Granulocyte-macrophage colony-stimulating factor reactivates fetal hemoglobin synthesis in erythroblast clones from normal adults

Reactivation of fetal hemoglobin (HbF, alpha 2 gamma 2) synthesis was previously reported in normal human adult erythroblast colonies ("bursts") generated by erythroid progenitors (BFU-E) in fetal calf serum-supplemented (FCS+) semisolid cultures stimulated with erythropoietin (Ep). Our studies focused on the reactivation of HbF synthesis in normal adult erythroid bursts generated by peripheral blood mononuclear cells (PBMCs) seeded in FCS+ methylcellulose culture. Reactivation is almost totally suppressed when (a) PBMCs are grown in optimized FCS- culture, or (b) PBMCs are first stringently depleted of monocytes and then plated in FCS+ medium (ie, BFU-E growth in FCS+ Mo- culture). In both experimental conditions, the proliferation of lymphocytes and macrophages interspersed among colonies is drastically reduced, and the cloning efficiency of granulocyte-macrophage (GM) progenitors is sharply diminished. In either case, addition of biosynthetic GM colony-stimulating factor (GM-CSF) induces a dose-related increase of HbF synthesis up to the level in FCS+ culture, with even more elevated values on delayed addition of Ep. A dose-related increase was also observed in erythroblast clones generated by highly purified BFU-E. These results suggest that reactivation of HbF synthesis in normal adults is at least in part mediated by GM-CSF. Furthermore, they imply intriguing hypotheses on the mechanism(s) of perinatal Hb switching. Finally, they raise the possibility of reactivation of HbF synthesis in beta-thalassemia and sickle cell anemia by GM-CSF therapy.     

Fetal calf serum contains activities that induce fetal hemoglobin in adult erythroid cell cultures.

Cultures of peripheral blood or bone marrow erythroid progenitors display stimulated production of fetal hemoglobin. We investigated whether this stimulation is due to factors contained in the sera of the culture medium. Comparisons of gamma/gamma + beta biosynthetic ratios in erythroid colonies grown in fetal calf serum (FCS) or in charcoal treated FCS (C-FCS) showed that FCS-grown cells had significantly higher gamma/gamma + beta ratios. This increase in globin chain biosynthesis was reflected by an increase in relative amounts of steady-state gamma-globin mRNA. In contrast to its effect on adult cells, FCS failed to influence gamma-chain synthesis in fetal burst forming units-erythroid (BFU-E) colonies. There was a high correlation of gamma-globin expression in paired cultures done with C-FCS or fetal sheep serum. Dose-response experiments showed that the induction of gamma-globin expression is dependent on the concentration of FCS. These results indicate that FCS contains an activity that induces gamma-globin expression in adult erythroid progenitor cell cultures.     

Reactivation of HbF synthesis in normal adult erythroid bursts by IL-3

Reactivation of HbF synthesis has been reported in normal adult erythroblast colonies ('burst') generated by erythroid progenitors (BFU-E) after seeding peripheral blood mononuclear cells (PBMC) in fetal calf serum-supplemented (FCS+) semisolid cultures stimulated by erythropoietin (Ep). Reactivation is almost totally suppressed when: (i) PMBC are grown in optimized FCS- culture or (ii) PBMC are first stringently depleted of monocytes and then plated in FCS+ medium (i.e. BFU-E growth in FCS+Mo- culture). In either case, addition of biosynthetic granulocyte-macrophage colony stimulating factor (GM-CSF) induces a dose-related increase of relative HbF synthesis up to the level in FCS+ culture. We report that, in FCS- culture of partially purified adult blood BFU-E, treatment with biosynthetic interleukin 3 (IL-3) causes a dose-related rise of relative HbF production in the bursts. A similar phenomenon is observed in FCS+ culture of highly purified BFU-E. The rise of HbF synthesis is seemingly mediated, at least in part, by a direct effect of IL-3 at BFU-E level. It is tentatively concluded that reactivation of HbF in vitro, as well as in a variety of in vivo conditions (i.e. stress erythropoiesis, marrow regeneration), may be at least in part mediated by IL-3 and GM-CSF.     

Evidence for direct action of human biosynthetic (recombinant) GM-CSF on erythroid progenitors in serum-free culture

The biologic activity of human biosynthetic granulocyte-monocyte colony stimulating factor (GM-CSF) was investigated in serum-free culture of erythroid progenitors derived from adult peripheral blood. The morphology of erythroid bursts and the cloning efficiency of BFU-E under serum-free conditions were similar to those observed in dishes with fetal bovine serum (FBS). For these experiments, progenitor cells were partially purified by Ficoll-Paque density centrifugation, adherence to a plastic surface, and complement-mediated cytotoxicity of Leu-1+ elements. For some studies, blastlike cells were harvested directly from 6-day-old semisolid cultures. In serum-free culture of the light-density cell fraction, biosynthetic erythropoietin (Ep) was sufficient for formation of pure and mixed erythroid colonies whereas GM-CSF was required for granulocyte-monocytic colonies. When adherent and Leu-1+ cells were removed, or when in vitro differentiated blast cells were used as a source of progenitors, neither Ep or GM-CSF alone induced colony formation. In dishes supplemented with both growth factors, erythroid bursts were detected. Although the presence of GM- CSF alone did not induce formation of any colony or clusters, BFU-E were recorded when Ep was added 8 days later, suggesting that BFU-E could be maintained. Terminal maturation of the resulting erythroid bursts was delayed by 8 days. These results provide evidence that GM- CSF acts directly on early erythroid progenitors. Furthermore, they suggest that both Ep and GM-CSF are necessary to start the differentiation process.     

Globin synthesis in erythroid bursts that mature sequentially in culture. I. Studies in cultures of adult peripheral blood BFU-Es

To study whether the culture time at which the burst populations mature influences the expression of fetal hemoglobin in bursts, we measured hemoglobin synthesis in cohorts of fully hemoglobinized erythroid bursts maturing sequentially in cultures of adult peripheral blood BFU- Es. In 13 of 15 experiments, a decline in gamma/gamma + beta ratio was noted as the culture time advanced. On the average, erythroid bursts that mature during the third culture week showed lower levels of fetal Hb synthesis compared to bursts that are already mature in the second culture week. The decline of gamma/gamma + beta ratio with culture time was also noted in erythroid bursts composed of immature erythroblasts. The enhanced HbF formation in peripheral blood BFU-E cultures is thus most pronounced among the bursts that become hemoglobinized early, and there is a tendency for normalization of HbF synthesis in bursts that mature in late culture days. These results can be interpreted by several alternatives, including the possibility that the expression of high HbF levels in the early days of adult BFU-E cultures is a reflection of premature commitment to terminal differentiation of progenitors that possess an active HbF program. The present data indicate that the variation of HbF synthesis with culture time should be taken into consideration when the influence of various culture conditions of HbF synthesis is studied in BFU-E cultures.     

Role of cyclic nucleotides in fetal hemoglobin induction in cultured CD34+ cells

OBJECTIVE: In vivo, several drugs have been shown to increase fetal hemoglobin (HbF), including 5-azacytidine (AZA), sodium butyrate (SB), and hydroxyurea (HU). Studies in K562 cells suggest that cyclic guanosine monophosphate (cGMP) is required for HbF induction; however, the role of cyclic nucleotides in HbF induction in primary erythroid cultures has not been established. METHODS: CD34-selected peripheral blood monocytes cultured in a semi-solid serum-free system that mimics in vivo F-cell production are utilized to explore the role of cyclic adenosine monophosphate (cAMP) and cGMP in HbF induction in response to HU, AZA, and SB. RESULTS: In serum-free CD34 cultures, HU, SB, and AZA all markedly stimulate FNRBC production up to 30-fold, associated with induction of gamma-globin mRNA and total HbF protein. Guanylate cyclase inhibition results in only minimal blunting of HbF induction by each agent. In contrast, adenylate cyclase inhibition markedly reduces HU, SB, and AZA-mediated FNRBC induction and gamma-globin mRNA induction. The adenylate cyclase activator forskolin modestly induces FNRBC production and augments the action of standard induction agents. HU, AZA, and SB, however, fail to significantly stimulate adenylate cyclase themselves. CONCLUSIONS: In human CD34(+) cultures, cAMP production is required for full induction of HbF by HU, SB, and AZA, while perturbation of cGMP production has only minimal effects. These findings are in marked contrast to data in K562 cells where cGMP production is critical for HbF induction while cAMP stimulation blunts HbF response, and suggest that these agents may share a common induction pathway.     

Erythropoietin alone induces erythroid burst formation by human embryonic but not adult BFU-E in unicellular serum-free culture [see comments]

Erythroid progenitors (BFU-E) from adult human peripheral blood generate erythroid bursts in semisolid culture supplemented with at least two growth factors, ie, erythropoietin (Ep) and interleukin-3 (IL- 3) or granulocyte-macrophage colony-stimulating factor (GM-CSF). We have analyzed the hematopoietin(s) requirement of human embryonic BFU- E, as compared to that of adult peripheral blood progenitors: This was basically evaluated in fetal calf serum-free (FCS-) methylcellulose culture of partially or highly purified progenitors treated with human recombinant hemopoietins. At a low seeding concentration (2 x 10(3) cells/dish) purified embryonic BFU-E generated erythroid bursts when treated only with Ep: Further addition of IL-3 or GM-CSF had no effect on BFU-E cloning efficiency, although the size of bursts was increased in a dose-dependent manner, particularly with IL-3. At a similar seeding concentration (ie, 10(3) cells/dish), purified adult BFU-E efficiently generated erythroid bursts in the presence of Ep and GM-CSF or IL-3, while only few small erythroid colonies were observed in the presence of Ep alone. In a final series of experiments, unicellular FCS- cultures of purified embryonic BFU-E gave rise to erythroid bursts in the presence of Ep alone. Furthermore, the cloning efficiency induced by Ep was unmodified by further addition of GM-CSF or IL-3. Unicellular FCS- cultures of highly purified adult peripheral blood progenitors generated no erythroid bursts in the presence of Ep alone. The addition of GM-CSF or IL-3 was required to generate BFU-E colonies. These studies indicate that in human embryonic life, BFU-E require only Ep for efficient erythroid burst formation, while IL-3 and GM-CSF essentially enhance the proliferation of early erythropoietic precursors.     

c-kit ligand reactivates fetal hemoglobin synthesis in serum-free culture of stringently purified normal adult burst-forming unit- erythroid

We have analyzed the reactivation of fetal hemoglobin (HbF) synthesis under rigorous in vitro conditions, ie, in mature erythroblasts generated by erythroid burst-forming units (BFU-E) stringently purified from normal adult peripheral blood and grown in fetal calf serum(FCS)- free semisolid or liquid phase culture. In clonogenetic dishes, graded amounts of c-kit ligand (KL) were added together with saturating levels of erythropoietin (Ep) and variable amounts of interleukin-3 and granulocyte-macrophage colony stimulating factor (IL-3/GM-CSF), ie, high or low level, or no IL-3/GM-CSF addition. In all conditions, KL induced a sharp, dose-dependent increase in the percentage of F cells and HbF content from nearly normal levels (< 10% and < 2.5%, respectively, at 0.1 and 1 ng/mL) up to 40% to 50% and 10% to 15% at 100 to 200 ng/mL. This increase was not associated with significant differences of burst number or stage of maturation at the time of analysis (as evaluated on the basis of percent mature erythroblasts and Hb content per cell). However, the KL-induced reactivation of HbF synthesis was strictly and directly correlated with a sharp increase of colony size, ie, cell number per burst. Addition of large amounts of IL- 3 and GM-CSF (10 to 100 U and 1 to 10 ng/mL, respectively) significantly potentiated the KL-induced reactivation of HbF, as compared with low levels (0.1 U and 0.01 to 0.1 ng) or no addition of these growth factors: this increase was highly significant at low KL doses (ie, 1 to 10 ng/mL). Single-burst analysis showed that the KL- induced HbF reactivation occurs homogeneously in the erythroid colonies within each of these culture conditions. We have analyzed the effect of KL in liquid phase BFU-E culture treated with the IL-3/GM-CSF/Ep combination at sequential times until terminal erythroid maturation: KL causes a sharp increase in the percentage of F cells and HbF content in all stages of maturation, whereas the IL-3/GM-CSF/Ep combination alone has a markedly lower effect. These results suggest that KL plays a key role in the reactivation of HbF synthesis in adult life, whereas IL- 3/GM-CSF potentiate this effect at low KL levels. The KL-induced HbF reactivation is seemingly related to an enhanced proliferation of erythroid progenitors in the erythropoietic differentiation pathway.     

Fetal hemoglobin production in cultures of primitive and mature human erythroid progenitors: differentiation affects the quantity of fetal hemoglobin produced per fetal-hemoglobin-containing cell

Single-cell microscopic immunodiffusion assays were used to determine the cellular mechanisms that regulate fetal hemoglobin (HbF) levels in cultures of primitive and late erythroid precursors obtained from human adult bone marrow. Two variables--the percentage of cells containing HbF (F cells) and the picograms (pg) of HbF/F cell--were assayed in cells derived from erythroid colony-forming units (CFU-E) and from erythroid burst-forming units (BFU-E) at 7 and 14 days in culture, respectively. The percentage of F cells among all nucleated cells from CFU-E-derived colonies (29.4% +/- 12.5%, mean +/- SD) was not significantly different (p = 0.2) from the percentage of F cells from BFU-E-derived bursts (37.3% +/- 10.1%). Serial daily assays of all cells in cultures on days 3 through 7 and on day 14 revealed a marked increase in F cells between days 4 and 6 in culture. The average amount of HbF/F cell was less in CFU-E-derived F cells than in BFU-E-derived cells (3.5 +/- 0.3 pg versus 6.2 +/- 3.3 pg; p less than 0.01), while adult hemoglobin (HbA) levels in CFU-E-and BFU-E-derived cells remained comparable (19.9 +/- 2.2 pg versus 21.9 +/- 5.3 pg, p = 0.3). These findings indicate that F cell number in culture is not significantly influenced by the relative maturity of the erythroid precursors from which the cells are derived. Differences in the levels of HbF between CFU-E-and BFU-E-derived cells are due to differences in the amount of HbF per F cell, not F cell number.     

HbF reactivation in sibling BFU-E colonies: synergistic interaction of kit ligand with low-dose dexamethasone

Mechanisms underlying fetal hemoglobin (HbF) reactivation in stress erythropoiesis have not been fully elucidated. We suggested that a key role is played by kit ligand (KL). Because glucocorticoids (GCs) mediate stress erythropoiesis, we explored their capacity to potentiate the stimulatory effect of KL on HbF reactivation, as evaluated in unilineage erythropoietic culture of purified adult progenitors (erythroid burst-forming units [BFU-Es]). The GC derivative dexamethasone (Dex) was tested in minibulk cultures at graded dosages within the therapeutical range (10(-6) to 10(-9) M). Dex did not exert significant effects alone, but synergistically it potentiated the action of KL in a dose-dependent fashion. Specifically, Dex induced delayed erythroid maturation coupled with a 2-log increased number of generated erythroblasts and enhanced HbF synthesis up to 85% F cells and 55% gamma-globin content at terminal maturation (ie, in more than 80%-90% mature erythroblasts). Equivalent results were obtained in unicellular erythroid cultures of sibling BFU-Es treated with KL alone or combined with graded amounts of Dex. These results indicate that the stimulatory effect of KL + Dex is related to the modulation of gamma-globin expression rather than to recruitment of BFU-Es with elevated HbF synthetic potential. At the molecular level, Id2 expression is totally suppressed in control erythroid culture but is sustained in KL + Dex culture. Hypothetically, Id2 may mediate the expansion of early erythroid cells, which correlates with HbF reactivation. These studies indicate that GCs play an important role in HbF reactivation. Because Dex acts at dosages used in immunologic disease therapy, KL + Dex administration may be considered to develop preclinical models for beta-hemoglobinopathy treatment.     

Effective erythropoiesis and HbF reactivation induced by kit ligand in beta-thalassemia

In human beta-thalassemia, the imbalance between alpha- and non-alpha-globin chains causes ineffective erythropoiesis, hemolysis, and anemia: this condition is effectively treated by an enhanced level of fetal hemoglobin (HbF). In spite of extensive studies on pharmacologic induction of HbF synthesis, clinical trials based on HbF reactivation in beta-thalassemia produced inconsistent results. Here, we investigated the in vitro response of beta-thalassemic erythroid progenitors to kit ligand (KL) in terms of HbF reactivation, stimulation of effective erythropoiesis, and inhibition of apoptosis. In unilineage erythroid cultures of 20 patients with intermedia or major beta-thalassemia, addition of KL, alone or combined with dexamethasone (Dex), remarkably stimulated cell proliferation (3-4 logs more than control cultures), while decreasing the percentage of apoptotic and dyserythropoietic cells (<5%). More important, in both thalassemic groups, addition of KL or KL plus Dex induced a marked increase of gamma-globin synthesis, thus reaching HbF levels 3-fold higher than in con-trol cultures (eg, from 27% to 75% or 81%, respectively, in beta-thalassemia major). These studies indicate that in beta-thalassemia, KL, alone or combined with Dex, induces an expansion of effective erythropoiesis and the reactivation of gamma-globin genes up to fetal levels and may hence be considered as a potential therapeutic agent for this disease.     

Fetal hemoglobin analysis of erythroid bursts in patients with chronic myelogenous leukemia (CML)

Early erythroid progenitors (BFUE) form colonies of mature progeny in culture. The development of hemoglobinized red cells within multilineage colonies (CFUGEMM) and erythroid bursts is dependent upon exogenously added erythropoietin and molecules released by hemopoietic subpopulations. Mixed colonies and erythroid bursts were grown from 3 patients with Ph' chronic myelogenous leukemia (CML). It was found that some mixed hemopoietic colonies and erythroid bursts did not require exogenously added erythropoietin. An increase of the plating efficiency of BFUE could be observed when erythropoietin was added. Erythroid bursts grown without added Ep from samples of the patients with chronic myelogenous leukemia have a higher probability to contain HbF than clones grown in the presence of Ep. The data support the view of a phenotypical heterogeneity among clonal descendents of a common ancestor as previously postulated for CML.     

Fetal hemoglobin modulation during human erythropoiesis: stem cell factor has "late" effects related to the expression pattern of CD117

A cytokine-screening assay of cultured peripheral blood cells obtained using immune rosetting and separation of progenitors was developed to identify determinants of fetal hemoglobin (HbF) modulation during adult erythropoiesis. Among the 12 erythroid growth-promoting cytokines tested, stem cell factor (SCF) at a concentration of 50 ng/mL resulted in the most significant increase in cell proliferation and HbF content. The average HbF/hemoglobin A (HbA) ratio was 30.9% +/- 18.7% in cultures containing SCF compared with 4.1% +/- 2.2% in those grown with erythropoietin (EPO) alone (P = 8.5E-8). To further investigate the hemoglobin-modulating effects of SCF, we examined the surface expression pattern of the SCF receptor, CD117, among maturing erythroblasts. CD117 expression increased during the first week of culture and peaked on culture days 7 to 9. After culture day 9, the level of CD117 declined to lower levels. The rise in CD117 expression to high levels mirrored that of the transferrin receptor (CD71), and the subsequent reduction in CD117 was inversely related to increases in expression of glycophorin A. SCF-related increases in the HbF/HbA ratio correlated with the expression pattern of CD117. SCF added during days 7 to 14 resulted in a more pancellular distribution of HbF on day 14 compared with the heterocellular distribution present in cultures supplemented with SCF on days 0 to 7. A significant SCF-mediated increase in HbF was also measured using progenitors derived from cord blood. These results suggest that the HbF response to SCF is greatest at the late progenitor stage as a function of surface CD117 expression.     

Influence of recombinant hematopoietins and of fetal bovine serum on the globin synthetic pattern of human BFUe.

We have studied the effects of recombinant hematopoietic growth factors, granulocyte-macrophage colony-stimulating factor (GM-CSF) and/or interleukin-3 (IL-3) on the globin program of adult human erythroid progenitors (BFUe) stimulated to terminal differentiation by erythropoietin under fetal bovine serum (FBS)-supplemented or FBS-deprived culture conditions. Fetal globin production by BFUe-derived erythroblasts was assessed at the protein and mRNA level and its cellular distribution was evaluated by immunofluorescence. Although hemoglobinization and maturation of BFUe-derived erythroblasts was by and large comparable in FBS-replete versus FBS-deprived cultures, the latter had significantly less (up to 20-fold) gamma-globin and gamma-globin mRNA levels. Reduced gamma-globin in serum-deprived cultures was also reflected by a smaller proportion of erythroblasts with detectable gamma-globin by immunofluorescence. Erythroid bursts induced by either GM-CSF or IL-3 produced similar levels of gamma-globin both in FBS-supplemented and in FBS-deprived cultures. These results, obtained even in cultures of highly enriched BFUe, suggest that GM-CSF and IL-3, although they significantly increase the number and size of erythroid bursts, do not by themselves exert a direct influence on the level of fetal globin synthesis. By contrast, factor(s) present in FBS appear to exert a dominant influence on fetal globin synthesis in vitro. Although FBS-deprived conditions appear to largely abrogate the in vitro activation of fetal hemoglobin (Hb F) in normal samples, they do support increased Hb F production in samples from patients with hereditary persistence of fetal hemoglobin or from cord blood.     

Individual BFU-E in polycythemia vera produce both erythropoietin dependent and independent progeny

Erythropoietic progenitors from peripheral blood of normal individuals or patients with polycythemia vera (PV) were cultured in methylcellulose medium containing 2.5 U/ml of erythropoietin (Ep). After 7-9 days, colonies considered to be early stage large bursts were individually removed, resuspended in a small volume of fresh methylcellulose medium, and then divided between 2 dishes. To one of these secondary cultures, sufficient Ep was added to bring the concentration of Ep up to approximately 3 U/ml. To the other was added an equal volume of medium but no Ep. The final concentration of Ep in these cultures was determined to be less than 0.01 U/ml. Nine days later, both types of secondary cultures were scored for the presence of colonies containing 8 or more hemoglobinized erythroblasts. Of 90 primary colonies from 3 normal individuals assessed in this way, 59 gave secondary erythroid colonies in the high Ep cultures, while none gave secondary erythroid colonies in the low Ep cultures. Additional control experiments in which primary colonies from normal individuals were divided into duplicate high Ep cultures showed that on average, the procedure used divided primary colonies equally. Of 109 primary colonies from 5 PV patients that yielded secondary erythroid colonies in the high Ep cultures, 21 yielded no secondary erythroid colonies in the low Ep cultures. The other 88 yielded erythroid colonies in both, but the secondary colonies in the low Ep cultures were consistently smaller in size and significantly fewer in number. Similar results were obtained when primary colonies were generated in cultures to which no Ep was added. These findings indicate that primitive BFU-E in patients with PV can be subdivided into 2 populations: a minor population restricted to the production of erythroid colony-forming cells (Ep- dependent progenitors) that require Ep for their detection, and a major population that is not restricted in this way. In addition, these experiments show that most of the primitive BFU-E that generate Ep- independent progenitors also produce significant numbers of cells that are Ep-dependent.     

alpha-Amino-N-butyric acid stimulates fetal hemoglobin in the adult

The effect of alpha-amino-N-butyric acid (alpha ABA) on fetal hemoglobin production in the adult was examined in vivo after being administered to normal and anemic baboons and in erythroid progenitor cell cultures. Infusion of alpha ABA for five days resulted in four- to fivefold increases in the level of F reticulocytes of normal or chronically anemic baboons. The induction of HbF by alpha ABA was strikingly enhanced by the administration of 5-azacytidine. The addition of alpha ABA in culture produced a concentration-related increase of HbF in baboon CFUe and e-cluster colonies. In addition to the induction of HbF, alpha ABA stimulated the growth of all classes of erythroid progenitors in vivo or in culture. The activation of gamma-globin gene expression by alpha ABA is attributed to an interaction between regulatory sites of globin chromatin modified by alpha ABA and the immature intracellular environment of the expanding erythropoiesis. The combination of chromatin modification, DNA methylation, and the immature intracellular environment of rapid erythroid regeneration may explain the synergistic induction of HbF by alpha ABA and 5-azacytidine.     

Erythroid burst-promoting activity of purified recombinant human GM-CSF and interleukin-3: studies with anti-GM-CSF and anti-IL-3 sera and studies in serum-free cultures

We studied the erythroid burst-promoting activity (BPA) of recombinant human granulocyte/macrophage colony-stimulating factor (GM-CSF) and Interleukin-3 (IL-3) with two experimental approaches. First we studied the effects of polyclonal antisera prepared against human GM-CSF and gibbon IL-3 on colony formation from 1,000 bone marrow null cells/dish in serum-containing culture. Both GM-CSF and IL-3 independently enhanced erythroid burst formation; however, IL-3 showed more BPA activity than GM-CSF. These data are in agreement with an emerging view that the primary targets of IL-3 are primitive progenitors and that the targets of GM-CSF are intermediate progenitors, including erythroid burst-forming units (BFU-E). The proliferation of one population of BFU- E was independent of GM-CSF or IL-3. To characterize this population of BFU-E further, we developed a serum-free culture assay for the purified progenitors by incorporating insulin-like growth factor-1 (IGF-1) to the serum-free medium. The development of erythroid bursts was supported by IL-3, IGF-1, and erythropoietin (Ep) in a serum-free culture system and to a lesser extent by the combination of GM-CSF, IGF- 1, and Ep. Although the burst-promoting ability of GM-CSF and IL-3 was again demonstrated in this system, unlike serum-containing culture Ep alone did not support burst formation. These results indicate that when fetal calf serum (FCS) is present, the culture system contains BPA that is not GM-CSF or IL-3.