erythropoietin and G-CSF on myelodysplastic bone marrow precursors and to assess the relation between in vitro and in vivo observations. The increased apoptosis in vitro before clinical treatment of patients was normalized upon incubation of cells with growth factors. After successful treatment, apoptosis variables approached normal levels and no longer responded to G-CSF. Our analyses thus reveal a high degree of concordance between in vitro and in vivo effects of growth factor administration. In addition, our data show that erythropoietin dramatically promotes the growth of normal progenitors during erythroid differentiation of progenitors from patients with 5q– syndrome, an effect that was enhanced by G-CSF, whereas no such effect was observed in monosomy 7 or trisomy 8 cultures. Taken together, the current data support the conjecture that growth factors can exert dual effects in MDS, with inhibition of apoptosis of myelodysplastic erythroid precursors, and selection of cytogenetically normal progenitor cells.

Patients. The diagnostic procedure was done according to the criteria put forth by the Nordic MDS Group (6). Informed consent was obtained from patients and controls, and the study followed the guidelines of the Ethical Committee for Research at Karolinska Institutet. Patients were classified according to the FAB system (Table 1). Fifteen RA patients were included, seven of which had the 5q– aberration, two with monosomy 7, one with trisomy 8, one with monosomy 7 plus trisomy 8, one with trisomy 15, and three with normal karyotype. All of the patients with the 5q– aberration had the breakpoint between q13-33 and six fulfilled the criteria for 5q– syndrome according to the WHO classification (14). We also studied 11 RARS patients, including one patient with trisomy 8 and 10 with normal karyotype. The cohort of healthy adult controls was the same as described in a previous study (7). Response categories were complete response (CR; stable hemoglobin >115 g/L), partial response (PR; an increase in hemoglobin with >15 g/L or stable hemoglobin level with no further requirement for transfusions), and no response (not fulfilling the criteria for PR; ref. 3). Twelve patients were transfusion-dependent at the time of sampling and 14 had stable anemia. Three RA and six RARS patients were resampled during successful clinical treatment with erythropoietin ± G-CSF. The repeat samples were taken within 2 months from the achievement of a response. Seven of these patients had CR and two PR. One patient with 5q– syndrome was sampled during remission to erythropoietin treatment, another in relapse after previous successful erythropoietin treatment, and in CR after treatment with erythropoietin + G-CSF. One patient with RARS was sampled at three occasions: before treatment, after erythropoietin treatment (in PR), and after erythropoietin + G-CSF treatment (in CR). Fifteen patients were investigated for cytochrome c release, seven of which were followed after successful clinical treatment with erythropoietin ± G-CSF. Among these patients, eight were investigated for caspase-3 activity before and after clinical treatment (see below).

Bone marrow aspiration and mononuclear cell cultures. Bone marrow needle aspiration and mononuclear cell isolation and cultivation were done as previously described (16, 17). Mononuclear cells were cultured in medium alone or in the presence of agonistic anti-Fas antibodies (1 μg/mL; clone CH-11; Medical & Biological Laboratories Ltd., Nagoya, Japan) and/or G-CSF (100 ng/mL; Neupogen, Amgen, Twelve Oaks, CA) for the indicated time points.

Cell separation and flow cytometry. CD34⁺ cells were separated from mononuclear cells using the MiniMacs system (Miltenyi Biotec, Bergisch Gladbach, Germany). The purity of CD34⁺ cells (>95%) was assessed by flow cytometry using anti-CD34 monoclonal antibodies (BD Biosciences, San Jose, CA). Phenotype and erythroid maturation were analyzed at days 4, 7, 11, and 14 using anti-CD36 (ImmunoTech, Marseille, France) and anti-GpA (DAKO, Copenhagen, Denmark) antibodies to define immature, early erythroid, and late erythroid cells, respectively, whereas CD45 was used as a marker of nonerythroid cells. Antibodies against CD114 (BD Biosciences) were applied to detect G-CSF receptor (G-CSFR) expression during erythroid maturation at the indicated time points. All analyses were done on a FACSCalibur (Becton Dickinson, Mountain View, CA) equipped with a 488-nm argon laser operating with the CellQuest software.

Erythroblast cultures. The erythroblast culture method has been described elsewhere (7). Briefly, following positive selection for CD34, the cells (0.1 × 10⁶/mL) were cultivated for 14 days in Iscove's medium (Sigma, Saint Louis, MO) with BIT supplements plus recombinant human interleukin-3 (10 ng/mL; PeproTech House, London, United Kingdom), recombinant human interleukin-6 (10 ng/mL; PeproTech House), recombinant human-stem cell factor (25 ng/mL; Medical & Biological Laboratories, Ltd.). The medium was replenished every second day as above to maintain the cultures at the same cell concentration. Erythropoietin (2 IU/mL; Roche, Basel, Switzerland) was added to the medium in the second week, beginning at day 7 and was replenished at days 9 and 11. At day 14, the expression of GpA⁺ cells was checked by flow cytometry [84.4% (78.9-90.6%)].

Cytochrome c release. Translocation of cytochrome c from mitochondria to cytosol was determined at the indicated time points using a previously established method for the concomitant localization of mitochondria and cytochrome c in single cells (7). Briefly, cells were stained with MitoTracker Red CMXRos (Molecular Probes, Eugene, OR), cytocentrifuged onto glass slides, and fixed and permeabilized in paraformaldehyde and Triton X-100, respectively. Cells were then stained with an anti-cytochrome c antibody (BD Biosciences), followed by a FITC-conjugated goat anti-mouse antibody (Sigma). Images were analyzed on a Leica DM RXA digital confocal microscope and further processed using the Slide Book 4.0.0.6 analysis software. A minimum of 200 cells were evaluated for each position.

Caspase activation. Caspase-3-like activity in hematopoietic progenitor cells was determined before and after successful clinical treatment of patients, as previously described (16). Briefly, cell lysates and substrate [Asp-Glu-Val-Asp (DEVD)-7-amino-4-methyl-coumarin (AMC), 50 μmol/L; Peptide Institute, Osaka, Japan] were combined in a standard reaction buffer and added to a 96-well plate. Enzyme-catalyzed release of AMC was measured every 70 seconds during a 30-minute period in a Fluoroscan II plate reader (Labsystems, Stockholm, Sweden) using 355-nm excitation and 460-nm emission wavelengths. Fluorescence units were converted to pmol of AMC using a standard curve generated with free AMC. Data are depicted as pmol AMC release per minute per 10⁶ cells.

Fluorescence in situ hybridization. To evaluate the relative growth of erythroid progenitors with cytogenetic aberrations, cells were cytocentrifuged after CD34 separation, as well as on days 4, 7, 11, and 14. Slides were pretreated with pepsin and fixed with formaldehyde/MgCl₂. To detect deletions of 5q, an EGR1/D5S721, D5S23 Dual Color Probe (Vysis, Inc., Downers Grove, IL) was used; EGR1 detects deletions of 5q31 and D5S721; D5S23 (5p15.2) serves as an internal control. For monosomy 7 and trisomy 8, α-satellite probes (Cytocell Ltd., Banbury, United Kingdom) for each centromere were used. Probes were applied as recommended by the manufacturer.

Statistical analysis. Statview 5.0 analysis software was used for analysis of quantitative data. Results are presented as mean values ± SE, or as median values (range), when appropriate. Paired Student's t tests were used for comparison of related samples. Ps < 0.05 were considered statistically significant.

Spontaneous cytochrome c release during erythroid differentiation of myelodysplastic progenitors. We have previously reported that myelodysplastic erythroid cells show a constitutive release of cytochrome c from mitochondria during the first week of differentiation (7). To further explore this event during the whole culture period, we analyzed cytochrome c release at days 2, 4, 7, 9, 11, and 14 in RA (n = 7) and RARS (n = 8) patients. As seen in Fig. 1A, cytochrome c release occurred in MDS progenitors at all stages of maturation. Normal erythroid progenitors never showed >5% of positive cells at any time point. The pattern of spontaneous cytochrome c release differed between RA and RARS subgroups, particularly during the second week of culture, insofar as cytochrome c release was pronounced in RARS cultures during the entire culture period, whereas RA cultures displayed a more fluctuating pattern (Fig. 1B-C).

Effects of in vitro administration of growth factors during erythroid maturation. Following a brief (4 hours) in vitro incubation with G-CSF, a significant decrease of spontaneous cytochrome c release was seen in MDS erythroblasts not only at days 4 and 7, as previously described, but also at days 11 and 14 (Fig. 1A-C). The inhibitory effect was thus observed during the whole period of erythroid maturation, albeit with a maximal effect (P <.001) on day 7. The relative decrease of cytochrome c release upon in vitro addition of G-CSF was >40% during the first week of culture (46%, day 4, P < 0.0001; 47%, day 7, P < 0.0001) and 29% to 36% during the second week of culture.

We also explored the in vitro effect of erythropoietin on cytochrome c release in a small subset of patients (RA, n = 1 and RARS, n = 3) before initiation of clinical treatment. Erythropoietin did not influence cytochrome c release during the first week of culture. During the second week, although erythropoietin was already present in the medium, incubation with erythropoietin in a higher concentration reduced cytochrome c release with 32%, 25%, and 20% at days 9, 11, and 14, respectively (data not shown).

Granulocyte colony-stimulating factor receptor expression on erythroid progenitors. The G-CSFR has been reported to be present on myeloid-committed progenitors, neutrophils, monocytes, and some lymphocyte subsets and may also be expressed on erythroid progenitors, at least at the transcriptional level (18–20). To explain the unanticipated effect of G-CSF on late erythroid progenitors, we assessed the protein expression of G-CSFR on erythroid progenitors during culture and also on freshly isolated GpA⁺ bone marrow cells. G-CSFR, as determined by flow cytometric detection of CD114, was seen in 100% of CD36⁺ erythroblasts from MDS patients at days 4 and 7 compared with 86% and 79% of CD36⁺, respectively, in controls (data not shown). For the GpA⁺ population, expression of CD114 was seen in 94% of freshly isolated cells in one control. At day 14, expression of G-CSFR was detected in 95% (75-98%) of cells in MDS samples (two RA and three RARS) and in 82% (54-87%) of cells in three normal bone marrow samples (for representative data, see Fig. 2).

Myelodysplastic syndrome progenitor responses approach the pattern of normal progenitors after successful clinical growth factor treatment. G-CSF and erythropoietin are known to synergistically improve hemoglobin values and reduce bone marrow apoptosis in low-risk MDS patients (5). To assess whether the in vitro effects of G-CSF in erythroid progenitors from untreated MDS patients are reflective of the biological effects of clinical growth factor treatment, we studied spontaneous cytochrome c release in nine patients, RA (n = 3), and RARS (n = 6) following successful growth factor administration leading to CR (n = 7) and PR (n = 2). As seen in Fig. 3A, cells obtained from patients after administration of erythropoietin ± G-CSF displayed significantly less cytochrome c release at all time points. Indeed, the pattern of cytochrome c release following in vitro incubation with G-CSF in untreated patients (see previous section) matched the spontaneous release of cytochrome c in progenitor cells obtained during successful clinical treatment (Fig. 3A). There were no clear differences in cytochrome c release between patients in PR versus patients in CR (data not shown).

We have previously shown that progenitor cells from low-risk MDS patients display an increase in spontaneous caspase-3 activity and show hypersensitivity to Fas ligation and that G-CSF inhibits these processes (17). Here, we assessed caspase-3-like activity in mononuclear cell obtained from three RA patients and five RARS patients before and after successful treatment. Spontaneous caspase-3-like activity was significantly lower after clinical treatment (AMC release 2.86 ± 0.5 pmol before treatment versus 1.17 ± 0.3 pmol AMC release after treatment; P = 0.03, Fig. 3B). The decrease in spontaneous caspase activity significantly correlated with diminished release of cytochrome c in patients sampled during successful treatment (correlation, 0.93; P = 0.004). Similarly, cells from treated patients were less sensitive to agonistic anti-Fas antibodies (before treatment, 10.9 ± 1.6 pmol AMC release; after treatment, 3.7 ± 0.9 pmol AMC release; P = 0.009; Fig. 3C).

Effects of growth factors on myelodysplastic progenitors obtained after clinical treatment. Seven of the transfusion-dependent patients (Table 1) were analyzed only at one occasion. They were either nonresponders to growth factors or had lost a previous response. The cytochrome c and caspase activity pattern of these patients did not differ from that of patients analyzed before a successful treatment (data not shown). Hence, an in vitro effect of G-CSF could be obtained also in patients refractory to clinical treatment, which suggests that the loss of clinical response to growth factors probably is multifactorial. Following successful treatment, we found that additional in vitro incubation with G-CSF had minimal effects on cytochrome c release in the erythroblast cultures (nonsignificant decrease at days 2-11; P = 0.02 at day 14). Moreover, G-CSF did not serve to further suppress Fas-induced caspase-3-like activity in cells derived from treated patients (data not shown). Two patients were analyzed both after erythropoietin treatment (one PR, one with relapse but still on erythropoietin treatment) and after erythropoietin + G-CSF treatment. We could show a further decrease in spontaneous cytochrome c release after the combined treatment (Table 2). Similarly, caspase-3 activity assessed during erythropoietin treatment was further decreased during treatment with erythropoietin in combination with G-CSF (Table 2).

Erythropoietin promotes growth of cytogenetically normal progenitor cells from myelodysplastic syndrome patients. Clinical observations have suggested that the percentages of abnormal metaphases in the bone marrow may decrease in MDS patients who respond to proerythroid growth factor administration (11, 15). We therefore assessed the growth of normal versus abnormal progenitors in the erythroblast culture system, in which erythropoietin was added from day 7. Sustained incubation with G-CSF shifted the cells towards myeloid differentiation; therefore, the effects of this growth factor could not be evaluated. The percentage of erythroblasts carrying the cytogenetic aberration (5q–, monosomy 7, or trisomy 8) was determined in freshly isolated CD34⁺ cells, and at days 4, 7, 11, and 14 of erythroblast culture. Two patients with monosomy 7, two with trisomy 8, and one patient with monosomy 7 plus trisomy 8 were analyzed. The relative size of the clone was unchanged during culture in four of these and, in fact, increased in one patient sampled during PR to clinical treatment (data not shown). There was no specific effect of adding erythropoietin at day 7. We then analyzed six 5q– syndrome patients; in one of these individuals, the analysis was done twice (after failing erythropoietin treatment and in CR following treatment with erythropoietin + G-CSF). In contrast to the monosomy 7 and trisomy 8 cultures, erythropoietin dramatically reduced the percentage of the aberrant clone in the 5q– cultures. Hence, whereas the percentage of 5q– cells (median, 97%; range, 73-99%) did not change during the first week, the addition of erythropoietin at day 7 resulted in a reduction of the 5q– clone to a median of 35% (2-77%) of all cells at day 14 (Fig. 4A). To be sure that this was not due to a general effect of the culture conditions, we also assessed cultures without the addition of erythropoietin from day 7. These cultures gave rise to a lower percentage of GpA⁺ cells at day 14 (9-15%) and thus a less pronounced erythroid differentiation. However, in cultures without erythropoietin from day 7, the percentage of 5q– positive cells was unchanged and above 83% at day 14 (data not shown). We then analyzed the proliferation patterns of 5q– positive versus cytogenetically normal cells and found a striking difference between the two cohorts. During the first week of cultivation, there was no difference in proliferation index between 5q– cells and cytogenetically normal cells (data not shown). During the second week, however, the cytogenetically normal cells showed a median expansion of 88.5-fold (range, 26- to 265-fold) and a median doubling time of 29 hours (range, 21-36 hours). By contrast, the 5q– positive cells grew poorly (median expansion, 2.5-fold; range, 0.38-7.3; Fig. 4B-C). In two cultures, the total number of 5q– positive cells was reduced during the second week. The additional effect of G-CSF to erythropoietin was shown by patient 6 (Table 2). After unsuccessful erythropoietin treatment, cells from this patient displayed 77% abnormalities at day 14 compared with 35% after successful treatment with G-CSF + erythropoietin (Fig. 4A; Table 2).

Erythropoietin and G-CSF are used for the treatment of anemia in low-risk MDS patients. The mechanism underlying the proerythroid effect of these growth factors is not fully understood. Our previous studies have shown that G-CSF inhibits mitochondrial cytochrome c release and subsequent apoptosis during the first week of in vitro differentiation of erythroblasts (7). In the present study, we show that G-CSF can inhibit cytochrome c release at all stages of erythroid differentiation. We also show, for the first time, that both normal and myelodysplastic erythroblasts retain expression of the G-CSFR until very mature stages. Thus, the fact that normal erythroblasts show no reaction to G-CSF suggests that the antiapoptotic effect of G-CSF in MDS progenitors is related to specific downstream signaling events in these cells. The number of cells that could be obtained from erythroblast cultures of MDS patients was insufficient for detailed mechanistic and biochemical studies of apoptosis signaling pathways. Notwithstanding these limitations, it is reasonable to speculate that G-CSF (and perhaps also erythropoietin) exerts its antiapoptotic effects, at least in part, through an inhibition of the mitochondrial release of apoptogenic factors, including cytochrome c (7). G-CSF has been shown to block constitutive apoptosis of peripheral blood neutrophils obtained from healthy controls, (21, 22) and recent studies suggest that G-CSF may exert its antiapoptotic effects through modulation of proapoptotic Bax, a protein that promotes the release of cytochrome c from the intermembrane space of mitochondria (23, 24). Moreover, our recent studies of patients with severe congenital neutropenia have indicated that G-CSF may act via effects on antiapoptotic members of the Bcl-2 family, including Bcl-2 itself (25). Similarly, previous studies have provided evidence that erythropoietin, a growth factor known to retard apoptosis of erythroid progenitor cells (26) may target members of the Bcl-2 family, affecting, e.g., the phosphorylation of Bcl-2 (27) or the expression of Bcl-2 and/or Bcl-X_L (28, 29). Taken together, these findings may serve to explain how, in erythroid progenitor cells derived from MDS patients, G-CSF and erythropoietin may prevent cytochrome c release and promote cell survival.

The current data show that erythropoietin alone has no significant effect of cytochrome c release during the first week of erythroid differentiation but exerts a potent inhibitory (i.e., antiapoptotic) effect during the second week of culture. We surmise that in cases with pronounced early erythroid apoptosis, erythropoietin alone may not be sufficient to promote survival and continuous differentiation of progenitor cells; in such cases, the potent antiapoptotic effect of G-CSF may be required to allow erythroid progenitors to survive until they become erythropoietin responsive. Importantly, these findings are reflective of the synergistic effect of erythropoietin and G-CSF in the treatment of MDS patients. The present study also provides evidence that constitutive apoptosis persists through all stages of erythroid differentiation in MDS patients. Thus, it is unlikely that the decreased erythroid maturation and subsequent anemia seen in low-risk MDS is caused by a specific block in differentiation; instead, anemia may result from an accumulation of events leading to a reduced potential for survival of progenitor cells. Of note, we observed that constitutive cytochrome c release is more pronounced in RARS cultures compared with cultures of RA progenitors. From a clinical point of view, this is reflected by the common finding of a higher percentage of erythroblasts in the bone marrow of RARS patients compared with RA patients.

The cohort of patients with 5q– syndrome described herein allowed us to investigate the relative growth of erythroblasts with normal versus abnormal karyotype. We observed that cytogenetically normal erythroblasts display a remarkable growth potential in comparison with the 5q– cells and a similar growth rate during the first week of differentiation. This may explain the profound anemia and erythroid hypoplasia of patients with 5q– syndrome but not why 5q– positive bone marrow progenitors have a growth advantage over normal cells allowing them to expand in vivo and cause disease. It is possible that the growth advantage is only present at a very early stem cell level, and could not be revealed by our culture model. It remains to be clarified whether the relative outgrowth of cytogenetically normal cells in our system is caused only by a stimulation of these cells or whether erythropoietin may have an inhibitory effect on 5q– cells. Interestingly, Sloand et al. (30) previously showed that trisomy 8 cells were more prone to apoptosis than normal cells. It is thus conceivable that the addition of antiapoptotic growth factors preferentially enhanced survival of these abnormal cells. Further studies are warranted to determine which signaling pathways that dictate the susceptibility of abnormal cells to apoptosis and responsiveness to proerythroid growth factors.

We also observed that the percentage of cytogenetically normal cells at day 14 was greater in 5q– patients responding to growth factors than in those with no response. Moreover, in one patient analyzed on two occasions, more normal cells were seen after successful treatment with erythropoietin + G-CSF than in the preceding erythropoietin-resistant phase. Previous clinical studies of erythropoietin treatment of MDS patients have failed to show a clear relationship between a terminated erythroid response and general disease progression or leukemic evolution (6, 31). On the contrary, relapse of anemia is often seen in these patients without obvious morphologic changes in the bone marrow. One possible explanation for the development of resistance to growth factors could therefore be that the remaining normal cells loose their propensity to outgrow the abnormal cells.

To conclude, the current studies show that constitutive cytochrome c release, an indicator of mitochondria-dependent apoptosis, occurs in all stages of erythroid differentiation in MDS patients. This event is blocked by G-CSF and, in more mature erythroid progenitors, also by erythropoietin. Moreover, in vitro addition of growth factors did not further promote survival of progenitors obtained from successfully treated patients. These findings further enhance our understanding of the mechanism of action of proerythroid growth factors in MDS. We have also shown that erythropoietin promotes the survival of cytogenetically normal erythroid progenitor cells derived from MDS patients. The development of novel drugs for the treatment of anemia in MDS patients requires in vitro models for the evaluation of the various effects on erythropoiesis. The erythroblast model used herein may prove useful in the testing of other proerythroid drugs.

We thank Monika Jansson for assistance in fluorescence in situ hybridization analyses.